slab.c 98.3 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 53 54 55 56 57
/*
 * 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.
 *
 * The c_cpuarray may not be read with enabled local interrupts - 
 * it's changed with a smp_call_function().
 *
 * SMP synchronization:
 *  constructors and destructors are called without any locking.
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 91 92 93 94 95 96 97 98 99 100 101 102 103
 */

#include	<linux/config.h>
#include	<linux/slab.h>
#include	<linux/mm.h>
#include	<linux/swap.h>
#include	<linux/cache.h>
#include	<linux/interrupt.h>
#include	<linux/init.h>
#include	<linux/compiler.h>
#include	<linux/seq_file.h>
#include	<linux/notifier.h>
#include	<linux/kallsyms.h>
#include	<linux/cpu.h>
#include	<linux/sysctl.h>
#include	<linux/module.h>
#include	<linux/rcupdate.h>
104
#include	<linux/string.h>
105
#include	<linux/nodemask.h>
106
#include	<linux/mempolicy.h>
I
Ingo Molnar 已提交
107
#include	<linux/mutex.h>
L
Linus Torvalds 已提交
108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202

#include	<asm/uaccess.h>
#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 | \
			 SLAB_NO_REAP | SLAB_CACHE_DMA | \
			 SLAB_MUST_HWCACHE_ALIGN | SLAB_STORE_USER | \
			 SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
			 SLAB_DESTROY_BY_RCU)
#else
# define CREATE_MASK	(SLAB_HWCACHE_ALIGN | SLAB_NO_REAP | \
			 SLAB_CACHE_DMA | SLAB_MUST_HWCACHE_ALIGN | \
			 SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
			 SLAB_DESTROY_BY_RCU)
#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.
 */

203
typedef unsigned int kmem_bufctl_t;
L
Linus Torvalds 已提交
204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220
#define BUFCTL_END	(((kmem_bufctl_t)(~0U))-0)
#define BUFCTL_FREE	(((kmem_bufctl_t)(~0U))-1)
#define	SLAB_LIMIT	(((kmem_bufctl_t)(~0U))-2)

/* Max number of objs-per-slab for caches which use off-slab slabs.
 * Needed to avoid a possible looping condition in cache_grow().
 */
static unsigned long offslab_limit;

/*
 * 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;
268 269 270 271 272 273 274
	spinlock_t lock;
	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
275 276 277 278 279 280 281 282
};

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

/*
287
 * The slab lists for all objects.
L
Linus Torvalds 已提交
288 289
 */
struct kmem_list3 {
P
Pekka Enberg 已提交
290 291 292 293 294 295 296 297 298 299
	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 long next_reap;
	int free_touched;
	unsigned int free_limit;
	spinlock_t list_lock;
	struct array_cache *shared;	/* shared per node */
	struct array_cache **alien;	/* on other nodes */
L
Linus Torvalds 已提交
300 301
};

302 303 304 305 306 307 308 309 310 311
/*
 * 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)

/*
312
 * This function must be completely optimized away if
313 314 315 316
 * a constant is passed to it. Mostly the same as
 * what is in linux/slab.h except it returns an
 * index.
 */
317
static __always_inline int index_of(const size_t size)
318
{
319 320
	extern void __bad_size(void);

321 322 323 324 325 326 327 328 329 330
	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
331
		__bad_size();
332
	} else
333
		__bad_size();
334 335 336 337 338
	return 0;
}

#define INDEX_AC index_of(sizeof(struct arraycache_init))
#define INDEX_L3 index_of(sizeof(struct kmem_list3))
L
Linus Torvalds 已提交
339

P
Pekka Enberg 已提交
340
static void kmem_list3_init(struct kmem_list3 *parent)
341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363
{
	INIT_LIST_HEAD(&parent->slabs_full);
	INIT_LIST_HEAD(&parent->slabs_partial);
	INIT_LIST_HEAD(&parent->slabs_free);
	parent->shared = NULL;
	parent->alien = NULL;
	spin_lock_init(&parent->list_lock);
	parent->free_objects = 0;
	parent->free_touched = 0;
}

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

#define	MAKE_ALL_LISTS(cachep, ptr, nodeid)			\
	do {					\
	MAKE_LIST((cachep), (&(ptr)->slabs_full), slabs_full, nodeid);	\
	MAKE_LIST((cachep), (&(ptr)->slabs_partial), slabs_partial, nodeid); \
	MAKE_LIST((cachep), (&(ptr)->slabs_free), slabs_free, nodeid);	\
	} while (0)
L
Linus Torvalds 已提交
364 365

/*
366
 * struct kmem_cache
L
Linus Torvalds 已提交
367 368 369
 *
 * manages a cache.
 */
P
Pekka Enberg 已提交
370

371
struct kmem_cache {
L
Linus Torvalds 已提交
372
/* 1) per-cpu data, touched during every alloc/free */
P
Pekka Enberg 已提交
373 374 375 376
	struct array_cache *array[NR_CPUS];
	unsigned int batchcount;
	unsigned int limit;
	unsigned int shared;
377
	unsigned int buffer_size;
378
/* 2) touched by every alloc & free from the backend */
P
Pekka Enberg 已提交
379 380 381 382
	struct kmem_list3 *nodelists[MAX_NUMNODES];
	unsigned int flags;	/* constant flags */
	unsigned int num;	/* # of objs per slab */
	spinlock_t spinlock;
L
Linus Torvalds 已提交
383 384 385

/* 3) cache_grow/shrink */
	/* order of pgs per slab (2^n) */
P
Pekka Enberg 已提交
386
	unsigned int gfporder;
L
Linus Torvalds 已提交
387 388

	/* force GFP flags, e.g. GFP_DMA */
P
Pekka Enberg 已提交
389
	gfp_t gfpflags;
L
Linus Torvalds 已提交
390

P
Pekka Enberg 已提交
391 392 393
	size_t colour;		/* cache colouring range */
	unsigned int colour_off;	/* colour offset */
	unsigned int colour_next;	/* cache colouring */
394
	struct kmem_cache *slabp_cache;
P
Pekka Enberg 已提交
395 396
	unsigned int slab_size;
	unsigned int dflags;	/* dynamic flags */
L
Linus Torvalds 已提交
397 398

	/* constructor func */
399
	void (*ctor) (void *, struct kmem_cache *, unsigned long);
L
Linus Torvalds 已提交
400 401

	/* de-constructor func */
402
	void (*dtor) (void *, struct kmem_cache *, unsigned long);
L
Linus Torvalds 已提交
403 404

/* 4) cache creation/removal */
P
Pekka Enberg 已提交
405 406
	const char *name;
	struct list_head next;
L
Linus Torvalds 已提交
407 408 409

/* 5) statistics */
#if STATS
P
Pekka Enberg 已提交
410 411 412 413 414 415 416 417 418 419 420 421 422
	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;
	atomic_t allochit;
	atomic_t allocmiss;
	atomic_t freehit;
	atomic_t freemiss;
L
Linus Torvalds 已提交
423 424
#endif
#if DEBUG
425 426 427 428 429 430 431 432
	/*
	 * 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
433 434 435 436 437 438 439 440 441 442
#endif
};

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

#define BATCHREFILL_LIMIT	16
/* Optimization question: fewer reaps means less 
 * probability for unnessary cpucache drain/refill cycles.
 *
A
Adrian Bunk 已提交
443
 * OTOH the cpuarrays can contain lots of objects,
L
Linus Torvalds 已提交
444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
 * which could lock up otherwise freeable slabs.
 */
#define REAPTIMEOUT_CPUC	(2*HZ)
#define REAPTIMEOUT_LIST3	(4*HZ)

#if STATS
#define	STATS_INC_ACTIVE(x)	((x)->num_active++)
#define	STATS_DEC_ACTIVE(x)	((x)->num_active--)
#define	STATS_INC_ALLOCED(x)	((x)->num_allocations++)
#define	STATS_INC_GROWN(x)	((x)->grown++)
#define	STATS_INC_REAPED(x)	((x)->reaped++)
#define	STATS_SET_HIGH(x)	do { if ((x)->num_active > (x)->high_mark) \
					(x)->high_mark = (x)->num_active; \
				} while (0)
#define	STATS_INC_ERR(x)	((x)->errors++)
#define	STATS_INC_NODEALLOCS(x)	((x)->node_allocs++)
460
#define	STATS_INC_NODEFREES(x)	((x)->node_frees++)
L
Linus Torvalds 已提交
461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478
#define	STATS_SET_FREEABLE(x, i) \
				do { if ((x)->max_freeable < i) \
					(x)->max_freeable = i; \
				} while (0)

#define STATS_INC_ALLOCHIT(x)	atomic_inc(&(x)->allochit)
#define STATS_INC_ALLOCMISS(x)	atomic_inc(&(x)->allocmiss)
#define STATS_INC_FREEHIT(x)	atomic_inc(&(x)->freehit)
#define STATS_INC_FREEMISS(x)	atomic_inc(&(x)->freemiss)
#else
#define	STATS_INC_ACTIVE(x)	do { } while (0)
#define	STATS_DEC_ACTIVE(x)	do { } while (0)
#define	STATS_INC_ALLOCED(x)	do { } while (0)
#define	STATS_INC_GROWN(x)	do { } while (0)
#define	STATS_INC_REAPED(x)	do { } while (0)
#define	STATS_SET_HIGH(x)	do { } while (0)
#define	STATS_INC_ERR(x)	do { } while (0)
#define	STATS_INC_NODEALLOCS(x)	do { } while (0)
479
#define	STATS_INC_NODEFREES(x)	do { } while (0)
L
Linus Torvalds 已提交
480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502
#define	STATS_SET_FREEABLE(x, i) \
				do { } while (0)

#define STATS_INC_ALLOCHIT(x)	do { } while (0)
#define STATS_INC_ALLOCMISS(x)	do { } while (0)
#define STATS_INC_FREEHIT(x)	do { } while (0)
#define STATS_INC_FREEMISS(x)	do { } while (0)
#endif

#if DEBUG
/* Magic nums for obj red zoning.
 * Placed in the first word before and the first word after an obj.
 */
#define	RED_INACTIVE	0x5A2CF071UL	/* when obj is inactive */
#define	RED_ACTIVE	0x170FC2A5UL	/* when obj is active */

/* ...and for poisoning */
#define	POISON_INUSE	0x5a	/* for use-uninitialised poisoning */
#define POISON_FREE	0x6b	/* for use-after-free poisoning */
#define	POISON_END	0xa5	/* end-byte of poisoning */

/* memory layout of objects:
 * 0		: objp
503
 * 0 .. cachep->obj_offset - BYTES_PER_WORD - 1: padding. This ensures that
L
Linus Torvalds 已提交
504 505
 * 		the end of an object is aligned with the end of the real
 * 		allocation. Catches writes behind the end of the allocation.
506
 * cachep->obj_offset - BYTES_PER_WORD .. cachep->obj_offset - 1:
L
Linus Torvalds 已提交
507
 * 		redzone word.
508 509 510
 * cachep->obj_offset: The real object.
 * cachep->buffer_size - 2* BYTES_PER_WORD: redzone word [BYTES_PER_WORD long]
 * cachep->buffer_size - 1* BYTES_PER_WORD: last caller address [BYTES_PER_WORD long]
L
Linus Torvalds 已提交
511
 */
512
static int obj_offset(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
513
{
514
	return cachep->obj_offset;
L
Linus Torvalds 已提交
515 516
}

517
static int obj_size(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
518
{
519
	return cachep->obj_size;
L
Linus Torvalds 已提交
520 521
}

522
static unsigned long *dbg_redzone1(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
523 524
{
	BUG_ON(!(cachep->flags & SLAB_RED_ZONE));
525
	return (unsigned long*) (objp+obj_offset(cachep)-BYTES_PER_WORD);
L
Linus Torvalds 已提交
526 527
}

528
static unsigned long *dbg_redzone2(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
529 530 531
{
	BUG_ON(!(cachep->flags & SLAB_RED_ZONE));
	if (cachep->flags & SLAB_STORE_USER)
532
		return (unsigned long *)(objp + cachep->buffer_size -
P
Pekka Enberg 已提交
533
					 2 * BYTES_PER_WORD);
534
	return (unsigned long *)(objp + cachep->buffer_size - BYTES_PER_WORD);
L
Linus Torvalds 已提交
535 536
}

537
static void **dbg_userword(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
538 539
{
	BUG_ON(!(cachep->flags & SLAB_STORE_USER));
540
	return (void **)(objp + cachep->buffer_size - BYTES_PER_WORD);
L
Linus Torvalds 已提交
541 542 543 544
}

#else

545 546
#define obj_offset(x)			0
#define obj_size(cachep)		(cachep->buffer_size)
L
Linus Torvalds 已提交
547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574
#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

/*
 * Maximum size of an obj (in 2^order pages)
 * and absolute limit for the gfp order.
 */
#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;

575
/* Functions for storing/retrieving the cachep and or slab from the
L
Linus Torvalds 已提交
576 577 578
 * global 'mem_map'. 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.
 */
579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597
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)
{
	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)
{
	return (struct slab *)page->lru.prev;
}
L
Linus Torvalds 已提交
598

599 600 601 602 603 604 605 606 607 608 609 610
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);
}

L
Linus Torvalds 已提交
611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628
/* These are the default caches for kmalloc. Custom caches can have other sizes. */
struct cache_sizes malloc_sizes[] = {
#define CACHE(x) { .cs_size = (x) },
#include <linux/kmalloc_sizes.h>
	CACHE(ULONG_MAX)
#undef CACHE
};
EXPORT_SYMBOL(malloc_sizes);

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

static struct cache_names __initdata cache_names[] = {
#define CACHE(x) { .name = "size-" #x, .name_dma = "size-" #x "(DMA)" },
#include <linux/kmalloc_sizes.h>
P
Pekka Enberg 已提交
629
	{NULL,}
L
Linus Torvalds 已提交
630 631 632 633
#undef CACHE
};

static struct arraycache_init initarray_cache __initdata =
P
Pekka Enberg 已提交
634
    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
L
Linus Torvalds 已提交
635
static struct arraycache_init initarray_generic =
P
Pekka Enberg 已提交
636
    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
L
Linus Torvalds 已提交
637 638

/* internal cache of cache description objs */
639
static struct kmem_cache cache_cache = {
P
Pekka Enberg 已提交
640 641 642
	.batchcount = 1,
	.limit = BOOT_CPUCACHE_ENTRIES,
	.shared = 1,
643
	.buffer_size = sizeof(struct kmem_cache),
P
Pekka Enberg 已提交
644 645 646
	.flags = SLAB_NO_REAP,
	.spinlock = SPIN_LOCK_UNLOCKED,
	.name = "kmem_cache",
L
Linus Torvalds 已提交
647
#if DEBUG
648
	.obj_size = sizeof(struct kmem_cache),
L
Linus Torvalds 已提交
649 650 651 652
#endif
};

/* Guard access to the cache-chain. */
I
Ingo Molnar 已提交
653
static DEFINE_MUTEX(cache_chain_mutex);
L
Linus Torvalds 已提交
654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669
static struct list_head cache_chain;

/*
 * vm_enough_memory() looks at this to determine how many
 * slab-allocated pages are possibly freeable under pressure
 *
 * SLAB_RECLAIM_ACCOUNT turns this on per-slab
 */
atomic_t slab_reclaim_pages;

/*
 * chicken and egg problem: delay the per-cpu array allocation
 * until the general caches are up.
 */
static enum {
	NONE,
670 671
	PARTIAL_AC,
	PARTIAL_L3,
L
Linus Torvalds 已提交
672 673 674 675 676
	FULL
} g_cpucache_up;

static DEFINE_PER_CPU(struct work_struct, reap_work);

