slub.c 142 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>
22
#include <linux/notifier.h>
Christoph Lameter's avatar
Christoph Lameter committed
23
#include <linux/seq_file.h>
24
#include <linux/kasan.h>
Vegard Nossum's avatar
Vegard Nossum committed
25
#include <linux/kmemcheck.h>
Christoph Lameter's avatar
Christoph Lameter committed
26
27
28
29
#include <linux/cpu.h>
#include <linux/cpuset.h>
#include <linux/mempolicy.h>
#include <linux/ctype.h>
30
#include <linux/debugobjects.h>
Christoph Lameter's avatar
Christoph Lameter committed
31
#include <linux/kallsyms.h>
32
#include <linux/memory.h>
Roman Zippel's avatar
Roman Zippel committed
33
#include <linux/math64.h>
Akinobu Mita's avatar
Akinobu Mita committed
34
#include <linux/fault-inject.h>
35
#include <linux/stacktrace.h>
36
#include <linux/prefetch.h>
37
#include <linux/memcontrol.h>
38
#include <linux/random.h>
Christoph Lameter's avatar
Christoph Lameter committed
39

40
41
#include <trace/events/kmem.h>

42
43
#include "internal.h"

Christoph Lameter's avatar
Christoph Lameter committed
44
45
/*
 * Lock order:
46
 *   1. slab_mutex (Global Mutex)
47
48
 *   2. node->list_lock
 *   3. slab_lock(page) (Only on some arches and for debugging)
Christoph Lameter's avatar
Christoph Lameter committed
49
 *
50
 *   slab_mutex
51
 *
52
 *   The role of the slab_mutex is to protect the list of all the slabs
53
54
55
56
57
58
59
60
61
62
63
64
65
66
 *   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
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
 *
 *   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
87
88
 * 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
89
 * freed then the slab will show up again on the partial lists.
Christoph Lameter's avatar
Christoph Lameter committed
90
91
 * We track full slabs for debugging purposes though because otherwise we
 * cannot scan all objects.
Christoph Lameter's avatar
Christoph Lameter committed
92
93
94
95
96
97
98
 *
 * 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.
 *
99
100
101
102
103
104
105
106
107
108
109
110
 * 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
111
 * 			freelist that allows lockless access to
112
113
 * 			free objects in addition to the regular freelist
 * 			that requires the slab lock.
Christoph Lameter's avatar
Christoph Lameter committed
114
115
116
 *
 * PageError		Slab requires special handling due to debug
 * 			options set. This moves	slab handling out of
117
 * 			the fast path and disables lockless freelists.
Christoph Lameter's avatar
Christoph Lameter committed
118
119
 */

120
121
static inline int kmem_cache_debug(struct kmem_cache *s)
{
122
#ifdef CONFIG_SLUB_DEBUG
123
	return unlikely(s->flags & SLAB_DEBUG_FLAGS);
124
#else
125
	return 0;
126
#endif
127
}
128

129
void *fixup_red_left(struct kmem_cache *s, void *p)
Joonsoo Kim's avatar
Joonsoo Kim committed
130
131
132
133
134
135
136
{
	if (kmem_cache_debug(s) && s->flags & SLAB_RED_ZONE)
		p += s->red_left_pad;

	return p;
}

137
138
139
140
141
142
143
144
145
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
146
147
148
149
150
151
152
153
154
155
156
/*
 * 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

157
158
159
/* Enable to log cmpxchg failures */
#undef SLUB_DEBUG_CMPXCHG

160
161
162
163
/*
 * Mininum number of partial slabs. These will be left on the partial
 * lists even if they are empty. kmem_cache_shrink may reclaim them.
 */
164
#define MIN_PARTIAL 5
Christoph Lameter's avatar
Christoph Lameter committed
165

166
167
168
/*
 * Maximum number of desirable partial slabs.
 * The existence of more partial slabs makes kmem_cache_shrink
169
 * sort the partial list by the number of objects in use.
170
171
172
 */
#define MAX_PARTIAL 10

173
#define DEBUG_DEFAULT_FLAGS (SLAB_CONSISTENCY_CHECKS | SLAB_RED_ZONE | \
Christoph Lameter's avatar
Christoph Lameter committed
174
				SLAB_POISON | SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
175

176
177
178
179
180
181
182
183
/*
 * 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)


