Re: [PATCH 1/2] virtio-pci: add setup_vqs flag in vp_try_to_find_vqs

[Rusty Russell <rusty@rustcorp.com.au>]

On Wed, 1 Feb 2012 11:54:09 +0200, "Michael S. Tsirkin" <mst@redhat.com> wrote:
> On Thu, Jan 12, 2012 at 09:20:06AM +0800, zanghongyong@huawei.com wrote:
> > From: Hongyong Zang <zanghongyong@huawei.com>
> > 
> > changes in vp_try_to_find_vqs:
> > Virtio-serial's probe() calls it to request irqs and setup vqs of port0 and
> > controls; add_port() calls it to set up vqs of io_port.
> > it will not create virtqueue if the name is null.
> > 
> > Signed-off-by: Hongyong Zang <zanghongyong@huawei.com>
> 
> This looks like a fragile hack, to me.
> 
> virtio spec also implied that VQ initialization is done
> before status is set to OK (during probe) and
> devices might have relied on that. So if we want to change
> that, I think we need a feature bit.

If virtio_serial doesn't use those virtqueues, I think it's OK.

I think I'd rather have vp_find_vqs take an offset, however.

> Besides, what about documentation? non pci transports?

I think this is a device-specific issue, but re-calling find_vqs will
have to be audited.  Which adding an offset arg should lead to anyway.

Thanks,
Rusty.