677 678
static void free_block(struct kmem_cache *cachep, void **objpp, int len, int node);
static void enable_cpucache(struct kmem_cache *cachep);
P
Pekka Enberg 已提交
679
static void cache_reap(void *unused);
680
static int __node_shrink(struct kmem_cache *cachep, int node);
L
Linus Torvalds 已提交
681

682
static inline struct array_cache *cpu_cache_get(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
683 684 685 686
{
	return cachep->array[smp_processor_id()];
}

687
static inline struct kmem_cache *__find_general_cachep(size_t size, gfp_t gfpflags)
L
Linus Torvalds 已提交
688 689 690 691 692
{
	struct cache_sizes *csizep = malloc_sizes;

#if DEBUG
	/* This happens if someone tries to call
P
Pekka Enberg 已提交
693 694 695
	 * kmem_cache_create(), or __kmalloc(), before
	 * the generic caches are initialized.
	 */
696
	BUG_ON(malloc_sizes[INDEX_AC].cs_cachep == NULL);
L
Linus Torvalds 已提交
697 698 699 700 701
#endif
	while (size > csizep->cs_size)
		csizep++;

	/*
702
	 * Really subtle: The last entry with cs->cs_size==ULONG_MAX
L
Linus Torvalds 已提交
703 704 705 706 707 708 709 710
	 * 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;
}

711
struct kmem_cache *kmem_find_general_cachep(size_t size, gfp_t gfpflags)
712 713 714 715 716
{
	return __find_general_cachep(size, gfpflags);
}
EXPORT_SYMBOL(kmem_find_general_cachep);

717
static size_t slab_mgmt_size(size_t nr_objs, size_t align)
L
Linus Torvalds 已提交
718
{
719 720
	return ALIGN(sizeof(struct slab)+nr_objs*sizeof(kmem_bufctl_t), align);
}
L
Linus Torvalds 已提交
721

722 723 724 725 726 727 728 729 730
/* Calculate the number of objects and left-over bytes for a given
   buffer size. */
static void cache_estimate(unsigned long gfporder, size_t buffer_size,
			   size_t align, int flags, size_t *left_over,
			   unsigned int *num)
{
	int nr_objs;
	size_t mgmt_size;
	size_t slab_size = PAGE_SIZE << gfporder;
L
Linus Torvalds 已提交
731

732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779
	/*
	 * 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
780 781 782 783
}

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

784
static void __slab_error(const char *function, struct kmem_cache *cachep, char *msg)
L
Linus Torvalds 已提交
785 786
{
	printk(KERN_ERR "slab error in %s(): cache `%s': %s\n",
P
Pekka Enberg 已提交
787
	       function, cachep->name, msg);
L
Linus Torvalds 已提交
788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812
	dump_stack();
}

/*
 * 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)
{
	struct work_struct *reap_work = &per_cpu(reap_work, cpu);

	/*
	 * When this gets called from do_initcalls via cpucache_init(),
	 * init_workqueues() has already run, so keventd will be setup
	 * at that time.
	 */
	if (keventd_up() && reap_work->func == NULL) {
		INIT_WORK(reap_work, cache_reap, NULL);
		schedule_delayed_work_on(cpu, reap_work, HZ + 3 * cpu);
	}
}

813
static struct array_cache *alloc_arraycache(int node, int entries,
P
Pekka Enberg 已提交
814
					    int batchcount)
L
Linus Torvalds 已提交
815
{
P
Pekka Enberg 已提交
816
	int memsize = sizeof(void *) * entries + sizeof(struct array_cache);
L
Linus Torvalds 已提交
817 818
	struct array_cache *nc = NULL;

819
	nc = kmalloc_node(memsize, GFP_KERNEL, node);
L
Linus Torvalds 已提交
820 821 822 823 824
	if (nc) {
		nc->avail = 0;
		nc->limit = entries;
		nc->batchcount = batchcount;
		nc->touched = 0;
825
		spin_lock_init(&nc->lock);
L
Linus Torvalds 已提交
826 827 828 829
	}
	return nc;
}

830
#ifdef CONFIG_NUMA
831
static void *__cache_alloc_node(struct kmem_cache *, gfp_t, int);
832

P
Pekka Enberg 已提交
833
static struct array_cache **alloc_alien_cache(int node, int limit)
834 835
{
	struct array_cache **ac_ptr;
P
Pekka Enberg 已提交
836
	int memsize = sizeof(void *) * MAX_NUMNODES;
837 838 839 840 841 842 843 844 845 846 847 848 849
	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
850
				for (i--; i <= 0; i--)
851 852 853 854 855 856 857 858 859
					kfree(ac_ptr[i]);
				kfree(ac_ptr);
				return NULL;
			}
		}
	}
	return ac_ptr;
}

P
Pekka Enberg 已提交
860
static void free_alien_cache(struct array_cache **ac_ptr)
861 862 863 864 865 866 867
{
	int i;

	if (!ac_ptr)
		return;

	for_each_node(i)
P
Pekka Enberg 已提交
868
	    kfree(ac_ptr[i]);
869 870 871 872

	kfree(ac_ptr);
}

873
static void __drain_alien_cache(struct kmem_cache *cachep,
P
Pekka Enberg 已提交
874
				struct array_cache *ac, int node)
875 876 877 878 879
{
	struct kmem_list3 *rl3 = cachep->nodelists[node];

	if (ac->avail) {
		spin_lock(&rl3->list_lock);
880
		free_block(cachep, ac->entry, ac->avail, node);
881 882 883 884 885
		ac->avail = 0;
		spin_unlock(&rl3->list_lock);
	}
}

886
static void drain_alien_cache(struct kmem_cache *cachep, struct kmem_list3 *l3)
887
{
P
Pekka Enberg 已提交
888
	int i = 0;
889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906
	struct array_cache *ac;
	unsigned long flags;

	for_each_online_node(i) {
		ac = l3->alien[i];
		if (ac) {
			spin_lock_irqsave(&ac->lock, flags);
			__drain_alien_cache(cachep, ac, i);
			spin_unlock_irqrestore(&ac->lock, flags);
		}
	}
}
#else
#define alloc_alien_cache(node, limit) do { } while (0)
#define free_alien_cache(ac_ptr) do { } while (0)
#define drain_alien_cache(cachep, l3) do { } while (0)
#endif

L
Linus Torvalds 已提交
907
static int __devinit cpuup_callback(struct notifier_block *nfb,
P
Pekka Enberg 已提交
908
				    unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
909 910
{
	long cpu = (long)hcpu;
911
	struct kmem_cache *cachep;
912 913 914
	struct kmem_list3 *l3 = NULL;
	int node = cpu_to_node(cpu);
	int memsize = sizeof(struct kmem_list3);
L
Linus Torvalds 已提交
915 916 917

	switch (action) {
	case CPU_UP_PREPARE:
I
Ingo Molnar 已提交
918
		mutex_lock(&cache_chain_mutex);
919 920 921 922 923 924
		/* we need to do this right in the beginning since
		 * alloc_arraycache's are going to use this list.
		 * kmalloc_node allows us to add the slab to the right
		 * kmem_list3 and not this cpu's kmem_list3
		 */

L
Linus Torvalds 已提交
925
		list_for_each_entry(cachep, &cache_chain, next) {
926 927 928 929 930 931
			/* setup the size64 kmemlist for cpu before we can
			 * begin anything. Make sure some other cpu on this
			 * node has not already allocated this
			 */
			if (!cachep->nodelists[node]) {
				if (!(l3 = kmalloc_node(memsize,
P
Pekka Enberg 已提交
932
							GFP_KERNEL, node)))
933 934 935
					goto bad;
				kmem_list3_init(l3);
				l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
P
Pekka Enberg 已提交
936
				    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
937 938 939

				cachep->nodelists[node] = l3;
			}
L
Linus Torvalds 已提交
940

941 942
			spin_lock_irq(&cachep->nodelists[node]->list_lock);
			cachep->nodelists[node]->free_limit =
P
Pekka Enberg 已提交
943 944
			    (1 + nr_cpus_node(node)) *
			    cachep->batchcount + cachep->num;
945 946 947 948
			spin_unlock_irq(&cachep->nodelists[node]->list_lock);
		}

		/* Now we can go ahead with allocating the shared array's
P
Pekka Enberg 已提交
949
		   & array cache's */
950
		list_for_each_entry(cachep, &cache_chain, next) {
951 952
			struct array_cache *nc;

953
			nc = alloc_arraycache(node, cachep->limit,
P
Pekka Enberg 已提交
954
					      cachep->batchcount);
L
Linus Torvalds 已提交
955 956 957 958
			if (!nc)
				goto bad;
			cachep->array[cpu] = nc;

959 960 961 962
			l3 = cachep->nodelists[node];
			BUG_ON(!l3);
			if (!l3->shared) {
				if (!(nc = alloc_arraycache(node,
P
Pekka Enberg 已提交
963 964 965 966
							    cachep->shared *
							    cachep->batchcount,
							    0xbaadf00d)))
					goto bad;
967 968

				/* we are serialised from CPU_DEAD or
P
Pekka Enberg 已提交
969
				   CPU_UP_CANCELLED by the cpucontrol lock */
970 971
				l3->shared = nc;
			}
L
Linus Torvalds 已提交
972
		}
I
Ingo Molnar 已提交
973
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
974 975 976 977 978 979 980 981
		break;
	case CPU_ONLINE:
		start_cpu_timer(cpu);
		break;
#ifdef CONFIG_HOTPLUG_CPU
	case CPU_DEAD:
		/* fall thru */
	case CPU_UP_CANCELED:
I
Ingo Molnar 已提交
982
		mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
983 984 985

		list_for_each_entry(cachep, &cache_chain, next) {
			struct array_cache *nc;
986
			cpumask_t mask;
L
Linus Torvalds 已提交
987

988
			mask = node_to_cpumask(node);
L
Linus Torvalds 已提交
989 990 991 992
			spin_lock_irq(&cachep->spinlock);
			/* cpu is dead; no one can alloc from it. */
			nc = cachep->array[cpu];
			cachep->array[cpu] = NULL;
993 994 995 996 997 998 999 1000 1001 1002
			l3 = cachep->nodelists[node];

			if (!l3)
				goto unlock_cache;

			spin_lock(&l3->list_lock);

			/* Free limit for this kmem_list3 */
			l3->free_limit -= cachep->batchcount;
			if (nc)
1003
				free_block(cachep, nc->entry, nc->avail, node);
1004 1005

			if (!cpus_empty(mask)) {
P
Pekka Enberg 已提交
1006 1007 1008
				spin_unlock(&l3->list_lock);
				goto unlock_cache;
			}
1009 1010 1011

			if (l3->shared) {
				free_block(cachep, l3->shared->entry,
P
Pekka Enberg 已提交
1012
					   l3->shared->avail, node);
1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029
				kfree(l3->shared);
				l3->shared = NULL;
			}
			if (l3->alien) {
				drain_alien_cache(cachep, l3);
				free_alien_cache(l3->alien);
				l3->alien = NULL;
			}

			/* free slabs belonging to this node */
			if (__node_shrink(cachep, node)) {
				cachep->nodelists[node] = NULL;
				spin_unlock(&l3->list_lock);
				kfree(l3);
			} else {
				spin_unlock(&l3->list_lock);
			}
P
Pekka Enberg 已提交
1030
		      unlock_cache:
L
Linus Torvalds 已提交
1031 1032 1033
			spin_unlock_irq(&cachep->spinlock);
			kfree(nc);
		}
I
Ingo Molnar 已提交
1034
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1035 1036 1037 1038
		break;
#endif
	}
	return NOTIFY_OK;
P
Pekka Enberg 已提交
1039
      bad:
I
Ingo Molnar 已提交
1040
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1041 1042 1043 1044 1045
	return NOTIFY_BAD;
}

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

1046 1047 1048
/*
 * swap the static kmem_list3 with kmalloced memory
 */
1049
static void init_list(struct kmem_cache *cachep, struct kmem_list3 *list, int nodeid)
1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063
{
	struct kmem_list3 *ptr;

	BUG_ON(cachep->nodelists[nodeid] != list);
	ptr = kmalloc_node(sizeof(struct kmem_list3), GFP_KERNEL, nodeid);
	BUG_ON(!ptr);

	local_irq_disable();
	memcpy(ptr, list, sizeof(struct kmem_list3));
	MAKE_ALL_LISTS(cachep, ptr, nodeid);
	cachep->nodelists[nodeid] = ptr;
	local_irq_enable();
}

L
Linus Torvalds 已提交
1064 1065 1066 1067 1068 1069 1070 1071
/* Initialisation.
 * Called after the gfp() functions have been enabled, and before smp_init().
 */
void __init kmem_cache_init(void)
{
	size_t left_over;
	struct cache_sizes *sizes;
	struct cache_names *names;
1072 1073 1074 1075 1076 1077 1078
	int i;

	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
1079 1080 1081 1082 1083 1084 1085 1086 1087 1088

	/*
	 * 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:
1089
	 * 1) initialize the cache_cache cache: it contains the struct kmem_cache
L
Linus Torvalds 已提交
1090 1091
	 *    structures of all caches, except cache_cache itself: cache_cache
	 *    is statically allocated.
1092 1093 1094
	 *    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
1095
	 * 2) Create the first kmalloc cache.
1096
	 *    The struct kmem_cache for the new cache is allocated normally.
1097 1098 1099
	 *    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
1100 1101
	 * 4) Replace the __init data head arrays for cache_cache and the first
	 *    kmalloc cache with kmalloc allocated arrays.
1102 1103 1104
	 * 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
1105 1106 1107 1108 1109 1110 1111
	 */

	/* 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;
1112
	cache_cache.nodelists[numa_node_id()] = &initkmem_list3[CACHE_CACHE];
L
Linus Torvalds 已提交
1113

1114
	cache_cache.buffer_size = ALIGN(cache_cache.buffer_size, cache_line_size());
L
Linus Torvalds 已提交
1115

1116
	cache_estimate(0, cache_cache.buffer_size, cache_line_size(), 0,
P
Pekka Enberg 已提交
1117
		       &left_over, &cache_cache.num);
L
Linus Torvalds 已提交
1118 1119 1120
	if (!cache_cache.num)
		BUG();

P
Pekka Enberg 已提交
1121
	cache_cache.colour = left_over / cache_cache.colour_off;
L
Linus Torvalds 已提交
1122
	cache_cache.colour_next = 0;
P
Pekka Enberg 已提交
1123 1124
	cache_cache.slab_size = ALIGN(cache_cache.num * sizeof(kmem_bufctl_t) +
				      sizeof(struct slab), cache_line_size());
L
Linus Torvalds 已提交
1125 1126 1127 1128 1129

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

1130 1131 1132 1133 1134 1135
	/* Initialize the caches that provide memory for the array cache
	 * and the kmem_list3 structures first.
	 * Without this, further allocations will bug
	 */

	sizes[INDEX_AC].cs_cachep = kmem_cache_create(names[INDEX_AC].name,
P
Pekka Enberg 已提交
1136 1137 1138 1139
						      sizes[INDEX_AC].cs_size,
						      ARCH_KMALLOC_MINALIGN,
						      (ARCH_KMALLOC_FLAGS |
						       SLAB_PANIC), NULL, NULL);
1140 1141 1142

	if (INDEX_AC != INDEX_L3)
		sizes[INDEX_L3].cs_cachep =
P
Pekka Enberg 已提交
1143 1144 1145 1146 1147
		    kmem_cache_create(names[INDEX_L3].name,
				      sizes[INDEX_L3].cs_size,
				      ARCH_KMALLOC_MINALIGN,
				      (ARCH_KMALLOC_FLAGS | SLAB_PANIC), NULL,
				      NULL);
1148

L
Linus Torvalds 已提交
1149
	while (sizes->cs_size != ULONG_MAX) {
1150 1151
		/*
		 * For performance, all the general caches are L1 aligned.
L
Linus Torvalds 已提交
1152 1153 1154
		 * This should be particularly beneficial on SMP boxes, as it
		 * eliminates "false sharing".
		 * Note for systems short on memory removing the alignment will
1155 1156
		 * allow tighter packing of the smaller caches.
		 */
P
Pekka Enberg 已提交
1157
		if (!sizes->cs_cachep)
1158
			sizes->cs_cachep = kmem_cache_create(names->name,
P
Pekka Enberg 已提交
1159 1160 1161 1162 1163
							     sizes->cs_size,
							     ARCH_KMALLOC_MINALIGN,
							     (ARCH_KMALLOC_FLAGS
							      | SLAB_PANIC),
							     NULL, NULL);
L
Linus Torvalds 已提交
1164 1165 1166

		/* Inc off-slab bufctl limit until the ceiling is hit. */
		if (!(OFF_SLAB(sizes->cs_cachep))) {
P
Pekka Enberg 已提交
1167
			offslab_limit = sizes->cs_size - sizeof(struct slab);
L
Linus Torvalds 已提交
1168 1169 1170 1171
			offslab_limit /= sizeof(kmem_bufctl_t);
		}

		sizes->cs_dmacachep = kmem_cache_create(names->name_dma,
P
Pekka Enberg 已提交
1172 1173 1174 1175 1176 1177
							sizes->cs_size,
							ARCH_KMALLOC_MINALIGN,
							(ARCH_KMALLOC_FLAGS |
							 SLAB_CACHE_DMA |
							 SLAB_PANIC), NULL,
							NULL);
L
Linus Torvalds 已提交
1178 1179 1180 1181 1182 1183

		sizes++;
		names++;
	}
	/* 4) Replace the bootstrap head arrays */
	{
P
Pekka Enberg 已提交
1184
		void *ptr;
1185

L
Linus Torvalds 已提交
1186
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
1187

L
Linus Torvalds 已提交
1188
		local_irq_disable();
1189 1190
		BUG_ON(cpu_cache_get(&cache_cache) != &initarray_cache.cache);
		memcpy(ptr, cpu_cache_get(&cache_cache),
P
Pekka Enberg 已提交
1191
		       sizeof(struct arraycache_init));
L
Linus Torvalds 已提交
1192 1193
		cache_cache.array[smp_processor_id()] = ptr;
		local_irq_enable();
1194

L
Linus Torvalds 已提交
1195
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
1196

L
Linus Torvalds 已提交
1197
		local_irq_disable();
1198
		BUG_ON(cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep)
P
Pekka Enberg 已提交
1199
		       != &initarray_generic.cache);
1200
		memcpy(ptr, cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep),
P
Pekka Enberg 已提交
1201
		       sizeof(struct arraycache_init));
1202
		malloc_sizes[INDEX_AC].cs_cachep->array[smp_processor_id()] =
P
Pekka Enberg 已提交
1203
		    ptr;
L
Linus Torvalds 已提交
1204 1205
		local_irq_enable();
	}
1206 1207 1208 1209 1210
	/* 5) Replace the bootstrap kmem_list3's */
	{
		int node;
		/* Replace the static kmem_list3 structures for the boot cpu */
		init_list(&cache_cache, &initkmem_list3[CACHE_CACHE],
P
Pekka Enberg 已提交
1211
			  numa_node_id());
1212 1213 1214

		for_each_online_node(node) {
			init_list(malloc_sizes[INDEX_AC].cs_cachep,
P
Pekka Enberg 已提交
1215
				  &initkmem_list3[SIZE_AC + node], node);
1216 1217 1218

			if (INDEX_AC != INDEX_L3) {
				init_list(malloc_sizes[INDEX_L3].cs_cachep,
P
Pekka Enberg 已提交
1219 1220
					  &initkmem_list3[SIZE_L3 + node],
					  node);
1221 1222 1223
			}
		}
	}
L
Linus Torvalds 已提交
1224

1225
	/* 6) resize the head arrays to their final sizes */
L
Linus Torvalds 已提交
1226
	{
1227
		struct kmem_cache *cachep;
I
Ingo Molnar 已提交
1228
		mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1229
		list_for_each_entry(cachep, &cache_chain, next)
P
Pekka Enberg 已提交
1230
		    enable_cpucache(cachep);
I
Ingo Molnar 已提交
1231
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1232 1233 1234 1235 1236 1237
	}

	/* Done! */
	g_cpucache_up = FULL;

	/* Register a cpu startup notifier callback
1238
	 * that initializes cpu_cache_get for all new cpus
L
Linus Torvalds 已提交
1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254
	 */
	register_cpu_notifier(&cpucache_notifier);

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

static int __init cpucache_init(void)
{
	int cpu;

	/* 
	 * Register the timers that return unneeded
	 * pages to gfp.
	 */
1255
	for_each_online_cpu(cpu)
P
Pekka Enberg 已提交
1256
	    start_cpu_timer(cpu);
L
Linus Torvalds 已提交
1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269

	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.
 */
1270
static void *kmem_getpages(struct kmem_cache *cachep, gfp_t flags, int nodeid)
L
Linus Torvalds 已提交
1271 1272 1273 1274 1275 1276
{
	struct page *page;
	void *addr;
	int i;

	flags |= cachep->gfpflags;
1277
	page = alloc_pages_node(nodeid, flags, cachep->gfporder);
L
Linus Torvalds 已提交
1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295
	if (!page)
		return NULL;
	addr = page_address(page);

	i = (1 << cachep->gfporder);
	if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
		atomic_add(i, &slab_reclaim_pages);
	add_page_state(nr_slab, i);
	while (i--) {
		SetPageSlab(page);
		page++;
	}
	return addr;
}

/*
 * Interface to system's page release.
 */
1296
static void kmem_freepages(struct kmem_cache *cachep, void *addr)
L
Linus Torvalds 已提交
1297
{
P
Pekka Enberg 已提交
1298
	unsigned long i = (1 << cachep->gfporder);
L
Linus Torvalds 已提交
1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310
	struct page *page = virt_to_page(addr);
	const unsigned long nr_freed = i;

	while (i--) {
		if (!TestClearPageSlab(page))
			BUG();
		page++;
	}
	sub_page_state(nr_slab, nr_freed);
	if (current->reclaim_state)
		current->reclaim_state->reclaimed_slab += nr_freed;
	free_pages((unsigned long)addr, cachep->gfporder);
P
Pekka Enberg 已提交
1311 1312
	if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
		atomic_sub(1 << cachep->gfporder, &slab_reclaim_pages);
L
Linus Torvalds 已提交
1313 1314 1315 1316
}

static void kmem_rcu_free(struct rcu_head *head)
{
P
Pekka Enberg 已提交
1317
	struct slab_rcu *slab_rcu = (struct slab_rcu *)head;
1318
	struct kmem_cache *cachep = slab_rcu->cachep;
L
Linus Torvalds 已提交
1319 1320 1321 1322 1323 1324 1325 1326 1327

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

#if DEBUG

#ifdef CONFIG_DEBUG_PAGEALLOC
1328
static void store_stackinfo(struct kmem_cache *cachep, unsigned long *addr,
P
Pekka Enberg 已提交
1329
			    unsigned long caller)
L
Linus Torvalds 已提交
1330
{
1331
	int size = obj_size(cachep);
L
Linus Torvalds 已提交
1332

1333
	addr = (unsigned long *)&((char *)addr)[obj_offset(cachep)];
L
Linus Torvalds 已提交
1334

P
Pekka Enberg 已提交
1335
	if (size < 5 * sizeof(unsigned long))
L
Linus Torvalds 已提交
1336 1337
		return;

P
Pekka Enberg 已提交
1338 1339 1340 1341
	*addr++ = 0x12345678;
	*addr++ = caller;
	*addr++ = smp_processor_id();
	size -= 3 * sizeof(unsigned long);
L
Linus Torvalds 已提交
1342 1343 1344 1345 1346 1347 1348
	{
		unsigned long *sptr = &caller;
		unsigned long svalue;

		while (!kstack_end(sptr)) {
			svalue = *sptr++;
			if (kernel_text_address(svalue)) {
P
Pekka Enberg 已提交
1349
				*addr++ = svalue;
L
Linus Torvalds 已提交
1350 1351 1352 1353 1354 1355 1356
				size -= sizeof(unsigned long);
				if (size <= sizeof(unsigned long))
					break;
			}
		}

	}
P
Pekka Enberg 已提交
1357
	*addr++ = 0x87654321;
L
Linus Torvalds 已提交
1358 1359 1360
}
#endif

1361
static void poison_obj(struct kmem_cache *cachep, void *addr, unsigned char val)
L
Linus Torvalds 已提交
1362
{
1363 1364
	int size = obj_size(cachep);
	addr = &((char *)addr)[obj_offset(cachep)];
L
Linus Torvalds 已提交
1365 1366

	memset(addr, val, size);
P
Pekka Enberg 已提交
1367
	*(unsigned char *)(addr + size - 1) = POISON_END;
L
Linus Torvalds 已提交
1368 1369 1370 1371 1372 1373
}

static void dump_line(char *data, int offset, int limit)
{
	int i;
	printk(KERN_ERR "%03x:", offset);
P
Pekka Enberg 已提交
1374 1375
	for (i = 0; i < limit; i++) {
		printk(" %02x", (unsigned char)data[offset + i]);
L
Linus Torvalds 已提交
1376 1377 1378 1379 1380 1381 1382
	}
	printk("\n");
}
#endif

#if DEBUG

1383
static void print_objinfo(struct kmem_cache *cachep, void *objp, int lines)
L
Linus Torvalds 已提交
1384 1385 1386 1387 1388 1389
{
	int i, size;
	char *realobj;

	if (cachep->flags & SLAB_RED_ZONE) {
		printk(KERN_ERR "Redzone: 0x%lx/0x%lx.\n",
P
Pekka Enberg 已提交
1390 1391
		       *dbg_redzone1(cachep, objp),
		       *dbg_redzone2(cachep, objp));
L
Linus Torvalds 已提交
1392 1393 1394 1395
	}

	if (cachep->flags & SLAB_STORE_USER) {
		printk(KERN_ERR "Last user: [<%p>]",
P
Pekka Enberg 已提交
1396
		       *dbg_userword(cachep, objp));
L
Linus Torvalds 已提交
1397
		print_symbol("(%s)",
P
Pekka Enberg 已提交
1398
			     (unsigned long)*dbg_userword(cachep, objp));
L
Linus Torvalds 已提交
1399 1400
		printk("\n");
	}
1401 1402
	realobj = (char *)objp + obj_offset(cachep);
	size = obj_size(cachep);
P
Pekka Enberg 已提交
1403
	for (i = 0; i < size && lines; i += 16, lines--) {
L
Linus Torvalds 已提交
1404 1405
		int limit;
		limit = 16;
P
Pekka Enberg 已提交
1406 1407
		if (i + limit > size)
			limit = size - i;
L
Linus Torvalds 已提交
1408 1409 1410 1411
		dump_line(realobj, i, limit);
	}
}

1412
static void check_poison_obj(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
1413 1414 1415 1416 1417
{
	char *realobj;
	int size, i;
	int lines = 0;

1418 1419
	realobj = (char *)objp + obj_offset(cachep);
	size = obj_size(cachep);
L
Linus Torvalds 已提交
1420

P
Pekka Enberg 已提交
1421
	for (i = 0; i < size; i++) {
L
Linus Torvalds 已提交
1422
		char exp = POISON_FREE;
P
Pekka Enberg 已提交
1423
		if (i == size - 1)
L
Linus Torvalds 已提交
1424 1425 1426 1427 1428 1429
			exp = POISON_END;
		if (realobj[i] != exp) {
			int limit;
			/* Mismatch ! */
			/* Print header */
			if (lines == 0) {
P
Pekka Enberg 已提交
1430 1431 1432
				printk(KERN_ERR
				       "Slab corruption: start=%p, len=%d\n",
				       realobj, size);
L
Linus Torvalds 已提交
1433 1434 1435
				print_objinfo(cachep, objp, 0);
			}
			/* Hexdump the affected line */
P
Pekka Enberg 已提交
1436
			i = (i / 16) * 16;
L
Linus Torvalds 已提交
1437
			limit = 16;
P
Pekka Enberg 已提交
1438 1439
			if (i + limit > size)
				limit = size - i;
L
Linus Torvalds 已提交
1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451
			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:
		 */
1452
		struct slab *slabp = virt_to_slab(objp);
L
Linus Torvalds 已提交
1453 1454
		int objnr;

1455
		objnr = (unsigned)(objp - slabp->s_mem) / cachep->buffer_size;
L
Linus Torvalds 已提交
1456
		if (objnr) {
1457 1458
			objp = slabp->s_mem + (objnr - 1) * cachep->buffer_size;
			realobj = (char *)objp + obj_offset(cachep);
L
Linus Torvalds 已提交
1459
			printk(KERN_ERR "Prev obj: start=%p, len=%d\n",
P
Pekka Enberg 已提交
1460
			       realobj, size);
L
Linus Torvalds 已提交
1461 1462
			print_objinfo(cachep, objp, 2);
		}
P
Pekka Enberg 已提交
1463
		if (objnr + 1 < cachep->num) {
1464 1465
			objp = slabp->s_mem + (objnr + 1) * cachep->buffer_size;
			realobj = (char *)objp + obj_offset(cachep);
L
Linus Torvalds 已提交
1466
			printk(KERN_ERR "Next obj: start=%p, len=%d\n",
P
Pekka Enberg 已提交
1467
			       realobj, size);
L
Linus Torvalds 已提交
1468 1469 1470 1471 1472 1473
			print_objinfo(cachep, objp, 2);
		}
	}
}
#endif

1474 1475 1476 1477
#if DEBUG
/**
 * slab_destroy_objs - call the registered destructor for each object in
 *      a slab that is to be destroyed.
L
Linus Torvalds 已提交
1478
 */
1479
static void slab_destroy_objs(struct kmem_cache *cachep, struct slab *slabp)
L
Linus Torvalds 已提交
1480 1481 1482
{
	int i;
	for (i = 0; i < cachep->num; i++) {
1483
		void *objp = slabp->s_mem + cachep->buffer_size * i;
L
Linus Torvalds 已提交
1484 1485 1486

		if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
1487
			if ((cachep->buffer_size % PAGE_SIZE) == 0
P
Pekka Enberg 已提交
1488 1489
			    && OFF_SLAB(cachep))
				kernel_map_pages(virt_to_page(objp),
1490
						 cachep->buffer_size / PAGE_SIZE,
P
Pekka Enberg 已提交
1491
						 1);
L
Linus Torvalds 已提交
1492 1493 1494 1495 1496 1497 1498 1499 1500
			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
1501
					   "was overwritten");
L
Linus Torvalds 已提交
1502 1503
			if (*dbg_redzone2(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "end of a freed object "
P
Pekka Enberg 已提交
1504
					   "was overwritten");
L
Linus Torvalds 已提交
1505 1506
		}
		if (cachep->dtor && !(cachep->flags & SLAB_POISON))
1507
			(cachep->dtor) (objp + obj_offset(cachep), cachep, 0);
L
Linus Torvalds 已提交
1508
	}
1509
}
L
Linus Torvalds 已提交
1510
#else
1511
static void slab_destroy_objs(struct kmem_cache *cachep, struct slab *slabp)
1512
{
L
Linus Torvalds 已提交
1513 1514 1515
	if (cachep->dtor) {
		int i;
		for (i = 0; i < cachep->num; i++) {
1516
			void *objp = slabp->s_mem + cachep->buffer_size * i;
P
Pekka Enberg 已提交
1517
			(cachep->dtor) (objp, cachep, 0);
L
Linus Torvalds 已提交
1518 1519
		}
	}
1520
}
L
Linus Torvalds 已提交
1521 1522
#endif

1523 1524 1525 1526 1527
/**
 * Destroy all the objs in a slab, and release the mem back to the system.
 * Before calling the slab must have been unlinked from the cache.
 * The cache-lock is not held/needed.
 */
1528
static void slab_destroy(struct kmem_cache *cachep, struct slab *slabp)
1529 1530 1531 1532
{
	void *addr = slabp->s_mem - slabp->colouroff;

	slab_destroy_objs(cachep, slabp);
L
Linus Torvalds 已提交
1533 1534 1535
	if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU)) {
		struct slab_rcu *slab_rcu;

P
Pekka Enberg 已提交
1536
		slab_rcu = (struct slab_rcu *)slabp;
L
Linus Torvalds 已提交
1537 1538 1539 1540 1541 1542 1543 1544 1545 1546
		slab_rcu->cachep = cachep;
		slab_rcu->addr = addr;
		call_rcu(&slab_rcu->head, kmem_rcu_free);
	} else {
		kmem_freepages(cachep, addr);
		if (OFF_SLAB(cachep))
			kmem_cache_free(cachep->slabp_cache, slabp);
	}
}

1547
/* For setting up all the kmem_list3s for cache whose buffer_size is same
1548
   as size of kmem_list3. */
1549
static void set_up_list3s(struct kmem_cache *cachep, int index)
1550 1551 1552 1553
{
	int node;

	for_each_online_node(node) {
P
Pekka Enberg 已提交
1554
		cachep->nodelists[node] = &initkmem_list3[index + node];
1555
		cachep->nodelists[node]->next_reap = jiffies +
P
Pekka Enberg 已提交
1556 1557
		    REAPTIMEOUT_LIST3 +
		    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
1558 1559 1560
	}
}

1561 1562 1563 1564 1565 1566 1567 1568
/**
 * calculate_slab_order - calculate size (page order) of slabs and the number
 *                        of objects per slab.
 *
 * This could be made much more intelligent.  For now, try to avoid using
 * high order pages for slabs.  When the gfp() functions are more friendly
 * towards high-order requests, this should be changed.
 */
1569
static inline size_t calculate_slab_order(struct kmem_cache *cachep, size_t size,
1570 1571 1572 1573
					  size_t align, gfp_t flags)
{
	size_t left_over = 0;

P
Pekka Enberg 已提交
1574
	for (;; cachep->gfporder++) {
1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607
		unsigned int num;
		size_t remainder;

		if (cachep->gfporder > MAX_GFP_ORDER) {
			cachep->num = 0;
			break;
		}

		cache_estimate(cachep->gfporder, size, align, flags,
			       &remainder, &num);
		if (!num)
			continue;
		/* More than offslab_limit objects will cause problems */
		if (flags & CFLGS_OFF_SLAB && cachep->num > offslab_limit)
			break;

		cachep->num = num;
		left_over = remainder;

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

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

L
Linus Torvalds 已提交
1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640
/**
 * 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
 * the module calling this has to destroy the cache before getting 
 * unloaded.
 * 
 * The flags are
 *
 * %SLAB_POISON - Poison the slab with a known test pattern (a5a5a5a5)
 * to catch references to uninitialised memory.
 *
 * %SLAB_RED_ZONE - Insert `Red' zones around the allocated memory to check
 * for buffer overruns.
 *
 * %SLAB_NO_REAP - Don't automatically reap this cache when we're under
 * memory pressure.
 *
 * %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.
 */
1641
struct kmem_cache *
L
Linus Torvalds 已提交
1642
kmem_cache_create (const char *name, size_t size, size_t align,
1643 1644
	unsigned long flags, void (*ctor)(void*, struct kmem_cache *, unsigned long),
	void (*dtor)(void*, struct kmem_cache *, unsigned long))
L
Linus Torvalds 已提交
1645 1646
{
	size_t left_over, slab_size, ralign;
1647
	struct kmem_cache *cachep = NULL;
1648
	struct list_head *p;
L
Linus Torvalds 已提交
1649 1650 1651 1652 1653

	/*
	 * Sanity checks... these are all serious usage bugs.
	 */
	if ((!name) ||
P
Pekka Enberg 已提交
1654 1655 1656 1657 1658 1659 1660
	    in_interrupt() ||
	    (size < BYTES_PER_WORD) ||
	    (size > (1 << MAX_OBJ_ORDER) * PAGE_SIZE) || (dtor && !ctor)) {
		printk(KERN_ERR "%s: Early error in slab %s\n",
		       __FUNCTION__, name);
		BUG();
	}
L
Linus Torvalds 已提交
1661

I
Ingo Molnar 已提交
1662
	mutex_lock(&cache_chain_mutex);
1663 1664

	list_for_each(p, &cache_chain) {
1665
		struct kmem_cache *pc = list_entry(p, struct kmem_cache, next);
1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679
		mm_segment_t old_fs = get_fs();
		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.
		 */
		set_fs(KERNEL_DS);
		res = __get_user(tmp, pc->name);
		set_fs(old_fs);
		if (res) {
			printk("SLAB: cache with size %d has lost its name\n",
1680
			       pc->buffer_size);
1681 1682 1683
			continue;
		}

P
Pekka Enberg 已提交
1684
		if (!strcmp(pc->name, name)) {
1685 1686 1687 1688 1689 1690
			printk("kmem_cache_create: duplicate cache %s\n", name);
			dump_stack();
			goto oops;
		}
	}

L
Linus Torvalds 已提交
1691 1692 1693 1694 1695
#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
1696
		       "requested - %s\n", __FUNCTION__, name);
L
Linus Torvalds 已提交
1697 1698 1699 1700 1701 1702 1703 1704 1705
		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.
	 */
P
Pekka Enberg 已提交
1706 1707 1708
	if ((size < 4096
	     || fls(size - 1) == fls(size - 1 + 3 * BYTES_PER_WORD)))
		flags |= SLAB_RED_ZONE | SLAB_STORE_USER;
L
Linus Torvalds 已提交
1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728
	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);

	/*
	 * Always checks flags, a caller might be expecting debug
	 * support which isn't available.
	 */
	if (flags & ~CREATE_MASK)
		BUG();

	/* Check that size is in terms of words.  This is needed to avoid
	 * unaligned accesses for some archs when redzoning is used, and makes
	 * sure any on-slab bufctl's are also correctly aligned.
	 */
P
Pekka Enberg 已提交
1729 1730 1731
	if (size & (BYTES_PER_WORD - 1)) {
		size += (BYTES_PER_WORD - 1);
		size &= ~(BYTES_PER_WORD - 1);
L
Linus Torvalds 已提交
1732 1733 1734 1735 1736 1737 1738 1739 1740 1741
	}

	/* calculate out the final buffer alignment: */
	/* 1) arch recommendation: can be overridden for debug */
	if (flags & SLAB_HWCACHE_ALIGN) {
		/* Default alignment: as specified by the arch code.
		 * Except if an object is really small, then squeeze multiple
		 * objects into one cacheline.
		 */
		ralign = cache_line_size();
P
Pekka Enberg 已提交
1742
		while (size <= ralign / 2)
L
Linus Torvalds 已提交
1743 1744 1745 1746 1747 1748 1749 1750
			ralign /= 2;
	} else {
		ralign = BYTES_PER_WORD;
	}
	/* 2) arch mandated alignment: disables debug if necessary */
	if (ralign < ARCH_SLAB_MINALIGN) {
		ralign = ARCH_SLAB_MINALIGN;
		if (ralign > BYTES_PER_WORD)
P
Pekka Enberg 已提交
1751
			flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
L
Linus Torvalds 已提交
1752 1753 1754 1755 1756
	}
	/* 3) caller mandated alignment: disables debug if necessary */
	if (ralign < align) {
		ralign = align;
		if (ralign > BYTES_PER_WORD)
P
Pekka Enberg 已提交
1757
			flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
L
Linus Torvalds 已提交
1758 1759 1760 1761 1762 1763 1764
	}
	/* 4) Store it. Note that the debug code below can reduce
	 *    the alignment to BYTES_PER_WORD.
	 */
	align = ralign;

	/* Get cache's description obj. */
1765
	cachep = kmem_cache_alloc(&cache_cache, SLAB_KERNEL);
L
Linus Torvalds 已提交
1766
	if (!cachep)
1767
		goto oops;
1768
	memset(cachep, 0, sizeof(struct kmem_cache));
L
Linus Torvalds 已提交
1769 1770

#if DEBUG
1771
	cachep->obj_size = size;
L
Linus Torvalds 已提交
1772 1773 1774 1775 1776 1777

	if (flags & SLAB_RED_ZONE) {
		/* redzoning only works with word aligned caches */
		align = BYTES_PER_WORD;

		/* add space for red zone words */
1778
		cachep->obj_offset += BYTES_PER_WORD;
P
Pekka Enberg 已提交
1779
		size += 2 * BYTES_PER_WORD;
L
Linus Torvalds 已提交
1780 1781 1782 1783 1784 1785 1786 1787 1788 1789
	}
	if (flags & SLAB_STORE_USER) {
		/* user store requires word alignment and
		 * one word storage behind the end of the real
		 * object.
		 */
		align = BYTES_PER_WORD;
		size += BYTES_PER_WORD;
	}
#if FORCED_DEBUG && defined(CONFIG_DEBUG_PAGEALLOC)
P
Pekka Enberg 已提交
1790
	if (size >= malloc_sizes[INDEX_L3 + 1].cs_size
1791 1792
	    && cachep->obj_size > cache_line_size() && size < PAGE_SIZE) {
		cachep->obj_offset += PAGE_SIZE - size;
L
Linus Torvalds 已提交
1793 1794 1795 1796 1797 1798
		size = PAGE_SIZE;
	}
#endif
#endif

	/* Determine if the slab management is 'on' or 'off' slab. */
P
Pekka Enberg 已提交
1799
	if (size >= (PAGE_SIZE >> 3))
L
Linus Torvalds 已提交
1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815
		/*
		 * 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);

	if ((flags & SLAB_RECLAIM_ACCOUNT) && size <= PAGE_SIZE) {
		/*
		 * 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.
		 */
		cachep->gfporder = 0;
		cache_estimate(cachep->gfporder, size, align, flags,
P
Pekka Enberg 已提交
1816
			       &left_over, &cachep->num);
1817 1818
	} else
		left_over = calculate_slab_order(cachep, size, align, flags);
L
Linus Torvalds 已提交
1819 1820 1821 1822 1823

	if (!cachep->num) {
		printk("kmem_cache_create: couldn't create cache %s.\n", name);
		kmem_cache_free(&cache_cache, cachep);
		cachep = NULL;
1824
		goto oops;
L
Linus Torvalds 已提交
1825
	}
P
Pekka Enberg 已提交
1826 1827
	slab_size = ALIGN(cachep->num * sizeof(kmem_bufctl_t)
			  + sizeof(struct slab), align);
L
Linus Torvalds 已提交
1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839

	/*
	 * 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
1840 1841
		slab_size =
		    cachep->num * sizeof(kmem_bufctl_t) + sizeof(struct slab);
L
Linus Torvalds 已提交
1842 1843 1844 1845 1846 1847
	}

	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
1848
	cachep->colour = left_over / cachep->colour_off;
L
Linus Torvalds 已提交
1849 1850 1851 1852 1853 1854
	cachep->slab_size = slab_size;
	cachep->flags = flags;
	cachep->gfpflags = 0;
	if (flags & SLAB_CACHE_DMA)
		cachep->gfpflags |= GFP_DMA;
	spin_lock_init(&cachep->spinlock);
1855
	cachep->buffer_size = size;
L
Linus Torvalds 已提交
1856 1857

	if (flags & CFLGS_OFF_SLAB)
1858
		cachep->slabp_cache = kmem_find_general_cachep(slab_size, 0u);
L
Linus Torvalds 已提交
1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873
	cachep->ctor = ctor;
	cachep->dtor = dtor;
	cachep->name = name;

	/* Don't let CPUs to come and go */
	lock_cpu_hotplug();

	if (g_cpucache_up == FULL) {
		enable_cpucache(cachep);
	} else {
		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().
			 */
1874
			cachep->array[smp_processor_id()] =
P
Pekka Enberg 已提交
1875
			    &initarray_generic.cache;
1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886

			/* 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;
L
Linus Torvalds 已提交
1887
		} else {
1888
			cachep->array[smp_processor_id()] =
P
Pekka Enberg 已提交
1889
			    kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
1890 1891 1892 1893 1894 1895 1896 1897 1898

			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] =
P
Pekka Enberg 已提交
1899 1900 1901
					    kmalloc_node(sizeof
							 (struct kmem_list3),
							 GFP_KERNEL, node);
1902
					BUG_ON(!cachep->nodelists[node]);
P
Pekka Enberg 已提交
1903 1904
					kmem_list3_init(cachep->
							nodelists[node]);
1905 1906
				}
			}
L
Linus Torvalds 已提交
1907
		}
1908
		cachep->nodelists[numa_node_id()]->next_reap =
P
Pekka Enberg 已提交
1909 1910
		    jiffies + REAPTIMEOUT_LIST3 +
		    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
1911

1912 1913 1914 1915 1916
		BUG_ON(!cpu_cache_get(cachep));
		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;
L
Linus Torvalds 已提交
1917 1918
		cachep->batchcount = 1;
		cachep->limit = BOOT_CPUCACHE_ENTRIES;
P
Pekka Enberg 已提交
1919
	}
L
Linus Torvalds 已提交
1920 1921 1922 1923

	/* cache setup completed, link it into the list */
	list_add(&cachep->next, &cache_chain);
	unlock_cpu_hotplug();
P
Pekka Enberg 已提交
1924
      oops:
L
Linus Torvalds 已提交
1925 1926
	if (!cachep && (flags & SLAB_PANIC))
		panic("kmem_cache_create(): failed to create slab `%s'\n",
P
Pekka Enberg 已提交
1927
		      name);
I
Ingo Molnar 已提交
1928
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943
	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());
}

1944
static void check_spinlock_acquired(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
1945 1946 1947
{
#ifdef CONFIG_SMP
	check_irq_off();
1948
	assert_spin_locked(&cachep->nodelists[numa_node_id()]->list_lock);
L
Linus Torvalds 已提交
1949 1950
#endif
}
1951

1952
static void check_spinlock_acquired_node(struct kmem_cache *cachep, int node)
1953 1954 1955 1956 1957 1958 1959
{
#ifdef CONFIG_SMP
	check_irq_off();
	assert_spin_locked(&cachep->nodelists[node]->list_lock);
#endif
}

L
Linus Torvalds 已提交
1960 1961 1962 1963
#else
#define check_irq_off()	do { } while(0)
#define check_irq_on()	do { } while(0)
#define check_spinlock_acquired(x) do { } while(0)
1964
#define check_spinlock_acquired_node(x, y) do { } while(0)
L
Linus Torvalds 已提交
1965 1966 1967 1968 1969
#endif

/*
 * Waits for all CPUs to execute func().
 */
P
Pekka Enberg 已提交
1970
static void smp_call_function_all_cpus(void (*func)(void *arg), void *arg)
L
Linus Torvalds 已提交
1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984
{
	check_irq_on();
	preempt_disable();

	local_irq_disable();
	func(arg);
	local_irq_enable();

	if (smp_call_function(func, arg, 1, 1))
		BUG();

	preempt_enable();
}

1985
static void drain_array_locked(struct kmem_cache *cachep, struct array_cache *ac,
P
Pekka Enberg 已提交
1986
				int force, int node);
L
Linus Torvalds 已提交
1987 1988 1989

static void do_drain(void *arg)
{
1990
	struct kmem_cache *cachep = (struct kmem_cache *) arg;
L
Linus Torvalds 已提交
1991
	struct array_cache *ac;
1992
	int node = numa_node_id();
L
Linus Torvalds 已提交
1993 1994

	check_irq_off();
1995
	ac = cpu_cache_get(cachep);
1996 1997 1998
	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
1999 2000 2001
	ac->avail = 0;
}

2002
static void drain_cpu_caches(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2003
{
2004 2005 2006
	struct kmem_list3 *l3;
	int node;

L
Linus Torvalds 已提交
2007 2008 2009
	smp_call_function_all_cpus(do_drain, cachep);
	check_irq_on();
	spin_lock_irq(&cachep->spinlock);
P
Pekka Enberg 已提交
2010
	for_each_online_node(node) {
2011 2012 2013 2014 2015 2016 2017 2018 2019
		l3 = cachep->nodelists[node];
		if (l3) {
			spin_lock(&l3->list_lock);
			drain_array_locked(cachep, l3->shared, 1, node);
			spin_unlock(&l3->list_lock);
			if (l3->alien)
				drain_alien_cache(cachep, l3);
		}
	}
L
Linus Torvalds 已提交
2020 2021 2022
	spin_unlock_irq(&cachep->spinlock);
}

2023
static int __node_shrink(struct kmem_cache *cachep, int node)
L
Linus Torvalds 已提交
2024 2025
{
	struct slab *slabp;
2026
	struct kmem_list3 *l3 = cachep->nodelists[node];
L
Linus Torvalds 已提交
2027 2028
	int ret;

2029
	for (;;) {
L
Linus Torvalds 已提交
2030 2031
		struct list_head *p;

2032 2033
		p = l3->slabs_free.prev;
		if (p == &l3->slabs_free)
L
Linus Torvalds 已提交
2034 2035
			break;

2036
		slabp = list_entry(l3->slabs_free.prev, struct slab, list);
L
Linus Torvalds 已提交
2037 2038 2039 2040 2041 2042
#if DEBUG
		if (slabp->inuse)
			BUG();
#endif
		list_del(&slabp->list);

2043 2044
		l3->free_objects -= cachep->num;
		spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
2045
		slab_destroy(cachep, slabp);
2046
		spin_lock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
2047
	}
P
Pekka Enberg 已提交
2048
	ret = !list_empty(&l3->slabs_full) || !list_empty(&l3->slabs_partial);
L
Linus Torvalds 已提交
2049 2050 2051
	return ret;
}

2052
static int __cache_shrink(struct kmem_cache *cachep)
2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070
{
	int ret = 0, i = 0;
	struct kmem_list3 *l3;

	drain_cpu_caches(cachep);

	check_irq_on();
	for_each_online_node(i) {
		l3 = cachep->nodelists[i];
		if (l3) {
			spin_lock_irq(&l3->list_lock);
			ret += __node_shrink(cachep, i);
			spin_unlock_irq(&l3->list_lock);
		}
	}
	return (ret ? 1 : 0);
}

L
Linus Torvalds 已提交
2071 2072 2073 2074 2075 2076 2077
/**
 * 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.
 */
2078
int kmem_cache_shrink(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090
{
	if (!cachep || in_interrupt())
		BUG();

	return __cache_shrink(cachep);
}
EXPORT_SYMBOL(kmem_cache_shrink);

/**
 * kmem_cache_destroy - delete a cache
 * @cachep: the cache to destroy
 *
2091
 * Remove a struct kmem_cache object from the slab cache.
L
Linus Torvalds 已提交
2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103
 * Returns 0 on success.
 *
 * 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().
 */
2104
int kmem_cache_destroy(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2105 2106
{
	int i;
2107
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
2108 2109 2110 2111 2112 2113 2114 2115

	if (!cachep || in_interrupt())
		BUG();

	/* Don't let CPUs to come and go */
	lock_cpu_hotplug();

	/* Find the cache in the chain of caches. */
I
Ingo Molnar 已提交
2116
	mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2117 2118 2119 2120
	/*
	 * the chain is never empty, cache_cache is never destroyed
	 */
	list_del(&cachep->next);
I
Ingo Molnar 已提交
2121
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2122 2123 2124

	if (__cache_shrink(cachep)) {
		slab_error(cachep, "Can't free all objects");
I
Ingo Molnar 已提交
2125
		mutex_lock(&cache_chain_mutex);
P
Pekka Enberg 已提交
2126
		list_add(&cachep->next, &cache_chain);
I
Ingo Molnar 已提交
2127
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2128 2129 2130 2131 2132
		unlock_cpu_hotplug();
		return 1;
	}

	if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU))
2133
		synchronize_rcu();
L
Linus Torvalds 已提交
2134

2135
	for_each_online_cpu(i)
P
Pekka Enberg 已提交
2136
	    kfree(cachep->array[i]);
L
Linus Torvalds 已提交
2137 2138

	/* NUMA: free the list3 structures */
2139 2140 2141 2142 2143 2144 2145
	for_each_online_node(i) {
		if ((l3 = cachep->nodelists[i])) {
			kfree(l3->shared);
			free_alien_cache(l3->alien);
			kfree(l3);
		}
	}
L
Linus Torvalds 已提交
2146 2147 2148 2149 2150 2151 2152 2153 2154
	kmem_cache_free(&cache_cache, cachep);

	unlock_cpu_hotplug();

	return 0;
}
EXPORT_SYMBOL(kmem_cache_destroy);

/* Get the memory for a slab management obj. */
2155
static struct slab *alloc_slabmgmt(struct kmem_cache *cachep, void *objp,
P
Pekka Enberg 已提交
2156
				   int colour_off, gfp_t local_flags)
L
Linus Torvalds 已提交
2157 2158
{
	struct slab *slabp;
P
Pekka Enberg 已提交
2159

L
Linus Torvalds 已提交
2160 2161 2162 2163 2164 2165
	if (OFF_SLAB(cachep)) {
		/* Slab management obj is off-slab. */
		slabp = kmem_cache_alloc(cachep->slabp_cache, local_flags);
		if (!slabp)
			return NULL;
	} else {
P
Pekka Enberg 已提交
2166
		slabp = objp + colour_off;
L
Linus Torvalds 已提交
2167 2168 2169 2170
		colour_off += cachep->slab_size;
	}
	slabp->inuse = 0;
	slabp->colouroff = colour_off;
P
Pekka Enberg 已提交
2171
	slabp->s_mem = objp + colour_off;
L
Linus Torvalds 已提交
2172 2173 2174 2175 2176 2177

	return slabp;
}

static inline kmem_bufctl_t *slab_bufctl(struct slab *slabp)
{
P
Pekka Enberg 已提交
2178
	return (kmem_bufctl_t *) (slabp + 1);
L
Linus Torvalds 已提交
2179 2180
}

2181
static void cache_init_objs(struct kmem_cache *cachep,
P
Pekka Enberg 已提交
2182
			    struct slab *slabp, unsigned long ctor_flags)
L
Linus Torvalds 已提交
2183 2184 2185 2186
{
	int i;

	for (i = 0; i < cachep->num; i++) {
2187
		void *objp = slabp->s_mem + cachep->buffer_size * i;
L
Linus Torvalds 已提交
2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204
#if DEBUG
		/* need to poison the objs? */
		if (cachep->flags & SLAB_POISON)
			poison_obj(cachep, objp, POISON_FREE);
		if (cachep->flags & SLAB_STORE_USER)
			*dbg_userword(cachep, objp) = NULL;

		if (cachep->flags & SLAB_RED_ZONE) {
			*dbg_redzone1(cachep, objp) = RED_INACTIVE;
			*dbg_redzone2(cachep, objp) = RED_INACTIVE;
		}
		/*
		 * Constructors are not allowed to allocate memory from
		 * the same cache which they are a constructor for.
		 * Otherwise, deadlock. They must also be threaded.
		 */
		if (cachep->ctor && !(cachep->flags & SLAB_POISON))
2205
			cachep->ctor(objp + obj_offset(cachep), cachep,
P
Pekka Enberg 已提交
2206
				     ctor_flags);
L
Linus Torvalds 已提交
2207 2208 2209 2210

		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
2211
					   " end of an object");
L
Linus Torvalds 已提交
2212 2213
			if (*dbg_redzone1(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "constructor overwrote the"
P
Pekka Enberg 已提交
2214
					   " start of an object");
L
Linus Torvalds 已提交
2215
		}
2216
		if ((cachep->buffer_size % PAGE_SIZE) == 0 && OFF_SLAB(cachep)
P
Pekka Enberg 已提交
2217 2218
		    && cachep->flags & SLAB_POISON)
			kernel_map_pages(virt_to_page(objp),
2219
					 cachep->buffer_size / PAGE_SIZE, 0);
L
Linus Torvalds 已提交
2220 2221 2222 2223
#else
		if (cachep->ctor)
			cachep->ctor(objp, cachep, ctor_flags);
#endif
P
Pekka Enberg 已提交
2224
		slab_bufctl(slabp)[i] = i + 1;
L
Linus Torvalds 已提交
2225
	}
P
Pekka Enberg 已提交
2226
	slab_bufctl(slabp)[i - 1] = BUFCTL_END;
L
Linus Torvalds 已提交
2227 2228 2229
	slabp->free = 0;
}

2230
static void kmem_flagcheck(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2231 2232 2233 2234 2235 2236 2237 2238 2239 2240
{
	if (flags & SLAB_DMA) {
		if (!(cachep->gfpflags & GFP_DMA))
			BUG();
	} else {
		if (cachep->gfpflags & GFP_DMA)
			BUG();
	}
}

2241
static void *slab_get_obj(struct kmem_cache *cachep, struct slab *slabp, int nodeid)
2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256
{
	void *objp = slabp->s_mem + (slabp->free * cachep->buffer_size);
	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;
}

2257
static void slab_put_obj(struct kmem_cache *cachep, struct slab *slabp, void *objp,
2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276
			  int nodeid)
{
	unsigned int objnr = (unsigned)(objp-slabp->s_mem) / cachep->buffer_size;

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

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

2277
static void set_slab_attr(struct kmem_cache *cachep, struct slab *slabp, void *objp)
L
Linus Torvalds 已提交
2278 2279 2280 2281 2282 2283 2284 2285
{
	int i;
	struct page *page;

	/* Nasty!!!!!! I hope this is OK. */
	i = 1 << cachep->gfporder;
	page = virt_to_page(objp);
	do {
2286 2287
		page_set_cache(page, cachep);
		page_set_slab(page, slabp);
L
Linus Torvalds 已提交
2288 2289 2290 2291 2292 2293 2294 2295
		page++;
	} while (--i);
}

/*
 * 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.
 */
2296
static int cache_grow(struct kmem_cache *cachep, gfp_t flags, int nodeid)
L
Linus Torvalds 已提交
2297
{
P
Pekka Enberg 已提交
2298 2299 2300 2301 2302
	struct slab *slabp;
	void *objp;
	size_t offset;
	gfp_t local_flags;
	unsigned long ctor_flags;
2303
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
2304 2305

	/* Be lazy and only check for valid flags here,
P
Pekka Enberg 已提交
2306
	 * keeping it out of the critical path in kmem_cache_alloc().
L
Linus Torvalds 已提交
2307
	 */
P
Pekka Enberg 已提交
2308
	if (flags & ~(SLAB_DMA | SLAB_LEVEL_MASK | SLAB_NO_GROW))
L
Linus Torvalds 已提交
2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334
		BUG();
	if (flags & SLAB_NO_GROW)
		return 0;

	ctor_flags = SLAB_CTOR_CONSTRUCTOR;
	local_flags = (flags & SLAB_LEVEL_MASK);
	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;

	/* About to mess with non-constant members - lock. */
	check_irq_off();
	spin_lock(&cachep->spinlock);

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

	spin_unlock(&cachep->spinlock);

2335
	check_irq_off();
L
Linus Torvalds 已提交
2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346
	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);

2347 2348 2349
	/* Get mem for the objs.
	 * Attempt to allocate a physical page from 'nodeid',
	 */
L
Linus Torvalds 已提交
2350 2351 2352 2353 2354 2355 2356
	if (!(objp = kmem_getpages(cachep, flags, nodeid)))
		goto failed;

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

2357
	slabp->nodeid = nodeid;
L
Linus Torvalds 已提交
2358 2359 2360 2361 2362 2363 2364
	set_slab_attr(cachep, slabp, objp);

	cache_init_objs(cachep, slabp, ctor_flags);

	if (local_flags & __GFP_WAIT)
		local_irq_disable();
	check_irq_off();
2365 2366
	l3 = cachep->nodelists[nodeid];
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2367 2368

	/* Make slab active. */
2369
	list_add_tail(&slabp->list, &(l3->slabs_free));
L
Linus Torvalds 已提交
2370
	STATS_INC_GROWN(cachep);
2371 2372
	l3->free_objects += cachep->num;
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2373
	return 1;
P
Pekka Enberg 已提交
2374
      opps1:
L
Linus Torvalds 已提交
2375
	kmem_freepages(cachep, objp);
P
Pekka Enberg 已提交
2376
      failed:
L
Linus Torvalds 已提交
2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395
	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
2396 2397
		       (unsigned long)objp);
		BUG();
L
Linus Torvalds 已提交
2398 2399 2400
	}
	page = virt_to_page(objp);
	if (!PageSlab(page)) {
P
Pekka Enberg 已提交
2401 2402
		printk(KERN_ERR "kfree_debugcheck: bad ptr %lxh.\n",
		       (unsigned long)objp);
L
Linus Torvalds 已提交
2403 2404 2405 2406
		BUG();
	}
}

2407
static void *cache_free_debugcheck(struct kmem_cache *cachep, void *objp,
P
Pekka Enberg 已提交
2408
				   void *caller)
L
Linus Torvalds 已提交
2409 2410 2411 2412 2413
{
	struct page *page;
	unsigned int objnr;
	struct slab *slabp;

2414
	objp -= obj_offset(cachep);
L
Linus Torvalds 已提交
2415 2416 2417
	kfree_debugcheck(objp);
	page = virt_to_page(objp);

2418
	if (page_get_cache(page) != cachep) {
P
Pekka Enberg 已提交
2419 2420 2421
		printk(KERN_ERR
		       "mismatch in kmem_cache_free: expected cache %p, got %p\n",
		       page_get_cache(page), cachep);
L
Linus Torvalds 已提交
2422
		printk(KERN_ERR "%p is %s.\n", cachep, cachep->name);
P
Pekka Enberg 已提交
2423 2424
		printk(KERN_ERR "%p is %s.\n", page_get_cache(page),
		       page_get_cache(page)->name);
L
Linus Torvalds 已提交
2425 2426
		WARN_ON(1);
	}
2427
	slabp = page_get_slab(page);
L
Linus Torvalds 已提交
2428 2429

	if (cachep->flags & SLAB_RED_ZONE) {
P
Pekka Enberg 已提交
2430 2431 2432 2433 2434 2435 2436 2437 2438
		if (*dbg_redzone1(cachep, objp) != RED_ACTIVE
		    || *dbg_redzone2(cachep, objp) != RED_ACTIVE) {
			slab_error(cachep,
				   "double free, or memory outside"
				   " object was overwritten");
			printk(KERN_ERR
			       "%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
2439 2440 2441 2442 2443 2444 2445
		}
		*dbg_redzone1(cachep, objp) = RED_INACTIVE;
		*dbg_redzone2(cachep, objp) = RED_INACTIVE;
	}
	if (cachep->flags & SLAB_STORE_USER)
		*dbg_userword(cachep, objp) = caller;

2446
	objnr = (unsigned)(objp - slabp->s_mem) / cachep->buffer_size;
L
Linus Torvalds 已提交
2447 2448

	BUG_ON(objnr >= cachep->num);
2449
	BUG_ON(objp != slabp->s_mem + objnr * cachep->buffer_size);
L
Linus Torvalds 已提交
2450 2451 2452 2453 2454 2455

	if (cachep->flags & SLAB_DEBUG_INITIAL) {
		/* Need to call the slab's constructor so the
		 * caller can perform a verify of its state (debugging).
		 * Called without the cache-lock held.
		 */
2456
		cachep->ctor(objp + obj_offset(cachep),
P
Pekka Enberg 已提交
2457
			     cachep, SLAB_CTOR_CONSTRUCTOR | SLAB_CTOR_VERIFY);
L
Linus Torvalds 已提交
2458 2459 2460 2461 2462
	}
	if (cachep->flags & SLAB_POISON && cachep->dtor) {
		/* we want to cache poison the object,
		 * call the destruction callback
		 */
2463
		cachep->dtor(objp + obj_offset(cachep), cachep, 0);
L
Linus Torvalds 已提交
2464 2465 2466
	}
	if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
2467
		if ((cachep->buffer_size % PAGE_SIZE) == 0 && OFF_SLAB(cachep)) {
L
Linus Torvalds 已提交
2468
			store_stackinfo(cachep, objp, (unsigned long)caller);
P
Pekka Enberg 已提交
2469
			kernel_map_pages(virt_to_page(objp),
2470
					 cachep->buffer_size / PAGE_SIZE, 0);
L
Linus Torvalds 已提交
2471 2472 2473 2474 2475 2476 2477 2478 2479 2480
		} else {
			poison_obj(cachep, objp, POISON_FREE);
		}
#else
		poison_obj(cachep, objp, POISON_FREE);
#endif
	}
	return objp;
}

2481
static void check_slabp(struct kmem_cache *cachep, struct slab *slabp)
L
Linus Torvalds 已提交
2482 2483 2484
{
	kmem_bufctl_t i;
	int entries = 0;
P
Pekka Enberg 已提交
2485

L
Linus Torvalds 已提交
2486 2487 2488 2489 2490 2491 2492
	/* 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) {
P
Pekka Enberg 已提交
2493 2494 2495 2496 2497 2498 2499 2500
	      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);
		for (i = 0;
		     i < sizeof(slabp) + cachep->num * sizeof(kmem_bufctl_t);
		     i++) {
			if ((i % 16) == 0)
L
Linus Torvalds 已提交
2501
				printk("\n%03x:", i);
P
Pekka Enberg 已提交
2502
			printk(" %02x", ((unsigned char *)slabp)[i]);
L
Linus Torvalds 已提交
2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513
		}
		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

2514
static void *cache_alloc_refill(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2515 2516 2517 2518 2519 2520
{
	int batchcount;
	struct kmem_list3 *l3;
	struct array_cache *ac;

	check_irq_off();
2521
	ac = cpu_cache_get(cachep);
P
Pekka Enberg 已提交
2522
      retry:
L
Linus Torvalds 已提交
2523 2524 2525 2526 2527 2528 2529 2530
	batchcount = ac->batchcount;
	if (!ac->touched && batchcount > BATCHREFILL_LIMIT) {
		/* if there was little recent activity on this
		 * cache, then perform only a partial refill.
		 * Otherwise we could generate refill bouncing.
		 */
		batchcount = BATCHREFILL_LIMIT;
	}
2531 2532 2533 2534
	l3 = cachep->nodelists[numa_node_id()];

	BUG_ON(ac->avail > 0 || !l3);
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2535 2536 2537 2538 2539 2540 2541 2542

	if (l3->shared) {
		struct array_cache *shared_array = l3->shared;
		if (shared_array->avail) {
			if (batchcount > shared_array->avail)
				batchcount = shared_array->avail;
			shared_array->avail -= batchcount;
			ac->avail = batchcount;
2543
			memcpy(ac->entry,
P
Pekka Enberg 已提交
2544 2545
			       &(shared_array->entry[shared_array->avail]),
			       sizeof(void *) * batchcount);
L
Linus Torvalds 已提交
2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569
			shared_array->touched = 1;
			goto alloc_done;
		}
	}
	while (batchcount > 0) {
		struct list_head *entry;
		struct slab *slabp;
		/* Get slab alloc is to come from. */
		entry = l3->slabs_partial.next;
		if (entry == &l3->slabs_partial) {
			l3->free_touched = 1;
			entry = l3->slabs_free.next;
			if (entry == &l3->slabs_free)
				goto must_grow;
		}

		slabp = list_entry(entry, struct slab, list);
		check_slabp(cachep, slabp);
		check_spinlock_acquired(cachep);
		while (slabp->inuse < cachep->num && batchcount--) {
			STATS_INC_ALLOCED(cachep);
			STATS_INC_ACTIVE(cachep);
			STATS_SET_HIGH(cachep);

2570 2571
			ac->entry[ac->avail++] = slab_get_obj(cachep, slabp,
							    numa_node_id());
L
Linus Torvalds 已提交
2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582
		}
		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);
	}

P
Pekka Enberg 已提交
2583
      must_grow:
L
Linus Torvalds 已提交
2584
	l3->free_objects -= ac->avail;
P
Pekka Enberg 已提交
2585
      alloc_done:
2586
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2587 2588 2589

	if (unlikely(!ac->avail)) {
		int x;
2590 2591
		x = cache_grow(cachep, flags, numa_node_id());

L
Linus Torvalds 已提交
2592
		// cache_grow can reenable interrupts, then ac could change.
2593
		ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
2594 2595 2596
		if (!x && ac->avail == 0)	// no objects in sight? abort
			return NULL;

P
Pekka Enberg 已提交
2597
		if (!ac->avail)	// objects refilled by interrupt?
L
Linus Torvalds 已提交
2598 2599 2600
			goto retry;
	}
	ac->touched = 1;
2601
	return ac->entry[--ac->avail];
L
Linus Torvalds 已提交
2602 2603 2604
}

static inline void
2605
cache_alloc_debugcheck_before(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2606 2607 2608 2609 2610 2611 2612 2613
{
	might_sleep_if(flags & __GFP_WAIT);
#if DEBUG
	kmem_flagcheck(cachep, flags);
#endif
}

#if DEBUG
2614
static void *cache_alloc_debugcheck_after(struct kmem_cache *cachep, gfp_t flags,
P
Pekka Enberg 已提交
2615
					void *objp, void *caller)
L
Linus Torvalds 已提交
2616
{
P
Pekka Enberg 已提交
2617
	if (!objp)
L
Linus Torvalds 已提交
2618
		return objp;
P
Pekka Enberg 已提交
2619
	if (cachep->flags & SLAB_POISON) {
L
Linus Torvalds 已提交
2620
#ifdef CONFIG_DEBUG_PAGEALLOC
2621
		if ((cachep->buffer_size % PAGE_SIZE) == 0 && OFF_SLAB(cachep))
P
Pekka Enberg 已提交
2622
			kernel_map_pages(virt_to_page(objp),
2623
					 cachep->buffer_size / PAGE_SIZE, 1);
L
Linus Torvalds 已提交
2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634
		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) {
P
Pekka Enberg 已提交
2635 2636 2637 2638 2639 2640 2641 2642 2643
		if (*dbg_redzone1(cachep, objp) != RED_INACTIVE
		    || *dbg_redzone2(cachep, objp) != RED_INACTIVE) {
			slab_error(cachep,
				   "double free, or memory outside"
				   " object was overwritten");
			printk(KERN_ERR
			       "%p: redzone 1: 0x%lx, redzone 2: 0x%lx.\n",
			       objp, *dbg_redzone1(cachep, objp),
			       *dbg_redzone2(cachep, objp));
L
Linus Torvalds 已提交
2644 2645 2646 2647
		}
		*dbg_redzone1(cachep, objp) = RED_ACTIVE;
		*dbg_redzone2(cachep, objp) = RED_ACTIVE;
	}
2648
	objp += obj_offset(cachep);
L
Linus Torvalds 已提交
2649
	if (cachep->ctor && cachep->flags & SLAB_POISON) {
P
Pekka Enberg 已提交
2650
		unsigned long ctor_flags = SLAB_CTOR_CONSTRUCTOR;
L
Linus Torvalds 已提交
2651 2652 2653 2654 2655

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

		cachep->ctor(objp, cachep, ctor_flags);
P
Pekka Enberg 已提交
2656
	}
L
Linus Torvalds 已提交
2657 2658 2659 2660 2661 2662
	return objp;
}
#else
#define cache_alloc_debugcheck_after(a,b,objp,d) (objp)
#endif

2663
static inline void *____cache_alloc(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2664
{
P
Pekka Enberg 已提交
2665
	void *objp;
L
Linus Torvalds 已提交
2666 2667
	struct array_cache *ac;

2668
#ifdef CONFIG_NUMA
2669
	if (unlikely(current->mempolicy && !in_interrupt())) {
2670 2671 2672 2673 2674 2675 2676
		int nid = slab_node(current->mempolicy);

		if (nid != numa_node_id())
			return __cache_alloc_node(cachep, flags, nid);
	}
#endif

2677
	check_irq_off();
2678
	ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
2679 2680 2681
	if (likely(ac->avail)) {
		STATS_INC_ALLOCHIT(cachep);
		ac->touched = 1;
2682
		objp = ac->entry[--ac->avail];
L
Linus Torvalds 已提交
2683 2684 2685 2686
	} else {
		STATS_INC_ALLOCMISS(cachep);
		objp = cache_alloc_refill(cachep, flags);
	}
2687 2688 2689
	return objp;
}

2690 2691
static __always_inline void *
__cache_alloc(struct kmem_cache *cachep, gfp_t flags, void *caller)
2692 2693
{
	unsigned long save_flags;
P
Pekka Enberg 已提交
2694
	void *objp;
2695 2696 2697 2698 2699

	cache_alloc_debugcheck_before(cachep, flags);

	local_irq_save(save_flags);
	objp = ____cache_alloc(cachep, flags);
L
Linus Torvalds 已提交
2700
	local_irq_restore(save_flags);
2701
	objp = cache_alloc_debugcheck_after(cachep, flags, objp,
2702
					    caller);
2703
	prefetchw(objp);
L
Linus Torvalds 已提交
2704 2705 2706
	return objp;
}

2707 2708 2709
#ifdef CONFIG_NUMA
/*
 * A interface to enable slab creation on nodeid
L
Linus Torvalds 已提交
2710
 */
2711
static void *__cache_alloc_node(struct kmem_cache *cachep, gfp_t flags, int nodeid)
2712 2713
{
	struct list_head *entry;
P
Pekka Enberg 已提交
2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741
	struct slab *slabp;
	struct kmem_list3 *l3;
	void *obj;
	int x;

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

      retry:
	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);

2742
	obj = slab_get_obj(cachep, slabp, nodeid);
P
Pekka Enberg 已提交
2743 2744 2745 2746 2747 2748 2749 2750 2751 2752
	check_slabp(cachep, slabp);
	l3->free_objects--;
	/* move slabp to correct slabp list: */
	list_del(&slabp->list);

	if (slabp->free == BUFCTL_END) {
		list_add(&slabp->list, &l3->slabs_full);
	} else {
		list_add(&slabp->list, &l3->slabs_partial);
	}
2753

P
Pekka Enberg 已提交
2754 2755
	spin_unlock(&l3->list_lock);
	goto done;
2756

P
Pekka Enberg 已提交
2757 2758 2759
      must_grow:
	spin_unlock(&l3->list_lock);
	x = cache_grow(cachep, flags, nodeid);
L
Linus Torvalds 已提交
2760

P
Pekka Enberg 已提交
2761 2762
	if (!x)
		return NULL;
2763

P
Pekka Enberg 已提交
2764 2765 2766
	goto retry;
      done:
	return obj;
2767 2768 2769 2770 2771 2772
}
#endif

/*
 * Caller needs to acquire correct kmem_list's list_lock
 */
2773
static void free_block(struct kmem_cache *cachep, void **objpp, int nr_objects,
P
Pekka Enberg 已提交
2774
		       int node)
L
Linus Torvalds 已提交
2775 2776
{
	int i;
2777
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
2778 2779 2780 2781 2782

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

2783
		slabp = virt_to_slab(objp);
2784
		l3 = cachep->nodelists[node];
L
Linus Torvalds 已提交
2785
		list_del(&slabp->list);
2786
		check_spinlock_acquired_node(cachep, node);
L
Linus Torvalds 已提交
2787
		check_slabp(cachep, slabp);
2788
		slab_put_obj(cachep, slabp, objp, node);
L
Linus Torvalds 已提交
2789
		STATS_DEC_ACTIVE(cachep);
2790
		l3->free_objects++;
L
Linus Torvalds 已提交
2791 2792 2793 2794
		check_slabp(cachep, slabp);

		/* fixup slab chains */
		if (slabp->inuse == 0) {
2795 2796
			if (l3->free_objects > l3->free_limit) {
				l3->free_objects -= cachep->num;
L
Linus Torvalds 已提交
2797 2798
				slab_destroy(cachep, slabp);
			} else {
2799
				list_add(&slabp->list, &l3->slabs_free);
L
Linus Torvalds 已提交
2800 2801 2802 2803 2804 2805
			}
		} else {
			/* Unconditionally move a slab to the end of the
			 * partial list on free - maximum time for the
			 * other objects to be freed, too.
			 */
2806
			list_add_tail(&slabp->list, &l3->slabs_partial);
L
Linus Torvalds 已提交
2807 2808 2809 2810
		}
	}
}

2811
static void cache_flusharray(struct kmem_cache *cachep, struct array_cache *ac)
L
Linus Torvalds 已提交
2812 2813
{
	int batchcount;
2814
	struct kmem_list3 *l3;
2815
	int node = numa_node_id();
L
Linus Torvalds 已提交
2816 2817 2818 2819 2820 2821

	batchcount = ac->batchcount;
#if DEBUG
	BUG_ON(!batchcount || batchcount > ac->avail);
#endif
	check_irq_off();
2822
	l3 = cachep->nodelists[node];
2823 2824 2825
	spin_lock(&l3->list_lock);
	if (l3->shared) {
		struct array_cache *shared_array = l3->shared;
P
Pekka Enberg 已提交
2826
		int max = shared_array->limit - shared_array->avail;
L
Linus Torvalds 已提交
2827 2828 2829
		if (max) {
			if (batchcount > max)
				batchcount = max;
2830
			memcpy(&(shared_array->entry[shared_array->avail]),
P
Pekka Enberg 已提交
2831
			       ac->entry, sizeof(void *) * batchcount);
L
Linus Torvalds 已提交
2832 2833 2834 2835 2836
			shared_array->avail += batchcount;
			goto free_done;
		}
	}

2837
	free_block(cachep, ac->entry, batchcount, node);
P
Pekka Enberg 已提交
2838
      free_done:
L
Linus Torvalds 已提交
2839 2840 2841 2842 2843
#if STATS
	{
		int i = 0;
		struct list_head *p;

2844 2845
		p = l3->slabs_free.next;
		while (p != &(l3->slabs_free)) {
L
Linus Torvalds 已提交
2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856
			struct slab *slabp;

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

			i++;
			p = p->next;
		}
		STATS_SET_FREEABLE(cachep, i);
	}
#endif
2857
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2858
	ac->avail -= batchcount;
2859
	memmove(ac->entry, &(ac->entry[batchcount]),
P
Pekka Enberg 已提交
2860
		sizeof(void *) * ac->avail);
L
Linus Torvalds 已提交
2861 2862 2863 2864 2865 2866 2867 2868 2869
}

/*
 * __cache_free
 * 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.
 */
2870
static inline void __cache_free(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
2871
{
2872
	struct array_cache *ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
2873 2874 2875 2876

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

2877 2878 2879 2880 2881 2882
	/* Make sure we are not freeing a object from another
	 * node to the array cache on this cpu.
	 */
#ifdef CONFIG_NUMA
	{
		struct slab *slabp;
2883
		slabp = virt_to_slab(objp);
2884 2885 2886
		if (unlikely(slabp->nodeid != numa_node_id())) {
			struct array_cache *alien = NULL;
			int nodeid = slabp->nodeid;
P
Pekka Enberg 已提交
2887 2888
			struct kmem_list3 *l3 =
			    cachep->nodelists[numa_node_id()];
2889 2890 2891 2892 2893 2894 2895

