tuntap: fix a possible race between queue selection and changing queues

Complier may generate codes that re-read the tun->numqueues during
tun_select_queue(). This may be a race if vlan->numqueues were changed in the
same time and can lead unexpected result (e.g. very huge value).

We need prevent the compiler from generating such codes by adding an
ACCESS_ONCE() to make sure tun->numqueues were only read once.

Bug were introduced by commit c8d68e6be1c3b242f1c598595830890b65cea64a
(tuntap: multiqueue support).

Reported-by: Michael S. Tsirkin <mst@redhat.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index 89776c5..b1cbfbc 100644 (file)
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -352,7 +352,7 @@ static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb)
        u32 numqueues = 0;
 
        rcu_read_lock();
-       numqueues = tun->numqueues;
+       numqueues = ACCESS_ONCE(tun->numqueues);
 
        txq = skb_get_rxhash(skb);
        if (txq) {