Skip to contentSkip to conversation

mptcp: fix 3rd ack rtx timer

PATCH Applied Paolo Abeni 3 participants 5 messages Nov 19, 2021 – Nov 20, 2021

Subsystems (3)
lore
Share & tools
lore

Paolo Abeni

author maintainer
#0 lore raw

cover mptcp: fix 3rd ack rtx timer

Eric noted that the MPTCP code do the wrong thing to schedule the MPJ 3rd ack timer. He also provided a patch to address the issues (patch 1/2).

To fix for good the MPJ 3rd ack retransmission timer, we additionally need to set it after the current ack is transmitted (patch 2/2)

Note that the bug went unnotice so far because all the related tests required some running data transfer, and that causes MPTCP-level ack even on the opening MPJ subflow. We now have explicit packet drill coverage for this code path.

Eric Dumazet (1):
mptcp: fix delack timer

Paolo Abeni (1):
mptcp: use delegate action to schedule 3rd ack retrans

 net/mptcp/options.c  | 32 ++++++++-------------------
 net/mptcp/protocol.c | 51 ++++++++++++++++++++++++++++++++++++--------
 net/mptcp/protocol.h | 17 ++++++++-------
 3 files changed, 60 insertions(+), 40 deletions(-)
Signature · 1 line
2.33.1
author maintainer
#1 lore raw

PATCH 1/2 mptcp: fix delack timer

From: Eric Dumazet <[email protected]>

To compute the rtx timeout schedule_3rdack_retransmission() does multiple things in the wrong way: srtt_us is measured in usec/8 and the timeout itself is an absolute value.

Fixes ec3edaa7ca6ce02f ("mptcp: Add handling of outgoing MP_JOIN requests")
---
 net/mptcp/options.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
net/mptcp/options.c +2 -1
index 7c3420afb1a0..2e9b73eeeeb5 100644
@@ -434,9 +434,10 @@ static void schedule_3rdack_retransmission(struct sock *sk)
 
 	/* reschedule with a timeout above RTT, as we must look only for drop */
 	if (tp->srtt_us)
-		timeout = tp->srtt_us << 1;
+		timeout = usecs_to_jiffies(tp->srtt_us >> (3 - 1));
 	else
 		timeout = TCP_TIMEOUT_INIT;
+	timeout += jiffies;
 
 	WARN_ON_ONCE(icsk->icsk_ack.pending & ICSK_ACK_TIMER);
 	icsk->icsk_ack.pending |= ICSK_ACK_SCHED | ICSK_ACK_TIMER;
Signature · 1 line
2.33.1
author maintainer
#2 lore raw

PATCH 2/2 mptcp: use delegate action to schedule 3rd ack retrans

In reply to Paolo Abeni

Scheduling a delack in mptcp_established_options_mp() is not a good idea: such function is called by tcp_send_ack() and the pending delayed ack will be cleared shortly after by the tcp_event_ack_sent() call in __tcp_transmit_skb().

Instead use the mptcp delegated action infrastructure to schedule the delayed ack after the current bh processing completes.

Additionally moves the schedule_3rdack_retransmission() helper into protocol.c to avoid making it visible in a different compilation unit.

Fixes ec3edaa7ca6ce02f ("mptcp: Add handling of outgoing MP_JOIN requests")
---
 net/mptcp/options.c  | 33 ++++++++--------------------
 net/mptcp/protocol.c | 51 ++++++++++++++++++++++++++++++++++++--------
 net/mptcp/protocol.h | 17 ++++++++-------
 3 files changed, 60 insertions(+), 41 deletions(-)
net/mptcp/options.c +9 -24
index 2e9b73eeeeb5..fe98e4f475ba 100644
@@ -422,29 +422,6 @@ bool mptcp_syn_options(struct sock *sk, const struct sk_buff *skb,
 	return false;
 }
 
-/* MP_JOIN client subflow must wait for 4th ack before sending any data:
- * TCP can't schedule delack timer before the subflow is fully established.
- * MPTCP uses the delack timer to do 3rd ack retransmissions
- */
-static void schedule_3rdack_retransmission(struct sock *sk)
-{
-	struct inet_connection_sock *icsk = inet_csk(sk);
-	struct tcp_sock *tp = tcp_sk(sk);
-	unsigned long timeout;
-
-	/* reschedule with a timeout above RTT, as we must look only for drop */
-	if (tp->srtt_us)
-		timeout = usecs_to_jiffies(tp->srtt_us >> (3 - 1));
-	else
-		timeout = TCP_TIMEOUT_INIT;
-	timeout += jiffies;
-
-	WARN_ON_ONCE(icsk->icsk_ack.pending & ICSK_ACK_TIMER);
-	icsk->icsk_ack.pending |= ICSK_ACK_SCHED | ICSK_ACK_TIMER;
-	icsk->icsk_ack.timeout = timeout;
-	sk_reset_timer(sk, &icsk->icsk_delack_timer, timeout);
-}
-
 static void clear_3rdack_retransmission(struct sock *sk)
 {
 	struct inet_connection_sock *icsk = inet_csk(sk);
@@ -527,7 +504,15 @@ static bool mptcp_established_options_mp(struct sock *sk, struct sk_buff *skb,
 		*size = TCPOLEN_MPTCP_MPJ_ACK;
 		pr_debug("subflow=%p", subflow);
 
