1. Patchew
  2. MPTCP
  3. View series

[PATCH mptcp-net v2] mptcp: fix data-race in __mptcp_retrans / mptcp_incoming_options

Matthieu Baerts (NGI0) posted 1 patch 4 days, 19 hours ago
Patches applied successfully (tree, apply log)
git fetch https://github.com/multipath-tcp/mptcp_net-next tags/patchew/20260317-mptcp-data-race-snd._5Funa-v2-1-2caac60de92a@kernel.org
net/mptcp/protocol.h | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
[PATCH mptcp-net v2] mptcp: fix data-race in __mptcp_retrans / mptcp_incoming_options
Posted by Matthieu Baerts (NGI0) 4 days, 19 hours ago 
SyzKaller found this data-race:

  BUG: KCSAN: data-race in __mptcp_retrans / mptcp_incoming_options

  write (marked) to 0xffff888015e8e5f0 of 8 bytes by interrupt on cpu 0:
   __mptcp_snd_una_update net/mptcp/options.c:1055 [inline]
   mptcp_incoming_options+0x6a3/0x1ac0 net/mptcp/options.c:1183
   tcp_data_queue+0x101b/0x2440 net/ipv4/tcp_input.c:5583
   tcp_rcv_established+0x684/0x1fc0 net/ipv4/tcp_input.c:6654
   tcp_v4_do_rcv+0x35c/0x690 net/ipv4/tcp_ipv4.c:1866
   tcp_v4_rcv+0x1d91/0x25a0 net/ipv4/tcp_ipv4.c:2263
   ip_protocol_deliver_rcu+0x46/0x280 net/ipv4/ip_input.c:207
   ip_local_deliver_finish+0x190/0x270 net/ipv4/ip_input.c:241
   NF_HOOK include/linux/netfilter.h:318 [inline]
   NF_HOOK include/linux/netfilter.h:312 [inline]
   ip_local_deliver+0xe3/0x210 net/ipv4/ip_input.c:262
   dst_input include/net/dst.h:480 [inline]
   ip_rcv_finish net/ipv4/ip_input.c:492 [inline]
   NF_HOOK include/linux/netfilter.h:318 [inline]
   NF_HOOK include/linux/netfilter.h:312 [inline]
   ip_rcv+0x200/0x220 net/ipv4/ip_input.c:612
   __netif_receive_skb_one_core+0xeb/0x110 net/core/dev.c:6178
   __netif_receive_skb+0x1f/0xc0 net/core/dev.c:6291
   process_backlog+0x168/0x360 net/core/dev.c:6642
   __napi_poll+0x71/0x460 net/core/dev.c:7706
   napi_poll net/core/dev.c:7769 [inline]
   net_rx_action+0x6f8/0x810 net/core/dev.c:7926
   handle_softirqs+0xc9/0x2e0 kernel/softirq.c:622
   run_ksoftirqd kernel/softirq.c:1063 [inline]
   run_ksoftirqd+0x20/0x30 kernel/softirq.c:1055
   smpboot_thread_fn+0x287/0x520 kernel/smpboot.c:160
   kthread+0x1f2/0x240 kernel/kthread.c:436
   ret_from_fork+0x321/0x440 arch/x86/kernel/process.c:158
   ret_from_fork_asm+0x1a/0x30 arch/x86/entry/entry_64.S:245

  read to 0xffff888015e8e5f0 of 8 bytes by task 24 on cpu 1:
   mptcp_rtx_head net/mptcp/protocol.h:487 [inline]
   __mptcp_retrans+0x169/0x8f0 net/mptcp/protocol.c:2759
   mptcp_worker+0x6a6/0xb30 net/mptcp/protocol.c:2980
   process_one_work+0x3ee/0x970 kernel/workqueue.c:3275
   process_scheduled_works kernel/workqueue.c:3358 [inline]
   worker_thread+0x3c3/0x730 kernel/workqueue.c:3439
   kthread+0x1f2/0x240 kernel/kthread.c:436
   ret_from_fork+0x321/0x440 arch/x86/kernel/process.c:158
   ret_from_fork_asm+0x1a/0x30 arch/x86/entry/entry_64.S:245

  value changed: 0x0b17a4285ae6137d -> 0x0b17a4285b078905

It looks like msk->snd_una was being modified in __mptcp_snd_una_update
under the msk data lock (spin lock), while being accessed in
mptcp_rtx_head() under a different lock: the msk socket lock.

Annotate access to msk->snd_una in mptcp_rtx_head() to prevent such
issue.

Fixes: 64b9cea7a0af ("mptcp: fix spurious retransmissions")
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
---
- Link to v1: https://lore.kernel.org/r/20260316-mptcp-data-race-snd_una-v1-1-221704522c67@kernel.org
---
v2:
 - mention under which locks the field is being updated and read (Paolo)
 - add comment where snd_una is defined to help future changes (Paolo)
---
 net/mptcp/protocol.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
index f5d4d7d030f2..9ff23c858077 100644
--- a/net/mptcp/protocol.h
+++ b/net/mptcp/protocol.h
@@ -300,7 +300,7 @@ struct mptcp_sock {
 						 * protection
 						 */
 	u64		bytes_acked;
-	u64		snd_una;
+	u64		snd_una;		/* updated under the msk data lock, lockless read */
 	u64		wnd_end;
 	u32		last_data_sent;
 	u32		last_data_recv;
@@ -484,7 +484,7 @@ static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
 {
 	struct mptcp_sock *msk = mptcp_sk(sk);
 
-	if (msk->snd_una == msk->snd_nxt)
+	if (READ_ONCE(msk->snd_una) == msk->snd_nxt)
 		return NULL;
 
 	return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);

---
base-commit: 767b686e12d42196211fe3efef07310ae2ce382a
change-id: 20260316-mptcp-data-race-snd_una-6bbe4139b228

Best regards,
-- 
Matthieu Baerts (NGI0) <matttbe@kernel.org>
Re: [PATCH mptcp-net v2] mptcp: fix data-race in __mptcp_retrans / mptcp_incoming_options
Posted by Geliang Tang 3 days, 22 hours ago 
Hi Matt,

Thanks for this v2.

On Tue, 2026-03-17 at 11:52 +0100, Matthieu Baerts (NGI0) wrote:
> SyzKaller found this data-race:
> 
>   BUG: KCSAN: data-race in __mptcp_retrans / mptcp_incoming_options
> 
>   write (marked) to 0xffff888015e8e5f0 of 8 bytes by interrupt on cpu
> 0:
>    __mptcp_snd_una_update net/mptcp/options.c:1055 [inline]
>    mptcp_incoming_options+0x6a3/0x1ac0 net/mptcp/options.c:1183
>    tcp_data_queue+0x101b/0x2440 net/ipv4/tcp_input.c:5583
>    tcp_rcv_established+0x684/0x1fc0 net/ipv4/tcp_input.c:6654
>    tcp_v4_do_rcv+0x35c/0x690 net/ipv4/tcp_ipv4.c:1866
>    tcp_v4_rcv+0x1d91/0x25a0 net/ipv4/tcp_ipv4.c:2263
>    ip_protocol_deliver_rcu+0x46/0x280 net/ipv4/ip_input.c:207
>    ip_local_deliver_finish+0x190/0x270 net/ipv4/ip_input.c:241
>    NF_HOOK include/linux/netfilter.h:318 [inline]
>    NF_HOOK include/linux/netfilter.h:312 [inline]
>    ip_local_deliver+0xe3/0x210 net/ipv4/ip_input.c:262
>    dst_input include/net/dst.h:480 [inline]
>    ip_rcv_finish net/ipv4/ip_input.c:492 [inline]
>    NF_HOOK include/linux/netfilter.h:318 [inline]
>    NF_HOOK include/linux/netfilter.h:312 [inline]
>    ip_rcv+0x200/0x220 net/ipv4/ip_input.c:612
>    __netif_receive_skb_one_core+0xeb/0x110 net/core/dev.c:6178
>    __netif_receive_skb+0x1f/0xc0 net/core/dev.c:6291
>    process_backlog+0x168/0x360 net/core/dev.c:6642
>    __napi_poll+0x71/0x460 net/core/dev.c:7706
>    napi_poll net/core/dev.c:7769 [inline]
>    net_rx_action+0x6f8/0x810 net/core/dev.c:7926
>    handle_softirqs+0xc9/0x2e0 kernel/softirq.c:622
>    run_ksoftirqd kernel/softirq.c:1063 [inline]
>    run_ksoftirqd+0x20/0x30 kernel/softirq.c:1055
>    smpboot_thread_fn+0x287/0x520 kernel/smpboot.c:160
>    kthread+0x1f2/0x240 kernel/kthread.c:436
>    ret_from_fork+0x321/0x440 arch/x86/kernel/process.c:158
>    ret_from_fork_asm+0x1a/0x30 arch/x86/entry/entry_64.S:245
> 
>   read to 0xffff888015e8e5f0 of 8 bytes by task 24 on cpu 1:
>    mptcp_rtx_head net/mptcp/protocol.h:487 [inline]
>    __mptcp_retrans+0x169/0x8f0 net/mptcp/protocol.c:2759
>    mptcp_worker+0x6a6/0xb30 net/mptcp/protocol.c:2980
>    process_one_work+0x3ee/0x970 kernel/workqueue.c:3275
>    process_scheduled_works kernel/workqueue.c:3358 [inline]
>    worker_thread+0x3c3/0x730 kernel/workqueue.c:3439
>    kthread+0x1f2/0x240 kernel/kthread.c:436
>    ret_from_fork+0x321/0x440 arch/x86/kernel/process.c:158
>    ret_from_fork_asm+0x1a/0x30 arch/x86/entry/entry_64.S:245
> 
>   value changed: 0x0b17a4285ae6137d -> 0x0b17a4285b078905
> 
> It looks like msk->snd_una was being modified in
> __mptcp_snd_una_update
> under the msk data lock (spin lock), while being accessed in
> mptcp_rtx_head() under a different lock: the msk socket lock.
> 
> Annotate access to msk->snd_una in mptcp_rtx_head() to prevent such
> issue.
> 
> Fixes: 64b9cea7a0af ("mptcp: fix spurious retransmissions")
> Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>

