Re: [patch] latency tracer, 2.6.15-rc7

[Dipankar Sarma <dipankar@in.ibm.com>]

On Sat, Dec 31, 2005 at 12:14:26PM -0800, Paul E. McKenney wrote:
> So it seems to me that Linus's patch is part of the solution, but
> needs to also have a global component, perhaps as follows:
> 
> 	if (unlikely(rdp->count > 100)) {
> 		set_need_resched();
> 		if (unlikely(rdp->count - rdp->last_rs_count > 1000)) {
> 			int cpu;
> 
> 			rdp->last_rs_count = rdp->count;
> 			spin_lock_bh(&rcu_bh_state.lock);
> 			for_each_cpu_mask(cpu, rdp->rcu_bh_state.cpumask)
> 				smp_send_reschedule(cpu);
> 			spin_unlock_bh(&rcu_bh_state.lock);
> 		}
> 	}

Yes, something like this that covers corner cases and forces
queiscent state in all cpus, would be ideal.

> I am sure that I am missing some interaction or another with tickless
> idle and CPU hotplug covered.

It would be safe to miss a cpu or two while sending the resched
interrupt. So, I don't think we need to worry about tickless
idle and cpu hotplug.

> There also needs to be some adjustment in rcu_do_batch(), which will
> have to somehow get back to a quiescent state periodically.  Dipankar,
> Vatsa, thoughts?

My original thought was to make maxbatch dynamic and automatically
adjust it depending on the situation. I can try that approach.

Thanks
Dipankar