slab.c 102 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
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.
 *
Andrew Morton's avatar
Andrew Morton committed
53
 * The c_cpuarray may not be read with enabled local interrupts -
Linus Torvalds's avatar
Linus Torvalds committed
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
Linus Torvalds's avatar
Linus Torvalds committed
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
Ingo Molnar's avatar
Ingo Molnar committed
71
 *	The global cache-chain is protected by the mutex 'cache_chain_mutex'.
Linus Torvalds's avatar
Linus Torvalds committed
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.
Linus Torvalds's avatar
Linus Torvalds committed
87
88
89
90
91
92
93
94
95
96
 */

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

#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 | \
174
			 SLAB_CACHE_DMA | \
Linus Torvalds's avatar
Linus Torvalds committed
175
176
			 SLAB_MUST_HWCACHE_ALIGN | SLAB_STORE_USER | \
			 SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
177
			 SLAB_DESTROY_BY_RCU | SLAB_MEM_SPREAD)
Linus Torvalds's avatar
Linus Torvalds committed
178
#else
179
# define CREATE_MASK	(SLAB_HWCACHE_ALIGN | \
Linus Torvalds's avatar
Linus Torvalds committed
180
181
			 SLAB_CACHE_DMA | SLAB_MUST_HWCACHE_ALIGN | \
			 SLAB_RECLAIM_ACCOUNT | SLAB_PANIC | \
182
			 SLAB_DESTROY_BY_RCU | SLAB_MEM_SPREAD)
Linus Torvalds's avatar
Linus Torvalds committed
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
#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.
 */

204
typedef unsigned int kmem_bufctl_t;
Linus Torvalds's avatar
Linus Torvalds committed
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
#define BUFCTL_END	(((kmem_bufctl_t)(~0U))-0)
#define BUFCTL_FREE	(((kmem_bufctl_t)(~0U))-1)
#define	SLAB_LIMIT	(((kmem_bufctl_t)(~0U))-2)

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

/*
 * struct slab
 *
 * Manages the objs in a slab. Placed either at the beginning of mem allocated
 * for a slab, or allocated from an general cache.
 * Slabs are chained into three list: fully used, partial, fully free slabs.
 */
struct slab {
222
223
224
225
226
227
	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;
Linus Torvalds's avatar
Linus Torvalds committed
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
};

/*
 * 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 {
247
	struct rcu_head head;
248
	struct kmem_cache *cachep;
249
	void *addr;
Linus Torvalds's avatar
Linus Torvalds committed
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
};

/*
 * 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;
269
	spinlock_t lock;
Andrew Morton's avatar
Andrew Morton committed
270
271
272
273
274
275
	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 [].
			 */
Linus Torvalds's avatar
Linus Torvalds committed
276
277
};

Andrew Morton's avatar
Andrew Morton committed
278
279
280
/*
 * bootstrap: The caches do not work without cpuarrays anymore, but the
 * cpuarrays are allocated from the generic caches...
Linus Torvalds's avatar
Linus Torvalds committed
281
282
283
284
 */
#define BOOT_CPUCACHE_ENTRIES	1
struct arraycache_init {
	struct array_cache cache;
285
	void *entries[BOOT_CPUCACHE_ENTRIES];
Linus Torvalds's avatar
Linus Torvalds committed
286
287
288
};

/*
289
 * The slab lists for all objects.
Linus Torvalds's avatar
Linus Torvalds committed
290
291
 */
struct kmem_list3 {
292
293
294
295
296
	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;
297
	unsigned int colour_next;	/* Per-node cache coloring */
298
299
300
	spinlock_t list_lock;
	struct array_cache *shared;	/* shared per node */
	struct array_cache **alien;	/* on other nodes */
301
302
	unsigned long next_reap;	/* updated without locking */
	int free_touched;		/* updated without locking */
Linus Torvalds's avatar
Linus Torvalds committed
303
304
};

305
306
307
308
309
310
311
312
313
314
/*
 * 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)

/*
Andrew Morton's avatar
Andrew Morton committed
315
316
 * 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.
317
 */
318
static __always_inline int index_of(const size_t size)
319
{
320
321
	extern void __bad_size(void);

322
323
324
325
326
327
328
329
330
331
	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
332
		__bad_size();
333
	} else
334
		__bad_size();
335
336
337
338
339
	return 0;
}

#define INDEX_AC index_of(sizeof(struct arraycache_init))
#define INDEX_L3 index_of(sizeof(struct kmem_list3))
Linus Torvalds's avatar
Linus Torvalds committed
340

Pekka Enberg's avatar
Pekka Enberg committed
341
static void kmem_list3_init(struct kmem_list3 *parent)
342
343
344
345
346
347
{
	INIT_LIST_HEAD(&parent->slabs_full);
	INIT_LIST_HEAD(&parent->slabs_partial);
	INIT_LIST_HEAD(&parent->slabs_free);
	parent->shared = NULL;
	parent->alien = NULL;
348
	parent->colour_next = 0;
349
350
351
352
353
	spin_lock_init(&parent->list_lock);
	parent->free_objects = 0;
	parent->free_touched = 0;
}

Andrew Morton's avatar
Andrew Morton committed
354
355
356
357
#define MAKE_LIST(cachep, listp, slab, nodeid)				\
	do {								\
		INIT_LIST_HEAD(listp);					\
		list_splice(&(cachep->nodelists[nodeid]->slab), listp);	\
358
359
	} while (0)

Andrew Morton's avatar
Andrew Morton committed
360
361
#define	MAKE_ALL_LISTS(cachep, ptr, nodeid)				\
	do {								\
362
363
364
365
	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)
Linus Torvalds's avatar
Linus Torvalds committed
366
367

/*
368
 * struct kmem_cache
Linus Torvalds's avatar
Linus Torvalds committed
369
370
371
 *
 * manages a cache.
 */
372

373
struct kmem_cache {
Linus Torvalds's avatar
Linus Torvalds committed
374
/* 1) per-cpu data, touched during every alloc/free */
375
	struct array_cache *array[NR_CPUS];
376
/* 2) Cache tunables. Protected by cache_chain_mutex */
377
378
379
	unsigned int batchcount;
	unsigned int limit;
	unsigned int shared;
380

381
	unsigned int buffer_size;
382
/* 3) touched by every alloc & free from the backend */
383
	struct kmem_list3 *nodelists[MAX_NUMNODES];
384

Andrew Morton's avatar
Andrew Morton committed
385
386
	unsigned int flags;		/* constant flags */
	unsigned int num;		/* # of objs per slab */
Linus Torvalds's avatar
Linus Torvalds committed
387

388
/* 4) cache_grow/shrink */
Linus Torvalds's avatar
Linus Torvalds committed
389
	/* order of pgs per slab (2^n) */
390
	unsigned int gfporder;
Linus Torvalds's avatar
Linus Torvalds committed
391
392

	/* force GFP flags, e.g. GFP_DMA */
393
	gfp_t gfpflags;
Linus Torvalds's avatar
Linus Torvalds committed
394

Andrew Morton's avatar
Andrew Morton committed
395
	size_t colour;			/* cache colouring range */
396
	unsigned int colour_off;	/* colour offset */
397
	struct kmem_cache *slabp_cache;
398
	unsigned int slab_size;
Andrew Morton's avatar
Andrew Morton committed
399
	unsigned int dflags;		/* dynamic flags */
Linus Torvalds's avatar
Linus Torvalds committed
400
401

	/* constructor func */
402
	void (*ctor) (void *, struct kmem_cache *, unsigned long);
Linus Torvalds's avatar
Linus Torvalds committed
403
404

	/* de-constructor func */
405
	void (*dtor) (void *, struct kmem_cache *, unsigned long);
Linus Torvalds's avatar
Linus Torvalds committed
406

407
/* 5) cache creation/removal */
408
409
	const char *name;
	struct list_head next;
Linus Torvalds's avatar
Linus Torvalds committed
410

411
/* 6) statistics */
Linus Torvalds's avatar
Linus Torvalds committed
412
#if STATS
413
414
415
416
417
418
419
420
421
422
423
424
425
	unsigned long num_active;
	unsigned long num_allocations;
	unsigned long high_mark;
	unsigned long grown;
	unsigned long reaped;
	unsigned long errors;
	unsigned long max_freeable;
	unsigned long node_allocs;
	unsigned long node_frees;
	atomic_t allochit;
	atomic_t allocmiss;
	atomic_t freehit;
	atomic_t freemiss;
Linus Torvalds's avatar
Linus Torvalds committed
426
427
#endif
#if DEBUG
428
429
430
431
432
433
434
435
	/*
	 * 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;
Linus Torvalds's avatar
Linus Torvalds committed
436
437
438
439
440
441
442
#endif
};

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

#define BATCHREFILL_LIMIT	16
Andrew Morton's avatar
Andrew Morton committed
443
444
445
/*
 * Optimization question: fewer reaps means less probability for unnessary
 * cpucache drain/refill cycles.
Linus Torvalds's avatar
Linus Torvalds committed
446
 *
Adrian Bunk's avatar
Adrian Bunk committed
447
 * OTOH the cpuarrays can contain lots of objects,
Linus Torvalds's avatar
Linus Torvalds committed
448
449
450
451
452
453
454
455
456
457
458
 * which could lock up otherwise freeable slabs.
 */
#define REAPTIMEOUT_CPUC	(2*HZ)
#define REAPTIMEOUT_LIST3	(4*HZ)

#if STATS
#define	STATS_INC_ACTIVE(x)	((x)->num_active++)
#define	STATS_DEC_ACTIVE(x)	((x)->num_active--)
#define	STATS_INC_ALLOCED(x)	((x)->num_allocations++)
#define	STATS_INC_GROWN(x)	((x)->grown++)
#define	STATS_INC_REAPED(x)	((x)->reaped++)
Andrew Morton's avatar
Andrew Morton committed
459
460
461
462
463
#define	STATS_SET_HIGH(x)						\
	do {								\
		if ((x)->num_active > (x)->high_mark)			\
			(x)->high_mark = (x)->num_active;		\
	} while (0)
Linus Torvalds's avatar
Linus Torvalds committed
464
465
#define	STATS_INC_ERR(x)	((x)->errors++)
#define	STATS_INC_NODEALLOCS(x)	((x)->node_allocs++)
466
#define	STATS_INC_NODEFREES(x)	((x)->node_frees++)
Andrew Morton's avatar
Andrew Morton committed
467
468
469
470
471
#define	STATS_SET_FREEABLE(x, i)					\
	do {								\
		if ((x)->max_freeable < i)				\
			(x)->max_freeable = i;				\
	} while (0)
Linus Torvalds's avatar
Linus Torvalds committed
472
473
474
475
476
477
478
479
480
481
482
483
484
#define STATS_INC_ALLOCHIT(x)	atomic_inc(&(x)->allochit)
#define STATS_INC_ALLOCMISS(x)	atomic_inc(&(x)->allocmiss)
#define STATS_INC_FREEHIT(x)	atomic_inc(&(x)->freehit)
#define STATS_INC_FREEMISS(x)	atomic_inc(&(x)->freemiss)
#else
#define	STATS_INC_ACTIVE(x)	do { } while (0)
#define	STATS_DEC_ACTIVE(x)	do { } while (0)
#define	STATS_INC_ALLOCED(x)	do { } while (0)
#define	STATS_INC_GROWN(x)	do { } while (0)
#define	STATS_INC_REAPED(x)	do { } while (0)
#define	STATS_SET_HIGH(x)	do { } while (0)
#define	STATS_INC_ERR(x)	do { } while (0)
#define	STATS_INC_NODEALLOCS(x)	do { } while (0)
485
#define	STATS_INC_NODEFREES(x)	do { } while (0)
Andrew Morton's avatar
Andrew Morton committed
486
#define	STATS_SET_FREEABLE(x, i) do { } while (0)
Linus Torvalds's avatar
Linus Torvalds committed
487
488
489
490
491
492
493
#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
Andrew Morton's avatar
Andrew Morton committed
494
495
/*
 * Magic nums for obj red zoning.
Linus Torvalds's avatar
Linus Torvalds committed
496
497
498
499
500
501
502
503
504
505
 * Placed in the first word before and the first word after an obj.
 */
#define	RED_INACTIVE	0x5A2CF071UL	/* when obj is inactive */
#define	RED_ACTIVE	0x170FC2A5UL	/* when obj is active */

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

Andrew Morton's avatar
Andrew Morton committed
506
507
/*
 * memory layout of objects:
Linus Torvalds's avatar
Linus Torvalds committed
508
 * 0		: objp
509
 * 0 .. cachep->obj_offset - BYTES_PER_WORD - 1: padding. This ensures that
Linus Torvalds's avatar
Linus Torvalds committed
510
511
 * 		the end of an object is aligned with the end of the real
 * 		allocation. Catches writes behind the end of the allocation.
512
 * cachep->obj_offset - BYTES_PER_WORD .. cachep->obj_offset - 1:
Linus Torvalds's avatar
Linus Torvalds committed
513
 * 		redzone word.
514
515
 * cachep->obj_offset: The real object.
 * cachep->buffer_size - 2* BYTES_PER_WORD: redzone word [BYTES_PER_WORD long]
Andrew Morton's avatar
Andrew Morton committed
516
517
 * cachep->buffer_size - 1* BYTES_PER_WORD: last caller address
 *					[BYTES_PER_WORD long]
Linus Torvalds's avatar
Linus Torvalds committed
518
 */
519
static int obj_offset(struct kmem_cache *cachep)
Linus Torvalds's avatar
Linus Torvalds committed
520
{
521
	return cachep->obj_offset;
Linus Torvalds's avatar
Linus Torvalds committed
522
523
}

524
static int obj_size(struct kmem_cache *cachep)
Linus Torvalds's avatar
Linus Torvalds committed
525
{
526
	return cachep->obj_size;
Linus Torvalds's avatar
Linus Torvalds committed
527
528
}

529
static unsigned long *dbg_redzone1(struct kmem_cache *cachep, void *objp)
Linus Torvalds's avatar
Linus Torvalds committed
530
531
{
	BUG_ON(!(cachep->flags & SLAB_RED_ZONE));
532
	return (unsigned long*) (objp+obj_offset(cachep)-BYTES_PER_WORD);
Linus Torvalds's avatar
Linus Torvalds committed
533
534
}

535
static unsigned long *dbg_redzone2(struct kmem_cache *cachep, void *objp)
Linus Torvalds's avatar
Linus Torvalds committed
536
537
538
{
	BUG_ON(!(cachep->flags & SLAB_RED_ZONE));
	if (cachep->flags & SLAB_STORE_USER)
539
		return (unsigned long *)(objp + cachep->buffer_size -
540
					 2 * BYTES_PER_WORD);
541
	return (unsigned long *)(objp + cachep->buffer_size - BYTES_PER_WORD);
Linus Torvalds's avatar
Linus Torvalds committed
542
543
}

544
static void **dbg_userword(struct kmem_cache *cachep, void *objp)
Linus Torvalds's avatar
Linus Torvalds committed
545
546
{
	BUG_ON(!(cachep->flags & SLAB_STORE_USER));
547
	return (void **)(objp + cachep->buffer_size - BYTES_PER_WORD);
Linus Torvalds's avatar
Linus Torvalds committed
548
549
550
551
}

#else

552
553
#define obj_offset(x)			0
#define obj_size(cachep)		(cachep->buffer_size)
Linus Torvalds's avatar
Linus Torvalds committed
554
555
556
557
558
559
560
#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

/*
Andrew Morton's avatar
Andrew Morton committed
561
562
 * Maximum size of an obj (in 2^order pages) and absolute limit for the gfp
 * order.
Linus Torvalds's avatar
Linus Torvalds committed
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
 */
#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;

Andrew Morton's avatar
Andrew Morton committed
582
583
584
585
/*
 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
586
 */
587
588
589
590
591
592
593
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)
{
594
595
	if (unlikely(PageCompound(page)))
		page = (struct page *)page_private(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
609
	return (struct slab *)page->lru.prev;
}
Linus Torvalds's avatar
Linus Torvalds committed
610

611
612
613
614
615
616
617
618
619
620
621
622
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);
}

623
624
625
626
627
628
629
630
631
632
633
634
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;
}

Andrew Morton's avatar
Andrew Morton committed
635
636
637
/*
 * These are the default caches for kmalloc. Custom caches can have other sizes.
 */
Linus Torvalds's avatar
Linus Torvalds committed
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
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>
655
	{NULL,}
Linus Torvalds's avatar
Linus Torvalds committed
656
657
658
659
#undef CACHE
};

static struct arraycache_init initarray_cache __initdata =
660
    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
Linus Torvalds's avatar
Linus Torvalds committed
661
static struct arraycache_init initarray_generic =
662
    { {0, BOOT_CPUCACHE_ENTRIES, 1, 0} };
Linus Torvalds's avatar
Linus Torvalds committed
663
664

/* internal cache of cache description objs */
665
static struct kmem_cache cache_cache = {
666
667
668
	.batchcount = 1,
	.limit = BOOT_CPUCACHE_ENTRIES,
	.shared = 1,
669
	.buffer_size = sizeof(struct kmem_cache),
670
	.name = "kmem_cache",
Linus Torvalds's avatar
Linus Torvalds committed
671
#if DEBUG
672
	.obj_size = sizeof(struct kmem_cache),
Linus Torvalds's avatar
Linus Torvalds committed
673
674
675
676
#endif
};

/* Guard access to the cache-chain. */
Ingo Molnar's avatar
Ingo Molnar committed
677
static DEFINE_MUTEX(cache_chain_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
678
679
680
static struct list_head cache_chain;

/*
Andrew Morton's avatar
Andrew Morton committed
681
682
 * vm_enough_memory() looks at this to determine how many slab-allocated pages
 * are possibly freeable under pressure
Linus Torvalds's avatar
Linus Torvalds committed
683
684
685
686
687
688
689
690
691
692
693
 *
 * SLAB_RECLAIM_ACCOUNT turns this on per-slab
 */
atomic_t slab_reclaim_pages;

/*
 * chicken and egg problem: delay the per-cpu array allocation
 * until the general caches are up.
 */
static enum {
	NONE,
694
695
	PARTIAL_AC,
	PARTIAL_L3,
Linus Torvalds's avatar
Linus Torvalds committed
696
697
698
699
700
	FULL
} g_cpucache_up;

static DEFINE_PER_CPU(struct work_struct, reap_work);

Andrew Morton's avatar
Andrew Morton committed
701
702
static void free_block(struct kmem_cache *cachep, void **objpp, int len,
			int node);
703
static void enable_cpucache(struct kmem_cache *cachep);
704
static void cache_reap(void *unused);
705
static int __node_shrink(struct kmem_cache *cachep, int node);
Linus Torvalds's avatar
Linus Torvalds committed
706

707
static inline struct array_cache *cpu_cache_get(struct kmem_cache *cachep)
Linus Torvalds's avatar
Linus Torvalds committed
708
709
710
711
{
	return cachep->array[smp_processor_id()];
}

Andrew Morton's avatar
Andrew Morton committed
712
713
static inline struct kmem_cache *__find_general_cachep(size_t size,
							gfp_t gfpflags)
Linus Torvalds's avatar
Linus Torvalds committed
714
715
716
717
718
{
	struct cache_sizes *csizep = malloc_sizes;

#if DEBUG
	/* This happens if someone tries to call
719
720
721
	 * kmem_cache_create(), or __kmalloc(), before
	 * the generic caches are initialized.
	 */
722
	BUG_ON(malloc_sizes[INDEX_AC].cs_cachep == NULL);
Linus Torvalds's avatar
Linus Torvalds committed
723
724
725
726
727
#endif
	while (size > csizep->cs_size)
		csizep++;

	/*
728
	 * Really subtle: The last entry with cs->cs_size==ULONG_MAX
Linus Torvalds's avatar
Linus Torvalds committed
729
730
731
732
733
734
735
736
	 * 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;
}

737
struct kmem_cache *kmem_find_general_cachep(size_t size, gfp_t gfpflags)
738
739
740
741
742
{
	return __find_general_cachep(size, gfpflags);
}
EXPORT_SYMBOL(kmem_find_general_cachep);

743
static size_t slab_mgmt_size(size_t nr_objs, size_t align)
Linus Torvalds's avatar
Linus Torvalds committed
744
{
745
746
	return ALIGN(sizeof(struct slab)+nr_objs*sizeof(kmem_bufctl_t), align);
}
Linus Torvalds's avatar
Linus Torvalds committed
747

Andrew Morton's avatar
Andrew Morton committed
748
749
750
/*
 * Calculate the number of objects and left-over bytes for a given buffer size.
 */
751
752
753
754
755
756
757
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;
Linus Torvalds's avatar
Linus Torvalds committed
758

759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
	/*
	 * 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;
Linus Torvalds's avatar
Linus Torvalds committed
807
808
809
810
}

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

Andrew Morton's avatar
Andrew Morton committed
811
812
static void __slab_error(const char *function, struct kmem_cache *cachep,
			char *msg)
Linus Torvalds's avatar
Linus Torvalds committed
813
814
{
	printk(KERN_ERR "slab error in %s(): cache `%s': %s\n",
815
	       function, cachep->name, msg);
Linus Torvalds's avatar
Linus Torvalds committed
816
817
818
	dump_stack();
}

819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
#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)
834
		node = first_node(node_online_map);
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859

	__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

Linus Torvalds's avatar
Linus Torvalds committed
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
/*
 * 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) {
877
		init_reap_node(cpu);
Linus Torvalds's avatar
Linus Torvalds committed
878
879
880
881
882
		INIT_WORK(reap_work, cache_reap, NULL);
		schedule_delayed_work_on(cpu, reap_work, HZ + 3 * cpu);
	}
}

883
static struct array_cache *alloc_arraycache(int node, int entries,
884
					    int batchcount)
Linus Torvalds's avatar
Linus Torvalds committed
885
{
886
	int memsize = sizeof(void *) * entries + sizeof(struct array_cache);
Linus Torvalds's avatar
Linus Torvalds committed
887
888
	struct array_cache *nc = NULL;

889
	nc = kmalloc_node(memsize, GFP_KERNEL, node);
Linus Torvalds's avatar
Linus Torvalds committed
890
891
892
893
894
	if (nc) {
		nc->avail = 0;
		nc->limit = entries;
		nc->batchcount = batchcount;
		nc->touched = 0;
895
		spin_lock_init(&nc->lock);
Linus Torvalds's avatar
Linus Torvalds committed
896
897
898
899
	}
	return nc;
}

900
#ifdef CONFIG_NUMA
901
static void *__cache_alloc_node(struct kmem_cache *, gfp_t, int);
902

Pekka Enberg's avatar
Pekka Enberg committed
903
static struct array_cache **alloc_alien_cache(int node, int limit)
904
905
{
	struct array_cache **ac_ptr;
906
	int memsize = sizeof(void *) * MAX_NUMNODES;
907
908
909
910
911
912
913
914
915
916
917
918
919
	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]) {
920
				for (i--; i <= 0; i--)
921
922
923
924
925
926
927
928
929
					kfree(ac_ptr[i]);
				kfree(ac_ptr);
				return NULL;
			}
		}
	}
	return ac_ptr;
}

Pekka Enberg's avatar
Pekka Enberg committed
930
static void free_alien_cache(struct array_cache **ac_ptr)
931
932
933
934
935
936
{
	int i;

	if (!ac_ptr)
		return;
	for_each_node(i)
937
	    kfree(ac_ptr[i]);
938
939
940
	kfree(ac_ptr);
}

941
static void __drain_alien_cache(struct kmem_cache *cachep,
Pekka Enberg's avatar
Pekka Enberg committed
942
				struct array_cache *ac, int node)
943
944
945
946
947
{
	struct kmem_list3 *rl3 = cachep->nodelists[node];

	if (ac->avail) {
		spin_lock(&rl3->list_lock);
948
		free_block(cachep, ac->entry, ac->avail, node);
949
950
951
952
953
		ac->avail = 0;
		spin_unlock(&rl3->list_lock);
	}
}

954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
/*
 * 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];
		if (ac && ac->avail) {
			spin_lock_irq(&ac->lock);
			__drain_alien_cache(cachep, ac, node);
			spin_unlock_irq(&ac->lock);
		}
	}
}

