tcp/dccp: fix timewait races in timer handling

When creating a timewait socket, we need to arm the timer before
allowing other cpus to find it. The signal allowing cpus to find
the socket is setting tw_refcnt to non zero value.

As we set tw_refcnt in __inet_twsk_hashdance(), we therefore need to
call inet_twsk_schedule() first.

This also means we need to remove tw_refcnt changes from
inet_twsk_schedule() and let the caller handle it.

Note that because we use mod_timer_pinned(), we have the guarantee
the timer wont expire before we set tw_refcnt as we run in BH context.

To make things more readable I introduced inet_twsk_reschedule() helper.

When rearming the timer, we can use mod_timer_pending() to make sure
we do not rearm a canceled timer.

Note: This bug can possibly trigger if packets of a flow can hit
multiple cpus. This does not normally happen, unless flow steering
is broken somehow. This explains this bug was spotted ~5 months after
its introduction.

A similar fix is needed for SYN_RECV sockets in reqsk_queue_hash_req(),
but will be provided in a separate patch for proper tracking.

Fixes: 789f558cfb36 ("tcp/dccp: get rid of central timewait timer")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Ying Cai <ycai@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

3 files changed, 18 insertions, 15 deletions

diff --git a/net/dccp/minisocks.c b/net/dccp/minisocks.c
index 30addee..838f524 100644
--- a/net/dccp/minisocks.c
+++ b/net/dccp/minisocks.c
@@ -48,8 +48,6 @@ void dccp_time_wait(struct sock *sk, int state, int timeo)
 			tw->tw_ipv6only = sk->sk_ipv6only;
 		}
 #endif
-		/* Linkage updates. */
-		__inet_twsk_hashdance(tw, sk, &dccp_hashinfo);
 
 		/* Get the TIME_WAIT timeout firing. */
 		if (timeo < rto)
@@ -60,6 +58,8 @@ void dccp_time_wait(struct sock *sk, int state, int timeo)
 			timeo = DCCP_TIMEWAIT_LEN;
 
 		inet_twsk_schedule(tw, timeo);
+		/* Linkage updates. */
+		__inet_twsk_hashdance(tw, sk, &dccp_hashinfo);
 		inet_twsk_put(tw);
 	} else {
 		/* Sorry, if we're out of memory, just CLOSE this
diff --git a/net/ipv4/inet_timewait_sock.c b/net/ipv4/inet_timewait_sock.c
index ae22cc2..c67f9bd 100644
--- a/net/ipv4/inet_timewait_sock.c
+++ b/net/ipv4/inet_timewait_sock.c
@@ -123,13 +123,15 @@ void __inet_twsk_hashdance(struct inet_timewait_sock *tw, struct sock *sk,
 	/*
 	 * Step 2: Hash TW into tcp ehash chain.
 	 * Notes :
-	 * - tw_refcnt is set to 3 because :
+	 * - tw_refcnt is set to 4 because :
 	 * - We have one reference from bhash chain.
 	 * - We have one reference from ehash chain.
+	 * - We have one reference from timer.
+	 * - One reference for ourself (our caller will release it).
 	 * We can use atomic_set() because prior spin_lock()/spin_unlock()
 	 * committed into memory all tw fields.
 	 */
-	atomic_set(&tw->tw_refcnt, 1 + 1 + 1);
+	atomic_set(&tw->tw_refcnt, 4);
 	inet_twsk_add_node_rcu(tw, &ehead->chain);
 
 	/* Step 3: Remove SK from hash chain */
@@ -217,7 +219,7 @@ void inet_twsk_deschedule_put(struct inet_timewait_sock *tw)
 }
 EXPORT_SYMBOL(inet_twsk_deschedule_put);
 
-void inet_twsk_schedule(struct inet_timewait_sock *tw, const int timeo)
+void __inet_twsk_schedule(struct inet_timewait_sock *tw, int timeo, bool rearm)
 {
 	/* timeout := RTO * 3.5
 	 *
@@ -245,12 +247,14 @@ void inet_twsk_schedule(struct inet_timewait_sock *tw, const int timeo)
 	 */
 
 	tw->tw_kill = timeo <= 4*HZ;
-	if (!mod_timer_pinned(&tw->tw_timer, jiffies + timeo)) {
-		atomic_inc(&tw->tw_refcnt);
+	if (!rearm) {
+		BUG_ON(mod_timer_pinned(&tw->tw_timer, jiffies + timeo));
 		atomic_inc(&tw->tw_dr->tw_count);
+	} else {
+		mod_timer_pending(&tw->tw_timer, jiffies + timeo);
 	}
 }
-EXPORT_SYMBOL_GPL(inet_twsk_schedule);
+EXPORT_SYMBOL_GPL(__inet_twsk_schedule);
 
 void inet_twsk_purge(struct inet_hashinfo *hashinfo,
 		     struct inet_timewait_death_row *twdr, int family)
diff --git a/net/ipv4/tcp_minisocks.c b/net/ipv4/tcp_minisocks.c
index 6d8795b..def7659 100644
--- a/net/ipv4/tcp_minisocks.c
+++ b/net/ipv4/tcp_minisocks.c
@@ -162,9 +162,9 @@ kill_with_rst:
 		if (tcp_death_row.sysctl_tw_recycle &&
 		    tcptw->tw_ts_recent_stamp &&
 		    tcp_tw_remember_stamp(tw))
-			inet_twsk_schedule(tw, tw->tw_timeout);
+			inet_twsk_reschedule(tw, tw->tw_timeout);
 		else
-			inet_twsk_schedule(tw, TCP_TIMEWAIT_LEN);
+			inet_twsk_reschedule(tw, TCP_TIMEWAIT_LEN);
 		return TCP_TW_ACK;
 	}
 
@@ -201,7 +201,7 @@ kill:
 				return TCP_TW_SUCCESS;
 			}
 		}
-		inet_twsk_schedule(tw, TCP_TIMEWAIT_LEN);
+		inet_twsk_reschedule(tw, TCP_TIMEWAIT_LEN);
 
 		if (tmp_opt.saw_tstamp) {
 			tcptw->tw_ts_recent	  = tmp_opt.rcv_tsval;
@@ -251,7 +251,7 @@ kill:
 		 * Do not reschedule in the last case.
 		 */
 		if (paws_reject || th->ack)
-			inet_twsk_schedule(tw, TCP_TIMEWAIT_LEN);
+			inet_twsk_reschedule(tw, TCP_TIMEWAIT_LEN);
 
 		return tcp_timewait_check_oow_rate_limit(
 			tw, skb, LINUX_MIB_TCPACKSKIPPEDTIMEWAIT);
@@ -322,9 +322,6 @@ void tcp_time_wait(struct sock *sk, int state, int timeo)
 		} while (0);
 #endif
 
-		/* Linkage updates. */
-		__inet_twsk_hashdance(tw, sk, &tcp_hashinfo);
-
 		/* Get the TIME_WAIT timeout firing. */
 		if (timeo < rto)
 			timeo = rto;
@@ -338,6 +335,8 @@ void tcp_time_wait(struct sock *sk, int state, int timeo)
 		}
 
 		inet_twsk_schedule(tw, timeo);
+		/* Linkage updates. */
+		__inet_twsk_hashdance(tw, sk, &tcp_hashinfo);
 		inet_twsk_put(tw);
 	} else {
 		/* Sorry, if we're out of memory, just CLOSE this