Skip to contentSkip to conversation

Fix race in tcp_poll

PATCH In review Tom Marshall 3 participants 3 messages Sep 20, 2010

Subsystems (3)
Networking Network drivers TCP
lore
Share & tools
lore

Tom Marshall

author
#0 lore raw

PATCH 1/1 Fix race in tcp_poll

If a RST comes in immediately after checking sk->sk_err, tcp_poll will return POLLIN but not POLLOUT. Fix this by checking sk->sk_err at the end of tcp_poll. Additionally, ensure the correct order of operations on SMP machines with memory barriers.

Signed-off-by Tom Marshall <[email protected]>
net/ipv4/tcp.c +5 -2
index 3fb1428..95d75d4 100644
@@ -386,8 +386,6 @@ unsigned int tcp_poll(struct file *file, struct socket *sock, poll_table *wait)
 	 */
 
 	mask = 0;
-	if (sk->sk_err)
-		mask = POLLERR;
 
 	/*
 	 * POLLHUP is certainly not done right. But poll() doesn't

@@ -457,6 +455,11 @@ unsigned int tcp_poll(struct file *file, struct socket *sock, poll_table *wait)
 		if (tp->urg_data & TCP_URG_VALID)
 			mask |= POLLPRI;
 	}
+	/* This barrier is coupled with smp_wmb() in tcp_reset() */
+	smp_rmb();
+	if (sk->sk_err)
+		mask |= POLLERR;
+
 	return mask;
 }
 EXPORT_SYMBOL(tcp_poll);
net/ipv4/tcp_input.c +2 -0
index e663b78..149e79a 100644
@@ -4048,6 +4048,8 @@ static void tcp_reset(struct sock *sk)
 	default:
 		sk->sk_err = ECONNRESET;
 	}
+	/* This barrier is coupled with smp_rmb() in tcp_poll() */
+	smp_wmb();
 
 	if (!sock_flag(sk, SOCK_DEAD))
 		sk->sk_error_report(sk);
#1 lore raw
Quoted 38 lines · Tom Marshall

If a RST comes in immediately after checking sk->sk_err, tcp_poll will return POLLIN but not POLLOUT. Fix this by checking sk->sk_err at the end of tcp_poll. Additionally, ensure the correct order of operations on SMP machines with memory barriers.

Signed-off-by Tom Marshall <[email protected]>
net/ipv4/tcp.c +5 -2
index 3fb1428..95d75d4 100644
@@ -386,8 +386,6 @@ unsigned int tcp_poll(struct file *file, struct socket *sock, poll_table *wait)
 	 */
 
 	mask = 0;
-	if (sk->sk_err)
-		mask = POLLERR;
 
 	/*
 	 * POLLHUP is certainly not done right. But poll() doesn't

@@ -457,6 +455,11 @@ unsigned int tcp_poll(struct file *file, struct socket *sock, poll_table *wait)
 		if (tp->urg_data & TCP_URG_VALID)
 			mask |= POLLPRI;
 	}
+	/* This barrier is coupled with smp_wmb() in tcp_reset() */
+	smp_rmb();
+	if (sk->sk_err)
+		mask |= POLLERR;
+
 	return mask;
 }
 EXPORT_SYMBOL(tcp_poll);
net/ipv4/tcp_input.c +0 -0
index e663b78..149e79a 100644

net/ipv4/tcp_input.c static void tcp_reset(struct sock *sk) · line 4055

Quoted 8 lines
net/ipv4/tcp_input.c +2 -0
@@ -4048,6 +4048,8 @@ static void tcp_reset(struct sock *sk)
 	default:
 		sk->sk_err = ECONNRESET;
 	}
+	/* This barrier is coupled with smp_rmb() in tcp_poll() */
+	smp_wmb();
 
 	if (!sock_flag(sk, SOCK_DEAD))
 		sk->sk_error_report(sk);
Signed-off-by Eric Dumazet <[email protected]>

Maybe this is the last time we can avoid taking socket lock in tcp_poll() ;)

maintainer
#2 lore raw

From: Eric Dumazet <[email protected]>
Date: Mon, 20 Sep 2010 23:54:06 +0200

Quoted 7 lines · Tom Marshall
Quoted 6 lines

If a RST comes in immediately after checking sk->sk_err, tcp_poll will return POLLIN but not POLLOUT. Fix this by checking sk->sk_err at the end of tcp_poll. Additionally, ensure the correct order of operations on SMP machines with memory barriers.

Signed-off-by Tom Marshall <[email protected]>

...

Quoted 4 lines
Signed-off-by Eric Dumazet <[email protected]>

Maybe this is the last time we can avoid taking socket lock in tcp_poll() ;)

:-) Applied, thanks everyone.