Re: [Xen-devel] XCI: can we get to the demo state?

[Phung Te Ha <phungte@xxxxxxxxx>]

Do you pass the graphics controller through? Which guest are you using? Do you mind publish your config file?

Thank you.

Phung-Te

On Fri, Aug 14, 2009 at 11:08 AM, Ross Philipson <Ross.Philipson@xxxxxxxxxx> wrote:

Well we don’t have a official list we can push but we run on the same HP 6930p you have all the time. We have also had it running on Intel Q35 systems. So in general it should work on these systems.

 

Thanks

Ross

 

From: Phung Te Ha [mailto:phungte@xxxxxxxxx]
Sent: Friday, August 14, 2009 11:32 AM
To: Ross Philipson
Cc: Vincent Hanquez; xen-devel@xxxxxxxxxxxxxxxxxxx; Kamala Narasimhan

Subject: Re: [Xen-devel] XCI: can we get to the demo state?

 

Hello all,

The only thing I passed through was the graphics controller.
After starting the guest, the system is still there running, xenops list_domains shows 2 doms running, but nothing from the guest is showing.

The board I am using is the Intel DQ35OJ and the graphics controller is the onboard one. Same operation has also been done on a HP6930p, same result. Can you publish the list of hardware you guys are using, MB, graphics board..., the config that most likely to get a graphics controller passed through.

Thank you.

Phung-Te

On Fri, Aug 14, 2009 at 7:13 AM, Ross Philipson <Ross.Philipson@xxxxxxxxxx> wrote:

Yeah, these occur when the iommu context for the device in question (in this case your iGfx card) has been torn down but the card is still active (still attempting DMA). The code in xen to perform an function level resets (FLR) on the devices takes care of the vast majority of these but there still remain one or two that occur even thought the FLR has been done. This issue was never resolved but these few remaining faults don't seem to cause a problem. So you are experiencing a hang as the vm starts I take it. Is the vtd fault trace the last thing you see? Are you passing anything else through?

Thanks
Ross

-----Original Message-----
From: Vincent Hanquez
Sent: Friday, August 14, 2009 7:05 AM
To: Phung Te Ha

Cc: xen-devel@xxxxxxxxxxxxxxxxxxx; Ross Philipson; Kamala Narasimhan
Subject: Re: [Xen-devel] XCI: can we get to the demo state?

Phung Te Ha wrote:
> Hi Vincent,
>
> I tried your suggestion, adding pci = 0,bind,0000:00:02.0 in the config
> on 2 different machines, the HP6930 and on an Intel MB DQ35OJ.
>
> On both machines, I got the samething, the guest run as shown by
> xenops list_domains:
>
>  id |  state |    cpu_time |                                 uuid
>  0 |     R  | 54203470658 | 00000000-0000-0000-0000-000000000000
>  1 |    B H | 12214238864 | 00000000-0000-0000-0000-000000000002
>
> But nothing shows on the screen.
>
> I checked the logs and found in xenops dmesg:
>
> ...
> (XEN) Intel VT-d Snoop Control not supported.
> (XEN) Intel VT-d DMA Passthrough not supported.
> (XEN) Intel VT-d Queued Invalidation not supported.
> (XEN) Intel VT-d Interrupt Remapping not supported.
> (XEN) I/O virtualisation enabled
> (XEN) I/O virtualisation for PV guests disabled
>
>
> And when the guest is started:
>
> ...
>  _fault:DMA Write: 0:2.0 addr faddf7000 REASON 5 iommu->reg =
> ffff828bfff56000
> (XEN) print_vtd_entries: iommu = ffff83007dcfb320 bdf = 0:2:0 gmfn =
> faddf7
> (XEN)     root_entry = ffff83007e0ce000
> (XEN)     root_entry[0] = 7e066001
> (XEN)     context = ffff83007e066000
> (XEN)     context[10] = 201_7da02001
> (XEN)     l3 = ffff83007da02000
> (XEN)     l3_index = 3e
> (XEN)     l3[3e] = 0
> (XEN)     l3[3e] not present
> (XEN) [VT-D]iommu.c:722: iommu_page_fault: iommu->reg = ffff828bfff56000
> (XEN) [VT-D]iommu.c:691: iommu_fault_status: Fault Overflow
> (XEN) [VT-D]iommu.c:694: iommu_fault_status: Primary Pending Fault
> (XEN) [VT-D]iommu.c:676: iommu_fault:DMA Write: 0:2.0 addr faddf7000
> REASON 5 iommu->reg = ffff828bfff56000
> (XEN) print_vtd_entries: iommu = ffff83007dcfb320 bdf = 0:2:0 gmfn =
> faddf7
> (XEN)     root_entry = ffff83007e0ce000
> (XEN)     root_entry[0] = 7e066001
> (XEN)     context = ffff83007e066000
> ...
>
> Do you think these 2 machines have enough of VT-d features to get the
> graphics passed-through?
> Do you have a preferred desktop board, known to work for graphisc
> passed-through?
I'm not really familiar with thoses VT-d errors. Kamala/Ross do you have
any idea, what this previous errors means ?

thanks,
--
Vincent

 

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