Re: [PATCH mptcp-next 4/4] selftests: mptcp: get stats just before timing out

[Matthieu Baerts <matttbe@kernel.org>]

Hi Paolo,

On 13/11/2025 08:25, Paolo Abeni wrote:
> On 11/8/25 3:20 PM, Matthieu Baerts (NGI0) wrote:
>> Recently, some debugging happened around a test that was timing out. The
>> stats were showing connections being closed which was confusing because
>> the closing state was caused by the timeout stopping the transfer.
>>
>> To avoid such confusion, the timeout is no longer done per mptcp_connect
>> process, but separately. In case of timeout, the stats are now printed,
>> then the apps are killed.
>>
>> The stats will still be printed after the kill, but that's fine, and
>> this might even be useful, just in case. Timeout should be exceptional.
>>
>> Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
> 
> This is very useful, thanks!
> 
>> ---
>>  tools/testing/selftests/net/mptcp/mptcp_connect.sh | 21 +++++++++-------
>>  tools/testing/selftests/net/mptcp/mptcp_join.sh    | 29 +++++++++++-----------
>>  tools/testing/selftests/net/mptcp/mptcp_lib.sh     | 13 ++++++++++
>>  tools/testing/selftests/net/mptcp/mptcp_sockopt.sh | 21 +++++++++-------
>>  tools/testing/selftests/net/mptcp/simult_flows.sh  | 24 ++++++++++--------
>>  5 files changed, 66 insertions(+), 42 deletions(-)
>>
>> diff --git a/tools/testing/selftests/net/mptcp/mptcp_connect.sh b/tools/testing/selftests/net/mptcp/mptcp_connect.sh
>> index 1149bd150d6a..6bff77ed0a74 100755
>> --- a/tools/testing/selftests/net/mptcp/mptcp_connect.sh
>> +++ b/tools/testing/selftests/net/mptcp/mptcp_connect.sh
>> @@ -391,26 +391,29 @@ do_transfer()
>>  		mptcp_lib_nstat_init "${connector_ns}"
>>  	fi
>>  
>> -	timeout ${timeout_test} \
>> -		ip netns exec ${listener_ns} \
>> -			./mptcp_connect -t ${timeout_poll} -l -p $port -s ${srv_proto} \
>> -				$extra_args $local_addr < "$sin" > "$sout" &
>> +	ip netns exec ${listener_ns} \
>> +		./mptcp_connect -t ${timeout_poll} -l -p $port -s ${srv_proto} \
>> +			$extra_args $local_addr < "$sin" > "$sout" &
>>  	local spid=$!
>>  
>>  	mptcp_lib_wait_local_port_listen "${listener_ns}" "${port}"
>>  
>>  	local start
>>  	start=$(date +%s%3N)
>> -	timeout ${timeout_test} \
>> -		ip netns exec ${connector_ns} \
>> -			./mptcp_connect -t ${timeout_poll} -p $port -s ${cl_proto} \
>> -				$extra_args $connect_addr < "$cin" > "$cout" &
>> +	ip netns exec ${connector_ns} \
>> +		./mptcp_connect -t ${timeout_poll} -p $port -s ${cl_proto} \
>> +			$extra_args $connect_addr < "$cin" > "$cout" &
>>  	local cpid=$!
>>  
>> +	mptcp_lib_wait_timeout "${timeout_test}" "${listener_ns}" \
>> +		"${connector_ns}" "${port}" "${cpid}" "${spid}" &
> 
> I'm still debugging some timeouts caused by `${timeout_poll}`. In such
> cases the process still terminates and closes the relevant fd. I'm
> wondering additionally setting ${timeout_test} to a lower value of
> ${timeout_poll}?
${timeout_test} is supposed to be a safeguard, just in case
${timeout_poll} is not enough. I think it might be better to call 'ss'
from mptcp_connect.c in case of timeout: WDYT?

> BTW I'm testing with CONFIG_PREEMPT=y and possibly I'm seeing the
> timeout due to such setting triggering some race we don't observe with
> vng defaults: it's very easily reproducible on mptcp_connect.sh
> "disconnect" case.

Arf, maybe not a new issue... Thank you for looking at that!

The MPTCP CI and my syzkaller instances are all using
CONFIG_PREEMPT_VOLUNTARY=y. Should I change that? Or at least have a mix
in the Syzkaller instances?

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