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

163
#define DEBUG_DEFAULT_FLAGS (SLAB_CONSISTENCY_CHECKS | SLAB_RED_ZONE | \
Christoph Lameter's avatar
Christoph Lameter committed
164
				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
static inline int order_objects(int order, unsigned long size, int reserved)
{
	return ((PAGE_SIZE << order) - reserved) / size;
}

292
static inline struct kmem_cache_order_objects oo_make(int order,
293
		unsigned long size, int reserved)
294
295
{
	struct kmem_cache_order_objects x = {
296
		(order << OO_SHIFT) + order_objects(order, size, reserved)
297
298
299
300
301
302
303
	};

	return x;
}

static inline int oo_order(struct kmem_cache_order_objects x)
{
304
	return x.x >> OO_SHIFT;
305
306
307
308
}

static inline int oo_objects(struct kmem_cache_order_objects x)
{
309
	return x.x & OO_MASK;
310
311
}

312
313
314
315
316
/*
 * Per slab locking using the pagelock
 */
static __always_inline void slab_lock(struct page *page)
{
317
	VM_BUG_ON_PAGE(PageTail(page), page);
318
319
320
321
322
	bit_spin_lock(PG_locked, &page->flags);
}

static __always_inline void slab_unlock(struct page *page)
{
323
	VM_BUG_ON_PAGE(PageTail(page), page);
324
325
326
	__bit_spin_unlock(PG_locked, &page->flags);
}

327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
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;
}

342
343
344
345
346
347
348
/* 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());
349
350
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
351
	if (s->flags & __CMPXCHG_DOUBLE) {
352
		if (cmpxchg_double(&page->freelist, &page->counters,
353
354
				   freelist_old, counters_old,
				   freelist_new, counters_new))
355
			return true;
356
357
358
359
	} else
#endif
	{
		slab_lock(page);
360
361
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
362
			page->freelist = freelist_new;
363
			set_page_slub_counters(page, counters_new);
364
			slab_unlock(page);
365
			return true;
366
367
368
369
370
371
372
373
		}
		slab_unlock(page);
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
374
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
375
376
#endif

377
	return false;
378
379
}

380
381
382
383
384
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)
{
385
386
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
387
	if (s->flags & __CMPXCHG_DOUBLE) {
388
		if (cmpxchg_double(&page->freelist, &page->counters,
389
390
				   freelist_old, counters_old,
				   freelist_new, counters_new))
391
			return true;
392
393
394
	} else
#endif
	{
395
396
397
		unsigned long flags;

		local_irq_save(flags);
398
		slab_lock(page);
399
400
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
401
			page->freelist = freelist_new;
402
			set_page_slub_counters(page, counters_new);
403
			slab_unlock(page);
404
			local_irq_restore(flags);
405
			return true;
406
		}
407
		slab_unlock(page);
408
		local_irq_restore(flags);
409
410
411
412
413
414
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
415
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
416
417
#endif

418
	return false;
419
420
}

421
#ifdef CONFIG_SLUB_DEBUG
422
423
424
/*
 * Determine a map of object in use on a page.
 *
425
 * Node listlock must be held to guarantee that the page does
426
427
428
429
430
431
432
433
434
435
436
 * 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);
}

437
438
439
/*
 * Debug settings:
 */
440
#if defined(CONFIG_SLUB_DEBUG_ON)
441
static int slub_debug = DEBUG_DEFAULT_FLAGS;
442
443
#elif defined(CONFIG_KASAN)
static int slub_debug = SLAB_STORE_USER;
444
#else
445
static int slub_debug;
446
#endif
447
448

static char *slub_debug_slabs;
449
static int disable_higher_order_debug;
450

451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
/*
 * 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
467
468
469
470
471
/*
 * Object debugging
 */
static void print_section(char *text, u8 *addr, unsigned int length)
{
472
	metadata_access_enable();
473
474
	print_hex_dump(KERN_ERR, text, DUMP_PREFIX_ADDRESS, 16, 1, addr,
			length, 1);
475
	metadata_access_disable();
Christoph Lameter's avatar
Christoph Lameter committed
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
}

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,
492
			enum track_item alloc, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
493
{
Akinobu Mita's avatar
Akinobu Mita committed
494
	struct track *p = get_track(s, object, alloc);
Christoph Lameter's avatar
Christoph Lameter committed
495
496

	if (addr) {
497
498
499
500
501
502
503
504
#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;
505
		metadata_access_enable();
506
		save_stack_trace(&trace);
507
		metadata_access_disable();
508
509
510
511
512
513
514
515
516

		/* 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
517
518
		p->addr = addr;
		p->cpu = smp_processor_id();
519
		p->pid = current->pid;
Christoph Lameter's avatar
Christoph Lameter committed
520
521
522
523
524
525
526
		p->when = jiffies;
	} else
		memset(p, 0, sizeof(struct track));
}

static void init_tracking(struct kmem_cache *s, void *object)
{
527
528
529
	if (!(s->flags & SLAB_STORE_USER))
		return;

530
531
	set_track(s, object, TRACK_FREE, 0UL);
	set_track(s, object, TRACK_ALLOC, 0UL);
Christoph Lameter's avatar
Christoph Lameter committed
532
533
534
535
536
537
538
}

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

539
540
	pr_err("INFO: %s in %pS age=%lu cpu=%u pid=%d\n",
	       s, (void *)t->addr, jiffies - t->when, t->cpu, t->pid);
541
542
543
544
545
#ifdef CONFIG_STACKTRACE
	{
		int i;
		for (i = 0; i < TRACK_ADDRS_COUNT; i++)
			if (t->addrs[i])
546
				pr_err("\t%pS\n", (void *)t->addrs[i]);
547
548
549
550
			else
				break;
	}
#endif
551
552
553
554
555
556
557
558
559
560
561
562
563
}

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)
{
564
	pr_err("INFO: Slab 0x%p objects=%u used=%u fp=0x%p flags=0x%04lx\n",
565
	       page, page->objects, page->inuse, page->freelist, page->flags);
566
567
568
569
570

}

static void slab_bug(struct kmem_cache *s, char *fmt, ...)
{
571
	struct va_format vaf;
572
573
574
	va_list args;

	va_start(args, fmt);
575
576
	vaf.fmt = fmt;
	vaf.va = &args;
577
	pr_err("=============================================================================\n");
578
	pr_err("BUG %s (%s): %pV\n", s->name, print_tainted(), &vaf);
579
	pr_err("-----------------------------------------------------------------------------\n\n");
580

581
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
582
	va_end(args);
Christoph Lameter's avatar
Christoph Lameter committed
583
584
}

585
586
static void slab_fix(struct kmem_cache *s, char *fmt, ...)
{
587
	struct va_format vaf;
588
589
590
	va_list args;

	va_start(args, fmt);
591
592
593
	vaf.fmt = fmt;
	vaf.va = &args;
	pr_err("FIX %s: %pV\n", s->name, &vaf);
594
595
596
597
	va_end(args);
}

static void print_trailer(struct kmem_cache *s, struct page *page, u8 *p)
Christoph Lameter's avatar
Christoph Lameter committed
598
599
{
	unsigned int off;	/* Offset of last byte */
600
	u8 *addr = page_address(page);
601
602
603
604
605

	print_tracking(s, p);

	print_page_info(page);

606
607
	pr_err("INFO: Object 0x%p @offset=%tu fp=0x%p\n\n",
	       p, p - addr, get_freepointer(s, p));
608
609

	if (p > addr + 16)
610
		print_section("Bytes b4 ", p - 16, 16);
Christoph Lameter's avatar
Christoph Lameter committed
611

612
	print_section("Object ", p, min_t(unsigned long, s->object_size,
613
				PAGE_SIZE));
Christoph Lameter's avatar
Christoph Lameter committed
614
	if (s->flags & SLAB_RED_ZONE)
615
616
		print_section("Redzone ", p + s->object_size,
			s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
617
618
619
620
621
622

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

623
	if (s->flags & SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
624
625
626
627
		off += 2 * sizeof(struct track);

	if (off != s->size)
		/* Beginning of the filler is the free pointer */
628
		print_section("Padding ", p + off, s->size - off);
629
630

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

633
void object_err(struct kmem_cache *s, struct page *page,
Christoph Lameter's avatar
Christoph Lameter committed
634
635
			u8 *object, char *reason)
{
636
	slab_bug(s, "%s", reason);
637
	print_trailer(s, page, object);
Christoph Lameter's avatar
Christoph Lameter committed
638
639
}

640
641
static void slab_err(struct kmem_cache *s, struct page *page,
			const char *fmt, ...)
Christoph Lameter's avatar
Christoph Lameter committed
642
643
644
645
{
	va_list args;
	char buf[100];

646
647
	va_start(args, fmt);
	vsnprintf(buf, sizeof(buf), fmt, args);
Christoph Lameter's avatar
Christoph Lameter committed
648
	va_end(args);
649
	slab_bug(s, "%s", buf);
650
	print_page_info(page);
Christoph Lameter's avatar
Christoph Lameter committed
651
652
653
	dump_stack();
}

654
static void init_object(struct kmem_cache *s, void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
655
656
657
658
{
	u8 *p = object;

	if (s->flags & __OBJECT_POISON) {
659
660
		memset(p, POISON_FREE, s->object_size - 1);
		p[s->object_size - 1] = POISON_END;
Christoph Lameter's avatar
Christoph Lameter committed
661
662
663
	}

	if (s->flags & SLAB_RED_ZONE)
664
		memset(p + s->object_size, val, s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
665
666
}

667
668
669
670
671
672
673
674
675
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,
676
			u8 *start, unsigned int value, unsigned int bytes)
677
678
679
680
{
	u8 *fault;
	u8 *end;

681
	metadata_access_enable();
682
	fault = memchr_inv(start, value, bytes);
683
	metadata_access_disable();
684
685
686
687
688
689
690
691
	if (!fault)
		return 1;

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

	slab_bug(s, "%s overwritten", what);
692
	pr_err("INFO: 0x%p-0x%p. First byte 0x%x instead of 0x%x\n",
693
694
695
696
697
					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
698
699
700
701
702
703
704
705
706
}

/*
 * 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
707
 *
Christoph Lameter's avatar
Christoph Lameter committed
708
709
710
 * 	Poisoning uses 0x6b (POISON_FREE) and the last byte is
 * 	0xa5 (POISON_END)
 *
711
 * object + s->object_size
Christoph Lameter's avatar
Christoph Lameter committed
712
 * 	Padding to reach word boundary. This is also used for Redzoning.
Christoph Lameter's avatar
Christoph Lameter committed
713
 * 	Padding is extended by another word if Redzoning is enabled and
714
 * 	object_size == inuse.
Christoph Lameter's avatar
Christoph Lameter committed
715
 *
Christoph Lameter's avatar
Christoph Lameter committed
716
717
718
719
 * 	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
720
721
 * 	Meta data starts here.
 *
Christoph Lameter's avatar
Christoph Lameter committed
722
723
 * 	A. Free pointer (if we cannot overwrite object on free)
 * 	B. Tracking data for SLAB_STORE_USER
Christoph Lameter's avatar
Christoph Lameter committed
724
 * 	C. Padding to reach required alignment boundary or at mininum
Christoph Lameter's avatar
Christoph Lameter committed
725
 * 		one word if debugging is on to be able to detect writes
Christoph Lameter's avatar
Christoph Lameter committed
726
727
728
 * 		before the word boundary.
 *
 *	Padding is done using 0x5a (POISON_INUSE)
Christoph Lameter's avatar
Christoph Lameter committed
729
730
 *
 * object + s->size
Christoph Lameter's avatar
Christoph Lameter committed
731
 * 	Nothing is used beyond s->size.
Christoph Lameter's avatar
Christoph Lameter committed
732
 *
733
 * If slabcaches are merged then the object_size and inuse boundaries are mostly
Christoph Lameter's avatar
Christoph Lameter committed
734
 * ignored. And therefore no slab options that rely on these boundaries
Christoph Lameter's avatar
Christoph Lameter committed
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
 * 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;

753
754
	return check_bytes_and_report(s, page, p, "Object padding",
				p + off, POISON_INUSE, s->size - off);
Christoph Lameter's avatar
Christoph Lameter committed
755
756
}

757
/* Check the pad bytes at the end of a slab page */
Christoph Lameter's avatar
Christoph Lameter committed
758
759
static int slab_pad_check(struct kmem_cache *s, struct page *page)
{
760
761
762
763
764
	u8 *start;
	u8 *fault;
	u8 *end;
	int length;
	int remainder;
Christoph Lameter's avatar
Christoph Lameter committed
765
766
767
768

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

769
	start = page_address(page);
770
	length = (PAGE_SIZE << compound_order(page)) - s->reserved;
771
772
	end = start + length;
	remainder = length % s->size;
Christoph Lameter's avatar
Christoph Lameter committed
773
774
775
	if (!remainder)
		return 1;

776
	metadata_access_enable();
777
	fault = memchr_inv(end - remainder, POISON_INUSE, remainder);
778
	metadata_access_disable();
779
780
781
782
783
784
	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);
785
	print_section("Padding ", end - remainder, remainder);
786

Eric Dumazet's avatar
Eric Dumazet committed
787
	restore_bytes(s, "slab padding", POISON_INUSE, end - remainder, end);
788
	return 0;
Christoph Lameter's avatar
Christoph Lameter committed
789
790
791
}

static int check_object(struct kmem_cache *s, struct page *page,
792
					void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
793
794
{
	u8 *p = object;
795
	u8 *endobject = object + s->object_size;
Christoph Lameter's avatar
Christoph Lameter committed
796
797

	if (s->flags & SLAB_RED_ZONE) {
798
		if (!check_bytes_and_report(s, page, object, "Redzone",
799
			endobject, val, s->inuse - s->object_size))
Christoph Lameter's avatar
Christoph Lameter committed
800
801
			return 0;
	} else {
802
		if ((s->flags & SLAB_POISON) && s->object_size < s->inuse) {
Ingo Molnar's avatar
Ingo Molnar committed
803
			check_bytes_and_report(s, page, p, "Alignment padding",
804
805
				endobject, POISON_INUSE,
				s->inuse - s->object_size);
Ingo Molnar's avatar
Ingo Molnar committed
806
		}
Christoph Lameter's avatar
Christoph Lameter committed
807
808
809
	}

	if (s->flags & SLAB_POISON) {
810
		if (val != SLUB_RED_ACTIVE && (s->flags & __OBJECT_POISON) &&
811
			(!check_bytes_and_report(s, page, p, "Poison", p,
812
					POISON_FREE, s->object_size - 1) ||
813
			 !check_bytes_and_report(s, page, p, "Poison",
814
				p + s->object_size - 1, POISON_END, 1)))
Christoph Lameter's avatar
Christoph Lameter committed
815
816
817
818
819
820
821
			return 0;
		/*
		 * check_pad_bytes cleans up on its own.
		 */
		check_pad_bytes(s, page, p);
	}

822
	if (!s->offset && val == SLUB_RED_ACTIVE)
Christoph Lameter's avatar
Christoph Lameter committed
823
824
825
826
827
828
829
830
831
832
		/*
		 * 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
833
		 * No choice but to zap it and thus lose the remainder
Christoph Lameter's avatar
Christoph Lameter committed
834
		 * of the free objects in this slab. May cause
Christoph Lameter's avatar
Christoph Lameter committed
835
		 * another error because the object count is now wrong.
Christoph Lameter's avatar
Christoph Lameter committed
836
		 */
837
		set_freepointer(s, p, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
838
839
840
841
842
843
844
		return 0;
	}
	return 1;
}

static int check_slab(struct kmem_cache *s, struct page *page)
{
845
846
	int maxobj;

Christoph Lameter's avatar
Christoph Lameter committed
847
848
849
	VM_BUG_ON(!irqs_disabled());

	if (!PageSlab(page)) {
850
		slab_err(s, page, "Not a valid slab page");
Christoph Lameter's avatar
Christoph Lameter committed
851
852
		return 0;
	}
853

854
	maxobj = order_objects(compound_order(page), s->size, s->reserved);
855
856
	if (page->objects > maxobj) {
		slab_err(s, page, "objects %u > max %u",
857
			page->objects, maxobj);
858
859
860
		return 0;
	}
	if (page->inuse > page->objects) {
861
		slab_err(s, page, "inuse %u > max %u",
862
			page->inuse, page->objects);
Christoph Lameter's avatar
Christoph Lameter committed
863
864
865
866
867
868
869
870
		return 0;
	}
	/* Slab_pad_check fixes things up after itself */
	slab_pad_check(s, page);
	return 1;
}

/*
Christoph Lameter's avatar
Christoph Lameter committed
871
872
 * 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
873
874
875
876
 */
static int on_freelist(struct kmem_cache *s, struct page *page, void *search)
{
	int nr = 0;
877
	void *fp;
Christoph Lameter's avatar
Christoph Lameter committed
878
	void *object = NULL;
879
	int max_objects;
Christoph Lameter's avatar
Christoph Lameter committed
880

881
	fp = page->freelist;
882
	while (fp && nr <= page->objects) {
Christoph Lameter's avatar
Christoph Lameter committed
883
884
885
886
887
888
		if (fp == search)
			return 1;
		if (!check_valid_pointer(s, page, fp)) {
			if (object) {
				object_err(s, page, object,
					"Freechain corrupt");
889
				set_freepointer(s, object, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
890
			} else {
891
				slab_err(s, page, "Freepointer corrupt");
892
				page->freelist = NULL;
893
				page->inuse = page->objects;
894
				slab_fix(s, "Freelist cleared");
Christoph Lameter's avatar
Christoph Lameter committed
895
896
897
898
899
900
901
902
903
				return 0;
			}
			break;
		}
		object = fp;
		fp = get_freepointer(s, object);
		nr++;
	}

904
	max_objects = order_objects(compound_order(page), s->size, s->reserved);
905
906
	if (max_objects > MAX_OBJS_PER_PAGE)
		max_objects = MAX_OBJS_PER_PAGE;
907
908
909
910
911
912
913

	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.");
	}
914
	if (page->inuse != page->objects - nr) {
915
		slab_err(s, page, "Wrong object count. Counter is %d but "
916
917
			"counted were %d", page->inuse, page->objects - nr);
		page->inuse = page->objects - nr;
918
		slab_fix(s, "Object count adjusted.");
Christoph Lameter's avatar
Christoph Lameter committed
919
920
921
922
	}
	return search == NULL;
}

923
924
static void trace(struct kmem_cache *s, struct page *page, void *object,
								int alloc)
Christoph Lameter's avatar
Christoph Lameter committed
925
926
{
	if (s->flags & SLAB_TRACE) {
927
		pr_info("TRACE %s %s 0x%p inuse=%d fp=0x%p\n",
Christoph Lameter's avatar
Christoph Lameter committed
928
929
930
931
932
933
			s->name,
			alloc ? "alloc" : "free",
			object, page->inuse,
			page->freelist);

		if (!alloc)
934
935
			print_section("Object ", (void *)object,
					s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
936
937
938
939
940

		dump_stack();
	}
}

941
/*
Christoph Lameter's avatar
Christoph Lameter committed
942
 * Tracking of fully allocated slabs for debugging purposes.
943
 */
944
945
static void add_full(struct kmem_cache *s,
	struct kmem_cache_node *n, struct page *page)
946
{
947
948
949
	if (!(s->flags & SLAB_STORE_USER))
		return;

950
	lockdep_assert_held(&n->list_lock);
951
952
953
	list_add(&page->lru, &n->full);
}

Peter Zijlstra's avatar
Peter Zijlstra committed
954
static void remove_full(struct kmem_cache *s, struct kmem_cache_node *n, struct page *page)
955
956
957
958
{
	if (!(s->flags & SLAB_STORE_USER))
		return;

959
	lockdep_assert_held(&n->list_lock);
960
961
962
	list_del(&page->lru);
}

963
964
965
966
967
968
969
970
/* 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);
}

971
972
973
974
975
static inline unsigned long node_nr_slabs(struct kmem_cache_node *n)
{
	return atomic_long_read(&n->nr_slabs);
}

976
static inline void inc_slabs_node(struct kmem_cache *s, int node, int objects)
977
978
979
980
981
982
983
984
985
{
	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).
	 */
986
	if (likely(n)) {
987
		atomic_long_inc(&n->nr_slabs);
988
989
		atomic_long_add(objects, &n->total_objects);
	}
990
}
991
static inline void dec_slabs_node(struct kmem_cache *s, int node, int objects)
992
993
994
995
{
	struct kmem_cache_node *n = get_node(s, node);

	atomic_long_dec(&n->nr_slabs);
996
	atomic_long_sub(objects, &n->total_objects);
997
998
999
}

/* Object debug checks for alloc/free paths */
Christoph Lameter's avatar
Christoph Lameter committed
1000
1001
1002
1003
1004
1005
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;

1006
	init_object(s, object, SLUB_RED_INACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1007
1008
1009
	init_tracking(s, object);
}

1010
static inline int alloc_consistency_checks(struct kmem_cache *s,
1011
					struct page *page,
1012
					void *object, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
1013
1014
{
	if (!check_slab(s, page))
1015
		return 0;
Christoph Lameter's avatar
Christoph Lameter committed
1016
1017
1018

	if (!check_valid_pointer(s, page, object)) {
		object_err(s, page, object, "Freelist Pointer check fails");
1019
		return 0;
Christoph Lameter's avatar
Christoph Lameter committed
1020
1021
	}

1022
	if (!check_object(s, page, object, SLUB_RED_INACTIVE))
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
		return 0;

	return 1;
}

static noinline int alloc_debug_processing(struct kmem_cache *s,
					struct page *page,
					void *object, unsigned long addr)
{
	if (s->flags & SLAB_CONSISTENCY_CHECKS) {
		if (!alloc_consistency_checks(s, page, object, addr))
			goto bad;
	}
Christoph Lameter's avatar
Christoph Lameter committed
1036

Christoph Lameter's avatar
Christoph Lameter committed
1037
1038
1039
1040
	/* 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);
1041
	init_object(s, object, SLUB_RED_ACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1042
	return 1;
Christoph Lameter's avatar
Christoph Lameter committed
1043

Christoph Lameter's avatar
Christoph Lameter committed
1044
1045
1046
1047
1048
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
1049
		 * as used avoids touching the remaining objects.
Christoph Lameter's avatar
Christoph Lameter committed
1050
		 */
1051
		slab_fix(s, "Marking all objects used");
1052
		page->inuse = page->objects;
1053
		page->freelist = NULL;
Christoph Lameter's avatar
Christoph Lameter committed
1054
1055
1056
1057
	}
	return 0;
}

1058
1059
static inline int free_consistency_checks(struct kmem_cache *s,
		struct page *page, void *object, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
1060
1061
{
	if (!check_valid_pointer(s, page, object)) {
1062
		slab_err(s, page, "Invalid object pointer 0x%p", object);
1063
		return 0;
Christoph Lameter's avatar
Christoph Lameter committed
1064
1065
1066
	}

	if (on_freelist(s, page, object)) {
1067
		object_err(s, page, object, "Object already free");
1068
		return 0;
Christoph Lameter's avatar
Christoph Lameter committed
1069
1070
	}

1071
	if (!check_object(s, page, object, SLUB_RED_ACTIVE))
1072
		return 0;
Christoph Lameter's avatar
Christoph Lameter committed
1073

1074
	if (unlikely(s != page->slab_cache)) {
Ingo Molnar's avatar
Ingo Molnar committed
1075
		if (!PageSlab(page)) {
1076
1077
			slab_err(s, page, "Attempt to free object(0x%p) "
				"outside of slab", object);
1078
		} else if (!page->