On Wed, Sep 13, 2006 at 11:58:50AM +1000, David Chinner wrote:
> Call Trace:
> [<c013ae74>] lock_release_non_nested+0xd8/0x143
> [<c013b291>] lock_release+0x178/0x19f
> [<c02f7dc5>] __mutex_unlock_slowpath+0xbb/0x131
> [<c02f7e43>] mutex_unlock+0x8/0xa
> [<c017655f>] generic_shutdown_super+0x9c/0xd9
> [<c01765bc>] kill_block_super+0x20/0x32
> [<c017667c>] deactivate_super+0x5d/0x6f
> [<c01892bc>] mntput_no_expire+0x52/0x85
> [<c017b2c9>] path_release_on_umount+0x15/0x18
> [<c018a469>] sys_umount+0x1e1/0x215
> [<c018a4aa>] sys_oldumount+0xd/0xf
> [<c0103156>] sysenter_past_esp+0x5f/0x99
>
> I'm not sure why XFS would cause this - the crash is outside XFS releasing
> a mutex (sb->s_lock) that XFS code has never touched. I doubt anyone
> in the XFS team has done any testing on this -mm kernel...
>
> What is the test case, Michal? Can you post the script you used?
I've booted 2.6.18-rc6-mm2 and mounted and unmounted several xfs
filesystems. I'm currently running xfsqa on it, and I haven't seen
any failures on unmount yet.
That test case would be really handy, Michal.
Cheers,
Dave.
--
Dave Chinner
Principal Engineer
SGI Australian Software Group
|