[LTP] [PATCH v5 3/6] Refactor mqns_03 using new LTP API

[Andrea Cervesato]

Hi Petr,

On 3/3/23 10:25, Petr Vorel wrote:
> Hi Andrea,
>
> ...
>> +++ b/runtest/containers
>> @@ -22,7 +22,8 @@ mqns_02 mqns_02
>>   mqns_02_clone mqns_02 -m clone
>>   mqns_02_unshare mqns_02 -m unshare
>>   mqns_03 mqns_03
>> -mqns_03_clone mqns_03 -clone
>> +mqns_03_clone mqns_03 -m clone
>> +mqns_03_unshare mqns_03 -m unshare
> Interesting, it was not run with unshare. +1 for adding it.
>
>>   mqns_04 mqns_04
>>   mqns_04_clone mqns_04 -clone
>> diff --git a/testcases/kernel/containers/mqns/mqns_03.c b/testcases/kernel/containers/mqns/mqns_03.c
> ...
>> +/*\
>> + * [Description]
>> + *
>> + * Test mqueuefs from an isolated/forked process namespace.
> Again, I'd mention supported modes here.
>> + *
>> + * [Algorithm]
>> + *
>> + * - create /MQ1
>> + * - mount mqueue inside the temporary folder
>> + * - check for /MQ1 existance
>> + * - create /MQ2 inside the temporary folder
>> + * - umount
>> + * - mount mqueue inside the temporary folder
>> + * - check /MQ1 existance
>> + * - check /MQ2 existance
>> + * - umount
>> + */
>>   #define _GNU_SOURCE
> I wonder why _GNU_SOURCE was used here and in mqns_04.c. I don't see anything
> in man mq_open(3), both tests are working without it.
Because of MAP_ANONYMOUS for mmap
> ...
>> +static void cleanup(void)
>> +{
> mqns_04.c checks here:
> 	if (!devdir)
> 		return;
>
> Although it's unlikely that SAFE_ASPRINTF(&devdir, "%s/mqueue", tmpdir); in
> setup() would fail, I'd add it also here.
>
> Kind regards,
> Petr

Andrea