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
static int sysfs_slab_add(struct kmem_cache *);
static int sysfs_slab_alias(struct kmem_cache *, const char *);
213
static void memcg_propagate_slab_attrs(struct kmem_cache *s);
Christoph Lameter's avatar
Christoph Lameter committed
214
#else
215
216
217
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; }
218
static inline void memcg_propagate_slab_attrs(struct kmem_cache *s) { }
Christoph Lameter's avatar
Christoph Lameter committed
219
220
#endif

221
static inline void stat(const struct kmem_cache *s, enum stat_item si)
222
223
{
#ifdef CONFIG_SLUB_STATS
224
225
226
227
228
	/*
	 * The rmw is racy on a preemptible kernel but this is acceptable, so
	 * avoid this_cpu_add()'s irq-disable overhead.
	 */
	raw_cpu_inc(s->cpu_slab->stat[si]);
229
230
231
#endif
}

Christoph Lameter's avatar
Christoph Lameter committed
232
233
234
235
236
237
238
239
240
/********************************************************************
 * 			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
241
/* Verify that a pointer has an address that is valid within a slab page */
242
243
244
245
246
static inline int check_valid_pointer(struct kmem_cache *s,
				struct page *page, const void *object)
{
	void *base;

247
	if (!object)
248
249
		return 1;

250
	base = page_address(page);
251
	if (object < base || object >= base + page->objects * s->size ||
252
253
254
255
256
257
258
		(object - base) % s->size) {
		return 0;
	}

	return 1;
}

259
260
261
262
263
static inline void *get_freepointer(struct kmem_cache *s, void *object)
{
	return *(void **)(object + s->offset);
}

264
265
266
267
268
static void prefetch_freepointer(const struct kmem_cache *s, void *object)
{
	prefetch(object + s->offset);
}

269
270
271
272
273
274
275
276
277
278
279
280
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;
}

281
282
283
284
285
286
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 */
287
288
#define for_each_object(__p, __s, __addr, __objects) \
	for (__p = (__addr); __p < (__addr) + (__objects) * (__s)->size;\
289
290
291
292
293
294
295
296
			__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;
}

297
298
299
300
301
302
303
304
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))
305
		return s->object_size;
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320

#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;
}

321
322
323
324
325
static inline int order_objects(int order, unsigned long size, int reserved)
{
	return ((PAGE_SIZE << order) - reserved) / size;
}

326
static inline struct kmem_cache_order_objects oo_make(int order,
327
		unsigned long size, int reserved)
328
329
{
	struct kmem_cache_order_objects x = {
330
		(order << OO_SHIFT) + order_objects(order, size, reserved)
331
332
333
334
335
336
337
	};

	return x;
}

static inline int oo_order(struct kmem_cache_order_objects x)
{
338
	return x.x >> OO_SHIFT;
339
340
341
342
}

static inline int oo_objects(struct kmem_cache_order_objects x)
{
343
	return x.x & OO_MASK;
344
345
}

346
347
348
349
350
351
352
353
354
355
356
357
358
/*
 * 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);
}

359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
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;
}

374
375
376
377
378
379
380
/* 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());
381
382
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
383
	if (s->flags & __CMPXCHG_DOUBLE) {
384
		if (cmpxchg_double(&page->freelist, &page->counters,
385
386
387
388
389
390
391
			freelist_old, counters_old,
			freelist_new, counters_new))
		return 1;
	} else
#endif
	{
		slab_lock(page);
392
393
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
394
			page->freelist = freelist_new;
395
			set_page_slub_counters(page, counters_new);
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
			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;
}

412
413
414
415
416
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)
{
417
418
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
419
	if (s->flags & __CMPXCHG_DOUBLE) {
420
		if (cmpxchg_double(&page->freelist, &page->counters,
421
422
423
424
425
426
			freelist_old, counters_old,
			freelist_new, counters_new))
		return 1;
	} else
#endif
	{
427
428
429
		unsigned long flags;

		local_irq_save(flags);
430
		slab_lock(page);
431
432
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
433
			page->freelist = freelist_new;
434
			set_page_slub_counters(page, counters_new);
435
			slab_unlock(page);
436
			local_irq_restore(flags);
437
438
			return 1;
		}
439
		slab_unlock(page);
440
		local_irq_restore(flags);
441
442
443
444
445
446
447
448
449
450
451
452
	}

	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;
}

453
#ifdef CONFIG_SLUB_DEBUG
454
455
456
/*
 * Determine a map of object in use on a page.
 *
457
 * Node listlock must be held to guarantee that the page does
458
459
460
461
462
463
464
465
466
467
468
 * 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);
}

469
470
471
/*
 * Debug settings:
 */
472
473
474
#ifdef CONFIG_SLUB_DEBUG_ON
static int slub_debug = DEBUG_DEFAULT_FLAGS;
#else
475
static int slub_debug;
476
#endif
477
478

static char *slub_debug_slabs;
479
static int disable_higher_order_debug;
480

Christoph Lameter's avatar
Christoph Lameter committed
481
482
483
484
485
/*
 * Object debugging
 */
static void print_section(char *text, u8 *addr, unsigned int length)
{
486
487
	print_hex_dump(KERN_ERR, text, DUMP_PREFIX_ADDRESS, 16, 1, addr,
			length, 1);
Christoph Lameter's avatar
Christoph Lameter committed
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
}

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,
504
			enum track_item alloc, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
505
{
Akinobu Mita's avatar
Akinobu Mita committed
506
	struct track *p = get_track(s, object, alloc);
Christoph Lameter's avatar
Christoph Lameter committed
507
508

	if (addr) {
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
#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
527
528
		p->addr = addr;
		p->cpu = smp_processor_id();
529
		p->pid = current->pid;
Christoph Lameter's avatar
Christoph Lameter committed
530
531
532
533
534
535
536
		p->when = jiffies;
	} else
		memset(p, 0, sizeof(struct track));
}

static void init_tracking(struct kmem_cache *s, void *object)
{
537
538
539
	if (!(s->flags & SLAB_STORE_USER))
		return;

540
541
	set_track(s, object, TRACK_FREE, 0UL);
	set_track(s, object, TRACK_ALLOC, 0UL);
Christoph Lameter's avatar
Christoph Lameter committed
542
543
544
545
546
547
548
}

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

549
	printk(KERN_ERR "INFO: %s in %pS age=%lu cpu=%u pid=%d\n",
550
		s, (void *)t->addr, jiffies - t->when, t->cpu, t->pid);
551
552
553
554
555
556
557
558
559
560
#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
561
562
563
564
565
566
567
568
569
570
571
572
573
}

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)
{
574
575
576
	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);
577
578
579
580
581
582
583
584
585
586
587
588
589

}

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");
590
	printk(KERN_ERR "BUG %s (%s): %s\n", s->name, print_tainted(), buf);
591
592
	printk(KERN_ERR "----------------------------------------"
			"-------------------------------------\n\n");
593

594
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
Christoph Lameter's avatar
Christoph Lameter committed
595
596
}

597
598
599
600
601
602
603
604
605
606
607
608
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
609
610
{
	unsigned int off;	/* Offset of last byte */
611
	u8 *addr = page_address(page);
612
613
614
615
616
617
618
619
620

	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)
621
		print_section("Bytes b4 ", p - 16, 16);
Christoph Lameter's avatar
Christoph Lameter committed
622

623
	print_section("Object ", p, min_t(unsigned long, s->object_size,
624
				PAGE_SIZE));
Christoph Lameter's avatar
Christoph Lameter committed
625
	if (s->flags & SLAB_RED_ZONE)
626
627
		print_section("Redzone ", p + s->object_size,
			s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
628
629
630
631
632
633

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

634
	if (s->flags & SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
635
636
637
638
		off += 2 * sizeof(struct track);

	if (off != s->size)
		/* Beginning of the filler is the free pointer */
639
		print_section("Padding ", p + off, s->size - off);
640
641

	dump_stack();
Christoph Lameter's avatar
Christoph Lameter committed
642
643
644
645
646
}

static void object_err(struct kmem_cache *s, struct page *page,
			u8 *object, char *reason)
{
647
	slab_bug(s, "%s", reason);
648
	print_trailer(s, page, object);
Christoph Lameter's avatar
Christoph Lameter committed
649
650
}

651
652
static void slab_err(struct kmem_cache *s, struct page *page,
			const char *fmt, ...)
Christoph Lameter's avatar
Christoph Lameter committed
653
654
655
656
{
	va_list args;
	char buf[100];

657
658
	va_start(args, fmt);
	vsnprintf(buf, sizeof(buf), fmt, args);
Christoph Lameter's avatar
Christoph Lameter committed
659
	va_end(args);
660
	slab_bug(s, "%s", buf);
661
	print_page_info(page);
Christoph Lameter's avatar
Christoph Lameter committed
662
663
664
	dump_stack();
}

665
static void init_object(struct kmem_cache *s, void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
666
667
668
669
{
	u8 *p = object;

	if (s->flags & __OBJECT_POISON) {
670
671
		memset(p, POISON_FREE, s->object_size - 1);
		p[s->object_size - 1] = POISON_END;
Christoph Lameter's avatar
Christoph Lameter committed
672
673
674
	}

	if (s->flags & SLAB_RED_ZONE)
675
		memset(p + s->object_size, val, s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
676
677
}

678
679
680
681
682
683
684
685
686
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,
687
			u8 *start, unsigned int value, unsigned int bytes)
688
689
690
691
{
	u8 *fault;
	u8 *end;

692
	fault = memchr_inv(start, value, bytes);
693
694
695
696
697
698
699
700
701
702
703
704
705
706
	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
707
708
709
710
711
712
713
714
715
}

/*
 * 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
716
 *
Christoph Lameter's avatar
Christoph Lameter committed
717
718
719
 * 	Poisoning uses 0x6b (POISON_FREE) and the last byte is
 * 	0xa5 (POISON_END)
 *
720
 * object + s->object_size
Christoph Lameter's avatar
Christoph Lameter committed
721
 * 	Padding to reach word boundary. This is also used for Redzoning.
Christoph Lameter's avatar
Christoph Lameter committed
722
 * 	Padding is extended by another word if Redzoning is enabled and
723
 * 	object_size == inuse.
Christoph Lameter's avatar
Christoph Lameter committed
724
 *
Christoph Lameter's avatar
Christoph Lameter committed
725
726
727
728
 * 	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
729
730
 * 	Meta data starts here.
 *
Christoph Lameter's avatar
Christoph Lameter committed
731
732
 * 	A. Free pointer (if we cannot overwrite object on free)
 * 	B. Tracking data for SLAB_STORE_USER
Christoph Lameter's avatar
Christoph Lameter committed
733
 * 	C. Padding to reach required alignment boundary or at mininum
Christoph Lameter's avatar
Christoph Lameter committed
734
 * 		one word if debugging is on to be able to detect writes
Christoph Lameter's avatar
Christoph Lameter committed
735
736
737
 * 		before the word boundary.
 *
 *	Padding is done using 0x5a (POISON_INUSE)
Christoph Lameter's avatar
Christoph Lameter committed
738
739
 *
 * object + s->size
Christoph Lameter's avatar
Christoph Lameter committed
740
 * 	Nothing is used beyond s->size.
Christoph Lameter's avatar
Christoph Lameter committed
741
 *
742
 * If slabcaches are merged then the object_size and inuse boundaries are mostly
Christoph Lameter's avatar
Christoph Lameter committed
743
 * ignored. And therefore no slab options that rely on these boundaries
Christoph Lameter's avatar
Christoph Lameter committed
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
 * 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;

762
763
	return check_bytes_and_report(s, page, p, "Object padding",
				p + off, POISON_INUSE, s->size - off);
Christoph Lameter's avatar
Christoph Lameter committed
764
765
}

766
/* Check the pad bytes at the end of a slab page */
Christoph Lameter's avatar
Christoph Lameter committed
767
768
static int slab_pad_check(struct kmem_cache *s, struct page *page)
{
769
770
771
772
773
	u8 *start;
	u8 *fault;
	u8 *end;
	int length;
	int remainder;
Christoph Lameter's avatar
Christoph Lameter committed
774
775
776
777

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

778
	start = page_address(page);
779
	length = (PAGE_SIZE << compound_order(page)) - s->reserved;
780
781
	end = start + length;
	remainder = length % s->size;
Christoph Lameter's avatar
Christoph Lameter committed
782
783
784
	if (!remainder)
		return 1;

785
	fault = memchr_inv(end - remainder, POISON_INUSE, remainder);
786
787
788
789
790
791
	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);
792
	print_section("Padding ", end - remainder, remainder);
793

Eric Dumazet's avatar
Eric Dumazet committed
794
	restore_bytes(s, "slab padding", POISON_INUSE, end - remainder, end);
795
	return 0;
Christoph Lameter's avatar
Christoph Lameter committed
796
797
798
}

static int check_object(struct kmem_cache *s, struct page *page,
799
					void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
800
801
{
	u8 *p = object;
802
	u8 *endobject = object + s->object_size;
Christoph Lameter's avatar
Christoph Lameter committed
803
804

	if (s->flags & SLAB_RED_ZONE) {
805
		if (!check_bytes_and_report(s, page, object, "Redzone",
806
			endobject, val, s->inuse - s->object_size))
Christoph Lameter's avatar
Christoph Lameter committed
807
808
			return 0;
	} else {
809
		if ((s->flags & SLAB_POISON) && s->object_size < s->inuse) {
Ingo Molnar's avatar
Ingo Molnar committed
810
			check_bytes_and_report(s, page, p, "Alignment padding",
811
812
				endobject, POISON_INUSE,
				s->inuse - s->object_size);
Ingo Molnar's avatar
Ingo Molnar committed
813
		}
Christoph Lameter's avatar
Christoph Lameter committed
814
815
816
	}

	if (s->flags & SLAB_POISON) {
817
		if (val != SLUB_RED_ACTIVE && (s->flags & __OBJECT_POISON) &&
818
			(!check_bytes_and_report(s, page, p, "Poison", p,
819
					POISON_FREE, s->object_size - 1) ||
820
			 !check_bytes_and_report(s, page, p, "Poison",
821
				p + s->object_size - 1, POISON_END, 1)))
Christoph Lameter's avatar
Christoph Lameter committed
822
823
824
825
826
827
828
			return 0;
		/*
		 * check_pad_bytes cleans up on its own.
		 */
		check_pad_bytes(s, page, p);
	}

829
	if (!s->offset && val == SLUB_RED_ACTIVE)
Christoph Lameter's avatar
Christoph Lameter committed
830
831
832
833
834
835
836
837
838
839
		/*
		 * 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
840
		 * No choice but to zap it and thus lose the remainder
Christoph Lameter's avatar
Christoph Lameter committed
841
		 * of the free objects in this slab. May cause
Christoph Lameter's avatar
Christoph Lameter committed
842
		 * another error because the object count is now wrong.
Christoph Lameter's avatar
Christoph Lameter committed
843
		 */
844
		set_freepointer(s, p, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
845
846
847
848
849
850
851
		return 0;
	}
	return 1;
}

static int check_slab(struct kmem_cache *s, struct page *page)
{
852
853
	int maxobj;

Christoph Lameter's avatar
Christoph Lameter committed
854
855
856
	VM_BUG_ON(!irqs_disabled());

	if (!PageSlab(page)) {
857
		slab_err(s, page, "Not a valid slab page");
Christoph Lameter's avatar
Christoph Lameter committed
858
859
		return 0;
	}
860

861
	maxobj = order_objects(compound_order(page), s->size, s->reserved);
862
863
864
865
866
867
	if (page->objects > maxobj) {
		slab_err(s, page, "objects %u > max %u",
			s->name, page->objects, maxobj);
		return 0;
	}
	if (page->inuse > page->objects) {
868
		slab_err(s, page, "inuse %u > max %u",
869
			s->name, page->inuse, page->objects);
Christoph Lameter's avatar
Christoph Lameter committed
870
871
872
873
874
875
876
877
		return 0;
	}
	/* Slab_pad_check fixes things up after itself */
	slab_pad_check(s, page);
	return 1;
}

/*
Christoph Lameter's avatar
Christoph Lameter committed
878
879
 * 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
880
881
882
883
 */
static int on_freelist(struct kmem_cache *s, struct page *page, void *search)
{
	int nr = 0;
884
	void *fp;
Christoph Lameter's avatar
Christoph Lameter committed
885
	void *object = NULL;
886
	unsigned long max_objects;
Christoph Lameter's avatar
Christoph Lameter committed
887

888
	fp = page->freelist;
889
	while (fp && nr <= page->objects) {
Christoph Lameter's avatar
Christoph Lameter committed
890
891
892
893
894
895
		if (fp == search)
			return 1;
		if (!check_valid_pointer(s, page, fp)) {
			if (object) {
				object_err(s, page, object,
					"Freechain corrupt");
896
				set_freepointer(s, object, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
897
			} else {
898
				slab_err(s, page, "Freepointer corrupt");
899
				page->freelist = NULL;
900
				page->inuse = page->objects;
901
				slab_fix(s, "Freelist cleared");
Christoph Lameter's avatar
Christoph Lameter committed
902
903
904
905
906
907
908
909
910
				return 0;
			}
			break;
		}
		object = fp;
		fp = get_freepointer(s, object);
		nr++;
	}

911
	max_objects = order_objects(compound_order(page), s->size, s->reserved);
912
913
	if (max_objects > MAX_OBJS_PER_PAGE)
		max_objects = MAX_OBJS_PER_PAGE;
914
915
916
917
918
919
920

	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.");
	}
921
	if (page->inuse != page->objects - nr) {
922
		slab_err(s, page, "Wrong object count. Counter is %d but "
923
924
			"counted were %d", page->inuse, page->objects - nr);
		page->inuse = page->objects - nr;
925
		slab_fix(s, "Object count adjusted.");
Christoph Lameter's avatar
Christoph Lameter committed
926
927
928
929
	}
	return search == NULL;
}

930
931
static void trace(struct kmem_cache *s, struct page *page, void *object,
								int alloc)
Christoph Lameter's avatar
Christoph Lameter committed
932
933
934
935
936
937
938
939
940
{
	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)
941
942
			print_section("Object ", (void *)object,
					s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
943
944
945
946
947

		dump_stack();
	}
}

948
949
950
951
/*
 * Hooks for other subsystems that check memory allocations. In a typical
 * production configuration these hooks all should produce no code at all.
 */
952
953
954
955
956
957
958
959
960
961
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);
}

962
963
static inline int slab_pre_alloc_hook(struct kmem_cache *s, gfp_t flags)
{
964
	flags &= gfp_allowed_mask;
965
966
967
	lockdep_trace_alloc(flags);
	might_sleep_if(flags & __GFP_WAIT);

968
	return should_failslab(s->object_size, flags, s->flags);
969
970
}

971
972
static inline void slab_post_alloc_hook(struct kmem_cache *s,
					gfp_t flags, void *object)
973
{
974
	flags &= gfp_allowed_mask;
975
	kmemcheck_slab_alloc(s, flags, object, slab_ksize(s));
976
	kmemleak_alloc_recursive(object, s->object_size, 1, s->flags, flags);
977
978
979
980
981
982
}

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

983
	/*
984
	 * Trouble is that we may no longer disable interrupts in the fast path
985
986
987
988
989
990
991
992
	 * 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);
993
994
		kmemcheck_slab_free(s, x, s->object_size);
		debug_check_no_locks_freed(x, s->object_size);
995
996
997
		local_irq_restore(flags);
	}
#endif
998
	if (!(s->flags & SLAB_DEBUG_OBJECTS))
999
		debug_check_no_obj_freed(x, s->object_size);
1000
1001
}

1002
/*
Christoph Lameter's avatar
Christoph Lameter committed
1003
 * Tracking of fully allocated slabs for debugging purposes.
1004
 */
1005
1006
static void add_full(struct kmem_cache *s,
	struct kmem_cache_node *n, struct page *page)
1007
{
1008
1009
1010
	if (!(s->flags & SLAB_STORE_USER))
		return;

1011
	lockdep_assert_held(&n->list_lock);
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
1017
1018
1019
{
	if (!(s->flags & SLAB_STORE_USER))
		return;

1020
	lockdep_assert_held(&n->list_lock);
1021
1022
1023
	list_del(&page->lru);
}

1024
1025
1026
1027
1028
1029
1030
1031
/* 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);
}

1032
1033
1034
1035
1036
static inline unsigned long node_nr_slabs(struct kmem_cache_node *n)
{
	return atomic_long_read(&n->nr_slabs);
}

1037
static inline void inc_slabs_node(struct kmem_cache *s, int node, int objects)
1038
1039
1040
1041
1042
1043
1044
1045
1046
{
	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).
	 */
1047
	if (likely(n)) {
1048
		atomic_long_inc(&n->nr_slabs);
1049
1050
		atomic_long_add(objects, &n->total_objects);
	}
1051
}
1052
static inline void dec_slabs_node(struct kmem_cache *s, int node, int objects)
1053
1054
1055
1056
{
	struct kmem_cache_node *n = get_node(s, node);

	atomic_long_dec(&n->nr_slabs);
1057
	atomic_long_sub(objects, &n->total_objects);
1058
1059
1060
}

/* Object debug checks for alloc/free paths */
Christoph Lameter's avatar
Christoph Lameter committed
1061
1062
1063
1064
1065
1066
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;

1067
	init_object(s, object, SLUB_RED_INACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1068
1069
1070
	init_tracking(s, object);
}

1071
1072
static noinline int alloc_debug_processing(struct kmem_cache *s,
					struct page *page,
1073
					void *object, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
1074
1075
1076
1077
1078
1079
{
	if (!check_slab(s, page))
		goto bad;

	if (!check_valid_pointer(s, page, object)) {
		object_err(s, page, object, "Freelist Pointer check fails");
1080
		goto bad;
Christoph Lameter's avatar
Christoph Lameter committed
1081
1082
	}

1083
	if (!check_object(s, page, object, SLUB_RED_INACTIVE))
Christoph Lameter's avatar
Christoph Lameter committed
1084
1085
		goto bad;

Christoph Lameter's avatar
Christoph Lameter committed
1086
1087
1088
1089
	/* 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);
1090
	init_object(s, object, SLUB_RED_ACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1091
	return 1;
Christoph Lameter's avatar
Christoph Lameter committed
1092

Christoph Lameter's avatar
Christoph Lameter committed
1093
1094
1095
1096
1097
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
1098
		 * as used avoids touching the remaining objects.
Christoph Lameter's avatar
Christoph Lameter committed
1099
		 */
1100
		slab_fix(s, "Marking all objects used");
1101
		page->inuse = page->objects;
1102
		page->freelist = NULL;
Christoph Lameter's avatar
Christoph Lameter committed
1103
1104
1105
1106
	}
	return 0;
}

1107
1108
1109
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
1110
{
1111
	struct kmem_cache_node *n = get_node(s, page_to_nid(page));
1112

1113
	spin_lock_irqsave(&n->list_lock, *flags);
1114
1115
	slab_lock(page);

Christoph Lameter's avatar
Christoph Lameter committed
1116
1117
1118
1119
	if (!check_slab(s, page))
		goto fail;

	if (!check_valid_pointer(s, page, object)) {
1120
		slab_err(s, page, "Invalid object pointer 0x%p", object);
Christoph Lameter's avatar
Christoph Lameter committed
1121
1122
1123
1124
		goto fail;
	}

	if (on_freelist(s, page, object)) {
1125
		object_err(s, page, object, "Object already free");
Christoph Lameter's avatar
Christoph Lameter committed
1126
1127
1128
		goto fail;
	}

1129
	if (!check_object(s, page, object, SLUB_RED_ACTIVE))
1130
		goto out;
Christoph Lameter's avatar
Christoph Lameter committed
1131

1132
	if (unlikely(s != page->slab_cache)) {
Ingo Molnar's avatar
Ingo Molnar committed
1133
		if (!PageSlab(page)) {
1134
1135
			slab_err(s, page, "Attempt to free object(0x%p) "
				"outside of slab", object);
1136
		} else if (!page->slab_cache) {
Christoph Lameter's avatar
Christoph Lameter committed
1137
			printk(KERN_ERR
1138
				"SLUB <none>: no slab for object 0x%p.\n",
Christoph Lameter's avatar
Christoph Lameter committed
1139
						object);
1140
			dump_stack();
1141
		} else
1142
1143
			object_err(s, page, object,
					"page slab pointer corrupt.");
Christoph Lameter's avatar
Christoph Lameter committed
1144
1145
		goto fail;
	}
Christoph Lameter's avatar
Christoph Lameter committed
1146
1147
1148
1149

	if (s->flags & SLAB_STORE_USER)
		set_track(s, object, TRACK_FREE, addr);
	trace(s, page, object, 0);
1150
	init_object(s, object, SLUB_RED_INACTIVE);
1151
out:
1152
	slab_unlock(page);
1153
1154
1155
1156
1157
	/*
	 * Keep node_lock to preserve integrity
	 * until the object is actually freed
	 */
	return n;
Christoph Lameter's avatar
Christoph Lameter committed
1158

Christoph Lameter's avatar
Christoph Lameter committed
1159