lkml.org 
[lkml]   [2006]   [Jul]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: 2.6.17-mm6
From
Date
On Wed, 2006-07-05 at 17:25 -0700, Andrew Morton wrote:
> On Wed, 5 Jul 2006 17:05:49 -0700
> "Keith Mannthey" <kmannth@gmail.com> wrote:
>
> > On 7/5/06, Andrew Morton <akpm@osdl.org> wrote:
> > > On Wed, 5 Jul 2006 16:44:57 -0700
> > > Andrew Morton <akpm@osdl.org> wrote:
> > >
> > > > I guess a medium-term fix would be to add a boot parameter to override
> > > > PERCPU_ENOUGH_ROOM - it's hard to justify increasing it permanently just
> > > > for the benefit of the tiny minority of kernels which are hand-built with
> > > > lots of drivers in vmlinux.
> >

[snip]

>
> So you've been hit by the expansion of NR_IRQS which bloats kernel_stat
> which gobbles per-cpu data.
>
> In 2.6.17 NR_IRQS is 244. In -mm (due to the x86_64 genirq conversion)
> NR_IRQS became (256 + 32 * NR_CPUS). Hence the kstat "array" became
> two-dimensional. It's now O(NR_CPUS^2).
>
> I don't know what's a sane max for NR_CPUS on x86_64, but that'll sure be a
> showstopper if the ia64 guys try the same trick.
>
> I guess one fix would be to de-percpuify kernel_stat.irqs[]. Or
> dynamically allocate it with alloc_percpu().

And people wondered why I'm fighting for the robust per_cpu variables.

http://marc.theaimsgroup.com/?l=linux-kernel&m=114785997413023&w=2

Yes there's still problems with this. But if I ever get some more time
to work on it, I would like to solve those issues. Having that
PERCPU_ENOUGH_ROOM laying around in the kernel is just disgusting ;)

Sorry, for the noise, I have another 2288 more LKML emails to read :)

-- Steve


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2006-07-12 05:59    [W:0.107 / U:1.172 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site