slub.c 144 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0
Christoph Lameter's avatar
Christoph Lameter committed
2
3
4
5
/*
 * SLUB: A slab allocator that limits cache line use instead of queuing
 * objects in per cpu and per node lists.
 *
6
7
 * 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
8
 *
Christoph Lameter's avatar
Christoph Lameter committed
9
 * (C) 2007 SGI, Christoph Lameter
10
 * (C) 2011 Linux Foundation, Christoph Lameter
Christoph Lameter's avatar
Christoph Lameter committed
11
12
13
 */

#include <linux/mm.h>
Nick Piggin's avatar
Nick Piggin committed
14
#include <linux/swap.h> /* struct reclaim_state */
Christoph Lameter's avatar
Christoph Lameter committed
15
16
17
18
19
#include <linux/module.h>
#include <linux/bit_spinlock.h>
#include <linux/interrupt.h>
#include <linux/bitops.h>
#include <linux/slab.h>
20
#include "slab.h"
21
#include <linux/proc_fs.h>
Christoph Lameter's avatar
Christoph Lameter committed
22
#include <linux/seq_file.h>
23
#include <linux/kasan.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>
36
#include <linux/random.h>
Christoph Lameter's avatar
Christoph Lameter committed
37

38
39
#include <trace/events/kmem.h>

40
41
#include "internal.h"

Christoph Lameter's avatar
Christoph Lameter committed
42
43
/*
 * Lock order:
44
 *   1. slab_mutex (Global Mutex)
45
46
 *   2. node->list_lock
 *   3. slab_lock(page) (Only on some arches and for debugging)
Christoph Lameter's avatar
Christoph Lameter committed
47
 *
48
 *   slab_mutex
49
 *
50
 *   The role of the slab_mutex is to protect the list of all the slabs
51
52
53
 *   and to synchronize major metadata changes to slab cache structures.
 *
 *   The slab_lock is only used for debugging and on arches that do not
54
 *   have the ability to do a cmpxchg_double. It only protects:
55
 *	A. page->freelist	-> List of object free in a page
56
57
58
 *	B. page->inuse		-> Number of objects in use
 *	C. page->objects	-> Number of objects in page
 *	D. page->frozen		-> frozen state
59
60
 *
 *   If a slab is frozen then it is exempt from list management. It is not
61
62
63
64
65
 *   on any list except per cpu partial 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 */
195
/* Poison object */
196
#define __OBJECT_POISON		((slab_flags_t __force)0x80000000U)
197
/* Use cmpxchg_double */
198
#define __CMPXCHG_DOUBLE	((slab_flags_t __force)0x40000000U)
Christoph Lameter's avatar
Christoph Lameter committed
199

200
201
202
/*
 * Tracking user of a slab.
 */
203
#define TRACK_ADDRS_COUNT 16
204
struct track {
205
	unsigned long addr;	/* Called from address */
206
207
208
#ifdef CONFIG_STACKTRACE
	unsigned long addrs[TRACK_ADDRS_COUNT];	/* Called from address */
#endif
209
210
211
212
213
214
215
	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 };

216
#ifdef CONFIG_SYSFS
Christoph Lameter's avatar
Christoph Lameter committed
217
218
static int sysfs_slab_add(struct kmem_cache *);
static int sysfs_slab_alias(struct kmem_cache *, const char *);
219
static void memcg_propagate_slab_attrs(struct kmem_cache *s);
220
static void sysfs_slab_remove(struct kmem_cache *s);
Christoph Lameter's avatar
Christoph Lameter committed
221
#else
222
223
224
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; }
225
static inline void memcg_propagate_slab_attrs(struct kmem_cache *s) { }
226
static inline void sysfs_slab_remove(struct kmem_cache *s) { }
Christoph Lameter's avatar
Christoph Lameter committed
227
228
#endif

229
static inline void stat(const struct kmem_cache *s, enum stat_item si)
230
231
{
#ifdef CONFIG_SLUB_STATS
232
233
234
235
236
	/*
	 * 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]);
237
238
239
#endif
}

Christoph Lameter's avatar
Christoph Lameter committed
240
241
242
243
/********************************************************************
 * 			Core slab cache functions
 *******************************************************************/

