Re: [PATCH 2/2] cputime: introduce thread_group_times()

[Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>]

Stanislaw Gruszka wrote:
> Idea is very good IMHO.

Thank you very much!

>>  #ifndef CONFIG_VIRT_CPU_ACCOUNTING
>> -	cputime_t prev_utime, prev_stime;
>> +	cputime_t prev_utime, prev_stime, prev_tgutime, prev_tgstime;
>>  #endif
> 
> I think the new values should be part of struct_signal (see below)

Good point.  I'll update patch to do so.

>>  /*
>> + * Must be called with siglock held.
>> + */
>> +void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
>> +{
>> +	struct task_cputime cputime;
>> +
>> +	__thread_group_times(p, &cputime);
>> +
>> +	if (ut)
>> +		*ut = cputime.utime;
>> +	if (st)
>> +		*st = cputime.stime;
> 
> No thread_group_times() nor task_times() is called with NULL arguments, we
> can get rid of "if ({u,s}t)" checks. Perhaps thread_group_times() should
> have "struct task_cputime" argument as it is wrapper for
> thread_group_cputime();

Removing "if ({u,s}t)" is OK with me.
I expect all thread_group_times() user should have no interest in members
of struct task_cputime other than {u,s}time, so I'd like to keep the
argument as is.


Thanks,
H.Seto