1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or https://opensource.org/licenses/CDDL-1.0. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23 * Copyright (c) 2011, 2024 by Delphix. All rights reserved. 24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 25 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. 26 * Copyright 2013 Saso Kiselkov. All rights reserved. 27 * Copyright (c) 2014 Integros [integros.com] 28 * Copyright 2017 Joyent, Inc. 29 * Copyright (c) 2017, Intel Corporation. 30 * Copyright (c) 2019, Allan Jude 31 * Copyright (c) 2019, Klara Inc. 32 * Copyright (c) 2019, Datto Inc. 33 */ 34 35 #ifndef _SYS_SPA_H 36 #define _SYS_SPA_H 37 38 #include <sys/zfs_context.h> 39 #include <sys/avl.h> 40 #include <sys/kstat.h> 41 #include <sys/nvpair.h> 42 #include <sys/types.h> 43 #include <sys/fs/zfs.h> 44 #include <sys/spa_checksum.h> 45 #include <sys/dmu.h> 46 #include <sys/space_map.h> 47 #include <sys/bitops.h> 48 49 #ifdef __cplusplus 50 extern "C" { 51 #endif 52 53 /* 54 * Forward references that lots of things need. 55 */ 56 typedef struct brt_vdev brt_vdev_t; 57 typedef struct spa spa_t; 58 typedef struct vdev vdev_t; 59 typedef struct metaslab metaslab_t; 60 typedef struct metaslab_group metaslab_group_t; 61 typedef struct metaslab_class metaslab_class_t; 62 typedef struct zio zio_t; 63 typedef struct zilog zilog_t; 64 typedef struct spa_aux_vdev spa_aux_vdev_t; 65 typedef struct zbookmark_phys zbookmark_phys_t; 66 typedef struct zbookmark_err_phys zbookmark_err_phys_t; 67 68 struct bpobj; 69 struct bplist; 70 struct dsl_pool; 71 struct dsl_dataset; 72 struct dsl_crypto_params; 73 74 /* 75 * Alignment Shift (ashift) is an immutable, internal top-level vdev property 76 * which can only be set at vdev creation time. Physical writes are always done 77 * according to it, which makes 2^ashift the smallest possible IO on a vdev. 78 * 79 * We currently allow values ranging from 512 bytes (2^9 = 512) to 64 KiB 80 * (2^16 = 65,536). 81 */ 82 #define ASHIFT_MIN 9 83 #define ASHIFT_MAX 16 84 85 /* 86 * Size of block to hold the configuration data (a packed nvlist) 87 */ 88 #define SPA_CONFIG_BLOCKSIZE (1ULL << 14) 89 90 /* 91 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. 92 * The ASIZE encoding should be at least 64 times larger (6 more bits) 93 * to support up to 4-way RAID-Z mirror mode with worst-case gang block 94 * overhead, three DVAs per bp, plus one more bit in case we do anything 95 * else that expands the ASIZE. 96 */ 97 #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ 98 #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ 99 #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ 100 101 #define SPA_COMPRESSBITS 7 102 #define SPA_VDEVBITS 24 103 #define SPA_COMPRESSMASK ((1U << SPA_COMPRESSBITS) - 1) 104 105 /* 106 * All SPA data is represented by 128-bit data virtual addresses (DVAs). 107 * The members of the dva_t should be considered opaque outside the SPA. 108 */ 109 typedef struct dva { 110 uint64_t dva_word[2]; 111 } dva_t; 112 113 114 /* 115 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept 116 * secret and is suitable for use in MAC algorithms as the key. 117 */ 118 typedef struct zio_cksum_salt { 119 uint8_t zcs_bytes[32]; 120 } zio_cksum_salt_t; 121 122 /* 123 * Each block is described by its DVAs, time of birth, checksum, etc. 124 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 125 * 126 * 64 56 48 40 32 24 16 8 0 127 * +-------+-------+-------+-------+-------+-------+-------+-------+ 128 * 0 | pad | vdev1 | pad | ASIZE | 129 * +-------+-------+-------+-------+-------+-------+-------+-------+ 130 * 1 |G| offset1 | 131 * +-------+-------+-------+-------+-------+-------+-------+-------+ 132 * 2 | pad | vdev2 | pad | ASIZE | 133 * +-------+-------+-------+-------+-------+-------+-------+-------+ 134 * 3 |G| offset2 | 135 * +-------+-------+-------+-------+-------+-------+-------+-------+ 136 * 4 | pad | vdev3 | pad | ASIZE | 137 * +-------+-------+-------+-------+-------+-------+-------+-------+ 138 * 5 |G| offset3 | 139 * +-------+-------+-------+-------+-------+-------+-------+-------+ 140 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 141 * +-------+-------+-------+-------+-------+-------+-------+-------+ 142 * 7 | padding | 143 * +-------+-------+-------+-------+-------+-------+-------+-------+ 144 * 8 | padding | 145 * +-------+-------+-------+-------+-------+-------+-------+-------+ 146 * 9 | physical birth txg | 147 * +-------+-------+-------+-------+-------+-------+-------+-------+ 148 * a | logical birth txg | 149 * +-------+-------+-------+-------+-------+-------+-------+-------+ 150 * b | fill count | 151 * +-------+-------+-------+-------+-------+-------+-------+-------+ 152 * c | checksum[0] | 153 * +-------+-------+-------+-------+-------+-------+-------+-------+ 154 * d | checksum[1] | 155 * +-------+-------+-------+-------+-------+-------+-------+-------+ 156 * e | checksum[2] | 157 * +-------+-------+-------+-------+-------+-------+-------+-------+ 158 * f | checksum[3] | 159 * +-------+-------+-------+-------+-------+-------+-------+-------+ 160 * 161 * Legend: 162 * 163 * vdev virtual device ID 164 * offset offset into virtual device 165 * LSIZE logical size 166 * PSIZE physical size (after compression) 167 * ASIZE allocated size (including RAID-Z parity and gang block headers) 168 * cksum checksum function 169 * comp compression function 170 * G gang block indicator 171 * B byteorder (endianness) 172 * D dedup 173 * X encryption 174 * E blkptr_t contains embedded data (see below) 175 * lvl level of indirection 176 * type DMU object type 177 * phys birth txg when dva[0] was written; zero if same as logical birth txg 178 * note that typically all the dva's would be written in this 179 * txg, but they could be different if they were moved by 180 * device removal. 181 * log. birth transaction group in which the block was logically born 182 * fill count number of non-zero blocks under this bp 183 * checksum[4] 256-bit checksum of the data this bp describes 184 */ 185 186 /* 187 * The blkptr_t's of encrypted blocks also need to store the encryption 188 * parameters so that the block can be decrypted. This layout is as follows: 189 * 190 * 64 56 48 40 32 24 16 8 0 191 * +-------+-------+-------+-------+-------+-------+-------+-------+ 192 * 0 | vdev1 | pad | ASIZE | 193 * +-------+-------+-------+-------+-------+-------+-------+-------+ 194 * 1 |G| offset1 | 195 * +-------+-------+-------+-------+-------+-------+-------+-------+ 196 * 2 | vdev2 | pad | ASIZE | 197 * +-------+-------+-------+-------+-------+-------+-------+-------+ 198 * 3 |G| offset2 | 199 * +-------+-------+-------+-------+-------+-------+-------+-------+ 200 * 4 | salt | 201 * +-------+-------+-------+-------+-------+-------+-------+-------+ 202 * 5 | IV1 | 203 * +-------+-------+-------+-------+-------+-------+-------+-------+ 204 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 205 * +-------+-------+-------+-------+-------+-------+-------+-------+ 206 * 7 | padding | 207 * +-------+-------+-------+-------+-------+-------+-------+-------+ 208 * 8 | padding | 209 * +-------+-------+-------+-------+-------+-------+-------+-------+ 210 * 9 | physical birth txg | 211 * +-------+-------+-------+-------+-------+-------+-------+-------+ 212 * a | logical birth txg | 213 * +-------+-------+-------+-------+-------+-------+-------+-------+ 214 * b | IV2 | fill count | 215 * +-------+-------+-------+-------+-------+-------+-------+-------+ 216 * c | checksum[0] | 217 * +-------+-------+-------+-------+-------+-------+-------+-------+ 218 * d | checksum[1] | 219 * +-------+-------+-------+-------+-------+-------+-------+-------+ 220 * e | MAC[0] | 221 * +-------+-------+-------+-------+-------+-------+-------+-------+ 222 * f | MAC[1] | 223 * +-------+-------+-------+-------+-------+-------+-------+-------+ 224 * 225 * Legend: 226 * 227 * salt Salt for generating encryption keys 228 * IV1 First 64 bits of encryption IV 229 * X Block requires encryption handling (set to 1) 230 * E blkptr_t contains embedded data (set to 0, see below) 231 * fill count number of non-zero blocks under this bp (truncated to 32 bits) 232 * IV2 Last 32 bits of encryption IV 233 * checksum[2] 128-bit checksum of the data this bp describes 234 * MAC[2] 128-bit message authentication code for this data 235 * 236 * The X bit being set indicates that this block is one of 3 types. If this is 237 * a level 0 block with an encrypted object type, the block is encrypted 238 * (see BP_IS_ENCRYPTED()). If this is a level 0 block with an unencrypted 239 * object type, this block is authenticated with an HMAC (see 240 * BP_IS_AUTHENTICATED()). Otherwise (if level > 0), this bp will use the MAC 241 * words to store a checksum-of-MACs from the level below (see 242 * BP_HAS_INDIRECT_MAC_CKSUM()). For convenience in the code, BP_IS_PROTECTED() 243 * refers to both encrypted and authenticated blocks and BP_USES_CRYPT() 244 * refers to any of these 3 kinds of blocks. 245 * 246 * The additional encryption parameters are the salt, IV, and MAC which are 247 * explained in greater detail in the block comment at the top of zio_crypt.c. 248 * The MAC occupies half of the checksum space since it serves a very similar 249 * purpose: to prevent data corruption on disk. The only functional difference 250 * is that the checksum is used to detect on-disk corruption whether or not the 251 * encryption key is loaded and the MAC provides additional protection against 252 * malicious disk tampering. We use the 3rd DVA to store the salt and first 253 * 64 bits of the IV. As a result encrypted blocks can only have 2 copies 254 * maximum instead of the normal 3. The last 32 bits of the IV are stored in 255 * the upper bits of what is usually the fill count. Note that only blocks at 256 * level 0 or -2 are ever encrypted, which allows us to guarantee that these 257 * 32 bits are not trampled over by other code (see zio_crypt.c for details). 258 * The salt and IV are not used for authenticated bps or bps with an indirect 259 * MAC checksum, so these blocks can utilize all 3 DVAs and the full 64 bits 260 * for the fill count. 261 */ 262 263 /* 264 * "Embedded" blkptr_t's don't actually point to a block, instead they 265 * have a data payload embedded in the blkptr_t itself. See the comment 266 * in blkptr.c for more details. 267 * 268 * The blkptr_t is laid out as follows: 269 * 270 * 64 56 48 40 32 24 16 8 0 271 * +-------+-------+-------+-------+-------+-------+-------+-------+ 272 * 0 | payload | 273 * 1 | payload | 274 * 2 | payload | 275 * 3 | payload | 276 * 4 | payload | 277 * 5 | payload | 278 * +-------+-------+-------+-------+-------+-------+-------+-------+ 279 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | 280 * +-------+-------+-------+-------+-------+-------+-------+-------+ 281 * 7 | payload | 282 * 8 | payload | 283 * 9 | payload | 284 * +-------+-------+-------+-------+-------+-------+-------+-------+ 285 * a | logical birth txg | 286 * +-------+-------+-------+-------+-------+-------+-------+-------+ 287 * b | payload | 288 * c | payload | 289 * d | payload | 290 * e | payload | 291 * f | payload | 292 * +-------+-------+-------+-------+-------+-------+-------+-------+ 293 * 294 * Legend: 295 * 296 * payload contains the embedded data 297 * B (byteorder) byteorder (endianness) 298 * D (dedup) padding (set to zero) 299 * X encryption (set to zero) 300 * E (embedded) set to one 301 * lvl indirection level 302 * type DMU object type 303 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) 304 * comp compression function of payload 305 * PSIZE size of payload after compression, in bytes 306 * LSIZE logical size of payload, in bytes 307 * note that 25 bits is enough to store the largest 308 * "normal" BP's LSIZE (2^16 * 2^9) in bytes 309 * log. birth transaction group in which the block was logically born 310 * 311 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded 312 * bp's they are stored in units of SPA_MINBLOCKSHIFT. 313 * Generally, the generic BP_GET_*() macros can be used on embedded BP's. 314 * The B, D, X, lvl, type, and comp fields are stored the same as with normal 315 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must 316 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before 317 * other macros, as they assert that they are only used on BP's of the correct 318 * "embedded-ness". Encrypted blkptr_t's cannot be embedded because they use 319 * the payload space for encryption parameters (see the comment above on 320 * how encryption parameters are stored). 321 */ 322 323 #define BPE_GET_ETYPE(bp) \ 324 (ASSERT(BP_IS_EMBEDDED(bp)), \ 325 BF64_GET((bp)->blk_prop, 40, 8)) 326 #define BPE_SET_ETYPE(bp, t) do { \ 327 ASSERT(BP_IS_EMBEDDED(bp)); \ 328 BF64_SET((bp)->blk_prop, 40, 8, t); \ 329 } while (0) 330 331 #define BPE_GET_LSIZE(bp) \ 332 (ASSERT(BP_IS_EMBEDDED(bp)), \ 333 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) 334 #define BPE_SET_LSIZE(bp, x) do { \ 335 ASSERT(BP_IS_EMBEDDED(bp)); \ 336 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ 337 } while (0) 338 339 #define BPE_GET_PSIZE(bp) \ 340 (ASSERT(BP_IS_EMBEDDED(bp)), \ 341 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) 342 #define BPE_SET_PSIZE(bp, x) do { \ 343 ASSERT(BP_IS_EMBEDDED(bp)); \ 344 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ 345 } while (0) 346 347 typedef enum bp_embedded_type { 348 BP_EMBEDDED_TYPE_DATA, 349 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for Delphix byteswap feature. */ 350 BP_EMBEDDED_TYPE_REDACTED, 351 NUM_BP_EMBEDDED_TYPES 352 } bp_embedded_type_t; 353 354 #define BPE_NUM_WORDS 14 355 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 356 #define BPE_IS_PAYLOADWORD(bp, wp) \ 357 ((wp) != &(bp)->blk_prop && (wp) != (&(bp)->blk_birth_word[1])) 358 359 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 360 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 361 #define SPA_SYNC_MIN_VDEVS 3 /* min vdevs to update during sync */ 362 363 /* 364 * A block is a hole when it has either 1) never been written to, or 365 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads 366 * without physically allocating disk space. Holes are represented in the 367 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is 368 * done through the BP_IS_HOLE macro. For holes, the logical size, level, 369 * DMU object type, and birth times are all also stored for holes that 370 * were written to at some point (i.e. were punched after having been filled). 371 */ 372 typedef struct blkptr { 373 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 374 uint64_t blk_prop; /* size, compression, type, etc */ 375 uint64_t blk_pad[2]; /* Extra space for the future */ 376 uint64_t blk_birth_word[2]; 377 uint64_t blk_fill; /* fill count */ 378 zio_cksum_t blk_cksum; /* 256-bit checksum */ 379 } blkptr_t; 380 381 /* 382 * Macros to get and set fields in a bp or DVA. 383 */ 384 385 /* 386 * Note, for gang blocks, DVA_GET_ASIZE() is the total space allocated for 387 * this gang DVA including its children BP's. The space allocated at this 388 * DVA's vdev/offset is vdev_gang_header_asize(vdev). 389 */ 390 #define DVA_GET_ASIZE(dva) \ 391 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 392 #define DVA_SET_ASIZE(dva, x) \ 393 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 394 SPA_MINBLOCKSHIFT, 0, x) 395 396 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, SPA_VDEVBITS) 397 #define DVA_SET_VDEV(dva, x) \ 398 BF64_SET((dva)->dva_word[0], 32, SPA_VDEVBITS, x) 399 400 #define DVA_GET_OFFSET(dva) \ 401 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 402 #define DVA_SET_OFFSET(dva, x) \ 403 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 404 405 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 406 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 407 408 #define BP_GET_LSIZE(bp) \ 409 (BP_IS_EMBEDDED(bp) ? \ 410 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ 411 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 412 #define BP_SET_LSIZE(bp, x) do { \ 413 ASSERT(!BP_IS_EMBEDDED(bp)); \ 414 BF64_SET_SB((bp)->blk_prop, \ 415 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 416 } while (0) 417 418 #define BP_GET_PSIZE(bp) \ 419 (BP_IS_EMBEDDED(bp) ? 0 : \ 420 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 421 #define BP_SET_PSIZE(bp, x) do { \ 422 ASSERT(!BP_IS_EMBEDDED(bp)); \ 423 BF64_SET_SB((bp)->blk_prop, \ 424 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 425 } while (0) 426 427 #define BP_GET_COMPRESS(bp) \ 428 BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS) 429 #define BP_SET_COMPRESS(bp, x) \ 430 BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x) 431 432 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) 433 #define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x) 434 435 #define BP_GET_CHECKSUM(bp) \ 436 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \ 437 BF64_GET((bp)->blk_prop, 40, 8)) 438 #define BP_SET_CHECKSUM(bp, x) do { \ 439 ASSERT(!BP_IS_EMBEDDED(bp)); \ 440 BF64_SET((bp)->blk_prop, 40, 8, x); \ 441 } while (0) 442 443 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 444 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 445 446 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 447 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 448 449 /* encrypted, authenticated, and MAC cksum bps use the same bit */ 450 #define BP_USES_CRYPT(bp) BF64_GET((bp)->blk_prop, 61, 1) 451 #define BP_SET_CRYPT(bp, x) BF64_SET((bp)->blk_prop, 61, 1, x) 452 453 #define BP_IS_ENCRYPTED(bp) \ 454 (BP_USES_CRYPT(bp) && \ 455 BP_GET_LEVEL(bp) <= 0 && \ 456 DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp))) 457 458 #define BP_IS_AUTHENTICATED(bp) \ 459 (BP_USES_CRYPT(bp) && \ 460 BP_GET_LEVEL(bp) <= 0 && \ 461 !DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp))) 462 463 #define BP_HAS_INDIRECT_MAC_CKSUM(bp) \ 464 (BP_USES_CRYPT(bp) && BP_GET_LEVEL(bp) > 0) 465 466 #define BP_IS_PROTECTED(bp) \ 467 (BP_IS_ENCRYPTED(bp) || BP_IS_AUTHENTICATED(bp)) 468 469 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 470 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 471 472 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 473 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 474 475 #define BP_GET_FREE(bp) BF64_GET((bp)->blk_fill, 0, 1) 476 #define BP_SET_FREE(bp, x) BF64_SET((bp)->blk_fill, 0, 1, x) 477 478 #define BP_GET_LOGICAL_BIRTH(bp) (bp)->blk_birth_word[1] 479 #define BP_SET_LOGICAL_BIRTH(bp, x) ((bp)->blk_birth_word[1] = (x)) 480 481 #define BP_GET_PHYSICAL_BIRTH(bp) (bp)->blk_birth_word[0] 482 #define BP_SET_PHYSICAL_BIRTH(bp, x) ((bp)->blk_birth_word[0] = (x)) 483 484 #define BP_GET_BIRTH(bp) \ 485 (BP_IS_EMBEDDED(bp) ? 0 : \ 486 BP_GET_PHYSICAL_BIRTH(bp) ? BP_GET_PHYSICAL_BIRTH(bp) : \ 487 BP_GET_LOGICAL_BIRTH(bp)) 488 489 #define BP_SET_BIRTH(bp, logical, physical) \ 490 { \ 491 ASSERT(!BP_IS_EMBEDDED(bp)); \ 492 BP_SET_LOGICAL_BIRTH(bp, logical); \ 493 BP_SET_PHYSICAL_BIRTH(bp, \ 494 ((logical) == (physical) ? 0 : (physical))); \ 495 } 496 497 #define BP_GET_FILL(bp) \ 498 ((BP_IS_ENCRYPTED(bp)) ? BF64_GET((bp)->blk_fill, 0, 32) : \ 499 ((BP_IS_EMBEDDED(bp)) ? 1 : (bp)->blk_fill)) 500 501 #define BP_SET_FILL(bp, fill) \ 502 { \ 503 if (BP_IS_ENCRYPTED(bp)) \ 504 BF64_SET((bp)->blk_fill, 0, 32, fill); \ 505 else \ 506 (bp)->blk_fill = fill; \ 507 } 508 509 #define BP_GET_IV2(bp) \ 510 (ASSERT(BP_IS_ENCRYPTED(bp)), \ 511 BF64_GET((bp)->blk_fill, 32, 32)) 512 #define BP_SET_IV2(bp, iv2) \ 513 { \ 514 ASSERT(BP_IS_ENCRYPTED(bp)); \ 515 BF64_SET((bp)->blk_fill, 32, 32, iv2); \ 516 } 517 518 #define BP_IS_METADATA(bp) \ 519 (BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) 520 521 #define BP_GET_ASIZE(bp) \ 522 (BP_IS_EMBEDDED(bp) ? 0 : \ 523 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 524 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 525 (DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp))) 526 527 #define BP_GET_UCSIZE(bp) \ 528 (BP_IS_METADATA(bp) ? BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)) 529 530 #define BP_GET_NDVAS(bp) \ 531 (BP_IS_EMBEDDED(bp) ? 0 : \ 532 !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 533 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 534 (!!DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp))) 535 536 #define BP_COUNT_GANG(bp) \ 537 (BP_IS_EMBEDDED(bp) ? 0 : \ 538 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \ 539 DVA_GET_GANG(&(bp)->blk_dva[1]) + \ 540 (DVA_GET_GANG(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))) 541 542 #define DVA_EQUAL(dva1, dva2) \ 543 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 544 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 545 546 #define BP_EQUAL(bp1, bp2) \ 547 (BP_GET_BIRTH(bp1) == BP_GET_BIRTH(bp2) && \ 548 BP_GET_LOGICAL_BIRTH(bp1) == BP_GET_LOGICAL_BIRTH(bp2) && \ 549 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \ 550 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \ 551 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2])) 552 553 554 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 555 556 #define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0]) 557 #define BP_IS_GANG(bp) \ 558 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp))) 559 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ 560 (dva)->dva_word[1] == 0ULL) 561 #define BP_IS_HOLE(bp) \ 562 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp))) 563 564 #define BP_SET_REDACTED(bp) \ 565 { \ 566 BP_SET_EMBEDDED(bp, B_TRUE); \ 567 BPE_SET_ETYPE(bp, BP_EMBEDDED_TYPE_REDACTED); \ 568 } 569 #define BP_IS_REDACTED(bp) \ 570 (BP_IS_EMBEDDED(bp) && BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_REDACTED) 571 572 /* BP_IS_RAIDZ(bp) assumes no block compression */ 573 #define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \ 574 BP_GET_PSIZE(bp)) 575 576 #define BP_ZERO_DVAS(bp) \ 577 { \ 578 (bp)->blk_dva[0].dva_word[0] = 0; \ 579 (bp)->blk_dva[0].dva_word[1] = 0; \ 580 (bp)->blk_dva[1].dva_word[0] = 0; \ 581 (bp)->blk_dva[1].dva_word[1] = 0; \ 582 (bp)->blk_dva[2].dva_word[0] = 0; \ 583 (bp)->blk_dva[2].dva_word[1] = 0; \ 584 } 585 586 #define BP_ZERO(bp) \ 587 { \ 588 BP_ZERO_DVAS(bp); \ 589 (bp)->blk_prop = 0; \ 590 (bp)->blk_pad[0] = 0; \ 591 (bp)->blk_pad[1] = 0; \ 592 (bp)->blk_birth_word[0] = 0; \ 593 (bp)->blk_birth_word[1] = 0; \ 594 (bp)->blk_fill = 0; \ 595 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 596 } 597 598 #ifdef _ZFS_BIG_ENDIAN 599 #define ZFS_HOST_BYTEORDER (0ULL) 600 #else 601 #define ZFS_HOST_BYTEORDER (1ULL) 602 #endif 603 604 #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) 605 606 #define BP_SPRINTF_LEN 400 607 608 /* 609 * This macro allows code sharing between zfs, libzpool, and mdb. 610 * 'func' is either kmem_scnprintf() or mdb_snprintf(). 611 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line. 612 */ 613 614 #define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \ 615 { \ 616 static const char *const copyname[] = \ 617 { "zero", "single", "double", "triple" }; \ 618 int len = 0; \ 619 int copies = 0; \ 620 const char *crypt_type; \ 621 if (bp != NULL) { \ 622 if (BP_IS_ENCRYPTED(bp)) { \ 623 crypt_type = "encrypted"; \ 624 /* LINTED E_SUSPICIOUS_COMPARISON */ \ 625 } else if (BP_IS_AUTHENTICATED(bp)) { \ 626 crypt_type = "authenticated"; \ 627 } else if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) { \ 628 crypt_type = "indirect-MAC"; \ 629 } else { \ 630 crypt_type = "unencrypted"; \ 631 } \ 632 } \ 633 if (bp == NULL) { \ 634 len += func(buf + len, size - len, "<NULL>"); \ 635 } else if (BP_IS_HOLE(bp)) { \ 636 len += func(buf + len, size - len, \ 637 "HOLE [L%llu %s] " \ 638 "size=%llxL birth=%lluL", \ 639 (u_longlong_t)BP_GET_LEVEL(bp), \ 640 type, \ 641 (u_longlong_t)BP_GET_LSIZE(bp), \ 642 (u_longlong_t)BP_GET_LOGICAL_BIRTH(bp)); \ 643 } else if (BP_IS_EMBEDDED(bp)) { \ 644 len = func(buf + len, size - len, \ 645 "EMBEDDED [L%llu %s] et=%u %s " \ 646 "size=%llxL/%llxP birth=%lluL", \ 647 (u_longlong_t)BP_GET_LEVEL(bp), \ 648 type, \ 649 (int)BPE_GET_ETYPE(bp), \ 650 compress, \ 651 (u_longlong_t)BPE_GET_LSIZE(bp), \ 652 (u_longlong_t)BPE_GET_PSIZE(bp), \ 653 (u_longlong_t)BP_GET_LOGICAL_BIRTH(bp)); \ 654 } else if (BP_IS_REDACTED(bp)) { \ 655 len += func(buf + len, size - len, \ 656 "REDACTED [L%llu %s] size=%llxL birth=%lluL", \ 657 (u_longlong_t)BP_GET_LEVEL(bp), \ 658 type, \ 659 (u_longlong_t)BP_GET_LSIZE(bp), \ 660 (u_longlong_t)BP_GET_LOGICAL_BIRTH(bp)); \ 661 } else { \ 662 for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \ 663 const dva_t *dva = &bp->blk_dva[d]; \ 664 if (DVA_IS_VALID(dva)) \ 665 copies++; \ 666 len += func(buf + len, size - len, \ 667 "DVA[%d]=<%llu:%llx:%llx>%c", d, \ 668 (u_longlong_t)DVA_GET_VDEV(dva), \ 669 (u_longlong_t)DVA_GET_OFFSET(dva), \ 670 (u_longlong_t)DVA_GET_ASIZE(dva), \ 671 ws); \ 672 } \ 673 ASSERT3S(copies, >, 0); \ 674 if (BP_IS_ENCRYPTED(bp)) { \ 675 len += func(buf + len, size - len, \ 676 "salt=%llx iv=%llx:%llx%c", \ 677 (u_longlong_t)bp->blk_dva[2].dva_word[0], \ 678 (u_longlong_t)bp->blk_dva[2].dva_word[1], \ 679 (u_longlong_t)BP_GET_IV2(bp), \ 680 ws); \ 681 } \ 682 if (BP_IS_GANG(bp) && \ 683 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \ 684 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \ 685 copies--; \ 686 len += func(buf + len, size - len, \ 687 "[L%llu %s] %s %s %s %s %s %s %s%c" \ 688 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \ 689 "cksum=%016llx:%016llx:%016llx:%016llx", \ 690 (u_longlong_t)BP_GET_LEVEL(bp), \ 691 type, \ 692 checksum, \ 693 compress, \ 694 crypt_type, \ 695 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \ 696 BP_IS_GANG(bp) ? "gang" : "contiguous", \ 697 BP_GET_DEDUP(bp) ? "dedup" : "unique", \ 698 copyname[copies], \ 699 ws, \ 700 (u_longlong_t)BP_GET_LSIZE(bp), \ 701 (u_longlong_t)BP_GET_PSIZE(bp), \ 702 (u_longlong_t)BP_GET_LOGICAL_BIRTH(bp), \ 703 (u_longlong_t)BP_GET_BIRTH(bp), \ 704 (u_longlong_t)BP_GET_FILL(bp), \ 705 ws, \ 706 (u_longlong_t)bp->blk_cksum.zc_word[0], \ 707 (u_longlong_t)bp->blk_cksum.zc_word[1], \ 708 (u_longlong_t)bp->blk_cksum.zc_word[2], \ 709 (u_longlong_t)bp->blk_cksum.zc_word[3]); \ 710 } \ 711 ASSERT(len < size); \ 712 } 713 714 #define BP_GET_BUFC_TYPE(bp) \ 715 (BP_IS_METADATA(bp) ? ARC_BUFC_METADATA : ARC_BUFC_DATA) 716 717 typedef enum spa_import_type { 718 SPA_IMPORT_EXISTING, 719 SPA_IMPORT_ASSEMBLE 720 } spa_import_type_t; 721 722 typedef enum spa_mode { 723 SPA_MODE_UNINIT = 0, 724 SPA_MODE_READ = 1, 725 SPA_MODE_WRITE = 2, 726 } spa_mode_t; 727 728 /* 729 * Send TRIM commands in-line during normal pool operation while deleting. 730 * OFF: no 731 * ON: yes 732 */ 733 typedef enum { 734 SPA_AUTOTRIM_OFF = 0, /* default */ 735 SPA_AUTOTRIM_ON, 736 } spa_autotrim_t; 737 738 /* 739 * Reason TRIM command was issued, used internally for accounting purposes. 740 */ 741 typedef enum trim_type { 742 TRIM_TYPE_MANUAL = 0, 743 TRIM_TYPE_AUTO = 1, 744 TRIM_TYPE_SIMPLE = 2 745 } trim_type_t; 746 747 /* state manipulation functions */ 748 extern int spa_open(const char *pool, spa_t **, const void *tag); 749 extern int spa_open_rewind(const char *pool, spa_t **, const void *tag, 750 nvlist_t *policy, nvlist_t **config); 751 extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot, 752 size_t buflen); 753 extern int spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props, 754 nvlist_t *zplprops, struct dsl_crypto_params *dcp); 755 extern int spa_import(char *pool, nvlist_t *config, nvlist_t *props, 756 uint64_t flags); 757 extern nvlist_t *spa_tryimport(nvlist_t *tryconfig); 758 extern int spa_destroy(const char *pool); 759 extern int spa_checkpoint(const char *pool); 760 extern int spa_checkpoint_discard(const char *pool); 761 extern int spa_export(const char *pool, nvlist_t **oldconfig, boolean_t force, 762 boolean_t hardforce); 763 extern int spa_reset(const char *pool); 764 extern void spa_async_request(spa_t *spa, int flag); 765 extern void spa_async_unrequest(spa_t *spa, int flag); 766 extern void spa_async_suspend(spa_t *spa); 767 extern void spa_async_resume(spa_t *spa); 768 extern int spa_async_tasks(spa_t *spa); 769 extern spa_t *spa_inject_addref(char *pool); 770 extern void spa_inject_delref(spa_t *spa); 771 extern void spa_scan_stat_init(spa_t *spa); 772 extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps); 773 extern int bpobj_enqueue_alloc_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx); 774 extern int bpobj_enqueue_free_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx); 775 776 #define SPA_ASYNC_CONFIG_UPDATE 0x01 777 #define SPA_ASYNC_REMOVE 0x02 778 #define SPA_ASYNC_FAULT_VDEV 0x04 779 #define SPA_ASYNC_RESILVER_DONE 0x08 780 #define SPA_ASYNC_RESILVER 0x10 781 #define SPA_ASYNC_AUTOEXPAND 0x20 782 #define SPA_ASYNC_REMOVE_DONE 0x40 783 #define SPA_ASYNC_REMOVE_STOP 0x80 784 #define SPA_ASYNC_INITIALIZE_RESTART 0x100 785 #define SPA_ASYNC_TRIM_RESTART 0x200 786 #define SPA_ASYNC_AUTOTRIM_RESTART 0x400 787 #define SPA_ASYNC_L2CACHE_REBUILD 0x800 788 #define SPA_ASYNC_L2CACHE_TRIM 0x1000 789 #define SPA_ASYNC_REBUILD_DONE 0x2000 790 #define SPA_ASYNC_DETACH_SPARE 0x4000 791 792 /* device manipulation */ 793 extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot, boolean_t ashift_check); 794 extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, 795 int replacing, int rebuild); 796 extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, 797 int replace_done); 798 extern int spa_vdev_alloc(spa_t *spa, uint64_t guid); 799 extern int spa_vdev_noalloc(spa_t *spa, uint64_t guid); 800 extern boolean_t spa_vdev_remove_active(spa_t *spa); 801 extern int spa_vdev_initialize(spa_t *spa, nvlist_t *nv, uint64_t cmd_type, 802 nvlist_t *vdev_errlist); 803 extern int spa_vdev_trim(spa_t *spa, nvlist_t *nv, uint64_t cmd_type, 804 uint64_t rate, boolean_t partial, boolean_t secure, nvlist_t *vdev_errlist); 805 extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath); 806 extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru); 807 extern int spa_vdev_split_mirror(spa_t *spa, const char *newname, 808 nvlist_t *config, nvlist_t *props, boolean_t exp); 809 810 /* spare state (which is global across all pools) */ 811 extern void spa_spare_add(vdev_t *vd); 812 extern void spa_spare_remove(vdev_t *vd); 813 extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt); 814 extern void spa_spare_activate(vdev_t *vd); 815 816 /* L2ARC state (which is global across all pools) */ 817 extern void spa_l2cache_add(vdev_t *vd); 818 extern void spa_l2cache_remove(vdev_t *vd); 819 extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool); 820 extern void spa_l2cache_activate(vdev_t *vd); 821 extern void spa_l2cache_drop(spa_t *spa); 822 823 /* scanning */ 824 extern int spa_scan(spa_t *spa, pool_scan_func_t func); 825 extern int spa_scan_range(spa_t *spa, pool_scan_func_t func, uint64_t txgstart, 826 uint64_t txgend); 827 extern int spa_scan_stop(spa_t *spa); 828 extern int spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t flag); 829 830 /* spa syncing */ 831 extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */ 832 extern void spa_sync_allpools(void); 833 834 extern uint_t zfs_sync_pass_deferred_free; 835 836 /* spa sync taskqueues */ 837 taskq_t *spa_sync_tq_create(spa_t *spa, const char *name); 838 void spa_sync_tq_destroy(spa_t *spa); 839 uint_t spa_acq_allocator(spa_t *spa); 840 void spa_rel_allocator(spa_t *spa, uint_t allocator); 841 void spa_select_allocator(zio_t *zio); 842 843 /* spa namespace global mutex */ 844 extern kmutex_t spa_namespace_lock; 845 extern avl_tree_t spa_namespace_avl; 846 extern kcondvar_t spa_namespace_cv; 847 848 /* 849 * SPA configuration functions in spa_config.c 850 */ 851 852 #define SPA_CONFIG_UPDATE_POOL 0 853 #define SPA_CONFIG_UPDATE_VDEVS 1 854 855 extern void spa_write_cachefile(spa_t *, boolean_t, boolean_t, boolean_t); 856 extern void spa_config_load(void); 857 extern int spa_all_configs(uint64_t *generation, nvlist_t **pools); 858 extern void spa_config_set(spa_t *spa, nvlist_t *config); 859 extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, 860 int getstats); 861 extern void spa_config_update(spa_t *spa, int what); 862 extern int spa_config_parse(spa_t *spa, vdev_t **vdp, nvlist_t *nv, 863 vdev_t *parent, uint_t id, int atype); 864 865 866 /* 867 * Miscellaneous SPA routines in spa_misc.c 868 */ 869 870 /* Namespace manipulation */ 871 extern spa_t *spa_lookup(const char *name); 872 extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot); 873 extern void spa_remove(spa_t *spa); 874 extern spa_t *spa_next(spa_t *prev); 875 876 /* Refcount functions */ 877 extern void spa_open_ref(spa_t *spa, const void *tag); 878 extern void spa_close(spa_t *spa, const void *tag); 879 extern void spa_async_close(spa_t *spa, const void *tag); 880 extern boolean_t spa_refcount_zero(spa_t *spa); 881 882 #define SCL_NONE 0x00 883 #define SCL_CONFIG 0x01 884 #define SCL_STATE 0x02 885 #define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */ 886 #define SCL_ALLOC 0x08 887 #define SCL_ZIO 0x10 888 #define SCL_FREE 0x20 889 #define SCL_VDEV 0x40 890 #define SCL_LOCKS 7 891 #define SCL_ALL ((1 << SCL_LOCKS) - 1) 892 #define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO) 893 894 /* Historical pool statistics */ 895 typedef struct spa_history_kstat { 896 kmutex_t lock; 897 uint64_t count; 898 uint64_t size; 899 kstat_t *kstat; 900 void *priv; 901 list_t list; 902 } spa_history_kstat_t; 903 904 typedef struct spa_history_list { 905 uint64_t size; 906 procfs_list_t procfs_list; 907 } spa_history_list_t; 908 909 typedef struct spa_stats { 910 spa_history_list_t read_history; 911 spa_history_list_t txg_history; 912 spa_history_kstat_t tx_assign_histogram; 913 spa_history_list_t mmp_history; 914 spa_history_kstat_t state; /* pool state */ 915 spa_history_kstat_t guid; /* pool guid */ 916 spa_history_kstat_t iostats; 917 } spa_stats_t; 918 919 typedef enum txg_state { 920 TXG_STATE_BIRTH = 0, 921 TXG_STATE_OPEN = 1, 922 TXG_STATE_QUIESCED = 2, 923 TXG_STATE_WAIT_FOR_SYNC = 3, 924 TXG_STATE_SYNCED = 4, 925 TXG_STATE_COMMITTED = 5, 926 } txg_state_t; 927 928 typedef struct txg_stat { 929 vdev_stat_t vs1; 930 vdev_stat_t vs2; 931 uint64_t txg; 932 uint64_t ndirty; 933 } txg_stat_t; 934 935 /* Assorted pool IO kstats */ 936 typedef struct spa_iostats { 937 kstat_named_t trim_extents_written; 938 kstat_named_t trim_bytes_written; 939 kstat_named_t trim_extents_skipped; 940 kstat_named_t trim_bytes_skipped; 941 kstat_named_t trim_extents_failed; 942 kstat_named_t trim_bytes_failed; 943 kstat_named_t autotrim_extents_written; 944 kstat_named_t autotrim_bytes_written; 945 kstat_named_t autotrim_extents_skipped; 946 kstat_named_t autotrim_bytes_skipped; 947 kstat_named_t autotrim_extents_failed; 948 kstat_named_t autotrim_bytes_failed; 949 kstat_named_t simple_trim_extents_written; 950 kstat_named_t simple_trim_bytes_written; 951 kstat_named_t simple_trim_extents_skipped; 952 kstat_named_t simple_trim_bytes_skipped; 953 kstat_named_t simple_trim_extents_failed; 954 kstat_named_t simple_trim_bytes_failed; 955 kstat_named_t arc_read_count; 956 kstat_named_t arc_read_bytes; 957 kstat_named_t arc_write_count; 958 kstat_named_t arc_write_bytes; 959 kstat_named_t direct_read_count; 960 kstat_named_t direct_read_bytes; 961 kstat_named_t direct_write_count; 962 kstat_named_t direct_write_bytes; 963 } spa_iostats_t; 964 965 extern void spa_stats_init(spa_t *spa); 966 extern void spa_stats_destroy(spa_t *spa); 967 extern void spa_read_history_add(spa_t *spa, const zbookmark_phys_t *zb, 968 uint32_t aflags); 969 extern void spa_txg_history_add(spa_t *spa, uint64_t txg, hrtime_t birth_time); 970 extern int spa_txg_history_set(spa_t *spa, uint64_t txg, 971 txg_state_t completed_state, hrtime_t completed_time); 972 extern txg_stat_t *spa_txg_history_init_io(spa_t *, uint64_t, 973 struct dsl_pool *); 974 extern void spa_txg_history_fini_io(spa_t *, txg_stat_t *); 975 extern void spa_tx_assign_add_nsecs(spa_t *spa, uint64_t nsecs); 976 extern int spa_mmp_history_set_skip(spa_t *spa, uint64_t mmp_kstat_id); 977 extern int spa_mmp_history_set(spa_t *spa, uint64_t mmp_kstat_id, int io_error, 978 hrtime_t duration); 979 extern void spa_mmp_history_add(spa_t *spa, uint64_t txg, uint64_t timestamp, 980 uint64_t mmp_delay, vdev_t *vd, int label, uint64_t mmp_kstat_id, 981 int error); 982 extern void spa_iostats_trim_add(spa_t *spa, trim_type_t type, 983 uint64_t extents_written, uint64_t bytes_written, 984 uint64_t extents_skipped, uint64_t bytes_skipped, 985 uint64_t extents_failed, uint64_t bytes_failed); 986 extern void spa_iostats_read_add(spa_t *spa, uint64_t size, uint64_t iops, 987 uint32_t flags); 988 extern void spa_iostats_write_add(spa_t *spa, uint64_t size, uint64_t iops, 989 uint32_t flags); 990 extern void spa_import_progress_add(spa_t *spa); 991 extern void spa_import_progress_remove(uint64_t spa_guid); 992 extern int spa_import_progress_set_mmp_check(uint64_t pool_guid, 993 uint64_t mmp_sec_remaining); 994 extern int spa_import_progress_set_max_txg(uint64_t pool_guid, 995 uint64_t max_txg); 996 extern int spa_import_progress_set_state(uint64_t pool_guid, 997 spa_load_state_t spa_load_state); 998 extern void spa_import_progress_set_notes(spa_t *spa, 999 const char *fmt, ...) __printflike(2, 3); 1000 extern void spa_import_progress_set_notes_nolog(spa_t *spa, 1001 const char *fmt, ...) __printflike(2, 3); 1002 1003 /* Pool configuration locks */ 1004 extern int spa_config_tryenter(spa_t *spa, int locks, const void *tag, 1005 krw_t rw); 1006 extern void spa_config_enter(spa_t *spa, int locks, const void *tag, krw_t rw); 1007 extern void spa_config_enter_mmp(spa_t *spa, int locks, const void *tag, 1008 krw_t rw); 1009 extern void spa_config_exit(spa_t *spa, int locks, const void *tag); 1010 extern int spa_config_held(spa_t *spa, int locks, krw_t rw); 1011 1012 /* Pool vdev add/remove lock */ 1013 extern uint64_t spa_vdev_enter(spa_t *spa); 1014 extern uint64_t spa_vdev_detach_enter(spa_t *spa, uint64_t guid); 1015 extern uint64_t spa_vdev_config_enter(spa_t *spa); 1016 extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, 1017 int error, const char *tag); 1018 extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error); 1019 1020 /* Pool vdev state change lock */ 1021 extern void spa_vdev_state_enter(spa_t *spa, int oplock); 1022 extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error); 1023 1024 /* Log state */ 1025 typedef enum spa_log_state { 1026 SPA_LOG_UNKNOWN = 0, /* unknown log state */ 1027 SPA_LOG_MISSING, /* missing log(s) */ 1028 SPA_LOG_CLEAR, /* clear the log(s) */ 1029 SPA_LOG_GOOD, /* log(s) are good */ 1030 } spa_log_state_t; 1031 1032 extern spa_log_state_t spa_get_log_state(spa_t *spa); 1033 extern void spa_set_log_state(spa_t *spa, spa_log_state_t state); 1034 extern int spa_reset_logs(spa_t *spa); 1035 1036 /* Log claim callback */ 1037 extern void spa_claim_notify(zio_t *zio); 1038 extern void spa_deadman(void *); 1039 1040 /* Accessor functions */ 1041 extern boolean_t spa_shutting_down(spa_t *spa); 1042 extern struct dsl_pool *spa_get_dsl(spa_t *spa); 1043 extern boolean_t spa_is_initializing(spa_t *spa); 1044 extern boolean_t spa_indirect_vdevs_loaded(spa_t *spa); 1045 extern blkptr_t *spa_get_rootblkptr(spa_t *spa); 1046 extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp); 1047 extern void spa_altroot(spa_t *, char *, size_t); 1048 extern uint32_t spa_sync_pass(spa_t *spa); 1049 extern char *spa_name(spa_t *spa); 1050 extern uint64_t spa_guid(spa_t *spa); 1051 extern uint64_t spa_load_guid(spa_t *spa); 1052 extern uint64_t spa_last_synced_txg(spa_t *spa); 1053 extern uint64_t spa_first_txg(spa_t *spa); 1054 extern uint64_t spa_syncing_txg(spa_t *spa); 1055 extern uint64_t spa_final_dirty_txg(spa_t *spa); 1056 extern uint64_t spa_version(spa_t *spa); 1057 extern pool_state_t spa_state(spa_t *spa); 1058 extern spa_load_state_t spa_load_state(spa_t *spa); 1059 extern uint64_t spa_freeze_txg(spa_t *spa); 1060 extern uint64_t spa_get_worst_case_asize(spa_t *spa, uint64_t lsize); 1061 extern uint64_t spa_get_dspace(spa_t *spa); 1062 extern uint64_t spa_get_checkpoint_space(spa_t *spa); 1063 extern uint64_t spa_get_slop_space(spa_t *spa); 1064 extern void spa_update_dspace(spa_t *spa); 1065 extern uint64_t spa_version(spa_t *spa); 1066 extern boolean_t spa_deflate(spa_t *spa); 1067 extern metaslab_class_t *spa_normal_class(spa_t *spa); 1068 extern metaslab_class_t *spa_log_class(spa_t *spa); 1069 extern metaslab_class_t *spa_embedded_log_class(spa_t *spa); 1070 extern metaslab_class_t *spa_special_class(spa_t *spa); 1071 extern metaslab_class_t *spa_dedup_class(spa_t *spa); 1072 extern metaslab_class_t *spa_preferred_class(spa_t *spa, const zio_t *zio); 1073 extern boolean_t spa_special_has_ddt(spa_t *spa); 1074 1075 extern void spa_evicting_os_register(spa_t *, objset_t *os); 1076 extern void spa_evicting_os_deregister(spa_t *, objset_t *os); 1077 extern void spa_evicting_os_wait(spa_t *spa); 1078 extern int spa_max_replication(spa_t *spa); 1079 extern int spa_prev_software_version(spa_t *spa); 1080 extern uint64_t spa_get_failmode(spa_t *spa); 1081 extern uint64_t spa_get_deadman_failmode(spa_t *spa); 1082 extern void spa_set_deadman_failmode(spa_t *spa, const char *failmode); 1083 extern boolean_t spa_suspended(spa_t *spa); 1084 extern uint64_t spa_bootfs(spa_t *spa); 1085 extern uint64_t spa_get_last_scrubbed_txg(spa_t *spa); 1086 extern uint64_t spa_delegation(spa_t *spa); 1087 extern objset_t *spa_meta_objset(spa_t *spa); 1088 extern space_map_t *spa_syncing_log_sm(spa_t *spa); 1089 extern uint64_t spa_deadman_synctime(spa_t *spa); 1090 extern uint64_t spa_deadman_ziotime(spa_t *spa); 1091 extern uint64_t spa_dirty_data(spa_t *spa); 1092 extern spa_autotrim_t spa_get_autotrim(spa_t *spa); 1093 extern int spa_get_allocator(spa_t *spa); 1094 extern void spa_set_allocator(spa_t *spa, const char *allocator); 1095 1096 /* Miscellaneous support routines */ 1097 extern void spa_load_failed(spa_t *spa, const char *fmt, ...) 1098 __attribute__((format(printf, 2, 3))); 1099 extern void spa_load_note(spa_t *spa, const char *fmt, ...) 1100 __attribute__((format(printf, 2, 3))); 1101 extern void spa_activate_mos_feature(spa_t *spa, const char *feature, 1102 dmu_tx_t *tx); 1103 extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature); 1104 extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid); 1105 extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid); 1106 extern char *spa_strdup(const char *); 1107 extern void spa_strfree(char *); 1108 extern uint64_t spa_generate_guid(spa_t *spa); 1109 extern uint64_t spa_generate_load_guid(void); 1110 extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp); 1111 extern void spa_freeze(spa_t *spa); 1112 extern int spa_change_guid(spa_t *spa, const uint64_t *guidp); 1113 extern void spa_upgrade(spa_t *spa, uint64_t version); 1114 extern void spa_evict_all(void); 1115 extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid, 1116 boolean_t l2cache); 1117 extern boolean_t spa_has_l2cache(spa_t *, uint64_t guid); 1118 extern boolean_t spa_has_spare(spa_t *, uint64_t guid); 1119 extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva); 1120 extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp); 1121 extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp); 1122 extern boolean_t spa_has_slogs(spa_t *spa); 1123 extern boolean_t spa_is_root(spa_t *spa); 1124 extern boolean_t spa_writeable(spa_t *spa); 1125 extern boolean_t spa_has_pending_synctask(spa_t *spa); 1126 extern int spa_maxblocksize(spa_t *spa); 1127 extern int spa_maxdnodesize(spa_t *spa); 1128 extern boolean_t spa_has_checkpoint(spa_t *spa); 1129 extern boolean_t spa_importing_readonly_checkpoint(spa_t *spa); 1130 extern boolean_t spa_suspend_async_destroy(spa_t *spa); 1131 extern uint64_t spa_min_claim_txg(spa_t *spa); 1132 extern boolean_t zfs_dva_valid(spa_t *spa, const dva_t *dva, 1133 const blkptr_t *bp); 1134 typedef void (*spa_remap_cb_t)(uint64_t vdev, uint64_t offset, uint64_t size, 1135 void *arg); 1136 extern boolean_t spa_remap_blkptr(spa_t *spa, blkptr_t *bp, 1137 spa_remap_cb_t callback, void *arg); 1138 extern uint64_t spa_get_last_removal_txg(spa_t *spa); 1139 extern boolean_t spa_trust_config(spa_t *spa); 1140 extern uint64_t spa_missing_tvds_allowed(spa_t *spa); 1141 extern void spa_set_missing_tvds(spa_t *spa, uint64_t missing); 1142 extern boolean_t spa_top_vdevs_spacemap_addressable(spa_t *spa); 1143 extern uint64_t spa_total_metaslabs(spa_t *spa); 1144 extern boolean_t spa_multihost(spa_t *spa); 1145 extern uint32_t spa_get_hostid(spa_t *spa); 1146 extern void spa_activate_allocation_classes(spa_t *, dmu_tx_t *); 1147 extern boolean_t spa_livelist_delete_check(spa_t *spa); 1148 1149 extern boolean_t spa_mmp_remote_host_activity(spa_t *spa); 1150 1151 extern spa_mode_t spa_mode(spa_t *spa); 1152 extern uint64_t zfs_strtonum(const char *str, char **nptr); 1153 1154 extern char *spa_his_ievent_table[]; 1155 1156 extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx); 1157 extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read, 1158 char *his_buf); 1159 extern int spa_history_log(spa_t *spa, const char *his_buf); 1160 extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl); 1161 extern void spa_history_log_version(spa_t *spa, const char *operation, 1162 dmu_tx_t *tx); 1163 extern void spa_history_log_internal(spa_t *spa, const char *operation, 1164 dmu_tx_t *tx, const char *fmt, ...) __printflike(4, 5); 1165 extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op, 1166 dmu_tx_t *tx, const char *fmt, ...) __printflike(4, 5); 1167 extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation, 1168 dmu_tx_t *tx, const char *fmt, ...) __printflike(4, 5); 1169 1170 extern const char *spa_state_to_name(spa_t *spa); 1171 1172 /* error handling */ 1173 struct zbookmark_phys; 1174 extern void spa_log_error(spa_t *spa, const zbookmark_phys_t *zb, 1175 const uint64_t birth); 1176 extern void spa_remove_error(spa_t *spa, zbookmark_phys_t *zb, 1177 uint64_t birth); 1178 extern int zfs_ereport_post(const char *clazz, spa_t *spa, vdev_t *vd, 1179 const zbookmark_phys_t *zb, zio_t *zio, uint64_t state); 1180 extern boolean_t zfs_ereport_is_valid(const char *clazz, spa_t *spa, vdev_t *vd, 1181 zio_t *zio); 1182 extern void zfs_ereport_taskq_fini(void); 1183 extern void zfs_ereport_clear(spa_t *spa, vdev_t *vd); 1184 extern nvlist_t *zfs_event_create(spa_t *spa, vdev_t *vd, const char *type, 1185 const char *name, nvlist_t *aux); 1186 extern void zfs_post_remove(spa_t *spa, vdev_t *vd); 1187 extern void zfs_post_state_change(spa_t *spa, vdev_t *vd, uint64_t laststate); 1188 extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd); 1189 extern uint64_t spa_approx_errlog_size(spa_t *spa); 1190 extern int spa_get_errlog(spa_t *spa, void *uaddr, uint64_t *count); 1191 extern uint64_t spa_get_last_errlog_size(spa_t *spa); 1192 extern void spa_errlog_rotate(spa_t *spa); 1193 extern void spa_errlog_drain(spa_t *spa); 1194 extern void spa_errlog_sync(spa_t *spa, uint64_t txg); 1195 extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub); 1196 extern void spa_delete_dataset_errlog(spa_t *spa, uint64_t ds, dmu_tx_t *tx); 1197 extern void spa_swap_errlog(spa_t *spa, uint64_t new_head_ds, 1198 uint64_t old_head_ds, dmu_tx_t *tx); 1199 extern void sync_error_list(spa_t *spa, avl_tree_t *t, uint64_t *obj, 1200 dmu_tx_t *tx); 1201 extern void spa_upgrade_errlog(spa_t *spa, dmu_tx_t *tx); 1202 extern int find_top_affected_fs(spa_t *spa, uint64_t head_ds, 1203 zbookmark_err_phys_t *zep, uint64_t *top_affected_fs); 1204 extern int find_birth_txg(struct dsl_dataset *ds, zbookmark_err_phys_t *zep, 1205 uint64_t *birth_txg); 1206 extern void zep_to_zb(uint64_t dataset, zbookmark_err_phys_t *zep, 1207 zbookmark_phys_t *zb); 1208 extern void name_to_errphys(char *buf, zbookmark_err_phys_t *zep); 1209 1210 /* vdev mirror */ 1211 extern void vdev_mirror_stat_init(void); 1212 extern void vdev_mirror_stat_fini(void); 1213 1214 /* Initialization and termination */ 1215 extern void spa_init(spa_mode_t mode); 1216 extern void spa_fini(void); 1217 extern void spa_boot_init(void); 1218 1219 /* properties */ 1220 extern int spa_prop_set(spa_t *spa, nvlist_t *nvp); 1221 extern int spa_prop_get(spa_t *spa, nvlist_t *nvp); 1222 extern int spa_prop_get_nvlist(spa_t *spa, char **props, 1223 unsigned int n_props, nvlist_t *outnvl); 1224 extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx); 1225 extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t); 1226 1227 /* asynchronous event notification */ 1228 extern void spa_event_notify(spa_t *spa, vdev_t *vdev, nvlist_t *hist_nvl, 1229 const char *name); 1230 extern void zfs_ereport_zvol_post(const char *subclass, const char *name, 1231 const char *device_name, const char *raw_name); 1232 1233 /* waiting for pool activities to complete */ 1234 extern int spa_wait(const char *pool, zpool_wait_activity_t activity, 1235 boolean_t *waited); 1236 extern int spa_wait_tag(const char *name, zpool_wait_activity_t activity, 1237 uint64_t tag, boolean_t *waited); 1238 extern void spa_notify_waiters(spa_t *spa); 1239 extern void spa_wake_waiters(spa_t *spa); 1240 1241 extern void spa_import_os(spa_t *spa); 1242 extern void spa_export_os(spa_t *spa); 1243 extern void spa_activate_os(spa_t *spa); 1244 extern void spa_deactivate_os(spa_t *spa); 1245 1246 /* module param call functions */ 1247 int param_set_deadman_ziotime(ZFS_MODULE_PARAM_ARGS); 1248 int param_set_deadman_synctime(ZFS_MODULE_PARAM_ARGS); 1249 int param_set_slop_shift(ZFS_MODULE_PARAM_ARGS); 1250 int param_set_deadman_failmode(ZFS_MODULE_PARAM_ARGS); 1251 int param_set_active_allocator(ZFS_MODULE_PARAM_ARGS); 1252 1253 #ifdef ZFS_DEBUG 1254 #define dprintf_bp(bp, fmt, ...) do { \ 1255 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 1256 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \ 1257 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \ 1258 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \ 1259 kmem_free(__blkbuf, BP_SPRINTF_LEN); \ 1260 } \ 1261 } while (0) 1262 #else 1263 #define dprintf_bp(bp, fmt, ...) 1264 #endif 1265 1266 extern spa_mode_t spa_mode_global; 1267 extern int zfs_deadman_enabled; 1268 extern uint64_t zfs_deadman_synctime_ms; 1269 extern uint64_t zfs_deadman_ziotime_ms; 1270 extern uint64_t zfs_deadman_checktime_ms; 1271 1272 extern kmem_cache_t *zio_buf_cache[]; 1273 extern kmem_cache_t *zio_data_buf_cache[]; 1274 1275 #ifdef __cplusplus 1276 } 1277 #endif 1278 1279 #endif /* _SYS_SPA_H */ 1280