Re: RCU latency regression in 2.6.16-rc1

[Ingo Molnar <mingo@elte.hu>]

* Lee Revell <rlrevell@joe-job.com> wrote:

> > Here is an updated version of that patch against 2.6.16-rc1. I have
> > sanity-tested it on ppc64 and x86_64 using dbench and kernbench.
> > I have also tested this for OOM situations - open()/close() in
> > a tight loop in my x86_64 which earlier used to reach file limit
> > if I set batch limit to 10 and found no problem. This patch does set 
> > default RCU batch limit to 10 and changes it only when there is an RCU
> > flood.
> 
> OK this seems to work, I can't tell yet whether it help the latency I 
> reported, but rt_run_flush still produces terrible latencies.
> 
> Ingo, should I try the softirq preemption patch + Dipankar's patch + 
> latency tracing patch?

yes, that would be a nice test. (I'm busy now with mutex stuff to be 
able to do a working softirq-preemption patch, but i sent you my current 
patches off-list - if you want to give it a shot. Be warned though, 
there will likely be quite some merging work to do, so it's definitely 
not for the faint hearted.)

	Ingo