Re: [PATCH mptcp-next] selftests: mptcp: join: no SKIP mark for group checks

[Matthieu Baerts <matttbe@kernel.org>]

Hi Geliang,

Thank you for the review! (and sorry for the late reply)

On 28/12/2025 04:06, Geliang Tang wrote:
> Hi Matt,
> 
> Thanks for this patch.
> 
> On Thu, 2025-12-11 at 19:09 +0100, Matthieu Baerts (NGI0) wrote:
>> When executing the last MPTCP selftests on older kernels, this output
>> is
>> printed:
>>
>>   # 001 no JOIN
>>   #       join Rx                             [SKIP]
>>   #       join Tx                             [SKIP]
>>   #       fallback                            [SKIP]
>>
>> In fact, behind each line, a few counters are checked, and likely not
>> all of them have been skipped because the they are not available on
>> these kernels. Instead, "new" and unsupported counters for these
>> groups
>> are now ignored, and [ OK ] will be printed instead of [SKIP].
>>
>> Note that on the MPTCP CI, when validating the dev versions, any
>> unsupported counter will cause the tests to fail. So this is safe not
>> to
>> print 'SKIP' for these group checks.
>>
>> Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
>> ---
>>  tools/testing/selftests/net/mptcp/mptcp_join.sh | 30 ++++++++++++---
>> ----------
>>  1 file changed, 15 insertions(+), 15 deletions(-)
>>
>> diff --git a/tools/testing/selftests/net/mptcp/mptcp_join.sh
>> b/tools/test[ -z "$count" ]ing/selftests/net/mptcp/mptcp_join.sh
>> index b2e6e548f796..77370b10651a 100755
>> --- a/tools/testing/selftests/net/mptcp/mptcp_join.sh
>> +++ b/tools/testing/selftests/net/mptcp/mptcp_join.sh
>> @@ -1407,7 +1407,7 @@ chk_join_tx_nr()
>>  
>>  	count=$(mptcp_lib_get_counter ${ns2}
>> "MPTcpExtMPJoinSynTxCreatSkErr")
>>  	if [ -z "$count" ]; then
>> -		rc=${KSFT_SKIP}
>> +		: # ignore skip
>>  	elif [ "$count" != "$create" ]; then
> 
> In that case, wouldn't it be better to remove 'if [ -z "$count" ];
> then' and directly check with 'if [ "$count" != "$create" ]; then'?

No, we cannot: if $count is empty, then "$count" != "$create", and the
test will be marked as failed instead of being ignored.

We could have:

  if [ -n "$count" ] && [ "$count" != "$create" ]

But I think it is clearer to have an explicit comment for this special
case than multiple conditions. WDYT?

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