Re: [PATCH] kmalloc_percpu

[William Lee Irwin III]

"David S. Miller" <[e-mail]> wrote:
>> As just pointed out by dipankar the only issue is NUMA...
>> so it has to be something more sophisticated than simply
>> kmalloc()[smp_processor_id];

On Tue, May 06, 2003 at 12:34:12AM -0700, Andrew Morton wrote:
> The proposed patch doesn't do anything about that either.
> +	ptr = alloc_bootmem(PERCPU_POOL_SIZE * NR_CPUS);
> So yes, we need an api which could be extended to use node-affine memory at
> some time in the future.  I think we have that.

IIRC that can be overridden; I wrote something to do node-local per-cpu
areas for i386 for some prior incarnations of per-cpu stuff and even
posted it, and this looks like it bootstraps at the same time (before
free_area_init_core() that is) and has the same override hook.

-- wli