Re: [Qemu-devel] [Xen-devel] [PATCH][XSA-126] xen: limit guest control of PCI command register

[Jan Beulich]

>>> On 08.06.15 at 11:30, <address@hidden> wrote:
> On Mon, Jun 08, 2015 at 08:42:57AM +0100, Jan Beulich wrote:
>> Otoh the respective root port also has
>> - Received Master Abort set in its Secondary Status register (but
>>   that's also already the case in the log that we have before the UR
>>   occurs, i.e. that doesn't mean all that much),
>> - Received System Error set in its Secondary Status register (and
>>   after the UR the sibling endpoint [UR originating from 83:00.0,
>>   sibling being 83:00.1] also shows Signaled System Error set).
> 
> It's another function of the same physical device, correct?

Yes (obviously with their BDF only differing in the function).

> So is this sibling the only function sending SERR?

Yes, albeit I can't tell whether the root port generated SERR on
its own or in response to the endpoint doing so.

> What happens if you disable SERR# in the command register
> of 83:00.1?

Any experiments with that system will be quite difficult, as they're
only accessible by partners or ours. But I'll ask to try this if you
think this can provide useful insight.

>> > Do we can chalk this up to hardware bugs on a specific box?
>> 
>> I have to admit that I'm still very uncertain whether to consider all
>> this correct behavior, a firmware flaw, or a hardware bug.
> 
> Questions:
> 1.  Does this only happen with a specific endpoint?
>     What if you add another endpoint to the same system?

We've got reports of this for two systems (two different vendors)
using a Broadcomm NIC in one case and an Intel one in the other.

> 2.  Has a driver initialized this endpoint? What if you don't
>     load a driver before sending the transaction resulting in the UR?

I'd have to ask for this to be tried too, and getting an answer
may take some time.

Jan