summaryrefslogtreecommitdiffstats
path: root/cli/src/cli-cmd-parser.c
blob: aec5384bd01821bec295fed73319d05ff637d9e4 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
4849
4850
4851
4852
4853
4854
4855
4856
4857
4858
4859
4860
4861
4862
4863
4864
4865
4866
4867
4868
4869
4870
4871
4872
4873
4874
4875
4876
4877
4878
4879
4880
4881
4882
4883
4884
4885
4886
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
4901
4902
4903
4904
4905
4906
4907
4908
4909
4910
4911
4912
4913
4914
4915
4916
4917
4918
4919
4920
4921
4922
4923
4924
4925
4926
4927
4928
4929
4930
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
4944
4945
4946
4947
4948
4949
4950
4951
4952
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
4967
4968
4969
4970
4971
4972
4973
4974
4975
4976
4977
4978
4979
4980
4981
4982
4983
4984
4985
4986
4987
4988
4989
4990
4991
4992
4993
4994
4995
4996
4997
4998
4999
5000
5001
5002
5003
5004
5005
5006
5007
5008
5009
5010
5011
5012
5013
5014
5015
5016
5017
5018
5019
5020
5021
5022
5023
5024
5025
5026
5027
5028
5029
5030
5031
5032
5033
5034
5035
5036
5037
5038
5039
5040
5041
5042
5043
5044
5045
5046
5047
5048
5049
5050
5051
5052
5053
5054
5055
5056
5057
5058
5059
5060
5061
5062
5063
5064
5065
5066
5067
5068
5069
5070
5071
5072
5073
5074
5075
5076
5077
5078
5079
5080
5081
5082
5083
5084
5085
5086
5087
5088
5089
5090
5091
5092
5093
5094
5095
5096
5097
5098
5099
5100
5101
5102
5103
5104
5105
5106
5107
5108
5109
5110
5111
5112
5113
5114
5115
5116
5117
5118
5119
5120
5121
5122
5123
5124
5125
5126
5127
5128
5129
5130
5131
5132
5133
5134
5135
5136
5137
5138
5139
5140
5141
5142
5143
5144
5145
5146
5147
5148
5149
5150
5151
5152
5153
5154
5155
5156
5157
5158
5159
5160
5161
5162
5163
5164
5165
5166
5167
5168
5169
5170
5171
5172
5173
5174
5175
5176
5177
5178
5179
5180
5181
5182
5183
5184
5185
5186
5187
5188
5189
5190
5191
5192
5193
5194
5195
5196
5197
5198
5199
5200
5201
5202
5203
5204
5205
5206
5207
5208
5209
5210
5211
5212
5213
5214
5215
5216
5217
5218
5219
5220
/*
   Copyright (c) 2010-2013 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.

   This file is licensed to you under your choice of the GNU Lesser
   General Public License, version 3 or any later version (LGPLv3 or
   later), or the GNU General Public License, version 2 (GPLv2), in all
   cases as published by the Free Software Foundation.
*/
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include <stdint.h>
#include <pthread.h>

#ifndef _CONFIG_H
#define _CONFIG_H
#include "config.h"
#endif

#include "cli.h"
#include "cli-cmd.h"
#include "cli-mem-types.h"
#include "dict.h"

#include "protocol-common.h"
#include "cli1-xdr.h"

#define MAX_SNAP_DESCRIPTION_LEN 1024

struct snap_config_opt_vals_ snap_confopt_vals[] = {
        {.op_name        = "snap-max-hard-limit",
         .question       = "Changing snapshot-max-hard-limit "
                           "will limit the creation of new snapshots "
                           "if they exceed the new limit.\n"
                           "Do you want to continue?"
        },
        {.op_name        = "snap-max-soft-limit",
         .question       = "If Auto-delete is enabled, snap-max-soft-limit will"
                           " trigger deletion of oldest snapshot, on the "
                           "creation of new snapshot, when the "
                           "snap-max-soft-limit is reached.\n"
                           "Do you want to change the snap-max-soft-limit?"
        },
        {.op_name        = "both",
        .question        = "Changing snapshot-max-hard-limit "
                           "will limit the creation of new snapshots "
                           "if they exceed the new snapshot-max-hard-limit.\n"
                           "If Auto-delete is enabled, snap-max-soft-limit will"
                           " trigger deletion of oldest snapshot, on the "
                           "creation of new snapshot, when the "
                           "snap-max-soft-limit is reached.\n"
                           "Do you want to continue?"
        },
        {.op_name        = NULL,
        }
};

enum cli_snap_config_set_types {
        GF_SNAP_CONFIG_SET_HARD = 0,
        GF_SNAP_CONFIG_SET_SOFT = 1,
        GF_SNAP_CONFIG_SET_BOTH = 2,
};
typedef enum cli_snap_config_set_types cli_snap_config_set_types;

static const char *
id_sel (void *wcon)
{
        return (const char *)wcon;
}

static char *
str_getunamb (const char *tok, char **opwords)
{
        return (char *)cli_getunamb (tok, (void **)opwords, id_sel);
}

int32_t
cli_cmd_bricks_parse (const char **words, int wordcount, int brick_index,
                      char **bricks, int *brick_count)
{
        int                    ret = 0;
        char             *tmp_list = NULL;
        char    brick_list[120000] = {0,};
        char                *space = " ";
        char            *delimiter = NULL;
        char            *host_name = NULL;
        char        *free_list_ptr = NULL;
        char               *tmpptr = NULL;
        int                      j = 0;
        int         brick_list_len = 0;
        char             *tmp_host = NULL;

        GF_ASSERT (words);
        GF_ASSERT (wordcount);
        GF_ASSERT (bricks);
        GF_ASSERT (brick_index > 0);
        GF_ASSERT (brick_index < wordcount);

        strncpy (brick_list, space, strlen (space));
        brick_list_len++;
        while (brick_index < wordcount) {
                if (validate_brick_name ((char *)words[brick_index])) {
                        cli_err ("Wrong brick type: %s, use <HOSTNAME>:"
                                 "<export-dir-abs-path>", words[brick_index]);
                        ret = -1;
                        goto out;
                } else {
                        delimiter = strrchr (words[brick_index], ':');
                        ret = gf_canonicalize_path (delimiter + 1);
                        if (ret)
                                goto out;
                }

                if ((brick_list_len + strlen (words[brick_index]) + 1) > sizeof (brick_list)) {
                        cli_err ("Total brick list is larger than a request. "
                                 "Can take (brick_count %d)", *brick_count);
                        ret = -1;
                        goto out;
                }

                tmp_host = gf_strdup ((char *)words[brick_index]);
                if (!tmp_host) {
                        gf_log ("cli", GF_LOG_ERROR, "Out of memory");
                        ret = -1;
                        goto out;
                }
                get_host_name (tmp_host, &host_name);
                if (!host_name) {
                        ret = -1;
                        gf_log("cli",GF_LOG_ERROR, "Unable to allocate "
                               "memory");
                        goto out;
                }

                if (!(strcmp (host_name, "localhost") &&
                      strcmp (host_name, "127.0.0.1") &&
                      strncmp (host_name, "0.", 2))) {
                        cli_err ("Please provide a valid hostname/ip other "
                                 "than localhost, 127.0.0.1 or loopback "
                                 "address (0.0.0.0 to 0.255.255.255).");
                        ret = -1;
                        GF_FREE (tmp_host);
                        goto out;
                }
                if (!valid_internet_address (host_name, _gf_false)) {
                        cli_err ("internet address '%s' does not conform to "
                                 "standards", host_name);
                }
                GF_FREE (tmp_host);
                tmp_list = gf_strdup (brick_list + 1);
                if (free_list_ptr) {
                        GF_FREE (free_list_ptr);
                        free_list_ptr = NULL;
                }
                free_list_ptr = tmp_list;
                j = 0;
                while(j < *brick_count) {
                        strtok_r (tmp_list, " ", &tmpptr);
                        if (!(strcmp (tmp_list, words[brick_index]))) {
                                ret = -1;
                                cli_err ("Found duplicate"
                                         " exports %s",words[brick_index]);
                                goto out;
                       }
                       tmp_list = tmpptr;
                       j++;
                }
                strcat (brick_list, words[brick_index]);
                strcat (brick_list, " ");
                brick_list_len += (strlen (words[brick_index]) + 1);
                ++(*brick_count);
                ++brick_index;
        }

        *bricks = gf_strdup (brick_list);
        if (!*bricks)
                ret = -1;
out:
        GF_FREE (free_list_ptr);
        return ret;
}

int32_t
cli_cmd_create_disperse_check (struct cli_state *state, int *disperse,
                               int *redundancy, int *data, int count)
{
        int i = 0;
        int tmp = 0;
        gf_answer_t answer = GF_ANSWER_NO;
        char question[128];

        const char *question1 = "There isn't an optimal redundancy value "
                                "for this configuration. Do you want to "
                                "create the volume with redundancy 1 ?";

        const char *question2 = "The optimal redundancy for this "
                                "configuration is %d. Do you want to create "
                                "the volume with this value ?";

        const char *question3 = "This configuration is not optimal on most "
                                "workloads. Do you want to use it ?";

        const char *question4 = "Redundancy for this configuration is %d. "
                                "Do you want to create "
                                "the volume with this value ?";

        if (*data > 0) {
                if (*disperse > 0 && *redundancy > 0) {
                        if (*disperse != (*data + *redundancy)) {
                                cli_err ("Disperse count(%d) should be equal "
                                         "to sum of disperse-data count(%d) and "
                                         "redundancy count(%d)", *disperse,
                                         *data, *redundancy);
                                return -1;
                        }
                } else if (*redundancy > 0) {
                        *disperse = *data + *redundancy;
                } else if (*disperse > 0) {
                        *redundancy = *disperse - *data;
                } else {
                        if ((count - *data) >= *data) {
                                cli_err ("Please provide redundancy count "
                                         "along with disperse-data count");
                                return -1;
                        } else {
                                sprintf (question, question4, count - *data);
                                answer = cli_cmd_get_confirmation (state,
                                                                   question);
                                if (answer == GF_ANSWER_NO)
                                        return -1;
                                *redundancy = count - *data;
                                *disperse = count;
                        }
                }
        }

        if (*disperse <= 0) {
                if (count < 3) {
                        cli_err ("number of bricks must be greater "
                                 "than 2");

                        return -1;
                }
                *disperse = count;
        }

        if (*redundancy == -1) {
                tmp = *disperse - 1;
                for (i = tmp / 2;
                     (i > 0) && ((tmp & -tmp) != tmp);
                     i--, tmp--);

                if (i == 0) {
                        answer = cli_cmd_get_confirmation(state, question1);
                        if (answer == GF_ANSWER_NO)
                                return -1;

                        *redundancy = 1;
                }
                else
                {
                        *redundancy = *disperse - tmp;
                        if (*redundancy > 1) {
                                sprintf(question, question2, *redundancy);
                                answer = cli_cmd_get_confirmation(state,
                                                                  question);
                                if (answer == GF_ANSWER_NO)
                                        return -1;
                        }
                }

                tmp = 0;
        } else {
                tmp = *disperse - *redundancy;
        }

        if (*redundancy > (*disperse - 1) / 2) {
                cli_err ("redundancy must be less than %d for a "
                         "disperse %d volume",
                         (*disperse + 1) / 2, *disperse);

                return -1;
        }

        if ((tmp & -tmp) != tmp) {
                answer = cli_cmd_get_confirmation(state, question3);
                if (answer == GF_ANSWER_NO)
                        return -1;
        }

        return 0;
}

static int32_t
cli_validate_disperse_volume (char *word, gf1_cluster_type type,
                              const char **words, int32_t wordcount,
                              int32_t index, int32_t *disperse_count,
                              int32_t *redundancy_count,
                              int32_t *data_count)
{
        int     ret = -1;

        switch (type) {
        case GF_CLUSTER_TYPE_NONE:
        case GF_CLUSTER_TYPE_DISPERSE:
                if (strcmp (word, "disperse") == 0) {
                        if (*disperse_count >= 0) {
                                cli_err ("disperse option given twice");
                                goto out;
                        }
                        if (wordcount < (index+2)) {
                                goto out;
                        }
                        ret = gf_string2int (words[index + 1], disperse_count);
                        if (ret == -1 && errno == EINVAL) {
                                *disperse_count = 0;
                                ret = 1;
                        } else if (ret == -1) {
                                goto out;
                        } else {
                                if (*disperse_count < 3) {
                                        cli_err ("disperse count must "
                                                 "be greater than 2");
                                        goto out;
                                }
                                ret = 2;
                        }
                } else if (strcmp (word, "disperse-data") == 0) {
                        if (*data_count >= 0) {
                                cli_err ("disperse-data option given twice");
                                goto out;
                        }
                        if (wordcount < (index+2)) {
                                goto out;
                        }
                        ret = gf_string2int (words[index+1], data_count);
                        if (ret == -1 || *data_count < 2) {
                                cli_err ("disperse-data must be greater than 1");
                                goto out;
                        }
                        ret = 2;
                } else if (strcmp (word, "redundancy") == 0) {
                        if (*redundancy_count >= 0) {
                                cli_err ("redundancy option given twice");
                                goto out;
                        }
                        if (wordcount < (index+2)) {
                                goto out;
                        }
                        ret = gf_string2int (words[index+1], redundancy_count);
                        if (ret == -1 || *redundancy_count < 1) {
                                cli_err ("redundancy must be greater than 0");
                                goto out;
                        }
                        ret = 2;
                }
                break;
        case GF_CLUSTER_TYPE_STRIPE_REPLICATE:
                cli_err ("striped-replicated-dispersed volume "
                         "is not supported");
                goto out;
        case GF_CLUSTER_TYPE_TIER:
                cli_err ("tier-dispersed volume is not "
                         "supported");
                goto out;
        case GF_CLUSTER_TYPE_STRIPE:
                cli_err ("striped-dispersed volume is not "
                         "supported");
                goto out;
        case GF_CLUSTER_TYPE_REPLICATE:
                cli_err ("replicated-dispersed volume is not "
                         "supported");
                goto out;
        default:
                cli_err ("Invalid type given");
                break;
        }
out:
        return ret;
}

int32_t
cli_validate_volname (const char *volname)
{
        int32_t            ret                       = -1;
        int32_t            i                         = -1;
        static const char * const invalid_volnames[] = {
                                      "volume", "type", "subvolumes", "option",
                                      "end-volume", "all", "volume_not_in_ring",
                                      "description", "force",
                                      "snap-max-hard-limit",
                                      "snap-max-soft-limit", "auto-delete",
                                      "activate-on-create", NULL};

        if (volname[0] == '-')
                goto out;

        for (i = 0; invalid_volnames[i]; i++) {
                if (!strcmp (volname, invalid_volnames[i])) {
                        cli_err ("\"%s\" cannot be the name of a volume.",
                                 volname);
                        goto out;
                }
        }

        if (strchr (volname, '/'))
                goto out;

        if (strlen (volname) > GD_VOLUME_NAME_MAX) {
                cli_err("Volume name exceeds %d characters.",
                        GD_VOLUME_NAME_MAX);
                goto out;
        }

        for (i = 0; i < strlen (volname); i++) {
                if (!isalnum (volname[i]) && (volname[i] != '_') &&
                   (volname[i] != '-')) {
                        cli_err ("Volume name should not contain \"%c\""
                                 " character.\nVolume names can only"
                                 "contain alphanumeric, '-' and '_' "
                                 "characters.", volname[i]);
                        goto out;
                }
        }

        ret = 0;
out:
        return ret;
}

int32_t
cli_cmd_volume_create_parse (struct cli_state *state, const char **words,
                             int wordcount, dict_t **options)
{
        dict_t  *dict = NULL;
        char    *volname = NULL;
        int     ret = -1;
        gf1_cluster_type type = GF_CLUSTER_TYPE_NONE;
        int     count = 1;
        int     sub_count = 1;
        int     brick_index = 0;
        char    *trans_type = NULL;
        int32_t index = 0;
        char    *bricks = NULL;
        int32_t brick_count = 0;
        char    *opwords[] = { "replica", "stripe", "transport", "disperse",
                               "redundancy", "disperse-data", "arbiter", NULL };

        char    *w = NULL;
        char    *ptr = NULL;
        int      op_count = 0;
        int32_t  replica_count = 1;
        int32_t  arbiter_count = 0;
        int32_t  stripe_count = 1;
        int32_t  disperse_count = -1;
        int32_t  redundancy_count = -1;
        int32_t  disperse_data_count = -1;
        gf_boolean_t is_force = _gf_false;
        int wc = wordcount;

        GF_ASSERT (words);
        GF_ASSERT (options);

        dict = dict_new ();

        if (!dict)
                goto out;

        if (wordcount < 3)
                goto out;

        volname = (char *)words[2];

        GF_ASSERT (volname);

        /* Validate the volume name here itself */
        if (cli_validate_volname (volname) < 0)
                goto out;

        if (wordcount < 4) {
                ret = -1;
                goto out;
        }

        type = GF_CLUSTER_TYPE_NONE;
        index = 3;

        while (op_count < 3) {
                ret = -1;
                w = str_getunamb (words[index], opwords);
                if (!w) {
                        break;
                } else if ((strcmp (w, "replica")) == 0) {
                        switch (type) {
                        case GF_CLUSTER_TYPE_STRIPE_REPLICATE:
                        case GF_CLUSTER_TYPE_REPLICATE:
                                cli_err ("replica option given twice");
                                goto out;
                        case GF_CLUSTER_TYPE_NONE:
                                type = GF_CLUSTER_TYPE_REPLICATE;
                                break;
                        case GF_CLUSTER_TYPE_STRIPE:
                                type = GF_CLUSTER_TYPE_STRIPE_REPLICATE;
                                break;
                        case GF_CLUSTER_TYPE_TIER:
                                cli_err ("replicated-tiered volume is not "
                                         "supported");
                                goto out;
                                break;
                        case GF_CLUSTER_TYPE_DISPERSE:
                                cli_err ("replicated-dispersed volume is not "
                                         "supported");
                                goto out;
                        default:
                                cli_err ("Invalid type given");
                                goto out;
                        }

                        if (wordcount < (index+2)) {
                                ret = -1;
                                goto out;
                        }
                        replica_count = strtol (words[index+1], NULL, 0);
                        if (replica_count < 2) {
                                cli_err ("replica count should be greater"
                                         " than 1");
                                ret = -1;
                                goto out;
                        }
                        ret = dict_set_int32 (dict, "replica-count", replica_count);
                        if (ret)
                                goto out;

                        index += 2;
                        if (words[index]) {
                                if (!strcmp (words[index], "arbiter")) {
                                        ret = gf_string2int (words[index+1],
                                                             &arbiter_count);
                                        if (ret == -1 || arbiter_count != 1 ||
                                            replica_count != 3) {
                                                cli_err ("For arbiter "
                                                         "configuration, "
                                                         "replica count must be"
                                                         " 3 and arbiter count "
                                                         "must be 1. The 3rd "
                                                         "brick of the replica "
                                                         "will be the arbiter");
                                                ret = -1;
                                                goto out;
                                        }
                                        ret = dict_set_int32 (dict, "arbiter-count",
                                                              arbiter_count);
                                        if (ret)
                                                goto out;
                                        index += 2;
                                }
                        }

                } else if ((strcmp (w, "stripe")) == 0) {
                        switch (type) {
                        case GF_CLUSTER_TYPE_STRIPE_REPLICATE:
                        case GF_CLUSTER_TYPE_STRIPE:
                                cli_err ("stripe option given twice");
                                goto out;
                        case GF_CLUSTER_TYPE_NONE:
                                type = GF_CLUSTER_TYPE_STRIPE;
                                break;
                        case GF_CLUSTER_TYPE_REPLICATE:
                                type = GF_CLUSTER_TYPE_STRIPE_REPLICATE;
                                break;
                        case GF_CLUSTER_TYPE_DISPERSE:
                                cli_err ("striped-dispersed volume is not "
                                         "supported");
                                goto out;
                        case GF_CLUSTER_TYPE_TIER:
                                cli_err ("striped-tier volume is not "
                                         "supported");
                                goto out;
                        default:
                                cli_err ("Invalid type given");
                                goto out;
                        }
                        if (wordcount < (index + 2)) {
                                ret = -1;
                                goto out;
                        }
                        stripe_count = strtol (words[index+1], NULL, 0);
                        if (stripe_count < 2) {
                                cli_err ("stripe count should be greater"
                                         " than 1");
                                ret = -1;
                                goto out;
                        }
                        ret = dict_set_int32 (dict, "stripe-count", stripe_count);
                        if (ret)
                                goto out;

                        index += 2;

                } else if ((strcmp (w, "transport")) == 0) {
                        if (trans_type) {
                                cli_err ("'transport' option given more"
                                         " than one time");
                                goto out;
                        }
                        if ((strcasecmp (words[index+1], "tcp") == 0)) {
                                trans_type = gf_strdup ("tcp");
                        } else if ((strcasecmp (words[index+1], "rdma") == 0)) {
                                trans_type = gf_strdup ("rdma");
                        } else if ((strcasecmp (words[index+1], "tcp,rdma") == 0) ||
                                   (strcasecmp (words[index+1], "rdma,tcp") == 0)) {
                                trans_type = gf_strdup ("tcp,rdma");
                        } else {
                                gf_log ("", GF_LOG_ERROR, "incorrect transport"
                                        " protocol specified");
                                ret = -1;
                                goto out;
                        }
                        index += 2;

                } else if ((strcmp (w, "disperse") == 0) ||
                           (strcmp (w, "redundancy") == 0) ||
                           (strcmp (w, "disperse-data") == 0)) {
                        ret = cli_validate_disperse_volume (w, type, words,
                                      wordcount, index, &disperse_count,
                                      &redundancy_count, &disperse_data_count);
                        if (ret < 0)
                                goto out;
                        index += ret;
                        type = GF_CLUSTER_TYPE_DISPERSE;
                } else {
                        GF_ASSERT (!"opword mismatch");
                        ret = -1;
                        goto out;
                }
                op_count++;
        }

        if (!trans_type)
                trans_type = gf_strdup ("tcp");

        /* reset the count value now */
        count = 1;

        if (index >= wordcount) {
                ret = -1;
                goto out;
        }

        brick_index = index;

        if (strcmp (words[wordcount - 1], "force") == 0) {
                is_force = _gf_true;
                wc = wordcount - 1;
        }

        ret = cli_cmd_bricks_parse (words, wc, brick_index, &bricks,
                                    &brick_count);
        if (ret)
                goto out;

        /* If brick-count is not valid when replica or stripe is
           given, exit here */
        if (!brick_count) {
                cli_err ("No bricks specified");
                ret = -1;
                goto out;
        }

        if (type == GF_CLUSTER_TYPE_DISPERSE) {
                ret = cli_cmd_create_disperse_check (state, &disperse_count,
                                                     &redundancy_count,
                                                     &disperse_data_count,
                                                     brick_count);
                if (!ret)
                        ret = dict_set_int32 (dict, "disperse-count",
                                              disperse_count);
                if (!ret)
                        ret = dict_set_int32 (dict, "redundancy-count",
                                              redundancy_count);
                if (ret)
                        goto out;

                sub_count = disperse_count;
        } else
                sub_count = stripe_count * replica_count;

        if (brick_count % sub_count) {
                if (type == GF_CLUSTER_TYPE_STRIPE)
                        cli_err ("number of bricks is not a multiple of "
                                 "stripe count");
                else if (type == GF_CLUSTER_TYPE_REPLICATE)
                        cli_err ("number of bricks is not a multiple of "
                                 "replica count");
                else if (type == GF_CLUSTER_TYPE_DISPERSE)
                        cli_err ("number of bricks is not a multiple of "
                                 "disperse count");
                else
                        cli_err ("number of bricks given doesn't match "
                                 "required count");

                ret = -1;
                goto out;
        }

        /* Everything is parsed fine. start setting info in dict */
        ret = dict_set_str (dict, "volname", volname);
        if (ret)
                goto out;

        ret = dict_set_int32 (dict, "type", type);
        if (ret)
                goto out;

        ret = dict_set_dynstr (dict, "transport", trans_type);
        if (ret)
                goto out;
        trans_type = NULL;

        ret = dict_set_dynstr (dict, "bricks", bricks);
        if (ret)
                goto out;

        ret = dict_set_int32 (dict, "count", brick_count);
        if (ret)
                goto out;

        ret = dict_set_int32 (dict, "force", is_force);
        if (ret)
                goto out;

        *options = dict;

out:
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Unable to parse create volume CLI");
                if (dict)
                        dict_destroy (dict);
        }

        GF_FREE (trans_type);

        return ret;
}

int32_t
cli_cmd_volume_reset_parse (const char **words, int wordcount, dict_t **options)
{
        dict_t  *dict = NULL;
        char    *volname = NULL;
        int     ret = -1;

        GF_ASSERT (words);
        GF_ASSERT (options);

        dict = dict_new ();

        if (!dict)
                goto out;

        if (wordcount < 3)
                goto out;

        if (wordcount > 5)
                goto out;

        volname = (char *)words[2];

        if (!volname) {
                ret = -1;
                goto out;
        }

        ret = dict_set_str (dict, "volname", volname);
        if (ret)
                goto out;

        if (wordcount == 3) {
                ret = dict_set_str (dict, "key", "all");
                if (ret)
                        goto out;
        }

        if (wordcount >= 4) {
                if (!strcmp ("force", (char*)words[3])) {
                        ret = dict_set_int32 (dict, "force", 1);
                        if (ret)
                                goto out;
                        ret = dict_set_str (dict, "key", "all");
                        if (ret)
                                goto out;
                } else {
                        ret = dict_set_str (dict, "key", (char *)words[3]);
                        if (ret)
                                goto out;
                }
        }

        if (wordcount == 5) {
                if (strcmp ("force", (char*)words[4])) {
                        ret = -1;
                        goto out;
                } else {
                        ret = dict_set_int32 (dict, "force", 1);
                        if (ret)
                                goto out;
                }
        }

        *options = dict;

out:
        if (ret && dict) {
                dict_destroy (dict);
        }

                return ret;
}

/* Parsing global option for NFS-Ganesha config
 *  gluster nfs-ganesha enable/disable */

int32_t
cli_cmd_ganesha_parse (struct cli_state *state,
                       const char **words, int wordcount,
                       dict_t **options, char **op_errstr)
{
        dict_t  *dict     =       NULL;
        int     ret       =       -1;
        int     flags     =       0;
        char    *key      =       NULL;
        char    *value    =       NULL;
        int     i         =       0;
        char    *w        =       NULL;
        char   *opwords[] =      { "enable", "disable", NULL };
        const char      *question       =       NULL;
        gf_answer_t     answer          =       GF_ANSWER_NO;


        GF_ASSERT (words);
        GF_ASSERT (options);

        dict = dict_new ();

        if (!dict)
                goto out;

        if (wordcount != 2)
                goto out;

        key   = (char *) words[0];
        value = (char *) words[1];

        if (!key || !value) {
                cli_out ("Usage : nfs-ganesha <enable/disable>");
                ret = -1;
                goto out;
        }

        ret = gf_strip_whitespace (value, strlen (value));
        if (ret == -1)
                goto out;

        if (strcmp (key, "nfs-ganesha")) {
                gf_asprintf (op_errstr, "Global option: error: ' %s '"
                          "is not a valid global option.", key);
                ret = -1;
                goto out;
        }

        w = str_getunamb (value, opwords);
        if (!w) {
                cli_out ("Invalid global option \n"
                         "Usage : nfs-ganesha <enable/disable>");
                ret = -1;
                goto out;
        }

        question = "Enabling NFS-Ganesha requires Gluster-NFS to be"
                   " disabled across the trusted pool. Do you "
                   "still want to continue?\n";

        if (strcmp (value, "enable") == 0) {
                answer = cli_cmd_get_confirmation (state, question);
                if (GF_ANSWER_NO == answer) {
                        gf_log ("cli", GF_LOG_ERROR, "Global operation "
                                "cancelled, exiting");
                        ret = -1;
                        goto out;
                }
        }
        cli_out ("This will take a few minutes to complete. Please wait ..");

        ret = dict_set_str (dict, "key", key);
        if (ret) {
               gf_log (THIS->name, GF_LOG_ERROR, "dict set on key failed");
                goto out;
        }

        ret = dict_set_str (dict, "value", value);
        if (ret) {
               gf_log (THIS->name, GF_LOG_ERROR, "dict set on value failed");
                goto out;
        }

        ret = dict_set_str (dict, "globalname", "All");
        if (ret) {
                gf_log (THIS->name, GF_LOG_ERROR, "dict set on global"
                        " key failed.");
                goto out;
        }

        ret = dict_set_int32 (dict, "hold_global_locks", _gf_true);
        if (ret) {
                gf_log (THIS->name, GF_LOG_ERROR, "dict set on global key "
                        "failed.");
                goto out;
        }

        *options = dict;
out:
        if (ret)
                dict_unref (dict);

        return ret;
}

int32_t
cli_cmd_inode_quota_parse (const char **words, int wordcount, dict_t **options)
{
        dict_t          *dict    = NULL;
        char            *volname = NULL;
        int              ret     = -1;

        GF_ASSERT (words);
        GF_ASSERT (options);

        dict = dict_new ();
        if (!dict) {
                gf_log ("cli", GF_LOG_ERROR, "dict_new failed");
                goto out;
        }

        if (wordcount != 4)
                goto out;

        volname = (char *)words[2];
        if (!volname) {
                ret = -1;
                goto out;
        }

        /* Validate the volume name here itself */
        if (cli_validate_volname (volname) < 0)
                goto out;

        ret = dict_set_str (dict, "volname", volname);
        if (ret < 0)
                goto out;

        if (strcmp (words[3], "enable") != 0) {
                cli_out ("Invalid quota option : %s", words[3]);
                ret = -1;
                goto out;
        }

        ret = dict_set_int32 (dict, "type",
                              GF_QUOTA_OPTION_TYPE_ENABLE_OBJECTS);
        if (ret < 0)
                goto out;

        *options = dict;
out:
        if (ret < 0) {
                if (dict)
                        dict_destroy (dict);
        }

        return ret;
}

int32_t
cli_cmd_quota_parse (const char **words, int wordcount, dict_t **options)
{
        dict_t          *dict    = NULL;
        char            *volname = NULL;
        int              ret     = -1;
        int              i       = -1;
        char             key[20] = {0, };
        int64_t          value   = 0;
        gf_quota_type    type    = GF_QUOTA_OPTION_TYPE_NONE;
        char           *opwords[] = { "enable", "disable", "limit-usage",
                                      "remove", "list", "alert-time",
                                      "soft-timeout", "hard-timeout",
                                      "default-soft-limit", "limit-objects",
                                      "list-objects", "remove-objects", NULL};
        char            *w       = NULL;
        uint32_t         time    = 0;
        double           percent = 0;
        char            *end_ptr = NULL;
        int64_t          limit   = 0;

        GF_ASSERT (words);
        GF_ASSERT (options);

        dict = dict_new ();
        if (!dict) {
                gf_log ("cli", GF_LOG_ERROR, "dict_new failed");
                goto out;
        }

        if (wordcount < 4)
                goto out;

        volname = (char *)words[2];
        if (!volname) {
                ret = -1;
                goto out;
        }

        /* Validate the volume name here itself */
        if (cli_validate_volname (volname) < 0)
                goto out;

        ret = dict_set_str (dict, "volname", volname);
        if (ret < 0)
                goto out;

        w = str_getunamb (words[3], opwords);
        if (!w) {
                cli_out ("Invalid quota option : %s", words[3]);
                ret = - 1;
                goto out;
        }

        if (strcmp (w, "enable") == 0) {
                if (wordcount == 4) {
                        type = GF_QUOTA_OPTION_TYPE_ENABLE;
                        ret = 0;
                        goto set_type;
                } else {
                        ret = -1;
                        goto out;
                }
        }

        if (strcmp (w, "disable") == 0) {
                if (wordcount == 4) {
                        type = GF_QUOTA_OPTION_TYPE_DISABLE;
                        ret = 0;
                        goto set_type;
                } else {
                        ret = -1;
                        goto out;
                }
        }

        if (strcmp (w, "limit-usage") == 0) {
                type = GF_QUOTA_OPTION_TYPE_LIMIT_USAGE;
        } else if (strcmp (w, "limit-objects") == 0) {
                type = GF_QUOTA_OPTION_TYPE_LIMIT_OBJECTS;
        }

        if (type == GF_QUOTA_OPTION_TYPE_LIMIT_USAGE ||
            type == GF_QUOTA_OPTION_TYPE_LIMIT_OBJECTS) {

                if (wordcount < 6 || wordcount > 7) {
                        ret = -1;
                        goto out;
                }

                if (words[4][0] != '/') {
                        cli_err ("Please enter absolute path");
                        ret = -1;
                        goto out;
                }
                ret = dict_set_str (dict, "path", (char *) words[4]);
                if (ret)
                        goto out;

                if (!words[5]) {
                        cli_err ("Please enter the limit value to be set");
                        ret = -1;
                        goto out;
                }

                if (type == GF_QUOTA_OPTION_TYPE_LIMIT_USAGE) {
                        ret = gf_string2bytesize_int64 (words[5], &value);
                        if (ret != 0 || value < 0) {
                                if (errno == ERANGE || value < 0)
                                        cli_err ("Value out of range "
                                                 "(0 - %"PRId64 "): %s",
                                                 INT64_MAX, words[5]);
                                else
                                        cli_err ("Please enter a correct "
                                                 "value");
                                goto out;
                        }
                } else {
                        errno = 0;
                        limit = strtol (words[5], &end_ptr, 10);
                        if (errno == ERANGE || errno == EINVAL || limit <= 0
                                            || strcmp (end_ptr, "") != 0) {
                                ret = -1;
                                cli_err ("Please enter an interger value in "
                                         "the range 1 - %"PRId64, INT64_MAX);
                                goto out;
                        }
                }

                ret  = dict_set_str (dict, "hard-limit", (char *) words[5]);
                if (ret < 0)
                        goto out;

                if (wordcount == 7) {

                        ret = gf_string2percent (words[6], &percent);
                        if (ret != 0) {
                                cli_err ("Please enter a correct value");
                                goto out;
                        }

                        ret = dict_set_str (dict, "soft-limit",
                                            (char *) words[6]);
                        if (ret < 0)
                                goto out;
                }

                goto set_type;
        }

        if (strcmp (w, "remove") == 0) {
                if (wordcount != 5) {
                        ret = -1;
                        goto out;
                }

                type = GF_QUOTA_OPTION_TYPE_REMOVE;

                if (words[4][0] != '/') {
                        cli_err ("Please enter absolute path");
                        ret = -1;
                        goto out;
                }

                ret = dict_set_str (dict, "path", (char *) words[4]);
                if (ret < 0)
                        goto out;
                goto set_type;
        }

        if (strcmp (w, "remove-objects") == 0) {
                if (wordcount != 5) {
                        ret = -1;
                        goto out;
                }

                type = GF_QUOTA_OPTION_TYPE_REMOVE_OBJECTS;

                if (words[4][0] != '/') {
                        cli_err ("Please enter absolute path");
                        ret = -1;
                        goto out;
                }

                ret = dict_set_str (dict, "path", (char *) words[4]);
                if (ret < 0)
                        goto out;
                goto set_type;
        }

        if (strcmp (w, "list") == 0) {

                type = GF_QUOTA_OPTION_TYPE_LIST;

                i = 4;
                while (i < wordcount) {
                        snprintf (key, 20, "path%d", i-4);

                        ret = dict_set_str (dict, key, (char *) words [i++]);
                        if (ret < 0)
                                goto out;
                }

                ret = dict_set_int32 (dict, "count", i - 4);
                if (ret < 0)
                        goto out;

                goto set_type;
        }

        if (strcmp (w, "list-objects") == 0) {
                if (wordcount < 4) {
                        ret = -1;
                        goto out;
                }

                type = GF_QUOTA_OPTION_TYPE_LIST_OBJECTS;

                i = 4;
                while (i < wordcount) {
                        snprintf (key, 20, "path%d", i-4);

                        ret = dict_set_str (dict, key, (char *) words[i++]);
                        if (ret < 0) {
                                gf_log ("cli", GF_LOG_ERROR, "Failed to set "
                                        "quota patch in request dictionary");
                                goto out;
                        }
                }

                ret = dict_set_int32 (dict, "count", i - 4);
                if (ret < 0) {
                        gf_log ("cli", GF_LOG_ERROR, "Failed to set quota "
                                "limit count in request dictionary");
                        goto out;
                }

                goto set_type;
        }

        if (strcmp (w, "alert-time") == 0) {
                if (wordcount != 5) {
                        ret = -1;
                        goto out;
                }
                type = GF_QUOTA_OPTION_TYPE_ALERT_TIME;

                ret = gf_string2time (words[4], &time);
                if (ret) {
                        cli_err ("Invalid argument %s. Please enter a valid "
                                 "string", words[4]);
                        goto out;
                }

                ret = dict_set_str (dict, "value", (char *)words[4]);
                if (ret < 0)
                        goto out;
                goto set_type;
        }

        if (strcmp (w, "soft-timeout") == 0) {
                if (wordcount != 5) {
                        ret = -1;
                        goto out;
                }
                type = GF_QUOTA_OPTION_TYPE_SOFT_TIMEOUT;

                ret = gf_string2time (words[4], &time);
                if (ret) {
                        cli_err ("Invalid argument %s. Please enter a valid "
                                 "string", words[4]);
                        goto out;
                }

                ret = dict_set_str (dict, "value", (char *)words[4]);
                if (ret < 0)
                        goto out;
                goto set_type;
        }

        if (strcmp (w, "hard-timeout") == 0) {
                if(wordcount != 5) {
                        ret = -1;
                        goto out;
                }
                type = GF_QUOTA_OPTION_TYPE_HARD_TIMEOUT;

                ret = gf_string2time (words[4], &time);
                if (ret) {
                        cli_err ("Invalid argument %s. Please enter a valid "
                                 "string", words[4]);
                        goto out;
                }

                ret = dict_set_str (dict, "value", (char *)words[4]);
                if (ret < 0)
                        goto out;
                goto set_type;
        }
        if (strcmp (w, "default-soft-limit") == 0) {
                if(wordcount != 5) {
                        ret = -1;
                        goto out;
                }
                type = GF_QUOTA_OPTION_TYPE_DEFAULT_SOFT_LIMIT;

                ret = dict_set_str (dict, "value", (char *)words[4]);
                if (ret < 0)
                        goto out;
                goto set_type;
        } else {
                GF_ASSERT (!"opword mismatch");
        }

set_type:
        ret = dict_set_int32 (dict, "type", type);
        if (ret < 0)
                goto out;

        *options = dict;
out:
        if (ret < 0) {
                if (dict)
                        dict_destroy (dict);
        }

        return ret;
}

static inline gf_boolean_t
cli_is_key_spl (char *key)
{
        return (strcmp (key, "group") == 0);
}

static int
cli_add_key_group (dict_t *dict, char *key, char *value, char **op_errstr)
{
        int             ret = -1;
        int             opt_count = 0;
        char            iter_key[1024] = {0,};
        char            iter_val[1024] = {0,};
        char            *saveptr = NULL;
        char            *tok_key = NULL;
        char            *tok_val = NULL;
        char            *dkey = NULL;
        char            *dval = NULL;
        char            *tagpath = NULL;
        char            *buf = NULL;
        char            line[PATH_MAX + 256] = {0,};
        char            errstr[2048] = "";
        FILE            *fp = NULL;

        ret = gf_asprintf (&tagpath, "%s/groups/%s",
                           GLUSTERD_DEFAULT_WORKDIR, value);
        if (ret == -1) {
                tagpath = NULL;
                goto out;
        }

        fp = fopen (tagpath, "r");
        if (!fp) {
                ret = -1;
                snprintf(errstr, sizeof(errstr), "Unable to open file '%s'."
                         " Error: %s", tagpath, strerror (errno));
                if (op_errstr)
                        *op_errstr = gf_strdup(errstr);
                goto out;
        }

        opt_count = 0;
        buf = line;
        while (fscanf (fp, "%s", buf) != EOF) {

                opt_count++;
                tok_key = strtok_r (line, "=", &saveptr);
                tok_val = strtok_r (NULL, "=", &saveptr);
                if (!tok_key || !tok_val) {
                        ret = -1;
                        snprintf(errstr, sizeof(errstr), "'%s' file format "
                                 "not valid.", tagpath);
                        if (op_errstr)
                                *op_errstr = gf_strdup(errstr);
                        goto out;
                }

                snprintf (iter_key, sizeof (iter_key), "key%d", opt_count);
                dkey = gf_strdup (tok_key);
                ret = dict_set_dynstr (dict, iter_key, dkey);
                if (ret)
                        goto out;
                dkey = NULL;

                snprintf (iter_val, sizeof (iter_val), "value%d", opt_count);
                dval = gf_strdup (tok_val);
                ret = dict_set_dynstr (dict, iter_val, dval);
                if (ret)
                        goto out;
                dval = NULL;

        }

        if (!opt_count) {
                ret = -1;
                snprintf(errstr, sizeof(errstr), "'%s' file format "
                         "not valid.", tagpath);
                if (op_errstr)
                        *op_errstr = gf_strdup(errstr);
                goto out;
        }
        ret = dict_set_int32 (dict, "count", opt_count);
out:

        GF_FREE (tagpath);

        if (ret) {
                GF_FREE (dkey);
                GF_FREE (dval);
        }

        if (fp)
                fclose (fp);

        return ret;
}

int32_t
cli_cmd_volume_set_parse (struct cli_state *state, const char **words,
                          int wordcount, dict_t **options, char **op_errstr)
{
        dict_t                 *dict      = NULL;
        char                   *volname   = NULL;
        int                     ret       = -1;
        int                     count     = 0;
        char                   *key       = NULL;
        char                   *value     = NULL;
        int                     i         = 0;
        char                    str[50]   = {0,};
        const char             *question  = NULL;
        gf_answer_t             answer    = GF_ANSWER_NO;

        GF_ASSERT (words);
        GF_ASSERT (options);

        dict = dict_new ();

        if (!dict)
                goto out;

        if (wordcount < 3)
                goto out;

        volname = (char *)words[2];

        GF_ASSERT (volname);

        ret = dict_set_str (dict, "volname", volname);

        if (ret)
                goto out;

        if (!strcmp (volname, "all")) {
                ret = dict_set_str (dict, "globalname", "All");
                if (ret) {
                        gf_log (THIS->name, GF_LOG_ERROR,
                                "dict set on global key failed.");
                        goto out;
                }

                ret = dict_set_int32 (dict, "hold_global_locks", _gf_true);
                if (ret) {
                        gf_log (THIS->name, GF_LOG_ERROR,
                                "dict set on global key failed.");
                        goto out;
                }
        }

        if ((!strcmp (volname, "help") || !strcmp (volname, "help-xml"))
            && wordcount == 3 ) {
                ret = dict_set_str (dict, volname, volname);
                if (ret)
                        goto out;

        } else if (wordcount < 5) {
                ret = -1;
                goto out;

        } else if (wordcount == 5  && cli_is_key_spl ((char *)words[3])) {
                key = (char *) words[3];
                value = (char *) words[4];
                if ( !key || !value) {
                        ret = -1;
                        goto out;
                }

                ret = gf_strip_whitespace (value, strlen (value));
                if (ret == -1)
                        goto out;

                if (strlen (value) == 0) {
                        ret = -1;
                        goto out;
                }

                ret = cli_add_key_group (dict, key, value, op_errstr);
                if (ret == 0)
                        *options = dict;
                goto out;
        }

        for (i = 3; i < wordcount; i+=2) {

                key = (char *) words[i];
                value = (char *) words[i+1];

                if ( !key || !value) {
                        ret = -1;
                        goto out;
                }

                count++;

                ret = gf_strip_whitespace (value, strlen (value));
                if (ret == -1)
                        goto out;

                if (strlen (value) == 0) {
                        ret = -1;
                        goto out;
                }

                if (cli_is_key_spl (key)) {
                        ret = -1;
                        goto out;
                }

                sprintf (str, "key%d", count);
                ret = dict_set_str (dict, str, key);
                if (ret)
                        goto out;

                sprintf (str, "value%d", count);
                ret = dict_set_str (dict, str, value);

                if (ret)
                        goto out;

                if ((!strcmp (key, "cluster.enable-shared-storage")) &&
                    (!strcmp (value, "disable"))) {
                        question = "Disabling cluster.enable-shared-storage "
                                   "will delete the shared storage volume"
                                   "(gluster_shared_storage), which is used "
                                   "by snapshot scheduler, geo-replication "
                                   "and NFS-Ganesha. Do you still want to "
                                   "continue?";
                        answer = cli_cmd_get_confirmation (state, question);
                        if (GF_ANSWER_NO == answer) {
                                gf_log ("cli", GF_LOG_ERROR, "Operation "
                                        "cancelled, exiting");
                                *op_errstr = gf_strdup ("Aborted by user.");
                                ret = -1;
                                goto out;
                        }
                }
        }

        ret = dict_set_int32 (dict, "count", wordcount-3);

        if (ret)
                goto out;

        *options = dict;

out:
        if (ret)
                dict_destroy (dict);

        return ret;
}

int32_t
cli_cmd_volume_add_brick_parse (const char **words, int wordcount,
                                dict_t **options, int *ret_type)
{
        dict_t  *dict = NULL;
        char    *volname = NULL;
        int     ret = -1;
        int     brick_count = 0, brick_index = 0;
        char    *bricks = NULL;
        char    *opwords_cl[] = { "replica", "stripe", NULL };
        gf1_cluster_type type = GF_CLUSTER_TYPE_NONE;
        int     count = 1;
        char    *w = NULL;
        int     index;
        gf_boolean_t is_force = _gf_false;
        int wc = wordcount;

        GF_ASSERT (words);
        GF_ASSERT (options);

        dict = dict_new ();

        if (!dict)
                goto out;

        if (wordcount < 3)
                goto out;

        volname = (char *)words[2];

        GF_ASSERT (volname);

        ret = dict_set_str (dict, "volname", volname);

        if (ret)
                goto out;

        if (wordcount < 4) {
                ret = -1;
                goto out;
        }
        if (wordcount < 6) {
                /* seems no options are given, go directly to the parse_brick */
                brick_index = 3;
                type = GF_CLUSTER_TYPE_NONE;
                goto parse_bricks;
        }

        w = str_getunamb (words[3], opwords_cl);
        if (!w) {
                type = GF_CLUSTER_TYPE_NONE;
                index = 3;
        } else if ((strcmp (w, "replica")) == 0) {
                type = GF_CLUSTER_TYPE_REPLICATE;
                count = strtol (words[4], NULL, 0);
                if (!count || (count < 2)) {
                        cli_err ("replica count should be greater than 1");
                        ret = -1;
                        goto out;
                }
                ret = dict_set_int32 (dict, "replica-count", count);
                if (ret)
                        goto out;
                index = 5;
        } else if ((strcmp (w, "stripe")) == 0) {
                type = GF_CLUSTER_TYPE_STRIPE;
                count = strtol (words[4], NULL, 0);
                if (!count || (count < 2)) {
                        cli_err ("stripe count should be greater than 1");
                        ret = -1;
                        goto out;
                }
                ret = dict_set_int32 (dict, "stripe-count", count);
                if (ret)
                        goto out;
                index = 5;
        } else {
                GF_ASSERT (!"opword mismatch");
                ret = -1;
                goto out;
        }

        brick_index = index;

parse_bricks:

        if (strcmp (words[wordcount - 1], "force") == 0) {
                is_force = _gf_true;
                wc = wordcount - 1;
        }

        ret = cli_cmd_bricks_parse (words, wc, brick_index, &bricks,
                                    &brick_count);
        if (ret)
                goto out;

        ret = dict_set_dynstr (dict, "bricks", bricks);
        if (ret)
                goto out;

        ret = dict_set_int32 (dict, "count", brick_count);

        if (ret)
                goto out;

        ret = dict_set_int32 (dict, "force", is_force);
        if (ret)
                goto out;

        *options = dict;

out:
        if (ret_type)
                *ret_type = type;

        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Unable to parse add-brick CLI");
                if (dict)
                        dict_destroy (dict);
        }

        return ret;
}

int32_t
cli_cmd_volume_detach_tier_parse (const char **words, int wordcount,
                                  dict_t **options)
{
        int      ret = -1;
        char    *word = NULL;
        dict_t  *dict = NULL;
        int32_t  command = GF_OP_CMD_NONE;
        int      force = 0;

        dict = dict_new ();
        if (!dict)
                goto out;

        ret = dict_set_str (dict, "volname", (char *)words[2]);
        if (ret)
                goto out;

        if (wordcount == 3 && !strcmp ((char *)words[2], "help")) {
                return -1;
        }

        if (!((wordcount == 4) || (wordcount == 5))) {
                ret = -1;
                goto out;
        }

        if (wordcount == 5) {
                word = (char *)words[4];
                if (!strcmp(word, "force"))
                        force = 1;
                else {
                        ret = -1;
                        goto out;
                }
        }

        word = (char *)words[3];

        ret = -1;

        if (!strcmp(word, "start")) {
                command = GF_OP_CMD_DETACH_START;
        } else if (!strcmp(word, "commit")) {
                if (force)
                        command = GF_OP_CMD_DETACH_COMMIT_FORCE;
                else
                        command = GF_OP_CMD_DETACH_COMMIT;
        } else if (!strcmp(word, "stop"))
                command = GF_OP_CMD_STOP_DETACH_TIER;
        else if (!strcmp(word, "status"))
                command = GF_OP_CMD_STATUS;
        else
                goto out;

        ret = dict_set_int32 (dict, "command", command);
        if (ret)
                goto out;

        *options = dict;
        ret = 0;
out:
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Unable to parse detach-tier CLI");
                if (dict)
                        dict_unref (dict);
        }

        return ret;
}

int32_t
cli_cmd_volume_remove_brick_parse (const char **words, int wordcount,
                                   dict_t **options, int *question)
{
        dict_t  *dict = NULL;
        char    *volname = NULL;
        char    *delimiter = NULL;
        int     ret = -1;
        char    key[50];
        int     brick_count = 0, brick_index = 0;
        int32_t tmp_index = 0;
        int32_t j = 0;
        char    *tmp_brick = NULL;
        char    *tmp_brick1 = NULL;
        char    *type_opword[] = { "replica", NULL };
        char    *opwords[] = { "start", "commit", "stop", "status",
                               "force", NULL };
        char    *w = NULL;
        int32_t  command = GF_OP_CMD_NONE;
        long     count = 0;

        GF_ASSERT (words);
        GF_ASSERT (options);

        if (wordcount < 5)
                goto out;

        dict = dict_new ();
        if (!dict)
                goto out;

        volname = (char *)words[2];

        GF_ASSERT (volname);

        ret = dict_set_str (dict, "volname", volname);
        if (ret)
                goto out;

        brick_index = 3;
        w = str_getunamb (words[3], type_opword);
        if (w && !strcmp ("replica", w)) {
                if (wordcount < 6) {
                        ret = -1;
                        goto out;
                }
                count = strtol (words[4], NULL, 0);
                if (count < 1) {
                        cli_err ("replica count should be greater than 0 in "
                                 "case of remove-brick");
                        ret = -1;
                        goto out;
                }

                ret = dict_set_int32 (dict, "replica-count", count);
                if (ret)
                        goto out;
                brick_index = 5;
        } else if (w) {
                GF_ASSERT (!"opword mismatch");
        }

        w = str_getunamb (words[wordcount - 1], opwords);
        if (!w) {
                ret = -1;
                goto out;
        } else {
                /* handled this option */
                wordcount--;
                if (!strcmp ("start", w)) {
                        command = GF_OP_CMD_START;
                } else if (!strcmp ("commit", w)) {
                        command = GF_OP_CMD_COMMIT;
                        if (question)
                                *question = 1;
                } else if (!strcmp ("stop", w)) {
                        command = GF_OP_CMD_STOP;
                } else if (!strcmp ("status", w)) {
                        command = GF_OP_CMD_STATUS;
                } else if (!strcmp ("force", w)) {
                        command = GF_OP_CMD_COMMIT_FORCE;
                        if (question)
                                *question = 1;
                } else {
                        GF_ASSERT (!"opword mismatch");
                        ret = -1;
                        goto out;
                }
        }

        ret = dict_set_int32 (dict, "command", command);
        if (ret)
                gf_log ("cli", GF_LOG_INFO, "failed to set 'command' %d",
                        command);


        tmp_index = brick_index;
        tmp_brick = GF_MALLOC(2048 * sizeof(*tmp_brick), gf_common_mt_char);

        if (!tmp_brick) {
                gf_log ("",GF_LOG_ERROR,"cli_cmd_volume_remove_brick_parse: "
                        "Unable to get memory");
                ret = -1;
                goto out;
        }

        tmp_brick1 = GF_MALLOC(2048 * sizeof(*tmp_brick1), gf_common_mt_char);

        if (!tmp_brick1) {
                gf_log ("",GF_LOG_ERROR,"cli_cmd_volume_remove_brick_parse: "
                        "Unable to get memory");
                ret = -1;
                goto out;
        }

        while (brick_index < wordcount) {
                if (validate_brick_name ((char *)words[brick_index])) {
                        cli_err ("wrong brick type: %s, use <HOSTNAME>:"
                                 "<export-dir-abs-path>", words[brick_index]);
                        ret = -1;
                        goto out;
                } else {
                        delimiter = strrchr(words[brick_index], ':');
                        ret = gf_canonicalize_path (delimiter + 1);
                        if (ret)
                                goto out;
                }

                j = tmp_index;
                strcpy(tmp_brick, words[brick_index]);
                while ( j < brick_index) {
                        strcpy(tmp_brick1, words[j]);
                        if (!(strcmp (tmp_brick, tmp_brick1))) {
                                gf_log("",GF_LOG_ERROR, "Duplicate bricks"
                                       " found %s", words[brick_index]);
                                cli_err("Duplicate bricks found %s",
                                        words[brick_index]);
                                ret = -1;
                                goto out;
                        }
                        j++;
                }
                snprintf (key, 50, "brick%d", ++brick_count);
                ret = dict_set_str (dict, key, (char *)words[brick_index++]);

                if (ret)
                        goto out;
        }

        if (command != GF_OP_CMD_STATUS && command != GF_OP_CMD_STOP) {
                ret = dict_set_int32 (dict, "count", brick_count);
                if (ret)
                        goto out;
        }

        *options = dict;

out:
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Unable to parse remove-brick CLI");
                if (dict)
                        dict_destroy (dict);
        }

        GF_FREE (tmp_brick);
        GF_FREE (tmp_brick1);

        return ret;
}


int32_t
cli_cmd_volume_replace_brick_parse (const char **words, int wordcount,
                                    dict_t **options)
{
        int                   ret        = -1;
        char                 *volname    = NULL;
        char                 *delimiter  = NULL;
        dict_t               *dict       = NULL;

        GF_ASSERT (words);
        GF_ASSERT (options);

        if (wordcount != 7) {
                ret = -1;
                goto out;
        }

        dict = dict_new ();

        if (!dict) {
                gf_log ("cli", GF_LOG_ERROR, "Failed to allocate dictionary");
                goto out;
        }

        volname = (char *)words[2];

        GF_ASSERT (volname);

        ret = dict_set_str (dict, "volname", volname);
        if (ret)
                goto out;

        if (validate_brick_name ((char *)words[3])) {
                cli_err ("wrong brick type: %s, use "
                         "<HOSTNAME>:<export-dir-abs-path>", words[3]);
                ret = -1;
                goto out;
        } else {
                delimiter = strrchr ((char *)words[3], ':');
                ret = gf_canonicalize_path (delimiter + 1);
                if (ret)
                        goto out;
        }

        ret = dict_set_str (dict, "src-brick", (char *)words[3]);
        if (ret)
                goto out;

        if (validate_brick_name ((char *)words[4])) {
                cli_err ("wrong brick type: %s, use "
                         "<HOSTNAME>:<export-dir-abs-path>", words[4]);
                ret = -1;
                goto out;
        } else {
                delimiter = strrchr ((char *)words[4], ':');
                ret = gf_canonicalize_path (delimiter + 1);
                if (ret)
                        goto out;
        }

        ret = dict_set_str (dict, "dst-brick", (char *)words[4]);
        if (ret)
                goto out;

        /* commit force option */
        if (strcmp ("commit", words[5]) || strcmp ("force", words[6])) {
                cli_err ("Invalid option '%s' '%s' for replace-brick. Please "
                         "enter valid replace-brick command", words[5],
                         words[6]);
                ret = -1;
                goto out;
        }

        ret = dict_set_str (dict, "operation", "GF_REPLACE_OP_COMMIT_FORCE");
        if (ret)
                goto out;

        *options = dict;

out:
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Unable to parse replace-brick CLI");
                if (dict)
                        dict_destroy (dict);
        }

        return ret;
}

int32_t
cli_cmd_log_filename_parse (const char **words, int wordcount, dict_t **options)
{
        dict_t  *dict = NULL;
        char    *volname = NULL;
        char    *str = NULL;
        int     ret = -1;
        char    *delimiter = NULL;

        GF_ASSERT (words);
        GF_ASSERT (options);

        dict = dict_new ();
        if (!dict)
                goto out;

        volname = (char *)words[3];
        GF_ASSERT (volname);

        ret = dict_set_str (dict, "volname", volname);
        if (ret)
                goto out;

        str = (char *)words[4];
        if (strchr (str, ':')) {
                delimiter = strchr (words[4], ':');
                if (!delimiter || delimiter == words[4]
                    || *(delimiter+1) != '/') {
                        cli_err ("wrong brick type: %s, use <HOSTNAME>:"
                                 "<export-dir-abs-path>", words[4]);
                        ret = -1;
                        goto out;
                } else {
                        ret = gf_canonicalize_path (delimiter + 1);
                        if (ret)
                                goto out;
                }
                ret = dict_set_str (dict, "brick", str);
                if (ret)
                        goto out;
                /* Path */
                str = (char *)words[5];
                ret = dict_set_str (dict, "path", str);
                if (ret)
                        goto out;
        } else {
                ret = dict_set_str (dict, "path", str);
                if (ret)
                        goto out;
        }

        *options = dict;

out:
        if (ret && dict)
                dict_destroy (dict);

        return ret;
}

int32_t
cli_cmd_log_level_parse (const char **words, int worcount, dict_t **options)
{
        dict_t *dict            = NULL;
        int     ret             = -1;

        GF_ASSERT (words);
        GF_ASSERT (options);

        /*
         * loglevel command format:
         *  > volume log level <VOL> <XLATOR[*]> <LOGLEVEL>
         *  > volume log level colon-o posix WARNING
         *  > volume log level colon-o replicate* DEBUG
         *  > volume log level coon-o * TRACE
         */

        GF_ASSERT ((strncmp(words[0], "volume", 6) == 0));
        GF_ASSERT ((strncmp(words[1], "log", 3) == 0));
        GF_ASSERT ((strncmp(words[2], "level", 5) == 0));

        ret = glusterd_check_log_level(words[5]);
        if (ret == -1) {
                cli_err("Invalid log level [%s] specified", words[5]);
                cli_err("Valid values for loglevel: (DEBUG|WARNING|ERROR"
                        "|CRITICAL|NONE|TRACE)");
                goto out;
        }

        dict = dict_new ();
        if (!dict)
                goto out;

        GF_ASSERT(words[3]);
        GF_ASSERT(words[4]);

        ret = dict_set_str (dict, "volname", (char *)words[3]);
        if (ret)
                goto out;

        ret = dict_set_str (dict, "xlator", (char *)words[4]);
        if (ret)
                goto out;

        ret = dict_set_str (dict, "loglevel", (char *)words[5]);
        if (ret)
                goto out;

        *options = dict;

 out:
        if (ret && dict)
                dict_destroy (dict);

        return ret;
}

int32_t
cli_cmd_log_locate_parse (const char **words, int wordcount, dict_t **options)
{
        dict_t  *dict = NULL;
        char    *volname = NULL;
        char    *str = NULL;
        int     ret = -1;
        char    *delimiter = NULL;

        GF_ASSERT (words);
        GF_ASSERT (options);

        dict = dict_new ();
        if (!dict)
                goto out;

        volname = (char *)words[3];
        GF_ASSERT (volname);

        ret = dict_set_str (dict, "volname", volname);
        if (ret)
                goto out;

        if (words[4]) {
                delimiter = strchr (words[4], ':');
                if (!delimiter || delimiter == words[4]
                    || *(delimiter+1) != '/') {
                        cli_err ("wrong brick type: %s, use <HOSTNAME>:"
                                 "<export-dir-abs-path>", words[4]);
                        ret = -1;
                        goto out;
                } else {
                        ret = gf_canonicalize_path (delimiter + 1);
                        if (ret)
                                goto out;
                }
                str = (char *)words[4];
                ret = dict_set_str (dict, "brick", str);
                if (ret)
                        goto out;
        }

        *options = dict;

out:
        if (ret && dict)
                dict_destroy (dict);

        return ret;
}

int32_t
cli_cmd_log_rotate_parse (const char **words, int wordcount, dict_t **options)
{
        dict_t  *dict = NULL;
        char    *volname = NULL;
        char    *str = NULL;
        int     ret = -1;
        char    *delimiter = NULL;

        GF_ASSERT (words);
        GF_ASSERT (options);

        dict = dict_new ();
        if (!dict)
                goto out;

        if (strcmp ("rotate", words[3]) == 0)
                volname = (char *)words[2];
        else if (strcmp ("rotate", words[2]) == 0)
                volname = (char *)words[3];
        GF_ASSERT (volname);

        ret = dict_set_str (dict, "volname", volname);
        if (ret)
                goto out;

        if (words[4]) {
                delimiter = strchr (words[4], ':');
                if (!delimiter || delimiter == words[4]
                    || *(delimiter+1) != '/') {
                        cli_err ("wrong brick type: %s, use <HOSTNAME>:"
                                 "<export-dir-abs-path>", words[4]);
                        ret = -1;
                        goto out;
                } else {
                        ret = gf_canonicalize_path (delimiter + 1);
                        if (ret)
                                goto out;
                }
                str = (char *)words[4];
                ret = dict_set_str (dict, "brick", str);
                if (ret)
                        goto out;
        }

        *options = dict;

out:
        if (ret && dict)
                dict_destroy (dict);

        return ret;
}

static gf_boolean_t
gsyncd_url_check (const char *w)
{
        return !!strpbrk (w, ":/");
}

static gf_boolean_t
gsyncd_glob_check (const char *w)
{
        return !!strpbrk (w, "*?[");
}

static int
config_parse (const char **words, int wordcount, dict_t *dict,
              unsigned cmdi, unsigned glob)
{
        int32_t            ret     = -1;
        int32_t            i       = -1;
        char               *append_str = NULL;
        size_t             append_len = 0;
        char               *subop = NULL;

        switch ((wordcount - 1) - cmdi) {
        case 0:
                subop = gf_strdup ("get-all");
                break;
        case 1:
                if (words[cmdi + 1][0] == '!') {
                        (words[cmdi + 1])++;
                        if (gf_asprintf (&subop, "del%s",
                                         glob ? "-glob" : "") == -1)
                                subop = NULL;
                } else
                        subop = gf_strdup ("get");

                ret = dict_set_str (dict, "op_name", ((char *)words[cmdi + 1]));
                if (ret < 0)
                        goto out;
                break;
        default:
                if (gf_asprintf (&subop, "set%s", glob ? "-glob" : "") == -1)
                        subop = NULL;

                ret = dict_set_str (dict, "op_name", ((char *)words[cmdi + 1]));
                if (ret < 0)
                        goto out;

                /* join the varargs by spaces to get the op_value */

                for (i = cmdi + 2; i < wordcount; i++)
                        append_len += (strlen (words[i]) + 1);
                /* trailing strcat will add two bytes, make space for that */
                append_len++;

                append_str = GF_CALLOC (1, append_len, cli_mt_append_str);
                if (!append_str) {
                        ret = -1;
                        goto out;
                }

                for (i = cmdi + 2; i < wordcount; i++) {
                        strcat (append_str, words[i]);
                        strcat (append_str, " ");
                }
                append_str[append_len - 2] = '\0';
                /* "checkpoint now" is special: we resolve that "now" */
                if ((strcmp (words[cmdi + 1], "checkpoint") == 0) &&
                    (strcmp (append_str, "now") == 0)) {
                        struct timeval tv = {0,};

                        ret = gettimeofday (&tv, NULL);
                        if (ret == -1)
                                goto out;

                        GF_FREE (append_str);
                        append_str = GF_CALLOC (1, 300, cli_mt_append_str);
                        if (!append_str) {
                                ret = -1;
                                goto out;
                        }
                        snprintf (append_str, 300, "%" GF_PRI_SECOND,
                                  tv.tv_sec);
                }

                ret = dict_set_dynstr (dict, "op_value", append_str);
        }

        ret = -1;
        if (subop) {
                ret = dict_set_dynstr (dict, "subop", subop);
                if (!ret)
                      subop = NULL;
        }

out:
        if (ret && append_str)
                GF_FREE (append_str);

        GF_FREE (subop);

        gf_log ("cli", GF_LOG_DEBUG, "Returning %d", ret);
        return ret;
}

static int32_t
force_push_pem_no_verify_parse (const char **words, int wordcount,
                      dict_t *dict, unsigned *cmdi)
{
        int32_t            ret     = 0;

        if (!strcmp ((char *)words[wordcount-1], "force")) {
                if ((strcmp ((char *)words[wordcount-2], "start")) &&
                    (strcmp ((char *)words[wordcount-2], "stop")) &&
                    (strcmp ((char *)words[wordcount-2], "create")) &&
                    (strcmp ((char *)words[wordcount-2], "no-verify")) &&
                    (strcmp ((char *)words[wordcount-2], "push-pem")) &&
                    (strcmp ((char *)words[wordcount-2], "pause")) &&
                    (strcmp ((char *)words[wordcount-2], "resume"))) {
                        ret = -1;
                        goto out;
                }
                ret = dict_set_uint32 (dict, "force",
                                       _gf_true);
                if (ret)
                        goto out;
                (*cmdi)++;

                if (!strcmp ((char *)words[wordcount-2], "push-pem")) {
                        if (strcmp ((char *)words[wordcount-3], "create")) {
                                ret = -1;
                                goto out;
                        }
                        ret = dict_set_int32 (dict, "push_pem", 1);
                        if (ret)
                                goto out;
                        (*cmdi)++;
                } else if (!strcmp ((char *)words[wordcount-2], "no-verify")) {
                        if (strcmp ((char *)words[wordcount-3], "create")) {
                                ret = -1;
                                goto out;
                        }
                        ret = dict_set_uint32 (dict, "no_verify",
                                               _gf_true);
                        if (ret)
                                goto out;
                        (*cmdi)++;
                }
        } else if (!strcmp ((char *)words[wordcount-1], "push-pem")) {
                if (strcmp ((char *)words[wordcount-2], "create")) {
                        ret = -1;
                        goto out;
                }
                ret = dict_set_int32 (dict, "push_pem", 1);
                if (ret)
                        goto out;
                (*cmdi)++;
        } else if (!strcmp ((char *)words[wordcount-1], "no-verify")) {
                if ((strcmp ((char *)words[wordcount-2], "create"))) {
                        ret = -1;
                        goto out;
                }
                ret = dict_set_uint32 (dict, "no_verify",
                                       _gf_true);
                if (ret)
                        goto out;
                (*cmdi)++;
        }

out:
        gf_log ("cli", GF_LOG_DEBUG, "Returning %d", ret);
        return ret;
}


int32_t
cli_cmd_gsync_set_parse (const char **words, int wordcount, dict_t **options)
{
        int32_t            ret     = -1;
        dict_t             *dict   = NULL;
        gf1_cli_gsync_set  type    = GF_GSYNC_OPTION_TYPE_NONE;
        int                i       = 0;
        unsigned           masteri = 0;
        unsigned           slavei  = 0;
        unsigned           glob    = 0;
        unsigned           cmdi    = 0;
        char               *opwords[] = { "create", "status", "start", "stop",
                                          "config", "force", "delete", "no-verify"
                                          "push-pem", "detail", "pause",
                                          "resume", NULL };
        char               *w = NULL;
        char               *save_ptr   = NULL;
        char               *slave_temp = NULL;
        char               *token      = NULL;

        GF_ASSERT (words);
        GF_ASSERT (options);

        dict = dict_new ();
        if (!dict)
                goto out;

        /* new syntax:
         *
         * volume geo-replication $m $s create [[no-verify] | [push-pem]] [force]
         * volume geo-replication [$m [$s]] status [detail]
         * volume geo-replication [$m] $s config [[!]$opt [$val]]
         * volume geo-replication $m $s start|stop [force]
         * volume geo-replication $m $s delete
         * volume geo-replication $m $s pause [force]
         * volume geo-replication $m $s resume [force]
         */

        if (wordcount < 3)
                goto out;

        for (i = 2; i <= 3 && i < wordcount - 1; i++) {
                if (gsyncd_glob_check (words[i]))
                        glob = i;
                if (gsyncd_url_check (words[i])) {
                        slavei = i;
                        break;
                }
        }

        if (glob && !slavei)
                /* glob is allowed only for config, thus it implies there is a
                 * slave argument; but that might have not been recognized on
                 * the first scan as it's url characteristics has been covered
                 * by the glob syntax.
                 *
                 * In this case, the slave is perforce the last glob-word -- the
                 * upcoming one is neither glob, nor url, so it's definitely not
                 * the slave.
                 */
                slavei = glob;
        if (slavei) {
                cmdi = slavei + 1;
                if (slavei == 3)
                        masteri = 2;
        } else if (i <= 3) {
                if (!strcmp ((char *)words[wordcount-1], "detail")) {
                        /* For status detail it is mandatory to provide
                         * both master and slave */
                        ret = -1;
                        goto out;
                }

                /* no $s, can only be status cmd
                 * (with either a single $m before it or nothing)
                 * -- these conditions imply that i <= 3 after
                 * the iteration and that i is the successor of
                 * the (0 or 1 length) sequence of $m-s.
                 */
                cmdi = i;
                if (i == 3)
                        masteri = 2;
        } else
                goto out;

        /* now check if input really complies syntax
         * (in a somewhat redundant way, in favor
         * transparent soundness)
         */

        if (masteri && gsyncd_url_check (words[masteri]))
                goto out;
        if (slavei && !glob && !gsyncd_url_check (words[slavei]))
                goto out;

        w = str_getunamb (words[cmdi], opwords);
        if (!w)
                goto out;

        if (strcmp (w, "create") == 0) {
                type = GF_GSYNC_OPTION_TYPE_CREATE;

                if (!masteri || !slavei)
                        goto out;
        } else if (strcmp (w, "status") == 0) {
                type = GF_GSYNC_OPTION_TYPE_STATUS;

                if (slavei && !masteri)
                        goto out;
        } else if (strcmp (w, "config") == 0) {
                type = GF_GSYNC_OPTION_TYPE_CONFIG;

                if (!slavei)
                        goto out;
        } else if (strcmp (w, "start") == 0) {
                type = GF_GSYNC_OPTION_TYPE_START;

                if (!masteri || !slavei)
                        goto out;
        } else if (strcmp (w, "stop") == 0) {
                type = GF_GSYNC_OPTION_TYPE_STOP;

                if (!masteri || !slavei)
                        goto out;
        } else if (strcmp (w, "delete") == 0) {
                type = GF_GSYNC_OPTION_TYPE_DELETE;

                if (!masteri || !slavei)
                        goto out;
        } else if (strcmp (w, "pause") == 0) {
                type = GF_GSYNC_OPTION_TYPE_PAUSE;

                if (!masteri || !slavei)
                        goto out;
        } else if (strcmp (w, "resume") == 0) {
                type = GF_GSYNC_OPTION_TYPE_RESUME;

                if (!masteri || !slavei)
                        goto out;
        } else
                GF_ASSERT (!"opword mismatch");

        ret = force_push_pem_no_verify_parse (words, wordcount, dict, &cmdi);
        if (ret)
                goto out;

        if (!strcmp ((char *)words[wordcount-1], "detail")) {
                if (strcmp ((char *)words[wordcount-2], "status")) {
                        ret = -1;
                        goto out;
                }
                if (!slavei || !masteri) {
                        ret = -1;
                        goto out;
                }
                ret = dict_set_uint32 (dict, "status-detail", _gf_true);
                if (ret)
                        goto out;
                cmdi++;
        }

        if (type != GF_GSYNC_OPTION_TYPE_CONFIG &&
            (cmdi < wordcount - 1 || glob))
                goto out;

        /* If got so far, input is valid, assemble the message */

        ret = 0;

        if (masteri) {
                ret = dict_set_str (dict, "master", (char *)words[masteri]);
                if (!ret)
                        ret = dict_set_str (dict, "volname",
                                            (char *)words[masteri]);
        }
        if (!ret && slavei) {
                /* If geo-rep is created with root user using the syntax
                 * gluster vol geo-rep <mastervol> root@<slavehost> ...
                 * pass down only <slavehost> else pass as it is.
                 */
                slave_temp = gf_strdup (words[slavei]);
                token = strtok_r (slave_temp, "@", &save_ptr);
                if (token && !strcmp (token, "root")) {
                        ret = dict_set_str (dict, "slave",
                                            (char *)words[slavei]+5);
                } else {
                        ret = dict_set_str (dict, "slave",
                                            (char *)words[slavei]);
                }
        }
        if (!ret)
                ret = dict_set_int32 (dict, "type", type);
        if (!ret && type == GF_GSYNC_OPTION_TYPE_CONFIG)
                ret = config_parse (words, wordcount, dict, cmdi, glob);

out:
        if (slave_temp)
                GF_FREE (slave_temp);
        if (ret) {
                if (dict)
                        dict_destroy (dict);
        } else
                *options = dict;


        return ret;
}

int32_t
cli_cmd_volume_profile_parse (const char **words, int wordcount,
                              dict_t **options)
{
        dict_t              *dict       = NULL;
        char                *volname    = NULL;
        int                 ret         = -1;
        gf1_cli_stats_op    op          = GF_CLI_STATS_NONE;
        gf1_cli_info_op     info_op     = GF_CLI_INFO_NONE;
        gf_boolean_t        is_peek     = _gf_false;

        char      *opwords[] = { "start", "stop", "info", NULL };
        char      *w = NULL;

        GF_ASSERT (words);
        GF_ASSERT (options);

        dict = dict_new ();
        if (!dict)
                goto out;

        if (wordcount < 4)
                goto out;

        volname = (char *)words[2];

        ret = dict_set_str (dict, "volname", volname);
        if (ret)
                goto out;

        w = str_getunamb (words[3], opwords);
        if (!w) {
                ret = -1;
                goto out;
        }

        if ((strcmp (w, "start") == 0 || strcmp (w, "stop") == 0) &&
            wordcount > 5)
                goto out;

        if (strcmp (w, "info") == 0 && wordcount > 7)
                goto out;

        if (strcmp (w, "start") == 0) {
                op = GF_CLI_STATS_START;
        } else if (strcmp (w, "stop") == 0) {
                op = GF_CLI_STATS_STOP;
        } else if (strcmp (w, "info") == 0) {
                op = GF_CLI_STATS_INFO;
                info_op = GF_CLI_INFO_ALL;
                if (wordcount > 4) {
                        if (strcmp (words[4], "incremental") == 0) {
                                info_op = GF_CLI_INFO_INCREMENTAL;
                                if (wordcount > 5 &&
                                    strcmp (words[5], "peek") == 0) {
                                        is_peek = _gf_true;
                                }
                        } else if (strcmp (words[4], "cumulative") == 0) {
                                info_op = GF_CLI_INFO_CUMULATIVE;
                        } else if (strcmp (words[4], "clear") == 0) {
                                info_op = GF_CLI_INFO_CLEAR;
                        } else if (strcmp (words[4], "peek") == 0) {
                                is_peek = _gf_true;
                        }
                }
        } else
                GF_ASSERT (!"opword mismatch");

        ret = dict_set_int32 (dict, "op", (int32_t)op);
        if (ret)
                goto out;

        ret = dict_set_int32 (dict, "info-op", (int32_t)info_op);
        if (ret)
                goto out;

        ret = dict_set_int32 (dict, "peek", is_peek);
        if (ret)
                goto out;

        if (!strcmp (words[wordcount - 1], "nfs")) {
                ret = dict_set_int32 (dict, "nfs", _gf_true);
                if (ret)
                        goto out;
        }

        *options = dict;
out:
        if (ret && dict)
                dict_destroy (dict);
        return ret;
}

int32_t
cli_cmd_volume_top_parse (const char **words, int wordcount,
                              dict_t **options)
{
        dict_t  *dict           = NULL;
        char    *volname        = NULL;
        char    *value          = NULL;
        char    *key            = NULL;
        int      ret            = -1;
        gf1_cli_stats_op op = GF_CLI_STATS_NONE;
        gf1_cli_top_op    top_op = GF_CLI_TOP_NONE;
        int32_t  list_cnt       = -1;
        int      index          = 0;
        int      perf           = 0;
        int32_t  blk_size      = 0;
        uint32_t  count         = 0;
        gf_boolean_t nfs        = _gf_false;
        char    *delimiter      = NULL;
        char    *opwords[]      = { "open", "read", "write", "opendir",
                                    "readdir", "read-perf", "write-perf",
                                    "clear", NULL };
        char    *w = NULL;

        GF_ASSERT (words);
        GF_ASSERT (options);

        dict = dict_new ();
        if (!dict)
                goto out;

        if (wordcount < 4)
                goto out;

        volname = (char *)words[2];

        ret = dict_set_str (dict, "volname", volname);
        if (ret)
                goto out;

        op = GF_CLI_STATS_TOP;
        ret = dict_set_int32 (dict, "op", (int32_t)op);
        if (ret)
                goto out;

        w = str_getunamb (words[3], opwords);
        if (!w) {
                ret = -1;
                goto out;
        }
        if (strcmp (w, "open") == 0) {
                top_op = GF_CLI_TOP_OPEN;
        } else if (strcmp (w, "read") == 0) {
                top_op = GF_CLI_TOP_READ;
        } else if (strcmp (w, "write") == 0) {
                top_op = GF_CLI_TOP_WRITE;
        } else if (strcmp (w, "opendir") == 0) {
                top_op = GF_CLI_TOP_OPENDIR;
        } else if (strcmp (w, "readdir") == 0) {
                top_op = GF_CLI_TOP_READDIR;
        } else if (strcmp (w, "read-perf") == 0) {
                top_op = GF_CLI_TOP_READ_PERF;
                perf = 1;
        } else if (strcmp (w, "write-perf") == 0) {
                top_op = GF_CLI_TOP_WRITE_PERF;
                perf = 1;
        } else if (strcmp (w, "clear") == 0) {
                ret = dict_set_int32 (dict, "clear-stats", 1);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR,
                                "Could not set clear-stats in dict");
                        goto out;
                }
        } else
                GF_ASSERT (!"opword mismatch");
        ret = dict_set_int32 (dict, "top-op", (int32_t)top_op);
        if (ret)
                goto out;

        if ((wordcount > 4) && !strcmp (words[4], "nfs")) {
                nfs = _gf_true;
                ret = dict_set_int32 (dict, "nfs", nfs);
                if (ret)
                        goto out;
                index = 5;
        } else {
                index = 4;
        }

        for (; index < wordcount; index+=2) {

                key = (char *) words[index];
                value = (char *) words[index+1];

                if ( key && !value ) {
                        ret = -1;
                        goto out;
                }
                if (!strcmp (key, "brick")) {
                        delimiter = strchr (value, ':');
                        if (!delimiter || delimiter == value
                            || *(delimiter+1) != '/') {
                                cli_err ("wrong brick type: %s, use <HOSTNAME>:"
                                         "<export-dir-abs-path>", value);
                                ret = -1;
                                goto out;
                        } else {
                                ret = gf_canonicalize_path (delimiter + 1);
                                if (ret)
                                        goto out;
                        }
                        ret = dict_set_str (dict, "brick", value);

                } else if (!strcmp (key, "list-cnt")) {
                        ret = gf_is_str_int (value);
                        if (!ret)
                                list_cnt = atoi (value);
                        if (ret || (list_cnt < 0) || (list_cnt > 100)) {
                                cli_err ("list-cnt should be between 0 to 100");
                                ret = -1;
                                goto out;
                        }
                } else if (perf && !nfs && !strcmp (key, "bs")) {
                        ret = gf_is_str_int (value);
                        if (!ret)
                                blk_size = atoi (value);
                        if (ret || (blk_size <= 0)) {
                                if (blk_size < 0)
                                        cli_err ("block size is an invalid"
                                                 " number");
                                else
                                        cli_err ("block size should be an "
                                                 "integer greater than zero");
                                ret = -1;
                                goto out;
                        }
                        ret = dict_set_uint32 (dict, "blk-size",
                                                        (uint32_t)blk_size);
                } else if (perf && !nfs && !strcmp (key, "count")) {
                        ret = gf_is_str_int (value);
                        if (!ret)
                                count = atoi(value);
                        if (ret || (count <= 0)) {
                                if (count < 0)
                                        cli_err ("count is an invalid number");
                                else
                                        cli_err ("count should be an integer "
                                                 "greater than zero");

                                ret = -1;
                                goto out;
                        }
                        ret = dict_set_uint32 (dict, "blk-cnt", count);
                } else {
                        ret = -1;
                        goto out;
                }
                if (ret) {
                        gf_log ("", GF_LOG_WARNING, "Dict set failed for "
                                "key %s", key);
                        goto out;
                }
        }
        if (list_cnt == -1)
                list_cnt = 100;
        ret = dict_set_int32 (dict, "list-cnt", list_cnt);
        if (ret) {
                gf_log ("", GF_LOG_WARNING, "Dict set failed for list_cnt");
                goto out;
        }

        if ((blk_size > 0) ^ (count > 0)) {
                cli_err ("Need to give both 'bs' and 'count'");
                ret = -1;
                goto out;
        } else if (((uint64_t)blk_size * count) > (10 * GF_UNIT_GB)) {
                cli_err ("'bs * count' value %"PRIu64" is greater than "
                         "maximum allowed value of 10GB",
                         ((uint64_t)blk_size * count));
                ret = -1;
                goto out;
        }

        *options = dict;
out:
        if (ret && dict)
                dict_destroy (dict);
        return ret;
}

uint32_t
cli_cmd_get_statusop (const char *arg)
{
        int        i         = 0;
        uint32_t   ret       = GF_CLI_STATUS_NONE;
        char      *w         = NULL;
        char      *opwords[] = {"detail", "mem", "clients", "fd",
                                "inode", "callpool", "tasks", NULL};
        struct {
                char      *opname;
                uint32_t   opcode;
        } optable[] = {
                { "detail",   GF_CLI_STATUS_DETAIL   },
                { "mem",      GF_CLI_STATUS_MEM      },
                { "clients",  GF_CLI_STATUS_CLIENTS  },
                { "fd",       GF_CLI_STATUS_FD       },
                { "inode",    GF_CLI_STATUS_INODE    },
                { "callpool", GF_CLI_STATUS_CALLPOOL },
                { "tasks",    GF_CLI_STATUS_TASKS    },
                { NULL }
        };

        w = str_getunamb (arg, opwords);
        if (!w) {
                gf_log ("cli", GF_LOG_DEBUG,
                        "Not a status op  %s", arg);
                goto out;
        }

        for (i = 0; optable[i].opname; i++) {
                if (!strcmp (w, optable[i].opname)) {
                        ret = optable[i].opcode;
                        break;
                }
        }

 out:
        return ret;
}

int
cli_cmd_volume_status_parse (const char **words, int wordcount,
                             dict_t **options)
{
        dict_t    *dict            = NULL;
        int        ret             = -1;
        uint32_t   cmd             = 0;

        GF_ASSERT (options);

        dict = dict_new ();
        if (!dict)
                goto out;

        switch (wordcount) {

        case 2:
                cmd = GF_CLI_STATUS_ALL;
                ret = 0;
                break;

        case 3:
                if (!strcmp (words[2], "all")) {
                        cmd = GF_CLI_STATUS_ALL;
                        ret = 0;

                } else {
                        cmd = GF_CLI_STATUS_VOL;
                        ret = dict_set_str (dict, "volname", (char *)words[2]);
                }

                break;

        case 4:
                cmd = cli_cmd_get_statusop (words[3]);

                if (!strcmp (words[2], "all")) {
                        if (cmd == GF_CLI_STATUS_NONE) {
                                cli_err ("%s is not a valid status option",
                                         words[3]);
                                ret = -1;
                                goto out;
                        }
                        cmd |= GF_CLI_STATUS_ALL;
                        ret  = 0;

                } else {
                        ret = dict_set_str (dict, "volname",
                                            (char *)words[2]);
                        if (ret)
                                goto out;

                        if (cmd == GF_CLI_STATUS_NONE) {
                                if (!strcmp (words[3], "nfs")) {
                                        cmd |= GF_CLI_STATUS_NFS;
                                } else if (!strcmp (words[3], "shd")) {
                                        cmd |= GF_CLI_STATUS_SHD;
                                } else if (!strcmp (words[3], "quotad")) {
                                        cmd |= GF_CLI_STATUS_QUOTAD;
                                } else if (!strcmp (words[3], "snapd")) {
                                        cmd |= GF_CLI_STATUS_SNAPD;
                                } else if (!strcmp (words[3], "bitd")) {
                                        cmd |= GF_CLI_STATUS_BITD;
                                } else if (!strcmp (words[3], "scrub")) {
                                        cmd |= GF_CLI_STATUS_SCRUB;
                                } else {
                                        cmd = GF_CLI_STATUS_BRICK;
                                        ret = dict_set_str (dict, "brick",
                                                            (char *)words[3]);
                                }

                        } else {
                                cmd |= GF_CLI_STATUS_VOL;
                                ret  = 0;
                        }
                }

                break;

        case 5:
                if (!strcmp (words[2], "all")) {
                        cli_err ("Cannot specify brick/nfs for \"all\"");
                        ret = -1;
                        goto out;
                }

                cmd = cli_cmd_get_statusop (words[4]);
                if (cmd == GF_CLI_STATUS_NONE) {
                        cli_err ("%s is not a valid status option",
                                 words[4]);
                        ret = -1;
                        goto out;
                }


                ret = dict_set_str (dict, "volname", (char *)words[2]);
                if (ret)
                        goto out;

                if (!strcmp (words[3], "nfs")) {
                        if (cmd == GF_CLI_STATUS_FD ||
                            cmd == GF_CLI_STATUS_DETAIL ||
                            cmd == GF_CLI_STATUS_TASKS) {
                                cli_err ("Detail/FD/Tasks status not available"
                                         " for NFS Servers");
                                ret = -1;
                                goto out;
                        }
                        cmd |= GF_CLI_STATUS_NFS;
                } else if (!strcmp (words[3], "shd")){
                        if (cmd == GF_CLI_STATUS_FD ||
                            cmd == GF_CLI_STATUS_CLIENTS ||
                            cmd == GF_CLI_STATUS_DETAIL ||
                            cmd == GF_CLI_STATUS_TASKS) {
                                cli_err ("Detail/FD/Clients/Tasks status not "
                                         "available for Self-heal Daemons");
                                ret = -1;
                                goto out;
                        }
                        cmd |= GF_CLI_STATUS_SHD;
                } else if (!strcmp (words[3], "quotad")) {
                        if (cmd == GF_CLI_STATUS_FD ||
                            cmd == GF_CLI_STATUS_CLIENTS ||
                            cmd == GF_CLI_STATUS_DETAIL ||
                            cmd == GF_CLI_STATUS_INODE) {
                                cli_err ("Detail/FD/Clients/Inode status not "
                                         "available for Quota Daemon");
                                ret = -1;
                                goto out;
                        }
                        cmd |= GF_CLI_STATUS_QUOTAD;
                } else if  (!strcmp (words[3], "snapd")) {
                        if (cmd == GF_CLI_STATUS_FD ||
                            cmd == GF_CLI_STATUS_CLIENTS ||
                            cmd == GF_CLI_STATUS_DETAIL ||
                            cmd == GF_CLI_STATUS_INODE) {
                                cli_err ("Detail/FD/Clients/Inode status not "
                                         "available for snap daemon");
                                ret = -1;
                                goto out;
                        }
                        cmd |= GF_CLI_STATUS_SNAPD;
                } else {
                        if (cmd == GF_CLI_STATUS_TASKS) {
                                cli_err ("Tasks status not available for "
                                         "bricks");
                                ret = -1;
                                goto out;
                        }
                        cmd |= GF_CLI_STATUS_BRICK;
                        ret = dict_set_str (dict, "brick", (char *)words[3]);
                }
                break;

        default:
                goto out;
        }

        if (ret)
                goto out;

        ret = dict_set_int32 (dict, "cmd", cmd);
        if (ret)
                goto out;

        *options = dict;

 out:
        if (ret && dict)
                dict_destroy (dict);

        return ret;
}

gf_boolean_t
cli_cmd_validate_dumpoption (const char *arg, char **option)
{
        char    *opwords[] = {"all", "nfs", "mem", "iobuf", "callpool", "priv",
                              "fd", "inode", "history", "inodectx", "fdctx",
                              "quotad", NULL};
        char    *w = NULL;

        w = str_getunamb (arg, opwords);
        if (!w) {
                gf_log ("cli", GF_LOG_DEBUG, "Unknown statedump option  %s",
                        arg);
                return _gf_false;
        }
        *option = w;
        return _gf_true;
}

int
cli_cmd_volume_statedump_options_parse (const char **words, int wordcount,
                                        dict_t **options)
{
        int     ret = 0;
        int     i = 0;
        dict_t  *dict = NULL;
        int     option_cnt = 0;
        char    *option = NULL;
        char    option_str[100] = {0,};

        for (i = 3; i < wordcount; i++, option_cnt++) {
                if (!cli_cmd_validate_dumpoption (words[i], &option)) {
                        ret = -1;
                        goto out;
                }
                strncat (option_str, option, strlen (option));
                strncat (option_str, " ", 1);
        }
        if((strstr (option_str, "nfs")) && strstr (option_str, "quotad")) {
                ret = -1;
                goto out;
        }

        dict = dict_new ();
        if (!dict)
                goto out;

        ret = dict_set_dynstr (dict, "options", gf_strdup (option_str));
        if (ret)
                goto out;

        ret = dict_set_int32 (dict, "option_cnt", option_cnt);
        if (ret)
                goto out;

        *options = dict;
out:
        if (ret && dict)
                dict_destroy (dict);
        if (ret)
                gf_log ("cli", GF_LOG_ERROR, "Error parsing dumpoptions");
        return ret;
}

int
cli_cmd_volume_clrlks_opts_parse (const char **words, int wordcount,
                                  dict_t **options)
{
        int     ret = -1;
        int     i = 0;
        dict_t  *dict = NULL;
        char    *kind_opts[4] = {"blocked", "granted", "all", NULL};
        char    *types[4] = {"inode", "entry", "posix", NULL};
        char    *free_ptr = NULL;

        dict = dict_new ();
        if (!dict)
                goto out;

        if (strcmp (words[4], "kind"))
                goto out;

        for (i = 0; kind_opts[i]; i++) {
               if (!strcmp (words[5], kind_opts[i])) {
                       free_ptr = gf_strdup (words[5]);
                       ret = dict_set_dynstr (dict, "kind", free_ptr);
                       if (ret)
                               goto out;
                       free_ptr = NULL;
                       break;
               }
        }
        if (i == 3)
                goto out;

        ret = -1;
        for (i = 0; types[i]; i++) {
               if (!strcmp (words[6], types[i])) {
                       free_ptr = gf_strdup (words[6]);
                       ret = dict_set_dynstr (dict, "type", free_ptr);
                       if (ret)
                               goto out;
                       free_ptr = NULL;
                       break;
               }
        }
        if (i == 3)
                goto out;

        if (wordcount == 8) {
                free_ptr = gf_strdup (words[7]);
                ret = dict_set_dynstr (dict, "opts", free_ptr);
                if (ret)
                        goto out;
                free_ptr = NULL;
        }

        ret = 0;
        *options = dict;
out:
       if (ret) {
               GF_FREE (free_ptr);
               dict_unref (dict);
       }

       return ret;
}

static int
extract_hostname_path_from_token (const char *tmp_words, char **hostname,
                                  char **path)
{
        int ret = 0;
        char *delimiter = NULL;
        char *tmp_host = NULL;
        char *host_name = NULL;
        char *words = NULL;

        *hostname = NULL;
        *path = NULL;

        words = GF_CALLOC (1, strlen (tmp_words) + 1, gf_common_mt_char);
        if (!words){
                ret = -1;
                goto out;
        }

        strncpy (words, tmp_words, strlen (tmp_words) + 1);

        if (validate_brick_name (words)) {
                cli_err ("Wrong brick type: %s, use <HOSTNAME>:"
                        "<export-dir-abs-path>", words);
                ret = -1;
                goto out;
        } else {
                delimiter = strrchr (words, ':');
                ret = gf_canonicalize_path (delimiter + 1);
                if (ret) {
                        goto out;
                } else {
                        *path = GF_CALLOC (1, strlen (delimiter+1) +1,
                                           gf_common_mt_char);
                        if (!*path) {
                           ret = -1;
                                goto out;

                        }
                        strncpy (*path, delimiter +1,
                                 strlen(delimiter + 1) + 1);
                }
        }

        tmp_host = gf_strdup (words);
        if (!tmp_host) {
                gf_log ("cli", GF_LOG_ERROR, "Out of memory");
                ret = -1;
                goto out;
        }
        get_host_name (tmp_host, &host_name);
        if (!host_name) {
                ret = -1;
                gf_log("cli",GF_LOG_ERROR, "Unable to allocate "
                        "memory");
                goto out;
        }
        if (!(strcmp (host_name, "localhost") &&
            strcmp (host_name, "127.0.0.1") &&
            strncmp (host_name, "0.", 2))) {
                cli_err ("Please provide a valid hostname/ip other "
                         "than localhost, 127.0.0.1 or loopback "
                         "address (0.0.0.0 to 0.255.255.255).");
                ret = -1;
                goto out;
        }
        if (!valid_internet_address (host_name, _gf_false)) {
                cli_err ("internet address '%s' does not conform to "
                          "standards", host_name);
                ret = -1;
                goto out;
        }

        *hostname = GF_CALLOC (1, strlen (host_name) + 1,
                                       gf_common_mt_char);
        if (!*hostname) {
                ret = -1;
                goto out;
        }
        strncpy (*hostname, host_name, strlen (host_name) + 1);
        ret = 0;

out:
        GF_FREE (words);
        GF_FREE (tmp_host);
        return ret;
}

static int
set_hostname_path_in_dict (const char *token, dict_t *dict, int heal_op)
{
        char *hostname = NULL;
        char *path     = NULL;
        int   ret      = 0;

        ret = extract_hostname_path_from_token (token, &hostname, &path);
        if (ret)
                goto out;

        switch (heal_op) {
        case GF_SHD_OP_SBRAIN_HEAL_FROM_BRICK:
                ret = dict_set_dynstr (dict, "heal-source-hostname",
                                       hostname);
                if (ret)
                        goto out;
                ret = dict_set_dynstr (dict, "heal-source-brickpath",
                                       path);
                break;
        case GF_SHD_OP_STATISTICS_HEAL_COUNT_PER_REPLICA:
                ret = dict_set_dynstr (dict, "per-replica-cmd-hostname",
                                       hostname);
                if (ret)
                        goto out;
                ret = dict_set_dynstr (dict, "per-replica-cmd-path",
                                       path);
                break;
        default:
                ret = -1;
                break;
        }

out:
        return ret;
}

static int
heal_command_type_get (const char *command)
{
        int     i = 0;
        /* subcommands are set as NULL */
        char    *heal_cmds[GF_SHD_OP_HEAL_DISABLE + 1] = {
                [GF_SHD_OP_INVALID]                            = NULL,
                [GF_SHD_OP_HEAL_INDEX]                         = NULL,
                [GF_SHD_OP_HEAL_FULL]                          = "full",
                [GF_SHD_OP_INDEX_SUMMARY]                      = "info",
                [GF_SHD_OP_HEALED_FILES]                       = NULL,
                [GF_SHD_OP_HEAL_FAILED_FILES]                  = NULL,
                [GF_SHD_OP_SPLIT_BRAIN_FILES]                  = NULL,
                [GF_SHD_OP_STATISTICS]                         = "statistics",
                [GF_SHD_OP_STATISTICS_HEAL_COUNT]              = NULL,
                [GF_SHD_OP_STATISTICS_HEAL_COUNT_PER_REPLICA]  = NULL,
                [GF_SHD_OP_SBRAIN_HEAL_FROM_BIGGER_FILE]       = NULL,
                [GF_SHD_OP_SBRAIN_HEAL_FROM_BRICK]             = NULL,
                [GF_SHD_OP_HEAL_ENABLE]                        = "enable",
                [GF_SHD_OP_HEAL_DISABLE]                       = "disable",
        };

        for (i = 0; i <= GF_SHD_OP_HEAL_DISABLE; i++) {
                if (heal_cmds[i] && (strcmp (heal_cmds[i], command) == 0))
                        return i;
        }

        return GF_SHD_OP_INVALID;
}

int
cli_cmd_volume_heal_options_parse (const char **words, int wordcount,
                                   dict_t **options)
{
        int     ret = 0;
        dict_t  *dict = NULL;
        char    *hostname = NULL;
        char    *path = NULL;
        gf_xl_afr_op_t op = GF_SHD_OP_INVALID;

        dict = dict_new ();
        if (!dict)
                goto out;

        ret = dict_set_str (dict, "volname", (char *) words[2]);
        if (ret) {
                gf_log (THIS->name, GF_LOG_ERROR, "failed to set volname");
                goto out;
        }

        if (wordcount == 3) {
                ret = dict_set_int32 (dict, "heal-op", GF_SHD_OP_HEAL_INDEX);
                goto done;
        }

        if (wordcount == 4) {
                op = heal_command_type_get (words[3]);
                if (op == GF_SHD_OP_INVALID) {
                        ret = -1;
                        goto out;
                }

                ret = dict_set_int32 (dict, "heal-op", op);
                goto done;
        }

        if (wordcount == 5) {
                if (strcmp (words[3], "info") &&
                    strcmp (words[3], "statistics")) {
                        ret = -1;
                        goto out;
                }

                if (!strcmp (words[3], "info")) {
                        if (!strcmp (words[4], "healed")) {
                                ret = dict_set_int32 (dict, "heal-op",
                                                      GF_SHD_OP_HEALED_FILES);
                                goto done;
                        }
                        if (!strcmp (words[4], "heal-failed")) {
                                ret = dict_set_int32 (dict, "heal-op",
                                                   GF_SHD_OP_HEAL_FAILED_FILES);
                                goto done;
                        }
                        if (!strcmp (words[4], "split-brain")) {
                                ret = dict_set_int32 (dict, "heal-op",
                                                   GF_SHD_OP_SPLIT_BRAIN_FILES);
                                goto done;
                        }
                }

                if (!strcmp (words[3], "statistics")) {
                        if (!strcmp (words[4], "heal-count")) {
                                ret = dict_set_int32 (dict, "heal-op",
                                               GF_SHD_OP_STATISTICS_HEAL_COUNT);
                                goto done;
                        }
                }
                ret = -1;
                goto out;
        }
        if (wordcount == 6) {
                if (strcmp (words[3], "split-brain")) {
                        ret = -1;
                        goto out;
                }
                if (!strcmp (words[4], "bigger-file")) {
                        ret = dict_set_int32 (dict, "heal-op",
                                        GF_SHD_OP_SBRAIN_HEAL_FROM_BIGGER_FILE);
                        if (ret)
                                goto out;
                        ret = dict_set_str (dict, "file", (char *)words[5]);
                        if (ret)
                                goto out;
                        goto done;
                }
                if (!strcmp (words[4], "source-brick")) {
                        ret = dict_set_int32 (dict, "heal-op",
                                              GF_SHD_OP_SBRAIN_HEAL_FROM_BRICK);
                        if (ret)
                                goto out;
                        ret = set_hostname_path_in_dict (words[5], dict,
                                              GF_SHD_OP_SBRAIN_HEAL_FROM_BRICK);
                        if (ret)
                                goto out;
                        goto done;
                }
                ret = -1;
                goto out;
        }
        if (wordcount == 7) {
                if (!strcmp (words[3], "statistics")
                    && !strcmp (words[4], "heal-count")
                    && !strcmp (words[5], "replica")) {

                        ret = dict_set_int32 (dict, "heal-op",
                                   GF_SHD_OP_STATISTICS_HEAL_COUNT_PER_REPLICA);
                        if (ret)
                                goto out;
                        ret = set_hostname_path_in_dict (words[6], dict,
                                   GF_SHD_OP_STATISTICS_HEAL_COUNT_PER_REPLICA);
                        if (ret)
                                goto out;
                        goto done;

                }
                if (!strcmp (words[3], "split-brain") &&
                    !strcmp (words[4], "source-brick")) {
                        ret = dict_set_int32 (dict, "heal-op",
                                              GF_SHD_OP_SBRAIN_HEAL_FROM_BRICK);
                        ret = set_hostname_path_in_dict (words[5], dict,
                                              GF_SHD_OP_SBRAIN_HEAL_FROM_BRICK);
                        if (ret)
                                goto out;
                        ret = dict_set_str (dict, "file",
                                            (char *) words[6]);
                        if (ret)
                                goto out;
                        goto done;
                }
        }
        ret = -1;
        goto out;
done:
        *options = dict;
out:
        if (ret && dict) {
                dict_unref (dict);
                *options = NULL;
        }

        return ret;
}

int
cli_cmd_volume_defrag_parse (const char **words, int wordcount,
                             dict_t **options)
{
        dict_t                 *dict = NULL;
        int                      ret = -1;
        char                *option  = NULL;
        char                *volname = NULL;
        char                *command = NULL;
        gf_cli_defrag_type       cmd = 0;

        GF_ASSERT (words);
        GF_ASSERT (options);

        dict = dict_new ();
        if (!dict)
                goto out;

        if (!((wordcount == 4) || (wordcount == 5)))
                goto out;

        if (wordcount == 4) {
                if (strcmp (words[3], "start") && strcmp (words[3], "stop") &&
                    strcmp (words[3], "status"))
                            goto out;
        } else if ((strcmp (words[3], "tier") == 0) &&
               (strcmp (words[4], "start") == 0)) {
                volname = (char *) words[2];
                cmd = GF_DEFRAG_CMD_START_TIER;
                goto done;
        } else if ((strcmp (words[3], "tier") == 0) &&
               (strcmp (words[4], "status") == 0)) {
                volname = (char *) words[2];
                cmd = GF_DEFRAG_CMD_STATUS_TIER;
                goto done;
        } else {
                if (strcmp (words[3], "fix-layout") &&
                    strcmp (words[3], "start"))
                        goto out;
        }

        volname = (char *) words[2];

        if (wordcount == 4) {
                command = (char *) words[3];
        }
        if (wordcount == 5) {
               if ((strcmp (words[3], "fix-layout") ||
                    strcmp (words[4], "start")) &&
                    (strcmp (words[3], "start") ||
                    strcmp (words[4], "force"))) {
                        ret = -1;
                        goto out;
                }
                command = (char *) words[3];
                option = (char *) words[4];
        }

        if (strcmp (command, "start") == 0) {
                cmd = GF_DEFRAG_CMD_START;
                if (option && strcmp (option, "force") == 0) {
                                cmd = GF_DEFRAG_CMD_START_FORCE;
                        }
                goto done;
        }

        if (strcmp (command, "fix-layout") == 0) {
                cmd = GF_DEFRAG_CMD_START_LAYOUT_FIX;
                goto done;
        }
        if (strcmp (command, "stop") == 0) {
                cmd = GF_DEFRAG_CMD_STOP;
                goto done;
        }
        if (strcmp (command, "status") == 0) {
                cmd = GF_DEFRAG_CMD_STATUS;
        }

done:
        ret = dict_set_str (dict, "volname", volname);

        if (ret) {
                gf_log (THIS->name, GF_LOG_ERROR, "failed to set dict");
                goto out;
        }

        ret = dict_set_int32 (dict, "rebalance-command", (int32_t) cmd);

        if (ret) {
                gf_log (THIS->name, GF_LOG_ERROR, "failed to set dict");
                goto out;
        }

        *options = dict;

out:
        if (ret && dict)
                dict_destroy (dict);

        return ret;
}

int32_t
cli_snap_create_desc_parse (dict_t *dict, const char **words,
                            size_t wordcount, int32_t desc_opt_loc)
{
        int32_t        ret      = -1;
        char          *desc     = NULL;
        int32_t        desc_len = 0;

        desc = GF_CALLOC (MAX_SNAP_DESCRIPTION_LEN + 1, sizeof(char),
                          gf_common_mt_char);
        if (!desc) {
                ret = -1;
                goto out;
        }


        if (strlen (words[desc_opt_loc]) >= MAX_SNAP_DESCRIPTION_LEN) {
                cli_out ("snapshot create: description truncated: "
                         "Description provided is longer than 1024 characters");
                desc_len = MAX_SNAP_DESCRIPTION_LEN;
        } else {
                desc_len = strlen (words[desc_opt_loc]);
        }

        strncpy (desc, words[desc_opt_loc], desc_len);
        desc[desc_len] = '\0';
        /* Calculating the size of the description as given by the user */

        ret = dict_set_dynstr (dict, "description", desc);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Unable to save snap "
                        "description");
                goto out;
        }

        ret = 0;
out:
        if (ret && desc)
                GF_FREE (desc);

        return ret;
}

/* Function to check whether the Volume name is repeated */
int
cli_check_if_volname_repeated (const char **words, unsigned int start_index,
                           uint64_t cur_index) {
        uint64_t        i       =       -1;
        int             ret     =        0;

        GF_ASSERT (words);

        for (i = start_index ; i < cur_index ; i++) {
                if (strcmp (words[i], words[cur_index]) == 0) {
                        ret = -1;
                        goto out;
                }
        }
out:
        return ret;
}

/* snapshot clone <clonename> <snapname>
 * @arg-0, dict     : Request Dictionary to be sent to server side.
 * @arg-1, words    : Contains individual words of CLI command.
 * @arg-2, wordcount: Contains number of words present in the CLI command.
 *
 * return value : -1 on failure
 *                 0 on success
 */
int
cli_snap_clone_parse (dict_t *dict, const char **words, int wordcount) {
        uint64_t        i               =       0;
        int             ret             =       -1;
        char            key[PATH_MAX]   =       "";
        char            *clonename      =       NULL;
        unsigned int    cmdi            =       2;
        int             flags           =       0;
        /* cmdi is command index, here cmdi is "2" (gluster snapshot clone)*/

        GF_ASSERT (words);
        GF_ASSERT (dict);

        if (wordcount == cmdi + 1) {
                cli_err ("Invalid Syntax.");
                gf_log ("cli", GF_LOG_ERROR,
                        "Invalid number of  words for snap clone command");
                goto out;
        }

        if (strlen(words[cmdi]) >= GLUSTERD_MAX_SNAP_NAME) {
                cli_err ("snapshot clone: failed: clonename cannot exceed "
                         "255 characters.");
                gf_log ("cli", GF_LOG_ERROR, "Clone name too long");

                goto out;
        }

        clonename = (char *) words[cmdi];
        for (i = 0 ; i < strlen (clonename); i++) {
                /* Following volume name convention */
                if (!isalnum (clonename[i]) && (clonename[i] != '_'
                                           && (clonename[i] != '-'))) {
                        /* TODO : Is this message enough?? */
                        cli_err ("Clonename can contain only alphanumeric, "
                                 "\"-\" and \"_\" characters");
                        goto out;
                }
        }

        ret = dict_set_int32 (dict, "volcount", 1);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Could not save volcount");
                goto out;
        }

        ret = dict_set_str (dict, "clonename", (char *)words[cmdi]);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Could not save clone "
                        "name(%s)", (char *)words[cmdi]);
                goto out;
        }

        /* Filling snap name in the dictionary */
        ret = dict_set_str (dict, "snapname", (char *)words[cmdi+1]);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Could not "
                        "save snap name(%s)", (char *)words[cmdi+1]);
                goto out;
        }


        ret = 0;

out:
        return ret;
}


/* snapshot create <snapname> <vol-name(s)> [description <description>]
 *                                           [force]
 * @arg-0, dict     : Request Dictionary to be sent to server side.
 * @arg-1, words    : Contains individual words of CLI command.
 * @arg-2, wordcount: Contains number of words present in the CLI command.
 *
 * return value : -1 on failure
 *                 0 on success
 */
int
cli_snap_create_parse (dict_t *dict, const char **words, int wordcount) {
        uint64_t        i               =       0;
        int             ret             =       -1;
        uint64_t        volcount        =       0;
        char            key[PATH_MAX]   =       "";
        char            *snapname       =       NULL;
        unsigned int    cmdi            =       2;
        int             flags           =       0;
        /* cmdi is command index, here cmdi is "2" (gluster snapshot create)*/

        GF_ASSERT (words);
        GF_ASSERT (dict);

        if (wordcount <= cmdi + 1) {
                cli_err ("Invalid Syntax.");
                gf_log ("cli", GF_LOG_ERROR,
                        "Too less words for snap create command");
                goto out;
        }

        if (strlen(words[cmdi]) >= GLUSTERD_MAX_SNAP_NAME) {
                cli_err ("snapshot create: failed: snapname cannot exceed "
                         "255 characters.");
                gf_log ("cli", GF_LOG_ERROR, "Snapname too long");

                goto out;
        }

        snapname = (char *) words[cmdi];
        for (i = 0 ; i < strlen (snapname); i++) {
                /* Following volume name convention */
                if (!isalnum (snapname[i]) && (snapname[i] != '_'
                                           && (snapname[i] != '-'))) {
                        /* TODO : Is this message enough?? */
                        cli_err ("Snapname can contain only alphanumeric, "
                                 "\"-\" and \"_\" characters");
                        goto out;
                }
        }

        ret = dict_set_str (dict, "snapname", (char *)words[cmdi]);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Could not save snap "
                        "name(%s)", (char *)words[cmdi]);
                goto out;
        }

        /* Filling volume name in the dictionary */
        for (i = cmdi + 1 ; i < wordcount
                            && (strcmp (words[i], "description")) != 0
                            && (strcmp (words[i], "force") != 0)
                            && (strcmp (words[i], "no-timestamp") != 0);
                            i++) {
                volcount++;
                /* volume index starts from 1 */
                ret = snprintf (key, sizeof (key), "volname%"PRIu64, volcount);
                if (ret < 0) {
                        goto out;
                }

                ret = dict_set_str (dict, key, (char *)words[i]);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Could not "
                                "save volume name(%s)", (char *)words[i]);
                        goto out;
                }

                if (i >= cmdi + 2) {
                        ret = -1;
                        cli_err("Creating multiple volume snapshot is not "
                                "supported as of now");
                        goto out;
                }
                /* TODO : remove this above condition check once
                 * multiple volume snapshot is supported */
        }

        if (volcount == 0) {
                ret = -1;
                cli_err ("Please provide the volume name");
                gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                goto out;
        }

        ret = dict_set_int32 (dict, "volcount", volcount);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Could not save volcount");
                goto out;
        }

        /* Verify how we got out of "for" loop,
         * if it is by reaching wordcount limit then goto "out",
         * because we need not parse for "description","force" and
         * "no-timestamp" after this.
         */
        if (i == wordcount) {
                goto out;
        }

        if (strcmp (words[i], "no-timestamp") == 0) {
                ret = dict_set_str (dict, "no-timestamp", "true");
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Could not save "
                                "time-stamp option");
                }
                if (i == (wordcount-1))
                        goto out;
                i++;
        }

        if ((strcmp (words[i], "description")) == 0) {
                ++i;
                if (i > (wordcount - 1)) {
                        ret = -1;
                        cli_err ("Please provide a description");
                        gf_log ("cli", GF_LOG_ERROR,
                                "Description not provided");
                        goto out;
                }

                ret = cli_snap_create_desc_parse(dict, words, wordcount, i);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Could not save snap "
                                "description");
                        goto out;
                }

                if (i == (wordcount - 1))
                        goto out;
                i++;
                /* point the index to next word.
                 * As description might be follwed by force option.
                 * Before that, check if wordcount limit is reached
                 */
        }

        if (strcmp (words[i], "force") == 0) {
                flags = GF_CLI_FLAG_OP_FORCE;

        } else {
                ret = -1;
                cli_err ("Invalid Syntax.");
                gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                goto out;
        }

        /* Check if the command has anything after "force" keyword */
        if (++i < wordcount) {
                ret = -1;
                gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                goto out;
        }

        ret = 0;

out:
        if(ret == 0) {
                /*Adding force flag in either of the case i.e force set
                 * or unset*/
                ret = dict_set_int32 (dict, "flags", flags);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Could not save "
                                "snap force option");
                }
        }
        return ret;
}

/* snapshot list [volname]
 * @arg-0, dict     : Request Dictionary to be sent to server side.
 * @arg-1, words    : Contains individual words of CLI command.
 * @arg-2, wordcount: Contains number of words present in the CLI command.
 *
 * return value : -1 on failure
 *                 0 on success
 */
int
cli_snap_list_parse (dict_t *dict, const char **words, int wordcount) {
        int             ret     =       -1;

        GF_ASSERT (words);
        GF_ASSERT (dict);

        if (wordcount < 2 || wordcount > 3) {
                gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                goto out;
        }

        if (wordcount == 2) {
                ret = 0;
                goto out;
        }

        ret = dict_set_str (dict, "volname", (char *)words[2]);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR,
                        "Failed to save volname in dictionary");
                goto out;
        }
out:
        return ret;
}

/* snapshot info [(snapname |  volume <volname>)]
 * @arg-0, dict     : Request Dictionary to be sent to server side.
 * @arg-1, words    : Contains individual words of CLI command.
 * @arg-2, wordcount: Contains number of words present in the CLI command.
 *
 * return value : -1 on failure
 *                 0 on success
 */
int
cli_snap_info_parse (dict_t *dict, const char **words, int wordcount)
{

        int             ret             =       -1;
        int32_t         cmd             =       GF_SNAP_INFO_TYPE_ALL;
        unsigned int    cmdi            =        2;
        /* cmdi is command index, here cmdi is "2" (gluster snapshot info)*/

        GF_ASSERT (words);
        GF_ASSERT (dict);

        if (wordcount > 4 || wordcount < cmdi) {
                gf_log ("cli", GF_LOG_ERROR, "Invalid syntax");
                goto out;
        }

        if (wordcount == cmdi) {
                ret = 0;
                goto out;
        }

        /* If 3rd word is not "volume", then it must
         * be snapname.
         */
        if (strcmp (words[cmdi], "volume") != 0) {
                ret = dict_set_str (dict, "snapname",
                                   (char *)words[cmdi]);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Unable to save "
                                "snapname %s", words[cmdi]);
                        goto out;
                }

                /* Once snap name is parsed, if we encounter any other
                 * word then fail it. Invalid Syntax.
                 * example : snapshot info <snapname> word
                 */
                if ((cmdi + 1) != wordcount) {
                        ret = -1;
                        gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                        goto out;
                }

                cmd = GF_SNAP_INFO_TYPE_SNAP;
                ret = 0;
                goto out;
                /* No need to continue the parsing once we
                 * get the snapname
                 */
        }

        /* If 3rd word is "volume", then check if next word
         * is present. As, "snapshot info volume" is an
         * invalid command.
         */
        if ((cmdi + 1) == wordcount) {
                ret = -1;
                gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                goto out;
        }

        ret = dict_set_str (dict, "volname", (char *)words[wordcount - 1]);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Could not save "
                        "volume name %s", words[wordcount - 1]);
                goto out;
        }
        cmd = GF_SNAP_INFO_TYPE_VOL;
out:
        if (ret == 0) {
                ret = dict_set_int32 (dict, "cmd", cmd);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Could not save "
                                "type of snapshot info");
                }
        }
        return ret;
}



/* snapshot restore <snapname>
 * @arg-0, dict     : Request Dictionary to be sent to server side.
 * @arg-1, words    : Contains individual words of CLI command.
 * @arg-2, wordcount: Contains number of words present in the CLI command.
 *
 * return value : -1 on failure
 *                 0 on success
 */
int
cli_snap_restore_parse (dict_t *dict, const char **words, int wordcount,
                        struct cli_state *state)
{

        int             ret             =       -1;
        const char      *question       =       NULL;
        gf_answer_t     answer          =       GF_ANSWER_NO;

        GF_ASSERT (words);
        GF_ASSERT (dict);

        if (wordcount != 3) {
                gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                goto out;
        }

        ret = dict_set_str (dict, "snapname", (char *)words[2]);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Unable to save snap-name %s",
                        words[2]);
                goto out;
        }

        question = "Restore operation will replace the "
                   "original volume with the snapshotted volume. "
                   "Do you still want to continue?";

        answer = cli_cmd_get_confirmation (state, question);
        if (GF_ANSWER_NO == answer) {
                ret = 1;
                gf_log ("cli", GF_LOG_ERROR, "User cancelled a snapshot "
                        "restore operation for snap %s", (char *)words[2]);
                goto out;
        }
out:
        return ret;
}

/* snapshot activate <snapname> [force]
 * @arg-0, dict     : Request Dictionary to be sent to server side.
 * @arg-1, words    : Contains individual words of CLI command.
 * @arg-2, wordcount: Contains number of words present in the CLI command.
 *
 * return value : -1 on failure
 *                 0 on success
 */
int
cli_snap_activate_parse (dict_t *dict, const char **words, int wordcount)
{

        int ret = -1;
        int flags = 0;

        GF_ASSERT (words);
        GF_ASSERT (dict);

        if ((wordcount < 3) || (wordcount > 4)) {
                gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                goto out;
        }

        ret = dict_set_str (dict, "snapname", (char *)words[2]);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Unable to save snap-name %s",
                        words[2]);
                goto out;
        }

        if (wordcount == 4) {
                if (!strcmp("force", (char *)words[3])) {
                        flags = GF_CLI_FLAG_OP_FORCE;
                } else {
                        gf_log ("cli", GF_LOG_ERROR, "Invalid option");
                        ret = -1;
                        goto out;
                }
        }
        ret = dict_set_int32 (dict, "flags", flags);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Unable to save force option");
                goto out;
        }
out:
        return ret;
}

/* snapshot deactivate <snapname>
 * @arg-0, dict     : Request Dictionary to be sent to server side.
 * @arg-1, words    : Contains individual words of CLI command.
 * @arg-2, wordcount: Contains number of words present in the CLI command.
 *
 * return value : -1 on failure
 *                 0 on success
 *                 1 if user cancelled the request
 */
int
cli_snap_deactivate_parse (dict_t *dict, const char **words, int wordcount,
                        struct cli_state *state)
{

        int             ret             = -1;
        gf_answer_t     answer          = GF_ANSWER_NO;
        const char     *question        = "Deactivating snap will make its "
                                          "data inaccessible. Do you want to "
                                          "continue?";


        GF_ASSERT (words);
        GF_ASSERT (dict);

        if ((wordcount != 3)) {
                gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                goto out;
        }

        ret = dict_set_str (dict, "snapname", (char *)words[2]);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Unable to save snap-name %s",
                        words[2]);
                goto out;
        }

        answer = cli_cmd_get_confirmation (state, question);
        if (GF_ANSWER_NO == answer) {
                ret = 1;
                gf_log ("cli", GF_LOG_DEBUG, "User cancelled "
                        "snapshot deactivate operation");
                goto out;
        }

out:
        return ret;
}

/* snapshot delete (all | snapname | volume <volname>)
 * @arg-0, dict     : Request Dictionary to be sent to server side.
 * @arg-1, words    : Contains individual words of CLI command.
 * @arg-2, wordcount: Contains number of words present in the CLI command.
 *
 * return value : -1 on failure
 *                 0 on success
 *                 1 if user cancel the operation
 */
int
cli_snap_delete_parse (dict_t *dict, const char **words, int wordcount,
                       struct cli_state *state) {

        int             ret             =       -1;
        const char      *question       =       NULL;
        int32_t         cmd             =       -1;
        unsigned int    cmdi            =       2;
        gf_answer_t     answer          =       GF_ANSWER_NO;

        GF_ASSERT (words);
        GF_ASSERT (dict);

        if (wordcount > 4 || wordcount <= cmdi) {
                gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                goto out;
        }

        question = "Deleting snap will erase all the information about "
                   "the snap. Do you still want to continue?";

        if (strcmp (words [cmdi], "all") == 0) {
                ret = 0;
                cmd = GF_SNAP_DELETE_TYPE_ALL;
        } else if (strcmp (words [cmdi], "volume") == 0) {
                if (++cmdi == wordcount) {
                        ret = -1;
                        gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                        goto out;
                }

                ret = dict_set_str (dict, "volname",
                                    (char *)words[cmdi]);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Could not save "
                                "volume name %s", words[wordcount - 1]);
                        goto out;
                }
                cmd = GF_SNAP_DELETE_TYPE_VOL;
        } else {
                ret = dict_set_str (dict, "snapname", (char *)words[cmdi]);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Unable to save "
                                "snapname %s", words[2]);
                        goto out;
                }
                cmd = GF_SNAP_DELETE_TYPE_SNAP;
        }

        if ((cmdi + 1) != wordcount) {
                ret = -1;
                gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                goto out;
        }

        if (cmd == GF_SNAP_DELETE_TYPE_SNAP) {
                answer = cli_cmd_get_confirmation (state, question);
                if (GF_ANSWER_NO == answer) {
                        ret = 1;
                        gf_log ("cli", GF_LOG_DEBUG, "User cancelled "
                                "snapshot delete operation for snap %s",
                                (char *)words[2]);
                        goto out;
                }
        }

        ret = dict_set_int32 (dict, "delete-cmd", cmd);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Could not save "
                        "type of snapshot delete");
        }
out:
        return ret;
}

/* snapshot status [(snapname | volume <volname>)]
 * @arg-0, dict     : Request Dictionary to be sent to server side.
 * @arg-1, words    : Contains individual words of CLI command.
 * @arg-2, wordcount: Contains number of words present in the CLI command.
 *
 * return value : -1 on failure
 *                 0 on success
 */
int
cli_snap_status_parse (dict_t *dict, const char **words, int wordcount)
{

        int             ret  =        -1;
        int32_t         cmd  =       GF_SNAP_STATUS_TYPE_ALL;
        unsigned int    cmdi =        2;
        /* cmdi is command index, here cmdi is "2" (gluster snapshot status)*/

        GF_ASSERT (words);
        GF_ASSERT (dict);

        if (wordcount > 4 || wordcount < cmdi) {
                gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                goto out;
        }

        if (wordcount == cmdi) {
                ret = 0;
                goto out;
        }

        /* if 3rd word is not "volume", then it must be "snapname"
        */
        if (strcmp (words[cmdi], "volume") != 0) {
                ret = dict_set_str (dict, "snapname",
                                   (char *)words[cmdi]);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Count not save "
                                "snap name %s", words[cmdi]);
                        goto out;
                }

                if ((cmdi + 1) != wordcount) {
                        ret = -1;
                        gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                        goto out;
                }

                ret = 0;
                cmd = GF_SNAP_STATUS_TYPE_SNAP;
                goto out;
        }

        /* If 3rd word is "volume", then check if next word is present.
         * As, "snapshot info volume" is an invalid command
         */
        if ((cmdi + 1) == wordcount) {
                ret = -1;
                gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                goto out;
        }

        ret = dict_set_str (dict, "volname", (char *)words [wordcount - 1]);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Count not save "
                        "volume name %s", words[wordcount - 1]);
                goto out;
        }
        cmd = GF_SNAP_STATUS_TYPE_VOL;

out:
        if (ret == 0) {
                ret = dict_set_int32 (dict, "status-cmd", cmd);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Could not save cmd "
                                "of snapshot status");
                }
        }

        return ret;
}


/* return value:
 *      -1  in case of failure.
 *       0  in case of success.
 */
int32_t
cli_snap_config_limit_parse (const char **words, dict_t *dict,
                             unsigned int wordcount, unsigned int index,
                             char *key)
{
        int             ret             = -1;
        int             limit           = 0;
        char            *end_ptr        = NULL;

        GF_ASSERT (words);
        GF_ASSERT (dict);
        GF_ASSERT (key);

        if (index >= wordcount) {
                ret = -1;
                cli_err ("Please provide a value for %s.", key);
                gf_log ("cli", GF_LOG_ERROR, "Value not provided for %s", key);
                goto out;
        }

        limit = strtol (words[index], &end_ptr, 10);

        if (limit <= 0 || strcmp (end_ptr, "") != 0) {
                ret = -1;
                cli_err("Please enter an integer value "
                        "greater than zero for %s", key);
                goto out;
        }

        ret = dict_set_int32 (dict, key, limit);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Could not set "
                        "%s in dictionary", key);
                goto out;
        }

        ret = dict_set_dynstr_with_alloc (dict, "globalname", "All");
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Could not set global key");
                goto out;
        }
        ret = dict_set_int32 (dict, "hold_global_locks", _gf_true);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Could not set global locks");
                goto out;
        }

out:
        return ret;
}

/* function cli_snap_config_parse
 * Config Syntax : gluster snapshot config [volname]
 *                                         [snap-max-hard-limit <count>]
 *                                         [snap-max-soft-limit <count>]
 *
   return value: <0  on failure
                  1  if user cancels the operation, or limit value is out of
                                                                      range
                  0  on success

  NOTE : snap-max-soft-limit can only be set for system.
*/
int32_t
cli_snap_config_parse (const char **words, int wordcount, dict_t *dict,
                       struct cli_state *state)
{
        int                            ret                 = -1;
        gf_answer_t                    answer              = GF_ANSWER_NO;
        gf_boolean_t                   vol_presence        = _gf_false;
        struct snap_config_opt_vals_  *conf_vals           = NULL;
        int8_t                         hard_limit          = 0;
        int8_t                         soft_limit          = 0;
        int8_t                         config_type         = -1;
        const char                    *question            = NULL;
        unsigned int                   cmdi                = 2;
        /* cmdi is command index, here cmdi is "2" (gluster snapshot config)*/

        GF_ASSERT (words);
        GF_ASSERT (dict);
        GF_ASSERT (state);

        if ((wordcount < 2) || (wordcount > 7)) {
                gf_log ("cli", GF_LOG_ERROR,
                        "Invalid wordcount(%d)", wordcount);
                goto out;
        }

        if (wordcount == 2) {
                config_type = GF_SNAP_CONFIG_DISPLAY;
                ret = 0;
                goto set;
        }

        /* auto-delete cannot be a volume name */
        /* Check whether the 3rd word is volname */
        if (strcmp (words[cmdi], "snap-max-hard-limit") != 0
             && strcmp (words[cmdi], "snap-max-soft-limit") != 0
             && strcmp (words[cmdi], "auto-delete") != 0
             && strcmp (words[cmdi], "activate-on-create") != 0) {
                ret = dict_set_str (dict, "volname", (char *)words[cmdi]);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Failed to set volname");
                        goto out;
                }
                cmdi++;
                vol_presence = _gf_true;

                if (cmdi == wordcount) {
                        config_type = GF_SNAP_CONFIG_DISPLAY;
                        ret = 0;
                        goto set;
                }
        }

        config_type = GF_SNAP_CONFIG_TYPE_SET;

        if (strcmp (words[cmdi], "snap-max-hard-limit") == 0) {
                ret = cli_snap_config_limit_parse (words, dict, wordcount,
                                                ++cmdi, "snap-max-hard-limit");
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Failed to parse snap "
                                "config hard limit");
                        goto out;
                }
                hard_limit = 1;

                if (++cmdi == wordcount) {
                        ret = 0;
                        goto set;
                }
        }

        if (strcmp (words[cmdi], "snap-max-soft-limit") == 0) {
                if (vol_presence == 1) {
                        ret = -1;
                        cli_err ("Soft limit cannot be set to individual "
                                  "volumes.");
                        gf_log ("cli", GF_LOG_ERROR, "Soft limit cannot be "
                                "set to volumes");
                        goto out;
                }

                ret = cli_snap_config_limit_parse (words, dict, wordcount,
                                                ++cmdi, "snap-max-soft-limit");
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Failed to parse snap "
                                "config soft limit");
                        goto out;
                }

                if (++cmdi != wordcount) {
                        ret = -1;
                        gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                        goto out;
                }
                soft_limit = 1;
        }

        if (hard_limit || soft_limit)
                goto set;

        if (strcmp(words[cmdi], "auto-delete") == 0) {
                if (vol_presence == 1) {
                        ret = -1;
                        cli_err ("As of now, auto-delete option cannot be set "
                                "to volumes");
                        gf_log ("cli", GF_LOG_ERROR, "auto-delete option "
                                "cannot be set to volumes");
                        goto out;
                }

                if (++cmdi >= wordcount) {
                        ret = -1;
                        gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                        goto out;
                }

                ret = dict_set_str (dict, "auto-delete", (char *)words[cmdi]);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Failed to set "
                                "value of auto-delete in request "
                                "dictionary");
                        goto out;
                }

                if (++cmdi != wordcount) {
                        ret = -1;
                        gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                        goto out;
                }
        } else if (strcmp(words[cmdi], "activate-on-create") == 0) {
                if (vol_presence == 1) {
                        ret = -1;
                        cli_err ("As of now, activate-on-create option "
                                 "cannot be set to volumes");
                        gf_log ("cli", GF_LOG_ERROR, "activate-on-create "
                                "option cannot be set to volumes");
                        goto out;
                }

                if (++cmdi >= wordcount) {
                        ret = -1;
                        gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                        goto out;
                }

                ret = dict_set_str (dict, "snap-activate-on-create",
                                    (char *)words[cmdi]);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Failed to set value "
                                "of activate-on-create in request dictionary");
                        goto out;
                }

                if (++cmdi != wordcount) {
                        ret = -1;
                        gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                        goto out;
                }
        } else {
                ret = -1;
                gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                goto out;
        }

        ret = 0; /* Success */

set:
        ret = dict_set_int32 (dict, "config-command", config_type);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Unable to set "
                        "config-command");
                goto out;
        }

        if (config_type == GF_SNAP_CONFIG_TYPE_SET &&
           (hard_limit || soft_limit)) {
                conf_vals = snap_confopt_vals;
                if (hard_limit && soft_limit) {
                        question = conf_vals[GF_SNAP_CONFIG_SET_BOTH].question;
                } else if (soft_limit) {
                        question = conf_vals[GF_SNAP_CONFIG_SET_SOFT].question;
                } else if (hard_limit) {
                        question = conf_vals[GF_SNAP_CONFIG_SET_HARD].question;
                }

                answer = cli_cmd_get_confirmation (state, question);
                if (GF_ANSWER_NO == answer) {
                        ret = 1;
                        gf_log ("cli", GF_LOG_DEBUG, "User cancelled "
                        "snapshot config operation");
                }
        }

out:
        return ret;
}

int
validate_op_name (const char *op, const char *opname, char **opwords) {
        int     ret     =       -1;
        int     i       =       0;

        GF_ASSERT (opname);
        GF_ASSERT (opwords);

        for (i = 0 ; opwords[i] != NULL; i++) {
                if (strcmp (opwords[i], opname) == 0) {
                        cli_out ("\"%s\" cannot be a %s", opname, op);
                        goto out;
                }
        }
        ret = 0;
out:
        return ret;
}

int32_t
cli_cmd_snapshot_parse (const char **words, int wordcount, dict_t **options,
                        struct cli_state *state)
{
        int32_t            ret        = -1;
        dict_t             *dict      = NULL;
        gf1_cli_snapshot   type       = GF_SNAP_OPTION_TYPE_NONE;
        char               *w         = NULL;
        char               *opwords[] = {"create", "delete", "restore",
                                        "activate", "deactivate", "list",
                                        "status", "config", "info", "clone",
                                        NULL};
        char               *invalid_snapnames[] = {"description", "force",
                                                  "volume", "all", NULL};
        char               *invalid_volnames[]  = {"volume", "type",
                                                   "subvolumes", "option",
                                                   "end-volume", "all",
                                                   "volume_not_in_ring",
                                                   "description", "force",
                                                   "snap-max-hard-limit",
                                                   "snap-max-soft-limit",
                                                   "auto-delete",
                                                   "activate-on-create", NULL};

        GF_ASSERT (words);
        GF_ASSERT (options);
        GF_ASSERT (state);

        dict = dict_new ();
        if (!dict)
                goto out;

        /* Lowest wordcount possible */
        if (wordcount < 2) {
                gf_log ("", GF_LOG_ERROR,
                        "Invalid command: Not enough arguments");
                goto out;
        }

        w = str_getunamb (words[1], opwords);
        if (!w) {
                /* Checks if the operation is a valid operation */
                gf_log ("", GF_LOG_ERROR, "Opword Mismatch");
                goto out;
        }

        if (!strcmp (w, "create")) {
                type = GF_SNAP_OPTION_TYPE_CREATE;
        } else if (!strcmp (w, "list")) {
                type = GF_SNAP_OPTION_TYPE_LIST;
        } else if (!strcmp (w, "info")) {
                type = GF_SNAP_OPTION_TYPE_INFO;
        } else if (!strcmp (w, "delete")) {
                type = GF_SNAP_OPTION_TYPE_DELETE;
        } else if (!strcmp (w, "config")) {
                type = GF_SNAP_OPTION_TYPE_CONFIG;
        } else if (!strcmp (w, "restore")) {
                type = GF_SNAP_OPTION_TYPE_RESTORE;
        } else if (!strcmp (w, "status")) {
                type = GF_SNAP_OPTION_TYPE_STATUS;
        } else if (!strcmp (w, "activate")) {
                type = GF_SNAP_OPTION_TYPE_ACTIVATE;
        } else if (!strcmp (w, "deactivate")) {
                type = GF_SNAP_OPTION_TYPE_DEACTIVATE;
        } else if (!strcmp(w, "clone")) {
                type = GF_SNAP_OPTION_TYPE_CLONE;
        }

        if (type != GF_SNAP_OPTION_TYPE_CONFIG &&
            type != GF_SNAP_OPTION_TYPE_STATUS) {
                ret = dict_set_int32 (dict, "hold_snap_locks", _gf_true);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR,
                                "Unable to set hold-snap-locks value "
                                "as _gf_true");
                        goto out;
                }
        }

        /* Following commands does not require volume locks  */
        if (type == GF_SNAP_OPTION_TYPE_STATUS ||
            type == GF_SNAP_OPTION_TYPE_ACTIVATE ||
            type == GF_SNAP_OPTION_TYPE_DEACTIVATE) {
                ret = dict_set_int32 (dict, "hold_vol_locks", _gf_false);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Setting volume lock "
                                "flag failed");
                        goto out;
                }
        }

        /* Check which op is intended */
        switch (type) {
        case GF_SNAP_OPTION_TYPE_CREATE:
                /* Syntax :
                 * gluster snapshot create <snapname> <vol-name(s)>
                 *                         [no-timestamp]
                 *                         [description <description>]
                 *                         [force]
                 */
                /* In cases where the snapname is not given then
                 * parsing fails & snapname cannot be "description",
                 * "force" and "volume", that check is made here
                 */
                if (wordcount == 2){
                        ret = -1;
                        gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                        goto out;
                }

                ret = validate_op_name ("snapname", words[2],
                                        invalid_snapnames);
                if (ret) {
                        goto out;
                }

                ret = cli_snap_create_parse (dict, words, wordcount);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR,
                                "create command parsing failed.");
                        goto out;
                }
                break;

        case GF_SNAP_OPTION_TYPE_CLONE:
                /* Syntax :
                 * gluster snapshot clone <clonename> <snapname>
                 */
                /* In cases where the clonename is not given then
                 * parsing fails & snapname cannot be "description",
                 * "force" and "volume", that check is made here
                 */
                if (wordcount == 2) {
                        ret = -1;
                        gf_log ("cli", GF_LOG_ERROR, "Invalid Syntax");
                        goto out;
                }

                ret = validate_op_name ("clonename", words[2],
                                        invalid_volnames);
                if (ret) {
                        goto out;
                }

                ret = cli_snap_clone_parse (dict, words, wordcount);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR,
                                "clone command parsing failed.");
                        goto out;
                }
                break;


        case GF_SNAP_OPTION_TYPE_INFO:
                /* Syntax :
                 * gluster snapshot info [(snapname] | [vol <volname>)]
                 */
                ret = cli_snap_info_parse (dict, words, wordcount);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Failed to parse "
                                "snapshot info command");
                        goto out;
                }
                break;

        case GF_SNAP_OPTION_TYPE_LIST:
                /* Syntax :
                 * gluster snaphsot list [volname]
                 */

                ret = cli_snap_list_parse (dict, words, wordcount);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Failed to parse "
                                "snapshot list command");
                        goto out;
                }
                break;

        case GF_SNAP_OPTION_TYPE_DELETE:
                /* Syntax :
                 * snapshot delete (all | snapname | volume <volname>)
                 */
                ret = cli_snap_delete_parse (dict, words, wordcount, state);
                if (ret) {
                        /* A positive ret value means user cancelled
                        * the command */
                        if (ret < 0) {
                                gf_log ("cli", GF_LOG_ERROR, "Failed to parse "
                                        "snapshot delete command");
                        }
                        goto out;
                }
                break;

        case GF_SNAP_OPTION_TYPE_CONFIG:
                /* snapshot config [volname]  [snap-max-hard-limit <count>]
                 *                            [snap-max-soft-limit <percent>] */
                ret = cli_snap_config_parse (words, wordcount, dict, state);
                if (ret) {
                        if (ret < 0)
                                gf_log ("cli", GF_LOG_ERROR,
                                        "config command parsing failed.");
                        goto out;
                }

                ret = dict_set_int32 (dict, "type", GF_SNAP_OPTION_TYPE_CONFIG);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Unable to set "
                                "config type");
                        ret = -1;
                        goto out;
                }
                break;

        case GF_SNAP_OPTION_TYPE_STATUS:
                {
                        /* Syntax :
                         * gluster snapshot status [(snapname |
                         *                         volume <volname>)]
                         */
                        ret = cli_snap_status_parse (dict, words, wordcount);
                        if (ret) {
                                gf_log ("cli", GF_LOG_ERROR, "Failed to parse "
                                        "snapshot status command");
                                goto out;
                        }
                        break;
                }

        case GF_SNAP_OPTION_TYPE_RESTORE:
                /* Syntax:
                 * snapshot restore <snapname>
                 */
                ret = cli_snap_restore_parse (dict, words, wordcount, state);
                if (ret) {
                        gf_log ("cli", GF_LOG_ERROR, "Failed to parse "
                                "restore command");
                        goto out;
                }
                break;

                case GF_SNAP_OPTION_TYPE_ACTIVATE:
                        /* Syntax:
                        * snapshot activate <snapname> [force]
                        */
                        ret = cli_snap_activate_parse (dict, words, wordcount);
                        if (ret) {
                                gf_log ("cli", GF_LOG_ERROR, "Failed to parse "
                                        "start command");
                                goto out;
                        }
                        break;
                case GF_SNAP_OPTION_TYPE_DEACTIVATE:
                        /* Syntax:
                        * snapshot deactivate <snapname>
                        */
                        ret = cli_snap_deactivate_parse (dict, words, wordcount,
                                state);
                        if (ret) {
                                /* A positive ret value means user cancelled
                                 * the command */
                                if (ret < 0) {
                                        gf_log ("cli", GF_LOG_ERROR,
                                                "Failed to parse deactivate "
                                                "command");
                                }
                                goto out;
                        }
                        break;

        default:
                gf_log ("", GF_LOG_ERROR, "Opword Mismatch");
                goto out;
        }

        ret = dict_set_int32 (dict, "type", type);
        if (ret) {
                gf_log ("", GF_LOG_ERROR,
                        "Failed to set type.");
                goto out;
        }
        /* If you got so far, input is valid */
        ret = 0;
out:
        if (ret) {
                if (dict)
                        dict_destroy (dict);
        } else
                *options = dict;

        return ret;
}

int
cli_cmd_validate_volume (char *volname)
{
        int       i        =  0;
        int       ret      = -1;


        if (volname[0] == '-')
                return ret;

        if (!strcmp (volname, "all")) {
                cli_err ("\"all\" cannot be the name of a volume.");
                return ret;
        }

        if (strchr (volname, '/')) {
                cli_err ("Volume name should not contain \"/\" character.");
                return ret;
        }

        if (strlen (volname) > GD_VOLUME_NAME_MAX) {
                cli_err ("Volname can not exceed %d characters.",
                        GD_VOLUME_NAME_MAX);
                return ret;
        }

        for (i = 0; i < strlen (volname); i++)
                if (!isalnum (volname[i]) && (volname[i] != '_') &&
                    (volname[i] != '-')) {
                        cli_err ("Volume name should not contain \"%c\""
                                 " character.\nVolume names can only"
                                 "contain alphanumeric, '-' and '_' "
                                 "characters.", volname[i]);
                        return ret;
                }

        ret = 0;

        return ret;
}

int32_t
cli_cmd_bitrot_parse (const char **words, int wordcount, dict_t **options)
{
        int32_t            ret                    = -1;
        char               *w                     = NULL;
        char               *volname               = NULL;
        char               *opwords[]             = {"enable", "disable",
                                                     "scrub-throttle",
                                                     "scrub-frequency", "scrub",
                                                     "signing-time", NULL};
        char               *scrub_throt_values[]  = {"lazy", "normal",
                                                     "aggressive", NULL};
        char               *scrub_freq_values[]   = {"hourly",
                                                     "daily", "weekly",
                                                     "biweekly", "monthly",
                                                      NULL};
        char               *scrub_values[]        = {"pause", "resume",
                                                      NULL};
        dict_t             *dict                  = NULL;
        gf_bitrot_type     type                   = GF_BITROT_OPTION_TYPE_NONE;
        int32_t            expiry_time            = 0;

        GF_ASSERT (words);
        GF_ASSERT (options);

        dict = dict_new ();
        if (!dict)
                goto out;

        if (wordcount < 4 || wordcount > 5) {
                gf_log ("cli", GF_LOG_ERROR, "Invalid syntax");
                goto out;
        }

        volname = (char *)words[2];
        if (!volname) {
                ret = -1;
                goto out;
        }

        ret = cli_cmd_validate_volume (volname);
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Failed to validate volume name");
                goto out;
        }

        ret = dict_set_str (dict, "volname", volname);
        if (ret) {
                cli_out ("Failed to set volume name in dictionary ");
                goto out;
        }

        w = str_getunamb (words[3], opwords);
        if (!w) {
                cli_out ("Invalid bit rot option : %s", words[3]);
                ret = -1;
                goto out;
        }

        if (strcmp (w, "enable") == 0) {
                if (wordcount == 4) {
                        type = GF_BITROT_OPTION_TYPE_ENABLE;
                        ret = 0;
                        goto set_type;
                } else {
                        ret = -1;
                        goto out;
                }
        }

        if (strcmp (w, "disable") == 0) {
                if (wordcount == 4) {
                        type = GF_BITROT_OPTION_TYPE_DISABLE;
                        ret = 0;
                        goto set_type;
                } else {
                        ret = -1;
                        goto out;
                }
        }

        if (!strcmp (w, "scrub-throttle")) {
                if (!words[4]) {
                        cli_err ("Missing scrub-throttle value for bitrot "
                                 "option");
                        ret = -1;
                        goto out;
                } else {
                        w = str_getunamb (words[4], scrub_throt_values);
                        if (!w) {
                                cli_err ("Invalid scrub-throttle option for "
                                         "bitrot");
                                ret = -1;
                                goto out;
                        } else {
                                type = GF_BITROT_OPTION_TYPE_SCRUB_THROTTLE;
                                ret =  dict_set_str (dict,
                                                     "scrub-throttle-value",
                                                     (char *) words[4]);
                                if (ret) {
                                        cli_out ("Failed to set scrub-throttle "
                                                 "value in the dict");
                                        goto out;
                                }
                                goto set_type;
                        }
                }
        }

        if (!strcmp (words[3], "scrub-frequency")) {
                if (!words[4]) {
                        cli_err ("Missing scrub-frequency value");
                        ret = -1;
                        goto out;
                } else {
                        w = str_getunamb (words[4], scrub_freq_values);
                        if (!w) {
                                cli_err ("Invalid frequency option for bitrot");
                                ret = -1;
                                goto out;
                        } else {
                                type = GF_BITROT_OPTION_TYPE_SCRUB_FREQ;
                                ret = dict_set_str (dict,
                                                    "scrub-frequency-value",
                                                    (char *) words[4]);
                                if (ret) {
                                        cli_out ("Failed to set dict for "
                                                 "bitrot");
                                        goto out;
                                }
                                goto set_type;
                        }
                }
        }

        if (!strcmp (words[3], "scrub")) {
                if (!words[4]) {
                        cli_err ("Missing scrub value for bitrot option");
                        ret = -1;
                        goto out;
                } else {
                        w = str_getunamb (words[4], scrub_values);
                        if (!w) {
                                cli_err ("Invalid scrub option for bitrot");
                                ret = -1;
                                goto out;
                        } else {
                                type = GF_BITROT_OPTION_TYPE_SCRUB;
                                ret =  dict_set_str (dict, "scrub-value",
                                                    (char *) words[4]);
                                if (ret) {
                                        cli_out ("Failed to set dict for "
                                                 "bitrot");
                                        goto out;
                                }
                                goto set_type;
                        }
                }
        }

        if (!strcmp (words[3], "signing-time")) {
                if (!words[4]) {
                        cli_err ("Missing signing-time value for bitrot "
                                 "option");
                        ret = -1;
                        goto out;
                } else {
                        type = GF_BITROT_OPTION_TYPE_EXPIRY_TIME;

                        expiry_time = strtol (words[4], NULL, 0);
                        if (expiry_time < 1) {
                                cli_err ("Expiry time  value should not be less"
                                         " than 1");
                                ret = -1;
                                goto out;
                        }

                        ret = dict_set_uint32 (dict, "expiry-time",
                                               (unsigned int) expiry_time);
                        if (ret) {
                                cli_out ("Failed to set dict for bitrot");
                                goto out;
                        }
                        goto set_type;
                }
        } else {
                cli_err ("Invalid option %s for bitrot. Please enter valid "
                         "bitrot option", words[3]);
                ret = -1;
                goto out;
        }

set_type:
        ret = dict_set_int32 (dict, "type", type);
        if (ret < 0)
                goto out;

