CK5 [1/6] Use correct cpu_slab on dead cpu

CK5 [1/6] Use correct cpu_slab on dead cpu

[Christoph Lameter]

Pass a kmem_cache_cpu pointer into unfreeze partials so that a different
kmem_cache_cpu structure than the local one can be specified.

V1->V2:
 - Improve the comments

Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Christoph Lameter <cl@linux.com>

Index: linux/mm/slub.c
===================================================================
--- linux.orig/mm/slub.c	2012-11-05 09:05:23.572665087 -0600
+++ linux/mm/slub.c	2012-11-05 09:08:04.208740893 -0600
@@ -1869,12 +1869,14 @@ redo:
 /*
  * Unfreeze all the cpu partial slabs.
  *
- * This function must be called with interrupt disabled.
+ * This function must be called with interrupts disabled
+ * for the cpu using c (or some other guarantee must be there
+ * to guarantee no concurrent accesses).
  */
-static void unfreeze_partials(struct kmem_cache *s)
+static void unfreeze_partials(struct kmem_cache *s,
+		struct kmem_cache_cpu *c)
 {
 	struct kmem_cache_node *n = NULL, *n2 = NULL;
-	struct kmem_cache_cpu *c = this_cpu_ptr(s->cpu_slab);
 	struct page *page, *discard_page = NULL;
 
 	while ((page = c->partial)) {
@@ -1960,7 +1962,7 @@ static int put_cpu_partial(struct kmem_c
 				 * set to the per node partial list.
 				 */
 				local_irq_save(flags);
-				unfreeze_partials(s);
+				unfreeze_partials(s, this_cpu_ptr(s->cpu_slab));
 				local_irq_restore(flags);
 				oldpage = NULL;
 				pobjects = 0;
@@ -2003,7 +2005,7 @@ static inline void __flush_cpu_slab(stru
 		if (c->page)
 			flush_slab(s, c);
 
-		unfreeze_partials(s);
+		unfreeze_partials(s, c);
 	}
 }
 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

CK5 [2/6] slab: Simplify bootstrap

[Christoph Lameter]

The nodelists field in kmem_cache is pointing to the first unused
object in the array field when bootstrap is complete.

A problem with the current approach is that the statically sized
kmem_cache structure use on boot can only contain NR_CPUS entries.
If the number of nodes plus the number of cpus is greater then we
would overwrite memory following the kmem_cache_boot definition.

Increase the size of the array field to ensure that also the node
pointers fit into the array field.

Once we do that we no longer need the kmem_cache_nodelists
array and we can then also use that structure elsewhere.

V1->V2:
	- No need to zap kmem_cache->nodelists since it is allocated
		with kmem_cache_zalloc() [glommer]

Acked-by: Glauber Costa <glommer@parallels.com>
Signed-off-by: Christoph Lameter <cl@linux.com>
---
 include/linux/slab_def.h |    2 +-
 mm/slab.c                |   18 +++++++++++++-----
 2 files changed, 14 insertions(+), 6 deletions(-)

Index: linux/include/linux/slab_def.h
===================================================================
--- linux.orig/include/linux/slab_def.h	2012-11-15 15:26:24.940949492 -0600
+++ linux/include/linux/slab_def.h	2012-11-28 09:18:26.219105161 -0600
@@ -89,9 +89,13 @@ struct kmem_cache {
 	 * (see kmem_cache_init())
 	 * We still use [NR_CPUS] and not [1] or [0] because cache_cache
 	 * is statically defined, so we reserve the max number of cpus.
+	 *
+	 * We also need to guarantee that the list is able to accomodate a
+	 * pointer for each node since "nodelists" uses the remainder of
+	 * available pointers.
 	 */
 	struct kmem_list3 **nodelists;
-	struct array_cache *array[NR_CPUS];
+	struct array_cache *array[NR_CPUS + MAX_NUMNODES];
 	/*
 	 * Do not add fields after array[]
 	 */
Index: linux/mm/slab.c
===================================================================
--- linux.orig/mm/slab.c	2012-11-28 09:17:22.110761578 -0600
+++ linux/mm/slab.c	2012-11-28 09:18:26.219105161 -0600
@@ -553,9 +553,7 @@ static struct arraycache_init initarray_
     { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
 
 /* internal cache of cache description objs */
-static struct kmem_list3 *kmem_cache_nodelists[MAX_NUMNODES];
 static struct kmem_cache kmem_cache_boot = {
-	.nodelists = kmem_cache_nodelists,
 	.batchcount = 1,
 	.limit = BOOT_CPUCACHE_ENTRIES,
 	.shared = 1,
@@ -1560,6 +1558,15 @@ static void __init set_up_list3s(struct
 }
 
