nfs: do lookup on getattr after brick-status change

By doing a lookup, we get a chance to do all of the self-heal checks that
would occur if we were using native protocol, and return proper status if
the self-heal fails.  Best of all, we don't need to misrepresent times.

Change-Id: I76477d1e5fce4d83e4029e02fcdd71e81e23110d
BUG: 830134
Signed-off-by: Jeff Darcy <jdarcy@redhat.com>
Reviewed-on: http://review.gluster.org/4058
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Vijay Bellur <vbellur@redhat.com>

8 files changed, 171 insertions, 40 deletions

diff --git a/xlators/nfs/server/src/nfs-common.c b/xlators/nfs/server/src/nfs-common.c
index d35df87bc3b..67509b54647 100644
--- a/xlators/nfs/server/src/nfs-common.c
+++ b/xlators/nfs/server/src/nfs-common.c
@@ -432,3 +432,38 @@ nfs_hash_gfid (uuid_t gfid)
 }
 
 
+void
+nfs_fix_generation (xlator_t *this, inode_t *inode)
+{
+        uint64_t                 raw_ctx        = 0;
+        struct nfs_inode_ctx    *ictx           = NULL;
+        struct nfs_state        *priv           = NULL;
+        int                      ret            = -1;
+
+        if (!inode) {
+                return;
+        }
+        priv = this->private;
+
+        if (inode_ctx_get(inode,this,&raw_ctx) == 0) {
+                ictx = (struct nfs_inode_ctx *)raw_ctx;
+                ictx->generation = priv->generation;
+        }
+        else {
+                ictx = GF_CALLOC (1, sizeof (struct nfs_inode_ctx),
+                                  gf_nfs_mt_inode_ctx);
+                if (!ictx) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "could not allocate nfs inode ctx");
+                        return;
+                }
+                INIT_LIST_HEAD(&ictx->shares);
+                ictx->generation = priv->generation;
+                ret = inode_ctx_put (inode, this, (uint64_t)ictx);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "could not store nfs inode ctx");
+                        return;
+                }
+        }
+}
diff --git a/xlators/nfs/server/src/nfs-common.h b/xlators/nfs/server/src/nfs-common.h
index 88fc1496176..f74bb3187cd 100644
--- a/xlators/nfs/server/src/nfs-common.h
+++ b/xlators/nfs/server/src/nfs-common.h
@@ -84,4 +84,7 @@ nfs_hash_gfid (uuid_t gfid);
 
 extern int
 nfs_gfid_loc_fill (inode_table_t *itable, uuid_t gfid, loc_t *loc, int how);
+
+void
+nfs_fix_generation (xlator_t *this, inode_t *inode);
 #endif
diff --git a/xlators/nfs/server/src/nfs-fops.c b/xlators/nfs/server/src/nfs-fops.c
index e2eedf43398..5a2cf2862e4 100644
--- a/xlators/nfs/server/src/nfs-fops.c
+++ b/xlators/nfs/server/src/nfs-fops.c
@@ -353,9 +353,6 @@ out:
                 }                                                       \
         } while (0)                                                     \
 
-
-
-
 /* Fops Layer Explained
  * The fops layer has three types of functions. They can all be identified by
  * their names. Here are the three patterns:
@@ -388,6 +385,23 @@ nfs_fop_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 {
         struct nfs_fop_local    *local = NULL;
         fop_lookup_cbk_t        progcbk;
+        char                    *sh_fail_val = NULL;
+
+        /*
+         * With native protocol, self-heal failures would be detected during
+         * open.  NFS doesn't issue that open when revalidating cache, so we
+         * have to check for failures here instead.
+         */
+        if (dict_get_str(xattr,"sh-failed",&sh_fail_val) == 0) {
+                if (strcmp(sh_fail_val,"1") == 0) {
+                        op_ret = -1;
+                        op_errno = EIO;
+                }
+        }
+
+        if (op_ret == 0) {
+                nfs_fix_generation(this,inode);
+        }
 
         nfl_to_prog_data (local, progcbk, frame);
         nfs_fop_restore_root_ino (local, op_ret, buf, NULL, NULL, postparent);
@@ -759,6 +773,10 @@ nfs_fop_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         struct nfs_fop_local    *nfl = NULL;
         fop_create_cbk_t        progcbk = NULL;
 
