[PATCH 5.10.y 0/3] mptcp: fix recent failed backports (20260919)

[PATCH 5.10.y 0/3] mptcp: fix recent failed backports (20260919)

[Matthieu Baerts (NGI0)]

The following patches could not be applied without conflicts in this
tree:

- 2b0f561f21b2 ("mptcp: avoid unneeded actions on subflow reset")
- 42064de57fb8 ("mptcp: close race between scheduler and state change")

I backported the following commit to avoid conflicts:

- ab82e996a1fa ("mptcp: hold mptcp socket before calling tcp_done")

Conflicts (if any) have been resolved, and documented in each patch.

Florian Westphal (1):
  mptcp: hold mptcp socket before calling tcp_done

Paolo Abeni (2):
  mptcp: avoid unneeded actions on subflow reset
  mptcp: close race between scheduler and state change

 net/mptcp/protocol.c | 10 +++++-----
 net/mptcp/protocol.h |  3 ++-
 net/mptcp/subflow.c  | 18 +++++++++++++++++-
 3 files changed, 24 insertions(+), 7 deletions(-)

-- 
2.55.0

[PATCH 5.10.y 1/3] mptcp: hold mptcp socket before calling tcp_done

[Matthieu Baerts (NGI0)]

From: Florian Westphal <fw@strlen.de>

commit ab82e996a1fa1b9ae514fa357d9ce8df62321157 upstream.

When processing options from tcp reset path its possible that
tcp_done(ssk) drops the last reference on the mptcp socket which
results in use-after-free.

Reviewed-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Stable-dep-of: 2b0f561f21b2 ("mptcp: avoid unneeded actions on subflow reset")
[ Conflicts in subflow.c, because commit 3ba14528684f ("mptcp: avoid
  setting TCP_CLOSE state twice") has already been backported and also
  had the same conflict: this commit here should have been backported
  first. Fixed now! ]
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
---
 net/mptcp/subflow.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
index c9cf0fdbfbdc..77cc4f585cd8 100644
--- a/net/mptcp/subflow.c
+++ b/net/mptcp/subflow.c
@@ -279,11 +279,16 @@ void mptcp_subflow_reset(struct sock *ssk)
 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
 	struct sock *sk = subflow->conn;
 
+	/* must hold: tcp_done() could drop last reference on parent */
+	sock_hold(sk);
+
 	tcp_send_active_reset(ssk, GFP_ATOMIC);
 	tcp_done(ssk);
 	if (!test_and_set_bit(MPTCP_WORK_CLOSE_SUBFLOW, &mptcp_sk(sk)->flags) &&
 	    schedule_work(&mptcp_sk(sk)->work))
-		sock_hold(sk);
+		return; /* worker will put sk for us */
+
+	sock_put(sk);
 }
 
 static void subflow_finish_connect(struct sock *sk, const struct sk_buff *skb)
-- 
2.55.0

[PATCH 5.10.y 2/3] mptcp: avoid unneeded actions on subflow reset

[Matthieu Baerts (NGI0)]

From: Paolo Abeni <pabeni@redhat.com>

commit 2b0f561f21b27c40c91ea4975268a06092bd7e9c upstream.

Once in a blue moon, the mptcp receive path can recursively call
mptcp_data_ready() via state change under unlucky error conditions, and
then try to hold the data lock again.

Break the recursion loop explicitly checking for the exceptional
condition.

Add a new flag instead of using an existing one like 'closing', to exit
early in subflow_state_change(), and explicitly flush the RX queue at
reset time.

This avoids unneeded processing to check for available data -- calling
get_mapping_status() and more on a dying subflow -- but also in error
reporting and worker scheduling.

Note that we must consume the currently peeked skb before invoking
mptcp_dss_corruption to avoid consuming it again after the eventual
reset has freed it.

Fixes: e32d262c89e2 ("mptcp: handle consistently DSS corruption")
Cc: stable@vger.kernel.org
Reported-by: Xinyang Ge <xinyang@anthropic.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260917-net-mptcp-misc-fixes-7-3-rc4-v2-1-0cf5c72667c8@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
[ Note: conflict in protocol.c, because commit e0ca4057e0ec ("mptcp:
  micro-optimize __mptcp_move_skb()") is not in this version, and is
  part of a consequent rx path refactor. The conflict is in the context,
  and is easy to resolve, "done = true" can be moved along without
  consequences. Also, the context is slightly different because there is
  no DEBUG_NET_WARN_ON_ONCE in this version, see the backport commit
  12c1676d598e ("mptcp: handle consistently DSS corruption").
  Also a conflict in protocol.h, because __unused is at a different
  number. Decrement the one from this version and add the new flag
  above. The context is also a bit different with data_avail being an
  enum, but that's without consequences here.
  Also a conflict in subflow.c, because commit 71154bbe4942 ("mptcp:
  fallback earlier on simult connection") was not needed in this
  version, and cause conflicts in the context, but that's without
  consequences here. Also, a conflict in the context, because commit
  81c1d0290160 ("mptcp: consolidate fallback and non fallback state
  machine") was not needed in this version. ]
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
---
 net/mptcp/protocol.c |  6 +++---
 net/mptcp/protocol.h |  3 ++-
 net/mptcp/subflow.c  | 11 +++++++++++
 3 files changed, 16 insertions(+), 4 deletions(-)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index 2c6aef813473..292c21713eb7 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -581,11 +581,11 @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
 			if (unlikely(map_remaining < len))
 				mptcp_dss_corruption(msk, ssk);
 		} else {
-			if (unlikely(!fin))
-				mptcp_dss_corruption(msk, ssk);
-
 			sk_eat_skb(ssk, skb);
 			done = true;
+
+			if (unlikely(!fin))
+				mptcp_dss_corruption(msk, ssk);
 		}
 
 		WRITE_ONCE(tp->copied_seq, seq);
diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
index ac064c44079d..b64a50b22d62 100644
--- a/net/mptcp/protocol.h
+++ b/net/mptcp/protocol.h
@@ -314,7 +314,8 @@ struct mptcp_subflow_context {
 		mpc_map : 1,
 		backup : 1,
 		rx_eof : 1,
-		can_ack : 1;	    /* only after processing the remote a key */
+		can_ack : 1,	    /* only after processing the remote a key */
+		resetting : 1;	    /* subflow is resetting */
 	enum mptcp_data_avail data_avail;
 	u32	remote_nonce;
 	u64	thmac;
diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
index 77cc4f585cd8..fff70a5b06db 100644
--- a/net/mptcp/subflow.c
+++ b/net/mptcp/subflow.c
@@ -282,6 +282,10 @@ void mptcp_subflow_reset(struct sock *ssk)
 	/* must hold: tcp_done() could drop last reference on parent */
 	sock_hold(sk);
 
+	subflow->resetting = 1;
+
+	/* No need to delay the actual close for to-be discarded data. */
+	__skb_queue_purge(&ssk->sk_receive_queue);
 	tcp_send_active_reset(ssk, GFP_ATOMIC);
 	tcp_done(ssk);
 	if (!test_and_set_bit(MPTCP_WORK_CLOSE_SUBFLOW, &mptcp_sk(sk)->flags) &&
@@ -1302,6 +1306,13 @@ static void subflow_state_change(struct sock *sk)
 
 	__subflow_state_change(sk);
 
+	/* Rx queue processing is unneeded, error reporting will take place at
+	 * __mptcp_close_ssk() time and subflow reset can't happen in case of
+	 * fallback: subflow_sched_work_if_closed() would be a no-op.
+	 */
+	if (subflow->resetting)
+		return;
+
 	if (subflow_simultaneous_connect(sk)) {
 		mptcp_do_fallback(sk);
 		mptcp_rcv_space_init(mptcp_sk(parent), sk);
-- 
2.55.0

[PATCH 5.10.y 3/3] mptcp: close race between scheduler and state change

[Matthieu Baerts (NGI0)]

From: Paolo Abeni <pabeni@redhat.com>

commit 42064de57fb83231fcc89663a94885f228a1ee53 upstream.

The mptcp scheduler may race with subflow sockets state change: data
transmission on the selected socket may fail and a later release could
try to use mss_now reset to 0 for a divide operation.

Address the issue by explicitly checking for the critical scenario.

