slab.c 109.5 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52
/*
 * linux/mm/slab.c
 * Written by Mark Hemment, 1996/97.
 * (markhe@nextd.demon.co.uk)
 *
 * kmem_cache_destroy() + some cleanup - 1999 Andrea Arcangeli
 *
 * Major cleanup, different bufctl logic, per-cpu arrays
 *	(c) 2000 Manfred Spraul
 *
 * Cleanup, make the head arrays unconditional, preparation for NUMA
 * 	(c) 2002 Manfred Spraul
 *
 * An implementation of the Slab Allocator as described in outline in;
 *	UNIX Internals: The New Frontiers by Uresh Vahalia
 *	Pub: Prentice Hall	ISBN 0-13-101908-2
 * or with a little more detail in;
 *	The Slab Allocator: An Object-Caching Kernel Memory Allocator
 *	Jeff Bonwick (Sun Microsystems).
 *	Presented at: USENIX Summer 1994 Technical Conference
 *
 * The memory is organized in caches, one cache for each object type.
 * (e.g. inode_cache, dentry_cache, buffer_head, vm_area_struct)
 * Each cache consists out of many slabs (they are small (usually one
 * page long) and always contiguous), and each slab contains multiple
 * initialized objects.
 *
 * This means, that your constructor is used only for newly allocated
 * slabs and you must pass objects with the same intializations to
 * kmem_cache_free.
 *
 * Each cache can only support one memory type (GFP_DMA, GFP_HIGHMEM,
 * normal). If you need a special memory type, then must create a new
 * cache for that memory type.
 *
 * In order to reduce fragmentation, the slabs are sorted in 3 groups:
 *   full slabs with 0 free objects
 *   partial slabs
 *   empty slabs with no allocated objects
 *
 * If partial slabs exist, then new allocations come from these slabs,
 * otherwise from empty slabs or new slabs are allocated.
 *
 * kmem_cache_destroy() CAN CRASH if you try to allocate from the cache
 * during kmem_cache_destroy(). The caller must prevent concurrent allocs.
 *
 * Each cache has a short per-cpu head array, most allocs
 * and frees go into that array, and if that array overflows, then 1/2
 * of the entries in the array are given back into the global cache.
 * The head array is strictly LIFO and should improve the cache hit rates.
 * On SMP, it additionally reduces the spinlock operations.
 *
A
Andrew Morton 已提交
53
 * The c_cpuarray may not be read with enabled local interrupts -
L
Linus Torvalds 已提交
54 55 56 57
 * it's changed with a smp_call_function().
 *
 * SMP synchronization:
 *  constructors and destructors are called without any locking.
58
 *  Several members in struct kmem_cache and struct slab never change, they
L
Linus Torvalds 已提交
59 60 61 62 63 64 65 66 67 68 69 70
 *	are accessed without any locking.
 *  The per-cpu arrays are never accessed from the wrong cpu, no locking,
 *  	and local interrupts are disabled so slab code is preempt-safe.
 *  The non-constant members are protected with a per-cache irq spinlock.
 *
 * Many thanks to Mark Hemment, who wrote another per-cpu slab patch
 * in 2000 - many ideas in the current implementation are derived from
 * his patch.
 *
 * Further notes from the original documentation:
 *
 * 11 April '97.  Started multi-threading - markhe
I
Ingo Molnar 已提交
71
 *	The global cache-chain is protected by the mutex 'cache_chain_mutex'.
L
Linus Torvalds 已提交
72 73 74 75 76 77
 *	The sem is only needed when accessing/extending the cache-chain, which
 *	can never happen inside an interrupt (kmem_cache_create(),
 *	kmem_cache_shrink() and kmem_cache_reap()).
 *
 *	At present, each engine can be growing a cache.  This should be blocked.
 *
78 79 80 81 82 83 84 85 86
 * 15 March 2005. NUMA slab allocator.
 *	Shai Fultheim <shai@scalex86.org>.
 *	Shobhit Dayal <shobhit@calsoftinc.com>
 *	Alok N Kataria <alokk@calsoftinc.com>
 *	Christoph Lameter <christoph@lameter.com>
 *
 *	Modified the slab allocator to be node aware on NUMA systems.
 *	Each node has its own list of partial, free and full slabs.
 *	All object allocations for a node occur from node specific slab lists.
L
Linus Torvalds 已提交
87 88 89 90 91
 */

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

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

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

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

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

#ifndef cache_line_size
#define cache_line_size()	L1_CACHE_BYTES
#endif

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

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

#ifndef ARCH_KMALLOC_FLAGS
#define ARCH_KMALLOC_FLAGS SLAB_HWCACHE_ALIGN
#endif

/* Legal flag mask for kmem_cache_create(). */
#if DEBUG
# define CREATE_MASK	(SLAB_DEBUG_INITIAL | SLAB_RED_ZONE | \
			 SLAB_POISON | SLAB_HWCACHE_ALIGN | \
176
			 SLAB_CACHE_DMA | \
L
Linus Torvalds 已提交
177 178
			 SLAB_MUST_HWCACHE_ALIGN | SLAB_STORE_USER | \
			 SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
179
			 SLAB_DESTROY_BY_RCU | SLAB_MEM_SPREAD)
L
Linus Torvalds 已提交
180
#else
181
# define CREATE_MASK	(SLAB_HWCACHE_ALIGN | \
L
Linus Torvalds 已提交
182 183
			 SLAB_CACHE_DMA | SLAB_MUST_HWCACHE_ALIGN | \
			 SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
184
			 SLAB_DESTROY_BY_RCU | SLAB_MEM_SPREAD)
L
Linus Torvalds 已提交
185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205
#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.
 */

206
typedef unsigned int kmem_bufctl_t;
L
Linus Torvalds 已提交
207 208
#define BUFCTL_END	(((kmem_bufctl_t)(~0U))-0)
#define BUFCTL_FREE	(((kmem_bufctl_t)(~0U))-1)
209 210
#define	BUFCTL_ACTIVE	(((kmem_bufctl_t)(~0U))-2)
#define	SLAB_LIMIT	(((kmem_bufctl_t)(~0U))-3)
L
Linus Torvalds 已提交
211 212 213 214 215 216 217 218 219

/*
 * struct slab
 *
 * Manages the objs in a slab. Placed either at the beginning of mem allocated
 * for a slab, or allocated from an general cache.
 * Slabs are chained into three list: fully used, partial, fully free slabs.
 */
struct slab {
P
Pekka Enberg 已提交
220 221 222 223 224 225
	struct list_head list;
	unsigned long colouroff;
	void *s_mem;		/* including colour offset */
	unsigned int inuse;	/* num of objs active in slab */
	kmem_bufctl_t free;
	unsigned short nodeid;
L
Linus Torvalds 已提交
226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244
};

/*
 * struct slab_rcu
 *
 * slab_destroy on a SLAB_DESTROY_BY_RCU cache uses this structure to
 * arrange for kmem_freepages to be called via RCU.  This is useful if
 * we need to approach a kernel structure obliquely, from its address
 * obtained without the usual locking.  We can lock the structure to
 * stabilize it and check it's still at the given address, only if we
 * can be sure that the memory has not been meanwhile reused for some
 * other kind of object (which our subsystem's lock might corrupt).
 *
 * rcu_read_lock before reading the address, then rcu_read_unlock after
 * taking the spinlock within the structure expected at that address.
 *
 * We assume struct slab_rcu can overlay struct slab when destroying.
 */
struct slab_rcu {
P
Pekka Enberg 已提交
245
	struct rcu_head head;
246
	struct kmem_cache *cachep;
P
Pekka Enberg 已提交
247
	void *addr;
L
Linus Torvalds 已提交
248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266
};

/*
 * 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;
267
	spinlock_t lock;
A
Andrew Morton 已提交
268 269 270 271 272 273
	void *entry[0];	/*
			 * Must have this definition in here for the proper
			 * alignment of array_cache. Also simplifies accessing
			 * the entries.
			 * [0] is for gcc 2.95. It should really be [].
			 */
L
Linus Torvalds 已提交
274 275
};

A
Andrew Morton 已提交
276 277 278
/*
 * bootstrap: The caches do not work without cpuarrays anymore, but the
 * cpuarrays are allocated from the generic caches...
L
Linus Torvalds 已提交
279 280 281 282
 */
#define BOOT_CPUCACHE_ENTRIES	1
struct arraycache_init {
	struct array_cache cache;
P
Pekka Enberg 已提交
283
	void *entries[BOOT_CPUCACHE_ENTRIES];
L
Linus Torvalds 已提交
284 285 286
};

/*
287
 * The slab lists for all objects.
L
Linus Torvalds 已提交
288 289
 */
struct kmem_list3 {
P
Pekka Enberg 已提交
290 291 292 293 294
	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;
295
	unsigned int colour_next;	/* Per-node cache coloring */
P
Pekka Enberg 已提交
296 297 298
	spinlock_t list_lock;
	struct array_cache *shared;	/* shared per node */
	struct array_cache **alien;	/* on other nodes */
299 300
	unsigned long next_reap;	/* updated without locking */
	int free_touched;		/* updated without locking */
L
Linus Torvalds 已提交
301 302
};

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

312 313 314 315
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);
316
static int enable_cpucache(struct kmem_cache *cachep);
317 318
static void cache_reap(void *unused);

319
/*
A
Andrew Morton 已提交
320 321
 * 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.
322
 */
323
static __always_inline int index_of(const size_t size)
324
{
325 326
	extern void __bad_size(void);

327 328 329 330 331 332 333 334 335 336
	if (__builtin_constant_p(size)) {
		int i = 0;

#define CACHE(x) \
	if (size <=x) \
		return i; \
	else \
		i++;
#include "linux/kmalloc_sizes.h"
#undef CACHE
337
		__bad_size();
338
	} else
339
		__bad_size();
340 341 342
	return 0;
}

343 344
static int slab_early_init = 1;

345 346
#define INDEX_AC index_of(sizeof(struct arraycache_init))
#define INDEX_L3 index_of(sizeof(struct kmem_list3))
L
Linus Torvalds 已提交
347

P
Pekka Enberg 已提交
348
static void kmem_list3_init(struct kmem_list3 *parent)
349 350 351 352 353 354
{
	INIT_LIST_HEAD(&parent->slabs_full);
	INIT_LIST_HEAD(&parent->slabs_partial);
	INIT_LIST_HEAD(&parent->slabs_free);
	parent->shared = NULL;
	parent->alien = NULL;
355
	parent->colour_next = 0;
356 357 358 359 360
	spin_lock_init(&parent->list_lock);
	parent->free_objects = 0;
	parent->free_touched = 0;
}

A
Andrew Morton 已提交
361 362 363 364
#define MAKE_LIST(cachep, listp, slab, nodeid)				\
	do {								\
		INIT_LIST_HEAD(listp);					\
		list_splice(&(cachep->nodelists[nodeid]->slab), listp);	\
365 366
	} while (0)

A
Andrew Morton 已提交
367 368
#define	MAKE_ALL_LISTS(cachep, ptr, nodeid)				\
	do {								\
369 370 371 372
	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
373 374

/*
375
 * struct kmem_cache
L
Linus Torvalds 已提交
376 377 378
 *
 * manages a cache.
 */
P
Pekka Enberg 已提交
379

380
struct kmem_cache {
L
Linus Torvalds 已提交
381
/* 1) per-cpu data, touched during every alloc/free */
P
Pekka Enberg 已提交
382
	struct array_cache *array[NR_CPUS];
383
/* 2) Cache tunables. Protected by cache_chain_mutex */
P
Pekka Enberg 已提交
384 385 386
	unsigned int batchcount;
	unsigned int limit;
	unsigned int shared;
387

388
	unsigned int buffer_size;
389
/* 3) touched by every alloc & free from the backend */
P
Pekka Enberg 已提交
390
	struct kmem_list3 *nodelists[MAX_NUMNODES];
391

A
Andrew Morton 已提交
392 393
	unsigned int flags;		/* constant flags */
	unsigned int num;		/* # of objs per slab */
L
Linus Torvalds 已提交
394

395
/* 4) cache_grow/shrink */
L
Linus Torvalds 已提交
396
	/* order of pgs per slab (2^n) */
P
Pekka Enberg 已提交
397
	unsigned int gfporder;
L
Linus Torvalds 已提交
398 399

	/* force GFP flags, e.g. GFP_DMA */
P
Pekka Enberg 已提交
400
	gfp_t gfpflags;
L
Linus Torvalds 已提交
401

A
Andrew Morton 已提交
402
	size_t colour;			/* cache colouring range */
P
Pekka Enberg 已提交
403
	unsigned int colour_off;	/* colour offset */
404
	struct kmem_cache *slabp_cache;
P
Pekka Enberg 已提交
405
	unsigned int slab_size;
A
Andrew Morton 已提交
406
	unsigned int dflags;		/* dynamic flags */
L
Linus Torvalds 已提交
407 408

	/* constructor func */
409
	void (*ctor) (void *, struct kmem_cache *, unsigned long);
L
Linus Torvalds 已提交
410 411

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

414
/* 5) cache creation/removal */
P
Pekka Enberg 已提交
415 416
	const char *name;
	struct list_head next;
L
Linus Torvalds 已提交
417

418
/* 6) statistics */
L
Linus Torvalds 已提交
419
#if STATS
P
Pekka Enberg 已提交
420 421 422 423 424 425 426 427 428
	unsigned long num_active;
	unsigned long num_allocations;
	unsigned long high_mark;
	unsigned long grown;
	unsigned long reaped;
	unsigned long errors;
	unsigned long max_freeable;
	unsigned long node_allocs;
	unsigned long node_frees;
429
	unsigned long node_overflow;
P
Pekka Enberg 已提交
430 431 432 433
	atomic_t allochit;
	atomic_t allocmiss;
	atomic_t freehit;
	atomic_t freemiss;
L
Linus Torvalds 已提交
434 435
#endif
#if DEBUG
436 437 438 439 440 441 442 443
	/*
	 * If debugging is enabled, then the allocator can add additional
	 * fields and/or padding to every object. buffer_size contains the total
	 * object size including these internal fields, the following two
	 * variables contain the offset to the user object and its size.
	 */
	int obj_offset;
	int obj_size;
L
Linus Torvalds 已提交
444 445 446 447 448 449 450
#endif
};

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

#define BATCHREFILL_LIMIT	16
A
Andrew Morton 已提交
451 452 453
/*
 * Optimization question: fewer reaps means less probability for unnessary
 * cpucache drain/refill cycles.
L
Linus Torvalds 已提交
454
 *
A
Adrian Bunk 已提交
455
 * OTOH the cpuarrays can contain lots of objects,
L
Linus Torvalds 已提交
456 457 458 459 460 461 462 463 464 465
 * 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++)
466
#define	STATS_ADD_REAPED(x,y)	((x)->reaped += (y))
A
Andrew Morton 已提交
467 468 469 470 471
#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
472 473
#define	STATS_INC_ERR(x)	((x)->errors++)
#define	STATS_INC_NODEALLOCS(x)	((x)->node_allocs++)
474
#define	STATS_INC_NODEFREES(x)	((x)->node_frees++)
475
#define STATS_INC_ACOVERFLOW(x)   ((x)->node_overflow++)
A
Andrew Morton 已提交
476 477 478 479 480
#define	STATS_SET_FREEABLE(x, i)					\
	do {								\
		if ((x)->max_freeable < i)				\
			(x)->max_freeable = i;				\
	} while (0)
L
Linus Torvalds 已提交
481 482 483 484 485 486 487 488 489
#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)
490
#define	STATS_ADD_REAPED(x,y)	do { } while (0)
L
Linus Torvalds 已提交
491 492 493
#define	STATS_SET_HIGH(x)	do { } while (0)
#define	STATS_INC_ERR(x)	do { } while (0)
#define	STATS_INC_NODEALLOCS(x)	do { } while (0)
494
#define	STATS_INC_NODEFREES(x)	do { } while (0)
495
#define STATS_INC_ACOVERFLOW(x)   do { } while (0)
A
Andrew Morton 已提交
496
#define	STATS_SET_FREEABLE(x, i) do { } while (0)
L
Linus Torvalds 已提交
497 498 499 500 501 502 503 504
#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
505 506
/*
 * memory layout of objects:
L
Linus Torvalds 已提交
507
 * 0		: objp
508
 * 0 .. cachep->obj_offset - BYTES_PER_WORD - 1: padding. This ensures that
L
Linus Torvalds 已提交
509 510
 * 		the end of an object is aligned with the end of the real
 * 		allocation. Catches writes behind the end of the allocation.
511
 * cachep->obj_offset - BYTES_PER_WORD .. cachep->obj_offset - 1:
L
Linus Torvalds 已提交
512
 * 		redzone word.
513 514
 * cachep->obj_offset: The real object.
 * cachep->buffer_size - 2* BYTES_PER_WORD: redzone word [BYTES_PER_WORD long]
A
Andrew Morton 已提交
515 516
 * cachep->buffer_size - 1* BYTES_PER_WORD: last caller address
 *					[BYTES_PER_WORD long]
L
Linus Torvalds 已提交
517
 */
518
static int obj_offset(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
519
{
520
	return cachep->obj_offset;
L
Linus Torvalds 已提交
521 522
}

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

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

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

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

#else

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

#endif

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

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

A
Andrew Morton 已提交
581 582 583 584
/*
 * Functions for storing/retrieving the cachep and or slab from the page
 * allocator.  These are used to find the slab an obj belongs to.  With kfree(),
 * these are used to find the cache which an obj belongs to.
L
Linus Torvalds 已提交
585
 */
586 587 588 589 590 591 592
static inline void page_set_cache(struct page *page, struct kmem_cache *cache)
{
	page->lru.next = (struct list_head *)cache;
}

static inline struct kmem_cache *page_get_cache(struct page *page)
{
593 594
	if (unlikely(PageCompound(page)))
		page = (struct page *)page_private(page);
595
	BUG_ON(!PageSlab(page));
596 597 598 599 600 601 602 603 604 605
	return (struct kmem_cache *)page->lru.next;
}

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

static inline struct slab *page_get_slab(struct page *page)
{
606 607
	if (unlikely(PageCompound(page)))
		page = (struct page *)page_private(page);
608
	BUG_ON(!PageSlab(page));
609 610
	return (struct slab *)page->lru.prev;
}
L
Linus Torvalds 已提交
611

612 613 614 615 616 617 618 619 620 621 622 623
static inline struct kmem_cache *virt_to_cache(const void *obj)
{
	struct page *page = virt_to_page(obj);
	return page_get_cache(page);
}

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

624 625 626 627 628 629 630 631 632 633 634 635
static inline void *index_to_obj(struct kmem_cache *cache, struct slab *slab,
				 unsigned int idx)
{
	return slab->s_mem + cache->buffer_size * idx;
}

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

A
Andrew Morton 已提交
636 637 638
/*
 * These are the default caches for kmalloc. Custom caches can have other sizes.
 */
L
Linus Torvalds 已提交
639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655
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
656
	{NULL,}
L
Linus Torvalds 已提交
657 658 659 660
#undef CACHE
};

static struct arraycache_init initarray_cache __initdata =
P
Pekka Enberg 已提交
661
    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
L
Linus Torvalds 已提交
662
static struct arraycache_init initarray_generic =
P
Pekka Enberg 已提交
663
    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
L
Linus Torvalds 已提交
664 665

/* internal cache of cache description objs */
666
static struct kmem_cache cache_cache = {
P
Pekka Enberg 已提交
667 668 669
	.batchcount = 1,
	.limit = BOOT_CPUCACHE_ENTRIES,
	.shared = 1,
670
	.buffer_size = sizeof(struct kmem_cache),
P
Pekka Enberg 已提交
671
	.name = "kmem_cache",
L
Linus Torvalds 已提交
672
#if DEBUG
673
	.obj_size = sizeof(struct kmem_cache),
L
Linus Torvalds 已提交
674 675 676
#endif
};

677 678
#define BAD_ALIEN_MAGIC 0x01020304ul

679 680 681 682 683 684 685 686
#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.
687 688 689 690
 *
 * 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
691
 */
692 693 694 695
static struct lock_class_key on_slab_l3_key;
static struct lock_class_key on_slab_alc_key;

static inline void init_lock_keys(void)
696 697 698

{
	int q;
699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725
	struct cache_sizes *s = malloc_sizes;

	while (s->cs_size != ULONG_MAX) {
		for_each_node(q) {
			struct array_cache **alc;
			int r;
			struct kmem_list3 *l3 = s->cs_cachep->nodelists[q];
			if (!l3 || OFF_SLAB(s->cs_cachep))
				continue;
			lockdep_set_class(&l3->list_lock, &on_slab_l3_key);
			alc = l3->alien;
			/*
			 * FIXME: This check for BAD_ALIEN_MAGIC
			 * should go away when common slab code is taught to
			 * work even without alien caches.
			 * Currently, non NUMA code returns BAD_ALIEN_MAGIC
			 * for alloc_alien_cache,
			 */
			if (!alc || (unsigned long)alc == BAD_ALIEN_MAGIC)
				continue;
			for_each_node(r) {
				if (alc[r])
					lockdep_set_class(&alc[r]->lock,
					     &on_slab_alc_key);
			}
		}
		s++;
726 727 728
	}
}
#else
729
static inline void init_lock_keys(void)
730 731 732 733
{
}
#endif

L
Linus Torvalds 已提交
734
/* Guard access to the cache-chain. */
I
Ingo Molnar 已提交
735
static DEFINE_MUTEX(cache_chain_mutex);
L
Linus Torvalds 已提交
736 737 738 739 740 741 742 743
static struct list_head cache_chain;

/*
 * chicken and egg problem: delay the per-cpu array allocation
 * until the general caches are up.
 */
static enum {
	NONE,
744 745
	PARTIAL_AC,
	PARTIAL_L3,
L
Linus Torvalds 已提交
746 747 748
	FULL
} g_cpucache_up;

749 750 751 752 753 754 755 756
/*
 * used by boot code to determine if it can use slab based allocator
 */
int slab_is_available(void)
{
	return g_cpucache_up == FULL;
}

L
Linus Torvalds 已提交
757 758
static DEFINE_PER_CPU(struct work_struct, reap_work);

759
static inline struct array_cache *cpu_cache_get(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
760 761 762 763
{
	return cachep->array[smp_processor_id()];
}

A
Andrew Morton 已提交
764 765
static inline struct kmem_cache *__find_general_cachep(size_t size,
							gfp_t gfpflags)
L
Linus Torvalds 已提交
766 767 768 769 770
{
	struct cache_sizes *csizep = malloc_sizes;

#if DEBUG
	/* This happens if someone tries to call
P
Pekka Enberg 已提交
771 772 773
	 * kmem_cache_create(), or __kmalloc(), before
	 * the generic caches are initialized.
	 */
774
	BUG_ON(malloc_sizes[INDEX_AC].cs_cachep == NULL);
L
Linus Torvalds 已提交
775 776 777 778 779
#endif
	while (size > csizep->cs_size)
		csizep++;

	/*
780
	 * Really subtle: The last entry with cs->cs_size==ULONG_MAX
L
Linus Torvalds 已提交
781 782 783 784 785 786 787 788
	 * has cs_{dma,}cachep==NULL. Thus no special case
	 * for large kmalloc calls required.
	 */
	if (unlikely(gfpflags & GFP_DMA))
		return csizep->cs_dmacachep;
	return csizep->cs_cachep;
}

A
Adrian Bunk 已提交
789
static struct kmem_cache *kmem_find_general_cachep(size_t size, gfp_t gfpflags)
790 791 792 793
{
	return __find_general_cachep(size, gfpflags);
}

794
static size_t slab_mgmt_size(size_t nr_objs, size_t align)
L
Linus Torvalds 已提交
795
{
796 797
	return ALIGN(sizeof(struct slab)+nr_objs*sizeof(kmem_bufctl_t), align);
}
L
Linus Torvalds 已提交
798

A
Andrew Morton 已提交
799 800 801
/*
 * Calculate the number of objects and left-over bytes for a given buffer size.
 */
802 803 804 805 806 807 808
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
809

810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857
	/*
	 * 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
858 859 860 861
}

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

A
Andrew Morton 已提交
862 863
static void __slab_error(const char *function, struct kmem_cache *cachep,
			char *msg)
L
Linus Torvalds 已提交
864 865
{
	printk(KERN_ERR "slab error in %s(): cache `%s': %s\n",
P
Pekka Enberg 已提交
866
	       function, cachep->name, msg);
L
Linus Torvalds 已提交
867 868 869
	dump_stack();
}

870 871 872 873 874 875 876 877 878 879 880 881 882 883 884
#ifdef CONFIG_NUMA
/*
 * Special reaping functions for NUMA systems called from cache_reap().
 * These take care of doing round robin flushing of alien caches (containing
 * objects freed on different nodes from which they were allocated) and the
 * flushing of remote pcps by calling drain_node_pages.
 */
static DEFINE_PER_CPU(unsigned long, reap_node);

static void init_reap_node(int cpu)
{
	int node;

	node = next_node(cpu_to_node(cpu), node_online_map);
	if (node == MAX_NUMNODES)
885
		node = first_node(node_online_map);
886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910

	__get_cpu_var(reap_node) = node;
}

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

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

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

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

L
Linus Torvalds 已提交
911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927
/*
 * Initiate the reap timer running on the target CPU.  We run at around 1 to 2Hz
 * via the workqueue/eventd.
 * Add the CPU number into the expiration time to minimize the possibility of
 * the CPUs getting into lockstep and contending for the global cache chain
 * lock.
 */
static void __devinit start_cpu_timer(int cpu)
{
	struct work_struct *reap_work = &per_cpu(reap_work, cpu);

	/*
	 * When this gets called from do_initcalls via cpucache_init(),
	 * init_workqueues() has already run, so keventd will be setup
	 * at that time.
	 */
	if (keventd_up() && reap_work->func == NULL) {
928
		init_reap_node(cpu);
L
Linus Torvalds 已提交
929 930 931 932 933
		INIT_WORK(reap_work, cache_reap, NULL);
		schedule_delayed_work_on(cpu, reap_work, HZ + 3 * cpu);
	}
}

934
static struct array_cache *alloc_arraycache(int node, int entries,
P
Pekka Enberg 已提交
935
					    int batchcount)
L
Linus Torvalds 已提交
936
{
P
Pekka Enberg 已提交
937
	int memsize = sizeof(void *) * entries + sizeof(struct array_cache);
L
Linus Torvalds 已提交
938 939
	struct array_cache *nc = NULL;

940
	nc = kmalloc_node(memsize, GFP_KERNEL, node);
L
Linus Torvalds 已提交
941 942 943 944 945
	if (nc) {
		nc->avail = 0;
		nc->limit = entries;
		nc->batchcount = batchcount;
		nc->touched = 0;
946
		spin_lock_init(&nc->lock);
L
Linus Torvalds 已提交
947 948 949 950
	}
	return nc;
}

951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974
/*
 * Transfer objects in one arraycache to another.
 * Locking must be handled by the caller.
 *
 * Return the number of entries transferred.
 */
static int transfer_objects(struct array_cache *to,
		struct array_cache *from, unsigned int max)
{
	/* Figure out how many entries to transfer */
	int nr = min(min(from->avail, max), to->limit - to->avail);

	if (!nr)
		return 0;

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

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

975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007
#ifndef CONFIG_NUMA

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

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

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

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

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

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

#else	/* CONFIG_NUMA */

1008
static void *__cache_alloc_node(struct kmem_cache *, gfp_t, int);
1009
static void *alternate_node_alloc(struct kmem_cache *, gfp_t);
1010

P
Pekka Enberg 已提交
1011
static struct array_cache **alloc_alien_cache(int node, int limit)
1012 1013
{
	struct array_cache **ac_ptr;
P
Pekka Enberg 已提交
1014
	int memsize = sizeof(void *) * MAX_NUMNODES;
1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027
	int i;

	if (limit > 1)
		limit = 12;
	ac_ptr = kmalloc_node(memsize, GFP_KERNEL, node);
	if (ac_ptr) {
		for_each_node(i) {
			if (i == node || !node_online(i)) {
				ac_ptr[i] = NULL;
				continue;
			}
			ac_ptr[i] = alloc_arraycache(node, limit, 0xbaadf00d);
			if (!ac_ptr[i]) {
P
Pekka Enberg 已提交
1028
				for (i--; i <= 0; i--)
1029 1030 1031 1032 1033 1034 1035 1036 1037
					kfree(ac_ptr[i]);
				kfree(ac_ptr);
				return NULL;
			}
		}
	}
	return ac_ptr;
}

P
Pekka Enberg 已提交
1038
static void free_alien_cache(struct array_cache **ac_ptr)
1039 1040 1041 1042 1043 1044
{
	int i;

	if (!ac_ptr)
		return;
	for_each_node(i)
P
Pekka Enberg 已提交
1045
	    kfree(ac_ptr[i]);
1046 1047 1048
	kfree(ac_ptr);
}

1049
static void __drain_alien_cache(struct kmem_cache *cachep,
P
Pekka Enberg 已提交
1050
				struct array_cache *ac, int node)
1051 1052 1053 1054 1055
{
	struct kmem_list3 *rl3 = cachep->nodelists[node];

	if (ac->avail) {
		spin_lock(&rl3->list_lock);
1056 1057 1058 1059 1060
		/*
		 * 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.
		 */
1061 1062
		if (rl3->shared)
			transfer_objects(rl3->shared, ac, ac->limit);
1063

1064
		free_block(cachep, ac->entry, ac->avail, node);
1065 1066 1067 1068 1069
		ac->avail = 0;
		spin_unlock(&rl3->list_lock);
	}
}

1070 1071 1072 1073 1074 1075 1076 1077 1078
/*
 * Called from cache_reap() to regularly drain alien caches round robin.
 */
static void reap_alien(struct kmem_cache *cachep, struct kmem_list3 *l3)
{
	int node = __get_cpu_var(reap_node);

	if (l3->alien) {
		struct array_cache *ac = l3->alien[node];
1079 1080

		if (ac && ac->avail && spin_trylock_irq(&ac->lock)) {
1081 1082 1083 1084 1085 1086
			__drain_alien_cache(cachep, ac, node);
			spin_unlock_irq(&ac->lock);
		}
	}
}

A
Andrew Morton 已提交
1087 1088
static void drain_alien_cache(struct kmem_cache *cachep,
				struct array_cache **alien)
1089
{
P
Pekka Enberg 已提交
1090
	int i = 0;
1091 1092 1093 1094
	struct array_cache *ac;
	unsigned long flags;

	for_each_online_node(i) {
1095
		ac = alien[i];
1096 1097 1098 1099 1100 1101 1102
		if (ac) {
			spin_lock_irqsave(&ac->lock, flags);
			__drain_alien_cache(cachep, ac, i);
			spin_unlock_irqrestore(&ac->lock, flags);
		}
	}
}
1103

1104
static inline int cache_free_alien(struct kmem_cache *cachep, void *objp)
1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121
{
	struct slab *slabp = virt_to_slab(objp);
	int nodeid = slabp->nodeid;
	struct kmem_list3 *l3;
	struct array_cache *alien = NULL;

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

	l3 = cachep->nodelists[numa_node_id()];
	STATS_INC_NODEFREES(cachep);
	if (l3->alien && l3->alien[nodeid]) {
		alien = l3->alien[nodeid];
1122
		spin_lock(&alien->lock);
1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135
		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;
}
1136 1137
#endif

1138
static int __cpuinit cpuup_callback(struct notifier_block *nfb,
P
Pekka Enberg 已提交
1139
				    unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
1140 1141
{
	long cpu = (long)hcpu;
1142
	struct kmem_cache *cachep;
1143 1144 1145
	struct kmem_list3 *l3 = NULL;
	int node = cpu_to_node(cpu);
	int memsize = sizeof(struct kmem_list3);
L
Linus Torvalds 已提交
1146 1147 1148

	switch (action) {
	case CPU_UP_PREPARE:
I
Ingo Molnar 已提交
1149
		mutex_lock(&cache_chain_mutex);
A
Andrew Morton 已提交
1150 1151
		/*
		 * We need to do this right in the beginning since
1152 1153 1154 1155 1156
		 * alloc_arraycache's are going to use this list.
		 * kmalloc_node allows us to add the slab to the right
		 * kmem_list3 and not this cpu's kmem_list3
		 */

L
Linus Torvalds 已提交
1157
		list_for_each_entry(cachep, &cache_chain, next) {
A
Andrew Morton 已提交
1158 1159
			/*
			 * Set up the size64 kmemlist for cpu before we can
1160 1161 1162 1163
			 * begin anything. Make sure some other cpu on this
			 * node has not already allocated this
			 */
			if (!cachep->nodelists[node]) {
A
Andrew Morton 已提交
1164 1165
				l3 = kmalloc_node(memsize, GFP_KERNEL, node);
				if (!l3)
1166 1167 1168
					goto bad;
				kmem_list3_init(l3);
				l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
P
Pekka Enberg 已提交
1169
				    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
1170

1171 1172 1173 1174 1175
				/*
				 * The l3s don't come and go as CPUs come and
				 * go.  cache_chain_mutex is sufficient
				 * protection here.
				 */
1176 1177
				cachep->nodelists[node] = l3;
			}
L
Linus Torvalds 已提交
1178

1179 1180
			spin_lock_irq(&cachep->nodelists[node]->list_lock);
			cachep->nodelists[node]->free_limit =
A
Andrew Morton 已提交
1181 1182
				(1 + nr_cpus_node(node)) *
				cachep->batchcount + cachep->num;
1183 1184 1185
			spin_unlock_irq(&cachep->nodelists[node]->list_lock);
		}

A
Andrew Morton 已提交
1186 1187 1188 1189
		/*
		 * Now we can go ahead with allocating the shared arrays and
		 * array caches
		 */
1190
		list_for_each_entry(cachep, &cache_chain, next) {
1191
			struct array_cache *nc;
1192 1193
			struct array_cache *shared;
			struct array_cache **alien;
1194

1195
			nc = alloc_arraycache(node, cachep->limit,
1196
						cachep->batchcount);
L
Linus Torvalds 已提交
1197 1198
			if (!nc)
				goto bad;
1199 1200 1201 1202 1203
			shared = alloc_arraycache(node,
					cachep->shared * cachep->batchcount,
					0xbaadf00d);
			if (!shared)
				goto bad;
1204

1205 1206 1207
			alien = alloc_alien_cache(node, cachep->limit);
			if (!alien)
				goto bad;
L
Linus Torvalds 已提交
1208
			cachep->array[cpu] = nc;
1209 1210 1211
			l3 = cachep->nodelists[node];
			BUG_ON(!l3);

1212 1213 1214 1215 1216 1217 1218 1219
			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;
1220
			}
1221 1222 1223 1224 1225 1226 1227 1228 1229
#ifdef CONFIG_NUMA
			if (!l3->alien) {
				l3->alien = alien;
				alien = NULL;
			}
#endif
			spin_unlock_irq(&l3->list_lock);
			kfree(shared);
			free_alien_cache(alien);
L
Linus Torvalds 已提交
1230
		}
I
Ingo Molnar 已提交
1231
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1232 1233 1234 1235 1236 1237
		break;
	case CPU_ONLINE:
		start_cpu_timer(cpu);
		break;
#ifdef CONFIG_HOTPLUG_CPU
	case CPU_DEAD:
1238 1239 1240 1241 1242 1243 1244 1245
		/*
		 * Even if all the cpus of a node are down, we don't free the
		 * kmem_list3 of any cache. This to avoid a race between
		 * cpu_down, and a kmalloc allocation from another cpu for
		 * memory from the node of the cpu going down.  The list3
		 * structure is usually allocated from kmem_cache_create() and
		 * gets destroyed at kmem_cache_destroy().
		 */
L
Linus Torvalds 已提交
1246 1247
		/* fall thru */
	case CPU_UP_CANCELED:
I
Ingo Molnar 已提交
1248
		mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1249 1250
		list_for_each_entry(cachep, &cache_chain, next) {
			struct array_cache *nc;
1251 1252
			struct array_cache *shared;
			struct array_cache **alien;
1253
			cpumask_t mask;
L
Linus Torvalds 已提交
1254

1255
			mask = node_to_cpumask(node);
L
Linus Torvalds 已提交
1256 1257 1258
			/* cpu is dead; no one can alloc from it. */
			nc = cachep->array[cpu];
			cachep->array[cpu] = NULL;
1259 1260 1261
			l3 = cachep->nodelists[node];

			if (!l3)
1262
				goto free_array_cache;
1263

1264
			spin_lock_irq(&l3->list_lock);
1265 1266 1267 1268

			/* Free limit for this kmem_list3 */
			l3->free_limit -= cachep->batchcount;
			if (nc)
1269
				free_block(cachep, nc->entry, nc->avail, node);
1270 1271

			if (!cpus_empty(mask)) {
1272
				spin_unlock_irq(&l3->list_lock);
1273
				goto free_array_cache;
P
Pekka Enberg 已提交
1274
			}
1275

1276 1277
			shared = l3->shared;
			if (shared) {
1278
				free_block(cachep, l3->shared->entry,
P
Pekka Enberg 已提交
1279
					   l3->shared->avail, node);
1280 1281 1282
				l3->shared = NULL;
			}

1283 1284 1285 1286 1287 1288 1289 1290 1291
			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);
1292
			}
1293
free_array_cache:
L
Linus Torvalds 已提交
1294 1295
			kfree(nc);
		}
1296 1297 1298 1299 1300 1301 1302 1303 1304
		/*
		 * In the previous loop, all the objects were freed to
		 * the respective cache's slabs,  now we can go ahead and
		 * shrink each nodelist to its limit.
		 */
		list_for_each_entry(cachep, &cache_chain, next) {
			l3 = cachep->nodelists[node];
			if (!l3)
				continue;
1305
			drain_freelist(cachep, l3, l3->free_objects);
1306
		}
I
Ingo Molnar 已提交
1307
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1308 1309 1310 1311
		break;
#endif
	}
	return NOTIFY_OK;
A
Andrew Morton 已提交
1312
bad:
I
Ingo Molnar 已提交
1313
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1314 1315 1316
	return NOTIFY_BAD;
}

1317 1318 1319
static struct notifier_block __cpuinitdata cpucache_notifier = {
	&cpuup_callback, NULL, 0
};
L
Linus Torvalds 已提交
1320

1321 1322 1323
/*
 * swap the static kmem_list3 with kmalloced memory
 */
A
Andrew Morton 已提交
1324 1325
static void init_list(struct kmem_cache *cachep, struct kmem_list3 *list,
			int nodeid)
1326 1327 1328 1329 1330 1331 1332 1333 1334
{
	struct kmem_list3 *ptr;

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

	local_irq_disable();
	memcpy(ptr, list, sizeof(struct kmem_list3));
1335 1336 1337 1338 1339
	/*
	 * Do not assume that spinlocks can be initialized via memcpy:
	 */
	spin_lock_init(&ptr->list_lock);

1340 1341 1342 1343 1344
	MAKE_ALL_LISTS(cachep, ptr, nodeid);
	cachep->nodelists[nodeid] = ptr;
	local_irq_enable();
}

A
Andrew Morton 已提交
1345 1346 1347
/*
 * Initialisation.  Called after the page allocator have been initialised and
 * before smp_init().
L
Linus Torvalds 已提交
1348 1349 1350 1351 1352 1353
 */
void __init kmem_cache_init(void)
{
	size_t left_over;
	struct cache_sizes *sizes;
	struct cache_names *names;
1354
	int i;
1355
	int order;
1356 1357 1358 1359 1360 1361

	for (i = 0; i < NUM_INIT_LISTS; i++) {
		kmem_list3_init(&initkmem_list3[i]);
		if (i < MAX_NUMNODES)
			cache_cache.nodelists[i] = NULL;
	}
L
Linus Torvalds 已提交
1362 1363 1364 1365 1366 1367 1368 1369 1370 1371

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

	/* Bootstrap is tricky, because several objects are allocated
	 * from caches that do not exist yet:
A
Andrew Morton 已提交
1372 1373 1374
	 * 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.
1375 1376 1377
	 *    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
1378
	 * 2) Create the first kmalloc cache.
1379
	 *    The struct kmem_cache for the new cache is allocated normally.
1380 1381 1382
	 *    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
1383 1384
	 * 4) Replace the __init data head arrays for cache_cache and the first
	 *    kmalloc cache with kmalloc allocated arrays.
1385 1386 1387
	 * 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
1388 1389 1390 1391 1392 1393 1394
	 */

	/* 1) create the cache_cache */
	INIT_LIST_HEAD(&cache_chain);
	list_add(&cache_cache.next, &cache_chain);
	cache_cache.colour_off = cache_line_size();
	cache_cache.array[smp_processor_id()] = &initarray_cache.cache;
1395
	cache_cache.nodelists[numa_node_id()] = &initkmem_list3[CACHE_CACHE];
L
Linus Torvalds 已提交
1396

A
Andrew Morton 已提交
1397 1398
	cache_cache.buffer_size = ALIGN(cache_cache.buffer_size,
					cache_line_size());
L
Linus Torvalds 已提交
1399

1400 1401 1402 1403 1404 1405
	for (order = 0; order < MAX_ORDER; order++) {
		cache_estimate(order, cache_cache.buffer_size,
			cache_line_size(), 0, &left_over, &cache_cache.num);
		if (cache_cache.num)
			break;
	}
1406
	BUG_ON(!cache_cache.num);
1407
	cache_cache.gfporder = order;
P
Pekka Enberg 已提交
1408 1409 1410
	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
1411 1412 1413 1414 1415

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

A
Andrew Morton 已提交
1416 1417 1418 1419
	/*
	 * Initialize the caches that provide memory for the array cache and the
	 * kmem_list3 structures first.  Without this, further allocations will
	 * bug.
1420 1421 1422
	 */

	sizes[INDEX_AC].cs_cachep = kmem_cache_create(names[INDEX_AC].name,
A
Andrew Morton 已提交
1423 1424 1425 1426
					sizes[INDEX_AC].cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_PANIC,
					NULL, NULL);
1427

A
Andrew Morton 已提交
1428
	if (INDEX_AC != INDEX_L3) {
1429
		sizes[INDEX_L3].cs_cachep =
A
Andrew Morton 已提交
1430 1431 1432 1433 1434 1435
			kmem_cache_create(names[INDEX_L3].name,
				sizes[INDEX_L3].cs_size,
				ARCH_KMALLOC_MINALIGN,
				ARCH_KMALLOC_FLAGS|SLAB_PANIC,
				NULL, NULL);
	}
1436

1437 1438
	slab_early_init = 0;

L
Linus Torvalds 已提交
1439
	while (sizes->cs_size != ULONG_MAX) {
1440 1441
		/*
		 * For performance, all the general caches are L1 aligned.
L
Linus Torvalds 已提交
1442 1443 1444
		 * This should be particularly beneficial on SMP boxes, as it
		 * eliminates "false sharing".
		 * Note for systems short on memory removing the alignment will
1445 1446
		 * allow tighter packing of the smaller caches.
		 */
A
Andrew Morton 已提交
1447
		if (!sizes->cs_cachep) {
1448
			sizes->cs_cachep = kmem_cache_create(names->name,
A
Andrew Morton 已提交
1449 1450 1451 1452 1453
					sizes->cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_PANIC,
					NULL, NULL);
		}
L
Linus Torvalds 已提交
1454 1455

		sizes->cs_dmacachep = kmem_cache_create(names->name_dma,
A
Andrew Morton 已提交
1456 1457 1458 1459 1460
					sizes->cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_CACHE_DMA|
						SLAB_PANIC,
					NULL, NULL);
L
Linus Torvalds 已提交
1461 1462 1463 1464 1465
		sizes++;
		names++;
	}
	/* 4) Replace the bootstrap head arrays */
	{
1466
		struct array_cache *ptr;
1467

L
Linus Torvalds 已提交
1468
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
1469

L
Linus Torvalds 已提交
1470
		local_irq_disable();
1471 1472
		BUG_ON(cpu_cache_get(&cache_cache) != &initarray_cache.cache);
		memcpy(ptr, cpu_cache_get(&cache_cache),
P
Pekka Enberg 已提交
1473
		       sizeof(struct arraycache_init));
1474 1475 1476 1477 1478
		/*
		 * Do not assume that spinlocks can be initialized via memcpy:
		 */
		spin_lock_init(&ptr->lock);

L
Linus Torvalds 已提交
1479 1480
		cache_cache.array[smp_processor_id()] = ptr;
		local_irq_enable();
1481

L
Linus Torvalds 已提交
1482
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
1483

L
Linus Torvalds 已提交
1484
		local_irq_disable();
1485
		BUG_ON(cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep)
P
Pekka Enberg 已提交
1486
		       != &initarray_generic.cache);
1487
		memcpy(ptr, cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep),
P
Pekka Enberg 已提交
1488
		       sizeof(struct arraycache_init));
1489 1490 1491 1492 1493
		/*
		 * Do not assume that spinlocks can be initialized via memcpy:
		 */
		spin_lock_init(&ptr->lock);

1494
		malloc_sizes[INDEX_AC].cs_cachep->array[smp_processor_id()] =
P
Pekka Enberg 已提交
1495
		    ptr;
L
Linus Torvalds 已提交
1496 1497
		local_irq_enable();
	}
1498 1499 1500 1501 1502
	/* 5) Replace the bootstrap kmem_list3's */
	{
		int node;
		/* Replace the static kmem_list3 structures for the boot cpu */
		init_list(&cache_cache, &initkmem_list3[CACHE_CACHE],
P
Pekka Enberg 已提交
1503
			  numa_node_id());
1504 1505 1506

		for_each_online_node(node) {
			init_list(malloc_sizes[INDEX_AC].cs_cachep,
P
Pekka Enberg 已提交
1507
				  &initkmem_list3[SIZE_AC + node], node);
1508 1509 1510

			if (INDEX_AC != INDEX_L3) {
				init_list(malloc_sizes[INDEX_L3].cs_cachep,
P
Pekka Enberg 已提交
1511 1512
					  &initkmem_list3[SIZE_L3 + node],
					  node);
1513 1514 1515
			}
		}
	}
L
Linus Torvalds 已提交
1516

1517
	/* 6) resize the head arrays to their final sizes */
L
Linus Torvalds 已提交
1518
	{
1519
		struct kmem_cache *cachep;
I
Ingo Molnar 已提交
1520
		mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1521
		list_for_each_entry(cachep, &cache_chain, next)
1522 1523
			if (enable_cpucache(cachep))
				BUG();
I
Ingo Molnar 已提交
1524
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
1525 1526
	}

1527 1528 1529 1530
	/* Annotate slab for lockdep -- annotate the malloc caches */
	init_lock_keys();


L
Linus Torvalds 已提交
1531 1532 1533
	/* Done! */
	g_cpucache_up = FULL;

A
Andrew Morton 已提交
1534 1535 1536
	/*
	 * Register a cpu startup notifier callback that initializes
	 * cpu_cache_get for all new cpus
L
Linus Torvalds 已提交
1537 1538 1539
	 */
	register_cpu_notifier(&cpucache_notifier);

A
Andrew Morton 已提交
1540 1541 1542
	/*
	 * 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
1543 1544 1545 1546 1547 1548 1549
	 */
}

static int __init cpucache_init(void)
{
	int cpu;

A
Andrew Morton 已提交
1550 1551
	/*
	 * Register the timers that return unneeded pages to the page allocator
L
Linus Torvalds 已提交
1552
	 */
1553
	for_each_online_cpu(cpu)
A
Andrew Morton 已提交
1554
		start_cpu_timer(cpu);
L
Linus Torvalds 已提交
1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565
	return 0;
}
__initcall(cpucache_init);

/*
 * Interface to system's page allocator. No need to hold the cache-lock.
 *
 * If we requested dmaable memory, we will get it. Even if we
 * did not request dmaable memory, we might get it, but that
 * would be relatively rare and ignorable.
 */
1566
static void *kmem_getpages(struct kmem_cache *cachep, gfp_t flags, int nodeid)
L
Linus Torvalds 已提交
1567 1568
{
	struct page *page;
1569
	int nr_pages;
L
Linus Torvalds 已提交
1570 1571
	int i;

1572
#ifndef CONFIG_MMU
1573 1574 1575
	/*
	 * Nommu uses slab's for process anonymous memory allocations, and thus
	 * requires __GFP_COMP to properly refcount higher order allocations
1576
	 */
1577
	flags |= __GFP_COMP;
1578
#endif
1579 1580 1581 1582 1583 1584 1585

	/*
	 * Under NUMA we want memory on the indicated node. We will handle
	 * the needed fallback ourselves since we want to serve from our
	 * per node object lists first for other nodes.
	 */
	flags |= cachep->gfpflags | GFP_THISNODE;
1586 1587

	page = alloc_pages_node(nodeid, flags, cachep->gfporder);
L
Linus Torvalds 已提交
1588 1589 1590
	if (!page)
		return NULL;

1591
	nr_pages = (1 << cachep->gfporder);
L
Linus Torvalds 已提交
1592
	if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
1593 1594 1595 1596 1597
		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);
1598 1599 1600
	for (i = 0; i < nr_pages; i++)
		__SetPageSlab(page + i);
	return page_address(page);
L
Linus Torvalds 已提交
1601 1602 1603 1604 1605
}

/*
 * Interface to system's page release.
 */
1606
static void kmem_freepages(struct kmem_cache *cachep, void *addr)
L
Linus Torvalds 已提交
1607
{
P
Pekka Enberg 已提交
1608
	unsigned long i = (1 << cachep->gfporder);
L
Linus Torvalds 已提交
1609 1610 1611
	struct page *page = virt_to_page(addr);
	const unsigned long nr_freed = i;

1612 1613 1614 1615 1616 1617
	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
1618
	while (i--) {
N
Nick Piggin 已提交
1619 1620
		BUG_ON(!PageSlab(page));
		__ClearPageSlab(page);
L
Linus Torvalds 已提交
1621 1622 1623 1624 1625 1626 1627 1628 1629
		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
1630
	struct slab_rcu *slab_rcu = (struct slab_rcu *)head;
1631
	struct kmem_cache *cachep = slab_rcu->cachep;
L
Linus Torvalds 已提交
1632 1633 1634 1635 1636 1637 1638 1639 1640

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

#if DEBUG

#ifdef CONFIG_DEBUG_PAGEALLOC
1641
static void store_stackinfo(struct kmem_cache *cachep, unsigned long *addr,
P
Pekka Enberg 已提交
1642
			    unsigned long caller)
L
Linus Torvalds 已提交
1643
{
1644
	int size = obj_size(cachep);
L
Linus Torvalds 已提交
1645

1646
	addr = (unsigned long *)&((char *)addr)[obj_offset(cachep)];
L
Linus Torvalds 已提交
1647

P
Pekka Enberg 已提交
1648
	if (size < 5 * sizeof(unsigned long))
L
Linus Torvalds 已提交
1649 1650
		return;

P
Pekka Enberg 已提交
1651 1652 1653 1654
	*addr++ = 0x12345678;
	*addr++ = caller;
	*addr++ = smp_processor_id();
	size -= 3 * sizeof(unsigned long);
L
Linus Torvalds 已提交
1655 1656 1657 1658 1659 1660 1661
	{
		unsigned long *sptr = &caller;
		unsigned long svalue;

		while (!kstack_end(sptr)) {
			svalue = *sptr++;
			if (kernel_text_address(svalue)) {
P
Pekka Enberg 已提交
1662
				*addr++ = svalue;
L
Linus Torvalds 已提交
1663 1664 1665 1666 1667 1668 1669
				size -= sizeof(unsigned long);
				if (size <= sizeof(unsigned long))
					break;
			}
		}

	}
P
Pekka Enberg 已提交
1670
	*addr++ = 0x87654321;
L
Linus Torvalds 已提交
1671 1672 1673
}
#endif

1674
static void poison_obj(struct kmem_cache *cachep, void *addr, unsigned char val)
L
Linus Torvalds 已提交
1675
{
1676 1677
	int size = obj_size(cachep);
	addr = &((char *)addr)[obj_offset(cachep)];
L
Linus Torvalds 已提交
1678 1679

	memset(addr, val, size);
P
Pekka Enberg 已提交
1680
	*(unsigned char *)(addr + size - 1) = POISON_END;
L
Linus Torvalds 已提交
1681 1682 1683 1684 1685 1686
}

static void dump_line(char *data, int offset, int limit)
{
	int i;
	printk(KERN_ERR "%03x:", offset);
A
Andrew Morton 已提交
1687
	for (i = 0; i < limit; i++)
P
Pekka Enberg 已提交
1688
		printk(" %02x", (unsigned char)data[offset + i]);
L
Linus Torvalds 已提交
1689 1690 1691 1692 1693 1694
	printk("\n");
}
#endif

#if DEBUG

1695
static void print_objinfo(struct kmem_cache *cachep, void *objp, int lines)
L
Linus Torvalds 已提交
1696 1697 1698 1699 1700 1701
{
	int i, size;
	char *realobj;

	if (cachep->flags & SLAB_RED_ZONE) {
		printk(KERN_ERR "Redzone: 0x%lx/0x%lx.\n",
A
Andrew Morton 已提交
1702 1703
			*dbg_redzone1(cachep, objp),
			*dbg_redzone2(cachep, objp));
L
Linus Torvalds 已提交
1704 1705 1706 1707
	}

	if (cachep->flags & SLAB_STORE_USER) {
		printk(KERN_ERR "Last user: [<%p>]",
A
Andrew Morton 已提交
1708
			*dbg_userword(cachep, objp));
L
Linus Torvalds 已提交
1709
		print_symbol("(%s)",
A
Andrew Morton 已提交
1710
				(unsigned long)*dbg_userword(cachep, objp));
L
Linus Torvalds 已提交
1711 1712
		printk("\n");
	}
1713 1714
	realobj = (char *)objp + obj_offset(cachep);
	size = obj_size(cachep);
P
Pekka Enberg 已提交
1715
	for (i = 0; i < size && lines; i += 16, lines--) {
L
Linus Torvalds 已提交
1716 1717
		int limit;
		limit = 16;
P
Pekka Enberg 已提交
1718 1719
		if (i + limit > size)
			limit = size - i;
L
Linus Torvalds 已提交
1720 1721 1722 1723
		dump_line(realobj, i, limit);
	}
}

1724
static void check_poison_obj(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
1725 1726 1727 1728 1729
{
	char *realobj;
	int size, i;
	int lines = 0;

1730 1731
	realobj = (char *)objp + obj_offset(cachep);
	size = obj_size(cachep);
L
Linus Torvalds 已提交
1732

P
Pekka Enberg 已提交
1733
	for (i = 0; i < size; i++) {
L
Linus Torvalds 已提交
1734
		char exp = POISON_FREE;
P
Pekka Enberg 已提交
1735
		if (i == size - 1)
L
Linus Torvalds 已提交
1736 1737 1738 1739 1740 1741
			exp = POISON_END;
		if (realobj[i] != exp) {
			int limit;
			/* Mismatch ! */
			/* Print header */
			if (lines == 0) {
P
Pekka Enberg 已提交
1742
				printk(KERN_ERR
A
Andrew Morton 已提交
1743 1744
					"Slab corruption: start=%p, len=%d\n",
					realobj, size);
L
Linus Torvalds 已提交
1745 1746 1747
				print_objinfo(cachep, objp, 0);
			}
			/* Hexdump the affected line */
P
Pekka Enberg 已提交
1748
			i = (i / 16) * 16;
L
Linus Torvalds 已提交
1749
			limit = 16;
P
Pekka Enberg 已提交
1750 1751
			if (i + limit > size)
				limit = size - i;
L
Linus Torvalds 已提交
1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763
			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:
		 */
1764
		struct slab *slabp = virt_to_slab(objp);
1765
		unsigned int objnr;
L
Linus Torvalds 已提交
1766

1767
		objnr = obj_to_index(cachep, slabp, objp);
L
Linus Torvalds 已提交
1768
		if (objnr) {
1769
			objp = index_to_obj(cachep, slabp, objnr - 1);
1770
			realobj = (char *)objp + obj_offset(cachep);
L
Linus Torvalds 已提交
1771
			printk(KERN_ERR "Prev obj: start=%p, len=%d\n",
P
Pekka Enberg 已提交
1772
			       realobj, size);
L
Linus Torvalds 已提交
1773 1774
			print_objinfo(cachep, objp, 2);
		}
P
Pekka Enberg 已提交
1775
		if (objnr + 1 < cachep->num) {
1776
			objp = index_to_obj(cachep, slabp, objnr + 1);
1777
			realobj = (char *)objp + obj_offset(cachep);
L
Linus Torvalds 已提交
1778
			printk(KERN_ERR "Next obj: start=%p, len=%d\n",
P
Pekka Enberg 已提交
1779
			       realobj, size);
L
Linus Torvalds 已提交
1780 1781 1782 1783 1784 1785
			print_objinfo(cachep, objp, 2);
		}
	}
}
#endif

1786 1787
#if DEBUG
/**
1788 1789 1790 1791 1792 1793
 * slab_destroy_objs - destroy a slab and its objects
 * @cachep: cache pointer being destroyed
 * @slabp: slab pointer being destroyed
 *
 * Call the registered destructor for each object in a slab that is being
 * destroyed.
L
Linus Torvalds 已提交
1794
 */
1795
static void slab_destroy_objs(struct kmem_cache *cachep, struct slab *slabp)
L
Linus Torvalds 已提交
1796 1797 1798
{
	int i;
	for (i = 0; i < cachep->num; i++) {
1799
		void *objp = index_to_obj(cachep, slabp, i);
L
Linus Torvalds 已提交
1800 1801 1802

		if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
A
Andrew Morton 已提交
1803 1804
			if (cachep->buffer_size % PAGE_SIZE == 0 &&
					OFF_SLAB(cachep))
P
Pekka Enberg 已提交
1805
				kernel_map_pages(virt_to_page(objp),
A
Andrew Morton 已提交
1806
					cachep->buffer_size / PAGE_SIZE, 1);
L
Linus Torvalds 已提交
1807 1808 1809 1810 1811 1812 1813 1814 1815
			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
1816
					   "was overwritten");
L
Linus Torvalds 已提交
1817 1818
			if (*dbg_redzone2(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "end of a freed object "
P
Pekka Enberg 已提交
1819
					   "was overwritten");
L
Linus Torvalds 已提交
1820 1821
		}
		if (cachep->dtor && !(cachep->flags & SLAB_POISON))
1822
			(cachep->dtor) (objp + obj_offset(cachep), cachep, 0);
L
Linus Torvalds 已提交
1823
	}
1824
}
L
Linus Torvalds 已提交
1825
#else
1826
static void slab_destroy_objs(struct kmem_cache *cachep, struct slab *slabp)
1827
{
L
Linus Torvalds 已提交
1828 1829 1830
	if (cachep->dtor) {
		int i;
		for (i = 0; i < cachep->num; i++) {
1831
			void *objp = index_to_obj(cachep, slabp, i);
P
Pekka Enberg 已提交
1832
			(cachep->dtor) (objp, cachep, 0);
L
Linus Torvalds 已提交
1833 1834
		}
	}
1835
}
L
Linus Torvalds 已提交
1836 1837
#endif

1838 1839 1840 1841 1842
/**
 * slab_destroy - destroy and release all objects in a slab
 * @cachep: cache pointer being destroyed
 * @slabp: slab pointer being destroyed
 *
1843
 * Destroy all the objs in a slab, and release the mem back to the system.
A
Andrew Morton 已提交
1844 1845
 * Before calling the slab must have been unlinked from the cache.  The
 * cache-lock is not held/needed.
1846
 */
1847
static void slab_destroy(struct kmem_cache *cachep, struct slab *slabp)
1848 1849 1850 1851
{
	void *addr = slabp->s_mem - slabp->colouroff;

	slab_destroy_objs(cachep, slabp);
L
Linus Torvalds 已提交
1852 1853 1854
	if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU)) {
		struct slab_rcu *slab_rcu;

P
Pekka Enberg 已提交
1855
		slab_rcu = (struct slab_rcu *)slabp;
L
Linus Torvalds 已提交
1856 1857 1858 1859 1860
		slab_rcu->cachep = cachep;
		slab_rcu->addr = addr;
		call_rcu(&slab_rcu->head, kmem_rcu_free);
	} else {
		kmem_freepages(cachep, addr);
1861 1862
		if (OFF_SLAB(cachep))
			kmem_cache_free(cachep->slabp_cache, slabp);
L
Linus Torvalds 已提交
1863 1864 1865
	}
}

A
Andrew Morton 已提交
1866 1867 1868 1869
/*
 * For setting up all the kmem_list3s for cache whose buffer_size is same as
 * size of kmem_list3.
 */
1870
static void set_up_list3s(struct kmem_cache *cachep, int index)
1871 1872 1873 1874
{
	int node;

	for_each_online_node(node) {
P
Pekka Enberg 已提交
1875
		cachep->nodelists[node] = &initkmem_list3[index + node];
1876
		cachep->nodelists[node]->next_reap = jiffies +
P
Pekka Enberg 已提交
1877 1878
		    REAPTIMEOUT_LIST3 +
		    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
1879 1880 1881
	}
}

1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902
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);
}


1903
/**
1904 1905 1906 1907 1908 1909 1910
 * 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.
1911 1912 1913 1914 1915
 *
 * 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
1916
static size_t calculate_slab_order(struct kmem_cache *cachep,
R
Randy Dunlap 已提交
1917
			size_t size, size_t align, unsigned long flags)
1918
{
1919
	unsigned long offslab_limit;
1920
	size_t left_over = 0;
1921
	int gfporder;
1922

A
Andrew Morton 已提交
1923
	for (gfporder = 0; gfporder <= MAX_GFP_ORDER; gfporder++) {
1924 1925 1926
		unsigned int num;
		size_t remainder;

1927
		cache_estimate(gfporder, size, align, flags, &remainder, &num);
1928 1929
		if (!num)
			continue;
1930

1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942
		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;
		}
1943

1944
		/* Found something acceptable - save it away */
1945
		cachep->num = num;
1946
		cachep->gfporder = gfporder;
1947 1948
		left_over = remainder;

1949 1950 1951 1952 1953 1954 1955 1956
		/*
		 * 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;

1957 1958 1959 1960
		/*
		 * Large number of objects is good, but very large slabs are
		 * currently bad for the gfp()s.
		 */
1961
		if (gfporder >= slab_break_gfp_order)
1962 1963
			break;

1964 1965 1966
		/*
		 * Acceptable internal fragmentation?
		 */
A
Andrew Morton 已提交
1967
		if (left_over * 8 <= (PAGE_SIZE << gfporder))
1968 1969 1970 1971 1972
			break;
	}
	return left_over;
}

1973
static int setup_cpu_cache(struct kmem_cache *cachep)
1974
{
1975 1976 1977
	if (g_cpucache_up == FULL)
		return enable_cpucache(cachep);

1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023
	if (g_cpucache_up == NONE) {
		/*
		 * Note: the first kmem_cache_create must create the cache
		 * that's used by kmalloc(24), otherwise the creation of
		 * further caches will BUG().
		 */
		cachep->array[smp_processor_id()] = &initarray_generic.cache;

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

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

	cpu_cache_get(cachep)->avail = 0;
	cpu_cache_get(cachep)->limit = BOOT_CPUCACHE_ENTRIES;
	cpu_cache_get(cachep)->batchcount = 1;
	cpu_cache_get(cachep)->touched = 0;
	cachep->batchcount = 1;
	cachep->limit = BOOT_CPUCACHE_ENTRIES;
2024
	return 0;
2025 2026
}

L
Linus Torvalds 已提交
2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041
/**
 * kmem_cache_create - Create a cache.
 * @name: A string which is used in /proc/slabinfo to identify this cache.
 * @size: The size of objects to be created in this cache.
 * @align: The required alignment for the objects.
 * @flags: SLAB flags
 * @ctor: A constructor for the objects.
 * @dtor: A destructor for the objects.
 *
 * Returns a ptr to the cache on success, NULL on failure.
 * Cannot be called within a int, but can be interrupted.
 * The @ctor is run when new pages are allocated by the cache
 * and the @dtor is run before the pages are handed back.
 *
 * @name must be valid until the cache is destroyed. This implies that
A
Andrew Morton 已提交
2042 2043
 * the module calling this has to destroy the cache before getting unloaded.
 *
L
Linus Torvalds 已提交
2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055
 * 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.
 */
2056
struct kmem_cache *
L
Linus Torvalds 已提交
2057
kmem_cache_create (const char *name, size_t size, size_t align,
A
Andrew Morton 已提交
2058 2059
	unsigned long flags,
	void (*ctor)(void*, struct kmem_cache *, unsigned long),
2060
	void (*dtor)(void*, struct kmem_cache *, unsigned long))
L
Linus Torvalds 已提交
2061 2062
{
	size_t left_over, slab_size, ralign;
2063
	struct kmem_cache *cachep = NULL, *pc;
L
Linus Torvalds 已提交
2064 2065 2066 2067

	/*
	 * Sanity checks... these are all serious usage bugs.
	 */
A
Andrew Morton 已提交
2068
	if (!name || in_interrupt() || (size < BYTES_PER_WORD) ||
P
Pekka Enberg 已提交
2069
	    (size > (1 << MAX_OBJ_ORDER) * PAGE_SIZE) || (dtor && !ctor)) {
A
Andrew Morton 已提交
2070 2071
		printk(KERN_ERR "%s: Early error in slab %s\n", __FUNCTION__,
				name);
P
Pekka Enberg 已提交
2072 2073
		BUG();
	}
L
Linus Torvalds 已提交
2074

2075 2076 2077 2078 2079 2080
	/*
	 * Prevent CPUs from coming and going.
	 * lock_cpu_hotplug() nests outside cache_chain_mutex
	 */
	lock_cpu_hotplug();

I
Ingo Molnar 已提交
2081
	mutex_lock(&cache_chain_mutex);
2082

2083
	list_for_each_entry(pc, &cache_chain, next) {
2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097
		mm_segment_t old_fs = get_fs();
		char tmp;
		int res;

		/*
		 * This happens when the module gets unloaded and doesn't
		 * destroy its slab cache and no-one else reuses the vmalloc
		 * area of the module.  Print a warning.
		 */
		set_fs(KERNEL_DS);
		res = __get_user(tmp, pc->name);
		set_fs(old_fs);
		if (res) {
			printk("SLAB: cache with size %d has lost its name\n",
2098
			       pc->buffer_size);
2099 2100 2101
			continue;
		}

P
Pekka Enberg 已提交
2102
		if (!strcmp(pc->name, name)) {
2103 2104 2105 2106 2107 2108
			printk("kmem_cache_create: duplicate cache %s\n", name);
			dump_stack();
			goto oops;
		}
	}

L
Linus Torvalds 已提交
2109 2110 2111 2112 2113
#if DEBUG
	WARN_ON(strchr(name, ' '));	/* It confuses parsers */
	if ((flags & SLAB_DEBUG_INITIAL) && !ctor) {
		/* No constructor, but inital state check requested */
		printk(KERN_ERR "%s: No con, but init state check "
P
Pekka Enberg 已提交
2114
		       "requested - %s\n", __FUNCTION__, name);
L
Linus Torvalds 已提交
2115 2116 2117 2118 2119 2120 2121 2122 2123
		flags &= ~SLAB_DEBUG_INITIAL;
	}
#if FORCED_DEBUG
	/*
	 * Enable redzoning and last user accounting, except for caches with
	 * large objects, if the increased size would increase the object size
	 * above the next power of two: caches with object sizes just above a
	 * power of two have a significant amount of internal fragmentation.
	 */
A
Andrew Morton 已提交
2124
	if (size < 4096 || fls(size - 1) == fls(size-1 + 3 * BYTES_PER_WORD))
P
Pekka Enberg 已提交
2125
		flags |= SLAB_RED_ZONE | SLAB_STORE_USER;
L
Linus Torvalds 已提交
2126 2127 2128 2129 2130 2131 2132 2133 2134 2135
	if (!(flags & SLAB_DESTROY_BY_RCU))
		flags |= SLAB_POISON;
#endif
	if (flags & SLAB_DESTROY_BY_RCU)
		BUG_ON(flags & SLAB_POISON);
#endif
	if (flags & SLAB_DESTROY_BY_RCU)
		BUG_ON(dtor);

	/*
A
Andrew Morton 已提交
2136 2137
	 * Always checks flags, a caller might be expecting debug support which
	 * isn't available.
L
Linus Torvalds 已提交
2138
	 */
2139
	BUG_ON(flags & ~CREATE_MASK);
L
Linus Torvalds 已提交
2140

A
Andrew Morton 已提交
2141 2142
	/*
	 * Check that size is in terms of words.  This is needed to avoid
L
Linus Torvalds 已提交
2143 2144 2145
	 * 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
2146 2147 2148
	if (size & (BYTES_PER_WORD - 1)) {
		size += (BYTES_PER_WORD - 1);
		size &= ~(BYTES_PER_WORD - 1);
L
Linus Torvalds 已提交
2149 2150
	}

A
Andrew Morton 已提交
2151 2152
	/* calculate the final buffer alignment: */

L
Linus Torvalds 已提交
2153 2154
	/* 1) arch recommendation: can be overridden for debug */
	if (flags & SLAB_HWCACHE_ALIGN) {
A
Andrew Morton 已提交
2155 2156 2157 2158
		/*
		 * 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
2159 2160
		 */
		ralign = cache_line_size();
P
Pekka Enberg 已提交
2161
		while (size <= ralign / 2)
L
Linus Torvalds 已提交
2162 2163 2164 2165
			ralign /= 2;
	} else {
		ralign = BYTES_PER_WORD;
	}
2166 2167 2168 2169 2170 2171 2172 2173 2174

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

L
Linus Torvalds 已提交
2175 2176 2177 2178
	/* 2) arch mandated alignment: disables debug if necessary */
	if (ralign < ARCH_SLAB_MINALIGN) {
		ralign = ARCH_SLAB_MINALIGN;
		if (ralign > BYTES_PER_WORD)
P
Pekka Enberg 已提交
2179
			flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
L
Linus Torvalds 已提交
2180 2181 2182 2183 2184
	}
	/* 3) caller mandated alignment: disables debug if necessary */
	if (ralign < align) {
		ralign = align;
		if (ralign > BYTES_PER_WORD)
P
Pekka Enberg 已提交
2185
			flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
L
Linus Torvalds 已提交
2186
	}
A
Andrew Morton 已提交
2187
	/*
2188
	 * 4) Store it.
L
Linus Torvalds 已提交
2189 2190 2191 2192
	 */
	align = ralign;

	/* Get cache's description obj. */
P
Pekka Enberg 已提交
2193
	cachep = kmem_cache_zalloc(&cache_cache, SLAB_KERNEL);
L
Linus Torvalds 已提交
2194
	if (!cachep)
2195
		goto oops;
L
Linus Torvalds 已提交
2196 2197

#if DEBUG
2198
	cachep->obj_size = size;
L
Linus Torvalds 已提交
2199

2200 2201 2202 2203
	/*
	 * Both debugging options require word-alignment which is calculated
	 * into align above.
	 */
L
Linus Torvalds 已提交
2204 2205
	if (flags & SLAB_RED_ZONE) {
		/* add space for red zone words */
2206
		cachep->obj_offset += BYTES_PER_WORD;
P
Pekka Enberg 已提交
2207
		size += 2 * BYTES_PER_WORD;
L
Linus Torvalds 已提交
2208 2209
	}
	if (flags & SLAB_STORE_USER) {
2210 2211
		/* user store requires one word storage behind the end of
		 * the real object.
L
Linus Torvalds 已提交
2212 2213 2214 2215
		 */
		size += BYTES_PER_WORD;
	}
#if FORCED_DEBUG && defined(CONFIG_DEBUG_PAGEALLOC)
P
Pekka Enberg 已提交
2216
	if (size >= malloc_sizes[INDEX_L3 + 1].cs_size
2217 2218
	    && cachep->obj_size > cache_line_size() && size < PAGE_SIZE) {
		cachep->obj_offset += PAGE_SIZE - size;
L
Linus Torvalds 已提交
2219 2220 2221 2222 2223
		size = PAGE_SIZE;
	}
#endif
#endif

2224 2225 2226 2227 2228 2229
	/*
	 * Determine if the slab management is 'on' or 'off' slab.
	 * (bootstrapping cannot cope with offslab caches so don't do
	 * it too early on.)
	 */
	if ((size >= (PAGE_SIZE >> 3)) && !slab_early_init)
L
Linus Torvalds 已提交
2230 2231 2232 2233 2234 2235 2236 2237
		/*
		 * 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);

2238
	left_over = calculate_slab_order(cachep, size, align, flags);
L
Linus Torvalds 已提交
2239 2240 2241 2242 2243

	if (!cachep->num) {
		printk("kmem_cache_create: couldn't create cache %s.\n", name);
		kmem_cache_free(&cache_cache, cachep);
		cachep = NULL;
2244
		goto oops;
L
Linus Torvalds 已提交
2245
	}
P
Pekka Enberg 已提交
2246 2247
	slab_size = ALIGN(cachep->num * sizeof(kmem_bufctl_t)
			  + sizeof(struct slab), align);
L
Linus Torvalds 已提交
2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259

	/*
	 * 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
2260 2261
		slab_size =
		    cachep->num * sizeof(kmem_bufctl_t) + sizeof(struct slab);
L
Linus Torvalds 已提交
2262 2263 2264 2265 2266 2267
	}

	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
2268
	cachep->colour = left_over / cachep->colour_off;
L
Linus Torvalds 已提交
2269 2270 2271 2272 2273
	cachep->slab_size = slab_size;
	cachep->flags = flags;
	cachep->gfpflags = 0;
	if (flags & SLAB_CACHE_DMA)
		cachep->gfpflags |= GFP_DMA;
2274
	cachep->buffer_size = size;
L
Linus Torvalds 已提交
2275

2276
	if (flags & CFLGS_OFF_SLAB) {
2277
		cachep->slabp_cache = kmem_find_general_cachep(slab_size, 0u);
2278 2279 2280 2281 2282 2283 2284 2285 2286
		/*
		 * This is a possibility for one of the malloc_sizes caches.
		 * But since we go off slab only for object size greater than
		 * PAGE_SIZE/8, and malloc_sizes gets created in ascending order,
		 * this should not happen at all.
		 * But leave a BUG_ON for some lucky dude.
		 */
		BUG_ON(!cachep->slabp_cache);
	}
L
Linus Torvalds 已提交
2287 2288 2289 2290
	cachep->ctor = ctor;
	cachep->dtor = dtor;
	cachep->name = name;

2291 2292 2293 2294 2295
	if (setup_cpu_cache(cachep)) {
		__kmem_cache_destroy(cachep);
		cachep = NULL;
		goto oops;
	}
L
Linus Torvalds 已提交
2296 2297 2298

	/* cache setup completed, link it into the list */
	list_add(&cachep->next, &cache_chain);
A
Andrew Morton 已提交
2299
oops:
L
Linus Torvalds 已提交
2300 2301
	if (!cachep && (flags & SLAB_PANIC))
		panic("kmem_cache_create(): failed to create slab `%s'\n",
P
Pekka Enberg 已提交
2302
		      name);
I
Ingo Molnar 已提交
2303
	mutex_unlock(&cache_chain_mutex);
2304
	unlock_cpu_hotplug();
L
Linus Torvalds 已提交
2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319
	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());
}

2320
static void check_spinlock_acquired(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2321 2322 2323
{
#ifdef CONFIG_SMP
	check_irq_off();
2324
	assert_spin_locked(&cachep->nodelists[numa_node_id()]->list_lock);
L
Linus Torvalds 已提交
2325 2326
#endif
}
2327

2328
static void check_spinlock_acquired_node(struct kmem_cache *cachep, int node)
2329 2330 2331 2332 2333 2334 2335
{
#ifdef CONFIG_SMP
	check_irq_off();
	assert_spin_locked(&cachep->nodelists[node]->list_lock);
#endif
}

L
Linus Torvalds 已提交
2336 2337 2338 2339
#else
#define check_irq_off()	do { } while(0)
#define check_irq_on()	do { } while(0)
#define check_spinlock_acquired(x) do { } while(0)
2340
#define check_spinlock_acquired_node(x, y) do { } while(0)
L
Linus Torvalds 已提交
2341 2342
#endif

2343 2344 2345 2346
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
2347 2348
static void do_drain(void *arg)
{
A
Andrew Morton 已提交
2349
	struct kmem_cache *cachep = arg;
L
Linus Torvalds 已提交
2350
	struct array_cache *ac;
2351
	int node = numa_node_id();
L
Linus Torvalds 已提交
2352 2353

	check_irq_off();
2354
	ac = cpu_cache_get(cachep);
2355 2356 2357
	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
2358 2359 2360
	ac->avail = 0;
}

2361
static void drain_cpu_caches(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2362
{
2363 2364 2365
	struct kmem_list3 *l3;
	int node;

A
Andrew Morton 已提交
2366
	on_each_cpu(do_drain, cachep, 1, 1);
L
Linus Torvalds 已提交
2367
	check_irq_on();
P
Pekka Enberg 已提交
2368
	for_each_online_node(node) {
2369
		l3 = cachep->nodelists[node];
2370 2371 2372 2373 2374 2375 2376
		if (l3 && l3->alien)
			drain_alien_cache(cachep, l3->alien);
	}

	for_each_online_node(node) {
		l3 = cachep->nodelists[node];
		if (l3)
2377
			drain_array(cachep, l3, l3->shared, 1, node);
2378
	}
L
Linus Torvalds 已提交
2379 2380
}

2381 2382 2383 2384 2385 2386 2387 2388
/*
 * 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
2389
{
2390 2391
	struct list_head *p;
	int nr_freed;
L
Linus Torvalds 已提交
2392 2393
	struct slab *slabp;

2394 2395
	nr_freed = 0;
	while (nr_freed < tofree && !list_empty(&l3->slabs_free)) {
L
Linus Torvalds 已提交
2396

2397
		spin_lock_irq(&l3->list_lock);
2398
		p = l3->slabs_free.prev;
2399 2400 2401 2402
		if (p == &l3->slabs_free) {
			spin_unlock_irq(&l3->list_lock);
			goto out;
		}
L
Linus Torvalds 已提交
2403

2404
		slabp = list_entry(p, struct slab, list);
L
Linus Torvalds 已提交
2405
#if DEBUG
2406
		BUG_ON(slabp->inuse);
L
Linus Torvalds 已提交
2407 2408
#endif
		list_del(&slabp->list);
2409 2410 2411 2412 2413
		/*
		 * Safe to drop the lock. The slab is no longer linked
		 * to the cache.
		 */
		l3->free_objects -= cache->num;
2414
		spin_unlock_irq(&l3->list_lock);
2415 2416
		slab_destroy(cache, slabp);
		nr_freed++;
L
Linus Torvalds 已提交
2417
	}
2418 2419
out:
	return nr_freed;
L
Linus Torvalds 已提交
2420 2421
}

2422
static int __cache_shrink(struct kmem_cache *cachep)
2423 2424 2425 2426 2427 2428 2429 2430 2431
{
	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];
2432 2433 2434 2435 2436 2437 2438
		if (!l3)
			continue;

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

		ret += !list_empty(&l3->slabs_full) ||
			!list_empty(&l3->slabs_partial);
2439 2440 2441 2442
	}
	return (ret ? 1 : 0);
}

L
Linus Torvalds 已提交
2443 2444 2445 2446 2447 2448 2449
/**
 * 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.
 */
2450
int kmem_cache_shrink(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2451
{
2452
	BUG_ON(!cachep || in_interrupt());
L
Linus Torvalds 已提交
2453 2454 2455 2456 2457 2458 2459 2460 2461

	return __cache_shrink(cachep);
}
EXPORT_SYMBOL(kmem_cache_shrink);

/**
 * kmem_cache_destroy - delete a cache
 * @cachep: the cache to destroy
 *
2462
 * Remove a struct kmem_cache object from the slab cache.
L
Linus Torvalds 已提交
2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473
 *
 * It is expected this function will be called by a module when it is
 * unloaded.  This will remove the cache completely, and avoid a duplicate
 * cache being allocated each time a module is loaded and unloaded, if the
 * module doesn't have persistent in-kernel storage across loads and unloads.
 *
 * The cache must be empty before calling this function.
 *
 * The caller must guarantee that noone will allocate memory from the cache
 * during the kmem_cache_destroy().
 */
2474
void kmem_cache_destroy(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
2475
{
2476
	BUG_ON(!cachep || in_interrupt());
L
Linus Torvalds 已提交
2477 2478 2479 2480 2481

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

	/* Find the cache in the chain of caches. */
I
Ingo Molnar 已提交
2482
	mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2483 2484 2485 2486
	/*
	 * the chain is never empty, cache_cache is never destroyed
	 */
	list_del(&cachep->next);
I
Ingo Molnar 已提交
2487
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2488 2489 2490

	if (__cache_shrink(cachep)) {
		slab_error(cachep, "Can't free all objects");
I
Ingo Molnar 已提交
2491
		mutex_lock(&cache_chain_mutex);
P
Pekka Enberg 已提交
2492
		list_add(&cachep->next, &cache_chain);
I
Ingo Molnar 已提交
2493
		mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
2494
		unlock_cpu_hotplug();
2495
		return;
L
Linus Torvalds 已提交
2496 2497 2498
	}

	if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU))
2499
		synchronize_rcu();
L
Linus Torvalds 已提交
2500

2501
	__kmem_cache_destroy(cachep);
L
Linus Torvalds 已提交
2502 2503 2504 2505
	unlock_cpu_hotplug();
}
EXPORT_SYMBOL(kmem_cache_destroy);

2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516
/*
 * 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.
 */
2517
static struct slab *alloc_slabmgmt(struct kmem_cache *cachep, void *objp,
2518 2519
				   int colour_off, gfp_t local_flags,
				   int nodeid)
L
Linus Torvalds 已提交
2520 2521
{
	struct slab *slabp;
P
Pekka Enberg 已提交
2522

L
Linus Torvalds 已提交
2523 2524
	if (OFF_SLAB(cachep)) {
		/* Slab management obj is off-slab. */
2525 2526
		slabp = kmem_cache_alloc_node(cachep->slabp_cache,
					      local_flags, nodeid);
L
Linus Torvalds 已提交
2527 2528 2529
		if (!slabp)
			return NULL;
	} else {
P
Pekka Enberg 已提交
2530
		slabp = objp + colour_off;
L
Linus Torvalds 已提交
2531 2532 2533 2534
		colour_off += cachep->slab_size;
	}
	slabp->inuse = 0;
	slabp->colouroff = colour_off;
P
Pekka Enberg 已提交
2535
	slabp->s_mem = objp + colour_off;
2536
	slabp->nodeid = nodeid;
L
Linus Torvalds 已提交
2537 2538 2539 2540 2541
	return slabp;
}

static inline kmem_bufctl_t *slab_bufctl(struct slab *slabp)
{
P
Pekka Enberg 已提交
2542
	return (kmem_bufctl_t *) (slabp + 1);
L
Linus Torvalds 已提交
2543 2544
}

2545
static void cache_init_objs(struct kmem_cache *cachep,
P
Pekka Enberg 已提交
2546
			    struct slab *slabp, unsigned long ctor_flags)
L
Linus Torvalds 已提交
2547 2548 2549 2550
{
	int i;

	for (i = 0; i < cachep->num; i++) {
2551
		void *objp = index_to_obj(cachep, slabp, i);
L
Linus Torvalds 已提交
2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563
#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
2564 2565 2566
		 * 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
2567 2568
		 */
		if (cachep->ctor && !(cachep->flags & SLAB_POISON))
2569
			cachep->ctor(objp + obj_offset(cachep), cachep,
P
Pekka Enberg 已提交
2570
				     ctor_flags);
L
Linus Torvalds 已提交
2571 2572 2573 2574

		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
2575
					   " end of an object");
L
Linus Torvalds 已提交
2576 2577
			if (*dbg_redzone1(cachep, objp) != RED_INACTIVE)
				slab_error(cachep, "constructor overwrote the"
P
Pekka Enberg 已提交
2578
					   " start of an object");
L
Linus Torvalds 已提交
2579
		}
A
Andrew Morton 已提交
2580 2581
		if ((cachep->buffer_size % PAGE_SIZE) == 0 &&
			    OFF_SLAB(cachep) && cachep->flags & SLAB_POISON)
P
Pekka Enberg 已提交
2582
			kernel_map_pages(virt_to_page(objp),
2583
					 cachep->buffer_size / PAGE_SIZE, 0);
L
Linus Torvalds 已提交
2584 2585 2586 2587
#else
		if (cachep->ctor)
			cachep->ctor(objp, cachep, ctor_flags);
#endif
P
Pekka Enberg 已提交
2588
		slab_bufctl(slabp)[i] = i + 1;
L
Linus Torvalds 已提交
2589
	}
P
Pekka Enberg 已提交
2590
	slab_bufctl(slabp)[i - 1] = BUFCTL_END;
L
Linus Torvalds 已提交
2591 2592 2593
	slabp->free = 0;
}

2594
static void kmem_flagcheck(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2595
{
A
Andrew Morton 已提交
2596 2597 2598 2599
	if (flags & SLAB_DMA)
		BUG_ON(!(cachep->gfpflags & GFP_DMA));
	else
		BUG_ON(cachep->gfpflags & GFP_DMA);
L
Linus Torvalds 已提交
2600 2601
}

A
Andrew Morton 已提交
2602 2603
static void *slab_get_obj(struct kmem_cache *cachep, struct slab *slabp,
				int nodeid)
2604
{
2605
	void *objp = index_to_obj(cachep, slabp, slabp->free);
2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618
	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
2619 2620
static void slab_put_obj(struct kmem_cache *cachep, struct slab *slabp,
				void *objp, int nodeid)
2621
{
2622
	unsigned int objnr = obj_to_index(cachep, slabp, objp);
2623 2624 2625 2626 2627

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

2628
	if (slab_bufctl(slabp)[objnr] + 1 <= SLAB_LIMIT + 1) {
2629
		printk(KERN_ERR "slab: double free detected in cache "
A
Andrew Morton 已提交
2630
				"'%s', objp %p\n", cachep->name, objp);
2631 2632 2633 2634 2635 2636 2637 2638
		BUG();
	}
#endif
	slab_bufctl(slabp)[objnr] = slabp->free;
	slabp->free = objnr;
	slabp->inuse--;
}

2639 2640 2641 2642 2643 2644 2645
/*
 * Map pages beginning at addr to the given cache and slab. This is required
 * for the slab allocator to be able to lookup the cache and slab of a
 * virtual address for kfree, ksize, kmem_ptr_validate, and slab debugging.
 */
static void slab_map_pages(struct kmem_cache *cache, struct slab *slab,
			   void *addr)
L
Linus Torvalds 已提交
2646
{
2647
	int nr_pages;
L
Linus Torvalds 已提交
2648 2649
	struct page *page;

2650
	page = virt_to_page(addr);
2651

2652
	nr_pages = 1;
2653
	if (likely(!PageCompound(page)))
2654 2655
		nr_pages <<= cache->gfporder;

L
Linus Torvalds 已提交
2656
	do {
2657 2658
		page_set_cache(page, cache);
		page_set_slab(page, slab);
L
Linus Torvalds 已提交
2659
		page++;
2660
	} while (--nr_pages);
L
Linus Torvalds 已提交
2661 2662 2663 2664 2665 2666
}

/*
 * 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.
 */
2667
static int cache_grow(struct kmem_cache *cachep, gfp_t flags, int nodeid)
L
Linus Torvalds 已提交
2668
{
P
Pekka Enberg 已提交
2669 2670 2671 2672 2673
	struct slab *slabp;
	void *objp;
	size_t offset;
	gfp_t local_flags;
	unsigned long ctor_flags;
2674
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
2675

A
Andrew Morton 已提交
2676 2677 2678
	/*
	 * 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
2679
	 */
2680
	BUG_ON(flags & ~(SLAB_DMA | SLAB_LEVEL_MASK | SLAB_NO_GROW));
L
Linus Torvalds 已提交
2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692
	if (flags & SLAB_NO_GROW)
		return 0;

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

2693
	/* Take the l3 list lock to change the colour_next on this node */
L
Linus Torvalds 已提交
2694
	check_irq_off();
2695 2696
	l3 = cachep->nodelists[nodeid];
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2697 2698

	/* Get colour for the slab, and cal the next value. */
2699 2700 2701 2702 2703
	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
2704

2705
	offset *= cachep->colour_off;
L
Linus Torvalds 已提交
2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717

	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
2718 2719 2720
	/*
	 * Get mem for the objs.  Attempt to allocate a physical page from
	 * 'nodeid'.
2721
	 */
A
Andrew Morton 已提交
2722 2723
	objp = kmem_getpages(cachep, flags, nodeid);
	if (!objp)
L
Linus Torvalds 已提交
2724 2725 2726
		goto failed;

	/* Get slab management. */
2727
	slabp = alloc_slabmgmt(cachep, objp, offset, local_flags, nodeid);
A
Andrew Morton 已提交
2728
	if (!slabp)
L
Linus Torvalds 已提交
2729 2730
		goto opps1;

2731
	slabp->nodeid = nodeid;
2732
	slab_map_pages(cachep, slabp, objp);
L
Linus Torvalds 已提交
2733 2734 2735 2736 2737 2738

	cache_init_objs(cachep, slabp, ctor_flags);

	if (local_flags & __GFP_WAIT)
		local_irq_disable();
	check_irq_off();
2739
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2740 2741

	/* Make slab active. */
2742
	list_add_tail(&slabp->list, &(l3->slabs_free));
L
Linus Torvalds 已提交
2743
	STATS_INC_GROWN(cachep);
2744 2745
	l3->free_objects += cachep->num;
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2746
	return 1;
A
Andrew Morton 已提交
2747
opps1:
L
Linus Torvalds 已提交
2748
	kmem_freepages(cachep, objp);
A
Andrew Morton 已提交
2749
failed:
L
Linus Torvalds 已提交
2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768
	if (local_flags & __GFP_WAIT)
		local_irq_disable();
	return 0;
}

#if DEBUG

/*
 * Perform extra freeing checks:
 * - detect bad pointers.
 * - POISON/RED_ZONE checking
 * - destructor calls, for caches with POISON+dtor
 */
static void kfree_debugcheck(const void *objp)
{
	struct page *page;

	if (!virt_addr_valid(objp)) {
		printk(KERN_ERR "kfree_debugcheck: out of range ptr %lxh.\n",
P
Pekka Enberg 已提交
2769 2770
		       (unsigned long)objp);
		BUG();
L
Linus Torvalds 已提交
2771 2772 2773
	}
	page = virt_to_page(objp);
	if (!PageSlab(page)) {
P
Pekka Enberg 已提交
2774 2775
		printk(KERN_ERR "kfree_debugcheck: bad ptr %lxh.\n",
		       (unsigned long)objp);
L
Linus Torvalds 已提交
2776 2777 2778 2779
		BUG();
	}
}

2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801
static inline void verify_redzone_free(struct kmem_cache *cache, void *obj)
{
	unsigned long redzone1, redzone2;

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

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

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

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

2802
static void *cache_free_debugcheck(struct kmem_cache *cachep, void *objp,
P
Pekka Enberg 已提交
2803
				   void *caller)
L
Linus Torvalds 已提交
2804 2805 2806 2807 2808
{
	struct page *page;
	unsigned int objnr;
	struct slab *slabp;

2809
	objp -= obj_offset(cachep);
L
Linus Torvalds 已提交
2810 2811 2812
	kfree_debugcheck(objp);
	page = virt_to_page(objp);

2813
	slabp = page_get_slab(page);
L
Linus Torvalds 已提交
2814 2815

	if (cachep->flags & SLAB_RED_ZONE) {
2816
		verify_redzone_free(cachep, objp);
L
Linus Torvalds 已提交
2817 2818 2819 2820 2821 2822
		*dbg_redzone1(cachep, objp) = RED_INACTIVE;
		*dbg_redzone2(cachep, objp) = RED_INACTIVE;
	}
	if (cachep->flags & SLAB_STORE_USER)
		*dbg_userword(cachep, objp) = caller;

2823
	objnr = obj_to_index(cachep, slabp, objp);
L
Linus Torvalds 已提交
2824 2825

	BUG_ON(objnr >= cachep->num);
2826
	BUG_ON(objp != index_to_obj(cachep, slabp, objnr));
L
Linus Torvalds 已提交
2827 2828

	if (cachep->flags & SLAB_DEBUG_INITIAL) {
A
Andrew Morton 已提交
2829 2830 2831 2832
		/*
		 * Need to call the slab's constructor so the caller can
		 * perform a verify of its state (debugging).  Called without
		 * the cache-lock held.
L
Linus Torvalds 已提交
2833
		 */
2834
		cachep->ctor(objp + obj_offset(cachep),
P
Pekka Enberg 已提交
2835
			     cachep, SLAB_CTOR_CONSTRUCTOR | SLAB_CTOR_VERIFY);
L
Linus Torvalds 已提交
2836 2837 2838 2839 2840
	}
	if (cachep->flags & SLAB_POISON && cachep->dtor) {
		/* we want to cache poison the object,
		 * call the destruction callback
		 */
2841
		cachep->dtor(objp + obj_offset(cachep), cachep, 0);
L
Linus Torvalds 已提交
2842
	}
2843 2844 2845
#ifdef CONFIG_DEBUG_SLAB_LEAK
	slab_bufctl(slabp)[objnr] = BUFCTL_FREE;
#endif
L
Linus Torvalds 已提交
2846 2847
	if (cachep->flags & SLAB_POISON) {
#ifdef CONFIG_DEBUG_PAGEALLOC
A
Andrew Morton 已提交
2848
		if ((cachep->buffer_size % PAGE_SIZE)==0 && OFF_SLAB(cachep)) {
L
Linus Torvalds 已提交
2849
			store_stackinfo(cachep, objp, (unsigned long)caller);
P
Pekka Enberg 已提交
2850
			kernel_map_pages(virt_to_page(objp),
2851
					 cachep->buffer_size / PAGE_SIZE, 0);
L
Linus Torvalds 已提交
2852 2853 2854 2855 2856 2857 2858 2859 2860 2861
		} else {
			poison_obj(cachep, objp, POISON_FREE);
		}
#else
		poison_obj(cachep, objp, POISON_FREE);
#endif
	}
	return objp;
}

2862
static void check_slabp(struct kmem_cache *cachep, struct slab *slabp)
L
Linus Torvalds 已提交
2863 2864 2865
{
	kmem_bufctl_t i;
	int entries = 0;
P
Pekka Enberg 已提交
2866

L
Linus Torvalds 已提交
2867 2868 2869 2870 2871 2872 2873
	/* 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
2874 2875 2876 2877
bad:
		printk(KERN_ERR "slab: Internal list corruption detected in "
				"cache '%s'(%d), slabp %p(%d). Hexdump:\n",
			cachep->name, cachep->num, slabp, slabp->inuse);
P
Pekka Enberg 已提交
2878
		for (i = 0;
2879
		     i < sizeof(*slabp) + cachep->num * sizeof(kmem_bufctl_t);
P
Pekka Enberg 已提交
2880
		     i++) {
A
Andrew Morton 已提交
2881
			if (i % 16 == 0)
L
Linus Torvalds 已提交
2882
				printk("\n%03x:", i);
P
Pekka Enberg 已提交
2883
			printk(" %02x", ((unsigned char *)slabp)[i]);
L
Linus Torvalds 已提交
2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894
		}
		printk("\n");
		BUG();
	}
}
#else
#define kfree_debugcheck(x) do { } while(0)
#define cache_free_debugcheck(x,objp,z) (objp)
#define check_slabp(x,y) do { } while(0)
#endif

2895
static void *cache_alloc_refill(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
2896 2897 2898 2899 2900 2901
{
	int batchcount;
	struct kmem_list3 *l3;
	struct array_cache *ac;

	check_irq_off();
2902
	ac = cpu_cache_get(cachep);
A
Andrew Morton 已提交
2903
retry:
L
Linus Torvalds 已提交
2904 2905
	batchcount = ac->batchcount;
	if (!ac->touched && batchcount > BATCHREFILL_LIMIT) {
A
Andrew Morton 已提交
2906 2907 2908 2909
		/*
		 * 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
2910 2911 2912
		 */
		batchcount = BATCHREFILL_LIMIT;
	}
2913 2914 2915 2916
	l3 = cachep->nodelists[numa_node_id()];

	BUG_ON(ac->avail > 0 || !l3);
	spin_lock(&l3->list_lock);
L
Linus Torvalds 已提交
2917

2918 2919 2920 2921
	/* See if we can refill from the shared array */
	if (l3->shared && transfer_objects(ac, l3->shared, batchcount))
		goto alloc_done;

L
Linus Torvalds 已提交
2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941
	while (batchcount > 0) {
		struct list_head *entry;
		struct slab *slabp;
		/* Get slab alloc is to come from. */
		entry = l3->slabs_partial.next;
		if (entry == &l3->slabs_partial) {
			l3->free_touched = 1;
			entry = l3->slabs_free.next;
			if (entry == &l3->slabs_free)
				goto must_grow;
		}

		slabp = list_entry(entry, struct slab, list);
		check_slabp(cachep, slabp);
		check_spinlock_acquired(cachep);
		while (slabp->inuse < cachep->num && batchcount--) {
			STATS_INC_ALLOCED(cachep);
			STATS_INC_ACTIVE(cachep);
			STATS_SET_HIGH(cachep);

2942 2943
			ac->entry[ac->avail++] = slab_get_obj(cachep, slabp,
							    numa_node_id());
L
Linus Torvalds 已提交
2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954
		}
		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
2955
must_grow:
L
Linus Torvalds 已提交
2956
	l3->free_objects -= ac->avail;
A
Andrew Morton 已提交
2957
alloc_done:
2958
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
2959 2960 2961

	if (unlikely(!ac->avail)) {
		int x;
2962 2963
		x = cache_grow(cachep, flags, numa_node_id());

A
Andrew Morton 已提交
2964
		/* cache_grow can reenable interrupts, then ac could change. */
2965
		ac = cpu_cache_get(cachep);
A
Andrew Morton 已提交
2966
		if (!x && ac->avail == 0)	/* no objects in sight? abort */
L
Linus Torvalds 已提交
2967 2968
			return NULL;

A
Andrew Morton 已提交
2969
		if (!ac->avail)		/* objects refilled by interrupt? */
L
Linus Torvalds 已提交
2970 2971 2972
			goto retry;
	}
	ac->touched = 1;
2973
	return ac->entry[--ac->avail];
L
Linus Torvalds 已提交
2974 2975
}

A
Andrew Morton 已提交
2976 2977
static inline void cache_alloc_debugcheck_before(struct kmem_cache *cachep,
						gfp_t flags)
L
Linus Torvalds 已提交
2978 2979 2980 2981 2982 2983 2984 2985
{
	might_sleep_if(flags & __GFP_WAIT);
#if DEBUG
	kmem_flagcheck(cachep, flags);
#endif
}

#if DEBUG
A
Andrew Morton 已提交
2986 2987
static void *cache_alloc_debugcheck_after(struct kmem_cache *cachep,
				gfp_t flags, void *objp, void *caller)
L
Linus Torvalds 已提交
2988
{
P
Pekka Enberg 已提交
2989
	if (!objp)
L
Linus Torvalds 已提交
2990
		return objp;
P
Pekka Enberg 已提交
2991
	if (cachep->flags & SLAB_POISON) {
L
Linus Torvalds 已提交
2992
#ifdef CONFIG_DEBUG_PAGEALLOC
2993
		if ((cachep->buffer_size % PAGE_SIZE) == 0 && OFF_SLAB(cachep))
P
Pekka Enberg 已提交
2994
			kernel_map_pages(virt_to_page(objp),
2995
					 cachep->buffer_size / PAGE_SIZE, 1);
L
Linus Torvalds 已提交
2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006
		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
3007 3008 3009 3010
		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
3011
			printk(KERN_ERR
A
Andrew Morton 已提交
3012 3013 3014
				"%p: redzone 1:0x%lx, redzone 2:0x%lx\n",
				objp, *dbg_redzone1(cachep, objp),
				*dbg_redzone2(cachep, objp));
L
Linus Torvalds 已提交
3015 3016 3017 3018
		}
		*dbg_redzone1(cachep, objp) = RED_ACTIVE;
		*dbg_redzone2(cachep, objp) = RED_ACTIVE;
	}
3019 3020 3021 3022 3023 3024 3025 3026 3027 3028
#ifdef CONFIG_DEBUG_SLAB_LEAK
	{
		struct slab *slabp;
		unsigned objnr;

		slabp = page_get_slab(virt_to_page(objp));
		objnr = (unsigned)(objp - slabp->s_mem) / cachep->buffer_size;
		slab_bufctl(slabp)[objnr] = BUFCTL_ACTIVE;
	}
#endif
3029
	objp += obj_offset(cachep);
L
Linus Torvalds 已提交
3030
	if (cachep->ctor && cachep->flags & SLAB_POISON) {
P
Pekka Enberg 已提交
3031
		unsigned long ctor_flags = SLAB_CTOR_CONSTRUCTOR;
L
Linus Torvalds 已提交
3032 3033 3034 3035 3036

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

		cachep->ctor(objp, cachep, ctor_flags);
P
Pekka Enberg 已提交
3037
	}
L
Linus Torvalds 已提交
3038 3039 3040 3041 3042 3043
	return objp;
}
#else
#define cache_alloc_debugcheck_after(a,b,objp,d) (objp)
#endif

3044
static inline void *____cache_alloc(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
3045
{
P
Pekka Enberg 已提交
3046
	void *objp;
L
Linus Torvalds 已提交
3047 3048
	struct array_cache *ac;

3049
	check_irq_off();
3050
	ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
3051 3052 3053
	if (likely(ac->avail)) {
		STATS_INC_ALLOCHIT(cachep);
		ac->touched = 1;
3054
		objp = ac->entry[--ac->avail];
L
Linus Torvalds 已提交
3055 3056 3057 3058
	} else {
		STATS_INC_ALLOCMISS(cachep);
		objp = cache_alloc_refill(cachep, flags);
	}
3059 3060 3061
	return objp;
}

A
Andrew Morton 已提交
3062 3063
static __always_inline void *__cache_alloc(struct kmem_cache *cachep,
						gfp_t flags, void *caller)
3064 3065
{
	unsigned long save_flags;
3066
	void *objp = NULL;
3067 3068 3069 3070

	cache_alloc_debugcheck_before(cachep, flags);

	local_irq_save(save_flags);
3071

3072 3073
	if (unlikely(NUMA_BUILD &&
			current->flags & (PF_SPREAD_SLAB | PF_MEMPOLICY)))
3074 3075 3076 3077
		objp = alternate_node_alloc(cachep, flags);

	if (!objp)
		objp = ____cache_alloc(cachep, flags);
3078 3079 3080 3081 3082 3083
	/*
	 * We may just have run out of memory on the local node.
	 * __cache_alloc_node() knows how to locate memory on other nodes
	 */
 	if (NUMA_BUILD && !objp)
 		objp = __cache_alloc_node(cachep, flags, numa_node_id());
L
Linus Torvalds 已提交
3084
	local_irq_restore(save_flags);
3085
	objp = cache_alloc_debugcheck_after(cachep, flags, objp,
3086
					    caller);
3087
	prefetchw(objp);
L
Linus Torvalds 已提交
3088 3089 3090
	return objp;
}

3091
#ifdef CONFIG_NUMA
3092
/*
3093
 * Try allocating on another node if PF_SPREAD_SLAB|PF_MEMPOLICY.
3094 3095 3096 3097 3098 3099 3100 3101
 *
 * 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;

3102
	if (in_interrupt() || (flags & __GFP_THISNODE))
3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113
		return NULL;
	nid_alloc = nid_here = numa_node_id();
	if (cpuset_do_slab_mem_spread() && (cachep->flags & SLAB_MEM_SPREAD))
		nid_alloc = cpuset_mem_spread_node();
	else if (current->mempolicy)
		nid_alloc = slab_node(current->mempolicy);
	if (nid_alloc != nid_here)
		return __cache_alloc_node(cachep, flags, nid_alloc);
	return NULL;
}

3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135
/*
 * Fallback function if there was no memory available and no objects on a
 * certain node and we are allowed to fall back. We mimick the behavior of
 * the page allocator. We fall back according to a zonelist determined by
 * the policy layer while obeying cpuset constraints.
 */
void *fallback_alloc(struct kmem_cache *cache, gfp_t flags)
{
	struct zonelist *zonelist = &NODE_DATA(slab_node(current->mempolicy))
					->node_zonelists[gfp_zone(flags)];
	struct zone **z;
	void *obj = NULL;

	for (z = zonelist->zones; *z && !obj; z++)
		if (zone_idx(*z) <= ZONE_NORMAL &&
				cpuset_zone_allowed(*z, flags))
			obj = __cache_alloc_node(cache,
					flags | __GFP_THISNODE,
					zone_to_nid(*z));
	return obj;
}

3136 3137
/*
 * A interface to enable slab creation on nodeid
L
Linus Torvalds 已提交
3138
 */
A
Andrew Morton 已提交
3139 3140
static void *__cache_alloc_node(struct kmem_cache *cachep, gfp_t flags,
				int nodeid)
3141 3142
{
	struct list_head *entry;
P
Pekka Enberg 已提交
3143 3144 3145 3146 3147 3148 3149 3150
	struct slab *slabp;
	struct kmem_list3 *l3;
	void *obj;
	int x;

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

A
Andrew Morton 已提交
3151
retry:
3152
	check_irq_off();
P
Pekka Enberg 已提交
3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171
	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);

3172
	obj = slab_get_obj(cachep, slabp, nodeid);
P
Pekka Enberg 已提交
3173 3174 3175 3176 3177
	check_slabp(cachep, slabp);
	l3->free_objects--;
	/* move slabp to correct slabp list: */
	list_del(&slabp->list);

A
Andrew Morton 已提交
3178
	if (slabp->free == BUFCTL_END)
P
Pekka Enberg 已提交
3179
		list_add(&slabp->list, &l3->slabs_full);
A
Andrew Morton 已提交
3180
	else
P
Pekka Enberg 已提交
3181
		list_add(&slabp->list, &l3->slabs_partial);
3182

P
Pekka Enberg 已提交
3183 3184
	spin_unlock(&l3->list_lock);
	goto done;
3185

A
Andrew Morton 已提交
3186
must_grow:
P
Pekka Enberg 已提交
3187 3188
	spin_unlock(&l3->list_lock);
	x = cache_grow(cachep, flags, nodeid);
3189 3190
	if (x)
		goto retry;
L
Linus Torvalds 已提交
3191

3192 3193 3194 3195 3196
	if (!(flags & __GFP_THISNODE))
		/* Unable to grow the cache. Fall back to other nodes. */
		return fallback_alloc(cachep, flags);

	return NULL;
3197

A
Andrew Morton 已提交
3198
done:
P
Pekka Enberg 已提交
3199
	return obj;
3200 3201 3202 3203 3204 3205
}
#endif

/*
 * Caller needs to acquire correct kmem_list's list_lock
 */
3206
static void free_block(struct kmem_cache *cachep, void **objpp, int nr_objects,
P
Pekka Enberg 已提交
3207
		       int node)
L
Linus Torvalds 已提交
3208 3209
{
	int i;
3210
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
3211 3212 3213 3214 3215

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

3216
		slabp = virt_to_slab(objp);
3217
		l3 = cachep->nodelists[node];
L
Linus Torvalds 已提交
3218
		list_del(&slabp->list);
3219
		check_spinlock_acquired_node(cachep, node);
L
Linus Torvalds 已提交
3220
		check_slabp(cachep, slabp);
3221
		slab_put_obj(cachep, slabp, objp, node);
L
Linus Torvalds 已提交
3222
		STATS_DEC_ACTIVE(cachep);
3223
		l3->free_objects++;
L
Linus Torvalds 已提交
3224 3225 3226 3227
		check_slabp(cachep, slabp);

		/* fixup slab chains */
		if (slabp->inuse == 0) {
3228 3229
			if (l3->free_objects > l3->free_limit) {
				l3->free_objects -= cachep->num;
3230 3231 3232 3233 3234 3235
				/* 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
3236 3237
				slab_destroy(cachep, slabp);
			} else {
3238
				list_add(&slabp->list, &l3->slabs_free);
L
Linus Torvalds 已提交
3239 3240 3241 3242 3243 3244
			}
		} else {
			/* Unconditionally move a slab to the end of the
			 * partial list on free - maximum time for the
			 * other objects to be freed, too.
			 */
3245
			list_add_tail(&slabp->list, &l3->slabs_partial);
L
Linus Torvalds 已提交
3246 3247 3248 3249
		}
	}
}

3250
static void cache_flusharray(struct kmem_cache *cachep, struct array_cache *ac)
L
Linus Torvalds 已提交
3251 3252
{
	int batchcount;
3253
	struct kmem_list3 *l3;
3254
	int node = numa_node_id();
L
Linus Torvalds 已提交
3255 3256 3257 3258 3259 3260

	batchcount = ac->batchcount;
#if DEBUG
	BUG_ON(!batchcount || batchcount > ac->avail);
#endif
	check_irq_off();
3261
	l3 = cachep->nodelists[node];
3262
	spin_lock(&l3->list_lock);
3263 3264
	if (l3->shared) {
		struct array_cache *shared_array = l3->shared;
P
Pekka Enberg 已提交
3265
		int max = shared_array->limit - shared_array->avail;
L
Linus Torvalds 已提交
3266 3267 3268
		if (max) {
			if (batchcount > max)
				batchcount = max;
3269
			memcpy(&(shared_array->entry[shared_array->avail]),
P
Pekka Enberg 已提交
3270
			       ac->entry, sizeof(void *) * batchcount);
L
Linus Torvalds 已提交
3271 3272 3273 3274 3275
			shared_array->avail += batchcount;
			goto free_done;
		}
	}

3276
	free_block(cachep, ac->entry, batchcount, node);
A
Andrew Morton 已提交
3277
free_done:
L
Linus Torvalds 已提交
3278 3279 3280 3281 3282
#if STATS
	{
		int i = 0;
		struct list_head *p;

3283 3284
		p = l3->slabs_free.next;
		while (p != &(l3->slabs_free)) {
L
Linus Torvalds 已提交
3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295
			struct slab *slabp;

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

			i++;
			p = p->next;
		}
		STATS_SET_FREEABLE(cachep, i);
	}
#endif
3296
	spin_unlock(&l3->list_lock);
L
Linus Torvalds 已提交
3297
	ac->avail -= batchcount;
A
Andrew Morton 已提交
3298
	memmove(ac->entry, &(ac->entry[batchcount]), sizeof(void *)*ac->avail);
L
Linus Torvalds 已提交
3299 3300 3301
}

/*
A
Andrew Morton 已提交
3302 3303
 * 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
3304
 */
3305
static inline void __cache_free(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
3306
{
3307
	struct array_cache *ac = cpu_cache_get(cachep);
L
Linus Torvalds 已提交
3308 3309 3310 3311

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

3312
	if (cache_free_alien(cachep, objp))
3313 3314
		return;

L
Linus Torvalds 已提交
3315 3316
	if (likely(ac->avail < ac->limit)) {
		STATS_INC_FREEHIT(cachep);
3317
		ac->entry[ac->avail++] = objp;
L
Linus Torvalds 已提交
3318 3319 3320 3321
		return;
	} else {
		STATS_INC_FREEMISS(cachep);
		cache_flusharray(cachep, ac);
3322
		ac->entry[ac->avail++] = objp;
L
Linus Torvalds 已提交
3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333
	}
}

/**
 * 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.
 */
3334
void *kmem_cache_alloc(struct kmem_cache *cachep, gfp_t flags)
L
Linus Torvalds 已提交
3335
{
3336
	return __cache_alloc(cachep, flags, __builtin_return_address(0));
L
Linus Torvalds 已提交
3337 3338 3339
}
EXPORT_SYMBOL(kmem_cache_alloc);

3340
/**
3341
 * kmem_cache_zalloc - Allocate an object. The memory is set to zero.
3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356
 * @cache: The cache to allocate from.
 * @flags: See kmalloc().
 *
 * Allocate an object from this cache and set the allocated memory to zero.
 * The flags are only relevant if the cache has no available objects.
 */
void *kmem_cache_zalloc(struct kmem_cache *cache, gfp_t flags)
{
	void *ret = __cache_alloc(cache, flags, __builtin_return_address(0));
	if (ret)
		memset(ret, 0, obj_size(cache));
	return ret;
}
EXPORT_SYMBOL(kmem_cache_zalloc);

L
Linus Torvalds 已提交
3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370
/**
 * kmem_ptr_validate - check if an untrusted pointer might
 *	be a slab entry.
 * @cachep: the cache we're checking against
 * @ptr: pointer to validate
 *
 * This verifies that the untrusted pointer looks sane:
 * it is _not_ a guarantee that the pointer is actually
 * part of the slab cache in question, but it at least
 * validates that the pointer can be dereferenced and
 * looks half-way sane.
 *
 * Currently only used for dentry validation.
 */
3371
int fastcall kmem_ptr_validate(struct kmem_cache *cachep, void *ptr)
L
Linus Torvalds 已提交
3372
{
P
Pekka Enberg 已提交
3373
	unsigned long addr = (unsigned long)ptr;
L
Linus Torvalds 已提交
3374
	unsigned long min_addr = PAGE_OFFSET;
P
Pekka Enberg 已提交
3375
	unsigned long align_mask = BYTES_PER_WORD - 1;
3376
	unsigned long size = cachep->buffer_size;
L
Linus Torvalds 已提交
3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391
	struct page *page;

	if (unlikely(addr < min_addr))
		goto out;
	if (unlikely(addr > (unsigned long)high_memory - size))
		goto out;
	if (unlikely(addr & align_mask))
		goto out;
	if (unlikely(!kern_addr_valid(addr)))
		goto out;
	if (unlikely(!kern_addr_valid(addr + size - 1)))
		goto out;
	page = virt_to_page(ptr);
	if (unlikely(!PageSlab(page)))
		goto out;
3392
	if (unlikely(page_get_cache(page) != cachep))
L
Linus Torvalds 已提交
3393 3394
		goto out;
	return 1;
A
Andrew Morton 已提交
3395
out:
L
Linus Torvalds 已提交
3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408
	return 0;
}

#ifdef CONFIG_NUMA
/**
 * kmem_cache_alloc_node - Allocate an object on the specified node
 * @cachep: The cache to allocate from.
 * @flags: See kmalloc().
 * @nodeid: node number of the target node.
 *
 * Identical to kmem_cache_alloc, except that this function is slow
 * and can sleep. And it will allocate memory on the given node, which
 * can improve the performance for cpu bound structures.
3409 3410
 * New and improved: it will now make sure that the object gets
 * put on the correct node list so that there is no false sharing.
L
Linus Torvalds 已提交
3411
 */
3412
void *kmem_cache_alloc_node(struct kmem_cache *cachep, gfp_t flags, int nodeid)
L
Linus Torvalds 已提交
3413
{
3414 3415
	unsigned long save_flags;
	void *ptr;
L
Linus Torvalds 已提交
3416

3417 3418
	cache_alloc_debugcheck_before(cachep, flags);
	local_irq_save(save_flags);
3419 3420

	if (nodeid == -1 || nodeid == numa_node_id() ||
A
Andrew Morton 已提交
3421
			!cachep->nodelists[nodeid])
3422 3423 3424
		ptr = ____cache_alloc(cachep, flags);
	else
		ptr = __cache_alloc_node(cachep, flags, nodeid);
3425
	local_irq_restore(save_flags);
3426 3427 3428

	ptr = cache_alloc_debugcheck_after(cachep, flags, ptr,
					   __builtin_return_address(0));
L
Linus Torvalds 已提交
3429

3430
	return ptr;
L
Linus Torvalds 已提交
3431 3432 3433
}
EXPORT_SYMBOL(kmem_cache_alloc_node);

3434
void *__kmalloc_node(size_t size, gfp_t flags, int node)
3435
{
3436
	struct kmem_cache *cachep;
3437 3438 3439 3440 3441 3442

	cachep = kmem_find_general_cachep(size, flags);
	if (unlikely(cachep == NULL))
		return NULL;
	return kmem_cache_alloc_node(cachep, flags, node);
}
3443
EXPORT_SYMBOL(__kmalloc_node);
L
Linus Torvalds 已提交
3444 3445 3446
#endif

/**
3447
 * __do_kmalloc - allocate memory
L
Linus Torvalds 已提交
3448
 * @size: how many bytes of memory are required.
3449
 * @flags: the type of memory to allocate (see kmalloc).
3450
 * @caller: function caller for debug tracking of the caller
L
Linus Torvalds 已提交
3451
 */
3452 3453
static __always_inline void *__do_kmalloc(size_t size, gfp_t flags,
					  void *caller)
L
Linus Torvalds 已提交
3454
{
3455
	struct kmem_cache *cachep;
L
Linus Torvalds 已提交
3456

3457 3458 3459 3460 3461 3462
	/* 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);
3463 3464
	if (unlikely(cachep == NULL))
		return NULL;
3465 3466 3467 3468 3469 3470
	return __cache_alloc(cachep, flags, caller);
}


void *__kmalloc(size_t size, gfp_t flags)
{
3471
#ifndef CONFIG_DEBUG_SLAB
3472
	return __do_kmalloc(size, flags, NULL);
3473 3474 3475
#else
	return __do_kmalloc(size, flags, __builtin_return_address(0));
#endif
L
Linus Torvalds 已提交
3476 3477 3478
}
EXPORT_SYMBOL(__kmalloc);

3479
#ifdef CONFIG_DEBUG_SLAB
3480 3481 3482 3483 3484 3485 3486
void *__kmalloc_track_caller(size_t size, gfp_t flags, void *caller)
{
	return __do_kmalloc(size, flags, caller);
}
EXPORT_SYMBOL(__kmalloc_track_caller);
#endif

L
Linus Torvalds 已提交
3487 3488 3489 3490 3491 3492 3493 3494
/**
 * 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.
 */
3495
void kmem_cache_free(struct kmem_cache *cachep, void *objp)
L
Linus Torvalds 已提交
3496 3497 3498
{
	unsigned long flags;

3499 3500
	BUG_ON(virt_to_cache(objp) != cachep);

L
Linus Torvalds 已提交
3501
	local_irq_save(flags);
3502
	__cache_free(cachep, objp);
L
Linus Torvalds 已提交
3503 3504 3505 3506 3507 3508 3509 3510
	local_irq_restore(flags);
}
EXPORT_SYMBOL(kmem_cache_free);

/**
 * kfree - free previously allocated memory
 * @objp: pointer returned by kmalloc.
 *
3511 3512
 * If @objp is NULL, no operation is performed.
 *
L
Linus Torvalds 已提交
3513 3514 3515 3516 3517
 * Don't free memory not originally allocated by kmalloc()
 * or you will run into trouble.
 */
void kfree(const void *objp)
{
3518
	struct kmem_cache *c;
L
Linus Torvalds 已提交
3519 3520 3521 3522 3523 3524
	unsigned long flags;

	if (unlikely(!objp))
		return;
	local_irq_save(flags);
	kfree_debugcheck(objp);
3525
	c = virt_to_cache(objp);
3526
	debug_check_no_locks_freed(objp, obj_size(c));
3527
	__cache_free(c, (void *)objp);
L
Linus Torvalds 已提交
3528 3529 3530 3531
	local_irq_restore(flags);
}
EXPORT_SYMBOL(kfree);

3532
unsigned int kmem_cache_size(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
3533
{
3534
	return obj_size(cachep);
L
Linus Torvalds 已提交
3535 3536 3537
}
EXPORT_SYMBOL(kmem_cache_size);

3538
const char *kmem_cache_name(struct kmem_cache *cachep)
3539 3540 3541 3542 3543
{
	return cachep->name;
}
EXPORT_SYMBOL_GPL(kmem_cache_name);

3544
/*
3545
 * This initializes kmem_list3 or resizes varioius caches for all nodes.
3546
 */
3547
static int alloc_kmemlist(struct kmem_cache *cachep)
3548 3549 3550
{
	int node;
	struct kmem_list3 *l3;
3551 3552
	struct array_cache *new_shared;
	struct array_cache **new_alien;
3553 3554

	for_each_online_node(node) {
3555

A
Andrew Morton 已提交
3556 3557
		new_alien = alloc_alien_cache(node, cachep->limit);
		if (!new_alien)
3558
			goto fail;
3559

3560 3561
		new_shared = alloc_arraycache(node,
				cachep->shared*cachep->batchcount,
A
Andrew Morton 已提交
3562
					0xbaadf00d);
3563 3564
		if (!new_shared) {
			free_alien_cache(new_alien);
3565
			goto fail;
3566
		}
3567

A
Andrew Morton 已提交
3568 3569
		l3 = cachep->nodelists[node];
		if (l3) {
3570 3571
			struct array_cache *shared = l3->shared;

3572 3573
			spin_lock_irq(&l3->list_lock);

3574
			if (shared)
3575 3576
				free_block(cachep, shared->entry,
						shared->avail, node);
3577

3578 3579
			l3->shared = new_shared;
			if (!l3->alien) {
3580 3581 3582
				l3->alien = new_alien;
				new_alien = NULL;
			}
P
Pekka Enberg 已提交
3583
			l3->free_limit = (1 + nr_cpus_node(node)) *
A
Andrew Morton 已提交
3584
					cachep->batchcount + cachep->num;
3585
			spin_unlock_irq(&l3->list_lock);
3586
			kfree(shared);
3587 3588 3589
			free_alien_cache(new_alien);
			continue;
		}
A
Andrew Morton 已提交
3590
		l3 = kmalloc_node(sizeof(struct kmem_list3), GFP_KERNEL, node);
3591 3592 3593
		if (!l3) {
			free_alien_cache(new_alien);
			kfree(new_shared);
3594
			goto fail;
3595
		}
3596 3597 3598

		kmem_list3_init(l3);
		l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
A
Andrew Morton 已提交
3599
				((unsigned long)cachep) % REAPTIMEOUT_LIST3;
3600
		l3->shared = new_shared;
3601
		l3->alien = new_alien;
P
Pekka Enberg 已提交
3602
		l3->free_limit = (1 + nr_cpus_node(node)) *
A
Andrew Morton 已提交
3603
					cachep->batchcount + cachep->num;
3604 3605
		cachep->nodelists[node] = l3;
	}
3606
	return 0;
3607

A
Andrew Morton 已提交
3608
fail:
3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623
	if (!cachep->next.next) {
		/* Cache is not active yet. Roll back what we did */
		node--;
		while (node >= 0) {
			if (cachep->nodelists[node]) {
				l3 = cachep->nodelists[node];

				kfree(l3->shared);
				free_alien_cache(l3->alien);
				kfree(l3);
				cachep->nodelists[node] = NULL;
			}
			node--;
		}
	}
3624
	return -ENOMEM;
3625 3626
}

L
Linus Torvalds 已提交
3627
struct ccupdate_struct {
3628
	struct kmem_cache *cachep;
L
Linus Torvalds 已提交
3629 3630 3631 3632 3633
	struct array_cache *new[NR_CPUS];
};

static void do_ccupdate_local(void *info)
{
A
Andrew Morton 已提交
3634
	struct ccupdate_struct *new = info;
L
Linus Torvalds 已提交
3635 3636 3637
	struct array_cache *old;

	check_irq_off();
3638
	old = cpu_cache_get(new->cachep);
3639

L
Linus Torvalds 已提交
3640 3641 3642 3643
	new->cachep->array[smp_processor_id()] = new->new[smp_processor_id()];
	new->new[smp_processor_id()] = old;
}

3644
/* Always called with the cache_chain_mutex held */
A
Andrew Morton 已提交
3645 3646
static int do_tune_cpucache(struct kmem_cache *cachep, int limit,
				int batchcount, int shared)
L
Linus Torvalds 已提交
3647
{
3648
	struct ccupdate_struct *new;
3649
	int i;
L
Linus Torvalds 已提交
3650

