telldir()/seekdir() portability fixes

POSIX says that an offset obtained from telldir() can only be used
on the same DIR *. Linux is abls to reuse the offset accross
closedir()/opendir() for a given directory, but this is not portable
and such a behavior should be fixed.

An incomplete fix for the posix xlator was merged in
http://review.gluster.org/8933
This change set completes it.

- Perform the same fix index xlator.
- Use appropriate casts and variable types so that 32 bit signed
  offsets obtained by telldir() do not get clobbered when copied into
  64 bit signed types.
- modify afr-self-heald.c so that it does not use anonymous fd,
  since this will cause closedir()/opendir() between each
  syncop_readdir(). On failure we fallback to anonymous fs
  only for Linux so that we can cope with updated client vs not
  updated brick.
- Avoid sending an EINVAL when the client request for the EOF offset.
  Here we fix an error in previous fix for posix xlator: since we
  fill each directory entry with the offset of the next entry, we
  must consider as EOF the offset of the last entry, and not the
  value of telldir() after we read it.

This is a backport of I59fb7f06a872c4f98987105792d648141c258c6a

BUG: 1138897
Change-Id: I1e9f3e4a7d780b98adf6d9f197ee2198d43ef94d
Signed-off-by: Emmanuel Dreyfus <manu@netbsd.org>
Reviewed-on: http://review.gluster.org/9084
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Raghavendra Bhat <raghavendra@redhat.com>

4 files changed, 87 insertions, 28 deletions

diff --git a/xlators/cluster/afr/src/afr-self-heald.c b/xlators/cluster/afr/src/afr-self-heald.c
index 10e19964ed9..6ce52aad1c4 100644
--- a/xlators/cluster/afr/src/afr-self-heald.c
+++ b/xlators/cluster/afr/src/afr-self-heald.c
@@ -276,6 +276,8 @@ afr_shd_index_opendir (xlator_t *this, int child)
 #ifdef GF_LINUX_HOST_OS
 		fd_unref (fd);
 		fd = fd_anonymous (inode);
+		if (!fd)
+		        goto out;
 #else /* GF_LINUX_HOST_OS */
 		gf_log(this->name, GF_LOG_ERROR,
 		       "opendir of %s for %s failed: %s",
@@ -429,7 +431,7 @@ afr_shd_index_sweep (struct subvol_healer *healer)
 	fd_t *fd = NULL;
 	xlator_t *subvol = NULL;
 	afr_private_t *priv = NULL;
-	off_t offset = 0;
+	uint64_t offset = 0;
 	gf_dirent_t entries;
 	gf_dirent_t *entry = NULL;
 	uuid_t gfid;
@@ -503,11 +505,12 @@ afr_shd_index_sweep (struct subvol_healer *healer)
 int
 afr_shd_full_sweep (struct subvol_healer *healer, inode_t *inode)
 {
+	loc_t loc = {0, };
 	fd_t *fd = NULL;
 	xlator_t *this = NULL;
 	xlator_t *subvol = NULL;
 	afr_private_t *priv = NULL;
-	off_t offset = 0;
+	uint64_t offset = 0;
 	gf_dirent_t entries;
 	gf_dirent_t *entry = NULL;
 	int ret = 0;
@@ -516,9 +519,38 @@ afr_shd_full_sweep (struct subvol_healer *healer, inode_t *inode)
 	priv = this->private;
 	subvol = priv->children[healer->subvol];
 
-	fd = fd_anonymous (inode);
-	if (!fd)
-		return -errno;
+	uuid_copy (loc.gfid, inode->gfid);
+	loc.inode = inode_ref(inode);
+
+	fd = fd_create (inode, GF_CLIENT_PID_AFR_SELF_HEALD);
+	if (!fd) {
+		gf_log(this->name, GF_LOG_ERROR,
+		       "fd_create of %s failed: %s",
+		       uuid_utoa (loc.gfid), strerror(errno));
+		ret = -errno;
+		goto out;
+	}
+
+	ret = syncop_opendir (subvol, &loc, fd);
+	if (ret) {
+#ifdef GF_LINUX_HOST_OS /* See comment in afr_shd_index_opendir() */
+		fd_unref(fd);
+		fd = fd_anonymous (inode);
+		if (!fd) {
+			gf_log(this->name, GF_LOG_ERROR,
+			       "fd_anonymous of %s failed: %s",
+			       uuid_utoa (loc.gfid), strerror(errno));
+			ret = -errno;
+			goto out;
+		}
+#else /* GF_LINUX_HOST_OS */
+		gf_log(this->name, GF_LOG_ERROR,
+		       "opendir of %s failed: %s",
+		       uuid_utoa (loc.gfid), strerror(errno));
+		ret = -errno;
+		goto out;
+#endif /* GF_LINUX_HOST_OS */
+	}
 
 	INIT_LIST_HEAD (&entries.list);
 
