From: Geliang Tang <tanggeliang@kylinos.cn>

Currently the ADD_ADDR option is retransmitted with a fixed timeout. This
patch makes the retransmission timeout adaptive by using the maximum RTO
among all the subflows, while still capping it at the configured maximum
value (add_addr_timeout_max). This improves responsiveness when
establishing new subflows.

Specifically:
1. Adds mptcp_adjust_add_addr_timeout() helper to compute the adaptive
timeout.
2. Uses maximum subflow RTO (icsk_rto) when available.
3. Applies exponential backoff based on retransmission count.
4. Maintains fallback to configured max timeout when no RTO data exists.

Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/576
Reviewed-by: Christoph Paasch <cpaasch@openai.com>
Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 Documentation/networking/mptcp-sysctl.rst |  4 ++--
 net/mptcp/pm.c                            | 28 +++++++++++++++++++----
 2 files changed, 26 insertions(+), 6 deletions(-)

diff --git a/Documentation/networking/mptcp-sysctl.rst b/Documentation/networking/mptcp-sysctl.rst
index 1683c139821e..2b8e8fb99ee2 100644
--- a/Documentation/networking/mptcp-sysctl.rst
+++ b/Documentation/networking/mptcp-sysctl.rst
@@ -8,8 +8,8 @@ MPTCP Sysfs variables
 ===============================
 
 add_addr_timeout - INTEGER (seconds)
-	Set the timeout after which an ADD_ADDR control message will be
-	resent to an MPTCP peer that has not acknowledged a previous
+	Set the maximum value of timeout after which an ADD_ADDR control message
+	will be resent to an MPTCP peer that has not acknowledged a previous
 	ADD_ADDR message.
 
 	Do not retransmit if set to 0.
diff --git a/net/mptcp/pm.c b/net/mptcp/pm.c
index e6dcc5b62645..02dfb379417e 100644
--- a/net/mptcp/pm.c
+++ b/net/mptcp/pm.c
@@ -268,6 +268,27 @@ int mptcp_pm_mp_prio_send_ack(struct mptcp_sock *msk,
 	return -EINVAL;
 }
 
+static unsigned int mptcp_adjust_add_addr_timeout(struct mptcp_sock *msk)
+{
+	const struct net *net = sock_net((struct sock *)msk);
+	unsigned int rto = mptcp_get_add_addr_timeout(net);
+	struct mptcp_subflow_context *subflow;
+	unsigned int max = 0;
+
+	mptcp_for_each_subflow(msk, subflow) {
+		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
+		struct inet_connection_sock *icsk = inet_csk(ssk);
+
+		if (icsk->icsk_rto > max)
+			max = icsk->icsk_rto;
+	}
+
+	if (max && max < rto)
+		rto = max;
+
+	return rto;
+}
+
 static void mptcp_pm_add_timer(struct timer_list *timer)
 {
 	struct mptcp_pm_add_entry *entry = timer_container_of(entry, timer,
@@ -292,7 +313,7 @@ static void mptcp_pm_add_timer(struct timer_list *timer)
 		goto out;
 	}
 
-	timeout = mptcp_get_add_addr_timeout(sock_net(sk));
+	timeout = mptcp_adjust_add_addr_timeout(msk);
 	if (!timeout)
 		goto out;
 
@@ -307,7 +328,7 @@ static void mptcp_pm_add_timer(struct timer_list *timer)
 
 	if (entry->retrans_times < ADD_ADDR_RETRANS_MAX)
 		sk_reset_timer(sk, timer,
-			       jiffies + timeout);
+			       jiffies + (timeout << entry->retrans_times));
 
 	spin_unlock_bh(&msk->pm.lock);
 
@@ -348,7 +369,6 @@ bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
 {
 	struct mptcp_pm_add_entry *add_entry = NULL;
 	struct sock *sk = (struct sock *)msk;
-	struct net *net = sock_net(sk);
 	unsigned int timeout;
 
 	lockdep_assert_held(&msk->pm.lock);
@@ -374,7 +394,7 @@ bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
 
 	timer_setup(&add_entry->add_timer, mptcp_pm_add_timer, 0);
 reset_timer:
-	timeout = mptcp_get_add_addr_timeout(net);
+	timeout = mptcp_adjust_add_addr_timeout(msk);
 	if (timeout)
 		sk_reset_timer(sk, &add_entry->add_timer, jiffies + timeout);
 
-- 
2.48.1

Hi Geliang,

On 12/08/2025 10:39, Geliang Tang wrote:
> From: Geliang Tang <tanggeliang@kylinos.cn>
> 
> Currently the ADD_ADDR option is retransmitted with a fixed timeout. This
> patch makes the retransmission timeout adaptive by using the maximum RTO
> among all the subflows, while still capping it at the configured maximum
> value (add_addr_timeout_max). This improves responsiveness when
> establishing new subflows.
> 
> Specifically:
> 1. Adds mptcp_adjust_add_addr_timeout() helper to compute the adaptive
> timeout.
> 2. Uses maximum subflow RTO (icsk_rto) when available.
> 3. Applies exponential backoff based on retransmission count.
> 4. Maintains fallback to configured max timeout when no RTO data exists.

I might be good to add a note about the behavioural change:

  This slightly change the behaviour of the MPTCP "add_addr_timeout"
  sysctl knob to be used as a maximum instead of a fixed value, but this
  is seen as an improvement: the ADD_ADDR might be sent quicker than
  before to improve the overall MPTCP connection. Also, the default
  value is set at 2 min, which was way too big, and caused the ADD_ADDR
  not to be retransmitted for connections shorter than 2 minutes.

> Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/576
> Reviewed-by: Christoph Paasch <cpaasch@openai.com>
> Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
> ---
>  Documentation/networking/mptcp-sysctl.rst |  4 ++--
>  net/mptcp/pm.c                            | 28 +++++++++++++++++++----
>  2 files changed, 26 insertions(+), 6 deletions(-)
> 
> diff --git a/Documentation/networking/mptcp-sysctl.rst b/Documentation/networking/mptcp-sysctl.rst
> index 1683c139821e..2b8e8fb99ee2 100644
> --- a/Documentation/networking/mptcp-sysctl.rst
> +++ b/Documentation/networking/mptcp-sysctl.rst
> @@ -8,8 +8,8 @@ MPTCP Sysfs variables
>  ===============================
>  
>  add_addr_timeout - INTEGER (seconds)
> -	Set the timeout after which an ADD_ADDR control message will be
> -	resent to an MPTCP peer that has not acknowledged a previous
> +	Set the maximum value of timeout after which an ADD_ADDR control message
> +	will be resent to an MPTCP peer that has not acknowledged a previous
>  	ADD_ADDR message.

I suggest adding this at the end:

  A dynamically estimated retransmission timeout based on the estimated
  connection round-trip-time is used if this value is lower than the
  maximum one.

I can do that when applying the patch. I can update the text if needed
later on.

Cheers,
Matt
-- 
Sponsored by the NGI0 Core fund.