slub.c 142 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>
23
#include <linux/kasan.h>
Vegard Nossum's avatar
Vegard Nossum committed
24
#include <linux/kmemcheck.h>
Christoph Lameter's avatar
Christoph Lameter committed
25
26
27
28
#include <linux/cpu.h>
#include <linux/cpuset.h>
#include <linux/mempolicy.h>
#include <linux/ctype.h>
29
#include <linux/debugobjects.h>
Christoph Lameter's avatar
Christoph Lameter committed
30
#include <linux/kallsyms.h>
31
#include <linux/memory.h>
Roman Zippel's avatar
Roman Zippel committed
32
#include <linux/math64.h>
Akinobu Mita's avatar
Akinobu Mita committed
33
#include <linux/fault-inject.h>
34
#include <linux/stacktrace.h>
35
#include <linux/prefetch.h>
36
#include <linux/memcontrol.h>
37
#include <linux/random.h>
Christoph Lameter's avatar
Christoph Lameter committed
38

39
40
#include <trace/events/kmem.h>

41
42
#include "internal.h"

Christoph Lameter's avatar
Christoph Lameter committed
43
44
/*
 * Lock order:
45
 *   1. slab_mutex (Global Mutex)
46
47
 *   2. node->list_lock
 *   3. slab_lock(page) (Only on some arches and for debugging)
Christoph Lameter's avatar
Christoph Lameter committed
48
 *
49
 *   slab_mutex
50
 *
51
 *   The role of the slab_mutex is to protect the list of all the slabs
52
53
54
55
56
57
58
59
60
61
62
63
64
65
 *   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
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
 *
 *   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
86
87
 * 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
88
 * freed then the slab will show up again on the partial lists.
Christoph Lameter's avatar
Christoph Lameter committed
89
90
 * We track full slabs for debugging purposes though because otherwise we
 * cannot scan all objects.
Christoph Lameter's avatar
Christoph Lameter committed
91
92
93
94
95
96
97
 *
 * 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.
 *
98
99
100
101
102
103
104
105
106
107
108
109
 * 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
110
 * 			freelist that allows lockless access to
111
112
 * 			free objects in addition to the regular freelist
 * 			that requires the slab lock.
Christoph Lameter's avatar
Christoph Lameter committed
113
114
115
 *
 * PageError		Slab requires special handling due to debug
 * 			options set. This moves	slab handling out of
116
 * 			the fast path and disables lockless freelists.
Christoph Lameter's avatar
Christoph Lameter committed
117
118
 */

119
120
static inline int kmem_cache_debug(struct kmem_cache *s)
{
121
#ifdef CONFIG_SLUB_DEBUG
122
	return unlikely(s->flags & SLAB_DEBUG_FLAGS);
123
#else
124
	return 0;
125
#endif
126
}
127

128
void *fixup_red_left(struct kmem_cache *s, void *p)
Joonsoo Kim's avatar
Joonsoo Kim committed
129
130
131
132
133
134
135
{
	if (kmem_cache_debug(s) && s->flags & SLAB_RED_ZONE)
		p += s->red_left_pad;

	return p;
}

136
137
138
139
140
141
142
143
144
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
145
146
147
148
149
150
151
152
153
154
155
/*
 * 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

156
157
158
/* Enable to log cmpxchg failures */
#undef SLUB_DEBUG_CMPXCHG

159
160
161
162
/*
 * Mininum number of partial slabs. These will be left on the partial
 * lists even if they are empty. kmem_cache_shrink may reclaim them.
 */
163
#define MIN_PARTIAL 5
Christoph Lameter's avatar
Christoph Lameter committed
164

165
166
167
/*
 * Maximum number of desirable partial slabs.
 * The existence of more partial slabs makes kmem_cache_shrink
168
 * sort the partial list by the number of objects in use.
169
170
171
 */
#define MAX_PARTIAL 10

172
#define DEBUG_DEFAULT_FLAGS (SLAB_CONSISTENCY_CHECKS | SLAB_RED_ZONE | \
Christoph Lameter's avatar
Christoph Lameter committed
173
				SLAB_POISON | SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
174

175
176
177
178
179
180
181
182
/*
 * These debug flags cannot use CMPXCHG because there might be consistency
 * issues when checking or reading debug information
 */
#define SLAB_NO_CMPXCHG (SLAB_CONSISTENCY_CHECKS | SLAB_STORE_USER | \
				SLAB_TRACE)


