Re: [Xen-devel] [TestDay] Xen crash on boot

[Eric Houby <ehouby@xxxxxxxxx>]

On Wed, 2014-01-22 at 08:30 +0000, Jan Beulich wrote:
> >>> On 21.01.14 at 11:39, "Eric Houby" <ehouby@xxxxxxxxx> wrote:
> > Xen console logs are attached.
> 
> Thanks. This
> 
> (XEN) ..MP-BIOS bug: 8254 timer not connected to IO-APIC
> (XEN) ...trying to set up timer (IRQ0) through the 8259A ...  failed.
> (XEN) ...trying to set up timer as Virtual Wire IRQ... works.
> 
> is the first thing you need to deal with. You may want to check
> how a native kernel handles that (they have some DMI based
> stuff in place to automatically deal with broken firmware, which
> on Xen requires command line overrides like
> "acpi_skip_timer_override").
> 
> That alone may already explain
> 
> (XEN) AMD-Vi: IO_PAGE_FAULT: domain = 0, device id = 0xa0, fault address = 
> 0xfdf8010140, flags = 0x8
> 
> since these come from the HPET (which may be generating legacy
> timer interrupts):
> 
> (XEN) AMD-Vi: IVHD Special: 0000:00:14.0 variety 0x2 handle 0
> 
> The final crash could easily be a consequence of the earlier
> problems. In any event you ought to make sure you run up to
> date firmware on your system, as there clearly are issues with it.
> 
> What would also be interesting to know is whether any older Xen
> version ever booted okay on that system.
> 
> Jan
> 

Jan,

The system I am testing with is a GA-890FXA-UD5 and it has been doing
secondary VGA passthrough to Win7/8.1 guests since mid 2010.  I believe
it was XSA-36 that was not friendly to the board and I have not been
able to get this functionality since xen 4.1.3.  In mid 2010, when I
last tested available BIOSs, I only found one version to support IOMMU
with Xen.  Since then, one more BIOS version has been released.  It may
be worth loading but my confidence in it fixing this issue is low.

Adding acpi_skip_timer_override to the xen command line did allow Dom0
to boot but the video display for Dom0 did not work.  The following log
was still seen the logs.

(XEN) AMD-Vi: IO_PAGE_FAULT: domain = 0, device id = 0xa0, fault address
= 0xfdf8020140, flags = 0x8


Including iommu=no-amd-iommu-perdev-intremap along with
acpi_skip_timer_override cleared the above errors and Dom0 video was
functional.  Attempts to passthrough a secondary VGA to a Win8.1 guest
was not successful with these logs in the xen console:

(XEN) AMD-Vi: IO_PAGE_FAULT: domain = 1, device id = 0x700, fault
address = 0x11620df40, flags = 0

Attached are two xen console boot logs, one with just
acpi_skip_timer_override in the xen command line and the other with
acpi_skip_timer_override and iommu=no-amd-iommu-perdev-intremap.  The
long files names I used make it clear which is which.  The log with both
xen command options also includes booting the Win8.1 guest with
secondary VGA passthrough.

I tried to see how a native kernel handled the timer bug you mentioned
but I could not find a corresponding log when booting 3.12.7 on bare
hardware. This boot log is attached as barehw.txt, maybe there is
something you can find.

Thanks,

Eric

Attachment: barehw.txt
Description: Text document

Attachment: xen console acpi-skip.txt
Description: Text document

Attachment: xen console acpi-skip no-amd-iommu.txt
Description: Text document

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel