Re: [PATCH] x86: use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz" again

[Ingo Molnar <mingo@kernel.org>]

* Rafael J. Wysocki <rjw@rjwysocki.net> wrote:

> On Wednesday, November 15, 2017 1:06:12 AM CET Linus Torvalds wrote:
> > On Tue, Nov 14, 2017 at 4:04 PM, Linus Torvalds
> > <torvalds@linux-foundation.org> wrote:
> > > On Tue, Nov 14, 2017 at 3:53 PM, Thomas Gleixner <tglx@linutronix.de> wrote:
> > >> Current head + Raphaels patch:
> > >>
> > >> real    0m0.029s
> > >> user    0m0.000s
> > >> sys     0m0.010s
> > >>
> > >> So that patch is actually slower.
> > >
> > > Oh it definitely is expected to be slower, because it does the IPI to
> > > all the cores and actually gets their frequency right.
> > >
> > > It was the old one that we had to revert (because it did so
> > > sequentially) that was really bad, and took something like 2+ seconds
> > > on Ingo's 160-core thing, iirc.
> > 
> > Looked it up. Ingo's machine "only" had 120 cores, and he said
> > 
> >     fomalhaut:~> time cat /proc/cpuinfo  >/dev/null
> >     real    0m2.689s
> > 
> > for the bad serial case, so yeah, it looks "a bit" better than it was ;)
> 
> OK, so may I queue it up?
> 
> I don't think I can get that to work substantially faster anyway ...

The new version is OK I suppose:

  Acked-by: Ingo Molnar <mingo@kernel.org>

I also think that /proc/cpuinfo is a pretty bad interface for many uses - I 
personally only very rarely need the cpuinfo of _all_ CPUs.

We we should eventually have /proc/cpu/N/info or so, so that 99% of the times 
cpuinfo is needed to report bugs we can do:

	cat /proc/cpu/0/info

With maybe also the following variants:

	/proc/cpu/first/
	/proc/cpu/last/
	/proc/cpu/current/

... to the first/last/current CPUs.

Thanks,

	Ingo