Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1374394

Re: [PATCH 2/3] sched: Correctly handle nohz ticks cpu load accounting

Show all headers | View raw

On Fri, Apr 08, 2016 at 02:53:21PM +0200, Frederic Weisbecker wrote:
> On Fri, Apr 08, 2016 at 11:41:54AM +0200, Peter Zijlstra wrote:
> > On Fri, Apr 08, 2016 at 03:07:12AM +0200, Frederic Weisbecker wrote:
> > > +void cpu_load_update_nohz_start(void)
> > >  {
> > >  	struct rq *this_rq = this_rq();
> > > +
> > > +	/*
> > > +	 * This is all lockless but should be fine. If weighted_cpuload changes
> > > +	 * concurrently we'll exit nohz. And cpu_load write can race with
> > > +	 * cpu_load_update_idle() but both updater would be writing the same.
> > > +	 */
> > > +	this_rq->cpu_load[0] = weighted_cpuload(cpu_of(this_rq));
> > > +}
> > 
> > There is more to this; this also updates ->cpu_load[0] at possibly much
> > higher frequency than we've done before, while not updating the other
> > ->cpu_load[] members.
> > 
> > Now, I'm not sure we care, but it is a bit odd.
> 
> This is right. cpu_load[0] is aimed at showing the most recent load,
> without knowing when was the last update (the last tick/update could
> have been recent or not, the readers shouldn't care). Now we can
> indeed worry about it if this field is read altogether with the other
> indexes and those are put into some relation. But it seems that each
> of the rq->cpu_load[i] are read independently without relation or
> comparison. Now really I'm saying that without much assurance as I
> don't know the details of the load balancing code.
> 
> If in doubt I can create a field in struct rq to record the tickless
> load instead of storing it in cpu_load[0]. That was in fact the first
> direction I took in my drafts.

Yeah, no I think this is fine, just maybe wants a comment.

> > > +/*
> > > + * Account the tickless load in the end of a nohz frame.
> > > + */
> > > +void cpu_load_update_nohz_stop(void)
> > > +{
> > >  	unsigned long curr_jiffies = READ_ONCE(jiffies);
> > > +	struct rq *this_rq = this_rq();
> > > +	unsigned long load;
> > >  
> > >  	if (curr_jiffies == this_rq->last_load_update_tick)
> > >  		return;
> > >  
> > > +	load = weighted_cpuload(cpu_of(this_rq));
> > >  	raw_spin_lock(&this_rq->lock);
> > > +	cpu_load_update_nohz(this_rq, curr_jiffies, load);
> > >  	raw_spin_unlock(&this_rq->lock);
> > >  }
> > 
> > And this makes us take rq->lock when waking from nohz; a bit
> > unfortunate. Do we really need this though? 
> 
> Note it was already the case before this patchset, the function was called
> cpu_load_update_nohz() instead.

Ah, ok. I got lost in the whole rename + nohz maze (again). OK no
problem then.

Back to linux.kernel | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

[PATCH 0/3] sched: Fix/improve nohz cpu load updates v2 Frederic Weisbecker <fweisbec@gmail.com> - 2016-04-08 03:10 +0200
  [PATCH 2/3] sched: Correctly handle nohz ticks cpu load accounting Frederic Weisbecker <fweisbec@gmail.com> - 2016-04-08 03:10 +0200
    Re: [PATCH 2/3] sched: Correctly handle nohz ticks cpu load  accounting Peter Zijlstra <peterz@infradead.org> - 2016-04-08 11:50 +0200
      Re: [PATCH 2/3] sched: Correctly handle nohz ticks cpu load  accounting Frederic Weisbecker <fweisbec@gmail.com> - 2016-04-08 15:00 +0200
        Re: [PATCH 2/3] sched: Correctly handle nohz ticks cpu load  accounting Peter Zijlstra <peterz@infradead.org> - 2016-04-08 19:50 +0200
  [PATCH 1/3] sched: Gather cpu load functions under a more conventional namespace Frederic Weisbecker <fweisbec@gmail.com> - 2016-04-08 03:10 +0200
  [PATCH 3/3] sched: Optimize !CONFIG_NO_HZ_COMMON cpu load updates Frederic Weisbecker <fweisbec@gmail.com> - 2016-04-08 03:10 +0200
    Re: [PATCH 3/3] sched: Optimize !CONFIG_NO_HZ_COMMON cpu load updates Peter Zijlstra <peterz@infradead.org> - 2016-04-08 12:50 +0200
      Re: [PATCH 3/3] sched: Optimize !CONFIG_NO_HZ_COMMON cpu load updates Frederic Weisbecker <fweisbec@gmail.com> - 2016-04-08 15:00 +0200
        Re: [PATCH 3/3] sched: Optimize !CONFIG_NO_HZ_COMMON cpu load updates Peter Zijlstra <peterz@infradead.org> - 2016-04-08 19:50 +0200
          Re: [PATCH 3/3] sched: Optimize !CONFIG_NO_HZ_COMMON cpu load updates Frederic Weisbecker <fweisbec@gmail.com> - 2016-04-11 15:20 +0200
            Re: [PATCH 3/3] sched: Optimize !CONFIG_NO_HZ_COMMON cpu load updates Frederic Weisbecker <fweisbec@gmail.com> - 2016-04-11 20:30 +0200
              Re: [PATCH 3/3] sched: Optimize !CONFIG_NO_HZ_COMMON cpu load updates Peter Zijlstra <peterz@infradead.org> - 2016-04-12 16:30 +0200

csiph-web