diff options
author | Raghavendra Bhat <raghavendra@redhat.com> | 2014-07-03 17:13:38 +0530 |
---|---|---|
committer | Vijay Bellur <vbellur@redhat.com> | 2014-07-16 02:27:50 -0700 |
commit | 1dea949cb60c3814c9206df6ba8dddec8d471a94 (patch) | |
tree | 4685ce644e5d4f813eb4dd9309b7b335e20ffd6c /xlators | |
parent | dcc1696045f12127ff37e6312a04c0024c8a4e24 (diff) |
make snapview-server more compatible with NFS server
* There was no handle based API for listxattr. With this change, glfs_h_getxattrs
also handles the listxattr functionality by checking whether the name is NULL
or not (like posix). But all the gfapi functions for listxattr
(glfs_h_getxattrs AND glfs_listxattr AND glfs_flistxattr) returns the names of
the xattrs in a buffer provided by the caller. But snapview-server has to
return the list of xattrs in a dict itself (similar to posix xlator). But
the buffer just contains the names of the xattrs. So for each xattr, a zero
byte value is set (i.e. "") into the dict and sent back. Translators which
do xattr caching (as of now md-cache which caches selinux and acl related
xattrs) should not cache those xattrs whose value is a zero byte data ("").
So made changes in md-cache to ignore zero byte values.
* NFS server was not linking the inodes to inode table in readdirp. This was
leading to applications getting errors. The below set of operations would
lead to applications getting error
1) ls -l in one of the snaopshots (snapview-server would generate gfids for
each entry on the fly and link the inodes associated with those entries)
2) NFS server upon getting readdirp reply would not link the inodes of the
entries. But it used to generate filehandles for each entry and associate
the gfid of that entry with the filehandle and send it as part of the
reply to nfs client.
3) NFS client would send the filehandle of one of those entries when some
activity is done on it.
4) NFS server would not be able to find the inode for the gfid present in the
filehandle (as the inode was not linked) and would go for hard resolution
by sending a lookup on the gfid by creating a new inode.
5) snapview-client will not able to identify whether the inode is a real inode
existing in the main volume or a virtual inode existing in the snapshots
as there would not be any inode context.
6) Since the gfid upon which lookup is sent is a virtual gfid which is not
present in the disk, lookup would fail and the application would get an
error.
To handle above situation, now nfs server also does inode linking in readdirp.
Change-Id: Ibb191408347b6b5f21cff72319ccee619ea77bcd
BUG: 1115949
Signed-off-by: Raghavendra Bhat <raghavendra@redhat.com>
Reviewed-on: http://review.gluster.org/8230
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Niels de Vos <ndevos@redhat.com>
Reviewed-by: Raghavendra G <rgowdapp@redhat.com>
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
Diffstat (limited to 'xlators')
-rw-r--r-- | xlators/features/snapview-client/src/snapview-client.c | 1 | ||||
-rw-r--r-- | xlators/features/snapview-server/src/snapview-server.c | 282 | ||||
-rw-r--r-- | xlators/nfs/server/src/nfs3.c | 4 | ||||
-rw-r--r-- | xlators/performance/md-cache/src/md-cache.c | 19 |
4 files changed, 228 insertions, 78 deletions
diff --git a/xlators/features/snapview-client/src/snapview-client.c b/xlators/features/snapview-client/src/snapview-client.c index 344a8c9562a..ad022101715 100644 --- a/xlators/features/snapview-client/src/snapview-client.c +++ b/xlators/features/snapview-client/src/snapview-client.c @@ -161,6 +161,7 @@ svc_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this, do_unwind = _gf_false; } } + gf_log (this->name, GF_LOG_WARNING, "Lookup on normal graph failed"); goto out; diff --git a/xlators/features/snapview-server/src/snapview-server.c b/xlators/features/snapview-server/src/snapview-server.c index 47344113406..99965a4ccef 100644 --- a/xlators/features/snapview-server/src/snapview-server.c +++ b/xlators/features/snapview-server/src/snapview-server.c @@ -1627,18 +1627,73 @@ out: return 0; } +/* + * This function adds the xattr keys present in the list (@list) to the dict. + * But the list contains only the names of the xattrs (and no value, as + * the gfapi functions for the listxattr operations would return only the + * names of the xattrs in the buffer provided by the caller, though they had + * got the values of those xattrs from posix) as described in the man page of + * listxattr. But before unwinding snapview-server has to put those names + * back into the dict. But to get the values for those xattrs it has to do the + * getxattr operation on each xattr which might turn out to be a costly + * operation. So for each of the xattrs present in the list, a 0 byte value + * ("") is set into the dict before unwinding. This can be treated as an + * indicator to other xlators which want to cache the xattrs (as of now, + * md-cache which caches acl and selinux related xattrs) to not to cache the + * values of the xattrs present in the dict. + */ +int32_t +svs_add_xattrs_to_dict (xlator_t *this, dict_t *dict, char *list, ssize_t size) +{ + char keybuffer[4096] = {0,}; + size_t remaining_size = 0; + int32_t list_offset = 0; + int32_t ret = -1; + + GF_VALIDATE_OR_GOTO ("snapview-daemon", this, out); + GF_VALIDATE_OR_GOTO (this->name, dict, out); + GF_VALIDATE_OR_GOTO (this->name, list, out); + + remaining_size = size; + list_offset = 0; + while (remaining_size > 0) { + strcpy (keybuffer, list + list_offset); +#ifdef GF_DARWIN_HOST_OS + /* The protocol expect namespace for now */ + char *newkey = NULL; + gf_add_prefix (XATTR_USER_PREFIX, keybuffer, &newkey); + strcpy (keybuffer, newkey); + GF_FREE (newkey); +#endif + ret = dict_set_str (dict, keybuffer, ""); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "dict set operation " + "for the key %s failed.", keybuffer); + goto out; + } + + remaining_size -= strlen (keybuffer) + 1; + list_offset += strlen (keybuffer) + 1; + } /* while (remaining_size > 0) */ + + ret = 0; + +out: + return ret; +} + int32_t svs_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, const char *name, dict_t *xdata) { - svs_inode_t *inode_ctx = NULL; - int32_t op_ret = -1; - int32_t op_errno = EINVAL; - glfs_t *fs = NULL; - glfs_object_t *object = NULL; - char *value = 0; - ssize_t size = 0; - dict_t *dict = NULL; + svs_inode_t *inode_ctx = NULL; + int32_t op_ret = -1; + int32_t op_errno = EINVAL; + glfs_t *fs = NULL; + glfs_object_t *object = NULL; + char *value = 0; + ssize_t size = 0; + dict_t *dict = NULL; GF_VALIDATE_OR_GOTO ("snap-view-daemon", this, out); GF_VALIDATE_OR_GOTO ("snap-view-daemon", frame, out); @@ -1654,8 +1709,11 @@ svs_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, const char *name, goto out; } - /* Fake success is sent if the getxattr is on entry point directory - or the inode is SNAP_VIEW_ENTRY_POINT_INODE + /* EINVAL is sent if the getxattr is on entry point directory + or the inode is SNAP_VIEW_ENTRY_POINT_INODE. Entry point is + a virtual directory on which setxattr operations are not + allowed. If getxattr has to be faked as success, then a value + for the name of the xattr has to be sent which we dont have. */ if (inode_ctx->type == SNAP_VIEW_ENTRY_POINT_INODE) { op_ret = -1; @@ -1664,19 +1722,30 @@ svs_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, const char *name, } else if (inode_ctx->type == SNAP_VIEW_VIRTUAL_INODE) { fs = inode_ctx->fs; object = inode_ctx->object; + dict = dict_new (); + if (!dict) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "allocate dict"); + op_ret = -1; + op_errno = ENOMEM; + goto out; + } + size = glfs_h_getxattrs (fs, object, name, NULL, 0); if (size == -1) { - gf_log (this->name, GF_LOG_ERROR, "getxattr on %s " - "failed (key: %s)", loc->name, name); + gf_log (this->name, GF_LOG_ERROR, "getxattr " + "on %s failed (key: %s)", loc->name, + name); op_ret = -1; op_errno = errno; goto