Commit e2cb96b7 authored Aug 19, 2008 by Christoph Lameter Committed by Pekka Enberg Aug 20, 2008

slub: Disable NUMA remote node defragmentation by default

Switch remote node defragmentation off by default. The current settings can
cause excessive node local allocations with hackbench:

  SLAB:

    % cat /proc/meminfo
    MemTotal:        7701760 kB
    MemFree:         5940096 kB
    Slab:             123840 kB

  SLUB:

    % cat /proc/meminfo
    MemTotal:        7701376 kB
    MemFree:         4740928 kB
    Slab:            1591680 kB

[Note: this feature is not related to slab defragmentation.]

You can find the original discussion here:

  http://lkml.org/lkml/2008/8/4/308



Reported-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Tested-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Christoph Lameter <cl@linux-foundation.org>
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>

parent 30a2f3c6

mm/slub.c

+2 −2

Original line number	Diff line number	Diff line
		@@ -2312,7 +2312,7 @@ static int kmem_cache_open(struct kmem_cache *s, gfp_t gfpflags,

		s->refcount = 1;
		#ifdef CONFIG_NUMA
		s->remote_node_defrag_ratio = 100;
		s->remote_node_defrag_ratio = 1000;
		#endif
		if (!init_kmem_cache_nodes(s, gfpflags & ~SLUB_DMA))
		goto error;
		@@ -4058,7 +4058,7 @@ static ssize_t remote_node_defrag_ratio_store(struct kmem_cache *s,
		if (err)
		return err;

		if (ratio < 100)
		if (ratio <= 100)
		s->remote_node_defrag_ratio = ratio * 10;

		return length;

Admin message