cluster/afr: Fix heal-info slow response while IO is in progress

        Backport of: http://review.gluster.org/#/c/13326/

Now heal-info does an open() on the file being examined so that
the client at some point sees open-fd count being > 1 and releases
the eager-lock so that heal-info doesn't remain blocked forever
until IO completes.

Change-Id: I7d4a8aa4de459216408b666894ee7bb42e406547
BUG: 1303899
Signed-off-by: Krutika Dhananjay <kdhananj@redhat.com>
Reviewed-on: http://review.gluster.org/13348
Smoke: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Pranith Kumar Karampuri <pkarampu@redhat.com>
Tested-by: Pranith Kumar Karampuri <pkarampu@redhat.com>
NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>
CentOS-regression: Gluster Build System <jenkins@build.gluster.com>

3 files changed, 39 insertions, 19 deletions

diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index e457dbce73c..ccdaaaaf052 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -4515,12 +4515,13 @@ afr_selfheal_locked_data_inspect (call_frame_t *frame, xlator_t *this,
                                   gf_boolean_t *pflag)
 {
         int ret = -1;
-        afr_private_t   *priv = NULL;
         unsigned char *locked_on = NULL;
         unsigned char *data_lock = NULL;
         unsigned char *sources = NULL;
         unsigned char *sinks = NULL;
         unsigned char *healed_sinks = NULL;
+        afr_private_t   *priv = NULL;
+        fd_t          *fd = NULL;
         struct afr_reply *locked_replies = NULL;
 
         priv = this->private;
@@ -4530,6 +4531,18 @@ afr_selfheal_locked_data_inspect (call_frame_t *frame, xlator_t *this,
         sinks = alloca0 (priv->child_count);
         healed_sinks = alloca0 (priv->child_count);
 
+        /* Heal-info does an open() on the file being examined so that the
+         * current eager-lock holding client, if present, at some point sees
+         * open-fd count being > 1 and releases the eager-lock so that heal-info
+         * doesn't remain blocked forever until IO completes.
+         */
+        ret = afr_selfheal_data_open (this, inode, &fd);
+        if (ret < 0) {
+                gf_msg_debug (this->name, -ret, "%s: Failed to open",
+                              uuid_utoa (inode->gfid));
+                goto out;
+        }
+
         locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
 
         ret = afr_selfheal_tryinodelk (frame, this, inode, priv->sh_domain,
@@ -4567,6 +4580,8 @@ unlock:
 out:
         if (locked_replies)
                 afr_replies_wipe (locked_replies, priv->child_count);
+        if (fd)
+                fd_unref (fd);
         return ret;
 }
 
diff --git a/xlators/cluster/afr/src/afr-self-heal-data.c b/xlators/cluster/afr/src/afr-self-heal-data.c
index e04b6575e60..ebf262e4f36 100644
--- a/xlators/cluster/afr/src/afr-self-heal-data.c
+++ b/xlators/cluster/afr/src/afr-self-heal-data.c
@@ -803,31 +803,32 @@ out:
 }
 
 
-static fd_t *
-afr_selfheal_data_open (xlator_t *this, inode_t *inode)
+int
+afr_selfheal_data_open (xlator_t *this, inode_t *inode, fd_t **fd)
 {
-	loc_t loc = {0,};
-	int ret = 0;
-	fd_t *fd = NULL;
+	int         ret    = 0;
+        fd_t       *fd_tmp = NULL;
+	loc_t       loc    = {0,};
 
-	fd = fd_create (inode, 0);
-	if (!fd)
-		return NULL;
+	fd_tmp = fd_create (inode, 0);
+	if (!fd_tmp)
+		return -ENOMEM;
 
 	loc.inode = inode_ref (inode);
 	gf_uuid_copy (loc.gfid, inode->gfid);
 
-	ret = syncop_open (this, &loc, O_RDWR|O_LARGEFILE, fd, NULL, NULL);
-	if (ret) {
-		fd_unref (fd);
-		fd = NULL;
+	ret = syncop_open (this, &loc, O_RDWR|O_LARGEFILE, fd_tmp, NULL, NULL);
+	if (ret < 0) {
+		fd_unref (fd_tmp);
+                loc_wipe (&loc);
+                goto out;
 	} else {
-		fd_bind (fd);
+		fd_bind (fd_tmp);
 	}
 
-	loc_wipe (&loc);
-
-	return fd;
+        *fd = fd_tmp;
+out:
+	return ret;
 }
 
 int
@@ -840,9 +841,9 @@ afr_selfheal_data (call_frame_t *frame, xlator_t *this, inode_t *inode)
 
 	priv = this->private;
 
-	fd = afr_selfheal_data_open (this, inode);
+	ret = afr_selfheal_data_open (this, inode, &fd);
 	if (!fd) {
-                gf_msg_debug (this->name, 0, "%s: Failed to open",
+                gf_msg_debug (this->name, -ret, "%s: Failed to open",
                               uuid_utoa (inode->gfid));
                 return -EIO;
         }
diff --git a/xlators/cluster/afr/src/afr.h b/xlators/cluster/afr/src/afr.h
index 8e0a1f18816..52f9c513a9e 100644
--- a/xlators/cluster/afr/src/afr.h
+++ b/xlators/cluster/afr/src/afr.h
@@ -1097,4 +1097,8 @@ afr_get_need_heal (xlator_t *this);
 
 void
 afr_set_need_heal (xlator_t *this, afr_local_t *local);
+
+int
+afr_selfheal_data_open (xlator_t *this, inode_t *inode, fd_t **fd);
+
 #endif /* __AFR_H__ */