cluster/afr: Make entry-self-heal in afr-v2 compatible with afr-v1

        Backport of http://review.gluster.org/9227

Problem:
entry self-heal in 3.6 and above, takes full lock on the directory only for the
duration of figuring out the xattrs of the directories where as 3.5 takes locks
through out the entry-self-heal. If the cluster is heterogeneous then there is
a chance that 3.6 self-heal is triggered and then 3.5 self-heal will also
triggered and both the self-heal daemons of 3.5 and 3.6 do self-heal.

Fix:
In 3.6.x and above get an entry lock on a very long name before entry self-heal
begins so that 3.5 entry self-heal will not get locks until 3.6.x entry
self-heal completes.

BUG: 1177418
Change-Id: Iecf49d794c6b480e38563e39599a40067b3a21cb
Signed-off-by: Pranith Kumar K <pkarampu@redhat.com>
Reviewed-on: http://review.gluster.org/9355
Reviewed-by: Krutika Dhananjay <kdhananj@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Raghavendra Bhat <raghavendra@redhat.com>

1 files changed, 28 insertions, 1 deletions

diff --git a/xlators/cluster/afr/src/afr-self-heal-entry.c b/xlators/cluster/afr/src/afr-self-heal-entry.c
index 3f753251e7c..4b513ffc73d 100644
--- a/xlators/cluster/afr/src/afr-self-heal-entry.c
+++ b/xlators/cluster/afr/src/afr-self-heal-entry.c
@@ -19,6 +19,15 @@
 #include "byte-order.h"
 #include "afr-transaction.h"
 
+/* Max file name length is 255 this filename is of length 256. No file with
+ * this name can ever come, entry-lock with this name is going to prevent
+ * self-heals from older versions while the granular entry-self-heal is going
+ * on in newer version.*/
+#define LONG_FILENAME "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"\
+                      "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"\
+                      "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"\
+                      "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"\
+                      "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
 
 static int
 afr_selfheal_entry_delete (xlator_t *this, inode_t *dir, const char *name,
@@ -679,6 +688,7 @@ afr_selfheal_entry (call_frame_t *frame, xlator_t *this, inode_t *inode)
 {
 	afr_private_t *priv = NULL;
 	unsigned char *locked_on = NULL;
+        unsigned char *long_name_locked = NULL;
 	fd_t *fd = NULL;
 	int ret = 0;
 
@@ -689,6 +699,7 @@ afr_selfheal_entry (call_frame_t *frame, xlator_t *this, inode_t *inode)
 		return -EIO;
 
 	locked_on = alloca0 (priv->child_count);
+	long_name_locked = alloca0 (priv->child_count);
 
 	ret = afr_selfheal_tryentrylk (frame, this, inode, priv->sh_domain, NULL,
 				       locked_on);
@@ -707,7 +718,23 @@ afr_selfheal_entry (call_frame_t *frame, xlator_t *this, inode_t *inode)
 			goto unlock;
 		}
 
-		ret = __afr_selfheal_entry (frame, this, fd, locked_on);
+                ret = afr_selfheal_tryentrylk (frame, this, inode, this->name,
+                                               LONG_FILENAME, long_name_locked);
+                {
+                        if (ret < 1) {
+                                gf_log (this->name, GF_LOG_DEBUG, "%s: Skipping"
+                                        " entry self-heal as only %d "
+                                        "sub-volumes could be locked in "
+                                        "special-filename domain",
+                                        uuid_utoa (fd->inode->gfid),
+                                        ret);
+                                ret = -ENOTCONN;
+                                goto unlock;
+                        }
+                        ret = __afr_selfheal_entry (frame, this, fd, locked_on);
+                }
+                afr_selfheal_unentrylk (frame, this, inode, this->name,
+                                        LONG_FILENAME, long_name_locked);
 	}
 unlock:
 	afr_selfheal_unentrylk (frame, this, inode, priv->sh_domain, NULL, locked_on);