udp: fix poll()
authorPaolo Abeni <pabeni@redhat.com>
Fri, 23 Jun 2017 12:19:51 +0000 (14:19 +0200)
committerDavid S. Miller <davem@davemloft.net>
Fri, 23 Jun 2017 15:18:43 +0000 (11:18 -0400)
Michael reported an UDP breakage caused by the commit b65ac44674dd
("udp: try to avoid 2 cache miss on dequeue").
The function __first_packet_length() can update the checksum bits
of the pending skb, making the scratched area out-of-sync, and
setting skb->csum, if the skb was previously in need of checksum
validation.

On later recvmsg() for such skb, checksum validation will be
invoked again - due to the wrong udp_skb_csum_unnecessary()
value - and will fail, causing the valid skb to be dropped.

This change addresses the issue refreshing the scratch area in
__first_packet_length() after the possible checksum update.

Fixes: b65ac44674dd ("udp: try to avoid 2 cache miss on dequeue")
Reported-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/udp.c

index 067a607917f95683b646e86d6f045ea67e026c3b..47c7aa0501af70e53332cc9e4ae79151f548d6f9 100644 (file)
@@ -1446,16 +1446,23 @@ static struct sk_buff *__first_packet_length(struct sock *sk,
 {
        struct sk_buff *skb;
 
-       while ((skb = skb_peek(rcvq)) != NULL &&
-              udp_lib_checksum_complete(skb)) {
-               __UDP_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS,
-                               IS_UDPLITE(sk));
-               __UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
-                               IS_UDPLITE(sk));
-               atomic_inc(&sk->sk_drops);
-               __skb_unlink(skb, rcvq);
-               *total += skb->truesize;
-               kfree_skb(skb);
+       while ((skb = skb_peek(rcvq)) != NULL) {
+               if (udp_lib_checksum_complete(skb)) {
+                       __UDP_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS,
+                                       IS_UDPLITE(sk));
+                       __UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
+                                       IS_UDPLITE(sk));
+                       atomic_inc(&sk->sk_drops);
+                       __skb_unlink(skb, rcvq);
+                       *total += skb->truesize;
+                       kfree_skb(skb);
+               } else {
+                       /* the csum related bits could be changed, refresh
+                        * the scratch area
+                        */
+                       udp_set_dev_scratch(skb);
+                       break;
+               }
        }
        return skb;
 }