Patch "mptcp: annotate data-races around subflow->fully_established" has been added to the 6.12-stable tree

Posted by gregkh@linuxfoundation.org 5 days, 2 hours ago


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

    mptcp: annotate data-races around subflow->fully_established

to the 6.12-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-annotate-data-races-around-subflow-fully_established.patch
and it can be found in the queue-6.12 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-338598-greg=kroah.com@vger.kernel.org Sat Sep 19 21:57:10 2026
From: "Matthieu Baerts (NGI0)" <matttbe@kernel.org>
Date: Sat, 19 Sep 2026 21:56:44 +0200
Subject: mptcp: annotate data-races around subflow->fully_established
To: mptcp@lists.linux.dev, stable@vger.kernel.org, gregkh@linuxfoundation.org
Cc: Gang Yan <yangang@kylinos.cn>, sashal@kernel.org, "Matthieu Baerts (NGI0)" <matttbe@kernel.org>, Jakub Kicinski <kuba@kernel.org>
Message-ID: <20260919195642.1977885-8-matttbe@kernel.org>

From: Gang Yan <yangang@kylinos.cn>

commit 581c8cbfa934aaa555daa4e843242fcecc160f05 upstream.

We introduce the same handling for potential data races with the
'fully_established' flag in subflow as previously done for
msk->fully_established.

Additionally, we make a crucial change: convert the subflow's
'fully_established' from 'bit_field' to 'bool' type. This is
necessary because methods for avoiding data races don't work well
with 'bit_field'. Specifically, the 'READ_ONCE' needs to know
the size of the variable being accessed, which is not supported in
'bit_field'. Also, 'test_bit' expect the address of 'bit_field'.

Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/516
Signed-off-by: Gang Yan <yangang@kylinos.cn>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20241021-net-next-mptcp-misc-6-13-v1-2-1ef02746504a@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Stable-dep-of: 2b0f561f21b2 ("mptcp: avoid unneeded actions on subflow reset")
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
 net/mptcp/options.c  |    4 ++--
 net/mptcp/protocol.c |    2 +-
 net/mptcp/protocol.h |    2 +-
 3 files changed, 4 insertions(+), 4 deletions(-)

--- a/net/mptcp/options.c
+++ b/net/mptcp/options.c
@@ -532,7 +532,7 @@ static bool mptcp_established_options_mp
 		return false;
 
 	/* MPC/MPJ needed only on 3rd ack packet, DATA_FIN and TCP shutdown take precedence */
-	if (subflow->fully_established || snd_data_fin_enable ||
+	if (READ_ONCE(subflow->fully_established) || snd_data_fin_enable ||
 	    subflow->snd_isn != TCP_SKB_CB(skb)->seq ||
 	    sk->sk_state != TCP_ESTABLISHED)
 		return false;
@@ -999,7 +999,7 @@ static bool check_fully_established(stru
 	/* here we can process OoO, in-window pkts, only in-sequence 4th ack
 	 * will make the subflow fully established
 	 */
-	if (likely(subflow->fully_established)) {
+	if (likely(READ_ONCE(subflow->fully_established))) {
 		/* on passive sockets, check for 3rd ack retransmission
 		 * note that msk is always set by subflow_syn_recv_sock()
 		 * for mp_join subflows
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -3685,7 +3685,7 @@ static void schedule_3rdack_retransmissi
 	struct tcp_sock *tp = tcp_sk(ssk);
 	unsigned long timeout;
 
-	if (mptcp_subflow_ctx(ssk)->fully_established)
+	if (READ_ONCE(mptcp_subflow_ctx(ssk)->fully_established))
 		return;
 
 	/* reschedule with a timeout above RTT, as we must look only for drop */
--- a/net/mptcp/protocol.h
+++ b/net/mptcp/protocol.h
@@ -545,7 +545,7 @@ struct mptcp_subflow_context {
 		is_mptfo : 1,	    /* subflow is doing TFO */
 		close_event_done : 1,       /* has done the post-closed part */
 		mpc_drop : 1,	    /* the MPC option has been dropped in a rtx */
-		__unused : 8;
+		__unused : 9;
 	bool	data_avail;
 	bool	scheduled;
 	bool	pm_listener;	    /* a listener managed by the kernel PM? */


