slub.c 138 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>
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
54
55
56
57
58
59
60
61
62
63
64
 *   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
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
 *
 *   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
85
86
 * 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
87
 * freed then the slab will show up again on the partial lists.
Christoph Lameter's avatar
Christoph Lameter committed
88
89
 * We track full slabs for debugging purposes though because otherwise we
 * cannot scan all objects.
Christoph Lameter's avatar
Christoph Lameter committed
90
91
92
93
94
95
96
 *
 * 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.
 *
97
98
99
100
101
102
103
104
105
106
107
108
 * 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
109
 * 			freelist that allows lockless access to
110
111
 * 			free objects in addition to the regular freelist
 * 			that requires the slab lock.
Christoph Lameter's avatar
Christoph Lameter committed
112
113
114
 *
 * PageError		Slab requires special handling due to debug
 * 			options set. This moves	slab handling out of
115
 * 			the fast path and disables lockless freelists.
Christoph Lameter's avatar
Christoph Lameter committed
116
117
 */

118
119
static inline int kmem_cache_debug(struct kmem_cache *s)
{
120
#ifdef CONFIG_SLUB_DEBUG
121
	return unlikely(s->flags & SLAB_DEBUG_FLAGS);
122
#else
123
	return 0;
124
#endif
125
}
126

127
void *fixup_red_left(struct kmem_cache *s, void *p)
Joonsoo Kim's avatar
Joonsoo Kim committed
128
129
130
131
132
133
134
{
	if (kmem_cache_debug(s) && s->flags & SLAB_RED_ZONE)
		p += s->red_left_pad;

	return p;
}

135
136
137
138
139
140
141
142
143
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
144
145
146
147
148
149
150
151
152
153
154
/*
 * 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

155
156
157
/* Enable to log cmpxchg failures */
#undef SLUB_DEBUG_CMPXCHG

158
159
160
161
/*
 * Mininum number of partial slabs. These will be left on the partial
 * lists even if they are empty. kmem_cache_shrink may reclaim them.
 */
162
#define MIN_PARTIAL 5
Christoph Lameter's avatar
Christoph Lameter committed
163

164
165
166
/*
 * Maximum number of desirable partial slabs.
 * The existence of more partial slabs makes kmem_cache_shrink
167
 * sort the partial list by the number of objects in use.
168
169
170
 */
#define MAX_PARTIAL 10

171
#define DEBUG_DEFAULT_FLAGS (SLAB_CONSISTENCY_CHECKS | SLAB_RED_ZONE | \
Christoph Lameter's avatar
Christoph Lameter committed
172
				SLAB_POISON | SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
173

174
175
176
177
178
179
180
181
/*
 * 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)


182
/*
183
184
185
 * 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.
186
 */
187
#define DEBUG_METADATA_FLAGS (SLAB_RED_ZONE | SLAB_POISON | SLAB_STORE_USER)
188

189
190
#define OO_SHIFT	16
#define OO_MASK		((1 << OO_SHIFT) - 1)
191
#define MAX_OBJS_PER_PAGE	32767 /* since page.objects is u15 */
192

Christoph Lameter's avatar
Christoph Lameter committed
193
/* Internal SLUB flags */
Christoph Lameter's avatar
Christoph Lameter committed
194
#define __OBJECT_POISON		0x80000000UL /* Poison object */
195
#define __CMPXCHG_DOUBLE	0x40000000UL /* Use cmpxchg_double */
Christoph Lameter's avatar
Christoph Lameter committed
196

197
198
199
/*
 * Tracking user of a slab.
 */
200
#define TRACK_ADDRS_COUNT 16
201
struct track {
202
	unsigned long addr;	/* Called from address */
203
204
205
#ifdef CONFIG_STACKTRACE
	unsigned long addrs[TRACK_ADDRS_COUNT];	/* Called from address */
#endif
206
207
208
209
210
211
212
	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 };

