slub.c 134 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

Joonsoo Kim's avatar
Joonsoo Kim committed
127
128
129
130
131
132
133
134
static inline void *fixup_red_left(struct kmem_cache *s, void *p)
{
	if (kmem_cache_debug(s) && s->flags & SLAB_RED_ZONE)
		p += s->red_left_pad;

	return p;
}

135
136
137
138
139
140
141
142
143
static inline bool kmem_cache_has_cpu_partial(struct kmem_cache *s)
{
#ifdef CONFIG_SLUB_CPU_PARTIAL
	return !kmem_cache_debug(s);
#else
	return false;
#endif
}

Christoph Lameter's avatar
Christoph Lameter committed
144
145
146
147
148
149
150
151
152
153
154
/*
 * Issues still to be resolved:
 *
 * - Support PAGE_ALLOC_DEBUG. Should be easy to do.
 *
 * - Variable sizing of the per node arrays
 */

/* Enable to test recovery from slab corruption on boot */
#undef SLUB_RESILIENCY_TEST

155
156
157
/* Enable to log cmpxchg failures */
#undef SLUB_DEBUG_CMPXCHG

158
159
160
161
/*
 * Mininum number of partial slabs. These will be left on the partial
 * lists even if they are empty. kmem_cache_shrink may reclaim them.
 */
162
#define MIN_PARTIAL 5
Christoph Lameter's avatar
Christoph Lameter committed
163

164
165
166
/*
 * Maximum number of desirable partial slabs.
 * The existence of more partial slabs makes kmem_cache_shrink
167
 * sort the partial list by the number of objects in use.
168
169
170
 */
#define MAX_PARTIAL 10

171
#define DEBUG_DEFAULT_FLAGS (SLAB_CONSISTENCY_CHECKS | SLAB_RED_ZONE | \
Christoph Lameter's avatar
Christoph Lameter committed
172
				SLAB_POISON | SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
173

174
175
176
177
178
179
180
181
/*
 * These debug flags cannot use CMPXCHG because there might be consistency
 * issues when checking or reading debug information
 */
#define SLAB_NO_CMPXCHG (SLAB_CONSISTENCY_CHECKS | SLAB_STORE_USER | \
				SLAB_TRACE)


182
/*
183
184
185
 * Debugging flags that require metadata to be stored in the slab.  These get
 * disabled when slub_debug=O is used and a cache's min order increases with
 * metadata.
186
 */
187
#define DEBUG_METADATA_FLAGS (SLAB_RED_ZONE | SLAB_POISON | SLAB_STORE_USER)
188

189
190
#define OO_SHIFT	16
#define OO_MASK		((1 << OO_SHIFT) - 1)
191
#define MAX_OBJS_PER_PAGE	32767 /* since page.objects is u15 */
192

Christoph Lameter's avatar
Christoph Lameter committed
193
/* Internal SLUB flags */
Christoph Lameter's avatar
Christoph Lameter committed
194
#define __OBJECT_POISON		0x80000000UL /* Poison object */
195
#define __CMPXCHG_DOUBLE	0x40000000UL /* Use cmpxchg_double */
Christoph Lameter's avatar
Christoph Lameter committed
196
197
198
199
200

#ifdef CONFIG_SMP
static struct notifier_block slab_notifier;
#endif

201
202
203
/*
 * Tracking user of a slab.
 */
204
#define TRACK_ADDRS_COUNT 16
205
struct track {
206
	unsigned long addr;	/* Called from address */
207
208
209
#ifdef CONFIG_STACKTRACE
	unsigned long addrs[TRACK_ADDRS_COUNT];	/* Called from address */
#endif
210
211
212
213
214
215
216
	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 };

217
#ifdef CONFIG_SYSFS
Christoph Lameter's avatar
Christoph Lameter committed
218
219
static int sysfs_slab_add(struct kmem_cache *);
static int sysfs_slab_alias(struct kmem_cache *, const char *);
220
static void memcg_propagate_slab_attrs(struct kmem_cache *s);
Christoph Lameter's avatar
Christoph Lameter committed
221
#else
222
223
224
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; }
225
static inline void memcg_propagate_slab_attrs(struct kmem_cache *s) { }
Christoph Lameter's avatar
Christoph Lameter committed
226
227
#endif

228
static inline void stat(const struct kmem_cache *s, enum stat_item si)
229
230
{
#ifdef CONFIG_SLUB_STATS
231
232
233
234
235
	/*
	 * 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]);
236
237
238
#endif
}

Christoph Lameter's avatar
Christoph Lameter committed
239
240
241
242
/********************************************************************
 * 			Core slab cache functions
 *******************************************************************/

243
244
245
246
247
static inline void *get_freepointer(struct kmem_cache *s, void *object)
{
	return *(void **)(object + s->offset);
}

248
249
250
251
252
static void prefetch_freepointer(const struct kmem_cache *s, void *object)
{
	prefetch(object + s->offset);
}

253
254
255
256
257
258
259
260
261
262
263
264
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;
}

265
266
267
268
269
270
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 */
271
#define for_each_object(__p, __s, __addr, __objects) \
Joonsoo Kim's avatar
Joonsoo Kim committed
272
273
274
	for (__p = fixup_red_left(__s, __addr); \
		__p < (__addr) + (__objects) * (__s)->size; \
		__p += (__s)->size)
275

276
#define for_each_object_idx(__p, __idx, __s, __addr, __objects) \
Joonsoo Kim's avatar
Joonsoo Kim committed
277
278
279
	for (__p = fixup_red_left(__s, __addr), __idx = 1; \
		__idx <= __objects; \
		__p += (__s)->size, __idx++)
280

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);
}

Joonsoo Kim's avatar
Joonsoo Kim committed
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
static inline int size_from_object(struct kmem_cache *s)
{
	if (s->flags & SLAB_RED_ZONE)
		return s->size - s->red_left_pad;

	return s->size;
}

static inline void *restore_red_left(struct kmem_cache *s, void *p)
{
	if (s->flags & SLAB_RED_ZONE)
		p -= s->red_left_pad;

	return p;
}

453
454
455
/*
 * Debug settings:
 */
456
#if defined(CONFIG_SLUB_DEBUG_ON)
457
static int slub_debug = DEBUG_DEFAULT_FLAGS;
458
459
#elif defined(CONFIG_KASAN)
static int slub_debug = SLAB_STORE_USER;
460
#else
461
static int slub_debug;
462
#endif
463
464

static char *slub_debug_slabs;
465
static int disable_higher_order_debug;
466

467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
/*
 * 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
483
484
485
/*
 * Object debugging
 */
Joonsoo Kim's avatar
Joonsoo Kim committed
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505

/* Verify that a pointer has an address that is valid within a slab page */
static inline int check_valid_pointer(struct kmem_cache *s,
				struct page *page, void *object)
{
	void *base;

	if (!object)
		return 1;

	base = page_address(page);
	object = restore_red_left(s, object);
	if (object < base || object >= base + page->objects * s->size ||
		(object - base) % s->size) {
		return 0;
	}

	return 1;
}

Christoph Lameter's avatar
Christoph Lameter committed
506
507
static void print_section(char *text, u8 *addr, unsigned int length)
{
508
	metadata_access_enable();
509
510
	print_hex_dump(KERN_ERR, text, DUMP_PREFIX_ADDRESS, 16, 1, addr,
			length, 1);
511
	metadata_access_disable();
Christoph Lameter's avatar
Christoph Lameter committed
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
}

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,
528
			enum track_item alloc, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
529
{
Akinobu Mita's avatar
Akinobu Mita committed
530
	struct track *p = get_track(s, object, alloc);
Christoph Lameter's avatar
Christoph Lameter committed
531
532

	if (addr) {
533
534
535
536
537
538
539
540
#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;
541
		metadata_access_enable();
542
		save_stack_trace(&trace);
543
		metadata_access_disable();
544
545
546
547
548
549
550
551
552