 /*
+ * The memory after the last cpu cache pointer is used for the
+ * the nodelists pointer.
+ */
+static void setup_nodelists_pointer(struct kmem_cache *cachep)
+{
+	cachep->nodelists = (struct kmem_list3 **)&cachep->array[nr_cpu_ids];
+}
+
+/*
  * Initialisation.  Called after the page allocator have been initialised and
  * before smp_init().
  */
@@ -1573,15 +1580,14 @@ void __init kmem_cache_init(void)
 	int node;
 
 	kmem_cache = &kmem_cache_boot;
+	setup_nodelists_pointer(kmem_cache);
 
 	if (num_possible_nodes() == 1)
 		use_alien_caches = 0;
 
-	for (i = 0; i < NUM_INIT_LISTS; i++) {
+	for (i = 0; i < NUM_INIT_LISTS; i++)
 		kmem_list3_init(&initkmem_list3[i]);
-		if (i < MAX_NUMNODES)
-			kmem_cache->nodelists[i] = NULL;
-	}
+
 	set_up_list3s(kmem_cache, CACHE_CACHE);
 
 	/*
@@ -1619,7 +1625,6 @@ void __init kmem_cache_init(void)
 	list_add(&kmem_cache->list, &slab_caches);
 	kmem_cache->colour_off = cache_line_size();
 	kmem_cache->array[smp_processor_id()] = &initarray_cache.cache;
-	kmem_cache->nodelists[node] = &initkmem_list3[CACHE_CACHE + node];
 
 	/*
 	 * struct kmem_cache size depends on nr_node_ids & nr_cpu_ids
@@ -2422,7 +2427,7 @@ __kmem_cache_create (struct kmem_cache *
 	else
 		gfp = GFP_NOWAIT;
 
-	cachep->nodelists = (struct kmem_list3 **)&cachep->array[nr_cpu_ids];
+	setup_nodelists_pointer(cachep);
 #if DEBUG
 
 	/*

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

CK5 [3/6] create common functions for boot slab creation

[Christoph Lameter]

Use a special function to create kmalloc caches and use that function in
SLAB and SLUB.

V1->V2:
	Do check for slasb state in slub's __kmem_cache_create to avoid
	unlocking a lock that was not taken
V2->V3:
	Remove slab_state check from sysfs_slab_add(). [Joonsoo]

V3->V4:
	- Use %zd instead of %td for size info.
	- Do not add slab caches to the list of slab caches during early
	  boot.

Acked-by: Joonsoo Kim <js1304@gmail.com>
Reviewed-by: Glauber Costa <glommer@parallels.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Christoph Lameter <cl@linux.com>
---
 mm/slab.c        |   48 ++++++++++++++----------------------------------
 mm/slab.h        |    5 +++++
 mm/slab_common.c |   32 ++++++++++++++++++++++++++++++++
 mm/slub.c        |   36 +++---------------------------------
 4 files changed, 54 insertions(+), 67 deletions(-)

Index: linux/mm/slab.c
===================================================================
--- linux.orig/mm/slab.c	2012-11-05 09:08:34.692753484 -0600
+++ linux/mm/slab.c	2012-11-05 09:11:52.020821613 -0600
@@ -1659,23 +1659,13 @@ void __init kmem_cache_init(void)
 	 * bug.
 	 */
 
-	sizes[INDEX_AC].cs_cachep = kmem_cache_zalloc(kmem_cache, GFP_NOWAIT);
-	sizes[INDEX_AC].cs_cachep->name = names[INDEX_AC].name;
-	sizes[INDEX_AC].cs_cachep->size = sizes[INDEX_AC].cs_size;
-	sizes[INDEX_AC].cs_cachep->object_size = sizes[INDEX_AC].cs_size;
-	sizes[INDEX_AC].cs_cachep->align = ARCH_KMALLOC_MINALIGN;
-	__kmem_cache_create(sizes[INDEX_AC].cs_cachep, ARCH_KMALLOC_FLAGS|SLAB_PANIC);
-	list_add(&sizes[INDEX_AC].cs_cachep->list, &slab_caches);
+	sizes[INDEX_AC].cs_cachep = create_kmalloc_cache(names[INDEX_AC].name,
+					sizes[INDEX_AC].cs_size, ARCH_KMALLOC_FLAGS);
 
-	if (INDEX_AC != INDEX_L3) {
-		sizes[INDEX_L3].cs_cachep = kmem_cache_zalloc(kmem_cache, GFP_NOWAIT);
-		sizes[INDEX_L3].cs_cachep->name = names[INDEX_L3].name;
-		sizes[INDEX_L3].cs_cachep->size = sizes[INDEX_L3].cs_size;
-		sizes[INDEX_L3].cs_cachep->object_size = sizes[INDEX_L3].cs_size;
-		sizes[INDEX_L3].cs_cachep->align = ARCH_KMALLOC_MINALIGN;
-		__kmem_cache_create(sizes[INDEX_L3].cs_cachep, ARCH_KMALLOC_FLAGS|SLAB_PANIC);
-		list_add(&sizes[INDEX_L3].cs_cachep->list, &slab_caches);
-	}
+	if (INDEX_AC != INDEX_L3)
+		sizes[INDEX_L3].cs_cachep =
+			create_kmalloc_cache(names[INDEX_L3].name,
+				sizes[INDEX_L3].cs_size, ARCH_KMALLOC_FLAGS);
 
 	slab_early_init = 0;
 
@@ -1687,24 +1677,14 @@ void __init kmem_cache_init(void)
 		 * Note for systems short on memory removing the alignment will
 		 * allow tighter packing of the smaller caches.
 		 */
-		if (!sizes->cs_cachep) {
-			sizes->cs_cachep = kmem_cache_zalloc(kmem_cache, GFP_NOWAIT);
-			sizes->cs_cachep->name = names->name;
-			sizes->cs_cachep->size = sizes->cs_size;
-			sizes->cs_cachep->object_size = sizes->cs_size;
-			sizes->cs_cachep->align = ARCH_KMALLOC_MINALIGN;
-			__kmem_cache_create(sizes->cs_cachep, ARCH_KMALLOC_FLAGS|SLAB_PANIC);
-			list_add(&sizes->cs_cachep->list, &slab_caches);
-		}
+		if (!sizes->cs_cachep)
+			sizes->cs_cachep = create_kmalloc_cache(names->name,
+					sizes->cs_size, ARCH_KMALLOC_FLAGS);
+
 #ifdef CONFIG_ZONE_DMA
-		sizes->cs_dmacachep = kmem_cache_zalloc(kmem_cache, GFP_NOWAIT);
-		sizes->cs_dmacachep->name = names->name_dma;
-		sizes->cs_dmacachep->size = sizes->cs_size;
-		sizes->cs_dmacachep->object_size = sizes->cs_size;
-		sizes->cs_dmacachep->align = ARCH_KMALLOC_MINALIGN;
-		__kmem_cache_create(sizes->cs_dmacachep,
-			       ARCH_KMALLOC_FLAGS|SLAB_CACHE_DMA| SLAB_PANIC);
-		list_add(&sizes->cs_dmacachep->list, &slab_caches);
+		sizes->cs_dmacachep = create_kmalloc_cache(
+			names->name_dma, sizes->cs_size,
+			SLAB_CACHE_DMA|ARCH_KMALLOC_FLAGS);
 #endif
 		sizes++;
 		names++;
Index: linux/mm/slab.h
===================================================================
--- linux.orig/mm/slab.h	2012-11-05 09:05:23.568665085 -0600
+++ linux/mm/slab.h	2012-11-05 09:11:52.020821613 -0600
@@ -35,6 +35,11 @@ extern struct kmem_cache *kmem_cache;
 /* Functions provided by the slab allocators */
 extern int __kmem_cache_create(struct kmem_cache *, unsigned long flags);
 
+extern struct kmem_cache *create_kmalloc_cache(const char *name, size_t size,
+			unsigned long flags);
+extern void create_boot_cache(struct kmem_cache *, const char *name,
+			size_t size, unsigned long flags);
+
 #ifdef CONFIG_SLUB
 struct kmem_cache *__kmem_cache_alias(const char *name, size_t size,
 	size_t align, unsigned long flags, void (*ctor)(void *));
Index: linux/mm/slab_common.c
===================================================================
--- linux.orig/mm/slab_common.c	2012-11-05 09:05:23.568665085 -0600
+++ linux/mm/slab_common.c	2012-11-05 09:11:52.020821613 -0600
@@ -202,6 +202,42 @@ int slab_is_available(void)
 	return slab_state >= UP;
 }
 
+#ifndef CONFIG_SLOB
+/* Create a cache during boot when no slab services are available yet */
+void __init create_boot_cache(struct kmem_cache *s, const char *name, size_t size,
+		unsigned long flags)
+{
+	int err;
+
+	s->name = name;
+	s->size = s->object_size = size;
+	s->align = ARCH_KMALLOC_MINALIGN;
+	err = __kmem_cache_create(s, flags);
+
+	if (err)
+		panic("Creation of kmalloc slab %s size=%zd failed. Reason %d\n",
+					name, size, err);
+
+	s->refcount = -1;	/* Exempt from merging for now */
+}
+
+struct kmem_cache *__init create_kmalloc_cache(const char *name, size_t size,
+				unsigned long flags)
+{
+	struct kmem_cache *s = kmem_cache_zalloc(kmem_cache, GFP_NOWAIT);
+
+	if (!s)
+		panic("Out of memory when creating slab %s\n", name);
+
+	create_boot_cache(s, name, size, flags);
+	list_add(&s->list, &slab_caches);
+	s->refcount = 1;
+	return s;
+}
+
+#endif /* !CONFIG_SLOB */
+
+
 #ifdef CONFIG_SLABINFO
 static void print_slabinfo_header(struct seq_file *m)
 {
Index: linux/mm/slub.c
===================================================================
--- linux.orig/mm/slub.c	2012-11-05 09:08:04.208740893 -0600
+++ linux/mm/slub.c	2012-11-05 09:11:52.020821613 -0600
@@ -3245,32 +3245,6 @@ static int __init setup_slub_nomerge(cha
 
 __setup("slub_nomerge", setup_slub_nomerge);
 
-static struct kmem_cache *__init create_kmalloc_cache(const char *name,
-						int size, unsigned int flags)
-{
-	struct kmem_cache *s;
-
-	s = kmem_cache_zalloc(kmem_cache, GFP_NOWAIT);
-
-	s->name = name;
-	s->size = s->object_size = size;
-	s->align = ARCH_KMALLOC_MINALIGN;
-
-	/*
-	 * This function is called with IRQs disabled during early-boot on
-	 * single CPU so there's no need to take slab_mutex here.
-	 */
-	if (kmem_cache_open(s, flags))
-		goto panic;
-
-	list_add(&s->list, &slab_caches);
-	return s;
-
-panic:
-	panic("Creation of kmalloc slab %s size=%d failed.\n", name, size);
-	return NULL;
-}
-
 /*
  * Conversion table for small slabs sizes / 8 to the index in the
  * kmalloc array. This is necessary for slabs < 192 since we have non power
@@ -3948,6 +3922,10 @@ int __kmem_cache_create(struct kmem_cach
 	if (err)
 		return err;
 
+	/* Mutex is not taken during early boot */
+	if (slab_state <= UP)
+		return 0;
+
 	mutex_unlock(&slab_mutex);
 	err = sysfs_slab_add(s);
 	mutex_lock(&slab_mutex);
@@ -5249,13 +5227,8 @@ static int sysfs_slab_add(struct kmem_ca
 {
 	int err;
 	const char *name;
-	int unmergeable;
-
-	if (slab_state < FULL)
-		/* Defer until later */
-		return 0;
+	int unmergeable = slab_unmergeable(s);
 
-	unmergeable = slab_unmergeable(s);
 	if (unmergeable) {
 		/*
 		 * Slabcache can never be merged so we can use the name proper.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

CK5 [4/6] slub: Use a statically allocated kmem_cache boot structure for bootstrap

[Christoph Lameter]

Simplify bootstrap by statically allocated two kmem_cache structures. These are
freed after bootup is complete. Allows us to no longer worry about calculations
of sizes of kmem_cache structures during bootstrap.

V1->V2:
	- Use kmem_cache_zalloc to properly zero structures.
	- Simplify setup by introducing a new boottime
		function "bootstrap()".

V2->V3:
	- bootstrap function adds slab to list of slab caches
	 *after* performing relocation.
	- Get of of kmem_size.
	- Make static boot structures local to kmem_cache_init

Reviewed-by: Glauber Costa <glommer@parallels.com>
Signed-off-by: Christoph Lameter <cl@linux.com>
---
 mm/slub.c |   41 +++++++++++------------------------------
 1 file changed, 11 insertions(+), 30 deletions(-)

Index: linux/mm/slub.c
===================================================================
--- linux.orig/mm/slub.c	2012-11-28 09:18:31.339135471 -0600
+++ linux/mm/slub.c	2012-11-28 09:18:33.419143788 -0600
@@ -176,8 +176,6 @@ static inline int kmem_cache_debug(struc
 #define __OBJECT_POISON		0x80000000UL /* Poison object */
 #define __CMPXCHG_DOUBLE	0x40000000UL /* Use cmpxchg_double */
 
-static int kmem_size = sizeof(struct kmem_cache);
-
 #ifdef CONFIG_SMP
 static struct notifier_block slab_notifier;
 #endif
@@ -3634,15 +3632,16 @@ static int slab_memory_callback(struct n
 
 /*
  * Used for early kmem_cache structures that were allocated using
- * the page allocator
+ * the page allocator. Allocate them properly then fix up the pointers
+ * that may be pointing to the wrong kmem_cache structure.
  */
 
-static void __init kmem_cache_bootstrap_fixup(struct kmem_cache *s)
+static struct kmem_cache * __init bootstrap(struct kmem_cache *static_cache)
 {
 	int node;
+	struct kmem_cache *s = kmem_cache_zalloc(kmem_cache, GFP_NOWAIT);
 
-	list_add(&s->list, &slab_caches);
-	s->refcount = -1;
+	memcpy(s, static_cache, kmem_cache->object_size);
 
 	for_each_node_state(node, N_NORMAL_MEMORY) {
 		struct kmem_cache_node *n = get_node(s, node);
@@ -3658,70 +3657,44 @@ static void __init kmem_cache_bootstrap_
 #endif
 		}
 	}
+	list_add(&s->list, &slab_caches);
+	return s;
 }
 
 void __init kmem_cache_init(void)
 {
+	static __initdata struct kmem_cache boot_kmem_cache,
+		boot_kmem_cache_node;
 	int i;
-	int caches = 0;
-	struct kmem_cache *temp_kmem_cache;
-	int order;
-	struct kmem_cache *temp_kmem_cache_node;
-	unsigned long kmalloc_size;
+	int caches = 2;
 
 	if (debug_guardpage_minorder())
 		slub_max_order = 0;
 
-	kmem_size = offsetof(struct kmem_cache, node) +
-			nr_node_ids * sizeof(struct kmem_cache_node *);
+	kmem_cache_node = &boot_kmem_cache_node;
+	kmem_cache = &boot_kmem_cache;
 
-	/* Allocate two kmem_caches from the page allocator */
-	kmalloc_size = ALIGN(kmem_size, cache_line_size());
-	order = get_order(2 * kmalloc_size);
-	kmem_cache = (void *)__get_free_pages(GFP_NOWAIT | __GFP_ZERO, order);
-
-	/*
-	 * Must first have the slab cache available for the allocations of the
-	 * struct kmem_cache_node's. There is special bootstrap code in
-	 * kmem_cache_open for slab_state == DOWN.
-	 */
-	kmem_cache_node = (void *)kmem_cache + kmalloc_size;
-
-	kmem_cache_node->name = "kmem_cache_node";
-	kmem_cache_node->size = kmem_cache_node->object_size =
-		sizeof(struct kmem_cache_node);
-	kmem_cache_open(kmem_cache_node, SLAB_HWCACHE_ALIGN | SLAB_PANIC);
+	create_boot_cache(kmem_cache_node, "kmem_cache_node",
+		sizeof(struct kmem_cache_node), SLAB_HWCACHE_ALIGN);
 
 	hotplug_memory_notifier(slab_memory_callback, SLAB_CALLBACK_PRI);
 
 	/* Able to allocate the per node structures */
 	slab_state = PARTIAL;
 
-	temp_kmem_cache = kmem_cache;
-	kmem_cache->name = "kmem_cache";
-	kmem_cache->size = kmem_cache->object_size = kmem_size;
-	kmem_cache_open(kmem_cache, SLAB_HWCACHE_ALIGN | SLAB_PANIC);
+	create_boot_cache(kmem_cache, "kmem_cache",
+			offsetof(struct kmem_cache, node) +
+				nr_node_ids * sizeof(struct kmem_cache_node *),
+		       SLAB_HWCACHE_ALIGN);
 
-	kmem_cache = kmem_cache_alloc(kmem_cache, GFP_NOWAIT);
-	memcpy(kmem_cache, temp_kmem_cache, kmem_size);
+	kmem_cache = bootstrap(&boot_kmem_cache);
 
 	/*
 	 * Allocate kmem_cache_node properly from the kmem_cache slab.
 	 * kmem_cache_node is separately allocated so no need to
 	 * update any list pointers.
 	 */
-	temp_kmem_cache_node = kmem_cache_node;
-
-	kmem_cache_node = kmem_cache_alloc(kmem_cache, GFP_NOWAIT);
-	memcpy(kmem_cache_node, temp_kmem_cache_node, kmem_size);
-
-	kmem_cache_bootstrap_fixup(kmem_cache_node);
-
-	caches++;
-	kmem_cache_bootstrap_fixup(kmem_cache);
-	caches++;
-	/* Free temporary boot structure */
-	free_pages((unsigned long)temp_kmem_cache, order);
+	kmem_cache_node = bootstrap(&boot_kmem_cache_node);
 
 	/* Now we can use the kmem_cache to allocate kmalloc slabs */
 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

CK5 [5/6] slab: Use the new create_boot_cache function to simplify bootstrap

[Christoph Lameter]

Simplify setup and reduce code in kmem_cache_init(). This allows us to
get rid of initarray_cache as well as the manual setup code for
the kmem_cache and kmem_cache_node arrays during bootstrap.

We introduce a new bootstrap state "PARTIAL" for slab that signals the
creation of a kmem_cache boot cache.

V1->V2: Get rid of initarray_cache as well.
V2->V3: Drop the setting of slab_state to PARTIAL [glommer]

Signed-off-by: Christoph Lameter <cl@linux.com>
---
 mm/slab.c |   51 ++++++++++++++++++---------------------------------
 1 file changed, 18 insertions(+), 33 deletions(-)

Index: linux/mm/slab.c
===================================================================
--- linux.orig/mm/slab.c	2012-11-28 09:18:31.339135471 -0600
+++ linux/mm/slab.c	2012-11-28 09:18:37.803167246 -0600
@@ -547,8 +547,6 @@ static struct cache_names __initdata cac
 #undef CACHE
 };
 
-static struct arraycache_init initarray_cache __initdata =
-    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
 static struct arraycache_init initarray_generic =
     { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
 