Andrew Morton's avatar
Andrew Morton committed
971
972
static void drain_alien_cache(struct kmem_cache *cachep,
				struct array_cache **alien)
973
{
974
	int i = 0;
975
976
977
978
	struct array_cache *ac;
	unsigned long flags;

	for_each_online_node(i) {
979
		ac = alien[i];
980
981
982
983
984
985
986
987
		if (ac) {
			spin_lock_irqsave(&ac->lock, flags);
			__drain_alien_cache(cachep, ac, i);
			spin_unlock_irqrestore(&ac->lock, flags);
		}
	}
}
#else
988

989
#define drain_alien_cache(cachep, alien) do { } while (0)
990
#define reap_alien(cachep, l3) do { } while (0)
991

992
993
994
995
996
static inline struct array_cache **alloc_alien_cache(int node, int limit)
{
	return (struct array_cache **) 0x01020304ul;
}

997
998
999
static inline void free_alien_cache(struct array_cache **ac_ptr)
{
}
1000

1001
1002
#endif

Linus Torvalds's avatar
Linus Torvalds committed
1003
static int __devinit cpuup_callback(struct notifier_block *nfb,
1004
				    unsigned long action, void *hcpu)
Linus Torvalds's avatar
Linus Torvalds committed
1005
1006
{
	long cpu = (long)hcpu;
1007
	struct kmem_cache *cachep;
1008
1009
1010
	struct kmem_list3 *l3 = NULL;
	int node = cpu_to_node(cpu);
	int memsize = sizeof(struct kmem_list3);
Linus Torvalds's avatar
Linus Torvalds committed
1011
1012
1013

	switch (action) {
	case CPU_UP_PREPARE:
Ingo Molnar's avatar
Ingo Molnar committed
1014
		mutex_lock(&cache_chain_mutex);
Andrew Morton's avatar
Andrew Morton committed
1015
1016
		/*
		 * We need to do this right in the beginning since
1017
1018
1019
1020
1021
		 * 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
		 */

Linus Torvalds's avatar
Linus Torvalds committed
1022
		list_for_each_entry(cachep, &cache_chain, next) {
Andrew Morton's avatar
Andrew Morton committed
1023
1024
			/*
			 * Set up the size64 kmemlist for cpu before we can
1025
1026
1027
1028
			 * begin anything. Make sure some other cpu on this
			 * node has not already allocated this
			 */
			if (!cachep->nodelists[node]) {
Andrew Morton's avatar
Andrew Morton committed
1029
1030
				l3 = kmalloc_node(memsize, GFP_KERNEL, node);
				if (!l3)
1031
1032
1033
					goto bad;
				kmem_list3_init(l3);
				l3->next_reap = jiffies + REAPTIMEOUT_LIST3 +
1034
				    ((unsigned long)cachep) % REAPTIMEOUT_LIST3;
1035

1036
1037
1038
1039
1040
				/*
				 * The l3s don't come and go as CPUs come and
				 * go.  cache_chain_mutex is sufficient
				 * protection here.
				 */
1041
1042
				cachep->nodelists[node] = l3;
			}
Linus Torvalds's avatar
Linus Torvalds committed
1043

1044
1045
			spin_lock_irq(&cachep->nodelists[node]->list_lock);
			cachep->nodelists[node]->free_limit =
Andrew Morton's avatar
Andrew Morton committed
1046
1047
				(1 + nr_cpus_node(node)) *
				cachep->batchcount + cachep->num;
1048
1049
1050
			spin_unlock_irq(&cachep->nodelists[node]->list_lock);
		}

Andrew Morton's avatar
Andrew Morton committed
1051
1052
1053
1054
		/*
		 * Now we can go ahead with allocating the shared arrays and
		 * array caches
		 */
1055
		list_for_each_entry(cachep, &cache_chain, next) {
1056
			struct array_cache *nc;
1057
1058
			struct array_cache *shared;
			struct array_cache **alien;
1059

1060
			nc = alloc_arraycache(node, cachep->limit,
1061
						cachep->batchcount);
Linus Torvalds's avatar
Linus Torvalds committed
1062
1063
			if (!nc)
				goto bad;
1064
1065
1066
1067
1068
			shared = alloc_arraycache(node,
					cachep->shared * cachep->batchcount,
					0xbaadf00d);
			if (!shared)
				goto bad;
1069

1070
1071
1072
			alien = alloc_alien_cache(node, cachep->limit);
			if (!alien)
				goto bad;
Linus Torvalds's avatar
Linus Torvalds committed
1073
			cachep->array[cpu] = nc;
1074
1075
1076
			l3 = cachep->nodelists[node];
			BUG_ON(!l3);

1077
1078
1079
1080
1081
1082
1083
1084
			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;
1085
			}
1086
1087
1088
1089
1090
1091
1092
1093
1094
#ifdef CONFIG_NUMA
			if (!l3->alien) {
				l3->alien = alien;
				alien = NULL;
			}
#endif
			spin_unlock_irq(&l3->list_lock);
			kfree(shared);
			free_alien_cache(alien);
Linus Torvalds's avatar
Linus Torvalds committed
1095
		}
Ingo Molnar's avatar
Ingo Molnar committed
1096
		mutex_unlock(&cache_chain_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
1097
1098
1099
1100
1101
1102
		break;
	case CPU_ONLINE:
		start_cpu_timer(cpu);
		break;
#ifdef CONFIG_HOTPLUG_CPU
	case CPU_DEAD:
1103
1104
1105
1106
1107
1108
1109
1110
		/*
		 * 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().
		 */
Linus Torvalds's avatar
Linus Torvalds committed
1111
1112
		/* fall thru */
	case CPU_UP_CANCELED:
Ingo Molnar's avatar
Ingo Molnar committed
1113
		mutex_lock(&cache_chain_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
1114
1115
		list_for_each_entry(cachep, &cache_chain, next) {
			struct array_cache *nc;
1116
1117
			struct array_cache *shared;
			struct array_cache **alien;
1118
			cpumask_t mask;
Linus Torvalds's avatar
Linus Torvalds committed
1119

1120
			mask = node_to_cpumask(node);
Linus Torvalds's avatar
Linus Torvalds committed
1121
1122
1123
			/* cpu is dead; no one can alloc from it. */
			nc = cachep->array[cpu];
			cachep->array[cpu] = NULL;
1124
1125
1126
			l3 = cachep->nodelists[node];

			if (!l3)
1127
				goto free_array_cache;
1128

1129
			spin_lock_irq(&l3->list_lock);
1130
1131
1132
1133

			/* Free limit for this kmem_list3 */
			l3->free_limit -= cachep->batchcount;
			if (nc)
1134
				free_block(cachep, nc->entry, nc->avail, node);
1135
1136

			if (!cpus_empty(mask)) {
1137
				spin_unlock_irq(&l3->list_lock);
1138
				goto free_array_cache;
1139
			}
1140

1141
1142
			shared = l3->shared;
			if (shared) {
1143
				free_block(cachep, l3->shared->entry,
1144
					   l3->shared->avail, node);
1145
1146
1147
				l3->shared = NULL;
			}

1148
1149
1150
1151
1152
1153
1154
1155
1156
			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);
1157
			}
1158
free_array_cache:
Linus Torvalds's avatar
Linus Torvalds committed
1159
1160
			kfree(nc);
		}
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
		/*
		 * 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;
			spin_lock_irq(&l3->list_lock);
			/* free slabs belonging to this node */
			__node_shrink(cachep, node);
			spin_unlock_irq(&l3->list_lock);
		}
Ingo Molnar's avatar
Ingo Molnar committed
1175
		mutex_unlock(&cache_chain_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
1176
1177
1178
1179
		break;
#endif
	}
	return NOTIFY_OK;
Andrew Morton's avatar
Andrew Morton committed
1180
bad:
Ingo Molnar's avatar
Ingo Molnar committed
1181
	mutex_unlock(&cache_chain_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
1182
1183
1184
1185
1186
	return NOTIFY_BAD;
}

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

1187
1188
1189
/*
 * swap the static kmem_list3 with kmalloced memory
 */
Andrew Morton's avatar
Andrew Morton committed
1190
1191
static void init_list(struct kmem_cache *cachep, struct kmem_list3 *list,
			int nodeid)
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
{
	struct kmem_list3 *ptr;

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

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

Andrew Morton's avatar
Andrew Morton committed
1206
1207
1208
/*
 * Initialisation.  Called after the page allocator have been initialised and
 * before smp_init().
Linus Torvalds's avatar
Linus Torvalds committed
1209
1210
1211
1212
1213
1214
 */
void __init kmem_cache_init(void)
{
	size_t left_over;
	struct cache_sizes *sizes;
	struct cache_names *names;
1215
	int i;
1216
	int order;
1217
1218
1219
1220
1221
1222

	for (i = 0; i < NUM_INIT_LISTS; i++) {
		kmem_list3_init(&initkmem_list3[i]);
		if (i < MAX_NUMNODES)
			cache_cache.nodelists[i] = NULL;
	}
Linus Torvalds's avatar
Linus Torvalds committed
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232

	/*
	 * 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:
Andrew Morton's avatar
Andrew Morton committed
1233
1234
1235
	 * 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.
1236
1237
1238
	 *    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.
Linus Torvalds's avatar
Linus Torvalds committed
1239
	 * 2) Create the first kmalloc cache.
1240
	 *    The struct kmem_cache for the new cache is allocated normally.
1241
1242
1243
	 *    An __init data area is used for the head array.
	 * 3) Create the remaining kmalloc caches, with minimally sized
	 *    head arrays.
Linus Torvalds's avatar
Linus Torvalds committed
1244
1245
	 * 4) Replace the __init data head arrays for cache_cache and the first
	 *    kmalloc cache with kmalloc allocated arrays.
1246
1247
1248
	 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
1249
1250
1251
1252
1253
1254
1255
	 */