183
/*
184
185
186
 * 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.
187
 */
188
#define DEBUG_METADATA_FLAGS (SLAB_RED_ZONE | SLAB_POISON | SLAB_STORE_USER)
189

190
191
#define OO_SHIFT	16
#define OO_MASK		((1 << OO_SHIFT) - 1)
192
#define MAX_OBJS_PER_PAGE	32767 /* since page.objects is u15 */
193

Christoph Lameter's avatar
Christoph Lameter committed
194
/* Internal SLUB flags */
Christoph Lameter's avatar
Christoph Lameter committed
195
#define __OBJECT_POISON		0x80000000UL /* Poison object */
196
#define __CMPXCHG_DOUBLE	0x40000000UL /* Use cmpxchg_double */
Christoph Lameter's avatar
Christoph Lameter committed
197

198
199
200
/*
 * Tracking user of a slab.
 */
201
#define TRACK_ADDRS_COUNT 16
202
struct track {
203
	unsigned long addr;	/* Called from address */
204
205
206
#ifdef CONFIG_STACKTRACE
	unsigned long addrs[TRACK_ADDRS_COUNT];	/* Called from address */
#endif
207
208
209
210
211
212
213
	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 };

214
#ifdef CONFIG_SYSFS
Christoph Lameter's avatar
Christoph Lameter committed
215
216
static int sysfs_slab_add(struct kmem_cache *);
static int sysfs_slab_alias(struct kmem_cache *, const char *);
217
static void memcg_propagate_slab_attrs(struct kmem_cache *s);
218
static void sysfs_slab_remove(struct kmem_cache *s);
Christoph Lameter's avatar
Christoph Lameter committed
219
#else
220
221
222
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; }
223
static inline void memcg_propagate_slab_attrs(struct kmem_cache *s) { }
224
static inline void sysfs_slab_remove(struct kmem_cache *s) { }
Christoph Lameter's avatar
Christoph Lameter committed
225
226
#endif

227
static inline void stat(const struct kmem_cache *s, enum stat_item si)
228
229
{
#ifdef CONFIG_SLUB_STATS
230
231
232
233
234
	/*
	 * 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]);
235
236
237
#endif
}

Christoph Lameter's avatar
Christoph Lameter committed
238
239
240
241
/********************************************************************
 * 			Core slab cache functions
 *******************************************************************/

242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
/*
 * Returns freelist pointer (ptr). With hardening, this is obfuscated
 * with an XOR of the address where the pointer is held and a per-cache
 * random number.
 */
static inline void *freelist_ptr(const struct kmem_cache *s, void *ptr,
				 unsigned long ptr_addr)
{
#ifdef CONFIG_SLAB_FREELIST_HARDENED
	return (void *)((unsigned long)ptr ^ s->random ^ ptr_addr);
#else
	return ptr;
#endif
}

/* Returns the freelist pointer recorded at location ptr_addr. */
static inline void *freelist_dereference(const struct kmem_cache *s,
					 void *ptr_addr)
{
	return freelist_ptr(s, (void *)*(unsigned long *)(ptr_addr),
			    (unsigned long)ptr_addr);
}

265
266
static inline void *get_freepointer(struct kmem_cache *s, void *object)
{
267
	return freelist_dereference(s, object + s->offset);
268
269
}

270
271
static void prefetch_freepointer(const struct kmem_cache *s, void *object)
{
272
273
	if (object)
		prefetch(freelist_dereference(s, object + s->offset));
274
275
}

276
277
static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
{
278
	unsigned long freepointer_addr;
279
280
	void *p;

281
282
283
	if (!debug_pagealloc_enabled())
		return get_freepointer(s, object);

284
285
286
	freepointer_addr = (unsigned long)object + s->offset;
	probe_kernel_read(&p, (void **)freepointer_addr, sizeof(p));
	return freelist_ptr(s, p, freepointer_addr);
287
288
}

289
290
static inline void set_freepointer(struct kmem_cache *s, void *object, void *fp)
{
291
292
293
	unsigned long freeptr_addr = (unsigned long)object + s->offset;

	*(void **)freeptr_addr = freelist_ptr(s, fp, freeptr_addr);
294
295
296
}

/* Loop over all objects in a slab */
297
#define for_each_object(__p, __s, __addr, __objects) \
Joonsoo Kim's avatar
Joonsoo Kim committed
298
299
300
	for (__p = fixup_red_left(__s, __addr); \
		__p < (__addr) + (__objects) * (__s)->size; \
		__p += (__s)->size)
301

302
#define for_each_object_idx(__p, __idx, __s, __addr, __objects) \
Joonsoo Kim's avatar
Joonsoo Kim committed
303
304
305
	for (__p = fixup_red_left(__s, __addr), __idx = 1; \
		__idx <= __objects; \
		__p += (__s)->size, __idx++)
306

307
308
309
310
311
312
/* 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;
}

313
314
315
316
317
static inline int order_objects(int order, unsigned long size, int reserved)
{
	return ((PAGE_SIZE << order) - reserved) / size;
}

318
static inline struct kmem_cache_order_objects oo_make(int order,
319
		unsigned long size, int reserved)
320
321
{
	struct kmem_cache_order_objects x = {
322
		(order << OO_SHIFT) + order_objects(order, size, reserved)
323
324
325
326
327
328
329
	};

	return x;
}

static inline int oo_order(struct kmem_cache_order_objects x)
{
330
	return x.x >> OO_SHIFT;
331
332
333
334
}

static inline int oo_objects(struct kmem_cache_order_objects x)
{
335
	return x.x & OO_MASK;
336
337
}

338
339
340
341
342
/*
 * Per slab locking using the pagelock
 */
static __always_inline void slab_lock(struct page *page)
{
343
	VM_BUG_ON_PAGE(PageTail(page), page);
344
345
346
347
348
	bit_spin_lock(PG_locked, &page->flags);
}

static __always_inline void slab_unlock(struct page *page)
{
349
	VM_BUG_ON_PAGE(PageTail(page), page);
350
351
352
	__bit_spin_unlock(PG_locked, &page->flags);
}

353
354
355
356
357
358
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
359
360
	 * as page->_refcount.  If we assign to ->counters directly
	 * we run the risk of losing updates to page->_refcount, so
361
362
363
364
365
366
367
	 * be careful and only assign to the fields we need.
	 */
	page->frozen  = tmp.frozen;
	page->inuse   = tmp.inuse;
	page->objects = tmp.objects;
}

368
369
370
371
372
373
374
/* 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());
375
376
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
377
	if (s->flags & __CMPXCHG_DOUBLE) {
378
		if (cmpxchg_double(&page->freelist, &page->counters,
379
380
				   freelist_old, counters_old,
				   freelist_new, counters_new))
381
			return true;
382
383
384
385
	} else
#endif
	{
		slab_lock(page);
386
387
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
388
			page->freelist = freelist_new;
389
			set_page_slub_counters(page, counters_new);
390
			slab_unlock(page);
391
			return true;
392
393
394
395
396
397
398
399
		}
		slab_unlock(page);
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
400
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
401
402
#endif

403
	return false;
404
405
}

406
407
408
409
410
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)
{
411
412
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
413
	if (s->flags & __CMPXCHG_DOUBLE) {
414
		if (cmpxchg_double(&page->freelist, &page->counters,
415
416
				   freelist_old, counters_old,
				   freelist_new, counters_new))
417
			return true;
418
419
420
	} else
#endif
	{
421
422
423
		unsigned long flags;

		local_irq_save(flags);
424
		slab_lock(page);
425
426
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
427
			page->freelist = freelist_new;
428
			set_page_slub_counters(page, counters_new);
429
			slab_unlock(page);
430
			local_irq_restore(flags);
431
			return true;
432
		}
433
		slab_unlock(page);
434
		local_irq_restore(flags);
435
436
437
438
439
440
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
441
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
442
443
#endif

444
	return false;
445
446
}

447
#ifdef CONFIG_SLUB_DEBUG
448
449
450
/*
 * Determine a map of object in use on a page.
 *
451
 * Node listlock must be held to guarantee that the page does
452
453
454
455
456
457
458
459
460
461
462
 * 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);
}

Joonsoo Kim's avatar
Joonsoo Kim committed
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
static inline int size_from_object(struct kmem_cache *s)
{
	if (s->flags & SLAB_RED_ZONE)
		return s->size - s->red_left_pad;

	return s->size;
}

static inline void *restore_red_left(struct kmem_cache *s, void *p)
{
	if (s->flags & SLAB_RED_ZONE)
		p -= s->red_left_pad;

	return p;
}

479
480
481
/*
 * Debug settings:
 */
482
#if defined(CONFIG_SLUB_DEBUG_ON)
483
484
static int slub_debug = DEBUG_DEFAULT_FLAGS;
#else
485
static int slub_debug;
486
#endif
487
488

static char *slub_debug_slabs;
489
static int disable_higher_order_debug;
490

491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
/*
 * slub is about to manipulate internal object metadata.  This memory lies
 * outside the range of the allocated object, so accessing it would normally
 * be reported by kasan as a bounds error.  metadata_access_enable() is used
 * to tell kasan that these accesses are OK.
 */
static inline void metadata_access_enable(void)
{
	kasan_disable_current();
}

static inline void metadata_access_disable(void)
{
	kasan_enable_current();
}

Christoph Lameter's avatar
Christoph Lameter committed
507
508
509
/*
 * Object debugging
 */
Joonsoo Kim's avatar
Joonsoo Kim committed
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529

/* Verify that a pointer has an address that is valid within a slab page */
static inline int check_valid_pointer(struct kmem_cache *s,
				struct page *page, void *object)
{
	void *base;

	if (!object)
		return 1;

	base = page_address(page);
	object = restore_red_left(s, object);
	if (object < base || object >= base + page->objects * s->size ||
		(object - base) % s->size) {
		return 0;
	}

	return 1;
}

530
531
static void print_section(char *level, char *text, u8 *addr,
			  unsigned int length)
Christoph Lameter's avatar
Christoph Lameter committed
532
{
533
	metadata_access_enable();
534
	print_hex_dump(level, text, DUMP_PREFIX_ADDRESS, 16, 1, addr,
535
			length, 1);
536
	metadata_access_disable();
Christoph Lameter's avatar
Christoph Lameter committed
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
}

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,
553
			enum track_item alloc, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
554
{
Akinobu Mita's avatar
Akinobu Mita committed
555
	struct track *p = get_track(s, object, alloc);
Christoph Lameter's avatar
Christoph Lameter committed
556
557

	if (addr) {
558
559
560
561
562
563
564
565
#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;
566
		metadata_access_enable();
567
		save_stack_trace(&trace);
568
		metadata_access_disable();
569
570
571
572
573
574
575
576
577

		/* 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
578
579
		p->addr = addr;
		p->cpu = smp_processor_id();
580
		p->pid = current->pid;
Christoph Lameter's avatar
Christoph Lameter committed
581
582
583
584
585
586
587
		p->when = jiffies;
	} else
		memset(p, 0, sizeof(struct track));
}

static void init_tracking(struct kmem_cache *s, void *object)
{
588
589
590
	if (!(s->flags & SLAB_STORE_USER))
		return;

591
592
	set_track(s, object, TRACK_FREE, 0UL);
	set_track(s, object, TRACK_ALLOC, 0UL);
Christoph Lameter's avatar
Christoph Lameter committed
593
594
595
596
597
598
599
}

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

600
601
	pr_err("INFO: %s in %pS age=%lu cpu=%u pid=%d\n",
	       s, (void *)t->addr, jiffies - t->when, t->cpu, t->pid);
602
603
604
605
606
#ifdef CONFIG_STACKTRACE
	{
		int i;
		for (i = 0; i < TRACK_ADDRS_COUNT; i++)
			if (t->addrs[i])
607
				pr_err("\t%pS\n", (void *)t->addrs[i]);
608
609
610
611
			else
				break;
	}
#endif
612
613
614
615
616
617
618
619
620
621
622
623
624
}

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)
{
625
	pr_err("INFO: Slab 0x%p objects=%u used=%u fp=0x%p flags=0x%04lx\n",
626
	       page, page->objects, page->inuse, page->freelist, page->flags);
627
628
629
630
631

}

static void slab_bug(struct kmem_cache *s, char *fmt, ...)
{
632
	struct va_format vaf;
633
634
635
	va_list args;

	va_start(args, fmt);
636
637
	vaf.fmt = fmt;
	vaf.va = &args;
638
	pr_err("=============================================================================\n");
639
	pr_err("BUG %s (%s): %pV\n", s->name, print_tainted(), &vaf);
640
	pr_err("-----------------------------------------------------------------------------\n\n");
641

642
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
643
	va_end(args);
Christoph Lameter's avatar
Christoph Lameter committed
644
645
}

646
647
static void slab_fix(struct kmem_cache *s, char *fmt, ...)
{
648
	struct va_format vaf;
649
650
651
	va_list args;

	va_start(args, fmt);
652
653
654
	vaf.fmt = fmt;
	vaf.va = &args;
	pr_err("FIX %s: %pV\n", s->name, &vaf);
655
656
657
658
	va_end(args);
}

static void print_trailer(struct kmem_cache *s, struct page *page, u8 *p)
Christoph Lameter's avatar
Christoph Lameter committed
659
660
{
	unsigned int off;	/* Offset of last byte */
661
	u8 *addr = page_address(page);
662
663
664
665
666

	print_tracking(s, p);

	print_page_info(page);

667
668
	pr_err("INFO: Object 0x%p @offset=%tu fp=0x%p\n\n",
	       p, p - addr, get_freepointer(s, p));
669

Joonsoo Kim's avatar
Joonsoo Kim committed
670
	if (s->flags & SLAB_RED_ZONE)
671
672
		print_section(KERN_ERR, "Redzone ", p - s->red_left_pad,
			      s->red_left_pad);
Joonsoo Kim's avatar
Joonsoo Kim committed
673
	else if (p > addr + 16)
674
		print_section(KERN_ERR, "Bytes b4 ", p - 16, 16);
Christoph Lameter's avatar
Christoph Lameter committed
675

676
677
	print_section(KERN_ERR, "Object ", p,
		      min_t(unsigned long, s->object_size, PAGE_SIZE));
Christoph Lameter's avatar
Christoph Lameter committed
678
	if (s->flags & SLAB_RED_ZONE)
679
		print_section(KERN_ERR, "Redzone ", p + s->object_size,
680
			s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
681
682
683
684
685
686

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

687
	if (s->flags & SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
688
689
		off += 2 * sizeof(struct track);

690
691
	off += kasan_metadata_size(s);

Joonsoo Kim's avatar
Joonsoo Kim committed
692
	if (off != size_from_object(s))
Christoph Lameter's avatar
Christoph Lameter committed
693
		/* Beginning of the filler is the free pointer */
694
695
		print_section(KERN_ERR, "Padding ", p + off,
			      size_from_object(s) - off);
696
697

	dump_stack();
Christoph Lameter's avatar
Christoph Lameter committed
698
699
}

700
void object_err(struct kmem_cache *s, struct page *page,
Christoph Lameter's avatar
Christoph Lameter committed
701
702
			u8 *object, char *reason)
{
703
	slab_bug(s, "%s", reason);
704
	print_trailer(s, page, object);
Christoph Lameter's avatar
Christoph Lameter committed
705
706
}

707
708
static void slab_err(struct kmem_cache *s, struct page *page,
			const char *fmt, ...)
Christoph Lameter's avatar
Christoph Lameter committed
709
710
711
712
{
	va_list args;
	char buf[100];

713
714
	va_start(args, fmt);
	vsnprintf(buf, sizeof(buf), fmt, args);
Christoph Lameter's avatar
Christoph Lameter committed
715
	va_end(args);
716
	slab_bug(s, "%s", buf);
717
	print_page_info(page);
Christoph Lameter's avatar
Christoph Lameter committed
718
719
720
	dump_stack();
}

721
static void init_object(struct kmem_cache *s, void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
722
723
724
{
	u8 *p = object;

Joonsoo Kim's avatar
Joonsoo Kim committed
725
726
727
	if (s->flags & SLAB_RED_ZONE)
		memset(p - s->red_left_pad, val, s->red_left_pad);

Christoph Lameter's avatar
Christoph Lameter committed
728
	if (s->flags & __OBJECT_POISON) {
729
730
		memset(p, POISON_FREE, s->object_size - 1);
		p[s->object_size - 1] = POISON_END;
Christoph Lameter's avatar
Christoph Lameter committed
731
732
733
	}

	if (s->flags & SLAB_RED_ZONE)
734
		memset(p + s->object_size, val, s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
735
736
}

737
738
739
740
741
742
743
744
745
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,
746
			u8 *start, unsigned int value, unsigned int bytes)
747
748
749
750
{
	u8 *fault;
	u8 *end;

751
	metadata_access_enable();
752
	fault = memchr_inv(start, value, bytes);
753
	metadata_access_disable();
754
755
756
757
758
759
760
761
	if (!fault)
		return 1;

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

	slab_bug(s, "%s overwritten", what);
762
	pr_err("INFO: 0x%p-0x%p. First byte 0x%x instead of 0x%x\n",
763
764
765
766
767
					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
768
769
770
771
772
773
774
775
776
}

/*
 * 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
777
 *
Christoph Lameter's avatar
Christoph Lameter committed
778
779
780
 * 	Poisoning uses 0x6b (POISON_FREE) and the last byte is
 * 	0xa5 (POISON_END)
 *
781
 * object + s->object_size
Christoph Lameter's avatar
Christoph Lameter committed
782
 * 	Padding to reach word boundary. This is also used for Redzoning.
Christoph Lameter's avatar
Christoph Lameter committed
783
 * 	Padding is extended by another word if Redzoning is enabled and
784
 * 	object_size == inuse.
Christoph Lameter's avatar
Christoph Lameter committed
785
 *
Christoph Lameter's avatar
Christoph Lameter committed
786
787
788
789
 * 	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
790
791
 * 	Meta data starts here.
 *
Christoph Lameter's avatar
Christoph Lameter committed
792
793
 * 	A. Free pointer (if we cannot overwrite object on free)
 * 	B. Tracking data for SLAB_STORE_USER
Christoph Lameter's avatar
Christoph Lameter committed
794
 * 	C. Padding to reach required alignment boundary or at mininum
Christoph Lameter's avatar
Christoph Lameter committed
795
 * 		one word if debugging is on to be able to detect writes
Christoph Lameter's avatar
Christoph Lameter committed
796
797
798
 * 		before the word boundary.
 *
 *	Padding is done using 0x5a (POISON_INUSE)
Christoph Lameter's avatar
Christoph Lameter committed
799
800
 *
 * object + s->size
Christoph Lameter's avatar
Christoph Lameter committed
801
 * 	Nothing is used beyond s->size.
Christoph Lameter's avatar
Christoph Lameter committed
802
 *
803
 * If slabcaches are merged then the object_size and inuse boundaries are mostly
Christoph Lameter's avatar
Christoph Lameter committed
804
 * ignored. And therefore no slab options that rely on these boundaries
Christoph Lameter's avatar
Christoph Lameter committed
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
 * 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);

820
821
	off += kasan_metadata_size(s);

Joonsoo Kim's avatar
Joonsoo Kim committed
822
	if (size_from_object(s) == off)
Christoph Lameter's avatar
Christoph Lameter committed
823
824
		return 1;

825
	return check_bytes_and_report(s, page, p, "Object padding",
Joonsoo Kim's avatar
Joonsoo Kim committed
826
			p + off, POISON_INUSE, size_from_object(s) - off);
Christoph Lameter's avatar
Christoph Lameter committed
827
828
}

829
/* Check the pad bytes at the end of a slab page */
Christoph Lameter's avatar
Christoph Lameter committed
830
831
static int slab_pad_check(struct kmem_cache *s, struct page *page)
{
832
833
834
835
836
	u8 *start;
	u8 *fault;
	u8 *end;
	int length;
	int remainder;
Christoph Lameter's avatar
Christoph Lameter committed
837
838
839
840

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

841
	start = page_address(page);
842
	length = (PAGE_SIZE << compound_order(page)) - s->reserved;
843
844
	end = start + length;
	remainder = length % s->size;
Christoph Lameter's avatar
Christoph Lameter committed
845
846
847
	if (!remainder)
		return 1;

848
	metadata_access_enable();
849
	fault = memchr_inv(end - remainder, POISON_INUSE, remainder);
850
	metadata_access_disable();
851
852
853
854
855
856
	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);
857
	print_section(KERN_ERR, "Padding ", end - remainder, remainder);
858

