Re: [Qemu-devel] [PATCH v2] net: Add support of VIRTIO_NET_F_GUEST_ANNOUNCE for vhost-net/vhost-user

["Michael S. Tsirkin" <mst@redhat.com>]

On Mon, Jun 08, 2015 at 12:01:38PM +0200, Thibaut Collet wrote:
> Hi,
> 
> I agree that my patch is OK only for recent drivers. To have a simple patch
> with only few modifications I do not implement a solution for old driver but I
> can be done later.
> 
> For legacy guest without VIRTIO_NET_F_GUEST_ANNOUNCE the problem is more
> complex. The RARP must be sent by the vhost client/backend. This component is
> outside QEMU (the reference implementation is snabbswitch: http://
> www.virtualopensystems.com/en/solutions/guides/snabbswitch-qemu/). To do that:
> - a receive function must be defined for the vhost user.
> - a message must be added between QEMU and vapp. This message will be sent only
> for old guest driver to avoid GARP duplication.
> - the added self_announce_disabled must be removed (decision to send or not the
> RARP is done later by the backend and not by the generic migration method)
> 
> Do you agree with this solution ?
> 
> 
> Regards.

I don't get it.  Why do you need any extra messages for old drivers?  To
detect old drivers, simply have backend check whether
VIRTIO_NET_F_GUEST_ANNOUNCE is set.

But I don't see this as a blocker for this patch,
this can be added separately as needed.


> On Mon, Jun 8, 2015 at 11:14 AM, Jason Wang <jasowang@redhat.com> wrote:
> 
>    
> 
>     On 06/08/2015 04:21 PM, Thibaut Collet wrote:
>     > Hi,
>     >
>     > My understanding of gratuitous packet with virtio for any backend
>     > (vhost user or other):
>     > - When the VM is loaded (first start or migration) the virtio net
>     > interfaces are loaded ( virtio_net_load_device function in
>     > hw/net/virtio-net.c)
>     > - If the guest has the VIRTIO_NET_F_GUEST_ANNOUNCE capability, request
>     > to send gratuitous packet is done.
>     >
>     > 1. To enable gratuitous packet through this mechanism I have
>     > added VIRTIO_NET_F_GUEST_ANNOUNCE capability to hw/net/vhost_net.c. So
>     > host and guest can negotiate this feature when vhost-user is used.
>     >
>     > 2. self announce occurs in case of live migration. During a live
>     > migration a GARP is sent to all net backend through a queue dedicated
>     > to the net backend.
>     >    But for vhost-user:
>     >        - this operation is not possible (vhost-user has no queue)
>     >        - it is already done with the previous mechanism.
>     >    Rather to define a queue to vhost user and notify twice the guest
>     > to send gratuitous packet I have disable GARP from self announce and
>     > use only the first mechanism for that.
>     >
>     > I have tested my modifications with guest that supports
>     > VIRTIO_NET_F_GUEST_ANNOUNCE and vhost-user on the host. After a live
>     > migration I have the GARP from the guest.
> 
>     Yes, your patch works well for recent drivers. But the problem is legacy
>     guest/driver without VIRTIO_NET_F_GUEST_ANNOUNCE. In this case there
>     will be no GARP sent after migration.
> 
>     Thanks
> 
>