Re: [RFC] schedule_timeout_range()

[David Woodhouse <dwmw2@infradead.org>]

On Tue, 2008-07-22 at 14:33 +1000, Nick Piggin wrote:
> The only thing I dislike about explicit times is that when a driver or
> someone doesn't _really_ know how much to specify. Do you say 10s, 100s?

This is true, but they certainly have a _better_ idea than we do. If the
individual callers can't even come up with an answer, how are we ever
going to come up with a generic policy that does the right thing? 

I really don't think that applying this kind of policy in generic code
is useful -- I'd like the callers to provide numbers even if they _do_
pull it out of their wossname.

The number they provide is the _maximum_ amount of time they should be
prepared to wait (let's assume for a moment that they stayed sober and
remembered Linux isn't a real-time kernel, so all guarantees are taken
with a pinch of salt. Let's not get bogged down in nomenclature).

In practice, they'll almost always get called before that maximum time
expires -- that's the whole _point_, of course. But we can't _invent_
that maximum in generic code; that's really up to the caller.

> Some upper bound would be nice, which basically would not have to ever
> fire by itself unless there is some CPU activity (so you don't have to
> set two timers as a bonus). After that, I wonder, perhaps some "maximum
> power savings value but not completely deferred"? Say give it a max of
> 30s? Or perhaps even that is not future-proof enough if we one day want
> to suspend most of the system between external IOs?

I _really_ don't think we want to go there. Let the caller set a maximum
amount of time they're prepared to wait, and that's it.

-- 
dwmw2