@@ -560,6 +592,8 @@ afr_shd_full_sweep (struct subvol_healer *healer, inode_t *inode)
 			break;
 	}
 
+out:
+	loc_wipe (&loc);
 	if (fd)
 		fd_unref (fd);
 	return ret;
@@ -949,7 +983,7 @@ afr_shd_gather_index_entries (xlator_t *this, int child, dict_t *output)
 	fd_t *fd = NULL;
 	xlator_t *subvol = NULL;
 	afr_private_t *priv = NULL;
-	off_t offset = 0;
+	uint64_t offset = 0;
 	gf_dirent_t entries;
 	gf_dirent_t *entry = NULL;
 	uuid_t gfid;
diff --git a/xlators/features/index/src/index.c b/xlators/features/index/src/index.c
index c342da8274c..50e2e0ab54d 100644
--- a/xlators/features/index/src/index.c
+++ b/xlators/features/index/src/index.c
@@ -268,10 +268,11 @@ check_delete_stale_index_file (xlator_t *this, char *filename)
 }
 
 static int
-index_fill_readdir (fd_t *fd, DIR *dir, off_t off,
+index_fill_readdir (fd_t *fd, index_fd_ctx_t *fctx, DIR *dir, off_t off,
                     size_t size, gf_dirent_t *entries)
 {
         off_t     in_case = -1;
+        off_t     last_off = 0;
         size_t    filled = 0;
         int       count = 0;
         char      entrybuf[sizeof(struct dirent) + 256 + 8];
@@ -286,11 +287,11 @@ index_fill_readdir (fd_t *fd, DIR *dir, off_t off,
         } else {
                 seekdir (dir, off);
 #ifndef GF_LINUX_HOST_OS
-                if (telldir(dir) != off) {
+                if ((u_long)telldir(dir) != off && off != fctx->dir_eof) {
                         gf_log (THIS->name, GF_LOG_ERROR,
-                                "seekdir(%ld) failed on dir=%p: "
+                                "seekdir(0x%llx) failed on dir=%p: "
 				"Invalid argument (offset reused from "
-				"another DIR * structure?)", (long)off, dir);
+				"another DIR * structure?)", off, dir);
                         errno = EINVAL;
                         count = -1;
                         goto out;
@@ -299,7 +300,7 @@ index_fill_readdir (fd_t *fd, DIR *dir, off_t off,
         }
 
         while (filled <= size) {
-                in_case = telldir (dir);
+                in_case = (u_long)telldir (dir);
 
                 if (in_case == -1) {
                         gf_log (THIS->name, GF_LOG_ERROR,
@@ -335,12 +336,13 @@ index_fill_readdir (fd_t *fd, DIR *dir, off_t off,
                 if (this_size + filled > size) {
                         seekdir (dir, in_case);
 #ifndef GF_LINUX_HOST_OS
-                        if (telldir(dir) != in_case) {
+                        if ((u_long)telldir(dir) != in_case &&
+                            in_case != fctx->dir_eof) {
 				gf_log (THIS->name, GF_LOG_ERROR,
-					"seekdir(%ld) failed on dir=%p: "
+					"seekdir(0x%llx) failed on dir=%p: "
 					"Invalid argument (offset reused from "
 					"another DIR * structure?)",
-					(long)in_case, dir);
+					in_case, dir);
 				errno = EINVAL;
 				count = -1;
 				goto out;
@@ -357,7 +359,14 @@ index_fill_readdir (fd_t *fd, DIR *dir, off_t off,
                                 entry->d_name, strerror (errno));
                         goto out;
                 }
-                this_entry->d_off = telldir (dir);
+                /*
+                 * we store the offset of next entry here, which is
+                 * probably not intended, but code using syncop_readdir()
+                 * (glfs-heal.c, afr-self-heald.c, pump.c) rely on it
+                 * for directory read resumption.
+                 */
+                last_off = (u_long)telldir(dir);
+                this_entry->d_off = last_off;
                 this_entry->d_ino = entry->d_ino;
 
                 list_add_tail (&this_entry->list, &entries->list);
@@ -366,9 +375,12 @@ index_fill_readdir (fd_t *fd, DIR *dir, off_t off,
                 count ++;
         }
 
-        if ((!readdir (dir) && (errno == 0)))
+        if ((!readdir (dir) && (errno == 0))) {
                 /* Indicate EOF */
                 errno = ENOENT;
+                /* Remember EOF offset for later detection */
+                fctx->dir_eof = last_off;
+        }
 out:
         return count;
 }
@@ -581,6 +593,7 @@ __index_fd_ctx_get (fd_t *fd, xlator_t *this, index_fd_ctx_t **ctx)
                 fctx = NULL;
                 goto out;
         }
+        fctx->dir_eof = -1;
 
         ret = __fd_ctx_set (fd, this, (uint64_t)(long)fctx);
         if (ret) {
@@ -949,7 +962,7 @@ index_readdir_wrapper (call_frame_t *frame, xlator_t *this,
                 goto done;
         }
 
