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 http://www.opensolaris.org/os/licensing. 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, 2019 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 2019 Joyent, Inc. 29 * Copyright (c) 2017, 2019, Datto Inc. All rights reserved. 30 * Copyright (c) 2017, Intel Corporation. 31 * Copyright 2020 Joshua M. Clulow <josh@sysmgr.org> 32 */ 33 34 #ifndef _SYS_SPA_H 35 #define _SYS_SPA_H 36 37 #include <sys/avl.h> 38 #include <sys/zfs_context.h> 39 #include <sys/nvpair.h> 40 #include <sys/sysevent.h> 41 #include <sys/sysmacros.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 spa spa_t; 57 typedef struct vdev vdev_t; 58 typedef struct metaslab metaslab_t; 59 typedef struct metaslab_group metaslab_group_t; 60 typedef struct metaslab_class metaslab_class_t; 61 typedef struct zio zio_t; 62 typedef struct zilog zilog_t; 63 typedef struct spa_aux_vdev spa_aux_vdev_t; 64 typedef struct ddt ddt_t; 65 typedef struct ddt_entry ddt_entry_t; 66 struct dsl_pool; 67 struct dsl_dataset; 68 struct dsl_crypto_params; 69 70 /* 71 * We currently support block sizes from 512 bytes to 16MB. 72 * The benefits of larger blocks, and thus larger IO, need to be weighed 73 * against the cost of COWing a giant block to modify one byte, and the 74 * large latency of reading or writing a large block. 75 * 76 * Note that although blocks up to 16MB are supported, the recordsize 77 * property can not be set larger than zfs_max_recordsize (default 1MB). 78 * See the comment near zfs_max_recordsize in dsl_dataset.c for details. 79 * 80 * Note that although the LSIZE field of the blkptr_t can store sizes up 81 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to 82 * 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB. 83 */ 84 #define SPA_MINBLOCKSHIFT 9 85 #define SPA_OLD_MAXBLOCKSHIFT 17 86 #define SPA_MAXBLOCKSHIFT 24 87 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) 88 #define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT) 89 #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) 90 91 /* 92 * Alignment Shift (ashift) is an immutable, internal top-level vdev property 93 * which can only be set at vdev creation time. Physical writes are always done 94 * according to it, which makes 2^ashift the smallest possible IO on a vdev. 95 * 96 * We currently allow values ranging from 512 bytes (2^9 = 512) to 64 KiB 97 * (2^16 = 65,536). 98 */ 99 #define ASHIFT_MIN 9 100 #define ASHIFT_MAX 16 101 102 /* 103 * Size of block to hold the configuration data (a packed nvlist) 104 */ 105 #define SPA_CONFIG_BLOCKSIZE (1ULL << 14) 106 107 /* 108 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. 109 * The ASIZE encoding should be at least 64 times larger (6 more bits) 110 * to support up to 4-way RAID-Z mirror mode with worst-case gang block 111 * overhead, three DVAs per bp, plus one more bit in case we do anything 112 * else that expands the ASIZE. 113 */ 114 #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ 115 #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ 116 #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ 117 118 #define SPA_COMPRESSBITS 7 119 #define SPA_VDEVBITS 24 120 121 /* 122 * All SPA data is represented by 128-bit data virtual addresses (DVAs). 123 * The members of the dva_t should be considered opaque outside the SPA. 124 */ 125 typedef struct dva { 126 uint64_t dva_word[2]; 127 } dva_t; 128 129 130 /* 131 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept 132 * secret and is suitable for use in MAC algorithms as the key. 133 */ 134 typedef struct zio_cksum_salt { 135 uint8_t zcs_bytes[32]; 136 } zio_cksum_salt_t; 137 138 /* 139 * Each block is described by its DVAs, time of birth, checksum, etc. 140 * The word-by-word, bit-by-bit layout of the blkptr is as follows: 141 * 142 * 64 56 48 40 32 24 16 8 0 143 * +-------+-------+-------+-------+-------+-------+-------+-------+ 144 * 0 | pad | vdev1 | GRID | ASIZE | 145 * +-------+-------+-------+-------+-------+-------+-------+-------+ 146 * 1 |G| offset1 | 147 * +-------+-------+-------+-------+-------+-------+-------+-------+ 148 * 2 | pad | vdev2 | GRID | ASIZE | 149 * +-------+-------+-------+-------+-------+-------+-------+-------+ 150 * 3 |G| offset2 | 151 * +-------+-------+-------+-------+-------+-------+-------+-------+ 152 * 4 | pad | vdev3 | GRID | ASIZE | 153 * +-------+-------+-------+-------+-------+-------+-------+-------+ 154 * 5 |G| offset3 | 155 * +-------+-------+-------+-------+-------+-------+-------+-------+ 156 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 157 * +-------+-------+-------+-------+-------+-------+-------+-------+ 158 * 7 | padding | 159 * +-------+-------+-------+-------+-------+-------+-------+-------+ 160 * 8 | padding | 161 * +-------+-------+-------+-------+-------+-------+-------+-------+ 162 * 9 | physical birth txg | 163 * +-------+-------+-------+-------+-------+-------+-------+-------+ 164 * a | logical