184
/*
185
186
187
 * 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.
188
 */
189
#define DEBUG_METADATA_FLAGS (SLAB_RED_ZONE | SLAB_POISON | SLAB_STORE_USER)
190

191
192
#define OO_SHIFT	16
#define OO_MASK		((1 << OO_SHIFT) - 1)
193
#define MAX_OBJS_PER_PAGE	32767 /* since page.objects is u15 */
194

Christoph Lameter's avatar
Christoph Lameter committed
195
/* Internal SLUB flags */
196
197
198
199
/* Poison object */
#define __OBJECT_POISON		((slab_flags_t __force)0x80000000UL)
/* Use cmpxchg_double */
#define __CMPXCHG_DOUBLE	((slab_flags_t __force)0x40000000UL)
Christoph Lameter's avatar
Christoph Lameter committed
200

201
202
203
/*
 * Tracking user of a slab.
 */
204
#define TRACK_ADDRS_COUNT 16
205
struct track {
206
	unsigned long addr;	/* Called from address */
207
208
209
#ifdef CONFIG_STACKTRACE
	unsigned long addrs[TRACK_ADDRS_COUNT];	/* Called from address */
#endif
210
211
212
213
214
215
216
	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 };

217
#ifdef CONFIG_SYSFS
Christoph Lameter's avatar
Christoph Lameter committed
218
219
static int sysfs_slab_add(struct kmem_cache *);
static int sysfs_slab_alias(struct kmem_cache *, const char *);
220
static void memcg_propagate_slab_attrs(struct kmem_cache *s);
221
static void sysfs_slab_remove(struct kmem_cache *s);
Christoph Lameter's avatar
Christoph Lameter committed
222
#else
223
224
225
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; }
226
static inline void memcg_propagate_slab_attrs(struct kmem_cache *s) { }
227
static inline void sysfs_slab_remove(struct kmem_cache *s) { }
Christoph Lameter's avatar
Christoph Lameter committed
228
229
#endif

230
static inline void stat(const struct kmem_cache *s, enum stat_item si)
231
232
{
#ifdef CONFIG_SLUB_STATS
233
234
235
236
237
	/*
	 * 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]);
238
239
240
#endif
}

Christoph Lameter's avatar
Christoph Lameter committed
241
242
243
244
/********************************************************************
 * 			Core slab cache functions
 *******************************************************************/

245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
/*
 * 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);
}

268
269
static inline void *get_freepointer(struct kmem_cache *s, void *object)
{
270
	return freelist_dereference(s, object + s->offset);
271
272
}

273
274
static void prefetch_freepointer(const struct kmem_cache *s, void *object)
{
275
276
	if (object)
		prefetch(freelist_dereference(s, object + s->offset));
277
278
}

279
280
static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
{
281
	unsigned long freepointer_addr;
282
283
	void *p;

284
285
286
	if (!debug_pagealloc_enabled())
		return get_freepointer(s, object);

287
288
289
	freepointer_addr = (unsigned long)object + s->offset;
	probe_kernel_read(&p, (void **)freepointer_addr, sizeof(p));
	return freelist_ptr(s, p, freepointer_addr);
290
291
}

292
293
static inline void set_freepointer(struct kmem_cache *s, void *object, void *fp)
{
294
295
	unsigned long freeptr_addr = (unsigned long)object + s->offset;

296
297
298
299
#ifdef CONFIG_SLAB_FREELIST_HARDENED
	BUG_ON(object == fp); /* naive detection of double free or corruption */
#endif

300
	*(void **)freeptr_addr = freelist_ptr(s, fp, freeptr_addr);
301
302
303
}

/* Loop over all objects in a slab */
304
#define for_each_object(__p, __s, __addr, __objects) \
Joonsoo Kim's avatar
Joonsoo Kim committed
305
306
307
	for (__p = fixup_red_left(__s, __addr); \
		__p < (__addr) + (__objects) * (__s)->size; \
		__p += (__s)->size)
308

309
#define for_each_object_idx(__p, __idx, __s, __addr, __objects) \
Joonsoo Kim's avatar
Joonsoo Kim committed
310
311
312
	for (__p = fixup_red_left(__s, __addr), __idx = 1; \
		__idx <= __objects; \
		__p += (__s)->size, __idx++)
313

314
315
316
317
318
319
/* 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;
}

