afr, index: Clean up stale directory and file indices in granular entry sh

Specifically when a directory tree is removed (rm -rf)
while a brick is down, both the directory index and the
name indices of the files and subdirs under it will remain.
Self-heal will need to pick up these and remove them.

Towards this, afr sh will now also crawl indices/entry-changes
and call an rmdir on the dir if the directory index is stale.

On the brick side, rmdir fop has been implemented for index xl,
which would delete the directory index and its contents if present
in a synctask.

Change-Id: I8b527331c2547e6c141db6c57c14055ad1198a7e
BUG: 1331323
Signed-off-by: Krutika Dhananjay <kdhananj@redhat.com>
Reviewed-on: http://review.gluster.org/14832
Reviewed-by: Ravishankar N <ravishankar@redhat.com>
Smoke: Gluster Build System <jenkins@build.gluster.org>
NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>
Reviewed-by: Pranith Kumar Karampuri <pkarampu@redhat.com>
CentOS-regression: Gluster Build System <jenkins@build.gluster.org>

3 files changed, 53 insertions, 13 deletions

diff --git a/xlators/cluster/afr/src/afr-self-heal-entry.c b/xlators/cluster/afr/src/afr-self-heal-entry.c
index c9a27bf8e20..985cebe76b9 100644
--- a/xlators/cluster/afr/src/afr-self-heal-entry.c
+++ b/xlators/cluster/afr/src/afr-self-heal-entry.c
@@ -567,7 +567,7 @@ afr_selfheal_entry_dirent (call_frame_t *frame, xlator_t *this,
 
                 if ((ret == 0) && (priv->esh_granular) && parent_idx_inode) {
                         ret = afr_shd_index_purge (subvol, parent_idx_inode,
-                                                   name);
+                                                   name, inode->ia_type);
                         /* Why is ret force-set to 0? We do not care about
                          * index purge failing for full heal as it is quite
                          * possible during replace-brick that not all files
@@ -748,7 +748,11 @@ afr_selfheal_entry_granular_dirent (xlator_t *subvol, gf_dirent_t *entry,
 
         ret = syncop_lookup (args->xl, &loc, &iatt, NULL, NULL, NULL);
         if ((ret == -ENOENT) || (ret == -ESTALE)) {
-                afr_shd_index_purge (subvol, parent->inode, entry->d_name);
+                /* The name indices under the pgfid index dir are guaranteed
+                 * to be regular files. Hence the hardcoding.
+                 */
+                afr_shd_index_purge (subvol, parent->inode, entry->d_name,
+                                     IA_IFREG);
                 ret = 0;
                 goto out;
         }
diff --git a/xlators/cluster/afr/src/afr-self-heald.c b/xlators/cluster/afr/src/afr-self-heald.c
index b29a822568d..7ccac919769 100644
--- a/xlators/cluster/afr/src/afr-self-heald.c
+++ b/xlators/cluster/afr/src/afr-self-heald.c
@@ -153,10 +153,13 @@ unlock:
 inode_t *
 afr_shd_inode_find (xlator_t *this, xlator_t *subvol, uuid_t gfid)
 {
-	inode_t *inode = NULL;
-	int ret = 0;
-	loc_t loc = {0, };
-	struct iatt iatt = {0, };
+	int           ret      = 0;
+        uint64_t       val     = IA_INVAL;
+	loc_t         loc      = {0, };
+        dict_t       *xdata    = NULL;
+        dict_t       *rsp_dict = NULL;
+	inode_t      *inode    = NULL;
+	struct iatt   iatt     = {0, };
 
 	inode = inode_find (this->itable, gfid);
 	if (inode)
@@ -167,12 +170,32 @@ afr_shd_inode_find (xlator_t *this, xlator_t *subvol, uuid_t gfid)
 		goto out;
 	gf_uuid_copy (loc.gfid, gfid);
 
