slub.c 127 KB
Newer Older
Christoph Lameter's avatar
Christoph Lameter committed
1
2
3
4
/*
 * SLUB: A slab allocator that limits cache line use instead of queuing
 * objects in per cpu and per node lists.
 *
5
6
 * The allocator synchronizes using per slab locks or atomic operatios
 * and only uses a centralized lock to manage a pool of partial slabs.
Christoph Lameter's avatar
Christoph Lameter committed
7
 *
Christoph Lameter's avatar
Christoph Lameter committed
8
 * (C) 2007 SGI, Christoph Lameter
9
 * (C) 2011 Linux Foundation, Christoph Lameter
Christoph Lameter's avatar
Christoph Lameter committed
10
11
12
 */

#include <linux/mm.h>
Nick Piggin's avatar
Nick Piggin committed
13
#include <linux/swap.h> /* struct reclaim_state */
Christoph Lameter's avatar
Christoph Lameter committed
14
15
16
17
18
#include <linux/module.h>
#include <linux/bit_spinlock.h>
#include <linux/interrupt.h>
#include <linux/bitops.h>
#include <linux/slab.h>
19
#include "slab.h"
20
#include <linux/proc_fs.h>
21
#include <linux/notifier.h>
Christoph Lameter's avatar
Christoph Lameter committed
22
#include <linux/seq_file.h>
Vegard Nossum's avatar
Vegard Nossum committed
23
#include <linux/kmemcheck.h>
Christoph Lameter's avatar
Christoph Lameter committed
24
25
26
27
#include <linux/cpu.h>
#include <linux/cpuset.h>
#include <linux/mempolicy.h>
#include <linux/ctype.h>
28
#include <linux/debugobjects.h>
Christoph Lameter's avatar
Christoph Lameter committed
29
#include <linux/kallsyms.h>
30
#include <linux/memory.h>
Roman Zippel's avatar
Roman Zippel committed
31
#include <linux/math64.h>
Akinobu Mita's avatar
Akinobu Mita committed
32
#include <linux/fault-inject.h>
33
#include <linux/stacktrace.h>
34
#include <linux/prefetch.h>
35
#include <linux/memcontrol.h>
Christoph Lameter's avatar
Christoph Lameter committed
36

37
38
#include <trace/events/kmem.h>

39
40
#include "internal.h"

Christoph Lameter's avatar
Christoph Lameter committed
41
42
/*
 * Lock order:
43
 *   1. slab_mutex (Global Mutex)
44
45
 *   2. node->list_lock
 *   3. slab_lock(page) (Only on some arches and for debugging)
Christoph Lameter's avatar
Christoph Lameter committed
46
 *
47
 *   slab_mutex
48
 *
49
 *   The role of the slab_mutex is to protect the list of all the slabs
50
51
52
53
54
55
56
57
58
59
60
61
62
63
 *   and to synchronize major metadata changes to slab cache structures.
 *
 *   The slab_lock is only used for debugging and on arches that do not
 *   have the ability to do a cmpxchg_double. It only protects the second
 *   double word in the page struct. Meaning
 *	A. page->freelist	-> List of object free in a page
 *	B. page->counters	-> Counters of objects
 *	C. page->frozen		-> frozen state
 *
 *   If a slab is frozen then it is exempt from list management. It is not
 *   on any list. The processor that froze the slab is the one who can
 *   perform list operations on the page. Other processors may put objects
 *   onto the freelist but the processor that froze the slab is the only
 *   one that can retrieve the objects from the page's freelist.
Christoph Lameter's avatar
Christoph Lameter committed
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
 *
 *   The list_lock protects the partial and full list on each node and
 *   the partial slab counter. If taken then no new slabs may be added or
 *   removed from the lists nor make the number of partial slabs be modified.
 *   (Note that the total number of slabs is an atomic value that may be
 *   modified without taking the list lock).
 *
 *   The list_lock is a centralized lock and thus we avoid taking it as
 *   much as possible. As long as SLUB does not have to handle partial
 *   slabs, operations can continue without any centralized lock. F.e.
 *   allocating a long series of objects that fill up slabs does not require
 *   the list lock.
 *   Interrupts are disabled during allocation and deallocation in order to
 *   make the slab allocator safe to use in the context of an irq. In addition
 *   interrupts are disabled to ensure that the processor does not change
 *   while handling per_cpu slabs, due to kernel preemption.
 *
 * SLUB assigns one slab for allocation to each processor.
 * Allocations only occur from these slabs called cpu slabs.
 *
Christoph Lameter's avatar
Christoph Lameter committed
84
85
 * Slabs with free elements are kept on a partial list and during regular
 * operations no list for full slabs is used. If an object in a full slab is
Christoph Lameter's avatar
Christoph Lameter committed
86
 * freed then the slab will show up again on the partial lists.
Christoph Lameter's avatar
Christoph Lameter committed
87
88
 * We track full slabs for debugging purposes though because otherwise we
 * cannot scan all objects.
Christoph Lameter's avatar
Christoph Lameter committed
89
90
91
92
93
94
95
 *
 * Slabs are freed when they become empty. Teardown and setup is
 * minimal so we rely on the page allocators per cpu caches for
 * fast frees and allocs.
 *
 * Overloading of page flags that are otherwise used for LRU management.
 *
96
97
98
99
100
101
102
103
104
105
106
107
 * PageActive 		The slab is frozen and exempt from list processing.
 * 			This means that the slab is dedicated to a purpose
 * 			such as satisfying allocations for a specific
 * 			processor. Objects may be freed in the slab while
 * 			it is frozen but slab_free will then skip the usual
 * 			list operations. It is up to the processor holding
 * 			the slab to integrate the slab into the slab lists
 * 			when the slab is no longer needed.
 *
 * 			One use of this flag is to mark slabs that are
 * 			used for allocations. Then such a slab becomes a cpu
 * 			slab. The cpu slab may be equipped with an additional
108
 * 			freelist that allows lockless access to
109
110
 * 			free objects in addition to the regular freelist
 * 			that requires the slab lock.
Christoph Lameter's avatar
Christoph Lameter committed
111
112
113
 *
 * PageError		Slab requires special handling due to debug
 * 			options set. This moves	slab handling out of
114
 * 			the fast path and disables lockless freelists.
Christoph Lameter's avatar
Christoph Lameter committed
115
116
 */

117
118
static inline int kmem_cache_debug(struct kmem_cache *s)
{
119
#ifdef CONFIG_SLUB_DEBUG
120
	return unlikely(s->flags & SLAB_DEBUG_FLAGS);
121
#else
122
	return 0;
123
#endif
124
}
125

126
127
128
129
130
131
132
133
134
static inline bool kmem_cache_has_cpu_partial(struct kmem_cache *s)
{
#ifdef CONFIG_SLUB_CPU_PARTIAL
	return !kmem_cache_debug(s);
#else
	return false;
#endif
}

Christoph Lameter's avatar
Christoph Lameter committed
135
136
137
138
139
140
141
142
143
144
145
/*
 * Issues still to be resolved:
 *
 * - Support PAGE_ALLOC_DEBUG. Should be easy to do.
 *
 * - Variable sizing of the per node arrays
 */

/* Enable to test recovery from slab corruption on boot */
#undef SLUB_RESILIENCY_TEST

146
147
148
/* Enable to log cmpxchg failures */
#undef SLUB_DEBUG_CMPXCHG

149
150
151
152
/*
 * Mininum number of partial slabs. These will be left on the partial
 * lists even if they are empty. kmem_cache_shrink may reclaim them.
 */
153
#define MIN_PARTIAL 5
Christoph Lameter's avatar
Christoph Lameter committed
154

155
156
157
/*
 * Maximum number of desirable partial slabs.
 * The existence of more partial slabs makes kmem_cache_shrink
158
 * sort the partial list by the number of objects in use.
159
160
161
 */
#define MAX_PARTIAL 10

Christoph Lameter's avatar
Christoph Lameter committed
162
163
#define DEBUG_DEFAULT_FLAGS (SLAB_DEBUG_FREE | SLAB_RED_ZONE | \
				SLAB_POISON | SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
164

165
/*
166
167
168
 * Debugging flags that require metadata to be stored in the slab.  These get
 * disabled when slub_debug=O is used and a cache's min order increases with
 * metadata.
169
 */
170
#define DEBUG_METADATA_FLAGS (SLAB_RED_ZONE | SLAB_POISON | SLAB_STORE_USER)
171

Christoph Lameter's avatar
Christoph Lameter committed
172
173
174
175
/*
 * Set of flags that will prevent slab merging
 */
#define SLUB_NEVER_MERGE (SLAB_RED_ZONE | SLAB_POISON | SLAB_STORE_USER | \
176
177
		SLAB_TRACE | SLAB_DESTROY_BY_RCU | SLAB_NOLEAKTRACE | \
		SLAB_FAILSLAB)
Christoph Lameter's avatar
Christoph Lameter committed
178
179

#define SLUB_MERGE_SAME (SLAB_DEBUG_FREE | SLAB_RECLAIM_ACCOUNT | \
Vegard Nossum's avatar
Vegard Nossum committed
180
		SLAB_CACHE_DMA | SLAB_NOTRACK)
Christoph Lameter's avatar
Christoph Lameter committed
181

182
183
#define OO_SHIFT	16
#define OO_MASK		((1 << OO_SHIFT) - 1)
184
#define MAX_OBJS_PER_PAGE	32767 /* since page.objects is u15 */
185

Christoph Lameter's avatar
Christoph Lameter committed
186
/* Internal SLUB flags */
Christoph Lameter's avatar
Christoph Lameter committed
187
#define __OBJECT_POISON		0x80000000UL /* Poison object */
188
#define __CMPXCHG_DOUBLE	0x40000000UL /* Use cmpxchg_double */
Christoph Lameter's avatar
Christoph Lameter committed
189
190
191
192
193

#ifdef CONFIG_SMP
static struct notifier_block slab_notifier;
#endif

194
195
196
/*
 * Tracking user of a slab.
 */
197
#define TRACK_ADDRS_COUNT 16
198
struct track {
199
	unsigned long addr;	/* Called from address */
200
201
202
#ifdef CONFIG_STACKTRACE
	unsigned long addrs[TRACK_ADDRS_COUNT];	/* Called from address */
#endif
203
204
205
206
207
208
209
	int cpu;		/* Was running on cpu */
	int pid;		/* Pid context */
	unsigned long when;	/* When did the operation occur */
};

enum track_item { TRACK_ALLOC, TRACK_FREE };

210
#ifdef CONFIG_SYSFS
Christoph Lameter's avatar
Christoph Lameter committed
211
212
213
static int sysfs_slab_add(struct kmem_cache *);
static int sysfs_slab_alias(struct kmem_cache *, const char *);
static void sysfs_slab_remove(struct kmem_cache *);
214
static void memcg_propagate_slab_attrs(struct kmem_cache *s);
Christoph Lameter's avatar
Christoph Lameter committed
215
#else
216
217
218
static inline int sysfs_slab_add(struct kmem_cache *s) { return 0; }
static inline int sysfs_slab_alias(struct kmem_cache *s, const char *p)
							{ return 0; }
219
static inline void sysfs_slab_remove(struct kmem_cache *s) { }
220

221
static inline void memcg_propagate_slab_attrs(struct kmem_cache *s) { }
Christoph Lameter's avatar
Christoph Lameter committed
222
223
#endif

224
static inline void stat(const struct kmem_cache *s, enum stat_item si)
225
226
{
#ifdef CONFIG_SLUB_STATS
227
	__this_cpu_inc(s->cpu_slab->stat[si]);
228
229
230
#endif
}

Christoph Lameter's avatar
Christoph Lameter committed
231
232
233
234
235
236
237
238
239
/********************************************************************
 * 			Core slab cache functions
 *******************************************************************/

static inline struct kmem_cache_node *get_node(struct kmem_cache *s, int node)
{
	return s->node[node];
}

Christoph Lameter's avatar
Christoph Lameter committed
240
/* Verify that a pointer has an address that is valid within a slab page */
241
242
243
244
245
static inline int check_valid_pointer(struct kmem_cache *s,
				struct page *page, const void *object)
{
	void *base;

246
	if (!object)
247
248
		return 1;

249
	base = page_address(page);
250
	if (object < base || object >= base + page->objects * s->size ||
251
252
253
254
255
256
257
		(object - base) % s->size) {
		return 0;
	}

	return 1;
}

258
259
260
261
262
static inline void *get_freepointer(struct kmem_cache *s, void *object)
{
	return *(void **)(object + s->offset);
}

263
264
265
266
267
static void prefetch_freepointer(const struct kmem_cache *s, void *object)
{
	prefetch(object + s->offset);
}

268
269
270
271
272
273
274
275
276
277
278
279
static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
{
	void *p;

#ifdef CONFIG_DEBUG_PAGEALLOC
	probe_kernel_read(&p, (void **)(object + s->offset), sizeof(p));
#else
	p = get_freepointer(s, object);
#endif
	return p;
}

280
281
282
283
284
285
static inline void set_freepointer(struct kmem_cache *s, void *object, void *fp)
{
	*(void **)(object + s->offset) = fp;
}

/* Loop over all objects in a slab */
286
287
#define for_each_object(__p, __s, __addr, __objects) \
	for (__p = (__addr); __p < (__addr) + (__objects) * (__s)->size;\
288
289
290
291
292
293
294
295
			__p += (__s)->size)

/* Determine object index from a given position */
static inline int slab_index(void *p, struct kmem_cache *s, void *addr)
{
	return (p - addr) / s->size;
}

296
297
298
299
300
301
302
303
static inline size_t slab_ksize(const struct kmem_cache *s)
{
#ifdef CONFIG_SLUB_DEBUG
	/*
	 * Debugging requires use of the padding between object
	 * and whatever may come after it.
	 */
	if (s->flags & (SLAB_RED_ZONE | SLAB_POISON))
304
		return s->object_size;
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319

#endif
	/*
	 * If we have the need to store the freelist pointer
	 * back there or track user information then we can
	 * only use the space before that information.
	 */
	if (s->flags & (SLAB_DESTROY_BY_RCU | SLAB_STORE_USER))
		return s->inuse;
	/*
	 * Else we can use all the padding etc for the allocation
	 */
	return s->size;
}

320
321
322
323
324
static inline int order_objects(int order, unsigned long size, int reserved)
{
	return ((PAGE_SIZE << order) - reserved) / size;
}

325
static inline struct kmem_cache_order_objects oo_make(int order,
326
		unsigned long size, int reserved)
327
328
{
	struct kmem_cache_order_objects x = {
329
		(order << OO_SHIFT) + order_objects(order, size, reserved)
330
331
332
333
334
335
336
	};

	return x;
}

static inline int oo_order(struct kmem_cache_order_objects x)
{
337
	return x.x >> OO_SHIFT;
338
339
340
341
}

static inline int oo_objects(struct kmem_cache_order_objects x)
{
342
	return x.x & OO_MASK;
343
344
}

345
346
347
348
349
350
351
352
353
354
355
356
357
/*
 * Per slab locking using the pagelock
 */
static __always_inline void slab_lock(struct page *page)
{
	bit_spin_lock(PG_locked, &page->flags);
}

static __always_inline void slab_unlock(struct page *page)
{
	__bit_spin_unlock(PG_locked, &page->flags);
}

358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
static inline void set_page_slub_counters(struct page *page, unsigned long counters_new)
{
	struct page tmp;
	tmp.counters = counters_new;
	/*
	 * page->counters can cover frozen/inuse/objects as well
	 * as page->_count.  If we assign to ->counters directly
	 * we run the risk of losing updates to page->_count, so
	 * be careful and only assign to the fields we need.
	 */
	page->frozen  = tmp.frozen;
	page->inuse   = tmp.inuse;
	page->objects = tmp.objects;
}

373
374
375
376
377
378
379
/* Interrupts must be disabled (for the fallback code to work right) */
static inline bool __cmpxchg_double_slab(struct kmem_cache *s, struct page *page,
		void *freelist_old, unsigned long counters_old,
		void *freelist_new, unsigned long counters_new,
		const char *n)
{
	VM_BUG_ON(!irqs_disabled());
380
381
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
382
	if (s->flags & __CMPXCHG_DOUBLE) {
383
		if (cmpxchg_double(&page->freelist, &page->counters,
384
385
386
387
388
389
390
			freelist_old, counters_old,
			freelist_new, counters_new))
		return 1;
	} else
#endif
	{
		slab_lock(page);
391
392
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
393
			page->freelist = freelist_new;
394
			set_page_slub_counters(page, counters_new);
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
			slab_unlock(page);
			return 1;
		}
		slab_unlock(page);
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
	printk(KERN_INFO "%s %s: cmpxchg double redo ", n, s->name);
#endif

	return 0;
}

411
412
413
414
415
static inline bool cmpxchg_double_slab(struct kmem_cache *s, struct page *page,
		void *freelist_old, unsigned long counters_old,
		void *freelist_new, unsigned long counters_new,
		const char *n)
{
416
417
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
418
	if (s->flags & __CMPXCHG_DOUBLE) {
419
		if (cmpxchg_double(&page->freelist, &page->counters,
420
421
422
423
424
425
			freelist_old, counters_old,
			freelist_new, counters_new))
		return 1;
	} else
#endif
	{
426
427
428
		unsigned long flags;

		local_irq_save(flags);
429
		slab_lock(page);
430
431
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
432
			page->freelist = freelist_new;
433
			set_page_slub_counters(page, counters_new);
434
			slab_unlock(page);
435
			local_irq_restore(flags);
436
437
			return 1;
		}
438
		slab_unlock(page);
439
		local_irq_restore(flags);
440
441
442
443
444
445
446
447
448
449
450
451
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
	printk(KERN_INFO "%s %s: cmpxchg double redo ", n, s->name);
#endif

	return 0;
}

452
#ifdef CONFIG_SLUB_DEBUG
453
454
455
/*
 * Determine a map of object in use on a page.
 *
456
 * Node listlock must be held to guarantee that the page does
457
458
459
460
461
462
463
464
465
466
467
 * not vanish from under us.
 */
static void get_map(struct kmem_cache *s, struct page *page, unsigned long *map)
{
	void *p;
	void *addr = page_address(page);

	for (p = page->freelist; p; p = get_freepointer(s, p))
		set_bit(slab_index(p, s, addr), map);
}

468
469
470
/*
 * Debug settings:
 */
471
472
473
#ifdef CONFIG_SLUB_DEBUG_ON
static int slub_debug = DEBUG_DEFAULT_FLAGS;
#else
474
static int slub_debug;
475
#endif
476
477

static char *slub_debug_slabs;
478
static int disable_higher_order_debug;
479

Christoph Lameter's avatar
Christoph Lameter committed
480
481
482
483
484
/*
 * Object debugging
 */
static void print_section(char *text, u8 *addr, unsigned int length)
{
485
486
	print_hex_dump(KERN_ERR, text, DUMP_PREFIX_ADDRESS, 16, 1, addr,
			length, 1);
Christoph Lameter's avatar
Christoph Lameter committed
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
}

static struct track *get_track(struct kmem_cache *s, void *object,
	enum track_item alloc)
{
	struct track *p;

	if (s->offset)
		p = object + s->offset + sizeof(void *);
	else
		p = object + s->inuse;

	return p + alloc;
}

static void set_track(struct kmem_cache *s, void *object,
503
			enum track_item alloc, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
504
{
Akinobu Mita's avatar
Akinobu Mita committed
505
	struct track *p = get_track(s, object, alloc);
Christoph Lameter's avatar
Christoph Lameter committed
506
507

	if (addr) {
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
#ifdef CONFIG_STACKTRACE
		struct stack_trace trace;
		int i;

		trace.nr_entries = 0;
		trace.max_entries = TRACK_ADDRS_COUNT;
		trace.entries = p->addrs;
		trace.skip = 3;
		save_stack_trace(&trace);

		/* See rant in lockdep.c */
		if (trace.nr_entries != 0 &&
		    trace.entries[trace.nr_entries - 1] == ULONG_MAX)
			trace.nr_entries--;

		for (i = trace.nr_entries; i < TRACK_ADDRS_COUNT; i++)
			p->addrs[i] = 0;
#endif
Christoph Lameter's avatar
Christoph Lameter committed
526
527
		p->addr = addr;
		p->cpu = smp_processor_id();
528
		p->pid = current->pid;
Christoph Lameter's avatar
Christoph Lameter committed
529
530
531
532
533
534
535
		p->when = jiffies;
	} else
		memset(p, 0, sizeof(struct track));
}

static void init_tracking(struct kmem_cache *s, void *object)
{
536
537
538
	if (!(s->flags & SLAB_STORE_USER))
		return;

539
540
	set_track(s, object, TRACK_FREE, 0UL);
	set_track(s, object, TRACK_ALLOC, 0UL);
Christoph Lameter's avatar
Christoph Lameter committed
541
542
543
544
545
546
547
}

static void print_track(const char *s, struct track *t)
{
	if (!t->addr)
		return;

548
	printk(KERN_ERR "INFO: %s in %pS age=%lu cpu=%u pid=%d\n",
549
		s, (void *)t->addr, jiffies - t->when, t->cpu, t->pid);
550
551
552
553
554
555
556
557
558
559
#ifdef CONFIG_STACKTRACE
	{
		int i;
		for (i = 0; i < TRACK_ADDRS_COUNT; i++)
			if (t->addrs[i])
				printk(KERN_ERR "\t%pS\n", (void *)t->addrs[i]);
			else
				break;
	}
#endif
560
561
562
563
564
565
566
567
568
569
570
571
572
}

static void print_tracking(struct kmem_cache *s, void *object)
{
	if (!(s->flags & SLAB_STORE_USER))
		return;

	print_track("Allocated", get_track(s, object, TRACK_ALLOC));
	print_track("Freed", get_track(s, object, TRACK_FREE));
}

static void print_page_info(struct page *page)
{
573
574
575
	printk(KERN_ERR
	       "INFO: Slab 0x%p objects=%u used=%u fp=0x%p flags=0x%04lx\n",
	       page, page->objects, page->inuse, page->freelist, page->flags);
576
577
578
579
580
581
582
583
584
585
586
587
588

}

static void slab_bug(struct kmem_cache *s, char *fmt, ...)
{
	va_list args;
	char buf[100];

	va_start(args, fmt);
	vsnprintf(buf, sizeof(buf), fmt, args);
	va_end(args);
	printk(KERN_ERR "========================================"
			"=====================================\n");
589
	printk(KERN_ERR "BUG %s (%s): %s\n", s->name, print_tainted(), buf);
590
591
	printk(KERN_ERR "----------------------------------------"
			"-------------------------------------\n\n");
592

593
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
Christoph Lameter's avatar
Christoph Lameter committed
594
595
}

596
597
598
599
600
601
602
603
604
605
606
607
static void slab_fix(struct kmem_cache *s, char *fmt, ...)
{
	va_list args;
	char buf[100];

	va_start(args, fmt);
	vsnprintf(buf, sizeof(buf), fmt, args);
	va_end(args);
	printk(KERN_ERR "FIX %s: %s\n", s->name, buf);
}

static void print_trailer(struct kmem_cache *s, struct page *page, u8 *p)
Christoph Lameter's avatar
Christoph Lameter committed
608
609
{
	unsigned int off;	/* Offset of last byte */
610
	u8 *addr = page_address(page);
611
612
613
614
615
616
617
618
619

	print_tracking(s, p);

	print_page_info(page);

	printk(KERN_ERR "INFO: Object 0x%p @offset=%tu fp=0x%p\n\n",
			p, p - addr, get_freepointer(s, p));

	if (p > addr + 16)
620
		print_section("Bytes b4 ", p - 16, 16);
Christoph Lameter's avatar
Christoph Lameter committed
621

622
	print_section("Object ", p, min_t(unsigned long, s->object_size,
623
				PAGE_SIZE));
Christoph Lameter's avatar
Christoph Lameter committed
624
	if (s->flags & SLAB_RED_ZONE)
625
626
		print_section("Redzone ", p + s->object_size,
			s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
627
628
629
630
631
632

	if (s->offset)
		off = s->offset + sizeof(void *);
	else
		off = s->inuse;

633
	if (s->flags & SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
634
635
636
637
		off += 2 * sizeof(struct track);

	if (off != s->size)
		/* Beginning of the filler is the free pointer */
638
		print_section("Padding ", p + off, s->size - off);
639
640

	dump_stack();
Christoph Lameter's avatar
Christoph Lameter committed
641
642
643
644
645
}

static void object_err(struct kmem_cache *s, struct page *page,
			u8 *object, char *reason)
{
646
	slab_bug(s, "%s", reason);
647
	print_trailer(s, page, object);
Christoph Lameter's avatar
Christoph Lameter committed
648
649
}

650
651
static void slab_err(struct kmem_cache *s, struct page *page,
			const char *fmt, ...)
Christoph Lameter's avatar
Christoph Lameter committed
652
653
654
655
{
	va_list args;
	char buf[100];

656
657
	va_start(args, fmt);
	vsnprintf(buf, sizeof(buf), fmt, args);
Christoph Lameter's avatar
Christoph Lameter committed
658
	va_end(args);
659
	slab_bug(s, "%s", buf);
660
	print_page_info(page);
Christoph Lameter's avatar
Christoph Lameter committed
661
662
663
	dump_stack();
}

664
static void init_object(struct kmem_cache *s, void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
665
666
667
668
{
	u8 *p = object;

	if (s->flags & __OBJECT_POISON) {
669
670
		memset(p, POISON_FREE, s->object_size - 1);
		p[s->object_size - 1] = POISON_END;
Christoph Lameter's avatar
Christoph Lameter committed
671
672
673
	}

	if (s->flags & SLAB_RED_ZONE)
674
		memset(p + s->object_size, val, s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
675
676
}

677
678
679
680
681
682
683
684
685
static void restore_bytes(struct kmem_cache *s, char *message, u8 data,
						void *from, void *to)
{
	slab_fix(s, "Restoring 0x%p-0x%p=0x%x\n", from, to - 1, data);
	memset(from, data, to - from);
}

static int check_bytes_and_report(struct kmem_cache *s, struct page *page,
			u8 *object, char *what,
686
			u8 *start, unsigned int value, unsigned int bytes)
687
688
689
690
{
	u8 *fault;
	u8 *end;

691
	fault = memchr_inv(start, value, bytes);
692
693
694
695
696
697
698
699
700
701
702
703
704
705
	if (!fault)
		return 1;

	end = start + bytes;
	while (end > fault && end[-1] == value)
		end--;

	slab_bug(s, "%s overwritten", what);
	printk(KERN_ERR "INFO: 0x%p-0x%p. First byte 0x%x instead of 0x%x\n",
					fault, end - 1, fault[0], value);
	print_trailer(s, page, object);

	restore_bytes(s, what, value, fault, end);
	return 0;
Christoph Lameter's avatar
Christoph Lameter committed
706
707
708
709
710
711
712
713
714
}

/*
 * Object layout:
 *
 * object address
 * 	Bytes of the object to be managed.
 * 	If the freepointer may overlay the object then the free
 * 	pointer is the first word of the object.
Christoph Lameter's avatar
Christoph Lameter committed
715
 *
Christoph Lameter's avatar
Christoph Lameter committed
716
717
718
 * 	Poisoning uses 0x6b (POISON_FREE) and the last byte is
 * 	0xa5 (POISON_END)
 *
719
 * object + s->object_size
Christoph Lameter's avatar
Christoph Lameter committed
720
 * 	Padding to reach word boundary. This is also used for Redzoning.
Christoph Lameter's avatar
Christoph Lameter committed
721
 * 	Padding is extended by another word if Redzoning is enabled and
722
 * 	object_size == inuse.
Christoph Lameter's avatar
Christoph Lameter committed
723
 *
Christoph Lameter's avatar
Christoph Lameter committed
724
725
726
727
 * 	We fill with 0xbb (RED_INACTIVE) for inactive objects and with
 * 	0xcc (RED_ACTIVE) for objects in use.
 *
 * object + s->inuse
Christoph Lameter's avatar
Christoph Lameter committed
728
729
 * 	Meta data starts here.
 *
Christoph Lameter's avatar
Christoph Lameter committed
730
731
 * 	A. Free pointer (if we cannot overwrite object on free)
 * 	B. Tracking data for SLAB_STORE_USER
Christoph Lameter's avatar
Christoph Lameter committed
732
 * 	C. Padding to reach required alignment boundary or at mininum
Christoph Lameter's avatar
Christoph Lameter committed
733
 * 		one word if debugging is on to be able to detect writes
Christoph Lameter's avatar
Christoph Lameter committed
734
735
736
 * 		before the word boundary.
 *
 *	Padding is done using 0x5a (POISON_INUSE)
Christoph Lameter's avatar
Christoph Lameter committed
737
738
 *
 * object + s->size
Christoph Lameter's avatar
Christoph Lameter committed
739
 * 	Nothing is used beyond s->size.
Christoph Lameter's avatar
Christoph Lameter committed
740
 *
741
 * If slabcaches are merged then the object_size and inuse boundaries are mostly
Christoph Lameter's avatar
Christoph Lameter committed
742
 * ignored. And therefore no slab options that rely on these boundaries
Christoph Lameter's avatar
Christoph Lameter committed
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
 * may be used with merged slabcaches.
 */

static int check_pad_bytes(struct kmem_cache *s, struct page *page, u8 *p)
{
	unsigned long off = s->inuse;	/* The end of info */

	if (s->offset)
		/* Freepointer is placed after the object. */
		off += sizeof(void *);

	if (s->flags & SLAB_STORE_USER)
		/* We also have user information there */
		off += 2 * sizeof(struct track);

	if (s->size == off)
		return 1;

761
762
	return check_bytes_and_report(s, page, p, "Object padding",
				p + off, POISON_INUSE, s->size - off);
Christoph Lameter's avatar
Christoph Lameter committed
763
764
}

765
/* Check the pad bytes at the end of a slab page */
Christoph Lameter's avatar
Christoph Lameter committed
766
767
static int slab_pad_check(struct kmem_cache *s, struct page *page)
{
768
769
770
771
772
	u8 *start;
	u8 *fault;
	u8 *end;
	int length;
	int remainder;
Christoph Lameter's avatar
Christoph Lameter committed
773
774
775
776

	if (!(s->flags & SLAB_POISON))
		return 1;

777
	start = page_address(page);
778
	length = (PAGE_SIZE << compound_order(page)) - s->reserved;
779
780
	end = start + length;
	remainder = length % s->size;
Christoph Lameter's avatar
Christoph Lameter committed
781
782
783
	if (!remainder)
		return 1;

784
	fault = memchr_inv(end - remainder, POISON_INUSE, remainder);
785
786
787
788
789
790
	if (!fault)
		return 1;
	while (end > fault && end[-1] == POISON_INUSE)
		end--;

	slab_err(s, page, "Padding overwritten. 0x%p-0x%p", fault, end - 1);
791
	print_section("Padding ", end - remainder, remainder);
792

Eric Dumazet's avatar
Eric Dumazet committed
793
	restore_bytes(s, "slab padding", POISON_INUSE, end - remainder, end);
794
	return 0;
Christoph Lameter's avatar
Christoph Lameter committed
795
796
797
}

static int check_object(struct kmem_cache *s, struct page *page,
798
					void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
799
800
{
	u8 *p = object;
801
	u8 *endobject = object + s->object_size;
Christoph Lameter's avatar
Christoph Lameter committed
802
803

	if (s->flags & SLAB_RED_ZONE) {
804
		if (!check_bytes_and_report(s, page, object, "Redzone",
805
			endobject, val, s->inuse - s->object_size))
Christoph Lameter's avatar
Christoph Lameter committed
806
807
			return 0;
	} else {
808
		if ((s->flags & SLAB_POISON) && s->object_size < s->inuse) {
Ingo Molnar's avatar
Ingo Molnar committed
809
			check_bytes_and_report(s, page, p, "Alignment padding",
810
811
				endobject, POISON_INUSE,
				s->inuse - s->object_size);
Ingo Molnar's avatar
Ingo Molnar committed
812
		}
Christoph Lameter's avatar
Christoph Lameter committed
813
814
815
	}

	if (s->flags & SLAB_POISON) {
816
		if (val != SLUB_RED_ACTIVE && (s->flags & __OBJECT_POISON) &&
817
			(!check_bytes_and_report(s, page, p, "Poison", p,
818
					POISON_FREE, s->object_size - 1) ||
819
			 !check_bytes_and_report(s, page, p, "Poison",
820
				p + s->object_size - 1, POISON_END, 1)))
Christoph Lameter's avatar
Christoph Lameter committed
821
822
823
824
825
826
827
			return 0;
		/*
		 * check_pad_bytes cleans up on its own.
		 */
		check_pad_bytes(s, page, p);
	}

828
	if (!s->offset && val == SLUB_RED_ACTIVE)
Christoph Lameter's avatar
Christoph Lameter committed
829
830
831
832
833
834
835
836
837
838
		/*
		 * Object and freepointer overlap. Cannot check
		 * freepointer while object is allocated.
		 */
		return 1;

	/* Check free pointer validity */
	if (!check_valid_pointer(s, page, get_freepointer(s, p))) {
		object_err(s, page, p, "Freepointer corrupt");
		/*
Nick Andrew's avatar
Nick Andrew committed
839
		 * No choice but to zap it and thus lose the remainder
Christoph Lameter's avatar
Christoph Lameter committed
840
		 * of the free objects in this slab. May cause
Christoph Lameter's avatar
Christoph Lameter committed
841
		 * another error because the object count is now wrong.
Christoph Lameter's avatar
Christoph Lameter committed
842
		 */
843
		set_freepointer(s, p, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
844
845
846
847
848
849
850
		return 0;
	}
	return 1;
}

static int check_slab(struct kmem_cache *s, struct page *page)
{
851
852
	int maxobj;

Christoph Lameter's avatar
Christoph Lameter committed
853
854
855
	VM_BUG_ON(!irqs_disabled());

	if (!PageSlab(page)) {
856
		slab_err(s, page, "Not a valid slab page");
Christoph Lameter's avatar
Christoph Lameter committed
857
858
		return 0;
	}
859

860
	maxobj = order_objects(compound_order(page), s->size, s->reserved);
861
862
863
864
865
866
	if (page->objects > maxobj) {
		slab_err(s, page, "objects %u > max %u",
			s->name, page->objects, maxobj);
		return 0;
	}
	if (page->inuse > page->objects) {
867
		slab_err(s, page, "inuse %u > max %u",
868
			s->name, page->inuse, page->objects);
Christoph Lameter's avatar
Christoph Lameter committed
869
870
871
872
873
874
875
876
		return 0;
	}
	/* Slab_pad_check fixes things up after itself */
	slab_pad_check(s, page);
	return 1;
}

