2.6.27-based kernel failing xfstests xfs/182

2.6.27-based kernel failing xfstests xfs/182

[Chris Friesen]

Hi,

I'm working on an embedded system running a 2.6.27-based kernel.

I'm digging into some intermittent xfs corruption issues.  Yes, I 
realize the right answer is probably to upgrade, but for the usual 
embedded reasons that's tricky.

I'm seeing test xfs/182 fail with 40 or so files showing messages like:

file /opt/xfstest2/960 has incorrect size - sync failed

In our case the xfs filesystems used for testing are on top of lvm, 
which is in turn on top of two SAS drives in RAID1.

Any suggestions on where I should start looking?  I'm looking at the 
"do_sync() and XFSQA test 182 failures...." thread on the mailing list 
from 2008, can someone point me to the commit that eventually resulted 
from that?

Thanks,
Chris


-- 

Chris Friesen
Software Designer
www.genband.com
office:+1.343.883.2717

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

Re: 2.6.27-based kernel failing xfstests xfs/182

[Dave Chinner]

On Mon, May 13, 2013 at 05:27:35PM -0600, Chris Friesen wrote:
> 
> Hi,
> 
> I'm working on an embedded system running a 2.6.27-based kernel.
> 
> I'm digging into some intermittent xfs corruption issues.  Yes, I
> realize the right answer is probably to upgrade, but for the usual
> embedded reasons that's tricky.
> 
> I'm seeing test xfs/182 fail with 40 or so files showing messages like:
> 
> file /opt/xfstest2/960 has incorrect size - sync failed
> 
> In our case the xfs filesystems used for testing are on top of lvm,
> which is in turn on top of two SAS drives in RAID1.
> 
> Any suggestions on where I should start looking?  I'm looking at the
> "do_sync() and XFSQA test 182 failures...." thread on the mailing
> list from 2008, can someone point me to the commit that eventually
> resulted from that?

I doubt anyone can pinpoint a fix for you. I'd suggest that you find
the kernel release where the test starts passing, and the work
backwards from there...

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

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

Re: 2.6.27-based kernel failing xfstests xfs/182 -- resolved, I think

[Chris Friesen]

On 05/13/2013 05:27 PM, Chris Friesen wrote:
>
> Hi,
>
> I'm working on an embedded system running a 2.6.27-based kernel.
>
> I'm digging into some intermittent xfs corruption issues. Yes, I realize
> the right answer is probably to upgrade, but for the usual embedded
> reasons that's tricky.
>
> I'm seeing test xfs/182 fail with 40 or so files showing messages like:
>
> file /opt/xfstest2/960 has incorrect size - sync failed
>
> In our case the xfs filesystems used for testing are on top of lvm,
> which is in turn on top of two SAS drives in RAID1.
>
> Any suggestions on where I should start looking? I'm looking at the
> "do_sync() and XFSQA test 182 failures...." thread on the mailing list
> from 2008, can someone point me to the commit that eventually resulted
> from that?

I think I may have figured it out.

I assume this was part of the impetus for the patch series from Jan Kara 
starting with "[PATCH 0/8] Sync fixes and cleanups (version 4)" that 
went into 2.6.31.

Thanks,
Chris

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