[PATCH mptcp-next v2] mptcp: use configured TCP RTO min/max values

[PATCH mptcp-next v2] mptcp: use configured TCP RTO min/max values

[Kalpan Jani]

On the TCP side, the min/max RTO can be configured via the route
rto_min option, the TCP_BPF_RTO_MIN and TCP_RTO_MIN_US socket options,
and the tcp_rto_min_us / tcp_rto_max_ms sysctls, in that order of
precedence. MPTCP did not honour any of these because its retransmit
logic still used the hard-coded TCP_RTO_MIN / TCP_RTO_MAX constants.

Replace the constants with the tcp_rto_min() / tcp_rto_max() helpers
in the three MPTCP-level retransmit paths:

  - mptcp_set_datafin_timeout(): both the backoff cap computation
    and the resulting timer_ival now follow the configured values.
    Guard against a pathological rto_min >= rto_max (e.g. via BPF
    or racing sysctl writers) which would otherwise feed ilog2(0).

  - __mptcp_set_timeout(): the fallback when no subflow timeout is
    available now uses tcp_rto_min().

  - __mptcp_init_sock(): MPTCP does not invoke tcp_init_sock() on
    the msk, so inet_csk(sk)->icsk_rto_min and icsk_rto_max remain
    zero by default. Seed them from the per-netns sysctls before
    using them. The initial timer_ival reads icsk_rto_min directly
    rather than going through tcp_rto_min(sk): at socket init time
    sk_dst_cache is not yet under RCU/lock protection, and the dst
    lookup inside tcp_rto_min() would otherwise trip
    lockdep_rcu_suspicious() via __sk_dst_get() (reported by the
    mptcp CI on v1).

The remaining uses of TCP_RTO_MAX in net/mptcp/ctrl.c (ADD_ADDR
default add_addr_timeout) and net/mptcp/subflow.c (mptcp_subflow_fail()
MP_FAIL timeout) are intentionally left unchanged: they use the
constant as a default duration, not as an RTO bound on a retransmit
timer.

Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/618
Signed-off-by: Kalpan Jani <kalpan.jani@mpiricsoftware.com>
---
Link to v1: https://lore.kernel.org/mptcp/20260610101123.765958-1-kalpan.jani@mpiricsoftware.com/

Changes since v1:
 - __mptcp_init_sock(): seed icsk_rto_min / icsk_rto_max from the
   per-netns sysctls so that the helpers return meaningful values
   on the msk (MPTCP does not call tcp_init_sock() on the msk).
 - __mptcp_init_sock(): use icsk->icsk_rto_min directly for the
   initial timer_ival instead of tcp_rto_min(sk), to avoid a
   lockdep_rcu_suspicious() splat from __sk_dst_get() at socket
   init time. Reported by the mptcp CI on v1.
 - mptcp_set_datafin_timeout(): add an ilog2(0) shift-safety guard
   for the rto_min >= rto_max corner case.

 net/mptcp/protocol.c | 16 ++++++++++++----
 1 file changed, 12 insertions(+), 4 deletions(-)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index a4f7e99b30db..878a0c023f90 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -563,17 +563,19 @@ static bool mptcp_pending_data_fin(struct sock *sk, u64 *seq)
 static void mptcp_set_datafin_timeout(struct sock *sk)
 {
 	struct inet_connection_sock *icsk = inet_csk(sk);
+	u32 rto_min = tcp_rto_min(sk);
+	u32 rto_max = tcp_rto_max(sk);
 	u32 retransmits;
 
 	retransmits = min_t(u32, icsk->icsk_retransmits,
-			    ilog2(TCP_RTO_MAX / TCP_RTO_MIN));
+			    ilog2(max_t(u32, rto_max / rto_min, 1)));
 
-	mptcp_sk(sk)->timer_ival = TCP_RTO_MIN << retransmits;
+	mptcp_sk(sk)->timer_ival = rto_min << retransmits;
 }
 
 static void __mptcp_set_timeout(struct sock *sk, long tout)
 {
-	mptcp_sk(sk)->timer_ival = tout > 0 ? tout : TCP_RTO_MIN;
+	mptcp_sk(sk)->timer_ival = tout > 0 ? tout : tcp_rto_min(sk);
 }
 
 static long mptcp_timeout_from_subflow(const struct mptcp_subflow_context *subflow)
@@ -3150,6 +3152,8 @@ static void mptcp_worker(struct work_struct *work)
 static void __mptcp_init_sock(struct sock *sk)
 {
 	struct mptcp_sock *msk = mptcp_sk(sk);
+	struct inet_connection_sock *icsk = inet_csk(sk);
+	struct net *net = sock_net(sk);
 
 	INIT_LIST_HEAD(&msk->conn_list);
 	INIT_LIST_HEAD(&msk->join_list);
@@ -3158,7 +3162,11 @@ static void __mptcp_init_sock(struct sock *sk)
 	INIT_WORK(&msk->work, mptcp_worker);
 	msk->out_of_order_queue = RB_ROOT;
 	msk->first_pending = NULL;
-	msk->timer_ival = TCP_RTO_MIN;
+
+	/* msk does not go through tcp_init_sock(); seed RTO bounds. */
+	icsk->icsk_rto_min = usecs_to_jiffies(READ_ONCE(net->ipv4.sysctl_tcp_rto_min_us));
+	icsk->icsk_rto_max = msecs_to_jiffies(READ_ONCE(net->ipv4.sysctl_tcp_rto_max_ms));
+	msk->timer_ival = icsk->icsk_rto_min;
 	msk->scaling_ratio = TCP_DEFAULT_SCALING_RATIO;
 	msk->backlog_len = 0;
 	mptcp_init_rtt_est(msk);
-- 
2.43.0

Re: [PATCH mptcp-next v2] mptcp: use configured TCP RTO min/max values

[MPTCP CI]

Hi Kalpan,

Thank you for your modifications, that's great!

Our CI did some validations and here is its report:

- KVM Validation: normal (except selftest_mptcp_join): Success! ✅
- KVM Validation: normal (only selftest_mptcp_join): Success! ✅
- KVM Validation: debug (except selftest_mptcp_join): Success! ✅
- KVM Validation: debug (only selftest_mptcp_join): Success! ✅
- KVM Validation: btf-normal (only bpftest_all): Success! ✅
- KVM Validation: btf-debug (only bpftest_all): Success! ✅
- Task: https://github.com/multipath-tcp/mptcp_net-next/actions/runs/27330400492

Initiator: Patchew Applier
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/b86218ed0f1d
Patchwork: https://patchwork.kernel.org/project/mptcp/list/?series=1109791


If there are some issues, you can reproduce them using the same environment as
the one used by the CI thanks to a docker image, e.g.:

    $ cd [kernel source code]
    $ docker run -v "${PWD}:${PWD}:rw" -w "${PWD}" --privileged --rm -it \
        --pull always mptcp/mptcp-upstream-virtme-docker:latest \
        auto-normal

For more details:

    https://github.com/multipath-tcp/mptcp-upstream-virtme-docker


Please note that despite all the efforts that have been already done to have a
stable tests suite when executed on a public CI like here, it is possible some
reported issues are not due to your modifications. Still, do not hesitate to
help us improve that ;-)

Cheers,
MPTCP GH Action bot
Bot operated by Matthieu Baerts (NGI0 Core)