	/* 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;
1256
	cache_cache.nodelists[numa_node_id()] = &initkmem_list3[CACHE_CACHE];
Linus Torvalds's avatar
Linus Torvalds committed
1257

Andrew Morton's avatar
Andrew Morton committed
1258
1259
	cache_cache.buffer_size = ALIGN(cache_cache.buffer_size,
					cache_line_size());
Linus Torvalds's avatar
Linus Torvalds committed
1260

1261
1262
1263
1264
1265
1266
	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;
	}
Linus Torvalds's avatar
Linus Torvalds committed
1267
1268
	if (!cache_cache.num)
		BUG();
1269
	cache_cache.gfporder = order;
1270
1271
1272
	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());
Linus Torvalds's avatar
Linus Torvalds committed
1273
1274
1275
1276
1277

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

Andrew Morton's avatar
Andrew Morton committed
1278
1279
1280
1281
	/*
	 * Initialize the caches that provide memory for the array cache and the
	 * kmem_list3 structures first.  Without this, further allocations will
	 * bug.
1282
1283
1284
	 */

	sizes[INDEX_AC].cs_cachep = kmem_cache_create(names[INDEX_AC].name,
Andrew Morton's avatar
Andrew Morton committed
1285
1286
1287
1288
					sizes[INDEX_AC].cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_PANIC,
					NULL, NULL);
1289

Andrew Morton's avatar
Andrew Morton committed
1290
	if (INDEX_AC != INDEX_L3) {
1291
		sizes[INDEX_L3].cs_cachep =
Andrew Morton's avatar
Andrew Morton committed
1292
1293
1294
1295
1296
1297
			kmem_cache_create(names[INDEX_L3].name,
				sizes[INDEX_L3].cs_size,
				ARCH_KMALLOC_MINALIGN,
				ARCH_KMALLOC_FLAGS|SLAB_PANIC,
				NULL, NULL);
	}
1298

Linus Torvalds's avatar
Linus Torvalds committed
1299
	while (sizes->cs_size != ULONG_MAX) {
1300
1301
		/*
		 * For performance, all the general caches are L1 aligned.
Linus Torvalds's avatar
Linus Torvalds committed
1302
1303
1304
		 * This should be particularly beneficial on SMP boxes, as it
		 * eliminates "false sharing".
		 * Note for systems short on memory removing the alignment will
1305
1306
		 * allow tighter packing of the smaller caches.
		 */
Andrew Morton's avatar
Andrew Morton committed
1307
		if (!sizes->cs_cachep) {
1308
			sizes->cs_cachep = kmem_cache_create(names->name,
Andrew Morton's avatar
Andrew Morton committed
1309
1310
1311
1312
1313
					sizes->cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_PANIC,
					NULL, NULL);
		}
Linus Torvalds's avatar
Linus Torvalds committed
1314
1315
1316

		/* Inc off-slab bufctl limit until the ceiling is hit. */
		if (!(OFF_SLAB(sizes->cs_cachep))) {
1317
			offslab_limit = sizes->cs_size - sizeof(struct slab);
Linus Torvalds's avatar
Linus Torvalds committed
1318
1319
1320
1321
			offslab_limit /= sizeof(kmem_bufctl_t);
		}

		sizes->cs_dmacachep = kmem_cache_create(names->name_dma,
Andrew Morton's avatar
Andrew Morton committed
1322
1323
1324
1325
1326
					sizes->cs_size,
					ARCH_KMALLOC_MINALIGN,
					ARCH_KMALLOC_FLAGS|SLAB_CACHE_DMA|
						SLAB_PANIC,
					NULL, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
1327
1328
1329
1330
1331
		sizes++;
		names++;
	}
	/* 4) Replace the bootstrap head arrays */
	{
1332
		void *ptr;
1333

Linus Torvalds's avatar
Linus Torvalds committed
1334
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
1335

Linus Torvalds's avatar
Linus Torvalds committed
1336
		local_irq_disable();
1337
1338
		BUG_ON(cpu_cache_get(&cache_cache) != &initarray_cache.cache);
		memcpy(ptr, cpu_cache_get(&cache_cache),
1339
		       sizeof(struct arraycache_init));
Linus Torvalds's avatar
Linus Torvalds committed
1340
1341
		cache_cache.array[smp_processor_id()] = ptr;
		local_irq_enable();
1342

Linus Torvalds's avatar
Linus Torvalds committed
1343
		ptr = kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
1344

Linus Torvalds's avatar
Linus Torvalds committed
1345
		local_irq_disable();
1346
		BUG_ON(cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep)
1347
		       != &initarray_generic.cache);
1348
		memcpy(ptr, cpu_cache_get(malloc_sizes[INDEX_AC].cs_cachep),
1349
		       sizeof(struct arraycache_init));
1350
		malloc_sizes[INDEX_AC].cs_cachep->array[smp_processor_id()] =
1351
		    ptr;
Linus Torvalds's avatar
Linus Torvalds committed
1352
1353
		local_irq_enable();
	}
1354
1355
1356
1357
1358
	/* 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],