slub.c 133 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
128
129
130
131
132
133
134
135
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
136
137
138
139
140
141
142
143
144
145
146
/*
 * 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

147
148
149
/* Enable to log cmpxchg failures */
#undef SLUB_DEBUG_CMPXCHG

150
151
152
153
/*
 * Mininum number of partial slabs. These will be left on the partial
 * lists even if they are empty. kmem_cache_shrink may reclaim them.
 */
154
#define MIN_PARTIAL 5
Christoph Lameter's avatar
Christoph Lameter committed
155

156
157
158
/*
 * Maximum number of desirable partial slabs.
 * The existence of more partial slabs makes kmem_cache_shrink
159
 * sort the partial list by the number of objects in use.
160
161
162
 */
#define MAX_PARTIAL 10

Christoph Lameter's avatar
Christoph Lameter committed
163
164
#define DEBUG_DEFAULT_FLAGS (SLAB_DEBUG_FREE | SLAB_RED_ZONE | \
				SLAB_POISON | SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
165

166
/*
167
168
169
 * 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.
170
 */
171
#define DEBUG_METADATA_FLAGS (SLAB_RED_ZONE | SLAB_POISON | SLAB_STORE_USER)
172

173
174
#define OO_SHIFT	16
#define OO_MASK		((1 << OO_SHIFT) - 1)
175
#define MAX_OBJS_PER_PAGE	32767 /* since page.objects is u15 */
176

Christoph Lameter's avatar
Christoph Lameter committed
177
/* Internal SLUB flags */
Christoph Lameter's avatar
Christoph Lameter committed
178
#define __OBJECT_POISON		0x80000000UL /* Poison object */
179
#define __CMPXCHG_DOUBLE	0x40000000UL /* Use cmpxchg_double */
Christoph Lameter's avatar
Christoph Lameter committed
180
181
182
183
184

#ifdef CONFIG_SMP
static struct notifier_block slab_notifier;
#endif

185
186
187
/*
 * Tracking user of a slab.
 */
188
#define TRACK_ADDRS_COUNT 16
189
struct track {
190
	unsigned long addr;	/* Called from address */
191
192
193
#ifdef CONFIG_STACKTRACE
	unsigned long addrs[TRACK_ADDRS_COUNT];	/* Called from address */
#endif
194
195
196
197
198
199
200
	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 };

201
#ifdef CONFIG_SYSFS
Christoph Lameter's avatar
Christoph Lameter committed
202
203
static int sysfs_slab_add(struct kmem_cache *);
static int sysfs_slab_alias(struct kmem_cache *, const char *);
204
static void memcg_propagate_slab_attrs(struct kmem_cache *s);
Christoph Lameter's avatar
Christoph Lameter committed
205
#else
206
207
208
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; }
209
static inline void memcg_propagate_slab_attrs(struct kmem_cache *s) { }
Christoph Lameter's avatar
Christoph Lameter committed
210
211
#endif

212
static inline void stat(const struct kmem_cache *s, enum stat_item si)
213
214
{
#ifdef CONFIG_SLUB_STATS
215
216
217
218
219
	/*
	 * 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]);
220
221
222
#endif
}

Christoph Lameter's avatar
Christoph Lameter committed
223
224
225
226
/********************************************************************
 * 			Core slab cache functions
 *******************************************************************/

Christoph Lameter's avatar
Christoph Lameter committed
227
/* Verify that a pointer has an address that is valid within a slab page */
228
229
230
231
232
static inline int check_valid_pointer(struct kmem_cache *s,
				struct page *page, const void *object)
{
	void *base;

233
	if (!object)
234
235
		return 1;

236
	base = page_address(page);
237
	if (object < base || object >= base + page->objects * s->size ||
238
239
240
241
242
243
244
		(object - base) % s->size) {
		return 0;
	}

	return 1;
}

245
246
247
248
249
static inline void *get_freepointer(struct kmem_cache *s, void *object)
{
	return *(void **)(object + s->offset);
}

250
251
252
253
254
static void prefetch_freepointer(const struct kmem_cache *s, void *object)
{
	prefetch(object + s->offset);
}

255
256
257
258
259
260
261
262
263
264
265
266
static inline void *get_freepointer_safe(struct kmem_cache *s, void *object)
{
	void *p;

