Re: 2.6.17-mm6

[Eric W. Biederman]

Ingo Molnar <mingo@xxxxxxx> writes:

> * Eric W. Biederman <ebiederm@xxxxxxxxxxxx> wrote:
>
>> What is scary is that at 1K cpus if we wind up using all of the irqs 
>> we start consuming 1Gig of RAM.  At only 128 cpus we are still in the 
>> 2M-15M territory, so that isn't too scary.  The point is that after a 
>> certain put the memory usage for all of those counters goes insane.
>
> we just need to move kernel_stat.irqs out of the per-cpu area and 
> alloc_percpu() a counter pointer for each IRQ that is truly set up. If 
> someone ends up using more than say 10,000 irqs we can reconsider. With 
> 10K irqs we'd have 10MB of stat counter footprint - that's reasonable.

Hmm.  at 10,000 IRQs and 128 cpus I got about 5MB.  But close enough.

I'm convinced at least for now.

The case I have seen are sparsely populated irqs.

And we really do need the high IRQ counts because the potential irqs
really do go up at 15 or so IRQs per cpu.

Still if we are doing this generically let's please put in a
big fat comment describing the situation, and maybe a compile warning,
if things get too big.

Eric
-
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/