syzbot reported a KMSAN uninit-value in mptcp_write_data_fin() reached
via tcp_send_ack() from the subflow shutdown path:

  BUG: KMSAN: uninit-value in mptcp_write_data_fin net/mptcp/options.c:542
   mptcp_established_options_dss net/mptcp/options.c:590
   mptcp_established_options net/mptcp/options.c:874
   tcp_established_options net/ipv4/tcp_output.c:1192
   __tcp_transmit_skb net/ipv4/tcp_output.c:1575
   __tcp_send_ack net/ipv4/tcp_output.c:4499
   tcp_send_ack net/ipv4/tcp_output.c:4505
   mptcp_subflow_shutdown net/mptcp/protocol.c:3137
   mptcp_check_send_data_fin net/mptcp/protocol.c:3218
   __mptcp_wr_shutdown net/mptcp/protocol.c:3234
   __mptcp_close net/mptcp/protocol.c:3313

  Local variable opts created at:
   __tcp_transmit_skb net/ipv4/tcp_output.c:1536

__tcp_transmit_skb() partially clears its on-stack tcp_out_options via
"memset(&opts.cleared, ...)" which does not cover opts.mptcp, so MPTCP
is responsible for initializing opts->ext_copy before use.

mptcp_established_options_dss() only writes opts->ext_copy when an
MPTCP extension is attached to the skb ("opts->ext_copy = *mpext").
When tcp_send_ack() builds a fresh pure-ACK skb during subflow
shutdown, the skb has no extension and that assignment is skipped, but
the function still calls mptcp_write_data_fin(&opts->ext_copy) because
snd_data_fin_enable is set. mptcp_write_data_fin() then reads
ext->use_map to choose between writing a synthetic DATA_FIN mapping or
extending an existing one, and that read hits uninitialized stack
memory.

Zero opts->ext_copy at the top of mptcp_established_options_dss() so
mptcp_write_data_fin() takes its synthetic-mapping branch on clean
memory.

Fixes: 9c29e3615274 ("mptcp: fix DSS map generation on fin retransmission")
Reported-by: syzbot+ff020673c5e3d94d9478@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=ff020673c5e3d94d9478
Signed-off-by: Deepanshu Kartikey <kartikey406@gmail.com>
---
 net/mptcp/options.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/net/mptcp/options.c b/net/mptcp/options.c
index 8a1c5698983c..3fb914196987 100644
--- a/net/mptcp/options.c
+++ b/net/mptcp/options.c
@@ -572,6 +572,7 @@ static bool mptcp_established_options_dss(struct sock *sk, struct sk_buff *skb,
 	bool ret = false;
 	u64 ack_seq;
 
+	memset(&opts->ext_copy, 0, sizeof(opts->ext_copy));
 	opts->csum_reqd = READ_ONCE(msk->csum_enabled);
 	mpext = skb ? mptcp_get_ext(skb) : NULL;
 
-- 
2.43.0

Hi Deepanshu,

On 04/05/2026 02:31, Deepanshu Kartikey wrote:
> syzbot reported a KMSAN uninit-value in mptcp_write_data_fin() reached
> via tcp_send_ack() from the subflow shutdown path:
> 
>   BUG: KMSAN: uninit-value in mptcp_write_data_fin net/mptcp/options.c:542
>    mptcp_established_options_dss net/mptcp/options.c:590
>    mptcp_established_options net/mptcp/options.c:874
>    tcp_established_options net/ipv4/tcp_output.c:1192
>    __tcp_transmit_skb net/ipv4/tcp_output.c:1575
>    __tcp_send_ack net/ipv4/tcp_output.c:4499
>    tcp_send_ack net/ipv4/tcp_output.c:4505
>    mptcp_subflow_shutdown net/mptcp/protocol.c:3137
>    mptcp_check_send_data_fin net/mptcp/protocol.c:3218
>    __mptcp_wr_shutdown net/mptcp/protocol.c:3234
>    __mptcp_close net/mptcp/protocol.c:3313
> 
>   Local variable opts created at:
>    __tcp_transmit_skb net/ipv4/tcp_output.c:1536
> 
> __tcp_transmit_skb() partially clears its on-stack tcp_out_options via
> "memset(&opts.cleared, ...)" which does not cover opts.mptcp, so MPTCP
> is responsible for initializing opts->ext_copy before use.
> 
> mptcp_established_options_dss() only writes opts->ext_copy when an
> MPTCP extension is attached to the skb ("opts->ext_copy = *mpext").
> When tcp_send_ack() builds a fresh pure-ACK skb during subflow
> shutdown, the skb has no extension and that assignment is skipped, but
> the function still calls mptcp_write_data_fin(&opts->ext_copy) because
> snd_data_fin_enable is set. mptcp_write_data_fin() then reads
> ext->use_map to choose between writing a synthetic DATA_FIN mapping or
> extending an existing one, and that read hits uninitialized stack
> memory.
> 
> Zero opts->ext_copy at the top of mptcp_established_options_dss() so
> mptcp_write_data_fin() takes its synthetic-mapping branch on clean
> memory.
Thank you for sharing this patch!

Kuniyuki already shared the same patch on the MPTCP ML a few days ago:

https://lore.kernel.org/mptcp/20260501061939.1808489-1-kuniyu@google.com/T/#u

A v2 has been shared a few hours ago:

https://lore.kernel.org/mptcp/20260504044051.3725846-1-kuniyu@google.com/T/#u

(and apparently, a v3 will be needed)

Cheers,
Matt
-- 
pw-bot: Rejected
Sponsored by the NGI0 Core fund.

Hi Deepanshu,

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/25295668463

Initiator: Patchew Applier
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/3318919851c0
Patchwork: https://patchwork.kernel.org/project/mptcp/list/?series=1089047


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)