slab.c 115.3 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52
/*
 * linux/mm/slab.c
 * Written by Mark Hemment, 1996/97.
 * (markhe@nextd.demon.co.uk)
 *
 * 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 intializations 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.
 *
A
Andrew Morton 已提交
53
 * The c_cpuarray may not be read with enabled local interrupts -
L
Linus Torvalds 已提交
54 55 56 57
 * it's changed with a smp_call_function().
 *
 * SMP synchronization:
 *  constructors and destructors are called without any locking.
58
 *  Several members in struct kmem_cache and struct slab never change, they
L
Linus Torvalds 已提交
59 60 61 62 63 64 65 66 67 68 69 70
 *	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
I
Ingo Molnar 已提交
71
 *	The global cache-chain is protected by the mutex 'cache_chain_mutex'.
L
Linus Torvalds 已提交
72 73 74 75 76 77
 *	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.
 *
78 79 80 81 82 83 84 85 86
 * 15 March 2005. NUMA slab allocator.
 *	Shai Fultheim <shai@scalex86.org>.
 *	Shobhit Dayal <shobhit@calsoftinc.com>
 *	Alok N Kataria <alokk@calsoftinc.com>
 *	Christoph Lameter <christoph@lameter.com>
 *
 *	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.
L
Linus Torvalds 已提交
87 88 89 90
 */

#include	<linux/slab.h>
#include	<linux/mm.h>
91
#include	<linux/poison.h>
L
Linus Torvalds 已提交
92 93 94 95 96
#include	<linux/swap.h>
#include	<linux/cache.h>
#include	<linux/interrupt.h>
#include	<linux/init.h>
#include	<linux/compiler.h>
97
#include	<linux/cpuset.h>
L
Linus Torvalds 已提交
98 99 100 101 102 103 104
#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>
105
#include	<linux/string.h>
106
#include	<linux/uaccess.h>
107
#include	<linux/nodemask.h>
108
#include	<linux/mempolicy.h>
I
Ingo Molnar 已提交
109
#include	<linux/mutex.h>
110
#include	<linux/fault-inject.h>
I
Ingo Molnar 已提交
111
#include	<linux/rtmutex.h>
112
#include	<linux/reciprocal_div.h>
L
Linus Torvalds 已提交
113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176

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

/*
 * DEBUG	- 1 for kmem_cache_create() to honour; SLAB_DEBUG_INITIAL,
 *		  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 *)

#ifndef cache_line_size
#define cache_line_size()	L1_CACHE_BYTES
#endif

#ifndef ARCH_KMALLOC_MINALIGN
/*
 * Enforce a minimum alignment for the kmalloc caches.
 * Usually, the kmalloc caches are cache_line_size() aligned, except when
 * DEBUG and FORCED_DEBUG are enabled, then they are BYTES_PER_WORD aligned.
 * Some archs want to perform DMA into kmalloc caches and need a guaranteed
 * alignment larger than BYTES_PER_WORD. ARCH_KMALLOC_MINALIGN allows that.
 * Note that this flag disables some debug features.
 */
#define ARCH_KMALLOC_MINALIGN 0
#endif

#ifndef ARCH_SLAB_MINALIGN
/*
 * Enforce a minimum alignment for all caches.
 * Intended for archs that get misalignment faults even for BYTES_PER_WORD
 * aligned buffers. Includes ARCH_KMALLOC_MINALIGN.
 * If possible: Do not enable this flag for CONFIG_DEBUG_SLAB, it disables
 * some debug features.
 */
#define ARCH_SLAB_MINALIGN 0
#endif

#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_DEBUG_INITIAL | SLAB_RED_ZONE | \
			 SLAB_POISON | SLAB_HWCACHE_ALIGN | \
177
			 SLAB_CACHE_DMA | \
L
Linus Torvalds 已提交
178 179
			 SLAB_MUST_HWCACHE_ALIGN | SLAB_STORE_USER | \
			 SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
180
			 SLAB_DESTROY_BY_RCU | SLAB_MEM_SPREAD)
L
Linus Torvalds 已提交
181
#else
182
# define CREATE_MASK	(SLAB_HWCACHE_ALIGN | \
L
Linus Torvalds 已提交
183 184
			 SLAB_CACHE_DMA | SLAB_MUST_HWCACHE_ALIGN | \
			 SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
185
			 SLAB_DESTROY_BY_RCU | SLAB_MEM_SPREAD)
L
Linus Torvalds 已提交
186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206
#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.
 */

207
typedef unsigned int kmem_bufctl_t;
L
Linus Torvalds 已提交
208 209
#define BUFCTL_END	(((kmem_bufctl_t)(~0U))-0)
#define BUFCTL_FREE	(((kmem_bufctl_t)(~0U))-1)
210 211
#define	BUFCTL_ACTIVE	(((kmem_bufctl_t)(~0U))-2)
#define	SLAB_LIMIT	(((kmem_bufctl_t)(~0U))-3)
L
Linus Torvalds 已提交
212 213 214 215 216 217 218 219 220

/*
 * 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 {
P
Pekka Enberg 已提交
221 222 223 224 225 226
	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;
L
Linus Torvalds 已提交
227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245
};

/*
 * 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.
 *
 * We assume struct slab_rcu can overlay struct slab when destroying.
 */
struct slab_rcu {
P
Pekka Enberg 已提交
246
	struct rcu_head head;
247
	struct kmem_cache *cachep;
P
Pekka Enberg 已提交
248
	void *addr;
L
Linus Torvalds 已提交
249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267
};

/*
 * 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;
268
	spinlock_t lock;
A
Andrew Morton 已提交
269 270 271 272 273 274
	void *entry[0];	/*
			 * Must have this definition in here for the proper
			 * alignment of array_cache. Also simplifies accessing
			 * the entries.
			 * [0] is for gcc 2.95. It should really be [].
			 */
L
Linus Torvalds 已提交
275 276
};

A
Andrew Morton 已提交
277 278 279
/*
 * bootstrap: The caches do not work without cpuarrays anymore, but the
 * cpuarrays are allocated from the generic caches...
L
Linus Torvalds 已提交
280 281 282 283
 */
#define BOOT_CPUCACHE_ENTRIES	1
struct arraycache_init {
	struct array_cache cache;
P
Pekka Enberg 已提交
284
	void *entries[BOOT_CPUCACHE_ENTRIES];
L
Linus Torvalds 已提交
285 286 287
};

/*
288
 * The slab lists for all objects.
L
Linus Torvalds 已提交
289 290
 */
struct kmem_list3 {
P
Pekka Enberg 已提交
291 292 293 294 295
	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;
296
	unsigned int colour_next;	/* Per-node cache coloring */
P
Pekka Enberg 已提交
297 298 299
	spinlock_t list_lock;
	struct array_cache *shared;	/* shared per node */
	struct array_cache **alien;	/* on other nodes */
300 301
	unsigned long next_reap;	/* updated without locking */
	int free_touched;		/* updated without locking */
L
Linus Torvalds 已提交
302 303
};

304 305 306 307 308 309 310 311 312
/*
 * Need this for bootstrapping a per node allocator.
 */
#define NUM_INIT_LISTS (2 * MAX_NUMNODES + 1)
struct kmem_list3 __initdata initkmem_list3[NUM_INIT_LISTS];
#define	CACHE_CACHE 0
#define	SIZE_AC 1
#define	SIZE_L3 (1 + MAX_NUMNODES)

313 314 315 316
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);
317
static int enable_cpucache(struct kmem_cache *cachep);
318
static void cache_reap(struct work_struct *unused);
319

320
/*
A
Andrew Morton 已提交
321 322
 * 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.
323
 */
324
static __always_inline int index_of(const size_t size)
325
{
326 327
	extern void __bad_size(void);

328 329 330 331 332 333 334 335 336 337
	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
338
		__bad_size();
339
	} else
340
		__bad_size();
341 342 343
	return 0;
}

344 345
static int slab_early_init = 1;

346 347
#define INDEX_AC index_of(sizeof(struct arraycache_init))
#define INDEX_L3 index_of(sizeof(struct kmem_list3))
L
Linus Torvalds 已提交
348

P
Pekka Enberg 已提交
349
static void kmem_list3_init(struct kmem_list3 *parent)
350 351 352 353 354 355
{
	INIT_LIST_HEAD(&parent->slabs_full);
	INIT_LIST_HEAD(&parent->slabs_partial);
	INIT_LIST_HEAD(&parent->slabs_free);
	parent->shared = NULL;
	parent->alien = NULL;
356
	parent->colour_next = 0;
357 358 359 360 361
	spin_lock_init(&parent->list_lock);
	parent->free_objects = 0;
	parent->free_touched = 0;
}

A
Andrew Morton 已提交
362 363 364 365
#define MAKE_LIST(cachep, listp, slab, nodeid)				\
	do {								\
		INIT_LIST_HEAD(listp);					\
		list_splice(&(cachep->nodelists[nodeid]->slab), listp);	\
366 367
	} while (0)

A
Andrew Morton 已提交
368 369
#define	MAKE_ALL_LISTS(cachep, ptr, nodeid)				\
	do {								\
370 371 372 373
	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)
L
Linus Torvalds 已提交
374 375

/*
376
 * struct kmem_cache
L
Linus Torvalds 已提交
377 378 379
 *
 * manages a cache.
 */
P
Pekka Enberg 已提交
380

381
struct kmem_cache {
L
Linus Torvalds 已提交
382
/* 1) per-cpu data, touched during every alloc/free */
P
Pekka Enberg 已提交
383
	struct array_cache *array[NR_CPUS];
384
/* 2) Cache tunables. Protected by cache_chain_mutex */
P
Pekka Enberg 已提交
385 386 387
	unsigned int batchcount;
	unsigned int limit;
	unsigned int shared;
388

389
	unsigned int buffer_size;
390
	u32 reciprocal_buffer_size;
391
/* 3) touched by every alloc & free from the backend */
P
Pekka Enberg 已提交
392
	struct kmem_list3 *nodelists[MAX_NUMNODES];
393

A
Andrew Morton 已提交
394 395
	unsigned int flags;		/* constant flags */
	unsigned int num;		/* # of objs per slab */
L
Linus Torvalds 已提交
396

397
/* 4) cache_grow/shrink */
L
Linus Torvalds 已提交
398
	/* order of pgs per slab (2^n) */
P
Pekka Enberg 已提交
399
	unsigned int gfporder;
L
Linus Torvalds 已提交
400 401

	/* force GFP flags, e.g. GFP_DMA */
P
Pekka Enberg 已提交
402
	gfp_t gfpflags;
L
Linus Torvalds 已提交
403

A
Andrew Morton 已提交
404
	size_t colour;			/* cache colouring range */
P
Pekka Enberg 已提交
405
	unsigned int colour_off;	/* colour offset */
406
	struct kmem_cache *slabp_cache;
P
Pekka Enberg 已提交
407
	unsigned int slab_size;
A
Andrew Morton 已提交
408
	unsigned int dflags;		/* dynamic flags */
L
Linus Torvalds 已提交
409 410

	/* constructor func */
411
	void (*ctor) (void *, struct kmem_cache *, unsigned long);
L
Linus Torvalds 已提交
412 413

	/* de-constructor func */
414
	void (*dtor) (void *, struct kmem_cache *, unsigned long);
L
Linus Torvalds 已提交
415

416
/* 5) cache creation/removal */
P
Pekka Enberg 已提交
417 418
	const char *name;
	struct list_head next;
L
Linus Torvalds 已提交
419

420
/* 6) statistics */
L
Linus Torvalds 已提交
421
#if STATS
P
Pekka Enberg 已提交
422 423 424 425 426 427 428 429 430
	unsigned long num_active;
	unsigned long num_allocations;
	unsigned long high_mark;
	unsigned long grown;
	unsigned long reaped;
	unsigned long errors;
	unsigned long max_freeable;
	unsigned long node_allocs;
	unsigned long node_frees;
431
	unsigned long node_overflow;
P
Pekka Enberg 已提交
432 433 434 435
	atomic_t allochit;
	atomic_t allocmiss;
	atomic_t freehit;
	atomic_t freemiss;
L
Linus Torvalds 已提交
436 437
#endif
#if DEBUG
438 439 440 441 442 443 444 445
	/*
	 * If debugging is enabled, then the allocator can add additional
	 * fields and/or padding to every object. buffer_size contains the total
	 * object size including these internal fields, the following two
	 * variables contain the offset to the user object and its size.
	 */
	int obj_offset;
	int obj_size;
L
Linus Torvalds 已提交
446 447 448 449 450 451 452
#endif
};

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

#define BATCHREFILL_LIMIT	16
A
Andrew Morton 已提交
453 454 455
/*
 * Optimization question: fewer reaps means less probability for unnessary
 * cpucache drain/refill cycles.
L
Linus Torvalds 已提交
456
 *
A
Adrian Bunk 已提交
457
 * OTOH the cpuarrays can contain lots of objects,
L
Linus Torvalds 已提交
458 459 460 461 462 463 464 465 466 467
 * 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++)
468
#define	STATS_ADD_REAPED(x,y)	((x)->reaped += (y))
A
Andrew Morton 已提交
469 470 471 472 473
#define	STATS_SET_HIGH(x)						\
	do {								\
		if ((x)->num_active > (x)->high_mark)			\
			(x)->high_mark = (x)->num_active;		\
	} while (0)
L
Linus Torvalds 已提交
474 475
#define	STATS_INC_ERR(x)	((x)->errors++)
#define	STATS_INC_NODEALLOCS(x)	((x)->node_allocs++)
476
#define	STATS_INC_NODEFREES(x)	((x)->node_frees++)
477
#define STATS_INC_ACOVERFLOW(x)   ((x)->node_overflow++)
A
Andrew Morton 已提交
478 479 480 481 482
#define	STATS_SET_FREEABLE(x, i)					\
	do {								\
		if ((x)->max_freeable < i)				\
			(x)->max_freeable = i;				\
	} while (0)
L
Linus Torvalds 已提交
483 484 485 486 487 488 489 490 491
#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)
492
#define	STATS_ADD_REAPED(x,y)	do { } while (0)
L
Linus Torvalds 已提交
493 494 495
#define	STATS_SET_HIGH(x)	do { } while (0)
#define	STATS_INC_ERR(x)	do { } while (0)
#define	STATS_INC_NODEALLOCS(x)	do { } while (0)
496
#define	STATS_INC_NODEFREES(x)	do { } while (0)
497
#define STATS_INC_ACOVERFLOW(x)   do { } while (0)
A
Andrew Morton 已提交
498
#define	STATS_SET_FREEABLE(x, i) do { } while (0)
L
Linus Torvalds 已提交
499 500 501 502 503 504 505 506
#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

A
Andrew Morton 已提交
507 508
/*
 * memory layout of objects:
L
Linus Torvalds 已提交
509
 * 0		: objp
510
 * 0 .. cachep->obj_offset - BYTES_PER_WORD - 1: padding. This ensures that
L
Linus Torvalds 已提交
511 512
 * 		the end of an object is aligned with the end of the real
 * 		allocation. Catches writes behind the end of the allocation.
513
 * cachep->obj_offset - BYTES_PER_WORD .. cachep->obj_offset - 1:
L
Linus Torvalds 已提交
514
 * 		redzone word.
515 516
 * cachep->obj_offset: The real object.
 * cachep->buffer_size - 2* BYTES_PER_WORD: redzone word [BYTES_PER_WORD long]
A
Andrew Morton 已提交
517 518
 * cachep->buffer_size - 1* BYTES_PER_WORD: last caller address
 *					[BYTES_PER_WORD long]
L
Linus Torvalds 已提交
519
 */
520
static int obj_offset(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
521
{
522
	return cachep->obj_offset;
L
Linus Torvalds 已提交
523 524
}

525
static int obj_size(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
526
{
527
	return cachep->obj_size;
L
Linus Torvalds 已提交
528 529
}

530
static unsigned long *dbg_redzone1(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
531 532
{
	BUG_ON(!(cachep->flags & SLAB_RED_ZONE));
533
	return (unsigned long*) (objp+obj_offset(cachep)-BYTES_PER_WORD);
L
Linus Torvalds 已提交
534 535
}

536
static unsigned long *dbg_redzone2(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
537 538 539
{
	BUG_ON(!(cachep->flags & SLAB_RED_ZONE));
	if (cachep->flags & SLAB_STORE_USER)
540
		return (unsigned long *)(objp + cachep->buffer_size -
P
Pekka Enberg 已提交
541
					 2 * BYTES_PER_WORD);
542
	return (unsigned long *)(objp + cachep->buffer_size - BYTES_PER_WORD);
L
Linus Torvalds 已提交
543 544
}

545
static void **dbg_userword(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
546 547
{
	BUG_ON(!(cachep->flags & SLAB_STORE_USER));
548
	return (void **)(objp + cachep->buffer_size - BYTES_PER_WORD);
L
Linus Torvalds 已提交
549 550 551 552
}

#else

553 554
#define obj_offset(x)			0
#define obj_size(cachep)		(cachep->buffer_size)
L
Linus Torvalds 已提交
555 556 557 558 559 560 561
#define dbg_redzone1(cachep, objp)	({BUG(); (unsigned long *)NULL;})
#define dbg_redzone2(cachep, objp)	({BUG(); (unsigned long *)NULL;})
#define dbg_userword(cachep, objp)	({BUG(); (void **)NULL;})

#endif

/*
A
Andrew Morton 已提交
562 563
 * Maximum size of an obj (in 2^order pages) and absolute limit for the gfp
 * order.
L
Linus Torvalds 已提交
564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582
 */
#if defined(CONFIG_LARGE_ALLOCS)
#define	MAX_OBJ_ORDER	13	/* up to 32Mb */
#define	MAX_GFP_ORDER	13	/* up to 32Mb */
#elif defined(CONFIG_MMU)
#define	MAX_OBJ_ORDER	5	/* 32 pages */
#define	MAX_GFP_ORDER	5	/* 32 pages */
#else
#define	MAX_OBJ_ORDER	8	/* up to 1Mb */
#define	MAX_GFP_ORDER	8	/* up to 1Mb */
#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;

A
Andrew Morton 已提交
583 584 585 586
/*
 * 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.
L
Linus Torvalds 已提交
587
 */
588 589 590 591 592 593 594
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)
{
595 596
	if (unlikely(PageCompound(page)))
		page = (struct page *)page_private(page);
597
	BUG_ON(!PageSlab(page));
598 599 600 601 602 603 604 605 606 607
	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)
{
608 609
	if (unlikely(PageCompound(page)))
		page = (struct page *)page_private(page);
610
	BUG_ON(!PageSlab(page));
611 612
	return (struct slab *)page->lru.prev;
}
L
Linus Torvalds 已提交
613