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

queue-6.12/mptcp-annotate-data-races-around-subflow-fully_established.patch
queue-6.12/selftests-mptcp-fix-an-uaf-in-mptcp_connect.c.patch
queue-6.12/mptcp-fix-bad-accounting-in-__mptcp_subflow_push_pending.patch
queue-6.12/mptcp-syncookies-remember-the-request-backup-flag.patch
queue-6.12/mptcp-remove-unneeded-read_once-annotation.patch
queue-6.12/mptcp-close-race-between-scheduler-and-state-change.patch
queue-6.12/tcp-use-gfp_atomic-in-tcp_send_active_reset.patch
queue-6.12/mptcp-avoid-unneeded-actions-on-subflow-reset.patch
queue-6.12/mptcp-options-handle-mpc-data-csum-reqd-no-csum.patch
queue-6.12/mptcp-subflow-no-need-to-copy-thmac-during-ulp_clone.patch

Re: Patch "mptcp: annotate data-races around subflow->fully_established" has been added to the 6.12-stable tree

Posted by sashiko-bot@kernel.org 5 days, 1 hour ago

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Low] mptcp: build failure due to READ_ONCE() on bit-field
--

commit 2cede4d0675228e6db5b209f56d2a298cf7cb2e9
Author: gregkh@linuxfoundation.org <gregkh@linuxfoundation.org>
Subject: Patch "mptcp: annotate data-races around subflow->fully_established" has been added to the 6.12-stable tree

This commit annotates data races around the fully_established flag in the
mptcp subflow context. It aims to convert the flag from a bit-field to a
boolean type to safely use READ_ONCE() and WRITE_ONCE().

Link: https://patch.msgid.link/20241021-net-next-mptcp-misc-6-13-v1-2-1ef02746504a@kernel.org

>     Additionally, we make a crucial change: convert the subflow's
>     'fully_established' from 'bit_field' to 'bool' type. This is
>     necessary because methods for avoiding data races don't work well
>     with 'bit_field'. Specifically, the 'READ_ONCE' needs to know
>     the size of the variable being accessed, which is not supported in
>     'bit_field'. Also, 'test_bit' expect the address of 'bit_field'.

[Severity: Low]
Was the hunk converting this field to a boolean dropped during the backport?
It appears fully_established is still defined as a bit-field in
struct mptcp_subflow_context.

Additionally, the write side update was missed in the backport:

net/mptcp/subflow.c:__mptcp_subflow_fully_established() {
    ...
	subflow->fully_established = 1;
    ...
}

This lacks the WRITE_ONCE() annotation and continues to perform an unsafe
read-modify-write on the shared bit-field word.

> diff --git a/net/mptcp/options.c b/net/mptcp/options.c
> index 370c3836b7712..1603b3702e220 100644
> --- a/net/mptcp/options.c
> +++ b/net/mptcp/options.c
> @@ -461,7 +461,7 @@ static bool mptcp_established_options_mp(struct sock *sk, struct sk_buff *skb,
>  		return false;
>  
>  	/* MPC/MPJ needed only on 3rd ack packet, DATA_FIN and TCP shutdown take precedence */
> -	if (subflow->fully_established || snd_data_fin_enable ||
> +	if (READ_ONCE(subflow->fully_established) || snd_data_fin_enable ||

[Severity: Low]
Does this trigger a compiler error? Since the struct modification was missed
and fully_established remains a bit-field, applying READ_ONCE() here attempts
to take the address of a bit-field, which is syntactically invalid in C.

[ ... ]

> diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
> index 568a72702b080..e9dba0708ee70 100644
> --- a/net/mptcp/protocol.h
> +++ b/net/mptcp/protocol.h
> @@ -532,7 +532,7 @@ struct mptcp_subflow_context {
>  		is_mptfo : 1,	    /* subflow is doing TFO */
>  		close_event_done : 1,       /* has done the post-closed part */
>  		mpc_drop : 1,	    /* the MPC option has been dropped in a rtx */
> -		__unused : 8;
> +		__unused : 9;

[Severity: Low]
Does this increase the total bit-field group to 33 bits, causing a boundary
overflow? 

Since fully_established wasn't actually removed from the bit-field group in
this backport, incrementing the unused bits to 9 pushes the total beyond a
32-bit word size.

>  	bool	data_avail;
>  	bool	scheduled;
>  	bool	pm_listener;	    /* a listener managed by the kernel PM? */

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/2026092037-islamic-default-0aca@gregkh?part=1

Re: Patch "mptcp: annotate data-races around subflow->fully_established" has been added to the 6.12-stable tree

Posted by Matthieu Baerts 5 days ago

Hello,

On 20/09/2026 09:48, sashiko-bot@kernel.org wrote:
> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
> - [Low] mptcp: build failure due to READ_ONCE() on bit-field

Mmh, I missed the fact this commit has already been backported in commit
477aa7116a76 ("mptcp: annotate data-races around subflow
->fully_established"), but mostly reverted by caa4a79f74f3 ("mptcp:
remove unneeded READ_ONCE() annotation").

I don't know what's better here: either we leave it like that, and I can
fix the size of the bit-field. Or we drop patch 1 and 2, and I resolve
the small conflict on patch 3/5: commit 2b0f561f21b2 ("mptcp: avoid
unneeded actions on subflow reset"). I can send a v2, and we can drop it
if the other solution has been preferred.

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

Re: Patch "mptcp: annotate data-races around subflow->fully_established" has been added to the 6.12-stable tree

Posted by Greg KH 4 days, 20 hours ago

On Sun, Sep 20, 2026 at 11:33:47AM +0200, Matthieu Baerts wrote:
> Hello,
> 
> On 20/09/2026 09:48, sashiko-bot@kernel.org wrote:
> > Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
> > - [Low] mptcp: build failure due to READ_ONCE() on bit-field
> 
> Mmh, I missed the fact this commit has already been backported in commit
> 477aa7116a76 ("mptcp: annotate data-races around subflow
> ->fully_established"), but mostly reverted by caa4a79f74f3 ("mptcp:
> remove unneeded READ_ONCE() annotation").
> 
> I don't know what's better here: either we leave it like that, and I can
> fix the size of the bit-field. Or we drop patch 1 and 2, and I resolve
> the small conflict on patch 3/5: commit 2b0f561f21b2 ("mptcp: avoid
> unneeded actions on subflow reset"). I can send a v2, and we can drop it
> if the other solution has been preferred.

Your v2 looks good, thanks for that, I'll go replace the patches with
those.

greg k-h

Re: Patch "mptcp: annotate data-races around subflow->fully_established" has been added to the 6.12-stable tree

Posted by Matthieu Baerts 4 days, 20 hours ago

On 20/09/2026 15:10, Greg KH wrote:
> On Sun, Sep 20, 2026 at 11:33:47AM +0200, Matthieu Baerts wrote:
>> Hello,
>>
>> On 20/09/2026 09:48, sashiko-bot@kernel.org wrote:
>>> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
>>> - [Low] mptcp: build failure due to READ_ONCE() on bit-field
>>
>> Mmh, I missed the fact this commit has already been backported in commit
>> 477aa7116a76 ("mptcp: annotate data-races around subflow
>> ->fully_established"), but mostly reverted by caa4a79f74f3 ("mptcp:
>> remove unneeded READ_ONCE() annotation").
>>
>> I don't know what's better here: either we leave it like that, and I can
>> fix the size of the bit-field. Or we drop patch 1 and 2, and I resolve
>> the small conflict on patch 3/5: commit 2b0f561f21b2 ("mptcp: avoid
>> unneeded actions on subflow reset"). I can send a v2, and we can drop it
>> if the other solution has been preferred.
> 
> Your v2 looks good, thanks for that, I'll go replace the patches with
> those.

Thank you! And sorry for the extra work :)

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