On Tue, 27 May 2014 17:09:58 -0700 Nishanth Aravamudan [off-list ref] wrote:

On 28.05.2014 [09:56:14 +1000], Benjamin Herrenschmidt wrote:

quoted

On Tue, 2014-05-27 at 16:44 -0700, Nishanth Aravamudan wrote:

quoted

quoted

Signed-off-by: Nishanth Aravamudan <redacted>

Ping on this and patch 2/2. Ben, would you be willing to pull these
into
your -next branch so they'd get some testing?

http://patchwork.ozlabs.org/patch/350368/
http://patchwork.ozlabs.org/patch/349838/

Without any further changes, these two help quite a bit with the slab
consumption on CONFIG_SLUB kernels when memoryless nodes are present.

I don't mind at all :-) I haven't really been following that story
so I was waiting for the dust to settle and maybe acks from MM people
but if you tell me they are good I'm prepared to trust you.

The patches themselves are pretty minimal and similar to the ia64
changes (and the affected code seems like it hasn't changed in some
time, beyond in the common code). I'd mostly like to get some
broad-range build & boot testing.

Also, is NUMA a sufficient symbol to depend, you think? I figure most of
the PPC NUMA systems are the pSeries/IBM variety, which is where I've
run into memoryless nodes in the first place.

It's a powerpc-only patchset so I didn't do anything.

Nits:

- When referring to git commits, use 12 digits of hash and include
  the name of the patch as well (because patches can have different
  hashes in different trees).  So

       Based on 3bccd996276b ("numa: ia64: use generic percpu
       var numa_node_id() implementation") for ia64.

- It's nice to include a diffstat.  If there's a [0/n] patch then
  that's a great place for it, so people can see the overall impact at
  a glance.