Re: [GIT PULL] cputime patch for 2.6.30-rc6

[Peter Zijlstra]

On Wed, 2009-05-20 at 10:09 +0200, Martin Schwidefsky wrote:
> On Tue, 19 May 2009 11:31:28 +0200
> Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> 
> > On Tue, 2009-05-19 at 11:00 +0200, Martin Schwidefsky wrote:
> > > I don't see a problem here. In an idle multiple cpu system there IS
> > > more idle time than elapsed time. What would makes sense is to compare
> > > elapsed time * #cpus with the idle time. But then there is cpu hotplug
> > > which forces you to look at the delta of two measuring points where the
> > > number of cpus did not change.
> > 
> > Sure, this one case isn't that bad, esp. as you note its about idle
> > time. However, see for example /proc/stat and fs/proc/stat.c:
> > 
> >         for_each_possible_cpu(i) {
> >                 user = cputime64_add(user, kstat_cpu(i).cpustat.user);
> >                 nice = cputime64_add(nice, kstat_cpu(i).cpustat.nice);
> >                 system = cputime64_add(system, kstat_cpu(i).cpustat.system);
> >                 idle = cputime64_add(idle, kstat_cpu(i).cpustat.idle);
> >                 idle = cputime64_add(idle, arch_idle_time(i));
> >                 iowait = cputime64_add(iowait, kstat_cpu(i).cpustat.iowait);
> >                 irq = cputime64_add(irq, kstat_cpu(i).cpustat.irq);
> >                 softirq = cputime64_add(softirq, kstat_cpu(i).cpustat.softirq);
> >                 steal = cputime64_add(steal, kstat_cpu(i).cpustat.steal);
> >                 guest = cputime64_add(guest, kstat_cpu(i).cpustat.guest);
> >                 for_each_irq_nr(j) {
> >                         sum += kstat_irqs_cpu(j, i);
> >                 }
> >                 sum += arch_irq_stat_cpu(i);
> >         }
> > 
> > If that isn't a problem on a large machine, then I don't know what is.
> 
> Well, we better distinguish between the semantical problem and the
> performance consideration, no? One thing is what the proc field is
> supposed to contain, the other is how fast you can do it.
> 
> I have been refering to the semantical problem, but your point with the
> performance is very valid as well. So
> 1) are we agreed that the second field of /proc/uptime should contain
> the aggregate idle time of all cpus?

More or less, yeah ;-)

> 2) I agree that an endless loop of 'cat /proc/uptime' or
> 'cat /proc/stat' can have negative performance impact on a system with
> many cpus. One way to deal with it would be to restrict access to the
> interface. I do not like that idea to much. Another way would be to
> limit the number of for_each_possible_cpu loops per second. Create
> global variables that contain the aggregate values for the different
> fields. If the last update has been too recent (e.g. less than 0.1
> seconds ago), just print the old values again.

Right, I was thinking about using things like percpu_counter and vmstat
which get more inaccurate the larger your machine is, that is inaccurate
in the temporal sense, they do converge to the actual result in that
nothing is lost, it might just take a while to show up.


--
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/