@@ -1572,12 +1570,9 @@ static void setup_nodelists_pointer(stru
  */
 void __init kmem_cache_init(void)
 {
-	size_t left_over;
 	struct cache_sizes *sizes;
 	struct cache_names *names;
 	int i;
-	int order;
-	int node;
 
 	kmem_cache = &kmem_cache_boot;
 	setup_nodelists_pointer(kmem_cache);
@@ -1618,36 +1613,16 @@ void __init kmem_cache_init(void)
 	 * 6) Resize the head arrays of the kmalloc caches to their final sizes.
 	 */
 
-	node = numa_mem_id();
-
 	/* 1) create the kmem_cache */
-	INIT_LIST_HEAD(&slab_caches);
-	list_add(&kmem_cache->list, &slab_caches);
-	kmem_cache->colour_off = cache_line_size();
-	kmem_cache->array[smp_processor_id()] = &initarray_cache.cache;
 
 	/*
 	 * struct kmem_cache size depends on nr_node_ids & nr_cpu_ids
 	 */
-	kmem_cache->size = offsetof(struct kmem_cache, array[nr_cpu_ids]) +
-				  nr_node_ids * sizeof(struct kmem_list3 *);
-	kmem_cache->object_size = kmem_cache->size;
-	kmem_cache->size = ALIGN(kmem_cache->object_size,
-					cache_line_size());
-	kmem_cache->reciprocal_buffer_size =
-		reciprocal_value(kmem_cache->size);
-
-	for (order = 0; order < MAX_ORDER; order++) {
-		cache_estimate(order, kmem_cache->size,
-			cache_line_size(), 0, &left_over, &kmem_cache->num);
-		if (kmem_cache->num)
-			break;
-	}
-	BUG_ON(!kmem_cache->num);
-	kmem_cache->gfporder = order;
-	kmem_cache->colour = left_over / kmem_cache->colour_off;
-	kmem_cache->slab_size = ALIGN(kmem_cache->num * sizeof(kmem_bufctl_t) +
-				      sizeof(struct slab), cache_line_size());
+	create_boot_cache(kmem_cache, "kmem_cache",
+		offsetof(struct kmem_cache, array[nr_cpu_ids]) +
+				  nr_node_ids * sizeof(struct kmem_list3 *),
+				  SLAB_HWCACHE_ALIGN);
+	list_add(&kmem_cache->list, &slab_caches);
 
 	/* 2+3) create the kmalloc caches */
 	sizes = malloc_sizes;
@@ -1695,7 +1670,6 @@ void __init kmem_cache_init(void)
 
 		ptr = kmalloc(sizeof(struct arraycache_init), GFP_NOWAIT);
 
-		BUG_ON(cpu_cache_get(kmem_cache) != &initarray_cache.cache);
 		memcpy(ptr, cpu_cache_get(kmem_cache),
 		       sizeof(struct arraycache_init));
 		/*
@@ -2250,7 +2224,15 @@ static int __init_refok setup_cpu_cache(
 
 	if (slab_state == DOWN) {
 		/*
-		 * Note: the first kmem_cache_create must create the cache
+		 * Note: Creation of first cache (kmem_cache).
+		 * The setup_list3s is taken care
+		 * of by the caller of __kmem_cache_create
+		 */
+		cachep->array[smp_processor_id()] = &initarray_generic.cache;
+		slab_state = PARTIAL;
+	} else if (slab_state == PARTIAL) {
+		/*
+		 * Note: the second kmem_cache_create must create the cache
 		 * that's used by kmalloc(24), otherwise the creation of
 		 * further caches will BUG().
 		 */
@@ -2258,7 +2240,7 @@ static int __init_refok setup_cpu_cache(
 
 		/*
 		 * If the cache that's used by kmalloc(sizeof(kmem_list3)) is
-		 * the first cache, then we need to set up all its list3s,
+		 * the second cache, then we need to set up all its list3s,
 		 * otherwise the creation of further caches will BUG().
 		 */
 		set_up_list3s(cachep, SIZE_AC);
@@ -2267,6 +2249,7 @@ static int __init_refok setup_cpu_cache(
 		else
 			slab_state = PARTIAL_ARRAYCACHE;
 	} else {
+		/* Remaining boot caches */
 		cachep->array[smp_processor_id()] =
 			kmalloc(sizeof(struct arraycache_init), gfp);
 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: CK5 [5/6] slab: Use the new create_boot_cache function to simplify bootstrap

[Glauber Costa]

On 11/28/2012 08:23 PM, Christoph Lameter wrote:
> Simplify setup and reduce code in kmem_cache_init(). This allows us to
> get rid of initarray_cache as well as the manual setup code for
> the kmem_cache and kmem_cache_node arrays during bootstrap.
> 
> We introduce a new bootstrap state "PARTIAL" for slab that signals the
> creation of a kmem_cache boot cache.
> 
> V1->V2: Get rid of initarray_cache as well.
> V2->V3: Drop the setting of slab_state to PARTIAL [glommer]
> 
> Signed-off-by: Christoph Lameter <cl@linux.com>

It looks good to me.

