Re: [PATCH v4 mptcp-net 0/5] mptcp: fix fallback-related races

[Matthieu Baerts <matttbe@kernel.org>]

Hi Paolo,

On 11/07/2025 15:51, Paolo Abeni wrote:
> On 7/11/25 11:36 AM, Matthieu Baerts wrote:
>> On 11/07/2025 11:29, Paolo Abeni wrote:
>>> This series contains 3 fixes somewhat related to various races we have
>>> while handling fallback and 2 small follow-up likely more suited for
>>> net-next.
>>>
>>> The root cause of the issues addressed here is that the check for
>>> "we can fallback to tcp now" and the related action are not atomic. That
>>> also applies to fallback due to MP_FAIL - where the window race is even
>>> wider.
>>>
>>> Address the issue introducing an additional spinlock to bundle together
>>> all the relevant events, as per patch 1 and 2.
>>>
>>> Note that mptcp_disconnect() unconditionally
>>> clears the fallback status (zeroing msk->flags) but don't tuch the
>>> `allows_infinite_fallback` flag. Such issue is addressed in patch 3.
>>>
>>> Patch 4 cleans up a bit the fallback code, introducing specific MIB for
>>> each FB reason, and patch 5 drops the, hopefully now redundant
>>> pr_fallback().
>>> ---
>>> v3 -> v4:
>>>  - fic MIB exposed names
>>
>> Thank you for the new version, it looks good to me!
>>
>> Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
> 
> Thank you!
> 
>> I can apply this series, restart syzkaller on top of it, and send
>> patches 1-3 on Monday if everything is OK. WDYT?
> 
> I would wait a little more for syzkaller. The underlying issue is not
> very recent and the fix is not that invasive, I guess it's not a tragedy
> if it reaches Linus' tree the following week.
> 
> But no strong opinion either ways.

Sure, I can wait, but it was just to be able to flush the new MIB
counters in next before the new release :)

Note that syzkaller was validating your previous versions since Monday,
not only from today then :)

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