slab.c 95.4 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70
/*
 * linux/mm/slab.c
 * Written by Mark Hemment, 1996/97.
 * (markhe@nextd.demon.co.uk)
 *
 * kmem_cache_destroy() + some cleanup - 1999 Andrea Arcangeli
 *
 * Major cleanup, different bufctl logic, per-cpu arrays
 *	(c) 2000 Manfred Spraul
 *
 * Cleanup, make the head arrays unconditional, preparation for NUMA
 * 	(c) 2002 Manfred Spraul
 *
 * An implementation of the Slab Allocator as described in outline in;
 *	UNIX Internals: The New Frontiers by Uresh Vahalia
 *	Pub: Prentice Hall	ISBN 0-13-101908-2
 * or with a little more detail in;
 *	The Slab Allocator: An Object-Caching Kernel Memory Allocator
 *	Jeff Bonwick (Sun Microsystems).
 *	Presented at: USENIX Summer 1994 Technical Conference
 *
 * The memory is organized in caches, one cache for each object type.
 * (e.g. inode_cache, dentry_cache, buffer_head, vm_area_struct)
 * Each cache consists out of many slabs (they are small (usually one
 * page long) and always contiguous), and each slab contains multiple
 * initialized objects.
 *
 * This means, that your constructor is used only for newly allocated
 * slabs and you must pass objects with the same intializations to
 * kmem_cache_free.
 *
 * Each cache can only support one memory type (GFP_DMA, GFP_HIGHMEM,
 * normal). If you need a special memory type, then must create a new
 * cache for that memory type.
 *
 * In order to reduce fragmentation, the slabs are sorted in 3 groups:
 *   full slabs with 0 free objects
 *   partial slabs
 *   empty slabs with no allocated objects
 *
 * If partial slabs exist, then new allocations come from these slabs,
 * otherwise from empty slabs or new slabs are allocated.
 *
 * kmem_cache_destroy() CAN CRASH if you try to allocate from the cache
 * during kmem_cache_destroy(). The caller must prevent concurrent allocs.
 *
 * Each cache has a short per-cpu head array, most allocs
 * and frees go into that array, and if that array overflows, then 1/2
 * of the entries in the array are given back into the global cache.
 * The head array is strictly LIFO and should improve the cache hit rates.
 * On SMP, it additionally reduces the spinlock operations.
 *
 * The c_cpuarray may not be read with enabled local interrupts - 
 * it's changed with a smp_call_function().
 *
 * SMP synchronization:
 *  constructors and destructors are called without any locking.
 *  Several members in kmem_cache_t and struct slab never change, they
 *	are accessed without any locking.
 *  The per-cpu arrays are never accessed from the wrong cpu, no locking,
 *  	and local interrupts are disabled so slab code is preempt-safe.
 *  The non-constant members are protected with a per-cache irq spinlock.
 *
 * Many thanks to Mark Hemment, who wrote another per-cpu slab patch
 * in 2000 - many ideas in the current implementation are derived from
 * his patch.
 *
 * Further notes from the original documentation:
 *
 * 11 April '97.  Started multi-threading - markhe
I
Ingo Molnar 已提交
71
 *	The global cache-chain is protected by the mutex 'cache_chain_mutex'.
L
Linus Torvalds 已提交
72 73 74 75 76 77
 *	The sem is only needed when accessing/extending the cache-chain, which
 *	can never happen inside an interrupt (kmem_cache_create(),
 *	kmem_cache_shrink() and kmem_cache_reap()).
 *
 *	At present, each engine can be growing a cache.  This should be blocked.
 *
78 79 80 81 82 83 84 85 86
 * 15 March 2005. NUMA slab allocator.
 *	Shai Fultheim <shai@scalex86.org>.
 *	Shobhit Dayal <shobhit@calsoftinc.com>
 *	Alok N Kataria <alokk@calsoftinc.com>
 *	Christoph Lameter <christoph@lameter.com>
 *
 *	Modified the slab allocator to be node aware on NUMA systems.
 *	Each node has its own list of partial, free and full slabs.
 *	All object allocations for a node occur from node specific slab lists.
L
Linus Torvalds 已提交
87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103
 */

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

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

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

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

/* Shouldn't this be in a header file somewhere? */
#define	BYTES_PER_WORD		sizeof(void *)

#ifndef cache_line_size
#define cache_line_size()	L1_CACHE_BYTES
#endif

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

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

#ifndef ARCH_KMALLOC_FLAGS
#define ARCH_KMALLOC_FLAGS SLAB_HWCACHE_ALIGN
#endif

/* Legal flag mask for kmem_cache_create(). */
#if DEBUG
# define CREATE_MASK	(SLAB_DEBUG_INITIAL | SLAB_RED_ZONE | \
			 SLAB_POISON | SLAB_HWCACHE_ALIGN | \
			 SLAB_NO_REAP | SLAB_CACHE_DMA | \
			 SLAB_MUST_HWCACHE_ALIGN | SLAB_STORE_USER | \
			 SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
			 SLAB_DESTROY_BY_RCU)
#else
# define CREATE_MASK	(SLAB_HWCACHE_ALIGN | SLAB_NO_REAP | \
			 SLAB_CACHE_DMA | SLAB_MUST_HWCACHE_ALIGN | \
			 SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
			 SLAB_DESTROY_BY_RCU)
#endif

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

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

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

/*
 * struct slab
 *
 * Manages the objs in a slab. Placed either at the beginning of mem allocated
 * for a slab, or allocated from an general cache.
 * Slabs are chained into three list: fully used, partial, fully free slabs.
 */
struct slab {
P
Pekka Enberg 已提交
221 222 223 224 225 226
	struct list_head list;
	unsigned long colouroff;
	void *s_mem;		/* including colour offset */
	unsigned int inuse;	/* num of objs active in slab */
	kmem_bufctl_t free;
	unsigned short nodeid;
L
Linus Torvalds 已提交
227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245
};

/*
 * struct slab_rcu
 *
 * slab_destroy on a SLAB_DESTROY_BY_RCU cache uses this structure to
 * arrange for kmem_freepages to be called via RCU.  This is useful if
 * we need to approach a kernel structure obliquely, from its address
 * obtained without the usual locking.  We can lock the structure to
 * stabilize it and check it's still at the given address, only if we
 * can be sure that the memory has not been meanwhile reused for some
 * other kind of object (which our subsystem's lock might corrupt).
 *
 * rcu_read_lock before reading the address, then rcu_read_unlock after
 * taking the spinlock within the structure expected at that address.
 *
 * We assume struct slab_rcu can overlay struct slab when destroying.
 */
struct slab_rcu {
P
Pekka Enberg 已提交
246 247 248
	struct rcu_head head;
	kmem_cache_t *cachep;
	void *addr;
L
Linus Torvalds 已提交
249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267
};

/*
 * struct array_cache
 *
 * Purpose:
 * - LIFO ordering, to hand out cache-warm objects from _alloc
 * - reduce the number of linked list operations
 * - reduce spinlock operations
 *
 * The limit is stored in the per-cpu structure to reduce the data cache
 * footprint.
 *
 */
struct array_cache {
	unsigned int avail;
	unsigned int limit;
	unsigned int batchcount;
	unsigned int touched;
268 269 270 271 272 273 274
	spinlock_t lock;
	void *entry[0];		/*
				 * Must have this definition in here for the proper
				 * alignment of array_cache. Also simplifies accessing
				 * the entries.
				 * [0] is for gcc 2.95. It should really be [].
				 */
L
Linus Torvalds 已提交
275 276 277 278 279 280 281 282
};

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

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

302 303 304 305 306 307 308 309 310 311
/*
 * Need this for bootstrapping a per node allocator.
 */
#define NUM_INIT_LISTS (2 * MAX_NUMNODES + 1)
struct kmem_list3 __initdata initkmem_list3[NUM_INIT_LISTS];
#define	CACHE_CACHE 0
#define	SIZE_AC 1
#define	SIZE_L3 (1 + MAX_NUMNODES)

/*
312
 * This function must be completely optimized away if
313 314 315 316
 * a constant is passed to it. Mostly the same as
 * what is in linux/slab.h except it returns an
 * index.
 */
317
static __always_inline int index_of(const size_t size)
318 319 320 321 322 323 324 325 326 327 328 329 330 331 332
{
	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
		{
			extern void __bad_size(void);
			__bad_size();
		}
333 334
	} else
		BUG();
335 336 337 338 339
	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
340

341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364
static inline void kmem_list3_init(struct kmem_list3 *parent)
{
	INIT_LIST_HEAD(&parent->slabs_full);
	INIT_LIST_HEAD(&parent->slabs_partial);
	INIT_LIST_HEAD(&parent->slabs_free);
	parent->shared = NULL;
	parent->alien = NULL;
	spin_lock_init(&parent->list_lock);
	parent->free_objects = 0;
	parent->free_touched = 0;
}

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

#define	MAKE_ALL_LISTS(cachep, ptr, nodeid)			\
	do {					\
	MAKE_LIST((cachep), (&(ptr)->slabs_full), slabs_full, nodeid);	\
	MAKE_LIST((cachep), (&(ptr)->slabs_partial), slabs_partial, nodeid); \
	MAKE_LIST((cachep), (&(ptr)->slabs_free), slabs_free, nodeid);	\
	} while (0)
L
Linus Torvalds 已提交
365 366 367 368 369 370

/*
 * kmem_cache_t
 *
 * 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
374 375 376 377 378
	struct array_cache *array[NR_CPUS];
	unsigned int batchcount;
	unsigned int limit;
	unsigned int shared;
	unsigned int objsize;
379
/* 2) touched by every alloc & free from the backend */
P
Pekka Enberg 已提交
380 381 382 383
	struct kmem_list3 *nodelists[MAX_NUMNODES];
	unsigned int flags;	/* constant flags */
	unsigned int num;	/* # of objs per slab */
	spinlock_t spinlock;
L
Linus Torvalds 已提交
384 385 386

/* 3) cache_grow/shrink */
	/* order of pgs per slab (2^n) */
P
Pekka Enberg 已提交
387
	unsigned int gfporder;
L
Linus Torvalds 已提交
388 389

	/* force GFP flags, e.g. GFP_DMA */
P
Pekka Enberg 已提交
390
	gfp_t gfpflags;
L
Linus Torvalds 已提交
391

P
Pekka Enberg 已提交
392 393 394 395 396 397
	size_t colour;		/* cache colouring range */
	unsigned int colour_off;	/* colour offset */
	unsigned int colour_next;	/* cache colouring */
	kmem_cache_t *slabp_cache;
	unsigned int slab_size;
	unsigned int dflags;	/* dynamic flags */
L
Linus Torvalds 已提交
398 399

	/* constructor func */
P
Pekka Enberg 已提交
400
	void (*ctor) (void *, kmem_cache_t *, unsigned long);
L
Linus Torvalds 已提交
401 402

	/* de-constructor func */
P
Pekka Enberg 已提交
403
	void (*dtor) (void *, kmem_cache_t *, unsigned long);
L
Linus Torvalds 已提交
404 405

/* 4) cache creation/removal */
P
Pekka Enberg 已提交
406 407
	const char *name;
	struct list_head next;
L
Linus Torvalds 已提交
408 409 410

/* 5) statistics */
#if STATS
P
Pekka Enberg 已提交
411 412 413 414 415 416 417 418 419 420 421 422 423
	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
424 425
#endif
#if DEBUG
P
Pekka Enberg 已提交
426 427
	int dbghead;
	int reallen;
L
Linus Torvalds 已提交
428 429 430 431 432 433 434 435 436 437
#endif
};

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

#define BATCHREFILL_LIMIT	16
/* Optimization question: fewer reaps means less 
 * probability for unnessary cpucache drain/refill cycles.
 *
A
Adrian Bunk 已提交
438
 * OTOH the cpuarrays can contain lots of objects,
L
Linus Torvalds 已提交
439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454
 * which could lock up otherwise freeable slabs.
 */
#define REAPTIMEOUT_CPUC	(2*HZ)
#define REAPTIMEOUT_LIST3	(4*HZ)

#if STATS
#define	STATS_INC_ACTIVE(x)	((x)->num_active++)
#define	STATS_DEC_ACTIVE(x)	((x)->num_active--)
#define	STATS_INC_ALLOCED(x)	((x)->num_allocations++)
#define	STATS_INC_GROWN(x)	((x)->grown++)
#define	STATS_INC_REAPED(x)	((x)->reaped++)
#define	STATS_SET_HIGH(x)	do { if ((x)->num_active > (x)->high_mark) \
					(x)->high_mark = (x)->num_active; \
				} while (0)
#define	STATS_INC_ERR(x)	((x)->errors++)
#define	STATS_INC_NODEALLOCS(x)	((x)->node_allocs++)
455
#define	STATS_INC_NODEFREES(x)	((x)->node_frees++)
L
Linus Torvalds 已提交
456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473
#define	STATS_SET_FREEABLE(x, i) \
				do { if ((x)->max_freeable < i) \
					(x)->max_freeable = i; \
				} while (0)

#define STATS_INC_ALLOCHIT(x)	atomic_inc(&(x)->allochit)
#define STATS_INC_ALLOCMISS(x)	atomic_inc(&(x)->allocmiss)
#define STATS_INC_FREEHIT(x)	atomic_inc(&(x)->freehit)
#define STATS_INC_FREEMISS(x)	atomic_inc(&(x)->freemiss)
#else
#define	STATS_INC_ACTIVE(x)	do { } while (0)
#define	STATS_DEC_ACTIVE(x)	do { } while (0)
#define	STATS_INC_ALLOCED(x)	do { } while (0)
#define	STATS_INC_GROWN(x)	do { } while (0)
#define	STATS_INC_REAPED(x)	do { } while (0)
#define	STATS_SET_HIGH(x)	do { } while (0)
#define	STATS_INC_ERR(x)	do { } while (0)
#define	STATS_INC_NODEALLOCS(x)	do { } while (0)
474
#define	STATS_INC_NODEFREES(x)	do { } while (0)
L
Linus Torvalds 已提交
475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526
#define	STATS_SET_FREEABLE(x, i) \
				do { } while (0)

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

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

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

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

static int obj_reallen(kmem_cache_t *cachep)
{
	return cachep->reallen;
}

static unsigned long *dbg_redzone1(kmem_cache_t *cachep, void *objp)
{
	BUG_ON(!(cachep->flags & SLAB_RED_ZONE));
	return (unsigned long*) (objp+obj_dbghead(cachep)-BYTES_PER_WORD);
}

static unsigned long *dbg_redzone2(kmem_cache_t *cachep, void *objp)
{
	BUG_ON(!(cachep->flags & SLAB_RED_ZONE));
	if (cachep->flags & SLAB_STORE_USER)
P
Pekka Enberg 已提交
527 528 529
		return (unsigned long *)(objp + cachep->objsize -
					 2 * BYTES_PER_WORD);
	return (unsigned long *)(objp + cachep->objsize - BYTES_PER_WORD);
L
Linus Torvalds 已提交
530 531 532 533 534
}

static void **dbg_userword(kmem_cache_t *cachep, void *objp)
{
	BUG_ON(!(cachep->flags & SLAB_STORE_USER));
P
Pekka Enberg 已提交
535
	return (void **)(objp + cachep->objsize - BYTES_PER_WORD);
L
Linus Torvalds 已提交
536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569
}

#else

#define obj_dbghead(x)			0
#define obj_reallen(cachep)		(cachep->objsize)
#define dbg_redzone1(cachep, objp)	({BUG(); (unsigned long *)NULL;})
#define dbg_redzone2(cachep, objp)	({BUG(); (unsigned long *)NULL;})
#define dbg_userword(cachep, objp)	({BUG(); (void **)NULL;})

#endif

/*
 * Maximum size of an obj (in 2^order pages)
 * and absolute limit for the gfp order.
 */
#if defined(CONFIG_LARGE_ALLOCS)
#define	MAX_OBJ_ORDER	13	/* up to 32Mb */
#define	MAX_GFP_ORDER	13	/* up to 32Mb */
#elif defined(CONFIG_MMU)
#define	MAX_OBJ_ORDER	5	/* 32 pages */
#define	MAX_GFP_ORDER	5	/* 32 pages */
#else
#define	MAX_OBJ_ORDER	8	/* up to 1Mb */
#define	MAX_GFP_ORDER	8	/* up to 1Mb */
#endif

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

570
/* Functions for storing/retrieving the cachep and or slab from the
L
Linus Torvalds 已提交
571 572 573
 * global 'mem_map'. These are used to find the slab an obj belongs to.
 * With kfree(), these are used to find the cache which an obj belongs to.
 */
574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592
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)
{
	return (struct kmem_cache *)page->lru.next;
}

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

static inline struct slab *page_get_slab(struct page *page)
{
	return (struct slab *)page->lru.prev;
}
L
Linus Torvalds 已提交
593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611

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

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

static struct cache_names __initdata cache_names[] = {
#define CACHE(x) { .name = "size-" #x, .name_dma = "size-" #x "(DMA)" },
#include <linux/kmalloc_sizes.h>
P
Pekka Enberg 已提交
612
	{NULL,}
L
Linus Torvalds 已提交
613 614 615 616
#undef CACHE
};

static struct arraycache_init initarray_cache __initdata =
P
Pekka Enberg 已提交
617
    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
L
Linus Torvalds 已提交
618
static struct arraycache_init initarray_generic =
P
Pekka Enberg 已提交
619
    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
L
Linus Torvalds 已提交
620 621 622

/* internal cache of cache description objs */
static kmem_cache_t cache_cache = {
P
Pekka Enberg 已提交
623 624 625 626 627 628 629
	.batchcount = 1,
	.limit = BOOT_CPUCACHE_ENTRIES,
	.shared = 1,
	.objsize = sizeof(kmem_cache_t),
	.flags = SLAB_NO_REAP,
	.spinlock = SPIN_LOCK_UNLOCKED,
	.name = "kmem_cache",
L
Linus Torvalds 已提交
630
#if DEBUG
P
Pekka Enberg 已提交
631
	.reallen = sizeof(kmem_cache_t),
L
Linus Torvalds 已提交
632 633 634 635
#endif
};

/* Guard access to the cache-chain. */
I
Ingo Molnar 已提交
636
static DEFINE_MUTEX(cache_chain_mutex);
L
Linus Torvalds 已提交
637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652
static struct list_head cache_chain;

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

/*
 * chicken and egg problem: delay the per-cpu array allocation
 * until the general caches are up.
 */
static enum {
	NONE,
653 654
	PARTIAL_AC,
	PARTIAL_L3,
L
Linus Torvalds 已提交
655 656 657 658 659
	FULL
} g_cpucache_up;

static DEFINE_PER_CPU(struct work_struct, reap_work);

P
Pekka Enberg 已提交
660 661 662
static void free_block(kmem_cache_t *cachep, void **objpp, int len, int node);
static void enable_cpucache(kmem_cache_t *cachep);
static void cache_reap(void *unused);
663
static int __node_shrink(kmem_cache_t *cachep, int node);
L
Linus Torvalds 已提交
664 665 666 667 668 669

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

A
Al Viro 已提交
670
static inline kmem_cache_t *__find_general_cachep(size_t size, gfp_t gfpflags)
L
Linus Torvalds 已提交
671 672 673 674 675
{
	struct cache_sizes *csizep = malloc_sizes;

#if DEBUG
	/* This happens if someone tries to call
P
Pekka Enberg 已提交
676 677 678
	 * kmem_cache_create(), or __kmalloc(), before
	 * the generic caches are initialized.
	 */
679
	BUG_ON(malloc_sizes[INDEX_AC].cs_cachep == NULL);
L
Linus Torvalds 已提交
680 681 682 683 684
#endif
	while (size > csizep->cs_size)
		csizep++;

	/*
685
	 * Really subtle: The last entry with cs->cs_size==ULONG_MAX
L
Linus Torvalds 已提交
686 687 688 689 690 691 692 693
	 * has cs_{dma,}cachep==NULL. Thus no special case
	 * for large kmalloc calls required.
	 */
	if (unlikely(gfpflags & GFP_DMA))
		return csizep->cs_dmacachep;
	return csizep->cs_cachep;
}

A
Al Viro 已提交
694
kmem_cache_t *kmem_find_general_cachep(size_t size, gfp_t gfpflags)
695 696 697 698 699
{
	return __find_general_cachep(size, gfpflags);
}
EXPORT_SYMBOL(kmem_find_general_cachep);

L
Linus Torvalds 已提交
700 701
/* Cal the num objs, wastage, and bytes left over for a given slab size. */
static void cache_estimate(unsigned long gfporder, size_t size, size_t align,
P
Pekka Enberg 已提交
702
			   int flags, size_t *left_over, unsigned int *num)
L
Linus Torvalds 已提交
703 704
{
	int i;
P
Pekka Enberg 已提交
705
	size_t wastage = PAGE_SIZE << gfporder;
L
Linus Torvalds 已提交
706 707 708 709 710 711 712 713
	size_t extra = 0;
	size_t base = 0;

	if (!(flags & CFLGS_OFF_SLAB)) {
		base = sizeof(struct slab);
		extra = sizeof(kmem_bufctl_t);
	}
	i = 0;
P
Pekka Enberg 已提交
714
	while (i * size + ALIGN(base + i * extra, align) <= wastage)
L
Linus Torvalds 已提交
715 716 717 718 719 720 721 722
		i++;
	if (i > 0)
		i--;

	if (i > SLAB_LIMIT)
		i = SLAB_LIMIT;

	*num = i;
P
Pekka Enberg 已提交
723 724
	wastage -= i * size;
	wastage -= ALIGN(base + i * extra, align);
L
Linus Torvalds 已提交
725 726 727 728 729 730 731 732
	*left_over = wastage;
}

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

static void __slab_error(const char *function, kmem_cache_t *cachep, char *msg)
{
	printk(KERN_ERR "slab error in %s(): cache `%s': %s\n",
P
Pekka Enberg 已提交
733
	       function, cachep->name, msg);
L
Linus Torvalds 已提交
734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758
	dump_stack();
}

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

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

759
static struct array_cache *alloc_arraycache(int node, int entries,
P
Pekka Enberg 已提交
760
					    int batchcount)
L
Linus Torvalds 已提交
761
{
P
Pekka Enberg 已提交
762
	int memsize = sizeof(void *) * entries + sizeof(struct array_cache);
L
Linus Torvalds 已提交
763 764
	struct array_cache *nc = NULL;

765
	nc = kmalloc_node(memsize, GFP_KERNEL, node);
L
Linus Torvalds 已提交
766 767 768 769 770
	if (nc) {
		nc->avail = 0;
		nc->limit = entries;
		nc->batchcount = batchcount;
		nc->touched = 0;
771
		spin_lock_init(&nc->lock);
L
Linus Torvalds 已提交
772 773 774 775
	}
	return nc;
}

776
#ifdef CONFIG_NUMA
777 778
static void *__cache_alloc_node(kmem_cache_t *, gfp_t, int);

779 780 781
static inline struct array_cache **alloc_alien_cache(int node, int limit)
{
	struct array_cache **ac_ptr;
P
Pekka Enberg 已提交
782
	int memsize = sizeof(void *) * MAX_NUMNODES;
783 784 785 786 787 788 789 790 791 792 793 794 795
	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
796
				for (i--; i <= 0; i--)
797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813
					kfree(ac_ptr[i]);
				kfree(ac_ptr);
				return NULL;
			}
		}
	}
	return ac_ptr;
}

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

	if (!ac_ptr)
		return;

	for_each_node(i)
P
Pekka Enberg 已提交
814
	    kfree(ac_ptr[i]);
815 816 817 818

	kfree(ac_ptr);
}

P
Pekka Enberg 已提交
819 820
static inline void __drain_alien_cache(kmem_cache_t *cachep,
				       struct array_cache *ac, int node)
821 822 823 824 825
{
	struct kmem_list3 *rl3 = cachep->nodelists[node];

	if (ac->avail) {
		spin_lock(&rl3->list_lock);
826
		free_block(cachep, ac->entry, ac->avail, node);
827 828 829 830 831 832 833
		ac->avail = 0;
		spin_unlock(&rl3->list_lock);
	}
}

static void drain_alien_cache(kmem_cache_t *cachep, struct kmem_list3 *l3)
{
P
Pekka Enberg 已提交
834
	int i = 0;
835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852
	struct array_cache *ac;
	unsigned long flags;

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

L
Linus Torvalds 已提交
853
static int __devinit cpuup_callback(struct notifier_block *nfb,
P
Pekka Enberg 已提交
854
				    unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
855 856
{
	long cpu = (long)hcpu;
P
Pekka Enberg 已提交
857
	kmem_cache_t *cachep;
858 859 860
	struct kmem_list3 *l3 = NULL;
	int node = cpu_to_node(cpu);
	int memsize = sizeof(struct kmem_list3);
L
Linus Torvalds 已提交
861 862 863

	switch (action) {
	case CPU_UP_PREPARE:
I
Ingo Molnar 已提交
864
		mutex_lock(&cache_chain_mutex);
865 866 867 868 869 870
		/* we need to do this right in the beginning since
		 * alloc_arraycache's are going to use this list.
		 * kmalloc_node allows us to add the slab to the right
		 * kmem_list3 and not this cpu's kmem_list3
		 */

L
Linus Torvalds 已提交
871
		list_for_each_entry(cachep, &cache_chain, next) {
872 873 874 875 876 877
			/* setup the size64 kmemlist for cpu before we can
			 * begin anything. Make sure some other cpu on this
			 * node has not already allocated this
			 */
			if (!cachep->nodelists[node]) {
				if (!(l3 = kmalloc_node(memsize,
P
Pekka Enberg 已提交
878
							GFP_KERNEL, node)))
879 880 881
					goto bad;
				kmem_list3_init(l3);
				l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
P
Pekka Enberg 已提交
882
				    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
883 884 885

				cachep->nodelists[node] = l3;
			}
L
Linus Torvalds 已提交
886

887 888
			spin_lock_irq(&cachep->nodelists[node]->list_lock);
			cachep->nodelists[node]->free_limit =
P
Pekka Enberg 已提交
889 890
			    (1 + nr_cpus_node(node)) *
			    cachep->batchcount + cachep->num;
891 892 893 894
			spin_unlock_irq(&cachep->nodelists[node]->list_lock);
		}

		/* Now we can go ahead with allocating the shared array's
P
Pekka Enberg 已提交
895
		   & array cache's */
896
		list_for_each_entry(cachep, &cache_chain, next) {
897 898
			struct array_cache *nc;

899
			nc = alloc_arraycache(node, cachep->limit,
P
Pekka Enberg 已提交
900
					      cachep->batchcount);
L
Linus Torvalds 已提交
901 902 903 904
			if (!nc)
				goto bad;
			cachep->array[cpu] = nc;

905 906 907 908
			l3 = cachep->nodelists[node];
			BUG_ON(!l3);
			if (!l3->shared) {
				if (!(nc = alloc_arraycache(node,
P
Pekka Enberg 已提交
909 910 911 912
							    cachep->shared *
							    cachep->batchcount,
							    0xbaadf00d)))
					goto bad;
913 914

				/* we are serialised from CPU_DEAD or
P
Pekka Enberg 已提交
915
				   CPU_UP_CANCELLED by the cpucontrol lock */
916 917
				l3->shared = nc;
			}
L
Linus Torvalds 已提交
918
		}
I
Ingo Molnar 已提交
919
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
920 921 922 923 924 925 926 927
		break;
	case CPU_ONLINE:
		start_cpu_timer(cpu);
		break;
#ifdef CONFIG_HOTPLUG_CPU
	case CPU_DEAD:
		/* fall thru */
	case CPU_UP_CANCELED:
I
Ingo Molnar 已提交
928
		mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
929 930 931

		list_for_each_entry(cachep, &cache_chain, next) {
			struct array_cache *nc;
932
			cpumask_t mask;
L
Linus Torvalds 已提交
933

934
			mask = node_to_cpumask(node);
L
Linus Torvalds 已提交
935 936 937 938
			spin_lock_irq(&cachep->spinlock);
			/* cpu is dead; no one can alloc from it. */
			nc = cachep->array[cpu];
			cachep->array[cpu] = NULL;
939 940 941 942 943 944 945 946 947 948
			l3 = cachep->nodelists[node];

			if (!l3)
				goto unlock_cache;

			spin_lock(&l3->list_lock);

			/* Free limit for this kmem_list3 */
			l3->free_limit -= cachep->batchcount;
			if (nc)
949
				free_block(cachep, nc->entry, nc->avail, node);
950 951

			if (!cpus_empty(mask)) {
P
Pekka Enberg 已提交
952 953 954
				spin_unlock(&l3->list_lock);
				goto unlock_cache;
			}
955 956 957

			if (l3->shared) {
				free_block(cachep, l3->shared->entry,
P
Pekka Enberg 已提交
958
					   l3->shared->avail, node);
959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975
				kfree(l3->shared);
				l3->shared = NULL;
			}
			if (l3->alien) {
				drain_alien_cache(cachep, l3);
				free_alien_cache(l3->alien);
				l3->alien = NULL;
			}

			/* free slabs belonging to this node */
			if (__node_shrink(cachep, node)) {
				cachep->nodelists[node] = NULL;
				spin_unlock(&l3->list_lock);
				kfree(l3);
			} else {
				spin_unlock(&l3->list_lock);
			}
P
Pekka Enberg 已提交
976
		      unlock_cache:
L
Linus Torvalds 已提交
977 978 979
			spin_unlock_irq(&cachep->spinlock);
			kfree(nc);
		}
I
Ingo Molnar 已提交
980
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
981 982 983 984
		break;
#endif
	}
	return NOTIFY_OK;
P
Pekka Enberg 已提交
985
      bad:
I
Ingo Molnar 已提交
986
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
987 988 989 990 991
	return NOTIFY_BAD;
}

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

992 993 994
/*
 * swap the static kmem_list3 with kmalloced memory
 */
P
Pekka Enberg 已提交
995
static void init_list(kmem_cache_t *cachep, struct kmem_list3 *list, int nodeid)
996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009
{
	struct kmem_list3 *ptr;

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

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

L
Linus Torvalds 已提交
1010 1011 1012 1013 1014 1015 1016 1017
/* Initialisation.
 * Called after the gfp() functions have been enabled, and before smp_init().
 */
void __init kmem_cache_init(void)
{
	size_t left_over;
	struct cache_sizes *sizes;
	struct cache_names *names;
1018 1019 1020 1021 1022 1023 1024
	int i;

	for (i = 0; i < NUM_INIT_LISTS; i++) {
		kmem_list3_init(&initkmem_list3[i]);
		if (i < MAX_NUMNODES)
			cache_cache.nodelists[i] = NULL;
	}
L
Linus Torvalds 已提交
1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037

	/*
	 * 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:
	 * 1) initialize the cache_cache cache: it contains the kmem_cache_t
	 *    structures of all caches, except cache_cache itself: cache_cache
	 *    is statically allocated.
1038 1039 1040
	 *    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
1041
	 * 2) Create the first kmalloc cache.
1042 1043 1044 1045
	 *    The kmem_cache_t for the new cache is allocated normally.
	 *    An __init data area is used for the head array.
	 * 3) Create the remaining kmalloc caches, with minimally sized
	 *    head arrays.
L
Linus Torvalds 已提交
1046 1047
	 * 4) Replace the __init data head arrays for cache_cache and the first
	 *    kmalloc cache with kmalloc allocated arrays.
1048 1049 1050
	 * 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
1051 1052 1053 1054 1055 1056 1057
	 */

	/* 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;
1058
	cache_cache.nodelists[numa_node_id()] = &initkmem_list3[CACHE_CACHE];
L
Linus Torvalds 已提交
1059 1060 1061 1062

	cache_cache.objsize = ALIGN(cache_cache.objsize, cache_line_size());

	cache_estimate(0, cache_cache.objsize, cache_line_size(), 0,
P
Pekka Enberg 已提交
1063
		       &left_over, &cache_cache.num);
L
Linus Torvalds 已提交
1064 1065 1066
	if (!cache_cache.num)
		BUG();

P
Pekka Enberg 已提交
1067
	cache_cache.colour = left_over / cache_cache.colour_off;
L
Linus Torvalds 已提交
1068
	cache_cache.colour_next = 0;
P
Pekka Enberg 已提交
1069 1070
	cache_cache.slab_size = ALIGN(cache_cache.num * sizeof(kmem_bufctl_t) +
				      sizeof(struct slab), cache_line_size());
L
Linus Torvalds 已提交
1071 1072 1073 1074 1075

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

1076 1077 1078 1079 1080 1081
	/* Initialize the caches that provide memory for the array cache
	 * and the kmem_list3 structures first.
	 * Without this, further allocations will bug
	 */

	sizes[INDEX_AC].cs_cachep = kmem_cache_create(names[INDEX_AC].name,
P
Pekka Enberg 已提交
1082 1083 1084 1085
						      sizes[INDEX_AC].cs_size,
						      ARCH_KMALLOC_MINALIGN,
						      (ARCH_KMALLOC_FLAGS |
						       SLAB_PANIC), NULL, NULL);
1086 1087 1088

	if (INDEX_AC != INDEX_L3)
		sizes[INDEX_L3].cs_cachep =
P
Pekka Enberg 已提交
1089 1090 1091 1092 1093
		    kmem_cache_create(names[INDEX_L3].name,
				      sizes[INDEX_L3].cs_size,
				      ARCH_KMALLOC_MINALIGN,
				      (ARCH_KMALLOC_FLAGS | SLAB_PANIC), NULL,
				      NULL);
1094

L
Linus Torvalds 已提交
1095
	while (sizes->cs_size != ULONG_MAX) {
1096 1097
		/*
		 * For performance, all the general caches are L1 aligned.
L
Linus Torvalds 已提交
1098 1099 1100
		 * This should be particularly beneficial on SMP boxes, as it
		 * eliminates "false sharing".
		 * Note for systems short on memory removing the alignment will
1101 1102
		 * allow tighter packing of the smaller caches.
		 */
P
Pekka Enberg 已提交
1103
		if (!sizes->cs_cachep)
1104
			sizes->cs_cachep = kmem_cache_create(names->name,
P
Pekka Enberg 已提交
1105 1106 1107 1108 1109
							     sizes->cs_size,
							     ARCH_KMALLOC_MINALIGN,
							     (ARCH_KMALLOC_FLAGS
							      | SLAB_PANIC),
							     NULL, NULL);
L
Linus Torvalds 已提交
1110 1111 1112

		/* Inc off-slab bufctl limit until the ceiling is hit. */
		if (!(OFF_SLAB(sizes->cs_cachep))) {
P
Pekka Enberg 已提交
1113
			offslab_limit = sizes->cs_size - sizeof(struct slab);
L
Linus Torvalds 已提交
1114 1115 1116 1117
			offslab_limit /= sizeof(kmem_bufctl_t);
		}

		sizes->cs_dmacachep = kmem_cache_create(names->name_dma,
P
Pekka Enberg 已提交
1118 1119 1120 1121 1122 1123
							sizes->cs_size,
							ARCH_KMALLOC_MINALIGN,
							(ARCH_KMALLOC_FLAGS |
							 SLAB_CACHE_DMA |
							 SLAB_PANIC), NULL,
							NULL);
L
Linus Torvalds 已提交
1124 1125 1126 1127 1128 1129

		sizes++;
		names++;
	}
	/* 4) Replace the bootstrap head arrays */
	{
P
Pekka Enberg 已提交
1130
		void *ptr;
1131

L
Linus Torvalds 已提交
1132
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
1133

L
Linus Torvalds 已提交
1134 1135
		local_irq_disable();
		BUG_ON(ac_data(&cache_cache) != &initarray_cache.cache);
1136
		memcpy(ptr, ac_data(&cache_cache),
P
Pekka Enberg 已提交
1137
		       sizeof(struct arraycache_init));
L
Linus Torvalds 已提交
1138 1139
		cache_cache.array[smp_processor_id()] = ptr;
		local_irq_enable();
1140

L
Linus Torvalds 已提交
1141
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
1142

L
Linus Torvalds 已提交
1143
		local_irq_disable();
1144
		BUG_ON(ac_data(malloc_sizes[INDEX_AC].cs_cachep)
P
Pekka Enberg 已提交
1145
		       != &initarray_generic.cache);
1146
		memcpy(ptr, ac_data(malloc_sizes[INDEX_AC].cs_cachep),
P
Pekka Enberg 已提交
1147
		       sizeof(struct arraycache_init));
1148
		malloc_sizes[INDEX_AC].cs_cachep->array[smp_processor_id()] =
P
Pekka Enberg 已提交
1149
		    ptr;
L
Linus Torvalds 已提交
1150 1151
		local_irq_enable();
	}
1152 1153 1154 1155 1156
	/* 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
1157
			  numa_node_id());
1158 1159 1160

		for_each_online_node(node) {
			init_list(malloc_sizes[INDEX_AC].cs_cachep,
P
Pekka Enberg 已提交
1161
				  &initkmem_list3[SIZE_AC + node], node);
1162 1163 1164

			if (INDEX_AC != INDEX_L3) {
				init_list(malloc_sizes[INDEX_L3].cs_cachep,
P
Pekka Enberg 已提交
1165 1166
					  &initkmem_list3[SIZE_L3 + node],
					  node);
1167 1168 1169
			}
		}
	}
L
Linus Torvalds 已提交
1170

1171
	/* 6) resize the head arrays to their final sizes */
L
Linus Torvalds 已提交
1172 1173
	{
		kmem_cache_t *cachep;
I
Ingo Molnar 已提交
1174
		mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1175
		list_for_each_entry(cachep, &cache_chain, next)
P
Pekka Enberg 已提交
1176
		    enable_cpucache(cachep);
I
Ingo Molnar 已提交
1177
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200
	}

	/* Done! */
	g_cpucache_up = FULL;

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

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

static int __init cpucache_init(void)
{
	int cpu;

	/* 
	 * Register the timers that return unneeded
	 * pages to gfp.
	 */
1201
	for_each_online_cpu(cpu)
P
Pekka Enberg 已提交
1202
	    start_cpu_timer(cpu);
L
Linus Torvalds 已提交
1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215

	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.
 */
A
Al Viro 已提交
1216
static void *kmem_getpages(kmem_cache_t *cachep, gfp_t flags, int nodeid)
L
Linus Torvalds 已提交
1217 1218 1219 1220 1221 1222
{
	struct page *page;
	void *addr;
	int i;

	flags |= cachep->gfpflags;
1223
	page = alloc_pages_node(nodeid, flags, cachep->gfporder);
L
Linus Torvalds 已提交
1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243
	if (!page)
		return NULL;
	addr = page_address(page);

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

/*
 * Interface to system's page release.
 */
static void kmem_freepages(kmem_cache_t *cachep, void *addr)
{
P
Pekka Enberg 已提交
1244
	unsigned long i = (1 << cachep->gfporder);
L
Linus Torvalds 已提交
1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256
	struct page *page = virt_to_page(addr);
	const unsigned long nr_freed = i;

	while (i--) {
		if (!TestClearPageSlab(page))
			BUG();
		page++;
	}
	sub_page_state(nr_slab, nr_freed);
	if (current->reclaim_state)
		current->reclaim_state->reclaimed_slab += nr_freed;
	free_pages((unsigned long)addr, cachep->gfporder);
P
Pekka Enberg 已提交
1257 1258
	if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
		atomic_sub(1 << cachep->gfporder, &slab_reclaim_pages);
L
Linus Torvalds 已提交
1259 1260 1261 1262
}

static void kmem_rcu_free(struct rcu_head *head)
{
P
Pekka Enberg 已提交
1263
	struct slab_rcu *slab_rcu = (struct slab_rcu *)head;
L
Linus Torvalds 已提交
1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274
	kmem_cache_t *cachep = slab_rcu->cachep;

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

#if DEBUG

#ifdef CONFIG_DEBUG_PAGEALLOC
static void store_stackinfo(kmem_cache_t *cachep, unsigned long *addr,
P
Pekka Enberg 已提交
1275
			    unsigned long caller)
L
Linus Torvalds 已提交
1276 1277 1278
{
	int size = obj_reallen(cachep);

P
Pekka Enberg 已提交
1279
	addr = (unsigned long *)&((char *)addr)[obj_dbghead(cachep)];
L
Linus Torvalds 已提交
1280

P
Pekka Enberg 已提交
1281
	if (size < 5 * sizeof(unsigned long))
L
Linus Torvalds 已提交
1282 1283
		return;

P
Pekka Enberg 已提交
1284 1285 1286 1287
	*addr++ = 0x12345678;
	*addr++ = caller;
	*addr++ = smp_processor_id();
	size -= 3 * sizeof(unsigned long);
L
Linus Torvalds 已提交
1288 1289 1290 1291 1292 1293 1294
	{
		unsigned long *sptr = &caller;
		unsigned long svalue;

		while (!kstack_end(sptr)) {
			svalue = *sptr++;
			if (kernel_text_address(svalue)) {
P
Pekka Enberg 已提交
1295
				*addr++ = svalue;
L
Linus Torvalds 已提交
1296 1297 1298 1299 1300 1301 1302
				size -= sizeof(unsigned long);
				if (size <= sizeof(unsigned long))
					break;
			}
		}

	}
P
Pekka Enberg 已提交
1303
	*addr++ = 0x87654321;
L
Linus Torvalds 已提交
1304 1305 1306 1307 1308 1309
}
#endif

static void poison_obj(kmem_cache_t *cachep, void *addr, unsigned char val)
{
	int size = obj_reallen(cachep);
P
Pekka Enberg 已提交
1310
	addr = &((char *)addr)[obj_dbghead(cachep)];
L
Linus Torvalds 已提交
1311 1312

	memset(addr, val, size);
P
Pekka Enberg 已提交
1313
	*(unsigned char *)(addr + size - 1) = POISON_END;
L
Linus Torvalds 已提交
1314 1315 1316 1317 1318 1319
}

static void dump_line(char *data, int offset, int limit)
{
	int i;
	printk(KERN_ERR "%03x:", offset);
P
Pekka Enberg 已提交
1320 1321
	for (i = 0; i < limit; i++) {
		printk(" %02x", (unsigned char)data[offset + i]);
L
Linus Torvalds 已提交
1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335
	}
	printk("\n");
}
#endif

#if DEBUG

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

	if (cachep->flags & SLAB_RED_ZONE) {
		printk(KERN_ERR "Redzone: 0x%lx/0x%lx.\n",
P
Pekka Enberg 已提交
1336 1337
		       *dbg_redzone1(cachep, objp),
		       *dbg_redzone2(cachep, objp));
L
Linus Torvalds 已提交
1338 1339 1340 1341
	}

	if (cachep->flags & SLAB_STORE_USER) {
		printk(KERN_ERR "Last user: [<%p>]",
P
Pekka Enberg 已提交
1342
		       *dbg_userword(cachep, objp));
L
Linus Torvalds 已提交
1343
		print_symbol("(%s)",
P
Pekka Enberg 已提交
1344
			     (unsigned long)*dbg_userword(cachep, objp));
L
Linus Torvalds 已提交
1345 1346
		printk("\n");
	}
P
Pekka Enberg 已提交
1347
	realobj = (char *)objp + obj_dbghead(cachep);
L
Linus Torvalds 已提交
1348
	size = obj_reallen(cachep);
P
Pekka Enberg 已提交
1349
	for (i = 0; i < size && lines; i += 16, lines--) {
L
Linus Torvalds 已提交
1350 1351
		int limit;
		limit = 16;
P
Pekka Enberg 已提交
1352 1353
		if (i + limit > size)
			limit = size - i;
L
Linus Torvalds 已提交
1354 1355 1356 1357 1358 1359 1360 1361 1362 1363
		dump_line(realobj, i, limit);
	}
}

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

P
Pekka Enberg 已提交
1364
	realobj = (char *)objp + obj_dbghead(cachep);
L
Linus Torvalds 已提交
1365 1366
	size = obj_reallen(cachep);

P
Pekka Enberg 已提交
1367
	for (i = 0; i < size; i++) {
L
Linus Torvalds 已提交
1368
		char exp = POISON_FREE;
P
Pekka Enberg 已提交
1369
		if (i == size - 1)
L
Linus Torvalds 已提交
1370 1371 1372 1373 1374 1375
			exp = POISON_END;
		if (realobj[i] != exp) {
			int limit;
			/* Mismatch ! */
			/* Print header */
			if (lines == 0) {
P
Pekka Enberg 已提交
1376 1377 1378
				printk(KERN_ERR
				       "Slab corruption: start=%p, len=%d\n",
				       realobj, size);
L
Linus Torvalds 已提交
1379 1380 1381
				print_objinfo(cachep, objp, 0);
			}
			/* Hexdump the affected line */
P
Pekka Enberg 已提交
1382
			i = (i / 16) * 16;
L
Linus Torvalds 已提交
1383
			limit = 16;
P
Pekka Enberg 已提交
1384 1385
			if (i + limit > size)
				limit = size - i;
L
Linus Torvalds 已提交
1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397
			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:
		 */
1398
		struct slab *slabp = page_get_slab(virt_to_page(objp));
L
Linus Torvalds 已提交
1399 1400
		int objnr;

