aboutsummaryrefslogtreecommitdiffstats
path: root/module/zstd/zfs_zstd.c
blob: dfcd938aea323e3e8b496a1765ba67253b950af7 (plain)
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
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
/*
 * BSD 3-Clause New License (https://spdx.org/licenses/BSD-3-Clause.html)
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are met:
 *
 * 1. Redistributions of source code must retain the above copyright notice,
 * this list of conditions and the following disclaimer.
 *
 * 2. Redistributions in binary form must reproduce the above copyright notice,
 * this list of conditions and the following disclaimer in the documentation
 * and/or other materials provided with the distribution.
 *
 * 3. Neither the name of the copyright holder nor the names of its
 * contributors may be used to endorse or promote products derived from this
 * software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
 * POSSIBILITY OF SUCH DAMAGE.
 */

/*
 * Copyright (c) 2016-2018, Klara Inc.
 * Copyright (c) 2016-2018, Allan Jude
 * Copyright (c) 2018-2020, Sebastian Gottschall
 * Copyright (c) 2019-2020, Michael Niewöhner
 * Copyright (c) 2020, The FreeBSD Foundation [1]
 *
 * [1] Portions of this software were developed by Allan Jude
 *     under sponsorship from the FreeBSD Foundation.
 */

#include <sys/param.h>
#include <sys/sysmacros.h>
#include <sys/zfs_context.h>
#include <sys/zio_compress.h>
#include <sys/spa.h>
#include <sys/zstd/zstd.h>

#define	ZSTD_STATIC_LINKING_ONLY
#include "lib/zstd.h"
#include "lib/zstd_errors.h"

kstat_t *zstd_ksp = NULL;

typedef struct zstd_stats {
	kstat_named_t	zstd_stat_alloc_fail;
	kstat_named_t	zstd_stat_alloc_fallback;
	kstat_named_t	zstd_stat_com_alloc_fail;
	kstat_named_t	zstd_stat_dec_alloc_fail;
	kstat_named_t	zstd_stat_com_inval;
	kstat_named_t	zstd_stat_dec_inval;
	kstat_named_t	zstd_stat_dec_header_inval;
	kstat_named_t	zstd_stat_com_fail;
	kstat_named_t	zstd_stat_dec_fail;
	kstat_named_t	zstd_stat_buffers;
	kstat_named_t	zstd_stat_size;
} zstd_stats_t;

static zstd_stats_t zstd_stats = {
	{ "alloc_fail",			KSTAT_DATA_UINT64 },
	{ "alloc_fallback",		KSTAT_DATA_UINT64 },
	{ "compress_alloc_fail",	KSTAT_DATA_UINT64 },
	{ "decompress_alloc_fail",	KSTAT_DATA_UINT64 },
	{ "compress_level_invalid",	KSTAT_DATA_UINT64 },
	{ "decompress_level_invalid",	KSTAT_DATA_UINT64 },
	{ "decompress_header_invalid",	KSTAT_DATA_UINT64 },
	{ "compress_failed",		KSTAT_DATA_UINT64 },
	{ "decompress_failed",		KSTAT_DATA_UINT64 },
	{ "buffers",			KSTAT_DATA_UINT64 },
	{ "size",			KSTAT_DATA_UINT64 },
};

/* Enums describing the allocator type specified by kmem_type in zstd_kmem */
enum zstd_kmem_type {
	ZSTD_KMEM_UNKNOWN = 0,
	/* Allocation type using kmem_vmalloc */
	ZSTD_KMEM_DEFAULT,
	/* Pool based allocation using mempool_alloc */
	ZSTD_KMEM_POOL,
	/* Reserved fallback memory for decompression only */
	ZSTD_KMEM_DCTX,
	ZSTD_KMEM_COUNT,
};

/* Structure for pooled memory objects */
struct zstd_pool {
	void *mem;
	size_t size;
	kmutex_t barrier;
	hrtime_t timeout;
};

/* Global structure for handling memory allocations */
struct zstd_kmem {
	enum zstd_kmem_type kmem_type;
	size_t kmem_size;
	struct zstd_pool *pool;
};

/* Fallback memory structure used for decompression only if memory runs out */
struct zstd_fallback_mem {
	size_t mem_size;
	void *mem;
	kmutex_t barrier;
};

struct zstd_levelmap {
	int16_t zstd_level;
	enum zio_zstd_levels level;
};

/*
 * ZSTD memory handlers
 *
 * For decompression we use a different handler which also provides fallback
 * memory allocation in case memory runs out.
 *
 * The ZSTD handlers were split up for the most simplified implementation.
 */
static void *zstd_alloc(void *opaque, size_t size);
static void *zstd_dctx_alloc(void *opaque, size_t size);
static void zstd_free(void *opaque, void *ptr);

/* Compression memory handler */
static const ZSTD_customMem zstd_malloc = {
	zstd_alloc,
	zstd_free,
	NULL,
};

/* Decompression memory handler */
static const ZSTD_customMem zstd_dctx_malloc = {
	zstd_dctx_alloc,
	zstd_free,
	NULL,
};

/* Level map for converting ZFS internal levels to ZSTD levels and vice versa */
static struct zstd_levelmap zstd_levels[] = {
	{ZIO_ZSTD_LEVEL_1, ZIO_ZSTD_LEVEL_1},
	{ZIO_ZSTD_LEVEL_2, ZIO_ZSTD_LEVEL_2},
	{ZIO_ZSTD_LEVEL_3, ZIO_ZSTD_LEVEL_3},
	{ZIO_ZSTD_LEVEL_4, ZIO_ZSTD_LEVEL_4},
	{ZIO_ZSTD_LEVEL_5, ZIO_ZSTD_LEVEL_5},
	{ZIO_ZSTD_LEVEL_6, ZIO_ZSTD_LEVEL_6},
	{ZIO_ZSTD_LEVEL_7, ZIO_ZSTD_LEVEL_7},
	{ZIO_ZSTD_LEVEL_8, ZIO_ZSTD_LEVEL_8},
	{ZIO_ZSTD_LEVEL_9, ZIO_ZSTD_LEVEL_9},
	{ZIO_ZSTD_LEVEL_10, ZIO_ZSTD_LEVEL_10},
	{ZIO_ZSTD_LEVEL_11, ZIO_ZSTD_LEVEL_11},
	{ZIO_ZSTD_LEVEL_12, ZIO_ZSTD_LEVEL_12},
	{ZIO_ZSTD_LEVEL_13, ZIO_ZSTD_LEVEL_13},
	{ZIO_ZSTD_LEVEL_14, ZIO_ZSTD_LEVEL_14},
	{ZIO_ZSTD_LEVEL_15, ZIO_ZSTD_LEVEL_15},
	{ZIO_ZSTD_LEVEL_16, ZIO_ZSTD_LEVEL_16},
	{ZIO_ZSTD_LEVEL_17, ZIO_ZSTD_LEVEL_17},
	{ZIO_ZSTD_LEVEL_18, ZIO_ZSTD_LEVEL_18},
	{ZIO_ZSTD_LEVEL_19, ZIO_ZSTD_LEVEL_19},
	{-1, ZIO_ZSTD_LEVEL_FAST_1},
	{-2, ZIO_ZSTD_LEVEL_FAST_2},
	{-3, ZIO_ZSTD_LEVEL_FAST_3},
	{-4, ZIO_ZSTD_LEVEL_FAST_4},
	{-5, ZIO_ZSTD_LEVEL_FAST_5},
	{-6, ZIO_ZSTD_LEVEL_FAST_6},
	{-7, ZIO_ZSTD_LEVEL_FAST_7},
	{-8, ZIO_ZSTD_LEVEL_FAST_8},
	{-9, ZIO_ZSTD_LEVEL_FAST_9},
	{-10, ZIO_ZSTD_LEVEL_FAST_10},
	{-20, ZIO_ZSTD_LEVEL_FAST_20},
	{-30, ZIO_ZSTD_LEVEL_FAST_30},
	{-40, ZIO_ZSTD_LEVEL_FAST_40},
	{-50, ZIO_ZSTD_LEVEL_FAST_50},
	{-60, ZIO_ZSTD_LEVEL_FAST_60},
	{-70, ZIO_ZSTD_LEVEL_FAST_70},
	{-80, ZIO_ZSTD_LEVEL_FAST_80},
	{-90, ZIO_ZSTD_LEVEL_FAST_90},
	{-100, ZIO_ZSTD_LEVEL_FAST_100},
	{-500, ZIO_ZSTD_LEVEL_FAST_500},
	{-1000, ZIO_ZSTD_LEVEL_FAST_1000},
};

/*
 * This variable represents the maximum count of the pool based on the number
 * of CPUs plus some buffer. We default to cpu count * 4, see init_zstd.
 */
static int pool_count = 16;

#define	ZSTD_POOL_MAX		pool_count
#define	ZSTD_POOL_TIMEOUT	60 * 2

static struct zstd_fallback_mem zstd_dctx_fallback;
static struct zstd_pool *zstd_mempool_cctx;
static struct zstd_pool *zstd_mempool_dctx;


static void
zstd_mempool_reap(struct zstd_pool *zstd_mempool)
{
	struct zstd_pool *pool;

	if (!zstd_mempool || !ZSTDSTAT(zstd_stat_buffers)) {
		return;
	}

	/* free obsolete slots */
	for (int i = 0; i < ZSTD_POOL_MAX; i++) {
		pool = &zstd_mempool[i];
		if (pool->mem && mutex_tryenter(&pool->barrier)) {
			/* Free memory if unused object older than 2 minutes */
			if (pool->mem && gethrestime_sec() > pool->timeout) {
				vmem_free(pool->mem, pool->size);
				ZSTDSTAT_SUB(zstd_stat_buffers, 1);
				ZSTDSTAT_SUB(zstd_stat_size, pool->size);
				pool->mem = NULL;
				pool->size = 0;
				pool->timeout = 0;
			}
			mutex_exit(&pool->barrier);
		}
	}
}

/*
 * Try to get a cached allocated buffer from memory pool or allocate a new one
 * if necessary. If a object is older than 2 minutes and does not fit the
 * requested size, it will be released and a new cached entry will be allocated.
 * If other pooled objects are detected without being used for 2 minutes, they
 * will be released, too.
 *
 * The concept is that high frequency memory allocations of bigger objects are
 * expensive. So if a lot of work is going on, allocations will be kept for a
 * while and can be reused in that time frame.
 *
 * The scheduled release will be updated every time a object is reused.
 */

static void *
zstd_mempool_alloc(struct zstd_pool *zstd_mempool, size_t size)
{
	struct zstd_pool *pool;
	struct zstd_kmem *mem = NULL;

	if (!zstd_mempool) {
		return (NULL);
	}

	/* Seek for preallocated memory slot and free obsolete slots */
	for (int i = 0; i < ZSTD_POOL_MAX; i++) {
		pool = &zstd_mempool[i];
		/*
		 * This lock is simply a marker for a pool object beeing in use.
		 * If it's already hold, it will be skipped.
		 *
		 * We need to create it before checking it to avoid race
		 * conditions caused by running in a threaded context.
		 *
		 * The lock is later released by zstd_mempool_free.
		 */
		if (mutex_tryenter(&pool->barrier)) {
			/*
			 * Check if objects fits the size, if so we take it and
			 * update the timestamp.
			 */
			if (pool->mem && size <= pool->size) {
				pool->timeout = gethrestime_sec() +
				    ZSTD_POOL_TIMEOUT;
				mem = pool->mem;
				return (mem);
			}
			mutex_exit(&pool->barrier);
		}
	}

	/*
	 * If no preallocated slot was found, try to fill in a new one.
	 *
	 * We run a similar algorithm twice here to avoid pool fragmentation.
	 * The first one may generate holes in the list if objects get released.
	 * We always make sure that these holes get filled instead of adding new
	 * allocations constantly at the end.
	 */
	for (int i = 0; i < ZSTD_POOL_MAX; i++) {
		pool = &zstd_mempool[i];
		if (mutex_tryenter(&pool->barrier)) {
			/* Object is free, try to allocate new one */
			if (!pool->mem) {
				mem = vmem_alloc(size, KM_SLEEP);
				if (mem) {
					ZSTDSTAT_ADD(zstd_stat_buffers, 1);
					ZSTDSTAT_ADD(zstd_stat_size, size);
					pool->mem = mem;
					pool->size = size;
					/* Keep track for later release */
					mem->pool = pool;
					mem->kmem_type = ZSTD_KMEM_POOL;
					mem->kmem_size = size;
				}
			}

			if (size <= pool->size) {
				/* Update timestamp */
				pool->timeout = gethrestime_sec() +
				    ZSTD_POOL_TIMEOUT;

				return (pool->mem);
			}

			mutex_exit(&pool->barrier);
		}
	}

	/*
	 * If the pool is full or the allocation failed, try lazy allocation
	 * instead.
	 */
	if (!mem) {
		mem = vmem_alloc(size, KM_NOSLEEP);
		if (mem) {
			mem->pool = NULL;
			mem->kmem_type = ZSTD_KMEM_DEFAULT;
			mem->kmem_size = size;
		}
	}

	return (mem);
}

/* Mark object as released by releasing the barrier mutex */
static void
zstd_mempool_free(struct zstd_kmem *z)
{
	mutex_exit(&z->pool->barrier);
}

/* Convert ZFS internal enum to ZSTD level */
static int
zstd_enum_to_level(enum zio_zstd_levels level, int16_t *zstd_level)
{
	if (level > 0 && level <= ZIO_ZSTD_LEVEL_19) {
		*zstd_level = zstd_levels[level - 1].zstd_level;
		return (0);
	}
	if (level >= ZIO_ZSTD_LEVEL_FAST_1 &&
	    level <= ZIO_ZSTD_LEVEL_FAST_1000) {
		*zstd_level = zstd_levels[level - ZIO_ZSTD_LEVEL_FAST_1
		    + ZIO_ZSTD_LEVEL_19].zstd_level;
		return (0);
	}

	/* Invalid/unknown zfs compression enum - this should never happen. */
	return (1);
}

/* Compress block using zstd */
size_t
zfs_zstd_compress(void *s_start, void *d_start, size_t s_len, size_t d_len,
    int level)
{
	size_t c_len;
	int16_t zstd_level;
	zfs_zstdhdr_t *hdr;
	ZSTD_CCtx *cctx;

	hdr = (zfs_zstdhdr_t *)d_start;

	/* Skip compression if the specified level is invalid */
	if (zstd_enum_to_level(level, &zstd_level)) {
		ZSTDSTAT_BUMP(zstd_stat_com_inval);
		return (s_len);
	}

	ASSERT3U(d_len, >=, sizeof (*hdr));
	ASSERT3U(d_len, <=, s_len);
	ASSERT3U(zstd_level, !=, 0);

	cctx = ZSTD_createCCtx_advanced(zstd_malloc);

	/*
	 * Out of kernel memory, gently fall through - this will disable
	 * compression in zio_compress_data
	 */
	if (!cctx) {
		ZSTDSTAT_BUMP(zstd_stat_com_alloc_fail);
		return (s_len);
	}

	/* Set the compression level */
	ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, zstd_level);

	/* Use the "magicless" zstd header which saves us 4 header bytes */
	ZSTD_CCtx_setParameter(cctx, ZSTD_c_format, ZSTD_f_zstd1_magicless);

	/*
	 * Disable redundant checksum calculation and content size storage since
	 * this is already done by ZFS itself.
	 */
	ZSTD_CCtx_setParameter(cctx, ZSTD_c_checksumFlag, 0);
	ZSTD_CCtx_setParameter(cctx, ZSTD_c_contentSizeFlag, 0);

	c_len = ZSTD_compress2(cctx,
	    hdr->data,
	    d_len - sizeof (*hdr),
	    s_start, s_len);

	ZSTD_freeCCtx(cctx);

	/* Error in the compression routine, disable compression. */
	if (ZSTD_isError(c_len)) {
		/*
		 * If we are aborting the compression because the saves are
		 * too small, that is not a failure. Everything else is a
		 * failure, so increment the compression failure counter.
		 */
		if (ZSTD_getErrorCode(c_len) != ZSTD_error_dstSize_tooSmall) {
			ZSTDSTAT_BUMP(zstd_stat_com_fail);
		}
		return (s_len);
	}

	/*
	 * Encode the compressed buffer size at the start. We'll need this in
	 * decompression to counter the effects of padding which might be added
	 * to the compressed buffer and which, if unhandled, would confuse the
	 * hell out of our decompression function.
	 */
	hdr->c_len = BE_32(c_len);

	/*
	 * Check version for overflow.
	 * The limit of 24 bits must not be exceeded. This allows a maximum
	 * version 1677.72.15 which we don't expect to be ever reached.
	 */
	ASSERT3U(ZSTD_VERSION_NUMBER, <=, 0xFFFFFF);

	/*
	 * Encode the compression level as well. We may need to know the
	 * original compression level if compressed_arc is disabled, to match
	 * the compression settings to write this block to the L2ARC.
	 *
	 * Encode the actual level, so if the enum changes in the future, we
	 * will be compatible.
	 *
	 * The upper 24 bits store the ZSTD version to be able to provide
	 * future compatibility, since new versions might enhance the
	 * compression algorithm in a way, where the compressed data will
	 * change.
	 *
	 * As soon as such incompatibility occurs, handling code needs to be
	 * added, differentiating between the versions.
	 */
	hdr->version = ZSTD_VERSION_NUMBER;
	hdr->level = level;
	hdr->raw_version_level = BE_32(hdr->raw_version_level);

	return (c_len + sizeof (*hdr));
}

/* Decompress block using zstd and return its stored level */
int
zfs_zstd_decompress_level(void *s_start, void *d_start, size_t s_len,
    size_t d_len, uint8_t *level)
{
	ZSTD_DCtx *dctx;
	size_t result;
	int16_t zstd_level;
	uint32_t c_len;
	const zfs_zstdhdr_t *hdr;
	zfs_zstdhdr_t hdr_copy;

	hdr = (const zfs_zstdhdr_t *)s_start;
	c_len = BE_32(hdr->c_len);

	/*
	 * Make a copy instead of directly converting the header, since we must
	 * not modify the original data that may be used again later.
	 */
	hdr_copy.raw_version_level = BE_32(hdr->raw_version_level);

	/*
	 * NOTE: We ignore the ZSTD version for now. As soon as any
	 * incompatibility occurrs, it has to be handled accordingly.
	 * The version can be accessed via `hdr_copy.version`.
	 */

	/*
	 * Convert and check the level
	 * An invalid level is a strong indicator for data corruption! In such
	 * case return an error so the upper layers can try to fix it.
	 */
	if (zstd_enum_to_level(hdr_copy.level, &zstd_level)) {
		ZSTDSTAT_BUMP(zstd_stat_dec_inval);
		return (1);
	}

	ASSERT3U(d_len, >=, s_len);
	ASSERT3U(hdr_copy.level, !=, ZIO_COMPLEVEL_INHERIT);

	/* Invalid compressed buffer size encoded at start */
	if (c_len + sizeof (*hdr) > s_len) {
		ZSTDSTAT_BUMP(zstd_stat_dec_header_inval);
		return (1);
	}

	dctx = ZSTD_createDCtx_advanced(zstd_dctx_malloc);
	if (!dctx) {
		ZSTDSTAT_BUMP(zstd_stat_dec_alloc_fail);
		return (1);
	}

	/* Set header type to "magicless" */
	ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, ZSTD_f_zstd1_magicless);

	/* Decompress the data and release the context */
	result = ZSTD_decompressDCtx(dctx, d_start, d_len, hdr->data, c_len);
	ZSTD_freeDCtx(dctx);

	/*
	 * Returns 0 on success (decompression function returned non-negative)
	 * and non-zero on failure (decompression function returned negative.
	 */
	if (ZSTD_isError(result)) {
		ZSTDSTAT_BUMP(zstd_stat_dec_fail);
		return (1);
	}

	if (level) {
		*level = hdr_copy.level;
	}

	return (0);
}

/* Decompress datablock using zstd */
int
zfs_zstd_decompress(void *s_start, void *d_start, size_t s_len, size_t d_len,
    int level __maybe_unused)
{

	return (zfs_zstd_decompress_level(s_start, d_start, s_len, d_len,
	    NULL));
}

/* Allocator for zstd compression context using mempool_allocator */
static void *
zstd_alloc(void *opaque __maybe_unused, size_t size)
{
	size_t nbytes = sizeof (struct zstd_kmem) + size;
	struct zstd_kmem *z = NULL;

	z = (struct zstd_kmem *)zstd_mempool_alloc(zstd_mempool_cctx, nbytes);

	if (!z) {
		ZSTDSTAT_BUMP(zstd_stat_alloc_fail);
		return (NULL);
	}

	return ((void*)z + (sizeof (struct zstd_kmem)));
}

/*
 * Allocator for zstd decompression context using mempool_allocator with
 * fallback to reserved memory if allocation fails
 */
static void *
zstd_dctx_alloc(void *opaque __maybe_unused, size_t size)
{
	size_t nbytes = sizeof (struct zstd_kmem) + size;
	struct zstd_kmem *z = NULL;
	enum zstd_kmem_type type = ZSTD_KMEM_DEFAULT;

	z = (struct zstd_kmem *)zstd_mempool_alloc(zstd_mempool_dctx, nbytes);
	if (!z) {
		/* Try harder, decompression shall not fail */
		z = vmem_alloc(nbytes, KM_SLEEP);
		if (z) {
			z->pool = NULL;
		}
		ZSTDSTAT_BUMP(zstd_stat_alloc_fail);
	} else {
		return ((void*)z + (sizeof (struct zstd_kmem)));
	}

	/* Fallback if everything fails */
	if (!z) {
		/*
		 * Barrier since we only can handle it in a single thread. All
		 * other following threads need to wait here until decompression
		 * is completed. zstd_free will release this barrier later.
		 */
		mutex_enter(&zstd_dctx_fallback.barrier);

		z = zstd_dctx_fallback.mem;
		type = ZSTD_KMEM_DCTX;
		ZSTDSTAT_BUMP(zstd_stat_alloc_fallback);
	}

	/* Allocation should always be successful */
	if (!z) {
		return (NULL);
	}

	z->kmem_type = type;
	z->kmem_size = nbytes;

	return ((void*)z + (sizeof (struct zstd_kmem)));
}

/* Free allocated memory by its specific type */
static void
zstd_free(void *opaque __maybe_unused, void *ptr)
{
	struct zstd_kmem *z = (ptr - sizeof (struct zstd_kmem));
	enum zstd_kmem_type type;

	ASSERT3U(z->kmem_type, <, ZSTD_KMEM_COUNT);
	ASSERT3U(z->kmem_type, >, ZSTD_KMEM_UNKNOWN);

	type = z->kmem_type;
	switch (type) {
	case ZSTD_KMEM_DEFAULT:
		vmem_free(z, z->kmem_size);
		break;
	case ZSTD_KMEM_POOL:
		zstd_mempool_free(z);
		break;
	case ZSTD_KMEM_DCTX:
		mutex_exit(&zstd_dctx_fallback.barrier);
		break;
	default:
		break;
	}
}

/* Allocate fallback memory to ensure safe decompression */
static void __init
create_fallback_mem(struct zstd_fallback_mem *mem, size_t size)
{
	mem->mem_size = size;
	mem->mem = vmem_zalloc(mem->mem_size, KM_SLEEP);
	mutex_init(&mem->barrier, NULL, MUTEX_DEFAULT, NULL);
}

/* Initialize memory pool barrier mutexes */
static void __init
zstd_mempool_init(void)
{
	zstd_mempool_cctx = (struct zstd_pool *)
	    kmem_zalloc(ZSTD_POOL_MAX * sizeof (struct zstd_pool), KM_SLEEP);
	zstd_mempool_dctx = (struct zstd_pool *)
	    kmem_zalloc(ZSTD_POOL_MAX * sizeof (struct zstd_pool), KM_SLEEP);

	for (int i = 0; i < ZSTD_POOL_MAX; i++) {
		mutex_init(&zstd_mempool_cctx[i].barrier, NULL,
		    MUTEX_DEFAULT, NULL);
		mutex_init(&zstd_mempool_dctx[i].barrier, NULL,
		    MUTEX_DEFAULT, NULL);
	}
}

/* Initialize zstd-related memory handling */
static int __init
zstd_meminit(void)
{
	zstd_mempool_init();

	/*
	 * Estimate the size of the fallback decompression context.
	 * The expected size on x64 with current ZSTD should be about 160 KB.
	 */
	create_fallback_mem(&zstd_dctx_fallback,
	    P2ROUNDUP(ZSTD_estimateDCtxSize() + sizeof (struct zstd_kmem),
	    PAGESIZE));

	return (0);
}

/* Release object from pool and free memory */
static void __exit
release_pool(struct zstd_pool *pool)
{
	mutex_destroy(&pool->barrier);
	vmem_free(pool->mem, pool->size);
	pool->mem = NULL;
	pool->size = 0;
}

/* Release memory pool objects */
static void __exit
zstd_mempool_deinit(void)
{
	for (int i = 0; i < ZSTD_POOL_MAX; i++) {
		release_pool(&zstd_mempool_cctx[i]);
		release_pool(&zstd_mempool_dctx[i]);
	}

	kmem_free(zstd_mempool_dctx, ZSTD_POOL_MAX * sizeof (struct zstd_pool));
	kmem_free(zstd_mempool_cctx, ZSTD_POOL_MAX * sizeof (struct zstd_pool));
	zstd_mempool_dctx = NULL;
	zstd_mempool_cctx = NULL;
}

/* release unused memory from pool */

void
zfs_zstd_cache_reap_now(void)
{
	/*
	 * calling alloc with zero size seeks
	 * and releases old unused objects
	 */
	zstd_mempool_reap(zstd_mempool_cctx);
	zstd_mempool_reap(zstd_mempool_dctx);
}

extern int __init
zstd_init(void)
{
	/* Set pool size by using maximum sane thread count * 4 */
	pool_count = (boot_ncpus * 4);
	zstd_meminit();

	/* Initialize kstat */
	zstd_ksp = kstat_create("zfs", 0, "zstd", "misc",
	    KSTAT_TYPE_NAMED, sizeof (zstd_stats) / sizeof (kstat_named_t),
	    KSTAT_FLAG_VIRTUAL);
	if (zstd_ksp != NULL) {
		zstd_ksp->ks_data = &zstd_stats;
		kstat_install(zstd_ksp);
	}

	return (0);
}

extern void __exit
zstd_fini(void)
{
	/* Deinitialize kstat */
	if (zstd_ksp != NULL) {
		kstat_delete(zstd_ksp);
		zstd_ksp = NULL;
	}

	/* Release fallback memory */
	vmem_free(zstd_dctx_fallback.mem, zstd_dctx_fallback.mem_size);
	mutex_destroy(&zstd_dctx_fallback.barrier);

	/* Deinit memory pool */
	zstd_mempool_deinit();
}

#if defined(_KERNEL)
module_init(zstd_init);
module_exit(zstd_fini);

ZFS_MODULE_DESCRIPTION("ZSTD Compression for ZFS");
ZFS_MODULE_LICENSE("Dual BSD/GPL");
ZFS_MODULE_VERSION(ZSTD_VERSION_STRING);

EXPORT_SYMBOL(zfs_zstd_compress);
EXPORT_SYMBOL(zfs_zstd_decompress_level);
EXPORT_SYMBOL(zfs_zstd_decompress);
EXPORT_SYMBOL(zfs_zstd_cache_reap_now);
#endif