        *options = dict;

out:
        if (ret) {
                gf_log ("cli", GF_LOG_ERROR, "Unable to parse bitrot command");
                if (dict)
                        dict_destroy (dict);
        }

        return ret;
}
s='add'>+ int32_t *success_children = NULL;
+ struct iatt *bufs = NULL;
+ int i = 0;
+ int child = 0;
- child_index = (long) cookie;
priv = this->private;
+ success_children = local->cont.lookup.success_children;
+ bufs = local->cont.lookup.bufs;
+ memcpy (local->fresh_children, success_children,
+ sizeof (*success_children) * priv->child_count);
+ for (i = 0; i < priv->child_count; i++) {
+ child = local->fresh_children[i];
+ if (child == -1)
+ break;
+ if (child == subvol)
+ continue;
+ if (uuid_is_null (bufs[child].ia_gfid) &&
+ (bufs[child].ia_type == bufs[subvol].ia_type))
+ continue;
+ afr_children_rm_child (success_children, child,
+ priv->child_count);
+ local->success_count--;
+ }
+ afr_reset_children (local->fresh_children, priv->child_count);
+}
- LOCK (&frame->lock);
- {
- local = frame->local;
+void
+afr_succeed_lookup_on_latest_iatt (afr_local_t *local, xlator_t *this)
+{
+ int lsubvol = 0;
- lookup_buf = &local->cont.lookup.buf;
+ if (!afr_lookup_conflicting_entries (local, this))
+ goto out;
- if (op_ret == -1) {
- if (op_errno == ENOENT)
- local->enoent_count++;
+ lsubvol = afr_lookup_get_latest_subvol (local, this);
+ if (lsubvol < 0)
+ goto out;
+ afr_lookup_mark_other_entries_stale (local, this, lsubvol);
+out:
+ return;
+}
- if (__error_more_important (local->op_errno, op_errno))
- local->op_errno = op_errno;
+gf_boolean_t
+afr_is_entry_possibly_under_creation (afr_local_t *local, xlator_t *this)
+{
+ /*
+ * We need to perform this test in lookup done and treat on going
+ * create/DELETE as ENOENT.
+ * Reason:
+ Multiple clients A, B and C are attempting 'mkdir -p /mnt/a/b/c'
+
+ 1 Client A is in the middle of mkdir(/a). It has acquired lock.
+ It has performed mkdir(/a) on one subvol, and second one is still
+ in progress
+ 2 Client B performs a lookup, sees directory /a on one,
+ ENOENT on the other, succeeds lookup.
+ 3 Client B performs lookup on /a/b on both subvols, both return ENOENT
+ (one subvol because /a/b does not exist, another because /a
+ itself does not exist)
+ 4 Client B proceeds to mkdir /a/b. It obtains entrylk on inode=/a with
+ basename=b on one subvol, but fails on other subvol as /a is yet to
+ be created by Client A.
+ 5 Client A finishes mkdir of /a on other subvol
+ 6 Client C also attempts to create /a/b, lookup returns ENOENT on
+ both subvols.
+ 7 Client C tries to obtain entrylk on on inode=/a with basename=b,
+ obtains on one subvol (where B had failed), and waits for B to unlock
+ on other subvol.
+ 8 Client B finishes mkdir() on one subvol with GFID-1 and completes
+ transaction and unlocks
+ 9 Client C gets the lock on the second subvol, At this stage second
+ subvol already has /a/b created from Client B, but Client C does not
+ check that in the middle of mkdir transaction
+ 10 Client C attempts mkdir /a/b on both subvols. It succeeds on
+ ONLY ONE (where Client B could not get lock because of
+ missing parent /a dir) with GFID-2, and gets EEXIST from ONE subvol.
+ This way we have /a/b in GFID mismatch. One subvol got GFID-1 because
+ Client B performed transaction on only one subvol (because entrylk()
+ could not be obtained on second subvol because of missing parent dir --
+ caused by premature/speculative succeeding of lookup() on /a when locks
+ are detected). Other subvol gets GFID-2 from Client C because while
+ it was waiting for entrylk() on both subvols, Client B was in the
+ middle of creating mkdir() on only one subvol, and Client C does not
+ "expect" this when it is between lock() and pre-op()/op() phase of the
+ transaction.
+ */
+ if (local->cont.lookup.parent_entrylk && local->enoent_count)
+ return _gf_true;
+
+ return _gf_false;
+}
- if (local->op_errno == ESTALE) {
- local->op_ret = -1;
- }
- goto unlock;
- }
+static void
+afr_lookup_done (call_frame_t *frame, xlator_t *this)
+{
+ int unwind = 1;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int ret = -1;
+ gf_boolean_t sh_launched = _gf_false;
+ gf_boolean_t fail_conflict = _gf_false;
+ int gfid_miss_count = 0;
+ int enotconn_count = 0;
+ int up_children_count = 0;
- afr_lookup_collect_xattr (local, this, child_index, xattr);
+ priv = this->private;
+ local = frame->local;
- first_up_child = afr_first_up_child (priv);
+ if (afr_is_entry_possibly_under_creation (local, this)) {
+ local->op_ret = -1;
+ local->op_errno = ENOENT;
+ goto unwind;
+ }
- if (child_index == first_up_child) {
- local->cont.lookup.ino =
- afr_itransform (buf->ia_ino,
- priv->child_count,
- first_up_child);
- }
+ if (local->op_ret < 0)
+ goto unwind;
- /* in case of revalidate, we need to send stat of the
- * child whose stat was sent during the first lookup.
- * (so that time stamp does not vary with revalidate.
- * in case it is down, stat of the fist success will
- * be replied */
+ if (local->cont.lookup.parent_entrylk && local->success_count > 1)
+ afr_succeed_lookup_on_latest_iatt (local, this);
+
+ gfid_miss_count = afr_lookup_gfid_missing_count (local, this);
+ up_children_count = afr_up_children_count (local->child_up,
+ priv->child_count);
+ enotconn_count = priv->child_count - up_children_count;
+ if ((gfid_miss_count == local->success_count) &&
+ (enotconn_count > 0)) {
+ local->op_ret = -1;
+ local->op_errno = EIO;
+ gf_log (this->name, GF_LOG_ERROR, "Failing lookup for %s, "
+ "LOOKUP on a file without gfid is not allowed when "
+ "some of the children are down", local->loc.path);
+ goto unwind;
+ }
- /* inode number should be preserved across revalidates */
+ if ((gfid_miss_count == local->success_count) &&
+ uuid_is_null (local->cont.lookup.gfid_req)) {
+ local->op_ret = -1;
+ local->op_errno = ENODATA;
+ gf_log (this->name, GF_LOG_ERROR, "%s: No gfid present",
+ local->loc.path);
+ goto unwind;
+ }
- if (local->success_count == 0) {
- if (local->op_errno != ESTALE)
- local->op_ret = op_ret;
+ if (gfid_miss_count && uuid_is_null (local->cont.lookup.gfid_req))
+ fail_conflict = _gf_true;
+ ret = afr_lookup_done_success_action (frame, this, fail_conflict);
+ if (ret)
+ goto unwind;
+ uuid_copy (local->self_heal.sh_gfid_req, local->cont.lookup.gfid_req);
- local->cont.lookup.inode = inode_ref (inode);
- local->cont.lookup.xattr = dict_ref (xattr);
- local->cont.lookup.xattrs[child_index] = dict_ref (xattr);
- local->cont.lookup.postparent = *postparent;
+ afr_lookup_perform_self_heal (frame, this, &sh_launched);
+ if (sh_launched) {
+ unwind = 0;
+ goto unwind;
+ }
- *lookup_buf = *buf;
+ unwind:
+ if (unwind) {
+ AFR_STACK_UNWIND (lookup, frame, local->op_ret,
+ local->op_errno, local->cont.lookup.inode,
+ &local->cont.lookup.buf,
+ local->cont.lookup.xattr,
+ &local->cont.lookup.postparent);
+ }
+}
- lookup_buf->ia_ino = afr_itransform (buf->ia_ino,
- priv->child_count,
- child_index);
+/*
+ * During a lookup, some errors are more "important" than
+ * others in that they must be given higher priority while
+ * returning to the user.
+ *
+ * The hierarchy is ESTALE > EIO > ENOENT > others
+ */
+int32_t
+afr_most_important_error(int32_t old_errno, int32_t new_errno,
+ gf_boolean_t eio)
+{
+ if (old_errno == ESTALE || new_errno == ESTALE)
+ return ESTALE;
+ if (eio && (old_errno == EIO || new_errno == EIO))
+ return EIO;
+ if (old_errno == ENOENT || new_errno == ENOENT)
+ return ENOENT;
+
+ return new_errno;
+}
- if (priv->read_child >= 0) {
- afr_set_read_child (this,
- local->cont.lookup.inode,
- priv->read_child);
- } else {
- afr_set_read_child (this,
- local->cont.lookup.inode,
- child_index);
- }
+int32_t
+afr_resultant_errno_get (int32_t *children,
+ int *child_errno, unsigned int child_count)
+{
+ int i = 0;
+ int32_t op_errno = 0;
+ int child = 0;
+ for (i = 0; i < child_count; i++) {
+ if (children) {
+ child = children[i];
+ if (child == -1)
+ break;
} else {
- afr_lookup_self_heal_check (this, local, buf, lookup_buf);
+ child = i;
+ }
+ op_errno = afr_most_important_error(op_errno,
+ child_errno[child],
+ _gf_false);
+ }
+ return op_errno;
+}
- if (child_index == local->read_child_index) {
+static void
+afr_lookup_handle_error (afr_local_t *local, int32_t op_ret, int32_t op_errno)
+{
+ GF_ASSERT (local);
+ if (op_errno == ENOENT)
+ local->enoent_count++;
- /*
- lookup has succeeded on the read child.
- So use its inode number
- */
+ local->op_errno = afr_most_important_error(local->op_errno, op_errno,
+ _gf_false);
- if (local->cont.lookup.xattr)
- dict_unref (local->cont.lookup.xattr);
+ if (local->op_errno == ESTALE) {
+ local->op_ret = -1;
+ }
+}
- local->cont.lookup.xattr = dict_ref (xattr);
- local->cont.lookup.xattrs[child_index] = dict_ref (xattr);
- local->cont.lookup.postparent = *postparent;
+static void
+afr_set_root_inode_on_first_lookup (afr_local_t *local, xlator_t *this,
+ inode_t *inode)
+{
+ afr_private_t *priv = NULL;
+ GF_ASSERT (inode);
- *lookup_buf = *buf;
+ if (!__is_root_gfid (inode->gfid))
+ goto out;
+ if (!afr_is_fresh_lookup (&local->loc, this))
+ goto out;
+ priv = this->private;
+ if ((priv->first_lookup)) {
+ gf_log (this->name, GF_LOG_INFO, "added root inode");
+ priv->root_inode = inode_ref (inode);
+ priv->first_lookup = 0;
+ }
+out:
+ return;
+}
- if (priv->read_child >= 0) {
- afr_set_read_child (this,
- local->cont.lookup.inode,
- priv->read_child);
- } else {
- afr_set_read_child (this,
- local->cont.lookup.inode,
- local->read_child_index);
- }
- }
+static void
+afr_lookup_cache_args (afr_local_t *local, int child_index, dict_t *xattr,
+ struct iatt *buf, struct iatt *postparent)
+{
+ GF_ASSERT (child_index >= 0);
+ local->cont.lookup.xattrs[child_index] = dict_ref (xattr);
+ local->cont.lookup.postparents[child_index] = *postparent;
+ local->cont.lookup.bufs[child_index] = *buf;
+}
+
+static void
+afr_lookup_handle_first_success (afr_local_t *local, xlator_t *this,
+ inode_t *inode, struct iatt *buf)
+{
+ local->cont.lookup.inode = inode_ref (inode);
+ local->cont.lookup.buf = *buf;
+ afr_set_root_inode_on_first_lookup (local, this, inode);
+}
+
+static int32_t
+afr_discovery_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
+{
+ int ret = 0;
+ char *pathinfo = NULL;
+ gf_boolean_t is_local = _gf_false;
+ afr_private_t *priv = NULL;
+ int32_t child_index = -1;
+
+ if (op_ret != 0) {
+ goto out;
+ }
+
+ ret = dict_get_str (dict, GF_XATTR_PATHINFO_KEY, &pathinfo);
+ if (ret != 0) {
+ goto out;
+ }
+
+ ret = afr_local_pathinfo (pathinfo, &is_local);
+ if (ret) {
+ goto out;
+ }
+
+ priv = this->private;
+ /*
+ * Note that one local subvolume will override another here. The only
+ * way to avoid that would be to retain extra information about whether
+ * the previous read_child is local, and it's just not worth it. Even
+ * the slowest local subvolume is far preferable to a remote one.
+ */
+ if (is_local) {
+ child_index = (int32_t)(long)cookie;
+ gf_log (this->name, GF_LOG_INFO,
+ "selecting local read_child %s",
+ priv->children[child_index]->name);
+ priv->read_child = child_index;
+ }
+
+out:
+ STACK_DESTROY(frame->root);
+ return 0;
+}
+
+static void
+afr_attempt_local_discovery (xlator_t *this, int32_t child_index)
+{
+ call_frame_t *newframe = NULL;
+ loc_t tmploc = {0,};
+ afr_private_t *priv = this->private;
+ newframe = create_frame(this,this->ctx->pool);
+ if (!newframe) {
+ return;
+ }
+
+ tmploc.gfid[sizeof(tmploc.gfid)-1] = 1;
+ STACK_WIND_COOKIE (newframe, afr_discovery_cbk,
+ (void *)(long)child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->getxattr,
+ &tmploc, GF_XATTR_PATHINFO_KEY, NULL);
+}
+
+static void
+afr_lookup_handle_success (afr_local_t *local, xlator_t *this, int32_t child_index,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xattr,
+ struct iatt *postparent)
+{
+ afr_private_t *priv = this->private;
+
+ if (local->success_count == 0) {
+ if (local->op_errno != ESTALE) {
+ local->op_ret = op_ret;
+ local->op_errno = 0;
}
+ afr_lookup_handle_first_success (local, this, inode, buf);
+ }
+ afr_lookup_update_lk_counts (local, this,
+ child_index, xattr);
+
+ afr_lookup_cache_args (local, child_index, xattr,
+ buf, postparent);
- local->success_count++;
+ if (local->do_discovery && (priv->read_child == (-1))) {
+ afr_attempt_local_discovery(this,child_index);
}
+
+ local->cont.lookup.success_children[local->success_count] = child_index;
+ local->success_count++;
+}
+
+int
+afr_lookup_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ inode_t *inode, struct iatt *buf, dict_t *xattr,
+ struct iatt *postparent)
+{
+ afr_local_t * local = NULL;
+ int call_count = -1;
+ int child_index = -1;
+
+ child_index = (long) cookie;
+
+ LOCK (&frame->lock);
+ {
+ local = frame->local;
+
+ if (op_ret == -1) {
+ afr_lookup_handle_error (local, op_ret, op_errno);
+ goto unlock;
+ }
+ afr_lookup_handle_success (local, this, child_index, op_ret,
+ op_errno, inode, buf, xattr,
+ postparent);
+
+ }
unlock:
UNLOCK (&frame->lock);
call_count = afr_frame_return (frame);
-
if (call_count == 0) {
- afr_lookup_done (frame, this, lookup_buf);
+ afr_lookup_done (frame, this);
}
- return 0;
+ return 0;
}
-
int
-afr_lookup (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xattr_req)
+afr_lookup_cont_init (afr_local_t *local, unsigned int child_count)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int ret = -1;
- int i = 0;
+ int ret = -ENOMEM;
+ struct iatt *iatts = NULL;
+ int32_t *success_children = NULL;
+ int32_t *sources = NULL;
+ int32_t **pending_matrix = NULL;
+
+ GF_ASSERT (local);
+ local->cont.lookup.xattrs = GF_CALLOC (child_count,
+ sizeof (*local->cont.lookup.xattr),
+ gf_afr_mt_dict_t);
+ if (NULL == local->cont.lookup.xattrs)
+ goto out;
- fop_lookup_cbk_t callback;
+ iatts = GF_CALLOC (child_count, sizeof (*iatts), gf_afr_mt_iatt);
+ if (NULL == iatts)
+ goto out;
+ local->cont.lookup.postparents = iatts;
- int call_count = 0;
+ iatts = GF_CALLOC (child_count, sizeof (*iatts), gf_afr_mt_iatt);
+ if (NULL == iatts)
+ goto out;
+ local->cont.lookup.bufs = iatts;
- uint64_t ctx;
+ success_children = afr_children_create (child_count);
+ if (NULL == success_children)
+ goto out;
+ local->cont.lookup.success_children = success_children;
+
+ local->fresh_children = afr_children_create (child_count);
+ if (NULL == local->fresh_children)
+ goto out;
+
+ sources = GF_CALLOC (sizeof (*sources), child_count, gf_afr_mt_int32_t);
+ if (NULL == sources)
+ goto out;
+ local->cont.lookup.sources = sources;
+
+ pending_matrix = afr_matrix_create (child_count, child_count);
+ if (NULL == pending_matrix)
+ goto out;
+ local->cont.lookup.pending_matrix = pending_matrix;
- int32_t op_errno = 0;
+ ret = 0;
+out:
+ return ret;
+}
+int
+afr_lookup (call_frame_t *frame, xlator_t *this,
+ loc_t *loc, dict_t *xattr_req)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ void *gfid_req = NULL;
+ int ret = -1;
+ int i = 0;
+ int call_count = 0;
+ uint64_t ctx = 0;
+ int32_t op_errno = 0;
+ int allow_sh = 0;
priv = this->private;
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (local, out);
local->op_ret = -1;
frame->local = local;
+ local->fop = GF_FOP_LOOKUP;
- if (!strcmp (loc->path, "/" GF_REPLICATE_TRASH_DIR)) {
- op_errno = ENOENT;
+ loc_copy (&local->loc, loc);
+ ret = loc_path (&local->loc, NULL);
+ if (ret < 0) {
+ op_errno = EINVAL;
goto out;
}
- loc_copy (&local->loc, loc);
+ if (local->loc.path &&
+ (strcmp (local->loc.path, "/" GF_REPLICATE_TRASH_DIR) == 0)) {
+ op_errno = EPERM;
+ ret = -1;
+ goto out;
+ }
- ret = inode_ctx_get (loc->inode, this, &ctx);
+ ret = inode_ctx_get (local->loc.inode, this, &ctx);
if (ret == 0) {
/* lookup is a revalidate */
- callback = afr_revalidate_lookup_cbk;
-
- local->cont.lookup.is_revalidate = _gf_true;
- local->read_child_index = afr_read_child (this,
- loc->inode);
+ local->read_child_index = afr_inode_get_read_ctx (this,
+ local->loc.inode,
+ NULL);
} else {
- callback = afr_fresh_lookup_cbk;
-
LOCK (&priv->read_child_lock);
{
- local->read_child_index = (++priv->read_child_rr)
- % (priv->child_count);
+ if (priv->hash_mode) {
+ local->read_child_index = -1;
+ }
+ else {
+ local->read_child_index =
+ (++priv->read_child_rr) %
+ (priv->child_count);
+ }
}
UNLOCK (&priv->read_child_lock);
+ local->cont.lookup.fresh_lookup = _gf_true;
}
- if (loc->parent)
- local->cont.lookup.parent_ino = loc->parent->ino;
-
- local->child_up = memdup (priv->child_up, priv->child_count);
+ local->child_up = memdup (priv->child_up,
+ sizeof (*local->child_up) * priv->child_count);
+ if (NULL == local->child_up) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- local->cont.lookup.xattrs = GF_CALLOC (priv->child_count,
- sizeof (*local->cont.lookup.xattr),
- gf_afr_mt_dict_t);
+ ret = afr_lookup_cont_init (local, priv->child_count);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- local->call_count = afr_up_children_count (priv->child_count,
- local->child_up);
+ local->call_count = afr_up_children_count (local->child_up,
+ priv->child_count);
call_count = local->call_count;
-
if (local->call_count == 0) {
ret = -1;
op_errno = ENOTCONN;
@@ -1083,38 +2500,33 @@ afr_lookup (call_frame_t *frame, xlator_t *this,
/* By default assume ENOTCONN. On success it will be set to 0. */
local->op_errno = ENOTCONN;
- if (xattr_req == NULL)
- local->xattr_req = dict_new ();
- else
- local->xattr_req = dict_ref (xattr_req);
+ ret = dict_get_int32 (xattr_req, "allow-sh-for-running-transaction",
+ &allow_sh);
+ dict_del (xattr_req, "allow-sh-for-running-transaction");
+ local->allow_sh_for_running_transaction = allow_sh;
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_set_uint64 (local->xattr_req, priv->pending_key[i],
- 3 * sizeof(int32_t));
- if (ret < 0)
- gf_log (this->name, GF_LOG_WARNING,
- "Unable to set dict value.");
- /* 3 = data+metadata+entry */
- }
-
- ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_INODELK_COUNT, 0);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Unable to set dict value.");
+ ret = afr_lookup_xattr_req_prepare (local, this, xattr_req, &local->loc,
+ &gfid_req);
+ if (ret) {
+ local->op_errno = -ret;
+ goto out;
}
-
- ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_ENTRYLK_COUNT, 0);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Unable to set dict value.");
+ afr_lookup_save_gfid (local->cont.lookup.gfid_req, gfid_req,
+ &local->loc);
+ local->fop = GF_FOP_LOOKUP;
+ if (priv->choose_local && !priv->did_discovery) {
+ if (gfid_req && __is_root_gfid(gfid_req)) {
+ local->do_discovery = _gf_true;
+ priv->did_discovery = _gf_true;
+ }
}
-
for (i = 0; i < priv->child_count; i++) {
if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, callback, (void *) (long) i,
+ STACK_WIND_COOKIE (frame, afr_lookup_cbk,
+ (void *) (long) i,
priv->children[i],
priv->children[i]->fops->lookup,
- loc, local->xattr_req);
+ &local->loc, local->xattr_req);
if (!--call_count)
break;
}
@@ -1122,7 +2534,7 @@ afr_lookup (call_frame_t *frame, xlator_t *this,
ret = 0;
out:
- if (ret == -1)
+ if (ret)
AFR_STACK_UNWIND (lookup, frame, -1, op_errno,
NULL, NULL, NULL, NULL);
@@ -1133,231 +2545,179 @@ out:
/* {{{ open */
int
-afr_fd_ctx_set (xlator_t *this, fd_t *fd)
+__afr_fd_ctx_set (xlator_t *this, fd_t *fd)
{
- afr_private_t * priv = NULL;
-
- int ret = 0;
-
- uint64_t ctx;
- afr_fd_ctx_t * fd_ctx = NULL;
+ afr_private_t * priv = NULL;
+ int ret = -1;
+ uint64_t ctx = 0;
+ afr_fd_ctx_t * fd_ctx = NULL;
VALIDATE_OR_GOTO (this->private, out);
VALIDATE_OR_GOTO (fd, out);
priv = this->private;
- LOCK (&fd->lock);
- {
- ret = __fd_ctx_get (fd, this, &ctx);
-
- if (ret == 0)
- goto unlock;
+ ret = __fd_ctx_get (fd, this, &ctx);
- fd_ctx = GF_CALLOC (1, sizeof (afr_fd_ctx_t),
- gf_afr_mt_afr_fd_ctx_t);
- if (!fd_ctx) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
+ if (ret == 0)
+ goto out;
- ret = -ENOMEM;
- goto unlock;
- }
+ fd_ctx = GF_CALLOC (1, sizeof (afr_fd_ctx_t),
+ gf_afr_mt_afr_fd_ctx_t);
+ if (!fd_ctx) {
+ ret = -ENOMEM;
+ goto out;
+ }
- fd_ctx->pre_op_done = GF_CALLOC (sizeof (*fd_ctx->pre_op_done),
- priv->child_count,
- gf_afr_mt_char);
- if (!fd_ctx->pre_op_done) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- ret = -ENOMEM;
- goto unlock;
- }
+ fd_ctx->pre_op_done = GF_CALLOC (sizeof (*fd_ctx->pre_op_done),
+ priv->child_count,
+ gf_afr_mt_char);
+ if (!fd_ctx->pre_op_done) {
+ ret = -ENOMEM;
+ goto out;
+ }
- fd_ctx->pre_op_piggyback = GF_CALLOC (sizeof (*fd_ctx->pre_op_piggyback),
- priv->child_count,
- gf_afr_mt_char);
- if (!fd_ctx->pre_op_piggyback) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- ret = -ENOMEM;
- goto unlock;
- }
+ fd_ctx->pre_op_piggyback = GF_CALLOC (sizeof (*fd_ctx->pre_op_piggyback),
+ priv->child_count,
+ gf_afr_mt_char);
+ if (!fd_ctx->pre_op_piggyback) {
+ ret = -ENOMEM;
+ goto out;
+ }
- fd_ctx->opened_on = GF_CALLOC (sizeof (*fd_ctx->opened_on),
- priv->child_count,
- gf_afr_mt_char);
- if (!fd_ctx->opened_on) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- ret = -ENOMEM;
- goto unlock;
- }
+ fd_ctx->opened_on = GF_CALLOC (sizeof (*fd_ctx->opened_on),
+ priv->child_count,
+ gf_afr_mt_int32_t);
+ if (!fd_ctx->opened_on) {
+ ret = -ENOMEM;
+ goto out;
+ }
- fd_ctx->up_count = priv->up_count;
- fd_ctx->down_count = priv->down_count;
+ fd_ctx->lock_piggyback = GF_CALLOC (sizeof (*fd_ctx->lock_piggyback),
+ priv->child_count,
+ gf_afr_mt_char);
+ if (!fd_ctx->lock_piggyback) {
+ ret = -ENOMEM;
+ goto out;
+ }
- fd_ctx->locked_on = GF_CALLOC (sizeof (*fd_ctx->locked_on),
- priv->child_count,
- gf_afr_mt_char);
- if (!fd_ctx->locked_on) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- ret = -ENOMEM;
- goto unlock;
- }
+ fd_ctx->lock_acquired = GF_CALLOC (sizeof (*fd_ctx->lock_acquired),
+ priv->child_count,
+ gf_afr_mt_char);
+ if (!fd_ctx->lock_acquired) {
+ ret = -ENOMEM;
+ goto out;
+ }
- ret = __fd_ctx_set (fd, this, (uint64_t)(long) fd_ctx);
+ fd_ctx->up_count = priv->up_count;
+ fd_ctx->down_count = priv->down_count;
- INIT_LIST_HEAD (&fd_ctx->entries);
+ fd_ctx->locked_on = GF_CALLOC (sizeof (*fd_ctx->locked_on),
+ priv->child_count,
+ gf_afr_mt_char);
+ if (!fd_ctx->locked_on) {
+ ret = -ENOMEM;
+ goto out;
}
-unlock:
- UNLOCK (&fd->lock);
+
+ pthread_mutex_init (&fd_ctx->delay_lock, NULL);
+ INIT_LIST_HEAD (&fd_ctx->entries);
+ fd_ctx->call_child = -1;
+
+ INIT_LIST_HEAD (&fd_ctx->eager_locked);
+
+ ret = __fd_ctx_set (fd, this, (uint64_t)(long) fd_ctx);
+ if (ret)
+ gf_log (this->name, GF_LOG_DEBUG,
+ "failed to set fd ctx (%p)", fd);
out:
return ret;
}
-/* {{{ flush */
int
-afr_flush_unwind (call_frame_t *frame, xlator_t *this)
+afr_fd_ctx_set (xlator_t *this, fd_t *fd)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ int ret = -1;
- LOCK (&frame->lock);
+ LOCK (&fd->lock);
{
- if (local->transaction.main_frame)
- main_frame = local->transaction.main_frame;
- local->transaction.main_frame = NULL;
- }
- UNLOCK (&frame->lock);
-
- if (main_frame) {
- AFR_STACK_UNWIND (flush, main_frame,
- local->op_ret, local->op_errno);
+ ret = __afr_fd_ctx_set (this, fd);
}
+ UNLOCK (&fd->lock);
- return 0;
+ return ret;
}
+/* {{{ flush */
int
-afr_flush_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+afr_flush_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
-
int call_count = -1;
- int child_index = (long) cookie;
- int need_unwind = 0;
local = frame->local;
- priv = this->private;
LOCK (&frame->lock);
{
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
if (op_ret != -1) {
if (local->success_count == 0) {
local->op_ret = op_ret;
}
local->success_count++;
-
- if (local->success_count == priv->wait_count) {
- need_unwind = 1;
- }
}
local->op_errno = op_errno;
}
UNLOCK (&frame->lock);
- if (need_unwind)
- afr_flush_unwind (frame, this);
+ call_count = afr_frame_return (frame);
- call_count = afr_frame_return (frame);
-
- if (call_count == 0) {
- local->transaction.resume (frame, this);
- }
+ if (call_count == 0)
+ AFR_STACK_UNWIND(flush, frame, local->op_ret,
+ local->op_errno, NULL);
return 0;
}
-
-int
-afr_flush_wind (call_frame_t *frame, xlator_t *this)
+static int
+afr_flush_wrapper (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- int i = 0;
- int call_count = -1;
+ int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
- local = frame->local;
priv = this->private;
-
- call_count = afr_up_children_count (priv->child_count, local->child_up);
-
- if (call_count == 0) {
- local->transaction.resume (frame, this);
- return 0;
- }
-
- local->call_count = call_count;
+ local = frame->local;
+ call_count = local->call_count;
for (i = 0; i < priv->child_count; i++) {
if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_flush_wind_cbk,
+ STACK_WIND_COOKIE (frame, afr_flush_cbk,
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->flush,
- local->fd);
-
+ local->fd, NULL);
if (!--call_count)
break;
+
}
}
return 0;
}
-
-int
-afr_flush_done (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
-
- local = frame->local;
-
- local->transaction.unwind (frame, this);
-
- AFR_STACK_DESTROY (frame);
-
- return 0;
-}
-
-
int
-afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
+afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
-
- call_frame_t * transaction_frame = NULL;
-
- int ret = -1;
-
- int op_ret = -1;
- int op_errno = 0;
-
- int call_count = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ int ret = -1;
+ int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -1365,49 +2725,27 @@ afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
priv = this->private;
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_local_init(local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
- call_count = afr_up_children_count (priv->child_count, local->child_up);
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame) {
+ local->fd = fd_ref(fd);
+ stub = fop_flush_stub (frame, afr_flush_wrapper, fd, xdata);
+ if (!stub) {
+ ret = -1;
op_errno = ENOMEM;
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
goto out;
}
- transaction_frame->local = local;
-
- local->op = GF_FOP_FLUSH;
-
- local->transaction.fop = afr_flush_wind;
- local->transaction.done = afr_flush_done;
- local->transaction.unwind = afr_flush_unwind;
-
- local->fd = fd_ref (fd);
-
- local->transaction.main_frame = frame;
- local->transaction.start = 0;
- local->transaction.len = 0;
-
- afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
-
+ afr_delayed_changelog_wake_resume (this, fd, stub);
+ ret = 0;
- op_ret = 0;
out:
- if (op_ret == -1) {
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
-
- AFR_STACK_UNWIND (flush, frame, op_ret, op_errno);
- }
+ if (ret < 0)
+ AFR_STACK_UNWIND(flush, frame, -1, op_errno, NULL);
return 0;
}
@@ -1423,21 +2761,24 @@ afr_cleanup_fd_ctx (xlator_t *this, fd_t *fd)
int ret = 0;
ret = fd_ctx_get (fd, this, &ctx);
-
if (ret < 0)
goto out;
fd_ctx = (afr_fd_ctx_t *)(long) ctx;
if (fd_ctx) {
- if (fd_ctx->pre_op_done)
- GF_FREE (fd_ctx->pre_op_done);
+ GF_FREE (fd_ctx->pre_op_done);
+
+ GF_FREE (fd_ctx->opened_on);
+
+ GF_FREE (fd_ctx->locked_on);
- if (fd_ctx->opened_on)
- GF_FREE (fd_ctx->opened_on);
+ GF_FREE (fd_ctx->pre_op_piggyback);
+ GF_FREE (fd_ctx->lock_piggyback);
- if (fd_ctx->locked_on)
- GF_FREE (fd_ctx->locked_on);
+ GF_FREE (fd_ctx->lock_acquired);
+
+ pthread_mutex_destroy (&fd_ctx->delay_lock);
GF_FREE (fd_ctx);
}
@@ -1450,8 +2791,24 @@ out:
int
afr_release (xlator_t *this, fd_t *fd)
{
+ afr_locked_fd_t *locked_fd = NULL;
+ afr_locked_fd_t *tmp = NULL;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+
afr_cleanup_fd_ctx (this, fd);
+ list_for_each_entry_safe (locked_fd, tmp, &priv->saved_fds,
+ list) {
+
+ if (locked_fd->fd == fd) {
+ list_del_init (&locked_fd->list);
+ GF_FREE (locked_fd);
+ }
+
+ }
+
return 0;
}
@@ -1459,20 +2816,29 @@ afr_release (xlator_t *this, fd_t *fd)
/* {{{ fsync */
int
+afr_fsync_unwind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ AFR_STACK_UNWIND (fsync, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
+}
+
+int
afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
+ struct iatt *postbuf, dict_t *xdata)
{
afr_local_t *local = NULL;
-
int call_count = -1;
-
int child_index = (long) cookie;
int read_child = 0;
+ call_stub_t *stub = NULL;
local = frame->local;
- read_child = afr_read_child (this, local->fd->inode);
+ read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
LOCK (&frame->lock);
{
@@ -1484,13 +2850,13 @@ afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->op_ret = 0;
if (local->success_count == 0) {
- local->cont.fsync.prebuf = *prebuf;
- local->cont.fsync.postbuf = *postbuf;
+ local->cont.inode_wfop.prebuf = *prebuf;
+ local->cont.inode_wfop.postbuf = *postbuf;
}
if (child_index == read_child) {
- local->cont.fsync.prebuf = *prebuf;
- local->cont.fsync.postbuf = *postbuf;
+ local->cont.inode_wfop.prebuf = *prebuf;
+ local->cont.inode_wfop.postbuf = *postbuf;
}
local->success_count++;
@@ -1503,12 +2869,32 @@ afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
call_count = afr_frame_return (frame);
if (call_count == 0) {
- local->cont.fsync.prebuf.ia_ino = local->cont.fsync.ino;
- local->cont.fsync.postbuf.ia_ino = local->cont.fsync.ino;
-
- AFR_STACK_UNWIND (fsync, frame, local->op_ret, local->op_errno,
- &local->cont.fsync.prebuf,
- &local->cont.fsync.postbuf);
+ /* Make a stub out of the frame, and register it
+ with the waking up post-op. When the call-stub resumes,
+ we are guaranteed that there was no post-op pending
+ (i.e changelogs were unset in the server). This is an
+ essential "guarantee", that fsync() returns only after
+ completely finishing EVERYTHING, including the delayed
+ post-op. This guarantee is expected by FUSE graph switching
+ for example.
+ */
+ stub = fop_fsync_cbk_stub (frame, afr_fsync_unwind_cbk,
+ local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf,
+ xdata);
+ if (!stub) {
+ AFR_STACK_UNWIND (fsync, frame, -1, ENOMEM, 0, 0, 0);
+ return 0;
+ }
+
+ /* If no new unstable writes happened between the
+ time we cleared the unstable write witness flag in afr_fsync
+ and now, calling afr_delayed_changelog_wake_up() should
+ wake up and skip over the fsync phase and go straight to
+ afr_changelog_post_op_now()
+ */
+ afr_delayed_changelog_wake_resume (this, local->fd, stub);
}
return 0;
@@ -1517,16 +2903,13 @@ afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int
afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int32_t datasync)
+ int32_t datasync, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
-
int ret = -1;
-
int i = 0;
int32_t call_count = 0;
- int32_t op_ret = -1;
int32_t op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
@@ -1535,19 +2918,20 @@ afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
priv = this->private;
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
goto out;
- }
call_count = local->call_count;
- frame->local = local;
local->fd = fd_ref (fd);
- local->cont.fsync.ino = fd->inode->ino;
+
+ if (afr_fd_has_witnessed_unstable_write (this, fd)) {
+ /* don't care. we only wanted to CLEAR the bit */
+ }
for (i = 0; i < priv->child_count; i++) {
if (local->child_up[i]) {
@@ -1555,17 +2939,16 @@ afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->fsync,
- fd, datasync);
+ fd, datasync, xdata);
if (!--call_count)
break;
}
}
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (fsync, frame, op_ret, op_errno, NULL, NULL);
- }
+ if (ret < 0)
+ AFR_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL, NULL);
return 0;
}
@@ -1575,10 +2958,10 @@ out:
int32_t
afr_fsyncdir_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *xdata)
{
afr_local_t *local = NULL;
-
int call_count = -1;
local = frame->local;
@@ -1596,7 +2979,7 @@ afr_fsyncdir_cbk (call_frame_t *frame, void *cookie,
if (call_count == 0)
AFR_STACK_UNWIND (fsyncdir, frame, local->op_ret,
- local->op_errno);
+ local->op_errno, xdata);
return 0;
}
@@ -1604,16 +2987,13 @@ afr_fsyncdir_cbk (call_frame_t *frame, void *cookie,
int32_t
afr_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int32_t datasync)
+ int32_t datasync, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
-
int ret = -1;
-
int i = 0;
int32_t call_count = 0;
- int32_t op_ret = -1;
int32_t op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
@@ -1622,33 +3002,30 @@ afr_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
priv = this->private;
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
goto out;
- }
call_count = local->call_count;
- frame->local = local;
for (i = 0; i < priv->child_count; i++) {
if (local->child_up[i]) {
STACK_WIND (frame, afr_fsyncdir_cbk,
priv->children[i],
priv->children[i]->fops->fsyncdir,
- fd, datasync);
+ fd, datasync, xdata);
if (!--call_count)
break;
}
}
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (fsyncdir, frame, op_ret, op_errno);
- }
+ if (ret < 0)
+ AFR_STACK_UNWIND (fsyncdir, frame, -1, op_errno, NULL);
return 0;
}
@@ -1659,18 +3036,20 @@ out:
int32_t
afr_xattrop_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xattr)
+ dict_t *xattr, dict_t *xdata)
{
afr_local_t *local = NULL;
-
int call_count = -1;
local = frame->local;
LOCK (&frame->lock);
{
- if (op_ret == 0)
+ if (op_ret == 0) {
+ if (!local->cont.xattrop.xattr)
+ local->cont.xattrop.xattr = dict_ref (xattr);
local->op_ret = 0;
+ }
local->op_errno = op_errno;
}
@@ -1680,7 +3059,7 @@ afr_xattrop_cbk (call_frame_t *frame, void *cookie,
if (call_count == 0)
AFR_STACK_UNWIND (xattrop, frame, local->op_ret, local->op_errno,
- xattr);
+ local->cont.xattrop.xattr, xdata);
return 0;
}
@@ -1688,16 +3067,13 @@ afr_xattrop_cbk (call_frame_t *frame, void *cookie,
int32_t
afr_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t optype, dict_t *xattr)
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
-
int ret = -1;
-
int i = 0;
int32_t call_count = 0;
- int32_t op_ret = -1;
int32_t op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
@@ -1706,33 +3082,30 @@ afr_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
priv = this->private;
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
goto out;
- }
call_count = local->call_count;
- frame->local = local;
for (i = 0; i < priv->child_count; i++) {
if (local->child_up[i]) {
STACK_WIND (frame, afr_xattrop_cbk,
priv->children[i],
priv->children[i]->fops->xattrop,
- loc, optype, xattr);
+ loc, optype, xattr, xdata);
if (!--call_count)
break;
}
}
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (xattrop, frame, op_ret, op_errno, NULL);
- }
+ if (ret < 0)
+ AFR_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL, NULL);
return 0;
}
@@ -1743,7 +3116,7 @@ out:
int32_t
afr_fxattrop_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xattr)
+ dict_t *xattr, dict_t *xdata)
{
afr_local_t *local = NULL;
@@ -1753,8 +3126,12 @@ afr_fxattrop_cbk (call_frame_t *frame, void *cookie,
LOCK (&frame->lock);
{
- if (op_ret == 0)
+ if (op_ret == 0) {
+ if (!local->cont.fxattrop.xattr)
+ local->cont.fxattrop.xattr = dict_ref (xattr);
+
local->op_ret = 0;
+ }
local->op_errno = op_errno;
}
@@ -1764,7 +3141,7 @@ afr_fxattrop_cbk (call_frame_t *frame, void *cookie,
if (call_count == 0)
AFR_STACK_UNWIND (fxattrop, frame, local->op_ret, local->op_errno,
- xattr);
+ local->cont.fxattrop.xattr, xdata);
return 0;
}
@@ -1772,16 +3149,13 @@ afr_fxattrop_cbk (call_frame_t *frame, void *cookie,
int32_t
afr_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
- gf_xattrop_flags_t optype, dict_t *xattr)
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
-
int ret = -1;
-
int i = 0;
int32_t call_count = 0;
- int32_t op_ret = -1;
int32_t op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
@@ -1790,33 +3164,30 @@ afr_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
priv = this->private;
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
goto out;
- }
call_count = local->call_count;
- frame->local = local;
for (i = 0; i < priv->child_count; i++) {
if (local->child_up[i]) {
STACK_WIND (frame, afr_fxattrop_cbk,
priv->children[i],
priv->children[i]->fops->fxattrop,
- fd, optype, xattr);
+ fd, optype, xattr, xdata);
if (!--call_count)
break;
}
}
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, NULL);
- }
+ if (ret < 0)
+ AFR_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL, NULL);
return 0;
}
@@ -1825,11 +3196,10 @@ out:
int32_t
afr_inodelk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
+ xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_local_t *local = NULL;
-
int call_count = -1;
local = frame->local;
@@ -1847,7 +3217,7 @@ afr_inodelk_cbk (call_frame_t *frame, void *cookie,
if (call_count == 0)
AFR_STACK_UNWIND (inodelk, frame, local->op_ret,
- local->op_errno);
+ local->op_errno, xdata);
return 0;
}
@@ -1855,16 +3225,14 @@ afr_inodelk_cbk (call_frame_t *frame, void *cookie,
int32_t
afr_inodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, int32_t cmd, struct flock *flock)
+ const char *volume, loc_t *loc, int32_t cmd,
+ struct gf_flock *flock, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
-
int ret = -1;
-
int i = 0;
int32_t call_count = 0;
- int32_t op_ret = -1;
int32_t op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
@@ -1873,45 +3241,42 @@ afr_inodelk (call_frame_t *frame, xlator_t *this,
priv = this->private;
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
goto out;
- }
call_count = local->call_count;
- frame->local = local;
for (i = 0; i < priv->child_count; i++) {
if (local->child_up[i]) {
STACK_WIND (frame, afr_inodelk_cbk,
priv->children[i],
priv->children[i]->fops->inodelk,
- volume, loc, cmd, flock);
+ volume, loc, cmd, flock, xdata);
if (!--call_count)
break;
}
}
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (inodelk, frame, op_ret, op_errno);
- }
+ if (ret < 0)
+ AFR_STACK_UNWIND (inodelk, frame, -1, op_errno, NULL);
return 0;
}
int32_t
afr_finodelk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *xdata)
{
afr_local_t *local = NULL;
-
int call_count = -1;
local = frame->local;
@@ -1929,7 +3294,7 @@ afr_finodelk_cbk (call_frame_t *frame, void *cookie,
if (call_count == 0)
AFR_STACK_UNWIND (finodelk, frame, local->op_ret,
- local->op_errno);
+ local->op_errno, xdata);
return 0;
}
@@ -1937,16 +3302,14 @@ afr_finodelk_cbk (call_frame_t *frame, void *cookie,
int32_t
afr_finodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, fd_t *fd, int32_t cmd, struct flock *flock)
+ const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *flock,
+ dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
-
int ret = -1;
-
int i = 0;
int32_t call_count = 0;
- int32_t op_ret = -1;
int32_t op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
@@ -1955,45 +3318,40 @@ afr_finodelk (call_frame_t *frame, xlator_t *this,
priv = this->private;
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
goto out;
- }
call_count = local->call_count;
- frame->local = local;
for (i = 0; i < priv->child_count; i++) {
if (local->child_up[i]) {
STACK_WIND (frame, afr_finodelk_cbk,
priv->children[i],
priv->children[i]->fops->finodelk,
- volume, fd, cmd, flock);
+ volume, fd, cmd, flock, xdata);
if (!--call_count)
break;
}
}
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (finodelk, frame, op_ret, op_errno);
- }
+ if (ret < 0)
+ AFR_STACK_UNWIND (finodelk, frame, -1, op_errno, NULL);
return 0;
}
int32_t
-afr_entrylk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
-
+afr_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_local_t *local = NULL;
-
int call_count = -1;
local = frame->local;
@@ -2011,7 +3369,7 @@ afr_entrylk_cbk (call_frame_t *frame, void *cookie,
if (call_count == 0)
AFR_STACK_UNWIND (entrylk, frame, local->op_ret,
- local->op_errno);
+ local->op_errno, xdata);
return 0;
}
@@ -2020,16 +3378,14 @@ afr_entrylk_cbk (call_frame_t *frame, void *cookie,
int32_t
afr_entrylk (call_frame_t *frame, xlator_t *this,
const char *volume, loc_t *loc,
- const char *basename, entrylk_cmd cmd, entrylk_type type)
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
-
int ret = -1;
-
int i = 0;
int32_t call_count = 0;
- int32_t op_ret = -1;
int32_t op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
@@ -2038,34 +3394,31 @@ afr_entrylk (call_frame_t *frame, xlator_t *this,
priv = this->private;
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
goto out;
- }
call_count = local->call_count;
- frame->local = local;
for (i = 0; i < priv->child_count; i++) {
if (local->child_up[i]) {
STACK_WIND (frame, afr_entrylk_cbk,
priv->children[i],
priv->children[i]->fops->entrylk,
- volume, loc, basename, cmd, type);
+ volume, loc, basename, cmd, type, xdata);
if (!--call_count)
break;
}
}
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (entrylk, frame, op_ret, op_errno);
- }
+ if (ret < 0)
+ AFR_STACK_UNWIND (entrylk, frame, -1, op_errno, NULL);
return 0;
}
@@ -2073,11 +3426,10 @@ out:
int32_t
afr_fentrylk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
+ xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_local_t *local = NULL;
-
int call_count = -1;
local = frame->local;
@@ -2095,7 +3447,7 @@ afr_fentrylk_cbk (call_frame_t *frame, void *cookie,
if (call_count == 0)
AFR_STACK_UNWIND (fentrylk, frame, local->op_ret,
- local->op_errno);
+ local->op_errno, xdata);
return 0;
}
@@ -2104,16 +3456,14 @@ afr_fentrylk_cbk (call_frame_t *frame, void *cookie,
int32_t
afr_fentrylk (call_frame_t *frame, xlator_t *this,
const char *volume, fd_t *fd,
- const char *basename, entrylk_cmd cmd, entrylk_type type)
+ const char *basename, entrylk_cmd cmd,
+ entrylk_type type, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
-
int ret = -1;
-
int i = 0;
int32_t call_count = 0;
- int32_t op_ret = -1;
int32_t op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
@@ -2122,44 +3472,40 @@ afr_fentrylk (call_frame_t *frame, xlator_t *this,
priv = this->private;
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
goto out;
- }
call_count = local->call_count;
- frame->local = local;
for (i = 0; i < priv->child_count; i++) {
if (local->child_up[i]) {
STACK_WIND (frame, afr_fentrylk_cbk,
priv->children[i],
priv->children[i]->fops->fentrylk,
- volume, fd, basename, cmd, type);
+ volume, fd, basename, cmd, type, xdata);
if (!--call_count)
break;
}
}
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (fentrylk, frame, op_ret, op_errno);
- }
+ if (ret < 0)
+ AFR_STACK_UNWIND (fentrylk, frame, -1, op_errno, NULL);
return 0;
}
int32_t
afr_statfs_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct statvfs *statvfs)
+ struct statvfs *statvfs, dict_t *xdata)
{
afr_local_t *local = NULL;
-
int call_count = 0;
LOCK (&frame->lock);
@@ -2188,7 +3534,7 @@ afr_statfs_cbk (call_frame_t *frame, void *cookie,
if (call_count == 0)
AFR_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno,
- &local->cont.statfs.buf);
+ &local->cont.statfs.buf, xdata);
return 0;
}
@@ -2196,16 +3542,14 @@ afr_statfs_cbk (call_frame_t *frame, void *cookie,
int32_t
afr_statfs (call_frame_t *frame, xlator_t *this,
- loc_t *loc)
+ loc_t *loc, dict_t *xdata)
{
afr_private_t * priv = NULL;
int child_count = 0;
afr_local_t * local = NULL;
int i = 0;
-
- int ret = -1;
+ int ret = -1;
int call_count = 0;
- int32_t op_ret = -1;
int32_t op_errno = 0;
VALIDATE_OR_GOTO (this, out);
@@ -2215,15 +3559,13 @@ afr_statfs (call_frame_t *frame, xlator_t *this,
priv = this->private;
child_count = priv->child_count;
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
goto out;
- }
- frame->local = local;
call_count = local->call_count;
for (i = 0; i < child_count; i++) {
@@ -2231,27 +3573,26 @@ afr_statfs (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, afr_statfs_cbk,
priv->children[i],
priv->children[i]->fops->statfs,
- loc);
+ loc, xdata);
if (!--call_count)
break;
}
}
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (statfs, frame, op_ret, op_errno, NULL);
- }
+ if (ret < 0)
+ AFR_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
return 0;
}
int32_t
afr_lk_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct flock *lock)
+ int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+ dict_t *xdata)
{
afr_local_t * local = NULL;
-
int call_count = -1;
local = frame->local;
@@ -2259,7 +3600,7 @@ afr_lk_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
if (call_count == 0)
AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
- lock);
+ lock, xdata);
return 0;
}
@@ -2270,8 +3611,7 @@ afr_lk_unlock (call_frame_t *frame, xlator_t *this)
{
afr_local_t * local = NULL;
afr_private_t * priv = NULL;
-
- int i;
+ int i = 0;
int call_count = 0;
local = frame->local;
@@ -2282,7 +3622,7 @@ afr_lk_unlock (call_frame_t *frame, xlator_t *this)
if (call_count == 0) {
AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
- &local->cont.lk.ret_flock);
+ &local->cont.lk.ret_flock, NULL);
return 0;
}
@@ -2296,7 +3636,7 @@ afr_lk_unlock (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->lk,
local->fd, F_SETLK,
- &local->cont.lk.user_flock);
+ &local->cont.lk.user_flock, NULL);
if (!--call_count)
break;
@@ -2309,13 +3649,13 @@ afr_lk_unlock (call_frame_t *frame, xlator_t *this)
int32_t
afr_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct flock *lock)
+ int32_t op_ret, int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int ret = 0;
-
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
int child_index = -1;
+/* int ret = 0; */
+
local = frame->local;
priv = this->private;
@@ -2344,28 +3684,30 @@ afr_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
priv->children[child_index],
priv->children[child_index]->fops->lk,
local->fd, local->cont.lk.cmd,
- &local->cont.lk.user_flock);
+ &local->cont.lk.user_flock, xdata);
} else if (local->op_ret == -1) {
/* all nodes have gone down */
AFR_STACK_UNWIND (lk, frame, -1, ENOTCONN,
- &local->cont.lk.ret_flock);
+ &local->cont.lk.ret_flock, NULL);
} else {
/* locking has succeeded on all nodes that are up */
- ret = afr_mark_locked_nodes (this, local->fd,
- local->cont.lk.locked_nodes);
- if (ret)
- gf_log (this->name, GF_LOG_DEBUG,
- "Could not save locked nodes info in fdctx");
+ /* temporarily
+ ret = afr_mark_locked_nodes (this, local->fd,
+ local->cont.lk.locked_nodes);
+ if (ret)
+ gf_log (this->name, GF_LOG_DEBUG,
+ "Could not save locked nodes info in fdctx");
- ret = afr_save_locked_fd (this, local->fd);
- if (ret)
- gf_log (this->name, GF_LOG_DEBUG,
- "Could not save locked fd");
+ ret = afr_save_locked_fd (this, local->fd);
+ if (ret)
+ gf_log (this->name, GF_LOG_DEBUG,
+ "Could not save locked fd");
- AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
- &local->cont.lk.ret_flock);
+ */
+ AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
+ &local->cont.lk.ret_flock, NULL);
}
return 0;
@@ -2374,16 +3716,13 @@ afr_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int
afr_lk (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int32_t cmd,
- struct flock *flock)
+ fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
-
int i = 0;
-
- int32_t op_ret = -1;
int32_t op_errno = 0;
+ int ret = -1;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -2391,17 +3730,18 @@ afr_lk (call_frame_t *frame, xlator_t *this,
priv = this->private;
- ALLOC_OR_GOTO (local, afr_local_t, out);
- AFR_LOCAL_INIT (local, priv);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- frame->local = local;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
local->cont.lk.locked_nodes = GF_CALLOC (priv->child_count,
- sizeof (*local->cont.lk.locked_nodes),
- gf_afr_mt_char);
+ sizeof (*local->cont.lk.locked_nodes),
+ gf_afr_mt_char);
if (!local->cont.lk.locked_nodes) {
- gf_log (this->name, GF_LOG_ERROR, "Out of memory");
op_errno = ENOMEM;
goto out;
}
@@ -2414,13 +3754,30 @@ afr_lk (call_frame_t *frame, xlator_t *this,
STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) 0,
priv->children[i],
priv->children[i]->fops->lk,
- fd, cmd, flock);
+ fd, cmd, flock, xdata);
- op_ret = 0;
+ ret = 0;
+out:
+ if (ret < 0)
+ AFR_STACK_UNWIND (lk, frame, -1, op_errno, NULL, NULL);
+ return 0;
+}
+
+int
+afr_forget (xlator_t *this, inode_t *inode)
+{
+ uint64_t ctx_addr = 0;
+ afr_inode_ctx_t *ctx = NULL;
+
+ inode_ctx_get (inode, this, &ctx_addr);
+
+ if (!ctx_addr)
+ goto out;
+
+ ctx = (afr_inode_ctx_t *)(long)ctx_addr;
+ GF_FREE (ctx->fresh_children);
+ GF_FREE (ctx);
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (lk, frame, op_ret, op_errno, NULL);
- }
return 0;
}
@@ -2433,47 +3790,29 @@ afr_priv_dump (xlator_t *this)
int i = 0;
- assert(this);
+ GF_ASSERT (this);
priv = this->private;
- assert(priv);
+ GF_ASSERT (priv);
snprintf(key_prefix, GF_DUMP_MAX_BUF_LEN, "%s.%s", this->type, this->name);
gf_proc_dump_add_section(key_prefix);
- gf_proc_dump_build_key(key, key_prefix, "child_count");
- gf_proc_dump_write(key, "%u", priv->child_count);
- gf_proc_dump_build_key(key, key_prefix, "read_child_rr");
- gf_proc_dump_write(key, "%u", priv->read_child_rr);
+ gf_proc_dump_write("child_count", "%u", priv->child_count);
+ gf_proc_dump_write("read_child_rr", "%u", priv->read_child_rr);
for (i = 0; i < priv->child_count; i++) {
- gf_proc_dump_build_key(key, key_prefix, "child_up[%d]", i);
+ sprintf (key, "child_up[%d]", i);
gf_proc_dump_write(key, "%d", priv->child_up[i]);
- gf_proc_dump_build_key(key, key_prefix,
- "pending_key[%d]", i);
+ sprintf (key, "pending_key[%d]", i);
gf_proc_dump_write(key, "%s", priv->pending_key[i]);
}
- gf_proc_dump_build_key(key, key_prefix, "data_self_heal");
- gf_proc_dump_write(key, "%d", priv->data_self_heal);
- gf_proc_dump_build_key(key, key_prefix, "metadata_self_heal");
- gf_proc_dump_write(key, "%d", priv->metadata_self_heal);
- gf_proc_dump_build_key(key, key_prefix, "entry_self_heal");
- gf_proc_dump_write(key, "%d", priv->entry_self_heal);
- gf_proc_dump_build_key(key, key_prefix, "data_change_log");
- gf_proc_dump_write(key, "%d", priv->data_change_log);
- gf_proc_dump_build_key(key, key_prefix, "metadata_change_log");
- gf_proc_dump_write(key, "%d", priv->metadata_change_log);
- gf_proc_dump_build_key(key, key_prefix, "entry_change_log");
- gf_proc_dump_write(key, "%d", priv->entry_change_log);
- gf_proc_dump_build_key(key, key_prefix, "read_child");
- gf_proc_dump_write(key, "%d", priv->read_child);
- gf_proc_dump_build_key(key, key_prefix, "favorite_child");
- gf_proc_dump_write(key, "%u", priv->favorite_child);
- gf_proc_dump_build_key(key, key_prefix, "data_lock_server_count");
- gf_proc_dump_write(key, "%u", priv->data_lock_server_count);
- gf_proc_dump_build_key(key, key_prefix, "metadata_lock_server_count");
- gf_proc_dump_write(key, "%u", priv->metadata_lock_server_count);
- gf_proc_dump_build_key(key, key_prefix, "entry_lock_server_count");
- gf_proc_dump_write(key, "%u", priv->entry_lock_server_count);
- gf_proc_dump_build_key(key, key_prefix, "wait_count");
- gf_proc_dump_write(key, "%u", priv->wait_count);
+ gf_proc_dump_write("data_self_heal", "%s", priv->data_self_heal);
+ gf_proc_dump_write("metadata_self_heal", "%d", priv->metadata_self_heal);
+ gf_proc_dump_write("entry_self_heal", "%d", priv->entry_self_heal);
+ gf_proc_dump_write("data_change_log", "%d", priv->data_change_log);
+ gf_proc_dump_write("metadata_change_log", "%d", priv->metadata_change_log);
+ gf_proc_dump_write("entry-change_log", "%d", priv->entry_change_log);
+ gf_proc_dump_write("read_child", "%d", priv->read_child);
+ gf_proc_dump_write("favorite_child", "%d", priv->favorite_child);
+ gf_proc_dump_write("wait_count", "%u", priv->wait_count);
return 0;
}
@@ -2489,7 +3828,6 @@ static int
find_child_index (xlator_t *this, xlator_t *child)
{
afr_private_t *priv = NULL;
-
int i = -1;
priv = this->private;
@@ -2504,86 +3842,750 @@ find_child_index (xlator_t *this, xlator_t *child)
int32_t
afr_notify (xlator_t *this, int32_t event,
- void *data, ...)
+ void *data, void *data2)
{
- afr_private_t * priv = NULL;
- unsigned char * child_up = NULL;
-
- int i = -1;
- int up_children = 0;
+ afr_private_t *priv = NULL;
+ int i = -1;
+ int up_children = 0;
+ int down_children = 0;
+ int propagate = 0;
+ int had_heard_from_all = 0;
+ int have_heard_from_all = 0;
+ int idx = -1;
+ int ret = -1;
+ int call_psh = 0;
+ int up_child = AFR_ALL_CHILDREN;
+ dict_t *input = NULL;
+ dict_t *output = NULL;
priv = this->private;
if (!priv)
return 0;
- child_up = priv->child_up;
+ /*
+ * We need to reset this in case children come up in "staggered"
+ * fashion, so that we discover a late-arriving local subvolume. Note
+ * that we could end up issuing N lookups to the first subvolume, and
+ * O(N^2) overall, but N is small for AFR so it shouldn't be an issue.
+ */
+ priv->did_discovery = _gf_false;
+
+ had_heard_from_all = 1;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!priv->last_event[i]) {
+ had_heard_from_all = 0;
+ }
+ }
+
+ /* parent xlators dont need to know about every child_up, child_down
+ * because of afr ha. If all subvolumes go down, child_down has
+ * to be triggered. In that state when 1 subvolume comes up child_up
+ * needs to be triggered. dht optimizes revalidate lookup by sending
+ * it only to one of its subvolumes. When child up/down happens
+ * for afr's subvolumes dht should be notified by child_modified. The
+ * subsequent revalidate lookup happens on all the dht's subvolumes
+ * which triggers afr self-heals if any.
+ */
+ idx = find_child_index (this, data);
+ if (idx < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "Received child_up "
+ "from invalid subvolume");
+ goto out;
+ }
switch (event) {
case GF_EVENT_CHILD_UP:
- i = find_child_index (this, data);
+ LOCK (&priv->lock);
+ {
+ /*
+ * This only really counts if the child was never up
+ * (value = -1) or had been down (value = 0). See
+ * comment at GF_EVENT_CHILD_DOWN for a more detailed
+ * explanation.
+ */
+ if (priv->child_up[idx] != 1) {
+ priv->up_count++;
+ }
+ priv->child_up[idx] = 1;
+
+ call_psh = 1;
+ up_child = idx;
+ for (i = 0; i < priv->child_count; i++)
+ if (priv->child_up[i] == 1)
+ up_children++;
+ if (up_children == 1) {
+ gf_log (this->name, GF_LOG_INFO,
+ "Subvolume '%s' came back up; "
+ "going online.", ((xlator_t *)data)->name);
+ } else {
+ event = GF_EVENT_CHILD_MODIFIED;
+ }
+
+ priv->last_event[idx] = event;
+ }
+ UNLOCK (&priv->lock);
- child_up[i] = 1;
+ break;
+ case GF_EVENT_CHILD_DOWN:
LOCK (&priv->lock);
{
- priv->up_count++;
+ /*
+ * If a brick is down when we start, we'll get a
+ * CHILD_DOWN to indicate its initial state. There
+ * was never a CHILD_UP in this case, so if we
+ * increment "down_count" the difference between than
+ * and "up_count" will no longer be the number of
+ * children that are currently up. This has serious
+ * implications e.g. for quorum enforcement, so we
+ * don't increment these values unless the event
+ * represents an actual state transition between "up"
+ * (value = 1) and anything else.
+ */
+ if (priv->child_up[idx] == 1) {
+ priv->down_count++;
+ }
+ priv->child_up[idx] = 0;
+
+ for (i = 0; i < priv->child_count; i++)
+ if (priv->child_up[i] == 0)
+ down_children++;
+ if (down_children == priv->child_count) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "All subvolumes are down. Going offline "
+ "until atleast one of them comes back up.");
+ } else {
+ event = GF_EVENT_CHILD_MODIFIED;
+ }
+
+ priv->last_event[idx] = event;
}
UNLOCK (&priv->lock);
- /*
- if all the children were down, and one child came up,
- send notify to parent
- */
+ break;
- for (i = 0; i < priv->child_count; i++)
- if (child_up[i])
- up_children++;
+ case GF_EVENT_CHILD_CONNECTING:
+ LOCK (&priv->lock);
+ {
+ priv->last_event[idx] = event;
+ }
+ UNLOCK (&priv->lock);
- if (up_children == 1) {
- gf_log (this->name, GF_LOG_NORMAL,
- "Subvolume '%s' came back up; "
- "going online.", ((xlator_t *)data)->name);
+ break;
- default_notify (this, event, data);
- }
+ case GF_EVENT_TRANSLATOR_OP:
+ input = data;
+ output = data2;
+ ret = afr_xl_op (this, input, output);
+ goto out;
+ break;
+ default:
+ propagate = 1;
break;
+ }
- case GF_EVENT_CHILD_DOWN:
- i = find_child_index (this, data);
+ /* have all subvolumes reported status once by now? */
+ have_heard_from_all = 1;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!priv->last_event[i])
+ have_heard_from_all = 0;
+ }
- child_up[i] = 0;
+ /* if all subvols have reported status, no need to hide anything
+ or wait for anything else. Just propagate blindly */
+ if (have_heard_from_all)
+ propagate = 1;
+
+ if (!had_heard_from_all && have_heard_from_all) {
+ /* This is the first event which completes aggregation
+ of events from all subvolumes. If at least one subvol
+ had come up, propagate CHILD_UP, but only this time
+ */
+ event = GF_EVENT_CHILD_DOWN;
LOCK (&priv->lock);
{
- priv->down_count++;
+ up_children = afr_up_children_count (priv->child_up,
+ priv->child_count);
+ for (i = 0; i < priv->child_count; i++) {
+ if (priv->last_event[i] == GF_EVENT_CHILD_UP) {
+ event = GF_EVENT_CHILD_UP;
+ break;
+ }
+
+ if (priv->last_event[i] ==
+ GF_EVENT_CHILD_CONNECTING) {
+ event = GF_EVENT_CHILD_CONNECTING;
+ /* continue to check other events for CHILD_UP */
+ }
+ }
}
UNLOCK (&priv->lock);
+ }
- /*
- if all children are down, and this was the last to go down,
- send notify to parent
- */
+ ret = 0;
+ if (propagate)
+ ret = default_notify (this, event, data);
+ if (call_psh && priv->shd.iamshd)
+ afr_proactive_self_heal ((void*) (long) up_child);
+
+out:
+ return ret;
+}
+
+int
+afr_first_up_child (unsigned char *child_up, size_t child_count)
+{
+ int ret = -1;
+ int i = 0;
+
+ GF_ASSERT (child_up);
+
+ for (i = 0; i < child_count; i++) {
+ if (child_up[i]) {
+ ret = i;
+ break;
+ }
+ }
+
+ return ret;
+}
+
+int
+afr_local_init (afr_local_t *local, afr_private_t *priv, int32_t *op_errno)
+{
+ int ret = -1;
+
+ local->op_ret = -1;
+ local->op_errno = EUCLEAN;
+
+ local->child_up = GF_CALLOC (priv->child_count,
+ sizeof (*local->child_up),
+ gf_afr_mt_char);
+ if (!local->child_up) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ memcpy (local->child_up, priv->child_up,
+ sizeof (*local->child_up) * priv->child_count);
+ local->call_count = afr_up_children_count (local->child_up,
+ priv->child_count);
+ if (local->call_count == 0) {
+ gf_log (THIS->name, GF_LOG_INFO, "no subvolumes up");
+ if (op_errno)
+ *op_errno = ENOTCONN;
+ goto out;
+ }
+
+ local->child_errno = GF_CALLOC (priv->child_count,
+ sizeof (*local->child_errno),
+ gf_afr_mt_int32_t);
+ if (!local->child_errno) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ local->transaction.postop_piggybacked = GF_CALLOC (priv->child_count,
+ sizeof (int),
+ gf_afr_mt_int32_t);
+ if (!local->transaction.postop_piggybacked) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ local->append_write = _gf_false;
+
+ ret = 0;
+out:
+ return ret;
+}
+
+int
+afr_internal_lock_init (afr_internal_lock_t *lk, size_t child_count,
+ transaction_lk_type_t lk_type)
+{
+ int ret = -ENOMEM;
+
+ lk->locked_nodes = GF_CALLOC (sizeof (*lk->locked_nodes),
+ child_count, gf_afr_mt_char);
+ if (NULL == lk->locked_nodes)
+ goto out;
+
+ lk->lower_locked_nodes = GF_CALLOC (sizeof (*lk->lower_locked_nodes),
+ child_count, gf_afr_mt_char);
+ if (NULL == lk->lower_locked_nodes)
+ goto out;
+
+ lk->lock_op_ret = -1;
+ lk->lock_op_errno = EUCLEAN;
+ lk->transaction_lk_type = lk_type;
+
+ ret = 0;
+out:
+ return ret;
+}
+
+void
+afr_matrix_cleanup (int32_t **matrix, unsigned int m)
+{
+ int i = 0;
+
+ if (!matrix)
+ goto out;
+ for (i = 0; i < m; i++) {
+ GF_FREE (matrix[i]);
+ }
+
+ GF_FREE (matrix);
+out:
+ return;
+}
+
+int32_t**
+afr_matrix_create (unsigned int m, unsigned int n)
+{
+ int32_t **matrix = NULL;
+ int i = 0;
+
+ matrix = GF_CALLOC (sizeof (*matrix), m, gf_afr_mt_int32_t);
+ if (!matrix)
+ goto out;
+
+ for (i = 0; i < m; i++) {
+ matrix[i] = GF_CALLOC (sizeof (*matrix[i]), n,
+ gf_afr_mt_int32_t);
+ if (!matrix[i])
+ goto out;
+ }
+ return matrix;
+out:
+ afr_matrix_cleanup (matrix, m);
+ return NULL;
+}
+
+int
+afr_inodelk_init (afr_inodelk_t *lk, char *dom, size_t child_count)
+{
+ int ret = -ENOMEM;
+
+ lk->domain = dom;
+ lk->locked_nodes = GF_CALLOC (sizeof (*lk->locked_nodes),
+ child_count, gf_afr_mt_char);
+ if (NULL == lk->locked_nodes)
+ goto out;
+ ret = 0;
+out:
+ return ret;
+}
+
+int
+afr_transaction_local_init (afr_local_t *local, xlator_t *this)
+{
+ int child_up_count = 0;
+ int ret = -ENOMEM;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ ret = afr_internal_lock_init (&local->internal_lock, priv->child_count,
+ AFR_TRANSACTION_LK);
+ if (ret < 0)
+ goto out;
+
+ if ((local->transaction.type == AFR_DATA_TRANSACTION) ||
+ (local->transaction.type == AFR_METADATA_TRANSACTION)) {
+ ret = afr_inodelk_init (&local->internal_lock.inodelk[0],
+ this->name, priv->child_count);
+ if (ret < 0)
+ goto out;
+ }
+
+ ret = -ENOMEM;
+ child_up_count = afr_up_children_count (local->child_up,
+ priv->child_count);
+ if (priv->optimistic_change_log && child_up_count == priv->child_count)
+ local->optimistic_change_log = 1;
+
+ local->first_up_child = afr_first_up_child (local->child_up,
+ priv->child_count);
+
+ local->transaction.eager_lock =
+ GF_CALLOC (sizeof (*local->transaction.eager_lock),
+ priv->child_count,
+ gf_afr_mt_int32_t);
+
+ if (!local->transaction.eager_lock)
+ goto out;
+
+ local->fresh_children = afr_children_create (priv->child_count);
+ if (!local->fresh_children)
+ goto out;
+
+ local->transaction.pre_op = GF_CALLOC (sizeof (*local->transaction.pre_op),
+ priv->child_count,
+ gf_afr_mt_char);
+ if (!local->transaction.pre_op)
+ goto out;
+
+ local->pending = afr_matrix_create (priv->child_count,
+ AFR_NUM_CHANGE_LOGS);
+ if (!local->pending)
+ goto out;
+
+ local->transaction.txn_changelog = afr_matrix_create (priv->child_count,
+ AFR_NUM_CHANGE_LOGS);
+ if (!local->transaction.txn_changelog)
+ goto out;
+
+ INIT_LIST_HEAD (&local->transaction.eager_locked);
+
+ ret = 0;
+out:
+ return ret;
+}
+
+void
+afr_reset_children (int32_t *fresh_children, int32_t child_count)
+{
+ unsigned int i = 0;
+ for (i = 0; i < child_count; i++)
+ fresh_children[i] = -1;
+}
+
+int32_t*
+afr_children_create (int32_t child_count)
+{
+ int32_t *children = NULL;
+ int i = 0;
+
+ GF_ASSERT (child_count > 0);
+
+ children = GF_CALLOC (child_count, sizeof (*children),
+ gf_afr_mt_int32_t);
+ if (NULL == children)
+ goto out;
+ for (i = 0; i < child_count; i++)
+ children[i] = -1;
+out:
+ return children;
+}
+
+void
+afr_children_add_child (int32_t *children, int32_t child,
+ int32_t child_count)
+{
+ gf_boolean_t child_found = _gf_false;
+ int i = 0;
+
+ for (i = 0; i < child_count; i++) {
+ if (children[i] == -1)
+ break;
+ if (children[i] == child) {
+ child_found = _gf_true;
+ break;
+ }
+ }
+
+ if (!child_found) {
+ GF_ASSERT (i < child_count);
+ children[i] = child;
+ }
+}
+
+void
+afr_children_rm_child (int32_t *children, int32_t child, int32_t child_count)
+{
+ int i = 0;
+
+ GF_ASSERT ((child >= 0) && (child < child_count));
+ for (i = 0; i < child_count; i++) {
+ if (children[i] == -1)
+ break;
+ if (children[i] == child) {
+ if (i != (child_count - 1))
+ memmove (children + i, children + i + 1,
+ sizeof (*children)*(child_count - i - 1));
+ children[child_count - 1] = -1;
+ break;
+ }
+ }
+}
+
+int
+afr_get_children_count (int32_t *children, unsigned int child_count)
+{
+ int count = 0;
+ int i = 0;
+
+ for (i = 0; i < child_count; i++) {
+ if (children[i] == -1)
+ break;
+ count++;
+ }
+ return count;
+}
+
+void
+afr_set_low_priority (call_frame_t *frame)
+{
+ frame->root->pid = LOW_PRIO_PROC_PID;
+}
+
+int
+afr_child_fd_ctx_set (xlator_t *this, fd_t *fd, int32_t child,
+ int flags)
+{
+ int ret = 0;
+ uint64_t ctx = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+
+ GF_ASSERT (fd && fd->inode);
+ ret = afr_fd_ctx_set (this, fd);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "could not set fd ctx for fd=%p", fd);
+ goto out;
+ }
+
+ ret = fd_ctx_get (fd, this, &ctx);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "could not get fd ctx for fd=%p", fd);
+ goto out;
+ }
+
+ fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+ fd_ctx->opened_on[child] = AFR_FD_OPENED;
+ if (!IA_ISDIR (fd->inode->ia_type)) {
+ fd_ctx->flags = flags;
+ }
+ ret = 0;
+out:
+ return ret;
+}
+
+gf_boolean_t
+afr_have_quorum (char *logname, afr_private_t *priv)
+{
+ unsigned int quorum = 0;
+
+ GF_VALIDATE_OR_GOTO(logname,priv,out);
+
+ quorum = priv->quorum_count;
+ if (quorum != AFR_QUORUM_AUTO) {
+ return (priv->up_count >= (priv->down_count + quorum));
+ }
+
+ quorum = priv->child_count / 2 + 1;
+ if (priv->up_count >= (priv->down_count + quorum)) {
+ return _gf_true;
+ }
+
+ /*
+ * Special case for even numbers of nodes: if we have exactly half
+ * and that includes the first ("senior-most") node, then that counts
+ * as quorum even if it wouldn't otherwise. This supports e.g. N=2
+ * while preserving the critical property that there can only be one
+ * such group.
+ */
+ if ((priv->child_count % 2) == 0) {
+ quorum = priv->child_count / 2;
+ if (priv->up_count >= (priv->down_count + quorum)) {
+ if (priv->child_up[0]) {
+ return _gf_true;
+ }
+ }
+ }
+
+out:
+ return _gf_false;
+}
+
+void
+afr_priv_destroy (afr_private_t *priv)
+{
+ int i = 0;
+
+ if (!priv)
+ goto out;
+ inode_unref (priv->root_inode);
+ GF_FREE (priv->shd.pos);
+ GF_FREE (priv->shd.pending);
+ GF_FREE (priv->shd.inprogress);
+// for (i = 0; i < priv->child_count; i++)
+// if (priv->shd.timer && priv->shd.timer[i])
+// gf_timer_call_cancel (this->ctx, priv->shd.timer[i]);
+ GF_FREE (priv->shd.timer);
+
+ if (priv->shd.healed)
+ eh_destroy (priv->shd.healed);
+
+ if (priv->shd.heal_failed)
+ eh_destroy (priv->shd.heal_failed);
+
+ if (priv->shd.split_brain)
+ eh_destroy (priv->shd.split_brain);
+
+ for (i = 0; i < priv->child_count; i++)
+ {
+ if (priv->shd.statistics[i])
+ eh_destroy (priv->shd.statistics[i]);
+ }
+
+ GF_FREE (priv->shd.statistics);
+ GF_FREE (priv->shd.crawl_events);
+
+ GF_FREE (priv->last_event);
+ if (priv->pending_key) {
for (i = 0; i < priv->child_count; i++)
- if (child_up[i])
- up_children++;
+ GF_FREE (priv->pending_key[i]);
+ }
+ GF_FREE (priv->pending_key);
+ GF_FREE (priv->children);
+ GF_FREE (priv->child_up);
+ LOCK_DESTROY (&priv->lock);
+ LOCK_DESTROY (&priv->read_child_lock);
+ pthread_mutex_destroy (&priv->mutex);
+ GF_FREE (priv);
+out:
+ return;
+}
+
+int
+xlator_subvolume_count (xlator_t *this)
+{
+ int i = 0;
+ xlator_list_t *list = NULL;
- if (up_children == 0) {
- gf_log (this->name, GF_LOG_ERROR,
- "All subvolumes are down. Going offline "
- "until atleast one of them comes back up.");
+ for (list = this->children; list; list = list->next)
+ i++;
+ return i;
+}
+
+inline gf_boolean_t
+afr_is_errno_set (int *child_errno, int child)
+{
+ return child_errno[child];
+}
- default_notify (this, event, data);
+inline gf_boolean_t
+afr_is_errno_unset (int *child_errno, int child)
+{
+ return !afr_is_errno_set (child_errno, child);
+}
+
+void
+afr_prepare_new_entry_pending_matrix (int32_t **pending,
+ gf_boolean_t (*is_pending) (int *, int),
+ int *ctx, struct iatt *buf,
+ unsigned int child_count)
+{
+ int midx = 0;
+ int idx = 0;
+ int i = 0;
+
+ midx = afr_index_for_transaction_type (AFR_METADATA_TRANSACTION);
+ if (IA_ISDIR (buf->ia_type))
+ idx = afr_index_for_transaction_type (AFR_ENTRY_TRANSACTION);
+ else if (IA_ISREG (buf->ia_type))
+ idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
+ else
+ idx = -1;
+ for (i = 0; i < child_count; i++) {
+ if (is_pending (ctx, i)) {
+ pending[i][midx] = hton32 (1);
+ if (idx == -1)
+ continue;
+ pending[i][idx] = hton32 (1);
}
+ }
+}
- break;
+gf_boolean_t
+afr_is_fd_fixable (fd_t *fd)
+{
+ if (!fd || !fd->inode)
+ return _gf_false;
+ else if (fd_is_anonymous (fd))
+ return _gf_false;
+ else if (uuid_is_null (fd->inode->gfid))
+ return _gf_false;
+
+ return _gf_true;
+}
- default:
- default_notify (this, event, data);
+void
+afr_handle_open_fd_count (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ inode_t *inode = NULL;
+ afr_inode_ctx_t *ctx = NULL;
+
+ local = frame->local;
+
+ if (local->fd)
+ inode = local->fd->inode;
+ else
+ inode = local->loc.inode;
+
+ if (!inode)
+ return;
+
+ LOCK (&inode->lock);
+ {
+ ctx = __afr_inode_ctx_get (inode, this);
+ ctx->open_fd_count = local->open_fd_count;
}
+ UNLOCK (&inode->lock);
+}
- return 0;
+int
+afr_initialise_statistics (xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ int ret = -1;
+ int i = 0;
+ int child_count = 0;
+ eh_t *stats_per_brick = NULL;
+ shd_crawl_event_t ***shd_crawl_events = NULL;
+ priv = this->private;
+
+ priv->shd.statistics = GF_CALLOC (sizeof(eh_t *), priv->child_count,
+ gf_common_mt_eh_t);
+ if (!priv->shd.statistics) {
+ ret = -1;
+ goto out;
+ }
+ child_count = priv->child_count;
+ for (i=0; i < child_count ; i++) {
+ stats_per_brick = eh_new (AFR_STATISTICS_HISTORY_SIZE,
+ _gf_false,
+ _destroy_crawl_event_data);
+ if (!stats_per_brick) {
+ ret = -1;
+ goto out;
+ }
+ priv->shd.statistics[i] = stats_per_brick;
+
+ }
+
+ shd_crawl_events = (shd_crawl_event_t***)(&priv->shd.crawl_events);
+ *shd_crawl_events = GF_CALLOC (sizeof(shd_crawl_event_t*),
+ priv->child_count,
+ gf_afr_mt_shd_crawl_event_t);
+
+ if (!priv->shd.crawl_events) {
+ ret = -1;
+ goto out;
+ }
+ ret = 0;
+out:
+ return ret;
}
diff --git a/xlators/cluster/afr/src/afr-dir-read.c b/xlators/cluster/afr/src/afr-dir-read.c
index c393ece98..689dd84e6 100644
--- a/xlators/cluster/afr/src/afr-dir-read.c
+++ b/xlators/cluster/afr/src/afr-dir-read.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -49,9 +40,9 @@
#include "afr-self-heal.h"
#include "afr-self-heal-common.h"
-
int
-afr_examine_dir_sh_unwind (call_frame_t *frame, xlator_t *this)
+afr_examine_dir_sh_unwind (call_frame_t *frame, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, int32_t sh_failed)
{
afr_local_t *local = NULL;
@@ -60,7 +51,7 @@ afr_examine_dir_sh_unwind (call_frame_t *frame, xlator_t *this)
afr_set_opendir_done (this, local->fd->inode);
AFR_STACK_UNWIND (opendir, frame, local->op_ret,
- local->op_errno, local->fd);
+ local->op_errno, local->fd, NULL);
return 0;
}
@@ -70,16 +61,19 @@ gf_boolean_t
__checksums_differ (uint32_t *checksum, int child_count,
unsigned char *child_up)
{
- int ret = _gf_false;
- int i = 0;
-
- uint32_t cksum;
-
- cksum = checksum[0];
+ int ret = _gf_false;
+ int i = 0;
+ uint32_t cksum = 0;
+ gf_boolean_t activate_check = _gf_false;
for (i = 0; i < child_count; i++) {
if (!child_up[i])
continue;
+ if (_gf_false == activate_check) {
+ cksum = checksum[i];
+ activate_check = _gf_true;
+ continue;
+ }
if (cksum != checksum[i]) {
ret = _gf_true;
@@ -96,40 +90,45 @@ __checksums_differ (uint32_t *checksum, int child_count,
int32_t
afr_examine_dir_readdir_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- gf_dirent_t *entries)
+ gf_dirent_t *entries, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- afr_self_heal_t * sh = NULL;
-
- gf_dirent_t * entry = NULL;
- gf_dirent_t * tmp = NULL;
-
- int child_index = 0;
-
- uint32_t entry_cksum;
-
- int call_count = 0;
- off_t last_offset = 0;
- char sh_type_str[256] = {0,};
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ afr_self_heal_t * sh = NULL;
+ gf_dirent_t * entry = NULL;
+ gf_dirent_t * tmp = NULL;
+ char *reason = NULL;
+ int child_index = 0;
+ uint32_t entry_cksum = 0;
+ int call_count = 0;
+ off_t last_offset = 0;
+ inode_t *inode = NULL;
priv = this->private;
local = frame->local;
sh = &local->self_heal;
+ inode = local->fd->inode;
child_index = (long) cookie;
if (op_ret == -1) {
+ gf_log (this->name, GF_LOG_INFO,
+ "%s: failed to do opendir on %s",
+ local->loc.path, priv->children[child_index]->name);
local->op_ret = -1;
local->op_ret = op_errno;
goto out;
}
- if (op_ret == 0)
+ if (op_ret == 0) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "%s: no entries found in %s",
+ local->loc.path, priv->children[child_index]->name);
goto out;
+ }
list_for_each_entry_safe (entry, tmp, &entries->list, list) {
- entry_cksum = gf_rsync_weak_checksum (entry->d_name,
+ entry_cksum = gf_rsync_weak_checksum ((unsigned char *)entry->d_name,
strlen (entry->d_name));
local->cont.opendir.checksum[child_index] ^= entry_cksum;
}
@@ -144,39 +143,30 @@ afr_examine_dir_readdir_cbk (call_frame_t *frame, void *cookie,
(void *) (long) child_index,
priv->children[child_index],
priv->children[child_index]->fops->readdir,
- local->fd, 131072, last_offset);
+ local->fd, 131072, last_offset, NULL);
+
+ return 0;
out:
- if ((op_ret == 0) || (op_ret == -1)) {
- call_count = afr_frame_return (frame);
-
- if (call_count == 0) {
- if (__checksums_differ (local->cont.opendir.checksum,
- priv->child_count,
- local->child_up)) {
-
- sh->need_entry_self_heal = _gf_true;
- sh->forced_merge = _gf_true;
- sh->type = local->fd->inode->ia_type;
- sh->background = _gf_false;
- sh->unwind = afr_examine_dir_sh_unwind;
-
- afr_self_heal_type_str_get(&local->self_heal,
- sh_type_str,
- sizeof(sh_type_str));
- gf_log (this->name, GF_LOG_NORMAL,
- "%s self-heal triggered. path: %s, "
- "reason: checksums of directory differ,"
- " forced merge option set",
- sh_type_str, local->loc.path);
-
- afr_self_heal (frame, this);
- } else {
- afr_set_opendir_done (this, local->fd->inode);
-
- AFR_STACK_UNWIND (opendir, frame, local->op_ret,
- local->op_errno, local->fd);
- }
+ call_count = afr_frame_return (frame);
+
+ if (call_count == 0) {
+ if (__checksums_differ (local->cont.opendir.checksum,
+ priv->child_count,
+ local->child_up)) {
+
+ sh->do_entry_self_heal = _gf_true;
+ sh->forced_merge = _gf_true;
+
+ reason = "checksums of directory differ";
+ afr_launch_self_heal (frame, this, inode, _gf_false,
+ inode->ia_type, reason, NULL,
+ afr_examine_dir_sh_unwind);
+ } else {
+ afr_set_opendir_done (this, inode);
+
+ AFR_STACK_UNWIND (opendir, frame, local->op_ret,
+ local->op_errno, local->fd, NULL);
}
}
@@ -187,20 +177,19 @@ out:
int
afr_examine_dir (call_frame_t *frame, xlator_t *this)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
-
- int i;
- int call_count = 0;
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ int i = 0;
+ int call_count = 0;
local = frame->local;
priv = this->private;
local->cont.opendir.checksum = GF_CALLOC (priv->child_count,
- sizeof (*local->cont.opendir.checksum),
- gf_afr_mt_int32_t);
+ sizeof (*local->cont.opendir.checksum),
+ gf_afr_mt_int32_t);
- call_count = afr_up_children_count (priv->child_count, local->child_up);
+ call_count = afr_up_children_count (local->child_up, priv->child_count);
local->call_count = call_count;
@@ -210,7 +199,7 @@ afr_examine_dir (call_frame_t *frame, xlator_t *this)
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->readdir,
- local->fd, 131072, 0);
+ local->fd, 131072, 0, NULL);
if (!--call_count)
break;
@@ -223,142 +212,139 @@ afr_examine_dir (call_frame_t *frame, xlator_t *this)
int32_t
afr_opendir_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- fd_t *fd)
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ fd_t *fd, dict_t *xdata)
{
afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
int32_t up_children_count = 0;
- int ret = -1;
-
- int call_count = -1;
+ int ret = -1;
+ int call_count = -1;
+ int32_t child_index = 0;
priv = this->private;
local = frame->local;
+ child_index = (long) cookie;
+
+ up_children_count = afr_up_children_count (local->child_up,
+ priv->child_count);
+
+ LOCK (&frame->lock);
+ {
+ if (op_ret >= 0) {
+ local->op_ret = op_ret;
+ ret = afr_child_fd_ctx_set (this, fd, child_index, 0);
+ if (ret) {
+ local->op_ret = -1;
+ local->op_errno = -ret;
+ goto unlock;
+ }
+ }
- up_children_count = afr_up_children_count (priv->child_count,
- local->child_up);
+ local->op_errno = op_errno;
+ }
+unlock:
+ UNLOCK (&frame->lock);
- LOCK (&frame->lock);
- {
- if (op_ret >= 0)
- local->op_ret = op_ret;
+ call_count = afr_frame_return (frame);
- local->op_errno = op_errno;
- }
- UNLOCK (&frame->lock);
+ if (call_count == 0) {
+ if (local->op_ret != 0)
+ goto out;
- call_count = afr_frame_return (frame);
+ if (!afr_is_opendir_done (this, local->fd->inode) &&
+ up_children_count > 1 && priv->entry_self_heal) {
- if (call_count == 0) {
- if (local->op_ret == 0) {
+ /*
+ * This is the first opendir on this inode. We need
+ * to check if the directory's entries are the same
+ * on all subvolumes. This is needed in addition
+ * to regular entry self-heal because the readdir
+ * call is sent only to the first subvolume, and
+ * thus files that exist only there will never be healed
+ * otherwise (assuming changelog shows no anomalies).
+ */
- ret = afr_fd_ctx_set (this, local->fd);
+ gf_log (this->name, GF_LOG_TRACE,
+ "reading contents of directory %s looking for mismatch",
+ local->loc.path);
+
+ afr_examine_dir (frame, this);
- if (ret) {
- local->op_ret = -1;
- local->op_errno = -1;
- gf_log (this->name, GF_LOG_ERROR, " failed to "
- "set fd ctx for fd %d", local->fd);
- goto out;
- }
- if (!afr_is_opendir_done (this, local->fd->inode) &&
- up_children_count > 1) {
-
- /*
- * This is the first opendir on this inode. We need
- * to check if the directory's entries are the same
- * on all subvolumes. This is needed in addition
- * to regular entry self-heal because the readdir
- * call is sent only to the first subvolume, and
- * thus files that exist only there will never be healed
- * otherwise (assuming changelog shows no anamolies).
- */
-
- gf_log (this->name, GF_LOG_TRACE,
- "reading contents of directory %s looking for mismatch",
- local->loc.path);
-
- afr_examine_dir (frame, this);
-
- } else {
- AFR_STACK_UNWIND (opendir, frame, local->op_ret,
- local->op_errno, local->fd);
- }
} else {
-out:
- AFR_STACK_UNWIND (opendir, frame, local->op_ret,
- local->op_errno, local->fd);
+ /* do the unwind */
+ goto out;
}
}
- return 0;
+ return 0;
+
+out:
+ AFR_STACK_UNWIND (opendir, frame, local->op_ret,
+ local->op_errno, local->fd, NULL);
+
+ return 0;
}
-int32_t
+int32_t
afr_opendir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, fd_t *fd)
+ loc_t *loc, fd_t *fd)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
-
- int child_count = 0;
- int i = 0;
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ int child_count = 0;
+ int i = 0;
+ int ret = -1;
+ int call_count = -1;
+ int32_t op_errno = 0;
- int ret = -1;
- int call_count = -1;
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
- int32_t op_ret = -1;
- int32_t op_errno = 0;
+ priv = this->private;
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
+ child_count = priv->child_count;
- priv = this->private;
-
- child_count = priv->child_count;
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- ALLOC_OR_GOTO (local, afr_local_t, out);
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
loc_copy (&local->loc, loc);
- frame->local = local;
- local->fd = fd_ref (fd);
+ local->fd = fd_ref (fd);
- call_count = local->call_count;
-
- for (i = 0; i < child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND (frame, afr_opendir_cbk,
- priv->children[i],
- priv->children[i]->fops->opendir,
- loc, fd);
+ call_count = local->call_count;
- if (!--call_count)
- break;
- }
- }
+ for (i = 0; i < child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE (frame, afr_opendir_cbk,
+ (void*) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->opendir,
+ loc, fd, NULL);
- op_ret = 0;
+ if (!--call_count)
+ break;
+ }
+ }
+
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (opendir, frame, op_ret, op_errno, fd);
- }
+ if (ret < 0)
+ AFR_STACK_UNWIND (opendir, frame, -1, op_errno, fd, NULL);
- return 0;
+ return 0;
}
/**
* Common algorithm for directory read calls:
- *
+ *
* - Try the fop on the first child that is up
* - if we have failed due to ENOTCONN:
* try the next child
@@ -372,373 +358,179 @@ struct entry_name {
struct list_head list;
};
-
-static gf_boolean_t
-remembered_name (const char *name, struct list_head *entries)
-{
- struct entry_name *e;
- gf_boolean_t ret = _gf_false;
-
- list_for_each_entry (e, entries, list) {
- if (!strcmp (name, e->name)) {
- ret = _gf_true;
- goto out;
- }
- }
-
-out:
- return ret;
-}
-
-
static void
-afr_remember_entries (gf_dirent_t *entries, fd_t *fd)
-{
- struct entry_name *n = NULL;
- gf_dirent_t * entry = NULL;
-
- int ret = 0;
-
- uint64_t ctx;
- afr_fd_ctx_t *fd_ctx;
-
- ret = fd_ctx_get (fd, THIS, &ctx);
- if (ret < 0) {
- gf_log (THIS->name, GF_LOG_DEBUG,
- "could not get fd ctx for fd=%p", fd);
- return;
- }
-
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
- list_for_each_entry (entry, &entries->list, list) {
- n = GF_CALLOC (1, sizeof (*n), gf_afr_mt_entry_name);
- n->name = gf_strdup (entry->d_name);
- INIT_LIST_HEAD (&n->list);
-
- list_add (&n->list, &fd_ctx->entries);
- }
-}
-
-
-static off_t
-afr_filter_entries (gf_dirent_t *entries, fd_t *fd)
+afr_forget_entries (fd_t *fd)
{
- gf_dirent_t *entry, *tmp;
- int ret = 0;
-
- uint64_t ctx;
- afr_fd_ctx_t *fd_ctx;
-
- off_t offset = 0;
-
- ret = fd_ctx_get (fd, THIS, &ctx);
- if (ret < 0) {
- gf_log (THIS->name, GF_LOG_DEBUG,
- "could not get fd ctx for fd=%p", fd);
- return -1;
- }
+ struct entry_name *entry = NULL;
+ struct entry_name *tmp = NULL;
+ int ret = 0;
+ uint64_t ctx = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+
+ ret = fd_ctx_get (fd, THIS, &ctx);
+ if (ret < 0) {
+ gf_log (THIS->name, GF_LOG_INFO,
+ "could not get fd ctx for fd=%p", fd);
+ return;
+ }
fd_ctx = (afr_fd_ctx_t *)(long) ctx;
- list_for_each_entry_safe (entry, tmp, &entries->list, list) {
- offset = entry->d_off;
-
- if (remembered_name (entry->d_name, &fd_ctx->entries)) {
- list_del (&entry->list);
- GF_FREE (entry);
- }
- }
-
- return offset;
+ list_for_each_entry_safe (entry, tmp, &fd_ctx->entries, list) {
+ GF_FREE (entry->name);
+ list_del (&entry->list);
+ GF_FREE (entry);
+ }
}
-
static void
-afr_forget_entries (fd_t *fd)
+afr_readdir_filter_trash_dir (gf_dirent_t *entries, fd_t *fd)
{
- struct entry_name *entry, *tmp;
- int ret = 0;
-
- uint64_t ctx;
- afr_fd_ctx_t *fd_ctx;
-
- ret = fd_ctx_get (fd, THIS, &ctx);
- if (ret < 0) {
- gf_log (THIS->name, GF_LOG_DEBUG,
- "could not get fd ctx for fd=%p", fd);
- return;
- }
-
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+ gf_dirent_t * entry = NULL;
+ gf_dirent_t * tmp = NULL;
- list_for_each_entry_safe (entry, tmp, &fd_ctx->entries, list) {
- GF_FREE (entry->name);
- list_del (&entry->list);
- GF_FREE (entry);
- }
+ list_for_each_entry_safe (entry, tmp, &entries->list, list) {
+ if (__is_root_gfid (fd->inode->gfid) &&
+ !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR)) {
+ list_del_init (&entry->list);
+ GF_FREE (entry);
+ }
+ }
}
-
int32_t
afr_readdir_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- gf_dirent_t *entries)
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ gf_dirent_t *entries, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
+ afr_local_t *local = NULL;
- gf_dirent_t * entry = NULL;
- gf_dirent_t * tmp = NULL;
-
- int child_index = -1;
-
- priv = this->private;
- local = frame->local;
- child_index = (long) cookie;
-
- if (op_ret != -1) {
- list_for_each_entry_safe (entry, tmp, &entries->list, list) {
- entry->d_ino = afr_itransform (entry->d_ino,
- priv->child_count,
- child_index);
-
- if ((local->fd->inode == local->fd->inode->table->root)
- && !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR)) {
- list_del_init (&entry->list);
- GF_FREE (entry);
- }
- }
- }
+ if (op_ret == -1)
+ goto out;
- AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, entries);
+ local = frame->local;
+ afr_readdir_filter_trash_dir (entries, local->fd);
- return 0;
+out:
+ AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, entries, NULL);
+ return 0;
}
int32_t
afr_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *entries)
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+ dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- xlator_t ** children = NULL;
- ino_t inum = 0;
-
- int call_child = 0;
- int ret = 0;
-
- gf_dirent_t * entry = NULL;
- gf_dirent_t * tmp = NULL;
-
- int child_index = -1;
+ afr_local_t *local = NULL;
- uint64_t ctx = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
-
- off_t offset = 0;
-
- priv = this->private;
- children = priv->children;
+ if (op_ret == -1)
+ goto out;
local = frame->local;
-
- child_index = (long) cookie;
-
- if (priv->strict_readdir) {
- ret = fd_ctx_get (local->fd, this, &ctx);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "could not get fd ctx for fd=%p", local->fd);
- op_ret = -1;
- op_errno = -ret;
- goto out;
- }
-
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
- if (child_went_down (op_ret, op_errno)) {
- if (all_tried (child_index, priv->child_count)) {
- goto out;
- }
-
- call_child = ++child_index;
-
- gf_log (this->name, GF_LOG_TRACE,
- "starting readdir afresh on child %d, offset %"PRId64,
- call_child, (uint64_t) 0);
-
- fd_ctx->failed_over = _gf_true;
-
- STACK_WIND_COOKIE (frame, afr_readdirp_cbk,
- (void *) (long) call_child,
- children[call_child],
- children[call_child]->fops->readdirp, local->fd,
- local->cont.readdir.size, 0);
- return 0;
- }
- }
-
- if (op_ret != -1) {
- list_for_each_entry_safe (entry, tmp, &entries->list, list) {
- inum = afr_itransform (entry->d_ino, priv->child_count,
- child_index);
- entry->d_ino = inum;
- inum = afr_itransform (entry->d_stat.ia_ino,
- priv->child_count, child_index);
- entry->d_stat.ia_ino = inum;
-
- if ((local->fd->inode == local->fd->inode->table->root)
- && !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR)) {
- list_del_init (&entry->list);
- GF_FREE (entry);
- }
- }
- }
-
- if (priv->strict_readdir) {
- if (fd_ctx->failed_over) {
- if (list_empty (&entries->list)) {
- goto out;
- }
-
- offset = afr_filter_entries (entries, local->fd);
-
- afr_remember_entries (entries, local->fd);
-
- if (list_empty (&entries->list)) {
- /* All the entries we got were duplicate. We
- shouldn't send an empty list now, because
- that'll make the application stop reading. So
- try to get more entries */
-
- gf_log (this->name, GF_LOG_TRACE,
- "trying to fetch non-duplicate entries from offset %"PRId64", child %s",
- offset, children[child_index]->name);
-
- STACK_WIND_COOKIE (frame, afr_readdirp_cbk,
- (void *) (long) child_index,
- children[child_index],
- children[child_index]->fops->readdirp,
- local->fd, local->cont.readdir.size, offset);
- return 0;
- }
- } else {
- afr_remember_entries (entries, local->fd);
- }
- }
+ afr_readdir_filter_trash_dir (entries, local->fd);
out:
- AFR_STACK_UNWIND (readdirp, frame, op_ret, op_errno, entries);
-
+ AFR_STACK_UNWIND (readdirp, frame, op_ret, op_errno, entries, NULL);
return 0;
}
-
int32_t
afr_do_readdir (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset, int whichop)
+ fd_t *fd, size_t size, off_t offset, int whichop, dict_t *dict)
{
- afr_private_t * priv = NULL;
- xlator_t ** children = NULL;
- int call_child = 0;
- afr_local_t *local = NULL;
-
- uint64_t ctx;
- afr_fd_ctx_t *fd_ctx;
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ int call_child = 0;
+ afr_local_t *local = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int ret = -1;
+ int32_t op_errno = 0;
+ uint64_t read_child = 0;
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
- int ret = -1;
+ priv = this->private;
+ children = priv->children;
- int32_t op_ret = -1;
- int32_t op_errno = 0;
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
- priv = this->private;
- children = priv->children;
+ local->fresh_children = afr_children_create (priv->child_count);
+ if (!local->fresh_children) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- ALLOC_OR_GOTO (local, afr_local_t, out);
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ read_child = afr_inode_get_read_ctx (this, fd->inode,
+ local->fresh_children);
+ ret = afr_get_call_child (this, local->child_up, read_child,
+ local->fresh_children,
+ &call_child,
+ &local->cont.readdir.last_index);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- frame->local = local;
+ fd_ctx = afr_fd_ctx_get (fd, this);
+ if (!fd_ctx) {
+ op_errno = EBADF;
+ goto out;
+ }
- call_child = afr_first_up_child (priv);
- if (call_child == -1) {
- op_errno = ENOTCONN;
- gf_log (this->name, GF_LOG_DEBUG,
- "no child is up");
- goto out;
- }
+ if ((offset == 0) || (fd_ctx->call_child == -1)) {
+ fd_ctx->call_child = call_child;
+ } else if ((priv->readdir_failover == _gf_false) &&
+ (call_child != fd_ctx->call_child)) {
+ op_errno = EBADF;
+ goto out;
+ }
local->fd = fd_ref (fd);
local->cont.readdir.size = size;
-
- if (priv->strict_readdir) {
- ret = fd_ctx_get (fd, this, &ctx);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "could not get fd ctx for fd=%p", fd);
- op_errno = -ret;
- goto out;
- }
-
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
- if (fd_ctx->last_tried != call_child) {
- gf_log (this->name, GF_LOG_TRACE,
- "first up child has changed from %d to %d, restarting readdir from offset 0",
- fd_ctx->last_tried, call_child);
-
- fd_ctx->failed_over = _gf_true;
- offset = 0;
- }
-
- fd_ctx->last_tried = call_child;
- }
+ local->cont.readdir.dict = (dict)? dict_ref (dict) : NULL;
if (whichop == GF_FOP_READDIR)
STACK_WIND_COOKIE (frame, afr_readdir_cbk,
(void *) (long) call_child,
children[call_child],
children[call_child]->fops->readdir, fd,
- size, offset);
+ size, offset, dict);
else
STACK_WIND_COOKIE (frame, afr_readdirp_cbk,
(void *) (long) call_child,
children[call_child],
children[call_child]->fops->readdirp, fd,
- size, offset);
+ size, offset, dict);
- op_ret = 0;
+ return 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, NULL);
- }
- return 0;
+ AFR_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
+ return 0;
}
int32_t
afr_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset)
+ off_t offset, dict_t *xdata)
{
- afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIR);
+ afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIR, xdata);
return 0;
}
int32_t
afr_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset)
+ off_t offset, dict_t *dict)
{
- afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIRP);
+ afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIRP, dict);
return 0;
}
@@ -746,8 +538,8 @@ afr_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
int32_t
afr_releasedir (xlator_t *this, fd_t *fd)
{
- afr_forget_entries (fd);
+ afr_forget_entries (fd);
afr_cleanup_fd_ctx (this, fd);
- return 0;
+ return 0;
}
diff --git a/xlators/cluster/afr/src/afr-dir-read.h b/xlators/cluster/afr/src/afr-dir-read.h
index abde2534d..09456d159 100644
--- a/xlators/cluster/afr/src/afr-dir-read.h
+++ b/xlators/cluster/afr/src/afr-dir-read.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __DIR_READ_H__
@@ -23,28 +14,23 @@
int32_t
afr_opendir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, fd_t *fd);
+ loc_t *loc, fd_t *fd, dict_t *xdata);
int32_t
afr_releasedir (xlator_t *this, fd_t *fd);
int32_t
afr_readdir (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset);
+ fd_t *fd, size_t size, off_t offset, dict_t *xdata);
int32_t
afr_readdirp (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset);
-
-int32_t
-afr_getdents (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset, int32_t flag);
-
+ fd_t *fd, size_t size, off_t offset, dict_t *dict);
int32_t
afr_checksum (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags);
+ loc_t *loc, int32_t flags, dict_t *xdata);
#endif /* __DIR_READ_H__ */
diff --git a/xlators/cluster/afr/src/afr-dir-write.c b/xlators/cluster/afr/src/afr-dir-write.c
index 1aa268f66..1943b719b 100644
--- a/xlators/cluster/afr/src/afr-dir-write.c
+++ b/xlators/cluster/afr/src/afr-dir-write.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -47,28 +38,222 @@
#include "afr.h"
#include "afr-transaction.h"
+int
+afr_build_parent_loc (loc_t *parent, loc_t *child, int32_t *op_errno)
+{
+ int ret = -1;
+ char *child_path = NULL;
+
+ if (!child->parent) {
+ if (op_errno)
+ *op_errno = EINVAL;
+ goto out;
+ }
+
+ child_path = gf_strdup (child->path);
+ if (!child_path) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+ parent->path = gf_strdup( dirname (child_path) );
+ if (!parent->path) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+ parent->inode = inode_ref (child->parent);
+ uuid_copy (parent->gfid, child->pargfid);
+
+ ret = 0;
+out:
+ GF_FREE(child_path);
+
+ return ret;
+}
void
-afr_build_parent_loc (loc_t *parent, loc_t *child)
+__dir_entry_fop_common_cbk (call_frame_t *frame, int child_index,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, struct iatt *prenewparent,
+ struct iatt *postnewparent)
{
- char *tmp = NULL;
+ afr_local_t *local = NULL;
+
+ local = frame->local;
+
+ if (afr_fop_failed (op_ret, op_errno))
+ afr_transaction_fop_failed (frame, this, child_index);
+
+ if (op_ret > -1) {
+ local->op_ret = op_ret;
+
+ if ((local->success_count == 0) ||
+ (child_index == local->read_child_index)) {
+ local->cont.dir_fop.preparent = *preparent;
+ local->cont.dir_fop.postparent = *postparent;
+ if (buf)
+ local->cont.dir_fop.buf = *buf;
+ if (prenewparent)
+ local->cont.dir_fop.prenewparent = *prenewparent;
+ if (postnewparent)
+ local->cont.dir_fop.postnewparent = *postnewparent;
+ }
- if (!child->parent) {
- loc_copy (parent, child);
- return;
- }
+ local->cont.dir_fop.inode = inode;
- tmp = gf_strdup (child->path);
- parent->path = gf_strdup (dirname (tmp));
- GF_FREE (tmp);
+ local->fresh_children[local->success_count] = child_index;
+ local->success_count++;
+ local->child_errno[child_index] = 0;
+ } else {
+ local->child_errno[child_index] = op_errno;
+ }
- parent->name = strrchr (parent->path, '/');
- if (parent->name)
- parent->name++;
+ local->op_errno = op_errno;
+}
- parent->inode = inode_ref (child->parent);
- parent->parent = inode_parent (parent->inode, 0, NULL);
- parent->ino = parent->inode->ino;
+int
+afr_mark_new_entry_changelog_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ dict_t *xattr, dict_t *xdata)
+{
+ int call_count = 0;
+
+ call_count = afr_frame_return (frame);
+ if (call_count == 0) {
+ AFR_STACK_DESTROY (frame);
+ }
+ return 0;
+}
+
+void
+afr_mark_new_entry_changelog (call_frame_t *frame, xlator_t *this)
+{
+ call_frame_t *new_frame = NULL;
+ afr_local_t *local = NULL;
+ afr_local_t *new_local = NULL;
+ afr_private_t *priv = NULL;
+ dict_t **xattr = NULL;
+ int32_t **changelog = NULL;
+ int i = 0;
+ GF_UNUSED int op_errno = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ new_frame = copy_frame (frame);
+ if (!new_frame) {
+ goto out;
+ }
+
+ AFR_LOCAL_ALLOC_OR_GOTO (new_frame->local, out);
+ new_local = new_frame->local;
+ changelog = afr_matrix_create (priv->child_count, AFR_NUM_CHANGE_LOGS);
+ if (!changelog)
+ goto out;
+
+ xattr = GF_CALLOC (priv->child_count, sizeof (*xattr),
+ gf_afr_mt_dict_t);
+ if (!xattr)
+ goto out;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_errno[i])
+ continue;
+ xattr[i] = dict_new ();
+ if (!xattr[i])
+ goto out;
+ }
+
+ afr_prepare_new_entry_pending_matrix (changelog,
+ afr_is_errno_set,
+ local->child_errno,
+ &local->cont.dir_fop.buf,
+ priv->child_count);
+
+ new_local->pending = changelog;
+ uuid_copy (new_local->loc.gfid, local->cont.dir_fop.buf.ia_gfid);
+ new_local->loc.inode = inode_ref (local->cont.dir_fop.inode);
+ new_local->call_count = local->success_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_errno[i])
+ continue;
+
+ afr_set_pending_dict (priv, xattr[i], changelog, i, LOCAL_LAST);
+ STACK_WIND_COOKIE (new_frame, afr_mark_new_entry_changelog_cbk,
+ (void *) (long) i, priv->children[i],
+ priv->children[i]->fops->xattrop,
+ &new_local->loc, GF_XATTROP_ADD_ARRAY,
+ xattr[i], NULL);
+ }
+ new_frame = NULL;
+out:
+ if (new_frame)
+ AFR_STACK_DESTROY (new_frame);
+ afr_xattr_array_destroy (xattr, priv->child_count);
+ return;
+}
+
+gf_boolean_t
+afr_is_new_entry_changelog_needed (glusterfs_fop_t fop)
+{
+ glusterfs_fop_t fops[] = {GF_FOP_CREATE, GF_FOP_MKNOD, GF_FOP_NULL};
+ int i = 0;
+
+ for (i = 0; fops[i] != GF_FOP_NULL; i++) {
+ if (fop == fops[i])
+ return _gf_true;
+ }
+ return _gf_false;
+}
+
+void
+afr_dir_fop_mark_entry_pending_changelog (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ if (local->op_ret < 0)
+ goto out;
+
+ if (local->success_count == priv->child_count)
+ goto out;
+
+ if (!afr_is_new_entry_changelog_needed (local->op))
+ goto out;
+
+ afr_mark_new_entry_changelog (frame, this);
+
+out:
+ return;
+}
+
+void
+afr_dir_fop_done (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ if (local->cont.dir_fop.inode == NULL)
+ goto done;
+ afr_set_read_ctx_from_policy (this, local->cont.dir_fop.inode,
+ local->fresh_children,
+ local->read_child_index,
+ priv->read_child,
+ local->cont.dir_fop.buf.ia_gfid);
+done:
+ local->transaction.unwind (frame, this);
+ afr_dir_fop_mark_entry_pending_changelog (frame, this);
+ local->transaction.resume (frame, this);
}
/* {{{ create */
@@ -76,77 +261,57 @@ afr_build_parent_loc (loc_t *parent, loc_t *child)
int
afr_create_unwind (call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
- struct iatt *unwind_buf = NULL;
-
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- if (local->transaction.main_frame) {
- main_frame = local->transaction.main_frame;
- }
- local->transaction.main_frame = NULL;
- }
- UNLOCK (&frame->lock);
-
- if (main_frame) {
- if (local->cont.create.read_child_buf.ia_ino) {
- unwind_buf = &local->cont.create.read_child_buf;
- } else {
- unwind_buf = &local->cont.create.buf;
- }
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- unwind_buf->ia_ino = local->cont.create.ino;
+ local = frame->local;
- local->cont.create.preparent.ia_ino = local->cont.create.parent_ino;
- local->cont.create.postparent.ia_ino = local->cont.create.parent_ino;
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame) {
+ main_frame = local->transaction.main_frame;
+ }
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK (&frame->lock);
- AFR_STACK_UNWIND (create, main_frame,
+ if (main_frame) {
+ AFR_STACK_UNWIND (create, main_frame,
local->op_ret, local->op_errno,
- local->cont.create.fd,
- local->cont.create.inode,
- unwind_buf, &local->cont.create.preparent,
- &local->cont.create.postparent);
+ local->cont.create.fd,
+ local->cont.dir_fop.inode,
+ &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ local->xdata_rsp);
}
-
- return 0;
+
+ return 0;
}
int
-afr_create_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- fd_t *fd, inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent)
+afr_create_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ fd_t *fd, inode_t *inode, struct iatt *buf,
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
-
- uint64_t ctx;
- afr_fd_ctx_t *fd_ctx;
-
- int ret = 0;
-
- int call_count = -1;
- int child_index = -1;
-
- local = frame->local;
- priv = this->private;
+ afr_local_t *local = NULL;
+ uint64_t ctx = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int ret = 0;
+ int call_count = -1;
+ int child_index = -1;
- child_index = (long) cookie;
+ local = frame->local;
- LOCK (&frame->lock);
- {
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
- if (op_ret != -1) {
- local->op_ret = op_ret;
+ child_index = (long) cookie;
+ LOCK (&frame->lock);
+ {
+ if (op_ret > -1) {
ret = afr_fd_ctx_set (this, fd);
-
if (ret < 0) {
gf_log (this->name, GF_LOG_ERROR,
"could not set ctx on fd=%p", fd);
@@ -157,7 +322,6 @@ afr_create_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
}
ret = fd_ctx_get (fd, this, &ctx);
-
if (ret < 0) {
gf_log (this->name, GF_LOG_ERROR,
"could not get fd ctx for fd=%p", fd);
@@ -168,194 +332,176 @@ afr_create_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
fd_ctx = (afr_fd_ctx_t *)(long) ctx;
- fd_ctx->opened_on[child_index] = 1;
+ fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
fd_ctx->flags = local->cont.create.flags;
if (local->success_count == 0) {
- local->cont.create.buf = *buf;
-
- local->cont.create.ino =
- afr_itransform (buf->ia_ino,
- priv->child_count,
- child_index);
-
- if (priv->read_child >= 0) {
- afr_set_read_child (this, inode,
- priv->read_child);
- } else {
- afr_set_read_child (this, inode,
- local->read_child_index);
- }
+ if (xdata)
+ local->xdata_rsp = dict_ref(xdata);
}
-
- if (child_index == local->first_up_child) {
- local->cont.create.ino =
- afr_itransform (buf->ia_ino,
- priv->child_count,
- local->first_up_child);
- }
-
- if (child_index == local->read_child_index) {
- local->cont.create.read_child_buf = *buf;
- local->cont.create.preparent = *preparent;
- local->cont.create.postparent = *postparent;
- }
-
- local->cont.create.inode = inode;
-
- local->success_count++;
- }
-
- local->op_errno = op_errno;
- }
+ }
+ __dir_entry_fop_common_cbk (frame, child_index, this,
+ op_ret, op_errno, inode, buf,
+ preparent, postparent, NULL, NULL);
+ }
unlock:
- UNLOCK (&frame->lock);
+ UNLOCK (&frame->lock);
- call_count = afr_frame_return (frame);
+ call_count = afr_frame_return (frame);
- if (call_count == 0) {
- local->transaction.unwind (frame, this);
+ if (call_count == 0)
+ afr_dir_fop_done (frame, this);
- local->transaction.resume (frame, this);
- }
-
- return 0;
+ return 0;
}
int
afr_create_wind (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- int call_count = -1;
- int i = 0;
-
- local = frame->local;
- priv = this->private;
-
- call_count = afr_up_children_count (priv->child_count, local->child_up);
-
- if (call_count == 0) {
- local->transaction.resume (frame, this);
- return 0;
- }
-
- local->call_count = call_count;
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_create_wind_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->create,
- &local->loc,
- local->cont.create.flags,
- local->cont.create.mode,
- local->cont.create.fd,
- local->cont.create.params);
- if (!--call_count)
- break;
- }
- }
-
- return 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
+
+ local->call_count = call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_create_wind_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->create,
+ &local->loc,
+ local->cont.create.flags,
+ local->cont.create.mode,
+ local->umask,
+ local->cont.create.fd,
+ local->xdata_req);
+ if (!--call_count)
+ break;
+ }
+ }
+
+ return 0;
}
int
afr_create_done (call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
+ afr_local_t * local = NULL;
- local = frame->local;
+ local = frame->local;
- local->transaction.unwind (frame, this);
+ local->transaction.unwind (frame, this);
- AFR_STACK_DESTROY (frame);
+ AFR_STACK_DESTROY (frame);
- return 0;
+ return 0;
}
int
afr_create (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, mode_t mode,
- fd_t *fd, dict_t *params)
+ loc_t *loc, int32_t flags, mode_t mode,
+ mode_t umask, fd_t *fd, dict_t *params)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
-
- int ret = -1;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
- int op_ret = -1;
- int op_errno = 0;
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
+ priv = this->private;
- priv = this->private;
+ QUORUM_CHECK(create,out);
- transaction_frame = copy_frame (frame);
- if (!transaction_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
- goto out;
- }
-
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+ local = transaction_frame->local;
- transaction_frame->local = local;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
- loc_copy (&local->loc, loc);
+ loc_copy (&local->loc, loc);
LOCK (&priv->read_child_lock);
{
- local->read_child_index = (++priv->read_child_rr)
+ local->read_child_index = (++priv->read_child_rr)
% (priv->child_count);
}
UNLOCK (&priv->read_child_lock);
- local->cont.create.flags = flags;
- local->cont.create.mode = mode;
- local->cont.create.fd = fd_ref (fd);
+ local->op = GF_FOP_CREATE;
+ local->cont.create.flags = flags;
+ local->cont.create.mode = mode;
+ local->cont.create.fd = fd_ref (fd);
+ local->umask = umask;
if (params)
- local->cont.create.params = dict_ref (params);
-
- if (loc->parent)
- local->cont.create.parent_ino = loc->parent->ino;
-
- local->transaction.fop = afr_create_wind;
- local->transaction.done = afr_create_done;
- local->transaction.unwind = afr_create_unwind;
-
- afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
-
- afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ local->xdata_req = dict_ref (params);
+
+ local->transaction.fop = afr_create_wind;
+ local->transaction.done = afr_create_done;
+ local->transaction.unwind = afr_create_unwind;
+
+ ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+ &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ int_lock->lockee_count++;
+ ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (create, frame, op_ret, op_errno,
- NULL, NULL, NULL, NULL, NULL);
- }
+ if (ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (create, frame, -1, op_errno,
+ NULL, NULL, NULL, NULL, NULL, NULL);
+ }
- return 0;
+ return 0;
}
/* }}} */
@@ -365,42 +511,31 @@ out:
int
afr_mknod_unwind (call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- struct iatt *unwind_buf = NULL;
-
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- if (local->transaction.main_frame) {
- main_frame = local->transaction.main_frame;
- }
- local->transaction.main_frame = NULL;
- }
- UNLOCK (&frame->lock);
-
- if (main_frame) {
- if (local->cont.mknod.read_child_buf.ia_ino) {
- unwind_buf = &local->cont.mknod.read_child_buf;
- } else {
- unwind_buf = &local->cont.mknod.buf;
- }
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- unwind_buf->ia_ino = local->cont.mknod.ino;
+ local = frame->local;
- local->cont.mknod.preparent.ia_ino = local->cont.mknod.parent_ino;
- local->cont.mknod.postparent.ia_ino = local->cont.mknod.parent_ino;
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame) {
+ main_frame = local->transaction.main_frame;
+ }
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK (&frame->lock);
- AFR_STACK_UNWIND (mknod, main_frame,
+ if (main_frame) {
+ AFR_STACK_UNWIND (mknod, main_frame,
local->op_ret, local->op_errno,
- local->cont.mknod.inode,
- unwind_buf, &local->cont.mknod.preparent,
- &local->cont.mknod.postparent);
+ local->cont.dir_fop.inode,
+ &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ NULL);
}
- return 0;
+ return 0;
}
@@ -408,203 +543,169 @@ int
afr_mknod_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, inode_t *inode,
struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
-
- int call_count = -1;
- int child_index = -1;
-
- local = frame->local;
- priv = this->private;
-
- child_index = (long) cookie;
-
- LOCK (&frame->lock);
- {
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
- if (op_ret != -1) {
- local->op_ret = op_ret;
-
- if (local->success_count == 0){
- local->cont.mknod.buf = *buf;
- local->cont.mknod.ino =
- afr_itransform (buf->ia_ino,
- priv->child_count,
- child_index);
-
- if (priv->read_child >= 0) {
- afr_set_read_child (this, inode,
- priv->read_child);
- } else {
- afr_set_read_child (this, inode,
- local->read_child_index);
- }
- }
+ int call_count = -1;
+ int child_index = -1;
- if (child_index == local->first_up_child) {
- local->cont.mknod.ino =
- afr_itransform (buf->ia_ino,
- priv->child_count,
- local->first_up_child);
- }
-
- if (child_index == local->read_child_index) {
- local->cont.mknod.read_child_buf = *buf;
- local->cont.mknod.preparent = *preparent;
- local->cont.mknod.postparent = *postparent;
- }
-
- local->cont.mknod.inode = inode;
+ child_index = (long) cookie;
- local->success_count++;
- }
-
- local->op_errno = op_errno;
- }
- UNLOCK (&frame->lock);
+ LOCK (&frame->lock);
+ {
+ __dir_entry_fop_common_cbk (frame, child_index, this,
+ op_ret, op_errno, inode, buf,
+ preparent, postparent, NULL, NULL);
+ }
+ UNLOCK (&frame->lock);
- call_count = afr_frame_return (frame);
+ call_count = afr_frame_return (frame);
- if (call_count == 0) {
- local->transaction.unwind (frame, this);
+ if (call_count == 0)
+ afr_dir_fop_done (frame, this);
- local->transaction.resume (frame, this);
- }
-
- return 0;
+ return 0;
}
int32_t
afr_mknod_wind (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- int call_count = -1;
- int i = 0;
-
- local = frame->local;
- priv = this->private;
-
- call_count = afr_up_children_count (priv->child_count, local->child_up);
-
- if (call_count == 0) {
- local->transaction.resume (frame, this);
- return 0;
- }
-
- local->call_count = call_count;
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_mknod_wind_cbk, (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->mknod,
- &local->loc, local->cont.mknod.mode,
- local->cont.mknod.dev,
- local->cont.mknod.params);
- if (!--call_count)
- break;
- }
- }
-
- return 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
+
+ local->call_count = call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_mknod_wind_cbk, (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->mknod,
+ &local->loc, local->cont.mknod.mode,
+ local->cont.mknod.dev,
+ local->umask,
+ local->xdata_req);
+ if (!--call_count)
+ break;
+ }
+ }
+
+ return 0;
}
int
afr_mknod_done (call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
+ afr_local_t * local = NULL;
- local = frame->local;
+ local = frame->local;
- local->transaction.unwind (frame, this);
- AFR_STACK_DESTROY (frame);
+ local->transaction.unwind (frame, this);
+ AFR_STACK_DESTROY (frame);
- return 0;
+ return 0;
}
int
-afr_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t dev, dict_t *params)
+afr_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t dev, mode_t umask, dict_t *params)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
-
- int ret = -1;
-
- int op_ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
- priv = this->private;
+ priv = this->private;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
- goto out;
- }
+ QUORUM_CHECK(mknod,out);
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+ local = transaction_frame->local;
- transaction_frame->local = local;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
- loc_copy (&local->loc, loc);
+ loc_copy (&local->loc, loc);
LOCK (&priv->read_child_lock);
{
- local->read_child_index = (++priv->read_child_rr)
+ local->read_child_index = (++priv->read_child_rr)
% (priv->child_count);
}
UNLOCK (&priv->read_child_lock);
- local->cont.mknod.mode = mode;
- local->cont.mknod.dev = dev;
+ local->op = GF_FOP_MKNOD;
+ local->cont.mknod.mode = mode;
+ local->cont.mknod.dev = dev;
+ local->umask = umask;
if (params)
- local->cont.mknod.params = dict_ref (params);
-
- if (loc->parent)
- local->cont.mknod.parent_ino = loc->parent->ino;
-
- local->transaction.fop = afr_mknod_wind;
- local->transaction.done = afr_mknod_done;
- local->transaction.unwind = afr_mknod_unwind;
-
- afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
-
- afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ local->xdata_req = dict_ref (params);
+
+ local->transaction.fop = afr_mknod_wind;
+ local->transaction.done = afr_mknod_done;
+ local->transaction.unwind = afr_mknod_unwind;
+
+ ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+ &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ int_lock->lockee_count++;
+ ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (mknod, frame, op_ret, op_errno,
- NULL, NULL, NULL, NULL);
- }
-
- return 0;
+ if (ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (mknod, frame, -1, op_errno,
+ NULL, NULL, NULL, NULL, NULL);
+ }
+
+ return 0;
}
/* }}} */
@@ -615,42 +716,31 @@ out:
int
afr_mkdir_unwind (call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- struct iatt *unwind_buf = NULL;
-
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- if (local->transaction.main_frame) {
- main_frame = local->transaction.main_frame;
- }
- local->transaction.main_frame = NULL;
- }
- UNLOCK (&frame->lock);
-
- if (main_frame) {
- if (local->cont.mkdir.read_child_buf.ia_ino) {
- unwind_buf = &local->cont.mkdir.read_child_buf;
- } else {
- unwind_buf = &local->cont.mkdir.buf;
- }
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- unwind_buf->ia_ino = local->cont.mkdir.ino;
+ local = frame->local;
- local->cont.mkdir.preparent.ia_ino = local->cont.mkdir.parent_ino;
- local->cont.mkdir.postparent.ia_ino = local->cont.mkdir.parent_ino;
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame) {
+ main_frame = local->transaction.main_frame;
+ }
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK (&frame->lock);
- AFR_STACK_UNWIND (mkdir, main_frame,
+ if (main_frame) {
+ AFR_STACK_UNWIND (mkdir, main_frame,
local->op_ret, local->op_errno,
- local->cont.mkdir.inode,
- unwind_buf, &local->cont.mkdir.preparent,
- &local->cont.mkdir.postparent);
+ local->cont.dir_fop.inode,
+ &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ NULL);
}
- return 0;
+ return 0;
}
@@ -658,205 +748,169 @@ int
afr_mkdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, inode_t *inode,
struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
-
- int call_count = -1;
- int child_index = -1;
-
- local = frame->local;
- priv = this->private;
-
- child_index = (long) cookie;
-
- LOCK (&frame->lock);
- {
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
+ int call_count = -1;
+ int child_index = -1;
- if (op_ret != -1) {
- local->op_ret = op_ret;
+ child_index = (long) cookie;
- if (local->success_count == 0) {
- local->cont.mkdir.buf = *buf;
-
- local->cont.mkdir.ino =
- afr_itransform (buf->ia_ino,
- priv->child_count,
- child_index);
+ LOCK (&frame->lock);
+ {
+ __dir_entry_fop_common_cbk (frame, child_index, this,
+ op_ret, op_errno, inode, buf,
+ preparent, postparent, NULL, NULL);
+ }
+ UNLOCK (&frame->lock);
- if (priv->read_child >= 0) {
- afr_set_read_child (this, inode,
- priv->read_child);
- } else {
- afr_set_read_child (this, inode,
- local->read_child_index);
- }
- }
+ call_count = afr_frame_return (frame);
- if (child_index == local->first_up_child) {
- local->cont.mkdir.ino =
- afr_itransform (buf->ia_ino,
- priv->child_count,
- local->first_up_child);
- }
-
- if (child_index == local->read_child_index) {
- local->cont.mkdir.read_child_buf = *buf;
- local->cont.mkdir.preparent = *preparent;
- local->cont.mkdir.postparent = *postparent;
- }
+ if (call_count == 0)
+ afr_dir_fop_done (frame, this);
- local->cont.mkdir.inode = inode;
+ return 0;
+}
- local->success_count++;
- }
- local->op_errno = op_errno;
- }
- UNLOCK (&frame->lock);
+int
+afr_mkdir_wind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
- call_count = afr_frame_return (frame);
+ local = frame->local;
+ priv = this->private;
- if (call_count == 0) {
- local->transaction.unwind (frame, this);
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
- local->transaction.resume (frame, this);
- }
-
- return 0;
-}
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
+ local->call_count = call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_mkdir_wind_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->mkdir,
+ &local->loc, local->cont.mkdir.mode,
+ local->umask,
+ local->xdata_req);
+ if (!--call_count)
+ break;
+ }
+ }
-int
-afr_mkdir_wind (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- int call_count = -1;
- int i = 0;
-
- local = frame->local;
- priv = this->private;
-
- call_count = afr_up_children_count (priv->child_count, local->child_up);
-
- if (call_count == 0) {
- local->transaction.resume (frame, this);
- return 0;
- }
-
- local->call_count = call_count;
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_mkdir_wind_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->mkdir,
- &local->loc, local->cont.mkdir.mode,
- local->cont.mkdir.params);
- if (!--call_count)
- break;
- }
- }
-
- return 0;
+ return 0;
}
int
afr_mkdir_done (call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
+ afr_local_t * local = NULL;
- local = frame->local;
+ local = frame->local;
- local->transaction.unwind (frame, this);
+ local->transaction.unwind (frame, this);
- AFR_STACK_DESTROY (frame);
+ AFR_STACK_DESTROY (frame);
- return 0;
+ return 0;
}
-
int
afr_mkdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dict_t *params)
+ loc_t *loc, mode_t mode, mode_t umask, dict_t *params)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
-
- int ret = -1;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
- int op_ret = -1;
- int op_errno = 0;
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
+ priv = this->private;
- priv = this->private;
+ QUORUM_CHECK(mkdir,out);
- transaction_frame = copy_frame (frame);
- if (!transaction_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
- goto out;
- }
-
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+ local = transaction_frame->local;
- transaction_frame->local = local;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
- loc_copy (&local->loc, loc);
+ loc_copy (&local->loc, loc);
LOCK (&priv->read_child_lock);
{
- local->read_child_index = (++priv->read_child_rr)
+ local->read_child_index = (++priv->read_child_rr)
% (priv->child_count);
}
UNLOCK (&priv->read_child_lock);
- local->cont.mkdir.mode = mode;
+ local->cont.mkdir.mode = mode;
+ local->umask = umask;
if (params)
- local->cont.mkdir.params = dict_ref (params);
-
- if (loc->parent)
- local->cont.mkdir.parent_ino = loc->parent->ino;
-
- local->transaction.fop = afr_mkdir_wind;
- local->transaction.done = afr_mkdir_done;
- local->transaction.unwind = afr_mkdir_unwind;
-
- afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
-
- afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ local->xdata_req = dict_ref (params);
+
+ local->op = GF_FOP_MKDIR;
+ local->transaction.fop = afr_mkdir_wind;
+ local->transaction.done = afr_mkdir_done;
+ local->transaction.unwind = afr_mkdir_unwind;
+
+ ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+ &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ int_lock->lockee_count++;
+ ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (mkdir, frame, op_ret, op_errno,
- NULL, NULL, NULL, NULL);
- }
+ AFR_STACK_UNWIND (mkdir, frame, -1, op_errno,
+ NULL, NULL, NULL, NULL, NULL);
+ }
- return 0;
+ return 0;
}
/* }}} */
@@ -867,233 +921,196 @@ out:
int
afr_link_unwind (call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- struct iatt *unwind_buf = NULL;
-
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- if (local->transaction.main_frame) {
- main_frame = local->transaction.main_frame;
- }
- local->transaction.main_frame = NULL;
- }
- UNLOCK (&frame->lock);
-
- if (main_frame) {
- if (local->cont.link.read_child_buf.ia_ino) {
- unwind_buf = &local->cont.link.read_child_buf;
- } else {
- unwind_buf = &local->cont.link.buf;
- }
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- unwind_buf->ia_ino = local->cont.link.ino;
+ local = frame->local;
- local->cont.link.preparent.ia_ino = local->cont.link.parent_ino;
- local->cont.link.postparent.ia_ino = local->cont.link.parent_ino;
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame) {
+ main_frame = local->transaction.main_frame;
+ }
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK (&frame->lock);
- AFR_STACK_UNWIND (link, main_frame,
- local->op_ret, local->op_errno,
- local->cont.link.inode,
- unwind_buf, &local->cont.link.preparent,
- &local->cont.link.postparent);
- }
+ if (main_frame) {
+ AFR_STACK_UNWIND (link, main_frame,
+ local->op_ret, local->op_errno,
+ local->cont.dir_fop.inode,
+ &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ NULL);
+ }
- return 0;
+ return 0;
}
int
-afr_link_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
+afr_link_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
-
- int call_count = -1;
- int child_index = -1;
-
- local = frame->local;
- priv = this->private;
+ int call_count = -1;
+ int child_index = -1;
- child_index = (long) cookie;
+ child_index = (long) cookie;
- LOCK (&frame->lock);
- {
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
+ LOCK (&frame->lock);
+ {
+ __dir_entry_fop_common_cbk (frame, child_index, this,
+ op_ret, op_errno, inode, buf,
+ preparent, postparent, NULL, NULL);
+ }
+ UNLOCK (&frame->lock);
- if (op_ret != -1) {
- local->op_ret = op_ret;
+ call_count = afr_frame_return (frame);
- if (local->success_count == 0) {
- local->cont.link.buf = *buf;
+ if (call_count == 0)
+ afr_dir_fop_done (frame, this);
- if (priv->read_child >= 0) {
- afr_set_read_child (this, inode,
- priv->read_child);
- } else {
- afr_set_read_child (this, inode,
- local->read_child_index);
- }
- }
+ return 0;
+}
- if (child_index == local->read_child_index) {
- local->cont.link.read_child_buf = *buf;
- local->cont.link.preparent = *preparent;
- local->cont.link.postparent = *postparent;
- }
- local->cont.link.inode = inode;
+int
+afr_link_wind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
- local->success_count++;
- }
+ local = frame->local;
+ priv = this->private;
- local->op_errno = op_errno;
- }
- UNLOCK (&frame->lock);
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
- call_count = afr_frame_return (frame);
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
- if (call_count == 0) {
- local->transaction.unwind (frame, this);
+ local->call_count = call_count;
- local->transaction.resume (frame, this);
- }
-
- return 0;
-}
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_link_wind_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->link,
+ &local->loc,
+ &local->newloc, local->xdata_req);
+ if (!--call_count)
+ break;
+ }
+ }
-int
-afr_link_wind (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- int call_count = -1;
- int i = 0;
-
- local = frame->local;
- priv = this->private;
-
- call_count = afr_up_children_count (priv->child_count, local->child_up);
-
- if (call_count == 0) {
- local->transaction.resume (frame, this);
- return 0;
- }
-
- local->call_count = call_count;
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_link_wind_cbk, (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->link,
- &local->loc,
- &local->newloc);
-
- if (!--call_count)
- break;
- }
- }
-
- return 0;
+ return 0;
}
int
afr_link_done (call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = frame->local;
+ afr_local_t * local = frame->local;
- local->transaction.unwind (frame, this);
+ local->transaction.unwind (frame, this);
- AFR_STACK_DESTROY (frame);
+ AFR_STACK_DESTROY (frame);
- return 0;
+ return 0;
}
int
afr_link (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc)
+ loc_t *oldloc, loc_t *newloc, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
-
- int ret = -1;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
- int op_ret = -1;
- int op_errno = 0;
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
+ priv = this->private;
- priv = this->private;
+ QUORUM_CHECK(link,out);
- transaction_frame = copy_frame (frame);
- if (!transaction_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
- goto out;
- }
-
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+ local = transaction_frame->local;
- transaction_frame->local = local;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
- loc_copy (&local->loc, oldloc);
- loc_copy (&local->newloc, newloc);
+ loc_copy (&local->loc, oldloc);
+ loc_copy (&local->newloc, newloc);
+ if (xdata)
+ local->xdata_req = dict_ref (xdata);
LOCK (&priv->read_child_lock);
{
- local->read_child_index = (++priv->read_child_rr)
+ local->read_child_index = (++priv->read_child_rr)
% (priv->child_count);
}
UNLOCK (&priv->read_child_lock);
- local->cont.link.ino = oldloc->inode->ino;
-
- if (oldloc->parent)
- local->cont.link.parent_ino = newloc->parent->ino;
-
- local->transaction.fop = afr_link_wind;
- local->transaction.done = afr_link_done;
- local->transaction.unwind = afr_link_unwind;
-
- afr_build_parent_loc (&local->transaction.parent_loc, oldloc);
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (oldloc->path);
- local->transaction.new_basename = AFR_BASENAME (newloc->path);
-
- afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
-
- op_ret = 0;
+ local->op = GF_FOP_LINK;
+ local->transaction.fop = afr_link_wind;
+ local->transaction.done = afr_link_done;
+ local->transaction.unwind = afr_link_unwind;
+
+ ret = afr_build_parent_loc (&local->transaction.parent_loc, newloc,
+ &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME (newloc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ int_lock->lockee_count++;
+ ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+ ret = 0;
out:
- if (op_ret == -1) {
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (link, frame, op_ret, op_errno,
- NULL, NULL, NULL, NULL);
- }
-
- return 0;
+ if (ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (link, frame, -1, op_errno,
+ NULL, NULL, NULL, NULL, NULL);
+ }
+
+ return 0;
}
/* }}} */
@@ -1104,246 +1121,202 @@ out:
int
afr_symlink_unwind (call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- struct iatt *unwind_buf = NULL;
-
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- if (local->transaction.main_frame) {
- main_frame = local->transaction.main_frame;
- }
- local->transaction.main_frame = NULL;
- }
- UNLOCK (&frame->lock);
-
- if (main_frame) {
- if (local->cont.symlink.read_child_buf.ia_ino) {
- unwind_buf = &local->cont.symlink.read_child_buf;
- } else {
- unwind_buf = &local->cont.symlink.buf;
- }
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- unwind_buf->ia_ino = local->cont.symlink.ino;
+ local = frame->local;
- local->cont.symlink.preparent.ia_ino = local->cont.symlink.parent_ino;
- local->cont.symlink.postparent.ia_ino = local->cont.symlink.parent_ino;
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame) {
+ main_frame = local->transaction.main_frame;
+ }
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK (&frame->lock);
- AFR_STACK_UNWIND (symlink, main_frame,
+ if (main_frame) {
+ AFR_STACK_UNWIND (symlink, main_frame,
local->op_ret, local->op_errno,
- local->cont.symlink.inode,
- unwind_buf, &local->cont.symlink.preparent,
- &local->cont.symlink.postparent);
+ local->cont.dir_fop.inode,
+ &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ NULL);
}
- return 0;
+ return 0;
}
int
-afr_symlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
+afr_symlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
-
- int call_count = -1;
- int child_index = -1;
-
- local = frame->local;
- priv = this->private;
-
- child_index = (long) cookie;
-
- LOCK (&frame->lock);
- {
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
- if (op_ret != -1) {
- local->op_ret = op_ret;
-
- if (local->success_count == 0) {
- local->cont.symlink.buf = *buf;
- local->cont.symlink.ino =
- afr_itransform (buf->ia_ino, priv->child_count,
- child_index);
-
- if (priv->read_child >= 0) {
- afr_set_read_child (this, inode,
- priv->read_child);
- } else {
- afr_set_read_child (this, inode,
- local->read_child_index);
- }
- }
-
- if (child_index == local->first_up_child) {
- local->cont.symlink.ino =
- afr_itransform (buf->ia_ino,
- priv->child_count,
- local->first_up_child);
- }
-
- if (child_index == local->read_child_index) {
- local->cont.symlink.read_child_buf = *buf;
- local->cont.symlink.preparent = *preparent;
- local->cont.symlink.postparent = *postparent;
- }
-
- local->cont.symlink.inode = inode;
+ int call_count = -1;
+ int child_index = -1;
- local->success_count++;
- }
+ child_index = (long) cookie;
- local->op_errno = op_errno;
- }
- UNLOCK (&frame->lock);
+ LOCK (&frame->lock);
+ {
+ __dir_entry_fop_common_cbk (frame, child_index, this,
+ op_ret, op_errno, inode, buf,
+ preparent, postparent, NULL, NULL);
+ }
+ UNLOCK (&frame->lock);
- call_count = afr_frame_return (frame);
+ call_count = afr_frame_return (frame);
- if (call_count == 0) {
- local->transaction.unwind (frame, this);
+ if (call_count == 0)
+ afr_dir_fop_done (frame, this);
- local->transaction.resume (frame, this);
- }
-
- return 0;
+ return 0;
}
int
afr_symlink_wind (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
- int call_count = -1;
- int i = 0;
+ local = frame->local;
+ priv = this->private;
- local = frame->local;
- priv = this->private;
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
- call_count = afr_up_children_count (priv->child_count, local->child_up);
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
- if (call_count == 0) {
- local->transaction.resume (frame, this);
- return 0;
- }
+ local->call_count = call_count;
- local->call_count = call_count;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_symlink_wind_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->symlink,
+ local->cont.symlink.linkpath,
+ &local->loc,
+ local->umask,
+ local->xdata_req);
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_symlink_wind_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->symlink,
- local->cont.symlink.linkpath,
- &local->loc,
- local->cont.symlink.params);
+ if (!--call_count)
+ break;
- if (!--call_count)
- break;
+ }
+ }
- }
- }
-
- return 0;
+ return 0;
}
int
afr_symlink_done (call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = frame->local;
+ afr_local_t * local = frame->local;
- local->transaction.unwind (frame, this);
+ local->transaction.unwind (frame, this);
- AFR_STACK_DESTROY (frame);
-
- return 0;
+ AFR_STACK_DESTROY (frame);
+
+ return 0;
}
int
afr_symlink (call_frame_t *frame, xlator_t *this,
- const char *linkpath, loc_t *loc, dict_t *params)
+ const char *linkpath, loc_t *loc, mode_t umask, dict_t *params)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
-
- int ret = -1;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
- int op_ret = -1;
- int op_errno = 0;
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
+ priv = this->private;
- priv = this->private;
+ QUORUM_CHECK(symlink,out);
- transaction_frame = copy_frame (frame);
- if (!transaction_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
- goto out;
- }
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+ local = transaction_frame->local;
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
- transaction_frame->local = local;
-
- loc_copy (&local->loc, loc);
+ loc_copy (&local->loc, loc);
LOCK (&priv->read_child_lock);
{
- local->read_child_index = (++priv->read_child_rr)
+ local->read_child_index = (++priv->read_child_rr)
% (priv->child_count);
}
UNLOCK (&priv->read_child_lock);
- local->cont.symlink.linkpath = gf_strdup (linkpath);
+ local->cont.symlink.linkpath = gf_strdup (linkpath);
+ local->umask = umask;
if (params)
- local->cont.symlink.params = dict_ref (params);
-
- if (loc->parent)
- local->cont.symlink.parent_ino = loc->parent->ino;
-
- local->transaction.fop = afr_symlink_wind;
- local->transaction.done = afr_symlink_done;
- local->transaction.unwind = afr_symlink_unwind;
-
- afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
-
- afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ local->xdata_req = dict_ref (params);
+
+ local->op = GF_FOP_SYMLINK;
+ local->transaction.fop = afr_symlink_wind;
+ local->transaction.done = afr_symlink_done;
+ local->transaction.unwind = afr_symlink_unwind;
+
+ ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+ &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ int_lock->lockee_count++;
+ ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (symlink, frame, op_ret, op_errno,
- NULL, NULL, NULL, NULL);
- }
-
- return 0;
+ if (ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (symlink, frame, -1, op_errno,
+ NULL, NULL, NULL, NULL, NULL);
+ }
+
+ return 0;
}
/* }}} */
@@ -1353,228 +1326,231 @@ out:
int
afr_rename_unwind (call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- struct iatt *unwind_buf = NULL;
-
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- if (local->transaction.main_frame) {
- main_frame = local->transaction.main_frame;
- }
- local->transaction.main_frame = NULL;
- }
- UNLOCK (&frame->lock);
-
- if (main_frame) {
- if (local->cont.rename.read_child_buf.ia_ino) {
- unwind_buf = &local->cont.rename.read_child_buf;
- } else {
- unwind_buf = &local->cont.rename.buf;
- }
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- unwind_buf->ia_ino = local->cont.rename.ino;
+ local = frame->local;
- local->cont.rename.preoldparent.ia_ino = local->cont.rename.oldparent_ino;
- local->cont.rename.postoldparent.ia_ino = local->cont.rename.oldparent_ino;
- local->cont.rename.prenewparent.ia_ino = local->cont.rename.newparent_ino;
- local->cont.rename.postnewparent.ia_ino = local->cont.rename.newparent_ino;
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame) {
+ main_frame = local->transaction.main_frame;
+ }
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK (&frame->lock);
- AFR_STACK_UNWIND (rename, main_frame,
- local->op_ret, local->op_errno,
- unwind_buf,
- &local->cont.rename.preoldparent,
- &local->cont.rename.postoldparent,
- &local->cont.rename.prenewparent,
- &local->cont.rename.postnewparent);
- }
+ if (main_frame) {
+ AFR_STACK_UNWIND (rename, main_frame,
+ local->op_ret, local->op_errno,
+ &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ &local->cont.dir_fop.prenewparent,
+ &local->cont.dir_fop.postnewparent,
+ NULL);
+ }
- return 0;
+ return 0;
}
int
-afr_rename_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
+afr_rename_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent)
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
{
- afr_local_t * local = NULL;
-
- int call_count = -1;
- int child_index = -1;
-
- local = frame->local;
-
- child_index = (long) cookie;
+ afr_local_t * local = NULL;
+ int call_count = -1;
+ int child_index = -1;
- LOCK (&frame->lock);
- {
- if (afr_fop_failed (op_ret, op_errno) && op_errno != ENOTEMPTY)
- afr_transaction_fop_failed (frame, this, child_index);
+ local = frame->local;
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
-
- if (buf) {
- local->cont.rename.buf = *buf;
- }
+ child_index = (long) cookie;
- local->success_count++;
- }
+ LOCK (&frame->lock);
+ {
+ if (afr_fop_failed (op_ret, op_errno) && op_errno != ENOTEMPTY)
+ afr_transaction_fop_failed (frame, this, child_index);
+ local->op_errno = op_errno;
+ local->child_errno[child_index] = op_errno;
- if (child_index == local->read_child_index) {
- local->cont.rename.read_child_buf = *buf;
+ if (op_ret > -1)
+ __dir_entry_fop_common_cbk (frame, child_index, this,
+ op_ret, op_errno, NULL, buf,
+ preoldparent, postoldparent,
+ prenewparent, postnewparent);
- local->cont.rename.preoldparent = *preoldparent;
- local->cont.rename.postoldparent = *postoldparent;
- local->cont.rename.prenewparent = *prenewparent;
- local->cont.rename.postnewparent = *postnewparent;
- }
- }
+ }
+ UNLOCK (&frame->lock);
- local->op_errno = op_errno;
- }
- UNLOCK (&frame->lock);
+ call_count = afr_frame_return (frame);
- call_count = afr_frame_return (frame);
+ if (call_count == 0)
+ afr_dir_fop_done (frame, this);
- if (call_count == 0) {
- local->transaction.unwind (frame, this);
- local->transaction.resume (frame, this);
- }
-
- return 0;
+ return 0;
}
int32_t
afr_rename_wind (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- int call_count = -1;
- int i = 0;
-
- local = frame->local;
- priv = this->private;
-
- call_count = afr_up_children_count (priv->child_count, local->child_up);
-
- if (call_count == 0) {
- local->transaction.resume (frame, this);
- return 0;
- }
-
- local->call_count = call_count;
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_rename_wind_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->rename,
- &local->loc,
- &local->newloc);
- if (!--call_count)
- break;
- }
- }
-
- return 0;
-}
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
+ local = frame->local;
+ priv = this->private;
-int
-afr_rename_done (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t * local = frame->local;
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
- local->transaction.unwind (frame, this);
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
- AFR_STACK_DESTROY (frame);
-
- return 0;
+ local->call_count = call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_rename_wind_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->rename,
+ &local->loc,
+ &local->newloc, NULL);
+ if (!--call_count)
+ break;
+ }
+ }
+
+ return 0;
}
int
-afr_rename (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc)
+afr_rename_done (call_frame_t *frame, xlator_t *this)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
-
- int ret = -1;
-
- int op_ret = -1;
- int op_errno = 0;
-
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
- goto out;
- }
-
- ALLOC_OR_GOTO (local, afr_local_t, out);
-
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- transaction_frame->local = local;
-
- loc_copy (&local->loc, oldloc);
- loc_copy (&local->newloc, newloc);
-
- local->read_child_index = afr_read_child (this, oldloc->inode);
+ afr_local_t * local = frame->local;
- local->cont.rename.ino = oldloc->inode->ino;
+ local->transaction.unwind (frame, this);
- if (oldloc->parent)
- local->cont.rename.oldparent_ino = oldloc->parent->ino;
- if (newloc->parent)
- local->cont.rename.newparent_ino = newloc->parent->ino;
+ AFR_STACK_DESTROY (frame);
- local->transaction.fop = afr_rename_wind;
- local->transaction.done = afr_rename_done;
- local->transaction.unwind = afr_rename_unwind;
+ return 0;
+}
- afr_build_parent_loc (&local->transaction.parent_loc, oldloc);
- afr_build_parent_loc (&local->transaction.new_parent_loc, newloc);
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (oldloc->path);
- local->transaction.new_basename = AFR_BASENAME (newloc->path);
