aboutsummaryrefslogtreecommitdiffstats
path: root/module/os/linux/spl/spl-kmem.c
blob: 943966cbb17a81d41838f93bd6368e7475391e14 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
/*
 *  Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
 *  Copyright (C) 2007 The Regents of the University of California.
 *  Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
 *  Written by Brian Behlendorf <behlendorf1@llnl.gov>.
 *  UCRL-CODE-235197
 *
 *  This file is part of the SPL, Solaris Porting Layer.
 *
 *  The SPL is free software; you can redistribute it and/or modify it
 *  under the terms of the GNU General Public License as published by the
 *  Free Software Foundation; either version 2 of the License, or (at your
 *  option) any later version.
 *
 *  The SPL is distributed in the hope that it will be useful, but WITHOUT
 *  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 *  FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 *  for more details.
 *
 *  You should have received a copy of the GNU General Public License along
 *  with the SPL.  If not, see <http://www.gnu.org/licenses/>.
 */

#include <sys/debug.h>
#include <sys/sysmacros.h>
#include <sys/kmem.h>
#include <sys/vmem.h>

/*
 * As a general rule kmem_alloc() allocations should be small, preferably
 * just a few pages since they must by physically contiguous.  Therefore, a
 * rate limited warning will be printed to the console for any kmem_alloc()
 * which exceeds a reasonable threshold.
 *
 * The default warning threshold is set to sixteen pages but capped at 64K to
 * accommodate systems using large pages.  This value was selected to be small
 * enough to ensure the largest allocations are quickly noticed and fixed.
 * But large enough to avoid logging any warnings when a allocation size is
 * larger than optimal but not a serious concern.  Since this value is tunable,
 * developers are encouraged to set it lower when testing so any new largish
 * allocations are quickly caught.  These warnings may be disabled by setting
 * the threshold to zero.
 */
/* BEGIN CSTYLED */
unsigned int spl_kmem_alloc_warn = MIN(16 * PAGE_SIZE, 64 * 1024);
module_param(spl_kmem_alloc_warn, uint, 0644);
MODULE_PARM_DESC(spl_kmem_alloc_warn,
	"Warning threshold in bytes for a kmem_alloc()");
EXPORT_SYMBOL(spl_kmem_alloc_warn);

/*
 * Large kmem_alloc() allocations will fail if they exceed KMALLOC_MAX_SIZE.
 * Allocations which are marginally smaller than this limit may succeed but
 * should still be avoided due to the expense of locating a contiguous range
 * of free pages.  Therefore, a maximum kmem size with reasonable safely
 * margin of 4x is set.  Kmem_alloc() allocations larger than this maximum
 * will quickly fail.  Vmem_alloc() allocations less than or equal to this
 * value will use kmalloc(), but shift to vmalloc() when exceeding this value.
 */
unsigned int spl_kmem_alloc_max = (KMALLOC_MAX_SIZE >> 2);
module_param(spl_kmem_alloc_max, uint, 0644);
MODULE_PARM_DESC(spl_kmem_alloc_max,
	"Maximum size in bytes for a kmem_alloc()");
EXPORT_SYMBOL(spl_kmem_alloc_max);
/* END CSTYLED */

int
kmem_debugging(void)
{
	return (0);
}
EXPORT_SYMBOL(kmem_debugging);

char *
kmem_vasprintf(const char *fmt, va_list ap)
{
	va_list aq;
	char *ptr;

	do {
		va_copy(aq, ap);
		ptr = kvasprintf(kmem_flags_convert(KM_SLEEP), fmt, aq);
		va_end(aq);
	} while (ptr == NULL);

	return (ptr);
}
EXPORT_SYMBOL(kmem_vasprintf);

char *
kmem_asprintf(const char *fmt, ...)
{
	va_list ap;
	char *ptr;

	do {
		va_start(ap, fmt);
		ptr = kvasprintf(kmem_flags_convert(KM_SLEEP), fmt, ap);
		va_end(ap);
	} while (ptr == NULL);

	return (ptr);
}
EXPORT_SYMBOL(kmem_asprintf);

static char *
__strdup(const char *str, int flags)
{
	char *ptr;
	int n;

	n = strlen(str);
	ptr = kmalloc(n + 1, kmem_flags_convert(flags));
	if (ptr)
		memcpy(ptr, str, n + 1);

	return (ptr);
}

char *
kmem_strdup(const char *str)
{
	return (__strdup(str, KM_SLEEP));
}
EXPORT_SYMBOL(kmem_strdup);

void
kmem_strfree(char *str)
{
	kfree(str);
}
EXPORT_SYMBOL(kmem_strfree);

void *
spl_kvmalloc(size_t size, gfp_t lflags)
{
#ifdef HAVE_KVMALLOC
	/*
	 * GFP_KERNEL allocations can safely use kvmalloc which may
	 * improve performance by avoiding a) high latency caused by
	 * vmalloc's on-access allocation, b) performance loss due to
	 * MMU memory address mapping and c) vmalloc locking overhead.
	 * This has the side-effect that the slab statistics will
	 * incorrectly report this as a vmem allocation, but that is
	 * purely cosmetic.
	 */
	if ((lflags & GFP_KERNEL) == GFP_KERNEL)
		return (kvmalloc(size, lflags));
#endif

	gfp_t kmalloc_lflags = lflags;

	if (size > PAGE_SIZE) {
		/*
		 * We need to set __GFP_NOWARN here since spl_kvmalloc is not
		 * only called by spl_kmem_alloc_impl but can be called
		 * directly with custom lflags, too. In that case
		 * kmem_flags_convert does not get called, which would
		 * implicitly set __GFP_NOWARN.
		 */
		kmalloc_lflags |= __GFP_NOWARN;

		/*
		 * N.B. __GFP_RETRY_MAYFAIL is supported only for large
		 * e (>32kB) allocations.
		 *
		 * We have to override __GFP_RETRY_MAYFAIL by __GFP_NORETRY
		 * for !costly requests because there is no other way to tell
		 * the allocator that we want to fail rather than retry
		 * endlessly.
		 */
		if (!(kmalloc_lflags & __GFP_RETRY_MAYFAIL) ||
		    (size <= PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) {
			kmalloc_lflags |= __GFP_NORETRY;
		}
	}

	/*
	 * We first try kmalloc - even for big sizes - and fall back to
	 * spl_vmalloc if that fails.
	 *
	 * For non-__GFP-RECLAIM allocations we always stick to
	 * kmalloc_node, and fail when kmalloc is not successful (returns
	 * NULL).
	 * We cannot fall back to spl_vmalloc in this case because spl_vmalloc
	 * internally uses GPF_KERNEL allocations.
	 */
	void *ptr = kmalloc_node(size, kmalloc_lflags, NUMA_NO_NODE);
	if (ptr || size <= PAGE_SIZE ||
	    (lflags & __GFP_RECLAIM) != __GFP_RECLAIM) {
		return (ptr);
	}

	return (spl_vmalloc(size, lflags | __GFP_HIGHMEM));
}

/*
 * General purpose unified implementation of kmem_alloc(). It is an
 * amalgamation of Linux and Illumos allocator design. It should never be
 * exported to ensure that code using kmem_alloc()/kmem_zalloc() remains
 * relatively portable.  Consumers may only access this function through
 * wrappers that enforce the common flags to ensure portability.
 */
inline void *
spl_kmem_alloc_impl(size_t size, int flags, int node)
{
	gfp_t lflags = kmem_flags_convert(flags);
	void *ptr;

	/*
	 * Log abnormally large allocations and rate limit the console output.
	 * Allocations larger than spl_kmem_alloc_warn should be performed
	 * through the vmem_alloc()/vmem_zalloc() interfaces.
	 */
	if ((spl_kmem_alloc_warn > 0) && (size > spl_kmem_alloc_warn) &&
	    !(flags & KM_VMEM)) {
		printk(KERN_WARNING
		    "Large kmem_alloc(%lu, 0x%x), please file an issue at:\n"
		    "https://github.com/openzfs/zfs/issues/new\n",
		    (unsigned long)size, flags);
		dump_stack();
	}

	/*
	 * Use a loop because kmalloc_node() can fail when GFP_KERNEL is used
	 * unlike kmem_alloc() with KM_SLEEP on Illumos.
	 */
	do {
		/*
		 * Calling kmalloc_node() when the size >= spl_kmem_alloc_max
		 * is unsafe.  This must fail for all for kmem_alloc() and
		 * kmem_zalloc() callers.
		 *
		 * For vmem_alloc() and vmem_zalloc() callers it is permissible
		 * to use spl_vmalloc().  However, in general use of
		 * spl_vmalloc() is strongly discouraged because a global lock
		 * must be acquired.  Contention on this lock can significantly
		 * impact performance so frequently manipulating the virtual
		 * address space is strongly discouraged.
		 */
		if (size > spl_kmem_alloc_max) {
			if (flags & KM_VMEM) {
				ptr = spl_vmalloc(size, lflags | __GFP_HIGHMEM);
			} else {
				return (NULL);
			}
		} else {
			if (flags & KM_VMEM) {
				ptr = spl_kvmalloc(size, lflags);
			} else {
				ptr = kmalloc_node(size, lflags, node);
			}
		}

		if (likely(ptr) || (flags & KM_NOSLEEP))
			return (ptr);

		/*
		 * Try hard to satisfy the allocation. However, when progress
		 * cannot be made, the allocation is allowed to fail.
		 */
		if ((lflags & GFP_KERNEL) == GFP_KERNEL)
			lflags |= __GFP_RETRY_MAYFAIL;

		/*
		 * Use cond_resched() instead of congestion_wait() to avoid
		 * deadlocking systems where there are no block devices.
		 */
		cond_resched();
	} while (1);

	return (NULL);
}

inline void
spl_kmem_free_impl(const void *buf, size_t size)
{
	if (is_vmalloc_addr(buf))
		vfree(buf);
	else
		kfree(buf);
}

/*
 * Memory allocation and accounting for kmem_* * style allocations.  When
 * DEBUG_KMEM is enabled the total memory allocated will be tracked and
 * any memory leaked will be reported during module unload.
 *
 * ./configure --enable-debug-kmem
 */
#ifdef DEBUG_KMEM

/* Shim layer memory accounting */
#ifdef HAVE_ATOMIC64_T
atomic64_t kmem_alloc_used = ATOMIC64_INIT(0);
unsigned long long kmem_alloc_max = 0;
#else  /* HAVE_ATOMIC64_T */
atomic_t kmem_alloc_used = ATOMIC_INIT(0);
unsigned long long kmem_alloc_max = 0;
#endif /* HAVE_ATOMIC64_T */

EXPORT_SYMBOL(kmem_alloc_used);
EXPORT_SYMBOL(kmem_alloc_max);

inline void *
spl_kmem_alloc_debug(size_t size, int flags, int node)
{
	void *ptr;

	ptr = spl_kmem_alloc_impl(size, flags, node);
	if (ptr) {
		kmem_alloc_used_add(size);
		if (unlikely(kmem_alloc_used_read() > kmem_alloc_max))
			kmem_alloc_max = kmem_alloc_used_read();
	}

	return (ptr);
}

inline void
spl_kmem_free_debug(const void *ptr, size_t size)
{
	kmem_alloc_used_sub(size);
	spl_kmem_free_impl(ptr, size);
}

/*
 * When DEBUG_KMEM_TRACKING is enabled not only will total bytes be tracked
 * but also the location of every alloc and free.  When the SPL module is
 * unloaded a list of all leaked addresses and where they were allocated
 * will be dumped to the console.  Enabling this feature has a significant
 * impact on performance but it makes finding memory leaks straight forward.
 *
 * Not surprisingly with debugging enabled the xmem_locks are very highly
 * contended particularly on xfree().  If we want to run with this detailed
 * debugging enabled for anything other than debugging  we need to minimize
 * the contention by moving to a lock per xmem_table entry model.
 *
 * ./configure --enable-debug-kmem-tracking
 */
#ifdef DEBUG_KMEM_TRACKING

#include <linux/hash.h>
#include <linux/ctype.h>

#define	KMEM_HASH_BITS		10
#define	KMEM_TABLE_SIZE		(1 << KMEM_HASH_BITS)

typedef struct kmem_debug {
	struct hlist_node kd_hlist;	/* Hash node linkage */
	struct list_head kd_list;	/* List of all allocations */
	void *kd_addr;			/* Allocation pointer */
	size_t kd_size;			/* Allocation size */
	const char *kd_func;		/* Allocation function */
	int kd_line;			/* Allocation line */
} kmem_debug_t;

static spinlock_t kmem_lock;
static struct hlist_head kmem_table[KMEM_TABLE_SIZE];
static struct list_head kmem_list;

static kmem_debug_t *
kmem_del_init(spinlock_t *lock, struct hlist_head *table,
    int bits, const void *addr)
{
	struct hlist_head *head;
	struct hlist_node *node = NULL;
	struct kmem_debug *p;
	unsigned long flags;

	spin_lock_irqsave(lock, flags);

	head = &table[hash_ptr((void *)addr, bits)];
	hlist_for_each(node, head) {
		p = list_entry(node, struct kmem_debug, kd_hlist);
		if (p->kd_addr == addr) {
			hlist_del_init(&p->kd_hlist);
			list_del_init(&p->kd_list);
			spin_unlock_irqrestore(lock, flags);
			return (p);
		}
	}

	spin_unlock_irqrestore(lock, flags);

	return (NULL);
}

inline void *
spl_kmem_alloc_track(size_t size, int flags,
    const char *func, int line, int node)
{
	void *ptr = NULL;
	kmem_debug_t *dptr;
	unsigned long irq_flags;

	dptr = kmalloc(sizeof (kmem_debug_t), kmem_flags_convert(flags));
	if (dptr == NULL)
		return (NULL);

	dptr->kd_func = __strdup(func, flags);
	if (dptr->kd_func == NULL) {
		kfree(dptr);
		return (NULL);
	}

	ptr = spl_kmem_alloc_debug(size, flags, node);
	if (ptr == NULL) {
		kfree(dptr->kd_func);
		kfree(dptr);
		return (NULL);
	}

	INIT_HLIST_NODE(&dptr->kd_hlist);
	INIT_LIST_HEAD(&dptr->kd_list);

	dptr->kd_addr = ptr;
	dptr->kd_size = size;
	dptr->kd_line = line;

	spin_lock_irqsave(&kmem_lock, irq_flags);
	hlist_add_head(&dptr->kd_hlist,
	    &kmem_table[hash_ptr(ptr, KMEM_HASH_BITS)]);
	list_add_tail(&dptr->kd_list, &kmem_list);
	spin_unlock_irqrestore(&kmem_lock, irq_flags);

	return (ptr);
}

inline void
spl_kmem_free_track(const void *ptr, size_t size)
{
	kmem_debug_t *dptr;

	/* Ignore NULL pointer since we haven't tracked it at all */
	if (ptr == NULL)
		return;

	/* Must exist in hash due to kmem_alloc() */
	dptr = kmem_del_init(&kmem_lock, kmem_table, KMEM_HASH_BITS, ptr);
	ASSERT3P(dptr, !=, NULL);
	ASSERT3S(dptr->kd_size, ==, size);

	kfree(dptr->kd_func);
	kfree(dptr);

	spl_kmem_free_debug(ptr, size);
}
#endif /* DEBUG_KMEM_TRACKING */
#endif /* DEBUG_KMEM */

/*
 * Public kmem_alloc(), kmem_zalloc() and kmem_free() interfaces.
 */
void *
spl_kmem_alloc(size_t size, int flags, const char *func, int line)
{
	ASSERT0(flags & ~KM_PUBLIC_MASK);

#if !defined(DEBUG_KMEM)
	return (spl_kmem_alloc_impl(size, flags, NUMA_NO_NODE));
#elif !defined(DEBUG_KMEM_TRACKING)
	return (spl_kmem_alloc_debug(size, flags, NUMA_NO_NODE));
#else
	return (spl_kmem_alloc_track(size, flags, func, line, NUMA_NO_NODE));
#endif
}
EXPORT_SYMBOL(spl_kmem_alloc);

void *
spl_kmem_zalloc(size_t size, int flags, const char *func, int line)
{
	ASSERT0(flags & ~KM_PUBLIC_MASK);

	flags |= KM_ZERO;

#if !defined(DEBUG_KMEM)
	return (spl_kmem_alloc_impl(size, flags, NUMA_NO_NODE));
#elif !defined(DEBUG_KMEM_TRACKING)
	return (spl_kmem_alloc_debug(size, flags, NUMA_NO_NODE));
#else
	return (spl_kmem_alloc_track(size, flags, func, line, NUMA_NO_NODE));
#endif
}
EXPORT_SYMBOL(spl_kmem_zalloc);

void
spl_kmem_free(const void *buf, size_t size)
{
#if !defined(DEBUG_KMEM)
	return (spl_kmem_free_impl(buf, size));
#elif !defined(DEBUG_KMEM_TRACKING)
	return (spl_kmem_free_debug(buf, size));
#else
	return (spl_kmem_free_track(buf, size));
#endif
}
EXPORT_SYMBOL(spl_kmem_free);

#if defined(DEBUG_KMEM) && defined(DEBUG_KMEM_TRACKING)
static char *
spl_sprintf_addr(kmem_debug_t *kd, char *str, int len, int min)
{
	int size = ((len - 1) < kd->kd_size) ? (len - 1) : kd->kd_size;
	int i, flag = 1;

	ASSERT(str != NULL && len >= 17);
	memset(str, 0, len);

	/*
	 * Check for a fully printable string, and while we are at
	 * it place the printable characters in the passed buffer.
	 */
	for (i = 0; i < size; i++) {
		str[i] = ((char *)(kd->kd_addr))[i];
		if (isprint(str[i])) {
			continue;
		} else {
			/*
			 * Minimum number of printable characters found
			 * to make it worthwhile to print this as ascii.
			 */
			if (i > min)
				break;

			flag = 0;
			break;
		}
	}

	if (!flag) {
		sprintf(str, "%02x%02x%02x%02x%02x%02x%02x%02x",
		    *((uint8_t *)kd->kd_addr),
		    *((uint8_t *)kd->kd_addr + 2),
		    *((uint8_t *)kd->kd_addr + 4),
		    *((uint8_t *)kd->kd_addr + 6),
		    *((uint8_t *)kd->kd_addr + 8),
		    *((uint8_t *)kd->kd_addr + 10),
		    *((uint8_t *)kd->kd_addr + 12),
		    *((uint8_t *)kd->kd_addr + 14));
	}

	return (str);
}

static int
spl_kmem_init_tracking(struct list_head *list, spinlock_t *lock, int size)
{
	int i;

	spin_lock_init(lock);
	INIT_LIST_HEAD(list);

	for (i = 0; i < size; i++)
		INIT_HLIST_HEAD(&kmem_table[i]);

	return (0);
}

static void
spl_kmem_fini_tracking(struct list_head *list, spinlock_t *lock)
{
	unsigned long flags;
	kmem_debug_t *kd = NULL;
	char str[17];

	spin_lock_irqsave(lock, flags);
	if (!list_empty(list))
		printk(KERN_WARNING "%-16s %-5s %-16s %s:%s\n", "address",
		    "size", "data", "func", "line");

	list_for_each_entry(kd, list, kd_list) {
		printk(KERN_WARNING "%p %-5d %-16s %s:%d\n", kd->kd_addr,
		    (int)kd->kd_size, spl_sprintf_addr(kd, str, 17, 8),
		    kd->kd_func, kd->kd_line);
	}

	spin_unlock_irqrestore(lock, flags);
}
#endif /* DEBUG_KMEM && DEBUG_KMEM_TRACKING */

int
spl_kmem_init(void)
{

#ifdef DEBUG_KMEM
	kmem_alloc_used_set(0);



#ifdef DEBUG_KMEM_TRACKING
	spl_kmem_init_tracking(&kmem_list, &kmem_lock, KMEM_TABLE_SIZE);
#endif /* DEBUG_KMEM_TRACKING */
#endif /* DEBUG_KMEM */

	return (0);
}

void
spl_kmem_fini(void)
{
#ifdef DEBUG_KMEM
	/*
	 * Display all unreclaimed memory addresses, including the
	 * allocation size and the first few bytes of what's located
	 * at that address to aid in debugging.  Performance is not
	 * a serious concern here since it is module unload time.
	 */
	if (kmem_alloc_used_read() != 0)
		printk(KERN_WARNING "kmem leaked %ld/%llu bytes\n",
		    (unsigned long)kmem_alloc_used_read(), kmem_alloc_max);

#ifdef DEBUG_KMEM_TRACKING
	spl_kmem_fini_tracking(&kmem_list, &kmem_lock);
#endif /* DEBUG_KMEM_TRACKING */
#endif /* DEBUG_KMEM */
}