linux/mm/slab.c

//对请求页进行管理

/*
 * linux/mm/slab.c
 * Written by Mark Hemment, 1996/97.
 * ([e-mail])
 *
 * kmem_cache_destroy() + some cleanup - 1999 Andrea Arcangeli
 *
 * Major cleanup, different bufctl logic, per-cpu arrays
 * (c) 2000 Manfred Spraul
 *
 * Cleanup, make the head arrays unconditional, preparation for NUMA
 *  (c) 2002 Manfred Spraul
 *
 * An implementation of the Slab Allocator as described in outline in;
 * UNIX Internals: The New Frontiers by Uresh Vahalia
 * Pub: Prentice Hall ISBN 0-13-101908-2
 * or with a little more detail in;
 * The Slab Allocator: An Object-Caching Kernel Memory Allocator
 * Jeff Bonwick (Sun Microsystems).
 * Presented at: USENIX Summer 1994 Technical Conference
 *
 * The memory is organized in caches, one cache for each object type.
 * (e.g. inode_cache, dentry_cache, buffer_head, vm_area_struct)
 * Each cache consists out of many slabs (they are small (usually one
 * page long) and always contiguous), and each slab contains multiple
 * initialized objects.
 *
 * This means, that your constructor is used only for newly allocated
 * slabs and you must pass objects with the same initializations to
 * kmem_cache_free.
 *
 * Each cache can only support one memory type (GFP_DMA, GFP_HIGHMEM,
 * normal). If you need a special memory type, then must create a new
 * cache for that memory type.
 *
 * In order to reduce fragmentation, the slabs are sorted in 3 groups:
 *   full slabs with 0 free objects
 *   partial slabs
 *   empty slabs with no allocated objects
 *
 * If partial slabs exist, then new allocations come from these slabs,
 * otherwise from empty slabs or new slabs are allocated.
 *
 * kmem_cache_destroy() CAN CRASH if you try to allocate from the cache
 * during kmem_cache_destroy(). The caller must prevent concurrent allocs.
 *
 * Each cache has a short per-cpu head array, most allocs
 * and frees go into that array, and if that array overflows, then 1/2
 * of the entries in the array are given back into the global cache.
 * The head array is strictly LIFO and should improve the cache hit rates.
 * On SMP, it additionally reduces the spinlock operations.
 *
 * The c_cpuarray may not be read with enabled local interrupts -
 * it's changed with a smp_call_function().
 *
 * SMP synchronization:
 *  constructors and destructors are called without any locking.
 *  Several members in struct kmem_cache and struct slab never change, they
 * are accessed without any locking.
 *  The per-cpu arrays are never accessed from the wrong cpu, no locking,
 *   and local interrupts are disabled so slab code is preempt-safe.
 *  The non-constant members are protected with a per-cache irq spinlock.
 *
 * Many thanks to Mark Hemment, who wrote another per-cpu slab patch
 * in 2000 - many ideas in the current implementation are derived from
 * his patch.
 *
 * Further notes from the original documentation:
 *
 * 11 April '97.  Started multi-threading - markhe
 * The global cache-chain is protected by the mutex 'cache_chain_mutex'.
 * The sem is only needed when accessing/extending the cache-chain, which
 * can never happen inside an interrupt (kmem_cache_create(),
 * kmem_cache_shrink() and kmem_cache_reap()).
 *
 * At present, each engine can be growing a cache.  This should be blocked.
 *
 * 15 March 2005. NUMA slab allocator.
 * Shai Fultheim <[e-mail]>.
 * Shobhit Dayal <[e-mail]>
 * Alok N Kataria <[e-mail]>
 * Christoph Lameter <[e-mail]>
 *
 * Modified the slab allocator to be node aware on NUMA systems.
 * Each node has its own list of partial, free and full slabs.
 * All object allocations for a node occur from node specific slab lists.
 */

#include <linux/slab.h>
#include <linux/mm.h>
#include <linux/poison.h>
#include <linux/swap.h>
#include <linux/cache.h>
#include <linux/interrupt.h>
#include <linux/init.h>
#include <linux/compiler.h>
#include <linux/cpuset.h>
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
#include <linux/notifier.h>
#include <linux/kallsyms.h>
#include <linux/cpu.h>
#include <linux/sysctl.h>
#include <linux/module.h>
#include <linux/rcupdate.h>
#include <linux/string.h>
#include <linux/uaccess.h>
#include <linux/nodemask.h>
#include <linux/kmemleak.h>
#include <linux/mempolicy.h>
#include <linux/mutex.h>
#include <linux/fault-inject.h>
#include <linux/rtmutex.h>
#include <linux/reciprocal_div.h>
#include <linux/debugobjects.h>
#include <linux/kmemcheck.h>
#include <linux/memory.h>
#include <linux/prefetch.h>

#include <asm/cacheflush.h>
#include <asm/tlbflush.h>
#include <asm/page.h>

/*
 * DEBUG - 1 for kmem_cache_create() to honour; SLAB_RED_ZONE & SLAB_POISON.
 *    0 for faster, smaller code (especially in the critical paths).
 *
 * STATS - 1 to collect stats for /proc/slabinfo.
 *    0 for faster, smaller code (especially in the critical paths).
 *
 * FORCED_DEBUG - 1 enables SLAB_RED_ZONE and SLAB_POISON (if possible)
 */

#ifdef CONFIG_DEBUG_SLAB
#define DEBUG  1
#define STATS  1
#define FORCED_DEBUG 1
#else
#define DEBUG  0
#define STATS  0
#define FORCED_DEBUG 0
#endif

/* Shouldn't this be in a header file somewhere? */
#define BYTES_PER_WORD  sizeof(void *)
#define REDZONE_ALIGN  max(BYTES_PER_WORD, __alignof__(unsigned long long))

#ifndef ARCH_KMALLOC_FLAGS
#define ARCH_KMALLOC_FLAGS SLAB_HWCACHE_ALIGN
#endif

/* Legal flag mask for kmem_cache_create(). */
#if DEBUG
# define CREATE_MASK (SLAB_RED_ZONE | \
    SLAB_POISON | SLAB_HWCACHE_ALIGN | \
    SLAB_CACHE_DMA | \
    SLAB_STORE_USER | \
    SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
    SLAB_DESTROY_BY_RCU | SLAB_MEM_SPREAD | \
    SLAB_DEBUG_OBJECTS | SLAB_NOLEAKTRACE | SLAB_NOTRACK)
#else
# define CREATE_MASK (SLAB_HWCACHE_ALIGN | \
    SLAB_CACHE_DMA | \
    SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
    SLAB_DESTROY_BY_RCU | SLAB_MEM_SPREAD | \
    SLAB_DEBUG_OBJECTS | SLAB_NOLEAKTRACE | SLAB_NOTRACK)
#endif

/*
 * kmem_bufctl_t:
 *
 * Bufctl's are used for linking objs within a slab
 * linked offsets.
 *
 * This implementation relies on "struct page" for locating the cache &
 * slab an object belongs to.
 * This allows the bufctl structure to be small (one int), but limits
 * the number of objects a slab (not a cache) can contain when off-slab
 * bufctls are used. The limit is the size of the largest general cache
 * that does not use off-slab slabs.
 * For 32bit archs with 4 kB pages, is this 56.
 * This is not serious, as it is only for large objects, when it is unwise
 * to have too many per slab.
 * Note: This limit can be raised by introducing a general cache whose size
 * is less than 512 (PAGE_SIZE<<3), but greater than 256.
 */

typedef unsigned int kmem_bufctl_t;
#define BUFCTL_END (((kmem_bufctl_t)(~0U))-0)
#define BUFCTL_FREE (((kmem_bufctl_t)(~0U))-1)
#define BUFCTL_ACTIVE (((kmem_bufctl_t)(~0U))-2)
#define SLAB_LIMIT (((kmem_bufctl_t)(~0U))-3)

/*
 * struct slab_rcu
 *
 * slab_destroy on a SLAB_DESTROY_BY_RCU cache uses this structure to
 * arrange for kmem_freepages to be called via RCU.  This is useful if
 * we need to approach a kernel structure obliquely, from its address
 * obtained without the usual locking.  We can lock the structure to
 * stabilize it and check it's still at the given address, only if we
 * can be sure that the memory has not been meanwhile reused for some
 * other kind of object (which our subsystem's lock might corrupt).
 *
 * rcu_read_lock before reading the address, then rcu_read_unlock after
 * taking the spinlock within the structure expected at that address.
 */
struct slab_rcu {
 struct rcu_head head;
 struct kmem_cache *cachep;
 void *addr;
};

/*
 * struct slab
 *
 * Manages the objs in a slab. Placed either at the beginning of mem allocated
 * for a slab, or allocated from an general cache.
 * Slabs are chained into three list: fully used, partial, fully free slabs.
 */
struct slab {
 union {
  struct {
   struct list_head list;
   unsigned long colouroff;
   void *s_mem;  /* including colour offset */
   unsigned int inuse; /* num of objs active in slab */
   kmem_bufctl_t free;
   unsigned short nodeid;
  };
  struct slab_rcu __slab_cover_slab_rcu;
 };
};

/*
 * struct array_cache
 *
 * Purpose:
 * - LIFO ordering, to hand out cache-warm objects from _alloc
 * - reduce the number of linked list operations
 * - reduce spinlock operations
 *
 * The limit is stored in the per-cpu structure to reduce the data cache
 * footprint.
 *
 */
struct array_cache {
 unsigned int avail;
 unsigned int limit;
 unsigned int batchcount;
 unsigned int touched;
 spinlock_t lock;
 void *entry[]; /*
    * Must have this definition in here for the proper
    * alignment of array_cache. Also simplifies accessing
    * the entries.
    */
};

/*
 * bootstrap: The caches do not work without cpuarrays anymore, but the
 * cpuarrays are allocated from the generic caches...
 */
#define BOOT_CPUCACHE_ENTRIES 1
struct arraycache_init {
 struct array_cache cache;
 void *entries[BOOT_CPUCACHE_ENTRIES];
};

/*
 * The slab lists for all objects.
 */
struct kmem_list3 {
 struct list_head slabs_partial; /* partial list first, better asm code */
 struct list_head slabs_full;
 struct list_head slabs_free;
 unsigned long free_objects;
 unsigned int free_limit;
 unsigned int colour_next; /* Per-node cache coloring */
 spinlock_t list_lock;
 struct array_cache *shared; /* shared per node */
 struct array_cache **alien; /* on other nodes */
 unsigned long next_reap; /* updated without locking */
 int free_touched;  /* updated without locking */
};

/*
 * Need this for bootstrapping a per node allocator.
 */
#define NUM_INIT_LISTS (3 * MAX_NUMNODES)
static struct kmem_list3 __initdata initkmem_list3[NUM_INIT_LISTS];
#define CACHE_CACHE 0
#define SIZE_AC MAX_NUMNODES
#define SIZE_L3 (2 * MAX_NUMNODES)

static int drain_freelist(struct kmem_cache *cache,
   struct kmem_list3 *l3, int tofree);
static void free_block(struct kmem_cache *cachep, void **objpp, int len,
   int node);
static int enable_cpucache(struct kmem_cache *cachep, gfp_t gfp);
static void cache_reap(struct work_struct *unused);

/*
 * This function must be completely optimized away if a constant is passed to
 * it.  Mostly the same as what is in linux/slab.h except it returns an index.
 */
static __always_inline int index_of(const size_t size)
{
 extern void __bad_size(void);

 if (__builtin_constant_p(size)) {
  int i = 0;

#define CACHE(x) \
 if (size <=x) \
  return i; \
 else \
  i++;
#include <linux/kmalloc_sizes.h>
#undef CACHE
  __bad_size();
 } else
  __bad_size();
 return 0;
}

static int slab_early_init = 1;

#define INDEX_AC index_of(sizeof(struct arraycache_init))
#define INDEX_L3 index_of(sizeof(struct kmem_list3))

static void kmem_list3_init(struct kmem_list3 *parent)
{
 INIT_LIST_HEAD(&parent->slabs_full);
 INIT_LIST_HEAD(&parent->slabs_partial);
 INIT_LIST_HEAD(&parent->slabs_free);
 parent->shared = NULL;
 parent->alien = NULL;
 parent->colour_next = 0;
 spin_lock_init(&parent->list_lock);
 parent->free_objects = 0;
 parent->free_touched = 0;
}

#define MAKE_LIST(cachep, listp, slab, nodeid)    \
 do {        \
  INIT_LIST_HEAD(listp);     \
  list_splice(&(cachep->nodelists[nodeid]->slab), listp); \
 } while (0)

#define MAKE_ALL_LISTS(cachep, ptr, nodeid)    \
 do {        \
 MAKE_LIST((cachep), (&(ptr)->slabs_full), slabs_full, nodeid); \
 MAKE_LIST((cachep), (&(ptr)->slabs_partial), slabs_partial, nodeid); \
 MAKE_LIST((cachep), (&(ptr)->slabs_free), slabs_free, nodeid); \
 } while (0)

#define CFLGS_OFF_SLAB  (0x80000000UL)
#define OFF_SLAB(x) ((x)->flags & CFLGS_OFF_SLAB)

#define BATCHREFILL_LIMIT 16
/*
 * Optimization question: fewer reaps means less probability for unnessary
 * cpucache drain/refill cycles.
 *
 * OTOH the cpuarrays can contain lots of objects,
 * which could lock up otherwise freeable slabs.
 */
#define REAPTIMEOUT_CPUC (2*HZ)
#define REAPTIMEOUT_LIST3 (4*HZ)

#if STATS
#define STATS_INC_ACTIVE(x) ((x)->num_active++)
#define STATS_DEC_ACTIVE(x) ((x)->num_active--)
#define STATS_INC_ALLOCED(x) ((x)->num_allocations++)
#define STATS_INC_GROWN(x) ((x)->grown++)
#define STATS_ADD_REAPED(x,y) ((x)->reaped += (y))
#define STATS_SET_HIGH(x)      \
 do {        \
  if ((x)->num_active > (x)->high_mark)   \
   (x)->high_mark = (x)->num_active;  \
 } while (0)
#define STATS_INC_ERR(x) ((x)->errors++)
#define STATS_INC_NODEALLOCS(x) ((x)->node_allocs++)
#define STATS_INC_NODEFREES(x) ((x)->node_frees++)
#define STATS_INC_ACOVERFLOW(x)   ((x)->node_overflow++)
#define STATS_SET_FREEABLE(x, i)     \
 do {        \
  if ((x)->max_freeable < i)    \
   (x)->max_freeable = i;    \
 } while (0)
#define STATS_INC_ALLOCHIT(x) atomic_inc(&(x)->allochit)
#define STATS_INC_ALLOCMISS(x) atomic_inc(&(x)->allocmiss)
#define STATS_INC_FREEHIT(x) atomic_inc(&(x)->freehit)
#define STATS_INC_FREEMISS(x) atomic_inc(&(x)->freemiss)
#else
#define STATS_INC_ACTIVE(x) do { } while (0)
#define STATS_DEC_ACTIVE(x) do { } while (0)
#define STATS_INC_ALLOCED(x) do { } while (0)
#define STATS_INC_GROWN(x) do { } while (0)
#define STATS_ADD_REAPED(x,y) do { (void)(y); } while (0)
#define STATS_SET_HIGH(x) do { } while (0)
#define STATS_INC_ERR(x) do { } while (0)
#define STATS_INC_NODEALLOCS(x) do { } while (0)
#define STATS_INC_NODEFREES(x) do { } while (0)
#define STATS_INC_ACOVERFLOW(x)   do { } while (0)
#define STATS_SET_FREEABLE(x, i) do { } while (0)
#define STATS_INC_ALLOCHIT(x) do { } while (0)
#define STATS_INC_ALLOCMISS(x) do { } while (0)
#define STATS_INC_FREEHIT(x) do { } while (0)
#define STATS_INC_FREEMISS(x) do { } while (0)
#endif

#if DEBUG

/*
 * memory layout of objects:
 * 0  : objp
 * 0 .. cachep->obj_offset - BYTES_PER_WORD - 1: padding. This ensures that
 *   the end of an object is aligned with the end of the real
 *   allocation. Catches writes behind the end of the allocation.
 * cachep->obj_offset - BYTES_PER_WORD .. cachep->obj_offset - 1:
 *   redzone word.
 * cachep->obj_offset: The real object.
 * cachep->buffer_size - 2* BYTES_PER_WORD: redzone word [BYTES_PER_WORD long]
 * cachep->buffer_size - 1* BYTES_PER_WORD: last caller address
 *     [BYTES_PER_WORD long]
 */
static int obj_offset(struct kmem_cache *cachep)
{
 return cachep->obj_offset;
}

static int obj_size(struct kmem_cache *cachep)
{
 return cachep->obj_size;
}

static unsigned long long *dbg_redzone1(struct kmem_cache *cachep, void *objp)
{
 BUG_ON(!(cachep->flags & SLAB_RED_ZONE));
 return (unsigned long long*) (objp + obj_offset(cachep) -
          sizeof(unsigned long long));
}

static unsigned long long *dbg_redzone2(struct kmem_cache *cachep, void *objp)
{
 BUG_ON(!(cachep->flags & SLAB_RED_ZONE));
 if (cachep->flags & SLAB_STORE_USER)
  return (unsigned long long *)(objp + cachep->buffer_size -
           sizeof(unsigned long long) -
           REDZONE_ALIGN);
 return (unsigned long long *) (objp + cachep->buffer_size -
           sizeof(unsigned long long));
}

static void **dbg_userword(struct kmem_cache *cachep, void *objp)
{
 BUG_ON(!(cachep->flags & SLAB_STORE_USER));
 return (void **)(objp + cachep->buffer_size - BYTES_PER_WORD);
}

#else

#define obj_offset(x)   0
#define obj_size(cachep)  (cachep->buffer_size)
#define dbg_redzone1(cachep, objp) ({BUG(); (unsigned long long *)NULL;})
#define dbg_redzone2(cachep, objp) ({BUG(); (unsigned long long *)NULL;})
#define dbg_userword(cachep, objp) ({BUG(); (void **)NULL;})

#endif

#ifdef CONFIG_TRACING
size_t slab_buffer_size(struct kmem_cache *cachep)
{
 return cachep->buffer_size;
}
EXPORT_SYMBOL(slab_buffer_size);
#endif

/*
 * Do not go above this order unless 0 objects fit into the slab.
 */
#define BREAK_GFP_ORDER_HI 1
#define BREAK_GFP_ORDER_LO 0
static int slab_break_gfp_order = BREAK_GFP_ORDER_LO;

/*
 * Functions for storing/retrieving the cachep and or slab from the page
 * allocator.  These are used to find the slab an obj belongs to.  With kfree(),
 * these are used to find the cache which an obj belongs to.
 */
static inline void page_set_cache(struct page *page, struct kmem_cache *cache)
{
 page->lru.next = (struct list_head *)cache;
}

static inline struct kmem_cache *page_get_cache(struct page *page)
{
 page = compound_head(page);
 BUG_ON(!PageSlab(page));
 return (struct kmem_cache *)page->lru.next;
}

static inline void page_set_slab(struct page *page, struct slab *slab)
{
 page->lru.prev = (struct list_head *)slab;
}

static inline struct slab *page_get_slab(struct page *page)
{
 BUG_ON(!PageSlab(page));
 return (struct slab *)page->lru.prev;
}

static inline struct kmem_cache *virt_to_cache(const void *obj)
{
 struct page *page = virt_to_head_page(obj);
 return page_get_cache(page);
}

static inline struct slab *virt_to_slab(const void *obj)
{
 struct page *page = virt_to_head_page(obj);
 return page_get_slab(page);
}

static inline void *index_to_obj(struct kmem_cache *cache, struct slab *slab,
     unsigned int idx)
{
 return slab->s_mem + cache->buffer_size * idx;
}

/*
 * We want to avoid an expensive divide : (offset / cache->buffer_size)
 *   Using the fact that buffer_size is a constant for a particular cache,
 *   we can replace (offset / cache->buffer_size) by
 *   reciprocal_divide(offset, cache->reciprocal_buffer_size)
 */
static inline unsigned int obj_to_index(const struct kmem_cache *cache,
     const struct slab *slab, void *obj)
{
 u32 offset = (obj - slab->s_mem);
 return reciprocal_divide(offset, cache->reciprocal_buffer_size);
}

/*
 * These are the default caches for kmalloc. Custom caches can have other sizes.
 */
struct cache_sizes malloc_sizes[] = {
#define CACHE(x) { .cs_size = (x) },
#include <linux/kmalloc_sizes.h>
 CACHE(ULONG_MAX)
#undef CACHE
};
EXPORT_SYMBOL(malloc_sizes);

/* Must match cache_sizes above. Out of line to keep cache footprint low. */
struct cache_names {
 char *name;
 char *name_dma;
};

static struct cache_names __initdata cache_names[] = {
#define CACHE(x) { .name = "size-" #x, .name_dma = "size-" #x "(DMA)" },
#include <linux/kmalloc_sizes.h>
 {NULL,}
#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} };

/* internal cache of cache description objs */
static struct kmem_list3 *cache_cache_nodelists[MAX_NUMNODES];
static struct kmem_cache cache_cache = {
 .nodelists = cache_cache_nodelists,
 .batchcount = 1,
 .limit = BOOT_CPUCACHE_ENTRIES,
 .shared = 1,
 .buffer_size = sizeof(struct kmem_cache),
 .name = "kmem_cache",
};

#define BAD_ALIEN_MAGIC 0x01020304ul

/*
 * chicken and egg problem: delay the per-cpu array allocation
 * until the general caches are up.
 */
static enum {
 NONE,
 PARTIAL_AC,
 PARTIAL_L3,
 EARLY,
 LATE,
 FULL
} g_cpucache_up;

/*
 * used by boot code to determine if it can use slab based allocator
 */
int slab_is_available(void)
{
 return g_cpucache_up >= EARLY;
}

#ifdef CONFIG_LOCKDEP

/*
 * Slab sometimes uses the kmalloc slabs to store the slab headers
 * for other slabs "off slab".
 * The locking for this is tricky in that it nests within the locks
 * of all other slabs in a few places; to deal with this special
 * locking we put on-slab caches into a separate lock-class.
 *
 * We set lock class for alien array caches which are up during init.
 * The lock annotation will be lost if all cpus of a node goes down and
 * then comes back up during hotplug
 */
static struct lock_class_key on_slab_l3_key;
static struct lock_class_key on_slab_alc_key;

static struct lock_class_key debugobj_l3_key;
static struct lock_class_key debugobj_alc_key;

static void slab_set_lock_classes(struct kmem_cache *cachep,
  struct lock_class_key *l3_key, struct lock_class_key *alc_key,
  int q)
{
 struct array_cache **alc;
 struct kmem_list3 *l3;
 int r;

 l3 = cachep->nodelists[q];
 if (!l3)
  return;

 lockdep_set_class(&l3->list_lock, l3_key);
 alc = l3->alien;
 /*
  * FIXME: This check for BAD_ALIEN_MAGIC
  * should go away when common slab code is taught to
  * work even without alien caches.
  * Currently, non NUMA code returns BAD_ALIEN_MAGIC
  * for alloc_alien_cache,
  */
 if (!alc || (unsigned long)alc == BAD_ALIEN_MAGIC)
  return;
 for_each_node(r) {
  if (alc[r])
   lockdep_set_class(&alc[r]->lock, alc_key);
 }
}

static void slab_set_debugobj_lock_classes_node(struct kmem_cache *cachep, int node)
{
 slab_set_lock_classes(cachep, &debugobj_l3_key, &debugobj_alc_key, node);
}

static void slab_set_debugobj_lock_classes(struct kmem_cache *cachep)
{
 int node;

 for_each_online_node(node)
  slab_set_debugobj_lock_classes_node(cachep, node);
}

static void init_node_lock_keys(int q)
{
 struct cache_sizes *s = malloc_sizes;

 if (g_cpucache_up < LATE)
  return;

 for (s = malloc_sizes; s->cs_size != ULONG_MAX; s++) {
  struct kmem_list3 *l3;

  l3 = s->cs_cachep->nodelists[q];
  if (!l3 || OFF_SLAB(s->cs_cachep))
   continue;

  slab_set_lock_classes(s->cs_cachep, &on_slab_l3_key,
    &on_slab_alc_key, q);
 }
}

static inline void init_lock_keys(void)
{
 int node;

 for_each_node(node)
  init_node_lock_keys(node);
}
#else
static void init_node_lock_keys(int q)
{
}

static inline void init_lock_keys(void)
{
}

static void slab_set_debugobj_lock_classes_node(struct kmem_cache *cachep, int node)
{
}

static void slab_set_debugobj_lock_classes(struct kmem_cache *cachep)
{
}
#endif

/*
 * Guard access to the cache-chain.
 */
static DEFINE_MUTEX(cache_chain_mutex);
static struct list_head cache_chain;

static DEFINE_PER_CPU(struct delayed_work, slab_reap_work);

static inline struct array_cache *cpu_cache_get(struct kmem_cache *cachep)
{
 return cachep->array[smp_processor_id()];
}

static inline struct kmem_cache *__find_general_cachep(size_t size,
       gfp_t gfpflags)
{
 struct cache_sizes *csizep = malloc_sizes;

#if DEBUG
 /* This happens if someone tries to call
  * kmem_cache_create(), or __kmalloc(), before
  * the generic caches are initialized.
  */
 BUG_ON(malloc_sizes[INDEX_AC].cs_cachep == NULL);
#endif
 if (!size)
  return ZERO_SIZE_PTR;

 while (size > csizep->cs_size)
  csizep++;

 /*
  * Really subtle: The last entry with cs->cs_size==ULONG_MAX
  * has cs_{dma,}cachep==NULL. Thus no special case
  * for large kmalloc calls required.
  */
#ifdef CONFIG_ZONE_DMA
 if (unlikely(gfpflags & GFP_DMA))
  return csizep->cs_dmacachep;
#endif
 return csizep->cs_cachep;
}

static struct kmem_cache *kmem_find_general_cachep(size_t size, gfp_t gfpflags)
{
 return __find_general_cachep(size, gfpflags);
}

static size_t slab_mgmt_size(size_t nr_objs, size_t align)
{
 return ALIGN(sizeof(struct slab)+nr_objs*sizeof(kmem_bufctl_t), align);
}

/*
 * Calculate the number of objects and left-over bytes for a given buffer size.
 */
static void cache_estimate(unsigned long gfporder, size_t buffer_size,
      size_t align, int flags, size_t *left_over,
      unsigned int *num)
{
 int nr_objs;
 size_t mgmt_size;
 size_t slab_size = PAGE_SIZE << gfporder;

 /*
  * The slab management structure can be either off the slab or
  * on it. For the latter case, the memory allocated for a
  * slab is used for:
  *
  * - The struct slab
  * - One kmem_bufctl_t for each object
  * - Padding to respect alignment of @align
  * - @buffer_size bytes for each object
  *
  * If the slab management structure is off the slab, then the
  * alignment will already be calculated into the size. Because
  * the slabs are all pages aligned, the objects will be at the
  * correct alignment when allocated.
  */
 if (flags & CFLGS_OFF_SLAB) {
  mgmt_size = 0;
  nr_objs = slab_size / buffer_size;

  if (nr_objs > SLAB_LIMIT)
   nr_objs = SLAB_LIMIT;
 } else {
  /*
   * Ignore padding for the initial guess. The padding
   * is at most @align-1 bytes, and @buffer_size is at
   * least @align. In the worst case, this result will
   * be one greater than the number of objects that fit
   * into the memory allocation when taking the padding
   * into account.
   */
  nr_objs = (slab_size - sizeof(struct slab)) /
     (buffer_size + sizeof(kmem_bufctl_t));

  /*
   * This calculated number will be either the right
   * amount, or one greater than what we want.
   */
  if (slab_mgmt_size(nr_objs, align) + nr_objs*buffer_size
         > slab_size)
   nr_objs--;

  if (nr_objs > SLAB_LIMIT)
   nr_objs = SLAB_LIMIT;

  mgmt_size = slab_mgmt_size(nr_objs, align);
 }
 *num = nr_objs;
 *left_over = slab_size - nr_objs*buffer_size - mgmt_size;
}

#define slab_error(cachep, msg) __slab_error(__func__, cachep, msg)

static void __slab_error(const char *function, struct kmem_cache *cachep,
   char *msg)
{
 printk(KERN_ERR "slab error in %s(): cache `%s': %s\n",
        function, cachep->name, msg);
 dump_stack();
}

/*
 * By default on NUMA we use alien caches to stage the freeing of
 * objects allocated from other nodes. This causes massive memory
 * inefficiencies when using fake NUMA setup to split memory into a
 * large number of small nodes, so it can be disabled on the command
 * line
  */

static int use_alien_caches __read_mostly = 1;
static int __init noaliencache_setup(char *s)
{
 use_alien_caches = 0;
 return 1;
}
__setup("noaliencache", noaliencache_setup);

#ifdef CONFIG_NUMA
/*
 * Special reaping functions for NUMA systems called from cache_reap().
 * These take care of doing round robin flushing of alien caches (containing
 * objects freed on different nodes from which they were allocated) and the
 * flushing of remote pcps by calling drain_node_pages.
 */
static DEFINE_PER_CPU(unsigned long, slab_reap_node);

static void init_reap_node(int cpu)
{
 int node;

 node = next_node(cpu_to_mem(cpu), node_online_map);
 if (node == MAX_NUMNODES)
  node = first_node(node_online_map);

 per_cpu(slab_reap_node, cpu) = node;
}

static void next_reap_node(void)
{
 int node = __this_cpu_read(slab_reap_node);

 node = next_node(node, node_online_map);
 if (unlikely(node >= MAX_NUMNODES))
  node = first_node(node_online_map);
 __this_cpu_write(slab_reap_node, node);
}

#else
#define init_reap_node(cpu) do { } while (0)
#define next_reap_node(void) do { } while (0)
#endif

/*
 * Initiate the reap timer running on the target CPU.  We run at around 1 to 2Hz
 * via the workqueue/eventd.
 * Add the CPU number into the expiration time to minimize the possibility of
 * the CPUs getting into lockstep and contending for the global cache chain
 * lock.
 */
static void __cpuinit start_cpu_timer(int cpu)
{
 struct delayed_work *reap_work = &per_cpu(slab_reap_work, cpu);

 /*
  * When this gets called from do_initcalls via cpucache_init(),
  * init_workqueues() has already run, so keventd will be setup
  * at that time.
  */
 if (keventd_up() && reap_work->work.func == NULL) {
  init_reap_node(cpu);
  INIT_DELAYED_WORK_DEFERRABLE(reap_work, cache_reap);
  schedule_delayed_work_on(cpu, reap_work,
     __round_jiffies_relative(HZ, cpu));
 }
}

static struct array_cache *alloc_arraycache(int node, int entries,
         int batchcount, gfp_t gfp)
{
 int memsize = sizeof(void *) * entries + sizeof(struct array_cache);
 struct array_cache *nc = NULL;

 nc = kmalloc_node(memsize, gfp, node);
 /*
  * The array_cache structures contain pointers to free object.
  * However, when such objects are allocated or transferred to another
  * cache the pointers are not cleared and they could be counted as
  * valid references during a kmemleak scan. Therefore, kmemleak must
  * not scan such objects.
  */
 kmemleak_no_scan(nc);
 if (nc) {
  nc->avail = 0;
  nc->limit = entries;
  nc->batchcount = batchcount;
  nc->touched = 0;
  spin_lock_init(&nc->lock);
 }
 return nc;
}

/*
 * Transfer objects in one arraycache to another.
 * Locking must be handled by the caller.
 *
 * Return the number of entries transferred.
 */
static int transfer_objects(struct array_cache *to,
  struct array_cache *from, unsigned int max)
{
 /* Figure out how many entries to transfer */
 int nr = min3(from->avail, max, to->limit - to->avail);

 if (!nr)
  return 0;

 memcpy(to->entry + to->avail, from->entry + from->avail -nr,
   sizeof(void *) *nr);

 from->avail -= nr;
 to->avail += nr;
 return nr;
}

#ifndef CONFIG_NUMA

#define drain_alien_cache(cachep, alien) do { } while (0)
#define reap_alien(cachep, l3) do { } while (0)

static inline struct array_cache **alloc_alien_cache(int node, int limit, gfp_t gfp)
{
 return (struct array_cache **)BAD_ALIEN_MAGIC;
}

static inline void free_alien_cache(struct array_cache **ac_ptr)
{
}

static inline int cache_free_alien(struct kmem_cache *cachep, void *objp)
{
 return 0;
}

static inline void *alternate_node_alloc(struct kmem_cache *cachep,
  gfp_t flags)
{
 return NULL;
}

static inline void *____cache_alloc_node(struct kmem_cache *cachep,
   gfp_t flags, int nodeid)
{
 return NULL;
}

#else /* CONFIG_NUMA */

static void *____cache_alloc_node(struct kmem_cache *, gfp_t, int);
static void *alternate_node_alloc(struct kmem_cache *, gfp_t);

static struct array_cache **alloc_alien_cache(int node, int limit, gfp_t gfp)
{
 struct array_cache **ac_ptr;
 int memsize = sizeof(void *) * nr_node_ids;
 int i;

 if (limit > 1)
  limit = 12;
 ac_ptr = kzalloc_node(memsize, gfp, node);
 if (ac_ptr) {
  for_each_node(i) {
   if (i == node || !node_online(i))
    continue;
   ac_ptr[i] = alloc_arraycache(node, limit, 0xbaadf00d, gfp);
   if (!ac_ptr[i]) {
    for (i--; i >= 0; i--)
     kfree(ac_ptr[i]);
    kfree(ac_ptr);
    return NULL;
   }
  }
 }
 return ac_ptr;
}

static void free_alien_cache(struct array_cache **ac_ptr)
{
 int i;

 if (!ac_ptr)
  return;
 for_each_node(i)
     kfree(ac_ptr[i]);
 kfree(ac_ptr);
}

static void __drain_alien_cache(struct kmem_cache *cachep,
    struct array_cache *ac, int node)
{
 struct kmem_list3 *rl3 = cachep->nodelists[node];

 if (ac->avail) {
  spin_lock(&rl3->list_lock);
  /*
   * Stuff objects into the remote nodes shared array first.
   * That way we could avoid the overhead of putting the objects
   * into the free lists and getting them back later.
   */
  if (rl3->shared)
   transfer_objects(rl3->shared, ac, ac->limit);

  free_block(cachep, ac->entry, ac->avail, node);
  ac->avail = 0;
  spin_unlock(&rl3->list_lock);
 }
}

/*
 * Called from cache_reap() to regularly drain alien caches round robin.
 */
static void reap_alien(struct kmem_cache *cachep, struct kmem_list3 *l3)
{
 int node = __this_cpu_read(slab_reap_node);

 if (l3->alien) {
  struct array_cache *ac = l3->alien[node];

  if (ac && ac->avail && spin_trylock_irq(&ac->lock)) {
   __drain_alien_cache(cachep, ac, node);
   spin_unlock_irq(&ac->lock);
  }
 }
}

static void drain_alien_cache(struct kmem_cache *cachep,
    struct array_cache **alien)
{
 int i = 0;
 struct array_cache *ac;
 unsigned long flags;

 for_each_online_node(i) {
  ac = alien[i];
  if (ac) {
   spin_lock_irqsave(&ac->lock, flags);
   __drain_alien_cache(cachep, ac, i);
   spin_unlock_irqrestore(&ac->lock, flags);
  }
 }
}

static inline int cache_free_alien(struct kmem_cache *cachep, void *objp)
{
 struct slab *slabp = virt_to_slab(objp);
 int nodeid = slabp->nodeid;
 struct kmem_list3 *l3;
 struct array_cache *alien = NULL;
 int node;

 node = numa_mem_id();

 /*
  * Make sure we are not freeing a object from another node to the array
  * cache on this cpu.
  */
 if (likely(slabp->nodeid == node))
  return 0;

 l3 = cachep->nodelists[node];
 STATS_INC_NODEFREES(cachep);
 if (l3->alien && l3->alien[nodeid]) {
  alien = l3->alien[nodeid];
  spin_lock(&alien->lock);
  if (unlikely(alien->avail == alien->limit)) {
   STATS_INC_ACOVERFLOW(cachep);
   __drain_alien_cache(cachep, alien, nodeid);
  }
  alien->entry[alien->avail++] = objp;
  spin_unlock(&alien->lock);
 } else {
  spin_lock(&(cachep->nodelists[nodeid])->list_lock);
  free_block(cachep, &objp, 1, nodeid);
  spin_unlock(&(cachep->nodelists[nodeid])->list_lock);
 }
 return 1;
}
#endif

/*
 * Allocates and initializes nodelists for a node on each slab cache, used for
 * either memory or cpu hotplug.  If memory is being hot-added, the kmem_list3
 * will be allocated off-node since memory is not yet online for the new node.
 * When hotplugging memory or a cpu, existing nodelists are not replaced if
 * already in use.
 *
 * Must hold cache_chain_mutex.
 */
static int init_cache_nodelists_node(int node)
{
 struct kmem_cache *cachep;
 struct kmem_list3 *l3;
 const int memsize = sizeof(struct kmem_list3);

 list_for_each_entry(cachep, &cache_chain, next) {
  /*
   * Set up the size64 kmemlist for cpu before we can
   * begin anything. Make sure some other cpu on this
   * node has not already allocated this
   */
  if (!cachep->nodelists[node]) {
   l3 = kmalloc_node(memsize, GFP_KERNEL, node);
   if (!l3)
    return -ENOMEM;
   kmem_list3_init(l3);
   l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
       ((unsigned long)cachep) % REAPTIMEOUT_LIST3;

   /*
    * The l3s don't come and go as CPUs come and
    * go.  cache_chain_mutex is sufficient
    * protection here.
    */
   cachep->nodelists[node] = l3;
  }

  spin_lock_irq(&cachep->nodelists[node]->list_lock);
  cachep->nodelists[node]->free_limit =
   (1 + nr_cpus_node(node)) *
   cachep->batchcount + cachep->num;
  spin_unlock_irq(&cachep->nodelists[node]->list_lock);
 }
 return 0;
}

static void __cpuinit cpuup_canceled(long cpu)
{
 struct kmem_cache *cachep;
 struct kmem_list3 *l3 = NULL;
 int node = cpu_to_mem(cpu);
 const struct cpumask *mask = cpumask_of_node(node);

 list_for_each_entry(cachep, &cache_chain, next) {
  struct array_cache *nc;
  struct array_cache *shared;
  struct array_cache **alien;

  /* cpu is dead; no one can alloc from it. */
  nc = cachep->array[cpu];
  cachep->array[cpu] = NULL;
  l3 = cachep->nodelists[node];

  if (!l3)
   goto free_array_cache;

  spin_lock_irq(&l3->list_lock);

  /* Free limit for this kmem_list3 */
  l3->free_limit -= cachep->batchcount;
  if (nc)
   free_block(cachep, nc->entry, nc->avail, node);

  if (!cpumask_empty(mask)) {
   spin_unlock_irq(&l3->list_lock);
   goto free_array_cache;
  }

  shared = l3->shared;
  if (shared) {
   free_block(cachep, shared->entry,
       shared->avail, node);
   l3->shared = NULL;
  }

  alien = l3->alien;
  l3->alien = NULL;

  spin_unlock_irq(&l3->list_lock);

  kfree(shared);
  if (alien) {
   drain_alien_cache(cachep, alien);
   free_alien_cache(alien);
  }
free_array_cache:
  kfree(nc);
 }
 /*
  * In the previous loop, all the objects were freed to
  * the respective cache's slabs,  now we can go ahead and
  * shrink each nodelist to its limit.
  */
 list_for_each_entry(cachep, &cache_chain, next) {
  l3 = cachep->nodelists[node];
  if (!l3)
   continue;
  drain_freelist(cachep, l3, l3->free_objects);
 }
}

static int __cpuinit cpuup_prepare(long cpu)
{
 struct kmem_cache *cachep;
 struct kmem_list3 *l3 = NULL;
 int node = cpu_to_mem(cpu);
 int err;

 /*
  * We need to do this right in the beginning since
  * alloc_arraycache's are going to use this list.
  * kmalloc_node allows us to add the slab to the right
  * kmem_list3 and not this cpu's kmem_list3
  */
 err = init_cache_nodelists_node(node);
 if (err < 0)
  goto bad;

 /*
  * Now we can go ahead with allocating the shared arrays and
  * array caches
  */
 list_for_each_entry(cachep, &cache_chain, next) {
  struct array_cache *nc;
  struct array_cache *shared = NULL;
  struct array_cache **alien = NULL;

  nc = alloc_arraycache(node, cachep->limit,
     cachep->batchcount, GFP_KERNEL);
  if (!nc)
   goto bad;
  if (cachep->shared) {
   shared = alloc_arraycache(node,
    cachep->shared * cachep->batchcount,
    0xbaadf00d, GFP_KERNEL);
   if (!shared) {
    kfree(nc);
    goto bad;
   }
  }
  if (use_alien_caches) {
   alien = alloc_alien_cache(node, cachep->limit, GFP_KERNEL);
   if (!alien) {
    kfree(shared);
    kfree(nc);
    goto bad;
   }
  }
  cachep->array[cpu] = nc;
  l3 = cachep->nodelists[node];
  BUG_ON(!l3);

  spin_lock_irq(&l3->list_lock);
  if (!l3->shared) {
   /*
    * We are serialised from CPU_DEAD or
    * CPU_UP_CANCELLED by the cpucontrol lock
    */
   l3->shared = shared;
   shared = NULL;
  }
#ifdef CONFIG_NUMA
  if (!l3->alien) {
   l3->alien = alien;
   alien = NULL;
  }
#endif
  spin_unlock_irq(&l3->list_lock);
  kfree(shared);
  free_alien_cache(alien);
  if (cachep->flags & SLAB_DEBUG_OBJECTS)
   slab_set_debugobj_lock_classes_node(cachep, node);
 }
 init_node_lock_keys(node);

 return 0;
bad:
 cpuup_canceled(cpu);
 return -ENOMEM;
}

static int __cpuinit cpuup_callback(struct notifier_block *nfb,
        unsigned long action, void *hcpu)
{
 long cpu = (long)hcpu;
 int err = 0;

 switch (action) {
 case CPU_UP_PREPARE:
 case CPU_UP_PREPARE_FROZEN:
  mutex_lock(&cache_chain_mutex);
  err = cpuup_prepare(cpu);
  mutex_unlock(&cache_chain_mutex);
  break;
 case CPU_ONLINE:
 case CPU_ONLINE_FROZEN:
  start_cpu_timer(cpu);
  break;
#ifdef CONFIG_HOTPLUG_CPU
   case CPU_DOWN_PREPARE:
   case CPU_DOWN_PREPARE_FROZEN:
  /*
   * Shutdown cache reaper. Note that the cache_chain_mutex is
   * held so that if cache_reap() is invoked it cannot do
   * anything expensive but will only modify reap_work
   * and reschedule the timer.
  */
  cancel_delayed_work_sync(&per_cpu(slab_reap_work, cpu));
  /* Now the cache_reaper is guaranteed to be not running. */
  per_cpu(slab_reap_work, cpu).work.func = NULL;
    break;
   case CPU_DOWN_FAILED:
   case CPU_DOWN_FAILED_FROZEN:
  start_cpu_timer(cpu);
    break;
 case CPU_DEAD:
 case CPU_DEAD_FROZEN:
  /*
   * Even if all the cpus of a node are down, we don't free the
   * kmem_list3 of any cache. This to avoid a race between
   * cpu_down, and a kmalloc allocation from another cpu for
   * memory from the node of the cpu going down.  The list3
   * structure is usually allocated from kmem_cache_create() and
   * gets destroyed at kmem_cache_destroy().
   */
  /* fall through */
#endif
 case CPU_UP_CANCELED:
 case CPU_UP_CANCELED_FROZEN:
  mutex_lock(&cache_chain_mutex);
  cpuup_canceled(cpu);
  mutex_unlock(&cache_chain_mutex);
  break;
 }
 return notifier_from_errno(err);
}

static struct notifier_block __cpuinitdata cpucache_notifier = {
 &cpuup_callback, NULL, 0
};

#if defined(CONFIG_NUMA) && defined(CONFIG_MEMORY_HOTPLUG)
/*
 * Drains freelist for a node on each slab cache, used for memory hot-remove.
 * Returns -EBUSY if all objects cannot be drained so that the node is not
 * removed.
 *
 * Must hold cache_chain_mutex.
 */
static int __meminit drain_cache_nodelists_node(int node)
{
 struct kmem_cache *cachep;
 int ret = 0;

 list_for_each_entry(cachep, &cache_chain, next) {
  struct kmem_list3 *l3;

  l3 = cachep->nodelists[node];
  if (!l3)
   continue;

  drain_freelist(cachep, l3, l3->free_objects);

  if (!list_empty(&l3->slabs_full) ||
      !list_empty(&l3->slabs_partial)) {
   ret = -EBUSY;
   break;
  }
 }
 return ret;
}

static int __meminit slab_memory_callback(struct notifier_block *self,
     unsigned long action, void *arg)
{
 struct memory_notify *mnb = arg;
 int ret = 0;
 int nid;

 nid = mnb->status_change_nid;
 if (nid < 0)
  goto out;

 switch (action) {
 case MEM_GOING_ONLINE:
  mutex_lock(&cache_chain_mutex);
  ret = init_cache_nodelists_node(nid);
  mutex_unlock(&cache_chain_mutex);
  break;
 case MEM_GOING_OFFLINE:
  mutex_lock(&cache_chain_mutex);
  ret = drain_cache_nodelists_node(nid);
  mutex_unlock(&cache_chain_mutex);
  break;
 case MEM_ONLINE:
 case MEM_OFFLINE:
 case MEM_CANCEL_ONLINE:
 case MEM_CANCEL_OFFLINE:
  break;
 }
out:
 return notifier_from_errno(ret);
}
#endif /* CONFIG_NUMA && CONFIG_MEMORY_HOTPLUG */

/*
 * swap the static kmem_list3 with kmalloced memory
 */
static void __init init_list(struct kmem_cache *cachep, struct kmem_list3 *list,
    int nodeid)
{
 struct kmem_list3 *ptr;

 ptr = kmalloc_node(sizeof(struct kmem_list3), GFP_NOWAIT, nodeid);
 BUG_ON(!ptr);

 memcpy(ptr, list, sizeof(struct kmem_list3));
 /*
  * Do not assume that spinlocks can be initialized via memcpy:
  */
 spin_lock_init(&ptr->list_lock);

 MAKE_ALL_LISTS(cachep, ptr, nodeid);
 cachep->nodelists[nodeid] = ptr;
}

/*
 * For setting up all the kmem_list3s for cache whose buffer_size is same as
 * size of kmem_list3.
 */
static void __init set_up_list3s(struct kmem_cache *cachep, int index)
{
 int node;

 for_each_online_node(node) {
  cachep->nodelists[node] = &initkmem_list3[index + node];
  cachep->nodelists[node]->next_reap = jiffies +
      REAPTIMEOUT_LIST3 +
      ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
 }
}

/*
 * Initialisation.  Called after the page allocator have been initialised and
 * before smp_init().
 */
void __init kmem_cache_init(void)
{
 size_t left_over;
 struct cache_sizes *sizes;
 struct cache_names *names;
 int i;
 int order;
 int node;

 if (num_possible_nodes() == 1)
  use_alien_caches = 0;

 for (i = 0; i < NUM_INIT_LISTS; i++) {
  kmem_list3_init(&initkmem_list3[i]);
  if (i < MAX_NUMNODES)
   cache_cache.nodelists[i] = NULL;
 }
 set_up_list3s(&cache_cache, CACHE_CACHE);

 /*
  * Fragmentation resistance on low memory - only use bigger
  * page orders on machines with more than 32MB of memory.
  */
 if (totalram_pages > (32 << 20) >> PAGE_SHIFT)
  slab_break_gfp_order = BREAK_GFP_ORDER_HI;

 /* Bootstrap is tricky, because several objects are allocated
  * from caches that do not exist yet:
  * 1) initialize the cache_cache cache: it contains the struct
  *    kmem_cache structures of all caches, except cache_cache itself:
  *    cache_cache is statically allocated.
  *    Initially an __init data area is used for the head array and the
  *    kmem_list3 structures, it's replaced with a kmalloc allocated
  *    array at the end of the bootstrap.
  * 2) Create the first kmalloc cache.
  *    The struct kmem_cache for the new cache is allocated normally.
  *    An __init data area is used for the head array.
  * 3) Create the remaining kmalloc caches, with minimally sized
  *    head arrays.
  * 4) Replace the __init data head arrays for cache_cache and the first
  *    kmalloc cache with kmalloc allocated arrays.
  * 5) Replace the __init data for kmem_list3 for cache_cache and
  *    the other cache's with kmalloc allocated memory.
  * 6) Resize the head arrays of the kmalloc caches to their final sizes.
  */

 node = numa_mem_id();

 /* 1) create the cache_cache */
 INIT_LIST_HEAD(&cache_chain);
 list_add(&cache_cache.next, &cache_chain);
 cache_cache.colour_off = cache_line_size();
 cache_cache.array[smp_processor_id()] = &initarray_cache.cache;
 cache_cache.nodelists[node] = &initkmem_list3[CACHE_CACHE + node];

 /*
  * struct kmem_cache size depends on nr_node_ids & nr_cpu_ids
  */
 cache_cache.buffer_size = offsetof(struct kmem_cache, array[nr_cpu_ids]) +
      nr_node_ids * sizeof(struct kmem_list3 *);
#if DEBUG
 cache_cache.obj_size = cache_cache.buffer_size;
#endif
 cache_cache.buffer_size = ALIGN(cache_cache.buffer_size,
     cache_line_size());
 cache_cache.reciprocal_buffer_size =
  reciprocal_value(cache_cache.buffer_size);

 for (order = 0; order < MAX_ORDER; order++) {
  cache_estimate(order, cache_cache.buffer_size,
   cache_line_size(), 0, &left_over, &cache_cache.num);
  if (cache_cache.num)
   break;
 }
 BUG_ON(!cache_cache.num);
 cache_cache.gfporder = order;
 cache_cache.colour = left_over / cache_cache.colour_off;
 cache_cache.slab_size = ALIGN(cache_cache.num * sizeof(kmem_bufctl_t) +
          sizeof(struct slab), cache_line_size());

 /* 2+3) create the kmalloc caches */
 sizes = malloc_sizes;
 names = cache_names;

 /*
  * Initialize the caches that provide memory for the array cache and the
  * kmem_list3 structures first.  Without this, further allocations will
  * bug.
  */

 sizes[INDEX_AC].cs_cachep = kmem_cache_create(names[INDEX_AC].name,
     sizes[INDEX_AC].cs_size,
     ARCH_KMALLOC_MINALIGN,
     ARCH_KMALLOC_FLAGS|SLAB_PANIC,
     NULL);

 if (INDEX_AC != INDEX_L3) {
  sizes[INDEX_L3].cs_cachep =
   kmem_cache_create(names[INDEX_L3].name,
    sizes[INDEX_L3].cs_size,
    ARCH_KMALLOC_MINALIGN,
    ARCH_KMALLOC_FLAGS|SLAB_PANIC,
    NULL);
 }

 slab_early_init = 0;

 while (sizes->cs_size != ULONG_MAX) {
  /*
   * For performance, all the general caches are L1 aligned.
   * This should be particularly beneficial on SMP boxes, as it
   * eliminates "false sharing".
   * 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_create(names->name,
     sizes->cs_size,
     ARCH_KMALLOC_MINALIGN,
     ARCH_KMALLOC_FLAGS|SLAB_PANIC,
     NULL);
  }
#ifdef CONFIG_ZONE_DMA
  sizes->cs_dmacachep = kmem_cache_create(
     names->name_dma,
     sizes->cs_size,
     ARCH_KMALLOC_MINALIGN,
     ARCH_KMALLOC_FLAGS|SLAB_CACHE_DMA|
      SLAB_PANIC,
     NULL);
#endif
  sizes++;
  names++;
 }
 /* 4) Replace the bootstrap head arrays */
 {
  struct array_cache *ptr;

  ptr = kmalloc(sizeof(struct arraycache_init), GFP_NOWAIT);

  BUG_ON(cpu_cache_get(&cache_cache) != &initarray_cache.cache);
  memcpy(ptr, cpu_cache_get(&cache_cache),
         sizeof(struct arraycache_init));
  /*
   * Do not assume that spinlocks can be initialized via memcpy:
   */
  spin_lock_init(&ptr->lock);

  cache_cache.array[smp_processor_id()] = ptr;

  ptr = kmalloc(sizeof(struct arraycache_init), GFP_NOWAIT);

  BUG_ON(cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep)
         != &initarray_generic.cache);
  memcpy(ptr, cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep),
         sizeof(struct arraycache_init));
  /*
   * Do not assume that spinlocks can be initialized via memcpy:
   */
  spin_lock_init(&ptr->lock);

  malloc_sizes[INDEX_AC].cs_cachep->array[smp_processor_id()] =
      ptr;
 }
 /* 5) Replace the bootstrap kmem_list3's */
 {
  int nid;

  for_each_online_node(nid) {
   init_list(&cache_cache, &initkmem_list3[CACHE_CACHE + nid], nid);

   init_list(malloc_sizes[INDEX_AC].cs_cachep,
      &initkmem_list3[SIZE_AC + nid], nid);

   if (INDEX_AC != INDEX_L3) {
    init_list(malloc_sizes[INDEX_L3].cs_cachep,
       &initkmem_list3[SIZE_L3 + nid], nid);
   }
  }
 }

 g_cpucache_up = EARLY;
}

void __init kmem_cache_init_late(void)
{
 struct kmem_cache *cachep;

 g_cpucache_up = LATE;

 /* Annotate slab for lockdep -- annotate the malloc caches */
 init_lock_keys();

 /* 6) resize the head arrays to their final sizes */
 mutex_lock(&cache_chain_mutex);
 list_for_each_entry(cachep, &cache_chain, next)
  if (enable_cpucache(cachep, GFP_NOWAIT))
   BUG();
 mutex_unlock(&cache_chain_mutex);

 /* Done! */
 g_cpucache_up = FULL;

 /*
  * Register a cpu startup notifier callback that initializes
  * cpu_cache_get for all new cpus
  */
 register_cpu_notifier(&cpucache_notifier);

#ifdef CONFIG_NUMA
 /*
  * Register a memory hotplug callback that initializes and frees
  * nodelists.
  */
 hotplug_memory_notifier(slab_memory_callback, SLAB_CALLBACK_PRI);
#endif

 /*
  * The reap timers are started later, with a module init call: That part
  * of the kernel is not yet operational.
  */
}

static int __init cpucache_init(void)
{
 int cpu;

 /*
  * Register the timers that return unneeded pages to the page allocator
  */
 for_each_online_cpu(cpu)
  start_cpu_timer(cpu);
 return 0;
}
__initcall(cpucache_init);

/*
 * Interface to system's page allocator. No need to hold the cache-lock.
 *
 * If we requested dmaable memory, we will get it. Even if we
 * did not request dmaable memory, we might get it, but that
 * would be relatively rare and ignorable.
 */
static void *kmem_getpages(struct kmem_cache *cachep, gfp_t flags, int nodeid)
{
 struct page *page;
 int nr_pages;
 int i;

#ifndef CONFIG_MMU
 /*
  * Nommu uses slab's for process anonymous memory allocations, and thus
  * requires __GFP_COMP to properly refcount higher order allocations
  */
 flags |= __GFP_COMP;
#endif

 flags |= cachep->gfpflags;
 if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
  flags |= __GFP_RECLAIMABLE;

 page = alloc_pages_exact_node(nodeid, flags | __GFP_NOTRACK, cachep->gfporder);
 if (!page)
  return NULL;

 nr_pages = (1 << cachep->gfporder);
 if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
  add_zone_page_state(page_zone(page),
   NR_SLAB_RECLAIMABLE, nr_pages);
 else
  add_zone_page_state(page_zone(page),
   NR_SLAB_UNRECLAIMABLE, nr_pages);
 for (i = 0; i < nr_pages; i++)
  __SetPageSlab(page + i);

 if (kmemcheck_enabled && !(cachep->flags & SLAB_NOTRACK)) {
  kmemcheck_alloc_shadow(page, cachep->gfporder, flags, nodeid);

  if (cachep->ctor)
   kmemcheck_mark_uninitialized_pages(page, nr_pages);
  else
   kmemcheck_mark_unallocated_pages(page, nr_pages);
 }

 return page_address(page);
}

/*
 * Interface to system's page release.
 */
static void kmem_freepages(struct kmem_cache *cachep, void *addr)
{
 unsigned long i = (1 << cachep->gfporder);
 struct page *page = virt_to_page(addr);
 const unsigned long nr_freed = i;

 kmemcheck_free_shadow(page, cachep->gfporder);

 if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
  sub_zone_page_state(page_zone(page),
    NR_SLAB_RECLAIMABLE, nr_freed);
 else
  sub_zone_page_state(page_zone(page),
    NR_SLAB_UNRECLAIMABLE, nr_freed);
 while (i--) {
  BUG_ON(!PageSlab(page));
  __ClearPageSlab(page);
  page++;
 }
 if (current->reclaim_state)
  current->reclaim_state->reclaimed_slab += nr_freed;
 free_pages((unsigned long)addr, cachep->gfporder);
}

static void kmem_rcu_free(struct rcu_head *head)
{
 struct slab_rcu *slab_rcu = (struct slab_rcu *)head;
 struct kmem_cache *cachep = slab_rcu->cachep;

 kmem_freepages(cachep, slab_rcu->addr);
 if (OFF_SLAB(cachep))
  kmem_cache_free(cachep->slabp_cache, slab_rcu);
}

#if DEBUG

#ifdef CONFIG_DEBUG_PAGEALLOC
static void store_stackinfo(struct kmem_cache *cachep, unsigned long *addr,
       unsigned long caller)
{
 int size = obj_size(cachep);

 addr = (unsigned long *)&((char *)addr)[obj_offset(cachep)];

 if (size < 5 * sizeof(unsigned long))
  return;

 *addr++ = 0x12345678;
 *addr++ = caller;
 *addr++ = smp_processor_id();
 size -= 3 * sizeof(unsigned long);
 {
  unsigned long *sptr = &caller;
  unsigned long svalue;

  while (!kstack_end(sptr)) {
   svalue = *sptr++;
   if (kernel_text_address(svalue)) {
    *addr++ = svalue;
    size -= sizeof(unsigned long);
    if (size <= sizeof(unsigned long))
     break;
   }
  }

 }
 *addr++ = 0x87654321;
}
#endif

static void poison_obj(struct kmem_cache *cachep, void *addr, unsigned char val)
{
 int size = obj_size(cachep);
 addr = &((char *)addr)[obj_offset(cachep)];

 memset(addr, val, size);
 *(unsigned char *)(addr + size - 1) = POISON_END;
}

static void dump_line(char *data, int offset, int limit)
{
 int i;
 unsigned char error = 0;
 int bad_count = 0;

 printk(KERN_ERR "%03x: ", offset);
 for (i = 0; i < limit; i++) {
  if (data[offset + i] != POISON_FREE) {
   error = data[offset + i];
   bad_count++;
  }
 }
 print_hex_dump(KERN_CONT, "", 0, 16, 1,
   &data[offset], limit, 1);

 if (bad_count == 1) {
  error ^= POISON_FREE;
  if (!(error & (error - 1))) {
   printk(KERN_ERR "Single bit error detected. Probably "
     "bad RAM.\n");
#ifdef CONFIG_X86
   printk(KERN_ERR "Run memtest86+ or a similar memory "
     "test tool.\n");
#else
   printk(KERN_ERR "Run a memory test tool.\n");
#endif
  }
 }
}
#endif

#if DEBUG

static void print_objinfo(struct kmem_cache *cachep, void *objp, int lines)
{
 int i, size;
 char *realobj;

 if (cachep->flags & SLAB_RED_ZONE) {
  printk(KERN_ERR "Redzone: 0x%llx/0x%llx.\n",
   *dbg_redzone1(cachep, objp),
   *dbg_redzone2(cachep, objp));
 }

 if (cachep->flags & SLAB_STORE_USER) {
  printk(KERN_ERR "Last user: [<%p>]",
   *dbg_userword(cachep, objp));
  print_symbol("(%s)",
    (unsigned long)*dbg_userword(cachep, objp));
  printk("\n");
 }
 realobj = (char *)objp + obj_offset(cachep);
 size = obj_size(cachep);
 for (i = 0; i < size && lines; i += 16, lines--) {
  int limit;
  limit = 16;
  if (i + limit > size)
   limit = size - i;
  dump_line(realobj, i, limit);
 }
}

static void check_poison_obj(struct kmem_cache *cachep, void *objp)
{
 char *realobj;
 int size, i;
 int lines = 0;

 realobj = (char *)objp + obj_offset(cachep);
 size = obj_size(cachep);

 for (i = 0; i < size; i++) {
  char exp = POISON_FREE;
  if (i == size - 1)
   exp = POISON_END;
  if (realobj[i] != exp) {
   int limit;
   /* Mismatch ! */
   /* Print header */
   if (lines == 0) {
    printk(KERN_ERR
     "Slab corruption: %s start=%p, len=%d\n",
     cachep->name, realobj, size);
    print_objinfo(cachep, objp, 0);
   }
   /* Hexdump the affected line */
   i = (i / 16) * 16;
   limit = 16;
   if (i + limit > size)
    limit = size - i;
   dump_line(realobj, i, limit);
   i += 16;
   lines++;
   /* Limit to 5 lines */
   if (lines > 5)
    break;
  }
 }
 if (lines != 0) {
  /* Print some data about the neighboring objects, if they
   * exist:
   */
  struct slab *slabp = virt_to_slab(objp);
  unsigned int objnr;

  objnr = obj_to_index(cachep, slabp, objp);
  if (objnr) {
   objp = index_to_obj(cachep, slabp, objnr - 1);
   realobj = (char *)objp + obj_offset(cachep);
   printk(KERN_ERR "Prev obj: start=%p, len=%d\n",
          realobj, size);
   print_objinfo(cachep, objp, 2);
  }
  if (objnr + 1 < cachep->num) {
   objp = index_to_obj(cachep, slabp, objnr + 1);
   realobj = (char *)objp + obj_offset(cachep);
   printk(KERN_ERR "Next obj: start=%p, len=%d\n",
          realobj, size);
   print_objinfo(cachep, objp, 2);
  }
 }
}
#endif

#if DEBUG
static void slab_destroy_debugcheck(struct kmem_cache *cachep, struct slab *slabp)
{
 int i;
 for (i = 0; i < cachep->num; i++) {
  void *objp = index_to_obj(cachep, slabp, i);

  if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
   if (cachep->buffer_size % PAGE_SIZE == 0 &&
     OFF_SLAB(cachep))
    kernel_map_pages(virt_to_page(objp),
     cachep->buffer_size / PAGE_SIZE, 1);
   else
    check_poison_obj(cachep, objp);
#else
   check_poison_obj(cachep, objp);
#endif
  }
  if (cachep->flags & SLAB_RED_ZONE) {
   if (*dbg_redzone1(cachep, objp) != RED_INACTIVE)
    slab_error(cachep, "start of a freed object "
        "was overwritten");
   if (*dbg_redzone2(cachep, objp) != RED_INACTIVE)
    slab_error(cachep, "end of a freed object "
        "was overwritten");
  }
 }
}
#else
static void slab_destroy_debugcheck(struct kmem_cache *cachep, struct slab *slabp)
{
}
#endif

/**
 * slab_destroy - destroy and release all objects in a slab
 * @cachep: cache pointer being destroyed
 * @slabp: slab pointer being destroyed
 *
 * Destroy all the objs in a slab, and release the mem back to the system.
 * Before calling the slab must have been unlinked from the cache.  The
 * cache-lock is not held/needed.
 */
static void slab_destroy(struct kmem_cache *cachep, struct slab *slabp)
{
 void *addr = slabp->s_mem - slabp->colouroff;

 slab_destroy_debugcheck(cachep, slabp);
 if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU)) {
  struct slab_rcu *slab_rcu;

  slab_rcu = (struct slab_rcu *)slabp;
  slab_rcu->cachep = cachep;
  slab_rcu->addr = addr;
  call_rcu(&slab_rcu->head, kmem_rcu_free);
 } else {
  kmem_freepages(cachep, addr);
  if (OFF_SLAB(cachep))
   kmem_cache_free(cachep->slabp_cache, slabp);
 }
}

static void __kmem_cache_destroy(struct kmem_cache *cachep)
{
 int i;
 struct kmem_list3 *l3;

 for_each_online_cpu(i)
     kfree(cachep->array[i]);

 /* NUMA: free the list3 structures */
 for_each_online_node(i) {
  l3 = cachep->nodelists[i];
  if (l3) {
   kfree(l3->shared);
   free_alien_cache(l3->alien);
   kfree(l3);
  }
 }
 kmem_cache_free(&cache_cache, cachep);
}

/**
 * calculate_slab_order - calculate size (page order) of slabs
 * @cachep: pointer to the cache that is being created
 * @size: size of objects to be created in this cache.
 * @align: required alignment for the objects.
 * @flags: slab allocation flags
 *
 * Also calculates the number of objects per slab.
 *
 * This could be made much more intelligent.  For now, try to avoid using
 * high order pages for slabs.  When the gfp() functions are more friendly
 * towards high-order requests, this should be changed.
 */
static size_t calculate_slab_order(struct kmem_cache *cachep,
   size_t size, size_t align, unsigned long flags)
{
 unsigned long offslab_limit;
 size_t left_over = 0;
 int gfporder;

 for (gfporder = 0; gfporder <= KMALLOC_MAX_ORDER; gfporder++) {
  unsigned int num;
  size_t remainder;

  cache_estimate(gfporder, size, align, flags, &remainder, &num);
  if (!num)
   continue;

  if (flags & CFLGS_OFF_SLAB) {
   /*
    * Max number of objs-per-slab for caches which
    * use off-slab slabs. Needed to avoid a possible
    * looping condition in cache_grow().
    */
   offslab_limit = size - sizeof(struct slab);
   offslab_limit /= sizeof(kmem_bufctl_t);

    if (num > offslab_limit)
    break;
  }

  /* Found something acceptable - save it away */
  cachep->num = num;
  cachep->gfporder = gfporder;
  left_over = remainder;

  /*
   * A VFS-reclaimable slab tends to have most allocations
   * as GFP_NOFS and we really don't want to have to be allocating
   * higher-order pages when we are unable to shrink dcache.
   */
  if (flags & SLAB_RECLAIM_ACCOUNT)
   break;

  /*
   * Large number of objects is good, but very large slabs are
   * currently bad for the gfp()s.
   */
  if (gfporder >= slab_break_gfp_order)
   break;

  /*
   * Acceptable internal fragmentation?
   */
  if (left_over * 8 <= (PAGE_SIZE << gfporder))
   break;
 }
 return left_over;
}

static int __init_refok setup_cpu_cache(struct kmem_cache *cachep, gfp_t gfp)
{
 if (g_cpucache_up == FULL)
  return enable_cpucache(cachep, gfp);

 if (g_cpucache_up == NONE) {
  /*
   * Note: the first kmem_cache_create must create the cache
   * that's used by kmalloc(24), otherwise the creation of
   * further caches will BUG().
   */
  cachep->array[smp_processor_id()] = &initarray_generic.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,
   * otherwise the creation of further caches will BUG().
   */
  set_up_list3s(cachep, SIZE_AC);
  if (INDEX_AC == INDEX_L3)
   g_cpucache_up = PARTIAL_L3;
  else
   g_cpucache_up = PARTIAL_AC;
 } else {
  cachep->array[smp_processor_id()] =
   kmalloc(sizeof(struct arraycache_init), gfp);

  if (g_cpucache_up == PARTIAL_AC) {
   set_up_list3s(cachep, SIZE_L3);
   g_cpucache_up = PARTIAL_L3;
  } else {
   int node;
   for_each_online_node(node) {
    cachep->nodelists[node] =
        kmalloc_node(sizeof(struct kmem_list3),
      gfp, node);
    BUG_ON(!cachep->nodelists[node]);
    kmem_list3_init(cachep->nodelists[node]);
   }
  }
 }
 cachep->nodelists[numa_mem_id()]->next_reap =
   jiffies + REAPTIMEOUT_LIST3 +
   ((unsigned long)cachep) % REAPTIMEOUT_LIST3;

 cpu_cache_get(cachep)->avail = 0;
 cpu_cache_get(cachep)->limit = BOOT_CPUCACHE_ENTRIES;
 cpu_cache_get(cachep)->batchcount = 1;
 cpu_cache_get(cachep)->touched = 0;
 cachep->batchcount = 1;
 cachep->limit = BOOT_CPUCACHE_ENTRIES;
 return 0;
}

/**
 * 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.
 * @align: The required alignment for the objects.
 * @flags: SLAB flags
 * @ctor: A constructor for the objects.
 *
 * Returns a ptr to the cache on success, NULL on failure.
 * Cannot be called within a int, but can be interrupted.
 * The @ctor is run when new pages are allocated by the cache.
 *
 * @name must be valid until the cache is destroyed. This implies that
 * the module calling this has to destroy the cache before getting unloaded.
 *
 * The flags are
 *
 * %SLAB_POISON - Poison the slab with a known test pattern (a5a5a5a5)
 * to catch references to uninitialised memory.
 *
 * %SLAB_RED_ZONE - Insert `Red' zones around the allocated memory to check
 * for buffer overruns.
 *
 * %SLAB_HWCACHE_ALIGN - Align the objects in this cache to a hardware
 * cacheline.  This can be beneficial if you're counting cycles as closely
 * as davem.
 */
struct kmem_cache *
kmem_cache_create (const char *name, size_t size, size_t align,
 unsigned long flags, void (*ctor)(void *))
{
 size_t left_over, slab_size, ralign;
 struct kmem_cache *cachep = NULL, *pc;
 gfp_t gfp;

 /*
  * Sanity checks... these are all serious usage bugs.
  */
 if (!name || in_interrupt() || (size < BYTES_PER_WORD) ||
     size > KMALLOC_MAX_SIZE) {
  printk(KERN_ERR "%s: Early error in slab %s\n", __func__,
    name);
  BUG();
 }

 /*
  * We use cache_chain_mutex to ensure a consistent view of
  * cpu_online_mask as well.  Please see cpuup_callback
  */
 if (slab_is_available()) {
  get_online_cpus();
  mutex_lock(&cache_chain_mutex);
 }

 list_for_each_entry(pc, &cache_chain, next) {
  char tmp;
  int res;

  /*
   * This happens when the module gets unloaded and doesn't
   * destroy its slab cache and no-one else reuses the vmalloc
   * area of the module.  Print a warning.
   */
  res = probe_kernel_address(pc->name, tmp);
  if (res) {
   printk(KERN_ERR
          "SLAB: cache with size %d has lost its name\n",
          pc->buffer_size);
   continue;
  }

  if (!strcmp(pc->name, name)) {
   printk(KERN_ERR
          "kmem_cache_create: duplicate cache %s\n", name);
   dump_stack();
   goto oops;
  }
 }

#if DEBUG
 WARN_ON(strchr(name, ' ')); /* It confuses parsers */
#if FORCED_DEBUG
 /*
  * Enable redzoning and last user accounting, except for caches with
  * large objects, if the increased size would increase the object size
  * above the next power of two: caches with object sizes just above a
  * power of two have a significant amount of internal fragmentation.
  */
 if (size < 4096 || fls(size - 1) == fls(size-1 + REDZONE_ALIGN +
      2 * sizeof(unsigned long long)))
  flags |= SLAB_RED_ZONE | SLAB_STORE_USER;
 if (!(flags & SLAB_DESTROY_BY_RCU))
  flags |= SLAB_POISON;
#endif
 if (flags & SLAB_DESTROY_BY_RCU)
  BUG_ON(flags & SLAB_POISON);
#endif
 /*
  * Always checks flags, a caller might be expecting debug support which
  * isn't available.
  */
 BUG_ON(flags & ~CREATE_MASK);

 /*
  * Check that size is in terms of words.  This is needed to avoid
  * unaligned accesses for some archs when redzoning is used, and makes
  * sure any on-slab bufctl's are also correctly aligned.
  */
 if (size & (BYTES_PER_WORD - 1)) {
  size += (BYTES_PER_WORD - 1);
  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
  * alignment if either is greater than BYTES_PER_WORD.
  */
 if (flags & SLAB_STORE_USER)
  ralign = BYTES_PER_WORD;

 if (flags & SLAB_RED_ZONE) {
  ralign = REDZONE_ALIGN;
  /* If redzoning, ensure that the second redzone is suitably
   * aligned, by adjusting the object size accordingly. */
  size += REDZONE_ALIGN - 1;
  size &= ~(REDZONE_ALIGN - 1);
 }

 /* 2) arch mandated alignment */
 if (ralign < ARCH_SLAB_MINALIGN) {
  ralign = ARCH_SLAB_MINALIGN;
 }
 /* 3) caller mandated alignment */
 if (ralign < align) {
  ralign = align;
 }
 /* disable debug if necessary */
 if (ralign > __alignof__(unsigned long long))
  flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
 /*
  * 4) Store it.
  */
 align = ralign;

 if (slab_is_available())
  gfp = GFP_KERNEL;
 else
  gfp = GFP_NOWAIT;

 /* Get cache's description obj. */
 cachep = kmem_cache_zalloc(&cache_cache, gfp);
 if (!cachep)
  goto oops;

 cachep->nodelists = (struct kmem_list3 **)&cachep->array[nr_cpu_ids];
#if DEBUG
 cachep->obj_size = size;

 /*
  * Both debugging options require word-alignment which is calculated
  * into align above.
  */
 if (flags & SLAB_RED_ZONE) {
  /* add space for red zone words */
  cachep->obj_offset += sizeof(unsigned long long);
  size += 2 * sizeof(unsigned long long);
 }
 if (flags & SLAB_STORE_USER) {
  /* user store requires one word storage behind the end of
   * the real object. But if the second red zone needs to be
   * aligned to 64 bits, we must allow that much space.
   */
  if (flags & SLAB_RED_ZONE)
   size += REDZONE_ALIGN;
  else
   size += BYTES_PER_WORD;
 }
#if FORCED_DEBUG && defined(CONFIG_DEBUG_PAGEALLOC)
 if (size >= malloc_sizes[INDEX_L3 + 1].cs_size
     && cachep->obj_size > cache_line_size() && ALIGN(size, align) < PAGE_SIZE) {
  cachep->obj_offset += PAGE_SIZE - ALIGN(size, align);
  size = PAGE_SIZE;
 }
#endif
#endif

 /*
  * Determine if the slab management is 'on' or 'off' slab.
  * (bootstrapping cannot cope with offslab caches so don't do
  * it too early on. Always use on-slab management when
  * SLAB_NOLEAKTRACE to avoid recursive calls into kmemleak)
  */
 if ((size >= (PAGE_SIZE >> 3)) && !slab_early_init &&
     !(flags & SLAB_NOLEAKTRACE))
  /*
   * Size is large, assume best to place the slab management obj
   * off-slab (should allow better packing of objs).
   */
  flags |= CFLGS_OFF_SLAB;

 size = ALIGN(size, align);

 left_over = calculate_slab_order(cachep, size, align, flags);

 if (!cachep->num) {
  printk(KERN_ERR
         "kmem_cache_create: couldn't create cache %s.\n", name);
  kmem_cache_free(&cache_cache, cachep);
  cachep = NULL;
  goto oops;
 }
 slab_size = ALIGN(cachep->num * sizeof(kmem_bufctl_t)
     + sizeof(struct slab), align);

 /*
  * If the slab has been placed off-slab, and we have enough space then
  * move it on-slab. This is at the expense of any extra colouring.
  */
 if (flags & CFLGS_OFF_SLAB && left_over >= slab_size) {
  flags &= ~CFLGS_OFF_SLAB;
  left_over -= slab_size;
 }

 if (flags & CFLGS_OFF_SLAB) {
  /* really off slab. No need for manual alignment */
  slab_size =
      cachep->num * sizeof(kmem_bufctl_t) + sizeof(struct slab);

#ifdef CONFIG_PAGE_POISONING
  /* If we're going to use the generic kernel_map_pages()
   * poisoning, then it's going to smash the contents of
   * the redzone and userword anyhow, so switch them off.
   */
  if (size % PAGE_SIZE == 0 && flags & SLAB_POISON)
   flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
#endif
 }

 cachep->colour_off = cache_line_size();
 /* Offset must be a multiple of the alignment. */
 if (cachep->colour_off < align)
  cachep->colour_off = align;
 cachep->colour = left_over / cachep->colour_off;
 cachep->slab_size = slab_size;
 cachep->flags = flags;
 cachep->gfpflags = 0;
 if (CONFIG_ZONE_DMA_FLAG && (flags & SLAB_CACHE_DMA))
  cachep->gfpflags |= GFP_DMA;
 cachep->buffer_size = size;
 cachep->reciprocal_buffer_size = reciprocal_value(size);

 if (flags & CFLGS_OFF_SLAB) {
  cachep->slabp_cache = kmem_find_general_cachep(slab_size, 0u);
  /*
   * This is a possibility for one of the malloc_sizes caches.
   * But since we go off slab only for object size greater than
   * PAGE_SIZE/8, and malloc_sizes gets created in ascending order,
   * this should not happen at all.
   * But leave a BUG_ON for some lucky dude.
   */
  BUG_ON(ZERO_OR_NULL_PTR(cachep->slabp_cache));
 }
 cachep->ctor = ctor;
 cachep->name = name;

 if (setup_cpu_cache(cachep, gfp)) {
  __kmem_cache_destroy(cachep);
  cachep = NULL;
  goto oops;
 }

 if (flags & SLAB_DEBUG_OBJECTS) {
  /*
   * Would deadlock through slab_destroy()->call_rcu()->
   * debug_object_activate()->kmem_cache_alloc().
   */
  WARN_ON_ONCE(flags & SLAB_DESTROY_BY_RCU);

  slab_set_debugobj_lock_classes(cachep);
 }

 /* cache setup completed, link it into the list */
 list_add(&cachep->next, &cache_chain);
oops:
 if (!cachep && (flags & SLAB_PANIC))
  panic("kmem_cache_create(): failed to create slab `%s'\n",
        name);
 if (slab_is_available()) {
  mutex_unlock(&cache_chain_mutex);
  put_online_cpus();
 }
 return cachep;
}
EXPORT_SYMBOL(kmem_cache_create);

#if DEBUG
static void check_irq_off(void)
{
 BUG_ON(!irqs_disabled());
}

static void check_irq_on(void)
{
 BUG_ON(irqs_disabled());
}

static void check_spinlock_acquired(struct kmem_cache *cachep)
{
#ifdef CONFIG_SMP
 check_irq_off();
 assert_spin_locked(&cachep->nodelists[numa_mem_id()]->list_lock);
#endif
}

static void check_spinlock_acquired_node(struct kmem_cache *cachep, int node)
{
#ifdef CONFIG_SMP
 check_irq_off();
 assert_spin_locked(&cachep->nodelists[node]->list_lock);
#endif
}

#else
#define check_irq_off() do { } while(0)
#define check_irq_on() do { } while(0)
#define check_spinlock_acquired(x) do { } while(0)
#define check_spinlock_acquired_node(x, y) do { } while(0)
#endif

static void drain_array(struct kmem_cache *cachep, struct kmem_list3 *l3,
   struct array_cache *ac,
   int force, int node);

static void do_drain(void *arg)
{
 struct kmem_cache *cachep = arg;
 struct array_cache *ac;
 int node = numa_mem_id();

 check_irq_off();
 ac = cpu_cache_get(cachep);
 spin_lock(&cachep->nodelists[node]->list_lock);
 free_block(cachep, ac->entry, ac->avail, node);
 spin_unlock(&cachep->nodelists[node]->list_lock);
 ac->avail = 0;
}

static void drain_cpu_caches(struct kmem_cache *cachep)
{
 struct kmem_list3 *l3;
 int node;

 on_each_cpu(do_drain, cachep, 1);
 check_irq_on();
 for_each_online_node(node) {
  l3 = cachep->nodelists[node];
  if (l3 && l3->alien)
   drain_alien_cache(cachep, l3->alien);
 }

 for_each_online_node(node) {
  l3 = cachep->nodelists[node];
  if (l3)
   drain_array(cachep, l3, l3->shared, 1, node);
 }
}

/*
 * Remove slabs from the list of free slabs.
 * Specify the number of slabs to drain in tofree.
 *
 * Returns the actual number of slabs released.
 */
static int drain_freelist(struct kmem_cache *cache,
   struct kmem_list3 *l3, int tofree)
{
 struct list_head *p;
 int nr_freed;
 struct slab *slabp;

 nr_freed = 0;
 while (nr_freed < tofree && !list_empty(&l3->slabs_free)) {

  spin_lock_irq(&l3->list_lock);
  p = l3->slabs_free.prev;
  if (p == &l3->slabs_free) {
   spin_unlock_irq(&l3->list_lock);
   goto out;
  }

  slabp = list_entry(p, struct slab, list);
#if DEBUG
  BUG_ON(slabp->inuse);
#endif
  list_del(&slabp->list);
  /*
   * Safe to drop the lock. The slab is no longer linked
   * to the cache.
   */
  l3->free_objects -= cache->num;
  spin_unlock_irq(&l3->list_lock);
  slab_destroy(cache, slabp);
  nr_freed++;
 }
out:
 return nr_freed;
}

/* Called with cache_chain_mutex held to protect against cpu hotplug */
static int __cache_shrink(struct kmem_cache *cachep)
{
 int ret = 0, i = 0;
 struct kmem_list3 *l3;

 drain_cpu_caches(cachep);

 check_irq_on();
 for_each_online_node(i) {
  l3 = cachep->nodelists[i];
  if (!l3)
   continue;

  drain_freelist(cachep, l3, l3->free_objects);

  ret += !list_empty(&l3->slabs_full) ||
   !list_empty(&l3->slabs_partial);
 }
 return (ret ? 1 : 0);
}

/**
 * kmem_cache_shrink - Shrink a cache.
 * @cachep: The cache to shrink.
 *
 * Releases as many slabs as possible for a cache.
 * To help debugging, a zero exit status indicates all slabs were released.
 */
int kmem_cache_shrink(struct kmem_cache *cachep)
{
 int ret;
 BUG_ON(!cachep || in_interrupt());

 get_online_cpus();
 mutex_lock(&cache_chain_mutex);
 ret = __cache_shrink(cachep);
 mutex_unlock(&cache_chain_mutex);
 put_online_cpus();
 return ret;
}
EXPORT_SYMBOL(kmem_cache_shrink);

/**
 * kmem_cache_destroy - delete a cache
 * @cachep: the cache to destroy
 *
 * Remove a &struct kmem_cache object from the slab cache.
 *
 * It is expected this function will be called by a module when it is
 * unloaded.  This will remove the cache completely, and avoid a duplicate
 * cache being allocated each time a module is loaded and unloaded, if the
 * module doesn't have persistent in-kernel storage across loads and unloads.
 *
 * The cache must be empty before calling this function.
 *
 * The caller must guarantee that no one will allocate memory from the cache
 * during the kmem_cache_destroy().
 */
void kmem_cache_destroy(struct kmem_cache *cachep)
{
 BUG_ON(!cachep || in_interrupt());

 /* Find the cache in the chain of caches. */
 get_online_cpus();
 mutex_lock(&cache_chain_mutex);
 /*
  * the chain is never empty, cache_cache is never destroyed
  */
 list_del(&cachep->next);
 if (__cache_shrink(cachep)) {
  slab_error(cachep, "Can't free all objects");
  list_add(&cachep->next, &cache_chain);
  mutex_unlock(&cache_chain_mutex);
  put_online_cpus();
  return;
 }

 if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU))
  rcu_barrier();

 __kmem_cache_destroy(cachep);
 mutex_unlock(&cache_chain_mutex);
 put_online_cpus();
}
EXPORT_SYMBOL(kmem_cache_destroy);

/*
 * Get the memory for a slab management obj.
 * For a slab cache when the slab descriptor is off-slab, slab descriptors
 * always come from malloc_sizes caches.  The slab descriptor cannot
 * come from the same cache which is getting created because,
 * when we are searching for an appropriate cache for these
 * descriptors in kmem_cache_create, we search through the malloc_sizes array.
 * If we are creating a malloc_sizes cache here it would not be visible to
 * kmem_find_general_cachep till the initialization is complete.
 * Hence we cannot have slabp_cache same as the original cache.
 */
static struct slab *alloc_slabmgmt(struct kmem_cache *cachep, void *objp,
       int colour_off, gfp_t local_flags,
       int nodeid)
{
 struct slab *slabp;

 if (OFF_SLAB(cachep)) {
  /* Slab management obj is off-slab. */
  slabp = kmem_cache_alloc_node(cachep->slabp_cache,
           local_flags, nodeid);
  /*
   * If the first object in the slab is leaked (it's allocated
   * but no one has a reference to it), we want to make sure
   * kmemleak does not treat the ->s_mem pointer as a reference
   * to the object. Otherwise we will not report the leak.
   */
  kmemleak_scan_area(&slabp->list, sizeof(struct list_head),
       local_flags);
  if (!slabp)
   return NULL;
 } else {
  slabp = objp + colour_off;
  colour_off += cachep->slab_size;
 }
 slabp->inuse = 0;
 slabp->colouroff = colour_off;
 slabp->s_mem = objp + colour_off;
 slabp->nodeid = nodeid;
 slabp->free = 0;
 return slabp;
}

static inline kmem_bufctl_t *slab_bufctl(struct slab *slabp)
{
 return (kmem_bufctl_t *) (slabp + 1);
}

static void cache_init_objs(struct kmem_cache *cachep,
       struct slab *slabp)
{
 int i;

 for (i = 0; i < cachep->num; i++) {
  void *objp = index_to_obj(cachep, slabp, i);
#if DEBUG
  /* need to poison the objs? */
  if (cachep->flags & SLAB_POISON)
   poison_obj(cachep, objp, POISON_FREE);
  if (cachep->flags & SLAB_STORE_USER)
   *dbg_userword(cachep, objp) = NULL;

  if (cachep->flags & SLAB_RED_ZONE) {
   *dbg_redzone1(cachep, objp) = RED_INACTIVE;
   *dbg_redzone2(cachep, objp) = RED_INACTIVE;
  }
  /*
   * Constructors are not allowed to allocate memory from the same
   * cache which they are a constructor for.  Otherwise, deadlock.
   * They must also be threaded.
   */
  if (cachep->ctor && !(cachep->flags & SLAB_POISON))
   cachep->ctor(objp + obj_offset(cachep));

  if (cachep->flags & SLAB_RED_ZONE) {
   if (*dbg_redzone2(cachep, objp) != RED_INACTIVE)
    slab_error(cachep, "constructor overwrote the"
        " end of an object");
   if (*dbg_redzone1(cachep, objp) != RED_INACTIVE)
    slab_error(cachep, "constructor overwrote the"
        " start of an object");
  }
  if ((cachep->buffer_size % PAGE_SIZE) == 0 &&
       OFF_SLAB(cachep) && cachep->flags & SLAB_POISON)
   kernel_map_pages(virt_to_page(objp),
      cachep->buffer_size / PAGE_SIZE, 0);
#else
  if (cachep->ctor)
   cachep->ctor(objp);
#endif
  slab_bufctl(slabp)[i] = i + 1;
 }
 slab_bufctl(slabp)[i - 1] = BUFCTL_END;
}

static void kmem_flagcheck(struct kmem_cache *cachep, gfp_t flags)
{
 if (CONFIG_ZONE_DMA_FLAG) {
  if (flags & GFP_DMA)
   BUG_ON(!(cachep->gfpflags & GFP_DMA));
  else
   BUG_ON(cachep->gfpflags & GFP_DMA);
 }
}

static void *slab_get_obj(struct kmem_cache *cachep, struct slab *slabp,
    int nodeid)
{
 void *objp = index_to_obj(cachep, slabp, slabp->free);
 kmem_bufctl_t next;

 slabp->inuse++;
 next = slab_bufctl(slabp)[slabp->free];
#if DEBUG
 slab_bufctl(slabp)[slabp->free] = BUFCTL_FREE;
 WARN_ON(slabp->nodeid != nodeid);
#endif
 slabp->free = next;

 return objp;
}

static void slab_put_obj(struct kmem_cache *cachep, struct slab *slabp,
    void *objp, int nodeid)
{
 unsigned int objnr = obj_to_index(cachep, slabp, objp);

#if DEBUG
 /* Verify that the slab belongs to the intended node */
 WARN_ON(slabp->nodeid != nodeid);

 if (slab_bufctl(slabp)[objnr] + 1 <= SLAB_LIMIT + 1) {
  printk(KERN_ERR "slab: double free detected in cache "
    "'%s', objp %p\n", cachep->name, objp);
  BUG();
 }
#endif
 slab_bufctl(slabp)[objnr] = slabp->free;
 slabp->free = objnr;
 slabp->inuse--;
}

/*
 * Map pages beginning at addr to the given cache and slab. This is required
 * for the slab allocator to be able to lookup the cache and slab of a
 * virtual address for kfree, ksize, and slab debugging.
 */
static void slab_map_pages(struct kmem_cache *cache, struct slab *slab,
      void *addr)
{
 int nr_pages;
 struct page *page;

 page = virt_to_page(addr);

 nr_pages = 1;
 if (likely(!PageCompound(page)))
  nr_pages <<= cache->gfporder;

 do {
  page_set_cache(page, cache);
  page_set_slab(page, slab);
  page++;
 } while (--nr_pages);
}

/*
 * Grow (by 1) the number of slabs within a cache.  This is called by
 * kmem_cache_alloc() when there are no active objs left in a cache.
 */
static int cache_grow(struct kmem_cache *cachep,
  gfp_t flags, int nodeid, void *objp)
{
 struct slab *slabp;
 size_t offset;
 gfp_t local_flags;
 struct kmem_list3 *l3;

 /*
  * Be lazy and only check for valid flags here,  keeping it out of the
  * critical path in kmem_cache_alloc().
  */
 BUG_ON(flags & GFP_SLAB_BUG_MASK);
 local_flags = flags & (GFP_CONSTRAINT_MASK|GFP_RECLAIM_MASK);

 /* Take the l3 list lock to change the colour_next on this node */
 check_irq_off();
 l3 = cachep->nodelists[nodeid];
 spin_lock(&l3->list_lock);

 /* Get colour for the slab, and cal the next value. */
 offset = l3->colour_next;
 l3->colour_next++;
 if (l3->colour_next >= cachep->colour)
  l3->colour_next = 0;
 spin_unlock(&l3->list_lock);

 offset *= cachep->colour_off;

 if (local_flags & __GFP_WAIT)
  local_irq_enable();

 /*
  * The test for missing atomic flag is performed here, rather than
  * the more obvious place, simply to reduce the critical path length
  * in kmem_cache_alloc(). If a caller is seriously mis-behaving they
  * will eventually be caught here (where it matters).
  */
 kmem_flagcheck(cachep, flags);

 /*
  * Get mem for the objs.  Attempt to allocate a physical page from
  * 'nodeid'.
  */
 if (!objp)
  objp = kmem_getpages(cachep, local_flags, nodeid);
 if (!objp)
  goto failed;

 /* Get slab management. */
 slabp = alloc_slabmgmt(cachep, objp, offset,
   local_flags & ~GFP_CONSTRAINT_MASK, nodeid);
 if (!slabp)
  goto opps1;

 slab_map_pages(cachep, slabp, objp);

 cache_init_objs(cachep, slabp);

 if (local_flags & __GFP_WAIT)
  local_irq_disable();
 check_irq_off();
 spin_lock(&l3->list_lock);

 /* Make slab active. */
 list_add_tail(&slabp->list, &(l3->slabs_free));
 STATS_INC_GROWN(cachep);
 l3->free_objects += cachep->num;
 spin_unlock(&l3->list_lock);
 return 1;
opps1:
 kmem_freepages(cachep, objp);
failed:
 if (local_flags & __GFP_WAIT)
  local_irq_disable();
 return 0;
}

#if DEBUG

/*
 * Perform extra freeing checks:
 * - detect bad pointers.
 * - POISON/RED_ZONE checking
 */
static void kfree_debugcheck(const void *objp)
{
 if (!virt_addr_valid(objp)) {
  printk(KERN_ERR "kfree_debugcheck: out of range ptr %lxh.\n",
         (unsigned long)objp);
  BUG();
 }
}

static inline void verify_redzone_free(struct kmem_cache *cache, void *obj)
{
 unsigned long long redzone1, redzone2;

 redzone1 = *dbg_redzone1(cache, obj);
 redzone2 = *dbg_redzone2(cache, obj);

 /*
  * Redzone is ok.
  */
 if (redzone1 == RED_ACTIVE && redzone2 == RED_ACTIVE)
  return;

 if (redzone1 == RED_INACTIVE && redzone2 == RED_INACTIVE)
  slab_error(cache, "double free detected");
 else
  slab_error(cache, "memory outside object was overwritten");

 printk(KERN_ERR "%p: redzone 1:0x%llx, redzone 2:0x%llx.\n",
   obj, redzone1, redzone2);
}

static void *cache_free_debugcheck(struct kmem_cache *cachep, void *objp,
       void *caller)
{
 struct page *page;
 unsigned int objnr;
 struct slab *slabp;

 BUG_ON(virt_to_cache(objp) != cachep);

 objp -= obj_offset(cachep);
 kfree_debugcheck(objp);
 page = virt_to_head_page(objp);

 slabp = page_get_slab(page);

 if (cachep->flags & SLAB_RED_ZONE) {
  verify_redzone_free(cachep, objp);
  *dbg_redzone1(cachep, objp) = RED_INACTIVE;
  *dbg_redzone2(cachep, objp) = RED_INACTIVE;
 }
 if (cachep->flags & SLAB_STORE_USER)
  *dbg_userword(cachep, objp) = caller;

 objnr = obj_to_index(cachep, slabp, objp);

 BUG_ON(objnr >= cachep->num);
 BUG_ON(objp != index_to_obj(cachep, slabp, objnr));

#ifdef CONFIG_DEBUG_SLAB_LEAK
 slab_bufctl(slabp)[objnr] = BUFCTL_FREE;
#endif
 if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
  if ((cachep->buffer_size % PAGE_SIZE)==0 && OFF_SLAB(cachep)) {
   store_stackinfo(cachep, objp, (unsigned long)caller);
   kernel_map_pages(virt_to_page(objp),
      cachep->buffer_size / PAGE_SIZE, 0);
  } else {
   poison_obj(cachep, objp, POISON_FREE);
  }
#else
  poison_obj(cachep, objp, POISON_FREE);
#endif
 }
 return objp;
}

static void check_slabp(struct kmem_cache *cachep, struct slab *slabp)
{
 kmem_bufctl_t i;
 int entries = 0;

 /* Check slab's freelist to see if this obj is there. */
 for (i = slabp->free; i != BUFCTL_END; i = slab_bufctl(slabp)[i]) {
  entries++;
  if (entries > cachep->num || i >= cachep->num)
   goto bad;
 }
 if (entries != cachep->num - slabp->inuse) {
bad:
  printk(KERN_ERR "slab: Internal list corruption detected in "
    "cache '%s'(%d), slabp %p(%d). Hexdump:\n",
   cachep->name, cachep->num, slabp, slabp->inuse);
  print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 16, 1, slabp,
   sizeof(*slabp) + cachep->num * sizeof(kmem_bufctl_t),
   1);
  BUG();
 }
}
#else
#define kfree_debugcheck(x) do { } while(0)
#define cache_free_debugcheck(x,objp,z) (objp)
#define check_slabp(x,y) do { } while(0)
#endif

static void *cache_alloc_refill(struct kmem_cache *cachep, gfp_t flags)
{
 int batchcount;
 struct kmem_list3 *l3;
 struct array_cache *ac;
 int node;

retry:
 check_irq_off();
 node = numa_mem_id();
 ac = cpu_cache_get(cachep);
 batchcount = ac->batchcount;
 if (!ac->touched && batchcount > BATCHREFILL_LIMIT) {
  /*
   * If there was little recent activity on this cache, then
   * perform only a partial refill.  Otherwise we could generate
   * refill bouncing.
   */
  batchcount = BATCHREFILL_LIMIT;
 }
 l3 = cachep->nodelists[node];

 BUG_ON(ac->avail > 0 || !l3);
 spin_lock(&l3->list_lock);

 /* See if we can refill from the shared array */
 if (l3->shared && transfer_objects(ac, l3->shared, batchcount)) {
  l3->shared->touched = 1;
  goto alloc_done;
 }

 while (batchcount > 0) {
  struct list_head *entry;
  struct slab *slabp;
  /* Get slab alloc is to come from. */
  entry = l3->slabs_partial.next;
  if (entry == &l3->slabs_partial) {
   l3->free_touched = 1;
   entry = l3->slabs_free.next;
   if (entry == &l3->slabs_free)
    goto must_grow;
  }

  slabp = list_entry(entry, struct slab, list);
  check_slabp(cachep, slabp);
  check_spinlock_acquired(cachep);

  /*
   * The slab was either on partial or free list so
   * there must be at least one object available for
   * allocation.
   */
  BUG_ON(slabp->inuse >= cachep->num);

  while (slabp->inuse < cachep->num && batchcount--) {
   STATS_INC_ALLOCED(cachep);
   STATS_INC_ACTIVE(cachep);
   STATS_SET_HIGH(cachep);

   ac->entry[ac->avail++] = slab_get_obj(cachep, slabp,
           node);
  }
  check_slabp(cachep, slabp);

  /* move slabp to correct slabp list: */
  list_del(&slabp->list);
  if (slabp->free == BUFCTL_END)
   list_add(&slabp->list, &l3->slabs_full);
  else
   list_add(&slabp->list, &l3->slabs_partial);
 }

must_grow:
 l3->free_objects -= ac->avail;
alloc_done:
 spin_unlock(&l3->list_lock);

 if (unlikely(!ac->avail)) {
  int x;
  x = cache_grow(cachep, flags | GFP_THISNODE, node, NULL);

  /* cache_grow can reenable interrupts, then ac could change. */
  ac = cpu_cache_get(cachep);
  if (!x && ac->avail == 0) /* no objects in sight? abort */
   return NULL;

  if (!ac->avail)  /* objects refilled by interrupt? */
   goto retry;
 }
 ac->touched = 1;
 return ac->entry[--ac->avail];
}

static inline void cache_alloc_debugcheck_before(struct kmem_cache *cachep,
      gfp_t flags)
{
 might_sleep_if(flags & __GFP_WAIT);
#if DEBUG
 kmem_flagcheck(cachep, flags);
#endif
}

#if DEBUG
static void *cache_alloc_debugcheck_after(struct kmem_cache *cachep,
    gfp_t flags, void *objp, void *caller)
{
 if (!objp)
  return objp;
 if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
  if ((cachep->buffer_size % PAGE_SIZE) == 0 && OFF_SLAB(cachep))
   kernel_map_pages(virt_to_page(objp),
      cachep->buffer_size / PAGE_SIZE, 1);
  else
   check_poison_obj(cachep, objp);
#else
  check_poison_obj(cachep, objp);
#endif
  poison_obj(cachep, objp, POISON_INUSE);
 }
 if (cachep->flags & SLAB_STORE_USER)
  *dbg_userword(cachep, objp) = caller;

 if (cachep->flags & SLAB_RED_ZONE) {
  if (*dbg_redzone1(cachep, objp) != RED_INACTIVE ||
    *dbg_redzone2(cachep, objp) != RED_INACTIVE) {
   slab_error(cachep, "double free, or memory outside"
      " object was overwritten");
   printk(KERN_ERR
    "%p: redzone 1:0x%llx, redzone 2:0x%llx\n",
    objp, *dbg_redzone1(cachep, objp),
    *dbg_redzone2(cachep, objp));
  }
  *dbg_redzone1(cachep, objp) = RED_ACTIVE;
  *dbg_redzone2(cachep, objp) = RED_ACTIVE;
 }
#ifdef CONFIG_DEBUG_SLAB_LEAK
 {
  struct slab *slabp;
  unsigned objnr;

  slabp = page_get_slab(virt_to_head_page(objp));
  objnr = (unsigned)(objp - slabp->s_mem) / cachep->buffer_size;
  slab_bufctl(slabp)[objnr] = BUFCTL_ACTIVE;
 }
#endif
 objp += obj_offset(cachep);
 if (cachep->ctor && cachep->flags & SLAB_POISON)
  cachep->ctor(objp);
 if (ARCH_SLAB_MINALIGN &&
     ((unsigned long)objp & (ARCH_SLAB_MINALIGN-1))) {
  printk(KERN_ERR "0x%p: not aligned to ARCH_SLAB_MINALIGN=%d\n",
         objp, (int)ARCH_SLAB_MINALIGN);
 }
 return objp;
}
#else
#define cache_alloc_debugcheck_after(a,b,objp,d) (objp)
#endif

static bool slab_should_failslab(struct kmem_cache *cachep, gfp_t flags)
{
 if (cachep == &cache_cache)
  return false;

 return should_failslab(obj_size(cachep), flags, cachep->flags);
}

static inline void *____cache_alloc(struct kmem_cache *cachep, gfp_t flags)
{
 void *objp;
 struct array_cache *ac;

 check_irq_off();

 ac = cpu_cache_get(cachep);
 if (likely(ac->avail)) {
  STATS_INC_ALLOCHIT(cachep);
  ac->touched = 1;
  objp = ac->entry[--ac->avail];
 } else {
  STATS_INC_ALLOCMISS(cachep);
  objp = cache_alloc_refill(cachep, flags);
  /*
   * the 'ac' may be updated by cache_alloc_refill(),
   * and kmemleak_erase() requires its correct value.
   */
  ac = cpu_cache_get(cachep);
 }
 /*
  * To avoid a false negative, if an object that is in one of the
  * per-CPU caches is leaked, we need to make sure kmemleak doesn't
  * treat the array pointers as a reference to the object.
  */
 if (objp)
  kmemleak_erase(&ac->entry[ac->avail]);
 return objp;
}

#ifdef CONFIG_NUMA
/*
 * Try allocating on another node if PF_SPREAD_SLAB|PF_MEMPOLICY.
 *
 * If we are in_interrupt, then process context, including cpusets and
 * mempolicy, may not apply and should not be used for allocation policy.
 */
static void *alternate_node_alloc(struct kmem_cache *cachep, gfp_t flags)
{
 int nid_alloc, nid_here;

 if (in_interrupt() || (flags & __GFP_THISNODE))
  return NULL;
 nid_alloc = nid_here = numa_mem_id();
 get_mems_allowed();
 if (cpuset_do_slab_mem_spread() && (cachep->flags & SLAB_MEM_SPREAD))
  nid_alloc = cpuset_slab_spread_node();
 else if (current->mempolicy)
  nid_alloc = slab_node(current->mempolicy);
 put_mems_allowed();
 if (nid_alloc != nid_here)
  return ____cache_alloc_node(cachep, flags, nid_alloc);
 return NULL;
}

/*
 * Fallback function if there was no memory available and no objects on a
 * certain node and fall back is permitted. First we scan all the
 * available nodelists for available objects. If that fails then we
 * perform an allocation without specifying a node. This allows the page
 * allocator to do its reclaim / fallback magic. We then insert the
 * slab into the proper nodelist and then allocate from it.
 */
static void *fallback_alloc(struct kmem_cache *cache, gfp_t flags)
{
 struct zonelist *zonelist;
 gfp_t local_flags;
 struct zoneref *z;
 struct zone *zone;
 enum zone_type high_zoneidx = gfp_zone(flags);
 void *obj = NULL;
 int nid;

 if (flags & __GFP_THISNODE)
  return NULL;

 get_mems_allowed();
 zonelist = node_zonelist(slab_node(current->mempolicy), flags);
 local_flags = flags & (GFP_CONSTRAINT_MASK|GFP_RECLAIM_MASK);

retry:
 /*
  * Look through allowed nodes for objects available
  * from existing per node queues.
  */
 for_each_zone_zonelist(zone, z, zonelist, high_zoneidx) {
  nid = zone_to_nid(zone);

  if (cpuset_zone_allowed_hardwall(zone, flags) &&
   cache->nodelists[nid] &&
   cache->nodelists[nid]->free_objects) {
    obj = ____cache_alloc_node(cache,
     flags | GFP_THISNODE, nid);
    if (obj)
     break;
  }
 }

 if (!obj) {
  /*
   * This allocation will be performed within the constraints
   * of the current cpuset / memory policy requirements.
   * We may trigger various forms of reclaim on the allowed
   * set and go into memory reserves if necessary.
   */
  if (local_flags & __GFP_WAIT)
   local_irq_enable();
  kmem_flagcheck(cache, flags);
  obj = kmem_getpages(cache, local_flags, numa_mem_id());
  if (local_flags & __GFP_WAIT)
   local_irq_disable();
  if (obj) {
   /*
    * Insert into the appropriate per node queues
    */
   nid = page_to_nid(virt_to_page(obj));
   if (cache_grow(cache, flags, nid, obj)) {
    obj = ____cache_alloc_node(cache,
     flags | GFP_THISNODE, nid);
    if (!obj)
     /*
      * Another processor may allocate the
      * objects in the slab since we are
      * not holding any locks.
      */
     goto retry;
   } else {
    /* cache_grow already freed obj */
    obj = NULL;
   }
  }
 }
 put_mems_allowed();
 return obj;
}

/*
 * A interface to enable slab creation on nodeid
 */
static void *____cache_alloc_node(struct kmem_cache *cachep, gfp_t flags,
    int nodeid)
{
 struct list_head *entry;
 struct slab *slabp;
 struct kmem_list3 *l3;
 void *obj;
 int x;

 l3 = cachep->nodelists[nodeid];
 BUG_ON(!l3);

retry:
 check_irq_off();
 spin_lock(&l3->list_lock);
 entry = l3->slabs_partial.next;
 if (entry == &l3->slabs_partial) {
  l3->free_touched = 1;
  entry = l3->slabs_free.next;
  if (entry == &l3->slabs_free)
   goto must_grow;
 }

 slabp = list_entry(entry, struct slab, list);
 check_spinlock_acquired_node(cachep, nodeid);
 check_slabp(cachep, slabp);

 STATS_INC_NODEALLOCS(cachep);
 STATS_INC_ACTIVE(cachep);
 STATS_SET_HIGH(cachep);

 BUG_ON(slabp->inuse == cachep->num);

 obj = slab_get_obj(cachep, slabp, nodeid);
 check_slabp(cachep, slabp);
 l3->free_objects--;
 /* move slabp to correct slabp list: */
 list_del(&slabp->list);

 if (slabp->free == BUFCTL_END)
  list_add(&slabp->list, &l3->slabs_full);
 else
  list_add(&slabp->list, &l3->slabs_partial);

 spin_unlock(&l3->list_lock);
 goto done;

must_grow:
 spin_unlock(&l3->list_lock);
 x = cache_grow(cachep, flags | GFP_THISNODE, nodeid, NULL);
 if (x)
  goto retry;

 return fallback_alloc(cachep, flags);

done:
 return obj;
}

/**
 * kmem_cache_alloc_node - Allocate an object on the specified node
 * @cachep: The cache to allocate from.
 * @flags: See kmalloc().
 * @nodeid: node number of the target node.
 * @caller: return address of caller, used for debug information
 *
 * Identical to kmem_cache_alloc but it will allocate memory on the given
 * node, which can improve the performance for cpu bound structures.
 *
 * Fallback to other node is possible if __GFP_THISNODE is not set.
 */
static __always_inline void *
__cache_alloc_node(struct kmem_cache *cachep, gfp_t flags, int nodeid,
     void *caller)
{
 unsigned long save_flags;
 void *ptr;
 int slab_node = numa_mem_id();

 flags &= gfp_allowed_mask;

 lockdep_trace_alloc(flags);

 if (slab_should_failslab(cachep, flags))
  return NULL;

 cache_alloc_debugcheck_before(cachep, flags);
 local_irq_save(save_flags);

 if (nodeid == NUMA_NO_NODE)
  nodeid = slab_node;

 if (unlikely(!cachep->nodelists[nodeid])) {
  /* Node not bootstrapped yet */
  ptr = fallback_alloc(cachep, flags);
  goto out;
 }

 if (nodeid == slab_node) {
  /*
   * Use the locally cached objects if possible.
   * However ____cache_alloc does not allow fallback
   * to other nodes. It may fail while we still have
   * objects on other nodes available.
   */
  ptr = ____cache_alloc(cachep, flags);
  if (ptr)
   goto out;
 }
 /* ___cache_alloc_node can fall back to other nodes */
 ptr = ____cache_alloc_node(cachep, flags, nodeid);
  out:
 local_irq_restore(save_flags);
 ptr = cache_alloc_debugcheck_after(cachep, flags, ptr, caller);
 kmemleak_alloc_recursive(ptr, obj_size(cachep), 1, cachep->flags,
     flags);

 if (likely(ptr))
  kmemcheck_slab_alloc(cachep, flags, ptr, obj_size(cachep));

 if (unlikely((flags & __GFP_ZERO) && ptr))
  memset(ptr, 0, obj_size(cachep));

 return ptr;
}

static __always_inline void *
__do_cache_alloc(struct kmem_cache *cache, gfp_t flags)
{
 void *objp;

 if (unlikely(current->flags & (PF_SPREAD_SLAB | PF_MEMPOLICY))) {
  objp = alternate_node_alloc(cache, flags);
  if (objp)
   goto out;
 }
 objp = ____cache_alloc(cache, flags);

 /*
  * We may just have run out of memory on the local node.
  * ____cache_alloc_node() knows how to locate memory on other nodes
  */
 if (!objp)
  objp = ____cache_alloc_node(cache, flags, numa_mem_id());

  out:
 return objp;
}
#else

static __always_inline void *
__do_cache_alloc(struct kmem_cache *cachep, gfp_t flags)
{
 return ____cache_alloc(cachep, flags);
}

#endif /* CONFIG_NUMA */

static __always_inline void *
__cache_alloc(struct kmem_cache *cachep, gfp_t flags, void *caller)
{
 unsigned long save_flags;
 void *objp;

 flags &= gfp_allowed_mask;

 lockdep_trace_alloc(flags);

 if (slab_should_failslab(cachep, flags))
  return NULL;

 cache_alloc_debugcheck_before(cachep, flags);
 local_irq_save(save_flags);
 objp = __do_cache_alloc(cachep, flags);
 local_irq_restore(save_flags);
 objp = cache_alloc_debugcheck_after(cachep, flags, objp, caller);
 kmemleak_alloc_recursive(objp, obj_size(cachep), 1, cachep->flags,
     flags);
 prefetchw(objp);

 if (likely(objp))
  kmemcheck_slab_alloc(cachep, flags, objp, obj_size(cachep));

 if (unlikely((flags & __GFP_ZERO) && objp))
  memset(objp, 0, obj_size(cachep));

 return objp;
}

/*
 * Caller needs to acquire correct kmem_list's list_lock
 */
static void free_block(struct kmem_cache *cachep, void **objpp, int nr_objects,
         int node)
{
 int i;
 struct kmem_list3 *l3;

 for (i = 0; i < nr_objects; i++) {
  void *objp = objpp[i];
  struct slab *slabp;

  slabp = virt_to_slab(objp);
  l3 = cachep->nodelists[node];
  list_del(&slabp->list);
  check_spinlock_acquired_node(cachep, node);
  check_slabp(cachep, slabp);
  slab_put_obj(cachep, slabp, objp, node);
  STATS_DEC_ACTIVE(cachep);
  l3->free_objects++;
  check_slabp(cachep, slabp);

  /* fixup slab chains */
  if (slabp->inuse == 0) {
   if (l3->free_objects > l3->free_limit) {
    l3->free_objects -= cachep->num;
    /* No need to drop any previously held
     * lock here, even if we have a off-slab slab
     * descriptor it is guaranteed to come from
     * a different cache, refer to comments before
     * alloc_slabmgmt.
     */
    slab_destroy(cachep, slabp);
   } else {
    list_add(&slabp->list, &l3->slabs_free);
   }
  } else {
   /* Unconditionally move a slab to the end of the
    * partial list on free - maximum time for the
    * other objects to be freed, too.
    */
   list_add_tail(&slabp->list, &l3->slabs_partial);
  }
 }
}

static void cache_flusharray(struct kmem_cache *cachep, struct array_cache *ac)
{
 int batchcount;
 struct kmem_list3 *l3;
 int node = numa_mem_id();

 batchcount = ac->batchcount;
#if DEBUG
 BUG_ON(!batchcount || batchcount > ac->avail);
#endif
 check_irq_off();
 l3 = cachep->nodelists[node];
 spin_lock(&l3->list_lock);
 if (l3->shared) {
  struct array_cache *shared_array = l3->shared;
  int max = shared_array->limit - shared_array->avail;
  if (max) {
   if (batchcount > max)
    batchcount = max;
   memcpy(&(shared_array->entry[shared_array->avail]),
          ac->entry, sizeof(void *) * batchcount);
   shared_array->avail += batchcount;
   goto free_done;
  }
 }

 free_block(cachep, ac->entry, batchcount, node);
free_done:
#if STATS
 {
  int i = 0;
  struct list_head *p;

  p = l3->slabs_free.next;
  while (p != &(l3->slabs_free)) {
   struct slab *slabp;

   slabp = list_entry(p, struct slab, list);
   BUG_ON(slabp->inuse);

   i++;
   p = p->next;
  }
  STATS_SET_FREEABLE(cachep, i);
 }
#endif
 spin_unlock(&l3->list_lock);
 ac->avail -= batchcount;
 memmove(ac->entry, &(ac->entry[batchcount]), sizeof(void *)*ac->avail);
}

/*
 * Release an obj back to its cache. If the obj has a constructed state, it must
 * be in this state _before_ it is released.  Called with disabled ints.
 */
static inline void __cache_free(struct kmem_cache *cachep, void *objp,
    void *caller)
{
 struct array_cache *ac = cpu_cache_get(cachep);

 check_irq_off();
 kmemleak_free_recursive(objp, cachep->flags);
 objp = cache_free_debugcheck(cachep, objp, caller);

 kmemcheck_slab_free(cachep, objp, obj_size(cachep));

 /*
  * Skip calling cache_free_alien() when the platform is not numa.
  * This will avoid cache misses that happen while accessing slabp (which
  * is per page memory  reference) to get nodeid. Instead use a global
  * variable to skip the call, which is mostly likely to be present in
  * the cache.
  */
 if (nr_online_nodes > 1 && cache_free_alien(cachep, objp))
  return;

 if (likely(ac->avail < ac->limit)) {
  STATS_INC_FREEHIT(cachep);
  ac->entry[ac->avail++] = objp;
  return;
 } else {
  STATS_INC_FREEMISS(cachep);
  cache_flusharray(cachep, ac);
  ac->entry[ac->avail++] = objp;
 }
}

/**
 * kmem_cache_alloc - Allocate an object
 * @cachep: The cache to allocate from.
 * @flags: See kmalloc().
 *
 * Allocate an object from this cache.  The flags are only relevant
 * if the cache has no available objects.
 */
void *kmem_cache_alloc(struct kmem_cache *cachep, gfp_t flags)
{
 void *ret = __cache_alloc(cachep, flags, __builtin_return_address(0));

 trace_kmem_cache_alloc(_RET_IP_, ret,
          obj_size(cachep), cachep->buffer_size, flags);

 return ret;
}
EXPORT_SYMBOL(kmem_cache_alloc);

#ifdef CONFIG_TRACING
void *
kmem_cache_alloc_trace(size_t size, struct kmem_cache *cachep, gfp_t flags)
{
 void *ret;

 ret = __cache_alloc(cachep, flags, __builtin_return_address(0));

 trace_kmalloc(_RET_IP_, ret,
        size, slab_buffer_size(cachep), flags);
 return ret;
}
EXPORT_SYMBOL(kmem_cache_alloc_trace);
#endif

#ifdef CONFIG_NUMA
void *kmem_cache_alloc_node(struct kmem_cache *cachep, gfp_t flags, int nodeid)
{
 void *ret = __cache_alloc_node(cachep, flags, nodeid,
           __builtin_return_address(0));

 trace_kmem_cache_alloc_node(_RET_IP_, ret,
        obj_size(cachep), cachep->buffer_size,
        flags, nodeid);

 return ret;
}
EXPORT_SYMBOL(kmem_cache_alloc_node);

#ifdef CONFIG_TRACING
void *kmem_cache_alloc_node_trace(size_t size,
      struct kmem_cache *cachep,
      gfp_t flags,
      int nodeid)
{
 void *ret;

 ret = __cache_alloc_node(cachep, flags, nodeid,
      __builtin_return_address(0));
 trace_kmalloc_node(_RET_IP_, ret,
      size, slab_buffer_size(cachep),
      flags, nodeid);
 return ret;
}
EXPORT_SYMBOL(kmem_cache_alloc_node_trace);
#endif

static __always_inline void *
__do_kmalloc_node(size_t size, gfp_t flags, int node, void *caller)
{
 struct kmem_cache *cachep;

 cachep = kmem_find_general_cachep(size, flags);
 if (unlikely(ZERO_OR_NULL_PTR(cachep)))
  return cachep;
 return kmem_cache_alloc_node_trace(size, cachep, flags, node);
}

#if defined(CONFIG_DEBUG_SLAB) || defined(CONFIG_TRACING)
void *__kmalloc_node(size_t size, gfp_t flags, int node)
{
 return __do_kmalloc_node(size, flags, node,
   __builtin_return_address(0));
}
EXPORT_SYMBOL(__kmalloc_node);

void *__kmalloc_node_track_caller(size_t size, gfp_t flags,
  int node, unsigned long caller)
{
 return __do_kmalloc_node(size, flags, node, (void *)caller);
}
EXPORT_SYMBOL(__kmalloc_node_track_caller);
#else
void *__kmalloc_node(size_t size, gfp_t flags, int node)
{
 return __do_kmalloc_node(size, flags, node, NULL);
}
EXPORT_SYMBOL(__kmalloc_node);
#endif /* CONFIG_DEBUG_SLAB || CONFIG_TRACING */
#endif /* CONFIG_NUMA */

/**
 * __do_kmalloc - allocate memory
 * @size: how many bytes of memory are required.
 * @flags: the type of memory to allocate (see kmalloc).
 * @caller: function caller for debug tracking of the caller
 */
static __always_inline void *__do_kmalloc(size_t size, gfp_t flags,
       void *caller)
{
 struct kmem_cache *cachep;
 void *ret;

 /* If you want to save a few bytes .text space: replace
  * __ with kmem_.
  * Then kmalloc uses the uninlined functions instead of the inline
  * functions.
  */
 cachep = __find_general_cachep(size, flags);
 if (unlikely(ZERO_OR_NULL_PTR(cachep)))
  return cachep;
 ret = __cache_alloc(cachep, flags, caller);

 trace_kmalloc((unsigned long) caller, ret,
        size, cachep->buffer_size, flags);

 return ret;
}

#if defined(CONFIG_DEBUG_SLAB) || defined(CONFIG_TRACING)
void *__kmalloc(size_t size, gfp_t flags)
{
 return __do_kmalloc(size, flags, __builtin_return_address(0));
}
EXPORT_SYMBOL(__kmalloc);

void *__kmalloc_track_caller(size_t size, gfp_t flags, unsigned long caller)
{
 return __do_kmalloc(size, flags, (void *)caller);
}
EXPORT_SYMBOL(__kmalloc_track_caller);

#else
void *__kmalloc(size_t size, gfp_t flags)
{
 return __do_kmalloc(size, flags, NULL);
}
EXPORT_SYMBOL(__kmalloc);
#endif

/**
 * kmem_cache_free - Deallocate an object
 * @cachep: The cache the allocation was from.
 * @objp: The previously allocated object.
 *
 * Free an object which was previously allocated from this
 * cache.
 */
void kmem_cache_free(struct kmem_cache *cachep, void *objp)
{
 unsigned long flags;

 local_irq_save(flags);
 debug_check_no_locks_freed(objp, obj_size(cachep));
 if (!(cachep->flags & SLAB_DEBUG_OBJECTS))
  debug_check_no_obj_freed(objp, obj_size(cachep));
 __cache_free(cachep, objp, __builtin_return_address(0));
 local_irq_restore(flags);

 trace_kmem_cache_free(_RET_IP_, objp);
}
EXPORT_SYMBOL(kmem_cache_free);

/**
 * kfree - free previously allocated memory
 * @objp: pointer returned by kmalloc.
 *
 * If @objp is NULL, no operation is performed.
 *
 * Don't free memory not originally allocated by kmalloc()
 * or you will run into trouble.
 */
void kfree(const void *objp)
{
 struct kmem_cache *c;
 unsigned long flags;

 trace_kfree(_RET_IP_, objp);

 if (unlikely(ZERO_OR_NULL_PTR(objp)))
  return;
 local_irq_save(flags);
 kfree_debugcheck(objp);
 c = virt_to_cache(objp);
 debug_check_no_locks_freed(objp, obj_size(c));
 debug_check_no_obj_freed(objp, obj_size(c));
 __cache_free(c, (void *)objp, __builtin_return_address(0));
 local_irq_restore(flags);
}
EXPORT_SYMBOL(kfree);

unsigned int kmem_cache_size(struct kmem_cache *cachep)
{
 return obj_size(cachep);
}
EXPORT_SYMBOL(kmem_cache_size);

/*
 * This initializes kmem_list3 or resizes various caches for all nodes.
 */
static int alloc_kmemlist(struct kmem_cache *cachep, gfp_t gfp)
{
 int node;
 struct kmem_list3 *l3;
 struct array_cache *new_shared;
 struct array_cache **new_alien = NULL;

 for_each_online_node(node) {

                if (use_alien_caches) {
                        new_alien = alloc_alien_cache(node, cachep->limit, gfp);
                        if (!new_alien)
                                goto fail;
                }

  new_shared = NULL;
  if (cachep->shared) {
   new_shared = alloc_arraycache(node,
    cachep->shared*cachep->batchcount,
     0xbaadf00d, gfp);
   if (!new_shared) {
    free_alien_cache(new_alien);
    goto fail;
   }
  }

  l3 = cachep->nodelists[node];
  if (l3) {
   struct array_cache *shared = l3->shared;

   spin_lock_irq(&l3->list_lock);

   if (shared)
    free_block(cachep, shared->entry,
      shared->avail, node);

   l3->shared = new_shared;
   if (!l3->alien) {
    l3->alien = new_alien;
    new_alien = NULL;
   }
   l3->free_limit = (1 + nr_cpus_node(node)) *
     cachep->batchcount + cachep->num;
   spin_unlock_irq(&l3->list_lock);
   kfree(shared);
   free_alien_cache(new_alien);
   continue;
  }
  l3 = kmalloc_node(sizeof(struct kmem_list3), gfp, node);
  if (!l3) {
   free_alien_cache(new_alien);
   kfree(new_shared);
   goto fail;
  }

  kmem_list3_init(l3);
  l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
  l3->shared = new_shared;
  l3->alien = new_alien;
  l3->free_limit = (1 + nr_cpus_node(node)) *
     cachep->batchcount + cachep->num;
  cachep->nodelists[node] = l3;
 }
 return 0;

fail:
 if (!cachep->next.next) {
  /* Cache is not active yet. Roll back what we did */
  node--;
  while (node >= 0) {
   if (cachep->nodelists[node]) {
    l3 = cachep->nodelists[node];

    kfree(l3->shared);
    free_alien_cache(l3->alien);
    kfree(l3);
    cachep->nodelists[node] = NULL;
   }
   node--;
  }
 }
 return -ENOMEM;
}

struct ccupdate_struct {
 struct kmem_cache *cachep;
 struct array_cache *new[0];
};

static void do_ccupdate_local(void *info)
{
 struct ccupdate_struct *new = info;
 struct array_cache *old;

 check_irq_off();
 old = cpu_cache_get(new->cachep);

 new->cachep->array[smp_processor_id()] = new->new[smp_processor_id()];
 new->new[smp_processor_id()] = old;
}

/* Always called with the cache_chain_mutex held */
static int do_tune_cpucache(struct kmem_cache *cachep, int limit,
    int batchcount, int shared, gfp_t gfp)
{
 struct ccupdate_struct *new;
 int i;

 new = kzalloc(sizeof(*new) + nr_cpu_ids * sizeof(struct array_cache *),
        gfp);
 if (!new)
  return -ENOMEM;

 for_each_online_cpu(i) {
  new->new[i] = alloc_arraycache(cpu_to_mem(i), limit,
      batchcount, gfp);
  if (!new->new[i]) {
   for (i--; i >= 0; i--)
    kfree(new->new[i]);
   kfree(new);
   return -ENOMEM;
  }
 }
 new->cachep = cachep;

 on_each_cpu(do_ccupdate_local, (void *)new, 1);

 check_irq_on();
 cachep->batchcount = batchcount;
 cachep->limit = limit;
 cachep->shared = shared;

 for_each_online_cpu(i) {
  struct array_cache *ccold = new->new[i];
  if (!ccold)
   continue;
  spin_lock_irq(&cachep->nodelists[cpu_to_mem(i)]->list_lock);
  free_block(cachep, ccold->entry, ccold->avail, cpu_to_mem(i));
  spin_unlock_irq(&cachep->nodelists[cpu_to_mem(i)]->list_lock);
  kfree(ccold);
 }
 kfree(new);
 return alloc_kmemlist(cachep, gfp);
}

/* Called with cache_chain_mutex held always */
static int enable_cpucache(struct kmem_cache *cachep, gfp_t gfp)
{
 int err;
 int limit, shared;

 /*
  * The head array serves three purposes:
  * - create a LIFO ordering, i.e. return objects that are cache-warm
  * - reduce the number of spinlock operations.
  * - reduce the number of linked list operations on the slab and
  *   bufctl chains: array operations are cheaper.
  * The numbers are guessed, we should auto-tune as described by
  * Bonwick.
  */
 if (cachep->buffer_size > 131072)
  limit = 1;
 else if (cachep->buffer_size > PAGE_SIZE)
  limit = 8;
 else if (cachep->buffer_size > 1024)
  limit = 24;
 else if (cachep->buffer_size > 256)
  limit = 54;
 else
  limit = 120;

 /*
  * CPU bound tasks (e.g. network routing) can exhibit cpu bound
  * allocation behaviour: Most allocs on one cpu, most free operations
  * on another cpu. For these cases, an efficient object passing between
  * cpus is necessary. This is provided by a shared array. The array
  * replaces Bonwick's magazine layer.
  * On uniprocessor, it's functionally equivalent (but less efficient)
  * to a larger limit. Thus disabled by default.
  */
 shared = 0;
 if (cachep->buffer_size <= PAGE_SIZE && num_possible_cpus() > 1)
  shared = 8;

#if DEBUG
 /*
  * With debugging enabled, large batchcount lead to excessively long
  * periods with disabled local interrupts. Limit the batchcount
  */
 if (limit > 32)
  limit = 32;
#endif
 err = do_tune_cpucache(cachep, limit, (limit + 1) / 2, shared, gfp);
 if (err)
  printk(KERN_ERR "enable_cpucache failed for %s, error %d.\n",
         cachep->name, -err);
 return err;
}

/*
 * Drain an array if it contains any elements taking the l3 lock only if
 * necessary. Note that the l3 listlock also protects the array_cache
 * if drain_array() is used on the shared array.
 */
static void drain_array(struct kmem_cache *cachep, struct kmem_list3 *l3,
    struct array_cache *ac, int force, int node)
{
 int tofree;

 if (!ac || !ac->avail)
  return;
 if (ac->touched && !force) {
  ac->touched = 0;
 } else {
  spin_lock_irq(&l3->list_lock);
  if (ac->avail) {
   tofree = force ? ac->avail : (ac->limit + 4) / 5;
   if (tofree > ac->avail)
    tofree = (ac->avail + 1) / 2;
   free_block(cachep, ac->entry, tofree, node);
   ac->avail -= tofree;
   memmove(ac->entry, &(ac->entry[tofree]),
    sizeof(void *) * ac->avail);
  }
  spin_unlock_irq(&l3->list_lock);
 }
}

/**
 * cache_reap - Reclaim memory from caches.
 * @w: work descriptor
 *
 * Called from workqueue/eventd every few seconds.
 * Purpose:
 * - clear the per-cpu caches for this CPU.
 * - return freeable pages to the main free memory pool.
 *
 * If we cannot acquire the cache chain mutex then just give up - we'll try
 * again on the next iteration.
 */
static void cache_reap(struct work_struct *w)
{
 struct kmem_cache *searchp;
 struct kmem_list3 *l3;
 int node = numa_mem_id();
 struct delayed_work *work = to_delayed_work(w);

 if (!mutex_trylock(&cache_chain_mutex))
  /* Give up. Setup the next iteration. */
  goto out;

 list_for_each_entry(searchp, &cache_chain, next) {
  check_irq_on();

  /*
   * We only take the l3 lock if absolutely necessary and we
   * have established with reasonable certainty that
   * we can do some work if the lock was obtained.
   */
  l3 = searchp->nodelists[node];

  reap_alien(searchp, l3);

  drain_array(searchp, l3, cpu_cache_get(searchp), 0, node);

  /*
   * These are racy checks but it does not matter
   * if we skip one check or scan twice.
   */
  if (time_after(l3->next_reap, jiffies))
   goto next;

  l3->next_reap = jiffies + REAPTIMEOUT_LIST3;

  drain_array(searchp, l3, l3->shared, 0, node);

  if (l3->free_touched)
   l3->free_touched = 0;
  else {
   int freed;

   freed = drain_freelist(searchp, l3, (l3->free_limit +
    5 * searchp->num - 1) / (5 * searchp->num));
   STATS_ADD_REAPED(searchp, freed);
  }
next:
  cond_resched();
 }
 check_irq_on();
 mutex_unlock(&cache_chain_mutex);
 next_reap_node();
out:
 /* Set up the next iteration */
 schedule_delayed_work(work, round_jiffies_relative(REAPTIMEOUT_CPUC));
}

#ifdef CONFIG_SLABINFO

static void print_slabinfo_header(struct seq_file *m)
{
 /*
  * Output format version, so at least we can change it
  * without _too_ many complaints.
  */
#if STATS
 seq_puts(m, "slabinfo - version: 2.1 (statistics)\n");
#else
 seq_puts(m, "slabinfo - version: 2.1\n");
#endif
 seq_puts(m, "# name            <active_objs> <num_objs> <objsize> "
   "<objperslab> <pagesperslab>");
 seq_puts(m, " : tunables <limit> <batchcount> <sharedfactor>");
 seq_puts(m, " : slabdata <active_slabs> <num_slabs> <sharedavail>");
#if STATS
 seq_puts(m, " : globalstat <listallocs> <maxobjs> <grown> <reaped> "
   "<error> <maxfreeable> <nodeallocs> <remotefrees> <alienoverflow>");
 seq_puts(m, " : cpustat <allochit> <allocmiss> <freehit> <freemiss>");
#endif
 seq_putc(m, '\n');
}

static void *s_start(struct seq_file *m, loff_t *pos)
{
 loff_t n = *pos;

 mutex_lock(&cache_chain_mutex);
 if (!n)
  print_slabinfo_header(m);

 return seq_list_start(&cache_chain, *pos);
}

static void *s_next(struct seq_file *m, void *p, loff_t *pos)
{
 return seq_list_next(p, &cache_chain, pos);
}

static void s_stop(struct seq_file *m, void *p)
{
 mutex_unlock(&cache_chain_mutex);
}

static int s_show(struct seq_file *m, void *p)
{
 struct kmem_cache *cachep = list_entry(p, struct kmem_cache, next);
 struct slab *slabp;
 unsigned long active_objs;
 unsigned long num_objs;
 unsigned long active_slabs = 0;
 unsigned long num_slabs, free_objects = 0, shared_avail = 0;
 const char *name;
 char *error = NULL;
 int node;
 struct kmem_list3 *l3;

 active_objs = 0;
 num_slabs = 0;
 for_each_online_node(node) {
  l3 = cachep->nodelists[node];
  if (!l3)
   continue;

  check_irq_on();
  spin_lock_irq(&l3->list_lock);

  list_for_each_entry(slabp, &l3->slabs_full, list) {
   if (slabp->inuse != cachep->num && !error)
    error = "slabs_full accounting error";
   active_objs += cachep->num;
   active_slabs++;
  }
  list_for_each_entry(slabp, &l3->slabs_partial, list) {
   if (slabp->inuse == cachep->num && !error)
    error = "slabs_partial inuse accounting error";
   if (!slabp->inuse && !error)
    error = "slabs_partial/inuse accounting error";
   active_objs += slabp->inuse;
   active_slabs++;
  }
  list_for_each_entry(slabp, &l3->slabs_free, list) {
   if (slabp->inuse && !error)
    error = "slabs_free/inuse accounting error";
   num_slabs++;
  }
  free_objects += l3->free_objects;
  if (l3->shared)
   shared_avail += l3->shared->avail;

  spin_unlock_irq(&l3->list_lock);
 }
 num_slabs += active_slabs;
 num_objs = num_slabs * cachep->num;
 if (num_objs - active_objs != free_objects && !error)
  error = "free_objects accounting error";

 name = cachep->name;
 if (error)
  printk(KERN_ERR "slab: cache %s error: %s\n", name, error);

 seq_printf(m, "%-17s %6lu %6lu %6u %4u %4d",
     name, active_objs, num_objs, cachep->buffer_size,
     cachep->num, (1 << cachep->gfporder));
 seq_printf(m, " : tunables %4u %4u %4u",
     cachep->limit, cachep->batchcount, cachep->shared);
 seq_printf(m, " : slabdata %6lu %6lu %6lu",
     active_slabs, num_slabs, shared_avail);
#if STATS
 {   /* list3 stats */
  unsigned long high = cachep->high_mark;
  unsigned long allocs = cachep->num_allocations;
  unsigned long grown = cachep->grown;
  unsigned long reaped = cachep->reaped;
  unsigned long errors = cachep->errors;
  unsigned long max_freeable = cachep->max_freeable;
  unsigned long node_allocs = cachep->node_allocs;
  unsigned long node_frees = cachep->node_frees;
  unsigned long overflows = cachep->node_overflow;

  seq_printf(m, " : globalstat %7lu %6lu %5lu %4lu "
      "%4lu %4lu %4lu %4lu %4lu",
      allocs, high, grown,
      reaped, errors, max_freeable, node_allocs,
      node_frees, overflows);
 }
 /* cpu stats */
 {
  unsigned long allochit = atomic_read(&cachep->allochit);
  unsigned long allocmiss = atomic_read(&cachep->allocmiss);
  unsigned long freehit = atomic_read(&cachep->freehit);
  unsigned long freemiss = atomic_read(&cachep->freemiss);

  seq_printf(m, " : cpustat %6lu %6lu %6lu %6lu",
      allochit, allocmiss, freehit, freemiss);
 }
#endif
 seq_putc(m, '\n');
 return 0;
}

/*
 * slabinfo_op - iterator that generates /proc/slabinfo
 *
 * Output layout:
 * cache-name
 * num-active-objs
 * total-objs
 * object size
 * num-active-slabs
 * total-slabs
 * num-pages-per-slab
 * + further values on SMP and with statistics enabled
 */

static const struct seq_operations slabinfo_op = {
 .start = s_start,
 .next = s_next,
 .stop = s_stop,
 .show = s_show,
};

#define MAX_SLABINFO_WRITE 128
/**
 * slabinfo_write - Tuning for the slab allocator
 * @file: unused
 * @buffer: user buffer
 * @count: data length
 * @ppos: unused
 */
static ssize_t slabinfo_write(struct file *file, const char __user *buffer,
         size_t count, loff_t *ppos)
{
 char kbuf[MAX_SLABINFO_WRITE + 1], *tmp;
 int limit, batchcount, shared, res;
 struct kmem_cache *cachep;

 if (count > MAX_SLABINFO_WRITE)
  return -EINVAL;
 if (copy_from_user(&kbuf, buffer, count))
  return -EFAULT;
 kbuf[MAX_SLABINFO_WRITE] = '\0';

 tmp = strchr(kbuf, ' ');
 if (!tmp)
  return -EINVAL;
 *tmp = '\0';
 tmp++;
 if (sscanf(tmp, " %d %d %d", &limit, &batchcount, &shared) != 3)
  return -EINVAL;

 /* Find the cache in the chain of caches. */
 mutex_lock(&cache_chain_mutex);
 res = -EINVAL;
 list_for_each_entry(cachep, &cache_chain, next) {
  if (!strcmp(cachep->name, kbuf)) {
   if (limit < 1 || batchcount < 1 ||
     batchcount > limit || shared < 0) {
    res = 0;
   } else {
    res = do_tune_cpucache(cachep, limit,
             batchcount, shared,
             GFP_KERNEL);
   }
   break;
  }
 }
 mutex_unlock(&cache_chain_mutex);
 if (res >= 0)
  res = count;
 return res;
}

static int slabinfo_open(struct inode *inode, struct file *file)
{
 return seq_open(file, &slabinfo_op);
}

static const struct file_operations proc_slabinfo_operations = {
 .open  = slabinfo_open,
 .read  = seq_read,
 .write  = slabinfo_write,
 .llseek  = seq_lseek,
 .release = seq_release,
};

#ifdef CONFIG_DEBUG_SLAB_LEAK

static void *leaks_start(struct seq_file *m, loff_t *pos)
{
 mutex_lock(&cache_chain_mutex);
 return seq_list_start(&cache_chain, *pos);
}

static inline int add_caller(unsigned long *n, unsigned long v)
{
 unsigned long *p;
 int l;
 if (!v)
  return 1;
 l = n[1];
 p = n + 2;
 while (l) {
  int i = l/2;
  unsigned long *q = p + 2 * i;
  if (*q == v) {
   q[1]++;
   return 1;
  }
  if (*q > v) {
   l = i;
  } else {
   p = q + 2;
   l -= i + 1;
  }
 }
 if (++n[1] == n[0])
  return 0;
 memmove(p + 2, p, n[1] * 2 * sizeof(unsigned long) - ((void *)p - (void *)n));
 p[0] = v;
 p[1] = 1;
 return 1;
}

static void handle_slab(unsigned long *n, struct kmem_cache *c, struct slab *s)
{
 void *p;
 int i;
 if (n[0] == n[1])
  return;
 for (i = 0, p = s->s_mem; i < c->num; i++, p += c->buffer_size) {
  if (slab_bufctl(s)[i] != BUFCTL_ACTIVE)
   continue;
  if (!add_caller(n, (unsigned long)*dbg_userword(c, p)))
   return;
 }
}

static void show_symbol(struct seq_file *m, unsigned long address)
{
#ifdef CONFIG_KALLSYMS
 unsigned long offset, size;
 char modname[MODULE_NAME_LEN], name[KSYM_NAME_LEN];

 if (lookup_symbol_attrs(address, &size, &offset, modname, name) == 0) {
  seq_printf(m, "%s+%#lx/%#lx", name, offset, size);
  if (modname[0])
   seq_printf(m, " [%s]", modname);
  return;
 }
#endif
 seq_printf(m, "%p", (void *)address);
}

static int leaks_show(struct seq_file *m, void *p)
{
 struct kmem_cache *cachep = list_entry(p, struct kmem_cache, next);
 struct slab *slabp;
 struct kmem_list3 *l3;
 const char *name;
 unsigned long *n = m->private;
 int node;
 int i;

 if (!(cachep->flags & SLAB_STORE_USER))
  return 0;
 if (!(cachep->flags & SLAB_RED_ZONE))
  return 0;

 /* OK, we can do it */

 n[1] = 0;

 for_each_online_node(node) {
  l3 = cachep->nodelists[node];
  if (!l3)
   continue;

  check_irq_on();
  spin_lock_irq(&l3->list_lock);

  list_for_each_entry(slabp, &l3->slabs_full, list)
   handle_slab(n, cachep, slabp);
  list_for_each_entry(slabp, &l3->slabs_partial, list)
   handle_slab(n, cachep, slabp);
  spin_unlock_irq(&l3->list_lock);
 }
 name = cachep->name;
 if (n[0] == n[1]) {
  /* Increase the buffer size */
  mutex_unlock(&cache_chain_mutex);
  m->private = kzalloc(n[0] * 4 * sizeof(unsigned long), GFP_KERNEL);
  if (!m->private) {
   /* Too bad, we are really out */
   m->private = n;
   mutex_lock(&cache_chain_mutex);
   return -ENOMEM;
  }
  *(unsigned long *)m->private = n[0] * 2;
  kfree(n);
  mutex_lock(&cache_chain_mutex);
  /* Now make sure this entry will be retried */
  m->count = m->size;
  return 0;
 }
 for (i = 0; i < n[1]; i++) {
  seq_printf(m, "%s: %lu ", name, n[2*i+3]);
  show_symbol(m, n[2*i+2]);
  seq_putc(m, '\n');
 }

 return 0;
}

static const struct seq_operations slabstats_op = {
 .start = leaks_start,
 .next = s_next,
 .stop = s_stop,
 .show = leaks_show,
};

static int slabstats_open(struct inode *inode, struct file *file)
{
 unsigned long *n = kzalloc(PAGE_SIZE, GFP_KERNEL);
 int ret = -ENOMEM;
 if (n) {
  ret = seq_open(file, &slabstats_op);
  if (!ret) {
   struct seq_file *m = file->private_data;
   *n = PAGE_SIZE / (2 * sizeof(unsigned long));
   m->private = n;
   n = NULL;
  }
  kfree(n);
 }
 return ret;
}

static const struct file_operations proc_slabstats_operations = {
 .open  = slabstats_open,
 .read  = seq_read,
 .llseek  = seq_lseek,
 .release = seq_release_private,
};
#endif

static int __init slab_proc_init(void)
{
 proc_create("slabinfo",S_IWUSR|S_IRUSR,NULL,&proc_slabinfo_operations);
#ifdef CONFIG_DEBUG_SLAB_LEAK
 proc_create("slab_allocators", 0, NULL, &proc_slabstats_operations);
#endif
 return 0;
}
module_init(slab_proc_init);
#endif

/**
 * ksize - get the actual amount of memory allocated for a given object
 * @objp: Pointer to the object
 *
 * kmalloc may internally round up allocations and return more memory
 * than requested. ksize() can be used to determine the actual amount of
 * memory allocated. The caller may use this additional memory, even though
 * a smaller amount of memory was initially specified with the kmalloc call.
 * The caller must guarantee that objp points to a valid object previously
 * allocated with either kmalloc() or kmem_cache_alloc(). The object
 * must not be freed during the duration of the call.
 */
size_t ksize(const void *objp)
{
 BUG_ON(!objp);
 if (unlikely(objp == ZERO_SIZE_PTR))
  return 0;

 return obj_size(virt_to_cache(objp));
}
EXPORT_SYMBOL(ksize);