Bug#579005: XFS internal error xfs_da_do_buf(2) at line 2085 of file fs/xfs/xfs_da_btree.c.

[Jonathan Nieder <jrnieder@gmail.com>]

Rudy Gevaert wrote:

> Hi Jonathan, thanks for getting back on this so quickly

>> If this is reproducible, would you be able to test patches to debug
>> this if provided?
>
> When I do something in a specific directory I get the backtrace.  So
> the current corruption (that I'm aware of) is limited to a specific
> directory.

Interesting.  So this might be a case of the kernel not coping well
with previous filesystem corruption.

[...]
> I can export this lun to an other server that is booted with the
> newer kernel.  I guess I'll have to do a repair when I remount the
> lun.  If the lun isn't damaged (or it is ver limited) I will need to
> bring the service back online.
>
> (I can failover but the idea is to run not to long without a replica...)
>
> Also I will have to schedule a maintenance window for this as this
> is production.  Also normally I can't schedule a window now because
> of university exam period...
>
> Your input is very much appreciated.

Feel free to take your time; there's no hurry.  Anyway, it's probably
best to get in touch with the XFS maintainers first in case they have
ideas for tracking it down more easily.

Please send a summary of the symptoms so far to xfs@oss.sgi.com,
cc-ing either me or this bug log so we can track it.  Be sure to
mention:

 - what steps you perform to reproduce this, what happens, and how
   that differs from what you expected (should be easy in this
   example)
 - which kernel versions you have tried and results from each
 - "dmesg" output from booting up as an attachment, if you happen to
   have it
 - what kind of further debugging you would be able and not able to do
 - a link to this bug log, in case someone wants to read the full
   story

If we are lucky, someone on that list might suggest commands to help
diagnose it, which should make it easier for others to artificially
reproduce, see if 3.x.y is affected, and make sure it is fixed in
3.x.y and 2.6.32.y.

Thanks and hope that helps.
Jonathan