213
#ifdef CONFIG_SYSFS
Christoph Lameter's avatar
Christoph Lameter committed
214
215
static int sysfs_slab_add(struct kmem_cache *);
static int sysfs_slab_alias(struct kmem_cache *, const char *);
216
static void memcg_propagate_slab_attrs(struct kmem_cache *s);
Christoph Lameter's avatar
Christoph Lameter committed
217
#else
218
219
220
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; }
221
static inline void memcg_propagate_slab_attrs(struct kmem_cache *s) { }
Christoph Lameter's avatar
Christoph Lameter committed
222
223
#endif

224
static inline void stat(const struct kmem_cache *s, enum stat_item si)
225
226
{
#ifdef CONFIG_SLUB_STATS
227
228
229
230
231
	/*
	 * 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]);
232
233
234
#endif
}

Christoph Lameter's avatar
Christoph Lameter committed
235
236
237
238
/********************************************************************
 * 			Core slab cache functions
 *******************************************************************/

239
240
241
242
243
static inline void *get_freepointer(struct kmem_cache *s, void *object)
{
	return *(void **)(object + s->offset);
}

244
245
246
247
248
static void prefetch_freepointer(const struct kmem_cache *s, void *object)
{
	prefetch(object + s->offset);
}

249
250
251
252
static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
{
	void *p;

253
254
255
	if (!debug_pagealloc_enabled())
		return get_freepointer(s, object);

256
257
258
259
	probe_kernel_read(&p, (void **)(object + s->offset), sizeof(p));
	return p;
}

260
261
262
263
264
265
static inline void set_freepointer(struct kmem_cache *s, void *object, void *fp)
{
	*(void **)(object + s->offset) = fp;
}

/* Loop over all objects in a slab */
266
#define for_each_object(__p, __s, __addr, __objects) \
Joonsoo Kim's avatar
Joonsoo Kim committed
267
268
269
	for (__p = fixup_red_left(__s, __addr); \
		__p < (__addr) + (__objects) * (__s)->size; \
		__p += (__s)->size)
270

271
#define for_each_object_idx(__p, __idx, __s, __addr, __objects) \
Joonsoo Kim's avatar
Joonsoo Kim committed
272
273
274
	for (__p = fixup_red_left(__s, __addr), __idx = 1; \
		__idx <= __objects; \
		__p += (__s)->size, __idx++)
275

276
277
278
279
280
281
/* 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;
}

282
283
284
285
286
static inline int order_objects(int order, unsigned long size, int reserved)
{
	return ((PAGE_SIZE << order) - reserved) / size;
}

287
static inline struct kmem_cache_order_objects oo_make(int order,
288
		unsigned long size, int reserved)
289
290
{
	struct kmem_cache_order_objects x = {
291
		(order << OO_SHIFT) + order_objects(order, size, reserved)
292
293
294
295
296
297
298
	};

	return x;
}

static inline int oo_order(struct kmem_cache_order_objects x)
{
299
	return x.x >> OO_SHIFT;
300
301
302
303
}

static inline int oo_objects(struct kmem_cache_order_objects x)
{
304
	return x.x & OO_MASK;
305
306
}

307
308
309
310
311
/*
 * Per slab locking using the pagelock
 */
static __always_inline void slab_lock(struct page *page)
{
312
	VM_BUG_ON_PAGE(PageTail(page), page);
313
314
315
316
317
	bit_spin_lock(PG_locked, &page->flags);
}

static __always_inline void slab_unlock(struct page *page)
{
318
	VM_BUG_ON_PAGE(PageTail(page), page);
319
320
321
	__bit_spin_unlock(PG_locked, &page->flags);
}

322
323
324
325
326
327
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
328
329
	 * as page->_refcount.  If we assign to ->counters directly
	 * we run the risk of losing updates to page->_refcount, so
330
331
332
333
334
335
336
	 * be careful and only assign to the fields we need.
	 */
	page->frozen  = tmp.frozen;
	page->inuse   = tmp.inuse;
	page->objects = tmp.objects;
}

