Re: Xfs_repair segfaults.

[Filippo Stenico <filippo.stenico@xxxxxxxxx>]

On Mon, May 6, 2013 at 4:34 PM, Eric Sandeen <sandeen@xxxxxxxxxxx> wrote:

On 5/6/13 7:06 AM, Filippo Stenico wrote:
> Hello, I've had an issue on a raid5 volume (2 disks failing at same
> time, and buggy NAS firmware trying hard to sync then stop for I/O
> error then retryes to sync leading (!?) to lost raid and an xfs data
> corruption.

Sooo bad storage hardware, first off \o/  :(

 

> I've recreated raid, mounted the xfs filesystem, copied about 6TB of
> data out of 8TB, the rest of data being unreachable with "structrure
> needs cleaning" error. Thus I tried to unmount/mount and still got
> same error, then dumped metadata and tried various xfs_repair with
> different options but it always reaches a same point where it always
> segfaults.

OK, what version of xfsprogs?
If not latest usptream, please try that next.

It was the one included in debian squeeze, I believe v.3.1.4
I figured segfault could be fixed in newer releases, so I added testing and unstable sources, and tryied with those (3.1.7 and 3.1.9) but same segfault.
 

If upstream still segfaults, you could provide a core and associated
binary debug info, or (maybe) an xfs_metadump for analysis.  It sounds
like this fs is in pretty bad shape though, so even the xfs_metadump
might fail and/or not gather enough information.

As suggested on IRC chat, I built from xfsprogs-3.1.9 source, I will restore "original" metadata and make same steps I took first time, so that you can get a detailed report.
Of course I will include as much as info I can (metadump - gdb backtrace on eventual segfault).

Thanx for now,

-Eric

> If someone is interested of investigating what's going on here and
> helping me recover more data I would be happy to send in in more
> details....
>
> Cheers
>
> -- F
>
>

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

-- 
F