Re: [PATCH] kmalloc_percpu

[Andrew Morton]

"David S. Miller" <[e-mail]> wrote:
>
> Make kmalloc_per_cpu() merely a convenience macro, made up of existing
> non-percpu primitives.

I think we're agreeing here.

The current kmalloc_percpu() is a wrapper around kmalloc.  That seems OK to
me.

What we _do_ want to solve is the problem that DEFINE_PERCPU() does not work
in modules.  Rusty's patch (reworked to not alter kmalloc_percpu) would suit
that requirement.

(kiran has a new version of kmalloc_percpu() which may be faster than the
current one, but for the purposes of this discussion it's equivalent).