virtio-net: set queues after driver_ok
authorJason Wang <jasowang@redhat.com>
Thu, 10 Aug 2023 03:12:56 +0000 (23:12 -0400)
committerDavid S. Miller <davem@davemloft.net>
Fri, 11 Aug 2023 08:50:14 +0000 (09:50 +0100)
Commit 25266128fe16 ("virtio-net: fix race between set queues and
probe") tries to fix the race between set queues and probe by calling
_virtnet_set_queues() before DRIVER_OK is set. This violates virtio
spec. Fixing this by setting queues after virtio_device_ready().

Note that rtnl needs to be held for userspace requests to change the
number of queues. So we are serialized in this way.

Fixes: 25266128fe16 ("virtio-net: fix race between set queues and probe")
Reported-by: Dragos Tatulea <dtatulea@nvidia.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/virtio_net.c

index 1270c8d23463fa35849dd290b7df7496198aaf9b..ff03921e46dfb7b5e5bce6ac78fc6cda93692736 100644 (file)
@@ -4219,8 +4219,6 @@ static int virtnet_probe(struct virtio_device *vdev)
        if (vi->has_rss || vi->has_rss_hash_report)
                virtnet_init_default_rss(vi);
 
-       _virtnet_set_queues(vi, vi->curr_queue_pairs);
-
        /* serialize netdev register + virtio_device_ready() with ndo_open() */
        rtnl_lock();
 
@@ -4233,6 +4231,8 @@ static int virtnet_probe(struct virtio_device *vdev)
 
        virtio_device_ready(vdev);
 
+       _virtnet_set_queues(vi, vi->curr_queue_pairs);
+
        /* a random MAC address has been assigned, notify the device.
         * We don't fail probe if VIRTIO_NET_F_CTRL_MAC_ADDR is not there
         * because many devices work fine without getting MAC explicitly