Re: PV passthrough of sibling igbvf's

[Laszlo Ersek <lersek@redhat.com>]

Hi,

On 10/16/12 16:42, Andrew Cooper wrote:

> I would hazard a guess that the real bug is trying to fake up 8
> individual virtual functions as an 8-fuction device, which seems like a
> toolstack bug to me.

I believe that comes from:

>> The VPCI
>> implementation of pciback_add_pci_dev() [drivers/xen/pciback/vpci.c]
>> will assign these sibling functions to the same virtual slot. In other
>> words, VFs that are siblings in dom0 end up as siblings in the PV domU.

This grouping-together of virtual functions is the same in current
upstream Linux:

>> (Upstream path and function: "drivers/xen/xen-pciback/vpci.c",
>> __xen_pcibk_add_pci_dev().)

(+ match_slot())

>>
>> This logic appears to date back to
>> <http://xenbits.xensource.com/xen-unstable.hg/rev/5b433b4fca34>.

A closer pointer into the changeset:

http://xenbits.xensource.com/xen-unstable.hg/rev/5b433b4fca34#l16.85

The patch doesn't seem to justify the grouping specifically, thus I did
not even try to refute it.

Now that you point it out, match_slot() is probably insufficient grounds
to group functions together. Maybe we should check *additionally* if the
device being passed through is multi-function. I'll try it.

Thanks!
Laszlo