Bug#695872: [squeeze] kernel panic in sunrpc module when using automount daemon

[Tom Downes <tpdownes@gravity.phys.uwm.edu>]

Jonathan:

I meant to include a URL to that particular patch. Thanks for pointing it out.

I know next-to-nothing about kernel programming at this level - should
I consider this a real fix in that a problem is actually resolved, or
should I deploy this patch on my systems knowing that it's more of a
band-aid?

Thanks for getting back quickly.

--
Tom Downes
Associate Scientist and Data Center Manager
Center for Gravitation, Cosmology and Astrophysics
University of Wisconsin-Milwaukee
414.229.2678


On Thu, Dec 13, 2012 at 1:34 PM, Jonathan Nieder <jrnieder@gmail.com> wrote:
> tags 695872 + upstream patch unreproducible
> quit
>
> Hi Tom,
>
> Tom Downes wrote:
>
>> While running a shell script that looped over a couple hundred
>> filesystems accessed via automoutning NFS exports (unmounted prior
>> to running the script), there was a kernel panic that prevented
>> further automounts from occurring.
>
> Thanks for reporting.
>
>> This problem seems very similar to bug #691764 filed under the
>> autofs package but I believe the bug is really in the sunrpc module.
>
> In that report the panic is in different code, so I think they are
> distinct.
>
>> I've found a patch described here, applied it to my system and have
>> not experienced a crash since. I only ever experienced one crash,
>> however, and unpatched systems are yet to crash.
>
> For reference:
>
>   v2.6.38-rc1~407^2~38 kernel panic when mount NFSv4, 2010-12-20
>
> Trond explains:
>
> | The panic place is:
> |   rpc_mkpipe
> |       __rpc_lookup_create()          <=== find pipefile *idmap*
> |       __rpc_mkpipe()                 <=== pipefile is *idmap*
> |         __rpc_create_common()
> |          ******  BUG_ON(!d_unhashed(dentry)); ****** *panic*
> |
> | The test is wrong: we can find ourselves with a hashed negative dentry here
> | if the idmapper tried to look up the file before we got round to creating
> | it.
>
> It seems like a match and I think you're right that the patch would
> fix it.  Thanks for finding it.
>
> Hope that helps,
> Jonathan