Fixes: c886d70286bf ("mptcp: do not queue data on closed subflows")
Cc: stable@vger.kernel.org
Reported-by: Shardul Bankar <shardul.b@mpiricsoftware.com>
Reported-by: Xinyang Ge <xinyang@anthropic.com>
Closes: https://lore.kernel.org/20260525194828.1137119-1-shardul.b@mpiricsoftware.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260917-net-mptcp-misc-fixes-7-3-rc4-v2-2-0cf5c72667c8@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
[ Note: moved the mss_now check in previous places before commit
  d9ca1de8c0cd ("mptcp: move page frag allocation in mptcp_sendmsg()")
  which is not in this version. ]
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
---
 net/mptcp/protocol.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index 292c21713eb7..fdebc451e23a 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -1381,7 +1381,7 @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 	}
 
 	mptcp_set_timeout(sk, ssk);
-	if (copied) {
+	if (copied && mss_now) {
 		tcp_push(ssk, msg->msg_flags, mss_now, tcp_sk(ssk)->nonagle,
 			 size_goal);
 
@@ -1891,7 +1891,7 @@ static void mptcp_worker(struct work_struct *work)
 		if (!mptcp_ext_cache_refill(msk))
 			break;
 	}
-	if (copied) {
+	if (copied && mss_now) {
 		tcp_push(ssk, msg.msg_flags, mss_now, tcp_sk(ssk)->nonagle,
 			 size_goal);
 		WRITE_ONCE(msk->allow_infinite_fallback, false);
-- 
2.55.0

Re: [PATCH 5.10.y 3/3] mptcp: close race between scheduler and state change

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [High] mptcp: skipped timer reset when mss_now is 0 leads to connection stall
- [High] mptcp: allow_infinite_fallback state corruption when mss_now is 0
--

commit 4c31a1df5ba815467dd012d7f0c9b7bc2457aca9
Author: Paolo Abeni <pabeni@redhat.com>

mptcp: close race between scheduler and state change

This commit adds a check for mss_now being 0 to prevent a divide by zero
operation when selecting a subflow socket. It modifies mptcp_sendmsg and
mptcp_worker to only trigger tcp_push if both copied and mss_now are non-zero.

Link: https://patch.msgid.link/20260917-net-mptcp-misc-fixes-7-3-rc4-v2-2-0cf5c72667c8@kernel.org

> diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
> index 292c21713eb7f..fdebc451e23aa 100644
> --- a/net/mptcp/protocol.c
> +++ b/net/mptcp/protocol.c
> @@ -1381,7 +1381,7 @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
>  	}
>  
>  	mptcp_set_timeout(sk, ssk);
> -	if (copied) {
> +	if (copied && mss_now) {
>  		tcp_push(ssk, msg->msg_flags, mss_now, tcp_sk(ssk)->nonagle,
>  			 size_goal);
>  
>  		/* start the timer, if it's not pending */
>  		if (!mptcp_timer_pending(sk))
>  			mptcp_reset_timer(sk);
>  	}

[Severity: High]
Does this unintentionally skip resetting the retransmit timer when mss_now
is 0?

If multiple fragments are sent and a subsequent fragment fails, it can leave
copied > 0 but mss_now == 0. By bundling the timer initialization with the
mss_now check, this might prevent data queued to the MPTCP socket from being
retransmitted if the current subflow fails, potentially stalling the
connection.

[ ... ]

> @@ -1891,7 +1891,7 @@ static void mptcp_worker(struct work_struct *work)
>  		if (!mptcp_ext_cache_refill(msk))
>  			break;
>  	}
> -	if (copied) {
> +	if (copied && mss_now) {
>  		tcp_push(ssk, msg.msg_flags, mss_now, tcp_sk(ssk)->nonagle,
>  			 size_goal);
>  		WRITE_ONCE(msk->allow_infinite_fallback, false);
>  	}

[Severity: High]
Could skipping this block when mss_now is 0 cause allow_infinite_fallback
state corruption?

In mptcp_worker(), if a subflow failure occurs during retransmission causing
mss_now == 0, copied might still be > 0. Skipping the update to
msk->allow_infinite_fallback could leave it as true even after MPTCP data
was successfully queued. A subsequent error might then trigger an invalid
fallback to plain TCP, corrupting the connection state and stream data.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260919204002.2106015-5-matttbe@kernel.org?part=3

Patch "mptcp: close race between scheduler and state change" has been added to the 5.10-stable tree

[gregkh]

This is a note to let you know that I've just added the patch titled

    mptcp: close race between scheduler and state change

to the 5.10-stable tree which can be found at:
    http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary

The filename of the patch is:
     mptcp-close-race-between-scheduler-and-state-change.patch
and it can be found in the queue-5.10 subdirectory.