Reviewed-by: Glauber Costa <glommer@parallels.com>


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

CK5 [6/6] Common alignment code

[Christoph Lameter]

Extract the code to do object alignment from the allocators.
Do the alignment calculations in slab_common so that the
__kmem_cache_create functions of the allocators do not have
to deal with alignment.

Signed-off-by: Christoph Lameter <cl@linux.com>
---
 mm/slab.c        |   20 --------------------
 mm/slab.h        |    3 +++
 mm/slab_common.c |   32 ++++++++++++++++++++++++++++++--
 mm/slob.c        |   10 ----------
 mm/slub.c        |   38 +-------------------------------------
 5 files changed, 34 insertions(+), 69 deletions(-)

Index: linux/mm/slab.h
===================================================================
--- linux.orig/mm/slab.h	2012-11-28 09:18:31.339135471 -0600
+++ linux/mm/slab.h	2012-11-28 09:18:41.347186238 -0600
@@ -32,6 +32,9 @@ extern struct list_head slab_caches;
 /* The slab cache that manages slab cache information */
 extern struct kmem_cache *kmem_cache;
 
+unsigned long calculate_alignment(unsigned long flags,
+		unsigned long align, unsigned long size);
+
 /* Functions provided by the slab allocators */
 extern int __kmem_cache_create(struct kmem_cache *, unsigned long flags);
 
Index: linux/mm/slab_common.c
===================================================================
--- linux.orig/mm/slab_common.c	2012-11-28 09:18:31.339135471 -0600
+++ linux/mm/slab_common.c	2012-11-28 09:18:41.347186238 -0600
@@ -73,6 +73,34 @@ static inline int kmem_cache_sanity_chec
 #endif
 
 /*
+ * Figure out what the alignment of the objects will be given a set of
+ * flags, a user specified alignment and the size of the objects.
+ */
+unsigned long calculate_alignment(unsigned long flags,
+		unsigned long align, unsigned long size)
+{
+	/*
+	 * If the user wants hardware cache aligned objects then follow that
+	 * suggestion if the object is sufficiently large.
+	 *
+	 * The hardware cache alignment cannot override the specified
+	 * alignment though. If that is greater then use it.
+	 */
+	if (flags & SLAB_HWCACHE_ALIGN) {
+		unsigned long ralign = cache_line_size();
+		while (size <= ralign / 2)
+			ralign /= 2;
+		align = max(align, ralign);
+	}
+
+	if (align < ARCH_SLAB_MINALIGN)
+		align = ARCH_SLAB_MINALIGN;
+
+	return ALIGN(align, sizeof(void *));
+}
+
+
+/*
  * kmem_cache_create - Create a cache.
  * @name: A string which is used in /proc/slabinfo to identify this cache.
  * @size: The size of objects to be created in this cache.
@@ -124,7 +152,7 @@ struct kmem_cache *kmem_cache_create(con
 	s = kmem_cache_zalloc(kmem_cache, GFP_KERNEL);
 	if (s) {
 		s->object_size = s->size = size;
-		s->align = align;
+		s->align = calculate_alignment(flags, align, size);
 		s->ctor = ctor;
 		s->name = kstrdup(name, GFP_KERNEL);
 		if (!s->name) {
@@ -211,7 +239,7 @@ void __init create_boot_cache(struct kme
 
 	s->name = name;
 	s->size = s->object_size = size;
-	s->align = ARCH_KMALLOC_MINALIGN;
+	s->align = calculate_alignment(flags, ARCH_KMALLOC_MINALIGN, size);
 	err = __kmem_cache_create(s, flags);
 
 	if (err)
Index: linux/mm/slob.c
===================================================================
--- linux.orig/mm/slob.c	2012-11-28 09:18:18.583064160 -0600
+++ linux/mm/slob.c	2012-11-28 09:18:41.347186238 -0600
@@ -123,7 +123,6 @@ static inline void clear_slob_page_free(
 
 #define SLOB_UNIT sizeof(slob_t)
 #define SLOB_UNITS(size) (((size) + SLOB_UNIT - 1)/SLOB_UNIT)
-#define SLOB_ALIGN L1_CACHE_BYTES
 
 /*
  * struct slob_rcu is inserted at the tail of allocated slob blocks, which
@@ -527,20 +526,11 @@ EXPORT_SYMBOL(ksize);
 
 int __kmem_cache_create(struct kmem_cache *c, unsigned long flags)
 {
-	size_t align = c->size;
-
 	if (flags & SLAB_DESTROY_BY_RCU) {
 		/* leave room for rcu footer at the end of object */
 		c->size += sizeof(struct slob_rcu);
 	}
 	c->flags = flags;
-	/* ignore alignment unless it's forced */
-	c->align = (flags & SLAB_HWCACHE_ALIGN) ? SLOB_ALIGN : 0;
-	if (c->align < ARCH_SLAB_MINALIGN)
-		c->align = ARCH_SLAB_MINALIGN;
-	if (c->align < align)
-		c->align = align;
-
 	return 0;
 }
 