Eric Dumazet's avatar
Eric Dumazet committed
859
	restore_bytes(s, "slab padding", POISON_INUSE, end - remainder, end);
860
	return 0;
Christoph Lameter's avatar
Christoph Lameter committed
861
862
863
}

static int check_object(struct kmem_cache *s, struct page *page,
864
					void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
865
866
{
	u8 *p = object;
867
	u8 *endobject = object + s->object_size;
Christoph Lameter's avatar
Christoph Lameter committed
868
869

	if (s->flags & SLAB_RED_ZONE) {
Joonsoo Kim's avatar
Joonsoo Kim committed
870
871
872
873
		if (!check_bytes_and_report(s, page, object, "Redzone",
			object - s->red_left_pad, val, s->red_left_pad))
			return 0;

874
		if (!check_bytes_and_report(s, page, object, "Redzone",
875
			endobject, val, s->inuse - s->object_size))
Christoph Lameter's avatar
Christoph Lameter committed
876
877
			return 0;
	} else {
878
		if ((s->flags & SLAB_POISON) && s->object_size < s->inuse) {
Ingo Molnar's avatar
Ingo Molnar committed
879
			check_bytes_and_report(s, page, p, "Alignment padding",
880
881
				endobject, POISON_INUSE,
				s->inuse - s->object_size);
Ingo Molnar's avatar
Ingo Molnar committed
882
		}
Christoph Lameter's avatar
Christoph Lameter committed
883
884
885
	}

	if (s->flags & SLAB_POISON) {
886
		if (val != SLUB_RED_ACTIVE && (s->flags & __OBJECT_POISON) &&
887
			(!check_bytes_and_report(s, page, p, "Poison", p,
888
					POISON_FREE, s->object_size - 1) ||
889
			 !check_bytes_and_report(s, page, p, "Poison",
890
				p + s->object_size - 1, POISON_END, 1)))
Christoph Lameter's avatar
Christoph Lameter committed
891
892
893
894
895
896
897
			return 0;
		/*
		 * check_pad_bytes cleans up on its own.
		 */
		check_pad_bytes(s, page, p);
	}

898
	if (!s->offset && val == SLUB_RED_ACTIVE)
Christoph Lameter's avatar
Christoph Lameter committed
899
900
901
902
903
904
905
906
907
908
		/*
		 * 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
909
		 * No choice but to zap it and thus lose the remainder
Christoph Lameter's avatar
Christoph Lameter committed
910
		 * of the free objects in this slab. May cause
Christoph Lameter's avatar
Christoph Lameter committed
911
		 * another error because the object count is now wrong.
Christoph Lameter's avatar
Christoph Lameter committed
912
		 */
913
		set_freepointer(s, p, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
914
915
916
917
918
919
920
		return 0;
	}
	return 1;
}

static int check_slab(struct kmem_cache *s, struct page *page)
{
921
922
	int maxobj;

Christoph Lameter's avatar
Christoph Lameter committed
923
924
925
	VM_BUG_ON(!irqs_disabled());

	if (!PageSlab(page)) {
926
		slab_err(s, page, "Not a valid slab page");
Christoph Lameter's avatar
Christoph Lameter committed
927
928
		return 0;
	}
929

930
	maxobj = order_objects(compound_order(page), s->size, s->reserved);
931
932
	if (page->objects > maxobj) {
		slab_err(s, page, "objects %u > max %u",
933
			page->objects, maxobj);
934
935
936
		return 0;
	}
	if (page->inuse > page->objects) {
937
		slab_err(s, page, "inuse %u > max %u",
938
			page->inuse, page->objects);
Christoph Lameter's avatar
Christoph Lameter committed
939
940
941
942
943
944
945
946
		return 0;
	}
	/* Slab_pad_check fixes things up after itself */
	slab_pad_check(s, page);
	return 1;
}

/*
Christoph Lameter's avatar
Christoph Lameter committed
947
948
 * 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
949
950
951
952
 */
static int on_freelist(struct kmem_cache *s, struct page *page, void *search)
{
	int nr = 0;
953
	void *fp;
Christoph Lameter's avatar
Christoph Lameter committed
954
	void *object = NULL;
955
	int max_objects;
Christoph Lameter's avatar
Christoph Lameter committed
956

957
	fp = page->freelist;
958
	while (fp && nr <= page->objects) {
Christoph Lameter's avatar
Christoph Lameter committed
959
960
961
962
963
964
		if (fp == search)
			return 1;
		if (!check_valid_pointer(s, page, fp)) {
			if (object) {
				object_err(s, page, object,
					"Freechain corrupt");
965
				set_freepointer(s, object, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
966
			} else {
967
				slab_err(s, page, "Freepointer corrupt");
968
				page->freelist = NULL;
969
				page->inuse = page->objects;
970
				slab_fix(s, "Freelist cleared");
Christoph Lameter's avatar
Christoph Lameter committed
971
972
973
974
975
976
977
978
979
				return 0;
			}
			break;
		}
		object = fp;
		fp = get_freepointer(s, object);
		nr++;
	}

980
	max_objects = order_objects(compound_order(page), s->size, s->reserved);
981
982
	if (max_objects > MAX_OBJS_PER_PAGE)
		max_objects = MAX_OBJS_PER_PAGE;
983
984

	if (page->objects != max_objects) {
Joe Perches's avatar
Joe Perches committed
985
986
		slab_err(s, page, "Wrong number of objects. Found %d but should be %d",
			 page->objects, max_objects);
987
988
989
		page->objects = max_objects;
		slab_fix(s, "Number of objects adjusted.");
	}
990
	if (page->inuse != page->objects - nr) {
Joe Perches's avatar
Joe Perches committed
991
992
		slab_err(s, page, "Wrong object count. Counter is %d but counted were %d",
			 page->inuse, page->objects - nr);
993
		page->inuse = page->objects - nr;
994
		slab_fix(s, "Object count adjusted.");
Christoph Lameter's avatar
Christoph Lameter committed
995
996
997
998
	}
	return search == NULL;
}

999
1000
static void trace(struct kmem_cache *s, struct page *page, void *object,
								int alloc)
Christoph Lameter's avatar
Christoph Lameter committed
1001
1002
{
	if (s->flags & SLAB_TRACE) {
1003
		pr_info("TRACE %s %s 0x%p inuse=%d fp=0x%p\n",
Christoph Lameter's avatar
Christoph Lameter committed
1004
1005
1006
1007
1008
1009
			s->name,
			alloc ? "alloc" : "free",
			object, page->inuse,
			page->freelist);

		if (!alloc)
1010
			print_section(KERN_INFO, "Object ", (void *)object,
1011
					s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
1012
1013
1014
1015
1016

		dump_stack();
	}
}

1017
/*
Christoph Lameter's avatar
Christoph Lameter committed
1018
 * Tracking of fully allocated slabs for debugging purposes.
1019
 */
1020
1021
static void add_full(struct kmem_cache *s,
	struct kmem_cache_node *n, struct page *page)
1022
{
1023
1024
1025
	if (!(s->flags & SLAB_STORE_USER))
		return;

1026
	lockdep_assert_held(&n->list_lock);
1027
1028
1029
	list_add(&page->lru, &n->full);
}

Peter Zijlstra's avatar
Peter Zijlstra committed
1030
static void remove_full(struct kmem_cache *s, struct kmem_cache_node *n, struct page *page)
1031
1032
1033
1034
{
	if (!(s->flags & SLAB_STORE_USER))
		return;

1035
	lockdep_assert_held(&n->list_lock);
1036
1037
1038
	list_del(&page->lru);
}

1039
1040
1041
1042
1043
1044
1045
1046
/* 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);
}

1047
1048
1049
1050
1051
static inline unsigned long node_nr_slabs(struct kmem_cache_node *n)
{
	return atomic_long_read(&n->nr_slabs);
}

1052
static inline void inc_slabs_node(struct kmem_cache *s, int node, int objects)
1053
1054
1055
1056
1057
1058
1059
1060
1061
{
	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).
	 */
1062
	if (likely(n)) {
1063
		atomic_long_inc(&n->nr_slabs);
1064
1065
		atomic_long_add(objects, &n->total_objects);
	}
1066
}
1067
static inline void dec_slabs_node(struct kmem_cache *s, int node, int objects)
1068
1069
1070
1071
{
	struct kmem_cache_node *n = get_node(s, node);

	atomic_long_dec(&n->nr_slabs);
1072
	atomic_long_sub(objects, &n->total_objects);
1073
1074
1075
}

/* Object debug checks for alloc/free paths */
Christoph Lameter's avatar
Christoph Lameter committed
1076
1077
1078
1079
1080
1081
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;

1082
	init_object(s, object, SLUB_RED_INACTIVE);