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
170
171
172
173
/*
 * 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)


174
/*
175
176
177
 * 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.
178
 */
179
#define DEBUG_METADATA_FLAGS (SLAB_RED_ZONE | SLAB_POISON | SLAB_STORE_USER)
180

181
182
#define OO_SHIFT	16
#define OO_MASK		((1 << OO_SHIFT) - 1)
183
#define MAX_OBJS_PER_PAGE	32767 /* since page.objects is u15 */
184

Christoph Lameter's avatar
Christoph Lameter committed
185
/* Internal SLUB flags */
Christoph Lameter's avatar
Christoph Lameter committed
186
#define __OBJECT_POISON		0x80000000UL /* Poison object */
187
#define __CMPXCHG_DOUBLE	0x40000000UL /* Use cmpxchg_double */
Christoph Lameter's avatar
Christoph Lameter committed
188
189
190
191
192

#ifdef CONFIG_SMP
static struct notifier_block slab_notifier;
#endif

193
194
195
/*
 * Tracking user of a slab.
 */
196
#define TRACK_ADDRS_COUNT 16
197
struct track {
198
	unsigned long addr;	/* Called from address */
199
200
201
#ifdef CONFIG_STACKTRACE
	unsigned long addrs[TRACK_ADDRS_COUNT];	/* Called from address */
#endif
202
203
204
205
206
207
208
	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 };

209
#ifdef CONFIG_SYSFS
Christoph Lameter's avatar
Christoph Lameter committed
210
211
static int sysfs_slab_add(struct kmem_cache *);
static int sysfs_slab_alias(struct kmem_cache *, const char *);
212
static void memcg_propagate_slab_attrs(struct kmem_cache *s);
Christoph Lameter's avatar
Christoph Lameter committed
213
#else
214
215
216
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; }
217
static inline void memcg_propagate_slab_attrs(struct kmem_cache *s) { }
Christoph Lameter's avatar
Christoph Lameter committed
218
219
#endif

220
static inline void stat(const struct kmem_cache *s, enum stat_item si)
221
222
{
#ifdef CONFIG_SLUB_STATS
223
224
225
226
227
	/*
	 * 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]);
228
229
230
#endif
}

Christoph Lameter's avatar
Christoph Lameter committed
231
232
233
234
/********************************************************************
 * 			Core slab cache functions
 *******************************************************************/

Christoph Lameter's avatar
Christoph Lameter committed
235
/* Verify that a pointer has an address that is valid within a slab page */
236
237
238
239
240
static inline int check_valid_pointer(struct kmem_cache *s,
				struct page *page, const void *object)
{
	void *base;

241
	if (!object)
242
243
		return 1;

244
	base = page_address(page);
245
	if (object < base || object >= base + page->objects * s->size ||
246
247
248
249
250
251
252
		(object - base) % s->size) {
		return 0;
	}

	return 1;
}

253
254
255
256
257
static inline void *get_freepointer(struct kmem_cache *s, void *object)
{
	return *(void **)(object + s->offset);
}

258
259
260
261
262
static void prefetch_freepointer(const struct kmem_cache *s, void *object)
{
	prefetch(object + s->offset);
}

263
264
265
266
267
268
269
270
271
272
273
274
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;
}

275
276
277
278
279
280
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 */
281
282
#define for_each_object(__p, __s, __addr, __objects) \
	for (__p = (__addr); __p < (__addr) + (__objects) * (__s)->size;\
283
284
			__p += (__s)->size)

285
286
287
288
#define for_each_object_idx(__p, __idx, __s, __addr, __objects) \
	for (__p = (__addr), __idx = 1; __idx <= __objects;\
			__p += (__s)->size, __idx++)

289
290
291
292
293
294
/* 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;
}

295
296
297
298
299
static inline int order_objects(int order, unsigned long size, int reserved)
{
	return ((PAGE_SIZE << order) - reserved) / size;
}

