slab.c 116.9 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52
/*
 * linux/mm/slab.c
 * Written by Mark Hemment, 1996/97.
 * (markhe@nextd.demon.co.uk)
 *
 * kmem_cache_destroy() + some cleanup - 1999 Andrea Arcangeli
 *
 * Major cleanup, different bufctl logic, per-cpu arrays
 *	(c) 2000 Manfred Spraul
 *
 * Cleanup, make the head arrays unconditional, preparation for NUMA
 * 	(c) 2002 Manfred Spraul
 *
 * An implementation of the Slab Allocator as described in outline in;
 *	UNIX Internals: The New Frontiers by Uresh Vahalia
 *	Pub: Prentice Hall	ISBN 0-13-101908-2
 * or with a little more detail in;
 *	The Slab Allocator: An Object-Caching Kernel Memory Allocator
 *	Jeff Bonwick (Sun Microsystems).
 *	Presented at: USENIX Summer 1994 Technical Conference
 *
 * The memory is organized in caches, one cache for each object type.
 * (e.g. inode_cache, dentry_cache, buffer_head, vm_area_struct)
 * Each cache consists out of many slabs (they are small (usually one
 * page long) and always contiguous), and each slab contains multiple
 * initialized objects.
 *
 * This means, that your constructor is used only for newly allocated
 * slabs and you must pass objects with the same intializations to
 * kmem_cache_free.
 *
 * Each cache can only support one memory type (GFP_DMA, GFP_HIGHMEM,
 * normal). If you need a special memory type, then must create a new
 * cache for that memory type.
 *
 * In order to reduce fragmentation, the slabs are sorted in 3 groups:
 *   full slabs with 0 free objects
 *   partial slabs
 *   empty slabs with no allocated objects
 *
 * If partial slabs exist, then new allocations come from these slabs,
 * otherwise from empty slabs or new slabs are allocated.
 *
 * kmem_cache_destroy() CAN CRASH if you try to allocate from the cache
 * during kmem_cache_destroy(). The caller must prevent concurrent allocs.
 *
 * Each cache has a short per-cpu head array, most allocs
 * and frees go into that array, and if that array overflows, then 1/2
 * of the entries in the array are given back into the global cache.
 * The head array is strictly LIFO and should improve the cache hit rates.
 * On SMP, it additionally reduces the spinlock operations.
 *
A
Andrew Morton 已提交
53
 * The c_cpuarray may not be read with enabled local interrupts -
L
Linus Torvalds 已提交
54 55 56 57
 * it's changed with a smp_call_function().
 *
 * SMP synchronization:
 *  constructors and destructors are called without any locking.
58
 *  Several members in struct kmem_cache and struct slab never change, they
L
Linus Torvalds 已提交
59 60 61 62 63 64 65 66 67 68 69 70
 *	are accessed without any locking.
 *  The per-cpu arrays are never accessed from the wrong cpu, no locking,
 *  	and local interrupts are disabled so slab code is preempt-safe.
 *  The non-constant members are protected with a per-cache irq spinlock.
 *
 * Many thanks to Mark Hemment, who wrote another per-cpu slab patch
 * in 2000 - many ideas in the current implementation are derived from
 * his patch.
 *
 * Further notes from the original documentation:
 *
 * 11 April '97.  Started multi-threading - markhe
I
Ingo Molnar 已提交
71
 *	The global cache-chain is protected by the mutex 'cache_chain_mutex'.
L
Linus Torvalds 已提交
72 73 74 75 76 77
 *	The sem is only needed when accessing/extending the cache-chain, which
 *	can never happen inside an interrupt (kmem_cache_create(),
 *	kmem_cache_shrink() and kmem_cache_reap()).
 *
 *	At present, each engine can be growing a cache.  This should be blocked.
 *
78 79 80 81 82 83 84 85 86
 * 15 March 2005. NUMA slab allocator.
 *	Shai Fultheim <shai@scalex86.org>.
 *	Shobhit Dayal <shobhit@calsoftinc.com>
 *	Alok N Kataria <alokk@calsoftinc.com>
 *	Christoph Lameter <christoph@lameter.com>
 *
 *	Modified the slab allocator to be node aware on NUMA systems.
 *	Each node has its own list of partial, free and full slabs.
 *	All object allocations for a node occur from node specific slab lists.
L
Linus Torvalds 已提交
87 88 89 90
 */

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

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

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

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

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

#ifndef cache_line_size
#define cache_line_size()	L1_CACHE_BYTES
#endif

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

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

#ifndef ARCH_KMALLOC_FLAGS
#define ARCH_KMALLOC_FLAGS SLAB_HWCACHE_ALIGN
#endif

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

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

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

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

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

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

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

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

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

313 314 315 316
static int drain_freelist(struct kmem_cache *cache,
			struct kmem_list3 *l3, int tofree);
static void free_block(struct kmem_cache *cachep, void **objpp, int len,
			int node);
317
static int enable_cpucache(struct kmem_cache *cachep);
318
static void cache_reap(struct work_struct *unused);
319

320
/*
A
Andrew Morton 已提交
321 322
 * This function must be completely optimized away if a constant is passed to
 * it.  Mostly the same as what is in linux/slab.h except it returns an index.
323
 */
324
static __always_inline int index_of(const size_t size)
325
{
326 327
	extern void __bad_size(void);

328 329 330 331 332 333 334 335 336 337
	if (__builtin_constant_p(size)) {
		int i = 0;

#define CACHE(x) \
	if (size <=x) \
		return i; \
	else \
		i++;
#include "linux/kmalloc_sizes.h"
#undef CACHE
338
		__bad_size();
339
	} else
340
		__bad_size();
341 342 343
	return 0;
}

344 345
static int slab_early_init = 1;

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

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

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

A
Andrew Morton 已提交
368 369
#define	MAKE_ALL_LISTS(cachep, ptr, nodeid)				\
	do {								\
370 371 372 373
	MAKE_LIST((cachep), (&(ptr)->slabs_full), slabs_full, nodeid);	\
	MAKE_LIST((cachep), (&(ptr)->slabs_partial), slabs_partial, nodeid); \
	MAKE_LIST((cachep), (&(ptr)->slabs_free), slabs_free, nodeid);	\
	} while (0)
L
Linus Torvalds 已提交
374 375

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

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

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

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

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

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

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

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

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

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

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

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

#define BATCHREFILL_LIMIT	16
A
Andrew Morton 已提交
453 454 455
/*
 * Optimization question: fewer reaps means less probability for unnessary
 * cpucache drain/refill cycles.
L
Linus Torvalds 已提交
456
 *
A
Adrian Bunk 已提交
457
 * OTOH the cpuarrays can contain lots of objects,
L
Linus Torvalds 已提交
458 459 460 461 462 463 464 465 466 467
 * which could lock up otherwise freeable slabs.
 */
#define REAPTIMEOUT_CPUC	(2*HZ)
#define REAPTIMEOUT_LIST3	(4*HZ)

#if STATS
#define	STATS_INC_ACTIVE(x)	((x)->num_active++)
#define	STATS_DEC_ACTIVE(x)	((x)->num_active--)
#define	STATS_INC_ALLOCED(x)	((x)->num_allocations++)
#define	STATS_INC_GROWN(x)	((x)->grown++)
468
#define	STATS_ADD_REAPED(x,y)	((x)->reaped += (y))
A
Andrew Morton 已提交
469 470 471 472 473
#define	STATS_SET_HIGH(x)						\
	do {								\
		if ((x)->num_active > (x)->high_mark)			\
			(x)->high_mark = (x)->num_active;		\
	} while (0)
L
Linus Torvalds 已提交
474 475
#define	STATS_INC_ERR(x)	((x)->errors++)
#define	STATS_INC_NODEALLOCS(x)	((x)->node_allocs++)
476
#define	STATS_INC_NODEFREES(x)	((x)->node_frees++)
477
#define STATS_INC_ACOVERFLOW(x)   ((x)->node_overflow++)
A
Andrew Morton 已提交
478 479 480 481 482
#define	STATS_SET_FREEABLE(x, i)					\
	do {								\
		if ((x)->max_freeable < i)				\
			(x)->max_freeable = i;				\
	} while (0)
L
Linus Torvalds 已提交
483 484 485 486 487 488 489 490 491
#define STATS_INC_ALLOCHIT(x)	atomic_inc(&(x)->allochit)
#define STATS_INC_ALLOCMISS(x)	atomic_inc(&(x)->allocmiss)
#define STATS_INC_FREEHIT(x)	atomic_inc(&(x)->freehit)
#define STATS_INC_FREEMISS(x)	atomic_inc(&(x)->freemiss)
#else
#define	STATS_INC_ACTIVE(x)	do { } while (0)
#define	STATS_DEC_ACTIVE(x)	do { } while (0)
#define	STATS_INC_ALLOCED(x)	do { } while (0)
#define	STATS_INC_GROWN(x)	do { } while (0)
492
#define	STATS_ADD_REAPED(x,y)	do { } while (0)
L
Linus Torvalds 已提交
493 494 495
#define	STATS_SET_HIGH(x)	do { } while (0)
#define	STATS_INC_ERR(x)	do { } while (0)
#define	STATS_INC_NODEALLOCS(x)	do { } while (0)
496
#define	STATS_INC_NODEFREES(x)	do { } while (0)
497
#define STATS_INC_ACOVERFLOW(x)   do { } while (0)
A
Andrew Morton 已提交
498
#define	STATS_SET_FREEABLE(x, i) do { } while (0)
L
Linus Torvalds 已提交
499 500 501 502 503 504 505 506
#define STATS_INC_ALLOCHIT(x)	do { } while (0)
#define STATS_INC_ALLOCMISS(x)	do { } while (0)
#define STATS_INC_FREEHIT(x)	do { } while (0)
#define STATS_INC_FREEMISS(x)	do { } while (0)
#endif

#if DEBUG

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

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

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

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

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

#else

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

#endif

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

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

A
Andrew Morton 已提交
583 584 585 586
/*
 * Functions for storing/retrieving the cachep and or slab from the page
 * allocator.  These are used to find the slab an obj belongs to.  With kfree(),
 * these are used to find the cache which an obj belongs to.
L
Linus Torvalds 已提交
587
 */
588 589 590 591 592 593 594
static inline void page_set_cache(struct page *page, struct kmem_cache *cache)
{
	page->lru.next = (struct list_head *)cache;
}

static inline struct kmem_cache *page_get_cache(struct page *page)
{
595 596
	if (unlikely(PageCompound(page)))
		page = (struct page *)page_private(page);
597
	BUG_ON(!PageSlab(page));
598 599 600 601 602 603 604 605 606 607
	return (struct kmem_cache *)page->lru.next;
}

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

static inline struct slab *page_get_slab(struct page *page)
{
608 609
	if (unlikely(PageCompound(page)))
		page = (struct page *)page_private(page);
610
	BUG_ON(!PageSlab(page));
611 612
	return (struct slab *)page->lru.prev;
}
L
Linus Torvalds 已提交
613

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

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

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

632 633 634 635 636 637 638 639
/*
 * We want to avoid an expensive divide : (offset / cache->buffer_size)
 *   Using the fact that buffer_size is a constant for a particular cache,
 *   we can replace (offset / cache->buffer_size) by
 *   reciprocal_divide(offset, cache->reciprocal_buffer_size)
 */
static inline unsigned int obj_to_index(const struct kmem_cache *cache,
					const struct slab *slab, void *obj)
640
{
641 642
	u32 offset = (obj - slab->s_mem);
	return reciprocal_divide(offset, cache->reciprocal_buffer_size);
643 644
}

A
Andrew Morton 已提交
645 646 647
/*
 * These are the default caches for kmalloc. Custom caches can have other sizes.
 */
L
Linus Torvalds 已提交
648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664
struct cache_sizes malloc_sizes[] = {
#define CACHE(x) { .cs_size = (x) },
#include <linux/kmalloc_sizes.h>
	CACHE(ULONG_MAX)
#undef CACHE
};
EXPORT_SYMBOL(malloc_sizes);

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

static struct cache_names __initdata cache_names[] = {
#define CACHE(x) { .name = "size-" #x, .name_dma = "size-" #x "(DMA)" },
#include <linux/kmalloc_sizes.h>
P
Pekka Enberg 已提交
665
	{NULL,}
L
Linus Torvalds 已提交
666 667 668 669
#undef CACHE
};

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

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

686 687
#define BAD_ALIEN_MAGIC 0x01020304ul

688 689 690 691 692 693 694 695
#ifdef CONFIG_LOCKDEP

/*
 * Slab sometimes uses the kmalloc slabs to store the slab headers
 * for other slabs "off slab".
 * The locking for this is tricky in that it nests within the locks
 * of all other slabs in a few places; to deal with this special
 * locking we put on-slab caches into a separate lock-class.
696 697 698 699
 *
 * We set lock class for alien array caches which are up during init.
 * The lock annotation will be lost if all cpus of a node goes down and
 * then comes back up during hotplug
700
 */
701 702 703 704
static struct lock_class_key on_slab_l3_key;
static struct lock_class_key on_slab_alc_key;

static inline void init_lock_keys(void)
705 706 707

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

	while (s->cs_size != ULONG_MAX) {
		for_each_node(q) {
			struct array_cache **alc;
			int r;
			struct kmem_list3 *l3 = s->cs_cachep->nodelists[q];
			if (!l3 || OFF_SLAB(s->cs_cachep))
				continue;
			lockdep_set_class(&l3->list_lock, &on_slab_l3_key);
			alc = l3->alien;
			/*
			 * FIXME: This check for BAD_ALIEN_MAGIC
			 * should go away when common slab code is taught to
			 * work even without alien caches.
			 * Currently, non NUMA code returns BAD_ALIEN_MAGIC
			 * for alloc_alien_cache,
			 */
			if (!alc || (unsigned long)alc == BAD_ALIEN_MAGIC)
				continue;
			for_each_node(r) {
				if (alc[r])
					lockdep_set_class(&alc[r]->lock,
					     &on_slab_alc_key);
			}
		}
		s++;
735 736 737
	}
}
#else
738
static inline void init_lock_keys(void)
739 740 741 742
{
}
#endif

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

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

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

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

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

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

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

	/*
792
	 * Really subtle: The last entry with cs->cs_size==ULONG_MAX
L
Linus Torvalds 已提交
793 794 795
	 * has cs_{dma,}cachep==NULL. Thus no special case
	 * for large kmalloc calls required.
	 */
796
#ifdef CONFIG_ZONE_DMA
L
Linus Torvalds 已提交
797 798
	if (unlikely(gfpflags & GFP_DMA))
		return csizep->cs_dmacachep;
799
#endif
L
Linus Torvalds 已提交
800 801 802
	return csizep->cs_cachep;
}

A
Adrian Bunk 已提交
803
static struct kmem_cache *kmem_find_general_cachep(size_t size, gfp_t gfpflags)
804 805 806 807
{
	return __find_general_cachep(size, gfpflags);
}

808
static size_t slab_mgmt_size(size_t nr_objs, size_t align)
L
Linus Torvalds 已提交
809
{
810 811
	return ALIGN(sizeof(struct slab)+nr_objs*sizeof(kmem_bufctl_t), align);
}
L
Linus Torvalds 已提交
812

A
Andrew Morton 已提交
813 814 815
/*
 * Calculate the number of objects and left-over bytes for a given buffer size.
 */
816 817 818 819 820 821 822
static void cache_estimate(unsigned long gfporder, size_t buffer_size,
			   size_t align, int flags, size_t *left_over,
			   unsigned int *num)
{
	int nr_objs;
	size_t mgmt_size;
	size_t slab_size = PAGE_SIZE << gfporder;
L
Linus Torvalds 已提交
823

824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871
	/*
	 * The slab management structure can be either off the slab or
	 * on it. For the latter case, the memory allocated for a
	 * slab is used for:
	 *
	 * - The struct slab
	 * - One kmem_bufctl_t for each object
	 * - Padding to respect alignment of @align
	 * - @buffer_size bytes for each object
	 *
	 * If the slab management structure is off the slab, then the
	 * alignment will already be calculated into the size. Because
	 * the slabs are all pages aligned, the objects will be at the
	 * correct alignment when allocated.
	 */
	if (flags & CFLGS_OFF_SLAB) {
		mgmt_size = 0;
		nr_objs = slab_size / buffer_size;

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

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

		if (nr_objs > SLAB_LIMIT)
			nr_objs = SLAB_LIMIT;

		mgmt_size = slab_mgmt_size(nr_objs, align);
	}
	*num = nr_objs;
	*left_over = slab_size - nr_objs*buffer_size - mgmt_size;
L
Linus Torvalds 已提交
872 873 874 875
}

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

A
Andrew Morton 已提交
876 877
static void __slab_error(const char *function, struct kmem_cache *cachep,
			char *msg)
L
Linus Torvalds 已提交
878 879
{
	printk(KERN_ERR "slab error in %s(): cache `%s': %s\n",
P
Pekka Enberg 已提交
880
	       function, cachep->name, msg);
L
Linus Torvalds 已提交
881 882 883
	dump_stack();
}

884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899
/*
 * By default on NUMA we use alien caches to stage the freeing of
 * objects allocated from other nodes. This causes massive memory
 * inefficiencies when using fake NUMA setup to split memory into a
 * large number of small nodes, so it can be disabled on the command
 * line
  */

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

900 901 902 903 904 905 906 907 908 909 910 911 912 913 914
#ifdef CONFIG_NUMA
/*
 * Special reaping functions for NUMA systems called from cache_reap().
 * These take care of doing round robin flushing of alien caches (containing
 * objects freed on different nodes from which they were allocated) and the
 * flushing of remote pcps by calling drain_node_pages.
 */
static DEFINE_PER_CPU(unsigned long, reap_node);

static void init_reap_node(int cpu)
{
	int node;

	node = next_node(cpu_to_node(cpu), node_online_map);
	if (node == MAX_NUMNODES)
915
		node = first_node(node_online_map);
916

917
	per_cpu(reap_node, cpu) = node;
918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940
}

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

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

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

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

L
Linus Torvalds 已提交
941 942 943 944 945 946 947 948 949
/*
 * 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)
{
950
	struct delayed_work *reap_work = &per_cpu(reap_work, cpu);
L
Linus Torvalds 已提交
951 952 953 954 955 956

	/*
	 * When this gets called from do_initcalls via cpucache_init(),
	 * init_workqueues() has already run, so keventd will be setup
	 * at that time.
	 */
957
	if (keventd_up() && reap_work->work.func == NULL) {
958
		init_reap_node(cpu);
959
		INIT_DELAYED_WORK(reap_work, cache_reap);
960 961
		schedule_delayed_work_on(cpu, reap_work,
					__round_jiffies_relative(HZ, cpu));
L
Linus Torvalds 已提交
962 963 964
	}
}

965
static struct array_cache *alloc_arraycache(int node, int entries,
P
Pekka Enberg 已提交
966
					    int batchcount)
L
Linus Torvalds 已提交
967
{
P
Pekka Enberg 已提交
968
	int memsize = sizeof(void *) * entries + sizeof(struct array_cache);
L
Linus Torvalds 已提交
969 970
	struct array_cache *nc = NULL;

971
	nc = kmalloc_node(memsize, GFP_KERNEL, node);
L
Linus Torvalds 已提交
972 973 974 975 976
	if (nc) {
		nc->avail = 0;
		nc->limit = entries;
		nc->batchcount = batchcount;
		nc->touched = 0;
977
		spin_lock_init(&nc->lock);
L
Linus Torvalds 已提交
978 979 980 981
	}
	return nc;
}

982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005
/*
 * Transfer objects in one arraycache to another.
 * Locking must be handled by the caller.
 *
 * Return the number of entries transferred.
 */
static int transfer_objects(struct array_cache *to,
		struct array_cache *from, unsigned int max)
{
	/* Figure out how many entries to transfer */
	int nr = min(min(from->avail, max), to->limit - to->avail);

	if (!nr)
		return 0;

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

	from->avail -= nr;
	to->avail += nr;
	to->touched = 1;
	return nr;
}

1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030
#ifndef CONFIG_NUMA

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

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

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

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

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

1031
static inline void *____cache_alloc_node(struct kmem_cache *cachep,
1032 1033 1034 1035 1036 1037 1038
		 gfp_t flags, int nodeid)
{
	return NULL;
}

#else	/* CONFIG_NUMA */

1039
static void *____cache_alloc_node(struct kmem_cache *, gfp_t, int);
1040
static void *alternate_node_alloc(struct kmem_cache *, gfp_t);
1041

P
Pekka Enberg 已提交
1042
static struct array_cache **alloc_alien_cache(int node, int limit)
1043 1044
{
	struct array_cache **ac_ptr;
1045
	int memsize = sizeof(void *) * nr_node_ids;
1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058
	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
1059
				for (i--; i <= 0; i--)
1060 1061 1062 1063 1064 1065 1066 1067 1068
					kfree(ac_ptr[i]);
				kfree(ac_ptr);
				return NULL;
			}
		}
	}
	return ac_ptr;
}

P
Pekka Enberg 已提交
1069
static void free_alien_cache(struct array_cache **ac_ptr)
1070 1071 1072 1073 1074 1075
{
	int i;

	if (!ac_ptr)
		return;
	for_each_node(i)
P
Pekka Enberg 已提交
1076
	    kfree(ac_ptr[i]);
1077 1078 1079
	kfree(ac_ptr);
}

1080
static void __drain_alien_cache(struct kmem_cache *cachep,
P
Pekka Enberg 已提交
1081
				struct array_cache *ac, int node)
1082 1083 1084 1085 1086
{
	struct kmem_list3 *rl3 = cachep->nodelists[node];

	if (ac->avail) {
		spin_lock(&rl3->list_lock);
1087 1088 1089 1090 1091
		/*
		 * Stuff objects into the remote nodes shared array first.
		 * That way we could avoid the overhead of putting the objects
		 * into the free lists and getting them back later.
		 */
1092 1093
		if (rl3->shared)
			transfer_objects(rl3->shared, ac, ac->limit);
1094

1095
		free_block(cachep, ac->entry, ac->avail, node);
1096 1097 1098 1099 1100
		ac->avail = 0;
		spin_unlock(&rl3->list_lock);
	}
}

1101 1102 1103 1104 1105 1106 1107 1108 1109
/*
 * Called from cache_reap() to regularly drain alien caches round robin.
 */
static void reap_alien(struct kmem_cache *cachep, struct kmem_list3 *l3)
{
	int node = __get_cpu_var(reap_node);

	if (l3->alien) {
		struct array_cache *ac = l3->alien[node];
1110 1111

		if (ac && ac->avail && spin_trylock_irq(&ac->lock)) {
1112 1113 1114 1115 1116 1117
			__drain_alien_cache(cachep, ac, node);
			spin_unlock_irq(&ac->lock);
		}
	}
}

A
Andrew Morton 已提交
1118 1119
static void drain_alien_cache(struct kmem_cache *cachep,
				struct array_cache **alien)
1120
{
P
Pekka Enberg 已提交
1121
	int i = 0;
1122 1123 1124 1125
	struct array_cache *ac;
	unsigned long flags;

	for_each_online_node(i) {
1126
		ac = alien[i];
1127 1128 1129 1130 1131 1132 1133
		if (ac) {
			spin_lock_irqsave(&ac->lock, flags);
			__drain_alien_cache(cachep, ac, i);
			spin_unlock_irqrestore(&ac->lock, flags);
		}
	}
}
1134

1135
static inline int cache_free_alien(struct kmem_cache *cachep, void *objp)
1136 1137 1138 1139 1140
{
	struct slab *slabp = virt_to_slab(objp);
	int nodeid = slabp->nodeid;
	struct kmem_list3 *l3;
	struct array_cache *alien = NULL;
P
Pekka Enberg 已提交
1141 1142 1143
	int node;

	node = numa_node_id();
1144 1145 1146 1147 1148

	/*
	 * Make sure we are not freeing a object from another node to the array
	 * cache on this cpu.
	 */
1149
	if (likely(slabp->nodeid == node))
1150 1151
		return 0;

P
Pekka Enberg 已提交
1152
	l3 = cachep->nodelists[node];
1153 1154 1155
	STATS_INC_NODEFREES(cachep);
	if (l3->alien && l3->alien[nodeid]) {
		alien = l3->alien[nodeid];
1156
		spin_lock(&alien->lock);
1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169
		if (unlikely(alien->avail == alien->limit)) {
			STATS_INC_ACOVERFLOW(cachep);
			__drain_alien_cache(cachep, alien, nodeid);
		}
		alien->entry[alien->avail++] = objp;
		spin_unlock(&alien->lock);
	} else {
		spin_lock(&(cachep->nodelists[nodeid])->list_lock);
		free_block(cachep, &objp, 1, nodeid);
		spin_unlock(&(cachep->nodelists[nodeid])->list_lock);
	}
	return 1;
}
1170 1171
#endif

1172
static int __cpuinit cpuup_callback(struct notifier_block *nfb,
P
Pekka Enberg 已提交
1173
				    unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
1174 1175
{
	long cpu = (long)hcpu;
1176
	struct kmem_cache *cachep;
1177 1178 1179
	struct kmem_list3 *l3 = NULL;
	int node = cpu_to_node(cpu);
	int memsize = sizeof(struct kmem_list3);
L
Linus Torvalds 已提交
1180 1181 1182

	switch (action) {
	case CPU_UP_PREPARE:
I
Ingo Molnar 已提交
1183
		mutex_lock(&cache_chain_mutex);
A
Andrew Morton 已提交
1184 1185
		/*
		 * We need to do this right in the beginning since
1186 1187 1188 1189 1190
		 * 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
1191
		list_for_each_entry(cachep, &cache_chain, next) {
A
Andrew Morton 已提交
1192 1193
			/*
			 * Set up the size64 kmemlist for cpu before we can
1194 1195 1196 1197
			 * begin anything. Make sure some other cpu on this
			 * node has not already allocated this
			 */
			if (!cachep->nodelists[node]) {
A
Andrew Morton 已提交
1198 1199
				l3 = kmalloc_node(memsize, GFP_KERNEL, node);
				if (!l3)
1200 1201 1202
					goto bad;
				kmem_list3_init(l3);
				l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
P
Pekka Enberg 已提交
1203
				    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
1204

1205 1206 1207 1208 1209
				/*
				 * The l3s don't come and go as CPUs come and
				 * go.  cache_chain_mutex is sufficient
				 * protection here.
				 */
1210 1211
				cachep->nodelists[node] = l3;
			}
L
Linus Torvalds 已提交
1212

1213 1214
			spin_lock_irq(&cachep->nodelists[node]->list_lock);
			cachep->nodelists[node]->free_limit =
A
Andrew Morton 已提交
1215 1216
				(1 + nr_cpus_node(node)) *
				cachep->batchcount + cachep->num;
1217 1218 1219
			spin_unlock_irq(&cachep->nodelists[node]->list_lock);
		}

A
Andrew Morton 已提交
1220 1221 1222 1223
		/*
		 * Now we can go ahead with allocating the shared arrays and
		 * array caches
		 */
1224
		list_for_each_entry(cachep, &cache_chain, next) {
1225
			struct array_cache *nc;
1226
			struct array_cache *shared;
1227
			struct array_cache **alien = NULL;
1228

1229
			nc = alloc_arraycache(node, cachep->limit,
1230
						cachep->batchcount);
L
Linus Torvalds 已提交
1231 1232
			if (!nc)
				goto bad;
1233 1234 1235 1236 1237
			shared = alloc_arraycache(node,
					cachep->shared * cachep->batchcount,
					0xbaadf00d);
			if (!shared)
				goto bad;
1238

1239 1240 1241 1242 1243
			if (use_alien_caches) {
                                alien = alloc_alien_cache(node, cachep->limit);
                                if (!alien)
                                        goto bad;
                        }
L
Linus Torvalds 已提交
1244
			cachep->array[cpu] = nc;
1245 1246 1247
			l3 = cachep->nodelists[node];
			BUG_ON(!l3);

1248 1249 1250 1251 1252 1253 1254 1255
			spin_lock_irq(&l3->list_lock);
			if (!l3->shared) {
				/*
				 * We are serialised from CPU_DEAD or
				 * CPU_UP_CANCELLED by the cpucontrol lock
				 */
				l3->shared = shared;
				shared = NULL;
1256
			}
1257 1258 1259 1260 1261 1262 1263 1264 1265
#ifdef CONFIG_NUMA
			if (!l3->alien) {
				l3->alien = alien;
				alien = NULL;
			}
#endif
			spin_unlock_irq(&l3->list_lock);
			kfree(shared);
			free_alien_cache(alien);
L
Linus Torvalds 已提交
1266 1267 1268
		}
		break;
	case CPU_ONLINE:
1269
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1270 1271 1272
		start_cpu_timer(cpu);
		break;
#ifdef CONFIG_HOTPLUG_CPU
1273 1274 1275 1276 1277 1278
	case CPU_DOWN_PREPARE:
		mutex_lock(&cache_chain_mutex);
		break;
	case CPU_DOWN_FAILED:
		mutex_unlock(&cache_chain_mutex);
		break;
L
Linus Torvalds 已提交
1279
	case CPU_DEAD:
1280 1281 1282 1283 1284 1285 1286 1287
		/*
		 * Even if all the cpus of a node are down, we don't free the
		 * kmem_list3 of any cache. This to avoid a race between
		 * cpu_down, and a kmalloc allocation from another cpu for
		 * memory from the node of the cpu going down.  The list3
		 * structure is usually allocated from kmem_cache_create() and
		 * gets destroyed at kmem_cache_destroy().
		 */
L
Linus Torvalds 已提交
1288
		/* fall thru */
1289
#endif
L
Linus Torvalds 已提交
1290 1291 1292
	case CPU_UP_CANCELED:
		list_for_each_entry(cachep, &cache_chain, next) {
			struct array_cache *nc;
1293 1294
			struct array_cache *shared;
			struct array_cache **alien;
1295
			cpumask_t mask;
L
Linus Torvalds 已提交
1296

1297
			mask = node_to_cpumask(node);
L
Linus Torvalds 已提交
1298 1299 1300
			/* cpu is dead; no one can alloc from it. */
			nc = cachep->array[cpu];
			cachep->array[cpu] = NULL;
1301 1302 1303
			l3 = cachep->nodelists[node];

			if (!l3)
1304
				goto free_array_cache;
1305

1306
			spin_lock_irq(&l3->list_lock);
1307 1308 1309 1310

			/* Free limit for this kmem_list3 */
			l3->free_limit -= cachep->batchcount;
			if (nc)
1311
				free_block(cachep, nc->entry, nc->avail, node);
1312 1313

			if (!cpus_empty(mask)) {
1314
				spin_unlock_irq(&l3->list_lock);
1315
				goto free_array_cache;
P
Pekka Enberg 已提交
1316
			}
1317

1318 1319
			shared = l3->shared;
			if (shared) {
1320
				free_block(cachep, l3->shared->entry,
P
Pekka Enberg 已提交
1321
					   l3->shared->avail, node);
1322 1323 1324
				l3->shared = NULL;
			}

1325 1326 1327 1328 1329 1330 1331 1332 1333
			alien = l3->alien;
			l3->alien = NULL;

			spin_unlock_irq(&l3->list_lock);

			kfree(shared);
			if (alien) {
				drain_alien_cache(cachep, alien);
				free_alien_cache(alien);
1334
			}
1335
free_array_cache:
L
Linus Torvalds 已提交
1336 1337
			kfree(nc);
		}
1338 1339 1340 1341 1342 1343 1344 1345 1346
		/*
		 * In the previous loop, all the objects were freed to
		 * the respective cache's slabs,  now we can go ahead and
		 * shrink each nodelist to its limit.
		 */
		list_for_each_entry(cachep, &cache_chain, next) {
			l3 = cachep->nodelists[node];
			if (!l3)
				continue;
1347
			drain_freelist(cachep, l3, l3->free_objects);
1348
		}
I
Ingo Molnar 已提交
1349
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1350 1351 1352
		break;
	}
	return NOTIFY_OK;
A
Andrew Morton 已提交
1353
bad:
L
Linus Torvalds 已提交
1354 1355 1356
	return NOTIFY_BAD;
}

1357 1358 1359
static struct notifier_block __cpuinitdata cpucache_notifier = {
	&cpuup_callback, NULL, 0
};
L
Linus Torvalds 已提交
1360

1361 1362 1363
/*
 * swap the static kmem_list3 with kmalloced memory
 */
A
Andrew Morton 已提交
1364 1365
static void init_list(struct kmem_cache *cachep, struct kmem_list3 *list,
			int nodeid)
1366 1367 1368 1369 1370 1371 1372 1373
{
	struct kmem_list3 *ptr;

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

	local_irq_disable();
	memcpy(ptr, list, sizeof(struct kmem_list3));
1374 1375 1376 1377 1378
	/*
	 * Do not assume that spinlocks can be initialized via memcpy:
	 */
	spin_lock_init(&ptr->list_lock);

1379 1380 1381 1382 1383
	MAKE_ALL_LISTS(cachep, ptr, nodeid);
	cachep->nodelists[nodeid] = ptr;
	local_irq_enable();
}

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

1397 1398 1399
	if (num_possible_nodes() == 1)
		use_alien_caches = 0;

1400 1401 1402 1403 1404
	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
1405 1406 1407 1408 1409 1410 1411 1412 1413 1414

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

	/* Bootstrap is tricky, because several objects are allocated
	 * from caches that do not exist yet:
A
Andrew Morton 已提交
1415 1416 1417
	 * 1) initialize the cache_cache cache: it contains the struct
	 *    kmem_cache structures of all caches, except cache_cache itself:
	 *    cache_cache is statically allocated.
1418 1419 1420
	 *    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
1421
	 * 2) Create the first kmalloc cache.
1422
	 *    The struct kmem_cache for the new cache is allocated normally.
1423 1424 1425
	 *    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
1426 1427
	 * 4) Replace the __init data head arrays for cache_cache and the first
	 *    kmalloc cache with kmalloc allocated arrays.
1428 1429 1430
	 * 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
1431 1432
	 */

P
Pekka Enberg 已提交
1433 1434
	node = numa_node_id();

L
Linus Torvalds 已提交
1435 1436 1437 1438 1439
	/* 1) create the cache_cache */
	INIT_LIST_HEAD(&cache_chain);
	list_add(&cache_cache.next, &cache_chain);
	cache_cache.colour_off = cache_line_size();
	cache_cache.array[smp_processor_id()] = &initarray_cache.cache;
P
Pekka Enberg 已提交
1440
	cache_cache.nodelists[node] = &initkmem_list3[CACHE_CACHE];
L
Linus Torvalds 已提交
1441

A
Andrew Morton 已提交
1442 1443
	cache_cache.buffer_size = ALIGN(cache_cache.buffer_size,
					cache_line_size());
1444 1445
	cache_cache.reciprocal_buffer_size =
		reciprocal_value(cache_cache.buffer_size);
L
Linus Torvalds 已提交
1446

1447 1448 1449 1450 1451 1452
	for (order = 0; order < MAX_ORDER; order++) {
		cache_estimate(order, cache_cache.buffer_size,
			cache_line_size(), 0, &left_over, &cache_cache.num);
		if (cache_cache.num)
			break;
	}
1453
	BUG_ON(!cache_cache.num);
1454
	cache_cache.gfporder = order;
P
Pekka Enberg 已提交
1455 1456 1457
	cache_cache.colour = left_over / cache_cache.colour_off;
	cache_cache.slab_size = ALIGN(cache_cache.num * sizeof(kmem_bufctl_t) +
				      sizeof(struct slab), cache_line_size());
L
Linus Torvalds 已提交
1458 1459 1460 1461 1462

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

A
Andrew Morton 已提交
1463 1464 1465 1466
	/*
	 * Initialize the caches that provide memory for the array cache and the
	 * kmem_list3 structures first.  Without this, further allocations will
	 * bug.
1467 1468 1469
	 */

	sizes[INDEX_AC].cs_cachep = kmem_cache_create(names[INDEX_AC].name,
A
Andrew Morton 已提交
1470 1471 1472 1473
					sizes[INDEX_AC].cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_PANIC,
					NULL, NULL);
1474

A
Andrew Morton 已提交
1475
	if (INDEX_AC != INDEX_L3) {
1476
		sizes[INDEX_L3].cs_cachep =
A
Andrew Morton 已提交
1477 1478 1479 1480 1481 1482
			kmem_cache_create(names[INDEX_L3].name,
				sizes[INDEX_L3].cs_size,
				ARCH_KMALLOC_MINALIGN,
				ARCH_KMALLOC_FLAGS|SLAB_PANIC,
				NULL, NULL);
	}
1483

1484 1485
	slab_early_init = 0;

L
Linus Torvalds 已提交
1486
	while (sizes->cs_size != ULONG_MAX) {
1487 1488
		/*
		 * For performance, all the general caches are L1 aligned.
L
Linus Torvalds 已提交
1489 1490 1491
		 * This should be particularly beneficial on SMP boxes, as it
		 * eliminates "false sharing".
		 * Note for systems short on memory removing the alignment will
1492 1493
		 * allow tighter packing of the smaller caches.
		 */
A
Andrew Morton 已提交
1494
		if (!sizes->cs_cachep) {
1495
			sizes->cs_cachep = kmem_cache_create(names->name,
A
Andrew Morton 已提交
1496 1497 1498 1499 1500
					sizes->cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_PANIC,
					NULL, NULL);
		}
1501 1502 1503
#ifdef CONFIG_ZONE_DMA
		sizes->cs_dmacachep = kmem_cache_create(
					names->name_dma,
A
Andrew Morton 已提交
1504 1505 1506 1507 1508
					sizes->cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_CACHE_DMA|
						SLAB_PANIC,
					NULL, NULL);
1509
#endif
L
Linus Torvalds 已提交
1510 1511 1512 1513 1514
		sizes++;
		names++;
	}
	/* 4) Replace the bootstrap head arrays */
	{
1515
		struct array_cache *ptr;
1516

L
Linus Torvalds 已提交
1517
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
1518

L
Linus Torvalds 已提交
1519
		local_irq_disable();
1520 1521
		BUG_ON(cpu_cache_get(&cache_cache) != &initarray_cache.cache);
		memcpy(ptr, cpu_cache_get(&cache_cache),
P
Pekka Enberg 已提交
1522
		       sizeof(struct arraycache_init));
1523 1524 1525 1526 1527
		/*
		 * Do not assume that spinlocks can be initialized via memcpy:
		 */
		spin_lock_init(&ptr->lock);

L
Linus Torvalds 已提交
1528 1529
		cache_cache.array[smp_processor_id()] = ptr;
		local_irq_enable();
1530

L
Linus Torvalds 已提交
1531
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
1532

L
Linus Torvalds 已提交
1533
		local_irq_disable();
1534
		BUG_ON(cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep)
P
Pekka Enberg 已提交
1535
		       != &initarray_generic.cache);
1536
		memcpy(ptr, cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep),
P
Pekka Enberg 已提交
1537
		       sizeof(struct arraycache_init));
1538 1539 1540 1541 1542
		/*
		 * Do not assume that spinlocks can be initialized via memcpy:
		 */
		spin_lock_init(&ptr->lock);

1543
		malloc_sizes[INDEX_AC].cs_cachep->array[smp_processor_id()] =
P
Pekka Enberg 已提交
1544
		    ptr;
L
Linus Torvalds 已提交
1545 1546
		local_irq_enable();
	}
1547 1548
	/* 5) Replace the bootstrap kmem_list3's */
	{
P
Pekka Enberg 已提交
1549 1550
		int nid;

1551
		/* Replace the static kmem_list3 structures for the boot cpu */
P
Pekka Enberg 已提交
1552
		init_list(&cache_cache, &initkmem_list3[CACHE_CACHE], node);
1553

P
Pekka Enberg 已提交
1554
		for_each_online_node(nid) {
1555
			init_list(malloc_sizes[INDEX_AC].cs_cachep,
P
Pekka Enberg 已提交
1556
				  &initkmem_list3[SIZE_AC + nid], nid);
1557 1558 1559

			if (INDEX_AC != INDEX_L3) {
				init_list(malloc_sizes[INDEX_L3].cs_cachep,
P
Pekka Enberg 已提交
1560
					  &initkmem_list3[SIZE_L3 + nid], nid);
1561 1562 1563
			}
		}
	}
L
Linus Torvalds 已提交
1564

1565
	/* 6) resize the head arrays to their final sizes */
L
Linus Torvalds 已提交
1566
	{
1567
		struct kmem_cache *cachep;
I
Ingo Molnar 已提交
1568
		mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1569
		list_for_each_entry(cachep, &cache_chain, next)
1570 1571
			if (enable_cpucache(cachep))
				BUG();
I
Ingo Molnar 已提交
1572
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1573 1574
	}

1575 1576 1577 1578
	/* Annotate slab for lockdep -- annotate the malloc caches */
	init_lock_keys();


L
Linus Torvalds 已提交
1579 1580 1581
	/* Done! */
	g_cpucache_up = FULL;

A
Andrew Morton 已提交
1582 1583 1584
	/*
	 * Register a cpu startup notifier callback that initializes
	 * cpu_cache_get for all new cpus
L
Linus Torvalds 已提交
1585 1586 1587
	 */
	register_cpu_notifier(&cpucache_notifier);

A
Andrew Morton 已提交
1588 1589 1590
	/*
	 * The reap timers are started later, with a module init call: That part
	 * of the kernel is not yet operational.
L
Linus Torvalds 已提交
1591 1592 1593 1594 1595 1596 1597
	 */
}

static int __init cpucache_init(void)
{
	int cpu;

A
Andrew Morton 已提交
1598 1599
	/*
	 * Register the timers that return unneeded pages to the page allocator
L
Linus Torvalds 已提交
1600
	 */
1601
	for_each_online_cpu(cpu)
A
Andrew Morton 已提交
1602
		start_cpu_timer(cpu);
L
Linus Torvalds 已提交
1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613
	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.
 */
1614
static void *kmem_getpages(struct kmem_cache *cachep, gfp_t flags, int nodeid)
L
Linus Torvalds 已提交
1615 1616
{
	struct page *page;
1617
	int nr_pages;
L
Linus Torvalds 已提交
1618 1619
	int i;

1620
#ifndef CONFIG_MMU
1621 1622 1623
	/*
	 * Nommu uses slab's for process anonymous memory allocations, and thus
	 * requires __GFP_COMP to properly refcount higher order allocations
1624
	 */
1625
	flags |= __GFP_COMP;
1626
#endif
1627

1628
	flags |= cachep->gfpflags;
1629 1630

	page = alloc_pages_node(nodeid, flags, cachep->gfporder);
L
Linus Torvalds 已提交
1631 1632 1633
	if (!page)
		return NULL;

1634
	nr_pages = (1 << cachep->gfporder);
L
Linus Torvalds 已提交
1635
	if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
1636 1637 1638 1639 1640
		add_zone_page_state(page_zone(page),
			NR_SLAB_RECLAIMABLE, nr_pages);
	else
		add_zone_page_state(page_zone(page),
			NR_SLAB_UNRECLAIMABLE, nr_pages);
1641 1642 1643
	for (i = 0; i < nr_pages; i++)
		__SetPageSlab(page + i);
	return page_address(page);
L
Linus Torvalds 已提交
1644 1645 1646 1647 1648
}

/*
 * Interface to system's page release.
 */
1649
static void kmem_freepages(struct kmem_cache *cachep, void *addr)
L
Linus Torvalds 已提交
1650
{
P
Pekka Enberg 已提交
1651
	unsigned long i = (1 << cachep->gfporder);
L
Linus Torvalds 已提交
1652 1653 1654
	struct page *page = virt_to_page(addr);
	const unsigned long nr_freed = i;

1655 1656 1657 1658 1659 1660
	if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
		sub_zone_page_state(page_zone(page),
				NR_SLAB_RECLAIMABLE, nr_freed);
	else
		sub_zone_page_state(page_zone(page),
				NR_SLAB_UNRECLAIMABLE, nr_freed);
L
Linus Torvalds 已提交
1661
	while (i--) {
N
Nick Piggin 已提交
1662 1663
		BUG_ON(!PageSlab(page));
		__ClearPageSlab(page);
L
Linus Torvalds 已提交
1664 1665 1666 1667 1668 1669 1670 1671 1672
		page++;
	}
	if (current->reclaim_state)
		current->reclaim_state->reclaimed_slab += nr_freed;
	free_pages((unsigned long)addr, cachep->gfporder);
}

static void kmem_rcu_free(struct rcu_head *head)
{
P
Pekka Enberg 已提交
1673
	struct slab_rcu *slab_rcu = (struct slab_rcu *)head;
1674
	struct kmem_cache *cachep = slab_rcu->cachep;
L
Linus Torvalds 已提交
1675 1676 1677 1678 1679 1680 1681 1682 1683

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

#if DEBUG

#ifdef CONFIG_DEBUG_PAGEALLOC
1684
static void store_stackinfo(struct kmem_cache *cachep, unsigned long *addr,
P
Pekka Enberg 已提交
1685
			    unsigned long caller)
L
Linus Torvalds 已提交
1686
{
1687
	int size = obj_size(cachep);
L
Linus Torvalds 已提交
1688

1689
	addr = (unsigned long *)&((char *)addr)[obj_offset(cachep)];
L
Linus Torvalds 已提交
1690

P
Pekka Enberg 已提交
1691
	if (size < 5 * sizeof(unsigned long))
L
Linus Torvalds 已提交
1692 1693
		return;

P
Pekka Enberg 已提交
1694 1695 1696 1697
	*addr++ = 0x12345678;
	*addr++ = caller;
	*addr++ = smp_processor_id();
	size -= 3 * sizeof(unsigned long);
L
Linus Torvalds 已提交
1698 1699 1700 1701 1702 1703 1704
	{
		unsigned long *sptr = &caller;
		unsigned long svalue;

		while (!kstack_end(sptr)) {
			svalue = *sptr++;
			if (kernel_text_address(svalue)) {
P
Pekka Enberg 已提交
1705
				*addr++ = svalue;
L
Linus Torvalds 已提交
1706 1707 1708 1709 1710 1711 1712
				size -= sizeof(unsigned long);
				if (size <= sizeof(unsigned long))
					break;
			}
		}

	}
P
Pekka Enberg 已提交
1713
	*addr++ = 0x87654321;
L
Linus Torvalds 已提交
1714 1715 1716
}
#endif

1717
static void poison_obj(struct kmem_cache *cachep, void *addr, unsigned char val)
L
Linus Torvalds 已提交
1718
{
1719 1720
	int size = obj_size(cachep);
	addr = &((char *)addr)[obj_offset(cachep)];
L
Linus Torvalds 已提交
1721 1722

	memset(addr, val, size);
P
Pekka Enberg 已提交
1723
	*(unsigned char *)(addr + size - 1) = POISON_END;
L
Linus Torvalds 已提交
1724 1725 1726 1727 1728
}

static void dump_line(char *data, int offset, int limit)
{
	int i;
D
Dave Jones 已提交
1729 1730 1731
	unsigned char error = 0;
	int bad_count = 0;

L
Linus Torvalds 已提交
1732
	printk(KERN_ERR "%03x:", offset);
D
Dave Jones 已提交
1733 1734 1735 1736 1737
	for (i = 0; i < limit; i++) {
		if (data[offset + i] != POISON_FREE) {
			error = data[offset + i];
			bad_count++;
		}
P
Pekka Enberg 已提交
1738
		printk(" %02x", (unsigned char)data[offset + i]);
D
Dave Jones 已提交
1739
	}
L
Linus Torvalds 已提交
1740
	printk("\n");
D
Dave Jones 已提交
1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754

	if (bad_count == 1) {
		error ^= POISON_FREE;
		if (!(error & (error - 1))) {
			printk(KERN_ERR "Single bit error detected. Probably "
					"bad RAM.\n");
#ifdef CONFIG_X86
			printk(KERN_ERR "Run memtest86+ or a similar memory "
					"test tool.\n");
#else
			printk(KERN_ERR "Run a memory test tool.\n");
#endif
		}
	}
L
Linus Torvalds 已提交
1755 1756 1757 1758 1759
}
#endif

#if DEBUG

1760
static void print_objinfo(struct kmem_cache *cachep, void *objp, int lines)
L
Linus Torvalds 已提交
1761 1762 1763 1764 1765 1766
{
	int i, size;
	char *realobj;

	if (cachep->flags & SLAB_RED_ZONE) {
		printk(KERN_ERR "Redzone: 0x%lx/0x%lx.\n",
A
Andrew Morton 已提交
1767 1768
			*dbg_redzone1(cachep, objp),
			*dbg_redzone2(cachep, objp));
L
Linus Torvalds 已提交
1769 1770 1771 1772
	}

	if (cachep->flags & SLAB_STORE_USER) {
		printk(KERN_ERR "Last user: [<%p>]",
A
Andrew Morton 已提交
1773
			*dbg_userword(cachep, objp));
L
Linus Torvalds 已提交
1774
		print_symbol("(%s)",
A
Andrew Morton 已提交
1775
				(unsigned long)*dbg_userword(cachep, objp));
L
Linus Torvalds 已提交
1776 1777
		printk("\n");
	}
1778 1779
	realobj = (char *)objp + obj_offset(cachep);
	size = obj_size(cachep);
P
Pekka Enberg 已提交
1780
	for (i = 0; i < size && lines; i += 16, lines--) {
L
Linus Torvalds 已提交
1781 1782
		int limit;
		limit = 16;
P
Pekka Enberg 已提交
1783 1784
		if (i + limit > size)
			limit = size - i;
L
Linus Torvalds 已提交
1785 1786 1787 1788
		dump_line(realobj, i, limit);
	}
}

1789
static void check_poison_obj(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
1790 1791 1792 1793 1794
{
	char *realobj;
	int size, i;
	int lines = 0;

1795 1796
	realobj = (char *)objp + obj_offset(cachep);
	size = obj_size(cachep);
L
Linus Torvalds 已提交
1797

P
Pekka Enberg 已提交
1798
	for (i = 0; i < size; i++) {
L
Linus Torvalds 已提交
1799
		char exp = POISON_FREE;
P
Pekka Enberg 已提交
1800
		if (i == size - 1)
L
Linus Torvalds 已提交
1801 1802 1803 1804 1805 1806
			exp = POISON_END;
		if (realobj[i] != exp) {
			int limit;
			/* Mismatch ! */
			/* Print header */
			if (lines == 0) {
P
Pekka Enberg 已提交
1807
				printk(KERN_ERR
1808 1809
					"Slab corruption: %s start=%p, len=%d\n",
					cachep->name, realobj, size);
L
Linus Torvalds 已提交
1810 1811 1812
				print_objinfo(cachep, objp, 0);
			}
			/* Hexdump the affected line */
P
Pekka Enberg 已提交
1813
			i = (i / 16) * 16;
L
Linus Torvalds 已提交
1814
			limit = 16;
P
Pekka Enberg 已提交
1815 1816
			if (i + limit > size)
				limit = size - i;
L
Linus Torvalds 已提交
1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828
			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:
		 */
1829
		struct slab *slabp = virt_to_slab(objp);
1830
		unsigned int objnr;
L
Linus Torvalds 已提交
1831

1832
		objnr = obj_to_index(cachep, slabp, objp);
L
Linus Torvalds 已提交
1833
		if (objnr) {
1834
			objp = index_to_obj(cachep, slabp, objnr - 1);
1835
			realobj = (char *)objp + obj_offset(cachep);
L
Linus Torvalds 已提交
1836
			printk(KERN_ERR "Prev obj: start=%p, len=%d\n",
P
Pekka Enberg 已提交
1837
			       realobj, size);
L
Linus Torvalds 已提交
1838 1839
			print_objinfo(cachep, objp, 2);
		}
P
Pekka Enberg 已提交
1840
		if (objnr + 1 < cachep->num) {
1841
			objp = index_to_obj(cachep, slabp, objnr + 1);
1842
			realobj = (char *)objp + obj_offset(cachep);
L
Linus Torvalds 已提交
1843
			printk(KERN_ERR "Next obj: start=%p, len=%d\n",
P
Pekka Enberg 已提交
1844
			       realobj, size);
L
Linus Torvalds 已提交
1845 1846 1847 1848 1849 1850
			print_objinfo(cachep, objp, 2);
		}
	}
}
#endif

1851 1852
#if DEBUG
/**
1853 1854 1855 1856 1857 1858
 * slab_destroy_objs - destroy a slab and its objects
 * @cachep: cache pointer being destroyed
 * @slabp: slab pointer being destroyed
 *
 * Call the registered destructor for each object in a slab that is being
 * destroyed.
L
Linus Torvalds 已提交
1859
 */
1860
static void slab_destroy_objs(struct kmem_cache *cachep, struct slab *slabp)
L
Linus Torvalds 已提交
1861 1862 1863
{
	int i;
	for (i = 0; i < cachep->num; i++) {
1864
		void *objp = index_to_obj(cachep, slabp, i);
L
Linus Torvalds 已提交
1865 1866 1867

		if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
A
Andrew Morton 已提交
1868 1869
			if (cachep->buffer_size % PAGE_SIZE == 0 &&
					OFF_SLAB(cachep))
P
Pekka Enberg 已提交
1870
				kernel_map_pages(virt_to_page(objp),
A
Andrew Morton 已提交
1871
					cachep->buffer_size / PAGE_SIZE, 1);
L
Linus Torvalds 已提交
1872 1873 1874 1875 1876 1877 1878 1879 1880
			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
1881
					   "was overwritten");
L
Linus Torvalds 已提交
1882 1883
			if (*dbg_redzone2(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "end of a freed object "
P
Pekka Enberg 已提交
1884
					   "was overwritten");
L
Linus Torvalds 已提交
1885 1886
		}
		if (cachep->dtor && !(cachep->flags & SLAB_POISON))
1887
			(cachep->dtor) (objp + obj_offset(cachep), cachep, 0);
L
Linus Torvalds 已提交
1888
	}
1889
}
L
Linus Torvalds 已提交
1890
#else
1891
static void slab_destroy_objs(struct kmem_cache *cachep, struct slab *slabp)
1892
{
L
Linus Torvalds 已提交
1893 1894 1895
	if (cachep->dtor) {
		int i;
		for (i = 0; i < cachep->num; i++) {
1896
			void *objp = index_to_obj(cachep, slabp, i);
P
Pekka Enberg 已提交
1897
			(cachep->dtor) (objp, cachep, 0);
L
Linus Torvalds 已提交
1898 1899
		}
	}
1900
}
L
Linus Torvalds 已提交
1901 1902
#endif

1903 1904 1905 1906 1907
/**
 * slab_destroy - destroy and release all objects in a slab
 * @cachep: cache pointer being destroyed
 * @slabp: slab pointer being destroyed
 *
1908
 * Destroy all the objs in a slab, and release the mem back to the system.
A
Andrew Morton 已提交
1909 1910
 * Before calling the slab must have been unlinked from the cache.  The
 * cache-lock is not held/needed.
1911
 */
1912
static void slab_destroy(struct kmem_cache *cachep, struct slab *slabp)
1913 1914 1915 1916
{
	void *addr = slabp->s_mem - slabp->colouroff;

	slab_destroy_objs(cachep, slabp);
L
Linus Torvalds 已提交
1917 1918 1919
	if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU)) {
		struct slab_rcu *slab_rcu;

P
Pekka Enberg 已提交
1920
		slab_rcu = (struct slab_rcu *)slabp;
L
Linus Torvalds 已提交
1921 1922 1923 1924 1925
		slab_rcu->cachep = cachep;
		slab_rcu->addr = addr;
		call_rcu(&slab_rcu->head, kmem_rcu_free);
	} else {
		kmem_freepages(cachep, addr);
1926 1927
		if (OFF_SLAB(cachep))
			kmem_cache_free(cachep->slabp_cache, slabp);
L
Linus Torvalds 已提交
1928 1929 1930
	}
}

A
Andrew Morton 已提交
1931 1932 1933 1934
/*
 * For setting up all the kmem_list3s for cache whose buffer_size is same as
 * size of kmem_list3.
 */
1935
static void set_up_list3s(struct kmem_cache *cachep, int index)
1936 1937 1938 1939
{
	int node;

	for_each_online_node(node) {
P
Pekka Enberg 已提交
1940
		cachep->nodelists[node] = &initkmem_list3[index + node];
1941
		cachep->nodelists[node]->next_reap = jiffies +
P
Pekka Enberg 已提交
1942 1943
		    REAPTIMEOUT_LIST3 +
		    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
1944 1945 1946
	}
}

1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967
static void __kmem_cache_destroy(struct kmem_cache *cachep)
{
	int i;
	struct kmem_list3 *l3;

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

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


1968
/**
1969 1970 1971 1972 1973 1974 1975
 * calculate_slab_order - calculate size (page order) of slabs
 * @cachep: pointer to the cache that is being created
 * @size: size of objects to be created in this cache.
 * @align: required alignment for the objects.
 * @flags: slab allocation flags
 *
 * Also calculates the number of objects per slab.
1976 1977 1978 1979 1980
 *
 * This could be made much more intelligent.  For now, try to avoid using
 * high order pages for slabs.  When the gfp() functions are more friendly
 * towards high-order requests, this should be changed.
 */
A
Andrew Morton 已提交
1981
static size_t calculate_slab_order(struct kmem_cache *cachep,
R
Randy Dunlap 已提交
1982
			size_t size, size_t align, unsigned long flags)
1983
{
1984
	unsigned long offslab_limit;
1985
	size_t left_over = 0;
1986
	int gfporder;
1987

A
Andrew Morton 已提交
1988
	for (gfporder = 0; gfporder <= MAX_GFP_ORDER; gfporder++) {
1989 1990 1991
		unsigned int num;
		size_t remainder;

1992
		cache_estimate(gfporder, size, align, flags, &remainder, &num);
1993 1994
		if (!num)
			continue;
1995

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
		if (flags & CFLGS_OFF_SLAB) {
			/*
			 * Max number of objs-per-slab for caches which
			 * use off-slab slabs. Needed to avoid a possible
			 * looping condition in cache_grow().
			 */
			offslab_limit = size - sizeof(struct slab);
			offslab_limit /= sizeof(kmem_bufctl_t);

 			if (num > offslab_limit)
				break;
		}
2008

2009
		/* Found something acceptable - save it away */
2010
		cachep->num = num;
2011
		cachep->gfporder = gfporder;
2012 2013
		left_over = remainder;

2014 2015 2016 2017 2018 2019 2020 2021
		/*
		 * A VFS-reclaimable slab tends to have most allocations
		 * as GFP_NOFS and we really don't want to have to be allocating
		 * higher-order pages when we are unable to shrink dcache.
		 */
		if (flags & SLAB_RECLAIM_ACCOUNT)
			break;

2022 2023 2024 2025
		/*
		 * Large number of objects is good, but very large slabs are
		 * currently bad for the gfp()s.
		 */
2026
		if (gfporder >= slab_break_gfp_order)
2027 2028
			break;

2029 2030 2031
		/*
		 * Acceptable internal fragmentation?
		 */
A
Andrew Morton 已提交
2032
		if (left_over * 8 <= (PAGE_SIZE << gfporder))
2033 2034 2035 2036 2037
			break;
	}
	return left_over;
}

2038
static int setup_cpu_cache(struct kmem_cache *cachep)
2039
{
2040 2041 2042
	if (g_cpucache_up == FULL)
		return enable_cpucache(cachep);

2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088
	if (g_cpucache_up == NONE) {
		/*
		 * Note: the first kmem_cache_create must create the cache
		 * that's used by kmalloc(24), otherwise the creation of
		 * further caches will BUG().
		 */
		cachep->array[smp_processor_id()] = &initarray_generic.cache;

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

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

	cpu_cache_get(cachep)->avail = 0;
	cpu_cache_get(cachep)->limit = BOOT_CPUCACHE_ENTRIES;
	cpu_cache_get(cachep)->batchcount = 1;
	cpu_cache_get(cachep)->touched = 0;
	cachep->batchcount = 1;
	cachep->limit = BOOT_CPUCACHE_ENTRIES;
2089
	return 0;
2090 2091
}

L
Linus Torvalds 已提交
2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106
/**
 * kmem_cache_create - Create a cache.
 * @name: A string which is used in /proc/slabinfo to identify this cache.
 * @size: The size of objects to be created in this cache.
 * @align: The required alignment for the objects.
 * @flags: SLAB flags
 * @ctor: A constructor for the objects.
 * @dtor: A destructor for the objects.
 *
 * Returns a ptr to the cache on success, NULL on failure.
 * Cannot be called within a int, but can be interrupted.
 * The @ctor is run when new pages are allocated by the cache
 * and the @dtor is run before the pages are handed back.
 *
 * @name must be valid until the cache is destroyed. This implies that
A
Andrew Morton 已提交
2107 2108
 * the module calling this has to destroy the cache before getting unloaded.
 *
L
Linus Torvalds 已提交
2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120
 * The flags are
 *
 * %SLAB_POISON - Poison the slab with a known test pattern (a5a5a5a5)
 * to catch references to uninitialised memory.
 *
 * %SLAB_RED_ZONE - Insert `Red' zones around the allocated memory to check
 * for buffer overruns.
 *
 * %SLAB_HWCACHE_ALIGN - Align the objects in this cache to a hardware
 * cacheline.  This can be beneficial if you're counting cycles as closely
 * as davem.
 */
2121
struct kmem_cache *
L
Linus Torvalds 已提交
2122
kmem_cache_create (const char *name, size_t size, size_t align,
A
Andrew Morton 已提交
2123 2124
	unsigned long flags,
	void (*ctor)(void*, struct kmem_cache *, unsigned long),
2125
	void (*dtor)(void*, struct kmem_cache *, unsigned long))
L
Linus Torvalds 已提交
2126 2127
{
	size_t left_over, slab_size, ralign;
2128
	struct kmem_cache *cachep = NULL, *pc;
L
Linus Torvalds 已提交
2129 2130 2131 2132

	/*
	 * Sanity checks... these are all serious usage bugs.
	 */
A
Andrew Morton 已提交
2133
	if (!name || in_interrupt() || (size < BYTES_PER_WORD) ||
P
Pekka Enberg 已提交
2134
	    (size > (1 << MAX_OBJ_ORDER) * PAGE_SIZE) || (dtor && !ctor)) {
A
Andrew Morton 已提交
2135 2136
		printk(KERN_ERR "%s: Early error in slab %s\n", __FUNCTION__,
				name);
P
Pekka Enberg 已提交
2137 2138
		BUG();
	}
L
Linus Torvalds 已提交
2139

2140
	/*
2141 2142
	 * We use cache_chain_mutex to ensure a consistent view of
	 * cpu_online_map as well.  Please see cpuup_callback
2143
	 */
I
Ingo Molnar 已提交
2144
	mutex_lock(&cache_chain_mutex);
2145

2146
	list_for_each_entry(pc, &cache_chain, next) {
2147 2148 2149 2150 2151 2152 2153 2154
		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.
		 */
2155
		res = probe_kernel_address(pc->name, tmp);
2156 2157
		if (res) {
			printk("SLAB: cache with size %d has lost its name\n",
2158
			       pc->buffer_size);
2159 2160 2161
			continue;
		}

P
Pekka Enberg 已提交
2162
		if (!strcmp(pc->name, name)) {
2163 2164 2165 2166 2167 2168
			printk("kmem_cache_create: duplicate cache %s\n", name);
			dump_stack();
			goto oops;
		}
	}

L
Linus Torvalds 已提交
2169 2170 2171 2172 2173
#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
2174
		       "requested - %s\n", __FUNCTION__, name);
L
Linus Torvalds 已提交
2175 2176 2177 2178 2179 2180 2181 2182 2183
		flags &= ~SLAB_DEBUG_INITIAL;
	}
#if FORCED_DEBUG
	/*
	 * Enable redzoning and last user accounting, except for caches with
	 * large objects, if the increased size would increase the object size
	 * above the next power of two: caches with object sizes just above a
	 * power of two have a significant amount of internal fragmentation.
	 */
A
Andrew Morton 已提交
2184
	if (size < 4096 || fls(size - 1) == fls(size-1 + 3 * BYTES_PER_WORD))
P
Pekka Enberg 已提交
2185
		flags |= SLAB_RED_ZONE | SLAB_STORE_USER;
L
Linus Torvalds 已提交
2186 2187 2188 2189 2190 2191 2192 2193 2194 2195
	if (!(flags & SLAB_DESTROY_BY_RCU))
		flags |= SLAB_POISON;
#endif
	if (flags & SLAB_DESTROY_BY_RCU)
		BUG_ON(flags & SLAB_POISON);
#endif
	if (flags & SLAB_DESTROY_BY_RCU)
		BUG_ON(dtor);

	/*
A
Andrew Morton 已提交
2196 2197
	 * Always checks flags, a caller might be expecting debug support which
	 * isn't available.
L
Linus Torvalds 已提交
2198
	 */
2199
	BUG_ON(flags & ~CREATE_MASK);
L
Linus Torvalds 已提交
2200

A
Andrew Morton 已提交
2201 2202
	/*
	 * Check that size is in terms of words.  This is needed to avoid
L
Linus Torvalds 已提交
2203 2204 2205
	 * 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
2206 2207 2208
	if (size & (BYTES_PER_WORD - 1)) {
		size += (BYTES_PER_WORD - 1);
		size &= ~(BYTES_PER_WORD - 1);
L
Linus Torvalds 已提交
2209 2210
	}

A
Andrew Morton 已提交
2211 2212
	/* calculate the final buffer alignment: */

L
Linus Torvalds 已提交
2213 2214
	/* 1) arch recommendation: can be overridden for debug */
	if (flags & SLAB_HWCACHE_ALIGN) {
A
Andrew Morton 已提交
2215 2216 2217 2218
		/*
		 * Default alignment: as specified by the arch code.  Except if
		 * an object is really small, then squeeze multiple objects into
		 * one cacheline.
L
Linus Torvalds 已提交
2219 2220
		 */
		ralign = cache_line_size();
P
Pekka Enberg 已提交
2221
		while (size <= ralign / 2)
L
Linus Torvalds 已提交
2222 2223 2224 2225
			ralign /= 2;
	} else {
		ralign = BYTES_PER_WORD;
	}
2226 2227 2228 2229 2230 2231 2232 2233 2234

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

2235
	/* 2) arch mandated alignment */
L
Linus Torvalds 已提交
2236 2237 2238
	if (ralign < ARCH_SLAB_MINALIGN) {
		ralign = ARCH_SLAB_MINALIGN;
	}
2239
	/* 3) caller mandated alignment */
L
Linus Torvalds 已提交
2240 2241 2242
	if (ralign < align) {
		ralign = align;
	}
2243 2244 2245
	/* disable debug if necessary */
	if (ralign > BYTES_PER_WORD)
		flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
A
Andrew Morton 已提交
2246
	/*
2247
	 * 4) Store it.
L
Linus Torvalds 已提交
2248 2249 2250 2251
	 */
	align = ralign;

	/* Get cache's description obj. */
2252
	cachep = kmem_cache_zalloc(&cache_cache, GFP_KERNEL);
L
Linus Torvalds 已提交
2253
	if (!cachep)
2254
		goto oops;
L
Linus Torvalds 已提交
2255 2256

#if DEBUG
2257
	cachep->obj_size = size;
L
Linus Torvalds 已提交
2258

2259 2260 2261 2262
	/*
	 * Both debugging options require word-alignment which is calculated
	 * into align above.
	 */
L
Linus Torvalds 已提交
2263 2264
	if (flags & SLAB_RED_ZONE) {
		/* add space for red zone words */
2265
		cachep->obj_offset += BYTES_PER_WORD;
P
Pekka Enberg 已提交
2266
		size += 2 * BYTES_PER_WORD;
L
Linus Torvalds 已提交
2267 2268
	}
	if (flags & SLAB_STORE_USER) {
2269 2270
		/* user store requires one word storage behind the end of
		 * the real object.
L
Linus Torvalds 已提交
2271 2272 2273 2274
		 */
		size += BYTES_PER_WORD;
	}
#if FORCED_DEBUG && defined(CONFIG_DEBUG_PAGEALLOC)
P
Pekka Enberg 已提交
2275
	if (size >= malloc_sizes[INDEX_L3 + 1].cs_size
2276 2277
	    && cachep->obj_size > cache_line_size() && size < PAGE_SIZE) {
		cachep->obj_offset += PAGE_SIZE - size;
L
Linus Torvalds 已提交
2278 2279 2280 2281 2282
		size = PAGE_SIZE;
	}
#endif
#endif

2283 2284 2285 2286 2287 2288
	/*
	 * Determine if the slab management is 'on' or 'off' slab.
	 * (bootstrapping cannot cope with offslab caches so don't do
	 * it too early on.)
	 */
	if ((size >= (PAGE_SIZE >> 3)) && !slab_early_init)
L
Linus Torvalds 已提交
2289 2290 2291 2292 2293 2294 2295 2296
		/*
		 * 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);

2297
	left_over = calculate_slab_order(cachep, size, align, flags);
L
Linus Torvalds 已提交
2298 2299 2300 2301 2302

	if (!cachep->num) {
		printk("kmem_cache_create: couldn't create cache %s.\n", name);
		kmem_cache_free(&cache_cache, cachep);
		cachep = NULL;
2303
		goto oops;
L
Linus Torvalds 已提交
2304
	}
P
Pekka Enberg 已提交
2305 2306
	slab_size = ALIGN(cachep->num * sizeof(kmem_bufctl_t)
			  + sizeof(struct slab), align);
L
Linus Torvalds 已提交
2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318

	/*
	 * 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
2319 2320
		slab_size =
		    cachep->num * sizeof(kmem_bufctl_t) + sizeof(struct slab);
L
Linus Torvalds 已提交
2321 2322 2323 2324 2325 2326
	}

	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
2327
	cachep->colour = left_over / cachep->colour_off;
L
Linus Torvalds 已提交
2328 2329 2330
	cachep->slab_size = slab_size;
	cachep->flags = flags;
	cachep->gfpflags = 0;
2331
	if (CONFIG_ZONE_DMA_FLAG && (flags & SLAB_CACHE_DMA))
L
Linus Torvalds 已提交
2332
		cachep->gfpflags |= GFP_DMA;
2333
	cachep->buffer_size = size;
2334
	cachep->reciprocal_buffer_size = reciprocal_value(size);
L
Linus Torvalds 已提交
2335

2336
	if (flags & CFLGS_OFF_SLAB) {
2337
		cachep->slabp_cache = kmem_find_general_cachep(slab_size, 0u);
2338 2339 2340 2341 2342 2343 2344 2345 2346
		/*
		 * This is a possibility for one of the malloc_sizes caches.
		 * But since we go off slab only for object size greater than
		 * PAGE_SIZE/8, and malloc_sizes gets created in ascending order,
		 * this should not happen at all.
		 * But leave a BUG_ON for some lucky dude.
		 */
		BUG_ON(!cachep->slabp_cache);
	}
L
Linus Torvalds 已提交
2347 2348 2349 2350
	cachep->ctor = ctor;
	cachep->dtor = dtor;
	cachep->name = name;

2351 2352 2353 2354 2355
	if (setup_cpu_cache(cachep)) {
		__kmem_cache_destroy(cachep);
		cachep = NULL;
		goto oops;
	}
L
Linus Torvalds 已提交
2356 2357 2358

	/* cache setup completed, link it into the list */
	list_add(&cachep->next, &cache_chain);
A
Andrew Morton 已提交
2359
oops:
L
Linus Torvalds 已提交
2360 2361
	if (!cachep && (flags & SLAB_PANIC))
		panic("kmem_cache_create(): failed to create slab `%s'\n",
P
Pekka Enberg 已提交
2362
		      name);
I
Ingo Molnar 已提交
2363
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378
	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());
}

2379
static void check_spinlock_acquired(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2380 2381 2382
{
#ifdef CONFIG_SMP
	check_irq_off();
2383
	assert_spin_locked(&cachep->nodelists[numa_node_id()]->list_lock);
L
Linus Torvalds 已提交
2384 2385
#endif
}
2386

2387
static void check_spinlock_acquired_node(struct kmem_cache *cachep, int node)
2388 2389 2390 2391 2392 2393 2394
{
#ifdef CONFIG_SMP
	check_irq_off();
	assert_spin_locked(&cachep->nodelists[node]->list_lock);
#endif
}

L
Linus Torvalds 已提交
2395 2396 2397 2398
#else
#define check_irq_off()	do { } while(0)
#define check_irq_on()	do { } while(0)
#define check_spinlock_acquired(x) do { } while(0)
2399
#define check_spinlock_acquired_node(x, y) do { } while(0)
L
Linus Torvalds 已提交
2400 2401
#endif

2402 2403 2404 2405
static void drain_array(struct kmem_cache *cachep, struct kmem_list3 *l3,
			struct array_cache *ac,
			int force, int node);

L
Linus Torvalds 已提交
2406 2407
static void do_drain(void *arg)
{
A
Andrew Morton 已提交
2408
	struct kmem_cache *cachep = arg;
L
Linus Torvalds 已提交
2409
	struct array_cache *ac;
2410
	int node = numa_node_id();
L
Linus Torvalds 已提交
2411 2412

	check_irq_off();
2413
	ac = cpu_cache_get(cachep);
2414 2415 2416
	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
2417 2418 2419
	ac->avail = 0;
}

2420
static void drain_cpu_caches(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2421
{
2422 2423 2424
	struct kmem_list3 *l3;
	int node;

A
Andrew Morton 已提交
2425
	on_each_cpu(do_drain, cachep, 1, 1);
L
Linus Torvalds 已提交
2426
	check_irq_on();
P
Pekka Enberg 已提交
2427
	for_each_online_node(node) {
2428
		l3 = cachep->nodelists[node];
2429 2430 2431 2432 2433 2434 2435
		if (l3 && l3->alien)
			drain_alien_cache(cachep, l3->alien);
	}

	for_each_online_node(node) {
		l3 = cachep->nodelists[node];
		if (l3)
2436
			drain_array(cachep, l3, l3->shared, 1, node);
2437
	}
L
Linus Torvalds 已提交
2438 2439
}

2440 2441 2442 2443 2444 2445 2446 2447
/*
 * Remove slabs from the list of free slabs.
 * Specify the number of slabs to drain in tofree.
 *
 * Returns the actual number of slabs released.
 */
static int drain_freelist(struct kmem_cache *cache,
			struct kmem_list3 *l3, int tofree)
L
Linus Torvalds 已提交
2448
{
2449 2450
	struct list_head *p;
	int nr_freed;
L
Linus Torvalds 已提交
2451 2452
	struct slab *slabp;

2453 2454
	nr_freed = 0;
	while (nr_freed < tofree && !list_empty(&l3->slabs_free)) {
L
Linus Torvalds 已提交
2455

2456
		spin_lock_irq(&l3->list_lock);
2457
		p = l3->slabs_free.prev;
2458 2459 2460 2461
		if (p == &l3->slabs_free) {
			spin_unlock_irq(&l3->list_lock);
			goto out;
		}
L
Linus Torvalds 已提交
2462

2463
		slabp = list_entry(p, struct slab, list);
L
Linus Torvalds 已提交
2464
#if DEBUG
2465
		BUG_ON(slabp->inuse);
L
Linus Torvalds 已提交
2466 2467
#endif
		list_del(&slabp->list);
2468 2469 2470 2471 2472
		/*
		 * Safe to drop the lock. The slab is no longer linked
		 * to the cache.
		 */
		l3->free_objects -= cache->num;
2473
		spin_unlock_irq(&l3->list_lock);
2474 2475
		slab_destroy(cache, slabp);
		nr_freed++;
L
Linus Torvalds 已提交
2476
	}
2477 2478
out:
	return nr_freed;
L
Linus Torvalds 已提交
2479 2480
}

2481
/* Called with cache_chain_mutex held to protect against cpu hotplug */
2482
static int __cache_shrink(struct kmem_cache *cachep)
2483 2484 2485 2486 2487 2488 2489 2490 2491
{
	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];
2492 2493 2494 2495 2496 2497 2498
		if (!l3)
			continue;

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

		ret += !list_empty(&l3->slabs_full) ||
			!list_empty(&l3->slabs_partial);
2499 2500 2501 2502
	}
	return (ret ? 1 : 0);
}

L
Linus Torvalds 已提交
2503 2504 2505 2506 2507 2508 2509
/**
 * 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.
 */
2510
int kmem_cache_shrink(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2511
{
2512
	int ret;
2513
	BUG_ON(!cachep || in_interrupt());
L
Linus Torvalds 已提交
2514

2515 2516 2517 2518
	mutex_lock(&cache_chain_mutex);
	ret = __cache_shrink(cachep);
	mutex_unlock(&cache_chain_mutex);
	return ret;
L
Linus Torvalds 已提交
2519 2520 2521 2522 2523 2524 2525
}
EXPORT_SYMBOL(kmem_cache_shrink);

/**
 * kmem_cache_destroy - delete a cache
 * @cachep: the cache to destroy
 *
2526
 * Remove a &struct kmem_cache object from the slab cache.
L
Linus Torvalds 已提交
2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537
 *
 * 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().
 */
2538
void kmem_cache_destroy(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2539
{
2540
	BUG_ON(!cachep || in_interrupt());
L
Linus Torvalds 已提交
2541 2542

	/* Find the cache in the chain of caches. */
I
Ingo Molnar 已提交
2543
	mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2544 2545 2546 2547 2548 2549
	/*
	 * the chain is never empty, cache_cache is never destroyed
	 */
	list_del(&cachep->next);
	if (__cache_shrink(cachep)) {
		slab_error(cachep, "Can't free all objects");
P
Pekka Enberg 已提交
2550
		list_add(&cachep->next, &cache_chain);
I
Ingo Molnar 已提交
2551
		mutex_unlock(&cache_chain_mutex);
2552
		return;
L
Linus Torvalds 已提交
2553 2554 2555
	}

	if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU))
2556
		synchronize_rcu();
L
Linus Torvalds 已提交
2557

2558
	__kmem_cache_destroy(cachep);
2559
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2560 2561 2562
}
EXPORT_SYMBOL(kmem_cache_destroy);

2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573
/*
 * Get the memory for a slab management obj.
 * For a slab cache when the slab descriptor is off-slab, slab descriptors
 * always come from malloc_sizes caches.  The slab descriptor cannot
 * come from the same cache which is getting created because,
 * when we are searching for an appropriate cache for these
 * descriptors in kmem_cache_create, we search through the malloc_sizes array.
 * If we are creating a malloc_sizes cache here it would not be visible to
 * kmem_find_general_cachep till the initialization is complete.
 * Hence we cannot have slabp_cache same as the original cache.
 */
2574
static struct slab *alloc_slabmgmt(struct kmem_cache *cachep, void *objp,
2575 2576
				   int colour_off, gfp_t local_flags,
				   int nodeid)
L
Linus Torvalds 已提交
2577 2578
{
	struct slab *slabp;
P
Pekka Enberg 已提交
2579

L
Linus Torvalds 已提交
2580 2581
	if (OFF_SLAB(cachep)) {
		/* Slab management obj is off-slab. */
2582
		slabp = kmem_cache_alloc_node(cachep->slabp_cache,
2583
					      local_flags & ~GFP_THISNODE, nodeid);
L
Linus Torvalds 已提交
2584 2585 2586
		if (!slabp)
			return NULL;
	} else {
P
Pekka Enberg 已提交
2587
		slabp = objp + colour_off;
L
Linus Torvalds 已提交
2588 2589 2590 2591
		colour_off += cachep->slab_size;
	}
	slabp->inuse = 0;
	slabp->colouroff = colour_off;
P
Pekka Enberg 已提交
2592
	slabp->s_mem = objp + colour_off;
2593
	slabp->nodeid = nodeid;
L
Linus Torvalds 已提交
2594 2595 2596 2597 2598
	return slabp;
}

static inline kmem_bufctl_t *slab_bufctl(struct slab *slabp)
{
P
Pekka Enberg 已提交
2599
	return (kmem_bufctl_t *) (slabp + 1);
L
Linus Torvalds 已提交
2600 2601
}

2602
static void cache_init_objs(struct kmem_cache *cachep,
P
Pekka Enberg 已提交
2603
			    struct slab *slabp, unsigned long ctor_flags)
L
Linus Torvalds 已提交
2604 2605 2606 2607
{
	int i;

	for (i = 0; i < cachep->num; i++) {
2608
		void *objp = index_to_obj(cachep, slabp, i);
L
Linus Torvalds 已提交
2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620
#if DEBUG
		/* need to poison the objs? */
		if (cachep->flags & SLAB_POISON)
			poison_obj(cachep, objp, POISON_FREE);
		if (cachep->flags & SLAB_STORE_USER)
			*dbg_userword(cachep, objp) = NULL;

		if (cachep->flags & SLAB_RED_ZONE) {
			*dbg_redzone1(cachep, objp) = RED_INACTIVE;
			*dbg_redzone2(cachep, objp) = RED_INACTIVE;
		}
		/*
A
Andrew Morton 已提交
2621 2622 2623
		 * Constructors are not allowed to allocate memory from the same
		 * cache which they are a constructor for.  Otherwise, deadlock.
		 * They must also be threaded.
L
Linus Torvalds 已提交
2624 2625
		 */
		if (cachep->ctor && !(cachep->flags & SLAB_POISON))
2626
			cachep->ctor(objp + obj_offset(cachep), cachep,
P
Pekka Enberg 已提交
2627
				     ctor_flags);
L
Linus Torvalds 已提交
2628 2629 2630 2631

		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
2632
					   " end of an object");
L
Linus Torvalds 已提交
2633 2634
			if (*dbg_redzone1(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "constructor overwrote the"
P
Pekka Enberg 已提交
2635
					   " start of an object");
L
Linus Torvalds 已提交
2636
		}
A
Andrew Morton 已提交
2637 2638
		if ((cachep->buffer_size % PAGE_SIZE) == 0 &&
			    OFF_SLAB(cachep) && cachep->flags & SLAB_POISON)
P
Pekka Enberg 已提交
2639
			kernel_map_pages(virt_to_page(objp),
2640
					 cachep->buffer_size / PAGE_SIZE, 0);
L
Linus Torvalds 已提交
2641 2642 2643 2644
#else
		if (cachep->ctor)
			cachep->ctor(objp, cachep, ctor_flags);
#endif
P
Pekka Enberg 已提交
2645
		slab_bufctl(slabp)[i] = i + 1;
L
Linus Torvalds 已提交
2646
	}
P
Pekka Enberg 已提交
2647
	slab_bufctl(slabp)[i - 1] = BUFCTL_END;
L
Linus Torvalds 已提交
2648 2649 2650
	slabp->free = 0;
}

2651
static void kmem_flagcheck(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2652
{
2653 2654 2655 2656 2657 2658
	if (CONFIG_ZONE_DMA_FLAG) {
		if (flags & GFP_DMA)
			BUG_ON(!(cachep->gfpflags & GFP_DMA));
		else
			BUG_ON(cachep->gfpflags & GFP_DMA);
	}
L
Linus Torvalds 已提交
2659 2660
}

A
Andrew Morton 已提交
2661 2662
static void *slab_get_obj(struct kmem_cache *cachep, struct slab *slabp,
				int nodeid)
2663
{
2664
	void *objp = index_to_obj(cachep, slabp, slabp->free);
2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677
	kmem_bufctl_t next;

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

	return objp;
}

A
Andrew Morton 已提交
2678 2679
static void slab_put_obj(struct kmem_cache *cachep, struct slab *slabp,
				void *objp, int nodeid)
2680
{
2681
	unsigned int objnr = obj_to_index(cachep, slabp, objp);
2682 2683 2684 2685 2686

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

2687
	if (slab_bufctl(slabp)[objnr] + 1 <= SLAB_LIMIT + 1) {
2688
		printk(KERN_ERR "slab: double free detected in cache "
A
Andrew Morton 已提交
2689
				"'%s', objp %p\n", cachep->name, objp);
2690 2691 2692 2693 2694 2695 2696 2697
		BUG();
	}
#endif
	slab_bufctl(slabp)[objnr] = slabp->free;
	slabp->free = objnr;
	slabp->inuse--;
}

2698 2699 2700 2701 2702 2703 2704
/*
 * Map pages beginning at addr to the given cache and slab. This is required
 * for the slab allocator to be able to lookup the cache and slab of a
 * virtual address for kfree, ksize, kmem_ptr_validate, and slab debugging.
 */
static void slab_map_pages(struct kmem_cache *cache, struct slab *slab,
			   void *addr)
L
Linus Torvalds 已提交
2705
{
2706
	int nr_pages;
L
Linus Torvalds 已提交
2707 2708
	struct page *page;

2709
	page = virt_to_page(addr);
2710

2711
	nr_pages = 1;
2712
	if (likely(!PageCompound(page)))
2713 2714
		nr_pages <<= cache->gfporder;

L
Linus Torvalds 已提交
2715
	do {
2716 2717
		page_set_cache(page, cache);
		page_set_slab(page, slab);
L
Linus Torvalds 已提交
2718
		page++;
2719
	} while (--nr_pages);
L
Linus Torvalds 已提交
2720 2721 2722 2723 2724 2725
}

/*
 * 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.
 */
2726 2727
static int cache_grow(struct kmem_cache *cachep,
		gfp_t flags, int nodeid, void *objp)
L
Linus Torvalds 已提交
2728
{
P
Pekka Enberg 已提交
2729 2730 2731 2732
	struct slab *slabp;
	size_t offset;
	gfp_t local_flags;
	unsigned long ctor_flags;
2733
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
2734

A
Andrew Morton 已提交
2735 2736 2737
	/*
	 * Be lazy and only check for valid flags here,  keeping it out of the
	 * critical path in kmem_cache_alloc().
L
Linus Torvalds 已提交
2738
	 */
C
Christoph Lameter 已提交
2739
	BUG_ON(flags & ~(GFP_DMA | GFP_LEVEL_MASK | __GFP_NO_GROW));
2740
	if (flags & __GFP_NO_GROW)
L
Linus Torvalds 已提交
2741 2742 2743
		return 0;

	ctor_flags = SLAB_CTOR_CONSTRUCTOR;
2744
	local_flags = (flags & GFP_LEVEL_MASK);
L
Linus Torvalds 已提交
2745 2746 2747 2748 2749 2750 2751
	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;

2752
	/* Take the l3 list lock to change the colour_next on this node */
L
Linus Torvalds 已提交
2753
	check_irq_off();
2754 2755
	l3 = cachep->nodelists[nodeid];
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2756 2757

	/* Get colour for the slab, and cal the next value. */
2758 2759 2760 2761 2762
	offset = l3->colour_next;
	l3->colour_next++;
	if (l3->colour_next >= cachep->colour)
		l3->colour_next = 0;
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2763

2764
	offset *= cachep->colour_off;
L
Linus Torvalds 已提交
2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776

	if (local_flags & __GFP_WAIT)
		local_irq_enable();

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

A
Andrew Morton 已提交
2777 2778 2779
	/*
	 * Get mem for the objs.  Attempt to allocate a physical page from
	 * 'nodeid'.
2780
	 */
2781 2782
	if (!objp)
		objp = kmem_getpages(cachep, flags, nodeid);
A
Andrew Morton 已提交
2783
	if (!objp)
L
Linus Torvalds 已提交
2784 2785 2786
		goto failed;

	/* Get slab management. */
2787 2788
	slabp = alloc_slabmgmt(cachep, objp, offset,
			local_flags & ~GFP_THISNODE, nodeid);
A
Andrew Morton 已提交
2789
	if (!slabp)
L
Linus Torvalds 已提交
2790 2791
		goto opps1;

2792
	slabp->nodeid = nodeid;
2793
	slab_map_pages(cachep, slabp, objp);
L
Linus Torvalds 已提交
2794 2795 2796 2797 2798 2799

	cache_init_objs(cachep, slabp, ctor_flags);

	if (local_flags & __GFP_WAIT)
		local_irq_disable();
	check_irq_off();
2800
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2801 2802

	/* Make slab active. */
2803
	list_add_tail(&slabp->list, &(l3->slabs_free));
L
Linus Torvalds 已提交
2804
	STATS_INC_GROWN(cachep);
2805 2806
	l3->free_objects += cachep->num;
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2807
	return 1;
A
Andrew Morton 已提交
2808
opps1:
L
Linus Torvalds 已提交
2809
	kmem_freepages(cachep, objp);
A
Andrew Morton 已提交
2810
failed:
L
Linus Torvalds 已提交
2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827
	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)
{
	if (!virt_addr_valid(objp)) {
		printk(KERN_ERR "kfree_debugcheck: out of range ptr %lxh.\n",
P
Pekka Enberg 已提交
2828 2829
		       (unsigned long)objp);
		BUG();
L
Linus Torvalds 已提交
2830 2831 2832
	}
}

2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854
static inline void verify_redzone_free(struct kmem_cache *cache, void *obj)
{
	unsigned long redzone1, redzone2;

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

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

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

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

2855
static void *cache_free_debugcheck(struct kmem_cache *cachep, void *objp,
P
Pekka Enberg 已提交
2856
				   void *caller)
L
Linus Torvalds 已提交
2857 2858 2859 2860 2861
{
	struct page *page;
	unsigned int objnr;
	struct slab *slabp;

2862
	objp -= obj_offset(cachep);
L
Linus Torvalds 已提交
2863 2864 2865
	kfree_debugcheck(objp);
	page = virt_to_page(objp);

2866
	slabp = page_get_slab(page);
L
Linus Torvalds 已提交
2867 2868

	if (cachep->flags & SLAB_RED_ZONE) {
2869
		verify_redzone_free(cachep, objp);
L
Linus Torvalds 已提交
2870 2871 2872 2873 2874 2875
		*dbg_redzone1(cachep, objp) = RED_INACTIVE;
		*dbg_redzone2(cachep, objp) = RED_INACTIVE;
	}
	if (cachep->flags & SLAB_STORE_USER)
		*dbg_userword(cachep, objp) = caller;

2876
	objnr = obj_to_index(cachep, slabp, objp);
L
Linus Torvalds 已提交
2877 2878

	BUG_ON(objnr >= cachep->num);
2879
	BUG_ON(objp != index_to_obj(cachep, slabp, objnr));
L
Linus Torvalds 已提交
2880 2881

	if (cachep->flags & SLAB_DEBUG_INITIAL) {
A
Andrew Morton 已提交
2882 2883 2884 2885
		/*
		 * Need to call the slab's constructor so the caller can
		 * perform a verify of its state (debugging).  Called without
		 * the cache-lock held.
L
Linus Torvalds 已提交
2886
		 */
2887
		cachep->ctor(objp + obj_offset(cachep),
P
Pekka Enberg 已提交
2888
			     cachep, SLAB_CTOR_CONSTRUCTOR | SLAB_CTOR_VERIFY);
L
Linus Torvalds 已提交
2889 2890 2891 2892 2893
	}
	if (cachep->flags & SLAB_POISON && cachep->dtor) {
		/* we want to cache poison the object,
		 * call the destruction callback
		 */
2894
		cachep->dtor(objp + obj_offset(cachep), cachep, 0);
L
Linus Torvalds 已提交
2895
	}
2896 2897 2898
#ifdef CONFIG_DEBUG_SLAB_LEAK
	slab_bufctl(slabp)[objnr] = BUFCTL_FREE;
#endif
L
Linus Torvalds 已提交
2899 2900
	if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
A
Andrew Morton 已提交
2901
		if ((cachep->buffer_size % PAGE_SIZE)==0 && OFF_SLAB(cachep)) {
L
Linus Torvalds 已提交
2902
			store_stackinfo(cachep, objp, (unsigned long)caller);
P
Pekka Enberg 已提交
2903
			kernel_map_pages(virt_to_page(objp),
2904
					 cachep->buffer_size / PAGE_SIZE, 0);
L
Linus Torvalds 已提交
2905 2906 2907 2908 2909 2910 2911 2912 2913 2914
		} else {
			poison_obj(cachep, objp, POISON_FREE);
		}
#else
		poison_obj(cachep, objp, POISON_FREE);
#endif
	}
	return objp;
}

2915
static void check_slabp(struct kmem_cache *cachep, struct slab *slabp)
L
Linus Torvalds 已提交
2916 2917 2918
{
	kmem_bufctl_t i;
	int entries = 0;
P
Pekka Enberg 已提交
2919

L
Linus Torvalds 已提交
2920 2921 2922 2923 2924 2925 2926
	/* Check slab's freelist to see if this obj is there. */
	for (i = slabp->free; i != BUFCTL_END; i = slab_bufctl(slabp)[i]) {
		entries++;
		if (entries > cachep->num || i >= cachep->num)
			goto bad;
	}
	if (entries != cachep->num - slabp->inuse) {
A
Andrew Morton 已提交
2927 2928 2929 2930
bad:
		printk(KERN_ERR "slab: Internal list corruption detected in "
				"cache '%s'(%d), slabp %p(%d). Hexdump:\n",
			cachep->name, cachep->num, slabp, slabp->inuse);
P
Pekka Enberg 已提交
2931
		for (i = 0;
2932
		     i < sizeof(*slabp) + cachep->num * sizeof(kmem_bufctl_t);
P
Pekka Enberg 已提交
2933
		     i++) {
A
Andrew Morton 已提交
2934
			if (i % 16 == 0)
L
Linus Torvalds 已提交
2935
				printk("\n%03x:", i);
P
Pekka Enberg 已提交
2936
			printk(" %02x", ((unsigned char *)slabp)[i]);
L
Linus Torvalds 已提交
2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947
		}
		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

2948
static void *cache_alloc_refill(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2949 2950 2951 2952
{
	int batchcount;
	struct kmem_list3 *l3;
	struct array_cache *ac;
P
Pekka Enberg 已提交
2953 2954 2955
	int node;

	node = numa_node_id();
L
Linus Torvalds 已提交
2956 2957

	check_irq_off();
2958
	ac = cpu_cache_get(cachep);
A
Andrew Morton 已提交
2959
retry:
L
Linus Torvalds 已提交
2960 2961
	batchcount = ac->batchcount;
	if (!ac->touched && batchcount > BATCHREFILL_LIMIT) {
A
Andrew Morton 已提交
2962 2963 2964 2965
		/*
		 * If there was little recent activity on this cache, then
		 * perform only a partial refill.  Otherwise we could generate
		 * refill bouncing.
L
Linus Torvalds 已提交
2966 2967 2968
		 */
		batchcount = BATCHREFILL_LIMIT;
	}
P
Pekka Enberg 已提交
2969
	l3 = cachep->nodelists[node];
2970 2971 2972

	BUG_ON(ac->avail > 0 || !l3);
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2973

2974 2975 2976 2977
	/* See if we can refill from the shared array */
	if (l3->shared && transfer_objects(ac, l3->shared, batchcount))
		goto alloc_done;

L
Linus Torvalds 已提交
2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992
	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);
2993 2994 2995 2996 2997 2998 2999 3000

		/*
		 * The slab was either on partial or free list so
		 * there must be at least one object available for
		 * allocation.
		 */
		BUG_ON(slabp->inuse < 0 || slabp->inuse >= cachep->num);

L
Linus Torvalds 已提交
3001 3002 3003 3004 3005
		while (slabp->inuse < cachep->num && batchcount--) {
			STATS_INC_ALLOCED(cachep);
			STATS_INC_ACTIVE(cachep);
			STATS_SET_HIGH(cachep);

3006
			ac->entry[ac->avail++] = slab_get_obj(cachep, slabp,
P
Pekka Enberg 已提交
3007
							    node);
L
Linus Torvalds 已提交
3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018
		}
		check_slabp(cachep, slabp);

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

A
Andrew Morton 已提交
3019
must_grow:
L
Linus Torvalds 已提交
3020
	l3->free_objects -= ac->avail;
A
Andrew Morton 已提交
3021
alloc_done:
3022
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
3023 3024 3025

	if (unlikely(!ac->avail)) {
		int x;
3026
		x = cache_grow(cachep, flags | GFP_THISNODE, node, NULL);
3027

A
Andrew Morton 已提交
3028
		/* cache_grow can reenable interrupts, then ac could change. */
3029
		ac = cpu_cache_get(cachep);
A
Andrew Morton 已提交
3030
		if (!x && ac->avail == 0)	/* no objects in sight? abort */
L
Linus Torvalds 已提交
3031 3032
			return NULL;

A
Andrew Morton 已提交
3033
		if (!ac->avail)		/* objects refilled by interrupt? */
L
Linus Torvalds 已提交
3034 3035 3036
			goto retry;
	}
	ac->touched = 1;
3037
	return ac->entry[--ac->avail];
L
Linus Torvalds 已提交
3038 3039
}

A
Andrew Morton 已提交
3040 3041
static inline void cache_alloc_debugcheck_before(struct kmem_cache *cachep,
						gfp_t flags)
L
Linus Torvalds 已提交
3042 3043 3044 3045 3046 3047 3048 3049
{
	might_sleep_if(flags & __GFP_WAIT);
#if DEBUG
	kmem_flagcheck(cachep, flags);
#endif
}

#if DEBUG
A
Andrew Morton 已提交
3050 3051
static void *cache_alloc_debugcheck_after(struct kmem_cache *cachep,
				gfp_t flags, void *objp, void *caller)
L
Linus Torvalds 已提交
3052
{
P
Pekka Enberg 已提交
3053
	if (!objp)
L
Linus Torvalds 已提交
3054
		return objp;
P
Pekka Enberg 已提交
3055
	if (cachep->flags & SLAB_POISON) {
L
Linus Torvalds 已提交
3056
#ifdef CONFIG_DEBUG_PAGEALLOC
3057
		if ((cachep->buffer_size % PAGE_SIZE) == 0 && OFF_SLAB(cachep))
P
Pekka Enberg 已提交
3058
			kernel_map_pages(virt_to_page(objp),
3059
					 cachep->buffer_size / PAGE_SIZE, 1);
L
Linus Torvalds 已提交
3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070
		else
			check_poison_obj(cachep, objp);
#else
		check_poison_obj(cachep, objp);
#endif
		poison_obj(cachep, objp, POISON_INUSE);
	}
	if (cachep->flags & SLAB_STORE_USER)
		*dbg_userword(cachep, objp) = caller;

	if (cachep->flags & SLAB_RED_ZONE) {
A
Andrew Morton 已提交
3071 3072 3073 3074
		if (*dbg_redzone1(cachep, objp) != RED_INACTIVE ||
				*dbg_redzone2(cachep, objp) != RED_INACTIVE) {
			slab_error(cachep, "double free, or memory outside"
						" object was overwritten");
P
Pekka Enberg 已提交
3075
			printk(KERN_ERR
A
Andrew Morton 已提交
3076 3077 3078
				"%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
3079 3080 3081 3082
		}
		*dbg_redzone1(cachep, objp) = RED_ACTIVE;
		*dbg_redzone2(cachep, objp) = RED_ACTIVE;
	}
3083 3084 3085 3086 3087 3088 3089 3090 3091 3092
#ifdef CONFIG_DEBUG_SLAB_LEAK
	{
		struct slab *slabp;
		unsigned objnr;

		slabp = page_get_slab(virt_to_page(objp));
		objnr = (unsigned)(objp - slabp->s_mem) / cachep->buffer_size;
		slab_bufctl(slabp)[objnr] = BUFCTL_ACTIVE;
	}
#endif
3093
	objp += obj_offset(cachep);
L
Linus Torvalds 已提交
3094
	if (cachep->ctor && cachep->flags & SLAB_POISON) {
P
Pekka Enberg 已提交
3095
		unsigned long ctor_flags = SLAB_CTOR_CONSTRUCTOR;
L
Linus Torvalds 已提交
3096 3097 3098 3099 3100

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

		cachep->ctor(objp, cachep, ctor_flags);
P
Pekka Enberg 已提交
3101
	}
3102 3103 3104 3105 3106 3107
#if ARCH_SLAB_MINALIGN
	if ((u32)objp & (ARCH_SLAB_MINALIGN-1)) {
		printk(KERN_ERR "0x%p: not aligned to ARCH_SLAB_MINALIGN=%d\n",
		       objp, ARCH_SLAB_MINALIGN);
	}
#endif
L
Linus Torvalds 已提交
3108 3109 3110 3111 3112 3113
	return objp;
}
#else
#define cache_alloc_debugcheck_after(a,b,objp,d) (objp)
#endif

3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126
#ifdef CONFIG_FAILSLAB

static struct failslab_attr {

	struct fault_attr attr;

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

} failslab = {
	.attr = FAULT_ATTR_INITIALIZER,
3127
	.ignore_gfp_wait = 1,
3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186
};

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

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

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

#ifdef CONFIG_FAULT_INJECTION_DEBUG_FS

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

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

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

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

	return err;
}

late_initcall(failslab_debugfs);

#endif /* CONFIG_FAULT_INJECTION_DEBUG_FS */

#else /* CONFIG_FAILSLAB */

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

#endif /* CONFIG_FAILSLAB */

3187
static inline void *____cache_alloc(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
3188
{
P
Pekka Enberg 已提交
3189
	void *objp;
L
Linus Torvalds 已提交
3190 3191
	struct array_cache *ac;

3192
	check_irq_off();
3193 3194 3195 3196

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

3197
	ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
3198 3199 3200
	if (likely(ac->avail)) {
		STATS_INC_ALLOCHIT(cachep);
		ac->touched = 1;
3201
		objp = ac->entry[--ac->avail];
L
Linus Torvalds 已提交
3202 3203 3204 3205
	} else {
		STATS_INC_ALLOCMISS(cachep);
		objp = cache_alloc_refill(cachep, flags);
	}
3206 3207 3208
	return objp;
}

3209
#ifdef CONFIG_NUMA
3210
/*
3211
 * Try allocating on another node if PF_SPREAD_SLAB|PF_MEMPOLICY.
3212 3213 3214 3215 3216 3217 3218 3219
 *
 * If we are in_interrupt, then process context, including cpusets and
 * mempolicy, may not apply and should not be used for allocation policy.
 */
static void *alternate_node_alloc(struct kmem_cache *cachep, gfp_t flags)
{
	int nid_alloc, nid_here;

3220
	if (in_interrupt() || (flags & __GFP_THISNODE))
3221 3222 3223 3224 3225 3226 3227
		return NULL;
	nid_alloc = nid_here = numa_node_id();
	if (cpuset_do_slab_mem_spread() && (cachep->flags & SLAB_MEM_SPREAD))
		nid_alloc = cpuset_mem_spread_node();
	else if (current->mempolicy)
		nid_alloc = slab_node(current->mempolicy);
	if (nid_alloc != nid_here)
3228
		return ____cache_alloc_node(cachep, flags, nid_alloc);
3229 3230 3231
	return NULL;
}

3232 3233
/*
 * Fallback function if there was no memory available and no objects on a
3234 3235 3236 3237 3238
 * certain node and fall back is permitted. First we scan all the
 * available nodelists for available objects. If that fails then we
 * perform an allocation without specifying a node. This allows the page
 * allocator to do its reclaim / fallback magic. We then insert the
 * slab into the proper nodelist and then allocate from it.
3239
 */
3240
static void *fallback_alloc(struct kmem_cache *cache, gfp_t flags)
3241
{
3242 3243
	struct zonelist *zonelist;
	gfp_t local_flags;
3244 3245
	struct zone **z;
	void *obj = NULL;
3246
	int nid;
3247 3248 3249 3250 3251 3252 3253

	if (flags & __GFP_THISNODE)
		return NULL;

	zonelist = &NODE_DATA(slab_node(current->mempolicy))
			->node_zonelists[gfp_zone(flags)];
	local_flags = (flags & GFP_LEVEL_MASK);
3254

3255 3256 3257 3258 3259
retry:
	/*
	 * Look through allowed nodes for objects available
	 * from existing per node queues.
	 */
3260
	for (z = zonelist->zones; *z && !obj; z++) {
3261
		nid = zone_to_nid(*z);
3262

3263
		if (cpuset_zone_allowed_hardwall(*z, flags) &&
3264 3265 3266 3267 3268 3269
			cache->nodelists[nid] &&
			cache->nodelists[nid]->free_objects)
				obj = ____cache_alloc_node(cache,
					flags | GFP_THISNODE, nid);
	}

3270
	if (!obj && !(flags & __GFP_NO_GROW)) {
3271 3272 3273 3274 3275 3276
		/*
		 * This allocation will be performed within the constraints
		 * of the current cpuset / memory policy requirements.
		 * We may trigger various forms of reclaim on the allowed
		 * set and go into memory reserves if necessary.
		 */
3277 3278 3279
		if (local_flags & __GFP_WAIT)
			local_irq_enable();
		kmem_flagcheck(cache, flags);
3280
		obj = kmem_getpages(cache, flags, -1);
3281 3282
		if (local_flags & __GFP_WAIT)
			local_irq_disable();
3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298
		if (obj) {
			/*
			 * Insert into the appropriate per node queues
			 */
			nid = page_to_nid(virt_to_page(obj));
			if (cache_grow(cache, flags, nid, obj)) {
				obj = ____cache_alloc_node(cache,
					flags | GFP_THISNODE, nid);
				if (!obj)
					/*
					 * Another processor may allocate the
					 * objects in the slab since we are
					 * not holding any locks.
					 */
					goto retry;
			} else {
3299
				/* cache_grow already freed obj */
3300 3301 3302
				obj = NULL;
			}
		}
3303
	}
3304 3305 3306
	return obj;
}

3307 3308
/*
 * A interface to enable slab creation on nodeid
L
Linus Torvalds 已提交
3309
 */
3310
static void *____cache_alloc_node(struct kmem_cache *cachep, gfp_t flags,
A
Andrew Morton 已提交
3311
				int nodeid)
3312 3313
{
	struct list_head *entry;
P
Pekka Enberg 已提交
3314 3315 3316 3317 3318 3319 3320 3321
	struct slab *slabp;
	struct kmem_list3 *l3;
	void *obj;
	int x;

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

A
Andrew Morton 已提交
3322
retry:
3323
	check_irq_off();
P
Pekka Enberg 已提交
3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342
	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);

3343
	obj = slab_get_obj(cachep, slabp, nodeid);
P
Pekka Enberg 已提交
3344 3345 3346 3347 3348
	check_slabp(cachep, slabp);
	l3->free_objects--;
	/* move slabp to correct slabp list: */
	list_del(&slabp->list);

A
Andrew Morton 已提交
3349
	if (slabp->free == BUFCTL_END)
P
Pekka Enberg 已提交
3350
		list_add(&slabp->list, &l3->slabs_full);
A
Andrew Morton 已提交
3351
	else
P
Pekka Enberg 已提交
3352
		list_add(&slabp->list, &l3->slabs_partial);
3353

P
Pekka Enberg 已提交
3354 3355
	spin_unlock(&l3->list_lock);
	goto done;
3356

A
Andrew Morton 已提交
3357
must_grow:
P
Pekka Enberg 已提交
3358
	spin_unlock(&l3->list_lock);
3359
	x = cache_grow(cachep, flags | GFP_THISNODE, nodeid, NULL);
3360 3361
	if (x)
		goto retry;
L
Linus Torvalds 已提交
3362

3363
	return fallback_alloc(cachep, flags);
3364

A
Andrew Morton 已提交
3365
done:
P
Pekka Enberg 已提交
3366
	return obj;
3367
}
3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466

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

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

	if (unlikely(nodeid == -1))
		nodeid = numa_node_id();

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

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

	return ptr;
}

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

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

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

  out:
	return objp;
}
#else

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

#endif /* CONFIG_NUMA */

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

	cache_alloc_debugcheck_before(cachep, flags);
	local_irq_save(save_flags);
	objp = __do_cache_alloc(cachep, flags);
	local_irq_restore(save_flags);
	objp = cache_alloc_debugcheck_after(cachep, flags, objp, caller);
	prefetchw(objp);

	return objp;
}
3467 3468 3469 3470

/*
 * Caller needs to acquire correct kmem_list's list_lock
 */
3471
static void free_block(struct kmem_cache *cachep, void **objpp, int nr_objects,
P
Pekka Enberg 已提交
3472
		       int node)
L
Linus Torvalds 已提交
3473 3474
{
	int i;
3475
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
3476 3477 3478 3479 3480

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

3481
		slabp = virt_to_slab(objp);
3482
		l3 = cachep->nodelists[node];
L
Linus Torvalds 已提交
3483
		list_del(&slabp->list);
3484
		check_spinlock_acquired_node(cachep, node);
L
Linus Torvalds 已提交
3485
		check_slabp(cachep, slabp);
3486
		slab_put_obj(cachep, slabp, objp, node);
L
Linus Torvalds 已提交
3487
		STATS_DEC_ACTIVE(cachep);
3488
		l3->free_objects++;
L
Linus Torvalds 已提交
3489 3490 3491 3492
		check_slabp(cachep, slabp);

		/* fixup slab chains */
		if (slabp->inuse == 0) {
3493 3494
			if (l3->free_objects > l3->free_limit) {
				l3->free_objects -= cachep->num;
3495 3496 3497 3498 3499 3500
				/* No need to drop any previously held
				 * lock here, even if we have a off-slab slab
				 * descriptor it is guaranteed to come from
				 * a different cache, refer to comments before
				 * alloc_slabmgmt.
				 */
L
Linus Torvalds 已提交
3501 3502
				slab_destroy(cachep, slabp);
			} else {
3503
				list_add(&slabp->list, &l3->slabs_free);
L
Linus Torvalds 已提交
3504 3505 3506 3507 3508 3509
			}
		} else {
			/* Unconditionally move a slab to the end of the
			 * partial list on free - maximum time for the
			 * other objects to be freed, too.
			 */
3510
			list_add_tail(&slabp->list, &l3->slabs_partial);
L
Linus Torvalds 已提交
3511 3512 3513 3514
		}
	}
}

3515
static void cache_flusharray(struct kmem_cache *cachep, struct array_cache *ac)
L
Linus Torvalds 已提交
3516 3517
{
	int batchcount;
3518
	struct kmem_list3 *l3;
3519
	int node = numa_node_id();
L
Linus Torvalds 已提交
3520 3521 3522 3523 3524 3525

	batchcount = ac->batchcount;
#if DEBUG
	BUG_ON(!batchcount || batchcount > ac->avail);
#endif
	check_irq_off();
3526
	l3 = cachep->nodelists[node];
3527
	spin_lock(&l3->list_lock);
3528 3529
	if (l3->shared) {
		struct array_cache *shared_array = l3->shared;
P
Pekka Enberg 已提交
3530
		int max = shared_array->limit - shared_array->avail;
L
Linus Torvalds 已提交
3531 3532 3533
		if (max) {
			if (batchcount > max)
				batchcount = max;
3534
			memcpy(&(shared_array->entry[shared_array->avail]),
P
Pekka Enberg 已提交
3535
			       ac->entry, sizeof(void *) * batchcount);
L
Linus Torvalds 已提交
3536 3537 3538 3539 3540
			shared_array->avail += batchcount;
			goto free_done;
		}
	}

3541
	free_block(cachep, ac->entry, batchcount, node);
A
Andrew Morton 已提交
3542
free_done:
L
Linus Torvalds 已提交
3543 3544 3545 3546 3547
#if STATS
	{
		int i = 0;
		struct list_head *p;

3548 3549
		p = l3->slabs_free.next;
		while (p != &(l3->slabs_free)) {
L
Linus Torvalds 已提交
3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560
			struct slab *slabp;

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

			i++;
			p = p->next;
		}
		STATS_SET_FREEABLE(cachep, i);
	}
#endif
3561
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
3562
	ac->avail -= batchcount;
A
Andrew Morton 已提交
3563
	memmove(ac->entry, &(ac->entry[batchcount]), sizeof(void *)*ac->avail);
L
Linus Torvalds 已提交
3564 3565 3566
}

/*
A
Andrew Morton 已提交
3567 3568
 * Release an obj back to its cache. If the obj has a constructed state, it must
 * be in this state _before_ it is released.  Called with disabled ints.
L
Linus Torvalds 已提交
3569
 */
3570
static inline void __cache_free(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
3571
{
3572
	struct array_cache *ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
3573 3574 3575 3576

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

3577
	if (use_alien_caches && cache_free_alien(cachep, objp))
3578 3579
		return;

L
Linus Torvalds 已提交
3580 3581
	if (likely(ac->avail < ac->limit)) {
		STATS_INC_FREEHIT(cachep);
3582
		ac->entry[ac->avail++] = objp;
L
Linus Torvalds 已提交
3583 3584 3585 3586
		return;
	} else {
		STATS_INC_FREEMISS(cachep);
		cache_flusharray(cachep, ac);
3587
		ac->entry[ac->avail++] = objp;
L
Linus Torvalds 已提交
3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598
	}
}

/**
 * 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.
 */
3599
void *kmem_cache_alloc(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
3600
{
3601
	return __cache_alloc(cachep, flags, __builtin_return_address(0));
L
Linus Torvalds 已提交
3602 3603 3604
}
EXPORT_SYMBOL(kmem_cache_alloc);

3605
/**
3606
 * kmem_cache_zalloc - Allocate an object. The memory is set to zero.
3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621
 * @cache: The cache to allocate from.
 * @flags: See kmalloc().
 *
 * Allocate an object from this cache and set the allocated memory to zero.
 * The flags are only relevant if the cache has no available objects.
 */
void *kmem_cache_zalloc(struct kmem_cache *cache, gfp_t flags)
{
	void *ret = __cache_alloc(cache, flags, __builtin_return_address(0));
	if (ret)
		memset(ret, 0, obj_size(cache));
	return ret;
}
EXPORT_SYMBOL(kmem_cache_zalloc);

L
Linus Torvalds 已提交
3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635
/**
 * 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.
 */
3636
int kmem_ptr_validate(struct kmem_cache *cachep, const void *ptr)
L
Linus Torvalds 已提交
3637
{
P
Pekka Enberg 已提交
3638
	unsigned long addr = (unsigned long)ptr;
L
Linus Torvalds 已提交
3639
	unsigned long min_addr = PAGE_OFFSET;
P
Pekka Enberg 已提交
3640
	unsigned long align_mask = BYTES_PER_WORD - 1;
3641
	unsigned long size = cachep->buffer_size;
L
Linus Torvalds 已提交
3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656
	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;
3657
	if (unlikely(page_get_cache(page) != cachep))
L
Linus Torvalds 已提交
3658 3659
		goto out;
	return 1;
A
Andrew Morton 已提交
3660
out:
L
Linus Torvalds 已提交
3661 3662 3663 3664
	return 0;
}

#ifdef CONFIG_NUMA
3665 3666 3667 3668 3669
void *kmem_cache_alloc_node(struct kmem_cache *cachep, gfp_t flags, int nodeid)
{
	return __cache_alloc_node(cachep, flags, nodeid,
			__builtin_return_address(0));
}
L
Linus Torvalds 已提交
3670 3671
EXPORT_SYMBOL(kmem_cache_alloc_node);

3672 3673
static __always_inline void *
__do_kmalloc_node(size_t size, gfp_t flags, int node, void *caller)
3674
{
3675
	struct kmem_cache *cachep;
3676 3677 3678 3679 3680 3681

	cachep = kmem_find_general_cachep(size, flags);
	if (unlikely(cachep == NULL))
		return NULL;
	return kmem_cache_alloc_node(cachep, flags, node);
}
3682 3683 3684 3685 3686 3687 3688

#ifdef CONFIG_DEBUG_SLAB
void *__kmalloc_node(size_t size, gfp_t flags, int node)
{
	return __do_kmalloc_node(size, flags, node,
			__builtin_return_address(0));
}
3689
EXPORT_SYMBOL(__kmalloc_node);
3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704

void *__kmalloc_node_track_caller(size_t size, gfp_t flags,
		int node, void *caller)
{
	return __do_kmalloc_node(size, flags, node, caller);
}
EXPORT_SYMBOL(__kmalloc_node_track_caller);
#else
void *__kmalloc_node(size_t size, gfp_t flags, int node)
{
	return __do_kmalloc_node(size, flags, node, NULL);
}
EXPORT_SYMBOL(__kmalloc_node);
#endif /* CONFIG_DEBUG_SLAB */
#endif /* CONFIG_NUMA */
L
Linus Torvalds 已提交
3705 3706

/**
3707
 * __do_kmalloc - allocate memory
L
Linus Torvalds 已提交
3708
 * @size: how many bytes of memory are required.
3709
 * @flags: the type of memory to allocate (see kmalloc).
3710
 * @caller: function caller for debug tracking of the caller
L
Linus Torvalds 已提交
3711
 */
3712 3713
static __always_inline void *__do_kmalloc(size_t size, gfp_t flags,
					  void *caller)
L
Linus Torvalds 已提交
3714
{
3715
	struct kmem_cache *cachep;
L
Linus Torvalds 已提交
3716

3717 3718 3719 3720 3721 3722
	/* 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);
3723 3724
	if (unlikely(cachep == NULL))
		return NULL;
3725 3726 3727 3728
	return __cache_alloc(cachep, flags, caller);
}


3729
#ifdef CONFIG_DEBUG_SLAB
3730 3731
void *__kmalloc(size_t size, gfp_t flags)
{
3732
	return __do_kmalloc(size, flags, __builtin_return_address(0));
L
Linus Torvalds 已提交
3733 3734 3735
}
EXPORT_SYMBOL(__kmalloc);

3736 3737 3738 3739 3740
void *__kmalloc_track_caller(size_t size, gfp_t flags, void *caller)
{
	return __do_kmalloc(size, flags, caller);
}
EXPORT_SYMBOL(__kmalloc_track_caller);
3741 3742 3743 3744 3745 3746 3747

#else
void *__kmalloc(size_t size, gfp_t flags)
{
	return __do_kmalloc(size, flags, NULL);
}
EXPORT_SYMBOL(__kmalloc);
3748 3749
#endif

P
Pekka Enberg 已提交
3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796
/**
 * krealloc - reallocate memory. The contents will remain unchanged.
 *
 * @p: object to reallocate memory for.
 * @new_size: how many bytes of memory are required.
 * @flags: the type of memory to allocate.
 *
 * The contents of the object pointed to are preserved up to the
 * lesser of the new and old sizes.  If @p is %NULL, krealloc()
 * behaves exactly like kmalloc().  If @size is 0 and @p is not a
 * %NULL pointer, the object pointed to is freed.
 */
void *krealloc(const void *p, size_t new_size, gfp_t flags)
{
	struct kmem_cache *cache, *new_cache;
	void *ret;

	if (unlikely(!p))
		return kmalloc_track_caller(new_size, flags);

	if (unlikely(!new_size)) {
		kfree(p);
		return NULL;
	}

	cache = virt_to_cache(p);
	new_cache = __find_general_cachep(new_size, flags);

	/*
 	 * If new size fits in the current cache, bail out.
 	 */
	if (likely(cache == new_cache))
		return (void *)p;

	/*
 	 * We are on the slow-path here so do not use __cache_alloc
 	 * because it bloats kernel text.
 	 */
	ret = kmalloc_track_caller(new_size, flags);
	if (ret) {
		memcpy(ret, p, min(new_size, ksize(p)));
		kfree(p);
	}
	return ret;
}
EXPORT_SYMBOL(krealloc);

L
Linus Torvalds 已提交
3797 3798 3799 3800 3801 3802 3803 3804
/**
 * 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.
 */
3805
void kmem_cache_free(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
3806 3807 3808
{
	unsigned long flags;

3809 3810
	BUG_ON(virt_to_cache(objp) != cachep);

L
Linus Torvalds 已提交
3811
	local_irq_save(flags);
3812
	debug_check_no_locks_freed(objp, obj_size(cachep));
3813
	__cache_free(cachep, objp);
L
Linus Torvalds 已提交
3814 3815 3816 3817 3818 3819 3820 3821
	local_irq_restore(flags);
}
EXPORT_SYMBOL(kmem_cache_free);

/**
 * kfree - free previously allocated memory
 * @objp: pointer returned by kmalloc.
 *
3822 3823
 * If @objp is NULL, no operation is performed.
 *
L
Linus Torvalds 已提交
3824 3825 3826 3827 3828
 * Don't free memory not originally allocated by kmalloc()
 * or you will run into trouble.
 */
void kfree(const void *objp)
{
3829
	struct kmem_cache *c;
L
Linus Torvalds 已提交
3830 3831 3832 3833 3834 3835
	unsigned long flags;

	if (unlikely(!objp))
		return;
	local_irq_save(flags);
	kfree_debugcheck(objp);
3836
	c = virt_to_cache(objp);
3837
	debug_check_no_locks_freed(objp, obj_size(c));
3838
	__cache_free(c, (void *)objp);
L
Linus Torvalds 已提交
3839 3840 3841 3842
	local_irq_restore(flags);
}
EXPORT_SYMBOL(kfree);

3843
unsigned int kmem_cache_size(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
3844
{
3845
	return obj_size(cachep);
L
Linus Torvalds 已提交
3846 3847 3848
}
EXPORT_SYMBOL(kmem_cache_size);

3849
const char *kmem_cache_name(struct kmem_cache *cachep)
3850 3851 3852 3853 3854
{
	return cachep->name;
}
EXPORT_SYMBOL_GPL(kmem_cache_name);

3855
/*
3856
 * This initializes kmem_list3 or resizes varioius caches for all nodes.
3857
 */
3858
static int alloc_kmemlist(struct kmem_cache *cachep)
3859 3860 3861
{
	int node;
	struct kmem_list3 *l3;
3862
	struct array_cache *new_shared;
3863
	struct array_cache **new_alien = NULL;
3864 3865

	for_each_online_node(node) {
3866

3867 3868 3869 3870 3871
                if (use_alien_caches) {
                        new_alien = alloc_alien_cache(node, cachep->limit);
                        if (!new_alien)
                                goto fail;
                }
3872

3873 3874
		new_shared = alloc_arraycache(node,
				cachep->shared*cachep->batchcount,
A
Andrew Morton 已提交
3875
					0xbaadf00d);
3876 3877
		if (!new_shared) {
			free_alien_cache(new_alien);
3878
			goto fail;
3879
		}
3880

A
Andrew Morton 已提交
3881 3882
		l3 = cachep->nodelists[node];
		if (l3) {
3883 3884
			struct array_cache *shared = l3->shared;

3885 3886
			spin_lock_irq(&l3->list_lock);

3887
			if (shared)
3888 3889
				free_block(cachep, shared->entry,
						shared->avail, node);
3890

3891 3892
			l3->shared = new_shared;
			if (!l3->alien) {
3893 3894 3895
				l3->alien = new_alien;
				new_alien = NULL;
			}
P
Pekka Enberg 已提交
3896
			l3->free_limit = (1 + nr_cpus_node(node)) *
A
Andrew Morton 已提交
3897
					cachep->batchcount + cachep->num;
3898
			spin_unlock_irq(&l3->list_lock);
3899
			kfree(shared);
3900 3901 3902
			free_alien_cache(new_alien);
			continue;
		}
A
Andrew Morton 已提交
3903
		l3 = kmalloc_node(sizeof(struct kmem_list3), GFP_KERNEL, node);
3904 3905 3906
		if (!l3) {
			free_alien_cache(new_alien);
			kfree(new_shared);
3907
			goto fail;
3908
		}
3909 3910 3911

		kmem_list3_init(l3);
		l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
A
Andrew Morton 已提交
3912
				((unsigned long)cachep) % REAPTIMEOUT_LIST3;
3913
		l3->shared = new_shared;
3914
		l3->alien = new_alien;
P
Pekka Enberg 已提交
3915
		l3->free_limit = (1 + nr_cpus_node(node)) *
A
Andrew Morton 已提交
3916
					cachep->batchcount + cachep->num;
3917 3918
		cachep->nodelists[node] = l3;
	}
3919
	return 0;
3920

A
Andrew Morton 已提交
3921
fail:
3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936
	if (!cachep->next.next) {
		/* Cache is not active yet. Roll back what we did */
		node--;
		while (node >= 0) {
			if (cachep->nodelists[node]) {
				l3 = cachep->nodelists[node];

				kfree(l3->shared);
				free_alien_cache(l3->alien);
				kfree(l3);
				cachep->nodelists[node] = NULL;
			}
			node--;
		}
	}
3937
	return -ENOMEM;
3938 3939
}

L
Linus Torvalds 已提交
3940
struct ccupdate_struct {
3941
	struct kmem_cache *cachep;
L
Linus Torvalds 已提交
3942 3943 3944 3945 3946
	struct array_cache *new[NR_CPUS];
};

static void do_ccupdate_local(void *info)
{
A
Andrew Morton 已提交
3947
	struct ccupdate_struct *new = info;
L
Linus Torvalds 已提交
3948 3949 3950
	struct array_cache *old;

	check_irq_off();
3951
	old = cpu_cache_get(new->cachep);
3952

L
Linus Torvalds 已提交
3953 3954 3955 3956
	new->cachep->array[smp_processor_id()] = new->new[smp_processor_id()];
	new->new[smp_processor_id()] = old;
}

3957
/* Always called with the cache_chain_mutex held */
A
Andrew Morton 已提交
3958 3959
static int do_tune_cpucache(struct kmem_cache *cachep, int limit,
				int batchcount, int shared)
L
Linus Torvalds 已提交
3960
{
3961
	struct ccupdate_struct *new;
3962
	int i;
L
Linus Torvalds 已提交
3963

3964 3965 3966 3967
	new = kzalloc(sizeof(*new), GFP_KERNEL);
	if (!new)
		return -ENOMEM;

3968
	for_each_online_cpu(i) {
3969
		new->new[i] = alloc_arraycache(cpu_to_node(i), limit,
A
Andrew Morton 已提交
3970
						batchcount);
3971
		if (!new->new[i]) {
P
Pekka Enberg 已提交
3972
			for (i--; i >= 0; i--)
3973 3974
				kfree(new->new[i]);
			kfree(new);
3975
			return -ENOMEM;
L
Linus Torvalds 已提交
3976 3977
		}
	}
3978
	new->cachep = cachep;
L
Linus Torvalds 已提交
3979

3980
	on_each_cpu(do_ccupdate_local, (void *)new, 1, 1);
3981

L
Linus Torvalds 已提交
3982 3983 3984
	check_irq_on();
	cachep->batchcount = batchcount;
	cachep->limit = limit;
3985
	cachep->shared = shared;
L
Linus Torvalds 已提交
3986

3987
	for_each_online_cpu(i) {
3988
		struct array_cache *ccold = new->new[i];
L
Linus Torvalds 已提交
3989 3990
		if (!ccold)
			continue;
3991
		spin_lock_irq(&cachep->nodelists[cpu_to_node(i)]->list_lock);
3992
		free_block(cachep, ccold->entry, ccold->avail, cpu_to_node(i));
3993
		spin_unlock_irq(&cachep->nodelists[cpu_to_node(i)]->list_lock);
L
Linus Torvalds 已提交
3994 3995
		kfree(ccold);
	}
3996
	kfree(new);
3997
	return alloc_kmemlist(cachep);
L
Linus Torvalds 已提交
3998 3999
}

4000
/* Called with cache_chain_mutex held always */
4001
static int enable_cpucache(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
4002 4003 4004 4005
{
	int err;
	int limit, shared;

A
Andrew Morton 已提交
4006 4007
	/*
	 * The head array serves three purposes:
L
Linus Torvalds 已提交
4008 4009
	 * - create a LIFO ordering, i.e. return objects that are cache-warm
	 * - reduce the number of spinlock operations.
A
Andrew Morton 已提交
4010
	 * - reduce the number of linked list operations on the slab and
L
Linus Torvalds 已提交
4011 4012 4013 4014
	 *   bufctl chains: array operations are cheaper.
	 * The numbers are guessed, we should auto-tune as described by
	 * Bonwick.
	 */
4015
	if (cachep->buffer_size > 131072)
L
Linus Torvalds 已提交
4016
		limit = 1;
4017
	else if (cachep->buffer_size > PAGE_SIZE)
L
Linus Torvalds 已提交
4018
		limit = 8;
4019
	else if (cachep->buffer_size > 1024)
L
Linus Torvalds 已提交
4020
		limit = 24;
4021
	else if (cachep->buffer_size > 256)
L
Linus Torvalds 已提交
4022 4023 4024 4025
		limit = 54;
	else
		limit = 120;

A
Andrew Morton 已提交
4026 4027
	/*
	 * CPU bound tasks (e.g. network routing) can exhibit cpu bound
L
Linus Torvalds 已提交
4028 4029 4030 4031 4032 4033 4034 4035
	 * 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;
4036
	if (cachep->buffer_size <= PAGE_SIZE && num_possible_cpus() > 1)
L
Linus Torvalds 已提交
4037 4038 4039
		shared = 8;

#if DEBUG
A
Andrew Morton 已提交
4040 4041 4042
	/*
	 * With debugging enabled, large batchcount lead to excessively long
	 * periods with disabled local interrupts. Limit the batchcount
L
Linus Torvalds 已提交
4043 4044 4045 4046
	 */
	if (limit > 32)
		limit = 32;
#endif
P
Pekka Enberg 已提交
4047
	err = do_tune_cpucache(cachep, limit, (limit + 1) / 2, shared);
L
Linus Torvalds 已提交
4048 4049
	if (err)
		printk(KERN_ERR "enable_cpucache failed for %s, error %d.\n",
P
Pekka Enberg 已提交
4050
		       cachep->name, -err);
4051
	return err;
L
Linus Torvalds 已提交
4052 4053
}

4054 4055
/*
 * Drain an array if it contains any elements taking the l3 lock only if
4056 4057
 * necessary. Note that the l3 listlock also protects the array_cache
 * if drain_array() is used on the shared array.
4058 4059 4060
 */
void drain_array(struct kmem_cache *cachep, struct kmem_list3 *l3,
			 struct array_cache *ac, int force, int node)
L
Linus Torvalds 已提交
4061 4062 4063
{
	int tofree;

4064 4065
	if (!ac || !ac->avail)
		return;
L
Linus Torvalds 已提交
4066 4067
	if (ac->touched && !force) {
		ac->touched = 0;
4068
	} else {
4069
		spin_lock_irq(&l3->list_lock);
4070 4071 4072 4073 4074 4075 4076 4077 4078
		if (ac->avail) {
			tofree = force ? ac->avail : (ac->limit + 4) / 5;
			if (tofree > ac->avail)
				tofree = (ac->avail + 1) / 2;
			free_block(cachep, ac->entry, tofree, node);
			ac->avail -= tofree;
			memmove(ac->entry, &(ac->entry[tofree]),
				sizeof(void *) * ac->avail);
		}
4079
		spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
4080 4081 4082 4083 4084
	}
}

/**
 * cache_reap - Reclaim memory from caches.
4085
 * @w: work descriptor
L
Linus Torvalds 已提交
4086 4087 4088 4089 4090 4091
 *
 * Called from workqueue/eventd every few seconds.
 * Purpose:
 * - clear the per-cpu caches for this CPU.
 * - return freeable pages to the main free memory pool.
 *
A
Andrew Morton 已提交
4092 4093
 * If we cannot acquire the cache chain mutex then just give up - we'll try
 * again on the next iteration.
L
Linus Torvalds 已提交
4094
 */
4095
static void cache_reap(struct work_struct *w)
L
Linus Torvalds 已提交
4096
{
4097
	struct kmem_cache *searchp;
4098
	struct kmem_list3 *l3;
4099
	int node = numa_node_id();
4100 4101
	struct delayed_work *work =
		container_of(w, struct delayed_work, work);
L
Linus Torvalds 已提交
4102

4103
	if (!mutex_trylock(&cache_chain_mutex))
L
Linus Torvalds 已提交
4104
		/* Give up. Setup the next iteration. */
4105
		goto out;
L
Linus Torvalds 已提交
4106

4107
	list_for_each_entry(searchp, &cache_chain, next) {
L
Linus Torvalds 已提交
4108 4109
		check_irq_on();

4110 4111 4112 4113 4114
		/*
		 * We only take the l3 lock if absolutely necessary and we
		 * have established with reasonable certainty that
		 * we can do some work if the lock was obtained.
		 */
4115
		l3 = searchp->nodelists[node];
4116

4117
		reap_alien(searchp, l3);
L
Linus Torvalds 已提交
4118

4119
		drain_array(searchp, l3, cpu_cache_get(searchp), 0, node);
L
Linus Torvalds 已提交
4120

4121 4122 4123 4124
		/*
		 * These are racy checks but it does not matter
		 * if we skip one check or scan twice.
		 */
4125
		if (time_after(l3->next_reap, jiffies))
4126
			goto next;
L
Linus Torvalds 已提交
4127

4128
		l3->next_reap = jiffies + REAPTIMEOUT_LIST3;
L
Linus Torvalds 已提交
4129

4130
		drain_array(searchp, l3, l3->shared, 0, node);
L
Linus Torvalds 已提交
4131

4132
		if (l3->free_touched)
4133
			l3->free_touched = 0;
4134 4135
		else {
			int freed;
L
Linus Torvalds 已提交
4136

4137 4138 4139 4140
			freed = drain_freelist(searchp, l3, (l3->free_limit +
				5 * searchp->num - 1) / (5 * searchp->num));
			STATS_ADD_REAPED(searchp, freed);
		}
4141
next:
L
Linus Torvalds 已提交
4142 4143 4144
		cond_resched();
	}
	check_irq_on();
I
Ingo Molnar 已提交
4145
	mutex_unlock(&cache_chain_mutex);
4146
	next_reap_node();
4147
	refresh_cpu_vm_stats(smp_processor_id());
4148
out:
A
Andrew Morton 已提交
4149
	/* Set up the next iteration */
4150
	schedule_delayed_work(work, round_jiffies_relative(REAPTIMEOUT_CPUC));
L
Linus Torvalds 已提交
4151 4152 4153 4154
}

#ifdef CONFIG_PROC_FS

4155
static void print_slabinfo_header(struct seq_file *m)
L
Linus Torvalds 已提交
4156
{
4157 4158 4159 4160
	/*
	 * Output format version, so at least we can change it
	 * without _too_ many complaints.
	 */
L
Linus Torvalds 已提交
4161
#if STATS
4162
	seq_puts(m, "slabinfo - version: 2.1 (statistics)\n");
L
Linus Torvalds 已提交
4163
#else
4164
	seq_puts(m, "slabinfo - version: 2.1\n");
L
Linus Torvalds 已提交
4165
#endif
4166 4167 4168 4169
	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
4170
#if STATS
4171
	seq_puts(m, " : globalstat <listallocs> <maxobjs> <grown> <reaped> "
4172
		 "<error> <maxfreeable> <nodeallocs> <remotefrees> <alienoverflow>");
4173
	seq_puts(m, " : cpustat <allochit> <allocmiss> <freehit> <freemiss>");
L
Linus Torvalds 已提交
4174
#endif
4175 4176 4177 4178 4179 4180 4181 4182
	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
4183
	mutex_lock(&cache_chain_mutex);
4184 4185
	if (!n)
		print_slabinfo_header(m);
L
Linus Torvalds 已提交
4186 4187 4188 4189 4190 4191
	p = cache_chain.next;
	while (n--) {
		p = p->next;
		if (p == &cache_chain)
			return NULL;
	}
4192
	return list_entry(p, struct kmem_cache, next);
L
Linus Torvalds 已提交
4193 4194 4195 4196
}

static void *s_next(struct seq_file *m, void *p, loff_t *pos)
{
4197
	struct kmem_cache *cachep = p;
L
Linus Torvalds 已提交
4198
	++*pos;
A
Andrew Morton 已提交
4199 4200
	return cachep->next.next == &cache_chain ?
		NULL : list_entry(cachep->next.next, struct kmem_cache, next);
L
Linus Torvalds 已提交
4201 4202 4203 4204
}

static void s_stop(struct seq_file *m, void *p)
{
I
Ingo Molnar 已提交
4205
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
4206 4207 4208 4209
}

static int s_show(struct seq_file *m, void *p)
{
4210
	struct kmem_cache *cachep = p;
P
Pekka Enberg 已提交
4211 4212 4213 4214 4215
	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;
4216
	const char *name;
L
Linus Torvalds 已提交
4217
	char *error = NULL;
4218 4219
	int node;
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
4220 4221 4222

	active_objs = 0;
	num_slabs = 0;
4223 4224 4225 4226 4227
	for_each_online_node(node) {
		l3 = cachep->nodelists[node];
		if (!l3)
			continue;

4228 4229
		check_irq_on();
		spin_lock_irq(&l3->list_lock);
4230

4231
		list_for_each_entry(slabp, &l3->slabs_full, list) {
4232 4233 4234 4235 4236
			if (slabp->inuse != cachep->num && !error)
				error = "slabs_full accounting error";
			active_objs += cachep->num;
			active_slabs++;
		}
4237
		list_for_each_entry(slabp, &l3->slabs_partial, list) {
4238 4239 4240 4241 4242 4243 4244
			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++;
		}
4245
		list_for_each_entry(slabp, &l3->slabs_free, list) {
4246 4247 4248 4249 4250
			if (slabp->inuse && !error)
				error = "slabs_free/inuse accounting error";
			num_slabs++;
		}
		free_objects += l3->free_objects;
4251 4252
		if (l3->shared)
			shared_avail += l3->shared->avail;
4253

4254
		spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
4255
	}
P
Pekka Enberg 已提交
4256 4257
	num_slabs += active_slabs;
	num_objs = num_slabs * cachep->num;
4258
	if (num_objs - active_objs != free_objects && !error)
L
Linus Torvalds 已提交
4259 4260
		error = "free_objects accounting error";

P
Pekka Enberg 已提交
4261
	name = cachep->name;
L
Linus Torvalds 已提交
4262 4263 4264 4265
	if (error)
		printk(KERN_ERR "slab: cache %s error: %s\n", name, error);

	seq_printf(m, "%-17s %6lu %6lu %6u %4u %4d",
4266
		   name, active_objs, num_objs, cachep->buffer_size,
P
Pekka Enberg 已提交
4267
		   cachep->num, (1 << cachep->gfporder));
L
Linus Torvalds 已提交
4268
	seq_printf(m, " : tunables %4u %4u %4u",
P
Pekka Enberg 已提交
4269
		   cachep->limit, cachep->batchcount, cachep->shared);
4270
	seq_printf(m, " : slabdata %6lu %6lu %6lu",
P
Pekka Enberg 已提交
4271
		   active_slabs, num_slabs, shared_avail);
L
Linus Torvalds 已提交
4272
#if STATS
P
Pekka Enberg 已提交
4273
	{			/* list3 stats */
L
Linus Torvalds 已提交
4274 4275 4276 4277 4278 4279 4280
		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;
4281
		unsigned long node_frees = cachep->node_frees;
4282
		unsigned long overflows = cachep->node_overflow;
L
Linus Torvalds 已提交
4283

4284
		seq_printf(m, " : globalstat %7lu %6lu %5lu %4lu \
4285
				%4lu %4lu %4lu %4lu %4lu", allocs, high, grown,
A
Andrew Morton 已提交
4286
				reaped, errors, max_freeable, node_allocs,
4287
				node_frees, overflows);
L
Linus Torvalds 已提交
4288 4289 4290 4291 4292 4293 4294 4295 4296
	}
	/* 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
4297
			   allochit, allocmiss, freehit, freemiss);
L
Linus Torvalds 已提交
4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317
	}
#endif
	seq_putc(m, '\n');
	return 0;
}

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

4318
const struct seq_operations slabinfo_op = {
P
Pekka Enberg 已提交
4319 4320 4321 4322
	.start = s_start,
	.next = s_next,
	.stop = s_stop,
	.show = s_show,
L
Linus Torvalds 已提交
4323 4324 4325 4326 4327 4328 4329 4330 4331 4332
};

#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
4333 4334
ssize_t slabinfo_write(struct file *file, const char __user * buffer,
		       size_t count, loff_t *ppos)
L
Linus Torvalds 已提交
4335
{
P
Pekka Enberg 已提交
4336
	char kbuf[MAX_SLABINFO_WRITE + 1], *tmp;
L
Linus Torvalds 已提交
4337
	int limit, batchcount, shared, res;
4338
	struct kmem_cache *cachep;
P
Pekka Enberg 已提交
4339

L
Linus Torvalds 已提交
4340 4341 4342 4343
	if (count > MAX_SLABINFO_WRITE)
		return -EINVAL;
	if (copy_from_user(&kbuf, buffer, count))
		return -EFAULT;
P
Pekka Enberg 已提交
4344
	kbuf[MAX_SLABINFO_WRITE] = '\0';
L
Linus Torvalds 已提交
4345 4346 4347 4348 4349 4350 4351 4352 4353 4354

	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
4355
	mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
4356
	res = -EINVAL;
4357
	list_for_each_entry(cachep, &cache_chain, next) {
L
Linus Torvalds 已提交
4358
		if (!strcmp(cachep->name, kbuf)) {
A
Andrew Morton 已提交
4359 4360
			if (limit < 1 || batchcount < 1 ||
					batchcount > limit || shared < 0) {
4361
				res = 0;
L
Linus Torvalds 已提交
4362
			} else {
4363
				res = do_tune_cpucache(cachep, limit,
P
Pekka Enberg 已提交
4364
						       batchcount, shared);
L
Linus Torvalds 已提交
4365 4366 4367 4368
			}
			break;
		}
	}
I
Ingo Molnar 已提交
4369
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
4370 4371 4372 4373
	if (res >= 0)
		res = count;
	return res;
}
4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482

#ifdef CONFIG_DEBUG_SLAB_LEAK

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

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

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

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

static void show_symbol(struct seq_file *m, unsigned long address)
{
#ifdef CONFIG_KALLSYMS
	char *modname;
	const char *name;
	unsigned long offset, size;
	char namebuf[KSYM_NAME_LEN+1];

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

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

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

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

	/* OK, we can do it */

	n[1] = 0;

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

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

4483
		list_for_each_entry(slabp, &l3->slabs_full, list)
4484
			handle_slab(n, cachep, slabp);
4485
		list_for_each_entry(slabp, &l3->slabs_partial, list)
4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511
			handle_slab(n, cachep, slabp);
		spin_unlock_irq(&l3->list_lock);
	}
	name = cachep->name;
	if (n[0] == n[1]) {
		/* Increase the buffer size */
		mutex_unlock(&cache_chain_mutex);
		m->private = kzalloc(n[0] * 4 * sizeof(unsigned long), GFP_KERNEL);
		if (!m->private) {
			/* Too bad, we are really out */
			m->private = n;
			mutex_lock(&cache_chain_mutex);
			return -ENOMEM;
		}
		*(unsigned long *)m->private = n[0] * 2;
		kfree(n);
		mutex_lock(&cache_chain_mutex);
		/* Now make sure this entry will be retried */
		m->count = m->size;
		return 0;
	}
	for (i = 0; i < n[1]; i++) {
		seq_printf(m, "%s: %lu ", name, n[2*i+3]);
		show_symbol(m, n[2*i+2]);
		seq_putc(m, '\n');
	}
4512

4513 4514 4515
	return 0;
}

4516
const struct seq_operations slabstats_op = {
4517 4518 4519 4520 4521 4522
	.start = leaks_start,
	.next = s_next,
	.stop = s_stop,
	.show = leaks_show,
};
#endif
L
Linus Torvalds 已提交
4523 4524
#endif

4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536
/**
 * 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.
 */
P
Pekka Enberg 已提交
4537
size_t ksize(const void *objp)
L
Linus Torvalds 已提交
4538
{
4539 4540
	if (unlikely(objp == NULL))
		return 0;
L
Linus Torvalds 已提交
4541

4542
	return obj_size(virt_to_cache(objp));
L
Linus Torvalds 已提交
4543
}