[Cluster-devel] [GFS2 PATCH] gfs2: Only do glock put in gfs2_create_inode for free inodes

[Bob Peterson <rpeterso@redhat.com>]

Hi,

Before this patch, the error path of function gfs2_create_inode would
always calls gfs2_glock_put for the inode glock. That's good for inodes
that are free. But after they've been added to the vfs inodes, errors
will cause the inode to be evicted, and the evict will do the glock
put for us. If we do a glock put again, we can try to free the glock
while there are still references to it, e.g. revokes pending for
the transaction that created it.

This patch adds a check: if (free_vfs_inode) before the put, thus
solving the problem.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/inode.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index 5acd3ce30759..e3a27fd284dd 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -780,7 +780,8 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 fail_free_inode:
 	if (ip->i_gl) {
 		glock_clear_object(ip->i_gl, ip);
-		gfs2_glock_put(ip->i_gl);
+		if (free_vfs_inode) /* else evict will do the put for us */
+			gfs2_glock_put(ip->i_gl);
 	}
 	gfs2_rs_delete(ip, NULL);
 	gfs2_qa_put(ip);