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

Christoph Lameter's avatar
Christoph Lameter committed
172
173
174
175
/*
 * Set of flags that will prevent slab merging
 */
#define SLUB_NEVER_MERGE (SLAB_RED_ZONE | SLAB_POISON | SLAB_STORE_USER | \
176
177
		SLAB_TRACE | SLAB_DESTROY_BY_RCU | SLAB_NOLEAKTRACE | \
		SLAB_FAILSLAB)
Christoph Lameter's avatar
Christoph Lameter committed
178
179

#define SLUB_MERGE_SAME (SLAB_DEBUG_FREE | SLAB_RECLAIM_ACCOUNT | \
Vegard Nossum's avatar
Vegard Nossum committed
180
		SLAB_CACHE_DMA | SLAB_NOTRACK)
Christoph Lameter's avatar
Christoph Lameter committed
181

182
183
#define OO_SHIFT	16
#define OO_MASK		((1 << OO_SHIFT) - 1)
184
#define MAX_OBJS_PER_PAGE	32767 /* since page.objects is u15 */
185

Christoph Lameter's avatar
Christoph Lameter committed
186
/* Internal SLUB flags */
Christoph Lameter's avatar
Christoph Lameter committed
187
#define __OBJECT_POISON		0x80000000UL /* Poison object */
188
#define __CMPXCHG_DOUBLE	0x40000000UL /* Use cmpxchg_double */
Christoph Lameter's avatar
Christoph Lameter committed
189
190
191
192
193

#ifdef CONFIG_SMP
static struct notifier_block slab_notifier;
#endif

194
195
196
/*
 * Tracking user of a slab.
 */
197
#define TRACK_ADDRS_COUNT 16
198
struct track {
199
	unsigned long addr;	/* Called from address */
200
201
202
#ifdef CONFIG_STACKTRACE
	unsigned long addrs[TRACK_ADDRS_COUNT];	/* Called from address */
#endif
203
204
205
206
207
208
209
	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 };

210
#ifdef CONFIG_SYSFS
Christoph Lameter's avatar
Christoph Lameter committed
211
212
static int sysfs_slab_add(struct kmem_cache *);
static int sysfs_slab_alias(struct kmem_cache *, const char *);
213
static void memcg_propagate_slab_attrs(struct kmem_cache *s);
Christoph Lameter's avatar
Christoph Lameter committed
214
#else
215
216
217
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; }
218
static inline void memcg_propagate_slab_attrs(struct kmem_cache *s) { }
Christoph Lameter's avatar
Christoph Lameter committed
219
220
#endif

221
static inline void stat(const struct kmem_cache *s, enum stat_item si)
222
223
{
#ifdef CONFIG_SLUB_STATS
224
225
226
227
228
	/*
	 * 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]);
229
230
231
#endif
}

Christoph Lameter's avatar
Christoph Lameter committed
232
233
234
235
/********************************************************************
 * 			Core slab cache functions
 *******************************************************************/

Christoph Lameter's avatar
Christoph Lameter committed
236
/* Verify that a pointer has an address that is valid within a slab page */
237
238
239
240
241
static inline int check_valid_pointer(struct kmem_cache *s,
				struct page *page, const void *object)
{
	void *base;

242
	if (!object)
243
244
		return 1;

245
	base = page_address(page);
246
	if (object < base || object >= base + page->objects * s->size ||
247
248
249
250
251
252
253
		(object - base) % s->size) {
		return 0;
	}

	return 1;
}

254
255
256
257
258
static inline void *get_freepointer(struct kmem_cache *s, void *object)
{
	return *(void **)(object + s->offset);
}

259
260
261
262
263
static void prefetch_freepointer(const struct kmem_cache *s, void *object)
{
	prefetch(object + s->offset);
}

264
265
266
267
268
269
270
271
272
273
274
275
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;
}

276
277
278
279
280
281
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 */
282
283
#define for_each_object(__p, __s, __addr, __objects) \
	for (__p = (__addr); __p < (__addr) + (__objects) * (__s)->size;\
284
285
			__p += (__s)->size)

286
287
288
289
#define for_each_object_idx(__p, __idx, __s, __addr, __objects) \
	for (__p = (__addr), __idx = 1; __idx <= __objects;\
			__p += (__s)->size, __idx++)

290
291
292
293
294
295
/* 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;
}

296
297
298
299
300
301
302
303
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))
304
		return s->object_size;
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319

#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;
}

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

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

	return x;
}

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

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

345
346
347
348
349
350
351
352
353
354
355
356
357
/*
 * 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);
}

358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
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;
}

373
374
375
376
377
378
379
/* 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());
380
381
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
382
	if (s->flags & __CMPXCHG_DOUBLE) {
383
		if (cmpxchg_double(&page->freelist, &page->counters,
384
385
386
387
388
389
390
			freelist_old, counters_old,
			freelist_new, counters_new))
		return 1;
	} else
#endif
	{
		slab_lock(page);
391
392
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
393
			page->freelist = freelist_new;
394
			set_page_slub_counters(page, counters_new);
395
396
397
398
399
400
401
402
403
404
			slab_unlock(page);
			return 1;
		}
		slab_unlock(page);
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
405
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
406
407
408
409
410
#endif

	return 0;
}

411
412
413
414
415
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)
{
416
417
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
418
	if (s->flags & __CMPXCHG_DOUBLE) {
419
		if (cmpxchg_double(&page->freelist, &page->counters,
420
421
422
423
424
425
			freelist_old, counters_old,
			freelist_new, counters_new))
		return 1;
	} else
#endif
	{
426
427
428
		unsigned long flags;

		local_irq_save(flags);
429
		slab_lock(page);
430
431
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
432
			page->freelist = freelist_new;
433
			set_page_slub_counters(page, counters_new);
434
			slab_unlock(page);
435
			local_irq_restore(flags);
436
437
			return 1;
		}
438
		slab_unlock(page);
439
		local_irq_restore(flags);
440
441
442
443
444
445
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
446
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
447
448
449
450
451
#endif

	return 0;
}

452
#ifdef CONFIG_SLUB_DEBUG
453
454
455
/*
 * Determine a map of object in use on a page.
 *
456
 * Node listlock must be held to guarantee that the page does
457
458
459
460
461
462
463
464
465
466
467
 * 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);
}

468
469
470
/*
 * Debug settings:
 */
471
472
473
#ifdef CONFIG_SLUB_DEBUG_ON
static int slub_debug = DEBUG_DEFAULT_FLAGS;
#else
474
static int slub_debug;
475
#endif
476
477

static char *slub_debug_slabs;
478
static int disable_higher_order_debug;
479

Christoph Lameter's avatar
Christoph Lameter committed
480
481
482
483
484
/*
 * Object debugging
 */
static void print_section(char *text, u8 *addr, unsigned int length)
{
485
486
	print_hex_dump(KERN_ERR, text, DUMP_PREFIX_ADDRESS, 16, 1, addr,
			length, 1);
Christoph Lameter's avatar
Christoph Lameter committed
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
}

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,
503
			enum track_item alloc, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
504
{
Akinobu Mita's avatar
Akinobu Mita committed
505
	struct track *p = get_track(s, object, alloc);
Christoph Lameter's avatar
Christoph Lameter committed
506
507

	if (addr) {
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
#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
526
527
		p->addr = addr;
		p->cpu = smp_processor_id();
528
		p->pid = current->pid;
Christoph Lameter's avatar
Christoph Lameter committed
529
530
531
532
533
534
535
		p->when = jiffies;
	} else
		memset(p, 0, sizeof(struct track));
}

static void init_tracking(struct kmem_cache *s, void *object)
{
536
537
538
	if (!(s->flags & SLAB_STORE_USER))
		return;

539
540
	set_track(s, object, TRACK_FREE, 0UL);
	set_track(s, object, TRACK_ALLOC, 0UL);
Christoph Lameter's avatar
Christoph Lameter committed
541
542
543
544
545
546
547
}

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

548
549
	pr_err("INFO: %s in %pS age=%lu cpu=%u pid=%d\n",
	       s, (void *)t->addr, jiffies - t->when, t->cpu, t->pid);
550
551
552
553
554
#ifdef CONFIG_STACKTRACE
	{
		int i;
		for (i = 0; i < TRACK_ADDRS_COUNT; i++)
			if (t->addrs[i])
555
				pr_err("\t%pS\n", (void *)t->addrs[i]);
556
557
558
559
			else
				break;
	}
#endif
560
561
562
563
564
565
566
567
568
569
570
571
572
}

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)
{
573
	pr_err("INFO: Slab 0x%p objects=%u used=%u fp=0x%p flags=0x%04lx\n",
574
	       page, page->objects, page->inuse, page->freelist, page->flags);
575
576
577
578
579

}

static void slab_bug(struct kmem_cache *s, char *fmt, ...)
{
580
	struct va_format vaf;
581
582
583
	va_list args;

	va_start(args, fmt);
584
585
	vaf.fmt = fmt;
	vaf.va = &args;
586
	pr_err("=============================================================================\n");
587
	pr_err("BUG %s (%s): %pV\n", s->name, print_tainted(), &vaf);
588
	pr_err("-----------------------------------------------------------------------------\n\n");
589

590
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
591
	va_end(args);
Christoph Lameter's avatar
Christoph Lameter committed
592
593
}

594
595
static void slab_fix(struct kmem_cache *s, char *fmt, ...)
{
596
	struct va_format vaf;
597
598
599
	va_list args;

	va_start(args, fmt);
600
601
602
	vaf.fmt = fmt;
	vaf.va = &args;
	pr_err("FIX %s: %pV\n", s->name, &vaf);
603
604
605
606
	va_end(args);
}

static void print_trailer(struct kmem_cache *s, struct page *page, u8 *p)
Christoph Lameter's avatar
Christoph Lameter committed
607
608
{
	unsigned int off;	/* Offset of last byte */
609
	u8 *addr = page_address(page);
610
611
612
613
614

	print_tracking(s, p);

	print_page_info(page);

615
616
	pr_err("INFO: Object 0x%p @offset=%tu fp=0x%p\n\n",
	       p, p - addr, get_freepointer(s, p));
617
618

	if (p > addr + 16)
619
		print_section("Bytes b4 ", p - 16, 16);
Christoph Lameter's avatar
Christoph Lameter committed
620

621
	print_section("Object ", p, min_t(unsigned long, s->object_size,
622
				PAGE_SIZE));
Christoph Lameter's avatar
Christoph Lameter committed
623
	if (s->flags & SLAB_RED_ZONE)
624
625
		print_section("Redzone ", p + s->object_size,
			s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
626
627
628
629
630
631

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

632
	if (s->flags & SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
633
634
635
636
		off += 2 * sizeof(struct track);

	if (off != s->size)
		/* Beginning of the filler is the free pointer */
637
		print_section("Padding ", p + off, s->size - off);
638
639

	dump_stack();
Christoph Lameter's avatar
Christoph Lameter committed
640
641
642
643
644
}

static void object_err(struct kmem_cache *s, struct page *page,
			u8 *object, char *reason)
{
645
	slab_bug(s, "%s", reason);
646
	print_trailer(s, page, object);
Christoph Lameter's avatar
Christoph Lameter committed
647
648
}

649
650
static void slab_err(struct kmem_cache *s, struct page *page,
			const char *fmt, ...)
Christoph Lameter's avatar
Christoph Lameter committed
651
652
653
654
{
	va_list args;
	char buf[100];

655
656
	va_start(args, fmt);
	vsnprintf(buf, sizeof(buf), fmt, args);
Christoph Lameter's avatar
Christoph Lameter committed
657
	va_end(args);
658
	slab_bug(s, "%s", buf);
659
	print_page_info(page);
Christoph Lameter's avatar
Christoph Lameter committed
660
661
662
	dump_stack();
}

663
static void init_object(struct kmem_cache *s, void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
664
665
666
667
{
	u8 *p = object;

	if (s->flags & __OBJECT_POISON) {
668
669
		memset(p, POISON_FREE, s->object_size - 1);
		p[s->object_size - 1] = POISON_END;
Christoph Lameter's avatar
Christoph Lameter committed
670
671
672
	}

	if (s->flags & SLAB_RED_ZONE)
673
		memset(p + s->object_size, val, s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
674
675
}

676
677
678
679
680
681
682
683
684
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,
685
			u8 *start, unsigned int value, unsigned int bytes)
686
687
688
689
{
	u8 *fault;
	u8 *end;

690
	fault = memchr_inv(start, value, bytes);
691
692
693
694
695
696
697
698
	if (!fault)
		return 1;

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

	slab_bug(s, "%s overwritten", what);
699
	pr_err("INFO: 0x%p-0x%p. First byte 0x%x instead of 0x%x\n",
700
701
702
703
704
					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
705
706
707
708
709
710
711
712
713
}

/*
 * 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
714
 *
Christoph Lameter's avatar
Christoph Lameter committed
715
716
717
 * 	Poisoning uses 0x6b (POISON_FREE) and the last byte is
 * 	0xa5 (POISON_END)
 *
718
 * object + s->object_size
Christoph Lameter's avatar
Christoph Lameter committed
719
 * 	Padding to reach word boundary. This is also used for Redzoning.
Christoph Lameter's avatar
Christoph Lameter committed
720
 * 	Padding is extended by another word if Redzoning is enabled and
721
 * 	object_size == inuse.
Christoph Lameter's avatar
Christoph Lameter committed
722
 *
Christoph Lameter's avatar
Christoph Lameter committed
723
724
725
726
 * 	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
727
728
 * 	Meta data starts here.
 *
Christoph Lameter's avatar
Christoph Lameter committed
729
730
 * 	A. Free pointer (if we cannot overwrite object on free)
 * 	B. Tracking data for SLAB_STORE_USER
Christoph Lameter's avatar
Christoph Lameter committed
731
 * 	C. Padding to reach required alignment boundary or at mininum
Christoph Lameter's avatar
Christoph Lameter committed
732
 * 		one word if debugging is on to be able to detect writes
Christoph Lameter's avatar
Christoph Lameter committed
733
734
735
 * 		before the word boundary.
 *
 *	Padding is done using 0x5a (POISON_INUSE)
Christoph Lameter's avatar
Christoph Lameter committed
736
737
 *
 * object + s->size
Christoph Lameter's avatar
Christoph Lameter committed
738
 * 	Nothing is used beyond s->size.
Christoph Lameter's avatar
Christoph Lameter committed
739
 *
740
 * If slabcaches are merged then the object_size and inuse boundaries are mostly
Christoph Lameter's avatar
Christoph Lameter committed
741
 * ignored. And therefore no slab options that rely on these boundaries
Christoph Lameter's avatar
Christoph Lameter committed
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
 * 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;

760
761
	return check_bytes_and_report(s, page, p, "Object padding",
				p + off, POISON_INUSE, s->size - off);
Christoph Lameter's avatar
Christoph Lameter committed
762
763
}

764
/* Check the pad bytes at the end of a slab page */
Christoph Lameter's avatar
Christoph Lameter committed
765
766
static int slab_pad_check(struct kmem_cache *s, struct page *page)
{
767
768
769
770
771
	u8 *start;
	u8 *fault;
	u8 *end;
	int length;
	int remainder;
Christoph Lameter's avatar
Christoph Lameter committed
772
773
774
775

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

776
	start = page_address(page);
777
	length = (PAGE_SIZE << compound_order(page)) - s->reserved;
778
779
	end = start + length;
	remainder = length % s->size;
Christoph Lameter's avatar
Christoph Lameter committed
780
781
782
	if (!remainder)
		return 1;

783
	fault = memchr_inv(end - remainder, POISON_INUSE, remainder);
784
785
786
787
788
789
	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);
790
	print_section("Padding ", end - remainder, remainder);
791

Eric Dumazet's avatar
Eric Dumazet committed
792
	restore_bytes(s, "slab padding", POISON_INUSE, end - remainder, end);
793
	return 0;
Christoph Lameter's avatar
Christoph Lameter committed
794
795
796
}

static int check_object(struct kmem_cache *s, struct page *page,
797
					void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
798
799
{
	u8 *p = object;
800
	u8 *endobject = object + s->object_size;
Christoph Lameter's avatar
Christoph Lameter committed
801
802

	if (s->flags & SLAB_RED_ZONE) {
803
		if (!check_bytes_and_report(s, page, object, "Redzone",
804
			endobject, val, s->inuse - s->object_size))
Christoph Lameter's avatar
Christoph Lameter committed
805
806
			return 0;
	} else {
807
		if ((s->flags & SLAB_POISON) && s->object_size < s->inuse) {
Ingo Molnar's avatar
Ingo Molnar committed
808
			check_bytes_and_report(s, page, p, "Alignment padding",
809
810
				endobject, POISON_INUSE,
				s->inuse - s->object_size);
Ingo Molnar's avatar
Ingo Molnar committed
811
		}
Christoph Lameter's avatar
Christoph Lameter committed
812
813
814
	}

	if (s->flags & SLAB_POISON) {
815
		if (val != SLUB_RED_ACTIVE && (s->flags & __OBJECT_POISON) &&
816
			(!check_bytes_and_report(s, page, p, "Poison", p,
817
					POISON_FREE, s->object_size - 1) ||
818
			 !check_bytes_and_report(s, page, p, "Poison",
819
				p + s->object_size - 1, POISON_END, 1)))
Christoph Lameter's avatar
Christoph Lameter committed
820
821
822
823
824
825
826
			return 0;
		/*
		 * check_pad_bytes cleans up on its own.
		 */
		check_pad_bytes(s, page, p);
	}

827
	if (!s->offset && val == SLUB_RED_ACTIVE)
Christoph Lameter's avatar
Christoph Lameter committed
828
829
830
831
832
833
834
835
836
837
		/*
		 * 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
838
		 * No choice but to zap it and thus lose the remainder
Christoph Lameter's avatar
Christoph Lameter committed
839
		 * of the free objects in this slab. May cause
Christoph Lameter's avatar
Christoph Lameter committed
840
		 * another error because the object count is now wrong.
Christoph Lameter's avatar
Christoph Lameter committed
841
		 */
842
		set_freepointer(s, p, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
843
844
845
846
847
848
849
		return 0;
	}
	return 1;
}

static int check_slab(struct kmem_cache *s, struct page *page)
{
850
851
	int maxobj;

Christoph Lameter's avatar
Christoph Lameter committed
852
853
854
	VM_BUG_ON(!irqs_disabled());

	if (!PageSlab(page)) {
855
		slab_err(s, page, "Not a valid slab page");
Christoph Lameter's avatar
Christoph Lameter committed
856
857
		return 0;
	}
858

859
	maxobj = order_objects(compound_order(page), s->size, s->reserved);
860
861
862
863
864
865
	if (page->objects > maxobj) {
		slab_err(s, page, "objects %u > max %u",
			s->name, page->objects, maxobj);
		return 0;
	}
	if (page->inuse > page->objects) {
866
		slab_err(s, page, "inuse %u > max %u",
867
			s->name, page->inuse, page->objects);
Christoph Lameter's avatar
Christoph Lameter committed
868
869
870
871
872
873
874
875
		return 0;
	}
	/* Slab_pad_check fixes things up after itself */
	slab_pad_check(s, page);
	return 1;
}

/*
Christoph Lameter's avatar
Christoph Lameter committed
876
877
 * 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
878
879
880
881
 */
static int on_freelist(struct kmem_cache *s, struct page *page, void *search)
{
	int nr = 0;
882
	void *fp;
Christoph Lameter's avatar
Christoph Lameter committed
883
	void *object = NULL;
884
	unsigned long max_objects;
Christoph Lameter's avatar
Christoph Lameter committed
885

886
	fp = page->freelist;
887
	while (fp && nr <= page->objects) {
Christoph Lameter's avatar
Christoph Lameter committed
888
889
890
891
892
893
		if (fp == search)
			return 1;
		if (!check_valid_pointer(s, page, fp)) {
			if (object) {
				object_err(s, page, object,
					"Freechain corrupt");
894
				set_freepointer(s, object, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
895
			} else {
896
				slab_err(s, page, "Freepointer corrupt");
897
				page->freelist = NULL;
898
				page->inuse = page->objects;
899
				slab_fix(s, "Freelist cleared");
Christoph Lameter's avatar
Christoph Lameter committed
900
901
902
903
904
905
906
907
908
				return 0;
			}
			break;
		}
		object = fp;
		fp = get_freepointer(s, object);
		nr++;
	}

909
	max_objects = order_objects(compound_order(page), s->size, s->reserved);
910
911
	if (max_objects > MAX_OBJS_PER_PAGE)
		max_objects = MAX_OBJS_PER_PAGE;
912
913
914
915
916
917
918

	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.");
	}
919
	if (page->inuse != page->objects - nr) {
920
		slab_err(s, page, "Wrong object count. Counter is %d but "
921
922
			"counted were %d", page->inuse, page->objects - nr);
		page->inuse = page->objects - nr;
923
		slab_fix(s, "Object count adjusted.");
Christoph Lameter's avatar
Christoph Lameter committed
924
925
926
927
	}
	return search == NULL;
}

928
929
static void trace(struct kmem_cache *s, struct page *page, void *object,
								int alloc)
Christoph Lameter's avatar
Christoph Lameter committed
930
931
{
	if (s->flags & SLAB_TRACE) {
932
		pr_info("TRACE %s %s 0x%p inuse=%d fp=0x%p\n",
Christoph Lameter's avatar
Christoph Lameter committed
933
934
935
936
937
938
			s->name,
			alloc ? "alloc" : "free",
			object, page->inuse,
			page->freelist);

		if (!alloc)
939
940
			print_section("Object ", (void *)object,
					s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
941
942
943
944
945

		dump_stack();
	}
}

946
/*
Christoph Lameter's avatar
Christoph Lameter committed
947
 * Tracking of fully allocated slabs for debugging purposes.
948
 */
949
950
static void add_full(struct kmem_cache *s,
	struct kmem_cache_node *n, struct page *page)
951
{
952
953
954
	if (!(s->flags & SLAB_STORE_USER))
		return;

955
	lockdep_assert_held(&n->list_lock);
956
957
958
	list_add(&page->lru, &n->full);
}

Peter Zijlstra's avatar
Peter Zijlstra committed
959
static void remove_full(struct kmem_cache *s, struct kmem_cache_node *n, struct page *page)
960
961
962
963
{
	if (!(s->flags & SLAB_STORE_USER))
		return;

964
	lockdep_assert_held(&n->list_lock);
965
966
967
	list_del(&page->lru);
}

968
969
970
971
972
973
974
975
/* 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);
}

976
977
978
979
980
static inline unsigned long node_nr_slabs(struct kmem_cache_node *n)
{
	return atomic_long_read(&n->nr_slabs);
}

981
static inline void inc_slabs_node(struct kmem_cache *s, int node, int objects)
982
983
984
985
986
987
988
989
990
{
	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).
	 */
991
	if (likely(n)) {
992
		atomic_long_inc(&n->nr_slabs);
993
994
		atomic_long_add(objects, &n->total_objects);
	}
995
}
996
static inline void dec_slabs_node(struct kmem_cache *s, int node, int objects)
997
998
999
1000
{
	struct kmem_cache_node *n = get_node(s, node);

	atomic_long_dec(&n->nr_slabs);
1001
	atomic_long_sub(objects, &n->total_objects);
1002
1003
1004
}

/* Object debug checks for alloc/free paths */
Christoph Lameter's avatar
Christoph Lameter committed
1005
1006
1007
1008
1009
1010
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;

1011
	init_object(s, object, SLUB_RED_INACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1012
1013
1014
	init_tracking(s, object);
}