-		schedule_3rdack_retransmission(sk);
+		/* we can use the full delegate action helper only from BH context
+		 * If we are in process context - sk is flushing the backlog at
+		 * socket lock release time - just set the appropriate flag, will
+		 * be handled by the release callback
+		 */
+		if (sock_owned_by_user(sk))
+			set_bit(MPTCP_DELEGATE_ACK, &subflow->delegated_status);
+		else
+			mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_ACK);
 		return true;
 	}
 	return false;
net/mptcp/protocol.c +42 -9

74 lines not on this page · raw message

net/mptcp/protocol.h +9 -8
index 67a61ac48b20..d87cc040352e 100644
@@ -387,6 +387,7 @@ struct mptcp_delegated_action {
 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
 
 #define MPTCP_DELEGATE_SEND		0
+#define MPTCP_DELEGATE_ACK		1
 
 /* MPTCP subflow context */
 struct mptcp_subflow_context {
@@ -492,23 +493,23 @@ static inline void mptcp_add_pending_subflow(struct mptcp_sock *msk,
 
 void mptcp_subflow_process_delegated(struct sock *ssk);
 
-static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow)
+static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
 {
 	struct mptcp_delegated_action *delegated;
 	bool schedule;
 
+	/* the caller held the subflow bh socket lock */
+	lockdep_assert_in_softirq();
+
 	/* The implied barrier pairs with mptcp_subflow_delegated_done(), and
 	 * ensures the below list check sees list updates done prior to status
 	 * bit changes
 	 */
-	if (!test_and_set_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status)) {
+	if (!test_and_set_bit(action, &subflow->delegated_status)) {
 		/* still on delegated list from previous scheduling */
 		if (!list_empty(&subflow->delegated_node))
 			return;
 
-		/* the caller held the subflow bh socket lock */
-		lockdep_assert_in_softirq();
-
 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
 		schedule = list_empty(&delegated->head);
 		list_add_tail(&subflow->delegated_node, &delegated->head);
@@ -533,16 +534,16 @@ mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
 
 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow)
 {
-	return test_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status);
+	return !!READ_ONCE(subflow->delegated_status);
 }
 
-static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow)
+static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow, int action)
 {
 	/* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before
 	 * touching the status bit
 	 */
 	smp_wmb();
-	clear_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status);
+	clear_bit(action, &subflow->delegated_status);
 }
 
 int mptcp_is_enabled(const struct net *net);
Signature · 1 line
2.33.1
maintainer Acked-by on the v1 series
#3 lore raw
In reply to Paolo Abeni

From: Paolo Abeni <[email protected]>
Date: Fri, 19 Nov 2021 15:27:53 +0100

Quoted 11 lines

Eric noted that the MPTCP code do the wrong thing to schedule the MPJ 3rd ack timer. He also provided a patch to address the issues (patch 1/2).

To fix for good the MPJ 3rd ack retransmission timer, we additionally need to set it after the current ack is transmitted (patch 2/2)

Note that the bug went unnotice so far because all the related tests required some running data transfer, and that causes MPTCP-level ack even on the opening MPJ subflow. We now have explicit packet drill coverage for this code path.

#4 lore raw
In reply to Paolo Abeni
This series was applied to netdev/net.git (master) by David S. Miller <[email protected]>:

Hello:

This series was applied to netdev/net.git (master) by David S. Miller <[email protected]>:

Quoted 8 lines · On Fri, 19 Nov 2021 15:27:53 +0100 you wrote:

Eric noted that the MPTCP code do the wrong thing to schedule the MPJ 3rd ack timer. He also provided a patch to address the issues (patch 1/2).

To fix for good the MPJ 3rd ack retransmission timer, we additionally need to set it after the current ack is transmitted (patch 2/2)

[...]

Here is the summary with links:
- [net,1/2] mptcp: fix delack timer

    https://git.kernel.org/netdev/net/c/ee50e67ba0e1
  - [net,2/2] mptcp: use delegate action to schedule 3rd ack retrans
    https://git.kernel.org/netdev/net/c/bcd97734318d

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html