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

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

	return x;
}

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

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

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

static __always_inline void slab_unlock(struct page *page)
{
356
	VM_BUG_ON_PAGE(PageTail(page), page);
357
358
359
	__bit_spin_unlock(PG_locked, &page->flags);
}

360
361
362
363
364
365
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
366
367
	 * as page->_refcount.  If we assign to ->counters directly
	 * we run the risk of losing updates to page->_refcount, so
368
369
370
371
372
373
374
	 * be careful and only assign to the fields we need.
	 */
	page->frozen  = tmp.frozen;
	page->inuse   = tmp.inuse;
	page->objects = tmp.objects;
}

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

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
407
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
408
409
#endif

410
	return false;
411
412
}

413
414
415
416
417
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)
{
418
419
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
420
	if (s->flags & __CMPXCHG_DOUBLE) {
421
		if (cmpxchg_double(&page->freelist, &page->counters,
422
423
				   freelist_old, counters_old,
				   freelist_new, counters_new))
424
			return true;
425
426
427
	} else
#endif
	{
428
429
430
		unsigned long flags;

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

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
448
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
449
450
#endif

451
	return false;
452
453
}

454
#ifdef CONFIG_SLUB_DEBUG
455
456
457
/*
 * Determine a map of object in use on a page.
 *
458
 * Node listlock must be held to guarantee that the page does
459
460
461
462
463
464
465
466
467
468
469
 * 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
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
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;
}

486
487
488
/*
 * Debug settings:
 */
489
#if defined(CONFIG_SLUB_DEBUG_ON)
490
static slab_flags_t slub_debug = DEBUG_DEFAULT_FLAGS;
491
#else
492
static slab_flags_t slub_debug;
493
#endif
494
495

static char *slub_debug_slabs;
496
static int disable_higher_order_debug;
497

498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
/*
 * 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
514
515
516
/*
 * Object debugging
 */
Joonsoo Kim's avatar
Joonsoo Kim committed
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536

/* 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;
}

537
538
static void print_section(char *level, char *text, u8 *addr,
			  unsigned int length)
Christoph Lameter's avatar
Christoph Lameter committed
539
{
540
	metadata_access_enable();
541
	print_hex_dump(level, text, DUMP_PREFIX_ADDRESS, 16, 1, addr,
542
			length, 1);
543
	metadata_access_disable();
Christoph Lameter's avatar
Christoph Lameter committed
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
}

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,
560
			enum track_item alloc, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
561
{
Akinobu Mita's avatar
Akinobu Mita committed
562
	struct track *p = get_track(s, object, alloc);
Christoph Lameter's avatar
Christoph Lameter committed
563
564

	if (addr) {
565
566
567
568
569
570
571
572
#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;
573
		metadata_access_enable();
574
		save_stack_trace(&trace);
575
		metadata_access_disable();
576
577
578
579
580
581
582
583
584

		/* 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
585
586
		p->addr = addr;
		p->cpu = smp_processor_id();
587
		p->pid = current->pid;
Christoph Lameter's avatar
Christoph Lameter committed
588
589
590
591
592
593
594
		p->when = jiffies;
	} else
		memset(p, 0, sizeof(struct track));
}

static void init_tracking(struct kmem_cache *s, void *object)
{
595
596
597
	if (!(s->flags & SLAB_STORE_USER))
		return;

598
599
	set_track(s, object, TRACK_FREE, 0UL);
	set_track(s, object, TRACK_ALLOC, 0UL);
Christoph Lameter's avatar
Christoph Lameter committed
600
601
602
603
604
605
606
}

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

607
608
	pr_err("INFO: %s in %pS age=%lu cpu=%u pid=%d\n",
	       s, (void *)t->addr, jiffies - t->when, t->cpu, t->pid);
609
610
611
612
613
#ifdef CONFIG_STACKTRACE
	{
		int i;
		for (i = 0; i < TRACK_ADDRS_COUNT; i++)
			if (t->addrs[i])
614
				pr_err("\t%pS\n", (void *)t->addrs[i]);
615
616
617
618
			else
				break;
	}
#endif
619
620
621
622
623
624
625
626
627
628
629
630
631
}

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)
{
632
	pr_err("INFO: Slab 0x%p objects=%u used=%u fp=0x%p flags=0x%04lx\n",
633
	       page, page->objects, page->inuse, page->freelist, page->flags);
634
635
636
637
638

}

static void slab_bug(struct kmem_cache *s, char *fmt, ...)
{
639
	struct va_format vaf;
640
641
642
	va_list args;

	va_start(args, fmt);
643
644
	vaf.fmt = fmt;
	vaf.va = &args;
645
	pr_err("=============================================================================\n");
646
	pr_err("BUG %s (%s): %pV\n", s->name, print_tainted(), &vaf);
647
	pr_err("-----------------------------------------------------------------------------\n\n");
648

649
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
650
	va_end(args);
Christoph Lameter's avatar
Christoph Lameter committed
651
652
}

653
654
static void slab_fix(struct kmem_cache *s, char *fmt, ...)
{
655
	struct va_format vaf;
656
657
658
	va_list args;

	va_start(args, fmt);
659
660
661
	vaf.fmt = fmt;
	vaf.va = &args;
	pr_err("FIX %s: %pV\n", s->name, &vaf);
662
663
664
665
	va_end(args);
}

static void print_trailer(struct kmem_cache *s, struct page *page, u8 *p)
Christoph Lameter's avatar
Christoph Lameter committed
666
667
{
	unsigned int off;	/* Offset of last byte */
668
	u8 *addr = page_address(page);
669
670
671
672
673

	print_tracking(s, p);

	print_page_info(page);

674
675
	pr_err("INFO: Object 0x%p @offset=%tu fp=0x%p\n\n",
	       p, p - addr, get_freepointer(s, p));
676

Joonsoo Kim's avatar
Joonsoo Kim committed
677
	if (s->flags & SLAB_RED_ZONE)
678
679
		print_section(KERN_ERR, "Redzone ", p - s->red_left_pad,
			      s->red_left_pad);
Joonsoo Kim's avatar
Joonsoo Kim committed
680
	else if (p > addr + 16)
681
		print_section(KERN_ERR, "Bytes b4 ", p - 16, 16);
Christoph Lameter's avatar
Christoph Lameter committed
682

683
684
	print_section(KERN_ERR, "Object ", p,
		      min_t(unsigned long, s->object_size, PAGE_SIZE));
Christoph Lameter's avatar
Christoph Lameter committed
685
	if (s->flags & SLAB_RED_ZONE)
686
		print_section(KERN_ERR, "Redzone ", p + s->object_size,
687
			s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
688
689
690
691
692
693

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

694
	if (s->flags & SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
695
696
		off += 2 * sizeof(struct track);

697
698
	off += kasan_metadata_size(s);

Joonsoo Kim's avatar
Joonsoo Kim committed
699
	if (off != size_from_object(s))
Christoph Lameter's avatar
Christoph Lameter committed
700
		/* Beginning of the filler is the free pointer */
701
702
		print_section(KERN_ERR, "Padding ", p + off,
			      size_from_object(s) - off);
703
704

	dump_stack();
Christoph Lameter's avatar
Christoph Lameter committed
705
706
}

707
void object_err(struct kmem_cache *s, struct page *page,
Christoph Lameter's avatar
Christoph Lameter committed
708
709
			u8 *object, char *reason)
{
710
	slab_bug(s, "%s", reason);
711
	print_trailer(s, page, object);
Christoph Lameter's avatar
Christoph Lameter committed
712
713
}

714
715
static void slab_err(struct kmem_cache *s, struct page *page,
			const char *fmt, ...)
Christoph Lameter's avatar
Christoph Lameter committed
716
717
718
719
{
	va_list args;
	char buf[100];

720
721
	va_start(args, fmt);
	vsnprintf(buf, sizeof(buf), fmt, args);
Christoph Lameter's avatar
Christoph Lameter committed
722
	va_end(args);
723
	slab_bug(s, "%s", buf);
724
	print_page_info(page);
Christoph Lameter's avatar
Christoph Lameter committed
725
726
727
	dump_stack();
}

728
static void init_object(struct kmem_cache *s, void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
729
730
731
{
	u8 *p = object;

Joonsoo Kim's avatar
Joonsoo Kim committed
732
733
734
	if (s->flags & SLAB_RED_ZONE)
		memset(p - s->red_left_pad, val, s->red_left_pad);