birth txg | 165 * +-------+-------+-------+-------+-------+-------+-------+-------+ 166 * b | fill count | 167 * +-------+-------+-------+-------+-------+-------+-------+-------+ 168 * c | checksum[0] | 169 * +-------+-------+-------+-------+-------+-------+-------+-------+ 170 * d | checksum[1] | 171 * +-------+-------+-------+-------+-------+-------+-------+-------+ 172 * e | checksum[2] | 173 * +-------+-------+-------+-------+-------+-------+-------+-------+ 174 * f | checksum[3] | 175 * +-------+-------+-------+-------+-------+-------+-------+-------+ 176 * 177 * Legend: 178 * 179 * vdev virtual device ID 180 * offset offset into virtual device 181 * LSIZE logical size 182 * PSIZE physical size (after compression) 183 * ASIZE allocated size (including RAID-Z parity and gang block headers) 184 * GRID RAID-Z layout information (reserved for future use) 185 * cksum checksum function 186 * comp compression function 187 * G gang block indicator 188 * B byteorder (endianness) 189 * D dedup 190 * X encryption 191 * E blkptr_t contains embedded data (see below) 192 * lvl level of indirection 193 * type DMU object type 194 * phys birth txg when dva[0] was written; zero if same as logical birth txg 195 * note that typically all the dva's would be written in this 196 * txg, but they could be different if they were moved by 197 * device removal. 198 * log. birth transaction group in which the block was logically born 199 * fill count number of non-zero blocks under this bp 200 * checksum[4] 256-bit checksum of the data this bp describes 201 */ 202 203 /* 204 * The blkptr_t's of encrypted blocks also need to store the encryption 205 * parameters so that the block can be decrypted. This layout is as follows: 206 * 207 * 64 56 48 40 32 24 16 8 0 208 * +-------+-------+-------+-------+-------+-------+-------+-------+ 209 * 0 | vdev1 | GRID | ASIZE | 210 * +-------+-------+-------+-------+-------+-------+-------+-------+ 211 * 1 |G| offset1 | 212 * +-------+-------+-------+-------+-------+-------+-------+-------+ 213 * 2 | vdev2 | GRID | ASIZE | 214 * +-------+-------+-------+-------+-------+-------+-------+-------+ 215 * 3 |G| offset2 | 216 * +-------+-------+-------+-------+-------+-------+-------+-------+ 217 * 4 | salt | 218 * +-------+-------+-------+-------+-------+-------+-------+-------+ 219 * 5 | IV1 | 220 * +-------+-------+-------+-------+-------+-------+-------+-------+ 221 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | 222 * +-------+-------+-------+-------+-------+-------+-------+-------+ 223 * 7 | padding | 224 * +-------+-------+-------+-------+-------+-------+-------+-------+ 225 * 8 | padding | 226 * +-------+-------+-------+-------+-------+-------+-------+-------+ 227 * 9 | physical birth txg | 228 * +-------+-------+-------+-------+-------+-------+-------+-------+ 229 * a | logical birth txg | 230 * +-------+-------+-------+-------+-------+-------+-------+-------+ 231 * b | IV2 | fill count | 232 * +-------+-------+-------+-------+-------+-------+-------+-------+ 233 * c | checksum[0] | 234 * +-------+-------+-------+-------+-------+-------+-------+-------+ 235 * d | checksum[1] | 236 * +-------+-------+-------+-------+-------+-------+-------+-------+ 237 * e | MAC[0] | 238 * +-------+-------+-------+-------+-------+-------+-------+-------+ 239 * f | MAC[1] | 240 * +-------+-------+-------+-------+-------+-------+-------+-------+ 241 * 242 * Legend: 243 * 244 * salt Salt for generating encryption keys 245 * IV1 First 64 bits of encryption IV 246 * X Block requires encryption handling (set to 1) 247 * E blkptr_t contains embedded data (set to 0, see below) 248 * fill count number of non-zero blocks under this bp (truncated to 32 bits) 249 * IV2 Last 32 bits of encryption IV 250 * checksum[2] 128-bit checksum of the data this bp describes 251 * MAC[2] 128-bit message authentication code for this data 252 * 253 * The X bit being set indicates that this block is one of 3 types. If this is 254 * a level 0 block with an encrypted object type, the block is encrypted 255 * (see BP_IS_ENCRYPTED()). If this is a level 0 block with an unencrypted 256 * object type, this block is authenticated with an HMAC (see 257 * BP_IS_AUTHENTICATED()). Otherwise (if level > 0), this bp will use the MAC 258 * words to store a checksum-of-MACs from the level below (see 259 * BP_HAS_INDIRECT_MAC_CKSUM()). For convenience in the code, BP_IS_PROTECTED() 260 * refers to both encrypted and authenticated blocks and BP_USES_CRYPT() 261 * refers to any of these 3 kinds of blocks. 