Re: [PATCH 1/3] signal: sys_rt_sigtimedwait: simplify the timeout logic

[Linus Torvalds]

On Mon, Apr 25, 2011 at 10:26 AM, Oleg Nesterov <oleg@xxxxxxxxxx> wrote:
>
> If only I knew what this comment could say except
>
>        /* Why do we add (tv_sec || tv_nsec) ?  */
>
> I'd better send 4/3 which simply removes this (I hope) unneeded code.

It's to guarantee that timeout is at least one tick more than asked
for, because the rule is that you really have to wait for AT LEAST the
time asked for. With the "zero timeout" being special, since that is
"immediate".

Imagine that you ask for one timer tick - but that you're in the
_middle_ of the current one. Waiting for the next timer is going to be
too short - you'll only get half a timer tick. So we need to ask for
"ceiling(nanoseconds / nanosecondspertick) + 1" to make sure that we
really wait _longer_ than asked for.

So "+ (tv_sec || tv_nsec)" is just the "+1" for the "not zero timeout" case.

                     Linus
--
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/