Re: [Qemu-devel] [PATCH v3 1/2] vhost user: add support of live migratio

On Wed, Jun 10, 2015 at 4:27 PM, Michael S. Tsirkin <address@hidden> wrote:

On Wed, Jun 10, 2015 at 04:22:10PM +0200, Thibaut Collet wrote:
> I am not sure to understand your comments.
>
> Could you confirm that the useless warning is:
> + fprintf(stderr,
> + "Warning: Guest with no VIRTIO_NET_F_GUEST_ANNOUNCE support.
> RARP must be sent by vhost-user backend\n");

Yes.

> > Can you move this out to vhost-user? It's not a generic function, is it?
>
> Which part of the code do you think it must be moved in a specific file ?

vhost_net_inject_rarp is vhost-user specific.

>
> On Wed, Jun 10, 2015 at 3:52 PM, Michael S. Tsirkin <address@hidden> wrote:
>
> On Wed, Jun 10, 2015 at 03:43:02PM +0200, Thibaut Collet wrote:
> > Some vhost client/backend are able to support live migration.
> > To provide this service the following features must be added:
> > 1. GARP from guest after live migration:
> > the VIRTIO_NET_F_GUEST_ANNOUNCE capability is added to vhost-net when
> netdev
> > backend is vhost-user to let virtio-net NIC manages GARP after
> migration.
> > 2. RARP on vhost-user:
> > After a live migration RARP are automatically sent to any interfaces.
> A
> > receive callback is added to vhost-user to catch this packet. If guest
> > supports VIRTIO_NET_F_GUEST_ANNOUNCE this packet is discarded to avoid
> > message duplication (already done by virtio-net NIC). For legacy guest
> > without VIRTIO_NET_F_GUEST_ANNOUNCE a warning message is displayed to
> > alert the user. RARP must be sent to the vhost client/backend.
> >
> > Signed-off-by: Thibaut Collet <address@hidden>
> > ---
> > hw/net/vhost_net.c | 15 +++++++++++++++
> > include/net/vhost_net.h | 1 +
> > net/vhost-user.c | 21 +++++++++++++++++++--
> > 3 files changed, 35 insertions(+), 2 deletions(-)
> >
> > diff --git a/hw/net/vhost_net.c b/hw/net/vhost_net.c
> > index 426b23e..4a42325 100644
> > --- a/hw/net/vhost_net.c
> > +++ b/hw/net/vhost_net.c
> > @@ -82,6 +82,8 @@ static const int user_feature_bits[] = {
> > VIRTIO_NET_F_CTRL_MAC_ADDR,
> > VIRTIO_NET_F_CTRL_GUEST_OFFLOADS,
> >
> > + VIRTIO_NET_F_GUEST_ANNOUNCE,
> > +
> > VIRTIO_NET_F_MQ,
> >
> > VHOST_INVALID_FEATURE_BIT
> > @@ -365,6 +367,15 @@ void vhost_net_stop(VirtIODevice *dev,
> NetClientState *ncs,
> > assert(r >= 0);
> > }
> >
> > +void vhost_net_inject_rarp(struct vhost_net *net, const uint8_t *buf,
> size_t size)
> > +{
> > + if ((net->dev.acked_features & (1 << VIRTIO_NET_F_GUEST_ANNOUNCE)) =
> = 0) {
> > + fprintf(stderr,
> > + "Warning: Guest with no VIRTIO_NET_F_GUEST_ANNOUNCE
> support. RARP must be sent by vhost-user backend\n");
> > + fflush(stderr);
>
> And maybe it does, then you are just filling log up with useless
> warnings. Pls add some ifdef so this isn't normally compiled in.
>
> > + }
>
> Can you move this out to vhost-user? It's not a generic function, is it?
>
> > +}
> > +
> > void vhost_net_cleanup(struct vhost_net *net)
> > {
> > vhost_dev_cleanup(&net->dev);
> > @@ -427,6 +438,10 @@ void vhost_net_stop(VirtIODevice *dev,
> > {
> > }
> >
> > +void vhost_net_inject_rarp(struct vhost_net *net, const uint8_t *buf,
> size_t size);
> > +{
> > +}
> > +
> > void vhost_net_cleanup(struct vhost_net *net)
> > {
> > }
> > diff --git a/include/net/vhost_net.h b/include/net/vhost_net.h
> > index b1c18a3..e82a0f9 100644
> > --- a/include/net/vhost_net.h
> > +++ b/include/net/vhost_net.h
> > @@ -20,6 +20,7 @@ bool vhost_net_query(VHostNetState *net, VirtIODevice
> *dev);
> > int vhost_net_start(VirtIODevice *dev, NetClientState *ncs, int
> total_queues);
> > void vhost_net_stop(VirtIODevice *dev, NetClientState *ncs, int
> total_queues);
> >
> > +void vhost_net_inject_rarp(VHostNetState *net, const uint8_t *buf,
> size_t size);
> > void vhost_net_cleanup(VHostNetState *net);
> >
> > unsigned vhost_net_get_features(VHostNetState *net, unsigned features);
> > diff --git a/net/vhost-user.c b/net/vhost-user.c
> > index 8d26728..69c2313 100644
> > --- a/net/vhost-user.c
> > +++ b/net/vhost-user.c
> > @@ -66,6 +66,24 @@ static void vhost_user_stop(VhostUserState *s)
> > s->vhost_net = 0;
> > }
> >
> > +static ssize_t vhost_user_receive(NetClientState *nc, const uint8_t
> *buf,
> > + size_t size)
> > +{
> > + if (size == 60) {
> > + /* Assume it is a RARP request sent automatically after a
> > + * live migration */
> > + /* The RARP must be sent if guest does not support
> > + * VIRTIO_NET_F_GUEST_ANNOUNCE */
> > + VhostUserState *s = DO_UPCAST(VhostUserState, nc, nc);
> > +
> > + vhost_net_inject_rarp(s->vhost_net, buf, size);
> > + } else {
> > + fprintf(stderr,"Vhost user receives unexpected packets\n");
> > + fflush(stderr);
> > + }
> > + return size;
> > +}
> > +
> > static void vhost_user_cleanup(NetClientState *nc)
> > {
> > VhostUserState *s = DO_UPCAST(VhostUserState, nc, nc);
> > @@ -91,6 +109,7 @@ static bool vhost_user_has_ufo(NetClientState *nc)
> > static NetClientInfo net_vhost_user_info = {
> > .type = NET_CLIENT_OPTIONS_KIND_VHOST_USER,
> > .size = sizeof(VhostUserState),
> > + .receive = vhost_user_receive,
> > .cleanup = vhost_user_cleanup,
> > .has_vnet_hdr = vhost_user_has_vnet_hdr,
> > .has_ufo = vhost_user_has_ufo,
> > @@ -147,8 +166,6 @@ static int net_vhost_user_init(NetClientState *peer,
> const char *device,
> >
> > s = DO_UPCAST(VhostUserState, nc, nc);
> >
> > - /* We don't provide a receive callback */
> > - s->nc.receive_disabled = 1;
> > s->chr = chr;
> > s->nc.queue_index = i;
> >
> > --
> > 1.7.10.4
>
>

From:	Thibaut Collet
Subject:	Re: [Qemu-devel] [PATCH v3 1/2] vhost user: add support of live migration
Date:	Wed, 10 Jun 2015 17:24:55 +0200