netdev - Re: [RFC PATCH] net: enable SO_REUSEPORT for AF

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <4ed5bfc9-dcdb-4f43-8be0-9d7315a88583@suse.com>
Date: Wed, 12 Mar 2025 20:45:17 +0100
From: Nicolas Morey <nicolas.morey@...e.com>
To: Kuniyuki Iwashima <kuniyu@...zon.com>
Cc: edumazet@...gle.com, netdev@...r.kernel.org
Subject: Re: [RFC PATCH] net: enable SO_REUSEPORT for AF_TIPC sockets

On 2025-03-12 19:20, Kuniyuki Iwashima wrote:
> From: Nicolas Morey <nicolas.morey@...e.com>
> Date: Wed, 12 Mar 2025 18:44:10 +0100
>> On 2025-03-12 17:35, Kuniyuki Iwashima wrote:
>>> From: Nicolas Morey <nicolas.morey@...e.com>
>>> Date: Wed, 12 Mar 2025 14:48:01 +0100
>>>> Commit 5b0af621c3f6 ("net: restrict SO_REUSEPORT to inet sockets") disabled
>>>> SO_REUSEPORT for all non inet sockets, including AF_TIPC sockets which broke
>>>> one of our customer applications.
>>>> Re-enable SO_REUSEPORT for AF_TIPC to restore the original behaviour.
>>>
>>> AFAIU, AF_TIPC does not actually implement SO_REUSEPORT logic, no ?
>>> If so, please tell your customer not to set it on AF_TIPC sockets.
>>>
>>> There were similar reports about AF_VSOCK and AF_UNIX, and we told
>>> that the userspace should not set SO_REUSEPORT for such sockets
>>> that do not support the option.
>>>
>>> https://lore.kernel.org/stable/CAGxU2F57EgVGbPifRuCvrUVjx06mrOXNdLcPdqhV9bdM0VqGvg@mail.gmail.com/
>>> https://github.com/amazonlinux/amazon-linux-2023/issues/901
>>>
>>>
>> Isn't the sk_reuseport inherited/used by the underlying UDP socket ?
> 
> tipc_udp_enable() calls udp_sock_create() and udp_sock_create[46]()
> creates a new UDP socket and bind()s without setting SO_REUSEPORT.

Thanks for the feedback. We'll advise to fix the userland code then.

Nicolas