On Tue, Sep 01, 2015 at 12:41:49PM +0300, Michael S. Tsirkin wrote:
> On Tue, Sep 01, 2015 at 05:20:06PM +0800, Yuanhan Liu wrote:
> > On Mon, Aug 31, 2015 at 02:30:06PM +0300, Michael S. Tsirkin wrote:
> > [snip..]
> > > > Then how about the other case:
> > > > When startup the vhost backend, user specify 8 queues, like:
> > > > Vhost-user ... -queues 8 # Y=8
> > > > When startup the guest, user specify 4 queues, like:
> > > > Qemu-kvm ... -queues 4 # X=4
> > > >
> > > > Vhost backend CAN support the queues number as much as the guest/qemu
> > > > requires,
> > > > And qemu don't want enable all of them(8 in the case), just part of
> > > > them(only 4 from 8).
> > > > Will qemu exit or go forward(of course with 4 queues)?
> > >
> > > I think it must go forward since guest does not have to
> > > configure all queues anyway - and that's built in to virtio spec.
> > >
> > > Note that # of queues in actual use in fact changes dynamically
> > > too, without restaring the device.
> >
> > May I know how do you do that? I mean, is there a public interface, like
> > by some commands? Or, by sending vhost messages?
> >
> > --yliu
>
> Using ethtool within guest. In fact, guests boot with a single queue,
> you need to use ethtool to enable mq, specifying the # of queues.
Thanks for the info.
--yliu