300
static inline struct kmem_cache_order_objects oo_make(int order,
301
		unsigned long size, int reserved)
302
303
{
	struct kmem_cache_order_objects x = {
304
		(order << OO_SHIFT) + order_objects(order, size, reserved)
305
306
307
308
309
310
311
	};

	return x;
}

static inline int oo_order(struct kmem_cache_order_objects x)
{
312
	return x.x >> OO_SHIFT;
313
314
315
316
}

static inline int oo_objects(struct kmem_cache_order_objects x)
{
317
	return x.x & OO_MASK;
318
319
}

320
321
322
323
324
/*
 * Per slab locking using the pagelock
 */
static __always_inline void slab_lock(struct page *page)
{
325
	VM_BUG_ON_PAGE(PageTail(page), page);
326
327
328
329
330
	bit_spin_lock(PG_locked, &page->flags);
}

static __always_inline void slab_unlock(struct page *page)
{
331
	VM_BUG_ON_PAGE(PageTail(page), page);
332
333
334
	__bit_spin_unlock(PG_locked, &page->flags);
}

335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
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;
}

350
351
352
353
354
355
356
/* 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());
357
358
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
359
	if (s->flags & __CMPXCHG_DOUBLE) {
360
		if (cmpxchg_double(&page->freelist, &page->counters,
361
362
				   freelist_old, counters_old,
				   freelist_new, counters_new))
363
			return true;
364
365
366
367
	} else
#endif
	{
		slab_lock(page);
368
369
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
370
			page->freelist = freelist_new;
371
			set_page_slub_counters(page, counters_new);
372
			slab_unlock(page);
373
			return true;
374
375
376
377
378
379
380
381
		}
		slab_unlock(page);
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
382
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
383
384
#endif

385
	return false;
386
387
}

388
389
390
391
392
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)
{
393
394
#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
    defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
395
	if (s->flags & __CMPXCHG_DOUBLE) {
396
		if (cmpxchg_double(&page->freelist, &page->counters,
397
398
				   freelist_old, counters_old,
				   freelist_new, counters_new))
399
			return true;
400
401
402
	} else
#endif
	{
403
404
405
		unsigned long flags;

		local_irq_save(flags);
406
		slab_lock(page);
407
408
		if (page->freelist == freelist_old &&
					page->counters == counters_old) {
409
			page->freelist = freelist_new;
410
			set_page_slub_counters(page, counters_new);
411
			slab_unlock(page);
412
			local_irq_restore(flags);
413
			return true;
414
		}
415
		slab_unlock(page);
416
		local_irq_restore(flags);
417
418
419
420
421
422
	}

	cpu_relax();
	stat(s, CMPXCHG_DOUBLE_FAIL);

#ifdef SLUB_DEBUG_CMPXCHG
423
	pr_info("%s %s: cmpxchg double redo ", n, s->name);
424
425
#endif

426
	return false;
427
428
}

429
#ifdef CONFIG_SLUB_DEBUG
430
431
432
/*
 * Determine a map of object in use on a page.
 *
433
 * Node listlock must be held to guarantee that the page does
434
435
436
437
438
439
440
441
442
443
444
 * 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);
}

445
446
447
/*
 * Debug settings:
 */
448
#if defined(CONFIG_SLUB_DEBUG_ON)
449
static int slub_debug = DEBUG_DEFAULT_FLAGS;
450
451
#elif defined(CONFIG_KASAN)
static int slub_debug = SLAB_STORE_USER;
452
#else
453
static int slub_debug;
454
#endif
455
456

static char *slub_debug_slabs;
457
static int disable_higher_order_debug;
458

459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
/*
 * 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
475
476
477
478
479
/*
 * Object debugging
 */
static void print_section(char *text, u8 *addr, unsigned int length)
{
480
	metadata_access_enable();
481
482
	print_hex_dump(KERN_ERR, text, DUMP_PREFIX_ADDRESS, 16, 1, addr,
			length, 1);
483
	metadata_access_disable();
Christoph Lameter's avatar
Christoph Lameter committed
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
}