+int
+afr_rename (call_frame_t *frame, xlator_t *this,
+ loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
+ int nlockee = 0;
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
+
+ priv = this->private;
+
+ QUORUM_CHECK(rename,out);
+
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- afr_transaction (transaction_frame, this, AFR_ENTRY_RENAME_TRANSACTION);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+ local = transaction_frame->local;
+
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
+
+ loc_copy (&local->loc, oldloc);
+ loc_copy (&local->newloc, newloc);
+
+ local->read_child_index = afr_inode_get_read_ctx (this, oldloc->inode, NULL);
+
+ local->op = GF_FOP_RENAME;
+ local->transaction.fop = afr_rename_wind;
+ local->transaction.done = afr_rename_done;
+ local->transaction.unwind = afr_rename_unwind;
+
+ ret = afr_build_parent_loc (&local->transaction.parent_loc, oldloc,
+ &op_errno);
+ if (ret)
+ goto out;
+ ret = afr_build_parent_loc (&local->transaction.new_parent_loc, newloc,
+ &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME (oldloc->path);
+ local->transaction.new_basename = AFR_BASENAME (newloc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = nlockee = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+ &local->transaction.new_parent_loc,
+ local->transaction.new_basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ if (local->newloc.inode && IA_ISDIR (local->newloc.inode->ia_type)) {
+ ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+ &local->newloc,
+ NULL,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ }
+ qsort (int_lock->lockee, nlockee, sizeof (*int_lock->lockee),
+ afr_entry_lockee_cmp);
+ int_lock->lockee_count = nlockee;
+
+ ret = afr_transaction (transaction_frame, this, AFR_ENTRY_RENAME_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (rename, frame, op_ret, op_errno,
- NULL, NULL, NULL, NULL, NULL);
- }
+ AFR_STACK_UNWIND (rename, frame, -1, op_errno,
+ NULL, NULL, NULL, NULL, NULL, NULL);
+ }
- return 0;
+ return 0;
}
/* }}} */
@@ -1584,201 +1560,190 @@ out:
int
afr_unlink_unwind (call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- LOCK (&frame->lock);
- {
- if (local->transaction.main_frame) {
- main_frame = local->transaction.main_frame;
- }
- local->transaction.main_frame = NULL;
- }
- UNLOCK (&frame->lock);
-
- if (main_frame) {
- local->cont.unlink.preparent.ia_ino = local->cont.unlink.parent_ino;
- local->cont.unlink.postparent.ia_ino = local->cont.unlink.parent_ino;
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame) {
+ main_frame = local->transaction.main_frame;
+ }
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK (&frame->lock);
- AFR_STACK_UNWIND (unlink, main_frame,
+ if (main_frame) {
+ AFR_STACK_UNWIND (unlink, main_frame,
local->op_ret, local->op_errno,
- &local->cont.unlink.preparent,
- &local->cont.unlink.postparent);
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ NULL);
}
- return 0;
+ return 0;
}
int
-afr_unlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent)
+afr_unlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
-
- int call_count = -1;
- int child_index = (long) cookie;
+ afr_local_t * local = NULL;
+ int call_count = -1;
+ int child_index = (long) cookie;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
- LOCK (&frame->lock);
- {
+ LOCK (&frame->lock);
+ {
if (child_index == local->read_child_index) {
local->read_child_returned = _gf_true;
}
+ __dir_entry_fop_common_cbk (frame, child_index, this,
+ op_ret, op_errno, NULL, NULL,
+ preparent, postparent, NULL, NULL);
+ }
+ UNLOCK (&frame->lock);
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
+ call_count = afr_frame_return (frame);
+ if (call_count == 0)
+ afr_dir_fop_done (frame, this);
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- local->cont.unlink.preparent = *preparent;
- local->cont.unlink.postparent = *postparent;
- }
+ return 0;
+}
- if (child_index == local->read_child_index) {
- local->cont.unlink.preparent = *preparent;
- local->cont.unlink.postparent = *postparent;
- }
- local->success_count++;
- }
+int32_t
+afr_unlink_wind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
- local->op_errno = op_errno;
- }
- UNLOCK (&frame->lock);
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
- call_count = afr_frame_return (frame);
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
- if (call_count == 0) {
- local->transaction.unwind (frame, this);
+ local->call_count = call_count;
- local->transaction.resume (frame, this);
- }
-
- return 0;
-}
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_unlink_wind_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->unlink,
+ &local->loc, local->xflag,
+ local->xdata_req);
+ if (!--call_count)
+ break;
+ }
+ }
-int32_t
-afr_unlink_wind (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- int call_count = -1;
- int i = 0;
-
- local = frame->local;
- priv = this->private;
-
- call_count = afr_up_children_count (priv->child_count, local->child_up);
-
- if (call_count == 0) {
- local->transaction.resume (frame, this);
- return 0;
- }
-
- local->call_count = call_count;
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_unlink_wind_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->unlink,
- &local->loc);
-
- if (!--call_count)
- break;
- }
- }
-
- return 0;
+ return 0;
}
int32_t
afr_unlink_done (call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = frame->local;
+ afr_local_t * local = frame->local;
- local->transaction.unwind (frame, this);
+ local->transaction.unwind (frame, this);
- AFR_STACK_DESTROY (frame);
-
- return 0;
+ AFR_STACK_DESTROY (frame);
+
+ return 0;
}
int32_t
afr_unlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc)
+ loc_t *loc, int xflag, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
-
- int ret = -1;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
- int op_ret = -1;
- int op_errno = 0;
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
+ priv = this->private;
- priv = this->private;
+ QUORUM_CHECK(unlink,out);
- transaction_frame = copy_frame (frame);
- if (!transaction_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
- goto out;
- }
-
- ALLOC_OR_GOTO (local, afr_local_t, out);
-
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- transaction_frame->local = local;
-
- loc_copy (&local->loc, loc);
-
- if (loc->parent)
- local->cont.unlink.parent_ino = loc->parent->ino;
-
- local->transaction.fop = afr_unlink_wind;
- local->transaction.done = afr_unlink_done;
- local->transaction.unwind = afr_unlink_unwind;
-
- afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+ local = transaction_frame->local;
+
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
+
+ loc_copy (&local->loc, loc);
+ local->xflag = xflag;
+ if (xdata)
+ local->xdata_req = dict_ref (xdata);
+
+ local->op = GF_FOP_UNLINK;
+ local->transaction.fop = afr_unlink_wind;
+ local->transaction.done = afr_unlink_done;
+ local->transaction.unwind = afr_unlink_unwind;
+
+ ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+ &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ int_lock->lockee_count++;
+ ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (unlink, frame, op_ret, op_errno,
- NULL, NULL);
- }
-
- return 0;
+ if (ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (unlink, frame, -1, op_errno,
+ NULL, NULL, NULL);
+ }
+
+ return 0;
}
/* }}} */
@@ -1790,203 +1755,208 @@ out:
int
afr_rmdir_unwind (call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- local = frame->local;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- LOCK (&frame->lock);
- {
- if (local->transaction.main_frame) {
- main_frame = local->transaction.main_frame;
- }
- local->transaction.main_frame = NULL;
- }
- UNLOCK (&frame->lock);
+ local = frame->local;
- if (main_frame) {
- local->cont.rmdir.preparent.ia_ino = local->cont.rmdir.parent_ino;
- local->cont.rmdir.postparent.ia_ino = local->cont.rmdir.parent_ino;
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame) {
+ main_frame = local->transaction.main_frame;
+ }
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK (&frame->lock);
- AFR_STACK_UNWIND (rmdir, main_frame,
+ if (main_frame) {
+ AFR_STACK_UNWIND (rmdir, main_frame,
local->op_ret, local->op_errno,
- &local->cont.rmdir.preparent,
- &local->cont.rmdir.postparent);
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ NULL);
}
- return 0;
+ return 0;
}
int
-afr_rmdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent)
+afr_rmdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
-
- int call_count = -1;
- int child_index = (long) cookie;
+ afr_local_t * local = NULL;
+ int call_count = -1;
+ int child_index = (long) cookie;
int read_child = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
- LOCK (&frame->lock);
- {
+ LOCK (&frame->lock);
+ {
if (child_index == read_child) {
local->read_child_returned = _gf_true;
}
+ if (afr_fop_failed (op_ret, op_errno) && (op_errno != ENOTEMPTY))
+ afr_transaction_fop_failed (frame, this, child_index);
+ local->op_errno = op_errno;
+ local->child_errno[child_index] = op_errno;
+ if (op_ret > -1)
+ __dir_entry_fop_common_cbk (frame, child_index, this,
+ op_ret, op_errno, NULL, NULL,
+ preparent, postparent, NULL,
+ NULL);
- if (afr_fop_failed (op_ret, op_errno) && (op_errno != ENOTEMPTY))
- afr_transaction_fop_failed (frame, this, child_index);
+ }
+ UNLOCK (&frame->lock);
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- local->cont.rmdir.preparent = *preparent;
- local->cont.rmdir.postparent = *postparent;
+ call_count = afr_frame_return (frame);
+ if (call_count == 0)
+ afr_dir_fop_done (frame, this);
- }
+ return 0;
+}
- if (child_index == read_child) {
- local->cont.rmdir.preparent = *preparent;
- local->cont.rmdir.postparent = *postparent;
- }
- local->success_count++;
- }
+int
+afr_rmdir_wind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
- local->op_errno = op_errno;
- }
- UNLOCK (&frame->lock);
+ local = frame->local;
+ priv = this->private;
- call_count = afr_frame_return (frame);
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
- if (call_count == 0) {
- local->transaction.unwind (frame, this);
- local->transaction.resume (frame, this);
- }
-
- return 0;
-}
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
+ local->call_count = call_count;
-int
-afr_rmdir_wind (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- int call_count = -1;
- int i = 0;
-
- local = frame->local;
- priv = this->private;
-
- call_count = afr_up_children_count (priv->child_count, local->child_up);
-
- if (call_count == 0) {
- local->transaction.resume (frame, this);
- return 0;
- }
-
- local->call_count = call_count;
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_rmdir_wind_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->rmdir,
- &local->loc);
-
- if (!--call_count)
- break;
- }
- }
-
- return 0;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_rmdir_wind_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->rmdir,
+ &local->loc, local->cont.rmdir.flags,
+ NULL);
+
+ if (!--call_count)
+ break;
+ }
+ }
+
+ return 0;
}
int
afr_rmdir_done (call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = frame->local;
+ afr_local_t * local = frame->local;
- local->transaction.unwind (frame, this);
+ local->transaction.unwind (frame, this);
- AFR_STACK_DESTROY (frame);
-
- return 0;
+ AFR_STACK_DESTROY (frame);
+
+ return 0;
}
int
afr_rmdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc)
+ loc_t *loc, int flags, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
-
- int ret = -1;
-
- int op_ret = -1;
- int op_errno = 0;
-
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
- goto out;
- }
-
- ALLOC_OR_GOTO (local, afr_local_t, out);
-
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- transaction_frame->local = local;
-
- loc_copy (&local->loc, loc);
-
- if (loc->parent)
- local->cont.rmdir.parent_ino = loc->parent->ino;
-
- local->transaction.fop = afr_rmdir_wind;
- local->transaction.done = afr_rmdir_done;
- local->transaction.unwind = afr_rmdir_unwind;
-
- afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
+ int nlockee = 0;
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
+
+ priv = this->private;
+
+ QUORUM_CHECK(rmdir,out);
+
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+ local = transaction_frame->local;
+
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
+
+ local->cont.rmdir.flags = flags;
+ loc_copy (&local->loc, loc);
+
+ local->op = GF_FOP_RMDIR;
+ local->transaction.fop = afr_rmdir_wind;
+ local->transaction.done = afr_rmdir_done;
+ local->transaction.unwind = afr_rmdir_unwind;
+
+ ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+ &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = nlockee = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+ &local->loc,
+ NULL,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ qsort (int_lock->lockee, nlockee, sizeof (*int_lock->lockee),
+ afr_entry_lockee_cmp);
+ int_lock->lockee_count = nlockee;
+
+ ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (rmdir, frame, op_ret, op_errno,
- NULL, NULL);
- }
-
- return 0;
+ if (ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (rmdir, frame, -1, op_errno, NULL, NULL, NULL);
+ }
+
+ return 0;
}
/* }}} */
-
diff --git a/xlators/cluster/afr/src/afr-dir-write.h b/xlators/cluster/afr/src/afr-dir-write.h
index e2ed8c759..02f0a3682 100644
--- a/xlators/cluster/afr/src/afr-dir-write.h
+++ b/xlators/cluster/afr/src/afr-dir-write.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __DIR_WRITE_H__
@@ -23,38 +14,34 @@
int32_t
afr_create (call_frame_t *frame, xlator_t *this,
loc_t *loc, int32_t flags, mode_t mode,
- fd_t *fd, dict_t *params);
+ mode_t umask, fd_t *fd, dict_t *xdata);
int32_t
afr_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t dev, dict_t *params);
+ loc_t *loc, mode_t mode, dev_t dev, mode_t umask, dict_t *xdata);
int32_t
afr_mkdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dict_t *params);
+ loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata);
int32_t
afr_unlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc);
+ loc_t *loc, int xflag, dict_t *xdata);
int32_t
afr_rmdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc);
+ loc_t *loc, int flags, dict_t *xdata);
int32_t
afr_link (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc);
+ loc_t *oldloc, loc_t *newloc, dict_t *xdata);
int32_t
afr_rename (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc);
+ loc_t *oldloc, loc_t *newloc, dict_t *xdata);
int
afr_symlink (call_frame_t *frame, xlator_t *this,
- const char *linkpath, loc_t *oldloc, dict_t *params);
-
-int32_t
-afr_setdents (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int32_t flags, dir_entry_t *entries, int32_t count);
+ const char *linkpath, loc_t *oldloc, mode_t umask, dict_t *params);
#endif /* __DIR_WRITE_H__ */
diff --git a/xlators/cluster/afr/src/afr-inode-read.c b/xlators/cluster/afr/src/afr-inode-read.c
index 8d6e12a93..e06e3b2f2 100644
--- a/xlators/cluster/afr/src/afr-inode-read.c
+++ b/xlators/cluster/afr/src/afr-inode-read.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -44,9 +35,6 @@
#include "compat-errno.h"
#include "compat.h"
-#include "afr.h"
-
-
/**
* Common algorithm for inode read calls:
*
@@ -61,114 +49,115 @@
int32_t
afr_access_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
+ xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- xlator_t ** children = NULL;
-
- int unwind = 1;
- int last_tried = -1;
- int this_try = -1;
- int read_child = -1;
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ xlator_t ** children = NULL;
+ int unwind = 1;
+ int32_t *last_index = NULL;
+ int32_t next_call_child = -1;
+ int32_t read_child = -1;
+ int32_t *fresh_children = NULL;
- priv = this->private;
- children = priv->children;
+ priv = this->private;
+ children = priv->children;
- local = frame->local;
+ local = frame->local;
read_child = (long) cookie;
- if (op_ret == -1) {
- retry:
- last_tried = local->cont.access.last_tried;
-
- if (all_tried (last_tried, priv->child_count)) {
- goto out;
- }
- this_try = ++local->cont.access.last_tried;
-
- if (this_try == read_child) {
- goto retry;
- }
+ if (op_ret == -1) {
+ last_index = &local->cont.access.last_index;
+ fresh_children = local->fresh_children;
+ next_call_child = afr_next_call_child (fresh_children,
+ local->child_up,
+ priv->child_count,
+ last_index, read_child);
+ if (next_call_child < 0)
+ goto out;
- unwind = 0;
+ unwind = 0;
- STACK_WIND_COOKIE (frame, afr_access_cbk,
- (void *) (long) read_child,
- children[this_try],
- children[this_try]->fops->access,
- &local->loc, local->cont.access.mask);
- }
+ STACK_WIND_COOKIE (frame, afr_access_cbk,
+ (void *) (long) read_child,
+ children[next_call_child],
+ children[next_call_child]->fops->access,
+ &local->loc, local->cont.access.mask,
+ NULL);
+ }
out:
- if (unwind) {
- AFR_STACK_UNWIND (access, frame, op_ret, op_errno);
- }
+ if (unwind) {
+ AFR_STACK_UNWIND (access, frame, op_ret, op_errno, xdata);
+ }
- return 0;
+ return 0;
}
int32_t
-afr_access (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t mask)
+afr_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+ dict_t *xdata)
{
- afr_private_t * priv = NULL;
- xlator_t ** children = NULL;
- int call_child = 0;
- afr_local_t *local = NULL;
-
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ int call_child = 0;
+ afr_local_t *local = NULL;
+ int32_t op_errno = 0;
int32_t read_child = -1;
+ int ret = -1;
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
- int32_t op_ret = -1;
- int32_t op_errno = 0;
-
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
+ priv = this->private;
+ VALIDATE_OR_GOTO (priv->children, out);
- priv = this->private;
- VALIDATE_OR_GOTO (priv->children, out);
+ children = priv->children;
- children = priv->children;
+ AFR_SBRAIN_CHECK_LOC (loc, out);
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- read_child = afr_read_child (this, loc->inode);
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
- if ((read_child >= 0) && (priv->child_up[read_child])) {
- call_child = read_child;
+ local->fresh_children = afr_children_create (priv->child_count);
+ if (!local->fresh_children) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- local->cont.access.last_tried = -1;
- } else {
- call_child = afr_first_up_child (priv);
- if (call_child == -1) {
- op_errno = ENOTCONN;
- gf_log (this->name, GF_LOG_DEBUG,
- "no child is up");
- goto out;
- }
-
- local->cont.access.last_tried = call_child;
+ read_child = afr_inode_get_read_ctx (this, loc->inode,
+ local->fresh_children);
+ ret = afr_get_call_child (this, local->child_up, read_child,
+ local->fresh_children,
+ &call_child,
+ &local->cont.access.last_index);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
}
- loc_copy (&local->loc, loc);
- local->cont.access.mask = mask;
+ loc_copy (&local->loc, loc);
+ local->cont.access.mask = mask;
- STACK_WIND_COOKIE (frame, afr_access_cbk,
- (void *) (long) call_child,
- children[call_child], children[call_child]->fops->access,
- loc, mask);
+ STACK_WIND_COOKIE (frame, afr_access_cbk,
+ (void *) (long) call_child,
+ children[call_child],
+ children[call_child]->fops->access,
+ loc, mask, xdata);
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (access, frame, op_ret, op_errno);
- }
- return 0;
+ if (ret < 0)
+ AFR_STACK_UNWIND (access, frame, -1, op_errno, NULL);
+ return 0;
}
@@ -178,121 +167,111 @@ out:
int32_t
afr_stat_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct iatt *buf)
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ struct iatt *buf, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- xlator_t ** children = NULL;
-
- int unwind = 1;
- int last_tried = -1;
- int this_try = -1;
- int read_child = -1;
-
- priv = this->private;
- children = priv->children;
-
- read_child = (long) cookie;
-
- local = frame->local;
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ xlator_t ** children = NULL;
+ int unwind = 1;
+ int32_t *last_index = NULL;
+ int32_t next_call_child = -1;
+ int32_t read_child = -1;
+ int32_t *fresh_children = NULL;
+
+ priv = this->private;
+ children = priv->children;
- if (op_ret == -1) {
- retry:
- last_tried = local->cont.stat.last_tried;
+ read_child = (long) cookie;
- if (all_tried (last_tried, priv->child_count)) {
- goto out;
- }
- this_try = ++local->cont.stat.last_tried;
+ local = frame->local;
- if (this_try == read_child) {
- goto retry;
- }
+ if (op_ret == -1) {
+ last_index = &local->cont.stat.last_index;
+ fresh_children = local->fresh_children;
+ next_call_child = afr_next_call_child (fresh_children,
+ local->child_up,
+ priv->child_count,
+ last_index, read_child);
+ if (next_call_child < 0)
+ goto out;
- unwind = 0;
+ unwind = 0;
- STACK_WIND_COOKIE (frame, afr_stat_cbk,
- (void *) (long) read_child,
- children[this_try],
- children[this_try]->fops->stat,
- &local->loc);
- }
+ STACK_WIND_COOKIE (frame, afr_stat_cbk,
+ (void *) (long) read_child,
+ children[next_call_child],
+ children[next_call_child]->fops->stat,
+ &local->loc, NULL);
+ }
out:
- if (unwind) {
- if (buf)
- buf->ia_ino = local->cont.stat.ino;
-
- AFR_STACK_UNWIND (stat, frame, op_ret, op_errno, buf);
- }
+ if (unwind) {
+ AFR_STACK_UNWIND (stat, frame, op_ret, op_errno, buf, xdata);
+ }
- return 0;
+ return 0;
}
int32_t
-afr_stat (call_frame_t *frame, xlator_t *this,
- loc_t *loc)
+afr_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- xlator_t ** children = NULL;
-
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ xlator_t **children = NULL;
+ int call_child = 0;
+ int32_t op_errno = 0;
int32_t read_child = -1;
- int call_child = 0;
-
- int32_t op_ret = -1;
- int32_t op_errno = 0;
-
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
-
- priv = this->private;
- VALIDATE_OR_GOTO (priv->children, out);
-
- children = priv->children;
-
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ int ret = -1;
- frame->local = local;
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
- read_child = afr_read_child (this, loc->inode);
+ priv = this->private;
+ VALIDATE_OR_GOTO (priv->children, out);
- if ((read_child >= 0) && (priv->child_up[read_child])) {
- call_child = read_child;
+ children = priv->children;
- local->cont.stat.last_tried = -1;
+ AFR_SBRAIN_CHECK_LOC (loc, out);
- } else {
- call_child = afr_first_up_child (priv);
- if (call_child == -1) {
- op_errno = ENOTCONN;
- gf_log (this->name, GF_LOG_DEBUG,
- "no child is up");
- goto out;
- }
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- local->cont.stat.last_tried = call_child;
- }
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
- loc_copy (&local->loc, loc);
+ local->fresh_children = afr_children_create (priv->child_count);
+ if (!local->fresh_children) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- local->cont.stat.ino = loc->inode->ino;
+ read_child = afr_inode_get_read_ctx (this, loc->inode,
+ local->fresh_children);
+ ret = afr_get_call_child (this, local->child_up, read_child,
+ local->fresh_children,
+ &call_child,
+ &local->cont.stat.last_index);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+ loc_copy (&local->loc, loc);
- STACK_WIND_COOKIE (frame, afr_stat_cbk, (void *) (long) call_child,
- children[call_child],
- children[call_child]->fops->stat,
- loc);
+ STACK_WIND_COOKIE (frame, afr_stat_cbk, (void *) (long) call_child,
+ children[call_child],
+ children[call_child]->fops->stat,
+ loc, xdata);
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (stat, frame, op_ret, op_errno, NULL);
- }
+ if (ret < 0)
+ AFR_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
@@ -301,124 +280,122 @@ out:
/* {{{ fstat */
int32_t
-afr_fstat_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct iatt *buf)
+afr_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- xlator_t ** children = NULL;
-
- int unwind = 1;
- int last_tried = -1;
- int this_try = -1;
- int read_child = -1;
-
- priv = this->private;
- children = priv->children;
-
- local = frame->local;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ xlator_t **children = NULL;
+ int unwind = 1;
+ int32_t *last_index = NULL;
+ int32_t next_call_child = -1;
+ int32_t read_child = -1;
+ int32_t *fresh_children = NULL;
- read_child = (long) cookie;
+ priv = this->private;
+ children = priv->children;
- if (op_ret == -1) {
- retry:
- last_tried = local->cont.fstat.last_tried;
+ local = frame->local;
- if (all_tried (last_tried, priv->child_count)) {
- goto out;
- }
- this_try = ++local->cont.fstat.last_tried;
+ read_child = (long) cookie;
- if (this_try == read_child) {
- goto retry;
- }
+ if (op_ret == -1) {
+ last_index = &local->cont.fstat.last_index;
+ fresh_children = local->fresh_children;
+ next_call_child = afr_next_call_child (fresh_children,
+ local->child_up,
+ priv->child_count,
+ last_index, read_child);
+ if (next_call_child < 0)
+ goto out;
- unwind = 0;
+ unwind = 0;
- STACK_WIND_COOKIE (frame, afr_fstat_cbk,
- (void *) (long) read_child,
- children[this_try],
- children[this_try]->fops->fstat,
- local->fd);
- }
+ STACK_WIND_COOKIE (frame, afr_fstat_cbk,
+ (void *) (long) read_child,
+ children[next_call_child],
+ children[next_call_child]->fops->fstat,
+ local->fd, NULL);
+ }
out:
- if (unwind) {
- if (buf)
- buf->ia_ino = local->cont.fstat.ino;
-
- AFR_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf);
- }
+ if (unwind) {
+ AFR_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf, xdata);
+ }
- return 0;
+ return 0;
}
int32_t
afr_fstat (call_frame_t *frame, xlator_t *this,
- fd_t *fd)
+ fd_t *fd, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- xlator_t ** children = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ xlator_t **children = NULL;
+ int call_child = 0;
+ int32_t op_errno = 0;
+ int32_t read_child = 0;
+ int ret = -1;
- int call_child = 0;
- int32_t read_child = -1;
-
- int32_t op_ret = -1;
- int32_t op_errno = 0;
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (fd, out);
+ VALIDATE_OR_GOTO (this->private, out);
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (fd, out);
- VALIDATE_OR_GOTO (this->private, out);
+ priv = this->private;
+ VALIDATE_OR_GOTO (priv->children, out);
- priv = this->private;
- VALIDATE_OR_GOTO (priv->children, out);
+ children = priv->children;
- children = priv->children;
+ VALIDATE_OR_GOTO (fd->inode, out);
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_SBRAIN_CHECK_FD (fd, out);
- frame->local = local;
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- VALIDATE_OR_GOTO (fd->inode, out);
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
- read_child = afr_read_child (this, fd->inode);
+ local->fresh_children = afr_children_create (priv->child_count);
+ if (!local->fresh_children) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- if ((read_child >= 0) && (priv->child_up[read_child])) {
- call_child = read_child;
+ read_child = afr_inode_get_read_ctx (this, fd->inode,
+ local->fresh_children);
- local->cont.fstat.last_tried = -1;
- } else {
- call_child = afr_first_up_child (priv);
- if (call_child == -1) {
- op_errno = ENOTCONN;
- gf_log (this->name, GF_LOG_DEBUG,
- "no child is up");
- goto out;
- }
- local->cont.fstat.last_tried = call_child;
+ ret = afr_get_call_child (this, local->child_up, read_child,
+ local->fresh_children,
+ &call_child,
+ &local->cont.fstat.last_index);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
}
- local->cont.fstat.ino = fd->inode->ino;
- local->fd = fd_ref (fd);
+ local->fd = fd_ref (fd);
+
+ afr_open_fd_fix (fd, this);
- STACK_WIND_COOKIE (frame, afr_fstat_cbk, (void *) (long) call_child,
- children[call_child],
- children[call_child]->fops->fstat,
- fd);
+ STACK_WIND_COOKIE (frame, afr_fstat_cbk, (void *) (long) call_child,
+ children[call_child],
+ children[call_child]->fops->fstat,
+ fd, xdata);
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (fstat, frame, op_ret, op_errno, NULL);
- }
+ if (ret < 0)
+ AFR_STACK_UNWIND (fstat, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
/* }}} */
@@ -427,122 +404,115 @@ out:
int32_t
afr_readlink_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- const char *buf, struct iatt *sbuf)
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ const char *buf, struct iatt *sbuf, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- xlator_t ** children = NULL;
-
- int unwind = 1;
- int last_tried = -1;
- int this_try = -1;
- int read_child = -1;
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ xlator_t ** children = NULL;
+ int unwind = 1;
+ int32_t *last_index = NULL;
+ int32_t next_call_child = -1;
+ int32_t read_child = -1;
+ int32_t *fresh_children = NULL;
- priv = this->private;
- children = priv->children;
+ priv = this->private;
+ children = priv->children;
- local = frame->local;
+ local = frame->local;
read_child = (long) cookie;
- if (op_ret == -1) {
- retry:
- last_tried = local->cont.readlink.last_tried;
-
- if (all_tried (last_tried, priv->child_count)) {
- goto out;
- }
- this_try = ++local->cont.readlink.last_tried;
-
- if (this_try == read_child) {
- goto retry;
- }
+ if (op_ret == -1) {
+ last_index = &local->cont.readlink.last_index;
+ fresh_children = local->fresh_children;
+ next_call_child = afr_next_call_child (fresh_children,
+ local->child_up,
+ priv->child_count,
+ last_index, read_child);
+ if (next_call_child < 0)
+ goto out;
- unwind = 0;
- STACK_WIND_COOKIE (frame, afr_readlink_cbk,
- (void *) (long) read_child,
- children[this_try],
- children[this_try]->fops->readlink,
- &local->loc,
- local->cont.readlink.size);
- }
+ unwind = 0;
+ STACK_WIND_COOKIE (frame, afr_readlink_cbk,
+ (void *) (long) read_child,
+ children[next_call_child],
+ children[next_call_child]->fops->readlink,
+ &local->loc,
+ local->cont.readlink.size, NULL);
+ }
out:
- if (unwind) {
- if (sbuf)
- sbuf->ia_ino = local->cont.readlink.ino;
-
- AFR_STACK_UNWIND (readlink, frame, op_ret, op_errno, buf, sbuf);
- }
+ if (unwind) {
+ AFR_STACK_UNWIND (readlink, frame, op_ret, op_errno, buf, sbuf,
+ xdata);
+ }
- return 0;
+ return 0;
}
int32_t
afr_readlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc, size_t size)
+ loc_t *loc, size_t size, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- xlator_t ** children = NULL;
- int call_child = 0;
- afr_local_t *local = NULL;
-
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ int call_child = 0;
+ afr_local_t *local = NULL;
+ int32_t op_errno = 0;
int32_t read_child = -1;
+ int ret = -1;
- int32_t op_ret = -1;
- int32_t op_errno = 0;
-
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
-
- priv = this->private;
- VALIDATE_OR_GOTO (priv->children, out);
-
- children = priv->children;
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ priv = this->private;
+ VALIDATE_OR_GOTO (priv->children, out);
- frame->local = local;
+ children = priv->children;
- read_child = afr_read_child (this, loc->inode);
+ AFR_SBRAIN_CHECK_LOC (loc, out);
- if ((read_child >= 0) && (priv->child_up[read_child])) {
- call_child = read_child;
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- local->cont.readlink.last_tried = -1;
-
- } else {
- call_child = afr_first_up_child (priv);
-
- if (call_child == -1) {
- op_errno = ENOTCONN;
- gf_log (this->name, GF_LOG_DEBUG,
- "no child is up");
- goto out;
- }
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
- local->cont.readlink.last_tried = call_child;
+ local->fresh_children = afr_children_create (priv->child_count);
+ if (!local->fresh_children) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+ read_child = afr_inode_get_read_ctx (this, loc->inode,
+ local->fresh_children);
+ ret = afr_get_call_child (this, local->child_up, read_child,
+ local->fresh_children,
+ &call_child,
+ &local->cont.readlink.last_index);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
}
- loc_copy (&local->loc, loc);
+ loc_copy (&local->loc, loc);
- local->cont.readlink.size = size;
- local->cont.readlink.ino = loc->inode->ino;
+ local->cont.readlink.size = size;
- STACK_WIND_COOKIE (frame, afr_readlink_cbk,
- (void *) (long) call_child,
- children[call_child], children[call_child]->fops->readlink,
- loc, size);
+ STACK_WIND_COOKIE (frame, afr_readlink_cbk,
+ (void *) (long) call_child,
+ children[call_child],
+ children[call_child]->fops->readlink,
+ loc, size, xdata);
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (readlink, frame, op_ret, op_errno, NULL, NULL);
- }
- return 0;
+ if (ret < 0)
+ AFR_STACK_UNWIND (readlink, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
@@ -556,7 +526,7 @@ struct _xattr_key {
};
-void
+int
__gather_xattr_keys (dict_t *dict, char *key, data_t *value,
void *data)
{
@@ -568,23 +538,23 @@ __gather_xattr_keys (dict_t *dict, char *key, data_t *value,
xkey = GF_CALLOC (1, sizeof (*xkey), gf_afr_mt_xattr_key);
if (!xkey)
- return;
+ return -1;
xkey->key = key;
INIT_LIST_HEAD (&xkey->list);
list_add_tail (&xkey->list, list);
}
+ return 0;
}
void
__filter_xattrs (dict_t *dict)
{
- struct list_head keys;
-
- struct _xattr_key *key;
- struct _xattr_key *tmp;
+ struct list_head keys = {0,};
+ struct _xattr_key *key = NULL;
+ struct _xattr_key *tmp = NULL;
INIT_LIST_HEAD (&keys);
@@ -604,129 +574,1222 @@ __filter_xattrs (dict_t *dict)
int32_t
afr_getxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict)
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- xlator_t ** children = NULL;
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ xlator_t ** children = NULL;
+ int unwind = 1;
+ int32_t *last_index = NULL;
+ int32_t next_call_child = -1;
+ int32_t read_child = -1;
+ int32_t *fresh_children = NULL;
- int unwind = 1;
- int last_tried = -1;
- int this_try = -1;
- int read_child = -1;
+ priv = this->private;
+ children = priv->children;
- priv = this->private;
- children = priv->children;
-
- local = frame->local;
+ local = frame->local;
read_child = (long) cookie;
- if (op_ret == -1) {
- retry:
- last_tried = local->cont.getxattr.last_tried;
-
- if (all_tried (last_tried, priv->child_count)) {
- goto out;
- }
- this_try = ++local->cont.getxattr.last_tried;
-
- if (this_try == read_child) {
- goto retry;
- }
+ if (op_ret == -1) {
+ last_index = &local->cont.getxattr.last_index;
+ fresh_children = local->fresh_children;
+ next_call_child = afr_next_call_child (fresh_children,
+ local->child_up,
+ priv->child_count,
+ last_index, read_child);
+ if (next_call_child < 0)
+ goto out;
- unwind = 0;
- STACK_WIND_COOKIE (frame, afr_getxattr_cbk,
- (void *) (long) read_child,
- children[this_try],
- children[this_try]->fops->getxattr,
- &local->loc,
- local->cont.getxattr.name);
- }
+ unwind = 0;
+ STACK_WIND_COOKIE (frame, afr_getxattr_cbk,
+ (void *) (long) read_child,
+ children[next_call_child],
+ children[next_call_child]->fops->getxattr,
+ &local->loc,
+ local->cont.getxattr.name,
+ NULL);
+ }
out:
- if (unwind) {
+ if (unwind) {
if (op_ret >= 0 && dict)
__filter_xattrs (dict);
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
- }
+ AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
+ }
- return 0;
+ return 0;
}
int32_t
-afr_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name)
+afr_getxattr_unwind (call_frame_t *frame, int op_ret, int op_errno,
+ dict_t *dict, dict_t *xdata)
+
+{
+ AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
+ return 0;
+}
+
+int32_t
+afr_fgetxattr_clrlk_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- xlator_t ** children = NULL;
- int call_child = 0;
- afr_local_t * local = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ dict_t *xattr = NULL;
+ char *tmp_report = NULL;
+ char lk_summary[1024] = {0,};
+ int serz_len = 0;
+ int32_t callcnt = 0;
+ long int cky = 0;
+ int ret = 0;
+
+ priv = this->private;
+ children = priv->children;
+
+ local = frame->local;
+ cky = (long) cookie;
+
+ LOCK (&frame->lock);
+ {
+ callcnt = --local->call_count;
+ if (op_ret == -1)
+ local->child_errno[cky] = op_errno;
+
+ if (!local->dict)
+ local->dict = dict_new ();
+ if (local->dict) {
+ ret = dict_get_str (dict, local->cont.getxattr.name,
+ &tmp_report);
+ if (ret)
+ goto unlock;
+ ret = dict_set_dynstr (local->dict,
+ children[cky]->name,
+ gf_strdup (tmp_report));
+ if (ret)
+ goto unlock;
+ }
+ }
+unlock:
+ UNLOCK (&frame->lock);
+
+ if (!callcnt) {
+ xattr = dict_new ();
+ if (!xattr) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ ret = dict_serialize_value_with_delim (local->dict,
+ lk_summary,
+ &serz_len, '\n');
+ if (ret) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ gf_log (this->name, GF_LOG_ERROR,
+ "Error serializing dictionary");
+ goto unwind;
+ }
+ if (serz_len == -1)
+ snprintf (lk_summary, sizeof (lk_summary),
+ "No locks cleared.");
+ ret = dict_set_dynstr (xattr, local->cont.getxattr.name,
+ gf_strdup (lk_summary));
+ if (ret) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ gf_log (this->name, GF_LOG_ERROR,
+ "Error setting dictionary");
+ goto unwind;
+ }
+
+ unwind:
+ // Updating child_errno with more recent 'events'
+ local->child_errno[cky] = op_errno;
+ op_errno = afr_resultant_errno_get (NULL, local->child_errno,
+ priv->child_count);
+ AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, xattr,
+ xdata);
+
+ if (xattr)
+ dict_unref (xattr);
+ }
+
+ return ret;
+}
+
+int32_t
+afr_getxattr_clrlk_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ dict_t *xattr = NULL;
+ char *tmp_report = NULL;
+ char lk_summary[1024] = {0,};
+ int serz_len = 0;
+ int32_t callcnt = 0;
+ long int cky = 0;
+ int ret = 0;
+
+ priv = this->private;
+ children = priv->children;
+
+ local = frame->local;
+ cky = (long) cookie;
+
+ LOCK (&frame->lock);
+ {
+ callcnt = --local->call_count;
+ if (op_ret == -1)
+ local->child_errno[cky] = op_errno;
+
+ if (!local->dict)
+ local->dict = dict_new ();
+ if (local->dict) {
+ ret = dict_get_str (dict, local->cont.getxattr.name,
+ &tmp_report);
+ if (ret)
+ goto unlock;
+ ret = dict_set_dynstr (local->dict,
+ children[cky]->name,
+ gf_strdup (tmp_report));
+ if (ret)
+ goto unlock;
+ }
+ }
+unlock:
+ UNLOCK (&frame->lock);
+
+ if (!callcnt) {
+ xattr = dict_new ();
+ if (!xattr) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ ret = dict_serialize_value_with_delim (local->dict,
+ lk_summary,
+ &serz_len, '\n');
+ if (ret) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ gf_log (this->name, GF_LOG_ERROR,
+ "Error serializing dictionary");
+ goto unwind;
+ }
+ if (serz_len == -1)
+ snprintf (lk_summary, sizeof (lk_summary),
+ "No locks cleared.");
+ ret = dict_set_dynstr (xattr, local->cont.getxattr.name,
+ gf_strdup (lk_summary));
+ if (ret) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ gf_log (this->name, GF_LOG_ERROR,
+ "Error setting dictionary");
+ goto unwind;
+ }
+
+ unwind:
+ // Updating child_errno with more recent 'events'
+ local->child_errno[cky] = op_errno;
+ op_errno = afr_resultant_errno_get (NULL, local->child_errno,
+ priv->child_count);
+ AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr, xdata);
+
+ if (xattr)
+ dict_unref (xattr);
+ }
+
+ return ret;
+}
+
+/**
+ * node-uuid cbk uses next child querying mechanism
+ */
+int32_t
+afr_getxattr_node_uuid_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ xlator_t **children = NULL;
+ int unwind = 1;
+ int curr_call_child = 0;
+
+ priv = this->private;
+ children = priv->children;
+
+ local = frame->local;
+
+ if (op_ret == -1) { /** query the _next_ child */
+
+ /**
+ * _current_ becomes _next_
+ * If done with all childs and yet no success; give up !
+ */
+ curr_call_child = (int) ((long)cookie);
+ if (++curr_call_child == priv->child_count)
+ goto unwind;
+
+ gf_log (this->name, GF_LOG_WARNING,
+ "op_ret (-1): Re-querying afr-child (%d/%d)",
+ curr_call_child, priv->child_count);
+
+ unwind = 0;
+ STACK_WIND_COOKIE (frame, afr_getxattr_node_uuid_cbk,
+ (void *) (long) curr_call_child,
+ children[curr_call_child],
+ children[curr_call_child]->fops->getxattr,
+ &local->loc,
+ local->cont.getxattr.name,
+ NULL);
+ }
+
+ unwind:
+ if (unwind)
+ AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict,
+ NULL);
+
+ return 0;
+}
+
+int32_t
+afr_getxattr_lockinfo_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
+{
+ int call_cnt = 0, len = 0;
+ char *lockinfo_buf = NULL;
+ dict_t *lockinfo = NULL, *newdict = NULL;
+ afr_local_t *local = NULL;
+
+ LOCK (&frame->lock);
+ {
+ local = frame->local;
+
+ call_cnt = --local->call_count;
+
+ if ((op_ret < 0) || (!dict && !xdata)) {
+ goto unlock;
+ }
+
+ if (xdata) {
+ if (!local->xdata_rsp) {
+ local->xdata_rsp = dict_new ();
+ if (!local->xdata_rsp) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unlock;
+ }
+ }
+ }
+
+ if (!dict) {
+ goto unlock;
+ }
+
+ op_ret = dict_get_ptr_and_len (dict, GF_XATTR_LOCKINFO_KEY,
+ (void **)&lockinfo_buf, &len);
+
+ if (!lockinfo_buf) {
+ goto unlock;
+ }
+
+ if (!local->dict) {
+ local->dict = dict_new ();
+ if (!local->dict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unlock;
+ }
+ }
+ }
+unlock:
+ UNLOCK (&frame->lock);
+
+ if (lockinfo_buf != NULL) {
+ lockinfo = dict_new ();
+ if (lockinfo == NULL) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ } else {
+ op_ret = dict_unserialize (lockinfo_buf, len,
+ &lockinfo);
+
+ if (lockinfo && local->dict) {
+ dict_copy (lockinfo, local->dict);
+ }
+ }
+ }
+
+ if (xdata && local->xdata_rsp) {
+ dict_copy (xdata, local->xdata_rsp);
+ }
+
+ if (!call_cnt) {
+ newdict = dict_new ();
+ if (!newdict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ len = dict_serialized_length (local->dict);
+ if (len == 0) {
+ goto unwind;
+ }
+
+ lockinfo_buf = GF_CALLOC (1, len, gf_common_mt_char);
+ if (!lockinfo_buf) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ op_ret = dict_serialize (local->dict, lockinfo_buf);
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -op_ret;
+ }
+
+ op_ret = dict_set_dynptr (newdict, GF_XATTR_LOCKINFO_KEY,
+ (void *)lockinfo_buf, len);
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -op_ret;
+ goto unwind;
+ }
+
+ unwind:
+ AFR_STACK_UNWIND (getxattr, frame, op_ret,
+ op_errno, newdict,
+ local->xdata_rsp);
+ }
+
+ dict_unref (lockinfo);
+
+ return 0;
+}
+
+int32_t
+afr_fgetxattr_lockinfo_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
+{
+ int call_cnt = 0, len = 0;
+ char *lockinfo_buf = NULL;
+ dict_t *lockinfo = NULL, *newdict = NULL;
+ afr_local_t *local = NULL;
+
+ LOCK (&frame->lock);
+ {
+ local = frame->local;
+
+ call_cnt = --local->call_count;
+
+ if ((op_ret < 0) || (!dict && !xdata)) {
+ goto unlock;
+ }
+
+ if (xdata) {
+ if (!local->xdata_rsp) {
+ local->xdata_rsp = dict_new ();
+ if (!local->xdata_rsp) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unlock;
+ }
+ }
+ }
+
+ if (!dict) {
+ goto unlock;
+ }
+
+ op_ret = dict_get_ptr_and_len (dict, GF_XATTR_LOCKINFO_KEY,
+ (void **)&lockinfo_buf, &len);
+
+ if (!lockinfo_buf) {
+ goto unlock;
+ }
+
+ if (!local->dict) {
+ local->dict = dict_new ();
+ if (!local->dict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unlock;
+ }
+ }
+ }
+unlock:
+ UNLOCK (&frame->lock);
+
+ if (lockinfo_buf != NULL) {
+ lockinfo = dict_new ();
+ if (lockinfo == NULL) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ } else {
+ op_ret = dict_unserialize (lockinfo_buf, len,
+ &lockinfo);
+
+ if (lockinfo && local->dict) {
+ dict_copy (lockinfo, local->dict);
+ }
+ }
+ }
+
+ if (xdata && local->xdata_rsp) {
+ dict_copy (xdata, local->xdata_rsp);
+ }
+
+ if (!call_cnt) {
+ newdict = dict_new ();
+ if (!newdict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ len = dict_serialized_length (local->dict);
+ if (len <= 0) {
+ goto unwind;
+ }
- int read_child = -1;
+ lockinfo_buf = GF_CALLOC (1, len, gf_common_mt_char);
+ if (!lockinfo_buf) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
- int32_t op_ret = -1;
- int32_t op_errno = 0;
+ op_ret = dict_serialize (local->dict, lockinfo_buf);
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -op_ret;
+ }
+ op_ret = dict_set_dynptr (newdict, GF_XATTR_LOCKINFO_KEY,
+ (void *)lockinfo_buf, len);
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -op_ret;
+ goto unwind;
+ }
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
+ unwind:
+ AFR_STACK_UNWIND (fgetxattr, frame, op_ret,
+ op_errno, newdict,
+ local->xdata_rsp);
+ }
- priv = this->private;
- VALIDATE_OR_GOTO (priv->children, out);
+ dict_unref (lockinfo);
- children = priv->children;
+ return 0;
+}
- ALLOC_OR_GOTO (local, afr_local_t, out);
- frame->local = local;
+int32_t
+afr_fgetxattr_pathinfo_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ int32_t callcnt = 0;
+ int ret = 0;
+ char *xattr = NULL;
+ char *xattr_serz = NULL;
+ char xattr_cky[1024] = {0,};
+ dict_t *nxattr = NULL;
+ long cky = 0;
+ int32_t padding = 0;
+ int32_t tlen = 0;
+
+ if (!frame || !frame->local || !this) {
+ gf_log ("", GF_LOG_ERROR, "possible NULL deref");
+ goto out;
+ }
- if (name) {
- if (!strncmp (name, AFR_XATTR_PREFIX,
- strlen (AFR_XATTR_PREFIX))) {
+ local = frame->local;
+ cky = (long) cookie;
- op_errno = ENODATA;
+ LOCK (&frame->lock);
+ {
+ callcnt = --local->call_count;
+
+ if (!dict || (op_ret < 0))
goto out;
+
+ if (!local->dict)
+ local->dict = dict_new ();
+
+ if (local->dict) {
+ ret = dict_get_str (dict,
+ local->cont.getxattr.name,
+ &xattr);
+ if (ret)
+ goto out;
+
+ xattr = gf_strdup (xattr);
+
+ (void)snprintf (xattr_cky, 1024, "%s-%ld",
+ local->cont.getxattr.name, cky);
+ ret = dict_set_dynstr (local->dict,
+ xattr_cky, xattr);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "Cannot set xattr cookie key");
+ goto out;
+ }
+
+ local->cont.getxattr.xattr_len
+ += strlen (xattr) + 1;
+ }
+ }
+out:
+ UNLOCK (&frame->lock);
+
+ if (!callcnt) {
+ if (!local->cont.getxattr.xattr_len)
+ goto unwind;
+
+ nxattr = dict_new ();
+ if (!nxattr)
+ goto unwind;
+
+ /* extra bytes for decorations (brackets and <>'s) */
+ padding += strlen (this->name)
+ + strlen (AFR_PATHINFO_HEADER) + 4;
+ local->cont.getxattr.xattr_len += (padding + 2);
+
+ xattr_serz = GF_CALLOC (local->cont.getxattr.xattr_len,
+ sizeof (char), gf_common_mt_char);
+
+ if (!xattr_serz)
+ goto unwind;
+
+ /* the xlator info */
+ (void) sprintf (xattr_serz, "(<"AFR_PATHINFO_HEADER"%s> ",
+ this->name);
+
+ /* actual series of pathinfo */
+ ret = dict_serialize_value_with_delim (local->dict,
+ xattr_serz
+ + strlen (xattr_serz),
+ &tlen, ' ');
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Error serializing"
+ " dictionary");
+ goto unwind;
+ }
+
+ /* closing part */
+ *(xattr_serz + padding + tlen) = ')';
+ *(xattr_serz + padding + tlen + 1) = '\0';
+
+ ret = dict_set_dynstr (nxattr, local->cont.getxattr.name,
+ xattr_serz);
+ if (ret)
+ gf_log (this->name, GF_LOG_ERROR, "Cannot set pathinfo"
+ " key in dict");
+
+ unwind:
+ AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, nxattr,
+ xdata);
+
+ if (nxattr)
+ dict_unref (nxattr);
+ }
+
+ return ret;
+}
+
+int32_t
+afr_getxattr_pathinfo_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ int32_t callcnt = 0;
+ int ret = 0;
+ char *xattr = NULL;
+ char *xattr_serz = NULL;
+ char xattr_cky[1024] = {0,};
+ dict_t *nxattr = NULL;
+ long cky = 0;
+ int32_t padding = 0;
+ int32_t tlen = 0;
+
+ if (!frame || !frame->local || !this) {
+ gf_log ("", GF_LOG_ERROR, "possible NULL deref");
+ goto out;
+ }
+
+ local = frame->local;
+ cky = (long) cookie;
+
+ LOCK (&frame->lock);
+ {
+ callcnt = --local->call_count;
+
+ if (!dict || (op_ret < 0))
+ goto out;
+
+ if (!local->dict)
+ local->dict = dict_new ();
+
+ if (local->dict) {
+ ret = dict_get_str (dict,
+ local->cont.getxattr.name,
+ &xattr);
+ if (ret)
+ goto out;
+
+ xattr = gf_strdup (xattr);
+
+ (void)snprintf (xattr_cky, 1024, "%s-%ld",
+ local->cont.getxattr.name, cky);
+ ret = dict_set_dynstr (local->dict,
+ xattr_cky, xattr);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "Cannot set xattr cookie key");
+ goto out;
+ }
+
+ local->cont.getxattr.xattr_len += strlen (xattr) + 1;
+ }
+ }
+ out:
+ UNLOCK (&frame->lock);
+
+ if (!callcnt) {
+ if (!local->cont.getxattr.xattr_len)
+ goto unwind;
+
+ nxattr = dict_new ();
+ if (!nxattr)
+ goto unwind;
+
+ /* extra bytes for decorations (brackets and <>'s) */
+ padding += strlen (this->name) + strlen (AFR_PATHINFO_HEADER) + 4;
+ local->cont.getxattr.xattr_len += (padding + 2);
+
+ xattr_serz = GF_CALLOC (local->cont.getxattr.xattr_len,
+ sizeof (char), gf_common_mt_char);
+
+ if (!xattr_serz)
+ goto unwind;
+
+ /* the xlator info */
+ (void) sprintf (xattr_serz, "(<"AFR_PATHINFO_HEADER"%s> ",
+ this->name);
+
+ /* actual series of pathinfo */
+ ret = dict_serialize_value_with_delim (local->dict,
+ xattr_serz + strlen (xattr_serz),
+ &tlen, ' ');
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Error serializing"
+ " dictionary");
+ goto unwind;
+ }
+
+ /* closing part */
+ *(xattr_serz + padding + tlen) = ')';
+ *(xattr_serz + padding + tlen + 1) = '\0';
+
+ ret = dict_set_dynstr (nxattr, local->cont.getxattr.name,
+ xattr_serz);
+ if (ret)
+ gf_log (this->name, GF_LOG_ERROR, "Cannot set pathinfo"
+ " key in dict");
+
+ unwind:
+ AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, nxattr,
+ xdata);
+
+ if (nxattr)
+ dict_unref (nxattr);
+ }
+
+ return ret;
+}
+
+static int
+afr_aggregate_stime_xattr (dict_t *this, char *key, data_t *value, void *data)
+{
+ int ret = 0;
+
+ if (fnmatch (GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0)
+ ret = gf_get_min_stime (THIS, data, key, value);
+
+ return ret;
+}
+
+int32_t
+afr_common_getxattr_stime_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ int32_t callcnt = 0;
+
+ if (!frame || !frame->local || !this) {
+ gf_log ("", GF_LOG_ERROR, "possible NULL deref");
+ goto out;
+ }
+
+ local = frame->local;
+
+ LOCK (&frame->lock);
+ {
+ callcnt = --local->call_count;
+
+ if (!dict || (op_ret < 0)) {
+ local->op_errno = op_errno;
+ goto cleanup;
}
+ if (!local->dict)
+ local->dict = dict_copy_with_ref (dict, NULL);
+ else
+ dict_foreach (dict, afr_aggregate_stime_xattr,
+ local->dict);
+ local->op_ret = 0;
+ }
+
+cleanup:
+ UNLOCK (&frame->lock);
+
+ if (!callcnt) {
+ AFR_STACK_UNWIND (getxattr, frame, local->op_ret,
+ local->op_errno, local->dict, xdata);
}
- read_child = afr_read_child (this, loc->inode);
+out:
+ return 0;
+}
- if ((read_child >= 0) && (priv->child_up[read_child])) {
- call_child = read_child;
- local->cont.getxattr.last_tried = -1;
+static gf_boolean_t
+afr_is_special_xattr (const char *name, fop_getxattr_cbk_t *cbk,
+ gf_boolean_t is_fgetxattr)
+{
+ gf_boolean_t is_spl = _gf_true;
+
+ GF_ASSERT (cbk);
+ if (!cbk) {
+ is_spl = _gf_false;
+ goto out;
+ }
+
+ if (!strcmp (name, GF_XATTR_PATHINFO_KEY)) {
+ if (is_fgetxattr) {
+ *cbk = afr_fgetxattr_pathinfo_cbk;
+ } else {
+ *cbk = afr_getxattr_pathinfo_cbk;
+ }
+ } else if (!strncmp (name, GF_XATTR_CLRLK_CMD,
+ strlen (GF_XATTR_CLRLK_CMD))) {
+ if (is_fgetxattr) {
+ *cbk = afr_fgetxattr_clrlk_cbk;
+ } else {
+ *cbk = afr_getxattr_clrlk_cbk;
+ }
+ } else if (!strncmp (name, GF_XATTR_LOCKINFO_KEY,
+ strlen (GF_XATTR_LOCKINFO_KEY))) {
+ if (is_fgetxattr) {
+ *cbk = afr_fgetxattr_lockinfo_cbk;
+ } else {
+ *cbk = afr_getxattr_lockinfo_cbk;
+ }
+ } else if (fnmatch (GF_XATTR_STIME_PATTERN, name, FNM_NOESCAPE) == 0) {
+ *cbk = afr_common_getxattr_stime_cbk;
} else {
- call_child = afr_first_up_child (priv);
+ is_spl = _gf_false;
+ }
+
+out:
+ return is_spl;
+}
+
+static void
+afr_getxattr_frm_all_children (xlator_t *this, call_frame_t *frame,
+ const char *name, loc_t *loc,
+ fop_getxattr_cbk_t cbk)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ xlator_t **children = NULL;
+ int i = 0;
+
+ priv = this->private;
+ children = priv->children;
+
+ local = frame->local;
+ local->call_count = priv->child_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ STACK_WIND_COOKIE (frame, cbk,
+ (void *) (long) i,
+ children[i], children[i]->fops->getxattr,
+ loc, name, NULL);
+ }
+ return;
+}
+
+int32_t
+afr_getxattr (call_frame_t *frame, xlator_t *this,
+ loc_t *loc, const char *name, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ int call_child = 0;
+ afr_local_t *local = NULL;
+ xlator_list_t *trav = NULL;
+ xlator_t **sub_volumes = NULL;
+ int i = 0;
+ int32_t op_errno = 0;
+ int32_t read_child = -1;
+ int ret = -1;
+ fop_getxattr_cbk_t cbk = NULL;
+ int afr_xtime_gauge[MCNT_MAX] = {0,};
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
+
+ priv = this->private;
+ VALIDATE_OR_GOTO (priv->children, out);
+
+ children = priv->children;
+
+ AFR_SBRAIN_CHECK_LOC (loc, out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
+
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
+
+ loc_copy (&local->loc, loc);
+ if (!name)
+ goto no_name;
+
+ local->cont.getxattr.name = gf_strdup (name);
- if (call_child == -1) {
- op_errno = ENOTCONN;
- gf_log (this->name, GF_LOG_DEBUG,
- "no child is up");
+ if (!strncmp (name, AFR_XATTR_PREFIX,
+ strlen (AFR_XATTR_PREFIX))) {
+ gf_log (this->name, GF_LOG_INFO,
+ "%s: no data present for key %s",
+ loc->path, name);
+ op_errno = ENODATA;
+ goto out;
+ }
+ if ((strcmp (GF_XATTR_MARKER_KEY, name) == 0)
+ && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+
+ local->marker.call_count = priv->child_count;
+
+ sub_volumes = alloca ( priv->child_count * sizeof (xlator_t *));
+ for (i = 0, trav = this->children; trav ;
+ trav = trav->next, i++) {
+
+ *(sub_volumes + i) = trav->xlator;
+ }
+
+ if (cluster_getmarkerattr (frame, this, loc, name,
+ local, afr_getxattr_unwind,
+ sub_volumes,
+ priv->child_count,
+ MARKER_UUID_TYPE,
+ marker_uuid_default_gauge,
+ priv->vol_uuid)) {
+
+ gf_log (this->name, GF_LOG_INFO,
+ "%s: failed to get marker attr (%s)",
+ loc->path, name);
+ op_errno = EINVAL;
goto out;
}
- local->cont.getxattr.last_tried = call_child;
+ return 0;
+ }
+
+ /*
+ * if we are doing getxattr with pathinfo as the key then we
+ * collect information from all childs
+ */
+ if (afr_is_special_xattr (name, &cbk, 0)) {
+ afr_getxattr_frm_all_children (this, frame, name,
+ loc, cbk);
+ return 0;
+ }
+
+ if (XATTR_IS_NODE_UUID (name)) {
+ i = 0;
+ STACK_WIND_COOKIE (frame, afr_getxattr_node_uuid_cbk,
+ (void *) (long) i,
+ children[i],
+ children[i]->fops->getxattr,
+ loc, name, xdata);
+ return 0;
+ }
+
+ if (*priv->vol_uuid) {
+ if ((match_uuid_local (name, priv->vol_uuid) == 0)
+ && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+ local->marker.call_count = priv->child_count;
+
+ sub_volumes = alloca ( priv->child_count
+ * sizeof (xlator_t *));
+ for (i = 0, trav = this->children; trav ;
+ trav = trav->next, i++) {
+
+ *(sub_volumes + i) = trav->xlator;
+
+ }
+
+ /* don't err out on getting ENOTCONN (brick down)
+ * from a subset of the bricks
+ */
+ memcpy (afr_xtime_gauge, marker_xtime_default_gauge,
+ sizeof (afr_xtime_gauge));
+ afr_xtime_gauge[MCNT_NOTFOUND] = 0;
+ afr_xtime_gauge[MCNT_ENOTCONN] = 0;
+ if (cluster_getmarkerattr (frame, this, loc,
+ name, local,
+ afr_getxattr_unwind,
+ sub_volumes,
+ priv->child_count,
+ MARKER_XTIME_TYPE,
+ afr_xtime_gauge,
+ priv->vol_uuid)) {
+ gf_log (this->name, GF_LOG_INFO,
+ "%s: failed to get marker attr (%s)",
+ loc->path, name);
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ return 0;
+ }
+ }
+
+no_name:
+ local->fresh_children = afr_children_create (priv->child_count);
+ if (!local->fresh_children) {
+ op_errno = ENOMEM;
+ goto out;
}
- loc_copy (&local->loc, loc);
- if (name)
- local->cont.getxattr.name = gf_strdup (name);
+ read_child = afr_inode_get_read_ctx (this, loc->inode,
+ local->fresh_children);
+ ret = afr_get_call_child (this, local->child_up, read_child,
+ local->fresh_children,
+ &call_child,
+ &local->cont.getxattr.last_index);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ STACK_WIND_COOKIE (frame, afr_getxattr_cbk,
+ (void *) (long) call_child,
+ children[call_child],
+ children[call_child]->fops->getxattr,
+ loc, name, xdata);
+
+ ret = 0;
+out:
+ if (ret < 0)
+ AFR_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
+ return 0;
+}
+
+/* {{{ fgetxattr */
- STACK_WIND_COOKIE (frame, afr_getxattr_cbk,
- (void *) (long) call_child,
- children[call_child], children[call_child]->fops->getxattr,
- loc, name);
- op_ret = 0;
+int32_t
+afr_fgetxattr_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
+{
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ xlator_t ** children = NULL;
+ int unwind = 1;
+ int32_t *last_index = NULL;
+ int32_t next_call_child = -1;
+ int32_t read_child = -1;
+ int32_t *fresh_children = NULL;
+
+ priv = this->private;
+ children = priv->children;
+
+ local = frame->local;
+
+ read_child = (long) cookie;
+
+ if (op_ret == -1) {
+ last_index = &local->cont.getxattr.last_index;
+ fresh_children = local->fresh_children;
+ next_call_child = afr_next_call_child (fresh_children,
+ local->child_up,
+ priv->child_count,
+ last_index, read_child);
+ if (next_call_child < 0)
+ goto out;
+
+ unwind = 0;
+ STACK_WIND_COOKIE (frame, afr_fgetxattr_cbk,
+ (void *) (long) read_child,
+ children[next_call_child],
+ children[next_call_child]->fops->fgetxattr,
+ local->fd,
+ local->cont.getxattr.name,
+ NULL);
+ }
+
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, NULL);
- }
- return 0;
+ if (unwind) {
+ if (op_ret >= 0 && dict)
+ __filter_xattrs (dict);
+
+ AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, dict,
+ xdata);
+ }
+
+ return 0;
+}
+
+int32_t
+afr_fgetxattr_unwind (call_frame_t *frame,
+ int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
+
+{
+ AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, dict, xdata);
+ return 0;
+}
+
+static void
+afr_fgetxattr_frm_all_children (xlator_t *this, call_frame_t *frame,
+ const char *name, fd_t *fd,
+ fop_fgetxattr_cbk_t cbk)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ xlator_t **children = NULL;
+ int i = 0;
+
+ priv = this->private;
+ children = priv->children;
+
+ local = frame->local;
+ local->call_count = priv->child_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ STACK_WIND_COOKIE (frame, cbk,
+ (void *) (long) i,
+ children[i], children[i]->fops->fgetxattr,
+ fd, name, NULL);
+ }
+
+ return;
+}
+
+int32_t
+afr_fgetxattr (call_frame_t *frame, xlator_t *this,
+ fd_t *fd, const char *name, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ int call_child = 0;
+ afr_local_t *local = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = 0;
+ int32_t read_child = -1;
+ fop_fgetxattr_cbk_t cbk = NULL;
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
+
+ priv = this->private;
+ VALIDATE_OR_GOTO (priv->children, out);
+
+ children = priv->children;
+
+ AFR_SBRAIN_CHECK_FD (fd, out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (local, out);
+ frame->local = local;
+
+ op_ret = afr_local_init (local, priv, &op_errno);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto out;
+ }
+
+ local->fd = fd_ref (fd);
+ if (name)
+ local->cont.getxattr.name = gf_strdup (name);
+
+ /* pathinfo gets handled only in getxattr(), but we need to handle
+ * lockinfo.
+ * If we are doing fgetxattr with lockinfo as the key then we
+ * collect information from all children.
+ */
+ if (afr_is_special_xattr (name, &cbk, 1)) {
+ afr_fgetxattr_frm_all_children (this, frame, name,
+ fd, cbk);
+ return 0;
+ }
+
+
+ local->fresh_children = afr_children_create (priv->child_count);
+ if (!local->fresh_children) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ read_child = afr_inode_get_read_ctx (this, fd->inode,
+ local->fresh_children);
+ op_ret = afr_get_call_child (this, local->child_up, read_child,
+ local->fresh_children,
+ &call_child,
+ &local->cont.getxattr.last_index);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ op_ret = -1;
+ goto out;
+ }
+
+ STACK_WIND_COOKIE (frame, afr_fgetxattr_cbk,
+ (void *) (long) call_child,
+ children[call_child],
+ children[call_child]->fops->fgetxattr,
+ fd, name, xdata);
+
+ op_ret = 0;
+out:
+ if (op_ret == -1) {
+ AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, NULL,
+ NULL);
+ }
+ return 0;
}
@@ -748,140 +1811,130 @@ out:
int32_t
afr_readv_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct iovec *vector, int32_t count, struct iatt *buf,
- struct iobref *iobref)
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ struct iovec *vector, int32_t count, struct iatt *buf,
+ struct iobref *iobref, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- xlator_t ** children = NULL;
-
- int unwind = 1;
- int last_tried = -1;
- int this_try = -1;
- int read_child = -1;
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ xlator_t ** children = NULL;
+ int unwind = 1;
+ int32_t *last_index = NULL;
+ int32_t next_call_child = -1;
+ int32_t *fresh_children = NULL;
+ int32_t read_child = -1;
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
- priv = this->private;
- VALIDATE_OR_GOTO (priv->children, out);
+ priv = this->private;
+ VALIDATE_OR_GOTO (priv->children, out);
- children = priv->children;
+ children = priv->children;
- local = frame->local;
+ local = frame->local;
read_child = (long) cookie;
- if (op_ret == -1) {
- retry:
- last_tried = local->cont.readv.last_tried;
-
- if (all_tried (last_tried, priv->child_count)) {
- goto out;
- }
- this_try = ++local->cont.readv.last_tried;
-
- if (this_try == read_child) {
- /*
- skip the read child since if we are here
- we must have already tried that child
- */
- goto retry;
- }
-
- unwind = 0;
-
- STACK_WIND_COOKIE (frame, afr_readv_cbk,
- (void *) (long) read_child,
- children[this_try],
- children[this_try]->fops->readv,
- local->fd, local->cont.readv.size,
- local->cont.readv.offset);
- }
+ if (op_ret == -1) {
+ last_index = &local->cont.readv.last_index;
+ fresh_children = local->fresh_children;
+ next_call_child = afr_next_call_child (fresh_children,
+ local->child_up,
+ priv->child_count,
+ last_index, read_child);
+ if (next_call_child < 0)
+ goto out;
-out:
- if (unwind) {
- if (buf && local)
- buf->ia_ino = local->cont.readv.ino;
+ unwind = 0;
- AFR_STACK_UNWIND (readv, frame, op_ret, op_errno,
- vector, count, buf, iobref);
- }
+ STACK_WIND_COOKIE (frame, afr_readv_cbk,
+ (void *) (long) read_child,
+ children[next_call_child],
+ children[next_call_child]->fops->readv,
+ local->fd, local->cont.readv.size,
+ local->cont.readv.offset,
+ local->cont.readv.flags,
+ NULL);
+ }
- return 0;
+out:
+ if (unwind) {
+ AFR_STACK_UNWIND (readv, frame, op_ret, op_errno,
+ vector, count, buf, iobref, xdata);
+ }
+
+ return 0;
}
int32_t
afr_readv (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset)
+ fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- xlator_t ** children = NULL;
-
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ xlator_t ** children = NULL;
+ int call_child = 0;
+ int32_t op_errno = 0;
int32_t read_child = -1;
- int call_child = 0;
+ int ret = -1;
- int32_t op_ret = -1;
- int32_t op_errno = 0;
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
+ VALIDATE_OR_GOTO (fd, out);
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
- VALIDATE_OR_GOTO (fd, out);
+ priv = this->private;
+ children = priv->children;
- priv = this->private;
- children = priv->children;
+ AFR_SBRAIN_CHECK_FD (fd, out);
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- frame->local = local;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
- read_child = afr_read_child (this, fd->inode);
+ local->fresh_children = afr_children_create (priv->child_count);
+ if (!local->fresh_children) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- if ((read_child >= 0) && (priv->child_up[read_child])) {
- call_child = read_child;
+ read_child = afr_inode_get_read_ctx (this, fd->inode, local->fresh_children);
+ ret = afr_get_call_child (this, local->child_up, read_child,
+ local->fresh_children,
+ &call_child,
+ &local->cont.readv.last_index);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- /*
- if read fails from the read child, we try
- all children starting with the first one
- */
- local->cont.readv.last_tried = -1;
+ local->fd = fd_ref (fd);
- } else {
- call_child = afr_first_up_child (priv);
- if (call_child == -1) {
- op_errno = ENOTCONN;
- gf_log (this->name, GF_LOG_DEBUG,
- "no child is up");
- goto out;
- }
-
- local->cont.readv.last_tried = call_child;
- }
-
- local->fd = fd_ref (fd);
-
- local->cont.readv.ino = fd->inode->ino;
- local->cont.readv.size = size;
- local->cont.readv.offset = offset;
-
- STACK_WIND_COOKIE (frame, afr_readv_cbk,
- (void *) (long) call_child,
- children[call_child],
- children[call_child]->fops->readv,
- fd, size, offset);
-
- op_ret = 0;
+ local->cont.readv.size = size;
+ local->cont.readv.offset = offset;
+ local->cont.readv.flags = flags;
+
+ afr_open_fd_fix (fd, this);
+
+ STACK_WIND_COOKIE (frame, afr_readv_cbk,
+ (void *) (long) call_child,
+ children[call_child],
+ children[call_child]->fops->readv,
+ fd, size, offset, flags, xdata);
+
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (readv, frame, op_ret, op_errno, NULL, 0, NULL,
- NULL);
- }
- return 0;
+ if (ret < 0) {
+ AFR_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL,
+ NULL, NULL);
+ }
+ return 0;
}
/* }}} */
diff --git a/xlators/cluster/afr/src/afr-inode-read.h b/xlators/cluster/afr/src/afr-inode-read.h
index 2b22131db..e4091a793 100644
--- a/xlators/cluster/afr/src/afr-inode-read.h
+++ b/xlators/cluster/afr/src/afr-inode-read.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __INODE_READ_H__
@@ -22,26 +13,30 @@
int32_t
afr_access (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t mask);
+ loc_t *loc, int32_t mask, dict_t *xdata);
int32_t
afr_stat (call_frame_t *frame, xlator_t *this,
- loc_t *loc);
+ loc_t *loc, dict_t *xdata);
int32_t
afr_fstat (call_frame_t *frame, xlator_t *this,
- fd_t *fd);
+ fd_t *fd, dict_t *xdata);
int32_t
afr_readlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc, size_t size);
+ loc_t *loc, size_t size, dict_t *xdata);
int32_t
afr_readv (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset);
+ fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata);
int32_t
afr_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name);
+ loc_t *loc, const char *name, dict_t *xdata);
+
+int32_t
+afr_fgetxattr (call_frame_t *frame, xlator_t *this,
+ fd_t *fd, const char *name, dict_t *xdata);
#endif /* __INODE_READ_H__ */
diff --git a/xlators/cluster/afr/src/afr-inode-write.c b/xlators/cluster/afr/src/afr-inode-write.c
index 5a2c40cc4..c1ec69a55 100644
--- a/xlators/cluster/afr/src/afr-inode-write.c
+++ b/xlators/cluster/afr/src/afr-inode-write.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -46,52 +37,155 @@
#include "afr.h"
#include "afr-transaction.h"
+#include "afr-self-heal-common.h"
+
+void
+__inode_write_fop_cbk (call_frame_t *frame, int child_index, int read_child,
+ xlator_t *this, int32_t *op_ret, int32_t *op_errno,
+ struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+
+ local = frame->local;
+
+ if (afr_fop_failed (*op_ret, *op_errno)) {
+ local->child_errno[child_index] = *op_errno;
+
+ switch (local->op) {
+ case GF_FOP_TRUNCATE:
+ case GF_FOP_FTRUNCATE:
+ if (*op_errno != EFBIG)
+ afr_transaction_fop_failed (frame, this,
+ child_index);
+ break;
+ default:
+ afr_transaction_fop_failed (frame, this, child_index);
+ break;
+ }
+ local->op_errno = *op_errno;
+ goto out;
+ }
+
+ if ((local->success_count == 0) || (read_child == child_index)) {
+ local->op_ret = *op_ret;
+ if (prebuf)
+ local->cont.inode_wfop.prebuf = *prebuf;
+ if (postbuf)
+ local->cont.inode_wfop.postbuf = *postbuf;
+ }
+
+ local->success_count++;
+out:
+ return;
+}
/* {{{ writev */
-int
+void
+afr_writev_copy_outvars (call_frame_t *src_frame, call_frame_t *dst_frame)
+{
+ afr_local_t *src_local = NULL;
+ afr_local_t *dst_local = NULL;
+
+ src_local = src_frame->local;
+ dst_local = dst_frame->local;
+
+ dst_local->op_ret = src_local->op_ret;
+ dst_local->op_errno = src_local->op_errno;
+ dst_local->cont.inode_wfop.prebuf = src_local->cont.inode_wfop.prebuf;
+ dst_local->cont.inode_wfop.postbuf = src_local->cont.inode_wfop.postbuf;
+}
+
+void
afr_writev_unwind (call_frame_t *frame, xlator_t *this)
{
afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
+ local = frame->local;
+
+ AFR_STACK_UNWIND (writev, frame,
+ local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf,
+ NULL);
+}
+
+call_frame_t*
+afr_transaction_detach_fop_frame (call_frame_t *frame)
+{
+ afr_local_t * local = NULL;
+ call_frame_t *fop_frame = NULL;
local = frame->local;
LOCK (&frame->lock);
{
- if (local->transaction.main_frame)
- main_frame = local->transaction.main_frame;
+ fop_frame = local->transaction.main_frame;
local->transaction.main_frame = NULL;
}
UNLOCK (&frame->lock);
- if (main_frame) {
- local->cont.writev.prebuf.ia_ino = local->cont.writev.ino;
- local->cont.writev.postbuf.ia_ino = local->cont.writev.ino;
+ return fop_frame;
+}
- AFR_STACK_UNWIND (writev, main_frame,
- local->op_ret, local->op_errno,
- &local->cont.writev.prebuf,
- &local->cont.writev.postbuf);
+int
+afr_transaction_writev_unwind (call_frame_t *frame, xlator_t *this)
+{
+ call_frame_t *fop_frame = NULL;
+
+ fop_frame = afr_transaction_detach_fop_frame (frame);
+
+ if (fop_frame) {
+ afr_writev_copy_outvars (frame, fop_frame);
+ afr_writev_unwind (fop_frame, this);
}
return 0;
}
+static void
+afr_writev_handle_short_writes (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+ /*
+ * We already have the best case result of the writev calls staged
+ * as the return value. Any writev that returns some value less
+ * than the best case is now out of sync, so mark the fop as
+ * failed. Note that fops that have returned with errors have
+ * already been marked as failed.
+ */
+ for (i = 0; i < priv->child_count; i++) {
+ if ((!local->replies[i].valid) ||
+ (local->replies[i].op_ret == -1))
+ continue;
+
+ if (local->replies[i].op_ret < local->op_ret)
+ afr_transaction_fop_failed(frame, this, i);
+ }
+}
int
afr_writev_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
+ struct iatt *postbuf, dict_t *xdata)
{
afr_local_t * local = NULL;
-
+ afr_private_t *priv = NULL;
+ call_frame_t *fop_frame = NULL;
int child_index = (long) cookie;
int call_count = -1;
int read_child = 0;
+ int ret = 0;
+ uint32_t open_fd_count = 0;
+ uint32_t write_is_append = 0;
local = frame->local;
+ priv = this->private;
- read_child = afr_read_child (this, local->fd->inode);
+ read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
LOCK (&frame->lock);
{
@@ -99,50 +193,100 @@ afr_writev_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->read_child_returned = _gf_true;
}
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- local->cont.writev.prebuf = *prebuf;
- local->cont.writev.postbuf = *postbuf;
- }
-
- if (child_index == read_child) {
- local->cont.writev.prebuf = *prebuf;
- local->cont.writev.postbuf = *postbuf;
+ __inode_write_fop_cbk (frame, child_index, read_child, this,
+ &op_ret, &op_errno, prebuf, postbuf,
+ xdata);
+
+ local->replies[child_index].valid = 1;
+ local->replies[child_index].op_ret = op_ret;
+ local->replies[child_index].op_errno = op_errno;
+
+
+ /* stage the best case return value for unwind */
+ if ((local->success_count == 0) || (op_ret > local->op_ret)) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ }
+
+ if (op_ret != -1) {
+ if (xdata) {
+ ret = dict_get_uint32 (xdata,
+ GLUSTERFS_OPEN_FD_COUNT,
+ &open_fd_count);
+ if ((ret == 0) &&
+ (open_fd_count > local->open_fd_count)) {
+ local->open_fd_count = open_fd_count;
+ local->update_open_fd_count = _gf_true;
+ }
+
+ write_is_append = 0;
+ ret = dict_get_uint32 (xdata,
+ GLUSTERFS_WRITE_IS_APPEND,
+ &write_is_append);
+ if (ret || !write_is_append)
+ local->append_write = _gf_false;
}
- }
- local->op_errno = op_errno;
+ }
}
UNLOCK (&frame->lock);
call_count = afr_frame_return (frame);
if (call_count == 0) {
- local->transaction.unwind (frame, this);
- local->transaction.resume (frame, this);
+ if (local->update_open_fd_count)
+ afr_handle_open_fd_count (frame, this);
+
+ if (!local->stable_write && !local->append_write)
+ /* An appended write removes the necessity to
+ fsync() the file. This is because self-heal
+ has the logic to check for larger file when
+ the xattrs are not reliably pointing at
+ a stale file.
+ */
+ afr_fd_report_unstable_write (this, local->fd);
+
+ afr_writev_handle_short_writes (frame, this);
+ if (afr_any_fops_failed (local, priv)) {
+ //Don't unwind until post-op is complete
+ local->transaction.resume (frame, this);
+ } else {
+ /*
+ * Generally inode-write fops do transaction.unwind then
+ * transaction.resume, but writev needs to make sure that
+ * delayed post-op frame is placed in fdctx before unwind
+ * happens. This prevents the race of flush doing the
+ * changelog wakeup first in fuse thread and then this
+ * writev placing its delayed post-op frame in fdctx.
+ * This helps flush make sure all the delayed post-ops are
+ * completed.
+ */
+
+ fop_frame = afr_transaction_detach_fop_frame (frame);
+ afr_writev_copy_outvars (frame, fop_frame);
+ local->transaction.resume (frame, this);
+ afr_writev_unwind (fop_frame, this);
+ }
}
return 0;
}
-
int
afr_writev_wind (call_frame_t *frame, xlator_t *this)
{
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
-
int i = 0;
int call_count = -1;
+ dict_t *xdata = NULL;
+ GF_UNUSED int ret = 0;
local = frame->local;
priv = this->private;
- call_count = afr_up_children_count (priv->child_count, local->child_up);
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
if (call_count == 0) {
local->transaction.resume (frame, this);
@@ -150,9 +294,31 @@ afr_writev_wind (call_frame_t *frame, xlator_t *this)
}
local->call_count = call_count;
+ local->replies = GF_CALLOC(priv->child_count, sizeof(*local->replies),
+ gf_afr_mt_reply_t);
+ if (!local->replies) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ local->transaction.unwind(frame, this);
+ local->transaction.resume(frame, this);
+ return 0;
+ }
+
+ xdata = dict_new ();
+ if (xdata) {
+ ret = dict_set_uint32 (xdata, GLUSTERFS_OPEN_FD_COUNT,
+ sizeof (uint32_t));
+ ret = dict_set_uint32 (xdata, GLUSTERFS_WRITE_IS_APPEND,
+ 0);
+ /* Set append_write to be true speculatively. If on any
+ server it turns not be true, we unset it in the
+ callback.
+ */
+ local->append_write = _gf_true;
+ }
for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
+ if (local->transaction.pre_op[i]) {
STACK_WIND_COOKIE (frame, afr_writev_wind_cbk,
(void *) (long) i,
priv->children[i],
@@ -161,13 +327,18 @@ afr_writev_wind (call_frame_t *frame, xlator_t *this)
local->cont.writev.vector,
local->cont.writev.count,
local->cont.writev.offset,
- local->cont.writev.iobref);
+ local->cont.writev.flags,
+ local->cont.writev.iobref,
+ xdata);
if (!--call_count)
break;
}
}
+ if (xdata)
+ dict_unref (xdata);
+
return 0;
}
@@ -193,24 +364,21 @@ afr_writev_done (call_frame_t *frame, xlator_t *this)
int
afr_do_writev (call_frame_t *frame, xlator_t *this)
{
- call_frame_t * transaction_frame = NULL;
- afr_local_t * local = NULL;
-
- int op_ret = -1;
- int op_errno = 0;
+ call_frame_t *transaction_frame = NULL;
+ afr_local_t *local = NULL;
+ int op_ret = -1;
+ int op_errno = 0;
local = frame->local;
transaction_frame = copy_frame (frame);
if (!transaction_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
op_errno = ENOMEM;
goto out;
}
transaction_frame->local = local;
- frame->local = NULL;
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local->op = GF_FOP_WRITE;
@@ -218,10 +386,17 @@ afr_do_writev (call_frame_t *frame, xlator_t *this)
local->transaction.fop = afr_writev_wind;
local->transaction.done = afr_writev_done;
- local->transaction.unwind = afr_writev_unwind;
+ local->transaction.unwind = afr_transaction_writev_unwind;
local->transaction.main_frame = frame;
if (local->fd->flags & O_APPEND) {
+ /*
+ * Backend vfs ignores the 'offset' for append mode fd so
+ * locking just the region provided for the writev does not
+ * give consistency gurantee. The actual write may happen at a
+ * completely different range than the one provided by the
+ * offset, len in the fop. So lock the entire file.
+ */
local->transaction.start = 0;
local->transaction.len = 0;
} else {
@@ -230,79 +405,179 @@ afr_do_writev (call_frame_t *frame, xlator_t *this)
local->cont.writev.count);
}
- afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ op_ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto out;
+ }
op_ret = 0;
out:
- if (op_ret == -1) {
+ if (op_ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (writev, frame, op_ret, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (writev, frame, op_ret, op_errno, NULL, NULL, NULL);
}
return 0;
}
+static void
+afr_trigger_open_fd_self_heal (fd_t *fd, xlator_t *this)
+{
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ char *reason = NULL;
+ int32_t op_errno = 0;
+ int ret = 0;
+
+ if (!fd || !fd->inode || uuid_is_null (fd->inode->gfid)) {
+ gf_log_callingfn (this->name, GF_LOG_ERROR, "Invalid args: "
+ "fd: %p, inode: %p", fd,
+ fd ? fd->inode : NULL);
+ goto out;
+ }
+
+ frame = create_frame (this, this->ctx->pool);
+ if (!frame)
+ goto out;
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
+ ret = afr_local_init (local, this->private, &op_errno);
+ if (ret < 0)
+ goto out;
+
+ local->loc.inode = inode_ref (fd->inode);
+ ret = loc_path (&local->loc, NULL);
+ if (ret < 0)
+ goto out;
+
+ sh = &local->self_heal;
+ sh->do_metadata_self_heal = _gf_true;
+ if (fd->inode->ia_type == IA_IFREG)
+ sh->do_data_self_heal = _gf_true;
+ else if (fd->inode->ia_type == IA_IFDIR)
+ sh->do_entry_self_heal = _gf_true;
+
+ reason = "subvolume came online";
+ afr_launch_self_heal (frame, this, fd->inode, _gf_true,
+ fd->inode->ia_type, reason, NULL, NULL);
+ return;
+out:
+ AFR_STACK_DESTROY (frame);
+}
+
+void
+afr_open_fd_fix (fd_t *fd, xlator_t *this)
+{
+ int ret = 0;
+ int i = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ gf_boolean_t need_self_heal = _gf_false;
+ int *need_open = NULL;
+ size_t need_open_count = 0;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+
+ if (!afr_is_fd_fixable (fd))
+ goto out;
+
+ fd_ctx = afr_fd_ctx_get (fd, this);
+ if (!fd_ctx)
+ goto out;
+
+ LOCK (&fd->lock);
+ {
+ if (fd_ctx->up_count < priv->up_count) {
+ need_self_heal = _gf_true;
+ fd_ctx->up_count = priv->up_count;
+ fd_ctx->down_count = priv->down_count;
+ }
+
+ need_open = alloca (priv->child_count * sizeof (*need_open));
+ for (i = 0; i < priv->child_count; i++) {
+ need_open[i] = 0;
+ if (fd_ctx->opened_on[i] != AFR_FD_NOT_OPENED)
+ continue;
+
+ if (!priv->child_up[i])
+ continue;
+
+ fd_ctx->opened_on[i] = AFR_FD_OPENING;
+
+ need_open[i] = 1;
+ need_open_count++;
+ }
+ }
+ UNLOCK (&fd->lock);
+ if (ret)
+ goto out;
+
+ if (need_self_heal)
+ afr_trigger_open_fd_self_heal (fd, this);
+
+ if (!need_open_count)
+ goto out;
+
+ afr_fix_open (this, fd, need_open_count, need_open);
+out:
+ return;
+}
int
afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
struct iovec *vector, int32_t count, off_t offset,
- struct iobref *iobref)
+ uint32_t flags, struct iobref *iobref, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
-
int ret = -1;
-
- int op_ret = -1;
int op_errno = 0;
- uint64_t ctx;
- afr_fd_ctx_t *fd_ctx = NULL;
-
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
VALIDATE_OR_GOTO (this->private, out);
priv = this->private;
- ALLOC_OR_GOTO (local, afr_local_t, out);
-
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
+ if (afr_is_split_brain (this, fd->inode)) {
+ op_errno = EIO;
goto out;
}
- frame->local = local;
+ QUORUM_CHECK(writev,out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
+
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
local->cont.writev.vector = iov_dup (vector, count);
local->cont.writev.count = count;
local->cont.writev.offset = offset;
- local->cont.writev.ino = fd->inode->ino;
+ local->cont.writev.flags = flags;
local->cont.writev.iobref = iobref_ref (iobref);
local->fd = fd_ref (fd);
- ret = fd_ctx_get (fd, this, &ctx);
- if (ret < 0) {
- goto out;
- }
+ /* detect here, but set it in writev_wind_cbk *after* the unstable
+ write is performed
+ */
+ local->stable_write = !!((fd->flags|flags)&(O_SYNC|O_DSYNC));
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+ afr_open_fd_fix (fd, this);
- if (fd_ctx->up_count < priv->up_count) {
- local->openfd_flush_cbk = afr_do_writev;
- afr_openfd_flush (frame, this, fd);
- } else {
- afr_do_writev (frame, this);
- }
+ afr_do_writev (frame, this);
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (writev, frame, op_ret, op_errno, NULL, NULL);
- }
+ if (ret < 0)
+ AFR_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
return 0;
}
@@ -329,13 +604,11 @@ afr_truncate_unwind (call_frame_t *frame, xlator_t *this)
UNLOCK (&frame->lock);
if (main_frame) {
- local->cont.truncate.prebuf.ia_ino = local->cont.truncate.ino;
- local->cont.truncate.postbuf.ia_ino = local->cont.truncate.ino;
-
AFR_STACK_UNWIND (truncate, main_frame, local->op_ret,
local->op_errno,
- &local->cont.truncate.prebuf,
- &local->cont.truncate.postbuf);
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf,
+ NULL);
}
return 0;
@@ -345,20 +618,16 @@ afr_truncate_unwind (call_frame_t *frame, xlator_t *this)
int
afr_truncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
+ struct iatt *postbuf, dict_t *xdata)
{
afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
-
int child_index = (long) cookie;
int read_child = 0;
int call_count = -1;
- int need_unwind = 0;
local = frame->local;
- priv = this->private;
- read_child = afr_read_child (this, local->loc.inode);
+ read_child = afr_inode_get_read_ctx (this, local->loc.inode, NULL);
LOCK (&frame->lock);
{
@@ -366,38 +635,22 @@ afr_truncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->read_child_returned = _gf_true;
}
- if (afr_fop_failed (op_ret, op_errno) && op_errno != EFBIG)
- afr_transaction_fop_failed (frame, this, child_index);
-
if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- local->cont.truncate.prebuf = *prebuf;
- local->cont.truncate.postbuf = *postbuf;
- }
-
- if (child_index == read_child) {
- local->cont.truncate.prebuf = *prebuf;
- local->cont.truncate.postbuf = *postbuf;
- }
-
- local->success_count++;
-
- if ((local->success_count >= priv->wait_count)
- && local->read_child_returned) {
- need_unwind = 1;
- }
+ if (prebuf->ia_size != postbuf->ia_size)
+ local->stable_write = _gf_false;
}
- local->op_errno = op_errno;
+ __inode_write_fop_cbk (frame, child_index, read_child, this,
+ &op_ret, &op_errno, prebuf, postbuf,
+ xdata);
}
UNLOCK (&frame->lock);
- if (need_unwind)
- local->transaction.unwind (frame, this);
-
call_count = afr_frame_return (frame);
if (call_count == 0) {
+ if (local->stable_write && afr_txn_nothing_failed (frame, this))
+ local->transaction.unwind (frame, this);
+
local->transaction.resume (frame, this);
}
@@ -410,14 +663,14 @@ afr_truncate_wind (call_frame_t *frame, xlator_t *this)
{
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
-
int call_count = -1;
int i = 0;
local = frame->local;
priv = this->private;
- call_count = afr_up_children_count (priv->child_count, local->child_up);
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
if (call_count == 0) {
local->transaction.resume (frame, this);
@@ -425,15 +678,17 @@ afr_truncate_wind (call_frame_t *frame, xlator_t *this)
}
local->call_count = call_count;
+ local->stable_write = _gf_true;
for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
+ if (local->transaction.pre_op[i]) {
STACK_WIND_COOKIE (frame, afr_truncate_wind_cbk,
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->truncate,
&local->loc,
- local->cont.truncate.offset);
+ local->cont.truncate.offset,
+ NULL);
if (!--call_count)
break;
@@ -461,15 +716,12 @@ afr_truncate_done (call_frame_t *frame, xlator_t *this)
int
afr_truncate (call_frame_t *frame, xlator_t *this,
- loc_t *loc, off_t offset)
+ loc_t *loc, off_t offset, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
call_frame_t *transaction_frame = NULL;
-
int ret = -1;
-
- int op_ret = -1;
int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
@@ -478,27 +730,22 @@ afr_truncate (call_frame_t *frame, xlator_t *this,
priv = this->private;
+ QUORUM_CHECK(truncate,out);
+
transaction_frame = copy_frame (frame);
if (!transaction_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
+ op_errno = ENOMEM;
goto out;
}
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+ local = transaction_frame->local;
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
goto out;
- }
-
- transaction_frame->local = local;
-
- local->op_ret = -1;
local->cont.truncate.offset = offset;
- local->cont.truncate.ino = loc->inode->ino;
local->transaction.fop = afr_truncate_wind;
local->transaction.done = afr_truncate_done;
@@ -507,17 +754,21 @@ afr_truncate (call_frame_t *frame, xlator_t *this,
loc_copy (&local->loc, loc);
local->transaction.main_frame = frame;
- local->transaction.start = 0;
- local->transaction.len = offset;
+ local->transaction.start = offset;
+ local->transaction.len = 0;
- afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
+ if (ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (truncate, frame, op_ret, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
}
return 0;
@@ -546,13 +797,11 @@ afr_ftruncate_unwind (call_frame_t *frame, xlator_t *this)
UNLOCK (&frame->lock);
if (main_frame) {
- local->cont.ftruncate.prebuf.ia_ino = local->cont.ftruncate.ino;
- local->cont.ftruncate.postbuf.ia_ino = local->cont.ftruncate.ino;
-
AFR_STACK_UNWIND (ftruncate, main_frame, local->op_ret,
local->op_errno,
- &local->cont.ftruncate.prebuf,
- &local->cont.ftruncate.postbuf);
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf,
+ NULL);
}
return 0;
}
@@ -561,20 +810,16 @@ afr_ftruncate_unwind (call_frame_t *frame, xlator_t *this)
int
afr_ftruncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
+ struct iatt *postbuf, dict_t *xdata)
{
afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
-
int child_index = (long) cookie;
int call_count = -1;
- int need_unwind = 0;
int read_child = 0;
local = frame->local;
- priv = this->private;
- read_child = afr_read_child (this, local->fd->inode);
+ read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
LOCK (&frame->lock);
{
@@ -582,38 +827,22 @@ afr_ftruncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->read_child_returned = _gf_true;
}
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- local->cont.ftruncate.prebuf = *prebuf;
- local->cont.ftruncate.postbuf = *postbuf;
- }
-
- if (child_index == read_child) {
- local->cont.ftruncate.prebuf = *prebuf;
- local->cont.ftruncate.postbuf = *postbuf;
- }
-
- local->success_count++;
-
- if ((local->success_count >= priv->wait_count)
- && local->read_child_returned) {
- need_unwind = 1;
- }
+ if (prebuf->ia_size != postbuf->ia_size)
+ local->stable_write = _gf_false;
}
- local->op_errno = op_errno;
+ __inode_write_fop_cbk (frame, child_index, read_child, this,
+ &op_ret, &op_errno, prebuf, postbuf,
+ xdata);
}
UNLOCK (&frame->lock);
- if (need_unwind)
- local->transaction.unwind (frame, this);
-
call_count = afr_frame_return (frame);
if (call_count == 0) {
+ if (local->stable_write && afr_txn_nothing_failed (frame, this))
+ local->transaction.unwind (frame, this);
+
local->transaction.resume (frame, this);
}
@@ -626,14 +855,14 @@ afr_ftruncate_wind (call_frame_t *frame, xlator_t *this)
{
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
-
int call_count = -1;
int i = 0;
local = frame->local;
priv = this->private;
- call_count = afr_up_children_count (priv->child_count, local->child_up);
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
if (call_count == 0) {
local->transaction.resume (frame, this);
@@ -641,14 +870,17 @@ afr_ftruncate_wind (call_frame_t *frame, xlator_t *this)
}
local->call_count = call_count;
+ local->stable_write = _gf_true;
for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
+ if (local->transaction.pre_op[i]) {
STACK_WIND_COOKIE (frame, afr_ftruncate_wind_cbk,
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->ftruncate,
- local->fd, local->cont.ftruncate.offset);
+ local->fd,
+ local->cont.ftruncate.offset,
+ NULL);
if (!--call_count)
break;
@@ -679,7 +911,6 @@ afr_do_ftruncate (call_frame_t *frame, xlator_t *this)
{
call_frame_t * transaction_frame = NULL;
afr_local_t * local = NULL;
-
int op_ret = -1;
int op_errno = 0;
@@ -687,8 +918,6 @@ afr_do_ftruncate (call_frame_t *frame, xlator_t *this)
transaction_frame = copy_frame (frame);
if (!transaction_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
goto out;
}
@@ -703,17 +932,22 @@ afr_do_ftruncate (call_frame_t *frame, xlator_t *this)
local->transaction.main_frame = frame;
- local->transaction.start = 0;
- local->transaction.len = local->cont.ftruncate.offset;
+ local->transaction.start = local->cont.ftruncate.offset;
+ local->transaction.len = 0;
- afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ op_ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto out;
+ }
op_ret = 0;
out:
- if (op_ret == -1) {
+ if (op_ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (ftruncate, frame, op_ret, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (ftruncate, frame, op_ret, op_errno, NULL,
+ NULL, NULL);
}
return 0;
@@ -722,61 +956,47 @@ out:
int
afr_ftruncate (call_frame_t *frame, xlator_t *this,
- fd_t *fd, off_t offset)
+ fd_t *fd, off_t offset, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
call_frame_t *transaction_frame = NULL;
-
int ret = -1;
-
- int op_ret = -1;
int op_errno = 0;
- uint64_t ctx;
- afr_fd_ctx_t *fd_ctx = NULL;
-
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
VALIDATE_OR_GOTO (this->private, out);
priv = this->private;
- ALLOC_OR_GOTO (local, afr_local_t, out);
-
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
+ if (afr_is_split_brain (this, fd->inode)) {
+ op_errno = EIO;
goto out;
}
+ QUORUM_CHECK(ftruncate,out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- frame->local = local;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
local->cont.ftruncate.offset = offset;
- local->cont.ftruncate.ino = fd->inode->ino;
local->fd = fd_ref (fd);
- ret = fd_ctx_get (fd, this, &ctx);
- if (ret < 0) {
- goto out;
- }
+ afr_open_fd_fix (fd, this);
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+ afr_do_ftruncate (frame, this);
- if (fd_ctx->up_count < priv->up_count) {
- local->openfd_flush_cbk = afr_do_ftruncate;
- afr_openfd_flush (frame, this, fd);
- } else {
- afr_do_ftruncate (frame, this);
- }
-
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
+ if (ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (ftruncate, frame, op_ret, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
}
return 0;
@@ -803,13 +1023,11 @@ afr_setattr_unwind (call_frame_t *frame, xlator_t *this)
UNLOCK (&frame->lock);
if (main_frame) {
- local->cont.setattr.preop_buf.ia_ino = local->cont.setattr.ino;
- local->cont.setattr.postop_buf.ia_ino = local->cont.setattr.ino;
-
AFR_STACK_UNWIND (setattr, main_frame, local->op_ret,
local->op_errno,
- &local->cont.setattr.preop_buf,
- &local->cont.setattr.postop_buf);
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf,
+ NULL);
}
return 0;
@@ -819,11 +1037,10 @@ afr_setattr_unwind (call_frame_t *frame, xlator_t *this)
int
afr_setattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop)
+ struct iatt *preop, struct iatt *postop, dict_t *xdata)
{
afr_local_t * local = NULL;
afr_private_t * priv = NULL;
-
int child_index = (long) cookie;
int read_child = 0;
int call_count = -1;
@@ -832,7 +1049,7 @@ afr_setattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local = frame->local;
priv = this->private;
- read_child = afr_read_child (this, local->loc.inode);
+ read_child = afr_inode_get_read_ctx (this, local->loc.inode, NULL);
LOCK (&frame->lock);
{
@@ -840,29 +1057,14 @@ afr_setattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->read_child_returned = _gf_true;
}
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- local->cont.setattr.preop_buf = *preop;
- local->cont.setattr.postop_buf = *postop;
- }
-
- if (child_index == read_child) {
- local->cont.setattr.preop_buf = *preop;
- local->cont.setattr.postop_buf = *postop;
- }
-
- local->success_count++;
+ __inode_write_fop_cbk (frame, child_index, read_child, this,
+ &op_ret, &op_errno, preop, postop,
+ xdata);
- if ((local->success_count >= priv->wait_count)
- && local->read_child_returned) {
- need_unwind = 1;
- }
+ if ((local->success_count >= priv->wait_count)
+ && local->read_child_returned) {
+ need_unwind = 1;
}
- local->op_errno = op_errno;
}
UNLOCK (&frame->lock);
@@ -884,14 +1086,14 @@ afr_setattr_wind (call_frame_t *frame, xlator_t *this)
{
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
-
int call_count = -1;
int i = 0;
local = frame->local;
priv = this->private;
- call_count = afr_up_children_count (priv->child_count, local->child_up);
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
if (call_count == 0) {
local->transaction.resume (frame, this);
@@ -901,14 +1103,15 @@ afr_setattr_wind (call_frame_t *frame, xlator_t *this)
local->call_count = call_count;
for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
+ if (local->transaction.pre_op[i]) {
STACK_WIND_COOKIE (frame, afr_setattr_wind_cbk,
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->setattr,
&local->loc,
&local->cont.setattr.in_buf,
- local->cont.setattr.valid);
+ local->cont.setattr.valid,
+ NULL);
if (!--call_count)
break;
@@ -936,15 +1139,12 @@ afr_setattr_done (call_frame_t *frame, xlator_t *this)
int
afr_setattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, struct iatt *buf, int32_t valid)
+ loc_t *loc, struct iatt *buf, int32_t valid, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
call_frame_t *transaction_frame = NULL;
-
int ret = -1;
-
- int op_ret = -1;
int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
@@ -953,26 +1153,20 @@ afr_setattr (call_frame_t *frame, xlator_t *this,
priv = this->private;
+ QUORUM_CHECK(setattr,out);
+
transaction_frame = copy_frame (frame);
if (!transaction_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
+ op_errno = ENOMEM;
goto out;
}
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+ local = transaction_frame->local;
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
goto out;
- }
-
- transaction_frame->local = local;
-
- local->op_ret = -1;
-
- local->cont.setattr.ino = loc->inode->ino;
local->cont.setattr.in_buf = *buf;
local->cont.setattr.valid = valid;
@@ -987,14 +1181,18 @@ afr_setattr (call_frame_t *frame, xlator_t *this,
local->transaction.start = LLONG_MAX - 1;
local->transaction.len = 0;
- afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
+ if (ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (setattr, frame, op_ret, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL, NULL);
}
return 0;
@@ -1019,15 +1217,11 @@ afr_fsetattr_unwind (call_frame_t *frame, xlator_t *this)
UNLOCK (&frame->lock);
if (main_frame) {
- local->cont.fsetattr.preop_buf.ia_ino =
- local->cont.fsetattr.ino;
- local->cont.fsetattr.postop_buf.ia_ino =
- local->cont.fsetattr.ino;
-
AFR_STACK_UNWIND (fsetattr, main_frame, local->op_ret,
local->op_errno,
- &local->cont.fsetattr.preop_buf,
- &local->cont.fsetattr.postop_buf);
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf,
+ NULL);
}
return 0;
@@ -1036,12 +1230,11 @@ afr_fsetattr_unwind (call_frame_t *frame, xlator_t *this)
int
afr_fsetattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop)
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *preop, struct iatt *postop, dict_t *xdata)
{
afr_local_t * local = NULL;
afr_private_t * priv = NULL;
-
int child_index = (long) cookie;
int read_child = 0;
int call_count = -1;
@@ -1050,7 +1243,7 @@ afr_fsetattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local = frame->local;
priv = this->private;
- read_child = afr_read_child (this, local->fd->inode);
+ read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
LOCK (&frame->lock);
{
@@ -1058,29 +1251,14 @@ afr_fsetattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->read_child_returned = _gf_true;
}
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- local->cont.fsetattr.preop_buf = *preop;
- local->cont.fsetattr.postop_buf = *postop;
- }
-
- if (child_index == read_child) {
- local->cont.fsetattr.preop_buf = *preop;
- local->cont.fsetattr.postop_buf = *postop;
- }
+ __inode_write_fop_cbk (frame, child_index, read_child, this,
+ &op_ret, &op_errno, preop, postop,
+ xdata);
- local->success_count++;
-
- if ((local->success_count >= priv->wait_count)
- && local->read_child_returned) {
- need_unwind = 1;
- }
+ if ((local->success_count >= priv->wait_count)
+ && local->read_child_returned) {
+ need_unwind = 1;
}
- local->op_errno = op_errno;
}
UNLOCK (&frame->lock);
@@ -1102,14 +1280,14 @@ afr_fsetattr_wind (call_frame_t *frame, xlator_t *this)
{
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
-
int call_count = -1;
int i = 0;
local = frame->local;
priv = this->private;
- call_count = afr_up_children_count (priv->child_count, local->child_up);
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
if (call_count == 0) {
local->transaction.resume (frame, this);
@@ -1119,14 +1297,15 @@ afr_fsetattr_wind (call_frame_t *frame, xlator_t *this)
local->call_count = call_count;
for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
+ if (local->transaction.pre_op[i]) {
STACK_WIND_COOKIE (frame, afr_fsetattr_wind_cbk,
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->fsetattr,
local->fd,
&local->cont.fsetattr.in_buf,
- local->cont.fsetattr.valid);
+ local->cont.fsetattr.valid,
+ NULL);
if (!--call_count)
break;
@@ -1151,18 +1330,14 @@ afr_fsetattr_done (call_frame_t *frame, xlator_t *this)
return 0;
}
-
int
afr_fsetattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, struct iatt *buf, int32_t valid)
+ fd_t *fd, struct iatt *buf, int32_t valid, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
call_frame_t *transaction_frame = NULL;
-
int ret = -1;
-
- int op_ret = -1;
int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
@@ -1171,26 +1346,25 @@ afr_fsetattr (call_frame_t *frame, xlator_t *this,
priv = this->private;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
+ if (afr_is_split_brain (this, fd->inode)) {
+ op_errno = EIO;
goto out;
}
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ QUORUM_CHECK(fsetattr,out);
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ op_errno = ENOMEM;
goto out;
}
- transaction_frame->local = local;
-
- local->op_ret = -1;
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+ local = transaction_frame->local;
- local->cont.fsetattr.ino = fd->inode->ino;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
local->cont.fsetattr.in_buf = *buf;
local->cont.fsetattr.valid = valid;
@@ -1201,18 +1375,24 @@ afr_fsetattr (call_frame_t *frame, xlator_t *this,
local->fd = fd_ref (fd);
+ afr_open_fd_fix (fd, this);
+
local->transaction.main_frame = frame;
local->transaction.start = LLONG_MAX - 1;
local->transaction.len = 0;
- afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
+ if (ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (fsetattr, frame, op_ret, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
}
return 0;
@@ -1240,7 +1420,8 @@ afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
if (main_frame) {
AFR_STACK_UNWIND (setxattr, main_frame,
- local->op_ret, local->op_errno)
+ local->op_ret, local->op_errno,
+ NULL);
}
return 0;
}
@@ -1248,31 +1429,25 @@ afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
int
afr_setxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
-
- int call_count = -1;
- int need_unwind = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int need_unwind = 0;
+ int child_index = (long) cookie;
local = frame->local;
priv = this->private;
LOCK (&frame->lock);
{
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- }
- local->success_count++;
-
- if (local->success_count == priv->child_count) {
- need_unwind = 1;
- }
+ __inode_write_fop_cbk (frame, child_index, -1, this,
+ &op_ret, &op_errno, NULL, NULL,
+ xdata);
+ if (local->success_count == priv->child_count) {
+ need_unwind = 1;
}
-
- local->op_errno = op_errno;
}
UNLOCK (&frame->lock);
@@ -1292,16 +1467,16 @@ afr_setxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int
afr_setxattr_wind (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- int call_count = -1;
- int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
local = frame->local;
priv = this->private;
- call_count = afr_up_children_count (priv->child_count, local->child_up);
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
if (call_count == 0) {
local->transaction.resume (frame, this);
@@ -1311,14 +1486,15 @@ afr_setxattr_wind (call_frame_t *frame, xlator_t *this)
local->call_count = call_count;
for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
+ if (local->transaction.pre_op[i]) {
STACK_WIND_COOKIE (frame, afr_setxattr_wind_cbk,
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->setxattr,
&local->loc,
local->cont.setxattr.dict,
- local->cont.setxattr.flags);
+ local->cont.setxattr.flags,
+ NULL);
if (!--call_count)
break;
@@ -1332,7 +1508,7 @@ afr_setxattr_wind (call_frame_t *frame, xlator_t *this)
int
afr_setxattr_done (call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = frame->local;
+ afr_local_t *local = frame->local;
local->transaction.unwind (frame, this);
@@ -1343,41 +1519,40 @@ afr_setxattr_done (call_frame_t *frame, xlator_t *this)
int
afr_setxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *dict, int32_t flags)
+ loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = EINVAL;
- int ret = -1;
+ VALIDATE_OR_GOTO (this, out);
- int op_ret = -1;
- int op_errno = 0;
+ GF_IF_INTERNAL_XATTR_GOTO ("trusted.afr.*", dict,
+ op_errno, out);
+
+ GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.afr.*", dict,
+ op_errno, out);
VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
VALIDATE_OR_GOTO (this->private, out);
priv = this->private;
- ALLOC_OR_GOTO (local, afr_local_t, out);
-
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
+ QUORUM_CHECK(setxattr,out);
transaction_frame = copy_frame (frame);
if (!transaction_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
+ op_errno = ENOMEM;
goto out;
}
- transaction_frame->local = local;
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+ local = transaction_frame->local;
- local->op_ret = -1;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
local->cont.setxattr.dict = dict_ref (dict);
local->cont.setxattr.flags = flags;
@@ -1392,14 +1567,211 @@ afr_setxattr (call_frame_t *frame, xlator_t *this,
local->transaction.start = LLONG_MAX - 1;
local->transaction.len = 0;
- afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- op_ret = 0;
+ ret = 0;
+out:
+ if (ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
+ }
+
+ return 0;
+}
+
+/* {{{ fsetxattr */
+
+
+int
+afr_fsetxattr_unwind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
+
+ local = frame->local;
+
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame)
+ main_frame = local->transaction.main_frame;
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK (&frame->lock);
+
+ if (main_frame) {
+ AFR_STACK_UNWIND (fsetxattr, main_frame,
+ local->op_ret, local->op_errno,
+ NULL);
+ }
+ return 0;
+}
+
+
+int
+afr_fsetxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int need_unwind = 0;
+ int child_index = (long) cookie;
+
+ local = frame->local;
+ priv = this->private;
+
+ LOCK (&frame->lock);
+ {
+
+ __inode_write_fop_cbk (frame, child_index, -1, this,
+ &op_ret, &op_errno, NULL, NULL,
+ xdata);
+ if (local->success_count == priv->child_count) {
+ need_unwind = 1;
+ }
+ }
+ UNLOCK (&frame->lock);
+
+ if (need_unwind)
+ local->transaction.unwind (frame, this);
+
+ call_count = afr_frame_return (frame);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ }
+
+ return 0;
+}
+
+
+int
+afr_fsetxattr_wind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
+
+ local->call_count = call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_fsetxattr_wind_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->fsetxattr,
+ local->fd,
+ local->cont.fsetxattr.dict,
+ local->cont.fsetxattr.flags,
+ NULL);
+
+ if (!--call_count)
+ break;
+ }
+ }
+
+ return 0;
+}
+
+
+int
+afr_fsetxattr_done (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = frame->local;
+
+ local->transaction.unwind (frame, this);
+
+ AFR_STACK_DESTROY (frame);
+
+ return 0;
+}
+
+int
+afr_fsetxattr (call_frame_t *frame, xlator_t *this,
+ fd_t *fd, dict_t *dict, int32_t flags, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = EINVAL;
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
+
+ GF_IF_INTERNAL_XATTR_GOTO ("trusted.afr.*", dict,
+ op_errno, out);
+
+ GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.afr.*", dict,
+ op_errno, out);
+
+ priv = this->private;
+
+ if (afr_is_split_brain (this, fd->inode)) {
+ op_errno = EIO;
+ goto out;
+ }
+
+ QUORUM_CHECK(fsetxattr,out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (local, out);
+
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
+
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ goto out;
+ }
+
+ transaction_frame->local = local;
+
+ local->op_ret = -1;
+
+ local->cont.fsetxattr.dict = dict_ref (dict);
+ local->cont.fsetxattr.flags = flags;
+
+ local->transaction.fop = afr_fsetxattr_wind;
+ local->transaction.done = afr_fsetxattr_done;
+ local->transaction.unwind = afr_fsetxattr_unwind;
+
+ local->fd = fd_ref (fd);
+
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
+
+ ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ ret = 0;
out:
- if (op_ret == -1) {
+ if (ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (setxattr, frame, op_ret, op_errno);
+ AFR_STACK_UNWIND (fsetxattr, frame, -1, op_errno, NULL);
}
return 0;
@@ -1407,6 +1779,7 @@ out:
/* }}} */
+
/* {{{ removexattr */
@@ -1428,7 +1801,8 @@ afr_removexattr_unwind (call_frame_t *frame, xlator_t *this)
if (main_frame) {
AFR_STACK_UNWIND (removexattr, main_frame,
- local->op_ret, local->op_errno)
+ local->op_ret, local->op_errno,
+ NULL);
}
return 0;
}
@@ -1436,31 +1810,25 @@ afr_removexattr_unwind (call_frame_t *frame, xlator_t *this)
int
afr_removexattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
-
- int call_count = -1;
- int need_unwind = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int need_unwind = 0;
+ int child_index = (long) cookie;
local = frame->local;
priv = this->private;
LOCK (&frame->lock);
{
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- }
- local->success_count++;
-
- if (local->success_count == priv->wait_count) {
- need_unwind = 1;
- }
+ __inode_write_fop_cbk (frame, child_index, -1, this,
+ &op_ret, &op_errno, NULL, NULL,
+ xdata);
+ if (local->success_count == priv->wait_count) {
+ need_unwind = 1;
}
-
- local->op_errno = op_errno;
}
UNLOCK (&frame->lock);
@@ -1482,14 +1850,14 @@ afr_removexattr_wind (call_frame_t *frame, xlator_t *this)
{
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
-
int call_count = -1;
int i = 0;
local = frame->local;
priv = this->private;
- call_count = afr_up_children_count (priv->child_count, local->child_up);
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
if (call_count == 0) {
local->transaction.resume (frame, this);
@@ -1499,13 +1867,14 @@ afr_removexattr_wind (call_frame_t *frame, xlator_t *this)
local->call_count = call_count;
for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
+ if (local->transaction.pre_op[i]) {
STACK_WIND_COOKIE (frame, afr_removexattr_wind_cbk,
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->removexattr,
&local->loc,
- local->cont.removexattr.name);
+ local->cont.removexattr.name,
+ NULL);
if (!--call_count)
break;
@@ -1531,34 +1900,227 @@ afr_removexattr_done (call_frame_t *frame, xlator_t *this)
int
afr_removexattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name)
+ loc_t *loc, const char *name, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
+
+ VALIDATE_OR_GOTO (this, out);
+
+ GF_IF_NATIVE_XATTR_GOTO ("trusted.afr.*",
+ name, op_errno, out);
+
+ GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.afr.*",
+ name, op_errno, out);
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this->private, out);
+ VALIDATE_OR_GOTO (loc, out);
+
+ priv = this->private;
+
+ QUORUM_CHECK(removexattr,out);
+
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+ local = transaction_frame->local;
+
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
+
+ local->cont.removexattr.name = gf_strdup (name);
+
+ local->transaction.fop = afr_removexattr_wind;
+ local->transaction.done = afr_removexattr_done;
+ local->transaction.unwind = afr_removexattr_unwind;
+
+ loc_copy (&local->loc, loc);
+
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
+
+ ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ ret = 0;
+out:
+ if (ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
+ }
+
+ return 0;
+}
+
+/* ffremovexattr */
+int
+afr_fremovexattr_unwind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t * local = NULL;
+ call_frame_t *main_frame = NULL;
+
+ local = frame->local;
+
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame)
+ main_frame = local->transaction.main_frame;
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK (&frame->lock);
+
+ if (main_frame) {
+ AFR_STACK_UNWIND (fremovexattr, main_frame,
+ local->op_ret, local->op_errno,
+ NULL);
+ }
+ return 0;
+}
+
+
+int
+afr_fremovexattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+ afr_local_t * local = NULL;
+ afr_private_t * priv = NULL;
+ int call_count = -1;
+ int need_unwind = 0;
+ int child_index = (long) cookie;
+
+ local = frame->local;
+ priv = this->private;
+
+ LOCK (&frame->lock);
+ {
+ __inode_write_fop_cbk (frame, child_index, -1, this,
+ &op_ret, &op_errno, NULL, NULL,
+ xdata);
+
+ if (local->success_count == priv->wait_count) {
+ need_unwind = 1;
+ }
+ }
+ UNLOCK (&frame->lock);
+
+ if (need_unwind)
+ local->transaction.unwind (frame, this);
+
+ call_count = afr_frame_return (frame);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ }
+
+ return 0;
+}
+
+
+int32_t
+afr_fremovexattr_wind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
+
+ local->call_count = call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_fremovexattr_wind_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->fremovexattr,
+ local->fd,
+ local->cont.removexattr.name,
+ NULL);
+
+ if (!--call_count)
+ break;
+ }
+ }
+
+ return 0;
+}
+
+
+int
+afr_fremovexattr_done (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t * local = frame->local;
+
+ local->transaction.unwind (frame, this);
+
+ AFR_STACK_DESTROY (frame);
+
+ return 0;
+}
+
+
+int
+afr_fremovexattr (call_frame_t *frame, xlator_t *this,
+ fd_t *fd, const char *name, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
call_frame_t *transaction_frame = NULL;
-
int ret = -1;
-
int op_ret = -1;
int op_errno = 0;
- VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
+
+ GF_IF_NATIVE_XATTR_GOTO ("trusted.afr.*",
+ name, op_errno, out);
+
+ GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.afr.*",
+ name, op_errno, out);
+
+ VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this->private, out);
- VALIDATE_OR_GOTO (loc, out);
priv = this->private;
+ if (afr_is_split_brain (this, fd->inode)) {
+ op_errno = EIO;
+ goto out;
+ }
+
+ QUORUM_CHECK(fremovexattr, out);
transaction_frame = copy_frame (frame);
if (!transaction_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
goto out;
}
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (local, out);
- ret = AFR_LOCAL_INIT (local, priv);
+ ret = afr_local_init (local, priv, &op_errno);
if (ret < 0) {
op_errno = -ret;
goto out;
@@ -1570,25 +2132,730 @@ afr_removexattr (call_frame_t *frame, xlator_t *this,
local->cont.removexattr.name = gf_strdup (name);
- local->transaction.fop = afr_removexattr_wind;
- local->transaction.done = afr_removexattr_done;
- local->transaction.unwind = afr_removexattr_unwind;
+ local->transaction.fop = afr_fremovexattr_wind;
+ local->transaction.done = afr_fremovexattr_done;
+ local->transaction.unwind = afr_fremovexattr_unwind;
- loc_copy (&local->loc, loc);
+ local->fd = fd_ref (fd);
local->transaction.main_frame = frame;
local->transaction.start = LLONG_MAX - 1;
local->transaction.len = 0;
- afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ op_ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto out;
+ }
+
+ op_ret = 0;
+out:
+ if (op_ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, NULL);
+ }
+
+ return 0;
+}
+
+static int
+afr_fallocate_unwind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t * local = NULL;
+ call_frame_t *main_frame = NULL;
+
+ local = frame->local;
+
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame)
+ main_frame = local->transaction.main_frame;
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK (&frame->lock);
+
+ if (main_frame) {
+ AFR_STACK_UNWIND (fallocate, main_frame, local->op_ret,
+ local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf,
+ NULL);
+ }
+ return 0;
+}
+
+static int
+afr_fallocate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ afr_local_t * local = NULL;
+ afr_private_t * priv = NULL;
+ int child_index = (long) cookie;
+ int call_count = -1;
+ int need_unwind = 0;
+ int read_child = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
+
+ LOCK (&frame->lock);
+ {
+ if (child_index == read_child) {
+ local->read_child_returned = _gf_true;
+ }
+
+ __inode_write_fop_cbk (frame, child_index, read_child, this,
+ &op_ret, &op_errno, prebuf, postbuf,
+ xdata);
+
+ if ((local->success_count >= priv->wait_count)
+ && local->read_child_returned) {
+ need_unwind = 1;
+ }
+ }
+ UNLOCK (&frame->lock);
+
+ if (need_unwind)
+ local->transaction.unwind (frame, this);
+
+ call_count = afr_frame_return (frame);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ }
+
+ return 0;
+}
+
+static int
+afr_fallocate_wind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
+
+ local->call_count = call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_fallocate_wind_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->fallocate,
+ local->fd,
+ local->cont.fallocate.mode,
+ local->cont.fallocate.offset,
+ local->cont.fallocate.len,
+ NULL);
+
+ if (!--call_count)
+ break;
+ }
+ }
+
+ return 0;
+}
+
+static int
+afr_fallocate_done (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+
+ local = frame->local;
+
+ local->transaction.unwind (frame, this);
+
+ AFR_STACK_DESTROY (frame);
+
+ return 0;
+}
+
+static int
+afr_do_fallocate (call_frame_t *frame, xlator_t *this)
+{
+ call_frame_t * transaction_frame = NULL;
+ afr_local_t * local = NULL;
+ int op_ret = -1;
+ int op_errno = 0;
+
+ local = frame->local;
+
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ goto out;
+ }
+
+ transaction_frame->local = local;
+ frame->local = NULL;
+
+ local->op = GF_FOP_FALLOCATE;
+
+ local->transaction.fop = afr_fallocate_wind;
+ local->transaction.done = afr_fallocate_done;
+ local->transaction.unwind = afr_fallocate_unwind;
+
+ local->transaction.main_frame = frame;
+
+ local->transaction.start = local->cont.fallocate.offset;
+ local->transaction.len = 0;
+
+ /* fallocate can modify the file size */
+ op_ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto out;
+ }
+
+ op_ret = 0;
+out:
+ if (op_ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (fallocate, frame, op_ret, op_errno, NULL,
+ NULL, NULL);
+ }
+
+ return 0;
+}
+
+int
+afr_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+ off_t offset, size_t len, dict_t *xdata)
+{
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
+
+ priv = this->private;
+
+ if (afr_is_split_brain (this, fd->inode)) {
+ op_errno = EIO;
+ goto out;
+ }
+ QUORUM_CHECK(fallocate,out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
+
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
+
+ local->cont.fallocate.mode = mode;
+ local->cont.fallocate.offset = offset;
+ local->cont.fallocate.len = len;
+
+ local->fd = fd_ref (fd);
+
+ afr_open_fd_fix (fd, this);
+
+ afr_do_fallocate (frame, this);
+
+ ret = 0;
+out:
+ if (ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
+ }
+
+ return 0;
+}
+
+/* }}} */
+
+/* {{{ discard */
+
+static int
+afr_discard_unwind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t * local = NULL;
+ call_frame_t *main_frame = NULL;
+
+ local = frame->local;
+
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame)
+ main_frame = local->transaction.main_frame;
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK (&frame->lock);
+
+ if (main_frame) {
+ AFR_STACK_UNWIND (discard, main_frame, local->op_ret,
+ local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf,
+ NULL);
+ }
+ return 0;
+}
+
+static int
+afr_discard_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ afr_local_t * local = NULL;
+ afr_private_t * priv = NULL;
+ int child_index = (long) cookie;
+ int call_count = -1;
+ int need_unwind = 0;
+ int read_child = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
+
+ LOCK (&frame->lock);
+ {
+ if (child_index == read_child) {
+ local->read_child_returned = _gf_true;
+ }
+
+ __inode_write_fop_cbk (frame, child_index, read_child, this,
+ &op_ret, &op_errno, prebuf, postbuf,
+ xdata);
+
+ if ((local->success_count >= priv->wait_count)
+ && local->read_child_returned) {
+ need_unwind = 1;
+ }
+ }
+ UNLOCK (&frame->lock);
+
+ if (need_unwind)
+ local->transaction.unwind (frame, this);
+
+ call_count = afr_frame_return (frame);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ }
+
+ return 0;
+}
+
+static int
+afr_discard_wind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
+
+ local->call_count = call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_discard_wind_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->discard,
+ local->fd,
+ local->cont.discard.offset,
+ local->cont.discard.len,
+ NULL);
+
+ if (!--call_count)
+ break;
+ }
+ }
+
+ return 0;
+}
+
+static int
+afr_discard_done (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+
+ local = frame->local;
+
+ local->transaction.unwind (frame, this);
+
+ AFR_STACK_DESTROY (frame);
+
+ return 0;
+}
+
+static int
+afr_do_discard (call_frame_t *frame, xlator_t *this)
+{
+ call_frame_t * transaction_frame = NULL;
+ afr_local_t * local = NULL;
+ int op_ret = -1;
+ int op_errno = 0;
+
+ local = frame->local;
+
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ goto out;
+ }
+
+ transaction_frame->local = local;
+ frame->local = NULL;
+
+ local->op = GF_FOP_DISCARD;
+
+ local->transaction.fop = afr_discard_wind;
+ local->transaction.done = afr_discard_done;
+ local->transaction.unwind = afr_discard_unwind;
+
+ local->transaction.main_frame = frame;
+
+ local->transaction.start = local->cont.discard.offset;
+ local->transaction.len = 0;
+
+ op_ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto out;
+ }
op_ret = 0;
out:
- if (op_ret == -1) {
+ if (op_ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (removexattr, frame, op_ret, op_errno);
+ AFR_STACK_UNWIND (discard, frame, op_ret, op_errno, NULL,
+ NULL, NULL);
+ }
+
+ return 0;
+}
+
+int
+afr_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
+{
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
+
+ priv = this->private;
+
+ if (afr_is_split_brain (this, fd->inode)) {
+ op_errno = EIO;
+ goto out;
+ }
+ QUORUM_CHECK(discard, out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
+
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
+
+ local->cont.discard.offset = offset;
+ local->cont.discard.len = len;
+
+ local->fd = fd_ref (fd);
+
+ afr_open_fd_fix (fd, this);
+
+ afr_do_discard(frame, this);
+
+ ret = 0;
+out:
+ if (ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (discard, frame, -1, op_errno, NULL, NULL, NULL);
+ }
+
+ return 0;
+}
+
+
+/* {{{ zerofill */
+
+static int
+afr_zerofill_unwind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
+
+ local = frame->local;
+
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame) {
+ main_frame = local->transaction.main_frame;
+ }
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK (&frame->lock);
+
+ if (main_frame) {
+ AFR_STACK_UNWIND (zerofill, main_frame, local->op_ret,
+ local->op_errno,
+ &local->cont.zerofill.prebuf,
+ &local->cont.zerofill.postbuf,
+ NULL);
+ }
+ return 0;
+}
+
+static int
+afr_zerofill_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int child_index = (long) cookie;
+ int call_count = -1;
+ int need_unwind = 0;
+ int read_child = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
+
+ LOCK (&frame->lock);
+ {
+ if (child_index == read_child) {
+ local->read_child_returned = _gf_true;
+ }
+
+ if (afr_fop_failed (op_ret, op_errno)) {
+ afr_transaction_fop_failed (frame, this, child_index);
+ }
+
+ if (op_ret != -1) {
+ if (local->success_count == 0) {
+ local->op_ret = op_ret;
+ local->cont.zerofill.prebuf = *prebuf;
+ local->cont.zerofill.postbuf = *postbuf;
+ }
+
+ if (child_index == read_child) {
+ local->cont.zerofill.prebuf = *prebuf;
+ local->cont.zerofill.postbuf = *postbuf;
+ }
+
+ local->success_count++;
+
+ if ((local->success_count >= priv->wait_count)
+ && local->read_child_returned) {
+ need_unwind = 1;
+ }
+ }
+ local->op_errno = op_errno;
+ }
+ UNLOCK (&frame->lock);
+
+ if (need_unwind) {
+ local->transaction.unwind (frame, this);
+ }
+ call_count = afr_frame_return (frame);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ }
+
+ return 0;
+}
+
+static int
+afr_zerofill_wind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
+
+ local->call_count = call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_zerofill_wind_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->zerofill,
+ local->fd,
+ local->cont.zerofill.offset,
+ local->cont.zerofill.len,
+ NULL);
+
+ if (!--call_count)
+ break;
+ }
+ }
+
+ return 0;
+}
+
+static int
+afr_zerofill_done (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+
+ local = frame->local;
+
+ local->transaction.unwind (frame, this);
+
+ AFR_STACK_DESTROY (frame);
+
+ return 0;
+}
+
+static int
+afr_do_zerofill(call_frame_t *frame, xlator_t *this)
+{
+ call_frame_t *transaction_frame = NULL;
+ afr_local_t *local = NULL;
+ int op_ret = -1;
+ int op_errno = 0;
+
+ local = frame->local;
+
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ goto out;
+ }
+
+ transaction_frame->local = local;
+ frame->local = NULL;
+
+ local->op = GF_FOP_ZEROFILL;
+
+ local->transaction.fop = afr_zerofill_wind;
+ local->transaction.done = afr_zerofill_done;
+ local->transaction.unwind = afr_zerofill_unwind;
+
+ local->transaction.main_frame = frame;
+
+ local->transaction.start = local->cont.zerofill.offset;
+ local->transaction.len = 0;
+
+ op_ret = afr_transaction (transaction_frame, this,
+ AFR_DATA_TRANSACTION);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto out;
+ }
+
+ op_ret = 0;
+out:
+ if (op_ret < 0) {
+ if (transaction_frame) {
+ AFR_STACK_DESTROY (transaction_frame);
+ }
+ AFR_STACK_UNWIND (zerofill, frame, op_ret, op_errno, NULL,
+ NULL, NULL);
}
return 0;
}
+
+int
+afr_zerofill (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
+
+ priv = this->private;
+
+ if (afr_is_split_brain (this, fd->inode)) {
+ op_errno = EIO;
+ goto out;
+ }
+ QUORUM_CHECK(zerofill, out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
+
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0) {
+ goto out;
+ }
+ local->cont.zerofill.offset = offset;
+ local->cont.zerofill.len = len;
+
+ local->fd = fd_ref (fd);
+
+ afr_open_fd_fix (fd, this);
+
+ afr_do_zerofill(frame, this);
+
+ ret = 0;
+out:
+ if (ret < 0) {
+ if (transaction_frame) {
+ AFR_STACK_DESTROY (transaction_frame);
+ }
+ AFR_STACK_UNWIND (zerofill, frame, -1, op_errno, NULL,
+ NULL, NULL);
+ }
+
+ return 0;
+}
+
+/* }}} */
+
+
diff --git a/xlators/cluster/afr/src/afr-inode-write.h b/xlators/cluster/afr/src/afr-inode-write.h
index f0c2fbe00..8e93ca44a 100644
--- a/xlators/cluster/afr/src/afr-inode-write.h
+++ b/xlators/cluster/afr/src/afr-inode-write.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __INODE_WRITE_H__
@@ -22,51 +13,70 @@
int32_t
afr_chmod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode);
+ loc_t *loc, mode_t mode, dict_t *xdata);
int32_t
afr_chown (call_frame_t *frame, xlator_t *this,
- loc_t *loc, uid_t uid, gid_t gid);
+ loc_t *loc, uid_t uid, gid_t gid, dict_t *xdata);
int
afr_fchown (call_frame_t *frame, xlator_t *this,
- fd_t *fd, uid_t uid, gid_t gid);
+ fd_t *fd, uid_t uid, gid_t gid, dict_t *xdata);
int32_t
afr_fchmod (call_frame_t *frame, xlator_t *this,
- fd_t *fd, mode_t mode);
+ fd_t *fd, mode_t mode, dict_t *xdata);
int32_t
-afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
+afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
struct iovec *vector, int32_t count, off_t offset,
- struct iobref *iobref);
+ uint32_t flags, struct iobref *iobref, dict_t *xdata);
int32_t
afr_truncate (call_frame_t *frame, xlator_t *this,
- loc_t *loc, off_t offset);
+ loc_t *loc, off_t offset, dict_t *xdata);
int32_t
afr_ftruncate (call_frame_t *frame, xlator_t *this,
- fd_t *fd, off_t offset);
+ fd_t *fd, off_t offset, dict_t *xdata);
int32_t
afr_utimens (call_frame_t *frame, xlator_t *this,
- loc_t *loc, struct timespec tv[2]);
+ loc_t *loc, struct timespec tv[2], dict_t *xdata);
int
afr_setattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, struct iatt *buf, int32_t valid);
+ loc_t *loc, struct iatt *buf, int32_t valid, dict_t *xdata);
int
afr_fsetattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, struct iatt *buf, int32_t valid);
+ fd_t *fd, struct iatt *buf, int32_t valid, dict_t *xdata);
int32_t
afr_setxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *dict, int32_t flags);
+ loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata);
+
+int32_t
+afr_fsetxattr (call_frame_t *frame, xlator_t *this,
+ fd_t *fd, dict_t *dict, int32_t flags, dict_t *xdata);
int32_t
afr_removexattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name);
+ loc_t *loc, const char *name, dict_t *xdata);
+
+int32_t
+afr_fremovexattr (call_frame_t *frame, xlator_t *this,
+ fd_t *fd, const char *name, dict_t *xdata);
+int
+afr_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata);
+
+int
+afr_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+ off_t offset, size_t len, dict_t *xdata);
+
+int
+afr_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata);
#endif /* __INODE_WRITE_H__ */
diff --git a/xlators/cluster/afr/src/afr-lk-common.c b/xlators/cluster/afr/src/afr-lk-common.c
index de95a6c76..060d78f35 100644
--- a/xlators/cluster/afr/src/afr-lk-common.c
+++ b/xlators/cluster/afr/src/afr-lk-common.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include "dict.h"
@@ -31,8 +22,69 @@
#define LOCKED_YES 0x1 /* for DATA, METADATA, ENTRY and higher_path */
#define LOCKED_LOWER 0x2 /* for lower path */
+#define AFR_TRACE_INODELK_IN(frame, this, params ...) \
+ do { \
+ afr_private_t *_priv = this->private; \
+ if (!_priv->inodelk_trace) \
+ break; \
+ afr_trace_inodelk_in (frame, this, params); \
+ } while (0);
+
+#define AFR_TRACE_INODELK_OUT(frame, this, params ...) \
+ do { \
+ afr_private_t *_priv = this->private; \
+ if (!_priv->inodelk_trace) \
+ break; \
+ afr_trace_inodelk_out (frame, this, params); \
+ } while (0);
+
+#define AFR_TRACE_ENTRYLK_IN(frame, this, params ...) \
+ do { \
+ afr_private_t *_priv = this->private; \
+ if (!_priv->entrylk_trace) \
+ break; \
+ afr_trace_entrylk_in (frame, this, params); \
+ } while (0);
+
+#define AFR_TRACE_ENTRYLK_OUT(frame, this, params ...) \
+ do { \
+ afr_private_t *_priv = this->private; \
+ if (!_priv->entrylk_trace) \
+ break; \
+ afr_trace_entrylk_out (frame, this, params); \
+ } while (0);
+
int
-afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index);
+afr_entry_lockee_cmp (const void *l1, const void *l2)
+{
+ const afr_entry_lockee_t *r1 = l1;
+ const afr_entry_lockee_t *r2 = l2;
+ int ret = 0;
+ uuid_t gfid1 = {0};
+ uuid_t gfid2 = {0};
+
+ loc_gfid ((loc_t*)&r1->loc, gfid1);
+ loc_gfid ((loc_t*)&r2->loc, gfid2);
+ ret = uuid_compare (gfid1, gfid2);
+ /*Entrylks with NULL basename are the 'smallest'*/
+ if (ret == 0) {
+ if (!r1->basename)
+ return -1;
+ if (!r2->basename)
+ return 1;
+ ret = strcmp (r1->basename, r2->basename);
+ }
+
+ if (ret <= 0)
+ return -1;
+ else
+ return 1;
+}
+
+int afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index);
+
+static int
+afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this);
static uint64_t afr_lock_number = 1;
@@ -57,14 +109,13 @@ afr_set_lock_number (call_frame_t *frame, xlator_t *this)
}
void
-afr_set_lk_owner (call_frame_t *frame, xlator_t *this)
+afr_set_lk_owner (call_frame_t *frame, xlator_t *this, void *lk_owner)
{
- if (!frame->root->lk_owner) {
- gf_log (this->name, GF_LOG_TRACE,
- "Setting lk-owner=%llu",
- (unsigned long long) (unsigned long)frame->root);
- frame->root->lk_owner = (uint64_t) (unsigned long)frame->root;
- }
+ gf_log (this->name, GF_LOG_TRACE,
+ "Setting lk-owner=%llu",
+ (unsigned long long) (unsigned long)lk_owner);
+
+ set_lk_owner_from_ptr (&frame->root->lk_owner, lk_owner);
}
static int
@@ -90,29 +141,19 @@ is_afr_lock_selfheal (afr_local_t *local)
}
int32_t
-internal_lock_count (call_frame_t *frame, xlator_t *this,
- afr_fd_ctx_t *fd_ctx)
+internal_lock_count (call_frame_t *frame, xlator_t *this)
{
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
-
int32_t call_count = 0;
int i = 0;
local = frame->local;
priv = this->private;
- if (fd_ctx) {
- GF_ASSERT (local->fd);
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i] && fd_ctx->opened_on[i])
- ++call_count;
- }
- } else {
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i])
- ++call_count;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i])
+ ++call_count;
}
return call_count;
@@ -120,7 +161,7 @@ internal_lock_count (call_frame_t *frame, xlator_t *this,
static void
afr_print_inodelk (char *str, int size, int cmd,
- struct flock *flock, uint64_t owner)
+ struct gf_flock *flock, gf_lkowner_t *owner)
{
char *cmd_str = NULL;
char *type_str = NULL;
@@ -168,11 +209,11 @@ afr_print_inodelk (char *str, int size, int cmd,
}
snprintf (str, size, "lock=INODELK, cmd=%s, type=%s, "
- "start=%llu, len=%llu, pid=%llu, lk-owner=%llu",
+ "start=%llu, len=%llu, pid=%llu, lk-owner=%s",
cmd_str, type_str, (unsigned long long) flock->l_start,
(unsigned long long) flock->l_len,
(unsigned long long) flock->l_pid,
- (unsigned long long) owner);
+ lkowner_utoa (owner));
}
@@ -188,11 +229,11 @@ afr_print_lockee (char *str, int size, loc_t *loc, fd_t *fd,
void
afr_print_entrylk (char *str, int size, const char *basename,
- uint64_t owner)
+ gf_lkowner_t *owner)
{
- snprintf (str, size, "Basename=%s, lk-owner=%llu",
+ snprintf (str, size, "Basename=%s, lk-owner=%s",
basename ? basename : "<nul>",
- (unsigned long long)owner);
+ lkowner_utoa (owner));
}
static void
@@ -246,27 +287,20 @@ afr_set_lock_call_type (afr_lock_call_type_t lock_call_type,
}
static void
-afr_trace_inodelk_out (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
- afr_lock_op_type_t lk_op_type, struct flock *flock,
+afr_trace_inodelk_out (call_frame_t *frame, xlator_t *this,
+ afr_lock_call_type_t lock_call_type,
+ afr_lock_op_type_t lk_op_type, struct gf_flock *flock,
int op_ret, int op_errno, int32_t child_index)
{
- xlator_t *this = NULL;
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
char lockee[256];
char lock_call_type_str[256];
char verdict[16];
- this = THIS;
local = frame->local;
int_lock = &local->internal_lock;
- priv = this->private;
-
- if (!priv->inodelk_trace) {
- return;
- }
afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
@@ -274,45 +308,37 @@ afr_trace_inodelk_out (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
afr_print_verdict (op_ret, op_errno, verdict);
- gf_log (this->name, GF_LOG_NORMAL,
- "[%s %s] [%s] Lockee={%s} Number={%llu}",
+ gf_log (this->name, GF_LOG_INFO,
+ "[%s %s] [%s] lk-owner=%s Lockee={%s} Number={%llu}",
lock_call_type_str,
lk_op_type == AFR_LOCK_OP ? "LOCK REPLY" : "UNLOCK REPLY",
- verdict,
- lockee,
+ verdict, lkowner_utoa (&frame->root->lk_owner), lockee,
(unsigned long long) int_lock->lock_number);
}
static void
-afr_trace_inodelk_in (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
- afr_lock_op_type_t lk_op_type, struct flock *flock,
+afr_trace_inodelk_in (call_frame_t *frame, xlator_t *this,
+ afr_lock_call_type_t lock_call_type,
+ afr_lock_op_type_t lk_op_type, struct gf_flock *flock,
int32_t cmd, int32_t child_index)
{
- xlator_t *this = NULL;
afr_local_t *local = NULL;
afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
char lock[256];
char lockee[256];
char lock_call_type_str[256];
- this = THIS;
local = frame->local;
int_lock = &local->internal_lock;
- priv = this->private;
- if (!priv->inodelk_trace) {
- return;
- }
-
- afr_print_inodelk (lock, 256, cmd, flock, frame->root->lk_owner);
+ afr_print_inodelk (lock, 256, cmd, flock, &frame->root->lk_owner);
afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
afr_set_lock_call_type (lock_call_type, lock_call_type_str, int_lock);
- gf_log (this->name, GF_LOG_NORMAL,
+ gf_log (this->name, GF_LOG_INFO,
"[%s %s] Lock={%s} Lockee={%s} Number={%llu}",
lock_call_type_str,
lk_op_type == AFR_LOCK_OP ? "LOCK REQUEST" : "UNLOCK REQUEST",
@@ -322,20 +348,21 @@ afr_trace_inodelk_in (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
}
static void
-afr_trace_entrylk_in (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
+afr_trace_entrylk_in (call_frame_t *frame, xlator_t *this,
+ afr_lock_call_type_t lock_call_type,
afr_lock_op_type_t lk_op_type, const char *basename,
- int32_t child_index)
+ int32_t cookie)
{
- xlator_t *this = NULL;
afr_local_t *local = NULL;
afr_internal_lock_t *int_lock = NULL;
afr_private_t *priv = NULL;
+ int child_index = 0;
+ int lockee_no = 0;
char lock[256];
char lockee[256];
char lock_call_type_str[256];
- this = THIS;
local = frame->local;
int_lock = &local->internal_lock;
priv = this->private;
@@ -343,36 +370,41 @@ afr_trace_entrylk_in (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
if (!priv->entrylk_trace) {
return;
}
+ lockee_no = cookie / priv->child_count;
+ child_index = cookie % priv->child_count;
- afr_print_entrylk (lock, 256, basename, frame->root->lk_owner);
- afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
+ afr_print_entrylk (lock, 256, basename, &frame->root->lk_owner);
+ afr_print_lockee (lockee, 256, &int_lock->lockee[lockee_no].loc, local->fd,
+ child_index);
afr_set_lock_call_type (lock_call_type, lock_call_type_str, int_lock);
- gf_log (this->name, GF_LOG_NORMAL,
- "[%s %s] Lock={%s} Lockee={%s} Number={%llu}",
+ gf_log (this->name, GF_LOG_INFO,
+ "[%s %s] Lock={%s} Lockee={%s} Number={%llu}, Cookie={%d}",
lock_call_type_str,
lk_op_type == AFR_LOCK_OP ? "LOCK REQUEST" : "UNLOCK REQUEST",
lock, lockee,
- (unsigned long long) int_lock->lock_number);
+ (unsigned long long) int_lock->lock_number,
+ cookie);
}
static void
-afr_trace_entrylk_out (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
- afr_lock_op_type_t lk_op_type, const char *basename, int op_ret,
- int op_errno, int32_t child_index)
+afr_trace_entrylk_out (call_frame_t *frame, xlator_t *this,
+ afr_lock_call_type_t lock_call_type,
+ afr_lock_op_type_t lk_op_type, const char *basename,
+ int op_ret, int op_errno, int32_t cookie)
{
- xlator_t *this = NULL;
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
+ int lockee_no = 0;
+ int child_index = 0;
char lock[256];
char lockee[256];
char lock_call_type_str[256];
char verdict[16];
- this = THIS;
local = frame->local;
int_lock = &local->internal_lock;
priv = this->private;
@@ -380,20 +412,25 @@ afr_trace_entrylk_out (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
if (!priv->entrylk_trace) {
return;
}
+ lockee_no = cookie / priv->child_count;
+ child_index = cookie % priv->child_count;
- afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
+ afr_print_entrylk (lock, 256, basename, &frame->root->lk_owner);
+ afr_print_lockee (lockee, 256, &int_lock->lockee[lockee_no].loc, local->fd,
+ child_index);
afr_set_lock_call_type (lock_call_type, lock_call_type_str, int_lock);
afr_print_verdict (op_ret, op_errno, verdict);
- gf_log (this->name, GF_LOG_NORMAL,
- "[%s %s] [%s] Lock={%s} Lockee={%s} Number={%llu}",
+ gf_log (this->name, GF_LOG_INFO,
+ "[%s %s] [%s] Lock={%s} Lockee={%s} Number={%llu} Cookie={%d}",
lock_call_type_str,
lk_op_type == AFR_LOCK_OP ? "LOCK REPLY" : "UNLOCK REPLY",
verdict,
lock, lockee,
- (unsigned long long) int_lock->lock_number);
+ (unsigned long long) int_lock->lock_number,
+ cookie);
}
@@ -431,8 +468,8 @@ is_afr_lock_transaction (afr_local_t *local)
int ret = 0;
switch (local->transaction.type) {
- case AFR_DATA_TRANSACTION:
- case AFR_METADATA_TRANSACTION:
+ case AFR_DATA_TRANSACTION:
+ case AFR_METADATA_TRANSACTION:
ret = 1;
break;
@@ -446,6 +483,47 @@ is_afr_lock_transaction (afr_local_t *local)
return ret;
}
+int
+afr_init_entry_lockee (afr_entry_lockee_t *lockee, afr_local_t *local,
+ loc_t *loc, char *basename, int child_count)
+{
+ int ret = -1;
+
+ loc_copy (&lockee->loc, loc);
+ lockee->basename = (basename)? gf_strdup (basename): NULL;
