summaryrefslogtreecommitdiffstats
path: root/xlators/features/locks/src/locks.h
blob: 0ab2aa6cbaec8e6f5dfc22c4bfb49de470d2473f (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
/*
   Copyright (c) 2006-2012, 2015-2016 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 __POSIX_LOCKS_H__
#define __POSIX_LOCKS_H__

#include <glusterfs/compat-errno.h>
#include <glusterfs/stack.h>
#include <glusterfs/call-stub.h>
#include "locks-mem-types.h"
#include <glusterfs/client_t.h>

#include <glusterfs/lkowner.h>

typedef enum {
    MLK_NONE,
    MLK_FILE_BASED,
    MLK_FORCED,
    MLK_OPTIMAL
} mlk_mode_t; /* defines different mandatory locking modes*/

struct __pl_fd;

struct __posix_lock {
    struct list_head list;

    off_t fl_start;
    off_t fl_end;
    uint32_t lk_flags;

    short fl_type;
    short blocked;              /* waiting to acquire */
    struct gf_flock user_flock; /* the flock supplied by the user */
    xlator_t *this;             /* required for blocked locks */
    unsigned long fd_num;

    fd_t *fd;
    call_frame_t *frame;

    struct timeval blkd_time; /*time at which lock was queued into blkd list*/
    struct timeval
        granted_time; /*time at which lock was queued into active list*/

    /* These two together serve to uniquely identify each process
       across nodes */

    void *client; /* to identify client node */

    /* This field uniquely identifies the client the lock belongs to.  As
     * lock migration is handled by rebalance, the client_t object will be
     * overwritten by rebalance and can't be deemed as the owner of the
     * lock on destination. Hence, the below field is migrated from
     * source to destination by lock_migration_info_t and updated on the
     * destination. So that on client-server disconnection, server can
     * cleanup the locks proper;y.  */

    char *client_uid;
    gf_lkowner_t owner;
    pid_t client_pid; /* pid of client process */

    int blocking;
};
typedef struct __posix_lock posix_lock_t;

struct __pl_inode_lock {
    struct list_head list;
    struct list_head blocked_locks; /* list_head pointing to blocked_inodelks */
    struct list_head contend;       /* list of contending locks */
    int ref;

    off_t fl_start;
    off_t fl_end;

    const char *volume;

    struct gf_flock user_flock; /* the flock supplied by the user */
    xlator_t *this;             /* required for blocked locks */
    struct __pl_inode *pl_inode;

    call_frame_t *frame;

    struct timeval blkd_time; /*time at which lock was queued into blkd list*/
    struct timeval
        granted_time; /*time at which lock was queued into active list*/
    /*last time at which lock contention was detected and notified*/
    struct timespec contention_time;

    /* These two together serve to uniquely identify each process
       across nodes */

    void *client; /* to identify client node */
    gf_lkowner_t owner;
    pid_t client_pid; /* pid of client process */

    char *connection_id; /* stores the client connection id */

    struct list_head client_list; /* list of all locks from a client */
    short fl_type;
};
typedef struct __pl_inode_lock pl_inode_lock_t;

struct _pl_rw_req {
    struct list_head list;
    call_stub_t *stub;
    posix_lock_t region;
};
typedef struct _pl_rw_req pl_rw_req_t;

struct _pl_dom_list {
    struct list_head inode_list; /* list_head back to pl_inode_t */
    const char *domain;
    struct list_head entrylk_list;     /* List of entry locks */
    struct list_head blocked_entrylks; /* List of all blocked entrylks */
    struct list_head inodelk_list;     /* List of inode locks */
    struct list_head blocked_inodelks; /* List of all blocked inodelks */
};
typedef struct _pl_dom_list pl_dom_list_t;

struct __entry_lock {
    struct list_head domain_list;   /* list_head back to pl_dom_list_t */
    struct list_head blocked_locks; /* list_head back to blocked_entrylks */
    struct list_head contend;       /* list of contending locks */
    int ref;

    call_frame_t *frame;
    xlator_t *this;
    struct __pl_inode *pinode;

    const char *volume;

    const char *basename;

    struct timeval blkd_time; /*time at which lock was queued into blkd list*/
    struct timeval
        granted_time; /*time at which lock was queued into active list*/
    /*last time at which lock contention was detected and notified*/
    struct timespec contention_time;

    void *client;
    gf_lkowner_t owner;
    pid_t client_pid; /* pid of client process */

    char *connection_id; /* stores the client connection id */

    struct list_head client_list; /* list of all locks from a client */
    entrylk_type type;
};
typedef struct __entry_lock pl_entry_lock_t;