		/* 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
553
554
		p->addr = addr;
		p->cpu = smp_processor_id();
555
		p->pid = current->pid;
Christoph Lameter's avatar
Christoph Lameter committed
556
557
558
559
560
561
562
		p->when = jiffies;
	} else
		memset(p, 0, sizeof(struct track));
}

static void init_tracking(struct kmem_cache *s, void *object)
{
563
564
565
	if (!(s->flags & SLAB_STORE_USER))
		return;

566
567
	set_track(s, object, TRACK_FREE, 0UL);
	set_track(s, object, TRACK_ALLOC, 0UL);
Christoph Lameter's avatar
Christoph Lameter committed
568
569
570
571
572
573
574
}

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

575
576
	pr_err("INFO: %s in %pS age=%lu cpu=%u pid=%d\n",
	       s, (void *)t->addr, jiffies - t->when, t->cpu, t->pid);
577
578
579
580
581
#ifdef CONFIG_STACKTRACE
	{
		int i;
		for (i = 0; i < TRACK_ADDRS_COUNT; i++)
			if (t->addrs[i])
582
				pr_err("\t%pS\n", (void *)t->addrs[i]);
583
584
585
586
			else
				break;
	}
#endif
587
588
589
590
591
592
593
594
595
596
597
598
599
}

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)
{
600
	pr_err("INFO: Slab 0x%p objects=%u used=%u fp=0x%p flags=0x%04lx\n",
601
	       page, page->objects, page->inuse, page->freelist, page->flags);
602
603
604
605
606

}

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

	va_start(args, fmt);
611
612
	vaf.fmt = fmt;
	vaf.va = &args;
613
	pr_err("=============================================================================\n");
614
	pr_err("BUG %s (%s): %pV\n", s->name, print_tainted(), &vaf);
615
	pr_err("-----------------------------------------------------------------------------\n\n");
616

617
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
618
	va_end(args);
Christoph Lameter's avatar
Christoph Lameter committed
619
620
}

621
622
static void slab_fix(struct kmem_cache *s, char *fmt, ...)
{
623
	struct va_format vaf;
624
625
626
	va_list args;

	va_start(args, fmt);
627
628
629
	vaf.fmt = fmt;
	vaf.va = &args;
	pr_err("FIX %s: %pV\n", s->name, &vaf);
630
631
632
633
	va_end(args);
}

static void print_trailer(struct kmem_cache *s, struct page *page, u8 *p)
Christoph Lameter's avatar
Christoph Lameter committed
634
635
{
	unsigned int off;	/* Offset of last byte */
636
	u8 *addr = page_address(page);
637
638
639
640
641

	print_tracking(s, p);

	print_page_info(page);

642
643
	pr_err("INFO: Object 0x%p @offset=%tu fp=0x%p\n\n",
	       p, p - addr, get_freepointer(s, p));
644

Joonsoo Kim's avatar
Joonsoo Kim committed
645
646
647
	if (s->flags & SLAB_RED_ZONE)
		print_section("Redzone ", p - s->red_left_pad, s->red_left_pad);
	else if (p > addr + 16)
648
		print_section("Bytes b4 ", p - 16, 16);
Christoph Lameter's avatar
Christoph Lameter committed
649

650
	print_section("Object ", p, min_t(unsigned long, s->object_size,
651
				PAGE_SIZE));
Christoph Lameter's avatar
Christoph Lameter committed
652
	if (s->flags & SLAB_RED_ZONE)
653
654
		print_section("Redzone ", p + s->object_size,
			s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
655
656
657
658
659
660

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

661
	if (s->flags & SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
662
663
		off += 2 * sizeof(struct track);

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

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

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

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

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

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

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

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

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

708
709
710
711
712
713
714
715
716
static void restore_bytes(struct kmem_cache *s, char *message, u8 data,
						void *from, void *to)
{
	slab_fix(s, "Restoring 0x%p-0x%p=0x%x\n", from, to - 1, data);
	memset(from, data, to - from);
}

static int check_bytes_and_report(struct kmem_cache *s, struct page *page,
			u8 *object, char *what,
717
			u8 *start, unsigned int value, unsigned int bytes)
718
719
720
721
{
	u8 *fault;
	u8 *end;

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

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

	slab_bug(s, "%s overwritten", what);
733
	pr_err("INFO: 0x%p-0x%p. First byte 0x%x instead of 0x%x\n",
734
735
736
737
738
					fault, end - 1, fault[0], value);
	print_trailer(s, page, object);

	restore_bytes(s, what, value, fault, end);
	return 0;
Christoph Lameter's avatar
Christoph Lameter committed
739
740
741
742
743
744
745
746
747
}

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

static int check_pad_bytes(struct kmem_cache *s, struct page *page, u8 *p)
{
	unsigned long off = s->inuse;	/* The end of info */

	if (s->offset)
		/* Freepointer is placed after the object. */
		off += sizeof(void *);

	if (s->flags & SLAB_STORE_USER)
		/* We also have user information there */
		off += 2 * sizeof(struct track);

Joonsoo Kim's avatar
Joonsoo Kim committed
791
	if (size_from_object(s) == off)
Christoph Lameter's avatar
Christoph Lameter committed
792
793
		return 1;

794
	return check_bytes_and_report(s, page, p, "Object padding",
Joonsoo Kim's avatar
Joonsoo Kim committed
795
			p + off, POISON_INUSE, size_from_object(s) - off);
Christoph Lameter's avatar
Christoph Lameter committed
796
797
}

798
/* Check the pad bytes at the end of a slab page */
Christoph Lameter's avatar
Christoph Lameter committed
799
800
static int slab_pad_check(struct kmem_cache *s, struct page *page)
{
801
802
803
804
805
	u8 *start;
	u8 *fault;
	u8 *end;
	int length;
	int remainder;
Christoph Lameter's avatar
Christoph Lameter committed
806
807
808
809

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

810
	start = page_address(page);
811
	length = (PAGE_SIZE << compound_order(page)) - s->reserved;
812
813
	end = start + length;
	remainder = length % s->size;
Christoph Lameter's avatar
Christoph Lameter committed
814
815
816
	if (!remainder)
		return 1;

817
	metadata_access_enable();
818
	fault = memchr_inv(end - remainder, POISON_INUSE, remainder);
819
	metadata_access_disable();
820
821
822
823
824
825
	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);
826
	print_section("Padding ", end - remainder, remainder);
827

Eric Dumazet's avatar
Eric Dumazet committed
828
	restore_bytes(s, "slab padding", POISON_INUSE, end - remainder, end);
829
	return 0;
Christoph Lameter's avatar
Christoph Lameter committed
830
831
832
}

static int check_object(struct kmem_cache *s, struct page *page,
833
					void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
834
835
{
	u8 *p = object;
836
	u8 *endobject = object + s->object_size;
Christoph Lameter's avatar
Christoph Lameter committed
837
838

	if (s->flags & SLAB_RED_ZONE) {
Joonsoo Kim's avatar
Joonsoo Kim committed
839
840
841
842
		if (!check_bytes_and_report(s, page, object, "Redzone",
			object - s->red_left_pad, val, s->red_left_pad))
			return 0;

843
		if (!check_bytes_and_report(s, page, object, "Redzone",
844
			endobject, val, s->inuse - s->object_size))
Christoph Lameter's avatar
Christoph Lameter committed
845
846
			return 0;
	} else {
847
		if ((s->flags & SLAB_POISON) && s->object_size < s->inuse) {
Ingo Molnar's avatar
Ingo Molnar committed
848
			check_bytes_and_report(s, page, p, "Alignment padding",
849
850
				endobject, POISON_INUSE,
				s->inuse - s->object_size);
Ingo Molnar's avatar
Ingo Molnar committed
851
		}
Christoph Lameter's avatar
Christoph Lameter committed
852
853
854
	}

	if (s->flags & SLAB_POISON) {
855
		if (val != SLUB_RED_ACTIVE && (s->flags & __OBJECT_POISON) &&
856
			(!check_bytes_and_report(s, page, p, "Poison", p,
857
					POISON_FREE, s->object_size - 1) ||
858
			 !check_bytes_and_report(s, page, p, "Poison",
859
				p + s->object_size - 1, POISON_END, 1)))
Christoph Lameter's avatar
Christoph Lameter committed
860
861
862
863
864
865
866
			return 0;
		/*
		 * check_pad_bytes cleans up on its own.
		 */
		check_pad_bytes(s, page, p);
	}

867
	if (!s->offset && val == SLUB_RED_ACTIVE)
Christoph Lameter's avatar
Christoph Lameter committed
868
869
870
871
872
873
874
875
876
877
		/*
		 * 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
878
		 * No choice but to zap it and thus lose the remainder
Christoph Lameter's avatar
Christoph Lameter committed
879
		 * of the free objects in this slab. May cause
Christoph Lameter's avatar
Christoph Lameter committed
880
		 * another error because the object count is now wrong.
Christoph Lameter's avatar
Christoph Lameter committed
881
		 */
882
		set_freepointer(s, p, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
883
884
885
886
887
888
889
		return 0;
	}
	return 1;
}

static int check_slab(struct kmem_cache *s, struct page *page)
{
890
891
	int maxobj;

Christoph Lameter's avatar
Christoph Lameter committed
892
893
894
	VM_BUG_ON(!irqs_disabled());

	if (!PageSlab(page)) {
895
		slab_err(s, page, "Not a valid slab page");
Christoph Lameter's avatar
Christoph Lameter committed
896
897
		return 0;
	}
898

899
	maxobj = order_objects(compound_order(page), s->size, s->reserved);
900
901
	if (page->objects > maxobj) {
		slab_err(s, page, "objects %u > max %u",
902
			page->objects, maxobj);
903
904
905
		return 0;
	}
	if (page->inuse > page->objects) {
906
		slab_err(s, page, "inuse %u > max %u",
907
			page->inuse, page->objects);
Christoph Lameter's avatar
Christoph Lameter committed
908
909
910
911
912
913
914
915
		return 0;
	}
	/* Slab_pad_check fixes things up after itself */
	slab_pad_check(s, page);
	return 1;
}

/*
Christoph Lameter's avatar
Christoph Lameter committed
916
917
 * 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
918
919
920
921
 */
static int on_freelist(struct kmem_cache *s, struct page *page, void *search)
{
	int nr = 0;
922
	void *fp;
Christoph Lameter's avatar
Christoph Lameter committed
923
	void *object = NULL;
924
	int max_objects;
Christoph Lameter's avatar
Christoph Lameter committed
925

926
	fp = page->freelist;
927
	while (fp && nr <= page->objects) {
Christoph Lameter's avatar
Christoph Lameter committed
928
929
930
931
932
933
		if (fp == search)
			return 1;
		if (!check_valid_pointer(s, page, fp)) {
			if (object) {
				object_err(s, page, object,
					"Freechain corrupt");
934
				set_freepointer(s, object, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
935
			} else {
936
				slab_err(s, page, "Freepointer corrupt");
937
				page->freelist = NULL;
938
				page->inuse = page->objects;
939
				slab_fix(s, "Freelist cleared");
Christoph Lameter's avatar
Christoph Lameter committed
940
941
942
943
944
945
946
947
948
				return 0;
			}
			break;
		}
		object = fp;
		fp = get_freepointer(s, object);
		nr++;
	}

949
	max_objects = order_objects(compound_order(page), s->size, s->reserved);
950
951
	if (max_objects > MAX_OBJS_PER_PAGE)
		max_objects = MAX_OBJS_PER_PAGE;
952
953
954
955
956
957
958

	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.");
	}
959
	if (page->inuse != page->objects - nr) {
960
		slab_err(s, page, "Wrong object count. Counter is %d but "
961
962
			"counted were %d", page->inuse, page->objects - nr);
		page->inuse = page->objects - nr;
963
		slab_fix(s, "Object count adjusted.");
Christoph Lameter's avatar
Christoph Lameter committed
964
965
966
967
	}
	return search == NULL;
}

968
969
static void trace(struct kmem_cache *s, struct page *page, void *object,
								int alloc)
Christoph Lameter's avatar
Christoph Lameter committed
970
971
{
	if (s->flags & SLAB_TRACE) {
972
		pr_info("TRACE %s %s 0x%p inuse=%d fp=0x%p\n",
Christoph Lameter's avatar
Christoph Lameter committed
973
974
975
976
977
978
			s->name,
			alloc ? "alloc" : "free",
			object, page->inuse,
			page->freelist);

		if (!alloc)
979
980
			print_section("Object ", (void *)object,
					s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
981
982
983
984
985

		dump_stack();
	}
}

986
/*
Christoph Lameter's avatar
Christoph Lameter committed
987
 * Tracking of fully allocated slabs for debugging purposes.
988
 */
989
990
static void add_full(struct kmem_cache *s,
	struct kmem_cache_node *n, struct page *page)
991
{
992
993
994
	if (!(s->flags & SLAB_STORE_USER))
		return;

995
	lockdep_assert_held(&n->list_lock);
996
997
998
	list_add(&page->lru, &n->full);
}

Peter Zijlstra's avatar
Peter Zijlstra committed
999
static void remove_full(struct kmem_cache *s, struct kmem_cache_node *n, struct page *page)
1000
1001
1002
1003
{
	if (!(s->flags & SLAB_STORE_USER))
		return;

1004
	lockdep_assert_held(&n->list_lock);
1005
1006
1007
	list_del(&page->lru);
}

1008
1009
1010
1011
1012
1013
1014
1015
/* 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);
}

1016
1017
1018
1019
1020
static inline unsigned long node_nr_slabs(struct kmem_cache_node *n)
{
	return atomic_long_read(&n->nr_slabs);
}

1021
static inline void inc_slabs_node(struct kmem_cache *s, int node, int objects)
1022
1023
1024
1025
1026
1027
1028
1029
1030
{
	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).
	 */
1031
	if (likely(n)) {
1032
		atomic_long_inc(&n->nr_slabs);
1033
1034
		atomic_long_add(objects, &n->total_objects);
	}
1035
}
1036
static inline void dec_slabs_node(struct kmem_cache *s, int node, int objects)
1037
1038
1039
1040
{
	struct kmem_cache_node *n = get_node(s, node);

	atomic_long_dec(&n->nr_slabs);
1041
	atomic_long_sub(objects, &n->total_objects);
1042
1043
1044
}

/* Object debug checks for alloc/free paths */
Christoph Lameter's avatar
Christoph Lameter committed
1045
1046
1047
1048
1049
1050
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;

1051
	init_object(s, object, SLUB_RED_INACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1052
1053
1054
	init_tracking(s, object);
}

1055
static inline int alloc_consistency_checks(struct kmem_cache *s,
1056
					struct page *page,
1057
					void *object, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
1058
1059
{
	if (!check_slab(s, page))
1060
		return 0;
Christoph Lameter's avatar
Christoph Lameter committed
1061
1062
1063

	if (!check_valid_pointer(s, page, object)) {
		object_err(s, page, object, "Freelist Pointer check fails");
1064
		return 0;
Christoph Lameter's avatar
Christoph Lameter committed
1065
1066
	}

1067
	if (!check_object(s, page, object, SLUB_RED_INACTIVE))
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
		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
1081

Christoph Lameter's avatar
Christoph Lameter committed
1082
1083
1084
1085
	/* 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);
1086
	init_object(s, object, SLUB_RED_ACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1087
	return 1;
Christoph Lameter's avatar
Christoph Lameter committed
1088

Christoph Lameter's avatar
Christoph Lameter committed
1089
1090
1091
1092
1093
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
1094
		 * as used avoids touching the remaining objects.
Christoph Lameter's avatar
Christoph Lameter committed
1095
		 */
1096
		slab_fix(s, "Marking all objects used");
1097
		page->inuse = page->objects;
1098
		page->freelist = NULL;
Christoph Lameter's avatar
Christoph Lameter committed
1099
1100
1101
1102
	}
	return 0;
}