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,
500
			enum track_item alloc, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
501
{
Akinobu Mita's avatar
Akinobu Mita committed
502
	struct track *p = get_track(s, object, alloc);
Christoph Lameter's avatar
Christoph Lameter committed
503
504

	if (addr) {
505
506
507
508
509
510
511
512
#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;
513
		metadata_access_enable();
514
		save_stack_trace(&trace);
515
		metadata_access_disable();
516
517
518
519
520
521
522
523
524

		/* 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
525
526
		p->addr = addr;
		p->cpu = smp_processor_id();
527
		p->pid = current->pid;
Christoph Lameter's avatar
Christoph Lameter committed
528
529
530
531
532
533
534
		p->when = jiffies;
	} else
		memset(p, 0, sizeof(struct track));
}

static void init_tracking(struct kmem_cache *s, void *object)
{
535
536
537
	if (!(s->flags & SLAB_STORE_USER))
		return;

538
539
	set_track(s, object, TRACK_FREE, 0UL);
	set_track(s, object, TRACK_ALLOC, 0UL);
Christoph Lameter's avatar
Christoph Lameter committed
540
541
542
543
544
545
546
}

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

547
548
	pr_err("INFO: %s in %pS age=%lu cpu=%u pid=%d\n",
	       s, (void *)t->addr, jiffies - t->when, t->cpu, t->pid);
549
550
551
552
553
#ifdef CONFIG_STACKTRACE
	{
		int i;
		for (i = 0; i < TRACK_ADDRS_COUNT; i++)
			if (t->addrs[i])
554
				pr_err("\t%pS\n", (void *)t->addrs[i]);
555
556
557
558
			else
				break;
	}
#endif
559
560
561
562
563
564
565
566
567
568
569
570
571
}

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)
{
572
	pr_err("INFO: Slab 0x%p objects=%u used=%u fp=0x%p flags=0x%04lx\n",
573
	       page, page->objects, page->inuse, page->freelist, page->flags);
574
575
576
577
578

}

static void slab_bug(struct kmem_cache *s, char *fmt, ...)
{
579
	struct va_format vaf;
580
581
582
	va_list args;

	va_start(args, fmt);
583
584
	vaf.fmt = fmt;
	vaf.va = &args;
585
	pr_err("=============================================================================\n");
586
	pr_err("BUG %s (%s): %pV\n", s->name, print_tainted(), &vaf);
587
	pr_err("-----------------------------------------------------------------------------\n\n");
588

589
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
590
	va_end(args);
Christoph Lameter's avatar
Christoph Lameter committed
591
592
}

593
594
static void slab_fix(struct kmem_cache *s, char *fmt, ...)
{
595
	struct va_format vaf;
596
597
598
	va_list args;

	va_start(args, fmt);
599
600
601
	vaf.fmt = fmt;
	vaf.va = &args;
	pr_err("FIX %s: %pV\n", s->name, &vaf);
602
603
604
605
	va_end(args);
}

static void print_trailer(struct kmem_cache *s, struct page *page, u8 *p)
Christoph Lameter's avatar
Christoph Lameter committed
606
607
{
	unsigned int off;	/* Offset of last byte */
608
	u8 *addr = page_address(page);
609
610
611
612
613

	print_tracking(s, p);

	print_page_info(page);

614
615
	pr_err("INFO: Object 0x%p @offset=%tu fp=0x%p\n\n",
	       p, p - addr, get_freepointer(s, p));
616
617

	if (p > addr + 16)
618
		print_section("Bytes b4 ", p - 16, 16);
Christoph Lameter's avatar
Christoph Lameter committed
619

620
	print_section("Object ", p, min_t(unsigned long, s->object_size,
621
				PAGE_SIZE));
Christoph Lameter's avatar
Christoph Lameter committed
622
	if (s->flags & SLAB_RED_ZONE)
623
624
		print_section("Redzone ", p + s->object_size,
			s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
625
626
627
628
629
630

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

631
	if (s->flags & SLAB_STORE_USER)
Christoph Lameter's avatar
Christoph Lameter committed
632
633
634
635
		off += 2 * sizeof(struct track);

	if (off != s->size)
		/* Beginning of the filler is the free pointer */
636
		print_section("Padding ", p + off, s->size - off);
637
638

	dump_stack();
Christoph Lameter's avatar
Christoph Lameter committed
639
640
}

641
void object_err(struct kmem_cache *s, struct page *page,
Christoph Lameter's avatar
Christoph Lameter committed
642
643
			u8 *object, char *reason)
{
644
	slab_bug(s, "%s", reason);
645
	print_trailer(s, page, object);
Christoph Lameter's avatar
Christoph Lameter committed
646
647
}

648
649
static void slab_err(struct kmem_cache *s, struct page *page,
			const char *fmt, ...)
Christoph Lameter's avatar
Christoph Lameter committed
650
651
652
653
{
	va_list args;
	char buf[100];

654
655
	va_start(args, fmt);
	vsnprintf(buf, sizeof(buf), fmt, args);
Christoph Lameter's avatar
Christoph Lameter committed
656
	va_end(args);
657
	slab_bug(s, "%s", buf);
658
	print_page_info(page);
Christoph Lameter's avatar
Christoph Lameter committed
659
660
661
	dump_stack();
}

662
static void init_object(struct kmem_cache *s, void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
663
664
665
666
{
	u8 *p = object;

	if (s->flags & __OBJECT_POISON) {
667
668
		memset(p, POISON_FREE, s->object_size - 1);
		p[s->object_size - 1] = POISON_END;
Christoph Lameter's avatar
Christoph Lameter committed
669
670
671
	}

	if (s->flags & SLAB_RED_ZONE)
672
		memset(p + s->object_size, val, s->inuse - s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
673
674
}

675
676
677
678
679
680
681
682
683
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,
684
			u8 *start, unsigned int value, unsigned int bytes)
685
686
687
688
{
	u8 *fault;
	u8 *end;

689
	metadata_access_enable();
690
	fault = memchr_inv(start, value, bytes);
691
	metadata_access_disable();
692
693
694
695
696
697
698
699
	if (!fault)
		return 1;

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

	slab_bug(s, "%s overwritten", what);
700
	pr_err("INFO: 0x%p-0x%p. First byte 0x%x instead of 0x%x\n",
701
702
703
704
705
					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
706
707
708
709
710
711
712
713
714
}

/*
 * 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
715
 *
Christoph Lameter's avatar
Christoph Lameter committed
716
717
718
 * 	Poisoning uses 0x6b (POISON_FREE) and the last byte is
 * 	0xa5 (POISON_END)
 *
719
 * object + s->object_size
Christoph Lameter's avatar
Christoph Lameter committed
720
 * 	Padding to reach word boundary. This is also used for Redzoning.
Christoph Lameter's avatar
Christoph Lameter committed
721
 * 	Padding is extended by another word if Redzoning is enabled and
722
 * 	object_size == inuse.
Christoph Lameter's avatar
Christoph Lameter committed
723
 *
Christoph Lameter's avatar
Christoph Lameter committed
724
725
726
727
 * 	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
728
729
 * 	Meta data starts here.
 *
Christoph Lameter's avatar
Christoph Lameter committed
730
731
 * 	A. Free pointer (if we cannot overwrite object on free)
 * 	B. Tracking data for SLAB_STORE_USER
Christoph Lameter's avatar
Christoph Lameter committed
732
 * 	C. Padding to reach required alignment boundary or at mininum
Christoph Lameter's avatar
Christoph Lameter committed
733
 * 		one word if debugging is on to be able to detect writes
Christoph Lameter's avatar
Christoph Lameter committed
734
735
736
 * 		before the word boundary.
 *
 *	Padding is done using 0x5a (POISON_INUSE)
Christoph Lameter's avatar
Christoph Lameter committed
737
738
 *
 * object + s->size
Christoph Lameter's avatar
Christoph Lameter committed
739
 * 	Nothing is used beyond s->size.
Christoph Lameter's avatar
Christoph Lameter committed
740
 *
741
 * If slabcaches are merged then the object_size and inuse boundaries are mostly
Christoph Lameter's avatar
Christoph Lameter committed
742
 * ignored. And therefore no slab options that rely on these boundaries
Christoph Lameter's avatar
Christoph Lameter committed
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
 * 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;

761
762
	return check_bytes_and_report(s, page, p, "Object padding",
				p + off, POISON_INUSE, s->size - off);
Christoph Lameter's avatar
Christoph Lameter committed
763
764
}

765
/* Check the pad bytes at the end of a slab page */
Christoph Lameter's avatar
Christoph Lameter committed
766
767
static int slab_pad_check(struct kmem_cache *s, struct page *page)
{
768
769
770
771
772
	u8 *start;
	u8 *fault;
	u8 *end;
	int length;
	int remainder;
Christoph Lameter's avatar
Christoph Lameter committed
773
774
775
776

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

777
	start = page_address(page);
778
	length = (PAGE_SIZE << compound_order(page)) - s->reserved;
779
780
	end = start + length;
	remainder = length % s->size;
Christoph Lameter's avatar
Christoph Lameter committed
781
782
783
	if (!remainder)
		return 1;

784
	metadata_access_enable();
785
	fault = memchr_inv(end - remainder, POISON_INUSE, remainder);
786
	metadata_access_disable();
787
788
789
790
791
792
	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);
793
	print_section("Padding ", end - remainder, remainder);
794

Eric Dumazet's avatar
Eric Dumazet committed
795
	restore_bytes(s, "slab padding", POISON_INUSE, end - remainder, end);
796
	return 0;
Christoph Lameter's avatar
Christoph Lameter committed
797
798
799
}

static int check_object(struct kmem_cache *s, struct page *page,
800
					void *object, u8 val)
Christoph Lameter's avatar
Christoph Lameter committed
801
802
{
	u8 *p = object;
803
	u8 *endobject = object + s->object_size;
Christoph Lameter's avatar
Christoph Lameter committed
804
805

	if (s->flags & SLAB_RED_ZONE) {
806
		if (!check_bytes_and_report(s, page, object, "Redzone",
807
			endobject, val, s->inuse - s->object_size))
Christoph Lameter's avatar
Christoph Lameter committed
808
809
			return 0;
	} else {
810
		if ((s->flags & SLAB_POISON) && s->object_size < s->inuse) {
Ingo Molnar's avatar
Ingo Molnar committed
811
			check_bytes_and_report(s, page, p, "Alignment padding",
812
813
				endobject, POISON_INUSE,
				s->inuse - s->object_size);
Ingo Molnar's avatar
Ingo Molnar committed
814
		}
Christoph Lameter's avatar
Christoph Lameter committed
815
816
817
	}

	if (s->flags & SLAB_POISON) {
818
		if (val != SLUB_RED_ACTIVE && (s->flags & __OBJECT_POISON) &&
819
			(!check_bytes_and_report(s, page, p, "Poison", p,
820
					POISON_FREE, s->object_size - 1) ||
821
			 !check_bytes_and_report(s, page, p, "Poison",
822
				p + s->object_size - 1, POISON_END, 1)))
Christoph Lameter's avatar
Christoph Lameter committed
823
824
825
826
827
828
829
			return 0;
		/*
		 * check_pad_bytes cleans up on its own.
		 */
		check_pad_bytes(s, page, p);
	}

830
	if (!s->offset && val == SLUB_RED_ACTIVE)
Christoph Lameter's avatar
Christoph Lameter committed
831
832
833
834
835
836
837
838
839
840
		/*
		 * 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
841
		 * No choice but to zap it and thus lose the remainder
Christoph Lameter's avatar
Christoph Lameter committed
842
		 * of the free objects in this slab. May cause
Christoph Lameter's avatar
Christoph Lameter committed
843
		 * another error because the object count is now wrong.
Christoph Lameter's avatar
Christoph Lameter committed
844
		 */
845
		set_freepointer(s, p, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
846
847
848
849
850
851
852
		return 0;
	}
	return 1;
}

static int check_slab(struct kmem_cache *s, struct page *page)
{
853
854
	int maxobj;