244
245
246
247
248
249
250
251
252
/*
 * 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
253
254
255
256
257
258
259
260
261
262
263
264
	/*
	 * When CONFIG_KASAN_SW_TAGS is enabled, ptr_addr might be tagged.
	 * Normally, this doesn't cause any issues, as both set_freepointer()
	 * and get_freepointer() are called with a pointer with the same tag.
	 * However, there are some issues with CONFIG_SLUB_DEBUG code. For
	 * example, when __free_slub() iterates over objects in a cache, it
	 * passes untagged pointers to check_object(). check_object() in turns
	 * calls get_freepointer() with an untagged pointer, which causes the
	 * freepointer to be restored incorrectly.
	 */
	return (void *)((unsigned long)ptr ^ s->random ^
			(unsigned long)kasan_reset_tag((void *)ptr_addr));
265
266
267
268
269
270
271
272
273
274
275
276
277
#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);
}

278
279
static inline void *get_freepointer(struct kmem_cache *s, void *object)
{
280
	return freelist_dereference(s, object + s->offset);
281
282
}

283
284
static void prefetch_freepointer(const struct kmem_cache *s, void *object)
{
285
	prefetch(object + s->offset);
286
287
}

288
289
static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
{
290
	unsigned long freepointer_addr;
291
292
	void *p;

293
294
295
	if (!debug_pagealloc_enabled())
		return get_freepointer(s, object);

296
297
298
	freepointer_addr = (unsigned long)object + s->offset;
	probe_kernel_read(&p, (void **)freepointer_addr, sizeof(p));
	return freelist_ptr(s, p, freepointer_addr);
299
300
}

301
302
static inline void set_freepointer(struct kmem_cache *s, void *object, void *fp)
{
303
304
	unsigned long freeptr_addr = (unsigned long)object + s->offset;

305
306
307
308
#ifdef CONFIG_SLAB_FREELIST_HARDENED
	BUG_ON(object == fp); /* naive detection of double free or corruption */
#endif

309
	*(void **)freeptr_addr = freelist_ptr(s, fp, freeptr_addr);
310
311
312
}

/* Loop over all objects in a slab */
313
#define for_each_object(__p, __s, __addr, __objects) \
Joonsoo Kim's avatar
Joonsoo Kim committed
314
315
316
	for (__p = fixup_red_left(__s, __addr); \
		__p < (__addr) + (__objects) * (__s)->size; \
		__p += (__s)->size)
317
318

/* Determine object index from a given position */
319
static inline unsigned int slab_index(void *p, struct kmem_cache *s, void *addr)
320
{
321
	return (kasan_reset_tag(p) - addr) / s->size;
322
323
}

324
static inline unsigned int order_objects(unsigned int order, unsigned int size)
325
{
326
	return ((unsigned int)PAGE_SIZE << order) / size;
327
328
}

329
static inline struct kmem_cache_order_objects oo_make(unsigned int order,
330
		unsigned int size)
331
332
{
	struct kmem_cache_order_objects x = {
333
		(order << OO_SHIFT) + order_objects(order, size)
334
335
336
337
338
	};

	return x;
}

339
static inline unsigned int oo_order(struct kmem_cache_order_objects x)
340
{
341
	return x.x >> OO_SHIFT;
342
343
}

344
static inline unsigned int oo_objects(struct kmem_cache_order_objects x)
345
{
346
	return x.x & OO_MASK;
347
348
}

349
350
351
352
353
/*
 * Per slab locking using the pagelock
 */
static __always_inline void slab_lock(struct page *page)
{
354
	VM_BUG_ON_PAGE(PageTail(page), page);
355
356
357
358
359
	bit_spin_lock(PG_locked, &page->flags);
}

static __always_inline void slab_unlock(struct page *page)
{
360
	VM_BUG_ON_PAGE(PageTail(page), page);
361
362
363
	__bit_spin_unlock(PG_locked, &page->flags);
}

364
365
366
367
368
369
370
/* 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());
371
372
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
373
	if (s->flags & __CMPXCHG_DOUBLE) {
374
		if (cmpxchg_double(&page->freelist, &page->counters,
375
376
				   freelist_old, counters_old,
				   freelist_new, counters_new))
377
			return true;
378
379
380
381
	} else
#endif
	{
		slab_lock(page);
382
383
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
384
			page->freelist = freelist_new;
385
			page->counters = counters_new;
386
			slab_unlock(page);
387
			return true;
388
389
390
391
392
393
394
395
		}
		slab_unlock(page);
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
396
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
397
398
#endif

399
	return false;
400
401
}

402
403
404
405
406
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)
{
407
408
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
409
	if (s->flags & __CMPXCHG_DOUBLE) {
410
		if (cmpxchg_double(&page->freelist, &page->counters,
411
412
				   freelist_old, counters_old,
				   freelist_new, counters_new))
413
			return true;
414
415
416
	} else
#endif
	{
417
418
419
		unsigned long flags;

		local_irq_save(flags);
420
		slab_lock(page);
421
422
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
423
			page->freelist = freelist_new;
424
			page->counters = counters_new;
425
			slab_unlock(page);
426
			local_irq_restore(flags);
427
			return true;
428
		}
429
		slab_unlock(page);
430
		local_irq_restore(flags);
431
432
433
434
435
436
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
437
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
438
439
#endif

440
	return false;
441
442
}

443
#ifdef CONFIG_SLUB_DEBUG
444
445
446
/*
 * Determine a map of object in use on a page.
 *
447
 * Node listlock must be held to guarantee that the page does
448
449
450
451
452
453
454
455
456
457
458
 * 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);
}

459
static inline unsigned int size_from_object(struct kmem_cache *s)
Joonsoo Kim's avatar
Joonsoo Kim committed
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
{
	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;
}

475
476
477
/*
 * Debug settings:
 */
478
#if defined(CONFIG_SLUB_DEBUG_ON)
479
static slab_flags_t slub_debug = DEBUG_DEFAULT_FLAGS;
480
#else
481
static slab_flags_t slub_debug;
482
#endif
483
484

static char *slub_debug_slabs;
485
static int disable_higher_order_debug;
486

487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
/*
 * 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
503
504
505
/*
 * Object debugging
 */
Joonsoo Kim's avatar
Joonsoo Kim committed
506
507
508
509
510
511
512
513
514
515
516

/* 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);
517
	object = kasan_reset_tag(object);
Joonsoo Kim's avatar
Joonsoo Kim committed
518
519
520
521
522
523
524
525
526
	object = restore_red_left(s, object);
	if (object < base || object >= base + page->objects * s->size ||
		(object - base) % s->size) {
		return 0;
	}

	return 1;
}

527
528
static void print_section(char *level, char *text, u8 *addr,
			  unsigned int length)
Christoph Lameter's avatar
Christoph Lameter committed
529
{
530
	metadata_access_enable();
531
	print_hex_dump(level, text, DUMP_PREFIX_ADDRESS, 16, 1, addr,
532
			length, 1);
533
	metadata_access_disable();
Christoph Lameter's avatar
Christoph Lameter committed
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
}

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,
550
			enum track_item alloc, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
551
{
Akinobu Mita's avatar
Akinobu Mita committed
552
	struct track *p = get_track(s, object, alloc);
Christoph Lameter's avatar
Christoph Lameter committed
553
554

	if (addr) {
555
#ifdef CONFIG_STACKTRACE
556
		unsigned int nr_entries;
557

558
		metadata_access_enable();
559
		nr_entries = stack_trace_save(p->addrs, TRACK_ADDRS_COUNT, 3);
560
		metadata_access_disable();
561

562
563
		if (nr_entries < TRACK_ADDRS_COUNT)
			p->addrs[nr_entries] = 0;
564
#endif
Christoph Lameter's avatar
Christoph Lameter committed
565
566
		p->addr = addr;
		p->cpu = smp_processor_id();
567
		p->pid = current->pid;
Christoph Lameter's avatar
Christoph Lameter committed
568
		p->when = jiffies;
569
	} else {
Christoph Lameter's avatar
Christoph Lameter committed
570
		memset(p, 0, sizeof(struct track));
571
	}
Christoph Lameter's avatar
Christoph Lameter committed
572
573
574
575
}

static void init_tracking(struct kmem_cache *s, void *object)
{
576
577
578
	if (!(s->flags & SLAB_STORE_USER))
		return;

579
580
	set_track(s, object, TRACK_FREE, 0UL);
	set_track(s, object, TRACK_ALLOC, 0UL);
Christoph Lameter's avatar
Christoph Lameter committed
581
582
}

583
static void print_track(const char *s, struct track *t, unsigned long pr_time)
Christoph Lameter's avatar
Christoph Lameter committed
584
585
586
587
{
	if (!t->addr)
		return;

588
	pr_err("INFO: %s in %pS age=%lu cpu=%u pid=%d\n",
589
	       s, (void *)t->addr, pr_time - t->when, t->cpu, t->pid);
590
591
592
593
594
#ifdef CONFIG_STACKTRACE
	{
		int i;
		for (i = 0; i < TRACK_ADDRS_COUNT; i++)
			if (t->addrs[i])
595
				pr_err("\t%pS\n", (void *)t->addrs[i]);
596
597
598
599
			else
				break;
	}
#endif
600
601
602
603
}

static void print_tracking(struct kmem_cache *s, void *object)
{
604
	unsigned long pr_time = jiffies;
605
606
607
	if (!(s->flags & SLAB_STORE_USER))
		return;

608
609
	print_track("Allocated", get_track(s, object, TRACK_ALLOC), pr_time);
	print_track("Freed", get_track(s, object, TRACK_FREE), pr_time);
610
611
612
613
}

static void print_page_info(struct page *page)
{
614
	pr_err("INFO: Slab 0x%p objects=%u used=%u fp=0x%p flags=0x%04lx\n",
615
	       page, page->objects, page->inuse, page->freelist, page->flags);
616
617
618
619
620

}

static void slab_bug(struct kmem_cache *s, char *fmt, ...)
{
621
	struct va_format vaf;
622
623
624
	va_list args;

	va_start(args, fmt);
625
626
	vaf.fmt = fmt;
	vaf.va = &args;
627
	pr_err("=============================================================================\n");
628
	pr_err("BUG %s (%s): %pV\n", s->name, print_tainted(), &vaf);
629
	pr_err("-----------------------------------------------------------------------------\n\n");
630

631
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
632
	va_end(args);
Christoph Lameter's avatar
Christoph Lameter committed
633
634
}

635
636
static void slab_fix(struct kmem_cache *s, char *fmt, ...)
{
637
	struct va_format vaf;
638
639
640
	va_list args;

	va_start(args, fmt);
641
642
643
	vaf.fmt = fmt;
	vaf.va = &args;
	pr_err("FIX %s: %pV\n", s->name, &vaf);
644
645
646
647
	va_end(args);
}

static void print_trailer(struct kmem_cache *s, struct page *page, u8 *p)
Christoph Lameter's avatar
Christoph Lameter committed
648
649
{
	unsigned int off;	/* Offset of last byte */
650
	u8 *addr = page_address(page);
651
652
653
654
655

	print_tracking(s, p);

	print_page_info(page);

656
657
	pr_err("INFO: Object 0x%p @offset=%tu fp=0x%p\n\n",
	       p, p - addr, get_freepointer(s, p));
658

Joonsoo Kim's avatar
Joonsoo Kim committed
659
	if (s->flags & SLAB_RED_ZONE)
660
661
		print_section(KERN_ERR, "Redzone ", p - s->red_left_pad,
			      s->red_left_pad);
Joonsoo Kim's avatar
Joonsoo Kim committed
662
	else if (p > addr + 16)
663
		print_section(KERN_ERR, "Bytes b4 ", p - 16, 16);
Christoph Lameter's avatar
Christoph Lameter committed
664

665
	print_section(KERN_ERR, "Object ", p,
666
		      min_t(unsigned int, s->object_size, PAGE_SIZE));
Christoph Lameter's avatar
Christoph Lameter committed
667
	if (s->flags & SLAB_RED_ZONE)
668
		print_section(KERN_ERR, "Redzone ", p + s->object_size,
669
			s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
670
671
672
673
674
675

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

676
	if (s->flags & SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
677
678
		off += 2 * sizeof(struct track);

679
680
	off += kasan_metadata_size(s);

Joonsoo Kim's avatar
Joonsoo Kim committed
681
	if (off != size_from_object(s))
Christoph Lameter's avatar
Christoph Lameter committed
682
		/* Beginning of the filler is the free pointer */
683
684
		print_section(KERN_ERR, "Padding ", p + off,
			      size_from_object(s) - off);
685
686

	dump_stack();
Christoph Lameter's avatar
Christoph Lameter committed
687
688
}

689
void object_err(struct kmem_cache *s, struct page *page,
Christoph Lameter's avatar
Christoph Lameter committed
690
691
			u8 *object, char *reason)
{
692
	slab_bug(s, "%s", reason);
693
	print_trailer(s, page, object);
Christoph Lameter's avatar
Christoph Lameter committed
694
695
}

696
static __printf(3, 4) void slab_err(struct kmem_cache *s, struct page *page,
697
			const char *fmt, ...)
Christoph Lameter's avatar
Christoph Lameter committed
698
699
700
701
{
	va_list args;
	char buf[100];

702
703
	va_start(args, fmt);
	vsnprintf(buf, sizeof(buf), fmt, args);
Christoph Lameter's avatar
Christoph Lameter committed
704
	va_end(args);
705
	slab_bug(s, "%s", buf);
706
	print_page_info(page);
Christoph Lameter's avatar
Christoph Lameter committed
707
708
709
	dump_stack();
}

710
static void init_object(struct kmem_cache *s, void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
711
712
713
{
	u8 *p = object;

Joonsoo Kim's avatar
Joonsoo Kim committed
714
715
716
	if (s->flags & SLAB_RED_ZONE)
		memset(p - s->red_left_pad, val, s->red_left_pad);

Christoph Lameter's avatar
Christoph Lameter committed
717
	if (s->flags & __OBJECT_POISON) {
718
719
		memset(p, POISON_FREE, s->object_size - 1);
		p[s->object_size - 1] = POISON_END;
Christoph Lameter's avatar
Christoph Lameter committed
720
721
722
	}

	if (s->flags & SLAB_RED_ZONE)
723
		memset(p + s->object_size, val, s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
724
725
}

726
727
728
729
730
731
732
733
734
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,
735
			u8 *start, unsigned int value, unsigned int bytes)
736
737
738
739
{
	u8 *fault;
	u8 *end;

740
	metadata_access_enable();
741
	fault = memchr_inv(start, value, bytes);
742
	metadata_access_disable();
743
744
745
746
747
748
749
750
	if (!fault)
		return 1;

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

	slab_bug(s, "%s overwritten", what);
751
	pr_err("INFO: 0x%p-0x%p. First byte 0x%x instead of 0x%x\n",
752
753
754
755
756
					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
757
758
759
760
761
762
763
764
765
}

/*
 * 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
766
 *
Christoph Lameter's avatar
Christoph Lameter committed
767
768
769
 * 	Poisoning uses 0x6b (POISON_FREE) and the last byte is
 * 	0xa5 (POISON_END)
 *
770
 * object + s->object_size
Christoph Lameter's avatar
Christoph Lameter committed
771
 * 	Padding to reach word boundary. This is also used for Redzoning.
Christoph Lameter's avatar
Christoph Lameter committed
772
 * 	Padding is extended by another word if Redzoning is enabled and
773
 * 	object_size == inuse.
Christoph Lameter's avatar
Christoph Lameter committed
774
 *
Christoph Lameter's avatar
Christoph Lameter committed
775
776
777
778
 * 	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
779
780
 * 	Meta data starts here.
 *
Christoph Lameter's avatar
Christoph Lameter committed
781
782
 * 	A. Free pointer (if we cannot overwrite object on free)
 * 	B. Tracking data for SLAB_STORE_USER
Christoph Lameter's avatar
Christoph Lameter committed
783
 * 	C. Padding to reach required alignment boundary or at mininum
Christoph Lameter's avatar
Christoph Lameter committed
784
 * 		one word if debugging is on to be able to detect writes
Christoph Lameter's avatar
Christoph Lameter committed
785
786
787
 * 		before the word boundary.
 *
 *	Padding is done using 0x5a (POISON_INUSE)
Christoph Lameter's avatar
Christoph Lameter committed
788
789
 *
 * object + s->size
Christoph Lameter's avatar
Christoph Lameter committed
790
 * 	Nothing is used beyond s->size.
Christoph Lameter's avatar
Christoph Lameter committed
791
 *
792
 * If slabcaches are merged then the object_size and inuse boundaries are mostly
Christoph Lameter's avatar
Christoph Lameter committed
793
 * ignored. And therefore no slab options that rely on these boundaries
Christoph Lameter's avatar
Christoph Lameter committed
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
 * 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);