614 615 616 617 618 619 620 621 622 623 624 625
static inline struct kmem_cache *virt_to_cache(const void *obj)
{
	struct page *page = virt_to_page(obj);
	return page_get_cache(page);
}

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

626 627 628 629 630 631
static inline void *index_to_obj(struct kmem_cache *cache, struct slab *slab,
				 unsigned int idx)
{
	return slab->s_mem + cache->buffer_size * idx;
}

632 633 634 635 636 637 638 639
/*
 * 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)
640
{
641 642
	u32 offset = (obj - slab->s_mem);
	return reciprocal_divide(offset, cache->reciprocal_buffer_size);
643 644
}

A
Andrew Morton 已提交
645 646 647
/*
 * These are the default caches for kmalloc. Custom caches can have other sizes.
 */
L
Linus Torvalds 已提交
648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664
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>
P
Pekka Enberg 已提交
665
	{NULL,}
L
Linus Torvalds 已提交
666 667 668 669
#undef CACHE
};

static struct arraycache_init initarray_cache __initdata =
P
Pekka Enberg 已提交
670
    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
L
Linus Torvalds 已提交
671
static struct arraycache_init initarray_generic =
P
Pekka Enberg 已提交
672
    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
L
Linus Torvalds 已提交
673 674

/* internal cache of cache description objs */
675
static struct kmem_cache cache_cache = {
P
Pekka Enberg 已提交
676 677 678
	.batchcount = 1,
	.limit = BOOT_CPUCACHE_ENTRIES,
	.shared = 1,
679
	.buffer_size = sizeof(struct kmem_cache),
P
Pekka Enberg 已提交
680
	.name = "kmem_cache",
L
Linus Torvalds 已提交
681
#if DEBUG
682
	.obj_size = sizeof(struct kmem_cache),
L
Linus Torvalds 已提交
683 684 685
#endif
};

686 687
#define BAD_ALIEN_MAGIC 0x01020304ul

688 689 690 691 692 693 694 695
#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.
696 697 698 699
 *
 * 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
700
 */
701 702 703 704
static struct lock_class_key on_slab_l3_key;
static struct lock_class_key on_slab_alc_key;

static inline void init_lock_keys(void)
705 706 707

{
	int q;
708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734
	struct cache_sizes *s = malloc_sizes;

	while (s->cs_size != ULONG_MAX) {
		for_each_node(q) {
			struct array_cache **alc;
			int r;
			struct kmem_list3 *l3 = s->cs_cachep->nodelists[q];
			if (!l3 || OFF_SLAB(s->cs_cachep))
				continue;
			lockdep_set_class(&l3->list_lock, &on_slab_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)
				continue;
			for_each_node(r) {
				if (alc[r])
					lockdep_set_class(&alc[r]->lock,
					     &on_slab_alc_key);
			}
		}
		s++;
735 736 737
	}
}
#else
738
static inline void init_lock_keys(void)
739 740 741 742
{
}
#endif

743 744 745 746
/*
 * 1. Guard access to the cache-chain.
 * 2. Protect sanity of cpu_online_map against cpu hotplug events
 */
I
Ingo Molnar 已提交
747
static DEFINE_MUTEX(cache_chain_mutex);
L
Linus Torvalds 已提交
748 749 750 751 752 753 754 755
static struct list_head cache_chain;

/*
 * chicken and egg problem: delay the per-cpu array allocation
 * until the general caches are up.
 */
static enum {
	NONE,
756 757
	PARTIAL_AC,
	PARTIAL_L3,
L
Linus Torvalds 已提交
758 759 760
	FULL
} g_cpucache_up;

761 762 763 764 765 766 767 768
/*
 * used by boot code to determine if it can use slab based allocator
 */
int slab_is_available(void)
{
	return g_cpucache_up == FULL;
}

769
static DEFINE_PER_CPU(struct delayed_work, reap_work);
L
Linus Torvalds 已提交
770

771
static inline struct array_cache *cpu_cache_get(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
772 773 774 775
{
	return cachep->array[smp_processor_id()];
}

A
Andrew Morton 已提交
776 777
static inline struct kmem_cache *__find_general_cachep(size_t size,
							gfp_t gfpflags)
L
Linus Torvalds 已提交
778 779 780 781 782
{
	struct cache_sizes *csizep = malloc_sizes;

#if DEBUG
	/* This happens if someone tries to call
P
Pekka Enberg 已提交
783 784 785
	 * kmem_cache_create(), or __kmalloc(), before
	 * the generic caches are initialized.
	 */
786
	BUG_ON(malloc_sizes[INDEX_AC].cs_cachep == NULL);
L
Linus Torvalds 已提交
787 788 789 790 791
#endif
	while (size > csizep->cs_size)
		csizep++;

	/*
792
	 * Really subtle: The last entry with cs->cs_size==ULONG_MAX
L
Linus Torvalds 已提交
793 794 795 796 797 798 799 800
	 * has cs_{dma,}cachep==NULL. Thus no special case
	 * for large kmalloc calls required.
	 */
	if (unlikely(gfpflags & GFP_DMA))
		return csizep->cs_dmacachep;
	return csizep->cs_cachep;
}

A
Adrian Bunk 已提交
801
static struct kmem_cache *kmem_find_general_cachep(size_t size, gfp_t gfpflags)
802 803 804 805
{
	return __find_general_cachep(size, gfpflags);
}

806
static size_t slab_mgmt_size(size_t nr_objs, size_t align)
L
Linus Torvalds 已提交
807
{
808 809
	return ALIGN(sizeof(struct slab)+nr_objs*sizeof(kmem_bufctl_t), align);
}
L
Linus Torvalds 已提交
810

A
Andrew Morton 已提交
811 812 813
/*
 * Calculate the number of objects and left-over bytes for a given buffer size.
 */
814 815 816 817 818 819 820
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;
L
Linus Torvalds 已提交
821

822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869
	/*
	 * 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;
L
Linus Torvalds 已提交
870 871 872 873
}

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

A
Andrew Morton 已提交
874 875
static void __slab_error(const char *function, struct kmem_cache *cachep,
			char *msg)
L
Linus Torvalds 已提交
876 877
{
	printk(KERN_ERR "slab error in %s(): cache `%s': %s\n",
P
Pekka Enberg 已提交
878
	       function, cachep->name, msg);
L
Linus Torvalds 已提交
879 880 881
	dump_stack();
}

882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897
/*
 * 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);

898 899 900 901 902 903 904 905 906 907 908 909 910 911 912
#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, reap_node);

static void init_reap_node(int cpu)
{
	int node;

	node = next_node(cpu_to_node(cpu), node_online_map);
	if (node == MAX_NUMNODES)
913
		node = first_node(node_online_map);
914

915
	per_cpu(reap_node, cpu) = node;
916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938
}

static void next_reap_node(void)
{
	int node = __get_cpu_var(reap_node);

	/*
	 * Also drain per cpu pages on remote zones
	 */
	if (node != numa_node_id())
		drain_node_pages(node);

	node = next_node(node, node_online_map);
	if (unlikely(node >= MAX_NUMNODES))
		node = first_node(node_online_map);
	__get_cpu_var(reap_node) = node;
}

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

L
Linus Torvalds 已提交
939 940 941 942 943 944 945 946 947
/*
 * 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 __devinit start_cpu_timer(int cpu)
{
948
	struct delayed_work *reap_work = &per_cpu(reap_work, cpu);
L
Linus Torvalds 已提交
949 950 951 952 953 954

	/*
	 * When this gets called from do_initcalls via cpucache_init(),
	 * init_workqueues() has already run, so keventd will be setup
	 * at that time.
	 */
955
	if (keventd_up() && reap_work->work.func == NULL) {
956
		init_reap_node(cpu);
957
		INIT_DELAYED_WORK(reap_work, cache_reap);
958 959
		schedule_delayed_work_on(cpu, reap_work,
					__round_jiffies_relative(HZ, cpu));
L
Linus Torvalds 已提交
960 961 962
	}
}

963
static struct array_cache *alloc_arraycache(int node, int entries,
P
Pekka Enberg 已提交
964
					    int batchcount)
L
Linus Torvalds 已提交
965
{
P
Pekka Enberg 已提交
966
	int memsize = sizeof(void *) * entries + sizeof(struct array_cache);
L
Linus Torvalds 已提交
967 968
	struct array_cache *nc = NULL;

969
	nc = kmalloc_node(memsize, GFP_KERNEL, node);
L
Linus Torvalds 已提交
970 971 972 973 974
	if (nc) {
		nc->avail = 0;
		nc->limit = entries;
		nc->batchcount = batchcount;
		nc->touched = 0;
975
		spin_lock_init(&nc->lock);
L
Linus Torvalds 已提交
976 977 978 979
	}
	return nc;
}

980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003
/*
 * 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 = min(min(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;
	to->touched = 1;
	return nr;
}

1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028
#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)
{
	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;
}

1029
static inline void *____cache_alloc_node(struct kmem_cache *cachep,
1030 1031 1032 1033 1034 1035 1036
		 gfp_t flags, int nodeid)
{
	return NULL;
}

#else	/* CONFIG_NUMA */

1037
static void *____cache_alloc_node(struct kmem_cache *, gfp_t, int);
1038
static void *alternate_node_alloc(struct kmem_cache *, gfp_t);
1039

P
Pekka Enberg 已提交
1040
static struct array_cache **alloc_alien_cache(int node, int limit)
1041 1042
{
	struct array_cache **ac_ptr;
P
Pekka Enberg 已提交
1043
	int memsize = sizeof(void *) * MAX_NUMNODES;
1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056
	int i;

	if (limit > 1)
		limit = 12;
	ac_ptr = kmalloc_node(memsize, GFP_KERNEL, node);
	if (ac_ptr) {
		for_each_node(i) {
			if (i == node || !node_online(i)) {
				ac_ptr[i] = NULL;
				continue;
			}
			ac_ptr[i] = alloc_arraycache(node, limit, 0xbaadf00d);
			if (!ac_ptr[i]) {
P
Pekka Enberg 已提交
1057
				for (i--; i <= 0; i--)
1058 1059 1060 1061 1062 1063 1064 1065 1066
					kfree(ac_ptr[i]);
				kfree(ac_ptr);
				return NULL;
			}
		}
	}
	return ac_ptr;
}

P
Pekka Enberg 已提交
1067
static void free_alien_cache(struct array_cache **ac_ptr)
1068 1069 1070 1071 1072 1073
{
	int i;

	if (!ac_ptr)
		return;
	for_each_node(i)
P
Pekka Enberg 已提交
1074
	    kfree(ac_ptr[i]);
1075 1076 1077
	kfree(ac_ptr);
}

1078
static void __drain_alien_cache(struct kmem_cache *cachep,
P
Pekka Enberg 已提交
1079
				struct array_cache *ac, int node)
1080 1081 1082 1083 1084
{
	struct kmem_list3 *rl3 = cachep->nodelists[node];

	if (ac->avail) {
		spin_lock(&rl3->list_lock);
1085 1086 1087 1088 1089
		/*
		 * 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.
		 */
1090 1091
		if (rl3->shared)
			transfer_objects(rl3->shared, ac, ac->limit);
1092

1093
		free_block(cachep, ac->entry, ac->avail, node);
1094 1095 1096 1097 1098
		ac->avail = 0;
		spin_unlock(&rl3->list_lock);
	}
}

1099 1100 1101 1102 1103 1104 1105 1106 1107
/*
 * 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 = __get_cpu_var(reap_node);

	if (l3->alien) {
		struct array_cache *ac = l3->alien[node];
1108 1109

		if (ac && ac->avail && spin_trylock_irq(&ac->lock)) {
1110 1111 1112 1113 1114 1115
			__drain_alien_cache(cachep, ac, node);
			spin_unlock_irq(&ac->lock);
		}
	}
}

A
Andrew Morton 已提交
1116 1117
static void drain_alien_cache(struct kmem_cache *cachep,
				struct array_cache **alien)
1118
{
P
Pekka Enberg 已提交
1119
	int i = 0;
1120 1121 1122 1123
	struct array_cache *ac;
	unsigned long flags;

	for_each_online_node(i) {
1124
		ac = alien[i];
1125 1126 1127 1128 1129 1130 1131
		if (ac) {
			spin_lock_irqsave(&ac->lock, flags);
			__drain_alien_cache(cachep, ac, i);
			spin_unlock_irqrestore(&ac->lock, flags);
		}
	}
}
1132

1133
static inline int cache_free_alien(struct kmem_cache *cachep, void *objp)
1134 1135 1136 1137 1138
{
	struct slab *slabp = virt_to_slab(objp);
	int nodeid = slabp->nodeid;
	struct kmem_list3 *l3;
	struct array_cache *alien = NULL;
P
Pekka Enberg 已提交
1139 1140 1141
	int node;

	node = numa_node_id();
1142 1143 1144 1145 1146

	/*
	 * Make sure we are not freeing a object from another node to the array
	 * cache on this cpu.
	 */
1147
	if (likely(slabp->nodeid == node) || unlikely(!use_alien_caches))
1148 1149
		return 0;

P
Pekka Enberg 已提交
1150
	l3 = cachep->nodelists[node];
1151 1152 1153
	STATS_INC_NODEFREES(cachep);
	if (l3->alien && l3->alien[nodeid]) {
		alien = l3->alien[nodeid];
1154
		spin_lock(&alien->lock);
1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167
		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;
}
1168 1169
#endif

1170
static int __cpuinit cpuup_callback(struct notifier_block *nfb,
P
Pekka Enberg 已提交
1171
				    unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
1172 1173
{
	long cpu = (long)hcpu;
1174
	struct kmem_cache *cachep;
1175 1176 1177
	struct kmem_list3 *l3 = NULL;
	int node = cpu_to_node(cpu);
	int memsize = sizeof(struct kmem_list3);
L
Linus Torvalds 已提交
1178 1179 1180

	switch (action) {
	case CPU_UP_PREPARE:
I
Ingo Molnar 已提交
1181
		mutex_lock(&cache_chain_mutex);
A
Andrew Morton 已提交
1182 1183
		/*
		 * We need to do this right in the beginning since
1184 1185 1186 1187 1188
		 * 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
		 */

L
Linus Torvalds 已提交
1189
		list_for_each_entry(cachep, &cache_chain, next) {
A
Andrew Morton 已提交
1190 1191
			/*
			 * Set up the size64 kmemlist for cpu before we can
1192 1193 1194 1195
			 * begin anything. Make sure some other cpu on this
			 * node has not already allocated this
			 */
			if (!cachep->nodelists[node]) {
A
Andrew Morton 已提交
1196 1197
				l3 = kmalloc_node(memsize, GFP_KERNEL, node);
				if (!l3)
1198 1199 1200
					goto bad;
				kmem_list3_init(l3);
				l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
P
Pekka Enberg 已提交
1201
				    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
1202

1203 1204 1205 1206 1207
				/*
				 * The l3s don't come and go as CPUs come and
				 * go.  cache_chain_mutex is sufficient
				 * protection here.
				 */
1208 1209
				cachep->nodelists[node] = l3;
			}
L
Linus Torvalds 已提交
1210

1211 1212
			spin_lock_irq(&cachep->nodelists[node]->list_lock);
			cachep->nodelists[node]->free_limit =
A
Andrew Morton 已提交
1213 1214
				(1 + nr_cpus_node(node)) *
				cachep->batchcount + cachep->num;
1215 1216 1217
			spin_unlock_irq(&cachep->nodelists[node]->list_lock);
		}

A
Andrew Morton 已提交
1218 1219 1220 1221
		/*
		 * Now we can go ahead with allocating the shared arrays and
		 * array caches
		 */
1222
		list_for_each_entry(cachep, &cache_chain, next) {
1223
			struct array_cache *nc;
1224
			struct array_cache *shared;
1225
			struct array_cache **alien = NULL;
1226

1227
			nc = alloc_arraycache(node, cachep->limit,
1228
						cachep->batchcount);
L
Linus Torvalds 已提交
1229 1230
			if (!nc)
				goto bad;
1231 1232 1233 1234 1235
			shared = alloc_arraycache(node,
					cachep->shared * cachep->batchcount,
					0xbaadf00d);
			if (!shared)
				goto bad;
1236

1237 1238 1239 1240 1241
			if (use_alien_caches) {
                                alien = alloc_alien_cache(node, cachep->limit);
                                if (!alien)
                                        goto bad;
                        }
L
Linus Torvalds 已提交
1242
			cachep->array[cpu] = nc;
1243 1244 1245
			l3 = cachep->nodelists[node];
			BUG_ON(!l3);

1246 1247 1248 1249 1250 1251 1252 1253
			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;
1254
			}
1255 1256 1257 1258 1259 1260 1261 1262 1263
#ifdef CONFIG_NUMA
			if (!l3->alien) {
				l3->alien = alien;
				alien = NULL;
			}
#endif
			spin_unlock_irq(&l3->list_lock);
			kfree(shared);
			free_alien_cache(alien);
L
Linus Torvalds 已提交
1264 1265 1266
		}
		break;
	case CPU_ONLINE:
1267
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1268 1269 1270
		start_cpu_timer(cpu);
		break;
#ifdef CONFIG_HOTPLUG_CPU
1271 1272 1273 1274 1275 1276
	case CPU_DOWN_PREPARE:
		mutex_lock(&cache_chain_mutex);
		break;
	case CPU_DOWN_FAILED:
		mutex_unlock(&cache_chain_mutex);
		break;
L
Linus Torvalds 已提交
1277
	case CPU_DEAD:
1278 1279 1280 1281 1282 1283 1284 1285
		/*
		 * 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().
		 */
L
Linus Torvalds 已提交
1286
		/* fall thru */
1287
#endif
L
Linus Torvalds 已提交
1288 1289 1290
	case CPU_UP_CANCELED:
		list_for_each_entry(cachep, &cache_chain, next) {
			struct array_cache *nc;
1291 1292
			struct array_cache *shared;
			struct array_cache **alien;
1293
			cpumask_t mask;
L
Linus Torvalds 已提交
1294

1295
			mask = node_to_cpumask(node);
L
Linus Torvalds 已提交
1296 1297 1298
			/* cpu is dead; no one can alloc from it. */
			nc = cachep->array[cpu];
			cachep->array[cpu] = NULL;
1299 1300 1301
			l3 = cachep->nodelists[node];

			if (!l3)
1302
				goto free_array_cache;
1303

1304
			spin_lock_irq(&l3->list_lock);
1305 1306 1307 1308

			/* Free limit for this kmem_list3 */
			l3->free_limit -= cachep->batchcount;
			if (nc)
1309
				free_block(cachep, nc->entry, nc->avail, node);
1310 1311

			if (!cpus_empty(mask)) {
1312
				spin_unlock_irq(&l3->list_lock);
1313
				goto free_array_cache;
P
Pekka Enberg 已提交
1314
			}
1315

1316 1317
			shared = l3->shared;
			if (shared) {
1318
				free_block(cachep, l3->shared->entry,
P
Pekka Enberg 已提交
1319
					   l3->shared->avail, node);
1320 1321 1322
				l3->shared = NULL;
			}

1323 1324 1325 1326 1327 1328 1329 1330 1331
			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);
1332
			}
1333
free_array_cache:
L
Linus Torvalds 已提交
1334 1335
			kfree(nc);
		}
1336 1337 1338 1339 1340 1341 1342 1343 1344
		/*
		 * 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;
1345
			drain_freelist(cachep, l3, l3->free_objects);
1346
		}
I
Ingo Molnar 已提交
1347
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1348 1349 1350
		break;
	}
	return NOTIFY_OK;
A
Andrew Morton 已提交
1351
bad:
L
Linus Torvalds 已提交
1352 1353 1354
	return NOTIFY_BAD;
}

1355 1356 1357
static struct notifier_block __cpuinitdata cpucache_notifier = {
	&cpuup_callback, NULL, 0
};
L
Linus Torvalds 已提交
1358

1359 1360 1361
/*
 * swap the static kmem_list3 with kmalloced memory
 */
A
Andrew Morton 已提交
1362 1363
static void init_list(struct kmem_cache *cachep, struct kmem_list3 *list,
			int nodeid)
1364 1365 1366 1367 1368 1369 1370 1371
{
	struct kmem_list3 *ptr;

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

	local_irq_disable();
	memcpy(ptr, list, sizeof(struct kmem_list3));
1372 1373 1374 1375 1376
	/*
	 * Do not assume that spinlocks can be initialized via memcpy:
	 */
	spin_lock_init(&ptr->list_lock);

1377 1378 1379 1380 1381
	MAKE_ALL_LISTS(cachep, ptr, nodeid);
	cachep->nodelists[nodeid] = ptr;
	local_irq_enable();
}

A
Andrew Morton 已提交
1382 1383 1384
/*
 * Initialisation.  Called after the page allocator have been initialised and
 * before smp_init().
L
Linus Torvalds 已提交
1385 1386 1387 1388 1389 1390
 */
void __init kmem_cache_init(void)
{
	size_t left_over;
	struct cache_sizes *sizes;
	struct cache_names *names;
1391
	int i;
1392
	int order;
P
Pekka Enberg 已提交
1393
	int node;
1394 1395 1396 1397 1398 1399

	for (i = 0; i < NUM_INIT_LISTS; i++) {
		kmem_list3_init(&initkmem_list3[i]);
		if (i < MAX_NUMNODES)
			cache_cache.nodelists[i] = NULL;
	}
L
Linus Torvalds 已提交
1400 1401 1402 1403 1404 1405 1406 1407 1408 1409

	/*
	 * Fragmentation resistance on low memory - only use bigger
	 * page orders on machines with more than 32MB of memory.
	 */
	if (num_physpages > (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:
A
Andrew Morton 已提交
1410 1411 1412
	 * 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.
1413 1414 1415
	 *    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.
L
Linus Torvalds 已提交
1416
	 * 2) Create the first kmalloc cache.
1417
	 *    The struct kmem_cache for the new cache is allocated normally.
1418 1419 1420
	 *    An __init data area is used for the head array.
	 * 3) Create the remaining kmalloc caches, with minimally sized
	 *    head arrays.
L
Linus Torvalds 已提交
1421 1422
	 * 4) Replace the __init data head arrays for cache_cache and the first
	 *    kmalloc cache with kmalloc allocated arrays.
1423 1424 1425
	 * 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.
L
Linus Torvalds 已提交
1426 1427
	 */

P
Pekka Enberg 已提交
1428 1429
	node = numa_node_id();

L
Linus Torvalds 已提交
1430 1431 1432 1433 1434
	/* 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;
P
Pekka Enberg 已提交
1435
	cache_cache.nodelists[node] = &initkmem_list3[CACHE_CACHE];
L
Linus Torvalds 已提交
1436

A
Andrew Morton 已提交
1437 1438
	cache_cache.buffer_size = ALIGN(cache_cache.buffer_size,
					cache_line_size());
1439 1440
	cache_cache.reciprocal_buffer_size =
		reciprocal_value(cache_cache.buffer_size);
L
Linus Torvalds 已提交
1441

1442 1443 1444 1445 1446 1447
	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;
	}
1448
	BUG_ON(!cache_cache.num);
1449
	cache_cache.gfporder = order;
P
Pekka Enberg 已提交
1450 1451 1452
	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());
L
Linus Torvalds 已提交
1453 1454 1455 1456 1457

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

A
Andrew Morton 已提交
1458 1459 1460 1461
	/*
	 * Initialize the caches that provide memory for the array cache and the
	 * kmem_list3 structures first.  Without this, further allocations will
	 * bug.
1462 1463 1464
	 */

	sizes[INDEX_AC].cs_cachep = kmem_cache_create(names[INDEX_AC].name,
A
Andrew Morton 已提交
1465 1466 1467 1468
					sizes[INDEX_AC].cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_PANIC,
					NULL, NULL);
1469

A
Andrew Morton 已提交
1470
	if (INDEX_AC != INDEX_L3) {
1471
		sizes[INDEX_L3].cs_cachep =
A
Andrew Morton 已提交
1472 1473 1474 1475 1476 1477
			kmem_cache_create(names[INDEX_L3].name,
				sizes[INDEX_L3].cs_size,
				ARCH_KMALLOC_MINALIGN,
				ARCH_KMALLOC_FLAGS|SLAB_PANIC,
				NULL, NULL);
	}
1478

1479 1480
	slab_early_init = 0;

L
Linus Torvalds 已提交
1481
	while (sizes->cs_size != ULONG_MAX) {
1482 1483
		/*
		 * For performance, all the general caches are L1 aligned.
L
Linus Torvalds 已提交
1484 1485 1486
		 * This should be particularly beneficial on SMP boxes, as it
		 * eliminates "false sharing".
		 * Note for systems short on memory removing the alignment will
1487 1488
		 * allow tighter packing of the smaller caches.
		 */
A
Andrew Morton 已提交
1489
		if (!sizes->cs_cachep) {
1490
			sizes->cs_cachep = kmem_cache_create(names->name,
A
Andrew Morton 已提交
1491 1492 1493 1494 1495
					sizes->cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_PANIC,
					NULL, NULL);
		}
L
Linus Torvalds 已提交
1496 1497

		sizes->cs_dmacachep = kmem_cache_create(names->name_dma,
A
Andrew Morton 已提交
1498 1499 1500 1501 1502
					sizes->cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_CACHE_DMA|
						SLAB_PANIC,
					NULL, NULL);
L
Linus Torvalds 已提交
1503 1504 1505 1506 1507
		sizes++;
		names++;
	}
	/* 4) Replace the bootstrap head arrays */
	{
1508
		struct array_cache *ptr;
1509

L
Linus Torvalds 已提交
1510
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
1511

L
Linus Torvalds 已提交
1512
		local_irq_disable();
1513 1514
		BUG_ON(cpu_cache_get(&cache_cache) != &initarray_cache.cache);
		memcpy(ptr, cpu_cache_get(&cache_cache),
P
Pekka Enberg 已提交
1515
		       sizeof(struct arraycache_init));
1516 1517 1518 1519 1520
		/*
		 * Do not assume that spinlocks can be initialized via memcpy:
		 */
		spin_lock_init(&ptr->lock);

L
Linus Torvalds 已提交
1521 1522
		cache_cache.array[smp_processor_id()] = ptr;
		local_irq_enable();
1523

L
Linus Torvalds 已提交
1524
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
1525

L
Linus Torvalds 已提交
1526
		local_irq_disable();
1527
		BUG_ON(cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep)
P
Pekka Enberg 已提交
1528
		       != &initarray_generic.cache);
1529
		memcpy(ptr, cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep),
P
Pekka Enberg 已提交
1530
		       sizeof(struct arraycache_init));
1531 1532 1533 1534 1535
		/*
		 * Do not assume that spinlocks can be initialized via memcpy:
		 */
		spin_lock_init(&ptr->lock);

1536
		malloc_sizes[INDEX_AC].cs_cachep->array[smp_processor_id()] =
P
Pekka Enberg 已提交
1537
		    ptr;
L
Linus Torvalds 已提交
1538 1539
		local_irq_enable();
	}
1540 1541
	/* 5) Replace the bootstrap kmem_list3's */
	{
P
Pekka Enberg 已提交
1542 1543
		int nid;

1544
		/* Replace the static kmem_list3 structures for the boot cpu */
P
Pekka Enberg 已提交
1545
		init_list(&cache_cache, &initkmem_list3[CACHE_CACHE], node);
1546

P
Pekka Enberg 已提交
1547
		for_each_online_node(nid) {
1548
			init_list(malloc_sizes[INDEX_AC].cs_cachep,
P
Pekka Enberg 已提交
1549
				  &initkmem_list3[SIZE_AC + nid], nid);
1550 1551 1552

			if (INDEX_AC != INDEX_L3) {
				init_list(malloc_sizes[INDEX_L3].cs_cachep,
P
Pekka Enberg 已提交
1553
					  &initkmem_list3[SIZE_L3 + nid], nid);
1554 1555 1556
			}
		}
	}
L
Linus Torvalds 已提交
1557

1558
	/* 6) resize the head arrays to their final sizes */
L
Linus Torvalds 已提交
1559
	{
1560
		struct kmem_cache *cachep;
I
Ingo Molnar 已提交
1561
		mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1562
		list_for_each_entry(cachep, &cache_chain, next)
1563 1564
			if (enable_cpucache(cachep))
				BUG();
I
Ingo Molnar 已提交
1565
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1566 1567
	}

1568 1569 1570 1571
	/* Annotate slab for lockdep -- annotate the malloc caches */
	init_lock_keys();


L
Linus Torvalds 已提交
1572 1573 1574
	/* Done! */
	g_cpucache_up = FULL;

A
Andrew Morton 已提交
1575 1576 1577
	/*
	 * Register a cpu startup notifier callback that initializes
	 * cpu_cache_get for all new cpus
L
Linus Torvalds 已提交
1578 1579 1580
	 */
	register_cpu_notifier(&cpucache_notifier);

A
Andrew Morton 已提交
1581 1582 1583
	/*
	 * The reap timers are started later, with a module init call: That part
	 * of the kernel is not yet operational.
L
Linus Torvalds 已提交
1584 1585 1586 1587 1588 1589 1590
	 */
}

static int __init cpucache_init(void)
{
	int cpu;

A
Andrew Morton 已提交
1591 1592
	/*
	 * Register the timers that return unneeded pages to the page allocator
L
Linus Torvalds 已提交
1593
	 */
1594
	for_each_online_cpu(cpu)
A
Andrew Morton 已提交
1595
		start_cpu_timer(cpu);
L
Linus Torvalds 已提交
1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606
	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.
 */
1607
static void *kmem_getpages(struct kmem_cache *cachep, gfp_t flags, int nodeid)
L
Linus Torvalds 已提交
1608 1609
{
	struct page *page;
1610
	int nr_pages;
L
Linus Torvalds 已提交
1611 1612
	int i;

1613
#ifndef CONFIG_MMU
1614 1615 1616
	/*
	 * Nommu uses slab's for process anonymous memory allocations, and thus
	 * requires __GFP_COMP to properly refcount higher order allocations
1617
	 */
1618
	flags |= __GFP_COMP;
1619
#endif
1620

1621
	flags |= cachep->gfpflags;
1622 1623

	page = alloc_pages_node(nodeid, flags, cachep->gfporder);
L
Linus Torvalds 已提交
1624 1625 1626
	if (!page)
		return NULL;

1627
	nr_pages = (1 << cachep->gfporder);
L
Linus Torvalds 已提交
1628
	if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
1629 1630 1631 1632 1633
		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);
1634 1635 1636
	for (i = 0; i < nr_pages; i++)
		__SetPageSlab(page + i);
	return page_address(page);
L
Linus Torvalds 已提交
1637 1638 1639 1640 1641
}

/*
 * Interface to system's page release.
 */
1642
static void kmem_freepages(struct kmem_cache *cachep, void *addr)
L
Linus Torvalds 已提交
1643
{
P
Pekka Enberg 已提交
1644
	unsigned long i = (1 << cachep->gfporder);
L
Linus Torvalds 已提交
1645 1646 1647
	struct page *page = virt_to_page(addr);
	const unsigned long nr_freed = i;

1648 1649 1650 1651 1652 1653
	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);
L
Linus Torvalds 已提交
1654
	while (i--) {
N
Nick Piggin 已提交
1655 1656
		BUG_ON(!PageSlab(page));
		__ClearPageSlab(page);
L
Linus Torvalds 已提交
1657 1658 1659 1660 1661 1662 1663 1664 1665
		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)
{
P
Pekka Enberg 已提交
1666
	struct slab_rcu *slab_rcu = (struct slab_rcu *)head;
1667
	struct kmem_cache *cachep = slab_rcu->cachep;
L
Linus Torvalds 已提交
1668 1669 1670 1671 1672 1673 1674 1675 1676

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

#if DEBUG

#ifdef CONFIG_DEBUG_PAGEALLOC
1677
static void store_stackinfo(struct kmem_cache *cachep, unsigned long *addr,
P
Pekka Enberg 已提交
1678
			    unsigned long caller)
L
Linus Torvalds 已提交
1679
{
1680
	int size = obj_size(cachep);
L
Linus Torvalds 已提交
1681

1682
	addr = (unsigned long *)&((char *)addr)[obj_offset(cachep)];
L
Linus Torvalds 已提交
1683

P
Pekka Enberg 已提交
1684
	if (size < 5 * sizeof(unsigned long))
L
Linus Torvalds 已提交
1685 1686
		return;

P
Pekka Enberg 已提交
1687 1688 1689 1690
	*addr++ = 0x12345678;
	*addr++ = caller;
	*addr++ = smp_processor_id();
	size -= 3 * sizeof(unsigned long);
L
Linus Torvalds 已提交
1691 1692 1693 1694 1695 1696 1697
	{
		unsigned long *sptr = &caller;
		unsigned long svalue;

		while (!kstack_end(sptr)) {
			svalue = *sptr++;
			if (kernel_text_address(svalue)) {
P
Pekka Enberg 已提交
1698
				*addr++ = svalue;
L
Linus Torvalds 已提交
1699 1700 1701 1702 1703 1704 1705
				size -= sizeof(unsigned long);
				if (size <= sizeof(unsigned long))
					break;
			}
		}

	}
P
Pekka Enberg 已提交
1706
	*addr++ = 0x87654321;
L
Linus Torvalds 已提交
1707 1708 1709
}
#endif

1710
static void poison_obj(struct kmem_cache *cachep, void *addr, unsigned char val)
L
Linus Torvalds 已提交
1711
{
1712 1713
	int size = obj_size(cachep);
	addr = &((char *)addr)[obj_offset(cachep)];
L
Linus Torvalds 已提交
1714 1715

	memset(addr, val, size);
P
Pekka Enberg 已提交
1716
	*(unsigned char *)(addr + size - 1) = POISON_END;
L
Linus Torvalds 已提交
1717 1718 1719 1720 1721
}

static void dump_line(char *data, int offset, int limit)
{
	int i;
D
Dave Jones 已提交
1722 1723 1724
	unsigned char error = 0;
	int bad_count = 0;

L
Linus Torvalds 已提交
1725
	printk(KERN_ERR "%03x:", offset);
D
Dave Jones 已提交
1726 1727 1728 1729 1730
	for (i = 0; i < limit; i++) {
		if (data[offset + i] != POISON_FREE) {
			error = data[offset + i];
			bad_count++;
		}
P
Pekka Enberg 已提交
1731
		printk(" %02x", (unsigned char)data[offset + i]);
D
Dave Jones 已提交
1732
	}
L
Linus Torvalds 已提交
1733
	printk("\n");
D
Dave Jones 已提交
1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747

	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
		}
	}
L
Linus Torvalds 已提交
1748 1749 1750 1751 1752
}
#endif

#if DEBUG

1753
static void print_objinfo(struct kmem_cache *cachep, void *objp, int lines)
L
Linus Torvalds 已提交
1754 1755 1756 1757 1758 1759
{
	int i, size;
	char *realobj;

	if (cachep->flags & SLAB_RED_ZONE) {
		printk(KERN_ERR "Redzone: 0x%lx/0x%lx.\n",
A
Andrew Morton 已提交
1760 1761
			*dbg_redzone1(cachep, objp),
			*dbg_redzone2(cachep, objp));
L
Linus Torvalds 已提交
1762 1763 1764 1765
	}

	if (cachep->flags & SLAB_STORE_USER) {
		printk(KERN_ERR "Last user: [<%p>]",
A
Andrew Morton 已提交
1766
			*dbg_userword(cachep, objp));
L
Linus Torvalds 已提交
1767
		print_symbol("(%s)",
A
Andrew Morton 已提交
1768
				(unsigned long)*dbg_userword(cachep, objp));
L
Linus Torvalds 已提交
1769 1770
		printk("\n");
	}
1771 1772
	realobj = (char *)objp + obj_offset(cachep);
	size = obj_size(cachep);
P
Pekka Enberg 已提交
1773
	for (i = 0; i < size && lines; i += 16, lines--) {
L
Linus Torvalds 已提交
1774 1775
		int limit;
		limit = 16;
P
Pekka Enberg 已提交
1776 1777
		if (i + limit > size)
			limit = size - i;
L
Linus Torvalds 已提交
1778 1779 1780 1781
		dump_line(realobj, i, limit);
	}
}

1782
static void check_poison_obj(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
1783 1784 1785 1786 1787
{
	char *realobj;
	int size, i;
	int lines = 0;

1788 1789
	realobj = (char *)objp + obj_offset(cachep);
	size = obj_size(cachep);
L
Linus Torvalds 已提交
1790

P
Pekka Enberg 已提交
1791
	for (i = 0; i < size; i++) {
L
Linus Torvalds 已提交
1792
		char exp = POISON_FREE;
P
Pekka Enberg 已提交
1793
		if (i == size - 1)
L
Linus Torvalds 已提交
1794 1795 1796 1797 1798 1799
			exp = POISON_END;
		if (realobj[i] != exp) {
			int limit;
			/* Mismatch ! */
			/* Print header */
			if (lines == 0) {
P
Pekka Enberg 已提交
1800
				printk(KERN_ERR
A
Andrew Morton 已提交
1801 1802
					"Slab corruption: start=%p, len=%d\n",
					realobj, size);
L
Linus Torvalds 已提交
1803 1804 1805
				print_objinfo(cachep, objp, 0);
			}
			/* Hexdump the affected line */
P
Pekka Enberg 已提交
1806
			i = (i / 16) * 16;
L
Linus Torvalds 已提交
1807
			limit = 16;
P
Pekka Enberg 已提交
1808 1809
			if (i + limit > size)
				limit = size - i;
L
Linus Torvalds 已提交
1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821
			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:
		 */
1822
		struct slab *slabp = virt_to_slab(objp);
1823
		unsigned int objnr;
L
Linus Torvalds 已提交
1824

1825
		objnr = obj_to_index(cachep, slabp, objp);
L
Linus Torvalds 已提交
1826
		if (objnr) {
1827
			objp = index_to_obj(cachep, slabp, objnr - 1);
1828
			realobj = (char *)objp + obj_offset(cachep);
L
Linus Torvalds 已提交
1829
			printk(KERN_ERR "Prev obj: start=%p, len=%d\n",
P
Pekka Enberg 已提交
1830
			       realobj, size);
L
Linus Torvalds 已提交
1831 1832
			print_objinfo(cachep, objp, 2);
		}
P
Pekka Enberg 已提交
1833
		if (objnr + 1 < cachep->num) {
1834
			objp = index_to_obj(cachep, slabp, objnr + 1);
1835
			realobj = (char *)objp + obj_offset(cachep);
L
Linus Torvalds 已提交
1836
			printk(KERN_ERR "Next obj: start=%p, len=%d\n",
P
Pekka Enberg 已提交
1837
			       realobj, size);
L
Linus Torvalds 已提交
1838 1839 1840 1841 1842 1843
			print_objinfo(cachep, objp, 2);
		}
	}
}
#endif

1844 1845
#if DEBUG
/**
1846 1847 1848 1849 1850 1851
 * slab_destroy_objs - destroy a slab and its objects
 * @cachep: cache pointer being destroyed
 * @slabp: slab pointer being destroyed
 *
 * Call the registered destructor for each object in a slab that is being
 * destroyed.
L
Linus Torvalds 已提交
1852
 */
1853
static void slab_destroy_objs(struct kmem_cache *cachep, struct slab *slabp)
L
Linus Torvalds 已提交
1854 1855 1856
{
	int i;
	for (i = 0; i < cachep->num; i++) {
1857
		void *objp = index_to_obj(cachep, slabp, i);
L
Linus Torvalds 已提交
1858 1859 1860

		if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
A
Andrew Morton 已提交
1861 1862
			if (cachep->buffer_size % PAGE_SIZE == 0 &&
					OFF_SLAB(cachep))
P
Pekka Enberg 已提交
1863
				kernel_map_pages(virt_to_page(objp),
A
Andrew Morton 已提交
1864
					cachep->buffer_size / PAGE_SIZE, 1);
L
Linus Torvalds 已提交
1865 1866 1867 1868 1869 1870 1871 1872 1873
			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 "
P
Pekka Enberg 已提交
1874
					   "was overwritten");
L
Linus Torvalds 已提交
1875 1876
			if (*dbg_redzone2(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "end of a freed object "
P
Pekka Enberg 已提交
1877
					   "was overwritten");
L
Linus Torvalds 已提交
1878 1879
		}
		if (cachep->dtor && !(cachep->flags & SLAB_POISON))
1880
			(cachep->dtor) (objp + obj_offset(cachep), cachep, 0);
L
Linus Torvalds 已提交
1881
	}
1882
}
L
Linus Torvalds 已提交
1883
#else
1884
static void slab_destroy_objs(struct kmem_cache *cachep, struct slab *slabp)
1885
{
L
Linus Torvalds 已提交
1886 1887 1888
	if (cachep->dtor) {
		int i;
		for (i = 0; i < cachep->num; i++) {
1889
			void *objp = index_to_obj(cachep, slabp, i);
P
Pekka Enberg 已提交
1890
			(cachep->dtor) (objp, cachep, 0);
L
Linus Torvalds 已提交
1891 1892
		}
	}
1893
}
L
Linus Torvalds 已提交
1894 1895
#endif

1896 1897 1898 1899 1900
/**
 * slab_destroy - destroy and release all objects in a slab
 * @cachep: cache pointer being destroyed
 * @slabp: slab pointer being destroyed
 *
1901
 * Destroy all the objs in a slab, and release the mem back to the system.
A
Andrew Morton 已提交
1902 1903
 * Before calling the slab must have been unlinked from the cache.  The
 * cache-lock is not held/needed.
1904
 */
1905
static void slab_destroy(struct kmem_cache *cachep, struct slab *slabp)
1906 1907 1908 1909
{
	void *addr = slabp->s_mem - slabp->colouroff;

	slab_destroy_objs(cachep, slabp);
L
Linus Torvalds 已提交
1910 1911 1912
	if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU)) {
		struct slab_rcu *slab_rcu;

P
Pekka Enberg 已提交
1913
		slab_rcu = (struct slab_rcu *)slabp;
L
Linus Torvalds 已提交
1914 1915 1916 1917 1918
		slab_rcu->cachep = cachep;
		slab_rcu->addr = addr;
		call_rcu(&slab_rcu->head, kmem_rcu_free);
	} else {
		kmem_freepages(cachep, addr);
1919 1920
		if (OFF_SLAB(cachep))
			kmem_cache_free(cachep->slabp_cache, slabp);
L
Linus Torvalds 已提交
1921 1922 1923
	}
}

A
Andrew Morton 已提交
1924 1925 1926 1927
/*
 * For setting up all the kmem_list3s for cache whose buffer_size is same as
 * size of kmem_list3.
 */
1928
static void set_up_list3s(struct kmem_cache *cachep, int index)
1929 1930 1931 1932
{
	int node;

	for_each_online_node(node) {
P
Pekka Enberg 已提交
1933
		cachep->nodelists[node] = &initkmem_list3[index + node];
1934
		cachep->nodelists[node]->next_reap = jiffies +
P
Pekka Enberg 已提交
1935 1936
		    REAPTIMEOUT_LIST3 +
		    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
1937 1938 1939
	}
}

1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960
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);
}


1961
/**
1962 1963 1964 1965 1966 1967 1968
 * 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.
1969 1970 1971 1972 1973
 *
 * 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.
 */
A
Andrew Morton 已提交
1974
static size_t calculate_slab_order(struct kmem_cache *cachep,
R
Randy Dunlap 已提交
1975
			size_t size, size_t align, unsigned long flags)
1976
{
1977
	unsigned long offslab_limit;
1978
	size_t left_over = 0;
1979
	int gfporder;
1980

A
Andrew Morton 已提交
1981
	for (gfporder = 0; gfporder <= MAX_GFP_ORDER; gfporder++) {
1982 1983 1984
		unsigned int num;
		size_t remainder;

1985
		cache_estimate(gfporder, size, align, flags, &remainder, &num);
1986 1987
		if (!num)
			continue;
1988

1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
		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;
		}
2001

2002
		/* Found something acceptable - save it away */
2003
		cachep->num = num;
2004
		cachep->gfporder = gfporder;
2005 2006
		left_over = remainder;

2007 2008 2009 2010 2011 2012 2013 2014
		/*
		 * 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;

2015 2016 2017 2018
		/*
		 * Large number of objects is good, but very large slabs are
		 * currently bad for the gfp()s.
		 */
2019
		if (gfporder >= slab_break_gfp_order)
2020 2021
			break;

2022 2023 2024
		/*
		 * Acceptable internal fragmentation?
		 */
A
Andrew Morton 已提交
2025
		if (left_over * 8 <= (PAGE_SIZE << gfporder))
2026 2027 2028 2029 2030
			break;
	}
	return left_over;
}

2031
static int setup_cpu_cache(struct kmem_cache *cachep)
2032
{
2033 2034 2035
	if (g_cpucache_up == FULL)
		return enable_cpucache(cachep);

2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081
	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_KERNEL);

		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_KERNEL, node);
				BUG_ON(!cachep->nodelists[node]);
				kmem_list3_init(cachep->nodelists[node]);
			}
		}
	}
	cachep->nodelists[numa_node_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;
2082
	return 0;
2083 2084
}

L
Linus Torvalds 已提交
2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099
/**
 * 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.
 * @dtor: A destructor 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
 * and the @dtor is run before the pages are handed back.
 *
 * @name must be valid until the cache is destroyed. This implies that
A
Andrew Morton 已提交
2100 2101
 * the module calling this has to destroy the cache before getting unloaded.
 *
L
Linus Torvalds 已提交
2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113
 * 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.
 */
2114
struct kmem_cache *
L
Linus Torvalds 已提交
2115
kmem_cache_create (const char *name, size_t size, size_t align,
A
Andrew Morton 已提交
2116 2117
	unsigned long flags,
	void (*ctor)(void*, struct kmem_cache *, unsigned long),
2118
	void (*dtor)(void*, struct kmem_cache *, unsigned long))
L
Linus Torvalds 已提交
2119 2120
{
	size_t left_over, slab_size, ralign;
2121
	struct kmem_cache *cachep = NULL, *pc;
L
Linus Torvalds 已提交
2122 2123 2124 2125

	/*
	 * Sanity checks... these are all serious usage bugs.
	 */
A
Andrew Morton 已提交
2126
	if (!name || in_interrupt() || (size < BYTES_PER_WORD) ||
P
Pekka Enberg 已提交
2127
	    (size > (1 << MAX_OBJ_ORDER) * PAGE_SIZE) || (dtor && !ctor)) {
A
Andrew Morton 已提交
2128 2129
		printk(KERN_ERR "%s: Early error in slab %s\n", __FUNCTION__,
				name);
P
Pekka Enberg 已提交
2130 2131
		BUG();
	}
L
Linus Torvalds 已提交
2132

2133
	/*
2134 2135
	 * We use cache_chain_mutex to ensure a consistent view of
	 * cpu_online_map as well.  Please see cpuup_callback
2136
	 */
I
Ingo Molnar 已提交
2137
	mutex_lock(&cache_chain_mutex);
2138

2139
	list_for_each_entry(pc, &cache_chain, next) {
2140 2141 2142 2143 2144 2145 2146 2147
		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.
		 */
2148
		res = probe_kernel_address(pc->name, tmp);
2149 2150
		if (res) {
			printk("SLAB: cache with size %d has lost its name\n",
2151
			       pc->buffer_size);
2152 2153 2154
			continue;
		}

P
Pekka Enberg 已提交
2155
		if (!strcmp(pc->name, name)) {
2156 2157 2158 2159 2160 2161
			printk("kmem_cache_create: duplicate cache %s\n", name);
			dump_stack();
			goto oops;
		}
	}

L
Linus Torvalds 已提交
2162 2163 2164 2165 2166
#if DEBUG
	WARN_ON(strchr(name, ' '));	/* It confuses parsers */
	if ((flags & SLAB_DEBUG_INITIAL) && !ctor) {
		/* No constructor, but inital state check requested */
		printk(KERN_ERR "%s: No con, but init state check "
P
Pekka Enberg 已提交
2167
		       "requested - %s\n", __FUNCTION__, name);
L
Linus Torvalds 已提交
2168 2169 2170 2171 2172 2173 2174 2175 2176
		flags &= ~SLAB_DEBUG_INITIAL;
	}
#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.
	 */
A
Andrew Morton 已提交
2177
	if (size < 4096 || fls(size - 1) == fls(size-1 + 3 * BYTES_PER_WORD))
P
Pekka Enberg 已提交
2178
		flags |= SLAB_RED_ZONE | SLAB_STORE_USER;
L
Linus Torvalds 已提交
2179 2180 2181 2182 2183 2184 2185 2186 2187 2188
	if (!(flags & SLAB_DESTROY_BY_RCU))
		flags |= SLAB_POISON;
#endif
	if (flags & SLAB_DESTROY_BY_RCU)
		BUG_ON(flags & SLAB_POISON);
#endif
	if (flags & SLAB_DESTROY_BY_RCU)
		BUG_ON(dtor);

	/*
A
Andrew Morton 已提交
2189 2190
	 * Always checks flags, a caller might be expecting debug support which
	 * isn't available.
L
Linus Torvalds 已提交
2191
	 */
2192
	BUG_ON(flags & ~CREATE_MASK);
L
Linus Torvalds 已提交
2193

A
Andrew Morton 已提交
2194 2195
	/*
	 * Check that size is in terms of words.  This is needed to avoid
L
Linus Torvalds 已提交
2196 2197 2198
	 * unaligned accesses for some archs when redzoning is used, and makes
	 * sure any on-slab bufctl's are also correctly aligned.
	 */
P
Pekka Enberg 已提交
2199 2200 2201
	if (size & (BYTES_PER_WORD - 1)) {
		size += (BYTES_PER_WORD - 1);
		size &= ~(BYTES_PER_WORD - 1);
L
Linus Torvalds 已提交
2202 2203
	}

A
Andrew Morton 已提交
2204 2205
	/* calculate the final buffer alignment: */

L
Linus Torvalds 已提交
2206 2207
	/* 1) arch recommendation: can be overridden for debug */
	if (flags & SLAB_HWCACHE_ALIGN) {
A
Andrew Morton 已提交
2208 2209 2210 2211
		/*
		 * Default alignment: as specified by the arch code.  Except if
		 * an object is really small, then squeeze multiple objects into
		 * one cacheline.
L
Linus Torvalds 已提交
2212 2213
		 */
		ralign = cache_line_size();
P
Pekka Enberg 已提交
2214
		while (size <= ralign / 2)
L
Linus Torvalds 已提交
2215 2216 2217 2218
			ralign /= 2;
	} else {
		ralign = BYTES_PER_WORD;
	}
2219 2220 2221 2222 2223 2224 2225 2226 2227

	/*
	 * Redzoning and user store require word alignment. Note this will be
	 * overridden by architecture or caller mandated alignment if either
	 * is greater than BYTES_PER_WORD.
	 */
	if (flags & SLAB_RED_ZONE || flags & SLAB_STORE_USER)
		ralign = BYTES_PER_WORD;

2228
	/* 2) arch mandated alignment */
L
Linus Torvalds 已提交
2229 2230 2231
	if (ralign < ARCH_SLAB_MINALIGN) {
		ralign = ARCH_SLAB_MINALIGN;
	}
2232
	/* 3) caller mandated alignment */
L
Linus Torvalds 已提交
2233 2234 2235
	if (ralign < align) {
		ralign = align;
	}
2236 2237 2238
	/* disable debug if necessary */
	if (ralign > BYTES_PER_WORD)
		flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
A
Andrew Morton 已提交
2239
	/*
2240
	 * 4) Store it.
L
Linus Torvalds 已提交
2241 2242 2243 2244
	 */
	align = ralign;

	/* Get cache's description obj. */
2245
	cachep = kmem_cache_zalloc(&cache_cache, GFP_KERNEL);
L
Linus Torvalds 已提交
2246
	if (!cachep)
2247
		goto oops;
L
Linus Torvalds 已提交
2248 2249

#if DEBUG
2250
	cachep->obj_size = size;
L
Linus Torvalds 已提交
2251

2252 2253 2254 2255
	/*
	 * Both debugging options require word-alignment which is calculated
	 * into align above.
	 */
L
Linus Torvalds 已提交
2256 2257
	if (flags & SLAB_RED_ZONE) {
		/* add space for red zone words */
2258
		cachep->obj_offset += BYTES_PER_WORD;
P
Pekka Enberg 已提交
2259
		size += 2 * BYTES_PER_WORD;
L
Linus Torvalds 已提交
2260 2261
	}
	if (flags & SLAB_STORE_USER) {
2262 2263
		/* user store requires one word storage behind the end of
		 * the real object.
L
Linus Torvalds 已提交
2264 2265 2266 2267
		 */
		size += BYTES_PER_WORD;
	}
#if FORCED_DEBUG && defined(CONFIG_DEBUG_PAGEALLOC)
P
Pekka Enberg 已提交
2268
	if (size >= malloc_sizes[INDEX_L3 + 1].cs_size
2269 2270
	    && cachep->obj_size > cache_line_size() && size < PAGE_SIZE) {
		cachep->obj_offset += PAGE_SIZE - size;
L
Linus Torvalds 已提交
2271 2272 2273 2274 2275
		size = PAGE_SIZE;
	}
#endif
#endif

2276 2277 2278 2279 2280 2281
	/*
	 * Determine if the slab management is 'on' or 'off' slab.
	 * (bootstrapping cannot cope with offslab caches so don't do
	 * it too early on.)
	 */
	if ((size >= (PAGE_SIZE >> 3)) && !slab_early_init)
L
Linus Torvalds 已提交
2282 2283 2284 2285 2286 2287 2288 2289
		/*
		 * 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);

2290
	left_over = calculate_slab_order(cachep, size, align, flags);
L
Linus Torvalds 已提交
2291 2292 2293 2294 2295

	if (!cachep->num) {
		printk("kmem_cache_create: couldn't create cache %s.\n", name);
		kmem_cache_free(&cache_cache, cachep);
		cachep = NULL;
2296
		goto oops;
L
Linus Torvalds 已提交
2297
	}
P
Pekka Enberg 已提交
2298 2299
	slab_size = ALIGN(cachep->num * sizeof(kmem_bufctl_t)
			  + sizeof(struct slab), align);
L
Linus Torvalds 已提交
2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311

	/*
	 * 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 */
P
Pekka Enberg 已提交
2312 2313
		slab_size =
		    cachep->num * sizeof(kmem_bufctl_t) + sizeof(struct slab);
L
Linus Torvalds 已提交
2314 2315 2316 2317 2318 2319
	}

	cachep->colour_off = cache_line_size();
	/* Offset must be a multiple of the alignment. */
	if (cachep->colour_off < align)
		cachep->colour_off = align;
P
Pekka Enberg 已提交
2320
	cachep->colour = left_over / cachep->colour_off;
L
Linus Torvalds 已提交
2321 2322 2323 2324 2325
	cachep->slab_size = slab_size;
	cachep->flags = flags;
	cachep->gfpflags = 0;
	if (flags & SLAB_CACHE_DMA)
		cachep->gfpflags |= GFP_DMA;
2326
	cachep->buffer_size = size;
2327
	cachep->reciprocal_buffer_size = reciprocal_value(size);
L
Linus Torvalds 已提交
2328

2329
	if (flags & CFLGS_OFF_SLAB) {
2330
		cachep->slabp_cache = kmem_find_general_cachep(slab_size, 0u);
2331 2332 2333 2334 2335 2336 2337 2338 2339
		/*
		 * 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(!cachep->slabp_cache);
	}
L
Linus Torvalds 已提交
2340 2341 2342 2343
	cachep->ctor = ctor;
	cachep->dtor = dtor;
	cachep->name = name;

2344 2345 2346 2347 2348
	if (setup_cpu_cache(cachep)) {
		__kmem_cache_destroy(cachep);
		cachep = NULL;
		goto oops;
	}
L
Linus Torvalds 已提交
2349 2350 2351

	/* cache setup completed, link it into the list */
	list_add(&cachep->next, &cache_chain);
A
Andrew Morton 已提交
2352
oops:
L
Linus Torvalds 已提交
2353 2354
	if (!cachep && (flags & SLAB_PANIC))
		panic("kmem_cache_create(): failed to create slab `%s'\n",
P
Pekka Enberg 已提交
2355
		      name);
I
Ingo Molnar 已提交
2356
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371
	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());
}

2372
static void check_spinlock_acquired(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2373 2374 2375
{
#ifdef CONFIG_SMP
	check_irq_off();
2376
	assert_spin_locked(&cachep->nodelists[numa_node_id()]->list_lock);
L
Linus Torvalds 已提交
2377 2378
#endif
}
2379

2380
static void check_spinlock_acquired_node(struct kmem_cache *cachep, int node)
2381 2382 2383 2384 2385 2386 2387
{
#ifdef CONFIG_SMP
	check_irq_off();
	assert_spin_locked(&cachep->nodelists[node]->list_lock);
#endif
}

L
Linus Torvalds 已提交
2388 2389 2390 2391
#else
#define check_irq_off()	do { } while(0)
#define check_irq_on()	do { } while(0)
#define check_spinlock_acquired(x) do { } while(0)
2392
#define check_spinlock_acquired_node(x, y) do { } while(0)
L
Linus Torvalds 已提交
2393 2394
#endif

2395 2396 2397 2398
static void drain_array(struct kmem_cache *cachep, struct kmem_list3 *l3,
			struct array_cache *ac,
			int force, int node);

L
Linus Torvalds 已提交
2399 2400
static void do_drain(void *arg)
{
A
Andrew Morton 已提交
2401
	struct kmem_cache *cachep = arg;
L
Linus Torvalds 已提交
2402
	struct array_cache *ac;
2403
	int node = numa_node_id();
L
Linus Torvalds 已提交
2404 2405

	check_irq_off();
2406
	ac = cpu_cache_get(cachep);
2407 2408 2409
	spin_lock(&cachep->nodelists[node]->list_lock);
	free_block(cachep, ac->entry, ac->avail, node);
	spin_unlock(&cachep->nodelists[node]->list_lock);
L
Linus Torvalds 已提交
2410 2411 2412
	ac->avail = 0;
}

2413
static void drain_cpu_caches(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2414
{
2415 2416 2417
	struct kmem_list3 *l3;
	int node;

A
Andrew Morton 已提交
2418
	on_each_cpu(do_drain, cachep, 1, 1);
L
Linus Torvalds 已提交
2419
	check_irq_on();
P
Pekka Enberg 已提交
2420
	for_each_online_node(node) {
2421
		l3 = cachep->nodelists[node];
2422 2423 2424 2425 2426 2427 2428
		if (l3 && l3->alien)
			drain_alien_cache(cachep, l3->alien);
	}

	for_each_online_node(node) {
		l3 = cachep->nodelists[node];
		if (l3)
2429
			drain_array(cachep, l3, l3->shared, 1, node);
2430
	}
L
Linus Torvalds 已提交
2431 2432
}

2433 2434 2435 2436 2437 2438 2439 2440
/*
 * 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)
L
Linus Torvalds 已提交
2441
{
2442 2443
	struct list_head *p;
	int nr_freed;
L
Linus Torvalds 已提交
2444 2445
	struct slab *slabp;

2446 2447
	nr_freed = 0;
	while (nr_freed < tofree && !list_empty(&l3->slabs_free)) {
L
Linus Torvalds 已提交
2448

2449
		spin_lock_irq(&l3->list_lock);
2450
		p = l3->slabs_free.prev;
2451 2452 2453 2454
		if (p == &l3->slabs_free) {
			spin_unlock_irq(&l3->list_lock);
			goto out;
		}
L
Linus Torvalds 已提交
2455

2456
		slabp = list_entry(p, struct slab, list);
L
Linus Torvalds 已提交
2457
#if DEBUG
2458
		BUG_ON(slabp->inuse);
L
Linus Torvalds 已提交
2459 2460
#endif
		list_del(&slabp->list);
2461 2462 2463 2464 2465
		/*
		 * Safe to drop the lock. The slab is no longer linked
		 * to the cache.
		 */
		l3->free_objects -= cache->num;
2466
		spin_unlock_irq(&l3->list_lock);
2467 2468
		slab_destroy(cache, slabp);
		nr_freed++;
L
Linus Torvalds 已提交
2469
	}
2470 2471
out:
	return nr_freed;
L
Linus Torvalds 已提交
2472 2473
}

2474
/* Called with cache_chain_mutex held to protect against cpu hotplug */
2475
static int __cache_shrink(struct kmem_cache *cachep)
2476 2477 2478 2479 2480 2481 2482 2483 2484
{
	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];
2485 2486 2487 2488 2489 2490 2491
		if (!l3)
			continue;

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

		ret += !list_empty(&l3->slabs_full) ||
			!list_empty(&l3->slabs_partial);
2492 2493 2494 2495
	}
	return (ret ? 1 : 0);
}

L
Linus Torvalds 已提交
2496 2497 2498 2499 2500 2501 2502
/**
 * 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.
 */
2503
int kmem_cache_shrink(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2504
{
2505
	int ret;
2506
	BUG_ON(!cachep || in_interrupt());
L
Linus Torvalds 已提交
2507

2508 2509 2510 2511
	mutex_lock(&cache_chain_mutex);
	ret = __cache_shrink(cachep);
	mutex_unlock(&cache_chain_mutex);
	return ret;
L
Linus Torvalds 已提交
2512 2513 2514 2515 2516 2517 2518
}
EXPORT_SYMBOL(kmem_cache_shrink);

/**
 * kmem_cache_destroy - delete a cache
 * @cachep: the cache to destroy
 *
2519
 * Remove a struct kmem_cache object from the slab cache.
L
Linus Torvalds 已提交
2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530
 *
 * 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 noone will allocate memory from the cache
 * during the kmem_cache_destroy().
 */
2531
void kmem_cache_destroy(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2532
{
2533
	BUG_ON(!cachep || in_interrupt());
L
Linus Torvalds 已提交
2534 2535

	/* Find the cache in the chain of caches. */
I
Ingo Molnar 已提交
2536
	mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2537 2538 2539 2540 2541 2542
	/*
	 * 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");
P
Pekka Enberg 已提交
2543
		list_add(&cachep->next, &cache_chain);
I
Ingo Molnar 已提交
2544
		mutex_unlock(&cache_chain_mutex);
2545
		return;
L
Linus Torvalds 已提交
2546 2547 2548
	}

	if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU))
2549
		synchronize_rcu();
L
Linus Torvalds 已提交
2550

2551
	__kmem_cache_destroy(cachep);
2552
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2553 2554 2555
}
EXPORT_SYMBOL(kmem_cache_destroy);

2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566
/*
 * 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.
 */
2567
static struct slab *alloc_slabmgmt(struct kmem_cache *cachep, void *objp,
2568 2569
				   int colour_off, gfp_t local_flags,
				   int nodeid)
L
Linus Torvalds 已提交
2570 2571
{
	struct slab *slabp;
P
Pekka Enberg 已提交
2572

L
Linus Torvalds 已提交
2573 2574
	if (OFF_SLAB(cachep)) {
		/* Slab management obj is off-slab. */
2575
		slabp = kmem_cache_alloc_node(cachep->slabp_cache,
2576
					      local_flags & ~GFP_THISNODE, nodeid);
L
Linus Torvalds 已提交
2577 2578 2579
		if (!slabp)
			return NULL;
	} else {
P
Pekka Enberg 已提交
2580
		slabp = objp + colour_off;
L
Linus Torvalds 已提交
2581 2582 2583 2584
		colour_off += cachep->slab_size;
	}
	slabp->inuse = 0;
	slabp->colouroff = colour_off;
P
Pekka Enberg 已提交
2585
	slabp->s_mem = objp + colour_off;
2586
	slabp->nodeid = nodeid;
L
Linus Torvalds 已提交
2587 2588 2589 2590 2591
	return slabp;
}

static inline kmem_bufctl_t *slab_bufctl(struct slab *slabp)
{
P
Pekka Enberg 已提交
2592
	return (kmem_bufctl_t *) (slabp + 1);
L
Linus Torvalds 已提交
2593 2594
}

2595
static void cache_init_objs(struct kmem_cache *cachep,
P
Pekka Enberg 已提交
2596
			    struct slab *slabp, unsigned long ctor_flags)
L
Linus Torvalds 已提交
2597 2598 2599 2600
{
	int i;

	for (i = 0; i < cachep->num; i++) {
2601
		void *objp = index_to_obj(cachep, slabp, i);
L
Linus Torvalds 已提交
2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613
#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;
		}
		/*
A
Andrew Morton 已提交
2614 2615 2616
		 * Constructors are not allowed to allocate memory from the same
		 * cache which they are a constructor for.  Otherwise, deadlock.
		 * They must also be threaded.
L
Linus Torvalds 已提交
2617 2618
		 */
		if (cachep->ctor && !(cachep->flags & SLAB_POISON))
2619
			cachep->ctor(objp + obj_offset(cachep), cachep,
P
Pekka Enberg 已提交
2620
				     ctor_flags);
L
Linus Torvalds 已提交
2621 2622 2623 2624

		if (cachep->flags & SLAB_RED_ZONE) {
			if (*dbg_redzone2(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "constructor overwrote the"
P
Pekka Enberg 已提交
2625
					   " end of an object");
L
Linus Torvalds 已提交
2626 2627
			if (*dbg_redzone1(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "constructor overwrote the"
P
Pekka Enberg 已提交
2628
					   " start of an object");
L
Linus Torvalds 已提交
2629
		}
A
Andrew Morton 已提交
2630 2631
		if ((cachep->buffer_size % PAGE_SIZE) == 0 &&
			    OFF_SLAB(cachep) && cachep->flags & SLAB_POISON)
P
Pekka Enberg 已提交
2632
			kernel_map_pages(virt_to_page(objp),
2633
					 cachep->buffer_size / PAGE_SIZE, 0);
L
Linus Torvalds 已提交
2634 2635 2636 2637
#else
		if (cachep->ctor)
			cachep->ctor(objp, cachep, ctor_flags);
#endif
P
Pekka Enberg 已提交
2638
		slab_bufctl(slabp)[i] = i + 1;
L
Linus Torvalds 已提交
2639
	}
P
Pekka Enberg 已提交
2640
	slab_bufctl(slabp)[i - 1] = BUFCTL_END;
L
Linus Torvalds 已提交
2641 2642 2643
	slabp->free = 0;
}

2644
static void kmem_flagcheck(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2645
{
C
Christoph Lameter 已提交
2646
	if (flags & GFP_DMA)
A
Andrew Morton 已提交
2647 2648 2649
		BUG_ON(!(cachep->gfpflags & GFP_DMA));
	else
		BUG_ON(cachep->gfpflags & GFP_DMA);
L
Linus Torvalds 已提交
2650 2651
}

A
Andrew Morton 已提交
2652 2653
static void *slab_get_obj(struct kmem_cache *cachep, struct slab *slabp,
				int nodeid)
2654
{
2655
	void *objp = index_to_obj(cachep, slabp, slabp->free);
2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668
	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;
}

A
Andrew Morton 已提交
2669 2670
static void slab_put_obj(struct kmem_cache *cachep, struct slab *slabp,
				void *objp, int nodeid)
2671
{
2672
	unsigned int objnr = obj_to_index(cachep, slabp, objp);
2673 2674 2675 2676 2677

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

2678
	if (slab_bufctl(slabp)[objnr] + 1 <= SLAB_LIMIT + 1) {
2679
		printk(KERN_ERR "slab: double free detected in cache "
A
Andrew Morton 已提交
2680
				"'%s', objp %p\n", cachep->name, objp);
2681 2682 2683 2684 2685 2686 2687 2688
		BUG();
	}
#endif
	slab_bufctl(slabp)[objnr] = slabp->free;
	slabp->free = objnr;
	slabp->inuse--;
}

2689 2690 2691 2692 2693 2694 2695
/*
 * 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, kmem_ptr_validate, and slab debugging.
 */
static void slab_map_pages(struct kmem_cache *cache, struct slab *slab,
			   void *addr)
L
Linus Torvalds 已提交
2696
{
2697
	int nr_pages;
L
Linus Torvalds 已提交
2698 2699
	struct page *page;

2700
	page = virt_to_page(addr);
2701

2702
	nr_pages = 1;
2703
	if (likely(!PageCompound(page)))
2704 2705
		nr_pages <<= cache->gfporder;

L
Linus Torvalds 已提交
2706
	do {
2707 2708
		page_set_cache(page, cache);
		page_set_slab(page, slab);
L
Linus Torvalds 已提交
2709
		page++;
2710
	} while (--nr_pages);
L
Linus Torvalds 已提交
2711 2712 2713 2714 2715 2716
}

/*
 * 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.
 */
2717 2718
static int cache_grow(struct kmem_cache *cachep,
		gfp_t flags, int nodeid, void *objp)
L
Linus Torvalds 已提交
2719
{
P
Pekka Enberg 已提交
2720 2721 2722 2723
	struct slab *slabp;
	size_t offset;
	gfp_t local_flags;
	unsigned long ctor_flags;
2724
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
2725

A
Andrew Morton 已提交
2726 2727 2728
	/*
	 * Be lazy and only check for valid flags here,  keeping it out of the
	 * critical path in kmem_cache_alloc().
L
Linus Torvalds 已提交
2729
	 */
C
Christoph Lameter 已提交
2730
	BUG_ON(flags & ~(GFP_DMA | GFP_LEVEL_MASK | __GFP_NO_GROW));
2731
	if (flags & __GFP_NO_GROW)
L
Linus Torvalds 已提交
2732 2733 2734
		return 0;

	ctor_flags = SLAB_CTOR_CONSTRUCTOR;
2735
	local_flags = (flags & GFP_LEVEL_MASK);
L
Linus Torvalds 已提交
2736 2737 2738 2739 2740 2741 2742
	if (!(local_flags & __GFP_WAIT))
		/*
		 * Not allowed to sleep.  Need to tell a constructor about
		 * this - it might need to know...
		 */
		ctor_flags |= SLAB_CTOR_ATOMIC;

2743
	/* Take the l3 list lock to change the colour_next on this node */
L
Linus Torvalds 已提交
2744
	check_irq_off();
2745 2746
	l3 = cachep->nodelists[nodeid];
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2747 2748

	/* Get colour for the slab, and cal the next value. */
2749 2750 2751 2752 2753
	offset = l3->colour_next;
	l3->colour_next++;
	if (l3->colour_next >= cachep->colour)
		l3->colour_next = 0;
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2754

2755
	offset *= cachep->colour_off;
L
Linus Torvalds 已提交
2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767

	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);

A
Andrew Morton 已提交
2768 2769 2770
	/*
	 * Get mem for the objs.  Attempt to allocate a physical page from
	 * 'nodeid'.
2771
	 */
2772 2773
	if (!objp)
		objp = kmem_getpages(cachep, flags, nodeid);
A
Andrew Morton 已提交
2774
	if (!objp)
L
Linus Torvalds 已提交
2775 2776 2777
		goto failed;

	/* Get slab management. */
2778 2779
	slabp = alloc_slabmgmt(cachep, objp, offset,
			local_flags & ~GFP_THISNODE, nodeid);
A
Andrew Morton 已提交
2780
	if (!slabp)
L
Linus Torvalds 已提交
2781 2782
		goto opps1;

2783
	slabp->nodeid = nodeid;
2784
	slab_map_pages(cachep, slabp, objp);
L
Linus Torvalds 已提交
2785 2786 2787 2788 2789 2790

	cache_init_objs(cachep, slabp, ctor_flags);

	if (local_flags & __GFP_WAIT)
		local_irq_disable();
	check_irq_off();
2791
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2792 2793

	/* Make slab active. */
2794
	list_add_tail(&slabp->list, &(l3->slabs_free));
L
Linus Torvalds 已提交
2795
	STATS_INC_GROWN(cachep);
2796 2797
	l3->free_objects += cachep->num;
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2798
	return 1;
A
Andrew Morton 已提交
2799
opps1:
L
Linus Torvalds 已提交
2800
	kmem_freepages(cachep, objp);
A
Andrew Morton 已提交
2801
failed:
L
Linus Torvalds 已提交
2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820
	if (local_flags & __GFP_WAIT)
		local_irq_disable();
	return 0;
}

#if DEBUG

/*
 * Perform extra freeing checks:
 * - detect bad pointers.
 * - POISON/RED_ZONE checking
 * - destructor calls, for caches with POISON+dtor
 */
static void kfree_debugcheck(const void *objp)
{
	struct page *page;

	if (!virt_addr_valid(objp)) {
		printk(KERN_ERR "kfree_debugcheck: out of range ptr %lxh.\n",
P
Pekka Enberg 已提交
2821 2822
		       (unsigned long)objp);
		BUG();
L
Linus Torvalds 已提交
2823 2824 2825
	}
	page = virt_to_page(objp);
	if (!PageSlab(page)) {
P
Pekka Enberg 已提交
2826 2827
		printk(KERN_ERR "kfree_debugcheck: bad ptr %lxh.\n",
		       (unsigned long)objp);
L
Linus Torvalds 已提交
2828 2829 2830 2831
		BUG();
	}
}

2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853
static inline void verify_redzone_free(struct kmem_cache *cache, void *obj)
{
	unsigned 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%lx, redzone 2:0x%lx.\n",
			obj, redzone1, redzone2);
}

2854
static void *cache_free_debugcheck(struct kmem_cache *cachep, void *objp,
P
Pekka Enberg 已提交
2855
				   void *caller)
L
Linus Torvalds 已提交
2856 2857 2858 2859 2860
{
	struct page *page;
	unsigned int objnr;
	struct slab *slabp;

2861
	objp -= obj_offset(cachep);
L
Linus Torvalds 已提交
2862 2863 2864
	kfree_debugcheck(objp);
	page = virt_to_page(objp);

2865
	slabp = page_get_slab(page);
L
Linus Torvalds 已提交
2866 2867

	if (cachep->flags & SLAB_RED_ZONE) {
2868
		verify_redzone_free(cachep, objp);
L
Linus Torvalds 已提交
2869 2870 2871 2872 2873 2874
		*dbg_redzone1(cachep, objp) = RED_INACTIVE;
		*dbg_redzone2(cachep, objp) = RED_INACTIVE;
	}
	if (cachep->flags & SLAB_STORE_USER)
		*dbg_userword(cachep, objp) = caller;

2875
	objnr = obj_to_index(cachep, slabp, objp);
L
Linus Torvalds 已提交
2876 2877

	BUG_ON(objnr >= cachep->num);
2878
	BUG_ON(objp != index_to_obj(cachep, slabp, objnr));
L
Linus Torvalds 已提交
2879 2880

	if (cachep->flags & SLAB_DEBUG_INITIAL) {
A
Andrew Morton 已提交
2881 2882 2883 2884
		/*
		 * Need to call the slab's constructor so the caller can
		 * perform a verify of its state (debugging).  Called without
		 * the cache-lock held.
L
Linus Torvalds 已提交
2885
		 */
2886
		cachep->ctor(objp + obj_offset(cachep),
P
Pekka Enberg 已提交
2887
			     cachep, SLAB_CTOR_CONSTRUCTOR | SLAB_CTOR_VERIFY);
L
Linus Torvalds 已提交
2888 2889 2890 2891 2892
	}
	if (cachep->flags & SLAB_POISON && cachep->dtor) {
		/* we want to cache poison the object,
		 * call the destruction callback
		 */
2893
		cachep->dtor(objp + obj_offset(cachep), cachep, 0);
L
Linus Torvalds 已提交
2894
	}
2895 2896 2897
#ifdef CONFIG_DEBUG_SLAB_LEAK
	slab_bufctl(slabp)[objnr] = BUFCTL_FREE;
#endif
L
Linus Torvalds 已提交
2898 2899
	if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
A
Andrew Morton 已提交
2900
		if ((cachep->buffer_size % PAGE_SIZE)==0 && OFF_SLAB(cachep)) {
L
Linus Torvalds 已提交
2901
			store_stackinfo(cachep, objp, (unsigned long)caller);
P
Pekka Enberg 已提交
2902
			kernel_map_pages(virt_to_page(objp),
2903
					 cachep->buffer_size / PAGE_SIZE, 0);
L
Linus Torvalds 已提交
2904 2905 2906 2907 2908 2909 2910 2911 2912 2913
		} else {
			poison_obj(cachep, objp, POISON_FREE);
		}
#else
		poison_obj(cachep, objp, POISON_FREE);
#endif
	}
	return objp;
}

2914
static void check_slabp(struct kmem_cache *cachep, struct slab *slabp)
L
Linus Torvalds 已提交
2915 2916 2917
{
	kmem_bufctl_t i;
	int entries = 0;
P
Pekka Enberg 已提交
2918

L
Linus Torvalds 已提交
2919 2920 2921 2922 2923 2924 2925
	/* 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) {
A
Andrew Morton 已提交
2926 2927 2928 2929
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);
P
Pekka Enberg 已提交
2930
		for (i = 0;
2931
		     i < sizeof(*slabp) + cachep->num * sizeof(kmem_bufctl_t);
P
Pekka Enberg 已提交
2932
		     i++) {
A
Andrew Morton 已提交
2933
			if (i % 16 == 0)
L
Linus Torvalds 已提交
2934
				printk("\n%03x:", i);
P
Pekka Enberg 已提交
2935
			printk(" %02x", ((unsigned char *)slabp)[i]);
L
Linus Torvalds 已提交
2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946
		}
		printk("\n");
		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

2947
static void *cache_alloc_refill(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2948 2949 2950 2951
{
	int batchcount;
	struct kmem_list3 *l3;
	struct array_cache *ac;
P
Pekka Enberg 已提交
2952 2953 2954
	int node;

	node = numa_node_id();
L
Linus Torvalds 已提交
2955 2956

	check_irq_off();
2957
	ac = cpu_cache_get(cachep);
A
Andrew Morton 已提交
2958
retry:
L
Linus Torvalds 已提交
2959 2960
	batchcount = ac->batchcount;
	if (!ac->touched && batchcount > BATCHREFILL_LIMIT) {
A
Andrew Morton 已提交
2961 2962 2963 2964
		/*
		 * If there was little recent activity on this cache, then
		 * perform only a partial refill.  Otherwise we could generate
		 * refill bouncing.
L
Linus Torvalds 已提交
2965 2966 2967
		 */
		batchcount = BATCHREFILL_LIMIT;
	}
P
Pekka Enberg 已提交
2968
	l3 = cachep->nodelists[node];
2969 2970 2971

	BUG_ON(ac->avail > 0 || !l3);
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2972

2973 2974 2975 2976
	/* See if we can refill from the shared array */
	if (l3->shared && transfer_objects(ac, l3->shared, batchcount))
		goto alloc_done;

L
Linus Torvalds 已提交
2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996
	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);
		while (slabp->inuse < cachep->num && batchcount--) {
			STATS_INC_ALLOCED(cachep);
			STATS_INC_ACTIVE(cachep);
			STATS_SET_HIGH(cachep);

2997
			ac->entry[ac->avail++] = slab_get_obj(cachep, slabp,
P
Pekka Enberg 已提交
2998
							    node);
L
Linus Torvalds 已提交
2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009
		}
		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);
	}

A
Andrew Morton 已提交
3010
must_grow:
L
Linus Torvalds 已提交
3011
	l3->free_objects -= ac->avail;
A
Andrew Morton 已提交
3012
alloc_done:
3013
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
3014 3015 3016

	if (unlikely(!ac->avail)) {
		int x;
3017
		x = cache_grow(cachep, flags | GFP_THISNODE, node, NULL);
3018

A
Andrew Morton 已提交
3019
		/* cache_grow can reenable interrupts, then ac could change. */
3020
		ac = cpu_cache_get(cachep);
A
Andrew Morton 已提交
3021
		if (!x && ac->avail == 0)	/* no objects in sight? abort */
L
Linus Torvalds 已提交
3022 3023
			return NULL;

A
Andrew Morton 已提交
3024
		if (!ac->avail)		/* objects refilled by interrupt? */
L
Linus Torvalds 已提交
3025 3026 3027
			goto retry;
	}
	ac->touched = 1;
3028
	return ac->entry[--ac->avail];
L
Linus Torvalds 已提交
3029 3030
}

A
Andrew Morton 已提交
3031 3032
static inline void cache_alloc_debugcheck_before(struct kmem_cache *cachep,
						gfp_t flags)
L
Linus Torvalds 已提交
3033 3034 3035 3036 3037 3038 3039 3040
{
	might_sleep_if(flags & __GFP_WAIT);
#if DEBUG
	kmem_flagcheck(cachep, flags);
#endif
}

#if DEBUG
A
Andrew Morton 已提交
3041 3042
static void *cache_alloc_debugcheck_after(struct kmem_cache *cachep,
				gfp_t flags, void *objp, void *caller)
L
Linus Torvalds 已提交
3043
{
P
Pekka Enberg 已提交
3044
	if (!objp)
L
Linus Torvalds 已提交
3045
		return objp;
P
Pekka Enberg 已提交
3046
	if (cachep->flags & SLAB_POISON) {
L
Linus Torvalds 已提交
3047
#ifdef CONFIG_DEBUG_PAGEALLOC
3048
		if ((cachep->buffer_size % PAGE_SIZE) == 0 && OFF_SLAB(cachep))
P
Pekka Enberg 已提交
3049
			kernel_map_pages(virt_to_page(objp),
3050
					 cachep->buffer_size / PAGE_SIZE, 1);
L
Linus Torvalds 已提交
3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061
		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) {
A
Andrew Morton 已提交
3062 3063 3064 3065
		if (*dbg_redzone1(cachep, objp) != RED_INACTIVE ||
				*dbg_redzone2(cachep, objp) != RED_INACTIVE) {
			slab_error(cachep, "double free, or memory outside"
						" object was overwritten");
P
Pekka Enberg 已提交
3066
			printk(KERN_ERR
A
Andrew Morton 已提交
3067 3068 3069
				"%p: redzone 1:0x%lx, redzone 2:0x%lx\n",
				objp, *dbg_redzone1(cachep, objp),
				*dbg_redzone2(cachep, objp));
L
Linus Torvalds 已提交
3070 3071 3072 3073
		}
		*dbg_redzone1(cachep, objp) = RED_ACTIVE;
		*dbg_redzone2(cachep, objp) = RED_ACTIVE;
	}
3074 3075 3076 3077 3078 3079 3080 3081 3082 3083
#ifdef CONFIG_DEBUG_SLAB_LEAK
	{
		struct slab *slabp;
		unsigned objnr;

		slabp = page_get_slab(virt_to_page(objp));
		objnr = (unsigned)(objp - slabp->s_mem) / cachep->buffer_size;
		slab_bufctl(slabp)[objnr] = BUFCTL_ACTIVE;
	}
#endif
3084
	objp += obj_offset(cachep);
L
Linus Torvalds 已提交
3085
	if (cachep->ctor && cachep->flags & SLAB_POISON) {
P
Pekka Enberg 已提交
3086
		unsigned long ctor_flags = SLAB_CTOR_CONSTRUCTOR;
L
Linus Torvalds 已提交
3087 3088 3089 3090 3091

		if (!(flags & __GFP_WAIT))
			ctor_flags |= SLAB_CTOR_ATOMIC;

		cachep->ctor(objp, cachep, ctor_flags);
P
Pekka Enberg 已提交
3092
	}
3093 3094 3095 3096 3097 3098
#if ARCH_SLAB_MINALIGN
	if ((u32)objp & (ARCH_SLAB_MINALIGN-1)) {
		printk(KERN_ERR "0x%p: not aligned to ARCH_SLAB_MINALIGN=%d\n",
		       objp, ARCH_SLAB_MINALIGN);
	}
#endif
L
Linus Torvalds 已提交
3099 3100 3101 3102 3103 3104
	return objp;
}
#else
#define cache_alloc_debugcheck_after(a,b,objp,d) (objp)
#endif

3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117
#ifdef CONFIG_FAILSLAB

static struct failslab_attr {

	struct fault_attr attr;

	u32 ignore_gfp_wait;
#ifdef CONFIG_FAULT_INJECTION_DEBUG_FS
	struct dentry *ignore_gfp_wait_file;
#endif

} failslab = {
	.attr = FAULT_ATTR_INITIALIZER,
3118
	.ignore_gfp_wait = 1,
3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177
};

static int __init setup_failslab(char *str)
{
	return setup_fault_attr(&failslab.attr, str);
}
__setup("failslab=", setup_failslab);

static int should_failslab(struct kmem_cache *cachep, gfp_t flags)
{
	if (cachep == &cache_cache)
		return 0;
	if (flags & __GFP_NOFAIL)
		return 0;
	if (failslab.ignore_gfp_wait && (flags & __GFP_WAIT))
		return 0;

	return should_fail(&failslab.attr, obj_size(cachep));
}

#ifdef CONFIG_FAULT_INJECTION_DEBUG_FS

static int __init failslab_debugfs(void)
{
	mode_t mode = S_IFREG | S_IRUSR | S_IWUSR;
	struct dentry *dir;
	int err;

       	err = init_fault_attr_dentries(&failslab.attr, "failslab");
	if (err)
		return err;
	dir = failslab.attr.dentries.dir;

	failslab.ignore_gfp_wait_file =
		debugfs_create_bool("ignore-gfp-wait", mode, dir,
				      &failslab.ignore_gfp_wait);

	if (!failslab.ignore_gfp_wait_file) {
		err = -ENOMEM;
		debugfs_remove(failslab.ignore_gfp_wait_file);
		cleanup_fault_attr_dentries(&failslab.attr);
	}

	return err;
}

late_initcall(failslab_debugfs);

#endif /* CONFIG_FAULT_INJECTION_DEBUG_FS */

#else /* CONFIG_FAILSLAB */

static inline int should_failslab(struct kmem_cache *cachep, gfp_t flags)
{
	return 0;
}

#endif /* CONFIG_FAILSLAB */

3178
static inline void *____cache_alloc(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
3179
{
P
Pekka Enberg 已提交
3180
	void *objp;
L
Linus Torvalds 已提交
3181 3182
	struct array_cache *ac;

3183
	check_irq_off();
3184 3185 3186 3187

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

3188
	ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
3189 3190 3191
	if (likely(ac->avail)) {
		STATS_INC_ALLOCHIT(cachep);
		ac->touched = 1;
3192
		objp = ac->entry[--ac->avail];
L
Linus Torvalds 已提交
3193 3194 3195 3196
	} else {
		STATS_INC_ALLOCMISS(cachep);
		objp = cache_alloc_refill(cachep, flags);
	}
3197 3198 3199
	return objp;
}

A
Andrew Morton 已提交
3200 3201
static __always_inline void *__cache_alloc(struct kmem_cache *cachep,
						gfp_t flags, void *caller)
3202 3203
{
	unsigned long save_flags;
3204
	void *objp = NULL;
3205 3206 3207 3208

	cache_alloc_debugcheck_before(cachep, flags);

	local_irq_save(save_flags);
3209

3210 3211
	if (unlikely(NUMA_BUILD &&
			current->flags & (PF_SPREAD_SLAB | PF_MEMPOLICY)))
3212 3213 3214 3215
		objp = alternate_node_alloc(cachep, flags);

	if (!objp)
		objp = ____cache_alloc(cachep, flags);
3216 3217
	/*
	 * We may just have run out of memory on the local node.
3218
	 * ____cache_alloc_node() knows how to locate memory on other nodes
3219 3220
	 */
 	if (NUMA_BUILD && !objp)
3221
 		objp = ____cache_alloc_node(cachep, flags, numa_node_id());
L
Linus Torvalds 已提交
3222
	local_irq_restore(save_flags);
3223
	objp = cache_alloc_debugcheck_after(cachep, flags, objp,
3224
					    caller);
3225
	prefetchw(objp);
L
Linus Torvalds 已提交
3226 3227 3228
	return objp;
}

3229
#ifdef CONFIG_NUMA
3230
/*
3231
 * Try allocating on another node if PF_SPREAD_SLAB|PF_MEMPOLICY.
3232 3233 3234 3235 3236 3237 3238 3239
 *
 * 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;

3240
	if (in_interrupt() || (flags & __GFP_THISNODE))
3241 3242 3243 3244 3245 3246 3247
		return NULL;
	nid_alloc = nid_here = numa_node_id();
	if (cpuset_do_slab_mem_spread() && (cachep->flags & SLAB_MEM_SPREAD))
		nid_alloc = cpuset_mem_spread_node();
	else if (current->mempolicy)
		nid_alloc = slab_node(current->mempolicy);
	if (nid_alloc != nid_here)
3248
		return ____cache_alloc_node(cachep, flags, nid_alloc);
3249 3250 3251
	return NULL;
}

3252 3253
/*
 * Fallback function if there was no memory available and no objects on a
3254 3255 3256 3257 3258
 * 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.
3259 3260 3261 3262 3263 3264 3265
 */
void *fallback_alloc(struct kmem_cache *cache, gfp_t flags)
{
	struct zonelist *zonelist = &NODE_DATA(slab_node(current->mempolicy))
					->node_zonelists[gfp_zone(flags)];
	struct zone **z;
	void *obj = NULL;
3266
	int nid;
3267
	gfp_t local_flags = (flags & GFP_LEVEL_MASK);
3268

3269 3270 3271 3272 3273
retry:
	/*
	 * Look through allowed nodes for objects available
	 * from existing per node queues.
	 */
3274
	for (z = zonelist->zones; *z && !obj; z++) {
3275
		nid = zone_to_nid(*z);
3276

3277
		if (cpuset_zone_allowed_hardwall(*z, flags) &&
3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290
			cache->nodelists[nid] &&
			cache->nodelists[nid]->free_objects)
				obj = ____cache_alloc_node(cache,
					flags | GFP_THISNODE, nid);
	}

	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.
		 */
3291 3292 3293
		if (local_flags & __GFP_WAIT)
			local_irq_enable();
		kmem_flagcheck(cache, flags);
3294
		obj = kmem_getpages(cache, flags, -1);
3295 3296
		if (local_flags & __GFP_WAIT)
			local_irq_disable();
3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316
		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 {
				kmem_freepages(cache, obj);
				obj = NULL;
			}
		}
3317
	}
3318 3319 3320
	return obj;
}

3321 3322
/*
 * A interface to enable slab creation on nodeid
L
Linus Torvalds 已提交
3323
 */
3324
static void *____cache_alloc_node(struct kmem_cache *cachep, gfp_t flags,
A
Andrew Morton 已提交
3325
				int nodeid)
3326 3327
{
	struct list_head *entry;
P
Pekka Enberg 已提交
3328 3329 3330 3331 3332 3333 3334 3335
	struct slab *slabp;
	struct kmem_list3 *l3;
	void *obj;
	int x;

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

A
Andrew Morton 已提交
3336
retry:
3337
	check_irq_off();
P
Pekka Enberg 已提交
3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356
	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);

3357
	obj = slab_get_obj(cachep, slabp, nodeid);
P
Pekka Enberg 已提交
3358 3359 3360 3361 3362
	check_slabp(cachep, slabp);
	l3->free_objects--;
	/* move slabp to correct slabp list: */
	list_del(&slabp->list);

