oopses in xfs_attr_shortform_* on kerneloops.org

oopses in xfs_attr_shortform_* on kerneloops.org

[Eric Sandeen]

I noticed this on kerneloops.org and pinged Christoph about it but I'll
go ahead & point it out here & file a bug so it doesn't get lost:

http://www.kerneloops.org/search.php?search=xfs_attr_shortform&btnG=Function+Search

would seem to indicate a problem, though not a (recent) regression anyway.

Many traces like:

RIP: 0010:[<ffffffff8802e346>]  [<ffffffff8802e346>]
:xfs:xfs_attr_shortform_getvalue+0x27/0xf3
...
Call Trace:
 [<ffffffff8802ce07>] :xfs:xfs_attr_fetch+0xba/0x114
 [<ffffffff810cc227>] ? inotify_d_instantiate+0x1a/0x42
 [<ffffffff8802cf12>] :xfs:xfs_attr_get+0xb1/0xd1
 [<ffffffff8802cf50>] :xfs:attr_generic_get+0x1e/0x2e
 [<ffffffff880713fb>] :xfs:xfs_vn_getxattr+0x86/0x95
 [<ffffffff810be6ff>] vfs_getxattr+0xa1/0xb4
 [<ffffffff810be7ae>] getxattr+0x9c/0xfb
 [<ffffffff810b9f8a>] ? mntput_no_expire+0x20/0x85
 [<ffffffff810ab1c5>] ? path_put+0x2c/0x30
 [<ffffffff810adc45>] ? path_walk+0xb6/0xc3
 [<ffffffff810adfee>] ? do_path_lookup+0x1d6/0x231
 [<ffffffff810acb45>] ? putname+0x30/0x39
 [<ffffffff810aea41>] ? __user_walk_fd+0x4c/0x5c
 [<ffffffff810be8e2>] sys_lgetxattr+0x53/0x73
 [<ffffffff810a7343>] ? sys_newlstat+0x31/0x3c
 [<ffffffff8100bedb>] system_call_after_swapgs+0x7b/0x80

-Eric

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

Re: oopses in xfs_attr_shortform_* on kerneloops.org

[Christoph Hellwig]

On Thu, Jan 01, 2009 at 10:01:36PM -0600, Eric Sandeen wrote:
> I noticed this on kerneloops.org and pinged Christoph about it but I'll
> go ahead & point it out here & file a bug so it doesn't get lost:
> 
> http://www.kerneloops.org/search.php?search=xfs_attr_shortform&btnG=Function+Search
> 
> would seem to indicate a problem, though not a (recent) regression anyway.

Yeah, I've been looking into and the disassemly looks like we have
a NULL attribut fork pointer for some reason.  I did some poking
around in the code and couldn't a cause for it.  So if anyone of
the kerneloops submitters actually is on the list and can reproduce
it that would be extremly helpful.

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

Re: oopses in xfs_attr_shortform_* on kerneloops.org

[Eric Sandeen]

Christoph Hellwig wrote:
> On Thu, Jan 01, 2009 at 10:01:36PM -0600, Eric Sandeen wrote:
>> I noticed this on kerneloops.org and pinged Christoph about it but I'll
>> go ahead & point it out here & file a bug so it doesn't get lost:
>>
>> http://www.kerneloops.org/search.php?search=xfs_attr_shortform&btnG=Function+Search
>>
>> would seem to indicate a problem, though not a (recent) regression anyway.
> 
> Yeah, I've been looking into and the disassemly looks like we have
> a NULL attribut fork pointer for some reason.  I did some poking
> around in the code and couldn't a cause for it.  So if anyone of
> the kerneloops submitters actually is on the list and can reproduce
> it that would be extremly helpful.

This *might* be a clue:

http://article.gmane.org/gmane.comp.file-systems.xfs.general/22156

one of those oopses, and associated xfs_repair output for the fs in
question... doesn't make anything clearer to me, though, offhand.

-Eric

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