337
338
339
340
341
342
343
/* 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());
344
345
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
346
	if (s->flags & __CMPXCHG_DOUBLE) {
347
		if (cmpxchg_double(&page->freelist, &page->counters,
348
349
				   freelist_old, counters_old,
				   freelist_new, counters_new))
350
			return true;
351
352
353
354
	} else
#endif
	{
		slab_lock(page);
355
356
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
357
			page->freelist = freelist_new;
358
			set_page_slub_counters(page, counters_new);
359
			slab_unlock(page);
360
			return true;
361
362
363
364
365
366
367
368
		}
		slab_unlock(page);
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
369
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
370
371
#endif

372
	return false;
373
374
}

375
376
377
378
379
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)
{
380
381
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
382
	if (s->flags & __CMPXCHG_DOUBLE) {
383
		if (cmpxchg_double(&page->freelist, &page->counters,
384
385
				   freelist_old, counters_old,
				   freelist_new, counters_new))
386
			return true;
387
388
389
	} else
#endif
	{
390
391
392
		unsigned long flags;

		local_irq_save(flags);
393
		slab_lock(page);
394
395
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
396
			page->freelist = freelist_new;
397
			set_page_slub_counters(page, counters_new);
398
			slab_unlock(page);
399
			local_irq_restore(flags);
400
			return true;
401
		}
402
		slab_unlock(page);
403
		local_irq_restore(flags);
404
405
406
407
408
409
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
410
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
411
412
#endif

413
	return false;
414
415
}

416
#ifdef CONFIG_SLUB_DEBUG
417
418
419
/*
 * Determine a map of object in use on a page.
 *
420
 * Node listlock must be held to guarantee that the page does
421
422
423
424
425
426
427
428
429
430
431
 * 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
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
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;
}

448
449
450
/*
 * Debug settings:
 */
451
#if defined(CONFIG_SLUB_DEBUG_ON)
452
453
static int slub_debug = DEBUG_DEFAULT_FLAGS;
#else
454
static int slub_debug;
455
#endif
456
457

static char *slub_debug_slabs;
458
static int disable_higher_order_debug;
459

460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
/*
 * 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
476
477
478
/*
 * Object debugging
 */
Joonsoo Kim's avatar
Joonsoo Kim committed
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498

/* 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;
}

Christoph Lameter's avatar
Christoph Lameter committed
499
500
static void print_section(char *text, u8 *addr, unsigned int length)
{
501
	metadata_access_enable();
502
503
	print_hex_dump(KERN_ERR, text, DUMP_PREFIX_ADDRESS, 16, 1, addr,
			length, 1);
504
	metadata_access_disable();
Christoph Lameter's avatar
Christoph Lameter committed
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
}

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,
521
			enum track_item alloc, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
522
{
Akinobu Mita's avatar
Akinobu Mita committed
523
	struct track *p = get_track(s, object, alloc);
Christoph Lameter's avatar
Christoph Lameter committed
524
525

	if (addr) {
526
527
528
529
530
531
532
533
#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;
534
		metadata_access_enable();
535
		save_stack_trace(&trace);
536
		metadata_access_disable();
537
538
539
540
541
542
543
544
545

		/* 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
546
547
		p->addr = addr;
		p->cpu = smp_processor_id();
548
		p->pid = current->pid;
Christoph Lameter's avatar
Christoph Lameter committed
549
550
551
552
553
554
555
		p->when = jiffies;
	} else
		memset(p, 0, sizeof(struct track));
}

static void init_tracking(struct kmem_cache *s, void *object)
{
556
557
558
	if (!(s->flags & SLAB_STORE_USER))
		return;

559
560
	set_track(s, object, TRACK_FREE, 0UL);
	set_track(s, object, TRACK_ALLOC, 0UL);
Christoph Lameter's avatar
Christoph Lameter committed
561
562
563
564
565
566
567
}

