DHT/readdirp: Directory not shown/healed on mount point if exists on single brick(non first up subvolume).

Problem: If snapshot is taken, when mkdir has succeeded only on
hashed_subvolume, then after restoring snapshot the directory
is not shown on mount point.

Why:    dht_readdirp takes only those directory entries in to
account, which are present on first_up_subvolume. Hence, if the
"hashed subvolume" is not same as first_up_subvolume, it wont be listed
on mount point and also not healed.

Solution:
Case 1: (Rebalance not running)If hashed subvolume is NULL or down then
filter in first_up_subvolume. Other wise the corresponding hashed subvolume
will take care of the directory entry.

Case 2: If readdirp_optimize option is turned on then read from first_up_subvol

Change-Id: Idaad28f1c9f688dbfb1a8a3ab8b244510c02365e
BUG: 1139986
Signed-off-by: Susant Palai <spalai@redhat.com>
Reviewed-on: http://review.gluster.org/7599
Reviewed-by: Raghavendra G <rgowdapp@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
Reviewed-on: http://review.gluster.org/8671
Reviewed-by: Kaleb KEITHLEY <kkeithle@redhat.com>
Tested-by: Kaleb KEITHLEY <kkeithle@redhat.com>

3 files changed, 62 insertions, 5 deletions

diff --git a/xlators/cluster/dht/src/dht-common.c b/xlators/cluster/dht/src/dht-common.c
index 0234bbbd409..29a63437f7b 100644
--- a/xlators/cluster/dht/src/dht-common.c
+++ b/xlators/cluster/dht/src/dht-common.c
@@ -3035,7 +3035,6 @@ err:
         return 0;
 }
 
-
 int
 dht_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
                   int op_errno, gf_dirent_t *orig_entries, dict_t *xdata)
@@ -3051,7 +3050,9 @@ dht_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
         dht_layout_t *layout = 0;
         dht_conf_t   *conf   = NULL;
         xlator_t     *subvol = 0;
+        xlator_t     *hashed_subvol = 0;
         int           ret    = 0;
+        int           readdir_optimize = 0;
 
         INIT_LIST_HEAD (&entries.list);
         prev = cookie;
@@ -3066,15 +3067,47 @@ dht_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
 
         layout = local->layout;
 
+        if (conf->readdir_optimize == _gf_true)
+                 readdir_optimize = 1;
+
         list_for_each_entry (orig_entry, (&orig_entries->list), list) {
                 next_offset = orig_entry->d_off;
-                if ((check_is_dir (NULL, (&orig_entry->d_stat), NULL) &&
-		    (prev->this != local->first_up_subvol)) ||
-                    check_is_linkfile (NULL, (&orig_entry->d_stat),
+                if (check_is_dir (NULL, (&orig_entry->d_stat), NULL)) {
+
+                /*Directory entries filtering :
+                 * a) If rebalance is running, pick from first_up_subvol
+                 * b) (rebalance not running)hashed subvolume is NULL or
+                 * down then filter in first_up_subvolume. Other wise the
+                 * corresponding hashed subvolume will take care of the
+                 * directory entry.
+                 */
+
+                        if (readdir_optimize) {
+                                if (prev->this == local->first_up_subvol)
+                                        goto list;
+                                else
+                                        continue;
+
+                        }
+
+                        hashed_subvol = dht_layout_search (this, layout, \
+                                                           orig_entry->d_name);
+
+                        if (prev->this == hashed_subvol)
+                                goto list;
+                        if ((hashed_subvol
+                                && dht_subvol_status (conf, hashed_subvol))
+                                ||(prev->this != local->first_up_subvol))
+                                continue;
+
+                        goto list;
+                }
+
+                if (check_is_linkfile (NULL, (&orig_entry->d_stat),
                                        orig_entry->dict)) {
                         continue;
                 }
-
+list:
                 entry = gf_dirent_for_name (orig_entry->d_name);
                 if (!entry) {
 
diff --git a/xlators/cluster/dht/src/dht-common.h b/xlators/cluster/dht/src/dht-common.h
index 3e9fabdbd9b..49fbd3878a0 100644
--- a/xlators/cluster/dht/src/dht-common.h
+++ b/xlators/cluster/dht/src/dht-common.h
@@ -740,4 +740,15 @@ dht_subvol_maxspace_nonzeroinode (xlator_t *this, xlator_t *subvol,
                                   dht_layout_t *layout);
 int
 dht_linkfile_attr_heal (call_frame_t *frame, xlator_t *this);
+
+void
+dht_layout_dump (dht_layout_t  *layout, const char *prefix);
+int32_t
+dht_priv_dump (xlator_t *this);
+int32_t
+dht_inodectx_dump (xlator_t *this, inode_t *inode);
+
+int
+dht_subvol_status (dht_conf_t *conf, xlator_t *subvol);
+
 #endif/* _DHT_H */
diff --git a/xlators/cluster/dht/src/dht-helper.c b/xlators/cluster/dht/src/dht-helper.c
index dd8dc495d02..5a9047836cc 100644
--- a/xlators/cluster/dht/src/dht-helper.c
+++ b/xlators/cluster/dht/src/dht-helper.c
@@ -1073,3 +1073,16 @@ int dht_inode_ctx_set (inode_t *inode, xlator_t *this, dht_inode_ctx_t *ctx)
 out:
         return ret;
 }
+
+int
+dht_subvol_status (dht_conf_t *conf, xlator_t *subvol)
+{
+        int i;
+
+        for (i=0 ; i < conf->subvolume_cnt; i++) {
+                if (conf->subvolumes[i] == subvol) {
+                        return conf->subvolume_status[i];
+                }
+        }
+        return 0;
+}