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