/* The "simulated" inode. This contains a list of all the locks associated
   with this file */

struct __pl_inode {
    pthread_mutex_t mutex;

    struct list_head dom_list;           /* list of domains */
    struct list_head ext_list;           /* list of fcntl locks */
    struct list_head rw_list;            /* list of waiting r/w requests */
    struct list_head reservelk_list;     /* list of reservelks */
    struct list_head blocked_reservelks; /* list of blocked reservelks */
    struct list_head
        blocked_calls; /* List of blocked lock calls while a reserve is held*/
    struct list_head metalk_list; /* Meta lock list */
                                  /* This is to store the incoming lock
                                     requests while meta lock is enabled */
    struct list_head queued_locks;
    int mandatory; /* if mandatory locking is enabled */

    inode_t *refkeeper; /* hold refs on an inode while locks are
                           held to prevent pruning */
    uuid_t gfid;        /* placeholder for gfid of the inode */
    inode_t *inode;     /* pointer to be used for ref and unref
                           of inode_t as long as there are
                           locks on it */
    gf_boolean_t migrated;

    /* Flag to indicate whether to read mlock-enforce xattr from disk */
    gf_boolean_t check_mlock_info;

    /* Mandatory_lock enforce: IO will be allowed if and only if the lkowner has
       held the lock.

       Note: An xattr is set on the file to recover this information post
       reboot. If client does not want mandatory lock to be enforced, then it
       should remove this xattr explicitly
    */
    gf_boolean_t mlock_enforced;
    /* There are scenarios where mandatory lock is granted but there are IOs
       pending at posix level. To avoid this before preempting the previous lock
       owner, we wait for all the fops to be unwound.
    */
    int fop_wind_count;
    pthread_cond_t check_fop_wind_count;
    gf_boolean_t track_fop_wind_count;
};
typedef struct __pl_inode pl_inode_t;

struct __pl_metalk {
    pthread_mutex_t mutex;
    /* For pl_inode meta lock list */
    struct list_head list;
    /* For pl_ctx_t list */
    struct list_head client_list;
    char *client_uid;

    pl_inode_t *pl_inode;
    int ref;
};
typedef struct __pl_metalk pl_meta_lock_t;

typedef struct {
    char *brickname;
    uint32_t revocation_secs;
    uint32_t revocation_max_blocked;
    uint32_t notify_contention_delay;
    mlk_mode_t mandatory_mode; /* holds current mandatory locking mode */
    gf_boolean_t trace;        /* trace lock requests in and out */
    gf_boolean_t monkey_unlocking;
    gf_boolean_t revocation_clear_all;
    gf_boolean_t notify_contention;
    gf_boolean_t mlock_enforced;
} posix_locks_private_t;

typedef struct {
    data_t *inodelk_dom_count_req;

    dict_t *xdata;
    loc_t loc[2];
    fd_t *fd;
    inode_t *inode;
    off_t offset;
    glusterfs_fop_t op;
    gf_boolean_t entrylk_count_req;
    gf_boolean_t inodelk_count_req;
    gf_boolean_t posixlk_count_req;
    gf_boolean_t parent_entrylk_req;
    int update_mlock_enforced_flag;
} pl_local_t;

typedef struct {
    struct list_head locks_list;
} pl_fdctx_t;

struct _locker {
    struct list_head lockers;
    char *volume;
    inode_t *inode;
    gf_lkowner_t owner;
};

typedef struct _locks_ctx {
    pthread_mutex_t lock;
    struct list_head inodelk_lockers;
    struct list_head entrylk_lockers;
    struct list_head metalk_list;
} pl_ctx_t;

typedef enum { DECREMENT, INCREMENT } pl_count_op_t;

pl_ctx_t *
pl_ctx_get(client_t *client, xlator_t *xlator);

int
pl_inodelk_client_cleanup(xlator_t *this, pl_ctx_t *ctx);

int
pl_entrylk_client_cleanup(xlator_t *this, pl_ctx_t *ctx);

#endif /* __POSIX_LOCKS_H__ */