Re: Re: 2.6.37-rc1 mainline domU - BUG: unable to handle kernel paging request

[Bruce Edge <bruce.edge@gmail.com>]

On Thu, Dec 2, 2010 at 10:28 AM, Jeremy Fitzhardinge <jeremy@goop.org> wrote:
> On 12/02/2010 10:05 AM, Bruce Edge wrote:
>> On Thu, Dec 2, 2010 at 6:41 AM, Bruce Edge <bruce.edge@gmail.com> wrote:
>>>
>>> On Wed, Dec 1, 2010 at 10:33 PM, Jeremy Fitzhardinge <jeremy@goop.org>
>>> wrote:
>>>> On 12/01/2010 01:32 PM, Bruce Edge wrote:
>>>>> I just checked the recently released 2.6.73-rc4,
>>>> Do you mean mainline 2.6.37-rc4, or the one in xen/next-2.6.37?
>>> mainline
>>>
>>>>>  and while the BUG
>>>>> signature is different, it still fails under NFS accesses. This is
>>>>> 100% recreatable.
>>>> Please try this patch which is queued up for mainline.  It's already in
>>>> xen/next-2.6.37.
>>> After testing the mainline .37 I also tried xen/next-2.6.37 with the same
>>> result. At least, I still hit a BUG after accessing an NFS mount. I didn't
>>> compare the stack, but the BUG condition looked the same.
>>> I'll reconfirm the location and traceback of the xen/next kernel.
>>>
>>> -Bruce
>>>
>>>
>> I re-ran the test with xen/next-2.6.37. The fault was a bit different this time.
>> The dom0 started spewing these as the domU appeared to hang:
>
> OK, these look like exactly the sort of symptoms which should be fixed
> by that patch, so assuming you're actually running the right kernel
> (worth checking!  I get bitten by this occasionally), there's something
> else going on here.

You're right, my branch was a day out of date. I didn't have the last
changes to:

 arch/x86/xen/grant-table.c |   28 +++++---
 arch/x86/xen/mmu.c         |  124 ++++++++++--------------------
 include/linux/mm.h         |    6 ++
 include/linux/vmalloc.h    |    2 -
 mm/memory.c                |   56 ++++++++++----
 mm/vmalloc.c               |  182 ++++++++++++++------------------------------

rebuilding with great expectations now :-)

-Bruce

>
> What NFS version are you using, and other parameters?  What is the NFS
> server?  What's the file access pattern going on?
>
> Thanks,
>    J
>