Xen project Mailing List

On Fri, Feb 17, 2012 at 4:14 PM, Paul S <pstroud@xxxxxxxxx> wrote:

Andrew,
Thanks for the quick response! I was finally able to capture it and get a bit more information. The problem does not always happen, even on 4.1.2, I was able to get it to boot a few times, though I was never able to get 4.2-unstable to boot. There was lots of strangeness, such as after I built and installed 4.1.1, it worked, but then, so did 4.1.2, now neither is working.

That being said, I caught it on video, and captured a picture of the final dump. I watched the video frame by frame and the boot actually got much further than I thought it was.

I can see that even though it does not bring up two of the CPUs, the boot continues into the kernel and dies somewhere with a:

BUG: Unable to handle paging request at FFFc7f(maybe).

I have captured several images from the video that shows the progression of the boot before the crash. I think I have gotten things badly out of whack at this point, so I'm going to stop, reinstall everything and move back to 4.1.1 and see if I can get things working again.

Let me know if anything in the pictures helps, or if you would like the video as well.

Paul

On Fri, Feb 17, 2012 at 11:39 AM, Andrew Cooper <andrew.cooper3@xxxxxxxxxx> wrote:

On 17/02/12 16:21, Paul S wrote:
> Environment:
> ASRock Z86 Extreme4 - Intel i5 2500
> http://www.asrock.com/mb/overview.asp?Model=Z68%20Extreme4
>
> I have installed the latest UEFI patch(1.70) and all tests were run
> from a reset of everything to default. I have tested with both Ubuntu
> 11.10 and Fedora 16 and get the same results on both, including
> building xen 4.2-unstable. I bought this particular MB because it had
> a know working vt-d implementation via VMWare.
>
> NOTE: current unstable as of last night is still affected by this
> library error http://www.gossamer-threads.com/lists/xen/devel/234300
>
> However, after correcting the library error, I was able to complete
> the unstable build.
>
> In both cases, it either stops with a panic/blank screen, or just reboots.
>
> Here are some other notes:
>
> 1) If x2apic is enabled on the motherboard, it almost immediately
> reboots with any of the configurations above

There are some fairly strict set of requirements on what which MSRs you
can wrt xAPIC/x2APIC mode. It is possible that we are tickling an MSR
early in boot which results in a protection fault.

>
> 2) The boot gets to here:
>
> (XEN) HVM: VMX enabled
>
> (XEN) HVM: Hardware Assisted Paging detected.
>
>
> Then goes to a "(XEN)Failed to bring up CPU x(error -5)" and this is
> where it hangs with the blank/panic screen or reboots. This sometimes
> shows up on CPU 1, CPU2, or both.
>
>
> 3) There is no boot log written and the machine does not have a native
> serial port, so capturing anything may prove difficult, but I am open
> to suggestions.
>

Try booting with noreboot - that should keep any panic message on screen
until you manually choose to reset the machine. Seeing where in the
panic occurs would be very helpful, even if you just transcribe a few
lines manually.

>
> 4) Xen Live(http://wiki.xen.org/wiki/LiveCD) works ok, unless x2apic
> is enabled, then it simply panics and reboots.
>
> The logs(xen_live.tgz) are attached. Note this was captured with VT-d
> being enabled, but I did check it with VT-d enabled and the boot
> looked the same. I can capture it again if needed.
>
>
> 5) Xen Server 6.0.0 works, with Xen 4.1.1, with both VT-d
> enabled(which shows working in xl) and x2apic enabled.
>
> I have also attached(xen_server_6.0.0.tar.gz) these logs, which
> includes logs both with and without VT-d and x2apic enabled. I did not
> actually create any VMs, only booted this and checked what xl reported.
>
>

That is rather interesting - I am not aware of any XenServer specific
changes in this regard.

> If there is anything else I can add, please let me know. I am going to
> download 4.1.1 now, build it and see if it works, because it does on
> Xen Server. I am open to testing anything as this is an original build
> on this box, so I'm willing to knock it around if it will help.
>

> #
>

Knowing whether upstream 4.1.1 works will be very useful in workout out
whether it is a regression introduced into unstable, or whether there is
some XenServer specific change which we should upstream.

> Thanks,
>
> Paul
>

--
Andrew Cooper - Dom0 Kernel Engineer, Citrix XenServer
T: +44 (0)1223 225 900, http://www.citrix.com

Re: [Xen-devel] Reboots and Panics(4.1.2/4.2-unstable)