On Mon, 2013-02-11 at 15:21 +0100, Sebastian Kutsch wrote:
> On 10.02.2013 01:31, Ben Hutchings wrote:
> > Control: tag -1 moreinfo
> >
> > On Fri, 2013-02-08 at 19:29 +0100, Sebastian Kutsch wrote:
> >> Package: src:linux Version: 3.2.35-2 Severity: normal
> >>
> >> Dear Maintainer,
> >>
> >> as expected, creating virtual function NICs with modprobe igb
> >> max_vfs=3 gave Virtual Function Interfaces - with a random
> >> MAC-address. Then I choosed a MAC with 'ip link set eth1 vf 0 mac
> >> 00:11:22:33:44:55'. When using those virtual network cards in
> >> qemu-kvm, there are new random MACs assigned to the virtual
> >> function NICs. This makes virtual function NICs useless.
> >>
> >> The problem isn't new, and should already be fixed according to
> >> this:
> >> http://comments.gmane.org/gmane.comp.emulators.kvm.devel/70839
> > [...]
> >
> > Which kernel version is running in the guest?
> I tried with GRML:
> Linux grml 3.3.0-3-grml-amd64 #1 SMP Fri May 25 23:30:37 UTC 2012
> x86_64 GNU/Linux
> Linux grml 3.3.0-3-grml-486 #1 Sat May 26 00:55:55 UTC 2012 i686 GNU/Linux
>
> > Are you invoking qemu-kvm directly, or using a higher-level tool
> > such as virsh or virt-manager?
> virt-manager
>
> Sorry I bothered you - I was following Intels howto, which didn't
> mention that you have to bring up the PF / ethX in the host. If you
> don't, everything looks fine, no visible error message. So I missed
> this line in dmesg:
> "The VF MAC address has been set, but the PF device is not up"
As it happens, my day job involves another net driver with SR-IOV
support, and we made it possible to configure VFs before the PF's net
device is brought up. But it's not that clear whether this should be
expected to work in general. If not, then I would think that libvirt
should bring up a PF's net device before using an associated VF for
guest networking.
libvirt maintainers, can you comment on this?
Ben.
--
Ben Hutchings
We get into the habit of living before acquiring the habit of thinking.
- Albert Camus
Attachment:
signature.asc
Description: This is a digitally signed message part