1 /* 2 * BSD 3-Clause New License (https://spdx.org/licenses/BSD-3-Clause.html) 3 * 4 * Redistribution and use in source and binary forms, with or without 5 * modification, are permitted provided that the following conditions are met: 6 * 7 * 1. Redistributions of source code must retain the above copyright notice, 8 * this list of conditions and the following disclaimer. 9 * 10 * 2. Redistributions in binary form must reproduce the above copyright notice, 11 * this list of conditions and the following disclaimer in the documentation 12 * and/or other materials provided with the distribution. 13 * 14 * 3. Neither the name of the copyright holder nor the names of its 15 * contributors may be used to endorse or promote products derived from this 16 * software without specific prior written permission. 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 19 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 21 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE 22 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 23 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 24 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 25 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 26 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 27 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 28 * POSSIBILITY OF SUCH DAMAGE. 29 */ 30 31 /* 32 * Copyright (c) 2016-2018, Klara Inc. 33 * Copyright (c) 2016-2018, Allan Jude 34 * Copyright (c) 2018-2020, Sebastian Gottschall 35 * Copyright (c) 2019-2020, Michael Niewöhner 36 * Copyright (c) 2020, The FreeBSD Foundation [1] 37 * 38 * [1] Portions of this software were developed by Allan Jude 39 * under sponsorship from the FreeBSD Foundation. 40 */ 41 42 #include <sys/param.h> 43 #include <sys/sysmacros.h> 44 #include <sys/zfs_context.h> 45 #include <sys/zio_compress.h> 46 #include <sys/spa.h> 47 #include <sys/zstd/zstd.h> 48 49 #define ZSTD_STATIC_LINKING_ONLY 50 #include "lib/zstd.h" 51 #include "lib/zstd_errors.h" 52 53 kstat_t *zstd_ksp = NULL; 54 55 typedef struct zstd_stats { 56 kstat_named_t zstd_stat_alloc_fail; 57 kstat_named_t zstd_stat_alloc_fallback; 58 kstat_named_t zstd_stat_com_alloc_fail; 59 kstat_named_t zstd_stat_dec_alloc_fail; 60 kstat_named_t zstd_stat_com_inval; 61 kstat_named_t zstd_stat_dec_inval; 62 kstat_named_t zstd_stat_dec_header_inval; 63 kstat_named_t zstd_stat_com_fail; 64 kstat_named_t zstd_stat_dec_fail; 65 kstat_named_t zstd_stat_buffers; 66 kstat_named_t zstd_stat_size; 67 } zstd_stats_t; 68 69 static zstd_stats_t zstd_stats = { 70 { "alloc_fail", KSTAT_DATA_UINT64 }, 71 { "alloc_fallback", KSTAT_DATA_UINT64 }, 72 { "compress_alloc_fail", KSTAT_DATA_UINT64 }, 73 { "decompress_alloc_fail", KSTAT_DATA_UINT64 }, 74 { "compress_level_invalid", KSTAT_DATA_UINT64 }, 75 { "decompress_level_invalid", KSTAT_DATA_UINT64 }, 76 { "decompress_header_invalid", KSTAT_DATA_UINT64 }, 77 { "compress_failed", KSTAT_DATA_UINT64 }, 78 { "decompress_failed", KSTAT_DATA_UINT64 }, 79 { "buffers", KSTAT_DATA_UINT64 }, 80 { "size", KSTAT_DATA_UINT64 }, 81 }; 82 83 /* Enums describing the allocator type specified by kmem_type in zstd_kmem */ 84 enum zstd_kmem_type { 85 ZSTD_KMEM_UNKNOWN = 0, 86 /* Allocation type using kmem_vmalloc */ 87 ZSTD_KMEM_DEFAULT, 88 /* Pool based allocation using mempool_alloc */ 89 ZSTD_KMEM_POOL, 90 /* Reserved fallback memory for decompression only */ 91 ZSTD_KMEM_DCTX, 92 ZSTD_KMEM_COUNT, 93 }; 94 95 /* Structure for pooled memory objects */ 96 struct zstd_pool { 97 void *mem; 98 size_t size; 99 kmutex_t barrier; 100 hrtime_t timeout; 101 }; 102 103 /* Global structure for handling memory allocations */ 104 struct zstd_kmem { 105 enum zstd_kmem_type kmem_type; 106 size_t kmem_size; 107 struct zstd_pool *pool; 108 }; 109 110 /* Fallback memory structure used for decompression only if memory runs out */ 111 struct zstd_fallback_mem { 112 size_t mem_size; 113 void *mem; 114 kmutex_t barrier; 115 }; 116 117 struct zstd_levelmap { 118 int16_t zstd_level; 119 enum zio_zstd_levels level; 120 }; 121 122 /* 123 * ZSTD memory handlers 124 * 125 * For decompression we use a different handler which also provides fallback 126 * memory allocation in case memory runs out. 127 * 128 * The ZSTD handlers were split up for the most simplified implementation. 129 */ 130 static void *zstd_alloc(void *opaque, size_t size); 131 static void *zstd_dctx_alloc(void *opaque, size_t size); 132 static void zstd_free(void *opaque, void *ptr); 133 134 /* Compression memory handler */ 135 static const ZSTD_customMem zstd_malloc = { 136 zstd_alloc, 137 zstd_free, 138 NULL, 139 }; 140 141 /* Decompression memory handler */ 142 static const ZSTD_customMem zstd_dctx_malloc = { 143 zstd_dctx_alloc, 144 zstd_free, 145 NULL, 146 }; 147 148 /* Level map for converting ZFS internal levels to ZSTD levels and vice versa */ 149 static struct zstd_levelmap zstd_levels[] = { 150 {ZIO_ZSTD_LEVEL_1, ZIO_ZSTD_LEVEL_1}, 151 {ZIO_ZSTD_LEVEL_2, ZIO_ZSTD_LEVEL_2}, 152 {ZIO_ZSTD_LEVEL_3, ZIO_ZSTD_LEVEL_3}, 153 {ZIO_ZSTD_LEVEL_4, ZIO_ZSTD_LEVEL_4}, 154 {ZIO_ZSTD_LEVEL_5, ZIO_ZSTD_LEVEL_5}, 155 {ZIO_ZSTD_LEVEL_6, ZIO_ZSTD_LEVEL_6}, 156 {ZIO_ZSTD_LEVEL_7, ZIO_ZSTD_LEVEL_7}, 157 {ZIO_ZSTD_LEVEL_8, ZIO_ZSTD_LEVEL_8}, 158 {ZIO_ZSTD_LEVEL_9, ZIO_ZSTD_LEVEL_9}, 159 {ZIO_ZSTD_LEVEL_10, ZIO_ZSTD_LEVEL_10}, 160 {ZIO_ZSTD_LEVEL_11, ZIO_ZSTD_LEVEL_11}, 161 {ZIO_ZSTD_LEVEL_12, ZIO_ZSTD_LEVEL_12}, 162 {ZIO_ZSTD_LEVEL_13, ZIO_ZSTD_LEVEL_13}, 163 {ZIO_ZSTD_LEVEL_14, ZIO_ZSTD_LEVEL_14}, 164 {ZIO_ZSTD_LEVEL_15, ZIO_ZSTD_LEVEL_15}, 165 {ZIO_ZSTD_LEVEL_16, ZIO_ZSTD_LEVEL_16}, 166 {ZIO_ZSTD_LEVEL_17, ZIO_ZSTD_LEVEL_17}, 167 {ZIO_ZSTD_LEVEL_18, ZIO_ZSTD_LEVEL_18}, 168 {ZIO_ZSTD_LEVEL_19, ZIO_ZSTD_LEVEL_19}, 169 {-1, ZIO_ZSTD_LEVEL_FAST_1}, 170 {-2, ZIO_ZSTD_LEVEL_FAST_2}, 171 {-3, ZIO_ZSTD_LEVEL_FAST_3}, 172 {-4, ZIO_ZSTD_LEVEL_FAST_4}, 173 {-5, ZIO_ZSTD_LEVEL_FAST_5}, 174 {-6, ZIO_ZSTD_LEVEL_FAST_6}, 175 {-7, ZIO_ZSTD_LEVEL_FAST_7}, 176 {-8, ZIO_ZSTD_LEVEL_FAST_8}, 177 {-9, ZIO_ZSTD_LEVEL_FAST_9}, 178 {-10, ZIO_ZSTD_LEVEL_FAST_10}, 179 {-20, ZIO_ZSTD_LEVEL_FAST_20}, 180 {-30, ZIO_ZSTD_LEVEL_FAST_30}, 181 {-40, ZIO_ZSTD_LEVEL_FAST_40}, 182 {-50, ZIO_ZSTD_LEVEL_FAST_50}, 183 {-60, ZIO_ZSTD_LEVEL_FAST_60}, 184 {-70, ZIO_ZSTD_LEVEL_FAST_70}, 185 {-80, ZIO_ZSTD_LEVEL_FAST_80}, 186 {-90, ZIO_ZSTD_LEVEL_FAST_90}, 187 {-100, ZIO_ZSTD_LEVEL_FAST_100}, 188 {-500, ZIO_ZSTD_LEVEL_FAST_500}, 189 {-1000, ZIO_ZSTD_LEVEL_FAST_1000}, 190 }; 191 192 /* 193 * This variable represents the maximum count of the pool based on the number 194 * of CPUs plus some buffer. We default to cpu count * 4, see init_zstd. 195 */ 196 static int pool_count = 16; 197 198 #define ZSTD_POOL_MAX pool_count 199 #define ZSTD_POOL_TIMEOUT 60 * 2 200 201 static struct zstd_fallback_mem zstd_dctx_fallback; 202 static struct zstd_pool *zstd_mempool_cctx; 203 static struct zstd_pool *zstd_mempool_dctx; 204 205 206 static void 207 zstd_mempool_reap(struct zstd_pool *zstd_mempool) 208 { 209 struct zstd_pool *pool; 210 211 if (!zstd_mempool || !ZSTDSTAT(zstd_stat_buffers)) { 212 return; 213 } 214 215 /* free obsolete slots */ 216 for (int i = 0; i < ZSTD_POOL_MAX; i++) { 217 pool = &zstd_mempool[i]; 218 if (pool->mem && mutex_tryenter(&pool->barrier)) { 219 /* Free memory if unused object older than 2 minutes */ 220 if (pool->mem && gethrestime_sec() > pool->timeout) { 221 vmem_free(pool->mem, pool->size); 222 ZSTDSTAT_SUB(zstd_stat_buffers, 1); 223 ZSTDSTAT_SUB(zstd_stat_size, pool->size); 224 pool->mem = NULL; 225 pool->size = 0; 226 pool->timeout = 0; 227 } 228 mutex_exit(&pool->barrier); 229 } 230 } 231 } 232 233 /* 234 * Try to get a cached allocated buffer from memory pool or allocate a new one 235 * if necessary. If a object is older than 2 minutes and does not fit the 236 * requested size, it will be released and a new cached entry will be allocated. 237 * If other pooled objects are detected without being used for 2 minutes, they 238 * will be released, too. 239 * 240 * The concept is that high frequency memory allocations of bigger objects are 241 * expensive. So if a lot of work is going on, allocations will be kept for a 242 * while and can be reused in that time frame. 243 * 244 * The scheduled release will be updated every time a object is reused. 245 */ 246 247 static void * 248 zstd_mempool_alloc(struct zstd_pool *zstd_mempool, size_t size) 249 { 250 struct zstd_pool *pool; 251 struct zstd_kmem *mem = NULL; 252 253 if (!zstd_mempool) { 254 return (NULL); 255 } 256 257 /* Seek for preallocated memory slot and free obsolete slots */ 258 for (int i = 0; i < ZSTD_POOL_MAX; i++) { 259 pool = &zstd_mempool[i]; 260 /* 261 * This lock is simply a marker for a pool object beeing in use. 262 * If it's already hold, it will be skipped. 263 * 264 * We need to create it before checking it to avoid race 265 * conditions caused by running in a threaded context. 266 * 267 * The lock is later released by zstd_mempool_free. 268 */ 269 if (mutex_tryenter(&pool->barrier)) { 270 /* 271 * Check if objects fits the size, if so we take it and 272 * update the timestamp. 273 */ 274 if (pool->mem && size <= pool->size) { 275 pool->timeout = gethrestime_sec() + 276 ZSTD_POOL_TIMEOUT; 277 mem = pool->mem; 278 return (mem); 279 } 280 mutex_exit(&pool->barrier); 281 } 282 } 283 284 /* 285 * If no preallocated slot was found, try to fill in a new one. 286 * 287 * We run a similar algorithm twice here to avoid pool fragmentation. 288 * The first one may generate holes in the list if objects get released. 289 * We always make sure that these holes get filled instead of adding new 290 * allocations constantly at the end. 291 */ 292 for (int i = 0; i < ZSTD_POOL_MAX; i++) { 293 pool = &zstd_mempool[i]; 294 if (mutex_tryenter(&pool->barrier)) { 295 /* Object is free, try to allocate new one */ 296 if (!pool->mem) { 297 mem = vmem_alloc(size, KM_SLEEP); 298 if (mem) { 299 ZSTDSTAT_ADD(zstd_stat_buffers, 1); 300 ZSTDSTAT_ADD(zstd_stat_size, size); 301 pool->mem = mem; 302 pool->size = size; 303 /* Keep track for later release */ 304 mem->pool = pool; 305 mem->kmem_type = ZSTD_KMEM_POOL; 306 mem->kmem_size = size; 307 } 308 } 309 310 if (size <= pool->size) { 311 /* Update timestamp */ 312 pool->timeout = gethrestime_sec() + 313 ZSTD_POOL_TIMEOUT; 314 315 return (pool->mem); 316 } 317 318 mutex_exit(&pool->barrier); 319 } 320 } 321 322 /* 323 * If the pool is full or the allocation failed, try lazy allocation 324 * instead. 325 */ 326 if (!mem) { 327 mem = vmem_alloc(size, KM_NOSLEEP); 328 if (mem) { 329 mem->pool = NULL; 330 mem->kmem_type = ZSTD_KMEM_DEFAULT; 331 mem->kmem_size = size; 332 } 333 } 334 335 return (mem); 336 } 337 338 /* Mark object as released by releasing the barrier mutex */ 339 static void 340 zstd_mempool_free(struct zstd_kmem *z) 341 { 342 mutex_exit(&z->pool->barrier); 343 } 344 345 /* Convert ZFS internal enum to ZSTD level */ 346 static int 347 zstd_enum_to_level(enum zio_zstd_levels level, int16_t *zstd_level) 348 { 349 if (level > 0 && level <= ZIO_ZSTD_LEVEL_19) { 350 *zstd_level = zstd_levels[level - 1].zstd_level; 351 return (0); 352 } 353 if (level >= ZIO_ZSTD_LEVEL_FAST_1 && 354 level <= ZIO_ZSTD_LEVEL_FAST_1000) { 355 *zstd_level = zstd_levels[level - ZIO_ZSTD_LEVEL_FAST_1 356 + ZIO_ZSTD_LEVEL_19].zstd_level; 357 return (0); 358 } 359 360 /* Invalid/unknown zfs compression enum - this should never happen. */ 361 return (1); 362 } 363 364 /* Compress block using zstd */ 365 size_t 366 zfs_zstd_compress(void *s_start, void *d_start, size_t s_len, size_t d_len, 367 int level) 368 { 369 size_t c_len; 370 int16_t zstd_level; 371 zfs_zstdhdr_t *hdr; 372 ZSTD_CCtx *cctx; 373 374 hdr = (zfs_zstdhdr_t *)d_start; 375 376 /* Skip compression if the specified level is invalid */ 377 if (zstd_enum_to_level(level, &zstd_level)) { 378 ZSTDSTAT_BUMP(zstd_stat_com_inval); 379 return (s_len); 380 } 381 382 ASSERT3U(d_len, >=, sizeof (*hdr)); 383 ASSERT3U(d_len, <=, s_len); 384 ASSERT3U(zstd_level, !=, 0); 385 386 cctx = ZSTD_createCCtx_advanced(zstd_malloc); 387 388 /* 389 * Out of kernel memory, gently fall through - this will disable 390 * compression in zio_compress_data 391 */ 392 if (!cctx) { 393 ZSTDSTAT_BUMP(zstd_stat_com_alloc_fail); 394 return (s_len); 395 } 396 397 /* Set the compression level */ 398 ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, zstd_level); 399 400 /* Use the "magicless" zstd header which saves us 4 header bytes */ 401 ZSTD_CCtx_setParameter(cctx, ZSTD_c_format, ZSTD_f_zstd1_magicless); 402 403 /* 404 * Disable redundant checksum calculation and content size storage since 405 * this is already done by ZFS itself. 406 */ 407 ZSTD_CCtx_setParameter(cctx, ZSTD_c_checksumFlag, 0); 408 ZSTD_CCtx_setParameter(cctx, ZSTD_c_contentSizeFlag, 0); 409 410 c_len = ZSTD_compress2(cctx, 411 hdr->data, 412 d_len - sizeof (*hdr), 413 s_start, s_len); 414 415 ZSTD_freeCCtx(cctx); 416 417 /* Error in the compression routine, disable compression. */ 418 if (ZSTD_isError(c_len)) { 419 /* 420 * If we are aborting the compression because the saves are 421 * too small, that is not a failure. Everything else is a 422 * failure, so increment the compression failure counter. 423 */ 424 if (ZSTD_getErrorCode(c_len) != ZSTD_error_dstSize_tooSmall) { 425 ZSTDSTAT_BUMP(zstd_stat_com_fail); 426 } 427 return (s_len); 428 } 429 430 /* 431 * Encode the compressed buffer size at the start. We'll need this in 432 * decompression to counter the effects of padding which might be added 433 * to the compressed buffer and which, if unhandled, would confuse the 434 * hell out of our decompression function. 435 */ 436 hdr->c_len = BE_32(c_len); 437 438 /* 439 * Check version for overflow. 440 * The limit of 24 bits must not be exceeded. This allows a maximum 441 * version 1677.72.15 which we don't expect to be ever reached. 442 */ 443 ASSERT3U(ZSTD_VERSION_NUMBER, <=, 0xFFFFFF); 444 445 /* 446 * Encode the compression level as well. We may need to know the 447 * original compression level if compressed_arc is disabled, to match 448 * the compression settings to write this block to the L2ARC. 449 * 450 * Encode the actual level, so if the enum changes in the future, we 451 * will be compatible. 452 * 453 * The upper 24 bits store the ZSTD version to be able to provide 454 * future compatibility, since new versions might enhance the 455 * compression algorithm in a way, where the compressed data will 456 * change. 457 * 458 * As soon as such incompatibility occurs, handling code needs to be 459 * added, differentiating between the versions. 460 */ 461 hdr->version = ZSTD_VERSION_NUMBER; 462 hdr->level = level; 463 hdr->raw_version_level = BE_32(hdr->raw_version_level); 464 465 return (c_len + sizeof (*hdr)); 466 } 467 468 /* Decompress block using zstd and return its stored level */ 469 int 470 zfs_zstd_decompress_level(void *s_start, void *d_start, size_t s_len, 471 size_t d_len, uint8_t *level) 472 { 473 ZSTD_DCtx *dctx; 474 size_t result; 475 int16_t zstd_level; 476 uint32_t c_len; 477 const zfs_zstdhdr_t *hdr; 478 zfs_zstdhdr_t hdr_copy; 479 480 hdr = (const zfs_zstdhdr_t *)s_start; 481 c_len = BE_32(hdr->c_len); 482 483 /* 484 * Make a copy instead of directly converting the header, since we must 485 * not modify the original data that may be used again later. 486 */ 487 hdr_copy.raw_version_level = BE_32(hdr->raw_version_level); 488 489 /* 490 * NOTE: We ignore the ZSTD version for now. As soon as any 491 * incompatibility occurrs, it has to be handled accordingly. 492 * The version can be accessed via `hdr_copy.version`. 493 */ 494 495 /* 496 * Convert and check the level 497 * An invalid level is a strong indicator for data corruption! In such 498 * case return an error so the upper layers can try to fix it. 499 */ 500 if (zstd_enum_to_level(hdr_copy.level, &zstd_level)) { 501 ZSTDSTAT_BUMP(zstd_stat_dec_inval); 502 return (1); 503 } 504 505 ASSERT3U(d_len, >=, s_len); 506 ASSERT3U(hdr_copy.level, !=, ZIO_COMPLEVEL_INHERIT); 507 508 /* Invalid compressed buffer size encoded at start */ 509 if (c_len + sizeof (*hdr) > s_len) { 510 ZSTDSTAT_BUMP(zstd_stat_dec_header_inval); 511 return (1); 512 } 513 514 dctx = ZSTD_createDCtx_advanced(zstd_dctx_malloc); 515 if (!dctx) { 516 ZSTDSTAT_BUMP(zstd_stat_dec_alloc_fail); 517 return (1); 518 } 519 520 /* Set header type to "magicless" */ 521 ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, ZSTD_f_zstd1_magicless); 522 523 /* Decompress the data and release the context */ 524 result = ZSTD_decompressDCtx(dctx, d_start, d_len, hdr->data, c_len); 525 ZSTD_freeDCtx(dctx); 526 527 /* 528 * Returns 0 on success (decompression function returned non-negative) 529 * and non-zero on failure (decompression function returned negative. 530 */ 531 if (ZSTD_isError(result)) { 532 ZSTDSTAT_BUMP(zstd_stat_dec_fail); 533 return (1); 534 } 535 536 if (level) { 537 *level = hdr_copy.level; 538 } 539 540 return (0); 541 } 542 543 /* Decompress datablock using zstd */ 544 int 545 zfs_zstd_decompress(void *s_start, void *d_start, size_t s_len, size_t d_len, 546 int level __maybe_unused) 547 { 548 549 return (zfs_zstd_decompress_level(s_start, d_start, s_len, d_len, 550 NULL)); 551 } 552 553 /* Allocator for zstd compression context using mempool_allocator */ 554 static void * 555 zstd_alloc(void *opaque __maybe_unused, size_t size) 556 { 557 size_t nbytes = sizeof (struct zstd_kmem) + size; 558 struct zstd_kmem *z = NULL; 559 560 z = (struct zstd_kmem *)zstd_mempool_alloc(zstd_mempool_cctx, nbytes); 561 562 if (!z) { 563 ZSTDSTAT_BUMP(zstd_stat_alloc_fail); 564 return (NULL); 565 } 566 567 return ((void*)z + (sizeof (struct zstd_kmem))); 568 } 569 570 /* 571 * Allocator for zstd decompression context using mempool_allocator with 572 * fallback to reserved memory if allocation fails 573 */ 574 static void * 575 zstd_dctx_alloc(void *opaque __maybe_unused, size_t size) 576 { 577 size_t nbytes = sizeof (struct zstd_kmem) + size; 578 struct zstd_kmem *z = NULL; 579 enum zstd_kmem_type type = ZSTD_KMEM_DEFAULT; 580 581 z = (struct zstd_kmem *)zstd_mempool_alloc(zstd_mempool_dctx, nbytes); 582 if (!z) { 583 /* Try harder, decompression shall not fail */ 584 z = vmem_alloc(nbytes, KM_SLEEP); 585 if (z) { 586 z->pool = NULL; 587 } 588 ZSTDSTAT_BUMP(zstd_stat_alloc_fail); 589 } else { 590 return ((void*)z + (sizeof (struct zstd_kmem))); 591 } 592 593 /* Fallback if everything fails */ 594 if (!z) { 595 /* 596 * Barrier since we only can handle it in a single thread. All 597 * other following threads need to wait here until decompression 598 * is completed. zstd_free will release this barrier later. 599 */ 600 mutex_enter(&zstd_dctx_fallback.barrier); 601 602 z = zstd_dctx_fallback.mem; 603 type = ZSTD_KMEM_DCTX; 604 ZSTDSTAT_BUMP(zstd_stat_alloc_fallback); 605 } 606 607 /* Allocation should always be successful */ 608 if (!z) { 609 return (NULL); 610 } 611 612 z->kmem_type = type; 613 z->kmem_size = nbytes; 614 615 return ((void*)z + (sizeof (struct zstd_kmem))); 616 } 617 618 /* Free allocated memory by its specific type */ 619 static void 620 zstd_free(void *opaque __maybe_unused, void *ptr) 621 { 622 struct zstd_kmem *z = (ptr - sizeof (struct zstd_kmem)); 623 enum zstd_kmem_type type; 624 625 ASSERT3U(z->kmem_type, <, ZSTD_KMEM_COUNT); 626 ASSERT3U(z->kmem_type, >, ZSTD_KMEM_UNKNOWN); 627 628 type = z->kmem_type; 629 switch (type) { 630 case ZSTD_KMEM_DEFAULT: 631 vmem_free(z, z->kmem_size); 632 break; 633 case ZSTD_KMEM_POOL: 634 zstd_mempool_free(z); 635 break; 636 case ZSTD_KMEM_DCTX: 637 mutex_exit(&zstd_dctx_fallback.barrier); 638 break; 639 default: 640 break; 641 } 642 } 643 644 /* Allocate fallback memory to ensure safe decompression */ 645 static void __init 646 create_fallback_mem(struct zstd_fallback_mem *mem, size_t size) 647 { 648 mem->mem_size = size; 649 mem->mem = vmem_zalloc(mem->mem_size, KM_SLEEP); 650 mutex_init(&mem->barrier, NULL, MUTEX_DEFAULT, NULL); 651 } 652 653 /* Initialize memory pool barrier mutexes */ 654 static void __init 655 zstd_mempool_init(void) 656 { 657 zstd_mempool_cctx = (struct zstd_pool *) 658 kmem_zalloc(ZSTD_POOL_MAX * sizeof (struct zstd_pool), KM_SLEEP); 659 zstd_mempool_dctx = (struct zstd_pool *) 660 kmem_zalloc(ZSTD_POOL_MAX * sizeof (struct zstd_pool), KM_SLEEP); 661 662 for (int i = 0; i < ZSTD_POOL_MAX; i++) { 663 mutex_init(&zstd_mempool_cctx[i].barrier, NULL, 664 MUTEX_DEFAULT, NULL); 665 mutex_init(&zstd_mempool_dctx[i].barrier, NULL, 666 MUTEX_DEFAULT, NULL); 667 } 668 } 669 670 /* Initialize zstd-related memory handling */ 671 static int __init 672 zstd_meminit(void) 673 { 674 zstd_mempool_init(); 675 676 /* 677 * Estimate the size of the fallback decompression context. 678 * The expected size on x64 with current ZSTD should be about 160 KB. 679 */ 680 create_fallback_mem(&zstd_dctx_fallback, 681 P2ROUNDUP(ZSTD_estimateDCtxSize() + sizeof (struct zstd_kmem), 682 PAGESIZE)); 683 684 return (0); 685 } 686 687 /* Release object from pool and free memory */ 688 static void __exit 689 release_pool(struct zstd_pool *pool) 690 { 691 mutex_destroy(&pool->barrier); 692 vmem_free(pool->mem, pool->size); 693 pool->mem = NULL; 694 pool->size = 0; 695 } 696 697 /* Release memory pool objects */ 698 static void __exit 699 zstd_mempool_deinit(void) 700 { 701 for (int i = 0; i < ZSTD_POOL_MAX; i++) { 702 release_pool(&zstd_mempool_cctx[i]); 703 release_pool(&zstd_mempool_dctx[i]); 704 } 705 706 kmem_free(zstd_mempool_dctx, ZSTD_POOL_MAX * sizeof (struct zstd_pool)); 707 kmem_free(zstd_mempool_cctx, ZSTD_POOL_MAX * sizeof (struct zstd_pool)); 708 zstd_mempool_dctx = NULL; 709 zstd_mempool_cctx = NULL; 710 } 711 712 /* release unused memory from pool */ 713 714 void 715 zfs_zstd_cache_reap_now(void) 716 { 717 718 /* 719 * Short-circuit if there are no buffers to begin with. 720 */ 721 if (ZSTDSTAT(zstd_stat_buffers) == 0) 722 return; 723 724 /* 725 * calling alloc with zero size seeks 726 * and releases old unused objects 727 */ 728 zstd_mempool_reap(zstd_mempool_cctx); 729 zstd_mempool_reap(zstd_mempool_dctx); 730 } 731 732 extern int __init 733 zstd_init(void) 734 { 735 /* Set pool size by using maximum sane thread count * 4 */ 736 pool_count = (boot_ncpus * 4); 737 zstd_meminit(); 738 739 /* Initialize kstat */ 740 zstd_ksp = kstat_create("zfs", 0, "zstd", "misc", 741 KSTAT_TYPE_NAMED, sizeof (zstd_stats) / sizeof (kstat_named_t), 742 KSTAT_FLAG_VIRTUAL); 743 if (zstd_ksp != NULL) { 744 zstd_ksp->ks_data = &zstd_stats; 745 kstat_install(zstd_ksp); 746 } 747 748 return (0); 749 } 750 751 extern void __exit 752 zstd_fini(void) 753 { 754 /* Deinitialize kstat */ 755 if (zstd_ksp != NULL) { 756 kstat_delete(zstd_ksp); 757 zstd_ksp = NULL; 758 } 759 760 /* Release fallback memory */ 761 vmem_free(zstd_dctx_fallback.mem, zstd_dctx_fallback.mem_size); 762 mutex_destroy(&zstd_dctx_fallback.barrier); 763 764 /* Deinit memory pool */ 765 zstd_mempool_deinit(); 766 } 767 768 #if defined(_KERNEL) 769 module_init(zstd_init); 770 module_exit(zstd_fini); 771 772 ZFS_MODULE_DESCRIPTION("ZSTD Compression for ZFS"); 773 ZFS_MODULE_LICENSE("Dual BSD/GPL"); 774 ZFS_MODULE_VERSION(ZSTD_VERSION_STRING); 775 776 EXPORT_SYMBOL(zfs_zstd_compress); 777 EXPORT_SYMBOL(zfs_zstd_decompress_level); 778 EXPORT_SYMBOL(zfs_zstd_decompress); 779 EXPORT_SYMBOL(zfs_zstd_cache_reap_now); 780 #endif 781