Re: HVM driver domains do not appear to be usable with stubdomains

[Chris Brannon <cmb@prgmr.com>]

Christopher Clark <christopher.w.clark@gmail.com> writes:

> On Tue, Dec 4, 2018 at 10:11 AM Chris Brannon <cmb@prgmr.com> wrote:
>>
>> Hi,
>> I set up a network driver domain for a dom0; it uses HVM
>> virtualization.  It worked very well when not using a device model
>> stubdomain, but when I requested the use of a device model stubdomain in
>> my xl.cfg file, the domU refused to boot.  It gave the following error
>> message.
>>
>> [76594.195404] xen_pciback: vpci: 0000:05:00.0: assign to virtual slot 0
>> [76594.195927] pciback 0000:05:00.0: registering for 35
>> [76594.230592] xen-pciback pci-34-0: 22 Couldn't locate PCI device
>>     (0000:05:00.0)! perhaps already in-use?
>>
>> It seems to me that Xen is trying to pass the PCI device to both the
>> domU and the stubdomain, and that is why it is failing.  Has anyone
>> managed to get an HVM driver domain running with a stubdomain?  If so,
>> how did you do it?  Or is it expected that HVM driver domains won't work
>> with stubdomains?
>
> OpenXT uses a HVM driver domain with a stubdomain for the network
> device driver, using Linux stubdomains with code very similar to that
> Marek has posted on this list.
>
> These pull requests may be a good place to start looking.
> https://github.com/OpenXT/openxt/pull/311
> and
> https://github.com/OpenXT/xenclient-oe/pull/912
> which includes this patch to qemu (originally from Qubes, referencing:

Hi Christopher,
Thank you for the pointers.  That gives me a very good place to start.
So if I understand correctly, the problem is that PCI passthrough
doesn't work with stubdomains, unless qemu is patched?

-- Chris

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel