slub.c 126 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>
Vegard Nossum's avatar
Vegard Nossum committed
23
#include <linux/kmemcheck.h>
Christoph Lameter's avatar
Christoph Lameter committed
24
25
26
27
#include <linux/cpu.h>
#include <linux/cpuset.h>
#include <linux/mempolicy.h>
#include <linux/ctype.h>
28
#include <linux/debugobjects.h>
Christoph Lameter's avatar
Christoph Lameter committed
29
#include <linux/kallsyms.h>
30
#include <linux/memory.h>
Roman Zippel's avatar
Roman Zippel committed
31
#include <linux/math64.h>
Akinobu Mita's avatar
Akinobu Mita committed
32
#include <linux/fault-inject.h>
33
#include <linux/stacktrace.h>
34
#include <linux/prefetch.h>
35
#include <linux/memcontrol.h>
Christoph Lameter's avatar
Christoph Lameter committed
36

37
38
#include <trace/events/kmem.h>

39
40
#include "internal.h"

Christoph Lameter's avatar
Christoph Lameter committed
41
42
/*
 * Lock order:
43
 *   1. slab_mutex (Global Mutex)
44
45
 *   2. node->list_lock
 *   3. slab_lock(page) (Only on some arches and for debugging)
Christoph Lameter's avatar
Christoph Lameter committed
46
 *
47
 *   slab_mutex
48
 *
49
 *   The role of the slab_mutex is to protect the list of all the slabs
50
51
52
53
54
55
56
57
58
59
60
61
62
63
 *   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
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
 *
 *   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
84
85
 * 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
86
 * freed then the slab will show up again on the partial lists.
Christoph Lameter's avatar
Christoph Lameter committed
87
88
 * We track full slabs for debugging purposes though because otherwise we
 * cannot scan all objects.
Christoph Lameter's avatar
Christoph Lameter committed
89
90
91
92
93
94
95
 *
 * 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.
 *
96
97
98
99
100
101
102
103
104
105
106
107
 * 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
108
 * 			freelist that allows lockless access to
109
110
 * 			free objects in addition to the regular freelist
 * 			that requires the slab lock.
Christoph Lameter's avatar
Christoph Lameter committed
111
112
113
 *
 * PageError		Slab requires special handling due to debug
 * 			options set. This moves	slab handling out of
114
 * 			the fast path and disables lockless freelists.
Christoph Lameter's avatar
Christoph Lameter committed
115
116
 */

117
118
static inline int kmem_cache_debug(struct kmem_cache *s)
{
119
#ifdef CONFIG_SLUB_DEBUG
120
	return unlikely(s->flags & SLAB_DEBUG_FLAGS);
121
#else
122
	return 0;
123
#endif
124
}
125

126
127
128
129
130
131
132
133
134
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
135
136
137
138
139
140
141
142
143
144
145
/*
 * 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

146
147
148
/* Enable to log cmpxchg failures */
#undef SLUB_DEBUG_CMPXCHG

149
150
151
152
/*
 * Mininum number of partial slabs. These will be left on the partial
 * lists even if they are empty. kmem_cache_shrink may reclaim them.
 */
153
#define MIN_PARTIAL 5
Christoph Lameter's avatar
Christoph Lameter committed
154

155
156
157
/*
 * Maximum number of desirable partial slabs.
 * The existence of more partial slabs makes kmem_cache_shrink
158
 * sort the partial list by the number of objects in use.
159
160
161
 */
#define MAX_PARTIAL 10

Christoph Lameter's avatar
Christoph Lameter committed
162
163
#define DEBUG_DEFAULT_FLAGS (SLAB_DEBUG_FREE | SLAB_RED_ZONE | \
				SLAB_POISON | SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
164

165
/*
166
167
168
 * 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.
169
 */
170
#define DEBUG_METADATA_FLAGS (SLAB_RED_ZONE | SLAB_POISON | SLAB_STORE_USER)
171

172
173
#define OO_SHIFT	16
#define OO_MASK		((1 << OO_SHIFT) - 1)
174
#define MAX_OBJS_PER_PAGE	32767 /* since page.objects is u15 */
175

Christoph Lameter's avatar
Christoph Lameter committed
176
/* Internal SLUB flags */
Christoph Lameter's avatar
Christoph Lameter committed
177
#define __OBJECT_POISON		0x80000000UL /* Poison object */
178
#define __CMPXCHG_DOUBLE	0x40000000UL /* Use cmpxchg_double */
Christoph Lameter's avatar
Christoph Lameter committed
179
180
181
182
183

#ifdef CONFIG_SMP
static struct notifier_block slab_notifier;
#endif

184
185
186
/*
 * Tracking user of a slab.
 */
187
#define TRACK_ADDRS_COUNT 16
188
struct track {
189
	unsigned long addr;	/* Called from address */
190
191
192
#ifdef CONFIG_STACKTRACE
	unsigned long addrs[TRACK_ADDRS_COUNT];	/* Called from address */
#endif
193
194
195
196
197
198
199
	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 };

200
#ifdef CONFIG_SYSFS
Christoph Lameter's avatar
Christoph Lameter committed
201
202
static int sysfs_slab_add(struct kmem_cache *);
static int sysfs_slab_alias(struct kmem_cache *, const char *);
203
static void memcg_propagate_slab_attrs(struct kmem_cache *s);
Christoph Lameter's avatar
Christoph Lameter committed
204
#else
205
206
207
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; }
208
static inline void memcg_propagate_slab_attrs(struct kmem_cache *s) { }
Christoph Lameter's avatar
Christoph Lameter committed
209
210
#endif

211
static inline void stat(const struct kmem_cache *s, enum stat_item si)
212
213
{
#ifdef CONFIG_SLUB_STATS
214
215
216
217
218
	/*
	 * 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]);
219
220
221
#endif
}

Christoph Lameter's avatar
Christoph Lameter committed
222
223
224
225
/********************************************************************
 * 			Core slab cache functions
 *******************************************************************/

Christoph Lameter's avatar
Christoph Lameter committed
226
/* Verify that a pointer has an address that is valid within a slab page */
227
228
229
230
231
static inline int check_valid_pointer(struct kmem_cache *s,
				struct page *page, const void *object)
{
	void *base;

232
	if (!object)
233
234
		return 1;

235
	base = page_address(page);
236
	if (object < base || object >= base + page->objects * s->size ||
237
238
239
240
241
242
243
		(object - base) % s->size) {
		return 0;
	}

	return 1;
}

244
245
246
247
248
static inline void *get_freepointer(struct kmem_cache *s, void *object)
{
	return *(void **)(object + s->offset);
}

249
250
251
252
253
static void prefetch_freepointer(const struct kmem_cache *s, void *object)
{
	prefetch(object + s->offset);
}

254
255
256
257
258
259
260
261
262
263
264
265
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;
}

266
267
268
269
270
271
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 */
272
273
#define for_each_object(__p, __s, __addr, __objects) \
	for (__p = (__addr); __p < (__addr) + (__objects) * (__s)->size;\
274
275
			__p += (__s)->size)

276
277
278
279
#define for_each_object_idx(__p, __idx, __s, __addr, __objects) \
	for (__p = (__addr), __idx = 1; __idx <= __objects;\
			__p += (__s)->size, __idx++)

280
281
282
283
284
285
/* 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;
}

286
287
288
289
290
291
292
293
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))
294
		return s->object_size;
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309

#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;
}

310
311
312
313
314
static inline int order_objects(int order, unsigned long size, int reserved)
{
	return ((PAGE_SIZE << order) - reserved) / size;
}

315
static inline struct kmem_cache_order_objects oo_make(int order,
316
		unsigned long size, int reserved)
317
318
{
	struct kmem_cache_order_objects x = {
319
		(order << OO_SHIFT) + order_objects(order, size, reserved)
320
321
322
323
324
325
326
	};

	return x;
}

static inline int oo_order(struct kmem_cache_order_objects x)
{
327
	return x.x >> OO_SHIFT;
328
329
330
331
}

static inline int oo_objects(struct kmem_cache_order_objects x)
{
332
	return x.x & OO_MASK;
333
334
}

335
336
337
338
339
340
341
342
343
344
345
346
347
/*
 * 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);
}

348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
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;
}

363
364
365
366
367
368
369
/* 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());
370
371
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
372
	if (s->flags & __CMPXCHG_DOUBLE) {
373
		if (cmpxchg_double(&page->freelist, &page->counters,
374
375
376
				   freelist_old, counters_old,
				   freelist_new, counters_new))
			return 1;
377
378
379
380
	} else
#endif
	{
		slab_lock(page);
381
382
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
383
			page->freelist = freelist_new;
384
			set_page_slub_counters(page, counters_new);
385
386
387
388
389
390
391
392
393
394
			slab_unlock(page);
			return 1;
		}
		slab_unlock(page);
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
395
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
396
397
398
399
400
#endif

	return 0;
}

401
402
403
404
405
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)
{
406
407
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
408
	if (s->flags & __CMPXCHG_DOUBLE) {
409
		if (cmpxchg_double(&page->freelist, &page->counters,
410
411
412
				   freelist_old, counters_old,
				   freelist_new, counters_new))
			return 1;
413
414
415
	} else
#endif
	{
416
417
418
		unsigned long flags;

		local_irq_save(flags);
419
		slab_lock(page);
420
421
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
422
			page->freelist = freelist_new;
423
			set_page_slub_counters(page, counters_new);
424
			slab_unlock(page);
425
			local_irq_restore(flags);
426
427
			return 1;
		}
428
		slab_unlock(page);
429
		local_irq_restore(flags);
430
431
432
433
434
435
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
436
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
437
438
439
440
441
#endif

	return 0;
}

442
#ifdef CONFIG_SLUB_DEBUG
443
444
445
/*
 * Determine a map of object in use on a page.
 *
446
 * Node listlock must be held to guarantee that the page does
447
448
449
450
451
452
453
454
455
456
457
 * 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);
}

458
459
460
/*
 * Debug settings:
 */
461
462
463
#ifdef CONFIG_SLUB_DEBUG_ON
static int slub_debug = DEBUG_DEFAULT_FLAGS;
#else
464
static int slub_debug;
465
#endif
466
467

static char *slub_debug_slabs;
468
static int disable_higher_order_debug;
469

Christoph Lameter's avatar
Christoph Lameter committed
470
471
472
473
474
/*
 * Object debugging
 */
static void print_section(char *text, u8 *addr, unsigned int length)
{
475
476
	print_hex_dump(KERN_ERR, text, DUMP_PREFIX_ADDRESS, 16, 1, addr,
			length, 1);
Christoph Lameter's avatar
Christoph Lameter committed
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
}

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,
493
			enum track_item alloc, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
494
{
Akinobu Mita's avatar
Akinobu Mita committed
495
	struct track *p = get_track(s, object, alloc);
Christoph Lameter's avatar
Christoph Lameter committed
496
497

	if (addr) {
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
#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;
		save_stack_trace(&trace);

		/* 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
516
517
		p->addr = addr;
		p->cpu = smp_processor_id();
518
		p->pid = current->pid;
Christoph Lameter's avatar
Christoph Lameter committed
519
520
521
522
523
524
525
		p->when = jiffies;
	} else
		memset(p, 0, sizeof(struct track));
}

static void init_tracking(struct kmem_cache *s, void *object)
{
526
527
528
	if (!(s->flags & SLAB_STORE_USER))
		return;

529
530
	set_track(s, object, TRACK_FREE, 0UL);
	set_track(s, object, TRACK_ALLOC, 0UL);
Christoph Lameter's avatar
Christoph Lameter committed
531
532
533
534
535
536
537
}

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

538
539
	pr_err("INFO: %s in %pS age=%lu cpu=%u pid=%d\n",
	       s, (void *)t->addr, jiffies - t->when, t->cpu, t->pid);
540
541
542
543
544
#ifdef CONFIG_STACKTRACE
	{
		int i;
		for (i = 0; i < TRACK_ADDRS_COUNT; i++)
			if (t->addrs[i])
545
				pr_err("\t%pS\n", (void *)t->addrs[i]);
546
547
548
549
			else
				break;
	}
#endif
550
551
552
553
554
555
556
557
558
559
560
561
562
}

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)
{
563
	pr_err("INFO: Slab 0x%p objects=%u used=%u fp=0x%p flags=0x%04lx\n",
564
	       page, page->objects, page->inuse, page->freelist, page->flags);
565
566
567
568
569

}

static void slab_bug(struct kmem_cache *s, char *fmt, ...)
{
570
	struct va_format vaf;
571
572
573
	va_list args;

	va_start(args, fmt);
574
575
	vaf.fmt = fmt;
	vaf.va = &args;
576
	pr_err("=============================================================================\n");
577
	pr_err("BUG %s (%s): %pV\n", s->name, print_tainted(), &vaf);
578
	pr_err("-----------------------------------------------------------------------------\n\n");
579

580
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
581
	va_end(args);
Christoph Lameter's avatar
Christoph Lameter committed
582
583
}

584
585
static void slab_fix(struct kmem_cache *s, char *fmt, ...)
{
586
	struct va_format vaf;
587
588
589
	va_list args;

	va_start(args, fmt);
590
591
592
	vaf.fmt = fmt;
	vaf.va = &args;
	pr_err("FIX %s: %pV\n", s->name, &vaf);
593
594
595
596
	va_end(args);
}

static void print_trailer(struct kmem_cache *s, struct page *page, u8 *p)
Christoph Lameter's avatar
Christoph Lameter committed
597
598
{
	unsigned int off;	/* Offset of last byte */
599
	u8 *addr = page_address(page);
600
601
602
603
604

	print_tracking(s, p);

	print_page_info(page);

605
606
	pr_err("INFO: Object 0x%p @offset=%tu fp=0x%p\n\n",
	       p, p - addr, get_freepointer(s, p));
607
608

	if (p > addr + 16)
609
		print_section("Bytes b4 ", p - 16, 16);
Christoph Lameter's avatar
Christoph Lameter committed
610

611
	print_section("Object ", p, min_t(unsigned long, s->object_size,
612
				PAGE_SIZE));
Christoph Lameter's avatar
Christoph Lameter committed
613
	if (s->flags & SLAB_RED_ZONE)
614
615
		print_section("Redzone ", p + s->object_size,
			s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
616
617
618
619
620
621

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

622
	if (s->flags & SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
623
624
625
626
		off += 2 * sizeof(struct track);

	if (off != s->size)
		/* Beginning of the filler is the free pointer */
627
		print_section("Padding ", p + off, s->size - off);
628
629

	dump_stack();
Christoph Lameter's avatar
Christoph Lameter committed
630
631
632
633
634
}

static void object_err(struct kmem_cache *s, struct page *page,
			u8 *object, char *reason)
{
635
	slab_bug(s, "%s", reason);
636
	print_trailer(s, page, object);
Christoph Lameter's avatar
Christoph Lameter committed
637
638
}

639
640
static void slab_err(struct kmem_cache *s, struct page *page,
			const char *fmt, ...)
Christoph Lameter's avatar
Christoph Lameter committed
641
642
643
644
{
	va_list args;
	char buf[100];

645
646
	va_start(args, fmt);
	vsnprintf(buf, sizeof(buf), fmt, args);
Christoph Lameter's avatar
Christoph Lameter committed
647
	va_end(args);
648
	slab_bug(s, "%s", buf);
649
	print_page_info(page);
Christoph Lameter's avatar
Christoph Lameter committed
650
651
652
	dump_stack();
}

653
static void init_object(struct kmem_cache *s, void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
654
655
656
657
{
	u8 *p = object;

	if (s->flags & __OBJECT_POISON) {
658
659
		memset(p, POISON_FREE, s->object_size - 1);
		p[s->object_size - 1] = POISON_END;
Christoph Lameter's avatar
Christoph Lameter committed
660
661
662
	}

	if (s->flags & SLAB_RED_ZONE)
663
		memset(p + s->object_size, val, s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
664
665
}

666
667
668
669
670
671
672
673
674
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,
675
			u8 *start, unsigned int value, unsigned int bytes)
676
677
678
679
{
	u8 *fault;
	u8 *end;

680
	fault = memchr_inv(start, value, bytes);
681
682
683
684
685
686
687
688
	if (!fault)
		return 1;

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

	slab_bug(s, "%s overwritten", what);
689
	pr_err("INFO: 0x%p-0x%p. First byte 0x%x instead of 0x%x\n",
690
691
692
693
694
					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
695
696
697
698
699
700
701
702
703
}

/*
 * 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
704
 *
Christoph Lameter's avatar
Christoph Lameter committed
705
706
707
 * 	Poisoning uses 0x6b (POISON_FREE) and the last byte is
 * 	0xa5 (POISON_END)
 *
708
 * object + s->object_size
Christoph Lameter's avatar
Christoph Lameter committed
709
 * 	Padding to reach word boundary. This is also used for Redzoning.
Christoph Lameter's avatar
Christoph Lameter committed
710
 * 	Padding is extended by another word if Redzoning is enabled and
711
 * 	object_size == inuse.
Christoph Lameter's avatar
Christoph Lameter committed
712
 *
Christoph Lameter's avatar
Christoph Lameter committed
713
714
715
716
 * 	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
717
718
 * 	Meta data starts here.
 *
Christoph Lameter's avatar
Christoph Lameter committed
719
720
 * 	A. Free pointer (if we cannot overwrite object on free)
 * 	B. Tracking data for SLAB_STORE_USER
Christoph Lameter's avatar
Christoph Lameter committed
721
 * 	C. Padding to reach required alignment boundary or at mininum
Christoph Lameter's avatar
Christoph Lameter committed
722
 * 		one word if debugging is on to be able to detect writes
Christoph Lameter's avatar
Christoph Lameter committed
723
724
725
 * 		before the word boundary.
 *
 *	Padding is done using 0x5a (POISON_INUSE)
Christoph Lameter's avatar
Christoph Lameter committed
726
727
 *
 * object + s->size
Christoph Lameter's avatar
Christoph Lameter committed
728
 * 	Nothing is used beyond s->size.
Christoph Lameter's avatar
Christoph Lameter committed
729
 *
730
 * If slabcaches are merged then the object_size and inuse boundaries are mostly
Christoph Lameter's avatar
Christoph Lameter committed
731
 * ignored. And therefore no slab options that rely on these boundaries
Christoph Lameter's avatar
Christoph Lameter committed
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
 * 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;

750
751
	return check_bytes_and_report(s, page, p, "Object padding",
				p + off, POISON_INUSE, s->size - off);
Christoph Lameter's avatar
Christoph Lameter committed
752
753
}

754
/* Check the pad bytes at the end of a slab page */
Christoph Lameter's avatar
Christoph Lameter committed
755
756
static int slab_pad_check(struct kmem_cache *s, struct page *page)
{
757
758
759
760
761
	u8 *start;
	u8 *fault;
	u8 *end;
	int length;
	int remainder;
Christoph Lameter's avatar
Christoph Lameter committed
762
763
764
765

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

766
	start = page_address(page);
767
	length = (PAGE_SIZE << compound_order(page)) - s->reserved;
768
769
	end = start + length;
	remainder = length % s->size;
Christoph Lameter's avatar
Christoph Lameter committed
770
771
772
	if (!remainder)
		return 1;

773
	fault = memchr_inv(end - remainder, POISON_INUSE, remainder);
774
775
776
777
778
779
	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);
780
	print_section("Padding ", end - remainder, remainder);
781

Eric Dumazet's avatar
Eric Dumazet committed
782
	restore_bytes(s, "slab padding", POISON_INUSE, end - remainder, end);
783
	return 0;
Christoph Lameter's avatar
Christoph Lameter committed
784
785
786
}

static int check_object(struct kmem_cache *s, struct page *page,
787
					void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
788
789
{
	u8 *p = object;
790
	u8 *endobject = object + s->object_size;
Christoph Lameter's avatar
Christoph Lameter committed
791
792

	if (s->flags & SLAB_RED_ZONE) {
793
		if (!check_bytes_and_report(s, page, object, "Redzone",
794
			endobject, val, s->inuse - s->object_size))
Christoph Lameter's avatar
Christoph Lameter committed
795
796
			return 0;
	} else {
797
		if ((s->flags & SLAB_POISON) && s->object_size < s->inuse) {
Ingo Molnar's avatar
Ingo Molnar committed
798
			check_bytes_and_report(s, page, p, "Alignment padding",
799
800
				endobject, POISON_INUSE,
				s->inuse - s->object_size);
Ingo Molnar's avatar
Ingo Molnar committed
801
		}
Christoph Lameter's avatar
Christoph Lameter committed
802
803
804
	}

	if (s->flags & SLAB_POISON) {
805
		if (val != SLUB_RED_ACTIVE && (s->flags & __OBJECT_POISON) &&
806
			(!check_bytes_and_report(s, page, p, "Poison", p,
807
					POISON_FREE, s->object_size - 1) ||
808
			 !check_bytes_and_report(s, page, p, "Poison",
809
				p + s->object_size - 1, POISON_END, 1)))
Christoph Lameter's avatar
Christoph Lameter committed
810
811
812
813
814
815
816
			return 0;
		/*
		 * check_pad_bytes cleans up on its own.
		 */
		check_pad_bytes(s, page, p);
	}

817
	if (!s->offset && val == SLUB_RED_ACTIVE)
Christoph Lameter's avatar
Christoph Lameter committed
818
819
820
821
822
823
824
825
826
827
		/*
		 * 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
828
		 * No choice but to zap it and thus lose the remainder
Christoph Lameter's avatar
Christoph Lameter committed
829
		 * of the free objects in this slab. May cause
Christoph Lameter's avatar
Christoph Lameter committed
830
		 * another error because the object count is now wrong.
Christoph Lameter's avatar
Christoph Lameter committed
831
		 */
832
		set_freepointer(s, p, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
833
834
835
836
837
838
839
		return 0;
	}
	return 1;
}

static int check_slab(struct kmem_cache *s, struct page *page)
{
840
841
	int maxobj;

Christoph Lameter's avatar
Christoph Lameter committed
842
843
844
	VM_BUG_ON(!irqs_disabled());

	if (!PageSlab(page)) {
845
		slab_err(s, page, "Not a valid slab page");
Christoph Lameter's avatar
Christoph Lameter committed
846
847
		return 0;
	}
848

849
	maxobj = order_objects(compound_order(page), s->size, s->reserved);
850
851
	if (page->objects > maxobj) {
		slab_err(s, page, "objects %u > max %u",
852
			page->objects, maxobj);
853
854
855
		return 0;
	}
	if (page->inuse > page->objects) {
856
		slab_err(s, page, "inuse %u > max %u",
857
			page->inuse, page->objects);
Christoph Lameter's avatar
Christoph Lameter committed
858
859
860
861
862
863
864
865
		return 0;
	}
	/* Slab_pad_check fixes things up after itself */
	slab_pad_check(s, page);
	return 1;
}

/*
Christoph Lameter's avatar
Christoph Lameter committed
866
867
 * 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
868
869
870
871
 */
static int on_freelist(struct kmem_cache *s, struct page *page, void *search)
{
	int nr = 0;
872
	void *fp;
Christoph Lameter's avatar
Christoph Lameter committed
873
	void *object = NULL;
874
	int max_objects;
Christoph Lameter's avatar
Christoph Lameter committed
875

876
	fp = page->freelist;
877
	while (fp && nr <= page->objects) {
Christoph Lameter's avatar
Christoph Lameter committed
878
879
880
881
882
883
		if (fp == search)
			return 1;
		if (!check_valid_pointer(s, page, fp)) {
			if (object) {
				object_err(s, page, object,
					"Freechain corrupt");
884
				set_freepointer(s, object, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
885
			} else {
886
				slab_err(s, page, "Freepointer corrupt");
887
				page->freelist = NULL;
888
				page->inuse = page->objects;
889
				slab_fix(s, "Freelist cleared");
Christoph Lameter's avatar
Christoph Lameter committed
890
891
892
893
894
895
896
897
898
				return 0;
			}
			break;
		}
		object = fp;
		fp = get_freepointer(s, object);
		nr++;
	}

899
	max_objects = order_objects(compound_order(page), s->size, s->reserved);
900
901
	if (max_objects > MAX_OBJS_PER_PAGE)
		max_objects = MAX_OBJS_PER_PAGE;
902
903
904
905
906
907
908

	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.");
	}
909
	if (page->inuse != page->objects - nr) {
910
		slab_err(s, page, "Wrong object count. Counter is %d but "
911
912
			"counted were %d", page->inuse, page->objects - nr);
		page->inuse = page->objects - nr;
913
		slab_fix(s, "Object count adjusted.");
Christoph Lameter's avatar
Christoph Lameter committed
914
915
916
917
	}
	return search == NULL;
}

918
919
static void trace(struct kmem_cache *s, struct page *page, void *object,
								int alloc)
Christoph Lameter's avatar
Christoph Lameter committed
920
921
{
	if (s->flags & SLAB_TRACE) {
922
		pr_info("TRACE %s %s 0x%p inuse=%d fp=0x%p\n",
Christoph Lameter's avatar
Christoph Lameter committed
923
924
925
926
927
928
			s->name,
			alloc ? "alloc" : "free",
			object, page->inuse,
			page->freelist);

		if (!alloc)
929
930
			print_section("Object ", (void *)object,
					s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
931
932
933
934
935

		dump_stack();
	}
}

936
/*
Christoph Lameter's avatar
Christoph Lameter committed
937
 * Tracking of fully allocated slabs for debugging purposes.
938
 */
939
940
static void add_full(struct kmem_cache *s,
	struct kmem_cache_node *n, struct page *page)
941
{
942
943
944
	if (!(s->flags & SLAB_STORE_USER))
		return;

945
	lockdep_assert_held(&n->list_lock);
946
947
948
	list_add(&page->lru, &n->full);
}

Peter Zijlstra's avatar
Peter Zijlstra committed
949
static void remove_full(struct kmem_cache *s, struct kmem_cache_node *n, struct page *page)
950
951
952
953
{
	if (!(s->flags & SLAB_STORE_USER))
		return;

954
	lockdep_assert_held(&n->list_lock);
955
956
957
	list_del(&page->lru);
}

958
959
960
961
962
963
964
965
/* 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);
}

966
967
968
969
970
static inline unsigned long node_nr_slabs(struct kmem_cache_node *n)
{
	return atomic_long_read(&n->nr_slabs);
}

971
static inline void inc_slabs_node(struct kmem_cache *s, int node, int objects)
972
973
974
975
976
977
978
979
980
{
	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).
	 */
981
	if (likely(n)) {
982
		atomic_long_inc(&n->nr_slabs);
983
984
		atomic_long_add(objects, &n->total_objects);
	}
985
}
986
static inline void dec_slabs_node(struct kmem_cache *s, int node, int objects)
987
988
989
990
{
	struct kmem_cache_node *n = get_node(s, node);

	atomic_long_dec(&n->nr_slabs);
991
	atomic_long_sub(objects, &n->total_objects);
992
993
994
}

/* Object debug checks for alloc/free paths */
Christoph Lameter's avatar
Christoph Lameter committed
995
996
997
998
999
1000
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;

1001
	init_object(s, object, SLUB_RED_INACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1002
1003
1004
	init_tracking(s, object);
}

1005
1006
static noinline int alloc_debug_processing(struct kmem_cache *s,
					struct page *page,
1007
					void *object, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
1008
1009
1010
1011
1012
1013
{
	if (!check_slab(s, page))
		goto bad;

	if (!check_valid_pointer(s, page, object)) {
		object_err(s, page, object, "Freelist Pointer check fails");
1014
		goto bad;
Christoph Lameter's avatar
Christoph Lameter committed
1015
1016
	}

1017
	if (!check_object(s, page, object, SLUB_RED_INACTIVE))
Christoph Lameter's avatar
Christoph Lameter committed
1018
1019
		goto bad;

Christoph Lameter's avatar
Christoph Lameter committed
1020
1021
1022
1023
	/* 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);
1024
	init_object(s, object, SLUB_RED_ACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1025
	return 1;
Christoph Lameter's avatar
Christoph Lameter committed
1026

