Re: [PATCH] x86: Don't allow nr_irqs > NR_IRQS

["Yinghai Lu" <yinghai@kernel.org>]

On Tue, Nov 4, 2008 at 10:36 AM, Ben Hutchings
<bhutchings@solarflare.com> wrote:
> On Tue, 2008-11-04 at 21:00 +0300, Cyrill Gorcunov wrote:
> [...]
>> | I hit this when running net-next-2.6 (close to 2.6.28-rc3) on a
>> | Supermicro dual Xeon system.  NR_IRQS is 224 but probe_nr_irqs() detects
>> | 5 IOAPICs (!) and returns 240.  Here are the log messages:
>> |
>
> Well there must have been an earlier change that resulted in detecting 5
> IOAPICs instead of just 1, but that presumably would work as long as the
> irq_desc array was dynamically allocated.  This reversion breaks that.
>
> You don't really need to see the config; NR_IRQS is *always* 224 on
> normal x86-32 systems.
>

Yes. it should the same as setting to 64bit. we will cost some extra ram...

We can not make sparseirq/dyn_array to 2.6.28..., maybe 2.6.29 later.

Ingo,
Maybe we can put back dyn_array at first, that will still be 1:1
mapping..., and not waste memory..

YH