+ if (basename && !lockee->basename)
+ goto out;
+
+ lockee->locked_count = 0;
+ lockee->locked_nodes = GF_CALLOC (child_count,
+ sizeof (*lockee->locked_nodes),
+ gf_afr_mt_afr_node_character);
+
+ if (!lockee->locked_nodes)
+ goto out;
+
+ ret = 0;
+out:
+ return ret;
+
+}
+
+void
+afr_entry_lockee_cleanup (afr_internal_lock_t *int_lock)
+{
+ int i = 0;
+
+ for (i = 0; i < int_lock->lockee_count; i++) {
+ loc_wipe (&int_lock->lockee[i].loc);
+ if (int_lock->lockee[i].basename)
+ GF_FREE (int_lock->lockee[i].basename);
+ if (int_lock->lockee[i].locked_nodes)
+ GF_FREE (int_lock->lockee[i].locked_nodes);
+ }
+
+ return;
+}
+
static int
initialize_entrylk_variables (call_frame_t *frame, xlator_t *this)
{
@@ -463,8 +541,13 @@ initialize_entrylk_variables (call_frame_t *frame, xlator_t *this)
int_lock->lock_op_ret = -1;
int_lock->lock_op_errno = 0;
- for (i = 0; i < priv->child_count; i++) {
- int_lock->entry_locked_nodes[i] = 0;
+ for (i = 0; i < AFR_LOCKEE_COUNT_MAX; i++) {
+ if (!int_lock->lockee[i].locked_nodes)
+ break;
+ int_lock->lockee[i].locked_count = 0;
+ memset (int_lock->lockee[i].locked_nodes, 0,
+ sizeof (*int_lock->lockee[i].locked_nodes) *
+ priv->child_count);
}
return 0;
@@ -476,20 +559,23 @@ initialize_inodelk_variables (call_frame_t *frame, xlator_t *this)
afr_local_t *local = NULL;
afr_internal_lock_t *int_lock = NULL;
afr_private_t *priv = NULL;
-
- int i = 0;
+ afr_inodelk_t *inodelk = NULL;
priv = this->private;
local = frame->local;
int_lock = &local->internal_lock;
- int_lock->inodelk_lock_count = 0;
- int_lock->lock_op_ret = -1;
- int_lock->lock_op_errno = 0;
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
- for (i = 0; i < priv->child_count; i++) {
- int_lock->inode_locked_nodes[i] = 0;
- }
+ inodelk->lock_count = 0;
+ int_lock->lk_attempted_count = 0;
+ int_lock->lock_op_ret = -1;
+ int_lock->lock_op_errno = 0;
+
+ memset (inodelk->locked_nodes, 0,
+ sizeof (*inodelk->locked_nodes) * priv->child_count);
+ memset (int_lock->locked_nodes, 0,
+ sizeof (*int_lock->locked_nodes) * priv->child_count);
return 0;
}
@@ -497,24 +583,36 @@ initialize_inodelk_variables (call_frame_t *frame, xlator_t *this)
loc_t *
lower_path (loc_t *l1, const char *b1, loc_t *l2, const char *b2)
{
- int ret = 0;
+ int ret = 0;
- ret = strcmp (l1->path, l2->path);
+ ret = uuid_compare (l1->inode->gfid, l2->inode->gfid);
- if (ret == 0)
- ret = strcmp (b1, b2);
+ if (ret == 0)
+ ret = strcmp (b1, b2);
- if (ret <= 0)
- return l1;
- else
- return l2;
+ if (ret <= 0)
+ return l1;
+ else
+ return l2;
+}
+
+int
+afr_lockee_locked_nodes_count (afr_internal_lock_t *int_lock)
+{
+ int call_count = 0;
+ int i = 0;
+
+ for (i = 0; i < int_lock->lockee_count; i++)
+ call_count += int_lock->lockee[i].locked_count;
+
+ return call_count;
}
int
afr_locked_nodes_count (unsigned char *locked_nodes, int child_count)
{
- int i;
+ int i = 0;
int call_count = 0;
for (i = 0; i < child_count; i++) {
@@ -528,44 +626,63 @@ afr_locked_nodes_count (unsigned char *locked_nodes, int child_count)
/* FIXME: What if UNLOCK fails */
static int32_t
afr_unlock_common_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
afr_internal_lock_t *int_lock = NULL;
- int call_count = 0;
+ int call_count = 0;
- local = frame->local;
+ local = frame->local;
int_lock = &local->internal_lock;
- LOCK (&frame->lock);
- {
- call_count = --int_lock->lk_call_count;
- }
- UNLOCK (&frame->lock);
+ LOCK (&frame->lock);
+ {
+ call_count = --int_lock->lk_call_count;
+ }
+ UNLOCK (&frame->lock);
- if (call_count == 0) {
+ if (call_count == 0) {
gf_log (this->name, GF_LOG_TRACE,
"All internal locks unlocked");
int_lock->lock_cbk (frame, this);
}
- return 0;
+ return 0;
}
static int32_t
afr_unlock_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_trace_inodelk_out (frame, AFR_INODELK_TRANSACTION,
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
+ int32_t child_index = (long)cookie;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ int_lock = &local->internal_lock;
+
+ AFR_TRACE_INODELK_OUT (frame, this, AFR_INODELK_TRANSACTION,
AFR_UNLOCK_OP, NULL, op_ret,
- op_errno, (long) cookie);
+ op_errno, child_index);
+
+ priv = this->private;
if (op_ret < 0 && op_errno != ENOTCONN && op_errno != EBADFD) {
- gf_log (this->name, GF_LOG_TRACE,
- "Unlock failed for some reason");
+ gf_log (this->name, GF_LOG_INFO, "%s: unlock failed on subvolume %s "
+ "with lock owner %s", local->loc.path,
+ priv->children[child_index]->name,
+ lkowner_utoa (&frame->root->lk_owner));
}
- afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno);
+
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+ inodelk->locked_nodes[child_index] &= LOCKED_NO;
+ if (local->transaction.eager_lock)
+ local->transaction.eager_lock[child_index] = 0;
+
+ afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno, xdata);
return 0;
@@ -575,22 +692,30 @@ static int
afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
{
afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
-
- struct flock flock;
+ struct gf_flock flock = {0,};
+ struct gf_flock full_flock = {0,};
+ struct gf_flock *flock_use = NULL;
int call_count = 0;
int i = 0;
+ int piggyback = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+
local = frame->local;
int_lock = &local->internal_lock;
priv = this->private;
- flock.l_start = int_lock->lk_flock.l_start;
- flock.l_len = int_lock->lk_flock.l_len;
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+
+ flock.l_start = inodelk->flock.l_start;
+ flock.l_len = inodelk->flock.l_len;
flock.l_type = F_UNLCK;
- call_count = afr_locked_nodes_count (int_lock->inode_locked_nodes,
+ full_flock.l_type = F_UNLCK;
+ call_count = afr_locked_nodes_count (inodelk->locked_nodes,
priv->child_count);
int_lock->lk_call_count = call_count;
@@ -602,55 +727,107 @@ afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
goto out;
}
+ if (local->fd)
+ fd_ctx = afr_fd_ctx_get (local->fd, this);
+
for (i = 0; i < priv->child_count; i++) {
- if (int_lock->inode_locked_nodes[i] & LOCKED_YES) {
- if (local->fd) {
- afr_trace_inodelk_in (frame, AFR_INODELK_TRANSACTION,
- AFR_UNLOCK_OP, &flock, F_SETLK, i);
-
- STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
- (void *) (long)i,
- priv->children[i],
- priv->children[i]->fops->finodelk,
- this->name, local->fd,
- F_SETLK, &flock);
+ if ((inodelk->locked_nodes[i] & LOCKED_YES) != LOCKED_YES)
+ continue;
- if (!--call_count)
- break;
+ if (local->fd) {
+ flock_use = &flock;
+ if (!local->transaction.eager_lock[i]) {
+ goto wind;
+ }
- } else {
- afr_trace_inodelk_in (frame, AFR_INODELK_TRANSACTION,
- AFR_UNLOCK_OP, &flock, F_SETLK, i);
+ piggyback = 0;
- STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
- (void *) (long)i,
- priv->children[i],
- priv->children[i]->fops->inodelk,
- this->name, &local->loc,
- F_SETLK, &flock);
+ LOCK (&local->fd->lock);
+ {
+ if (fd_ctx->lock_piggyback[i]) {
+ fd_ctx->lock_piggyback[i]--;
+ piggyback = 1;
+ } else {
+ fd_ctx->lock_acquired[i]--;
+ }
+ }
+ UNLOCK (&local->fd->lock);
+ if (piggyback) {
+ afr_unlock_inodelk_cbk (frame, (void *) (long) i,
+ this, 1, 0, NULL);
if (!--call_count)
break;
-
+ continue;
}
- }
+ flock_use = &full_flock;
+ wind:
+ AFR_TRACE_INODELK_IN (frame, this,
+ AFR_INODELK_TRANSACTION,
+ AFR_UNLOCK_OP, flock_use, F_SETLK,
+ i);
- }
+ STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
+ (void *) (long)i,
+ priv->children[i],
+ priv->children[i]->fops->finodelk,
+ int_lock->domain, local->fd,
+ F_SETLK, flock_use, NULL);
+
+ if (!--call_count)
+ break;
+ } else {
+ AFR_TRACE_INODELK_IN (frame, this,
+ AFR_INODELK_TRANSACTION,
+ AFR_UNLOCK_OP, &flock, F_SETLK, i);
+
+ STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
+ (void *) (long)i,
+ priv->children[i],
+ priv->children[i]->fops->inodelk,
+ int_lock->domain, &local->loc,
+ F_SETLK, &flock, NULL);
+
+ if (!--call_count)
+ break;
+ }
+ }
out:
return 0;
}
static int32_t
afr_unlock_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_trace_entrylk_out (frame, AFR_ENTRYLK_TRANSACTION,
- AFR_UNLOCK_OP, NULL, op_ret,
- op_errno, (long) cookie);
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ int32_t child_index = 0;
+ int lockee_no = 0;
+
+ priv = this->private;
+ lockee_no = (int)((long) cookie) / priv->child_count;
+ child_index = (int) ((long) cookie) % priv->child_count;
+
+ local = frame->local;
+ int_lock = &local->internal_lock;
+
+ AFR_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
+ AFR_UNLOCK_OP,
+ int_lock->lockee[lockee_no].basename, op_ret,
+ op_errno, (int) ((long)cookie));
+
+ if (op_ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "%s: unlock failed on %d, reason: %s",
+ local->loc.path, child_index, strerror (op_errno));
+ }
- afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno);
+ int_lock->lockee[lockee_no].locked_nodes[child_index] &= LOCKED_NO;
+ afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno, NULL);
return 0;
}
@@ -658,25 +835,22 @@ afr_unlock_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
static int
afr_unlock_entrylk (call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- const char *basename = NULL;
- loc_t *loc = NULL;
-
- int call_count = 0;
- int i = -1;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int index = 0;
+ int lockee_no = 0;
+ int copies = 0;
+ int i = -1;
local = frame->local;
int_lock = &local->internal_lock;
priv = this->private;
+ copies = priv->child_count;
- basename = int_lock->lk_basename;
- if (int_lock->lk_loc)
- loc = int_lock->lk_loc;
+ call_count = afr_lockee_locked_nodes_count (int_lock);
- call_count = afr_locked_nodes_count (int_lock->entry_locked_nodes,
- priv->child_count);
int_lock->lk_call_count = call_count;
if (!call_count){
@@ -686,18 +860,23 @@ afr_unlock_entrylk (call_frame_t *frame, xlator_t *this)
goto out;
}
- for (i = 0; i < priv->child_count; i++) {
- if (int_lock->entry_locked_nodes[i] & LOCKED_YES) {
- afr_trace_entrylk_in (frame, AFR_ENTRYLK_NB_TRANSACTION,
- AFR_UNLOCK_OP, basename, i);
+ for (i = 0; i < int_lock->lockee_count * priv->child_count; i++) {
+ lockee_no = i / copies;
+ index = i % copies;
+ if (int_lock->lockee[lockee_no].locked_nodes[index] & LOCKED_YES) {
+ AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_NB_TRANSACTION,
+ AFR_UNLOCK_OP,
+ int_lock->lockee[lockee_no].basename,
+ i);
STACK_WIND_COOKIE (frame, afr_unlock_entrylk_cbk,
(void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->entrylk,
- this->name,
- loc, basename,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK);
+ priv->children[index],
+ priv->children[index]->fops->entrylk,
+ int_lock->domain,
+ &int_lock->lockee[lockee_no].loc,
+ int_lock->lockee[lockee_no].basename,
+ ENTRYLK_UNLOCK, ENTRYLK_WRLCK, NULL);
if (!--call_count)
break;
@@ -711,155 +890,85 @@ out:
static int32_t
afr_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- int done = 0;
- int child_index = (long) cookie;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- if (op_errno == ENOSYS) {
- /* return ENOTSUP */
- gf_log (this->name, GF_LOG_ERROR,
- "subvolume does not support locking. "
- "please load features/posix-locks xlator on server");
- local->op_ret = op_ret;
- int_lock->lock_op_ret = op_ret;
- done = 1;
- }
-
- local->child_up[child_index] = 0;
- local->op_errno = op_errno;
- int_lock->lock_op_errno = op_errno;
- }
- }
- UNLOCK (&frame->lock);
-
- if ((op_ret == -1) &&
- (op_errno == ENOSYS)) {
- afr_unlock (frame, this);
- } else {
- if (op_ret == 0) {
- int_lock->locked_nodes[child_index]
- |= LOCKED_YES;
- int_lock->lock_count++;
- }
- afr_lock_blocking (frame, this, child_index + 1);
- }
-
- return 0;
-}
-
-static int32_t
-afr_blocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
-{
- afr_trace_inodelk_out (frame, AFR_INODELK_TRANSACTION,
- AFR_LOCK_OP, NULL, op_ret,
- op_errno, (long) cookie);
-
- afr_lock_cbk (frame, cookie, this, op_ret, op_errno);
- return 0;
-
-}
-
-static int32_t
-afr_lock_lower_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- loc_t *lower = NULL;
- loc_t *higher = NULL;
- const char *lower_name = NULL;
- const char *higher_name = NULL;
-
- int child_index = (long) cookie;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int cky = (long) cookie;
+ int child_index = 0;
+ int lockee_no = 0;
priv = this->private;
local = frame->local;
int_lock = &local->internal_lock;
+ child_index = ((int)cky) % priv->child_count;
+ lockee_no = ((int)cky) / priv->child_count;
+
LOCK (&frame->lock);
{
if (op_ret == -1) {
if (op_errno == ENOSYS) {
/* return ENOTSUP */
-
gf_log (this->name, GF_LOG_ERROR,
"subvolume does not support locking. "
- "please load features/posix-locks xlator on server");
-
- local->op_ret = op_ret;
+ "please load features/locks xlator on server");
+ local->op_ret = op_ret;
+ int_lock->lock_op_ret = op_ret;
}
- local->child_up[child_index] = 0;
- local->op_errno = op_errno;
+ local->op_errno = op_errno;
+ int_lock->lock_op_errno = op_errno;
}
+
+ int_lock->lk_attempted_count++;
}
UNLOCK (&frame->lock);
- if (op_ret != 0) {
+ if ((op_ret == -1) &&
+ (op_errno == ENOSYS)) {
afr_unlock (frame, this);
- goto out;
} else {
- int_lock->lower_locked_nodes[child_index] |= LOCKED_LOWER;
- int_lock->lock_count++;
+ if (op_ret == 0) {
+ if (local->transaction.type == AFR_ENTRY_TRANSACTION ||
+ local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
+ int_lock->lockee[lockee_no].locked_nodes[child_index] |= LOCKED_YES;
+ int_lock->lockee[lockee_no].locked_count++;
+ int_lock->entrylk_lock_count++;
+ } else {
+ int_lock->locked_nodes[child_index] |= LOCKED_YES;
+ int_lock->lock_count++;
+ }
+ }
+ afr_lock_blocking (frame, this, cky + 1);
}
- /* The lower path has been locked. Now lock the higher path */
-
- lower = lower_path (&local->transaction.parent_loc,
- local->transaction.basename,
- &local->transaction.new_parent_loc,
- local->transaction.new_basename);
-
- lower_name = (lower == &local->transaction.parent_loc ?
- local->transaction.basename :
- local->transaction.new_basename);
-
- higher = (lower == &local->transaction.parent_loc ?
- &local->transaction.new_parent_loc :
- &local->transaction.parent_loc);
-
- higher_name = (higher == &local->transaction.parent_loc ?
- local->transaction.basename :
- local->transaction.new_basename);
-
- afr_trace_entrylk_in (frame, AFR_ENTRYLK_TRANSACTION,
- AFR_LOCK_OP, higher_name, child_index);
-
+ return 0;
+}
- STACK_WIND_COOKIE (frame, afr_lock_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->entrylk,
- this->name, higher, higher_name,
- ENTRYLK_LOCK, ENTRYLK_WRLCK);
+static int32_t
+afr_blocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+ AFR_TRACE_INODELK_OUT (frame, this, AFR_INODELK_TRANSACTION,
+ AFR_LOCK_OP, NULL, op_ret,
+ op_errno, (long) cookie);
-out:
+ afr_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
return 0;
+
}
static int32_t
afr_blocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_trace_entrylk_out (frame, AFR_ENTRYLK_TRANSACTION,
+ AFR_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
AFR_LOCK_OP, NULL, op_ret,
op_errno, (long)cookie);
- afr_lock_cbk (frame, cookie, this, op_ret, op_errno);
+ afr_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
return 0;
}
@@ -867,6 +976,7 @@ static int
afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this)
{
afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
@@ -875,20 +985,18 @@ afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this)
int_lock = &local->internal_lock;
switch (local->transaction.type) {
- case AFR_DATA_TRANSACTION:
- case AFR_METADATA_TRANSACTION:
- memcpy (int_lock->inode_locked_nodes,
- int_lock->locked_nodes,
- priv->child_count);
- int_lock->inodelk_lock_count = int_lock->lock_count;
+ case AFR_DATA_TRANSACTION:
+ case AFR_METADATA_TRANSACTION:
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+ memcpy (inodelk->locked_nodes, int_lock->locked_nodes,
+ sizeof (*inodelk->locked_nodes) * priv->child_count);
+ inodelk->lock_count = int_lock->lock_count;
break;
case AFR_ENTRY_RENAME_TRANSACTION:
case AFR_ENTRY_TRANSACTION:
- memcpy (int_lock->entry_locked_nodes,
- int_lock->locked_nodes,
- priv->child_count);
- int_lock->entrylk_lock_count = int_lock->lock_count;
+ /*entrylk_count is being used in both non-blocking and blocking
+ * modes */
break;
}
@@ -896,42 +1004,78 @@ afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this)
}
+static inline gf_boolean_t
+afr_is_entrylk (afr_internal_lock_t *int_lock,
+ afr_transaction_type trans_type)
+{
+ gf_boolean_t is_entrylk = _gf_false;
+
+ if ((int_lock->transaction_lk_type == AFR_SELFHEAL_LK) &&
+ int_lock->selfheal_lk_type == AFR_ENTRY_SELF_HEAL_LK) {
+
+ is_entrylk = _gf_true;
+
+ } else if ((int_lock->transaction_lk_type == AFR_TRANSACTION_LK) &&
+ (trans_type == AFR_ENTRY_TRANSACTION ||
+ trans_type == AFR_ENTRY_RENAME_TRANSACTION)) {
+
+ is_entrylk = _gf_true;
+
+ } else {
+ is_entrylk = _gf_false;
+ }
+
+ return is_entrylk;
+}
+
+static gf_boolean_t
+_is_lock_wind_needed (afr_local_t *local, int child_index)
+{
+ if (!local->child_up[child_index])
+ return _gf_false;
+
+ return _gf_true;
+}
+
int
-afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
+afr_lock_blocking (call_frame_t *frame, xlator_t *this, int cookie)
{
afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
- loc_t *lower = NULL;
- loc_t *higher = NULL;
- const char *lower_name = NULL;
- const char *higher_name = NULL;
-
- struct flock flock;
- uint64_t ctx;
+ afr_private_t *priv = NULL;
+ struct gf_flock flock = {0,};
+ uint64_t ctx = 0;
int ret = 0;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
-
- flock.l_start = int_lock->lk_flock.l_start;
- flock.l_len = int_lock->lk_flock.l_len;
- flock.l_type = int_lock->lk_flock.l_type;
+ int child_index = 0;
+ int lockee_no = 0;
+ gf_boolean_t is_entrylk = _gf_false;
+
+ local = frame->local;
+ int_lock = &local->internal_lock;
+ priv = this->private;
+ child_index = cookie % priv->child_count;
+ lockee_no = cookie / priv->child_count;
+ is_entrylk = afr_is_entrylk (int_lock, local->transaction.type);
+
+
+ if (!is_entrylk) {
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+ flock.l_start = inodelk->flock.l_start;
+ flock.l_len = inodelk->flock.l_len;
+ flock.l_type = inodelk->flock.l_type;
+ }
if (local->fd) {
ret = fd_ctx_get (local->fd, this, &ctx);
if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
+ gf_log (this->name, GF_LOG_INFO,
"unable to get fd ctx for fd=%p",
local->fd);
local->op_ret = -1;
int_lock->lock_op_ret = -1;
- local->op_errno = EINVAL;
- int_lock->lock_op_errno = EINVAL;
afr_copy_locked_nodes (frame, this);
@@ -939,49 +1083,27 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
return 0;
}
-
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
- /* skip over children that or down
- or don't have the fd open */
-
- while ((child_index < priv->child_count)
- && (!local->child_up[child_index]
- || !fd_ctx->opened_on[child_index]))
-
- child_index++;
- } else {
- /* skip over children that are down */
- while ((child_index < priv->child_count)
- && !local->child_up[child_index])
- child_index++;
}
- if ((child_index == priv->child_count) &&
- int_lock->lock_count == 0) {
+ if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
+ if ((is_entrylk && int_lock->entrylk_lock_count == 0) ||
+ (!is_entrylk && int_lock->lock_count == 0)) {
+ gf_log (this->name, GF_LOG_INFO,
+ "unable to lock on even one child");
- gf_log (this->name, GF_LOG_DEBUG,
- "unable to lock on even one child");
-
- local->op_ret = -1;
- int_lock->lock_op_ret = -1;
- local->op_errno = EAGAIN;
- int_lock->lock_op_errno = EAGAIN;
-
- afr_copy_locked_nodes (frame, this);
-
- afr_unlock(frame, this);
+ local->op_ret = -1;
+ int_lock->lock_op_ret = -1;
- return 0;
+ afr_copy_locked_nodes (frame, this);
- }
+ afr_unlock(frame, this);
- if ((child_index == priv->child_count)
- || (int_lock->lock_count ==
- afr_up_children_count (priv->child_count,
- local->child_up))) {
+ return 0;
+ }
+ }
- /* we're done locking */
+ if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
+ /* we're done locking */
gf_log (this->name, GF_LOG_DEBUG,
"we're done locking");
@@ -990,107 +1112,85 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
int_lock->lock_op_ret = 0;
int_lock->lock_cbk (frame, this);
- return 0;
- }
+ return 0;
+ }
- switch (local->transaction.type) {
- case AFR_DATA_TRANSACTION:
- case AFR_METADATA_TRANSACTION:
+ if (!_is_lock_wind_needed (local, child_index)) {
+ afr_lock_blocking (frame, this, cookie + 1);
+ return 0;
+ }
- if (local->fd) {
- afr_trace_inodelk_in (frame, AFR_INODELK_TRANSACTION,
+ switch (local->transaction.type) {
+ case AFR_DATA_TRANSACTION:
+ case AFR_METADATA_TRANSACTION:
+
+ if (local->fd) {
+ AFR_TRACE_INODELK_IN (frame, this,
+ AFR_INODELK_TRANSACTION,
AFR_LOCK_OP, &flock, F_SETLKW,
child_index);
- STACK_WIND_COOKIE (frame, afr_blocking_inodelk_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->finodelk,
- this->name, local->fd,
- F_SETLKW, &flock);
+ STACK_WIND_COOKIE (frame, afr_blocking_inodelk_cbk,
+ (void *) (long) child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->finodelk,
+ int_lock->domain, local->fd,
+ F_SETLKW, &flock, NULL);
- } else {
- afr_trace_inodelk_in (frame, AFR_INODELK_TRANSACTION,
+ } else {
+ AFR_TRACE_INODELK_IN (frame, this,
+ AFR_INODELK_TRANSACTION,
AFR_LOCK_OP, &flock, F_SETLKW,
child_index);
- STACK_WIND_COOKIE (frame, afr_blocking_inodelk_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->inodelk,
- this->name, &local->loc,
- F_SETLKW, &flock);
- }
-
- break;
-
- case AFR_ENTRY_RENAME_TRANSACTION:
- {
- lower = lower_path (&local->transaction.parent_loc,
- local->transaction.basename,
- &local->transaction.new_parent_loc,
- local->transaction.new_basename);
-
- lower_name = (lower == &local->transaction.parent_loc ?
- local->transaction.basename :
- local->transaction.new_basename);
-
- higher = (lower == &local->transaction.parent_loc ?
- &local->transaction.new_parent_loc :
- &local->transaction.parent_loc);
-
- higher_name = (higher == &local->transaction.parent_loc ?
- local->transaction.basename :
- local->transaction.new_basename);
-
- afr_trace_entrylk_in (frame, AFR_ENTRYLK_TRANSACTION,
- AFR_LOCK_OP, lower_name, child_index);
-
-
- STACK_WIND_COOKIE (frame, afr_lock_lower_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->entrylk,
- this->name, lower, lower_name,
- ENTRYLK_LOCK, ENTRYLK_WRLCK);
-
- break;
- }
+ STACK_WIND_COOKIE (frame, afr_blocking_inodelk_cbk,
+ (void *) (long) child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->inodelk,
+ int_lock->domain, &local->loc,
+ F_SETLKW, &flock, NULL);
+ }
- case AFR_ENTRY_TRANSACTION:
- if (local->fd) {
- afr_trace_entrylk_in (frame, AFR_ENTRYLK_TRANSACTION,
- AFR_LOCK_OP, local->transaction.basename,
- child_index);
+ break;
- STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->fentrylk,
- this->name, local->fd,
- local->transaction.basename,
- ENTRYLK_LOCK, ENTRYLK_WRLCK);
- } else {
- afr_trace_entrylk_in (frame, AFR_ENTRYLK_TRANSACTION,
+ case AFR_ENTRY_RENAME_TRANSACTION:
+ case AFR_ENTRY_TRANSACTION:
+ /*Accounting for child_index increments on 'down'
+ *and 'fd-less' children */
+
+ if (local->fd) {
+ AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_TRANSACTION,
+ AFR_LOCK_OP,
+ int_lock->lockee[lockee_no].basename,
+ cookie);
+
+ STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
+ (void *) (long) cookie,
+ priv->children[child_index],
+ priv->children[child_index]->fops->fentrylk,
+ int_lock->domain, local->fd,
+ int_lock->lockee[lockee_no].basename,
+ ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
+ } else {
+ AFR_TRACE_ENTRYLK_IN (frame, this,
+ AFR_ENTRYLK_TRANSACTION,
AFR_LOCK_OP, local->transaction.basename,
child_index);
- STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->entrylk,
- this->name,
- &local->transaction.parent_loc,
- local->transaction.basename,
- ENTRYLK_LOCK, ENTRYLK_WRLCK);
- }
-
- break;
- }
-
- return 0;
+ STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
+ (void *) (long) cookie,
+ priv->children[child_index],
+ priv->children[child_index]->fops->entrylk,
+ int_lock->domain,
+ &int_lock->lockee[lockee_no].loc,
+ int_lock->lockee[lockee_no].basename,
+ ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
+ }
+ break;
+ }
+ return 0;
}
int32_t
@@ -1099,19 +1199,25 @@ afr_blocking_lock (call_frame_t *frame, xlator_t *this)
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
+ int up_count = 0;
priv = this->private;
local = frame->local;
int_lock = &local->internal_lock;
switch (local->transaction.type) {
- case AFR_DATA_TRANSACTION:
- case AFR_METADATA_TRANSACTION:
+ case AFR_DATA_TRANSACTION:
+ case AFR_METADATA_TRANSACTION:
initialize_inodelk_variables (frame, this);
break;
case AFR_ENTRY_RENAME_TRANSACTION:
case AFR_ENTRY_TRANSACTION:
+ up_count = afr_up_children_count (local->child_up,
+ priv->child_count);
+ int_lock->lk_call_count = int_lock->lk_expected_count
+ = (int_lock->lockee_count *
+ up_count);
initialize_entrylk_variables (frame, this);
break;
}
@@ -1123,60 +1229,68 @@ afr_blocking_lock (call_frame_t *frame, xlator_t *this)
static int32_t
afr_nonblocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
+ afr_local_t *local = NULL;
int call_count = 0;
- int child_index = (long) cookie;
+ int child_index = (long) cookie;
+ int copies = 0;
+ int index = 0;
+ int lockee_no = 0;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
- local = frame->local;
+ copies = priv->child_count;
+ index = child_index % copies;
+ lockee_no = child_index / copies;
+
+ local = frame->local;
int_lock = &local->internal_lock;
- priv = this->private;
- afr_trace_entrylk_out (frame, AFR_ENTRYLK_TRANSACTION,
- AFR_LOCK_OP, NULL, op_ret,
+ AFR_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
+ AFR_LOCK_OP,
+ int_lock->lockee[lockee_no].basename, op_ret,
op_errno, (long) cookie);
- LOCK (&frame->lock);
- {
- call_count = --int_lock->lk_call_count;
- }
- UNLOCK (&frame->lock);
-
- if (op_ret < 0 ) {
+ LOCK (&frame->lock);
+ {
+ if (op_ret < 0 ) {
if (op_errno == ENOSYS) {
- /* return ENOTSUP */
+ /* return ENOTSUP */
gf_log (this->name, GF_LOG_ERROR,
"subvolume does not support locking. "
- "please load features/posix-locks xlator on server");
+ "please load features/locks xlator on server");
local->op_ret = op_ret;
- int_lock->lock_op_ret = op_ret;
+ int_lock->lock_op_ret = op_ret;
- local->child_up[child_index] = 0;
- int_lock->lock_op_errno = op_errno;
- local->op_errno = op_errno;
- }
- } else if (op_ret == 0) {
- int_lock->entry_locked_nodes[child_index]
- |= LOCKED_YES;
- int_lock->entrylk_lock_count++;
+ int_lock->lock_op_errno = op_errno;
+ local->op_errno = op_errno;
+ }
+ } else if (op_ret == 0) {
+ int_lock->lockee[lockee_no].locked_nodes[index] |= \
+ LOCKED_YES;
+ int_lock->lockee[lockee_no].locked_count++;
+ int_lock->entrylk_lock_count++;
+ }
+
+ call_count = --int_lock->lk_call_count;
}
+ UNLOCK (&frame->lock);
if (call_count == 0) {
gf_log (this->name, GF_LOG_TRACE,
"Last locking reply received");
- /* all locks successfull. Proceed to call FOP */
+ /* all locks successful. Proceed to call FOP */
if (int_lock->entrylk_lock_count ==
- afr_up_children_count (priv->child_count, local->child_up)) {
+ int_lock->lk_expected_count) {
gf_log (this->name, GF_LOG_TRACE,
"All servers locked. Calling the cbk");
int_lock->lock_op_ret = 0;
int_lock->lock_cbk (frame, this);
}
- /* Not all locks were successfull. Unlock and try locking
+ /* Not all locks were successful. Unlock and try locking
again, this time with serially blocking locks */
else {
gf_log (this->name, GF_LOG_TRACE,
@@ -1193,33 +1307,27 @@ afr_nonblocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int
afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
- const char *basename = NULL;
- loc_t *loc = NULL;
-
- int32_t call_count = 0;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int copies = 0;
+ int index = 0;
+ int lockee_no = 0;
+ int32_t call_count = 0;
int i = 0;
- uint64_t ctx;
- int ret = 0;
local = frame->local;
int_lock = &local->internal_lock;
priv = this->private;
+ copies = priv->child_count;
initialize_entrylk_variables (frame, this);
- basename = int_lock->lk_basename;
- if (int_lock->lk_loc)
- loc = int_lock->lk_loc;
-
if (local->fd) {
- ret = fd_ctx_get (local->fd, this, &ctx);
-
- if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
+ fd_ctx = afr_fd_ctx_get (local->fd, this);
+ if (!fd_ctx) {
+ gf_log (this->name, GF_LOG_INFO,
"unable to get fd ctx for fd=%p",
local->fd);
@@ -1228,113 +1336,147 @@ afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this)
local->op_errno = EINVAL;
int_lock->lock_op_errno = EINVAL;
+ afr_unlock (frame, this);
return -1;
}
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
- call_count = internal_lock_count (frame, this, fd_ctx);
+ call_count = int_lock->lockee_count * internal_lock_count (frame, this);
int_lock->lk_call_count = call_count;
+ int_lock->lk_expected_count = call_count;
+
+ if (!call_count) {
+ gf_log (this->name, GF_LOG_INFO,
+ "fd not open on any subvolumes. aborting.");
+ afr_unlock (frame, this);
+ goto out;
+ }
/* Send non-blocking entrylk calls only on up children
and where the fd has been opened */
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i] && fd_ctx->opened_on[i]) {
- afr_trace_entrylk_in (frame, AFR_ENTRYLK_NB_TRANSACTION,
- AFR_LOCK_OP, basename, i);
+ for (i = 0; i < int_lock->lockee_count*priv->child_count; i++) {
+ index = i%copies;
+ lockee_no = i/copies;
+ if (local->child_up[index]) {
+ AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_NB_TRANSACTION,
+ AFR_LOCK_OP,
+ int_lock->lockee[lockee_no].basename,
+ i);
STACK_WIND_COOKIE (frame, afr_nonblocking_entrylk_cbk,
(void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->fentrylk,
+ priv->children[index],
+ priv->children[index]->fops->fentrylk,
this->name, local->fd,
- basename,
- ENTRYLK_LOCK_NB, ENTRYLK_WRLCK);
+ int_lock->lockee[lockee_no].basename,
+ ENTRYLK_LOCK_NB, ENTRYLK_WRLCK,
+ NULL);
+ if (!--call_count)
+ break;
}
}
} else {
- GF_ASSERT (loc);
-
- call_count = internal_lock_count (frame, this, NULL);
+ call_count = int_lock->lockee_count * internal_lock_count (frame, this);
int_lock->lk_call_count = call_count;
+ int_lock->lk_expected_count = call_count;
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- afr_trace_entrylk_in (frame, AFR_ENTRYLK_NB_TRANSACTION,
- AFR_LOCK_OP, basename, i);
+ for (i = 0; i < int_lock->lockee_count*priv->child_count; i++) {
+ index = i%copies;
+ lockee_no = i/copies;
+ if (local->child_up[index]) {
+ AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_NB_TRANSACTION,
+ AFR_LOCK_OP,
+ int_lock->lockee[lockee_no].basename,
+ i);
STACK_WIND_COOKIE (frame, afr_nonblocking_entrylk_cbk,
(void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->entrylk,
- this->name, loc, basename,
- ENTRYLK_LOCK, ENTRYLK_WRLCK);
+ priv->children[index],
+ priv->children[index]->fops->entrylk,
+ this->name, &int_lock->lockee[lockee_no].loc,
+ int_lock->lockee[lockee_no].basename,
+ ENTRYLK_LOCK_NB, ENTRYLK_WRLCK,
+ NULL);
if (!--call_count)
break;
-
}
}
}
-
+out:
return 0;
}
int32_t
afr_nonblocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
+ afr_inodelk_t *inodelk = NULL;
+ afr_local_t *local = NULL;
int call_count = 0;
- int child_index = (long) cookie;
+ int child_index = (long) cookie;
+ afr_fd_ctx_t *fd_ctx = NULL;
+
- local = frame->local;
+ local = frame->local;
int_lock = &local->internal_lock;
- priv = this->private;
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
- afr_trace_inodelk_out (frame, AFR_INODELK_NB_TRANSACTION,
+ AFR_TRACE_INODELK_OUT (frame, this, AFR_INODELK_NB_TRANSACTION,
AFR_LOCK_OP, NULL, op_ret,
op_errno, (long) cookie);
+ if (local->fd)
+ fd_ctx = afr_fd_ctx_get (local->fd, this);
+
LOCK (&frame->lock);
{
- call_count = --int_lock->lk_call_count;
- }
- UNLOCK (&frame->lock);
-
- if (op_ret < 0 ) {
+ if (op_ret < 0) {
if (op_errno == ENOSYS) {
/* return ENOTSUP */
gf_log (this->name, GF_LOG_ERROR,
"subvolume does not support locking. "
- "please load features/posix-locks xlator on server");
+ "please load features/locks xlator on "
+ "server");
local->op_ret = op_ret;
- int_lock->lock_op_ret = op_ret;
- local->child_up[child_index] = 0;
- int_lock->lock_op_errno = op_errno;
- local->op_errno = op_errno;
- }
- } else if (op_ret == 0) {
- int_lock->inode_locked_nodes[child_index]
- |= LOCKED_YES;
- int_lock->inodelk_lock_count++;
+ int_lock->lock_op_ret = op_ret;
+ int_lock->lock_op_errno = op_errno;
+ local->op_errno = op_errno;
+ }
+ if (local->transaction.eager_lock)
+ local->transaction.eager_lock[child_index] = 0;
+ } else {
+ inodelk->locked_nodes[child_index] |= LOCKED_YES;
+ inodelk->lock_count++;
+
+ if (local->transaction.eager_lock &&
+ local->transaction.eager_lock[child_index] &&
+ local->fd) {
+ /* piggybacked */
+ if (op_ret == 1) {
+ /* piggybacked */
+ } else if (op_ret == 0) {
+ /* lock acquired from server */
+ fd_ctx->lock_acquired[child_index]++;
+ }
+ }
+ }
+
+ call_count = --int_lock->lk_call_count;
}
+ UNLOCK (&frame->lock);
if (call_count == 0) {
gf_log (this->name, GF_LOG_TRACE,
"Last inode locking reply received");
- /* all locks successfull. Proceed to call FOP */
- if (int_lock->inodelk_lock_count ==
- afr_up_children_count (priv->child_count, local->child_up)) {
+ /* all locks successful. Proceed to call FOP */
+ if (inodelk->lock_count == int_lock->lk_expected_count) {
gf_log (this->name, GF_LOG_TRACE,
"All servers locked. Calling the cbk");
int_lock->lock_op_ret = 0;
int_lock->lock_cbk (frame, this);
}
- /* Not all locks were successfull. Unlock and try locking
+ /* Not all locks were successful. Unlock and try locking
again, this time with serially blocking locks */
else {
gf_log (this->name, GF_LOG_TRACE,
@@ -1352,31 +1494,36 @@ int
afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
{
afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
afr_fd_ctx_t *fd_ctx = NULL;
-
- int32_t call_count = 0;
- uint64_t ctx = 0;
- int i = 0;
- int ret = 0;
- struct flock flock;
+ int32_t call_count = 0;
+ int i = 0;
+ int ret = 0;
+ struct gf_flock flock = {0,};
+ struct gf_flock full_flock = {0,};
+ struct gf_flock *flock_use = NULL;
+ int piggyback = 0;
local = frame->local;
int_lock = &local->internal_lock;
priv = this->private;
- flock.l_start = int_lock->lk_flock.l_start;
- flock.l_len = int_lock->lk_flock.l_len;
- flock.l_type = int_lock->lk_flock.l_type;
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+
+ flock.l_start = inodelk->flock.l_start;
+ flock.l_len = inodelk->flock.l_len;
+ flock.l_type = inodelk->flock.l_type;
+
+ full_flock.l_type = inodelk->flock.l_type;
initialize_inodelk_variables (frame, this);
if (local->fd) {
- ret = fd_ctx_get (local->fd, this, &ctx);
-
- if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
+ fd_ctx = afr_fd_ctx_get (local->fd, this);
+ if (!fd_ctx) {
+ gf_log (this->name, GF_LOG_INFO,
"unable to get fd ctx for fd=%p",
local->fd);
@@ -1385,397 +1532,643 @@ afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
local->op_errno = EINVAL;
int_lock->lock_op_errno = EINVAL;
+ afr_unlock (frame, this);
ret = -1;
goto out;
}
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
- call_count = internal_lock_count (frame, this, fd_ctx);
+ call_count = internal_lock_count (frame, this);
int_lock->lk_call_count = call_count;
+ int_lock->lk_expected_count = call_count;
+
+ if (!call_count) {
+ gf_log (this->name, GF_LOG_INFO,
+ "fd not open on any subvolumes. aborting.");
+ afr_unlock (frame, this);
+ goto out;
+ }
/* Send non-blocking inodelk calls only on up children
and where the fd has been opened */
for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i] && fd_ctx->opened_on[i]) {
- afr_trace_inodelk_in (frame, AFR_INODELK_NB_TRANSACTION,
- AFR_LOCK_OP, &flock, F_SETLK, i);
+ if (!local->child_up[i])
+ continue;
- STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->finodelk,
- this->name, local->fd,
- F_SETLK, &flock);
+ flock_use = &flock;
+ if (!local->transaction.eager_lock_on) {
+ goto wind;
+ }
+
+ piggyback = 0;
+ local->transaction.eager_lock[i] = 1;
+
+ afr_set_delayed_post_op (frame, this);
+ LOCK (&local->fd->lock);
+ {
+ if (fd_ctx->lock_acquired[i]) {
+ fd_ctx->lock_piggyback[i]++;
+ piggyback = 1;
+ }
+ }
+ UNLOCK (&local->fd->lock);
+
+ if (piggyback) {
+ /* (op_ret == 1) => indicate piggybacked lock */
+ afr_nonblocking_inodelk_cbk (frame, (void *) (long) i,
+ this, 1, 0, NULL);
if (!--call_count)
break;
-
+ continue;
}
+ flock_use = &full_flock;
+ wind:
+ AFR_TRACE_INODELK_IN (frame, this,
+ AFR_INODELK_NB_TRANSACTION,
+ AFR_LOCK_OP, flock_use, F_SETLK, i);
+ STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->finodelk,
+ int_lock->domain, local->fd,
+ F_SETLK, flock_use, NULL);
+
+ if (!--call_count)
+ break;
}
} else {
- call_count = internal_lock_count (frame, this, NULL);
+ call_count = internal_lock_count (frame, this);
int_lock->lk_call_count = call_count;
+ int_lock->lk_expected_count = call_count;
for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- afr_trace_inodelk_in (frame, AFR_INODELK_NB_TRANSACTION,
- AFR_LOCK_OP, &flock, F_SETLK, i);
+ if (!local->child_up[i])
+ continue;
+ AFR_TRACE_INODELK_IN (frame, this,
+ AFR_INODELK_NB_TRANSACTION,
+ AFR_LOCK_OP, &flock, F_SETLK, i);
- STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->inodelk,
- this->name, &local->loc,
- F_SETLK, &flock);
-
- if (!--call_count)
- break;
+ STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->inodelk,
+ int_lock->domain, &local->loc,
+ F_SETLK, &flock, NULL);
- }
+ if (!--call_count)
+ break;
}
}
-
out:
return ret;
}
-static int
-__is_lower_locked (call_frame_t *frame, xlator_t *this)
+int32_t
+afr_unlock (call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- int count = 0;
- int i = 0;
+ local = frame->local;
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
+ if (transaction_lk_op (local)) {
+ if (is_afr_lock_transaction (local))
+ afr_unlock_inodelk (frame, this);
+ else
+ afr_unlock_entrylk (frame, this);
- for (i = 0; i < priv->child_count; i++) {
- if (int_lock->lower_locked_nodes[i] & LOCKED_LOWER)
- count++;
+ } else {
+ if (is_afr_lock_selfheal (local))
+ afr_unlock_inodelk (frame, this);
+ else
+ afr_unlock_entrylk (frame, this);
}
- return count;
-
+ return 0;
}
-static int
-__is_higher_locked (call_frame_t *frame, xlator_t *this)
+int
+afr_mark_locked_nodes (xlator_t *this, fd_t *fd,
+ unsigned char *locked_nodes)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_fd_ctx_t *fdctx = NULL;
+ uint64_t tmp = 0;
+ int ret = 0;
- int count = 0;
- int i = 0;
+ priv = this->private;
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
+ ret = afr_fd_ctx_set (this, fd);
+ if (ret)
+ goto out;
- for (i = 0; i < priv->child_count; i++) {
- if (int_lock->locked_nodes[i] & LOCKED_YES)
- count++;
+ ret = fd_ctx_get (fd, this, &tmp);
+ if (ret) {
+ gf_log (this->name, GF_LOG_INFO,
+ "failed to get the fd ctx");
+ goto out;
}
+ fdctx = (afr_fd_ctx_t *) (long) tmp;
+
+ GF_ASSERT (fdctx->locked_on);
- return count;
+ memcpy (fdctx->locked_on, locked_nodes,
+ priv->child_count);
+out:
+ return ret;
}
static int
-afr_unlock_lower_entrylk (call_frame_t *frame, xlator_t *this)
+__is_fd_saved (xlator_t *this, fd_t *fd)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- const char *basename = NULL;
- loc_t *loc = NULL;
+ afr_locked_fd_t *locked_fd = NULL;
+ afr_private_t *priv = NULL;
+ int found = 0;
- int call_count = 0;
- int i = -1;
+ priv = this->private;
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
+ list_for_each_entry (locked_fd, &priv->saved_fds, list) {
+ if (locked_fd->fd == fd) {
+ found = 1;
+ break;
+ }
+ }
- basename = int_lock->lk_basename;
- if (int_lock->lk_loc)
- loc = int_lock->lk_loc;
+ return found;
+}
- call_count = __is_lower_locked (frame, this);
- int_lock->lk_call_count = call_count;
+static int
+__afr_save_locked_fd (xlator_t *this, fd_t *fd)
+{
+ afr_private_t *priv = NULL;
+ afr_locked_fd_t *locked_fd = NULL;
+ int ret = 0;
- if (!call_count){
- gf_log (this->name, GF_LOG_TRACE,
- "No internal locks unlocked");
- int_lock->lock_cbk (frame, this);
+ priv = this->private;
+
+ locked_fd = GF_CALLOC (1, sizeof (*locked_fd),
+ gf_afr_mt_locked_fd);
+ if (!locked_fd) {
+ ret = -1;
goto out;
}
- for (i = 0; i < priv->child_count; i++) {
- if (int_lock->lower_locked_nodes[i] & LOCKED_LOWER) {
- afr_trace_entrylk_in (frame, AFR_ENTRYLK_NB_TRANSACTION,
- AFR_UNLOCK_OP, basename, i);
+ locked_fd->fd = fd;
+ INIT_LIST_HEAD (&locked_fd->list);
- STACK_WIND_COOKIE (frame, afr_unlock_entrylk_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->entrylk,
- this->name,
- loc, basename,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK);
+ list_add_tail (&locked_fd->list, &priv->saved_fds);
- if (!--call_count)
- break;
+out:
+ return ret;
+}
+
+int
+afr_save_locked_fd (xlator_t *this, fd_t *fd)
+{
+ afr_private_t *priv = NULL;
+ int ret = 0;
+ priv = this->private;
+
+ pthread_mutex_lock (&priv->mutex);
+ {
+ if (__is_fd_saved (this, fd)) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "fd=%p already saved", fd);
+ goto unlock;
}
- }
-out:
- return 0;
+ ret = __afr_save_locked_fd (this, fd);
+ if (ret) {
+ gf_log (this->name, GF_LOG_INFO,
+ "fd=%p could not be saved", fd);
+ goto unlock;
+ }
+ }
+unlock:
+ pthread_mutex_unlock (&priv->mutex);
+ return ret;
}
-
static int
-afr_post_unlock_higher_cbk (call_frame_t *frame, xlator_t *this)
+afr_lock_recovery_cleanup (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
+ afr_locked_fd_t *locked_fd = NULL;
local = frame->local;
- local->transaction.done (frame, this);
+ locked_fd = local->locked_fd;
+
+ STACK_DESTROY (frame->root);
+ afr_local_cleanup (local, this);
+
+ afr_save_locked_fd (this, locked_fd->fd);
+
return 0;
+
}
static int
-afr_post_unlock_lower_cbk (call_frame_t *frame, xlator_t *this)
+afr_get_source_lock_recovery (xlator_t *this, fd_t *fd)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- loc_t *lower = NULL;
- loc_t *higher = NULL;
- const char *lower_name = NULL;
- const char *higher_name = NULL;
-
- local = frame->local;
- int_lock = &local->internal_lock;
+ afr_fd_ctx_t *fdctx = NULL;
+ afr_private_t *priv = NULL;
+ uint64_t tmp = 0;
+ int i = 0;
+ int source_child = -1;
+ int ret = 0;
- lower = lower_path (&local->transaction.parent_loc,
- local->transaction.basename,
- &local->transaction.new_parent_loc,
- local->transaction.new_basename);
-
- lower_name = (lower == &local->transaction.parent_loc ?
- local->transaction.basename :
- local->transaction.new_basename);
+ priv = this->private;
- higher = (lower == &local->transaction.parent_loc ?
- &local->transaction.new_parent_loc :
- &local->transaction.parent_loc);
+ ret = fd_ctx_get (fd, this, &tmp);
+ if (ret)
+ goto out;
- higher_name = (higher == &local->transaction.parent_loc ?
- local->transaction.basename :
- local->transaction.new_basename);
+ fdctx = (afr_fd_ctx_t *) (long) tmp;
- if (__is_higher_locked (frame, this)) {
- gf_log (this->name, GF_LOG_DEBUG,
- "unlocking higher");
- int_lock->lk_basename = higher_name;
- int_lock->lk_loc = higher;
- int_lock->lock_cbk = afr_post_unlock_higher_cbk;
+ for (i = 0; i < priv->child_count; i++) {
+ if (fdctx->locked_on[i]) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "Found lock recovery source=%d", i);
+ source_child = i;
+ break;
+ }
+ }
- afr_unlock_entrylk (frame, this);
- } else
- local->transaction.done (frame, this);
+out:
+ return source_child;
- return 0;
}
-static int
-afr_rename_unlock (call_frame_t *frame, xlator_t *this)
+int32_t
+afr_get_locks_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+ dict_t *xdata);
+int32_t
+afr_recover_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+ dict_t *xdata)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- loc_t *lower = NULL;
- loc_t *higher = NULL;
- const char *lower_name = NULL;
- const char *higher_name = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int32_t source_child = 0;
+ struct gf_flock flock = {0,};
- local = frame->local;
- int_lock = &local->internal_lock;
+ local = frame->local;
+ priv = this->private;
- lower = lower_path (&local->transaction.parent_loc,
- local->transaction.basename,
- &local->transaction.new_parent_loc,
- local->transaction.new_basename);
+ if (op_ret) {
+ gf_log (this->name, GF_LOG_INFO,
+ "lock recovery failed");
+ goto cleanup;
+ }
- lower_name = (lower == &local->transaction.parent_loc ?
- local->transaction.basename :
- local->transaction.new_basename);
+ source_child = local->source_child;
- higher = (lower == &local->transaction.parent_loc ?
- &local->transaction.new_parent_loc :
- &local->transaction.parent_loc);
+ memcpy (&flock, lock, sizeof (*lock));
- higher_name = (higher == &local->transaction.parent_loc ?
- local->transaction.basename :
- local->transaction.new_basename);
+ STACK_WIND_COOKIE (frame, afr_get_locks_fd_cbk,
+ (void *) (long) source_child,
+ priv->children[source_child],
+ priv->children[source_child]->fops->lk,
+ local->fd, F_GETLK_FD, &flock, NULL);
+ return 0;
- if (__is_lower_locked (frame, this)) {
- gf_log (this->name, GF_LOG_DEBUG,
- "unlocking lower");
- int_lock->lk_basename = lower_name;
- int_lock->lk_loc = lower;
- int_lock->lock_cbk = afr_post_unlock_lower_cbk;
+cleanup:
+ afr_lock_recovery_cleanup (frame, this);
+ return 0;
+}
- afr_unlock_lower_entrylk (frame, this);
- } else
- afr_post_unlock_lower_cbk (frame, this);
+int
+afr_recover_lock (call_frame_t *frame, xlator_t *this,
+ struct gf_flock *flock)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int32_t lock_recovery_child = 0;
+
+ priv = this->private;
+ local = frame->local;
+
+ lock_recovery_child = local->lock_recovery_child;
+
+ frame->root->lk_owner = flock->l_owner;
+
+ STACK_WIND_COOKIE (frame, afr_recover_lock_cbk,
+ (void *) (long) lock_recovery_child,
+ priv->children[lock_recovery_child],
+ priv->children[lock_recovery_child]->fops->lk,
+ local->fd, F_SETLK, flock, NULL);
return 0;
}
static int
-afr_rename_transaction (call_frame_t *frame, xlator_t *this)
+is_afr_lock_eol (struct gf_flock *lock)
{
- afr_local_t *local = NULL;
-
- local = frame->local;
+ int ret = 0;
- return (local->transaction.type ==
- AFR_ENTRY_RENAME_TRANSACTION);
+ if ((lock->l_type == GF_LK_EOL))
+ ret = 1;
+ return ret;
}
int32_t
-afr_unlock (call_frame_t *frame, xlator_t *this)
+afr_get_locks_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
+ if (op_ret) {
+ gf_log (this->name, GF_LOG_INFO,
+ "Failed to get locks on fd");
+ goto cleanup;
+ }
- local = frame->local;
+ gf_log (this->name, GF_LOG_DEBUG,
+ "Got a lock on fd");
- if (transaction_lk_op (local)) {
- if (is_afr_lock_transaction (local))
- afr_unlock_inodelk (frame, this);
- else
- if (!afr_rename_transaction (frame, this))
- afr_unlock_entrylk (frame, this);
- else
- afr_rename_unlock (frame, this);
- } else {
- if (is_afr_lock_selfheal (local))
- afr_unlock_inodelk (frame, this);
- else
- afr_unlock_entrylk (frame, this);
+ if (is_afr_lock_eol (lock)) {
+ gf_log (this->name, GF_LOG_INFO,
+ "Reached EOL on locks on fd");
+ goto cleanup;
}
+ afr_recover_lock (frame, this, lock);
+
+ return 0;
+
+cleanup:
+ afr_lock_recovery_cleanup (frame, this);
+
return 0;
}
-int
-afr_mark_locked_nodes (xlator_t *this, fd_t *fd,
- unsigned char *locked_nodes)
+static int
+afr_lock_recovery (call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_fd_ctx_t *fdctx = NULL;
- uint64_t tmp = 0;
- int ret = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ fd_t *fd = NULL;
+ int ret = 0;
+ int32_t source_child = 0;
+ struct gf_flock flock = {0,};
- priv = this->private;
+ priv = this->private;
+ local = frame->local;
- afr_fd_ctx_set (this, fd);
- if (ret < 0)
+ fd = local->fd;
+
+ source_child = afr_get_source_lock_recovery (this, fd);
+ if (source_child < 0) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "Could not recover locks due to lock "
+ "split brain");
+ ret = -1;
goto out;
+ }
+
+ local->source_child = source_child;
+
+ /* the flock can be zero filled as we're querying incrementally
+ the locks held on the fd.
+ */
+ STACK_WIND_COOKIE (frame, afr_get_locks_fd_cbk,
+ (void *) (long) source_child,
+ priv->children[source_child],
+ priv->children[source_child]->fops->lk,
+ local->fd, F_GETLK_FD, &flock, NULL);
+
+out:
+ return ret;
+}
+
+
+static int
+afr_mark_fd_opened (xlator_t *this, fd_t *fd, int32_t child_index)
+{
+ afr_fd_ctx_t *fdctx = NULL;
+ uint64_t tmp = 0;
+ int ret = 0;
ret = fd_ctx_get (fd, this, &tmp);
- fdctx = (afr_fd_ctx_t *) (long) tmp;
+ if (ret)
+ goto out;
- GF_ASSERT (fdctx->locked_on);
+ fdctx = (afr_fd_ctx_t *) (long) tmp;
- memcpy (fdctx->locked_on, locked_nodes,
- priv->child_count);
+ fdctx->opened_on[child_index] = AFR_FD_OPENED;
out:
return ret;
}
-static int
-__is_fd_saved (xlator_t *this, fd_t *fd)
+int32_t
+afr_lock_recovery_preopen_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd,
+ dict_t *xdata)
{
- afr_locked_fd_t *locked_fd = NULL;
- afr_private_t *priv = NULL;
- int found = 0;
+ int32_t child_index = (long )cookie;
+ int ret = 0;
- priv = this->private;
+ if (op_ret) {
+ gf_log (this->name, GF_LOG_INFO,
+ "Reopen during lock-recovery failed");
+ goto cleanup;
+ }
- list_for_each_entry (locked_fd, &priv->saved_fds, list) {
- if (locked_fd->fd == fd) {
- found = 1;
- break;
- }
+ gf_log (this->name, GF_LOG_DEBUG,
+ "Open succeeded => proceed to recover locks");
+
+ ret = afr_lock_recovery (frame, this);
+ if (ret) {
+ gf_log (this->name, GF_LOG_INFO,
+ "Lock recovery failed");
+ goto cleanup;
}
- return found;
+ ret = afr_mark_fd_opened (this, fd, child_index);
+ if (ret) {
+ gf_log (this->name, GF_LOG_INFO,
+ "Marking fd open failed");
+ goto cleanup;
+ }
+
+ return 0;
+
+cleanup:
+ afr_lock_recovery_cleanup (frame, this);
+ return 0;
}
static int
-__afr_save_locked_fd (xlator_t *this, fd_t *fd)
+afr_lock_recovery_preopen (call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_locked_fd_t *locked_fd = NULL;
- int ret = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ uint64_t tmp = 0;
+ afr_fd_ctx_t *fdctx = NULL;
+ loc_t loc = {0,};
+ int32_t child_index = 0;
+ int ret = 0;
- priv = this->private;
+ priv = this->private;
+ local = frame->local;
- locked_fd = GF_CALLOC (1, sizeof (*locked_fd),
- gf_afr_mt_locked_fd);
- if (!locked_fd) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- ret = -1;
+ GF_ASSERT (local && local->fd);
+
+ ret = fd_ctx_get (local->fd, this, &tmp);
+ if (ret)
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: failed to get the context of fd",
+ uuid_utoa (local->fd->inode->gfid));
+ fdctx = (afr_fd_ctx_t *) (long) tmp;
+ /* TODO: instead we should return from the function */
+ GF_ASSERT (fdctx);
+
+ child_index = local->lock_recovery_child;
+
+ inode_path (local->fd->inode, NULL, (char **)&loc.path);
+ loc.name = strrchr (loc.path, '/');
+ loc.inode = inode_ref (local->fd->inode);
+ loc.parent = inode_parent (local->fd->inode, 0, NULL);
+
+
+ STACK_WIND_COOKIE (frame, afr_lock_recovery_preopen_cbk,
+ (void *)(long) child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->open,
+ &loc, fdctx->flags, local->fd, NULL);
+
+ return 0;
+}
+
+static int
+is_fd_opened (fd_t *fd, int32_t child_index)
+{
+ afr_fd_ctx_t *fdctx = NULL;
+ uint64_t tmp = 0;
+ int ret = 0;
+
+ ret = fd_ctx_get (fd, THIS, &tmp);
+ if (ret)
goto out;
- }
- INIT_LIST_HEAD (&locked_fd->list);
+ fdctx = (afr_fd_ctx_t *) (long) tmp;
- list_add_tail (&locked_fd->list, &priv->saved_fds);
+ if (fdctx->opened_on[child_index] == AFR_FD_OPENED)
+ ret = 1;
out:
return ret;
}
int
-afr_save_locked_fd (xlator_t *this, fd_t *fd)
+afr_attempt_lock_recovery (xlator_t *this, int32_t child_index)
{
- afr_private_t *priv = NULL;
- int ret = 0;
+ call_frame_t *frame = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_locked_fd_t *locked_fd = NULL;
+ afr_locked_fd_t *tmp = NULL;
+ int ret = -1;
+ struct list_head locks_list = {0,};
+ int32_t op_errno = 0;
+
priv = this->private;
+ if (list_empty (&priv->saved_fds))
+ goto out;
+
+ frame = create_frame (this, this->ctx->pool);
+ if (!frame) {
+ ret = -1;
+ goto out;
+ }
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0) {
+ ret = -1;
+ goto out;
+ }
+
+ frame->local = local;
+
+ INIT_LIST_HEAD (&locks_list);
+
pthread_mutex_lock (&priv->mutex);
{
- if (__is_fd_saved (this, fd)) {
- gf_log (this->name, GF_LOG_DEBUG,
- "fd=%p already saved", fd);
- goto unlock;
- }
+ list_splice_init (&priv->saved_fds, &locks_list);
+ }
+ pthread_mutex_unlock (&priv->mutex);
- ret = __afr_save_locked_fd (this, fd);
- if (ret) {
+ list_for_each_entry_safe (locked_fd, tmp,
+ &locks_list, list) {
+
+ list_del_init (&locked_fd->list);
+
+ local->fd = fd_ref (locked_fd->fd);
+ local->lock_recovery_child = child_index;
+ local->locked_fd = locked_fd;
+
+ if (!is_fd_opened (locked_fd->fd, child_index)) {
gf_log (this->name, GF_LOG_DEBUG,
- "fd=%p could not be saved");
- goto unlock;
+ "attempting open before lock "
+ "recovery");
+ afr_lock_recovery_preopen (frame, this);
+ } else {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "attempting lock recovery "
+ "without a preopen");
+ afr_lock_recovery (frame, this);
}
}
-unlock:
- pthread_mutex_unlock (&priv->mutex);
+out:
+ if ((ret < 0) && frame)
+ AFR_STACK_DESTROY (frame);
+ return ret;
+}
+
+int
+afr_lk_transfer_datalock (call_frame_t *dst, call_frame_t *src, char *dom,
+ unsigned int child_count)
+{
+ afr_local_t *dst_local = NULL;
+ afr_local_t *src_local = NULL;
+ afr_internal_lock_t *dst_lock = NULL;
+ afr_internal_lock_t *src_lock = NULL;
+ afr_inodelk_t *dst_inodelk = NULL;
+ afr_inodelk_t *src_inodelk = NULL;
+ int ret = -1;
+
+ src_local = src->local;
+ src_lock = &src_local->internal_lock;
+ src_inodelk = afr_get_inodelk (src_lock, dom);
+ dst_local = dst->local;
+ dst_lock = &dst_local->internal_lock;
+ dst_inodelk = afr_get_inodelk (dst_lock, dom);
+ if (!dst_inodelk || !src_inodelk)
+ goto out;
+ if (src_inodelk->locked_nodes) {
+ memcpy (dst_inodelk->locked_nodes, src_inodelk->locked_nodes,
+ sizeof (*dst_inodelk->locked_nodes) * child_count);
+ memset (src_inodelk->locked_nodes, 0,
+ sizeof (*src_inodelk->locked_nodes) * child_count);
+ }
+
+ dst_lock->transaction_lk_type = src_lock->transaction_lk_type;
+ dst_lock->selfheal_lk_type = src_lock->selfheal_lk_type;
+ dst_inodelk->lock_count = src_inodelk->lock_count;
+ src_inodelk->lock_count = 0;
+ ret = 0;
+out:
return ret;
}
diff --git a/xlators/cluster/afr/src/afr-mem-types.h b/xlators/cluster/afr/src/afr-mem-types.h
index 31a80e8f5..73594f265 100644
--- a/xlators/cluster/afr/src/afr-mem-types.h
+++ b/xlators/cluster/afr/src/afr-mem-types.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -26,7 +17,6 @@
enum gf_afr_mem_types_ {
gf_afr_mt_iovec = gf_common_mt_end + 1,
gf_afr_mt_afr_fd_ctx_t,
- gf_afr_mt_afr_local_t,
gf_afr_mt_afr_private_t,
gf_afr_mt_int32_t,
gf_afr_mt_char,
@@ -42,6 +32,19 @@ enum gf_afr_mem_types_ {
gf_afr_mt_entry_name,
gf_afr_mt_pump_priv,
gf_afr_mt_locked_fd,
+ gf_afr_mt_inode_ctx_t,
+ gf_afr_fd_paused_call_t,
+ gf_afr_mt_crawl_data_t,
+ gf_afr_mt_brick_pos_t,
+ gf_afr_mt_shd_bool_t,
+ gf_afr_mt_shd_timer_t,
+ gf_afr_mt_shd_event_t,
+ gf_afr_mt_time_t,
+ gf_afr_mt_pos_data_t,
+ gf_afr_mt_reply_t,
+ gf_afr_mt_stats_t,
+ gf_afr_mt_shd_crawl_event_t,
+ gf_afr_mt_uint64_t,
gf_afr_mt_end
};
#endif
diff --git a/xlators/cluster/afr/src/afr-open.c b/xlators/cluster/afr/src/afr-open.c
index 1360dfa89..643a5d692 100644
--- a/xlators/cluster/afr/src/afr-open.c
+++ b/xlators/cluster/afr/src/afr-open.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include <libgen.h>
@@ -55,419 +46,337 @@
#include "afr-self-heal.h"
#include "afr-self-heal-common.h"
-
int
-afr_open_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
+afr_stale_child_up (afr_local_t *local, xlator_t *this)
{
- afr_local_t * local = frame->local;
-
- AFR_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
- local->fd);
- return 0;
-}
-
-
-int
-afr_open_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- fd_t *fd)
-{
- afr_local_t * local = NULL;
-
- int child_index = (long) cookie;
-
- uint64_t ctx;
- afr_fd_ctx_t *fd_ctx;
-
- int ret = 0;
-
- int call_count = -1;
-
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- }
-
- if (op_ret >= 0) {
- local->op_ret = op_ret;
- local->success_count++;
-
- ret = afr_fd_ctx_set (this, fd);
-
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR,
- "could not set fd ctx for fd=%p",
- fd);
-
- local->op_ret = -1;
- local->op_errno = -ret;
- goto unlock;
- }
-
- ret = fd_ctx_get (fd, this, &ctx);
+ int i = 0;
+ afr_private_t *priv = NULL;
+ int up = -1;
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR,
- "could not get fd ctx for fd=%p", fd);
- local->op_ret = -1;
- local->op_errno = -ret;
- goto unlock;
- }
+ priv = this->private;
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
- fd_ctx->opened_on[child_index] = 1;
- fd_ctx->flags = local->cont.open.flags;
- fd_ctx->wbflags = local->cont.open.wbflags;
- }
- }
-unlock:
- UNLOCK (&frame->lock);
-
- call_count = afr_frame_return (frame);
+ if (!local->fresh_children)
+ local->fresh_children = afr_children_create (priv->child_count);
+ if (!local->fresh_children)
+ goto out;
- if (call_count == 0) {
- if ((local->cont.open.flags & O_TRUNC)
- && (local->op_ret >= 0)) {
- STACK_WIND (frame, afr_open_ftruncate_cbk,
- this, this->fops->ftruncate,
- fd, 0);
- } else {
- AFR_STACK_UNWIND (open, frame, local->op_ret,
- local->op_errno, local->fd);
- }
- }
+ afr_inode_get_read_ctx (this, local->fd->inode, local->fresh_children);
+ if (priv->child_count == afr_get_children_count (local->fresh_children,
+ priv->child_count))
+ goto out;
- return 0;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->child_up[i])
+ continue;
+ if (afr_is_child_present (local->fresh_children,
+ priv->child_count, i))
+ continue;
+ up = i;
+ break;
+ }
+out:
+ return up;
}
-
-int
-afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- fd_t *fd, int32_t wbflags)
+void
+afr_perform_data_self_heal (call_frame_t *frame, xlator_t *this)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
-
- int i = 0;
- int ret = -1;
-
- int32_t call_count = 0;
- int32_t op_ret = -1;
- int32_t op_errno = 0;
- int32_t wind_flags = flags & (~O_TRUNC);
-
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
- VALIDATE_OR_GOTO (loc, out);
-
- priv = this->private;
-
- if (afr_is_split_brain (this, loc->inode)) {
- /* self-heal failed */
- op_errno = EIO;
- goto out;
- }
-
- ALLOC_OR_GOTO (local, afr_local_t, out);
-
- ret = AFR_LOCAL_INIT (local, priv);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- frame->local = local;
- call_count = local->call_count;
-
- loc_copy (&local->loc, loc);
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ inode_t *inode = NULL;
+ int st_child = -1;
+ char reason[64] = {0};
- local->cont.open.flags = flags;
- local->cont.open.wbflags = wbflags;
+ local = frame->local;
+ sh = &local->self_heal;
+ inode = local->fd->inode;
- local->fd = fd_ref (fd);
+ if (!IA_ISREG (inode->ia_type))
+ goto out;
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_open_cbk, (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->open,
- loc, wind_flags, fd, wbflags);
+ st_child = afr_stale_child_up (local, this);
+ if (st_child < 0)
+ goto out;
- if (!--call_count)
- break;
- }
- }
+ sh->do_data_self_heal = _gf_true;
+ sh->do_metadata_self_heal = _gf_true;
+ sh->do_gfid_self_heal = _gf_true;
+ sh->do_missing_entry_self_heal = _gf_true;
- op_ret = 0;
+ snprintf (reason, sizeof (reason), "stale subvolume %d detected",
+ st_child);
+ afr_launch_self_heal (frame, this, inode, _gf_true, inode->ia_type,
+ reason, NULL, NULL);
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (open, frame, op_ret, op_errno, fd);
- }
-
- return 0;
+ return;
}
-
int
-afr_openfd_sh_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd)
+afr_open_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t * local = frame->local;
+ afr_private_t *priv = NULL;
- int ret = 0;
+ priv = this->private;
+ if (afr_open_only_data_self_heal (priv->data_self_heal))
+ afr_perform_data_self_heal (frame, this);
+ AFR_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
+ local->fd, xdata);
+ return 0;
+}
- uint64_t ctx;
- afr_fd_ctx_t *fd_ctx;
- int call_count = 0;
- int child_index = (long) cookie;
+int
+afr_open_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ fd_t *fd, dict_t *xdata)
+{
+ afr_local_t * local = NULL;
+ int ret = 0;
+ int call_count = -1;
+ int child_index = (long) cookie;
+ afr_private_t *priv = NULL;
- priv = this->private;
- local = frame->local;
- int_lock = &local->internal_lock;
+ priv = this->private;
+ local = frame->local;
LOCK (&frame->lock);
{
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ }
+
if (op_ret >= 0) {
- ret = fd_ctx_get (fd, this, &ctx);
+ local->op_ret = op_ret;
+ local->success_count++;
- if (ret < 0) {
- goto out;
+ ret = afr_child_fd_ctx_set (this, fd, child_index,
+ local->cont.open.flags);
+ if (ret) {
+ local->op_ret = -1;
+ local->op_errno = -ret;
+ goto unlock;
}
-
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
- fd_ctx->opened_on[child_index] = 1;
-
- gf_log (this->name, GF_LOG_TRACE,
- "fd for %s opened successfully on subvolume %s",
- local->loc.path, priv->children[child_index]->name);
}
}
-out:
+unlock:
UNLOCK (&frame->lock);
call_count = afr_frame_return (frame);
if (call_count == 0) {
- int_lock->lock_cbk = local->transaction.done;
- local->transaction.resume (frame, this);
+ if ((local->cont.open.flags & O_TRUNC)
+ && (local->op_ret >= 0)) {
+ STACK_WIND (frame, afr_open_ftruncate_cbk,
+ this, this->fops->ftruncate,
+ fd, 0, NULL);
+ } else {
+ if (afr_open_only_data_self_heal (priv->data_self_heal))
+ afr_perform_data_self_heal (frame, this);
+ AFR_STACK_UNWIND (open, frame, local->op_ret,
+ local->op_errno, local->fd, xdata);
+ }
}
return 0;
}
-
-static int
-__unopened_count (int child_count, unsigned int *opened_on, unsigned char *child_up)
-{
- int i;
- int count = 0;
-
- for (i = 0; i < child_count; i++) {
- if (!opened_on[i] && child_up[i])
- count++;
- }
-
- return count;
-}
-
-
int
-afr_openfd_sh_unwind (call_frame_t *frame, xlator_t *this)
+afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- uint64_t ctx;
- afr_fd_ctx_t *fd_ctx;
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ int i = 0;
+ int ret = -1;
+ int32_t call_count = 0;
+ int32_t op_errno = 0;
+ int32_t wind_flags = flags & (~O_TRUNC);
+ //We can't let truncation to happen outside transaction.
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
+ VALIDATE_OR_GOTO (loc, out);
+
+ priv = this->private;
+
+ if (flags & (O_CREAT|O_TRUNC)) {
+ QUORUM_CHECK(open,out);
+ }
- int abandon = 0;
- int ret = 0;
- int i;
- int call_count = 0;
+ if (afr_is_split_brain (this, loc->inode)) {
+ /* self-heal failed */
+ gf_log (this->name, GF_LOG_WARNING,
+ "failed to open as split brain seen, returning EIO");
+ op_errno = EIO;
+ goto out;
+ }
- priv = this->private;
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
- /*
- * Some subvolumes might have come up on which we never
- * opened this fd in the first place. Re-open fd's on those
- * subvolumes now.
- */
-
- ret = fd_ctx_get (local->fd, this, &ctx);
-
- if (ret < 0) {
- abandon = 1;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
goto out;
- }
-
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
- LOCK (&local->fd->lock);
- {
- call_count = __unopened_count (priv->child_count,
- fd_ctx->opened_on,
- local->child_up);
- for (i = 0; i < priv->child_count; i++) {
- fd_ctx->pre_op_done[i] = 0;
- fd_ctx->pre_op_piggyback[i] = 0;
- }
- }
- UNLOCK (&local->fd->lock);
+ call_count = local->call_count;
+ loc_copy (&local->loc, loc);
- if (call_count == 0) {
- abandon = 1;
- goto out;
- }
+ local->cont.open.flags = flags;
- local->call_count = call_count;
+ local->fd = fd_ref (fd);
for (i = 0; i < priv->child_count; i++) {
- if (!fd_ctx->opened_on[i] && local->child_up[i]) {
- gf_log (this->name, GF_LOG_TRACE,
- "opening fd for %s on subvolume %s",
- local->loc.path, priv->children[i]->name);
-
- STACK_WIND_COOKIE (frame, afr_openfd_sh_open_cbk,
- (void *)(long) i,
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE (frame, afr_open_cbk, (void *) (long) i,
priv->children[i],
priv->children[i]->fops->open,
- &local->loc, fd_ctx->flags, local->fd,
- fd_ctx->wbflags);
+ loc, wind_flags, fd, xdata);
if (!--call_count)
break;
}
}
+ ret = 0;
out:
- if (abandon)
- local->transaction.resume (frame, this);
+ if (ret < 0)
+ AFR_STACK_UNWIND (open, frame, -1, op_errno, fd, xdata);
return 0;
}
-
int
-afr_openfd_sh (call_frame_t *frame, xlator_t *this)
+afr_openfd_fix_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- char sh_type_str[256] = {0,};
-
- local = frame->local;
- sh = &local->self_heal;
-
- inode_path (local->fd->inode, NULL, (char **)&local->loc.path);
- local->loc.name = strrchr (local->loc.path, '/');
- local->loc.inode = inode_ref (local->fd->inode);
- local->loc.parent = inode_parent (local->fd->inode, 0, NULL);
-
- /* forcibly trigger missing-entries self-heal */
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int call_count = 0;
+ int child_index = (long) cookie;
- local->success_count = 1;
- local->enoent_count = 1;
-
- sh->data_lock_held = _gf_true;
- sh->need_data_self_heal = _gf_true;
- sh->type = local->fd->inode->ia_type;
- sh->background = _gf_false;
- sh->unwind = afr_openfd_sh_unwind;
-
- afr_self_heal_type_str_get(&local->self_heal,
- sh_type_str,
- sizeof(sh_type_str));
- gf_log (this->name, GF_LOG_NORMAL, "%s self-heal triggered. "
- "path: %s, reason: Replicate up down flush, data lock is held",
- sh_type_str, local->loc.path);
-
- afr_self_heal (frame, this);
-
- return 0;
-}
-
-
-int
-afr_openfd_flush_done (call_frame_t *frame, xlator_t *this)
-{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
-
- uint64_t ctx;
- afr_fd_ctx_t * fd_ctx = NULL;
+ priv = this->private;
+ local = frame->local;
- int _ret = -1;
+ if (op_ret >= 0) {
+ gf_log (this->name, GF_LOG_DEBUG, "fd for %s opened "
+ "successfully on subvolume %s", local->loc.path,
+ priv->children[child_index]->name);
+ } else {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to open %s "
+ "on subvolume %s", local->loc.path,
+ priv->children[child_index]->name);
+ }
- priv = this->private;
- local = frame->local;
+ fd_ctx = afr_fd_ctx_get (local->fd, this);
+ if (!fd_ctx) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "failed to get fd context, %p", local->fd);
+ goto out;
+ }
LOCK (&local->fd->lock);
{
- _ret = __fd_ctx_get (local->fd, this, &ctx);
-
- if (_ret < 0) {
- goto out;
+ if (op_ret >= 0) {
+ fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
+ } else {
+ fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
}
-
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
- fd_ctx->down_count = priv->down_count;
- fd_ctx->up_count = priv->up_count;
}
-out:
UNLOCK (&local->fd->lock);
+out:
+ call_count = afr_frame_return (frame);
+ if (call_count == 0)
+ AFR_STACK_DESTROY (frame);
- afr_local_transaction_cleanup (local, this);
-
- gf_log (this->name, GF_LOG_TRACE,
- "The up/down flush is over");
+ return 0;
+}
- fd_unref (local->fd);
- local->openfd_flush_cbk (frame, this);
+void
+afr_fix_open (xlator_t *this, fd_t *fd, size_t need_open_count, int *need_open)
+{
+ afr_private_t *priv = NULL;
+ int i = 0;
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
+ int ret = -1;
+ int32_t op_errno = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
- return 0;
-}
+ priv = this->private;
+ if (!afr_is_fd_fixable (fd) || !need_open || !need_open_count)
+ goto out;
-int
-afr_openfd_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
-{
- afr_local_t * local = NULL;
+ fd_ctx = afr_fd_ctx_get (fd, this);
+ if (!fd_ctx) {
+ ret = -1;
+ goto out;
+ }
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
+ frame = create_frame (this, this->ctx->pool);
+ if (!frame) {
+ ret = -1;
+ goto out;
+ }
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
- local->op = GF_FOP_FLUSH;
+ local->loc.inode = inode_ref (fd->inode);
+ ret = loc_path (&local->loc, NULL);
+ if (ret < 0)
+ goto out;
local->fd = fd_ref (fd);
+ local->call_count = need_open_count;
+
+ gf_log (this->name, GF_LOG_DEBUG, "need open count: %zd",
+ need_open_count);
- local->transaction.fop = afr_openfd_sh;
- local->transaction.done = afr_openfd_flush_done;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!need_open[i])
+ continue;
- local->transaction.start = 0;
- local->transaction.len = 0;
+ if (IA_IFDIR == fd->inode->ia_type) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "opening fd for dir %s on subvolume %s",
+ local->loc.path, priv->children[i]->name);
- gf_log (this->name, GF_LOG_TRACE,
- "doing up/down flush on fd=%p",
- fd);
+ STACK_WIND_COOKIE (frame, afr_openfd_fix_open_cbk,
+ (void*) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->opendir,
+ &local->loc, local->fd,
+ NULL);
+ } else {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "opening fd for file %s on subvolume %s",
+ local->loc.path, priv->children[i]->name);
- afr_transaction (frame, this, AFR_DATA_TRANSACTION);
+ STACK_WIND_COOKIE (frame, afr_openfd_fix_open_cbk,
+ (void *)(long) i,
+ priv->children[i],
+ priv->children[i]->fops->open,
+ &local->loc,
+ fd_ctx->flags & (~O_TRUNC),
+ local->fd, NULL);
+ }
+ }
+ op_errno = 0;
+ ret = 0;
out:
- return 0;
+ if (op_errno)
+ ret = -1; //For handling ALLOC_OR_GOTO
+ if (ret && frame)
+ AFR_STACK_DESTROY (frame);
}
-
diff --git a/xlators/cluster/afr/src/afr-self-heal-algorithm.c b/xlators/cluster/afr/src/afr-self-heal-algorithm.c
index 0f7570dc2..83846f152 100644
--- a/xlators/cluster/afr/src/afr-self-heal-algorithm.c
+++ b/xlators/cluster/afr/src/afr-self-heal-algorithm.c
@@ -1,23 +1,15 @@
/*
- Copyright (c) 2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
+#include <openssl/md5.h>
#include "glusterfs.h"
#include "afr.h"
#include "xlator.h"
@@ -33,7 +25,6 @@
#include "compat-errno.h"
#include "compat.h"
#include "byte-order.h"
-#include "md5.h"
#include "afr-transaction.h"
#include "afr-self-heal.h"
@@ -44,751 +35,624 @@
This file contains the various self-heal algorithms
*/
+static int
+sh_loop_driver (call_frame_t *sh_frame, xlator_t *this,
+ gf_boolean_t is_first_call, call_frame_t *old_loop_frame);
+static int
+sh_loop_return (call_frame_t *sh_frame, xlator_t *this, call_frame_t *loop_frame,
+ int32_t op_ret, int32_t op_errno);
+static int
+sh_destroy_frame (call_frame_t *frame, xlator_t *this)
+{
+ if (!frame)
+ goto out;
-/*
- The "full" algorithm. Copies the entire file from
- source to sinks.
-*/
-
+ AFR_STACK_DESTROY (frame);
+out:
+ return 0;
+}
static void
-sh_full_private_cleanup (call_frame_t *frame, xlator_t *this)
+sh_private_cleanup (call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- afr_self_heal_t * sh = NULL;
- afr_sh_algo_full_private_t *sh_priv = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_sh_algo_private_t *sh_priv = NULL;
local = frame->local;
sh = &local->self_heal;
sh_priv = sh->private;
-
- if (sh_priv)
- GF_FREE (sh_priv);
+ GF_FREE (sh_priv);
}
-
-static int
-sh_full_loop_driver (call_frame_t *frame, xlator_t *this);
-
static int
-sh_full_loop_return (call_frame_t *rw_frame, xlator_t *this, off_t offset)
+sh_number_of_writes_needed (unsigned char *write_needed, int child_count)
{
- afr_local_t * rw_local = NULL;
- afr_self_heal_t * rw_sh = NULL;
-
- call_frame_t *sh_frame = NULL;
- afr_local_t * sh_local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_sh_algo_full_private_t *sh_priv = NULL;
-
- rw_local = rw_frame->local;
- rw_sh = &rw_local->self_heal;
-
- sh_frame = rw_sh->sh_frame;
- sh_local = sh_frame->local;
- sh = &sh_local->self_heal;
- sh_priv = sh->private;
+ int writes = 0;
+ int i = 0;
- LOCK (&sh_priv->lock);
- {
- sh_priv->loops_running--;
+ for (i = 0; i < child_count; i++) {
+ if (write_needed[i])
+ writes++;
}
- UNLOCK (&sh_priv->lock);
-
- gf_log (this->name, GF_LOG_TRACE,
- "loop for offset %"PRId64" returned", offset);
- AFR_STACK_DESTROY (rw_frame);
-
- sh_full_loop_driver (sh_frame, this);
-
- return 0;
+ return writes;
}
static int
-sh_full_write_cbk (call_frame_t *rw_frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
+sh_loop_driver_done (call_frame_t *sh_frame, xlator_t *this,
+ call_frame_t *last_loop_frame)
{
- afr_private_t * priv = NULL;
- afr_local_t * rw_local = NULL;
- afr_self_heal_t *rw_sh = NULL;
-
- call_frame_t *sh_frame = NULL;
- afr_local_t * sh_local = NULL;
- afr_self_heal_t *sh = NULL;
-
- int child_index = (long) cookie;
- int call_count = 0;
-
- priv = this->private;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_sh_algo_private_t *sh_priv = NULL;
+ int32_t total_blocks = 0;
+ int32_t diff_blocks = 0;
+
+ local = sh_frame->local;
+ sh = &local->self_heal;
+ sh_priv = sh->private;
+ if (sh_priv) {
+ total_blocks = sh_priv->total_blocks;
+ diff_blocks = sh_priv->diff_blocks;
+ }
- rw_local = rw_frame->local;
- rw_sh = &rw_local->self_heal;
+ sh_private_cleanup (sh_frame, this);
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+ GF_ASSERT (!last_loop_frame);
+ //loop_finish should have happened and the old_loop should be NULL
+ gf_log (this->name, GF_LOG_DEBUG,
+ "self-heal aborting on %s",
+ local->loc.path);
- sh_frame = rw_sh->sh_frame;
- sh_local = sh_frame->local;
- sh = &sh_local->self_heal;
+ local->self_heal.algo_abort_cbk (sh_frame, this);
+ } else {
+ GF_ASSERT (last_loop_frame);
+ if (diff_blocks == total_blocks) {
+ gf_log (this->name, GF_LOG_DEBUG, "full self-heal "
+ "completed on %s",local->loc.path);
+ } else {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "diff self-heal on %s: completed. "
+ "(%d blocks of %d were different (%.2f%%))",
+ local->loc.path, diff_blocks, total_blocks,
+ ((diff_blocks * 1.0)/total_blocks) * 100);
+ }
- gf_log (this->name, GF_LOG_TRACE,
- "wrote %d bytes of data from %s to child %d, offset %"PRId64"",
- op_ret, sh_local->loc.path, child_index,
- rw_sh->offset - op_ret);
+ sh->old_loop_frame = last_loop_frame;
+ local->self_heal.algo_completion_cbk (sh_frame, this);
+ }
- LOCK (&sh_frame->lock);
- {
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "write to %s failed on subvolume %s (%s)",
- sh_local->loc.path,
- priv->children[child_index]->name,
- strerror (op_errno));
+ return 0;
+}
- sh->op_failed = 1;
- }
- }
- UNLOCK (&sh_frame->lock);
+int
+sh_loop_finish (call_frame_t *loop_frame, xlator_t *this)
+{
+ afr_local_t *loop_local = NULL;
+ afr_self_heal_t *loop_sh = NULL;
- call_count = afr_frame_return (rw_frame);
+ if (!loop_frame)
+ goto out;
- if (call_count == 0) {
- sh_full_loop_return (rw_frame, this, rw_sh->offset - op_ret);
- }
+ loop_local = loop_frame->local;
+ if (loop_local) {
+ loop_sh = &loop_local->self_heal;
+ }
- return 0;
+ if (loop_sh && loop_sh->data_lock_held) {
+ afr_sh_data_unlock (loop_frame, this, this->name,
+ sh_destroy_frame);
+ } else {
+ sh_destroy_frame (loop_frame, this);
+ }
+out:
+ return 0;
}
-
static int
-sh_full_read_cbk (call_frame_t *rw_frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct iovec *vector, int32_t count, struct iatt *buf,
- struct iobref *iobref)
+sh_loop_lock_success (call_frame_t *loop_frame, xlator_t *this)
{
- afr_private_t * priv = NULL;
- afr_local_t * rw_local = NULL;
- afr_self_heal_t *rw_sh = NULL;
-
- call_frame_t *sh_frame = NULL;
- afr_local_t * sh_local = NULL;
- afr_self_heal_t *sh = NULL;
-
- int i = 0;
- int call_count = 0;
-
- off_t offset = (long) cookie;
-
- priv = this->private;
- rw_local = rw_frame->local;
- rw_sh = &rw_local->self_heal;
-
- sh_frame = rw_sh->sh_frame;
- sh_local = sh_frame->local;
- sh = &sh_local->self_heal;
-
- call_count = sh->active_sinks;
-
- rw_local->call_count = call_count;
-
- gf_log (this->name, GF_LOG_TRACE,
- "read %d bytes of data from %s, offset %"PRId64"",
- op_ret, sh_local->loc.path, offset);
-
- if (op_ret <= 0) {
- sh->op_failed = 1;
-
- sh_full_loop_return (rw_frame, this, offset);
- return 0;
- }
+ afr_local_t *loop_local = NULL;
+ afr_self_heal_t *loop_sh = NULL;
- rw_sh->offset += op_ret;
+ loop_local = loop_frame->local;
+ loop_sh = &loop_local->self_heal;
- if (sh->file_has_holes) {
- if (iov_0filled (vector, count) == 0) {
- /* the iter function depends on the
- sh->offset already being updated
- above
- */
+ sh_loop_finish (loop_sh->old_loop_frame, this);
+ loop_sh->old_loop_frame = NULL;
- sh_full_loop_return (rw_frame, this, offset);
- goto out;
- }
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (sh->sources[i] || !sh_local->child_up[i])
- continue;
-
- /* this is a sink, so write to it */
+ gf_log (this->name, GF_LOG_DEBUG, "Acquired lock for range %"PRIu64
+ " %"PRIu64, loop_sh->offset, loop_sh->block_size);
+ loop_sh->data_lock_held = _gf_true;
+ loop_sh->sh_data_algo_start (loop_frame, this);
+ return 0;
+}
- STACK_WIND_COOKIE (rw_frame, sh_full_write_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->writev,
- sh->healing_fd, vector, count, offset,
- iobref);
+static int
+sh_loop_lock_failure (call_frame_t *loop_frame, xlator_t *this)
+{
+ call_frame_t *sh_frame = NULL;
+ afr_local_t *loop_local = NULL;
+ afr_self_heal_t *loop_sh = NULL;
+
+ loop_local = loop_frame->local;
+ loop_sh = &loop_local->self_heal;
+ sh_frame = loop_sh->sh_frame;
+
+ gf_log (this->name, GF_LOG_ERROR, "failed lock for range %"PRIu64
+ " %"PRIu64, loop_sh->offset, loop_sh->block_size);
+ sh_loop_finish (loop_sh->old_loop_frame, this);
+ loop_sh->old_loop_frame = NULL;
+ sh_loop_return (sh_frame, this, loop_frame, -1, ENOTCONN);
+ return 0;
+}
- if (!--call_count)
- break;
- }
+static int
+sh_loop_frame_create (call_frame_t *sh_frame, xlator_t *this,
+ call_frame_t *old_loop_frame, call_frame_t **loop_frame)
+{
+ call_frame_t *new_loop_frame = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_local_t *new_loop_local = NULL;
+ afr_self_heal_t *new_loop_sh = NULL;
+ afr_private_t *priv = NULL;
+
+ GF_ASSERT (sh_frame);
+ GF_ASSERT (loop_frame);
+
+ *loop_frame = NULL;
+ local = sh_frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+
+ new_loop_frame = copy_frame (sh_frame);
+ if (!new_loop_frame)
+ goto out;
+ //We want the frame to have same lk_owner as sh_frame
+ //so that locks translator allows conflicting locks
+ new_loop_local = afr_self_heal_local_init (local, this);
+ if (!new_loop_local)
+ goto out;
+ new_loop_frame->local = new_loop_local;
+ new_loop_sh = &new_loop_local->self_heal;
+ new_loop_sh->sources = memdup (sh->sources,
+ priv->child_count * sizeof (*sh->sources));
+ if (!new_loop_sh->sources)
+ goto out;
+ new_loop_sh->write_needed = GF_CALLOC (priv->child_count,
+ sizeof (*new_loop_sh->write_needed),
+ gf_afr_mt_char);
+ if (!new_loop_sh->write_needed)
+ goto out;
+ new_loop_sh->checksum = GF_CALLOC (priv->child_count, MD5_DIGEST_LENGTH,
+ gf_afr_mt_uint8_t);
+ if (!new_loop_sh->checksum)
+ goto out;
+ new_loop_sh->inode = inode_ref (sh->inode);
+ new_loop_sh->sh_data_algo_start = sh->sh_data_algo_start;
+ new_loop_sh->source = sh->source;
+ new_loop_sh->active_sinks = sh->active_sinks;
+ new_loop_sh->healing_fd = fd_ref (sh->healing_fd);
+ new_loop_sh->file_has_holes = sh->file_has_holes;
+ new_loop_sh->old_loop_frame = old_loop_frame;
+ new_loop_sh->sh_frame = sh_frame;
+ *loop_frame = new_loop_frame;
+ return 0;
out:
- return 0;
+ sh_destroy_frame (new_loop_frame, this);
+ return -ENOMEM;
}
-
static int
-sh_full_read_write (call_frame_t *frame, xlator_t *this, off_t offset)
+sh_loop_start (call_frame_t *sh_frame, xlator_t *this, off_t offset,
+ call_frame_t *old_loop_frame)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- afr_local_t * rw_local = NULL;
- afr_self_heal_t *rw_sh = NULL;
- afr_self_heal_t *sh = NULL;
+ call_frame_t *new_loop_frame = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_local_t *new_loop_local = NULL;
+ afr_self_heal_t *new_loop_sh = NULL;
+ int ret = 0;
- call_frame_t *rw_frame = NULL;
+ GF_ASSERT (sh_frame);
- int32_t op_errno = 0;
+ local = sh_frame->local;
+ sh = &local->self_heal;
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
-
- rw_frame = copy_frame (frame);
- if (!rw_frame)
+ ret = sh_loop_frame_create (sh_frame, this, old_loop_frame,
+ &new_loop_frame);
+ if (ret)
goto out;
-
- ALLOC_OR_GOTO (rw_local, afr_local_t, out);
-
- rw_frame->local = rw_local;
- rw_sh = &rw_local->self_heal;
-
- rw_sh->offset = offset;
- rw_sh->sh_frame = frame;
-
- STACK_WIND_COOKIE (rw_frame, sh_full_read_cbk,
- (void *) (long) offset,
- priv->children[sh->source],
- priv->children[sh->source]->fops->readv,
- sh->healing_fd, sh->block_size,
- offset);
+ new_loop_local = new_loop_frame->local;
+ new_loop_sh = &new_loop_local->self_heal;
+ new_loop_sh->offset = offset;
+ new_loop_sh->block_size = sh->block_size;
+ afr_sh_data_lock (new_loop_frame, this, offset, new_loop_sh->block_size,
+ _gf_true, this->name, sh_loop_lock_success, sh_loop_lock_failure);
return 0;
-
out:
- sh->op_failed = 1;
-
- sh_full_loop_driver (frame, this);
-
- return 0;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ if (old_loop_frame)
+ sh_loop_finish (old_loop_frame, this);
+ sh_loop_return (sh_frame, this, new_loop_frame, -1, ENOMEM);
+ return 0;
}
-
static int
-sh_full_loop_driver (call_frame_t *frame, xlator_t *this)
+sh_loop_driver (call_frame_t *sh_frame, xlator_t *this,
+ gf_boolean_t is_first_call, call_frame_t *old_loop_frame)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_sh_algo_full_private_t *sh_priv = NULL;
-
- int loop = 0;
- int recurse = 0;
-
- off_t offset = 0;
-
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
+ afr_local_t * local = NULL;
+ afr_self_heal_t * sh = NULL;
+ afr_sh_algo_private_t *sh_priv = NULL;
+ gf_boolean_t is_driver_done = _gf_false;
+ blksize_t block_size = 0;
+ int loop = 0;
+ off_t offset = 0;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ local = sh_frame->local;
+ sh = &local->self_heal;
sh_priv = sh->private;
- if (sh->op_failed) {
- if (sh_priv->loops_running == 0) {
- gf_log (this->name, GF_LOG_TRACE,
- "full self-heal aborting on %s",
- local->loc.path);
-
- sh_full_private_cleanup (frame, this);
- local->self_heal.algo_abort_cbk (frame, this);
- }
-
- goto out;
- }
-
- if (sh_priv->offset >= sh->file_size) {
- if (sh_priv->loops_running == 0) {
-
- gf_log (this->name, GF_LOG_TRACE,
- "full self-heal completed on %s",
- local->loc.path);
-
- sh_full_private_cleanup (frame, this);
- local->self_heal.algo_completion_cbk (frame, this);
- }
-
- goto out;
- }
-
-spawn:
- loop = 0;
- recurse = 0;
-
LOCK (&sh_priv->lock);
{
- if ((sh_priv->loops_running < priv->data_self_heal_window_size)
- && (sh_priv->offset < sh->file_size)) {
-
- gf_log (this->name, GF_LOG_TRACE,
- "spawning a loop for offset %"PRId64,
- sh_priv->offset);
-
- offset = sh_priv->offset;
- sh_priv->offset += sh->block_size;
-
+ if (!is_first_call)
+ sh_priv->loops_running--;
+ offset = sh_priv->offset;
+ block_size = sh->block_size;
+ while ((!sh->eof_reached) &&
+ (!is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) &&
+ (sh_priv->loops_running < priv->data_self_heal_window_size)
+ && (sh_priv->offset < sh->file_size)) {
+
+ loop++;
+ sh_priv->offset += block_size;
sh_priv->loops_running++;
- loop = 1;
-
- if (sh_priv->offset < sh->file_size)
- recurse = 1;
+ if (!is_first_call)
+ break;
+ }
+ if (0 == sh_priv->loops_running) {
+ is_driver_done = _gf_true;
}
}
UNLOCK (&sh_priv->lock);
- if (loop) {
- sh_full_read_write (frame, this, offset);
- if (recurse)
- goto spawn;
+ if (0 == loop) {
+ //loop finish does unlock, but the erasing of the pending
+ //xattrs needs to happen before that so do not finish the loop
+ if (is_driver_done &&
+ !is_self_heal_failed (sh, AFR_CHECK_SPECIFIC))
+ goto driver_done;
+ if (old_loop_frame) {
+ sh_loop_finish (old_loop_frame, this);
+ old_loop_frame = NULL;
+ }
}
-out:
- return 0;
-}
-
-
-int
-afr_sh_algo_full (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t * local = NULL;
- afr_self_heal_t * sh = NULL;
- afr_sh_algo_full_private_t *sh_priv = NULL;
-
- local = frame->local;
- sh = &local->self_heal;
-
- sh_priv = GF_CALLOC (1, sizeof (*sh_priv),
- gf_afr_mt_afr_private_t);
-
- LOCK_INIT (&sh_priv->lock);
-
- sh->private = sh_priv;
-
- local->call_count = 0;
-
- sh_full_loop_driver (frame, this);
- return 0;
-}
-
-
-/*
- * The "diff" algorithm. Copies only those blocks whose checksums
- * don't match with those of source.
- */
-
-
-static void
-sh_diff_private_cleanup (call_frame_t *frame, xlator_t *this)
-{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- afr_self_heal_t * sh = NULL;
- afr_sh_algo_diff_private_t *sh_priv = NULL;
-
- int i;
-
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
-
- sh_priv = sh->private;
-
- for (i = 0; i < priv->data_self_heal_window_size; i++) {
- if (sh_priv->loops[i]) {
- if (sh_priv->loops[i]->write_needed)
- GF_FREE (sh_priv->loops[i]->write_needed);
-
- if (sh_priv->loops[i]->checksum)
- GF_FREE (sh_priv->loops[i]->checksum);
+ //If we have more loops to form we should finish previous loop after
+ //the next loop lock
+ while (loop--) {
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+ // op failed in other loop, stop spawning more loops
+ if (old_loop_frame) {
+ sh_loop_finish (old_loop_frame, this);
+ old_loop_frame = NULL;
+ }
+ sh_loop_driver (sh_frame, this, _gf_false, NULL);
+ } else {
+ gf_log (this->name, GF_LOG_TRACE, "spawning a loop "
+ "for offset %"PRId64, offset);
- GF_FREE (sh_priv->loops[i]);
+ sh_loop_start (sh_frame, this, offset, old_loop_frame);
+ old_loop_frame = NULL;
+ offset += block_size;
}
}
- if (sh_priv) {
- if (sh_priv->loops)
- GF_FREE (sh_priv->loops);
-
- GF_FREE (sh_priv);
+driver_done:
+ if (is_driver_done) {
+ sh_loop_driver_done (sh_frame, this, old_loop_frame);
}
-
-
-}
-
-
-static uint32_t
-__make_cookie (int loop_index, int child_index)
-{
- uint32_t ret = (loop_index << 16) | child_index;
- return ret;
-}
-
-
-static int
-__loop_index (uint32_t cookie)
-{
- return (cookie & 0xFFFF0000) >> 16;
+ return 0;
}
-
static int
-__child_index (uint32_t cookie)
+sh_loop_return (call_frame_t *sh_frame, xlator_t *this, call_frame_t *loop_frame,
+ int32_t op_ret, int32_t op_errno)
{
- return (cookie & 0x0000FFFF);
-}
-
+ afr_local_t * loop_local = NULL;
+ afr_self_heal_t * loop_sh = NULL;
+ afr_local_t * sh_local = NULL;
+ afr_self_heal_t *sh = NULL;
-static void
-sh_diff_loop_state_reset (struct sh_diff_loop_state *loop_state, int child_count)
-{
- loop_state->active = _gf_false;
-// loop_state->offset = 0;
-
- memset (loop_state->write_needed,
- 0, sizeof (*loop_state->write_needed) * child_count);
-
- memset (loop_state->checksum,
- 0, MD5_DIGEST_LEN * child_count);
-}
-
-
-static int
-sh_diff_number_of_writes_needed (unsigned char *write_needed, int child_count)
-{
- int writes = 0;
- int i;
+ sh_local = sh_frame->local;
+ sh = &sh_local->self_heal;
- for (i = 0; i < child_count; i++) {
- if (write_needed[i])
- writes++;
+ if (loop_frame) {
+ loop_local = loop_frame->local;
+ if (loop_local)
+ loop_sh = &loop_local->self_heal;
+ if (loop_sh)
+ gf_log (this->name, GF_LOG_TRACE, "loop for offset "
+ "%"PRId64" returned", loop_sh->offset);
}
- return writes;
-}
-
+ if (op_ret == -1) {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_set_error (sh, op_errno);
+ if (loop_frame) {
+ sh_loop_finish (loop_frame, this);
+ loop_frame = NULL;
+ }
+ }
-static int
-sh_diff_loop_driver (call_frame_t *frame, xlator_t *this);
+ sh_loop_driver (sh_frame, this, _gf_false, loop_frame);
+ return 0;
+}
static int
-sh_diff_loop_return (call_frame_t *rw_frame, xlator_t *this,
- struct sh_diff_loop_state *loop_state)
+sh_loop_write_cbk (call_frame_t *loop_frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ struct iatt *postbuf, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * rw_local = NULL;
- afr_self_heal_t * rw_sh = NULL;
-
- call_frame_t *sh_frame = NULL;
- afr_local_t * sh_local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_sh_algo_diff_private_t *sh_priv = NULL;
-
- priv = this->private;
-
- rw_local = rw_frame->local;
- rw_sh = &rw_local->self_heal;
-
- sh_frame = rw_sh->sh_frame;
+ afr_private_t * priv = NULL;
+ afr_local_t * loop_local = NULL;
+ afr_self_heal_t * loop_sh = NULL;
+ call_frame_t *sh_frame = NULL;
+ afr_local_t * sh_local = NULL;
+ afr_self_heal_t *sh = NULL;
+ int call_count = 0;
+ int child_index = 0;
+
+ priv = this->private;
+ loop_local = loop_frame->local;
+ loop_sh = &loop_local->self_heal;
+
+ sh_frame = loop_sh->sh_frame;
sh_local = sh_frame->local;
sh = &sh_local->self_heal;
- sh_priv = sh->private;
+
+ child_index = (long) cookie;
gf_log (this->name, GF_LOG_TRACE,
- "loop for offset %"PRId64" returned", loop_state->offset);
+ "wrote %d bytes of data from %s to child %d, offset %"PRId64"",
+ op_ret, sh_local->loc.path, child_index, loop_sh->offset);
- LOCK (&sh_priv->lock);
- {
- sh_priv->loops_running--;
- sh_diff_loop_state_reset (loop_state, priv->child_count);
+ if (op_ret == -1) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "write to %s failed on subvolume %s (%s)",
+ sh_local->loc.path,
+ priv->children[child_index]->name,
+ strerror (op_errno));
+
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_set_error (loop_sh, op_errno);
+ } else if (op_ret < loop_local->cont.writev.vector->iov_len) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "incomplete write to %s on subvolume %s "
+ "(expected %lu, returned %d)", sh_local->loc.path,
+ priv->children[child_index]->name,
+ loop_local->cont.writev.vector->iov_len, op_ret);
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
}
- UNLOCK (&sh_priv->lock);
- AFR_STACK_DESTROY (rw_frame);
+ call_count = afr_frame_return (loop_frame);
+
+ if (call_count == 0) {
+ iobref_unref(loop_local->cont.writev.iobref);
- sh_diff_loop_driver (sh_frame, this);
+ sh_loop_return (sh_frame, this, loop_frame,
+ loop_sh->op_ret, loop_sh->op_errno);
+ }
return 0;
}
-
-static int
-sh_diff_write_cbk (call_frame_t *rw_frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- struct iatt *postbuf)
+static void
+sh_prune_writes_needed (call_frame_t *sh_frame, call_frame_t *loop_frame,
+ afr_private_t *priv)
{
- afr_private_t * priv = NULL;
- afr_local_t * rw_local = NULL;
- afr_self_heal_t * rw_sh = NULL;
-
- call_frame_t *sh_frame = NULL;
- afr_local_t * sh_local = NULL;
- afr_self_heal_t *sh = NULL;
+ afr_local_t *sh_local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_local_t *loop_local = NULL;
+ afr_self_heal_t *loop_sh = NULL;
+ int i = 0;
- afr_sh_algo_diff_private_t *sh_priv;
- struct sh_diff_loop_state *loop_state;
+ sh_local = sh_frame->local;
+ sh = &sh_local->self_heal;
- int call_count = 0;
- int child_index = 0;
- int loop_index = 0;
+ if (!strcmp (sh->algo->name, "diff"))
+ return;
- priv = this->private;
- rw_local = rw_frame->local;
- rw_sh = &rw_local->self_heal;
+ loop_local = loop_frame->local;
+ loop_sh = &loop_local->self_heal;
- sh_frame = rw_sh->sh_frame;
- sh_local = sh_frame->local;
- sh = &sh_local->self_heal;
- sh_priv = sh->private;
-
- child_index = __child_index ((uint32_t) (long) cookie);
- loop_index = __loop_index ((uint32_t) (long) cookie);
- loop_state = sh_priv->loops[loop_index];
-
- gf_log (this->name, GF_LOG_TRACE,
- "wrote %d bytes of data from %s to child %d, offset %"PRId64"",
- op_ret, sh_local->loc.path, child_index,
- loop_state->offset);
-
- LOCK (&sh_frame->lock);
- {
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "write to %s failed on subvolume %s (%s)",
- sh_local->loc.path,
- priv->children[child_index]->name,
- strerror (op_errno));
-
- sh->op_failed = 1;
- }
- }
- UNLOCK (&sh_frame->lock);
-
- call_count = afr_frame_return (rw_frame);
-
- if (call_count == 0) {
- sh_diff_loop_return (rw_frame, this, loop_state);
- }
-
- return 0;
+ /* full self-heal guarantees there exists atleast 1 file with size 0
+ * That means for other files we can preserve holes that come after
+ * its size before 'trim'
+ */
+ for (i = 0; i < priv->child_count; i++) {
+ if (loop_sh->write_needed[i] &&
+ ((loop_sh->offset + 1) > sh->buf[i].ia_size))
+ loop_sh->write_needed[i] = 0;
+ }
}
-
static int
-sh_diff_read_cbk (call_frame_t *rw_frame, void *cookie,
+sh_loop_read_cbk (call_frame_t *loop_frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
struct iovec *vector, int32_t count, struct iatt *buf,
- struct iobref *iobref)
+ struct iobref *iobref, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * rw_local = NULL;
- afr_self_heal_t * rw_sh = NULL;
-
- afr_sh_algo_diff_private_t * sh_priv = NULL;
-
- call_frame_t *sh_frame = NULL;
- afr_local_t * sh_local = NULL;
- afr_self_heal_t *sh = NULL;
-
- int loop_index;
- struct sh_diff_loop_state *loop_state;
-
- uint32_t wcookie;
-
- int i = 0;
- int call_count = 0;
-
- priv = this->private;
- rw_local = rw_frame->local;
- rw_sh = &rw_local->self_heal;
-
- sh_frame = rw_sh->sh_frame;
+ afr_private_t * priv = NULL;
+ afr_local_t * loop_local = NULL;
+ afr_self_heal_t * loop_sh = NULL;
+ call_frame_t *sh_frame = NULL;
+ int i = 0;
+ int call_count = 0;
+ afr_local_t * sh_local = NULL;
+ afr_self_heal_t * sh = NULL;
+
+ priv = this->private;
+ loop_local = loop_frame->local;
+ loop_sh = &loop_local->self_heal;
+
+ sh_frame = loop_sh->sh_frame;
sh_local = sh_frame->local;
sh = &sh_local->self_heal;
- sh_priv = sh->private;
-
- loop_index = __loop_index ((uint32_t) (long) cookie);
- loop_state = sh_priv->loops[loop_index];
-
- call_count = sh_diff_number_of_writes_needed (loop_state->write_needed,
- priv->child_count);
- rw_local->call_count = call_count;
-
- gf_log (this->name, GF_LOG_TRACE,
- "read %d bytes of data from %s, offset %"PRId64"",
- op_ret, sh_local->loc.path, loop_state->offset);
-
- if ((op_ret <= 0) ||
- (call_count == 0)) {
- sh_diff_loop_return (rw_frame, this, loop_state);
+ gf_log (this->name, GF_LOG_TRACE,
+ "read %d bytes of data from %s, offset %"PRId64"",
+ op_ret, loop_local->loc.path, loop_sh->offset);
+
+ if (op_ret <= 0) {
+ if (op_ret < 0) {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ gf_log (this->name, GF_LOG_ERROR, "read failed on %d "
+ "for %s reason :%s", sh->source,
+ sh_local->loc.path, strerror (errno));
+ } else {
+ sh->eof_reached = _gf_true;
+ gf_log (this->name, GF_LOG_DEBUG, "Eof reached for %s",
+ sh_local->loc.path);
+ }
+ sh_loop_return (sh_frame, this, loop_frame, op_ret, op_errno);
+ goto out;
+ }
- return 0;
- }
+ if (loop_sh->file_has_holes && iov_0filled (vector, count) == 0)
+ sh_prune_writes_needed (sh_frame, loop_frame, priv);
- if (sh->file_has_holes) {
- if (iov_0filled (vector, count) == 0) {
+ call_count = sh_number_of_writes_needed (loop_sh->write_needed,
+ priv->child_count);
+ if (call_count == 0) {
+ sh_loop_return (sh_frame, this, loop_frame, 0, 0);
+ goto out;
+ }
- sh_diff_loop_return (rw_frame, this, loop_state);
- goto out;
- }
- }
+ loop_local->call_count = call_count;
- for (i = 0; i < priv->child_count; i++) {
- if (loop_state->write_needed[i]) {
- wcookie = __make_cookie (loop_index, i);
+ /*
+ * We only really need the request size at the moment, but the buffer
+ * is required if we want to issue a retry in the event of a short write.
+ * Therefore, we duplicate the vector and ref the iobref here...
+ */
+ loop_local->cont.writev.vector = iov_dup(vector, count);
+ loop_local->cont.writev.iobref = iobref_ref(iobref);
- STACK_WIND_COOKIE (rw_frame, sh_diff_write_cbk,
- (void *) (long) wcookie,
- priv->children[i],
- priv->children[i]->fops->writev,
- sh->healing_fd, vector, count,
- loop_state->offset, iobref);
+ for (i = 0; i < priv->child_count; i++) {
+ if (!loop_sh->write_needed[i])
+ continue;
+ STACK_WIND_COOKIE (loop_frame, sh_loop_write_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->writev,
+ loop_sh->healing_fd, vector, count,
+ loop_sh->offset, 0, iobref, NULL);
- if (!--call_count)
- break;
- }
+ if (!--call_count)
+ break;
}
out:
- return 0;
+ return 0;
}
static int
-sh_diff_read (call_frame_t *rw_frame, xlator_t *this,
- int loop_index)
+sh_loop_read (call_frame_t *loop_frame, xlator_t *this)
{
- afr_private_t * priv = NULL;
- afr_local_t * rw_local = NULL;
- afr_self_heal_t * rw_sh = NULL;
-
- afr_sh_algo_diff_private_t * sh_priv = NULL;
- struct sh_diff_loop_state *loop_state;
-
- call_frame_t *sh_frame = NULL;
- afr_local_t * sh_local = NULL;
- afr_self_heal_t *sh = NULL;
-
- uint32_t cookie;
-
- priv = this->private;
- rw_local = rw_frame->local;
- rw_sh = &rw_local->self_heal;
-
- sh_frame = rw_sh->sh_frame;
- sh_local = sh_frame->local;
- sh = &sh_local->self_heal;
- sh_priv = sh->private;
-
- loop_state = sh_priv->loops[loop_index];
+ afr_private_t *priv = NULL;
+ afr_local_t *loop_local = NULL;
+ afr_self_heal_t *loop_sh = NULL;
- cookie = __make_cookie (loop_index, sh->source);
+ priv = this->private;
+ loop_local = loop_frame->local;
+ loop_sh = &loop_local->self_heal;
- STACK_WIND_COOKIE (rw_frame, sh_diff_read_cbk,
- (void *) (long) cookie,
- priv->children[sh->source],
- priv->children[sh->source]->fops->readv,
- sh->healing_fd, sh_priv->block_size,
- loop_state->offset);
+ STACK_WIND_COOKIE (loop_frame, sh_loop_read_cbk,
+ (void *) (long) loop_sh->source,
+ priv->children[loop_sh->source],
+ priv->children[loop_sh->source]->fops->readv,
+ loop_sh->healing_fd, loop_sh->block_size,
+ loop_sh->offset, 0, NULL);
- return 0;
+ return 0;
}
static int
-sh_diff_checksum_cbk (call_frame_t *rw_frame, void *cookie, xlator_t *this,
+sh_diff_checksum_cbk (call_frame_t *loop_frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
- uint32_t weak_checksum, uint8_t *strong_checksum)
+ uint32_t weak_checksum, uint8_t *strong_checksum,
+ dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * rw_local = NULL;
- afr_self_heal_t *rw_sh = NULL;
-
- call_frame_t *sh_frame = NULL;
- afr_local_t * sh_local = NULL;
- afr_self_heal_t *sh = NULL;
-
- afr_sh_algo_diff_private_t * sh_priv = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *loop_local = NULL;
+ afr_self_heal_t *loop_sh = NULL;
+ call_frame_t *sh_frame = NULL;
+ afr_local_t *sh_local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_sh_algo_private_t *sh_priv = NULL;
+ int child_index = 0;
+ int call_count = 0;
+ int i = 0;
+ int write_needed = 0;
- int loop_index = 0;
- int child_index = 0;
- struct sh_diff_loop_state *loop_state;
-
- int call_count = 0;
- int i = 0;
- int write_needed = 0;
-
- priv = this->private;
+ priv = this->private;
- rw_local = rw_frame->local;
- rw_sh = &rw_local->self_heal;
+ loop_local = loop_frame->local;
+ loop_sh = &loop_local->self_heal;
- sh_frame = rw_sh->sh_frame;
+ sh_frame = loop_sh->sh_frame;
sh_local = sh_frame->local;
sh = &sh_local->self_heal;
sh_priv = sh->private;
- child_index = __child_index ((uint32_t) (long) cookie);
- loop_index = __loop_index ((uint32_t) (long) cookie);
-
- loop_state = sh_priv->loops[loop_index];
+ child_index = (long) cookie;
if (op_ret < 0) {
gf_log (this->name, GF_LOG_ERROR,
"checksum on %s failed on subvolume %s (%s)",
sh_local->loc.path, priv->children[child_index]->name,
strerror (op_errno));
-
- sh->op_failed = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
} else {
- memcpy (loop_state->checksum + child_index * MD5_DIGEST_LEN,
- strong_checksum,
- MD5_DIGEST_LEN);
+ memcpy (loop_sh->checksum + child_index * MD5_DIGEST_LENGTH,
+ strong_checksum, MD5_DIGEST_LENGTH);
}
- call_count = afr_frame_return (rw_frame);
+ call_count = afr_frame_return (loop_frame);
if (call_count == 0) {
for (i = 0; i < priv->child_count; i++) {
if (sh->sources[i] || !sh_local->child_up[i])
continue;
- if (memcmp (loop_state->checksum + (i * MD5_DIGEST_LEN),
- loop_state->checksum + (sh->source * MD5_DIGEST_LEN),
- MD5_DIGEST_LEN)) {
+ if (memcmp (loop_sh->checksum + (i * MD5_DIGEST_LENGTH),
+ loop_sh->checksum + (sh->source * MD5_DIGEST_LENGTH),
+ MD5_DIGEST_LENGTH)) {
/*
- Checksums differ, so this block
- must be written to this sink
+ Checksums differ, so this block
+ must be written to this sink
*/
- gf_log (this->name, GF_LOG_TRACE,
+ gf_log (this->name, GF_LOG_DEBUG,
"checksum on subvolume %s at offset %"
PRId64" differs from that on source",
- priv->children[i]->name, loop_state->offset);
+ priv->children[i]->name, loop_sh->offset);
- write_needed = loop_state->write_needed[i] = 1;
+ write_needed = loop_sh->write_needed[i] = 1;
}
}
@@ -800,270 +664,171 @@ sh_diff_checksum_cbk (call_frame_t *rw_frame, void *cookie, xlator_t *this,
}
UNLOCK (&sh_priv->lock);
- if (write_needed && !sh->op_failed) {
- sh_diff_read (rw_frame, this, loop_index);
+ if (write_needed &&
+ !is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+ sh_loop_read (loop_frame, this);
} else {
- sh->offset += sh_priv->block_size;
-
- sh_diff_loop_return (rw_frame, this, loop_state);
+ sh_loop_return (sh_frame, this, loop_frame,
+ op_ret, op_errno);
}
}
return 0;
}
-
-static int
-sh_diff_find_unused_loop (afr_sh_algo_diff_private_t *sh_priv, int max)
-{
- int i;
-
- LOCK (&sh_priv->lock);
- {
- for (i = 0; i < max; i++) {
- if (sh_priv->loops[i]->active == _gf_false) {
- sh_priv->loops[i]->active = _gf_true;
- break;
- }
- }
- }
- UNLOCK (&sh_priv->lock);
-
- if (i == max) {
- gf_log ("[sh-diff]", GF_LOG_ERROR,
- "no free loops found! This shouldn't happen. Please"
- " report this to gluster-devel@nongnu.org");
- }
-
- return i;
-}
-
-
static int
-sh_diff_checksum (call_frame_t *frame, xlator_t *this, off_t offset)
+sh_diff_checksum (call_frame_t *loop_frame, xlator_t *this)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- afr_local_t * rw_local = NULL;
- afr_self_heal_t * sh = NULL;
- afr_self_heal_t * rw_sh = NULL;
-
- afr_sh_algo_diff_private_t * sh_priv = NULL;
-
- call_frame_t *rw_frame = NULL;
-
- uint32_t cookie;
- int loop_index = 0;
- struct sh_diff_loop_state *loop_state = NULL;
-
- int32_t op_errno = 0;
-
- int call_count = 0;
- int i = 0;
-
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
+ afr_private_t *priv = NULL;
+ afr_local_t *loop_local = NULL;
+ afr_self_heal_t *loop_sh = NULL;
+ int call_count = 0;
+ int i = 0;
- sh_priv = sh->private;
-
- rw_frame = copy_frame (frame);
- if (!rw_frame)
- goto out;
-
- ALLOC_OR_GOTO (rw_local, afr_local_t, out);
-
- rw_frame->local = rw_local;
- rw_sh = &rw_local->self_heal;
-
- rw_sh->offset = sh->offset;
- rw_sh->sh_frame = frame;
+ priv = this->private;
+ loop_local = loop_frame->local;
+ loop_sh = &loop_local->self_heal;
- call_count = sh->active_sinks + 1; /* sinks and source */
+ call_count = loop_sh->active_sinks + 1; /* sinks and source */
- rw_local->call_count = call_count;
+ loop_local->call_count = call_count;
- loop_index = sh_diff_find_unused_loop (sh_priv, priv->data_self_heal_window_size);
-
- loop_state = sh_priv->loops[loop_index];
- loop_state->offset = offset;
-
- /* we need to send both the loop index and child index,
- so squeeze them both into a 32-bit number */
-
- cookie = __make_cookie (loop_index, sh->source);
-
- STACK_WIND_COOKIE (rw_frame, sh_diff_checksum_cbk,
- (void *) (long) cookie,
- priv->children[sh->source],
- priv->children[sh->source]->fops->rchecksum,
- sh->healing_fd,
- offset, sh_priv->block_size);
+ STACK_WIND_COOKIE (loop_frame, sh_diff_checksum_cbk,
+ (void *) (long) loop_sh->source,
+ priv->children[loop_sh->source],
+ priv->children[loop_sh->source]->fops->rchecksum,
+ loop_sh->healing_fd,
+ loop_sh->offset, loop_sh->block_size, NULL);
for (i = 0; i < priv->child_count; i++) {
- if (sh->sources[i] || !local->child_up[i])
+ if (loop_sh->sources[i] || !loop_local->child_up[i])
continue;
- cookie = __make_cookie (loop_index, i);
-
- STACK_WIND_COOKIE (rw_frame, sh_diff_checksum_cbk,
- (void *) (long) cookie,
+ STACK_WIND_COOKIE (loop_frame, sh_diff_checksum_cbk,
+ (void *) (long) i,
priv->children[i],
priv->children[i]->fops->rchecksum,
- sh->healing_fd,
- offset, sh_priv->block_size);
+ loop_sh->healing_fd,
+ loop_sh->offset, loop_sh->block_size, NULL);
if (!--call_count)
break;
}
return 0;
-
-out:
- sh->op_failed = 1;
-
- sh_diff_loop_driver (frame, this);
-
- return 0;
}
-
static int
-sh_diff_loop_driver (call_frame_t *frame, xlator_t *this)
+sh_full_read_write_to_sinks (call_frame_t *loop_frame, xlator_t *this)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- afr_self_heal_t * sh = NULL;
- afr_sh_algo_diff_private_t *sh_priv = NULL;
-
- int loop = 0;
- int recurse = 0;
-
- off_t offset = 0;
- char sh_type_str[256] = {0,};
-
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
- sh_priv = sh->private;
-
- afr_self_heal_type_str_get(sh, sh_type_str, sizeof(sh_type_str));
-
- if (sh->op_failed) {
- if (sh_priv->loops_running == 0) {
- gf_log (this->name, GF_LOG_ERROR,
- "diff %s self-heal aborting on %s",
- sh_type_str, local->loc.path);
-
- sh_diff_private_cleanup (frame, this);
- local->self_heal.algo_abort_cbk (frame, this);
- }
-
- goto out;
- }
-
- if (sh_priv->offset >= sh->file_size) {
- if (sh_priv->loops_running == 0) {
- gf_log (this->name, GF_LOG_TRACE,
- "diff %s self-heal completed on %s",
- sh_type_str, local->loc.path);
+ afr_private_t *priv = NULL;
+ afr_local_t *loop_local = NULL;
+ afr_self_heal_t *loop_sh = NULL;
+ int i = 0;
+ priv = this->private;
+ loop_local = loop_frame->local;
+ loop_sh = &loop_local->self_heal;
- gf_log (this->name, GF_LOG_NORMAL,
- "diff %s self-heal on %s: %d blocks of %d were different (%.2f%%)",
- sh_type_str, local->loc.path,
- sh_priv->diff_blocks, sh_priv->total_blocks,
- ((sh_priv->diff_blocks * 1.0)/sh_priv->total_blocks) * 100);
-
- sh_diff_private_cleanup (frame, this);
- local->self_heal.algo_completion_cbk (frame, this);
- }
-
- goto out;
- }
-
-spawn:
- loop = 0;
- recurse = 0;
-
- LOCK (&sh_priv->lock);
- {
- if ((sh_priv->loops_running < priv->data_self_heal_window_size)
- && (sh_priv->offset < sh->file_size)) {
-
- gf_log (this->name, GF_LOG_TRACE,
- "spawning a loop for offset %"PRId64,
- sh_priv->offset);
-
- offset = sh_priv->offset;
- sh_priv->offset += sh_priv->block_size;
-
- sh_priv->loops_running++;
-
- loop = 1;
-
- if (sh_priv->offset < sh->file_size)
- recurse = 1;
- }
- }
- UNLOCK (&sh_priv->lock);
-
- if (loop) {
- sh_diff_checksum (frame, this, offset);
- if (recurse)
- goto spawn;
+ for (i = 0; i < priv->child_count; i++) {
+ if (loop_sh->sources[i] || !loop_local->child_up[i])
+ continue;
+ loop_sh->write_needed[i] = 1;
}
-
-out:
+ sh_loop_read (loop_frame, this);
return 0;
}
-
-int
-afr_sh_algo_diff (call_frame_t *frame, xlator_t *this)
+afr_sh_algo_private_t*
+afr_sh_priv_init ()
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- afr_self_heal_t * sh = NULL;
- afr_sh_algo_diff_private_t *sh_priv = NULL;
-
- int i;
-
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
+ afr_sh_algo_private_t *sh_priv = NULL;
sh_priv = GF_CALLOC (1, sizeof (*sh_priv),
gf_afr_mt_afr_private_t);
-
- sh_priv->block_size = this->ctx->page_size;
-
- sh->private = sh_priv;
+ if (!sh_priv)
+ goto out;
LOCK_INIT (&sh_priv->lock);
+out:
+ return sh_priv;
+}
- local->call_count = 0;
+int
+afr_sh_transfer_lock (call_frame_t *dst, call_frame_t *src, char *dom,
+ unsigned int child_count)
+{
+ afr_local_t *dst_local = NULL;
+ afr_self_heal_t *dst_sh = NULL;
+ afr_local_t *src_local = NULL;
+ afr_self_heal_t *src_sh = NULL;
+ int ret = -1;
+
+ dst_local = dst->local;
+ dst_sh = &dst_local->self_heal;
+ src_local = src->local;
+ src_sh = &src_local->self_heal;
+ GF_ASSERT (src_sh->data_lock_held);
+ GF_ASSERT (!dst_sh->data_lock_held);
+ ret = afr_lk_transfer_datalock (dst, src, dom, child_count);
+ if (ret)
+ return ret;
+ src_sh->data_lock_held = _gf_false;
+ dst_sh->data_lock_held = _gf_true;
+ return 0;
+}
- sh_priv->loops = GF_CALLOC (priv->data_self_heal_window_size,
- sizeof (*sh_priv->loops),
- gf_afr_mt_sh_diff_loop_state);
+int
+afr_sh_start_loops (call_frame_t *sh_frame, xlator_t *this,
+ afr_sh_algo_fn sh_data_algo_start)
+{
+ call_frame_t *first_loop_frame = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ int ret = 0;
+ afr_private_t *priv = NULL;
- for (i = 0; i < priv->data_self_heal_window_size; i++) {
- sh_priv->loops[i] = GF_CALLOC (1, sizeof (*sh_priv->loops[i]),
- gf_afr_mt_sh_diff_loop_state);
+ local = sh_frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- sh_priv->loops[i]->checksum = GF_CALLOC (priv->child_count,
- MD5_DIGEST_LEN, gf_afr_mt_uint8_t);
- sh_priv->loops[i]->write_needed = GF_CALLOC (priv->child_count,
- sizeof (*sh_priv->loops[i]->write_needed),
- gf_afr_mt_char);
+ sh->sh_data_algo_start = sh_data_algo_start;
+ local->call_count = 0;
+ ret = sh_loop_frame_create (sh_frame, this, NULL, &first_loop_frame);
+ if (ret)
+ goto out;
+ ret = afr_sh_transfer_lock (first_loop_frame, sh_frame, this->name,
+ priv->child_count);
+ if (ret)
+ goto out;
+ sh->private = afr_sh_priv_init ();
+ if (!sh->private) {
+ ret = -1;
+ goto out;
}
+ sh_loop_driver (sh_frame, this, _gf_true, first_loop_frame);
+ ret = 0;
+out:
+ if (ret) {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ sh_loop_driver_done (sh_frame, this, NULL);
+ }
+ return 0;
+}
- sh_diff_loop_driver (frame, this);
-
+int
+afr_sh_algo_diff (call_frame_t *sh_frame, xlator_t *this)
+{
+ afr_sh_start_loops (sh_frame, this, sh_diff_checksum);
return 0;
}
+int
+afr_sh_algo_full (call_frame_t *sh_frame, xlator_t *this)
+{
+ afr_sh_start_loops (sh_frame, this, sh_full_read_write_to_sinks);
+ return 0;
+}
struct afr_sh_algorithm afr_self_heal_algorithms[] = {
{.name = "full", .fn = afr_sh_algo_full},
diff --git a/xlators/cluster/afr/src/afr-self-heal-algorithm.h b/xlators/cluster/afr/src/afr-self-heal-algorithm.h
index 0bdae3aa7..6b20789b1 100644
--- a/xlators/cluster/afr/src/afr-self-heal-algorithm.h
+++ b/xlators/cluster/afr/src/afr-self-heal-algorithm.h
@@ -1,26 +1,16 @@
/*
- Copyright (c) 2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __AFR_SELF_HEAL_ALGORITHM_H__
#define __AFR_SELF_HEAL_ALGORITHM_H__
-
typedef int (*afr_sh_algo_fn) (call_frame_t *frame,
xlator_t *this);
@@ -30,31 +20,13 @@ struct afr_sh_algorithm {
};
extern struct afr_sh_algorithm afr_self_heal_algorithms[3];
-
typedef struct {
gf_lock_t lock;
unsigned int loops_running;
off_t offset;
-} afr_sh_algo_full_private_t;
-
-struct sh_diff_loop_state {
- off_t offset;
- unsigned char *write_needed;
- uint8_t *checksum;
- gf_boolean_t active;
-};
-
-typedef struct {
- size_t block_size;
-
- gf_lock_t lock;
- unsigned int loops_running;
- off_t offset;
int32_t total_blocks;
int32_t diff_blocks;
-
- struct sh_diff_loop_state **loops;
-} afr_sh_algo_diff_private_t;
+} afr_sh_algo_private_t;
#endif /* __AFR_SELF_HEAL_ALGORITHM_H__ */
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.c b/xlators/cluster/afr/src/afr-self-heal-common.c
index d19f389ef..ef92b4205 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.c
+++ b/xlators/cluster/afr/src/afr-self-heal-common.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include "glusterfs.h"
@@ -27,6 +18,67 @@
#include "afr-self-heal.h"
#include "pump.h"
+#define ADD_FMT_STRING(msg, off, sh_str, status, print_log) \
+ do { \
+ if (AFR_SELF_HEAL_NOT_ATTEMPTED != status) { \
+ off += snprintf (msg + off, sizeof (msg) - off, \
+ " "sh_str" self heal %s,", \
+ get_sh_completion_status (status));\
+ print_log = 1; \
+ } \
+ } while (0)
+
+#define ADD_FMT_STRING_SYNC(msg, off, sh_str, status, print_log) \
+ do { \
+ if (AFR_SELF_HEAL_SYNC_BEGIN == status || \
+ AFR_SELF_HEAL_FAILED == status) { \
+ off += snprintf (msg + off, sizeof (msg) - off, \
+ " "sh_str" self heal %s,", \
+ get_sh_completion_status (status));\
+ print_log = 1; \
+ } \
+ } while (0)
+
+
+void
+afr_sh_reset (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+
+ memset (sh->child_errno, 0,
+ sizeof (*sh->child_errno) * priv->child_count);
+ memset (sh->buf, 0, sizeof (*sh->buf) * priv->child_count);
+ memset (sh->parentbufs, 0,
+ sizeof (*sh->parentbufs) * priv->child_count);
+ memset (sh->success, 0, sizeof (*sh->success) * priv->child_count);
+ memset (sh->locked_nodes, 0,
+ sizeof (*sh->locked_nodes) * priv->child_count);
+ sh->active_sinks = 0;
+
+ afr_reset_xattr (sh->xattr, priv->child_count);
+}
+
+//Intersection[child]=1 if child is part of intersection
+void
+afr_children_intersection_get (int32_t *set1, int32_t *set2,
+ int *intersection, unsigned int child_count)
+{
+ int i = 0;
+
+ memset (intersection, 0, sizeof (*intersection) * child_count);
+ for (i = 0; i < child_count; i++) {
+ intersection[i] = afr_is_child_present (set1, child_count, i)
+ && afr_is_child_present (set2, child_count,
+ i);
+ }
+}
+
/**
* select_source - select a source and return it
*/
@@ -34,118 +86,215 @@
int
afr_sh_select_source (int sources[], int child_count)
{
- int i;
- for (i = 0; i < child_count; i++)
- if (sources[i])
- return i;
+ int i = 0;
+ for (i = 0; i < child_count; i++)
+ if (sources[i])
+ return i;
- return -1;
+ return -1;
}
+void
+afr_sh_mark_source_sinks (call_frame_t *frame, xlator_t *this)
+{
+ int i = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int active_sinks = 0;
-/**
- * sink_count - return number of sinks in sources array
- */
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
-int
-afr_sh_sink_count (int sources[], int child_count)
-{
- int i;
- int sinks = 0;
- for (i = 0; i < child_count; i++)
- if (!sources[i])
- sinks++;
- return sinks;
+ for (i = 0; i < priv->child_count; i++) {
+ if (sh->sources[i] == 0 && local->child_up[i] == 1) {
+ active_sinks++;
+ sh->success[i] = 1;
+ } else if (sh->sources[i] == 1 && local->child_up[i] == 1) {
+ sh->success[i] = 1;
+ }
+ }
+ sh->active_sinks = active_sinks;
}
int
afr_sh_source_count (int sources[], int child_count)
{
- int i;
- int nsource = 0;
+ int i = 0;
+ int nsource = 0;
- for (i = 0; i < child_count; i++)
- if (sources[i])
- nsource++;
- return nsource;
+ for (i = 0; i < child_count; i++)
+ if (sources[i])
+ nsource++;
+ return nsource;
}
+void
+afr_sh_set_error (afr_self_heal_t *sh, int32_t op_errno)
+{
+ sh->op_ret = -1;
+ sh->op_errno = afr_most_important_error(sh->op_errno, op_errno,
+ _gf_false);
+}
-int
-afr_sh_supress_errenous_children (int sources[], int child_errno[],
- int child_count)
+void
+afr_sh_print_pending_matrix (int32_t *pending_matrix[], xlator_t *this)
{
- int i = 0;
+ afr_private_t * priv = this->private;
+ char *buf = NULL;
+ char *ptr = NULL;
+ int i = 0;
+ int j = 0;
- for (i = 0; i < child_count; i++) {
- if (child_errno[i] && sources[i]) {
- sources[i] = 0;
- }
- }
+ /* 10 digits per entry + 1 space + '[' and ']' */
+ buf = GF_MALLOC (priv->child_count * 11 + 8, gf_afr_mt_char);
- return 0;
+ for (i = 0; i < priv->child_count; i++) {
+ ptr = buf;
+ ptr += sprintf (ptr, "[ ");
+ for (j = 0; j < priv->child_count; j++) {
+ ptr += sprintf (ptr, "%d ", pending_matrix[i][j]);
+ }
+ sprintf (ptr, "]");
+ gf_log (this->name, GF_LOG_DEBUG, "pending_matrix: %s", buf);
+ }
+
+ GF_FREE (buf);
}
+char*
+afr_get_pending_matrix_str (int32_t *pending_matrix[], xlator_t *this)
+{
+ afr_private_t * priv = this->private;
+ char *buf = NULL;
+ char *ptr = NULL;
+ int i = 0;
+ int j = 0;
+ int child_count = priv->child_count;
+ char *matrix_begin = "[ [ ";
+ char *matrix_end = "] ]";
+ char *seperator = "] [ ";
+ int pending_entry_strlen = 12; //Including space after entry
+ int matrix_begin_strlen = 0;
+ int matrix_end_strlen = 0;
+ int seperator_strlen = 0;
+ int string_length = 0;
+ char *msg = "- Pending matrix: ";
+
+ /*
+ * for a list of lists of [ [ a b ] [ c d ] ]
+ * */
+
+ matrix_begin_strlen = strlen (matrix_begin);
+ matrix_end_strlen = strlen (matrix_end);
+ seperator_strlen = strlen (seperator);
+ string_length = matrix_begin_strlen + matrix_end_strlen
+ + (child_count -1) * seperator_strlen
+ + (child_count * child_count * pending_entry_strlen);
+
+ buf = GF_CALLOC (1, 1 + strlen (msg) + string_length , gf_afr_mt_char);
+ if (!buf)
+ goto out;
+
+ ptr = buf;
+ ptr += sprintf (ptr, "%s", msg);
+ ptr += sprintf (ptr, "%s", matrix_begin);
+ for (i = 0; i < priv->child_count; i++) {
+ for (j = 0; j < priv->child_count; j++) {
+ ptr += sprintf (ptr, "%d ", pending_matrix[i][j]);
+ }
+ if (i < priv->child_count -1)
+ ptr += sprintf (ptr, "%s", seperator);
+ }
+
+ ptr += sprintf (ptr, "%s", matrix_end);
+
+out:
+ return buf;
+}
void
-afr_sh_print_pending_matrix (int32_t *pending_matrix[], xlator_t *this)
+afr_sh_print_split_brain_log (int32_t *pending_matrix[], xlator_t *this,
+ const char *loc)
{
- afr_private_t * priv = this->private;
+ char *buf = NULL;
+ char *free_ptr = NULL;
- char *buf = NULL;
- char *ptr = NULL;
+ buf = afr_get_pending_matrix_str (pending_matrix, this);
+ if (buf)
+ free_ptr = buf;
+ else
+ buf = "";
- int i, j;
- /* 10 digits per entry + 1 space + '[' and ']' */
- buf = GF_MALLOC (priv->child_count * 11 + 8, gf_afr_mt_char);
+ gf_log (this->name, GF_LOG_ERROR, "Unable to self-heal contents of '%s'"
+ " (possible split-brain). Please delete the file from all but "
+ "the preferred subvolume.%s", loc, buf);
+ GF_FREE (free_ptr);
+ return;
+}
- for (i = 0; i < priv->child_count; i++) {
- ptr = buf;
- ptr += sprintf (ptr, "[ ");
- for (j = 0; j < priv->child_count; j++) {
- ptr += sprintf (ptr, "%d ", pending_matrix[i][j]);
- }
- sprintf (ptr, "]");
- gf_log (this->name, GF_LOG_TRACE,
- "pending_matrix: %s", buf);
- }
- GF_FREE (buf);
-}
+void
+afr_init_pending_matrix (int32_t **pending_matrix, size_t child_count)
+{
+ int i = 0;
+ int j = 0;
+
+ GF_ASSERT (pending_matrix);
+ for (i = 0; i < child_count; i++) {
+ for (j = 0; j < child_count; j++) {
+ pending_matrix[i][j] = 0;
+ }
+ }
+}
void
-afr_sh_build_pending_matrix (afr_private_t *priv,
- int32_t *pending_matrix[], dict_t *xattr[],
- int child_count, afr_transaction_type type)
+afr_mark_ignorant_subvols_as_pending (int32_t **pending_matrix,
+ unsigned char *ignorant_subvols,
+ size_t child_count)
{
- int i, j, k;
+ int i = 0;
+ int j = 0;
- /* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
- int32_t pending[3];
- void *pending_raw = NULL;
- int ret = -1;
+ GF_ASSERT (pending_matrix);
+ GF_ASSERT (ignorant_subvols);
- unsigned char *ignorant_subvols = NULL;
+ for (i = 0; i < child_count; i++) {
+ if (ignorant_subvols[i]) {
+ for (j = 0; j < child_count; j++) {
+ if (!ignorant_subvols[j])
+ pending_matrix[j][i] += 1;
+ }
+ }
+ }
+}
- ignorant_subvols = GF_CALLOC (sizeof (*ignorant_subvols), child_count,
- gf_afr_mt_char);
+int
+afr_build_pending_matrix (char **pending_key, int32_t **pending_matrix,
+ unsigned char *ignorant_subvols,
+ dict_t *xattr[], afr_transaction_type type,
+ size_t child_count)
+{
+ /* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
+ int32_t pending[3] = {0,};
+ void *pending_raw = NULL;
+ int ret = -1;
+ int i = 0;
+ int j = 0;
+ int k = 0;
- /* start clean */
- for (i = 0; i < child_count; i++) {
- for (j = 0; j < child_count; j++) {
- pending_matrix[i][j] = 0;
- }
- }
+ afr_init_pending_matrix (pending_matrix, child_count);
- for (i = 0; i < child_count; i++) {
- pending_raw = NULL;
+ for (i = 0; i < child_count; i++) {
+ pending_raw = NULL;
for (j = 0; j < child_count; j++) {
- ret = dict_get_ptr (xattr[i], priv->pending_key[j],
+ ret = dict_get_ptr (xattr[i], pending_key[j],
&pending_raw);
-
+
if (ret != 0) {
/*
* There is no xattr present. This means this
@@ -153,63 +302,26 @@ afr_sh_build_pending_matrix (afr_private_t *priv,
* subvolume.
*/
- ignorant_subvols[i] = 1;
+ if (ignorant_subvols)
+ ignorant_subvols[i] = 1;
continue;
}
- memcpy (pending, pending_raw, sizeof(pending));
+ memcpy (pending, pending_raw, sizeof(pending));
k = afr_index_for_transaction_type (type);
-
- pending_matrix[i][j] = ntoh32 (pending[k]);
- }
- }
- /*
- * Make all non-ignorant subvols point towards the ignorant
- * subvolumes.
- */
-
- for (i = 0; i < child_count; i++) {
- if (ignorant_subvols[i]) {
- for (j = 0; j < child_count; j++) {
- if (!ignorant_subvols[j])
- pending_matrix[j][i] += 1;
- }
+ pending_matrix[i][j] = ntoh32 (pending[k]);
}
}
- GF_FREE (ignorant_subvols);
+ return ret;
}
-
-/**
- * mark_sources: Mark all 'source' nodes and return number of source
- * nodes found
- *
- * A node (a row in the pending matrix) belongs to one of
- * three categories:
- *
- * M is the pending matrix.
- *
- * 'innocent' - M[i] is all zeroes
- * 'fool' - M[i] has i'th element = 1 (self-reference)
- * 'wise' - M[i] has i'th element = 0, others are 1 or 0.
- *
- * All 'innocent' nodes are sinks. If all nodes are innocent, no self-heal is
- * needed.
- *
- * A 'wise' node can be a source. If two 'wise' nodes conflict, it is
- * a split-brain. If one wise node refers to the other but the other doesn't
- * refer back, the referrer is a source.
- *
- * All fools are sinks, unless there are no 'wise' nodes. In that case,
- * one of the fools is made a source.
- */
-
typedef enum {
+ AFR_NODE_INVALID,
AFR_NODE_INNOCENT,
AFR_NODE_FOOL,
- AFR_NODE_WISE
+ AFR_NODE_WISE,
} afr_node_type;
typedef struct {
@@ -250,7 +362,7 @@ afr_sh_is_wise (int32_t *array, int i, int child_count)
static int
-afr_sh_all_nodes_innocent (afr_node_character *characters,
+afr_sh_all_nodes_innocent (afr_node_character *characters,
int child_count)
{
int i = 0;
@@ -286,10 +398,10 @@ afr_sh_wise_nodes_exist (afr_node_character *characters, int child_count)
/*
* The 'wisdom' of a wise node is 0 if any other wise node accuses it.
- * It is 1 if no other wise node accuses it.
+ * It is 1 if no other wise node accuses it.
* Only wise nodes with wisdom 1 are sources.
*
- * If no nodes with wisdom 1 exist, a split-brain has occured.
+ * If no nodes with wisdom 1 exist, a split-brain has occurred.
*/
static void
@@ -306,7 +418,7 @@ afr_sh_compute_wisdom (int32_t *pending_matrix[],
for (j = 0; j < child_count; j++) {
if ((characters[j].type == AFR_NODE_WISE)
&& pending_matrix[j][i]) {
-
+
characters[i].wisdom = 0;
}
}
@@ -316,7 +428,7 @@ afr_sh_compute_wisdom (int32_t *pending_matrix[],
static int
-afr_sh_wise_nodes_conflict (afr_node_character *characters,
+afr_sh_wise_nodes_conflict (afr_node_character *characters,
int child_count)
{
int i = 0;
@@ -337,13 +449,12 @@ afr_sh_wise_nodes_conflict (afr_node_character *characters,
static int
-afr_sh_mark_wisest_as_sources (int sources[],
- afr_node_character *characters,
+afr_sh_mark_wisest_as_sources (int sources[],
+ afr_node_character *characters,
int child_count)
{
int nsources = 0;
-
- int i = 0;
+ int i = 0;
for (i = 0; i < child_count; i++) {
if (characters[i].wisdom == 1) {
@@ -355,1108 +466,1778 @@ afr_sh_mark_wisest_as_sources (int sources[],
return nsources;
}
-
-static int
-afr_sh_mark_if_size_differs (afr_self_heal_t *sh, int child_count)
+static void
+afr_compute_witness_of_fools (int32_t *witnesses, int32_t **pending_matrix,
+ afr_node_character *characters,
+ int32_t child_count)
{
- int32_t ** pending_matrix;
- int i, j;
+ int i = 0;
+ int j = 0;
+ int witness = 0;
- int size_differs = 0;
-
- pending_matrix = sh->pending_matrix;
+ GF_ASSERT (witnesses);
+ GF_ASSERT (pending_matrix);
+ GF_ASSERT (characters);
+ GF_ASSERT (child_count > 0);
for (i = 0; i < child_count; i++) {
+ if (characters[i].type != AFR_NODE_FOOL)
+ continue;
+
+ witness = 0;
for (j = 0; j < child_count; j++) {
- if (!sh->buf)
- break;
+ if (i == j)
+ continue;
+ witness += pending_matrix[i][j];
+ }
+ witnesses[i] = witness;
+ }
+}
- if (SIZE_DIFFERS (&sh->buf[i], &sh->buf[j])
- && (pending_matrix[i][j] == 0)
- && (pending_matrix[j][i] == 0)) {
+static int32_t
+afr_find_biggest_witness_among_fools (int32_t *witnesses,
+ afr_node_character *characters,
+ int32_t child_count)
+{
+ int i = 0;
+ int biggest_witness = -1;
+ int biggest_witness_idx = -1;
+ int biggest_witness_cnt = -1;
- pending_matrix[i][j] = 1;
- pending_matrix[j][i] = 1;
+ GF_ASSERT (witnesses);
+ GF_ASSERT (characters);
+ GF_ASSERT (child_count > 0);
- size_differs = 1;
- }
- }
+ for (i = 0; i < child_count; i++) {
+ if (characters[i].type != AFR_NODE_FOOL)
+ continue;
+
+ if (biggest_witness < witnesses[i]) {
+ biggest_witness = witnesses[i];
+ biggest_witness_idx = i;
+ biggest_witness_cnt = 1;
+ continue;
+ }
+
+ if (biggest_witness == witnesses[i])
+ biggest_witness_cnt++;
}
- return size_differs;
+ if (biggest_witness_cnt != 1)
+ return -1;
+
+ return biggest_witness_idx;
}
-
-static int
-afr_sh_mark_biggest_fool_as_source (afr_self_heal_t *sh,
- afr_node_character *characters,
- int child_count)
+int
+afr_mark_fool_as_source_by_witness (int32_t *sources, int32_t *witnesses,
+ afr_node_character *characters,
+ int32_t child_count, int32_t witness)
{
- int i = 0;
- int biggest = 0;
+ int i = 0;
+ int nsources = 0;
- for (i = 0; i < child_count; i++) {
- if (characters[i].type == AFR_NODE_FOOL) {
- biggest = i;
- break;
- }
- }
+ GF_ASSERT (sources);
+ GF_ASSERT (witnesses);
+ GF_ASSERT (characters);
+ GF_ASSERT (child_count > 0);
for (i = 0; i < child_count; i++) {
if (characters[i].type != AFR_NODE_FOOL)
continue;
- if (!sh->buf)
- break;
-
- if (SIZE_GREATER (&sh->buf[i], &sh->buf[biggest])) {
- biggest = i;
+ if (witness == witnesses[i]) {
+ sources[i] = 1;
+ nsources++;
}
}
+ return nsources;
+}
+
+
+int
+afr_mark_fool_as_source_by_idx (int32_t *sources, int child_count, int idx)
+{
+ if (idx >= 0 && idx < child_count) {
+ sources[idx] = 1;
+ return 1;
+ }
+ return 0;
+}
+
+
+static int
+afr_find_largest_file_size (struct iatt *bufs, int32_t *success_children,
+ int child_count)
+{
+ int idx = -1;
+ int i = -1;
+ int child = -1;
+ uint64_t max_size = 0;
+ uint64_t min_size = 0;
+ int num_children = 0;
+
+ for (i = 0; i < child_count; i++) {
+ if (success_children[i] == -1)
+ break;
+
+ child = success_children[i];
+ if (bufs[child].ia_size > max_size) {
+ max_size = bufs[child].ia_size;
+ idx = child;
+ }
+
+ if ((num_children == 0) || (bufs[child].ia_size < min_size)) {
+ min_size = bufs[child].ia_size;
+ }
+
+ num_children++;
+ }
+
+ /* If sizes are same for all of them, finding sources will have to
+ * happen with pending changelog. So return -1
+ */
+ if ((num_children > 1) && (min_size == max_size))
+ return -1;
+ return idx;
+}
+
+
+static int
+afr_find_newest_file (struct iatt *bufs, int32_t *success_children,
+ int child_count)
+{
+ int idx = -1;
+ int i = -1;
+ int child = -1;
+ uint64_t max_ctime = 0;
+
+ for (i = 0; i < child_count; i++) {
+ if (success_children[i] == -1)
+ break;
- sh->sources[biggest] = 1;
+ child = success_children[i];
+ if (bufs[child].ia_ctime > max_ctime) {
+ max_ctime = bufs[child].ia_ctime;
+ idx = child;
+ }
+ }
- return 1;
+ return idx;
}
static int
-afr_sh_mark_biggest_as_source (afr_self_heal_t *sh, int child_count)
+afr_mark_biggest_of_fools_as_source (int32_t *sources, int32_t **pending_matrix,
+ afr_node_character *characters,
+ int32_t *success_children,
+ int child_count, struct iatt *bufs)
+{
+ int32_t biggest_witness = 0;
+ int nsources = 0;
+ int32_t *witnesses = NULL;
+
+ GF_ASSERT (child_count > 0);
+
+ biggest_witness = afr_find_largest_file_size (bufs, success_children,
+ child_count);
+ if (biggest_witness != -1)
+ goto found;
+
+ witnesses = GF_CALLOC (child_count, sizeof (*witnesses),
+ gf_afr_mt_int32_t);
+ if (NULL == witnesses) {
+ nsources = -1;
+ goto out;
+ }
+
+ afr_compute_witness_of_fools (witnesses, pending_matrix, characters,
+ child_count);
+ biggest_witness = afr_find_biggest_witness_among_fools (witnesses,
+ characters,
+ child_count);
+ if (biggest_witness != -1)
+ goto found;
+
+ biggest_witness = afr_find_newest_file (bufs, success_children,
+ child_count);
+
+found:
+ nsources = afr_mark_fool_as_source_by_idx (sources, child_count,
+ biggest_witness);
+out:
+ GF_FREE (witnesses);
+ return nsources;
+}
+
+int
+afr_mark_child_as_source_by_uid (int32_t *sources, struct iatt *bufs,
+ int32_t *success_children,
+ unsigned int child_count, uint32_t uid)
{
- int biggest = 0;
- int i;
+ int i = 0;
+ int nsources = 0;
+ int child = 0;
for (i = 0; i < child_count; i++) {
- if (!sh->buf)
+ if (-1 == success_children[i])
break;
- if (SIZE_GREATER (&sh->buf[i], &sh->buf[biggest])) {
- biggest = i;
+ child = success_children[i];
+ if (uid == bufs[child].ia_uid) {
+ sources[child] = 1;
+ nsources++;
}
}
+ return nsources;
+}
- sh->sources[biggest] = 1;
+int
+afr_get_child_with_lowest_uid (struct iatt *bufs, int32_t *success_children,
+ unsigned int child_count)
+{
+ int i = 0;
+ int smallest = -1;
+ int child = 0;
- return 1;
+ for (i = 0; i < child_count; i++) {
+ if (-1 == success_children[i])
+ break;
+ child = success_children[i];
+ if ((smallest == -1) ||
+ (bufs[child].ia_uid < bufs[smallest].ia_uid)) {
+ smallest = child;
+ }
+ }
+ return smallest;
}
-
static int
-afr_sh_mark_loweia_uid_as_source (afr_self_heal_t *sh, int child_count)
+afr_sh_mark_lowest_uid_as_source (struct iatt *bufs, int32_t *success_children,
+ int child_count, int32_t *sources)
{
- uid_t smallest = 0;
- int i;
+ int nsources = 0;
+ int smallest = 0;
- for (i = 0; i < child_count; i++) {
- if (!sh->buf)
+ smallest = afr_get_child_with_lowest_uid (bufs, success_children,
+ child_count);
+ if (smallest < 0) {
+ nsources = -1;
+ goto out;
+ }
+ nsources = afr_mark_child_as_source_by_uid (sources, bufs,
+ success_children, child_count,
+ bufs[smallest].ia_uid);
+out:
+ return nsources;
+}
+
+int
+afr_get_no_xattr_dir_read_child (xlator_t *this, int32_t *success_children,
+ struct iatt *bufs)
+{
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int child = -1;
+ int read_child = -1;
+
+ priv = this->private;
+ for (i = 0; i < priv->child_count; i++) {
+ child = success_children[i];
+ if (child < 0)
break;
+ if (read_child < 0)
+ read_child = child;
+ else if (bufs[read_child].ia_size < bufs[child].ia_size)
+ read_child = child;
+ }
+ return read_child;
+}
- if (sh->buf[i].ia_uid < sh->buf[smallest].ia_uid) {
- smallest = i;
+int
+afr_sh_mark_zero_size_file_as_sink (struct iatt *bufs, int32_t *success_children,
+ int child_count, int32_t *sources)
+{
+ int nsources = 0;
+ int i = 0;
+ int child = 0;
+ gf_boolean_t sink_exists = _gf_false;
+ gf_boolean_t source_exists = _gf_false;
+ int source = -1;
+
+ for (i = 0; i < child_count; i++) {
+ child = success_children[i];
+ if (child < 0)
+ break;
+ if (!bufs[child].ia_size) {
+ sink_exists = _gf_true;
+ continue;
+ }
+ if (!source_exists) {
+ source_exists = _gf_true;
+ source = child;
+ continue;
+ }
+ if (bufs[source].ia_size != bufs[child].ia_size) {
+ nsources = -1;
+ goto out;
}
}
+ if (!source_exists && !sink_exists) {
+ nsources = -1;
+ goto out;
+ }
+
+ if (!source_exists || !sink_exists)
+ goto out;
- sh->sources[smallest] = 1;
+ for (i = 0; i < child_count; i++) {
+ child = success_children[i];
+ if (child < 0)
+ break;
+ if (bufs[child].ia_size) {
+ sources[child] = 1;
+ nsources++;
+ }
+ }
+out:
+ return nsources;
+}
- return 1;
+char *
+afr_get_character_str (afr_node_type type)
+{
+ char *character = NULL;
+
+ switch (type) {
+ case AFR_NODE_INNOCENT:
+ character = "innocent";
+ break;
+ case AFR_NODE_FOOL:
+ character = "fool";
+ break;
+ case AFR_NODE_WISE:
+ character = "wise";
+ break;
+ default:
+ character = "invalid";
+ break;
+ }
+ return character;
}
+afr_node_type
+afr_find_child_character_type (int32_t *pending_row, int32_t child,
+ unsigned int child_count)
+{
+ afr_node_type type = AFR_NODE_INVALID;
+
+ GF_ASSERT ((child >= 0) && (child < child_count));
+
+ if (afr_sh_is_innocent (pending_row, child_count))
+ type = AFR_NODE_INNOCENT;
+ else if (afr_sh_is_fool (pending_row, child, child_count))
+ type = AFR_NODE_FOOL;
+ else if (afr_sh_is_wise (pending_row, child, child_count))
+ type = AFR_NODE_WISE;
+ return type;
+}
int
-afr_sh_mark_sources (afr_self_heal_t *sh, int child_count,
- afr_self_heal_type type)
+afr_build_sources (xlator_t *this, dict_t **xattr, struct iatt *bufs,
+ int32_t **pending_matrix, int32_t *sources,
+ int32_t *success_children, afr_transaction_type type,
+ int32_t *subvol_status, gf_boolean_t ignore_ignorant)
{
- int i = 0;
+ afr_private_t *priv = NULL;
+ afr_self_heal_type sh_type = AFR_SELF_HEAL_INVALID;
+ int nsources = -1;
+ unsigned char *ignorant_subvols = NULL;
+ unsigned int child_count = 0;
- int32_t ** pending_matrix;
- int * sources;
+ priv = this->private;
+ child_count = priv->child_count;
- int size_differs = 0;
+ if (afr_get_children_count (success_children, priv->child_count) == 0)
+ goto out;
- pending_matrix = sh->pending_matrix;
- sources = sh->sources;
+ if (!ignore_ignorant) {
+ ignorant_subvols = GF_CALLOC (sizeof (*ignorant_subvols),
+ child_count, gf_afr_mt_char);
+ if (NULL == ignorant_subvols)
+ goto out;
+ }
- int nsources = 0;
+ afr_build_pending_matrix (priv->pending_key, pending_matrix,
+ ignorant_subvols, xattr, type,
+ priv->child_count);
- /* stores the 'characters' (innocent, fool, wise) of the nodes */
- afr_node_character *
- characters = GF_CALLOC (sizeof (afr_node_character),
- child_count,
- gf_afr_mt_afr_node_character) ;
+ if (!ignore_ignorant)
+ afr_mark_ignorant_subvols_as_pending (pending_matrix,
+ ignorant_subvols,
+ priv->child_count);
+ sh_type = afr_self_heal_type_for_transaction (type);
+ if (AFR_SELF_HEAL_INVALID == sh_type)
+ goto out;
- /* start clean */
- for (i = 0; i < child_count; i++) {
- sources[i] = 0;
- }
-
- for (i = 0; i < child_count; i++) {
- if (afr_sh_is_innocent (pending_matrix[i], child_count)) {
- characters[i].type = AFR_NODE_INNOCENT;
+ afr_sh_print_pending_matrix (pending_matrix, this);
- } else if (afr_sh_is_fool (pending_matrix[i], i, child_count)) {
- characters[i].type = AFR_NODE_FOOL;
+ nsources = afr_mark_sources (this, sources, pending_matrix, bufs,
+ sh_type, success_children, subvol_status);
+out:
+ GF_FREE (ignorant_subvols);
+ return nsources;
+}
- } else if (afr_sh_is_wise (pending_matrix[i], i, child_count)) {
- characters[i].type = AFR_NODE_WISE;
+void
+afr_find_character_types (afr_node_character *characters,
+ int32_t **pending_matrix, int32_t *success_children,
+ unsigned int child_count)
+{
+ afr_node_type type = AFR_NODE_INVALID;
+ int child = 0;
+ int i = 0;
- } else {
- gf_log ("[module:replicate]", GF_LOG_ERROR,
- "Could not determine the state of subvolume %d!"
- " (This message should never appear."
- " Please file a bug report to "
- "<gluster-devel@nongnu.org>.)", i);
- }
+ for (i = 0; i < child_count; i++) {
+ child = success_children[i];
+ if (child == -1)
+ break;
+ type = afr_find_child_character_type (pending_matrix[child],
+ child, child_count);
+ characters[child].type = type;
}
+}
- if (type == AFR_SELF_HEAL_DATA) {
- size_differs = afr_sh_mark_if_size_differs (sh, child_count);
+void
+afr_mark_success_children_sources (int32_t *sources, int32_t *success_children,
+ unsigned int child_count)
+{
+ int i = 0;
+ for (i = 0; i < child_count; i++) {
+ if (success_children[i] == -1)
+ break;
+ sources[success_children[i]] = 1;
}
+}
+/**
+ * mark_sources: Mark all 'source' nodes and return number of source
+ * nodes found
+ *
+ * A node (a row in the pending matrix) belongs to one of
+ * three categories:
+ *
+ * M is the pending matrix.
+ *
+ * 'innocent' - M[i] is all zeroes
+ * 'fool' - M[i] has i'th element = 1 (self-reference)
+ * 'wise' - M[i] has i'th element = 0, others are 1 or 0.
+ *
+ * All 'innocent' nodes are sinks. If all nodes are innocent, no self-heal is
+ * needed.
+ *
+ * A 'wise' node can be a source. If two 'wise' nodes conflict, it is
+ * a split-brain. If one wise node refers to the other but the other doesn't
+ * refer back, the referrer is a source.
+ *
+ * All fools are sinks, unless there are no 'wise' nodes. In that case,
+ * one of the fools is made a source.
+ */
- if ((type == AFR_SELF_HEAL_METADATA)
- && afr_sh_all_nodes_innocent (characters, child_count)) {
+int
+afr_mark_sources (xlator_t *this, int32_t *sources, int32_t **pending_matrix,
+ struct iatt *bufs, afr_self_heal_type type,
+ int32_t *success_children, int32_t *subvol_status)
+{
+ /* stores the 'characters' (innocent, fool, wise) of the nodes */
+ afr_node_character *characters = NULL;
+ int nsources = -1;
+ unsigned int child_count = 0;
+ afr_private_t *priv = NULL;
- nsources = afr_sh_mark_loweia_uid_as_source (sh, child_count);
+ priv = this->private;
+ child_count = priv->child_count;
+ characters = GF_CALLOC (sizeof (afr_node_character),
+ child_count, gf_afr_mt_afr_node_character);
+ if (!characters)
goto out;
- }
+ this = THIS;
+
+ /* start clean */
+ memset (sources, 0, sizeof (*sources) * child_count);
+ nsources = 0;
+ afr_find_character_types (characters, pending_matrix, success_children,
+ child_count);
if (afr_sh_all_nodes_innocent (characters, child_count)) {
- if (size_differs) {
- nsources = afr_sh_mark_biggest_as_source (sh,
- child_count);
+ switch (type) {
+ case AFR_SELF_HEAL_METADATA:
+ nsources = afr_sh_mark_lowest_uid_as_source (bufs,
+ success_children,
+ child_count,
+ sources);
+ break;
+ case AFR_SELF_HEAL_DATA:
+ nsources = afr_sh_mark_zero_size_file_as_sink (bufs,
+ success_children,
+ child_count,
+ sources);
+ if ((nsources < 0) && subvol_status)
+ *subvol_status |= SPLIT_BRAIN;
+ break;
+ default:
+ break;
}
+ goto out;
+ }
- } else if (afr_sh_wise_nodes_exist (characters, child_count)) {
+ if (afr_sh_wise_nodes_exist (characters, child_count)) {
afr_sh_compute_wisdom (pending_matrix, characters, child_count);
