Re: [Weekly meetings] MoM - 2nd of December 2021

[Mat Martineau <mathew.j.martineau@linux.intel.com>]

On Thu, 2 Dec 2021, Matthieu Baerts wrote:

>
>    - news about Intel's kbuild? (Mat):
>        - Saw one mptcp_join.sh failure but probably already seen by the
> public CI (unstable test)
>

This failure was a missing ADD_ADDR echo:

user  :notice: [  501.625742] # 41 add multiple addresses IPv6          syn[ ok ] - synack[ ok ] - ack[ ok ]
user  :notice: [  501.676759] #                                         add[ ok ] - echo  [fail] got 1 ADD_ADDR echo[s] expected 2
user  :notice: [  501.690104] #
user  :notice: [  501.695183] # Server ns stats
user  :notice: [  501.705037] # MPTcpExtMPCapableSYNRX          1                  0.0
user  :notice: [  501.715549] # MPTcpExtMPCapableACKRX          1                  0.0
user  :notice: [  501.726495] # MPTcpExtMPJoinSynRx             2                  0.0
user  :notice: [  501.737151] # MPTcpExtMPJoinAckRx             2                  0.0
user  :notice: [  501.747762] # MPTcpExtEchoAdd                 1                  0.0
user  :notice: [  501.757498] # Client ns stats
user  :notice: [  501.764657] # MPTcpExtMPCapableSYNTX          1                  0.0
user  :notice: [  501.776142] # MPTcpExtMPCapableSYNACKRX       1                  0.0
user  :notice: [  501.786696] # MPTcpExtMPJoinSynAckRx          2                  0.0
user  :notice: [  501.797269] # MPTcpExtAddAddr                 2                  0.0


Given that the client received two ADD_ADDRs, it could have been a few 
different things:

* Echo packet was sent but dropped, and the test finished before the
   retransmit (the default add_addr timeout is quite long) (not a bug)

* Echo packet was never sent (maybe a bug, maybe not)


There isn't enough logged to narrow it down more, so I will watch for 
similar failures. Looks like most of the recent Cirrus CI mptcp_join.sh 
failures are related to missing RM_ADDRs.


--
Mat Martineau
Intel