809
810
	off += kasan_metadata_size(s);

Joonsoo Kim's avatar
Joonsoo Kim committed
811
	if (size_from_object(s) == off)
Christoph Lameter's avatar
Christoph Lameter committed
812
813
		return 1;

814
	return check_bytes_and_report(s, page, p, "Object padding",
Joonsoo Kim's avatar
Joonsoo Kim committed
815
			p + off, POISON_INUSE, size_from_object(s) - off);
Christoph Lameter's avatar
Christoph Lameter committed
816
817
}

818
/* Check the pad bytes at the end of a slab page */
Christoph Lameter's avatar
Christoph Lameter committed
819
820
static int slab_pad_check(struct kmem_cache *s, struct page *page)
{
821
822
823
	u8 *start;
	u8 *fault;
	u8 *end;
824
	u8 *pad;
825
826
	int length;
	int remainder;
Christoph Lameter's avatar
Christoph Lameter committed
827
828
829
830

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

831
	start = page_address(page);
832
	length = page_size(page);
833
834
	end = start + length;
	remainder = length % s->size;
Christoph Lameter's avatar
Christoph Lameter committed
835
836
837
	if (!remainder)
		return 1;

838
	pad = end - remainder;
839
	metadata_access_enable();
840
	fault = memchr_inv(pad, POISON_INUSE, remainder);
841
	metadata_access_disable();
842
843
844
845
846
847
	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);
848
	print_section(KERN_ERR, "Padding ", pad, remainder);
849

850
	restore_bytes(s, "slab padding", POISON_INUSE, fault, end);
851
	return 0;
Christoph Lameter's avatar
Christoph Lameter committed
852
853
854
}

static int check_object(struct kmem_cache *s, struct page *page,
855
					void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
856
857
{
	u8 *p = object;
858
	u8 *endobject = object + s->object_size;
Christoph Lameter's avatar
Christoph Lameter committed
859
860

	if (s->flags & SLAB_RED_ZONE) {
Joonsoo Kim's avatar
Joonsoo Kim committed
861
862
863
864
		if (!check_bytes_and_report(s, page, object, "Redzone",
			object - s->red_left_pad, val, s->red_left_pad))
			return 0;

865
		if (!check_bytes_and_report(s, page, object, "Redzone",
866
			endobject, val, s->inuse - s->object_size))
Christoph Lameter's avatar
Christoph Lameter committed
867
868
			return 0;
	} else {
869
		if ((s->flags & SLAB_POISON) && s->object_size < s->inuse) {
Ingo Molnar's avatar
Ingo Molnar committed
870
			check_bytes_and_report(s, page, p, "Alignment padding",
871
872
				endobject, POISON_INUSE,
				s->inuse - s->object_size);
Ingo Molnar's avatar
Ingo Molnar committed
873
		}
Christoph Lameter's avatar
Christoph Lameter committed
874
875
876
	}

	if (s->flags & SLAB_POISON) {
877
		if (val != SLUB_RED_ACTIVE && (s->flags & __OBJECT_POISON) &&
878
			(!check_bytes_and_report(s, page, p, "Poison", p,
879
					POISON_FREE, s->object_size - 1) ||
880
			 !check_bytes_and_report(s, page, p, "Poison",
881
				p + s->object_size - 1, POISON_END, 1)))
Christoph Lameter's avatar
Christoph Lameter committed
882
883
884
885
886
887
888
			return 0;
		/*
		 * check_pad_bytes cleans up on its own.
		 */
		check_pad_bytes(s, page, p);
	}

889
	if (!s->offset && val == SLUB_RED_ACTIVE)
Christoph Lameter's avatar
Christoph Lameter committed
890
891
892
893
894
895
896
897
898
899
		/*
		 * 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
900
		 * No choice but to zap it and thus lose the remainder
Christoph Lameter's avatar
Christoph Lameter committed
901
		 * of the free objects in this slab. May cause
Christoph Lameter's avatar
Christoph Lameter committed
902
		 * another error because the object count is now wrong.
Christoph Lameter's avatar
Christoph Lameter committed
903
		 */
904
		set_freepointer(s, p, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
905
906
907
908
909
910
911
		return 0;
	}
	return 1;
}

static int check_slab(struct kmem_cache *s, struct page *page)
{
912
913
	int maxobj;

