Re: [RFC v2 5/9] vhost-vdpa: implement vhost-vdpa backend

[Cindy Lu <lulu@redhat.com>]

On Thu, May 21, 2020 at 8:40 PM Stefan Hajnoczi <stefanha@gmail.com> wrote:
>
> On Sat, May 09, 2020 at 12:32:14AM +0800, Cindy Lu wrote:
> > From: Tiwei Bie <tiwei.bie@intel.com>
> >
> > Currently we have 2 types of vhost backends in QEMU: vhost kernel and
> > vhost-user. The above patch provides a generic device for vDPA purpose,
> > this vDPA device exposes to user space a non-vendor-specific configuration
> > interface for setting up a vhost HW accelerator, this patch set introduces
> > a third vhost backend called vhost-vdpa based on the vDPA interface.
> >
> > Vhost-vdpa usage:
> >
> >   qemu-system-x86_64 -cpu host -enable-kvm \
> >     ......
> >   -netdev type=vhost-vdpa,vhostdev=/dev/vhost-vdpa-id,id=vhost-vdpa0 \
> >   -device virtio-net-pci,netdev=vhost-vdpa0,page-per-vq=on \
>
> I haven't looked at vDPA in depth. What is different here compared to
> the existing vhost-backend.c kernel backend?
>
> It seems to be making the same ioctls calls so I wonder if it makes
> sense to share the vhost-backend.c kernel code?
>
> Stefan
Hi Stefan,
Sorry for the late reply and Thanks for these suggestions.
I think the most difference between vhost kernel and vdpa is vdpa
depends on a real hardware.
The point is that vDPA devices work as a virtio device, but vhost-vdpa
qemu must present a vhost-like device in qemu vhost layer.
The ioctl  calls are similar  with vhost-backend.c now, but after more
and more NIC support vdpa. The difference between vhost-backend.c and
vpda will become more and more big. It will make the code complicated
to share the  code with kernel code.

Thanks
Cindy