-	ret = syncop_lookup (subvol, &loc, &iatt, NULL, NULL, NULL);
+        xdata = dict_new ();
+        if (!xdata)
+                goto out;
+
+        ret = dict_set_int8 (xdata, GF_INDEX_IA_TYPE_GET_REQ, 1);
+        if (ret)
+                goto out;
+
+	ret = syncop_lookup (subvol, &loc, &iatt, NULL, xdata, &rsp_dict);
 	if (ret < 0)
 		goto out;
 
+        if (rsp_dict) {
+                ret = dict_get_uint64 (rsp_dict, GF_INDEX_IA_TYPE_GET_RSP,
+                                       &val);
+                if (ret)
+                        goto out;
+        }
+
 	inode = inode_link (loc.inode, NULL, NULL, &iatt);
+        ret = inode_ctx_set2 (inode, subvol, 0, &val);
 out:
+        if (xdata)
+                dict_unref (xdata);
+        if (rsp_dict)
+                dict_unref (rsp_dict);
 	loc_wipe (&loc);
 	return inode;
 }
@@ -215,15 +238,19 @@ out:
 }
 
 int
-afr_shd_index_purge (xlator_t *subvol, inode_t *inode, char *name)
+afr_shd_index_purge (xlator_t *subvol, inode_t *inode, char *name,
+                     ia_type_t type)
 {
-	loc_t loc = {0, };
-	int ret = 0;
+	int    ret = 0;
+	loc_t  loc = {0,};
 
 	loc.parent = inode_ref (inode);
 	loc.name = name;
 
-	ret = syncop_unlink (subvol, &loc, NULL, NULL);
+        if (IA_ISDIR (type))
+                ret = syncop_rmdir (subvol, &loc, 1, NULL, NULL);
+        else
+                ret = syncop_unlink (subvol, &loc, NULL, NULL);
 
 	loc_wipe (&loc);
 	return ret;
@@ -395,6 +422,7 @@ afr_shd_index_heal (xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
         afr_private_t        *priv   = NULL;
         uuid_t               gfid    = {0};
         int                  ret     = 0;
+        uint64_t             val     = IA_INVAL;
 
         priv = healer->this->private;
         if (!priv->shd.enabled)
@@ -407,10 +435,13 @@ afr_shd_index_heal (xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
         if (ret)
                 return 0;
 
+        inode_ctx_get2 (parent->inode, subvol, NULL, &val);
+
         ret = afr_shd_selfheal (healer, healer->subvol, gfid);
 
         if (ret == -ENOENT || ret == -ESTALE)
-                afr_shd_index_purge (subvol, parent->inode, entry->d_name);
+                afr_shd_index_purge (subvol, parent->inode, entry->d_name, val);
+
         if (ret == 2)
                 /* If bricks crashed in pre-op after creating indices/xattrop
                  * link but before setting afr changelogs, we end up with stale
@@ -489,6 +520,10 @@ afr_shd_index_sweep_all (struct subvol_healer *healer)
                 goto out;
         count += ret;
 
+        ret = afr_shd_index_sweep (healer, GF_XATTROP_ENTRY_CHANGES_GFID);
+        if (ret < 0)
+                goto out;
+        count += ret;
 out:
         if (ret < 0)
                 return ret;
diff --git a/xlators/cluster/afr/src/afr-self-heald.h b/xlators/cluster/afr/src/afr-self-heald.h
index f591515669c..c6ac5ebfd1b 100644
--- a/xlators/cluster/afr/src/afr-self-heald.h
+++ b/xlators/cluster/afr/src/afr-self-heald.h
@@ -75,5 +75,6 @@ afr_shd_gfid_to_path (xlator_t *this, xlator_t *subvol, uuid_t gfid,
                       char **path_p);
 
 int
-afr_shd_index_purge (xlator_t *subvol, inode_t *inode, char *name);
+afr_shd_index_purge (xlator_t *subvol, inode_t *inode, char *name,
+                     ia_type_t type);
 #endif /* !_AFR_SELF_HEALD_H */