ceph: fix iput race when queueing inode work
authorSage Weil <sage@newdream.net>
Sun, 6 Nov 2011 05:06:31 +0000 (22:06 -0700)
committerSage Weil <sage@newdream.net>
Sun, 6 Nov 2011 05:06:31 +0000 (22:06 -0700)
commit15a2015fbc692e1c97d7ce12d96e077f5ae7ea6d
treef902cc2b7cab475c8b220147bb786dc3a772888e
parent0c6d4b4e22a513f8563a2e00c5ab08e9f8796272
ceph: fix iput race when queueing inode work

If we queue a work item that calls iput(), make sure we ihold() before
attempting to queue work. Otherwise our queued work might miraculously run
before we notice the queue_work() succeeded and call ihold(), allowing the
inode to be destroyed.

That is, instead of

if (queue_work(...))
ihold();

we need to do

ihold();
if (!queue_work(...))
iput();

Reported-by: Amon Ott <a.ott@m-privacy.de>
Signed-off-by: Sage Weil <sage@newdream.net>
fs/ceph/inode.c