[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] vmx status report against cset 14912 - 1 fixed issue, 4 old issues


  • To: "Zhao, Yunfeng" <yunfeng.zhao@xxxxxxxxx>, <xen-devel@xxxxxxxxxxxxxxxxxxx>
  • From: Keir Fraser <keir@xxxxxxxxxxxxx>
  • Date: Wed, 25 Apr 2007 10:34:15 +0100
  • Delivery-date: Wed, 25 Apr 2007 02:33:04 -0700
  • List-id: Xen developer discussion <xen-devel.lists.xensource.com>
  • Thread-index: AceG3J3OlwL2jeGuSIK4ido1hBOBCQAOIa5QAACbto8AAKc0AAAArNhf
  • Thread-topic: [Xen-devel] vmx status report against cset 14912 - 1 fixed issue, 4 old issues

On 25/4/07 10:27, "Zhao, Yunfeng" <yunfeng.zhao@xxxxxxxxx> wrote:

>> To be clear: this is an unmodified RHEL4 release kernel, with *latest*
> PV
>> drivers from -unstable? Any idea if this issue is specific to the RHEL4
>> kernel, or can you reproduce with other kernels? Does it very
> definitely
>> seem to be specific to multiprocessor guests?
> [Yunfeng] The guest was booted with a 2.16.16 smp kernel which we built
> for PV test. It only happens on SMP HVM linux guest.I also saw the same
> issue on some SLES SMP HVM guests with SLES release kernel and PV
> drivers.

It would be interesting to see what happens if the platform-pci interrupt is
pinned strictly to vcpu0. So, what happens if you boot the guest with
'acpi=0'? Or, what about if you kill any irqbalance programs that are
running after boot, then 'cat /proc/interrupts' to find the xen-platform-pci
interrupt number (x) and then 'echo 1 >/proc/irq/<x>/smp_affinity'?

I couldn't repro just with SpecJBB but it's quite possible the kernel build
is really required to repro. Specjbb itself does very little I/O.

 -- Keir


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.