Christoph Lameter's avatar
Christoph Lameter committed
735
	if (s->flags & __OBJECT_POISON) {
736
737
		memset(p, POISON_FREE, s->object_size - 1);
		p[s->object_size - 1] = POISON_END;
Christoph Lameter's avatar
Christoph Lameter committed
738
739
740
	}

	if (s->flags & SLAB_RED_ZONE)
741
		memset(p + s->object_size, val, s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
742
743
}

744
745
746
747
748
749
750
751
752
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,
753
			u8 *start, unsigned int value, unsigned int bytes)
754
755
756
757
{
	u8 *fault;
	u8 *end;

758
	metadata_access_enable();
759
	fault = memchr_inv(start, value, bytes);
760
	metadata_access_disable();
761
762
763
764
765
766
767
768
	if (!fault)
		return 1;

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

	slab_bug(s, "%s overwritten", what);
769
	pr_err("INFO: 0x%p-0x%p. First byte 0x%x instead of 0x%x\n",
770
771
772
773
774
					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
775
776
777
778
779
780
781
782
783
}

/*
 * 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
784
 *
Christoph Lameter's avatar
Christoph Lameter committed
785
786
787
 * 	Poisoning uses 0x6b (POISON_FREE) and the last byte is
 * 	0xa5 (POISON_END)
 *
788
 * object + s->object_size
Christoph Lameter's avatar
Christoph Lameter committed
789
 * 	Padding to reach word boundary. This is also used for Redzoning.
Christoph Lameter's avatar
Christoph Lameter committed
790
 * 	Padding is extended by another word if Redzoning is enabled and
791
 * 	object_size == inuse.
Christoph Lameter's avatar
Christoph Lameter committed
792
 *
Christoph Lameter's avatar
Christoph Lameter committed
793
794
795
796
 * 	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
797
798
 * 	Meta data starts here.
 *
Christoph Lameter's avatar
Christoph Lameter committed
799
800
 * 	A. Free pointer (if we cannot overwrite object on free)
 * 	B. Tracking data for SLAB_STORE_USER
Christoph Lameter's avatar
Christoph Lameter committed
801
 * 	C. Padding to reach required alignment boundary or at mininum
Christoph Lameter's avatar
Christoph Lameter committed
802
 * 		one word if debugging is on to be able to detect writes
Christoph Lameter's avatar
Christoph Lameter committed
803
804
805
 * 		before the word boundary.
 *
 *	Padding is done using 0x5a (POISON_INUSE)
Christoph Lameter's avatar
Christoph Lameter committed
806
807
 *
 * object + s->size
Christoph Lameter's avatar
Christoph Lameter committed
808
 * 	Nothing is used beyond s->size.
Christoph Lameter's avatar
Christoph Lameter committed
809
 *
810
 * If slabcaches are merged then the object_size and inuse boundaries are mostly
Christoph Lameter's avatar
Christoph Lameter committed
811
 * ignored. And therefore no slab options that rely on these boundaries
Christoph Lameter's avatar
Christoph Lameter committed
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
 * 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);

827
828
	off += kasan_metadata_size(s);

Joonsoo Kim's avatar
Joonsoo Kim committed
829
	if (size_from_object(s) == off)
Christoph Lameter's avatar
Christoph Lameter committed
830
831
		return 1;

832
	return check_bytes_and_report(s, page, p, "Object padding",
Joonsoo Kim's avatar
Joonsoo Kim committed
833
			p + off, POISON_INUSE, size_from_object(s) - off);
Christoph Lameter's avatar
Christoph Lameter committed
834
835
}

836
/* Check the pad bytes at the end of a slab page */
Christoph Lameter's avatar
Christoph Lameter committed
837
838
static int slab_pad_check(struct kmem_cache *s, struct page *page)
{
839
840
841
842
843
	u8 *start;
	u8 *fault;
	u8 *end;
	int length;
	int remainder;
Christoph Lameter's avatar
Christoph Lameter committed
844
845
846
847

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

848
	start = page_address(page);
849
	length = (PAGE_SIZE << compound_order(page)) - s->reserved;
850
851
	end = start + length;
	remainder = length % s->size;
Christoph Lameter's avatar
Christoph Lameter committed
852
853
854
	if (!remainder)
		return 1;

855
	metadata_access_enable();
856
	fault = memchr_inv(end - remainder, POISON_INUSE, remainder);
857
	metadata_access_disable();
858
859
860
861
862
863
	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);
864
	print_section(KERN_ERR, "Padding ", end - remainder, remainder);
865

