Re: Corruption of root fs during git bisect of drm system hang

[Dave Chinner <david@fromorbit.com>]

On Thu, Jul 11, 2013 at 05:36:21AM +0200, Markus Trippelsdorf wrote:
> On 2013.07.11 at 10:31 +1000, Dave Chinner wrote:
> > On Wed, Jul 10, 2013 at 11:06:34AM +0200, Markus Trippelsdorf wrote:
> > > While bisecting a system hang, caused by the drm gpu subsystem, my root fs got
> > > corrupted:
> > 
> > I don't see any corruption being repaired....
> > 
> > > 
> > >  # xfs_repair /dev/sdc2
> > > Phase 1 - find and verify superblock...
> > > Phase 2 - using internal log
> > >         - zero log...
> > >         - scan filesystem freespace and inode maps...
> > > agi unlinked bucket 6 is 682886 in ag 3 (inode=101346182)
> > > agi unlinked bucket 7 is 11335 in ag 3 (inode=100674631)
> > > agi unlinked bucket 10 is 682890 in ag 3 (inode=101346186)
> > > agi unlinked bucket 21 is 981 in ag 3 (inode=100664277)
> > > agi unlinked bucket 23 is 5704343 in ag 3 (inode=106367639)
> > > agi unlinked bucket 29 is 211421 in ag 3 (inode=100874717)
> > > agi unlinked bucket 31 is 7681375 in ag 3 (inode=108344671)
> > > agi unlinked bucket 34 is 3480162 in ag 3 (inode=104143458)
> > > agi unlinked bucket 40 is 211432 in ag 3 (inode=100874728)
> > > agi unlinked bucket 41 is 2704937 in ag 3 (inode=103368233)
> > > agi unlinked bucket 45 is 594669 in ag 3 (inode=101257965)
> > > agi unlinked bucket 62 is 11902 in ag 3 (inode=100675198)
> > 
> > That's a filesystem that has unlinked inodes on the unlinked list.
> > They get cleaned up during log replay. All the other "errors" are
> > related to cleaning these up....
> > 
> > So what is making you think there is a corruption? What's the error
> > being reported when you are using the filesystem? i.e. what's the
> > entire process you go through before you get to finding this
> > problem?
> 
> I was loosing my KDE settings bit by bit with every reboot during the
> bisection. First my window-rules disappeared, then my desktop background
> changed to default, then my taskbar moved from top to the bottom, etc.
> In the end I had to restore all my .files from backup. 

That's not filesystem corruption. That sounds more like someone not
using fsync in the apropriate place when overwriting a file....

> And please note that xfs_repair unlinked the inodes _after_ the
> filesystem has been mounted and unmounted normally. 

Which means we might not be processing the unlinked lists correctly
and leaking them. If repair is finding the inodes in the AGI
unlinked lists, then recovery should be finding them, too. Not
processing them and not clearing the AGI bucket tends to imply that
recovery failed to read the AGI buffer.

What error messages are in dmesg, if any? And what kernel are you
running?

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs