Re: [PATCH] Fix hotplug cpu on x86_64

[Andi Kleen]

On Friday 07 October 2005 17:07, Protasevich, Natalie wrote:
> > Brian Gerst wrote:
> > > Brian Gerst wrote:
> > >> I've been seeing bogus values from /proc/loadavg on an x86-64 SMP
> > >> kernel (but not UP).
> > >>
> > >> $ cat /proc/loadavg
> > >> -1012098.26 922203.26 -982431.60 1/112 2688
> > >>
> > >> This is in the current git tree.  I'm also seeing strange values in
> > >> /proc/stat:
> > >>
> > >> cpu  2489 40 920 60530 9398 171 288 1844674407350 cpu0 2509 60 940
> > >> 60550 9418 191 308 0
> > >>
> > >> The first line is the sum of all cpus (I only have one), so it's
> > >> picking up up bad data from the non-present cpus.  The last value,
> > >> stolen time, is completely bogus since that value is only
> >
> > ever used
> >
> > >> on s390.
> > >>
> > >> It looks to me like there is some problem with how the per-cpu
> > >> structures are being initialized, or are getting
> >
> > corrupted.  I have
> >
> > >> not been able to test i386 SMP yet to see if the problem is x86_64
> > >> specific.
> > >
> > > I found the culprit: CPU hotplug.  The problem is that
> > > prefill_possible_map() is called after setup_per_cpu_areas().  This
> > > leaves the per-cpu data sections for the future cpus uninitialized
> > > (still pointing to the original per-cpu data, which is initmem).
> > > Since the cpus exists in cpu_possible_map, for_each_cpu
> >
> > will iterate
> >
> > > over them even though the per-cpu data is invalid.
>
> I had to do the same in i386, but initially I was trying to avoid the
> whole situation - allocating per_cpu data for all possible processors.
> It seemed wasteful that on the system with NR_CPU=256 or 512 and brought
> up as 4x everything per_cpu is (pre)allocated for all, although it's

It's quite. With NR_CPUS==128 the current x86-64 code wastes around 4MB
in per CPU data :/ 

> sure convenient. I though at the time it would be great if
> alloc_percpu() mechanism was able to dynamically re-create all the
> per_cpu's for new processors, that way cpu_possible_map woun't probably
> even be needed. Or is it too much trouble for too little gain...

The problem is to tell all subsystems to reinitialize the data for possible
CPUs (essentially the concept of "possible" CPUs would need to go)  It would 
need an reaudit of a lot of code. Probably quite some work.

I think it is better to try to figure out how many hotplug CPUs are supported,
otherwise use a small default.  My current code uses disabled CPUs 
as a heuristic, otherwise half of available CPUs and it can be overwritten
by the user.

-Andi
-
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/