Re: RT task scheduling

[Ingo Molnar]

* Bill Huey <billh@xxxxxxxxxxxxxxxxx> wrote:

> The last time I looked at it I thought it did something pretty 
> simplistic in that it just dumped any RT thread to another CPU but 
> didn't do it in a strict manner with regard to priority. Maybe that's 
> changed or else I didn't pay attention to it that as carefully as I 
> thought.

well as Darren's testcase shows, it might still have some bug - but the 
mechanism is intended to be strict. (the implementation had a couple of 
strictness bugs (they show up as long latencies on SMP) but those were 
ironed out months ago.)

> As far as CPU binding goes, I'm wanting a method of getting around the 
> latency of the rt overload logic in certain cases at the expense of 
> rebalancing. That's what I ment by it.

yeah, that certainly makes sense, and it's one reason why i'm thinking 
about the separate SCHED_FIFO_GLOBAL policy for 'globally scheduled' RT 
tasks, while still keeping the current lightweight non-global RT 
scheduling. Global scheduling either means a global lock, or as in the 
-rt implementation means a "global IPI", but there's always a nontrivial 
"global" cost involved.

	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/