[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <3a8ba87f-a6ab-4523-b0ce-8e9dbd5a923b@redhat.com>
Date: Wed, 1 Oct 2025 18:46:16 +0200
From: Paolo Abeni <pabeni@...hat.com>
To: Jakub Kicinski <kuba@...nel.org>, Eric Dumazet <edumazet@...gle.com>,
Florian Westphal <fw@...len.de>, Kuniyuki Iwashima <kuniyu@...gle.com>,
Willem de Bruijn <willemb@...gle.com>
Cc: netdev@...r.kernel.org
Subject: Re: deadlocks on pernet_ops_rwsem
On 10/1/25 5:20 PM, Jakub Kicinski wrote:
> We started hitting a deadlock on pernet_ops_rwsem, not very often,
> these are the CI branches were we saw it:
>
> 2025-09-18--03-00
> 2025-09-18--15-00
> 2025-09-18--18-00
> 2025-09-19--21-00
> 2025-09-21--15-00
> 2025-09-21--18-00
> 2025-09-22--00-00
> 2025-09-22--09-00
> 2025-09-23--06-00
> 2025-09-24--21-00
> 2025-09-27--21-00
> 2025-09-28--09-00
> 2025-10-01--00-00
> 2025-10-01--06-00
>
> First hit seems to be:
> https://netdev-3.bots.linux.dev/vmksft-net-dbg/results/302741/65-l2tp-sh/stderr
> Two most recent:
> https://netdev-3.bots.linux.dev/vmksft-net-dbg/results/321640/110-ip-local-port-range-sh/stderr
> https://netdev-3.bots.linux.dev/vmksft-net-dbg/results/321281/63-fdb-notify-sh/stderr
>
> The stack traces aren't very helpful, they mostly show the victim not
> who caused the deadlock :(
>
> Any ideas?
Not many here. The above are debug builds, so we should get a lockdep
splat on deadlock, the logs lack it. I guess the request_module() breaks
the lockdep checks?
IIRC syzkaller uses config with most/all builtins, so I guess it should
not be able to replicate this scenario (to possibly help creating a
repro), unless there is an allmodconfig instance?
Adding Willem, just in case he knows more about possible syzkaller usage
here.
/P
Powered by blists - more mailing lists