+        if (op_ret == 0) {
+                nfs_fix_generation(this,inode);
+        }
+
         nfl_to_prog_data (nfl, progcbk, frame);
         nfs_fop_restore_root_ino (nfl, op_ret, buf, NULL, preparent,
                                   postparent);
@@ -861,6 +879,10 @@ nfs_fop_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         struct nfs_fop_local    *nfl = NULL;
         fop_mkdir_cbk_t         progcbk = NULL;
 
+        if (op_ret == 0) {
+                nfs_fix_generation(this,inode);
+        }
+
         nfl_to_prog_data (nfl, progcbk, frame);
         nfs_fop_restore_root_ino (nfl, op_ret, buf, NULL,preparent, postparent);
         if (progcbk)
@@ -910,6 +932,10 @@ nfs_fop_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         struct nfs_fop_local    *nfl = NULL;
         fop_symlink_cbk_t       progcbk = NULL;
 
+        if (op_ret == 0) {
+                nfs_fix_generation(this,inode);
+        }
+
         nfl_to_prog_data (nfl, progcbk, frame);
         nfs_fop_restore_root_ino (nfl, op_ret,buf, NULL, preparent, postparent);
         if (progcbk)
@@ -1006,6 +1032,10 @@ nfs_fop_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         struct nfs_fop_local    *nfl = NULL;
         fop_mknod_cbk_t         progcbk = NULL;
 
+        if (op_ret == 0) {
+                nfs_fix_generation(this,inode);
+        }
+
         nfl_to_prog_data (nfl, progcbk, frame);
         nfs_fop_restore_root_ino (nfl, op_ret,buf, NULL, preparent, postparent);
         if (progcbk)
@@ -1153,6 +1183,10 @@ nfs_fop_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         struct nfs_fop_local    *nfl = NULL;
         fop_link_cbk_t          progcbk = NULL;
 
+        if (op_ret == 0) {
+                nfs_fix_generation(this,inode);
+        }
+
         nfl_to_prog_data (nfl, progcbk, frame);
         nfs_fop_restore_root_ino (nfl, op_ret, buf, NULL, preparent,
                                   postparent);
diff --git a/xlators/nfs/server/src/nfs-mem-types.h b/xlators/nfs/server/src/nfs-mem-types.h
index de25b08a826..005598c1f96 100644
--- a/xlators/nfs/server/src/nfs-mem-types.h
+++ b/xlators/nfs/server/src/nfs-mem-types.h
@@ -53,6 +53,7 @@ enum gf_nfs_mem_types_ {
         gf_nfs_mt_nlm4_nlmclnt,
         gf_nfs_mt_nlm4_share,
         gf_nfs_mt_aux_gids,
+        gf_nfs_mt_inode_ctx,
         gf_nfs_mt_end
 };
 #endif
diff --git a/xlators/nfs/server/src/nfs.c b/xlators/nfs/server/src/nfs.c
index 0a5a9d1e34a..033d95213df 100644
--- a/xlators/nfs/server/src/nfs.c
+++ b/xlators/nfs/server/src/nfs.c
@@ -759,6 +759,7 @@ nfs_init_state (xlator_t *this)
 
         this->private = (void *)nfs;
         INIT_LIST_HEAD (&nfs->versions);
+        nfs->generation = 1965;
 
         ret = 0;
 
@@ -843,24 +844,26 @@ int
 notify (xlator_t *this, int32_t event, void *data, ...)
 {
         xlator_t                *subvol = NULL;
+        struct nfs_state        *priv   = NULL;
 
         subvol = (xlator_t *)data;
 
         gf_log (GF_NFS, GF_LOG_TRACE, "Notification received: %d",
                 event);
-        switch (event)
-        {
-                case GF_EVENT_CHILD_UP:
-                {
-                        nfs_startup_subvolume (this, subvol);
-                        break;
-                }
 
-                case GF_EVENT_PARENT_UP:
-                {
-                        default_notify (this, GF_EVENT_PARENT_UP, data);
-                        break;
-                }
+        switch (event) {
+        case GF_EVENT_CHILD_UP:
+                nfs_startup_subvolume (this, subvol);
+                break;
+
+        case GF_EVENT_CHILD_MODIFIED:
+                priv = this->private;
+                ++(priv->generation);
+                break;
+
+        case GF_EVENT_PARENT_UP:
+                default_notify (this, GF_EVENT_PARENT_UP, data);
+                break;
         }
 
         return 0;
@@ -882,14 +885,14 @@ fini (xlator_t *this)
 int32_t
 nfs_forget (xlator_t *this, inode_t *inode)
 {
-        uint64_t           ctx        = 0;
-        struct list_head  *head       = NULL;
+        uint64_t                 ctx    = 0;
+        struct nfs_inode_ctx    *ictx   = NULL;
 
         if (inode_ctx_del (inode, this, &ctx))
                 return -1;
 
-        head = (struct list_head *)ctx;
-        GF_FREE (head);
+        ictx = (struct nfs_inode_ctx *)ctx;
+        GF_FREE (ictx);
 
         return 0;
 }
diff --git a/xlators/nfs/server/src/nfs.h b/xlators/nfs/server/src/nfs.h
index c3deba00a02..7d5163dfe0e 100644
--- a/xlators/nfs/server/src/nfs.h
+++ b/xlators/nfs/server/src/nfs.h
@@ -91,6 +91,12 @@ struct nfs_state {
         gf_boolean_t            server_aux_gids;
 	uint32_t		server_aux_gids_max_age;
 	gid_cache_t		gid_cache;
+        uint32_t                generation;
+};
+
+struct nfs_inode_ctx {
+        struct list_head        shares;
+        uint32_t                generation;
 };
 
 #define gf_nfs_dvm_on(nfsstt)   (((struct nfs_state *)nfsstt)->dynamicvolumes == GF_NFS_DVM_ON)
diff --git a/xlators/nfs/server/src/nfs3.c b/xlators/nfs/server/src/nfs3.c
index 1c379997266..7c01d0301ba 100644
--- a/xlators/nfs/server/src/nfs3.c
+++ b/xlators/nfs/server/src/nfs3.c
@@ -32,6 +32,7 @@
 #include "nfs3.h"
 #include "mem-pool.h"
 #include "logging.h"
+#include "nfs-common.h"
 #include "nfs-fops.h"
 #include "nfs-inodes.h"
 #include "nfs-generics.h"
@@ -692,12 +693,21 @@ nfs3svc_getattr_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         cs = frame->local;
 
+        /*
+         * Somewhat counter-intuitively, we don't need to look for sh-failed
+         * here. Failing this getattr will generate a new lookup from the
+         * client, and nfs_fop_lookup_cbk will detect any self-heal failures.
+         */
+
         if (op_ret == -1) {
                 gf_log (GF_NFS, GF_LOG_WARNING,
                         "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req),
                         cs->resolvedloc.path, strerror (op_errno));
                 status = nfs3_errno_to_nfsstat3 (op_errno);
         }
+        else {
+                nfs_fix_generation(this,inode);
+        }
 
         nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_GETATTR,
                              status, op_errno);
@@ -743,6 +753,9 @@ nfs3_getattr_resume (void *carg)
         int                             ret = -EFAULT;
         nfs_user_t                      nfu = {0, };
         nfs3_call_state_t               *cs = NULL;
+        uint64_t                         raw_ctx = 0;
+        struct nfs_inode_ctx            *ictx = NULL;
+        struct nfs_state                *priv = NULL;
 
         if (!carg)
                 return ret;
@@ -769,9 +782,28 @@ nfs3_getattr_resume (void *carg)
 		goto nfs3err;
 	}
 
+        /*
+         * If brick state changed, we need to force a proper lookup cycle (as
+         * would happen in native protocol) to do self-heal checks. We detect
+         * this by comparing the generation number for the last successful
+         * creation/lookup on the inode to the current number, so inodes that
+         * haven't been validated since the state change are affected.
+         */
+        if (inode_ctx_get(cs->resolvedloc.inode,cs->nfsx,&raw_ctx) == 0) {
+                ictx = (struct nfs_inode_ctx *)raw_ctx;
+                priv = cs->nfsx->private;
+                if (ictx->generation != priv->generation) {
+                        ret = nfs_lookup (cs->nfsx, cs->vol, &nfu,
+                                          &cs->resolvedloc,
+                                          nfs3svc_getattr_lookup_cbk, cs);
+                        goto check_err;
+                }
+        }
+
         ret = nfs_stat (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc,
                         nfs3svc_getattr_stat_cbk, cs);
 
+check_err:
         if (ret < 0) {
                 gf_log (GF_NFS3, GF_LOG_ERROR, "Stat fop failed: %s: %s",
                         cs->oploc.path, strerror (-ret));
diff --git a/xlators/nfs/server/src/nlm4.c b/xlators/nfs/server/src/nlm4.c
index 05de66d5633..2e02a5836c8 100644
--- a/xlators/nfs/server/src/nlm4.c
+++ b/xlators/nfs/server/src/nlm4.c
@@ -1789,25 +1789,38 @@ nlm4_add_share_to_inode (nlm_share_t *share)
         struct list_head      *head    = NULL;
         xlator_t              *this    = NULL;
         inode_t               *inode   = NULL;
+        struct nfs_inode_ctx  *ictx    = NULL;
+        struct nfs_state      *priv    = NULL;
 
         this = THIS;
+        priv = this->private;
         inode = share->inode;
         ret = inode_ctx_get (inode, this, &ctx);
 
-        head = (struct list_head *)ctx;
-
         if (ret || !head) {
-                head = GF_CALLOC (1, sizeof (struct list_head),
-                                  gf_common_mt_list_head);
-                if (!head ) {
+                ictx = GF_CALLOC (1, sizeof (struct nfs_inode_ctx),
+                                  gf_nfs_mt_inode_ctx);
+                if (!ictx ) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "could not allocate nfs inode ctx");
                         ret = -1;
                         goto out;
                 }
+                ictx->generation = priv->generation;
 
+                head = &ictx->shares;
                 INIT_LIST_HEAD (head);
-                ret = inode_ctx_put (inode, this, (uint64_t)head);
-                if (ret)
+
+                ret = inode_ctx_put (inode, this, (uint64_t)ictx);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "could not store share list");
                         goto out;
+                }
+        }
+        else {
+                ictx = (struct nfs_inode_ctx *)ctx;
+                head = &ictx->shares;
         }
 
         list_add (&share->inode_list, head);
@@ -1829,6 +1842,7 @@ nlm4_approve_share_reservation (nfs3_call_state_t *cs)
         inode_t                  *inode             = NULL;
         nlm_share_t              *share             = NULL;
         struct list_head         *head              = NULL;
+        struct nfs_inode_ctx     *ictx              = NULL;
 
         if (!cs)
                 goto out;
@@ -1840,8 +1854,9 @@ nlm4_approve_share_reservation (nfs3_call_state_t *cs)
                 ret = 0;
                 goto out;
         }
+        ictx = (struct nfs_inode_ctx *)ctx;
 
-        head = (struct list_head *)ctx;
+        head = &ictx->shares;
         if (!head || list_empty (head))
                 goto out;
 
@@ -2014,18 +2029,19 @@ nlm4svc_share (rpcsvc_request_t *req)
 int
 nlm4_remove_share_reservation (nfs3_call_state_t *cs)
 {
-        int                ret        = -1;
-        uint64_t           ctx        = 0;
-        fsh_mode           req_mode   = 0;
-        fsh_access         req_access = 0;
-        nlm_share_t       *share      = NULL;
-        nlm_share_t       *tmp        = NULL;
-        nlm_client_t      *client     = NULL;
-        char              *caller     = NULL;
-        inode_t           *inode      = NULL;
-        xlator_t          *this       = NULL;
-        struct list_head  *head       = NULL;
-        nlm4_shareargs    *args       = NULL;
+        int                      ret        = -1;
+        uint64_t                 ctx        = 0;
+        fsh_mode                 req_mode   = 0;
+        fsh_access               req_access = 0;
+        nlm_share_t             *share      = NULL;
+        nlm_share_t             *tmp        = NULL;
+        nlm_client_t            *client     = NULL;
+        char                    *caller     = NULL;
+        inode_t                 *inode      = NULL;
+        xlator_t                *this       = NULL;
+        struct list_head        *head       = NULL;
+        nlm4_shareargs          *args       = NULL;
+        struct nfs_inode_ctx    *ictx       = NULL;
 
         LOCK (&nlm_client_list_lk);
 
@@ -2055,8 +2071,9 @@ nlm4_remove_share_reservation (nfs3_call_state_t *cs)
                         inode->gfid, caller);
                 goto out;
         }
+        ictx = (struct nfs_inode_ctx *)ctx;
 
-        head = (struct list_head *)ctx;
+        head = &ictx->shares;
         if (list_empty (head)) {
                 ret = -1;
                 goto out;