Christoph Lameter's avatar
Christoph Lameter committed
855
856
857
	VM_BUG_ON(!irqs_disabled());

	if (!PageSlab(page)) {
858
		slab_err(s, page, "Not a valid slab page");
Christoph Lameter's avatar
Christoph Lameter committed
859
860
		return 0;
	}
861

862
	maxobj = order_objects(compound_order(page), s->size, s->reserved);
863
864
	if (page->objects > maxobj) {
		slab_err(s, page, "objects %u > max %u",
865
			page->objects, maxobj);
866
867
868
		return 0;
	}
	if (page->inuse > page->objects) {
869
		slab_err(s, page, "inuse %u > max %u",
870
			page->inuse, page->objects);
Christoph Lameter's avatar
Christoph Lameter committed
871
872
873
874
875
876
877
878
		return 0;
	}
	/* Slab_pad_check fixes things up after itself */
	slab_pad_check(s, page);
	return 1;
}

/*
Christoph Lameter's avatar
Christoph Lameter committed
879
880
 * 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
881
882
883
884
 */
static int on_freelist(struct kmem_cache *s, struct page *page, void *search)
{
	int nr = 0;
885
	void *fp;
Christoph Lameter's avatar
Christoph Lameter committed
886
	void *object = NULL;
887
	int max_objects;
Christoph Lameter's avatar
Christoph Lameter committed
888

889
	fp = page->freelist;
890
	while (fp && nr <= page->objects) {
Christoph Lameter's avatar
Christoph Lameter committed
891
892
893
894
895
896
		if (fp == search)
			return 1;
		if (!check_valid_pointer(s, page, fp)) {
			if (object) {
				object_err(s, page, object,
					"Freechain corrupt");
897
				set_freepointer(s, object, NULL);
Christoph Lameter's avatar
Christoph Lameter committed
898
			} else {
899
				slab_err(s, page, "Freepointer corrupt");
900
				page->freelist = NULL;
901
				page->inuse = page->objects;
902
				slab_fix(s, "Freelist cleared");
Christoph Lameter's avatar
Christoph Lameter committed
903
904
905
906
907
908
909
910
911
				return 0;
			}
			break;
		}
		object = fp;
		fp = get_freepointer(s, object);
		nr++;
	}

912
	max_objects = order_objects(compound_order(page), s->size, s->reserved);
913
914
	if (max_objects > MAX_OBJS_PER_PAGE)
		max_objects = MAX_OBJS_PER_PAGE;
915
916
917
918
919
920
921

	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.");
	}
922
	if (page->inuse != page->objects - nr) {
923
		slab_err(s, page, "Wrong object count. Counter is %d but "
924
925
			"counted were %d", page->inuse, page->objects - nr);
		page->inuse = page->objects - nr;
926
		slab_fix(s, "Object count adjusted.");
Christoph Lameter's avatar
Christoph Lameter committed
927
928
929
930
	}
	return search == NULL;
}

931
932
static void trace(struct kmem_cache *s, struct page *page, void *object,
								int alloc)
Christoph Lameter's avatar
Christoph Lameter committed
933
934
{
	if (s->flags & SLAB_TRACE) {
935
		pr_info("TRACE %s %s 0x%p inuse=%d fp=0x%p\n",
Christoph Lameter's avatar
Christoph Lameter committed
936
937
938
939
940
941
			s->name,
			alloc ? "alloc" : "free",
			object, page->inuse,
			page->freelist);

		if (!alloc)
942
943
			print_section("Object ", (void *)object,
					s->object_size);
Christoph Lameter's avatar
Christoph Lameter committed
944
945
946
947
948

		dump_stack();
	}
}

949
/*
Christoph Lameter's avatar
Christoph Lameter committed
950
 * Tracking of fully allocated slabs for debugging purposes.
951
 */
952
953
static void add_full(struct kmem_cache *s,
	struct kmem_cache_node *n, struct page *page)
954
{
955
956
957
	if (!(s->flags & SLAB_STORE_USER))
		return;

