Re: On migrate_disable() and latencies

[Peter Zijlstra <peterz@infradead.org>]

On Fri, 2011-07-22 at 19:45 +0200, Nicholas Mc Guire wrote:

> > Therefore the worst case latency is in the order of
> > max-migrate-disable-period * nr-cpus.
> 
> + something like sum of (interrupt rate [n] /
> max-migrate-disable-period * nr-cpus) * top-half handler [n]. if you
> go on with theoretical WCET analysis on multicore systems you will
> always end up at the conclusion that only UP is suitable for RT....

+ preemption cost + migration costs etc.. :-)

> > Currently we have no means of measuring these latencies, this is
> > something we need to grow, I think Steven can fairly easy craft a
> > migrate_disable runtime tracer -- it needs to use t->se.sum_exec_runtime
> > for measure so as to only count the actual time spend on the task and
> > ignore any time it was blocked.
>  
> well this is a similar problem as with the WCET "calculations" - you can
> calculate theoretical worst cases - but the question is what the actual
> distribution of "stacking" is and thus what the probability is that you
> manage to stack tasks in this way.

Sure, and yes those are thing to look at.

> One could I guess put some relatively simple instrumentation in to monitor
> this stacking problem - quit independant of actually measuring the times 

Yes, that would be very easy to do.

> > Once we have this, its back to the old game of 'lock'-breaking.
> >
> if the stacking problem does not practically exist then it might not be worth
> the effort to resolve it with elaborate lock breaking.

Fully agreed, its just that I wanted to share my understanding of the
problem space (both Thomas and I thought on-list email was a better
location that private IRC logs ;-), and we need to get the know the
problem before we can decide to ignore it ;-)

Ignoring unknowns always leads to surprises.