tcp: Fix for race due to temporary drop of the socket lock in skb_splice_bits.

skb_splice_bits temporary drops the socket lock while iterating over
the socket queue in order to break a reverse locking condition which
happens with sendfile. This, however, opens a window of opportunity
for tcp_collapse() to aggregate skbs and thus potentially free the
current skb used in skb_splice_bits and tcp_read_sock.

This patch fixes the problem by (re-)getting the same "logical skb"
after the lock has been temporary dropped.

Based on idea and initial patch from Evgeniy Polyakov.
Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Acked-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/core/skbuff.c

@@ -1445,6 +1445,7 @@ int skb_splice_bits(struct sk_buff *__skb, unsigned int offset,
 
 	if (spd.nr_pages) {
 		int ret;
+		struct sock *sk = __skb->sk;
 
 		/*
 		 * Drop the socket lock, otherwise we have reverse
@@ -1455,9 +1456,9 @@ int skb_splice_bits(struct sk_buff *__skb, unsigned int offset,
 		 * we call into ->sendpage() with the i_mutex lock held
 		 * and networking will grab the socket lock.
 		 */
-		release_sock(__skb->sk);
+		release_sock(sk);
 		ret = splice_to_pipe(pipe, &spd);
-		lock_sock(__skb->sk);
+		lock_sock(sk);
 		return ret;
 	}
 

net/ipv4/tcp.c

@@ -1227,7 +1227,14 @@ int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
 				copied += used;
 				offset += used;
 			}
-			if (offset != skb->len)
+			/*
+			 * If recv_actor drops the lock (e.g. TCP splice
+			 * receive) the skb pointer might be invalid when
+			 * getting here: tcp_collapse might have deleted it
+			 * while aggregating skbs from the socket queue.
+			 */
+			skb = tcp_recv_skb(sk, seq-1, &offset);
+			if (!skb || (offset+1 != skb->len))
 				break;
 		}
 		if (tcp_hdr(skb)->fin) {