1401
		objnr = (unsigned)(objp - slabp->s_mem) / cachep->objsize;
L
Linus Torvalds 已提交
1402
		if (objnr) {
P
Pekka Enberg 已提交
1403 1404
			objp = slabp->s_mem + (objnr - 1) * cachep->objsize;
			realobj = (char *)objp + obj_dbghead(cachep);
L
Linus Torvalds 已提交
1405
			printk(KERN_ERR "Prev obj: start=%p, len=%d\n",
P
Pekka Enberg 已提交
1406
			       realobj, size);
L
Linus Torvalds 已提交
1407 1408
			print_objinfo(cachep, objp, 2);
		}
P
Pekka Enberg 已提交
1409 1410 1411
		if (objnr + 1 < cachep->num) {
			objp = slabp->s_mem + (objnr + 1) * cachep->objsize;
			realobj = (char *)objp + obj_dbghead(cachep);
L
Linus Torvalds 已提交
1412
			printk(KERN_ERR "Next obj: start=%p, len=%d\n",
P
Pekka Enberg 已提交
1413
			       realobj, size);
L
Linus Torvalds 已提交
1414 1415 1416 1417 1418 1419 1420 1421 1422 1423
			print_objinfo(cachep, objp, 2);
		}
	}
}
#endif

/* Destroy all the objs in a slab, and release the mem back to the system.
 * Before calling the slab must have been unlinked from the cache.
 * The cache-lock is not held/needed.
 */
P
Pekka Enberg 已提交
1424
static void slab_destroy(kmem_cache_t *cachep, struct slab *slabp)
L
Linus Torvalds 已提交
1425 1426 1427 1428 1429 1430 1431 1432 1433 1434
{
	void *addr = slabp->s_mem - slabp->colouroff;

#if DEBUG
	int i;
	for (i = 0; i < cachep->num; i++) {
		void *objp = slabp->s_mem + cachep->objsize * i;

		if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
P
Pekka Enberg 已提交
1435 1436 1437 1438 1439
			if ((cachep->objsize % PAGE_SIZE) == 0
			    && OFF_SLAB(cachep))
				kernel_map_pages(virt_to_page(objp),
						 cachep->objsize / PAGE_SIZE,
						 1);
L
Linus Torvalds 已提交
1440 1441 1442 1443 1444 1445 1446 1447 1448
			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
1449
					   "was overwritten");
L
Linus Torvalds 已提交
1450 1451
			if (*dbg_redzone2(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "end of a freed object "
P
Pekka Enberg 已提交
1452
					   "was overwritten");
L
Linus Torvalds 已提交
1453 1454
		}
		if (cachep->dtor && !(cachep->flags & SLAB_POISON))
P
Pekka Enberg 已提交
1455
			(cachep->dtor) (objp + obj_dbghead(cachep), cachep, 0);
L
Linus Torvalds 已提交
1456 1457 1458 1459 1460
	}
#else
	if (cachep->dtor) {
		int i;
		for (i = 0; i < cachep->num; i++) {
P
Pekka Enberg 已提交
1461 1462
			void *objp = slabp->s_mem + cachep->objsize * i;
			(cachep->dtor) (objp, cachep, 0);
L
Linus Torvalds 已提交
1463 1464 1465 1466 1467 1468 1469
		}
	}
#endif

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

P
Pekka Enberg 已提交
1470
		slab_rcu = (struct slab_rcu *)slabp;
L
Linus Torvalds 已提交
1471 1472 1473 1474 1475 1476 1477 1478 1479 1480
		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);
	}
}

1481 1482 1483 1484 1485 1486 1487
/* For setting up all the kmem_list3s for cache whose objsize is same
   as size of kmem_list3. */
static inline void set_up_list3s(kmem_cache_t *cachep, int index)
{
	int node;

	for_each_online_node(node) {
P
Pekka Enberg 已提交
1488
		cachep->nodelists[node] = &initkmem_list3[index + node];
1489
		cachep->nodelists[node]->next_reap = jiffies +
P
Pekka Enberg 已提交
1490 1491
		    REAPTIMEOUT_LIST3 +
		    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
1492 1493 1494
	}
}

1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507
/**
 * calculate_slab_order - calculate size (page order) of slabs and the number
 *                        of objects per slab.
 *
 * This could be made much more intelligent.  For now, try to avoid using
 * high order pages for slabs.  When the gfp() functions are more friendly
 * towards high-order requests, this should be changed.
 */
static inline size_t calculate_slab_order(kmem_cache_t *cachep, size_t size,
					  size_t align, gfp_t flags)
{
	size_t left_over = 0;

P
Pekka Enberg 已提交
1508
	for (;; cachep->gfporder++) {
1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541
		unsigned int num;
		size_t remainder;

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

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

		cachep->num = num;
		left_over = remainder;

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

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

L
Linus Torvalds 已提交
1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581
/**
 * kmem_cache_create - Create a cache.
 * @name: A string which is used in /proc/slabinfo to identify this cache.
 * @size: The size of objects to be created in this cache.
 * @align: The required alignment for the objects.
 * @flags: SLAB flags
 * @ctor: A constructor for the objects.
 * @dtor: A destructor for the objects.
 *
 * Returns a ptr to the cache on success, NULL on failure.
 * Cannot be called within a int, but can be interrupted.
 * The @ctor is run when new pages are allocated by the cache
 * and the @dtor is run before the pages are handed back.
 *
 * @name must be valid until the cache is destroyed. This implies that
 * the module calling this has to destroy the cache before getting 
 * unloaded.
 * 
 * The flags are
 *
 * %SLAB_POISON - Poison the slab with a known test pattern (a5a5a5a5)
 * to catch references to uninitialised memory.
 *
 * %SLAB_RED_ZONE - Insert `Red' zones around the allocated memory to check
 * for buffer overruns.
 *
 * %SLAB_NO_REAP - Don't automatically reap this cache when we're under
 * memory pressure.
 *
 * %SLAB_HWCACHE_ALIGN - Align the objects in this cache to a hardware
 * cacheline.  This can be beneficial if you're counting cycles as closely
 * as davem.
 */
kmem_cache_t *
kmem_cache_create (const char *name, size_t size, size_t align,
	unsigned long flags, void (*ctor)(void*, kmem_cache_t *, unsigned long),
	void (*dtor)(void*, kmem_cache_t *, unsigned long))
{
	size_t left_over, slab_size, ralign;
	kmem_cache_t *cachep = NULL;
1582
	struct list_head *p;
L
Linus Torvalds 已提交
1583 1584 1585 1586 1587

	/*
	 * Sanity checks... these are all serious usage bugs.
	 */
	if ((!name) ||
P
Pekka Enberg 已提交
1588 1589 1590 1591 1592 1593 1594
	    in_interrupt() ||
	    (size < BYTES_PER_WORD) ||
	    (size > (1 << MAX_OBJ_ORDER) * PAGE_SIZE) || (dtor && !ctor)) {
		printk(KERN_ERR "%s: Early error in slab %s\n",
		       __FUNCTION__, name);
		BUG();
	}
L
Linus Torvalds 已提交
1595

I
Ingo Molnar 已提交
1596
	mutex_lock(&cache_chain_mutex);
1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613

	list_for_each(p, &cache_chain) {
		kmem_cache_t *pc = list_entry(p, kmem_cache_t, next);
		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",
P
Pekka Enberg 已提交
1614
			       pc->objsize);
1615 1616 1617
			continue;
		}

P
Pekka Enberg 已提交
1618
		if (!strcmp(pc->name, name)) {
1619 1620 1621 1622 1623 1624
			printk("kmem_cache_create: duplicate cache %s\n", name);
			dump_stack();
			goto oops;
		}
	}

L
Linus Torvalds 已提交
1625 1626 1627 1628 1629
#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
1630
		       "requested - %s\n", __FUNCTION__, name);
L
Linus Torvalds 已提交
1631 1632 1633 1634 1635 1636 1637 1638 1639
		flags &= ~SLAB_DEBUG_INITIAL;
	}
#if FORCED_DEBUG
	/*
	 * Enable redzoning and last user accounting, except for caches with
	 * large objects, if the increased size would increase the object size
	 * above the next power of two: caches with object sizes just above a
	 * power of two have a significant amount of internal fragmentation.
	 */
P
Pekka Enberg 已提交
1640 1641 1642
	if ((size < 4096
	     || fls(size - 1) == fls(size - 1 + 3 * BYTES_PER_WORD)))
		flags |= SLAB_RED_ZONE | SLAB_STORE_USER;
L
Linus Torvalds 已提交
1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662
	if (!(flags & SLAB_DESTROY_BY_RCU))
		flags |= SLAB_POISON;
#endif
	if (flags & SLAB_DESTROY_BY_RCU)
		BUG_ON(flags & SLAB_POISON);
#endif
	if (flags & SLAB_DESTROY_BY_RCU)
		BUG_ON(dtor);

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

	/* Check that size is in terms of words.  This is needed to avoid
	 * unaligned accesses for some archs when redzoning is used, and makes
	 * sure any on-slab bufctl's are also correctly aligned.
	 */
P
Pekka Enberg 已提交
1663 1664 1665
	if (size & (BYTES_PER_WORD - 1)) {
		size += (BYTES_PER_WORD - 1);
		size &= ~(BYTES_PER_WORD - 1);
L
Linus Torvalds 已提交
1666 1667 1668 1669 1670 1671 1672 1673 1674 1675
	}

	/* calculate out the final buffer alignment: */
	/* 1) arch recommendation: can be overridden for debug */
	if (flags & SLAB_HWCACHE_ALIGN) {
		/* Default alignment: as specified by the arch code.
		 * Except if an object is really small, then squeeze multiple
		 * objects into one cacheline.
		 */
		ralign = cache_line_size();
P
Pekka Enberg 已提交
1676
		while (size <= ralign / 2)
L
Linus Torvalds 已提交
1677 1678 1679 1680 1681 1682 1683 1684
			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
1685
			flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
L
Linus Torvalds 已提交
1686 1687 1688 1689 1690
	}
	/* 3) caller mandated alignment: disables debug if necessary */
	if (ralign < align) {
		ralign = align;
		if (ralign > BYTES_PER_WORD)
P
Pekka Enberg 已提交
1691
			flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
L
Linus Torvalds 已提交
1692 1693 1694 1695 1696 1697 1698 1699 1700
	}
	/* 4) Store it. Note that the debug code below can reduce
	 *    the alignment to BYTES_PER_WORD.
	 */
	align = ralign;

	/* Get cache's description obj. */
	cachep = (kmem_cache_t *) kmem_cache_alloc(&cache_cache, SLAB_KERNEL);
	if (!cachep)
1701
		goto oops;
L
Linus Torvalds 已提交
1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712
	memset(cachep, 0, sizeof(kmem_cache_t));

#if DEBUG
	cachep->reallen = size;

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

		/* add space for red zone words */
		cachep->dbghead += BYTES_PER_WORD;
P
Pekka Enberg 已提交
1713
		size += 2 * BYTES_PER_WORD;
L
Linus Torvalds 已提交
1714 1715 1716 1717 1718 1719 1720 1721 1722 1723
	}
	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
1724 1725
	if (size >= malloc_sizes[INDEX_L3 + 1].cs_size
	    && cachep->reallen > cache_line_size() && size < PAGE_SIZE) {
L
Linus Torvalds 已提交
1726 1727 1728 1729 1730 1731 1732
		cachep->dbghead += PAGE_SIZE - size;
		size = PAGE_SIZE;
	}
#endif
#endif

	/* Determine if the slab management is 'on' or 'off' slab. */
P
Pekka Enberg 已提交
1733
	if (size >= (PAGE_SIZE >> 3))
L
Linus Torvalds 已提交
1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749
		/*
		 * Size is large, assume best to place the slab management obj
		 * off-slab (should allow better packing of objs).
		 */
		flags |= CFLGS_OFF_SLAB;

	size = ALIGN(size, align);

	if ((flags & SLAB_RECLAIM_ACCOUNT) && size <= PAGE_SIZE) {
		/*
		 * A VFS-reclaimable slab tends to have most allocations
		 * as GFP_NOFS and we really don't want to have to be allocating
		 * higher-order pages when we are unable to shrink dcache.
		 */
		cachep->gfporder = 0;
		cache_estimate(cachep->gfporder, size, align, flags,
P
Pekka Enberg 已提交
1750
			       &left_over, &cachep->num);
1751 1752
	} else
		left_over = calculate_slab_order(cachep, size, align, flags);
L
Linus Torvalds 已提交
1753 1754 1755 1756 1757

	if (!cachep->num) {
		printk("kmem_cache_create: couldn't create cache %s.\n", name);
		kmem_cache_free(&cache_cache, cachep);
		cachep = NULL;
1758
		goto oops;
L
Linus Torvalds 已提交
1759
	}
P
Pekka Enberg 已提交
1760 1761
	slab_size = ALIGN(cachep->num * sizeof(kmem_bufctl_t)
			  + sizeof(struct slab), align);
L
Linus Torvalds 已提交
1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773

	/*
	 * 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
1774 1775
		slab_size =
		    cachep->num * sizeof(kmem_bufctl_t) + sizeof(struct slab);
L
Linus Torvalds 已提交
1776 1777 1778 1779 1780 1781
	}

	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
1782
	cachep->colour = left_over / cachep->colour_off;
L
Linus Torvalds 已提交
1783 1784 1785 1786 1787 1788 1789 1790 1791
	cachep->slab_size = slab_size;
	cachep->flags = flags;
	cachep->gfpflags = 0;
	if (flags & SLAB_CACHE_DMA)
		cachep->gfpflags |= GFP_DMA;
	spin_lock_init(&cachep->spinlock);
	cachep->objsize = size;

	if (flags & CFLGS_OFF_SLAB)
1792
		cachep->slabp_cache = kmem_find_general_cachep(slab_size, 0u);
L
Linus Torvalds 已提交
1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807
	cachep->ctor = ctor;
	cachep->dtor = dtor;
	cachep->name = name;

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

	if (g_cpucache_up == FULL) {
		enable_cpucache(cachep);
	} else {
		if (g_cpucache_up == NONE) {
			/* Note: the first kmem_cache_create must create
			 * the cache that's used by kmalloc(24), otherwise
			 * the creation of further caches will BUG().
			 */
1808
			cachep->array[smp_processor_id()] =
P
Pekka Enberg 已提交
1809
			    &initarray_generic.cache;
1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820

			/* If the cache that's used by
			 * kmalloc(sizeof(kmem_list3)) is the first cache,
			 * then we need to set up all its list3s, otherwise
			 * the creation of further caches will BUG().
			 */
			set_up_list3s(cachep, SIZE_AC);
			if (INDEX_AC == INDEX_L3)
				g_cpucache_up = PARTIAL_L3;
			else
				g_cpucache_up = PARTIAL_AC;
L
Linus Torvalds 已提交
1821
		} else {
1822
			cachep->array[smp_processor_id()] =
P
Pekka Enberg 已提交
1823
			    kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
1824 1825 1826 1827 1828 1829 1830 1831 1832

			if (g_cpucache_up == PARTIAL_AC) {
				set_up_list3s(cachep, SIZE_L3);
				g_cpucache_up = PARTIAL_L3;
			} else {
				int node;
				for_each_online_node(node) {

					cachep->nodelists[node] =
P
Pekka Enberg 已提交
1833 1834 1835
					    kmalloc_node(sizeof
							 (struct kmem_list3),
							 GFP_KERNEL, node);
1836
					BUG_ON(!cachep->nodelists[node]);
P
Pekka Enberg 已提交
1837 1838
					kmem_list3_init(cachep->
							nodelists[node]);
1839 1840
				}
			}
L
Linus Torvalds 已提交
1841
		}
1842
		cachep->nodelists[numa_node_id()]->next_reap =
P
Pekka Enberg 已提交
1843 1844
		    jiffies + REAPTIMEOUT_LIST3 +
		    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
1845

L
Linus Torvalds 已提交
1846 1847 1848 1849 1850 1851 1852
		BUG_ON(!ac_data(cachep));
		ac_data(cachep)->avail = 0;
		ac_data(cachep)->limit = BOOT_CPUCACHE_ENTRIES;
		ac_data(cachep)->batchcount = 1;
		ac_data(cachep)->touched = 0;
		cachep->batchcount = 1;
		cachep->limit = BOOT_CPUCACHE_ENTRIES;
P
Pekka Enberg 已提交
1853
	}
L
Linus Torvalds 已提交
1854 1855 1856 1857

	/* cache setup completed, link it into the list */
	list_add(&cachep->next, &cache_chain);
	unlock_cpu_hotplug();
P
Pekka Enberg 已提交
1858
      oops:
L
Linus Torvalds 已提交
1859 1860
	if (!cachep && (flags & SLAB_PANIC))
		panic("kmem_cache_create(): failed to create slab `%s'\n",
P
Pekka Enberg 已提交
1861
		      name);
I
Ingo Molnar 已提交
1862
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881
	return cachep;
}
EXPORT_SYMBOL(kmem_cache_create);

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

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

static void check_spinlock_acquired(kmem_cache_t *cachep)
{
#ifdef CONFIG_SMP
	check_irq_off();
1882
	assert_spin_locked(&cachep->nodelists[numa_node_id()]->list_lock);
L
Linus Torvalds 已提交
1883 1884
#endif
}
1885 1886 1887 1888 1889 1890 1891 1892 1893

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

L
Linus Torvalds 已提交
1894 1895 1896 1897
#else
#define check_irq_off()	do { } while(0)
#define check_irq_on()	do { } while(0)
#define check_spinlock_acquired(x) do { } while(0)
1898
#define check_spinlock_acquired_node(x, y) do { } while(0)
L
Linus Torvalds 已提交
1899 1900 1901 1902 1903
#endif

/*
 * Waits for all CPUs to execute func().
 */
P
Pekka Enberg 已提交
1904
static void smp_call_function_all_cpus(void (*func)(void *arg), void *arg)
L
Linus Torvalds 已提交
1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918
{
	check_irq_on();
	preempt_disable();

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

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

	preempt_enable();
}

P
Pekka Enberg 已提交
1919 1920
static void drain_array_locked(kmem_cache_t *cachep, struct array_cache *ac,
				int force, int node);
L
Linus Torvalds 已提交
1921 1922 1923

static void do_drain(void *arg)
{
P
Pekka Enberg 已提交
1924
	kmem_cache_t *cachep = (kmem_cache_t *) arg;
L
Linus Torvalds 已提交
1925
	struct array_cache *ac;
1926
	int node = numa_node_id();
L
Linus Torvalds 已提交
1927 1928 1929

	check_irq_off();
	ac = ac_data(cachep);
1930 1931 1932
	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
1933 1934 1935 1936 1937
	ac->avail = 0;
}

static void drain_cpu_caches(kmem_cache_t *cachep)
{
1938 1939 1940
	struct kmem_list3 *l3;
	int node;

L
Linus Torvalds 已提交
1941 1942 1943
	smp_call_function_all_cpus(do_drain, cachep);
	check_irq_on();
	spin_lock_irq(&cachep->spinlock);
P
Pekka Enberg 已提交
1944
	for_each_online_node(node) {
1945 1946 1947 1948 1949 1950 1951 1952 1953
		l3 = cachep->nodelists[node];
		if (l3) {
			spin_lock(&l3->list_lock);
			drain_array_locked(cachep, l3->shared, 1, node);
			spin_unlock(&l3->list_lock);
			if (l3->alien)
				drain_alien_cache(cachep, l3);
		}
	}
L
Linus Torvalds 已提交
1954 1955 1956
	spin_unlock_irq(&cachep->spinlock);
}

1957
static int __node_shrink(kmem_cache_t *cachep, int node)
L
Linus Torvalds 已提交
1958 1959
{
	struct slab *slabp;
1960
	struct kmem_list3 *l3 = cachep->nodelists[node];
L
Linus Torvalds 已提交
1961 1962
	int ret;

1963
	for (;;) {
L
Linus Torvalds 已提交
1964 1965
		struct list_head *p;

1966 1967
		p = l3->slabs_free.prev;
		if (p == &l3->slabs_free)
L
Linus Torvalds 已提交
1968 1969
			break;

1970
		slabp = list_entry(l3->slabs_free.prev, struct slab, list);
L
Linus Torvalds 已提交
1971 1972 1973 1974 1975 1976
#if DEBUG
		if (slabp->inuse)
			BUG();
#endif
		list_del(&slabp->list);

1977 1978
		l3->free_objects -= cachep->num;
		spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
1979
		slab_destroy(cachep, slabp);
1980
		spin_lock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
1981
	}
P
Pekka Enberg 已提交
1982
	ret = !list_empty(&l3->slabs_full) || !list_empty(&l3->slabs_partial);
L
Linus Torvalds 已提交
1983 1984 1985
	return ret;
}

1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
static int __cache_shrink(kmem_cache_t *cachep)
{
	int ret = 0, i = 0;
	struct kmem_list3 *l3;

	drain_cpu_caches(cachep);

	check_irq_on();
	for_each_online_node(i) {
		l3 = cachep->nodelists[i];
		if (l3) {
			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
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037
/**
 * kmem_cache_shrink - Shrink a cache.
 * @cachep: The cache to shrink.
 *
 * Releases as many slabs as possible for a cache.
 * To help debugging, a zero exit status indicates all slabs were released.
 */
int kmem_cache_shrink(kmem_cache_t *cachep)
{
	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
 *
 * Remove a kmem_cache_t object from the slab cache.
 * 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().
 */
P
Pekka Enberg 已提交
2038
int kmem_cache_destroy(kmem_cache_t *cachep)
L
Linus Torvalds 已提交
2039 2040
{
	int i;
2041
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
2042 2043 2044 2045 2046 2047 2048 2049

	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
2050
	mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2051 2052 2053 2054
	/*
	 * the chain is never empty, cache_cache is never destroyed
	 */
	list_del(&cachep->next);
I
Ingo Molnar 已提交
2055
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2056 2057 2058

	if (__cache_shrink(cachep)) {
		slab_error(cachep, "Can't free all objects");
I
Ingo Molnar 已提交
2059
		mutex_lock(&cache_chain_mutex);
P
Pekka Enberg 已提交
2060
		list_add(&cachep->next, &cache_chain);
I
Ingo Molnar 已提交
2061
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2062 2063 2064 2065 2066
		unlock_cpu_hotplug();
		return 1;
	}

	if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU))
2067
		synchronize_rcu();
L
Linus Torvalds 已提交
2068

2069
	for_each_online_cpu(i)
P
Pekka Enberg 已提交
2070
	    kfree(cachep->array[i]);
L
Linus Torvalds 已提交
2071 2072

	/* NUMA: free the list3 structures */
2073 2074 2075 2076 2077 2078 2079
	for_each_online_node(i) {
		if ((l3 = cachep->nodelists[i])) {
			kfree(l3->shared);
			free_alien_cache(l3->alien);
			kfree(l3);
		}
	}
L
Linus Torvalds 已提交
2080 2081 2082 2083 2084 2085 2086 2087 2088
	kmem_cache_free(&cache_cache, cachep);

	unlock_cpu_hotplug();

	return 0;
}
EXPORT_SYMBOL(kmem_cache_destroy);

/* Get the memory for a slab management obj. */
P
Pekka Enberg 已提交
2089 2090
static struct slab *alloc_slabmgmt(kmem_cache_t *cachep, void *objp,
				   int colour_off, gfp_t local_flags)
L
Linus Torvalds 已提交
2091 2092
{
	struct slab *slabp;
P
Pekka Enberg 已提交
2093

L
Linus Torvalds 已提交
2094 2095 2096 2097 2098 2099
	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
2100
		slabp = objp + colour_off;
L
Linus Torvalds 已提交
2101 2102 2103 2104
		colour_off += cachep->slab_size;
	}
	slabp->inuse = 0;
	slabp->colouroff = colour_off;
P
Pekka Enberg 已提交
2105
	slabp->s_mem = objp + colour_off;
L
Linus Torvalds 已提交
2106 2107 2108 2109 2110 2111

	return slabp;
}

static inline kmem_bufctl_t *slab_bufctl(struct slab *slabp)
{
P
Pekka Enberg 已提交
2112
	return (kmem_bufctl_t *) (slabp + 1);
L
Linus Torvalds 已提交
2113 2114 2115
}

static void cache_init_objs(kmem_cache_t *cachep,
P
Pekka Enberg 已提交
2116
			    struct slab *slabp, unsigned long ctor_flags)
L
Linus Torvalds 已提交
2117 2118 2119 2120
{
	int i;

	for (i = 0; i < cachep->num; i++) {
P
Pekka Enberg 已提交
2121
		void *objp = slabp->s_mem + cachep->objsize * i;
L
Linus Torvalds 已提交
2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138
#if DEBUG
		/* need to poison the objs? */
		if (cachep->flags & SLAB_POISON)
			poison_obj(cachep, objp, POISON_FREE);
		if (cachep->flags & SLAB_STORE_USER)
			*dbg_userword(cachep, objp) = NULL;

		if (cachep->flags & SLAB_RED_ZONE) {
			*dbg_redzone1(cachep, objp) = RED_INACTIVE;
			*dbg_redzone2(cachep, objp) = RED_INACTIVE;
		}
		/*
		 * Constructors are not allowed to allocate memory from
		 * the same cache which they are a constructor for.
		 * Otherwise, deadlock. They must also be threaded.
		 */
		if (cachep->ctor && !(cachep->flags & SLAB_POISON))
P
Pekka Enberg 已提交
2139 2140
			cachep->ctor(objp + obj_dbghead(cachep), cachep,
				     ctor_flags);
L
Linus Torvalds 已提交
2141 2142 2143 2144

		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
2145
					   " end of an object");
L
Linus Torvalds 已提交
2146 2147
			if (*dbg_redzone1(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "constructor overwrote the"
P
Pekka Enberg 已提交
2148
					   " start of an object");
L
Linus Torvalds 已提交
2149
		}
P
Pekka Enberg 已提交
2150 2151 2152 2153
		if ((cachep->objsize % PAGE_SIZE) == 0 && OFF_SLAB(cachep)
		    && cachep->flags & SLAB_POISON)
			kernel_map_pages(virt_to_page(objp),
					 cachep->objsize / PAGE_SIZE, 0);
L
Linus Torvalds 已提交
2154 2155 2156 2157
#else
		if (cachep->ctor)
			cachep->ctor(objp, cachep, ctor_flags);
#endif
P
Pekka Enberg 已提交
2158
		slab_bufctl(slabp)[i] = i + 1;
L
Linus Torvalds 已提交
2159
	}
P
Pekka Enberg 已提交
2160
	slab_bufctl(slabp)[i - 1] = BUFCTL_END;
L
Linus Torvalds 已提交
2161 2162 2163
	slabp->free = 0;
}

A
Al Viro 已提交
2164
static void kmem_flagcheck(kmem_cache_t *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183
{
	if (flags & SLAB_DMA) {
		if (!(cachep->gfpflags & GFP_DMA))
			BUG();
	} else {
		if (cachep->gfpflags & GFP_DMA)
			BUG();
	}
}

static void set_slab_attr(kmem_cache_t *cachep, struct slab *slabp, void *objp)
{
	int i;
	struct page *page;

	/* Nasty!!!!!! I hope this is OK. */
	i = 1 << cachep->gfporder;
	page = virt_to_page(objp);
	do {
2184 2185
		page_set_cache(page, cachep);
		page_set_slab(page, slabp);
L
Linus Torvalds 已提交
2186 2187 2188 2189 2190 2191 2192 2193
		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.
 */
A
Al Viro 已提交
2194
static int cache_grow(kmem_cache_t *cachep, gfp_t flags, int nodeid)
L
Linus Torvalds 已提交
2195
{
P
Pekka Enberg 已提交
2196 2197 2198 2199 2200
	struct slab *slabp;
	void *objp;
	size_t offset;
	gfp_t local_flags;
	unsigned long ctor_flags;
2201
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
2202 2203

	/* Be lazy and only check for valid flags here,
P
Pekka Enberg 已提交
2204
	 * keeping it out of the critical path in kmem_cache_alloc().
L
Linus Torvalds 已提交
2205
	 */
P
Pekka Enberg 已提交
2206
	if (flags & ~(SLAB_DMA | SLAB_LEVEL_MASK | SLAB_NO_GROW))
L
Linus Torvalds 已提交
2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232
		BUG();
	if (flags & SLAB_NO_GROW)
		return 0;

	ctor_flags = SLAB_CTOR_CONSTRUCTOR;
	local_flags = (flags & SLAB_LEVEL_MASK);
	if (!(local_flags & __GFP_WAIT))
		/*
		 * Not allowed to sleep.  Need to tell a constructor about
		 * this - it might need to know...
		 */
		ctor_flags |= SLAB_CTOR_ATOMIC;

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

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

	spin_unlock(&cachep->spinlock);

2233
	check_irq_off();
L
Linus Torvalds 已提交
2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244
	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);

2245 2246 2247
	/* Get mem for the objs.
	 * Attempt to allocate a physical page from 'nodeid',
	 */
L
Linus Torvalds 已提交
2248 2249 2250 2251 2252 2253 2254
	if (!(objp = kmem_getpages(cachep, flags, nodeid)))
		goto failed;

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

2255
	slabp->nodeid = nodeid;
L
Linus Torvalds 已提交
2256 2257 2258 2259 2260 2261 2262
	set_slab_attr(cachep, slabp, objp);

	cache_init_objs(cachep, slabp, ctor_flags);

	if (local_flags & __GFP_WAIT)
		local_irq_disable();
	check_irq_off();
2263 2264
	l3 = cachep->nodelists[nodeid];
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2265 2266

	/* Make slab active. */
2267
	list_add_tail(&slabp->list, &(l3->slabs_free));
L
Linus Torvalds 已提交
2268
	STATS_INC_GROWN(cachep);
2269 2270
	l3->free_objects += cachep->num;
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2271
	return 1;
P
Pekka Enberg 已提交
2272
      opps1:
L
Linus Torvalds 已提交
2273
	kmem_freepages(cachep, objp);
P
Pekka Enberg 已提交
2274
      failed:
L
Linus Torvalds 已提交
2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293
	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
2294 2295
		       (unsigned long)objp);
		BUG();
L
Linus Torvalds 已提交
2296 2297 2298
	}
	page = virt_to_page(objp);
	if (!PageSlab(page)) {
P
Pekka Enberg 已提交
2299 2300
		printk(KERN_ERR "kfree_debugcheck: bad ptr %lxh.\n",
		       (unsigned long)objp);
L
Linus Torvalds 已提交
2301 2302 2303 2304 2305
		BUG();
	}
}

static void *cache_free_debugcheck(kmem_cache_t *cachep, void *objp,
P
Pekka Enberg 已提交
2306
				   void *caller)
L
Linus Torvalds 已提交
2307 2308 2309 2310 2311 2312 2313 2314 2315
{
	struct page *page;
	unsigned int objnr;
	struct slab *slabp;

	objp -= obj_dbghead(cachep);
	kfree_debugcheck(objp);
	page = virt_to_page(objp);

2316
	if (page_get_cache(page) != cachep) {
P
Pekka Enberg 已提交
2317 2318 2319
		printk(KERN_ERR
		       "mismatch in kmem_cache_free: expected cache %p, got %p\n",
		       page_get_cache(page), cachep);
L
Linus Torvalds 已提交
2320
		printk(KERN_ERR "%p is %s.\n", cachep, cachep->name);
P
Pekka Enberg 已提交
2321 2322
		printk(KERN_ERR "%p is %s.\n", page_get_cache(page),
		       page_get_cache(page)->name);
L
Linus Torvalds 已提交
2323 2324
		WARN_ON(1);
	}
2325
	slabp = page_get_slab(page);
L
Linus Torvalds 已提交
2326 2327

	if (cachep->flags & SLAB_RED_ZONE) {
P
Pekka Enberg 已提交
2328 2329 2330 2331 2332 2333 2334 2335 2336
		if (*dbg_redzone1(cachep, objp) != RED_ACTIVE
		    || *dbg_redzone2(cachep, objp) != RED_ACTIVE) {
			slab_error(cachep,
				   "double free, or memory outside"
				   " object was overwritten");
			printk(KERN_ERR
			       "%p: redzone 1: 0x%lx, redzone 2: 0x%lx.\n",
			       objp, *dbg_redzone1(cachep, objp),
			       *dbg_redzone2(cachep, objp));
L
Linus Torvalds 已提交
2337 2338 2339 2340 2341 2342 2343
		}
		*dbg_redzone1(cachep, objp) = RED_INACTIVE;
		*dbg_redzone2(cachep, objp) = RED_INACTIVE;
	}
	if (cachep->flags & SLAB_STORE_USER)
		*dbg_userword(cachep, objp) = caller;

2344
	objnr = (unsigned)(objp - slabp->s_mem) / cachep->objsize;
L
Linus Torvalds 已提交
2345 2346

	BUG_ON(objnr >= cachep->num);
P
Pekka Enberg 已提交
2347
	BUG_ON(objp != slabp->s_mem + objnr * cachep->objsize);
L
Linus Torvalds 已提交
2348 2349 2350 2351 2352 2353

	if (cachep->flags & SLAB_DEBUG_INITIAL) {
		/* Need to call the slab's constructor so the
		 * caller can perform a verify of its state (debugging).
		 * Called without the cache-lock held.
		 */
P
Pekka Enberg 已提交
2354 2355
		cachep->ctor(objp + obj_dbghead(cachep),
			     cachep, SLAB_CTOR_CONSTRUCTOR | SLAB_CTOR_VERIFY);
L
Linus Torvalds 已提交
2356 2357 2358 2359 2360
	}
	if (cachep->flags & SLAB_POISON && cachep->dtor) {
		/* we want to cache poison the object,
		 * call the destruction callback
		 */
P
Pekka Enberg 已提交
2361
		cachep->dtor(objp + obj_dbghead(cachep), cachep, 0);
L
Linus Torvalds 已提交
2362 2363 2364 2365 2366
	}
	if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
		if ((cachep->objsize % PAGE_SIZE) == 0 && OFF_SLAB(cachep)) {
			store_stackinfo(cachep, objp, (unsigned long)caller);
P
Pekka Enberg 已提交
2367 2368
			kernel_map_pages(virt_to_page(objp),
					 cachep->objsize / PAGE_SIZE, 0);
L
Linus Torvalds 已提交
2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382
		} else {
			poison_obj(cachep, objp, POISON_FREE);
		}
#else
		poison_obj(cachep, objp, POISON_FREE);
#endif
	}
	return objp;
}

static void check_slabp(kmem_cache_t *cachep, struct slab *slabp)
{
	kmem_bufctl_t i;
	int entries = 0;
P
Pekka Enberg 已提交
2383

L
Linus Torvalds 已提交
2384 2385 2386 2387 2388 2389 2390
	/* Check slab's freelist to see if this obj is there. */
	for (i = slabp->free; i != BUFCTL_END; i = slab_bufctl(slabp)[i]) {
		entries++;
		if (entries > cachep->num || i >= cachep->num)
			goto bad;
	}
	if (entries != cachep->num - slabp->inuse) {
P
Pekka Enberg 已提交
2391 2392 2393 2394 2395 2396 2397 2398
	      bad:
		printk(KERN_ERR
		       "slab: Internal list corruption detected in cache '%s'(%d), slabp %p(%d). Hexdump:\n",
		       cachep->name, cachep->num, slabp, slabp->inuse);
		for (i = 0;
		     i < sizeof(slabp) + cachep->num * sizeof(kmem_bufctl_t);
		     i++) {
			if ((i % 16) == 0)
L
Linus Torvalds 已提交
2399
				printk("\n%03x:", i);
P
Pekka Enberg 已提交
2400
			printk(" %02x", ((unsigned char *)slabp)[i]);
L
Linus Torvalds 已提交
2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411
		}
		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

A
Al Viro 已提交
2412
static void *cache_alloc_refill(kmem_cache_t *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2413 2414 2415 2416 2417 2418 2419
{
	int batchcount;
	struct kmem_list3 *l3;
	struct array_cache *ac;

	check_irq_off();
	ac = ac_data(cachep);
P
Pekka Enberg 已提交
2420
      retry:
L
Linus Torvalds 已提交
2421 2422 2423 2424 2425 2426 2427 2428
	batchcount = ac->batchcount;
	if (!ac->touched && batchcount > BATCHREFILL_LIMIT) {
		/* if there was little recent activity on this
		 * cache, then perform only a partial refill.
		 * Otherwise we could generate refill bouncing.
		 */
		batchcount = BATCHREFILL_LIMIT;
	}
2429 2430 2431 2432
	l3 = cachep->nodelists[numa_node_id()];

	BUG_ON(ac->avail > 0 || !l3);
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2433 2434 2435 2436 2437 2438 2439 2440

	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;
2441
			memcpy(ac->entry,
P
Pekka Enberg 已提交
2442 2443
			       &(shared_array->entry[shared_array->avail]),
			       sizeof(void *) * batchcount);
L
Linus Torvalds 已提交
2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469
			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--) {
			kmem_bufctl_t next;
			STATS_INC_ALLOCED(cachep);
			STATS_INC_ACTIVE(cachep);
			STATS_SET_HIGH(cachep);

			/* get obj pointer */
2470
			ac->entry[ac->avail++] = slabp->s_mem +
P
Pekka Enberg 已提交
2471
			    slabp->free * cachep->objsize;
L
Linus Torvalds 已提交
2472 2473 2474 2475 2476

			slabp->inuse++;
			next = slab_bufctl(slabp)[slabp->free];
#if DEBUG
			slab_bufctl(slabp)[slabp->free] = BUFCTL_FREE;
2477
			WARN_ON(numa_node_id() != slabp->nodeid);
L
Linus Torvalds 已提交
2478
#endif
P
Pekka Enberg 已提交
2479
			slabp->free = next;
L
Linus Torvalds 已提交
2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490
		}
		check_slabp(cachep, slabp);

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

P
Pekka Enberg 已提交
2491
      must_grow:
L
Linus Torvalds 已提交
2492
	l3->free_objects -= ac->avail;
P
Pekka Enberg 已提交
2493
      alloc_done:
2494
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2495 2496 2497

	if (unlikely(!ac->avail)) {
		int x;
2498 2499
		x = cache_grow(cachep, flags, numa_node_id());

L
Linus Torvalds 已提交
2500 2501 2502 2503 2504
		// cache_grow can reenable interrupts, then ac could change.
		ac = ac_data(cachep);
		if (!x && ac->avail == 0)	// no objects in sight? abort
			return NULL;

P
Pekka Enberg 已提交
2505
		if (!ac->avail)	// objects refilled by interrupt?
L
Linus Torvalds 已提交
2506 2507 2508
			goto retry;
	}
	ac->touched = 1;
2509
	return ac->entry[--ac->avail];
L
Linus Torvalds 已提交
2510 2511 2512
}

static inline void
A
Al Viro 已提交
2513
cache_alloc_debugcheck_before(kmem_cache_t *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2514 2515 2516 2517 2518 2519 2520 2521
{
	might_sleep_if(flags & __GFP_WAIT);
#if DEBUG
	kmem_flagcheck(cachep, flags);
#endif
}

#if DEBUG
P
Pekka Enberg 已提交
2522 2523
static void *cache_alloc_debugcheck_after(kmem_cache_t *cachep, gfp_t flags,
					void *objp, void *caller)
L
Linus Torvalds 已提交
2524
{
P
Pekka Enberg 已提交
2525
	if (!objp)
L
Linus Torvalds 已提交
2526
		return objp;
P
Pekka Enberg 已提交
2527
	if (cachep->flags & SLAB_POISON) {
L
Linus Torvalds 已提交
2528 2529
#ifdef CONFIG_DEBUG_PAGEALLOC
		if ((cachep->objsize % PAGE_SIZE) == 0 && OFF_SLAB(cachep))
P
Pekka Enberg 已提交
2530 2531
			kernel_map_pages(virt_to_page(objp),
					 cachep->objsize / PAGE_SIZE, 1);
L
Linus Torvalds 已提交
2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542
		else
			check_poison_obj(cachep, objp);
#else
		check_poison_obj(cachep, objp);
#endif
		poison_obj(cachep, objp, POISON_INUSE);
	}
	if (cachep->flags & SLAB_STORE_USER)
		*dbg_userword(cachep, objp) = caller;

	if (cachep->flags & SLAB_RED_ZONE) {
P
Pekka Enberg 已提交
2543 2544 2545 2546 2547 2548 2549 2550 2551
		if (*dbg_redzone1(cachep, objp) != RED_INACTIVE
		    || *dbg_redzone2(cachep, objp) != RED_INACTIVE) {
			slab_error(cachep,
				   "double free, or memory outside"
				   " object was overwritten");
			printk(KERN_ERR
			       "%p: redzone 1: 0x%lx, redzone 2: 0x%lx.\n",
			       objp, *dbg_redzone1(cachep, objp),
			       *dbg_redzone2(cachep, objp));
L
Linus Torvalds 已提交
2552 2553 2554 2555 2556 2557
		}
		*dbg_redzone1(cachep, objp) = RED_ACTIVE;
		*dbg_redzone2(cachep, objp) = RED_ACTIVE;
	}
	objp += obj_dbghead(cachep);
	if (cachep->ctor && cachep->flags & SLAB_POISON) {
P
Pekka Enberg 已提交
2558
		unsigned long ctor_flags = SLAB_CTOR_CONSTRUCTOR;
L
Linus Torvalds 已提交
2559 2560 2561 2562 2563

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

		cachep->ctor(objp, cachep, ctor_flags);
P
Pekka Enberg 已提交
2564
	}
L
Linus Torvalds 已提交
2565 2566 2567 2568 2569 2570
	return objp;
}
#else
#define cache_alloc_debugcheck_after(a,b,objp,d) (objp)
#endif

A
Al Viro 已提交
2571
static inline void *____cache_alloc(kmem_cache_t *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2572
{
P
Pekka Enberg 已提交
2573
	void *objp;
L
Linus Torvalds 已提交
2574 2575
	struct array_cache *ac;

2576
#ifdef CONFIG_NUMA
2577
	if (unlikely(current->mempolicy && !in_interrupt())) {
2578 2579 2580 2581 2582 2583 2584
		int nid = slab_node(current->mempolicy);

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

2585
	check_irq_off();
L
Linus Torvalds 已提交
2586 2587 2588 2589
	ac = ac_data(cachep);
	if (likely(ac->avail)) {
		STATS_INC_ALLOCHIT(cachep);
		ac->touched = 1;
2590
		objp = ac->entry[--ac->avail];
L
Linus Torvalds 已提交
2591 2592 2593 2594
	} else {
		STATS_INC_ALLOCMISS(cachep);
		objp = cache_alloc_refill(cachep, flags);
	}
2595 2596 2597
	return objp;
}

A
Al Viro 已提交
2598
static inline void *__cache_alloc(kmem_cache_t *cachep, gfp_t flags)
2599 2600
{
	unsigned long save_flags;
P
Pekka Enberg 已提交
2601
	void *objp;
2602 2603 2604 2605 2606

	cache_alloc_debugcheck_before(cachep, flags);

	local_irq_save(save_flags);
	objp = ____cache_alloc(cachep, flags);
L
Linus Torvalds 已提交
2607
	local_irq_restore(save_flags);
2608
	objp = cache_alloc_debugcheck_after(cachep, flags, objp,
P
Pekka Enberg 已提交
2609
					    __builtin_return_address(0));
2610
	prefetchw(objp);
L
Linus Torvalds 已提交
2611 2612 2613
	return objp;
}

2614 2615 2616
#ifdef CONFIG_NUMA
/*
 * A interface to enable slab creation on nodeid
L
Linus Torvalds 已提交
2617
 */
A
Al Viro 已提交
2618
static void *__cache_alloc_node(kmem_cache_t *cachep, gfp_t flags, int nodeid)
2619 2620
{
	struct list_head *entry;
P
Pekka Enberg 已提交
2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653
	struct slab *slabp;
	struct kmem_list3 *l3;
	void *obj;
	kmem_bufctl_t next;
	int x;

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

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

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

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

	BUG_ON(slabp->inuse == cachep->num);

	/* get obj pointer */
	obj = slabp->s_mem + slabp->free * cachep->objsize;
	slabp->inuse++;
	next = slab_bufctl(slabp)[slabp->free];
2654
#if DEBUG
P
Pekka Enberg 已提交
2655
	slab_bufctl(slabp)[slabp->free] = BUFCTL_FREE;
2656
#endif
P
Pekka Enberg 已提交
2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667
	slabp->free = next;
	check_slabp(cachep, slabp);
	l3->free_objects--;
	/* move slabp to correct slabp list: */
	list_del(&slabp->list);

	if (slabp->free == BUFCTL_END) {
		list_add(&slabp->list, &l3->slabs_full);
	} else {
		list_add(&slabp->list, &l3->slabs_partial);
	}
2668

P
Pekka Enberg 已提交
2669 2670
	spin_unlock(&l3->list_lock);
	goto done;
2671

P
Pekka Enberg 已提交
2672 2673 2674
      must_grow:
	spin_unlock(&l3->list_lock);
	x = cache_grow(cachep, flags, nodeid);
L
Linus Torvalds 已提交
2675

P
Pekka Enberg 已提交
2676 2677
	if (!x)
		return NULL;
2678

P
Pekka Enberg 已提交
2679 2680 2681
	goto retry;
      done:
	return obj;
2682 2683 2684 2685 2686 2687
}
#endif

/*
 * Caller needs to acquire correct kmem_list's list_lock
 */
P
Pekka Enberg 已提交
2688 2689
static void free_block(kmem_cache_t *cachep, void **objpp, int nr_objects,
		       int node)
L
Linus Torvalds 已提交
2690 2691
{
	int i;
2692
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
2693 2694 2695 2696 2697 2698

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

2699
		slabp = page_get_slab(virt_to_page(objp));
2700
		l3 = cachep->nodelists[node];
L
Linus Torvalds 已提交
2701
		list_del(&slabp->list);
2702
		objnr = (unsigned)(objp - slabp->s_mem) / cachep->objsize;
2703
		check_spinlock_acquired_node(cachep, node);
L
Linus Torvalds 已提交
2704
		check_slabp(cachep, slabp);
2705

L
Linus Torvalds 已提交
2706
#if DEBUG
2707 2708 2709
		/* Verify that the slab belongs to the intended node */
		WARN_ON(slabp->nodeid != node);

L
Linus Torvalds 已提交
2710
		if (slab_bufctl(slabp)[objnr] != BUFCTL_FREE) {
2711
			printk(KERN_ERR "slab: double free detected in cache "
P
Pekka Enberg 已提交
2712
			       "'%s', objp %p\n", cachep->name, objp);
L
Linus Torvalds 已提交
2713 2714 2715 2716 2717 2718 2719
			BUG();
		}
#endif
		slab_bufctl(slabp)[objnr] = slabp->free;
		slabp->free = objnr;
		STATS_DEC_ACTIVE(cachep);
		slabp->inuse--;
2720
		l3->free_objects++;
L
Linus Torvalds 已提交
2721 2722 2723 2724
		check_slabp(cachep, slabp);

		/* fixup slab chains */
		if (slabp->inuse == 0) {
2725 2726
			if (l3->free_objects > l3->free_limit) {
				l3->free_objects -= cachep->num;
L
Linus Torvalds 已提交
2727 2728
				slab_destroy(cachep, slabp);
			} else {
2729
				list_add(&slabp->list, &l3->slabs_free);
L
Linus Torvalds 已提交
2730 2731 2732 2733 2734 2735
			}
		} else {
			/* Unconditionally move a slab to the end of the
			 * partial list on free - maximum time for the
			 * other objects to be freed, too.
			 */
2736
			list_add_tail(&slabp->list, &l3->slabs_partial);
L
Linus Torvalds 已提交
2737 2738 2739 2740 2741 2742 2743
		}
	}
}

static void cache_flusharray(kmem_cache_t *cachep, struct array_cache *ac)
{
	int batchcount;
2744
	struct kmem_list3 *l3;
2745
	int node = numa_node_id();
L
Linus Torvalds 已提交
2746 2747 2748 2749 2750 2751

	batchcount = ac->batchcount;
#if DEBUG
	BUG_ON(!batchcount || batchcount > ac->avail);
#endif
	check_irq_off();
2752
	l3 = cachep->nodelists[node];
2753 2754 2755
	spin_lock(&l3->list_lock);
	if (l3->shared) {
		struct array_cache *shared_array = l3->shared;
P
Pekka Enberg 已提交
2756
		int max = shared_array->limit - shared_array->avail;
L
Linus Torvalds 已提交
2757 2758 2759
		if (max) {
			if (batchcount > max)
				batchcount = max;
2760
			memcpy(&(shared_array->entry[shared_array->avail]),
P
Pekka Enberg 已提交
2761
			       ac->entry, sizeof(void *) * batchcount);
L
Linus Torvalds 已提交
2762 2763 2764 2765 2766
			shared_array->avail += batchcount;
			goto free_done;
		}
	}

2767
	free_block(cachep, ac->entry, batchcount, node);
P
Pekka Enberg 已提交
2768
      free_done:
L
Linus Torvalds 已提交
2769 2770 2771 2772 2773
#if STATS
	{
		int i = 0;
		struct list_head *p;

2774 2775
		p = l3->slabs_free.next;
		while (p != &(l3->slabs_free)) {
L
Linus Torvalds 已提交
2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786
			struct slab *slabp;

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

			i++;
			p = p->next;
		}
		STATS_SET_FREEABLE(cachep, i);
	}
#endif
2787
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2788
	ac->avail -= batchcount;
2789
	memmove(ac->entry, &(ac->entry[batchcount]),
P
Pekka Enberg 已提交
2790
		sizeof(void *) * ac->avail);
L
Linus Torvalds 已提交
2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806
}

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

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

2807 2808 2809 2810 2811 2812
	/* Make sure we are not freeing a object from another
	 * node to the array cache on this cpu.
	 */
#ifdef CONFIG_NUMA
	{
		struct slab *slabp;
2813
		slabp = page_get_slab(virt_to_page(objp));
2814 2815 2816
		if (unlikely(slabp->nodeid != numa_node_id())) {
			struct array_cache *alien = NULL;
			int nodeid = slabp->nodeid;
P
Pekka Enberg 已提交
2817 2818
			struct kmem_list3 *l3 =
			    cachep->nodelists[numa_node_id()];
2819 2820 2821 2822 2823 2824 2825

			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
2826
							    alien, nodeid);
2827 2828 2829 2830
				alien->entry[alien->avail++] = objp;
				spin_unlock(&alien->lock);
			} else {
				spin_lock(&(cachep->nodelists[nodeid])->
P
Pekka Enberg 已提交
2831
					  list_lock);
2832
				free_block(cachep, &objp, 1, nodeid);
2833
				spin_unlock(&(cachep->nodelists[nodeid])->
P
Pekka Enberg 已提交
2834
					    list_lock);
2835 2836 2837 2838 2839
			}
			return;
		}
	}
#endif
L
Linus Torvalds 已提交
2840 2841
	if (likely(ac->avail < ac->limit)) {
		STATS_INC_FREEHIT(cachep);
2842
		ac->entry[ac->avail++] = objp;
L
Linus Torvalds 已提交
2843 2844 2845 2846
		return;
	} else {
		STATS_INC_FREEMISS(cachep);
		cache_flusharray(cachep, ac);
2847
		ac->entry[ac->avail++] = objp;
L
Linus Torvalds 已提交
2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858
	}
}

/**
 * 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.
 */
A
Al Viro 已提交
2859
void *kmem_cache_alloc(kmem_cache_t *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880
{
	return __cache_alloc(cachep, flags);
}
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.
 */
int fastcall kmem_ptr_validate(kmem_cache_t *cachep, void *ptr)
{
P
Pekka Enberg 已提交
2881
	unsigned long addr = (unsigned long)ptr;
L
Linus Torvalds 已提交
2882
	unsigned long min_addr = PAGE_OFFSET;
P
Pekka Enberg 已提交
2883
	unsigned long align_mask = BYTES_PER_WORD - 1;
L
Linus Torvalds 已提交
2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899
	unsigned long size = cachep->objsize;
	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;
2900
	if (unlikely(page_get_cache(page) != cachep))
L
Linus Torvalds 已提交
2901 2902
		goto out;
	return 1;
P
Pekka Enberg 已提交
2903
      out:
L
Linus Torvalds 已提交
2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916
	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.
2917 2918
 * 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
2919
 */
A
Al Viro 已提交
2920
void *kmem_cache_alloc_node(kmem_cache_t *cachep, gfp_t flags, int nodeid)
L
Linus Torvalds 已提交
2921
{
2922 2923
	unsigned long save_flags;
	void *ptr;
L
Linus Torvalds 已提交
2924

2925
	if (nodeid == -1)
2926
		return __cache_alloc(cachep, flags);
L
Linus Torvalds 已提交
2927

2928 2929
	if (unlikely(!cachep->nodelists[nodeid])) {
		/* Fall back to __cache_alloc if we run into trouble */
P
Pekka Enberg 已提交
2930 2931 2932 2933
		printk(KERN_WARNING
		       "slab: not allocating in inactive node %d for cache %s\n",
		       nodeid, cachep->name);
		return __cache_alloc(cachep, flags);
L
Linus Torvalds 已提交
2934 2935
	}

2936 2937
	cache_alloc_debugcheck_before(cachep, flags);
	local_irq_save(save_flags);
2938 2939 2940 2941
	if (nodeid == numa_node_id())
		ptr = ____cache_alloc(cachep, flags);
	else
		ptr = __cache_alloc_node(cachep, flags, nodeid);
2942
	local_irq_restore(save_flags);
P
Pekka Enberg 已提交
2943 2944 2945
	ptr =
	    cache_alloc_debugcheck_after(cachep, flags, ptr,
					 __builtin_return_address(0));
L
Linus Torvalds 已提交
2946

2947
	return ptr;
L
Linus Torvalds 已提交
2948 2949 2950
}
EXPORT_SYMBOL(kmem_cache_alloc_node);

A
Al Viro 已提交
2951
void *kmalloc_node(size_t size, gfp_t flags, int node)
2952 2953 2954 2955 2956 2957 2958 2959 2960
{
	kmem_cache_t *cachep;

	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
2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983
#endif

/**
 * kmalloc - allocate memory
 * @size: how many bytes of memory are required.
 * @flags: the type of memory to allocate.
 *
 * kmalloc is the normal method of allocating memory
 * in the kernel.
 *
 * The @flags argument may be one of:
 *
 * %GFP_USER - Allocate memory on behalf of user.  May sleep.
 *
 * %GFP_KERNEL - Allocate normal kernel ram.  May sleep.
 *
 * %GFP_ATOMIC - Allocation will not sleep.  Use inside interrupt handlers.
 *
 * Additionally, the %GFP_DMA flag may be set to indicate the memory
 * must be suitable for DMA.  This can mean different things on different
 * platforms.  For example, on i386, it means that the memory must come
 * from the first 16MB.
 */
A
Al Viro 已提交
2984
void *__kmalloc(size_t size, gfp_t flags)
L
Linus Torvalds 已提交
2985 2986 2987
{
	kmem_cache_t *cachep;

2988 2989 2990 2991 2992 2993
	/* 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);
2994 2995
	if (unlikely(cachep == NULL))
		return NULL;
L
Linus Torvalds 已提交
2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007
	return __cache_alloc(cachep, flags);
}
EXPORT_SYMBOL(__kmalloc);

#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.
 */
3008
void *__alloc_percpu(size_t size)
L
Linus Torvalds 已提交
3009 3010
{
	int i;
P
Pekka Enberg 已提交
3011
	struct percpu_data *pdata = kmalloc(sizeof(*pdata), GFP_KERNEL);
L
Linus Torvalds 已提交
3012 3013 3014 3015

	if (!pdata)
		return NULL;

3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027
	/*
	 * 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
3028 3029 3030 3031 3032 3033 3034

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

	/* Catch derefs w/o wrappers */
P
Pekka Enberg 已提交
3035
	return (void *)(~(unsigned long)pdata);
L
Linus Torvalds 已提交
3036

P
Pekka Enberg 已提交
3037
      unwind_oom:
L
Linus Torvalds 已提交
3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070
	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.
 */
void kmem_cache_free(kmem_cache_t *cachep, void *objp)
{
	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.
 *
3071 3072
 * If @objp is NULL, no operation is performed.
 *
L
Linus Torvalds 已提交
3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084
 * Don't free memory not originally allocated by kmalloc()
 * or you will run into trouble.
 */
void kfree(const void *objp)
{
	kmem_cache_t *c;
	unsigned long flags;

	if (unlikely(!objp))
		return;
	local_irq_save(flags);
	kfree_debugcheck(objp);
3085
	c = page_get_cache(virt_to_page(objp));
3086
	mutex_debug_check_no_locks_freed(objp, obj_reallen(c));
P
Pekka Enberg 已提交
3087
	__cache_free(c, (void *)objp);
L
Linus Torvalds 已提交
3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099
	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
3100
void free_percpu(const void *objp)
L
Linus Torvalds 已提交
3101 3102
{
	int i;
P
Pekka Enberg 已提交
3103
	struct percpu_data *p = (struct percpu_data *)(~(unsigned long)objp);
L
Linus Torvalds 已提交
3104

3105 3106 3107 3108
	/*
	 * We allocate for all cpus so we cannot use for online cpu here.
	 */
	for_each_cpu(i)
P
Pekka Enberg 已提交
3109
	    kfree(p->ptrs[i]);
L
Linus Torvalds 已提交
3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120
	kfree(p);
}
EXPORT_SYMBOL(free_percpu);
#endif

unsigned int kmem_cache_size(kmem_cache_t *cachep)
{
	return obj_reallen(cachep);
}
EXPORT_SYMBOL(kmem_cache_size);

3121 3122 3123 3124 3125 3126
const char *kmem_cache_name(kmem_cache_t *cachep)
{
	return cachep->name;
}
EXPORT_SYMBOL_GPL(kmem_cache_name);

3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142
/*
 * This initializes kmem_list3 for all nodes.
 */
static int alloc_kmemlist(kmem_cache_t *cachep)
{
	int node;
	struct kmem_list3 *l3;
	int err = 0;

	for_each_online_node(node) {
		struct array_cache *nc = NULL, *new;
		struct array_cache **new_alien = NULL;
#ifdef CONFIG_NUMA
		if (!(new_alien = alloc_alien_cache(node, cachep->limit)))
			goto fail;
#endif
P
Pekka Enberg 已提交
3143 3144 3145
		if (!(new = alloc_arraycache(node, (cachep->shared *
						    cachep->batchcount),
					     0xbaadf00d)))
3146 3147 3148 3149 3150 3151
			goto fail;
		if ((l3 = cachep->nodelists[node])) {

			spin_lock_irq(&l3->list_lock);

			if ((nc = cachep->nodelists[node]->shared))
P
Pekka Enberg 已提交
3152
				free_block(cachep, nc->entry, nc->avail, node);
3153 3154 3155 3156 3157 3158

			l3->shared = new;
			if (!cachep->nodelists[node]->alien) {
				l3->alien = new_alien;
				new_alien = NULL;
			}
P
Pekka Enberg 已提交
3159 3160
			l3->free_limit = (1 + nr_cpus_node(node)) *
			    cachep->batchcount + cachep->num;
3161 3162 3163 3164 3165 3166
			spin_unlock_irq(&l3->list_lock);
			kfree(nc);
			free_alien_cache(new_alien);
			continue;
		}
		if (!(l3 = kmalloc_node(sizeof(struct kmem_list3),
P
Pekka Enberg 已提交
3167
					GFP_KERNEL, node)))
3168 3169 3170 3171
			goto fail;

		kmem_list3_init(l3);
		l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
P
Pekka Enberg 已提交
3172
		    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
3173 3174
		l3->shared = new;
		l3->alien = new_alien;
P
Pekka Enberg 已提交
3175 3176
		l3->free_limit = (1 + nr_cpus_node(node)) *
		    cachep->batchcount + cachep->num;
3177 3178 3179
		cachep->nodelists[node] = l3;
	}
	return err;
P
Pekka Enberg 已提交
3180
      fail:
3181 3182 3183 3184
	err = -ENOMEM;
	return err;
}

L
Linus Torvalds 已提交
3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196
struct ccupdate_struct {
	kmem_cache_t *cachep;
	struct array_cache *new[NR_CPUS];
};

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

	check_irq_off();
	old = ac_data(new->cachep);
3197

L
Linus Torvalds 已提交
3198 3199 3200 3201 3202
	new->cachep->array[smp_processor_id()] = new->new[smp_processor_id()];
	new->new[smp_processor_id()] = old;
}

static int do_tune_cpucache(kmem_cache_t *cachep, int limit, int batchcount,
P
Pekka Enberg 已提交
3203
			    int shared)
L
Linus Torvalds 已提交
3204 3205
{
	struct ccupdate_struct new;
3206
	int i, err;
L
Linus Torvalds 已提交
3207

P
Pekka Enberg 已提交
3208
	memset(&new.new, 0, sizeof(new.new));
3209
	for_each_online_cpu(i) {
P
Pekka Enberg 已提交
3210 3211
		new.new[i] =
		    alloc_arraycache(cpu_to_node(i), limit, batchcount);
3212
		if (!new.new[i]) {
P
Pekka Enberg 已提交
3213 3214
			for (i--; i >= 0; i--)
				kfree(new.new[i]);
3215
			return -ENOMEM;
L
Linus Torvalds 已提交
3216 3217 3218 3219 3220
		}
	}
	new.cachep = cachep;

	smp_call_function_all_cpus(do_ccupdate_local, (void *)&new);
3221

L
Linus Torvalds 已提交
3222 3223 3224 3225
	check_irq_on();
	spin_lock_irq(&cachep->spinlock);
	cachep->batchcount = batchcount;
	cachep->limit = limit;
3226
	cachep->shared = shared;
L
Linus Torvalds 已提交
3227 3228
	spin_unlock_irq(&cachep->spinlock);

3229
	for_each_online_cpu(i) {
L
Linus Torvalds 已提交
3230 3231 3232
		struct array_cache *ccold = new.new[i];
		if (!ccold)
			continue;
3233
		spin_lock_irq(&cachep->nodelists[cpu_to_node(i)]->list_lock);
3234
		free_block(cachep, ccold->entry, ccold->avail, cpu_to_node(i));
3235
		spin_unlock_irq(&cachep->nodelists[cpu_to_node(i)]->list_lock);
L
Linus Torvalds 已提交
3236 3237 3238
		kfree(ccold);
	}

3239 3240 3241
	err = alloc_kmemlist(cachep);
	if (err) {
		printk(KERN_ERR "alloc_kmemlist failed for %s, error %d.\n",
P
Pekka Enberg 已提交
3242
		       cachep->name, -err);
3243
		BUG();
L
Linus Torvalds 已提交
3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293
	}
	return 0;
}

static void enable_cpucache(kmem_cache_t *cachep)
{
	int err;
	int limit, shared;

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

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

#if DEBUG
	/* With debugging enabled, large batchcount lead to excessively
	 * long periods with disabled local interrupts. Limit the 
	 * batchcount
	 */
	if (limit > 32)
		limit = 32;
#endif
P
Pekka Enberg 已提交
3294
	err = do_tune_cpucache(cachep, limit, (limit + 1) / 2, shared);
L
Linus Torvalds 已提交
3295 3296
	if (err)
		printk(KERN_ERR "enable_cpucache failed for %s, error %d.\n",
P
Pekka Enberg 已提交
3297
		       cachep->name, -err);
L
Linus Torvalds 已提交
3298 3299
}

P
Pekka Enberg 已提交
3300 3301
static void drain_array_locked(kmem_cache_t *cachep, struct array_cache *ac,
				int force, int node)
L
Linus Torvalds 已提交
3302 3303 3304
{
	int tofree;

3305
	check_spinlock_acquired_node(cachep, node);
L
Linus Torvalds 已提交
3306 3307 3308
	if (ac->touched && !force) {
		ac->touched = 0;
	} else if (ac->avail) {
P
Pekka Enberg 已提交
3309
		tofree = force ? ac->avail : (ac->limit + 4) / 5;
L
Linus Torvalds 已提交
3310
		if (tofree > ac->avail) {
P
Pekka Enberg 已提交
3311
			tofree = (ac->avail + 1) / 2;
L
Linus Torvalds 已提交
3312
		}
3313
		free_block(cachep, ac->entry, tofree, node);
L
Linus Torvalds 已提交
3314
		ac->avail -= tofree;
3315
		memmove(ac->entry, &(ac->entry[tofree]),
P
Pekka Enberg 已提交
3316
			sizeof(void *) * ac->avail);
L
Linus Torvalds 已提交
3317 3318 3319 3320 3321
	}
}

/**
 * cache_reap - Reclaim memory from caches.
3322
 * @unused: unused parameter
L
Linus Torvalds 已提交
3323 3324 3325 3326 3327 3328
 *
 * Called from workqueue/eventd every few seconds.
 * Purpose:
 * - clear the per-cpu caches for this CPU.
 * - return freeable pages to the main free memory pool.
 *
I
Ingo Molnar 已提交
3329
 * If we cannot acquire the cache chain mutex then just give up - we'll
L
Linus Torvalds 已提交
3330 3331 3332 3333 3334
 * try again on the next iteration.
 */
static void cache_reap(void *unused)
{
	struct list_head *walk;
3335
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
3336

I
Ingo Molnar 已提交
3337
	if (!mutex_trylock(&cache_chain_mutex)) {
L
Linus Torvalds 已提交
3338
		/* Give up. Setup the next iteration. */
P
Pekka Enberg 已提交
3339 3340
		schedule_delayed_work(&__get_cpu_var(reap_work),
				      REAPTIMEOUT_CPUC);
L
Linus Torvalds 已提交
3341 3342 3343 3344 3345
		return;
	}

	list_for_each(walk, &cache_chain) {
		kmem_cache_t *searchp;
P
Pekka Enberg 已提交
3346
		struct list_head *p;
L
Linus Torvalds 已提交
3347 3348 3349 3350 3351 3352 3353 3354 3355 3356
		int tofree;
		struct slab *slabp;

		searchp = list_entry(walk, kmem_cache_t, next);

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

		check_irq_on();

3357 3358 3359 3360
		l3 = searchp->nodelists[numa_node_id()];
		if (l3->alien)
			drain_alien_cache(searchp, l3);
		spin_lock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
3361

3362
		drain_array_locked(searchp, ac_data(searchp), 0,
P
Pekka Enberg 已提交
3363
				   numa_node_id());
L
Linus Torvalds 已提交
3364

3365
		if (time_after(l3->next_reap, jiffies))
L
Linus Torvalds 已提交
3366 3367
			goto next_unlock;

3368
		l3->next_reap = jiffies + REAPTIMEOUT_LIST3;
L
Linus Torvalds 已提交
3369

3370 3371
		if (l3->shared)
			drain_array_locked(searchp, l3->shared, 0,
P
Pekka Enberg 已提交
3372
					   numa_node_id());
L
Linus Torvalds 已提交
3373

3374 3375
		if (l3->free_touched) {
			l3->free_touched = 0;
L
Linus Torvalds 已提交
3376 3377 3378
			goto next_unlock;
		}

P
Pekka Enberg 已提交
3379 3380 3381
		tofree =
		    (l3->free_limit + 5 * searchp->num -
		     1) / (5 * searchp->num);
L
Linus Torvalds 已提交
3382
		do {
3383 3384
			p = l3->slabs_free.next;
			if (p == &(l3->slabs_free))
L
Linus Torvalds 已提交
3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396
				break;

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

			/* Safe to drop the lock. The slab is no longer
			 * linked to the cache.
			 * searchp cannot disappear, we hold
			 * cache_chain_lock
			 */
3397 3398
			l3->free_objects -= searchp->num;
			spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
3399
			slab_destroy(searchp, slabp);
3400
			spin_lock_irq(&l3->list_lock);
P
Pekka Enberg 已提交
3401 3402
		} while (--tofree > 0);
	      next_unlock:
3403
		spin_unlock_irq(&l3->list_lock);
P
Pekka Enberg 已提交
3404
	      next:
L
Linus Torvalds 已提交
3405 3406 3407
		cond_resched();
	}
	check_irq_on();
I
Ingo Molnar 已提交
3408
	mutex_unlock(&cache_chain_mutex);
3409
	drain_remote_pages();
L
Linus Torvalds 已提交
3410
	/* Setup the next iteration */
3411
	schedule_delayed_work(&__get_cpu_var(reap_work), REAPTIMEOUT_CPUC);
L
Linus Torvalds 已提交
3412 3413 3414 3415
}

#ifdef CONFIG_PROC_FS

3416
static void print_slabinfo_header(struct seq_file *m)
L
Linus Torvalds 已提交
3417
{
3418 3419 3420 3421
	/*
	 * Output format version, so at least we can change it
	 * without _too_ many complaints.
	 */
L
Linus Torvalds 已提交
3422
#if STATS
3423
	seq_puts(m, "slabinfo - version: 2.1 (statistics)\n");
L
Linus Torvalds 已提交
3424
#else
3425
	seq_puts(m, "slabinfo - version: 2.1\n");
L
Linus Torvalds 已提交
3426
#endif
3427 3428 3429 3430
	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
3431
#if STATS
3432 3433 3434
	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
3435
#endif
3436 3437 3438 3439 3440 3441 3442 3443
	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
3444
	mutex_lock(&cache_chain_mutex);
3445 3446
	if (!n)
		print_slabinfo_header(m);
L
Linus Torvalds 已提交
3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460
	p = cache_chain.next;
	while (n--) {
		p = p->next;
		if (p == &cache_chain)
			return NULL;
	}
	return list_entry(p, kmem_cache_t, next);
}

static void *s_next(struct seq_file *m, void *p, loff_t *pos)
{
	kmem_cache_t *cachep = p;
	++*pos;
	return cachep->next.next == &cache_chain ? NULL
P
Pekka Enberg 已提交
3461
	    : list_entry(cachep->next.next, kmem_cache_t, next);
L
Linus Torvalds 已提交
3462 3463 3464 3465
}

static void s_stop(struct seq_file *m, void *p)
{
I
Ingo Molnar 已提交
3466
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
3467 3468 3469 3470 3471 3472
}

static int s_show(struct seq_file *m, void *p)
{
	kmem_cache_t *cachep = p;
	struct list_head *q;
P
Pekka Enberg 已提交
3473 3474 3475 3476 3477
	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;
3478
	const char *name;
L
Linus Torvalds 已提交
3479
	char *error = NULL;
3480 3481
	int node;
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
3482 3483 3484 3485 3486

	check_irq_on();
	spin_lock_irq(&cachep->spinlock);
	active_objs = 0;
	num_slabs = 0;
3487 3488 3489 3490 3491 3492 3493
	for_each_online_node(node) {
		l3 = cachep->nodelists[node];
		if (!l3)
			continue;

		spin_lock(&l3->list_lock);

P
Pekka Enberg 已提交
3494
		list_for_each(q, &l3->slabs_full) {
3495 3496 3497 3498 3499 3500
			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
3501
		list_for_each(q, &l3->slabs_partial) {
3502 3503 3504 3505 3506 3507 3508 3509
			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
3510
		list_for_each(q, &l3->slabs_free) {
3511 3512 3513 3514 3515 3516 3517 3518 3519
			slabp = list_entry(q, struct slab, list);
			if (slabp->inuse && !error)
				error = "slabs_free/inuse accounting error";
			num_slabs++;
		}
		free_objects += l3->free_objects;
		shared_avail += l3->shared->avail;

		spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
3520
	}
P
Pekka Enberg 已提交
3521 3522
	num_slabs += active_slabs;
	num_objs = num_slabs * cachep->num;
3523
	if (num_objs - active_objs != free_objects && !error)
L
Linus Torvalds 已提交
3524 3525
		error = "free_objects accounting error";

P
Pekka Enberg 已提交
3526
	name = cachep->name;
L
Linus Torvalds 已提交
3527 3528 3529 3530
	if (error)
		printk(KERN_ERR "slab: cache %s error: %s\n", name, error);

	seq_printf(m, "%-17s %6lu %6lu %6u %4u %4d",
P
Pekka Enberg 已提交
3531 3532
		   name, active_objs, num_objs, cachep->objsize,
		   cachep->num, (1 << cachep->gfporder));
L
Linus Torvalds 已提交
3533
	seq_printf(m, " : tunables %4u %4u %4u",
P
Pekka Enberg 已提交
3534
		   cachep->limit, cachep->batchcount, cachep->shared);
3535
	seq_printf(m, " : slabdata %6lu %6lu %6lu",
P
Pekka Enberg 已提交
3536
		   active_slabs, num_slabs, shared_avail);
L
Linus Torvalds 已提交
3537
#if STATS
P
Pekka Enberg 已提交
3538
	{			/* list3 stats */
L
Linus Torvalds 已提交
3539 3540 3541 3542 3543 3544 3545
		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;
3546
		unsigned long node_frees = cachep->node_frees;
L
Linus Torvalds 已提交
3547

3548
		seq_printf(m, " : globalstat %7lu %6lu %5lu %4lu \
P
Pekka Enberg 已提交
3549
				%4lu %4lu %4lu %4lu", allocs, high, grown, reaped, errors, max_freeable, node_allocs, node_frees);
L
Linus Torvalds 已提交
3550 3551 3552 3553 3554 3555 3556 3557 3558
	}
	/* 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
3559
			   allochit, allocmiss, freehit, freemiss);
L
Linus Torvalds 已提交
3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581
	}
#endif
	seq_putc(m, '\n');
	spin_unlock_irq(&cachep->spinlock);
	return 0;
}

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

struct seq_operations slabinfo_op = {
P
Pekka Enberg 已提交
3582 3583 3584 3585
	.start = s_start,
	.next = s_next,
	.stop = s_stop,
	.show = s_show,
L
Linus Torvalds 已提交
3586 3587 3588 3589 3590 3591 3592 3593 3594 3595
};

#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
3596 3597
ssize_t slabinfo_write(struct file *file, const char __user * buffer,
		       size_t count, loff_t *ppos)
L
Linus Torvalds 已提交
3598
{
P
Pekka Enberg 已提交
3599
	char kbuf[MAX_SLABINFO_WRITE + 1], *tmp;
L
Linus Torvalds 已提交
3600 3601
	int limit, batchcount, shared, res;
	struct list_head *p;
P
Pekka Enberg 已提交
3602

L
Linus Torvalds 已提交
3603 3604 3605 3606
	if (count > MAX_SLABINFO_WRITE)
		return -EINVAL;
	if (copy_from_user(&kbuf, buffer, count))
		return -EFAULT;
P
Pekka Enberg 已提交
3607
	kbuf[MAX_SLABINFO_WRITE] = '\0';
L
Linus Torvalds 已提交
3608 3609 3610 3611 3612 3613 3614 3615 3616 3617

	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
3618
	mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
3619
	res = -EINVAL;
P
Pekka Enberg 已提交
3620
	list_for_each(p, &cache_chain) {
L
Linus Torvalds 已提交
3621 3622 3623 3624 3625
		kmem_cache_t *cachep = list_entry(p, kmem_cache_t, next);

		if (!strcmp(cachep->name, kbuf)) {
			if (limit < 1 ||
			    batchcount < 1 ||
P
Pekka Enberg 已提交
3626
			    batchcount > limit || shared < 0) {
3627
				res = 0;
L
Linus Torvalds 已提交
3628
			} else {
3629
				res = do_tune_cpucache(cachep, limit,
P
Pekka Enberg 已提交
3630
						       batchcount, shared);
L
Linus Torvalds 已提交
3631 3632 3633 3634
			}
			break;
		}
	}
I
Ingo Molnar 已提交
3635
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
3636 3637 3638 3639 3640 3641
	if (res >= 0)
		res = count;
	return res;
}
#endif

3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653
/**
 * 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
3654 3655
unsigned int ksize(const void *objp)
{
3656 3657
	if (unlikely(objp == NULL))
		return 0;
L
Linus Torvalds 已提交
3658

3659
	return obj_reallen(page_get_cache(virt_to_page(objp)));
L
Linus Torvalds 已提交
3660
}