Christoph Lameter's avatar
Christoph Lameter committed
1027
1028
1029
1030
1031
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
1032
		 * as used avoids touching the remaining objects.
Christoph Lameter's avatar
Christoph Lameter committed
1033
		 */
1034
		slab_fix(s, "Marking all objects used");
1035
		page->inuse = page->objects;
1036
		page->freelist = NULL;
Christoph Lameter's avatar
Christoph Lameter committed
1037
1038
1039
1040
	}
	return 0;
}

1041
1042
1043
static noinline struct kmem_cache_node *free_debug_processing(
	struct kmem_cache *s, struct page *page, void *object,
	unsigned long addr, unsigned long *flags)
Christoph Lameter's avatar
Christoph Lameter committed
1044
{
1045
	struct kmem_cache_node *n = get_node(s, page_to_nid(page));
1046

1047
	spin_lock_irqsave(&n->list_lock, *flags);
1048
1049
	slab_lock(page);

Christoph Lameter's avatar
Christoph Lameter committed
1050
1051
1052
1053
	if (!check_slab(s, page))
		goto fail;

	if (!check_valid_pointer(s, page, object)) {
1054
		slab_err(s, page, "Invalid object pointer 0x%p", object);
Christoph Lameter's avatar
Christoph Lameter committed
1055
1056
1057
1058
		goto fail;
	}

	if (on_freelist(s, page, object)) {
1059
		object_err(s, page, object, "Object already free");
Christoph Lameter's avatar
Christoph Lameter committed
1060
1061
1062
		goto fail;
	}

1063
	if (!check_object(s, page, object, SLUB_RED_ACTIVE))
1064
		goto out;
Christoph Lameter's avatar
Christoph Lameter committed
1065

1066
	if (unlikely(s != page->slab_cache)) {
Ingo Molnar's avatar
Ingo Molnar committed
1067
		if (!PageSlab(page)) {
1068
1069
			slab_err(s, page, "Attempt to free object(0x%p) "
				"outside of slab", object);
1070
		} else if (!page->slab_cache) {
1071
1072
			pr_err("SLUB <none>: no slab for object 0x%p.\n",
			       object);
1073
			dump_stack();
1074
		} else
1075
1076
			object_err(s, page, object,
					"page slab pointer corrupt.");
Christoph Lameter's avatar
Christoph Lameter committed
1077
1078
		goto fail;
	}
Christoph Lameter's avatar
Christoph Lameter committed
1079
1080
1081
1082

	if (s->flags & SLAB_STORE_USER)
		set_track(s, object, TRACK_FREE, addr);
	trace(s, page, object, 0);
1083
	init_object(s, object, SLUB_RED_INACTIVE);
1084
out:
1085
	slab_unlock(page);
1086
1087
1088
1089
1090
	/*
	 * Keep node_lock to preserve integrity
	 * until the object is actually freed
	 */
	return n;
Christoph Lameter's avatar
Christoph Lameter committed
1091

Christoph Lameter's avatar
Christoph Lameter committed
1092
fail:
1093
1094
	slab_unlock(page);
	spin_unlock_irqrestore(&n->list_lock, *flags);
1095
	slab_fix(s, "Object at 0x%p not freed", object);
1096
	return NULL;
Christoph Lameter's avatar
Christoph Lameter committed
1097
1098
}

1099
1100
static int __init setup_slub_debug(char *str)
{
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
	slub_debug = DEBUG_DEFAULT_FLAGS;
	if (*str++ != '=' || !*str)
		/*
		 * No options specified. Switch on full debugging.
		 */
		goto out;

	if (*str == ',')
		/*
		 * No options but restriction on slabs. This means full
		 * debugging for slabs matching a pattern.
		 */
		goto check_slabs;

1115
1116
1117
1118
1119
1120
1121
1122
1123
	if (tolower(*str) == 'o') {
		/*
		 * Avoid enabling debugging on caches if its minimum order
		 * would increase as a result.
		 */
		disable_higher_order_debug = 1;
		goto out;
	}

1124
1125
1126
1127
1128
1129
1130
1131