slab.c 101.7 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52
/*
 * linux/mm/slab.c
 * Written by Mark Hemment, 1996/97.
 * (markhe@nextd.demon.co.uk)
 *
 * kmem_cache_destroy() + some cleanup - 1999 Andrea Arcangeli
 *
 * Major cleanup, different bufctl logic, per-cpu arrays
 *	(c) 2000 Manfred Spraul
 *
 * Cleanup, make the head arrays unconditional, preparation for NUMA
 * 	(c) 2002 Manfred Spraul
 *
 * An implementation of the Slab Allocator as described in outline in;
 *	UNIX Internals: The New Frontiers by Uresh Vahalia
 *	Pub: Prentice Hall	ISBN 0-13-101908-2
 * or with a little more detail in;
 *	The Slab Allocator: An Object-Caching Kernel Memory Allocator
 *	Jeff Bonwick (Sun Microsystems).
 *	Presented at: USENIX Summer 1994 Technical Conference
 *
 * The memory is organized in caches, one cache for each object type.
 * (e.g. inode_cache, dentry_cache, buffer_head, vm_area_struct)
 * Each cache consists out of many slabs (they are small (usually one
 * page long) and always contiguous), and each slab contains multiple
 * initialized objects.
 *
 * This means, that your constructor is used only for newly allocated
 * slabs and you must pass objects with the same intializations to
 * kmem_cache_free.
 *
 * Each cache can only support one memory type (GFP_DMA, GFP_HIGHMEM,
 * normal). If you need a special memory type, then must create a new
 * cache for that memory type.
 *
 * In order to reduce fragmentation, the slabs are sorted in 3 groups:
 *   full slabs with 0 free objects
 *   partial slabs
 *   empty slabs with no allocated objects
 *
 * If partial slabs exist, then new allocations come from these slabs,
 * otherwise from empty slabs or new slabs are allocated.
 *
 * kmem_cache_destroy() CAN CRASH if you try to allocate from the cache
 * during kmem_cache_destroy(). The caller must prevent concurrent allocs.
 *
 * Each cache has a short per-cpu head array, most allocs
 * and frees go into that array, and if that array overflows, then 1/2
 * of the entries in the array are given back into the global cache.
 * The head array is strictly LIFO and should improve the cache hit rates.
 * On SMP, it additionally reduces the spinlock operations.
 *
A
Andrew Morton 已提交
53
 * The c_cpuarray may not be read with enabled local interrupts -
L
Linus Torvalds 已提交
54 55 56 57
 * it's changed with a smp_call_function().
 *
 * SMP synchronization:
 *  constructors and destructors are called without any locking.
58
 *  Several members in struct kmem_cache and struct slab never change, they
L
Linus Torvalds 已提交
59 60 61 62 63 64 65 66 67 68 69 70
 *	are accessed without any locking.
 *  The per-cpu arrays are never accessed from the wrong cpu, no locking,
 *  	and local interrupts are disabled so slab code is preempt-safe.
 *  The non-constant members are protected with a per-cache irq spinlock.
 *
 * Many thanks to Mark Hemment, who wrote another per-cpu slab patch
 * in 2000 - many ideas in the current implementation are derived from
 * his patch.
 *
 * Further notes from the original documentation:
 *
 * 11 April '97.  Started multi-threading - markhe
I
Ingo Molnar 已提交
71
 *	The global cache-chain is protected by the mutex 'cache_chain_mutex'.
L
Linus Torvalds 已提交
72 73 74 75 76 77
 *	The sem is only needed when accessing/extending the cache-chain, which
 *	can never happen inside an interrupt (kmem_cache_create(),
 *	kmem_cache_shrink() and kmem_cache_reap()).
 *
 *	At present, each engine can be growing a cache.  This should be blocked.
 *
78 79 80 81 82 83 84 85 86
 * 15 March 2005. NUMA slab allocator.
 *	Shai Fultheim <shai@scalex86.org>.
 *	Shobhit Dayal <shobhit@calsoftinc.com>
 *	Alok N Kataria <alokk@calsoftinc.com>
 *	Christoph Lameter <christoph@lameter.com>
 *
 *	Modified the slab allocator to be node aware on NUMA systems.
 *	Each node has its own list of partial, free and full slabs.
 *	All object allocations for a node occur from node specific slab lists.
L
Linus Torvalds 已提交
87 88 89 90 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

#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 | \
173
			 SLAB_CACHE_DMA | \
L
Linus Torvalds 已提交
174 175 176 177
			 SLAB_MUST_HWCACHE_ALIGN | SLAB_STORE_USER | \
			 SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
			 SLAB_DESTROY_BY_RCU)
#else
178
# define CREATE_MASK	(SLAB_HWCACHE_ALIGN | \
L
Linus Torvalds 已提交
179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202
			 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
	spinlock_t lock;
A
Andrew Morton 已提交
269 270 271 272 273 274
	void *entry[0];	/*
			 * Must have this definition in here for the proper
			 * alignment of array_cache. Also simplifies accessing
			 * the entries.
			 * [0] is for gcc 2.95. It should really be [].
			 */
L
Linus Torvalds 已提交
275 276
};

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

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

304 305 306 307 308 309 310 311 312 313
/*
 * 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)

/*
A
Andrew Morton 已提交
314 315
 * This function must be completely optimized away if a constant is passed to
 * it.  Mostly the same as what is in linux/slab.h except it returns an index.
316
 */
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
{
	INIT_LIST_HEAD(&parent->slabs_full);
	INIT_LIST_HEAD(&parent->slabs_partial);
	INIT_LIST_HEAD(&parent->slabs_free);
	parent->shared = NULL;
	parent->alien = NULL;
347
	parent->colour_next = 0;
348 349 350 351 352
	spin_lock_init(&parent->list_lock);
	parent->free_objects = 0;
	parent->free_touched = 0;
}

A
Andrew Morton 已提交
353 354 355 356
#define MAKE_LIST(cachep, listp, slab, nodeid)				\
	do {								\
		INIT_LIST_HEAD(listp);					\
		list_splice(&(cachep->nodelists[nodeid]->slab), listp);	\
357 358
	} while (0)

A
Andrew Morton 已提交
359 360
#define	MAKE_ALL_LISTS(cachep, ptr, nodeid)				\
	do {								\
361 362 363 364
	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
365 366

/*
367
 * struct kmem_cache
L
Linus Torvalds 已提交
368 369 370
 *
 * manages a cache.
 */
P
Pekka Enberg 已提交
371

372
struct kmem_cache {
L
Linus Torvalds 已提交
373
/* 1) per-cpu data, touched during every alloc/free */
P
Pekka Enberg 已提交
374
	struct array_cache *array[NR_CPUS];
375
/* 2) Cache tunables. Protected by cache_chain_mutex */
P
Pekka Enberg 已提交
376 377 378
	unsigned int batchcount;
	unsigned int limit;
	unsigned int shared;
379

380
	unsigned int buffer_size;
381
/* 3) touched by every alloc & free from the backend */
P
Pekka Enberg 已提交
382
	struct kmem_list3 *nodelists[MAX_NUMNODES];
383

A
Andrew Morton 已提交
384 385
	unsigned int flags;		/* constant flags */
	unsigned int num;		/* # of objs per slab */
L
Linus Torvalds 已提交
386

387
/* 4) cache_grow/shrink */
L
Linus Torvalds 已提交
388
	/* order of pgs per slab (2^n) */
P
Pekka Enberg 已提交
389
	unsigned int gfporder;
L
Linus Torvalds 已提交
390 391

	/* force GFP flags, e.g. GFP_DMA */
P
Pekka Enberg 已提交
392
	gfp_t gfpflags;
L
Linus Torvalds 已提交
393

A
Andrew Morton 已提交
394
	size_t colour;			/* cache colouring range */
P
Pekka Enberg 已提交
395
	unsigned int colour_off;	/* colour offset */
396
	struct kmem_cache *slabp_cache;
P
Pekka Enberg 已提交
397
	unsigned int slab_size;
A
Andrew Morton 已提交
398
	unsigned int dflags;		/* dynamic flags */
L
Linus Torvalds 已提交
399 400

	/* constructor func */
401
	void (*ctor) (void *, struct kmem_cache *, unsigned long);
L
Linus Torvalds 已提交
402 403

	/* de-constructor func */
404
	void (*dtor) (void *, struct kmem_cache *, unsigned long);
L
Linus Torvalds 已提交
405

406
/* 5) cache creation/removal */
P
Pekka Enberg 已提交
407 408
	const char *name;
	struct list_head next;
L
Linus Torvalds 已提交
409

410
/* 6) statistics */
L
Linus Torvalds 已提交
411
#if STATS
P
Pekka Enberg 已提交
412 413 414 415 416 417 418 419 420 421 422 423 424
	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
425 426
#endif
#if DEBUG
427 428 429 430 431 432 433 434
	/*
	 * 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
435 436 437 438 439 440 441
#endif
};

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

#define BATCHREFILL_LIMIT	16
A
Andrew Morton 已提交
442 443 444
/*
 * Optimization question: fewer reaps means less probability for unnessary
 * cpucache drain/refill cycles.
L
Linus Torvalds 已提交
445
 *
A
Adrian Bunk 已提交
446
 * OTOH the cpuarrays can contain lots of objects,
L
Linus Torvalds 已提交
447 448 449 450 451 452 453 454 455 456 457
 * 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++)
A
Andrew Morton 已提交
458 459 460 461 462
#define	STATS_SET_HIGH(x)						\
	do {								\
		if ((x)->num_active > (x)->high_mark)			\
			(x)->high_mark = (x)->num_active;		\
	} while (0)
L
Linus Torvalds 已提交
463 464
#define	STATS_INC_ERR(x)	((x)->errors++)
#define	STATS_INC_NODEALLOCS(x)	((x)->node_allocs++)
465
#define	STATS_INC_NODEFREES(x)	((x)->node_frees++)
A
Andrew Morton 已提交
466 467 468 469 470
#define	STATS_SET_FREEABLE(x, i)					\
	do {								\
		if ((x)->max_freeable < i)				\
			(x)->max_freeable = i;				\
	} while (0)
L
Linus Torvalds 已提交
471 472 473 474 475 476 477 478 479 480 481 482 483
#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)
484
#define	STATS_INC_NODEFREES(x)	do { } while (0)
A
Andrew Morton 已提交
485
#define	STATS_SET_FREEABLE(x, i) do { } while (0)
L
Linus Torvalds 已提交
486 487 488 489 490 491 492
#define STATS_INC_ALLOCHIT(x)	do { } while (0)
#define STATS_INC_ALLOCMISS(x)	do { } while (0)
#define STATS_INC_FREEHIT(x)	do { } while (0)
#define STATS_INC_FREEMISS(x)	do { } while (0)
#endif

#if DEBUG
A
Andrew Morton 已提交
493 494
/*
 * Magic nums for obj red zoning.
L
Linus Torvalds 已提交
495 496 497 498 499 500 501 502 503 504
 * 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 */

A
Andrew Morton 已提交
505 506
/*
 * memory layout of objects:
L
Linus Torvalds 已提交
507
 * 0		: objp
508
 * 0 .. cachep->obj_offset - BYTES_PER_WORD - 1: padding. This ensures that
L
Linus Torvalds 已提交
509 510
 * 		the end of an object is aligned with the end of the real
 * 		allocation. Catches writes behind the end of the allocation.
511
 * cachep->obj_offset - BYTES_PER_WORD .. cachep->obj_offset - 1:
L
Linus Torvalds 已提交
512
 * 		redzone word.
513 514
 * cachep->obj_offset: The real object.
 * cachep->buffer_size - 2* BYTES_PER_WORD: redzone word [BYTES_PER_WORD long]
A
Andrew Morton 已提交
515 516
 * cachep->buffer_size - 1* BYTES_PER_WORD: last caller address
 *					[BYTES_PER_WORD long]
L
Linus Torvalds 已提交
517
 */
518
static int obj_offset(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
519
{
520
	return cachep->obj_offset;
L
Linus Torvalds 已提交
521 522
}

523
static int obj_size(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
524
{
525
	return cachep->obj_size;
L
Linus Torvalds 已提交
526 527
}

528
static unsigned long *dbg_redzone1(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
529 530
{
	BUG_ON(!(cachep->flags & SLAB_RED_ZONE));
531
	return (unsigned long*) (objp+obj_offset(cachep)-BYTES_PER_WORD);
L
Linus Torvalds 已提交
532 533
}

534
static unsigned long *dbg_redzone2(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
535 536 537
{
	BUG_ON(!(cachep->flags & SLAB_RED_ZONE));
	if (cachep->flags & SLAB_STORE_USER)
538
		return (unsigned long *)(objp + cachep->buffer_size -
P
Pekka Enberg 已提交
539
					 2 * BYTES_PER_WORD);
540
	return (unsigned long *)(objp + cachep->buffer_size - BYTES_PER_WORD);
L
Linus Torvalds 已提交
541 542
}

543
static void **dbg_userword(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
544 545
{
	BUG_ON(!(cachep->flags & SLAB_STORE_USER));
546
	return (void **)(objp + cachep->buffer_size - BYTES_PER_WORD);
L
Linus Torvalds 已提交
547 548 549 550
}

#else

551 552
#define obj_offset(x)			0
#define obj_size(cachep)		(cachep->buffer_size)
L
Linus Torvalds 已提交
553 554 555 556 557 558 559
#define dbg_redzone1(cachep, objp)	({BUG(); (unsigned long *)NULL;})
#define dbg_redzone2(cachep, objp)	({BUG(); (unsigned long *)NULL;})
#define dbg_userword(cachep, objp)	({BUG(); (void **)NULL;})

#endif

/*
A
Andrew Morton 已提交
560 561
 * Maximum size of an obj (in 2^order pages) and absolute limit for the gfp
 * order.
L
Linus Torvalds 已提交
562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580
 */
#if defined(CONFIG_LARGE_ALLOCS)
#define	MAX_OBJ_ORDER	13	/* up to 32Mb */
#define	MAX_GFP_ORDER	13	/* up to 32Mb */
#elif defined(CONFIG_MMU)
#define	MAX_OBJ_ORDER	5	/* 32 pages */
#define	MAX_GFP_ORDER	5	/* 32 pages */
#else
#define	MAX_OBJ_ORDER	8	/* up to 1Mb */
#define	MAX_GFP_ORDER	8	/* up to 1Mb */
#endif

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

A
Andrew Morton 已提交
581 582 583 584
/*
 * Functions for storing/retrieving the cachep and or slab from the page
 * allocator.  These are used to find the slab an obj belongs to.  With kfree(),
 * these are used to find the cache which an obj belongs to.
L
Linus Torvalds 已提交
585
 */
586 587 588 589 590 591 592
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)
{
593 594
	if (unlikely(PageCompound(page)))
		page = (struct page *)page_private(page);
595 596 597 598 599 600 601 602 603 604
	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)
{
605 606
	if (unlikely(PageCompound(page)))
		page = (struct page *)page_private(page);
607 608
	return (struct slab *)page->lru.prev;
}
L
Linus Torvalds 已提交
609

610 611 612 613 614 615 616 617 618 619 620 621
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);
}

622 623 624 625 626 627 628 629 630 631 632 633
static inline void *index_to_obj(struct kmem_cache *cache, struct slab *slab,
				 unsigned int idx)
{
	return slab->s_mem + cache->buffer_size * idx;
}

static inline unsigned int obj_to_index(struct kmem_cache *cache,
					struct slab *slab, void *obj)
{
	return (unsigned)(obj - slab->s_mem) / cache->buffer_size;
}

A
Andrew Morton 已提交
634 635 636
/*
 * These are the default caches for kmalloc. Custom caches can have other sizes.
 */
L
Linus Torvalds 已提交
637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653
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
654
	{NULL,}
L
Linus Torvalds 已提交
655 656 657 658
#undef CACHE
};

static struct arraycache_init initarray_cache __initdata =
P
Pekka Enberg 已提交
659
    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
L
Linus Torvalds 已提交
660
static struct arraycache_init initarray_generic =
P
Pekka Enberg 已提交
661
    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
L
Linus Torvalds 已提交
662 663

/* internal cache of cache description objs */
664
static struct kmem_cache cache_cache = {
P
Pekka Enberg 已提交
665 666 667
	.batchcount = 1,
	.limit = BOOT_CPUCACHE_ENTRIES,
	.shared = 1,
668
	.buffer_size = sizeof(struct kmem_cache),
P
Pekka Enberg 已提交
669
	.name = "kmem_cache",
L
Linus Torvalds 已提交
670
#if DEBUG
671
	.obj_size = sizeof(struct kmem_cache),
L
Linus Torvalds 已提交
672 673 674 675
#endif
};

/* Guard access to the cache-chain. */
I
Ingo Molnar 已提交
676
static DEFINE_MUTEX(cache_chain_mutex);
L
Linus Torvalds 已提交
677 678 679
static struct list_head cache_chain;

/*
A
Andrew Morton 已提交
680 681
 * vm_enough_memory() looks at this to determine how many slab-allocated pages
 * are possibly freeable under pressure
L
Linus Torvalds 已提交
682 683 684 685 686 687 688 689 690 691 692
 *
 * 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,
693 694
	PARTIAL_AC,
	PARTIAL_L3,
L
Linus Torvalds 已提交
695 696 697 698 699
	FULL
} g_cpucache_up;

static DEFINE_PER_CPU(struct work_struct, reap_work);

A
Andrew Morton 已提交
700 701
static void free_block(struct kmem_cache *cachep, void **objpp, int len,
			int node);
702
static void enable_cpucache(struct kmem_cache *cachep);
P
Pekka Enberg 已提交
703
static void cache_reap(void *unused);
704
static int __node_shrink(struct kmem_cache *cachep, int node);
L
Linus Torvalds 已提交
705

706
static inline struct array_cache *cpu_cache_get(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
707 708 709 710
{
	return cachep->array[smp_processor_id()];
}

A
Andrew Morton 已提交
711 712
static inline struct kmem_cache *__find_general_cachep(size_t size,
							gfp_t gfpflags)
L
Linus Torvalds 已提交
713 714 715 716 717
{
	struct cache_sizes *csizep = malloc_sizes;

#if DEBUG
	/* This happens if someone tries to call
P
Pekka Enberg 已提交
718 719 720
	 * kmem_cache_create(), or __kmalloc(), before
	 * the generic caches are initialized.
	 */
721
	BUG_ON(malloc_sizes[INDEX_AC].cs_cachep == NULL);
L
Linus Torvalds 已提交
722 723 724 725 726
#endif
	while (size > csizep->cs_size)
		csizep++;

	/*
727
	 * Really subtle: The last entry with cs->cs_size==ULONG_MAX
L
Linus Torvalds 已提交
728 729 730 731 732 733 734 735
	 * 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;
}

736
struct kmem_cache *kmem_find_general_cachep(size_t size, gfp_t gfpflags)
737 738 739 740 741
{
	return __find_general_cachep(size, gfpflags);
}
EXPORT_SYMBOL(kmem_find_general_cachep);

742
static size_t slab_mgmt_size(size_t nr_objs, size_t align)
L
Linus Torvalds 已提交
743
{
744 745
	return ALIGN(sizeof(struct slab)+nr_objs*sizeof(kmem_bufctl_t), align);
}
L
Linus Torvalds 已提交
746

A
Andrew Morton 已提交
747 748 749
/*
 * Calculate the number of objects and left-over bytes for a given buffer size.
 */
750 751 752 753 754 755 756
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
757

758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805
	/*
	 * 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
806 807 808 809
}

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

A
Andrew Morton 已提交
810 811
static void __slab_error(const char *function, struct kmem_cache *cachep,
			char *msg)
L
Linus Torvalds 已提交
812 813
{
	printk(KERN_ERR "slab error in %s(): cache `%s': %s\n",
P
Pekka Enberg 已提交
814
	       function, cachep->name, msg);
L
Linus Torvalds 已提交
815 816 817
	dump_stack();
}

818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858
#ifdef CONFIG_NUMA
/*
 * Special reaping functions for NUMA systems called from cache_reap().
 * These take care of doing round robin flushing of alien caches (containing
 * objects freed on different nodes from which they were allocated) and the
 * flushing of remote pcps by calling drain_node_pages.
 */
static DEFINE_PER_CPU(unsigned long, reap_node);

static void init_reap_node(int cpu)
{
	int node;

	node = next_node(cpu_to_node(cpu), node_online_map);
	if (node == MAX_NUMNODES)
		node = 0;

	__get_cpu_var(reap_node) = node;
}

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

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

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

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

L
Linus Torvalds 已提交
859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875
/*
 * 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) {
876
		init_reap_node(cpu);
L
Linus Torvalds 已提交
877 878 879 880 881
		INIT_WORK(reap_work, cache_reap, NULL);
		schedule_delayed_work_on(cpu, reap_work, HZ + 3 * cpu);
	}
}

882
static struct array_cache *alloc_arraycache(int node, int entries,
P
Pekka Enberg 已提交
883
					    int batchcount)
L
Linus Torvalds 已提交
884
{
P
Pekka Enberg 已提交
885
	int memsize = sizeof(void *) * entries + sizeof(struct array_cache);
L
Linus Torvalds 已提交
886 887
	struct array_cache *nc = NULL;

888
	nc = kmalloc_node(memsize, GFP_KERNEL, node);
L
Linus Torvalds 已提交
889 890 891 892 893
	if (nc) {
		nc->avail = 0;
		nc->limit = entries;
		nc->batchcount = batchcount;
		nc->touched = 0;
894
		spin_lock_init(&nc->lock);
L
Linus Torvalds 已提交
895 896 897 898
	}
	return nc;
}

899
#ifdef CONFIG_NUMA
900
static void *__cache_alloc_node(struct kmem_cache *, gfp_t, int);
901

P
Pekka Enberg 已提交
902
static struct array_cache **alloc_alien_cache(int node, int limit)
903 904
{
	struct array_cache **ac_ptr;
P
Pekka Enberg 已提交
905
	int memsize = sizeof(void *) * MAX_NUMNODES;
906 907 908 909 910 911 912 913 914 915 916 917 918
	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
919
				for (i--; i <= 0; i--)
920 921 922 923 924 925 926 927 928
					kfree(ac_ptr[i]);
				kfree(ac_ptr);
				return NULL;
			}
		}
	}
	return ac_ptr;
}

P
Pekka Enberg 已提交
929
static void free_alien_cache(struct array_cache **ac_ptr)
930 931 932 933 934 935
{
	int i;

	if (!ac_ptr)
		return;
	for_each_node(i)
P
Pekka Enberg 已提交
936
	    kfree(ac_ptr[i]);
937 938 939
	kfree(ac_ptr);
}

940
static void __drain_alien_cache(struct kmem_cache *cachep,
P
Pekka Enberg 已提交
941
				struct array_cache *ac, int node)
942 943 944 945 946
{
	struct kmem_list3 *rl3 = cachep->nodelists[node];

	if (ac->avail) {
		spin_lock(&rl3->list_lock);
947
		free_block(cachep, ac->entry, ac->avail, node);
948 949 950 951 952
		ac->avail = 0;
		spin_unlock(&rl3->list_lock);
	}
}

953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969
/*
 * Called from cache_reap() to regularly drain alien caches round robin.
 */
static void reap_alien(struct kmem_cache *cachep, struct kmem_list3 *l3)
{
	int node = __get_cpu_var(reap_node);

	if (l3->alien) {
		struct array_cache *ac = l3->alien[node];
		if (ac && ac->avail) {
			spin_lock_irq(&ac->lock);
			__drain_alien_cache(cachep, ac, node);
			spin_unlock_irq(&ac->lock);
		}
	}
}

A
Andrew Morton 已提交
970 971
static void drain_alien_cache(struct kmem_cache *cachep,
				struct array_cache **alien)
972
{
P
Pekka Enberg 已提交
973
	int i = 0;
974 975 976 977
	struct array_cache *ac;
	unsigned long flags;

	for_each_online_node(i) {
978
		ac = alien[i];
979 980 981 982 983 984 985 986
		if (ac) {
			spin_lock_irqsave(&ac->lock, flags);
			__drain_alien_cache(cachep, ac, i);
			spin_unlock_irqrestore(&ac->lock, flags);
		}
	}
}
#else
987

988
#define drain_alien_cache(cachep, alien) do { } while (0)
989
#define reap_alien(cachep, l3) do { } while (0)
990

991 992 993 994 995
static inline struct array_cache **alloc_alien_cache(int node, int limit)
{
	return (struct array_cache **) 0x01020304ul;
}

996 997 998
static inline void free_alien_cache(struct array_cache **ac_ptr)
{
}
999

1000 1001
#endif

L
Linus Torvalds 已提交
1002
static int __devinit cpuup_callback(struct notifier_block *nfb,
P
Pekka Enberg 已提交
1003
				    unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
1004 1005
{
	long cpu = (long)hcpu;
1006
	struct kmem_cache *cachep;
1007 1008 1009
	struct kmem_list3 *l3 = NULL;
	int node = cpu_to_node(cpu);
	int memsize = sizeof(struct kmem_list3);
L
Linus Torvalds 已提交
1010 1011 1012

	switch (action) {
	case CPU_UP_PREPARE:
I
Ingo Molnar 已提交
1013
		mutex_lock(&cache_chain_mutex);
A
Andrew Morton 已提交
1014 1015
		/*
		 * We need to do this right in the beginning since
1016 1017 1018 1019 1020
		 * 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
1021
		list_for_each_entry(cachep, &cache_chain, next) {
A
Andrew Morton 已提交
1022 1023
			/*
			 * Set up the size64 kmemlist for cpu before we can
1024 1025 1026 1027
			 * begin anything. Make sure some other cpu on this
			 * node has not already allocated this
			 */
			if (!cachep->nodelists[node]) {
A
Andrew Morton 已提交
1028 1029
				l3 = kmalloc_node(memsize, GFP_KERNEL, node);
				if (!l3)
1030 1031 1032
					goto bad;
				kmem_list3_init(l3);
				l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
P
Pekka Enberg 已提交
1033
				    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
1034

1035 1036 1037 1038 1039
				/*
				 * The l3s don't come and go as CPUs come and
				 * go.  cache_chain_mutex is sufficient
				 * protection here.
				 */
1040 1041
				cachep->nodelists[node] = l3;
			}
L
Linus Torvalds 已提交
1042

1043 1044
			spin_lock_irq(&cachep->nodelists[node]->list_lock);
			cachep->nodelists[node]->free_limit =
A
Andrew Morton 已提交
1045 1046
				(1 + nr_cpus_node(node)) *
				cachep->batchcount + cachep->num;
1047 1048 1049
			spin_unlock_irq(&cachep->nodelists[node]->list_lock);
		}

A
Andrew Morton 已提交
1050 1051 1052 1053
		/*
		 * Now we can go ahead with allocating the shared arrays and
		 * array caches
		 */
1054
		list_for_each_entry(cachep, &cache_chain, next) {
1055
			struct array_cache *nc;
1056 1057
			struct array_cache *shared;
			struct array_cache **alien;
1058

1059
			nc = alloc_arraycache(node, cachep->limit,
1060
						cachep->batchcount);
L
Linus Torvalds 已提交
1061 1062
			if (!nc)
				goto bad;
1063 1064 1065 1066 1067
			shared = alloc_arraycache(node,
					cachep->shared * cachep->batchcount,
					0xbaadf00d);
			if (!shared)
				goto bad;
1068

1069 1070 1071
			alien = alloc_alien_cache(node, cachep->limit);
			if (!alien)
				goto bad;
L
Linus Torvalds 已提交
1072
			cachep->array[cpu] = nc;
1073 1074 1075
			l3 = cachep->nodelists[node];
			BUG_ON(!l3);

1076 1077 1078 1079 1080 1081 1082 1083
			spin_lock_irq(&l3->list_lock);
			if (!l3->shared) {
				/*
				 * We are serialised from CPU_DEAD or
				 * CPU_UP_CANCELLED by the cpucontrol lock
				 */
				l3->shared = shared;
				shared = NULL;
1084
			}
1085 1086 1087 1088 1089 1090 1091 1092 1093
#ifdef CONFIG_NUMA
			if (!l3->alien) {
				l3->alien = alien;
				alien = NULL;
			}
#endif
			spin_unlock_irq(&l3->list_lock);
			kfree(shared);
			free_alien_cache(alien);
L
Linus Torvalds 已提交
1094
		}
I
Ingo Molnar 已提交
1095
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1096 1097 1098 1099 1100 1101
		break;
	case CPU_ONLINE:
		start_cpu_timer(cpu);
		break;
#ifdef CONFIG_HOTPLUG_CPU
	case CPU_DEAD:
1102 1103 1104 1105 1106 1107 1108 1109
		/*
		 * Even if all the cpus of a node are down, we don't free the
		 * kmem_list3 of any cache. This to avoid a race between
		 * cpu_down, and a kmalloc allocation from another cpu for
		 * memory from the node of the cpu going down.  The list3
		 * structure is usually allocated from kmem_cache_create() and
		 * gets destroyed at kmem_cache_destroy().
		 */
L
Linus Torvalds 已提交
1110 1111
		/* fall thru */
	case CPU_UP_CANCELED:
I
Ingo Molnar 已提交
1112
		mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1113 1114
		list_for_each_entry(cachep, &cache_chain, next) {
			struct array_cache *nc;
1115 1116
			struct array_cache *shared;
			struct array_cache **alien;
1117
			cpumask_t mask;
L
Linus Torvalds 已提交
1118

1119
			mask = node_to_cpumask(node);
L
Linus Torvalds 已提交
1120 1121 1122
			/* cpu is dead; no one can alloc from it. */
			nc = cachep->array[cpu];
			cachep->array[cpu] = NULL;
1123 1124 1125
			l3 = cachep->nodelists[node];

			if (!l3)
1126
				goto free_array_cache;
1127

1128
			spin_lock_irq(&l3->list_lock);
1129 1130 1131 1132

			/* Free limit for this kmem_list3 */
			l3->free_limit -= cachep->batchcount;
			if (nc)
1133
				free_block(cachep, nc->entry, nc->avail, node);
1134 1135

			if (!cpus_empty(mask)) {
1136
				spin_unlock_irq(&l3->list_lock);
1137
				goto free_array_cache;
P
Pekka Enberg 已提交
1138
			}
1139

1140 1141
			shared = l3->shared;
			if (shared) {
1142
				free_block(cachep, l3->shared->entry,
P
Pekka Enberg 已提交
1143
					   l3->shared->avail, node);
1144 1145 1146
				l3->shared = NULL;
			}

1147 1148 1149 1150 1151 1152 1153 1154 1155
			alien = l3->alien;
			l3->alien = NULL;

			spin_unlock_irq(&l3->list_lock);

			kfree(shared);
			if (alien) {
				drain_alien_cache(cachep, alien);
				free_alien_cache(alien);
1156
			}
1157
free_array_cache:
L
Linus Torvalds 已提交
1158 1159
			kfree(nc);
		}
1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173
		/*
		 * In the previous loop, all the objects were freed to
		 * the respective cache's slabs,  now we can go ahead and
		 * shrink each nodelist to its limit.
		 */
		list_for_each_entry(cachep, &cache_chain, next) {
			l3 = cachep->nodelists[node];
			if (!l3)
				continue;
			spin_lock_irq(&l3->list_lock);
			/* free slabs belonging to this node */
			__node_shrink(cachep, node);
			spin_unlock_irq(&l3->list_lock);
		}
I
Ingo Molnar 已提交
1174
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1175 1176 1177 1178
		break;
#endif
	}
	return NOTIFY_OK;
A
Andrew Morton 已提交
1179
bad:
I
Ingo Molnar 已提交
1180
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1181 1182 1183 1184 1185
	return NOTIFY_BAD;
}

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

1186 1187 1188
/*
 * swap the static kmem_list3 with kmalloced memory
 */
A
Andrew Morton 已提交
1189 1190
static void init_list(struct kmem_cache *cachep, struct kmem_list3 *list,
			int nodeid)
1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204
{
	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();
}

A
Andrew Morton 已提交
1205 1206 1207
/*
 * Initialisation.  Called after the page allocator have been initialised and
 * before smp_init().
L
Linus Torvalds 已提交
1208 1209 1210 1211 1212 1213
 */
void __init kmem_cache_init(void)
{
	size_t left_over;
	struct cache_sizes *sizes;
	struct cache_names *names;
1214
	int i;
1215
	int order;
1216 1217 1218 1219 1220 1221

	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
1222 1223 1224 1225 1226 1227 1228 1229 1230 1231

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

	/* Bootstrap is tricky, because several objects are allocated
	 * from caches that do not exist yet:
A
Andrew Morton 已提交
1232 1233 1234
	 * 1) initialize the cache_cache cache: it contains the struct
	 *    kmem_cache structures of all caches, except cache_cache itself:
	 *    cache_cache is statically allocated.
1235 1236 1237
	 *    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
1238
	 * 2) Create the first kmalloc cache.
1239
	 *    The struct kmem_cache for the new cache is allocated normally.
1240 1241 1242
	 *    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
1243 1244
	 * 4) Replace the __init data head arrays for cache_cache and the first
	 *    kmalloc cache with kmalloc allocated arrays.
1245 1246 1247
	 * 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
1248 1249 1250 1251 1252 1253 1254
	 */

	/* 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;
1255
	cache_cache.nodelists[numa_node_id()] = &initkmem_list3[CACHE_CACHE];
L
Linus Torvalds 已提交
1256

A
Andrew Morton 已提交
1257 1258
	cache_cache.buffer_size = ALIGN(cache_cache.buffer_size,
					cache_line_size());
L
Linus Torvalds 已提交
1259

1260 1261 1262 1263 1264 1265
	for (order = 0; order < MAX_ORDER; order++) {
		cache_estimate(order, cache_cache.buffer_size,
			cache_line_size(), 0, &left_over, &cache_cache.num);
		if (cache_cache.num)
			break;
	}
L
Linus Torvalds 已提交
1266 1267
	if (!cache_cache.num)
		BUG();
1268
	cache_cache.gfporder = order;
P
Pekka Enberg 已提交
1269 1270 1271
	cache_cache.colour = left_over / cache_cache.colour_off;
	cache_cache.slab_size = ALIGN(cache_cache.num * sizeof(kmem_bufctl_t) +
				      sizeof(struct slab), cache_line_size());
L
Linus Torvalds 已提交
1272 1273 1274 1275 1276

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

A
Andrew Morton 已提交
1277 1278 1279 1280
	/*
	 * Initialize the caches that provide memory for the array cache and the
	 * kmem_list3 structures first.  Without this, further allocations will
	 * bug.
1281 1282 1283
	 */

	sizes[INDEX_AC].cs_cachep = kmem_cache_create(names[INDEX_AC].name,
A
Andrew Morton 已提交
1284 1285 1286 1287
					sizes[INDEX_AC].cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_PANIC,
					NULL, NULL);
1288

A
Andrew Morton 已提交
1289
	if (INDEX_AC != INDEX_L3) {
1290
		sizes[INDEX_L3].cs_cachep =
A
Andrew Morton 已提交
1291 1292 1293 1294 1295 1296
			kmem_cache_create(names[INDEX_L3].name,
				sizes[INDEX_L3].cs_size,
				ARCH_KMALLOC_MINALIGN,
				ARCH_KMALLOC_FLAGS|SLAB_PANIC,
				NULL, NULL);
	}
1297

L
Linus Torvalds 已提交
1298
	while (sizes->cs_size != ULONG_MAX) {
1299 1300
		/*
		 * For performance, all the general caches are L1 aligned.
L
Linus Torvalds 已提交
1301 1302 1303
		 * This should be particularly beneficial on SMP boxes, as it
		 * eliminates "false sharing".
		 * Note for systems short on memory removing the alignment will
1304 1305
		 * allow tighter packing of the smaller caches.
		 */
A
Andrew Morton 已提交
1306
		if (!sizes->cs_cachep) {
1307
			sizes->cs_cachep = kmem_cache_create(names->name,
A
Andrew Morton 已提交
1308 1309 1310 1311 1312
					sizes->cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_PANIC,
					NULL, NULL);
		}
L
Linus Torvalds 已提交
1313 1314 1315

		/* Inc off-slab bufctl limit until the ceiling is hit. */
		if (!(OFF_SLAB(sizes->cs_cachep))) {
P
Pekka Enberg 已提交
1316
			offslab_limit = sizes->cs_size - sizeof(struct slab);
L
Linus Torvalds 已提交
1317 1318 1319 1320
			offslab_limit /= sizeof(kmem_bufctl_t);
		}

		sizes->cs_dmacachep = kmem_cache_create(names->name_dma,
A
Andrew Morton 已提交
1321 1322 1323 1324 1325
					sizes->cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_CACHE_DMA|
						SLAB_PANIC,
					NULL, NULL);
L
Linus Torvalds 已提交
1326 1327 1328 1329 1330
		sizes++;
		names++;
	}
	/* 4) Replace the bootstrap head arrays */
	{
P
Pekka Enberg 已提交
1331
		void *ptr;
1332

L
Linus Torvalds 已提交
1333
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
1334

L
Linus Torvalds 已提交
1335
		local_irq_disable();
1336 1337
		BUG_ON(cpu_cache_get(&cache_cache) != &initarray_cache.cache);
		memcpy(ptr, cpu_cache_get(&cache_cache),
P
Pekka Enberg 已提交
1338
		       sizeof(struct arraycache_init));
L
Linus Torvalds 已提交
1339 1340
		cache_cache.array[smp_processor_id()] = ptr;
		local_irq_enable();
1341

L
Linus Torvalds 已提交
1342
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
1343

L
Linus Torvalds 已提交
1344
		local_irq_disable();
1345
		BUG_ON(cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep)
P
Pekka Enberg 已提交
1346
		       != &initarray_generic.cache);
1347
		memcpy(ptr, cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep),
P
Pekka Enberg 已提交
1348
		       sizeof(struct arraycache_init));
1349
		malloc_sizes[INDEX_AC].cs_cachep->array[smp_processor_id()] =
P
Pekka Enberg 已提交
1350
		    ptr;
L
Linus Torvalds 已提交
1351 1352
		local_irq_enable();
	}
1353 1354 1355 1356 1357
	/* 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
1358
			  numa_node_id());
1359 1360 1361

		for_each_online_node(node) {
			init_list(malloc_sizes[INDEX_AC].cs_cachep,
P
Pekka Enberg 已提交
1362
				  &initkmem_list3[SIZE_AC + node], node);
1363 1364 1365

			if (INDEX_AC != INDEX_L3) {
				init_list(malloc_sizes[INDEX_L3].cs_cachep,
P
Pekka Enberg 已提交
1366 1367
					  &initkmem_list3[SIZE_L3 + node],
					  node);
1368 1369 1370
			}
		}
	}
L
Linus Torvalds 已提交
1371

1372
	/* 6) resize the head arrays to their final sizes */
L
Linus Torvalds 已提交
1373
	{
1374
		struct kmem_cache *cachep;
I
Ingo Molnar 已提交
1375
		mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1376
		list_for_each_entry(cachep, &cache_chain, next)
A
Andrew Morton 已提交
1377
			enable_cpucache(cachep);
I
Ingo Molnar 已提交
1378
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1379 1380 1381 1382 1383
	}

	/* Done! */
	g_cpucache_up = FULL;

A
Andrew Morton 已提交
1384 1385 1386
	/*
	 * Register a cpu startup notifier callback that initializes
	 * cpu_cache_get for all new cpus
L
Linus Torvalds 已提交
1387 1388 1389
	 */
	register_cpu_notifier(&cpucache_notifier);

A
Andrew Morton 已提交
1390 1391 1392
	/*
	 * The reap timers are started later, with a module init call: That part
	 * of the kernel is not yet operational.
L
Linus Torvalds 已提交
1393 1394 1395 1396 1397 1398 1399
	 */
}

static int __init cpucache_init(void)
{
	int cpu;

A
Andrew Morton 已提交
1400 1401
	/*
	 * Register the timers that return unneeded pages to the page allocator
L
Linus Torvalds 已提交
1402
	 */
1403
	for_each_online_cpu(cpu)
A
Andrew Morton 已提交
1404
		start_cpu_timer(cpu);
L
Linus Torvalds 已提交
1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415
	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.
 */
1416
static void *kmem_getpages(struct kmem_cache *cachep, gfp_t flags, int nodeid)
L
Linus Torvalds 已提交
1417 1418 1419 1420 1421 1422
{
	struct page *page;
	void *addr;
	int i;

	flags |= cachep->gfpflags;
1423
	page = alloc_pages_node(nodeid, flags, cachep->gfporder);
L
Linus Torvalds 已提交
1424 1425 1426 1427 1428 1429 1430 1431 1432
	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--) {
N
Nick Piggin 已提交
1433
		__SetPageSlab(page);
L
Linus Torvalds 已提交
1434 1435 1436 1437 1438 1439 1440 1441
		page++;
	}
	return addr;
}

/*
 * Interface to system's page release.
 */
1442
static void kmem_freepages(struct kmem_cache *cachep, void *addr)
L
Linus Torvalds 已提交
1443
{
P
Pekka Enberg 已提交
1444
	unsigned long i = (1 << cachep->gfporder);
L
Linus Torvalds 已提交
1445 1446 1447 1448
	struct page *page = virt_to_page(addr);
	const unsigned long nr_freed = i;

	while (i--) {
N
Nick Piggin 已提交
1449 1450
		BUG_ON(!PageSlab(page));
		__ClearPageSlab(page);
L
Linus Torvalds 已提交
1451 1452 1453 1454 1455 1456
		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
1457 1458
	if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
		atomic_sub(1 << cachep->gfporder, &slab_reclaim_pages);
L
Linus Torvalds 已提交
1459 1460 1461 1462
}

static void kmem_rcu_free(struct rcu_head *head)
{
P
Pekka Enberg 已提交
1463
	struct slab_rcu *slab_rcu = (struct slab_rcu *)head;
1464
	struct kmem_cache *cachep = slab_rcu->cachep;
L
Linus Torvalds 已提交
1465 1466 1467 1468 1469 1470 1471 1472 1473

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

#if DEBUG

#ifdef CONFIG_DEBUG_PAGEALLOC
1474
static void store_stackinfo(struct kmem_cache *cachep, unsigned long *addr,
P
Pekka Enberg 已提交
1475
			    unsigned long caller)
L
Linus Torvalds 已提交
1476
{
1477
	int size = obj_size(cachep);
L
Linus Torvalds 已提交
1478

1479
	addr = (unsigned long *)&((char *)addr)[obj_offset(cachep)];
L
Linus Torvalds 已提交
1480

P
Pekka Enberg 已提交
1481
	if (size < 5 * sizeof(unsigned long))
L
Linus Torvalds 已提交
1482 1483
		return;

P
Pekka Enberg 已提交
1484 1485 1486 1487
	*addr++ = 0x12345678;
	*addr++ = caller;
	*addr++ = smp_processor_id();
	size -= 3 * sizeof(unsigned long);
L
Linus Torvalds 已提交
1488 1489 1490 1491 1492 1493 1494
	{
		unsigned long *sptr = &caller;
		unsigned long svalue;

		while (!kstack_end(sptr)) {
			svalue = *sptr++;
			if (kernel_text_address(svalue)) {
P
Pekka Enberg 已提交
1495
				*addr++ = svalue;
L
Linus Torvalds 已提交
1496 1497 1498 1499 1500 1501 1502
				size -= sizeof(unsigned long);
				if (size <= sizeof(unsigned long))
					break;
			}
		}

	}
P
Pekka Enberg 已提交
1503
	*addr++ = 0x87654321;
L
Linus Torvalds 已提交
1504 1505 1506
}
#endif

1507
static void poison_obj(struct kmem_cache *cachep, void *addr, unsigned char val)
L
Linus Torvalds 已提交
1508
{
1509 1510
	int size = obj_size(cachep);
	addr = &((char *)addr)[obj_offset(cachep)];
L
Linus Torvalds 已提交
1511 1512

	memset(addr, val, size);
P
Pekka Enberg 已提交
1513
	*(unsigned char *)(addr + size - 1) = POISON_END;
L
Linus Torvalds 已提交
1514 1515 1516 1517 1518 1519
}

static void dump_line(char *data, int offset, int limit)
{
	int i;
	printk(KERN_ERR "%03x:", offset);
A
Andrew Morton 已提交
1520
	for (i = 0; i < limit; i++)
P
Pekka Enberg 已提交
1521
		printk(" %02x", (unsigned char)data[offset + i]);
L
Linus Torvalds 已提交
1522 1523 1524 1525 1526 1527
	printk("\n");
}
#endif

#if DEBUG

1528
static void print_objinfo(struct kmem_cache *cachep, void *objp, int lines)
L
Linus Torvalds 已提交
1529 1530 1531 1532 1533 1534
{
	int i, size;
	char *realobj;

	if (cachep->flags & SLAB_RED_ZONE) {
		printk(KERN_ERR "Redzone: 0x%lx/0x%lx.\n",
A
Andrew Morton 已提交
1535 1536
			*dbg_redzone1(cachep, objp),
			*dbg_redzone2(cachep, objp));
L
Linus Torvalds 已提交
1537 1538 1539 1540
	}

	if (cachep->flags & SLAB_STORE_USER) {
		printk(KERN_ERR "Last user: [<%p>]",
A
Andrew Morton 已提交
1541
			*dbg_userword(cachep, objp));
L
Linus Torvalds 已提交
1542
		print_symbol("(%s)",
A
Andrew Morton 已提交
1543
				(unsigned long)*dbg_userword(cachep, objp));
L
Linus Torvalds 已提交
1544 1545
		printk("\n");
	}
1546 1547
	realobj = (char *)objp + obj_offset(cachep);
	size = obj_size(cachep);
P
Pekka Enberg 已提交
1548
	for (i = 0; i < size && lines; i += 16, lines--) {
L
Linus Torvalds 已提交
1549 1550
		int limit;
		limit = 16;
P
Pekka Enberg 已提交
1551 1552
		if (i + limit > size)
			limit = size - i;
L
Linus Torvalds 已提交
1553 1554 1555 1556
		dump_line(realobj, i, limit);
	}
}

1557
static void check_poison_obj(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
1558 1559 1560 1561 1562
{
	char *realobj;
	int size, i;
	int lines = 0;

1563 1564
	realobj = (char *)objp + obj_offset(cachep);
	size = obj_size(cachep);
L
Linus Torvalds 已提交
1565

P
Pekka Enberg 已提交
1566
	for (i = 0; i < size; i++) {
L
Linus Torvalds 已提交
1567
		char exp = POISON_FREE;
P
Pekka Enberg 已提交
1568
		if (i == size - 1)
L
Linus Torvalds 已提交
1569 1570 1571 1572 1573 1574
			exp = POISON_END;
		if (realobj[i] != exp) {
			int limit;
			/* Mismatch ! */
			/* Print header */
			if (lines == 0) {
P
Pekka Enberg 已提交
1575
				printk(KERN_ERR
A
Andrew Morton 已提交
1576 1577
					"Slab corruption: start=%p, len=%d\n",
					realobj, size);
L
Linus Torvalds 已提交
1578 1579 1580
				print_objinfo(cachep, objp, 0);
			}
			/* Hexdump the affected line */
P
Pekka Enberg 已提交
1581
			i = (i / 16) * 16;
L
Linus Torvalds 已提交
1582
			limit = 16;
P
Pekka Enberg 已提交
1583 1584
			if (i + limit > size)
				limit = size - i;
L
Linus Torvalds 已提交
1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596
			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:
		 */
1597
		struct slab *slabp = virt_to_slab(objp);
1598
		unsigned int objnr;
L
Linus Torvalds 已提交
1599

1600
		objnr = obj_to_index(cachep, slabp, objp);
L
Linus Torvalds 已提交
1601
		if (objnr) {
1602
			objp = index_to_obj(cachep, slabp, objnr - 1);
1603
			realobj = (char *)objp + obj_offset(cachep);
L
Linus Torvalds 已提交
1604
			printk(KERN_ERR "Prev obj: start=%p, len=%d\n",
P
Pekka Enberg 已提交
1605
			       realobj, size);
L
Linus Torvalds 已提交
1606 1607
			print_objinfo(cachep, objp, 2);
		}
P
Pekka Enberg 已提交
1608
		if (objnr + 1 < cachep->num) {
1609
			objp = index_to_obj(cachep, slabp, objnr + 1);
1610
			realobj = (char *)objp + obj_offset(cachep);
L
Linus Torvalds 已提交
1611
			printk(KERN_ERR "Next obj: start=%p, len=%d\n",
P
Pekka Enberg 已提交
1612
			       realobj, size);
L
Linus Torvalds 已提交
1613 1614 1615 1616 1617 1618
			print_objinfo(cachep, objp, 2);
		}
	}
}
#endif

1619 1620
#if DEBUG
/**
1621 1622 1623 1624 1625 1626
 * slab_destroy_objs - destroy a slab and its objects
 * @cachep: cache pointer being destroyed
 * @slabp: slab pointer being destroyed
 *
 * Call the registered destructor for each object in a slab that is being
 * destroyed.
L
Linus Torvalds 已提交
1627
 */
1628
static void slab_destroy_objs(struct kmem_cache *cachep, struct slab *slabp)
L
Linus Torvalds 已提交
1629 1630 1631
{
	int i;
	for (i = 0; i < cachep->num; i++) {
1632
		void *objp = index_to_obj(cachep, slabp, i);
L
Linus Torvalds 已提交
1633 1634 1635

		if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
A
Andrew Morton 已提交
1636 1637
			if (cachep->buffer_size % PAGE_SIZE == 0 &&
					OFF_SLAB(cachep))
P
Pekka Enberg 已提交
1638
				kernel_map_pages(virt_to_page(objp),
A
Andrew Morton 已提交
1639
					cachep->buffer_size / PAGE_SIZE, 1);
L
Linus Torvalds 已提交
1640 1641 1642 1643 1644 1645 1646 1647 1648
			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
1649
					   "was overwritten");
L
Linus Torvalds 已提交
1650 1651
			if (*dbg_redzone2(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "end of a freed object "
P
Pekka Enberg 已提交
1652
					   "was overwritten");
L
Linus Torvalds 已提交
1653 1654
		}
		if (cachep->dtor && !(cachep->flags & SLAB_POISON))
1655
			(cachep->dtor) (objp + obj_offset(cachep), cachep, 0);
L
Linus Torvalds 已提交
1656
	}
1657
}
L
Linus Torvalds 已提交
1658
#else
1659
static void slab_destroy_objs(struct kmem_cache *cachep, struct slab *slabp)
1660
{
L
Linus Torvalds 已提交
1661 1662 1663
	if (cachep->dtor) {
		int i;
		for (i = 0; i < cachep->num; i++) {
1664
			void *objp = index_to_obj(cachep, slabp, i);
P
Pekka Enberg 已提交
1665
			(cachep->dtor) (objp, cachep, 0);
L
Linus Torvalds 已提交
1666 1667
		}
	}
1668
}
L
Linus Torvalds 已提交
1669 1670
#endif

1671 1672 1673 1674 1675
/**
 * slab_destroy - destroy and release all objects in a slab
 * @cachep: cache pointer being destroyed
 * @slabp: slab pointer being destroyed
 *
1676
 * Destroy all the objs in a slab, and release the mem back to the system.
A
Andrew Morton 已提交
1677 1678
 * Before calling the slab must have been unlinked from the cache.  The
 * cache-lock is not held/needed.
1679
 */
1680
static void slab_destroy(struct kmem_cache *cachep, struct slab *slabp)
1681 1682 1683 1684
{
	void *addr = slabp->s_mem - slabp->colouroff;

	slab_destroy_objs(cachep, slabp);
L
Linus Torvalds 已提交
1685 1686 1687
	if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU)) {
		struct slab_rcu *slab_rcu;

P
Pekka Enberg 已提交
1688
		slab_rcu = (struct slab_rcu *)slabp;
L
Linus Torvalds 已提交
1689 1690 1691 1692 1693 1694 1695 1696 1697 1698
		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);
	}
}

A
Andrew Morton 已提交
1699 1700 1701 1702
/*
 * For setting up all the kmem_list3s for cache whose buffer_size is same as
 * size of kmem_list3.
 */
1703
static void set_up_list3s(struct kmem_cache *cachep, int index)
1704 1705 1706 1707
{
	int node;

	for_each_online_node(node) {
P
Pekka Enberg 已提交
1708
		cachep->nodelists[node] = &initkmem_list3[index + node];
1709
		cachep->nodelists[node]->next_reap = jiffies +
P
Pekka Enberg 已提交
1710 1711
		    REAPTIMEOUT_LIST3 +
		    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
1712 1713 1714
	}
}

1715
/**
1716 1717 1718 1719 1720 1721 1722
 * calculate_slab_order - calculate size (page order) of slabs
 * @cachep: pointer to the cache that is being created
 * @size: size of objects to be created in this cache.
 * @align: required alignment for the objects.
 * @flags: slab allocation flags
 *
 * Also calculates the number of objects per slab.
1723 1724 1725 1726 1727
 *
 * This could be made much more intelligent.  For now, try to avoid using
 * high order pages for slabs.  When the gfp() functions are more friendly
 * towards high-order requests, this should be changed.
 */
A
Andrew Morton 已提交
1728
static size_t calculate_slab_order(struct kmem_cache *cachep,
R
Randy Dunlap 已提交
1729
			size_t size, size_t align, unsigned long flags)
1730 1731
{
	size_t left_over = 0;
1732
	int gfporder;
1733

A
Andrew Morton 已提交
1734
	for (gfporder = 0; gfporder <= MAX_GFP_ORDER; gfporder++) {
1735 1736 1737
		unsigned int num;
		size_t remainder;

1738
		cache_estimate(gfporder, size, align, flags, &remainder, &num);
1739 1740
		if (!num)
			continue;
1741

1742
		/* More than offslab_limit objects will cause problems */
1743
		if ((flags & CFLGS_OFF_SLAB) && num > offslab_limit)
1744 1745
			break;

1746
		/* Found something acceptable - save it away */
1747
		cachep->num = num;
1748
		cachep->gfporder = gfporder;
1749 1750
		left_over = remainder;

1751 1752 1753 1754 1755 1756 1757 1758
		/*
		 * A VFS-reclaimable slab tends to have most allocations
		 * as GFP_NOFS and we really don't want to have to be allocating
		 * higher-order pages when we are unable to shrink dcache.
		 */
		if (flags & SLAB_RECLAIM_ACCOUNT)
			break;

1759 1760 1761 1762
		/*
		 * Large number of objects is good, but very large slabs are
		 * currently bad for the gfp()s.
		 */
1763
		if (gfporder >= slab_break_gfp_order)
1764 1765
			break;

1766 1767 1768
		/*
		 * Acceptable internal fragmentation?
		 */
A
Andrew Morton 已提交
1769
		if (left_over * 8 <= (PAGE_SIZE << gfporder))
1770 1771 1772 1773 1774
			break;
	}
	return left_over;
}

1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828
static void setup_cpu_cache(struct kmem_cache *cachep)
{
	if (g_cpucache_up == FULL) {
		enable_cpucache(cachep);
		return;
	}
	if (g_cpucache_up == NONE) {
		/*
		 * Note: the first kmem_cache_create must create the cache
		 * that's used by kmalloc(24), otherwise the creation of
		 * further caches will BUG().
		 */
		cachep->array[smp_processor_id()] = &initarray_generic.cache;

		/*
		 * If the cache that's used by kmalloc(sizeof(kmem_list3)) is
		 * the first cache, then we need to set up all its list3s,
		 * otherwise the creation of further caches will BUG().
		 */
		set_up_list3s(cachep, SIZE_AC);
		if (INDEX_AC == INDEX_L3)
			g_cpucache_up = PARTIAL_L3;
		else
			g_cpucache_up = PARTIAL_AC;
	} else {
		cachep->array[smp_processor_id()] =
			kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);

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

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

L
Linus Torvalds 已提交
1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843
/**
 * kmem_cache_create - Create a cache.
 * @name: A string which is used in /proc/slabinfo to identify this cache.
 * @size: The size of objects to be created in this cache.
 * @align: The required alignment for the objects.
 * @flags: SLAB flags
 * @ctor: A constructor for the objects.
 * @dtor: A destructor for the objects.
 *
 * Returns a ptr to the cache on success, NULL on failure.
 * Cannot be called within a int, but can be interrupted.
 * The @ctor is run when new pages are allocated by the cache
 * and the @dtor is run before the pages are handed back.
 *
 * @name must be valid until the cache is destroyed. This implies that
A
Andrew Morton 已提交
1844 1845
 * the module calling this has to destroy the cache before getting unloaded.
 *
L
Linus Torvalds 已提交
1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857
 * The flags are
 *
 * %SLAB_POISON - Poison the slab with a known test pattern (a5a5a5a5)
 * to catch references to uninitialised memory.
 *
 * %SLAB_RED_ZONE - Insert `Red' zones around the allocated memory to check
 * for buffer overruns.
 *
 * %SLAB_HWCACHE_ALIGN - Align the objects in this cache to a hardware
 * cacheline.  This can be beneficial if you're counting cycles as closely
 * as davem.
 */
1858
struct kmem_cache *
L
Linus Torvalds 已提交
1859
kmem_cache_create (const char *name, size_t size, size_t align,
A
Andrew Morton 已提交
1860 1861
	unsigned long flags,
	void (*ctor)(void*, struct kmem_cache *, unsigned long),
1862
	void (*dtor)(void*, struct kmem_cache *, unsigned long))
L
Linus Torvalds 已提交
1863 1864
{
	size_t left_over, slab_size, ralign;
1865
	struct kmem_cache *cachep = NULL;
1866
	struct list_head *p;
L
Linus Torvalds 已提交
1867 1868 1869 1870

	/*
	 * Sanity checks... these are all serious usage bugs.
	 */
A
Andrew Morton 已提交
1871
	if (!name || in_interrupt() || (size < BYTES_PER_WORD) ||
P
Pekka Enberg 已提交
1872
	    (size > (1 << MAX_OBJ_ORDER) * PAGE_SIZE) || (dtor && !ctor)) {
A
Andrew Morton 已提交
1873 1874
		printk(KERN_ERR "%s: Early error in slab %s\n", __FUNCTION__,
				name);
P
Pekka Enberg 已提交
1875 1876
		BUG();
	}
L
Linus Torvalds 已提交
1877

1878 1879 1880 1881 1882 1883
	/*
	 * Prevent CPUs from coming and going.
	 * lock_cpu_hotplug() nests outside cache_chain_mutex
	 */
	lock_cpu_hotplug();

I
Ingo Molnar 已提交
1884
	mutex_lock(&cache_chain_mutex);
1885 1886

	list_for_each(p, &cache_chain) {
1887
		struct kmem_cache *pc = list_entry(p, struct kmem_cache, next);
1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901
		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",
1902
			       pc->buffer_size);
1903 1904 1905
			continue;
		}

P
Pekka Enberg 已提交
1906
		if (!strcmp(pc->name, name)) {
1907 1908 1909 1910 1911 1912
			printk("kmem_cache_create: duplicate cache %s\n", name);
			dump_stack();
			goto oops;
		}
	}

L
Linus Torvalds 已提交
1913 1914 1915 1916 1917
#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
1918
		       "requested - %s\n", __FUNCTION__, name);
L
Linus Torvalds 已提交
1919 1920 1921 1922 1923 1924 1925 1926 1927
		flags &= ~SLAB_DEBUG_INITIAL;
	}
#if FORCED_DEBUG
	/*
	 * Enable redzoning and last user accounting, except for caches with
	 * large objects, if the increased size would increase the object size
	 * above the next power of two: caches with object sizes just above a
	 * power of two have a significant amount of internal fragmentation.
	 */
A
Andrew Morton 已提交
1928
	if (size < 4096 || fls(size - 1) == fls(size-1 + 3 * BYTES_PER_WORD))
P
Pekka Enberg 已提交
1929
		flags |= SLAB_RED_ZONE | SLAB_STORE_USER;
L
Linus Torvalds 已提交
1930 1931 1932 1933 1934 1935 1936 1937 1938 1939
	if (!(flags & SLAB_DESTROY_BY_RCU))
		flags |= SLAB_POISON;
#endif
	if (flags & SLAB_DESTROY_BY_RCU)
		BUG_ON(flags & SLAB_POISON);
#endif
	if (flags & SLAB_DESTROY_BY_RCU)
		BUG_ON(dtor);

	/*
A
Andrew Morton 已提交
1940 1941
	 * Always checks flags, a caller might be expecting debug support which
	 * isn't available.
L
Linus Torvalds 已提交
1942 1943 1944 1945
	 */
	if (flags & ~CREATE_MASK)
		BUG();

A
Andrew Morton 已提交
1946 1947
	/*
	 * Check that size is in terms of words.  This is needed to avoid
L
Linus Torvalds 已提交
1948 1949 1950
	 * 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
1951 1952 1953
	if (size & (BYTES_PER_WORD - 1)) {
		size += (BYTES_PER_WORD - 1);
		size &= ~(BYTES_PER_WORD - 1);
L
Linus Torvalds 已提交
1954 1955
	}

A
Andrew Morton 已提交
1956 1957
	/* calculate the final buffer alignment: */

L
Linus Torvalds 已提交
1958 1959
	/* 1) arch recommendation: can be overridden for debug */
	if (flags & SLAB_HWCACHE_ALIGN) {
A
Andrew Morton 已提交
1960 1961 1962 1963
		/*
		 * Default alignment: as specified by the arch code.  Except if
		 * an object is really small, then squeeze multiple objects into
		 * one cacheline.
L
Linus Torvalds 已提交
1964 1965
		 */
		ralign = cache_line_size();
P
Pekka Enberg 已提交
1966
		while (size <= ralign / 2)
L
Linus Torvalds 已提交
1967 1968 1969 1970 1971 1972 1973 1974
			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
1975
			flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
L
Linus Torvalds 已提交
1976 1977 1978 1979 1980
	}
	/* 3) caller mandated alignment: disables debug if necessary */
	if (ralign < align) {
		ralign = align;
		if (ralign > BYTES_PER_WORD)
P
Pekka Enberg 已提交
1981
			flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
L
Linus Torvalds 已提交
1982
	}
A
Andrew Morton 已提交
1983 1984
	/*
	 * 4) Store it. Note that the debug code below can reduce
L
Linus Torvalds 已提交
1985 1986 1987 1988 1989
	 *    the alignment to BYTES_PER_WORD.
	 */
	align = ralign;

	/* Get cache's description obj. */
1990
	cachep = kmem_cache_alloc(&cache_cache, SLAB_KERNEL);
L
Linus Torvalds 已提交
1991
	if (!cachep)
1992
		goto oops;
1993
	memset(cachep, 0, sizeof(struct kmem_cache));
L
Linus Torvalds 已提交
1994 1995

#if DEBUG
1996
	cachep->obj_size = size;
L
Linus Torvalds 已提交
1997 1998 1999 2000 2001 2002

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

		/* add space for red zone words */
2003
		cachep->obj_offset += BYTES_PER_WORD;
P
Pekka Enberg 已提交
2004
		size += 2 * BYTES_PER_WORD;
L
Linus Torvalds 已提交
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
	}
	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
2015
	if (size >= malloc_sizes[INDEX_L3 + 1].cs_size
2016 2017
	    && cachep->obj_size > cache_line_size() && size < PAGE_SIZE) {
		cachep->obj_offset += PAGE_SIZE - size;
L
Linus Torvalds 已提交
2018 2019 2020 2021 2022 2023
		size = PAGE_SIZE;
	}
#endif
#endif

	/* Determine if the slab management is 'on' or 'off' slab. */
P
Pekka Enberg 已提交
2024
	if (size >= (PAGE_SIZE >> 3))
L
Linus Torvalds 已提交
2025 2026 2027 2028 2029 2030 2031 2032
		/*
		 * 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);

2033
	left_over = calculate_slab_order(cachep, size, align, flags);
L
Linus Torvalds 已提交
2034 2035 2036 2037 2038

	if (!cachep->num) {
		printk("kmem_cache_create: couldn't create cache %s.\n", name);
		kmem_cache_free(&cache_cache, cachep);
		cachep = NULL;
2039
		goto oops;
L
Linus Torvalds 已提交
2040
	}
P
Pekka Enberg 已提交
2041 2042
	slab_size = ALIGN(cachep->num * sizeof(kmem_bufctl_t)
			  + sizeof(struct slab), align);
L
Linus Torvalds 已提交
2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054

	/*
	 * 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
2055 2056
		slab_size =
		    cachep->num * sizeof(kmem_bufctl_t) + sizeof(struct slab);
L
Linus Torvalds 已提交
2057 2058 2059 2060 2061 2062
	}

	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
2063
	cachep->colour = left_over / cachep->colour_off;
L
Linus Torvalds 已提交
2064 2065 2066 2067 2068
	cachep->slab_size = slab_size;
	cachep->flags = flags;
	cachep->gfpflags = 0;
	if (flags & SLAB_CACHE_DMA)
		cachep->gfpflags |= GFP_DMA;
2069
	cachep->buffer_size = size;
L
Linus Torvalds 已提交
2070 2071

	if (flags & CFLGS_OFF_SLAB)
2072
		cachep->slabp_cache = kmem_find_general_cachep(slab_size, 0u);
L
Linus Torvalds 已提交
2073 2074 2075 2076 2077
	cachep->ctor = ctor;
	cachep->dtor = dtor;
	cachep->name = name;


2078
	setup_cpu_cache(cachep);
L
Linus Torvalds 已提交
2079 2080 2081

	/* cache setup completed, link it into the list */
	list_add(&cachep->next, &cache_chain);
A
Andrew Morton 已提交
2082
oops:
L
Linus Torvalds 已提交
2083 2084
	if (!cachep && (flags & SLAB_PANIC))
		panic("kmem_cache_create(): failed to create slab `%s'\n",
P
Pekka Enberg 已提交
2085
		      name);
I
Ingo Molnar 已提交
2086
	mutex_unlock(&cache_chain_mutex);
2087
	unlock_cpu_hotplug();
L
Linus Torvalds 已提交
2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102
	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());
}

2103
static void check_spinlock_acquired(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2104 2105 2106
{
#ifdef CONFIG_SMP
	check_irq_off();
2107
	assert_spin_locked(&cachep->nodelists[numa_node_id()]->list_lock);
L
Linus Torvalds 已提交
2108 2109
#endif
}
2110

2111
static void check_spinlock_acquired_node(struct kmem_cache *cachep, int node)
2112 2113 2114 2115 2116 2117 2118
{
#ifdef CONFIG_SMP
	check_irq_off();
	assert_spin_locked(&cachep->nodelists[node]->list_lock);
#endif
}

L
Linus Torvalds 已提交
2119 2120 2121 2122
#else
#define check_irq_off()	do { } while(0)
#define check_irq_on()	do { } while(0)
#define check_spinlock_acquired(x) do { } while(0)
2123
#define check_spinlock_acquired_node(x, y) do { } while(0)
L
Linus Torvalds 已提交
2124 2125
#endif

A
Andrew Morton 已提交
2126 2127
static void drain_array_locked(struct kmem_cache *cachep,
			struct array_cache *ac, int force, int node);
L
Linus Torvalds 已提交
2128

2129 2130 2131 2132
static void drain_array(struct kmem_cache *cachep, struct kmem_list3 *l3,
			struct array_cache *ac,
			int force, int node);

L
Linus Torvalds 已提交
2133 2134
static void do_drain(void *arg)
{
A
Andrew Morton 已提交
2135
	struct kmem_cache *cachep = arg;
L
Linus Torvalds 已提交
2136
	struct array_cache *ac;
2137
	int node = numa_node_id();
L
Linus Torvalds 已提交
2138 2139

	check_irq_off();
2140
	ac = cpu_cache_get(cachep);
2141 2142 2143
	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
2144 2145 2146
	ac->avail = 0;
}

2147
static void drain_cpu_caches(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2148
{
2149 2150 2151
	struct kmem_list3 *l3;
	int node;

A
Andrew Morton 已提交
2152
	on_each_cpu(do_drain, cachep, 1, 1);
L
Linus Torvalds 已提交
2153
	check_irq_on();
P
Pekka Enberg 已提交
2154
	for_each_online_node(node) {
2155 2156
		l3 = cachep->nodelists[node];
		if (l3) {
2157
			drain_array(cachep, l3, l3->shared, 1, node);
2158
			if (l3->alien)
2159
				drain_alien_cache(cachep, l3->alien);
2160 2161
		}
	}
L
Linus Torvalds 已提交
2162 2163
}

2164
static int __node_shrink(struct kmem_cache *cachep, int node)
L
Linus Torvalds 已提交
2165 2166
{
	struct slab *slabp;
2167
	struct kmem_list3 *l3 = cachep->nodelists[node];
L
Linus Torvalds 已提交
2168 2169
	int ret;

2170
	for (;;) {
L
Linus Torvalds 已提交
2171 2172
		struct list_head *p;

2173 2174
		p = l3->slabs_free.prev;
		if (p == &l3->slabs_free)
L
Linus Torvalds 已提交
2175 2176
			break;

2177
		slabp = list_entry(l3->slabs_free.prev, struct slab, list);
L
Linus Torvalds 已提交
2178 2179 2180 2181 2182 2183
#if DEBUG
		if (slabp->inuse)
			BUG();
#endif
		list_del(&slabp->list);

2184 2185
		l3->free_objects -= cachep->num;
		spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
2186
		slab_destroy(cachep, slabp);
2187
		spin_lock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
2188
	}
P
Pekka Enberg 已提交
2189
	ret = !list_empty(&l3->slabs_full) || !list_empty(&l3->slabs_partial);
L
Linus Torvalds 已提交
2190 2191 2192
	return ret;
}

2193
static int __cache_shrink(struct kmem_cache *cachep)
2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211
{
	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
2212 2213 2214 2215 2216 2217 2218
/**
 * 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.
 */
2219
int kmem_cache_shrink(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231
{
	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
 *
2232
 * Remove a struct kmem_cache object from the slab cache.
L
Linus Torvalds 已提交
2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244
 * 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().
 */
2245
int kmem_cache_destroy(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2246 2247
{
	int i;
2248
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
2249 2250 2251 2252 2253 2254 2255 2256

	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
2257
	mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2258 2259 2260 2261
	/*
	 * the chain is never empty, cache_cache is never destroyed
	 */
	list_del(&cachep->next);
I
Ingo Molnar 已提交
2262
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2263 2264 2265

	if (__cache_shrink(cachep)) {
		slab_error(cachep, "Can't free all objects");
I
Ingo Molnar 已提交
2266
		mutex_lock(&cache_chain_mutex);
P
Pekka Enberg 已提交
2267
		list_add(&cachep->next, &cache_chain);
I
Ingo Molnar 已提交
2268
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2269 2270 2271 2272 2273
		unlock_cpu_hotplug();
		return 1;
	}

	if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU))
2274
		synchronize_rcu();
L
Linus Torvalds 已提交
2275

2276
	for_each_online_cpu(i)
P
Pekka Enberg 已提交
2277
	    kfree(cachep->array[i]);
L
Linus Torvalds 已提交
2278 2279

	/* NUMA: free the list3 structures */
2280
	for_each_online_node(i) {
A
Andrew Morton 已提交
2281 2282
		l3 = cachep->nodelists[i];
		if (l3) {
2283 2284 2285 2286 2287
			kfree(l3->shared);
			free_alien_cache(l3->alien);
			kfree(l3);
		}
	}
L
Linus Torvalds 已提交
2288 2289 2290 2291 2292 2293 2294
	kmem_cache_free(&cache_cache, cachep);
	unlock_cpu_hotplug();
	return 0;
}
EXPORT_SYMBOL(kmem_cache_destroy);

/* Get the memory for a slab management obj. */
2295
static struct slab *alloc_slabmgmt(struct kmem_cache *cachep, void *objp,
P
Pekka Enberg 已提交
2296
				   int colour_off, gfp_t local_flags)
L
Linus Torvalds 已提交
2297 2298
{
	struct slab *slabp;
P
Pekka Enberg 已提交
2299

L
Linus Torvalds 已提交
2300 2301 2302 2303 2304 2305
	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
2306
		slabp = objp + colour_off;
L
Linus Torvalds 已提交
2307 2308 2309 2310
		colour_off += cachep->slab_size;
	}
	slabp->inuse = 0;
	slabp->colouroff = colour_off;
P
Pekka Enberg 已提交
2311
	slabp->s_mem = objp + colour_off;
L
Linus Torvalds 已提交
2312 2313 2314 2315 2316
	return slabp;
}

static inline kmem_bufctl_t *slab_bufctl(struct slab *slabp)
{
P
Pekka Enberg 已提交
2317
	return (kmem_bufctl_t *) (slabp + 1);
L
Linus Torvalds 已提交
2318 2319
}

2320
static void cache_init_objs(struct kmem_cache *cachep,
P
Pekka Enberg 已提交
2321
			    struct slab *slabp, unsigned long ctor_flags)
L
Linus Torvalds 已提交
2322 2323 2324 2325
{
	int i;

	for (i = 0; i < cachep->num; i++) {
2326
		void *objp = index_to_obj(cachep, slabp, i);
L
Linus Torvalds 已提交
2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338
#if DEBUG
		/* need to poison the objs? */
		if (cachep->flags & SLAB_POISON)
			poison_obj(cachep, objp, POISON_FREE);
		if (cachep->flags & SLAB_STORE_USER)
			*dbg_userword(cachep, objp) = NULL;

		if (cachep->flags & SLAB_RED_ZONE) {
			*dbg_redzone1(cachep, objp) = RED_INACTIVE;
			*dbg_redzone2(cachep, objp) = RED_INACTIVE;
		}
		/*
A
Andrew Morton 已提交
2339 2340 2341
		 * Constructors are not allowed to allocate memory from the same
		 * cache which they are a constructor for.  Otherwise, deadlock.
		 * They must also be threaded.
L
Linus Torvalds 已提交
2342 2343
		 */
		if (cachep->ctor && !(cachep->flags & SLAB_POISON))
2344
			cachep->ctor(objp + obj_offset(cachep), cachep,
P
Pekka Enberg 已提交
2345
				     ctor_flags);
L
Linus Torvalds 已提交
2346 2347 2348 2349

		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
2350
					   " end of an object");
L
Linus Torvalds 已提交
2351 2352
			if (*dbg_redzone1(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "constructor overwrote the"
P
Pekka Enberg 已提交
2353
					   " start of an object");
L
Linus Torvalds 已提交
2354
		}
A
Andrew Morton 已提交
2355 2356
		if ((cachep->buffer_size % PAGE_SIZE) == 0 &&
			    OFF_SLAB(cachep) && cachep->flags & SLAB_POISON)
P
Pekka Enberg 已提交
2357
			kernel_map_pages(virt_to_page(objp),
2358
					 cachep->buffer_size / PAGE_SIZE, 0);
L
Linus Torvalds 已提交
2359 2360 2361 2362
#else
		if (cachep->ctor)
			cachep->ctor(objp, cachep, ctor_flags);
#endif
P
Pekka Enberg 已提交
2363
		slab_bufctl(slabp)[i] = i + 1;
L
Linus Torvalds 已提交
2364
	}
P
Pekka Enberg 已提交
2365
	slab_bufctl(slabp)[i - 1] = BUFCTL_END;
L
Linus Torvalds 已提交
2366 2367 2368
	slabp->free = 0;
}

2369
static void kmem_flagcheck(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2370
{
A
Andrew Morton 已提交
2371 2372 2373 2374
	if (flags & SLAB_DMA)
		BUG_ON(!(cachep->gfpflags & GFP_DMA));
	else
		BUG_ON(cachep->gfpflags & GFP_DMA);
L
Linus Torvalds 已提交
2375 2376
}

A
Andrew Morton 已提交
2377 2378
static void *slab_get_obj(struct kmem_cache *cachep, struct slab *slabp,
				int nodeid)
2379
{
2380
	void *objp = index_to_obj(cachep, slabp, slabp->free);
2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393
	kmem_bufctl_t next;

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

	return objp;
}

A
Andrew Morton 已提交
2394 2395
static void slab_put_obj(struct kmem_cache *cachep, struct slab *slabp,
				void *objp, int nodeid)
2396
{
2397
	unsigned int objnr = obj_to_index(cachep, slabp, objp);
2398 2399 2400 2401 2402 2403 2404

#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 "
A
Andrew Morton 已提交
2405
				"'%s', objp %p\n", cachep->name, objp);
2406 2407 2408 2409 2410 2411 2412 2413
		BUG();
	}
#endif
	slab_bufctl(slabp)[objnr] = slabp->free;
	slabp->free = objnr;
	slabp->inuse--;
}

A
Andrew Morton 已提交
2414 2415
static void set_slab_attr(struct kmem_cache *cachep, struct slab *slabp,
			void *objp)
L
Linus Torvalds 已提交
2416 2417 2418 2419 2420 2421
{
	int i;
	struct page *page;

	/* Nasty!!!!!! I hope this is OK. */
	page = virt_to_page(objp);
2422 2423 2424 2425

	i = 1;
	if (likely(!PageCompound(page)))
		i <<= cachep->gfporder;
L
Linus Torvalds 已提交
2426
	do {
2427 2428
		page_set_cache(page, cachep);
		page_set_slab(page, slabp);
L
Linus Torvalds 已提交
2429 2430 2431 2432 2433 2434 2435 2436
		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.
 */
2437
static int cache_grow(struct kmem_cache *cachep, gfp_t flags, int nodeid)
L
Linus Torvalds 已提交
2438
{
P
Pekka Enberg 已提交
2439 2440 2441 2442 2443
	struct slab *slabp;
	void *objp;
	size_t offset;
	gfp_t local_flags;
	unsigned long ctor_flags;
2444
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
2445

A
Andrew Morton 已提交
2446 2447 2448
	/*
	 * Be lazy and only check for valid flags here,  keeping it out of the
	 * critical path in kmem_cache_alloc().
L
Linus Torvalds 已提交
2449
	 */
P
Pekka Enberg 已提交
2450
	if (flags & ~(SLAB_DMA | SLAB_LEVEL_MASK | SLAB_NO_GROW))
L
Linus Torvalds 已提交
2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463
		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;

2464
	/* Take the l3 list lock to change the colour_next on this node */
L
Linus Torvalds 已提交
2465
	check_irq_off();
2466 2467
	l3 = cachep->nodelists[nodeid];
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2468 2469

	/* Get colour for the slab, and cal the next value. */
2470 2471 2472 2473 2474
	offset = l3->colour_next;
	l3->colour_next++;
	if (l3->colour_next >= cachep->colour)
		l3->colour_next = 0;
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2475

2476
	offset *= cachep->colour_off;
L
Linus Torvalds 已提交
2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488

	if (local_flags & __GFP_WAIT)
		local_irq_enable();

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

A
Andrew Morton 已提交
2489 2490 2491
	/*
	 * Get mem for the objs.  Attempt to allocate a physical page from
	 * 'nodeid'.
2492
	 */
A
Andrew Morton 已提交
2493 2494
	objp = kmem_getpages(cachep, flags, nodeid);
	if (!objp)
L
Linus Torvalds 已提交
2495 2496 2497
		goto failed;

	/* Get slab management. */
A
Andrew Morton 已提交
2498 2499
	slabp = alloc_slabmgmt(cachep, objp, offset, local_flags);
	if (!slabp)
L
Linus Torvalds 已提交
2500 2501
		goto opps1;

2502
	slabp->nodeid = nodeid;
L
Linus Torvalds 已提交
2503 2504 2505 2506 2507 2508 2509
	set_slab_attr(cachep, slabp, objp);

	cache_init_objs(cachep, slabp, ctor_flags);

	if (local_flags & __GFP_WAIT)
		local_irq_disable();
	check_irq_off();
2510
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2511 2512

	/* Make slab active. */
2513
	list_add_tail(&slabp->list, &(l3->slabs_free));
L
Linus Torvalds 已提交
2514
	STATS_INC_GROWN(cachep);
2515 2516
	l3->free_objects += cachep->num;
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2517
	return 1;
A
Andrew Morton 已提交
2518
opps1:
L
Linus Torvalds 已提交
2519
	kmem_freepages(cachep, objp);
A
Andrew Morton 已提交
2520
failed:
L
Linus Torvalds 已提交
2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539
	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
2540 2541
		       (unsigned long)objp);
		BUG();
L
Linus Torvalds 已提交
2542 2543 2544
	}
	page = virt_to_page(objp);
	if (!PageSlab(page)) {
P
Pekka Enberg 已提交
2545 2546
		printk(KERN_ERR "kfree_debugcheck: bad ptr %lxh.\n",
		       (unsigned long)objp);
L
Linus Torvalds 已提交
2547 2548 2549 2550
		BUG();
	}
}

2551
static void *cache_free_debugcheck(struct kmem_cache *cachep, void *objp,
P
Pekka Enberg 已提交
2552
				   void *caller)
L
Linus Torvalds 已提交
2553 2554 2555 2556 2557
{
	struct page *page;
	unsigned int objnr;
	struct slab *slabp;

2558
	objp -= obj_offset(cachep);
L
Linus Torvalds 已提交
2559 2560 2561
	kfree_debugcheck(objp);
	page = virt_to_page(objp);

2562
	if (page_get_cache(page) != cachep) {
A
Andrew Morton 已提交
2563 2564
		printk(KERN_ERR "mismatch in kmem_cache_free: expected "
				"cache %p, got %p\n",
P
Pekka Enberg 已提交
2565
		       page_get_cache(page), cachep);
L
Linus Torvalds 已提交
2566
		printk(KERN_ERR "%p is %s.\n", cachep, cachep->name);
P
Pekka Enberg 已提交
2567 2568
		printk(KERN_ERR "%p is %s.\n", page_get_cache(page),
		       page_get_cache(page)->name);
L
Linus Torvalds 已提交
2569 2570
		WARN_ON(1);
	}
2571
	slabp = page_get_slab(page);
L
Linus Torvalds 已提交
2572 2573

	if (cachep->flags & SLAB_RED_ZONE) {
A
Andrew Morton 已提交
2574 2575 2576 2577 2578 2579
		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",
P
Pekka Enberg 已提交
2580 2581
			       objp, *dbg_redzone1(cachep, objp),
			       *dbg_redzone2(cachep, objp));
L
Linus Torvalds 已提交
2582 2583 2584 2585 2586 2587 2588
		}
		*dbg_redzone1(cachep, objp) = RED_INACTIVE;
		*dbg_redzone2(cachep, objp) = RED_INACTIVE;
	}
	if (cachep->flags & SLAB_STORE_USER)
		*dbg_userword(cachep, objp) = caller;

2589
	objnr = obj_to_index(cachep, slabp, objp);
L
Linus Torvalds 已提交
2590 2591

	BUG_ON(objnr >= cachep->num);
2592
	BUG_ON(objp != index_to_obj(cachep, slabp, objnr));
L
Linus Torvalds 已提交
2593 2594

	if (cachep->flags & SLAB_DEBUG_INITIAL) {
A
Andrew Morton 已提交
2595 2596 2597 2598
		/*
		 * Need to call the slab's constructor so the caller can
		 * perform a verify of its state (debugging).  Called without
		 * the cache-lock held.
L
Linus Torvalds 已提交
2599
		 */
2600
		cachep->ctor(objp + obj_offset(cachep),
P
Pekka Enberg 已提交
2601
			     cachep, SLAB_CTOR_CONSTRUCTOR | SLAB_CTOR_VERIFY);
L
Linus Torvalds 已提交
2602 2603 2604 2605 2606
	}
	if (cachep->flags & SLAB_POISON && cachep->dtor) {
		/* we want to cache poison the object,
		 * call the destruction callback
		 */
2607
		cachep->dtor(objp + obj_offset(cachep), cachep, 0);
L
Linus Torvalds 已提交
2608 2609 2610
	}
	if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
A
Andrew Morton 已提交
2611
		if ((cachep->buffer_size % PAGE_SIZE)==0 && OFF_SLAB(cachep)) {
L
Linus Torvalds 已提交
2612
			store_stackinfo(cachep, objp, (unsigned long)caller);
P
Pekka Enberg 已提交
2613
			kernel_map_pages(virt_to_page(objp),
2614
					 cachep->buffer_size / PAGE_SIZE, 0);
L
Linus Torvalds 已提交
2615 2616 2617 2618 2619 2620 2621 2622 2623 2624
		} else {
			poison_obj(cachep, objp, POISON_FREE);
		}
#else
		poison_obj(cachep, objp, POISON_FREE);
#endif
	}
	return objp;
}

2625
static void check_slabp(struct kmem_cache *cachep, struct slab *slabp)
L
Linus Torvalds 已提交
2626 2627 2628
{
	kmem_bufctl_t i;
	int entries = 0;
P
Pekka Enberg 已提交
2629

L
Linus Torvalds 已提交
2630 2631 2632 2633 2634 2635 2636
	/* Check slab's freelist to see if this obj is there. */
	for (i = slabp->free; i != BUFCTL_END; i = slab_bufctl(slabp)[i]) {
		entries++;
		if (entries > cachep->num || i >= cachep->num)
			goto bad;
	}
	if (entries != cachep->num - slabp->inuse) {
A
Andrew Morton 已提交
2637 2638 2639 2640
bad:
		printk(KERN_ERR "slab: Internal list corruption detected in "
				"cache '%s'(%d), slabp %p(%d). Hexdump:\n",
			cachep->name, cachep->num, slabp, slabp->inuse);
P
Pekka Enberg 已提交
2641
		for (i = 0;
2642
		     i < sizeof(*slabp) + cachep->num * sizeof(kmem_bufctl_t);
P
Pekka Enberg 已提交
2643
		     i++) {
A
Andrew Morton 已提交
2644
			if (i % 16 == 0)
L
Linus Torvalds 已提交
2645
				printk("\n%03x:", i);
P
Pekka Enberg 已提交
2646
			printk(" %02x", ((unsigned char *)slabp)[i]);
L
Linus Torvalds 已提交
2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657
		}
		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

2658
static void *cache_alloc_refill(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2659 2660 2661 2662 2663 2664
{
	int batchcount;
	struct kmem_list3 *l3;
	struct array_cache *ac;

	check_irq_off();
2665
	ac = cpu_cache_get(cachep);
A
Andrew Morton 已提交
2666
retry:
L
Linus Torvalds 已提交
2667 2668
	batchcount = ac->batchcount;
	if (!ac->touched && batchcount > BATCHREFILL_LIMIT) {
A
Andrew Morton 已提交
2669 2670 2671 2672
		/*
		 * If there was little recent activity on this cache, then
		 * perform only a partial refill.  Otherwise we could generate
		 * refill bouncing.
L
Linus Torvalds 已提交
2673 2674 2675
		 */
		batchcount = BATCHREFILL_LIMIT;
	}
2676 2677 2678 2679
	l3 = cachep->nodelists[numa_node_id()];

	BUG_ON(ac->avail > 0 || !l3);
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2680 2681 2682 2683 2684 2685 2686 2687

	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;
2688
			memcpy(ac->entry,
P
Pekka Enberg 已提交
2689 2690
			       &(shared_array->entry[shared_array->avail]),
			       sizeof(void *) * batchcount);
L
Linus Torvalds 已提交
2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714
			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);

2715 2716
			ac->entry[ac->avail++] = slab_get_obj(cachep, slabp,
							    numa_node_id());
L
Linus Torvalds 已提交
2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727
		}
		check_slabp(cachep, slabp);

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

A
Andrew Morton 已提交
2728
must_grow:
L
Linus Torvalds 已提交
2729
	l3->free_objects -= ac->avail;
A
Andrew Morton 已提交
2730
alloc_done:
2731
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2732 2733 2734

	if (unlikely(!ac->avail)) {
		int x;
2735 2736
		x = cache_grow(cachep, flags, numa_node_id());

A
Andrew Morton 已提交
2737
		/* cache_grow can reenable interrupts, then ac could change. */
2738
		ac = cpu_cache_get(cachep);
A
Andrew Morton 已提交
2739
		if (!x && ac->avail == 0)	/* no objects in sight? abort */
L
Linus Torvalds 已提交
2740 2741
			return NULL;

A
Andrew Morton 已提交
2742
		if (!ac->avail)		/* objects refilled by interrupt? */
L
Linus Torvalds 已提交
2743 2744 2745
			goto retry;
	}
	ac->touched = 1;
2746
	return ac->entry[--ac->avail];
L
Linus Torvalds 已提交
2747 2748
}

A
Andrew Morton 已提交
2749 2750
static inline void cache_alloc_debugcheck_before(struct kmem_cache *cachep,
						gfp_t flags)
L
Linus Torvalds 已提交
2751 2752 2753 2754 2755 2756 2757 2758
{
	might_sleep_if(flags & __GFP_WAIT);
#if DEBUG
	kmem_flagcheck(cachep, flags);
#endif
}

#if DEBUG
A
Andrew Morton 已提交
2759 2760
static void *cache_alloc_debugcheck_after(struct kmem_cache *cachep,
				gfp_t flags, void *objp, void *caller)
L
Linus Torvalds 已提交
2761
{
P
Pekka Enberg 已提交
2762
	if (!objp)
L
Linus Torvalds 已提交
2763
		return objp;
P
Pekka Enberg 已提交
2764
	if (cachep->flags & SLAB_POISON) {
L
Linus Torvalds 已提交
2765
#ifdef CONFIG_DEBUG_PAGEALLOC
2766
		if ((cachep->buffer_size % PAGE_SIZE) == 0 && OFF_SLAB(cachep))
P
Pekka Enberg 已提交
2767
			kernel_map_pages(virt_to_page(objp),
2768
					 cachep->buffer_size / PAGE_SIZE, 1);
L
Linus Torvalds 已提交
2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779
		else
			check_poison_obj(cachep, objp);
#else
		check_poison_obj(cachep, objp);
#endif
		poison_obj(cachep, objp, POISON_INUSE);
	}
	if (cachep->flags & SLAB_STORE_USER)
		*dbg_userword(cachep, objp) = caller;

	if (cachep->flags & SLAB_RED_ZONE) {
A
Andrew Morton 已提交
2780 2781 2782 2783
		if (*dbg_redzone1(cachep, objp) != RED_INACTIVE ||
				*dbg_redzone2(cachep, objp) != RED_INACTIVE) {
			slab_error(cachep, "double free, or memory outside"
						" object was overwritten");
P
Pekka Enberg 已提交
2784
			printk(KERN_ERR
A
Andrew Morton 已提交
2785 2786 2787
				"%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
2788 2789 2790 2791
		}
		*dbg_redzone1(cachep, objp) = RED_ACTIVE;
		*dbg_redzone2(cachep, objp) = RED_ACTIVE;
	}
2792
	objp += obj_offset(cachep);
L
Linus Torvalds 已提交
2793
	if (cachep->ctor && cachep->flags & SLAB_POISON) {
P
Pekka Enberg 已提交
2794
		unsigned long ctor_flags = SLAB_CTOR_CONSTRUCTOR;
L
Linus Torvalds 已提交
2795 2796 2797 2798 2799

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

		cachep->ctor(objp, cachep, ctor_flags);
P
Pekka Enberg 已提交
2800
	}
L
Linus Torvalds 已提交
2801 2802 2803 2804 2805 2806
	return objp;
}
#else
#define cache_alloc_debugcheck_after(a,b,objp,d) (objp)
#endif

2807
static inline void *____cache_alloc(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2808
{
P
Pekka Enberg 已提交
2809
	void *objp;
L
Linus Torvalds 已提交
2810 2811
	struct array_cache *ac;

2812
#ifdef CONFIG_NUMA
2813
	if (unlikely(current->mempolicy && !in_interrupt())) {
2814 2815 2816 2817 2818 2819 2820
		int nid = slab_node(current->mempolicy);

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

2821
	check_irq_off();
2822
	ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
2823 2824 2825
	if (likely(ac->avail)) {
		STATS_INC_ALLOCHIT(cachep);
		ac->touched = 1;
2826
		objp = ac->entry[--ac->avail];
L
Linus Torvalds 已提交
2827 2828 2829 2830
	} else {
		STATS_INC_ALLOCMISS(cachep);
		objp = cache_alloc_refill(cachep, flags);
	}
2831 2832 2833
	return objp;
}

A
Andrew Morton 已提交
2834 2835
static __always_inline void *__cache_alloc(struct kmem_cache *cachep,
						gfp_t flags, void *caller)
2836 2837
{
	unsigned long save_flags;
P
Pekka Enberg 已提交
2838
	void *objp;
2839 2840 2841 2842 2843

	cache_alloc_debugcheck_before(cachep, flags);

	local_irq_save(save_flags);
	objp = ____cache_alloc(cachep, flags);
L
Linus Torvalds 已提交
2844
	local_irq_restore(save_flags);
2845
	objp = cache_alloc_debugcheck_after(cachep, flags, objp,
2846
					    caller);
2847
	prefetchw(objp);
L
Linus Torvalds 已提交
2848 2849 2850
	return objp;
}

2851 2852 2853
#ifdef CONFIG_NUMA
/*
 * A interface to enable slab creation on nodeid
L
Linus Torvalds 已提交
2854
 */
A
Andrew Morton 已提交
2855 2856
static void *__cache_alloc_node(struct kmem_cache *cachep, gfp_t flags,
				int nodeid)
2857 2858
{
	struct list_head *entry;
P
Pekka Enberg 已提交
2859 2860 2861 2862 2863 2864 2865 2866
	struct slab *slabp;
	struct kmem_list3 *l3;
	void *obj;
	int x;

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

A
Andrew Morton 已提交
2867
retry:
2868
	check_irq_off();
P
Pekka Enberg 已提交
2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887
	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);

2888
	obj = slab_get_obj(cachep, slabp, nodeid);
P
Pekka Enberg 已提交
2889 2890 2891 2892 2893
	check_slabp(cachep, slabp);
	l3->free_objects--;
	/* move slabp to correct slabp list: */
	list_del(&slabp->list);

A
Andrew Morton 已提交
2894
	if (slabp->free == BUFCTL_END)
P
Pekka Enberg 已提交
2895
		list_add(&slabp->list, &l3->slabs_full);
A
Andrew Morton 已提交
2896
	else
P
Pekka Enberg 已提交
2897
		list_add(&slabp->list, &l3->slabs_partial);
2898

P
Pekka Enberg 已提交
2899 2900
	spin_unlock(&l3->list_lock);
	goto done;
2901

A
Andrew Morton 已提交
2902
must_grow:
P
Pekka Enberg 已提交
2903 2904
	spin_unlock(&l3->list_lock);
	x = cache_grow(cachep, flags, nodeid);
L
Linus Torvalds 已提交
2905

P
Pekka Enberg 已提交
2906 2907
	if (!x)
		return NULL;
2908

P
Pekka Enberg 已提交
2909
	goto retry;
A
Andrew Morton 已提交
2910
done:
P
Pekka Enberg 已提交
2911
	return obj;
2912 2913 2914 2915 2916 2917
}
#endif

/*
 * Caller needs to acquire correct kmem_list's list_lock
 */
2918
static void free_block(struct kmem_cache *cachep, void **objpp, int nr_objects,
P
Pekka Enberg 已提交
2919
		       int node)
L
Linus Torvalds 已提交
2920 2921
{
	int i;
2922
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
2923 2924 2925 2926 2927

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

2928
		slabp = virt_to_slab(objp);
2929
		l3 = cachep->nodelists[node];
L
Linus Torvalds 已提交
2930
		list_del(&slabp->list);
2931
		check_spinlock_acquired_node(cachep, node);
L
Linus Torvalds 已提交
2932
		check_slabp(cachep, slabp);
2933
		slab_put_obj(cachep, slabp, objp, node);
L
Linus Torvalds 已提交
2934
		STATS_DEC_ACTIVE(cachep);
2935
		l3->free_objects++;
L
Linus Torvalds 已提交
2936 2937 2938 2939
		check_slabp(cachep, slabp);

