Re: [PATCH net] mptcp: protect listener backlog reads in diagnostics

[Matthieu Baerts <matttbe@kernel.org>]

Hi Gang, Cen,

On 27/10/2026 11:41, gang.yan@linux.dev wrote:
> October 26, 2026 at 10:44 PM, "Cen Zhang" <zzzccc427@gmail.com mailto:zzzccc427@gmail.com?to=%22Cen%20Zhang%22%20%3Czzzccc427%40gmail.com%3E > wrote:
> 
> 
>>
>> Hi Eric,
>>
>> Eric Dumazet <edumazet@kernel.org> 于2026年10月26日周一 20:38写道：
>>
>>>
>>> Adding these stack traces without symbols is not really nice.
>>>
>>>  Time to use scripts/decode_stacktrace.sh
>>>
>> Sorry about that, v2 carries the report decoded with
>> scripts/decode_stacktrace.sh and trimmed to the relevant frames.
>>
>>>
>>> msk->first is a TCP socket: TCP listeners have SOCK_RCU_FREE, and
>>>  tcp_prot / tcpv6_prot use SLAB_TYPESAFE_BY_RCU. Reading two integers
>>>  under rcu_read_lock() is safe, and stale values are fine for diag.
>>>
>> Thanks for the suggestions.
> 
> Hi Cen,
> 
> Basically this looks fine to me, I don't see any issue with the overall
> approach.

@Gang: Thank you for your review! (Don't hesitate to send a
Reviewed-by/Acked-by on the v2 then ;) )

> Also, could you send v3 separately to the MPTCP mailing list? It seems the
> CI checks were not triggered this time. I guess it might be because the
> patch was not based on the export branch of the mptcp repo on GitHub, so
> you can generate the patch based on that branch
> (multipath-tcp/mptcp_net-next, branch export).

It's not clear to me why the CI got an issue to apply the v2:

  Applying: mptcp: protect listener backlog reads in diagnostics
  error: corrupt patch at line 38
  error: could not build fake ancestor

(Not sure what this "line 38" corresponds to)

https://patchew.org/MPTCP/pm-mptcp-objects-candidate-0066-v5-4d1b83d62fcde2d57e8d@gmail.com/logs/git/

But I manually applied the patch and asked the CI to validate it.

@Cen: just to be clear: don't send a v3 just to fix the CI issue, except
if a reviewer asks you too.

> Two notes for next time (Important):
> 1. For MPTCP-only patches, it is enough to send them to the MPTCP list
>    only, no need to CC netdev.

Small note: for "urgent" patches, it is fine to send them to Netdev as
well, following what's in the MAINTAINERS file. For new features and
non-urgent fixes, it is better to send them to the MPTCP list only to
reduce the traffic on the Netdev list.

> 2. If a new version is needed, please wait 24 hours before sending it.

(or more: wait for the Netdev Sashiko review)

Cheers,
Matt
-- 
Sponsored by the NGI0 Core fund.