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

568
569
	pr_err("INFO: %s in %pS age=%lu cpu=%u pid=%d\n",
	       s, (void *)t->addr, jiffies - t->when, t->cpu, t->pid);
570
571
572
573
574
#ifdef CONFIG_STACKTRACE
	{
		int i;
		for (i = 0; i < TRACK_ADDRS_COUNT; i++)
			if (t->addrs[i])
575
				pr_err("\t%pS\n", (void *)t->addrs[i]);
576
577
578
579
			else
				break;
	}
#endif
580
581
582
583
584
585
586
587
588
589
590
591
592
}

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)
{
593
	pr_err("INFO: Slab 0x%p objects=%u used=%u fp=0x%p flags=0x%04lx\n",
594
	       page, page->objects, page->inuse, page->freelist, page->flags);
595
596
597
598
599

}

static void slab_bug(struct kmem_cache *s, char *fmt, ...)
{
600
	struct va_format vaf;
601
602
603
	va_list args;

	va_start(args, fmt);
604
605
	vaf.fmt = fmt;
	vaf.va = &args;
606
	pr_err("=============================================================================\n");
607
	pr_err("BUG %s (%s): %pV\n", s->name, print_tainted(), &vaf);
608
	pr_err("-----------------------------------------------------------------------------\n\n");
609

610
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
611
	va_end(args);
Christoph Lameter's avatar
Christoph Lameter committed
612
613
}

614
615
static void slab_fix(struct kmem_cache *s, char *fmt, ...)
{
616
	struct va_format vaf;
617
618
619
	va_list args;

	va_start(args, fmt);
620
621
622
	vaf.fmt = fmt;
	vaf.va = &args;
	pr_err("FIX %s: %pV\n", s->name, &vaf);
623
624
625
626
	va_end(args);
}

static void print_trailer(struct kmem_cache *s, struct page *page, u8 *p)
Christoph Lameter's avatar
Christoph Lameter committed
627
628
{
	unsigned int off;	/* Offset of last byte */
629
	u8 *addr = page_address(page);
630
631
632
633
634

	print_tracking(s, p);

	print_page_info(page);

635
636
	pr_err("INFO: Object 0x%p @offset=%tu fp=0x%p\n\n",
	       p, p - addr, get_freepointer(s, p));
637

Joonsoo Kim's avatar
Joonsoo Kim committed
638
639
640
	if (s->flags & SLAB_RED_ZONE)
		print_section("Redzone ", p - s->red_left_pad, s->red_left_pad);
	else if (p > addr + 16)
641
		print_section("Bytes b4 ", p - 16, 16);
Christoph Lameter's avatar
Christoph Lameter committed
642

643
	print_section("Object ", p, min_t(unsigned long, s->object_size,
644
				PAGE_SIZE));
Christoph Lameter's avatar
Christoph Lameter committed
645
	if (s->flags & SLAB_RED_ZONE)
646
647
		print_section("Redzone ", p + s->object_size,
			s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
648
649
650
651
652
653

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

654
	if (s->flags & SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
655
656
		off += 2 * sizeof(struct track);

657
658
	off += kasan_metadata_size(s);

Joonsoo Kim's avatar
Joonsoo Kim committed
659
	if (off != size_from_object(s))
Christoph Lameter's avatar
Christoph Lameter committed
660
		/* Beginning of the filler is the free pointer */
Joonsoo Kim's avatar
Joonsoo Kim committed
661
		print_section("Padding ", p + off, size_from_object(s) - off);
662
663

	dump_stack();
Christoph Lameter's avatar
Christoph Lameter committed
664
665
}

666
void object_err(struct kmem_cache *s, struct page *page,
Christoph Lameter's avatar
Christoph Lameter committed
667
668
			u8 *object, char *reason)
{
669
	slab_bug(s, "%s", reason);
670
	print_trailer(s, page, object);
Christoph Lameter's avatar
Christoph Lameter committed
671
672
}

673
674
static void slab_err(struct kmem_cache *s, struct page *page,
			const char *fmt, ...)
Christoph Lameter's avatar
Christoph Lameter committed
675
676
677
678
{
	va_list args;
	char buf[100];

679
680
	va_start(args, fmt);
	vsnprintf(buf, sizeof(buf), fmt, args);
Christoph Lameter's avatar
Christoph Lameter committed
681
	va_end(args);
682
	slab_bug(s, "%s", buf);
683
	print_page_info(page);
Christoph Lameter's avatar
Christoph Lameter committed
684
685
686
	dump_stack();
}

687
static void init_object(struct kmem_cache *s, void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
688
689
690
{
	u8 *p = object;

Joonsoo Kim's avatar
Joonsoo Kim committed
691
692
693
	if (s->flags & SLAB_RED_ZONE)
		memset(p - s->red_left_pad, val, s->red_left_pad);

Christoph Lameter's avatar
Christoph Lameter committed
694
	if (s->flags & __OBJECT_POISON) {
695
696
		memset(p, POISON_FREE, s->object_size - 1);
		p[s->object_size - 1] = POISON_END;
Christoph Lameter's avatar
Christoph Lameter committed
697
698
699
	}

	if (s->flags & SLAB_RED_ZONE)
700
		memset(p + s->object_size, val, s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
701
702
}

703
704
705
706
707
708
709
710
711
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,
712
			u8 *start, unsigned int value, unsigned int bytes)
713
714
715
716
{
	u8 *fault;
	u8 *end;

717
	metadata_access_enable();
718
	fault = memchr_inv(start, value, bytes);
719
	metadata_access_disable();
720
721
722
723
724
725
726
727
	if (!fault)
		return 1;

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

	slab_bug(s, "%s overwritten", what);
728
	pr_err("INFO: 0x%p-0x%p. First byte 0x%x instead of 0x%x\n",
729
730
731
732
733
					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
734
735
736
737
738
739
740
741
742
}

/*
 * 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
743
 *
Christoph Lameter's avatar
Christoph Lameter committed
744
745
746
 * 	Poisoning uses 0x6b (POISON_FREE) and the last byte is
 * 	0xa5 (POISON_END)
 *
747
 * object + s->object_size
Christoph Lameter's avatar
Christoph Lameter committed
748
 * 	Padding to reach word boundary. This is also used for Redzoning.
Christoph Lameter's avatar
Christoph Lameter committed
749
 * 	Padding is extended by another word if Redzoning is enabled and
750
 * 	object_size == inuse.
Christoph Lameter's avatar
Christoph Lameter committed
751
 *
Christoph Lameter's avatar
Christoph Lameter committed
752
753
754
755
 * 	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
756
757
 * 	Meta data starts here.
 *
Christoph Lameter's avatar
Christoph Lameter committed
758
759
 * 	A. Free pointer (if we cannot overwrite object on free)
 * 	B. Tracking data for SLAB_STORE_USER
Christoph Lameter's avatar
Christoph Lameter committed
760
 * 	C. Padding to reach required alignment boundary or at mininum
Christoph Lameter's avatar
Christoph Lameter committed
761
 * 		one word if debugging is on to be able to detect writes
Christoph Lameter's avatar
Christoph Lameter committed
762
763
764
 * 		before the word boundary.
 *
 *	Padding is done using 0x5a (POISON_INUSE)
Christoph Lameter's avatar
Christoph Lameter committed
765
766
 *
 * object + s->size
Christoph Lameter's avatar
Christoph Lameter committed
767
 * 	Nothing is used beyond s->size.
Christoph Lameter's avatar
Christoph Lameter committed
768
 *
769
 * If slabcaches are merged then the object_size and inuse boundaries are mostly
Christoph Lameter's avatar
Christoph Lameter committed
770
 * ignored. And therefore no slab options that rely on these boundaries
Christoph Lameter's avatar
Christoph Lameter committed
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
 * 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);

786
787
	off += kasan_metadata_size(s);

Joonsoo Kim's avatar
Joonsoo Kim committed
788
	if (size_from_object(s) == off)
Christoph Lameter's avatar
Christoph Lameter committed
789
790
		return 1;

