According to Jans suggestion, I did a second build with xen up to the
changeset he suggested and attached another round of logs.

Also, to answer Yangs questions, the devices 00:12.0 and 00:13.0 are
regular noboard USB controller from the mainboard:

00:12.0 USB controller: Advanced Micro Devices, Inc. [AMD/ATI]
SB7x0/SB8x0/SB9x0 USB OHCI0 Controller
00:12.2 USB controller: Advanced Micro Devices, Inc. [AMD/ATI]
SB7x0/SB8x0/SB9x0 USB EHCI Controller
00:13.0 USB controller: Advanced Micro Devices, Inc. [AMD/ATI]
SB7x0/SB8x0/SB9x0 USB OHCI0 Controller
00:13.2 USB controller: Advanced Micro Devices, Inc. [AMD/ATI]
SB7x0/SB8x0/SB9x0 USB EHCI Controller

But I did do a second run without them passed through the domU and added
the log (xldmesg_secondrun.log). Since I'm seeing IO_PAGE_FAULTs there
(05:00.0 is the vga i'm passing), I would assume that basically any device
I pass through causes these page faults and it has nothing to do with
specific faulty devices.


2014-06-16 10:08 GMT+02:00 Jan Beulich <JBeulich@suse.com>:

> >>> On 16.06.14 at 09:47, <yang.z.zhang@intel.com> wrote:
> > Matthias wrote on 2014-06-12:
> >> domain = 0, device id = 0x98, fault address = 0xffffffc0, flags = 0
> >> (XEN) AMD-Vi: IO_PAGE_FAULT: domain = 0, device id = 0x90, fault
> >> address = 0xffffffc0, flags = 0
> >> (XEN) AMD-Vi: IO_PAGE_FAULT: domain = 0, device id = 0x98, fault
> >> address = 0xffffffc0, flags = 0
> >> (XEN) AMD-Vi: IO_PAGE_FAULT: domain = 0, device id = 0x90, fault
> >> address = 0xffffffc0, flags = 0
> >> (XEN) AMD-Vi: IO_PAGE_FAULT: domain = 0, device id = 0x98, fault
> >> address = 0xffffffc0, flags = 0
> >> (XEN) AMD-Vi: IO_PAGE_FAULT: domain = 0, device id = 0x90, fault
> >> address = 0xffffffc0, flags = 0
> >> (XEN) AMD-Vi: IO_PAGE_FAULT: domain = 0, device id = 0x98, fault
> >> address = 0xffffffc0, flags = 0
> >> (XEN) AMD-Vi: IO_PAGE_FAULT: domain = 0, device id = 0x90, fault
> >> address = 0xffffffc0, flags = 0
> >
> > This is really strange. The fault shows it is caused by dom0's activity.
> But
> > the device have assigned to the guest. Are you sure the devices(00:13.0
> and
> > 00:12.0) are hidden and dom0 is not using it? BTW, what are the devices
> of
> > 00:12.0 and 00:13.0?
>
> It's not just that which is odd. The fault address is suspicious too;
> I only ever saw such along with discovering what led to 6b4d71d0
> ("AMD IOMMU: don't free page table prematurely". Matthias - it
> certainly wouldn't hurt if you included that fix in whatever you build
> your hypervisor from (backports are already available in older trees).
>
> Jan
>
> Jan
>
>