slub.c 141 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);
217
static void sysfs_slab_remove(struct kmem_cache *s);
Christoph Lameter's avatar
Christoph Lameter committed
218
#else
219
220
221
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; }
222
static inline void memcg_propagate_slab_attrs(struct kmem_cache *s) { }
223
static inline void sysfs_slab_remove(struct kmem_cache *s) { }
Christoph Lameter's avatar
Christoph Lameter committed
224
225
#endif

226
static inline void stat(const struct kmem_cache *s, enum stat_item si)
227
228
{
#ifdef CONFIG_SLUB_STATS
229
230
231
232
233
	/*
	 * 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]);
234
235
236
#endif
}

Christoph Lameter's avatar
Christoph Lameter committed
237
238
239
240
/********************************************************************
 * 			Core slab cache functions
 *******************************************************************/

241
242
243
244
245
static inline void *get_freepointer(struct kmem_cache *s, void *object)
{
	return *(void **)(object + s->offset);
}

246
247
248
249
250
static void prefetch_freepointer(const struct kmem_cache *s, void *object)
{
	prefetch(object + s->offset);
}

251
252
253
254
static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
{
	void *p;

255
256
257
	if (!debug_pagealloc_enabled())
		return get_freepointer(s, object);

258
259
260
261
	probe_kernel_read(&p, (void **)(object + s->offset), sizeof(p));
	return p;
}

262
263
264
265
266
267
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 */
268
#define for_each_object(__p, __s, __addr, __objects) \
Joonsoo Kim's avatar
Joonsoo Kim committed
269
270
271
	for (__p = fixup_red_left(__s, __addr); \
		__p < (__addr) + (__objects) * (__s)->size; \
		__p += (__s)->size)
272

273
#define for_each_object_idx(__p, __idx, __s, __addr, __objects) \
Joonsoo Kim's avatar
Joonsoo Kim committed
274
275
276
	for (__p = fixup_red_left(__s, __addr), __idx = 1; \
		__idx <= __objects; \
		__p += (__s)->size, __idx++)
277

278
279
280
281
282
283
/* 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;
}

284
285
286
287
288
static inline int order_objects(int order, unsigned long size, int reserved)
{
	return ((PAGE_SIZE << order) - reserved) / size;
}

289
static inline struct kmem_cache_order_objects oo_make(int order,
290
		unsigned long size, int reserved)
291
292
{
	struct kmem_cache_order_objects x = {
293
		(order << OO_SHIFT) + order_objects(order, size, reserved)
294
295
296
297
298
299
300
	};

	return x;
}

static inline int oo_order(struct kmem_cache_order_objects x)
{
301
	return x.x >> OO_SHIFT;
302
303
304
305
}

static inline int oo_objects(struct kmem_cache_order_objects x)
{
306
	return x.x & OO_MASK;
307
308
}

309
310
311
312
313
/*
 * Per slab locking using the pagelock
 */
static __always_inline void slab_lock(struct page *page)
{
314
	VM_BUG_ON_PAGE(PageTail(page), page);
315
316
317
318
319
	bit_spin_lock(PG_locked, &page->flags);
}

static __always_inline void slab_unlock(struct page *page)
{
320
	VM_BUG_ON_PAGE(PageTail(page), page);
321
322
323
	__bit_spin_unlock(PG_locked, &page->flags);
}

324
325
326
327
328
329
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
330
331
	 * as page->_refcount.  If we assign to ->counters directly
	 * we run the risk of losing updates to page->_refcount, so
332
333
334
335
336
337
338
	 * be careful and only assign to the fields we need.
	 */
	page->frozen  = tmp.frozen;
	page->inuse   = tmp.inuse;
	page->objects = tmp.objects;
}

339
340
341
342
343
344
345
/* 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());
346
347
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
348
	if (s->flags & __CMPXCHG_DOUBLE) {
349
		if (cmpxchg_double(&page->freelist, &page->counters,
350
351
				   freelist_old, counters_old,
				   freelist_new, counters_new))
352
			return true;
353
354
355
356
	} else
#endif
	{
		slab_lock(page);
357
358
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
359
			page->freelist = freelist_new;
360
			set_page_slub_counters(page, counters_new);
361
			slab_unlock(page);
362
			return true;
363
364
365
366
367
368
369
370
		}
		slab_unlock(page);
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
371
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
372
373
#endif

374
	return false;
375
376
}

377
378
379
380
381
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)
{
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
	} else
#endif
	{
392
393
394
		unsigned long flags;

		local_irq_save(flags);
395
		slab_lock(page);
396
397
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
398
			page->freelist = freelist_new;
399
			set_page_slub_counters(page, counters_new);
400
			slab_unlock(page);
401
			local_irq_restore(flags);
402
			return true;
403
		}
404
		slab_unlock(page);
405
		local_irq_restore(flags);
406
407
408
409
410
411
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
412
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
413
414
#endif

415
	return false;
416
417
}

418
#ifdef CONFIG_SLUB_DEBUG
419
420
421
/*
 * Determine a map of object in use on a page.
 *
422
 * Node listlock must be held to guarantee that the page does
423
424
425
426
427
428
429
430
431
432
433
 * 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
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
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;
}

450
451
452
/*
 * Debug settings:
 */
453
#if defined(CONFIG_SLUB_DEBUG_ON)
454
455
static int slub_debug = DEBUG_DEFAULT_FLAGS;
#else
456
static int slub_debug;
457
#endif
458
459

static char *slub_debug_slabs;
460
static int disable_higher_order_debug;
461

462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
/*
 * 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
478
479
480
/*
 * Object debugging
 */
Joonsoo Kim's avatar
Joonsoo Kim committed
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500

/* 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;
}

501
502
static void print_section(char *level, char *text, u8 *addr,
			  unsigned int length)
Christoph Lameter's avatar
Christoph Lameter committed
503
{
504
	metadata_access_enable();
505
	print_hex_dump(level, text, DUMP_PREFIX_ADDRESS, 16, 1, addr,
506
			length, 1);
507
	metadata_access_disable();
Christoph Lameter's avatar
Christoph Lameter committed
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
}

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,
524
			enum track_item alloc, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
525
{
Akinobu Mita's avatar
Akinobu Mita committed
526
	struct track *p = get_track(s, object, alloc);
Christoph Lameter's avatar
Christoph Lameter committed
527
528

	if (addr) {
529
530
531
532
533
534
535
536
#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;
537
		metadata_access_enable();
538
		save_stack_trace(&trace);
539
		metadata_access_disable();
540
541
542
543
544
545
546
547
548

		/* 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
549
550
		p->addr = addr;
		p->cpu = smp_processor_id();
551
		p->pid = current->pid;
Christoph Lameter's avatar
Christoph Lameter committed
552
553
554
555
556
557
558
		p->when = jiffies;
	} else
		memset(p, 0, sizeof(struct track));
}

static void init_tracking(struct kmem_cache *s, void *object)
{
559
560
561
	if (!(s->flags & SLAB_STORE_USER))
		return;

562
563
	set_track(s, object, TRACK_FREE, 0UL);
	set_track(s, object, TRACK_ALLOC, 0UL);
Christoph Lameter's avatar
Christoph Lameter committed
564
565
566
567
568
569
570
}

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

571
572
	pr_err("INFO: %s in %pS age=%lu cpu=%u pid=%d\n",
	       s, (void *)t->addr, jiffies - t->when, t->cpu, t->pid);
573
574
575
576
577
#ifdef CONFIG_STACKTRACE
	{
		int i;
		for (i = 0; i < TRACK_ADDRS_COUNT; i++)
			if (t->addrs[i])
578
				pr_err("\t%pS\n", (void *)t->addrs[i]);
579
580
581
582
			else
				break;
	}
#endif
583
584
585
586
587
588
589
590
591
592
593
594
595
}

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)
{
596
	pr_err("INFO: Slab 0x%p objects=%u used=%u fp=0x%p flags=0x%04lx\n",
597
	       page, page->objects, page->inuse, page->freelist, page->flags);
598
599
600
601
602

}

static void slab_bug(struct kmem_cache *s, char *fmt, ...)
{
603
	struct va_format vaf;
604
605
606
	va_list args;

	va_start(args, fmt);
607
608
	vaf.fmt = fmt;
	vaf.va = &args;
609
	pr_err("=============================================================================\n");
610
	pr_err("BUG %s (%s): %pV\n", s->name, print_tainted(), &vaf);
611
	pr_err("-----------------------------------------------------------------------------\n\n");
612

613
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
614
	va_end(args);
Christoph Lameter's avatar
Christoph Lameter committed
615
616
}

617
618
static void slab_fix(struct kmem_cache *s, char *fmt, ...)
{
619
	struct va_format vaf;
620
621
622
	va_list args;

	va_start(args, fmt);
623
624
625
	vaf.fmt = fmt;
	vaf.va = &args;
	pr_err("FIX %s: %pV\n", s->name, &vaf);
626
627
628
629
	va_end(args);
}

static void print_trailer(struct kmem_cache *s, struct page *page, u8 *p)
Christoph Lameter's avatar
Christoph Lameter committed
630
631
{
	unsigned int off;	/* Offset of last byte */
632
	u8 *addr = page_address(page);
633
634
635
636
637

	print_tracking(s, p);

	print_page_info(page);

638
639
	pr_err("INFO: Object 0x%p @offset=%tu fp=0x%p\n\n",
	       p, p - addr, get_freepointer(s, p));
640

Joonsoo Kim's avatar
Joonsoo Kim committed
641
	if (s->flags & SLAB_RED_ZONE)
642
643
		print_section(KERN_ERR, "Redzone ", p - s->red_left_pad,
			      s->red_left_pad);
Joonsoo Kim's avatar
Joonsoo Kim committed
644
	else if (p > addr + 16)
645
		print_section(KERN_ERR, "Bytes b4 ", p - 16, 16);
Christoph Lameter's avatar
Christoph Lameter committed
646

647
648
	print_section(KERN_ERR, "Object ", p,
		      min_t(unsigned long, s->object_size, PAGE_SIZE));
Christoph Lameter's avatar
Christoph Lameter committed
649
	if (s->flags & SLAB_RED_ZONE)
650
		print_section(KERN_ERR, "Redzone ", p + s->object_size,
651
			s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
652
653
654
655
656
657

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

658
	if (s->flags & SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
659
660
		off += 2 * sizeof(struct track);

661
662
	off += kasan_metadata_size(s);

Joonsoo Kim's avatar
Joonsoo Kim committed
663
	if (off != size_from_object(s))
Christoph Lameter's avatar
Christoph Lameter committed
664
		/* Beginning of the filler is the free pointer */
665
666
		print_section(KERN_ERR, "Padding ", p + off,
			      size_from_object(s) - off);
667
668

	dump_stack();
Christoph Lameter's avatar
Christoph Lameter committed
669
670
}

671
void object_err(struct kmem_cache *s, struct page *page,
Christoph Lameter's avatar
Christoph Lameter committed
672
673
			u8 *object, char *reason)
{
674
	slab_bug(s, "%s", reason);
675
	print_trailer(s, page, object);
Christoph Lameter's avatar
Christoph Lameter committed
676
677
}

678
679
static void slab_err(struct kmem_cache *s, struct page *page,
			const char *fmt, ...)
Christoph Lameter's avatar
Christoph Lameter committed
680
681
682
683
{
	va_list args;
	char buf[100];

684
685
	va_start(args, fmt);
	vsnprintf(buf, sizeof(buf), fmt, args);
Christoph Lameter's avatar
Christoph Lameter committed
686
	va_end(args);
687
	slab_bug(s, "%s", buf);
688
	print_page_info(page);
Christoph Lameter's avatar
Christoph Lameter committed
689
690
691
	dump_stack();
}

692
static void init_object(struct kmem_cache *s, void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
693
694
695
{
	u8 *p = object;

Joonsoo Kim's avatar
Joonsoo Kim committed
696
697
698
	if (s->flags & SLAB_RED_ZONE)
		memset(p - s->red_left_pad, val, s->red_left_pad);

Christoph Lameter's avatar
Christoph Lameter committed
699
	if (s->flags & __OBJECT_POISON) {
700
701
		memset(p, POISON_FREE, s->object_size - 1);
		p[s->object_size - 1] = POISON_END;
Christoph Lameter's avatar
Christoph Lameter committed
702
703
704
	}

	if (s->flags & SLAB_RED_ZONE)
705
		memset(p + s->object_size, val, s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
706
707
}

708
709
710
711
712
713
714
715
716
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,
717
			u8 *start, unsigned int value, unsigned int bytes)
718
719
720
721
{
	u8 *fault;
	u8 *end;

722
	metadata_access_enable();
723
	fault = memchr_inv(start, value, bytes);
724
	metadata_access_disable();
725
726
727
728
729
730
731
732
	if (!fault)
		return 1;

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

	slab_bug(s, "%s overwritten", what);
733
	pr_err("INFO: 0x%p-0x%p. First byte 0x%x instead of 0x%x\n",
734
735
736
737
738
					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
739
740
741
742
743
744
745
746
747
}

/*
 * 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
748
 *
Christoph Lameter's avatar
Christoph Lameter committed
749
750
751
 * 	Poisoning uses 0x6b (POISON_FREE) and the last byte is
 * 	0xa5 (POISON_END)
 *
752
 * object + s->object_size
Christoph Lameter's avatar
Christoph Lameter committed
753
 * 	Padding to reach word boundary. This is also used for Redzoning.
Christoph Lameter's avatar
Christoph Lameter committed
754
 * 	Padding is extended by another word if Redzoning is enabled and
755
 * 	object_size == inuse.
Christoph Lameter's avatar
Christoph Lameter committed
756
 *
Christoph Lameter's avatar
Christoph Lameter committed
757
758
759
760
 * 	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
761
762
 * 	Meta data starts here.
 *
Christoph Lameter's avatar
Christoph Lameter committed
763
764
 * 	A. Free pointer (if we cannot overwrite object on free)
 * 	B. Tracking data for SLAB_STORE_USER
Christoph Lameter's avatar
Christoph Lameter committed
765
 * 	C. Padding to reach required alignment boundary or at mininum
Christoph Lameter's avatar
Christoph Lameter committed
766
 * 		one word if debugging is on to be able to detect writes
Christoph Lameter's avatar
Christoph Lameter committed
767
768
769
 * 		before the word boundary.
 *
 *	Padding is done using 0x5a (POISON_INUSE)
Christoph Lameter's avatar
Christoph Lameter committed
770
771
 *
 * object + s->size
Christoph Lameter's avatar
Christoph Lameter committed
772
 * 	Nothing is used beyond s->size.
Christoph Lameter's avatar
Christoph Lameter committed
773
 *
774
 * If slabcaches are merged then the object_size and inuse boundaries are mostly
Christoph Lameter's avatar
Christoph Lameter committed
775
 * ignored. And therefore no slab options that rely on these boundaries
Christoph Lameter's avatar
Christoph Lameter committed
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
 * 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);

791
792
	off += kasan_metadata_size(s);

Joonsoo Kim's avatar
Joonsoo Kim committed
793
	if (size_from_object(s) == off)
Christoph Lameter's avatar
Christoph Lameter committed
794
795
		return 1;

796
	return check_bytes_and_report(s, page, p, "Object padding",
Joonsoo Kim's avatar
Joonsoo Kim committed
797
			p + off, POISON_INUSE, size_from_object(s) - off);
Christoph Lameter's avatar
Christoph Lameter committed
798
799
}

800
/* Check the pad bytes at the end of a slab page */
Christoph Lameter's avatar
Christoph Lameter committed
801
802
static int slab_pad_check(struct kmem_cache *s, struct page *page)
{
803
804
805
806
807
	u8 *start;
	u8 *fault;
	u8 *end;
	int length;
	int remainder;
Christoph Lameter's avatar
Christoph Lameter committed
808
809
810
811

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

812
	start = page_address(page);
813
	length = (PAGE_SIZE << compound_order(page)) - s->reserved;
814
815
	end = start + length;
	remainder = length % s->size;
Christoph Lameter's avatar
Christoph Lameter committed
816
817
818
	if (!remainder)
		return 1;

819
	metadata_access_enable();
820
	fault = memchr_inv(end - remainder, POISON_INUSE, remainder);
821
	metadata_access_disable();
822
823
824
825
826
827
	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);
828
	print_section(KERN_ERR, "Padding ", end - remainder, remainder);
829

Eric Dumazet's avatar
Eric Dumazet committed
830
	restore_bytes(s, "slab padding", POISON_INUSE, end - remainder, end);
831
	return 0;
Christoph Lameter's avatar
Christoph Lameter committed
832
833
834
}

static int check_object(struct kmem_cache *s, struct page *page,
835
					void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
836
837
{
	u8 *p = object;
838
	u8 *endobject = object + s->object_size;
Christoph Lameter's avatar
Christoph Lameter committed
839
840

	if (s->flags & SLAB_RED_ZONE) {
Joonsoo Kim's avatar
Joonsoo Kim committed
841
842
843
844
		if (!check_bytes_and_report(s, page, object, "Redzone",
			object - s->red_left_pad, val, s->red_left_pad))
			return 0;

845
		if (!check_bytes_and_report(s, page, object, "Redzone",
846
			endobject, val, s->inuse - s->object_size))
Christoph Lameter's avatar
Christoph Lameter committed
847
848
			return 0;
	} else {
849
		if ((s->flags & SLAB_POISON) && s->object_size < s->inuse) {
Ingo Molnar's avatar
Ingo Molnar committed
850
			check_bytes_and_report(s, page, p, "Alignment padding",
851
852
				endobject, POISON_INUSE,
				s->inuse - s->object_size);
Ingo Molnar's avatar
Ingo Molnar committed
853
		}
Christoph Lameter's avatar
Christoph Lameter committed
854
855
856
	}

	if (s->flags & SLAB_POISON) {
857
		if (val != SLUB_RED_ACTIVE && (s->flags & __OBJECT_POISON) &&
858
			(!check_bytes_and_report(s, page, p, "Poison", p,
859
					POISON_FREE, s->object_size - 1) ||
860
			 !check_bytes_and_report(s, page, p, "Poison",
861
				p + s->object_size - 1, POISON_END, 1)))
Christoph Lameter's avatar
Christoph Lameter committed
862
863
864
865
866
867
868
			return 0;
		/*
		 * check_pad_bytes cleans up on its own.
		 */
		check_pad_bytes(s, page, p);
	}

869
	if (!s->offset && val == SLUB_RED_ACTIVE)
Christoph Lameter's avatar
Christoph Lameter committed
870
871
872
873
874
875
876
877
878
879
		/*
		 * 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
880
		 * No choice but to zap it and thus lose the remainder
Christoph Lameter's avatar
Christoph Lameter committed
881
		 * of the free objects in this slab. May cause
Christoph Lameter's avatar
Christoph Lameter committed
882
		 * another error because the object count is now wrong.
Christoph Lameter's avatar
Christoph Lameter committed
883
		 */
884
		set_freepointer(s, p, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
885
886
887
888
889
890
891
		return 0;
	}
	return 1;
}

static int check_slab(struct kmem_cache *s, struct page *page)
{
892
893
	int maxobj;

Christoph Lameter's avatar
Christoph Lameter committed
894
895
896
	VM_BUG_ON(!irqs_disabled());

	if (!PageSlab(page)) {
897
		slab_err(s, page, "Not a valid slab page");
Christoph Lameter's avatar
Christoph Lameter committed
898
899
		return 0;
	}
900

901
	maxobj = order_objects(compound_order(page), s->size, s->reserved);
902
903
	if (page->objects > maxobj) {
		slab_err(s, page, "objects %u > max %u",
904
			page->objects, maxobj);
905
906
907
		return 0;
	}
	if (page->inuse > page->objects) {
908
		slab_err(s, page, "inuse %u > max %u",
909
			page->inuse, page->objects);
Christoph Lameter's avatar
Christoph Lameter committed
910
911
912
913
914
915
916
917
		return 0;
	}
	/* Slab_pad_check fixes things up after itself */
	slab_pad_check(s, page);
	return 1;
}

/*
Christoph Lameter's avatar
Christoph Lameter committed
918
919
 * 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
920
921
922
923
 */
static int on_freelist(struct kmem_cache *s, struct page *page, void *search)
{
	int nr = 0;
924
	void *fp;
Christoph Lameter's avatar
Christoph Lameter committed
925
	void *object = NULL;
926
	int max_objects;
Christoph Lameter's avatar
Christoph Lameter committed
927

928
	fp = page->freelist;
929
	while (fp && nr <= page->objects) {
Christoph Lameter's avatar
Christoph Lameter committed
930
931
932
933
934
935
		if (fp == search)
			return 1;
		if (!check_valid_pointer(s, page, fp)) {
			if (object) {
				object_err(s, page, object,
					"Freechain corrupt");
936
				set_freepointer(s, object, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
937
			} else {
938
				slab_err(s, page, "Freepointer corrupt");
939
				page->freelist = NULL;
940
				page->inuse = page->objects;
941
				slab_fix(s, "Freelist cleared");
Christoph Lameter's avatar
Christoph Lameter committed
942
943
944
945
946
947
948
949
950
				return 0;
			}
			break;
		}
		object = fp;
		fp = get_freepointer(s, object);
		nr++;
	}

951
	max_objects = order_objects(compound_order(page), s->size, s->reserved);
952
953
	if (max_objects > MAX_OBJS_PER_PAGE)
		max_objects = MAX_OBJS_PER_PAGE;
954
955

	if (page->objects != max_objects) {
Joe Perches's avatar
Joe Perches committed
956
957
		slab_err(s, page, "Wrong number of objects. Found %d but should be %d",
			 page->objects, max_objects);
958
959
960
		page->objects = max_objects;
		slab_fix(s, "Number of objects adjusted.");
	}
961
	if (page->inuse != page->objects - nr) {
Joe Perches's avatar
Joe Perches committed
962
963
		slab_err(s, page, "Wrong object count. Counter is %d but counted were %d",
			 page->inuse, page->objects - nr);
964
		page->inuse = page->objects - nr;
965
		slab_fix(s, "Object count adjusted.");
Christoph Lameter's avatar
Christoph Lameter committed
966
967
968
969
	}
	return search == NULL;
}

970
971
static void trace(struct kmem_cache *s, struct page *page, void *object,
								int alloc)
Christoph Lameter's avatar
Christoph Lameter committed
972
973
{
	if (s->flags & SLAB_TRACE) {
974
		pr_info("TRACE %s %s 0x%p inuse=%d fp=0x%p\n",
Christoph Lameter's avatar
Christoph Lameter committed
975
976
977
978
979
980
			s->name,
			alloc ? "alloc" : "free",
			object, page->inuse,
			page->freelist);

		if (!alloc)
981
			print_section(KERN_INFO, "Object ", (void *)object,
982
					s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
983
984
985
986
987

		dump_stack();
	}
}

988
/*
Christoph Lameter's avatar
Christoph Lameter committed
989
 * Tracking of fully allocated slabs for debugging purposes.
990
 */
991
992
static void add_full(struct kmem_cache *s,
	struct kmem_cache_node *n, struct page *page)
993
{
994
995
996
	if (!(s->flags & SLAB_STORE_USER))
		return;

997
	lockdep_assert_held(&n->list_lock);
998
999
1000
	list_add(&page->lru, &n->full);
}

Peter Zijlstra's avatar
Peter Zijlstra committed
1001
static void remove_full(struct kmem_cache *s, struct kmem_cache_node *n, struct page *page)
1002
1003
1004
1005
{
	if (!(s->flags & SLAB_STORE_USER))
		return;

1006
	lockdep_assert_held(&n->list_lock);
1007
1008
1009
	list_del(&page->lru);
}

1010
1011
1012
1013
1014
1015
1016
1017
/* 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);
}

1018
1019
1020
1021
1022
static inline unsigned long node_nr_slabs(struct kmem_cache_node *n)
{
	return atomic_long_read(&n->nr_slabs);
}

1023
static inline void inc_slabs_node(struct kmem_cache *s, int node, int objects)
1024
1025
1026
1027
1028
1029
1030
1031
1032
{
	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).
	 */
1033
	if (likely(n)) {
1034
		atomic_long_inc(&n->nr_slabs);
1035
1036
		atomic_long_add(objects, &n->total_objects);
	}
1037
}
1038
static inline void dec_slabs_node(struct kmem_cache *s, int node, int objects)
1039
1040
1041
1042
{
	struct kmem_cache_node *n = get_node(s, node);

	atomic_long_dec(&n->nr_slabs);
1043
	atomic_long_sub(objects, &n->total_objects);
1044
1045
1046
}

/* Object debug checks for alloc/free paths */
Christoph Lameter's avatar
Christoph Lameter committed
1047
1048
1049
1050
1051
1052
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;

1053
	init_object(s, object, SLUB_RED_INACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1054
1055
1056
	init_tracking(s, object);
}

1057
static inline int alloc_consistency_checks(struct kmem_cache *s,
1058
					struct page *page,
1059
					void *object, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
1060
1061
{
	if (!check_slab(s, page))
1062
		return 0;
Christoph Lameter's avatar
Christoph Lameter committed
1063
1064
1065

	if (!check_valid_pointer(s, page, object)) {
		object_err(s, page, object, "Freelist Pointer check fails");
1066
		return 0;
Christoph Lameter's avatar
Christoph Lameter committed
1067
1068
	}

1069
	if (!check_object(s, page, object, SLUB_RED_INACTIVE))
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
		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
1083