/*
Christoph Lameter's avatar
Christoph Lameter committed
877
878
 * Determine if a certain object on a page is on the freelist. Must hold the
 * slab lock to guarantee that the chains are in a consistent state.
Christoph Lameter's avatar
Christoph Lameter committed
879
880
881
882
 */
static int on_freelist(struct kmem_cache *s, struct page *page, void *search)
{
	int nr = 0;
883
	void *fp;
Christoph Lameter's avatar
Christoph Lameter committed
884
	void *object = NULL;
885
	unsigned long max_objects;
Christoph Lameter's avatar
Christoph Lameter committed
886

887
	fp = page->freelist;
888
	while (fp && nr <= page->objects) {
Christoph Lameter's avatar
Christoph Lameter committed
889
890
891
892
893
894
		if (fp == search)
			return 1;
		if (!check_valid_pointer(s, page, fp)) {
			if (object) {
				object_err(s, page, object,
					"Freechain corrupt");
895
				set_freepointer(s, object, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
896
			} else {
897
				slab_err(s, page, "Freepointer corrupt");
898
				page->freelist = NULL;
899
				page->inuse = page->objects;
900
				slab_fix(s, "Freelist cleared");
Christoph Lameter's avatar
Christoph Lameter committed
901
902
903
904
905
906
907
908
909
				return 0;
			}
			break;
		}
		object = fp;
		fp = get_freepointer(s, object);
		nr++;
	}

910
	max_objects = order_objects(compound_order(page), s->size, s->reserved);
911
912
	if (max_objects > MAX_OBJS_PER_PAGE)
		max_objects = MAX_OBJS_PER_PAGE;
913
914
915
916
917
918
919

	if (page->objects != max_objects) {
		slab_err(s, page, "Wrong number of objects. Found %d but "
			"should be %d", page->objects, max_objects);
		page->objects = max_objects;
		slab_fix(s, "Number of objects adjusted.");
	}
920
	if (page->inuse != page->objects - nr) {
921
		slab_err(s, page, "Wrong object count. Counter is %d but "
922
923
			"counted were %d", page->inuse, page->objects - nr);
		page->inuse = page->objects - nr;
924
		slab_fix(s, "Object count adjusted.");
Christoph Lameter's avatar
Christoph Lameter committed
925
926
927
928
	}
	return search == NULL;
}

929
930
static void trace(struct kmem_cache *s, struct page *page, void *object,
								int alloc)
Christoph Lameter's avatar
Christoph Lameter committed
931
932
933
934
935
936
937
938
939
{
	if (s->flags & SLAB_TRACE) {
		printk(KERN_INFO "TRACE %s %s 0x%p inuse=%d fp=0x%p\n",
			s->name,
			alloc ? "alloc" : "free",
			object, page->inuse,
			page->freelist);

		if (!alloc)
940
941
			print_section("Object ", (void *)object,
					s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
942
943
944
945
946

		dump_stack();
	}
}

947
948
949
950
/*
 * Hooks for other subsystems that check memory allocations. In a typical
 * production configuration these hooks all should produce no code at all.
 */
951
952
953
954
955
956
957
958
959
960
static inline void kmalloc_large_node_hook(void *ptr, size_t size, gfp_t flags)
{
	kmemleak_alloc(ptr, size, 1, flags);
}

static inline void kfree_hook(const void *x)
{
	kmemleak_free(x);
}

961
962
static inline int slab_pre_alloc_hook(struct kmem_cache *s, gfp_t flags)
{
963
	flags &= gfp_allowed_mask;
964
965
966
	lockdep_trace_alloc(flags);
	might_sleep_if(flags & __GFP_WAIT);

967
	return should_failslab(s->object_size, flags, s->flags);
968
969
}

970
971
static inline void slab_post_alloc_hook(struct kmem_cache *s,
					gfp_t flags, void *object)
972
{
973
	flags &= gfp_allowed_mask;
974
	kmemcheck_slab_alloc(s, flags, object, slab_ksize(s));
975
	kmemleak_alloc_recursive(object, s->object_size, 1, s->flags, flags);
976
977
978
979
980
981
}

static inline void slab_free_hook(struct kmem_cache *s, void *x)
{
	kmemleak_free_recursive(x, s->flags);

982
	/*
983
	 * Trouble is that we may no longer disable interrupts in the fast path
984
985
986
987
988
989
990
991
	 * So in order to make the debug calls that expect irqs to be
	 * disabled we need to disable interrupts temporarily.
	 */
#if defined(CONFIG_KMEMCHECK) || defined(CONFIG_LOCKDEP)
	{
		unsigned long flags;

		local_irq_save(flags);
992
993
		kmemcheck_slab_free(s, x, s->object_size);
		debug_check_no_locks_freed(x, s->object_size);
994
995
996
		local_irq_restore(flags);
	}
#endif
997
	if (!(s->flags & SLAB_DEBUG_OBJECTS))
998
		debug_check_no_obj_freed(x, s->object_size);
999
1000
}

1001
/*
Christoph Lameter's avatar
Christoph Lameter committed
1002
 * Tracking of fully allocated slabs for debugging purposes.
1003
 */
1004
1005
static void add_full(struct kmem_cache *s,
	struct kmem_cache_node *n, struct page *page)
1006
{
Peter Zijlstra's avatar
Peter Zijlstra committed
1007
1008
	lockdep_assert_held(&n->list_lock);

1009
1010
1011
	if (!(s->flags & SLAB_STORE_USER))
		return;

1012
1013
1014
	list_add(&page->lru, &n->full);
}

Peter Zijlstra's avatar
Peter Zijlstra committed
1015
static void remove_full(struct kmem_cache *s, struct kmem_cache_node *n, struct page *page)
1016
{
Peter Zijlstra's avatar
Peter Zijlstra committed
1017
1018
	lockdep_assert_held(&n->list_lock);

