slab.c 119.7 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>
118
#include	<linux/prefetch.h>
L
Linus Torvalds 已提交
119 120 121 122 123

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

124 125
#include <trace/events/kmem.h>

L
Linus Torvalds 已提交
126
/*
127
 * DEBUG	- 1 for kmem_cache_create() to honour; SLAB_RED_ZONE & SLAB_POISON.
L
Linus Torvalds 已提交
128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147
 *		  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
148
#define	REDZONE_ALIGN		max(BYTES_PER_WORD, __alignof__(unsigned long long))
L
Linus Torvalds 已提交
149 150 151 152 153 154 155

#ifndef ARCH_KMALLOC_FLAGS
#define ARCH_KMALLOC_FLAGS SLAB_HWCACHE_ALIGN
#endif

/* Legal flag mask for kmem_cache_create(). */
#if DEBUG
156
# define CREATE_MASK	(SLAB_RED_ZONE | \
L
Linus Torvalds 已提交
157
			 SLAB_POISON | SLAB_HWCACHE_ALIGN | \
158
			 SLAB_CACHE_DMA | \
159
			 SLAB_STORE_USER | \
L
Linus Torvalds 已提交
160
			 SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
161
			 SLAB_DESTROY_BY_RCU | SLAB_MEM_SPREAD | \
P
Pekka Enberg 已提交
162
			 SLAB_DEBUG_OBJECTS | SLAB_NOLEAKTRACE | SLAB_NOTRACK)
L
Linus Torvalds 已提交
163
#else
164
# define CREATE_MASK	(SLAB_HWCACHE_ALIGN | \
165
			 SLAB_CACHE_DMA | \
L
Linus Torvalds 已提交
166
			 SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
167
			 SLAB_DESTROY_BY_RCU | SLAB_MEM_SPREAD | \
P
Pekka Enberg 已提交
168
			 SLAB_DEBUG_OBJECTS | SLAB_NOLEAKTRACE | SLAB_NOTRACK)
L
Linus Torvalds 已提交
169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189
#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.
 */

190
typedef unsigned int kmem_bufctl_t;
L
Linus Torvalds 已提交
191 192
#define BUFCTL_END	(((kmem_bufctl_t)(~0U))-0)
#define BUFCTL_FREE	(((kmem_bufctl_t)(~0U))-1)
193 194
#define	BUFCTL_ACTIVE	(((kmem_bufctl_t)(~0U))-2)
#define	SLAB_LIMIT	(((kmem_bufctl_t)(~0U))-3)
L
Linus Torvalds 已提交
195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210

/*
 * 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.
 */
struct slab_rcu {
P
Pekka Enberg 已提交
211
	struct rcu_head head;
212
	struct kmem_cache *cachep;
P
Pekka Enberg 已提交
213
	void *addr;
L
Linus Torvalds 已提交
214 215
};

216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236
/*
 * 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 {
	union {
		struct {
			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;
		};
		struct slab_rcu __slab_cover_slab_rcu;
	};
};

L
Linus Torvalds 已提交
237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253
/*
 * 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;
254
	spinlock_t lock;
255
	void *entry[];	/*
A
Andrew Morton 已提交
256 257 258 259
			 * Must have this definition in here for the proper
			 * alignment of array_cache. Also simplifies accessing
			 * the entries.
			 */
L
Linus Torvalds 已提交
260 261
};

A
Andrew Morton 已提交
262 263 264
/*
 * bootstrap: The caches do not work without cpuarrays anymore, but the
 * cpuarrays are allocated from the generic caches...
L
Linus Torvalds 已提交
265 266 267 268
 */
#define BOOT_CPUCACHE_ENTRIES	1
struct arraycache_init {
	struct array_cache cache;
P
Pekka Enberg 已提交
269
	void *entries[BOOT_CPUCACHE_ENTRIES];
L
Linus Torvalds 已提交
270 271 272
};

/*
273
 * The slab lists for all objects.
L
Linus Torvalds 已提交
274 275
 */
struct kmem_list3 {
P
Pekka Enberg 已提交
276 277 278 279 280
	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;
281
	unsigned int colour_next;	/* Per-node cache coloring */
P
Pekka Enberg 已提交
282 283 284
	spinlock_t list_lock;
	struct array_cache *shared;	/* shared per node */
	struct array_cache **alien;	/* on other nodes */
285 286
	unsigned long next_reap;	/* updated without locking */
	int free_touched;		/* updated without locking */
L
Linus Torvalds 已提交
287 288
};

289 290 291
/*
 * Need this for bootstrapping a per node allocator.
 */
292
#define NUM_INIT_LISTS (3 * MAX_NUMNODES)
293
static struct kmem_list3 __initdata initkmem_list3[NUM_INIT_LISTS];
294
#define	CACHE_CACHE 0
295 296
#define	SIZE_AC MAX_NUMNODES
#define	SIZE_L3 (2 * MAX_NUMNODES)
297

298 299 300 301
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);
302
static int enable_cpucache(struct kmem_cache *cachep, gfp_t gfp);
303
static void cache_reap(struct work_struct *unused);
304

305
/*
A
Andrew Morton 已提交
306 307
 * 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.
308
 */
309
static __always_inline int index_of(const size_t size)
310
{
311 312
	extern void __bad_size(void);

313 314 315 316 317 318 319 320
	if (__builtin_constant_p(size)) {
		int i = 0;

#define CACHE(x) \
	if (size <=x) \
		return i; \
	else \
		i++;
321
#include <linux/kmalloc_sizes.h>
322
#undef CACHE
323
		__bad_size();
324
	} else
325
		__bad_size();
326 327 328
	return 0;
}

329 330
static int slab_early_init = 1;

331 332
#define INDEX_AC index_of(sizeof(struct arraycache_init))
#define INDEX_L3 index_of(sizeof(struct kmem_list3))
L
Linus Torvalds 已提交
333

P
Pekka Enberg 已提交
334
static void kmem_list3_init(struct kmem_list3 *parent)
335 336 337 338 339 340
{
	INIT_LIST_HEAD(&parent->slabs_full);
	INIT_LIST_HEAD(&parent->slabs_partial);
	INIT_LIST_HEAD(&parent->slabs_free);
	parent->shared = NULL;
	parent->alien = NULL;
341
	parent->colour_next = 0;
342 343 344 345 346
	spin_lock_init(&parent->list_lock);
	parent->free_objects = 0;
	parent->free_touched = 0;
}

A
Andrew Morton 已提交
347 348 349 350
#define MAKE_LIST(cachep, listp, slab, nodeid)				\
	do {								\
		INIT_LIST_HEAD(listp);					\
		list_splice(&(cachep->nodelists[nodeid]->slab), listp);	\
351 352
	} while (0)

A
Andrew Morton 已提交
353 354
#define	MAKE_ALL_LISTS(cachep, ptr, nodeid)				\
	do {								\
355 356 357 358
	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
359 360 361 362 363

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

#define BATCHREFILL_LIMIT	16
A
Andrew Morton 已提交
364 365 366
/*
 * Optimization question: fewer reaps means less probability for unnessary
 * cpucache drain/refill cycles.
L
Linus Torvalds 已提交
367
 *
A
Adrian Bunk 已提交
368
 * OTOH the cpuarrays can contain lots of objects,
L
Linus Torvalds 已提交
369 370 371 372 373 374 375 376 377 378
 * 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++)
379
#define	STATS_ADD_REAPED(x,y)	((x)->reaped += (y))
A
Andrew Morton 已提交
380 381 382 383 384
#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
385 386
#define	STATS_INC_ERR(x)	((x)->errors++)
#define	STATS_INC_NODEALLOCS(x)	((x)->node_allocs++)
387
#define	STATS_INC_NODEFREES(x)	((x)->node_frees++)
388
#define STATS_INC_ACOVERFLOW(x)   ((x)->node_overflow++)
A
Andrew Morton 已提交
389 390 391 392 393
#define	STATS_SET_FREEABLE(x, i)					\
	do {								\
		if ((x)->max_freeable < i)				\
			(x)->max_freeable = i;				\
	} while (0)
L
Linus Torvalds 已提交
394 395 396 397 398 399 400 401 402
#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)
403
#define	STATS_ADD_REAPED(x,y)	do { (void)(y); } while (0)
L
Linus Torvalds 已提交
404 405 406
#define	STATS_SET_HIGH(x)	do { } while (0)
#define	STATS_INC_ERR(x)	do { } while (0)
#define	STATS_INC_NODEALLOCS(x)	do { } while (0)
407
#define	STATS_INC_NODEFREES(x)	do { } while (0)
408
#define STATS_INC_ACOVERFLOW(x)   do { } while (0)
A
Andrew Morton 已提交
409
#define	STATS_SET_FREEABLE(x, i) do { } while (0)
L
Linus Torvalds 已提交
410 411 412 413 414 415 416 417
#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
418 419
/*
 * memory layout of objects:
L
Linus Torvalds 已提交
420
 * 0		: objp
421
 * 0 .. cachep->obj_offset - BYTES_PER_WORD - 1: padding. This ensures that
L
Linus Torvalds 已提交
422 423
 * 		the end of an object is aligned with the end of the real
 * 		allocation. Catches writes behind the end of the allocation.
424
 * cachep->obj_offset - BYTES_PER_WORD .. cachep->obj_offset - 1:
L
Linus Torvalds 已提交
425
 * 		redzone word.
426
 * cachep->obj_offset: The real object.
427 428
 * cachep->size - 2* BYTES_PER_WORD: redzone word [BYTES_PER_WORD long]
 * cachep->size - 1* BYTES_PER_WORD: last caller address
A
Andrew Morton 已提交
429
 *					[BYTES_PER_WORD long]
L
Linus Torvalds 已提交
430
 */
431
static int obj_offset(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
432
{
433
	return cachep->obj_offset;
L
Linus Torvalds 已提交
434 435
}

436
static unsigned long long *dbg_redzone1(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
437 438
{
	BUG_ON(!(cachep->flags & SLAB_RED_ZONE));
439 440
	return (unsigned long long*) (objp + obj_offset(cachep) -
				      sizeof(unsigned long long));
L
Linus Torvalds 已提交
441 442
}

443
static unsigned long long *dbg_redzone2(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
444 445 446
{
	BUG_ON(!(cachep->flags & SLAB_RED_ZONE));
	if (cachep->flags & SLAB_STORE_USER)
447
		return (unsigned long long *)(objp + cachep->size -
448
					      sizeof(unsigned long long) -
D
David Woodhouse 已提交
449
					      REDZONE_ALIGN);
450
	return (unsigned long long *) (objp + cachep->size -
451
				       sizeof(unsigned long long));
L
Linus Torvalds 已提交
452 453
}

454
static void **dbg_userword(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
455 456
{
	BUG_ON(!(cachep->flags & SLAB_STORE_USER));
457
	return (void **)(objp + cachep->size - BYTES_PER_WORD);
L
Linus Torvalds 已提交
458 459 460 461
}

#else

462
#define obj_offset(x)			0
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
size_t slab_buffer_size(struct kmem_cache *cachep)
{
472
	return cachep->size;
E
Eduard - Gabriel Munteanu 已提交
473 474 475 476
}
EXPORT_SYMBOL(slab_buffer_size);
#endif

L
Linus Torvalds 已提交
477
/*
478 479
 * Do not go above this order unless 0 objects fit into the slab or
 * overridden on the command line.
L
Linus Torvalds 已提交
480
 */
481 482 483
#define	SLAB_MAX_ORDER_HI	1
#define	SLAB_MAX_ORDER_LO	0
static int slab_max_order = SLAB_MAX_ORDER_LO;
484
static bool slab_max_order_set __initdata;
L
Linus Torvalds 已提交
485

486 487
static inline struct kmem_cache *page_get_cache(struct page *page)
{
488
	page = compound_head(page);
489
	BUG_ON(!PageSlab(page));
490
	return page->slab_cache;
491 492
}

493 494
static inline struct kmem_cache *virt_to_cache(const void *obj)
{
495
	struct page *page = virt_to_head_page(obj);
C
Christoph Lameter 已提交
496
	return page->slab_cache;
497 498 499 500
}

static inline struct slab *virt_to_slab(const void *obj)
{
501
	struct page *page = virt_to_head_page(obj);
C
Christoph Lameter 已提交
502 503 504

	VM_BUG_ON(!PageSlab(page));
	return page->slab_page;
505 506
}

507 508 509
static inline void *index_to_obj(struct kmem_cache *cache, struct slab *slab,
				 unsigned int idx)
{
510
	return slab->s_mem + cache->size * idx;
511 512
}

513
/*
514 515 516
 * We want to avoid an expensive divide : (offset / cache->size)
 *   Using the fact that size is a constant for a particular cache,
 *   we can replace (offset / cache->size) by
517 518 519 520
 *   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)
521
{
522 523
	u32 offset = (obj - slab->s_mem);
	return reciprocal_divide(offset, cache->reciprocal_buffer_size);
524 525
}

A
Andrew Morton 已提交
526 527 528
/*
 * These are the default caches for kmalloc. Custom caches can have other sizes.
 */
L
Linus Torvalds 已提交
529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545
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
546
	{NULL,}
L
Linus Torvalds 已提交
547 548 549 550
#undef CACHE
};

static struct arraycache_init initarray_cache __initdata =
P
Pekka Enberg 已提交
551
    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
L
Linus Torvalds 已提交
552
static struct arraycache_init initarray_generic =
P
Pekka Enberg 已提交
553
    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
L
Linus Torvalds 已提交
554 555

/* internal cache of cache description objs */
556
static struct kmem_list3 *cache_cache_nodelists[MAX_NUMNODES];
557
static struct kmem_cache cache_cache = {
558
	.nodelists = cache_cache_nodelists,
P
Pekka Enberg 已提交
559 560 561
	.batchcount = 1,
	.limit = BOOT_CPUCACHE_ENTRIES,
	.shared = 1,
562
	.size = sizeof(struct kmem_cache),
P
Pekka Enberg 已提交
563
	.name = "kmem_cache",
L
Linus Torvalds 已提交
564 565
};

566 567
#define BAD_ALIEN_MAGIC 0x01020304ul

568 569 570 571 572 573 574 575 576
/*
 * chicken and egg problem: delay the per-cpu array allocation
 * until the general caches are up.
 */
static enum {
	NONE,
	PARTIAL_AC,
	PARTIAL_L3,
	EARLY,
P
Peter Zijlstra 已提交
577
	LATE,
578 579 580 581 582 583 584 585 586 587 588
	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;
}

589 590 591 592 593 594 595 596
#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.
597 598 599 600
 *
 * 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
601
 */
602 603 604
static struct lock_class_key on_slab_l3_key;
static struct lock_class_key on_slab_alc_key;

605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649
static struct lock_class_key debugobj_l3_key;
static struct lock_class_key debugobj_alc_key;

static void slab_set_lock_classes(struct kmem_cache *cachep,
		struct lock_class_key *l3_key, struct lock_class_key *alc_key,
		int q)
{
	struct array_cache **alc;
	struct kmem_list3 *l3;
	int r;

	l3 = cachep->nodelists[q];
	if (!l3)
		return;

	lockdep_set_class(&l3->list_lock, 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)
		return;
	for_each_node(r) {
		if (alc[r])
			lockdep_set_class(&alc[r]->lock, alc_key);
	}
}

static void slab_set_debugobj_lock_classes_node(struct kmem_cache *cachep, int node)
{
	slab_set_lock_classes(cachep, &debugobj_l3_key, &debugobj_alc_key, node);
}

static void slab_set_debugobj_lock_classes(struct kmem_cache *cachep)
{
	int node;

	for_each_online_node(node)
		slab_set_debugobj_lock_classes_node(cachep, node);
}

650
static void init_node_lock_keys(int q)
651
{
652 653
	struct cache_sizes *s = malloc_sizes;

P
Peter Zijlstra 已提交
654
	if (g_cpucache_up < LATE)
655 656 657 658 659 660 661
		return;

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

		l3 = s->cs_cachep->nodelists[q];
		if (!l3 || OFF_SLAB(s->cs_cachep))
662
			continue;
663 664 665

		slab_set_lock_classes(s->cs_cachep, &on_slab_l3_key,
				&on_slab_alc_key, q);
666 667
	}
}
668 669 670 671 672 673 674 675

static inline void init_lock_keys(void)
{
	int node;

	for_each_node(node)
		init_node_lock_keys(node);
}
676
#else
677 678 679 680
static void init_node_lock_keys(int q)
{
}

681
static inline void init_lock_keys(void)
682 683
{
}
684 685 686 687 688 689 690 691

static void slab_set_debugobj_lock_classes_node(struct kmem_cache *cachep, int node)
{
}

static void slab_set_debugobj_lock_classes(struct kmem_cache *cachep)
{
}
692 693
#endif

694
/*
695
 * Guard access to the cache-chain.
696
 */
I
Ingo Molnar 已提交
697
static DEFINE_MUTEX(cache_chain_mutex);
L
Linus Torvalds 已提交
698 699
static struct list_head cache_chain;

700
static DEFINE_PER_CPU(struct delayed_work, slab_reap_work);
L
Linus Torvalds 已提交
701

702
static inline struct array_cache *cpu_cache_get(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
703 704 705 706
{
	return cachep->array[smp_processor_id()];
}

A
Andrew Morton 已提交
707 708
static inline struct kmem_cache *__find_general_cachep(size_t size,
							gfp_t gfpflags)
L
Linus Torvalds 已提交
709 710 711 712 713
{
	struct cache_sizes *csizep = malloc_sizes;

#if DEBUG
	/* This happens if someone tries to call
P
Pekka Enberg 已提交
714 715 716
	 * kmem_cache_create(), or __kmalloc(), before
	 * the generic caches are initialized.
	 */
717
	BUG_ON(malloc_sizes[INDEX_AC].cs_cachep == NULL);
L
Linus Torvalds 已提交
718
#endif
719 720 721
	if (!size)
		return ZERO_SIZE_PTR;

L
Linus Torvalds 已提交
722 723 724 725
	while (size > csizep->cs_size)
		csizep++;

	/*
726
	 * Really subtle: The last entry with cs->cs_size==ULONG_MAX
L
Linus Torvalds 已提交
727 728 729
	 * has cs_{dma,}cachep==NULL. Thus no special case
	 * for large kmalloc calls required.
	 */
730
#ifdef CONFIG_ZONE_DMA
L
Linus Torvalds 已提交
731 732
	if (unlikely(gfpflags & GFP_DMA))
		return csizep->cs_dmacachep;
733
#endif
L
Linus Torvalds 已提交
734 735 736
	return csizep->cs_cachep;
}

A
Adrian Bunk 已提交
737
static struct kmem_cache *kmem_find_general_cachep(size_t size, gfp_t gfpflags)
738 739 740 741
{
	return __find_general_cachep(size, gfpflags);
}

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

A
Andrew Morton 已提交
747 748 749
/*
 * Calculate the number of objects and left-over bytes for a given buffer size.
 */
750 751 752 753 754 755 756
static void cache_estimate(unsigned long gfporder, size_t buffer_size,
			   size_t align, int flags, size_t *left_over,
			   unsigned int *num)
{
	int nr_objs;
	size_t mgmt_size;
	size_t slab_size = PAGE_SIZE << gfporder;
L
Linus Torvalds 已提交
757

758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805
	/*
	 * The slab management structure can be either off the slab or
	 * on it. For the latter case, the memory allocated for a
	 * slab is used for:
	 *
	 * - The struct slab
	 * - One kmem_bufctl_t for each object
	 * - Padding to respect alignment of @align
	 * - @buffer_size bytes for each object
	 *
	 * If the slab management structure is off the slab, then the
	 * alignment will already be calculated into the size. Because
	 * the slabs are all pages aligned, the objects will be at the
	 * correct alignment when allocated.
	 */
	if (flags & CFLGS_OFF_SLAB) {
		mgmt_size = 0;
		nr_objs = slab_size / buffer_size;

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

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

		if (nr_objs > SLAB_LIMIT)
			nr_objs = SLAB_LIMIT;

		mgmt_size = slab_mgmt_size(nr_objs, align);
	}
	*num = nr_objs;
	*left_over = slab_size - nr_objs*buffer_size - mgmt_size;
L
Linus Torvalds 已提交
806 807
}

808
#define slab_error(cachep, msg) __slab_error(__func__, cachep, msg)
L
Linus Torvalds 已提交
809

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

818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833
/*
 * 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);

834 835 836 837 838 839 840 841 842 843 844
static int __init slab_max_order_setup(char *str)
{
	get_option(&str, &slab_max_order);
	slab_max_order = slab_max_order < 0 ? 0 :
				min(slab_max_order, MAX_ORDER - 1);
	slab_max_order_set = true;

	return 1;
}
__setup("slab_max_order=", slab_max_order_setup);

845 846 847 848 849 850 851
#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.
 */
852
static DEFINE_PER_CPU(unsigned long, slab_reap_node);
853 854 855 856 857

static void init_reap_node(int cpu)
{
	int node;

858
	node = next_node(cpu_to_mem(cpu), node_online_map);
859
	if (node == MAX_NUMNODES)
860
		node = first_node(node_online_map);
861

862
	per_cpu(slab_reap_node, cpu) = node;
863 864 865 866
}

static void next_reap_node(void)
{
867
	int node = __this_cpu_read(slab_reap_node);
868 869 870 871

	node = next_node(node, node_online_map);
	if (unlikely(node >= MAX_NUMNODES))
		node = first_node(node_online_map);
872
	__this_cpu_write(slab_reap_node, node);
873 874 875 876 877 878 879
}

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

L
Linus Torvalds 已提交
880 881 882 883 884 885 886
/*
 * 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.
 */
887
static void __cpuinit start_cpu_timer(int cpu)
L
Linus Torvalds 已提交
888
{
889
	struct delayed_work *reap_work = &per_cpu(slab_reap_work, cpu);
L
Linus Torvalds 已提交
890 891 892 893 894 895

	/*
	 * When this gets called from do_initcalls via cpucache_init(),
	 * init_workqueues() has already run, so keventd will be setup
	 * at that time.
	 */
896
	if (keventd_up() && reap_work->work.func == NULL) {
897
		init_reap_node(cpu);
898
		INIT_DELAYED_WORK_DEFERRABLE(reap_work, cache_reap);
899 900
		schedule_delayed_work_on(cpu, reap_work,
					__round_jiffies_relative(HZ, cpu));
L
Linus Torvalds 已提交
901 902 903
	}
}

904
static struct array_cache *alloc_arraycache(int node, int entries,
905
					    int batchcount, gfp_t gfp)
L
Linus Torvalds 已提交
906
{
P
Pekka Enberg 已提交
907
	int memsize = sizeof(void *) * entries + sizeof(struct array_cache);
L
Linus Torvalds 已提交
908 909
	struct array_cache *nc = NULL;

910
	nc = kmalloc_node(memsize, gfp, node);
911 912
	/*
	 * The array_cache structures contain pointers to free object.
L
Lucas De Marchi 已提交
913
	 * However, when such objects are allocated or transferred to another
914 915 916 917 918
	 * 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
919 920 921 922 923
	if (nc) {
		nc->avail = 0;
		nc->limit = entries;
		nc->batchcount = batchcount;
		nc->touched = 0;
924
		spin_lock_init(&nc->lock);
L
Linus Torvalds 已提交
925 926 927 928
	}
	return nc;
}

929 930 931 932 933 934 935 936 937 938
/*
 * 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 */
939
	int nr = min3(from->avail, max, to->limit - to->avail);
940 941 942 943 944 945 946 947 948 949 950 951

	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;
}

952 953 954 955 956
#ifndef CONFIG_NUMA

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

957
static inline struct array_cache **alloc_alien_cache(int node, int limit, gfp_t gfp)
958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976
{
	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;
}

977
static inline void *____cache_alloc_node(struct kmem_cache *cachep,
978 979 980 981 982 983 984
		 gfp_t flags, int nodeid)
{
	return NULL;
}

#else	/* CONFIG_NUMA */

985
static void *____cache_alloc_node(struct kmem_cache *, gfp_t, int);
986
static void *alternate_node_alloc(struct kmem_cache *, gfp_t);
987

988
static struct array_cache **alloc_alien_cache(int node, int limit, gfp_t gfp)
989 990
{
	struct array_cache **ac_ptr;
991
	int memsize = sizeof(void *) * nr_node_ids;
992 993 994 995
	int i;

	if (limit > 1)
		limit = 12;
996
	ac_ptr = kzalloc_node(memsize, gfp, node);
997 998
	if (ac_ptr) {
		for_each_node(i) {
999
			if (i == node || !node_online(i))
1000
				continue;
1001
			ac_ptr[i] = alloc_arraycache(node, limit, 0xbaadf00d, gfp);
1002
			if (!ac_ptr[i]) {
1003
				for (i--; i >= 0; i--)
1004 1005 1006 1007 1008 1009 1010 1011 1012
					kfree(ac_ptr[i]);
				kfree(ac_ptr);
				return NULL;
			}
		}
	}
	return ac_ptr;
}

P
Pekka Enberg 已提交
1013
static void free_alien_cache(struct array_cache **ac_ptr)
1014 1015 1016 1017 1018 1019
{
	int i;

	if (!ac_ptr)
		return;
	for_each_node(i)
P
Pekka Enberg 已提交
1020
	    kfree(ac_ptr[i]);
1021 1022 1023
	kfree(ac_ptr);
}

1024
static void __drain_alien_cache(struct kmem_cache *cachep,
P
Pekka Enberg 已提交
1025
				struct array_cache *ac, int node)
1026 1027 1028 1029 1030
{
	struct kmem_list3 *rl3 = cachep->nodelists[node];

	if (ac->avail) {
		spin_lock(&rl3->list_lock);
1031 1032 1033 1034 1035
		/*
		 * 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.
		 */
1036 1037
		if (rl3->shared)
			transfer_objects(rl3->shared, ac, ac->limit);
1038

1039
		free_block(cachep, ac->entry, ac->avail, node);
1040 1041 1042 1043 1044
		ac->avail = 0;
		spin_unlock(&rl3->list_lock);
	}
}

1045 1046 1047 1048 1049
/*
 * Called from cache_reap() to regularly drain alien caches round robin.
 */
static void reap_alien(struct kmem_cache *cachep, struct kmem_list3 *l3)
{
1050
	int node = __this_cpu_read(slab_reap_node);
1051 1052 1053

	if (l3->alien) {
		struct array_cache *ac = l3->alien[node];
1054 1055

		if (ac && ac->avail && spin_trylock_irq(&ac->lock)) {
1056 1057 1058 1059 1060 1061
			__drain_alien_cache(cachep, ac, node);
			spin_unlock_irq(&ac->lock);
		}
	}
}

A
Andrew Morton 已提交
1062 1063
static void drain_alien_cache(struct kmem_cache *cachep,
				struct array_cache **alien)
1064
{
P
Pekka Enberg 已提交
1065
	int i = 0;
1066 1067 1068 1069
	struct array_cache *ac;
	unsigned long flags;

	for_each_online_node(i) {
1070
		ac = alien[i];
1071 1072 1073 1074 1075 1076 1077
		if (ac) {
			spin_lock_irqsave(&ac->lock, flags);
			__drain_alien_cache(cachep, ac, i);
			spin_unlock_irqrestore(&ac->lock, flags);
		}
	}
}
1078

1079
static inline int cache_free_alien(struct kmem_cache *cachep, void *objp)
1080 1081 1082 1083 1084
{
	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
1085 1086
	int node;

1087
	node = numa_mem_id();
1088 1089 1090 1091 1092

	/*
	 * Make sure we are not freeing a object from another node to the array
	 * cache on this cpu.
	 */
1093
	if (likely(slabp->nodeid == node))
1094 1095
		return 0;

P
Pekka Enberg 已提交
1096
	l3 = cachep->nodelists[node];
1097 1098 1099
	STATS_INC_NODEFREES(cachep);
	if (l3->alien && l3->alien[nodeid]) {
		alien = l3->alien[nodeid];
1100
		spin_lock(&alien->lock);
1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113
		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;
}
1114 1115
#endif

1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130
/*
 * 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);

1131
	list_for_each_entry(cachep, &cache_chain, list) {
1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161
		/*
		 * 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;
}

1162 1163 1164 1165
static void __cpuinit cpuup_canceled(long cpu)
{
	struct kmem_cache *cachep;
	struct kmem_list3 *l3 = NULL;
1166
	int node = cpu_to_mem(cpu);
1167
	const struct cpumask *mask = cpumask_of_node(node);
1168

1169
	list_for_each_entry(cachep, &cache_chain, list) {
1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188
		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);

1189
		if (!cpumask_empty(mask)) {
1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218
			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.
	 */
1219
	list_for_each_entry(cachep, &cache_chain, list) {
1220 1221 1222 1223 1224 1225 1226 1227
		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
1228
{
1229
	struct kmem_cache *cachep;
1230
	struct kmem_list3 *l3 = NULL;
1231
	int node = cpu_to_mem(cpu);
1232
	int err;
L
Linus Torvalds 已提交
1233

1234 1235 1236 1237 1238 1239
	/*
	 * 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
	 */
1240 1241 1242
	err = init_cache_nodelists_node(node);
	if (err < 0)
		goto bad;
1243 1244 1245 1246 1247

	/*
	 * Now we can go ahead with allocating the shared arrays and
	 * array caches
	 */
1248
	list_for_each_entry(cachep, &cache_chain, list) {
1249 1250 1251 1252 1253
		struct array_cache *nc;
		struct array_cache *shared = NULL;
		struct array_cache **alien = NULL;

		nc = alloc_arraycache(node, cachep->limit,
1254
					cachep->batchcount, GFP_KERNEL);
1255 1256 1257 1258 1259
		if (!nc)
			goto bad;
		if (cachep->shared) {
			shared = alloc_arraycache(node,
				cachep->shared * cachep->batchcount,
1260
				0xbaadf00d, GFP_KERNEL);
1261 1262
			if (!shared) {
				kfree(nc);
L
Linus Torvalds 已提交
1263
				goto bad;
1264
			}
1265 1266
		}
		if (use_alien_caches) {
1267
			alien = alloc_alien_cache(node, cachep->limit, GFP_KERNEL);
1268 1269 1270
			if (!alien) {
				kfree(shared);
				kfree(nc);
1271
				goto bad;
1272
			}
1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286
		}
		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;
		}
1287
#ifdef CONFIG_NUMA
1288 1289 1290
		if (!l3->alien) {
			l3->alien = alien;
			alien = NULL;
L
Linus Torvalds 已提交
1291
		}
1292 1293 1294 1295
#endif
		spin_unlock_irq(&l3->list_lock);
		kfree(shared);
		free_alien_cache(alien);
1296 1297
		if (cachep->flags & SLAB_DEBUG_OBJECTS)
			slab_set_debugobj_lock_classes_node(cachep, node);
1298
	}
1299 1300
	init_node_lock_keys(node);

1301 1302
	return 0;
bad:
1303
	cpuup_canceled(cpu);
1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315
	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:
1316
		mutex_lock(&cache_chain_mutex);
1317
		err = cpuup_prepare(cpu);
1318
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1319 1320
		break;
	case CPU_ONLINE:
1321
	case CPU_ONLINE_FROZEN:
L
Linus Torvalds 已提交
1322 1323 1324
		start_cpu_timer(cpu);
		break;
#ifdef CONFIG_HOTPLUG_CPU
1325
  	case CPU_DOWN_PREPARE:
1326
  	case CPU_DOWN_PREPARE_FROZEN:
1327 1328 1329 1330 1331 1332
		/*
		 * 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.
		*/
1333
		cancel_delayed_work_sync(&per_cpu(slab_reap_work, cpu));
1334
		/* Now the cache_reaper is guaranteed to be not running. */
1335
		per_cpu(slab_reap_work, cpu).work.func = NULL;
1336 1337
  		break;
  	case CPU_DOWN_FAILED:
1338
  	case CPU_DOWN_FAILED_FROZEN:
1339 1340
		start_cpu_timer(cpu);
  		break;
L
Linus Torvalds 已提交
1341
	case CPU_DEAD:
1342
	case CPU_DEAD_FROZEN:
1343 1344 1345 1346 1347 1348 1349 1350
		/*
		 * 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
1351
		/* fall through */
1352
#endif
L
Linus Torvalds 已提交
1353
	case CPU_UP_CANCELED:
1354
	case CPU_UP_CANCELED_FROZEN:
1355
		mutex_lock(&cache_chain_mutex);
1356
		cpuup_canceled(cpu);
I
Ingo Molnar 已提交
1357
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1358 1359
		break;
	}
1360
	return notifier_from_errno(err);
L
Linus Torvalds 已提交
1361 1362
}

1363 1364 1365
static struct notifier_block __cpuinitdata cpucache_notifier = {
	&cpuup_callback, NULL, 0
};
L
Linus Torvalds 已提交
1366

1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379
#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;

1380
	list_for_each_entry(cachep, &cache_chain, list) {
1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426
		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:
1427
	return notifier_from_errno(ret);
1428 1429 1430
}
#endif /* CONFIG_NUMA && CONFIG_MEMORY_HOTPLUG */

1431 1432 1433
/*
 * swap the static kmem_list3 with kmalloced memory
 */
1434 1435
static void __init init_list(struct kmem_cache *cachep, struct kmem_list3 *list,
				int nodeid)
1436 1437 1438
{
	struct kmem_list3 *ptr;

1439
	ptr = kmalloc_node(sizeof(struct kmem_list3), GFP_NOWAIT, nodeid);
1440 1441 1442
	BUG_ON(!ptr);

	memcpy(ptr, list, sizeof(struct kmem_list3));
1443 1444 1445 1446 1447
	/*
	 * Do not assume that spinlocks can be initialized via memcpy:
	 */
	spin_lock_init(&ptr->list_lock);

1448 1449 1450 1451
	MAKE_ALL_LISTS(cachep, ptr, nodeid);
	cachep->nodelists[nodeid] = ptr;
}

1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467
/*
 * 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
1468 1469 1470
/*
 * Initialisation.  Called after the page allocator have been initialised and
 * before smp_init().
L
Linus Torvalds 已提交
1471 1472 1473 1474 1475 1476
 */
void __init kmem_cache_init(void)
{
	size_t left_over;
	struct cache_sizes *sizes;
	struct cache_names *names;
1477
	int i;
1478
	int order;
P
Pekka Enberg 已提交
1479
	int node;
1480

1481
	if (num_possible_nodes() == 1)
1482 1483
		use_alien_caches = 0;

1484 1485 1486 1487 1488
	for (i = 0; i < NUM_INIT_LISTS; i++) {
		kmem_list3_init(&initkmem_list3[i]);
		if (i < MAX_NUMNODES)
			cache_cache.nodelists[i] = NULL;
	}
1489
	set_up_list3s(&cache_cache, CACHE_CACHE);
L
Linus Torvalds 已提交
1490 1491 1492

	/*
	 * Fragmentation resistance on low memory - only use bigger
1493 1494
	 * page orders on machines with more than 32MB of memory if
	 * not overridden on the command line.
L
Linus Torvalds 已提交
1495
	 */
1496
	if (!slab_max_order_set && totalram_pages > (32 << 20) >> PAGE_SHIFT)
1497
		slab_max_order = SLAB_MAX_ORDER_HI;
L
Linus Torvalds 已提交
1498 1499 1500

	/* Bootstrap is tricky, because several objects are allocated
	 * from caches that do not exist yet:
A
Andrew Morton 已提交
1501 1502 1503
	 * 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.
1504 1505 1506
	 *    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
1507
	 * 2) Create the first kmalloc cache.
1508
	 *    The struct kmem_cache for the new cache is allocated normally.
1509 1510 1511
	 *    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
1512 1513
	 * 4) Replace the __init data head arrays for cache_cache and the first
	 *    kmalloc cache with kmalloc allocated arrays.
1514 1515 1516
	 * 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
1517 1518
	 */

1519
	node = numa_mem_id();
P
Pekka Enberg 已提交
1520

L
Linus Torvalds 已提交
1521 1522
	/* 1) create the cache_cache */
	INIT_LIST_HEAD(&cache_chain);
1523
	list_add(&cache_cache.list, &cache_chain);
L
Linus Torvalds 已提交
1524 1525
	cache_cache.colour_off = cache_line_size();
	cache_cache.array[smp_processor_id()] = &initarray_cache.cache;
1526
	cache_cache.nodelists[node] = &initkmem_list3[CACHE_CACHE + node];
L
Linus Torvalds 已提交
1527

E
Eric Dumazet 已提交
1528
	/*
1529
	 * struct kmem_cache size depends on nr_node_ids & nr_cpu_ids
E
Eric Dumazet 已提交
1530
	 */
1531
	cache_cache.size = offsetof(struct kmem_cache, array[nr_cpu_ids]) +
1532
				  nr_node_ids * sizeof(struct kmem_list3 *);
1533 1534
	cache_cache.object_size = cache_cache.size;
	cache_cache.size = ALIGN(cache_cache.size,
A
Andrew Morton 已提交
1535
					cache_line_size());
1536
	cache_cache.reciprocal_buffer_size =
1537
		reciprocal_value(cache_cache.size);
L
Linus Torvalds 已提交
1538

1539
	for (order = 0; order < MAX_ORDER; order++) {
1540
		cache_estimate(order, cache_cache.size,
1541 1542 1543 1544
			cache_line_size(), 0, &left_over, &cache_cache.num);
		if (cache_cache.num)
			break;
	}
1545
	BUG_ON(!cache_cache.num);
1546
	cache_cache.gfporder = order;
P
Pekka Enberg 已提交
1547 1548 1549
	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
1550 1551 1552 1553 1554

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

A
Andrew Morton 已提交
1555 1556 1557 1558
	/*
	 * Initialize the caches that provide memory for the array cache and the
	 * kmem_list3 structures first.  Without this, further allocations will
	 * bug.
1559 1560 1561
	 */

	sizes[INDEX_AC].cs_cachep = kmem_cache_create(names[INDEX_AC].name,
A
Andrew Morton 已提交
1562 1563 1564
					sizes[INDEX_AC].cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_PANIC,
1565
					NULL);
1566

A
Andrew Morton 已提交
1567
	if (INDEX_AC != INDEX_L3) {
1568
		sizes[INDEX_L3].cs_cachep =
A
Andrew Morton 已提交
1569 1570 1571 1572
			kmem_cache_create(names[INDEX_L3].name,
				sizes[INDEX_L3].cs_size,
				ARCH_KMALLOC_MINALIGN,
				ARCH_KMALLOC_FLAGS|SLAB_PANIC,
1573
				NULL);
A
Andrew Morton 已提交
1574
	}
1575

1576 1577
	slab_early_init = 0;

L
Linus Torvalds 已提交
1578
	while (sizes->cs_size != ULONG_MAX) {
1579 1580
		/*
		 * For performance, all the general caches are L1 aligned.
L
Linus Torvalds 已提交
1581 1582 1583
		 * This should be particularly beneficial on SMP boxes, as it
		 * eliminates "false sharing".
		 * Note for systems short on memory removing the alignment will
1584 1585
		 * allow tighter packing of the smaller caches.
		 */
A
Andrew Morton 已提交
1586
		if (!sizes->cs_cachep) {
1587
			sizes->cs_cachep = kmem_cache_create(names->name,
A
Andrew Morton 已提交
1588 1589 1590
					sizes->cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_PANIC,
1591
					NULL);
A
Andrew Morton 已提交
1592
		}
1593 1594 1595
#ifdef CONFIG_ZONE_DMA
		sizes->cs_dmacachep = kmem_cache_create(
					names->name_dma,
A
Andrew Morton 已提交
1596 1597 1598 1599
					sizes->cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_CACHE_DMA|
						SLAB_PANIC,
1600
					NULL);
1601
#endif
L
Linus Torvalds 已提交
1602 1603 1604 1605 1606
		sizes++;
		names++;
	}
	/* 4) Replace the bootstrap head arrays */
	{
1607
		struct array_cache *ptr;
1608

1609
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_NOWAIT);
1610

1611 1612
		BUG_ON(cpu_cache_get(&cache_cache) != &initarray_cache.cache);
		memcpy(ptr, cpu_cache_get(&cache_cache),
P
Pekka Enberg 已提交
1613
		       sizeof(struct arraycache_init));
1614 1615 1616 1617 1618
		/*
		 * Do not assume that spinlocks can be initialized via memcpy:
		 */
		spin_lock_init(&ptr->lock);

L
Linus Torvalds 已提交
1619
		cache_cache.array[smp_processor_id()] = ptr;
1620

1621
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_NOWAIT);
1622

1623
		BUG_ON(cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep)
P
Pekka Enberg 已提交
1624
		       != &initarray_generic.cache);
1625
		memcpy(ptr, cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep),
P
Pekka Enberg 已提交
1626
		       sizeof(struct arraycache_init));
1627 1628 1629 1630 1631
		/*
		 * Do not assume that spinlocks can be initialized via memcpy:
		 */
		spin_lock_init(&ptr->lock);

1632
		malloc_sizes[INDEX_AC].cs_cachep->array[smp_processor_id()] =
P
Pekka Enberg 已提交
1633
		    ptr;
L
Linus Torvalds 已提交
1634
	}
1635 1636
	/* 5) Replace the bootstrap kmem_list3's */
	{
P
Pekka Enberg 已提交
1637 1638
		int nid;

1639
		for_each_online_node(nid) {
1640
			init_list(&cache_cache, &initkmem_list3[CACHE_CACHE + nid], nid);
1641

1642
			init_list(malloc_sizes[INDEX_AC].cs_cachep,
P
Pekka Enberg 已提交
1643
				  &initkmem_list3[SIZE_AC + nid], nid);
1644 1645 1646

			if (INDEX_AC != INDEX_L3) {
				init_list(malloc_sizes[INDEX_L3].cs_cachep,
P
Pekka Enberg 已提交
1647
					  &initkmem_list3[SIZE_L3 + nid], nid);
1648 1649 1650
			}
		}
	}
L
Linus Torvalds 已提交
1651

1652 1653 1654 1655 1656 1657 1658
	g_cpucache_up = EARLY;
}

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

P
Peter Zijlstra 已提交
1659 1660
	g_cpucache_up = LATE;

1661 1662 1663
	/* Annotate slab for lockdep -- annotate the malloc caches */
	init_lock_keys();

1664 1665
	/* 6) resize the head arrays to their final sizes */
	mutex_lock(&cache_chain_mutex);
1666
	list_for_each_entry(cachep, &cache_chain, list)
1667 1668 1669
		if (enable_cpucache(cachep, GFP_NOWAIT))
			BUG();
	mutex_unlock(&cache_chain_mutex);
1670

A
Andrew Morton 已提交
1671 1672 1673
	/*
	 * Register a cpu startup notifier callback that initializes
	 * cpu_cache_get for all new cpus
L
Linus Torvalds 已提交
1674 1675 1676
	 */
	register_cpu_notifier(&cpucache_notifier);

1677 1678 1679 1680 1681 1682 1683 1684
#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
1685 1686 1687
	/*
	 * 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
1688 1689 1690 1691 1692 1693 1694
	 */
}

static int __init cpucache_init(void)
{
	int cpu;

A
Andrew Morton 已提交
1695 1696
	/*
	 * Register the timers that return unneeded pages to the page allocator
L
Linus Torvalds 已提交
1697
	 */
1698
	for_each_online_cpu(cpu)
A
Andrew Morton 已提交
1699
		start_cpu_timer(cpu);
1700 1701 1702

	/* Done! */
	g_cpucache_up = FULL;
L
Linus Torvalds 已提交
1703 1704 1705 1706
	return 0;
}
__initcall(cpucache_init);

1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718
static noinline void
slab_out_of_memory(struct kmem_cache *cachep, gfp_t gfpflags, int nodeid)
{
	struct kmem_list3 *l3;
	struct slab *slabp;
	unsigned long flags;
	int node;

	printk(KERN_WARNING
		"SLAB: Unable to allocate memory on node %d (gfp=0x%x)\n",
		nodeid, gfpflags);
	printk(KERN_WARNING "  cache: %s, object size: %d, order: %d\n",
1719
		cachep->name, cachep->size, cachep->gfporder);
1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752

	for_each_online_node(node) {
		unsigned long active_objs = 0, num_objs = 0, free_objects = 0;
		unsigned long active_slabs = 0, num_slabs = 0;

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

		spin_lock_irqsave(&l3->list_lock, flags);
		list_for_each_entry(slabp, &l3->slabs_full, list) {
			active_objs += cachep->num;
			active_slabs++;
		}
		list_for_each_entry(slabp, &l3->slabs_partial, list) {
			active_objs += slabp->inuse;
			active_slabs++;
		}
		list_for_each_entry(slabp, &l3->slabs_free, list)
			num_slabs++;

		free_objects += l3->free_objects;
		spin_unlock_irqrestore(&l3->list_lock, flags);

		num_slabs += active_slabs;
		num_objs = num_slabs * cachep->num;
		printk(KERN_WARNING
			"  node %d: slabs: %ld/%ld, objs: %ld/%ld, free: %ld\n",
			node, active_slabs, num_slabs, active_objs, num_objs,
			free_objects);
	}
}

L
Linus Torvalds 已提交
1753 1754 1755 1756 1757 1758 1759
/*
 * 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.
 */
1760
static void *kmem_getpages(struct kmem_cache *cachep, gfp_t flags, int nodeid)
L
Linus Torvalds 已提交
1761 1762
{
	struct page *page;
1763
	int nr_pages;
L
Linus Torvalds 已提交
1764 1765
	int i;

1766
#ifndef CONFIG_MMU
1767 1768 1769
	/*
	 * Nommu uses slab's for process anonymous memory allocations, and thus
	 * requires __GFP_COMP to properly refcount higher order allocations
1770
	 */
1771
	flags |= __GFP_COMP;
1772
#endif
1773

1774
	flags |= cachep->allocflags;
1775 1776
	if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
		flags |= __GFP_RECLAIMABLE;
1777

L
Linus Torvalds 已提交
1778
	page = alloc_pages_exact_node(nodeid, flags | __GFP_NOTRACK, cachep->gfporder);
1779 1780 1781
	if (!page) {
		if (!(flags & __GFP_NOWARN) && printk_ratelimit())
			slab_out_of_memory(cachep, flags, nodeid);
L
Linus Torvalds 已提交
1782
		return NULL;
1783
	}
L
Linus Torvalds 已提交
1784

1785
	nr_pages = (1 << cachep->gfporder);
L
Linus Torvalds 已提交
1786
	if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
1787 1788 1789 1790 1791
		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);
1792 1793
	for (i = 0; i < nr_pages; i++)
		__SetPageSlab(page + i);
P
Pekka Enberg 已提交
1794

1795 1796 1797 1798 1799 1800 1801 1802
	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
1803

1804
	return page_address(page);
L
Linus Torvalds 已提交
1805 1806 1807 1808 1809
}

/*
 * Interface to system's page release.
 */
1810
static void kmem_freepages(struct kmem_cache *cachep, void *addr)
L
Linus Torvalds 已提交
1811
{
P
Pekka Enberg 已提交
1812
	unsigned long i = (1 << cachep->gfporder);
L
Linus Torvalds 已提交
1813 1814 1815
	struct page *page = virt_to_page(addr);
	const unsigned long nr_freed = i;

1816
	kmemcheck_free_shadow(page, cachep->gfporder);
P
Pekka Enberg 已提交
1817

1818 1819 1820 1821 1822 1823
	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
1824
	while (i--) {
N
Nick Piggin 已提交
1825 1826
		BUG_ON(!PageSlab(page));
		__ClearPageSlab(page);
L
Linus Torvalds 已提交
1827 1828 1829 1830 1831 1832 1833 1834 1835
		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
1836
	struct slab_rcu *slab_rcu = (struct slab_rcu *)head;
1837
	struct kmem_cache *cachep = slab_rcu->cachep;
L
Linus Torvalds 已提交
1838 1839 1840 1841 1842 1843 1844 1845 1846

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

#if DEBUG

#ifdef CONFIG_DEBUG_PAGEALLOC
1847
static void store_stackinfo(struct kmem_cache *cachep, unsigned long *addr,
P
Pekka Enberg 已提交
1848
			    unsigned long caller)
L
Linus Torvalds 已提交
1849
{
1850
	int size = cachep->object_size;
L
Linus Torvalds 已提交
1851

1852
	addr = (unsigned long *)&((char *)addr)[obj_offset(cachep)];
L
Linus Torvalds 已提交
1853

P
Pekka Enberg 已提交
1854
	if (size < 5 * sizeof(unsigned long))
L
Linus Torvalds 已提交
1855 1856
		return;

P
Pekka Enberg 已提交
1857 1858 1859 1860
	*addr++ = 0x12345678;
	*addr++ = caller;
	*addr++ = smp_processor_id();
	size -= 3 * sizeof(unsigned long);
L
Linus Torvalds 已提交
1861 1862 1863 1864 1865 1866 1867
	{
		unsigned long *sptr = &caller;
		unsigned long svalue;

		while (!kstack_end(sptr)) {
			svalue = *sptr++;
			if (kernel_text_address(svalue)) {
P
Pekka Enberg 已提交
1868
				*addr++ = svalue;
L
Linus Torvalds 已提交
1869 1870 1871 1872 1873 1874 1875
				size -= sizeof(unsigned long);
				if (size <= sizeof(unsigned long))
					break;
			}
		}

	}
P
Pekka Enberg 已提交
1876
	*addr++ = 0x87654321;
L
Linus Torvalds 已提交
1877 1878 1879
}
#endif

1880
static void poison_obj(struct kmem_cache *cachep, void *addr, unsigned char val)
L
Linus Torvalds 已提交
1881
{
1882
	int size = cachep->object_size;
1883
	addr = &((char *)addr)[obj_offset(cachep)];
L
Linus Torvalds 已提交
1884 1885

	memset(addr, val, size);
P
Pekka Enberg 已提交
1886
	*(unsigned char *)(addr + size - 1) = POISON_END;
L
Linus Torvalds 已提交
1887 1888 1889 1890 1891
}

static void dump_line(char *data, int offset, int limit)
{
	int i;
D
Dave Jones 已提交
1892 1893 1894
	unsigned char error = 0;
	int bad_count = 0;

1895
	printk(KERN_ERR "%03x: ", offset);
D
Dave Jones 已提交
1896 1897 1898 1899 1900 1901
	for (i = 0; i < limit; i++) {
		if (data[offset + i] != POISON_FREE) {
			error = data[offset + i];
			bad_count++;
		}
	}
1902 1903
	print_hex_dump(KERN_CONT, "", 0, 16, 1,
			&data[offset], limit, 1);
D
Dave Jones 已提交
1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917

	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
1918 1919 1920 1921 1922
}
#endif

#if DEBUG

1923
static void print_objinfo(struct kmem_cache *cachep, void *objp, int lines)
L
Linus Torvalds 已提交
1924 1925 1926 1927 1928
{
	int i, size;
	char *realobj;

	if (cachep->flags & SLAB_RED_ZONE) {
1929
		printk(KERN_ERR "Redzone: 0x%llx/0x%llx.\n",
A
Andrew Morton 已提交
1930 1931
			*dbg_redzone1(cachep, objp),
			*dbg_redzone2(cachep, objp));
L
Linus Torvalds 已提交
1932 1933 1934 1935
	}

	if (cachep->flags & SLAB_STORE_USER) {
		printk(KERN_ERR "Last user: [<%p>]",
A
Andrew Morton 已提交
1936
			*dbg_userword(cachep, objp));
L
Linus Torvalds 已提交
1937
		print_symbol("(%s)",
A
Andrew Morton 已提交
1938
				(unsigned long)*dbg_userword(cachep, objp));
L
Linus Torvalds 已提交
1939 1940
		printk("\n");
	}
1941
	realobj = (char *)objp + obj_offset(cachep);
1942
	size = cachep->object_size;
P
Pekka Enberg 已提交
1943
	for (i = 0; i < size && lines; i += 16, lines--) {
L
Linus Torvalds 已提交
1944 1945
		int limit;
		limit = 16;
P
Pekka Enberg 已提交
1946 1947
		if (i + limit > size)
			limit = size - i;
L
Linus Torvalds 已提交
1948 1949 1950 1951
		dump_line(realobj, i, limit);
	}
}

1952
static void check_poison_obj(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
1953 1954 1955 1956 1957
{
	char *realobj;
	int size, i;
	int lines = 0;

1958
	realobj = (char *)objp + obj_offset(cachep);
1959
	size = cachep->object_size;
L
Linus Torvalds 已提交
1960

P
Pekka Enberg 已提交
1961
	for (i = 0; i < size; i++) {
L
Linus Torvalds 已提交
1962
		char exp = POISON_FREE;
P
Pekka Enberg 已提交
1963
		if (i == size - 1)
L
Linus Torvalds 已提交
1964 1965 1966 1967 1968 1969
			exp = POISON_END;
		if (realobj[i] != exp) {
			int limit;
			/* Mismatch ! */
			/* Print header */
			if (lines == 0) {
P
Pekka Enberg 已提交
1970
				printk(KERN_ERR
1971 1972
					"Slab corruption (%s): %s start=%p, len=%d\n",
					print_tainted(), cachep->name, realobj, size);
L
Linus Torvalds 已提交
1973 1974 1975
				print_objinfo(cachep, objp, 0);
			}
			/* Hexdump the affected line */
P
Pekka Enberg 已提交
1976
			i = (i / 16) * 16;
L
Linus Torvalds 已提交
1977
			limit = 16;
P
Pekka Enberg 已提交
1978 1979
			if (i + limit > size)
				limit = size - i;
L
Linus Torvalds 已提交
1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991
			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:
		 */
1992
		struct slab *slabp = virt_to_slab(objp);
1993
		unsigned int objnr;
L
Linus Torvalds 已提交
1994

1995
		objnr = obj_to_index(cachep, slabp, objp);
L
Linus Torvalds 已提交
1996
		if (objnr) {
1997
			objp = index_to_obj(cachep, slabp, objnr - 1);
1998
			realobj = (char *)objp + obj_offset(cachep);
L
Linus Torvalds 已提交
1999
			printk(KERN_ERR "Prev obj: start=%p, len=%d\n",
P
Pekka Enberg 已提交
2000
			       realobj, size);
L
Linus Torvalds 已提交
2001 2002
			print_objinfo(cachep, objp, 2);
		}
P
Pekka Enberg 已提交
2003
		if (objnr + 1 < cachep->num) {
2004
			objp = index_to_obj(cachep, slabp, objnr + 1);
2005
			realobj = (char *)objp + obj_offset(cachep);
L
Linus Torvalds 已提交
2006
			printk(KERN_ERR "Next obj: start=%p, len=%d\n",
P
Pekka Enberg 已提交
2007
			       realobj, size);
L
Linus Torvalds 已提交
2008 2009 2010 2011 2012 2013
			print_objinfo(cachep, objp, 2);
		}
	}
}
#endif

2014
#if DEBUG
R
Rabin Vincent 已提交
2015
static void slab_destroy_debugcheck(struct kmem_cache *cachep, struct slab *slabp)
L
Linus Torvalds 已提交
2016 2017 2018
{
	int i;
	for (i = 0; i < cachep->num; i++) {
2019
		void *objp = index_to_obj(cachep, slabp, i);
L
Linus Torvalds 已提交
2020 2021 2022

		if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
2023
			if (cachep->size % PAGE_SIZE == 0 &&
A
Andrew Morton 已提交
2024
					OFF_SLAB(cachep))
P
Pekka Enberg 已提交
2025
				kernel_map_pages(virt_to_page(objp),
2026
					cachep->size / PAGE_SIZE, 1);
L
Linus Torvalds 已提交
2027 2028 2029 2030 2031 2032 2033 2034 2035
			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
2036
					   "was overwritten");
L
Linus Torvalds 已提交
2037 2038
			if (*dbg_redzone2(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "end of a freed object "
P
Pekka Enberg 已提交
2039
					   "was overwritten");
L
Linus Torvalds 已提交
2040 2041
		}
	}
2042
}
L
Linus Torvalds 已提交
2043
#else
R
Rabin Vincent 已提交
2044
static void slab_destroy_debugcheck(struct kmem_cache *cachep, struct slab *slabp)
2045 2046
{
}
L
Linus Torvalds 已提交
2047 2048
#endif

2049 2050 2051 2052 2053
/**
 * slab_destroy - destroy and release all objects in a slab
 * @cachep: cache pointer being destroyed
 * @slabp: slab pointer being destroyed
 *
2054
 * Destroy all the objs in a slab, and release the mem back to the system.
A
Andrew Morton 已提交
2055 2056
 * Before calling the slab must have been unlinked from the cache.  The
 * cache-lock is not held/needed.
2057
 */
2058
static void slab_destroy(struct kmem_cache *cachep, struct slab *slabp)
2059 2060 2061
{
	void *addr = slabp->s_mem - slabp->colouroff;

R
Rabin Vincent 已提交
2062
	slab_destroy_debugcheck(cachep, slabp);
L
Linus Torvalds 已提交
2063 2064 2065
	if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU)) {
		struct slab_rcu *slab_rcu;

P
Pekka Enberg 已提交
2066
		slab_rcu = (struct slab_rcu *)slabp;
L
Linus Torvalds 已提交
2067 2068 2069 2070 2071
		slab_rcu->cachep = cachep;
		slab_rcu->addr = addr;
		call_rcu(&slab_rcu->head, kmem_rcu_free);
	} else {
		kmem_freepages(cachep, addr);
2072 2073
		if (OFF_SLAB(cachep))
			kmem_cache_free(cachep->slabp_cache, slabp);
L
Linus Torvalds 已提交
2074 2075 2076
	}
}

2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097
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);
}


2098
/**
2099 2100 2101 2102 2103 2104 2105
 * 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.
2106 2107 2108 2109 2110
 *
 * 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
2111
static size_t calculate_slab_order(struct kmem_cache *cachep,
R
Randy Dunlap 已提交
2112
			size_t size, size_t align, unsigned long flags)
2113
{
2114
	unsigned long offslab_limit;
2115
	size_t left_over = 0;
2116
	int gfporder;
2117

2118
	for (gfporder = 0; gfporder <= KMALLOC_MAX_ORDER; gfporder++) {
2119 2120 2121
		unsigned int num;
		size_t remainder;

2122
		cache_estimate(gfporder, size, align, flags, &remainder, &num);
2123 2124
		if (!num)
			continue;
2125

2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137
		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;
		}
2138

2139
		/* Found something acceptable - save it away */
2140
		cachep->num = num;
2141
		cachep->gfporder = gfporder;
2142 2143
		left_over = remainder;

2144 2145 2146 2147 2148 2149 2150 2151
		/*
		 * 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;

2152 2153 2154 2155
		/*
		 * Large number of objects is good, but very large slabs are
		 * currently bad for the gfp()s.
		 */
2156
		if (gfporder >= slab_max_order)
2157 2158
			break;

2159 2160 2161
		/*
		 * Acceptable internal fragmentation?
		 */
A
Andrew Morton 已提交
2162
		if (left_over * 8 <= (PAGE_SIZE << gfporder))
2163 2164 2165 2166 2167
			break;
	}
	return left_over;
}

2168
static int __init_refok setup_cpu_cache(struct kmem_cache *cachep, gfp_t gfp)
2169
{
2170
	if (g_cpucache_up >= LATE)
2171
		return enable_cpucache(cachep, gfp);
2172

2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192
	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()] =
2193
			kmalloc(sizeof(struct arraycache_init), gfp);
2194 2195 2196 2197 2198 2199

		if (g_cpucache_up == PARTIAL_AC) {
			set_up_list3s(cachep, SIZE_L3);
			g_cpucache_up = PARTIAL_L3;
		} else {
			int node;
2200
			for_each_online_node(node) {
2201 2202
				cachep->nodelists[node] =
				    kmalloc_node(sizeof(struct kmem_list3),
2203
						gfp, node);
2204 2205 2206 2207 2208
				BUG_ON(!cachep->nodelists[node]);
				kmem_list3_init(cachep->nodelists[node]);
			}
		}
	}
2209
	cachep->nodelists[numa_mem_id()]->next_reap =
2210 2211 2212 2213 2214 2215 2216 2217 2218
			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;
2219
	return 0;
2220 2221
}

L
Linus Torvalds 已提交
2222 2223 2224 2225 2226 2227 2228 2229 2230 2231
/**
 * 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.
2232
 * The @ctor is run when new pages are allocated by the cache.
L
Linus Torvalds 已提交
2233 2234
 *
 * @name must be valid until the cache is destroyed. This implies that
A
Andrew Morton 已提交
2235 2236
 * the module calling this has to destroy the cache before getting unloaded.
 *
L
Linus Torvalds 已提交
2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248
 * 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.
 */
2249
struct kmem_cache *
L
Linus Torvalds 已提交
2250
kmem_cache_create (const char *name, size_t size, size_t align,
2251
	unsigned long flags, void (*ctor)(void *))
L
Linus Torvalds 已提交
2252 2253
{
	size_t left_over, slab_size, ralign;
2254
	struct kmem_cache *cachep = NULL, *pc;
2255
	gfp_t gfp;
L
Linus Torvalds 已提交
2256 2257 2258 2259

	/*
	 * Sanity checks... these are all serious usage bugs.
	 */
A
Andrew Morton 已提交
2260
	if (!name || in_interrupt() || (size < BYTES_PER_WORD) ||
2261
	    size > KMALLOC_MAX_SIZE) {
2262
		printk(KERN_ERR "%s: Early error in slab %s\n", __func__,
A
Andrew Morton 已提交
2263
				name);
P
Pekka Enberg 已提交
2264 2265
		BUG();
	}
L
Linus Torvalds 已提交
2266

2267
	/*
2268
	 * We use cache_chain_mutex to ensure a consistent view of
R
Rusty Russell 已提交
2269
	 * cpu_online_mask as well.  Please see cpuup_callback
2270
	 */
2271 2272 2273 2274
	if (slab_is_available()) {
		get_online_cpus();
		mutex_lock(&cache_chain_mutex);
	}
2275

2276
	list_for_each_entry(pc, &cache_chain, list) {
2277 2278 2279 2280 2281 2282 2283 2284
		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.
		 */
2285
		res = probe_kernel_address(pc->name, tmp);
2286
		if (res) {
2287 2288
			printk(KERN_ERR
			       "SLAB: cache with size %d has lost its name\n",
2289
			       pc->size);
2290 2291 2292
			continue;
		}

P
Pekka Enberg 已提交
2293
		if (!strcmp(pc->name, name)) {
2294 2295
			printk(KERN_ERR
			       "kmem_cache_create: duplicate cache %s\n", name);
2296 2297 2298 2299 2300
			dump_stack();
			goto oops;
		}
	}

L
Linus Torvalds 已提交
2301 2302 2303 2304 2305 2306 2307 2308 2309
#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
2310 2311
	if (size < 4096 || fls(size - 1) == fls(size-1 + REDZONE_ALIGN +
						2 * sizeof(unsigned long long)))
P
Pekka Enberg 已提交
2312
		flags |= SLAB_RED_ZONE | SLAB_STORE_USER;
L
Linus Torvalds 已提交
2313 2314 2315 2316 2317 2318 2319
	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
2320 2321
	 * Always checks flags, a caller might be expecting debug support which
	 * isn't available.
L
Linus Torvalds 已提交
2322
	 */
2323
	BUG_ON(flags & ~CREATE_MASK);
L
Linus Torvalds 已提交
2324

A
Andrew Morton 已提交
2325 2326
	/*
	 * Check that size is in terms of words.  This is needed to avoid
L
Linus Torvalds 已提交
2327 2328 2329
	 * 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
2330 2331 2332
	if (size & (BYTES_PER_WORD - 1)) {
		size += (BYTES_PER_WORD - 1);
		size &= ~(BYTES_PER_WORD - 1);
L
Linus Torvalds 已提交
2333 2334
	}

A
Andrew Morton 已提交
2335 2336
	/* calculate the final buffer alignment: */

L
Linus Torvalds 已提交
2337 2338
	/* 1) arch recommendation: can be overridden for debug */
	if (flags & SLAB_HWCACHE_ALIGN) {
A
Andrew Morton 已提交
2339 2340 2341 2342
		/*
		 * 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
2343 2344
		 */
		ralign = cache_line_size();
P
Pekka Enberg 已提交
2345
		while (size <= ralign / 2)
L
Linus Torvalds 已提交
2346 2347 2348 2349
			ralign /= 2;
	} else {
		ralign = BYTES_PER_WORD;
	}
2350 2351

	/*
D
David Woodhouse 已提交
2352 2353 2354
	 * 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.
2355
	 */
D
David Woodhouse 已提交
2356 2357 2358 2359 2360 2361 2362 2363 2364 2365
	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);
	}
2366

2367
	/* 2) arch mandated alignment */
L
Linus Torvalds 已提交
2368 2369 2370
	if (ralign < ARCH_SLAB_MINALIGN) {
		ralign = ARCH_SLAB_MINALIGN;
	}
2371
	/* 3) caller mandated alignment */
L
Linus Torvalds 已提交
2372 2373 2374
	if (ralign < align) {
		ralign = align;
	}
2375 2376
	/* disable debug if necessary */
	if (ralign > __alignof__(unsigned long long))
2377
		flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
A
Andrew Morton 已提交
2378
	/*
2379
	 * 4) Store it.
L
Linus Torvalds 已提交
2380 2381 2382
	 */
	align = ralign;

2383 2384 2385 2386 2387
	if (slab_is_available())
		gfp = GFP_KERNEL;
	else
		gfp = GFP_NOWAIT;

L
Linus Torvalds 已提交
2388
	/* Get cache's description obj. */
2389
	cachep = kmem_cache_zalloc(&cache_cache, gfp);
L
Linus Torvalds 已提交
2390
	if (!cachep)
2391
		goto oops;
L
Linus Torvalds 已提交
2392

2393
	cachep->nodelists = (struct kmem_list3 **)&cachep->array[nr_cpu_ids];
2394 2395
	cachep->object_size = size;
	cachep->align = align;
L
Linus Torvalds 已提交
2396 2397
#if DEBUG

2398 2399 2400 2401
	/*
	 * Both debugging options require word-alignment which is calculated
	 * into align above.
	 */
L
Linus Torvalds 已提交
2402 2403
	if (flags & SLAB_RED_ZONE) {
		/* add space for red zone words */
2404 2405
		cachep->obj_offset += sizeof(unsigned long long);
		size += 2 * sizeof(unsigned long long);
L
Linus Torvalds 已提交
2406 2407
	}
	if (flags & SLAB_STORE_USER) {
2408
		/* user store requires one word storage behind the end of
D
David Woodhouse 已提交
2409 2410
		 * 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
2411
		 */
D
David Woodhouse 已提交
2412 2413 2414 2415
		if (flags & SLAB_RED_ZONE)
			size += REDZONE_ALIGN;
		else
			size += BYTES_PER_WORD;
L
Linus Torvalds 已提交
2416 2417
	}
#if FORCED_DEBUG && defined(CONFIG_DEBUG_PAGEALLOC)
P
Pekka Enberg 已提交
2418
	if (size >= malloc_sizes[INDEX_L3 + 1].cs_size
2419
	    && cachep->object_size > cache_line_size() && ALIGN(size, align) < PAGE_SIZE) {
C
Carsten Otte 已提交
2420
		cachep->obj_offset += PAGE_SIZE - ALIGN(size, align);
L
Linus Torvalds 已提交
2421 2422 2423 2424 2425
		size = PAGE_SIZE;
	}
#endif
#endif

2426 2427 2428
	/*
	 * Determine if the slab management is 'on' or 'off' slab.
	 * (bootstrapping cannot cope with offslab caches so don't do
2429 2430
	 * it too early on. Always use on-slab management when
	 * SLAB_NOLEAKTRACE to avoid recursive calls into kmemleak)
2431
	 */
2432 2433
	if ((size >= (PAGE_SIZE >> 3)) && !slab_early_init &&
	    !(flags & SLAB_NOLEAKTRACE))
L
Linus Torvalds 已提交
2434 2435 2436 2437 2438 2439 2440 2441
		/*
		 * 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);

2442
	left_over = calculate_slab_order(cachep, size, align, flags);
L
Linus Torvalds 已提交
2443 2444

	if (!cachep->num) {
2445 2446
		printk(KERN_ERR
		       "kmem_cache_create: couldn't create cache %s.\n", name);
L
Linus Torvalds 已提交
2447 2448
		kmem_cache_free(&cache_cache, cachep);
		cachep = NULL;
2449
		goto oops;
L
Linus Torvalds 已提交
2450
	}
P
Pekka Enberg 已提交
2451 2452
	slab_size = ALIGN(cachep->num * sizeof(kmem_bufctl_t)
			  + sizeof(struct slab), align);
L
Linus Torvalds 已提交
2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464

	/*
	 * 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
2465 2466
		slab_size =
		    cachep->num * sizeof(kmem_bufctl_t) + sizeof(struct slab);
2467 2468 2469 2470 2471 2472 2473 2474 2475

#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
2476 2477 2478 2479 2480 2481
	}

	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
2482
	cachep->colour = left_over / cachep->colour_off;
L
Linus Torvalds 已提交
2483 2484
	cachep->slab_size = slab_size;
	cachep->flags = flags;
2485
	cachep->allocflags = 0;
2486
	if (CONFIG_ZONE_DMA_FLAG && (flags & SLAB_CACHE_DMA))
2487
		cachep->allocflags |= GFP_DMA;
2488
	cachep->size = size;
2489
	cachep->reciprocal_buffer_size = reciprocal_value(size);
L
Linus Torvalds 已提交
2490

2491
	if (flags & CFLGS_OFF_SLAB) {
2492
		cachep->slabp_cache = kmem_find_general_cachep(slab_size, 0u);
2493 2494 2495 2496 2497 2498 2499
		/*
		 * 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.
		 */
2500
		BUG_ON(ZERO_OR_NULL_PTR(cachep->slabp_cache));
2501
	}
L
Linus Torvalds 已提交
2502 2503 2504
	cachep->ctor = ctor;
	cachep->name = name;

2505
	if (setup_cpu_cache(cachep, gfp)) {
2506 2507 2508 2509
		__kmem_cache_destroy(cachep);
		cachep = NULL;
		goto oops;
	}
L
Linus Torvalds 已提交
2510

2511 2512 2513 2514 2515 2516 2517 2518 2519 2520
	if (flags & SLAB_DEBUG_OBJECTS) {
		/*
		 * Would deadlock through slab_destroy()->call_rcu()->
		 * debug_object_activate()->kmem_cache_alloc().
		 */
		WARN_ON_ONCE(flags & SLAB_DESTROY_BY_RCU);

		slab_set_debugobj_lock_classes(cachep);
	}

L
Linus Torvalds 已提交
2521
	/* cache setup completed, link it into the list */
2522
	list_add(&cachep->list, &cache_chain);
A
Andrew Morton 已提交
2523
oops:
L
Linus Torvalds 已提交
2524 2525
	if (!cachep && (flags & SLAB_PANIC))
		panic("kmem_cache_create(): failed to create slab `%s'\n",
P
Pekka Enberg 已提交
2526
		      name);
2527 2528 2529 2530
	if (slab_is_available()) {
		mutex_unlock(&cache_chain_mutex);
		put_online_cpus();
	}
L
Linus Torvalds 已提交
2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545
	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());
}

2546
static void check_spinlock_acquired(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2547 2548 2549
{
#ifdef CONFIG_SMP
	check_irq_off();
2550
	assert_spin_locked(&cachep->nodelists[numa_mem_id()]->list_lock);
L
Linus Torvalds 已提交
2551 2552
#endif
}
2553

2554
static void check_spinlock_acquired_node(struct kmem_cache *cachep, int node)
2555 2556 2557 2558 2559 2560 2561
{
#ifdef CONFIG_SMP
	check_irq_off();
	assert_spin_locked(&cachep->nodelists[node]->list_lock);
#endif
}

L
Linus Torvalds 已提交
2562 2563 2564 2565
#else
#define check_irq_off()	do { } while(0)
#define check_irq_on()	do { } while(0)
#define check_spinlock_acquired(x) do { } while(0)
2566
#define check_spinlock_acquired_node(x, y) do { } while(0)
L
Linus Torvalds 已提交
2567 2568
#endif

2569 2570 2571 2572
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
2573 2574
static void do_drain(void *arg)
{
A
Andrew Morton 已提交
2575
	struct kmem_cache *cachep = arg;
L
Linus Torvalds 已提交
2576
	struct array_cache *ac;
2577
	int node = numa_mem_id();
L
Linus Torvalds 已提交
2578 2579

	check_irq_off();
2580
	ac = cpu_cache_get(cachep);
2581 2582 2583
	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
2584 2585 2586
	ac->avail = 0;
}

2587
static void drain_cpu_caches(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2588
{
2589 2590 2591
	struct kmem_list3 *l3;
	int node;

2592
	on_each_cpu(do_drain, cachep, 1);
L
Linus Torvalds 已提交
2593
	check_irq_on();
P
Pekka Enberg 已提交
2594
	for_each_online_node(node) {
2595
		l3 = cachep->nodelists[node];
2596 2597 2598 2599 2600 2601 2602
		if (l3 && l3->alien)
			drain_alien_cache(cachep, l3->alien);
	}

	for_each_online_node(node) {
		l3 = cachep->nodelists[node];
		if (l3)
2603
			drain_array(cachep, l3, l3->shared, 1, node);
2604
	}
L
Linus Torvalds 已提交
2605 2606
}

2607 2608 2609 2610 2611 2612 2613 2614
/*
 * 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
2615
{
2616 2617
	struct list_head *p;
	int nr_freed;
L
Linus Torvalds 已提交
2618 2619
	struct slab *slabp;

2620 2621
	nr_freed = 0;
	while (nr_freed < tofree && !list_empty(&l3->slabs_free)) {
L
Linus Torvalds 已提交
2622

2623
		spin_lock_irq(&l3->list_lock);
2624
		p = l3->slabs_free.prev;
2625 2626 2627 2628
		if (p == &l3->slabs_free) {
			spin_unlock_irq(&l3->list_lock);
			goto out;
		}
L
Linus Torvalds 已提交
2629

2630
		slabp = list_entry(p, struct slab, list);
L
Linus Torvalds 已提交
2631
#if DEBUG
2632
		BUG_ON(slabp->inuse);
L
Linus Torvalds 已提交
2633 2634
#endif
		list_del(&slabp->list);
2635 2636 2637 2638 2639
		/*
		 * Safe to drop the lock. The slab is no longer linked
		 * to the cache.
		 */
		l3->free_objects -= cache->num;
2640
		spin_unlock_irq(&l3->list_lock);
2641 2642
		slab_destroy(cache, slabp);
		nr_freed++;
L
Linus Torvalds 已提交
2643
	}
2644 2645
out:
	return nr_freed;
L
Linus Torvalds 已提交
2646 2647
}

2648
/* Called with cache_chain_mutex held to protect against cpu hotplug */
2649
static int __cache_shrink(struct kmem_cache *cachep)
2650 2651 2652 2653 2654 2655 2656 2657 2658
{
	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];
2659 2660 2661 2662 2663 2664 2665
		if (!l3)
			continue;

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

		ret += !list_empty(&l3->slabs_full) ||
			!list_empty(&l3->slabs_partial);
2666 2667 2668 2669
	}
	return (ret ? 1 : 0);
}

L
Linus Torvalds 已提交
2670 2671 2672 2673 2674 2675 2676
/**
 * 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.
 */
2677
int kmem_cache_shrink(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2678
{
2679
	int ret;
2680
	BUG_ON(!cachep || in_interrupt());
L
Linus Torvalds 已提交
2681

2682
	get_online_cpus();
2683 2684 2685
	mutex_lock(&cache_chain_mutex);
	ret = __cache_shrink(cachep);
	mutex_unlock(&cache_chain_mutex);
2686
	put_online_cpus();
2687
	return ret;
L
Linus Torvalds 已提交
2688 2689 2690 2691 2692 2693 2694
}
EXPORT_SYMBOL(kmem_cache_shrink);

/**
 * kmem_cache_destroy - delete a cache
 * @cachep: the cache to destroy
 *
2695
 * Remove a &struct kmem_cache object from the slab cache.
L
Linus Torvalds 已提交
2696 2697 2698 2699 2700 2701 2702 2703
 *
 * 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.
 *
L
Lucas De Marchi 已提交
2704
 * The caller must guarantee that no one will allocate memory from the cache
L
Linus Torvalds 已提交
2705 2706
 * during the kmem_cache_destroy().
 */
2707
void kmem_cache_destroy(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2708
{
2709
	BUG_ON(!cachep || in_interrupt());
L
Linus Torvalds 已提交
2710 2711

	/* Find the cache in the chain of caches. */
2712
	get_online_cpus();
I
Ingo Molnar 已提交
2713
	mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2714 2715 2716
	/*
	 * the chain is never empty, cache_cache is never destroyed
	 */
2717
	list_del(&cachep->list);
L
Linus Torvalds 已提交
2718 2719
	if (__cache_shrink(cachep)) {
		slab_error(cachep, "Can't free all objects");
2720
		list_add(&cachep->list, &cache_chain);
I
Ingo Molnar 已提交
2721
		mutex_unlock(&cache_chain_mutex);
2722
		put_online_cpus();
2723
		return;
L
Linus Torvalds 已提交
2724 2725 2726
	}

	if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU))
2727
		rcu_barrier();
L
Linus Torvalds 已提交
2728

2729
	__kmem_cache_destroy(cachep);
2730
	mutex_unlock(&cache_chain_mutex);
2731
	put_online_cpus();
L
Linus Torvalds 已提交
2732 2733 2734
}
EXPORT_SYMBOL(kmem_cache_destroy);

2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745
/*
 * 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.
 */
2746
static struct slab *alloc_slabmgmt(struct kmem_cache *cachep, void *objp,
2747 2748
				   int colour_off, gfp_t local_flags,
				   int nodeid)
L
Linus Torvalds 已提交
2749 2750
{
	struct slab *slabp;
P
Pekka Enberg 已提交
2751

L
Linus Torvalds 已提交
2752 2753
	if (OFF_SLAB(cachep)) {
		/* Slab management obj is off-slab. */
2754
		slabp = kmem_cache_alloc_node(cachep->slabp_cache,
2755
					      local_flags, nodeid);
2756 2757 2758 2759 2760 2761
		/*
		 * 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.
		 */
2762 2763
		kmemleak_scan_area(&slabp->list, sizeof(struct list_head),
				   local_flags);
L
Linus Torvalds 已提交
2764 2765 2766
		if (!slabp)
			return NULL;
	} else {
P
Pekka Enberg 已提交
2767
		slabp = objp + colour_off;
L
Linus Torvalds 已提交
2768 2769 2770 2771
		colour_off += cachep->slab_size;
	}
	slabp->inuse = 0;
	slabp->colouroff = colour_off;
P
Pekka Enberg 已提交
2772
	slabp->s_mem = objp + colour_off;
2773
	slabp->nodeid = nodeid;
2774
	slabp->free = 0;
L
Linus Torvalds 已提交
2775 2776 2777 2778 2779
	return slabp;
}

static inline kmem_bufctl_t *slab_bufctl(struct slab *slabp)
{
P
Pekka Enberg 已提交
2780
	return (kmem_bufctl_t *) (slabp + 1);
L
Linus Torvalds 已提交
2781 2782
}

2783
static void cache_init_objs(struct kmem_cache *cachep,
C
Christoph Lameter 已提交
2784
			    struct slab *slabp)
L
Linus Torvalds 已提交
2785 2786 2787 2788
{
	int i;

	for (i = 0; i < cachep->num; i++) {
2789
		void *objp = index_to_obj(cachep, slabp, i);
L
Linus Torvalds 已提交
2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801
#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
2802 2803 2804
		 * 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
2805 2806
		 */
		if (cachep->ctor && !(cachep->flags & SLAB_POISON))
2807
			cachep->ctor(objp + obj_offset(cachep));
L
Linus Torvalds 已提交
2808 2809 2810 2811

		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
2812
					   " end of an object");
L
Linus Torvalds 已提交
2813 2814
			if (*dbg_redzone1(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "constructor overwrote the"
P
Pekka Enberg 已提交
2815
					   " start of an object");
L
Linus Torvalds 已提交
2816
		}
2817
		if ((cachep->size % PAGE_SIZE) == 0 &&
A
Andrew Morton 已提交
2818
			    OFF_SLAB(cachep) && cachep->flags & SLAB_POISON)
P
Pekka Enberg 已提交
2819
			kernel_map_pages(virt_to_page(objp),
2820
					 cachep->size / PAGE_SIZE, 0);
L
Linus Torvalds 已提交
2821 2822
#else
		if (cachep->ctor)
2823
			cachep->ctor(objp);
L
Linus Torvalds 已提交
2824
#endif
P
Pekka Enberg 已提交
2825
		slab_bufctl(slabp)[i] = i + 1;
L
Linus Torvalds 已提交
2826
	}
P
Pekka Enberg 已提交
2827
	slab_bufctl(slabp)[i - 1] = BUFCTL_END;
L
Linus Torvalds 已提交
2828 2829
}

2830
static void kmem_flagcheck(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2831
{
2832 2833
	if (CONFIG_ZONE_DMA_FLAG) {
		if (flags & GFP_DMA)
2834
			BUG_ON(!(cachep->allocflags & GFP_DMA));
2835
		else
2836
			BUG_ON(cachep->allocflags & GFP_DMA);
2837
	}
L
Linus Torvalds 已提交
2838 2839
}

A
Andrew Morton 已提交
2840 2841
static void *slab_get_obj(struct kmem_cache *cachep, struct slab *slabp,
				int nodeid)
2842
{
2843
	void *objp = index_to_obj(cachep, slabp, slabp->free);
2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856
	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
2857 2858
static void slab_put_obj(struct kmem_cache *cachep, struct slab *slabp,
				void *objp, int nodeid)
2859
{
2860
	unsigned int objnr = obj_to_index(cachep, slabp, objp);
2861 2862 2863 2864 2865

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

2866
	if (slab_bufctl(slabp)[objnr] + 1 <= SLAB_LIMIT + 1) {
2867
		printk(KERN_ERR "slab: double free detected in cache "
A
Andrew Morton 已提交
2868
				"'%s', objp %p\n", cachep->name, objp);
2869 2870 2871 2872 2873 2874 2875 2876
		BUG();
	}
#endif
	slab_bufctl(slabp)[objnr] = slabp->free;
	slabp->free = objnr;
	slabp->inuse--;
}

2877 2878 2879
/*
 * 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
2880
 * virtual address for kfree, ksize, and slab debugging.
2881 2882 2883
 */
static void slab_map_pages(struct kmem_cache *cache, struct slab *slab,
			   void *addr)
L
Linus Torvalds 已提交
2884
{
2885
	int nr_pages;
L
Linus Torvalds 已提交
2886 2887
	struct page *page;

2888
	page = virt_to_page(addr);
2889

2890
	nr_pages = 1;
2891
	if (likely(!PageCompound(page)))
2892 2893
		nr_pages <<= cache->gfporder;

L
Linus Torvalds 已提交
2894
	do {
C
Christoph Lameter 已提交
2895 2896
		page->slab_cache = cache;
		page->slab_page = slab;
L
Linus Torvalds 已提交
2897
		page++;
2898
	} while (--nr_pages);
L
Linus Torvalds 已提交
2899 2900 2901 2902 2903 2904
}

/*
 * 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.
 */
2905 2906
static int cache_grow(struct kmem_cache *cachep,
		gfp_t flags, int nodeid, void *objp)
L
Linus Torvalds 已提交
2907
{
P
Pekka Enberg 已提交
2908 2909 2910
	struct slab *slabp;
	size_t offset;
	gfp_t local_flags;
2911
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
2912

A
Andrew Morton 已提交
2913 2914 2915
	/*
	 * 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
2916
	 */
C
Christoph Lameter 已提交
2917 2918
	BUG_ON(flags & GFP_SLAB_BUG_MASK);
	local_flags = flags & (GFP_CONSTRAINT_MASK|GFP_RECLAIM_MASK);
L
Linus Torvalds 已提交
2919

2920
	/* Take the l3 list lock to change the colour_next on this node */
L
Linus Torvalds 已提交
2921
	check_irq_off();
2922 2923
	l3 = cachep->nodelists[nodeid];
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2924 2925

	/* Get colour for the slab, and cal the next value. */
2926 2927 2928 2929 2930
	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
2931

2932
	offset *= cachep->colour_off;
L
Linus Torvalds 已提交
2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944

	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
2945 2946 2947
	/*
	 * Get mem for the objs.  Attempt to allocate a physical page from
	 * 'nodeid'.
2948
	 */
2949
	if (!objp)
2950
		objp = kmem_getpages(cachep, local_flags, nodeid);
A
Andrew Morton 已提交
2951
	if (!objp)
L
Linus Torvalds 已提交
2952 2953 2954
		goto failed;

	/* Get slab management. */
2955
	slabp = alloc_slabmgmt(cachep, objp, offset,
C
Christoph Lameter 已提交
2956
			local_flags & ~GFP_CONSTRAINT_MASK, nodeid);
A
Andrew Morton 已提交
2957
	if (!slabp)
L
Linus Torvalds 已提交
2958 2959
		goto opps1;

2960
	slab_map_pages(cachep, slabp, objp);
L
Linus Torvalds 已提交
2961

C
Christoph Lameter 已提交
2962
	cache_init_objs(cachep, slabp);
L
Linus Torvalds 已提交
2963 2964 2965 2966

	if (local_flags & __GFP_WAIT)
		local_irq_disable();
	check_irq_off();
2967
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2968 2969

	/* Make slab active. */
2970
	list_add_tail(&slabp->list, &(l3->slabs_free));
L
Linus Torvalds 已提交
2971
	STATS_INC_GROWN(cachep);
2972 2973
	l3->free_objects += cachep->num;
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2974
	return 1;
A
Andrew Morton 已提交
2975
opps1:
L
Linus Torvalds 已提交
2976
	kmem_freepages(cachep, objp);
A
Andrew Morton 已提交
2977
failed:
L
Linus Torvalds 已提交
2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993
	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
2994 2995
		       (unsigned long)objp);
		BUG();
L
Linus Torvalds 已提交
2996 2997 2998
	}
}

2999 3000
static inline void verify_redzone_free(struct kmem_cache *cache, void *obj)
{
3001
	unsigned long long redzone1, redzone2;
3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016

	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");

3017
	printk(KERN_ERR "%p: redzone 1:0x%llx, redzone 2:0x%llx.\n",
3018 3019 3020
			obj, redzone1, redzone2);
}

3021
static void *cache_free_debugcheck(struct kmem_cache *cachep, void *objp,
P
Pekka Enberg 已提交
3022
				   void *caller)
L
Linus Torvalds 已提交
3023 3024 3025 3026 3027
{
	struct page *page;
	unsigned int objnr;
	struct slab *slabp;

3028 3029
	BUG_ON(virt_to_cache(objp) != cachep);

3030
	objp -= obj_offset(cachep);
L
Linus Torvalds 已提交
3031
	kfree_debugcheck(objp);
3032
	page = virt_to_head_page(objp);
L
Linus Torvalds 已提交
3033

C
Christoph Lameter 已提交
3034
	slabp = page->slab_page;
L
Linus Torvalds 已提交
3035 3036

	if (cachep->flags & SLAB_RED_ZONE) {
3037
		verify_redzone_free(cachep, objp);
L
Linus Torvalds 已提交
3038 3039 3040 3041 3042 3043
		*dbg_redzone1(cachep, objp) = RED_INACTIVE;
		*dbg_redzone2(cachep, objp) = RED_INACTIVE;
	}
	if (cachep->flags & SLAB_STORE_USER)
		*dbg_userword(cachep, objp) = caller;

3044
	objnr = obj_to_index(cachep, slabp, objp);
L
Linus Torvalds 已提交
3045 3046

	BUG_ON(objnr >= cachep->num);
3047
	BUG_ON(objp != index_to_obj(cachep, slabp, objnr));
L
Linus Torvalds 已提交
3048

3049 3050 3051
#ifdef CONFIG_DEBUG_SLAB_LEAK
	slab_bufctl(slabp)[objnr] = BUFCTL_FREE;
#endif
L
Linus Torvalds 已提交
3052 3053
	if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
3054
		if ((cachep->size % PAGE_SIZE)==0 && OFF_SLAB(cachep)) {
L
Linus Torvalds 已提交
3055
			store_stackinfo(cachep, objp, (unsigned long)caller);
P
Pekka Enberg 已提交
3056
			kernel_map_pages(virt_to_page(objp),
3057
					 cachep->size / PAGE_SIZE, 0);
L
Linus Torvalds 已提交
3058 3059 3060 3061 3062 3063 3064 3065 3066 3067
		} else {
			poison_obj(cachep, objp, POISON_FREE);
		}
#else
		poison_obj(cachep, objp, POISON_FREE);
#endif
	}
	return objp;
}

3068
static void check_slabp(struct kmem_cache *cachep, struct slab *slabp)
L
Linus Torvalds 已提交
3069 3070 3071
{
	kmem_bufctl_t i;
	int entries = 0;
P
Pekka Enberg 已提交
3072

L
Linus Torvalds 已提交
3073 3074 3075 3076 3077 3078 3079
	/* 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
3080 3081
bad:
		printk(KERN_ERR "slab: Internal list corruption detected in "
3082 3083 3084
			"cache '%s'(%d), slabp %p(%d). Tainted(%s). Hexdump:\n",
			cachep->name, cachep->num, slabp, slabp->inuse,
			print_tainted());
3085 3086 3087
		print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 16, 1, slabp,
			sizeof(*slabp) + cachep->num * sizeof(kmem_bufctl_t),
			1);
L
Linus Torvalds 已提交
3088 3089 3090 3091 3092 3093 3094 3095 3096
		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

3097
static void *cache_alloc_refill(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
3098 3099 3100 3101
{
	int batchcount;
	struct kmem_list3 *l3;
	struct array_cache *ac;
P
Pekka Enberg 已提交
3102 3103
	int node;

3104
retry:
L
Linus Torvalds 已提交
3105
	check_irq_off();
3106
	node = numa_mem_id();
3107
	ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
3108 3109
	batchcount = ac->batchcount;
	if (!ac->touched && batchcount > BATCHREFILL_LIMIT) {
A
Andrew Morton 已提交
3110 3111 3112 3113
		/*
		 * 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
3114 3115 3116
		 */
		batchcount = BATCHREFILL_LIMIT;
	}
P
Pekka Enberg 已提交
3117
	l3 = cachep->nodelists[node];
3118 3119 3120

	BUG_ON(ac->avail > 0 || !l3);
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
3121

3122
	/* See if we can refill from the shared array */
3123 3124
	if (l3->shared && transfer_objects(ac, l3->shared, batchcount)) {
		l3->shared->touched = 1;
3125
		goto alloc_done;
3126
	}
3127

L
Linus Torvalds 已提交
3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142
	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);
3143 3144 3145 3146 3147 3148

		/*
		 * The slab was either on partial or free list so
		 * there must be at least one object available for
		 * allocation.
		 */
3149
		BUG_ON(slabp->inuse >= cachep->num);
3150

L
Linus Torvalds 已提交
3151 3152 3153 3154 3155
		while (slabp->inuse < cachep->num && batchcount--) {
			STATS_INC_ALLOCED(cachep);
			STATS_INC_ACTIVE(cachep);
			STATS_SET_HIGH(cachep);

3156
			ac->entry[ac->avail++] = slab_get_obj(cachep, slabp,
P
Pekka Enberg 已提交
3157
							    node);
L
Linus Torvalds 已提交
3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168
		}
		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
3169
must_grow:
L
Linus Torvalds 已提交
3170
	l3->free_objects -= ac->avail;
A
Andrew Morton 已提交
3171
alloc_done:
3172
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
3173 3174 3175

	if (unlikely(!ac->avail)) {
		int x;
3176
		x = cache_grow(cachep, flags | GFP_THISNODE, node, NULL);
3177

A
Andrew Morton 已提交
3178
		/* cache_grow can reenable interrupts, then ac could change. */
3179
		ac = cpu_cache_get(cachep);
A
Andrew Morton 已提交
3180
		if (!x && ac->avail == 0)	/* no objects in sight? abort */
L
Linus Torvalds 已提交
3181 3182
			return NULL;

A
Andrew Morton 已提交
3183
		if (!ac->avail)		/* objects refilled by interrupt? */
L
Linus Torvalds 已提交
3184 3185 3186
			goto retry;
	}
	ac->touched = 1;
3187
	return ac->entry[--ac->avail];
L
Linus Torvalds 已提交
3188 3189
}

A
Andrew Morton 已提交
3190 3191
static inline void cache_alloc_debugcheck_before(struct kmem_cache *cachep,
						gfp_t flags)
L
Linus Torvalds 已提交
3192 3193 3194 3195 3196 3197 3198 3199
{
	might_sleep_if(flags & __GFP_WAIT);
#if DEBUG
	kmem_flagcheck(cachep, flags);
#endif
}

#if DEBUG
A
Andrew Morton 已提交
3200 3201
static void *cache_alloc_debugcheck_after(struct kmem_cache *cachep,
				gfp_t flags, void *objp, void *caller)
L
Linus Torvalds 已提交
3202
{
P
Pekka Enberg 已提交
3203
	if (!objp)
L
Linus Torvalds 已提交
3204
		return objp;
P
Pekka Enberg 已提交
3205
	if (cachep->flags & SLAB_POISON) {
L
Linus Torvalds 已提交
3206
#ifdef CONFIG_DEBUG_PAGEALLOC
3207
		if ((cachep->size % PAGE_SIZE) == 0 && OFF_SLAB(cachep))
P
Pekka Enberg 已提交
3208
			kernel_map_pages(virt_to_page(objp),
3209
					 cachep->size / PAGE_SIZE, 1);
L
Linus Torvalds 已提交
3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220
		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
3221 3222 3223 3224
		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
3225
			printk(KERN_ERR
3226
				"%p: redzone 1:0x%llx, redzone 2:0x%llx\n",
A
Andrew Morton 已提交
3227 3228
				objp, *dbg_redzone1(cachep, objp),
				*dbg_redzone2(cachep, objp));
L
Linus Torvalds 已提交
3229 3230 3231 3232
		}
		*dbg_redzone1(cachep, objp) = RED_ACTIVE;
		*dbg_redzone2(cachep, objp) = RED_ACTIVE;
	}
3233 3234 3235 3236 3237
#ifdef CONFIG_DEBUG_SLAB_LEAK
	{
		struct slab *slabp;
		unsigned objnr;

C
Christoph Lameter 已提交
3238
		slabp = virt_to_head_page(objp)->slab_page;
3239
		objnr = (unsigned)(objp - slabp->s_mem) / cachep->size;
3240 3241 3242
		slab_bufctl(slabp)[objnr] = BUFCTL_ACTIVE;
	}
#endif
3243
	objp += obj_offset(cachep);
3244
	if (cachep->ctor && cachep->flags & SLAB_POISON)
3245
		cachep->ctor(objp);
T
Tetsuo Handa 已提交
3246 3247
	if (ARCH_SLAB_MINALIGN &&
	    ((unsigned long)objp & (ARCH_SLAB_MINALIGN-1))) {
3248
		printk(KERN_ERR "0x%p: not aligned to ARCH_SLAB_MINALIGN=%d\n",
H
Hugh Dickins 已提交
3249
		       objp, (int)ARCH_SLAB_MINALIGN);
3250
	}
L
Linus Torvalds 已提交
3251 3252 3253 3254 3255 3256
	return objp;
}
#else
#define cache_alloc_debugcheck_after(a,b,objp,d) (objp)
#endif

A
Akinobu Mita 已提交
3257
static bool slab_should_failslab(struct kmem_cache *cachep, gfp_t flags)
3258 3259
{
	if (cachep == &cache_cache)
A
Akinobu Mita 已提交
3260
		return false;
3261

3262
	return should_failslab(cachep->object_size, flags, cachep->flags);
3263 3264
}

3265
static inline void *____cache_alloc(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
3266
{
P
Pekka Enberg 已提交
3267
	void *objp;
L
Linus Torvalds 已提交
3268 3269
	struct array_cache *ac;

3270
	check_irq_off();
3271

3272
	ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
3273 3274 3275
	if (likely(ac->avail)) {
		STATS_INC_ALLOCHIT(cachep);
		ac->touched = 1;
3276
		objp = ac->entry[--ac->avail];
L
Linus Torvalds 已提交
3277 3278 3279
	} else {
		STATS_INC_ALLOCMISS(cachep);
		objp = cache_alloc_refill(cachep, flags);
3280 3281 3282 3283 3284
		/*
		 * 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
3285
	}
3286 3287 3288 3289 3290
	/*
	 * 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.
	 */
3291 3292
	if (objp)
		kmemleak_erase(&ac->entry[ac->avail]);
3293 3294 3295
	return objp;
}

3296
#ifdef CONFIG_NUMA
3297
/*
3298
 * Try allocating on another node if PF_SPREAD_SLAB|PF_MEMPOLICY.
3299 3300 3301 3302 3303 3304 3305 3306
 *
 * 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;

3307
	if (in_interrupt() || (flags & __GFP_THISNODE))
3308
		return NULL;
3309
	nid_alloc = nid_here = numa_mem_id();
3310
	if (cpuset_do_slab_mem_spread() && (cachep->flags & SLAB_MEM_SPREAD))
3311
		nid_alloc = cpuset_slab_spread_node();
3312
	else if (current->mempolicy)
3313
		nid_alloc = slab_node();
3314
	if (nid_alloc != nid_here)
3315
		return ____cache_alloc_node(cachep, flags, nid_alloc);
3316 3317 3318
	return NULL;
}

3319 3320
/*
 * Fallback function if there was no memory available and no objects on a
3321 3322 3323 3324 3325
 * 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.
3326
 */
3327
static void *fallback_alloc(struct kmem_cache *cache, gfp_t flags)
3328
{
3329 3330
	struct zonelist *zonelist;
	gfp_t local_flags;
3331
	struct zoneref *z;
3332 3333
	struct zone *zone;
	enum zone_type high_zoneidx = gfp_zone(flags);
3334
	void *obj = NULL;
3335
	int nid;
3336
	unsigned int cpuset_mems_cookie;
3337 3338 3339 3340

	if (flags & __GFP_THISNODE)
		return NULL;

C
Christoph Lameter 已提交
3341
	local_flags = flags & (GFP_CONSTRAINT_MASK|GFP_RECLAIM_MASK);
3342

3343 3344
retry_cpuset:
	cpuset_mems_cookie = get_mems_allowed();
3345
	zonelist = node_zonelist(slab_node(), flags);
3346

3347 3348 3349 3350 3351
retry:
	/*
	 * Look through allowed nodes for objects available
	 * from existing per node queues.
	 */
3352 3353
	for_each_zone_zonelist(zone, z, zonelist, high_zoneidx) {
		nid = zone_to_nid(zone);
3354

3355
		if (cpuset_zone_allowed_hardwall(zone, flags) &&
3356
			cache->nodelists[nid] &&
3357
			cache->nodelists[nid]->free_objects) {
3358 3359
				obj = ____cache_alloc_node(cache,
					flags | GFP_THISNODE, nid);
3360 3361 3362
				if (obj)
					break;
		}
3363 3364
	}

3365
	if (!obj) {
3366 3367 3368 3369 3370 3371
		/*
		 * 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.
		 */
3372 3373 3374
		if (local_flags & __GFP_WAIT)
			local_irq_enable();
		kmem_flagcheck(cache, flags);
3375
		obj = kmem_getpages(cache, local_flags, numa_mem_id());
3376 3377
		if (local_flags & __GFP_WAIT)
			local_irq_disable();
3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393
		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 {
3394
				/* cache_grow already freed obj */
3395 3396 3397
				obj = NULL;
			}
		}
3398
	}
3399 3400 3401

	if (unlikely(!put_mems_allowed(cpuset_mems_cookie) && !obj))
		goto retry_cpuset;
3402 3403 3404
	return obj;
}

3405 3406
/*
 * A interface to enable slab creation on nodeid
L
Linus Torvalds 已提交
3407
 */
3408
static void *____cache_alloc_node(struct kmem_cache *cachep, gfp_t flags,
A
Andrew Morton 已提交
3409
				int nodeid)
3410 3411
{
	struct list_head *entry;
P
Pekka Enberg 已提交
3412 3413 3414 3415 3416 3417 3418 3419
	struct slab *slabp;
	struct kmem_list3 *l3;
	void *obj;
	int x;

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

A
Andrew Morton 已提交
3420
retry:
3421
	check_irq_off();
P
Pekka Enberg 已提交
3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440
	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);

3441
	obj = slab_get_obj(cachep, slabp, nodeid);
P
Pekka Enberg 已提交
3442 3443 3444 3445 3446
	check_slabp(cachep, slabp);
	l3->free_objects--;
	/* move slabp to correct slabp list: */
	list_del(&slabp->list);

A
Andrew Morton 已提交
3447
	if (slabp->free == BUFCTL_END)
P
Pekka Enberg 已提交
3448
		list_add(&slabp->list, &l3->slabs_full);
A
Andrew Morton 已提交
3449
	else
P
Pekka Enberg 已提交
3450
		list_add(&slabp->list, &l3->slabs_partial);
3451

P
Pekka Enberg 已提交
3452 3453
	spin_unlock(&l3->list_lock);
	goto done;
3454

A
Andrew Morton 已提交
3455
must_grow:
P
Pekka Enberg 已提交
3456
	spin_unlock(&l3->list_lock);
3457
	x = cache_grow(cachep, flags | GFP_THISNODE, nodeid, NULL);
3458 3459
	if (x)
		goto retry;
L
Linus Torvalds 已提交
3460

3461
	return fallback_alloc(cachep, flags);
3462

A
Andrew Morton 已提交
3463
done:
P
Pekka Enberg 已提交
3464
	return obj;
3465
}
3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484

/**
 * 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;
3485
	int slab_node = numa_mem_id();
3486

3487
	flags &= gfp_allowed_mask;
3488

3489 3490
	lockdep_trace_alloc(flags);

A
Akinobu Mita 已提交
3491
	if (slab_should_failslab(cachep, flags))
3492 3493
		return NULL;

3494 3495 3496
	cache_alloc_debugcheck_before(cachep, flags);
	local_irq_save(save_flags);

A
Andrew Morton 已提交
3497
	if (nodeid == NUMA_NO_NODE)
3498
		nodeid = slab_node;
3499 3500 3501 3502 3503 3504 3505

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

3506
	if (nodeid == slab_node) {
3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521
		/*
		 * 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);
3522
	kmemleak_alloc_recursive(ptr, cachep->object_size, 1, cachep->flags,
3523
				 flags);
3524

P
Pekka Enberg 已提交
3525
	if (likely(ptr))
3526
		kmemcheck_slab_alloc(cachep, flags, ptr, cachep->object_size);
P
Pekka Enberg 已提交
3527

3528
	if (unlikely((flags & __GFP_ZERO) && ptr))
3529
		memset(ptr, 0, cachep->object_size);
3530

3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549
	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
	 */
3550 3551
	if (!objp)
		objp = ____cache_alloc_node(cache, flags, numa_mem_id());
3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571

  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;

3572
	flags &= gfp_allowed_mask;
3573

3574 3575
	lockdep_trace_alloc(flags);

A
Akinobu Mita 已提交
3576
	if (slab_should_failslab(cachep, flags))
3577 3578
		return NULL;

3579 3580 3581 3582 3583
	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);
3584
	kmemleak_alloc_recursive(objp, cachep->object_size, 1, cachep->flags,
3585
				 flags);
3586 3587
	prefetchw(objp);

P
Pekka Enberg 已提交
3588
	if (likely(objp))
3589
		kmemcheck_slab_alloc(cachep, flags, objp, cachep->object_size);
P
Pekka Enberg 已提交
3590

3591
	if (unlikely((flags & __GFP_ZERO) && objp))
3592
		memset(objp, 0, cachep->object_size);
3593

3594 3595
	return objp;
}
3596 3597 3598 3599

/*
 * Caller needs to acquire correct kmem_list's list_lock
 */
3600
static void free_block(struct kmem_cache *cachep, void **objpp, int nr_objects,
P
Pekka Enberg 已提交
3601
		       int node)
L
Linus Torvalds 已提交
3602 3603
{
	int i;
3604
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
3605 3606 3607 3608 3609

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

3610
		slabp = virt_to_slab(objp);
3611
		l3 = cachep->nodelists[node];
L
Linus Torvalds 已提交
3612
		list_del(&slabp->list);
3613
		check_spinlock_acquired_node(cachep, node);
L
Linus Torvalds 已提交
3614
		check_slabp(cachep, slabp);
3615
		slab_put_obj(cachep, slabp, objp, node);
L
Linus Torvalds 已提交
3616
		STATS_DEC_ACTIVE(cachep);
3617
		l3->free_objects++;
L
Linus Torvalds 已提交
3618 3619 3620 3621
		check_slabp(cachep, slabp);

		/* fixup slab chains */
		if (slabp->inuse == 0) {
3622 3623
			if (l3->free_objects > l3->free_limit) {
				l3->free_objects -= cachep->num;
3624 3625 3626 3627 3628 3629
				/* 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
3630 3631
				slab_destroy(cachep, slabp);
			} else {
3632
				list_add(&slabp->list, &l3->slabs_free);
L
Linus Torvalds 已提交
3633 3634 3635 3636 3637 3638
			}
		} else {
			/* Unconditionally move a slab to the end of the
			 * partial list on free - maximum time for the
			 * other objects to be freed, too.
			 */
3639
			list_add_tail(&slabp->list, &l3->slabs_partial);
L
Linus Torvalds 已提交
3640 3641 3642 3643
		}
	}
}

3644
static void cache_flusharray(struct kmem_cache *cachep, struct array_cache *ac)
L
Linus Torvalds 已提交
3645 3646
{
	int batchcount;
3647
	struct kmem_list3 *l3;
3648
	int node = numa_mem_id();
L
Linus Torvalds 已提交
3649 3650 3651 3652 3653 3654

	batchcount = ac->batchcount;
#if DEBUG
	BUG_ON(!batchcount || batchcount > ac->avail);
#endif
	check_irq_off();
3655
	l3 = cachep->nodelists[node];
3656
	spin_lock(&l3->list_lock);
3657 3658
	if (l3->shared) {
		struct array_cache *shared_array = l3->shared;
P
Pekka Enberg 已提交
3659
		int max = shared_array->limit - shared_array->avail;
L
Linus Torvalds 已提交
3660 3661 3662
		if (max) {
			if (batchcount > max)
				batchcount = max;
3663
			memcpy(&(shared_array->entry[shared_array->avail]),
P
Pekka Enberg 已提交
3664
			       ac->entry, sizeof(void *) * batchcount);
L
Linus Torvalds 已提交
3665 3666 3667 3668 3669
			shared_array->avail += batchcount;
			goto free_done;
		}
	}

3670
	free_block(cachep, ac->entry, batchcount, node);
A
Andrew Morton 已提交
3671
free_done:
L
Linus Torvalds 已提交
3672 3673 3674 3675 3676
#if STATS
	{
		int i = 0;
		struct list_head *p;

3677 3678
		p = l3->slabs_free.next;
		while (p != &(l3->slabs_free)) {
L
Linus Torvalds 已提交
3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689
			struct slab *slabp;

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

			i++;
			p = p->next;
		}
		STATS_SET_FREEABLE(cachep, i);
	}
#endif
3690
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
3691
	ac->avail -= batchcount;
A
Andrew Morton 已提交
3692
	memmove(ac->entry, &(ac->entry[batchcount]), sizeof(void *)*ac->avail);
L
Linus Torvalds 已提交
3693 3694 3695
}

/*
A
Andrew Morton 已提交
3696 3697
 * 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
3698
 */
3699 3700
static inline void __cache_free(struct kmem_cache *cachep, void *objp,
    void *caller)
L
Linus Torvalds 已提交
3701
{
3702
	struct array_cache *ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
3703 3704

	check_irq_off();
3705
	kmemleak_free_recursive(objp, cachep->flags);
3706
	objp = cache_free_debugcheck(cachep, objp, caller);
L
Linus Torvalds 已提交
3707

3708
	kmemcheck_slab_free(cachep, objp, cachep->object_size);
P
Pekka Enberg 已提交
3709

3710 3711 3712 3713 3714 3715 3716
	/*
	 * 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.
	 */
3717
	if (nr_online_nodes > 1 && cache_free_alien(cachep, objp))
3718 3719
		return;

L
Linus Torvalds 已提交
3720 3721 3722 3723 3724 3725
	if (likely(ac->avail < ac->limit)) {
		STATS_INC_FREEHIT(cachep);
	} else {
		STATS_INC_FREEMISS(cachep);
		cache_flusharray(cachep, ac);
	}
Z
Zhao Jin 已提交
3726 3727

	ac->entry[ac->avail++] = objp;
L
Linus Torvalds 已提交
3728 3729 3730 3731 3732 3733 3734 3735 3736 3737
}

/**
 * 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.
 */
3738
void *kmem_cache_alloc(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
3739
{
E
Eduard - Gabriel Munteanu 已提交
3740 3741
	void *ret = __cache_alloc(cachep, flags, __builtin_return_address(0));

3742
	trace_kmem_cache_alloc(_RET_IP_, ret,
3743
			       cachep->object_size, cachep->size, flags);
E
Eduard - Gabriel Munteanu 已提交
3744 3745

	return ret;
L
Linus Torvalds 已提交
3746 3747 3748
}
EXPORT_SYMBOL(kmem_cache_alloc);

3749
#ifdef CONFIG_TRACING
3750 3751
void *
kmem_cache_alloc_trace(size_t size, struct kmem_cache *cachep, gfp_t flags)
E
Eduard - Gabriel Munteanu 已提交
3752
{
3753 3754 3755 3756 3757 3758 3759
	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
3760
}
3761
EXPORT_SYMBOL(kmem_cache_alloc_trace);
E
Eduard - Gabriel Munteanu 已提交
3762 3763
#endif

L
Linus Torvalds 已提交
3764
#ifdef CONFIG_NUMA
3765 3766
void *kmem_cache_alloc_node(struct kmem_cache *cachep, gfp_t flags, int nodeid)
{
E
Eduard - Gabriel Munteanu 已提交
3767 3768 3769
	void *ret = __cache_alloc_node(cachep, flags, nodeid,
				       __builtin_return_address(0));

3770
	trace_kmem_cache_alloc_node(_RET_IP_, ret,
3771
				    cachep->object_size, cachep->size,
3772
				    flags, nodeid);
E
Eduard - Gabriel Munteanu 已提交
3773 3774

	return ret;
3775
}
L
Linus Torvalds 已提交
3776 3777
EXPORT_SYMBOL(kmem_cache_alloc_node);

3778
#ifdef CONFIG_TRACING
3779 3780 3781 3782
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
3783
{
3784 3785 3786
	void *ret;

	ret = __cache_alloc_node(cachep, flags, nodeid,
E
Eduard - Gabriel Munteanu 已提交
3787
				  __builtin_return_address(0));
3788 3789 3790 3791
	trace_kmalloc_node(_RET_IP_, ret,
			   size, slab_buffer_size(cachep),
			   flags, nodeid);
	return ret;
E
Eduard - Gabriel Munteanu 已提交
3792
}
3793
EXPORT_SYMBOL(kmem_cache_alloc_node_trace);
E
Eduard - Gabriel Munteanu 已提交
3794 3795
#endif

3796 3797
static __always_inline void *
__do_kmalloc_node(size_t size, gfp_t flags, int node, void *caller)
3798
{
3799
	struct kmem_cache *cachep;
3800 3801

	cachep = kmem_find_general_cachep(size, flags);
3802 3803
	if (unlikely(ZERO_OR_NULL_PTR(cachep)))
		return cachep;
3804
	return kmem_cache_alloc_node_trace(size, cachep, flags, node);
3805
}
3806

3807
#if defined(CONFIG_DEBUG_SLAB) || defined(CONFIG_TRACING)
3808 3809 3810 3811 3812
void *__kmalloc_node(size_t size, gfp_t flags, int node)
{
	return __do_kmalloc_node(size, flags, node,
			__builtin_return_address(0));
}
3813
EXPORT_SYMBOL(__kmalloc_node);
3814 3815

void *__kmalloc_node_track_caller(size_t size, gfp_t flags,
3816
		int node, unsigned long caller)
3817
{
3818
	return __do_kmalloc_node(size, flags, node, (void *)caller);
3819 3820 3821 3822 3823 3824 3825 3826
}
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);
3827
#endif /* CONFIG_DEBUG_SLAB || CONFIG_TRACING */
3828
#endif /* CONFIG_NUMA */
L
Linus Torvalds 已提交
3829 3830

/**
3831
 * __do_kmalloc - allocate memory
L
Linus Torvalds 已提交
3832
 * @size: how many bytes of memory are required.
3833
 * @flags: the type of memory to allocate (see kmalloc).
3834
 * @caller: function caller for debug tracking of the caller
L
Linus Torvalds 已提交
3835
 */
3836 3837
static __always_inline void *__do_kmalloc(size_t size, gfp_t flags,
					  void *caller)
L
Linus Torvalds 已提交
3838
{
3839
	struct kmem_cache *cachep;
E
Eduard - Gabriel Munteanu 已提交
3840
	void *ret;
L
Linus Torvalds 已提交
3841

3842 3843 3844 3845 3846 3847
	/* 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);
3848 3849
	if (unlikely(ZERO_OR_NULL_PTR(cachep)))
		return cachep;
E
Eduard - Gabriel Munteanu 已提交
3850 3851
	ret = __cache_alloc(cachep, flags, caller);

3852
	trace_kmalloc((unsigned long) caller, ret,
3853
		      size, cachep->size, flags);
E
Eduard - Gabriel Munteanu 已提交
3854 3855

	return ret;
3856 3857 3858
}


3859
#if defined(CONFIG_DEBUG_SLAB) || defined(CONFIG_TRACING)
3860 3861
void *__kmalloc(size_t size, gfp_t flags)
{
3862
	return __do_kmalloc(size, flags, __builtin_return_address(0));
L
Linus Torvalds 已提交
3863 3864 3865
}
EXPORT_SYMBOL(__kmalloc);

3866
void *__kmalloc_track_caller(size_t size, gfp_t flags, unsigned long caller)
3867
{
3868
	return __do_kmalloc(size, flags, (void *)caller);
3869 3870
}
EXPORT_SYMBOL(__kmalloc_track_caller);
3871 3872 3873 3874 3875 3876 3877

#else
void *__kmalloc(size_t size, gfp_t flags)
{
	return __do_kmalloc(size, flags, NULL);
}
EXPORT_SYMBOL(__kmalloc);
3878 3879
#endif

L
Linus Torvalds 已提交
3880 3881 3882 3883 3884 3885 3886 3887
/**
 * 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.
 */
3888
void kmem_cache_free(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
3889 3890 3891 3892
{
	unsigned long flags;

	local_irq_save(flags);
3893
	debug_check_no_locks_freed(objp, cachep->object_size);
3894
	if (!(cachep->flags & SLAB_DEBUG_OBJECTS))
3895
		debug_check_no_obj_freed(objp, cachep->object_size);
3896
	__cache_free(cachep, objp, __builtin_return_address(0));
L
Linus Torvalds 已提交
3897
	local_irq_restore(flags);
E
Eduard - Gabriel Munteanu 已提交
3898

3899
	trace_kmem_cache_free(_RET_IP_, objp);
L
Linus Torvalds 已提交
3900 3901 3902 3903 3904 3905 3906
}
EXPORT_SYMBOL(kmem_cache_free);

/**
 * kfree - free previously allocated memory
 * @objp: pointer returned by kmalloc.
 *
3907 3908
 * If @objp is NULL, no operation is performed.
 *
L
Linus Torvalds 已提交
3909 3910 3911 3912 3913
 * Don't free memory not originally allocated by kmalloc()
 * or you will run into trouble.
 */
void kfree(const void *objp)
{
3914
	struct kmem_cache *c;
L
Linus Torvalds 已提交
3915 3916
	unsigned long flags;

3917 3918
	trace_kfree(_RET_IP_, objp);

3919
	if (unlikely(ZERO_OR_NULL_PTR(objp)))
L
Linus Torvalds 已提交
3920 3921 3922
		return;
	local_irq_save(flags);
	kfree_debugcheck(objp);
3923
	c = virt_to_cache(objp);
3924 3925 3926
	debug_check_no_locks_freed(objp, c->object_size);

	debug_check_no_obj_freed(objp, c->object_size);
3927
	__cache_free(c, (void *)objp, __builtin_return_address(0));
L
Linus Torvalds 已提交
3928 3929 3930 3931
	local_irq_restore(flags);
}
EXPORT_SYMBOL(kfree);

3932
unsigned int kmem_cache_size(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
3933
{
3934
	return cachep->object_size;
L
Linus Torvalds 已提交
3935 3936 3937
}
EXPORT_SYMBOL(kmem_cache_size);

3938
/*
S
Simon Arlott 已提交
3939
 * This initializes kmem_list3 or resizes various caches for all nodes.
3940
 */
3941
static int alloc_kmemlist(struct kmem_cache *cachep, gfp_t gfp)
3942 3943 3944
{
	int node;
	struct kmem_list3 *l3;
3945
	struct array_cache *new_shared;
3946
	struct array_cache **new_alien = NULL;
3947

3948
	for_each_online_node(node) {
3949

3950
                if (use_alien_caches) {
3951
                        new_alien = alloc_alien_cache(node, cachep->limit, gfp);
3952 3953 3954
                        if (!new_alien)
                                goto fail;
                }
3955

3956 3957 3958
		new_shared = NULL;
		if (cachep->shared) {
			new_shared = alloc_arraycache(node,
3959
				cachep->shared*cachep->batchcount,
3960
					0xbaadf00d, gfp);
3961 3962 3963 3964
			if (!new_shared) {
				free_alien_cache(new_alien);
				goto fail;
			}
3965
		}
3966

A
Andrew Morton 已提交
3967 3968
		l3 = cachep->nodelists[node];
		if (l3) {
3969 3970
			struct array_cache *shared = l3->shared;

3971 3972
			spin_lock_irq(&l3->list_lock);

3973
			if (shared)
3974 3975
				free_block(cachep, shared->entry,
						shared->avail, node);
3976

3977 3978
			l3->shared = new_shared;
			if (!l3->alien) {
3979 3980 3981
				l3->alien = new_alien;
				new_alien = NULL;
			}
P
Pekka Enberg 已提交
3982
			l3->free_limit = (1 + nr_cpus_node(node)) *
A
Andrew Morton 已提交
3983
					cachep->batchcount + cachep->num;
3984
			spin_unlock_irq(&l3->list_lock);
3985
			kfree(shared);
3986 3987 3988
			free_alien_cache(new_alien);
			continue;
		}
3989
		l3 = kmalloc_node(sizeof(struct kmem_list3), gfp, node);
3990 3991 3992
		if (!l3) {
			free_alien_cache(new_alien);
			kfree(new_shared);
3993
			goto fail;
3994
		}
3995 3996 3997

		kmem_list3_init(l3);
		l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
A
Andrew Morton 已提交
3998
				((unsigned long)cachep) % REAPTIMEOUT_LIST3;
3999
		l3->shared = new_shared;
4000
		l3->alien = new_alien;
P
Pekka Enberg 已提交
4001
		l3->free_limit = (1 + nr_cpus_node(node)) *
A
Andrew Morton 已提交
4002
					cachep->batchcount + cachep->num;
4003 4004
		cachep->nodelists[node] = l3;
	}
4005
	return 0;
4006

A
Andrew Morton 已提交
4007
fail:
4008
	if (!cachep->list.next) {
4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022
		/* 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--;
		}
	}
4023
	return -ENOMEM;
4024 4025
}

L
Linus Torvalds 已提交
4026
struct ccupdate_struct {
4027
	struct kmem_cache *cachep;
4028
	struct array_cache *new[0];
L
Linus Torvalds 已提交
4029 4030 4031 4032
};

static void do_ccupdate_local(void *info)
{
A
Andrew Morton 已提交
4033
	struct ccupdate_struct *new = info;
L
Linus Torvalds 已提交
4034 4035 4036
	struct array_cache *old;

	check_irq_off();
4037
	old = cpu_cache_get(new->cachep);
4038

L
Linus Torvalds 已提交
4039 4040 4041 4042
	new->cachep->array[smp_processor_id()] = new->new[smp_processor_id()];
	new->new[smp_processor_id()] = old;
}