out; } - value = GF_CALLOC (size + 1, sizeof (char), gf_common_mt_char); + value = GF_CALLOC (size + 1, sizeof (char), + gf_common_mt_char); if (!value) { - gf_log (this->name, GF_LOG_ERROR, "failed to allocate " - "memory for getxattr on %s (key: %s)", - loc->name, name); + gf_log (this->name, GF_LOG_ERROR, "failed to " + "allocate memory for getxattr on %s " + "(key: %s)", loc->name, name); op_ret = -1; op_errno = ENOMEM; goto out; @@ -1684,35 +1753,38 @@ svs_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, const char *name, size = glfs_h_getxattrs (fs, object, name, value, size); if (size == -1) { - gf_log (this->name, GF_LOG_ERROR, "failed to get the " - "xattr %s for entry %s", name, loc->name); + gf_log (this->name, GF_LOG_ERROR, "failed to " + "get the xattr %s for entry %s", name, + loc->name); op_ret = -1; op_errno = errno; goto out; } value[size] = '\0'; - dict = dict_new (); - if (!dict) { - gf_log (this->name, GF_LOG_ERROR, "failed to allocate " - "dict"); - op_ret = -1; - op_errno = ENOMEM; - goto out; - } - - op_ret = dict_set_dynptr (dict, (char *)name, value, size); - if (op_ret < 0) { - op_errno = -op_ret; - gf_log (this->name, GF_LOG_ERROR, "dict set operation " - "for %s for the key %s failed.", loc->path, - name); + if (name) { + op_ret = dict_set_dynptr (dict, (char *)name, value, + size); + if (op_ret < 0) { + op_errno = -op_ret; + gf_log (this->name, GF_LOG_ERROR, "dict set " + "operation for %s for the key %s " + "failed.", loc->path, name); + GF_FREE (value); + value = NULL; + goto out; + } + } else { + op_ret = svs_add_xattrs_to_dict (this, dict, value, + size); + if (op_ret == -1) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "add the xattrs from the list to dict"); + op_errno = ENOMEM; + goto out; + } GF_FREE (value); - goto out; } - - op_ret = 0; - op_errno = 0; } out: @@ -1721,6 +1793,9 @@ out: STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, NULL); + if (dict) + dict_unref (dict); + return 0; } @@ -1761,8 +1836,11 @@ svs_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name, } glfd = sfd->fd; - /* Fake success is sent if the getxattr is on entry point directory - or the inode is SNAP_VIEW_ENTRY_POINT_INODE + /* EINVAL is sent if the getxattr is on entry point directory + or the inode is SNAP_VIEW_ENTRY_POINT_INODE. Entry point is + a virtual directory on which setxattr operations are not + allowed. If getxattr has to be faked as success, then a value + for the name of the xattr has to be sent which we dont have. */ if (inode_ctx->type == SNAP_VIEW_ENTRY_POINT_INODE) { op_ret = -1; @@ -1771,53 +1849,98 @@ svs_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name, } if (inode_ctx->type == SNAP_VIEW_VIRTUAL_INODE) { - size = glfs_fgetxattr (glfd, name, NULL, 0); - if (size == -1) { - gf_log (this->name, GF_LOG_ERROR, "getxattr on %s " - "failed (key: %s)", uuid_utoa (fd->inode->gfid), - name); - op_ret = -1; - op_errno = errno; - goto out; - } - value = GF_CALLOC (size + 1, sizeof (char), gf_common_mt_char); - if (!value) { - gf_log (this->name, GF_LOG_ERROR, "failed to allocate " - "memory for getxattr on %s (key: %s)", - uuid_utoa (fd->inode->gfid), name); - op_ret = -1; - op_errno = ENOMEM; - goto out; - } - - size = glfs_fgetxattr (glfd, name, value, size); - if (size == -1) { - gf_log (this->name, GF_LOG_ERROR, "failed to get the " - "xattr %s for inode %s", name, - uuid_utoa (fd->inode->gfid)); - op_ret = -1; - op_errno = errno; - goto out; - } - value[size] = '\0'; - dict = dict_new (); if (!dict) { - gf_log (this->name, GF_LOG_ERROR, "failed to allocate " - "dict"); + gf_log (this->name, GF_LOG_ERROR, "failed to " + "allocate dict"); op_ret = -1; op_errno = ENOMEM; goto out; } - op_ret = dict_set_dynptr (dict, (char *)name, value, size); - if (op_ret < 0) { - op_errno = -op_ret; - gf_log (this->name, GF_LOG_ERROR, "dict set operation " - "for gfid %s for the key %s failed.", - uuid_utoa (fd->inode->gfid), name); + if (name) { + size = glfs_fgetxattr (glfd, name, NULL, 0); + if (size == -1) { + gf_log (this->name, GF_LOG_ERROR, "getxattr on " + "%s failed (key: %s)", + uuid_utoa (fd->inode->gfid), name); + op_ret = -1; + op_errno = errno; + goto out; + } + value = GF_CALLOC (size + 1, sizeof (char), + gf_common_mt_char); + if (!value) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "allocate memory for getxattr on %s " + "(key: %s)", + uuid_utoa (fd->inode->gfid), name); + op_ret = -1; + op_errno = ENOMEM; + goto out; + } + + size = glfs_fgetxattr (glfd, name, value, size); + if (size == -1) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "get the xattr %s for inode %s", name, + uuid_utoa (fd->inode->gfid)); + op_ret = -1; + op_errno = errno; + goto out; + } + value[size] = '\0'; + + op_ret = dict_set_dynptr (dict, (char *)name, value, + size); + if (op_ret < 0) { + op_errno = -op_ret; + gf_log (this->name, GF_LOG_ERROR, "dict set " + "operation for gfid %s for the key %s " + "failed.", + uuid_utoa (fd->inode->gfid), name); + GF_FREE (value); + goto out; + } + } else { + size = glfs_flistxattr (glfd, NULL, 0); + if (size == -1) { + gf_log (this->name, GF_LOG_ERROR, "listxattr " + "on %s failed", + uuid_utoa (fd->inode->gfid)); + goto out; + } + + value = GF_CALLOC (size + 1, sizeof (char), + gf_common_mt_char); + if (!value) { + op_ret = -1; + op_errno = ENOMEM; + gf_log (this->name, GF_LOG_ERROR, "failed to " + "allocate buffer for xattr list (%s)", + uuid_utoa (fd->inode->gfid)); + goto out; + } + + size = glfs_flistxattr (glfd, value, size); + if (size == -1) { + op_ret = -1; + op_errno = errno; + gf_log (this->name, GF_LOG_ERROR, "listxattr " + "on %s failed", + uuid_utoa (fd->inode->gfid)); + goto out; + } + + op_ret = svs_add_xattrs_to_dict (this, dict, value, + size); + if (op_ret == -1) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "add the xattrs from the list to dict"); + op_errno = ENOMEM; + goto out; + } GF_FREE (value); - goto out; } op_ret = 0; @@ -1830,6 +1953,9 @@ out: STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict, NULL); + if (dict) + dict_unref (dict); + return 0; } diff --git a/xlators/nfs/server/src/nfs3.c b/xlators/nfs/server/src/nfs3.c index 6361f9e20aa..7ecfc675b61 100644 --- a/xlators/nfs/server/src/nfs3.c +++ b/xlators/nfs/server/src/nfs3.c @@ -4099,6 +4099,10 @@ nfs3svc_readdir_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, } stat = NFS3_OK; + + /* do inode linking here */ + gf_link_inodes_from_dirent (this, cs->fd->inode, &cs->entries); + nfs3err: if (cs->maxcount == 0) { nfs3_log_readdir_res (rpcsvc_request_xid (cs->req), stat, diff --git a/xlators/performance/md-cache/src/md-cache.c b/xlators/performance/md-cache/src/md-cache.c index 7588463b891..782d258f47c 100644 --- a/xlators/performance/md-cache/src/md-cache.c +++ b/xlators/performance/md-cache/src/md-cache.c @@ -486,6 +486,25 @@ updatefn(dict_t *dict, char *key, data_t *value, void *data) } } + /* posix xlator as part of listxattr will send both names + * and values of the xattrs in the dict. But as per man page + * listxattr is mainly supposed to send names of the all the + * xattrs. gfapi, as of now will put all the keys it obtained + * in the dict (sent by posix) into a buffer provided by the + * caller (thus the values of those xattrs are lost). If some + * xlator makes gfapi based calls (ex: snapview-server), then + * it has to unwind the calls by putting those names it got + * in the buffer again into the dict. But now it would not be + * having the values for those xattrs. So it might just put + * a 0 byte value ("") into the dict for each xattr and unwind + * the call. So the xlators which cache the xattrs (as of now + * md-cache caches the acl and selinux related xattrs), should + * not update their cache if the value of a xattr is a 0 byte + * data (i.e. ""). + */ + if (!strcmp (value->data, "")) + continue; + if (dict_set(u->dict, key, value) < 0) { u->ret = -1; return -1; |