if (afr_sh_wise_nodes_conflict (characters, child_count)) {
- /* split-brain */
-
+ if (subvol_status)
+ *subvol_status |= SPLIT_BRAIN;
nsources = -1;
- goto out;
-
} else {
- nsources = afr_sh_mark_wisest_as_sources (sources,
+ nsources = afr_sh_mark_wisest_as_sources (sources,
characters,
child_count);
}
} else {
- nsources = afr_sh_mark_biggest_fool_as_source (sh, characters,
- child_count);
+ if (subvol_status)
+ *subvol_status |= ALL_FOOLS;
+ nsources = afr_mark_biggest_of_fools_as_source (sources,
+ pending_matrix,
+ characters,
+ success_children,
+ child_count, bufs);
}
out:
+ if (nsources == 0)
+ afr_mark_success_children_sources (sources, success_children,
+ child_count);
GF_FREE (characters);
- return nsources;
+ gf_log (this->name, GF_LOG_DEBUG, "Number of sources: %d", nsources);
+ return nsources;
}
-
void
afr_sh_pending_to_delta (afr_private_t *priv, dict_t **xattr,
- int32_t *delta_matrix[], int success[],
+ int32_t *delta_matrix[], unsigned char success[],
int child_count, afr_transaction_type type)
{
- int i = 0;
- int j = 0;
- int k = 0;
-
- /* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
- int32_t pending[3];
- void * pending_raw = NULL;
- int ret = 0;
-
- /* start clean */
- for (i = 0; i < child_count; i++) {
- for (j = 0; j < child_count; j++) {
- delta_matrix[i][j] = 0;
- }
- }
-
- for (i = 0; i < child_count; i++) {
- pending_raw = NULL;
-
- for (j = 0; j < child_count; j++) {
- ret = dict_get_ptr (xattr[i], priv->pending_key[j],
- &pending_raw);
- if (ret < 0)
- gf_log ("afr_sh_pending_to_delta",
- GF_LOG_WARNING,
- "Unable to get dict value.");
+ int tgt = 0;
+ int src = 0;
+ int value = 0;
- if (!success[j])
- continue;
-
- k = afr_index_for_transaction_type (type);
+ afr_build_pending_matrix (priv->pending_key, delta_matrix, NULL,
+ xattr, type, priv->child_count);
- if (pending_raw) {
- memcpy (pending, pending_raw, sizeof(pending));
- delta_matrix[i][j] = -(ntoh32 (pending[k]));
- } else {
- delta_matrix[i][j] = 0;
+ /*
+ * The algorithm here has two parts. First, for each subvol indexed
+ * as tgt, we try to figure out what count everyone should have for it.
+ * If the self-heal succeeded, that's easy; the value is zero.
+ * Otherwise, the value is the maximum of the succeeding nodes' counts.
+ * Once we know the value, we loop through (possibly for a second time)
+ * setting each count to the difference so that when we're done all
+ * succeeding nodes will have the same count for tgt.
+ */
+ for (tgt = 0; tgt < priv->child_count; ++tgt) {
+ value = 0;
+ if (!success[tgt]) {
+ /* Find the maximum. */
+ for (src = 0; src < priv->child_count; ++src) {
+ if (!success[src]) {
+ continue;
+ }
+ if (delta_matrix[src][tgt] > value) {
+ value = delta_matrix[src][tgt];
+ }
}
-
}
- }
+ /* Force everyone who succeeded to the chosen value. */
+ for (src = 0; src < priv->child_count; ++src) {
+ if (success[src]) {
+ delta_matrix[src][tgt] = value
+ - delta_matrix[src][tgt];
+ }
+ else {
+ delta_matrix[src][tgt] = 0;
+ }
+ }
+ }
}
int
-afr_sh_delta_to_xattr (afr_private_t *priv,
+afr_sh_delta_to_xattr (xlator_t *this,
int32_t *delta_matrix[], dict_t *xattr[],
- int child_count, afr_transaction_type type)
+ int child_count, afr_transaction_type type)
{
- int i = 0;
- int j = 0;
- int k = 0;
-
- int ret = 0;
+ int i = 0;
+ int j = 0;
+ int k = 0;
+ int ret = 0;
+ int32_t *pending = NULL;
+ int32_t *local_pending = NULL;
+ afr_private_t *priv = NULL;
- int32_t *pending = 0;
-
- for (i = 0; i < child_count; i++) {
- if (!xattr[i])
- continue;
+ priv = this->private;
+ for (i = 0; i < child_count; i++) {
+ if (!xattr[i])
+ continue;
- for (j = 0; j < child_count; j++) {
+ local_pending = NULL;
+ for (j = 0; j < child_count; j++) {
pending = GF_CALLOC (sizeof (int32_t), 3,
gf_afr_mt_int32_t);
+
+ if (!pending) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "failed to allocate pending entry "
+ "for %s[%d] on %s",
+ priv->pending_key[j], type,
+ priv->children[i]->name);
+ continue;
+ }
/* 3 = data+metadata+entry */
k = afr_index_for_transaction_type (type);
- pending[k] = hton32 (delta_matrix[i][j]);
+ pending[k] = hton32 (delta_matrix[i][j]);
+ if (j == i) {
+ local_pending = pending;
+ continue;
+ }
ret = dict_set_bin (xattr[i], priv->pending_key[j],
pending,
- 3 * sizeof (int32_t));
- if (ret < 0)
- gf_log ("afr_sh_delta_to_xattr",
- GF_LOG_WARNING,
+ AFR_NUM_CHANGE_LOGS * sizeof (int32_t));
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_WARNING,
"Unable to set dict value.");
- }
- }
- return 0;
+ GF_FREE (pending);
+ }
+ }
+ if (local_pending) {
+ ret = dict_set_bin (xattr[i], priv->pending_key[i],
+ local_pending,
+ AFR_NUM_CHANGE_LOGS * sizeof (int32_t));
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "Unable to set dict value.");
+ GF_FREE (local_pending);
+ }
+ }
+ }
+ return 0;
}
int
-afr_sh_has_metadata_pending (dict_t *xattr, int child_count, xlator_t *this)
+afr_sh_missing_entries_done (call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- /* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
- int32_t pending[3];
- void *pending_raw = NULL;
-
- int ret = -1;
- int i = 0;
- int j = 0;
-
- priv = this->private;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_get_ptr (xattr, priv->pending_key[i],
- &pending_raw);
+ local = frame->local;
+ sh = &local->self_heal;
- if (ret != 0)
- return 0;
+ afr_sh_reset (frame, this);
- memcpy (pending, pending_raw, sizeof(pending));
- j = afr_index_for_transaction_type (AFR_METADATA_TRANSACTION);
+ if (local->unhealable) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "split brain found, aborting selfheal of %s",
+ local->loc.path);
+ }
- if (pending[j])
- return 1;
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+ sh->completion_cbk (frame, this);
+ } else {
+ gf_log (this->name, GF_LOG_TRACE,
+ "proceeding to metadata check on %s",
+ local->loc.path);
+ afr_self_heal_metadata (frame, this);
}
- return 0;
+ return 0;
}
-int
-afr_sh_has_data_pending (dict_t *xattr, int child_count, xlator_t *this)
+static int
+afr_sh_missing_entries_finish (call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- /* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
- int32_t pending[3];
- void *pending_raw = NULL;
-
- int ret = -1;
- int i = 0;
- int j = 0;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_get_ptr (xattr, priv->pending_key[i],
- &pending_raw);
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
- if (ret != 0)
- return 0;
-
- memcpy (pending, pending_raw, sizeof(pending));
- j = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
+ local = frame->local;
+ int_lock = &local->internal_lock;
- if (pending[j])
- return 1;
- }
+ int_lock->lock_cbk = afr_sh_missing_entries_done;
+ afr_unlock (frame, this);
- return 0;
+ return 0;
}
-
int
-afr_sh_has_entry_pending (dict_t *xattr, int child_count, xlator_t *this)
+afr_sh_common_create (afr_self_heal_t *sh, unsigned int child_count)
{
- afr_private_t *priv = NULL;
- /* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
- int32_t pending[3];
- void *pending_raw = NULL;
-
- int ret = -1;
- int i = 0;
- int j = 0;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_get_ptr (xattr, priv->pending_key[i],
- &pending_raw);
+ int ret = -ENOMEM;
+ sh->buf = GF_CALLOC (child_count, sizeof (*sh->buf),
+ gf_afr_mt_iatt);
+ if (!sh->buf)
+ goto out;
+ sh->parentbufs = GF_CALLOC (child_count, sizeof (*sh->parentbufs),
+ gf_afr_mt_iatt);
+ if (!sh->parentbufs)
+ goto out;
+ sh->child_errno = GF_CALLOC (child_count, sizeof (*sh->child_errno),
+ gf_afr_mt_int);
+ if (!sh->child_errno)
+ goto out;
+ sh->success_children = afr_children_create (child_count);
+ if (!sh->success_children)
+ goto out;
+ sh->fresh_children = afr_children_create (child_count);
+ if (!sh->fresh_children)
+ goto out;
+ sh->xattr = GF_CALLOC (child_count, sizeof (*sh->xattr),
+ gf_afr_mt_dict_t);
+ if (!sh->xattr)
+ goto out;
+ ret = 0;
+out:
+ return ret;
+}
- if (ret != 0)
- return 0;
-
- memcpy (pending, pending_raw, sizeof(pending));
- j = afr_index_for_transaction_type (AFR_ENTRY_TRANSACTION);
+void
+afr_sh_common_lookup_resp_handler (call_frame_t *frame, void *cookie,
+ xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ inode_t *inode, struct iatt *buf,
+ dict_t *xattr, struct iatt *postparent,
+ loc_t *loc)
+{
+ int child_index = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_self_heal_t *sh = NULL;
- if (pending[j])
- return 1;
+ local = frame->local;
+ priv = this->private;
+ sh = &local->self_heal;
+ child_index = (long) cookie;
+
+ LOCK (&frame->lock);
+ {
+ if (op_ret == 0) {
+ sh->buf[child_index] = *buf;
+ sh->parentbufs[child_index] = *postparent;
+ sh->success_children[sh->success_count] = child_index;
+ sh->success_count++;
+ sh->xattr[child_index] = dict_ref (xattr);
+ } else {
+ gf_log (this->name, GF_LOG_DEBUG, "path %s on subvolume"
+ " %s => -1 (%s)", loc->path,
+ priv->children[child_index]->name,
+ strerror (op_errno));
+ local->self_heal.child_errno[child_index] = op_errno;
+ }
}
-
- return 0;
+ UNLOCK (&frame->lock);
+ return;
}
-
-/**
- * is_matrix_zero - return true if pending matrix is all zeroes
- */
-
-int
-afr_sh_is_matrix_zero (int32_t *pending_matrix[], int child_count)
+gf_boolean_t
+afr_valid_ia_type (ia_type_t ia_type)
{
- int i, j;
-
- for (i = 0; i < child_count; i++)
- for (j = 0; j < child_count; j++)
- if (pending_matrix[i][j])
- return 0;
- return 1;
+ switch (ia_type) {
+ case IA_IFSOCK:
+ case IA_IFREG:
+ case IA_IFBLK:
+ case IA_IFCHR:
+ case IA_IFIFO:
+ case IA_IFLNK:
+ case IA_IFDIR:
+ return _gf_true;
+ default:
+ return _gf_false;
+ }
+ return _gf_false;
}
-
int
-afr_sh_missing_entries_done (call_frame_t *frame, xlator_t *this)
+afr_impunge_frame_create (call_frame_t *frame, xlator_t *this,
+ int active_source, call_frame_t **impunge_frame)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
+ afr_local_t *local = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ int32_t op_errno = 0;
+ afr_private_t *priv = NULL;
+ int ret = 0;
+ call_frame_t *new_frame = NULL;
+
+ op_errno = ENOMEM;
+ priv = this->private;
+ new_frame = copy_frame (frame);
+ if (!new_frame) {
+ goto out;
+ }
+
+ AFR_LOCAL_ALLOC_OR_GOTO (impunge_local, out);
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ local = frame->local;
+ new_frame->local = impunge_local;
+ impunge_sh = &impunge_local->self_heal;
+ impunge_sh->sh_frame = frame;
+ impunge_sh->active_source = active_source;
+ impunge_local->child_up = memdup (local->child_up,
+ sizeof (*local->child_up) *
+ priv->child_count);
+ if (!impunge_local->child_up)
+ goto out;
-// memset (sh->child_errno, 0, sizeof (int) * priv->child_count);
- memset (sh->buf, 0, sizeof (struct iatt) * priv->child_count);
+ impunge_local->pending = afr_matrix_create (priv->child_count,
+ AFR_NUM_CHANGE_LOGS);
+ if (!impunge_local->pending)
+ goto out;
- for (i = 0; i < priv->child_count; i++) {
- sh->locked_nodes[i] = 0;
+ ret = afr_sh_common_create (impunge_sh, priv->child_count);
+ if (ret) {
+ op_errno = -ret;
+ goto out;
}
+ op_errno = 0;
+ *impunge_frame = new_frame;
+out:
+ if (op_errno && new_frame)
+ AFR_STACK_DESTROY (new_frame);
+ return -op_errno;
+}
- for (i = 0; i < priv->child_count; i++) {
- if (sh->xattr[i])
- dict_unref (sh->xattr[i]);
- sh->xattr[i] = NULL;
- }
+void
+afr_sh_missing_entry_call_impunge_recreate (call_frame_t *frame, xlator_t *this,
+ struct iatt *buf,
+ struct iatt *postparent,
+ afr_impunge_done_cbk_t impunge_done)
+{
+ call_frame_t *impunge_frame = NULL;
+ afr_local_t *local = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ int ret = 0;
+ unsigned int enoent_count = 0;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int32_t op_errno = 0;
- if (local->govinda_gOvinda) {
- gf_log (this->name, GF_LOG_TRACE,
- "aborting selfheal of %s",
- local->loc.path);
- sh->completion_cbk (frame, this);
- } else {
- gf_log (this->name, GF_LOG_TRACE,
- "proceeding to metadata check on %s",
- local->loc.path);
- afr_self_heal_metadata (frame, this);
- }
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- return 0;
+ enoent_count = afr_errno_count (NULL, sh->child_errno,
+ priv->child_count, ENOENT);
+ if (!enoent_count) {
+ gf_log (this->name, GF_LOG_INFO,
+ "no missing files - %s. proceeding to metadata check",
+ local->loc.path);
+ goto out;
+ }
+ sh->impunge_done = impunge_done;
+ ret = afr_impunge_frame_create (frame, this, sh->source, &impunge_frame);
+ if (ret)
+ goto out;
+ impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
+ loc_copy (&impunge_local->loc, &local->loc);
+ ret = afr_build_parent_loc (&impunge_sh->parent_loc,
+ &impunge_local->loc, &op_errno);
+ if (ret) {
+ ret = -op_errno;
+ goto out;
+ }
+ impunge_local->call_count = enoent_count;
+ impunge_sh->entrybuf = sh->buf[sh->source];
+ impunge_sh->parentbuf = sh->parentbufs[sh->source];
+ for (i = 0; i < priv->child_count; i++) {
+ if (!impunge_local->child_up[i]) {
+ impunge_sh->child_errno[i] = ENOTCONN;
+ continue;
+ }
+ if (sh->child_errno[i] != ENOENT) {
+ impunge_sh->child_errno[i] = EEXIST;
+ continue;
+ }
+ }
+ for (i = 0; i < priv->child_count; i++) {
+ if (sh->child_errno[i] != ENOENT)
+ continue;
+ afr_sh_entry_impunge_create (impunge_frame, this, i);
+ enoent_count--;
+ }
+ GF_ASSERT (!enoent_count);
+ return;
+out:
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "impunge of %s failed, "
+ "reason: %s", local->loc.path, strerror (-ret));
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ }
+ afr_sh_missing_entries_finish (frame, this);
}
-
-static int
-sh_missing_entries_finish (call_frame_t *frame, xlator_t *this)
+int
+afr_sh_create_entry_cbk (call_frame_t *frame, xlator_t *this,
+ int32_t op_ret, int32_t op_errno)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
-
- local = frame->local;
- int_lock = &local->internal_lock;
-
- int_lock->lock_cbk = afr_sh_missing_entries_done;
- afr_unlock (frame, this);
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
- return 0;
+ local = frame->local;
+ sh = &local->self_heal;
+ if (op_ret < 0)
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_missing_entries_finish (frame, this);
+ return 0;
}
-
static int
-sh_destroy_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int op_errno,
- struct iatt *preop, struct iatt *postop)
+sh_missing_entries_create (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
-
- loc_t *parent_loc = cookie;
-
- int call_count = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ int type = 0;
+ struct iatt *buf = NULL;
+ struct iatt *postparent = NULL;
local = frame->local;
+ sh = &local->self_heal;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "setattr on %s failed: %s",
- local->loc.path, strerror (op_errno));
- }
+ buf = &sh->buf[sh->source];
+ postparent = &sh->parentbufs[sh->source];
- if (parent_loc) {
- loc_wipe (parent_loc);
- GF_FREE (parent_loc);
+ type = buf->ia_type;
+ if (!afr_valid_ia_type (type)) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "%s: unknown file type: 0%o", local->loc.path, type);
+ afr_set_local_for_unhealable (local);
+ afr_sh_missing_entries_finish (frame, this);
+ goto out;
}
- call_count = afr_frame_return (frame);
-
- if (call_count == 0) {
- STACK_DESTROY (frame->root);
- }
-
- return 0;
+ afr_sh_missing_entry_call_impunge_recreate (frame, this,
+ buf, postparent,
+ afr_sh_create_entry_cbk);
+out:
+ return 0;
}
+void
+afr_sh_missing_entries_lookup_done (call_frame_t *frame, xlator_t *this,
+ int32_t op_ret, int32_t op_errno)
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ ia_type_t ia_type = IA_INVAL;
+ int32_t nsources = 0;
+ loc_t *loc = NULL;
+ int32_t subvol_status = 0;
+ afr_transaction_type txn_type = AFR_DATA_TRANSACTION;
+ gf_boolean_t split_brain = _gf_false;
+ int read_child = -1;
-static int
-sh_missing_entries_newentry_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf,
- struct iatt *preparent,
- struct iatt *postparent)
-{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- call_frame_t *setattr_frame = NULL;
- int call_count = 0;
- int child_index = 0;
-
- loc_t *parent_loc = NULL;
-
- struct iatt stbuf;
- int32_t valid;
-
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
-
- child_index = (long) cookie;
-
- stbuf.ia_atime = sh->buf[sh->source].ia_atime;
- stbuf.ia_atime_nsec = sh->buf[sh->source].ia_atime_nsec;
- stbuf.ia_mtime = sh->buf[sh->source].ia_mtime;
- stbuf.ia_mtime_nsec = sh->buf[sh->source].ia_mtime_nsec;
-
- stbuf.ia_uid = sh->buf[sh->source].ia_uid;
- stbuf.ia_gid = sh->buf[sh->source].ia_gid;
-
- valid = GF_SET_ATTR_UID | GF_SET_ATTR_GID |
- GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
-
- if (op_ret == 0) {
- setattr_frame = copy_frame (frame);
-
- setattr_frame->local = GF_CALLOC (1, sizeof (afr_local_t),
- gf_afr_mt_afr_local_t);
-
- ((afr_local_t *)setattr_frame->local)->call_count = 2;
-
- gf_log (this->name, GF_LOG_TRACE,
- "setattr (%s) on subvolume %s",
- local->loc.path, priv->children[child_index]->name);
-
- STACK_WIND_COOKIE (setattr_frame, sh_destroy_cbk,
- (void *) (long) 0,
- priv->children[child_index],
- priv->children[child_index]->fops->setattr,
- &local->loc, &stbuf, valid);
-
- valid = GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
- parent_loc = GF_CALLOC (1, sizeof (*parent_loc),
- gf_afr_mt_loc_t);
- afr_build_parent_loc (parent_loc, &local->loc);
-
- STACK_WIND_COOKIE (setattr_frame, sh_destroy_cbk,
- (void *) (long) parent_loc,
- priv->children[child_index],
- priv->children[child_index]->fops->setattr,
- parent_loc, &sh->parentbuf, valid);
- }
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+ loc = &local->loc;
- call_count = afr_frame_return (frame);
+ if (op_ret < 0) {
+ if (op_errno == EIO) {
+ afr_set_local_for_unhealable (local);
+ }
+ // EIO can happen if finding the fresh parent dir failed
+ goto out;
+ }
- if (call_count == 0) {
- sh_missing_entries_finish (frame, this);
- }
+ //now No chance for the ia_type to conflict
+ ia_type = sh->buf[sh->success_children[0]].ia_type;
+ txn_type = afr_transaction_type_get (ia_type);
+ nsources = afr_build_sources (this, sh->xattr, sh->buf,
+ sh->pending_matrix, sh->sources,
+ sh->success_children, txn_type,
+ &subvol_status, _gf_false);
+ if (nsources < 0) {
+ gf_log (this->name, GF_LOG_INFO, "No sources for dir of %s,"
+ " in missing entry self-heal, continuing with the rest"
+ " of the self-heals", local->loc.path);
+ if (subvol_status & SPLIT_BRAIN) {
+ split_brain = _gf_true;
+ switch (txn_type) {
+ case AFR_DATA_TRANSACTION:
+ nsources = 1;
+ sh->sources[sh->success_children[0]] = 1;
+ break;
+ case AFR_ENTRY_TRANSACTION:
+ read_child = afr_get_no_xattr_dir_read_child
+ (this,
+ sh->success_children,
+ sh->buf);
+ sh->sources[read_child] = 1;
+ nsources = 1;
+ break;
+ default:
+ op_errno = EIO;
+ goto out;
+ }
+ } else {
+ op_errno = EIO;
+ goto out;
+ }
+ }
- return 0;
-}
+ afr_get_fresh_children (sh->success_children, sh->sources,
+ sh->fresh_children, priv->child_count);
+ sh->source = sh->fresh_children[0];
+ if (sh->source == -1) {
+ gf_log (this->name, GF_LOG_DEBUG, "No active sources found.");
+ op_errno = EIO;
+ goto out;
+ }
+ if (sh->gfid_sh_success_cbk)
+ sh->gfid_sh_success_cbk (frame, this);
+ sh->type = sh->buf[sh->source].ia_type;
+ if (uuid_is_null (loc->inode->gfid))
+ uuid_copy (loc->gfid, sh->buf[sh->source].ia_gfid);
+ if (split_brain) {
+ afr_sh_missing_entries_finish (frame, this);
+ } else {
+ sh_missing_entries_create (frame, this);
+ }
+ return;
+out:
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_set_error (sh, op_errno);
+ afr_sh_missing_entries_finish (frame, this);
+ return;
+}
static int
-sh_missing_entries_mknod (call_frame_t *frame, xlator_t *this)
+afr_sh_common_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xattr,
+ struct iatt *postparent)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- int ret = 0;
- int enoent_count = 0;
- int call_count = 0;
- mode_t st_mode = 0;
- dev_t ia_dev = 0;
- dict_t *dict = NULL;
+ int call_count = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++)
- if (sh->child_errno[i] == ENOENT)
- enoent_count++;
+ afr_sh_common_lookup_resp_handler (frame, cookie, this, op_ret,
+ op_errno, inode, buf, xattr,
+ postparent, &sh->lookup_loc);
+ call_count = afr_frame_return (frame);
- call_count = enoent_count;
- local->call_count = call_count;
+ if (call_count)
+ goto out;
+ op_ret = -1;
+ if (!sh->success_count) {
+ op_errno = afr_resultant_errno_get (NULL, sh->child_errno,
+ priv->child_count);
+ gf_log (this->name, GF_LOG_ERROR, "Failed to lookup %s, "
+ "reason %s", sh->lookup_loc.path,
+ strerror (op_errno));
+ goto done;
+ }
- st_mode = st_mode_from_ia (sh->buf[sh->source].ia_prot,
- sh->buf[sh->source].ia_type);
- ia_dev = sh->buf[sh->source].ia_dev;
+ if ((sh->lookup_flags & AFR_LOOKUP_FAIL_CONFLICTS) &&
+ (afr_conflicting_iattrs (sh->buf, sh->success_children,
+ priv->child_count,
+ sh->lookup_loc.path, this->name))) {
+ op_errno = EIO;
+ gf_log (this->name, GF_LOG_ERROR, "Conflicting entries "
+ "for %s", sh->lookup_loc.path);
+ goto done;
+ }
- gf_log (this->name, GF_LOG_TRACE,
- "mknod %s mode 0%o on %d subvolumes",
- local->loc.path, st_mode, enoent_count);
+ if ((sh->lookup_flags & AFR_LOOKUP_FAIL_MISSING_GFIDS) &&
+ (afr_gfid_missing_count (this->name, sh->success_children,
+ sh->buf, priv->child_count,
+ sh->lookup_loc.path))) {
+ op_errno = ENODATA;
+ gf_log (this->name, GF_LOG_ERROR, "Missing Gfids "
+ "for %s", sh->lookup_loc.path);
+ goto done;
+ }
+ op_ret = 0;
- dict = dict_new ();
- if (!dict)
- gf_log (this->name, GF_LOG_ERROR, "out of memory");
+done:
+ sh->lookup_done (frame, this, op_ret, op_errno);
+out:
+ return 0;
+}
- ret = afr_set_dict_gfid (dict, sh->buf[sh->source].ia_gfid);
- if (ret)
- gf_log (this->name, GF_LOG_DEBUG, "gfid set failed");
-
- for (i = 0; i < priv->child_count; i++) {
- if (sh->child_errno[i] == ENOENT) {
- STACK_WIND_COOKIE (frame,
- sh_missing_entries_newentry_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->mknod,
- &local->loc, st_mode, ia_dev, dict);
- if (!--call_count)
- break;
- }
- }
+int
+afr_sh_remove_entry_cbk (call_frame_t *frame, xlator_t *this, int child,
+ int32_t op_ret, int32_t op_errno)
+{
+ int call_count = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
- if (dict)
- dict_unref (dict);
+ local = frame->local;
+ sh = &local->self_heal;
- return 0;
+ GF_ASSERT (sh->post_remove_call);
+ if ((op_ret == -1) && (op_errno != ENOENT)) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "purge entry %s failed, on child %d reason, %s",
+ local->loc.path, child, strerror (op_errno));
+ LOCK (&frame->lock);
+ {
+ afr_sh_set_error (sh, EIO);
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ }
+ UNLOCK (&frame->lock);
+ }
+ call_count = afr_frame_return (frame);
+ if (call_count == 0)
+ sh->post_remove_call (frame, this);
+ return 0;
}
-
-static int
-sh_missing_entries_mkdir (call_frame_t *frame, xlator_t *this)
+void
+afr_sh_call_entry_expunge_remove (call_frame_t *frame, xlator_t *this,
+ int child_index, struct iatt *buf,
+ struct iatt *parentbuf,
+ afr_expunge_done_cbk_t expunge_done)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- dict_t *dict = NULL;
- int i = 0;
- int ret = 0;
- int enoent_count = 0;
- int call_count = 0;
- mode_t st_mode = 0;
-
+ call_frame_t *expunge_frame = NULL;
+ afr_local_t *local = NULL;
+ afr_local_t *expunge_local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_self_heal_t *expunge_sh = NULL;
+ int32_t op_errno = 0;
+ int ret = 0;
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ expunge_frame = copy_frame (frame);
+ if (!expunge_frame) {
+ goto out;
+ }
- for (i = 0; i < priv->child_count; i++)
- if (sh->child_errno[i] == ENOENT)
- enoent_count++;
+ AFR_LOCAL_ALLOC_OR_GOTO (expunge_local, out);
- call_count = enoent_count;
- local->call_count = call_count;
+ local = frame->local;
+ sh = &local->self_heal;
+ expunge_frame->local = expunge_local;
+ expunge_sh = &expunge_local->self_heal;
+ expunge_sh->sh_frame = frame;
+ loc_copy (&expunge_local->loc, &local->loc);
+ ret = afr_build_parent_loc (&expunge_sh->parent_loc,
+ &expunge_local->loc, &op_errno);
+ if (ret) {
+ ret = -op_errno;
+ goto out;
+ }
+ sh->expunge_done = expunge_done;
+ afr_sh_entry_expunge_remove (expunge_frame, this, child_index, buf,
+ parentbuf);
+ return;
+out:
+ gf_log (this->name, GF_LOG_ERROR, "Expunge of %s failed, reason: %s",
+ local->loc.path, strerror (op_errno));
+ expunge_done (frame, this, child_index, -1, op_errno);
+}
- st_mode = st_mode_from_ia (sh->buf[sh->source].ia_prot,
- sh->buf[sh->source].ia_type);
+void
+afr_sh_remove_stale_lookup_info (afr_self_heal_t *sh, int32_t *success_children,
+ int32_t *fresh_children,
+ unsigned int child_count)
+{
+ int i = 0;
- dict = dict_new ();
- if (!dict) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- sh_missing_entries_finish (frame, this);
- return 0;
+ for (i = 0; i < child_count; i++) {
+ if (afr_is_child_present (success_children, child_count, i) &&
+ !afr_is_child_present (fresh_children, child_count, i)) {
+ sh->child_errno[i] = ENOENT;
+ GF_ASSERT (sh->xattr[i]);
+ dict_unref (sh->xattr[i]);
+ sh->xattr[i] = NULL;
+ }
}
+}
- ret = afr_set_dict_gfid (dict, sh->buf[sh->source].ia_gfid);
- if (ret)
- gf_log (this->name, GF_LOG_DEBUG,
- "inode gfid set failed");
+int
+afr_sh_purge_stale_entries_done (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- gf_log (this->name, GF_LOG_TRACE,
- "mkdir %s mode 0%o on %d subvolumes",
- local->loc.path, st_mode, enoent_count);
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+ afr_sh_missing_entries_finish (frame, this);
+ } else {
+ if (afr_gfid_missing_count (this->name, sh->fresh_children,
+ sh->buf, priv->child_count,
+ local->loc.path)) {
+ afr_sh_common_lookup (frame, this, &local->loc,
+ afr_sh_missing_entries_lookup_done,
+ sh->sh_gfid_req,
+ AFR_LOOKUP_FAIL_CONFLICTS|
+ AFR_LOOKUP_FAIL_MISSING_GFIDS,
+ NULL);
+ } else {
+ //No need to set gfid so goto missing entries lookup done
+ //Behave as if you have done the lookup
+ afr_sh_remove_stale_lookup_info (sh,
+ sh->success_children,
+ sh->fresh_children,
+ priv->child_count);
+ afr_children_copy (sh->success_children,
+ sh->fresh_children,
+ priv->child_count);
+ afr_sh_missing_entries_lookup_done (frame, this, 0, 0);
+ }
+ }
+ return 0;
+}
- for (i = 0; i < priv->child_count; i++) {
- if (sh->child_errno[i] == ENOENT) {
- if (!strcmp (local->loc.path, "/")) {
- /* We shouldn't try to create "/" */
+gf_boolean_t
+afr_sh_purge_entry_condition (afr_local_t *local, afr_private_t *priv,
+ int child)
+{
+ afr_self_heal_t *sh = NULL;
- sh_missing_entries_finish (frame, this);
+ sh = &local->self_heal;
- return 0;
- } else {
- STACK_WIND_COOKIE (frame,
- sh_missing_entries_newentry_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->mkdir,
- &local->loc, st_mode, dict);
- if (!--call_count)
- break;
- }
- }
- }
-
- if (dict)
- dict_unref (dict);
+ if (local->child_up[child] &&
+ (!afr_is_child_present (sh->fresh_parent_dirs, priv->child_count,
+ child))
+ && (sh->child_errno[child] != ENOENT))
+ return _gf_true;
- return 0;
+ return _gf_false;
}
+gf_boolean_t
+afr_sh_purge_stale_entry_condition (afr_local_t *local, afr_private_t *priv,
+ int child)
+{
+ afr_self_heal_t *sh = NULL;
-static int
-sh_missing_entries_symlink (call_frame_t *frame, xlator_t *this,
- const char *link, struct iatt *buf)
-{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- dict_t *dict = NULL;
- int i = 0;
- int ret = 0;
- int enoent_count = 0;
- int call_count = 0;
+ sh = &local->self_heal;
+ if (local->child_up[child] &&
+ (!afr_is_child_present (sh->fresh_children, priv->child_count,
+ child))
+ && (sh->child_errno[child] != ENOENT))
+ return _gf_true;
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ return _gf_false;
+}
- for (i = 0; i < priv->child_count; i++)
- if (sh->child_errno[i] == ENOENT)
- enoent_count++;
+void
+afr_sh_purge_entry_common (call_frame_t *frame, xlator_t *this,
+ gf_boolean_t purge_condition (afr_local_t *local,
+ afr_private_t *priv,
+ int child))
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_self_heal_t *sh = NULL;
+ int i = 0;
+ int call_count = 0;
- call_count = enoent_count;
- local->call_count = call_count;
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- dict = dict_new ();
- if (!dict) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- sh_missing_entries_finish (frame, this);
- return 0;
+ for (i = 0; i < priv->child_count; i++) {
+ if (purge_condition (local, priv, i))
+ call_count++;
}
- ret = afr_set_dict_gfid (dict, buf->ia_gfid);
- if (ret)
- gf_log (this->name, GF_LOG_DEBUG,
- "dict gfid set failed");
-
- gf_log (this->name, GF_LOG_TRACE,
- "symlink %s -> %s on %d subvolumes",
- local->loc.path, link, enoent_count);
-
- for (i = 0; i < priv->child_count; i++) {
- if (sh->child_errno[i] == ENOENT) {
- STACK_WIND_COOKIE (frame,
- sh_missing_entries_newentry_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->symlink,
- link, &local->loc, dict);
- if (!--call_count)
- break;
- }
- }
+ if (call_count == 0) {
+ sh->post_remove_call (frame, this);
+ goto out;
+ }
- return 0;
+ local->call_count = call_count;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!purge_condition (local, priv, i))
+ continue;
+ gf_log (this->name, GF_LOG_INFO, "purging the stale entry %s "
+ "on %s", local->loc.path, priv->children[i]->name);
+ afr_sh_call_entry_expunge_remove (frame, this,
+ (long) i, &sh->buf[i],
+ &sh->parentbufs[i],
+ afr_sh_remove_entry_cbk);
+ }
+out:
+ return;
}
-
-static int
-sh_missing_entries_readlink_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- const char *link, struct iatt *sbuf)
+void
+afr_sh_purge_entry (call_frame_t *frame, xlator_t *this)
{
- if (op_ret > 0)
- sh_missing_entries_symlink (frame, this, link, sbuf);
- else
- sh_missing_entries_finish (frame, this);
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
- return 0;
-}
+ local = frame->local;
+ sh = &local->self_heal;
+ sh->post_remove_call = afr_sh_missing_entries_finish;
+ afr_sh_purge_entry_common (frame, this, afr_sh_purge_entry_condition);
+}
-static int
-sh_missing_entries_readlink (call_frame_t *frame, xlator_t *this)
+void
+afr_sh_purge_stale_entry (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ sh->post_remove_call = afr_sh_purge_stale_entries_done;
- STACK_WIND (frame, sh_missing_entries_readlink_cbk,
- priv->children[sh->source],
- priv->children[sh->source]->fops->readlink,
- &local->loc, 4096);
+ for (i = 0; i < priv->child_count; i++) {
+ if (afr_is_child_present (sh->fresh_children,
+ priv->child_count, i))
+ continue;
- return 0;
-}
+ if ((!local->child_up[i]) || sh->child_errno[i] != 0)
+ continue;
+ GF_ASSERT (!uuid_is_null (sh->entrybuf.ia_gfid) ||
+ uuid_is_null (sh->buf[i].ia_gfid));
-static int
-sh_missing_entries_create (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- int type = 0;
- int i = 0;
- afr_private_t *priv = NULL;
- int enoent_count = 0;
- int govinda_gOvinda = 0;
+ if ((sh->entrybuf.ia_type != sh->buf[i].ia_type) ||
+ (uuid_compare (sh->buf[i].ia_gfid,
+ sh->entrybuf.ia_gfid)))
+ continue;
+ afr_children_add_child (sh->fresh_children, i,
+ priv->child_count);
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ }
+ afr_sh_purge_entry_common (frame, this,
+ afr_sh_purge_stale_entry_condition);
+}
- for (i = 0; i < priv->child_count; i++) {
- if (!local->child_up[i])
- continue;
+void
+afr_sh_save_child_iatts_from_policy (int32_t *children, struct iatt *bufs,
+ struct iatt *save,
+ unsigned int child_count)
+{
+ int i = 0;
+ int child = 0;
+ gf_boolean_t saved = _gf_false;
- if (sh->child_errno[i]) {
- if (sh->child_errno[i] == ENOENT)
- enoent_count++;
- } else {
- if (type) {
- if (type != sh->buf[i].ia_type) {
- gf_log (this->name, GF_LOG_TRACE,
- "file %s is govinda!",
- local->loc.path);
-
- govinda_gOvinda = 1;
- }
- } else {
- sh->source = i;
- type = sh->buf[i].ia_type;
- }
- }
- }
+ GF_ASSERT (save);
+ //if iatt buf with gfid exists sets it
+ for (i = 0; i < child_count; i++) {
+ child = children[i];
+ if (child == -1)
+ break;
+ *save = bufs[child];
+ saved = _gf_true;
+ if (!uuid_is_null (save->ia_gfid))
+ break;
+ }
+ GF_ASSERT (saved);
+}
- if (govinda_gOvinda) {
- gf_log (this->name, GF_LOG_ERROR,
- "conflicting filetypes exist for path %s. returning.",
- local->loc.path);
+void
+afr_get_children_of_fresh_parent_dirs (afr_self_heal_t *sh,
+ unsigned int child_count)
+{
+ afr_children_intersection_get (sh->success_children,
+ sh->fresh_parent_dirs,
+ sh->sources, child_count);
+ afr_get_fresh_children (sh->success_children, sh->sources,
+ sh->fresh_children, child_count);
+ memset (sh->sources, 0, sizeof (*sh->sources) * child_count);
+}
- local->govinda_gOvinda = 1;
- sh_missing_entries_finish (frame, this);
- return 0;
- }
+void
+afr_sh_children_lookup_done (call_frame_t *frame, xlator_t *this,
+ int32_t op_ret, int32_t op_errno)
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int32_t fresh_child_enoents = 0;
+ int32_t fresh_parent_count = 0;
- if (!type) {
- gf_log (this->name, GF_LOG_ERROR,
- "no source found for %s. all nodes down?. returning.",
- local->loc.path);
- /* subvolumes down and/or file does not exist */
- sh_missing_entries_finish (frame, this);
- return 0;
- }
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- if (enoent_count == 0) {
- gf_log (this->name, GF_LOG_ERROR,
- "no missing files - %s. proceeding to metadata check",
- local->loc.path);
- /* proceed to next step - metadata self-heal */
- sh_missing_entries_finish (frame, this);
- return 0;
- }
+ if (op_ret < 0)
+ goto fail;
+ afr_get_children_of_fresh_parent_dirs (sh, priv->child_count);
+ fresh_parent_count = afr_get_children_count (sh->fresh_parent_dirs,
+ priv->child_count);
+ //we need the enoent count of the subvols present in fresh_parent_dirs
+ fresh_child_enoents = afr_errno_count (sh->fresh_parent_dirs,
+ sh->child_errno,
+ priv->child_count, ENOENT);
+ if (fresh_child_enoents == fresh_parent_count) {
+ afr_sh_set_error (sh, ENOENT);
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_purge_entry (frame, this);
+ } else if (!afr_conflicting_iattrs (sh->buf, sh->fresh_children,
+ priv->child_count, local->loc.path,
+ this->name)) {
+ afr_sh_save_child_iatts_from_policy (sh->fresh_children,
+ sh->buf, &sh->entrybuf,
+ priv->child_count);
+ afr_update_gfid_from_iatts (sh->sh_gfid_req, sh->buf,
+ sh->fresh_children,
+ priv->child_count);
+ afr_sh_purge_stale_entry (frame, this);
+ } else {
+ op_errno = EIO;
+ afr_set_local_for_unhealable (local);
+ goto fail;
+ }
- switch (type) {
- case IA_IFSOCK:
- case IA_IFREG:
- case IA_IFBLK:
- case IA_IFCHR:
- case IA_IFIFO:
- sh_missing_entries_mknod (frame, this);
- break;
- case IA_IFLNK:
- sh_missing_entries_readlink (frame, this);
- break;
- case IA_IFDIR:
- sh_missing_entries_mkdir (frame, this);
- break;
- default:
- gf_log (this->name, GF_LOG_ERROR,
- "unknown file type: 0%o", type);
- local->govinda_gOvinda = 1;
- sh_missing_entries_finish (frame, this);
- }
+ return;
- return 0;
+fail:
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_set_error (sh, op_errno);
+ afr_sh_missing_entries_finish (frame, this);
+ return;
}
+static void
+afr_sh_find_fresh_parents (call_frame_t *frame, xlator_t *this,
+ int32_t op_ret, int32_t op_errno)
+{
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int enoent_count = 0;
+ int nsources = 0;
+ int source = -1;
+ int32_t subvol_status = 0;
-static int
-sh_missing_entries_lookup_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf, dict_t *xattr,
- struct iatt *postparent)
-{
- int child_index = 0;
- afr_local_t *local = NULL;
- int call_count = 0;
- afr_private_t *priv = NULL;
-
-
- local = frame->local;
- priv = this->private;
-
- child_index = (long) cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == 0) {
- gf_log (this->name, GF_LOG_TRACE,
- "path %s on subvolume %s is of mode 0%o",
- local->loc.path,
- priv->children[child_index]->name,
- buf->ia_type);
-
- local->self_heal.buf[child_index] = *buf;
- local->self_heal.parentbuf = *postparent;
- } else {
- gf_log (this->name, GF_LOG_TRACE,
- "path %s on subvolume %s => -1 (%s)",
- local->loc.path,
- priv->children[child_index]->name,
- strerror (op_errno));
-
- local->self_heal.child_errno[child_index] = op_errno;
- }
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- }
- UNLOCK (&frame->lock);
+ if (op_ret < 0)
+ goto out;
+ enoent_count = afr_errno_count (NULL, sh->child_errno,
+ priv->child_count, ENOENT);
+ if (enoent_count > 0) {
+ gf_log (this->name, GF_LOG_INFO, "Parent dir missing for %s,"
+ " in missing entry self-heal, aborting missing-entry "
+ "self-heal",
+ local->loc.path);
+ afr_sh_missing_entries_finish (frame, this);
+ return;
+ }
- call_count = afr_frame_return (frame);
+ nsources = afr_build_sources (this, sh->xattr, sh->buf,
+ sh->pending_matrix, sh->sources,
+ sh->success_children,
+ AFR_ENTRY_TRANSACTION, &subvol_status,
+ _gf_true);
+ if ((subvol_status & ALL_FOOLS) ||
+ (subvol_status & SPLIT_BRAIN)) {
+ gf_log (this->name, GF_LOG_INFO, "%s: Performing conservative "
+ "merge", sh->parent_loc.path);
+ afr_mark_success_children_sources (sh->sources,
+ sh->success_children,
+ priv->child_count);
+ } else if (nsources < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "No sources for dir "
+ "of %s, in missing entry self-heal, aborting "
+ "self-heal", local->loc.path);
+ op_errno = EIO;
+ goto out;
+ }
- if (call_count == 0) {
- sh_missing_entries_create (frame, this);
- }
+ source = afr_sh_select_source (sh->sources, priv->child_count);
+ if (source == -1) {
+ GF_ASSERT (0);
+ gf_log (this->name, GF_LOG_DEBUG, "No active sources found.");
+ op_errno = EIO;
+ goto out;
+ }
+ afr_get_fresh_children (sh->success_children, sh->sources,
+ sh->fresh_parent_dirs, priv->child_count);
+ afr_sh_common_lookup (frame, this, &local->loc,
+ afr_sh_children_lookup_done, NULL, 0,
+ NULL);
+ return;
- return 0;
+out:
+ afr_sh_set_error (sh, op_errno);
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_missing_entries_finish (frame, this);
+ return;
}
+void
+afr_sh_common_reset (afr_self_heal_t *sh, unsigned int child_count)
+{
+ int i = 0;
-static int
-sh_missing_entries_lookup (call_frame_t *frame, xlator_t *this)
+ for (i = 0; i < child_count; i++) {
+ memset (&sh->buf[i], 0, sizeof (sh->buf[i]));
+ memset (&sh->parentbufs[i], 0, sizeof (sh->parentbufs[i]));
+ sh->child_errno[i] = 0;
+ }
+ memset (&sh->parentbuf, 0, sizeof (sh->parentbuf));
+ sh->success_count = 0;
+ afr_reset_children (sh->success_children, child_count);
+ afr_reset_children (sh->fresh_children, child_count);
+ afr_reset_xattr (sh->xattr, child_count);
+ loc_wipe (&sh->lookup_loc);
+}
+
+/* afr self-heal state will be lost if this call is made
+ * please check the afr_sh_common_reset that is called in this function
+ */
+int
+afr_sh_common_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
+ afr_lookup_done_cbk_t lookup_done , uuid_t gfid,
+ int32_t flags, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int i = 0;
- int call_count = 0;
- afr_private_t *priv = NULL;
- dict_t *xattr_req = NULL;
- int ret = -1;
+ afr_local_t *local = NULL;
+ int i = 0;
+ int call_count = 0;
+ afr_private_t *priv = NULL;
+ dict_t *xattr_req = NULL;
+ afr_self_heal_t *sh = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
+ sh = &local->self_heal;
- call_count = afr_up_children_count (priv->child_count,
- local->child_up);
+ call_count = afr_up_children_count (local->child_up, priv->child_count);
- local->call_count = call_count;
+ local->call_count = call_count;
- xattr_req = dict_new();
+ xattr_req = dict_new();
- if (xattr_req) {
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_set_uint64 (xattr_req,
- priv->pending_key[i],
- 3 * sizeof(int32_t));
- if (ret < 0)
- gf_log (this->name, GF_LOG_WARNING,
- "Unable to set dict value.");
+ if (xattr_req) {
+ afr_xattr_req_prepare (this, xattr_req, loc->path);
+ if (gfid) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "looking up %s with gfid: %s",
+ loc->path, uuid_utoa (gfid));
+ GF_ASSERT (!uuid_is_null (gfid));
+ afr_set_dict_gfid (xattr_req, gfid);
}
}
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- gf_log (this->name, GF_LOG_TRACE,
- "looking up %s on subvolume %s",
- local->loc.path, priv->children[i]->name);
-
- STACK_WIND_COOKIE (frame,
- sh_missing_entries_lookup_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->lookup,
- &local->loc, xattr_req);
-
- if (!--call_count)
- break;
- }
- }
+ afr_sh_common_reset (sh, priv->child_count);
+ sh->lookup_done = lookup_done;
+ loc_copy (&sh->lookup_loc, loc);
+ sh->lookup_flags = flags;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "looking up %s on subvolume %s",
+ loc->path, priv->children[i]->name);
+
+ STACK_WIND_COOKIE (frame,
+ afr_sh_common_lookup_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->lookup,
+ loc, xattr_req);
+
+ if (!--call_count)
+ break;
+ }
+ }
- if (xattr_req)
- dict_unref (xattr_req);
+ if (xattr_req)
+ dict_unref (xattr_req);
- return 0;
+ return 0;
}
int
-afr_sh_post_nonblocking_entrylk_cbk (call_frame_t *frame, xlator_t *this)
+afr_sh_post_nb_entrylk_missing_entry_sh_cbk (call_frame_t *frame,
+ xlator_t *this)
{
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
local = frame->local;
int_lock = &local->internal_lock;
+ sh = &local->self_heal;
if (int_lock->lock_op_ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
+ gf_log (this->name, GF_LOG_INFO,
"Non blocking entrylks failed.");
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
afr_sh_missing_entries_done (frame, this);
} else {
gf_log (this->name, GF_LOG_DEBUG,
"Non blocking entrylks done. Proceeding to FOP");
- sh_missing_entries_lookup (frame, this);
+ afr_sh_common_lookup (frame, this, &sh->parent_loc,
+ afr_sh_find_fresh_parents,
+ NULL, AFR_LOOKUP_FAIL_CONFLICTS,
+ NULL);
}
return 0;
}
-static int
-afr_sh_entrylk (call_frame_t *frame, xlator_t *this)
+int
+afr_sh_entrylk (call_frame_t *frame, xlator_t *this, loc_t *loc,
+ char *base_name, afr_lock_cbk_t lock_cbk)
{
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ priv = this->private;
local = frame->local;
int_lock = &local->internal_lock;
- sh = &local->self_heal;
int_lock->transaction_lk_type = AFR_SELFHEAL_LK;
int_lock->selfheal_lk_type = AFR_ENTRY_SELF_HEAL_LK;
afr_set_lock_number (frame, this);
- int_lock->lk_basename = local->loc.name;
- int_lock->lk_loc = &sh->parent_loc;
- int_lock->lock_cbk = afr_sh_post_nonblocking_entrylk_cbk;
+ int_lock->lk_basename = base_name;
+ int_lock->lk_loc = loc;
+ int_lock->lock_cbk = lock_cbk;
+ int_lock->domain = this->name;
+ int_lock->lockee_count = 0;
+ afr_init_entry_lockee (&int_lock->lockee[0], local, loc,
+ base_name, priv->child_count);
+ int_lock->lockee_count++;
afr_nonblocking_entrylk (frame, this);
return 0;
}
static int
-afr_self_heal_missing_entries (call_frame_t *frame, xlator_t *this)
+afr_self_heal_parent_entrylk (call_frame_t *frame, xlator_t *this,
+ afr_lock_cbk_t lock_cbk)
{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
+ int ret = -1;
+ int32_t op_errno = 0;
- local = frame->local;
- int_lock = &local->internal_lock;
- sh = &local->self_heal;
- priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
- gf_log (this->name, GF_LOG_TRACE,
- "attempting to recreate missing entries for path=%s",
- local->loc.path);
+ gf_log (this->name, GF_LOG_TRACE,
+ "attempting to recreate missing entries for path=%s",
+ local->loc.path);
- afr_build_parent_loc (&sh->parent_loc, &local->loc);
+ ret = afr_build_parent_loc (&sh->parent_loc, &local->loc, &op_errno);
+ if (ret)
+ goto out;
- afr_sh_entrylk (frame, this);
- return 0;
+ afr_sh_entrylk (frame, this, &sh->parent_loc, NULL,
+ lock_cbk);
+ return 0;
+out:
+ int_lock = &local->internal_lock;
+ int_lock->lock_op_ret = -1;
+ lock_cbk (frame, this);
+ return 0;
}
-afr_local_t *afr_local_copy (afr_local_t *l, xlator_t *this)
+static int
+afr_self_heal_missing_entries (call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *lc = NULL;
+ afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
- afr_self_heal_t *shc = NULL;
+ local = frame->local;
+ sh = &local->self_heal;
+
+ sh->sh_type_in_action = AFR_SELF_HEAL_GFID_OR_MISSING_ENTRY;
+
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_STARTED);
+
+ afr_self_heal_parent_entrylk (frame, this,
+ afr_sh_post_nb_entrylk_missing_entry_sh_cbk);
+ return 0;
+}
+
+afr_local_t*
+afr_self_heal_local_init (afr_local_t *l, xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *lc = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_self_heal_t *shc = NULL;
+ int ret = 0;
priv = this->private;
sh = &l->self_heal;
- lc = GF_CALLOC (1, sizeof (afr_local_t),
- gf_afr_mt_afr_local_t);
+ lc = mem_get0 (this->local_pool);
+ if (!lc)
+ goto out;
shc = &lc->self_heal;
shc->unwind = sh->unwind;
- shc->need_data_self_heal = sh->need_data_self_heal;
- shc->need_metadata_self_heal = sh->need_metadata_self_heal;
- shc->need_entry_self_heal = sh->need_entry_self_heal;
+ shc->gfid_sh_success_cbk = sh->gfid_sh_success_cbk;
+ shc->do_missing_entry_self_heal = sh->do_missing_entry_self_heal;
+ shc->do_gfid_self_heal = sh->do_gfid_self_heal;
+ shc->do_data_self_heal = sh->do_data_self_heal;
+ shc->do_metadata_self_heal = sh->do_metadata_self_heal;
+ shc->do_entry_self_heal = sh->do_entry_self_heal;
+ shc->force_confirm_spb = sh->force_confirm_spb;
shc->forced_merge = sh->forced_merge;
- shc->healing_fd_opened = sh->healing_fd_opened;
- shc->data_lock_held = sh->data_lock_held;
- if (sh->healing_fd && !sh->healing_fd_opened)
- shc->healing_fd = fd_ref (sh->healing_fd);
- else
- shc->healing_fd = sh->healing_fd;
shc->background = sh->background;
shc->type = sh->type;
+ shc->data_sh_info = "";
+ shc->metadata_sh_info = "";
+
+ uuid_copy (shc->sh_gfid_req, sh->sh_gfid_req);
+ if (l->loc.path) {
+ ret = loc_copy (&lc->loc, &l->loc);
+ if (ret < 0)
+ goto out;
+ }
- if (l->loc.path)
- loc_copy (&lc->loc, &l->loc);
+ lc->child_up = memdup (l->child_up,
+ sizeof (*lc->child_up) * priv->child_count);
+ if (!lc->child_up) {
+ ret = -1;
+ goto out;
+ }
- lc->child_up = memdup (l->child_up, priv->child_count);
if (l->xattr_req)
lc->xattr_req = dict_ref (l->xattr_req);
@@ -1464,69 +2245,67 @@ afr_local_t *afr_local_copy (afr_local_t *l, xlator_t *this)
lc->cont.lookup.inode = inode_ref (l->cont.lookup.inode);
if (l->cont.lookup.xattr)
lc->cont.lookup.xattr = dict_ref (l->cont.lookup.xattr);
- if (l->internal_lock.inode_locked_nodes)
- lc->internal_lock.inode_locked_nodes =
- memdup (l->internal_lock.inode_locked_nodes,
- priv->child_count);
- else
- lc->internal_lock.inode_locked_nodes =
- GF_CALLOC (sizeof (*l->internal_lock.inode_locked_nodes),
- priv->child_count,
- gf_afr_mt_char);
- if (l->internal_lock.entry_locked_nodes)
- lc->internal_lock.entry_locked_nodes =
- memdup (l->internal_lock.entry_locked_nodes,
- priv->child_count);
- else
- lc->internal_lock.entry_locked_nodes =
- GF_CALLOC (sizeof (*l->internal_lock.entry_locked_nodes),
- priv->child_count,
- gf_afr_mt_char);
- if (l->internal_lock.locked_nodes)
- lc->internal_lock.locked_nodes =
- memdup (l->internal_lock.locked_nodes,
- priv->child_count);
- else
- lc->internal_lock.locked_nodes =
- GF_CALLOC (sizeof (*l->internal_lock.locked_nodes),
- priv->child_count,
- gf_afr_mt_char);
- lc->internal_lock.inodelk_lock_count =
- l->internal_lock.inodelk_lock_count;
- lc->internal_lock.entrylk_lock_count =
- l->internal_lock.entrylk_lock_count;
+ lc->internal_lock.locked_nodes =
+ GF_CALLOC (sizeof (*l->internal_lock.locked_nodes),
+ priv->child_count, gf_afr_mt_char);
+ if (!lc->internal_lock.locked_nodes) {
+ ret = -1;
+ goto out;
+ }
+
+ ret = afr_inodelk_init (&lc->internal_lock.inodelk[0],
+ this->name, priv->child_count);
+ if (ret)
+ goto out;
+out:
+ if (ret) {
+ afr_local_cleanup (lc, this);
+ lc = NULL;
+ }
return lc;
}
-
int
afr_self_heal_completion_cbk (call_frame_t *bgsh_frame, xlator_t *this)
{
afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
+ afr_local_t * local = NULL;
afr_self_heal_t * sh = NULL;
+ afr_local_t * orig_frame_local = NULL;
+ afr_self_heal_t * orig_frame_sh = NULL;
char sh_type_str[256] = {0,};
+ gf_loglevel_t loglevel = 0;
priv = this->private;
- local = bgsh_frame->local;
+ local = bgsh_frame->local;
sh = &local->self_heal;
- if (local->govinda_gOvinda) {
- afr_set_split_brain (this, local->cont.lookup.inode, _gf_true);
- } else {
- afr_set_split_brain (this, local->cont.lookup.inode, _gf_false);
- }
+ if (local->unhealable) {
+ afr_set_split_brain (this, sh->inode, SPB, SPB);
+ }
- afr_self_heal_type_str_get(sh, sh_type_str,
- sizeof(sh_type_str));
- gf_log (this->name, GF_LOG_NORMAL,
- "background %s self-heal completed on %s", sh_type_str,
- local->loc.path);
+ afr_self_heal_type_str_get (sh, sh_type_str,
+ sizeof(sh_type_str));
+ if (is_self_heal_failed (sh, AFR_CHECK_ALL) && !priv->shd.iamshd) {
+ loglevel = GF_LOG_ERROR;
+ } else if (!is_self_heal_failed (sh, AFR_CHECK_ALL)) {
+ loglevel = GF_LOG_INFO;
+ } else {
+ loglevel = GF_LOG_DEBUG;
+ }
+
+ afr_log_self_heal_completion_status (local, loglevel);
- if (!sh->unwound) {
- sh->unwind (sh->orig_frame, this);
+ FRAME_SU_UNDO (bgsh_frame, afr_local_t);
+
+ if (!sh->unwound && sh->unwind) {
+ orig_frame_local = sh->orig_frame->local;
+ orig_frame_sh = &orig_frame_local->self_heal;
+ orig_frame_sh->actual_sh_started = _gf_true;
+ sh->unwind (sh->orig_frame, this, sh->op_ret, sh->op_errno,
+ is_self_heal_failed (sh, AFR_CHECK_ALL));
}
if (sh->background) {
@@ -1539,119 +2318,495 @@ afr_self_heal_completion_cbk (call_frame_t *bgsh_frame, xlator_t *this)
AFR_STACK_DESTROY (bgsh_frame);
- return 0;
+ return 0;
}
int
-afr_self_heal (call_frame_t *frame, xlator_t *this)
+afr_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int32_t op_errno = 0;
+ int ret = 0;
+ afr_self_heal_t *orig_sh = NULL;
+ call_frame_t *sh_frame = NULL;
+ afr_local_t *sh_local = NULL;
+ loc_t *loc = NULL;
- call_frame_t *sh_frame = NULL;
- afr_local_t *sh_local = NULL;
+ local = frame->local;
+ orig_sh = &local->self_heal;
+ priv = this->private;
+
+ GF_ASSERT (local->loc.path);
+
+ gf_log (this->name, GF_LOG_TRACE,
+ "performing self heal on %s (metadata=%d data=%d entry=%d)",
+ local->loc.path,
+ local->self_heal.do_metadata_self_heal,
+ local->self_heal.do_data_self_heal,
+ local->self_heal.do_entry_self_heal);
+
+ op_errno = ENOMEM;
+ sh_frame = copy_frame (frame);
+ if (!sh_frame)
+ goto out;
+ afr_set_lk_owner (sh_frame, this, sh_frame->root);
+ afr_set_low_priority (sh_frame);
+
+ sh_local = afr_self_heal_local_init (local, this);
+ if (!sh_local)
+ goto out;
+ sh_frame->local = sh_local;
+ sh = &sh_local->self_heal;
+
+ sh->inode = inode_ref (inode);
+ sh->orig_frame = frame;
+ sh->completion_cbk = afr_self_heal_completion_cbk;
- local = frame->local;
- priv = this->private;
+ sh->success = GF_CALLOC (priv->child_count, sizeof (*sh->success),
+ gf_afr_mt_char);
+ if (!sh->success)
+ goto out;
+ sh->sources = GF_CALLOC (sizeof (*sh->sources), priv->child_count,
+ gf_afr_mt_int);
+ if (!sh->sources)
+ goto out;
+ sh->locked_nodes = GF_CALLOC (sizeof (*sh->locked_nodes),
+ priv->child_count,
+ gf_afr_mt_int);
+ if (!sh->locked_nodes)
+ goto out;
- afr_set_lk_owner (frame, this);
+ sh->pending_matrix = afr_matrix_create (priv->child_count,
+ priv->child_count);
+ if (!sh->pending_matrix)
+ goto out;
+
+ sh->delta_matrix = afr_matrix_create (priv->child_count,
+ priv->child_count);
+ if (!sh->delta_matrix)
+ goto out;
+
+ sh->fresh_parent_dirs = afr_children_create (priv->child_count);
+ if (!sh->fresh_parent_dirs)
+ goto out;
+ ret = afr_sh_common_create (sh, priv->child_count);
+ if (ret) {
+ op_errno = -ret;
+ goto out;
+ }
if (local->self_heal.background) {
LOCK (&priv->lock);
{
if (priv->background_self_heals_started
- > priv->background_self_heal_count) {
+ < priv->background_self_heal_count) {
+ priv->background_self_heals_started++;
- local->self_heal.background = _gf_false;
} else {
- priv->background_self_heals_started++;
+ local->self_heal.background = _gf_false;
+ sh->background = _gf_false;
}
}
UNLOCK (&priv->lock);
}
- gf_log (this->name, GF_LOG_TRACE,
- "performing self heal on %s (metadata=%d data=%d entry=%d)",
- local->loc.path,
- local->self_heal.need_metadata_self_heal,
- local->self_heal.need_data_self_heal,
- local->self_heal.need_entry_self_heal);
+ if (!local->loc.parent) {
+ sh->do_missing_entry_self_heal = _gf_false;
+ sh->do_gfid_self_heal = _gf_false;
+ }
- sh_frame = copy_frame (frame);
- sh_local = afr_local_copy (local, this);
- sh_frame->local = sh_local;
- sh = &sh_local->self_heal;
+ sh->sh_type_in_action = AFR_SELF_HEAL_INVALID;
- sh->orig_frame = frame;
+ FRAME_SU_DO (sh_frame, afr_local_t);
+ if (sh->do_missing_entry_self_heal || sh->do_gfid_self_heal) {
+ afr_self_heal_missing_entries (sh_frame, this);
+ } else {
+ loc = &sh_local->loc;
+ if (uuid_is_null (loc->inode->gfid) && uuid_is_null (loc->gfid)) {
+ if (!uuid_is_null (inode->gfid))
+ GF_ASSERT (!uuid_compare (inode->gfid,
+ sh->sh_gfid_req));
+ uuid_copy (loc->gfid, sh->sh_gfid_req);
+ }
+ gf_log (this->name, GF_LOG_TRACE,
+ "proceeding to metadata check on %s",
+ local->loc.path);
- sh->completion_cbk = afr_self_heal_completion_cbk;
+ afr_sh_missing_entries_done (sh_frame, this);
+ }
+ op_errno = 0;
+out:
+ if (op_errno) {
+ orig_sh->unwind (frame, this, -1, op_errno, 1);
+ if (sh_frame)
+ AFR_STACK_DESTROY (sh_frame);
+ }
+ return 0;
+}
- sh->buf = GF_CALLOC (priv->child_count, sizeof (struct iatt),
- gf_afr_mt_iatt);
- sh->child_errno = GF_CALLOC (priv->child_count, sizeof (int),
- gf_afr_mt_int);
- sh->success = GF_CALLOC (priv->child_count, sizeof (int),
- gf_afr_mt_int);
- sh->xattr = GF_CALLOC (priv->child_count, sizeof (dict_t *),
- gf_afr_mt_dict_t);
- sh->sources = GF_CALLOC (sizeof (*sh->sources), priv->child_count,
- gf_afr_mt_int);
- sh->locked_nodes = GF_CALLOC (sizeof (*sh->locked_nodes),
- priv->child_count,
- gf_afr_mt_int);
+void
+afr_self_heal_type_str_get (afr_self_heal_t *self_heal_p, char *str,
+ size_t size)
+{
+ GF_ASSERT (str && (size > strlen (" missing-entry gfid "
+ "meta-data data entry")));
- sh->pending_matrix = GF_CALLOC (sizeof (int32_t *), priv->child_count,
- gf_afr_mt_int32_t);
+ if (self_heal_p->do_metadata_self_heal) {
+ snprintf (str, size, " meta-data");
+ }
- for (i = 0; i < priv->child_count; i++) {
- sh->pending_matrix[i] = GF_CALLOC (sizeof (int32_t),
- priv->child_count,
- gf_afr_mt_int32_t);
- }
+ if (self_heal_p->do_data_self_heal) {
+ snprintf (str + strlen(str), size - strlen(str), " data");
+ }
- sh->delta_matrix = GF_CALLOC (sizeof (int32_t *), priv->child_count,
- gf_afr_mt_int32_t);
- for (i = 0; i < priv->child_count; i++) {
- sh->delta_matrix[i] = GF_CALLOC (sizeof (int32_t),
- priv->child_count,
- gf_afr_mt_int32_t);
- }
+ if (self_heal_p->do_entry_self_heal) {
+ snprintf (str + strlen(str), size - strlen(str), " entry");
+ }
- if (local->success_count && local->enoent_count) {
- afr_self_heal_missing_entries (sh_frame, this);
- } else {
- gf_log (this->name, GF_LOG_TRACE,
- "proceeding to metadata check on %s",
- local->loc.path);
+ if (self_heal_p->do_missing_entry_self_heal) {
+ snprintf (str + strlen(str), size - strlen(str),
+ " missing-entry");
+ }
- afr_sh_missing_entries_done (sh_frame, this);
- }
+ if (self_heal_p->do_gfid_self_heal) {
+ snprintf (str + strlen(str), size - strlen(str), " gfid");
+ }
+}
- return 0;
+afr_self_heal_type
+afr_self_heal_type_for_transaction (afr_transaction_type type)
+{
+ afr_self_heal_type sh_type = AFR_SELF_HEAL_INVALID;
+
+ switch (type) {
+ case AFR_DATA_TRANSACTION:
+ sh_type = AFR_SELF_HEAL_DATA;
+ break;
+ case AFR_METADATA_TRANSACTION:
+ sh_type = AFR_SELF_HEAL_METADATA;
+ break;
+ case AFR_ENTRY_TRANSACTION:
+ sh_type = AFR_SELF_HEAL_ENTRY;
+ break;
+ case AFR_ENTRY_RENAME_TRANSACTION:
+ GF_ASSERT (0);
+ break;
+ }
+ return sh_type;
+}
+
+int
+afr_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name)
+{
+ int ret = -1;
+ uuid_t pargfid = {0};
+
+ if (!child)
+ goto out;
+
+ if (!uuid_is_null (parent->inode->gfid))
+ uuid_copy (pargfid, parent->inode->gfid);
+ else if (!uuid_is_null (parent->gfid))
+ uuid_copy (pargfid, parent->gfid);
+
+ if (uuid_is_null (pargfid))
+ goto out;
+
+ if (strcmp (parent->path, "/") == 0)
+ ret = gf_asprintf ((char **)&child->path, "/%s", name);
+ else
+ ret = gf_asprintf ((char **)&child->path, "%s/%s", parent->path,
+ name);
+
+ if (-1 == ret) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "asprintf failed while setting child path");
+ }
+
+ child->name = strrchr (child->path, '/');
+ if (child->name)
+ child->name++;
+
+ child->parent = inode_ref (parent->inode);
+ child->inode = inode_new (parent->inode->table);
+ uuid_copy (child->pargfid, pargfid);
+
+ if (!child->inode) {
+ ret = -1;
+ goto out;
+ }
+
+ ret = 0;
+out:
+ if ((ret == -1) && child)
+ loc_wipe (child);
+
+ return ret;
+}
+
+int
+afr_sh_erase_pending (call_frame_t *frame, xlator_t *this,
+ afr_transaction_type type, afr_fxattrop_cbk_t cbk,
+ int (*finish)(call_frame_t *frame, xlator_t *this))
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int i = 0;
+ dict_t **erase_xattr = NULL;
+ int ret = -1;
+
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+
+ afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix,
+ sh->success, priv->child_count, type);
+
+ erase_xattr = GF_CALLOC (sizeof (*erase_xattr), priv->child_count,
+ gf_afr_mt_dict_t);
+ if (!erase_xattr)
+ goto out;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (sh->xattr[i]) {
+ call_count++;
+ erase_xattr[i] = dict_new ();
+ if (!erase_xattr[i])
+ goto out;
+ }
+ }
+
+ afr_sh_delta_to_xattr (this, sh->delta_matrix, erase_xattr,
+ priv->child_count, type);
+
+ gf_log (this->name, GF_LOG_DEBUG, "Delta matrix for: %s",
+ lkowner_utoa (&frame->root->lk_owner));
+ afr_sh_print_pending_matrix (sh->delta_matrix, this);
+ local->call_count = call_count;
+ if (call_count == 0) {
+ ret = 0;
+ finish (frame, this);
+ goto out;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!erase_xattr[i])
+ continue;
+
+ if (sh->healing_fd) {//true for ENTRY, reg file DATA transaction
+ STACK_WIND_COOKIE (frame, cbk, (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->fxattrop,
+ sh->healing_fd,
+ GF_XATTROP_ADD_ARRAY, erase_xattr[i],
+ NULL);
+ } else {
+ STACK_WIND_COOKIE (frame, cbk, (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->xattrop,
+ &local->loc,
+ GF_XATTROP_ADD_ARRAY, erase_xattr[i],
+ NULL);
+ }
+ }
+
+ ret = 0;
+out:
+ if (erase_xattr) {
+ for (i = 0; i < priv->child_count; i++) {
+ if (erase_xattr[i]) {
+ dict_unref (erase_xattr[i]);
+ }
+ }
+ }
+
+ GF_FREE (erase_xattr);
+
+ if (ret < 0) {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ finish (frame, this);
+ }
+
+ return 0;
}
void
-afr_self_heal_type_str_get (afr_self_heal_t *self_heal_p, char *str,
- size_t size)
+afr_set_self_heal_status(afr_self_heal_t *sh, afr_self_heal_status status)
+{
+ xlator_t *this = NULL;
+ afr_sh_status_for_all_type *sh_status = &(sh->afr_all_sh_status);
+ afr_self_heal_type sh_type_in_action = sh->sh_type_in_action;
+ this = THIS;
+
+ if (!sh) {
+ gf_log_callingfn (this->name, GF_LOG_ERROR, "Null self heal"
+ "Structure");
+ goto out;
+ }
+
+ switch (sh_type_in_action) {
+ case AFR_SELF_HEAL_GFID_OR_MISSING_ENTRY:
+ sh_status->gfid_or_missing_entry_self_heal = status;
+ break;
+ case AFR_SELF_HEAL_METADATA:
+ sh_status->metadata_self_heal = status;
+ break;
+ case AFR_SELF_HEAL_DATA:
+ sh_status->data_self_heal = status;
+ break;
+ case AFR_SELF_HEAL_ENTRY:
+ sh_status->entry_self_heal = status;
+ break;
+ case AFR_SELF_HEAL_INVALID:
+ gf_log_callingfn (this->name, GF_LOG_ERROR, "Invalid"
+ "self heal type in action");
+ break;
+ }
+out:
+ return;
+}
+
+void
+afr_set_local_for_unhealable (afr_local_t *local)
+{
+ afr_self_heal_t *sh = NULL;
+
+ sh = &local->self_heal;
+
+ local->unhealable = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+}
+
+int
+is_self_heal_failed (afr_self_heal_t *sh, afr_sh_fail_check_type type)
{
- assert(str && (size > 0));
+ afr_sh_status_for_all_type sh_status = sh->afr_all_sh_status;
+ afr_self_heal_type sh_type_in_action = AFR_SELF_HEAL_INVALID;
+ afr_self_heal_status status = AFR_SELF_HEAL_FAILED;
+ xlator_t *this = NULL;
+ int sh_failed = 0;
+
+ this = THIS;
+
+ if (!sh) {
+ gf_log_callingfn (this->name, GF_LOG_ERROR, "Null self heal "
+ "structure");
+ sh_failed = 1;
+ goto out;
+ }
+
+ if (type == AFR_CHECK_ALL) {
+ if ((sh_status.gfid_or_missing_entry_self_heal == AFR_SELF_HEAL_FAILED)
+ || (sh_status.metadata_self_heal == AFR_SELF_HEAL_FAILED)
+ || (sh_status.data_self_heal == AFR_SELF_HEAL_FAILED)
+ || (sh_status.entry_self_heal == AFR_SELF_HEAL_FAILED))
+ sh_failed = 1;
+ } else if (type == AFR_CHECK_SPECIFIC) {
+ sh_type_in_action = sh->sh_type_in_action;
+ switch (sh_type_in_action) {
+ case AFR_SELF_HEAL_GFID_OR_MISSING_ENTRY:
+ status = sh_status.gfid_or_missing_entry_self_heal;
+ break;
+ case AFR_SELF_HEAL_METADATA:
+ status = sh_status.metadata_self_heal;
+ break;
+ case AFR_SELF_HEAL_ENTRY:
+ status = sh_status.entry_self_heal;
+ break;
+ case AFR_SELF_HEAL_DATA:
+ status = sh_status.data_self_heal;
+ break;
+ case AFR_SELF_HEAL_INVALID:
+ status = AFR_SELF_HEAL_NOT_ATTEMPTED;
+ break;
+ }
+ if (status == AFR_SELF_HEAL_FAILED)
+ sh_failed = 1;
- if (self_heal_p->need_metadata_self_heal) {
- snprintf(str, size, " meta-data");
}
- if (self_heal_p->need_data_self_heal) {
- snprintf(str + strlen(str), size - strlen(str),
- " data");
+out:
+ return sh_failed;
+}
+
+char *
+get_sh_completion_status (afr_self_heal_status status)
+{
+
+ char *not_attempted = " is not attempted";
+ char *failed = " failed";
+ char *started = " is started";
+ char *sync_begin = " is successfully completed";
+ char *result = " has unknown status";
+
+ switch (status)
+ {
+ case AFR_SELF_HEAL_NOT_ATTEMPTED:
+ result = not_attempted;
+ break;
+ case AFR_SELF_HEAL_FAILED:
+ result = failed;
+ break;
+ case AFR_SELF_HEAL_STARTED:
+ result = started;
+ break;
+ case AFR_SELF_HEAL_SYNC_BEGIN:
+ result = sync_begin;
+ break;
}
- if (self_heal_p->need_entry_self_heal) {
- snprintf(str + strlen(str), size - strlen(str),
- " entry");
+ return result;
+
+}
+
+void
+afr_log_self_heal_completion_status (afr_local_t *local, gf_loglevel_t loglvl)
+{
+
+ char sh_log[4096] = {0};
+ afr_self_heal_t *sh = &local->self_heal;
+ afr_sh_status_for_all_type all_status = sh->afr_all_sh_status;
+ xlator_t *this = NULL;
+ size_t off = 0;
+ int data_sh = 0;
+ int metadata_sh = 0;
+ int print_log = 0;
+
+ this = THIS;
+
+ ADD_FMT_STRING (sh_log, off, "gfid or missing entry",
+ all_status.gfid_or_missing_entry_self_heal, print_log);
+ ADD_FMT_STRING_SYNC (sh_log, off, "metadata",
+ all_status.metadata_self_heal, print_log);
+ if (sh->background) {
+ ADD_FMT_STRING_SYNC (sh_log, off, "backgroung data",
+ all_status.data_self_heal, print_log);
+ } else {
+ ADD_FMT_STRING_SYNC (sh_log, off, "foreground data",
+ all_status.data_self_heal, print_log);
}
+ ADD_FMT_STRING_SYNC (sh_log, off, "entry", all_status.entry_self_heal,
+ print_log);
+
+ if (AFR_SELF_HEAL_SYNC_BEGIN == all_status.data_self_heal &&
+ strcmp (sh->data_sh_info, "") && sh->data_sh_info )
+ data_sh = 1;
+ if (AFR_SELF_HEAL_SYNC_BEGIN == all_status.metadata_self_heal &&
+ strcmp (sh->metadata_sh_info, "") && sh->metadata_sh_info)
+ metadata_sh = 1;
+
+ if (!print_log)
+ return;
+
+ gf_log (this->name, loglvl, "%s %s %s on %s", sh_log,
+ ((data_sh == 1) ? sh->data_sh_info : ""),
+ ((metadata_sh == 1) ? sh->metadata_sh_info : ""),
+ local->loc.path);
}
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.h b/xlators/cluster/afr/src/afr-self-heal-common.h
index 31f02f3f9..473264776 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.h
+++ b/xlators/cluster/afr/src/afr-self-heal-common.h
@@ -1,73 +1,144 @@
/*
- Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __AFR_SELF_HEAL_COMMON_H__
#define __AFR_SELF_HEAL_COMMON_H__
#define FILE_HAS_HOLES(buf) (((buf)->ia_size) > ((buf)->ia_blocks * 512))
+#define AFR_SH_MIN_PARTICIPANTS 2
typedef enum {
- AFR_SELF_HEAL_ENTRY,
- AFR_SELF_HEAL_METADATA,
- AFR_SELF_HEAL_DATA,
-} afr_self_heal_type;
+ AFR_LOOKUP_FAIL_CONFLICTS = 1,
+ AFR_LOOKUP_FAIL_MISSING_GFIDS = 2,
+} afr_lookup_flags_t;
int
afr_sh_select_source (int sources[], int child_count);
int
-afr_sh_sink_count (int sources[], int child_count);
-
-int
afr_sh_source_count (int sources[], int child_count);
-int
-afr_sh_supress_errenous_children (int sources[], int child_errno[],
- int child_count);
-
void
afr_sh_print_pending_matrix (int32_t *pending_matrix[], xlator_t *this);
void
-afr_sh_build_pending_matrix (afr_private_t *priv,
- int32_t *pending_matrix[], dict_t *xattr[],
- int child_count, afr_transaction_type type);
+afr_sh_print_split_brain_log (int32_t *pending_matrix[], xlator_t *this,
+ const char *loc);
+
+int
+afr_build_pending_matrix (char **pending_key, int32_t **pending_matrix,
+ unsigned char *ignorant_subvols,
+ dict_t *xattr[], afr_transaction_type type,
+ size_t child_count);
void
afr_sh_pending_to_delta (afr_private_t *priv, dict_t **xattr,
- int32_t *delta_matrix[], int success[],
+ int32_t *delta_matrix[], unsigned char success[],
int child_count, afr_transaction_type type);
int
-afr_sh_mark_sources (afr_self_heal_t *sh, int child_count,
- afr_self_heal_type type);
+afr_mark_sources (xlator_t *this, int32_t *sources, int32_t **pending_matrix,
+ struct iatt *bufs, afr_self_heal_type type,
+ int32_t *success_children, int32_t *subvol_status);
int
-afr_sh_delta_to_xattr (afr_private_t *priv,
+afr_sh_delta_to_xattr (xlator_t *this,
int32_t *delta_matrix[], dict_t *xattr[],
int child_count, afr_transaction_type type);
-int
-afr_sh_is_matrix_zero (int32_t *pending_matrix[], int child_count);
-
void
afr_self_heal_type_str_get (afr_self_heal_t *self_heal_p, char *str,
size_t size);
+afr_self_heal_type
+afr_self_heal_type_for_transaction (afr_transaction_type type);
+
+int
+afr_build_sources (xlator_t *this, dict_t **xattr, struct iatt *bufs,
+ int32_t **pending_matrix, int32_t *sources,
+ int32_t *success_children, afr_transaction_type type,
+ int32_t *subvol_status, gf_boolean_t ignore_ignorant);
+void
+afr_sh_common_reset (afr_self_heal_t *sh, unsigned int child_count);
+
+void
+afr_sh_common_lookup_resp_handler (call_frame_t *frame, void *cookie,
+ xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ inode_t *inode, struct iatt *buf,
+ dict_t *xattr, struct iatt *postparent,
+ loc_t *loc);
+
+int
+afr_sh_common_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
+ afr_lookup_done_cbk_t lookup_cbk, uuid_t uuid,
+ int32_t flags, dict_t *xdata);
+int
+afr_sh_entry_expunge_remove (call_frame_t *expunge_frame, xlator_t *this,
+ int active_src, struct iatt *buf,
+ struct iatt *parentbuf);
+int
+afr_sh_entrylk (call_frame_t *frame, xlator_t *this, loc_t *loc,
+ char *base_name, afr_lock_cbk_t lock_cbk);
+int
+afr_sh_entry_impunge_create (call_frame_t *impunge_frame, xlator_t *this,
+ int child_index);
+int
+afr_sh_data_unlock (call_frame_t *frame, xlator_t *this, char *dom,
+ afr_lock_cbk_t lock_cbk);
+afr_local_t *
+afr_self_heal_local_init (afr_local_t *l, xlator_t *this);
+int
+afr_sh_data_lock (call_frame_t *frame, xlator_t *this,
+ off_t start, off_t len, gf_boolean_t block, char *dom,
+ afr_lock_cbk_t success_handler,
+ afr_lock_cbk_t failure_handler);
+void
+afr_sh_set_error (afr_self_heal_t *sh, int32_t op_errno);
+void
+afr_sh_mark_source_sinks (call_frame_t *frame, xlator_t *this);
+typedef int
+(*afr_fxattrop_cbk_t) (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *xattr, dict_t *xdata);
+int
+afr_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name);
+int
+afr_impunge_frame_create (call_frame_t *frame, xlator_t *this,
+ int active_source, call_frame_t **impunge_frame);
+void
+afr_sh_reset (call_frame_t *frame, xlator_t *this);
+
+void
+afr_children_intersection_get (int32_t *set1, int32_t *set2,
+ int *intersection, unsigned int child_count);
+int
+afr_get_no_xattr_dir_read_child (xlator_t *this, int32_t *success_children,
+ struct iatt *bufs);
+int
+afr_sh_erase_pending (call_frame_t *frame, xlator_t *this,
+ afr_transaction_type type, afr_fxattrop_cbk_t cbk,
+ int (*finish)(call_frame_t *frame, xlator_t *this));
+
+void
+afr_set_local_for_unhealable (afr_local_t *local);
+
+int
+is_self_heal_failed (afr_self_heal_t *sh, afr_sh_fail_check_type type);
+
+void
+afr_set_self_heal_status (afr_self_heal_t *sh, afr_self_heal_status status);
+
+void
+afr_log_self_heal_completion_status (afr_local_t *local, gf_loglevel_t logl);
+
+char*
+afr_get_pending_matrix_str (int32_t *pending_matrix[], xlator_t *this);
#endif /* __AFR_SELF_HEAL_COMMON_H__ */
diff --git a/xlators/cluster/afr/src/afr-self-heal-data.c b/xlators/cluster/afr/src/afr-self-heal-data.c
index 0398d21b1..9de26ee56 100644
--- a/xlators/cluster/afr/src/afr-self-heal-data.c
+++ b/xlators/cluster/afr/src/afr-self-heal-data.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include <libgen.h>
@@ -49,421 +40,476 @@
#include "afr-self-heal-common.h"
#include "afr-self-heal-algorithm.h"
-
int
-afr_sh_data_done (call_frame_t *frame, xlator_t *this)
+afr_sh_data_fail (call_frame_t *frame, xlator_t *this);
+
+static inline gf_boolean_t
+afr_sh_data_proceed (unsigned int success_count)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
+ return (success_count >= AFR_SH_MIN_PARTICIPANTS);
+}
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+extern int
+sh_loop_finish (call_frame_t *loop_frame, xlator_t *this);
- /*
- TODO: cleanup sh->*
- */
+int
+afr_post_sh_big_lock_success (call_frame_t *frame, xlator_t *this);
- if (sh->healing_fd && !sh->healing_fd_opened) {
- /* unref only if we created the fd ourselves */
+int
+afr_post_sh_big_lock_failure (call_frame_t *frame, xlator_t *this);
- fd_unref (sh->healing_fd);
- sh->healing_fd = NULL;
- }
+int
+afr_sh_data_finish (call_frame_t *frame, xlator_t *this);
-/* for (i = 0; i < priv->child_count; i++) */
-/* sh->locked_nodes[i] = 0; */
+int
+afr_sh_data_done (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
- gf_log (this->name, GF_LOG_TRACE,
- "self heal of %s completed",
- local->loc.path);
+ local = frame->local;
+ sh = &local->self_heal;
- sh->completion_cbk (frame, this);
+ sh->completion_cbk (frame, this);
- return 0;
+ return 0;
}
int
afr_sh_data_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
-
- int child_index = (long) cookie;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int child_index = (long) cookie;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- LOCK (&frame->lock);
- {
+ LOCK (&frame->lock);
+ {
if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "flush or setattr failed on %s on subvolume %s: %s",
+ gf_log (this->name, GF_LOG_ERROR,
+ "flush failed on %s on subvolume %s: %s",
local->loc.path, priv->children[child_index]->name,
strerror (op_errno));
}
- }
- UNLOCK (&frame->lock);
+ }
+ UNLOCK (&frame->lock);
- call_count = afr_frame_return (frame);
+ call_count = afr_frame_return (frame);
- if (call_count == 0) {
- afr_sh_data_done (frame, this);
- }
+ if (call_count == 0) {
+ afr_sh_data_done (frame, this);
+ }
- return 0;
+ return 0;
}
-
int
-afr_sh_data_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *statpre, struct iatt *statpost)
+afr_sh_data_close (call_frame_t *frame, xlator_t *this)
{
- afr_sh_data_flush_cbk (frame, cookie, this, op_ret, op_errno);
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_self_heal_t *sh = NULL;
+ int i = 0;
+ int call_count = 0;
+
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+
+ if (!sh->healing_fd) {
+ //This happens when file is non-reg
+ afr_sh_data_done (frame, this);
+ return 0;
+ }
+ call_count = afr_set_elem_count_get (sh->success,
+ priv->child_count);
+ local->call_count = call_count;
+
+ if (call_count == 0) {
+ afr_sh_data_done (frame, this);
+ return 0;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sh->success[i])
+ continue;
+ gf_log (this->name, GF_LOG_DEBUG,
+ "closing fd of %s on %s",
+ local->loc.path, priv->children[i]->name);
+
+ STACK_WIND_COOKIE (frame, afr_sh_data_flush_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->flush,
+ sh->healing_fd, NULL);
+
+ if (!--call_count)
+ break;
+ }
return 0;
}
-
int
-afr_sh_data_close (call_frame_t *frame, xlator_t *this)
+afr_sh_dom_unlock (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_self_heal_t *sh = NULL;
-
- int i = 0;
- int call_count = 0;
- int source = 0;
- int32_t valid = 0;
-
- struct iatt stbuf = {0,};
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
local = frame->local;
sh = &local->self_heal;
priv = this->private;
- source = sh->source;
+ if (sh->sh_dom_lock_held)
+ afr_sh_data_unlock (frame, this, priv->sh_domain,
+ afr_sh_data_close);
+ else
+ afr_sh_data_close (frame, this);
+ return 0;
+}
- valid |= (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME);
+int
+afr_sh_data_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
+{
- stbuf.ia_atime = sh->buf[source].ia_atime;
- stbuf.ia_atime_nsec = sh->buf[source].ia_atime_nsec;
- stbuf.ia_mtime = sh->buf[source].ia_mtime;
- stbuf.ia_mtime_nsec = sh->buf[source].ia_mtime_nsec;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int child_index = (long) cookie;
- if (sh->healing_fd_opened) {
- /* not our job to close the fd */
+ local = frame->local;
+ priv = this->private;
- afr_sh_data_done (frame, this);
- return 0;
+ LOCK (&frame->lock);
+ {
+ if (op_ret == -1) {
+ gf_log (this->name, GF_LOG_INFO,
+ "setattr failed on %s on subvolume %s: %s",
+ local->loc.path, priv->children[child_index]->name,
+ strerror (op_errno));
+ }
}
+ UNLOCK (&frame->lock);
- if (!sh->healing_fd) {
- afr_sh_data_done (frame, this);
- return 0;
- }
-
- call_count = (sh->active_sinks + 1) * 2;
- local->call_count = call_count;
+ call_count = afr_frame_return (frame);
- /* closed source */
- gf_log (this->name, GF_LOG_TRACE,
- "closing fd of %s on %s",
- local->loc.path, priv->children[sh->source]->name);
-
- STACK_WIND_COOKIE (frame, afr_sh_data_flush_cbk,
- (void *) (long) sh->source,
- priv->children[sh->source],
- priv->children[sh->source]->fops->flush,
- sh->healing_fd);
- call_count--;
+ if (call_count == 0) {
+ afr_sh_data_finish (frame, this);
+ }
- STACK_WIND_COOKIE (frame, afr_sh_data_setattr_cbk,
- (void *) (long) sh->source,
- priv->children[sh->source],
- priv->children[sh->source]->fops->setattr,
- &local->loc, &stbuf, valid);
+ return 0;
+}
- call_count--;
+int
+afr_sh_data_setattr (call_frame_t *frame, xlator_t *this, struct iatt* stbuf)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_self_heal_t *sh = NULL;
+ int i = 0;
+ int call_count = 0;
+ int32_t valid = 0;
- if (call_count == 0)
- return 0;
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (sh->sources[i] || !local->child_up[i])
- continue;
+ valid = (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME);
- gf_log (this->name, GF_LOG_TRACE,
- "closing fd of %s on %s",
- local->loc.path, priv->children[i]->name);
+ call_count = afr_set_elem_count_get (sh->success,
+ priv->child_count);
+ local->call_count = call_count;
- STACK_WIND_COOKIE (frame, afr_sh_data_flush_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->flush,
- sh->healing_fd);
+ if (call_count == 0) {
+ GF_ASSERT (0);
+ afr_sh_data_finish (frame, this);
+ return 0;
+ }
- call_count--;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sh->success[i])
+ continue;
STACK_WIND_COOKIE (frame, afr_sh_data_setattr_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->setattr,
- &local->loc, &stbuf, valid);
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->setattr,
+ &local->loc, stbuf, valid, NULL);
- if (!--call_count)
- break;
- }
+ if (!--call_count)
+ break;
+ }
- return 0;
+ return 0;
}
-
int
-afr_sh_data_unlck_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+afr_sh_data_setattr_fstat_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ struct iatt *buf, dict_t *xdata)
{
- afr_local_t * local = NULL;
- int call_count = 0;
- int child_index = (long) cookie;
-
-
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "locking inode of %s on child %d failed: %s",
- local->loc.path, child_index,
- strerror (op_errno));
- } else {
- gf_log (this->name, GF_LOG_TRACE,
- "inode of %s on child %d locked",
- local->loc.path, child_index);
- }
- }
- UNLOCK (&frame->lock);
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ int child_index = (long) cookie;
- call_count = afr_frame_return (frame);
+ local = frame->local;
+ sh = &local->self_heal;
- if (call_count == 0) {
- afr_sh_data_close (frame, this);
- }
+ GF_ASSERT (sh->source == child_index);
+ if (op_ret != -1) {
+ sh->buf[child_index] = *buf;
+ afr_sh_data_setattr (frame, this, buf);
+ } else {
+ gf_log (this->name, GF_LOG_ERROR, "%s: Failed to set "
+ "time-stamps after self-heal", local->loc.path);
+ afr_sh_data_fail (frame, this);
+ }
- return 0;
+ return 0;
}
+/*
+ * If there are any writes after the self-heal is triggered then the
+ * stbuf stored in local->self_heal.buf[] will be invalid so we do one more
+ * stat on the source and then set the [am]times
+ */
+int
+afr_sh_set_timestamps (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_self_heal_t *sh = NULL;
+
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+
+ STACK_WIND_COOKIE (frame, afr_sh_data_setattr_fstat_cbk,
+ (void *) (long) sh->source,
+ priv->children[sh->source],
+ priv->children[sh->source]->fops->fstat,
+ sh->healing_fd, NULL);
+ return 0;
+}
+//Fun fact, lock_cbk is being used for both lock & unlock
int
-afr_sh_data_unlock (call_frame_t *frame, xlator_t *this)
+afr_sh_data_unlock (call_frame_t *frame, xlator_t *this, char *dom,
+ afr_lock_cbk_t lock_cbk)
{
afr_local_t *local = NULL;
afr_internal_lock_t *int_lock = NULL;
- afr_self_heal_t *sh = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int ret = 0;
local = frame->local;
int_lock = &local->internal_lock;
- sh = &local->self_heal;
-
- GF_ASSERT (!sh->data_lock_held);
+ sh = &local->self_heal;
+ priv = this->private;
- int_lock->lock_cbk = afr_sh_data_close;
+ if (strcmp (dom, this->name) == 0) {
+ sh->data_lock_held = _gf_false;
+ } else if (strcmp (dom, priv->sh_domain) == 0) {
+ sh->sh_dom_lock_held = _gf_false;
+ } else {
+ ret = -1;
+ goto out;
+ }
+ int_lock->lock_cbk = lock_cbk;
+ int_lock->domain = dom;
afr_unlock (frame, this);
- return 0;
+out:
+ if (ret) {
+ int_lock->lock_op_ret = -1;
+ int_lock->lock_cbk (frame, this);
+ }
+ return 0;
}
-
int
afr_sh_data_finish (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
- local = frame->local;
- sh = &local->self_heal;
+ local = frame->local;
+ sh = &local->self_heal;
- gf_log (this->name, GF_LOG_TRACE,
- "finishing data selfheal of %s", local->loc.path);
+ gf_log (this->name, GF_LOG_DEBUG,
+ "finishing data selfheal of %s", local->loc.path);
- if (!sh->data_lock_held)
- afr_sh_data_unlock (frame, this);
+ if (sh->data_lock_held)
+ afr_sh_data_unlock (frame, this, this->name, afr_sh_dom_unlock);
else
- afr_sh_data_close (frame, this);
+ afr_sh_dom_unlock (frame, this);
- return 0;
+ return 0;
}
-
int
-afr_sh_data_erase_pending_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xattr)
+afr_sh_data_fail (call_frame_t *frame, xlator_t *this)
{
- int call_count = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
- call_count = afr_frame_return (frame);
+ local = frame->local;
+ sh = &local->self_heal;
- if (call_count == 0)
- afr_sh_data_finish (frame, this);
+ gf_log (this->name, GF_LOG_DEBUG,
+ "finishing failed data selfheal of %s", local->loc.path);
- return 0;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_data_finish (frame, this);
+ return 0;
}
-
int
-afr_sh_data_erase_pending (call_frame_t *frame, xlator_t *this)
+afr_sh_data_erase_pending_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xattr, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
- int i = 0;
- dict_t **erase_xattr = NULL;
-
-
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
-
- afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix, sh->success,
- priv->child_count, AFR_DATA_TRANSACTION);
-
- erase_xattr = GF_CALLOC (sizeof (*erase_xattr), priv->child_count,
- gf_afr_mt_dict_t);
+ int call_count = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int32_t child_index = (long) cookie;
- for (i = 0; i < priv->child_count; i++) {
- if (sh->xattr[i]) {
- call_count++;
-
- erase_xattr[i] = get_new_dict();
- dict_ref (erase_xattr[i]);
- }
- }
-
- afr_sh_delta_to_xattr (priv, sh->delta_matrix, erase_xattr,
- priv->child_count, AFR_DATA_TRANSACTION);
-
- local->call_count = call_count;
- for (i = 0; i < priv->child_count; i++) {
- if (!erase_xattr[i])
- continue;
-
- gf_log (this->name, GF_LOG_TRACE,
- "erasing pending flags from %s on %s",
- local->loc.path, priv->children[i]->name);
-
- STACK_WIND_COOKIE (frame, afr_sh_data_erase_pending_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->fxattrop,
- sh->healing_fd,
- GF_XATTROP_ADD_ARRAY, erase_xattr[i]);
- if (!--call_count)
- break;
- }
+ priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
+ if (op_ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "Erasing of pending change "
+ "log failed on %s for subvol %s, reason: %s",
+ local->loc.path, priv->children[child_index]->name,
+ strerror (op_errno));
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (erase_xattr[i]) {
- dict_unref (erase_xattr[i]);
- }
- }
- GF_FREE (erase_xattr);
+ call_count = afr_frame_return (frame);
- return 0;
+ if (call_count == 0) {
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+ if (sh->old_loop_frame)
+ sh_loop_finish (sh->old_loop_frame, this);
+ sh->old_loop_frame = NULL;
+ afr_sh_data_fail (frame, this);
+ goto out;
+ }
+ if (!IA_ISREG (sh->type)) {
+ afr_sh_data_finish (frame, this);
+ goto out;
+ }
+ GF_ASSERT (sh->old_loop_frame);
+ afr_sh_data_lock (frame, this, 0, 0, _gf_true, this->name,
+ afr_post_sh_big_lock_success,
+ afr_post_sh_big_lock_failure);
+ }
+out:
+ return 0;
}
-
int
-afr_sh_data_trim_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
+afr_sh_data_erase_pending (call_frame_t *frame, xlator_t *this)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- int call_count = 0;
- int child_index = 0;
-
- priv = this->private;
- local = frame->local;
-
- child_index = (long) cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1)
- gf_log (this->name, GF_LOG_DEBUG,
- "ftruncate of %s on subvolume %s failed (%s)",
- local->loc.path,
- priv->children[child_index]->name,
- strerror (op_errno));
- else
- gf_log (this->name, GF_LOG_TRACE,
- "ftruncate of %s on subvolume %s completed",
- local->loc.path,
- priv->children[child_index]->name);
- }
- UNLOCK (&frame->lock);
-
- call_count = afr_frame_return (frame);
-
- if (call_count == 0) {
- afr_sh_data_erase_pending (frame, this);
- }
-
- return 0;
+ afr_sh_erase_pending (frame, this, AFR_DATA_TRANSACTION,
+ afr_sh_data_erase_pending_cbk,
+ afr_sh_data_finish);
+ return 0;
}
-
int
-afr_sh_data_trim_sinks (call_frame_t *frame, xlator_t *this)
+afr_sh_data_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *pre,
+ struct iatt *post, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- afr_self_heal_t *sh = NULL;
- int *sources = NULL;
- int call_count = 0;
- int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_self_heal_t *sh = NULL;
+ int call_count = 0;
+ int child_index = (long) cookie;
+ local = frame->local;
+ priv = this->private;
+ sh = &local->self_heal;
+
+ if (op_ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "%s: Failed to fsync on "
+ "%s - %s", local->loc.path,
+ priv->children[child_index]->name, strerror (op_errno));
+ LOCK (&frame->lock);
+ {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ }
+ UNLOCK (&frame->lock);
+ if (sh->old_loop_frame)
+ sh_loop_finish (sh->old_loop_frame, this);
+ sh->old_loop_frame = NULL;
+ }
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
+ call_count = afr_frame_return (frame);
+ if (call_count == 0) {
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC))
+ afr_sh_data_fail (frame, this);
+ else
+ afr_sh_data_erase_pending (frame, this);
+ }
+ return 0;
+}
- sources = sh->sources;
- call_count = sh->active_sinks;
+/*
+ * Before erasing xattrs, make sure the data is written to disk
+ */
+int
+afr_sh_data_fsync (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_self_heal_t *sh = NULL;
+ int i = 0;
+ int call_count = 0;
- local->call_count = call_count;
+ local = frame->local;
+ priv = this->private;
+ sh = &local->self_heal;
- for (i = 0; i < priv->child_count; i++) {
- if (sources[i] || !local->child_up[i])
- continue;
+ call_count = sh->active_sinks;
+ if (call_count == 0) {
+ afr_sh_data_erase_pending (frame, this);
+ return 0;
+ }
- STACK_WIND_COOKIE (frame, afr_sh_data_trim_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->ftruncate,
- sh->healing_fd, sh->file_size);
+ local->call_count = call_count;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sh->success[i] || sh->sources[i])
+ continue;
- if (!--call_count)
- break;
- }
+ STACK_WIND_COOKIE (frame, afr_sh_data_fsync_cbk,
+ (void *) (long) i, priv->children[i],
+ priv->children[i]->fops->fsync,
+ sh->healing_fd, 1, NULL);
+ }
- return 0;
+ return 0;
}
-
static struct afr_sh_algorithm *
sh_algo_from_name (xlator_t *this, char *name)
{
int i = 0;
+ if (name == NULL)
+ goto out;
+
while (afr_self_heal_algorithms[i].name) {
if (!strcmp (name, afr_self_heal_algorithms[i].name)) {
return &afr_self_heal_algorithms[i];
@@ -472,17 +518,22 @@ sh_algo_from_name (xlator_t *this, char *name)
i++;
}
+out:
return NULL;
}
static int
-sh_zero_byte_files_exist (afr_self_heal_t *sh, int child_count)
+sh_zero_byte_files_exist (afr_local_t *local, int child_count)
{
- int i;
- int ret = 0;
+ int i = 0;
+ int ret = 0;
+ afr_self_heal_t *sh = NULL;
+ sh = &local->self_heal;
for (i = 0; i < child_count; i++) {
+ if (!local->child_up[i] || sh->child_errno[i])
+ continue;
if (sh->buf[i].ia_size == 0) {
ret = 1;
break;
@@ -509,9 +560,9 @@ afr_sh_data_pick_algo (call_frame_t *frame, xlator_t *this)
if (algo == NULL) {
/* option not set, so fall back on heuristics */
- if ((local->enoent_count != 0)
- || sh_zero_byte_files_exist (sh, priv->child_count)
- || (sh->file_size <= (priv->data_self_heal_window_size * this->ctx->page_size))) {
+ if (sh_zero_byte_files_exist (local, priv->child_count)
+ || (sh->file_size <= (priv->data_self_heal_window_size *
+ this->ctx->page_size))) {
/*
* If the file does not exist on one of the subvolumes,
@@ -540,398 +591,860 @@ afr_sh_data_pick_algo (call_frame_t *frame, xlator_t *this)
int
afr_sh_data_sync_prepare (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int active_sinks = 0;
- int source = 0;
- int i = 0;
-
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
struct afr_sh_algorithm *sh_algo = NULL;
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
- source = sh->source;
+ sh->algo_completion_cbk = afr_sh_data_fsync;
+ sh->algo_abort_cbk = afr_sh_data_fail;
- for (i = 0; i < priv->child_count; i++) {
- if (sh->sources[i] == 0 && local->child_up[i] == 1) {
- active_sinks++;
- sh->success[i] = 1;
- }
- }
- sh->success[source] = 1;
-
- if (active_sinks == 0) {
- gf_log (this->name, GF_LOG_TRACE,
- "no active sinks for performing self-heal on file %s",
- local->loc.path);
- afr_sh_data_finish (frame, this);
- return 0;
- }
- sh->active_sinks = active_sinks;
+ sh_algo = afr_sh_data_pick_algo (frame, this);
- gf_log (this->name, GF_LOG_DEBUG,
- "self-healing file %s from subvolume %s to %d other",
- local->loc.path, priv->children[source]->name, active_sinks);
+ sh->algo = sh_algo;
+ sh_algo->fn (frame, this);
- sh->algo_completion_cbk = afr_sh_data_trim_sinks;
- sh->algo_abort_cbk = afr_sh_data_finish;
+ return 0;
+}
- sh_algo = afr_sh_data_pick_algo (frame, this);
+int
+afr_sh_data_trim_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ int call_count = 0;
+ int child_index = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
- sh_algo->fn (frame, this);
+ priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
+
+ child_index = (long) cookie;
+
+ LOCK (&frame->lock);
+ {
+ if (op_ret == -1) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "ftruncate of %s on subvolume %s failed (%s)",
+ local->loc.path,
+ priv->children[child_index]->name,
+ strerror (op_errno));
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ } else {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "ftruncate of %s on subvolume %s completed",
+ local->loc.path,
+ priv->children[child_index]->name);
+ }
+ }
+ UNLOCK (&frame->lock);
+
+ call_count = afr_frame_return (frame);
+
+ if (call_count == 0) {
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC))
+ afr_sh_data_fail (frame, this);
+ else
+ afr_sh_data_sync_prepare (frame, this);
+ }
- return 0;
+ return 0;
}
int
-afr_sh_data_fix (call_frame_t *frame, xlator_t *this)
+afr_sh_data_trim_sinks (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_local_t * orig_local = NULL;
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ afr_self_heal_t *sh = NULL;
+ int *sources = NULL;
+ int call_count = 0;
+ int i = 0;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int nsources = 0;
- int source = 0;
- int i = 0;
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
- afr_sh_build_pending_matrix (priv, sh->pending_matrix, sh->xattr,
- priv->child_count, AFR_DATA_TRANSACTION);
+ sources = sh->sources;
+ call_count = sh->active_sinks;
- afr_sh_print_pending_matrix (sh->pending_matrix, this);
+ local->call_count = call_count;
- nsources = afr_sh_mark_sources (sh, priv->child_count,
- AFR_SELF_HEAL_DATA);
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i] || !local->child_up[i])
+ continue;
- afr_sh_supress_errenous_children (sh->sources, sh->child_errno,
- priv->child_count);
+ STACK_WIND_COOKIE (frame, afr_sh_data_trim_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->ftruncate,
+ sh->healing_fd, sh->file_size,
+ NULL);
- if (nsources == 0) {
- gf_log (this->name, GF_LOG_TRACE,
- "No self-heal needed for %s",
- local->loc.path);
+ if (!--call_count)
+ break;
+ }
- afr_sh_data_finish (frame, this);
- return 0;
+ return 0;
+}
+
+int
+afr_sh_inode_set_read_ctx (afr_self_heal_t *sh, xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ int ret = 0;
+ int i = 0;
+
+ priv = this->private;
+ sh->source = afr_sh_select_source (sh->sources, priv->child_count);
+ if (sh->source < 0) {
+ ret = -1;
+ goto out;
}
- if ((nsources == -1)
- && (priv->favorite_child != -1)
- && (sh->child_errno[priv->favorite_child] == 0)) {
+ /* detect changes not visible through pending flags -- JIC */
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == sh->source || sh->child_errno[i])
+ continue;
+
+ if (SIZE_DIFFERS (&sh->buf[i], &sh->buf[sh->source]))
+ sh->sources[i] = 0;
+ }
- gf_log (this->name, GF_LOG_DEBUG,
- "Picking favorite child %s as authentic source to resolve conflicting data of %s",
- priv->children[priv->favorite_child]->name,
- local->loc.path);
+ afr_reset_children (sh->fresh_children, priv->child_count);
+ afr_get_fresh_children (sh->success_children, sh->sources,
+ sh->fresh_children, priv->child_count);
+ afr_inode_set_read_ctx (this, sh->inode, sh->source,
+ sh->fresh_children);
+out:
+ return ret;
+}
- sh->sources[priv->favorite_child] = 1;
+char*
+afr_get_sizes_str (afr_local_t *local, struct iatt *bufs, xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ int i = 0;
+ char num[1024] = {0};
+ size_t len = 0;
+ char *sizes_str = NULL;
+ size_t off = 0;
+ char *fmt_str = "%llu bytes on %s, ";
+ char *child_down = " %s,";
+ char *child_unknown = " %s,";
+ int down_child_present = 0;
+ int down_count = 0;
+ int unknown_count = 0;
+ int unknown_child_present = 0;
+ char *down_subvol_1 = " down subvolume is ";
+ char *unknown_subvol_1 = " unknown subvolume is ";
+ char *down_subvol_2 = " down subvolumes are ";
+ char *unknown_subvol_2 = " unknown subvolumes are ";
+
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i] == 1) {
+ len += snprintf (num, sizeof (num), fmt_str,
+ (unsigned long long) bufs[i].ia_size,
+ priv->children[i]->name);
+ } else if (local->child_up[i] == 0) {
+ len += snprintf (num, sizeof (num), child_down,
+ priv->children[i]->name);
+ if (!down_child_present)
+ down_child_present = 1;
+ down_count ++;
+ } else if (local->child_up[i] == -1) {
+ len += snprintf (num, sizeof (num), child_unknown,
+ priv->children[i]->name);
+ if (!unknown_child_present)
+ unknown_child_present = 1;
+ unknown_count++;
+ }
- nsources = afr_sh_source_count (sh->sources,
- priv->child_count);
- }
+ }
- if (nsources == -1) {
- gf_log (this->name, GF_LOG_ERROR,
- "Unable to self-heal contents of '%s' (possible split-brain). "
- "Please delete the file from all but the preferred "
- "subvolume.", local->loc.path);
+ if (down_child_present) {
+ if (down_count > 1)
+ len += snprintf (num, sizeof (num), "%s",
+ down_subvol_2);
+ else
+ len += snprintf (num, sizeof (num), "%s",
+ down_subvol_1);
+ }
+ if (unknown_child_present) {
+ if (unknown_count > 1)
+ len += snprintf (num, sizeof (num), "%s",
+ unknown_subvol_2);
+ else
+ len += snprintf (num, sizeof (num), "%s",
+ unknown_subvol_1);
+ }
- local->govinda_gOvinda = 1;
+ len++;//for '\0'
- afr_sh_data_finish (frame, this);
- return 0;
- }
+ sizes_str = GF_CALLOC (len, sizeof (char), gf_common_mt_char);
- source = afr_sh_select_source (sh->sources, priv->child_count);
-
- if (source == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "No active sources found.");
+ if (!sizes_str)
+ return NULL;
- afr_sh_data_finish (frame, this);
- return 0;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i] == 1) {
+ off += snprintf (sizes_str + off, len - off, fmt_str,
+ (unsigned long long) bufs[i].ia_size,
+ priv->children[i]->name);
+ }
}
- sh->source = source;
- sh->block_size = 65536;
- sh->file_size = sh->buf[source].ia_size;
+ if (down_child_present) {
+ if (down_count > 1) {
+ off += snprintf (sizes_str + off, len - off, "%s",
+ down_subvol_2);
+ } else {
+ off += snprintf (sizes_str + off, len - off, "%s",
+ down_subvol_1);
+ }
+ }
- if (FILE_HAS_HOLES (&sh->buf[source]))
- sh->file_has_holes = 1;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i] == 0) {
+ off += snprintf (sizes_str + off, len - off, child_down,
+ priv->children[i]->name);
+ }
+ }
- orig_local = sh->orig_frame->local;
- orig_local->cont.lookup.buf.ia_size = sh->buf[source].ia_size;
+ if (unknown_child_present) {
+ if (unknown_count > 1) {
+ off += snprintf (sizes_str + off, len - off, "%s",
+ unknown_subvol_2);
+ } else {
+ off += snprintf (sizes_str + off, len - off, "%s",
+ unknown_subvol_1);
+ }
+ }
- /* detect changes not visible through pending flags -- JIC */
- for (i = 0; i < priv->child_count; i++) {
- if (i == source || sh->child_errno[i])
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i] == -1) {
+ off += snprintf (sizes_str + off, len - off,
+ child_unknown,
+ priv->children[i]->name);
- if (SIZE_DIFFERS (&sh->buf[i], &sh->buf[source]))
- sh->sources[i] = 0;
- }
+ }
+ }
- afr_set_read_child (this, local->loc.inode, sh->source);
+ return sizes_str;
+}
- /*
- quick-read might have read the file, so send xattr from
- the source subvolume (http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=815)
- */
+char*
+afr_get_sinks_str (xlator_t *this, afr_local_t *local, afr_self_heal_t *sh)
+{
+ afr_private_t *priv = NULL;
+ int i = 0;
+ char num[1024] = {0};
+ size_t len = 0;
+ char *sinks_str = NULL;
+ char *temp_str = " to sinks ";
+ char *str_format = " %s,";
+ char off = 0;
+
+ priv = this->private;
+
+ len += snprintf (num, sizeof (num), "%s", temp_str);
+ for (i = 0; i < priv->child_count; i++) {
+ if ((sh->sources[i] == 0) && (local->child_up[i] == 1)) {
+ len += snprintf (num, sizeof (num), str_format,
+ priv->children[i]->name);
+ }
+ }
- dict_unref (orig_local->cont.lookup.xattr);
- if (orig_local->cont.lookup.xattrs)
- orig_local->cont.lookup.xattr = dict_ref (orig_local->cont.lookup.xattrs[sh->source]);
+ len ++;
- if (sh->background) {
- sh->unwind (sh->orig_frame, this);
- sh->unwound = _gf_true;
+ sinks_str = GF_CALLOC (len, sizeof (char), gf_common_mt_char);
+
+ if (!sinks_str)
+ return NULL;
+
+ off += snprintf (sinks_str + off, len - off, "%s", temp_str);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if ((sh->sources[i] == 0) && (local->child_up[i] == 1)) {
+ off += snprintf (sinks_str + off, len - off,
+ str_format,
+ priv->children[i]->name);
+ }
}
- afr_sh_data_sync_prepare (frame, this);
+ return sinks_str;
- return 0;
}
+void
+afr_set_data_sh_info_str (afr_local_t *local, afr_self_heal_t *sh, xlator_t *this)
+{
+ char *pending_matrix_str = NULL;
+ char *sizes_str = NULL;
+ char *sinks_str = NULL;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+
+ pending_matrix_str = afr_get_pending_matrix_str (sh->pending_matrix,
+ this);
+ if (!pending_matrix_str)
+ pending_matrix_str = "";
+
+ sizes_str = afr_get_sizes_str (local, sh->buf, this);
+ if (!sizes_str)
+ sizes_str = "";
+
+ sinks_str = afr_get_sinks_str (this, local, sh);
+ if (!sinks_str)
+ sinks_str = "";
+
+ gf_asprintf (&sh->data_sh_info, " data self heal from %s %s with "
+ "%s data %s", priv->children[sh->source]->name, sinks_str,
+ sizes_str, pending_matrix_str);
+
+ if (pending_matrix_str && strcmp (pending_matrix_str, ""))
+ GF_FREE (pending_matrix_str);
+
+ if (sizes_str && strcmp (sizes_str, ""))
+ GF_FREE (sizes_str);
+}
+
+void
+afr_sh_data_fix (call_frame_t *frame, xlator_t *this)
+{
+ int source = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+
+ source = sh->source;
+ sh->block_size = this->ctx->page_size;
+ sh->file_size = sh->buf[source].ia_size;
+
+ if (FILE_HAS_HOLES (&sh->buf[source]))
+ sh->file_has_holes = 1;
+
+ if (sh->background && sh->unwind && !sh->unwound) {
+ sh->unwind (sh->orig_frame, this, sh->op_ret, sh->op_errno,
+ is_self_heal_failed (sh, AFR_CHECK_SPECIFIC));
+ sh->unwound = _gf_true;
+ }
+
+ afr_sh_mark_source_sinks (frame, this);
+ if (sh->active_sinks == 0) {
+ gf_log (this->name, GF_LOG_INFO,
+ "no active sinks for performing self-heal on file %s",
+ local->loc.path);
+ afr_sh_data_finish (frame, this);
+ return;
+ }
+
+ gf_log (this->name, GF_LOG_DEBUG,
+ "self-healing file %s from subvolume %s to %d other",
+ local->loc.path, priv->children[sh->source]->name,
+ sh->active_sinks);
+
+ sh->actual_sh_started = _gf_true;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_SYNC_BEGIN);
+ afr_sh_data_trim_sinks (frame, this);
+}
+
int
-afr_self_heal_get_source (xlator_t *this, afr_local_t *local, dict_t **xattr)
+afr_sh_data_fxattrop_fstat_done (call_frame_t *frame, xlator_t *this)
{
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int nsources = 0;
+ int ret = 0;
+ int *old_sources = NULL;
+ int tstamp_source = 0;
+ int i = 0;
- int source = 0;
- int i = 0;
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+
+ gf_log (this->name, GF_LOG_DEBUG, "Pending matrix for: %s",
+ lkowner_utoa (&frame->root->lk_owner));
+ if (sh->sync_done) {
+ //store sources before sync so that mtime can be set using the
+ //iatt buf from one of them.
+ old_sources = alloca (priv->child_count*sizeof (*old_sources));
+ memcpy (old_sources, sh->sources,
+ priv->child_count * sizeof (*old_sources));
+ }
- sh = &local->self_heal;
- priv = this->private;
+ nsources = afr_build_sources (this, sh->xattr, sh->buf, sh->pending_matrix,
+ sh->sources, sh->success_children,
+ AFR_DATA_TRANSACTION, NULL, _gf_true);
+ if ((nsources == -1)
+ && (priv->favorite_child != -1)
+ && (sh->child_errno[priv->favorite_child] == 0)) {
- sh->pending_matrix = GF_CALLOC (sizeof (int32_t *), priv->child_count,
- gf_afr_mt_int32_t);
- for (i = 0; i < priv->child_count; i++) {
- sh->pending_matrix[i] = GF_CALLOC (sizeof (int32_t),
- priv->child_count,
- gf_afr_mt_int32_t);
- }
+ gf_log (this->name, GF_LOG_DEBUG,
+ "Picking favorite child %s as authentic source to "
+ "resolve conflicting data of %s",
+ priv->children[priv->favorite_child]->name,
+ local->loc.path);
+
+ sh->sources[priv->favorite_child] = 1;
+
+ nsources = afr_sh_source_count (sh->sources,
+ priv->child_count);
+ }
+
+ if (nsources == -1) {
+ afr_sh_print_split_brain_log (sh->pending_matrix, this,
+ local->loc.path);
+ afr_set_split_brain (this, sh->inode, DONT_KNOW, SPB);
- sh->sources = GF_CALLOC (priv->child_count, sizeof (*sh->sources),
- gf_afr_mt_int32_t);
+ afr_sh_data_fail (frame, this);
+ return 0;
+ }
+
+ afr_set_split_brain (this, sh->inode, DONT_KNOW, NO_SPB);
+
+ ret = afr_sh_inode_set_read_ctx (sh, this);
+ if (ret) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "No active sources found.");
- afr_sh_build_pending_matrix (priv, sh->pending_matrix, xattr,
- priv->child_count, AFR_DATA_TRANSACTION);
+ afr_sh_data_fail (frame, this);
+ return 0;
+ }
- (void)afr_sh_mark_sources (sh, priv->child_count, AFR_SELF_HEAL_DATA);
+ if (sh->sync_done) {
+ /* Perform setattr from one of the old_sources if possible
+ * Because only they have the correct mtime, the new sources
+ * (i.e. old sinks) have mtime from last writev in sync.
+ */
+ tstamp_source = sh->source;
+ for (i = 0; i < priv->child_count; i++) {
+ if (old_sources[i] && sh->sources[i])
+ tstamp_source = i;
+ }
+ afr_sh_data_setattr (frame, this, &sh->buf[tstamp_source]);
+ } else {
+ afr_set_data_sh_info_str (local, sh, this);
+ if (nsources == 0) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "No self-heal needed for %s",
+ local->loc.path);
- source = afr_sh_select_source (sh->sources, priv->child_count);
+ afr_sh_data_finish (frame, this);
+ return 0;
+ }
- return source;
+ if (sh->do_data_self_heal &&
+ afr_data_self_heal_enabled (priv->data_self_heal))
+ afr_sh_data_fix (frame, this);
+ else
+ afr_sh_data_finish (frame, this);
+ }
+ return 0;
}
+int
+afr_lookup_select_read_child_by_txn_type (xlator_t *this, afr_local_t *local,
+ dict_t **xattr,
+ afr_transaction_type txn_type,
+ uuid_t gfid)
+{
+ afr_private_t *priv = NULL;
+ int read_child = -1;
+ int32_t **pending_matrix = NULL;
+ int32_t *sources = NULL;
+ int32_t *success_children = NULL;
+ struct iatt *bufs = NULL;
+ int32_t nsources = 0;
+ int32_t prev_read_child = -1;
+ int32_t config_read_child = -1;
+ int32_t subvol_status = 0;
+
+ priv = this->private;
+ bufs = local->cont.lookup.bufs;
+ success_children = local->cont.lookup.success_children;
+
+ pending_matrix = local->cont.lookup.pending_matrix;
+ sources = local->cont.lookup.sources;
+ memset (sources, 0, sizeof (*sources) * priv->child_count);
+
+ nsources = afr_build_sources (this, xattr, bufs, pending_matrix,
+ sources, success_children, txn_type,
+ &subvol_status, _gf_false);
+ if (subvol_status & SPLIT_BRAIN) {
+ gf_log (this->name, GF_LOG_DEBUG, "%s: Possible split-brain",
+ local->loc.path);
+ switch (txn_type) {
+ case AFR_DATA_TRANSACTION:
+ local->cont.lookup.possible_spb = _gf_true;
+ nsources = 1;
+ sources[success_children[0]] = 1;
+ break;
+ case AFR_ENTRY_TRANSACTION:
+ read_child = afr_get_no_xattr_dir_read_child (this,
+ success_children,
+ bufs);
+ sources[read_child] = 1;
+ nsources = 1;
+ break;
+ default:
+ break;
+ }
+ }
+ if (nsources < 0)
+ goto out;
+
+ prev_read_child = local->read_child_index;
+ config_read_child = priv->read_child;
+ read_child = afr_select_read_child_from_policy (success_children,
+ priv->child_count,
+ prev_read_child,
+ config_read_child,
+ sources,
+ priv->hash_mode, gfid);
+out:
+ gf_log (this->name, GF_LOG_DEBUG, "returning read_child: %d",
+ read_child);
+ return read_child;
+}
int
afr_sh_data_fstat_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct iatt *buf)
+ struct iatt *buf, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
-
- int call_count = -1;
- int child_index = (long) cookie;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ int call_count = -1;
+ int child_index = (long) cookie;
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- LOCK (&frame->lock);
- {
- if (op_ret != -1) {
+ LOCK (&frame->lock);
+ {
+ if (op_ret != -1) {
gf_log (this->name, GF_LOG_TRACE,
"fstat of %s on %s succeeded",
local->loc.path,
priv->children[child_index]->name);
- sh->buf[child_index] = *buf;
- }
- }
- UNLOCK (&frame->lock);
-
- call_count = afr_frame_return (frame);
-
- if (call_count == 0) {
- afr_sh_data_fix (frame, this);
- }
-
- return 0;
+ sh->buf[child_index] = *buf;
+ sh->success_children[sh->success_count] = child_index;
+ sh->success_count++;
+ } else {
+ gf_log (this->name, GF_LOG_ERROR, "%s: fstat failed "
+ "on %s, reason %s", local->loc.path,
+ priv->children[child_index]->name,
+ strerror (op_errno));
+ sh->child_errno[child_index] = op_errno;
+ }
+ }
+ UNLOCK (&frame->lock);
+
+ call_count = afr_frame_return (frame);
+
+ if (call_count == 0) {
+ /* Previous versions of glusterfs might have set
+ * the pending data xattrs which need to be erased
+ */
+ if (!afr_sh_data_proceed (sh->success_count)) {
+ gf_log (this->name, GF_LOG_ERROR, "inspecting metadata "
+ "succeeded on < %d children, aborting "
+ "self-heal for %s", AFR_SH_MIN_PARTICIPANTS,
+ local->loc.path);
+ afr_sh_data_fail (frame, this);
+ goto out;
+ }
+ afr_sh_data_fxattrop_fstat_done (frame, this);
+ }
+out:
+ return 0;
}
int
afr_sh_data_fstat (call_frame_t *frame, xlator_t *this)
{
- afr_self_heal_t *sh = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int i = 0;
+ int child = 0;
+ int32_t *fstat_children = NULL;
- int call_count = 0;
- int i = 0;
-
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
-
- call_count = afr_up_children_count (priv->child_count,
- local->child_up);
-
- local->call_count = call_count;
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_sh_data_fstat_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->fstat,
- sh->healing_fd);
-
- if (!--call_count)
- break;
- }
- }
+ priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
- return 0;
+ fstat_children = memdup (sh->success_children,
+ sizeof (*fstat_children) * priv->child_count);
+ if (!fstat_children) {
+ afr_sh_data_fail (frame, this);
+ goto out;
+ }
+ call_count = sh->success_count;
+ local->call_count = call_count;
+
+ memset (sh->buf, 0, sizeof (*sh->buf) * priv->child_count);
+ afr_reset_children (sh->success_children, priv->child_count);
+ sh->success_count = 0;
+ for (i = 0; i < priv->child_count; i++) {
+ child = fstat_children[i];
+ if (child == -1)
+ break;
+ STACK_WIND_COOKIE (frame, afr_sh_data_fstat_cbk,
+ (void *) (long) child,
+ priv->children[child],
+ priv->children[child]->fops->fstat,
+ sh->healing_fd, NULL);
+ --call_count;
+ }
+ GF_ASSERT (!call_count);
+out:
+ GF_FREE (fstat_children);
+ return 0;
}
-
-int
-afr_sh_data_fxattrop_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xattr)
+void
+afr_sh_common_fxattrop_resp_handler (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xattr)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
-
- int call_count = -1;
- int child_index = (long) cookie;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ int child_index = (long) cookie;
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- LOCK (&frame->lock);
- {
- if (op_ret != -1) {
+ LOCK (&frame->lock);
+ {
+ if (op_ret != -1) {
gf_log (this->name, GF_LOG_TRACE,
"fxattrop of %s on %s succeeded",
local->loc.path,
priv->children[child_index]->name);
- sh->xattr[child_index] = dict_ref (xattr);
- }
- }
- UNLOCK (&frame->lock);
+ sh->xattr[child_index] = dict_ref (xattr);
+ sh->success_children[sh->success_count] = child_index;
+ sh->success_count++;
+ } else {
+ gf_log (this->name, GF_LOG_ERROR, "fxattrop of %s "
+ "failed on %s, reason %s", local->loc.path,
+ priv->children[child_index]->name,
+ strerror (op_errno));
+ sh->child_errno[child_index] = op_errno;
+ }
+ }
+ UNLOCK (&frame->lock);
+}
- call_count = afr_frame_return (frame);
+int
+afr_sh_data_fxattrop_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *xattr, dict_t *xdata)
+{
+ int call_count = -1;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
- if (call_count == 0) {
- afr_sh_data_fstat (frame, this);
- }
+ local = frame->local;
+ sh = &local->self_heal;
- return 0;
+ afr_sh_common_fxattrop_resp_handler (frame, cookie, this, op_ret,
+ op_errno, xattr);
+
+ call_count = afr_frame_return (frame);
+ if (call_count == 0) {
+ if (!afr_sh_data_proceed (sh->success_count)) {
+ gf_log (this->name, GF_LOG_ERROR, "%s, inspecting "
+ "change log succeeded on < %d children",
+ local->loc.path, AFR_SH_MIN_PARTICIPANTS);
+ afr_sh_data_fail (frame, this);
+ goto out;
+ }
+ afr_sh_data_fstat (frame, this);
+ }
+out:
+ return 0;
}
int
afr_sh_data_fxattrop (call_frame_t *frame, xlator_t *this)
{
- afr_self_heal_t *sh = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- dict_t *xattr_req = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ dict_t **xattr_req;
+ int32_t *zero_pending = NULL;
+ int call_count = 0;
+ int i = 0;
+ int ret = 0;
+ int j;
- int32_t zero_pending[3] = {0, 0, 0};
+ priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
- int call_count = 0;
- int i = 0;
- int ret = 0;
+ call_count = afr_up_children_count (local->child_up,
+ priv->child_count);
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
+ local->call_count = call_count;
- call_count = afr_up_children_count (priv->child_count,
- local->child_up);
+ xattr_req = GF_CALLOC(priv->child_count, sizeof(struct dict_t *),
+ gf_afr_mt_dict_t);
+ if (!xattr_req)
+ goto out;
- local->call_count = call_count;
+ for (i = 0; i < priv->child_count; i++) {
+ xattr_req[i] = dict_new();
+ if (!xattr_req[i]) {
+ ret = -1;
+ goto out;
+ }
+ }
- xattr_req = dict_new();
- if (xattr_req) {
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_set_static_bin (xattr_req, priv->pending_key[i],
- zero_pending, 3 * sizeof(int32_t));
- if (ret < 0)
+ for (i = 0; i < priv->child_count; i++) {
+ for (j = 0; j < priv->child_count; j++) {
+ zero_pending = GF_CALLOC (3, sizeof (*zero_pending),
+ gf_afr_mt_int32_t);
+ if (!zero_pending) {
+ ret = -1;
+ goto out;
+ }
+ ret = dict_set_dynptr (xattr_req[i], priv->pending_key[j],
+ zero_pending,
+ 3 * sizeof (*zero_pending));
+ if (ret < 0) {
gf_log (this->name, GF_LOG_WARNING,
"Unable to set dict value");
+ goto out;
+ } else {
+ zero_pending = NULL;
+ }
+ }
+ }
+
+ afr_reset_xattr (sh->xattr, priv->child_count);
+ afr_reset_children (sh->success_children, priv->child_count);
+ memset (sh->child_errno, 0,
+ sizeof (*sh->child_errno) * priv->child_count);
+ sh->success_count = 0;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE (frame, afr_sh_data_fxattrop_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->fxattrop,
+ sh->healing_fd, GF_XATTROP_ADD_ARRAY,
+ xattr_req[i], NULL);
+
+ if (!--call_count)
+ break;
}
}
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_sh_data_fxattrop_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->fxattrop,
- sh->healing_fd, GF_XATTROP_ADD_ARRAY,
- xattr_req);
-
- if (!--call_count)
- break;
- }
+out:
+ if (xattr_req) {
+ for (i = 0; i < priv->child_count; i++)
+ if (xattr_req[i])
+ dict_unref(xattr_req[i]);
+ GF_FREE(xattr_req);
}
- if (xattr_req)
- dict_unref (xattr_req);
+ if (ret) {
+ GF_FREE (zero_pending);
+ afr_sh_data_fail (frame, this);
+ }
- return 0;
+ return 0;
}
+int
+afr_sh_data_big_lock_success (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+
+ local = frame->local;
+ sh = &local->self_heal;
+
+ sh->data_lock_held = _gf_true;
+ afr_sh_data_fxattrop (frame, this);
+ return 0;
+}
int
-afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this);
+afr_sh_dom_lock_success (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+
+ local = frame->local;
+ sh = &local->self_heal;
+
+ sh->sh_dom_lock_held = _gf_true;
+ afr_sh_data_lock (frame, this, 0, 0, _gf_true, this->name,
+ afr_sh_data_big_lock_success,
+ afr_sh_data_fail);
+ return 0;
+}
int
-afr_sh_data_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
+afr_sh_data_post_blocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
{
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
local = frame->local;
int_lock = &local->internal_lock;
+ sh = &local->self_heal;
if (int_lock->lock_op_ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Non Blocking inodelks failed.");
- afr_sh_data_done (frame, this);
+ gf_log (this->name, GF_LOG_ERROR, "Blocking data inodelks "
+ "failed for %s. by %s",
+ local->loc.path, lkowner_utoa (&frame->root->lk_owner));
+
+ sh->data_lock_failure_handler (frame, this);
} else {
- gf_log (this->name, GF_LOG_DEBUG,
- "Non Blocking inodelks done. Proceeding to FOP");
- afr_sh_data_fxattrop (frame, this);
+ gf_log (this->name, GF_LOG_DEBUG, "Blocking data inodelks "
+ "done for %s by %s. Proceding to self-heal",
+ local->loc.path, lkowner_utoa (&frame->root->lk_owner));
+
+ sh->data_lock_success_handler (frame, this);
}
return 0;
}
int
-afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this)
+afr_sh_data_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
{
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
@@ -941,173 +1454,301 @@ afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this)
int_lock = &local->internal_lock;
sh = &local->self_heal;
+ if (int_lock->lock_op_ret < 0) {
+ gf_log (this->name, GF_LOG_DEBUG, "Non Blocking data inodelks "
+ "failed for %s. by %s",
+ local->loc.path, lkowner_utoa (&frame->root->lk_owner));
+
+ if (!sh->data_lock_block) {
+ sh->data_lock_failure_handler(frame, this);
+ } else {
+ int_lock->lock_cbk =
+ afr_sh_data_post_blocking_inodelk_cbk;
+ afr_blocking_lock (frame, this);
+ }
+ } else {
+
+ gf_log (this->name, GF_LOG_DEBUG, "Non Blocking data inodelks "
+ "done for %s by %s. Proceeding to self-heal",
+ local->loc.path, lkowner_utoa (&frame->root->lk_owner));
+ sh->data_lock_success_handler (frame, this);
+ }
+
+ return 0;
+}
+
+int
+afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this, char *dom,
+ off_t start, off_t len)
+{
+ afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
+ afr_local_t *local = NULL;
+
+ local = frame->local;
+ int_lock = &local->internal_lock;
+
int_lock->transaction_lk_type = AFR_SELFHEAL_LK;
int_lock->selfheal_lk_type = AFR_DATA_SELF_HEAL_LK;
afr_set_lock_number (frame, this);
- int_lock->lk_flock.l_start = 0;
- int_lock->lk_flock.l_len = 0;
- int_lock->lk_flock.l_type = F_WRLCK;
int_lock->lock_cbk = afr_sh_data_post_nonblocking_inodelk_cbk;
- afr_nonblocking_inodelk (frame, this);
+ int_lock->domain = dom;
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+ inodelk->flock.l_start = start;
+ inodelk->flock.l_len = len;
+ inodelk->flock.l_type = F_WRLCK;
+ afr_nonblocking_inodelk (frame, this);
- return 0;
+ return 0;
}
+int
+afr_post_sh_big_lock_success (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+
+ local = frame->local;
+ sh = &local->self_heal;
+
+ GF_ASSERT (sh->old_loop_frame);
+ sh_loop_finish (sh->old_loop_frame, this);
+ sh->old_loop_frame = NULL;
+ sh->data_lock_held = _gf_true;
+ sh->sync_done = _gf_true;
+ afr_sh_data_fxattrop (frame, this);
+ return 0;
+}
int
-afr_sh_data_lock (call_frame_t *frame, xlator_t *this)
+afr_post_sh_big_lock_failure (call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
- afr_self_heal_t * sh = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ local = frame->local;
+ sh = &local->self_heal;
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ GF_ASSERT (sh->old_loop_frame);
+ sh_loop_finish (sh->old_loop_frame, this);
+ sh->old_loop_frame = NULL;
+ afr_sh_set_timestamps (frame, this);
+ return 0;
+}
- if (sh->data_lock_held) {
- /* caller has held the lock already,
- so skip locking */
- afr_sh_data_fxattrop (frame, this);
- return 0;
- }
+int
+afr_sh_data_lock (call_frame_t *frame, xlator_t *this,
+ off_t start, off_t len, gf_boolean_t block,
+ char *dom, afr_lock_cbk_t success_handler,
+ afr_lock_cbk_t failure_handler)
+{
+ afr_local_t * local = NULL;
+ afr_self_heal_t * sh = NULL;
- return afr_sh_data_lock_rec (frame, this);
-}
+ local = frame->local;
+ sh = &local->self_heal;
+ sh->data_lock_success_handler = success_handler;
+ sh->data_lock_failure_handler = failure_handler;
+ sh->data_lock_block = block;
+ return afr_sh_data_lock_rec (frame, this, dom, start, len);
+}
int
afr_sh_data_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd)
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
- int child_index = 0;
-
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
-
- child_index = (long) cookie;
-
- /* TODO: some of the open's might fail.
- In that case, modify cleanup fn to send flush on those
- fd's which are already open */
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE,
- "open of %s failed on child %s (%s)",
- local->loc.path,
- priv->children[child_index]->name,
- strerror (op_errno));
- sh->op_failed = 1;
- }
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int child_index = 0;
- gf_log (this->name, GF_LOG_TRACE,
- "open of %s succeeded on child %s",
- local->loc.path,
- priv->children[child_index]->name);
- }
- UNLOCK (&frame->lock);
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- call_count = afr_frame_return (frame);
+ child_index = (long) cookie;
- if (call_count == 0) {
- if (sh->op_failed) {
- afr_sh_data_finish (frame, this);
- return 0;
- }
+ /* TODO: some of the open's might fail.
+ In that case, modify cleanup fn to send flush on those
+ fd's which are already open */
+
+ LOCK (&frame->lock);
+ {
+ if (op_ret == -1) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "open of %s failed on child %s (%s)",
+ local->loc.path,
+ priv->children[child_index]->name,
+ strerror (op_errno));
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ } else {
+ gf_log (this->name, GF_LOG_TRACE,
+ "open of %s succeeded on child %s",
+ local->loc.path,
+ priv->children[child_index]->name);
+ }
+ }
+ UNLOCK (&frame->lock);
- gf_log (this->name, GF_LOG_TRACE,
- "fd for %s opened, commencing sync",
- local->loc.path);
+ call_count = afr_frame_return (frame);
- afr_sh_data_lock (frame, this);
- }
+ if (call_count == 0) {
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+ afr_sh_data_fail (frame, this);
+ return 0;
+ }
+
+ gf_log (this->name, GF_LOG_TRACE,
+ "fd for %s opened, commencing sync",
+ local->loc.path);
+
+ afr_sh_data_lock (frame, this, 0, 0, _gf_true, priv->sh_domain,
+ afr_sh_dom_lock_success, afr_sh_data_fail);
+ }
- return 0;
+ return 0;
}
int
afr_sh_data_open (call_frame_t *frame, xlator_t *this)
{
- int i = 0;
- int call_count = 0;
+ int i = 0;
+ int call_count = 0;
+ fd_t *fd = NULL;
+ afr_local_t * local = NULL;
+ afr_private_t * priv = NULL;
+ afr_self_heal_t *sh = NULL;
+
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- fd_t *fd = NULL;
+ call_count = afr_up_children_count (local->child_up, priv->child_count);
+ local->call_count = call_count;
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
- afr_self_heal_t *sh = NULL;
+ fd = fd_create (local->loc.inode, frame->root->pid);
+ sh->healing_fd = fd;
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ /* open sinks */
+ for (i = 0; i < priv->child_count; i++) {
+ if(!local->child_up[i])
+ continue;
- if (sh->healing_fd_opened) {
- /* caller has opened the fd for us already, so skip open */
+ STACK_WIND_COOKIE (frame, afr_sh_data_open_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->open,
+ &local->loc,
+ O_RDWR|O_LARGEFILE, fd, NULL);
- afr_sh_data_lock (frame, this);
- return 0;
+ if (!--call_count)
+ break;
}
- call_count = afr_up_children_count (priv->child_count, local->child_up);
- local->call_count = call_count;
+ return 0;
+}
- fd = fd_create (local->loc.inode, frame->root->pid);
- sh->healing_fd = fd;
+void
+afr_sh_non_reg_fix (call_frame_t *frame, xlator_t *this,
+ int32_t op_ret, int32_t op_errno)
+{
+ afr_private_t *priv = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
+
+ if (op_ret < 0) {
+ afr_sh_data_fail (frame, this);
+ return;
+ }
- /* open sinks */
- for (i = 0; i < priv->child_count; i++) {
- if(!local->child_up[i])
- continue;
-
- STACK_WIND_COOKIE (frame, afr_sh_data_open_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->open,
- &local->loc,
- O_RDWR|O_LARGEFILE, fd, 0);
-
- if (!--call_count)
- break;
- }
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- return 0;
+ for (i = 0; i < priv->child_count ; i++) {
+ if (1 == local->child_up[i])
+ sh->success[i] = 1;
+ }
+
+ afr_sh_erase_pending (frame, this, AFR_DATA_TRANSACTION,
+ afr_sh_data_erase_pending_cbk,
+ afr_sh_data_finish);
+}
+
+int
+afr_sh_non_reg_lock_success (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+
+ local = frame->local;
+ sh = &local->self_heal;
+ sh->data_lock_held = _gf_true;
+ afr_sh_common_lookup (frame, this, &local->loc,
+ afr_sh_non_reg_fix, NULL,
+ AFR_LOOKUP_FAIL_CONFLICTS |
+ AFR_LOOKUP_FAIL_MISSING_GFIDS,
+ NULL);
+ return 0;
}
+gf_boolean_t
+afr_can_start_data_self_heal (afr_self_heal_t *sh, afr_private_t *priv)
+{
+ if (sh->force_confirm_spb)
+ return _gf_true;
+ if (sh->do_data_self_heal &&
+ afr_data_self_heal_enabled (priv->data_self_heal))
+ return _gf_true;
+ return _gf_false;
+}
int
afr_self_heal_data (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = this->private;
-
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = this->private;
+ int ret = -1;
- local = frame->local;
- sh = &local->self_heal;
+ local = frame->local;
+ sh = &local->self_heal;
+
+ sh->sh_type_in_action = AFR_SELF_HEAL_DATA;
+
+ if (afr_can_start_data_self_heal (sh, priv)) {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_STARTED);
+ ret = afr_inodelk_init (&local->internal_lock.inodelk[1],
+ priv->sh_domain, priv->child_count);
+ if (ret < 0) {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_data_done (frame, this);
+ return 0;
+ }
- if (sh->need_data_self_heal && priv->data_self_heal) {
- afr_sh_data_open (frame, this);
- } else {
- gf_log (this->name, GF_LOG_TRACE,
- "not doing data self heal on %s",
- local->loc.path);
- afr_sh_data_done (frame, this);
- }
+ if (IA_ISREG (sh->type)) {
+ afr_sh_data_open (frame, this);
+ } else {
+ afr_sh_data_lock (frame, this, 0, 0, _gf_true,
+ this->name,
+ afr_sh_non_reg_lock_success,
+ afr_sh_data_fail);
+ }
+ } else {
+ gf_log (this->name, GF_LOG_TRACE,
+ "not doing data self heal on %s",
+ local->loc.path);
+ afr_sh_data_done (frame, this);
+ }
- return 0;
+ return 0;
}
-
diff --git a/xlators/cluster/afr/src/afr-self-heal-entry.c b/xlators/cluster/afr/src/afr-self-heal-entry.c
index 446b3d868..53491a1d7 100644
--- a/xlators/cluster/afr/src/afr-self-heal-entry.c
+++ b/xlators/cluster/afr/src/afr-self-heal-entry.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include <libgen.h>
@@ -49,39 +40,30 @@
#include "afr-self-heal.h"
#include "afr-self-heal-common.h"
-int
-afr_sh_post_nonblocking_entrylk_cbk (call_frame_t *frame, xlator_t *this);
+#define AFR_INIT_SH_FRAME_VALS(_frame, _local, _sh, _sh_frame, _sh_local, _sh_sh)\
+ do {\
+ _local = _frame->local;\
+ _sh = &_local->self_heal;\
+ _sh_frame = _sh->sh_frame;\
+ _sh_local = _sh_frame->local;\
+ _sh_sh = &_sh_local->self_heal;\
+ } while (0);
int
+afr_sh_entry_impunge_create_file (call_frame_t *impunge_frame, xlator_t *this,
+ int child_index);
+int
afr_sh_entry_done (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
-
- /*
- TODO: cleanup sh->*
- */
-
- if (sh->healing_fd)
- fd_unref (sh->healing_fd);
- sh->healing_fd = NULL;
-
-/* for (i = 0; i < priv->child_count; i++) { */
-/* sh->locked_nodes[i] = 0; */
-/* } */
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
- gf_log (this->name, GF_LOG_TRACE,
- "self heal of %s completed",
- local->loc.path);
+ local = frame->local;
+ sh = &local->self_heal;
- sh->completion_cbk (frame, this);
+ sh->completion_cbk (frame, this);
- return 0;
+ return 0;
}
@@ -104,323 +86,268 @@ afr_sh_entry_unlock (call_frame_t *frame, xlator_t *this)
int
afr_sh_entry_finish (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- gf_log (this->name, GF_LOG_TRACE,
- "finishing entry selfheal of %s", local->loc.path);
+ gf_log (this->name, GF_LOG_TRACE,
+ "finishing entry selfheal of %s", local->loc.path);
- afr_sh_entry_unlock (frame, this);
+ afr_sh_entry_unlock (frame, this);
- return 0;
+ return 0;
}
int
afr_sh_entry_erase_pending_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xattr)
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xattr, dict_t *xdata)
{
- int call_count = 0;
+ long i = 0;
+ int call_count = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_local_t *orig_local = NULL;
+ call_frame_t *orig_frame = NULL;
+ afr_private_t *priv = NULL;
+ int32_t read_child = -1;
- call_count = afr_frame_return (frame);
+ local = frame->local;
+ priv = this->private;
+ sh = &local->self_heal;
+ i = (long)cookie;
- if (call_count == 0)
- afr_sh_entry_finish (frame, this);
- return 0;
+ afr_children_add_child (sh->fresh_children, i, priv->child_count);
+ if (op_ret == -1) {
+ gf_log (this->name, GF_LOG_INFO,
+ "%s: failed to erase pending xattrs on %s (%s)",
+ local->loc.path, priv->children[i]->name,
+ strerror (op_errno));
+ }
+
+ call_count = afr_frame_return (frame);
+
+ if (call_count == 0) {
+ if (sh->source == -1) {
+ //this happens if the forced merge option is set
+ read_child = sh->fresh_children[0];
+ } else {
+ read_child = sh->source;
+ }
+ afr_inode_set_read_ctx (this, sh->inode, read_child,
+ sh->fresh_children);
+ orig_frame = sh->orig_frame;
+ orig_local = orig_frame->local;
+
+ if (sh->source != -1) {
+ orig_local->cont.lookup.buf.ia_nlink = sh->buf[sh->source].ia_nlink;
+ }
+
+ afr_sh_entry_finish (frame, this);
+ }
+
+ return 0;
}
int
afr_sh_entry_erase_pending (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
- int i = 0;
- dict_t **erase_xattr = NULL;
- int need_unwind = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ local = frame->local;
+ sh = &local->self_heal;
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ if (sh->entries_skipped) {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ goto out;
+ }
+ afr_sh_erase_pending (frame, this, AFR_ENTRY_TRANSACTION,
+ afr_sh_entry_erase_pending_cbk,
+ afr_sh_entry_finish);
+ return 0;
+out:
+ afr_sh_entry_finish (frame, this);
+ return 0;
+}
- afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix, sh->success,
- priv->child_count, AFR_ENTRY_TRANSACTION);
- erase_xattr = GF_CALLOC (sizeof (*erase_xattr), priv->child_count,
- gf_afr_mt_dict_t);
- for (i = 0; i < priv->child_count; i++) {
- if (sh->xattr[i]) {
- call_count++;
+static int
+next_active_source (call_frame_t *frame, xlator_t *this,
+ int current_active_source)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ int source = -1;
+ int next_active_source = -1;
+ int i = 0;
- erase_xattr[i] = get_new_dict();
- dict_ref (erase_xattr[i]);
- }
- }
+ priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
- if (call_count == 0)
- need_unwind = 1;
-
- afr_sh_delta_to_xattr (priv, sh->delta_matrix, erase_xattr,
- priv->child_count, AFR_ENTRY_TRANSACTION);
-
- local->call_count = call_count;
- for (i = 0; i < priv->child_count; i++) {
- if (!erase_xattr[i])
- continue;
-
- gf_log (this->name, GF_LOG_TRACE,
- "erasing pending flags from %s on %s",
- local->loc.path, priv->children[i]->name);
-
- STACK_WIND_COOKIE (frame, afr_sh_entry_erase_pending_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->xattrop,
- &local->loc,
- GF_XATTROP_ADD_ARRAY, erase_xattr[i]);
- if (!--call_count)
- break;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (erase_xattr[i]) {
- dict_unref (erase_xattr[i]);
- }
- }
- GF_FREE (erase_xattr);
-
- if (need_unwind)
- afr_sh_entry_finish (frame, this);
+ source = sh->source;
- return 0;
-}
+ if (source != -1) {
+ if (current_active_source != source)
+ next_active_source = source;
+ goto out;
+ }
+ /*
+ the next active sink becomes the source for the
+ 'conservative decision' of merging all entries
+ */
+ for (i = 0; i < priv->child_count; i++) {
+ if ((sh->sources[i] == 0)
+ && (local->child_up[i] == 1)
+ && (i > current_active_source)) {
-static int
-next_active_source (call_frame_t *frame, xlator_t *this,
- int current_active_source)
-{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- int source = -1;
- int next_active_source = -1;
- int i = 0;
-
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
-
- source = sh->source;
-
- if (source != -1) {
- if (current_active_source != source)
- next_active_source = source;
- goto out;
- }
-
- /*
- the next active sink becomes the source for the
- 'conservative decision' of merging all entries
- */
-
- for (i = 0; i < priv->child_count; i++) {
- if ((sh->sources[i] == 0)
- && (local->child_up[i] == 1)
- && (i > current_active_source)) {
-
- next_active_source = i;
- break;
- }
- }
+ next_active_source = i;
+ break;
+ }
+ }
out:
- return next_active_source;
+ return next_active_source;
}
static int
next_active_sink (call_frame_t *frame, xlator_t *this,
- int current_active_sink)
+ int current_active_sink)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- int next_active_sink = -1;
- int i = 0;
-
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
-
- /*
- the next active sink becomes the source for the
- 'conservative decision' of merging all entries
- */
-
- for (i = 0; i < priv->child_count; i++) {
- if ((sh->sources[i] == 0)
- && (local->child_up[i] == 1)
- && (i > current_active_sink)) {
-
- next_active_sink = i;
- break;
- }
- }
-
- return next_active_sink;
-}
-
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ int next_active_sink = -1;
+ int i = 0;
-int
-build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name)
-{
- int ret = -1;
+ priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
- if (!child) {
- goto out;
- }
+ /*
+ the next active sink becomes the source for the
+ 'conservative decision' of merging all entries
+ */
- if (strcmp (parent->path, "/") == 0)
- ret = gf_asprintf ((char **)&child->path, "/%s", name);
- else
- ret = gf_asprintf ((char **)&child->path, "%s/%s", parent->path,
- name);
+ for (i = 0; i < priv->child_count; i++) {
+ if ((sh->sources[i] == 0)
+ && (local->child_up[i] == 1)
+ && (i > current_active_sink)) {
- if (-1 == ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "asprintf failed while setting child path");
+ next_active_sink = i;
+ break;
+ }
}
- if (!child->path) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
- goto out;
- }
-
- child->name = strrchr (child->path, '/');
- if (child->name)
- child->name++;
-
- child->parent = inode_ref (parent->inode);
- child->inode = inode_new (parent->inode->table);
-
- if (!child->inode) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
- goto out;
- }
-
- ret = 0;
-out:
- if (ret == -1)
- loc_wipe (child);
-
- return ret;
+ return next_active_sink;
}
+int
+afr_sh_entry_impunge_all (call_frame_t *frame, xlator_t *this);
+
+int
+afr_sh_entry_impunge_subvol (call_frame_t *frame, xlator_t *this);
int
afr_sh_entry_expunge_all (call_frame_t *frame, xlator_t *this);
int
afr_sh_entry_expunge_subvol (call_frame_t *frame, xlator_t *this,
- int active_src);
+ int active_src);
int
afr_sh_entry_expunge_entry_done (call_frame_t *frame, xlator_t *this,
- int active_src)
+ int active_src, int32_t op_ret,
+ int32_t op_errno)
{
- int call_count = 0;
+ int call_count = 0;
- call_count = afr_frame_return (frame);
+ call_count = afr_frame_return (frame);
- if (call_count == 0)
- afr_sh_entry_expunge_subvol (frame, this, active_src);
+ if (call_count == 0)
+ afr_sh_entry_expunge_subvol (frame, this, active_src);
- return 0;
+ return 0;
}
int
afr_sh_entry_expunge_parent_setattr_cbk (call_frame_t *expunge_frame,
void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop)
+ struct iatt *preop, struct iatt *postop,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *expunge_local = NULL;
- afr_self_heal_t *expunge_sh = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *expunge_local = NULL;
+ afr_self_heal_t *expunge_sh = NULL;
call_frame_t *frame = NULL;
-
- int active_src = (long) cookie;
+ int active_src = (long) cookie;
+ afr_self_heal_t *sh = NULL;
+ afr_local_t *local = NULL;
priv = this->private;
expunge_local = expunge_frame->local;
- expunge_sh = &expunge_local->self_heal;
- frame = expunge_sh->sh_frame;
+ expunge_sh = &expunge_local->self_heal;
+ frame = expunge_sh->sh_frame;
+ local = frame->local;
+ sh = &local->self_heal;
if (op_ret != 0) {
- gf_log (this->name, GF_LOG_DEBUG,
+ gf_log (this->name, GF_LOG_ERROR,
"setattr on parent directory of %s on subvolume %s failed: %s",
expunge_local->loc.path,
priv->children[active_src]->name, strerror (op_errno));
}
- AFR_STACK_DESTROY (expunge_frame);
- afr_sh_entry_expunge_entry_done (frame, this, active_src);
+ AFR_STACK_DESTROY (expunge_frame);
+ sh->expunge_done (frame, this, active_src, op_ret, op_errno);
return 0;
}
int
-afr_sh_entry_expunge_rename_cbk (call_frame_t *expunge_frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *buf,
- struct iatt *preoldparent,
- struct iatt *postoldparent,
- struct iatt *prenewparent,
- struct iatt *postnewparent)
+afr_sh_entry_expunge_remove_cbk (call_frame_t *expunge_frame, void *cookie,
+ xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *expunge_local = NULL;
- afr_self_heal_t *expunge_sh = NULL;
- int active_src = 0;
- call_frame_t *frame = NULL;
-
- int32_t valid = 0;
-
- priv = this->private;
- expunge_local = expunge_frame->local;
- expunge_sh = &expunge_local->self_heal;
- frame = expunge_sh->sh_frame;
-
- active_src = (long) cookie;
-
- if (op_ret == 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "removed %s on %s",
- expunge_local->loc.path,
- priv->children[active_src]->name);
- } else {
- gf_log (this->name, GF_LOG_DEBUG,
- "removing %s on %s failed (%s)",
- expunge_local->loc.path,
- priv->children[active_src]->name,
- strerror (op_errno));
- }
+ afr_private_t *priv = NULL;
+ afr_local_t *expunge_local = NULL;
+ afr_self_heal_t *expunge_sh = NULL;
+ int active_src = 0;
+ int32_t valid = 0;
+
+ priv = this->private;
+ expunge_local = expunge_frame->local;
+ expunge_sh = &expunge_local->self_heal;
+
+ active_src = (long) cookie;
+
+ if (op_ret == 0) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "removed %s on %s",
+ expunge_local->loc.path,
+ priv->children[active_src]->name);
+ } else {
+ gf_log (this->name, GF_LOG_INFO,
+ "removing %s on %s failed (%s)",
+ expunge_local->loc.path,
+ priv->children[active_src]->name,
+ strerror (op_errno));
+ }
valid = GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
- afr_build_parent_loc (&expunge_sh->parent_loc, &expunge_local->loc);
STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_parent_setattr_cbk,
(void *) (long) active_src,
@@ -428,915 +355,972 @@ afr_sh_entry_expunge_rename_cbk (call_frame_t *expunge_frame, void *cookie,
priv->children[active_src]->fops->setattr,
&expunge_sh->parent_loc,
&expunge_sh->parentbuf,
- valid);
+ valid, NULL);
return 0;
}
-static void
-init_trash_loc (loc_t *trash_loc, inode_table_t *table)
-{
- trash_loc->path = gf_strdup ("/" GF_REPLICATE_TRASH_DIR);
- trash_loc->name = GF_REPLICATE_TRASH_DIR;
- trash_loc->parent = table->root;
- trash_loc->inode = inode_new (table);
-}
-
-
-char *
-make_trash_path (const char *path)
+int
+afr_sh_entry_expunge_unlink (call_frame_t *expunge_frame, xlator_t *this,
+ int active_src)
{
- char *c = NULL;
- char *tp = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *expunge_local = NULL;
- tp = GF_CALLOC (strlen ("/" GF_REPLICATE_TRASH_DIR) + strlen (path) + 1, sizeof (char),
- gf_afr_mt_char);
+ priv = this->private;
+ expunge_local = expunge_frame->local;
- strcpy (tp, GF_REPLICATE_TRASH_DIR);
- strcat (tp, path);
+ gf_log (this->name, GF_LOG_TRACE,
+ "expunging file %s on %s",
+ expunge_local->loc.path, priv->children[active_src]->name);
- c = strchr (tp, '/') + 1;
- while (*c++)
- if (*c == '/')
- *c = '-';
+ STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_remove_cbk,
+ (void *) (long) active_src,
+ priv->children[active_src],
+ priv->children[active_src]->fops->unlink,
+ &expunge_local->loc, 0, NULL);
- return tp;
+ return 0;
}
+
int
-afr_sh_entry_expunge_rename (call_frame_t *expunge_frame, xlator_t *this,
- int active_src, inode_t *trash_inode)
+afr_sh_entry_expunge_rmdir (call_frame_t *expunge_frame, xlator_t *this,
+ int active_src)
{
- afr_private_t *priv = NULL;
- afr_local_t *expunge_local = NULL;
-
- loc_t rename_loc;
-
- priv = this->private;
- expunge_local = expunge_frame->local;
-
- rename_loc.inode = inode_ref (expunge_local->loc.inode);
- rename_loc.path = make_trash_path (expunge_local->loc.path);
- rename_loc.name = strrchr (rename_loc.path, '/') + 1;
- rename_loc.parent = trash_inode;
+ afr_private_t *priv = NULL;
+ afr_local_t *expunge_local = NULL;
- gf_log (this->name, GF_LOG_TRACE,
- "moving file/directory %s on %s to %s",
- expunge_local->loc.path, priv->children[active_src]->name,
- rename_loc.path);
+ priv = this->private;
+ expunge_local = expunge_frame->local;
- STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_rename_cbk,
- (void *) (long) active_src,
- priv->children[active_src],
- priv->children[active_src]->fops->rename,
- &expunge_local->loc, &rename_loc);
+ gf_log (this->name, GF_LOG_DEBUG,
+ "expunging directory %s on %s",