If you, or anyone else, feels it should not be added to the stable tree,
please let <stable@vger.kernel.org> know about it.


From stable+bounces-338617-greg=kroah.com@vger.kernel.org Sat Sep 19 22:40:32 2026
From: "Matthieu Baerts (NGI0)" <matttbe@kernel.org>
Date: Sat, 19 Sep 2026 22:40:06 +0200
Subject: mptcp: close race between scheduler and state change
To: mptcp@lists.linux.dev, stable@vger.kernel.org, gregkh@linuxfoundation.org
Cc: Paolo Abeni <pabeni@redhat.com>, sashal@kernel.org, Shardul Bankar <shardul.b@mpiricsoftware.com>, Xinyang Ge <xinyang@anthropic.com>, "Matthieu Baerts (NGI0)" <matttbe@kernel.org>, Jakub Kicinski <kuba@kernel.org>
Message-ID: <20260919204002.2106015-8-matttbe@kernel.org>

From: Paolo Abeni <pabeni@redhat.com>

commit 42064de57fb83231fcc89663a94885f228a1ee53 upstream.

The mptcp scheduler may race with subflow sockets state change: data
transmission on the selected socket may fail and a later release could
try to use mss_now reset to 0 for a divide operation.

Address the issue by explicitly checking for the critical scenario.

Fixes: c886d70286bf ("mptcp: do not queue data on closed subflows")
Cc: stable@vger.kernel.org
Reported-by: Shardul Bankar <shardul.b@mpiricsoftware.com>
Reported-by: Xinyang Ge <xinyang@anthropic.com>
Closes: https://lore.kernel.org/20260525194828.1137119-1-shardul.b@mpiricsoftware.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260917-net-mptcp-misc-fixes-7-3-rc4-v2-2-0cf5c72667c8@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
[ Note: moved the mss_now check in previous places before commit
  d9ca1de8c0cd ("mptcp: move page frag allocation in mptcp_sendmsg()")
  which is not in this version. ]
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
 net/mptcp/protocol.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -1381,7 +1381,7 @@ restart:
 	}
 
 	mptcp_set_timeout(sk, ssk);
-	if (copied) {
+	if (copied && mss_now) {
 		tcp_push(ssk, msg->msg_flags, mss_now, tcp_sk(ssk)->nonagle,
 			 size_goal);
 
@@ -1891,7 +1891,7 @@ static void mptcp_worker(struct work_str
 		if (!mptcp_ext_cache_refill(msk))
 			break;
 	}
-	if (copied) {
+	if (copied && mss_now) {
 		tcp_push(ssk, msg.msg_flags, mss_now, tcp_sk(ssk)->nonagle,
 			 size_goal);
 		WRITE_ONCE(msk->allow_infinite_fallback, false);


Patches currently in stable-queue which might be from matttbe@kernel.org are

queue-5.10/mptcp-hold-mptcp-socket-before-calling-tcp_done.patch
queue-5.10/mptcp-syncookies-remember-the-request-backup-flag.patch
queue-5.10/mptcp-close-race-between-scheduler-and-state-change.patch
queue-5.10/mptcp-avoid-unneeded-actions-on-subflow-reset.patch

Patch "mptcp: hold mptcp socket before calling tcp_done" has been added to the 5.10-stable tree

[gregkh]

This is a note to let you know that I've just added the patch titled

    mptcp: hold mptcp socket before calling tcp_done

to the 5.10-stable tree which can be found at:
    http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary

The filename of the patch is:
     mptcp-hold-mptcp-socket-before-calling-tcp_done.patch
and it can be found in the queue-5.10 subdirectory.

If you, or anyone else, feels it should not be added to the stable tree,
please let <stable@vger.kernel.org> know about it.


From stable+bounces-338615-greg=kroah.com@vger.kernel.org Sat Sep 19 22:40:23 2026
From: "Matthieu Baerts (NGI0)" <matttbe@kernel.org>
Date: Sat, 19 Sep 2026 22:40:04 +0200
Subject: mptcp: hold mptcp socket before calling tcp_done
To: mptcp@lists.linux.dev, stable@vger.kernel.org, gregkh@linuxfoundation.org
Cc: Florian Westphal <fw@strlen.de>, sashal@kernel.org, Matthieu Baerts <matthieu.baerts@tessares.net>, Mat Martineau <mathew.j.martineau@linux.intel.com>, Jakub Kicinski <kuba@kernel.org>, "Matthieu Baerts (NGI0)" <matttbe@kernel.org>
Message-ID: <20260919204002.2106015-6-matttbe@kernel.org>

From: Florian Westphal <fw@strlen.de>

commit ab82e996a1fa1b9ae514fa357d9ce8df62321157 upstream.

