2007-08-09 05:10:48

by Mitchell Erblich

Subject: Question : sched_rt.c : Loss of stats?? requeue_task_rt() does not call update_curr_rt() which updates stats

sched_rt.c : requeue_task_rt()

The comment states the problem requeue no dequeue.
Put task to the end of the run list without the overhead of dequeue
followed by enqueue.

dequeue_task_rt() updates stats. Where without calling
it will skip the stat update.

Thus, shouldn't requeue_task_rt() call
update_curr_rt(rq); ???

Mitchell Erblich.

2007-08-09 08:24:40

by Dmitry Adamushko

Subject: Re: Question : sched_rt.c : Loss of stats?? requeue_task_rt() does not call update_curr_rt() which updates stats

On 09/08/07, Mitchell Erblich <[email protected]> wrote:
> sched_rt.c : requeue_task_rt()
>
> The comment states the problem requeue no dequeue.
> Put task to the end of the run list without the overhead of dequeue
> followed by enqueue.
>
> dequeue_task_rt() updates stats. Where without calling
> it will skip the stat update.
>
> Thus, shouldn't requeue_task_rt() call
> update_curr_rt(rq); ???

No. As update_curr_rt() suggests it's only about collecting runtime
statistics, namely
(a) the longest interval of time a given task was consuming a CPU
(i.e. == 'current') and
(b) the total amount of time it was running.

To this goal, it would be enough to call it only in
put_prev_task_rt().. I recall, we do it in dequeue_task_rt() to catch
a case of SCHED_RR/FIFO --> SCHED_NORMAL transition (e.g. in
sched_setscheduler()).. so that may make accounting a bit less precise
for (a).


--
Best regards,
Dmitry Adamushko