1015
1016
static noinline int alloc_debug_processing(struct kmem_cache *s,
					struct page *page,
1017
					void *object, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
1018
1019
1020
1021
1022
1023
{
	if (!check_slab(s, page))
		goto bad;

	if (!check_valid_pointer(s, page, object)) {
		object_err(s, page, object, "Freelist Pointer check fails");
1024
		goto bad;
Christoph Lameter's avatar
Christoph Lameter committed
1025
1026
	}

1027
	if (!check_object(s, page, object, SLUB_RED_INACTIVE))
Christoph Lameter's avatar
Christoph Lameter committed
1028
1029
		goto bad;

Christoph Lameter's avatar
Christoph Lameter committed
1030
1031
1032
1033
	/* 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);
1034
	init_object(s, object, SLUB_RED_ACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1035
	return 1;
Christoph Lameter's avatar
Christoph Lameter committed
1036

Christoph Lameter's avatar
Christoph Lameter committed
1037
1038
1039
1040
1041
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
1042
		 * as used avoids touching the remaining objects.
Christoph Lameter's avatar
Christoph Lameter committed
1043
		 */
1044
		slab_fix(s, "Marking all objects used");
1045
		page->inuse = page->objects;
1046
		page->freelist = NULL;
Christoph Lameter's avatar
Christoph Lameter committed
1047
1048
1049
1050
	}
	return 0;
}

1051
1052
1053
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
1054
{
1055
	struct kmem_cache_node *n = get_node(s, page_to_nid(page));
1056

1057
	spin_lock_irqsave(&n->list_lock, *flags);
1058
1059
	slab_lock(page);

Christoph Lameter's avatar
Christoph Lameter committed
1060
1061
1062
1063
	if (!check_slab(s, page))
		goto fail;

	if (!check_valid_pointer(s, page, object)) {
1064
		slab_err(s, page, "Invalid object pointer 0x%p", object);
Christoph Lameter's avatar
Christoph Lameter committed
1065
1066
1067
1068
		goto fail;
	}

	if (on_freelist(s, page, object)) {
1069
		object_err(s, page, object, "Object already free");
Christoph Lameter's avatar
Christoph Lameter committed
1070
1071
1072
		goto fail;
	}

1073
	if (!check_object(s, page, object, SLUB_RED_ACTIVE))
1074
		goto out;
Christoph Lameter's avatar
Christoph Lameter committed
1075

1076
	if (unlikely(s != page->slab_cache)) {
Ingo Molnar's avatar
Ingo Molnar committed
1077
		if (!PageSlab(page)) {
1078
1079
			slab_err(s, page, "Attempt to free object(0x%p) "
				"outside of slab", object);
1080
		} else if (!page->slab_cache) {
1081
1082
			pr_err("SLUB <none>: no slab for object 0x%p.\n",
			       object);
1083
			dump_stack();
1084
		} else
1085
1086
			object_err(s, page, object,
					"page slab pointer corrupt.");
Christoph Lameter's avatar
Christoph Lameter committed
1087
1088
		goto fail;
	}
Christoph Lameter's avatar
Christoph Lameter committed
1089
1090
1091
1092

	if (s->flags & SLAB_STORE_USER)
		set_track(s, object, TRACK_FREE, addr);
	trace(s, page, object, 0);
1093
	init_object(s, object, SLUB_RED_INACTIVE);
1094
out:
1095
	slab_unlock(page);
1096
1097
1098
1099
1100
	/*
	 * Keep node_lock to preserve integrity
	 * until the object is actually freed
	 */
	return n;
Christoph Lameter's avatar
Christoph Lameter committed
1101

Christoph Lameter's avatar
Christoph Lameter committed
1102
fail:
1103
1104
	slab_unlock(page);
	spin_unlock_irqrestore(&n->list_lock, *flags);
1105
	slab_fix(s, "Object at 0x%p not freed", object);
1106
	return NULL;
Christoph Lameter's avatar
Christoph Lameter committed
1107
1108
}

1109
1110
static int __init setup_slub_debug(char *str)
{
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
	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;

1125
1126
1127
1128
1129
1130
1131
1132
1133
	if (tolower(*str) == 'o') {
		/*
		 * Avoid enabling debugging on caches if its minimum order
		 * would increase as a result.
		 */