Re: XFS causes Kernel trace with Jeremy's pv_ops kernel

[Ulf Kreutzberg <uk@hosteurope.de>]

Hi Jeremy,


thank you for your answer.
I wonder if it is a then a problem with the git kernel.
It seems only to occur if it is running under Xen (4). If I boot the
kernel on bare metal, anythig works fine. As an additional info, under
kernel 2.6.27.5 (formerly taken from git) under xen 3.4.0 (which is then
not a pv_ops kernel) all works fine, too.
So my question would be, how to get the issue solved? Do you have any Ideas?

Best regards,
Ulf


Am 07.10.2010 23:53, schrieb Jeremy Fitzhardinge:
>  On 10/07/2010 03:14 AM, Ulf Kreutzberg wrote:
>> Hi Jeremy,
>> hi all,
>>
>>
>> I have checked out Jeremy's last git built of dom0 pv_ops kernel 2.6.32.23:
>>
>> commit 41a85de5caef68bbd58e11ed7b781b7867d96eaa
>> Merge: 62bf426... 0ee0f94...
>> Author: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
>> Date:   Tue Sep 28 11:20:40 2010 -0700
>>
>> I have set up on a Dell PE1950 2xDualXeon 5130 an LVM2 partition
>> containing XFS. I have built Jeremy's kernel successfully, so the
>> running dom0 Kernel is the one mentioned above.
>> System is Debian-Lenny, Xen and Xen-tools are self-built.
>>
>> If I make some I/O, eg:
>> cd linux-2.6-xen-git (which is then on the lv with the XFS filesystem)
>> make -j 7 ; make install; make -j 7 modules; make modules_install
>>
>> in dmesg appears a kernel trace after a while. But compiling continues
>> and all keeps up and running. This only happens if the kernel is running
>> under xen 4.0.2-rc1-pre hypervisor. Nothing strange happens if I run it
>> on bare-metal (same machine) without hypervisor.
>> I have not tested bonnie or iozone yet.
>>
>> I have attached booting output of xen 4.0.2-rc1-pre
>> as well as the trace (file console-kerneltrace.txt).
>> Kernel config is also attached (config-2.6.32.23.txt)
> 
> This doesn't look obviously Xen specific.  At least at one point XFS was
> prone to spurious lockdep warnings that were hard to eliminate.  I
> wonder if its possible you're seeing this under Xen because of some kind
> of timing issue?
> 
> Of course it could be Xen related too, but as I said, nothing stands out.
> 
>     J
> 
>> Perhaps you can help. Please let me know if I have forgotten any
>> important information.
>>
>> Many thanks and best regards,
>> Ulf Kreutzberg
>>
>>
>> Viele Grüße,
>> Ulf
>>
>>
>>
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@lists.xensource.com
>> http://lists.xensource.com/xen-devel
> 
>