			STATS_INC_NODEFREES(cachep);
			if (l3->alien && l3->alien[nodeid]) {
				alien = l3->alien[nodeid];
				spin_lock(&alien->lock);
				if (unlikely(alien->avail == alien->limit))
					__drain_alien_cache(cachep,
P
Pekka Enberg 已提交
2896
							    alien, nodeid);
2897 2898 2899 2900
				alien->entry[alien->avail++] = objp;
				spin_unlock(&alien->lock);
			} else {
				spin_lock(&(cachep->nodelists[nodeid])->
P
Pekka Enberg 已提交
2901
					  list_lock);
2902
				free_block(cachep, &objp, 1, nodeid);
2903
				spin_unlock(&(cachep->nodelists[nodeid])->
P
Pekka Enberg 已提交
2904
					    list_lock);
2905 2906 2907 2908 2909
			}
			return;
		}
	}
#endif
L
Linus Torvalds 已提交
2910 2911
	if (likely(ac->avail < ac->limit)) {
		STATS_INC_FREEHIT(cachep);
2912
		ac->entry[ac->avail++] = objp;
L
Linus Torvalds 已提交
2913 2914 2915 2916
		return;
	} else {
		STATS_INC_FREEMISS(cachep);
		cache_flusharray(cachep, ac);
2917
		ac->entry[ac->avail++] = objp;
L
Linus Torvalds 已提交
2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928
	}
}

/**
 * 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.
 */
2929
void *kmem_cache_alloc(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2930
{
2931
	return __cache_alloc(cachep, flags, __builtin_return_address(0));
L
Linus Torvalds 已提交
2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948
}
EXPORT_SYMBOL(kmem_cache_alloc);

/**
 * 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.
 */
2949
int fastcall kmem_ptr_validate(struct kmem_cache *cachep, void *ptr)
L
Linus Torvalds 已提交
2950
{
P
Pekka Enberg 已提交
2951
	unsigned long addr = (unsigned long)ptr;
L
Linus Torvalds 已提交
2952
	unsigned long min_addr = PAGE_OFFSET;
P
Pekka Enberg 已提交
2953
	unsigned long align_mask = BYTES_PER_WORD - 1;
2954
	unsigned long size = cachep->buffer_size;
L
Linus Torvalds 已提交
2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969
	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;
2970
	if (unlikely(page_get_cache(page) != cachep))
L
Linus Torvalds 已提交
2971 2972
		goto out;
	return 1;
P
Pekka Enberg 已提交
2973
      out:
L
Linus Torvalds 已提交
2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986
	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.
 *
 * Identical to kmem_cache_alloc, except that this function is slow
 * and can sleep. And it will allocate memory on the given node, which
 * can improve the performance for cpu bound structures.
2987 2988
 * New and improved: it will now make sure that the object gets
 * put on the correct node list so that there is no false sharing.
L
Linus Torvalds 已提交
2989
 */
2990
void *kmem_cache_alloc_node(struct kmem_cache *cachep, gfp_t flags, int nodeid)
L
Linus Torvalds 已提交
2991
{
2992 2993
	unsigned long save_flags;
	void *ptr;
L
Linus Torvalds 已提交
2994

2995 2996
	cache_alloc_debugcheck_before(cachep, flags);
	local_irq_save(save_flags);
2997 2998 2999

	if (nodeid == -1 || nodeid == numa_node_id() ||
	    !cachep->nodelists[nodeid])
3000 3001 3002
		ptr = ____cache_alloc(cachep, flags);
	else
		ptr = __cache_alloc_node(cachep, flags, nodeid);
3003
	local_irq_restore(save_flags);
3004 3005 3006

	ptr = cache_alloc_debugcheck_after(cachep, flags, ptr,
					   __builtin_return_address(0));
L
Linus Torvalds 已提交
3007

3008
	return ptr;
L
Linus Torvalds 已提交
3009 3010 3011
}
EXPORT_SYMBOL(kmem_cache_alloc_node);

A
Al Viro 已提交
3012
void *kmalloc_node(size_t size, gfp_t flags, int node)
3013
{
3014
	struct kmem_cache *cachep;
3015 3016 3017 3018 3019 3020 3021

	cachep = kmem_find_general_cachep(size, flags);
	if (unlikely(cachep == NULL))
		return NULL;
	return kmem_cache_alloc_node(cachep, flags, node);
}
EXPORT_SYMBOL(kmalloc_node);
L
Linus Torvalds 已提交
3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044
#endif

/**
 * kmalloc - allocate memory
 * @size: how many bytes of memory are required.
 * @flags: the type of memory to allocate.
 *
 * kmalloc is the normal method of allocating memory
 * in the kernel.
 *
 * The @flags argument may be one of:
 *
 * %GFP_USER - Allocate memory on behalf of user.  May sleep.
 *
 * %GFP_KERNEL - Allocate normal kernel ram.  May sleep.
 *
 * %GFP_ATOMIC - Allocation will not sleep.  Use inside interrupt handlers.
 *
 * Additionally, the %GFP_DMA flag may be set to indicate the memory
 * must be suitable for DMA.  This can mean different things on different
 * platforms.  For example, on i386, it means that the memory must come
 * from the first 16MB.
 */
3045 3046
static __always_inline void *__do_kmalloc(size_t size, gfp_t flags,
					  void *caller)
L
Linus Torvalds 已提交
3047
{
3048
	struct kmem_cache *cachep;
L
Linus Torvalds 已提交
3049

3050 3051 3052 3053 3054 3055
	/* 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);
3056 3057
	if (unlikely(cachep == NULL))
		return NULL;
3058 3059 3060 3061 3062 3063 3064 3065
	return __cache_alloc(cachep, flags, caller);
}

#ifndef CONFIG_DEBUG_SLAB

void *__kmalloc(size_t size, gfp_t flags)
{
	return __do_kmalloc(size, flags, NULL);
L
Linus Torvalds 已提交
3066 3067 3068
}
EXPORT_SYMBOL(__kmalloc);

3069 3070 3071 3072 3073 3074 3075 3076 3077 3078
#else

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

#endif

L
Linus Torvalds 已提交
3079 3080 3081 3082 3083 3084 3085 3086
#ifdef CONFIG_SMP
/**
 * __alloc_percpu - allocate one copy of the object for every present
 * cpu in the system, zeroing them.
 * Objects should be dereferenced using the per_cpu_ptr macro only.
 *
 * @size: how many bytes of memory are required.
 */
3087
void *__alloc_percpu(size_t size)
L
Linus Torvalds 已提交
3088 3089
{
	int i;
P
Pekka Enberg 已提交
3090
	struct percpu_data *pdata = kmalloc(sizeof(*pdata), GFP_KERNEL);
L
Linus Torvalds 已提交
3091 3092 3093 3094

	if (!pdata)
		return NULL;

3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106
	/*
	 * Cannot use for_each_online_cpu since a cpu may come online
	 * and we have no way of figuring out how to fix the array
	 * that we have allocated then....
	 */
	for_each_cpu(i) {
		int node = cpu_to_node(i);

		if (node_online(node))
			pdata->ptrs[i] = kmalloc_node(size, GFP_KERNEL, node);
		else
			pdata->ptrs[i] = kmalloc(size, GFP_KERNEL);
L
Linus Torvalds 已提交
3107 3108 3109 3110 3111 3112 3113

		if (!pdata->ptrs[i])
			goto unwind_oom;
		memset(pdata->ptrs[i], 0, size);
	}

	/* Catch derefs w/o wrappers */
P
Pekka Enberg 已提交
3114
	return (void *)(~(unsigned long)pdata);
L
Linus Torvalds 已提交
3115

P
Pekka Enberg 已提交
3116
      unwind_oom:
L
Linus Torvalds 已提交
3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135
	while (--i >= 0) {
		if (!cpu_possible(i))
			continue;
		kfree(pdata->ptrs[i]);
	}
	kfree(pdata);
	return NULL;
}
EXPORT_SYMBOL(__alloc_percpu);
#endif

/**
 * kmem_cache_free - Deallocate an object
 * @cachep: The cache the allocation was from.
 * @objp: The previously allocated object.
 *
 * Free an object which was previously allocated from this
 * cache.
 */
3136
void kmem_cache_free(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149
{
	unsigned long flags;

	local_irq_save(flags);
	__cache_free(cachep, objp);
	local_irq_restore(flags);
}
EXPORT_SYMBOL(kmem_cache_free);

/**
 * kfree - free previously allocated memory
 * @objp: pointer returned by kmalloc.
 *
3150 3151
 * If @objp is NULL, no operation is performed.
 *
L
Linus Torvalds 已提交
3152 3153 3154 3155 3156
 * Don't free memory not originally allocated by kmalloc()
 * or you will run into trouble.
 */
void kfree(const void *objp)
{
3157
	struct kmem_cache *c;
L
Linus Torvalds 已提交
3158 3159 3160 3161 3162 3163
	unsigned long flags;

	if (unlikely(!objp))
		return;
	local_irq_save(flags);
	kfree_debugcheck(objp);
3164
	c = virt_to_cache(objp);
3165
	mutex_debug_check_no_locks_freed(objp, obj_size(c));
P
Pekka Enberg 已提交
3166
	__cache_free(c, (void *)objp);
L
Linus Torvalds 已提交
3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178
	local_irq_restore(flags);
}
EXPORT_SYMBOL(kfree);

#ifdef CONFIG_SMP
/**
 * free_percpu - free previously allocated percpu memory
 * @objp: pointer returned by alloc_percpu.
 *
 * Don't free memory not originally allocated by alloc_percpu()
 * The complemented objp is to check for that.
 */
P
Pekka Enberg 已提交
3179
void free_percpu(const void *objp)
L
Linus Torvalds 已提交
3180 3181
{
	int i;
P
Pekka Enberg 已提交
3182
	struct percpu_data *p = (struct percpu_data *)(~(unsigned long)objp);
L
Linus Torvalds 已提交
3183

3184 3185 3186 3187
	/*
	 * We allocate for all cpus so we cannot use for online cpu here.
	 */
	for_each_cpu(i)
P
Pekka Enberg 已提交
3188
	    kfree(p->ptrs[i]);
L
Linus Torvalds 已提交
3189 3190 3191 3192 3193
	kfree(p);
}
EXPORT_SYMBOL(free_percpu);
#endif

3194
unsigned int kmem_cache_size(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
3195
{
3196
	return obj_size(cachep);
L
Linus Torvalds 已提交
3197 3198 3199
}
EXPORT_SYMBOL(kmem_cache_size);

3200
const char *kmem_cache_name(struct kmem_cache *cachep)
3201 3202 3203 3204 3205
{
	return cachep->name;
}
EXPORT_SYMBOL_GPL(kmem_cache_name);

3206 3207 3208
/*
 * This initializes kmem_list3 for all nodes.
 */
3209
static int alloc_kmemlist(struct kmem_cache *cachep)
3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221
{
	int node;
	struct kmem_list3 *l3;
	int err = 0;

	for_each_online_node(node) {
		struct array_cache *nc = NULL, *new;
		struct array_cache **new_alien = NULL;
#ifdef CONFIG_NUMA
		if (!(new_alien = alloc_alien_cache(node, cachep->limit)))
			goto fail;
#endif
P
Pekka Enberg 已提交
3222 3223 3224
		if (!(new = alloc_arraycache(node, (cachep->shared *
						    cachep->batchcount),
					     0xbaadf00d)))
3225 3226 3227 3228 3229 3230
			goto fail;
		if ((l3 = cachep->nodelists[node])) {

			spin_lock_irq(&l3->list_lock);

			if ((nc = cachep->nodelists[node]->shared))
P
Pekka Enberg 已提交
3231
				free_block(cachep, nc->entry, nc->avail, node);
3232 3233 3234 3235 3236 3237

			l3->shared = new;
			if (!cachep->nodelists[node]->alien) {
				l3->alien = new_alien;
				new_alien = NULL;
			}
P
Pekka Enberg 已提交
3238 3239
			l3->free_limit = (1 + nr_cpus_node(node)) *
			    cachep->batchcount + cachep->num;
3240 3241 3242 3243 3244 3245
			spin_unlock_irq(&l3->list_lock);
			kfree(nc);
			free_alien_cache(new_alien);
			continue;
		}
		if (!(l3 = kmalloc_node(sizeof(struct kmem_list3),
P
Pekka Enberg 已提交
3246
					GFP_KERNEL, node)))
3247 3248 3249 3250
			goto fail;

		kmem_list3_init(l3);
		l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
P
Pekka Enberg 已提交
3251
		    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
3252 3253
		l3->shared = new;
		l3->alien = new_alien;
P
Pekka Enberg 已提交
3254 3255
		l3->free_limit = (1 + nr_cpus_node(node)) *
		    cachep->batchcount + cachep->num;
3256 3257 3258
		cachep->nodelists[node] = l3;
	}
	return err;
P
Pekka Enberg 已提交
3259
      fail:
3260 3261 3262 3263
	err = -ENOMEM;
	return err;
}

L
Linus Torvalds 已提交
3264
struct ccupdate_struct {
3265
	struct kmem_cache *cachep;
L
Linus Torvalds 已提交
3266 3267 3268 3269 3270 3271 3272 3273 3274
	struct array_cache *new[NR_CPUS];
};

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

	check_irq_off();
3275
	old = cpu_cache_get(new->cachep);
3276

L
Linus Torvalds 已提交
3277 3278 3279 3280
	new->cachep->array[smp_processor_id()] = new->new[smp_processor_id()];
	new->new[smp_processor_id()] = old;
}

3281
static int do_tune_cpucache(struct kmem_cache *cachep, int limit, int batchcount,
P
Pekka Enberg 已提交
3282
			    int shared)
L
Linus Torvalds 已提交
3283 3284
{
	struct ccupdate_struct new;
3285
	int i, err;
L
Linus Torvalds 已提交
3286

P
Pekka Enberg 已提交
3287
	memset(&new.new, 0, sizeof(new.new));
3288
	for_each_online_cpu(i) {
P
Pekka Enberg 已提交
3289 3290
		new.new[i] =
		    alloc_arraycache(cpu_to_node(i), limit, batchcount);
3291
		if (!new.new[i]) {
P
Pekka Enberg 已提交
3292 3293
			for (i--; i >= 0; i--)
				kfree(new.new[i]);
3294
			return -ENOMEM;
L
Linus Torvalds 已提交
3295 3296 3297 3298 3299
		}
	}
	new.cachep = cachep;

	smp_call_function_all_cpus(do_ccupdate_local, (void *)&new);
3300

L
Linus Torvalds 已提交
3301 3302 3303 3304
	check_irq_on();
	spin_lock_irq(&cachep->spinlock);
	cachep->batchcount = batchcount;
	cachep->limit = limit;
3305
	cachep->shared = shared;
L
Linus Torvalds 已提交
3306 3307
	spin_unlock_irq(&cachep->spinlock);

3308
	for_each_online_cpu(i) {
L
Linus Torvalds 已提交
3309 3310 3311
		struct array_cache *ccold = new.new[i];
		if (!ccold)
			continue;
3312
		spin_lock_irq(&cachep->nodelists[cpu_to_node(i)]->list_lock);
3313
		free_block(cachep, ccold->entry, ccold->avail, cpu_to_node(i));
3314
		spin_unlock_irq(&cachep->nodelists[cpu_to_node(i)]->list_lock);
L
Linus Torvalds 已提交
3315 3316 3317
		kfree(ccold);
	}

3318 3319 3320
	err = alloc_kmemlist(cachep);
	if (err) {
		printk(KERN_ERR "alloc_kmemlist failed for %s, error %d.\n",
P
Pekka Enberg 已提交
3321
		       cachep->name, -err);
3322
		BUG();
L
Linus Torvalds 已提交
3323 3324 3325 3326
	}
	return 0;
}

3327
static void enable_cpucache(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339
{
	int err;
	int limit, shared;

	/* The head array serves three purposes:
	 * - create a LIFO ordering, i.e. return objects that are cache-warm
	 * - reduce the number of spinlock operations.
	 * - reduce the number of linked list operations on the slab and 
	 *   bufctl chains: array operations are cheaper.
	 * The numbers are guessed, we should auto-tune as described by
	 * Bonwick.
	 */
3340
	if (cachep->buffer_size > 131072)
L
Linus Torvalds 已提交
3341
		limit = 1;
3342
	else if (cachep->buffer_size > PAGE_SIZE)
L
Linus Torvalds 已提交
3343
		limit = 8;
3344
	else if (cachep->buffer_size > 1024)
L
Linus Torvalds 已提交
3345
		limit = 24;
3346
	else if (cachep->buffer_size > 256)
L
Linus Torvalds 已提交
3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360
		limit = 54;
	else
		limit = 120;

	/* Cpu bound tasks (e.g. network routing) can exhibit cpu bound
	 * allocation behaviour: Most allocs on one cpu, most free operations
	 * on another cpu. For these cases, an efficient object passing between
	 * cpus is necessary. This is provided by a shared array. The array
	 * replaces Bonwick's magazine layer.
	 * On uniprocessor, it's functionally equivalent (but less efficient)
	 * to a larger limit. Thus disabled by default.
	 */
	shared = 0;
#ifdef CONFIG_SMP
3361
	if (cachep->buffer_size <= PAGE_SIZE)
L
Linus Torvalds 已提交
3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372
		shared = 8;
#endif

#if DEBUG
	/* With debugging enabled, large batchcount lead to excessively
	 * long periods with disabled local interrupts. Limit the 
	 * batchcount
	 */
	if (limit > 32)
		limit = 32;
#endif
P
Pekka Enberg 已提交
3373
	err = do_tune_cpucache(cachep, limit, (limit + 1) / 2, shared);
L
Linus Torvalds 已提交
3374 3375
	if (err)
		printk(KERN_ERR "enable_cpucache failed for %s, error %d.\n",
P
Pekka Enberg 已提交
3376
		       cachep->name, -err);
L
Linus Torvalds 已提交
3377 3378
}

3379
static void drain_array_locked(struct kmem_cache *cachep, struct array_cache *ac,
P
Pekka Enberg 已提交
3380
				int force, int node)
L
Linus Torvalds 已提交
3381 3382 3383
{
	int tofree;

3384
	check_spinlock_acquired_node(cachep, node);
L
Linus Torvalds 已提交
3385 3386 3387
	if (ac->touched && !force) {
		ac->touched = 0;
	} else if (ac->avail) {
P
Pekka Enberg 已提交
3388
		tofree = force ? ac->avail : (ac->limit + 4) / 5;
L
Linus Torvalds 已提交
3389
		if (tofree > ac->avail) {
P
Pekka Enberg 已提交
3390
			tofree = (ac->avail + 1) / 2;
L
Linus Torvalds 已提交
3391
		}
3392
		free_block(cachep, ac->entry, tofree, node);
L
Linus Torvalds 已提交
3393
		ac->avail -= tofree;
3394
		memmove(ac->entry, &(ac->entry[tofree]),
P
Pekka Enberg 已提交
3395
			sizeof(void *) * ac->avail);
L
Linus Torvalds 已提交
3396 3397 3398 3399 3400
	}
}

/**
 * cache_reap - Reclaim memory from caches.
3401
 * @unused: unused parameter
L
Linus Torvalds 已提交
3402 3403 3404 3405 3406 3407
 *
 * 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.
 *
I
Ingo Molnar 已提交
3408
 * If we cannot acquire the cache chain mutex then just give up - we'll
L
Linus Torvalds 已提交
3409 3410 3411 3412 3413
 * try again on the next iteration.
 */
static void cache_reap(void *unused)
{
	struct list_head *walk;
3414
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
3415

I
Ingo Molnar 已提交
3416
	if (!mutex_trylock(&cache_chain_mutex)) {
L
Linus Torvalds 已提交
3417
		/* Give up. Setup the next iteration. */
P
Pekka Enberg 已提交
3418 3419
		schedule_delayed_work(&__get_cpu_var(reap_work),
				      REAPTIMEOUT_CPUC);
L
Linus Torvalds 已提交
3420 3421 3422 3423
		return;
	}

	list_for_each(walk, &cache_chain) {
3424
		struct kmem_cache *searchp;
P
Pekka Enberg 已提交
3425
		struct list_head *p;
L
Linus Torvalds 已提交
3426 3427 3428
		int tofree;
		struct slab *slabp;

3429
		searchp = list_entry(walk, struct kmem_cache, next);
L
Linus Torvalds 已提交
3430 3431 3432 3433 3434 3435

		if (searchp->flags & SLAB_NO_REAP)
			goto next;

		check_irq_on();

3436 3437 3438 3439
		l3 = searchp->nodelists[numa_node_id()];
		if (l3->alien)
			drain_alien_cache(searchp, l3);
		spin_lock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
3440

3441
		drain_array_locked(searchp, cpu_cache_get(searchp), 0,
P
Pekka Enberg 已提交
3442
				   numa_node_id());
L
Linus Torvalds 已提交
3443

3444
		if (time_after(l3->next_reap, jiffies))
L
Linus Torvalds 已提交
3445 3446
			goto next_unlock;

3447
		l3->next_reap = jiffies + REAPTIMEOUT_LIST3;
L
Linus Torvalds 已提交
3448

3449 3450
		if (l3->shared)
			drain_array_locked(searchp, l3->shared, 0,
P
Pekka Enberg 已提交
3451
					   numa_node_id());
L
Linus Torvalds 已提交
3452

3453 3454
		if (l3->free_touched) {
			l3->free_touched = 0;
L
Linus Torvalds 已提交
3455 3456 3457
			goto next_unlock;
		}

P
Pekka Enberg 已提交
3458 3459 3460
		tofree =
		    (l3->free_limit + 5 * searchp->num -
		     1) / (5 * searchp->num);
L
Linus Torvalds 已提交
3461
		do {
3462 3463
			p = l3->slabs_free.next;
			if (p == &(l3->slabs_free))
L
Linus Torvalds 已提交
3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475
				break;

			slabp = list_entry(p, struct slab, list);
			BUG_ON(slabp->inuse);
			list_del(&slabp->list);
			STATS_INC_REAPED(searchp);

			/* Safe to drop the lock. The slab is no longer
			 * linked to the cache.
			 * searchp cannot disappear, we hold
			 * cache_chain_lock
			 */
3476 3477
			l3->free_objects -= searchp->num;
			spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
3478
			slab_destroy(searchp, slabp);
3479
			spin_lock_irq(&l3->list_lock);
P
Pekka Enberg 已提交
3480 3481
		} while (--tofree > 0);
	      next_unlock:
3482
		spin_unlock_irq(&l3->list_lock);
P
Pekka Enberg 已提交
3483
	      next:
L
Linus Torvalds 已提交
3484 3485 3486
		cond_resched();
	}
	check_irq_on();
I
Ingo Molnar 已提交
3487
	mutex_unlock(&cache_chain_mutex);
3488
	drain_remote_pages();
L
Linus Torvalds 已提交
3489
	/* Setup the next iteration */
3490
	schedule_delayed_work(&__get_cpu_var(reap_work), REAPTIMEOUT_CPUC);
L
Linus Torvalds 已提交
3491 3492 3493 3494
}

#ifdef CONFIG_PROC_FS

3495
static void print_slabinfo_header(struct seq_file *m)
L
Linus Torvalds 已提交
3496
{
3497 3498 3499 3500
	/*
	 * Output format version, so at least we can change it
	 * without _too_ many complaints.
	 */
L
Linus Torvalds 已提交
3501
#if STATS
3502
	seq_puts(m, "slabinfo - version: 2.1 (statistics)\n");
L
Linus Torvalds 已提交
3503
#else
3504
	seq_puts(m, "slabinfo - version: 2.1\n");
L
Linus Torvalds 已提交
3505
#endif
3506 3507 3508 3509
	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
3510
#if STATS
3511 3512 3513
	seq_puts(m, " : globalstat <listallocs> <maxobjs> <grown> <reaped> "
		 "<error> <maxfreeable> <nodeallocs> <remotefrees>");
	seq_puts(m, " : cpustat <allochit> <allocmiss> <freehit> <freemiss>");
L
Linus Torvalds 已提交
3514
#endif
3515 3516 3517 3518 3519 3520 3521 3522
	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
3523
	mutex_lock(&cache_chain_mutex);
3524 3525
	if (!n)
		print_slabinfo_header(m);
L
Linus Torvalds 已提交
3526 3527 3528 3529 3530 3531
	p = cache_chain.next;
	while (n--) {
		p = p->next;
		if (p == &cache_chain)
			return NULL;
	}
3532
	return list_entry(p, struct kmem_cache, next);
L
Linus Torvalds 已提交
3533 3534 3535 3536
}

static void *s_next(struct seq_file *m, void *p, loff_t *pos)
{
3537
	struct kmem_cache *cachep = p;
L
Linus Torvalds 已提交
3538 3539
	++*pos;
	return cachep->next.next == &cache_chain ? NULL
3540
	    : list_entry(cachep->next.next, struct kmem_cache, next);
L
Linus Torvalds 已提交
3541 3542 3543 3544
}

static void s_stop(struct seq_file *m, void *p)
{
I
Ingo Molnar 已提交
3545
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
3546 3547 3548 3549
}

static int s_show(struct seq_file *m, void *p)
{
3550
	struct kmem_cache *cachep = p;
L
Linus Torvalds 已提交
3551
	struct list_head *q;
P
Pekka Enberg 已提交
3552 3553 3554 3555 3556
	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;
3557
	const char *name;
L
Linus Torvalds 已提交
3558
	char *error = NULL;
3559 3560
	int node;
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
3561 3562 3563 3564 3565

	check_irq_on();
	spin_lock_irq(&cachep->spinlock);
	active_objs = 0;
	num_slabs = 0;
3566 3567 3568 3569 3570 3571 3572
	for_each_online_node(node) {
		l3 = cachep->nodelists[node];
		if (!l3)
			continue;

		spin_lock(&l3->list_lock);

P
Pekka Enberg 已提交
3573
		list_for_each(q, &l3->slabs_full) {
3574 3575 3576 3577 3578 3579
			slabp = list_entry(q, struct slab, list);
			if (slabp->inuse != cachep->num && !error)
				error = "slabs_full accounting error";
			active_objs += cachep->num;
			active_slabs++;
		}
P
Pekka Enberg 已提交
3580
		list_for_each(q, &l3->slabs_partial) {
3581 3582 3583 3584 3585 3586 3587 3588
			slabp = list_entry(q, struct slab, list);
			if (slabp->inuse == cachep->num && !error)
				error = "slabs_partial inuse accounting error";
			if (!slabp->inuse && !error)
				error = "slabs_partial/inuse accounting error";
			active_objs += slabp->inuse;
			active_slabs++;
		}
P
Pekka Enberg 已提交
3589
		list_for_each(q, &l3->slabs_free) {
3590 3591 3592 3593 3594 3595 3596 3597 3598
			slabp = list_entry(q, struct slab, list);
			if (slabp->inuse && !error)
				error = "slabs_free/inuse accounting error";
			num_slabs++;
		}
		free_objects += l3->free_objects;
		shared_avail += l3->shared->avail;

		spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
3599
	}
P
Pekka Enberg 已提交
3600 3601
	num_slabs += active_slabs;
	num_objs = num_slabs * cachep->num;
3602
	if (num_objs - active_objs != free_objects && !error)
L
Linus Torvalds 已提交
3603 3604
		error = "free_objects accounting error";

P
Pekka Enberg 已提交
3605
	name = cachep->name;
L
Linus Torvalds 已提交
3606 3607 3608 3609
	if (error)
		printk(KERN_ERR "slab: cache %s error: %s\n", name, error);

	seq_printf(m, "%-17s %6lu %6lu %6u %4u %4d",
3610
		   name, active_objs, num_objs, cachep->buffer_size,
P
Pekka Enberg 已提交
3611
		   cachep->num, (1 << cachep->gfporder));
L
Linus Torvalds 已提交
3612
	seq_printf(m, " : tunables %4u %4u %4u",
P
Pekka Enberg 已提交
3613
		   cachep->limit, cachep->batchcount, cachep->shared);
3614
	seq_printf(m, " : slabdata %6lu %6lu %6lu",
P
Pekka Enberg 已提交
3615
		   active_slabs, num_slabs, shared_avail);
L
Linus Torvalds 已提交
3616
#if STATS
P
Pekka Enberg 已提交
3617
	{			/* list3 stats */
L
Linus Torvalds 已提交
3618 3619 3620 3621 3622 3623 3624
		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;
3625
		unsigned long node_frees = cachep->node_frees;
L
Linus Torvalds 已提交
3626

3627
		seq_printf(m, " : globalstat %7lu %6lu %5lu %4lu \
P
Pekka Enberg 已提交
3628
				%4lu %4lu %4lu %4lu", allocs, high, grown, reaped, errors, max_freeable, node_allocs, node_frees);
L
Linus Torvalds 已提交
3629 3630 3631 3632 3633 3634 3635 3636 3637
	}
	/* 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
3638
			   allochit, allocmiss, freehit, freemiss);
L
Linus Torvalds 已提交
3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660
	}
#endif
	seq_putc(m, '\n');
	spin_unlock_irq(&cachep->spinlock);
	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
 */

struct seq_operations slabinfo_op = {
P
Pekka Enberg 已提交
3661 3662 3663 3664
	.start = s_start,
	.next = s_next,
	.stop = s_stop,
	.show = s_show,
L
Linus Torvalds 已提交
3665 3666 3667 3668 3669 3670 3671 3672 3673 3674
};

#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
3675 3676
ssize_t slabinfo_write(struct file *file, const char __user * buffer,
		       size_t count, loff_t *ppos)
L
Linus Torvalds 已提交
3677
{
P
Pekka Enberg 已提交
3678
	char kbuf[MAX_SLABINFO_WRITE + 1], *tmp;
L
Linus Torvalds 已提交
3679 3680
	int limit, batchcount, shared, res;
	struct list_head *p;
P
Pekka Enberg 已提交
3681

L
Linus Torvalds 已提交
3682 3683 3684 3685
	if (count > MAX_SLABINFO_WRITE)
		return -EINVAL;
	if (copy_from_user(&kbuf, buffer, count))
		return -EFAULT;
P
Pekka Enberg 已提交
3686
	kbuf[MAX_SLABINFO_WRITE] = '\0';
L
Linus Torvalds 已提交
3687 3688 3689 3690 3691 3692 3693 3694 3695 3696

	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
3697
	mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
3698
	res = -EINVAL;
P
Pekka Enberg 已提交
3699
	list_for_each(p, &cache_chain) {
3700 3701
		struct kmem_cache *cachep = list_entry(p, struct kmem_cache,
						       next);
L
Linus Torvalds 已提交
3702 3703 3704 3705

		if (!strcmp(cachep->name, kbuf)) {
			if (limit < 1 ||
			    batchcount < 1 ||
P
Pekka Enberg 已提交
3706
			    batchcount > limit || shared < 0) {
3707
				res = 0;
L
Linus Torvalds 已提交
3708
			} else {
3709
				res = do_tune_cpucache(cachep, limit,
P
Pekka Enberg 已提交
3710
						       batchcount, shared);
L
Linus Torvalds 已提交
3711 3712 3713 3714
			}
			break;
		}
	}
I
Ingo Molnar 已提交
3715
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
3716 3717 3718 3719 3720 3721
	if (res >= 0)
		res = count;
	return res;
}
#endif

3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733
/**
 * 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
3734 3735
unsigned int ksize(const void *objp)
{
3736 3737
	if (unlikely(objp == NULL))
		return 0;
L
Linus Torvalds 已提交
3738

3739
	return obj_size(virt_to_cache(objp));
L
Linus Torvalds 已提交
3740
}