4043
/* Always called with the cache_chain_mutex held */
A
Andrew Morton 已提交
4044
static int do_tune_cpucache(struct kmem_cache *cachep, int limit,
4045
				int batchcount, int shared, gfp_t gfp)
L
Linus Torvalds 已提交
4046
{
4047
	struct ccupdate_struct *new;
4048
	int i;
L
Linus Torvalds 已提交
4049

4050 4051
	new = kzalloc(sizeof(*new) + nr_cpu_ids * sizeof(struct array_cache *),
		      gfp);
4052 4053 4054
	if (!new)
		return -ENOMEM;

4055
	for_each_online_cpu(i) {
4056
		new->new[i] = alloc_arraycache(cpu_to_mem(i), limit,
4057
						batchcount, gfp);
4058
		if (!new->new[i]) {
P
Pekka Enberg 已提交
4059
			for (i--; i >= 0; i--)
4060 4061
				kfree(new->new[i]);
			kfree(new);
4062
			return -ENOMEM;
L
Linus Torvalds 已提交
4063 4064
		}
	}
4065
	new->cachep = cachep;
L
Linus Torvalds 已提交
4066

4067
	on_each_cpu(do_ccupdate_local, (void *)new, 1);
4068

L
Linus Torvalds 已提交
4069 4070 4071
	check_irq_on();
	cachep->batchcount = batchcount;
	cachep->limit = limit;
4072
	cachep->shared = shared;
L
Linus Torvalds 已提交
4073

4074
	for_each_online_cpu(i) {
4075
		struct array_cache *ccold = new->new[i];
L
Linus Torvalds 已提交
4076 4077
		if (!ccold)
			continue;
4078 4079 4080
		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
4081 4082
		kfree(ccold);
	}
4083
	kfree(new);
4084
	return alloc_kmemlist(cachep, gfp);
L
Linus Torvalds 已提交
4085 4086
}

4087
/* Called with cache_chain_mutex held always */
4088
static int enable_cpucache(struct kmem_cache *cachep, gfp_t gfp)
L
Linus Torvalds 已提交
4089 4090 4091 4092
{
	int err;
	int limit, shared;

A
Andrew Morton 已提交
4093 4094
	/*
	 * The head array serves three purposes:
L
Linus Torvalds 已提交
4095 4096
	 * - create a LIFO ordering, i.e. return objects that are cache-warm
	 * - reduce the number of spinlock operations.
A
Andrew Morton 已提交
4097
	 * - reduce the number of linked list operations on the slab and
L
Linus Torvalds 已提交
4098 4099 4100 4101
	 *   bufctl chains: array operations are cheaper.
	 * The numbers are guessed, we should auto-tune as described by
	 * Bonwick.
	 */
4102
	if (cachep->size > 131072)
L
Linus Torvalds 已提交
4103
		limit = 1;
4104
	else if (cachep->size > PAGE_SIZE)
L
Linus Torvalds 已提交
4105
		limit = 8;
4106
	else if (cachep->size > 1024)
L
Linus Torvalds 已提交
4107
		limit = 24;
4108
	else if (cachep->size > 256)
L
Linus Torvalds 已提交
4109 4110 4111 4112
		limit = 54;
	else
		limit = 120;

A
Andrew Morton 已提交
4113 4114
	/*
	 * CPU bound tasks (e.g. network routing) can exhibit cpu bound
L
Linus Torvalds 已提交
4115 4116 4117 4118 4119 4120 4121 4122
	 * 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;
4123
	if (cachep->size <= PAGE_SIZE && num_possible_cpus() > 1)
L
Linus Torvalds 已提交
4124 4125 4126
		shared = 8;

#if DEBUG
A
Andrew Morton 已提交
4127 4128 4129
	/*
	 * With debugging enabled, large batchcount lead to excessively long
	 * periods with disabled local interrupts. Limit the batchcount
L
Linus Torvalds 已提交
4130 4131 4132 4133
	 */
	if (limit > 32)
		limit = 32;
#endif
4134
	err = do_tune_cpucache(cachep, limit, (limit + 1) / 2, shared, gfp);
L
Linus Torvalds 已提交
4135 4136
	if (err)
		printk(KERN_ERR "enable_cpucache failed for %s, error %d.\n",
P
Pekka Enberg 已提交
4137
		       cachep->name, -err);
4138
	return err;
L
Linus Torvalds 已提交
4139 4140
}

4141 4142
/*
 * Drain an array if it contains any elements taking the l3 lock only if
4143 4144
 * necessary. Note that the l3 listlock also protects the array_cache
 * if drain_array() is used on the shared array.
4145
 */
4146
static void drain_array(struct kmem_cache *cachep, struct kmem_list3 *l3,
4147
			 struct array_cache *ac, int force, int node)
L
Linus Torvalds 已提交
4148 4149 4150
{
	int tofree;

4151 4152
	if (!ac || !ac->avail)
		return;
L
Linus Torvalds 已提交
4153 4154
	if (ac->touched && !force) {
		ac->touched = 0;
4155
	} else {
4156
		spin_lock_irq(&l3->list_lock);
4157 4158 4159 4160 4161 4162 4163 4164 4165
		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);
		}
4166
		spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
4167 4168 4169 4170 4171
	}
}

/**
 * cache_reap - Reclaim memory from caches.
4172
 * @w: work descriptor
L
Linus Torvalds 已提交
4173 4174 4175 4176 4177 4178
 *
 * 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
4179 4180
 * 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
4181
 */
4182
static void cache_reap(struct work_struct *w)
L
Linus Torvalds 已提交
4183
{
4184
	struct kmem_cache *searchp;
4185
	struct kmem_list3 *l3;
4186
	int node = numa_mem_id();
4187
	struct delayed_work *work = to_delayed_work(w);
L
Linus Torvalds 已提交
4188

4189
	if (!mutex_trylock(&cache_chain_mutex))
L
Linus Torvalds 已提交
4190
		/* Give up. Setup the next iteration. */
4191
		goto out;
L
Linus Torvalds 已提交
4192

4193
	list_for_each_entry(searchp, &cache_chain, list) {
L
Linus Torvalds 已提交
4194 4195
		check_irq_on();

4196 4197 4198 4199 4200
		/*
		 * 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.
		 */
4201
		l3 = searchp->nodelists[node];
4202

4203
		reap_alien(searchp, l3);
L
Linus Torvalds 已提交
4204

4205
		drain_array(searchp, l3, cpu_cache_get(searchp), 0, node);
L
Linus Torvalds 已提交
4206

4207 4208 4209 4210
		/*
		 * These are racy checks but it does not matter
		 * if we skip one check or scan twice.
		 */
4211
		if (time_after(l3->next_reap, jiffies))
4212
			goto next;
L
Linus Torvalds 已提交
4213

4214
		l3->next_reap = jiffies + REAPTIMEOUT_LIST3;
L
Linus Torvalds 已提交
4215

4216
		drain_array(searchp, l3, l3->shared, 0, node);
L
Linus Torvalds 已提交
4217

4218
		if (l3->free_touched)
4219
			l3->free_touched = 0;
4220 4221
		else {
			int freed;
L
Linus Torvalds 已提交
4222

4223 4224 4225 4226
			freed = drain_freelist(searchp, l3, (l3->free_limit +
				5 * searchp->num - 1) / (5 * searchp->num));
			STATS_ADD_REAPED(searchp, freed);
		}
4227
next:
L
Linus Torvalds 已提交
4228 4229 4230
		cond_resched();
	}
	check_irq_on();
I
Ingo Molnar 已提交
4231
	mutex_unlock(&cache_chain_mutex);
4232
	next_reap_node();
4233
out:
A
Andrew Morton 已提交
4234
	/* Set up the next iteration */
4235
	schedule_delayed_work(work, round_jiffies_relative(REAPTIMEOUT_CPUC));
L
Linus Torvalds 已提交
4236 4237
}

4238
#ifdef CONFIG_SLABINFO
L
Linus Torvalds 已提交
4239

4240
static void print_slabinfo_header(struct seq_file *m)
L
Linus Torvalds 已提交
4241
{
4242 4243 4244 4245
	/*
	 * Output format version, so at least we can change it
	 * without _too_ many complaints.
	 */
L
Linus Torvalds 已提交
4246
#if STATS
4247
	seq_puts(m, "slabinfo - version: 2.1 (statistics)\n");
L
Linus Torvalds 已提交
4248
#else
4249
	seq_puts(m, "slabinfo - version: 2.1\n");
L
Linus Torvalds 已提交
4250
#endif
4251 4252 4253 4254
	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
4255
#if STATS
4256
	seq_puts(m, " : globalstat <listallocs> <maxobjs> <grown> <reaped> "
4257
		 "<error> <maxfreeable> <nodeallocs> <remotefrees> <alienoverflow>");
4258
	seq_puts(m, " : cpustat <allochit> <allocmiss> <freehit> <freemiss>");
L
Linus Torvalds 已提交
4259
#endif
4260 4261 4262 4263 4264 4265 4266
	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
4267
	mutex_lock(&cache_chain_mutex);
4268 4269
	if (!n)
		print_slabinfo_header(m);
4270 4271

	return seq_list_start(&cache_chain, *pos);
L
Linus Torvalds 已提交
4272 4273 4274 4275
}

static void *s_next(struct seq_file *m, void *p, loff_t *pos)
{
4276
	return seq_list_next(p, &cache_chain, pos);
L
Linus Torvalds 已提交
4277 4278 4279 4280
}

static void s_stop(struct seq_file *m, void *p)
{
I
Ingo Molnar 已提交
4281
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
4282 4283 4284 4285
}

static int s_show(struct seq_file *m, void *p)
{
4286
	struct kmem_cache *cachep = list_entry(p, struct kmem_cache, list);
P
Pekka Enberg 已提交
4287 4288 4289 4290 4291
	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;
4292
	const char *name;
L
Linus Torvalds 已提交
4293
	char *error = NULL;
4294 4295
	int node;
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
4296 4297 4298

	active_objs = 0;
	num_slabs = 0;
4299 4300 4301 4302 4303
	for_each_online_node(node) {
		l3 = cachep->nodelists[node];
		if (!l3)
			continue;

4304 4305
		check_irq_on();
		spin_lock_irq(&l3->list_lock);
4306

4307
		list_for_each_entry(slabp, &l3->slabs_full, list) {
4308 4309 4310 4311 4312
			if (slabp->inuse != cachep->num && !error)
				error = "slabs_full accounting error";
			active_objs += cachep->num;
			active_slabs++;
		}
4313
		list_for_each_entry(slabp, &l3->slabs_partial, list) {
4314 4315 4316 4317 4318 4319 4320
			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++;
		}
4321
		list_for_each_entry(slabp, &l3->slabs_free, list) {
4322 4323 4324 4325 4326
			if (slabp->inuse && !error)
				error = "slabs_free/inuse accounting error";
			num_slabs++;
		}
		free_objects += l3->free_objects;
4327 4328
		if (l3->shared)
			shared_avail += l3->shared->avail;
4329

4330
		spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
4331
	}
P
Pekka Enberg 已提交
4332 4333
	num_slabs += active_slabs;
	num_objs = num_slabs * cachep->num;
4334
	if (num_objs - active_objs != free_objects && !error)
L
Linus Torvalds 已提交
4335 4336
		error = "free_objects accounting error";

P
Pekka Enberg 已提交
4337
	name = cachep->name;
L
Linus Torvalds 已提交
4338 4339 4340 4341
	if (error)
		printk(KERN_ERR "slab: cache %s error: %s\n", name, error);

	seq_printf(m, "%-17s %6lu %6lu %6u %4u %4d",
4342
		   name, active_objs, num_objs, cachep->size,
P
Pekka Enberg 已提交
4343
		   cachep->num, (1 << cachep->gfporder));
L
Linus Torvalds 已提交
4344
	seq_printf(m, " : tunables %4u %4u %4u",
P
Pekka Enberg 已提交
4345
		   cachep->limit, cachep->batchcount, cachep->shared);
4346
	seq_printf(m, " : slabdata %6lu %6lu %6lu",
P
Pekka Enberg 已提交
4347
		   active_slabs, num_slabs, shared_avail);
L
Linus Torvalds 已提交
4348
#if STATS
P
Pekka Enberg 已提交
4349
	{			/* list3 stats */
L
Linus Torvalds 已提交
4350 4351 4352 4353 4354 4355 4356
		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;
4357
		unsigned long node_frees = cachep->node_frees;
4358
		unsigned long overflows = cachep->node_overflow;
L
Linus Torvalds 已提交
4359

J
Joe Perches 已提交
4360 4361 4362 4363 4364
		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
4365 4366 4367 4368 4369 4370 4371 4372 4373
	}
	/* 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
4374
			   allochit, allocmiss, freehit, freemiss);
L
Linus Torvalds 已提交
4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394
	}
#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
 */

4395
static const struct seq_operations slabinfo_op = {
P
Pekka Enberg 已提交
4396 4397 4398 4399
	.start = s_start,
	.next = s_next,
	.stop = s_stop,
	.show = s_show,
L
Linus Torvalds 已提交
4400 4401 4402 4403 4404 4405 4406 4407 4408 4409
};

#define MAX_SLABINFO_WRITE 128
/**
 * slabinfo_write - Tuning for the slab allocator
 * @file: unused
 * @buffer: user buffer
 * @count: data length
 * @ppos: unused
 */
4410
static ssize_t slabinfo_write(struct file *file, const char __user *buffer,
P
Pekka Enberg 已提交
4411
		       size_t count, loff_t *ppos)
L
Linus Torvalds 已提交
4412
{
P
Pekka Enberg 已提交
4413
	char kbuf[MAX_SLABINFO_WRITE + 1], *tmp;
L
Linus Torvalds 已提交
4414
	int limit, batchcount, shared, res;
4415
	struct kmem_cache *cachep;
P
Pekka Enberg 已提交
4416

L
Linus Torvalds 已提交
4417 4418 4419 4420
	if (count > MAX_SLABINFO_WRITE)
		return -EINVAL;
	if (copy_from_user(&kbuf, buffer, count))
		return -EFAULT;
P
Pekka Enberg 已提交
4421
	kbuf[MAX_SLABINFO_WRITE] = '\0';
L
Linus Torvalds 已提交
4422 4423 4424 4425 4426 4427 4428 4429 4430 4431

	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
4432
	mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
4433
	res = -EINVAL;
4434
	list_for_each_entry(cachep, &cache_chain, list) {
L
Linus Torvalds 已提交
4435
		if (!strcmp(cachep->name, kbuf)) {
A
Andrew Morton 已提交
4436 4437
			if (limit < 1 || batchcount < 1 ||
					batchcount > limit || shared < 0) {
4438
				res = 0;
L
Linus Torvalds 已提交
4439
			} else {
4440
				res = do_tune_cpucache(cachep, limit,
4441 4442
						       batchcount, shared,
						       GFP_KERNEL);
L
Linus Torvalds 已提交
4443 4444 4445 4446
			}
			break;
		}
	}
I
Ingo Molnar 已提交
4447
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
4448 4449 4450 4451
	if (res >= 0)
		res = count;
	return res;
}
4452

4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465
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,
};

4466 4467 4468 4469 4470
#ifdef CONFIG_DEBUG_SLAB_LEAK

static void *leaks_start(struct seq_file *m, loff_t *pos)
{
	mutex_lock(&cache_chain_mutex);
4471
	return seq_list_start(&cache_chain, *pos);
4472 4473 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 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509
}

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;
4510
	for (i = 0, p = s->s_mem; i < c->num; i++, p += c->size) {
4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521
		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;
4522
	char modname[MODULE_NAME_LEN], name[KSYM_NAME_LEN];
4523

4524
	if (lookup_symbol_attrs(address, &size, &offset, modname, name) == 0) {
4525
		seq_printf(m, "%s+%#lx/%#lx", name, offset, size);
4526
		if (modname[0])
4527 4528 4529 4530 4531 4532 4533 4534 4535
			seq_printf(m, " [%s]", modname);
		return;
	}
#endif
	seq_printf(m, "%p", (void *)address);
}

static int leaks_show(struct seq_file *m, void *p)
{
4536
	struct kmem_cache *cachep = list_entry(p, struct kmem_cache, list);
4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560
	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);

4561
		list_for_each_entry(slabp, &l3->slabs_full, list)
4562
			handle_slab(n, cachep, slabp);
4563
		list_for_each_entry(slabp, &l3->slabs_partial, list)
4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589
			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');
	}
4590

4591 4592 4593
	return 0;
}

4594
static const struct seq_operations slabstats_op = {
4595 4596 4597 4598 4599
	.start = leaks_start,
	.next = s_next,
	.stop = s_stop,
	.show = leaks_show,
};
4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627

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)
{
4628
	proc_create("slabinfo",S_IWUSR|S_IRUSR,NULL,&proc_slabinfo_operations);
4629 4630
#ifdef CONFIG_DEBUG_SLAB_LEAK
	proc_create("slab_allocators", 0, NULL, &proc_slabstats_operations);
4631
#endif
4632 4633 4634
	return 0;
}
module_init(slab_proc_init);
L
Linus Torvalds 已提交
4635 4636
#endif

4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648
/**
 * 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
4649
size_t ksize(const void *objp)
L
Linus Torvalds 已提交
4650
{
4651 4652
	BUG_ON(!objp);
	if (unlikely(objp == ZERO_SIZE_PTR))
4653
		return 0;
L
Linus Torvalds 已提交
4654

4655
	return virt_to_cache(objp)->object_size;
L
Linus Torvalds 已提交
4656
}
K
Kirill A. Shutemov 已提交
4657
EXPORT_SYMBOL(ksize);