]> pilppa.com Git - linux-2.6-omap-h63xx.git/commit
[XFS] Fix race in xfs_write() between direct and buffered I/O with DMAPI
authorLachlan McIlroy <lachlan@redback.melbourne.sgi.com>
Wed, 24 Dec 2008 03:07:32 +0000 (14:07 +1100)
committerLachlan McIlroy <lachlan@redback.melbourne.sgi.com>
Wed, 24 Dec 2008 03:07:32 +0000 (14:07 +1100)
commit25051158bbed127e8672b43396c71c5eb610e5f1
treeee93995d39c50956334157ca723f980f5d9fa7d2
parentad1ad968f4e7b06c75741575ea077e25a87da49a
[XFS] Fix race in xfs_write() between direct and buffered I/O with DMAPI

The iolock is dropped and re-acquired around the call to XFS_SEND_NAMESP().
While the iolock is released the file can become cached.  We then
'goto retry' and - if we are doing direct I/O - mapping->nrpages may now be
non zero but need_i_mutex will be zero and we will hit the WARN_ON().

Since we have dropped the I/O lock then the file size may have also changed
so what we need to do here is 'goto start' like we do for the XFS_SEND_DATA()
DMAPI event.

We also need to update the filesize before releasing the iolock so that
needs to be done before the XFS_SEND_NAMESP event.  If we drop the iolock
before setting the filesize we could race with a truncate.

Reviewed-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
fs/xfs/linux-2.6/xfs_lrw.c