		/* fixup slab chains */
		if (slabp->inuse == 0) {
2940 2941
			if (l3->free_objects > l3->free_limit) {
				l3->free_objects -= cachep->num;
L
Linus Torvalds 已提交
2942 2943
				slab_destroy(cachep, slabp);
			} else {
2944
				list_add(&slabp->list, &l3->slabs_free);
L
Linus Torvalds 已提交
2945 2946 2947 2948 2949 2950
			}
		} else {
			/* Unconditionally move a slab to the end of the
			 * partial list on free - maximum time for the
			 * other objects to be freed, too.
			 */
2951
			list_add_tail(&slabp->list, &l3->slabs_partial);
L
Linus Torvalds 已提交
2952 2953 2954 2955
		}
	}
}

2956
static void cache_flusharray(struct kmem_cache *cachep, struct array_cache *ac)
L
Linus Torvalds 已提交
2957 2958
{
	int batchcount;
2959
	struct kmem_list3 *l3;
2960
	int node = numa_node_id();
L
Linus Torvalds 已提交
2961 2962 2963 2964 2965 2966

	batchcount = ac->batchcount;
#if DEBUG
	BUG_ON(!batchcount || batchcount > ac->avail);
#endif
	check_irq_off();
2967
	l3 = cachep->nodelists[node];
2968 2969 2970
	spin_lock(&l3->list_lock);
	if (l3->shared) {
		struct array_cache *shared_array = l3->shared;
P
Pekka Enberg 已提交
2971
		int max = shared_array->limit - shared_array->avail;
L
Linus Torvalds 已提交
2972 2973 2974
		if (max) {
			if (batchcount > max)
				batchcount = max;
2975
			memcpy(&(shared_array->entry[shared_array->avail]),
P
Pekka Enberg 已提交
2976
			       ac->entry, sizeof(void *) * batchcount);
L
Linus Torvalds 已提交
2977 2978 2979 2980 2981
			shared_array->avail += batchcount;
			goto free_done;
		}
	}

2982
	free_block(cachep, ac->entry, batchcount, node);
A
Andrew Morton 已提交
2983
free_done:
L
Linus Torvalds 已提交
2984 2985 2986 2987 2988
#if STATS
	{
		int i = 0;
		struct list_head *p;

2989 2990
		p = l3->slabs_free.next;
		while (p != &(l3->slabs_free)) {
L
Linus Torvalds 已提交
2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001
			struct slab *slabp;

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

			i++;
			p = p->next;
		}
		STATS_SET_FREEABLE(cachep, i);
	}
#endif
3002
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
3003
	ac->avail -= batchcount;
A
Andrew Morton 已提交
3004
	memmove(ac->entry, &(ac->entry[batchcount]), sizeof(void *)*ac->avail);
L
Linus Torvalds 已提交
3005 3006 3007
}

/*
A
Andrew Morton 已提交
3008 3009
 * Release an obj back to its cache. If the obj has a constructed state, it must
 * be in this state _before_ it is released.  Called with disabled ints.
L
Linus Torvalds 已提交
3010
 */
3011
static inline void __cache_free(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
3012
{
3013
	struct array_cache *ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
3014 3015 3016 3017

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

3018 3019 3020 3021 3022 3023
	/* Make sure we are not freeing a object from another
	 * node to the array cache on this cpu.
	 */
#ifdef CONFIG_NUMA
	{
		struct slab *slabp;
3024
		slabp = virt_to_slab(objp);
3025 3026 3027
		if (unlikely(slabp->nodeid != numa_node_id())) {
			struct array_cache *alien = NULL;
			int nodeid = slabp->nodeid;
A
Andrew Morton 已提交
3028
			struct kmem_list3 *l3;
3029

A
Andrew Morton 已提交
3030
			l3 = cachep->nodelists[numa_node_id()];
3031 3032 3033 3034 3035 3036
			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
3037
							    alien, nodeid);
3038 3039 3040 3041
				alien->entry[alien->avail++] = objp;
				spin_unlock(&alien->lock);
			} else {
				spin_lock(&(cachep->nodelists[nodeid])->
P
Pekka Enberg 已提交
3042
					  list_lock);
3043
				free_block(cachep, &objp, 1, nodeid);
3044
				spin_unlock(&(cachep->nodelists[nodeid])->
P
Pekka Enberg 已提交
3045
					    list_lock);
3046 3047 3048 3049 3050
			}
			return;
		}
	}
#endif
L
Linus Torvalds 已提交
3051 3052
	if (likely(ac->avail < ac->limit)) {
		STATS_INC_FREEHIT(cachep);
3053
		ac->entry[ac->avail++] = objp;
L
Linus Torvalds 已提交
3054 3055 3056 3057
		return;
	} else {
		STATS_INC_FREEMISS(cachep);
		cache_flusharray(cachep, ac);
3058
		ac->entry[ac->avail++] = objp;
L
Linus Torvalds 已提交
3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069
	}
}

/**
 * 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.
 */
3070
void *kmem_cache_alloc(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
3071
{
3072
	return __cache_alloc(cachep, flags, __builtin_return_address(0));
L
Linus Torvalds 已提交
3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089
}
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.
 */
3090
int fastcall kmem_ptr_validate(struct kmem_cache *cachep, void *ptr)
L
Linus Torvalds 已提交
3091
{
P
Pekka Enberg 已提交
3092
	unsigned long addr = (unsigned long)ptr;
L
Linus Torvalds 已提交
3093
	unsigned long min_addr = PAGE_OFFSET;
P
Pekka Enberg 已提交
3094
	unsigned long align_mask = BYTES_PER_WORD - 1;
3095
	unsigned long size = cachep->buffer_size;
L
Linus Torvalds 已提交
3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110
	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;
3111
	if (unlikely(page_get_cache(page) != cachep))
L
Linus Torvalds 已提交
3112 3113
		goto out;
	return 1;
A
Andrew Morton 已提交
3114
out:
L
Linus Torvalds 已提交
3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127
	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.
3128 3129
 * 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
3130
 */
3131
void *kmem_cache_alloc_node(struct kmem_cache *cachep, gfp_t flags, int nodeid)
L
Linus Torvalds 已提交
3132
{
3133 3134
	unsigned long save_flags;
	void *ptr;
L
Linus Torvalds 已提交
3135

3136 3137
	cache_alloc_debugcheck_before(cachep, flags);
	local_irq_save(save_flags);
3138 3139

	if (nodeid == -1 || nodeid == numa_node_id() ||
A
Andrew Morton 已提交
3140
			!cachep->nodelists[nodeid])
3141 3142 3143
		ptr = ____cache_alloc(cachep, flags);
	else
		ptr = __cache_alloc_node(cachep, flags, nodeid);
3144
	local_irq_restore(save_flags);
3145 3146 3147

	ptr = cache_alloc_debugcheck_after(cachep, flags, ptr,
					   __builtin_return_address(0));
L
Linus Torvalds 已提交
3148

3149
	return ptr;
L
Linus Torvalds 已提交
3150 3151 3152
}
EXPORT_SYMBOL(kmem_cache_alloc_node);

A
Al Viro 已提交
3153
void *kmalloc_node(size_t size, gfp_t flags, int node)
3154
{
3155
	struct kmem_cache *cachep;
3156 3157 3158 3159 3160 3161 3162

	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
3163 3164 3165 3166 3167 3168
#endif

/**
 * kmalloc - allocate memory
 * @size: how many bytes of memory are required.
 * @flags: the type of memory to allocate.
3169
 * @caller: function caller for debug tracking of the caller
L
Linus Torvalds 已提交
3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186
 *
 * 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.
 */
3187 3188
static __always_inline void *__do_kmalloc(size_t size, gfp_t flags,
					  void *caller)
L
Linus Torvalds 已提交
3189
{
3190
	struct kmem_cache *cachep;
L
Linus Torvalds 已提交
3191

3192 3193 3194 3195 3196 3197
	/* 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);
3198 3199
	if (unlikely(cachep == NULL))
		return NULL;
3200 3201 3202 3203 3204 3205 3206 3207
	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
3208 3209 3210
}
EXPORT_SYMBOL(__kmalloc);

3211 3212 3213 3214 3215 3216 3217 3218 3219 3220
#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
3221 3222 3223 3224 3225 3226 3227 3228
#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.
 */
3229
void *__alloc_percpu(size_t size)
L
Linus Torvalds 已提交
3230 3231
{
	int i;
P
Pekka Enberg 已提交
3232
	struct percpu_data *pdata = kmalloc(sizeof(*pdata), GFP_KERNEL);
L
Linus Torvalds 已提交
3233 3234 3235 3236

	if (!pdata)
		return NULL;

3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248
	/*
	 * 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
3249 3250 3251 3252 3253 3254 3255

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

	/* Catch derefs w/o wrappers */
P
Pekka Enberg 已提交
3256
	return (void *)(~(unsigned long)pdata);
L
Linus Torvalds 已提交
3257

A
Andrew Morton 已提交
3258
unwind_oom:
L
Linus Torvalds 已提交
3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277
	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.
 */
3278
void kmem_cache_free(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291
{
	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.
 *
3292 3293
 * If @objp is NULL, no operation is performed.
 *
L
Linus Torvalds 已提交
3294 3295 3296 3297 3298
 * Don't free memory not originally allocated by kmalloc()
 * or you will run into trouble.
 */
void kfree(const void *objp)
{
3299
	struct kmem_cache *c;
L
Linus Torvalds 已提交
3300 3301 3302 3303 3304 3305
	unsigned long flags;

	if (unlikely(!objp))
		return;
	local_irq_save(flags);
	kfree_debugcheck(objp);
3306
	c = virt_to_cache(objp);
3307
	mutex_debug_check_no_locks_freed(objp, obj_size(c));
P
Pekka Enberg 已提交
3308
	__cache_free(c, (void *)objp);
L
Linus Torvalds 已提交
3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320
	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
3321
void free_percpu(const void *objp)
L
Linus Torvalds 已提交
3322 3323
{
	int i;
P
Pekka Enberg 已提交
3324
	struct percpu_data *p = (struct percpu_data *)(~(unsigned long)objp);
L
Linus Torvalds 已提交
3325

3326 3327 3328 3329
	/*
	 * We allocate for all cpus so we cannot use for online cpu here.
	 */
	for_each_cpu(i)
P
Pekka Enberg 已提交
3330
	    kfree(p->ptrs[i]);
L
Linus Torvalds 已提交
3331 3332 3333 3334 3335
	kfree(p);
}
EXPORT_SYMBOL(free_percpu);
#endif

3336
unsigned int kmem_cache_size(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
3337
{
3338
	return obj_size(cachep);
L
Linus Torvalds 已提交
3339 3340 3341
}
EXPORT_SYMBOL(kmem_cache_size);

3342
const char *kmem_cache_name(struct kmem_cache *cachep)
3343 3344 3345 3346 3347
{
	return cachep->name;
}
EXPORT_SYMBOL_GPL(kmem_cache_name);

3348 3349 3350
/*
 * This initializes kmem_list3 for all nodes.
 */
3351
static int alloc_kmemlist(struct kmem_cache *cachep)
3352 3353 3354 3355 3356 3357 3358 3359 3360
{
	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
A
Andrew Morton 已提交
3361 3362
		new_alien = alloc_alien_cache(node, cachep->limit);
		if (!new_alien)
3363 3364
			goto fail;
#endif
A
Andrew Morton 已提交
3365 3366 3367
		new = alloc_arraycache(node, cachep->shared*cachep->batchcount,
					0xbaadf00d);
		if (!new)
3368
			goto fail;
A
Andrew Morton 已提交
3369 3370
		l3 = cachep->nodelists[node];
		if (l3) {
3371 3372
			spin_lock_irq(&l3->list_lock);

A
Andrew Morton 已提交
3373 3374
			nc = cachep->nodelists[node]->shared;
			if (nc)
P
Pekka Enberg 已提交
3375
				free_block(cachep, nc->entry, nc->avail, node);
3376 3377 3378 3379 3380 3381

			l3->shared = new;
			if (!cachep->nodelists[node]->alien) {
				l3->alien = new_alien;
				new_alien = NULL;
			}
P
Pekka Enberg 已提交
3382
			l3->free_limit = (1 + nr_cpus_node(node)) *
A
Andrew Morton 已提交
3383
					cachep->batchcount + cachep->num;
3384 3385 3386 3387 3388
			spin_unlock_irq(&l3->list_lock);
			kfree(nc);
			free_alien_cache(new_alien);
			continue;
		}
A
Andrew Morton 已提交
3389 3390
		l3 = kmalloc_node(sizeof(struct kmem_list3), GFP_KERNEL, node);
		if (!l3)
3391 3392 3393 3394
			goto fail;

		kmem_list3_init(l3);
		l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
A
Andrew Morton 已提交
3395
				((unsigned long)cachep) % REAPTIMEOUT_LIST3;
3396 3397
		l3->shared = new;
		l3->alien = new_alien;
P
Pekka Enberg 已提交
3398
		l3->free_limit = (1 + nr_cpus_node(node)) *
A
Andrew Morton 已提交
3399
					cachep->batchcount + cachep->num;
3400 3401 3402
		cachep->nodelists[node] = l3;
	}
	return err;
A
Andrew Morton 已提交
3403
fail:
3404 3405 3406 3407
	err = -ENOMEM;
	return err;
}

L
Linus Torvalds 已提交
3408
struct ccupdate_struct {
3409
	struct kmem_cache *cachep;
L
Linus Torvalds 已提交
3410 3411 3412 3413 3414
	struct array_cache *new[NR_CPUS];
};

static void do_ccupdate_local(void *info)
{
A
Andrew Morton 已提交
3415
	struct ccupdate_struct *new = info;
L
Linus Torvalds 已提交
3416 3417 3418
	struct array_cache *old;

	check_irq_off();
3419
	old = cpu_cache_get(new->cachep);
3420

L
Linus Torvalds 已提交
3421 3422 3423 3424
	new->cachep->array[smp_processor_id()] = new->new[smp_processor_id()];
	new->new[smp_processor_id()] = old;
}

3425
/* Always called with the cache_chain_mutex held */
A
Andrew Morton 已提交
3426 3427
static int do_tune_cpucache(struct kmem_cache *cachep, int limit,
				int batchcount, int shared)
L
Linus Torvalds 已提交
3428 3429
{
	struct ccupdate_struct new;
3430
	int i, err;
L
Linus Torvalds 已提交
3431

P
Pekka Enberg 已提交
3432
	memset(&new.new, 0, sizeof(new.new));
3433
	for_each_online_cpu(i) {
A
Andrew Morton 已提交
3434 3435
		new.new[i] = alloc_arraycache(cpu_to_node(i), limit,
						batchcount);
3436
		if (!new.new[i]) {
P
Pekka Enberg 已提交
3437 3438
			for (i--; i >= 0; i--)
				kfree(new.new[i]);
3439
			return -ENOMEM;
L
Linus Torvalds 已提交
3440 3441 3442 3443
		}
	}
	new.cachep = cachep;

A
Andrew Morton 已提交
3444
	on_each_cpu(do_ccupdate_local, (void *)&new, 1, 1);
3445

L
Linus Torvalds 已提交
3446 3447 3448
	check_irq_on();
	cachep->batchcount = batchcount;
	cachep->limit = limit;
3449
	cachep->shared = shared;
L
Linus Torvalds 已提交
3450

3451
	for_each_online_cpu(i) {
L
Linus Torvalds 已提交
3452 3453 3454
		struct array_cache *ccold = new.new[i];
		if (!ccold)
			continue;
3455
		spin_lock_irq(&cachep->nodelists[cpu_to_node(i)]->list_lock);
3456
		free_block(cachep, ccold->entry, ccold->avail, cpu_to_node(i));
3457
		spin_unlock_irq(&cachep->nodelists[cpu_to_node(i)]->list_lock);
L
Linus Torvalds 已提交
3458 3459 3460
		kfree(ccold);
	}

3461 3462 3463
	err = alloc_kmemlist(cachep);
	if (err) {
		printk(KERN_ERR "alloc_kmemlist failed for %s, error %d.\n",
P
Pekka Enberg 已提交
3464
		       cachep->name, -err);
3465
		BUG();
L
Linus Torvalds 已提交
3466 3467 3468 3469
	}
	return 0;
}

3470
/* Called with cache_chain_mutex held always */
3471
static void enable_cpucache(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
3472 3473 3474 3475
{
	int err;
	int limit, shared;

