Re: [PATCH mptcp-next v3 00/29] userspace pm enhancements

[Matthieu Baerts <matttbe@kernel.org>]

Hi Geliang,

On 28/09/2023 23:37, Matthieu Baerts wrote:
> Hi Geliang,
> 
> On 28/09/2023 22:54, Matthieu Baerts wrote:
>> Hi Geliang,
>>
>> On 25/09/2023 10:41, Geliang Tang wrote:
>>> v3:
>>>  - drop patch 4 in v2
>>>  - invert patch 10 and 11 in v2
>>>  - rebased
>>
>> Thank you for the v3! And again, sorry for the delays!
>>
>> I think patches 1-11 + 13-14 are OK to be applied (with two small
>> modifications for patch 7 and 11 that I can do when applying these patches):
>>
>> Reviewed-by: Matthieu Baerts <matthieu.baerts@tessares.net>
>>
>> Please note that patch 13 will go to "Fixes for -net". Patch 14 can stay
>> in "Features for net-next" due to other dependences.
> 
> I just did that:
> 
> New patches for t/upstream-net and t/upstream:
> - 505a33d76a32: mptcp: userspace pm allow creating id 0 subflow
> - Results: 05758f924fe3..999b9f7e47a5 (export-net)
> 
> New patches for t/upstream:
> - 1cb016e94837: mptcp: drop useless ssk in pm_subflow_check_next
> - 450deaf6dd71: mptcp: use mptcp_check_fallback helper
> - 6df74a0ced73: mptcp: use mptcp_get_ext helper
> - 03649553fd68: mptcp: move sk assignment statement ahead
> - 6bc1e5144525: mptcp: define more local variables sk
> - b7e3cc69d9fe: selftests: mptcp: sockopt: drop mptcp_connect var
> - 14c099dbd7a7: selftests: mptcp: display simult in extra_msg
> - 7d9e508e1117: mptcp: add mptcpi_subflows_total counter
> - af6ed0819599: selftests: mptcp: add evts_get_info helper
> - 9177dc0b474c: selftests: mptcp: add chk_subflows_total helper
> - bc1edd827cfa: selftests: mptcp: update userspace pm test helpers
> - 0dd61d4e14e8: selftests: mptcp: userspace pm create id 0 subflow
> - Results: 3be1c8ce544a..516baae7a220 (export)

I think these patches are introducing regressions because these tests
are now regularly failing on the public CI:

  - 112 - mptcp_join: userspace pm add & remove address
  - 113 - mptcp_join: userspace pm create destroy subflow

I didn't check in more details but I see they are failing since
export/20230928T213439 and only when running with a debug kconfig (slow
environment), e.g.:

  - https://cirrus-ci.com/task/5444189572300800
  -
https://api.cirrus-ci.com/v1/artifact/task/5444189572300800/summary/summary.txt

Do you mind having a look?

Maybe the connection has been closed when looking at the number of subflows?
Also, it looks like the helper looking at 'subflows_total' with older
versions ss is producing a lot of debug output that can be muted.

Cheers,
Matt