A
Andrew Morton 已提交
3363
	if (slabp->free == BUFCTL_END)
P
Pekka Enberg 已提交
3364
		list_add(&slabp->list, &l3->slabs_full);
A
Andrew Morton 已提交
3365
	else
P
Pekka Enberg 已提交
3366
		list_add(&slabp->list, &l3->slabs_partial);
3367

P
Pekka Enberg 已提交
3368 3369
	spin_unlock(&l3->list_lock);
	goto done;
3370

A
Andrew Morton 已提交
3371
must_grow:
P
Pekka Enberg 已提交
3372
	spin_unlock(&l3->list_lock);
3373
	x = cache_grow(cachep, flags | GFP_THISNODE, nodeid, NULL);
3374 3375
	if (x)
		goto retry;
L
Linus Torvalds 已提交
3376

3377 3378 3379 3380 3381
	if (!(flags & __GFP_THISNODE))
		/* Unable to grow the cache. Fall back to other nodes. */
		return fallback_alloc(cachep, flags);

	return NULL;
3382

A
Andrew Morton 已提交
3383
done:
P
Pekka Enberg 已提交
3384
	return obj;
3385 3386 3387 3388 3389 3390
}
#endif

/*
 * Caller needs to acquire correct kmem_list's list_lock
 */
3391
static void free_block(struct kmem_cache *cachep, void **objpp, int nr_objects,
P
Pekka Enberg 已提交
3392
		       int node)
L
Linus Torvalds 已提交
3393 3394
{
	int i;
3395
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
3396 3397 3398 3399 3400

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

3401
		slabp = virt_to_slab(objp);
3402
		l3 = cachep->nodelists[node];
L
Linus Torvalds 已提交
3403
		list_del(&slabp->list);
3404
		check_spinlock_acquired_node(cachep, node);
L
Linus Torvalds 已提交
3405
		check_slabp(cachep, slabp);
3406
		slab_put_obj(cachep, slabp, objp, node);
L
Linus Torvalds 已提交
3407
		STATS_DEC_ACTIVE(cachep);
3408
		l3->free_objects++;
L
Linus Torvalds 已提交
3409 3410 3411 3412
		check_slabp(cachep, slabp);

		/* fixup slab chains */
		if (slabp->inuse == 0) {
3413 3414
			if (l3->free_objects > l3->free_limit) {
				l3->free_objects -= cachep->num;
3415 3416 3417 3418 3419 3420
				/* 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.
				 */
L
Linus Torvalds 已提交
3421 3422
				slab_destroy(cachep, slabp);
			} else {
3423
				list_add(&slabp->list, &l3->slabs_free);
L
Linus Torvalds 已提交
3424 3425 3426 3427 3428 3429
			}
		} else {
			/* Unconditionally move a slab to the end of the
			 * partial list on free - maximum time for the
			 * other objects to be freed, too.
			 */
3430
			list_add_tail(&slabp->list, &l3->slabs_partial);
L
Linus Torvalds 已提交
3431 3432 3433 3434
		}
	}
}

3435
static void cache_flusharray(struct kmem_cache *cachep, struct array_cache *ac)
L
Linus Torvalds 已提交
3436 3437
{
	int batchcount;
3438
	struct kmem_list3 *l3;
3439
	int node = numa_node_id();
L
Linus Torvalds 已提交
3440 3441 3442 3443 3444 3445

	batchcount = ac->batchcount;
#if DEBUG
	BUG_ON(!batchcount || batchcount > ac->avail);
#endif
	check_irq_off();
3446
	l3 = cachep->nodelists[node];
3447
	spin_lock(&l3->list_lock);
3448 3449
	if (l3->shared) {
		struct array_cache *shared_array = l3->shared;
P
Pekka Enberg 已提交
3450
		int max = shared_array->limit - shared_array->avail;
L
Linus Torvalds 已提交
3451 3452 3453
		if (max) {
			if (batchcount > max)
				batchcount = max;
3454
			memcpy(&(shared_array->entry[shared_array->avail]),
P
Pekka Enberg 已提交
3455
			       ac->entry, sizeof(void *) * batchcount);
L
Linus Torvalds 已提交
3456 3457 3458 3459 3460
			shared_array->avail += batchcount;
			goto free_done;
		}
	}

3461
	free_block(cachep, ac->entry, batchcount, node);
A
Andrew Morton 已提交
3462
free_done:
L
Linus Torvalds 已提交
3463 3464 3465 3466 3467
#if STATS
	{
		int i = 0;
		struct list_head *p;

3468 3469
		p = l3->slabs_free.next;
		while (p != &(l3->slabs_free)) {
L
Linus Torvalds 已提交
3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480
			struct slab *slabp;

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

			i++;
			p = p->next;
		}
		STATS_SET_FREEABLE(cachep, i);
	}
#endif
3481
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
3482
	ac->avail -= batchcount;
A
Andrew Morton 已提交
3483
	memmove(ac->entry, &(ac->entry[batchcount]), sizeof(void *)*ac->avail);
L
Linus Torvalds 已提交
3484 3485 3486
}

/*
A
Andrew Morton 已提交
3487 3488
 * 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.
L
Linus Torvalds 已提交
3489
 */
3490
static inline void __cache_free(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
3491
{
3492
	struct array_cache *ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
3493 3494 3495 3496

	check_irq_off();
	objp = cache_free_debugcheck(cachep, objp, __builtin_return_address(0));

3497
	if (cache_free_alien(cachep, objp))
3498 3499
		return;

L
Linus Torvalds 已提交
3500 3501
	if (likely(ac->avail < ac->limit)) {
		STATS_INC_FREEHIT(cachep);
3502
		ac->entry[ac->avail++] = objp;
L
Linus Torvalds 已提交
3503 3504 3505 3506
		return;
	} else {
		STATS_INC_FREEMISS(cachep);
		cache_flusharray(cachep, ac);
3507
		ac->entry[ac->avail++] = objp;
L
Linus Torvalds 已提交
3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518
	}
}

/**
 * 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.
 */
3519
void *kmem_cache_alloc(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
3520
{
3521
	return __cache_alloc(cachep, flags, __builtin_return_address(0));
L
Linus Torvalds 已提交
3522 3523 3524
}
EXPORT_SYMBOL(kmem_cache_alloc);

3525
/**
3526
 * kmem_cache_zalloc - Allocate an object. The memory is set to zero.
3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541
 * @cache: The cache to allocate from.
 * @flags: See kmalloc().
 *
 * Allocate an object from this cache and set the allocated memory to zero.
 * The flags are only relevant if the cache has no available objects.
 */
void *kmem_cache_zalloc(struct kmem_cache *cache, gfp_t flags)
{
	void *ret = __cache_alloc(cache, flags, __builtin_return_address(0));
	if (ret)
		memset(ret, 0, obj_size(cache));
	return ret;
}
EXPORT_SYMBOL(kmem_cache_zalloc);

L
Linus Torvalds 已提交
3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555
/**
 * kmem_ptr_validate - check if an untrusted pointer might
 *	be a slab entry.
 * @cachep: the cache we're checking against
 * @ptr: pointer to validate
 *
 * This verifies that the untrusted pointer looks sane:
 * it is _not_ a guarantee that the pointer is actually
 * part of the slab cache in question, but it at least
 * validates that the pointer can be dereferenced and
 * looks half-way sane.
 *
 * Currently only used for dentry validation.
 */
3556
int kmem_ptr_validate(struct kmem_cache *cachep, const void *ptr)
L
Linus Torvalds 已提交
3557
{
P
Pekka Enberg 已提交
3558
	unsigned long addr = (unsigned long)ptr;
L
Linus Torvalds 已提交
3559
	unsigned long min_addr = PAGE_OFFSET;
P
Pekka Enberg 已提交
3560
	unsigned long align_mask = BYTES_PER_WORD - 1;
3561
	unsigned long size = cachep->buffer_size;
L
Linus Torvalds 已提交
3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576
	struct page *page;

	if (unlikely(addr < min_addr))
		goto out;
	if (unlikely(addr > (unsigned long)high_memory - size))
		goto out;
	if (unlikely(addr & align_mask))
		goto out;
	if (unlikely(!kern_addr_valid(addr)))
		goto out;
	if (unlikely(!kern_addr_valid(addr + size - 1)))
		goto out;
	page = virt_to_page(ptr);
	if (unlikely(!PageSlab(page)))
		goto out;
3577
	if (unlikely(page_get_cache(page) != cachep))
L
Linus Torvalds 已提交
3578 3579
		goto out;
	return 1;
A
Andrew Morton 已提交
3580
out:
L
Linus Torvalds 已提交
3581 3582 3583 3584 3585 3586 3587 3588 3589
	return 0;
}

#ifdef CONFIG_NUMA
/**
 * 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.
3590
 * @caller: return address of caller, used for debug information
L
Linus Torvalds 已提交
3591
 *
3592 3593 3594 3595
 * 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.
L
Linus Torvalds 已提交
3596
 */
3597 3598 3599
static __always_inline void *
__cache_alloc_node(struct kmem_cache *cachep, gfp_t flags,
		int nodeid, void *caller)
L
Linus Torvalds 已提交
3600
{
3601
	unsigned long save_flags;
3602
	void *ptr = NULL;
L
Linus Torvalds 已提交
3603

3604 3605
	cache_alloc_debugcheck_before(cachep, flags);
	local_irq_save(save_flags);
3606

3607 3608
	if (unlikely(nodeid == -1))
		nodeid = numa_node_id();
3609

3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630
	if (likely(cachep->nodelists[nodeid])) {
		if (nodeid == numa_node_id()) {
			/*
			 * 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) {
			/* ___cache_alloc_node can fall back to other nodes */
			ptr = ____cache_alloc_node(cachep, flags, nodeid);
		}
	} else {
		/* Node not bootstrapped yet */
		if (!(flags & __GFP_THISNODE))
			ptr = fallback_alloc(cachep, flags);
	}

	local_irq_restore(save_flags);
3631
	ptr = cache_alloc_debugcheck_after(cachep, flags, ptr, caller);
L
Linus Torvalds 已提交
3632

3633
	return ptr;
L
Linus Torvalds 已提交
3634
}
3635 3636 3637 3638 3639 3640

void *kmem_cache_alloc_node(struct kmem_cache *cachep, gfp_t flags, int nodeid)
{
	return __cache_alloc_node(cachep, flags, nodeid,
			__builtin_return_address(0));
}
L
Linus Torvalds 已提交
3641 3642
EXPORT_SYMBOL(kmem_cache_alloc_node);

3643 3644
static __always_inline void *
__do_kmalloc_node(size_t size, gfp_t flags, int node, void *caller)
3645
{
3646
	struct kmem_cache *cachep;
3647 3648 3649 3650 3651 3652

	cachep = kmem_find_general_cachep(size, flags);
	if (unlikely(cachep == NULL))
		return NULL;
	return kmem_cache_alloc_node(cachep, flags, node);
}
3653 3654 3655 3656 3657 3658 3659

#ifdef CONFIG_DEBUG_SLAB
void *__kmalloc_node(size_t size, gfp_t flags, int node)
{
	return __do_kmalloc_node(size, flags, node,
			__builtin_return_address(0));
}
3660
EXPORT_SYMBOL(__kmalloc_node);
3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675

void *__kmalloc_node_track_caller(size_t size, gfp_t flags,
		int node, void *caller)
{
	return __do_kmalloc_node(size, flags, node, 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 */
#endif /* CONFIG_NUMA */
L
Linus Torvalds 已提交
3676 3677

/**
3678
 * __do_kmalloc - allocate memory
L
Linus Torvalds 已提交
3679
 * @size: how many bytes of memory are required.
3680
 * @flags: the type of memory to allocate (see kmalloc).
3681
 * @caller: function caller for debug tracking of the caller
L
Linus Torvalds 已提交
3682
 */
3683 3684
static __always_inline void *__do_kmalloc(size_t size, gfp_t flags,
					  void *caller)
L
Linus Torvalds 已提交
3685
{
3686
	struct kmem_cache *cachep;
L
Linus Torvalds 已提交
3687

3688 3689 3690 3691 3692 3693
	/* 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);
3694 3695
	if (unlikely(cachep == NULL))
		return NULL;
3696 3697 3698 3699
	return __cache_alloc(cachep, flags, caller);
}


3700
#ifdef CONFIG_DEBUG_SLAB
3701 3702
void *__kmalloc(size_t size, gfp_t flags)
{
3703
	return __do_kmalloc(size, flags, __builtin_return_address(0));
L
Linus Torvalds 已提交
3704 3705 3706
}
EXPORT_SYMBOL(__kmalloc);

3707 3708 3709 3710 3711
void *__kmalloc_track_caller(size_t size, gfp_t flags, void *caller)
{
	return __do_kmalloc(size, flags, caller);
}
EXPORT_SYMBOL(__kmalloc_track_caller);
3712 3713 3714 3715 3716 3717 3718

#else
void *__kmalloc(size_t size, gfp_t flags)
{
	return __do_kmalloc(size, flags, NULL);
}
EXPORT_SYMBOL(__kmalloc);
3719 3720
#endif

L
Linus Torvalds 已提交
3721 3722 3723 3724 3725 3726 3727 3728
/**
 * 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.
 */
3729
void kmem_cache_free(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
3730 3731 3732
{
	unsigned long flags;

3733 3734
	BUG_ON(virt_to_cache(objp) != cachep);

L
Linus Torvalds 已提交
3735
	local_irq_save(flags);
3736
	__cache_free(cachep, objp);
L
Linus Torvalds 已提交
3737 3738 3739 3740 3741 3742 3743 3744
	local_irq_restore(flags);
}
EXPORT_SYMBOL(kmem_cache_free);

/**
 * kfree - free previously allocated memory
 * @objp: pointer returned by kmalloc.
 *
3745 3746
 * If @objp is NULL, no operation is performed.
 *
L
Linus Torvalds 已提交
3747 3748 3749 3750 3751
 * Don't free memory not originally allocated by kmalloc()
 * or you will run into trouble.
 */
void kfree(const void *objp)
{
3752
	struct kmem_cache *c;
L
Linus Torvalds 已提交
3753 3754 3755 3756 3757 3758
	unsigned long flags;

	if (unlikely(!objp))
		return;
	local_irq_save(flags);
	kfree_debugcheck(objp);
3759
	c = virt_to_cache(objp);
3760
	debug_check_no_locks_freed(objp, obj_size(c));
3761
	__cache_free(c, (void *)objp);
L
Linus Torvalds 已提交
3762 3763 3764 3765
	local_irq_restore(flags);
}
EXPORT_SYMBOL(kfree);

3766
unsigned int kmem_cache_size(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
3767
{
3768
	return obj_size(cachep);
L
Linus Torvalds 已提交
3769 3770 3771
}
EXPORT_SYMBOL(kmem_cache_size);

3772
const char *kmem_cache_name(struct kmem_cache *cachep)
3773 3774 3775 3776 3777
{
	return cachep->name;
}
EXPORT_SYMBOL_GPL(kmem_cache_name);

3778
/*
3779
 * This initializes kmem_list3 or resizes varioius caches for all nodes.
3780
 */
3781
static int alloc_kmemlist(struct kmem_cache *cachep)
3782 3783 3784
{
	int node;
	struct kmem_list3 *l3;
3785
	struct array_cache *new_shared;
3786
	struct array_cache **new_alien = NULL;
3787 3788

	for_each_online_node(node) {
3789

3790 3791 3792 3793 3794
                if (use_alien_caches) {
                        new_alien = alloc_alien_cache(node, cachep->limit);
                        if (!new_alien)
                                goto fail;
                }
3795

3796 3797
		new_shared = alloc_arraycache(node,
				cachep->shared*cachep->batchcount,
A
Andrew Morton 已提交
3798
					0xbaadf00d);
3799 3800
		if (!new_shared) {
			free_alien_cache(new_alien);
3801
			goto fail;
3802
		}
3803

A
Andrew Morton 已提交
3804 3805
		l3 = cachep->nodelists[node];
		if (l3) {
3806 3807
			struct array_cache *shared = l3->shared;

3808 3809
			spin_lock_irq(&l3->list_lock);

3810
			if (shared)
3811 3812
				free_block(cachep, shared->entry,
						shared->avail, node);
3813

3814 3815
			l3->shared = new_shared;
			if (!l3->alien) {
3816 3817 3818
				l3->alien = new_alien;
				new_alien = NULL;
			}
P
Pekka Enberg 已提交
3819
			l3->free_limit = (1 + nr_cpus_node(node)) *
A
Andrew Morton 已提交
3820
					cachep->batchcount + cachep->num;
3821
			spin_unlock_irq(&l3->list_lock);
3822
			kfree(shared);
3823 3824 3825
			free_alien_cache(new_alien);
			continue;
		}
A
Andrew Morton 已提交
3826
		l3 = kmalloc_node(sizeof(struct kmem_list3), GFP_KERNEL, node);
3827 3828 3829
		if (!l3) {
			free_alien_cache(new_alien);
			kfree(new_shared);
3830
			goto fail;
3831
		}
3832 3833 3834

		kmem_list3_init(l3);
		l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
A
Andrew Morton 已提交
3835
				((unsigned long)cachep) % REAPTIMEOUT_LIST3;
3836
		l3->shared = new_shared;
3837
		l3->alien = new_alien;
P
Pekka Enberg 已提交
3838
		l3->free_limit = (1 + nr_cpus_node(node)) *
A
Andrew Morton 已提交
3839
					cachep->batchcount + cachep->num;
3840 3841
		cachep->nodelists[node] = l3;
	}
3842
	return 0;
3843

A
Andrew Morton 已提交
3844
fail:
3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859
	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--;
		}
	}
3860
	return -ENOMEM;
3861 3862
}

L
Linus Torvalds 已提交
3863
struct ccupdate_struct {
3864
	struct kmem_cache *cachep;
L
Linus Torvalds 已提交
3865 3866 3867 3868 3869
	struct array_cache *new[NR_CPUS];
};

static void do_ccupdate_local(void *info)
{
A
Andrew Morton 已提交
3870
	struct ccupdate_struct *new = info;
L
Linus Torvalds 已提交
3871 3872 3873
	struct array_cache *old;

	check_irq_off();
3874
	old = cpu_cache_get(new->cachep);
3875

L
Linus Torvalds 已提交
3876 3877 3878 3879
	new->cachep->array[smp_processor_id()] = new->new[smp_processor_id()];
	new->new[smp_processor_id()] = old;
}