+ expunge_local->loc.path, priv->children[active_src]->name);
- loc_wipe (&rename_loc);
+ STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_remove_cbk,
+ (void *) (long) active_src,
+ priv->children[active_src],
+ priv->children[active_src]->fops->rmdir,
+ &expunge_local->loc, 1, NULL);
- return 0;
+ return 0;
}
int
-afr_sh_entry_expunge_mkdir_cbk (call_frame_t *expunge_frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent)
+afr_sh_entry_expunge_remove (call_frame_t *expunge_frame, xlator_t *this,
+ int active_src, struct iatt *buf,
+ struct iatt *parentbuf)
{
- afr_private_t *priv = NULL;
+ afr_private_t *priv = NULL;
afr_local_t *expunge_local = NULL;
- afr_self_heal_t *expunge_sh = NULL;
- call_frame_t *frame = NULL;
-
- int active_src = (long) cookie;
-
- inode_t *trash_inode = NULL;
-
- priv = this->private;
+ afr_self_heal_t *expunge_sh = NULL;
+ call_frame_t *frame = NULL;
+ int type = 0;
+ afr_self_heal_t *sh = NULL;
+ afr_local_t *local = NULL;
+ loc_t *loc = NULL;
+
+ priv = this->private;
expunge_local = expunge_frame->local;
- expunge_sh = &expunge_local->self_heal;
- frame = expunge_sh->sh_frame;
-
- if (op_ret != 0) {
+ expunge_sh = &expunge_local->self_heal;
+ frame = expunge_sh->sh_frame;
+ local = frame->local;
+ sh = &local->self_heal;
+ loc = &expunge_local->loc;
+
+ type = buf->ia_type;
+ if (loc->parent && uuid_is_null (loc->parent->gfid))
+ uuid_copy (loc->pargfid, parentbuf->ia_gfid);
+
+ switch (type) {
+ case IA_IFSOCK:
+ case IA_IFREG:
+ case IA_IFBLK:
+ case IA_IFCHR:
+ case IA_IFIFO:
+ case IA_IFLNK:
+ afr_sh_entry_expunge_unlink (expunge_frame, this, active_src);
+ break;
+ case IA_IFDIR:
+ afr_sh_entry_expunge_rmdir (expunge_frame, this, active_src);
+ break;
+ default:
gf_log (this->name, GF_LOG_ERROR,
- "mkdir of /" GF_REPLICATE_TRASH_DIR " failed on %s",
- priv->children[active_src]->name);
-
+ "%s has unknown file type on %s: 0%o",
+ expunge_local->loc.path,
+ priv->children[active_src]->name, type);
goto out;
+ break;
}
- /* mkdir successful */
-
- trash_inode = inode_link (inode, expunge_local->loc.inode->table->root,
- GF_REPLICATE_TRASH_DIR, buf);
-
- afr_sh_entry_expunge_rename (expunge_frame, this, active_src,
- trash_inode);
return 0;
out:
AFR_STACK_DESTROY (expunge_frame);
- afr_sh_entry_expunge_entry_done (frame, this, active_src);
+ sh->expunge_done (frame, this, active_src, -1, EINVAL);
+
return 0;
}
int
-afr_sh_entry_expunge_lookup_trash_cbk (call_frame_t *expunge_frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf,
- dict_t *xattr, struct iatt *postparent)
+afr_sh_entry_expunge_lookup_cbk (call_frame_t *expunge_frame, void *cookie,
+ xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ inode_t *inode, struct iatt *buf, dict_t *x,
+ struct iatt *postparent)
{
- afr_private_t *priv = NULL;
+ afr_private_t *priv = NULL;
afr_local_t *expunge_local = NULL;
- afr_self_heal_t *expunge_sh = NULL;
- call_frame_t *frame = NULL;
- dict_t *dict = NULL;
-
- int active_src = (long) cookie;
- int ret = 0;
-
- inode_t *trash_inode;
- loc_t trash_loc;
+ afr_self_heal_t *expunge_sh = NULL;
+ call_frame_t *frame = NULL;
+ int active_src = 0;
+ afr_self_heal_t *sh = NULL;
+ afr_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
expunge_local = expunge_frame->local;
- expunge_sh = &expunge_local->self_heal;
- frame = expunge_sh->sh_frame;
-
- if ((op_ret != 0) && (op_errno == ENOENT)) {
- init_trash_loc (&trash_loc, expunge_local->loc.inode->table);
-
- dict = dict_new ();
- if (!dict) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- goto out;
- }
-
- ret = afr_set_dict_gfid (dict, buf->ia_gfid);
- if (ret)
- gf_log (this->name, GF_LOG_DEBUG, "gfid set failed");
-
-
- gf_log (this->name, GF_LOG_TRACE,
- "creating directory " GF_REPLICATE_TRASH_DIR " on subvolume %s",
- priv->children[active_src]->name);
-
- STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_mkdir_cbk,
- (void *) (long) active_src,
- priv->children[active_src],
- priv->children[active_src]->fops->mkdir,
- &trash_loc, 0777, dict);
-
- loc_wipe (&trash_loc);
- if (dict)
- dict_unref (dict);
- return 0;
- }
+ expunge_sh = &expunge_local->self_heal;
+ frame = expunge_sh->sh_frame;
+ active_src = (long) cookie;
+ local = frame->local;
+ sh = &local->self_heal;
- if (op_ret != 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "lookup of /" GF_REPLICATE_TRASH_DIR " failed on %s",
- priv->children[active_src]->name);
+ if (op_ret == -1) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "lookup of %s on %s failed (%s)",
+ expunge_local->loc.path,
+ priv->children[active_src]->name,
+ strerror (op_errno));
goto out;
}
- /* lookup successful */
+ afr_sh_entry_expunge_remove (expunge_frame, this, active_src, buf,
+ postparent);
- trash_inode = inode_link (inode, expunge_local->loc.inode->table->root,
- GF_REPLICATE_TRASH_DIR, buf);
-
- afr_sh_entry_expunge_rename (expunge_frame, this, active_src,
- trash_inode);
return 0;
out:
AFR_STACK_DESTROY (expunge_frame);
- afr_sh_entry_expunge_entry_done (frame, this, active_src);
+ sh->expunge_done (frame, this, active_src, op_ret, op_errno);
+
return 0;
}
int
-afr_sh_entry_expunge_lookup_trash (call_frame_t *expunge_frame, xlator_t *this,
- int active_src)
+afr_sh_entry_expunge_purge (call_frame_t *expunge_frame, xlator_t *this,
+ int active_src)
{
- afr_private_t *priv = NULL;
- afr_local_t *expunge_local = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *expunge_local = NULL;
- inode_t *root = NULL;
- inode_t *trash = NULL;
- loc_t trash_loc;
+ priv = this->private;
+ expunge_local = expunge_frame->local;
- priv = this->private;
- expunge_local = expunge_frame->local;
+ gf_log (this->name, GF_LOG_TRACE,
+ "looking up %s on %s",
+ expunge_local->loc.path, priv->children[active_src]->name);
- root = expunge_local->loc.inode->table->root;
+ STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_lookup_cbk,
+ (void *) (long) active_src,
+ priv->children[active_src],
+ priv->children[active_src]->fops->lookup,
+ &expunge_local->loc, NULL);
- trash = inode_grep (root->table, root, GF_REPLICATE_TRASH_DIR);
+ return 0;
+}
- if (trash) {
- /* inode is in cache, so no need to mkdir */
+int
+afr_sh_entry_expunge_entry_cbk (call_frame_t *expunge_frame, void *cookie,
+ xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ inode_t *inode, struct iatt *buf, dict_t *x,
+ struct iatt *postparent)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *expunge_local = NULL;
+ afr_self_heal_t *expunge_sh = NULL;
+ int source = 0;
+ call_frame_t *frame = NULL;
+ int active_src = 0;
+ int need_expunge = 0;
+ afr_self_heal_t *sh = NULL;
+ afr_local_t *local = NULL;
+
+ priv = this->private;
+ expunge_local = expunge_frame->local;
+ expunge_sh = &expunge_local->self_heal;
+ frame = expunge_sh->sh_frame;
+ active_src = expunge_sh->active_source;
+ source = (long) cookie;
+ local = frame->local;
+ sh = &local->self_heal;
+
+ if (op_ret == -1 && op_errno == ENOENT)
+ need_expunge = 1;
+ else if (op_ret == -1)
+ goto out;
- afr_sh_entry_expunge_rename (expunge_frame, this, active_src,
- trash);
- return 0;
+ if (!uuid_is_null (expunge_sh->entrybuf.ia_gfid) &&
+ !uuid_is_null (buf->ia_gfid) &&
+ (uuid_compare (expunge_sh->entrybuf.ia_gfid, buf->ia_gfid) != 0)) {
+ char uuidbuf1[64];
+ char uuidbuf2[64];
+ gf_log (this->name, GF_LOG_DEBUG,
+ "entry %s found on %s with mismatching gfid (%s/%s)",
+ expunge_local->loc.path,
+ priv->children[source]->name,
+ uuid_utoa_r (expunge_sh->entrybuf.ia_gfid, uuidbuf1),
+ uuid_utoa_r (buf->ia_gfid, uuidbuf2));
+ need_expunge = 1;
}
- /* Not in cache, so look it up */
+ if (need_expunge) {
+ gf_log (this->name, GF_LOG_INFO,
+ "Entry %s is missing on %s and deleting from "
+ "replica's other bricks",
+ expunge_local->loc.path,
+ priv->children[source]->name);
+
+ if (postparent)
+ expunge_sh->parentbuf = *postparent;
- init_trash_loc (&trash_loc, expunge_local->loc.inode->table);
+ afr_sh_entry_expunge_purge (expunge_frame, this, active_src);
- gf_log (this->name, GF_LOG_TRACE,
- "looking up /" GF_REPLICATE_TRASH_DIR " on %s",
- priv->children[active_src]->name);
+ return 0;
+ }
- STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_lookup_trash_cbk,
- (void *) (long) active_src,
- priv->children[active_src],
- priv->children[active_src]->fops->lookup,
- &trash_loc, NULL);
+out:
+ if (op_ret == 0) {
+ gf_log (this->name, GF_LOG_TRACE,
+ "%s exists under %s",
+ expunge_local->loc.path,
+ priv->children[source]->name);
+ } else {
+ gf_log (this->name, GF_LOG_INFO,
+ "looking up %s under %s failed (%s)",
+ expunge_local->loc.path,
+ priv->children[source]->name,
+ strerror (op_errno));
+ }
- loc_wipe (&trash_loc);
+ AFR_STACK_DESTROY (expunge_frame);
+ sh->expunge_done (frame, this, active_src, op_ret, op_errno);
- return 0;
+ return 0;
}
-
-int
-afr_sh_entry_expunge_remove (call_frame_t *expunge_frame, xlator_t *this,
- int active_src, struct iatt *buf)
+static gf_boolean_t
+can_skip_entry_self_heal (char *name, loc_t *parent_loc)
{
- afr_private_t *priv = NULL;
- afr_local_t *expunge_local = NULL;
- afr_self_heal_t *expunge_sh = NULL;
- int source = 0;
- call_frame_t *frame = NULL;
- int type = 0;
-
- priv = this->private;
- expunge_local = expunge_frame->local;
- expunge_sh = &expunge_local->self_heal;
- frame = expunge_sh->sh_frame;
- source = expunge_sh->source;
-
- type = buf->ia_type;
-
- switch (type) {
- case IA_IFSOCK:
- case IA_IFREG:
- case IA_IFBLK:
- case IA_IFCHR:
- case IA_IFIFO:
- case IA_IFLNK:
- case IA_IFDIR:
- afr_sh_entry_expunge_lookup_trash (expunge_frame, this, active_src);
- break;
- default:
- gf_log (this->name, GF_LOG_ERROR,
- "%s has unknown file type on %s: 0%o",
- expunge_local->loc.path,
- priv->children[source]->name, type);
- goto out;
- break;
- }
-
- return 0;
-out:
- AFR_STACK_DESTROY (expunge_frame);
- afr_sh_entry_expunge_entry_done (frame, this, active_src);
-
- return 0;
+ if (strcmp (name, ".") == 0) {
+ return _gf_true;
+ } else if (strcmp (name, "..") == 0) {
+ return _gf_true;
+ } else if (loc_is_root (parent_loc) &&
+ (strcmp (name, GF_REPLICATE_TRASH_DIR) == 0)) {
+ return _gf_true;
+ }
+ return _gf_false;
}
-
int
-afr_sh_entry_expunge_lookup_cbk (call_frame_t *expunge_frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf, dict_t *x,
- struct iatt *postparent)
+afr_sh_entry_expunge_entry (call_frame_t *frame, xlator_t *this,
+ gf_dirent_t *entry)
{
- afr_private_t *priv = NULL;
- afr_local_t *expunge_local = NULL;
- afr_self_heal_t *expunge_sh = NULL;
- call_frame_t *frame = NULL;
- int active_src = 0;
-
- priv = this->private;
- expunge_local = expunge_frame->local;
- expunge_sh = &expunge_local->self_heal;
- frame = expunge_sh->sh_frame;
- active_src = (long) cookie;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE,
- "lookup of %s on %s failed (%s)",
- expunge_local->loc.path,
- priv->children[active_src]->name,
- strerror (op_errno));
- goto out;
- }
-
- afr_sh_entry_expunge_remove (expunge_frame, this, active_src, buf);
-
- return 0;
-out:
- AFR_STACK_DESTROY (expunge_frame);
- afr_sh_entry_expunge_entry_done (frame, this, active_src);
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ int ret = -1;
+ call_frame_t *expunge_frame = NULL;
+ afr_local_t *expunge_local = NULL;
+ afr_self_heal_t *expunge_sh = NULL;
+ int active_src = 0;
+ int source = 0;
+ int op_errno = 0;
+ char *name = NULL;
+ int op_ret = -1;
+
+ priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
- return 0;
-}
+ active_src = sh->active_source;
+ source = sh->source;
+ sh->expunge_done = afr_sh_entry_expunge_entry_done;
+ name = entry->d_name;
+ if (can_skip_entry_self_heal (name, &local->loc)) {
+ op_ret = 0;
+ goto out;
+ }
-int
-afr_sh_entry_expunge_purge (call_frame_t *expunge_frame, xlator_t *this,
- int active_src)
-{
- afr_private_t *priv = NULL;
- afr_local_t *expunge_local = NULL;
-
- priv = this->private;
- expunge_local = expunge_frame->local;
-
- gf_log (this->name, GF_LOG_TRACE,
- "looking up %s on %s",
- expunge_local->loc.path, priv->children[active_src]->name);
-
- STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_lookup_cbk,
- (void *) (long) active_src,
- priv->children[active_src],
- priv->children[active_src]->fops->lookup,
- &expunge_local->loc, 0);
-
- return 0;
-}
+ gf_log (this->name, GF_LOG_TRACE,
+ "inspecting existence of %s under %s",
+ name, local->loc.path);
+ expunge_frame = copy_frame (frame);
+ if (!expunge_frame) {
+ op_errno = ENOMEM;
+ goto out;
+ }
-int
-afr_sh_entry_expunge_entry_cbk (call_frame_t *expunge_frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf, dict_t *x,
- struct iatt *postparent)
-{
- afr_private_t *priv = NULL;
- afr_local_t *expunge_local = NULL;
- afr_self_heal_t *expunge_sh = NULL;
- int source = 0;
- call_frame_t *frame = NULL;
- int active_src = 0;
-
-
- priv = this->private;
- expunge_local = expunge_frame->local;
- expunge_sh = &expunge_local->self_heal;
- frame = expunge_sh->sh_frame;
- active_src = expunge_sh->active_source;
- source = (long) cookie;
-
- if (op_ret == -1 && op_errno == ENOENT && postparent) {
-
- gf_log (this->name, GF_LOG_TRACE,
- "missing entry %s on %s",
- expunge_local->loc.path,
- priv->children[source]->name);
-
- expunge_sh->parentbuf = *postparent;
-
- afr_sh_entry_expunge_purge (expunge_frame, this, active_src);
-
- return 0;
- }
-
- if (op_ret == 0) {
- gf_log (this->name, GF_LOG_TRACE,
- "%s exists under %s",
- expunge_local->loc.path,
- priv->children[source]->name);
- } else {
- gf_log (this->name, GF_LOG_TRACE,
- "looking up %s under %s failed (%s)",
- expunge_local->loc.path,
- priv->children[source]->name,
- strerror (op_errno));
- }
-
- AFR_STACK_DESTROY (expunge_frame);
- afr_sh_entry_expunge_entry_done (frame, this, active_src);
-
- return 0;
-}
+ AFR_LOCAL_ALLOC_OR_GOTO (expunge_local, out);
+ expunge_frame->local = expunge_local;
+ expunge_sh = &expunge_local->self_heal;
+ expunge_sh->sh_frame = frame;
+ expunge_sh->active_source = active_src;
+ expunge_sh->entrybuf = entry->d_stat;
+ loc_copy (&expunge_sh->parent_loc, &local->loc);
-int
-afr_sh_entry_expunge_entry (call_frame_t *frame, xlator_t *this,
- char *name)
-{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- int ret = -1;
- call_frame_t *expunge_frame = NULL;
- afr_local_t *expunge_local = NULL;
- afr_self_heal_t *expunge_sh = NULL;
- int active_src = 0;
- int source = 0;
- int op_errno = 0;
-
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
-
- active_src = sh->active_source;
- source = sh->source;
-
- if ((strcmp (name, ".") == 0)
- || (strcmp (name, "..") == 0)
- || ((strcmp (local->loc.path, "/") == 0)
- && (strcmp (name, GF_REPLICATE_TRASH_DIR) == 0))) {
+ ret = afr_build_child_loc (this, &expunge_local->loc, &local->loc,
+ name);
+ if (ret != 0) {
+ op_errno = EINVAL;
+ goto out;
+ }
- gf_log (this->name, GF_LOG_TRACE,
- "skipping inspection of %s under %s",
- name, local->loc.path);
- goto out;
- }
-
- gf_log (this->name, GF_LOG_TRACE,
- "inspecting existance of %s under %s",
- name, local->loc.path);
-
- expunge_frame = copy_frame (frame);
- if (!expunge_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
- goto out;
- }
-
- ALLOC_OR_GOTO (expunge_local, afr_local_t, out);
-
- expunge_frame->local = expunge_local;
- expunge_sh = &expunge_local->self_heal;
- expunge_sh->sh_frame = frame;
- expunge_sh->active_source = active_src;
-
- ret = build_child_loc (this, &expunge_local->loc, &local->loc, name);
- if (ret != 0) {
- goto out;
- }
-
- gf_log (this->name, GF_LOG_TRACE,
- "looking up %s on %s", expunge_local->loc.path,
- priv->children[source]->name);
-
- STACK_WIND_COOKIE (expunge_frame,
- afr_sh_entry_expunge_entry_cbk,
- (void *) (long) source,
- priv->children[source],
- priv->children[source]->fops->lookup,
- &expunge_local->loc, 0);
-
- ret = 0;
+ gf_log (this->name, GF_LOG_TRACE,
+ "looking up %s on %s", expunge_local->loc.path,
+ priv->children[source]->name);
+
+ STACK_WIND_COOKIE (expunge_frame,
+ afr_sh_entry_expunge_entry_cbk,
+ (void *) (long) source,
+ priv->children[source],
+ priv->children[source]->fops->lookup,
+ &expunge_local->loc, NULL);
+
+ ret = 0;
out:
- if (ret == -1)
- afr_sh_entry_expunge_entry_done (frame, this, active_src);
+ if (ret == -1)
+ sh->expunge_done (frame, this, active_src, op_ret, op_errno);
- return 0;
+ return 0;
}
int
afr_sh_entry_expunge_readdir_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- gf_dirent_t *entries)
+ xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ gf_dirent_t *entries, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- gf_dirent_t *entry = NULL;
- off_t last_offset = 0;
- int active_src = 0;
- int entry_count = 0;
-
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
-
- active_src = sh->active_source;
-
- if (op_ret <= 0) {
- if (op_ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "readdir of %s on subvolume %s failed (%s)",
- local->loc.path,
- priv->children[active_src]->name,
- strerror (op_errno));
- } else {
- gf_log (this->name, GF_LOG_TRACE,
- "readdir of %s on subvolume %s complete",
- local->loc.path,
- priv->children[active_src]->name);
- }
-
- afr_sh_entry_expunge_all (frame, this);
- return 0;
- }
-
- list_for_each_entry (entry, &entries->list, list) {
- last_offset = entry->d_off;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ gf_dirent_t *entry = NULL;
+ off_t last_offset = 0;
+ int active_src = 0;
+ int entry_count = 0;
+
+ priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
+
+ active_src = sh->active_source;
+
+ if (op_ret <= 0) {
+ if (op_ret < 0) {
+ gf_log (this->name, GF_LOG_INFO,
+ "readdir of %s on subvolume %s failed (%s)",
+ local->loc.path,
+ priv->children[active_src]->name,
+ strerror (op_errno));
+ } else {
+ gf_log (this->name, GF_LOG_TRACE,
+ "readdir of %s on subvolume %s complete",
+ local->loc.path,
+ priv->children[active_src]->name);
+ }
+
+ afr_sh_entry_expunge_all (frame, this);
+ return 0;
+ }
+
+ list_for_each_entry (entry, &entries->list, list) {
+ last_offset = entry->d_off;
entry_count++;
- }
+ }
- gf_log (this->name, GF_LOG_TRACE,
- "readdir'ed %d entries from %s",
- entry_count, priv->children[active_src]->name);
+ gf_log (this->name, GF_LOG_TRACE,
+ "readdir'ed %d entries from %s",
+ entry_count, priv->children[active_src]->name);
- sh->offset = last_offset;
- local->call_count = entry_count;
+ sh->offset = last_offset;
+ local->call_count = entry_count;
- list_for_each_entry (entry, &entries->list, list) {
- afr_sh_entry_expunge_entry (frame, this, entry->d_name);
- }
+ list_for_each_entry (entry, &entries->list, list) {
+ afr_sh_entry_expunge_entry (frame, this, entry);
+ }
- return 0;
+ return 0;
}
int
afr_sh_entry_expunge_subvol (call_frame_t *frame, xlator_t *this,
- int active_src)
+ int active_src)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
+ priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
- STACK_WIND (frame, afr_sh_entry_expunge_readdir_cbk,
- priv->children[active_src],
- priv->children[active_src]->fops->readdir,
- sh->healing_fd, sh->block_size, sh->offset);
+ STACK_WIND (frame, afr_sh_entry_expunge_readdir_cbk,
+ priv->children[active_src],
+ priv->children[active_src]->fops->readdirp,
+ sh->healing_fd, sh->block_size, sh->offset, NULL);
- return 0;
+ return 0;
}
int
afr_sh_entry_expunge_all (call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- int active_src = -1;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ int active_src = -1;
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
+ priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
- sh->offset = 0;
+ sh->offset = 0;
- if (sh->source == -1) {
- gf_log (this->name, GF_LOG_TRACE,
- "no active sources for %s to expunge entries",
- local->loc.path);
- goto out;
- }
+ if (sh->source == -1) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "no active sources for %s to expunge entries",
+ local->loc.path);
+ goto out;
+ }
- active_src = next_active_sink (frame, this, sh->active_source);
- sh->active_source = active_src;
+ active_src = next_active_sink (frame, this, sh->active_source);
+ sh->active_source = active_src;
- if (sh->op_failed) {
- goto out;
- }
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+ goto out;
+ }
- if (active_src == -1) {
- /* completed creating missing files on all subvolumes */
- goto out;
- }
+ if (active_src == -1) {
+ /* completed creating missing files on all subvolumes */
+ goto out;
+ }
- gf_log (this->name, GF_LOG_TRACE,
- "expunging entries of %s on %s to other sinks",
- local->loc.path, priv->children[active_src]->name);
+ gf_log (this->name, GF_LOG_TRACE,
+ "expunging entries of %s on %s to other sinks",
+ local->loc.path, priv->children[active_src]->name);
- afr_sh_entry_expunge_subvol (frame, this, active_src);
+ afr_sh_entry_expunge_subvol (frame, this, active_src);
- return 0;
+ return 0;
out:
- afr_sh_entry_erase_pending (frame, this);
- return 0;
+ afr_sh_entry_impunge_all (frame, this);
+ return 0;
}
int
-afr_sh_entry_impunge_all (call_frame_t *frame, xlator_t *this);
-
-int
-afr_sh_entry_impunge_subvol (call_frame_t *frame, xlator_t *this,
- int active_src);
-
-int
afr_sh_entry_impunge_entry_done (call_frame_t *frame, xlator_t *this,
- int active_src)
+ int32_t op_ret, int32_t op_errno)
{
- int call_count = 0;
-
- call_count = afr_frame_return (frame);
+ int call_count = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
- if (call_count == 0)
- afr_sh_entry_impunge_subvol (frame, this, active_src);
+ local = frame->local;
+ sh = &local->self_heal;
+ if (op_ret < 0)
+ sh->entries_skipped = _gf_true;
+ call_count = afr_frame_return (frame);
+ if (call_count == 0)
+ afr_sh_entry_impunge_subvol (frame, this);
- return 0;
+ return 0;
}
-
-int
-afr_sh_entry_impunge_setattr_cbk (call_frame_t *impunge_frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop)
+void
+afr_sh_entry_call_impunge_done (call_frame_t *impunge_frame, xlator_t *this,
+ int32_t op_ret, int32_t op_errno)
{
- int call_count = 0;
- afr_private_t *priv = NULL;
- afr_local_t *impunge_local = NULL;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_self_heal_t *impunge_sh = NULL;
- call_frame_t *frame = NULL;
- int active_src = 0;
- int child_index = 0;
-
- priv = this->private;
- impunge_local = impunge_frame->local;
- impunge_sh = &impunge_local->self_heal;
- frame = impunge_sh->sh_frame;
- local = frame->local;
- sh = &local->self_heal;
- active_src = sh->active_source;
- child_index = (long) cookie;
-
- if (op_ret == 0) {
- gf_log (this->name, GF_LOG_TRACE,
- "setattr done for %s on %s",
- impunge_local->loc.path,
- priv->children[child_index]->name);
- } else {
- gf_log (this->name, GF_LOG_DEBUG,
- "setattr (%s) on %s failed (%s)",
- impunge_local->loc.path,
- priv->children[child_index]->name,
- strerror (op_errno));
- }
-
- LOCK (&impunge_frame->lock);
- {
- call_count = --impunge_local->call_count;
- }
- UNLOCK (&impunge_frame->lock);
-
- if (call_count == 0) {
- AFR_STACK_DESTROY (impunge_frame);
- afr_sh_entry_impunge_entry_done (frame, this, active_src);
- }
-
- return 0;
-}
+ afr_local_t *impunge_local = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ call_frame_t *frame = NULL;
+ AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
+ frame, local, sh);
+
+ AFR_STACK_DESTROY (impunge_frame);
+ sh->impunge_done (frame, this, op_ret, op_errno);
+}
int
-afr_sh_entry_impunge_xattrop_cbk (call_frame_t *impunge_frame, void *cookie,
+afr_sh_entry_impunge_setattr_cbk (call_frame_t *impunge_frame, void *cookie,
xlator_t *this,
int32_t op_ret, int32_t op_errno,
- dict_t *xattr)
+ struct iatt *preop, struct iatt *postop,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *impunge_local = NULL;
- afr_self_heal_t *impunge_sh = NULL;
- int child_index = 0;
-
- struct iatt stbuf;
- int32_t valid = 0;
-
- priv = this->private;
- impunge_local = impunge_frame->local;
- impunge_sh = &impunge_local->self_heal;
-
- child_index = (long) cookie;
-
- gf_log (this->name, GF_LOG_TRACE,
- "setting ownership of %s on %s to %d/%d",
- impunge_local->loc.path,
- priv->children[child_index]->name,
- impunge_local->cont.lookup.buf.ia_uid,
- impunge_local->cont.lookup.buf.ia_gid);
-
- stbuf.ia_atime = impunge_local->cont.lookup.buf.ia_atime;
- stbuf.ia_atime_nsec = impunge_local->cont.lookup.buf.ia_atime_nsec;
- stbuf.ia_mtime = impunge_local->cont.lookup.buf.ia_mtime;
- stbuf.ia_mtime_nsec = impunge_local->cont.lookup.buf.ia_mtime_nsec;
-
- stbuf.ia_uid = impunge_local->cont.lookup.buf.ia_uid;
- stbuf.ia_gid = impunge_local->cont.lookup.buf.ia_gid;
-
- valid = GF_SET_ATTR_UID | GF_SET_ATTR_GID |
- GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
-
- STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_setattr_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->setattr,
- &impunge_local->loc,
- &stbuf, valid);
+ int call_count = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
+ int child_index = 0;
+
+ priv = this->private;
+ impunge_local = impunge_frame->local;
+ child_index = (long) cookie;
+
+ if (op_ret == 0) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "setattr done for %s on %s",
+ impunge_local->loc.path,
+ priv->children[child_index]->name);
+ } else {
+ gf_log (this->name, GF_LOG_INFO,
+ "setattr (%s) on %s failed (%s)",
+ impunge_local->loc.path,
+ priv->children[child_index]->name,
+ strerror (op_errno));
+ }
+
+ call_count = afr_frame_return (impunge_frame);
+ if (call_count == 0) {
+ afr_sh_entry_call_impunge_done (impunge_frame, this,
+ 0, op_errno);
+ }
+
return 0;
}
-
int
afr_sh_entry_impunge_parent_setattr_cbk (call_frame_t *setattr_frame,
void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop)
+ struct iatt *preop, struct iatt *postop,
+ dict_t *xdata)
{
- loc_t *parent_loc = cookie;
+ int call_count = 0;
+ afr_local_t *setattr_local = NULL;
+ setattr_local = setattr_frame->local;
if (op_ret != 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "setattr on parent directory failed: %s",
- strerror (op_errno));
+ gf_log (this->name, GF_LOG_INFO,
+ "setattr on parent directory (%s) failed: %s",
+ setattr_local->loc.path, strerror (op_errno));
}
- loc_wipe (parent_loc);
+ call_count = afr_frame_return (setattr_frame);
+ if (call_count == 0)
+ AFR_STACK_DESTROY (setattr_frame);
+ return 0;
+}
+
+int
+afr_sh_entry_impunge_setattr (call_frame_t *impunge_frame, xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_local_t *setattr_local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ call_frame_t *setattr_frame = NULL;
+ int32_t valid = 0;
+ int32_t op_errno = 0;
+ int child_index = 0;
+ int call_count = 0;
+ int i = 0;
+
+ priv = this->private;
+ impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
+
+ gf_log (this->name, GF_LOG_DEBUG,
+ "setting ownership of %s on %s to %d/%d",
+ impunge_local->loc.path,
+ priv->children[child_index]->name,
+ impunge_sh->entrybuf.ia_uid,
+ impunge_sh->entrybuf.ia_gid);
+
+ setattr_frame = copy_frame (impunge_frame);
+ if (!setattr_frame) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+ AFR_LOCAL_ALLOC_OR_GOTO (setattr_frame->local, out);
+ setattr_local = setattr_frame->local;
+ call_count = afr_errno_count (NULL, impunge_sh->child_errno,
+ priv->child_count, 0);
+ loc_copy (&setattr_local->loc, &impunge_sh->parent_loc);
+ impunge_local->call_count = call_count;
+ setattr_local->call_count = call_count;
+ for (i = 0; i < priv->child_count; i++) {
+ if (impunge_sh->child_errno[i])
+ continue;
+ valid = GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
+ STACK_WIND_COOKIE (setattr_frame,
+ afr_sh_entry_impunge_parent_setattr_cbk,
+ (void *) (long) i, priv->children[i],
+ priv->children[i]->fops->setattr,
+ &setattr_local->loc,
+ &impunge_sh->parentbuf, valid, NULL);
+
+ valid = GF_SET_ATTR_UID | GF_SET_ATTR_GID |
+ GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
+ STACK_WIND_COOKIE (impunge_frame,
+ afr_sh_entry_impunge_setattr_cbk,
+ (void *) (long) i, priv->children[i],
+ priv->children[i]->fops->setattr,
+ &impunge_local->loc,
+ &impunge_sh->entrybuf, valid, NULL);
+ call_count--;
+ }
+ GF_ASSERT (!call_count);
+ return 0;
+out:
+ if (setattr_frame)
+ AFR_STACK_DESTROY (setattr_frame);
+ afr_sh_entry_call_impunge_done (impunge_frame, this, 0, op_errno);
+ return 0;
+}
+
+int
+afr_sh_entry_impunge_xattrop_cbk (call_frame_t *impunge_frame, void *cookie,
+ xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ dict_t *xattr, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
+ int child_index = 0;
+
+ priv = this->private;
+ impunge_local = impunge_frame->local;
- GF_FREE (parent_loc);
+ child_index = (long) cookie;
- AFR_STACK_DESTROY (setattr_frame);
+ if (op_ret == -1) {
+ gf_log (this->name, GF_LOG_INFO,
+ "%s: failed to perform xattrop on %s (%s)",
+ impunge_local->loc.path,
+ priv->children[child_index]->name,
+ strerror (op_errno));
+ goto out;
+ }
+
+ afr_sh_entry_impunge_setattr (impunge_frame, this);
+ return 0;
+out:
+ afr_sh_entry_call_impunge_done (impunge_frame, this,
+ -1, op_errno);
return 0;
}
+int
+afr_sh_entry_impunge_perform_xattrop (call_frame_t *impunge_frame,
+ xlator_t *this)
+{
+ int active_src = 0;
+ dict_t *xattr = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ int32_t op_errno = 0;
+
+ priv = this->private;
+ impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
+ active_src = impunge_sh->active_source;
+
+ afr_prepare_new_entry_pending_matrix (impunge_local->pending,
+ afr_is_errno_unset,
+ impunge_sh->child_errno,
+ &impunge_sh->entrybuf,
+ priv->child_count);
+ xattr = dict_new ();
+ if (!xattr) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ afr_set_pending_dict (priv, xattr, impunge_local->pending, active_src,
+ LOCAL_LAST);
+
+ STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_xattrop_cbk,
+ (void *) (long) active_src,
+ priv->children[active_src],
+ priv->children[active_src]->fops->xattrop,
+ &impunge_local->loc, GF_XATTROP_ADD_ARRAY, xattr, NULL);
+
+ if (xattr)
+ dict_unref (xattr);
+ return 0;
+out:
+ afr_sh_entry_call_impunge_done (impunge_frame, this,
+ -1, op_errno);
+ return 0;
+}
int
afr_sh_entry_impunge_newfile_cbk (call_frame_t *impunge_frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
+ xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
inode_t *inode, struct iatt *stbuf,
struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
- int call_count = 0;
- afr_private_t *priv = NULL;
- afr_local_t *impunge_local = NULL;
- afr_self_heal_t *impunge_sh = NULL;
- call_frame_t *frame = NULL;
- int active_src = 0;
- int child_index = 0;
- int pending_array[3] = {0, };
- dict_t *xattr = NULL;
- int ret = 0;
- int idx = 0;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
-
- call_frame_t *setattr_frame = NULL;
- int32_t valid = 0;
- loc_t *parent_loc = NULL;
- struct iatt parentbuf;
+ int call_count = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ int child_index = 0;
- priv = this->private;
- impunge_local = impunge_frame->local;
- impunge_sh = &impunge_local->self_heal;
- frame = impunge_sh->sh_frame;
- local = frame->local;
- sh = &local->self_heal;
- active_src = sh->active_source;
+ priv = this->private;
+ impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
- child_index = (long) cookie;
+ child_index = (long) cookie;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "creation of %s on %s failed (%s)",
- impunge_local->loc.path,
- priv->children[child_index]->name,
- strerror (op_errno));
- goto out;
- }
+ if (op_ret == -1) {
+ impunge_sh->child_errno[child_index] = op_errno;
+ gf_log (this->name, GF_LOG_ERROR,
+ "creation of %s on %s failed (%s)",
+ impunge_local->loc.path,
+ priv->children[child_index]->name,
+ strerror (op_errno));
+ } else {
+ impunge_sh->child_errno[child_index] = 0;
+ }
- inode->ia_type = stbuf->ia_type;
+ call_count = afr_frame_return (impunge_frame);
+ if (call_count == 0) {
+ if (!afr_errno_count (NULL, impunge_sh->child_errno,
+ priv->child_count, 0)) {
+ // new_file creation failed every where
+ afr_sh_entry_call_impunge_done (impunge_frame, this,
+ -1, op_errno);
+ goto out;
+ }
+ afr_sh_entry_impunge_perform_xattrop (impunge_frame, this);
+ }
+out:
+ return 0;
+}
- xattr = get_new_dict ();
- dict_ref (xattr);
+int
+afr_sh_entry_impunge_hardlink_cbk (call_frame_t *impunge_frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ int call_count = 0;
+ afr_local_t *impunge_local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+
+ impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
+
+ if (IA_IFLNK == impunge_sh->entrybuf.ia_type) {
+ //For symlinks impunge is attempted un-conditionally
+ //So the file can already exist.
+ if ((op_ret < 0) && (op_errno == EEXIST))
+ op_ret = 0;
+ }
- idx = afr_index_for_transaction_type (AFR_METADATA_TRANSACTION);
- pending_array[idx] = hton32 (1);
- if (IA_ISDIR (stbuf->ia_type))
- idx = afr_index_for_transaction_type (AFR_ENTRY_TRANSACTION);
- else
- idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
- pending_array[idx] = hton32 (1);
+ call_count = afr_frame_return (impunge_frame);
+ if (call_count == 0)
+ afr_sh_entry_call_impunge_done (impunge_frame, this,
+ op_ret, op_errno);
- ret = dict_set_static_bin (xattr, priv->pending_key[child_index],
- pending_array, sizeof (pending_array));
- if (ret < 0)
- gf_log (this->name, GF_LOG_WARNING,
- "Unable to set dict value.");
+ return 0;
+}
- valid = GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
- parentbuf = impunge_sh->parentbuf;
- setattr_frame = copy_frame (impunge_frame);
+int
+afr_sh_entry_impunge_hardlink (call_frame_t *impunge_frame, xlator_t *this,
+ int child_index)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ loc_t *loc = NULL;
+ struct iatt *buf = NULL;
+ loc_t oldloc = {0};
- parent_loc = GF_CALLOC (1, sizeof (*parent_loc),
- gf_afr_mt_loc_t);
- afr_build_parent_loc (parent_loc, &impunge_local->loc);
+ priv = this->private;
+ impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
+ loc = &impunge_local->loc;
+ buf = &impunge_sh->entrybuf;
- STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_xattrop_cbk,
- (void *) (long) child_index,
- priv->children[active_src],
- priv->children[active_src]->fops->xattrop,
- &impunge_local->loc, GF_XATTROP_ADD_ARRAY, xattr);
+ oldloc.inode = inode_ref (loc->inode);
+ uuid_copy (oldloc.gfid, buf->ia_gfid);
+ gf_log (this->name, GF_LOG_DEBUG, "linking missing file %s on %s",
+ loc->path, priv->children[child_index]->name);
- STACK_WIND_COOKIE (setattr_frame, afr_sh_entry_impunge_parent_setattr_cbk,
- (void *) (long) parent_loc,
+ STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_hardlink_cbk,
+ (void *) (long) child_index,
priv->children[child_index],
- priv->children[child_index]->fops->setattr,
- parent_loc, &parentbuf, valid);
+ priv->children[child_index]->fops->link,
+ &oldloc, loc, NULL);
+ loc_wipe (&oldloc);
- dict_unref (xattr);
+ return 0;
+}
+
+int
+afr_sh_nameless_lookup_cbk (call_frame_t *impunge_frame, void *cookie,
+ xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xattr,
+ struct iatt *postparent)
+{
+ if (op_ret < 0) {
+ afr_sh_entry_impunge_create_file (impunge_frame, this,
+ (long)cookie);
+ } else {
+ afr_sh_entry_impunge_hardlink (impunge_frame, this,
+ (long)cookie);
+ }
+ return 0;
+}
- return 0;
+int
+afr_sh_entry_impunge_check_hardlink (call_frame_t *impunge_frame,
+ xlator_t *this,
+ int child_index, struct iatt *stbuf)
+{
+ afr_private_t *priv = NULL;
+ call_frame_t *frame = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ afr_self_heal_t *sh = NULL;
+ loc_t *loc = NULL;
+ dict_t *xattr_req = NULL;
+ loc_t oldloc = {0};
+ int ret = -1;
+
+ priv = this->private;
+ AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
+ frame, local, sh);
+ loc = &impunge_local->loc;
+
+ xattr_req = dict_new ();
+ if (!xattr_req)
+ goto out;
+ oldloc.inode = inode_ref (loc->inode);
+ uuid_copy (oldloc.gfid, stbuf->ia_gfid);
+ STACK_WIND_COOKIE (impunge_frame, afr_sh_nameless_lookup_cbk,
+ (void *) (long) child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->lookup,
+ &oldloc, xattr_req);
+ ret = 0;
out:
- LOCK (&impunge_frame->lock);
- {
- call_count = --impunge_local->call_count;
- }
- UNLOCK (&impunge_frame->lock);
-
- if (call_count == 0) {
- AFR_STACK_DESTROY (impunge_frame);
- afr_sh_entry_impunge_entry_done (frame, this, active_src);
- }
-
- return 0;
+ if (xattr_req)
+ dict_unref (xattr_req);
+ loc_wipe (&oldloc);
+ if (ret)
+ sh->impunge_done (frame, this, -1, ENOMEM);
+ return 0;
}
-
int
afr_sh_entry_impunge_mknod (call_frame_t *impunge_frame, xlator_t *this,
- int child_index, struct iatt *stbuf)
+ int child_index, struct iatt *stbuf)
{
- afr_private_t *priv = NULL;
- afr_local_t *impunge_local = NULL;
- dict_t *dict = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
+ dict_t *dict = NULL;
+ int ret = 0;
- int ret = 0;
+ priv = this->private;
+ impunge_local = impunge_frame->local;
- priv = this->private;
- impunge_local = impunge_frame->local;
-
- gf_log (this->name, GF_LOG_DEBUG,
- "creating missing file %s on %s",
- impunge_local->loc.path,
- priv->children[child_index]->name);
+ gf_log (this->name, GF_LOG_DEBUG,
+ "creating missing file %s on %s",
+ impunge_local->loc.path,
+ priv->children[child_index]->name);
dict = dict_new ();
if (!dict)
gf_log (this->name, GF_LOG_ERROR, "Out of memory");
+ GF_ASSERT (!uuid_is_null (stbuf->ia_gfid));
ret = afr_set_dict_gfid (dict, stbuf->ia_gfid);
if (ret)
- gf_log (this->name, GF_LOG_DEBUG, "gfid set failed");
+ gf_log (this->name, GF_LOG_INFO, "%s: gfid set failed",
+ impunge_local->loc.path);
+
+ /*
+ * Reason for adding GLUSTERFS_INTERNAL_FOP_KEY :
+ *
+ * Problem:
+ * While a brick is down in a replica pair, lets say the user creates
+ * one file(file-A) and a hard link to that file(h-file-A). After the
+ * brick comes back up, entry self-heal is attempted on parent dir of
+ * these two files. As part of readdir in self-heal it reads both the
+ * entries file-A and h-file-A for both of them it does name less lookup
+ * to check if there are any hardlinks already present in the
+ * destination brick. It finds that there are no hard links already
+ * present for files file-A, h-file-A. Self-heal does mknods for both
+ * file-A and h-file-A. This leads to file-A and h-file-A not being
+ * hardlinks anymore.
+ *
+ * Fix: (More like shrinking of race-window, the race itself is still
+ * present in posix-mknod).
+ * If mknod comes with the presence of GLUSTERFS_INTERNAL_FOP_KEY then
+ * posix_mknod checks if there are already any gfid-links and does
+ * link() instead of mknod. There still can be a race where two
+ * posix_mknods same gfid see that
+ * gfid-link file is not present and proceeds with mknods and result in
+ * two different files with same gfid.
+ */
+ ret = dict_set_str (dict, GLUSTERFS_INTERNAL_FOP_KEY, "yes");
+ if (ret)
+ gf_log (this->name, GF_LOG_INFO, "%s: %s set failed",
+ impunge_local->loc.path, GLUSTERFS_INTERNAL_FOP_KEY);
- STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_newfile_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->mknod,
- &impunge_local->loc,
- st_mode_from_ia (stbuf->ia_prot, stbuf->ia_type),
- stbuf->ia_rdev, dict);
+ STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_newfile_cbk,
+ (void *) (long) child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->mknod,
+ &impunge_local->loc,
+ st_mode_from_ia (stbuf->ia_prot, stbuf->ia_type),
+ makedev (ia_major (stbuf->ia_rdev),
+ ia_minor (stbuf->ia_rdev)), 0, dict);
if (dict)
dict_unref (dict);
- return 0;
+ return 0;
}
int
afr_sh_entry_impunge_mkdir (call_frame_t *impunge_frame, xlator_t *this,
- int child_index, struct iatt *stbuf)
+ int child_index, struct iatt *stbuf)
{
- afr_private_t *priv = NULL;
- afr_local_t *impunge_local = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
dict_t *dict = NULL;
int ret = 0;
- priv = this->private;
- impunge_local = impunge_frame->local;
+ priv = this->private;
+ impunge_local = impunge_frame->local;
dict = dict_new ();
if (!dict) {
@@ -1345,73 +1329,76 @@ afr_sh_entry_impunge_mkdir (call_frame_t *impunge_frame, xlator_t *this,
return 0;
}
+ GF_ASSERT (!uuid_is_null (stbuf->ia_gfid));
ret = afr_set_dict_gfid (dict, stbuf->ia_gfid);
if (ret)
- gf_log (this->name, GF_LOG_DEBUG, "gfid set failed");
-
- gf_log (this->name, GF_LOG_DEBUG,
- "creating missing directory %s on %s",
- impunge_local->loc.path,
- priv->children[child_index]->name);
-
- STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_newfile_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->mkdir,
- &impunge_local->loc,
+ gf_log (this->name, GF_LOG_INFO, "%s: gfid set failed",
+ impunge_local->loc.path);
+
+ gf_log (this->name, GF_LOG_DEBUG,
+ "creating missing directory %s on %s",
+ impunge_local->loc.path,
+ priv->children[child_index]->name);
+
+ STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_newfile_cbk,
+ (void *) (long) child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->mkdir,
+ &impunge_local->loc,
st_mode_from_ia (stbuf->ia_prot, stbuf->ia_type),
- dict);
+ 0, dict);
if (dict)
dict_unref (dict);
- return 0;
+ return 0;
}
int
afr_sh_entry_impunge_symlink (call_frame_t *impunge_frame, xlator_t *this,
- int child_index, const char *linkname)
+ int child_index, const char *linkname)
{
- afr_private_t *priv = NULL;
- afr_local_t *impunge_local = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
dict_t *dict = NULL;
struct iatt *buf = NULL;
+ int ret = 0;
- int ret = 0;
+ priv = this->private;
+ impunge_local = impunge_frame->local;
- priv = this->private;
- impunge_local = impunge_frame->local;
-
- buf = &impunge_local->cont.symlink.buf;
+ buf = &impunge_local->cont.dir_fop.buf;
dict = dict_new ();
if (!dict) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- afr_sh_entry_impunge_entry_done (impunge_frame, this, 0);
+ afr_sh_entry_call_impunge_done (impunge_frame, this,
+ -1, ENOMEM);
+ goto out;
}
+ GF_ASSERT (!uuid_is_null (buf->ia_gfid));
ret = afr_set_dict_gfid (dict, buf->ia_gfid);
if (ret)
- gf_log (this->name, GF_LOG_DEBUG,
- "dict set gfid failed");
+ gf_log (this->name, GF_LOG_INFO,
+ "%s: dict set gfid failed",
+ impunge_local->loc.path);
- gf_log (this->name, GF_LOG_DEBUG,
- "creating missing symlink %s -> %s on %s",
- impunge_local->loc.path, linkname,
- priv->children[child_index]->name);
+ gf_log (this->name, GF_LOG_DEBUG,
+ "creating missing symlink %s -> %s on %s",
+ impunge_local->loc.path, linkname,
+ priv->children[child_index]->name);
- STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_newfile_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->symlink,
- linkname, &impunge_local->loc, dict);
+ STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_newfile_cbk,
+ (void *) (long) child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->symlink,
+ linkname, &impunge_local->loc, 0, dict);
if (dict)
dict_unref (dict);
-
- return 0;
+out:
+ return 0;
}
@@ -1420,50 +1407,45 @@ afr_sh_entry_impunge_symlink_unlink_cbk (call_frame_t *impunge_frame,
void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
afr_private_t *priv = NULL;
- afr_local_t *impunge_local = NULL;
- afr_self_heal_t *impunge_sh = NULL;
- int child_index = -1;
- call_frame_t *frame = NULL;
- int call_count = -1;
- int active_src = -1;
-
- priv = this->private;
- impunge_local = impunge_frame->local;
- impunge_sh = &impunge_local->self_heal;
- frame = impunge_sh->sh_frame;
- active_src = impunge_sh->active_source;
-
- child_index = (long) cookie;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "unlink of %s on %s failed (%s)",
- impunge_local->loc.path,
- priv->children[child_index]->name,
- strerror (op_errno));
- goto out;
- }
+ afr_local_t *impunge_local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ int child_index = -1;
+ int call_count = -1;
+
+ priv = this->private;
+ impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
+
+ child_index = (long) cookie;
+
+ if (op_ret == -1) {
+ gf_log (this->name, GF_LOG_INFO,
+ "unlink of %s on %s failed (%s)",
+ impunge_local->loc.path,
+ priv->children[child_index]->name,
+ strerror (op_errno));
+ goto out;
+ }
afr_sh_entry_impunge_symlink (impunge_frame, this, child_index,
impunge_sh->linkname);
return 0;
out:
- LOCK (&impunge_frame->lock);
- {
- call_count = --impunge_local->call_count;
- }
- UNLOCK (&impunge_frame->lock);
-
- if (call_count == 0) {
- AFR_STACK_DESTROY (impunge_frame);
- afr_sh_entry_impunge_entry_done (frame, this, active_src);
- }
-
- return 0;
+ LOCK (&impunge_frame->lock);
+ {
+ call_count = --impunge_local->call_count;
+ }
+ UNLOCK (&impunge_frame->lock);
+
+ if (call_count == 0)
+ afr_sh_entry_call_impunge_done (impunge_frame, this,
+ op_ret, op_errno);
+
+ return 0;
}
@@ -1471,22 +1453,22 @@ int
afr_sh_entry_impunge_symlink_unlink (call_frame_t *impunge_frame, xlator_t *this,
int child_index)
{
- afr_private_t *priv = NULL;
- afr_local_t *impunge_local = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
- priv = this->private;
- impunge_local = impunge_frame->local;
+ priv = this->private;
+ impunge_local = impunge_frame->local;
- gf_log (this->name, GF_LOG_DEBUG,
- "unlinking symlink %s with wrong target on %s",
- impunge_local->loc.path,
- priv->children[child_index]->name);
+ gf_log (this->name, GF_LOG_DEBUG,
+ "unlinking symlink %s with wrong target on %s",
+ impunge_local->loc.path,
+ priv->children[child_index]->name);
STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_symlink_unlink_cbk,
(void *) (long) child_index,
priv->children[child_index],
priv->children[child_index]->fops->unlink,
- &impunge_local->loc);
+ &impunge_local->loc, 0, NULL);
return 0;
}
@@ -1496,32 +1478,30 @@ int
afr_sh_entry_impunge_readlink_sink_cbk (call_frame_t *impunge_frame, void *cookie,
xlator_t *this,
int32_t op_ret, int32_t op_errno,
- const char *linkname, struct iatt *sbuf)
+ const char *linkname, struct iatt *sbuf, dict_t *xdata)
{
afr_private_t *priv = NULL;
- afr_local_t *impunge_local = NULL;
- afr_self_heal_t *impunge_sh = NULL;
- int child_index = -1;
- call_frame_t *frame = NULL;
- int call_count = -1;
- int active_src = -1;
-
- priv = this->private;
- impunge_local = impunge_frame->local;
- impunge_sh = &impunge_local->self_heal;
- frame = impunge_sh->sh_frame;
- active_src = impunge_sh->active_source;
-
- child_index = (long) cookie;
-
- if ((op_ret == -1) && (op_errno != ENOENT)) {
- gf_log (this->name, GF_LOG_DEBUG,
- "readlink of %s on %s failed (%s)",
- impunge_local->loc.path,
- priv->children[active_src]->name,
- strerror (op_errno));
- goto out;
- }
+ afr_local_t *impunge_local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ int child_index = -1;
+ int call_count = -1;
+ int active_src = -1;
+
+ priv = this->private;
+ impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
+ active_src = impunge_sh->active_source;
+
+ child_index = (long) cookie;
+
+ if ((op_ret == -1) && (op_errno != ENOENT)) {
+ gf_log (this->name, GF_LOG_INFO,
+ "readlink of %s on %s failed (%s)",
+ impunge_local->loc.path,
+ priv->children[active_src]->name,
+ strerror (op_errno));
+ goto out;
+ }
/* symlink doesn't exist on the sink */
@@ -1548,18 +1528,17 @@ afr_sh_entry_impunge_readlink_sink_cbk (call_frame_t *impunge_frame, void *cooki
}
out:
- LOCK (&impunge_frame->lock);
- {
- call_count = --impunge_local->call_count;
- }
- UNLOCK (&impunge_frame->lock);
-
- if (call_count == 0) {
- AFR_STACK_DESTROY (impunge_frame);
- afr_sh_entry_impunge_entry_done (frame, this, active_src);
- }
-
- return 0;
+ LOCK (&impunge_frame->lock);
+ {
+ call_count = --impunge_local->call_count;
+ }
+ UNLOCK (&impunge_frame->lock);
+
+ if (call_count == 0)
+ afr_sh_entry_call_impunge_done (impunge_frame, this,
+ op_ret, op_errno);
+
+ return 0;
}
@@ -1567,821 +1546,770 @@ int
afr_sh_entry_impunge_readlink_sink (call_frame_t *impunge_frame, xlator_t *this,
int child_index)
{
- afr_private_t *priv = NULL;
- afr_local_t *impunge_local = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
- priv = this->private;
- impunge_local = impunge_frame->local;
+ priv = this->private;
+ impunge_local = impunge_frame->local;
- gf_log (this->name, GF_LOG_DEBUG,
- "checking symlink target of %s on %s",
- impunge_local->loc.path, priv->children[child_index]->name);
+ gf_log (this->name, GF_LOG_DEBUG,
+ "checking symlink target of %s on %s",
+ impunge_local->loc.path, priv->children[child_index]->name);
- STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_readlink_sink_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->readlink,
- &impunge_local->loc, 4096);
+ STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_readlink_sink_cbk,
+ (void *) (long) child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->readlink,
+ &impunge_local->loc, 4096, NULL);
- return 0;
+ return 0;
}
int
afr_sh_entry_impunge_readlink_cbk (call_frame_t *impunge_frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- const char *linkname, struct iatt *sbuf)
+ xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ const char *linkname, struct iatt *sbuf, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *impunge_local = NULL;
- afr_self_heal_t *impunge_sh = NULL;
- int child_index = -1;
- call_frame_t *frame = NULL;
- int call_count = -1;
- int active_src = -1;
-
- priv = this->private;
- impunge_local = impunge_frame->local;
- impunge_sh = &impunge_local->self_heal;
- frame = impunge_sh->sh_frame;
- active_src = impunge_sh->active_source;
-
- child_index = (long) cookie;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "readlink of %s on %s failed (%s)",
- impunge_local->loc.path,
- priv->children[active_src]->name,
- strerror (op_errno));
- goto out;
- }
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ int child_index = -1;
+ int call_count = -1;
+ int active_src = -1;
+
+ priv = this->private;
+ impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
+ active_src = impunge_sh->active_source;
+
+ child_index = (long) cookie;
+
+ if (op_ret == -1) {
+ gf_log (this->name, GF_LOG_INFO,
+ "readlink of %s on %s failed (%s)",
+ impunge_local->loc.path,
+ priv->children[active_src]->name,
+ strerror (op_errno));
+ goto out;
+ }
impunge_sh->linkname = gf_strdup (linkname);
- afr_sh_entry_impunge_readlink_sink (impunge_frame, this, child_index);
+ afr_sh_entry_impunge_readlink_sink (impunge_frame, this, child_index);
- return 0;
+ return 0;
out:
- LOCK (&impunge_frame->lock);
- {
- call_count = --impunge_local->call_count;
- }
- UNLOCK (&impunge_frame->lock);
-
- if (call_count == 0) {
- AFR_STACK_DESTROY (impunge_frame);
- afr_sh_entry_impunge_entry_done (frame, this, active_src);
- }
-
- return 0;
+ LOCK (&impunge_frame->lock);
+ {
+ call_count = --impunge_local->call_count;
+ }
+ UNLOCK (&impunge_frame->lock);
+
+ if (call_count == 0)
+ afr_sh_entry_call_impunge_done (impunge_frame, this,
+ op_ret, op_errno);
+
+ return 0;
}
int
afr_sh_entry_impunge_readlink (call_frame_t *impunge_frame, xlator_t *this,
- int child_index, struct iatt *stbuf)
+ int child_index, struct iatt *stbuf)
{
- afr_private_t *priv = NULL;
- afr_local_t *impunge_local = NULL;
- afr_self_heal_t *impunge_sh = NULL;
- int active_src = -1;
-
- priv = this->private;
- impunge_local = impunge_frame->local;
- impunge_sh = &impunge_local->self_heal;
- active_src = impunge_sh->active_source;
- impunge_local->cont.symlink.buf = *stbuf;
-
- STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_readlink_cbk,
- (void *) (long) child_index,
- priv->children[active_src],
- priv->children[active_src]->fops->readlink,
- &impunge_local->loc, 4096);
-
- return 0;
-}
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ int active_src = -1;
+
+ priv = this->private;
+ impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
+ active_src = impunge_sh->active_source;
+ impunge_local->cont.dir_fop.buf = *stbuf;
+ STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_readlink_cbk,
+ (void *) (long) child_index,
+ priv->children[active_src],
+ priv->children[active_src]->fops->readlink,
+ &impunge_local->loc, 4096, NULL);
+
+ return 0;
+}
int
-afr_sh_entry_impunge_recreate_lookup_cbk (call_frame_t *impunge_frame,
- void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf,
- dict_t *xattr,struct iatt *postparent)
+afr_sh_entry_impunge_create (call_frame_t *impunge_frame, xlator_t *this,
+ int child_index)
{
- afr_private_t *priv = NULL;
- afr_local_t *impunge_local = NULL;
- afr_self_heal_t *impunge_sh = NULL;
- int active_src = 0;
- int type = 0;
- int child_index = 0;
- call_frame_t *frame = NULL;
- int call_count = 0;
-
- priv = this->private;
- impunge_local = impunge_frame->local;
- impunge_sh = &impunge_local->self_heal;
- frame = impunge_sh->sh_frame;
-
- child_index = (long) cookie;
-
- active_src = impunge_sh->active_source;
-
- if (op_ret != 0) {
- gf_log (this->name, GF_LOG_TRACE,
- "looking up %s on %s (for %s) failed (%s)",
- impunge_local->loc.path,
- priv->children[active_src]->name,
- priv->children[child_index]->name,
- strerror (op_errno));
- goto out;
- }
-
- impunge_sh->parentbuf = *postparent;
-
- impunge_local->cont.lookup.buf = *buf;
- type = buf->ia_type;
-
- switch (type) {
- case IA_IFSOCK:
- case IA_IFREG:
- case IA_IFBLK:
- case IA_IFCHR:
- case IA_IFIFO:
- afr_sh_entry_impunge_mknod (impunge_frame, this,
- child_index, buf);
- break;
- case IA_IFLNK:
- afr_sh_entry_impunge_readlink (impunge_frame, this,
- child_index, buf);
- break;
- case IA_IFDIR:
- afr_sh_entry_impunge_mkdir (impunge_frame, this,
- child_index, buf);
- break;
- default:
- gf_log (this->name, GF_LOG_ERROR,
- "%s has unknown file type on %s: 0%o",
- impunge_local->loc.path,
- priv->children[active_src]->name, type);
- goto out;
- break;
- }
-
- return 0;
+ call_frame_t *frame = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ ia_type_t type = IA_INVAL;
+ int active_src = 0;
+ struct iatt *buf = NULL;
+
+ AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
+ frame, local, sh);
+ active_src = impunge_sh->active_source;
+ afr_update_loc_gfids (&impunge_local->loc, &impunge_sh->entrybuf,
+ &impunge_sh->parentbuf);
+
+ buf = &impunge_sh->entrybuf;
+ type = buf->ia_type;
+
+ switch (type) {
+ case IA_IFSOCK:
+ case IA_IFREG:
+ case IA_IFBLK:
+ case IA_IFCHR:
+ case IA_IFIFO:
+ case IA_IFLNK:
+ afr_sh_entry_impunge_check_hardlink (impunge_frame, this,
+ child_index, buf);
+ break;
+ case IA_IFDIR:
+ afr_sh_entry_impunge_mkdir (impunge_frame, this,
+ child_index, buf);
+ break;
+ default:
+ gf_log (this->name, GF_LOG_ERROR,
+ "%s has unknown file type on %s: 0%o",
+ impunge_local->loc.path,
+ priv->children[active_src]->name, type);
+ sh->impunge_done (frame, this, -1, EINVAL);
+ break;
+ }
-out:
- LOCK (&impunge_frame->lock);
- {
- call_count = --impunge_local->call_count;
- }
- UNLOCK (&impunge_frame->lock);
-
- if (call_count == 0) {
- AFR_STACK_DESTROY (impunge_frame);
- afr_sh_entry_impunge_entry_done (frame, this, active_src);
- }
-
- return 0;
+ return 0;
}
-
int
-afr_sh_entry_impunge_recreate (call_frame_t *impunge_frame, xlator_t *this,
- int child_index)
+afr_sh_entry_impunge_create_file (call_frame_t *impunge_frame, xlator_t *this,
+ int child_index)
{
- afr_private_t *priv = NULL;
- afr_local_t *impunge_local = NULL;
- afr_self_heal_t *impunge_sh = NULL;
- int active_src = 0;
+ call_frame_t *frame = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ ia_type_t type = IA_INVAL;
+ int active_src = 0;
+ struct iatt *buf = NULL;
+
+ AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
+ frame, local, sh);
+ active_src = impunge_sh->active_source;
+ buf = &impunge_sh->entrybuf;
+ type = buf->ia_type;
+
+ switch (type) {
+ case IA_IFSOCK:
+ case IA_IFREG:
+ case IA_IFBLK:
+ case IA_IFCHR:
+ case IA_IFIFO:
+ afr_sh_entry_impunge_mknod (impunge_frame, this,
+ child_index, buf);
+ break;
+ case IA_IFLNK:
+ afr_sh_entry_impunge_readlink (impunge_frame, this,
+ child_index, buf);
+ break;
+ default:
+ gf_log (this->name, GF_LOG_ERROR,
+ "%s has unknown file type on %s: 0%o",
+ impunge_local->loc.path,
+ priv->children[active_src]->name, type);
+ sh->impunge_done (frame, this, -1, EINVAL);
+ break;
+ }
+ return 0;
+}
- priv = this->private;
- impunge_local = impunge_frame->local;
- impunge_sh = &impunge_local->self_heal;
+gf_boolean_t
+afr_sh_need_recreate (afr_self_heal_t *impunge_sh, unsigned int child,
+ unsigned int child_count)
+{
+ gf_boolean_t recreate = _gf_false;
- active_src = impunge_sh->active_source;
+ GF_ASSERT (impunge_sh->child_errno);
- STACK_WIND_COOKIE (impunge_frame,
- afr_sh_entry_impunge_recreate_lookup_cbk,
- (void *) (long) child_index,
- priv->children[active_src],
- priv->children[active_src]->fops->lookup,
- &impunge_local->loc, 0);
+ if (child == impunge_sh->active_source)
+ goto out;
- return 0;
+ if (IA_IFLNK == impunge_sh->entrybuf.ia_type) {
+ recreate = _gf_true;
+ goto out;
+ }
+
+ if (impunge_sh->child_errno[child] == ENOENT)
+ recreate = _gf_true;
+out:
+ return recreate;
}
+unsigned int
+afr_sh_recreate_count (afr_self_heal_t *impunge_sh, int *sources,
+ unsigned int child_count)
+{
+ int count = 0;
+ int i = 0;
+
+ for (i = 0; i < child_count; i++) {
+ if (afr_sh_need_recreate (impunge_sh, i, child_count))
+ count++;
+ }
+
+ return count;
+}
int
-afr_sh_entry_impunge_entry_cbk (call_frame_t *impunge_frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf, dict_t *x,
- struct iatt *postparent)
+afr_sh_entry_call_impunge_recreate (call_frame_t *impunge_frame,
+ xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *impunge_local = NULL;
- afr_self_heal_t *impunge_sh = NULL;
- int call_count = 0;
- int child_index = 0;
- call_frame_t *frame = NULL;
- int active_src = 0;
-
- priv = this->private;
- impunge_local = impunge_frame->local;
- impunge_sh = &impunge_local->self_heal;
- frame = impunge_sh->sh_frame;
- child_index = (long) cookie;
- active_src = impunge_sh->active_source;
-
- if ((op_ret == -1 && op_errno == ENOENT)
- || (IA_ISLNK (impunge_sh->impunging_entry_mode))) {
-
- /*
- * A symlink's target might have changed, so
- * always go down the recreate path for them.
- */
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ unsigned int recreate_count = 0;
+ int i = 0;
+ int active_src = 0;
- /* decrease call_count in recreate-callback */
-
- gf_log (this->name, GF_LOG_TRACE,
- "missing entry %s on %s",
- impunge_local->loc.path,
- priv->children[child_index]->name);
-
- afr_sh_entry_impunge_recreate (impunge_frame, this,
- child_index);
- return 0;
- }
-
- if (op_ret == 0) {
- gf_log (this->name, GF_LOG_TRACE,
- "%s exists under %s",
- impunge_local->loc.path,
- priv->children[child_index]->name);
-
- impunge_sh->parentbuf = *postparent;
- } else {
- gf_log (this->name, GF_LOG_TRACE,
- "looking up %s under %s failed (%s)",
- impunge_local->loc.path,
- priv->children[child_index]->name,
- strerror (op_errno));
- }
-
- LOCK (&impunge_frame->lock);
- {
- call_count = --impunge_local->call_count;
- }
- UNLOCK (&impunge_frame->lock);
-
- if (call_count == 0) {
- AFR_STACK_DESTROY (impunge_frame);
- afr_sh_entry_impunge_entry_done (frame, this, active_src);
- }
-
- return 0;
+ priv = this->private;
+ AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
+ frame, local, sh);
+ active_src = impunge_sh->active_source;
+ impunge_sh->entrybuf = impunge_sh->buf[active_src];
+ impunge_sh->parentbuf = impunge_sh->parentbufs[active_src];
+ recreate_count = afr_sh_recreate_count (impunge_sh, sh->sources,
+ priv->child_count);
+ if (!recreate_count) {
+ afr_sh_entry_call_impunge_done (impunge_frame, this, 0, 0);
+ goto out;
+ }
+ impunge_local->call_count = recreate_count;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!impunge_local->child_up[i]) {
+ impunge_sh->child_errno[i] = ENOTCONN;
+ continue;
+ }
+ if (!afr_sh_need_recreate (impunge_sh, i, priv->child_count)) {
+ impunge_sh->child_errno[i] = EEXIST;
+ continue;
+ }
+ }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!afr_sh_need_recreate (impunge_sh, i, priv->child_count))
+ continue;
+ (void)afr_sh_entry_impunge_create (impunge_frame, this, i);
+ recreate_count--;
+ }
+ GF_ASSERT (!recreate_count);
+out:
+ return 0;
}
+void
+afr_sh_entry_common_lookup_done (call_frame_t *impunge_frame, xlator_t *this,
+ int32_t op_ret, int32_t op_errno)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ unsigned int gfid_miss_count = 0;
+ unsigned int children_up_count = 0;
+ uuid_t gfid = {0};
+ int active_src = 0;
+
+ priv = this->private;
+ AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
+ frame, local, sh);
+ active_src = impunge_sh->active_source;
+
+ if (op_ret < 0)
+ goto done;
+ if (impunge_sh->child_errno[active_src]) {
+ op_ret = -1;
+ op_errno = impunge_sh->child_errno[active_src];
+ goto done;
+ }
+
+ gfid_miss_count = afr_gfid_missing_count (this->name,
+ impunge_sh->success_children,
+ impunge_sh->buf, priv->child_count,
+ impunge_local->loc.path);
+ children_up_count = afr_up_children_count (impunge_local->child_up,
+ priv->child_count);
+ if ((gfid_miss_count == children_up_count) &&
+ (children_up_count < priv->child_count)) {
+ op_ret = -1;
+ op_errno = ENODATA;
+ gf_log (this->name, GF_LOG_ERROR, "Not all children are up, "
+ "gfid should not be assigned in this state for %s",
+ impunge_local->loc.path);
+ goto done;
+ }
+
+ if (gfid_miss_count) {
+ afr_update_gfid_from_iatts (gfid, impunge_sh->buf,
+ impunge_sh->success_children,
+ priv->child_count);
+ if (uuid_is_null (gfid)) {
+ sh->entries_skipped = _gf_true;
+ gf_log (this->name, GF_LOG_INFO, "%s: Skipping entry "
+ "self-heal because of gfid absence",
+ impunge_local->loc.path);
+ goto done;
+ }
+ afr_sh_common_lookup (impunge_frame, this, &impunge_local->loc,
+ afr_sh_entry_common_lookup_done, gfid,
+ AFR_LOOKUP_FAIL_CONFLICTS |
+ AFR_LOOKUP_FAIL_MISSING_GFIDS,
+ NULL);
+ } else {
+ afr_sh_entry_call_impunge_recreate (impunge_frame, this);
+ }
+ return;
+done:
+ afr_sh_entry_call_impunge_done (impunge_frame, this,
+ op_ret, op_errno);
+ return;
+}
int
afr_sh_entry_impunge_entry (call_frame_t *frame, xlator_t *this,
- gf_dirent_t *entry)
+ gf_dirent_t *entry)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- int ret = -1;
- call_frame_t *impunge_frame = NULL;
- afr_local_t *impunge_local = NULL;
- afr_self_heal_t *impunge_sh = NULL;
- int active_src = 0;
- int i = 0;
- int call_count = 0;
- int op_errno = 0;
-
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
-
- active_src = sh->active_source;
-
- if ((strcmp (entry->d_name, ".") == 0)
- || (strcmp (entry->d_name, "..") == 0)
- || ((strcmp (local->loc.path, "/") == 0)
- && (strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR) == 0))) {
-
- gf_log (this->name, GF_LOG_TRACE,
- "skipping inspection of %s under %s",
- entry->d_name, local->loc.path);
- goto out;
- }
-
- gf_log (this->name, GF_LOG_TRACE,
- "inspecting existance of %s under %s",
- entry->d_name, local->loc.path);
-
- impunge_frame = copy_frame (frame);
- if (!impunge_frame) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
- goto out;
- }
-
- ALLOC_OR_GOTO (impunge_local, afr_local_t, out);
-
- impunge_frame->local = impunge_local;
- impunge_sh = &impunge_local->self_heal;
- impunge_sh->sh_frame = frame;
- impunge_sh->active_source = active_src;
-
- impunge_sh->impunging_entry_mode =
- st_mode_from_ia (entry->d_stat.ia_prot, entry->d_stat.ia_type);
-
- ret = build_child_loc (this, &impunge_local->loc, &local->loc, entry->d_name);
- if (ret != 0) {
- goto out;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (i == active_src)
- continue;
- if (local->child_up[i] == 0)
- continue;
- if (sh->sources[i] == 1)
- continue;
- call_count++;
- }
-
- impunge_local->call_count = call_count;
-
- for (i = 0; i < priv->child_count; i++) {
- if (i == active_src)
- continue;
- if (local->child_up[i] == 0)
- continue;
- if (sh->sources[i] == 1)
- continue;
-
- gf_log (this->name, GF_LOG_TRACE,
- "looking up %s on %s", impunge_local->loc.path,
- priv->children[i]->name);
-
- STACK_WIND_COOKIE (impunge_frame,
- afr_sh_entry_impunge_entry_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->lookup,
- &impunge_local->loc, 0);
-
- if (!--call_count)
- break;
- }
-
- ret = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ int ret = -1;
+ call_frame_t *impunge_frame = NULL;
+ afr_local_t *impunge_local = NULL;
+ int active_src = 0;
+ int op_errno = 0;
+ int op_ret = -1;
+
+ local = frame->local;
+ sh = &local->self_heal;
+
+ active_src = sh->active_source;
+ sh->impunge_done = afr_sh_entry_impunge_entry_done;
+
+ if (can_skip_entry_self_heal (entry->d_name, &local->loc)) {
+ op_ret = 0;
+ goto out;
+ }
+
+ gf_log (this->name, GF_LOG_TRACE,
+ "inspecting existence of %s under %s",
+ entry->d_name, local->loc.path);
+
+ ret = afr_impunge_frame_create (frame, this, active_src,
+ &impunge_frame);
+ if (ret) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
+ ret = afr_build_child_loc (this, &impunge_local->loc, &local->loc,
+ entry->d_name);
+ loc_copy (&impunge_sh->parent_loc, &local->loc);
+ if (ret != 0) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ afr_sh_common_lookup (impunge_frame, this, &impunge_local->loc,
+ afr_sh_entry_common_lookup_done, NULL,
+ AFR_LOOKUP_FAIL_CONFLICTS, NULL);
+
+ op_ret = 0;
out:
- if (ret == -1)
- afr_sh_entry_impunge_entry_done (frame, this, active_src);
+ if (ret) {
+ if (impunge_frame)
+ AFR_STACK_DESTROY (impunge_frame);
+ sh->impunge_done (frame, this, op_ret, op_errno);
+ }
- return 0;
+ return 0;
}
int
afr_sh_entry_impunge_readdir_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- gf_dirent_t *entries)
+ xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ gf_dirent_t *entries, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- gf_dirent_t *entry = NULL;
- off_t last_offset = 0;
- int active_src = 0;
- int entry_count = 0;
-
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
-
- active_src = sh->active_source;
-
- if (op_ret <= 0) {
- if (op_ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "readdir of %s on subvolume %s failed (%s)",
- local->loc.path,
- priv->children[active_src]->name,
- strerror (op_errno));
- } else {
- gf_log (this->name, GF_LOG_TRACE,
- "readdir of %s on subvolume %s complete",
- local->loc.path,
- priv->children[active_src]->name);
- }
-
- afr_sh_entry_impunge_all (frame, this);
- return 0;
- }
-
- list_for_each_entry (entry, &entries->list, list) {
- last_offset = entry->d_off;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ gf_dirent_t *entry = NULL;
+ off_t last_offset = 0;
+ int active_src = 0;
+ int entry_count = 0;
+
+ priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
+
+ active_src = sh->active_source;
+
+ if (op_ret <= 0) {
+ if (op_ret < 0) {
+ gf_log (this->name, GF_LOG_INFO,
+ "readdir of %s on subvolume %s failed (%s)",
+ local->loc.path,
+ priv->children[active_src]->name,
+ strerror (op_errno));
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ } else {
+ gf_log (this->name, GF_LOG_TRACE,
+ "readdir of %s on subvolume %s complete",
+ local->loc.path,
+ priv->children[active_src]->name);
+ }
+
+ afr_sh_entry_impunge_all (frame, this);
+ return 0;
+ }
+
+ list_for_each_entry (entry, &entries->list, list) {
+ last_offset = entry->d_off;
entry_count++;
- }
+ }
- gf_log (this->name, GF_LOG_TRACE,
- "readdir'ed %d entries from %s",
- entry_count, priv->children[active_src]->name);
+ gf_log (this->name, GF_LOG_DEBUG,
+ "readdir'ed %d entries from %s",
+ entry_count, priv->children[active_src]->name);
- sh->offset = last_offset;
- local->call_count = entry_count;
+ sh->offset = last_offset;
+ local->call_count = entry_count;
- list_for_each_entry (entry, &entries->list, list) {
+ list_for_each_entry (entry, &entries->list, list) {
afr_sh_entry_impunge_entry (frame, this, entry);
- }
+ }
- return 0;
+ return 0;
}
-
+
int
-afr_sh_entry_impunge_subvol (call_frame_t *frame, xlator_t *this,
- int active_src)
+afr_sh_entry_impunge_subvol (call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ int32_t active_src = 0;
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
+ priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
+ active_src = sh->active_source;
+ gf_log (this->name, GF_LOG_DEBUG, "%s: readdir from offset %zd",
+ local->loc.path, sh->offset);
- STACK_WIND (frame, afr_sh_entry_impunge_readdir_cbk,
- priv->children[active_src],
- priv->children[active_src]->fops->readdirp,
- sh->healing_fd, sh->block_size, sh->offset);
+ STACK_WIND (frame, afr_sh_entry_impunge_readdir_cbk,
+ priv->children[active_src],
+ priv->children[active_src]->fops->readdirp,
+ sh->healing_fd, sh->block_size, sh->offset, NULL);
- return 0;
+ return 0;
}
int
afr_sh_entry_impunge_all (call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- int active_src = -1;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ int active_src = -1;
- priv = this->private;
- local = frame->local;
- sh = &local->self_heal;
+ priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
- sh->offset = 0;
+ sh->offset = 0;
- active_src = next_active_source (frame, this, sh->active_source);
- sh->active_source = active_src;
+ active_src = next_active_source (frame, this, sh->active_source);
+ sh->active_source = active_src;
- if (sh->op_failed) {
- afr_sh_entry_finish (frame, this);
- return 0;
- }
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+ afr_sh_entry_finish (frame, this);
+ return 0;
+ }
- if (active_src == -1) {
- /* completed creating missing files on all subvolumes */
- afr_sh_entry_expunge_all (frame, this);
- return 0;
- }
+ if (active_src == -1) {
+ /* completed creating missing files on all subvolumes */
+ afr_sh_entry_erase_pending (frame, this);
+ return 0;
+ }
- gf_log (this->name, GF_LOG_TRACE,
- "impunging entries of %s on %s to other sinks",
- local->loc.path, priv->children[active_src]->name);
+ gf_log (this->name, GF_LOG_TRACE,
+ "impunging entries of %s on %s to other sinks",
+ local->loc.path, priv->children[active_src]->name);
- afr_sh_entry_impunge_subvol (frame, this, active_src);
+ afr_sh_entry_impunge_subvol (frame, this);
- return 0;
+ return 0;
}
int
afr_sh_entry_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd)
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
- int child_index = 0;
-
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
-
- child_index = (long) cookie;
-
- /* TODO: some of the open's might fail.
- In that case, modify cleanup fn to send flush on those
- fd's which are already open */
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "opendir of %s failed on child %s (%s)",
- local->loc.path,
- priv->children[child_index]->name,
- strerror (op_errno));
- sh->op_failed = 1;
- }
- }
- UNLOCK (&frame->lock);
-
- call_count = afr_frame_return (frame);
-
- if (call_count == 0) {
- if (sh->op_failed) {
- afr_sh_entry_finish (frame, this);
- return 0;
- }
- gf_log (this->name, GF_LOG_TRACE,
- "fd for %s opened, commencing sync",
- local->loc.path);
-
- sh->active_source = -1;
- afr_sh_entry_impunge_all (frame, this);
- }
-
- return 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int child_index = 0;
+
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+
+ child_index = (long) cookie;
+
+ /* TODO: some of the open's might fail.
+ In that case, modify cleanup fn to send flush on those
+ fd's which are already open */
+
+ LOCK (&frame->lock);
+ {
+ if (op_ret == -1) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "opendir of %s failed on child %s (%s)",
+ local->loc.path,
+ priv->children[child_index]->name,
+ strerror (op_errno));
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ }
+ }
+ UNLOCK (&frame->lock);
+
+ call_count = afr_frame_return (frame);
+
+ if (call_count == 0) {
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+ afr_sh_entry_finish (frame, this);
+ return 0;
+ }
+ gf_log (this->name, GF_LOG_TRACE,
+ "fd for %s opened, commencing sync",
+ local->loc.path);
+
+ sh->active_source = -1;
+ afr_sh_entry_expunge_all (frame, this);
+ }
+
+ return 0;
}
int
afr_sh_entry_open (call_frame_t *frame, xlator_t *this)
{
- int i = 0;
- int call_count = 0;
+ int i = 0;
+ int call_count = 0;
- int source = -1;
- int *sources = NULL;
+ int source = -1;
+ int *sources = NULL;
- fd_t *fd = NULL;
+ fd_t *fd = NULL;
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
- afr_self_heal_t *sh = NULL;
-
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ afr_local_t * local = NULL;
+ afr_private_t * priv = NULL;
+ afr_self_heal_t *sh = NULL;
- source = local->self_heal.source;
- sources = local->self_heal.sources;
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- sh->block_size = 65536; //131072
- sh->offset = 0;
+ source = local->self_heal.source;
+ sources = local->self_heal.sources;
- call_count = sh->active_sinks;
- if (source != -1)
- call_count++;
+ sh->block_size = priv->sh_readdir_size;
+ sh->offset = 0;
- local->call_count = call_count;
+ call_count = sh->active_sinks;
+ if (source != -1)
+ call_count++;
- fd = fd_create (local->loc.inode, frame->root->pid);
- sh->healing_fd = fd;
+ local->call_count = call_count;
- if (source != -1) {
- gf_log (this->name, GF_LOG_TRACE,
- "opening directory %s on subvolume %s (source)",
- local->loc.path, priv->children[source]->name);
+ fd = fd_create (local->loc.inode, frame->root->pid);
+ sh->healing_fd = fd;
- /* open source */
- STACK_WIND_COOKIE (frame, afr_sh_entry_opendir_cbk,
- (void *) (long) source,
- priv->children[source],
- priv->children[source]->fops->opendir,
- &local->loc, fd);
- call_count--;
- }
+ if (source != -1) {
+ gf_log (this->name, GF_LOG_TRACE,
+ "opening directory %s on subvolume %s (source)",
+ local->loc.path, priv->children[source]->name);
+
+ /* open source */
+ STACK_WIND_COOKIE (frame, afr_sh_entry_opendir_cbk,
+ (void *) (long) source,
+ priv->children[source],
+ priv->children[source]->fops->opendir,
+ &local->loc, fd, NULL);
+ call_count--;
+ }
- /* open sinks */
- for (i = 0; i < priv->child_count; i++) {
- if (sources[i] || !local->child_up[i])
- continue;
+ /* open sinks */
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i] || !local->child_up[i])
+ continue;
- gf_log (this->name, GF_LOG_TRACE,
- "opening directory %s on subvolume %s (sink)",
- local->loc.path, priv->children[i]->name);
+ gf_log (this->name, GF_LOG_TRACE,
+ "opening directory %s on subvolume %s (sink)",
+ local->loc.path, priv->children[i]->name);
- STACK_WIND_COOKIE (frame, afr_sh_entry_opendir_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->opendir,
- &local->loc, fd);
+ STACK_WIND_COOKIE (frame, afr_sh_entry_opendir_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->opendir,
+ &local->loc, fd, NULL);
- if (!--call_count)
- break;
- }
+ if (!--call_count)
+ break;
+ }
- return 0;
+ return 0;
}
int
afr_sh_entry_sync_prepare (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int active_sinks = 0;
- int source = 0;
- int i = 0;
-
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
-
- source = sh->source;
-
- for (i = 0; i < priv->child_count; i++) {
- if (sh->sources[i] == 0 && local->child_up[i] == 1) {
- active_sinks++;
- sh->success[i] = 1;
- }
- }
- if (source != -1)
- sh->success[source] = 1;
-
- if (active_sinks == 0) {
- gf_log (this->name, GF_LOG_TRACE,
- "no active sinks for self-heal on dir %s",
- local->loc.path);
- afr_sh_entry_finish (frame, this);
- return 0;
- }
- if (source == -1 && active_sinks < 2) {
- gf_log (this->name, GF_LOG_TRACE,
- "cannot sync with 0 sources and 1 sink on dir %s",
- local->loc.path);
- afr_sh_entry_finish (frame, this);
- return 0;
- }
- sh->active_sinks = active_sinks;
-
- if (source != -1)
- gf_log (this->name, GF_LOG_DEBUG,
- "self-healing directory %s from subvolume %s to "
- "%d other",
- local->loc.path, priv->children[source]->name,
- active_sinks);
- else
- gf_log (this->name, GF_LOG_DEBUG,
- "no active sources for %s found. "
- "merging all entries as a conservative decision",
- local->loc.path);
-
- afr_sh_entry_open (frame, this);
-
- return 0;
-}
-
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int source = 0;
-int
-afr_sh_entry_fix (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int source = 0;
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- int nsources = 0;
+ source = sh->source;
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ afr_sh_mark_source_sinks (frame, this);
+ if (source != -1)
+ sh->success[source] = 1;
- if (sh->forced_merge) {
- sh->source = -1;
- goto heal;
+ if (sh->active_sinks == 0) {
+ gf_log (this->name, GF_LOG_TRACE,
+ "no active sinks for self-heal on dir %s",
+ local->loc.path);
+ afr_sh_entry_finish (frame, this);
+ return 0;
}
-
- afr_sh_build_pending_matrix (priv, sh->pending_matrix, sh->xattr,
- priv->child_count, AFR_ENTRY_TRANSACTION);
-
- afr_sh_print_pending_matrix (sh->pending_matrix, this);
-
- nsources = afr_sh_mark_sources (sh, priv->child_count,
- AFR_SELF_HEAL_ENTRY);
-
- if (nsources == 0) {
+ if (source == -1 && sh->active_sinks < 2) {
gf_log (this->name, GF_LOG_TRACE,
- "No self-heal needed for %s",
+ "cannot sync with 0 sources and 1 sink on dir %s",
local->loc.path);
-
afr_sh_entry_finish (frame, this);
return 0;
}
- afr_sh_supress_errenous_children (sh->sources, sh->child_errno,
- priv->child_count);
-
- source = afr_sh_select_source (sh->sources, priv->child_count);
-
- sh->source = source;
+ if (source != -1)
+ gf_log (this->name, GF_LOG_DEBUG,
+ "self-healing directory %s from subvolume %s to "
+ "%d other",
+ local->loc.path, priv->children[source]->name,
+ sh->active_sinks);
+ else
+ gf_log (this->name, GF_LOG_DEBUG,
+ "no active sources for %s found. "
+ "merging all entries as a conservative decision",
+ local->loc.path);
-heal:
- afr_sh_entry_sync_prepare (frame, this);
+ sh->actual_sh_started = _gf_true;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_SYNC_BEGIN);
+ afr_sh_entry_open (frame, this);
- return 0;
+ return 0;
}
-
-int
-afr_sh_entry_lookup_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf, dict_t *xattr,
- struct iatt *postparent)
+void
+afr_sh_entry_fix (call_frame_t *frame, xlator_t *this,
+ int32_t op_ret, int32_t op_errno)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
-
- int call_count = -1;
- int child_index = (long) cookie;
-
- local = frame->local;
- sh = &local->self_heal;
-
- LOCK (&frame->lock);
- {
- if (op_ret != -1) {
- sh->xattr[child_index] = dict_ref (xattr);
- sh->buf[child_index] = *buf;
- }
- }
- UNLOCK (&frame->lock);
-
- call_count = afr_frame_return (frame);
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int source = 0;
+ int nsources = 0;
+ int32_t subvol_status = 0;
- if (call_count == 0) {
- afr_sh_entry_fix (frame, this);
- }
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- return 0;
-}
+ if (op_ret < 0) {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_set_error (sh, op_errno);
+ afr_sh_entry_finish (frame, this);
+ goto out;
+ }
+ if (sh->forced_merge) {
+ sh->source = -1;
+ goto heal;
+ }
+ nsources = afr_build_sources (this, sh->xattr, sh->buf,
+ sh->pending_matrix, sh->sources,
+ sh->success_children,
+ AFR_ENTRY_TRANSACTION, &subvol_status,
+ _gf_true);
+ if ((subvol_status & ALL_FOOLS) ||
+ (subvol_status & SPLIT_BRAIN)) {
+ gf_log (this->name, GF_LOG_INFO, "%s: Performing conservative "
+ "merge", local->loc.path);
+ source = -1;
+ memset (sh->sources, 0,
+ sizeof (*sh->sources) * priv->child_count);
+ } else if (nsources == 0) {
+ gf_log (this->name, GF_LOG_TRACE,
+ "No self-heal needed for %s",
+ local->loc.path);
-int
-afr_sh_entry_lookup (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
- dict_t *xattr_req = NULL;
- int ret = 0;
- int call_count = 0;
- int i = 0;
-
- priv = this->private;
- local = frame->local;
-
- call_count = afr_up_children_count (priv->child_count,
- local->child_up);
-
- local->call_count = call_count;
-
- xattr_req = dict_new();
- if (xattr_req) {
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_set_uint64 (xattr_req,
- priv->pending_key[i],
- 3 * sizeof(int32_t));
- if (ret < 0)
- gf_log (this->name, GF_LOG_WARNING,
- "Unable to set dict value.");
- }
+ afr_sh_entry_finish (frame, this);
+ return;
+ } else {
+ source = afr_sh_select_source (sh->sources, priv->child_count);
}
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame,
- afr_sh_entry_lookup_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->lookup,
- &local->loc, xattr_req);
- if (!--call_count)
- break;
- }
- }
+ sh->source = source;
- if (xattr_req)
- dict_unref (xattr_req);
+ afr_reset_children (sh->fresh_children, priv->child_count);
+ afr_get_fresh_children (sh->success_children, sh->sources,
+ sh->fresh_children, priv->child_count);
+ if (sh->source >= 0)
+ afr_inode_set_read_ctx (this, sh->inode, sh->source,
+ sh->fresh_children);
- return 0;
+heal:
+ afr_sh_entry_sync_prepare (frame, this);
+out:
+ return;
}
int
@@ -2389,68 +2317,54 @@ afr_sh_post_nonblocking_entry_cbk (call_frame_t *frame, xlator_t *this)
{
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
local = frame->local;
int_lock = &local->internal_lock;
+ sh = &local->self_heal;
if (int_lock->lock_op_ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Non Blocking entrylks failed.");
+ gf_log (this->name, GF_LOG_ERROR, "Non Blocking entrylks "
+ "failed for %s.", local->loc.path);
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
afr_sh_entry_done (frame, this);
} else {
- gf_log (this->name, GF_LOG_DEBUG,
- "Non Blocking entrylks done. Proceeding to FOP");
- afr_sh_entry_lookup(frame, this);
+ gf_log (this->name, GF_LOG_DEBUG, "Non Blocking entrylks done "
+ "for %s. Proceeding to FOP", local->loc.path);
+ afr_sh_common_lookup (frame, this, &local->loc,
+ afr_sh_entry_fix, NULL,
+ AFR_LOOKUP_FAIL_CONFLICTS |
+ AFR_LOOKUP_FAIL_MISSING_GFIDS,
+ NULL);
}
return 0;
}
int
-afr_sh_entry_lock (call_frame_t *frame, xlator_t *this)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
-
- local = frame->local;
- int_lock = &local->internal_lock;
-
- int_lock->transaction_lk_type = AFR_SELFHEAL_LK;
- int_lock->selfheal_lk_type = AFR_ENTRY_SELF_HEAL_LK;
-
- afr_set_lock_number (frame, this);
-
- int_lock->lk_basename = NULL;
- int_lock->lk_loc = &local->loc;
- int_lock->lock_cbk = afr_sh_post_nonblocking_entry_cbk;
-
- afr_nonblocking_entrylk (frame, this);
-
-
- return 0;
-}
-
-
-int
afr_self_heal_entry (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_self_heal_t *sh = NULL;
+ priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
- priv = this->private;
- local = frame->local;
+ sh->sh_type_in_action = AFR_SELF_HEAL_ENTRY;
- if (local->self_heal.need_entry_self_heal && priv->entry_self_heal) {
- afr_sh_entry_lock (frame, this);
- } else {
- gf_log (this->name, GF_LOG_TRACE,
- "proceeding to completion on %s",
- local->loc.path);
- afr_sh_entry_done (frame, this);
- }
+ if (local->self_heal.do_entry_self_heal && priv->entry_self_heal) {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_STARTED);
+ afr_sh_entrylk (frame, this, &local->loc, NULL,
+ afr_sh_post_nonblocking_entry_cbk);
+ } else {
+ gf_log (this->name, GF_LOG_TRACE,
+ "proceeding to completion on %s",
+ local->loc.path);
+ afr_sh_entry_done (frame, this);
+ }
- return 0;
+ return 0;
}
-
diff --git a/xlators/cluster/afr/src/afr-self-heal-metadata.c b/xlators/cluster/afr/src/afr-self-heal-metadata.c
index bd70608a2..fd5da6cfd 100644
--- a/xlators/cluster/afr/src/afr-self-heal-metadata.c
+++ b/xlators/cluster/afr/src/afr-self-heal-metadata.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include <libgen.h>
@@ -52,73 +43,26 @@
int
afr_sh_metadata_done (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
-
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
-
-// memset (sh->child_errno, 0, sizeof (int) * priv->child_count);
- memset (sh->buf, 0, sizeof (struct iatt) * priv->child_count);
- memset (sh->success, 0, sizeof (int) * priv->child_count);
-
-/* for (i = 0; i < priv->child_count; i++) { */
-/* sh->locked_nodes[i] = 1; */
-/* } */
-
- for (i = 0; i < priv->child_count; i++) {
- if (sh->xattr[i])
- dict_unref (sh->xattr[i]);
- sh->xattr[i] = NULL;
- }
-
- if (local->govinda_gOvinda) {
- gf_log (this->name, GF_LOG_DEBUG,
- "aborting selfheal of %s",
- local->loc.path);
- sh->completion_cbk (frame, this);
- } else {
- if (IA_ISREG (sh->type)) {
- gf_log (this->name, GF_LOG_TRACE,
- "proceeding to data check on %s",
- local->loc.path);
- afr_self_heal_data (frame, this);
- return 0;
- }
-
- if (IA_ISDIR (sh->type)) {
- gf_log (this->name, GF_LOG_TRACE,
- "proceeding to entry check on %s",
- local->loc.path);
- afr_self_heal_entry (frame, this);
- return 0;
- }
- gf_log (this->name, GF_LOG_DEBUG,
- "completed self heal of %s",
- local->loc.path);
-
- sh->completion_cbk (frame, this);
- }
-
- return 0;
-}
-
-
-int
-afr_sh_metadata_unlck_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
-{
- int call_count = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
- call_count = afr_frame_return (frame);
+ local = frame->local;
+ sh = &local->self_heal;
- if (call_count == 0)
- afr_sh_metadata_done (frame, this);
+ afr_sh_reset (frame, this);
+ if (IA_ISDIR (sh->type)) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "proceeding to entry check on %s",
+ local->loc.path);
+ afr_self_heal_entry (frame, this);
+ } else {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "proceeding to data check on %s",
+ local->loc.path);
+ afr_self_heal_data (frame, this);
+ }
- return 0;
+ return 0;
}
int
@@ -141,530 +85,604 @@ afr_sh_metadata_finish (call_frame_t *frame, xlator_t *this)
{
afr_sh_inode_unlock (frame, this);
- return 0;
+ return 0;
}
-
int
-afr_sh_metadata_erase_pending_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xattr)
+afr_sh_metadata_fail (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- int call_count = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
- local = frame->local;
+ local = frame->local;
+ sh = &local->self_heal;
- LOCK (&frame->lock);
- {
- }
- UNLOCK (&frame->lock);
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_metadata_finish (frame, this);
+ return 0;
+}
- call_count = afr_frame_return (frame);
+int
+afr_sh_metadata_erase_pending_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xattr, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ int call_count = 0;
+ long i = 0;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = this->private;
+ sh = &local->self_heal;
+ i = (long)cookie;
+
+ if ((!IA_ISREG (sh->buf[sh->source].ia_type)) &&
+ (!IA_ISDIR (sh->buf[sh->source].ia_type))) {
+ afr_children_add_child (sh->fresh_children, i,
+ priv->child_count);
+ }
+ call_count = afr_frame_return (frame);
- if (call_count == 0)
- afr_sh_metadata_finish (frame, this);
+ if (call_count == 0) {
+ if ((!IA_ISREG (sh->buf[sh->source].ia_type)) &&
+ (!IA_ISDIR (sh->buf[sh->source].ia_type))) {
+ afr_inode_set_read_ctx (this, sh->inode, sh->source,
+ sh->fresh_children);
+ }
+ afr_sh_metadata_finish (frame, this);
+ }
- return 0;
+ return 0;
}
-
int
afr_sh_metadata_erase_pending (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
- int i = 0;
- dict_t **erase_xattr = NULL;
-
-
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
-
- afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix,
- sh->success, priv->child_count,
- AFR_METADATA_TRANSACTION);
-
- erase_xattr = GF_CALLOC (sizeof (*erase_xattr), priv->child_count,
- gf_afr_mt_dict_t);
-
- for (i = 0; i < priv->child_count; i++) {
- if (sh->xattr[i]) {
- call_count++;
-
- erase_xattr[i] = get_new_dict();
- dict_ref (erase_xattr[i]);
- }
- }
-
- afr_sh_delta_to_xattr (priv, sh->delta_matrix, erase_xattr,
- priv->child_count, AFR_METADATA_TRANSACTION);
-
- local->call_count = call_count;
-
- if (call_count == 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "metadata of %s not healed on any subvolume",
- local->loc.path);
-
- afr_sh_metadata_finish (frame, this);
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (!erase_xattr[i])
- continue;
-
- gf_log (this->name, GF_LOG_TRACE,
- "erasing pending flags from %s on %s",
- local->loc.path, priv->children[i]->name);
-
- STACK_WIND_COOKIE (frame, afr_sh_metadata_erase_pending_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->xattrop,
- &local->loc,
- GF_XATTROP_ADD_ARRAY, erase_xattr[i]);
- if (!--call_count)
- break;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (erase_xattr[i]) {
- dict_unref (erase_xattr[i]);
- }
- }
- GF_FREE (erase_xattr);
-
- return 0;
+ afr_sh_erase_pending (frame, this, AFR_METADATA_TRANSACTION,
+ afr_sh_metadata_erase_pending_cbk,
+ afr_sh_metadata_finish);
+ return 0;
}
int
afr_sh_metadata_sync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
- int child_index = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int child_index = 0;
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- child_index = (long) cookie;
+ child_index = (long) cookie;
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "setting attributes failed for %s on %s (%s)",
- local->loc.path,
- priv->children[child_index]->name,
- strerror (op_errno));
+ LOCK (&frame->lock);
+ {
+ if (op_ret == -1) {
+ gf_log (this->name, GF_LOG_INFO,
+ "setting attributes failed for %s on %s (%s)",
+ local->loc.path,
+ priv->children[child_index]->name,
+ strerror (op_errno));
- sh->success[child_index] = 0;
- }
- }
- UNLOCK (&frame->lock);
+ sh->success[child_index] = 0;
+ }
+ }
+ UNLOCK (&frame->lock);
- call_count = afr_frame_return (frame);
+ call_count = afr_frame_return (frame);
- if (call_count == 0)
- afr_sh_metadata_erase_pending (frame, this);
+ if (call_count == 0) {
+ if (local->xattr_req) {
+ dict_unref (local->xattr_req);
+ local->xattr_req = NULL;
+ }
+ afr_sh_metadata_erase_pending (frame, this);
+ }
- return 0;
+ return 0;
}
int
afr_sh_metadata_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop)
+ struct iatt *preop, struct iatt *postop, dict_t *xdata)
{
- afr_sh_metadata_sync_cbk (frame, cookie, this, op_ret, op_errno);
+ afr_sh_metadata_sync_cbk (frame, cookie, this, op_ret, op_errno, xdata);
- return 0;
+ return 0;
}
int
afr_sh_metadata_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_sh_metadata_sync_cbk (frame, cookie, this, op_ret, op_errno);
+ afr_sh_metadata_sync_cbk (frame, cookie, this, op_ret, op_errno, xdata);
- return 0;
+ return 0;
}
+int
+afr_sh_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ dict_t *xdata)
+{
+ int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ if (op_ret < 0) {
+ afr_sh_metadata_sync_cbk (frame, cookie,
+ this, -1, op_errno, xdata);
+ goto out;
+ }
+
+ i = (long) cookie;
+
+ STACK_WIND_COOKIE (frame, afr_sh_metadata_xattr_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->setxattr,
+ &local->loc, local->xattr_req, 0, NULL);
+
+ out:
+ return 0;
+}
+
+inline void
+afr_prune_special_keys (dict_t *xattr_dict)
+{
+ dict_del (xattr_dict, GF_SELINUX_XATTR_KEY);
+}
+
+inline void
+afr_prune_pending_keys (dict_t *xattr_dict, afr_private_t *priv)
+{
+ int i = 0;
+
+ for (; i < priv->child_count; i++) {
+ dict_del (xattr_dict, priv->pending_key[i]);
+ }
+}
+
+int
+afr_sh_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
+{
+ int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ if (op_ret < 0) {
+ afr_sh_metadata_sync_cbk (frame, cookie,
+ this, -1, op_errno, xdata);
+ goto out;
+ }
+
+ afr_prune_pending_keys (xattr, priv);
+
+ afr_prune_special_keys (xattr);
+
+ i = (long) cookie;
+
+ /* send removexattr in bulk via xdata */
+ STACK_WIND_COOKIE (frame, afr_sh_removexattr_cbk,
+ cookie,
+ priv->children[i],
+ priv->children[i]->fops->removexattr,
+ &local->loc, "", xattr);
+
+ out:
+ return 0;
+}
int
afr_sh_metadata_sync (call_frame_t *frame, xlator_t *this, dict_t *xattr)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int source = 0;
- int active_sinks = 0;
- int call_count = 0;
- int i = 0;
-
- struct iatt stbuf;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int source = 0;
+ int active_sinks = 0;
+ int call_count = 0;
+ int i = 0;
+
+ struct iatt stbuf = {0,};
int32_t valid = 0;
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- source = sh->source;
- active_sinks = sh->active_sinks;
+ source = sh->source;
+ active_sinks = sh->active_sinks;
- /*
- * 2 calls per sink - setattr, setxattr
- */
- if (xattr)
- call_count = active_sinks * 2;
- else
- call_count = active_sinks;
+ /*
+ * 2 calls per sink - setattr, setxattr
+ */
+ if (xattr) {
+ call_count = active_sinks * 2;
+ local->xattr_req = dict_ref (xattr);
+ } else
+ call_count = active_sinks;
- local->call_count = call_count;
+ local->call_count = call_count;
- stbuf.ia_atime = sh->buf[source].ia_atime;
- stbuf.ia_atime_nsec = sh->buf[source].ia_atime_nsec;
- stbuf.ia_mtime = sh->buf[source].ia_mtime;
- stbuf.ia_mtime_nsec = sh->buf[source].ia_mtime_nsec;
+ stbuf.ia_atime = sh->buf[source].ia_atime;
+ stbuf.ia_atime_nsec = sh->buf[source].ia_atime_nsec;
+ stbuf.ia_mtime = sh->buf[source].ia_mtime;
+ stbuf.ia_mtime_nsec = sh->buf[source].ia_mtime_nsec;
stbuf.ia_uid = sh->buf[source].ia_uid;
stbuf.ia_gid = sh->buf[source].ia_gid;
stbuf.ia_type = sh->buf[source].ia_type;
stbuf.ia_prot = sh->buf[source].ia_prot;
-
- valid = GF_SET_ATTR_MODE |
+
+ valid = GF_SET_ATTR_MODE |
GF_SET_ATTR_UID | GF_SET_ATTR_GID |
GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
- for (i = 0; i < priv->child_count; i++) {
- if (call_count == 0) {
- break;
- }
- if (sh->sources[i] || !local->child_up[i])
- continue;
-
- gf_log (this->name, GF_LOG_DEBUG,
- "self-healing metadata of %s from %s to %s",
- local->loc.path, priv->children[source]->name,
- priv->children[i]->name);
-
- STACK_WIND_COOKIE (frame, afr_sh_metadata_setattr_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->setattr,
- &local->loc, &stbuf, valid);
-
- call_count--;
-
- if (!xattr)
- continue;
-
- STACK_WIND_COOKIE (frame, afr_sh_metadata_xattr_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->setxattr,
- &local->loc, xattr, 0);
- call_count--;
- }
-
- return 0;
+ for (i = 0; i < priv->child_count; i++) {
+ if (call_count == 0) {
+ break;
+ }
+ if (sh->sources[i] || !local->child_up[i])
+ continue;
+
+ gf_log (this->name, GF_LOG_DEBUG,
+ "self-healing metadata of %s from %s to %s",
+ local->loc.path, priv->children[source]->name,
+ priv->children[i]->name);
+
+ STACK_WIND_COOKIE (frame, afr_sh_metadata_setattr_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->setattr,
+ &local->loc, &stbuf, valid, NULL);
+
+ call_count--;
+
+ if (!xattr)
+ continue;
+
+ STACK_WIND_COOKIE (frame, afr_sh_getxattr_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->getxattr,
+ &local->loc, NULL, NULL);
+ call_count--;
+ }
+
+ return 0;
}
int
-afr_sh_metadata_getxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xattr)
+afr_sh_metadata_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int source = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int source = 0;
+
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
- int i;
+ source = sh->source;
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ if (op_ret == -1) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "getxattr of %s failed on subvolume %s (%s). proceeding without xattr",
+ local->loc.path, priv->children[source]->name,
+ strerror (op_errno));
+
+ afr_sh_metadata_sync (frame, this, NULL);
+ } else {
+ afr_prune_pending_keys (xattr, priv);
+ afr_sh_metadata_sync (frame, this, xattr);
+ }
- source = sh->source;
+ return 0;
+}
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "getxattr of %s failed on subvolume %s (%s). proceeding without xattr",
- local->loc.path, priv->children[source]->name,
- strerror (op_errno));
+static void
+afr_set_metadata_sh_info_str (afr_local_t *local, afr_self_heal_t *sh,
+ xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ int i = 0;
+ char num[1024] = {0};
+ size_t len = 0;
+ char *string = NULL;
+ size_t off = 0;
+ char *source_child = " from source %s to";
+ char *format = " %s, ";
+ char *string_msg = " metadata self heal";
+ char *pending_matrix_str = NULL;
+ int down_child_present = 0;
+ int unknown_child_present = 0;
+ char *down_subvol_1 = " down subvolume is ";
+ char *unknown_subvol_1 = " unknown subvolume is";
+ char *down_subvol_2 = " down subvolumes are ";
+ char *unknown_subvol_2 = " unknown subvolumes are ";
+ int down_count = 0;
+ int unknown_count = 0;
+
+ priv = this->private;
+
+ pending_matrix_str = afr_get_pending_matrix_str (sh->pending_matrix,
+ this);
+
+ if (!pending_matrix_str)
+ pending_matrix_str = "";
+
+ len += snprintf (num, sizeof (num), "%s", string_msg);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if ((sh->source == i) && (local->child_up[i] == 1)) {
+ len += snprintf (num, sizeof (num), source_child,
+ priv->children[i]->name);
+ } else if ((local->child_up[i] == 1) && (sh->sources[i] == 0)) {
+ len += snprintf (num, sizeof (num), format,
+ priv->children[i]->name);
+ } else if (local->child_up[i] == 0) {
+ len += snprintf (num, sizeof (num), format,
+ priv->children[i]->name);
+ if (!down_child_present)
+ down_child_present = 1;
+ down_count++;
+ } else if (local->child_up[i] == -1) {
+ len += snprintf (num, sizeof (num), format,
+ priv->children[i]->name);
+ if (!unknown_child_present)
+ unknown_child_present = 1;
+ unknown_count++;
+ }
+ }
- afr_sh_metadata_sync (frame, this, NULL);
- } else {
- for (i = 0; i < priv->child_count; i++) {
- dict_del (xattr, priv->pending_key[i]);
+ if (down_child_present) {
+ if (down_count > 1) {
+ len += snprintf (num, sizeof (num), "%s",
+ down_subvol_2);
+ } else {
+ len += snprintf (num, sizeof (num), "%s",
+ down_subvol_1);
}
+ }
+ if (unknown_child_present) {
+ if (unknown_count > 1) {
+ len += snprintf (num, sizeof (num), "%s",
+ unknown_subvol_2);
+ } else {
+ len += snprintf (num, sizeof (num), "%s",
+ unknown_subvol_1);
+ }
+ }
- afr_sh_metadata_sync (frame, this, xattr);
- }
+ len ++;
- return 0;
-}
+ string = GF_CALLOC (len, sizeof (char), gf_common_mt_char);
+ if (!string)
+ return;
+ off += snprintf (string + off, len - off, "%s", string_msg);
+ for (i=0; i < priv->child_count; i++) {
+ if ((sh->source == i) && (local->child_up[i] == 1))
+ off += snprintf (string + off, len - off, source_child,
+ priv->children[i]->name);
+ }
-int
-afr_sh_metadata_sync_prepare (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int active_sinks = 0;
- int source = 0;
- int i = 0;
-
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
-
- source = sh->source;
-
- for (i = 0; i < priv->child_count; i++) {
- if (sh->sources[i] == 0 && local->child_up[i] == 1) {
- active_sinks++;
- sh->success[i] = 1;
- }
- }
- sh->success[source] = 1;
-
- if (active_sinks == 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "no active sinks for performing self-heal on file %s",
- local->loc.path);
- afr_sh_metadata_finish (frame, this);
- return 0;
- }
- sh->active_sinks = active_sinks;
-
- gf_log (this->name, GF_LOG_TRACE,
- "syncing metadata of %s from subvolume %s to %d active sinks",
- local->loc.path, priv->children[source]->name, active_sinks);
-
- STACK_WIND (frame, afr_sh_metadata_getxattr_cbk,
- priv->children[source],
- priv->children[source]->fops->getxattr,
- &local->loc, NULL);
-
- return 0;
-}
+ for (i = 0; i < priv->child_count; i++) {
+ if ((local->child_up[i] == 1)&& (sh->sources[i] == 0))
+ off += snprintf (string + off, len - off, format,
+ priv->children[i]->name);
+ }
+ if (down_child_present) {
+ if (down_count > 1) {
+ off += snprintf (string + off, len - off, "%s",
+ down_subvol_2);
+ } else {
+ off += snprintf (string + off, len - off, "%s",
+ down_subvol_1);
+ }
+ }
-int
-afr_sh_metadata_fix (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int nsources = 0;
- int source = 0;
- int i = 0;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i] == 0)
+ off += snprintf (string + off, len - off, format,
+ priv->children[i]->name);
+ }
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
+ if (unknown_child_present) {
+ if (unknown_count > 1) {
+ off += snprintf (string + off, len - off, "%s",
+ unknown_subvol_2);
+ } else {
+ off += snprintf (string + off, len - off, "%s",
+ unknown_subvol_1);
+ }
+ }
- afr_sh_build_pending_matrix (priv, sh->pending_matrix, sh->xattr,
- priv->child_count,
- AFR_METADATA_TRANSACTION);
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i] == -1)
+ off += snprintf (string + off, len - off, format,
+ priv->children[i]->name);
+ }
- afr_sh_print_pending_matrix (sh->pending_matrix, this);
+ gf_asprintf (&sh->metadata_sh_info, "%s metadata %s,", string,
+ pending_matrix_str);
- nsources = afr_sh_mark_sources (sh, priv->child_count,
- AFR_SELF_HEAL_METADATA);
+ if (pending_matrix_str && strcmp (pending_matrix_str, ""))
+ GF_FREE (pending_matrix_str);
- afr_sh_supress_errenous_children (sh->sources, sh->child_errno,
- priv->child_count);
+ if (string && strcmp (string, ""))
+ GF_FREE (string);
+}
- if (nsources == 0) {
- gf_log (this->name, GF_LOG_TRACE,
- "No self-heal needed for %s",
- local->loc.path);
+int
+afr_sh_metadata_sync_prepare (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int source = 0;
+
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+
+ source = sh->source;
+ afr_sh_mark_source_sinks (frame, this);
+ if (sh->active_sinks == 0) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "no active sinks for performing self-heal on file %s",
+ local->loc.path);
afr_sh_metadata_finish (frame, this);
return 0;
}
- if ((nsources == -1)
- && (priv->favorite_child != -1)
- && (sh->child_errno[priv->favorite_child] == 0)) {
+ gf_log (this->name, GF_LOG_TRACE,
+ "syncing metadata of %s from subvolume %s to %d active sinks",
+ local->loc.path, priv->children[source]->name,
+ sh->active_sinks);
- gf_log (this->name, GF_LOG_WARNING,
- "Picking favorite child %s as authentic source to resolve conflicting metadata of %s",
- priv->children[priv->favorite_child]->name,
- local->loc.path);
+ sh->actual_sh_started = _gf_true;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_SYNC_BEGIN);
+ afr_set_metadata_sh_info_str (local, sh, this);
+ STACK_WIND (frame, afr_sh_metadata_getxattr_cbk,
+ priv->children[source],
+ priv->children[source]->fops->getxattr,
+ &local->loc, NULL, NULL);
- sh->sources[priv->favorite_child] = 1;
+ return 0;
+}
- nsources = afr_sh_source_count (sh->sources,
- priv->child_count);
- }
- if (nsources == -1) {
- gf_log (this->name, GF_LOG_ERROR,
- "Unable to self-heal permissions/ownership of '%s' "
- "(possible split-brain). Please fix the file on "
- "all backend volumes", local->loc.path);
+void
+afr_sh_metadata_fix (call_frame_t *frame, xlator_t *this,
+ int32_t op_ret, int32_t op_errno)
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int nsources = 0;
+ int source = 0;
+ int i = 0;
+
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+
+ if (op_ret < 0) {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_set_error (sh, op_errno);
+ afr_sh_metadata_finish (frame, this);
+ goto out;
+ }
+ nsources = afr_build_sources (this, sh->xattr, sh->buf,
+ sh->pending_matrix, sh->sources,
+ sh->success_children,
+ AFR_METADATA_TRANSACTION, NULL, _gf_false);
+ if ((nsources == -1)
+ && (priv->favorite_child != -1)
+ && (sh->child_errno[priv->favorite_child] == 0)) {
+
+ gf_log (this->name, GF_LOG_WARNING,
+ "Picking favorite child %s as authentic source to resolve conflicting metadata of %s",
+ priv->children[priv->favorite_child]->name,
+ local->loc.path);
- local->govinda_gOvinda = 1;
+ sh->sources[priv->favorite_child] = 1;
- afr_sh_metadata_finish (frame, this);
- return 0;
- }
+ nsources = afr_sh_source_count (sh->sources,
+ priv->child_count);
+ }
- source = afr_sh_select_source (sh->sources, priv->child_count);
+ if (nsources == -1) {
+ afr_sh_print_split_brain_log (sh->pending_matrix, this,
+ local->loc.path);
+ afr_set_split_brain (this, sh->inode, SPB, DONT_KNOW);
+ afr_sh_metadata_fail (frame, this);
+ goto out;
+ }
- if (source == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "No active sources found.");
+ afr_set_split_brain (this, sh->inode, NO_SPB, DONT_KNOW);
+ if (nsources == 0) {
+ gf_log (this->name, GF_LOG_TRACE,
+ "No self-heal needed for %s",
+ local->loc.path);
afr_sh_metadata_finish (frame, this);
- return 0;
+ goto out;
}
- sh->source = source;
-
- /* detect changes not visible through pending flags -- JIC */
- for (i = 0; i < priv->child_count; i++) {
- if (i == source || sh->child_errno[i])
- continue;
+ source = afr_sh_select_source (sh->sources, priv->child_count);
- if (PERMISSION_DIFFERS (&sh->buf[i], &sh->buf[source]))
- sh->sources[i] = 0;
+ if (source == -1) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "No active sources found.");
- if (OWNERSHIP_DIFFERS (&sh->buf[i], &sh->buf[source]))
- sh->sources[i] = 0;
- }
+ afr_sh_metadata_finish (frame, this);
+ goto out;
+ }
- afr_sh_metadata_sync_prepare (frame, this);
+ sh->source = source;
- return 0;
-}
+ /* detect changes not visible through pending flags -- JIC */
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == source || sh->child_errno[i])
+ continue;
+ if (PERMISSION_DIFFERS (&sh->buf[i], &sh->buf[source]))
+ sh->sources[i] = 0;
-int
-afr_sh_metadata_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf, dict_t *xattr,
- struct iatt *postparent)
-{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
- int child_index = 0;
-
-
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
-
- child_index = (long) cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == 0) {
- gf_log (this->name, GF_LOG_TRACE,
- "path %s on subvolume %s is of mode 0%o",
- local->loc.path,
- priv->children[child_index]->name,
- buf->ia_type);
-
- sh->buf[child_index] = *buf;
- if (xattr)
- sh->xattr[child_index] = dict_ref (xattr);
- } else {
- gf_log (this->name, GF_LOG_DEBUG,
- "path %s on subvolume %s => -1 (%s)",
- local->loc.path,
- priv->children[child_index]->name,
- strerror (op_errno));
-
- sh->child_errno[child_index] = op_errno;
- }
- }
- UNLOCK (&frame->lock);
-
- call_count = afr_frame_return (frame);
-
- if (call_count == 0)
- afr_sh_metadata_fix (frame, this);
-
- return 0;
-}
-
+ if (OWNERSHIP_DIFFERS (&sh->buf[i], &sh->buf[source]))
+ sh->sources[i] = 0;
+ }
-int
-afr_sh_metadata_lookup (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- int call_count = 0;
- dict_t *xattr_req = NULL;
- int ret = 0;
-
- local = frame->local;
- priv = this->private;
-
- call_count = afr_up_children_count (priv->child_count,
- local->child_up);
- local->call_count = call_count;
-
- xattr_req = dict_new();
-
- if (xattr_req) {
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_set_uint64 (xattr_req,
- priv->pending_key[i],
- 3 * sizeof(int32_t));
- if (ret < 0)
- gf_log (this->name, GF_LOG_WARNING,
- "Unable to set dict value.");
- }
+ if ((!IA_ISREG (sh->buf[source].ia_type)) &&
+ (!IA_ISDIR (sh->buf[source].ia_type))) {
+ afr_reset_children (sh->fresh_children, priv->child_count);
+ afr_get_fresh_children (sh->success_children, sh->sources,
+ sh->fresh_children, priv->child_count);
+ afr_inode_set_read_ctx (this, sh->inode, sh->source,
+ sh->fresh_children);
}
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- gf_log (this->name, GF_LOG_TRACE,
- "looking up %s on %s",
- local->loc.path, priv->children[i]->name);
-
- STACK_WIND_COOKIE (frame, afr_sh_metadata_lookup_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->lookup,
- &local->loc, xattr_req);
- if (!--call_count)
- break;
- }
- }
-
- if (xattr_req)
- dict_unref (xattr_req);
-
- return 0;
+ if (sh->do_metadata_self_heal && priv->metadata_self_heal)
+ afr_sh_metadata_sync_prepare (frame, this);
+ else
+ afr_sh_metadata_finish (frame, this);
+out:
+ return;
}
int
-afr_sh_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
+afr_sh_metadata_post_nonblocking_inodelk_cbk (call_frame_t *frame,
+ xlator_t *this)
{
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
@@ -673,14 +691,21 @@ afr_sh_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
int_lock = &local->internal_lock;
if (int_lock->lock_op_ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Non Blocking inodelks failed.");
+ gf_log (this->name, GF_LOG_DEBUG, "Non Blocking metadata "
+ "inodelks failed for %s.", local->loc.path);
+ gf_log (this->name, GF_LOG_DEBUG, "Metadata self-heal "
+ "failed for %s.", local->loc.path);
afr_sh_metadata_done (frame, this);
} else {
- gf_log (this->name, GF_LOG_DEBUG,
- "Non Blocking inodelks done. Proceeding to FOP");
- afr_sh_metadata_lookup (frame, this);
+ gf_log (this->name, GF_LOG_DEBUG, "Non Blocking metadata "
+ "inodelks done for %s. Proceeding to FOP",
+ local->loc.path);
+ afr_sh_common_lookup (frame, this, &local->loc,
+ afr_sh_metadata_fix, NULL,
+ AFR_LOOKUP_FAIL_CONFLICTS |
+ AFR_LOOKUP_FAIL_MISSING_GFIDS,
+ NULL);
}
return 0;
@@ -690,42 +715,56 @@ int
afr_sh_metadata_lock (call_frame_t *frame, xlator_t *this)
{
afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
afr_local_t *local = NULL;
local = frame->local;
int_lock = &local->internal_lock;
+ int_lock->domain = this->name;
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
int_lock->transaction_lk_type = AFR_SELFHEAL_LK;
int_lock->selfheal_lk_type = AFR_METADATA_SELF_HEAL_LK;
afr_set_lock_number (frame, this);
- int_lock->lk_flock.l_start = 0;
- int_lock->lk_flock.l_len = 0;
- int_lock->lk_flock.l_type = F_WRLCK;
- int_lock->lock_cbk = afr_sh_post_nonblocking_inodelk_cbk;
+ inodelk->flock.l_start = LLONG_MAX - 1;
+ inodelk->flock.l_len = 0;
+ inodelk->flock.l_type = F_WRLCK;
+ int_lock->lock_cbk = afr_sh_metadata_post_nonblocking_inodelk_cbk;
afr_nonblocking_inodelk (frame, this);
- return 0;
+ return 0;
}
+gf_boolean_t
+afr_can_start_metadata_self_heal (afr_self_heal_t *sh, afr_private_t *priv)
+{
+ if (sh->force_confirm_spb)
+ return _gf_true;
+ if (sh->do_metadata_self_heal && priv->metadata_self_heal)
+ return _gf_true;
+ return _gf_false;
+}
int
afr_self_heal_metadata (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = this->private;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = this->private;
+ afr_self_heal_t *sh = &local->self_heal;
+ local = frame->local;
+ sh = &local->self_heal;
+ sh->sh_type_in_action = AFR_SELF_HEAL_METADATA;
- local = frame->local;
-
- if (local->self_heal.need_metadata_self_heal && priv->metadata_self_heal) {
- afr_sh_metadata_lock (frame, this);
- } else {
- afr_sh_metadata_done (frame, this);
- }
+ if (afr_can_start_metadata_self_heal (sh, priv)) {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_STARTED);
+ afr_sh_metadata_lock (frame, this);
+ } else {
+ afr_sh_metadata_done (frame, this);
+ }
- return 0;
+ return 0;
}
-
diff --git a/xlators/cluster/afr/src/afr-self-heal.h b/xlators/cluster/afr/src/afr-self-heal.h
index c43473a43..7c9bc8111 100644
--- a/xlators/cluster/afr/src/afr-self-heal.h
+++ b/xlators/cluster/afr/src/afr-self-heal.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __AFR_SELF_HEAL_H__
@@ -30,13 +21,6 @@
#define SIZE_GREATER(buf1,buf2) ((buf1)->ia_size > (buf2)->ia_size)
int
-afr_sh_has_metadata_pending (dict_t *xattr, int child_count, xlator_t *this);
-int
-afr_sh_has_entry_pending (dict_t *xattr, int child_count, xlator_t *this);
-int
-afr_sh_has_data_pending (dict_t *xattr, int child_count, xlator_t *this);
-
-int
afr_self_heal_entry (call_frame_t *frame, xlator_t *this);
int
@@ -49,6 +33,11 @@ int
afr_self_heal_get_source (xlator_t *this, afr_local_t *local, dict_t **xattr);
int
-afr_self_heal (call_frame_t *frame, xlator_t *this);
+afr_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode);
+int
+afr_lookup_select_read_child_by_txn_type (xlator_t *this, afr_local_t *local,
+ dict_t **xattr,
+ afr_transaction_type txn_type,
+ uuid_t gfid);
#endif /* __AFR_SELF_HEAL_H__ */
diff --git a/xlators/cluster/afr/src/afr-self-heald.c b/xlators/cluster/afr/src/afr-self-heald.c
new file mode 100644
index 000000000..1b48a1bca
--- /dev/null
+++ b/xlators/cluster/afr/src/afr-self-heald.c
@@ -0,0 +1,1787 @@
+/*
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+#include "afr.h"
+#include "syncop.h"
+#include "afr-self-heald.h"
+#include "afr-self-heal-common.h"
+#include "protocol-common.h"
+#include "event-history.h"
+
+typedef enum {
+ STOP_CRAWL_ON_SINGLE_SUBVOL = 1
+} afr_crawl_flags_t;
+
+typedef enum {
+ HEAL = 1,
+ INFO,
+ STATISTICS_TO_BE_HEALED,
+} shd_crawl_op;
+
+typedef struct shd_dump {
+ dict_t *dict;
+ xlator_t *this;
+ int child;
+} shd_dump_t;
+
+typedef struct shd_event_ {
+ int child;
+ char *path;
+} shd_event_t;
+
+typedef struct shd_pos_ {
+ int child;
+ xlator_t *this;
+ afr_child_pos_t pos;
+} shd_pos_t;
+
+typedef int
+(*afr_crawl_done_cbk_t) (int ret, call_frame_t *sync_frame, void *crawl_data);
+
+void
+afr_start_crawl (xlator_t *this, int idx, afr_crawl_type_t crawl,
+ process_entry_cbk_t process_entry, void *op_data,
+ gf_boolean_t exclusive, int crawl_flags,
+ afr_crawl_done_cbk_t crawl_done);
+
+static int
+_crawl_directory (fd_t *fd, loc_t *loc, afr_crawl_data_t *crawl_data);
+
+/* For calling straight through (e.g. already in a synctask). */
+int
+afr_find_child_position (xlator_t *this, int child, afr_child_pos_t *pos);
+
+/* For deferring through a new synctask. */
+int
+afr_syncop_find_child_position (void *data);
+
+static int
+_loc_assign_gfid_path (loc_t *loc)
+{
+ int ret = -1;
+ char gfid_path[64] = {0};
+
+ if (loc->inode && !uuid_is_null (loc->inode->gfid)) {
+ ret = inode_path (loc->inode, NULL, (char**)&loc->path);
+ } else if (!uuid_is_null (loc->gfid)) {
+ snprintf (gfid_path, sizeof (gfid_path), "<gfid:%s>",
+ uuid_utoa (loc->gfid));
+ loc->path = gf_strdup (gfid_path);
+ if (loc->path)
+ ret = 0;
+ }
+ return ret;
+}
+
+void
+_destroy_crawl_event_data (void *data)
+{
+ shd_crawl_event_t *crawl_event = NULL;
+
+ if (!data)
+ goto out;
+
+ crawl_event = (shd_crawl_event_t *)data;
+ GF_FREE (crawl_event->start_time_str);
+ GF_FREE (crawl_event->end_time_str);
+
+out:
+ return;
+}
+
+void
+_destroy_shd_event_data (void *data)
+{
+ shd_event_t *event = NULL;
+ if (!data)
+ goto out;
+ event = (shd_event_t*)data;
+ GF_FREE (event->path);
+out:
+ return;
+}
+void
+shd_cleanup_event (void *event)
+{
+ shd_event_t *shd_event = event;
+
+ if (!shd_event)
+ goto out;
+ GF_FREE (shd_event->path);
+ GF_FREE (shd_event);
+out:
+ return;
+}
+
+int
+afr_get_local_child (afr_self_heald_t *shd, unsigned int child_count)
+{
+ int i = 0;
+ int ret = -1;
+ for (i = 0; i < child_count; i++) {
+ if (shd->pos[i] == AFR_POS_LOCAL) {
+ ret = i;
+ break;
+ }
+ }
+ return ret;
+}
+
+static int
+_build_index_loc (xlator_t *this, loc_t *loc, char *name, loc_t *parent)
+{
+ int ret = 0;
+
+ uuid_copy (loc->pargfid, parent->inode->gfid);
+ loc->path = "";
+ loc->name = name;
+ loc->parent = inode_ref (parent->inode);
+ if (!loc->parent) {
+ loc->path = NULL;
+ loc_wipe (loc);
+ ret = -1;
+ }
+ return ret;
+}
+
+int
+_add_crawl_stats_to_dict (xlator_t *this, dict_t *output, int child,
+ shd_crawl_event_t *shd_event, struct timeval *tv)
+{
+ int ret = 0;
+ uint64_t count = 0;
+ char key[256] = {0};
+ int xl_id = 0;
+ uint64_t healed_count = 0;
+ uint64_t split_brain_count = 0;
+ uint64_t heal_failed_count = 0;
+ char *start_time_str = NULL;
+ char *end_time_str = NULL;
+ char *crawl_type = NULL;
+ int progress = -1;
+
+ healed_count = shd_event->healed_count;
+ split_brain_count = shd_event->split_brain_count;
+ heal_failed_count = shd_event->heal_failed_count;
+ start_time_str = shd_event->start_time_str;
+ end_time_str = shd_event->end_time_str;
+ crawl_type = shd_event->crawl_type;
+
+ if (!start_time_str) {
+ ret = -1;
+ goto out;
+ }
+
+
+ ret = dict_get_int32 (output, this->name, &xl_id);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "xl does not have id");
+ goto out;
+ }
+
+ snprintf (key, sizeof (key), "statistics-%d-%d-count", xl_id, child);
+ ret = dict_get_uint64 (output, key, &count);
+
+ snprintf (key, sizeof (key), "statistics_healed_cnt-%d-%d-%"PRIu64,
+ xl_id, child, count);
+ ret = dict_set_uint64(output, key, healed_count);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+ "healed_count to outout");
+ goto out;
+ }
+ snprintf (key, sizeof (key), "statistics_sb_cnt-%d-%d-%"PRIu64,
+ xl_id, child, count);
+ ret = dict_set_uint64 (output, key, split_brain_count);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+ "split_brain_count to outout");
+ goto out;
+ }
+ snprintf (key, sizeof (key), "statistics_crawl_type-%d-%d-%"PRIu64,
+ xl_id, child, count);
+ ret = dict_set_dynstr (output, key, gf_strdup (crawl_type));
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+ "crawl_type to output");
+ goto out;
+ }
+ snprintf (key, sizeof (key), "statistics_heal_failed_cnt-%d-%d-%"PRIu64,
+ xl_id, child, count);
+ ret = dict_set_uint64 (output, key, heal_failed_count);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+ "healed_failed_count to outout");
+ goto out;
+ }
+ snprintf (key, sizeof (key), "statistics_strt_time-%d-%d-%"PRIu64,
+ xl_id, child, count);
+ ret = dict_set_dynstr (output, key, gf_strdup(start_time_str));
+
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+ "crawl_start_time to outout");
+ goto out;
+ }
+
+ snprintf (key, sizeof (key), "statistics_end_time-%d-%d-%"PRIu64,
+ xl_id, child, count);
+
+ if (!end_time_str)
+ end_time_str = "Could not determine the end time";
+ ret = dict_set_dynstr (output, key, gf_strdup(end_time_str));
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+ "crawl_end_time to outout");
+ goto out;
+ }
+ snprintf (key, sizeof (key), "statistics_inprogress-%d-%d-%"PRIu64,
+ xl_id, child, count);
+
+ if (shd_event->crawl_inprogress == _gf_true)
+ progress = 1;
+ else
+ progress = 0;
+
+ ret = dict_set_int32 (output, key, progress);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+ "inprogress to outout");
+ goto out;
+ }
+
+ snprintf (key, sizeof (key), "statistics-%d-%d-count",xl_id, child);
+ ret = dict_set_uint64 (output, key, count + 1);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not increment the "
+ "counter.");
+ goto out;
+ }
+out:
+ return ret;
+}
+
+int
+_add_path_to_dict (xlator_t *this, dict_t *output, int child, char *path,
+ struct timeval *tv, gf_boolean_t dyn)
+{
+ //subkey not used for now
+ int ret = -1;
+ uint64_t count = 0;
+ char key[256] = {0};
+ int xl_id = 0;
+
+ ret = dict_get_int32 (output, this->name, &xl_id);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "xl does not have id");
+ goto out;
+ }
+
+ snprintf (key, sizeof (key), "%d-%d-count", xl_id, child);
+ ret = dict_get_uint64 (output, key, &count);
+
+ snprintf (key, sizeof (key), "%d-%d-%"PRIu64, xl_id, child, count);
+ if (dyn)
+ ret = dict_set_dynstr (output, key, path);
+ else
+ ret = dict_set_str (output, key, path);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "%s: Could not add to output",
+ path);
+ goto out;
+ }
+
+ if (!tv)
+ goto inc_count;
+ snprintf (key, sizeof (key), "%d-%d-%"PRIu64"-time", xl_id,
+ child, count);
+ ret = dict_set_uint32 (output, key, tv->tv_sec);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "%s: Could not set time",
+ path);
+ goto out;
+ }
+
+inc_count:
+ snprintf (key, sizeof (key), "%d-%d-count", xl_id, child);
+ ret = dict_set_uint64 (output, key, count + 1);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not increment count");
+ goto out;
+ }
+ ret = 0;
+out:
+ return ret;
+}
+
+int
+_get_path_from_gfid_loc (xlator_t *this, xlator_t *readdir_xl, loc_t *child,
+ char **fpath, gf_boolean_t *missing)
+{
+ dict_t *xattr = NULL;
+ char *path = NULL;
+ int ret = -1;
+
+ ret = syncop_getxattr (readdir_xl, child, &xattr, GFID_TO_PATH_KEY);
+ if (ret < 0) {
+ if ((errno == ENOENT) && missing)
+ *missing = _gf_true;
+ goto out;
+ }
+ ret = dict_get_str (xattr, GFID_TO_PATH_KEY, &path);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to get path for "
+ "gfid %s", uuid_utoa (child->gfid));
+ goto out;
+ }
+ path = gf_strdup (path);
+ if (!path) {
+ ret = -1;
+ goto out;
+ }
+ ret = 0;
+out:
+ if (!ret)
+ *fpath = path;
+ if (xattr)
+ dict_unref (xattr);
+ return ret;
+}
+
+int
+_add_event_to_dict (circular_buffer_t *cb, void *data)
+{
+ int ret = 0;
+ shd_dump_t *dump_data = NULL;
+ shd_event_t *shd_event = NULL;
+
+ dump_data = data;
+ shd_event = cb->data;
+ if (shd_event->child != dump_data->child)
+ goto out;
+ ret = _add_path_to_dict (dump_data->this, dump_data->dict,
+ dump_data->child, shd_event->path, &cb->tv,
+ _gf_false);
+out:
+ return ret;
+}
+
+int
+_add_crawl_event_statistics_to_dict (circular_buffer_t *cb, void *data)
+{
+ int ret = 0;
+ shd_dump_t *dump_data = NULL;
+ shd_crawl_event_t *shd_event = NULL;
+
+ dump_data = data;
+ shd_event = cb->data;
+ ret = _add_crawl_stats_to_dict (dump_data->this, dump_data->dict,
+ dump_data->child, shd_event, &cb->tv);
+ return ret;
+}
+
+int
+_add_eh_to_dict (xlator_t *this, eh_t *eh, dict_t *dict, int child)
+{
+ shd_dump_t dump_data = {0};
+
+ dump_data.this = this;
+ dump_data.dict = dict;
+ dump_data.child = child;
+ eh_dump (eh, &dump_data, _add_event_to_dict);
+ return 0;
+}
+
+
+int
+_add_statistics_to_dict (xlator_t *this, dict_t *dict, int child)
+{
+ shd_dump_t dump_data = {0};
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+
+ priv = this->private;
+ shd = &priv->shd;
+
+ dump_data.this = this;
+ dump_data.dict = dict;
+ dump_data.child = child;
+ eh_dump (shd->statistics[child], &dump_data,
+ _add_crawl_event_statistics_to_dict);
+ return 0;
+
+}
+
+void
+_remove_stale_index (xlator_t *this, xlator_t *readdir_xl,
+ loc_t *parent, char *fname)
+{
+ int ret = 0;
+ loc_t index_loc = {0};
+
+ ret = _build_index_loc (this, &index_loc, fname, parent);
+ if (ret)
+ goto out;
+ gf_log (this->name, GF_LOG_DEBUG, "Removing stale index "
+ "for %s on %s", index_loc.name, readdir_xl->name);
+ ret = syncop_unlink (readdir_xl, &index_loc);
+ if(ret && (errno != ENOENT)) {
+ gf_log(this->name, GF_LOG_ERROR, "%s: Failed to remove index "
+ "on %s - %s",index_loc.name, readdir_xl->name,
+ strerror (errno));
+ }
+ index_loc.path = NULL;
+ loc_wipe (&index_loc);
+out:
+ return;
+}
+
+int
+_count_hard_links_under_base_indices_dir (xlator_t *this,
+ afr_crawl_data_t *crawl_data,
+ gf_dirent_t *entry, loc_t *childloc,
+ loc_t *parentloc, struct iatt *iattr)
+{
+ xlator_t *readdir_xl = crawl_data->readdir_xl;
+ struct iatt parent = {0};
+ int ret = 0;
+ dict_t *output = NULL;
+ int xl_id = 0;
+ char key[256] = {0};
+ int child = -1;
+ uint64_t hardlinks = 0;
+
+ output = crawl_data->op_data;
+ child = crawl_data->child;
+
+ ret = syncop_lookup (readdir_xl, childloc, NULL, iattr, NULL, &parent);
+ if (ret)
+ goto out;
+
+ ret = dict_get_int32 (output, this->name, &xl_id);
+ if (ret)
+ goto out;
+
+ snprintf (key, sizeof (key), "%d-%d-hardlinks", xl_id, child);
+ ret = dict_get_uint64 (output, key, &hardlinks);
+
+ /*Removing the count of base_entry under indices/base_indicies and
+ * entry under indices/xattrop */
+ hardlinks = hardlinks + iattr->ia_nlink - 2;
+ ret = dict_set_uint64 (output, key, hardlinks);
+ if (ret)
+ goto out;
+
+out:
+ return ret;
+}
+
+int
+_add_summary_to_dict (xlator_t *this, afr_crawl_data_t *crawl_data,
+ gf_dirent_t *entry,
+ loc_t *childloc, loc_t *parentloc, struct iatt *iattr)
+{
+ dict_t *output = NULL;
+ xlator_t *readdir_xl = NULL;
+ int ret = -1;
+ char *path = NULL;
+ gf_boolean_t missing = _gf_false;
+ char gfid_str[64] = {0};
+
+ if (uuid_is_null (childloc->gfid))
+ goto out;
+
+ output = crawl_data->op_data;
+ readdir_xl = crawl_data->readdir_xl;
+
+ ret = _get_path_from_gfid_loc (this, readdir_xl, childloc, &path,
+ &missing);
+ if (ret == 0) {
+ ret = _add_path_to_dict (this, output, crawl_data->child, path,
+ NULL, _gf_true);
+ } else if (missing) {
+ _remove_stale_index (this, readdir_xl, parentloc,
+ uuid_utoa_r (childloc->gfid, gfid_str));
+ }
+
+out:
+ if (ret && path)
+ GF_FREE (path);
+ return ret;
+}
+
+void
+_crawl_post_sh_action (xlator_t *this, loc_t *parent, loc_t *child,
+ int32_t op_ret, int32_t op_errno, dict_t *xattr_rsp,
+ afr_crawl_data_t *crawl_data)
+{
+ int ret = 0;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ eh_t *eh = NULL;
+ char *path = NULL;
+ char gfid_str[64] = {0};
+ shd_event_t *event = NULL;
+ int32_t sh_failed = 0;
+ gf_boolean_t split_brain = 0;
+ int32_t actual_sh_done = 0;
+ shd_crawl_event_t **shd_crawl_event = NULL;
+
+ priv = this->private;
+ shd = &priv->shd;
+ if (crawl_data->crawl == INDEX) {
+ if ((op_ret < 0) && (op_errno == ENOENT)) {
+ _remove_stale_index (this, crawl_data->readdir_xl,
+ parent, uuid_utoa_r (child->gfid,
+ gfid_str));
+ goto out;
+ }
+ ret = _get_path_from_gfid_loc (this, crawl_data->readdir_xl,
+ child, &path, NULL);
+ if (ret)
+ goto out;
+ } else {
+ path = gf_strdup (child->path);
+ if (!path) {
+ ret = -1;
+ goto out;
+ }
+ }
+
+ if (xattr_rsp) {
+ ret = dict_get_int32 (xattr_rsp, "sh-failed", &sh_failed);
+ ret = dict_get_int32 (xattr_rsp, "actual-sh-done", &actual_sh_done);
+ }
+
+ shd_crawl_event = (shd_crawl_event_t**)(shd->crawl_events);
+
+ split_brain = afr_is_split_brain (this, child->inode);
+ if ((op_ret < 0 && op_errno == EIO) || split_brain) {
+ eh = shd->split_brain;
+ shd_crawl_event[crawl_data->child]->split_brain_count += 1;
+ } else if ((op_ret < 0) || sh_failed) {
+ eh = shd->heal_failed;
+ shd_crawl_event[crawl_data->child]->heal_failed_count += 1;
+ } else if (actual_sh_done == 1) {
+ eh = shd->healed;
+ shd_crawl_event[crawl_data->child]->healed_count += 1;
+ }
+ ret = -1;
+
+ if (eh != NULL) {
+ event = GF_CALLOC (1, sizeof (*event), gf_afr_mt_shd_event_t);
+ if (!event)
+ goto out;
+ event->child = crawl_data->child;
+ event->path = path;
+
+ ret = eh_save_history (eh, event);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "%s:Failed to save "
+ "to event history, (%d, %s)", path, op_ret,
+ strerror (op_errno));
+
+ goto out;
+ }
+ } else {
+ gf_log (this->name, GF_LOG_DEBUG, "%s:Self heal already done ",
+ path);
+
+ }
+ ret = 0;
+out:
+ if (ret && path)
+ GF_FREE (path);
+ return;
+}
+
+int
+_link_inode_update_loc (xlator_t *this, loc_t *loc, struct iatt *iattr)
+{
+ inode_t *link_inode = NULL;
+ int ret = -1;
+
+ link_inode = inode_link (loc->inode, NULL, NULL, iattr);
+ if (link_inode == NULL) {
+ gf_log (this->name, GF_LOG_ERROR, "inode link failed "
+ "on the inode (%s)", uuid_utoa (iattr->ia_gfid));
+ goto out;
+ }
+ inode_unref (loc->inode);
+ loc->inode = link_inode;
+ ret = 0;
+out:
+ return ret;
+}
+
+int
+_self_heal_entry (xlator_t *this, afr_crawl_data_t *crawl_data, gf_dirent_t *entry,
+ loc_t *child, loc_t *parent, struct iatt *iattr)
+{
+ struct iatt parentbuf = {0};
+ int ret = 0;
+ dict_t *xattr_rsp = NULL;
+ dict_t *xattr_req = NULL;
+
+ xattr_req = dict_new ();
+ if (!xattr_req) {
+ errno = ENOMEM;
+ ret = -1;
+ goto out;
+ }
+
+ ret = dict_set_int32 (xattr_req, "allow-sh-for-running-transaction", 1);
+
+ gf_log (this->name, GF_LOG_DEBUG, "lookup %s", child->path);
+
+ ret = syncop_lookup (this, child, xattr_req,
+ iattr, &xattr_rsp, &parentbuf);
+ _crawl_post_sh_action (this, parent, child, ret, errno, xattr_rsp,
+ crawl_data);
+ if (xattr_rsp)
+ dict_unref (xattr_rsp);
+ if (ret == 0)
+ ret = _link_inode_update_loc (this, child, iattr);
+
+out:
+ if (xattr_req)
+ dict_unref(xattr_req);
+ return ret;
+}
+
+static int
+afr_crawl_done (int ret, call_frame_t *sync_frame, void *data)
+{
+ GF_FREE (data);
+ STACK_DESTROY (sync_frame->root);
+ return 0;
+}
+
+void
+_do_self_heal_on_subvol (xlator_t *this, int child, afr_crawl_type_t crawl)
+{
+ afr_start_crawl (this, child, crawl, _self_heal_entry,
+ NULL, _gf_true, STOP_CRAWL_ON_SINGLE_SUBVOL,
+ afr_crawl_done);
+}
+
+gf_boolean_t
+_crawl_proceed (xlator_t *this, int child, int crawl_flags, char **reason)
+{
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ gf_boolean_t proceed = _gf_false;
+ char *msg = NULL;
+
+ priv = this->private;
+ shd = &priv->shd;
+ if (!shd->enabled) {
+ msg = "Self-heal daemon is not enabled";
+ gf_log (this->name, GF_LOG_DEBUG, "%s", msg);
+ goto out;
+ }
+ if (!priv->child_up[child]) {
+ gf_log (this->name, GF_LOG_DEBUG, "Stopping crawl for %s , "
+ "subvol went down", priv->children[child]->name);
+ msg = "Brick is Not connected";
+ goto out;
+ }
+
+ if (crawl_flags & STOP_CRAWL_ON_SINGLE_SUBVOL) {
+ if (afr_up_children_count (priv->child_up,
+ priv->child_count) < 2) {
+ gf_log (this->name, GF_LOG_DEBUG, "Stopping crawl as "
+ "< 2 children are up");
+ msg = "< 2 bricks in replica are running";
+ goto out;
+ }
+ }
+ proceed = _gf_true;
+out:
+ if (reason)
+ *reason = msg;
+ return proceed;
+}
+
+int
+_do_crawl_op_on_local_subvols (xlator_t *this, afr_crawl_type_t crawl,
+ shd_crawl_op op, dict_t *output)
+{
+ afr_private_t *priv = NULL;
+ char *status = NULL;
+ char *subkey = NULL;
+ char key[256] = {0};
+ shd_pos_t pos_data = {0};
+ int op_ret = -1;
+ int xl_id = -1;
+ int i = 0;
+ int ret = 0;
+ int crawl_flags = 0;
+
+ priv = this->private;
+ if (op == HEAL)
+ crawl_flags |= STOP_CRAWL_ON_SINGLE_SUBVOL;
+
+ if (output) {
+ ret = dict_get_int32 (output, this->name, &xl_id);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Invalid input, "
+ "translator-id is not available");
+ goto out;
+ }
+ }
+ pos_data.this = this;
+ subkey = "status";
+ for (i = 0; i < priv->child_count; i++) {
+ if (_crawl_proceed (this, i, crawl_flags, &status)) {
+ pos_data.child = i;
+ /*
+ * We're already in a synctask in this case, so we
+ * don't need to defer through a second (and in fact
+ * that can cause deadlock). Just call straight
+ * through instead.
+ */
+ ret = afr_find_child_position(pos_data.this,
+ pos_data.child,
+ &pos_data.pos);
+ if (ret) {
+ status = "Not able to find brick location";
+ } else if (pos_data.pos == AFR_POS_REMOTE) {
+ status = "brick is remote";
+ } else {
+ op_ret = 0;
+ if (op == HEAL) {
+ status = "Started self-heal";
+ _do_self_heal_on_subvol (this, i,
+ crawl);
+ } else if (output && (op == INFO)) {
+ status = "";
+ afr_start_crawl (this, i, INDEX,
+ _add_summary_to_dict,
+ output, _gf_false, 0,
+ NULL);
+ } else if (output &&
+ (op == STATISTICS_TO_BE_HEALED)) {
+ status = "";
+ afr_start_crawl (this, i,
+ INDEX_TO_BE_HEALED,
+ _count_hard_links_under_base_indices_dir,
+ output, _gf_false,
+ 0, NULL);
+ }
+ }
+ if (output) {
+ snprintf (key, sizeof (key), "%d-%d-%s", xl_id,
+ i, subkey);
+ ret = dict_set_str (output, key, status);
+ }
+ if (!op_ret && (crawl == FULL))
+ break;
+ }
+ if (output) {
+ snprintf (key, sizeof (key), "%d-%d-%s", xl_id, i,
+ subkey);
+ ret = dict_set_str (output, key, status);
+ }
+ }
+out:
+ return op_ret;
+}
+
+int
+_do_self_heal_on_local_subvols (xlator_t *this, afr_crawl_type_t crawl,
+ dict_t *output)
+{
+ return _do_crawl_op_on_local_subvols (this, crawl, HEAL, output);
+}
+
+int
+_get_index_summary_on_local_subvols (xlator_t *this, dict_t *output)
+{
+ return _do_crawl_op_on_local_subvols (this, INDEX, INFO, output);
+}
+
+void
+afr_fill_completed_crawl_statistics_to_dict (xlator_t *this, dict_t *dict)
+{
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ int i = 0;
+ priv = this->private;
+ shd= &priv->shd;
+ for (i = 0; i < priv->child_count; i++) {
+ if (shd->pos[i] != AFR_POS_LOCAL)
+ continue;
+ _add_statistics_to_dict (this, dict, i);
+ }
+
+ return ;
+}
+
+static void
+reset_crawl_event (shd_crawl_event_t *crawl_event)
+{
+ crawl_event->healed_count = 0;
+ crawl_event->split_brain_count = 0;
+ crawl_event->heal_failed_count = 0;
+ GF_FREE (crawl_event->start_time_str);
+ crawl_event->start_time_str = NULL;
+ crawl_event->end_time_str = NULL;
+ crawl_event->crawl_type = NULL;
+ crawl_event->crawl_inprogress = _gf_false;
+ return;
+}
+
+static void
+afr_copy_crawl_event_struct (shd_crawl_event_t *src, shd_crawl_event_t *dst)
+{
+ dst->healed_count = src->healed_count;
+ dst->split_brain_count = src->split_brain_count;
+ dst->heal_failed_count = src->heal_failed_count;
+ dst->start_time_str = gf_strdup (src->start_time_str);
+ dst->end_time_str = "Crawl is already in progress";
+ dst->crawl_type = src->crawl_type;
+ dst->crawl_inprogress = _gf_true;
+ return;
+}
+
+static int
+afr_fill_crawl_statistics_of_running_crawl(xlator_t *this, dict_t *dict)
+{
+ shd_crawl_event_t *evnt = NULL;
+ int ret = 0;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ int i = 0;
+ priv = this->private;
+ shd = &priv->shd;
+
+ evnt = GF_CALLOC (1, sizeof (shd_crawl_event_t),
+ gf_afr_mt_shd_crawl_event_t);
+ if (!evnt) {
+ ret = -1;
+ goto out;
+ }
+ LOCK (&priv->lock);
+ {
+ for (i = 0; i < priv->child_count; i++) {
+ if (shd->pos[i] != AFR_POS_LOCAL)
+ continue;
+
+ reset_crawl_event (evnt);
+
+ if (!shd->crawl_events[i]) {
+ continue;
+ }
+
+ afr_copy_crawl_event_struct (shd->crawl_events[i],
+ evnt);
+ _add_crawl_stats_to_dict (this, dict, i, evnt, NULL);
+
+ }
+ }
+ UNLOCK (&priv->lock);
+ reset_crawl_event (evnt);
+ GF_FREE (evnt);
+
+out:
+ return ret;
+}
+
+static int
+_add_local_subvols_crawl_statistics_to_dict (xlator_t *this, dict_t *dict)
+{
+ int ret = 0;
+ afr_fill_completed_crawl_statistics_to_dict (this, dict);
+ ret = afr_fill_crawl_statistics_of_running_crawl (this, dict);
+ return ret;
+}
+int
+_add_local_subvols_eh_to_dict (xlator_t *this, eh_t *eh, dict_t *dict)
+{
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ int i = 0;
+
+ priv = this->private;
+ shd = &priv->shd;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (shd->pos[i] != AFR_POS_LOCAL)
+ continue;
+ _add_eh_to_dict (this, eh, dict, i);
+ }
+ return 0;
+}
+
+int
+afr_xl_op (xlator_t *this, dict_t *input, dict_t *output)
+{
+ gf_xl_afr_op_t op = GF_AFR_OP_INVALID;
+ int ret = 0;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ int xl_id = 0;
+
+ priv = this->private;
+ shd = &priv->shd;
+
+ ret = dict_get_int32 (input, "xl-op", (int32_t*)&op);
+ if (ret)
+ goto out;
+ ret = dict_get_int32 (input, this->name, &xl_id);
+ if (ret)
+ goto out;
+ ret = dict_set_int32 (output, this->name, xl_id);
+ if (ret)
+ goto out;
+ switch (op) {
+ case GF_AFR_OP_HEAL_INDEX:
+ ret = _do_self_heal_on_local_subvols (this, INDEX, output);
+ break;
+ case GF_AFR_OP_HEAL_FULL:
+ ret = _do_self_heal_on_local_subvols (this, FULL, output);
+ break;
+ case GF_AFR_OP_INDEX_SUMMARY:
+ (void)_get_index_summary_on_local_subvols (this, output);
+ ret = 0;
+ break;
+ case GF_AFR_OP_HEALED_FILES:
+ ret = _add_local_subvols_eh_to_dict (this, shd->healed, output);
+ break;
+ case GF_AFR_OP_HEAL_FAILED_FILES:
+ ret = _add_local_subvols_eh_to_dict (this, shd->heal_failed,
+ output);
+ break;
+ case GF_AFR_OP_SPLIT_BRAIN_FILES:
+ ret = _add_local_subvols_eh_to_dict (this, shd->split_brain,
+ output);
+ break;
+ case GF_AFR_OP_STATISTICS:
+ ret = _add_local_subvols_crawl_statistics_to_dict (this, output);
+ break;
+ case GF_AFR_OP_STATISTICS_HEAL_COUNT:
+ case GF_AFR_OP_STATISTICS_HEAL_COUNT_PER_REPLICA:
+ ret = _do_crawl_op_on_local_subvols (this, INDEX_TO_BE_HEALED,
+ STATISTICS_TO_BE_HEALED,
+ output);
+ break;
+ default:
+ gf_log (this->name, GF_LOG_ERROR, "Unknown set op %d", op);
+ break;
+ }
+out:
+ dict_del (output, this->name);
+ return ret;
+}
+
+void
+afr_poll_self_heal (void *data)
+{
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ struct timespec timeout = {0};
+ xlator_t *this = NULL;
+ long child = (long)data;
+ gf_timer_t *old_timer = NULL;
+ gf_timer_t *new_timer = NULL;
+ shd_pos_t pos_data = {0};
+ int ret = 0;
+
+ this = THIS;
+ priv = this->private;
+ shd = &priv->shd;
+
+ if (shd->pos[child] == AFR_POS_UNKNOWN) {
+ pos_data.this = this;
+ pos_data.child = child;
+ ret = synctask_new (this->ctx->env,
+ afr_syncop_find_child_position,
+ NULL, NULL, &pos_data);
+ if (!ret)
+ shd->pos[child] = pos_data.pos;
+ }
+ if (shd->enabled && (shd->pos[child] == AFR_POS_LOCAL))
+ _do_self_heal_on_subvol (this, child, INDEX);
+ timeout.tv_sec = shd->timeout;
+ timeout.tv_nsec = 0;
+ //notify and previous timer should be synchronized.
+ LOCK (&priv->lock);
+ {
+ old_timer = shd->timer[child];
+ if (shd->pos[child] == AFR_POS_REMOTE)
+ goto unlock;
+ shd->timer[child] = gf_timer_call_after (this->ctx, timeout,
+ afr_poll_self_heal,
+ data);
+ new_timer = shd->timer[child];
+ }
+unlock:
+ UNLOCK (&priv->lock);
+
+ if (old_timer)
+ gf_timer_call_cancel (this->ctx, old_timer);
+ if (!new_timer && (shd->pos[child] != AFR_POS_REMOTE)) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "Could not create self-heal polling timer for %s",
+ priv->children[child]->name);
+ }
+ return;
+}
+
+static int
+afr_handle_child_up (int ret, call_frame_t *sync_frame, void *data)
+{
+ afr_self_heald_t *shd = NULL;
+ shd_pos_t *pos_data = data;
+ afr_private_t *priv = NULL;
+
+ if (ret)
+ goto out;
+
+ priv = pos_data->this->private;
+ shd = &priv->shd;
+ shd->pos[pos_data->child] = pos_data->pos;
+ if (pos_data->pos != AFR_POS_REMOTE)
+ afr_poll_self_heal ((void*)(long)pos_data->child);
+ _do_self_heal_on_local_subvols (THIS, INDEX, NULL);
+out:
+ GF_FREE (data);
+ return 0;
+}
+
+void
+afr_proactive_self_heal (void *data)
+{
+ xlator_t *this = NULL;
+ long child = (long)data;
+ shd_pos_t *pos_data = NULL;
+ int ret = 0;
+
+ this = THIS;
+
+ //Position of brick could have changed and it could be local now.
+ //Compute the position again
+ pos_data = GF_CALLOC (1, sizeof (*pos_data), gf_afr_mt_pos_data_t);
+ if (!pos_data)
+ goto out;
+ pos_data->this = this;
+ pos_data->child = child;
+ ret = synctask_new (this->ctx->env, afr_syncop_find_child_position,
+ afr_handle_child_up, NULL, pos_data);
+ if (ret)
