On Sun, Nov 14, 2010 at 8:56 AM, Sander Eikelenboom
> Hmmm have you tried do do a lot of I/O with something else as NFS ?
> That would perhaps pinpoint it to NFS doing something not completely
> compatible with Xen.
I have my own suspicions regarding the more recent NFS clients. Post
10.04 Ubuntu variants do not tolerate large NFS transfers even without
Xen. Any more than a few 100 Megs and you start getting 'task blocked
for more than 120 sec..." messages along with stack traces showing
part of the NFS call stack.
Perhaps a parallel effort could be to test the 2.6.37-rc1 kernel with
something other than NFS for remote filesystems. I'll see if I get the
same problems with glusterfs.
> I'm not using NFS (I still use file: based guests, and i use glusterfs (fuse
> based userspace cluster fs) to share diskspace to domU's via ethernet).
> I tried NFS in the past, but had some troubles setting it up, and even more
> problems with disconnects.
> I haven't seen any "unable to handle page request" problems with my mix of
> guest kernels, which includes some 2.6.37-rc1 kernels.
> Sunday, November 14, 2010, 5:37:59 PM, you wrote:
>> I've tested F14 DomU (kernel
>> vmlinuz-2.6.37-0.1.rc1.git8.xendom0.fc14.x86_64) as NFS client and Xen 4.0.1
>> F14 Dom0 (kernel vmlinuz-22.214.171.124-172.xendom0.fc14.x86_64) as NFS server .
>> Copied 700 MB ISO images from NFS folder at Dom0 to DomU and scp'ed them
>> back to Dom0. During about 30 - 40 min DomU ran pretty stable , regardless
>> kernel crash as "unable to handle page request" was reported once by F14
>> DomU, but it didn't actually crash DomU. Same excersises with replacement
>> F14 by Ubuntu 10.04 Server results DomU crash in about several minutes.
>> Dom0's instances dual boot on same development box ( Q9500,ASUS P5Q3,8GB)
>> --- On Fri, 11/12/10, Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx> wrote:
>> From: Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>
>> Subject: Re: [Xen-devel] Re: 2.6.37-rc1 mainline domU - BUG: unable to
>> handle kernel paging request
>> To: "Sander Eikelenboom" <linux@xxxxxxxxxxxxxx>
>> Cc: "Boris Derzhavets" <bderzhavets@xxxxxxxxx>,
>> xen-devel@xxxxxxxxxxxxxxxxxxx, "Bruce Edge" <bruce.edge@xxxxxxxxx>, "Jeremy
>> Fitzhardinge" <jeremy@xxxxxxxx>
>> Date: Friday, November 12, 2010, 12:01 PM
>> On Fri, Nov 12, 2010 at 05:27:43PM +0100, Sander Eikelenboom wrote:
>>> Hi Bruce,
>>> Perhaps handpick some kernels before and after the pulls of the xen patches
>>> (pv-on-hvm etc) to begin with ?
>>> When you let git choose, especially with rc-1 kernels, you will end up with
>>> kernels in between patch series, resulting in panics.
>> Well, just the bare-bone boot of PV guests with nothing fancy ought to work.
>> But that is the theory and ..
>>> > The git bisecting is slow going. I've never tried that before and I'm a
>>> > git
>>> > rookie.
>>> > I picked 2.6.36 - 2.6.37-rc1 as the bisect range and my first 2 bisects
>>> > all
>>> > panic at boot so I'm obviously doing something wrong.
>>> > I'll RTFM a bit more and keep at it.
>> .. as Bruce experiences this is not the case. Hmm..
>> Xen-devel mailing list
> Best regards,
> Sander mailto:linux@xxxxxxxxxxxxxx
Xen-devel mailing list