Netdev List
 help / color / mirror / Atom feed
* [PATCH net] tcp: cancel cloned request timers without waiting
@ 2026-11-19  2:56 Wentao Guan
  2026-11-19  3:03 ` netdev-bot+sinfo
  2026-11-19  6:22 ` Kuniyuki Iwashima
  0 siblings, 2 replies; 3+ messages in thread
From: Wentao Guan @ 2026-11-19  2:56 UTC (permalink / raw)
  To: edumazet, ncardwell, kuba, kuniyu; +Cc: netdev, melotti, Wentao Guan, stable

When a cloned request is migrated again, its timer has the same lockdep
class as the running request callback. Using timer_delete_sync() through
__inet_csk_reqsk_queue_drop(..., false) can therefore trigger a
recursive-locking splat and disable lockdep.

The clone is unhashed when inet_ehash_insert() fails, so
reqsk_queue_unlink() is a no-op. Its timer is pinned to the current CPU,
allowing it to be cancelled without waiting for its callback. Use
timer_delete() and drop the timer reference directly instead of
synchronously deleting it through the queue helper.

Fixes: 97c8a3c1f73d ("tcp: Fix potential UAF in reqsk_timer_handler().")
Cc: stable@vger.kernel.org
Signed-off-by: Wentao Guan <guanwentao@uniontech.com>
---
 net/ipv4/inet_connection_sock.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/net/ipv4/inet_connection_sock.c b/net/ipv4/inet_connection_sock.c
index 6c079929e06f6..346d0015a140d 100644
--- a/net/ipv4/inet_connection_sock.c
+++ b/net/ipv4/inet_connection_sock.c
@@ -1121,8 +1121,8 @@ static void reqsk_timer_handler(struct timer_list *t)
 			return;
 
 		if (!inet_ehash_insert(req_to_sk(nreq), req_to_sk(oreq), NULL)) {
-			/* delete timer */
-			__inet_csk_reqsk_queue_drop(sk_listener, nreq, false);
+			if (timer_delete(&nreq->rsk_timer))
+				reqsk_put(nreq);
 			goto no_ownership;
 		}
 
-- 
2.47.3


^ permalink raw reply related	[flat|nested] 3+ messages in thread

* Re: [PATCH net] tcp: cancel cloned request timers without waiting
  2026-11-19  2:56 [PATCH net] tcp: cancel cloned request timers without waiting Wentao Guan
@ 2026-11-19  3:03 ` netdev-bot+sinfo
  2026-11-19  6:22 ` Kuniyuki Iwashima
  1 sibling, 0 replies; 3+ messages in thread
From: netdev-bot+sinfo @ 2026-11-19  3:03 UTC (permalink / raw)
  To: Wentao Guan; +Cc: edumazet, ncardwell, kuba, kuniyu, netdev, melotti, stable

Hi!

This is an automated message. This series looks like a fix, but its
commit messages seem to be missing some information:

 - How the issue was discovered, e.g. hit in production, hit during
   development, syzbot report, manual code inspection, LLM or static
   analysis tool scan.

 - Whether the issue was actually triggered, or is only theoretical
   (e.g. found by LLM and not proven at runtime). If it was triggered
   please include the symptoms, like the stack trace or kernel messages.

Please do not repost the series just to address the above. Instead,
reply to this email with the missing information, so that reviewers
can take it into account.

*IMPORTANT*: If you do repost (for other reasons) - please include
the relevant info in the cover letter or commit messages.
-- 
The evaluation is done by an LLM so it may be wrong, if you think
that is the case please reply and explain.

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH net] tcp: cancel cloned request timers without waiting
  2026-11-19  2:56 [PATCH net] tcp: cancel cloned request timers without waiting Wentao Guan
  2026-11-19  3:03 ` netdev-bot+sinfo
@ 2026-11-19  6:22 ` Kuniyuki Iwashima
  1 sibling, 0 replies; 3+ messages in thread
From: Kuniyuki Iwashima @ 2026-11-19  6:22 UTC (permalink / raw)
  To: Wentao Guan; +Cc: edumazet, ncardwell, kuba, netdev, melotti, stable

On Thu, 19 Nov 2026 10:56:20 +0800, Wentao Guan wrote:
> When a cloned request is migrated again, its timer has the same lockdep
> class as the running request callback. Using timer_delete_sync() through
> __inet_csk_reqsk_queue_drop(..., false) can therefore trigger a
> recursive-locking splat and disable lockdep.

This is net-next material.


> 
> The clone is unhashed when inet_ehash_insert() fails, so
> reqsk_queue_unlink() is a no-op. Its timer is pinned to the current CPU,
> allowing it to be cancelled without waiting for its callback. Use
> timer_delete() and drop the timer reference directly instead of
> synchronously deleting it through the queue helper.

This change depnds on 4e2b3ef19e0e, without that, __reqsk_free()
frees reqsk immediately and timer triggers UAF.

So please mention the commit.


> 
> Fixes: 97c8a3c1f73d ("tcp: Fix potential UAF in reqsk_timer_handler().")
> Cc: stable@vger.kernel.org

Please drop two tags above.

---
pw-bot: cr

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2026-11-19  6:23 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-19  2:56 [PATCH net] tcp: cancel cloned request timers without waiting Wentao Guan
2026-11-19  3:03 ` netdev-bot+sinfo
2026-11-19  6:22 ` Kuniyuki Iwashima

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox