tcp: annotate sk->sk_rcvbuf lockless reads

For the sake of tcp_poll(), there are few places where we fetch sk->sk_rcvbuf while this field can change from IRQ or other cpu. We need to add READ_ONCE() annotations, and also make sure write sides use corresponding WRITE_ONCE() to avoid store-tearing. Note that other transports probably need similar fixes. Signed-off-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
mariux64 · Oct 13, 2019 · ebb3b78 · ebb3b78
1 parent d9b55bf
commit ebb3b78
Show file tree

Hide file tree

Showing 7 changed files with 15 additions and 12 deletions.
diff --git a/include/net/tcp.h b/include/net/tcp.h
@@ -1380,14 +1380,14 @@ static inline int tcp_win_from_space(const struct sock *sk, int space)
 /* Note: caller must be prepared to deal with negative returns */
 static inline int tcp_space(const struct sock *sk)
 {
-	return tcp_win_from_space(sk, sk->sk_rcvbuf -
+	return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) -
 				  READ_ONCE(sk->sk_backlog.len) -
 				  atomic_read(&sk->sk_rmem_alloc));
 }
 
 static inline int tcp_full_space(const struct sock *sk)
 {
-	return tcp_win_from_space(sk, sk->sk_rcvbuf);
+	return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf));
 }
 
 extern void tcp_openreq_init_rwin(struct request_sock *req,

diff --git a/include/trace/events/sock.h b/include/trace/events/sock.h
@@ -82,7 +82,7 @@ TRACE_EVENT(sock_rcvqueue_full,
 	TP_fast_assign(
 		__entry->rmem_alloc = atomic_read(&sk->sk_rmem_alloc);
 		__entry->truesize   = skb->truesize;
-		__entry->sk_rcvbuf  = sk->sk_rcvbuf;
+		__entry->sk_rcvbuf  = READ_ONCE(sk->sk_rcvbuf);
 	),
 
 	TP_printk("rmem_alloc=%d truesize=%u sk_rcvbuf=%d",

diff --git a/net/core/filter.c b/net/core/filter.c
@@ -4252,7 +4252,8 @@ BPF_CALL_5(bpf_setsockopt, struct bpf_sock_ops_kern *, bpf_sock,
 		case SO_RCVBUF:
 			val = min_t(u32, val, sysctl_rmem_max);
 			sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
-			sk->sk_rcvbuf = max_t(int, val * 2, SOCK_MIN_RCVBUF);
+			WRITE_ONCE(sk->sk_rcvbuf,
+				   max_t(int, val * 2, SOCK_MIN_RCVBUF));
 			break;
 		case SO_SNDBUF:
 			val = min_t(u32, val, sysctl_wmem_max);

diff --git a/net/core/skbuff.c b/net/core/skbuff.c
@@ -4415,7 +4415,7 @@ static void skb_set_err_queue(struct sk_buff *skb)
 int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
 {
 	if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
-	    (unsigned int)sk->sk_rcvbuf)
+	    (unsigned int)READ_ONCE(sk->sk_rcvbuf))
 		return -ENOMEM;
 
 	skb_orphan(skb);

diff --git a/net/core/sock.c b/net/core/sock.c
@@ -831,7 +831,8 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		 * returning the value we actually used in getsockopt
 		 * is the most desirable behavior.
 		 */
-		sk->sk_rcvbuf = max_t(int, val * 2, SOCK_MIN_RCVBUF);
+		WRITE_ONCE(sk->sk_rcvbuf,
+			   max_t(int, val * 2, SOCK_MIN_RCVBUF));
 		break;
 
 	case SO_RCVBUFFORCE:
@@ -3204,7 +3205,7 @@ void sk_get_meminfo(const struct sock *sk, u32 *mem)
 	memset(mem, 0, sizeof(*mem) * SK_MEMINFO_VARS);
 
 	mem[SK_MEMINFO_RMEM_ALLOC] = sk_rmem_alloc_get(sk);
-	mem[SK_MEMINFO_RCVBUF] = sk->sk_rcvbuf;
+	mem[SK_MEMINFO_RCVBUF] = READ_ONCE(sk->sk_rcvbuf);
 	mem[SK_MEMINFO_WMEM_ALLOC] = sk_wmem_alloc_get(sk);
 	mem[SK_MEMINFO_SNDBUF] = sk->sk_sndbuf;
 	mem[SK_MEMINFO_FWD_ALLOC] = sk->sk_forward_alloc;

diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
@@ -451,7 +451,7 @@ void tcp_init_sock(struct sock *sk)
 	icsk->icsk_sync_mss = tcp_sync_mss;
 
 	sk->sk_sndbuf = sock_net(sk)->ipv4.sysctl_tcp_wmem[1];
-	sk->sk_rcvbuf = sock_net(sk)->ipv4.sysctl_tcp_rmem[1];
+	WRITE_ONCE(sk->sk_rcvbuf, sock_net(sk)->ipv4.sysctl_tcp_rmem[1]);
 
 	sk_sockets_allocated_inc(sk);
 	sk->sk_route_forced_caps = NETIF_F_GSO;
@@ -1711,7 +1711,7 @@ int tcp_set_rcvlowat(struct sock *sk, int val)
 
 	val <<= 1;
 	if (val > sk->sk_rcvbuf) {
-		sk->sk_rcvbuf = val;
+		WRITE_ONCE(sk->sk_rcvbuf, val);
 		tcp_sk(sk)->window_clamp = tcp_win_from_space(sk, val);
 	}
 	return 0;

diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
@@ -483,8 +483,9 @@ static void tcp_clamp_window(struct sock *sk)
 	    !(sk->sk_userlocks & SOCK_RCVBUF_LOCK) &&
 	    !tcp_under_memory_pressure(sk) &&
 	    sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)) {
-		sk->sk_rcvbuf = min(atomic_read(&sk->sk_rmem_alloc),
-				    net->ipv4.sysctl_tcp_rmem[2]);
+		WRITE_ONCE(sk->sk_rcvbuf,
+			   min(atomic_read(&sk->sk_rmem_alloc),
+			       net->ipv4.sysctl_tcp_rmem[2]));
 	}
 	if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf)
 		tp->rcv_ssthresh = min(tp->window_clamp, 2U * tp->advmss);
@@ -648,7 +649,7 @@ void tcp_rcv_space_adjust(struct sock *sk)
 		rcvbuf = min_t(u64, rcvwin * rcvmem,
 			       sock_net(sk)->ipv4.sysctl_tcp_rmem[2]);
 		if (rcvbuf > sk->sk_rcvbuf) {
-			sk->sk_rcvbuf = rcvbuf;
+			WRITE_ONCE(sk->sk_rcvbuf, rcvbuf);
 
 			/* Make the window clamp follow along.  */
 			tp->window_clamp = tcp_win_from_space(sk, rcvbuf);