#ifdef CONFIG_DEBUG_PAGEALLOC
	probe_kernel_read(&p, (void **)(object + s->offset), sizeof(p));
#else
	p = get_freepointer(s, object);
#endif
	return p;
}

267
268
269
270
271
272
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 */
273
274
#define for_each_object(__p, __s, __addr, __objects) \
	for (__p = (__addr); __p < (__addr) + (__objects) * (__s)->size;\
275
276
			__p += (__s)->size)

277
278
279
280
#define for_each_object_idx(__p, __idx, __s, __addr, __objects) \
	for (__p = (__addr), __idx = 1; __idx <= __objects;\
			__p += (__s)->size, __idx++)

281
282
283
284
285
286
/* 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;
}

287
288
289
290
291
292
293
294
static inline size_t slab_ksize(const struct kmem_cache *s)
{
#ifdef CONFIG_SLUB_DEBUG
	/*
	 * Debugging requires use of the padding between object
	 * and whatever may come after it.
	 */
	if (s->flags & (SLAB_RED_ZONE | SLAB_POISON))
295
		return s->object_size;
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310

#endif
	/*
	 * If we have the need to store the freelist pointer
	 * back there or track user information then we can
	 * only use the space before that information.
	 */
	if (s->flags & (SLAB_DESTROY_BY_RCU | SLAB_STORE_USER))
		return s->inuse;
	/*
	 * Else we can use all the padding etc for the allocation
	 */
	return s->size;
}

311
312
313
314
315
static inline int order_objects(int order, unsigned long size, int reserved)
{
	return ((PAGE_SIZE << order) - reserved) / size;
}

316
static inline struct kmem_cache_order_objects oo_make(int order,
317
		unsigned long size, int reserved)
318
319
{
	struct kmem_cache_order_objects x = {
320
		(order << OO_SHIFT) + order_objects(order, size, reserved)
321
322
323
324
325
326
327
	};

	return x;
}

static inline int oo_order(struct kmem_cache_order_objects x)
{
328
	return x.x >> OO_SHIFT;
329
330
331
332
}

static inline int oo_objects(struct kmem_cache_order_objects x)
{
333
	return x.x & OO_MASK;
334
335
}

336
337
338
339
340
341
342
343
344
345
346
347
348
/*
 * Per slab locking using the pagelock
 */
static __always_inline void slab_lock(struct page *page)
{
	bit_spin_lock(PG_locked, &page->flags);
}

static __always_inline void slab_unlock(struct page *page)
{
	__bit_spin_unlock(PG_locked, &page->flags);
}

349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
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
	 * as page->_count.  If we assign to ->counters directly
	 * we run the risk of losing updates to page->_count, so
	 * be careful and only assign to the fields we need.
	 */
	page->frozen  = tmp.frozen;
	page->inuse   = tmp.inuse;
	page->objects = tmp.objects;
}

364
365
366
367
368
369
370
/* Interrupts must be disabled (for the fallback code to work right) */
static inline bool __cmpxchg_double_slab(struct kmem_cache *s, struct page *page,
		void *freelist_old, unsigned long counters_old,
		void *freelist_new, unsigned long counters_new,
		const char *n)
{
	VM_BUG_ON(!irqs_disabled());
371
372
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
373
	if (s->flags & __CMPXCHG_DOUBLE) {
374
		if (cmpxchg_double(&page->freelist, &page->counters,
375
376
				   freelist_old, counters_old,
				   freelist_new, counters_new))
377
			return true;
378
379
380
381
	} else
#endif
	{
		slab_lock(page);
382
383
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
384
			page->freelist = freelist_new;
385
			set_page_slub_counters(page, counters_new);
386
			slab_unlock(page);
387
			return true;
388
389
390
391
392
393
394
395
		}
		slab_unlock(page);
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

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

399
	return false;
400
401
}

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

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

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

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

440
	return false;
441
442
}

443
#ifdef CONFIG_SLUB_DEBUG
444
445
446
/*
 * Determine a map of object in use on a page.
 *
447
 * Node listlock must be held to guarantee that the page does
448
449
450
451
452
453
454
455
456
457
458
 * not vanish from under us.
 */
static void get_map(struct kmem_cache *s, struct page *page, unsigned long *map)
{
	void *p;
	void *addr = page_address(page);

	for (p = page->freelist; p; p = get_freepointer(s, p))
		set_bit(slab_index(p, s, addr), map);
}

459
460
461
/*
 * Debug settings:
 */
462
#if defined(CONFIG_SLUB_DEBUG_ON)
463
static int slub_debug = DEBUG_DEFAULT_FLAGS;
464
465
#elif defined(CONFIG_KASAN)
static int slub_debug = SLAB_STORE_USER;
466
#else
467
static int slub_debug;
468
#endif
469
470

static char *slub_debug_slabs;
471
static int disable_higher_order_debug;
472

473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
/*
 * 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
489
490
491
492
493
/*
 * Object debugging
 */
static void print_section(char *text, u8 *addr, unsigned int length)
{
494
	metadata_access_enable();
495
496
	print_hex_dump(KERN_ERR, text, DUMP_PREFIX_ADDRESS, 16, 1, addr,
			length, 1);
497
	metadata_access_disable();
Christoph Lameter's avatar
Christoph Lameter committed
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
}

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,
514
			enum track_item alloc, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
515
{
Akinobu Mita's avatar
Akinobu Mita committed
516
	struct track *p = get_track(s, object, alloc);
Christoph Lameter's avatar
Christoph Lameter committed
517
518

	if (addr) {
519
520
521
522
523
524
525
526
#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;
527
		metadata_access_enable();
528
		save_stack_trace(&trace);
529
		metadata_access_disable();
530
531
532
533
534
535
536
537
538

		/* 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
539
540
		p->addr = addr;
		p->cpu = smp_processor_id();
541
		p->pid = current->pid;
Christoph Lameter's avatar
Christoph Lameter committed
542
543
544
545
546
547
548
		p->when = jiffies;
	} else
		memset(p, 0, sizeof(struct track));
}

static void init_tracking(struct kmem_cache *s, void *object)
{
549
550
551
	if (!(s->flags & SLAB_STORE_USER))
		return;

552
553
	set_track(s, object, TRACK_FREE, 0UL);
	set_track(s, object, TRACK_ALLOC, 0UL);
Christoph Lameter's avatar
Christoph Lameter committed
554
555
556
557
558
559
560
}

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

561
562
	pr_err("INFO: %s in %pS age=%lu cpu=%u pid=%d\n",
	       s, (void *)t->addr, jiffies - t->when, t->cpu, t->pid);
563
564
565
566
567
#ifdef CONFIG_STACKTRACE
	{
		int i;
		for (i = 0; i < TRACK_ADDRS_COUNT; i++)
			if (t->addrs[i])
568
				pr_err("\t%pS\n", (void *)t->addrs[i]);
569
570
571
572
			else
				break;
	}
#endif
573
574
575
576
577
578
579
580
581
582
583
584
585
}

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)
{
586
	pr_err("INFO: Slab 0x%p objects=%u used=%u fp=0x%p flags=0x%04lx\n",
587
	       page, page->objects, page->inuse, page->freelist, page->flags);
588
589
590
591
592

}

static void slab_bug(struct kmem_cache *s, char *fmt, ...)
{
593
	struct va_format vaf;
594
595
596
	va_list args;

	va_start(args, fmt);
597
598
	vaf.fmt = fmt;
	vaf.va = &args;
599
	pr_err("=============================================================================\n");
600
	pr_err("BUG %s (%s): %pV\n", s->name, print_tainted(), &vaf);
601
	pr_err("-----------------------------------------------------------------------------\n\n");
602

603
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
604
	va_end(args);
Christoph Lameter's avatar
Christoph Lameter committed
605
606
}

607
608
static void slab_fix(struct kmem_cache *s, char *fmt, ...)
{
609
	struct va_format vaf;
610
611
612
	va_list args;

	va_start(args, fmt);
613
614
615
	vaf.fmt = fmt;
	vaf.va = &args;
	pr_err("FIX %s: %pV\n", s->name, &vaf);
616
617
618
619
	va_end(args);
}

static void print_trailer(struct kmem_cache *s, struct page *page, u8 *p)
Christoph Lameter's avatar
Christoph Lameter committed
620
621
{
	unsigned int off;	/* Offset of last byte */
622
	u8 *addr = page_address(page);
623
624
625
626
627

	print_tracking(s, p);

	print_page_info(page);

628
629
	pr_err("INFO: Object 0x%p @offset=%tu fp=0x%p\n\n",
	       p, p - addr, get_freepointer(s, p));
630
631

	if (p > addr + 16)
632
		print_section("Bytes b4 ", p - 16, 16);
Christoph Lameter's avatar
Christoph Lameter committed
633

634
	print_section("Object ", p, min_t(unsigned long, s->object_size,
635
				PAGE_SIZE));
Christoph Lameter's avatar
Christoph Lameter committed
636
	if (s->flags & SLAB_RED_ZONE)
637
638
		print_section("Redzone ", p + s->object_size,
			s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
639
640
641
642
643
644

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

645
	if (s->flags & SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
646
647
648
649
		off += 2 * sizeof(struct track);

	if (off != s->size)
		/* Beginning of the filler is the free pointer */
650
		print_section("Padding ", p + off, s->size - off);
651
652

	dump_stack();
Christoph Lameter's avatar
Christoph Lameter committed
653
654
}

655
void object_err(struct kmem_cache *s, struct page *page,
Christoph Lameter's avatar
Christoph Lameter committed
656
657
			u8 *object, char *reason)
{
658
	slab_bug(s, "%s", reason);
659
	print_trailer(s, page, object);
Christoph Lameter's avatar
Christoph Lameter committed
660
661
}

662
663
static void slab_err(struct kmem_cache *s, struct page *page,
			const char *fmt, ...)
Christoph Lameter's avatar
Christoph Lameter committed
664
665
666
667
{
	va_list args;
	char buf[100];

668
669
	va_start(args, fmt);
	vsnprintf(buf, sizeof(buf), fmt, args);
Christoph Lameter's avatar
Christoph Lameter committed
670
	va_end(args);
671
	slab_bug(s, "%s", buf);
672
	print_page_info(page);
Christoph Lameter's avatar
Christoph Lameter committed
673
674
675
	dump_stack();
}

676
static void init_object(struct kmem_cache *s, void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
677
678
679
680
{
	u8 *p = object;

	if (s->flags & __OBJECT_POISON) {
681
682
		memset(p, POISON_FREE, s->object_size - 1);
		p[s->object_size - 1] = POISON_END;
Christoph Lameter's avatar
Christoph Lameter committed
683
684
685
	}

	if (s->flags & SLAB_RED_ZONE)
686
		memset(p + s->object_size, val, s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
687
688
}

689
690
691
692
693
694
695
696
697
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,
698
			u8 *start, unsigned int value, unsigned int bytes)
699
700
701
702
{
	u8 *fault;
	u8 *end;

703
	metadata_access_enable();
704
	fault = memchr_inv(start, value, bytes);
705
	metadata_access_disable();
706
707
708
709
710
711
712
713
	if (!fault)
		return 1;

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

	slab_bug(s, "%s overwritten", what);
714
	pr_err("INFO: 0x%p-0x%p. First byte 0x%x instead of 0x%x\n",
715
716
717
718
719
					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
720
721
722
723
724
725
726
727
728
}

/*
 * 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
729
 *
Christoph Lameter's avatar
Christoph Lameter committed
730
731
732
 * 	Poisoning uses 0x6b (POISON_FREE) and the last byte is
 * 	0xa5 (POISON_END)
 *
733
 * object + s->object_size
Christoph Lameter's avatar
Christoph Lameter committed
734
 * 	Padding to reach word boundary. This is also used for Redzoning.
Christoph Lameter's avatar
Christoph Lameter committed
735
 * 	Padding is extended by another word if Redzoning is enabled and
736
 * 	object_size == inuse.
Christoph Lameter's avatar
Christoph Lameter committed
737
 *
Christoph Lameter's avatar
Christoph Lameter committed
738
739
740
741
 * 	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
742
743
 * 	Meta data starts here.
 *
Christoph Lameter's avatar
Christoph Lameter committed
744
745
 * 	A. Free pointer (if we cannot overwrite object on free)
 * 	B. Tracking data for SLAB_STORE_USER
Christoph Lameter's avatar
Christoph Lameter committed
746
 * 	C. Padding to reach required alignment boundary or at mininum
Christoph Lameter's avatar
Christoph Lameter committed
747
 * 		one word if debugging is on to be able to detect writes
Christoph Lameter's avatar
Christoph Lameter committed
748
749
750
 * 		before the word boundary.
 *
 *	Padding is done using 0x5a (POISON_INUSE)
Christoph Lameter's avatar
Christoph Lameter committed
751
752
 *
 * object + s->size
Christoph Lameter's avatar
Christoph Lameter committed
753
 * 	Nothing is used beyond s->size.
Christoph Lameter's avatar
Christoph Lameter committed
754
 *
755
 * If slabcaches are merged then the object_size and inuse boundaries are mostly
Christoph Lameter's avatar
Christoph Lameter committed
756
 * ignored. And therefore no slab options that rely on these boundaries
Christoph Lameter's avatar
Christoph Lameter committed
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
 * 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);

	if (s->size == off)
		return 1;

775
776
	return check_bytes_and_report(s, page, p, "Object padding",
				p + off, POISON_INUSE, s->size - off);
Christoph Lameter's avatar
Christoph Lameter committed
777
778
}

779
/* Check the pad bytes at the end of a slab page */
Christoph Lameter's avatar
Christoph Lameter committed
780
781
static int slab_pad_check(struct kmem_cache *s, struct page *page)
{
782
783
784
785
786
	u8 *start;
	u8 *fault;
	u8 *end;
	int length;
	int remainder;
Christoph Lameter's avatar
Christoph Lameter committed
787
788
789
790

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

791
	start = page_address(page);
792
	length = (PAGE_SIZE << compound_order(page)) - s->reserved;
793
794
	end = start + length;
	remainder = length % s->size;
Christoph Lameter's avatar
Christoph Lameter committed
795
796
797
	if (!remainder)
		return 1;

798
	metadata_access_enable();
799
	fault = memchr_inv(end - remainder, POISON_INUSE, remainder);
800
	metadata_access_disable();
801
802
803
804
805
806
	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);
807
	print_section("Padding ", end - remainder, remainder);
808

Eric Dumazet's avatar
Eric Dumazet committed
809
	restore_bytes(s, "slab padding", POISON_INUSE, end - remainder, end);
810
	return 0;
Christoph Lameter's avatar
Christoph Lameter committed
811
812
813
}

static int check_object(struct kmem_cache *s, struct page *page,
814
					void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
815
816
{
	u8 *p = object;
817
	u8 *endobject = object + s->object_size;
Christoph Lameter's avatar
Christoph Lameter committed
818
819

	if (s->flags & SLAB_RED_ZONE) {
820
		if (!check_bytes_and_report(s, page, object, "Redzone",
821
			endobject, val, s->inuse - s->object_size))
Christoph Lameter's avatar
Christoph Lameter committed
822
823
			return 0;
	} else {
824
		if ((s->flags & SLAB_POISON) && s->object_size < s->inuse) {
Ingo Molnar's avatar
Ingo Molnar committed
825
			check_bytes_and_report(s, page, p, "Alignment padding",
826
827
				endobject, POISON_INUSE,
				s->inuse - s->object_size);
Ingo Molnar's avatar
Ingo Molnar committed
828
		}
Christoph Lameter's avatar
Christoph Lameter committed
829
830
831
	}

	if (s->flags & SLAB_POISON) {
832
		if (val != SLUB_RED_ACTIVE && (s->flags & __OBJECT_POISON) &&
833
			(!check_bytes_and_report(s, page, p, "Poison", p,
834
					POISON_FREE, s->object_size - 1) ||
835
			 !check_bytes_and_report(s, page, p, "Poison",
836
				p + s->object_size - 1, POISON_END, 1)))
Christoph Lameter's avatar
Christoph Lameter committed
837
838
839
840
841
842
843
			return 0;
		/*
		 * check_pad_bytes cleans up on its own.
		 */
		check_pad_bytes(s, page, p);
	}

844
	if (!s->offset && val == SLUB_RED_ACTIVE)
Christoph Lameter's avatar
Christoph Lameter committed
845
846
847
848
849
850
851
852
853
854
		/*
		 * 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
855
		 * No choice but to zap it and thus lose the remainder
Christoph Lameter's avatar
Christoph Lameter committed
856
		 * of the free objects in this slab. May cause
Christoph Lameter's avatar
Christoph Lameter committed
857
		 * another error because the object count is now wrong.
Christoph Lameter's avatar
Christoph Lameter committed
858
		 */
859
		set_freepointer(s, p, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
860
861
862
863
864
865
866
		return 0;
	}
	return 1;
}

static int check_slab(struct kmem_cache *s, struct page *page)
{
867
868
	int maxobj;

Christoph Lameter's avatar
Christoph Lameter committed
869
870
871
	VM_BUG_ON(!irqs_disabled());

	if (!PageSlab(page)) {
872
		slab_err(s, page, "Not a valid slab page");
Christoph Lameter's avatar
Christoph Lameter committed
873
874
		return 0;
	}
875

876
	maxobj = order_objects(compound_order(page), s->size, s->reserved);
877
878
	if (page->objects > maxobj) {
		slab_err(s, page, "objects %u > max %u",
879
			page->objects, maxobj);
880
881
882
		return 0;
	}
	if (page->inuse > page->objects) {
883
		slab_err(s, page, "inuse %u > max %u",
884
			page->inuse, page->objects);
Christoph Lameter's avatar
Christoph Lameter committed
885
886
887
888
889
890
891
892
		return 0;
	}
	/* Slab_pad_check fixes things up after itself */
	slab_pad_check(s, page);
	return 1;
}

/*
Christoph Lameter's avatar
Christoph Lameter committed
893
894
 * 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
895
896
897
898
 */
static int on_freelist(struct kmem_cache *s, struct page *page, void *search)
{
	int nr = 0;
899
	void *fp;
Christoph Lameter's avatar
Christoph Lameter committed
900
	void *object = NULL;
901
	int max_objects;
Christoph Lameter's avatar
Christoph Lameter committed
902

903
	fp = page->freelist;
904
	while (fp && nr <= page->objects) {
Christoph Lameter's avatar
Christoph Lameter committed
905
906
907
908
909
910
		if (fp == search)
			return 1;
		if (!check_valid_pointer(s, page, fp)) {
			if (object) {
				object_err(s, page, object,
					"Freechain corrupt");
911
				set_freepointer(s, object, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
912
			} else {
913
				slab_err(s, page, "Freepointer corrupt");
914
				page->freelist = NULL;
915
				page->inuse = page->objects;
916
				slab_fix(s, "Freelist cleared");
Christoph Lameter's avatar
Christoph Lameter committed
917
918
919
920
921
922
923
924
925
				return 0;
			}
			break;
		}
		object = fp;
		fp = get_freepointer(s, object);
		nr++;
	}

926
	max_objects = order_objects(compound_order(page), s->size, s->reserved);
927
928
	if (max_objects > MAX_OBJS_PER_PAGE)
		max_objects = MAX_OBJS_PER_PAGE;
929
930
931
932
933
934
935

	if (page->objects != max_objects) {
		slab_err(s, page, "Wrong number of objects. Found %d but "
			"should be %d", page->objects, max_objects);
		page->objects = max_objects;
		slab_fix(s, "Number of objects adjusted.");
	}
936
	if (page->inuse != page->objects - nr) {
937
		slab_err(s, page, "Wrong object count. Counter is %d but "
938
939
			"counted were %d", page->inuse, page->objects - nr);
		page->inuse = page->objects - nr;
940
		slab_fix(s, "Object count adjusted.");
Christoph Lameter's avatar
Christoph Lameter committed
941
942
943
944
	}
	return search == NULL;
}

945
946
static void trace(struct kmem_cache *s, struct page *page, void *object,
								int alloc)
Christoph Lameter's avatar
Christoph Lameter committed
947
948
{
	if (s->flags & SLAB_TRACE) {
949
		pr_info("TRACE %s %s 0x%p inuse=%d fp=0x%p\n",
Christoph Lameter's avatar
Christoph Lameter committed
950
951
952
953
954
955
			s->name,
			alloc ? "alloc" : "free",
			object, page->inuse,
			page->freelist);

		if (!alloc)
956
957
			print_section("Object ", (void *)object,
					s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
958
959
960
961
962

		dump_stack();
	}
}

963
/*
Christoph Lameter's avatar
Christoph Lameter committed
964
 * Tracking of fully allocated slabs for debugging purposes.
965
 */
966
967
static void add_full(struct kmem_cache *s,
	struct kmem_cache_node *n, struct page *page)
968
{
969
970
971
	if (!(s->flags & SLAB_STORE_USER))
		return;

972
	lockdep_assert_held(&n->list_lock);
973
974
975
	list_add(&page->lru, &n->full);
}

Peter Zijlstra's avatar
Peter Zijlstra committed
976
static void remove_full(struct kmem_cache *s, struct kmem_cache_node *n, struct page *page)
977
978
979
980
{
	if (!(s->flags & SLAB_STORE_USER))
		return;

981
	lockdep_assert_held(&n->list_lock);
982
983
984
	list_del(&page->lru);
}

985
986
987
988
989
990
991
992
/* 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);
}

993
994
995
996
997
static inline unsigned long node_nr_slabs(struct kmem_cache_node *n)
{
	return atomic_long_read(&n->nr_slabs);
}

998
static inline void inc_slabs_node(struct kmem_cache *s, int node, int objects)
999
1000
1001
1002
1003
1004
1005
1006
1007
{
	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).
	 */
1008
	if (likely(n)) {
1009
		atomic_long_inc(&n->nr_slabs);
1010
1011
		atomic_long_add(objects, &n->total_objects);
	}
1012
}
1013
static inline void dec_slabs_node(struct kmem_cache *s, int node, int objects)
1014
1015
1016
1017
{
	struct kmem_cache_node *n = get_node(s, node);

	atomic_long_dec(&n->nr_slabs);
1018
	atomic_long_sub(objects, &n->total_objects);
1019
1020
1021
}

/* Object debug checks for alloc/free paths */
Christoph Lameter's avatar
Christoph Lameter committed
1022
1023
1024
1025
1026
1027
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;

1028
	init_object(s, object, SLUB_RED_INACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1029
1030
1031
	init_tracking(s, object);
}

1032
1033
static noinline int alloc_debug_processing(struct kmem_cache *s,
					struct page *page,
1034
					void *object, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
1035
1036
1037
1038
1039
1040
{
	if (!check_slab(s, page))
		goto bad;

	if (!check_valid_pointer(s, page, object)) {
		object_err(s, page, object, "Freelist Pointer check fails");
1041
		goto bad;
Christoph Lameter's avatar
Christoph Lameter committed
1042
1043
	}

1044
	if (!check_object(s, page, object, SLUB_RED_INACTIVE))
Christoph Lameter's avatar
Christoph Lameter committed
1045
1046
		goto bad;

Christoph Lameter's avatar
Christoph Lameter committed
1047
1048
1049
1050
	/* 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);
1051
	init_object(s, object, SLUB_RED_ACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1052
	return 1;
Christoph Lameter's avatar
Christoph Lameter committed
1053

Christoph Lameter's avatar
Christoph Lameter committed
1054
1055
1056
1057
1058
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
1059
		 * as used avoids touching the remaining objects.
Christoph Lameter's avatar
Christoph Lameter committed
1060
		 */
1061
		slab_fix(s, "Marking all objects used");
1062
		page->inuse = page->objects;
1063
		page->freelist = NULL;
Christoph Lameter's avatar
Christoph Lameter committed
1064
1065
1066
1067
	}
	return 0;
}

1068
/* Supports checking bulk free of a constructed freelist */
1069
static noinline struct kmem_cache_node *free_debug_processing(
1070
1071
	struct kmem_cache *s, struct page *page,
	void *head, void *tail, int bulk_cnt,
1072
	unsigned long addr, unsigned long *flags)
Christoph Lameter's avatar
Christoph Lameter committed
1073
{
1074
	struct kmem_cache_node *n = get_node(s, page_to_nid(page));
1075
1076
	void *object = head;
	int cnt = 0;
1077

1078
	spin_lock_irqsave(&n->list_lock, *flags);
1079
1080
	slab_lock(page);

Christoph Lameter's avatar
Christoph Lameter committed
1081
1082
1083
	if (!check_slab(s, page))
		goto fail;

1084
1085
1086
next_object:
	cnt++;

Christoph Lameter's avatar
Christoph Lameter committed
1087
	if (!check_valid_pointer(s, page, object)) {
1088
		slab_err(s, page, "Invalid object pointer 0x%p", object);
Christoph Lameter's avatar
Christoph Lameter committed
1089
1090
1091
1092
		goto fail;
	}

	if (on_freelist(s, page, object)) {
1093
		object_err(s, page, object, "Object already free");
Christoph Lameter's avatar
Christoph Lameter committed
1094
1095
1096
		goto fail;
	}

1097
	if (!check_object(