Re: [RFC PATCH] hrtimer: removing all ur callback modes

From: Oliver Hartkopp
Date: Tue Dec 09 2008 - 03:07:52 EST


Peter Zijlstra wrote:
On Sun, 2008-12-07 at 13:59 +0100, Oliver Hartkopp wrote:
Oliver Hartkopp wrote:
Peter Zijlstra wrote:
On Tue, 2008-11-25 at 12:43 +0100, Peter Zijlstra wrote:
Hi,

This is an attempt at removing some of the hrtimer complexity by
reducing the number of callback modes to 1.

This means that all hrtimer callback functions will be ran from HARD-irq
context.

Hi Peter,

i ran a heavy load test, which get's (reproducible) the attached outputs ...

Maybe it's not that good to define the hrtimer context to be always hard-irq.

Thing is, this 'cleanup' removes quite a bit of complexity from the core
hrtimer code, and afaict your bit is the only thing that cannot seem to
cope. So I'd rather look at fixing your site than re-introduce softirqs
to hrtimers.


I wouldn't be that sure that all the other sites can cope with it as i only detected the problems with my code under heavy load ...

Any idea?

What are the timing constraints of your problem? - I assume they are not
too aggressive, otherwise you'd not be able to run from softirq, could
you run from keventd?

I browsed some code that's using hrtimers and found some hopefully good example in drivers/s390/crypto/ap_bus.c : They are using the hrtimer to schedule a tasklet which is running in soft-irq context:

http://lxr.linux.no/linux+v2.6.27.8/drivers/s390/crypto/ap_bus.c#L1150

This could also be the correct approach for my sock_queue_rcv_skb() problem, right?

Regards,
Oliver

ps. What is the intended release for this hrtimer cleanup? 2.6.29?


--
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/