Christoph Lameter's avatar
Christoph Lameter committed
914
915
916
	VM_BUG_ON(!irqs_disabled());

	if (!PageSlab(page)) {
917
		slab_err(s, page, "Not a valid slab page");
Christoph Lameter's avatar
Christoph Lameter committed
918
919
		return 0;
	}
920

921
	maxobj = order_objects(compound_order(page), s->size);
922
923
	if (page->objects > maxobj) {
		slab_err(s, page, "objects %u > max %u",
924
			page->objects, maxobj);
925
926
927
		return 0;
	}
	if (page->inuse > page->objects) {
928
		slab_err(s, page, "inuse %u > max %u",
929
			page->inuse, page->objects);
Christoph Lameter's avatar
Christoph Lameter committed
930
931
932
933
934
935
936
937
		return 0;
	}
	/* Slab_pad_check fixes things up after itself */
	slab_pad_check(s, page);
	return 1;
}

/*
Christoph Lameter's avatar
Christoph Lameter committed
938
939
 * 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
940
941
942
943
 */
static int on_freelist(struct kmem_cache *s, struct page *page, void *search)
{
	int nr = 0;
944
	void *fp;
Christoph Lameter's avatar
Christoph Lameter committed
945
	void *object = NULL;
946
	int max_objects;
Christoph Lameter's avatar
Christoph Lameter committed
947

948
	fp = page->freelist;
949
	while (fp && nr <= page->objects) {
Christoph Lameter's avatar
Christoph Lameter committed
950
951
952
953
954
955
		if (fp == search)
			return 1;
		if (!check_valid_pointer(s, page, fp)) {
			if (object) {
				object_err(s, page, object,
					"Freechain corrupt");
956
				set_freepointer(s, object, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
957
			} else {
958
				slab_err(s, page, "Freepointer corrupt");
959
				page->freelist = NULL;
960
				page->inuse = page->objects;
961
				slab_fix(s, "Freelist cleared");
Christoph Lameter's avatar
Christoph Lameter committed
962
963
964
965
966
967
968
969
970
				return 0;
			}
			break;
		}
		object = fp;
		fp = get_freepointer(s, object);
		nr++;
	}

971
	max_objects = order_objects(compound_order(page), s->size);
972
973
	if (max_objects > MAX_OBJS_PER_PAGE)
		max_objects = MAX_OBJS_PER_PAGE;
974
975

	if (page->objects != max_objects) {
Joe Perches's avatar
Joe Perches committed
976
977
		slab_err(s, page, "Wrong number of objects. Found %d but should be %d",
			 page->objects, max_objects);
978
979
980
		page->objects = max_objects;
		slab_fix(s, "Number of objects adjusted.");
	}
981
	if (page->inuse != page->objects - nr) {
Joe Perches's avatar
Joe Perches committed
982
983
		slab_err(s, page, "Wrong object count. Counter is %d but counted were %d",
			 page->inuse, page->objects - nr);
984
		page->inuse = page->objects - nr;
985
		slab_fix(s, "Object count adjusted.");
Christoph Lameter's avatar
Christoph Lameter committed
986
987
988
989
	}
	return search == NULL;
}

990
991
static void trace(struct kmem_cache *s, struct page *page, void *object,
								int alloc)
Christoph Lameter's avatar
Christoph Lameter committed
992
993
{
	if (s->flags & SLAB_TRACE) {
994
		pr_info("TRACE %s %s 0x%p inuse=%d fp=0x%p\n",
Christoph Lameter's avatar
Christoph Lameter committed
995
996
997
998
999
1000
			s->name,
			alloc ? "alloc" : "free",
			object, page->inuse,
			page->freelist);

		if (!alloc)
1001
			print_section(KERN_INFO, "Object ", (void *)object,
1002
					s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
1003
1004
1005
1006
1007

		dump_stack();
	}
}

1008
/*
Christoph Lameter's avatar
Christoph Lameter committed
1009
 * Tracking of fully allocated slabs for debugging purposes.
1010
 */
1011
1012
static void add_full(struct kmem_cache *s,
	struct kmem_cache_node *n, struct page *page)
1013
{