Eric Dumazet's avatar
Eric Dumazet committed
866
	restore_bytes(s, "slab padding", POISON_INUSE, end - remainder, end);
867
	return 0;
Christoph Lameter's avatar
Christoph Lameter committed
868
869
870
}

static int check_object(struct kmem_cache *s, struct page *page,
871
					void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
872
873
{
	u8 *p = object;
874
	u8 *endobject = object + s->object_size;
Christoph Lameter's avatar
Christoph Lameter committed
875
876

	if (s->flags & SLAB_RED_ZONE) {
Joonsoo Kim's avatar
Joonsoo Kim committed
877
878
879
880
		if (!check_bytes_and_report(s, page, object, "Redzone",
			object - s->red_left_pad, val, s->red_left_pad))
			return 0;

881
		if (!check_bytes_and_report(s, page, object, "Redzone",
882
			endobject, val, s->inuse - s->object_size))
Christoph Lameter's avatar
Christoph Lameter committed
883
884
			return 0;
	} else {
885
		if ((s->flags & SLAB_POISON) && s->object_size < s->inuse) {
Ingo Molnar's avatar
Ingo Molnar committed
886
			check_bytes_and_report(s, page, p, "Alignment padding",
887
888
				endobject, POISON_INUSE,
				s->inuse - s->object_size);
Ingo Molnar's avatar
Ingo Molnar committed
889
		}
Christoph Lameter's avatar
Christoph Lameter committed
890
891
892
	}

	if (s->flags & SLAB_POISON) {
893
		if (val != SLUB_RED_ACTIVE && (s->flags & __OBJECT_POISON) &&
894
			(!check_bytes_and_report(s, page, p, "Poison", p,
895
					POISON_FREE, s->object_size - 1) ||
896
			 !check_bytes_and_report(s, page, p, "Poison",
897
				p + s->object_size - 1, POISON_END, 1)))
Christoph Lameter's avatar
Christoph Lameter committed
898
899
900
901
902
903
904
			return 0;
		/*
		 * check_pad_bytes cleans up on its own.
		 */
		check_pad_bytes(s, page, p);
	}

905
	if (!s->offset && val == SLUB_RED_ACTIVE)
Christoph Lameter's avatar
Christoph Lameter committed
906
907
908
909
910
911
912
913
914
915
		/*
		 * 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
916
		 * No choice but to zap it and thus lose the remainder
Christoph Lameter's avatar
Christoph Lameter committed
917
		 * of the free objects in this slab. May cause
Christoph Lameter's avatar
Christoph Lameter committed
918
		 * another error because the object count is now wrong.
Christoph Lameter's avatar
Christoph Lameter committed
919
		 */
920
		set_freepointer(s, p, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
921
922
923
924
925
926
927
		return 0;
	}
	return 1;
}

static int check_slab(struct kmem_cache *s, struct page *page)
{
928
929
	int maxobj;

Christoph Lameter's avatar
Christoph Lameter committed
930
931
932
	VM_BUG_ON(!irqs_disabled());

	if (!PageSlab(page)) {
933
		slab_err(s, page, "Not a valid slab page");
Christoph Lameter's avatar
Christoph Lameter committed
934
935
		return 0;
	}
936

937
	maxobj = order_objects(compound_order(page), s->size, s->reserved);
938
939
	if (page->objects > maxobj) {
		slab_err(s, page, "objects %u > max %u",
940
			page->objects, maxobj);
941
942
943
		return 0;
	}
	if (page->inuse > page->objects) {
944
		slab_err(s, page, "inuse %u > max %u",
945
			page->inuse, page->objects);
Christoph Lameter's avatar
Christoph Lameter committed
946
947
948
949
950
951
952
953
		return 0;
	}
	/* Slab_pad_check fixes things up after itself */
	slab_pad_check(s, page);
	return 1;
}

/*
Christoph Lameter's avatar
Christoph Lameter committed
954
955
 * 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
956
957
958
959
 */
static int on_freelist(struct kmem_cache *s, struct page *page, void *search)
{
	int nr = 0;
960
	void *fp;
Christoph Lameter's avatar
Christoph Lameter committed
961
	void *object = NULL;
962
	int max_objects;
Christoph Lameter's avatar
Christoph Lameter committed
963

