Re: [PATCH] hrtimers: Special-case zero length sleeps

[Thomas Gleixner <tglx@linutronix.de>]

On Thu, 29 Sep 2011, Matthew Garrett wrote:

> sleep(0) is a common construct used by applications that want to trigger
> the scheduler. sched_yield() might make more sense, but only appeared in
> POSIX.1-2001 and so plenty of example code still uses the sleep(0) form.
> This wouldn't normally be a problem, but it means that event-driven
> applications that are merely trying to avoid starving other processes may
> actually end up sleeping due to having large timer_slack values. Special-
> casing this seems reasonable.
> 
> Signed-off-by: Matthew Garrett <mjg@redhat.com>
> ---
>  kernel/hrtimer.c |    8 ++++++++
>  1 files changed, 8 insertions(+), 0 deletions(-)
> 
> diff --git a/kernel/hrtimer.c b/kernel/hrtimer.c
> index a9205e3..0bb70a7 100644
> --- a/kernel/hrtimer.c
> +++ b/kernel/hrtimer.c
> @@ -1566,6 +1566,14 @@ long hrtimer_nanosleep(struct timespec *rqtp, struct timespec __user *rmtp,
>  	if (rt_task(current))
>  		slack = 0;
>  
> +	/*
> +	 * Applications will often sleep(0) to indicate that they wish to
> +	 * be scheduled. Special case that to avoid actually putting them
> +	 * to sleep for the duration of the slack.
> +	 */
> +	if (rqtp->tv_sec == 0 && rqtp->tv_nsec == 0)
> +		slack = 0;

That's pretty pointless. You can simply return 0 here as
do_nanosleep() will not call the scheduler on an already expired
timer, which is always true for a relative timer with delta 0.

Thanks,

	tglx