Re: [PATCH] selftests: mm: restore settings from only parent process

[Muhammad Usama Anjum]

On 4/18/24 5:52 PM, Joey Gouly wrote:> Hi again,
> 
> On Thu, Mar 14, 2024 at 02:40:45PM +0500, Muhammad Usama Anjum wrote:
>> The atexit() is called from parent process as well as forked processes.
>> Hence the child restores the settings at exit while the parent is still
>> executing. Fix this by checking pid of atexit() calling process and only
>> restore THP number from parent process.
>>
>> Fixes: c23ea61726d5 ("selftests/mm: protection_keys: save/restore nr_hugepages settings")
>> Tested-by: Joey Gouly <joey.gouly@xxxxxxx>
>> Signed-off-by: Muhammad Usama Anjum <usama.anjum@xxxxxxxxxxxxx>
>> ---
>>  tools/testing/selftests/mm/protection_keys.c | 6 +++++-
>>  1 file changed, 5 insertions(+), 1 deletion(-)
>>
>> diff --git a/tools/testing/selftests/mm/protection_keys.c b/tools/testing/selftests/mm/protection_keys.c
>> index f822ae31af22e..374a308174d2b 100644
>> --- a/tools/testing/selftests/mm/protection_keys.c
>> +++ b/tools/testing/selftests/mm/protection_keys.c
>> @@ -1745,9 +1745,12 @@ void pkey_setup_shadow(void)
>>  	shadow_pkey_reg = __read_pkey_reg();
>>  }
>>  
>> +pid_t parent_pid;
>> +
>>  void restore_settings_atexit(void)
>>  {
>> -	cat_into_file(buf, "/proc/sys/vm/nr_hugepages");
>> +	if (parent_pid == getpid())
>> +		cat_into_file(buf, "/proc/sys/vm/nr_hugepages");
>>  }
>>  
>>  void save_settings(void)
>> @@ -1773,6 +1776,7 @@ void save_settings(void)
>>  		exit(__LINE__);
>>  	}
>>  
>> +	parent_pid = getpid();
>>  	atexit(restore_settings_atexit);
>>  	close(fd);
>>  }
> 
> After more testing, this is not actually enough. It passes sometimes, which is
> why I gave my Tested-by, but it can still fail the same way as I originally
> said.
> 
> assert() at protection_keys.c::812 test_nr: 19 iteration: 1
> running abort_hooks()...
> errno at assert: 12
> 
> I think the parent process needs to wait for all it's child processes (recursively).
> This is due to the test_pkey_alloc_exhaust() calling become_child(), where it
> exits, and that exit may be from the original PID.
Thanks for reporting. I'll have a look at it.

> 
> Thanks,
> Joey

-- 
BR,
Muhammad Usama Anjum