archive mirror
 help / color / mirror / Atom feed
From: Joonsoo Kim <>
To: Pekka Enberg <>
	Joonsoo Kim <>,
	David Miller <>, Neil Brown <>,
	Peter Zijlstra <>,
	Mike Christie <>,
	Eric B Munson <>,
	Eric Dumazet <>,
	Sebastian Andrzej Siewior <>,
	Mel Gorman <>,
	Christoph Lameter <>,
	Andrew Morton <>
Subject: [PATCH] slub: consider pfmemalloc_match() in get_partial_node()
Date: Sat, 25 Aug 2012 03:00:02 +0900	[thread overview]
Message-ID: <> (raw)

There is no consideration for pfmemalloc_match() in get_partial(). If we don't
consider that, we can't restrict access to PFMEMALLOC page mostly.

We may encounter following scenario.

Assume there is a request from normal allocation
and there is no objects in per cpu cache and no node partial slab.

In this case, slab_alloc go into slow-path and
new_slab_objects() is invoked. It may return PFMEMALLOC page.
Current user is not allowed to access PFMEMALLOC page,
deactivate_slab() is called (commit 5091b74a95d447e34530e713a8971450a45498b3),
then return object from PFMEMALLOC page.

Next time, when we meet another request from normal allocation,
slab_alloc() go into slow-path and re-go new_slab_objects().
In new_slab_objects(), we invoke get_partial() and we get a partial slab
which we have been deactivated just before, that is, PFMEMALLOC page.
We extract one object from it and re-deactivate.

"deactivate -> re-get in get_partial -> re-deactivate" occures repeatedly.

As a result, we can't restrict access to PFMEMALLOC page and
moreover, it introduce much performance degration to normal allocation
because of deactivation frequently.

Now, we need to consider pfmemalloc_match() in get_partial_node()
It prevent "deactivate -> re-get in get_partial".
Instead, new_slab() is called. It may return !PFMEMALLOC page,
so above situation will be suspended sometime.

Signed-off-by: Joonsoo Kim <>
Cc: David Miller <>
Cc: Neil Brown <>
Cc: Peter Zijlstra <>
Cc: Mike Christie <>
Cc: Eric B Munson <>
Cc: Eric Dumazet <>
Cc: Sebastian Andrzej Siewior <>
Cc: Mel Gorman <>
Cc: Christoph Lameter <>
Cc: Andrew Morton <>
This patch based on Pekka's slab/next tree with my two patches.

[PATCH 1/2] slub: rename cpu_partial to max_cpu_object

[PATCH 2/2] slub: correct the calculation of the number of cpu objects in get_partial_node

diff --git a/mm/slub.c b/mm/slub.c
index c96e0e4..a21da3a 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -1529,12 +1529,13 @@ static inline void *acquire_slab(struct kmem_cache *s,
 static int put_cpu_partial(struct kmem_cache *s, struct page *page, int drain);
+static inline bool pfmemalloc_match(struct page *page, gfp_t gfpflags);
  * Try to allocate a partial slab from a specific node.
-static void *get_partial_node(struct kmem_cache *s,
-		struct kmem_cache_node *n, struct kmem_cache_cpu *c)
+static void *get_partial_node(struct kmem_cache *s, struct kmem_cache_node *n,
+				struct kmem_cache_cpu *c, gfp_t flags)
 	struct page *page, *page2;
 	void *object = NULL;
@@ -1551,8 +1552,12 @@ static void *get_partial_node(struct kmem_cache *s,
 	list_for_each_entry_safe(page, page2, &n->partial, lru) {
-		void *t = acquire_slab(s, n, page, object == NULL);
+		void *t;
+		if (!pfmemalloc_match(page, flags))
+			continue;
+		t = acquire_slab(s, n, page, object == NULL);
 		if (!t)
@@ -1620,7 +1625,7 @@ static void *get_any_partial(struct kmem_cache *s, gfp_t flags,
 			if (n && cpuset_zone_allowed_hardwall(zone, flags) &&
 					n->nr_partial > s->min_partial) {
-				object = get_partial_node(s, n, c);
+				object = get_partial_node(s, n, c, flags);
 				if (object) {
 					 * Return the object even if
@@ -1649,7 +1654,7 @@ static void *get_partial(struct kmem_cache *s, gfp_t flags, int node,
 	void *object;
 	int searchnode = (node == NUMA_NO_NODE) ? numa_node_id() : node;
-	object = get_partial_node(s, get_node(s, searchnode), c);
+	object = get_partial_node(s, get_node(s, searchnode), c, flags);
 	if (object || node != NUMA_NO_NODE)
 		return object;

             reply	other threads:[~2012-08-24 18:01 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-08-24 18:00 Joonsoo Kim [this message]
2012-08-30 22:28 ` David Rientjes

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \
    --subject='Re: [PATCH] slub: consider pfmemalloc_match() in get_partial_node()' \

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).