l2tp: fix race in l2tp_recv_dequeue()

Misha Labjuk reported panics occurring in l2tp_recv_dequeue()

If we release reorder_q.lock, we must not keep a dangling pointer (tmp),
since another thread could manipulate reorder_q.

Instead we must restart the scan at beginning of list.

Reported-by: Misha Labjuk <spiked.yar@gmail.com>
Tested-by: Misha Labjuk <spiked.yar@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/net/l2tp/l2tp_core.c b/net/l2tp/l2tp_core.c
index 34b2dde..bf8d50c 100644 (file)
--- a/net/l2tp/l2tp_core.c
+++ b/net/l2tp/l2tp_core.c
@@ -397,6 +397,7 @@ static void l2tp_recv_dequeue(struct l2tp_session *session)
         * expect to send up next, dequeue it and any other
         * in-sequence packets behind it.
         */
+start:
        spin_lock_bh(&session->reorder_q.lock);
        skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
                if (time_after(jiffies, L2TP_SKB_CB(skb)->expires)) {
@@ -433,7 +434,7 @@ static void l2tp_recv_dequeue(struct l2tp_session *session)
                 */
                spin_unlock_bh(&session->reorder_q.lock);
                l2tp_recv_dequeue_skb(session, skb);
-               spin_lock_bh(&session->reorder_q.lock);
+               goto start;
        }
 
 out: