slab.c 117.7 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
/*
 * 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
S
Simon Arlott 已提交
29
 * slabs and you must pass objects with the same initializations to
L
Linus Torvalds 已提交
30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52
 * 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>
98
#include	<linux/proc_fs.h>
L
Linus Torvalds 已提交
99 100 101 102 103 104 105
#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>
106
#include	<linux/string.h>
107
#include	<linux/uaccess.h>
108
#include	<linux/nodemask.h>
109
#include	<linux/kmemleak.h>
110
#include	<linux/mempolicy.h>
I
Ingo Molnar 已提交
111
#include	<linux/mutex.h>
112
#include	<linux/fault-inject.h>
I
Ingo Molnar 已提交
113
#include	<linux/rtmutex.h>
114
#include	<linux/reciprocal_div.h>
115
#include	<linux/debugobjects.h>
P
Pekka Enberg 已提交
116
#include	<linux/kmemcheck.h>
117
#include	<linux/memory.h>
L
Linus Torvalds 已提交
118 119 120 121 122 123

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

/*
124
 * DEBUG	- 1 for kmem_cache_create() to honour; SLAB_RED_ZONE & SLAB_POISON.
L
Linus Torvalds 已提交
125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144
 *		  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 *)
D
David Woodhouse 已提交
145
#define	REDZONE_ALIGN		max(BYTES_PER_WORD, __alignof__(unsigned long long))
L
Linus Torvalds 已提交
146 147 148 149 150 151 152

#ifndef ARCH_KMALLOC_FLAGS
#define ARCH_KMALLOC_FLAGS SLAB_HWCACHE_ALIGN
#endif

/* Legal flag mask for kmem_cache_create(). */
#if DEBUG
153
# define CREATE_MASK	(SLAB_RED_ZONE | \
L
Linus Torvalds 已提交
154
			 SLAB_POISON | SLAB_HWCACHE_ALIGN | \
155
			 SLAB_CACHE_DMA | \
156
			 SLAB_STORE_USER | \
L
Linus Torvalds 已提交
157
			 SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
158
			 SLAB_DESTROY_BY_RCU | SLAB_MEM_SPREAD | \
P
Pekka Enberg 已提交
159
			 SLAB_DEBUG_OBJECTS | SLAB_NOLEAKTRACE | SLAB_NOTRACK)
L
Linus Torvalds 已提交
160
#else
161
# define CREATE_MASK	(SLAB_HWCACHE_ALIGN | \
162
			 SLAB_CACHE_DMA | \
L
Linus Torvalds 已提交
163
			 SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
164
			 SLAB_DESTROY_BY_RCU | SLAB_MEM_SPREAD | \
P
Pekka Enberg 已提交
165
			 SLAB_DEBUG_OBJECTS | SLAB_NOLEAKTRACE | SLAB_NOTRACK)
L
Linus Torvalds 已提交
166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186
#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.
 */

187
typedef unsigned int kmem_bufctl_t;
L
Linus Torvalds 已提交
188 189
#define BUFCTL_END	(((kmem_bufctl_t)(~0U))-0)
#define BUFCTL_FREE	(((kmem_bufctl_t)(~0U))-1)
190 191
#define	BUFCTL_ACTIVE	(((kmem_bufctl_t)(~0U))-2)
#define	SLAB_LIMIT	(((kmem_bufctl_t)(~0U))-3)
L
Linus Torvalds 已提交
192 193 194 195 196 197 198 199 200

/*
 * 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
201 202 203 204 205 206
	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
207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225
};

/*
 * 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
226
	struct rcu_head head;
227
	struct kmem_cache *cachep;
P
Pekka Enberg 已提交
228
	void *addr;
L
Linus Torvalds 已提交
229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247
};

/*
 * 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;
248
	spinlock_t lock;
249
	void *entry[];	/*
A
Andrew Morton 已提交
250 251 252 253
			 * Must have this definition in here for the proper
			 * alignment of array_cache. Also simplifies accessing
			 * the entries.
			 */
L
Linus Torvalds 已提交
254 255
};

A
Andrew Morton 已提交
256 257 258
/*
 * bootstrap: The caches do not work without cpuarrays anymore, but the
 * cpuarrays are allocated from the generic caches...
L
Linus Torvalds 已提交
259 260 261 262
 */
#define BOOT_CPUCACHE_ENTRIES	1
struct arraycache_init {
	struct array_cache cache;
P
Pekka Enberg 已提交
263
	void *entries[BOOT_CPUCACHE_ENTRIES];
L
Linus Torvalds 已提交
264 265 266
};

/*
267
 * The slab lists for all objects.
L
Linus Torvalds 已提交
268 269
 */
struct kmem_list3 {
P
Pekka Enberg 已提交
270 271 272 273 274
	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;
275
	unsigned int colour_next;	/* Per-node cache coloring */
P
Pekka Enberg 已提交
276 277 278
	spinlock_t list_lock;
	struct array_cache *shared;	/* shared per node */
	struct array_cache **alien;	/* on other nodes */
279 280
	unsigned long next_reap;	/* updated without locking */
	int free_touched;		/* updated without locking */
L
Linus Torvalds 已提交
281 282
};

283 284 285
/*
 * Need this for bootstrapping a per node allocator.
 */
286
#define NUM_INIT_LISTS (3 * MAX_NUMNODES)
287
static struct kmem_list3 __initdata initkmem_list3[NUM_INIT_LISTS];
288
#define	CACHE_CACHE 0
289 290
#define	SIZE_AC MAX_NUMNODES
#define	SIZE_L3 (2 * MAX_NUMNODES)
291

292 293 294 295
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);
296
static int enable_cpucache(struct kmem_cache *cachep, gfp_t gfp);
297
static void cache_reap(struct work_struct *unused);
298

299
/*
A
Andrew Morton 已提交
300 301
 * 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.
302
 */
303
static __always_inline int index_of(const size_t size)
304
{
305 306
	extern void __bad_size(void);

307 308 309 310 311 312 313 314
	if (__builtin_constant_p(size)) {
		int i = 0;

#define CACHE(x) \
	if (size <=x) \
		return i; \
	else \
		i++;
315
#include <linux/kmalloc_sizes.h>
316
#undef CACHE
317
		__bad_size();
318
	} else
319
		__bad_size();
320 321 322
	return 0;
}

323 324
static int slab_early_init = 1;

325 326
#define INDEX_AC index_of(sizeof(struct arraycache_init))
#define INDEX_L3 index_of(sizeof(struct kmem_list3))
L
Linus Torvalds 已提交
327

P
Pekka Enberg 已提交
328
static void kmem_list3_init(struct kmem_list3 *parent)
329 330 331 332 333 334
{
	INIT_LIST_HEAD(&parent->slabs_full);
	INIT_LIST_HEAD(&parent->slabs_partial);
	INIT_LIST_HEAD(&parent->slabs_free);
	parent->shared = NULL;
	parent->alien = NULL;
335
	parent->colour_next = 0;
336 337 338 339 340
	spin_lock_init(&parent->list_lock);
	parent->free_objects = 0;
	parent->free_touched = 0;
}

A
Andrew Morton 已提交
341 342 343 344
#define MAKE_LIST(cachep, listp, slab, nodeid)				\
	do {								\
		INIT_LIST_HEAD(listp);					\
		list_splice(&(cachep->nodelists[nodeid]->slab), listp);	\
345 346
	} while (0)

A
Andrew Morton 已提交
347 348
#define	MAKE_ALL_LISTS(cachep, ptr, nodeid)				\
	do {								\
349 350 351 352
	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
353 354 355 356 357

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

#define BATCHREFILL_LIMIT	16
A
Andrew Morton 已提交
358 359 360
/*
 * Optimization question: fewer reaps means less probability for unnessary
 * cpucache drain/refill cycles.
L
Linus Torvalds 已提交
361
 *
A
Adrian Bunk 已提交
362
 * OTOH the cpuarrays can contain lots of objects,
L
Linus Torvalds 已提交
363 364 365 366 367 368 369 370 371 372
 * 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++)
373
#define	STATS_ADD_REAPED(x,y)	((x)->reaped += (y))
A
Andrew Morton 已提交
374 375 376 377 378
#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
379 380
#define	STATS_INC_ERR(x)	((x)->errors++)
#define	STATS_INC_NODEALLOCS(x)	((x)->node_allocs++)
381
#define	STATS_INC_NODEFREES(x)	((x)->node_frees++)
382
#define STATS_INC_ACOVERFLOW(x)   ((x)->node_overflow++)
A
Andrew Morton 已提交
383 384 385 386 387
#define	STATS_SET_FREEABLE(x, i)					\
	do {								\
		if ((x)->max_freeable < i)				\
			(x)->max_freeable = i;				\
	} while (0)
L
Linus Torvalds 已提交
388 389 390 391 392 393 394 395 396
#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)
397
#define	STATS_ADD_REAPED(x,y)	do { (void)(y); } while (0)
L
Linus Torvalds 已提交
398 399 400
#define	STATS_SET_HIGH(x)	do { } while (0)
#define	STATS_INC_ERR(x)	do { } while (0)
#define	STATS_INC_NODEALLOCS(x)	do { } while (0)
401
#define	STATS_INC_NODEFREES(x)	do { } while (0)
402
#define STATS_INC_ACOVERFLOW(x)   do { } while (0)
A
Andrew Morton 已提交
403
#define	STATS_SET_FREEABLE(x, i) do { } while (0)
L
Linus Torvalds 已提交
404 405 406 407 408 409 410 411
#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
412 413
/*
 * memory layout of objects:
L
Linus Torvalds 已提交
414
 * 0		: objp
415
 * 0 .. cachep->obj_offset - BYTES_PER_WORD - 1: padding. This ensures that
L
Linus Torvalds 已提交
416 417
 * 		the end of an object is aligned with the end of the real
 * 		allocation. Catches writes behind the end of the allocation.
418
 * cachep->obj_offset - BYTES_PER_WORD .. cachep->obj_offset - 1:
L
Linus Torvalds 已提交
419
 * 		redzone word.
420 421
 * 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
422 423
 * cachep->buffer_size - 1* BYTES_PER_WORD: last caller address
 *					[BYTES_PER_WORD long]
L
Linus Torvalds 已提交
424
 */
425
static int obj_offset(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
426
{
427
	return cachep->obj_offset;
L
Linus Torvalds 已提交
428 429
}

430
static int obj_size(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
431
{
432
	return cachep->obj_size;
L
Linus Torvalds 已提交
433 434
}

435
static unsigned long long *dbg_redzone1(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
436 437
{
	BUG_ON(!(cachep->flags & SLAB_RED_ZONE));
438 439
	return (unsigned long long*) (objp + obj_offset(cachep) -
				      sizeof(unsigned long long));
L
Linus Torvalds 已提交
440 441
}

442
static unsigned long long *dbg_redzone2(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
443 444 445
{
	BUG_ON(!(cachep->flags & SLAB_RED_ZONE));
	if (cachep->flags & SLAB_STORE_USER)
446 447
		return (unsigned long long *)(objp + cachep->buffer_size -
					      sizeof(unsigned long long) -
D
David Woodhouse 已提交
448
					      REDZONE_ALIGN);
449 450
	return (unsigned long long *) (objp + cachep->buffer_size -
				       sizeof(unsigned long long));
L
Linus Torvalds 已提交
451 452
}

453
static void **dbg_userword(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
454 455
{
	BUG_ON(!(cachep->flags & SLAB_STORE_USER));
456
	return (void **)(objp + cachep->buffer_size - BYTES_PER_WORD);
L
Linus Torvalds 已提交
457 458 459 460
}

#else

461 462
#define obj_offset(x)			0
#define obj_size(cachep)		(cachep->buffer_size)
463 464
#define dbg_redzone1(cachep, objp)	({BUG(); (unsigned long long *)NULL;})
#define dbg_redzone2(cachep, objp)	({BUG(); (unsigned long long *)NULL;})
L
Linus Torvalds 已提交
465 466 467 468
#define dbg_userword(cachep, objp)	({BUG(); (void **)NULL;})

#endif

469
#ifdef CONFIG_TRACING
E
Eduard - Gabriel Munteanu 已提交
470 471 472 473 474 475 476
size_t slab_buffer_size(struct kmem_cache *cachep)
{
	return cachep->buffer_size;
}
EXPORT_SYMBOL(slab_buffer_size);
#endif

L
Linus Torvalds 已提交
477 478 479 480 481 482 483
/*
 * 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
484 485 486 487
/*
 * 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
488
 */
489 490 491 492 493 494 495
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)
{
496
	page = compound_head(page);
497
	BUG_ON(!PageSlab(page));
498 499 500 501 502 503 504 505 506 507
	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)
{
508
	BUG_ON(!PageSlab(page));
509 510
	return (struct slab *)page->lru.prev;
}
L
Linus Torvalds 已提交
511

512 513
static inline struct kmem_cache *virt_to_cache(const void *obj)
{
514
	struct page *page = virt_to_head_page(obj);
515 516 517 518 519
	return page_get_cache(page);
}

static inline struct slab *virt_to_slab(const void *obj)
{
520
	struct page *page = virt_to_head_page(obj);
521 522 523
	return page_get_slab(page);
}

524 525 526 527 528 529
static inline void *index_to_obj(struct kmem_cache *cache, struct slab *slab,
				 unsigned int idx)
{
	return slab->s_mem + cache->buffer_size * idx;
}

530 531 532 533 534 535 536 537
/*
 * 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)
538
{
539 540
	u32 offset = (obj - slab->s_mem);
	return reciprocal_divide(offset, cache->reciprocal_buffer_size);
541 542
}

A
Andrew Morton 已提交
543 544 545
/*
 * These are the default caches for kmalloc. Custom caches can have other sizes.
 */
L
Linus Torvalds 已提交
546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562
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
563
	{NULL,}
L
Linus Torvalds 已提交
564 565 566 567
#undef CACHE
};

static struct arraycache_init initarray_cache __initdata =
P
Pekka Enberg 已提交
568
    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
L
Linus Torvalds 已提交
569
static struct arraycache_init initarray_generic =
P
Pekka Enberg 已提交
570
    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
L
Linus Torvalds 已提交
571 572

/* internal cache of cache description objs */
573
static struct kmem_cache cache_cache = {
P
Pekka Enberg 已提交
574 575 576
	.batchcount = 1,
	.limit = BOOT_CPUCACHE_ENTRIES,
	.shared = 1,
577
	.buffer_size = sizeof(struct kmem_cache),
P
Pekka Enberg 已提交
578
	.name = "kmem_cache",
L
Linus Torvalds 已提交
579 580
};

581 582
#define BAD_ALIEN_MAGIC 0x01020304ul

583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602
/*
 * chicken and egg problem: delay the per-cpu array allocation
 * until the general caches are up.
 */
static enum {
	NONE,
	PARTIAL_AC,
	PARTIAL_L3,
	EARLY,
	FULL
} g_cpucache_up;

/*
 * used by boot code to determine if it can use slab based allocator
 */
int slab_is_available(void)
{
	return g_cpucache_up >= EARLY;
}

603 604 605 606 607 608 609 610
#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.
611 612 613 614
 *
 * 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
615
 */
616 617 618
static struct lock_class_key on_slab_l3_key;
static struct lock_class_key on_slab_alc_key;

619
static void init_node_lock_keys(int q)
620
{
621 622
	struct cache_sizes *s = malloc_sizes;

623 624 625 626 627 628 629 630 631 632
	if (g_cpucache_up != FULL)
		return;

	for (s = malloc_sizes; s->cs_size != ULONG_MAX; s++) {
		struct array_cache **alc;
		struct kmem_list3 *l3;
		int r;

		l3 = s->cs_cachep->nodelists[q];
		if (!l3 || OFF_SLAB(s->cs_cachep))
633
			continue;
634 635 636 637 638 639 640 641 642 643
		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)
644
			continue;
645 646 647 648
		for_each_node(r) {
			if (alc[r])
				lockdep_set_class(&alc[r]->lock,
					&on_slab_alc_key);
649
		}
650 651
	}
}
652 653 654 655 656 657 658 659

static inline void init_lock_keys(void)
{
	int node;

	for_each_node(node)
		init_node_lock_keys(node);
}
660
#else
661 662 663 664
static void init_node_lock_keys(int q)
{
}

665
static inline void init_lock_keys(void)
666 667 668 669
{
}
#endif

670
/*
671
 * Guard access to the cache-chain.
672
 */
I
Ingo Molnar 已提交
673
static DEFINE_MUTEX(cache_chain_mutex);
L
Linus Torvalds 已提交
674 675
static struct list_head cache_chain;

676
static DEFINE_PER_CPU(struct delayed_work, slab_reap_work);
L
Linus Torvalds 已提交
677

678
static inline struct array_cache *cpu_cache_get(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
679 680 681 682
{
	return cachep->array[smp_processor_id()];
}

A
Andrew Morton 已提交
683 684
static inline struct kmem_cache *__find_general_cachep(size_t size,
							gfp_t gfpflags)
L
Linus Torvalds 已提交
685 686 687 688 689
{
	struct cache_sizes *csizep = malloc_sizes;

#if DEBUG
	/* This happens if someone tries to call
P
Pekka Enberg 已提交
690 691 692
	 * kmem_cache_create(), or __kmalloc(), before
	 * the generic caches are initialized.
	 */
693
	BUG_ON(malloc_sizes[INDEX_AC].cs_cachep == NULL);
L
Linus Torvalds 已提交
694
#endif
695 696 697
	if (!size)
		return ZERO_SIZE_PTR;

L
Linus Torvalds 已提交
698 699 700 701
	while (size > csizep->cs_size)
		csizep++;

	/*
702
	 * Really subtle: The last entry with cs->cs_size==ULONG_MAX
L
Linus Torvalds 已提交
703 704 705
	 * has cs_{dma,}cachep==NULL. Thus no special case
	 * for large kmalloc calls required.
	 */
706
#ifdef CONFIG_ZONE_DMA
L
Linus Torvalds 已提交
707 708
	if (unlikely(gfpflags & GFP_DMA))
		return csizep->cs_dmacachep;
709
#endif
L
Linus Torvalds 已提交
710 711 712
	return csizep->cs_cachep;
}

A
Adrian Bunk 已提交
713
static struct kmem_cache *kmem_find_general_cachep(size_t size, gfp_t gfpflags)
714 715 716 717
{
	return __find_general_cachep(size, gfpflags);
}

718
static size_t slab_mgmt_size(size_t nr_objs, size_t align)
L
Linus Torvalds 已提交
719
{
720 721
	return ALIGN(sizeof(struct slab)+nr_objs*sizeof(kmem_bufctl_t), align);
}
L
Linus Torvalds 已提交
722

A
Andrew Morton 已提交
723 724 725
/*
 * Calculate the number of objects and left-over bytes for a given buffer size.
 */
726 727 728 729 730 731 732
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
733

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

784
#define slab_error(cachep, msg) __slab_error(__func__, cachep, msg)
L
Linus Torvalds 已提交
785

A
Andrew Morton 已提交
786 787
static void __slab_error(const char *function, struct kmem_cache *cachep,
			char *msg)
L
Linus Torvalds 已提交
788 789
{
	printk(KERN_ERR "slab error in %s(): cache `%s': %s\n",
P
Pekka Enberg 已提交
790
	       function, cachep->name, msg);
L
Linus Torvalds 已提交
791 792 793
	dump_stack();
}

794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809
/*
 * 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);

810 811 812 813 814 815 816
#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.
 */
817
static DEFINE_PER_CPU(unsigned long, slab_reap_node);
818 819 820 821 822

static void init_reap_node(int cpu)
{
	int node;

823
	node = next_node(cpu_to_mem(cpu), node_online_map);
824
	if (node == MAX_NUMNODES)
825
		node = first_node(node_online_map);
826

827
	per_cpu(slab_reap_node, cpu) = node;
828 829 830 831
}

static void next_reap_node(void)
{
832
	int node = __this_cpu_read(slab_reap_node);
833 834 835 836

	node = next_node(node, node_online_map);
	if (unlikely(node >= MAX_NUMNODES))
		node = first_node(node_online_map);
837
	__this_cpu_write(slab_reap_node, node);
838 839 840 841 842 843 844
}

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

L
Linus Torvalds 已提交
845 846 847 848 849 850 851
/*
 * 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.
 */
852
static void __cpuinit start_cpu_timer(int cpu)
L
Linus Torvalds 已提交
853
{
854
	struct delayed_work *reap_work = &per_cpu(slab_reap_work, cpu);
L
Linus Torvalds 已提交
855 856 857 858 859 860

	/*
	 * When this gets called from do_initcalls via cpucache_init(),
	 * init_workqueues() has already run, so keventd will be setup
	 * at that time.
	 */
861
	if (keventd_up() && reap_work->work.func == NULL) {
862
		init_reap_node(cpu);
863
		INIT_DELAYED_WORK_DEFERRABLE(reap_work, cache_reap);
864 865
		schedule_delayed_work_on(cpu, reap_work,
					__round_jiffies_relative(HZ, cpu));
L
Linus Torvalds 已提交
866 867 868
	}
}

869
static struct array_cache *alloc_arraycache(int node, int entries,
870
					    int batchcount, gfp_t gfp)
L
Linus Torvalds 已提交
871
{
P
Pekka Enberg 已提交
872
	int memsize = sizeof(void *) * entries + sizeof(struct array_cache);
L
Linus Torvalds 已提交
873 874
	struct array_cache *nc = NULL;

875
	nc = kmalloc_node(memsize, gfp, node);
876 877 878 879 880 881 882 883
	/*
	 * The array_cache structures contain pointers to free object.
	 * However, when such objects are allocated or transfered to another
	 * cache the pointers are not cleared and they could be counted as
	 * valid references during a kmemleak scan. Therefore, kmemleak must
	 * not scan such objects.
	 */
	kmemleak_no_scan(nc);
L
Linus Torvalds 已提交
884 885 886 887 888
	if (nc) {
		nc->avail = 0;
		nc->limit = entries;
		nc->batchcount = batchcount;
		nc->touched = 0;
889
		spin_lock_init(&nc->lock);
L
Linus Torvalds 已提交
890 891 892 893
	}
	return nc;
}

894 895 896 897 898 899 900 901 902 903
/*
 * 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 */
904
	int nr = min3(from->avail, max, to->limit - to->avail);
905 906 907 908 909 910 911 912 913 914 915 916

	if (!nr)
		return 0;

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

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

917 918 919 920 921
#ifndef CONFIG_NUMA

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

922
static inline struct array_cache **alloc_alien_cache(int node, int limit, gfp_t gfp)
923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941
{
	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;
}

942
static inline void *____cache_alloc_node(struct kmem_cache *cachep,
943 944 945 946 947 948 949
		 gfp_t flags, int nodeid)
{
	return NULL;
}

#else	/* CONFIG_NUMA */

950
static void *____cache_alloc_node(struct kmem_cache *, gfp_t, int);
951
static void *alternate_node_alloc(struct kmem_cache *, gfp_t);
952

953
static struct array_cache **alloc_alien_cache(int node, int limit, gfp_t gfp)
954 955
{
	struct array_cache **ac_ptr;
956
	int memsize = sizeof(void *) * nr_node_ids;
957 958 959 960
	int i;

	if (limit > 1)
		limit = 12;
961
	ac_ptr = kzalloc_node(memsize, gfp, node);
962 963
	if (ac_ptr) {
		for_each_node(i) {
964
			if (i == node || !node_online(i))
965
				continue;
966
			ac_ptr[i] = alloc_arraycache(node, limit, 0xbaadf00d, gfp);
967
			if (!ac_ptr[i]) {
968
				for (i--; i >= 0; i--)
969 970 971 972 973 974 975 976 977
					kfree(ac_ptr[i]);
				kfree(ac_ptr);
				return NULL;
			}
		}
	}
	return ac_ptr;
}

P
Pekka Enberg 已提交
978
static void free_alien_cache(struct array_cache **ac_ptr)
979 980 981 982 983 984
{
	int i;

	if (!ac_ptr)
		return;
	for_each_node(i)
P
Pekka Enberg 已提交
985
	    kfree(ac_ptr[i]);
986 987 988
	kfree(ac_ptr);
}

989
static void __drain_alien_cache(struct kmem_cache *cachep,
P
Pekka Enberg 已提交
990
				struct array_cache *ac, int node)
991 992 993 994 995
{
	struct kmem_list3 *rl3 = cachep->nodelists[node];

	if (ac->avail) {
		spin_lock(&rl3->list_lock);
996 997 998 999 1000
		/*
		 * 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.
		 */
1001 1002
		if (rl3->shared)
			transfer_objects(rl3->shared, ac, ac->limit);
1003

1004
		free_block(cachep, ac->entry, ac->avail, node);
1005 1006 1007 1008 1009
		ac->avail = 0;
		spin_unlock(&rl3->list_lock);
	}
}

1010 1011 1012 1013 1014
/*
 * Called from cache_reap() to regularly drain alien caches round robin.
 */
static void reap_alien(struct kmem_cache *cachep, struct kmem_list3 *l3)
{
1015
	int node = __this_cpu_read(slab_reap_node);
1016 1017 1018

	if (l3->alien) {
		struct array_cache *ac = l3->alien[node];
1019 1020

		if (ac && ac->avail && spin_trylock_irq(&ac->lock)) {
1021 1022 1023 1024 1025 1026
			__drain_alien_cache(cachep, ac, node);
			spin_unlock_irq(&ac->lock);
		}
	}
}

A
Andrew Morton 已提交
1027 1028
static void drain_alien_cache(struct kmem_cache *cachep,
				struct array_cache **alien)
1029
{
P
Pekka Enberg 已提交
1030
	int i = 0;
1031 1032 1033 1034
	struct array_cache *ac;
	unsigned long flags;

	for_each_online_node(i) {
1035
		ac = alien[i];
1036 1037 1038 1039 1040 1041 1042
		if (ac) {
			spin_lock_irqsave(&ac->lock, flags);
			__drain_alien_cache(cachep, ac, i);
			spin_unlock_irqrestore(&ac->lock, flags);
		}
	}
}
1043

1044
static inline int cache_free_alien(struct kmem_cache *cachep, void *objp)
1045 1046 1047 1048 1049
{
	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
1050 1051
	int node;

1052
	node = numa_mem_id();
1053 1054 1055 1056 1057

	/*
	 * Make sure we are not freeing a object from another node to the array
	 * cache on this cpu.
	 */
1058
	if (likely(slabp->nodeid == node))
1059 1060
		return 0;

P
Pekka Enberg 已提交
1061
	l3 = cachep->nodelists[node];
1062 1063 1064
	STATS_INC_NODEFREES(cachep);
	if (l3->alien && l3->alien[nodeid]) {
		alien = l3->alien[nodeid];
1065
		spin_lock(&alien->lock);
1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
		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;
}
1079 1080
#endif

1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126
/*
 * Allocates and initializes nodelists for a node on each slab cache, used for
 * either memory or cpu hotplug.  If memory is being hot-added, the kmem_list3
 * will be allocated off-node since memory is not yet online for the new node.
 * When hotplugging memory or a cpu, existing nodelists are not replaced if
 * already in use.
 *
 * Must hold cache_chain_mutex.
 */
static int init_cache_nodelists_node(int node)
{
	struct kmem_cache *cachep;
	struct kmem_list3 *l3;
	const int memsize = sizeof(struct kmem_list3);

	list_for_each_entry(cachep, &cache_chain, next) {
		/*
		 * Set up the size64 kmemlist for cpu before we can
		 * begin anything. Make sure some other cpu on this
		 * node has not already allocated this
		 */
		if (!cachep->nodelists[node]) {
			l3 = kmalloc_node(memsize, GFP_KERNEL, node);
			if (!l3)
				return -ENOMEM;
			kmem_list3_init(l3);
			l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
			    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;

			/*
			 * The l3s don't come and go as CPUs come and
			 * go.  cache_chain_mutex is sufficient
			 * protection here.
			 */
			cachep->nodelists[node] = l3;
		}

		spin_lock_irq(&cachep->nodelists[node]->list_lock);
		cachep->nodelists[node]->free_limit =
			(1 + nr_cpus_node(node)) *
			cachep->batchcount + cachep->num;
		spin_unlock_irq(&cachep->nodelists[node]->list_lock);
	}
	return 0;
}

1127 1128 1129 1130
static void __cpuinit cpuup_canceled(long cpu)
{
	struct kmem_cache *cachep;
	struct kmem_list3 *l3 = NULL;
1131
	int node = cpu_to_mem(cpu);
1132
	const struct cpumask *mask = cpumask_of_node(node);
1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153

	list_for_each_entry(cachep, &cache_chain, next) {
		struct array_cache *nc;
		struct array_cache *shared;
		struct array_cache **alien;

		/* cpu is dead; no one can alloc from it. */
		nc = cachep->array[cpu];
		cachep->array[cpu] = NULL;
		l3 = cachep->nodelists[node];

		if (!l3)
			goto free_array_cache;

		spin_lock_irq(&l3->list_lock);

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

1154
		if (!cpumask_empty(mask)) {
1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192
			spin_unlock_irq(&l3->list_lock);
			goto free_array_cache;
		}

		shared = l3->shared;
		if (shared) {
			free_block(cachep, shared->entry,
				   shared->avail, node);
			l3->shared = NULL;
		}

		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);
		}
free_array_cache:
		kfree(nc);
	}
	/*
	 * 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;
		drain_freelist(cachep, l3, l3->free_objects);
	}
}

static int __cpuinit cpuup_prepare(long cpu)
L
Linus Torvalds 已提交
1193
{
1194
	struct kmem_cache *cachep;
1195
	struct kmem_list3 *l3 = NULL;
1196
	int node = cpu_to_mem(cpu);
1197
	int err;
L
Linus Torvalds 已提交
1198

1199 1200 1201 1202 1203 1204
	/*
	 * We need to do this right in the beginning since
	 * alloc_arraycache's are going to use this list.
	 * kmalloc_node allows us to add the slab to the right
	 * kmem_list3 and not this cpu's kmem_list3
	 */
1205 1206 1207
	err = init_cache_nodelists_node(node);
	if (err < 0)
		goto bad;
1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218

	/*
	 * Now we can go ahead with allocating the shared arrays and
	 * array caches
	 */
	list_for_each_entry(cachep, &cache_chain, next) {
		struct array_cache *nc;
		struct array_cache *shared = NULL;
		struct array_cache **alien = NULL;

		nc = alloc_arraycache(node, cachep->limit,
1219
					cachep->batchcount, GFP_KERNEL);
1220 1221 1222 1223 1224
		if (!nc)
			goto bad;
		if (cachep->shared) {
			shared = alloc_arraycache(node,
				cachep->shared * cachep->batchcount,
1225
				0xbaadf00d, GFP_KERNEL);
1226 1227
			if (!shared) {
				kfree(nc);
L
Linus Torvalds 已提交
1228
				goto bad;
1229
			}
1230 1231
		}
		if (use_alien_caches) {
1232
			alien = alloc_alien_cache(node, cachep->limit, GFP_KERNEL);
1233 1234 1235
			if (!alien) {
				kfree(shared);
				kfree(nc);
1236
				goto bad;
1237
			}
1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251
		}
		cachep->array[cpu] = nc;
		l3 = cachep->nodelists[node];
		BUG_ON(!l3);

		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;
		}
1252
#ifdef CONFIG_NUMA
1253 1254 1255
		if (!l3->alien) {
			l3->alien = alien;
			alien = NULL;
L
Linus Torvalds 已提交
1256
		}
1257 1258 1259 1260 1261
#endif
		spin_unlock_irq(&l3->list_lock);
		kfree(shared);
		free_alien_cache(alien);
	}
1262 1263
	init_node_lock_keys(node);

1264 1265
	return 0;
bad:
1266
	cpuup_canceled(cpu);
1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278
	return -ENOMEM;
}

static int __cpuinit cpuup_callback(struct notifier_block *nfb,
				    unsigned long action, void *hcpu)
{
	long cpu = (long)hcpu;
	int err = 0;

	switch (action) {
	case CPU_UP_PREPARE:
	case CPU_UP_PREPARE_FROZEN:
1279
		mutex_lock(&cache_chain_mutex);
1280
		err = cpuup_prepare(cpu);
1281
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1282 1283
		break;
	case CPU_ONLINE:
1284
	case CPU_ONLINE_FROZEN:
L
Linus Torvalds 已提交
1285 1286 1287
		start_cpu_timer(cpu);
		break;
#ifdef CONFIG_HOTPLUG_CPU
1288
  	case CPU_DOWN_PREPARE:
1289
  	case CPU_DOWN_PREPARE_FROZEN:
1290 1291 1292 1293 1294 1295
		/*
		 * Shutdown cache reaper. Note that the cache_chain_mutex is
		 * held so that if cache_reap() is invoked it cannot do
		 * anything expensive but will only modify reap_work
		 * and reschedule the timer.
		*/
1296
		cancel_delayed_work_sync(&per_cpu(slab_reap_work, cpu));
1297
		/* Now the cache_reaper is guaranteed to be not running. */
1298
		per_cpu(slab_reap_work, cpu).work.func = NULL;
1299 1300
  		break;
  	case CPU_DOWN_FAILED:
1301
  	case CPU_DOWN_FAILED_FROZEN:
1302 1303
		start_cpu_timer(cpu);
  		break;
L
Linus Torvalds 已提交
1304
	case CPU_DEAD:
1305
	case CPU_DEAD_FROZEN:
1306 1307 1308 1309 1310 1311 1312 1313
		/*
		 * 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().
		 */
S
Simon Arlott 已提交
1314
		/* fall through */
1315
#endif
L
Linus Torvalds 已提交
1316
	case CPU_UP_CANCELED:
1317
	case CPU_UP_CANCELED_FROZEN:
1318
		mutex_lock(&cache_chain_mutex);
1319
		cpuup_canceled(cpu);
I
Ingo Molnar 已提交
1320
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1321 1322
		break;
	}
1323
	return notifier_from_errno(err);
L
Linus Torvalds 已提交
1324 1325
}

1326 1327 1328
static struct notifier_block __cpuinitdata cpucache_notifier = {
	&cpuup_callback, NULL, 0
};
L
Linus Torvalds 已提交
1329

1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393
#if defined(CONFIG_NUMA) && defined(CONFIG_MEMORY_HOTPLUG)
/*
 * Drains freelist for a node on each slab cache, used for memory hot-remove.
 * Returns -EBUSY if all objects cannot be drained so that the node is not
 * removed.
 *
 * Must hold cache_chain_mutex.
 */
static int __meminit drain_cache_nodelists_node(int node)
{
	struct kmem_cache *cachep;
	int ret = 0;

	list_for_each_entry(cachep, &cache_chain, next) {
		struct kmem_list3 *l3;

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

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

		if (!list_empty(&l3->slabs_full) ||
		    !list_empty(&l3->slabs_partial)) {
			ret = -EBUSY;
			break;
		}
	}
	return ret;
}

static int __meminit slab_memory_callback(struct notifier_block *self,
					unsigned long action, void *arg)
{
	struct memory_notify *mnb = arg;
	int ret = 0;
	int nid;

	nid = mnb->status_change_nid;
	if (nid < 0)
		goto out;

	switch (action) {
	case MEM_GOING_ONLINE:
		mutex_lock(&cache_chain_mutex);
		ret = init_cache_nodelists_node(nid);
		mutex_unlock(&cache_chain_mutex);
		break;
	case MEM_GOING_OFFLINE:
		mutex_lock(&cache_chain_mutex);
		ret = drain_cache_nodelists_node(nid);
		mutex_unlock(&cache_chain_mutex);
		break;
	case MEM_ONLINE:
	case MEM_OFFLINE:
	case MEM_CANCEL_ONLINE:
	case MEM_CANCEL_OFFLINE:
		break;
	}
out:
	return ret ? notifier_from_errno(ret) : NOTIFY_OK;
}
#endif /* CONFIG_NUMA && CONFIG_MEMORY_HOTPLUG */

1394 1395 1396
/*
 * swap the static kmem_list3 with kmalloced memory
 */
1397 1398
static void __init init_list(struct kmem_cache *cachep, struct kmem_list3 *list,
				int nodeid)
1399 1400 1401
{
	struct kmem_list3 *ptr;

1402
	ptr = kmalloc_node(sizeof(struct kmem_list3), GFP_NOWAIT, nodeid);
1403 1404 1405
	BUG_ON(!ptr);

	memcpy(ptr, list, sizeof(struct kmem_list3));
1406 1407 1408 1409 1410
	/*
	 * Do not assume that spinlocks can be initialized via memcpy:
	 */
	spin_lock_init(&ptr->list_lock);

1411 1412 1413 1414
	MAKE_ALL_LISTS(cachep, ptr, nodeid);
	cachep->nodelists[nodeid] = ptr;
}

1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430
/*
 * For setting up all the kmem_list3s for cache whose buffer_size is same as
 * size of kmem_list3.
 */
static void __init set_up_list3s(struct kmem_cache *cachep, int index)
{
	int node;

	for_each_online_node(node) {
		cachep->nodelists[node] = &initkmem_list3[index + node];
		cachep->nodelists[node]->next_reap = jiffies +
		    REAPTIMEOUT_LIST3 +
		    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
	}
}

A
Andrew Morton 已提交
1431 1432 1433
/*
 * Initialisation.  Called after the page allocator have been initialised and
 * before smp_init().
L
Linus Torvalds 已提交
1434 1435 1436 1437 1438 1439
 */
void __init kmem_cache_init(void)
{
	size_t left_over;
	struct cache_sizes *sizes;
	struct cache_names *names;
1440
	int i;
1441
	int order;
P
Pekka Enberg 已提交
1442
	int node;
1443

1444
	if (num_possible_nodes() == 1)
1445 1446
		use_alien_caches = 0;

1447 1448 1449 1450 1451
	for (i = 0; i < NUM_INIT_LISTS; i++) {
		kmem_list3_init(&initkmem_list3[i]);
		if (i < MAX_NUMNODES)
			cache_cache.nodelists[i] = NULL;
	}
1452
	set_up_list3s(&cache_cache, CACHE_CACHE);
L
Linus Torvalds 已提交
1453 1454 1455 1456 1457

	/*
	 * Fragmentation resistance on low memory - only use bigger
	 * page orders on machines with more than 32MB of memory.
	 */
1458
	if (totalram_pages > (32 << 20) >> PAGE_SHIFT)
L
Linus Torvalds 已提交
1459 1460 1461 1462
		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
1463 1464 1465
	 * 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.
1466 1467 1468
	 *    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
1469
	 * 2) Create the first kmalloc cache.
1470
	 *    The struct kmem_cache for the new cache is allocated normally.
1471 1472 1473
	 *    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
1474 1475
	 * 4) Replace the __init data head arrays for cache_cache and the first
	 *    kmalloc cache with kmalloc allocated arrays.
1476 1477 1478
	 * 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
1479 1480
	 */

1481
	node = numa_mem_id();
P
Pekka Enberg 已提交
1482

L
Linus Torvalds 已提交
1483 1484 1485 1486 1487
	/* 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;
1488
	cache_cache.nodelists[node] = &initkmem_list3[CACHE_CACHE + node];
L
Linus Torvalds 已提交
1489

E
Eric Dumazet 已提交
1490 1491 1492 1493 1494 1495 1496 1497 1498
	/*
	 * struct kmem_cache size depends on nr_node_ids, which
	 * can be less than MAX_NUMNODES.
	 */
	cache_cache.buffer_size = offsetof(struct kmem_cache, nodelists) +
				 nr_node_ids * sizeof(struct kmem_list3 *);
#if DEBUG
	cache_cache.obj_size = cache_cache.buffer_size;
#endif
A
Andrew Morton 已提交
1499 1500
	cache_cache.buffer_size = ALIGN(cache_cache.buffer_size,
					cache_line_size());
1501 1502
	cache_cache.reciprocal_buffer_size =
		reciprocal_value(cache_cache.buffer_size);
L
Linus Torvalds 已提交
1503

1504 1505 1506 1507 1508 1509
	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;
	}
1510
	BUG_ON(!cache_cache.num);
1511
	cache_cache.gfporder = order;
P
Pekka Enberg 已提交
1512 1513 1514
	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
1515 1516 1517 1518 1519

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

A
Andrew Morton 已提交
1520 1521 1522 1523
	/*
	 * Initialize the caches that provide memory for the array cache and the
	 * kmem_list3 structures first.  Without this, further allocations will
	 * bug.
1524 1525 1526
	 */

	sizes[INDEX_AC].cs_cachep = kmem_cache_create(names[INDEX_AC].name,
A
Andrew Morton 已提交
1527 1528 1529
					sizes[INDEX_AC].cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_PANIC,
1530
					NULL);
1531

A
Andrew Morton 已提交
1532
	if (INDEX_AC != INDEX_L3) {
1533
		sizes[INDEX_L3].cs_cachep =
A
Andrew Morton 已提交
1534 1535 1536 1537
			kmem_cache_create(names[INDEX_L3].name,
				sizes[INDEX_L3].cs_size,
				ARCH_KMALLOC_MINALIGN,
				ARCH_KMALLOC_FLAGS|SLAB_PANIC,
1538
				NULL);
A
Andrew Morton 已提交
1539
	}
1540

1541 1542
	slab_early_init = 0;

L
Linus Torvalds 已提交
1543
	while (sizes->cs_size != ULONG_MAX) {
1544 1545
		/*
		 * For performance, all the general caches are L1 aligned.
L
Linus Torvalds 已提交
1546 1547 1548
		 * This should be particularly beneficial on SMP boxes, as it
		 * eliminates "false sharing".
		 * Note for systems short on memory removing the alignment will
1549 1550
		 * allow tighter packing of the smaller caches.
		 */
A
Andrew Morton 已提交
1551
		if (!sizes->cs_cachep) {
1552
			sizes->cs_cachep = kmem_cache_create(names->name,
A
Andrew Morton 已提交
1553 1554 1555
					sizes->cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_PANIC,
1556
					NULL);
A
Andrew Morton 已提交
1557
		}
1558 1559 1560
#ifdef CONFIG_ZONE_DMA
		sizes->cs_dmacachep = kmem_cache_create(
					names->name_dma,
A
Andrew Morton 已提交
1561 1562 1563 1564
					sizes->cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_CACHE_DMA|
						SLAB_PANIC,
1565
					NULL);
1566
#endif
L
Linus Torvalds 已提交
1567 1568 1569 1570 1571
		sizes++;
		names++;
	}
	/* 4) Replace the bootstrap head arrays */
	{
1572
		struct array_cache *ptr;
1573

1574
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_NOWAIT);
1575

1576 1577
		BUG_ON(cpu_cache_get(&cache_cache) != &initarray_cache.cache);
		memcpy(ptr, cpu_cache_get(&cache_cache),
P
Pekka Enberg 已提交
1578
		       sizeof(struct arraycache_init));
1579 1580 1581 1582 1583
		/*
		 * Do not assume that spinlocks can be initialized via memcpy:
		 */
		spin_lock_init(&ptr->lock);

L
Linus Torvalds 已提交
1584
		cache_cache.array[smp_processor_id()] = ptr;
1585

1586
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_NOWAIT);
1587

1588
		BUG_ON(cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep)
P
Pekka Enberg 已提交
1589
		       != &initarray_generic.cache);
1590
		memcpy(ptr, cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep),
P
Pekka Enberg 已提交
1591
		       sizeof(struct arraycache_init));
1592 1593 1594 1595 1596
		/*
		 * Do not assume that spinlocks can be initialized via memcpy:
		 */
		spin_lock_init(&ptr->lock);

1597
		malloc_sizes[INDEX_AC].cs_cachep->array[smp_processor_id()] =
P
Pekka Enberg 已提交
1598
		    ptr;
L
Linus Torvalds 已提交
1599
	}
1600 1601
	/* 5) Replace the bootstrap kmem_list3's */
	{
P
Pekka Enberg 已提交
1602 1603
		int nid;

1604
		for_each_online_node(nid) {
1605
			init_list(&cache_cache, &initkmem_list3[CACHE_CACHE + nid], nid);
1606

1607
			init_list(malloc_sizes[INDEX_AC].cs_cachep,
P
Pekka Enberg 已提交
1608
				  &initkmem_list3[SIZE_AC + nid], nid);
1609 1610 1611

			if (INDEX_AC != INDEX_L3) {
				init_list(malloc_sizes[INDEX_L3].cs_cachep,
P
Pekka Enberg 已提交
1612
					  &initkmem_list3[SIZE_L3 + nid], nid);
1613 1614 1615
			}
		}
	}
L
Linus Torvalds 已提交
1616

1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629
	g_cpucache_up = EARLY;
}

void __init kmem_cache_init_late(void)
{
	struct kmem_cache *cachep;

	/* 6) resize the head arrays to their final sizes */
	mutex_lock(&cache_chain_mutex);
	list_for_each_entry(cachep, &cache_chain, next)
		if (enable_cpucache(cachep, GFP_NOWAIT))
			BUG();
	mutex_unlock(&cache_chain_mutex);
1630

L
Linus Torvalds 已提交
1631 1632 1633
	/* Done! */
	g_cpucache_up = FULL;

P
Pekka Enberg 已提交
1634 1635 1636
	/* Annotate slab for lockdep -- annotate the malloc caches */
	init_lock_keys();

A
Andrew Morton 已提交
1637 1638 1639
	/*
	 * Register a cpu startup notifier callback that initializes
	 * cpu_cache_get for all new cpus
L
Linus Torvalds 已提交
1640 1641 1642
	 */
	register_cpu_notifier(&cpucache_notifier);

1643 1644 1645 1646 1647 1648 1649 1650
#ifdef CONFIG_NUMA
	/*
	 * Register a memory hotplug callback that initializes and frees
	 * nodelists.
	 */
	hotplug_memory_notifier(slab_memory_callback, SLAB_CALLBACK_PRI);
#endif

A
Andrew Morton 已提交
1651 1652 1653
	/*
	 * 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
1654 1655 1656 1657 1658 1659 1660
	 */
}

static int __init cpucache_init(void)
{
	int cpu;

A
Andrew Morton 已提交
1661 1662
	/*
	 * Register the timers that return unneeded pages to the page allocator
L
Linus Torvalds 已提交
1663
	 */
1664
	for_each_online_cpu(cpu)
A
Andrew Morton 已提交
1665
		start_cpu_timer(cpu);
L
Linus Torvalds 已提交
1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676
	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.
 */
1677
static void *kmem_getpages(struct kmem_cache *cachep, gfp_t flags, int nodeid)
L
Linus Torvalds 已提交
1678 1679
{
	struct page *page;
1680
	int nr_pages;
L
Linus Torvalds 已提交
1681 1682
	int i;

1683
#ifndef CONFIG_MMU
1684 1685 1686
	/*
	 * Nommu uses slab's for process anonymous memory allocations, and thus
	 * requires __GFP_COMP to properly refcount higher order allocations
1687
	 */
1688
	flags |= __GFP_COMP;
1689
#endif
1690

1691
	flags |= cachep->gfpflags;
1692 1693
	if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
		flags |= __GFP_RECLAIMABLE;
1694

L
Linus Torvalds 已提交
1695
	page = alloc_pages_exact_node(nodeid, flags | __GFP_NOTRACK, cachep->gfporder);
L
Linus Torvalds 已提交
1696 1697 1698
	if (!page)
		return NULL;

1699
	nr_pages = (1 << cachep->gfporder);
L
Linus Torvalds 已提交
1700
	if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
1701 1702 1703 1704 1705
		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);
1706 1707
	for (i = 0; i < nr_pages; i++)
		__SetPageSlab(page + i);
P
Pekka Enberg 已提交
1708

1709 1710 1711 1712 1713 1714 1715 1716
	if (kmemcheck_enabled && !(cachep->flags & SLAB_NOTRACK)) {
		kmemcheck_alloc_shadow(page, cachep->gfporder, flags, nodeid);

		if (cachep->ctor)
			kmemcheck_mark_uninitialized_pages(page, nr_pages);
		else
			kmemcheck_mark_unallocated_pages(page, nr_pages);
	}
P
Pekka Enberg 已提交
1717

1718
	return page_address(page);
L
Linus Torvalds 已提交
1719 1720 1721 1722 1723
}

/*
 * Interface to system's page release.
 */
1724
static void kmem_freepages(struct kmem_cache *cachep, void *addr)
L
Linus Torvalds 已提交
1725
{
P
Pekka Enberg 已提交
1726
	unsigned long i = (1 << cachep->gfporder);
L
Linus Torvalds 已提交
1727 1728 1729
	struct page *page = virt_to_page(addr);
	const unsigned long nr_freed = i;

1730
	kmemcheck_free_shadow(page, cachep->gfporder);
P
Pekka Enberg 已提交
1731

1732 1733 1734 1735 1736 1737
	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
1738
	while (i--) {
N
Nick Piggin 已提交
1739 1740
		BUG_ON(!PageSlab(page));
		__ClearPageSlab(page);
L
Linus Torvalds 已提交
1741 1742 1743 1744 1745 1746 1747 1748 1749
		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
1750
	struct slab_rcu *slab_rcu = (struct slab_rcu *)head;
1751
	struct kmem_cache *cachep = slab_rcu->cachep;
L
Linus Torvalds 已提交
1752 1753 1754 1755 1756 1757 1758 1759 1760

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

#if DEBUG

#ifdef CONFIG_DEBUG_PAGEALLOC
1761
static void store_stackinfo(struct kmem_cache *cachep, unsigned long *addr,
P
Pekka Enberg 已提交
1762
			    unsigned long caller)
L
Linus Torvalds 已提交
1763
{
1764
	int size = obj_size(cachep);
L
Linus Torvalds 已提交
1765

1766
	addr = (unsigned long *)&((char *)addr)[obj_offset(cachep)];
L
Linus Torvalds 已提交
1767

P
Pekka Enberg 已提交
1768
	if (size < 5 * sizeof(unsigned long))
L
Linus Torvalds 已提交
1769 1770
		return;

P
Pekka Enberg 已提交
1771 1772 1773 1774
	*addr++ = 0x12345678;
	*addr++ = caller;
	*addr++ = smp_processor_id();
	size -= 3 * sizeof(unsigned long);
L
Linus Torvalds 已提交
1775 1776 1777 1778 1779 1780 1781
	{
		unsigned long *sptr = &caller;
		unsigned long svalue;

		while (!kstack_end(sptr)) {
			svalue = *sptr++;
			if (kernel_text_address(svalue)) {
P
Pekka Enberg 已提交
1782
				*addr++ = svalue;
L
Linus Torvalds 已提交
1783 1784 1785 1786 1787 1788 1789
				size -= sizeof(unsigned long);
				if (size <= sizeof(unsigned long))
					break;
			}
		}

	}
P
Pekka Enberg 已提交
1790
	*addr++ = 0x87654321;
L
Linus Torvalds 已提交
1791 1792 1793
}
#endif

1794
static void poison_obj(struct kmem_cache *cachep, void *addr, unsigned char val)
L
Linus Torvalds 已提交
1795
{
1796 1797
	int size = obj_size(cachep);
	addr = &((char *)addr)[obj_offset(cachep)];
L
Linus Torvalds 已提交
1798 1799

	memset(addr, val, size);
P
Pekka Enberg 已提交
1800
	*(unsigned char *)(addr + size - 1) = POISON_END;
L
Linus Torvalds 已提交
1801 1802 1803 1804 1805
}

static void dump_line(char *data, int offset, int limit)
{
	int i;
D
Dave Jones 已提交
1806 1807 1808
	unsigned char error = 0;
	int bad_count = 0;

L
Linus Torvalds 已提交
1809
	printk(KERN_ERR "%03x:", offset);
D
Dave Jones 已提交
1810 1811 1812 1813 1814
	for (i = 0; i < limit; i++) {
		if (data[offset + i] != POISON_FREE) {
			error = data[offset + i];
			bad_count++;
		}
P
Pekka Enberg 已提交
1815
		printk(" %02x", (unsigned char)data[offset + i]);
D
Dave Jones 已提交
1816
	}
L
Linus Torvalds 已提交
1817
	printk("\n");
D
Dave Jones 已提交
1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831

	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
1832 1833 1834 1835 1836
}
#endif

#if DEBUG

1837
static void print_objinfo(struct kmem_cache *cachep, void *objp, int lines)
L
Linus Torvalds 已提交
1838 1839 1840 1841 1842
{
	int i, size;
	char *realobj;

	if (cachep->flags & SLAB_RED_ZONE) {
1843
		printk(KERN_ERR "Redzone: 0x%llx/0x%llx.\n",
A
Andrew Morton 已提交
1844 1845
			*dbg_redzone1(cachep, objp),
			*dbg_redzone2(cachep, objp));
L
Linus Torvalds 已提交
1846 1847 1848 1849
	}

	if (cachep->flags & SLAB_STORE_USER) {
		printk(KERN_ERR "Last user: [<%p>]",
A
Andrew Morton 已提交
1850
			*dbg_userword(cachep, objp));
L
Linus Torvalds 已提交
1851
		print_symbol("(%s)",
A
Andrew Morton 已提交
1852
				(unsigned long)*dbg_userword(cachep, objp));
L
Linus Torvalds 已提交
1853 1854
		printk("\n");
	}
1855 1856
	realobj = (char *)objp + obj_offset(cachep);
	size = obj_size(cachep);
P
Pekka Enberg 已提交
1857
	for (i = 0; i < size && lines; i += 16, lines--) {
L
Linus Torvalds 已提交
1858 1859
		int limit;
		limit = 16;
P
Pekka Enberg 已提交
1860 1861
		if (i + limit > size)
			limit = size - i;
L
Linus Torvalds 已提交
1862 1863 1864 1865
		dump_line(realobj, i, limit);
	}
}

1866
static void check_poison_obj(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
1867 1868 1869 1870 1871
{
	char *realobj;
	int size, i;
	int lines = 0;

1872 1873
	realobj = (char *)objp + obj_offset(cachep);
	size = obj_size(cachep);
L
Linus Torvalds 已提交
1874

P
Pekka Enberg 已提交
1875
	for (i = 0; i < size; i++) {
L
Linus Torvalds 已提交
1876
		char exp = POISON_FREE;
P
Pekka Enberg 已提交
1877
		if (i == size - 1)
L
Linus Torvalds 已提交
1878 1879 1880 1881 1882 1883
			exp = POISON_END;
		if (realobj[i] != exp) {
			int limit;
			/* Mismatch ! */
			/* Print header */
			if (lines == 0) {
P
Pekka Enberg 已提交
1884
				printk(KERN_ERR
1885 1886
					"Slab corruption: %s start=%p, len=%d\n",
					cachep->name, realobj, size);
L
Linus Torvalds 已提交
1887 1888 1889
				print_objinfo(cachep, objp, 0);
			}
			/* Hexdump the affected line */
P
Pekka Enberg 已提交
1890
			i = (i / 16) * 16;
L
Linus Torvalds 已提交
1891
			limit = 16;
P
Pekka Enberg 已提交
1892 1893
			if (i + limit > size)
				limit = size - i;
L
Linus Torvalds 已提交
1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905
			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:
		 */
1906
		struct slab *slabp = virt_to_slab(objp);
1907
		unsigned int objnr;
L
Linus Torvalds 已提交
1908

1909
		objnr = obj_to_index(cachep, slabp, objp);
L
Linus Torvalds 已提交
1910
		if (objnr) {
1911
			objp = index_to_obj(cachep, slabp, objnr - 1);
1912
			realobj = (char *)objp + obj_offset(cachep);
L
Linus Torvalds 已提交
1913
			printk(KERN_ERR "Prev obj: start=%p, len=%d\n",
P
Pekka Enberg 已提交
1914
			       realobj, size);
L
Linus Torvalds 已提交
1915 1916
			print_objinfo(cachep, objp, 2);
		}
P
Pekka Enberg 已提交
1917
		if (objnr + 1 < cachep->num) {
1918
			objp = index_to_obj(cachep, slabp, objnr + 1);
1919
			realobj = (char *)objp + obj_offset(cachep);
L
Linus Torvalds 已提交
1920
			printk(KERN_ERR "Next obj: start=%p, len=%d\n",
P
Pekka Enberg 已提交
1921
			       realobj, size);
L
Linus Torvalds 已提交
1922 1923 1924 1925 1926 1927
			print_objinfo(cachep, objp, 2);
		}
	}
}
#endif

1928
#if DEBUG
R
Rabin Vincent 已提交
1929
static void slab_destroy_debugcheck(struct kmem_cache *cachep, struct slab *slabp)
L
Linus Torvalds 已提交
1930 1931 1932
{
	int i;
	for (i = 0; i < cachep->num; i++) {
1933
		void *objp = index_to_obj(cachep, slabp, i);
L
Linus Torvalds 已提交
1934 1935 1936

		if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
A
Andrew Morton 已提交
1937 1938
			if (cachep->buffer_size % PAGE_SIZE == 0 &&
					OFF_SLAB(cachep))
P
Pekka Enberg 已提交
1939
				kernel_map_pages(virt_to_page(objp),
A
Andrew Morton 已提交
1940
					cachep->buffer_size / PAGE_SIZE, 1);
L
Linus Torvalds 已提交
1941 1942 1943 1944 1945 1946 1947 1948 1949
			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
1950
					   "was overwritten");
L
Linus Torvalds 已提交
1951 1952
			if (*dbg_redzone2(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "end of a freed object "
P
Pekka Enberg 已提交
1953
					   "was overwritten");
L
Linus Torvalds 已提交
1954 1955
		}
	}
1956
}
L
Linus Torvalds 已提交
1957
#else
R
Rabin Vincent 已提交
1958
static void slab_destroy_debugcheck(struct kmem_cache *cachep, struct slab *slabp)
1959 1960
{
}
L
Linus Torvalds 已提交
1961 1962
#endif

1963 1964 1965 1966 1967
/**
 * slab_destroy - destroy and release all objects in a slab
 * @cachep: cache pointer being destroyed
 * @slabp: slab pointer being destroyed
 *
1968
 * Destroy all the objs in a slab, and release the mem back to the system.
A
Andrew Morton 已提交
1969 1970
 * Before calling the slab must have been unlinked from the cache.  The
 * cache-lock is not held/needed.
1971
 */
1972
static void slab_destroy(struct kmem_cache *cachep, struct slab *slabp)
1973 1974 1975
{
	void *addr = slabp->s_mem - slabp->colouroff;

R
Rabin Vincent 已提交
1976
	slab_destroy_debugcheck(cachep, slabp);
L
Linus Torvalds 已提交
1977 1978 1979
	if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU)) {
		struct slab_rcu *slab_rcu;

P
Pekka Enberg 已提交
1980
		slab_rcu = (struct slab_rcu *)slabp;
L
Linus Torvalds 已提交
1981 1982 1983 1984 1985
		slab_rcu->cachep = cachep;
		slab_rcu->addr = addr;
		call_rcu(&slab_rcu->head, kmem_rcu_free);
	} else {
		kmem_freepages(cachep, addr);
1986 1987
		if (OFF_SLAB(cachep))
			kmem_cache_free(cachep->slabp_cache, slabp);
L
Linus Torvalds 已提交
1988 1989 1990
	}
}

1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
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);
}


2012
/**
2013 2014 2015 2016 2017 2018 2019
 * 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.
2020 2021 2022 2023 2024
 *
 * 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
2025
static size_t calculate_slab_order(struct kmem_cache *cachep,
R
Randy Dunlap 已提交
2026
			size_t size, size_t align, unsigned long flags)
2027
{
2028
	unsigned long offslab_limit;
2029
	size_t left_over = 0;
2030
	int gfporder;
2031

2032
	for (gfporder = 0; gfporder <= KMALLOC_MAX_ORDER; gfporder++) {
2033 2034 2035
		unsigned int num;
		size_t remainder;

2036
		cache_estimate(gfporder, size, align, flags, &remainder, &num);
2037 2038
		if (!num)
			continue;
2039

2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051
		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;
		}
2052

2053
		/* Found something acceptable - save it away */
2054
		cachep->num = num;
2055
		cachep->gfporder = gfporder;
2056 2057
		left_over = remainder;

2058 2059 2060 2061 2062 2063 2064 2065
		/*
		 * 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;

2066 2067 2068 2069
		/*
		 * Large number of objects is good, but very large slabs are
		 * currently bad for the gfp()s.
		 */
2070
		if (gfporder >= slab_break_gfp_order)
2071 2072
			break;

2073 2074 2075
		/*
		 * Acceptable internal fragmentation?
		 */
A
Andrew Morton 已提交
2076
		if (left_over * 8 <= (PAGE_SIZE << gfporder))
2077 2078 2079 2080 2081
			break;
	}
	return left_over;
}

2082
static int __init_refok setup_cpu_cache(struct kmem_cache *cachep, gfp_t gfp)
2083
{
2084
	if (g_cpucache_up == FULL)
2085
		return enable_cpucache(cachep, gfp);
2086

2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106
	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()] =
2107
			kmalloc(sizeof(struct arraycache_init), gfp);
2108 2109 2110 2111 2112 2113

		if (g_cpucache_up == PARTIAL_AC) {
			set_up_list3s(cachep, SIZE_L3);
			g_cpucache_up = PARTIAL_L3;
		} else {
			int node;
2114
			for_each_online_node(node) {
2115 2116
				cachep->nodelists[node] =
				    kmalloc_node(sizeof(struct kmem_list3),
2117
						gfp, node);
2118 2119 2120 2121 2122
				BUG_ON(!cachep->nodelists[node]);
				kmem_list3_init(cachep->nodelists[node]);
			}
		}
	}
2123
	cachep->nodelists[numa_mem_id()]->next_reap =
2124 2125 2126 2127 2128 2129 2130 2131 2132
			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;
2133
	return 0;
2134 2135
}

L
Linus Torvalds 已提交
2136 2137 2138 2139 2140 2141 2142 2143 2144 2145
/**
 * 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.
 *
 * Returns a ptr to the cache on success, NULL on failure.
 * Cannot be called within a int, but can be interrupted.
2146
 * The @ctor is run when new pages are allocated by the cache.
L
Linus Torvalds 已提交
2147 2148
 *
 * @name must be valid until the cache is destroyed. This implies that
A
Andrew Morton 已提交
2149
 * the module calling this has to destroy the cache before getting unloaded.
2150 2151
 * Note that kmem_cache_name() is not guaranteed to return the same pointer,
 * therefore applications must manage it themselves.
A
Andrew Morton 已提交
2152
 *
L
Linus Torvalds 已提交
2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164
 * 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.
 */
2165
struct kmem_cache *
L
Linus Torvalds 已提交
2166
kmem_cache_create (const char *name, size_t size, size_t align,
2167
	unsigned long flags, void (*ctor)(void *))
L
Linus Torvalds 已提交
2168 2169
{
	size_t left_over, slab_size, ralign;
2170
	struct kmem_cache *cachep = NULL, *pc;
2171
	gfp_t gfp;
L
Linus Torvalds 已提交
2172 2173 2174 2175

	/*
	 * Sanity checks... these are all serious usage bugs.
	 */
A
Andrew Morton 已提交
2176
	if (!name || in_interrupt() || (size < BYTES_PER_WORD) ||
2177
	    size > KMALLOC_MAX_SIZE) {
2178
		printk(KERN_ERR "%s: Early error in slab %s\n", __func__,
A
Andrew Morton 已提交
2179
				name);
P
Pekka Enberg 已提交
2180 2181
		BUG();
	}
L
Linus Torvalds 已提交
2182

2183
	/*
2184
	 * We use cache_chain_mutex to ensure a consistent view of
R
Rusty Russell 已提交
2185
	 * cpu_online_mask as well.  Please see cpuup_callback
2186
	 */
2187 2188 2189 2190
	if (slab_is_available()) {
		get_online_cpus();
		mutex_lock(&cache_chain_mutex);
	}
2191

2192
	list_for_each_entry(pc, &cache_chain, next) {
2193 2194 2195 2196 2197 2198 2199 2200
		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.
		 */
2201
		res = probe_kernel_address(pc->name, tmp);
2202
		if (res) {
2203 2204
			printk(KERN_ERR
			       "SLAB: cache with size %d has lost its name\n",
2205
			       pc->buffer_size);
2206 2207 2208
			continue;
		}

P
Pekka Enberg 已提交
2209
		if (!strcmp(pc->name, name)) {
2210 2211
			printk(KERN_ERR
			       "kmem_cache_create: duplicate cache %s\n", name);
2212 2213 2214 2215 2216
			dump_stack();
			goto oops;
		}
	}

L
Linus Torvalds 已提交
2217 2218 2219 2220 2221 2222 2223 2224 2225
#if DEBUG
	WARN_ON(strchr(name, ' '));	/* It confuses parsers */
#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.
	 */
D
David Woodhouse 已提交
2226 2227
	if (size < 4096 || fls(size - 1) == fls(size-1 + REDZONE_ALIGN +
						2 * sizeof(unsigned long long)))
P
Pekka Enberg 已提交
2228
		flags |= SLAB_RED_ZONE | SLAB_STORE_USER;
L
Linus Torvalds 已提交
2229 2230 2231 2232 2233 2234 2235
	if (!(flags & SLAB_DESTROY_BY_RCU))
		flags |= SLAB_POISON;
#endif
	if (flags & SLAB_DESTROY_BY_RCU)
		BUG_ON(flags & SLAB_POISON);
#endif
	/*
A
Andrew Morton 已提交
2236 2237
	 * Always checks flags, a caller might be expecting debug support which
	 * isn't available.
L
Linus Torvalds 已提交
2238
	 */
2239
	BUG_ON(flags & ~CREATE_MASK);
L
Linus Torvalds 已提交
2240

A
Andrew Morton 已提交
2241 2242
	/*
	 * Check that size is in terms of words.  This is needed to avoid
L
Linus Torvalds 已提交
2243 2244 2245
	 * 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
2246 2247 2248
	if (size & (BYTES_PER_WORD - 1)) {
		size += (BYTES_PER_WORD - 1);
		size &= ~(BYTES_PER_WORD - 1);
L
Linus Torvalds 已提交
2249 2250
	}

A
Andrew Morton 已提交
2251 2252
	/* calculate the final buffer alignment: */

L
Linus Torvalds 已提交
2253 2254
	/* 1) arch recommendation: can be overridden for debug */
	if (flags & SLAB_HWCACHE_ALIGN) {
A
Andrew Morton 已提交
2255 2256 2257 2258
		/*
		 * 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
2259 2260
		 */
		ralign = cache_line_size();
P
Pekka Enberg 已提交
2261
		while (size <= ralign / 2)
L
Linus Torvalds 已提交
2262 2263 2264 2265
			ralign /= 2;
	} else {
		ralign = BYTES_PER_WORD;
	}
2266 2267

	/*
D
David Woodhouse 已提交
2268 2269 2270
	 * Redzoning and user store require word alignment or possibly larger.
	 * Note this will be overridden by architecture or caller mandated
	 * alignment if either is greater than BYTES_PER_WORD.
2271
	 */
D
David Woodhouse 已提交
2272 2273 2274 2275 2276 2277 2278 2279 2280 2281
	if (flags & SLAB_STORE_USER)
		ralign = BYTES_PER_WORD;

	if (flags & SLAB_RED_ZONE) {
		ralign = REDZONE_ALIGN;
		/* If redzoning, ensure that the second redzone is suitably
		 * aligned, by adjusting the object size accordingly. */
		size += REDZONE_ALIGN - 1;
		size &= ~(REDZONE_ALIGN - 1);
	}
2282

2283
	/* 2) arch mandated alignment */
L
Linus Torvalds 已提交
2284 2285 2286
	if (ralign < ARCH_SLAB_MINALIGN) {
		ralign = ARCH_SLAB_MINALIGN;
	}
2287
	/* 3) caller mandated alignment */
L
Linus Torvalds 已提交
2288 2289 2290
	if (ralign < align) {
		ralign = align;
	}
2291 2292
	/* disable debug if not aligning with REDZONE_ALIGN */
	if (ralign & (__alignof__(unsigned long long) - 1))
2293
		flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
A
Andrew Morton 已提交
2294
	/*
2295
	 * 4) Store it.
L
Linus Torvalds 已提交
2296 2297 2298
	 */
	align = ralign;

2299 2300 2301 2302 2303
	if (slab_is_available())
		gfp = GFP_KERNEL;
	else
		gfp = GFP_NOWAIT;

L
Linus Torvalds 已提交
2304
	/* Get cache's description obj. */
2305
	cachep = kmem_cache_zalloc(&cache_cache, gfp);
L
Linus Torvalds 已提交
2306
	if (!cachep)
2307
		goto oops;
L
Linus Torvalds 已提交
2308 2309

#if DEBUG
2310
	cachep->obj_size = size;
L
Linus Torvalds 已提交
2311

2312 2313 2314 2315
	/*
	 * Both debugging options require word-alignment which is calculated
	 * into align above.
	 */
L
Linus Torvalds 已提交
2316 2317
	if (flags & SLAB_RED_ZONE) {
		/* add space for red zone words */
2318 2319
		cachep->obj_offset += align;
		size += align + sizeof(unsigned long long);
L
Linus Torvalds 已提交
2320 2321
	}
	if (flags & SLAB_STORE_USER) {
2322
		/* user store requires one word storage behind the end of
D
David Woodhouse 已提交
2323 2324
		 * the real object. But if the second red zone needs to be
		 * aligned to 64 bits, we must allow that much space.
L
Linus Torvalds 已提交
2325
		 */
D
David Woodhouse 已提交
2326 2327 2328 2329
		if (flags & SLAB_RED_ZONE)
			size += REDZONE_ALIGN;
		else
			size += BYTES_PER_WORD;
L
Linus Torvalds 已提交
2330 2331
	}
#if FORCED_DEBUG && defined(CONFIG_DEBUG_PAGEALLOC)
P
Pekka Enberg 已提交
2332
	if (size >= malloc_sizes[INDEX_L3 + 1].cs_size
C
Carsten Otte 已提交
2333 2334
	    && cachep->obj_size > cache_line_size() && ALIGN(size, align) < PAGE_SIZE) {
		cachep->obj_offset += PAGE_SIZE - ALIGN(size, align);
L
Linus Torvalds 已提交
2335 2336 2337 2338 2339
		size = PAGE_SIZE;
	}
#endif
#endif

2340 2341 2342
	/*
	 * Determine if the slab management is 'on' or 'off' slab.
	 * (bootstrapping cannot cope with offslab caches so don't do
2343 2344
	 * it too early on. Always use on-slab management when
	 * SLAB_NOLEAKTRACE to avoid recursive calls into kmemleak)
2345
	 */
2346 2347
	if ((size >= (PAGE_SIZE >> 3)) && !slab_early_init &&
	    !(flags & SLAB_NOLEAKTRACE))
L
Linus Torvalds 已提交
2348 2349 2350 2351 2352 2353 2354 2355
		/*
		 * 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);

2356
	left_over = calculate_slab_order(cachep, size, align, flags);
L
Linus Torvalds 已提交
2357 2358

	if (!cachep->num) {
2359 2360
		printk(KERN_ERR
		       "kmem_cache_create: couldn't create cache %s.\n", name);
L
Linus Torvalds 已提交
2361 2362
		kmem_cache_free(&cache_cache, cachep);
		cachep = NULL;
2363
		goto oops;
L
Linus Torvalds 已提交
2364
	}
P
Pekka Enberg 已提交
2365 2366
	slab_size = ALIGN(cachep->num * sizeof(kmem_bufctl_t)
			  + sizeof(struct slab), align);
L
Linus Torvalds 已提交
2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378

	/*
	 * 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
2379 2380
		slab_size =
		    cachep->num * sizeof(kmem_bufctl_t) + sizeof(struct slab);
2381 2382 2383 2384 2385 2386 2387 2388 2389

#ifdef CONFIG_PAGE_POISONING
		/* If we're going to use the generic kernel_map_pages()
		 * poisoning, then it's going to smash the contents of
		 * the redzone and userword anyhow, so switch them off.
		 */
		if (size % PAGE_SIZE == 0 && flags & SLAB_POISON)
			flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
#endif
L
Linus Torvalds 已提交
2390 2391 2392 2393 2394 2395
	}

	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
2396
	cachep->colour = left_over / cachep->colour_off;
L
Linus Torvalds 已提交
2397 2398 2399
	cachep->slab_size = slab_size;
	cachep->flags = flags;
	cachep->gfpflags = 0;
2400
	if (CONFIG_ZONE_DMA_FLAG && (flags & SLAB_CACHE_DMA))
L
Linus Torvalds 已提交
2401
		cachep->gfpflags |= GFP_DMA;
2402
	cachep->buffer_size = size;
2403
	cachep->reciprocal_buffer_size = reciprocal_value(size);
L
Linus Torvalds 已提交
2404

2405
	if (flags & CFLGS_OFF_SLAB) {
2406
		cachep->slabp_cache = kmem_find_general_cachep(slab_size, 0u);
2407 2408 2409 2410 2411 2412 2413
		/*
		 * 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.
		 */
2414
		BUG_ON(ZERO_OR_NULL_PTR(cachep->slabp_cache));
2415
	}
L
Linus Torvalds 已提交
2416 2417 2418
	cachep->ctor = ctor;
	cachep->name = name;

2419
	if (setup_cpu_cache(cachep, gfp)) {
2420 2421 2422 2423
		__kmem_cache_destroy(cachep);
		cachep = NULL;
		goto oops;
	}
L
Linus Torvalds 已提交
2424 2425 2426

	/* cache setup completed, link it into the list */
	list_add(&cachep->next, &cache_chain);
A
Andrew Morton 已提交
2427
oops:
L
Linus Torvalds 已提交
2428 2429
	if (!cachep && (flags & SLAB_PANIC))
		panic("kmem_cache_create(): failed to create slab `%s'\n",
P
Pekka Enberg 已提交
2430
		      name);
2431 2432 2433 2434
	if (slab_is_available()) {
		mutex_unlock(&cache_chain_mutex);
		put_online_cpus();
	}
L
Linus Torvalds 已提交
2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449
	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());
}

2450
static void check_spinlock_acquired(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2451 2452 2453
{
#ifdef CONFIG_SMP
	check_irq_off();
2454
	assert_spin_locked(&cachep->nodelists[numa_mem_id()]->list_lock);
L
Linus Torvalds 已提交
2455 2456
#endif
}
2457

2458
static void check_spinlock_acquired_node(struct kmem_cache *cachep, int node)
2459 2460 2461 2462 2463 2464 2465
{
#ifdef CONFIG_SMP
	check_irq_off();
	assert_spin_locked(&cachep->nodelists[node]->list_lock);
#endif
}

L
Linus Torvalds 已提交
2466 2467 2468 2469
#else
#define check_irq_off()	do { } while(0)
#define check_irq_on()	do { } while(0)
#define check_spinlock_acquired(x) do { } while(0)
2470
#define check_spinlock_acquired_node(x, y) do { } while(0)
L
Linus Torvalds 已提交
2471 2472
#endif

2473 2474 2475 2476
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
2477 2478
static void do_drain(void *arg)
{
A
Andrew Morton 已提交
2479
	struct kmem_cache *cachep = arg;
L
Linus Torvalds 已提交
2480
	struct array_cache *ac;
2481
	int node = numa_mem_id();
L
Linus Torvalds 已提交
2482 2483

	check_irq_off();
2484
	ac = cpu_cache_get(cachep);
2485 2486 2487
	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
2488 2489 2490
	ac->avail = 0;
}

2491
static void drain_cpu_caches(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2492
{
2493 2494 2495
	struct kmem_list3 *l3;
	int node;

2496
	on_each_cpu(do_drain, cachep, 1);
L
Linus Torvalds 已提交
2497
	check_irq_on();
P
Pekka Enberg 已提交
2498
	for_each_online_node(node) {
2499
		l3 = cachep->nodelists[node];
2500 2501 2502 2503 2504 2505 2506
		if (l3 && l3->alien)
			drain_alien_cache(cachep, l3->alien);
	}

	for_each_online_node(node) {
		l3 = cachep->nodelists[node];
		if (l3)
2507
			drain_array(cachep, l3, l3->shared, 1, node);
2508
	}
L
Linus Torvalds 已提交
2509 2510
}

2511 2512 2513 2514 2515 2516 2517 2518
/*
 * 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
2519
{
2520 2521
	struct list_head *p;
	int nr_freed;
L
Linus Torvalds 已提交
2522 2523
	struct slab *slabp;

2524 2525
	nr_freed = 0;
	while (nr_freed < tofree && !list_empty(&l3->slabs_free)) {
L
Linus Torvalds 已提交
2526

2527
		spin_lock_irq(&l3->list_lock);
2528
		p = l3->slabs_free.prev;
2529 2530 2531 2532
		if (p == &l3->slabs_free) {
			spin_unlock_irq(&l3->list_lock);
			goto out;
		}
L
Linus Torvalds 已提交
2533

2534
		slabp = list_entry(p, struct slab, list);
L
Linus Torvalds 已提交
2535
#if DEBUG
2536
		BUG_ON(slabp->inuse);
L
Linus Torvalds 已提交
2537 2538
#endif
		list_del(&slabp->list);
2539 2540 2541 2542 2543
		/*
		 * Safe to drop the lock. The slab is no longer linked
		 * to the cache.
		 */
		l3->free_objects -= cache->num;
2544
		spin_unlock_irq(&l3->list_lock);
2545 2546
		slab_destroy(cache, slabp);
		nr_freed++;
L
Linus Torvalds 已提交
2547
	}
2548 2549
out:
	return nr_freed;
L
Linus Torvalds 已提交
2550 2551
}

2552
/* Called with cache_chain_mutex held to protect against cpu hotplug */
2553
static int __cache_shrink(struct kmem_cache *cachep)
2554 2555 2556 2557 2558 2559 2560 2561 2562
{
	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];
2563 2564 2565 2566 2567 2568 2569
		if (!l3)
			continue;

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

		ret += !list_empty(&l3->slabs_full) ||
			!list_empty(&l3->slabs_partial);
2570 2571 2572 2573
	}
	return (ret ? 1 : 0);
}

L
Linus Torvalds 已提交
2574 2575 2576 2577 2578 2579 2580
/**
 * 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.
 */
2581
int kmem_cache_shrink(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2582
{
2583
	int ret;
2584
	BUG_ON(!cachep || in_interrupt());
L
Linus Torvalds 已提交
2585

2586
	get_online_cpus();
2587 2588 2589
	mutex_lock(&cache_chain_mutex);
	ret = __cache_shrink(cachep);
	mutex_unlock(&cache_chain_mutex);
2590
	put_online_cpus();
2591
	return ret;
L
Linus Torvalds 已提交
2592 2593 2594 2595 2596 2597 2598
}
EXPORT_SYMBOL(kmem_cache_shrink);

/**
 * kmem_cache_destroy - delete a cache
 * @cachep: the cache to destroy
 *
2599
 * Remove a &struct kmem_cache object from the slab cache.
L
Linus Torvalds 已提交
2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610
 *
 * 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().
 */
2611
void kmem_cache_destroy(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2612
{
2613
	BUG_ON(!cachep || in_interrupt());
L
Linus Torvalds 已提交
2614 2615

	/* Find the cache in the chain of caches. */
2616
	get_online_cpus();
I
Ingo Molnar 已提交
2617
	mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2618 2619 2620 2621 2622 2623
	/*
	 * 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
2624
		list_add(&cachep->next, &cache_chain);
I
Ingo Molnar 已提交
2625
		mutex_unlock(&cache_chain_mutex);
2626
		put_online_cpus();
2627
		return;
L
Linus Torvalds 已提交
2628 2629 2630
	}

	if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU))
2631
		rcu_barrier();
L
Linus Torvalds 已提交
2632

2633
	__kmem_cache_destroy(cachep);
2634
	mutex_unlock(&cache_chain_mutex);
2635
	put_online_cpus();
L
Linus Torvalds 已提交
2636 2637 2638
}
EXPORT_SYMBOL(kmem_cache_destroy);

2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649
/*
 * 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.
 */
2650
static struct slab *alloc_slabmgmt(struct kmem_cache *cachep, void *objp,
2651 2652
				   int colour_off, gfp_t local_flags,
				   int nodeid)
L
Linus Torvalds 已提交
2653 2654
{
	struct slab *slabp;
P
Pekka Enberg 已提交
2655

L
Linus Torvalds 已提交
2656 2657
	if (OFF_SLAB(cachep)) {
		/* Slab management obj is off-slab. */
2658
		slabp = kmem_cache_alloc_node(cachep->slabp_cache,
2659
					      local_flags, nodeid);
2660 2661 2662 2663 2664 2665
		/*
		 * If the first object in the slab is leaked (it's allocated
		 * but no one has a reference to it), we want to make sure
		 * kmemleak does not treat the ->s_mem pointer as a reference
		 * to the object. Otherwise we will not report the leak.
		 */
2666 2667
		kmemleak_scan_area(&slabp->list, sizeof(struct list_head),
				   local_flags);
L
Linus Torvalds 已提交
2668 2669 2670
		if (!slabp)
			return NULL;
	} else {
P
Pekka Enberg 已提交
2671
		slabp = objp + colour_off;
L
Linus Torvalds 已提交
2672 2673 2674 2675
		colour_off += cachep->slab_size;
	}
	slabp->inuse = 0;
	slabp->colouroff = colour_off;
P
Pekka Enberg 已提交
2676
	slabp->s_mem = objp + colour_off;
2677
	slabp->nodeid = nodeid;
2678
	slabp->free = 0;
L
Linus Torvalds 已提交
2679 2680 2681 2682 2683
	return slabp;
}

static inline kmem_bufctl_t *slab_bufctl(struct slab *slabp)
{
P
Pekka Enberg 已提交
2684
	return (kmem_bufctl_t *) (slabp + 1);
L
Linus Torvalds 已提交
2685 2686
}

2687
static void cache_init_objs(struct kmem_cache *cachep,
C
Christoph Lameter 已提交
2688
			    struct slab *slabp)
L
Linus Torvalds 已提交
2689 2690 2691 2692
{
	int i;

	for (i = 0; i < cachep->num; i++) {
2693
		void *objp = index_to_obj(cachep, slabp, i);
L
Linus Torvalds 已提交
2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705
#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
2706 2707 2708
		 * 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
2709 2710
		 */
		if (cachep->ctor && !(cachep->flags & SLAB_POISON))
2711
			cachep->ctor(objp + obj_offset(cachep));
L
Linus Torvalds 已提交
2712 2713 2714 2715

		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
2716
					   " end of an object");
L
Linus Torvalds 已提交
2717 2718
			if (*dbg_redzone1(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "constructor overwrote the"
P
Pekka Enberg 已提交
2719
					   " start of an object");
L
Linus Torvalds 已提交
2720
		}
A
Andrew Morton 已提交
2721 2722
		if ((cachep->buffer_size % PAGE_SIZE) == 0 &&
			    OFF_SLAB(cachep) && cachep->flags & SLAB_POISON)
P
Pekka Enberg 已提交
2723
			kernel_map_pages(virt_to_page(objp),
2724
					 cachep->buffer_size / PAGE_SIZE, 0);
L
Linus Torvalds 已提交
2725 2726
#else
		if (cachep->ctor)
2727
			cachep->ctor(objp);
L
Linus Torvalds 已提交
2728
#endif
P
Pekka Enberg 已提交
2729
		slab_bufctl(slabp)[i] = i + 1;
L
Linus Torvalds 已提交
2730
	}
P
Pekka Enberg 已提交
2731
	slab_bufctl(slabp)[i - 1] = BUFCTL_END;
L
Linus Torvalds 已提交
2732 2733
}

2734
static void kmem_flagcheck(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2735
{
2736 2737 2738 2739 2740 2741
	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
2742 2743
}

A
Andrew Morton 已提交
2744 2745
static void *slab_get_obj(struct kmem_cache *cachep, struct slab *slabp,
				int nodeid)
2746
{
2747
	void *objp = index_to_obj(cachep, slabp, slabp->free);
2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760
	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
2761 2762
static void slab_put_obj(struct kmem_cache *cachep, struct slab *slabp,
				void *objp, int nodeid)
2763
{
2764
	unsigned int objnr = obj_to_index(cachep, slabp, objp);
2765 2766 2767 2768 2769

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

2770
	if (slab_bufctl(slabp)[objnr] + 1 <= SLAB_LIMIT + 1) {
2771
		printk(KERN_ERR "slab: double free detected in cache "
A
Andrew Morton 已提交
2772
				"'%s', objp %p\n", cachep->name, objp);
2773 2774 2775 2776 2777 2778 2779 2780
		BUG();
	}
#endif
	slab_bufctl(slabp)[objnr] = slabp->free;
	slabp->free = objnr;
	slabp->inuse--;
}

2781 2782 2783
/*
 * 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
2784
 * virtual address for kfree, ksize, and slab debugging.
2785 2786 2787
 */
static void slab_map_pages(struct kmem_cache *cache, struct slab *slab,
			   void *addr)
L
Linus Torvalds 已提交
2788
{
2789
	int nr_pages;
L
Linus Torvalds 已提交
2790 2791
	struct page *page;

2792
	page = virt_to_page(addr);
2793

2794
	nr_pages = 1;
2795
	if (likely(!PageCompound(page)))
2796 2797
		nr_pages <<= cache->gfporder;

L
Linus Torvalds 已提交
2798
	do {
2799 2800
		page_set_cache(page, cache);
		page_set_slab(page, slab);
L
Linus Torvalds 已提交
2801
		page++;
2802
	} while (--nr_pages);
L
Linus Torvalds 已提交
2803 2804 2805 2806 2807 2808
}

/*
 * 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.
 */
2809 2810
static int cache_grow(struct kmem_cache *cachep,
		gfp_t flags, int nodeid, void *objp)
L
Linus Torvalds 已提交
2811
{
P
Pekka Enberg 已提交
2812 2813 2814
	struct slab *slabp;
	size_t offset;
	gfp_t local_flags;
2815
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
2816

A
Andrew Morton 已提交
2817 2818 2819
	/*
	 * 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
2820
	 */
C
Christoph Lameter 已提交
2821 2822
	BUG_ON(flags & GFP_SLAB_BUG_MASK);
	local_flags = flags & (GFP_CONSTRAINT_MASK|GFP_RECLAIM_MASK);
L
Linus Torvalds 已提交
2823

2824
	/* Take the l3 list lock to change the colour_next on this node */
L
Linus Torvalds 已提交
2825
	check_irq_off();
2826 2827
	l3 = cachep->nodelists[nodeid];
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2828 2829

	/* Get colour for the slab, and cal the next value. */
2830 2831 2832 2833 2834
	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
2835

2836
	offset *= cachep->colour_off;
L
Linus Torvalds 已提交
2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848

	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
2849 2850 2851
	/*
	 * Get mem for the objs.  Attempt to allocate a physical page from
	 * 'nodeid'.
2852
	 */
2853
	if (!objp)
2854
		objp = kmem_getpages(cachep, local_flags, nodeid);
A
Andrew Morton 已提交
2855
	if (!objp)
L
Linus Torvalds 已提交
2856 2857 2858
		goto failed;

	/* Get slab management. */
2859
	slabp = alloc_slabmgmt(cachep, objp, offset,
C
Christoph Lameter 已提交
2860
			local_flags & ~GFP_CONSTRAINT_MASK, nodeid);
A
Andrew Morton 已提交
2861
	if (!slabp)
L
Linus Torvalds 已提交
2862 2863
		goto opps1;

2864
	slab_map_pages(cachep, slabp, objp);
L
Linus Torvalds 已提交
2865

C
Christoph Lameter 已提交
2866
	cache_init_objs(cachep, slabp);
L
Linus Torvalds 已提交
2867 2868 2869 2870

	if (local_flags & __GFP_WAIT)
		local_irq_disable();
	check_irq_off();
2871
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2872 2873

	/* Make slab active. */
2874
	list_add_tail(&slabp->list, &(l3->slabs_free));
L
Linus Torvalds 已提交
2875
	STATS_INC_GROWN(cachep);
2876 2877
	l3->free_objects += cachep->num;
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2878
	return 1;
A
Andrew Morton 已提交
2879
opps1:
L
Linus Torvalds 已提交
2880
	kmem_freepages(cachep, objp);
A
Andrew Morton 已提交
2881
failed:
L
Linus Torvalds 已提交
2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897
	if (local_flags & __GFP_WAIT)
		local_irq_disable();
	return 0;
}

#if DEBUG

/*
 * Perform extra freeing checks:
 * - detect bad pointers.
 * - POISON/RED_ZONE checking
 */
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
2898 2899
		       (unsigned long)objp);
		BUG();
L
Linus Torvalds 已提交
2900 2901 2902
	}
}

2903 2904
static inline void verify_redzone_free(struct kmem_cache *cache, void *obj)
{
2905
	unsigned long long redzone1, redzone2;
2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920

	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");

2921
	printk(KERN_ERR "%p: redzone 1:0x%llx, redzone 2:0x%llx.\n",
2922 2923 2924
			obj, redzone1, redzone2);
}

2925
static void *cache_free_debugcheck(struct kmem_cache *cachep, void *objp,
P
Pekka Enberg 已提交
2926
				   void *caller)
L
Linus Torvalds 已提交
2927 2928 2929 2930 2931
{
	struct page *page;
	unsigned int objnr;
	struct slab *slabp;

2932 2933
	BUG_ON(virt_to_cache(objp) != cachep);

2934
	objp -= obj_offset(cachep);
L
Linus Torvalds 已提交
2935
	kfree_debugcheck(objp);
2936
	page = virt_to_head_page(objp);
L
Linus Torvalds 已提交
2937

2938
	slabp = page_get_slab(page);
L
Linus Torvalds 已提交
2939 2940

	if (cachep->flags & SLAB_RED_ZONE) {
2941
		verify_redzone_free(cachep, objp);
L
Linus Torvalds 已提交
2942 2943 2944 2945 2946 2947
		*dbg_redzone1(cachep, objp) = RED_INACTIVE;
		*dbg_redzone2(cachep, objp) = RED_INACTIVE;
	}
	if (cachep->flags & SLAB_STORE_USER)
		*dbg_userword(cachep, objp) = caller;

2948
	objnr = obj_to_index(cachep, slabp, objp);
L
Linus Torvalds 已提交
2949 2950

	BUG_ON(objnr >= cachep->num);
2951
	BUG_ON(objp != index_to_obj(cachep, slabp, objnr));
L
Linus Torvalds 已提交
2952

2953 2954 2955
#ifdef CONFIG_DEBUG_SLAB_LEAK
	slab_bufctl(slabp)[objnr] = BUFCTL_FREE;
#endif
L
Linus Torvalds 已提交
2956 2957
	if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
A
Andrew Morton 已提交
2958
		if ((cachep->buffer_size % PAGE_SIZE)==0 && OFF_SLAB(cachep)) {
L
Linus Torvalds 已提交
2959
			store_stackinfo(cachep, objp, (unsigned long)caller);
P
Pekka Enberg 已提交
2960
			kernel_map_pages(virt_to_page(objp),
2961
					 cachep->buffer_size / PAGE_SIZE, 0);
L
Linus Torvalds 已提交
2962 2963 2964 2965 2966 2967 2968 2969 2970 2971
		} else {
			poison_obj(cachep, objp, POISON_FREE);
		}
#else
		poison_obj(cachep, objp, POISON_FREE);
#endif
	}
	return objp;
}

2972
static void check_slabp(struct kmem_cache *cachep, struct slab *slabp)
L
Linus Torvalds 已提交
2973 2974 2975
{
	kmem_bufctl_t i;
	int entries = 0;
P
Pekka Enberg 已提交
2976

L
Linus Torvalds 已提交
2977 2978 2979 2980 2981 2982 2983
	/* 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
2984 2985 2986 2987
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
2988
		for (i = 0;
2989
		     i < sizeof(*slabp) + cachep->num * sizeof(kmem_bufctl_t);
P
Pekka Enberg 已提交
2990
		     i++) {
A
Andrew Morton 已提交
2991
			if (i % 16 == 0)
L
Linus Torvalds 已提交
2992
				printk("\n%03x:", i);
P
Pekka Enberg 已提交
2993
			printk(" %02x", ((unsigned char *)slabp)[i]);
L
Linus Torvalds 已提交
2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004
		}
		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

3005
static void *cache_alloc_refill(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
3006 3007 3008 3009
{
	int batchcount;
	struct kmem_list3 *l3;
	struct array_cache *ac;
P
Pekka Enberg 已提交
3010 3011
	int node;

3012
retry:
L
Linus Torvalds 已提交
3013
	check_irq_off();
3014
	node = numa_mem_id();
3015
	ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
3016 3017
	batchcount = ac->batchcount;
	if (!ac->touched && batchcount > BATCHREFILL_LIMIT) {
A
Andrew Morton 已提交
3018 3019 3020 3021
		/*
		 * 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
3022 3023 3024
		 */
		batchcount = BATCHREFILL_LIMIT;
	}
P
Pekka Enberg 已提交
3025
	l3 = cachep->nodelists[node];
3026 3027 3028

	BUG_ON(ac->avail > 0 || !l3);
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
3029

3030
	/* See if we can refill from the shared array */
3031 3032
	if (l3->shared && transfer_objects(ac, l3->shared, batchcount)) {
		l3->shared->touched = 1;
3033
		goto alloc_done;
3034
	}
3035

L
Linus Torvalds 已提交
3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050
	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);
3051 3052 3053 3054 3055 3056

		/*
		 * The slab was either on partial or free list so
		 * there must be at least one object available for
		 * allocation.
		 */
3057
		BUG_ON(slabp->inuse >= cachep->num);
3058

L
Linus Torvalds 已提交
3059 3060 3061 3062 3063
		while (slabp->inuse < cachep->num && batchcount--) {
			STATS_INC_ALLOCED(cachep);
			STATS_INC_ACTIVE(cachep);
			STATS_SET_HIGH(cachep);

3064
			ac->entry[ac->avail++] = slab_get_obj(cachep, slabp,
P
Pekka Enberg 已提交
3065
							    node);
L
Linus Torvalds 已提交
3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076
		}
		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
3077
must_grow:
L
Linus Torvalds 已提交
3078
	l3->free_objects -= ac->avail;
A
Andrew Morton 已提交
3079
alloc_done:
3080
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
3081 3082 3083

	if (unlikely(!ac->avail)) {
		int x;
3084
		x = cache_grow(cachep, flags | GFP_THISNODE, node, NULL);
3085

A
Andrew Morton 已提交
3086
		/* cache_grow can reenable interrupts, then ac could change. */
3087
		ac = cpu_cache_get(cachep);
A
Andrew Morton 已提交
3088
		if (!x && ac->avail == 0)	/* no objects in sight? abort */
L
Linus Torvalds 已提交
3089 3090
			return NULL;

A
Andrew Morton 已提交
3091
		if (!ac->avail)		/* objects refilled by interrupt? */
L
Linus Torvalds 已提交
3092 3093 3094
			goto retry;
	}
	ac->touched = 1;
3095
	return ac->entry[--ac->avail];
L
Linus Torvalds 已提交
3096 3097
}

A
Andrew Morton 已提交
3098 3099
static inline void cache_alloc_debugcheck_before(struct kmem_cache *cachep,
						gfp_t flags)
L
Linus Torvalds 已提交
3100 3101 3102 3103 3104 3105 3106 3107
{
	might_sleep_if(flags & __GFP_WAIT);
#if DEBUG
	kmem_flagcheck(cachep, flags);
#endif
}

#if DEBUG
A
Andrew Morton 已提交
3108 3109
static void *cache_alloc_debugcheck_after(struct kmem_cache *cachep,
				gfp_t flags, void *objp, void *caller)
L
Linus Torvalds 已提交
3110
{
P
Pekka Enberg 已提交
3111
	if (!objp)
L
Linus Torvalds 已提交
3112
		return objp;
P
Pekka Enberg 已提交
3113
	if (cachep->flags & SLAB_POISON) {
L
Linus Torvalds 已提交
3114
#ifdef CONFIG_DEBUG_PAGEALLOC
3115
		if ((cachep->buffer_size % PAGE_SIZE) == 0 && OFF_SLAB(cachep))
P
Pekka Enberg 已提交
3116
			kernel_map_pages(virt_to_page(objp),
3117
					 cachep->buffer_size / PAGE_SIZE, 1);
L
Linus Torvalds 已提交
3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128
		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
3129 3130 3131 3132
		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
3133
			printk(KERN_ERR
3134
				"%p: redzone 1:0x%llx, redzone 2:0x%llx\n",
A
Andrew Morton 已提交
3135 3136
				objp, *dbg_redzone1(cachep, objp),
				*dbg_redzone2(cachep, objp));
L
Linus Torvalds 已提交
3137 3138 3139 3140
		}
		*dbg_redzone1(cachep, objp) = RED_ACTIVE;
		*dbg_redzone2(cachep, objp) = RED_ACTIVE;
	}
3141 3142 3143 3144 3145
#ifdef CONFIG_DEBUG_SLAB_LEAK
	{
		struct slab *slabp;
		unsigned objnr;

3146
		slabp = page_get_slab(virt_to_head_page(objp));
3147 3148 3149 3150
		objnr = (unsigned)(objp - slabp->s_mem) / cachep->buffer_size;
		slab_bufctl(slabp)[objnr] = BUFCTL_ACTIVE;
	}
#endif
3151
	objp += obj_offset(cachep);
3152
	if (cachep->ctor && cachep->flags & SLAB_POISON)
3153
		cachep->ctor(objp);
3154 3155 3156 3157 3158 3159
#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
3160 3161 3162 3163 3164 3165
	return objp;
}
#else
#define cache_alloc_debugcheck_after(a,b,objp,d) (objp)
#endif

A
Akinobu Mita 已提交
3166
static bool slab_should_failslab(struct kmem_cache *cachep, gfp_t flags)
3167 3168
{
	if (cachep == &cache_cache)
A
Akinobu Mita 已提交
3169
		return false;
3170

3171
	return should_failslab(obj_size(cachep), flags, cachep->flags);
3172 3173
}

3174
static inline void *____cache_alloc(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
3175
{
P
Pekka Enberg 已提交
3176
	void *objp;
L
Linus Torvalds 已提交
3177 3178
	struct array_cache *ac;

3179
	check_irq_off();
3180

3181
	ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
3182 3183 3184
	if (likely(ac->avail)) {
		STATS_INC_ALLOCHIT(cachep);
		ac->touched = 1;
3185
		objp = ac->entry[--ac->avail];
L
Linus Torvalds 已提交
3186 3187 3188
	} else {
		STATS_INC_ALLOCMISS(cachep);
		objp = cache_alloc_refill(cachep, flags);
3189 3190 3191 3192 3193
		/*
		 * the 'ac' may be updated by cache_alloc_refill(),
		 * and kmemleak_erase() requires its correct value.
		 */
		ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
3194
	}
3195 3196 3197 3198 3199
	/*
	 * To avoid a false negative, if an object that is in one of the
	 * per-CPU caches is leaked, we need to make sure kmemleak doesn't
	 * treat the array pointers as a reference to the object.
	 */
3200 3201
	if (objp)
		kmemleak_erase(&ac->entry[ac->avail]);
3202 3203 3204
	return objp;
}

3205
#ifdef CONFIG_NUMA
3206
/*
3207
 * Try allocating on another node if PF_SPREAD_SLAB|PF_MEMPOLICY.
3208 3209 3210 3211 3212 3213 3214 3215
 *
 * 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;

3216
	if (in_interrupt() || (flags & __GFP_THISNODE))
3217
		return NULL;
3218
	nid_alloc = nid_here = numa_mem_id();
3219
	get_mems_allowed();
3220
	if (cpuset_do_slab_mem_spread() && (cachep->flags & SLAB_MEM_SPREAD))
3221
		nid_alloc = cpuset_slab_spread_node();
3222 3223
	else if (current->mempolicy)
		nid_alloc = slab_node(current->mempolicy);
3224
	put_mems_allowed();
3225
	if (nid_alloc != nid_here)
3226
		return ____cache_alloc_node(cachep, flags, nid_alloc);
3227 3228 3229
	return NULL;
}

3230 3231
/*
 * Fallback function if there was no memory available and no objects on a
3232 3233 3234 3235 3236
 * 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.
3237
 */
3238
static void *fallback_alloc(struct kmem_cache *cache, gfp_t flags)
3239
{
3240 3241
	struct zonelist *zonelist;
	gfp_t local_flags;
3242
	struct zoneref *z;
3243 3244
	struct zone *zone;
	enum zone_type high_zoneidx = gfp_zone(flags);
3245
	void *obj = NULL;
3246
	int nid;
3247 3248 3249 3250

	if (flags & __GFP_THISNODE)
		return NULL;

3251
	get_mems_allowed();
3252
	zonelist = node_zonelist(slab_node(current->mempolicy), flags);
C
Christoph Lameter 已提交
3253
	local_flags = flags & (GFP_CONSTRAINT_MASK|GFP_RECLAIM_MASK);
3254

3255 3256 3257 3258 3259
retry:
	/*
	 * Look through allowed nodes for objects available
	 * from existing per node queues.
	 */
3260 3261
	for_each_zone_zonelist(zone, z, zonelist, high_zoneidx) {
		nid = zone_to_nid(zone);
3262

3263
		if (cpuset_zone_allowed_hardwall(zone, flags) &&
3264
			cache->nodelists[nid] &&
3265
			cache->nodelists[nid]->free_objects) {
3266 3267
				obj = ____cache_alloc_node(cache,
					flags | GFP_THISNODE, nid);
3268 3269 3270
				if (obj)
					break;
		}
3271 3272
	}

3273
	if (!obj) {
3274 3275 3276 3277 3278 3279
		/*
		 * 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.
		 */
3280 3281 3282
		if (local_flags & __GFP_WAIT)
			local_irq_enable();
		kmem_flagcheck(cache, flags);
3283
		obj = kmem_getpages(cache, local_flags, numa_mem_id());
3284 3285
		if (local_flags & __GFP_WAIT)
			local_irq_disable();
3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301
		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 {
3302
				/* cache_grow already freed obj */
3303 3304 3305
				obj = NULL;
			}
		}
3306
	}
3307
	put_mems_allowed();
3308 3309 3310
	return obj;
}

3311 3312
/*
 * A interface to enable slab creation on nodeid
L
Linus Torvalds 已提交
3313
 */
3314
static void *____cache_alloc_node(struct kmem_cache *cachep, gfp_t flags,
A
Andrew Morton 已提交
3315
				int nodeid)
3316 3317
{
	struct list_head *entry;
P
Pekka Enberg 已提交
3318 3319 3320 3321 3322 3323 3324 3325
	struct slab *slabp;
	struct kmem_list3 *l3;
	void *obj;
	int x;

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

A
Andrew Morton 已提交
3326
retry:
3327
	check_irq_off();
P
Pekka Enberg 已提交
3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346
	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);

3347
	obj = slab_get_obj(cachep, slabp, nodeid);
P
Pekka Enberg 已提交
3348 3349 3350 3351 3352
	check_slabp(cachep, slabp);
	l3->free_objects--;
	/* move slabp to correct slabp list: */
	list_del(&slabp->list);

A
Andrew Morton 已提交
3353
	if (slabp->free == BUFCTL_END)
P
Pekka Enberg 已提交
3354
		list_add(&slabp->list, &l3->slabs_full);
A
Andrew Morton 已提交
3355
	else
P
Pekka Enberg 已提交
3356
		list_add(&slabp->list, &l3->slabs_partial);
3357

P
Pekka Enberg 已提交
3358 3359
	spin_unlock(&l3->list_lock);
	goto done;
3360

A
Andrew Morton 已提交
3361
must_grow:
P
Pekka Enberg 已提交
3362
	spin_unlock(&l3->list_lock);
3363
	x = cache_grow(cachep, flags | GFP_THISNODE, nodeid, NULL);
3364 3365
	if (x)
		goto retry;
L
Linus Torvalds 已提交
3366

3367
	return fallback_alloc(cachep, flags);
3368

A
Andrew Morton 已提交
3369
done:
P
Pekka Enberg 已提交
3370
	return obj;
3371
}
3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390

/**
 * 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;
3391
	int slab_node = numa_mem_id();
3392

3393
	flags &= gfp_allowed_mask;
3394

3395 3396
	lockdep_trace_alloc(flags);

A
Akinobu Mita 已提交
3397
	if (slab_should_failslab(cachep, flags))
3398 3399
		return NULL;

3400 3401 3402
	cache_alloc_debugcheck_before(cachep, flags);
	local_irq_save(save_flags);

3403
	if (nodeid == -1)
3404
		nodeid = slab_node;
3405 3406 3407 3408 3409 3410 3411

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

3412
	if (nodeid == slab_node) {
3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427
		/*
		 * 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);
3428 3429
	kmemleak_alloc_recursive(ptr, obj_size(cachep), 1, cachep->flags,
				 flags);
3430

P
Pekka Enberg 已提交
3431 3432 3433
	if (likely(ptr))
		kmemcheck_slab_alloc(cachep, flags, ptr, obj_size(cachep));

3434 3435 3436
	if (unlikely((flags & __GFP_ZERO) && ptr))
		memset(ptr, 0, obj_size(cachep));

3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455
	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
	 */
3456 3457
	if (!objp)
		objp = ____cache_alloc_node(cache, flags, numa_mem_id());
3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477

  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;

3478
	flags &= gfp_allowed_mask;
3479

3480 3481
	lockdep_trace_alloc(flags);

A
Akinobu Mita 已提交
3482
	if (slab_should_failslab(cachep, flags))
3483 3484
		return NULL;

3485 3486 3487 3488 3489
	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);
3490 3491
	kmemleak_alloc_recursive(objp, obj_size(cachep), 1, cachep->flags,
				 flags);
3492 3493
	prefetchw(objp);

P
Pekka Enberg 已提交
3494 3495 3496
	if (likely(objp))
		kmemcheck_slab_alloc(cachep, flags, objp, obj_size(cachep));

3497 3498 3499
	if (unlikely((flags & __GFP_ZERO) && objp))
		memset(objp, 0, obj_size(cachep));

3500 3501
	return objp;
}
3502 3503 3504 3505

/*
 * Caller needs to acquire correct kmem_list's list_lock
 */
3506
static void free_block(struct kmem_cache *cachep, void **objpp, int nr_objects,
P
Pekka Enberg 已提交
3507
		       int node)
L
Linus Torvalds 已提交
3508 3509
{
	int i;
3510
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
3511 3512 3513 3514 3515

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

3516
		slabp = virt_to_slab(objp);
3517
		l3 = cachep->nodelists[node];
L
Linus Torvalds 已提交
3518
		list_del(&slabp->list);
3519
		check_spinlock_acquired_node(cachep, node);
L
Linus Torvalds 已提交
3520
		check_slabp(cachep, slabp);
3521
		slab_put_obj(cachep, slabp, objp, node);
L
Linus Torvalds 已提交
3522
		STATS_DEC_ACTIVE(cachep);
3523
		l3->free_objects++;
L
Linus Torvalds 已提交
3524 3525 3526 3527
		check_slabp(cachep, slabp);

		/* fixup slab chains */
		if (slabp->inuse == 0) {
3528 3529
			if (l3->free_objects > l3->free_limit) {
				l3->free_objects -= cachep->num;
3530 3531 3532 3533 3534 3535
				/* 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
3536 3537
				slab_destroy(cachep, slabp);
			} else {
3538
				list_add(&slabp->list, &l3->slabs_free);
L
Linus Torvalds 已提交
3539 3540 3541 3542 3543 3544
			}
		} else {
			/* Unconditionally move a slab to the end of the
			 * partial list on free - maximum time for the
			 * other objects to be freed, too.
			 */
3545
			list_add_tail(&slabp->list, &l3->slabs_partial);
L
Linus Torvalds 已提交
3546 3547 3548 3549
		}
	}
}

3550
static void cache_flusharray(struct kmem_cache *cachep, struct array_cache *ac)
L
Linus Torvalds 已提交
3551 3552
{
	int batchcount;
3553
	struct kmem_list3 *l3;
3554
	int node = numa_mem_id();
L
Linus Torvalds 已提交
3555 3556 3557 3558 3559 3560

	batchcount = ac->batchcount;
#if DEBUG
	BUG_ON(!batchcount || batchcount > ac->avail);
#endif
	check_irq_off();
3561
	l3 = cachep->nodelists[node];
3562
	spin_lock(&l3->list_lock);
3563 3564
	if (l3->shared) {
		struct array_cache *shared_array = l3->shared;
P
Pekka Enberg 已提交
3565
		int max = shared_array->limit - shared_array->avail;
L
Linus Torvalds 已提交
3566 3567 3568
		if (max) {
			if (batchcount > max)
				batchcount = max;
3569
			memcpy(&(shared_array->entry[shared_array->avail]),
P
Pekka Enberg 已提交
3570
			       ac->entry, sizeof(void *) * batchcount);
L
Linus Torvalds 已提交
3571 3572 3573 3574 3575
			shared_array->avail += batchcount;
			goto free_done;
		}
	}

3576
	free_block(cachep, ac->entry, batchcount, node);
A
Andrew Morton 已提交
3577
free_done:
L
Linus Torvalds 已提交
3578 3579 3580 3581 3582
#if STATS
	{
		int i = 0;
		struct list_head *p;

3583 3584
		p = l3->slabs_free.next;
		while (p != &(l3->slabs_free)) {
L
Linus Torvalds 已提交
3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595
			struct slab *slabp;

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

			i++;
			p = p->next;
		}
		STATS_SET_FREEABLE(cachep, i);
	}
#endif
3596
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
3597
	ac->avail -= batchcount;
A
Andrew Morton 已提交
3598
	memmove(ac->entry, &(ac->entry[batchcount]), sizeof(void *)*ac->avail);
L
Linus Torvalds 已提交
3599 3600 3601
}

/*
A
Andrew Morton 已提交
3602 3603
 * 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
3604
 */
3605
static inline void __cache_free(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
3606
{
3607
	struct array_cache *ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
3608 3609

	check_irq_off();
3610
	kmemleak_free_recursive(objp, cachep->flags);
L
Linus Torvalds 已提交
3611 3612
	objp = cache_free_debugcheck(cachep, objp, __builtin_return_address(0));

P
Pekka Enberg 已提交
3613 3614
	kmemcheck_slab_free(cachep, objp, obj_size(cachep));

3615 3616 3617 3618 3619 3620 3621
	/*
	 * Skip calling cache_free_alien() when the platform is not numa.
	 * This will avoid cache misses that happen while accessing slabp (which
	 * is per page memory  reference) to get nodeid. Instead use a global
	 * variable to skip the call, which is mostly likely to be present in
	 * the cache.
	 */
3622
	if (nr_online_nodes > 1 && cache_free_alien(cachep, objp))
3623 3624
		return;

L
Linus Torvalds 已提交
3625 3626
	if (likely(ac->avail < ac->limit)) {
		STATS_INC_FREEHIT(cachep);
3627
		ac->entry[ac->avail++] = objp;
L
Linus Torvalds 已提交
3628 3629 3630 3631
		return;
	} else {
		STATS_INC_FREEMISS(cachep);
		cache_flusharray(cachep, ac);
3632
		ac->entry[ac->avail++] = objp;
L
Linus Torvalds 已提交
3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643
	}
}

/**
 * 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.
 */
3644
void *kmem_cache_alloc(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
3645
{
E
Eduard - Gabriel Munteanu 已提交
3646 3647
	void *ret = __cache_alloc(cachep, flags, __builtin_return_address(0));

3648 3649
	trace_kmem_cache_alloc(_RET_IP_, ret,
			       obj_size(cachep), cachep->buffer_size, flags);
E
Eduard - Gabriel Munteanu 已提交
3650 3651

	return ret;
L
Linus Torvalds 已提交
3652 3653 3654
}
EXPORT_SYMBOL(kmem_cache_alloc);

3655
#ifdef CONFIG_TRACING
3656 3657
void *
kmem_cache_alloc_trace(size_t size, struct kmem_cache *cachep, gfp_t flags)
E
Eduard - Gabriel Munteanu 已提交
3658
{
3659 3660 3661 3662 3663 3664 3665
	void *ret;

	ret = __cache_alloc(cachep, flags, __builtin_return_address(0));

	trace_kmalloc(_RET_IP_, ret,
		      size, slab_buffer_size(cachep), flags);
	return ret;
E
Eduard - Gabriel Munteanu 已提交
3666
}
3667
EXPORT_SYMBOL(kmem_cache_alloc_trace);
E
Eduard - Gabriel Munteanu 已提交
3668 3669
#endif

L
Linus Torvalds 已提交
3670
#ifdef CONFIG_NUMA
3671 3672
void *kmem_cache_alloc_node(struct kmem_cache *cachep, gfp_t flags, int nodeid)
{
E
Eduard - Gabriel Munteanu 已提交
3673 3674 3675
	void *ret = __cache_alloc_node(cachep, flags, nodeid,
				       __builtin_return_address(0));

3676 3677 3678
	trace_kmem_cache_alloc_node(_RET_IP_, ret,
				    obj_size(cachep), cachep->buffer_size,
				    flags, nodeid);
E
Eduard - Gabriel Munteanu 已提交
3679 3680

	return ret;
3681
}
L
Linus Torvalds 已提交
3682 3683
EXPORT_SYMBOL(kmem_cache_alloc_node);

3684
#ifdef CONFIG_TRACING
3685 3686 3687 3688
void *kmem_cache_alloc_node_trace(size_t size,
				  struct kmem_cache *cachep,
				  gfp_t flags,
				  int nodeid)
E
Eduard - Gabriel Munteanu 已提交
3689
{
3690 3691 3692
	void *ret;

	ret = __cache_alloc_node(cachep, flags, nodeid,
E
Eduard - Gabriel Munteanu 已提交
3693
				  __builtin_return_address(0));
3694 3695 3696 3697
	trace_kmalloc_node(_RET_IP_, ret,
			   size, slab_buffer_size(cachep),
			   flags, nodeid);
	return ret;
E
Eduard - Gabriel Munteanu 已提交
3698
}
3699
EXPORT_SYMBOL(kmem_cache_alloc_node_trace);
E
Eduard - Gabriel Munteanu 已提交
3700 3701
#endif

3702 3703
static __always_inline void *
__do_kmalloc_node(size_t size, gfp_t flags, int node, void *caller)
3704
{
3705
	struct kmem_cache *cachep;
3706 3707

	cachep = kmem_find_general_cachep(size, flags);
3708 3709
	if (unlikely(ZERO_OR_NULL_PTR(cachep)))
		return cachep;
3710
	return kmem_cache_alloc_node_trace(size, cachep, flags, node);
3711
}
3712

3713
#if defined(CONFIG_DEBUG_SLAB) || defined(CONFIG_TRACING)
3714 3715 3716 3717 3718
void *__kmalloc_node(size_t size, gfp_t flags, int node)
{
	return __do_kmalloc_node(size, flags, node,
			__builtin_return_address(0));
}
3719
EXPORT_SYMBOL(__kmalloc_node);
3720 3721

void *__kmalloc_node_track_caller(size_t size, gfp_t flags,
3722
		int node, unsigned long caller)
3723
{
3724
	return __do_kmalloc_node(size, flags, node, (void *)caller);
3725 3726 3727 3728 3729 3730 3731 3732
}
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);
3733
#endif /* CONFIG_DEBUG_SLAB || CONFIG_TRACING */
3734
#endif /* CONFIG_NUMA */
L
Linus Torvalds 已提交
3735 3736

/**
3737
 * __do_kmalloc - allocate memory
L
Linus Torvalds 已提交
3738
 * @size: how many bytes of memory are required.
3739
 * @flags: the type of memory to allocate (see kmalloc).
3740
 * @caller: function caller for debug tracking of the caller
L
Linus Torvalds 已提交
3741
 */
3742 3743
static __always_inline void *__do_kmalloc(size_t size, gfp_t flags,
					  void *caller)
L
Linus Torvalds 已提交
3744
{
3745
	struct kmem_cache *cachep;
E
Eduard - Gabriel Munteanu 已提交
3746
	void *ret;
L
Linus Torvalds 已提交
3747

3748 3749 3750 3751 3752 3753
	/* 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);
3754 3755
	if (unlikely(ZERO_OR_NULL_PTR(cachep)))
		return cachep;
E
Eduard - Gabriel Munteanu 已提交
3756 3757
	ret = __cache_alloc(cachep, flags, caller);

3758 3759
	trace_kmalloc((unsigned long) caller, ret,
		      size, cachep->buffer_size, flags);
E
Eduard - Gabriel Munteanu 已提交
3760 3761

	return ret;
3762 3763 3764
}


3765
#if defined(CONFIG_DEBUG_SLAB) || defined(CONFIG_TRACING)
3766 3767
void *__kmalloc(size_t size, gfp_t flags)
{
3768
	return __do_kmalloc(size, flags, __builtin_return_address(0));
L
Linus Torvalds 已提交
3769 3770 3771
}
EXPORT_SYMBOL(__kmalloc);

3772
void *__kmalloc_track_caller(size_t size, gfp_t flags, unsigned long caller)
3773
{
3774
	return __do_kmalloc(size, flags, (void *)caller);
3775 3776
}
EXPORT_SYMBOL(__kmalloc_track_caller);
3777 3778 3779 3780 3781 3782 3783

#else
void *__kmalloc(size_t size, gfp_t flags)
{
	return __do_kmalloc(size, flags, NULL);
}
EXPORT_SYMBOL(__kmalloc);
3784 3785
#endif

L
Linus Torvalds 已提交
3786 3787 3788 3789 3790 3791 3792 3793
/**
 * 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.
 */
3794
void kmem_cache_free(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
3795 3796 3797 3798
{
	unsigned long flags;

	local_irq_save(flags);
3799
	debug_check_no_locks_freed(objp, obj_size(cachep));
3800 3801
	if (!(cachep->flags & SLAB_DEBUG_OBJECTS))
		debug_check_no_obj_freed(objp, obj_size(cachep));
3802
	__cache_free(cachep, objp);
L
Linus Torvalds 已提交
3803
	local_irq_restore(flags);
E
Eduard - Gabriel Munteanu 已提交
3804

3805
	trace_kmem_cache_free(_RET_IP_, objp);
L
Linus Torvalds 已提交
3806 3807 3808 3809 3810 3811 3812
}
EXPORT_SYMBOL(kmem_cache_free);

/**
 * kfree - free previously allocated memory
 * @objp: pointer returned by kmalloc.
 *
3813 3814
 * If @objp is NULL, no operation is performed.
 *
L
Linus Torvalds 已提交
3815 3816 3817 3818 3819
 * Don't free memory not originally allocated by kmalloc()
 * or you will run into trouble.
 */
void kfree(const void *objp)
{
3820
	struct kmem_cache *c;
L
Linus Torvalds 已提交
3821 3822
	unsigned long flags;

3823 3824
	trace_kfree(_RET_IP_, objp);

3825
	if (unlikely(ZERO_OR_NULL_PTR(objp)))
L
Linus Torvalds 已提交
3826 3827 3828
		return;
	local_irq_save(flags);
	kfree_debugcheck(objp);
3829
	c = virt_to_cache(objp);
3830
	debug_check_no_locks_freed(objp, obj_size(c));
3831
	debug_check_no_obj_freed(objp, obj_size(c));
3832
	__cache_free(c, (void *)objp);
L
Linus Torvalds 已提交
3833 3834 3835 3836
	local_irq_restore(flags);
}
EXPORT_SYMBOL(kfree);

3837
unsigned int kmem_cache_size(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
3838
{
3839
	return obj_size(cachep);
L
Linus Torvalds 已提交
3840 3841 3842
}
EXPORT_SYMBOL(kmem_cache_size);

3843
const char *kmem_cache_name(struct kmem_cache *cachep)
3844 3845 3846 3847 3848
{
	return cachep->name;
}
EXPORT_SYMBOL_GPL(kmem_cache_name);

3849
/*
S
Simon Arlott 已提交
3850
 * This initializes kmem_list3 or resizes various caches for all nodes.
3851
 */
3852
static int alloc_kmemlist(struct kmem_cache *cachep, gfp_t gfp)
3853 3854 3855
{
	int node;
	struct kmem_list3 *l3;
3856
	struct array_cache *new_shared;
3857
	struct array_cache **new_alien = NULL;
3858

3859
	for_each_online_node(node) {
3860

3861
                if (use_alien_caches) {
3862
                        new_alien = alloc_alien_cache(node, cachep->limit, gfp);
3863 3864 3865
                        if (!new_alien)
                                goto fail;
                }
3866

3867 3868 3869
		new_shared = NULL;
		if (cachep->shared) {
			new_shared = alloc_arraycache(node,
3870
				cachep->shared*cachep->batchcount,
3871
					0xbaadf00d, gfp);
3872 3873 3874 3875
			if (!new_shared) {
				free_alien_cache(new_alien);
				goto fail;
			}
3876
		}
3877

A
Andrew Morton 已提交
3878 3879
		l3 = cachep->nodelists[node];
		if (l3) {
3880 3881
			struct array_cache *shared = l3->shared;

3882 3883
			spin_lock_irq(&l3->list_lock);

3884
			if (shared)
3885 3886
				free_block(cachep, shared->entry,
						shared->avail, node);
3887

3888 3889
			l3->shared = new_shared;
			if (!l3->alien) {
3890 3891 3892
				l3->alien = new_alien;
				new_alien = NULL;
			}
P
Pekka Enberg 已提交
3893
			l3->free_limit = (1 + nr_cpus_node(node)) *
A
Andrew Morton 已提交
3894
					cachep->batchcount + cachep->num;
3895
			spin_unlock_irq(&l3->list_lock);
3896
			kfree(shared);
3897 3898 3899
			free_alien_cache(new_alien);
			continue;
		}
3900
		l3 = kmalloc_node(sizeof(struct kmem_list3), gfp, node);
3901 3902 3903
		if (!l3) {
			free_alien_cache(new_alien);
			kfree(new_shared);
3904
			goto fail;
3905
		}
3906 3907 3908

		kmem_list3_init(l3);
		l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
A
Andrew Morton 已提交
3909
				((unsigned long)cachep) % REAPTIMEOUT_LIST3;
3910
		l3->shared = new_shared;
3911
		l3->alien = new_alien;
P
Pekka Enberg 已提交
3912
		l3->free_limit = (1 + nr_cpus_node(node)) *
A
Andrew Morton 已提交
3913
					cachep->batchcount + cachep->num;
3914 3915
		cachep->nodelists[node] = l3;
	}
3916
	return 0;
3917

A
Andrew Morton 已提交
3918
fail:
3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933
	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--;
		}
	}
3934
	return -ENOMEM;
3935 3936
}

L
Linus Torvalds 已提交
3937
struct ccupdate_struct {
3938
	struct kmem_cache *cachep;
L
Linus Torvalds 已提交
3939 3940 3941 3942 3943
	struct array_cache *new[NR_CPUS];
};

static void do_ccupdate_local(void *info)
{
A
Andrew Morton 已提交
3944
	struct ccupdate_struct *new = info;
L
Linus Torvalds 已提交
3945 3946 3947
	struct array_cache *old;

	check_irq_off();
3948
	old = cpu_cache_get(new->cachep);
3949

L
Linus Torvalds 已提交
3950 3951 3952 3953
	new->cachep->array[smp_processor_id()] = new->new[smp_processor_id()];
	new->new[smp_processor_id()] = old;
}

3954
/* Always called with the cache_chain_mutex held */
A
Andrew Morton 已提交
3955
static int do_tune_cpucache(struct kmem_cache *cachep, int limit,
3956
				int batchcount, int shared, gfp_t gfp)
L
Linus Torvalds 已提交
3957
{
3958
	struct ccupdate_struct *new;
3959
	int i;
L
Linus Torvalds 已提交
3960

3961
	new = kzalloc(sizeof(*new), gfp);
3962 3963 3964
	if (!new)
		return -ENOMEM;

3965
	for_each_online_cpu(i) {
3966
		new->new[i] = alloc_arraycache(cpu_to_mem(i), limit,
3967
						batchcount, gfp);
3968
		if (!new->new[i]) {
P
Pekka Enberg 已提交
3969
			for (i--; i >= 0; i--)
3970 3971
				kfree(new->new[i]);
			kfree(new);
3972
			return -ENOMEM;
L
Linus Torvalds 已提交
3973 3974
		}
	}
3975
	new->cachep = cachep;
L
Linus Torvalds 已提交
3976

3977
	on_each_cpu(do_ccupdate_local, (void *)new, 1);
3978

L
Linus Torvalds 已提交
3979 3980 3981
	check_irq_on();
	cachep->batchcount = batchcount;
	cachep->limit = limit;
3982
	cachep->shared = shared;
L
Linus Torvalds 已提交
3983

3984
	for_each_online_cpu(i) {
3985
		struct array_cache *ccold = new->new[i];
L
Linus Torvalds 已提交
3986 3987
		if (!ccold)
			continue;
3988 3989 3990
		spin_lock_irq(&cachep->nodelists[cpu_to_mem(i)]->list_lock);
		free_block(cachep, ccold->entry, ccold->avail, cpu_to_mem(i));
		spin_unlock_irq(&cachep->nodelists[cpu_to_mem(i)]->list_lock);
L
Linus Torvalds 已提交
3991 3992
		kfree(ccold);
	}
3993
	kfree(new);
3994
	return alloc_kmemlist(cachep, gfp);
L
Linus Torvalds 已提交
3995 3996
}

3997
/* Called with cache_chain_mutex held always */
3998
static int enable_cpucache(struct kmem_cache *cachep, gfp_t gfp)
L
Linus Torvalds 已提交
3999 4000 4001 4002
{
	int err;
	int limit, shared;

A
Andrew Morton 已提交
4003 4004
	/*
	 * The head array serves three purposes:
L
Linus Torvalds 已提交
4005 4006
	 * - create a LIFO ordering, i.e. return objects that are cache-warm
	 * - reduce the number of spinlock operations.
A
Andrew Morton 已提交
4007
	 * - reduce the number of linked list operations on the slab and
L
Linus Torvalds 已提交
4008 4009 4010 4011
	 *   bufctl chains: array operations are cheaper.
	 * The numbers are guessed, we should auto-tune as described by
	 * Bonwick.
	 */
4012
	if (cachep->buffer_size > 131072)
L
Linus Torvalds 已提交
4013
		limit = 1;
4014
	else if (cachep->buffer_size > PAGE_SIZE)
L
Linus Torvalds 已提交
4015
		limit = 8;
4016
	else if (cachep->buffer_size > 1024)
L
Linus Torvalds 已提交
4017
		limit = 24;
4018
	else if (cachep->buffer_size > 256)
L
Linus Torvalds 已提交
4019 4020 4021 4022
		limit = 54;
	else
		limit = 120;

A
Andrew Morton 已提交
4023 4024
	/*
	 * CPU bound tasks (e.g. network routing) can exhibit cpu bound
L
Linus Torvalds 已提交
4025 4026 4027 4028 4029 4030 4031 4032
	 * 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;
4033
	if (cachep->buffer_size <= PAGE_SIZE && num_possible_cpus() > 1)
L
Linus Torvalds 已提交
4034 4035 4036
		shared = 8;

#if DEBUG
A
Andrew Morton 已提交
4037 4038 4039
	/*
	 * With debugging enabled, large batchcount lead to excessively long
	 * periods with disabled local interrupts. Limit the batchcount
L
Linus Torvalds 已提交
4040 4041 4042 4043
	 */
	if (limit > 32)
		limit = 32;
#endif
4044
	err = do_tune_cpucache(cachep, limit, (limit + 1) / 2, shared, gfp);
L
Linus Torvalds 已提交
4045 4046
	if (err)
		printk(KERN_ERR "enable_cpucache failed for %s, error %d.\n",
P
Pekka Enberg 已提交
4047
		       cachep->name, -err);
4048
	return err;
L
Linus Torvalds 已提交
4049 4050
}

4051 4052
/*
 * Drain an array if it contains any elements taking the l3 lock only if
4053 4054
 * necessary. Note that the l3 listlock also protects the array_cache
 * if drain_array() is used on the shared array.
4055
 */
4056
static void drain_array(struct kmem_cache *cachep, struct kmem_list3 *l3,
4057
			 struct array_cache *ac, int force, int node)
L
Linus Torvalds 已提交
4058 4059 4060
{
	int tofree;

4061 4062
	if (!ac || !ac->avail)
		return;
L
Linus Torvalds 已提交
4063 4064
	if (ac->touched && !force) {
		ac->touched = 0;
4065
	} else {
4066
		spin_lock_irq(&l3->list_lock);
4067 4068 4069 4070 4071 4072 4073 4074 4075
		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);
		}
4076
		spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
4077 4078 4079 4080 4081
	}
}

/**
 * cache_reap - Reclaim memory from caches.
4082
 * @w: work descriptor
L
Linus Torvalds 已提交
4083 4084 4085 4086 4087 4088
 *
 * 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
4089 4090
 * 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
4091
 */
4092
static void cache_reap(struct work_struct *w)
L
Linus Torvalds 已提交
4093
{
4094
	struct kmem_cache *searchp;
4095
	struct kmem_list3 *l3;
4096
	int node = numa_mem_id();
4097
	struct delayed_work *work = to_delayed_work(w);
L
Linus Torvalds 已提交
4098

4099
	if (!mutex_trylock(&cache_chain_mutex))
L
Linus Torvalds 已提交
4100
		/* Give up. Setup the next iteration. */
4101
		goto out;
L
Linus Torvalds 已提交
4102

4103
	list_for_each_entry(searchp, &cache_chain, next) {
L
Linus Torvalds 已提交
4104 4105
		check_irq_on();

4106 4107 4108 4109 4110
		/*
		 * 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.
		 */
4111
		l3 = searchp->nodelists[node];
4112

4113
		reap_alien(searchp, l3);
L
Linus Torvalds 已提交
4114

4115
		drain_array(searchp, l3, cpu_cache_get(searchp), 0, node);
L
Linus Torvalds 已提交
4116

4117 4118 4119 4120
		/*
		 * These are racy checks but it does not matter
		 * if we skip one check or scan twice.
		 */
4121
		if (time_after(l3->next_reap, jiffies))
4122
			goto next;
L
Linus Torvalds 已提交
4123

4124
		l3->next_reap = jiffies + REAPTIMEOUT_LIST3;
L
Linus Torvalds 已提交
4125

4126
		drain_array(searchp, l3, l3->shared, 0, node);
L
Linus Torvalds 已提交
4127

4128
		if (l3->free_touched)
4129
			l3->free_touched = 0;
4130 4131
		else {
			int freed;
L
Linus Torvalds 已提交
4132

4133 4134 4135 4136
			freed = drain_freelist(searchp, l3, (l3->free_limit +
				5 * searchp->num - 1) / (5 * searchp->num));
			STATS_ADD_REAPED(searchp, freed);
		}
4137
next:
L
Linus Torvalds 已提交
4138 4139 4140
		cond_resched();
	}
	check_irq_on();
I
Ingo Molnar 已提交
4141
	mutex_unlock(&cache_chain_mutex);
4142
	next_reap_node();
4143
out:
A
Andrew Morton 已提交
4144
	/* Set up the next iteration */
4145
	schedule_delayed_work(work, round_jiffies_relative(REAPTIMEOUT_CPUC));
L
Linus Torvalds 已提交
4146 4147
}

4148
#ifdef CONFIG_SLABINFO
L
Linus Torvalds 已提交
4149

4150
static void print_slabinfo_header(struct seq_file *m)
L
Linus Torvalds 已提交
4151
{
4152 4153 4154 4155
	/*
	 * Output format version, so at least we can change it
	 * without _too_ many complaints.
	 */
L
Linus Torvalds 已提交
4156
#if STATS
4157
	seq_puts(m, "slabinfo - version: 2.1 (statistics)\n");
L
Linus Torvalds 已提交
4158
#else
4159
	seq_puts(m, "slabinfo - version: 2.1\n");
L
Linus Torvalds 已提交
4160
#endif
4161 4162 4163 4164
	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
4165
#if STATS
4166
	seq_puts(m, " : globalstat <listallocs> <maxobjs> <grown> <reaped> "
4167
		 "<error> <maxfreeable> <nodeallocs> <remotefrees> <alienoverflow>");
4168
	seq_puts(m, " : cpustat <allochit> <allocmiss> <freehit> <freemiss>");
L
Linus Torvalds 已提交
4169
#endif
4170 4171 4172 4173 4174 4175 4176
	seq_putc(m, '\n');
}

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

I
Ingo Molnar 已提交
4177
	mutex_lock(&cache_chain_mutex);
4178 4179
	if (!n)
		print_slabinfo_header(m);
4180 4181

	return seq_list_start(&cache_chain, *pos);
L
Linus Torvalds 已提交
4182 4183 4184 4185
}

static void *s_next(struct seq_file *m, void *p, loff_t *pos)
{
4186
	return seq_list_next(p, &cache_chain, pos);
L
Linus Torvalds 已提交
4187 4188 4189 4190
}

static void s_stop(struct seq_file *m, void *p)
{
I
Ingo Molnar 已提交
4191
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
4192 4193 4194 4195
}

static int s_show(struct seq_file *m, void *p)
{
4196
	struct kmem_cache *cachep = list_entry(p, struct kmem_cache, next);
P
Pekka Enberg 已提交
4197 4198 4199 4200 4201
	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;
4202
	const char *name;
L
Linus Torvalds 已提交
4203
	char *error = NULL;
4204 4205
	int node;
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
4206 4207 4208

	active_objs = 0;
	num_slabs = 0;
4209 4210 4211 4212 4213
	for_each_online_node(node) {
		l3 = cachep->nodelists[node];
		if (!l3)
			continue;

4214 4215
		check_irq_on();
		spin_lock_irq(&l3->list_lock);
4216

4217
		list_for_each_entry(slabp, &l3->slabs_full, list) {
4218 4219 4220 4221 4222
			if (slabp->inuse != cachep->num && !error)
				error = "slabs_full accounting error";
			active_objs += cachep->num;
			active_slabs++;
		}
4223
		list_for_each_entry(slabp, &l3->slabs_partial, list) {
4224 4225 4226 4227 4228 4229 4230
			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++;
		}
4231
		list_for_each_entry(slabp, &l3->slabs_free, list) {
4232 4233 4234 4235 4236
			if (slabp->inuse && !error)
				error = "slabs_free/inuse accounting error";
			num_slabs++;
		}
		free_objects += l3->free_objects;
4237 4238
		if (l3->shared)
			shared_avail += l3->shared->avail;
4239

4240
		spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
4241
	}
P
Pekka Enberg 已提交
4242 4243
	num_slabs += active_slabs;
	num_objs = num_slabs * cachep->num;
4244
	if (num_objs - active_objs != free_objects && !error)
L
Linus Torvalds 已提交
4245 4246
		error = "free_objects accounting error";

P
Pekka Enberg 已提交
4247
	name = cachep->name;
L
Linus Torvalds 已提交
4248 4249 4250 4251
	if (error)
		printk(KERN_ERR "slab: cache %s error: %s\n", name, error);

	seq_printf(m, "%-17s %6lu %6lu %6u %4u %4d",
4252
		   name, active_objs, num_objs, cachep->buffer_size,
P
Pekka Enberg 已提交
4253
		   cachep->num, (1 << cachep->gfporder));
L
Linus Torvalds 已提交
4254
	seq_printf(m, " : tunables %4u %4u %4u",
P
Pekka Enberg 已提交
4255
		   cachep->limit, cachep->batchcount, cachep->shared);
4256
	seq_printf(m, " : slabdata %6lu %6lu %6lu",
P
Pekka Enberg 已提交
4257
		   active_slabs, num_slabs, shared_avail);
L
Linus Torvalds 已提交
4258
#if STATS
P
Pekka Enberg 已提交
4259
	{			/* list3 stats */
L
Linus Torvalds 已提交
4260 4261 4262 4263 4264 4265 4266
		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;
4267
		unsigned long node_frees = cachep->node_frees;
4268
		unsigned long overflows = cachep->node_overflow;
L
Linus Torvalds 已提交
4269

J
Joe Perches 已提交
4270 4271 4272 4273 4274
		seq_printf(m, " : globalstat %7lu %6lu %5lu %4lu "
			   "%4lu %4lu %4lu %4lu %4lu",
			   allocs, high, grown,
			   reaped, errors, max_freeable, node_allocs,
			   node_frees, overflows);
L
Linus Torvalds 已提交
4275 4276 4277 4278 4279 4280 4281 4282 4283
	}
	/* 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
4284
			   allochit, allocmiss, freehit, freemiss);
L
Linus Torvalds 已提交
4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304
	}
#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
 */

4305
static const struct seq_operations slabinfo_op = {
P
Pekka Enberg 已提交
4306 4307 4308 4309
	.start = s_start,
	.next = s_next,
	.stop = s_stop,
	.show = s_show,
L
Linus Torvalds 已提交
4310 4311 4312 4313 4314 4315 4316 4317 4318 4319
};

#define MAX_SLABINFO_WRITE 128
/**
 * slabinfo_write - Tuning for the slab allocator
 * @file: unused
 * @buffer: user buffer
 * @count: data length
 * @ppos: unused
 */
4320
static ssize_t slabinfo_write(struct file *file, const char __user *buffer,
P
Pekka Enberg 已提交
4321
		       size_t count, loff_t *ppos)
L
Linus Torvalds 已提交
4322
{
P
Pekka Enberg 已提交
4323
	char kbuf[MAX_SLABINFO_WRITE + 1], *tmp;
L
Linus Torvalds 已提交
4324
	int limit, batchcount, shared, res;
4325
	struct kmem_cache *cachep;
P
Pekka Enberg 已提交
4326

L
Linus Torvalds 已提交
4327 4328 4329 4330
	if (count > MAX_SLABINFO_WRITE)
		return -EINVAL;
	if (copy_from_user(&kbuf, buffer, count))
		return -EFAULT;
P
Pekka Enberg 已提交
4331
	kbuf[MAX_SLABINFO_WRITE] = '\0';
L
Linus Torvalds 已提交
4332 4333 4334 4335 4336 4337 4338 4339 4340 4341

	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
4342
	mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
4343
	res = -EINVAL;
4344
	list_for_each_entry(cachep, &cache_chain, next) {
L
Linus Torvalds 已提交
4345
		if (!strcmp(cachep->name, kbuf)) {
A
Andrew Morton 已提交
4346 4347
			if (limit < 1 || batchcount < 1 ||
					batchcount > limit || shared < 0) {
4348
				res = 0;
L
Linus Torvalds 已提交
4349
			} else {
4350
				res = do_tune_cpucache(cachep, limit,
4351 4352
						       batchcount, shared,
						       GFP_KERNEL);
L
Linus Torvalds 已提交
4353 4354 4355 4356
			}
			break;
		}
	}
I
Ingo Molnar 已提交
4357
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
4358 4359 4360 4361
	if (res >= 0)
		res = count;
	return res;
}
4362

4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375
static int slabinfo_open(struct inode *inode, struct file *file)
{
	return seq_open(file, &slabinfo_op);
}

static const struct file_operations proc_slabinfo_operations = {
	.open		= slabinfo_open,
	.read		= seq_read,
	.write		= slabinfo_write,
	.llseek		= seq_lseek,
	.release	= seq_release,
};

4376 4377 4378 4379 4380
#ifdef CONFIG_DEBUG_SLAB_LEAK

static void *leaks_start(struct seq_file *m, loff_t *pos)
{
	mutex_lock(&cache_chain_mutex);
4381
	return seq_list_start(&cache_chain, *pos);
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
}

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
	unsigned long offset, size;
4432
	char modname[MODULE_NAME_LEN], name[KSYM_NAME_LEN];
4433

4434
	if (lookup_symbol_attrs(address, &size, &offset, modname, name) == 0) {
4435
		seq_printf(m, "%s+%#lx/%#lx", name, offset, size);
4436
		if (modname[0])
4437 4438 4439 4440 4441 4442 4443 4444 4445
			seq_printf(m, " [%s]", modname);
		return;
	}
#endif
	seq_printf(m, "%p", (void *)address);
}

static int leaks_show(struct seq_file *m, void *p)
{
4446
	struct kmem_cache *cachep = list_entry(p, struct kmem_cache, next);
4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470
	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);

4471
		list_for_each_entry(slabp, &l3->slabs_full, list)
4472
			handle_slab(n, cachep, slabp);
4473
		list_for_each_entry(slabp, &l3->slabs_partial, list)
4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499
			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');
	}
4500

4501 4502 4503
	return 0;
}

4504
static const struct seq_operations slabstats_op = {
4505 4506 4507 4508 4509
	.start = leaks_start,
	.next = s_next,
	.stop = s_stop,
	.show = leaks_show,
};
4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537

static int slabstats_open(struct inode *inode, struct file *file)
{
	unsigned long *n = kzalloc(PAGE_SIZE, GFP_KERNEL);
	int ret = -ENOMEM;
	if (n) {
		ret = seq_open(file, &slabstats_op);
		if (!ret) {
			struct seq_file *m = file->private_data;
			*n = PAGE_SIZE / (2 * sizeof(unsigned long));
			m->private = n;
			n = NULL;
		}
		kfree(n);
	}
	return ret;
}

static const struct file_operations proc_slabstats_operations = {
	.open		= slabstats_open,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= seq_release_private,
};
#endif

static int __init slab_proc_init(void)
{
4538
	proc_create("slabinfo",S_IWUSR|S_IRUGO,NULL,&proc_slabinfo_operations);
4539 4540
#ifdef CONFIG_DEBUG_SLAB_LEAK
	proc_create("slab_allocators", 0, NULL, &proc_slabstats_operations);
4541
#endif
4542 4543 4544
	return 0;
}
module_init(slab_proc_init);
L
Linus Torvalds 已提交
4545 4546
#endif

4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558
/**
 * 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
4559
size_t ksize(const void *objp)
L
Linus Torvalds 已提交
4560
{
4561 4562
	BUG_ON(!objp);
	if (unlikely(objp == ZERO_SIZE_PTR))
4563
		return 0;
L
Linus Torvalds 已提交
4564

4565
	return obj_size(virt_to_cache(objp));
L
Linus Torvalds 已提交
4566
}
K
Kirill A. Shutemov 已提交
4567
EXPORT_SYMBOL(ksize);