Re: [PATCH 1/7] posix-cpu-timers: don't account cpu timer after stopped thread runtime accounting

[KOSAKI Motohiro <kosaki.motohiro@gmail.com>]

(5/7/13 11:24 AM), Frederic Weisbecker wrote:
> 2013/5/7 KOSAKI Motohiro <kosaki.motohiro@gmail.com>:
>>>> +     /*
>>>> +      * After turning over se.sum_exec_runtime to sig->sum_sched_runtime
>>>> +      * in __exit_signal(), we must not account exec_runtime for consistency.
>>>> +      */
>>>> +     if (unlikely(!tsk->sighand))
>>>> +             return;
>>>
>>> Ok, if we want the clock and timer to be consistent, do we also want the same check in
>>> account_group_user_time() and account_group_system_time()? The task can still account
>>> a tick after autoreaping itself between release_task() and the final schedule().
>>
>> You are right.
>>
>> That said, current the man pages don't describe this linux specific
>> extensions. So, nobody
>> (glibc, ltp, and me) tested them. Please give me a couple of days.
>> I'll test and fix this features
>> too.
>>
>> timer_create(2): http://man7.org/linux/man-pages/man2/timer_create.2.html
> 
> Ah, indeed timer_create() seem to only create CPUCLOCK_SCHED timers. So that
> issue with timer_gettime becoming asynchonous with clock_gettime can't happen
> with PROF and VIRT clocks
> 
> I see itimers can use those clocks. But there don't seem to be a
> similar issue with
> getitimer/setitimer as they don't have matching clock reads.

OK. I've found PROF and VIRT clock of posix timer have a bug and I could narrow down
and fixed it. Please see my next iteration.

Thanks.