Index: linux/mm/slub.c
===================================================================
--- linux.orig/mm/slub.c	2012-11-28 09:18:33.419143788 -0600
+++ linux/mm/slub.c	2012-11-28 09:18:41.347186238 -0600
@@ -2760,32 +2760,6 @@ static inline int calculate_order(int si
 	return -ENOSYS;
 }
 
-/*
- * Figure out what the alignment of the objects will be.
- */
-static unsigned long calculate_alignment(unsigned long flags,
-		unsigned long align, unsigned long size)
-{
-	/*
-	 * If the user wants hardware cache aligned objects then follow that
-	 * suggestion if the object is sufficiently large.
-	 *
-	 * The hardware cache alignment cannot override the specified
-	 * alignment though. If that is greater then use it.
-	 */
-	if (flags & SLAB_HWCACHE_ALIGN) {
-		unsigned long ralign = cache_line_size();
-		while (size <= ralign / 2)
-			ralign /= 2;
-		align = max(align, ralign);
-	}
-
-	if (align < ARCH_SLAB_MINALIGN)
-		align = ARCH_SLAB_MINALIGN;
-
-	return ALIGN(align, sizeof(void *));
-}
-
 static void
 init_kmem_cache_node(struct kmem_cache_node *n)
 {
@@ -2919,7 +2893,6 @@ static int calculate_sizes(struct kmem_c
 {
 	unsigned long flags = s->flags;
 	unsigned long size = s->object_size;
-	unsigned long align = s->align;
 	int order;
 
 	/*
@@ -2991,19 +2964,11 @@ static int calculate_sizes(struct kmem_c
 #endif
 
 	/*
-	 * Determine the alignment based on various parameters that the
-	 * user specified and the dynamic determination of cache line size
-	 * on bootup.
-	 */
-	align = calculate_alignment(flags, align, s->object_size);
-	s->align = align;
-
-	/*
 	 * SLUB stores one object immediately after another beginning from
 	 * offset 0. In order to align the objects we have to simply size
 	 * each object to conform to the alignment.
 	 */
-	size = ALIGN(size, align);
+	size = ALIGN(size, s->align);
 	s->size = size;
 	if (forced_order >= 0)
 		order = forced_order;
@@ -3032,7 +2997,6 @@ static int calculate_sizes(struct kmem_c
 		s->max = s->oo;
 
 	return !!oo_objects(s->oo);
-
 }
 
 static int kmem_cache_open(struct kmem_cache *s, unsigned long flags)
Index: linux/mm/slab.c
===================================================================
--- linux.orig/mm/slab.c	2012-11-28 09:18:37.803167246 -0600
+++ linux/mm/slab.c	2012-11-28 09:18:41.347186238 -0600
@@ -2337,22 +2337,6 @@ __kmem_cache_create (struct kmem_cache *
 		size &= ~(BYTES_PER_WORD - 1);
 	}
 
-	/* calculate the final buffer alignment: */
-
-	/* 1) arch recommendation: can be overridden for debug */
-	if (flags & SLAB_HWCACHE_ALIGN) {
-		/*
-		 * Default alignment: as specified by the arch code.  Except if
-		 * an object is really small, then squeeze multiple objects into
-		 * one cacheline.
-		 */
-		ralign = cache_line_size();
-		while (size <= ralign / 2)
-			ralign /= 2;
-	} else {
-		ralign = BYTES_PER_WORD;
-	}
-
 	/*
 	 * Redzoning and user store require word alignment or possibly larger.
 	 * Note this will be overridden by architecture or caller mandated
@@ -2369,10 +2353,6 @@ __kmem_cache_create (struct kmem_cache *
 		size &= ~(REDZONE_ALIGN - 1);
 	}
 
-	/* 2) arch mandated alignment */
-	if (ralign < ARCH_SLAB_MINALIGN) {
-		ralign = ARCH_SLAB_MINALIGN;
-	}
 	/* 3) caller mandated alignment */
 	if (ralign < cachep->align) {
 		ralign = cachep->align;

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: CK5 [6/6] Common alignment code

[Glauber Costa]

On 11/28/2012 08:23 PM, Christoph Lameter wrote:
> Extract the code to do object alignment from the allocators.
> Do the alignment calculations in slab_common so that the
> __kmem_cache_create functions of the allocators do not have
> to deal with alignment.
> 
> Signed-off-by: Christoph Lameter <cl@linux.com>

Looks decent and straightforward enough.

Reviewed-by: Glauber Costa <glommer@parallels.com>

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>