A
Andrew Morton 已提交
3476 3477
	/*
	 * The head array serves three purposes:
L
Linus Torvalds 已提交
3478 3479
	 * - create a LIFO ordering, i.e. return objects that are cache-warm
	 * - reduce the number of spinlock operations.
A
Andrew Morton 已提交
3480
	 * - reduce the number of linked list operations on the slab and
L
Linus Torvalds 已提交
3481 3482 3483 3484
	 *   bufctl chains: array operations are cheaper.
	 * The numbers are guessed, we should auto-tune as described by
	 * Bonwick.
	 */
3485
	if (cachep->buffer_size > 131072)
L
Linus Torvalds 已提交
3486
		limit = 1;
3487
	else if (cachep->buffer_size > PAGE_SIZE)
L
Linus Torvalds 已提交
3488
		limit = 8;
3489
	else if (cachep->buffer_size > 1024)
L
Linus Torvalds 已提交
3490
		limit = 24;
3491
	else if (cachep->buffer_size > 256)
L
Linus Torvalds 已提交
3492 3493 3494 3495
		limit = 54;
	else
		limit = 120;

A
Andrew Morton 已提交
3496 3497
	/*
	 * CPU bound tasks (e.g. network routing) can exhibit cpu bound
L
Linus Torvalds 已提交
3498 3499 3500 3501 3502 3503 3504 3505 3506
	 * 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
3507
	if (cachep->buffer_size <= PAGE_SIZE)
L
Linus Torvalds 已提交
3508 3509 3510 3511
		shared = 8;
#endif

#if DEBUG
A
Andrew Morton 已提交
3512 3513 3514
	/*
	 * With debugging enabled, large batchcount lead to excessively long
	 * periods with disabled local interrupts. Limit the batchcount
L
Linus Torvalds 已提交
3515 3516 3517 3518
	 */
	if (limit > 32)
		limit = 32;
#endif
P
Pekka Enberg 已提交
3519
	err = do_tune_cpucache(cachep, limit, (limit + 1) / 2, shared);
L
Linus Torvalds 已提交
3520 3521
	if (err)
		printk(KERN_ERR "enable_cpucache failed for %s, error %d.\n",
P
Pekka Enberg 已提交
3522
		       cachep->name, -err);
L
Linus Torvalds 已提交
3523 3524
}

A
Andrew Morton 已提交
3525 3526
static void drain_array_locked(struct kmem_cache *cachep,
				struct array_cache *ac, int force, int node)
L
Linus Torvalds 已提交
3527 3528 3529
{
	int tofree;

3530
	check_spinlock_acquired_node(cachep, node);
L
Linus Torvalds 已提交
3531 3532 3533
	if (ac->touched && !force) {
		ac->touched = 0;
	} else if (ac->avail) {
P
Pekka Enberg 已提交
3534
		tofree = force ? ac->avail : (ac->limit + 4) / 5;
A
Andrew Morton 已提交
3535
		if (tofree > ac->avail)
P
Pekka Enberg 已提交
3536
			tofree = (ac->avail + 1) / 2;
3537
		free_block(cachep, ac->entry, tofree, node);
L
Linus Torvalds 已提交
3538
		ac->avail -= tofree;
3539
		memmove(ac->entry, &(ac->entry[tofree]),
P
Pekka Enberg 已提交
3540
			sizeof(void *) * ac->avail);
L
Linus Torvalds 已提交
3541 3542 3543
	}
}

3544 3545 3546 3547 3548 3549

/*
 * Drain an array if it contains any elements taking the l3 lock only if
 * necessary.
 */
static void drain_array(struct kmem_cache *searchp, struct kmem_list3 *l3,
3550
			 struct array_cache *ac, int force, int node)
3551 3552 3553
{
	if (ac && ac->avail) {
		spin_lock_irq(&l3->list_lock);
3554
		drain_array_locked(searchp, ac, force, node);
3555 3556 3557 3558
		spin_unlock_irq(&l3->list_lock);
	}
}

L
Linus Torvalds 已提交
3559 3560
/**
 * cache_reap - Reclaim memory from caches.
3561
 * @unused: unused parameter
L
Linus Torvalds 已提交
3562 3563 3564 3565 3566 3567
 *
 * Called from workqueue/eventd every few seconds.
 * Purpose:
 * - clear the per-cpu caches for this CPU.
 * - return freeable pages to the main free memory pool.
 *
A
Andrew Morton 已提交
3568 3569
 * If we cannot acquire the cache chain mutex then just give up - we'll try
 * again on the next iteration.
L
Linus Torvalds 已提交
3570 3571 3572 3573
 */
static void cache_reap(void *unused)
{
	struct list_head *walk;
3574
	struct kmem_list3 *l3;
3575
	int node = numa_node_id();
L
Linus Torvalds 已提交
3576

I
Ingo Molnar 已提交
3577
	if (!mutex_trylock(&cache_chain_mutex)) {
L
Linus Torvalds 已提交
3578
		/* Give up. Setup the next iteration. */
P
Pekka Enberg 已提交
3579 3580
		schedule_delayed_work(&__get_cpu_var(reap_work),
				      REAPTIMEOUT_CPUC);
L
Linus Torvalds 已提交
3581 3582 3583 3584
		return;
	}

	list_for_each(walk, &cache_chain) {
3585
		struct kmem_cache *searchp;
P
Pekka Enberg 已提交
3586
		struct list_head *p;
L
Linus Torvalds 已提交
3587 3588 3589
		int tofree;
		struct slab *slabp;

3590
		searchp = list_entry(walk, struct kmem_cache, next);
L
Linus Torvalds 已提交
3591 3592
		check_irq_on();

3593 3594 3595 3596 3597
		/*
		 * We only take the l3 lock if absolutely necessary and we
		 * have established with reasonable certainty that
		 * we can do some work if the lock was obtained.
		 */
3598
		l3 = searchp->nodelists[node];
3599

3600
		reap_alien(searchp, l3);
L
Linus Torvalds 已提交
3601

3602
		drain_array(searchp, l3, cpu_cache_get(searchp), 0, node);
L
Linus Torvalds 已提交
3603

3604 3605 3606 3607
		/*
		 * These are racy checks but it does not matter
		 * if we skip one check or scan twice.
		 */
3608
		if (time_after(l3->next_reap, jiffies))
3609
			goto next;
L
Linus Torvalds 已提交
3610

3611
		l3->next_reap = jiffies + REAPTIMEOUT_LIST3;
L
Linus Torvalds 已提交
3612

3613
		drain_array(searchp, l3, l3->shared, 0, node);
L
Linus Torvalds 已提交
3614

3615 3616
		if (l3->free_touched) {
			l3->free_touched = 0;
3617
			goto next;
L
Linus Torvalds 已提交
3618 3619
		}

A
Andrew Morton 已提交
3620 3621
		tofree = (l3->free_limit + 5 * searchp->num - 1) /
				(5 * searchp->num);
L
Linus Torvalds 已提交
3622
		do {
3623 3624 3625 3626 3627 3628 3629
			/*
			 * Do not lock if there are no free blocks.
			 */
			if (list_empty(&l3->slabs_free))
				break;

			spin_lock_irq(&l3->list_lock);
3630
			p = l3->slabs_free.next;
3631 3632
			if (p == &(l3->slabs_free)) {
				spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
3633
				break;
3634
			}
L
Linus Torvalds 已提交
3635 3636 3637 3638 3639 3640

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

A
Andrew Morton 已提交
3641 3642 3643
			/*
			 * Safe to drop the lock. The slab is no longer linked
			 * to the cache. searchp cannot disappear, we hold
L
Linus Torvalds 已提交
3644 3645
			 * cache_chain_lock
			 */
3646 3647
			l3->free_objects -= searchp->num;
			spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
3648
			slab_destroy(searchp, slabp);
P
Pekka Enberg 已提交
3649
		} while (--tofree > 0);
3650
next:
L
Linus Torvalds 已提交
3651 3652 3653
		cond_resched();
	}
	check_irq_on();
I
Ingo Molnar 已提交
3654
	mutex_unlock(&cache_chain_mutex);
3655
	next_reap_node();
A
Andrew Morton 已提交
3656
	/* Set up the next iteration */
3657
	schedule_delayed_work(&__get_cpu_var(reap_work), REAPTIMEOUT_CPUC);
L
Linus Torvalds 已提交
3658 3659 3660 3661
}

#ifdef CONFIG_PROC_FS

3662
static void print_slabinfo_header(struct seq_file *m)
L
Linus Torvalds 已提交
3663
{
3664 3665 3666 3667
	/*
	 * Output format version, so at least we can change it
	 * without _too_ many complaints.
	 */
L
Linus Torvalds 已提交
3668
#if STATS
3669
	seq_puts(m, "slabinfo - version: 2.1 (statistics)\n");
L
Linus Torvalds 已提交
3670
#else
3671
	seq_puts(m, "slabinfo - version: 2.1\n");
L
Linus Torvalds 已提交
3672
#endif
3673 3674 3675 3676
	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
3677
#if STATS
3678 3679 3680
	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
3681
#endif
3682 3683 3684 3685 3686 3687 3688 3689
	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
3690
	mutex_lock(&cache_chain_mutex);
3691 3692
	if (!n)
		print_slabinfo_header(m);
L
Linus Torvalds 已提交
3693 3694 3695 3696 3697 3698
	p = cache_chain.next;
	while (n--) {
		p = p->next;
		if (p == &cache_chain)
			return NULL;
	}
3699
	return list_entry(p, struct kmem_cache, next);
L
Linus Torvalds 已提交
3700 3701 3702 3703
}

static void *s_next(struct seq_file *m, void *p, loff_t *pos)
{
3704
	struct kmem_cache *cachep = p;
L
Linus Torvalds 已提交
3705
	++*pos;
A
Andrew Morton 已提交
3706 3707
	return cachep->next.next == &cache_chain ?
		NULL : list_entry(cachep->next.next, struct kmem_cache, next);
L
Linus Torvalds 已提交
3708 3709 3710 3711
}

static void s_stop(struct seq_file *m, void *p)
{
I
Ingo Molnar 已提交
3712
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
3713 3714 3715 3716
}

static int s_show(struct seq_file *m, void *p)
{
3717
	struct kmem_cache *cachep = p;
L
Linus Torvalds 已提交
3718
	struct list_head *q;
P
Pekka Enberg 已提交
3719 3720 3721 3722 3723
	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;
3724
	const char *name;
L
Linus Torvalds 已提交
3725
	char *error = NULL;
3726 3727
	int node;
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
3728 3729 3730

	active_objs = 0;
	num_slabs = 0;
3731 3732 3733 3734 3735
	for_each_online_node(node) {
		l3 = cachep->nodelists[node];
		if (!l3)
			continue;

3736 3737
		check_irq_on();
		spin_lock_irq(&l3->list_lock);
3738

P
Pekka Enberg 已提交
3739
		list_for_each(q, &l3->slabs_full) {
3740 3741 3742 3743 3744 3745
			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
3746
		list_for_each(q, &l3->slabs_partial) {
3747 3748 3749 3750 3751 3752 3753 3754
			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
3755
		list_for_each(q, &l3->slabs_free) {
3756 3757 3758 3759 3760 3761
			slabp = list_entry(q, struct slab, list);
			if (slabp->inuse && !error)
				error = "slabs_free/inuse accounting error";
			num_slabs++;
		}
		free_objects += l3->free_objects;
3762 3763
		if (l3->shared)
			shared_avail += l3->shared->avail;
3764

3765
		spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
3766
	}
P
Pekka Enberg 已提交
3767 3768
	num_slabs += active_slabs;
	num_objs = num_slabs * cachep->num;
3769
	if (num_objs - active_objs != free_objects && !error)
L
Linus Torvalds 已提交
3770 3771
		error = "free_objects accounting error";

P
Pekka Enberg 已提交
3772
	name = cachep->name;
L
Linus Torvalds 已提交
3773 3774 3775 3776
	if (error)
		printk(KERN_ERR "slab: cache %s error: %s\n", name, error);

	seq_printf(m, "%-17s %6lu %6lu %6u %4u %4d",
3777
		   name, active_objs, num_objs, cachep->buffer_size,
P
Pekka Enberg 已提交
3778
		   cachep->num, (1 << cachep->gfporder));
L
Linus Torvalds 已提交
3779
	seq_printf(m, " : tunables %4u %4u %4u",
P
Pekka Enberg 已提交
3780
		   cachep->limit, cachep->batchcount, cachep->shared);
3781
	seq_printf(m, " : slabdata %6lu %6lu %6lu",
P
Pekka Enberg 已提交
3782
		   active_slabs, num_slabs, shared_avail);
L
Linus Torvalds 已提交
3783
#if STATS
P
Pekka Enberg 已提交
3784
	{			/* list3 stats */
L
Linus Torvalds 已提交
3785 3786 3787 3788 3789 3790 3791
		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;
3792
		unsigned long node_frees = cachep->node_frees;
L
Linus Torvalds 已提交
3793

3794
		seq_printf(m, " : globalstat %7lu %6lu %5lu %4lu \
A
Andrew Morton 已提交
3795 3796 3797
				%4lu %4lu %4lu %4lu", allocs, high, grown,
				reaped, errors, max_freeable, node_allocs,
				node_frees);
L
Linus Torvalds 已提交
3798 3799 3800 3801 3802 3803 3804 3805 3806
	}
	/* 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
3807
			   allochit, allocmiss, freehit, freemiss);
L
Linus Torvalds 已提交
3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828
	}
#endif
	seq_putc(m, '\n');
	return 0;
}

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

struct seq_operations slabinfo_op = {
P
Pekka Enberg 已提交
3829 3830 3831 3832
	.start = s_start,
	.next = s_next,
	.stop = s_stop,
	.show = s_show,
L
Linus Torvalds 已提交
3833 3834 3835 3836 3837 3838 3839 3840 3841 3842
};

#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
3843 3844
ssize_t slabinfo_write(struct file *file, const char __user * buffer,
		       size_t count, loff_t *ppos)
L
Linus Torvalds 已提交
3845
{
P
Pekka Enberg 已提交
3846
	char kbuf[MAX_SLABINFO_WRITE + 1], *tmp;
L
Linus Torvalds 已提交
3847 3848
	int limit, batchcount, shared, res;
	struct list_head *p;
P
Pekka Enberg 已提交
3849

L
Linus Torvalds 已提交
3850 3851 3852 3853
	if (count > MAX_SLABINFO_WRITE)
		return -EINVAL;
	if (copy_from_user(&kbuf, buffer, count))
		return -EFAULT;
P
Pekka Enberg 已提交
3854
	kbuf[MAX_SLABINFO_WRITE] = '\0';
L
Linus Torvalds 已提交
3855 3856 3857 3858 3859 3860 3861 3862 3863 3864

	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
3865
	mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
3866
	res = -EINVAL;
P
Pekka Enberg 已提交
3867
	list_for_each(p, &cache_chain) {
A
Andrew Morton 已提交
3868
		struct kmem_cache *cachep;
L
Linus Torvalds 已提交
3869

A
Andrew Morton 已提交
3870
		cachep = list_entry(p, struct kmem_cache, next);
L
Linus Torvalds 已提交
3871
		if (!strcmp(cachep->name, kbuf)) {
A
Andrew Morton 已提交
3872 3873
			if (limit < 1 || batchcount < 1 ||
					batchcount > limit || shared < 0) {
3874
				res = 0;
L
Linus Torvalds 已提交
3875
			} else {
3876
				res = do_tune_cpucache(cachep, limit,
P
Pekka Enberg 已提交
3877
						       batchcount, shared);
L
Linus Torvalds 已提交
3878 3879 3880 3881
			}
			break;
		}
	}
I
Ingo Molnar 已提交
3882
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
3883 3884 3885 3886 3887 3888
	if (res >= 0)
		res = count;
	return res;
}
#endif

3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900
/**
 * 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
3901 3902
unsigned int ksize(const void *objp)
{
3903 3904
	if (unlikely(objp == NULL))
		return 0;
L
Linus Torvalds 已提交
3905

3906
	return obj_size(virt_to_cache(objp));
L
Linus Torvalds 已提交
3907
}