Re: [PATCH 2/2] Use ktime_get() instead of sched_clock() for blkiocgroup stats.

From: Ingo Molnar
Date: Sat Jun 12 2010 - 03:19:23 EST



* Divyesh Shah <dpshah@xxxxxxxxxx> wrote:

> This will take care of the pre-emptive kernel issue and the unbounded
> TSC drift problem. We will lose resolution though in some cases.

> - blkg->stats.start_group_wait_time = sched_clock();
> + blkg->stats.start_group_wait_time = ktime_to_ns(ktime_get());

Ugh!

ktime_get() can have insanely high overhead. Peter has added local_clock(), if
then you should use that and apply checks to make sure the result isnt
negative.

Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/