vhost user: add support of live migration
authorThibaut Collet <thibaut.collet@6wind.com>
Fri, 9 Oct 2015 15:17:31 +0000 (17:17 +0200)
committerMichael S. Tsirkin <mst@redhat.com>
Thu, 22 Oct 2015 11:34:49 +0000 (14:34 +0300)
Some vhost user backends are able to support live migration.
To provide this service the following features must be added:
1. Add the VIRTIO_NET_F_GUEST_ANNOUNCE capability to vhost-net when netdev
   backend is vhost-user.
2. Provide a nop receive callback to vhost-user.
   This callback is called by:
    *  qemu_announce_self after a migration to send fake RARP to avoid network
       outage for peers talking to the migrated guest.
         - For guest with GUEST_ANNOUNCE capabilities, guest already sends GARP
           when the bit VIRTIO_NET_S_ANNOUNCE is set.
           => These packets must be discarded.
         - For guest without GUEST_ANNOUNCE capabilities, migration termination
           is notified when the guest sends packets.
           => These packets can be discarded.
    * virtio_net_tx_bh with a dummy boot to send fake bootp/dhcp request.
      BIOS guest manages virtio driver to send 4 bootp/dhcp request in case of
      dummy boot.
      => These packets must be discarded.

Signed-off-by: Thibaut Collet <thibaut.collet@6wind.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Thibaut Collet <thibaut.collet@6wind.com>
hw/net/vhost_net.c
net/vhost-user.c

index 1ab4133769e9c609a31f7bc0933e8b29cfe513c3..840f44306546422009ff12baa988d0e5fa3085ea 100644 (file)
@@ -85,6 +85,8 @@ 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
index 9b3843181909e546f7b0a5e1914714d91964e980..b067524f2b75888399d054797bcd3e1e89eae672 100644 (file)
@@ -103,6 +103,15 @@ err:
     return -1;
 }
 
+static ssize_t vhost_user_receive(NetClientState *nc, const uint8_t *buf,
+                                  size_t size)
+{
+    /* Discard the request that is received and managed by backend
+     * by an other way.
+     */
+    return size;
+}
+
 static void vhost_user_cleanup(NetClientState *nc)
 {
     VhostUserState *s = DO_UPCAST(VhostUserState, nc, nc);
@@ -132,6 +141,7 @@ 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,
@@ -182,8 +192,6 @@ static int net_vhost_user_init(NetClientState *peer, const char *device,
         snprintf(nc->info_str, sizeof(nc->info_str), "vhost-user%d to %s",
                  i, chr->label);
 
-        /* We don't provide a receive callback */
-        nc->receive_disabled = 1;
         nc->queue_index = i;
 
         s = DO_UPCAST(VhostUserState, nc, nc);