On Tue, Oct 26, 2010 at 03:35:38PM +0200, Bernhard Kohl wrote:
> Am 26.10.2010 15:07, schrieb ext Gleb Natapov:
> >On Tue, Oct 26, 2010 at 02:40:51PM +0200, Bernhard Kohl wrote:
> >>Am 26.10.2010 12:48, schrieb ext Gleb Natapov:
> >>>This is current sate of the patch series for people to comment on.
> >>>I dropped ioport double reservation checking from isa-bus and added
> >>>bus_id field for IDE bus since as Markus pointed out unit has different
> >>>meaning there.
> >>>
> >>>This patch series produce names like:
> >>>
> >>>i...@03f1-03f5,03f7/f...@a
> >>>i...@03f1-03f5,03f7/f...@b
> >>>p...@0000:00:01.1/i...@1:0
> >>>p...@0000:00:01.1/i...@1:1
> >>>p...@0000:00:03.0/virtio-...@0
> >>>p...@0000:00:04.0/virtio-...@0
> >>Are PCI pass-through devices (SR-IOV NICs) also supported?
> >Don't see any problem. They are regular PCI devices as seen by
> >guest, correct?
>
> Yes, that's true. And we add a bootrom on the qemu command line.
> But this is already working and would not be changed.
>
> >So I expect path to them will look like this:
> >p...@0000:00:04.0/n...@0
>
> Yes, that would be fine. I assume that each boot NIC,
> e.g. e1000 gets its name here.
>
Each boot nic sits on different PCI bus, so in fact nic (or
virtio-net) part is somewhat redundant. Each nic with specified
bootindex will be listed here.
> >>I assume that hw/device-assignment.c is missing from the changed files.
> >>Our systems highly depend on PXE booting from these devices.
> >>
> >The patch is only an example. I converted only virtio-net to support
> >bootindex ATM.
> >
> >--
> > Gleb.
> >
--
Gleb.