When processing options from tcp reset path its possible that
tcp_done(ssk) drops the last reference on the mptcp socket which
results in use-after-free.

Reviewed-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Stable-dep-of: 2b0f561f21b2 ("mptcp: avoid unneeded actions on subflow reset")
[ Conflicts in subflow.c, because commit 3ba14528684f ("mptcp: avoid
  setting TCP_CLOSE state twice") has already been backported and also
  had the same conflict: this commit here should have been backported
  first. Fixed now! ]
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
 net/mptcp/subflow.c |    7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

--- a/net/mptcp/subflow.c
+++ b/net/mptcp/subflow.c
@@ -279,11 +279,16 @@ void mptcp_subflow_reset(struct sock *ss
 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
 	struct sock *sk = subflow->conn;
 
+	/* must hold: tcp_done() could drop last reference on parent */
+	sock_hold(sk);
+
 	tcp_send_active_reset(ssk, GFP_ATOMIC);
 	tcp_done(ssk);
 	if (!test_and_set_bit(MPTCP_WORK_CLOSE_SUBFLOW, &mptcp_sk(sk)->flags) &&
 	    schedule_work(&mptcp_sk(sk)->work))
-		sock_hold(sk);
+		return; /* worker will put sk for us */
+
+	sock_put(sk);
 }
 
 static void subflow_finish_connect(struct sock *sk, const struct sk_buff *skb)


Patches currently in stable-queue which might be from matttbe@kernel.org are

queue-5.10/mptcp-hold-mptcp-socket-before-calling-tcp_done.patch
queue-5.10/mptcp-syncookies-remember-the-request-backup-flag.patch
queue-5.10/mptcp-close-race-between-scheduler-and-state-change.patch
queue-5.10/mptcp-avoid-unneeded-actions-on-subflow-reset.patch

Patch "mptcp: avoid unneeded actions on subflow reset" has been added to the 5.10-stable tree

[gregkh]

This is a note to let you know that I've just added the patch titled

    mptcp: avoid unneeded actions on subflow reset

to the 5.10-stable tree which can be found at:
    http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary

The filename of the patch is:
     mptcp-avoid-unneeded-actions-on-subflow-reset.patch
and it can be found in the queue-5.10 subdirectory.

If you, or anyone else, feels it should not be added to the stable tree,
please let <stable@vger.kernel.org> know about it.


From stable+bounces-338616-greg=kroah.com@vger.kernel.org Sat Sep 19 22:40:25 2026
From: "Matthieu Baerts (NGI0)" <matttbe@kernel.org>
Date: Sat, 19 Sep 2026 22:40:05 +0200
Subject: mptcp: avoid unneeded actions on subflow reset
To: mptcp@lists.linux.dev, stable@vger.kernel.org, gregkh@linuxfoundation.org
Cc: Paolo Abeni <pabeni@redhat.com>, sashal@kernel.org, Xinyang Ge <xinyang@anthropic.com>, "Matthieu Baerts (NGI0)" <matttbe@kernel.org>, Jakub Kicinski <kuba@kernel.org>
Message-ID: <20260919204002.2106015-7-matttbe@kernel.org>

From: Paolo Abeni <pabeni@redhat.com>

commit 2b0f561f21b27c40c91ea4975268a06092bd7e9c upstream.

Once in a blue moon, the mptcp receive path can recursively call
mptcp_data_ready() via state change under unlucky error conditions, and
then try to hold the data lock again.

Break the recursion loop explicitly checking for the exceptional
condition.

Add a new flag instead of using an existing one like 'closing', to exit
early in subflow_state_change(), and explicitly flush the RX queue at
reset time.

This avoids unneeded processing to check for available data -- calling
get_mapping_status() and more on a dying subflow -- but also in error
reporting and worker scheduling.

Note that we must consume the currently peeked skb before invoking
mptcp_dss_corruption to avoid consuming it again after the eventual
reset has freed it.