LGTM.

    Reviewed-by: Geliang Tang <geliang@kernel.org>

> ---
> - Link to v1:
> https://lore.kernel.org/r/20260316-mptcp-data-race-snd_una-v1-1-221704522c67@kernel.org
> ---
> v2:
>  - mention under which locks the field is being updated and read
> (Paolo)
>  - add comment where snd_una is defined to help future changes
> (Paolo)
> ---
>  net/mptcp/protocol.h | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
> index f5d4d7d030f2..9ff23c858077 100644
> --- a/net/mptcp/protocol.h
> +++ b/net/mptcp/protocol.h
> @@ -300,7 +300,7 @@ struct mptcp_sock {
>  						 * protection
>  						 */
>  	u64		bytes_acked;
> -	u64		snd_una;
> +	u64		snd_una;		/* updated under the
> msk data lock, lockless read */

CI complains that this line is too long:

 WARNING: line length of 100 exceeds 80 columns
 #78: FILE: net/mptcp/protocol.h:303:

Please fix it when merging this patch.

Thanks,
-Geliang

>  	u64		wnd_end;
>  	u32		last_data_sent;
>  	u32		last_data_recv;
> @@ -484,7 +484,7 @@ static inline struct mptcp_data_frag
> *mptcp_rtx_head(struct sock *sk)
>  {
>  	struct mptcp_sock *msk = mptcp_sk(sk);
>  
> -	if (msk->snd_una == msk->snd_nxt)
> +	if (READ_ONCE(msk->snd_una) == msk->snd_nxt)
>  		return NULL;
>  
>  	return list_first_entry_or_null(&msk->rtx_queue, struct
> mptcp_data_frag, list);
> 
> ---
> base-commit: 767b686e12d42196211fe3efef07310ae2ce382a
> change-id: 20260316-mptcp-data-race-snd_una-6bbe4139b228
> 
> Best regards,
Re: [PATCH mptcp-net v2] mptcp: fix data-race in __mptcp_retrans / mptcp_incoming_options
Posted by MPTCP CI 4 days, 18 hours ago 
Hi Matthieu,

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): Unstable: 2 failed test(s): packetdrill_dss packetdrill_mp_capable 🔴
- 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/23191463243

Initiator: Patchew Applier
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/6d0a7619439b
Patchwork: https://patchwork.kernel.org/project/mptcp/list/?series=1067924


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)

Patchew 2.1-devel-ea86937

© 2016 - 2026 Red Hat, Inc.