1019
1020
1021
1022
1023
1024
	if (!(s->flags & SLAB_STORE_USER))
		return;

	list_del(&page->lru);
}

1025
1026
1027
1028
1029
1030
1031
1032
/* Tracking of the number of slabs for debugging purposes */
static inline unsigned long slabs_node(struct kmem_cache *s, int node)
{
	struct kmem_cache_node *n = get_node(s, node);

	return atomic_long_read(&n->nr_slabs);
}

1033
1034
1035
1036
1037
static inline unsigned long node_nr_slabs(struct kmem_cache_node *n)
{
	return atomic_long_read(&n->nr_slabs);
}

1038
static inline void inc_slabs_node(struct kmem_cache *s, int node, int objects)
1039
1040
1041
1042
1043
1044
1045
1046
1047
{
	struct kmem_cache_node *n = get_node(s, node);

	/*
	 * May be called early in order to allocate a slab for the
	 * kmem_cache_node structure. Solve the chicken-egg
	 * dilemma by deferring the increment of the count during
	 * bootstrap (see early_kmem_cache_node_alloc).
	 */
1048
	if (likely(n)) {
1049
		atomic_long_inc(&n->nr_slabs);
1050
1051
		atomic_long_add(objects, &n->total_objects);
	}
1052
}
1053
static inline void dec_slabs_node(struct kmem_cache *s, int node, int objects)
1054
1055
1056
1057
{
	struct kmem_cache_node *n = get_node(s, node);

	atomic_long_dec(&n->nr_slabs);
1058
	atomic_long_sub(objects, &n->total_objects);
1059
1060
1061
}

/* Object debug checks for alloc/free paths */
Christoph Lameter's avatar
Christoph Lameter committed
1062
1063
1064
1065
1066
1067
static void setup_object_debug(struct kmem_cache *s, struct page *page,
								void *object)
{
	if (!(s->flags & (SLAB_STORE_USER|SLAB_RED_ZONE|__OBJECT_POISON)))
		return;

1068
	init_object(s, object, SLUB_RED_INACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1069
1070
1071
	init_tracking(s, object);
}

1072
1073
static noinline int alloc_debug_processing(struct kmem_cache *s,
					struct page *page,
1074
					void *object, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
1075
1076
1077
1078
1079
1080
{
	if (!check_slab(s, page))
		goto bad;

	if (!check_valid_pointer(s, page, object)) {
		object_err(s, page, object, "Freelist Pointer check fails");
1081
		goto bad;
Christoph Lameter's avatar
Christoph Lameter committed
1082
1083
	}

1084
	if (!check_object(s, page, object, SLUB_RED_INACTIVE))
Christoph Lameter's avatar
Christoph Lameter committed
1085
1086
		goto bad;

Christoph Lameter's avatar
Christoph Lameter committed
1087
1088
1089
1090
	/* Success perform special debug activities for allocs */
	if (s->flags & SLAB_STORE_USER)
		set_track(s, object, TRACK_ALLOC, addr);
	trace(s, page, object, 1);
1091
	init_object(s, object, SLUB_RED_ACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1092
	return 1;
Christoph Lameter's avatar
Christoph Lameter committed
1093

Christoph Lameter's avatar
Christoph Lameter committed
1094
1095
1096
1097
1098
bad:
	if (PageSlab(page)) {
		/*
		 * If this is a slab page then lets do the best we can
		 * to avoid issues in the future. Marking all objects
Christoph Lameter's avatar
Christoph Lameter committed
1099
		 * as used avoids touching the remaining objects.
Christoph Lameter's avatar
Christoph Lameter committed
1100
		 */
1101
		slab_fix(s, "Marking all objects used");
1102
		page->inuse = page->objects;
1103
		page->freelist = NULL;
Christoph Lameter's avatar
Christoph Lameter committed
1104
1105
1106
1107
	}
	return 0;
}

1108
1109
1110
static noinline struct kmem_cache_node *free_debug_processing(
	struct kmem_cache *s, struct page *page, void *object,
	unsigned long addr, unsigned long *flags)
Christoph Lameter's avatar
Christoph Lameter committed
1111
{
1112
	struct kmem_cache_node *n = get_node(s, page_to_nid(page));
1113

1114
	spin_lock_irqsave(&n->list_lock, *flags);
1115
1116
	slab_lock(page);

Christoph Lameter's avatar
Christoph Lameter committed
1117
1118
1119
1120
	if (!check_slab(s, page))
		goto fail;

	if (!check_valid_pointer(s, page, object)) {
1121
		slab_err(s, page, "Invalid object pointer 0x%p", object);
Christoph Lameter's avatar
Christoph Lameter committed
1122
1123
1124
1125
		goto fail;
	}

	if (on_freelist(s, page, object)) {
1126
		object_err(s, page, object, "Object already free");
Christoph Lameter's avatar
Christoph Lameter committed
1127
1128
1129
		goto fail;
	}

1130
	if (!check_object(s, page, object, SLUB_RED_ACTIVE))
1131
		goto out;
Christoph Lameter's avatar
Christoph Lameter committed
1132

1133
	if (unlikely(s != page->slab_cache)) {
Ingo Molnar's avatar
Ingo Molnar committed
1134
		if (!PageSlab(page)) {
1135
1136
			slab_err(s, page, "Attempt to free object(0x%p) "
				"outside of slab", object);
1137
		} else if (!page->slab_cache) {
Christoph Lameter's avatar
Christoph Lameter committed
1138
			printk(KERN_ERR
1139
				"SLUB <none>: no slab for object 0x%p.\n",
Christoph Lameter's avatar
Christoph Lameter committed
1140
						object);
1141
			dump_stack();
1142
		} else
1143
1144
			object_err(s, page, object,
					"page slab pointer corrupt.");
Christoph Lameter's avatar
Christoph Lameter committed
1145
1146
		goto fail;
	}
Christoph Lameter's avatar
Christoph Lameter committed
1147
1148
1149
1150

	if (s->flags & SLAB_STORE_USER)
		set_track(s, object, TRACK_FREE, addr);
	trace(s, page, object, 0);
1151
	init_object(s, object, SLUB_RED_INACTIVE);
1152
out:
1153
	slab_unlock(page);
1154
1155
1156
1157
1158