964
	fp = page->freelist;
965
	while (fp && nr <= page->objects) {
Christoph Lameter's avatar
Christoph Lameter committed
966
967
968
969
970
971
		if (fp == search)
			return 1;
		if (!check_valid_pointer(s, page, fp)) {
			if (object) {
				object_err(s, page, object,
					"Freechain corrupt");
972
				set_freepointer(s, object, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
973
			} else {
974
				slab_err(s, page, "Freepointer corrupt");
975
				page->freelist = NULL;
976
				page->inuse = page->objects;
977
				slab_fix(s, "Freelist cleared");
Christoph Lameter's avatar
Christoph Lameter committed
978
979
980
981
982
983
984
985
986
				return 0;
			}
			break;
		}
		object = fp;
		fp = get_freepointer(s, object);
		nr++;
	}

987
	max_objects = order_objects(compound_order(page), s->size, s->reserved);
988
989
	if (max_objects > MAX_OBJS_PER_PAGE)
		max_objects = MAX_OBJS_PER_PAGE;
990
991

	if (page->objects != max_objects) {
Joe Perches's avatar
Joe Perches committed
992
993
		slab_err(s, page, "Wrong number of objects. Found %d but should be %d",
			 page->objects, max_objects);
994
995
996
		page->objects = max_objects;
		slab_fix(s, "Number of objects adjusted.");
	}
997
	if (page->inuse != page->objects - nr) {
Joe Perches's avatar
Joe Perches committed
998
999
		slab_err(s, page, "Wrong object count. Counter is %d but counted were %d",
			 page->inuse, page->objects - nr);
1000
		page->inuse = page->objects - nr;
1001
		slab_fix(s, "Object count adjusted.");
Christoph Lameter's avatar
Christoph Lameter committed
1002
1003
1004
1005
	}
	return search == NULL;
}

1006
1007
static void trace(struct kmem_cache *s, struct page *page, void *object,
								int alloc)
Christoph Lameter's avatar
Christoph Lameter committed
1008
1009
{
	if (s->flags & SLAB_TRACE) {
1010
		pr_info("TRACE %s %s 0x%p inuse=%d fp=0x%p\n",
Christoph Lameter's avatar
Christoph Lameter committed
1011
1012
1013
1014
1015
1016
			s->name,
			alloc ? "alloc" : "free",
			object, page->inuse,
			page->freelist);

		if (!alloc)
1017
			print_section(KERN_INFO, "Object ", (void *)object,
1018
					s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
1019
1020
1021
1022
1023

		dump_stack();
	}
}

1024
/*
Christoph Lameter's avatar
Christoph Lameter committed
1025
 * Tracking of fully allocated slabs for debugging purposes.
1026
 */
1027
1028
static void add_full(struct kmem_cache *s,
	struct kmem_cache_node *n, struct page *page)
1029
{
1030
1031
1032
	if (!(s->flags & SLAB_STORE_USER))
		return;

1033
	lockdep_assert_held(&n->list_lock);
1034
1035
1036
	list_add(&page->lru, &n->full);
}

Peter Zijlstra's avatar
Peter Zijlstra committed
1037
static void remove_full(struct kmem_cache *s, struct kmem_cache_node *n, struct page *page)
1038
1039
1040
1041
{
	if (!(s->flags & SLAB_STORE_USER))
		return;

1042
	lockdep_assert_held(&n->list_lock);
1043
1044
1045
	list_del(&page->lru);
}

1046
1047
1048
1049
1050
1051
1052
1053
/* 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);
}

1054
1055
1056
1057
1058
static inline unsigned long node_nr_slabs(struct kmem_cache_node *n)
{
	return atomic_long_read(&n->nr_slabs);
}

1059
static inline void inc_slabs_node(struct kmem_cache *s, int node, int objects)
1060
1061
1062
1063
1064
1065
1066
1067
1068
{
	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).
	 */
1069
	if (likely(n)) {
1070
		atomic_long_inc(&n->nr_slabs);
1071
1072
		atomic_long_add(objects, &n->total_objects);
	}
1073
}
1074
static inline void dec_slabs_node(struct kmem_cache *s, int node, int objects)
1075
1076
1077
1078
{
	struct kmem_cache_node *n = get_node(s, node);

	atomic_long_dec(&n->nr_slabs);
1079
	atomic_long_sub(objects, &n->total_objects);
1080
1081
1082
}

/* Object debug checks for alloc/free paths */