3880
/* Always called with the cache_chain_mutex held */
A
Andrew Morton 已提交
3881 3882
static int do_tune_cpucache(struct kmem_cache *cachep, int limit,
				int batchcount, int shared)
L
Linus Torvalds 已提交
3883
{
3884
	struct ccupdate_struct *new;
3885
	int i;
L
Linus Torvalds 已提交
3886

3887 3888 3889 3890
	new = kzalloc(sizeof(*new), GFP_KERNEL);
	if (!new)
		return -ENOMEM;

3891
	for_each_online_cpu(i) {
3892
		new->new[i] = alloc_arraycache(cpu_to_node(i), limit,
A
Andrew Morton 已提交
3893
						batchcount);
3894
		if (!new->new[i]) {
P
Pekka Enberg 已提交
3895
			for (i--; i >= 0; i--)
3896 3897
				kfree(new->new[i]);
			kfree(new);
3898
			return -ENOMEM;
L
Linus Torvalds 已提交
3899 3900
		}
	}
3901
	new->cachep = cachep;
L
Linus Torvalds 已提交
3902

3903
	on_each_cpu(do_ccupdate_local, (void *)new, 1, 1);
3904

L
Linus Torvalds 已提交
3905 3906 3907
	check_irq_on();
	cachep->batchcount = batchcount;
	cachep->limit = limit;
3908
	cachep->shared = shared;
L
Linus Torvalds 已提交
3909

3910
	for_each_online_cpu(i) {
3911
		struct array_cache *ccold = new->new[i];
L
Linus Torvalds 已提交
3912 3913
		if (!ccold)
			continue;
3914
		spin_lock_irq(&cachep->nodelists[cpu_to_node(i)]->list_lock);
3915
		free_block(cachep, ccold->entry, ccold->avail, cpu_to_node(i));
3916
		spin_unlock_irq(&cachep->nodelists[cpu_to_node(i)]->list_lock);
L
Linus Torvalds 已提交
3917 3918
		kfree(ccold);
	}
3919
	kfree(new);
3920
	return alloc_kmemlist(cachep);
L
Linus Torvalds 已提交
3921 3922
}

3923
/* Called with cache_chain_mutex held always */
3924
static int enable_cpucache(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
3925 3926 3927 3928
{
	int err;
	int limit, shared;

A
Andrew Morton 已提交
3929 3930
	/*
	 * The head array serves three purposes:
L
Linus Torvalds 已提交
3931 3932
	 * - create a LIFO ordering, i.e. return objects that are cache-warm
	 * - reduce the number of spinlock operations.
A
Andrew Morton 已提交
3933
	 * - reduce the number of linked list operations on the slab and
L
Linus Torvalds 已提交
3934 3935 3936 3937
	 *   bufctl chains: array operations are cheaper.
	 * The numbers are guessed, we should auto-tune as described by
	 * Bonwick.
	 */
3938
	if (cachep->buffer_size > 131072)
L
Linus Torvalds 已提交
3939
		limit = 1;
3940
	else if (cachep->buffer_size > PAGE_SIZE)
L
Linus Torvalds 已提交
3941
		limit = 8;
3942
	else if (cachep->buffer_size > 1024)
L
Linus Torvalds 已提交
3943
		limit = 24;
3944
	else if (cachep->buffer_size > 256)
L
Linus Torvalds 已提交
3945 3946 3947 3948
		limit = 54;
	else
		limit = 120;

A
Andrew Morton 已提交
3949 3950
	/*
	 * CPU bound tasks (e.g. network routing) can exhibit cpu bound
L
Linus Torvalds 已提交
3951 3952 3953 3954 3955 3956 3957 3958 3959
	 * 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;
#ifdef CONFIG_SMP
3960
	if (cachep->buffer_size <= PAGE_SIZE)
L
Linus Torvalds 已提交
3961 3962 3963 3964
		shared = 8;
#endif

#if DEBUG
A
Andrew Morton 已提交
3965 3966 3967
	/*
	 * With debugging enabled, large batchcount lead to excessively long
	 * periods with disabled local interrupts. Limit the batchcount
L
Linus Torvalds 已提交
3968 3969 3970 3971
	 */
	if (limit > 32)
		limit = 32;
#endif
P
Pekka Enberg 已提交
3972
	err = do_tune_cpucache(cachep, limit, (limit + 1) / 2, shared);
L
Linus Torvalds 已提交
3973 3974
	if (err)
		printk(KERN_ERR "enable_cpucache failed for %s, error %d.\n",
P
Pekka Enberg 已提交
3975
		       cachep->name, -err);
3976
	return err;
L
Linus Torvalds 已提交
3977 3978
}

3979 3980
/*
 * Drain an array if it contains any elements taking the l3 lock only if
3981 3982
 * necessary. Note that the l3 listlock also protects the array_cache
 * if drain_array() is used on the shared array.
3983 3984 3985
 */
void drain_array(struct kmem_cache *cachep, struct kmem_list3 *l3,
			 struct array_cache *ac, int force, int node)
L
Linus Torvalds 已提交
3986 3987 3988
{
	int tofree;

3989 3990
	if (!ac || !ac->avail)
		return;
L
Linus Torvalds 已提交
3991 3992
	if (ac->touched && !force) {
		ac->touched = 0;
3993
	} else {
3994
		spin_lock_irq(&l3->list_lock);
3995 3996 3997 3998 3999 4000 4001 4002 4003
		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);
		}
4004
		spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
4005 4006 4007 4008 4009
	}
}

/**
 * cache_reap - Reclaim memory from caches.
4010
 * @unused: unused parameter
L
Linus Torvalds 已提交
4011 4012 4013 4014 4015 4016
 *
 * 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.
 *
A
Andrew Morton 已提交
4017 4018
 * If we cannot acquire the cache chain mutex then just give up - we'll try
 * again on the next iteration.
L
Linus Torvalds 已提交
4019
 */
4020
static void cache_reap(struct work_struct *unused)
L
Linus Torvalds 已提交
4021
{
4022
	struct kmem_cache *searchp;
4023
	struct kmem_list3 *l3;
4024
	int node = numa_node_id();
L
Linus Torvalds 已提交
4025

I
Ingo Molnar 已提交
4026
	if (!mutex_trylock(&cache_chain_mutex)) {
L
Linus Torvalds 已提交
4027
		/* Give up. Setup the next iteration. */
P
Pekka Enberg 已提交
4028
		schedule_delayed_work(&__get_cpu_var(reap_work),
4029
				      round_jiffies_relative(REAPTIMEOUT_CPUC));
L
Linus Torvalds 已提交
4030 4031 4032
		return;
	}

4033
	list_for_each_entry(searchp, &cache_chain, next) {
L
Linus Torvalds 已提交
4034 4035
		check_irq_on();

4036 4037 4038 4039 4040
		/*
		 * 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.
		 */
4041
		l3 = searchp->nodelists[node];
4042

4043
		reap_alien(searchp, l3);
L
Linus Torvalds 已提交
4044

4045
		drain_array(searchp, l3, cpu_cache_get(searchp), 0, node);
L
Linus Torvalds 已提交
4046

4047 4048 4049 4050
		/*
		 * These are racy checks but it does not matter
		 * if we skip one check or scan twice.
		 */
4051
		if (time_after(l3->next_reap, jiffies))
4052
			goto next;
L
Linus Torvalds 已提交
4053

4054
		l3->next_reap = jiffies + REAPTIMEOUT_LIST3;
L
Linus Torvalds 已提交
4055

4056
		drain_array(searchp, l3, l3->shared, 0, node);
L
Linus Torvalds 已提交
4057

4058
		if (l3->free_touched)
4059
			l3->free_touched = 0;
4060 4061
		else {
			int freed;
L
Linus Torvalds 已提交
4062

4063 4064 4065 4066
			freed = drain_freelist(searchp, l3, (l3->free_limit +
				5 * searchp->num - 1) / (5 * searchp->num));
			STATS_ADD_REAPED(searchp, freed);
		}
4067
next:
L
Linus Torvalds 已提交
4068 4069 4070
		cond_resched();
	}
	check_irq_on();
I
Ingo Molnar 已提交
4071
	mutex_unlock(&cache_chain_mutex);
4072
	next_reap_node();
4073
	refresh_cpu_vm_stats(smp_processor_id());
A
Andrew Morton 已提交
4074
	/* Set up the next iteration */
4075 4076
	schedule_delayed_work(&__get_cpu_var(reap_work),
		round_jiffies_relative(REAPTIMEOUT_CPUC));
L
Linus Torvalds 已提交
4077 4078 4079 4080
}

#ifdef CONFIG_PROC_FS

4081
static void print_slabinfo_header(struct seq_file *m)
L
Linus Torvalds 已提交
4082
{
4083 4084 4085 4086
	/*
	 * Output format version, so at least we can change it
	 * without _too_ many complaints.
	 */
L
Linus Torvalds 已提交
4087
#if STATS
4088
	seq_puts(m, "slabinfo - version: 2.1 (statistics)\n");
L
Linus Torvalds 已提交
4089
#else
4090
	seq_puts(m, "slabinfo - version: 2.1\n");
L
Linus Torvalds 已提交
4091
#endif
4092 4093 4094 4095
	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>");
L
Linus Torvalds 已提交
4096
#if STATS
4097
	seq_puts(m, " : globalstat <listallocs> <maxobjs> <grown> <reaped> "
4098
		 "<error> <maxfreeable> <nodeallocs> <remotefrees> <alienoverflow>");
4099
	seq_puts(m, " : cpustat <allochit> <allocmiss> <freehit> <freemiss>");
L
Linus Torvalds 已提交
4100
#endif
4101 4102 4103 4104 4105 4106 4107 4108
	seq_putc(m, '\n');
}

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

I
Ingo Molnar 已提交
4109
	mutex_lock(&cache_chain_mutex);
4110 4111
	if (!n)
		print_slabinfo_header(m);
L
Linus Torvalds 已提交
4112 4113 4114 4115 4116 4117
	p = cache_chain.next;
	while (n--) {
		p = p->next;
		if (p == &cache_chain)
			return NULL;
	}
4118
	return list_entry(p, struct kmem_cache, next);
L
Linus Torvalds 已提交
4119 4120 4121 4122
}

static void *s_next(struct seq_file *m, void *p, loff_t *pos)
{
4123
	struct kmem_cache *cachep = p;
L
Linus Torvalds 已提交
4124
	++*pos;
A
Andrew Morton 已提交
4125 4126
	return cachep->next.next == &cache_chain ?
		NULL : list_entry(cachep->next.next, struct kmem_cache, next);
L
Linus Torvalds 已提交
4127 4128 4129 4130
}

static void s_stop(struct seq_file *m, void *p)
{
I
Ingo Molnar 已提交
4131
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
4132 4133 4134 4135
}

static int s_show(struct seq_file *m, void *p)
{
4136
	struct kmem_cache *cachep = p;
P
Pekka Enberg 已提交
4137 4138 4139 4140 4141
	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;
4142
	const char *name;
L
Linus Torvalds 已提交
4143
	char *error = NULL;
4144 4145
	int node;
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
4146 4147 4148

	active_objs = 0;
	num_slabs = 0;
4149 4150 4151 4152 4153
	for_each_online_node(node) {
		l3 = cachep->nodelists[node];
		if (!l3)
			continue;

4154 4155
		check_irq_on();
		spin_lock_irq(&l3->list_lock);
4156

4157
		list_for_each_entry(slabp, &l3->slabs_full, list) {
4158 4159 4160 4161 4162
			if (slabp->inuse != cachep->num && !error)
				error = "slabs_full accounting error";
			active_objs += cachep->num;
			active_slabs++;
		}
4163
		list_for_each_entry(slabp, &l3->slabs_partial, list) {
4164 4165 4166 4167 4168 4169 4170
			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++;
		}
4171
		list_for_each_entry(slabp, &l3->slabs_free, list) {
4172 4173 4174 4175 4176
			if (slabp->inuse && !error)
				error = "slabs_free/inuse accounting error";
			num_slabs++;
		}
		free_objects += l3->free_objects;
4177 4178
		if (l3->shared)
			shared_avail += l3->shared->avail;
4179

4180
		spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
4181
	}
P
Pekka Enberg 已提交
4182 4183
	num_slabs += active_slabs;
	num_objs = num_slabs * cachep->num;
4184
	if (num_objs - active_objs != free_objects && !error)
L
Linus Torvalds 已提交
4185 4186
		error = "free_objects accounting error";

P
Pekka Enberg 已提交
4187
	name = cachep->name;
L
Linus Torvalds 已提交
4188 4189 4190 4191
	if (error)
		printk(KERN_ERR "slab: cache %s error: %s\n", name, error);

	seq_printf(m, "%-17s %6lu %6lu %6u %4u %4d",
4192
		   name, active_objs, num_objs, cachep->buffer_size,
P
Pekka Enberg 已提交
4193
		   cachep->num, (1 << cachep->gfporder));
L
Linus Torvalds 已提交
4194
	seq_printf(m, " : tunables %4u %4u %4u",
P
Pekka Enberg 已提交
4195
		   cachep->limit, cachep->batchcount, cachep->shared);
4196
	seq_printf(m, " : slabdata %6lu %6lu %6lu",
P
Pekka Enberg 已提交
4197
		   active_slabs, num_slabs, shared_avail);
L
Linus Torvalds 已提交
4198
#if STATS
P
Pekka Enberg 已提交
4199
	{			/* list3 stats */
L
Linus Torvalds 已提交
4200 4201 4202 4203 4204 4205 4206
		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;
4207
		unsigned long node_frees = cachep->node_frees;
4208
		unsigned long overflows = cachep->node_overflow;
L
Linus Torvalds 已提交
4209

4210
		seq_printf(m, " : globalstat %7lu %6lu %5lu %4lu \
4211
				%4lu %4lu %4lu %4lu %4lu", allocs, high, grown,
A
Andrew Morton 已提交
4212
				reaped, errors, max_freeable, node_allocs,
4213
				node_frees, overflows);
L
Linus Torvalds 已提交
4214 4215 4216 4217 4218 4219 4220 4221 4222
	}
	/* 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",
P
Pekka Enberg 已提交
4223
			   allochit, allocmiss, freehit, freemiss);
L
Linus Torvalds 已提交
4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243
	}
#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
 */

4244
const struct seq_operations slabinfo_op = {
P
Pekka Enberg 已提交
4245 4246 4247 4248
	.start = s_start,
	.next = s_next,
	.stop = s_stop,
	.show = s_show,
L
Linus Torvalds 已提交
4249 4250 4251 4252 4253 4254 4255 4256 4257 4258
};

#define MAX_SLABINFO_WRITE 128
/**
 * slabinfo_write - Tuning for the slab allocator
 * @file: unused
 * @buffer: user buffer
 * @count: data length
 * @ppos: unused
 */
P
Pekka Enberg 已提交
4259 4260
ssize_t slabinfo_write(struct file *file, const char __user * buffer,
		       size_t count, loff_t *ppos)
L
Linus Torvalds 已提交
4261
{
P
Pekka Enberg 已提交
4262
	char kbuf[MAX_SLABINFO_WRITE + 1], *tmp;
L
Linus Torvalds 已提交
4263
	int limit, batchcount, shared, res;
4264
	struct kmem_cache *cachep;
P
Pekka Enberg 已提交
4265

L
Linus Torvalds 已提交
4266 4267 4268 4269
	if (count > MAX_SLABINFO_WRITE)
		return -EINVAL;
	if (copy_from_user(&kbuf, buffer, count))
		return -EFAULT;
P
Pekka Enberg 已提交
4270
	kbuf[MAX_SLABINFO_WRITE] = '\0';
L
Linus Torvalds 已提交
4271 4272 4273 4274 4275 4276 4277 4278 4279 4280

	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. */
I
Ingo Molnar 已提交
4281
	mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
4282
	res = -EINVAL;
4283
	list_for_each_entry(cachep, &cache_chain, next) {
L
Linus Torvalds 已提交
4284
		if (!strcmp(cachep->name, kbuf)) {
A
Andrew Morton 已提交
4285 4286
			if (limit < 1 || batchcount < 1 ||
					batchcount > limit || shared < 0) {
4287
				res = 0;
L
Linus Torvalds 已提交
4288
			} else {
4289
				res = do_tune_cpucache(cachep, limit,
P
Pekka Enberg 已提交
4290
						       batchcount, shared);
L
Linus Torvalds 已提交
4291 4292 4293 4294
			}
			break;
		}
	}
I
Ingo Molnar 已提交
4295
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
4296 4297 4298 4299
	if (res >= 0)
		res = count;
	return res;
}
4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408

#ifdef CONFIG_DEBUG_SLAB_LEAK

static void *leaks_start(struct seq_file *m, loff_t *pos)
{
	loff_t n = *pos;
	struct list_head *p;

	mutex_lock(&cache_chain_mutex);
	p = cache_chain.next;
	while (n--) {
		p = p->next;
		if (p == &cache_chain)
			return NULL;
	}
	return list_entry(p, struct kmem_cache, next);
}

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
	char *modname;
	const char *name;
	unsigned long offset, size;
	char namebuf[KSYM_NAME_LEN+1];

	name = kallsyms_lookup(address, &size, &offset, &modname, namebuf);

	if (name) {
		seq_printf(m, "%s+%#lx/%#lx", name, offset, size);
		if (modname)
			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 = p;
	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);

4409
		list_for_each_entry(slabp, &l3->slabs_full, list)
4410
			handle_slab(n, cachep, slabp);
4411
		list_for_each_entry(slabp, &l3->slabs_partial, list)
4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437
			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');
	}
4438

4439 4440 4441
	return 0;
}

4442
const struct seq_operations slabstats_op = {
4443 4444 4445 4446 4447 4448
	.start = leaks_start,
	.next = s_next,
	.stop = s_stop,
	.show = leaks_show,
};
#endif
L
Linus Torvalds 已提交
4449 4450
#endif

4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462
/**
 * 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.
 */
L
Linus Torvalds 已提交
4463 4464
unsigned int ksize(const void *objp)
{
4465 4466
	if (unlikely(objp == NULL))
		return 0;
L
Linus Torvalds 已提交
4467

4468
	return obj_size(virt_to_cache(objp));
L
Linus Torvalds 已提交
4469
}