Fixes: e32d262c89e2 ("mptcp: handle consistently DSS corruption")
Cc: stable@vger.kernel.org
Reported-by: Xinyang Ge <xinyang@anthropic.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260917-net-mptcp-misc-fixes-7-3-rc4-v2-1-0cf5c72667c8@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
[ Note: conflict in protocol.c, because commit e0ca4057e0ec ("mptcp:
  micro-optimize __mptcp_move_skb()") is not in this version, and is
  part of a consequent rx path refactor. The conflict is in the context,
  and is easy to resolve, "done = true" can be moved along without
  consequences. Also, the context is slightly different because there is
  no DEBUG_NET_WARN_ON_ONCE in this version, see the backport commit
  12c1676d598e ("mptcp: handle consistently DSS corruption").
  Also a conflict in protocol.h, because __unused is at a different
  number. Decrement the one from this version and add the new flag
  above. The context is also a bit different with data_avail being an
  enum, but that's without consequences here.
  Also a conflict in subflow.c, because commit 71154bbe4942 ("mptcp:
  fallback earlier on simult connection") was not needed in this
  version, and cause conflicts in the context, but that's without
  consequences here. Also, a conflict in the context, because commit
  81c1d0290160 ("mptcp: consolidate fallback and non fallback state
  machine") was not needed in this version. ]
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
 net/mptcp/protocol.c |    6 +++---
 net/mptcp/protocol.h |    3 ++-
 net/mptcp/subflow.c  |   11 +++++++++++
 3 files changed, 16 insertions(+), 4 deletions(-)

--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -581,11 +581,11 @@ static bool __mptcp_move_skbs_from_subfl
 			if (unlikely(map_remaining < len))
 				mptcp_dss_corruption(msk, ssk);
 		} else {
-			if (unlikely(!fin))
-				mptcp_dss_corruption(msk, ssk);
-
 			sk_eat_skb(ssk, skb);
 			done = true;
+
+			if (unlikely(!fin))
+				mptcp_dss_corruption(msk, ssk);
 		}
 
 		WRITE_ONCE(tp->copied_seq, seq);
--- a/net/mptcp/protocol.h
+++ b/net/mptcp/protocol.h
@@ -314,7 +314,8 @@ struct mptcp_subflow_context {
 		mpc_map : 1,
 		backup : 1,
 		rx_eof : 1,
-		can_ack : 1;	    /* only after processing the remote a key */
+		can_ack : 1,	    /* only after processing the remote a key */
+		resetting : 1;	    /* subflow is resetting */
 	enum mptcp_data_avail data_avail;
 	u32	remote_nonce;
 	u64	thmac;
--- a/net/mptcp/subflow.c
+++ b/net/mptcp/subflow.c
@@ -282,6 +282,10 @@ void mptcp_subflow_reset(struct sock *ss
 	/* must hold: tcp_done() could drop last reference on parent */
 	sock_hold(sk);
 
+	subflow->resetting = 1;
+
+	/* No need to delay the actual close for to-be discarded data. */
+	__skb_queue_purge(&ssk->sk_receive_queue);
 	tcp_send_active_reset(ssk, GFP_ATOMIC);
 	tcp_done(ssk);
 	if (!test_and_set_bit(MPTCP_WORK_CLOSE_SUBFLOW, &mptcp_sk(sk)->flags) &&
@@ -1302,6 +1306,13 @@ static void subflow_state_change(struct
 
 	__subflow_state_change(sk);
 
+	/* Rx queue processing is unneeded, error reporting will take place at
+	 * __mptcp_close_ssk() time and subflow reset can't happen in case of
+	 * fallback: subflow_sched_work_if_closed() would be a no-op.
+	 */
+	if (subflow->resetting)
+		return;
+
 	if (subflow_simultaneous_connect(sk)) {
 		mptcp_do_fallback(sk);
 		mptcp_rcv_space_init(mptcp_sk(parent), sk);


Patches currently in stable-queue which might be from matttbe@kernel.org are

queue-5.10/mptcp-hold-mptcp-socket-before-calling-tcp_done.patch
queue-5.10/mptcp-syncookies-remember-the-request-backup-flag.patch
queue-5.10/mptcp-close-race-between-scheduler-and-state-change.patch
queue-5.10/mptcp-avoid-unneeded-actions-on-subflow-reset.patch