+ goto out;
+out:
+ return;
+}
+
+static int
+get_pathinfo_host (char *pathinfo, char *hostname, size_t size)
+{
+ char *start = NULL;
+ char *end = NULL;
+ int ret = -1;
+ int i = 0;
+
+ if (!pathinfo)
+ goto out;
+
+ start = strchr (pathinfo, ':');
+ if (!start)
+ goto out;
+ end = strrchr (pathinfo, ':');
+ if (start == end)
+ goto out;
+
+ memset (hostname, 0, size);
+ i = 0;
+ while (++start != end)
+ hostname[i++] = *start;
+ ret = 0;
+out:
+ return ret;
+}
+
+int
+afr_local_pathinfo (char *pathinfo, gf_boolean_t *local)
+{
+ int ret = 0;
+ char pathinfohost[1024] = {0};
+ char localhost[1024] = {0};
+ xlator_t *this = THIS;
+
+ *local = _gf_false;
+ ret = get_pathinfo_host (pathinfo, pathinfohost, sizeof (pathinfohost));
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Invalid pathinfo: %s",
+ pathinfo);
+ goto out;
+ }
+
+ ret = gethostname (localhost, sizeof (localhost));
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "gethostname() failed, "
+ "reason: %s", strerror (errno));
+ goto out;
+ }
+
+ if (!strcmp (localhost, pathinfohost))
+ *local = _gf_true;
+out:
+ return ret;
+}
+
+int
+afr_crawl_build_start_loc (xlator_t *this, afr_crawl_data_t *crawl_data,
+ loc_t *dirloc)
+{
+ afr_private_t *priv = NULL;
+ dict_t *xattr = NULL;
+ void *index_gfid = NULL;
+ void *base_indices_holder_vgfid = NULL;
+ loc_t rootloc = {0};
+ struct iatt iattr = {0};
+ struct iatt parent = {0};
+ int ret = 0;
+ xlator_t *readdir_xl = crawl_data->readdir_xl;
+
+ priv = this->private;
+ if (crawl_data->crawl == FULL) {
+ afr_build_root_loc (this, dirloc);
+ } else if (crawl_data->crawl == INDEX) {
+ afr_build_root_loc (this, &rootloc);
+ ret = syncop_getxattr (readdir_xl, &rootloc, &xattr,
+ GF_XATTROP_INDEX_GFID);
+ if (ret < 0)
+ goto out;
+ ret = dict_get_ptr (xattr, GF_XATTROP_INDEX_GFID, &index_gfid);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "failed to get index "
+ "dir gfid on %s", readdir_xl->name);
+ goto out;
+ }
+ if (!index_gfid) {
+ gf_log (this->name, GF_LOG_ERROR, "index gfid empty "
+ "on %s", readdir_xl->name);
+ ret = -1;
+ goto out;
+ }
+ uuid_copy (dirloc->gfid, index_gfid);
+ dirloc->path = "";
+ dirloc->inode = inode_new (priv->root_inode->table);
+ ret = syncop_lookup (readdir_xl, dirloc, NULL,
+ &iattr, NULL, &parent);
+ if (ret < 0) {
+ if (errno != ENOENT) {
+ gf_log (this->name, GF_LOG_ERROR, "lookup "
+ "failed on index dir on %s - (%s)",
+ readdir_xl->name, strerror (errno));
+ }
+ goto out;
+ }
+ ret = _link_inode_update_loc (this, dirloc, &iattr);
+ if (ret)
+ goto out;
+ } else if (crawl_data->crawl == INDEX_TO_BE_HEALED) {
+ afr_build_root_loc (this, &rootloc);
+ ret = syncop_getxattr (readdir_xl, &rootloc, &xattr,
+ GF_BASE_INDICES_HOLDER_GFID);
+ if (ret < 0)
+ goto out;
+ ret = dict_get_ptr (xattr, GF_BASE_INDICES_HOLDER_GFID,
+ &base_indices_holder_vgfid);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "index gfid empty "
+ "on %s", readdir_xl->name);
+ ret = -1;
+ goto out;
+ }
+ if (!base_indices_holder_vgfid) {
+ gf_log (this->name, GF_LOG_ERROR, "Base indices holder"
+ "virtual gfid is null on %s", readdir_xl->name);
+ ret = -1;
+ goto out;
+ }
+ uuid_copy (dirloc->gfid, base_indices_holder_vgfid);
+ dirloc->path = "";
+ dirloc->inode = inode_new (priv->root_inode->table);
+ ret = syncop_lookup (readdir_xl, dirloc, NULL, &iattr, NULL,
+ &parent);
+ if (ret < 0) {
+ if (errno != ENOENT) {
+ gf_log (this->name, GF_LOG_ERROR, "lookup "
+ "failed for base_indices_holder dir"
+ " on %s - (%s)", readdir_xl->name,
+ strerror (errno));
+
+ } else {
+ gf_log (this->name, GF_LOG_ERROR, "base_indices"
+ "_holder is not yet created.");
+ }
+ goto out;
+ }
+ ret = _link_inode_update_loc (this, dirloc, &iattr);
+ if (ret)
+ goto out;
+ }
+ ret = 0;
+out:
+ if (xattr)
+ dict_unref (xattr);
+ loc_wipe (&rootloc);
+ return ret;
+}
+
+int
+afr_crawl_opendir (xlator_t *this, afr_crawl_data_t *crawl_data, fd_t **dirfd,
+ loc_t *dirloc)
+{
+ fd_t *fd = NULL;
+ int ret = 0;
+
+ if (crawl_data->crawl == FULL) {
+ fd = fd_create (dirloc->inode, crawl_data->pid);
+ if (!fd) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "Failed to create fd for %s", dirloc->path);
+ ret = -1;
+ goto out;
+ }
+
+ ret = syncop_opendir (crawl_data->readdir_xl, dirloc, fd);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "opendir failed on %s", dirloc->path);
+ goto out;
+ }
+ } else {
+ fd = fd_anonymous (dirloc->inode);
+ }
+ ret = 0;
+out:
+ if (!ret)
+ *dirfd = fd;
+ return ret;
+}
+
+xlator_t*
+afr_crawl_readdir_xl_get (xlator_t *this, afr_crawl_data_t *crawl_data)
+{
+ afr_private_t *priv = this->private;
+
+ if (crawl_data->crawl == FULL) {
+ return this;
+ } else {
+ return priv->children[crawl_data->child];
+ }
+ return NULL;
+}
+
+int
+afr_crawl_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent,
+ gf_dirent_t *entry, afr_crawl_data_t *crawl_data)
+{
+ int ret = -1;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ if (crawl_data->crawl == FULL) {
+ ret = afr_build_child_loc (this, child, parent, entry->d_name);
+ } else if (crawl_data->crawl == INDEX_TO_BE_HEALED) {
+ ret = _build_index_loc (this, child, entry->d_name, parent);
+ if (ret)
+ goto out;
+ child->inode = inode_new (priv->root_inode->table);
+ if (!child->inode) {
+ ret = -1;
+ goto out;
+ }
+ child->path = NULL;
+ } else {
+ child->inode = inode_new (priv->root_inode->table);
+ if (!child->inode)
+ goto out;
+ uuid_parse (entry->d_name, child->gfid);
+ ret = _loc_assign_gfid_path (child);
+ }
+out:
+ return ret;
+}
+
+static int
+_process_entries (xlator_t *this, loc_t *parentloc, gf_dirent_t *entries,
+ off_t *offset, afr_crawl_data_t *crawl_data)
+{
+ gf_dirent_t *entry = NULL;
+ gf_dirent_t *tmp = NULL;
+ int ret = 0;
+ loc_t entry_loc = {0};
+ fd_t *fd = NULL;
+ struct iatt iattr = {0};
+
+ list_for_each_entry_safe (entry, tmp, &entries->list, list) {
+ if (!_crawl_proceed (this, crawl_data->child,
+ crawl_data->crawl_flags, NULL)) {
+ ret = -1;
+ goto out;
+ }
+ *offset = entry->d_off;
+ if (IS_ENTRY_CWD (entry->d_name) ||
+ IS_ENTRY_PARENT (entry->d_name))
+ continue;
+ if ((crawl_data->crawl == FULL) &&
+ uuid_is_null (entry->d_stat.ia_gfid)) {
+ gf_log (this->name, GF_LOG_WARNING, "%s/%s: No "
+ "gfid present skipping",
+ parentloc->path, entry->d_name);
+ continue;
+ }
+
+ loc_wipe (&entry_loc);
+ ret = afr_crawl_build_child_loc (this, &entry_loc, parentloc,
+ entry, crawl_data);
+ if (ret)
+ goto out;
+
+ ret = crawl_data->process_entry (this, crawl_data, entry,
+ &entry_loc, parentloc, &iattr);
+
+ if (crawl_data->crawl == INDEX_TO_BE_HEALED && ret) {
+ goto out;
+ } else if (ret) {
+ continue;
+ }
+
+ if ((crawl_data->crawl == INDEX) ||
+ (crawl_data->crawl == INDEX_TO_BE_HEALED))
+ continue;
+
+ if (!IA_ISDIR (iattr.ia_type))
+ continue;
+ fd = NULL;
+ ret = afr_crawl_opendir (this, crawl_data, &fd, &entry_loc);
+ if (ret)
+ continue;
+ ret = _crawl_directory (fd, &entry_loc, crawl_data);
+ if (fd)
+ fd_unref (fd);
+ }
+ ret = 0;
+out:
+ if ((crawl_data->crawl == INDEX_TO_BE_HEALED) && ret) {
+ gf_log (this->name, GF_LOG_ERROR,"Failed to get the hardlink "
+ "count");
+ }
+ loc_wipe (&entry_loc);
+ return ret;
+}
+
+static int
+_crawl_directory (fd_t *fd, loc_t *loc, afr_crawl_data_t *crawl_data)
+{
+ xlator_t *this = NULL;
+ off_t offset = 0;
+ gf_dirent_t entries;
+ int ret = 0;
+ gf_boolean_t free_entries = _gf_false;
+ xlator_t *readdir_xl = crawl_data->readdir_xl;
+
+ INIT_LIST_HEAD (&entries.list);
+ this = THIS;
+
+ GF_ASSERT (loc->inode);
+
+ if (crawl_data->crawl == FULL)
+ gf_log (this->name, GF_LOG_DEBUG, "crawling %s", loc->path);
+ else
+ gf_log (this->name, GF_LOG_DEBUG, "crawling INDEX %s",
+ uuid_utoa (loc->gfid));
+
+ while (1) {
+ if (crawl_data->crawl == FULL)
+ ret = syncop_readdirp (readdir_xl, fd, 131072, offset,
+ NULL, &entries);
+ else
+ ret = syncop_readdir (readdir_xl, fd, 131072, offset,
+ &entries);
+ if (ret <= 0)
+ break;
+ ret = 0;
+ free_entries = _gf_true;
+
+ if (!_crawl_proceed (this, crawl_data->child,
+ crawl_data->crawl_flags, NULL)) {
+ ret = -1;
+ goto out;
+ }
+ if (list_empty (&entries.list))
+ goto out;
+
+ ret = _process_entries (this, loc, &entries, &offset,
+ crawl_data);
+ if ((ret < 0) && (crawl_data->crawl == INDEX_TO_BE_HEALED)) {
+ goto out;
+ }
+ gf_dirent_free (&entries);
+ free_entries = _gf_false;
+ }
+ ret = 0;
+out:
+ if (free_entries)
+ gf_dirent_free (&entries);
+ return ret;
+}
+
+static char*
+position_str_get (afr_child_pos_t pos)
+{
+ switch (pos) {
+ case AFR_POS_UNKNOWN:
+ return "unknown";
+ case AFR_POS_LOCAL:
+ return "local";
+ case AFR_POS_REMOTE:
+ return "remote";
+ }
+ return NULL;
+}
+
+int
+afr_find_child_position (xlator_t *this, int child, afr_child_pos_t *pos)
+{
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ dict_t *xattr_rsp = NULL;
+ loc_t loc = {0};
+ int ret = 0;
+ char *node_uuid = NULL;
+
+ priv = this->private;
+ shd = &priv->shd;
+
+ afr_build_root_loc (this, &loc);
+
+ ret = syncop_getxattr (priv->children[child], &loc, &xattr_rsp,
+ GF_XATTR_NODE_UUID_KEY);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "getxattr failed on %s - "
+ "(%s)", priv->children[child]->name, strerror (errno));
+ goto out;
+ }
+
+ ret = dict_get_str (xattr_rsp, GF_XATTR_NODE_UUID_KEY, &node_uuid);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "node-uuid key not found on "
+ "child %s", priv->children[child]->name);
+ goto out;
+ }
+
+ if (!strcmp (node_uuid, shd->node_uuid))
+ *pos = AFR_POS_LOCAL;
+ else
+ *pos = AFR_POS_REMOTE;
+
+ gf_log (this->name, GF_LOG_DEBUG, "child %s is %s",
+ priv->children[child]->name, position_str_get (*pos));
+out:
+ if (ret)
+ *pos = AFR_POS_UNKNOWN;
+ loc_wipe (&loc);
+ return ret;
+}
+
+int
+afr_syncop_find_child_position (void *data)
+{
+ shd_pos_t *pos_data = data;
+ int ret = 0;
+
+ ret = afr_find_child_position (pos_data->this, pos_data->child,
+ &pos_data->pos);
+ return ret;
+}
+
+static int
+afr_dir_crawl (void *data)
+{
+ xlator_t *this = NULL;
+ int ret = -1;
+ xlator_t *readdir_xl = NULL;
+ fd_t *fd = NULL;
+ loc_t dirloc = {0};
+ afr_crawl_data_t *crawl_data = data;
+
+ this = THIS;
+
+ if (!_crawl_proceed (this, crawl_data->child, crawl_data->crawl_flags,
+ NULL))
+ goto out;
+
+ readdir_xl = afr_crawl_readdir_xl_get (this, crawl_data);
+ if (!readdir_xl)
+ goto out;
+ crawl_data->readdir_xl = readdir_xl;
+
+ ret = afr_crawl_build_start_loc (this, crawl_data, &dirloc);
+ if (ret)
+ goto out;
+
+ ret = afr_crawl_opendir (this, crawl_data, &fd, &dirloc);
+ if (ret) {
+ if (crawl_data->crawl == INDEX_TO_BE_HEALED) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to open base_"
+ "indices_holder");
+ }
+ goto out;
+ }
+
+ ret = _crawl_directory (fd, &dirloc, crawl_data);
+ if (ret)
+ gf_log (this->name, GF_LOG_ERROR, "Crawl failed on %s",
+ readdir_xl->name);
+ else
+ gf_log (this->name, GF_LOG_DEBUG, "Crawl completed "
+ "on %s", readdir_xl->name);
+ if (crawl_data->crawl == INDEX)
+ dirloc.path = NULL;
+out:
+ if (fd)
+ fd_unref (fd);
+ if ((crawl_data->crawl == INDEX) ||
+ (crawl_data->crawl == INDEX_TO_BE_HEALED ))
+ dirloc.path = NULL;
+ loc_wipe (&dirloc);
+ return ret;
+}
+
+char *
+get_crawl_type_in_string (afr_crawl_type_t crawl)
+{
+ char *index = "INDEX";
+ char *full = "FULL";
+ char *crawl_type = NULL;
+
+ if (crawl == INDEX){
+ crawl_type = index;
+ } else if (crawl == FULL) {
+ crawl_type = full;
+ }
+
+ return crawl_type;
+}
+
+static int
+afr_allocate_crawl_event (xlator_t *this, int child, afr_crawl_type_t crawl)
+{
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ int ret = 0;
+ shd_crawl_event_t *crawl_event = NULL;
+ time_t get_time = 0;
+
+ priv = this->private;
+ shd = &priv->shd;
+
+ crawl_event = GF_CALLOC (sizeof (shd_crawl_event_t), 1,
+ gf_afr_mt_shd_crawl_event_t);
+ if (!crawl_event) {
+ ret = -1;
+ goto out;
+ }
+
+ get_time = time(NULL);
+ if (get_time == ((time_t)-1)) {
+ ret = -1;
+ goto out;
+ }
+
+ crawl_event->start_time_str = gf_strdup (ctime(&get_time));
+
+ crawl_event->crawl_type = get_crawl_type_in_string (crawl);
+ if (!crawl_event->crawl_type) {
+ ret = -1;
+ goto out;
+ }
+ LOCK (&priv->lock);
+ {
+ shd->crawl_events[child] = crawl_event;
+ }
+ UNLOCK (&priv->lock);
+ ret = 0;
+out:
+ return ret;
+
+}
+
+static int
+afr_put_crawl_event_in_eh (xlator_t *this, int child)
+{
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ int ret = 0;
+ time_t get_time = 0;
+ shd_crawl_event_t **crawl_event = NULL;
+
+ priv = this->private;
+ shd = &priv->shd;
+
+ get_time = time(NULL);
+ if (get_time == ((time_t)-1)) {
+ ret = -1;
+ goto out;
+ }
+ crawl_event = (shd_crawl_event_t**)shd->crawl_events;
+ LOCK (&priv->lock);
+ {
+ crawl_event[child]->end_time_str = gf_strdup (ctime(&get_time));
+ ret = eh_save_history (shd->statistics[child],
+ crawl_event[child]);
+ crawl_event[child] = NULL;
+ }
+ UNLOCK (&priv->lock);
+out:
+ return ret;
+}
+
+static int
+afr_dir_exclusive_crawl (void *data)
+{
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ gf_boolean_t crawl = _gf_false;
+ int ret = 0;
+ int child = -1;
+ xlator_t *this = NULL;
+ afr_crawl_data_t *crawl_data = data;
+
+ this = THIS;
+ priv = this->private;
+ shd = &priv->shd;
+ child = crawl_data->child;
+
+ LOCK (&priv->lock);
+ {
+ if (shd->inprogress[child]) {
+ if (shd->pending[child] != FULL)
+ shd->pending[child] = crawl_data->crawl;
+ } else {
+ shd->inprogress[child] = _gf_true;
+ crawl = _gf_true;
+ }
+ }
+ UNLOCK (&priv->lock);
+
+ if (!crawl) {
+ gf_log (this->name, GF_LOG_INFO, "Another crawl is in progress "
+ "for %s", priv->children[child]->name);
+ goto out;
+ }
+
+ do {
+ ret = afr_allocate_crawl_event (this, child, crawl_data->crawl);
+ if (ret)
+ goto out;
+ afr_dir_crawl (data);
+
+ ret = afr_put_crawl_event_in_eh (this, child);
+ if (ret < 0)
+ goto out;
+
+ LOCK (&priv->lock);
+ {
+ if (shd->pending[child] != NONE) {
+ crawl_data->crawl = shd->pending[child];
+ shd->pending[child] = NONE;
+ } else {
+ shd->inprogress[child] = _gf_false;
+ crawl = _gf_false;
+ }
+ }
+ UNLOCK (&priv->lock);
+ } while (crawl);
+out:
+ return ret;
+}
+
+void
+afr_start_crawl (xlator_t *this, int idx, afr_crawl_type_t crawl,
+ process_entry_cbk_t process_entry, void *op_data,
+ gf_boolean_t exclusive, int crawl_flags,
+ afr_crawl_done_cbk_t crawl_done)
+{
+ afr_private_t *priv = NULL;
+ call_frame_t *frame = NULL;
+ afr_crawl_data_t *crawl_data = NULL;
+ int ret = 0;
+ int (*crawler) (void*) = NULL;
+
+ priv = this->private;
+
+ frame = create_frame (this, this->ctx->pool);
+ if (!frame)
+ goto out;
+
+ afr_set_lk_owner (frame, this, frame->root);
+ afr_set_low_priority (frame);
+ crawl_data = GF_CALLOC (1, sizeof (*crawl_data),
+ gf_afr_mt_crawl_data_t);
+ if (!crawl_data)
+ goto out;
+ crawl_data->process_entry = process_entry;
+ crawl_data->child = idx;
+ crawl_data->pid = frame->root->pid;
+ crawl_data->crawl = crawl;
+ crawl_data->op_data = op_data;
+ crawl_data->crawl_flags = crawl_flags;
+ gf_log (this->name, GF_LOG_DEBUG, "starting crawl %d for %s",
+ crawl_data->crawl, priv->children[idx]->name);
+
+ if (exclusive)
+ crawler = afr_dir_exclusive_crawl;
+ else
+ crawler = afr_dir_crawl;
+ ret = synctask_new (this->ctx->env, crawler,
+ crawl_done, frame, crawl_data);
+ if (ret)
+ gf_log (this->name, GF_LOG_ERROR, "afr crawl failed for child"
+ " %d with ret %d", idx, ret);
+out:
+ return;
+}
+
+void
+afr_build_root_loc (xlator_t *this, loc_t *loc)
+{
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ loc->path = gf_strdup ("/");
+ loc->name = "";
+ loc->inode = inode_ref (priv->root_inode);
+ uuid_copy (loc->gfid, loc->inode->gfid);
+}
+
+int
+afr_set_root_gfid (dict_t *dict)
+{
+ uuid_t gfid;
+ int ret = 0;
+
+ memset (gfid, 0, 16);
+ gfid[15] = 1;
+
+ ret = afr_set_dict_gfid (dict, gfid);
+
+ return ret;
+}
diff --git a/xlators/cluster/afr/src/afr-self-heald.h b/xlators/cluster/afr/src/afr-self-heald.h
new file mode 100644
index 000000000..e0c083754
--- /dev/null
+++ b/xlators/cluster/afr/src/afr-self-heald.h
@@ -0,0 +1,65 @@
+/*
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+*/
+
+#ifndef __AFR_SELF_HEALD_H__
+#define __AFR_SELF_HEALD_H__
+#include "xlator.h"
+
+#define IS_ROOT_PATH(path) (!strcmp (path, "/"))
+#define IS_ENTRY_CWD(entry) (!strcmp (entry, "."))
+#define IS_ENTRY_PARENT(entry) (!strcmp (entry, ".."))
+#define AFR_ALL_CHILDREN -1
+
+typedef struct afr_crawl_data_ {
+ int child;
+ pid_t pid;
+ afr_crawl_type_t crawl;
+ xlator_t *readdir_xl;
+ void *op_data;
+ int crawl_flags;
+ int (*process_entry) (xlator_t *this, struct afr_crawl_data_ *crawl_data,
+ gf_dirent_t *entry, loc_t *child, loc_t *parent,
+ struct iatt *iattr);
+} afr_crawl_data_t;
+
+typedef struct crawl_event_stats_ {
+ uint64_t healed_count;
+ uint64_t split_brain_count;
+ uint64_t heal_failed_count;
+ char *start_time_str;
+ char *end_time_str;
+ char *crawl_type;
+ gf_boolean_t crawl_inprogress;
+} shd_crawl_event_t;
+
+void _destroy_crawl_event_data (void *data);
+void _destroy_shd_event_data (void *data);
+
+typedef int (*process_entry_cbk_t) (xlator_t *this, afr_crawl_data_t *crawl_data,
+ gf_dirent_t *entry, loc_t *child, loc_t *parent,
+ struct iatt *iattr);
+
+void afr_build_root_loc (xlator_t *this, loc_t *loc);
+
+int afr_set_root_gfid (dict_t *dict);
+
+void
+afr_proactive_self_heal (void *data);
+
+int
+afr_xl_op (xlator_t *this, dict_t *input, dict_t *output);
+
+/*
+ * In addition to its self-heal use, this is used to find a local default
+ * read_child.
+ */
+int
+afr_local_pathinfo (char *pathinfo, gf_boolean_t *local);
+#endif /* __AFR_SELF_HEALD_H__ */
diff --git a/xlators/cluster/afr/src/afr-transaction.c b/xlators/cluster/afr/src/afr-transaction.c
index 3cdd1829d..20306e469 100644
--- a/xlators/cluster/afr/src/afr-transaction.c
+++ b/xlators/cluster/afr/src/afr-transaction.c
@@ -1,25 +1,17 @@
/*
- Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include "dict.h"
#include "byte-order.h"
#include "common-utils.h"
+#include "timer.h"
#include "afr.h"
#include "afr-transaction.h"
@@ -32,58 +24,85 @@
of RENAME */
#define LOCKED_LOWER 0x2 /* for lower_path of RENAME */
-
afr_fd_ctx_t *
-afr_fd_ctx_get (fd_t *fd, xlator_t *this)
+__afr_fd_ctx_get (fd_t *fd, xlator_t *this)
{
uint64_t ctx = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
int ret = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
- ret = fd_ctx_get (fd, this, &ctx);
+ priv = this->private;
- if (ret < 0)
- goto out;
+ ret = __fd_ctx_get (fd, this, &ctx);
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+ if (ret < 0 && fd_is_anonymous (fd)) {
+ ret = __afr_fd_ctx_set (this, fd);
+ if (ret < 0)
+ goto out;
+
+ ret = __fd_ctx_get (fd, this, &ctx);
+ if (ret < 0)
+ goto out;
+ fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+ for (i = 0; i < priv->child_count; i++)
+ fd_ctx->opened_on[i] = AFR_FD_OPENED;
+ }
+
+ fd_ctx = (afr_fd_ctx_t *)(long) ctx;
out:
return fd_ctx;
}
+afr_fd_ctx_t *
+afr_fd_ctx_get (fd_t *fd, xlator_t *this)
+{
+ afr_fd_ctx_t *fd_ctx = NULL;
+
+ LOCK(&fd->lock);
+ {
+ fd_ctx = __afr_fd_ctx_get (fd, this);
+ }
+ UNLOCK(&fd->lock);
+
+ return fd_ctx;
+}
+
+
static void
-afr_pid_save (call_frame_t *frame)
+afr_save_lk_owner (call_frame_t *frame)
{
afr_local_t * local = NULL;
local = frame->local;
- local->saved_pid = frame->root->pid;
+ local->saved_lk_owner = frame->root->lk_owner;
}
static void
-afr_pid_restore (call_frame_t *frame)
+afr_restore_lk_owner (call_frame_t *frame)
{
afr_local_t * local = NULL;
local = frame->local;
- frame->root->pid = local->saved_pid;
+ frame->root->lk_owner = local->saved_lk_owner;
}
-
static void
__mark_all_pending (int32_t *pending[], int child_count,
afr_transaction_type type)
{
- int i;
- int j;
+ int i = 0;
+ int j = 0;
for (i = 0; i < child_count; i++) {
j = afr_index_for_transaction_type (type);
- pending[i][j] = hton32 (1);
+ pending[i][j] = hton32 (1);
}
}
@@ -92,11 +111,11 @@ static void
__mark_child_dead (int32_t *pending[], int child_count, int child,
afr_transaction_type type)
{
- int j;
+ int j = 0;
j = afr_index_for_transaction_type (type);
- pending[child][j] = 0;
+ pending[child][j] = 0;
}
@@ -126,140 +145,127 @@ out:
return;
}
-
static void
-__mark_pre_op_undone_on_fd (call_frame_t *frame, xlator_t *this, int child_index)
+__mark_non_participant_children (int32_t *pending[], int child_count,
+ unsigned char *participants,
+ afr_transaction_type type)
{
- afr_local_t *local = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
-
- local = frame->local;
+ int i = 0;
+ int j = 0;
- if (!local->fd)
- return;
-
- fd_ctx = afr_fd_ctx_get (local->fd, this);
-
- if (!fd_ctx)
- goto out;
-
- LOCK (&local->fd->lock);
- {
- if (local->transaction.type == AFR_DATA_TRANSACTION)
- fd_ctx->pre_op_done[child_index]--;
- }
- UNLOCK (&local->fd->lock);
-out:
- return;
-}
-
-
-static void
-__mark_down_children (int32_t *pending[], int child_count,
- unsigned char *child_up, afr_transaction_type type)
-{
- int i;
- int j;
-
- for (i = 0; i < child_count; i++) {
- j = afr_index_for_transaction_type (type);
-
- if (!child_up[i])
- pending[i][j] = 0;
+ j = afr_index_for_transaction_type (type);
+ for (i = 0; i < child_count; i++) {
+ if (!participants[i])
+ pending[i][j] = 0;
}
}
-static void
+void
__mark_all_success (int32_t *pending[], int child_count,
afr_transaction_type type)
{
- int i;
+ int i;
int j;
- for (i = 0; i < child_count; i++) {
+ for (i = 0; i < child_count; i++) {
j = afr_index_for_transaction_type (type);
- pending[i][j] = hton32 (-1);
+ pending[i][j] = hton32 (-1);
}
}
-
-static int
-__changelog_enabled (afr_private_t *priv, afr_transaction_type type)
+void
+_set_all_child_errno (int *child_errno, unsigned int child_count)
{
- int ret = 0;
-
- switch (type) {
- case AFR_DATA_TRANSACTION:
- if (priv->data_change_log)
- ret = 1;
-
- break;
+ int i = 0;
- case AFR_METADATA_TRANSACTION:
- if (priv->metadata_change_log)
- ret = 1;
+ for (i = 0; i < child_count; i++)
+ if (child_errno[i] == 0)
+ child_errno[i] = ENOTCONN;
+}
- break;
+void
+afr_transaction_perform_fop (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ fd_t *fd = NULL;
- case AFR_ENTRY_TRANSACTION:
- case AFR_ENTRY_RENAME_TRANSACTION:
- if (priv->entry_change_log)
- ret = 1;
+ local = frame->local;
+ priv = this->private;
+ fd = local->fd;
- break;
- }
+ __mark_all_success (local->pending, priv->child_count,
+ local->transaction.type);
- return ret;
+ _set_all_child_errno (local->child_errno, priv->child_count);
+
+ /* Perform fops with the lk-owner from top xlator.
+ * Eg: lk-owner of posix-lk and flush should be same,
+ * flush cant clear the posix-lks without that lk-owner.
+ */
+ afr_save_lk_owner (frame);
+ frame->root->lk_owner =
+ local->transaction.main_frame->root->lk_owner;
+
+
+ /* The wake up needs to happen independent of
+ what type of fop arrives here. If it was
+ a write, then it has already inherited the
+ lock and changelog. If it was not a write,
+ then the presumption of the optimization (of
+ optimizing for successive write operations)
+ fails.
+ */
+ if (fd)
+ afr_delayed_changelog_wake_up (this, fd);
+ local->transaction.fop (frame, this);
}
static int
-__changelog_needed_pre_op (call_frame_t *frame, xlator_t *this)
+__changelog_enabled (afr_private_t *priv, afr_transaction_type type)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
+ int ret = 0;
- int op_ret = 0;
+ switch (type) {
+ case AFR_DATA_TRANSACTION:
+ if (priv->data_change_log)
+ ret = 1;
- priv = this->private;
- local = frame->local;
+ break;
- if (__changelog_enabled (priv, local->transaction.type)) {
- switch (local->op) {
+ case AFR_METADATA_TRANSACTION:
+ if (priv->metadata_change_log)
+ ret = 1;
- case GF_FOP_WRITE:
- case GF_FOP_FTRUNCATE:
- op_ret = 1;
- break;
+ break;
- case GF_FOP_FLUSH:
- op_ret = 0;
- break;
+ case AFR_ENTRY_TRANSACTION:
+ case AFR_ENTRY_RENAME_TRANSACTION:
+ if (priv->entry_change_log)
+ ret = 1;
- default:
- op_ret = 1;
- }
- }
+ break;
+ }
- return op_ret;
+ return ret;
}
static int
-__changelog_needed_post_op (call_frame_t *frame, xlator_t *this)
+__fop_changelog_needed (call_frame_t *frame, xlator_t *this)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ int op_ret = 0;
+ afr_transaction_type type = -1;
- int op_ret = 0;
- afr_transaction_type type = -1;
-
- priv = this->private;
- local = frame->local;
- type = local->transaction.type;
+ priv = this->private;
+ local = frame->local;
+ type = local->transaction.type;
- if (__changelog_enabled (priv, type)) {
+ if (__changelog_enabled (priv, type)) {
switch (local->op) {
case GF_FOP_WRITE:
@@ -276,285 +282,455 @@ __changelog_needed_post_op (call_frame_t *frame, xlator_t *this)
}
}
- return op_ret;
+ return op_ret;
}
-
-static int
-afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending)
+int
+afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending,
+ int child, afr_xattrop_type_t op)
{
- int i;
+ int i = 0;
int ret = 0;
+ if (op == LOCAL_FIRST) {
+ ret = dict_set_static_bin (xattr, priv->pending_key[child],
+ pending[child],
+ AFR_NUM_CHANGE_LOGS * sizeof (int32_t));
+ if (ret)
+ goto out;
+ }
for (i = 0; i < priv->child_count; i++) {
+ if (i == child)
+ continue;
ret = dict_set_static_bin (xattr, priv->pending_key[i],
- pending[i], 3 * sizeof (int32_t));
+ pending[i],
+ AFR_NUM_CHANGE_LOGS * sizeof (int32_t));
/* 3 = data+metadata+entry */
if (ret < 0)
goto out;
}
-
-out:
- return ret;
-}
-
-
-static int
-afr_set_piggyback_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending,
- afr_transaction_type type)
-{
- int i;
- int ret = 0;
- int *arr = NULL;
- int index = 0;
-
- index = afr_index_for_transaction_type (type);
-
- for (i = 0; i < priv->child_count; i++) {
- arr = GF_CALLOC (3 * sizeof (int32_t), priv->child_count,
- gf_afr_mt_char);
- if (!arr) {
- ret = -1;
- goto out;
- }
-
- memcpy (arr, pending[i], 3 * sizeof (int32_t));
-
- arr[index]++;
-
- ret = dict_set_bin (xattr, priv->pending_key[i],
- arr, 3 * sizeof (int32_t));
- /* 3 = data+metadata+entry */
-
- if (ret < 0)
+ if (op == LOCAL_LAST) {
+ ret = dict_set_static_bin (xattr, priv->pending_key[child],
+ pending[child],
+ AFR_NUM_CHANGE_LOGS * sizeof (int32_t));
+ if (ret)
goto out;
}
-
out:
return ret;
}
-
int
afr_lock_server_count (afr_private_t *priv, afr_transaction_type type)
{
- int ret = 0;
+ int ret = 0;
- switch (type) {
- case AFR_DATA_TRANSACTION:
- ret = priv->child_count;
- break;
+ switch (type) {
+ case AFR_DATA_TRANSACTION:
+ ret = priv->child_count;
+ break;
- case AFR_METADATA_TRANSACTION:
- ret = priv->child_count;
- break;
+ case AFR_METADATA_TRANSACTION:
+ ret = priv->child_count;
+ break;
- case AFR_ENTRY_TRANSACTION:
- case AFR_ENTRY_RENAME_TRANSACTION:
- ret = priv->child_count;
- break;
- }
+ case AFR_ENTRY_TRANSACTION:
+ case AFR_ENTRY_RENAME_TRANSACTION:
+ ret = priv->child_count;
+ break;
+ }
- return ret;
+ return ret;
}
/* {{{ pending */
int32_t
afr_changelog_post_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xattr)
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
{
afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int child_index = 0;
-
- int call_count = -1;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int call_count = -1;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
int_lock = &local->internal_lock;
- child_index = (long) cookie;
-
- if (op_ret == 1) {
- }
-
- if (op_ret == 0) {
- __mark_pre_op_undone_on_fd (frame, this, child_index);
+ LOCK (&frame->lock);
+ {
+ call_count = --local->call_count;
}
+ UNLOCK (&frame->lock);
- LOCK (&frame->lock);
- {
- call_count = --local->call_count;
- }
- UNLOCK (&frame->lock);
+ if (call_count == 0) {
+ if (local->transaction.resume_stub) {
+ call_resume (local->transaction.resume_stub);
+ local->transaction.resume_stub = NULL;
+ }
- if (call_count == 0) {
if (afr_lock_server_count (priv, local->transaction.type) == 0) {
local->transaction.done (frame, this);
} else {
int_lock->lock_cbk = local->transaction.done;
afr_unlock (frame, this);
}
- }
+ }
- return 0;
+ return 0;
}
void
-afr_update_read_child (call_frame_t *frame, xlator_t *this, inode_t *inode,
- afr_transaction_type type)
+afr_transaction_rm_stale_children (call_frame_t *frame, xlator_t *this,
+ inode_t *inode, afr_transaction_type type)
{
- int curr_read_child = -1;
- int new_read_child = -1;
+ int i = -1;
+ int count = 0;
+ int read_child = -1;
afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int **pending = NULL;
- int idx = 0;
+ afr_local_t *local = NULL;
+ int **pending = NULL;
+ int idx = 0;
+ int32_t *stale_children = NULL;
+ int32_t *fresh_children = NULL;
+ gf_boolean_t rm_stale_children = _gf_false;
idx = afr_index_for_transaction_type (type);
priv = this->private;
local = frame->local;
- curr_read_child = afr_read_child (this, inode);
pending = local->pending;
- if (pending[curr_read_child][idx] != 0)
- return;
-
- /* need to set new read_child */
- for (new_read_child = 0; new_read_child < priv->child_count;
- new_read_child++) {
+ if (local->op_ret < 0)
+ goto out;
+ fresh_children = local->fresh_children;
+ read_child = afr_inode_get_read_ctx (this, inode, fresh_children);
+ if (read_child < 0) {
+ gf_log (this->name, GF_LOG_DEBUG, "Possible split-brain "
+ "for %s", uuid_utoa (inode->gfid));
+ goto out;
+ }
- if (!priv->child_up[new_read_child])
- /* child is down */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!afr_is_child_present (fresh_children,
+ priv->child_count, i))
continue;
-
- if (pending[new_read_child][idx] == 0)
- /* op just failed */
+ if (pending[i][idx])
continue;
+ /* child is down or op failed on it */
+ if (!stale_children)
+ stale_children = afr_children_create (priv->child_count);
+ if (!stale_children)
+ goto out;
- break;
+ rm_stale_children = _gf_true;
+ stale_children[count++] = i;
+ gf_log (this->name, GF_LOG_DEBUG, "Removing stale child "
+ "%d for %s", i, uuid_utoa (inode->gfid));
}
- if (new_read_child == priv->child_count)
- /* all children uneligible. leave as-is */
+ if (!rm_stale_children)
+ goto out;
+
+ afr_inode_rm_stale_children (this, inode, stale_children);
+out:
+ GF_FREE (stale_children);
+ return;
+}
+
+afr_inodelk_t*
+afr_get_inodelk (afr_internal_lock_t *int_lock, char *dom)
+{
+ afr_inodelk_t *inodelk = NULL;
+ int i = 0;
+
+ for (i = 0; int_lock->inodelk[i].domain; i++) {
+ inodelk = &int_lock->inodelk[i];
+ if (strcmp (dom, inodelk->domain) == 0)
+ return inodelk;
+ }
+ return NULL;
+}
+
+unsigned char*
+afr_locked_nodes_get (afr_transaction_type type, afr_internal_lock_t *int_lock)
+{
+ unsigned char *locked_nodes = NULL;
+ afr_inodelk_t *inodelk = NULL;
+ switch (type) {
+ case AFR_DATA_TRANSACTION:
+ case AFR_METADATA_TRANSACTION:
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+ locked_nodes = inodelk->locked_nodes;
+ break;
+
+ case AFR_ENTRY_TRANSACTION:
+ case AFR_ENTRY_RENAME_TRANSACTION:
+ /*Because same set of subvols participate in all lockee
+ * entities*/
+ locked_nodes = int_lock->lockee[0].locked_nodes;
+ break;
+ }
+ return locked_nodes;
+}
+
+int
+afr_changelog_pre_op_call_count (afr_transaction_type type,
+ afr_internal_lock_t *int_lock,
+ unsigned int child_count)
+{
+ int call_count = 0;
+ unsigned char *locked_nodes = NULL;
+
+ locked_nodes = afr_locked_nodes_get (type, int_lock);
+ GF_ASSERT (locked_nodes);
+
+ call_count = afr_locked_children_count (locked_nodes, child_count);
+ if (type == AFR_ENTRY_RENAME_TRANSACTION)
+ call_count *= 2;
+
+ return call_count;
+}
+
+int
+afr_changelog_post_op_call_count (afr_transaction_type type,
+ unsigned char *pre_op,
+ unsigned int child_count)
+{
+ int call_count = 0;
+
+ call_count = afr_pre_op_done_children_count (pre_op, child_count);
+ if (type == AFR_ENTRY_RENAME_TRANSACTION)
+ call_count *= 2;
+
+ return call_count;
+}
+
+void
+afr_compute_txn_changelog (afr_local_t *local, afr_private_t *priv)
+{
+ int i = 0;
+ int index = 0;
+ int32_t postop = 0;
+ int32_t preop = 1;
+ int32_t **txn_changelog = NULL;
+
+ txn_changelog = local->transaction.txn_changelog;
+ index = afr_index_for_transaction_type (local->transaction.type);
+ for (i = 0; i < priv->child_count; i++) {
+ postop = ntoh32 (local->pending[i][index]);
+ txn_changelog[i][index] = hton32 (postop + preop);
+ }
+}
+
+afr_xattrop_type_t
+afr_get_postop_xattrop_type (int32_t **pending, int optimized, int child,
+ afr_transaction_type type)
+{
+ int index = 0;
+ afr_xattrop_type_t op = LOCAL_LAST;
+
+ index = afr_index_for_transaction_type (type);
+ if (optimized && !pending[child][index])
+ op = LOCAL_FIRST;
+ return op;
+}
+
+void
+afr_set_postop_dict (afr_local_t *local, xlator_t *this, dict_t *xattr,
+ int optimized, int child)
+{
+ int32_t **txn_changelog = NULL;
+ int32_t **changelog = NULL;
+ afr_private_t *priv = NULL;
+ int ret = 0;
+ afr_xattrop_type_t op = LOCAL_LAST;
+
+ priv = this->private;
+ txn_changelog = local->transaction.txn_changelog;
+ op = afr_get_postop_xattrop_type (local->pending, optimized, child,
+ local->transaction.type);
+ if (optimized)
+ changelog = txn_changelog;
+ else
+ changelog = local->pending;
+ ret = afr_set_pending_dict (priv, xattr, changelog, child, op);
+ if (ret < 0)
+ gf_log (this->name, GF_LOG_INFO,
+ "failed to set pending entry");
+}
+
+
+gf_boolean_t
+afr_txn_nothing_failed (call_frame_t *frame, xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int index = -1;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ index = afr_index_for_transaction_type (local->transaction.type);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->pending[i][index] == 0)
+ return _gf_false;
+ }
+
+ return _gf_true;
+}
+
+static void
+afr_dir_fop_handle_all_fop_failures (call_frame_t *frame)
+{
+ xlator_t *this = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ this = frame->this;
+ local = frame->local;
+ priv = this->private;
+
+ if ((local->transaction.type != AFR_ENTRY_TRANSACTION) &&
+ (local->transaction.type != AFR_ENTRY_RENAME_TRANSACTION))
return;
- afr_set_read_child (this, inode, new_read_child);
+ if (local->op_ret >= 0)
+ goto out;
+
+ __mark_all_success (local->pending, priv->child_count,
+ local->transaction.type);
+out:
+ return;
}
+static void
+afr_data_handle_quota_errors (call_frame_t *frame, xlator_t *this)
+{
+ int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ gf_boolean_t all_quota_failures = _gf_false;
+
+ local = frame->local;
+ priv = this->private;
+ if (local->transaction.type != AFR_DATA_TRANSACTION)
+ return;
+ /*
+ * Idea is to not leave the file in FOOL-FOOL scenario in case on
+ * all the bricks data transaction failed with EDQUOT to avoid
+ * increasing un-necessary load of self-heals in the system.
+ */
+ all_quota_failures = _gf_true;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i] &&
+ (local->child_errno[i] != EDQUOT)) {
+ all_quota_failures = _gf_false;
+ break;
+ }
+ }
+ if (all_quota_failures)
+ __mark_all_success (local->pending, priv->child_count,
+ local->transaction.type);
+}
-int
-afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
+int
+afr_changelog_post_op_now (call_frame_t *frame, xlator_t *this)
{
- afr_private_t * priv = this->private;
+ afr_private_t * priv = this->private;
afr_internal_lock_t *int_lock = NULL;
- int ret = 0;
- int i = 0;
- int call_count = 0;
+ int i = 0;
+ int call_count = 0;
- afr_local_t * local = NULL;
+ afr_local_t * local = NULL;
afr_fd_ctx_t *fdctx = NULL;
- dict_t **xattr = NULL;
+ dict_t **xattr = NULL;
int piggyback = 0;
- int index = 0;
int nothing_failed = 1;
- local = frame->local;
+ local = frame->local;
int_lock = &local->internal_lock;
- __mark_down_children (local->pending, priv->child_count,
- local->child_up, local->transaction.type);
+ __mark_non_participant_children (local->pending, priv->child_count,
+ local->transaction.pre_op,
+ local->transaction.type);
+
+ afr_data_handle_quota_errors (frame, this);
+ afr_dir_fop_handle_all_fop_failures (frame);
if (local->fd)
- afr_update_read_child (frame, this, local->fd->inode,
- local->transaction.type);
+ afr_transaction_rm_stale_children (frame, this,
+ local->fd->inode,
+ local->transaction.type);
xattr = alloca (priv->child_count * sizeof (*xattr));
memset (xattr, 0, (priv->child_count * sizeof (*xattr)));
- for (i = 0; i < priv->child_count; i++) {
- xattr[i] = get_new_dict ();
- dict_ref (xattr[i]);
- }
-
- call_count = afr_up_children_count (priv->child_count, local->child_up);
-
- if (local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
- call_count *= 2;
+ for (i = 0; i < priv->child_count; i++) {
+ xattr[i] = dict_new ();
}
- local->call_count = call_count;
+ call_count = afr_changelog_post_op_call_count (local->transaction.type,
+ local->transaction.pre_op,
+ priv->child_count);
+ local->call_count = call_count;
if (local->fd)
fdctx = afr_fd_ctx_get (local->fd, this);
- if (call_count == 0) {
- /* no child is up */
- for (i = 0; i < priv->child_count; i++) {
- dict_unref (xattr[i]);
- }
-
+ if (call_count == 0) {
+ /* no child is up */
int_lock->lock_cbk = local->transaction.done;
- afr_unlock (frame, this);
- return 0;
- }
-
- /* check if something has failed, to handle piggybacking */
- nothing_failed = 1;
- index = afr_index_for_transaction_type (local->transaction.type);
- for (i = 0; i < priv->child_count; i++) {
- if (local->pending[i][index] == 0) {
- nothing_failed = 0;
- break;
- }
+ afr_unlock (frame, this);
+ goto out;
}
- for (i = 0; i < priv->child_count; i++) {
- if (!local->child_up[i])
- continue;
+ nothing_failed = afr_txn_nothing_failed (frame, this);
- ret = afr_set_pending_dict (priv, xattr[i],
- local->pending);
-
- if (ret < 0)
- gf_log (this->name, GF_LOG_DEBUG,
- "failed to set pending entry");
+ afr_compute_txn_changelog (local , priv);
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.pre_op[i])
+ continue;
+ if (local->transaction.type != AFR_DATA_TRANSACTION)
+ afr_set_postop_dict (local, this, xattr[i],
+ local->optimistic_change_log, i);
switch (local->transaction.type) {
case AFR_DATA_TRANSACTION:
{
if (!fdctx) {
+ afr_set_postop_dict (local, this, xattr[i],
+ 0, i);
STACK_WIND (frame, afr_changelog_post_op_cbk,
priv->children[i],
priv->children[i]->fops->xattrop,
&local->loc,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
break;
}
- LOCK (&local->fd->lock);
- {
- piggyback = 0;
- if (fdctx->pre_op_piggyback[i]) {
- fdctx->pre_op_piggyback[i]--;
- piggyback = 1;
- }
- }
- UNLOCK (&local->fd->lock);
+ /* local->transaction.postop_piggybacked[] was
+ precomputed in is_piggyback_postop() when called from
+ afr_changelog_post_op_safe()
+ */
- if (piggyback && !nothing_failed)
- ret = afr_set_piggyback_dict (priv, xattr[i],
- local->pending,
- local->transaction.type);
+ piggyback = 0;
+ if (local->transaction.postop_piggybacked[i])
+ piggyback = 1;
+
+ afr_set_postop_dict (local, this, xattr[i],
+ piggyback, i);
if (nothing_failed && piggyback) {
afr_changelog_post_op_cbk (frame, (void *)(long)i,
- this, 1, 0, xattr[i]);
+ this, 1, 0, xattr[i], NULL);
} else {
STACK_WIND_COOKIE (frame,
afr_changelog_post_op_cbk,
@@ -562,35 +738,52 @@ afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->fxattrop,
local->fd,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
}
}
break;
case AFR_METADATA_TRANSACTION:
{
+ if (nothing_failed && local->optimistic_change_log) {
+ afr_changelog_post_op_cbk (frame, (void *)(long)i,
+ this, 1, 0, xattr[i],
+ NULL);
+ break;
+ }
+
if (local->fd)
STACK_WIND (frame, afr_changelog_post_op_cbk,
priv->children[i],
priv->children[i]->fops->fxattrop,
local->fd,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
else
STACK_WIND (frame, afr_changelog_post_op_cbk,
priv->children[i],
priv->children[i]->fops->xattrop,
&local->loc,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
}
break;
case AFR_ENTRY_RENAME_TRANSACTION:
{
- STACK_WIND_COOKIE (frame, afr_changelog_post_op_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->xattrop,
- &local->transaction.new_parent_loc,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ if (nothing_failed && local->optimistic_change_log) {
+ afr_changelog_post_op_cbk (frame, (void *)(long)i,
+ this, 1, 0, xattr[i],
+ NULL);
+ } else {
+ STACK_WIND_COOKIE (frame, afr_changelog_post_op_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->xattrop,
+ &local->transaction.new_parent_loc,
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
+ }
call_count--;
}
@@ -602,167 +795,157 @@ afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
value
*/
- ret = afr_set_pending_dict (priv, xattr[i],
- local->pending);
-
- if (ret < 0)
- gf_log (this->name, GF_LOG_DEBUG,
- "failed to set pending entry");
+ afr_set_postop_dict (local, this, xattr[i],
+ local->optimistic_change_log, i);
/* fall through */
case AFR_ENTRY_TRANSACTION:
{
+ if (nothing_failed && local->optimistic_change_log) {
+ afr_changelog_post_op_cbk (frame, (void *)(long)i,
+ this, 1, 0, xattr[i],
+ NULL);
+ break;
+ }
+
if (local->fd)
STACK_WIND (frame, afr_changelog_post_op_cbk,
priv->children[i],
priv->children[i]->fops->fxattrop,
local->fd,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
else
STACK_WIND (frame, afr_changelog_post_op_cbk,
priv->children[i],
priv->children[i]->fops->xattrop,
&local->transaction.parent_loc,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
}
break;
}
if (!--call_count)
break;
- }
+ }
+out:
for (i = 0; i < priv->child_count; i++) {
dict_unref (xattr[i]);
}
- return 0;
+ return 0;
}
int32_t
afr_changelog_pre_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xattr)
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
{
- afr_local_t * local = NULL;
- afr_private_t * priv = this->private;
- loc_t * loc = NULL;
+ afr_local_t * local = NULL;
+ afr_private_t * priv = this->private;
+ int call_count = -1;
+ int child_index = (long) cookie;
- int call_count = -1;
- int child_index = (long) cookie;
-
- local = frame->local;
- loc = &local->loc;
-
- LOCK (&frame->lock);
- {
- if (op_ret == 1) {
- /* special op_ret for piggyback */
- }
+ local = frame->local;
- if (op_ret == 0) {
+ LOCK (&frame->lock);
+ {
+ switch (op_ret) {
+ case 0:
__mark_pre_op_done_on_fd (frame, this, child_index);
+ //fallthrough we need to mark the pre_op
+ case 1:
+ local->transaction.pre_op[child_index] = 1;
+ /* special op_ret for piggyback */
+ break;
+ case -1:
+ if (op_errno == ENOTSUP) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "xattrop not supported by %s",
+ priv->children[child_index]->name);
+ local->op_ret = -1;
+
+ } else if (!child_went_down (op_ret, op_errno)) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "xattrop failed on child %s: %s",
+ priv->children[child_index]->name,
+ strerror (op_errno));
+ }
+ local->op_errno = op_errno;
+ break;
}
- if (op_ret == -1) {
- local->child_up[child_index] = 0;
-
- if (op_errno == ENOTSUP) {
- gf_log (this->name, GF_LOG_ERROR,
- "xattrop not supported by %s",
- priv->children[child_index]->name);
- local->op_ret = -1;
-
- } else if (!child_went_down (op_ret, op_errno)) {
- gf_log (this->name, GF_LOG_ERROR,
- "xattrop failed on child %s: %s",
- priv->children[child_index]->name,
- strerror (op_errno));
- }
- local->op_errno = op_errno;
- }
-
- call_count = --local->call_count;
- }
- UNLOCK (&frame->lock);
-
- if (call_count == 0) {
- if ((local->op_ret == -1) &&
- (local->op_errno == ENOTSUP)) {
- local->transaction.resume (frame, this);
- } else {
- __mark_all_success (local->pending, priv->child_count,
- local->transaction.type);
-
- afr_pid_restore (frame);
+ call_count = --local->call_count;
+ }
+ UNLOCK (&frame->lock);
- local->transaction.fop (frame, this);
- }
- }
+ if (call_count == 0) {
+ if ((local->op_ret == -1) &&
+ (local->op_errno == ENOTSUP)) {
+ local->transaction.resume (frame, this);
+ } else {
+ afr_transaction_perform_fop (frame, this);
+ }
+ }
- return 0;
+ return 0;
}
-
int
afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
{
- afr_private_t * priv = this->private;
- int i = 0;
- int ret = 0;
- int call_count = 0;
- dict_t **xattr = NULL;
+ afr_private_t * priv = this->private;
+ int i = 0;
+ int ret = 0;
+ int call_count = 0;
+ dict_t **xattr = NULL;
afr_fd_ctx_t *fdctx = NULL;
-
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
int piggyback = 0;
+ afr_internal_lock_t *int_lock = NULL;
+ unsigned char *locked_nodes = NULL;
- local = frame->local;
+ local = frame->local;
+ int_lock = &local->internal_lock;
xattr = alloca (priv->child_count * sizeof (*xattr));
memset (xattr, 0, (priv->child_count * sizeof (*xattr)));
- for (i = 0; i < priv->child_count; i++) {
- xattr[i] = get_new_dict ();
- dict_ref (xattr[i]);
+ for (i = 0; i < priv->child_count; i++) {
+ xattr[i] = dict_new ();
}
- call_count = afr_up_children_count (priv->child_count,
- local->child_up);
-
- if (local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
- call_count *= 2;
- }
-
- if (call_count == 0) {
- /* no child is up */
- for (i = 0; i < priv->child_count; i++) {
- dict_unref (xattr[i]);
- }
-
+ call_count = afr_changelog_pre_op_call_count (local->transaction.type,
+ int_lock,
+ priv->child_count);
+ if (call_count == 0) {
local->internal_lock.lock_cbk =
local->transaction.done;
- afr_unlock (frame, this);
- return 0;
- }
+ afr_unlock (frame, this);
+ goto out;
+ }
- local->call_count = call_count;
+ local->call_count = call_count;
- __mark_all_pending (local->pending, priv->child_count,
+ __mark_all_pending (local->pending, priv->child_count,
local->transaction.type);
if (local->fd)
fdctx = afr_fd_ctx_get (local->fd, this);
- for (i = 0; i < priv->child_count; i++) {
- if (!local->child_up[i])
+ locked_nodes = afr_locked_nodes_get (local->transaction.type, int_lock);
+ for (i = 0; i < priv->child_count; i++) {
+ if (!locked_nodes[i])
continue;
- ret = afr_set_pending_dict (priv, xattr[i],
- local->pending);
+ ret = afr_set_pending_dict (priv, xattr[i], local->pending,
+ i, LOCAL_FIRST);
if (ret < 0)
- gf_log (this->name, GF_LOG_DEBUG,
+ gf_log (this->name, GF_LOG_INFO,
"failed to set pending entry");
@@ -776,7 +959,8 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->xattrop,
&(local->loc),
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
break;
}
@@ -793,9 +977,12 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
}
UNLOCK (&local->fd->lock);
+ afr_set_delayed_post_op (frame, this);
+
if (piggyback)
afr_changelog_pre_op_cbk (frame, (void *)(long)i,
- this, 1, 0, xattr[i]);
+ this, 1, 0, xattr[i],
+ NULL);
else
STACK_WIND_COOKIE (frame,
afr_changelog_pre_op_cbk,
@@ -803,11 +990,19 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->fxattrop,
local->fd,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
}
break;
case AFR_METADATA_TRANSACTION:
{
+ if (local->optimistic_change_log) {
+ afr_changelog_pre_op_cbk (frame, (void *)(long)i,
+ this, 1, 0, xattr[i],
+ NULL);
+ break;
+ }
+
if (local->fd)
STACK_WIND_COOKIE (frame,
afr_changelog_pre_op_cbk,
@@ -815,7 +1010,8 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->fxattrop,
local->fd,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
else
STACK_WIND_COOKIE (frame,
afr_changelog_pre_op_cbk,
@@ -823,19 +1019,27 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->xattrop,
&(local->loc),
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
}
break;
case AFR_ENTRY_RENAME_TRANSACTION:
{
- STACK_WIND_COOKIE (frame,
- afr_changelog_pre_op_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->xattrop,
- &local->transaction.new_parent_loc,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ if (local->optimistic_change_log) {
+ afr_changelog_pre_op_cbk (frame, (void *)(long)i,
+ this, 1, 0, xattr[i],
+ NULL);
+ } else {
+ STACK_WIND_COOKIE (frame,
+ afr_changelog_pre_op_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->xattrop,
+ &local->transaction.new_parent_loc,
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
+ }
call_count--;
}
@@ -849,17 +1053,24 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
value
*/
- ret = afr_set_pending_dict (priv, xattr[i],
- local->pending);
+ ret = afr_set_pending_dict (priv, xattr[i], local->pending,
+ i, LOCAL_FIRST);
if (ret < 0)
- gf_log (this->name, GF_LOG_DEBUG,
+ gf_log (this->name, GF_LOG_INFO,
"failed to set pending entry");
/* fall through */
case AFR_ENTRY_TRANSACTION:
{
+ if (local->optimistic_change_log) {
+ afr_changelog_pre_op_cbk (frame, (void *)(long)i,
+ this, 1, 0, xattr[i],
+ NULL);
+ break;
+ }
+
if (local->fd)
STACK_WIND_COOKIE (frame,
afr_changelog_pre_op_cbk,
@@ -867,7 +1078,8 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->fxattrop,
local->fd,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
else
STACK_WIND_COOKIE (frame,
afr_changelog_pre_op_cbk,
@@ -875,20 +1087,21 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->xattrop,
&local->transaction.parent_loc,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
}
break;
}
if (!--call_count)
break;
- }
-
+ }
+out:
for (i = 0; i < priv->child_count; i++) {
dict_unref (xattr[i]);
}
- return 0;
+ return 0;
}
@@ -902,9 +1115,9 @@ afr_post_blocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
int_lock = &local->internal_lock;
if (int_lock->lock_op_ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
+ gf_log (this->name, GF_LOG_INFO,
"Blocking inodelks failed.");
- local->transaction.done (frame, this);
+ local->transaction.done (frame, this);
} else {
gf_log (this->name, GF_LOG_DEBUG,
@@ -952,9 +1165,9 @@ afr_post_blocking_entrylk_cbk (call_frame_t *frame, xlator_t *this)
int_lock = &local->internal_lock;
if (int_lock->lock_op_ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
+ gf_log (this->name, GF_LOG_INFO,
"Blocking entrylks failed.");
- local->transaction.done (frame, this);
+ local->transaction.done (frame, this);
} else {
gf_log (this->name, GF_LOG_DEBUG,
@@ -1002,9 +1215,9 @@ afr_post_blocking_rename_cbk (call_frame_t *frame, xlator_t *this)
int_lock = &local->internal_lock;
if (int_lock->lock_op_ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
+ gf_log (this->name, GF_LOG_INFO,
"Blocking entrylks failed.");
- local->transaction.done (frame, this);
+ local->transaction.done (frame, this);
} else {
gf_log (this->name, GF_LOG_DEBUG,
@@ -1037,12 +1250,14 @@ int
afr_set_transaction_flock (afr_local_t *local)
{
afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
int_lock = &local->internal_lock;
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
- int_lock->lk_flock.l_len = local->transaction.len;
- int_lock->lk_flock.l_start = local->transaction.start;
- int_lock->lk_flock.l_type = F_WRLCK;
+ inodelk->flock.l_len = local->transaction.len;
+ inodelk->flock.l_start = local->transaction.start;
+ inodelk->flock.l_type = F_WRLCK;
return 0;
}
@@ -1057,24 +1272,25 @@ afr_lock_rec (call_frame_t *frame, xlator_t *this)
int_lock = &local->internal_lock;
int_lock->transaction_lk_type = AFR_TRANSACTION_LK;
+ int_lock->domain = this->name;
- switch (local->transaction.type) {
- case AFR_DATA_TRANSACTION:
- case AFR_METADATA_TRANSACTION:
+ switch (local->transaction.type) {
+ case AFR_DATA_TRANSACTION:
+ case AFR_METADATA_TRANSACTION:
afr_set_transaction_flock (local);
int_lock->lock_cbk = afr_post_nonblocking_inodelk_cbk;
afr_nonblocking_inodelk (frame, this);
- break;
+ break;
case AFR_ENTRY_RENAME_TRANSACTION:
- int_lock->lock_cbk = afr_post_blocking_rename_cbk;
- afr_blocking_lock (frame, this);
+ int_lock->lock_cbk = afr_post_nonblocking_entrylk_cbk;
+ afr_nonblocking_entrylk (frame, this);
break;
- case AFR_ENTRY_TRANSACTION:
+ case AFR_ENTRY_TRANSACTION:
int_lock->lk_basename = local->transaction.basename;
if (&local->transaction.parent_loc)
int_lock->lk_loc = &local->transaction.parent_loc;
@@ -1093,15 +1309,9 @@ afr_lock_rec (call_frame_t *frame, xlator_t *this)
int
afr_lock (call_frame_t *frame, xlator_t *this)
{
- afr_pid_save (frame);
-
- frame->root->pid = (long) frame->root;
-
- afr_set_lk_owner (frame, this);
-
afr_set_lock_number (frame, this);
- return afr_lock_rec (frame, this);
+ return afr_lock_rec (frame, this);
}
@@ -1110,50 +1320,498 @@ afr_lock (call_frame_t *frame, xlator_t *this)
int
afr_internal_lock_finish (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ if (__fop_changelog_needed (frame, this)) {
+ afr_changelog_pre_op (frame, this);
+ } else {
+ afr_transaction_perform_fop (frame, this);
+ }
+
+ return 0;
+}
+
+
+void
+afr_set_delayed_post_op (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ /* call this function from any of the related optimizations
+ which benefit from delaying post op are enabled, namely:
+
+ - changelog piggybacking
+ - eager locking
+ */
+
+ priv = this->private;
+ if (!priv)
+ return;
+
+ if (!priv->post_op_delay_secs)
+ return;
+
+ local = frame->local;
+ if (!local->transaction.eager_lock_on)
+ return;
+
+ if (!local)
+ return;
+
+ if (!local->fd)
+ return;
+
+ if (local->op == GF_FOP_WRITE)
+ local->delayed_post_op = _gf_true;
+}
+
+gf_boolean_t
+afr_are_multiple_fds_opened (inode_t *inode, xlator_t *this)
+{
+ afr_inode_ctx_t *ictx = NULL;
+
+ if (!inode) {
+ /* If false is returned, it may keep on taking eager-lock
+ * which may lead to starvation, so return true to avoid that.
+ */
+ gf_log_callingfn (this->name, GF_LOG_ERROR, "Invalid inode");
+ return _gf_true;
+ }
+ /* Lets say mount1 has eager-lock(full-lock) and after the eager-lock
+ * is taken mount2 opened the same file, it won't be able to
+ * perform any data operations until mount1 releases eager-lock.
+ * To avoid such scenario do not enable eager-lock for this transaction
+ * if open-fd-count is > 1
+ */
+
+ ictx = afr_inode_ctx_get (inode, this);
+ if (!ictx)
+ return _gf_true;
+
+ if (ictx->open_fd_count > 1)
+ return _gf_true;
+
+ return _gf_false;
+}
+
+gf_boolean_t
+afr_any_fops_failed (afr_local_t *local, afr_private_t *priv)
+{
+ if (local->success_count != priv->child_count)
+ return _gf_true;
+ return _gf_false;
+}
+
+gf_boolean_t
+is_afr_delayed_changelog_post_op_needed (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ gf_boolean_t res = _gf_false;
+ afr_private_t *priv = NULL;
priv = this->private;
+
local = frame->local;
+ if (!local)
+ goto out;
- if (__changelog_needed_pre_op (frame, this)) {
- afr_changelog_pre_op (frame, this);
- } else {
- __mark_all_success (local->pending, priv->child_count,
- local->transaction.type);
+ if (!local->delayed_post_op)
+ goto out;
+
+ //Mark pending changelog ASAP
+ if (afr_any_fops_failed (local, priv))
+ goto out;
+
+ if (local->fd && afr_are_multiple_fds_opened (local->fd->inode, this))
+ goto out;
+
+ res = _gf_true;
+out:
+ return res;
+}
+
+
+void
+afr_delayed_changelog_post_op (xlator_t *this, call_frame_t *frame, fd_t *fd,
+ call_stub_t *stub);
+
+void
+afr_delayed_changelog_wake_up_cbk (void *data)
+{
+ fd_t *fd = NULL;
+
+ fd = data;
+
+ afr_delayed_changelog_wake_up (THIS, fd);
+}
+
+
+/*
+ Check if the frame is destined to get optimized away
+ with changelog piggybacking
+*/
+static gf_boolean_t
+is_piggyback_post_op (call_frame_t *frame, fd_t *fd)
+{
+ afr_fd_ctx_t *fdctx = NULL;
+ afr_local_t *local = NULL;
+ gf_boolean_t piggyback = _gf_true;
+ afr_private_t *priv = NULL;
+ int i = 0;
+
+ priv = frame->this->private;
+ local = frame->local;
+ fdctx = afr_fd_ctx_get (fd, frame->this);
+
+ LOCK(&fd->lock);
+ {
+ piggyback = _gf_true;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.pre_op[i])
+ continue;
+ if (fdctx->pre_op_piggyback[i]) {
+ fdctx->pre_op_piggyback[i]--;
+ local->transaction.postop_piggybacked[i] = 1;
+ } else {
+ /* For at least _one_ subvolume we cannot
+ piggyback on the changelog, and have to
+ perform a hard POST-OP and therefore fsync
+ if necesssary
+ */
+ piggyback = _gf_false;
+ GF_ASSERT (fdctx->pre_op_done[i]);
+ fdctx->pre_op_done[i]--;
+ }
+ }
+ }
+ UNLOCK(&fd->lock);
+
+ if (!afr_txn_nothing_failed (frame, frame->this)) {
+ /* something failed in this transaction,
+ we will be performing a hard post-op
+ */
+ return _gf_false;
+ }
+
+ return piggyback;
+}
+
+
+/* SET operation */
+int
+afr_fd_report_unstable_write (xlator_t *this, fd_t *fd)
+{
+ afr_fd_ctx_t *fdctx = NULL;
+
+ fdctx = afr_fd_ctx_get (fd, this);
+
+ LOCK(&fd->lock);
+ {
+ fdctx->witnessed_unstable_write = _gf_true;
+ }
+ UNLOCK(&fd->lock);
+
+ return 0;
+}
+
+/* TEST and CLEAR operation */
+gf_boolean_t
+afr_fd_has_witnessed_unstable_write (xlator_t *this, fd_t *fd)
+{
+ afr_fd_ctx_t *fdctx = NULL;
+ gf_boolean_t witness = _gf_false;
+
+ fdctx = afr_fd_ctx_get (fd, this);
+ if (!fdctx)
+ return _gf_true;
+
+ LOCK(&fd->lock);
+ {
+ if (fdctx->witnessed_unstable_write) {
+ witness = _gf_true;
+ fdctx->witnessed_unstable_write = _gf_false;
+ }
+ }
+ UNLOCK (&fd->lock);
+
+ return witness;
+}
+
+
+int
+afr_changelog_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *pre,
+ struct iatt *post, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ int child_index = (long) cookie;
+ int call_count = -1;
+ afr_local_t *local = NULL;
+
+ priv = this->private;
+ local = frame->local;
- afr_pid_restore (frame);
+ if (afr_fop_failed (op_ret, op_errno)) {
+ /* Failure of fsync() is as good as failure of previous
+ write(). So treat it like one.
+ */
+ gf_log (this->name, GF_LOG_WARNING,
+ "fsync(%s) failed on subvolume %s. Transaction was %s",
+ uuid_utoa (local->fd->inode->gfid),
+ priv->children[child_index]->name,
+ gf_fop_list[local->op]);
- local->transaction.fop (frame, this);
+ afr_transaction_fop_failed (frame, this, child_index);
}
+ call_count = afr_frame_return (frame);
+
+ if (call_count == 0)
+ afr_changelog_post_op_now (frame, this);
+
return 0;
}
int
+afr_changelog_fsync (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ int i = 0;
+ int call_count = 0;
+ afr_private_t *priv = NULL;
+ dict_t *xdata = NULL;
+ GF_UNUSED int ret = -1;
+
+ local = frame->local;
+ priv = this->private;
+
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
+
+ if (!call_count) {
+ /* will go straight to unlock */
+ afr_changelog_post_op_now (frame, this);
+ return 0;
+ }
+
+ local->call_count = call_count;
+
+ xdata = dict_new();
+ if (xdata)
+ ret = dict_set_int32 (xdata, "batch-fsync", 1);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.pre_op[i])
+ continue;
+
+ STACK_WIND_COOKIE (frame, afr_changelog_fsync_cbk,
+ (void *) (long) i, priv->children[i],
+ priv->children[i]->fops->fsync, local->fd,
+ 1, xdata);
+ if (!--call_count)
+ break;
+ }
+
+ if (xdata)
+ dict_unref (xdata);
+
+ return 0;
+}
+
+
+ int
+afr_changelog_post_op_safe (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ if (!local->fd || local->transaction.type != AFR_DATA_TRANSACTION) {
+ afr_changelog_post_op_now (frame, this);
+ return 0;
+ }
+
+ if (is_piggyback_post_op (frame, local->fd)) {
+ /* just detected that this post-op is about to
+ be optimized away as a new write() has
+ already piggybacked on this frame's changelog.
+ */
+ afr_changelog_post_op_now (frame, this);
+ return 0;
+ }
+
+ /* Calling afr_changelog_post_op_now() now will result in
+ issuing ->[f]xattrop().
+
+ Performing a hard POST-OP (->[f]xattrop() FOP) is a more
+ responsible operation that what it might appear on the surface.
+
+ The changelog of a file (in the xattr of the file on the server)
+ stores information (pending count) about the state of the file
+ on the OTHER server. This changelog is blindly trusted, and must
+ therefore be updated in such a way it remains trustworthy. This
+ implies that decrementing the pending count (essentially "clearing
+ the dirty flag") must be done STRICTLY after we are sure that the
+ operation on the other server has reached stable storage.
+
+ While the backend filesystem on that server will eventually flush
+ it to stable storage, we (being in userspace) have no mechanism
+ to get notified when the write became "stable".
+
+ This means we need take matter into our own hands and issue an
+ fsync() EVEN IF THE APPLICATION WAS PERFORMING UNSTABLE WRITES,
+ and get an acknowledgement for it. And we need to wait for the
+ fsync() acknowledgement before initiating the hard POST-OP.
+
+ However if the FD itself was opened in O_SYNC or O_DSYNC then
+ we are already guaranteed that the writes were made stable as
+ part of the FOP itself. The same holds true for NFS stable
+ writes which happen on an anonymous FD with O_DSYNC or O_SYNC
+ flag set in the writev() @flags param. For all other write types,
+ mark a flag in the fdctx whenever an unstable write is witnessed.
+ */
+
+ if (!afr_fd_has_witnessed_unstable_write (this, local->fd)) {
+ afr_changelog_post_op_now (frame, this);
+ return 0;
+ }
+
+ /* Check whether users want durability and perform fsync/post-op
+ * accordingly.
+ */
+ if (priv->ensure_durability) {
+ /* Time to fsync() */
+ afr_changelog_fsync (frame, this);
+ } else {
+ afr_changelog_post_op_now (frame, this);
+ }
+
+ return 0;
+}
+
+
+void
+afr_delayed_changelog_post_op (xlator_t *this, call_frame_t *frame, fd_t *fd,
+ call_stub_t *stub)
+{
+ afr_fd_ctx_t *fd_ctx = NULL;
+ call_frame_t *prev_frame = NULL;
+ struct timespec delta = {0, };
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+
+ priv = this->private;
+
+ fd_ctx = afr_fd_ctx_get (fd, this);
+ if (!fd_ctx)
+ goto out;
+
+ delta.tv_sec = priv->post_op_delay_secs;
+ delta.tv_nsec = 0;
+
+ pthread_mutex_lock (&fd_ctx->delay_lock);
+ {
+ prev_frame = fd_ctx->delay_frame;
+ fd_ctx->delay_frame = NULL;
+ if (fd_ctx->delay_timer)
+ gf_timer_call_cancel (this->ctx, fd_ctx->delay_timer);
+ fd_ctx->delay_timer = NULL;
+ if (!frame)
+ goto unlock;
+ fd_ctx->delay_timer = gf_timer_call_after (this->ctx, delta,
+ afr_delayed_changelog_wake_up_cbk,
+ fd);
+ fd_ctx->delay_frame = frame;
+ }
+unlock:
+ pthread_mutex_unlock (&fd_ctx->delay_lock);
+
+out:
+ if (prev_frame) {
+ local = prev_frame->local;
+ local->transaction.resume_stub = stub;
+ afr_changelog_post_op_safe (prev_frame, this);
+ } else if (stub) {
+ call_resume (stub);
+ }
+}
+
+
+void
+afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+
+ local = frame->local;
+
+ if (is_afr_delayed_changelog_post_op_needed (frame, this))
+ afr_delayed_changelog_post_op (this, frame, local->fd, NULL);
+ else
+ afr_changelog_post_op_safe (frame, this);
+}
+
+
+
+/* Wake up the sleeping/delayed post-op, and also register
+ a stub to have it resumed after this transaction
+ completely finishes.
+
+ The @stub gets saved in @local and gets resumed in
+ afr_local_cleanup()
+ */
+void
+afr_delayed_changelog_wake_resume (xlator_t *this, fd_t *fd, call_stub_t *stub)
+{
+ afr_delayed_changelog_post_op (this, NULL, fd, stub);
+}
+
+
+void
+afr_delayed_changelog_wake_up (xlator_t *this, fd_t *fd)
+{
+ afr_delayed_changelog_post_op (this, NULL, fd, NULL);
+}
+
+
+ int
afr_transaction_resume (call_frame_t *frame, xlator_t *this)
{
afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
+ local = frame->local;
int_lock = &local->internal_lock;
- priv = this->private;
-
- if (__changelog_needed_post_op (frame, this)) {
- afr_changelog_post_op (frame, this);
- } else {
- if (afr_lock_server_count (priv, local->transaction.type) == 0) {
- local->transaction.done (frame, this);
- } else {
+ priv = this->private;
+
+ if (local->transaction.eager_lock_on) {
+ /* We don't need to retain "local" in the
+ fd list anymore, writes to all subvols
+ are finished by now */
+ LOCK (&local->fd->lock);
+ {
+ list_del_init (&local->transaction.eager_locked);
+ }
+ UNLOCK (&local->fd->lock);
+ }
+
+ afr_restore_lk_owner (frame);
+
+ if (__fop_changelog_needed (frame, this)) {
+ afr_changelog_post_op (frame, this);
+ } else {
+ if (afr_lock_server_count (priv, local->transaction.type) == 0) {
+ local->transaction.done (frame, this);
+ } else {
int_lock->lock_cbk = local->transaction.done;
- afr_unlock (frame, this);
- }
- }
+ afr_unlock (frame, this);
+ }
+ }
- return 0;
+ return 0;
}
@@ -1162,38 +1820,144 @@ afr_transaction_resume (call_frame_t *frame, xlator_t *this)
*/
void
-afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this, int child_index)
+afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this,
+ int child_index)
{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ afr_private_t * priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
__mark_child_dead (local->pending, priv->child_count,
- child_index, local->transaction.type);
+ child_index, local->transaction.type);
+}
+
+
+
+ static gf_boolean_t
+afr_locals_overlap (afr_local_t *local1, afr_local_t *local2)
+{
+ uint64_t start1 = local1->transaction.start;
+ uint64_t start2 = local2->transaction.start;
+ uint64_t end1 = 0;
+ uint64_t end2 = 0;
+
+ if (local1->transaction.len)
+ end1 = start1 + local1->transaction.len - 1;
+ else
+ end1 = ULLONG_MAX;
+
+ if (local2->transaction.len)
+ end2 = start2 + local2->transaction.len - 1;
+ else
+ end2 = ULLONG_MAX;
+
+ return ((end1 >= start2) && (end2 >= start1));
+}
+
+void
+afr_transaction_eager_lock_init (afr_local_t *local, xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ afr_fd_ctx_t *fdctx = NULL;
+ afr_local_t *each = NULL;
+
+ priv = this->private;
+
+ if (!local->fd)
+ return;
+
+ if (local->transaction.type != AFR_DATA_TRANSACTION)
+ return;
+
+ if (!priv->eager_lock)
+ return;
+
+ fdctx = afr_fd_ctx_get (local->fd, this);
+ if (!fdctx)
+ return;
+
+ if (afr_are_multiple_fds_opened (local->fd->inode, this))
+ return;
+ /*
+ * Once full file lock is acquired in eager-lock phase, overlapping
+ * writes do not compete for inode-locks, instead are transferred to the
+ * next writes. Because of this overlapping writes are not ordered.
+ * This can cause inconsistencies in replication.
+ * Example:
+ * Two overlapping writes w1, w2 are sent in parallel on same fd
+ * in two threads t1, t2.
+ * Both threads can execute afr_writev_wind in the following manner.
+ * t1 winds w1 on brick-0
+ * t2 winds w2 on brick-0
+ * t2 winds w2 on brick-1
+ * t1 winds w1 on brick-1
+ *
+ * This check makes sure the locks are not transferred for
+ * overlapping writes.
+ */
+ LOCK (&local->fd->lock);
+ {
+ list_for_each_entry (each, &fdctx->eager_locked,
+ transaction.eager_locked) {
+ if (afr_locals_overlap (each, local)) {
+ local->transaction.eager_lock_on = _gf_false;
+ goto unlock;
+ }
+ }
+
+ local->transaction.eager_lock_on = _gf_true;
+ list_add_tail (&local->transaction.eager_locked,
+ &fdctx->eager_locked);
+ }
+unlock:
+ UNLOCK (&local->fd->lock);
}
int
afr_transaction (call_frame_t *frame, xlator_t *this, afr_transaction_type type)
{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ afr_private_t * priv = NULL;
+ fd_t *fd = NULL;
+ int ret = -1;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- afr_transaction_local_init (local, priv);
+ local->transaction.resume = afr_transaction_resume;
+ local->transaction.type = type;
- local->transaction.resume = afr_transaction_resume;
- local->transaction.type = type;
+ ret = afr_transaction_local_init (local, this);
+ if (ret < 0)
+ goto out;
- if (afr_lock_server_count (priv, local->transaction.type) == 0) {
- afr_internal_lock_finish (frame, this);
- } else {
- afr_lock (frame, this);
- }
+ afr_transaction_eager_lock_init (local, this);
+
+ if (local->fd && local->transaction.eager_lock_on)
+ afr_set_lk_owner (frame, this, local->fd);
+ else
+ afr_set_lk_owner (frame, this, frame->root);
- return 0;
+ if (!local->transaction.eager_lock_on && local->loc.inode) {
+ fd = fd_lookup (local->loc.inode, frame->root->pid);
+ if (fd == NULL)
+ fd = fd_lookup_anonymous (local->loc.inode);
+
+ if (fd) {
+ afr_delayed_changelog_wake_up (this, fd);
+ fd_unref (fd);
+ }
+ }
+
+ if (afr_lock_server_count (priv, local->transaction.type) == 0) {
+ afr_internal_lock_finish (frame, this);
+ } else {
+ afr_lock (frame, this);
+ }
+ ret = 0;
+out:
+ return ret;
}
diff --git a/xlators/cluster/afr/src/afr-transaction.h b/xlators/cluster/afr/src/afr-transaction.h
index 0d3d4443e..fa626fd0d 100644
--- a/xlators/cluster/afr/src/afr-transaction.h
+++ b/xlators/cluster/afr/src/afr-transaction.h
@@ -1,25 +1,21 @@
/*
- Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __TRANSACTION_H__
#define __TRANSACTION_H__
+typedef enum {
+ LOCAL_FIRST = 1,
+ LOCAL_LAST = 2
+} afr_xattrop_type_t;
+
void
afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this,
int child_index);
@@ -27,7 +23,29 @@ afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this,
int
afr_lock_server_count (afr_private_t *priv, afr_transaction_type type);
+afr_inodelk_t*
+afr_get_inodelk (afr_internal_lock_t *int_lock, char *dom);
+
int32_t
afr_transaction (call_frame_t *frame, xlator_t *this, afr_transaction_type type);
+afr_fd_ctx_t *
+afr_fd_ctx_get (fd_t *fd, xlator_t *this);
+int
+afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending,
+ int child, afr_xattrop_type_t op);
+void
+afr_set_delayed_post_op (call_frame_t *frame, xlator_t *this);
+
+void
+afr_delayed_changelog_wake_up (xlator_t *this, fd_t *fd);
+
+void
+__mark_all_success (int32_t *pending[], int child_count,
+ afr_transaction_type type);
+gf_boolean_t
+afr_any_fops_failed (afr_local_t *local, afr_private_t *priv);
+
+gf_boolean_t
+afr_txn_nothing_failed (call_frame_t *frame, xlator_t *this);
#endif /* __TRANSACTION_H__ */
diff --git a/xlators/cluster/afr/src/afr.c b/xlators/cluster/afr/src/afr.c
index 69b281d97..c724eb2ae 100644
--- a/xlators/cluster/afr/src/afr.c
+++ b/xlators/cluster/afr/src/afr.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include <libgen.h>
@@ -30,13 +21,25 @@
#endif
#include "afr-common.c"
+#define SHD_INODE_LRU_LIMIT 2048
+#define AFR_EH_HEALED_LIMIT 1024
+#define AFR_EH_HEAL_FAIL_LIMIT 1024
+#define AFR_EH_SPLIT_BRAIN_LIMIT 1024
+
+struct volume_options options[];
+
int32_t
notify (xlator_t *this, int32_t event,
- void *data, ...)
+ void *data, ...)
{
int ret = -1;
+ va_list ap;
+ void *data2 = NULL;
- ret = afr_notify (this, event, data);
+ va_start (ap, data);
+ data2 = va_arg (ap, dict_t*);
+ va_end (ap);
+ ret = afr_notify (this, event, data, data2);
return ret;
}
@@ -53,7 +56,7 @@ mem_acct_init (xlator_t *this)
if (ret != 0) {
gf_log(this->name, GF_LOG_ERROR, "Memory accounting init"
- "failed");
+ "failed");
return ret;
}
@@ -62,647 +65,476 @@ mem_acct_init (xlator_t *this)
int
+xlator_subvolume_index (xlator_t *this, xlator_t *subvol)
+{
+ int index = -1;
+ int i = 0;
+ xlator_list_t *list = NULL;
+
+ list = this->children;
+
+ while (list) {
+ if (subvol == list->xlator ||
+ strcmp (subvol->name, list->xlator->name) == 0) {
+ index = i;
+ break;
+ }
+ list = list->next;
+ i++;
+ }
+
+ return index;
+}
+
+void
+fix_quorum_options (xlator_t *this, afr_private_t *priv, char *qtype)
+{
+ if (priv->quorum_count && strcmp(qtype,"fixed")) {
+ gf_log(this->name,GF_LOG_WARNING,
+ "quorum-type %s overriding quorum-count %u",
+ qtype, priv->quorum_count);
+ }
+ if (!strcmp(qtype,"none")) {
+ priv->quorum_count = 0;
+ }
+ else if (!strcmp(qtype,"auto")) {
+ priv->quorum_count = AFR_QUORUM_AUTO;
+ }
+}
+
+int
reconfigure (xlator_t *this, dict_t *options)
{
+ afr_private_t *priv = NULL;
+ xlator_t *read_subvol = NULL;
+ int read_subvol_index = -1;
+ int ret = -1;
+ int index = -1;
+ char *qtype = NULL;
+
+ priv = this->private;
+
+ GF_OPTION_RECONF ("background-self-heal-count",
+ priv->background_self_heal_count, options, uint32,
+ out);
- gf_boolean_t metadata_self_heal; /* on/off */
- gf_boolean_t entry_self_heal;
- gf_boolean_t data_self_heal;
- gf_boolean_t data_change_log; /* on/off */
- gf_boolean_t metadata_change_log; /* on/off */
- gf_boolean_t entry_change_log; /* on/off */
- gf_boolean_t strict_readdir;
-
- afr_private_t * priv = NULL;
- xlator_list_t * trav = NULL;
-
- char * read_subvol = NULL;
- char * self_heal = NULL;
- char * change_log = NULL;
- char * str_readdir = NULL;
-
- int32_t background_count = 0;
- int32_t window_size = 0;
-
- int read_ret = -1;
- int dict_ret = -1;
- int flag = 1;
- int ret = 0;
- int temp_ret = -1;
-
- priv = this->private;
-
- dict_ret = dict_get_int32 (options, "background-self-heal-count",
- &background_count);
- if (dict_ret == 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Reconfiguring background self-heal count to %d",
- background_count);
-
- priv->background_self_heal_count = background_count;
- }
-
- dict_ret = dict_get_str (options, "metadata-self-heal",
- &self_heal);
- if (dict_ret == 0) {
- temp_ret = gf_string2boolean (self_heal, &metadata_self_heal);
- if (temp_ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Reconfiguration Invalid 'option metadata"
- "-self-heal %s'. Defaulting to old value.",
- self_heal);
- ret = -1;
- goto out;
- }
-
- priv->metadata_self_heal = metadata_self_heal;
- gf_log (this->name, GF_LOG_DEBUG,
- "Reconfiguring 'option metadata"
- "-self-heal %s'.",
- self_heal);
- }
-
- dict_ret = dict_get_str (options, "data-self-heal",
- &self_heal);
- if (dict_ret == 0) {
- temp_ret = gf_string2boolean (self_heal, &data_self_heal);
- if (temp_ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Reconfiguration Invalid 'option data"
- "-self-heal %s'. Defaulting to old value.",
- self_heal);
- ret = -1;
- goto out;
- }
-
- priv->data_self_heal = data_self_heal;
- gf_log (this->name, GF_LOG_DEBUG,
- "Reconfiguring 'option data"
- "-self-heal %s'.", self_heal);
- }
-
- dict_ret = dict_get_str (options, "entry-self-heal",
- &self_heal);
- if (dict_ret == 0) {
- temp_ret = gf_string2boolean (self_heal, &entry_self_heal);
- if (temp_ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Reconfiguration Invalid 'option data"
- "-self-heal %s'. Defaulting to old value.",
- self_heal);
- ret = -1;
- goto out;
- }
-
- priv->entry_self_heal = entry_self_heal;
- gf_log (this->name, GF_LOG_DEBUG,
- "Reconfiguring 'option entry"
- "-self-heal %s'.", self_heal);
- }
-
-
- dict_ret = dict_get_str (options, "strict-readdir",
- &str_readdir);
- if (dict_ret == 0) {
- temp_ret = gf_string2boolean (str_readdir, &strict_readdir);
- if (temp_ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Invalid 'option strict-readdir %s'. "
- "Defaulting to old value.",
- str_readdir);
- ret = -1;
- goto out;
- }
-
- priv->strict_readdir = strict_readdir;
- gf_log (this->name, GF_LOG_DEBUG,
- "Reconfiguring 'option strict"
- "-readdir %s'.", str_readdir);
- }
-
- dict_ret = dict_get_int32 (options, "data-self-heal-window-size",
- &window_size);
- if (dict_ret == 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Reconfiguring, Setting data self-heal window size to %d",
- window_size);
-
- priv->data_self_heal_window_size = window_size;
- }
-
- dict_ret = dict_get_str (options, "data-change-log",
- &change_log);
- if (dict_ret == 0) {
- temp_ret = gf_string2boolean (change_log, &data_change_log);
- if (temp_ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Reconfiguration Invalid 'option data-"
- "change-log %s'. Defaulting to old value.",
- change_log);
- ret = -1;
- goto out;
- }
-
- priv->data_change_log = data_change_log;
- gf_log (this->name, GF_LOG_DEBUG,
- "Reconfiguring 'option data-"
- "change-log %s'.", change_log);
- }
-
- dict_ret = dict_get_str (options, "metadata-change-log",
- &change_log);
- if (dict_ret == 0) {
- temp_ret = gf_string2boolean (change_log,
- &metadata_change_log);
- if (temp_ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Invalid 'option metadata-change-log %s'. "
- "Defaulting to metadata-change-log as 'off'.",
- change_log);
- ret = -1;
- goto out;
- }
-
- priv->metadata_change_log = metadata_change_log;
- gf_log (this->name, GF_LOG_DEBUG,
- "Reconfiguring 'option metadata-"
- "change-log %s'.", change_log);
- }
-
- dict_ret = dict_get_str (options, "entry-change-log",
- &change_log);
- if (dict_ret == 0) {
- temp_ret = gf_string2boolean (change_log, &entry_change_log);
- if (temp_ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Invalid 'option entry-change-log %s'. "
- "Defaulting to entry-change-log as 'on'.",
- change_log);
- ret = -1;
- goto out;
- }
-
- priv->entry_change_log = entry_change_log;
- gf_log (this->name, GF_LOG_DEBUG,
- "Reconfiguring 'option entry-"
- "change-log %s'.", change_log);
- }
-
- read_ret = dict_get_str (options, "read-subvolume", &read_subvol);
-
- if (read_ret < 0)
- goto next;// No need to traverse, hence set the next option
-
- trav = this->children;
- flag = 0;
- while (trav) {
- if (!read_ret && !strcmp (read_subvol, trav->xlator->name)) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Subvolume '%s' specified as read child.",
- trav->xlator->name);
-
- flag = 1;
- ret = -1;
- goto out;
- }
-
-
- trav = trav->next;
- }
-
- if (flag == 0 ) {
-
- gf_log (this->name, GF_LOG_ERROR,
- "Invalid 'option read-subvolume %s', no such subvolume"
- , read_subvol);
- ret = -1;
- goto out;
- }
-
-next:
+ GF_OPTION_RECONF ("metadata-self-heal",
+ priv->metadata_self_heal, options, bool, out);
+
+ GF_OPTION_RECONF ("data-self-heal", priv->data_self_heal, options, str,
+ out);
+
+ GF_OPTION_RECONF ("entry-self-heal", priv->entry_self_heal, options,
+ bool, out);
+
+ GF_OPTION_RECONF ("strict-readdir", priv->strict_readdir, options, bool,
+ out);
+
+ GF_OPTION_RECONF ("data-self-heal-window-size",
+ priv->data_self_heal_window_size, options,
+ uint32, out);
+
+ GF_OPTION_RECONF ("data-change-log", priv->data_change_log, options,
+ bool, out);
+
+ GF_OPTION_RECONF ("metadata-change-log",
+ priv->metadata_change_log, options, bool, out);
+
+ GF_OPTION_RECONF ("entry-change-log", priv->entry_change_log, options,
+ bool, out);
+
+ GF_OPTION_RECONF ("data-self-heal-algorithm",
+ priv->data_self_heal_algorithm, options, str, out);
+
+ GF_OPTION_RECONF ("self-heal-daemon", priv->shd.enabled, options, bool, out);
+
+ GF_OPTION_RECONF ("read-subvolume", read_subvol, options, xlator, out);
+
+ GF_OPTION_RECONF ("read-hash-mode", priv->hash_mode,
+ options, uint32, out);
+
+ if (read_subvol) {
+ index = xlator_subvolume_index (this, read_subvol);
+ if (index == -1) {
+ gf_log (this->name, GF_LOG_ERROR, "%s not a subvolume",
+ read_subvol->name);
+ goto out;
+ }
+ priv->read_child = index;
+ }
+
+ GF_OPTION_RECONF ("read-subvolume-index",read_subvol_index, options,int32,out);
+
+ if (read_subvol_index >-1) {
+ index=read_subvol_index;
+ if (index >= priv->child_count) {
+ gf_log (this->name, GF_LOG_ERROR, "%d not a subvolume-index",
+ index);
+ goto out;
+ }
+ priv->read_child = index;
+ }
+
+ GF_OPTION_RECONF ("eager-lock", priv->eager_lock, options, bool, out);
+ GF_OPTION_RECONF ("quorum-type", qtype, options, str, out);
+ GF_OPTION_RECONF ("quorum-count", priv->quorum_count, options,
+ uint32, out);
+ fix_quorum_options(this,priv,qtype);
+ GF_OPTION_RECONF ("heal-timeout", priv->shd.timeout, options,
+ int32, out);
+
+ GF_OPTION_RECONF ("post-op-delay-secs", priv->post_op_delay_secs, options,
+ uint32, out);
+
+ GF_OPTION_RECONF (AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size,
+ options, size, out);
+ /* Reset this so we re-discover in case the topology changed. */
+ GF_OPTION_RECONF ("readdir-failover", priv->readdir_failover, options,
+ bool, out);
+ GF_OPTION_RECONF ("ensure-durability", priv->ensure_durability, options,
+ bool, out);
+ priv->did_discovery = _gf_false;
+
+ ret = 0;
out:
- return ret;
+ return ret;
}
static const char *favorite_child_warning_str = "You have specified subvolume '%s' "
- "as the 'favorite child'. This means that if a discrepancy in the content "
- "or attributes (ownership, permission, etc.) of a file is detected among "
- "the subvolumes, the file on '%s' will be considered the definitive "
- "version and its contents will OVERWRITE the contents of the file on other "
- "subvolumes. All versions of the file except that on '%s' "
- "WILL BE LOST.";
-
-static const char *no_lock_servers_warning_str = "You have set lock-server-count = 0. "
- "This means correctness is NO LONGER GUARANTEED in all cases. If two or more "
- "applications write to the same region of a file, there is a possibility that "
- "its copies will be INCONSISTENT. Set it to a value greater than 0 unless you "
- "are ABSOLUTELY SURE of what you are doing and WILL NOT HOLD GlusterFS "
- "RESPONSIBLE for inconsistent data. If you are in doubt, set it to a value "
- "greater than 0.";
-
-int32_t
+ "as the 'favorite child'. This means that if a discrepancy in the content "
+ "or attributes (ownership, permission, etc.) of a file is detected among "
+ "the subvolumes, the file on '%s' will be considered the definitive "
+ "version and its contents will OVERWRITE the contents of the file on other "
+ "subvolumes. All versions of the file except that on '%s' "
+ "WILL BE LOST.";
+
+
+int32_t
init (xlator_t *this)
{
- afr_private_t * priv = NULL;
- int child_count = 0;
- xlator_list_t * trav = NULL;
- int i = 0;
- int ret = -1;
- int op_errno = 0;
-
- char * read_subvol = NULL;
- char * fav_child = NULL;
- char * self_heal = NULL;
- char * algo = NULL;
- char * change_log = NULL;
- char * strict_readdir = NULL;
- char * inodelk_trace = NULL;
- char * entrylk_trace = NULL;
-
- int32_t background_count = 0;
- int32_t lock_server_count = 1;
- int32_t window_size = 0;
-
- int fav_ret = -1;
- int read_ret = -1;
- int dict_ret = -1;
-
-
- if (!this->children) {
- gf_log (this->name, GF_LOG_ERROR,
- "replicate translator needs more than one "
+ afr_private_t *priv = NULL;
+ int child_count = 0;
+ xlator_list_t *trav = NULL;
+ int i = 0;
+ int ret = -1;
+ GF_UNUSED int op_errno = 0;
+ xlator_t *read_subvol = NULL;
+ int read_subvol_index = -1;
+ xlator_t *fav_child = NULL;
+ char *qtype = NULL;
+
+ if (!this->children) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "replicate translator needs more than one "
"subvolume defined.");
- return -1;
- }
-
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "Volume is dangling.");
- }
+ return -1;
+ }
+
+ if (!this->parents) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "Volume is dangling.");
+ }
+
+ this->private = GF_CALLOC (1, sizeof (afr_private_t),
+ gf_afr_mt_afr_private_t);
+ if (!this->private)
+ goto out;
+
+ priv = this->private;
+ LOCK_INIT (&priv->lock);
+ LOCK_INIT (&priv->read_child_lock);
+ //lock recovery is not done in afr
+ pthread_mutex_init (&priv->mutex, NULL);
+ INIT_LIST_HEAD (&priv->saved_fds);
+
+ child_count = xlator_subvolume_count (this);
+
+ priv->child_count = child_count;
+
+ priv->read_child = -1;
+ GF_OPTION_INIT ("read-subvolume", read_subvol, xlator, out);
+ if (read_subvol) {
+ priv->read_child = xlator_subvolume_index (this, read_subvol);
+ if (priv->read_child == -1) {
+ gf_log (this->name, GF_LOG_ERROR, "%s not a subvolume",
+ read_subvol->name);
+ goto out;
+ }
+ }
+ GF_OPTION_INIT ("read-subvolume-index",read_subvol_index,int32,out);
+ if (read_subvol_index > -1) {
+ if (read_subvol_index >= priv->child_count) {
+ gf_log (this->name, GF_LOG_ERROR, "%d not a subvolume-index",
+ read_subvol_index);
+ goto out;
+ }
+ priv->read_child = read_subvol_index;
+ }
+ GF_OPTION_INIT ("choose-local", priv->choose_local, bool, out);
- ALLOC_OR_GOTO (this->private, afr_private_t, out);
+ GF_OPTION_INIT ("read-hash-mode", priv->hash_mode, uint32, out);
- priv = this->private;
+ priv->favorite_child = -1;
+ GF_OPTION_INIT ("favorite-child", fav_child, xlator, out);
+ if (fav_child) {
+ priv->favorite_child = xlator_subvolume_index (this, fav_child);
+ if (priv->favorite_child == -1) {
+ gf_log (this->name, GF_LOG_ERROR, "%s not a subvolume",
+ fav_child->name);
+ goto out;
+ }
+ gf_log (this->name, GF_LOG_WARNING,
+ favorite_child_warning_str, fav_child->name,
+ fav_child->name, fav_child->name);
+ }
- read_ret = dict_get_str (this->options, "read-subvolume", &read_subvol);
- priv->read_child = -1;
- fav_ret = dict_get_str (this->options, "favorite-child", &fav_child);
- priv->favorite_child = -1;
+ GF_OPTION_INIT ("background-self-heal-count",
+ priv->background_self_heal_count, uint32, out);
- priv->background_self_heal_count = 16;
+ GF_OPTION_INIT ("data-self-heal", priv->data_self_heal, str, out);
- dict_ret = dict_get_int32 (this->options, "background-self-heal-count",
- &background_count);
- if (dict_ret == 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Setting background self-heal count to %d",
- background_count);
+ GF_OPTION_INIT ("data-self-heal-algorithm",
+ priv->data_self_heal_algorithm, str, out);
- priv->background_self_heal_count = background_count;
- }
+ GF_OPTION_INIT ("data-self-heal-window-size",
+ priv->data_self_heal_window_size, uint32, out);
- /* Default values */
+ GF_OPTION_INIT ("metadata-self-heal", priv->metadata_self_heal, bool,
+ out);
- priv->data_self_heal = 1;
- priv->metadata_self_heal = 1;
- priv->entry_self_heal = 1;
+ GF_OPTION_INIT ("entry-self-heal", priv->entry_self_heal, bool, out);
- dict_ret = dict_get_str (this->options, "data-self-heal", &self_heal);
- if (dict_ret == 0) {
- ret = gf_string2boolean (self_heal, &priv->data_self_heal);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Invalid 'option data-self-heal %s'. "
- "Defaulting to data-self-heal as 'on'",
- self_heal);
- priv->data_self_heal = 1;
- }
- }
+ GF_OPTION_INIT ("self-heal-daemon", priv->shd.enabled, bool, out);
- priv->data_self_heal_algorithm = "";
+ GF_OPTION_INIT ("iam-self-heal-daemon", priv->shd.iamshd, bool, out);
- dict_ret = dict_get_str (this->options, "data-self-heal-algorithm",
- &algo);
- if (dict_ret == 0) {
- priv->data_self_heal_algorithm = gf_strdup (algo);
- }
+ GF_OPTION_INIT ("data-change-log", priv->data_change_log, bool, out);
+ GF_OPTION_INIT ("metadata-change-log", priv->metadata_change_log, bool,
+ out);
- priv->data_self_heal_window_size = 16;
-
- dict_ret = dict_get_int32 (this->options, "data-self-heal-window-size",
- &window_size);
- if (dict_ret == 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Setting data self-heal window size to %d",
- window_size);
-
- priv->data_self_heal_window_size = window_size;
- }
-
- dict_ret = dict_get_str (this->options, "metadata-self-heal",
- &self_heal);
- if (dict_ret == 0) {
- ret = gf_string2boolean (self_heal, &priv->metadata_self_heal);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Invalid 'option metadata-self-heal %s'. "
- "Defaulting to metadata-self-heal as 'on'.",
- self_heal);
- priv->metadata_self_heal = 1;
- }
- }
-
- dict_ret = dict_get_str (this->options, "entry-self-heal", &self_heal);
- if (dict_ret == 0) {
- ret = gf_string2boolean (self_heal, &priv->entry_self_heal);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Invalid 'option entry-self-heal %s'. "
- "Defaulting to entry-self-heal as 'on'.",
- self_heal);
- priv->entry_self_heal = 1;
- }
- }
-
- /* Change log options */
-
- priv->data_change_log = 1;
- priv->metadata_change_log = 1;
- priv->entry_change_log = 1;
-
- dict_ret = dict_get_str (this->options, "data-change-log",
- &change_log);
- if (dict_ret == 0) {
- ret = gf_string2boolean (change_log, &priv->data_change_log);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Invalid 'option data-change-log %s'. "
- "Defaulting to data-change-log as 'on'.",
- change_log);
- priv->data_change_log = 1;
- }
- }
-
- dict_ret = dict_get_str (this->options, "metadata-change-log",
- &change_log);
- if (dict_ret == 0) {
- ret = gf_string2boolean (change_log,
- &priv->metadata_change_log);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Invalid 'option metadata-change-log %s'. "
- "Defaulting to metadata-change-log as 'off'.",
- change_log);
- priv->metadata_change_log = 0;
- }
- }
-
- dict_ret = dict_get_str (this->options, "entry-change-log",
- &change_log);
- if (dict_ret == 0) {
- ret = gf_string2boolean (change_log, &priv->entry_change_log);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Invalid 'option entry-change-log %s'. "
- "Defaulting to entry-change-log as 'on'.",
- change_log);
- priv->entry_change_log = 1;
- }
- }
-
- /* Locking options */
-
- priv->inodelk_trace = 0;
- priv->entrylk_trace = 0;
-
- dict_ret = dict_get_str (this->options, "inodelk-trace",
- &inodelk_trace);
- if (dict_ret == 0) {
- ret = gf_string2boolean (inodelk_trace, &priv->inodelk_trace);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Invalid 'option inodelk-trace %s' ",
- inodelk_trace);
-
- priv->inodelk_trace = 0;
- }
- }
-
-
- dict_ret = dict_get_str (this->options, "entrylk-trace",
- &entrylk_trace);
- if (dict_ret == 0) {
- ret = gf_string2boolean (entrylk_trace, &priv->entrylk_trace);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Invalid 'option entrylk-trace %s' ",
- inodelk_trace);
-
- priv->entrylk_trace = 0;
- }
- }
-
-
- priv->data_lock_server_count = 1;
- priv->metadata_lock_server_count = 0;
- priv->entry_lock_server_count = 1;
-
- dict_ret = dict_get_int32 (this->options, "data-lock-server-count",
- &lock_server_count);
- if (dict_ret == 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Setting data lock server count to %d.",
- lock_server_count);
-
- if (lock_server_count == 0)
- gf_log (this->name, GF_LOG_WARNING, "%s",
- no_lock_servers_warning_str);
-
- priv->data_lock_server_count = lock_server_count;
- }
-
-
- dict_ret = dict_get_int32 (this->options,
- "metadata-lock-server-count",
- &lock_server_count);
- if (dict_ret == 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Setting metadata lock server count to %d.",
- lock_server_count);
- priv->metadata_lock_server_count = lock_server_count;
- }
-
-
- dict_ret = dict_get_int32 (this->options, "entry-lock-server-count",
- &lock_server_count);
- if (dict_ret == 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Setting entry lock server count to %d.",
- lock_server_count);
-
- priv->entry_lock_server_count = lock_server_count;
- }
-
- priv->strict_readdir = _gf_false;
-
- dict_ret = dict_get_str (this->options, "strict-readdir",
- &strict_readdir);
- if (dict_ret == 0) {
- ret = gf_string2boolean (strict_readdir, &priv->strict_readdir);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Invalid 'option strict-readdir %s'. "
- "Defaulting to strict-readdir as 'off'.",
- strict_readdir);
- }
- }
-
- trav = this->children;
- while (trav) {
- if (!read_ret && !strcmp (read_subvol, trav->xlator->name)) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Subvolume '%s' specified as read child.",
- trav->xlator->name);
-
- priv->read_child = child_count;
- }
-
- if (fav_ret == 0 && !strcmp (fav_child, trav->xlator->name)) {
- gf_log (this->name, GF_LOG_WARNING,
- favorite_child_warning_str, trav->xlator->name,
- trav->xlator->name, trav->xlator->name);
- priv->favorite_child = child_count;
- }
-
- child_count++;
- trav = trav->next;
- }
-
- priv->wait_count = 1;
-
- priv->child_count = child_count;
-
- LOCK_INIT (&priv->lock);
- LOCK_INIT (&priv->read_child_lock);
+ GF_OPTION_INIT ("entry-change-log", priv->entry_change_log, bool, out);
+
+ GF_OPTION_INIT ("optimistic-change-log", priv->optimistic_change_log,
+ bool, out);
+
+ GF_OPTION_INIT ("inodelk-trace", priv->inodelk_trace, bool, out);
+
+ GF_OPTION_INIT ("entrylk-trace", priv->entrylk_trace, bool, out);
+
+ GF_OPTION_INIT ("strict-readdir", priv->strict_readdir, bool, out);
- priv->child_up = GF_CALLOC (sizeof (unsigned char), child_count,
+ GF_OPTION_INIT ("eager-lock", priv->eager_lock, bool, out);
+ GF_OPTION_INIT ("quorum-type", qtype, str, out);
+ GF_OPTION_INIT ("quorum-count", priv->quorum_count, uint32, out);
+ GF_OPTION_INIT (AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size, size,
+ out);
+ fix_quorum_options(this,priv,qtype);
+
+ GF_OPTION_INIT ("post-op-delay-secs", priv->post_op_delay_secs, uint32, out);
+ GF_OPTION_INIT ("readdir-failover", priv->readdir_failover, bool, out);
+ GF_OPTION_INIT ("ensure-durability", priv->ensure_durability, bool,
+ out);
+
+ priv->wait_count = 1;
+
+ priv->child_up = GF_CALLOC (sizeof (unsigned char), child_count,
gf_afr_mt_char);
- if (!priv->child_up) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
- ret = -ENOMEM;
- goto out;
- }
-
- priv->children = GF_CALLOC (sizeof (xlator_t *), child_count,
+ if (!priv->child_up) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ for (i = 0; i < child_count; i++)
+ priv->child_up[i] = -1; /* start with unknown state.
+ this initialization needed
+ for afr_notify() to work
+ reliably
+ */
+
+ priv->children = GF_CALLOC (sizeof (xlator_t *), child_count,
gf_afr_mt_xlator_t);
- if (!priv->children) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
- ret = -ENOMEM;
- goto out;
- }
-
- priv->pending_key = GF_CALLOC (sizeof (*priv->pending_key),
- child_count,
- gf_afr_mt_char);
+ if (!priv->children) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ priv->pending_key = GF_CALLOC (sizeof (*priv->pending_key),
+ child_count,
+ gf_afr_mt_char);
if (!priv->pending_key) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
ret = -ENOMEM;
goto out;
}
- trav = this->children;
- i = 0;
- while (i < child_count) {
- priv->children[i] = trav->xlator;
+ trav = this->children;
+ i = 0;
+ while (i < child_count) {
+ priv->children[i] = trav->xlator;
- ret = gf_asprintf (&priv->pending_key[i], "%s.%s",
+ ret = gf_asprintf (&priv->pending_key[i], "%s.%s",
AFR_XATTR_PREFIX,
trav->xlator->name);
if (-1 == ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "asprintf failed to set pending key");
ret = -ENOMEM;
goto out;
}
- trav = trav->next;
- i++;
- }
+ trav = trav->next;
+ i++;
+ }
+
+ ret = gf_asprintf (&priv->sh_domain, AFR_SH_DATA_DOMAIN_FMT,
+ this->name);
+ if (-1 == ret) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ priv->last_event = GF_CALLOC (child_count, sizeof (*priv->last_event),
+ gf_afr_mt_int32_t);
+ if (!priv->last_event) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ /* keep more local here as we may need them for self-heal etc */
+ this->local_pool = mem_pool_new (afr_local_t, 512);
+ if (!this->local_pool) {
+ ret = -1;
+ gf_log (this->name, GF_LOG_ERROR,
+ "failed to create local_t's memory pool");
+ goto out;
+ }
- LOCK_INIT (&priv->root_inode_lk);
priv->first_lookup = 1;
priv->root_inode = NULL;
- pthread_mutex_init (&priv->mutex, NULL);
- INIT_LIST_HEAD (&priv->saved_fds);
+ if (!priv->shd.iamshd) {
+ ret = 0;
+ goto out;
+ }
+
+ ret = -ENOMEM;
+ priv->shd.pos = GF_CALLOC (sizeof (*priv->shd.pos), child_count,
+ gf_afr_mt_brick_pos_t);
+ if (!priv->shd.pos)
+ goto out;
+
+ priv->shd.pending = GF_CALLOC (sizeof (*priv->shd.pending), child_count,
+ gf_afr_mt_int32_t);
+ if (!priv->shd.pending)
+ goto out;
+
+ priv->shd.inprogress = GF_CALLOC (sizeof (*priv->shd.inprogress),
+ child_count, gf_afr_mt_shd_bool_t);
+ if (!priv->shd.inprogress)
+ goto out;
+ priv->shd.timer = GF_CALLOC (sizeof (*priv->shd.timer), child_count,
+ gf_afr_mt_shd_timer_t);
+ if (!priv->shd.timer)
+ goto out;
+
+ priv->shd.healed = eh_new (AFR_EH_HEALED_LIMIT, _gf_false,
+ _destroy_shd_event_data);
+ if (!priv->shd.healed)
+ goto out;
+
+ priv->shd.heal_failed = eh_new (AFR_EH_HEAL_FAIL_LIMIT, _gf_false,
+ _destroy_shd_event_data);
+ if (!priv->shd.heal_failed)
+ goto out;
+
+ priv->shd.split_brain = eh_new (AFR_EH_SPLIT_BRAIN_LIMIT, _gf_false,
+ _destroy_shd_event_data);
+ if (!priv->shd.split_brain)
+ goto out;
- ret = 0;
+ this->itable = inode_table_new (SHD_INODE_LRU_LIMIT, this);
+ if (!this->itable)
+ goto out;
+ priv->root_inode = inode_ref (this->itable->root);
+ GF_OPTION_INIT ("node-uuid", priv->shd.node_uuid, str, out);
+ GF_OPTION_INIT ("heal-timeout", priv->shd.timeout, int32, out);
+ ret = afr_initialise_statistics (this);
+ if (ret)
+ goto out;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
fini (xlator_t *this)
{
- return 0;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ this->private = NULL;
+ afr_priv_destroy (priv);
+ if (this->itable);//I dont see any destroy func
+
+ return 0;
}
struct xlator_fops fops = {
- .lookup = afr_lookup,
- .open = afr_open,
- .lk = afr_lk,
- .flush = afr_flush,
- .statfs = afr_statfs,
- .fsync = afr_fsync,
- .fsyncdir = afr_fsyncdir,
- .xattrop = afr_xattrop,
- .fxattrop = afr_fxattrop,
- .inodelk = afr_inodelk,
- .finodelk = afr_finodelk,
- .entrylk = afr_entrylk,
- .fentrylk = afr_fentrylk,
-
- /* inode read */
- .access = afr_access,
- .stat = afr_stat,
- .fstat = afr_fstat,
- .readlink = afr_readlink,
- .getxattr = afr_getxattr,
- .readv = afr_readv,
-
- /* inode write */
- .writev = afr_writev,
- .truncate = afr_truncate,
- .ftruncate = afr_ftruncate,
- .setxattr = afr_setxattr,
+ .lookup = afr_lookup,
+ .open = afr_open,
+ .lk = afr_lk,
+ .flush = afr_flush,
+ .statfs = afr_statfs,
+ .fsync = afr_fsync,
+ .fsyncdir = afr_fsyncdir,
+ .xattrop = afr_xattrop,
+ .fxattrop = afr_fxattrop,
+ .inodelk = afr_inodelk,
+ .finodelk = afr_finodelk,
+ .entrylk = afr_entrylk,
+ .fentrylk = afr_fentrylk,
+ .fallocate = afr_fallocate,
+ .discard = afr_discard,
+ .zerofill = afr_zerofill,
+
+ /* inode read */
+ .access = afr_access,
+ .stat = afr_stat,
+ .fstat = afr_fstat,
+ .readlink = afr_readlink,
+ .getxattr = afr_getxattr,
+ .fgetxattr = afr_fgetxattr,
+ .readv = afr_readv,
+
+ /* inode write */
+ .writev = afr_writev,
+ .truncate = afr_truncate,
+ .ftruncate = afr_ftruncate,
+ .setxattr = afr_setxattr,
+ .fsetxattr = afr_fsetxattr,
.setattr = afr_setattr,
- .fsetattr = afr_fsetattr,
- .removexattr = afr_removexattr,
-
- /* dir read */
- .opendir = afr_opendir,
- .readdir = afr_readdir,
- .readdirp = afr_readdirp,
-
- /* dir write */
- .create = afr_create,
- .mknod = afr_mknod,
- .mkdir = afr_mkdir,
- .unlink = afr_unlink,
- .rmdir = afr_rmdir,
- .link = afr_link,
- .symlink = afr_symlink,
- .rename = afr_rename,
+ .fsetattr = afr_fsetattr,
+ .removexattr = afr_removexattr,
+ .fremovexattr = afr_fremovexattr,
+
+ /* dir read */
+ .opendir = afr_opendir,
+ .readdir = afr_readdir,
+ .readdirp = afr_readdirp,
+
+ /* dir write */
+ .create = afr_create,
+ .mknod = afr_mknod,
+ .mkdir = afr_mkdir,
+ .unlink = afr_unlink,
+ .rmdir = afr_rmdir,
+ .link = afr_link,
+ .symlink = afr_symlink,
+ .rename = afr_rename,
};
@@ -712,62 +544,250 @@ struct xlator_dumpops dumpops = {
struct xlator_cbks cbks = {
- .release = afr_release,
- .releasedir = afr_releasedir,
+ .release = afr_release,
+ .releasedir = afr_releasedir,
+ .forget = afr_forget,
};
struct volume_options options[] = {
- { .key = {"read-subvolume" },
- .type = GF_OPTION_TYPE_XLATOR
- },
- { .key = {"favorite-child"},
- .type = GF_OPTION_TYPE_XLATOR
- },
+ { .key = {"read-subvolume" },
+ .type = GF_OPTION_TYPE_XLATOR,
+ .description = "inode-read fops happen only on one of the bricks in "
+ "replicate. Afr will prefer the one specified using "
+ "this option if it is not stale. Option value must be "
+ "one of the xlator names of the children. "
+ "Ex: <volname>-client-0 till "
+ "<volname>-client-<number-of-bricks - 1>"
+ },
+ { .key = {"read-subvolume-index" },
+ .type = GF_OPTION_TYPE_INT,
+ .default_value = "-1",
+ .description = "inode-read fops happen only on one of the bricks in "
+ "replicate. AFR will prefer the one specified using "
+ "this option if it is not stale. allowed options"
+ " include -1 till replica-count - 1"
+ },
+ { .key = {"read-hash-mode" },
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max = 2,
+ .default_value = "0",
+ .description = "inode-read fops happen only on one of the bricks in "
+ "replicate. AFR will prefer the one computed using "
+ "the method specified using this option"
+ "0 = first responder, "
+ "1 = hash by GFID of file (all clients use "
+ "same subvolume), "
+ "2 = hash by GFID of file and client PID",
+ },
+ { .key = {"choose-local" },
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "true",
+ .description = "Choose a local subvolume(i.e. Brick) to read from if "
+ "read-subvolume is not explicitly set.",
+ },
+ { .key = {"favorite-child"},
+ .type = GF_OPTION_TYPE_XLATOR,
+ .description = "If a split-brain happens choose subvol/brick set by "
+ "this option as source."
+ },
{ .key = {"background-self-heal-count"},
.type = GF_OPTION_TYPE_INT,
- .min = 0
+ .min = 0,
+ .default_value = "16",
+ .validate = GF_OPT_VALIDATE_MIN,
+ .description = "This specifies the number of self-heals that can be "
+ " performed in background without blocking the fop"
+ },
+ { .key = {"data-self-heal"},
+ .type = GF_OPTION_TYPE_STR,
+ .value = {"1", "on", "yes", "true", "enable",
+ "0", "off", "no", "false", "disable",
+ "open"},
+ .default_value = "on",
+ .description = "Using this option we can enable/disable data "
+ "self-heal on the file. \"open\" means data "
+ "self-heal action will only be triggered by file "
+ "open operations."
},
- { .key = {"data-self-heal"},
- .type = GF_OPTION_TYPE_BOOL
- },
{ .key = {"data-self-heal-algorithm"},
- .type = GF_OPTION_TYPE_STR
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Select between \"full\", \"diff\". The "
+ "\"full\" algorithm copies the entire file from "
+ "source to sink. The \"diff\" algorithm copies to "
+ "sink only those blocks whose checksums don't match "
+ "with those of source. If no option is configured "
+ "the option is chosen dynamically as follows: "
+ "If the file does not exist on one of the sinks "
+ "or empty file exists or if the source file size is "
+ "about the same as page size the entire file will "
+ "be read and written i.e \"full\" algo, "
+ "otherwise \"diff\" algo is chosen.",
+ .value = { "diff", "full"}
},
{ .key = {"data-self-heal-window-size"},
.type = GF_OPTION_TYPE_INT,
.min = 1,
- .max = 1024
+ .max = 1024,
+ .default_value = "1",
+ .description = "Maximum number blocks per file for which self-heal "
+ "process would be applied simultaneously."
+ },
+ { .key = {"metadata-self-heal"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "Using this option we can enable/disable metadata "
+ "i.e. Permissions, ownerships, xattrs self-heal on "
+ "the file/directory."
+ },
+ { .key = {"entry-self-heal"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "Using this option we can enable/disable entry "
+ "self-heal on the directory."
+ },
+ { .key = {"data-change-log"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "Data fops like write/truncate will not perform "
+ "pre/post fop changelog operations in afr transaction "
+ "if this option is disabled"
+ },
+ { .key = {"metadata-change-log"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "Metadata fops like setattr/setxattr will not perform "
+ "pre/post fop changelog operations in afr transaction "
+ "if this option is disabled"
+ },
+ { .key = {"entry-change-log"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "Entry fops like create/unlink will not perform "
+ "pre/post fop changelog operations in afr transaction "
+ "if this option is disabled"
+ },
+ { .key = {"optimistic-change-log"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "Entry/Metadata fops will not perform "
+ "pre fop changelog operations in afr transaction "
+ "if this option is enabled."
+ },
+ { .key = {"strict-readdir"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ },
+ { .key = {"inodelk-trace"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "Enabling this option logs inode lock/unlocks"
+ },
+ { .key = {"entrylk-trace"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "Enabling this option logs entry lock/unlocks"
+ },
+ { .key = {"eager-lock"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "Lock phase of a transaction has two sub-phases. "
+ "First is an attempt to acquire locks in parallel by "
+ "broadcasting non-blocking lock requests. If lock "
+ "aquistion fails on any server, then the held locks "
+ "are unlocked and revert to a blocking locked mode "
+ "sequentially on one server after another. If this "
+ "option is enabled the initial broadcasting lock "
+ "request attempt to acquire lock on the entire file. "
+ "If this fails, we revert back to the sequential "
+ "\"regional\" blocking lock as before. In the case "
+ "where such an \"eager\" lock is granted in the "
+ "non-blocking phase, it gives rise to an opportunity "
+ "for optimization. i.e, if the next write transaction "
+ "on the same FD arrives before the unlock phase of "
+ "the first transaction, it \"takes over\" the full "
+ "file lock. Similarly if yet another data transaction "
+ "arrives before the unlock phase of the \"optimized\" "
+ "transaction, that in turn \"takes over\" the lock as "
+ "well. The actual unlock now happens at the end of "
+ "the last \"optimzed\" transaction."
+
+ },
+ { .key = {"self-heal-daemon"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "This option applies to only self-heal-daemon. "
+ "Index directory crawl and automatic healing of files"
+ "will not be performed if this option is turned off."
+ },
+ { .key = {"iam-self-heal-daemon"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "This option differentiates if the replicate "
+ "translator is running as part of self-heal-daemon "
+ "or not."
+ },
+ { .key = {"quorum-type"},
+ .type = GF_OPTION_TYPE_STR,
+ .value = { "none", "auto", "fixed"},
+ .default_value = "none",
+ .description = "If value is \"fixed\" only allow writes if "
+ "quorum-count bricks are present. If value is "
+ "\"auto\" only allow writes if more than half of "
+ "bricks, or exactly half including the first, are "
+ "present.",
+ },
+ { .key = {"quorum-count"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = INT_MAX,
+ .default_value = 0,
+ .description = "If quorum-type is \"fixed\" only allow writes if "
+ "this many bricks or present. Other quorum types "
+ "will OVERWRITE this value.",
+ },
+ { .key = {"node-uuid"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Local glusterd uuid string, used in starting "
+ "self-heal-daemon so that it can crawl only on "
+ "local index directories.",
+ },
+ { .key = {"heal-timeout"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 60,
+ .max = INT_MAX,
+ .default_value = "600",
+ .description = "time interval for checking the need to self-heal "
+ "in self-heal-daemon"
+ },
+ { .key = {"post-op-delay-secs"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max = INT_MAX,
+ .default_value = "1",
+ .description = "Time interval induced artificially before "
+ "post-operation phase of the transaction to "
+ "enhance overlap of adjacent write operations.",
+ },
+ { .key = {AFR_SH_READDIR_SIZE_KEY},
+ .type = GF_OPTION_TYPE_SIZET,
+ .description = "readdirp size for performing entry self-heal",
+ .min = 1024,
+ .max = 131072,
+ .default_value = "1KB",
+ },
+ { .key = {"readdir-failover"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .description = "readdir(p) will not failover if this option is off",
+ .default_value = "on",
+ },
+ { .key = {"ensure-durability"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .description = "Afr performs fsyncs for transactions if this "
+ "option is on to make sure the changelogs/data is "
+ "written to the disk",
+ .default_value = "on",
},
- { .key = {"metadata-self-heal"},
- .type = GF_OPTION_TYPE_BOOL
- },
- { .key = {"entry-self-heal"},
- .type = GF_OPTION_TYPE_BOOL
- },
- { .key = {"data-change-log"},
- .type = GF_OPTION_TYPE_BOOL
- },
- { .key = {"metadata-change-log"},
- .type = GF_OPTION_TYPE_BOOL
- },
- { .key = {"entry-change-log"},
- .type = GF_OPTION_TYPE_BOOL
- },
- { .key = {"data-lock-server-count"},
- .type = GF_OPTION_TYPE_INT,
- .min = 0
- },
- { .key = {"metadata-lock-server-count"},
- .type = GF_OPTION_TYPE_INT,
- .min = 0
- },
- { .key = {"entry-lock-server-count"},
- .type = GF_OPTION_TYPE_INT,
- .min = 0
- },
- { .key = {"strict-readdir"},
- .type = GF_OPTION_TYPE_BOOL,
- },
- { .key = {NULL} },
+ { .key = {NULL} },
};
diff --git a/xlators/cluster/afr/src/afr.h b/xlators/cluster/afr/src/afr.h
index 68b4a1e30..21064db58 100644
--- a/xlators/cluster/afr/src/afr.h
+++ b/xlators/cluster/afr/src/afr.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -29,154 +20,316 @@
#include "call-stub.h"
#include "compat-errno.h"
#include "afr-mem-types.h"
+#include "afr-self-heal-algorithm.h"
+
+#include "libxlator.h"
+#include "timer.h"
#define AFR_XATTR_PREFIX "trusted.afr"
+#define AFR_PATHINFO_HEADER "REPLICATE:"
+#define AFR_SH_READDIR_SIZE_KEY "self-heal-readdir-size"
+#define AFR_SH_DATA_DOMAIN_FMT "%s:self-heal"
+
+#define AFR_LOCKEE_COUNT_MAX 3
+#define AFR_DOM_COUNT_MAX 3
struct _pump_private;
+typedef int (*afr_expunge_done_cbk_t) (call_frame_t *frame, xlator_t *this,
+ int child, int32_t op_error,
+ int32_t op_errno);
+
+typedef int (*afr_impunge_done_cbk_t) (call_frame_t *frame, xlator_t *this,
+ int32_t op_error, int32_t op_errno);
+typedef int (*afr_post_remove_call_t) (call_frame_t *frame, xlator_t *this);
+
+typedef int (*afr_lock_cbk_t) (call_frame_t *frame, xlator_t *this);
+typedef void (*afr_lookup_done_cbk_t) (call_frame_t *frame, xlator_t *this,
+ int32_t op_ret, int32_t op_errno);
+
+typedef enum {
+ AFR_POS_UNKNOWN,
+ AFR_POS_LOCAL,
+ AFR_POS_REMOTE
+} afr_child_pos_t;
+
+typedef enum {
+ SPLIT_BRAIN = 1,
+ ALL_FOOLS = 2
+} afr_subvol_status_t;
+
+typedef enum {
+ AFR_INODE_SET_READ_CTX = 1,
+ AFR_INODE_RM_STALE_CHILDREN,
+ AFR_INODE_SET_OPENDIR_DONE,
+ AFR_INODE_GET_READ_CTX,
+ AFR_INODE_GET_OPENDIR_DONE,
+} afr_inode_op_t;
+
+typedef struct afr_inode_params_ {
+ afr_inode_op_t op;
+ union {
+ gf_boolean_t value;
+ struct {
+ int32_t read_child;
+ int32_t *children;
+ } read_ctx;
+ } u;
+} afr_inode_params_t;
+
+typedef enum afr_spb_state {
+ DONT_KNOW,
+ SPB,
+ NO_SPB
+} afr_spb_state_t;
+
+typedef struct afr_inode_ctx_ {
+ uint64_t masks;
+ int32_t *fresh_children;//increasing order of latency
+ afr_spb_state_t mdata_spb;
+ afr_spb_state_t data_spb;
+ uint32_t open_fd_count;
+} afr_inode_ctx_t;
+
+typedef enum {
+ NONE,
+ INDEX,
+ INDEX_TO_BE_HEALED,
+ FULL,
+} afr_crawl_type_t;
+
+typedef struct afr_self_heald_ {
+ gf_boolean_t enabled;
+ gf_boolean_t iamshd;
+ afr_crawl_type_t *pending;
+ gf_boolean_t *inprogress;
+ afr_child_pos_t *pos;
+ gf_timer_t **timer;
+ eh_t *healed;
+ eh_t *heal_failed;
+ eh_t *split_brain;
+ eh_t **statistics;
+ void **crawl_events;
+ char *node_uuid;
+ int timeout;
+} afr_self_heald_t;
+
typedef struct _afr_private {
- gf_lock_t lock; /* to guard access to child_count, etc */
- unsigned int child_count; /* total number of children */
+ gf_lock_t lock; /* to guard access to child_count, etc */
+ unsigned int child_count; /* total number of children */
unsigned int read_child_rr; /* round-robin index of the read_child */
gf_lock_t read_child_lock; /* lock to protect above */
- xlator_t **children;
+ xlator_t **children;
- gf_lock_t root_inode_lk;
int first_lookup;
inode_t *root_inode;
- unsigned char *child_up;
+ unsigned char *child_up;
char **pending_key;
- gf_boolean_t data_self_heal; /* on/off */
+ char *data_self_heal; /* on/off/open */
char * data_self_heal_algorithm; /* name of algorithm */
unsigned int data_self_heal_window_size; /* max number of pipelined
read/writes */
unsigned int background_self_heal_count;
unsigned int background_self_heals_started;
- gf_boolean_t metadata_self_heal; /* on/off */
- gf_boolean_t entry_self_heal; /* on/off */
-
- gf_boolean_t data_change_log; /* on/off */
- gf_boolean_t metadata_change_log; /* on/off */
- gf_boolean_t entry_change_log; /* on/off */
+ gf_boolean_t metadata_self_heal; /* on/off */
+ gf_boolean_t entry_self_heal; /* on/off */
- int read_child; /* read-subvolume */
- unsigned int favorite_child; /* subvolume to be preferred in resolving
- split-brain cases */
+ gf_boolean_t data_change_log; /* on/off */
+ gf_boolean_t metadata_change_log; /* on/off */
+ gf_boolean_t entry_change_log; /* on/off */
- unsigned int data_lock_server_count;
- unsigned int metadata_lock_server_count;
- unsigned int entry_lock_server_count;
+ int read_child; /* read-subvolume */
+ unsigned int hash_mode; /* for when read_child is not set */
+ int favorite_child; /* subvolume to be preferred in resolving
+ split-brain cases */
- gf_boolean_t inodelk_trace;
- gf_boolean_t entrylk_trace;
+ gf_boolean_t inodelk_trace;
+ gf_boolean_t entrylk_trace;
- gf_boolean_t strict_readdir;
+ gf_boolean_t strict_readdir;
- unsigned int wait_count; /* # of servers to wait for success */
+ unsigned int wait_count; /* # of servers to wait for success */
uint64_t up_count; /* number of CHILD_UPs we have seen */
uint64_t down_count; /* number of CHILD_DOWNs we have seen */
- struct _pump_private *pump_private; /* Set if we are loaded as pump */
+ struct _pump_private *pump_private; /* Set if we are loaded as pump */
int use_afr_in_pump;
pthread_mutex_t mutex;
struct list_head saved_fds; /* list of fds on which locks have succeeded */
+ gf_boolean_t optimistic_change_log;
+ gf_boolean_t eager_lock;
+ uint32_t post_op_delay_secs;
+ unsigned int quorum_count;
+
+ char vol_uuid[UUID_SIZE + 1];
+ int32_t *last_event;
+ afr_self_heald_t shd;
+ gf_boolean_t choose_local;
+ gf_boolean_t did_discovery;
+ gf_boolean_t readdir_failover;
+ uint64_t sh_readdir_size;
+ gf_boolean_t ensure_durability;
+ char *sh_domain;
} afr_private_t;
+typedef enum {
+ AFR_SELF_HEAL_NOT_ATTEMPTED,
+ AFR_SELF_HEAL_STARTED,
+ AFR_SELF_HEAL_FAILED,
+ AFR_SELF_HEAL_SYNC_BEGIN,
+} afr_self_heal_status;
+
typedef struct {
+ afr_self_heal_status gfid_or_missing_entry_self_heal;
+ afr_self_heal_status metadata_self_heal;
+ afr_self_heal_status data_self_heal;
+ afr_self_heal_status entry_self_heal;
+} afr_sh_status_for_all_type;
+
+typedef enum {
+ AFR_SELF_HEAL_ENTRY,
+ AFR_SELF_HEAL_METADATA,
+ AFR_SELF_HEAL_DATA,
+ AFR_SELF_HEAL_GFID_OR_MISSING_ENTRY,
+ AFR_SELF_HEAL_INVALID = -1,
+} afr_self_heal_type;
+
+typedef enum {
+ AFR_CHECK_ALL,
+ AFR_CHECK_SPECIFIC,
+} afr_sh_fail_check_type;
+
+struct afr_self_heal_ {
/* External interface: These are variables (some optional) that
are set by whoever has triggered self-heal */
- gf_boolean_t need_data_self_heal;
- gf_boolean_t need_metadata_self_heal;
- gf_boolean_t need_entry_self_heal;
+ gf_boolean_t do_data_self_heal;
+ gf_boolean_t do_metadata_self_heal;
+ gf_boolean_t do_entry_self_heal;
+ gf_boolean_t do_gfid_self_heal;
+ gf_boolean_t do_missing_entry_self_heal;
+ gf_boolean_t force_confirm_spb; /* Check for split-brains even when
+ self-heal is turned off */
gf_boolean_t forced_merge; /* Is this a self-heal triggered to
forcibly merge the directories? */
- gf_boolean_t healing_fd_opened; /* true if caller has already
- opened fd */
-
- gf_boolean_t data_lock_held; /* true if caller has already
- acquired 0-0 lock */
-
- fd_t *healing_fd; /* set if callers has opened fd */
-
gf_boolean_t background; /* do self-heal in background
if possible */
-
ia_type_t type; /* st_mode of the entry we're doing
self-heal on */
+ inode_t *inode; /* inode on which the self-heal is
+ performed on */
+ uuid_t sh_gfid_req; /* gfid self-heal needs to be done
+ with this gfid if it is not null */
/* Function to call to unwind. If self-heal is being done in the
background, this function will be called as soon as possible. */
- int (*unwind) (call_frame_t *frame, xlator_t *this);
+ int (*unwind) (call_frame_t *frame, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, int32_t sh_failed);
/* End of external interface members */
- /* array of stat's, one for each child */
- struct iatt *buf;
+ /* array of stat's, one for each child */
+ struct iatt *buf;
+ struct iatt *parentbufs;
struct iatt parentbuf;
-
- /* array of xattr's, one for each child */
- dict_t **xattr;
-
- /* array of errno's, one for each child */
- int *child_errno;
-
- int32_t **pending_matrix;
- int32_t **delta_matrix;
-
- int *sources;
- int source;
- int active_source;
- int active_sinks;
- int *success;
- unsigned char *locked_nodes;
+ struct iatt entrybuf;
+
+ afr_expunge_done_cbk_t expunge_done;
+ afr_impunge_done_cbk_t impunge_done;
+
+ /* array of xattr's, one for each child */
+ dict_t **xattr;
+
+ /* array containing if the lookups succeeded in the order of response
+ */
+ int32_t *success_children;
+ int success_count;
+ /* array containing the fresh children found in the self-heal process */
+ int32_t *fresh_children;
+ /* array containing the fresh children found in the parent lookup */
+ int32_t *fresh_parent_dirs;
+ /* array of errno's, one for each child */
+ int *child_errno;
+ /*loc used for lookup*/
+ loc_t lookup_loc;
+ int32_t lookup_flags;
+ afr_lookup_done_cbk_t lookup_done;
+
+ int32_t **pending_matrix;
+ int32_t **delta_matrix;
+
+ int32_t op_ret;
+ int32_t op_errno;
+
+ int *sources;
+ int source;
+ int active_source;
+ int active_sinks;
+ unsigned char *success;
+ unsigned char *locked_nodes;
int lock_count;
- mode_t impunging_entry_mode;
const char *linkname;
-
- int op_failed;
-
- int file_has_holes;
- blksize_t block_size;
- off_t file_size;
- off_t offset;
-
- loc_t parent_loc;
-
+ gf_boolean_t entries_skipped;
+
+ gf_boolean_t actual_sh_started;
+ gf_boolean_t sync_done;
+ gf_boolean_t data_lock_held;
+ gf_boolean_t sh_dom_lock_held;
+ gf_boolean_t eof_reached;
+ fd_t *healing_fd;
+ int file_has_holes;
+ blksize_t block_size;
+ off_t file_size;
+ off_t offset;
+ unsigned char *write_needed;
+ uint8_t *checksum;
+ afr_post_remove_call_t post_remove_call;
+
+ char *data_sh_info;
+ char *metadata_sh_info;
+
+ loc_t parent_loc;
call_frame_t *orig_frame;
+ call_frame_t *old_loop_frame;
gf_boolean_t unwound;
- /* private data for the particular self-heal algorithm */
- void *private;
+ afr_sh_algo_private_t *private;
+ afr_sh_status_for_all_type afr_all_sh_status;
+ afr_self_heal_type sh_type_in_action;
- int (*flush_self_heal_cbk) (call_frame_t *frame, xlator_t *this);
-
- int (*completion_cbk) (call_frame_t *frame, xlator_t *this);
+ struct afr_sh_algorithm *algo;
+ afr_lock_cbk_t data_lock_success_handler;
+ afr_lock_cbk_t data_lock_failure_handler;
+ gf_boolean_t data_lock_block;
+ int (*completion_cbk) (call_frame_t *frame, xlator_t *this);
+ int (*sh_data_algo_start) (call_frame_t *frame, xlator_t *this);
int (*algo_completion_cbk) (call_frame_t *frame, xlator_t *this);
int (*algo_abort_cbk) (call_frame_t *frame, xlator_t *this);
+ void (*gfid_sh_success_cbk) (call_frame_t *sh_frame, xlator_t *this);
- call_frame_t *sh_frame;
-} afr_self_heal_t;
+ call_frame_t *sh_frame;
+};
+typedef struct afr_self_heal_ afr_self_heal_t;
typedef enum {
- AFR_DATA_TRANSACTION, /* truncate, write, ... */
- AFR_METADATA_TRANSACTION, /* chmod, chown, ... */
- AFR_ENTRY_TRANSACTION, /* create, rmdir, ... */
- AFR_ENTRY_RENAME_TRANSACTION, /* rename */
+ AFR_DATA_TRANSACTION, /* truncate, write, ... */
+ AFR_METADATA_TRANSACTION, /* chmod, chown, ... */
+ AFR_ENTRY_TRANSACTION, /* create, rmdir, ... */
+ AFR_ENTRY_RENAME_TRANSACTION, /* rename */
} afr_transaction_type;
typedef enum {
@@ -232,11 +385,31 @@ afr_index_for_transaction_type (afr_transaction_type type)
return -1; /* make gcc happy */
}
+typedef struct {
+ loc_t loc;
+ char *basename;
+ unsigned char *locked_nodes;
+ int locked_count;
+
+} afr_entry_lockee_t;
+
+int
+afr_entry_lockee_cmp (const void *l1, const void *l2);
+
+typedef struct {
+ char *domain; /* Domain on which inodelk is taken */
+ struct gf_flock flock;
+ unsigned char *locked_nodes;
+ int32_t lock_count;
+} afr_inodelk_t;
typedef struct {
loc_t *lk_loc;
- struct flock lk_flock;
+ int lockee_count;
+ afr_entry_lockee_t lockee[AFR_LOCKEE_COUNT_MAX];
+
+ afr_inodelk_t inodelk[AFR_DOM_COUNT_MAX];
const char *lk_basename;
const char *lower_basename;
const char *higher_basename;
@@ -245,24 +418,22 @@ typedef struct {
unsigned char *locked_nodes;
unsigned char *lower_locked_nodes;
- unsigned char *inode_locked_nodes;
- unsigned char *entry_locked_nodes;
selfheal_lk_type_t selfheal_lk_type;
transaction_lk_type_t transaction_lk_type;
int32_t lock_count;
- int32_t inodelk_lock_count;
int32_t entrylk_lock_count;
uint64_t lock_number;
int32_t lk_call_count;
+ int32_t lk_expected_count;
+ int32_t lk_attempted_count;
int32_t lock_op_ret;
int32_t lock_op_errno;
-
- int (*lock_cbk) (call_frame_t *, xlator_t *);
-
+ afr_lock_cbk_t lock_cbk;
+ char *domain; /* Domain on which inode/entry lock/unlock in progress.*/
} afr_internal_lock_t;
typedef struct _afr_locked_fd {
@@ -270,355 +441,355 @@ typedef struct _afr_locked_fd {
struct list_head list;
} afr_locked_fd_t;
+struct afr_reply {
+ int valid;
+ int32_t op_ret;
+ int32_t op_errno;
+};
+
typedef struct _afr_local {
- unsigned int call_count;
- unsigned int success_count;
- unsigned int enoent_count;
+ int uid;
+ int gid;
+ unsigned int call_count;
+ unsigned int success_count;
+ unsigned int enoent_count;
+ uint32_t open_fd_count;
+ gf_boolean_t update_open_fd_count;
- unsigned int govinda_gOvinda;
+ unsigned int unhealable;
- unsigned int read_child_index;
+ unsigned int read_child_index;
unsigned char read_child_returned;
unsigned int first_up_child;
- pid_t saved_pid;
+ gf_lkowner_t saved_lk_owner;
- int32_t op_ret;
- int32_t op_errno;
+ int32_t op_ret;
+ int32_t op_errno;
- int32_t **pending;
+ int32_t **pending;
- loc_t loc;
- loc_t newloc;
+ loc_t loc;
+ loc_t newloc;
- fd_t *fd;
+ fd_t *fd;
- glusterfs_fop_t fop;
+ glusterfs_fop_t fop;
- unsigned char *child_up;
+ unsigned char *child_up;
+ int32_t *fresh_children; //in the order of response
- int32_t *child_errno;
+ int32_t *child_errno;
- dict_t *xattr_req;
+ dict_t *xattr_req;
- int32_t inodelk_count;
- int32_t entrylk_count;
+ int32_t inodelk_count;
+ int32_t entrylk_count;
afr_internal_lock_t internal_lock;
+ afr_locked_fd_t *locked_fd;
+ int32_t source_child;
+ int32_t lock_recovery_child;
+
dict_t *dict;
+ int optimistic_change_log;
+ gf_boolean_t delayed_post_op;
- int (*openfd_flush_cbk) (call_frame_t *frame, xlator_t *this);
- /*
- This struct contains the arguments for the "continuation"
- (scheme-like) of fops
+ /* Is the current writev() going to perform a stable write?
+ i.e, is fd->flags or @flags writev param have O_SYNC or
+ O_DSYNC?
*/
+ gf_boolean_t stable_write;
- int op;
- struct {
- struct {
- unsigned char buf_set;
- struct statvfs buf;
- } statfs;
+ /* This write appended to the file. Nnot necessarily O_APPEND,
+ just means the offset of write was at the end of file.
+ */
+ gf_boolean_t append_write;
- struct {
- inode_t *inode;
- struct iatt buf;
- struct iatt read_child_buf;
+ int allow_sh_for_running_transaction;
+
+
+ /* This struct contains the arguments for the "continuation"
+ (scheme-like) of fops
+ */
+
+ int op;
+ struct {
+ struct {
+ unsigned char buf_set;
+ struct statvfs buf;
+ } statfs;
+
+ struct {
+ uint32_t parent_entrylk;
+ uuid_t gfid_req;
+ inode_t *inode;
+ struct iatt buf;
struct iatt postparent;
- ino_t ino;
- uint64_t gen;
- ino_t parent_ino;
- dict_t *xattr;
- dict_t **xattrs;
- gf_boolean_t is_revalidate;
- } lookup;
+ dict_t **xattrs;
+ dict_t *xattr;
+ struct iatt *postparents;
+ struct iatt *bufs;
+ int32_t read_child;
+ int32_t *sources;
+ int32_t *success_children;
+ int32_t **pending_matrix;
+ gf_boolean_t fresh_lookup;
+ gf_boolean_t possible_spb;
+ } lookup;
- struct {
- int32_t flags;
- int32_t wbflags;
- } open;
+ struct {
+ int32_t flags;
+ } open;
- struct {
- int32_t cmd;
- struct flock user_flock;
- struct flock ret_flock;
- unsigned char *locked_nodes;
- } lk;
+ struct {
+ int32_t cmd;
+ struct gf_flock user_flock;
+ struct gf_flock ret_flock;
+ unsigned char *locked_nodes;
+ } lk;
- /* inode read */
+ /* inode read */
- struct {
- int32_t mask;
- int last_tried; /* index of the child we tried previously */
- } access;
+ struct {
+ int32_t mask;
+ int last_index; /* index of the child we tried previously */
+ } access;
- struct {
- int last_tried;
- ino_t ino;
- } stat;
+ struct {
+ int last_index;
+ } stat;
- struct {
- int last_tried;
- ino_t ino;
- } fstat;
+ struct {
+ int last_index;
+ } fstat;
- struct {
- size_t size;
- int last_tried;
- ino_t ino;
- } readlink;
+ struct {
+ size_t size;
+ int last_index;
+ } readlink;
- struct {
- char *name;
- int last_tried;
- } getxattr;
+ struct {
+ char *name;
+ int last_index;
+ long xattr_len;
+ } getxattr;
- struct {
- ino_t ino;
- size_t size;
- off_t offset;
- int last_tried;
- } readv;
+ struct {
+ size_t size;
+ off_t offset;
+ int last_index;
+ uint32_t flags;
+ } readv;
- /* dir read */
+ /* dir read */
- struct {
- int success_count;
- int32_t op_ret;
- int32_t op_errno;
+ struct {
+ int success_count;
+ int32_t op_ret;
+ int32_t op_errno;
uint32_t *checksum;
- } opendir;
-
- struct {
- int32_t op_ret;
- int32_t op_errno;
- size_t size;
- off_t offset;
+ } opendir;
+ struct {
+ int32_t op_ret;
+ int32_t op_errno;
+ size_t size;
+ off_t offset;
+ dict_t *dict;
gf_boolean_t failed;
- int last_tried;
- } readdir;
+ int last_index;
+ } readdir;
+ /* inode write */
- struct {
- int32_t op_ret;
- int32_t op_errno;
+ struct {
+ struct iatt prebuf;
+ struct iatt postbuf;
+ } inode_wfop; //common structure for all inode-write-fops
- size_t size;
- off_t offset;
- int32_t flag;
+ struct {
+ int32_t op_ret;
- int last_tried;
- } getdents;
+ struct iovec *vector;
+ struct iobref *iobref;
+ int32_t count;
+ off_t offset;
+ uint32_t flags;
+ } writev;
- /* inode write */
+ struct {
+ off_t offset;
+ } truncate;
- struct {
- ino_t ino;
- struct iatt prebuf;
- struct iatt postbuf;
+ struct {
+ off_t offset;
+ } ftruncate;
- int32_t op_ret;
+ struct {
+ struct iatt in_buf;
+ int32_t valid;
+ } setattr;
- struct iovec *vector;
- struct iobref *iobref;
- int32_t count;
- off_t offset;
- } writev;
+ struct {
+ struct iatt in_buf;
+ int32_t valid;
+ } fsetattr;
struct {
- ino_t ino;
- struct iatt prebuf;
- struct iatt postbuf;
- } fsync;
+ dict_t *dict;
+ int32_t flags;
+ } setxattr;
- struct {
- ino_t ino;
- off_t offset;
- struct iatt prebuf;
- struct iatt postbuf;
- } truncate;
+ struct {
+ dict_t *dict;
+ int32_t flags;
+ } fsetxattr;
- struct {
- ino_t ino;
- off_t offset;
- struct iatt prebuf;
- struct iatt postbuf;
- } ftruncate;
+ struct {
+ char *name;
+ } removexattr;
- struct {
- ino_t ino;
- struct iatt in_buf;
- int32_t valid;
- struct iatt preop_buf;
- struct iatt postop_buf;
- } setattr;
+ struct {
+ dict_t *xattr;
+ } xattrop;
- struct {
- ino_t ino;
- struct iatt in_buf;
- int32_t valid;
- struct iatt preop_buf;
- struct iatt postop_buf;
- } fsetattr;
+ struct {
+ dict_t *xattr;
+ } fxattrop;
- struct {
- dict_t *dict;
- int32_t flags;
- } setxattr;
+ /* dir write */
- struct {
- char *name;
- } removexattr;
+ struct {
+ inode_t *inode;
+ struct iatt buf;
+ struct iatt preparent;
+ struct iatt postparent;
+ struct iatt prenewparent;
+ struct iatt postnewparent;
+ } dir_fop; //common structure for all dir fops
- /* dir write */
+ struct {
+ fd_t *fd;
+ dict_t *params;
+ int32_t flags;
+ mode_t mode;
+ } create;
- struct {
- ino_t ino;
- uint64_t gen;
- ino_t parent_ino;
- fd_t *fd;
+ struct {
+ dev_t dev;
+ mode_t mode;
dict_t *params;
- int32_t flags;
- mode_t mode;
- inode_t *inode;
- struct iatt buf;
- struct iatt preparent;
- struct iatt postparent;
- struct iatt read_child_buf;
- } create;
+ } mknod;
- struct {
- ino_t ino;
- uint64_t gen;
- ino_t parent_ino;
- dev_t dev;
- mode_t mode;
+ struct {
+ int32_t mode;
dict_t *params;
- inode_t *inode;
- struct iatt buf;
- struct iatt preparent;
- struct iatt postparent;
- struct iatt read_child_buf;
- } mknod;
+ } mkdir;
- struct {
- ino_t ino;
- uint64_t gen;
- ino_t parent_ino;
- int32_t mode;
+ struct {
+ int flags;
+ } rmdir;
+
+ struct {
dict_t *params;
- inode_t *inode;
- struct iatt buf;
- struct iatt read_child_buf;
- struct iatt preparent;
- struct iatt postparent;
- } mkdir;
+ char *linkpath;
+ } symlink;
struct {
- ino_t parent_ino;
- int32_t op_ret;
- int32_t op_errno;
- struct iatt preparent;
- struct iatt postparent;
- } unlink;
+ int32_t mode;
+ off_t offset;
+ size_t len;
+ } fallocate;
struct {
- ino_t parent_ino;
- int32_t op_ret;
- int32_t op_errno;
- struct iatt preparent;
- struct iatt postparent;
- } rmdir;
+ off_t offset;
+ size_t len;
+ } discard;
- struct {
- ino_t oldparent_ino;
- ino_t newparent_ino;
- ino_t ino;
- struct iatt buf;
- struct iatt read_child_buf;
- struct iatt preoldparent;
- struct iatt prenewparent;
- struct iatt postoldparent;
- struct iatt postnewparent;
- } rename;
+ struct {
+ off_t offset;
+ size_t len;
+ struct iatt prebuf;
+ struct iatt postbuf;
+ } zerofill;
- struct {
- ino_t ino;
- uint64_t gen;
- ino_t parent_ino;
- inode_t *inode;
- struct iatt buf;
- struct iatt read_child_buf;
- struct iatt preparent;
- struct iatt postparent;
- } link;
- struct {
- ino_t ino;
- uint64_t gen;
- ino_t parent_ino;
- inode_t *inode;
- dict_t *params;
- struct iatt buf;
- struct iatt read_child_buf;
- char *linkpath;
- struct iatt preparent;
- struct iatt postparent;
- } symlink;
+ } cont;
- struct {
- int32_t flags;
- dir_entry_t *entries;
- int32_t count;
- } setdents;
- } cont;
+ struct {
+ off_t start, len;
- struct {
- off_t start, len;
+ gf_boolean_t eager_lock_on;
+ int *eager_lock;
- char *basename;
- char *new_basename;
+ char *basename;
+ char *new_basename;
- loc_t parent_loc;
- loc_t new_parent_loc;
+ loc_t parent_loc;
+ loc_t new_parent_loc;
- afr_transaction_type type;
+ afr_transaction_type type;
- int success_count;
- int erase_pending;
- int failure_count;
+ /* pre-compute the post piggyback status before
+ entering POST-OP phase
+ */
+ int *postop_piggybacked;
- int last_tried;
- int32_t *child_errno;
+ /* stub to resume on destruction
+ of the transaction frame */
+ call_stub_t *resume_stub;
- call_frame_t *main_frame;
+ struct list_head eager_locked;
- int (*fop) (call_frame_t *frame, xlator_t *this);
+ int32_t **txn_changelog;//changelog after pre+post ops
+ unsigned char *pre_op;
- int (*done) (call_frame_t *frame, xlator_t *this);
+ call_frame_t *main_frame;
- int (*resume) (call_frame_t *frame, xlator_t *this);
+ int (*fop) (call_frame_t *frame, xlator_t *this);
- int (*unwind) (call_frame_t *frame, xlator_t *this);
+ int (*done) (call_frame_t *frame, xlator_t *this);
+
+ int (*resume) (call_frame_t *frame, xlator_t *this);
+
+ int (*unwind) (call_frame_t *frame, xlator_t *this);
/* post-op hook */
- } transaction;
+ } transaction;
+
+ afr_self_heal_t self_heal;
+
+ struct marker_str marker;
- afr_self_heal_t self_heal;
+ /* extra data for fops */
+ dict_t *xdata_req;
+ dict_t *xdata_rsp;
+
+ mode_t umask;
+ int xflag;
+ gf_boolean_t do_discovery;
+ struct afr_reply *replies;
} afr_local_t;
+typedef enum {
+ AFR_FD_NOT_OPENED,
+ AFR_FD_OPENED,
+ AFR_FD_OPENING
+} afr_fd_open_status_t;
typedef struct {
unsigned int *pre_op_done;
- unsigned int *opened_on; /* which subvolumes the fd is open on */
+ afr_fd_open_status_t *opened_on; /* which subvolumes the fd is open on */
unsigned int *pre_op_piggyback;
+ unsigned int *lock_piggyback;
+ unsigned int *lock_acquired;
+
int flags;
- int32_t wbflags;
uint64_t up_count; /* number of CHILD_UPs this fd has seen */
uint64_t down_count; /* number of CHILD_DOWNs this fd has seen */
@@ -629,26 +800,39 @@ typedef struct {
struct list_head entries; /* needed for readdir failover */
unsigned char *locked_on; /* which subvolumes locks have been successful */
+
+ /* used for delayed-post-op optimization */
+ pthread_mutex_t delay_lock;
+ gf_timer_t *delay_timer;
+ call_frame_t *delay_frame;
+ int call_child;
+
+ /* set if any write on this fd was a non stable write
+ (i.e, without O_SYNC or O_DSYNC)
+ */
+ gf_boolean_t witnessed_unstable_write;
+
+ /* list of frames currently in progress */
+ struct list_head eager_locked;
} afr_fd_ctx_t;
/* try alloc and if it fails, goto label */
-#define ALLOC_OR_GOTO(var, type, label) do { \
- var = GF_CALLOC (sizeof (type), 1, \
- gf_afr_mt_##type); \
- if (!var) { \
- gf_log (this->name, GF_LOG_ERROR, \
- "out of memory :("); \
- op_errno = ENOMEM; \
- goto label; \
- } \
- } while (0);
+#define AFR_LOCAL_ALLOC_OR_GOTO(var, label) do { \
+ var = mem_get0 (THIS->local_pool); \
+ if (!var) { \
+ gf_log (this->name, GF_LOG_ERROR, \
+ "out of memory :("); \
+ op_errno = ENOMEM; \
+ goto label; \
+ } \
+ } while (0);
/* did a call fail due to a child failing? */
-#define child_went_down(op_ret, op_errno) (((op_ret) < 0) && \
- ((op_errno == ENOTCONN) || \
- (op_errno == EBADFD)))
+#define child_went_down(op_ret, op_errno) (((op_ret) < 0) && \
+ ((op_errno == ENOTCONN) || \
+ (op_errno == EBADFD)))
#define afr_fop_failed(op_ret, op_errno) ((op_ret) == -1)
@@ -662,8 +846,17 @@ int
pump_command_reply (call_frame_t *frame, xlator_t *this);
int32_t
-afr_notify (xlator_t *this, int32_t event,
- void *data, ...);
+afr_notify (xlator_t *this, int32_t event, void *data, void *data2);
+
+int
+afr_init_entry_lockee (afr_entry_lockee_t *lockee, afr_local_t *local,
+ loc_t *loc, char *basename, int child_count);
+
+void
+afr_entry_lockee_cleanup (afr_internal_lock_t *int_lock);
+
+int
+afr_attempt_lock_recovery (xlator_t *this, int32_t child_index);
int
afr_save_locked_fd (xlator_t *this, fd_t *fd);
@@ -673,7 +866,7 @@ afr_mark_locked_nodes (xlator_t *this, fd_t *fd,
unsigned char *locked_nodes);
void
-afr_set_lk_owner (call_frame_t *frame, xlator_t *this);
+afr_set_lk_owner (call_frame_t *frame, xlator_t *this, void *lk_owner);
int
afr_set_lock_number (call_frame_t *frame, xlator_t *this);
@@ -697,32 +890,46 @@ afr_blocking_lock (call_frame_t *frame, xlator_t *this);
int
afr_internal_lock_finish (call_frame_t *frame, xlator_t *this);
+int
+afr_lk_transfer_datalock (call_frame_t *dst, call_frame_t *src, char *dom,
+ unsigned int child_count);
int pump_start (call_frame_t *frame, xlator_t *this);
int
-afr_fd_ctx_set (xlator_t *this, fd_t *fd);
+__afr_fd_ctx_set (xlator_t *this, fd_t *fd);
-uint64_t
-afr_read_child (xlator_t *this, inode_t *inode);
+int
+afr_fd_ctx_set (xlator_t *this, fd_t *fd);
-void
-afr_set_read_child (xlator_t *this, inode_t *inode, int32_t read_child);
+int32_t
+afr_inode_get_read_ctx (xlator_t *this, inode_t *inode, int32_t *fresh_children);
void
-afr_build_parent_loc (loc_t *parent, loc_t *child);
+afr_inode_set_read_ctx (xlator_t *this, inode_t *inode, int32_t read_child,
+ int32_t *fresh_children);
int
-afr_up_children_count (int child_count, unsigned char *child_up);
+afr_build_parent_loc (loc_t *parent, loc_t *child, int32_t *op_errno);
-int
-afr_locked_nodes_count (unsigned char *locked_nodes, int child_count);
+unsigned int
+afr_up_children_count (unsigned char *child_up, unsigned int child_count);
+
+unsigned int
+afr_locked_children_count (unsigned char *children, unsigned int child_count);
+
+unsigned int
+afr_pre_op_done_children_count (unsigned char *pre_op,
+ unsigned int child_count);
-ino64_t
-afr_itransform (ino64_t ino, int child_count, int child_index);
+gf_boolean_t
+afr_is_fresh_lookup (loc_t *loc, xlator_t *this);
+
+void
+afr_update_loc_gfids (loc_t *loc, struct iatt *buf, struct iatt *postparent);
int
-afr_deitransform (ino64_t ino, int child_count);
+afr_locked_nodes_count (unsigned char *locked_nodes, int child_count);
void
afr_local_cleanup (afr_local_t *local, xlator_t *this);
@@ -730,20 +937,21 @@ afr_local_cleanup (afr_local_t *local, xlator_t *this);
int
afr_frame_return (call_frame_t *frame);
-uint64_t
+gf_boolean_t
afr_is_split_brain (xlator_t *this, inode_t *inode);
void
-afr_set_split_brain (xlator_t *this, inode_t *inode, gf_boolean_t set);
+afr_set_split_brain (xlator_t *this, inode_t *inode, afr_spb_state_t mdata_spb,
+ afr_spb_state_t data_spb);
int
afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- fd_t *fd, int32_t wbflags);
+ fd_t *fd, dict_t *xdata);
void
afr_set_opendir_done (xlator_t *this, inode_t *inode);
-uint64_t
+gf_boolean_t
afr_is_opendir_done (xlator_t *this, inode_t *inode);
void
@@ -753,162 +961,252 @@ int
afr_cleanup_fd_ctx (xlator_t *this, fd_t *fd);
int
-afr_openfd_flush (call_frame_t *frame, xlator_t *this, fd_t *fd);
-
-#define AFR_STACK_UNWIND(fop, frame, params ...) \
- do { \
- afr_local_t *__local = NULL; \
- xlator_t *__this = NULL; \
- if (frame) { \
- __local = frame->local; \
- __this = frame->this; \
- frame->local = NULL; \
+afr_launch_openfd_self_heal (call_frame_t *frame, xlator_t *this, fd_t *fd);
+
+#define AFR_STACK_UNWIND(fop, frame, params ...) \
+ do { \
+ afr_local_t *__local = NULL; \
+ xlator_t *__this = NULL; \
+ if (frame) { \
+ __local = frame->local; \
+ __this = frame->this; \
+ frame->local = NULL; \
+ } \
+ STACK_UNWIND_STRICT (fop, frame, params); \
+ if (__local) { \
+ afr_local_cleanup (__local, __this); \
+ mem_put (__local); \
+ } \
+ } while (0)
+
+#define AFR_STACK_DESTROY(frame) \
+ do { \
+ afr_local_t *__local = NULL; \
+ xlator_t *__this = NULL; \
+ __local = frame->local; \
+ __this = frame->this; \
+ frame->local = NULL; \
+ STACK_DESTROY (frame->root); \
+ if (__local) { \
+ afr_local_cleanup (__local, __this); \
+ mem_put (__local); \
} \
- STACK_UNWIND_STRICT (fop, frame, params); \
- afr_local_cleanup (__local, __this); \
- GF_FREE (__local); \
- } while (0);
-
-#define AFR_STACK_DESTROY(frame) \
- do { \
- afr_local_t *__local = NULL; \
- xlator_t *__this = NULL; \
- __local = frame->local; \
- __this = frame->this; \
- frame->local = NULL; \
- STACK_DESTROY (frame->root); \
- afr_local_cleanup (__local, __this); \
- GF_FREE (__local); \
} while (0);
+#define AFR_NUM_CHANGE_LOGS 3 /*data + metadata + entry*/
/* allocate and return a string that is the basename of argument */
static inline char *
AFR_BASENAME (const char *str)
{
- char *__tmp_str = NULL;
- char *__basename_str = NULL;
- __tmp_str = gf_strdup (str);
- __basename_str = gf_strdup (basename (__tmp_str));
- GF_FREE (__tmp_str);
- return __basename_str;
+ char *__tmp_str = NULL;
+ char *__basename_str = NULL;
+ __tmp_str = gf_strdup (str);
+ __basename_str = gf_strdup (basename (__tmp_str));
+ GF_FREE (__tmp_str);
+ return __basename_str;
}
-/* initialize local_t */
-static inline int