791
	return check_bytes_and_report(s, page, p, "Object padding",
Joonsoo Kim's avatar
Joonsoo Kim committed
792
			p + off, POISON_INUSE, size_from_object(s) - off);
Christoph Lameter's avatar
Christoph Lameter committed
793
794
}

795
/* Check the pad bytes at the end of a slab page */
Christoph Lameter's avatar
Christoph Lameter committed
796
797
static int slab_pad_check(struct kmem_cache *s, struct page *page)
{
798
799
800
801
802
	u8 *start;
	u8 *fault;
	u8 *end;
	int length;
	int remainder;
Christoph Lameter's avatar
Christoph Lameter committed
803
804
805
806

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

807
	start = page_address(page);
808
	length = (PAGE_SIZE << compound_order(page)) - s->reserved;
809
810
	end = start + length;
	remainder = length % s->size;
Christoph Lameter's avatar
Christoph Lameter committed
811
812
813
	if (!remainder)
		return 1;

814
	metadata_access_enable();
815
	fault = memchr_inv(end - remainder, POISON_INUSE, remainder);
816
	metadata_access_disable();
817
818
819
820
821
822
	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);
823
	print_section("Padding ", end - remainder, remainder);
824

Eric Dumazet's avatar
Eric Dumazet committed
825
	restore_bytes(s, "slab padding", POISON_INUSE, end - remainder, end);
826
	return 0;
Christoph Lameter's avatar
Christoph Lameter committed
827
828
829
}

static int check_object(struct kmem_cache *s, struct page *page,
830
					void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
831
832
{
	u8 *p = object;
833
	u8 *endobject = object + s->object_size;
Christoph Lameter's avatar
Christoph Lameter committed
834
835

	if (s->flags & SLAB_RED_ZONE) {
Joonsoo Kim's avatar
Joonsoo Kim committed
836
837
838
839
		if (!check_bytes_and_report(s, page, object, "Redzone",
			object - s->red_left_pad, val, s->red_left_pad))
			return 0;

840
		if (!check_bytes_and_report(s, page, object, "Redzone",
841
			endobject, val, s->inuse - s->object_size))
Christoph Lameter's avatar
Christoph Lameter committed
842
843
			return 0;
	} else {
844
		if ((s->flags & SLAB_POISON) && s->object_size < s->inuse) {
Ingo Molnar's avatar
Ingo Molnar committed
845
			check_bytes_and_report(s, page, p, "Alignment padding",
846
847
				endobject, POISON_INUSE,
				s->inuse - s->object_size);
Ingo Molnar's avatar
Ingo Molnar committed
848
		}
Christoph Lameter's avatar
Christoph Lameter committed
849
850
851
	}

	if (s->flags & SLAB_POISON) {
852
		if (val != SLUB_RED_ACTIVE && (s->flags & __OBJECT_POISON) &&
853
			(!check_bytes_and_report(s, page, p, "Poison", p,
854
					POISON_FREE, s->object_size - 1) ||
855
			 !check_bytes_and_report(s, page, p, "Poison",
856
				p + s->object_size - 1, POISON_END, 1)))
Christoph Lameter's avatar
Christoph Lameter committed
857
858
859
860
861
862
863
			return 0;
		/*
		 * check_pad_bytes cleans up on its own.
		 */
		check_pad_bytes(s, page, p);
	}

864
	if (!s->offset && val == SLUB_RED_ACTIVE)
Christoph Lameter's avatar
Christoph Lameter committed
865
866
867
868
869
870
871
872
873
874
		/*
		 * 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
875
		 * No choice but to zap it and thus lose the remainder
Christoph Lameter's avatar
Christoph Lameter committed
876
		 * of the free objects in this slab. May cause
Christoph Lameter's avatar
Christoph Lameter committed
877
		 * another error because the object count is now wrong.
Christoph Lameter's avatar
Christoph Lameter committed
878
		 */
879
		set_freepointer(s, p, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
880
881
882
883
884
885
886
		return 0;
	}
	return 1;
}

static int check_slab(struct kmem_cache *s, struct page *page)
{
887
888
	int maxobj;

Christoph Lameter's avatar
Christoph Lameter committed
889
890
891
	VM_BUG_ON(!irqs_disabled());

	if (!PageSlab(page)) {
892
		slab_err(s, page, "Not a valid slab page");
Christoph Lameter's avatar
Christoph Lameter committed
893
894
		return 0;
	}
895

896
	maxobj = order_objects(compound_order(page), s->size, s->reserved);
897
898
	if (page->objects > maxobj) {
		slab_err(s, page, "objects %u > max %u",
899
			page->objects, maxobj);
900
901
902
		return 0;
	}
	if (page->inuse > page->objects) {
903
		slab_err(s, page, "inuse %u > max %u",
904
			page->inuse, page->objects);
Christoph Lameter's avatar
Christoph Lameter committed
905
906
907
908
909
910
911
912
		return 0;
	}
	/* Slab_pad_check fixes things up after itself */
	slab_pad_check(s, page);
	return 1;
}

/*
Christoph Lameter's avatar
Christoph Lameter committed
913
914
 * 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
915
916
917
918
 */
static int on_freelist(struct kmem_cache *s, struct page *page, void *search)
{
	int nr = 0;
919
	void *fp;
Christoph Lameter's avatar
Christoph Lameter committed
920
	void *object = NULL;
921
	int max_objects;
Christoph Lameter's avatar
Christoph Lameter committed
922

923
	fp = page->freelist;
924
	while (fp && nr <= page->objects) {
Christoph Lameter's avatar
Christoph Lameter committed
925
926
927
928
929
930
		if (fp == search)
			return 1;
		if (!check_valid_pointer(s, page, fp)) {
			if (object) {
				object_err(s, page, object,
					"Freechain corrupt");
931
				set_freepointer(s, object, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
932
			} else {
933
				slab_err(s, page, "Freepointer corrupt");
934
				page->freelist = NULL;
935
				page->inuse = page->objects;
936
				slab_fix(s, "Freelist cleared");
Christoph Lameter's avatar
Christoph Lameter committed
937
938
939
940
941
942
943
944
945
				return 0;
			}
			break;
		}
		object = fp;
		fp = get_freepointer(s, object);
		nr++;
	}

946
	max_objects = order_objects(compound_order(page), s->size, s->reserved);
947
948
	if (max_objects > MAX_OBJS_PER_PAGE)
		max_objects = MAX_OBJS_PER_PAGE;
949
950

	if (page->objects != max_objects) {
Joe Perches's avatar
Joe Perches committed
951
952
		slab_err(s, page, "Wrong number of objects. Found %d but should be %d",
			 page->objects, max_objects);
953
954
955
		page->objects = max_objects;
		slab_fix(s, "Number of objects adjusted.");
	}
956
	if (page->inuse != page->objects - nr) {
Joe Perches's avatar
Joe Perches committed
957
958
		slab_err(s, page, "Wrong object count. Counter is %d but counted were %d",
			 page->inuse, page->objects - nr);
959
		page->inuse = page->objects - nr;
960
		slab_fix(s, "Object count adjusted.");
Christoph Lameter's avatar
Christoph Lameter committed
961
962
963
964
	}
	return search == NULL;
}

965
966
static void trace(struct kmem_cache *s, struct page *page, void *object,
								int alloc)
Christoph Lameter's avatar
Christoph Lameter committed
967
968
{
	if (s->flags & SLAB_TRACE) {
969
		pr_info("TRACE %s %s 0x%p inuse=%d fp=0x%p\n",
Christoph Lameter's avatar
Christoph Lameter committed
970
971
972
973
974
975
			s->name,
			alloc ? "alloc" : "free",
			object, page->inuse,
			page->freelist);

		if (!alloc)
976
977
			print_section("Object ", (void *)object,
					s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
978
979
980
981
982

		dump_stack();
	}
}

983
/*
Christoph Lameter's avatar
Christoph Lameter committed
984
 * Tracking of fully allocated slabs for debugging purposes.
985
 */
986
987
static void add_full(struct kmem_cache *s,
	struct kmem_cache_node *n, struct page *page)
988
{
989
990
991
	if (!(s->flags & SLAB_STORE_USER))
		return;