958
	lockdep_assert_held(&n->list_lock);
959
960
961
	list_add(&page->lru, &n->full);
}

Peter Zijlstra's avatar
Peter Zijlstra committed
962
static void remove_full(struct kmem_cache *s, struct kmem_cache_node *n, struct page *page)
963
964
965
966
{
	if (!(s->flags & SLAB_STORE_USER))
		return;

967
	lockdep_assert_held(&n->list_lock);
968
969
970
	list_del(&page->lru);
}

971
972
973
974
975
976
977
978
/* 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);
}

979
980
981
982
983
static inline unsigned long node_nr_slabs(struct kmem_cache_node *n)
{
	return atomic_long_read(&n->nr_slabs);
}

984
static inline void inc_slabs_node(struct kmem_cache *s, int node, int objects)
985
986
987
988
989
990
991
992
993
{
	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).
	 */
994
	if (likely(n)) {
995
		atomic_long_inc(&n->nr_slabs);
996
997
		atomic_long_add(objects, &n->total_objects);
	}
998
}
999
static inline void dec_slabs_node(struct kmem_cache *s, int node, int objects)
1000
1001
1002
1003
{
	struct kmem_cache_node *n = get_node(s, node);

	atomic_long_dec(&n->nr_slabs);
1004
	atomic_long_sub(objects, &n->total_objects);
1005
1006
1007
}

/* Object debug checks for alloc/free paths */
Christoph Lameter's avatar
Christoph Lameter committed
1008
1009
1010
1011
1012
1013
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;

1014
	init_object(s, object, SLUB_RED_INACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1015
1016
1017
	init_tracking(s, object);
}

1018
static inline int alloc_consistency_checks(struct kmem_cache *s,
1019
					struct page *page,
1020
					void *object, unsigned long addr)
Christoph Lameter's avatar
Christoph Lameter committed
1021
1022
{
	if (!check_slab(s, page))
1023
		return 0;
Christoph Lameter's avatar
Christoph Lameter committed
1024
1025
1026

	if (!check_valid_pointer(s, page, object)) {
		object_err(s, page, object, "Freelist Pointer check fails");
1027
		return 0;
Christoph Lameter's avatar
Christoph Lameter committed
1028
1029
	}

1030
	if (!check_object(s, page, object, SLUB_RED_INACTIVE))
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
		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
1044

Christoph Lameter's avatar
Christoph Lameter committed
1045
1046
1047
1048
	/* 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);
1049
	init_object(s, object, SLUB_RED_ACTIVE);
Christoph Lameter's avatar
Christoph Lameter committed
1050
	return 1;
Christoph Lameter's avatar
Christoph Lameter committed
1051

Christoph Lameter's avatar
Christoph Lameter committed
1052
1053
1054
1055
1056
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
1057
		 * as used avoids touching the remaining objects.
Christoph Lameter's avatar
Christoph Lameter committed
1058
		 */
1059
		slab_fix(s, "Marking all objects used");
1060
		page->inuse = page->objects;
1061
		page->freelist = NULL;
Christoph Lameter's avatar
Christoph Lameter committed
1062
1063
1064
1065
	}
	return 0;
}

1066
1067
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
1068
1069
{
	if (!check_valid_pointer(s, page, object)) {
1070
		slab_err(s, page, "Invalid object pointer 0x%p", object);
1071
		return 0;
Christoph Lameter's avatar
Christoph Lameter committed
1072
1073
1074
	}

	if (on_freelist(s, page, object)) {
1075
		object_err(s, page, object, "Object already free");
1076
		return 0;
Christoph Lameter's avatar
Christoph Lameter committed
1077
1078
	}

1079
	if (!check_object(s, page, object, SLUB_RED_ACTIVE))
1080
		return 0;
Christoph Lameter's avatar
Christoph Lameter committed
1081

1082
	if (unlikely(s != page->slab_cache)) {
Ingo Molnar's avatar
Ingo Molnar committed
1083
		if (!PageSlab(page)) {
Christoph Lameter's avatar