Skip to contentSkip to conversation

ocfs2: unlock rw lock if inode lock failed

PATCH New Joseph Qi 3 participants 5 messages May 6, 2013 – May 14, 2013 OCFS2

lore
Share & tools
lore

Joseph Qi

author
#0 lore raw

PATCH 1/1 ocfs2: unlock rw lock if inode lock failed

In ocfs2_file_aio_write, it does ocfs2_rw_lock first and then ocfs2_inode_lock. But if ocfs2_inode_lock failed, it goes to out_sems without unlocking rw lock. This will cause a bug in ocfs2_lock_res_free when testing res->l_ex_holders, which is increased in __ocfs2_cluster_lock and decreased in __ocfs2_cluster_unlock.

---
 fs/ocfs2/file.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/ocfs2/file.c +1 -1
index 6474cb4..e2cd7a8 100644
@@ -2290,7 +2290,7 @@ relock:
 		ret = ocfs2_inode_lock(inode, NULL, 1);
 		if (ret < 0) {
 			mlog_errno(ret);
-			goto out_sems;
+			goto out;
 		}
 
 		ocfs2_inode_unlock(inode, 1);
Signature · 1 line
1.7.9.7
#1 lore raw

PATCH 1/1 ocfs2: unlock rw lock if inode lock failed

Looks good to me.

Quoted 29 lines · Joseph Qi

In ocfs2_file_aio_write, it does ocfs2_rw_lock first and then ocfs2_inode_lock. But if ocfs2_inode_lock failed, it goes to out_sems without unlocking rw lock. This will cause a bug in ocfs2_lock_res_free when testing res->l_ex_holders, which is increased in __ocfs2_cluster_lock and decreased in __ocfs2_cluster_unlock.

---
 fs/ocfs2/file.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/ocfs2/file.c +1 -2
index 6474cb4..e2cd7a8 100644
@@ -2290,7 +2290,7 @@ relock:
                ret = ocfs2_inode_lock(inode, NULL, 1);
                if (ret < 0) {
                        mlog_errno(ret);
-                       goto out_sems;
+                       goto out;
                }
 
                ocfs2_inode_unlock(inode, 1);
--

1.7.9.7

Quoted 5 lines

_______________________________________________
Ocfs2-devel mailing list
Ocfs2-devel at oss.oracle.com
https://oss.oracle.com/mailman/listinfo/ocfs2-devel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://oss.oracle.com/pipermail/ocfs2-devel/attachments/20130506/5718074c/attachment.html

#2 lore raw

PATCH 1/1 ocfs2: unlock rw lock if inode lock failed

In reply to Joseph Qi
Quoted 11 lines · On Mon, 6 May 2013 22:43:39 +0800 Joseph Qi <[email protected]> wrote:

In ocfs2_file_aio_write, it does ocfs2_rw_lock first and then ocfs2_inode_lock. But if ocfs2_inode_lock failed, it goes to out_sems without unlocking rw lock. This will cause a bug in ocfs2_lock_res_free when testing res->l_ex_holders, which is increased in __ocfs2_cluster_lock and decreased in __ocfs2_cluster_unlock.

...

fs/ocfs2/file.c relock: · line 2296

Quoted 8 lines
fs/ocfs2/file.c +1 -1
@@ -2290,7 +2290,7 @@ relock:
 		ret = ocfs2_inode_lock(inode, NULL, 1);
 		if (ret < 0) {
 			mlog_errno(ret);
-			goto out_sems;
+			goto out;
 		}
 
 		ocfs2_inode_unlock(inode, 1);

That seems like a fairly serious bug. How long has it been there and what userspace actions are required to trigger it?

(I'm trying to work out which kernel versions we should merge the fix into, but the changelog didn't give me enough info to determine this)

#3 lore raw

PATCH 1/1 ocfs2: unlock rw lock if inode lock failed

It should almost never trigger. ocfs2_inode_lock() should always succeed and only return after it has gotten the required lock.

Quoted 25 lines · Andrew Morton
Quoted 11 lines · On Mon, 6 May 2013 22:43:39 +0800 Joseph Qi <[email protected]> wrote:

In ocfs2_file_aio_write, it does ocfs2_rw_lock first and then ocfs2_inode_lock. But if ocfs2_inode_lock failed, it goes to out_sems without unlocking rw lock. This will cause a bug in ocfs2_lock_res_free when testing res->l_ex_holders, which is increased in __ocfs2_cluster_lock and decreased in __ocfs2_cluster_unlock.

...

fs/ocfs2/file.c relock: · line 2296

Quoted 8 lines
fs/ocfs2/file.c +1 -1
@@ -2290,7 +2290,7 @@ relock:
              ret = ocfs2_inode_lock(inode, NULL, 1);
              if (ret < 0) {
                      mlog_errno(ret);
-                     goto out_sems;
+                     goto out;
              }
 
              ocfs2_inode_unlock(inode, 1);

That seems like a fairly serious bug. How long has it been there and what userspace actions are required to trigger it?

Quoted 9 lines

(I'm trying to work out which kernel versions we should merge the fix into, but the changelog didn't give me enough info to determine this)

_______________________________________________
Ocfs2-devel mailing list
Ocfs2-devel at oss.oracle.com
https://oss.oracle.com/mailman/listinfo/ocfs2-devel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://oss.oracle.com/pipermail/ocfs2-devel/attachments/20130508/4ec622c9/attachment.html

author
#4 lore raw

PATCH 1/1 ocfs2: unlock rw lock if inode lock failed

In reply to Andrew Morton
Quoted 25 lines · Andrew Morton
Quoted 11 lines · On Mon, 6 May 2013 22:43:39 +0800 Joseph Qi <[email protected]> wrote:

In ocfs2_file_aio_write, it does ocfs2_rw_lock first and then ocfs2_inode_lock. But if ocfs2_inode_lock failed, it goes to out_sems without unlocking rw lock. This will cause a bug in ocfs2_lock_res_free when testing res->l_ex_holders, which is increased in __ocfs2_cluster_lock and decreased in __ocfs2_cluster_unlock.

...

fs/ocfs2/file.c relock: · line 2296

Quoted 8 lines
fs/ocfs2/file.c +1 -1
@@ -2290,7 +2290,7 @@ relock:
 		ret = ocfs2_inode_lock(inode, NULL, 1);
 		if (ret < 0) {
 			mlog_errno(ret);
-			goto out_sems;
+			goto out;
 		}
 
 		ocfs2_inode_unlock(inode, 1);

That seems like a fairly serious bug. How long has it been there and what userspace actions are required to trigger it?

Quoted 6 lines

(I'm trying to work out which kernel versions we should merge the fix into, but the changelog didn't give me enough info to determine this)

.

Sorry for the delayed reply.
The reproducible case is lots of write IOs plus storage link down and then restore.
And my kernel is 3.0.13.