3651 3652 3653 3654
	new = kzalloc(sizeof(*new), GFP_KERNEL);
	if (!new)
		return -ENOMEM;

3655
	for_each_online_cpu(i) {
3656
		new->new[i] = alloc_arraycache(cpu_to_node(i), limit,
A
Andrew Morton 已提交
3657
						batchcount);
3658
		if (!new->new[i]) {
P
Pekka Enberg 已提交
3659
			for (i--; i >= 0; i--)
3660 3661
				kfree(new->new[i]);
			kfree(new);
3662
			return -ENOMEM;
L
Linus Torvalds 已提交
3663 3664
		}
	}
3665
	new->cachep = cachep;
L
Linus Torvalds 已提交
3666

3667
	on_each_cpu(do_ccupdate_local, (void *)new, 1, 1);
3668

L
Linus Torvalds 已提交
3669 3670 3671
	check_irq_on();
	cachep->batchcount = batchcount;
	cachep->limit = limit;
3672
	cachep->shared = shared;
L
Linus Torvalds 已提交
3673

3674
	for_each_online_cpu(i) {
3675
		struct array_cache *ccold = new->new[i];
L
Linus Torvalds 已提交
3676 3677
		if (!ccold)
			continue;
3678
		spin_lock_irq(&cachep->nodelists[cpu_to_node(i)]->list_lock);
3679
		free_block(cachep, ccold->entry, ccold->avail, cpu_to_node(i));
3680
		spin_unlock_irq(&cachep->nodelists[cpu_to_node(i)]->list_lock);
L
Linus Torvalds 已提交
3681 3682
		kfree(ccold);
	}
3683
	kfree(new);
3684
	return alloc_kmemlist(cachep);
L
Linus Torvalds 已提交
3685 3686
}

3687
/* Called with cache_chain_mutex held always */
3688
static int enable_cpucache(struct kmem_cache *cachep)
L
Linus Torvalds 已提交
3689 3690 3691 3692
{
	int err;
	int limit, shared;

A
Andrew Morton 已提交
3693 3694
	/*
	 * The head array serves three purposes:
L
Linus Torvalds 已提交
3695 3696
	 * - create a LIFO ordering, i.e. return objects that are cache-warm
	 * - reduce the number of spinlock operations.
A
Andrew Morton 已提交
3697
	 * - reduce the number of linked list operations on the slab and
L
Linus Torvalds 已提交
3698 3699 3700 3701
	 *   bufctl chains: array operations are cheaper.
	 * The numbers are guessed, we should auto-tune as described by
	 * Bonwick.
	 */
3702
	if (cachep->buffer_size > 131072)
L
Linus Torvalds 已提交
3703
		limit = 1;
3704
	else if (cachep->buffer_size > PAGE_SIZE)
L
Linus Torvalds 已提交
3705
		limit = 8;
3706
	else if (cachep->buffer_size > 1024)
L
Linus Torvalds 已提交
3707
		limit = 24;
3708
	else if (cachep->buffer_size > 256)
L
Linus Torvalds 已提交
3709 3710 3711 3712
		limit = 54;
	else
		limit = 120;

A
Andrew Morton 已提交
3713 3714
	/*
	 * CPU bound tasks (e.g. network routing) can exhibit cpu bound
L
Linus Torvalds 已提交
3715 3716 3717 3718 3719 3720 3721 3722 3723
	 * allocation behaviour: Most allocs on one cpu, most free operations
	 * on another cpu. For these cases, an efficient object passing between
	 * cpus is necessary. This is provided by a shared array. The array
	 * replaces Bonwick's magazine layer.
	 * On uniprocessor, it's functionally equivalent (but less efficient)
	 * to a larger limit. Thus disabled by default.
	 */
	shared = 0;
#ifdef CONFIG_SMP
3724
	if (cachep->buffer_size <= PAGE_SIZE)
L
Linus Torvalds 已提交
3725 3726 3727 3728
		shared = 8;
#endif

#if DEBUG
A
Andrew Morton 已提交
3729 3730 3731
	/*
	 * With debugging enabled, large batchcount lead to excessively long
	 * periods with disabled local interrupts. Limit the batchcount
L
Linus Torvalds 已提交
3732 3733 3734 3735
	 */
	if (limit > 32)
		limit = 32;
#endif
P
Pekka Enberg 已提交
3736
	err = do_tune_cpucache(cachep, limit, (limit + 1) / 2, shared);
L
Linus Torvalds 已提交
3737 3738
	if (err)
		printk(KERN_ERR "enable_cpucache failed for %s, error %d.\n",
P
Pekka Enberg 已提交
3739
		       cachep->name, -err);
3740
	return err;
L
Linus Torvalds 已提交
3741 3742
}

3743 3744
/*
 * Drain an array if it contains any elements taking the l3 lock only if
3745 3746
 * necessary. Note that the l3 listlock also protects the array_cache
 * if drain_array() is used on the shared array.
3747 3748 3749
 */
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
3750 3751 3752
{
	int tofree;

3753 3754
	if (!ac || !ac->avail)
		return;
L
Linus Torvalds 已提交
3755 3756
	if (ac->touched && !force) {
		ac->touched = 0;
3757
	} else {
3758
		spin_lock_irq(&l3->list_lock);
3759 3760 3761 3762 3763 3764 3765 3766 3767
		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);
		}
3768
		spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
3769 3770 3771 3772 3773
	}
}

/**
 * cache_reap - Reclaim memory from caches.
3774
 * @unused: unused parameter
L
Linus Torvalds 已提交
3775 3776 3777 3778 3779 3780
 *
 * 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
3781 3782
 * 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
3783 3784 3785
 */
static void cache_reap(void *unused)
{
3786
	struct kmem_cache *searchp;
3787
	struct kmem_list3 *l3;
3788
	int node = numa_node_id();
L
Linus Torvalds 已提交
3789

I
Ingo Molnar 已提交
3790
	if (!mutex_trylock(&cache_chain_mutex)) {
L
Linus Torvalds 已提交
3791
		/* Give up. Setup the next iteration. */
P
Pekka Enberg 已提交
3792 3793
		schedule_delayed_work(&__get_cpu_var(reap_work),
				      REAPTIMEOUT_CPUC);
L
Linus Torvalds 已提交
3794 3795 3796
		return;
	}

3797
	list_for_each_entry(searchp, &cache_chain, next) {
L
Linus Torvalds 已提交
3798 3799
		check_irq_on();

3800 3801 3802 3803 3804
		/*
		 * 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.
		 */
3805
		l3 = searchp->nodelists[node];
3806

3807
		reap_alien(searchp, l3);
L
Linus Torvalds 已提交
3808

3809
		drain_array(searchp, l3, cpu_cache_get(searchp), 0, node);
L
Linus Torvalds 已提交
3810

3811 3812 3813 3814
		/*
		 * These are racy checks but it does not matter
		 * if we skip one check or scan twice.
		 */
3815
		if (time_after(l3->next_reap, jiffies))
3816
			goto next;
L
Linus Torvalds 已提交
3817

3818
		l3->next_reap = jiffies + REAPTIMEOUT_LIST3;
L
Linus Torvalds 已提交
3819

3820
		drain_array(searchp, l3, l3->shared, 0, node);
L
Linus Torvalds 已提交
3821

3822
		if (l3->free_touched)
3823
			l3->free_touched = 0;
3824 3825
		else {
			int freed;
L
Linus Torvalds 已提交
3826

3827 3828 3829 3830
			freed = drain_freelist(searchp, l3, (l3->free_limit +
				5 * searchp->num - 1) / (5 * searchp->num));
			STATS_ADD_REAPED(searchp, freed);
		}
3831
next:
L
Linus Torvalds 已提交
3832 3833 3834
		cond_resched();
	}
	check_irq_on();
I
Ingo Molnar 已提交
3835
	mutex_unlock(&cache_chain_mutex);
3836
	next_reap_node();
3837
	refresh_cpu_vm_stats(smp_processor_id());
A
Andrew Morton 已提交
3838
	/* Set up the next iteration */
3839
	schedule_delayed_work(&__get_cpu_var(reap_work), REAPTIMEOUT_CPUC);
L
Linus Torvalds 已提交
3840 3841 3842 3843
}

#ifdef CONFIG_PROC_FS

3844
static void print_slabinfo_header(struct seq_file *m)
L
Linus Torvalds 已提交
3845
{
3846 3847 3848 3849
	/*
	 * Output format version, so at least we can change it
	 * without _too_ many complaints.
	 */
L
Linus Torvalds 已提交
3850
#if STATS
3851
	seq_puts(m, "slabinfo - version: 2.1 (statistics)\n");
L
Linus Torvalds 已提交
3852
#else
3853
	seq_puts(m, "slabinfo - version: 2.1\n");
L
Linus Torvalds 已提交
3854
#endif
3855 3856 3857 3858
	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
3859
#if STATS
3860
	seq_puts(m, " : globalstat <listallocs> <maxobjs> <grown> <reaped> "
3861
		 "<error> <maxfreeable> <nodeallocs> <remotefrees> <alienoverflow>");
3862
	seq_puts(m, " : cpustat <allochit> <allocmiss> <freehit> <freemiss>");
L
Linus Torvalds 已提交
3863
#endif
3864 3865 3866 3867 3868 3869 3870 3871
	seq_putc(m, '\n');
}

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

I
Ingo Molnar 已提交
3872
	mutex_lock(&cache_chain_mutex);
3873 3874
	if (!n)
		print_slabinfo_header(m);
L
Linus Torvalds 已提交
3875 3876 3877 3878 3879 3880
	p = cache_chain.next;
	while (n--) {
		p = p->next;
		if (p == &cache_chain)
			return NULL;
	}
3881
	return list_entry(p, struct kmem_cache, next);
L
Linus Torvalds 已提交
3882 3883 3884 3885
}

static void *s_next(struct seq_file *m, void *p, loff_t *pos)
{
3886
	struct kmem_cache *cachep = p;
L
Linus Torvalds 已提交
3887
	++*pos;
A
Andrew Morton 已提交
3888 3889
	return cachep->next.next == &cache_chain ?
		NULL : list_entry(cachep->next.next, struct kmem_cache, next);
L
Linus Torvalds 已提交
3890 3891 3892 3893
}

static void s_stop(struct seq_file *m, void *p)
{
I
Ingo Molnar 已提交
3894
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
3895 3896 3897 3898
}

static int s_show(struct seq_file *m, void *p)
{
3899
	struct kmem_cache *cachep = p;
P
Pekka Enberg 已提交
3900 3901 3902 3903 3904
	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;
3905
	const char *name;
L
Linus Torvalds 已提交
3906
	char *error = NULL;
3907 3908
	int node;
	struct kmem_list3 *l3;
L
Linus Torvalds 已提交
3909 3910 3911

	active_objs = 0;
	num_slabs = 0;
3912 3913 3914 3915 3916
	for_each_online_node(node) {
		l3 = cachep->nodelists[node];
		if (!l3)
			continue;

3917 3918
		check_irq_on();
		spin_lock_irq(&l3->list_lock);
3919

3920
		list_for_each_entry(slabp, &l3->slabs_full, list) {
3921 3922 3923 3924 3925
			if (slabp->inuse != cachep->num && !error)
				error = "slabs_full accounting error";
			active_objs += cachep->num;
			active_slabs++;
		}
3926
		list_for_each_entry(slabp, &l3->slabs_partial, list) {
3927 3928 3929 3930 3931 3932 3933
			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++;
		}
3934
		list_for_each_entry(slabp, &l3->slabs_free, list) {
3935 3936 3937 3938 3939
			if (slabp->inuse && !error)
				error = "slabs_free/inuse accounting error";
			num_slabs++;
		}
		free_objects += l3->free_objects;
3940 3941
		if (l3->shared)
			shared_avail += l3->shared->avail;
3942

3943
		spin_unlock_irq(&l3->list_lock);
L
Linus Torvalds 已提交
3944
	}
P
Pekka Enberg 已提交
3945 3946
	num_slabs += active_slabs;
	num_objs = num_slabs * cachep->num;
3947
	if (num_objs - active_objs != free_objects && !error)
L
Linus Torvalds 已提交
3948 3949
		error = "free_objects accounting error";

P
Pekka Enberg 已提交
3950
	name = cachep->name;
L
Linus Torvalds 已提交
3951 3952 3953 3954
	if (error)
		printk(KERN_ERR "slab: cache %s error: %s\n", name, error);

	seq_printf(m, "%-17s %6lu %6lu %6u %4u %4d",
3955
		   name, active_objs, num_objs, cachep->buffer_size,
P
Pekka Enberg 已提交
3956
		   cachep->num, (1 << cachep->gfporder));
L
Linus Torvalds 已提交
3957
	seq_printf(m, " : tunables %4u %4u %4u",
P
Pekka Enberg 已提交
3958
		   cachep->limit, cachep->batchcount, cachep->shared);
3959
	seq_printf(m, " : slabdata %6lu %6lu %6lu",
P
Pekka Enberg 已提交
3960
		   active_slabs, num_slabs, shared_avail);
L
Linus Torvalds 已提交
3961
#if STATS
P
Pekka Enberg 已提交
3962
	{			/* list3 stats */
L
Linus Torvalds 已提交
3963 3964 3965 3966 3967 3968 3969
		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;
3970
		unsigned long node_frees = cachep->node_frees;
3971
		unsigned long overflows = cachep->node_overflow;
L
Linus Torvalds 已提交
3972

3973
		seq_printf(m, " : globalstat %7lu %6lu %5lu %4lu \
3974
				%4lu %4lu %4lu %4lu %4lu", allocs, high, grown,
A
Andrew Morton 已提交
3975
				reaped, errors, max_freeable, node_allocs,
3976
				node_frees, overflows);
L
Linus Torvalds 已提交
3977 3978 3979 3980 3981 3982 3983 3984 3985
	}
	/* 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
3986
			   allochit, allocmiss, freehit, freemiss);
L
Linus Torvalds 已提交
3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007
	}
#endif
	seq_putc(m, '\n');
	return 0;
}

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

struct seq_operations slabinfo_op = {
P
Pekka Enberg 已提交
4008 4009 4010 4011
	.start = s_start,
	.next = s_next,
	.stop = s_stop,
	.show = s_show,
L
Linus Torvalds 已提交
4012 4013 4014 4015 4016 4017 4018 4019 4020 4021
};

#define MAX_SLABINFO_WRITE 128
/**
 * slabinfo_write - Tuning for the slab allocator
 * @file: unused
 * @buffer: user buffer
 * @count: data length
 * @ppos: unused
 */
P
Pekka Enberg 已提交
4022 4023
ssize_t slabinfo_write(struct file *file, const char __user * buffer,
		       size_t count, loff_t *ppos)
L
Linus Torvalds 已提交
4024
{
P
Pekka Enberg 已提交
4025
	char kbuf[MAX_SLABINFO_WRITE + 1], *tmp;
L
Linus Torvalds 已提交
4026
	int limit, batchcount, shared, res;
4027
	struct kmem_cache *cachep;
P
Pekka Enberg 已提交
4028

L
Linus Torvalds 已提交
4029 4030 4031 4032
	if (count > MAX_SLABINFO_WRITE)
		return -EINVAL;
	if (copy_from_user(&kbuf, buffer, count))
		return -EFAULT;
P
Pekka Enberg 已提交
4033
	kbuf[MAX_SLABINFO_WRITE] = '\0';
L
Linus Torvalds 已提交
4034 4035 4036 4037 4038 4039 4040 4041 4042 4043

	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
4044
	mutex_lock(&cache_chain_mutex);
L
Linus Torvalds 已提交
4045
	res = -EINVAL;
4046
	list_for_each_entry(cachep, &cache_chain, next) {
L
Linus Torvalds 已提交
4047
		if (!strcmp(cachep->name, kbuf)) {
A
Andrew Morton 已提交
4048 4049
			if (limit < 1 || batchcount < 1 ||
					batchcount > limit || shared < 0) {
4050
				res = 0;
L
Linus Torvalds 已提交
4051
			} else {
4052
				res = do_tune_cpucache(cachep, limit,
P
Pekka Enberg 已提交
4053
						       batchcount, shared);
L
Linus Torvalds 已提交
4054 4055 4056 4057
			}
			break;
		}
	}
I
Ingo Molnar 已提交
4058
	mutex_unlock(&cache_chain_mutex);
L
Linus Torvalds 已提交
4059 4060 4061 4062
	if (res >= 0)
		res = count;
	return res;
}
4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171

#ifdef CONFIG_DEBUG_SLAB_LEAK

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

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

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

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

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

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

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

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

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

	/* OK, we can do it */

	n[1] = 0;

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

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

4172
		list_for_each_entry(slabp, &l3->slabs_full, list)
4173
			handle_slab(n, cachep, slabp);
4174
		list_for_each_entry(slabp, &l3->slabs_partial, list)
4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200
			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');
	}
4201

4202 4203 4204 4205 4206 4207 4208 4209 4210 4211
	return 0;
}

struct seq_operations slabstats_op = {
	.start = leaks_start,
	.next = s_next,
	.stop = s_stop,
	.show = leaks_show,
};
#endif
L
Linus Torvalds 已提交
4212 4213
#endif

4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225
/**
 * ksize - get the actual amount of memory allocated for a given object
 * @objp: Pointer to the object
 *
 * kmalloc may internally round up allocations and return more memory
 * than requested. ksize() can be used to determine the actual amount of
 * memory allocated. The caller may use this additional memory, even though
 * a smaller amount of memory was initially specified with the kmalloc call.
 * The caller must guarantee that objp points to a valid object previously
 * allocated with either kmalloc() or kmem_cache_alloc(). The object
 * must not be freed during the duration of the call.
 */
L
Linus Torvalds 已提交
4226 4227
unsigned int ksize(const void *objp)
{
4228 4229
	if (unlikely(objp == NULL))
		return 0;
L
Linus Torvalds 已提交
4230

4231
	return obj_size(virt_to_cache(objp));
L
Linus Torvalds 已提交
4232
}