linux-kernel - Re: [PATCH] mm-slab: allocate kmem_cache with __GFP

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Wed, 20 Jul 2011 17:56:02 +0200
From:	Eric Dumazet <eric.dumazet@...il.com>
To:	Christoph Lameter <cl@...ux.com>
Cc:	Mel Gorman <mgorman@...e.de>, Pekka Enberg <penberg@...nel.org>,
	Konstantin Khlebnikov <khlebnikov@...nvz.org>,
	Andrew Morton <akpm@...ux-foundation.org>, linux-mm@...ck.org,
	linux-kernel@...r.kernel.org, Matt Mackall <mpm@...enic.com>
Subject: Re: [PATCH] mm-slab: allocate kmem_cache with __GFP_REPEAT

Le mercredi 20 juillet 2011 à 10:34 -0500, Christoph Lameter a écrit :
> On Wed, 20 Jul 2011, Eric Dumazet wrote:
> 
> > [PATCH] slab: remove one NR_CPUS dependency
> 
> Ok simple enough.
> 
> Acked-by: Christoph Lameter <cl@...ux.com>
> 

Thanks Christoph

Here is the second patch, also simple and working for me (tested on
x86_64, NR_CPUS=4096, on my 2x4x2 machine)

Eventually, we could avoid the extra 'array' pointer if NR_CPUS is known
to be a small value (<= 16 for example)

Note that adding ____cacheline_aligned_in_smp on nodelists[] actually
helps performance, as all following fields are readonly after kmem_cache
setup.

[PATCH] slab: shrinks sizeof(struct kmem_cache)

Reduce high order allocations for some setups.
(NR_CPUS=4096 -> we need 64KB per kmem_cache struct)


Reported-by: Konstantin Khlebnikov <khlebnikov@...nvz.org>
Signed-off-by: Eric Dumazet <eric.dumazet@...il.com>
CC: Pekka Enberg <penberg@...nel.org>
CC: Christoph Lameter <cl@...ux.com>
CC: Andrew Morton <akpm@...ux-foundation.org>
---
 include/linux/slab_def.h |    4 ++--
 mm/slab.c                |   10 ++++++----
 2 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/include/linux/slab_def.h b/include/linux/slab_def.h
index 83203ae..abedd8e 100644
--- a/include/linux/slab_def.h
+++ b/include/linux/slab_def.h
@@ -51,7 +51,7 @@
 
 struct kmem_cache {
 /* 1) per-cpu data, touched during every alloc/free */
-	struct array_cache *array[NR_CPUS];
+	struct array_cache **array;
 /* 2) Cache tunables. Protected by cache_chain_mutex */
 	unsigned int batchcount;
 	unsigned int limit;
@@ -118,7 +118,7 @@ struct kmem_cache {
 	 * We still use [MAX_NUMNODES] and not [1] or [0] because cache_cache
 	 * is statically defined, so we reserve the max number of nodes.
 	 */
-	struct kmem_list3 *nodelists[MAX_NUMNODES];
+	struct kmem_list3 *nodelists[MAX_NUMNODES] ____cacheline_aligned_in_smp;
 	/*
 	 * Do not add fields after nodelists[]
 	 */
diff --git a/mm/slab.c b/mm/slab.c
index d96e223..f951015 100644
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -574,7 +574,9 @@ static struct arraycache_init initarray_generic =
     { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
 
 /* internal cache of cache description objs */
+static struct array_cache *array_cache_cache[NR_CPUS];
 static struct kmem_cache cache_cache = {
+	.array = array_cache_cache,
 	.batchcount = 1,
 	.limit = BOOT_CPUCACHE_ENTRIES,
 	.shared = 1,
@@ -1492,11 +1494,10 @@ void __init kmem_cache_init(void)
 	cache_cache.nodelists[node] = &initkmem_list3[CACHE_CACHE + node];
 
 	/*
-	 * struct kmem_cache size depends on nr_node_ids, which
-	 * can be less than MAX_NUMNODES.
+	 * struct kmem_cache size depends on nr_node_ids & nr_cpu_ids
 	 */
-	cache_cache.buffer_size = offsetof(struct kmem_cache, nodelists) +
-				 nr_node_ids * sizeof(struct kmem_list3 *);
+	cache_cache.buffer_size = offsetof(struct kmem_cache, nodelists[nr_node_ids]) +
+				 nr_cpu_ids * sizeof(struct array_cache *);
 #if DEBUG
 	cache_cache.obj_size = cache_cache.buffer_size;
 #endif
@@ -2308,6 +2309,7 @@ kmem_cache_create (const char *name, size_t size, size_t align,
 	if (!cachep)
 		goto oops;
 
+	cachep->array = (struct array_cache **)&cachep->nodelists[nr_node_ids];
 #if DEBUG
 	cachep->obj_size = size;
 


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/