-AFR_LOCAL_INIT (afr_local_t *local, afr_private_t *priv)
-{
- local->child_up = GF_CALLOC (sizeof (*local->child_up),
- priv->child_count,
- gf_afr_mt_char);
- if (!local->child_up) {
- return -ENOMEM;
- }
+int
+afr_transaction_local_init (afr_local_t *local, xlator_t *this);
+
+int32_t
+afr_marker_getxattr (call_frame_t *frame, xlator_t *this,
+ loc_t *loc, const char *name,afr_local_t *local, afr_private_t *priv );
- memcpy (local->child_up, priv->child_up,
- sizeof (*local->child_up) * priv->child_count);
+int32_t *
+afr_children_create (int32_t child_count);
+int
+afr_local_init (afr_local_t *local, afr_private_t *priv, int32_t *op_errno);
- local->call_count = afr_up_children_count (priv->child_count, local->child_up);
- if (local->call_count == 0)
- return -ENOTCONN;
+int
+afr_internal_lock_init (afr_internal_lock_t *lk, size_t child_count,
+ transaction_lk_type_t lk_type);
- local->transaction.erase_pending = 1;
+int
+afr_first_up_child (unsigned char *child_up, size_t child_count);
- local->op_ret = -1;
- local->op_errno = EUCLEAN;
+int
+afr_select_read_child_from_policy (int32_t *fresh_children, int32_t child_count,
+ int32_t prev_read_child,
+ int32_t config_read_child, int32_t *sources,
+ unsigned int hmode, uuid_t gfid);
- local->internal_lock.lock_op_ret = -1;
- local->internal_lock.lock_op_errno = EUCLEAN;
+void
+afr_set_read_ctx_from_policy (xlator_t *this, inode_t *inode,
+ int32_t *fresh_children, int32_t prev_read_child,
+ int32_t config_read_child, uuid_t gfid);
+int32_t
+afr_get_call_child (xlator_t *this, unsigned char *child_up, int32_t read_child,
+ int32_t *fresh_children,
+ int32_t *call_child, int32_t *last_index);
- return 0;
-}
+int32_t
+afr_next_call_child (int32_t *fresh_children, unsigned char *child_up,
+ size_t child_count, int32_t *last_index,
+ int32_t read_child);
+void
+afr_get_fresh_children (int32_t *success_children, int32_t *sources,
+ int32_t *children, unsigned int child_count);
+void
+afr_children_add_child (int32_t *children, int32_t child,
+ int32_t child_count);
+void
+afr_children_rm_child (int32_t *children, int32_t child,
+ int32_t child_count);
+void
+afr_reset_children (int32_t *children, int32_t child_count);
+int32_t
+afr_most_important_error(int32_t old_errno, int32_t new_errno,
+ gf_boolean_t eio);
+int
+afr_errno_count (int32_t *children, int *child_errno,
+ unsigned int child_count, int32_t op_errno);
+int
+afr_get_children_count (int32_t *children, unsigned int child_count);
+gf_boolean_t
+afr_is_child_present (int32_t *success_children, int32_t child_count,
+ int32_t child);
+void
+afr_update_gfid_from_iatts (uuid_t uuid, struct iatt *bufs,
+ int32_t *success_children,
+ unsigned int child_count);
+void
+afr_reset_xattr (dict_t **xattr, unsigned int child_count);
+gf_boolean_t
+afr_conflicting_iattrs (struct iatt *bufs, int32_t *success_children,
+ unsigned int child_count, const char *path,
+ const char *xlator_name);
+unsigned int
+afr_gfid_missing_count (const char *xlator_name, int32_t *children,
+ struct iatt *bufs, unsigned int child_count,
+ const char *path);
+void
+afr_xattr_req_prepare (xlator_t *this, dict_t *xattr_req, const char *path);
+void
+afr_children_copy (int32_t *dst, int32_t *src, unsigned int child_count);
+afr_transaction_type
+afr_transaction_type_get (ia_type_t ia_type);
+int32_t
+afr_resultant_errno_get (int32_t *children,
+ int *child_errno, unsigned int child_count);
+void
+afr_inode_rm_stale_children (xlator_t *this, inode_t *inode,
+ int32_t *stale_children);
+void
+afr_launch_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode,
+ gf_boolean_t background, ia_type_t ia_type, char *reason,
+ void (*gfid_sh_success_cbk) (call_frame_t *sh_frame,
+ xlator_t *this),
+ int (*unwind) (call_frame_t *frame, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ int32_t sh_failed));
+void
+afr_fix_open (xlator_t *this, fd_t *fd, size_t need_open_count, int *need_open);
+void
+afr_open_fd_fix (fd_t *fd, xlator_t *this);
+int
+afr_set_elem_count_get (unsigned char *elems, int child_count);
-/**
- * first_up_child - return the index of the first child that is up
- */
+afr_fd_ctx_t *
+afr_fd_ctx_get (fd_t *fd, xlator_t *this);
-static inline int
-afr_first_up_child (afr_private_t *priv)
-{
- xlator_t ** children = NULL;
- int ret = -1;
- int i = 0;
-
- LOCK (&priv->lock);
- {
- children = priv->children;
- for (i = 0; i < priv->child_count; i++) {
- if (priv->child_up[i]) {
- ret = i;
- break;
- }
- }
- }
- UNLOCK (&priv->lock);
-
- return ret;
-}
+gf_boolean_t
+afr_open_only_data_self_heal (char *data_self_heal);
+gf_boolean_t
+afr_data_self_heal_enabled (char *data_self_heal);
-static inline int
-afr_transaction_local_init (afr_local_t *local, afr_private_t *priv)
-{
- int i;
-
- local->first_up_child = afr_first_up_child (priv);
-
- local->child_errno = GF_CALLOC (sizeof (*local->child_errno),
- priv->child_count,
- gf_afr_mt_int32_t);
- if (!local->child_errno) {
- return -ENOMEM;
- }
-
- local->pending = GF_CALLOC (sizeof (*local->pending),
- priv->child_count,
- gf_afr_mt_int32_t);
-
- if (!local->pending) {
- return -ENOMEM;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- local->pending[i] = GF_CALLOC (sizeof (*local->pending[i]),
- 3, /* data + metadata + entry */
- gf_afr_mt_int32_t);
- if (!local->pending[i])
- return -ENOMEM;
- }
+void
+afr_set_low_priority (call_frame_t *frame);
+int
+afr_child_fd_ctx_set (xlator_t *this, fd_t *fd, int32_t child,
+ int flags);
- local->internal_lock.inode_locked_nodes =
- GF_CALLOC (sizeof (*local->internal_lock.inode_locked_nodes),
- priv->child_count,
- gf_afr_mt_char);
+gf_boolean_t
+afr_have_quorum (char *logname, afr_private_t *priv);
- local->internal_lock.entry_locked_nodes =
- GF_CALLOC (sizeof (*local->internal_lock.entry_locked_nodes),
- priv->child_count,
- gf_afr_mt_char);
+void
+afr_matrix_cleanup (int32_t **pending, unsigned int m);
- local->internal_lock.locked_nodes =
- GF_CALLOC (sizeof (*local->internal_lock.locked_nodes),
- priv->child_count,
- gf_afr_mt_char);
+int32_t**
+afr_matrix_create (unsigned int m, unsigned int n);
- local->internal_lock.lower_locked_nodes
- = GF_CALLOC (sizeof (*local->internal_lock.lower_locked_nodes),
- priv->child_count,
- gf_afr_mt_char);
+gf_boolean_t
+afr_is_errno_set (int *child_errno, int child);
- local->transaction.child_errno = GF_CALLOC (sizeof (*local->transaction.child_errno),
- priv->child_count,
- gf_afr_mt_int32_t);
+gf_boolean_t
+afr_is_errno_unset (int *child_errno, int child);
- local->internal_lock.transaction_lk_type = AFR_TRANSACTION_LK;
+gf_boolean_t
+afr_is_fd_fixable (fd_t *fd);
- return 0;
-}
+void
+afr_prepare_new_entry_pending_matrix (int32_t **pending,
+ gf_boolean_t (*is_pending) (int *, int),
+ int *ctx, struct iatt *buf,
+ unsigned int child_count);
+void
+afr_xattr_array_destroy (dict_t **xattr, unsigned int child_count);
+/*
+ * Special value indicating we should use the "auto" quorum method instead of
+ * a fixed value (including zero to turn off quorum enforcement).
+ */
+#define AFR_QUORUM_AUTO INT_MAX
+
+/*
+ * Having this as a macro will make debugging a bit weirder, but does reduce
+ * the probability of functions handling this check inconsistently.
+ */
+#define QUORUM_CHECK(_func,_label) do { \
+ if (priv->quorum_count && !afr_have_quorum(this->name,priv)) { \
+ gf_log(this->name,GF_LOG_WARNING, \
+ "failing "#_func" due to lack of quorum"); \
+ op_errno = EROFS; \
+ goto _label; \
+ } \
+} while (0);
+
+
+#define AFR_SBRAIN_MSG "Failed on %s as split-brain is seen. Returning EIO."
+
+#define AFR_SBRAIN_CHECK_FD(fd, label) do { \
+ if (fd->inode && afr_is_split_brain (this, fd->inode)) { \
+ op_errno = EIO; \
+ gf_log (this->name, GF_LOG_WARNING, \
+ AFR_SBRAIN_MSG ,uuid_utoa (fd->inode->gfid)); \
+ goto label; \
+ } \
+} while (0)
+
+#define AFR_SBRAIN_CHECK_LOC(loc, label) do { \
+ if (loc->inode && afr_is_split_brain (this, loc->inode)) { \
+ op_errno = EIO; \
+ loc_path (loc, NULL); \
+ gf_log (this->name, GF_LOG_WARNING, \
+ AFR_SBRAIN_MSG , loc->path); \
+ goto label; \
+ } \
+} while (0)
+
+int
+afr_fd_report_unstable_write (xlator_t *this, fd_t *fd);
+
+gf_boolean_t
+afr_fd_has_witnessed_unstable_write (xlator_t *this, fd_t *fd);
+
+void
+afr_delayed_changelog_wake_resume (xlator_t *this, fd_t *fd, call_stub_t *stub);
+
+int
+afr_inodelk_init (afr_inodelk_t *lk, char *dom, size_t child_count);
+
+void
+afr_handle_open_fd_count (call_frame_t *frame, xlator_t *this);
+
+afr_inode_ctx_t*
+afr_inode_ctx_get (inode_t *inode, xlator_t *this);
#endif /* __AFR_H__ */
diff --git a/xlators/cluster/afr/src/pump.c b/xlators/cluster/afr/src/pump.c
index 45ae7db08..a7f72fb30 100644
--- a/xlators/cluster/afr/src/pump.c
+++ b/xlators/cluster/afr/src/pump.c
@@ -1,25 +1,17 @@
/*
- Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include <unistd.h>
#include <sys/time.h>
#include <stdlib.h>
+#include <fnmatch.h>
#ifndef _CONFIG_H
#define _CONFIG_H
@@ -28,6 +20,15 @@
#include "afr-common.c"
#include "defaults.c"
+#include "glusterfs.h"
+
+static uint64_t pump_pid = 0;
+static inline void
+pump_fill_loc_info (loc_t *loc, struct iatt *iatt, struct iatt *parent)
+{
+ afr_update_loc_gfids (loc, iatt, parent);
+ uuid_copy (loc->inode->gfid, iatt->ia_gfid);
+}
static int
pump_mark_start_pending (xlator_t *this)
@@ -135,92 +136,24 @@ pump_set_resume_path (xlator_t *this, const char *path)
priv = this->private;
pump_priv = priv->pump_private;
- assert (pump_priv);
+ GF_ASSERT (pump_priv);
LOCK (&pump_priv->resume_path_lock);
{
- pump_priv->resume_path = strdup (path);
- if (!pump_priv->resume_path)
- ret = -1;
+ strncpy (pump_priv->resume_path, path, strlen (path) + 1);
}
UNLOCK (&pump_priv->resume_path_lock);
return ret;
}
-static void
-build_child_loc (loc_t *parent, loc_t *child, char *path, char *name)
-{
- child->path = path;
- child->name = name;
-
- child->parent = inode_ref (parent->inode);
- child->inode = inode_new (parent->inode->table);
-}
-
-static char *
-build_file_path (loc_t *loc, gf_dirent_t *entry)
-{
- xlator_t *this = NULL;
- char *file_path = NULL;
- int pathlen = 0;
- int total_size = 0;
-
- this = THIS;
-
- pathlen = STRLEN_0 (loc->path);
-
- if (IS_ROOT_PATH (loc->path)) {
- total_size = pathlen + entry->d_len;
- file_path = GF_CALLOC (1, total_size, gf_afr_mt_char);
- if (!file_path) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- return NULL;
- }
-
- gf_log (this->name, GF_LOG_TRACE,
- "constructing file path of size=%d"
- "pathlen=%d, d_len=%d",
- total_size, pathlen,
- entry->d_len);
-
- snprintf(file_path, total_size, "%s%s", loc->path, entry->d_name);
-
- } else {
- total_size = pathlen + entry->d_len + 1; /* for the extra '/' in the path */
- file_path = GF_CALLOC (1, total_size + 1, gf_afr_mt_char);
- if (!file_path) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- return NULL;
- }
-
- gf_log (this->name, GF_LOG_TRACE,
- "constructing file path of size=%d"
- "pathlen=%d, d_len=%d",
- total_size, pathlen,
- entry->d_len);
-
- snprintf(file_path, total_size, "%s/%s", loc->path, entry->d_name);
- }
-
- gf_log (this->name, GF_LOG_TRACE,
- "path=%s and d_name=%s", loc->path, entry->d_name);
- gf_log (this->name, GF_LOG_TRACE,
- "constructed file_path=%s of size=%d", file_path, total_size);
-
- return file_path;
-}
-
static int
pump_save_path (xlator_t *this, const char *path)
{
afr_private_t *priv = NULL;
- pump_private_t *pump_priv = NULL;
pump_state_t state;
dict_t *dict = NULL;
- loc_t loc;
+ loc_t loc = {0};
int dict_ret = 0;
int ret = -1;
@@ -229,29 +162,30 @@ pump_save_path (xlator_t *this, const char *path)
return 0;
priv = this->private;
- pump_priv = priv->pump_private;
- assert (priv->root_inode);
+ GF_ASSERT (priv->root_inode);
- build_root_loc (priv->root_inode, &loc);
+ afr_build_root_loc (this, &loc);
dict = dict_new ();
dict_ret = dict_set_str (dict, PUMP_PATH, (char *)path);
+ if (dict_ret)
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: failed to set the key %s", path, PUMP_PATH);
ret = syncop_setxattr (PUMP_SOURCE_CHILD (this), &loc, dict, 0);
if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
+ gf_log (this->name, GF_LOG_INFO,
"setxattr failed - could not save path=%s", path);
} else {
gf_log (this->name, GF_LOG_DEBUG,
"setxattr succeeded - saved path=%s", path);
- gf_log (this->name, GF_LOG_DEBUG,
- "Saving path for status info");
}
dict_unref (dict);
+ loc_wipe (&loc);
return 0;
}
@@ -314,15 +248,9 @@ pump_get_resume_path (xlator_t *this)
static int
pump_update_resume_state (xlator_t *this, const char *path)
{
- afr_private_t *priv = NULL;
- pump_private_t *pump_priv = NULL;
-
pump_state_t state;
const char *resume_path = NULL;
- priv = this->private;
- pump_priv = priv->pump_private;
-
state = pump_get_state ();
if (state == PUMP_STATE_RESUME) {
@@ -350,16 +278,10 @@ pump_update_resume_state (xlator_t *this, const char *path)
static gf_boolean_t
is_pump_traversal_allowed (xlator_t *this, const char *path)
{
- afr_private_t *priv = NULL;
- pump_private_t *pump_priv = NULL;
-
pump_state_t state;
const char *resume_path = NULL;
gf_boolean_t ret = _gf_true;
- priv = this->private;
- pump_priv = priv->pump_private;
-
state = pump_get_state ();
if (state == PUMP_STATE_RESUME) {
@@ -402,38 +324,33 @@ pump_save_file_stats (xlator_t *this, const char *path)
static int
gf_pump_traverse_directory (loc_t *loc)
{
- xlator_t *this = NULL;
- afr_private_t *priv = NULL;
- fd_t *fd = NULL;
-
- off_t offset = 0;
- loc_t entry_loc;
- gf_dirent_t *entry = NULL;
- gf_dirent_t *tmp = NULL;
- gf_dirent_t entries;
-
- struct iatt iatt, parent;
- dict_t *xattr_rsp;
-
- int source = 0;
-
- char *file_path = NULL;
- int ret = 0;
+ xlator_t *this = NULL;
+ fd_t *fd = NULL;
+ off_t offset = 0;
+ loc_t entry_loc = {0};
+ gf_dirent_t *entry = NULL;
+ gf_dirent_t *tmp = NULL;
+ gf_dirent_t entries;
+ struct iatt iatt = {0};
+ struct iatt parent = {0};
+ dict_t *xattr_rsp = NULL;
+ int ret = 0;
+ gf_boolean_t is_directory_empty = _gf_true;
+ gf_boolean_t free_entries = _gf_false;
INIT_LIST_HEAD (&entries.list);
this = THIS;
- priv = this->private;
- assert (loc->inode);
+ GF_ASSERT (loc->inode);
- fd = fd_create (loc->inode, PUMP_PID);
+ fd = fd_create (loc->inode, pump_pid);
if (!fd) {
gf_log (this->name, GF_LOG_ERROR,
"Failed to create fd for %s", loc->path);
goto out;
}
- ret = syncop_opendir (priv->children[source], loc, fd);
+ ret = syncop_opendir (this, loc, fd);
if (ret < 0) {
gf_log (this->name, GF_LOG_DEBUG,
"opendir failed on %s", loc->path);
@@ -444,7 +361,8 @@ gf_pump_traverse_directory (loc_t *loc)
"pump opendir on %s returned=%d",
loc->path, ret);
- while (syncop_readdirp (priv->children[source], fd, 131072, offset, &entries)) {
+ while (syncop_readdirp (this, fd, 131072, offset, NULL, &entries)) {
+ free_entries = _gf_true;
if (list_empty (&entries.list)) {
gf_log (this->name, GF_LOG_TRACE,
@@ -456,110 +374,94 @@ gf_pump_traverse_directory (loc_t *loc)
gf_log (this->name, GF_LOG_DEBUG,
"found readdir entry=%s", entry->d_name);
- if ((strcmp (entry->d_name, ".") == 0) ||
- (strcmp (entry->d_name, "..") == 0))
+ offset = entry->d_off;
+ if (uuid_is_null (entry->d_stat.ia_gfid)) {
+ gf_log (this->name, GF_LOG_WARNING, "%s/%s: No "
+ "gfid present skipping",
+ loc->path, entry->d_name);
continue;
-
- file_path = build_file_path (loc, entry);
- if (!file_path) {
- gf_log (this->name, GF_LOG_DEBUG,
- "file path construction failed");
- goto out;
- }
-
- build_child_loc (loc, &entry_loc, file_path, entry->d_name);
-
- ret = syncop_lookup (this, &entry_loc, NULL,
- &iatt, &xattr_rsp, &parent);
-
- entry_loc.ino = iatt.ia_ino;
- entry_loc.inode->ino = iatt.ia_ino;
- memcpy (entry_loc.inode->gfid, iatt.ia_gfid, 16);
-
- gf_log (this->name, GF_LOG_DEBUG,
- "lookup %s => %"PRId64,
- entry_loc.path,
- iatt.ia_ino);
-
- ret = syncop_lookup (this, &entry_loc, NULL,
- &iatt, &xattr_rsp, &parent);
-
-
- gf_log (this->name, GF_LOG_DEBUG,
- "second lookup ret=%d: %s => %"PRId64,
- ret,
- entry_loc.path,
- iatt.ia_ino);
-
- pump_update_resume_state (this, entry_loc.path);
-
- if (!IS_ENTRY_CWD(entry->d_name) &&
- !IS_ENTRY_PARENT (entry->d_name)) {
- pump_save_path (this, entry_loc.path);
- pump_save_file_stats (this, entry_loc.path);
}
-
- ret = pump_check_and_update_status (this);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Pump beginning to exit out");
+ loc_wipe (&entry_loc);
+ ret = afr_build_child_loc (this, &entry_loc, loc,
+ entry->d_name);
+ if (ret)
goto out;
- }
-
- gf_log (this->name, GF_LOG_TRACE,
- "type of file=%d, IFDIR=%d",
- iatt.ia_type, IA_IFDIR);
- if (IA_ISDIR (iatt.ia_type) && !IS_ENTRY_CWD(entry->d_name) &&
+ if (!IS_ENTRY_CWD (entry->d_name) &&
!IS_ENTRY_PARENT (entry->d_name)) {
- if (is_pump_traversal_allowed (this, entry_loc.path)) {
- gf_log (this->name, GF_LOG_TRACE,
- "entering dir=%s",
- entry->d_name);
- gf_pump_traverse_directory (&entry_loc);
- }
- }
- offset = entry->d_off;
- loc_wipe (&entry_loc);
+ is_directory_empty = _gf_false;
+ gf_log (this->name, GF_LOG_DEBUG,
+ "lookup %s => %"PRId64,
+ entry_loc.path,
+ iatt.ia_ino);
+
+ ret = syncop_lookup (this, &entry_loc, NULL,
+ &iatt, &xattr_rsp, &parent);
+
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "%s: lookup failed",
+ entry_loc.path);
+ continue;
+ }
+ pump_fill_loc_info (&entry_loc, &iatt,
+ &parent);
+
+ pump_update_resume_state (this, entry_loc.path);
+
+ pump_save_path (this, entry_loc.path);
+ pump_save_file_stats (this, entry_loc.path);
+
+ ret = pump_check_and_update_status (this);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "Pump beginning to exit out");
+ goto out;
+ }
+
+ if (IA_ISDIR (iatt.ia_type)) {
+ if (is_pump_traversal_allowed (this, entry_loc.path)) {
+ gf_log (this->name, GF_LOG_TRACE,
+ "entering dir=%s",
+ entry->d_name);
+ gf_pump_traverse_directory (&entry_loc);
+ }
+ }
+ }
}
gf_dirent_free (&entries);
+ free_entries = _gf_false;
gf_log (this->name, GF_LOG_TRACE,
"offset incremented to %d",
(int32_t ) offset);
}
-out:
- return 0;
-
-}
+ ret = syncop_close (fd);
+ if (ret < 0)
+ gf_log (this->name, GF_LOG_DEBUG, "closing the fd failed");
-void
-build_root_loc (inode_t *inode, loc_t *loc)
-{
- loc->path = "/";
- loc->name = "";
- loc->inode = inode;
- loc->ino = 1;
- loc->inode->ino = 1;
- memset (loc->inode->gfid, 0, 16);
- loc->inode->gfid[15] = 1;
+ if (is_directory_empty && IS_ROOT_PATH (loc->path)) {
+ pump_change_state (this, PUMP_STATE_RUNNING);
+ gf_log (this->name, GF_LOG_INFO, "Empty source brick. "
+ "Nothing to be done.");
+ }
+out:
+ if (entry_loc.path)
+ loc_wipe (&entry_loc);
+ if (free_entries)
+ gf_dirent_free (&entries);
+ return 0;
}
static int
pump_update_resume_path (xlator_t *this)
{
- afr_private_t *priv = NULL;
- pump_private_t *pump_priv = NULL;
-
const char *resume_path = NULL;
- priv = this->private;
- pump_priv = priv->pump_private;
-
resume_path = pump_get_resume_path (this);
if (resume_path) {
@@ -578,6 +480,39 @@ pump_update_resume_path (xlator_t *this)
return 0;
}
+static int32_t
+pump_xattr_cleaner (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ loc_t loc = {0};
+ int i = 0;
+ int ret = 0;
+ int source = 0;
+ int sink = 1;
+
+ priv = this->private;
+
+ afr_build_root_loc (this, &loc);
+
+ ret = syncop_removexattr (priv->children[source], &loc,
+ PUMP_PATH);
+
+ ret = syncop_removexattr (priv->children[sink], &loc,
+ PUMP_SINK_COMPLETE);
+
+ for (i = 0; i < priv->child_count; i++) {
+ ret = syncop_removexattr (priv->children[i], &loc,
+ PUMP_SOURCE_COMPLETE);
+ if (ret)
+ gf_log (this->name, GF_LOG_DEBUG, "removexattr "
+ "failed with %s", strerror (errno));
+ }
+
+ loc_wipe (&loc);
+ return pump_command_reply (frame, this);
+}
+
static int
pump_complete_migration (xlator_t *this)
{
@@ -585,16 +520,16 @@ pump_complete_migration (xlator_t *this)
pump_private_t *pump_priv = NULL;
dict_t *dict = NULL;
pump_state_t state;
- loc_t loc;
+ loc_t loc = {0};
int dict_ret = 0;
int ret = -1;
priv = this->private;
pump_priv = priv->pump_private;
- assert (priv->root_inode);
+ GF_ASSERT (priv->root_inode);
- build_root_loc (priv->root_inode, &loc);
+ afr_build_root_loc (this, &loc);
dict = dict_new ();
@@ -606,6 +541,10 @@ pump_complete_migration (xlator_t *this)
pump_priv->pump_finished = _gf_true;
dict_ret = dict_set_str (dict, PUMP_SOURCE_COMPLETE, "jargon");
+ if (dict_ret)
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: failed to set the key %s",
+ loc.path, PUMP_SOURCE_COMPLETE);
ret = syncop_setxattr (PUMP_SOURCE_CHILD (this), &loc, dict, 0);
if (ret < 0) {
@@ -613,44 +552,97 @@ pump_complete_migration (xlator_t *this)
"setxattr failed - while notifying source complete");
}
dict_ret = dict_set_str (dict, PUMP_SINK_COMPLETE, "jargon");
+ if (dict_ret)
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: failed to set the key %s",
+ loc.path, PUMP_SINK_COMPLETE);
ret = syncop_setxattr (PUMP_SINK_CHILD (this), &loc, dict, 0);
if (ret < 0) {
gf_log (this->name, GF_LOG_DEBUG,
"setxattr failed - while notifying sink complete");
}
+
+ pump_save_path (this, "/");
+
+ } else if (state == PUMP_STATE_ABORT) {
+ gf_log (this->name, GF_LOG_DEBUG, "Starting cleanup "
+ "of pump internal xattrs");
+ call_resume (pump_priv->cleaner);
}
+ loc_wipe (&loc);
return 0;
}
static int
+pump_lookup_sink (loc_t *loc)
+{
+ xlator_t *this = NULL;
+ struct iatt iatt, parent;
+ dict_t *xattr_rsp;
+ dict_t *xattr_req = NULL;
+ int ret = 0;
+
+ this = THIS;
+
+ xattr_req = dict_new ();
+
+ ret = afr_set_root_gfid (xattr_req);
+ if (ret)
+ goto out;
+
+ ret = syncop_lookup (PUMP_SINK_CHILD (this), loc,
+ xattr_req, &iatt, &xattr_rsp, &parent);
+
+ if (ret) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "Lookup on sink child failed");
+ goto out;
+ }
+
+out:
+ if (xattr_req)
+ dict_unref (xattr_req);
+
+ return ret;
+}
+
+static int
pump_task (void *data)
{
xlator_t *this = NULL;
afr_private_t *priv = NULL;
- loc_t loc;
+ loc_t loc = {0};
struct iatt iatt, parent;
- dict_t *xattr_rsp;
+ dict_t *xattr_rsp = NULL;
+ dict_t *xattr_req = NULL;
int ret = -1;
this = THIS;
priv = this->private;
- assert (priv->root_inode);
+ GF_ASSERT (priv->root_inode);
- build_root_loc (priv->root_inode, &loc);
+ afr_build_root_loc (this, &loc);
+ xattr_req = dict_new ();
+ if (!xattr_req) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "Out of memory");
+ ret = -1;
+ goto out;
+ }
- ret = syncop_lookup (this, &loc, NULL,
+ afr_set_root_gfid (xattr_req);
+ ret = syncop_lookup (this, &loc, xattr_req,
&iatt, &xattr_rsp, &parent);
gf_log (this->name, GF_LOG_TRACE,
- "lookup: ino=%"PRId64", path=%s",
- loc.ino,
- loc.path);
+ "lookup: path=%s gfid=%s",
+ loc.path, uuid_utoa (loc.inode->gfid));
ret = pump_check_and_update_status (this);
if (ret < 0) {
@@ -659,30 +651,37 @@ pump_task (void *data)
pump_update_resume_path (this);
+ afr_set_root_gfid (xattr_req);
+ ret = pump_lookup_sink (&loc);
+ if (ret) {
+ pump_update_resume_path (this);
+ goto out;
+ }
+
gf_pump_traverse_directory (&loc);
pump_complete_migration (this);
out:
+ if (xattr_req)
+ dict_unref (xattr_req);
+
+ loc_wipe (&loc);
return 0;
}
static int
-pump_task_completion (int ret, void *data)
+pump_task_completion (int ret, call_frame_t *sync_frame, void *data)
{
xlator_t *this = NULL;
- call_frame_t *frame = NULL;
afr_private_t *priv = NULL;
- pump_private_t *pump_priv = NULL;
this = THIS;
- frame = (call_frame_t *) data;
-
priv = this->private;
- pump_priv = priv->pump_private;
inode_unref (priv->root_inode);
+ STACK_DESTROY (sync_frame->root);
gf_log (this->name, GF_LOG_DEBUG,
"Pump xlator exiting");
@@ -700,20 +699,22 @@ pump_start (call_frame_t *pump_frame, xlator_t *this)
priv = this->private;
pump_priv = priv->pump_private;
- if (!pump_frame->root->lk_owner)
- pump_frame->root->lk_owner = PUMP_LK_OWNER;
+ afr_set_lk_owner (pump_frame, this, pump_frame->root);
+ pump_pid = (uint64_t) (unsigned long)pump_frame->root;
- ret = synctask_new (pump_priv->env, pump_task, pump_task_completion,
- pump_frame);
+ ret = synctask_new (pump_priv->env, pump_task,
+ pump_task_completion,
+ pump_frame, NULL);
if (ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
+ gf_log (this->name, GF_LOG_ERROR,
"starting pump failed");
pump_change_state (this, PUMP_STATE_ABORT);
goto out;
}
- gf_log (this->name, GF_LOG_TRACE,
- "setting pump as started");
+ gf_log (this->name, GF_LOG_DEBUG,
+ "setting pump as started lk_owner: %s %"PRIu64,
+ lkowner_utoa (&pump_frame->root->lk_owner), pump_pid);
priv->use_afr_in_pump = 1;
out:
@@ -747,9 +748,10 @@ pump_cmd_start_setxattr_cbk (call_frame_t *frame,
void *cookie,
xlator_t *this,
int32_t op_ret,
- int32_t op_errno)
+ int32_t op_errno, dict_t *xdata)
{
+ call_frame_t *prev = NULL;
afr_local_t *local = NULL;
int ret = 0;
@@ -769,6 +771,11 @@ pump_cmd_start_setxattr_cbk (call_frame_t *frame,
pump_mark_start_pending (this);
+ /* send the PARENT_UP as pump is ready now */
+ prev = cookie;
+ if (prev && prev->this)
+ prev->this->notify (prev->this, GF_EVENT_PARENT_UP, this);
+
out:
local->op_ret = ret;
pump_command_reply (frame, this);
@@ -782,8 +789,9 @@ pump_initiate_sink_connect (call_frame_t *frame, xlator_t *this)
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
dict_t *dict = NULL;
- char *dst_brick = NULL;
- loc_t loc;
+ data_t *data = NULL;
+ char *clnt_cmd = NULL;
+ loc_t loc = {0};
int ret = 0;
@@ -792,10 +800,11 @@ pump_initiate_sink_connect (call_frame_t *frame, xlator_t *this)
GF_ASSERT (priv->root_inode);
- build_root_loc (priv->root_inode, &loc);
+ afr_build_root_loc (this, &loc);
- ret = dict_get_str (local->dict, PUMP_CMD_START, &dst_brick);
- if (ret < 0) {
+ data = data_ref (dict_get (local->dict, RB_PUMP_CMD_START));
+ if (!data) {
+ ret = -1;
gf_log (this->name, GF_LOG_ERROR,
"Could not get destination brick value");
goto out;
@@ -803,17 +812,22 @@ pump_initiate_sink_connect (call_frame_t *frame, xlator_t *this)
dict = dict_new ();
if (!dict) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
ret = -1;
goto out;
}
- GF_ASSERT (dst_brick);
- gf_log (this->name, GF_LOG_DEBUG,
- "Got destination brick as %s", dst_brick);
+ clnt_cmd = GF_CALLOC (1, data->len+1, gf_common_mt_char);
+ if (!clnt_cmd) {
+ ret = -1;
+ goto out;
+ }
+
+ memcpy (clnt_cmd, data->data, data->len);
+ clnt_cmd[data->len] = '\0';
+ gf_log (this->name, GF_LOG_DEBUG, "Got destination brick %s\n",
+ clnt_cmd);
- ret = dict_set_str (dict, CLIENT_CMD_CONNECT, dst_brick);
+ ret = dict_set_dynstr (dict, CLIENT_CMD_CONNECT, clnt_cmd);
if (ret < 0) {
gf_log (this->name, GF_LOG_ERROR,
"Could not inititiate destination brick "
@@ -827,12 +841,21 @@ pump_initiate_sink_connect (call_frame_t *frame, xlator_t *this)
PUMP_SINK_CHILD(this)->fops->setxattr,
&loc,
dict,
- 0);
+ 0, NULL);
ret = 0;
- dict_unref (dict);
out:
+ if (dict)
+ dict_unref (dict);
+
+ if (data)
+ data_unref (data);
+
+ if (ret && clnt_cmd)
+ GF_FREE (clnt_cmd);
+
+ loc_wipe (&loc);
return ret;
}
@@ -852,7 +875,7 @@ pump_cmd_start_getxattr_cbk (call_frame_t *frame,
xlator_t *this,
int32_t op_ret,
int32_t op_errno,
- dict_t *dict)
+ dict_t *dict, dict_t *xdata)
{
afr_local_t *local = NULL;
char *path = NULL;
@@ -896,6 +919,8 @@ pump_cmd_start_getxattr_cbk (call_frame_t *frame,
else {
/* We're re-starting pump from a previous
pause */
+ gf_log (this->name, GF_LOG_DEBUG,
+ "about to start synctask");
ret = pump_start_synctask (this);
need_unwind = 1;
}
@@ -917,6 +942,7 @@ pump_execute_status (call_frame_t *frame, xlator_t *this)
uint64_t number_files = 0;
char filename[PATH_MAX];
+ char summary[PATH_MAX+256];
char *dict_str = NULL;
int32_t op_ret = 0;
@@ -945,28 +971,35 @@ pump_execute_status (call_frame_t *frame, xlator_t *this)
}
if (pump_priv->pump_finished) {
- snprintf (dict_str, PATH_MAX + 256, "Number of files migrated = %"PRIu64" Migration complete ",
- number_files);
+ snprintf (summary, PATH_MAX+256,
+ "no_of_files=%"PRIu64, number_files);
} else {
- snprintf (dict_str, PATH_MAX + 256, "Number of files migrated = %"PRIu64" Current file= %s ",
- number_files, filename);
+ snprintf (summary, PATH_MAX+256,
+ "no_of_files=%"PRIu64":current_file=%s",
+ number_files, filename);
}
+ snprintf (dict_str, PATH_MAX+256, "status=%d:%s",
+ (pump_priv->pump_finished)?1:0, summary);
dict = dict_new ();
- ret = dict_set_str (dict, PUMP_CMD_STATUS, dict_str);
+ ret = dict_set_dynstr (dict, RB_PUMP_CMD_STATUS, dict_str);
if (ret < 0) {
gf_log (this->name, GF_LOG_DEBUG,
- "dict_set_str returned negative value");
+ "dict_set_dynstr returned negative value");
+ } else {
+ dict_str = NULL;
}
op_ret = 0;
out:
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
+ AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, NULL);
+
+ if (dict)
+ dict_unref (dict);
- dict_unref (dict);
GF_FREE (dict_str);
return 0;
@@ -994,7 +1027,7 @@ pump_execute_start (call_frame_t *frame, xlator_t *this)
afr_local_t *local = NULL;
int ret = 0;
- loc_t loc;
+ loc_t loc = {0};
priv = this->private;
local = frame->local;
@@ -1009,14 +1042,14 @@ pump_execute_start (call_frame_t *frame, xlator_t *this)
GF_ASSERT (priv->root_inode);
- build_root_loc (priv->root_inode, &loc);
+ afr_build_root_loc (this, &loc);
STACK_WIND (frame,
pump_cmd_start_getxattr_cbk,
PUMP_SOURCE_CHILD(this),
PUMP_SOURCE_CHILD(this)->fops->getxattr,
&loc,
- PUMP_PATH);
+ PUMP_PATH, NULL);
ret = 0;
@@ -1026,22 +1059,98 @@ out:
pump_command_reply (frame, this);
}
+ loc_wipe (&loc);
return 0;
}
+static int
+pump_cleanup_helper (void *data) {
+ call_frame_t *frame = data;
+
+ pump_xattr_cleaner (frame, 0, frame->this, 0, 0, NULL);
+
+ return 0;
+}
+
+static int
+pump_cleanup_done (int ret, call_frame_t *sync_frame, void *data)
+{
+ STACK_DESTROY (sync_frame->root);
+
+ return 0;
+}
+
+int
+pump_execute_commit (call_frame_t *frame, xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ pump_private_t *pump_priv = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *sync_frame = NULL;
+ int ret = 0;
+
+ priv = this->private;
+ pump_priv = priv->pump_private;
+ local = frame->local;
+
+ local->op_ret = 0;
+ if (pump_priv->pump_finished) {
+ pump_change_state (this, PUMP_STATE_COMMIT);
+ sync_frame = create_frame (this, this->ctx->pool);
+ ret = synctask_new (pump_priv->env, pump_cleanup_helper,
+ pump_cleanup_done, sync_frame, frame);
+ if (ret) {
+ gf_log (this->name, GF_LOG_DEBUG, "Couldn't create "
+ "synctask for cleaning up xattrs.");
+ }
+
+ } else {
+ gf_log (this->name, GF_LOG_ERROR, "Commit can't proceed. "
+ "Migration in progress");
+ local->op_ret = -1;
+ local->op_errno = EINPROGRESS;
+ pump_command_reply (frame, this);
+ }
+
+ return 0;
+}
int
pump_execute_abort (call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ pump_private_t *pump_priv = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *sync_frame = NULL;
+ int ret = 0;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ pump_priv = priv->pump_private;
+ local = frame->local;
pump_change_state (this, PUMP_STATE_ABORT);
+ LOCK (&pump_priv->resume_path_lock);
+ {
+ pump_priv->number_files_pumped = 0;
+ pump_priv->current_file[0] = '\0';
+ }
+ UNLOCK (&pump_priv->resume_path_lock);
+
local->op_ret = 0;
- pump_command_reply (frame, this);
+ if (pump_priv->pump_finished) {
+ sync_frame = create_frame (this, this->ctx->pool);
+ ret = synctask_new (pump_priv->env, pump_cleanup_helper,
+ pump_cleanup_done, sync_frame, frame);
+ if (ret) {
+ gf_log (this->name, GF_LOG_DEBUG, "Couldn't create "
+ "synctask for cleaning up xattrs.");
+ }
+
+ } else {
+ pump_priv->cleaner = fop_setxattr_cbk_stub (frame,
+ pump_xattr_cleaner,
+ 0, 0, NULL);
+ }
return 0;
}
@@ -1053,7 +1162,7 @@ pump_command_status (xlator_t *this, dict_t *dict)
int dict_ret = -1;
int ret = _gf_true;
- dict_ret = dict_get_str (dict, PUMP_CMD_STATUS, &cmd);
+ dict_ret = dict_get_str (dict, RB_PUMP_CMD_STATUS, &cmd);
if (dict_ret < 0) {
gf_log (this->name, GF_LOG_DEBUG,
"Not a pump status command");
@@ -1077,7 +1186,7 @@ pump_command_pause (xlator_t *this, dict_t *dict)
int dict_ret = -1;
int ret = _gf_true;
- dict_ret = dict_get_str (dict, PUMP_CMD_PAUSE, &cmd);
+ dict_ret = dict_get_str (dict, RB_PUMP_CMD_PAUSE, &cmd);
if (dict_ret < 0) {
gf_log (this->name, GF_LOG_DEBUG,
"Not a pump pause command");
@@ -1095,13 +1204,37 @@ out:
}
gf_boolean_t
+pump_command_commit (xlator_t *this, dict_t *dict)
+{
+ char *cmd = NULL;
+ int dict_ret = -1;
+ int ret = _gf_true;
+
+ dict_ret = dict_get_str (dict, RB_PUMP_CMD_COMMIT, &cmd);
+ if (dict_ret < 0) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "Not a pump commit command");
+ ret = _gf_false;
+ goto out;
+ }
+
+ gf_log (this->name, GF_LOG_DEBUG,
+ "Hit a pump command - commit");
+ ret = _gf_true;
+
+out:
+ return ret;
+
+}
+
+gf_boolean_t
pump_command_abort (xlator_t *this, dict_t *dict)
{
char *cmd = NULL;
int dict_ret = -1;
int ret = _gf_true;
- dict_ret = dict_get_str (dict, PUMP_CMD_ABORT, &cmd);
+ dict_ret = dict_get_str (dict, RB_PUMP_CMD_ABORT, &cmd);
if (dict_ret < 0) {
gf_log (this->name, GF_LOG_DEBUG,
"Not a pump abort command");
@@ -1125,7 +1258,7 @@ pump_command_start (xlator_t *this, dict_t *dict)
int dict_ret = -1;
int ret = _gf_true;
- dict_ret = dict_get_str (dict, PUMP_CMD_START, &cmd);
+ dict_ret = dict_get_str (dict, RB_PUMP_CMD_START, &cmd);
if (dict_ret < 0) {
gf_log (this->name, GF_LOG_DEBUG,
"Not a pump start command");
@@ -1147,7 +1280,7 @@ struct _xattr_key {
struct list_head list;
};
-static void
+static int
__gather_xattr_keys (dict_t *dict, char *key, data_t *value,
void *data)
{
@@ -1159,13 +1292,14 @@ __gather_xattr_keys (dict_t *dict, char *key, data_t *value,
xkey = GF_CALLOC (1, sizeof (*xkey), gf_afr_mt_xattr_key);
if (!xkey)
- return;
+ return -1;
xkey->key = key;
INIT_LIST_HEAD (&xkey->list);
list_add_tail (&xkey->list, list);
}
+ return 0;
}
static void
@@ -1193,16 +1327,17 @@ __filter_xattrs (dict_t *dict)
int32_t
pump_getxattr_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict)
+ dict_t *dict, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- xlator_t ** children = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ xlator_t **children = NULL;
+ int unwind = 1;
+ int32_t *last_index = NULL;
+ int32_t next_call_child = -1;
+ int32_t read_child = -1;
+ int32_t *fresh_children = NULL;
- int unwind = 1;
- int last_tried = -1;
- int this_try = -1;
- int read_child = -1;
priv = this->private;
children = priv->children;
@@ -1212,25 +1347,22 @@ pump_getxattr_cbk (call_frame_t *frame, void *cookie,
read_child = (long) cookie;
if (op_ret == -1) {
- retry:
- last_tried = local->cont.getxattr.last_tried;
-
- if (all_tried (last_tried, priv->child_count)) {
- goto out;
- }
- this_try = ++local->cont.getxattr.last_tried;
-
- if (this_try == read_child) {
- goto retry;
- }
+ last_index = &local->cont.getxattr.last_index;
+ fresh_children = local->fresh_children;
+ next_call_child = afr_next_call_child (fresh_children,
+ local->child_up,
+ priv->child_count,
+ last_index, read_child);
+ if (next_call_child < 0)
+ goto out;
unwind = 0;
STACK_WIND_COOKIE (frame, pump_getxattr_cbk,
(void *) (long) read_child,
- children[this_try],
- children[this_try]->fops->getxattr,
+ children[next_call_child],
+ children[next_call_child]->fops->getxattr,
&local->loc,
- local->cont.getxattr.name);
+ local->cont.getxattr.name, NULL);
}
out:
@@ -1238,7 +1370,7 @@ out:
if (op_ret >= 0 && dict)
__filter_xattrs (dict);
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
+ AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, NULL);
}
return 0;
@@ -1246,17 +1378,15 @@ out:
int32_t
pump_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name)
+ loc_t *loc, const char *name, dict_t *xdata)
{
afr_private_t * priv = NULL;
xlator_t ** children = NULL;
int call_child = 0;
- afr_local_t * local = NULL;
-
- int read_child = -1;
-
- int32_t op_ret = -1;
- int32_t op_errno = 0;
+ afr_local_t *local = NULL;
+ int32_t ret = -1;
+ int32_t op_errno = 0;
+ uint64_t read_child = 0;
VALIDATE_OR_GOTO (frame, out);
@@ -1267,9 +1397,21 @@ pump_getxattr (call_frame_t *frame, xlator_t *this,
VALIDATE_OR_GOTO (priv->children, out);
children = priv->children;
+ if (!priv->use_afr_in_pump) {
+ STACK_WIND (frame, default_getxattr_cbk,
+ FIRST_CHILD (this),
+ (FIRST_CHILD (this))->fops->getxattr,
+ loc, name, xdata);
+ return 0;
+ }
+
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- ALLOC_OR_GOTO (local, afr_local_t, out);
- frame->local = local;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
if (name) {
if (!strncmp (name, AFR_XATTR_PREFIX,
@@ -1279,42 +1421,33 @@ pump_getxattr (call_frame_t *frame, xlator_t *this,
goto out;
}
- if (!strcmp (name, PUMP_CMD_STATUS)) {
+ if (!strcmp (name, RB_PUMP_CMD_STATUS)) {
gf_log (this->name, GF_LOG_DEBUG,
"Hit pump command - status");
pump_execute_status (frame, this);
- op_ret = 0;
+ ret = 0;
goto out;
}
}
- if (!priv->use_afr_in_pump) {
- STACK_WIND (frame, default_getxattr_cbk,
- FIRST_CHILD (this),
- (FIRST_CHILD (this))->fops->getxattr,
- loc, name);
- return 0;
+ local->fresh_children = GF_CALLOC (priv->child_count,
+ sizeof (*local->fresh_children),
+ gf_afr_mt_int32_t);
+ if (!local->fresh_children) {
+ ret = -1;
+ op_errno = ENOMEM;
+ goto out;
}
- read_child = afr_read_child (this, loc->inode);
-
- if (read_child >= 0) {
- call_child = read_child;
-
- local->cont.getxattr.last_tried = -1;
- } else {
- call_child = afr_first_up_child (priv);
-
- if (call_child == -1) {
- op_errno = ENOTCONN;
- gf_log (this->name, GF_LOG_DEBUG,
- "no child is up");
- goto out;
- }
-
- local->cont.getxattr.last_tried = call_child;
+ read_child = afr_inode_get_read_ctx (this, loc->inode, local->fresh_children);
+ ret = afr_get_call_child (this, local->child_up, read_child,
+ local->fresh_children,
+ &call_child,
+ &local->cont.getxattr.last_index);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
}
-
loc_copy (&local->loc, loc);
if (name)
local->cont.getxattr.name = gf_strdup (name);
@@ -1322,13 +1455,12 @@ pump_getxattr (call_frame_t *frame, xlator_t *this,
STACK_WIND_COOKIE (frame, pump_getxattr_cbk,
(void *) (long) call_child,
children[call_child], children[call_child]->fops->getxattr,
- loc, name);
+ loc, name, xdata);
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, NULL);
- }
+ if (ret < 0)
+ AFR_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
return 0;
}
@@ -1336,11 +1468,9 @@ static int
afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
{
afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
call_frame_t *main_frame = NULL;
local = frame->local;
- priv = this->private;
LOCK (&frame->lock);
{
@@ -1352,14 +1482,14 @@ afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
if (main_frame) {
AFR_STACK_UNWIND (setxattr, main_frame,
- local->op_ret, local->op_errno)
+ local->op_ret, local->op_errno, NULL);
}
return 0;
}
static int
afr_setxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_local_t * local = NULL;
afr_private_t * priv = NULL;
@@ -1411,7 +1541,7 @@ afr_setxattr_wind (call_frame_t *frame, xlator_t *this)
local = frame->local;
priv = this->private;
- call_count = afr_up_children_count (priv->child_count, local->child_up);
+ call_count = afr_up_children_count (local->child_up, priv->child_count);
if (call_count == 0) {
local->transaction.resume (frame, this);
@@ -1428,7 +1558,7 @@ afr_setxattr_wind (call_frame_t *frame, xlator_t *this)
priv->children[i]->fops->setxattr,
&local->loc,
local->cont.setxattr.dict,
- local->cont.setxattr.flags);
+ local->cont.setxattr.flags, NULL);
if (!--call_count)
break;
@@ -1456,11 +1586,9 @@ pump_setxattr_cbk (call_frame_t *frame,
void *cookie,
xlator_t *this,
int32_t op_ret,
- int32_t op_errno)
+ int32_t op_errno, dict_t *xdata)
{
- STACK_UNWIND (frame,
- op_ret,
- op_errno);
+ AFR_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
return 0;
}
@@ -1472,18 +1600,16 @@ pump_command_reply (call_frame_t *frame, xlator_t *this)
local = frame->local;
if (local->op_ret < 0)
- gf_log (this->name, GF_LOG_NORMAL,
+ gf_log (this->name, GF_LOG_INFO,
"Command failed");
else
- gf_log (this->name, GF_LOG_NORMAL,
+ gf_log (this->name, GF_LOG_INFO,
"Command succeeded");
- dict_unref (local->dict);
-
AFR_STACK_UNWIND (setxattr,
frame,
local->op_ret,
- local->op_errno);
+ local->op_errno, NULL);
return 0;
}
@@ -1509,56 +1635,64 @@ pump_parse_command (call_frame_t *frame, xlator_t *this,
frame->local = local;
local->dict = dict_ref (dict);
ret = pump_execute_abort (frame, this);
+
+ } else if (pump_command_commit (this, dict)) {
+ frame->local = local;
+ local->dict = dict_ref (dict);
+ ret = pump_execute_commit (frame, this);
}
return ret;
}
int
pump_setxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *dict, int32_t flags)
+ loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
call_frame_t *transaction_frame = NULL;
-
int ret = -1;
-
- int op_ret = -1;
int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
VALIDATE_OR_GOTO (this->private, out);
+ GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.pump*", dict,
+ op_errno, out);
+
priv = this->private;
+ if (!priv->use_afr_in_pump) {
+ STACK_WIND (frame, default_setxattr_cbk,
+ FIRST_CHILD (this),
+ (FIRST_CHILD (this))->fops->setxattr,
+ loc, dict, flags, xdata);
+ return 0;
+ }
+
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (local, out);
- ret = AFR_LOCAL_INIT (local, priv);
+ ret = afr_local_init (local, priv, &op_errno);
if (ret < 0) {
- op_errno = -ret;
+ afr_local_cleanup (local, this);
goto out;
- }
+ }
ret = pump_parse_command (frame, this,
local, dict);
if (ret >= 0) {
- op_ret = 0;
+ ret = 0;
goto out;
}
- if (!priv->use_afr_in_pump) {
- STACK_WIND (frame, default_setxattr_cbk,
- FIRST_CHILD (this),
- (FIRST_CHILD (this))->fops->setxattr,
- loc, dict, flags);
- return 0;
- }
-
transaction_frame = copy_frame (frame);
if (!transaction_frame) {
gf_log (this->name, GF_LOG_ERROR,
"Out of memory.");
+ op_errno = ENOMEM;
+ ret = -1;
+ afr_local_cleanup (local, this);
goto out;
}
@@ -1581,12 +1715,12 @@ pump_setxattr (call_frame_t *frame, xlator_t *this,
afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
- op_ret = 0;
+ ret = 0;
out:
- if (op_ret == -1) {
+ if (ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (setxattr, frame, op_ret, op_errno);
+ AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
}
return 0;
@@ -1620,7 +1754,7 @@ static int32_t
pump_truncate (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
- off_t offset)
+ off_t offset, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1630,11 +1764,11 @@ pump_truncate (call_frame_t *frame,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->truncate,
loc,
- offset);
+ offset, xdata);
return 0;
}
- afr_truncate (frame, this, loc, offset);
+ afr_truncate (frame, this, loc, offset, xdata);
return 0;
}
@@ -1643,7 +1777,7 @@ static int32_t
pump_ftruncate (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
- off_t offset)
+ off_t offset, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1653,11 +1787,11 @@ pump_ftruncate (call_frame_t *frame,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->ftruncate,
fd,
- offset);
+ offset, xdata);
return 0;
}
- afr_ftruncate (frame, this, fd, offset);
+ afr_ftruncate (frame, this, fd, offset, xdata);
return 0;
}
@@ -1666,7 +1800,7 @@ pump_ftruncate (call_frame_t *frame,
int
pump_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t rdev, dict_t *parms)
+ loc_t *loc, mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1674,10 +1808,10 @@ pump_mknod (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, default_mknod_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->mknod,
- loc, mode, rdev, parms);
+ loc, mode, rdev, umask, xdata);
return 0;
}
- afr_mknod (frame, this, loc, mode, rdev, parms);
+ afr_mknod (frame, this, loc, mode, rdev, umask, xdata);
return 0;
}
@@ -1686,7 +1820,7 @@ pump_mknod (call_frame_t *frame, xlator_t *this,
int
pump_mkdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dict_t *params)
+ loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1694,10 +1828,10 @@ pump_mkdir (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, default_mkdir_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->mkdir,
- loc, mode, params);
+ loc, mode, umask, xdata);
return 0;
}
- afr_mkdir (frame, this, loc, mode, params);
+ afr_mkdir (frame, this, loc, mode, umask, xdata);
return 0;
}
@@ -1706,7 +1840,7 @@ pump_mkdir (call_frame_t *frame, xlator_t *this,
static int32_t
pump_unlink (call_frame_t *frame,
xlator_t *this,
- loc_t *loc)
+ loc_t *loc, int xflag, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1715,31 +1849,32 @@ pump_unlink (call_frame_t *frame,
default_unlink_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->unlink,
- loc);
+ loc, xflag, xdata);
return 0;
}
- afr_unlink (frame, this, loc);
+ afr_unlink (frame, this, loc, xflag, xdata);
return 0;
}
-static int32_t
-pump_rmdir (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc)
+static int
+pump_rmdir (call_frame_t *frame, xlator_t *this,
+ loc_t *loc, int flags, dict_t *xdata)
{
afr_private_t *priv = NULL;
+
priv = this->private;
+
if (!priv->use_afr_in_pump) {
- STACK_WIND (frame,
- default_rmdir_cbk,
+ STACK_WIND (frame, default_rmdir_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->rmdir,
- loc);
+ loc, flags, xdata);
return 0;
}
- afr_rmdir (frame, this, loc);
+
+ afr_rmdir (frame, this, loc, flags, xdata);
return 0;
}
@@ -1748,7 +1883,7 @@ pump_rmdir (call_frame_t *frame,
int
pump_symlink (call_frame_t *frame, xlator_t *this,
- const char *linkpath, loc_t *loc, dict_t *params)
+ const char *linkpath, loc_t *loc, mode_t umask, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1756,10 +1891,10 @@ pump_symlink (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, default_symlink_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->symlink,
- linkpath, loc, params);
+ linkpath, loc, umask, xdata);
return 0;
}
- afr_symlink (frame, this, linkpath, loc, params);
+ afr_symlink (frame, this, linkpath, loc, umask, xdata);
return 0;
}
@@ -1769,7 +1904,7 @@ static int32_t
pump_rename (call_frame_t *frame,
xlator_t *this,
loc_t *oldloc,
- loc_t *newloc)
+ loc_t *newloc, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1778,10 +1913,10 @@ pump_rename (call_frame_t *frame,
default_rename_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->rename,
- oldloc, newloc);
+ oldloc, newloc, xdata);
return 0;
}
- afr_rename (frame, this, oldloc, newloc);
+ afr_rename (frame, this, oldloc, newloc, xdata);
return 0;
}
@@ -1791,7 +1926,7 @@ static int32_t
pump_link (call_frame_t *frame,
xlator_t *this,
loc_t *oldloc,
- loc_t *newloc)
+ loc_t *newloc, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1800,10 +1935,10 @@ pump_link (call_frame_t *frame,
default_link_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->link,
- oldloc, newloc);
+ oldloc, newloc, xdata);
return 0;
}
- afr_link (frame, this, oldloc, newloc);
+ afr_link (frame, this, oldloc, newloc, xdata);
return 0;
}
@@ -1812,7 +1947,7 @@ pump_link (call_frame_t *frame,
static int32_t
pump_create (call_frame_t *frame, xlator_t *this,
loc_t *loc, int32_t flags, mode_t mode,
- fd_t *fd, dict_t *params)
+ mode_t umask, fd_t *fd, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1820,10 +1955,10 @@ pump_create (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, default_create_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->create,
- loc, flags, mode, fd, params);
+ loc, flags, mode, umask, fd, xdata);
return 0;
}
- afr_create (frame, this, loc, flags, mode, fd, params);
+ afr_create (frame, this, loc, flags, mode, umask, fd, xdata);
return 0;
}
@@ -1833,8 +1968,7 @@ static int32_t
pump_open (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
- int32_t flags, fd_t *fd,
- int32_t wbflags)
+ int32_t flags, fd_t *fd, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1843,10 +1977,10 @@ pump_open (call_frame_t *frame,
default_open_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->open,
- loc, flags, fd, wbflags);
+ loc, flags, fd, xdata);
return 0;
}
- afr_open (frame, this, loc, flags, fd, wbflags);
+ afr_open (frame, this, loc, flags, fd, xdata);
return 0;
}
@@ -1858,8 +1992,8 @@ pump_writev (call_frame_t *frame,
fd_t *fd,
struct iovec *vector,
int32_t count,
- off_t off,
- struct iobref *iobref)
+ off_t off, uint32_t flags,
+ struct iobref *iobref, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1871,20 +2005,20 @@ pump_writev (call_frame_t *frame,
fd,
vector,
count,
- off,
- iobref);
+ off, flags,
+ iobref, xdata);
return 0;
}
- afr_writev (frame, this, fd, vector, count, off, iobref);
- return 0;
+ afr_writev (frame, this, fd, vector, count, off, flags, iobref, xdata);
+ return 0;
}
static int32_t
pump_flush (call_frame_t *frame,
xlator_t *this,
- fd_t *fd)
+ fd_t *fd, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1893,10 +2027,10 @@ pump_flush (call_frame_t *frame,
default_flush_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->flush,
- fd);
+ fd, xdata);
return 0;
}
- afr_flush (frame, this, fd);
+ afr_flush (frame, this, fd, xdata);
return 0;
}
@@ -1906,7 +2040,7 @@ static int32_t
pump_fsync (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
- int32_t flags)
+ int32_t flags, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1916,10 +2050,10 @@ pump_fsync (call_frame_t *frame,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->fsync,
fd,
- flags);
+ flags, xdata);
return 0;
}
- afr_fsync (frame, this, fd, flags);
+ afr_fsync (frame, this, fd, flags, xdata);
return 0;
}
@@ -1928,7 +2062,7 @@ pump_fsync (call_frame_t *frame,
static int32_t
pump_opendir (call_frame_t *frame,
xlator_t *this,
- loc_t *loc, fd_t *fd)
+ loc_t *loc, fd_t *fd, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1937,10 +2071,10 @@ pump_opendir (call_frame_t *frame,
default_opendir_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->opendir,
- loc, fd);
+ loc, fd, xdata);
return 0;
}
- afr_opendir (frame, this, loc, fd);
+ afr_opendir (frame, this, loc, fd, xdata);
return 0;
}
@@ -1950,7 +2084,7 @@ static int32_t
pump_fsyncdir (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
- int32_t flags)
+ int32_t flags, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1960,10 +2094,10 @@ pump_fsyncdir (call_frame_t *frame,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->fsyncdir,
fd,
- flags);
+ flags, xdata);
return 0;
}
- afr_fsyncdir (frame, this, fd, flags);
+ afr_fsyncdir (frame, this, fd, flags, xdata);
return 0;
}
@@ -1974,7 +2108,7 @@ pump_xattrop (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
gf_xattrop_flags_t flags,
- dict_t *dict)
+ dict_t *dict, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1985,10 +2119,10 @@ pump_xattrop (call_frame_t *frame,
FIRST_CHILD(this)->fops->xattrop,
loc,
flags,
- dict);
+ dict, xdata);
return 0;
}
- afr_xattrop (frame, this, loc, flags, dict);
+ afr_xattrop (frame, this, loc, flags, dict, xdata);
return 0;
}
@@ -1998,7 +2132,7 @@ pump_fxattrop (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
gf_xattrop_flags_t flags,
- dict_t *dict)
+ dict_t *dict, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -2009,10 +2143,10 @@ pump_fxattrop (call_frame_t *frame,
FIRST_CHILD(this)->fops->fxattrop,
fd,
flags,
- dict);
+ dict, xdata);
return 0;
}
- afr_fxattrop (frame, this, fd, flags, dict);
+ afr_fxattrop (frame, this, fd, flags, dict, xdata);
return 0;
}
@@ -2022,9 +2156,17 @@ static int32_t
pump_removexattr (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
- const char *name)
+ const char *name, dict_t *xdata)
{
- afr_private_t *priv = NULL;
+ afr_private_t *priv = NULL;
+ int op_errno = -1;
+
+ VALIDATE_OR_GOTO (this, out);
+
+ GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.pump*",
+ name, op_errno, out);
+
+ op_errno = 0;
priv = this->private;
if (!priv->use_afr_in_pump) {
STACK_WIND (frame,
@@ -2032,10 +2174,14 @@ pump_removexattr (call_frame_t *frame,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->removexattr,
loc,
- name);
+ name, xdata);
return 0;
}
- afr_removexattr (frame, this, loc, name);
+ afr_removexattr (frame, this, loc, name, xdata);
+
+ out:
+ if (op_errno)
+ AFR_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
return 0;
}
@@ -2047,7 +2193,7 @@ pump_readdir (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
size_t size,
- off_t off)
+ off_t off, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -2056,21 +2202,18 @@ pump_readdir (call_frame_t *frame,
default_readdir_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->readdir,
- fd, size, off);
+ fd, size, off, xdata);
return 0;
}
- afr_readdir (frame, this, fd, size, off);
+ afr_readdir (frame, this, fd, size, off, xdata);
return 0;
}
static int32_t
-pump_readdirp (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- size_t size,
- off_t off)
+pump_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd,
+ size_t size, off_t off, dict_t *dict)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -2079,10 +2222,10 @@ pump_readdirp (call_frame_t *frame,
default_readdirp_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->readdirp,
- fd, size, off);
+ fd, size, off, dict);
return 0;
}
- afr_readdirp (frame, this, fd, size, off);
+ afr_readdirp (frame, this, fd, size, off, dict);
return 0;
}
@@ -2113,13 +2256,24 @@ pump_release (xlator_t *this,
}
+static int32_t
+pump_forget (xlator_t *this, inode_t *inode)
+{
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ if (priv->use_afr_in_pump)
+ afr_forget (this, inode);
+
+ return 0;
+}
static int32_t
pump_setattr (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
struct iatt *stbuf,
- int32_t valid)
+ int32_t valid, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -2128,10 +2282,10 @@ pump_setattr (call_frame_t *frame,
default_setattr_cbk,
FIRST_CHILD (this),
FIRST_CHILD (this)->fops->setattr,
- loc, stbuf, valid);
+ loc, stbuf, valid, xdata);
return 0;
}
- afr_setattr (frame, this, loc, stbuf, valid);
+ afr_setattr (frame, this, loc, stbuf, valid, xdata);
return 0;
}
@@ -2142,7 +2296,7 @@ pump_fsetattr (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
struct iatt *stbuf,
- int32_t valid)
+ int32_t valid, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -2151,10 +2305,10 @@ pump_fsetattr (call_frame_t *frame,
default_fsetattr_cbk,
FIRST_CHILD (this),
FIRST_CHILD (this)->fops->fsetattr,
- fd, stbuf, valid);
+ fd, stbuf, valid, xdata);
return 0;
}
- afr_fsetattr (frame, this, fd, stbuf, valid);
+ afr_fsetattr (frame, this, fd, stbuf, valid, xdata);
return 0;
}
@@ -2203,7 +2357,7 @@ notify (xlator_t *this, int32_t event,
child_xl = (xlator_t *) data;
- ret = afr_notify (this, event, data);
+ ret = afr_notify (this, event, data, NULL);
switch (event) {
case GF_EVENT_CHILD_DOWN:
@@ -2214,6 +2368,8 @@ notify (xlator_t *this, int32_t event,
case GF_EVENT_CHILD_UP:
if (is_xlator_pump_sink (child_xl))
if (is_pump_start_pending (this)) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "about to start synctask");
ret = pump_start_synctask (this);
if (ret < 0)
gf_log (this->name, GF_LOG_DEBUG,
@@ -2236,7 +2392,7 @@ init (xlator_t *this)
xlator_list_t * trav = NULL;
int i = 0;
int ret = -1;
- int op_errno = 0;
+ GF_UNUSED int op_errno = 0;
int source_child = 0;
@@ -2252,26 +2408,42 @@ init (xlator_t *this)
"Volume is dangling.");
}
- ALLOC_OR_GOTO (this->private, afr_private_t, out);
+ this->private = GF_CALLOC (1, sizeof (afr_private_t),
+ gf_afr_mt_afr_private_t);
+ if (!this->private)
+ goto out;
priv = this->private;
+ LOCK_INIT (&priv->lock);
+ LOCK_INIT (&priv->read_child_lock);
+ //lock recovery is not done in afr
+ pthread_mutex_init (&priv->mutex, NULL);
+ INIT_LIST_HEAD (&priv->saved_fds);
+
+ child_count = xlator_subvolume_count (this);
+ if (child_count != 2) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "There should be exactly 2 children - one source "
+ "and one sink");
+ return -1;
+ }
+ priv->child_count = child_count;
priv->read_child = source_child;
priv->favorite_child = source_child;
priv->background_self_heal_count = 0;
- priv->data_self_heal = 1;
+ priv->data_self_heal = "on";
priv->metadata_self_heal = 1;
priv->entry_self_heal = 1;
- priv->data_self_heal_algorithm = "";
-
priv->data_self_heal_window_size = 16;
priv->data_change_log = 1;
priv->metadata_change_log = 1;
priv->entry_change_log = 1;
priv->use_afr_in_pump = 1;
+ priv->sh_readdir_size = 65536;
/* Locking options */
@@ -2280,31 +2452,9 @@ init (xlator_t *this)
and the sink.
*/
- priv->data_lock_server_count = 2;
- priv->metadata_lock_server_count = 2;
- priv->entry_lock_server_count = 2;
-
priv->strict_readdir = _gf_false;
- trav = this->children;
- while (trav) {
- child_count++;
- trav = trav->next;
- }
-
priv->wait_count = 1;
-
- if (child_count != 2) {
- gf_log (this->name, GF_LOG_ERROR,
- "There should be exactly 2 children - one source "
- "and one sink");
- return -1;
- }
- priv->child_count = child_count;
-
- LOCK_INIT (&priv->lock);
- LOCK_INIT (&priv->read_child_lock);
-
priv->child_up = GF_CALLOC (sizeof (unsigned char), child_count,
gf_afr_mt_char);
if (!priv->child_up) {
@@ -2338,8 +2488,8 @@ init (xlator_t *this)
while (i < child_count) {
priv->children[i] = trav->xlator;
- ret = asprintf (&priv->pending_key[i], "%s.%s", AFR_XATTR_PREFIX,
- trav->xlator->name);
+ ret = gf_asprintf (&priv->pending_key[i], "%s.%s", AFR_XATTR_PREFIX,
+ trav->xlator->name);
if (-1 == ret) {
gf_log (this->name, GF_LOG_ERROR,
"asprintf failed to set pending key");
@@ -2351,9 +2501,22 @@ init (xlator_t *this)
i++;
}
+ ret = gf_asprintf (&priv->sh_domain, "%s-self-heal", this->name);
+ if (-1 == ret) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
priv->first_lookup = 1;
priv->root_inode = NULL;
+ priv->last_event = GF_CALLOC (child_count, sizeof (*priv->last_event),
+ gf_afr_mt_int32_t);
+ if (!priv->last_event) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
pump_priv = GF_CALLOC (1, sizeof (*pump_priv),
gf_afr_mt_pump_priv);
if (!pump_priv) {
@@ -2375,7 +2538,7 @@ init (xlator_t *this)
goto out;
}
- pump_priv->env = syncenv_new (0);
+ pump_priv->env = this->ctx->env;
if (!pump_priv->env) {
gf_log (this->name, GF_LOG_ERROR,
"Could not create new sync-environment");
@@ -2383,6 +2546,15 @@ init (xlator_t *this)
goto out;
}
+ /* keep more local here as we may need them for self-heal etc */
+ this->local_pool = mem_pool_new (afr_local_t, 128);
+ if (!this->local_pool) {
+ ret = -1;
+ gf_log (this->name, GF_LOG_ERROR,
+ "failed to create local_t's memory pool");
+ goto out;
+ }
+
priv->pump_private = pump_priv;
pump_change_state (this, PUMP_STATE_ABORT);
@@ -2395,6 +2567,25 @@ out:
int
fini (xlator_t *this)
{
+ afr_private_t * priv = NULL;
+ pump_private_t *pump_priv = NULL;
+
+ priv = this->private;
+ this->private = NULL;
+ if (!priv)
+ goto out;
+
+ pump_priv = priv->pump_private;
+ if (!pump_priv)
+ goto afr_priv;
+
+ GF_FREE (pump_priv->resume_path);
+ LOCK_DESTROY (&pump_priv->resume_path_lock);
+ LOCK_DESTROY (&pump_priv->pump_state_lock);
+ GF_FREE (pump_priv);
+afr_priv:
+ afr_priv_destroy (priv);
+out:
return 0;
}
@@ -2442,6 +2633,7 @@ struct xlator_dumpops dumpops = {
struct xlator_cbks cbks = {
.release = pump_release,
.releasedir = pump_releasedir,
+ .forget = pump_forget,
};
struct volume_options options[] = {
diff --git a/xlators/cluster/afr/src/pump.h b/xlators/cluster/afr/src/pump.h
index 65cfcb55b..bc4c31a78 100644
--- a/xlators/cluster/afr/src/pump.h
+++ b/xlators/cluster/afr/src/pump.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __PUMP_H__
@@ -26,18 +17,6 @@
#define CLIENT_CMD_CONNECT "trusted.glusterfs.client-connect"
#define CLIENT_CMD_DISCONNECT "trusted.glusterfs.client-disconnect"
-#define PUMP_PID 696969
-#define PUMP_LK_OWNER 696969
-
-#define IS_ROOT_PATH(path) (!strcmp (path, "/"))
-#define IS_ENTRY_CWD(entry) (!strcmp (entry, "."))
-#define IS_ENTRY_PARENT(entry) (!strcmp (entry, ".."))
-
-#define PUMP_CMD_START "trusted.glusterfs.pump.start"
-#define PUMP_CMD_ABORT "trusted.glusterfs.pump.abort"
-#define PUMP_CMD_PAUSE "trusted.glusterfs.pump.pause"
-#define PUMP_CMD_STATUS "trusted.glusterfs.pump.status"
-
#define PUMP_SOURCE_COMPLETE "trusted.glusterfs.pump-source-complete"
#define PUMP_SINK_COMPLETE "trusted.glusterfs.pump-sink-complete"
@@ -51,11 +30,12 @@ typedef enum {
PUMP_STATE_RESUME, /* Pump is resuming from a previous pause */
PUMP_STATE_PAUSE, /* Pump is paused */
PUMP_STATE_ABORT, /* Pump is aborted */
+ PUMP_STATE_COMMIT, /* Pump is commited */
} pump_state_t;
typedef struct _pump_private {
struct syncenv *env; /* The env pointer to the pump synctask */
- const char *resume_path; /* path to resume from the last pause */
+ char *resume_path; /* path to resume from the last pause */
gf_lock_t resume_path_lock; /* Synchronize resume_path changes */
gf_lock_t pump_state_lock; /* Synchronize pump_state changes */
pump_state_t pump_state; /* State of pump */
@@ -64,6 +44,7 @@ typedef struct _pump_private {
gf_boolean_t pump_finished; /* Boolean to indicate pump termination */
char pump_start_pending; /* Boolean to mark start pending until
CHILD_UP */
+ call_stub_t *cleaner;
} pump_private_t;
void
diff --git a/xlators/cluster/dht/src/Makefile.am b/xlators/cluster/dht/src/Makefile.am
index 4b69aa071..174bea841 100644
--- a/xlators/cluster/dht/src/Makefile.am
+++ b/xlators/cluster/dht/src/Makefile.am
@@ -2,33 +2,37 @@
xlator_LTLIBRARIES = dht.la nufa.la switch.la
xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/cluster
+dht_common_source = dht-layout.c dht-helper.c dht-linkfile.c dht-rebalance.c \
+ dht-selfheal.c dht-rename.c dht-hashfn.c dht-diskusage.c \
+ dht-common.c dht-inode-write.c dht-inode-read.c dht-shared.c \
+ $(top_builddir)/xlators/lib/src/libxlator.c
-dht_common_source = dht-layout.c dht-helper.c dht-linkfile.c \
- dht-selfheal.c dht-rename.c dht-hashfn.c dht-diskusage.c
-
-dht_la_SOURCES = $(dht_common_source) dht.c
+dht_la_SOURCES = $(dht_common_source) dht.c
nufa_la_SOURCES = $(dht_common_source) nufa.c
switch_la_SOURCES = $(dht_common_source) switch.c
-dht_la_LDFLAGS = -module -avoidversion
+dht_la_LDFLAGS = -module -avoid-version
dht_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-nufa_la_LDFLAGS = -module -avoidversion
+nufa_la_LDFLAGS = -module -avoid-version
nufa_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-switch_la_LDFLAGS = -module -avoidversion
+switch_la_LDFLAGS = -module -avoid-version
switch_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-noinst_HEADERS = dht-common.h dht-common.c dht-mem-types.h
+noinst_HEADERS = dht-common.h dht-mem-types.h \
+ $(top_builddir)/xlators/lib/src/libxlator.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
+ -I$(top_srcdir)/xlators/lib/src
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
- -I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CFLAGS = -Wall $(GF_CFLAGS)
-CLEANFILES =
+CLEANFILES =
uninstall-local:
rm -f $(DESTDIR)$(xlatordir)/distribute.so
install-data-hook:
- ln -sf dht.so $(DESTDIR)$(xlatordir)/distribute.so \ No newline at end of file
+ ln -sf dht.so $(DESTDIR)$(xlatordir)/distribute.so
diff --git a/xlators/cluster/dht/src/dht-common.c b/xlators/cluster/dht/src/dht-common.c
index 2c8a53547..8f61339e6 100644
--- a/xlators/cluster/dht/src/dht-common.c
+++ b/xlators/cluster/dht/src/dht-common.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2009-2009 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -27,12 +18,88 @@
#include "glusterfs.h"
#include "xlator.h"
+#include "libxlator.h"
#include "dht-common.h"
#include "defaults.h"
+#include "byte-order.h"
+#include "glusterfs-acl.h"
#include <sys/time.h>
#include <libgen.h>
+int
+dht_aggregate (dict_t *this, char *key, data_t *value, void *data)
+{
+ dict_t *dst = NULL;
+ int64_t *ptr = 0, *size = NULL;
+ int32_t ret = -1;
+ data_t *dict_data = NULL;
+
+ dst = data;
+
+ if (strcmp (key, GF_XATTR_QUOTA_SIZE_KEY) == 0) {
+ ret = dict_get_bin (dst, key, (void **)&size);
+ if (ret < 0) {
+ size = GF_CALLOC (1, sizeof (int64_t),
+ gf_common_mt_char);
+ if (size == NULL) {
+ gf_log ("dht", GF_LOG_WARNING,
+ "memory allocation failed");
+ return -1;
+ }
+ ret = dict_set_bin (dst, key, size, sizeof (int64_t));
+ if (ret < 0) {
+ gf_log ("dht", GF_LOG_WARNING,
+ "dht aggregate dict set failed");
+ GF_FREE (size);
+ return -1;
+ }
+ }
+
+ ptr = data_to_bin (value);
+ if (ptr == NULL) {
+ gf_log ("dht", GF_LOG_WARNING, "data to bin failed");
+ return -1;
+ }
+
+ *size = hton64 (ntoh64 (*size) + ntoh64 (*ptr));
+
+ } else if (fnmatch (GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0) {
+ ret = gf_get_min_stime (THIS, dst, key, value);
+ if (ret < 0)
+ return ret;
+ } else {
+ /* compare user xattrs only */
+ if (!strncmp (key, "user.", strlen ("user."))) {
+ ret = dict_lookup (dst, key, &dict_data);
+ if (!ret && dict_data && value) {
+ ret = is_data_equal (dict_data, value);
+ if (!ret)
+ gf_log ("dht", GF_LOG_DEBUG,
+ "xattr mismatch for %s", key);
+ }
+ }
+ ret = dict_set (dst, key, value);
+ if (ret)
+ gf_log ("dht", GF_LOG_WARNING, "xattr dict set failed");
+ }
+
+ return 0;
+}
+
+
+void
+dht_aggregate_xattr (dict_t *dst, dict_t *src)
+{
+ if ((dst == NULL) || (src == NULL)) {
+ goto out;
+ }
+
+ dict_foreach (src, dht_aggregate, dst);
+out:
+ return;
+}
+
/* TODO:
- use volumename in xattr instead of "dht"
- use NS locks
@@ -43,219 +110,369 @@
int
dht_lookup_selfheal_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int op_ret, int op_errno)
+ xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- dht_layout_t *layout = NULL;
- int ret = 0;
+ dht_local_t *local = NULL;
+ dht_layout_t *layout = NULL;
+ int ret = -1;
- local = frame->local;
- ret = op_ret;
+ GF_VALIDATE_OR_GOTO ("dht", frame, out);
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+ GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
- dht_frame_su_undo (frame);
+ local = frame->local;
+ ret = op_ret;
- if (ret == 0) {
- layout = local->selfheal.layout;
- ret = dht_layout_set (this, local->inode, layout);
+ FRAME_SU_UNDO (frame, dht_local_t);
- if (local->ia_ino) {
- local->stbuf.ia_ino = local->ia_ino;
- } else {
- gf_log (this->name, GF_LOG_DEBUG,
- "could not find hashed subvolume for %s",
- local->loc.path);
- }
+ if (ret == 0) {
+ layout = local->selfheal.layout;
+ ret = dht_layout_set (this, local->inode, layout);
+ }
- if (local->loc.parent)
- local->postparent.ia_ino = local->loc.parent->ino;
- }
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ &local->postparent, 1);
+ }
- WIPE (&local->postparent);
+ DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
- DHT_STACK_UNWIND (lookup, frame, ret, local->op_errno, local->inode,
- &local->stbuf, local->xattr, &local->postparent);
+ DHT_STACK_UNWIND (lookup, frame, ret, local->op_errno, local->inode,
+ &local->stbuf, local->xattr, &local->postparent);
- return 0;
+out:
+ return ret;
}
int
-dht_lookup_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, dict_t *xattr,
- struct iatt *postparent)
+dht_discover_complete (xlator_t *this, call_frame_t *discover_frame)
+{
+ dht_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
+ int op_errno = 0;
+ int ret = -1;
+ dht_layout_t *layout = NULL;
+ dht_conf_t *conf = NULL;
+
+ local = discover_frame->local;
+ layout = local->layout;
+ conf = this->private;
+
+ LOCK(&discover_frame->lock);
+ {
+ main_frame = local->main_frame;
+ local->main_frame = NULL;
+ }
+ UNLOCK(&discover_frame->lock);
+
+ if (!main_frame)
+ return 0;
+
+ if (local->file_count && local->dir_count) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "path %s exists as a file on one subvolume "
+ "and directory on another. "
+ "Please fix it manually",
+ local->loc.path);
+ op_errno = EIO;
+ goto out;
+ }
+
+ if (local->cached_subvol) {
+ ret = dht_layout_preset (this, local->cached_subvol,
+ local->inode);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "failed to set layout for subvolume %s",
+ local->cached_subvol ? local->cached_subvol->name : "<nil>");
+ op_errno = EINVAL;
+ goto out;
+ }
+ } else {
+ ret = dht_layout_normalize (this, &local->loc, layout);
+ if ((ret < 0) || ((ret > 0) && (local->op_ret != 0))) {
+ /* either the layout is incorrect or the directory is
+ * not found even in one subvolume.
+ */
+ gf_log (this->name, GF_LOG_DEBUG,
+ "normalizing failed on %s "
+ "(overlaps/holes present: %s, "
+ "ENOENT errors: %d)", local->loc.path,
+ (ret < 0) ? "yes" : "no", (ret > 0) ? ret : 0);
+ if ((ret > 0) && (ret == conf->subvolume_cnt)) {
+ op_errno = ESTALE;
+ goto out;
+ }
+ }
+
+ if (local->inode)
+ dht_layout_set (this, local->inode, layout);
+ }
+
+ DHT_STACK_UNWIND (lookup, main_frame, local->op_ret, local->op_errno,
+ local->inode, &local->stbuf, local->xattr,
+ &local->postparent);
+ return 0;
+out:
+ DHT_STACK_UNWIND (lookup, main_frame, -1, op_errno, NULL, NULL, NULL,
+ NULL);
+
+ return ret;
+}
+
+
+int
+dht_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno,
+ inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+ struct iatt *postparent)
{
- dht_conf_t *conf = NULL;
dht_local_t *local = NULL;
int this_call_cnt = 0;
call_frame_t *prev = NULL;
- dht_layout_t *layout = NULL;
- int ret = 0;
- int is_dir = 0;
+ dht_layout_t *layout = NULL;
+ int ret = -1;
+ int is_dir = 0;
+ int is_linkfile = 0;
+ int attempt_unwind = 0;
+ dht_conf_t *conf = 0;
+
+ GF_VALIDATE_OR_GOTO ("dht", frame, out);
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+ GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+ GF_VALIDATE_OR_GOTO ("dht", cookie, out);
- conf = this->private;
local = frame->local;
prev = cookie;
+ conf = this->private;
- layout = local->layout;
+ layout = local->layout;
+
+ /* Check if the gfid is different for file from other node */
+ if (!op_ret && uuid_compare (local->gfid, stbuf->ia_gfid)) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: gfid different on %s",
+ local->loc.path, prev->this->name);
+ }
- if (!op_ret && uuid_is_null (local->gfid))
- memcpy (local->gfid, stbuf->ia_gfid, 16);
LOCK (&frame->lock);
{
/* TODO: assert equal mode on stbuf->st_mode and
- local->stbuf->st_mode
-
- else mkdir/chmod/chown and fix
- */
- ret = dht_layout_merge (this, layout, prev->this,
- op_ret, op_errno, xattr);
-
- if (op_ret == -1) {
- local->op_errno = ENOENT;
- gf_log (this->name, GF_LOG_DEBUG,
- "lookup of %s on %s returned error (%s)",
- local->loc.path, prev->this->name,
- strerror (op_errno));
+ local->stbuf->st_mode
- goto unlock;
- }
+ else mkdir/chmod/chown and fix
+ */
+ ret = dht_layout_merge (this, layout, prev->this,
+ op_ret, op_errno, xattr);
+ if (ret)
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: failed to merge layouts", local->loc.path);
- is_dir = check_is_dir (inode, stbuf, xattr);
- if (!is_dir) {
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
gf_log (this->name, GF_LOG_DEBUG,
- "lookup of %s on %s returned non dir 0%o",
+ "lookup of %s on %s returned error (%s)",
local->loc.path, prev->this->name,
- stbuf->ia_type);
- local->need_selfheal = 1;
- goto unlock;
+ strerror (op_errno));
+
+ goto unlock;
}
- local->op_ret = 0;
- if (local->xattr == NULL)
- local->xattr = dict_ref (xattr);
- if (local->inode == NULL)
- local->inode = inode_ref (inode);
+ is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+ conf->link_xattr_name);
+ is_dir = check_is_dir (inode, stbuf, xattr);
+
+ if (is_dir) {
+ local->dir_count ++;
+ } else {
+ local->file_count ++;
- dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
- dht_iatt_merge (this, &local->postparent, postparent,
- prev->this);
+ if (!is_linkfile) {
+ /* real file */
+ local->cached_subvol = prev->this;
+ attempt_unwind = 1;
+ } else {
+ goto unlock;
+ }
+ }
+
+ local->op_ret = 0;
- if (prev->this == dht_first_up_subvol (this)) {
- local->ia_ino = local->stbuf.ia_ino;
+ if (local->xattr == NULL) {
+ local->xattr = dict_ref (xattr);
+ } else {
+ dht_aggregate_xattr (local->xattr, xattr);
}
+ if (local->inode == NULL)
+ local->inode = inode_ref (inode);
+
+ dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+ dht_iatt_merge (this, &local->postparent, postparent,
+ prev->this);
}
unlock:
UNLOCK (&frame->lock);
+out:
+ this_call_cnt = dht_frame_return (frame);
+ if (is_last_call (this_call_cnt) || attempt_unwind) {
+ dht_discover_complete (this, frame);
+ }
- this_call_cnt = dht_frame_return (frame);
+ if (is_last_call (this_call_cnt))
+ DHT_STACK_DESTROY (frame);
- if (is_last_call (this_call_cnt)) {
- if (local->need_selfheal) {
- local->need_selfheal = 0;
- dht_lookup_everywhere (frame, this, &local->loc);
- return 0;
- }
+ return 0;
+}
- if (local->op_ret == 0) {
- ret = dht_layout_normalize (this, &local->loc, layout);
- if (ret != 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "fixing assignment on %s",
- local->loc.path);
- goto selfheal;
- }
+int
+dht_discover (call_frame_t *frame, xlator_t *this, loc_t *loc)
+{
+ int ret;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int call_cnt = 0;
+ int op_errno = EINVAL;
+ int i = 0;
+ call_frame_t *discover_frame = NULL;
- dht_layout_set (this, local->inode, layout);
+ conf = this->private;
+ local = frame->local;
- if (local->ia_ino) {
- local->stbuf.ia_ino = local->ia_ino;
- } else {
- gf_log (this->name, GF_LOG_DEBUG,
- "could not find hashed subvol for %s",
- local->loc.path);
- }
+ ret = dict_set_uint32 (local->xattr_req, conf->xattr_name, 4 * 4);
+ if (ret)
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: failed to set '%s' key",
+ loc->path, conf->xattr_name);
- if (local->loc.parent)
- local->postparent.ia_ino =
- local->loc.parent->ino;
- }
+ ret = dict_set_uint32 (local->xattr_req, conf->link_xattr_name, 256);
+ if (ret)
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: failed to set '%s' key",
+ loc->path, conf->link_xattr_name);
- DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
- local->inode, &local->stbuf, local->xattr,
- &local->postparent);
+ call_cnt = conf->subvolume_cnt;
+ local->call_cnt = call_cnt;
+
+ local->layout = dht_layout_new (this, conf->subvolume_cnt);
+
+ if (!local->layout) {
+ op_errno = ENOMEM;
+ goto err;
}
- return 0;
+ uuid_copy (local->gfid, loc->gfid);
-selfheal:
- dht_frame_su_do (frame);
- ret = dht_selfheal_directory (frame, dht_lookup_selfheal_cbk,
- &local->loc, layout);
+ discover_frame = copy_frame (frame);
+ if (!discover_frame) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- return 0;
+ discover_frame->local = local;
+ frame->local = NULL;
+ local->main_frame = frame;
+
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND (discover_frame, dht_discover_cbk,
+ conf->subvolumes[i],
+ conf->subvolumes[i]->fops->lookup,
+ &local->loc, local->xattr_req);
+ }
+
+ return 0;
+
+err:
+ DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL,
+ NULL);
+
+ return 0;
}
+
int
-dht_lookup_root_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, dict_t *xattr,
- struct iatt *postparent)
+dht_lookup_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno,
+ inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+ struct iatt *postparent)
{
- dht_conf_t *conf = NULL;
dht_local_t *local = NULL;
int this_call_cnt = 0;
call_frame_t *prev = NULL;
- dht_layout_t *layout = NULL;
- int ret = 0;
- int is_dir = 0;
+ dht_layout_t *layout = NULL;
+ int ret = -1;
+ int is_dir = 0;
+
+ GF_VALIDATE_OR_GOTO ("dht", frame, out);
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+ GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+ GF_VALIDATE_OR_GOTO ("dht", cookie, out);
- conf = this->private;
local = frame->local;
prev = cookie;
- layout = local->layout;
+ layout = local->layout;
+
+ if (!op_ret && uuid_is_null (local->gfid))
+ memcpy (local->gfid, stbuf->ia_gfid, 16);
+
+ /* Check if the gfid is different for file from other node */
+ if (!op_ret && uuid_compare (local->gfid, stbuf->ia_gfid)) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: gfid different on %s",
+ local->loc.path, prev->this->name);
+ }
LOCK (&frame->lock);
{
- ret = dht_layout_merge (this, layout, prev->this,
- op_ret, op_errno, xattr);
-
- if (op_ret == -1) {
- local->op_errno = op_errno;
- gf_log (this->name, GF_LOG_ERROR,
- "lookup of %s on %s returned error (%s)",
- local->loc.path, prev->this->name,
- strerror (op_errno));
- goto unlock;
- }
+ /* TODO: assert equal mode on stbuf->st_mode and
+ local->stbuf->st_mode
+
+ else mkdir/chmod/chown and fix
+ */
+ ret = dht_layout_merge (this, layout, prev->this,
+ op_ret, op_errno, xattr);
+
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ gf_log (this->name, GF_LOG_DEBUG,
+ "lookup of %s on %s returned error (%s)",
+ local->loc.path, prev->this->name,
+ strerror (op_errno));
+
+ goto unlock;
+ }
is_dir = check_is_dir (inode, stbuf, xattr);
if (!is_dir) {
- gf_log (this->name, GF_LOG_CRITICAL,
+ gf_log (this->name, GF_LOG_DEBUG,
"lookup of %s on %s returned non dir 0%o",
local->loc.path, prev->this->name,
stbuf->ia_type);
+ local->need_selfheal = 1;
goto unlock;
}
local->op_ret = 0;
- if (local->xattr == NULL)
+ if (local->xattr == NULL) {
local->xattr = dict_ref (xattr);
+ } else {
+ dht_aggregate_xattr (local->xattr, xattr);
+ }
+
if (local->inode == NULL)
local->inode = inode_ref (inode);
- dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
-
- if (prev->this == dht_first_up_subvol (this)) {
- local->ia_ino = local->stbuf.ia_ino;
- }
-
+ dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+ dht_iatt_merge (this, &local->postparent, postparent,
+ prev->this);
}
unlock:
UNLOCK (&frame->lock);
@@ -264,72 +481,45 @@ unlock:
this_call_cnt = dht_frame_return (frame);
if (is_last_call (this_call_cnt)) {
- if (local->op_ret == 0) {
- ret = dht_layout_normalize (this, &local->loc, layout);
- if (ret != 0) {
- gf_log (this->name, GF_LOG_INFO,
- "fixing assignment on %s",
- local->loc.path);
- }
-
- dht_layout_set (this, local->inode, layout);
- }
-
- DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
- local->inode, &local->stbuf, local->xattr,
- &local->postparent);
- }
-
- return 0;
-}
-
-static int
-dht_do_fresh_lookup_on_root (xlator_t *this, call_frame_t *frame)
-{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- int ret = -1;
- int call_cnt = 0;
- int i = 0;
-
- local = frame->local;
- conf = this->private;
+ if (local->need_selfheal) {
+ local->need_selfheal = 0;
+ dht_lookup_everywhere (frame, this, &local->loc);
+ return 0;
+ }
- if (local->layout) {
- dht_layout_unref (this, local->layout);
- local->layout = NULL;
- }
+ if (local->op_ret == 0) {
+ ret = dht_layout_normalize (this, &local->loc, layout);
- ret = dict_set_uint32 (local->xattr_req,
- "trusted.glusterfs.dht", 4 * 4);
- if (ret)
- gf_log (this->name, GF_LOG_DEBUG,
- "failed to set the dict entry for dht");
+ if (ret != 0) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "fixing assignment on %s",
+ local->loc.path);
+ goto selfheal;
+ }
- call_cnt = local->call_cnt = conf->subvolume_cnt;
+ dht_layout_set (this, local->inode, layout);
+ }
- local->layout = dht_layout_new (this,
- conf->subvolume_cnt);
- if (!local->layout) {
- local->op_errno = ENOMEM;
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- goto err;
- }
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ &local->postparent, 1);
+ }
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND (frame, dht_lookup_root_dir_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->lookup,
- &local->loc, local->xattr_req);
+ DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+ DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
+ local->inode, &local->stbuf, local->xattr,
+ &local->postparent);
}
return 0;
-err:
- DHT_STACK_UNWIND (lookup, frame, -1, local->op_errno,
- local->inode, &local->stbuf, local->xattr,
- &local->postparent);
- return 0;
+
+selfheal:
+ FRAME_SU_DO (frame, dht_local_t);
+ uuid_copy (local->loc.gfid, local->gfid);
+ ret = dht_selfheal_directory (frame, dht_lookup_selfheal_cbk,
+ &local->loc, layout);
+out:
+ return ret;
}
int
@@ -341,148 +531,224 @@ dht_revalidate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
dht_local_t *local = NULL;
int this_call_cnt = 0;
call_frame_t *prev = NULL;
- dht_layout_t *layout = NULL;
- dht_conf_t *conf = NULL;
- int ret = -1;
- int is_dir = 0;
- int is_linkfile = 0;
- unsigned char root_gfid[16] = {0,};
+ dht_layout_t *layout = NULL;
+ dht_conf_t *conf = NULL;
+ int ret = -1;
+ int is_dir = 0;
+ int is_linkfile = 0;
+ call_frame_t *copy = NULL;
+ dht_local_t *copy_local = NULL;
+
+ GF_VALIDATE_OR_GOTO ("dht", frame, err);
+ GF_VALIDATE_OR_GOTO ("dht", this, err);
+ GF_VALIDATE_OR_GOTO ("dht", frame->local, err);
+ GF_VALIDATE_OR_GOTO ("dht", cookie, err);
local = frame->local;
prev = cookie;
- conf = this->private;
+ conf = this->private;
+ if (!conf)
+ goto out;
LOCK (&frame->lock);
{
- if (op_ret == -1) {
- local->op_errno = op_errno;
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
- if ((op_errno != ENOTCONN)
+ if ((op_errno != ENOTCONN)
&& (op_errno != ENOENT)
&& (op_errno != ESTALE)) {
gf_log (this->name, GF_LOG_INFO,
- "subvolume %s returned -1 (%s)",
- prev->this->name, strerror (op_errno));
+ "subvolume %s for %s returned -1 (%s)",
+ prev->this->name, local->loc.path,
+ strerror (op_errno));
}
-
if (op_errno == ESTALE) {
- /* propogate the ESTALE to parent.
- * setting local->layout_mismatch would send
+ /* propagate the ESTALE to parent.
+ * setting local->return_estale would send
* ESTALE to parent. */
- local->layout_mismatch = 1;
+ local->return_estale = 1;
}
- goto unlock;
- }
+ /* if it is ENOENT, we may have to do a
+ * 'lookup_everywhere()' to make sure
+ * the file is not migrated */
+ if (op_errno == ENOENT) {
+ if (IA_ISREG (local->loc.inode->ia_type)) {
+ local->need_lookup_everywhere = 1;
+ }
+ }
+ goto unlock;
+ }
- if (stbuf->ia_type != local->inode->ia_type) {
- gf_log (this->name, GF_LOG_INFO,
- "mismatching filetypes 0%o v/s 0%o for %s",
- (stbuf->ia_type), (local->inode->ia_type),
- local->loc.path);
+ if (stbuf->ia_type != local->inode->ia_type) {
+ gf_log (this->name, GF_LOG_INFO,
+ "mismatching filetypes 0%o v/s 0%o for %s",
+ (stbuf->ia_type), (local->inode->ia_type),
+ local->loc.path);
- local->op_ret = -1;
- local->op_errno = EINVAL;
+ local->op_ret = -1;
+ local->op_errno = EINVAL;
- goto unlock;
- }
+ goto unlock;
+ }
- layout = local->layout;
-
- is_dir = check_is_dir (inode, stbuf, xattr);
- is_linkfile = check_is_linkfile (inode, stbuf, xattr);
-
- if (is_linkfile) {
- gf_log (this->name, GF_LOG_INFO,
- "linkfile found in revalidate for %s",
- local->loc.path);
- local->layout_mismatch = 1;
-
- goto unlock;
- }
+ layout = local->layout;
- if (is_dir) {
- ret = dht_layout_dir_mismatch (this, layout,
- prev->this, &local->loc,
- xattr);
- if (ret != 0) {
- gf_log (this->name, GF_LOG_INFO,
- "mismatching layouts for %s",
- local->loc.path);
-
- local->layout_mismatch = 1;
+ is_dir = check_is_dir (inode, stbuf, xattr);
+ is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+ conf->link_xattr_name);
- goto unlock;
- }
- }
-
- dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+ if (is_linkfile) {
+ gf_log (this->name, GF_LOG_INFO,
+ "linkfile found in revalidate for %s",
+ local->loc.path);
+ local->return_estale = 1;
+
+ goto unlock;
+ }
+
+ if (is_dir) {
+ ret = dht_dir_has_layout (xattr, conf->xattr_name);
+ if (ret >= 0) {
+ if (is_greater_time(local->stbuf.ia_ctime,
+ local->stbuf.ia_ctime_nsec,
+ stbuf->ia_ctime,
+ stbuf->ia_ctime_nsec)) {
+ local->prebuf.ia_gid = stbuf->ia_gid;
+ local->prebuf.ia_uid = stbuf->ia_uid;
+ }
+ }
+ if (local->stbuf.ia_type != IA_INVAL)
+ {
+ if ((local->stbuf.ia_gid != stbuf->ia_gid) ||
+ (local->stbuf.ia_uid != stbuf->ia_uid)) {
+ local->need_selfheal = 1;
+ }
+ }
+ ret = dht_layout_dir_mismatch (this, layout,
+ prev->this, &local->loc,
+ xattr);
+ if (ret != 0) {
+ gf_log (this->name, GF_LOG_INFO,
+ "mismatching layouts for %s",
+ local->loc.path);
+
+ local->layout_mismatch = 1;
+
+ goto unlock;
+ }
+ }
+
+ dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
dht_iatt_merge (this, &local->postparent, postparent,
prev->this);
-
- local->op_ret = 0;
- local->stbuf.ia_ino = local->ia_ino;
- if (local->loc.parent)
- local->postparent.ia_ino = local->loc.parent->ino;
+ local->op_ret = 0;
- if (!local->xattr)
- local->xattr = dict_ref (xattr);
- }
+ if (!local->xattr) {
+ local->xattr = dict_ref (xattr);
+ } else if (is_dir) {
+ dht_aggregate_xattr (local->xattr, xattr);
+ }
+ }
unlock:
- UNLOCK (&frame->lock);
-
+ UNLOCK (&frame->lock);
+out:
this_call_cnt = dht_frame_return (frame);
if (is_last_call (this_call_cnt)) {
- if (!IA_ISDIR (local->stbuf.ia_type)
- && (local->hashed_subvol != local->cached_subvol)
- && (local->stbuf.ia_nlink == 1)
- && (conf->unhashed_sticky_bit)) {
- local->stbuf.ia_prot.sticky = 1;
- }
-
- if (local->layout_mismatch) {
- local->op_ret = -1;
- local->op_errno = ESTALE;
-
- /* Because for 'root' inode, there is no FRESH lookup
- * sent from FUSE layer upon ESTALE, we need to handle
- * that one case here */
- root_gfid[15] = 1;
- if (!local->loc.parent &&
- !uuid_compare (local->loc.inode->gfid, root_gfid)) {
- dht_do_fresh_lookup_on_root (this, frame);
- return 0;
+ if (!IA_ISDIR (local->stbuf.ia_type)
+ && (local->hashed_subvol != local->cached_subvol)
+ && (local->stbuf.ia_nlink == 1)
+ && (conf && conf->unhashed_sticky_bit)) {
+ local->stbuf.ia_prot.sticky = 1;
+ }
+ if (local->need_selfheal) {
+ local->need_selfheal = 0;
+ uuid_copy (local->gfid, local->stbuf.ia_gfid);
+ local->stbuf.ia_gid = local->prebuf.ia_gid;
+ local->stbuf.ia_uid = local->prebuf.ia_uid;
+ copy = create_frame (this, this->ctx->pool);
+ if (copy) {
+ copy_local = dht_local_init (copy, &local->loc,
+ NULL, 0);
+ if (!copy_local)
+ goto cont;
+ copy_local->stbuf = local->stbuf;
+ copy->local = copy_local;
+ FRAME_SU_DO (copy, dht_local_t);
+ ret = synctask_new (this->ctx->env,
+ dht_dir_attr_heal,
+ dht_dir_attr_heal_done,
+ copy, copy);
}
- }
+ }
+cont:
+ if (local->layout_mismatch) {
+ /* Found layout mismatch in the directory, need to
+ fix this in the inode context */
+ dht_layout_unref (this, local->layout);
+ local->layout = NULL;
+ dht_lookup_directory (frame, this, &local->loc);
+ return 0;
+ }
- WIPE (&local->postparent);
+ if (local->need_lookup_everywhere) {
+ /* As the current layout gave ENOENT error, we would
+ need a new layout */
+ dht_layout_unref (this, local->layout);
+ local->layout = NULL;
+
+ /* We know that current cached subvol is no more
+ valid, get the new one */
+ local->cached_subvol = NULL;
+ dht_lookup_everywhere (frame, this, &local->loc);
+ return 0;
+ }
+ if (local->return_estale) {
+ local->op_ret = -1;
+ local->op_errno = ESTALE;
+ }
+
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ &local->postparent, 1);
+ }
+ DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
- local->inode, &local->stbuf, local->xattr,
+ local->inode, &local->stbuf, local->xattr,
&local->postparent);
- }
+ }
- return 0;
+err:
+ return ret;
}
int
dht_lookup_linkfile_create_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
+ xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent)
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *cached_subvol = NULL;
- dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *cached_subvol = NULL;
+ dht_conf_t *conf = NULL;
int ret = -1;
- local = frame->local;
- cached_subvol = local->cached_subvol;
- conf = this->private;
+ GF_VALIDATE_OR_GOTO ("dht", frame, out);
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+ GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+ GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+ local = frame->local;
+ cached_subvol = local->cached_subvol;
+ conf = this->private;
ret = dht_layout_preset (this, local->cached_subvol, inode);
if (ret < 0) {
@@ -494,74 +760,232 @@ dht_lookup_linkfile_create_cbk (call_frame_t *frame, void *cookie,
goto unwind;
}
- local->op_ret = 0;
- if ((local->stbuf.ia_nlink == 1)
- && (conf->unhashed_sticky_bit)) {
- local->stbuf.ia_prot.sticky = 1;
- }
+ local->op_ret = 0;
+ if ((local->stbuf.ia_nlink == 1)
+ && (conf && conf->unhashed_sticky_bit)) {
+ local->stbuf.ia_prot.sticky = 1;
+ }
- if (local->loc.parent)
- local->postparent.ia_ino = local->loc.parent->ino;
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ postparent, 1);
+ }
unwind:
- WIPE (&local->postparent);
+ if (local->linked == _gf_true)
+ dht_linkfile_attr_heal (frame, this);
- DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
- local->inode, &local->stbuf, local->xattr,
+ DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+ DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
+ local->inode, &local->stbuf, local->xattr,
&local->postparent);
- return 0;
+out:
+ return ret;
+}
+
+
+int
+dht_lookup_everywhere_done (call_frame_t *frame, xlator_t *this)
+{
+ int ret = 0;
+ dht_local_t *local = NULL;
+ xlator_t *hashed_subvol = NULL;
+ xlator_t *cached_subvol = NULL;
+ dht_layout_t *layout = NULL;
+
+ local = frame->local;
+ hashed_subvol = local->hashed_subvol;
+ cached_subvol = local->cached_subvol;
+
+ if (local->file_count && local->dir_count) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "path %s exists as a file on one subvolume "
+ "and directory on another. "
+ "Please fix it manually",
+ local->loc.path);
+ DHT_STACK_UNWIND (lookup, frame, -1, EIO, NULL, NULL, NULL,
+ NULL);
+ return 0;
+ }
+
+ if (local->dir_count) {
+ dht_lookup_directory (frame, this, &local->loc);
+ return 0;
+ }
+
+ if (!cached_subvol) {
+ DHT_STACK_UNWIND (lookup, frame, -1, ENOENT, NULL, NULL, NULL,
+ NULL);
+ return 0;
+ }
+
+ if (local->need_lookup_everywhere) {
+ if (uuid_compare (local->gfid, local->inode->gfid)) {
+ /* GFID different, return error */
+ DHT_STACK_UNWIND (lookup, frame, -1, ENOENT, NULL,
+ NULL, NULL, NULL);
+ return 0;
+ }
+ local->op_ret = 0;
+ local->op_errno = 0;
+ layout = dht_layout_for_subvol (this, cached_subvol);
+ if (!layout) {
+ gf_log (this->name, GF_LOG_INFO,
+ "%s: no pre-set layout for subvolume %s",
+ local->loc.path, (cached_subvol ?
+ cached_subvol->name :
+ "<nil>"));
+ }
+
+ ret = dht_layout_set (this, local->inode, layout);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_INFO,
+ "%s: failed to set layout for subvol %s",
+ local->loc.path, (cached_subvol ?
+ cached_subvol->name :
+ "<nil>"));
+ }
+
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ &local->postparent, 1);
+ }
+
+ DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+ DHT_STACK_UNWIND (lookup, frame, local->op_ret,
+ local->op_errno, local->inode,
+ &local->stbuf, local->xattr,
+ &local->postparent);
+ return 0;
+ }
+
+ if (!hashed_subvol) {
+ gf_log (this->name, GF_LOG_INFO,
+ "cannot create linkfile file for %s on %s: "
+ "hashed subvolume cannot be found.",
+ local->loc.path, cached_subvol->name);
+
+ local->op_ret = 0;
+ local->op_errno = 0;
+
+ ret = dht_layout_preset (frame->this, cached_subvol,
+ local->inode);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_INFO,
+ "failed to set layout for subvol %s",
+ cached_subvol ? cached_subvol->name :
+ "<nil>");
+ local->op_ret = -1;
+ local->op_errno = EINVAL;
+ }
+
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ &local->postparent, 1);
+ }
+
+ DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+ DHT_STACK_UNWIND (lookup, frame, local->op_ret,
+ local->op_errno, local->inode,
+ &local->stbuf, local->xattr,
+ &local->postparent);
+ return 0;
+ }
+
+ gf_log (this->name, GF_LOG_DEBUG,
+ "linking file %s existing on %s to %s (hash)",
+ local->loc.path, cached_subvol->name,
+ hashed_subvol->name);
+
+ ret = dht_linkfile_create (frame,
+ dht_lookup_linkfile_create_cbk, this,
+ cached_subvol, hashed_subvol, &local->loc);
+
+ return ret;
+}
+
+
+int
+dht_lookup_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno,
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
+{
+ int this_call_cnt = 0;
+
+ this_call_cnt = dht_frame_return (frame);
+ if (is_last_call (this_call_cnt)) {
+ dht_lookup_everywhere_done (frame, this);
+ }
+
+ return 0;
}
int
dht_lookup_everywhere_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
+ int32_t op_ret, int32_t op_errno,
inode_t *inode, struct iatt *buf, dict_t *xattr,
struct iatt *postparent)
{
- dht_conf_t *conf = NULL;
dht_local_t *local = NULL;
int this_call_cnt = 0;
call_frame_t *prev = NULL;
- int is_linkfile = 0;
- int is_dir = 0;
- xlator_t *subvol = NULL;
- loc_t *loc = NULL;
- xlator_t *link_subvol = NULL;
- xlator_t *hashed_subvol = NULL;
- xlator_t *cached_subvol = NULL;
- int ret = -1;
+ int is_linkfile = 0;
+ int is_dir = 0;
+ xlator_t *subvol = NULL;
+ loc_t *loc = NULL;
+ xlator_t *link_subvol = NULL;
+ int ret = -1;
+ int32_t fd_count = 0;
+ dht_conf_t *conf = NULL;
- conf = this->private;
+ GF_VALIDATE_OR_GOTO ("dht", frame, out);
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+ GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+ GF_VALIDATE_OR_GOTO ("dht", this->private, out);
- local = frame->local;
- loc = &local->loc;
+ local = frame->local;
+ loc = &local->loc;
+ conf = this->private;
- prev = cookie;
- subvol = prev->this;
+ prev = cookie;
+ subvol = prev->this;
+
+ LOCK (&frame->lock);
+ {
+ if (op_ret == -1) {
+ if (op_errno != ENOENT)
+ local->op_errno = op_errno;
+ goto unlock;
+ }
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- if (op_errno != ENOENT)
- local->op_errno = op_errno;
- goto unlock;
- }
if (uuid_is_null (local->gfid))
- memcpy (local->gfid, buf->ia_gfid, 16);
-
- is_linkfile = check_is_linkfile (inode, buf, xattr);
- is_dir = check_is_dir (inode, buf, xattr);
-
- if (is_linkfile) {
- link_subvol = dht_linkfile_subvol (this, inode, buf,
- xattr);
- gf_log (this->name, GF_LOG_DEBUG,
- "found on %s linkfile %s (-> %s)",
- subvol->name, loc->path,
- link_subvol ? link_subvol->name : "''");
- goto unlock;
- }
+ uuid_copy (local->gfid, buf->ia_gfid);
+
+ if (uuid_compare (local->gfid, buf->ia_gfid)) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: gfid differs on subvolume %s",
+ loc->path, prev->this->name);
+ }
+
+ is_linkfile = check_is_linkfile (inode, buf, xattr,
+ conf->link_xattr_name);
+ is_dir = check_is_dir (inode, buf, xattr);
+
+ if (is_linkfile) {
+ link_subvol = dht_linkfile_subvol (this, inode, buf,
+ xattr);
+ gf_log (this->name, GF_LOG_DEBUG,
+ "found on %s linkfile %s (-> %s)",
+ subvol->name, loc->path,
+ link_subvol ? link_subvol->name : "''");
+ goto unlock;
+ }
+
+ /* non linkfile GFID takes precedence */
+ uuid_copy (local->gfid, buf->ia_gfid);
if (is_dir) {
local->dir_count++;
@@ -581,126 +1005,82 @@ dht_lookup_everywhere_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
gf_log (this->name, GF_LOG_DEBUG,
"found on %s file %s",
subvol->name, loc->path);
-
+
dht_iatt_merge (this, &local->postparent,
postparent, subvol);
} else {
- gf_log (this->name, GF_LOG_DEBUG,
+ /* This is where we need 'rename' both entries logic */
+ gf_log (this->name, GF_LOG_WARNING,
"multiple subvolumes (%s and %s) have "
- "file %s", local->cached_subvol->name,
+ "file %s (preferably rename the file "
+ "in the backend, and do a fresh lookup)",
+ local->cached_subvol->name,
subvol->name, local->loc.path);
}
}
- }
+ }
unlock:
- UNLOCK (&frame->lock);
-
- if (is_linkfile) {
- gf_log (this->name, GF_LOG_DEBUG,
- "deleting stale linkfile %s on %s",
- loc->path, subvol->name);
- dht_linkfile_unlink (frame, this, subvol, loc);
- }
-
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- hashed_subvol = local->hashed_subvol;
- cached_subvol = local->cached_subvol;
-
- if (local->file_count && local->dir_count) {
- gf_log (this->name, GF_LOG_ERROR,
- "path %s exists as a file on one subvolume "
- "and directory on another. "
- "Please fix it manually",
- loc->path);
- DHT_STACK_UNWIND (lookup, frame, -1, EIO, NULL, NULL, NULL,
- NULL);
- return 0;
- }
-
- if (local->dir_count) {
- dht_lookup_directory (frame, this, &local->loc);
- return 0;
- }
-
- if (!cached_subvol) {
- DHT_STACK_UNWIND (lookup, frame, -1, ENOENT, NULL, NULL, NULL,
- NULL);
- return 0;
- }
-
- if (!hashed_subvol) {
- gf_log (this->name, GF_LOG_DEBUG,
- "cannot create linkfile file for %s on %s: "
- "hashed subvolume cannot be found.",
- loc->path, cached_subvol->name);
-
- local->op_ret = 0;
- local->op_errno = 0;
-
- ret = dht_layout_preset (frame->this, cached_subvol,
- local->inode);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "failed to set layout for subvol %s",
- cached_subvol ? cached_subvol->name :
- "<nil>");
- local->op_ret = -1;
- local->op_errno = EINVAL;
- }
-
- if (local->loc.parent)
- local->postparent.ia_ino =
- local->loc.parent->ino;
-
- WIPE (&local->postparent);
+ UNLOCK (&frame->lock);
- DHT_STACK_UNWIND (lookup, frame, local->op_ret,
- local->op_errno, local->inode,
- &local->stbuf, local->xattr,
- &local->postparent);
+ if (is_linkfile) {
+ ret = dict_get_int32 (xattr, GLUSTERFS_OPEN_FD_COUNT, &fd_count);
+ /* Delete the linkfile only if there are no open fds on it.
+ if there is a open-fd, it may be in migration */
+ if (!ret && (fd_count == 0)) {
+ gf_log (this->name, GF_LOG_INFO,
+ "deleting stale linkfile %s on %s",
+ loc->path, subvol->name);
+ STACK_WIND (frame, dht_lookup_unlink_cbk,
+ subvol, subvol->fops->unlink, loc, 0, NULL);
return 0;
}
+ }
- gf_log (this->name, GF_LOG_DEBUG,
- "linking file %s existing on %s to %s (hash)",
- loc->path, cached_subvol->name,
- hashed_subvol->name);
-
- dht_linkfile_create (frame,
- dht_lookup_linkfile_create_cbk,
- cached_subvol, hashed_subvol, loc);
- }
+ this_call_cnt = dht_frame_return (frame);
+ if (is_last_call (this_call_cnt)) {
+ dht_lookup_everywhere_done (frame, this);
+ }
- return 0;
+out:
+ return ret;
}
int
dht_lookup_everywhere (call_frame_t *frame, xlator_t *this, loc_t *loc)
{
- dht_conf_t *conf = NULL;
- dht_local_t *local = NULL;
- int i = 0;
- int call_cnt = 0;
+ dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ int i = 0;
+ int call_cnt = 0;
- conf = this->private;
- local = frame->local;
+ GF_VALIDATE_OR_GOTO ("dht", frame, err);
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+ GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+ GF_VALIDATE_OR_GOTO ("dht", loc, out);
- call_cnt = conf->subvolume_cnt;
- local->call_cnt = call_cnt;
+ conf = this->private;
+ local = frame->local;
- if (!local->inode)
- local->inode = inode_ref (loc->inode);
+ call_cnt = conf->subvolume_cnt;
+ local->call_cnt = call_cnt;
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND (frame, dht_lookup_everywhere_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->lookup,
- loc, local->xattr_req);
- }
+ if (!local->inode)
+ local->inode = inode_ref (loc->inode);
- return 0;
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND (frame, dht_lookup_everywhere_cbk,
+ conf->subvolumes[i],
+ conf->subvolumes[i]->fops->lookup,
+ loc, local->xattr_req);
+ }
+
+ return 0;
+out:
+ DHT_STACK_UNWIND (lookup, frame, -1, EINVAL, NULL, NULL, NULL, NULL);
+err:
+ return -1;
}
@@ -711,60 +1091,82 @@ dht_lookup_linkfile_cbk (call_frame_t *frame, void *cookie,
struct iatt *postparent)
{
call_frame_t *prev = NULL;
- dht_local_t *local = NULL;
- xlator_t *subvol = NULL;
- loc_t *loc = NULL;
- dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+ loc_t *loc = NULL;
+ dht_conf_t *conf = NULL;
int ret = 0;
+ GF_VALIDATE_OR_GOTO ("dht", frame, out);
+ GF_VALIDATE_OR_GOTO ("dht", this, unwind);
+ GF_VALIDATE_OR_GOTO ("dht", frame->local, unwind);
+ GF_VALIDATE_OR_GOTO ("dht", this->private, unwind);
+ GF_VALIDATE_OR_GOTO ("dht", cookie, unwind);
+
prev = cookie;
- subvol = prev->this;
- conf = this->private;
- local = frame->local;
- loc = &local->loc;
+ subvol = prev->this;
+ conf = this->private;
+ local = frame->local;
+ loc = &local->loc;
if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "lookup of %s on %s (following linkfile) failed (%s)",
- local->loc.path, subvol->name, strerror (op_errno));
- goto err;
- }
+ gf_log (this->name, GF_LOG_INFO,
+ "lookup of %s on %s (following linkfile) failed (%s)",
+ local->loc.path, subvol->name, strerror (op_errno));
+
+ /* If cached subvol returned ENOTCONN, do not do
+ lookup_everywhere. We need to make sure linkfile does not get
+ removed, which can take away the namespace, and subvol is
+ anyways down. */
+
+ if (op_errno != ENOTCONN)
+ goto err;
+ else
+ goto unwind;
+ }
if (check_is_dir (inode, stbuf, xattr)) {
- gf_log (this->name, GF_LOG_DEBUG,
+ gf_log (this->name, GF_LOG_INFO,
"lookup of %s on %s (following linkfile) reached dir",
local->loc.path, subvol->name);
goto err;
}
- if (check_is_linkfile (inode, stbuf, xattr)) {
- gf_log (this->name, GF_LOG_DEBUG,
+ if (check_is_linkfile (inode, stbuf, xattr, conf->link_xattr_name)) {
+ gf_log (this->name, GF_LOG_INFO,
"lookup of %s on %s (following linkfile) reached link",
local->loc.path, subvol->name);
goto err;
}
- if ((stbuf->ia_nlink == 1)
- && (conf->unhashed_sticky_bit)) {
- stbuf->ia_prot.sticky = 1;
- }
- dht_itransform (this, prev->this, stbuf->ia_ino, &stbuf->ia_ino);
- if (local->loc.parent)
- postparent->ia_ino = local->loc.parent->ino;
-
- ret = dht_layout_preset (this, prev->this, inode);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
+ if (uuid_compare (local->gfid, stbuf->ia_gfid)) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: gfid different on data file on %s",
+ local->loc.path, subvol->name);
+ goto err;
+ }
+
+ if ((stbuf->ia_nlink == 1)
+ && (conf && conf->unhashed_sticky_bit)) {
+ stbuf->ia_prot.sticky = 1;
+ }
+
+ ret = dht_layout_preset (this, prev->this, inode);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_INFO,
"failed to set layout for subvolume %s",
prev->this->name);
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
+ op_ret = -1;
+ op_errno = EINVAL;
+ }
-out:
- WIPE (postparent);
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ postparent, 1);
+ }
+unwind:
+ DHT_STRIP_PHASE1_FLAGS (stbuf);
DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, stbuf, xattr,
postparent);
@@ -772,7 +1174,7 @@ out:
err:
dht_lookup_everywhere (frame, this, loc);
-
+out:
return 0;
}
@@ -784,21 +1186,38 @@ dht_lookup_directory (call_frame_t *frame, xlator_t *this, loc_t *loc)
int i = 0;
dht_conf_t *conf = NULL;
dht_local_t *local = NULL;
+ int ret = 0;
+
+ GF_VALIDATE_OR_GOTO ("dht", frame, out);
+ GF_VALIDATE_OR_GOTO ("dht", this, unwind);
+ GF_VALIDATE_OR_GOTO ("dht", frame->local, unwind);
+ GF_VALIDATE_OR_GOTO ("dht", this->private, unwind);
+ GF_VALIDATE_OR_GOTO ("dht", loc, unwind);
conf = this->private;
local = frame->local;
call_cnt = conf->subvolume_cnt;
local->call_cnt = call_cnt;
-
+
local->layout = dht_layout_new (this, conf->subvolume_cnt);
if (!local->layout) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- DHT_STACK_UNWIND (lookup, frame, -1, ENOMEM, NULL, NULL, NULL, NULL);
- return 0;
+ goto unwind;
+ }
+
+ if (local->xattr != NULL) {
+ dict_unref (local->xattr);
+ local->xattr = NULL;
}
-
+
+ if (!uuid_is_null (local->gfid)) {
+ ret = dict_set_static_bin (local->xattr_req, "gfid-req",
+ local->gfid, 16);
+ if (ret)
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: failed to set gfid", local->loc.path);
+ }
+
for (i = 0; i < call_cnt; i++) {
STACK_WIND (frame, dht_lookup_dir_cbk,
conf->subvolumes[i],
@@ -806,6 +1225,11 @@ dht_lookup_directory (call_frame_t *frame, xlator_t *this, loc_t *loc)
&local->loc, local->xattr_req);
}
return 0;
+unwind:
+ DHT_STACK_UNWIND (lookup, frame, -1, ENOMEM, NULL, NULL, NULL, NULL);
+out:
+ return 0;
+
}
@@ -823,9 +1247,14 @@ dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
loc_t *loc = NULL;
call_frame_t *prev = NULL;
int ret = 0;
- uint64_t tmp_layout = 0;
dht_layout_t *parent_layout = NULL;
+ GF_VALIDATE_OR_GOTO ("dht", frame, err);
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+ GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+ GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+
conf = this->private;
prev = cookie;
@@ -835,99 +1264,137 @@ dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
/* This is required for handling stale linkfile deletion,
* or any more call which happens from this 'loc'.
*/
- if (uuid_is_null (local->gfid) && !op_ret)
+ if (!op_ret && uuid_is_null (local->gfid))
memcpy (local->gfid, stbuf->ia_gfid, 16);
- if (ENTRY_MISSING (op_ret, op_errno)) {
+ if (ENTRY_MISSING (op_ret, op_errno)) {
+ gf_log (this->name, GF_LOG_TRACE, "Entry %s missing on subvol"
+ " %s", loc->path, prev->this->name);
if (conf->search_unhashed == GF_DHT_LOOKUP_UNHASHED_ON) {
- local->op_errno = ENOENT;
- dht_lookup_everywhere (frame, this, loc);
- return 0;
- }
+ local->op_errno = ENOENT;
+ dht_lookup_everywhere (frame, this, loc);
+ return 0;
+ }
if ((conf->search_unhashed == GF_DHT_LOOKUP_UNHASHED_AUTO) &&
(loc->parent)) {
- ret = inode_ctx_get (loc->parent, this, &tmp_layout);
- parent_layout = (dht_layout_t *)(long)tmp_layout;
+ ret = dht_inode_ctx_layout_get (loc->parent, this,
+ &parent_layout);
+ if (ret || !parent_layout)
+ goto out;
if (parent_layout->search_unhashed) {
local->op_errno = ENOENT;
dht_lookup_everywhere (frame, this, loc);
return 0;
}
}
- }
+ }
- if (op_ret == 0) {
- is_dir = check_is_dir (inode, stbuf, xattr);
- if (is_dir) {
- local->inode = inode_ref (inode);
- local->xattr = dict_ref (xattr);
- }
- }
+ if (op_ret == 0) {
+ is_dir = check_is_dir (inode, stbuf, xattr);
+ if (is_dir) {
+ local->inode = inode_ref (inode);
+ local->xattr = dict_ref (xattr);
+ }
+ }
- if (is_dir || (op_ret == -1 && op_errno == ENOTCONN)) {
+ if (is_dir || (op_ret == -1 && op_errno == ENOTCONN)) {
dht_lookup_directory (frame, this, &local->loc);
return 0;
- }
-
- if (op_ret == -1)
+ }
+
+ if (op_ret == -1) {
+ gf_log (this->name, GF_LOG_DEBUG, "Lookup of %s for subvolume"
+ " %s failed with error %s", loc->path, prev->this->name,
+ strerror (op_errno));
goto out;
+ }
- is_linkfile = check_is_linkfile (inode, stbuf, xattr);
- is_dir = check_is_dir (inode, stbuf, xattr);
+ is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+ conf->link_xattr_name);
- if (!is_dir && !is_linkfile) {
+ if (!is_linkfile) {
/* non-directory and not a linkfile */
- dht_itransform (this, prev->this, stbuf->ia_ino,
- &stbuf->ia_ino);
- if (loc->parent)
- postparent->ia_ino = loc->parent->ino;
-
- ret = dht_layout_preset (this, prev->this, inode);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "could not set pre-set layout for subvolume %s",
- prev->this->name);
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
- goto out;
- }
-
- if (is_linkfile) {
- subvol = dht_linkfile_subvol (this, inode, stbuf, xattr);
-
- if (!subvol) {
- gf_log (this->name, GF_LOG_DEBUG,
- "linkfile not having link subvolume. path=%s",
- loc->path);
- dht_lookup_everywhere (frame, this, loc);
- return 0;
+ ret = dht_layout_preset (this, prev->this, inode);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_INFO,
+ "could not set pre-set layout for subvolume %s",
+ prev->this->name);
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto out;
}
+ goto out;
+ }
- STACK_WIND (frame, dht_lookup_linkfile_cbk,
- subvol, subvol->fops->lookup,
- &local->loc, local->xattr_req);
+ subvol = dht_linkfile_subvol (this, inode, stbuf, xattr);
+ if (!subvol) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "linkfile not having link subvolume. path=%s",
+ loc->path);
+ dht_lookup_everywhere (frame, this, loc);
+ return 0;
}
+ STACK_WIND (frame, dht_lookup_linkfile_cbk,
+ subvol, subvol->fops->lookup,
+ &local->loc, local->xattr_req);
+
return 0;
out:
- /*
- * FIXME: postparent->ia_size and postparent->st_blocks do not have
- * correct values. since, postparent corresponds to a directory these
+ /*
+ * FIXME: postparent->ia_size and postparent->st_blocks do not have
+ * correct values. since, postparent corresponds to a directory these
* two members should have values equal to sum of corresponding values
* from each of the subvolume. See dht_iatt_merge for reference.
*/
- WIPE (postparent);
+ if (!op_ret && local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ postparent, 1);
+ }
+ DHT_STRIP_PHASE1_FLAGS (stbuf);
DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, stbuf, xattr,
postparent);
+err:
return 0;
}
+/* For directories, check if acl xattrs have been requested (by the acl xlator),
+ * if not, request for them. These xattrs are needed for dht dir self-heal to
+ * perform proper self-healing of dirs
+ */
+void
+dht_check_and_set_acl_xattr_req (inode_t *inode, dict_t *xattr_req)
+{
+ int ret = 0;
+
+ GF_ASSERT (inode);
+ GF_ASSERT (xattr_req);
+
+ if (inode->ia_type != IA_IFDIR)
+ return;
+
+ if (!dict_get (xattr_req, POSIX_ACL_ACCESS_XATTR)) {
+ ret = dict_set_int8 (xattr_req, POSIX_ACL_ACCESS_XATTR, 0);
+ if (ret)
+ gf_log (THIS->name, GF_LOG_WARNING,
+ "failed to set key %s",
+ POSIX_ACL_ACCESS_XATTR);
+ }
+
+ if (!dict_get (xattr_req, POSIX_ACL_DEFAULT_XATTR)) {
+ ret = dict_set_int8 (xattr_req, POSIX_ACL_DEFAULT_XATTR, 0);
+ if (ret)
+ gf_log (THIS->name, GF_LOG_WARNING,
+ "failed to set key %s",
+ POSIX_ACL_DEFAULT_XATTR);
+ }
+
+ return;
+}
int
dht_lookup (call_frame_t *frame, xlator_t *this,
@@ -935,34 +1402,40 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
{
xlator_t *subvol = NULL;
xlator_t *hashed_subvol = NULL;
- xlator_t *cached_subvol = NULL;
dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
+ dht_conf_t *conf = NULL;
int ret = -1;
int op_errno = -1;
- dht_layout_t *layout = NULL;
- int i = 0;
- int call_cnt = 0;
-
+ dht_layout_t *layout = NULL;
+ int i = 0;
+ int call_cnt = 0;
+ loc_t new_loc = {0,};
VALIDATE_OR_GOTO (frame, err);
VALIDATE_OR_GOTO (this, err);
VALIDATE_OR_GOTO (loc, err);
VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
- conf = this->private;
+ conf = this->private;
+ if (!conf)
+ goto err;
- local = dht_local_init (frame);
- if (!local) {
- op_errno = ENOMEM;
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- goto err;
- }
- if (!dht_filter_loc_subvol_key (this, loc, &local->loc,
- &hashed_subvol)) {
- ret = loc_dup (loc, &local->loc);
+ local = dht_local_init (frame, loc, NULL, GF_FOP_LOOKUP);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ ret = dht_filter_loc_subvol_key (this, loc, &new_loc,
+ &hashed_subvol);
+ if (ret) {
+ loc_wipe (&local->loc);
+ ret = loc_dup (&new_loc, &local->loc);
+
+ /* we no more need 'new_loc' entries */
+ loc_wipe (&new_loc);
+
+ /* check if loc_dup() is successful */
if (ret == -1) {
op_errno = errno;
gf_log (this->name, GF_LOG_DEBUG,
@@ -972,22 +1445,25 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
}
}
- if (xattr_req) {
- local->xattr_req = dict_ref (xattr_req);
- } else {
- local->xattr_req = dict_new ();
- }
+ if (xattr_req) {
+ local->xattr_req = dict_ref (xattr_req);
+ } else {
+ local->xattr_req = dict_new ();
+ }
+
+ if (uuid_is_null (loc->pargfid) && !uuid_is_null (loc->gfid) &&
+ !__is_root_gfid (loc->inode->gfid)) {
+ local->cached_subvol = NULL;
+ dht_discover (frame, this, loc);
+ return 0;
+ }
if (!hashed_subvol)
hashed_subvol = dht_subvol_get_hashed (this, loc);
- cached_subvol = dht_subvol_get_cached (this, loc->inode);
-
- local->cached_subvol = cached_subvol;
- local->hashed_subvol = hashed_subvol;
+ local->hashed_subvol = hashed_subvol;
if (is_revalidate (loc)) {
- local->layout = layout = dht_layout_get (this, loc->inode);
-
+ layout = local->layout;
if (!layout) {
gf_log (this->name, GF_LOG_DEBUG,
"revalidate without cache. path=%s",
@@ -996,71 +1472,93 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
goto err;
}
- if (layout->gen && (layout->gen < conf->gen)) {
- gf_log (this->name, GF_LOG_TRACE,
- "incomplete layout failure for path=%s",
- loc->path);
+ if (layout->gen && (layout->gen < conf->gen)) {
+ gf_log (this->name, GF_LOG_TRACE,
+ "incomplete layout failure for path=%s",
+ loc->path);
dht_layout_unref (this, local->layout);
local->layout = NULL;
- goto do_fresh_lookup;
- }
+ local->cached_subvol = NULL;
+ goto do_fresh_lookup;
+ }
+
+ local->inode = inode_ref (loc->inode);
+
+ /* NOTE: we don't require 'trusted.glusterfs.dht.linkto' attribute,
+ * revalidates directly go to the cached-subvolume.
+ */
+ ret = dict_set_uint32 (local->xattr_req,
+ conf->xattr_name, 4 * 4);
+
+ if (IA_ISDIR (local->inode->ia_type)) {
+ local->call_cnt = call_cnt = conf->subvolume_cnt;
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND (frame, dht_revalidate_cbk,
+ conf->subvolumes[i],
+ conf->subvolumes[i]->fops->lookup,
+ loc, local->xattr_req);
+ }
+ return 0;
+ }
- local->inode = inode_ref (loc->inode);
- local->ia_ino = loc->inode->ino;
-
- local->call_cnt = layout->cnt;
- call_cnt = local->call_cnt;
-
- /* NOTE: we don't require 'trusted.glusterfs.dht.linkto' attribute,
- * revalidates directly go to the cached-subvolume.
- */
- ret = dict_set_uint32 (local->xattr_req,
- "trusted.glusterfs.dht", 4 * 4);
-
- for (i = 0; i < layout->cnt; i++) {
+ call_cnt = local->call_cnt = layout->cnt;
+
+ /* need it for self-healing linkfiles which is
+ 'in-migration' state */
+ ret = dict_set_uint32 (local->xattr_req,
+ GLUSTERFS_OPEN_FD_COUNT, 4);
+
+ /* need it for dir self-heal */
+ dht_check_and_set_acl_xattr_req (loc->inode, local->xattr_req);
+
+ for (i = 0; i < call_cnt; i++) {
subvol = layout->list[i].xlator;
-
+
STACK_WIND (frame, dht_revalidate_cbk,
subvol, subvol->fops->lookup,
&local->loc, local->xattr_req);
- if (!--call_cnt)
- break;
}
} else {
do_fresh_lookup:
- /* TODO: remove the hard-coding */
- ret = dict_set_uint32 (local->xattr_req,
- "trusted.glusterfs.dht", 4 * 4);
+ /* TODO: remove the hard-coding */
+ ret = dict_set_uint32 (local->xattr_req,
+ conf->xattr_name, 4 * 4);
+
+ ret = dict_set_uint32 (local->xattr_req,
+ conf->link_xattr_name, 256);
- ret = dict_set_uint32 (local->xattr_req,
- "trusted.glusterfs.dht.linkto", 256);
+ /* need it for self-healing linkfiles which is
+ 'in-migration' state */
+ ret = dict_set_uint32 (local->xattr_req,
+ GLUSTERFS_OPEN_FD_COUNT, 4);
+
+ /* need it for dir self-heal */
+ dht_check_and_set_acl_xattr_req (loc->inode, local->xattr_req);
if (!hashed_subvol) {
- gf_log (this->name, GF_LOG_DEBUG,
- "no subvolume in layout for path=%s, "
- "checking on all the subvols to see if "
- "it is a directory", loc->path);
- call_cnt = conf->subvolume_cnt;
- local->call_cnt = call_cnt;
-
- local->layout = dht_layout_new (this,
+ gf_log (this->name, GF_LOG_DEBUG,
+ "no subvolume in layout for path=%s, "
+ "checking on all the subvols to see if "
+ "it is a directory", loc->path);
+ call_cnt = conf->subvolume_cnt;
+ local->call_cnt = call_cnt;
+
+ local->layout = dht_layout_new (this,
conf->subvolume_cnt);
- if (!local->layout) {
- op_errno = ENOMEM;
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- goto err;
- }
-
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND (frame, dht_lookup_dir_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->lookup,
- &local->loc, local->xattr_req);
- }
- return 0;
+ if (!local->layout) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND (frame, dht_lookup_dir_cbk,
+ conf->subvolumes[i],
+ conf->subvolumes[i]->fops->lookup,
+ &local->loc, local->xattr_req);
+ }
+ return 0;
}
STACK_WIND (frame, dht_lookup_cbk,
@@ -1071,336 +1569,52 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
- return 0;
-}
-
-
-int
-dht_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
-{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- call_frame_t *prev = NULL;
-
-
- local = frame->local;
- prev = cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- local->op_ret = -1;
- gf_log (this->name, GF_LOG_DEBUG,
- "subvolume %s returned -1 (%s)",
- prev->this->name, strerror (op_errno));
- goto unlock;
- }
-
- dht_iatt_merge (this, &local->prebuf, prebuf, prev->this);
- dht_iatt_merge (this, &local->stbuf, postbuf, prev->this);
-
- if (local->inode) {
- local->stbuf.ia_ino = local->inode->ino;
- local->prebuf.ia_ino = local->inode->ino;
- }
-
- local->op_ret = 0;
- }
-unlock:
- UNLOCK (&frame->lock);
-
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt))
- DHT_STACK_UNWIND (truncate, frame, local->op_ret, local->op_errno,
- &local->prebuf, &local->stbuf);
-
- return 0;
-}
-
-
-
-int
-dht_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *stbuf)
-{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- call_frame_t *prev = NULL;
-
-
- local = frame->local;
- prev = cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- gf_log (this->name, GF_LOG_DEBUG,
- "subvolume %s returned -1 (%s)",
- prev->this->name, strerror (op_errno));
- goto unlock;
- }
-
- dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
-
- if (local->inode)
- local->stbuf.ia_ino = local->inode->ino;
- local->op_ret = 0;
- }
-unlock:
- UNLOCK (&frame->lock);
-
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt))
- DHT_STACK_UNWIND (stat, frame, local->op_ret, local->op_errno,
- &local->stbuf);
-
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL,
+ NULL);
return 0;
}
int
-dht_stat (call_frame_t *frame, xlator_t *this,
- loc_t *loc)
-{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
- dht_layout_t *layout = NULL;
- int i = 0;
-
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
-
-
- local = dht_local_init (frame);
- if (!local) {
- op_errno = ENOMEM;
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- goto err;
- }
-
- local->layout = layout = dht_layout_get (this, loc->inode);
- if (!layout) {
- gf_log (this->name, GF_LOG_DEBUG,
- "no layout for path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
-
- local->inode = inode_ref (loc->inode);
- local->call_cnt = layout->cnt;
-
- for (i = 0; i < layout->cnt; i++) {
- subvol = layout->list[i].xlator;
-
- STACK_WIND (frame, dht_attr_cbk,
- subvol, subvol->fops->stat,
- loc);
- }
-
- return 0;
-
-err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL);
-
- return 0;
-}
-
-
-int
-dht_fstat (call_frame_t *frame, xlator_t *this,
- fd_t *fd)
-{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
- dht_layout_t *layout = NULL;
- int i = 0;
-
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- local = dht_local_init (frame);
- if (!local) {
- op_errno = ENOMEM;
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- goto err;
- }
-
- local->layout = layout = dht_layout_get (this, fd->inode);
- if (!layout) {
- gf_log (this->name, GF_LOG_DEBUG,
- "no layout for fd=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
-
- local->inode = inode_ref (fd->inode);
- local->call_cnt = layout->cnt;;
-
- for (i = 0; i < layout->cnt; i++) {
- subvol = layout->list[i].xlator;
- STACK_WIND (frame, dht_attr_cbk,
- subvol, subvol->fops->fstat,
- fd);
- }
-
- return 0;
-
-err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fstat, frame, -1, op_errno, NULL);
-
- return 0;
-}
-
-
-int
-dht_truncate (call_frame_t *frame, xlator_t *this,
- loc_t *loc, off_t offset)
-{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
-
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
-
- subvol = dht_subvol_get_cached (this, loc->inode);
- if (!subvol) {
- gf_log (this->name, GF_LOG_DEBUG,
- "no cached subvolume for path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
-
- local = dht_local_init (frame);
- if (!local) {
- op_errno = ENOMEM;
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- goto err;
- }
-
- local->inode = inode_ref (loc->inode);
- local->call_cnt = 1;
-
- STACK_WIND (frame, dht_truncate_cbk,
- subvol, subvol->fops->truncate,
- loc, offset);
-
- return 0;
-
-err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL);
-
- return 0;
-}
-
-
-int
-dht_ftruncate (call_frame_t *frame, xlator_t *this,
- fd_t *fd, off_t offset)
-{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
-
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- subvol = dht_subvol_get_cached (this, fd->inode);
- if (!subvol) {
- gf_log (this->name, GF_LOG_DEBUG,
- "no cached subvolume for fd=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
-
- local = dht_local_init (frame);
- if (!local) {
- op_errno = ENOMEM;
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- goto err;
- }
-
- local->inode = inode_ref (fd->inode);
- local->call_cnt = 1;
-
- STACK_WIND (frame, dht_truncate_cbk,
- subvol, subvol->fops->ftruncate,
- fd, offset);
-
- return 0;
-
-err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL);
-
- return 0;
-}
-
-
-int
dht_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
- dht_local_t *local = NULL;
- call_frame_t *prev = NULL;
+ dht_local_t *local = NULL;
+ call_frame_t *prev = NULL;
- local = frame->local;
- prev = cookie;
+ local = frame->local;
+ prev = cookie;
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
+ LOCK (&frame->lock);
+ {
+ if (op_ret == -1) {
local->op_ret = -1;
- local->op_errno = op_errno;
- gf_log (this->name, GF_LOG_DEBUG,
- "subvolume %s returned -1 (%s)",
- prev->this->name, strerror (op_errno));
- goto unlock;
- }
+ local->op_errno = op_errno;
+ gf_log (this->name, GF_LOG_DEBUG,
+ "subvolume %s returned -1 (%s)",
+ prev->this->name, strerror (op_errno));
+ goto unlock;
+ }
- preparent->ia_ino = local->loc.parent->ino;
- postparent->ia_ino = local->loc.parent->ino;
- local->op_ret = 0;
+ local->op_ret = 0;
local->postparent = *postparent;
local->preparent = *preparent;
- WIPE (&local->postparent);
- WIPE (&local->preparent);
- }
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ &local->preparent, 0);
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ &local->postparent, 1);
+ }
+ }
unlock:
- UNLOCK (&frame->lock);
+ UNLOCK (&frame->lock);
DHT_STACK_UNWIND (unlink, frame, local->op_ret, local->op_errno,
- &local->preparent, &local->postparent);
+ &local->preparent, &local->postparent, NULL);
return 0;
}
@@ -1409,32 +1623,33 @@ unlock:
int
dht_unlink_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
- dht_local_t *local = NULL;
- call_frame_t *prev = NULL;
+ dht_local_t *local = NULL;
+ call_frame_t *prev = NULL;
xlator_t *cached_subvol = NULL;
- local = frame->local;
- prev = cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- gf_log (this->name, GF_LOG_DEBUG,
- "subvolume %s returned -1 (%s)",
- prev->this->name, strerror (op_errno));
- goto unlock;
- }
+ local = frame->local;
+ prev = cookie;
- local->op_ret = 0;
- }
+ LOCK (&frame->lock);
+ {
+ if ((op_ret == -1) && !((op_errno == ENOENT) ||
+ (op_errno == ENOTCONN))) {
+ local->op_errno = op_errno;
+ gf_log (this->name, GF_LOG_DEBUG,
+ "subvolume %s returned -1 (%s)",
+ prev->this->name, strerror (op_errno));
+ goto unlock;
+ }
+
+ local->op_ret = 0;
+ }
unlock:
- UNLOCK (&frame->lock);
+ UNLOCK (&frame->lock);
- if (op_ret == -1)
+ if (local->op_ret == -1)
goto err;
cached_subvol = dht_subvol_get_cached (this, local->loc.inode);
@@ -1448,321 +1663,311 @@ unlock:
STACK_WIND (frame, dht_unlink_cbk,
cached_subvol, cached_subvol->fops->unlink,
- &local->loc);
+ &local->loc, local->flags, NULL);
return 0;
err:
DHT_STACK_UNWIND (unlink, frame, -1, local->op_errno,
- NULL, NULL);
+ NULL, NULL, NULL);
return 0;
}
-
int
-dht_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
- int op_errno, struct iatt *prebuf, struct iatt *postbuf)
+dht_err_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- call_frame_t *prev = NULL;
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ call_frame_t *prev = NULL;
+ local = frame->local;
+ prev = cookie;
- local = frame->local;
- prev = cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- gf_log (this->name, GF_LOG_DEBUG,
- "subvolume %s returned -1 (%s)",
- prev->this->name, strerror (op_errno));
- goto unlock;
- }
+ LOCK (&frame->lock);
+ {
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ gf_log (this->name, GF_LOG_DEBUG,
+ "subvolume %s returned -1 (%s)",
+ prev->this->name, strerror (op_errno));
+ goto unlock;
+ }
- local->op_ret = 0;
- }
+ local->op_ret = 0;
+ }
unlock:
- UNLOCK (&frame->lock);
+ UNLOCK (&frame->lock);
- if (local && (op_ret == 0)) {
- prebuf->ia_ino = local->ia_ino;
- postbuf->ia_ino = local->ia_ino;
+ this_call_cnt = dht_frame_return (frame);
+ if (is_last_call (this_call_cnt)) {
+ DHT_STACK_UNWIND (setxattr, frame, local->op_ret,
+ local->op_errno, NULL);
}
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt))
- DHT_STACK_UNWIND (fsync, frame, local->op_ret, local->op_errno,
- prebuf, postbuf);
-
return 0;
}
-
-
-int
-dht_err_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno)
+static void
+fill_layout_info (dht_layout_t *layout, char *buf)
{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- call_frame_t *prev = NULL;
-
-
- local = frame->local;
- prev = cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- gf_log (this->name, GF_LOG_DEBUG,
- "subvolume %s returned -1 (%s)",
- prev->this->name, strerror (op_errno));
- goto unlock;
- }
-
- local->op_ret = 0;
- }
-unlock:
- UNLOCK (&frame->lock);
+ int i = 0;
+ char tmp_buf[128] = {0,};
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- DHT_STACK_UNWIND (setxattr, frame, local->op_ret, local->op_errno);
+ for (i = 0; i < layout->cnt; i++) {
+ snprintf (tmp_buf, 128, "(%s %u %u)",
+ layout->list[i].xlator->name,
+ layout->list[i].start,
+ layout->list[i].stop);
+ if (i)
+ strcat (buf, " ");
+ strcat (buf, tmp_buf);
}
-
- return 0;
}
+void
+dht_fill_pathinfo_xattr (xlator_t *this, dht_local_t *local,
+ char *xattr_buf, int32_t alloc_len,
+ int flag, char *layout_buf)
+{
+ if (flag && local->xattr_val)
+ snprintf (xattr_buf, alloc_len,
+ "((<"DHT_PATHINFO_HEADER"%s> %s) (%s-layout %s))",
+ this->name, local->xattr_val, this->name,
+ layout_buf);
+ else if (local->xattr_val)
+ snprintf (xattr_buf, alloc_len,
+ "(<"DHT_PATHINFO_HEADER"%s> %s)",
+ this->name, local->xattr_val);
+ else if (flag)
+ snprintf (xattr_buf, alloc_len, "(%s-layout %s)",
+ this->name, layout_buf);
+}
int
-dht_access (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t mask)
+dht_vgetxattr_alloc_and_fill (dht_local_t *local, dict_t *xattr, xlator_t *this,
+ int op_errno)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
-
+ int ret = -1;
+ char *value = NULL;
+ int32_t plen = 0;
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
-
- subvol = dht_subvol_get_cached (this, loc->inode);
- if (!subvol) {
- gf_log (this->name, GF_LOG_DEBUG,
- "no cached subvolume for path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
+ ret = dict_get_str (xattr, local->xsel, &value);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "Subvolume %s returned -1 (%s)", this->name,
+ strerror (op_errno));
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ goto out;
+ }
- local = dht_local_init (frame);
- if (!local) {
- op_errno = ENOMEM;
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- goto err;
- }
+ local->alloc_len += strlen(value);
- local->call_cnt = 1;
+ if (!local->xattr_val) {
+ local->alloc_len += (strlen (DHT_PATHINFO_HEADER) + 10);
+ local->xattr_val = GF_CALLOC (local->alloc_len, sizeof (char),
+ gf_common_mt_char);
+ if (!local->xattr_val) {
+ ret = -1;
+ goto out;
+ }
+ }
- STACK_WIND (frame, dht_err_cbk,
- subvol, subvol->fops->access,
- loc, mask);
+ if (local->xattr_val) {
+ plen = strlen (local->xattr_val);
+ if (plen) {
+ /* extra byte(s) for \0 to be safe */
+ local->alloc_len += (plen + 2);
+ local->xattr_val = GF_REALLOC (local->xattr_val,
+ local->alloc_len);
+ if (!local->xattr_val) {
+ ret = -1;
+ goto out;
+ }
+ }
- return 0;
+ (void) strcat (local->xattr_val, value);
+ (void) strcat (local->xattr_val, " ");
+ local->op_ret = 0;
+ }
-err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (access, frame, -1, op_errno);
+ ret = 0;
- return 0;
+ out:
+ return ret;
}
-
int
-dht_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, const char *path, struct iatt *sbuf)
+dht_vgetxattr_fill_and_set (dht_local_t *local, dict_t **dict, xlator_t *this,
+ gf_boolean_t flag)
{
- dht_local_t *local = NULL;
+ int ret = -1;
+ char *xattr_buf = NULL;
+ char layout_buf[8192] = {0,};
- local = frame->local;
- if (op_ret == -1)
- goto err;
+ if (flag)
+ fill_layout_info (local->layout, layout_buf);
+
+ *dict = dict_new ();
+ if (!*dict)
+ goto out;
+
+ local->xattr_val[strlen (local->xattr_val) - 1] = '\0';
+
+ /* we would need max this many bytes to create xattr string
+ * extra 40 bytes is just an estimated amount of additional
+ * space required as we include translator name and some
+ * spaces, brackets etc. when forming the pathinfo string.
+ *
+ * For node-uuid we just don't have all the pretty formatting,
+ * but since this is a generic routine for pathinfo & node-uuid
+ * we dont have conditional space allocation and try to be
+ * generic
+ */
+ local->alloc_len += (2 * strlen (this->name))
+ + strlen (layout_buf)
+ + 40;
+ xattr_buf = GF_CALLOC (local->alloc_len, sizeof (char),
+ gf_common_mt_char);
+ if (!xattr_buf)
+ goto out;
- if (local) {
- sbuf->ia_ino = local->ia_ino;
+ if (XATTR_IS_PATHINFO (local->xsel)) {
+ (void) dht_fill_pathinfo_xattr (this, local, xattr_buf,
+ local->alloc_len, flag,
+ layout_buf);
+ } else if (XATTR_IS_NODE_UUID (local->xsel)) {
+ (void) snprintf (xattr_buf, local->alloc_len, "%s",
+ local->xattr_val);
} else {
- op_ret = -1;
- op_errno = EINVAL;
+ gf_log (this->name, GF_LOG_WARNING,
+ "Unknown local->xsel (%s)", local->xsel);
+ goto out;
}
-err:
- DHT_STACK_UNWIND (readlink, frame, op_ret, op_errno, path, sbuf);
+ ret = dict_set_dynstr (*dict, local->xsel, xattr_buf);
+ GF_FREE (local->xattr_val);
- return 0;
+ out:
+ return ret;
}
-
int
-dht_readlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc, size_t size)
+dht_vgetxattr_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
+ int ret = 0;
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ dict_t *dict = NULL;
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (frame->local, out);
- subvol = dht_subvol_get_cached (this, loc->inode);
- if (!subvol) {
- gf_log (this->name, GF_LOG_DEBUG,
- "no cached subvolume for path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
+ local = frame->local;
- local = dht_local_init (frame);
- if (!local) {
- op_errno = ENOMEM;
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory");
- goto err;
- }
+ LOCK (&frame->lock);
+ {
+ this_call_cnt = --local->call_cnt;
+ if (op_ret < 0) {
+ if (op_errno != ENOTCONN) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "getxattr err (%s) for dir",
+ strerror (op_errno));
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ }
- local->ia_ino = loc->inode->ino;
-
- STACK_WIND (frame, dht_readlink_cbk,
- subvol, subvol->fops->readlink,
- loc, size);
+ goto unlock;
+ }
- return 0;
+ ret = dht_vgetxattr_alloc_and_fill (local, xattr, this,
+ op_errno);
+ if (ret)
+ gf_log (this->name, GF_LOG_ERROR,
+ "alloc or fill failure");
+ }
+ unlock:
+ UNLOCK (&frame->lock);
-err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (readlink, frame, -1, op_errno, NULL, NULL);
+ if (!is_last_call (this_call_cnt))
+ goto out;
- return 0;
-}
+ /* -- last call: do patch ups -- */
+ if (local->op_ret == -1) {
+ goto unwind;
+ }
-int
-dht_fix_layout_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
-{
- DHT_STACK_UNWIND (getxattr, frame, -1, ENODATA, NULL);
+ ret = dht_vgetxattr_fill_and_set (local, &dict, this, _gf_true);
+ if (ret)
+ goto unwind;
- return 0;
-}
+ DHT_STACK_UNWIND (getxattr, frame, 0, 0, dict, xdata);
+ goto cleanup;
-static void
-fill_layout_info (dht_layout_t *layout, char *buf)
-{
- int i = 0;
- char tmp_buf[128] = {0,};
-
- for (i = 0; i < layout->cnt; i++) {
- snprintf (tmp_buf, 128, "(%s %u %u)",
- layout->list[i].xlator->name,
- layout->list[i].start,
- layout->list[i].stop);
- if (i)
- strcat (buf, " ");
- strcat (buf, tmp_buf);
- }
+ unwind:
+ DHT_STACK_UNWIND (getxattr, frame, -1, local->op_errno, NULL, NULL);
+ cleanup:
+ if (dict)
+ dict_unref (dict);
+ out:
+ return 0;
}
int
-dht_pathinfo_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr)
+dht_vgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int ret = 0;
- int flag = 0;
- int this_call_cnt = 0;
- char *value_got = NULL;
- char layout_buf[8192] = {0,};
- char xattr_buf[8192 + 1024] = {0,};
- dict_t *dict = NULL;
+ dht_local_t *local = NULL;
+ int ret = 0;
+ dict_t *dict = NULL;
+ call_frame_t *prev = NULL;
+ gf_boolean_t flag = _gf_true;
local = frame->local;
+ prev = cookie;
- if (op_ret != -1) {
- ret = dict_get_str (xattr, GF_XATTR_PATHINFO_KEY, &value_got);
- if (!ret) {
- if (!local->pathinfo)
- local->pathinfo = GF_CALLOC (8192, sizeof (char),
- gf_common_mt_char);
- if (local->pathinfo)
- strcat (local->pathinfo, value_got);
- }
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ gf_log (this->name, GF_LOG_ERROR, "Subvolume %s returned -1 "
+ "(%s)", prev->this->name, strerror (op_errno));
+ goto unwind;
}
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- if (local->layout->cnt > 1) {
- /* Set it for directory */
- fill_layout_info (local->layout, layout_buf);
- flag = 1;
- }
-
- dict = dict_new ();
-
- if (flag && local->pathinfo)
- snprintf (xattr_buf, 9216, "((%s %s) (%s-layout %s))",
- this->name, local->pathinfo, this->name,
- layout_buf);
- else if (local->pathinfo)
- snprintf (xattr_buf, 9216, "(%s %s)",
- this->name, local->pathinfo);
- else if (flag)
- snprintf (xattr_buf, 9216, "(%s-layout %s)",
- this->name, layout_buf);
-
- ret = dict_set_str (dict, GF_XATTR_PATHINFO_KEY,
- xattr_buf);
-
- if (local->pathinfo)
- GF_FREE (local->pathinfo);
- GF_FREE (local->key);
-
- DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
+ ret = dht_vgetxattr_alloc_and_fill (local, xattr, this,
+ op_errno);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "alloc or fill failure");
+ goto unwind;
+ }
- if (dict)
- dict_unref (dict);
+ flag = (local->layout->cnt > 1) ? _gf_true : _gf_false;
- return 0;
- }
+ ret = dht_vgetxattr_fill_and_set (local, &dict, this, flag);
+ if (ret)
+ goto unwind;
- if (local->pathinfo)
- strcat (local->pathinfo, " Link: ");
- if (local->hashed_subvol) {
- /* This will happen if there pending */
- STACK_WIND (frame, dht_pathinfo_getxattr_cbk, local->hashed_subvol,
- local->hashed_subvol->fops->getxattr,
- &local->loc, local->key);
+ DHT_STACK_UNWIND (getxattr, frame, 0, 0, dict, xdata);
+ goto cleanup;
- return 0;
- }
+ unwind:
+ DHT_STACK_UNWIND (getxattr, frame, -1, local->op_errno,
+ NULL, NULL);
+ cleanup:
+ if (dict)
+ dict_unref (dict);
- gf_log ("this->name", GF_LOG_ERROR, "Unable to find hashed_subvol for path"
- " %s", local->pathinfo);
-
- DHT_STACK_UNWIND (getxattr, frame, -1, op_errno, dict);
return 0;
}
int
dht_linkinfo_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr)
+ int op_ret, int op_errno, dict_t *xattr,
+ dict_t *xdata)
{
int ret = 0;
char *value = NULL;
@@ -1777,92 +1982,234 @@ dht_linkinfo_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
}
}
- DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr);
+ DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr, xdata);
return 0;
}
int
dht_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr)
+ int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
{
- if (op_ret != -1) {
- if (dict_get (xattr, "trusted.glusterfs.dht")) {
- dict_del (xattr, "trusted.glusterfs.dht");
- }
+ int this_call_cnt = 0;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (frame->local, out);
+ VALIDATE_OR_GOTO (this->private, out);
+
+ conf = this->private;
+ local = frame->local;
+
+ this_call_cnt = dht_frame_return (frame);
+
+ if (!xattr || (op_ret == -1))
+ goto out;
+
+ if (dict_get (xattr, conf->xattr_name)) {
+ dict_del (xattr, conf->xattr_name);
}
+ local->op_ret = 0;
- DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr);
+ if (!local->xattr) {
+ local->xattr = dict_copy_with_ref (xattr, NULL);
+ } else {
+ /* first aggregate everything into xattr and then copy into
+ * local->xattr. This is required as we want to have
+ * 'local->xattr' as the proper final dictionary passed above
+ * distribute xlator.
+ */
+ dht_aggregate_xattr (xattr, local->xattr);
+ local->xattr = dict_copy (xattr, local->xattr);
+ }
+out:
+ if (is_last_call (this_call_cnt)) {
+ DHT_STACK_UNWIND (getxattr, frame, local->op_ret, op_errno,
+ local->xattr, NULL);
+ }
+ return 0;
+}
+int32_t
+dht_getxattr_unwind (call_frame_t *frame,
+ int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
+{
+ DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
return 0;
}
int
+dht_getxattr_get_real_filename_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int op_ret, int op_errno,
+ dict_t *xattr, dict_t *xdata)
+{
+ int this_call_cnt = 0;
+ dht_local_t *local = NULL;
+
+
+ local = frame->local;
+
+ if (op_ret != -1) {
+ if (local->xattr)
+ dict_unref (local->xattr);
+ local->xattr = dict_ref (xattr);
+
+ if (local->xattr_req)
+ dict_unref (local->xattr_req);
+ local->xattr_req = dict_ref (xdata);
+ }
+
+ this_call_cnt = dht_frame_return (frame);
+ if (is_last_call (this_call_cnt)) {
+ DHT_STACK_UNWIND (getxattr, frame, local->op_ret, op_errno,
+ local->xattr, local->xattr_req);
+ }
+
+ return 0;
+}
+
+
+int
+dht_getxattr_get_real_filename (call_frame_t *frame, xlator_t *this,
+ loc_t *loc, const char *key, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ int i = 0;
+ dht_layout_t *layout = NULL;
+ int cnt = 0;
+ xlator_t *subvol = NULL;
+
+
+ local = frame->local;
+ layout = local->layout;
+
+ cnt = local->call_cnt = layout->cnt;
+
+ local->op_ret = -1;
+ local->op_errno = ENODATA;
+
+ for (i = 0; i < cnt; i++) {
+ subvol = layout->list[i].xlator;
+ STACK_WIND (frame, dht_getxattr_get_real_filename_cbk,
+ subvol, subvol->fops->getxattr,
+ loc, key, xdata);
+ }
+
+ return 0;
+}
+
+