992
	lockdep_assert_held(&n->list_lock);
993
994
995
	list_add(&page->lru, &n->full);
}

Peter Zijlstra's avatar
Peter Zijlstra committed
996
static void remove_full(struct kmem_cache *s, struct kmem_cache_node *n, struct page *page)
997
998
999
1000
{
	if (!(s->flags & SLAB_STORE_USER))
		return;

1001
	lockdep_assert_held(&n->list_lock);
1002
1003
1004
	list_del(&page->lru);
}

1005
1006
1007
1008
1009
1010
1011
1012
/* 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);
}

1013
1014
1015
1016
1017
static inline unsigned long node_nr_slabs(struct kmem_cache_node *n)
{
	return atomic_long_read(&n->nr_slabs);
}

1018
static inline void inc_slabs_node(struct kmem_cache *s, int node, int objects)
1019
1020
1021
1022
1023
1024
1025
1026
1027
{
	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).
	 */
1028
	if (likely(n)) {
1029
		atomic_long_inc(&n->nr_slabs);
1030
1031
		atomic_long_add(objects, &n->total_objects);
	}
1032
}
1033
static inline void dec_slabs_node(struct kmem_cache *s, int node, int objects)
1034
1035
1036
1037
{
	struct kmem_cache_node *n = get_node(s, node);

	atomic_long_dec(&n->nr_slabs);
1038
	atomic_long_sub(objects, &n->total_objects);
1039
1040
1041
}

/* Object debug checks for alloc/free paths */
Christoph Lameter's avatar
Christoph Lameter committed
1042
1043
1044
1045
1046
1047
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;

1048
	init_object(s, object, SLUB_RED_INACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1049
1050
1051
	init_tracking(s, object);
}

1052
static inline int alloc_consistency_checks(struct kmem_cache *s,
1053
					struct page *page,
1054
					void *object, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
1055
1056
{
	if (!check_slab(s, page))
1057
		return 0;
Christoph Lameter's avatar
Christoph Lameter committed
1058
1059
1060

	if (!check_valid_pointer(s, page, object)) {
		object_err(s, page, object, "Freelist Pointer check fails");
1061
		return 0;
Christoph Lameter's avatar
Christoph Lameter committed
1062
1063
	}

1064
	if (!check_object(s, page, object, SLUB_RED_INACTIVE))
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
		return 0;

	return 1;
}

static noinline int alloc_debug_processing(struct kmem_cache *s,
					struct page *page,
					void *object, unsigned long addr)
{
	if (s->flags & SLAB_CONSISTENCY_CHECKS) {
		if (!alloc_consistency_checks(s, page, object, addr))
			goto bad;
	}
Christoph Lameter's avatar
Christoph Lameter committed
1078

Christoph Lameter's avatar
Christoph Lameter committed
1079
1080
1081
1082
	/* Success perform special debug activities for allocs */
	if (s->flags & SLAB_STORE_USER)
		set_track(s, object, TRACK_ALLOC, addr);
	trace(s, page, object, 1);
1083
	init_object(s, object, SLUB_RED_ACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1084
	return 1;
Christoph Lameter's avatar
Christoph Lameter committed
1085

Christoph Lameter's avatar
Christoph Lameter committed
1086
1087
1088
1089
1090
bad:
	if (PageSlab(page)) {
		/*
		 * If this is a slab page then lets do the best we can
		 * to avoid issues in the future. Marking all objects
Christoph Lameter's avatar
Christoph Lameter committed
1091
		 * as used avoids touching the remaining objects.
Christoph Lameter's avatar