-        count = index_fill_readdir (fd, dir, off, size, &entries);
+        count = index_fill_readdir (fd, fctx, dir, off, size, &entries);
 
         /* pick ENOENT to indicate EOF */
         op_errno = errno;
@@ -1270,8 +1283,11 @@ index_releasedir (xlator_t *this, fd_t *fd)
                 goto out;
 
         fctx = (index_fd_ctx_t*) (long) ctx;
-        if (fctx->dir)
-                closedir (fctx->dir);
+        if (fctx->dir) {
+                ret = closedir (fctx->dir);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR, "closedir error: %s", strerror (errno));
+        }
 
         GF_FREE (fctx);
 out:
diff --git a/xlators/features/index/src/index.h b/xlators/features/index/src/index.h
index 661dcdbc417..206d280e519 100644
--- a/xlators/features/index/src/index.h
+++ b/xlators/features/index/src/index.h
@@ -34,6 +34,7 @@ typedef struct index_inode_ctx {
 
 typedef struct index_fd_ctx {
         DIR *dir;
+        off_t dir_eof;
 } index_fd_ctx_t;
 
 typedef struct index_priv {
diff --git a/xlators/storage/posix/src/posix.c b/xlators/storage/posix/src/posix.c
index d8599881fb7..c1730e8d337 100644
--- a/xlators/storage/posix/src/posix.c
+++ b/xlators/storage/posix/src/posix.c
@@ -4853,6 +4853,7 @@ posix_fill_readdir (fd_t *fd, DIR *dir, off_t off, size_t size,
                     gf_dirent_t *entries, xlator_t *this, int32_t skip_dirs)
 {
         off_t     in_case = -1;
+        off_t     last_off = 0;
         size_t    filled = 0;
         int             count = 0;
         char      entrybuf[sizeof(struct dirent) + 256 + 8];
@@ -4888,11 +4889,11 @@ posix_fill_readdir (fd_t *fd, DIR *dir, off_t off, size_t size,
         } else {
                 seekdir (dir, off);
 #ifndef GF_LINUX_HOST_OS
-                if (telldir(dir) != (long)off && off != pfd->dir_eof) {
+                if ((u_long)telldir(dir) != off && off != pfd->dir_eof) {
                         gf_log (THIS->name, GF_LOG_ERROR,
-                                "seekdir(%ld) failed on dir=%p: "
+                                "seekdir(0x%llx) failed on dir=%p: "
                                 "Invalid argument (offset reused from "
-                                "another DIR * structure?)", (long)off, dir);
+                                "another DIR * structure?)", off, dir);
                         errno = EINVAL;
                         count = -1;
                         goto out;
@@ -4901,7 +4902,7 @@ posix_fill_readdir (fd_t *fd, DIR *dir, off_t off, size_t size,
         }
 
         while (filled <= size) {
-                in_case = telldir (dir);
+                in_case = (u_long)telldir (dir);
 
                 if (in_case == -1) {
                         gf_log (THIS->name, GF_LOG_ERROR,
@@ -4961,13 +4962,13 @@ posix_fill_readdir (fd_t *fd, DIR *dir, off_t off, size_t size,
                 if (this_size + filled > size) {
                         seekdir (dir, in_case);
 #ifndef GF_LINUX_HOST_OS
-                        if (telldir(dir) != (long)in_case &&
+                        if ((u_long)telldir(dir) != in_case &&
                             in_case != pfd->dir_eof) {
                                 gf_log (THIS->name, GF_LOG_ERROR,
-                                        "seekdir(%ld) failed on dir=%p: "
+                                        "seekdir(0x%llx) failed on dir=%p: "
                                         "Invalid argument (offset reused from "
                                         "another DIR * structure?)",
-                                        (long)in_case, dir);
+                                        in_case, dir);
                                 errno = EINVAL;
                                 count = -1;
                                 goto out;
@@ -4984,7 +4985,14 @@ posix_fill_readdir (fd_t *fd, DIR *dir, off_t off, size_t size,
                                 entry->d_name, strerror (errno));
                         goto out;
                 }
-                this_entry->d_off = telldir (dir);
+                /*
+                 * we store the offset of next entry here, which is
+                 * probably not intended, but code using syncop_readdir()
+                 * (glfs-heal.c, afr-self-heald.c, pump.c) rely on it
+                 * for directory read resumption.
+                 */
+                last_off = (u_long)telldir(dir);
+                this_entry->d_off = last_off;
                 this_entry->d_ino = entry->d_ino;
                 this_entry->d_type = entry->d_type;
 
@@ -4998,7 +5006,7 @@ posix_fill_readdir (fd_t *fd, DIR *dir, off_t off, size_t size,
                 /* Indicate EOF */
                 errno = ENOENT;
                 /* Remember EOF offset for later detection */
-                pfd->dir_eof = telldir (dir);
+                pfd->dir_eof = (u_long)last_off;
         }
 out:
         return count;