262 * 263 * The additional encryption parameters are the salt, IV, and MAC which are 264 * explained in greater detail in the block comment at the top of zio_crypt.c. 265 * The MAC occupies half of the checksum space since it serves a very similar 266 * purpose: to prevent data corruption on disk. The only functional difference 267 * is that the checksum is used to detect on-disk corruption whether or not the 268 * encryption key is loaded and the MAC provides additional protection against 269 * malicious disk tampering. We use the 3rd DVA to store the salt and first 270 * 64 bits of the IV. As a result encrypted blocks can only have 2 copies 271 * maximum instead of the normal 3. The last 32 bits of the IV are stored in 272 * the upper bits of what is usually the fill count. Note that only blocks at 273 * level 0 or -2 are ever encrypted, which allows us to guarantee that these 274 * 32 bits are not trampled over by other code (see zio_crypt.c for details). 275 * The salt and IV are not used for authenticated bps or bps with an indirect 276 * MAC checksum, so these blocks can utilize all 3 DVAs and the full 64 bits 277 * for the fill count. 278 */ 279 280 /* 281 * "Embedded" blkptr_t's don't actually point to a block, instead they 282 * have a data payload embedded in the blkptr_t itself. See the comment 283 * in blkptr.c for more details. 284 * 285 * The blkptr_t is laid out as follows: 286 * 287 * 64 56 48 40 32 24 16 8 0 288 * +-------+-------+-------+-------+-------+-------+-------+-------+ 289 * 0 | payload | 290 * 1 | payload | 291 * 2 | payload | 292 * 3 | payload | 293 * 4 | payload | 294 * 5 | payload | 295 * +-------+-------+-------+-------+-------+-------+-------+-------+ 296 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | 297 * +-------+-------+-------+-------+-------+-------+-------+-------+ 298 * 7 | payload | 299 * 8 | payload | 300 * 9 | payload | 301 * +-------+-------+-------+-------+-------+-------+-------+-------+ 302 * a | logical birth txg | 303 * +-------+-------+-------+-------+-------+-------+-------+-------+ 304 * b | payload | 305 * c | payload | 306 * d | payload | 307 * e | payload | 308 * f | payload | 309 * +-------+-------+-------+-------+-------+-------+-------+-------+ 310 * 311 * Legend: 312 * 313 * payload contains the embedded data 314 * B (byteorder) byteorder (endianness) 315 * D (dedup) padding (set to zero) 316 * X encryption (set to zero; see above) 317 * E (embedded) set to one 318 * lvl indirection level 319 * type DMU object type 320 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) 321 * comp compression function of payload 322 * PSIZE size of payload after compression, in bytes 323 * LSIZE logical size of payload, in bytes 324 * note that 25 bits is enough to store the largest 325 * "normal" BP's LSIZE (2^16 * 2^9) in bytes 326 * log. birth transaction group in which the block was logically born 327 * 328 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded 329 * bp's they are stored in units of SPA_MINBLOCKSHIFT. 330 * Generally, the generic BP_GET_*() macros can be used on embedded BP's. 331 * The B, D, X, lvl, type, and comp fields are stored the same as with normal 332 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must 333 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before 334 * other macros, as they assert that they are only used on BP's of the correct 335 * "embedded-ness". Encrypted blkptr_t's cannot be embedded because they use 336 * the payload space for encryption parameters (see the comment above on 337 * how encryption parameters are stored). 338 */ 339 340 #define BPE_GET_ETYPE(bp) \ 341 (ASSERT(BP_IS_EMBEDDED(bp)), \ 342 BF64_GET((bp)->blk_prop, 40, 8)) 343 #define BPE_SET_ETYPE(bp, t) do { \ 344 ASSERT(BP_IS_EMBEDDED(bp)); \ 345 BF64_SET((bp)->blk_prop, 40, 8, t); \ 346 _NOTE(CONSTCOND) } while (0) 347 348 #define BPE_GET_LSIZE(bp) \ 349 (ASSERT(BP_IS_EMBEDDED(bp)), \ 350 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) 351 #define BPE_SET_LSIZE(bp, x) do { \ 352 ASSERT(BP_IS_EMBEDDED(bp)); \ 353 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ 354 _NOTE(CONSTCOND) } while (0) 355 356 #define BPE_GET_PSIZE(bp) \ 357 (ASSERT(BP_IS_EMBEDDED(bp)), \ 358 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) 359 #define BPE_SET_PSIZE(bp, x) do { \ 360 ASSERT(BP_IS_EMBEDDED(bp)); \ 361 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ 362 _NOTE(CONSTCOND) } while (0) 363 364 typedef enum bp_embedded_type { 365 BP_EMBEDDED_TYPE_DATA, 366 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ 367 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED 368 } bp_embedded_type_t; 369 370 #define BPE_NUM_WORDS 14 371 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) 372 #define BPE_IS_PAYLOADWORD(bp, wp) \ 373 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) 374 375 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ 376 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ 377 #define SPA_SYNC_MIN_VDEVS 3 /* min vdevs to update during sync */ 378 379 /* 380 * A block is a hole when it has either 1) never been written to, or 381 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads 382 * without physically allocating disk space. Holes are represented in the 383 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is 384 * done through the BP_IS_HOLE macro. For holes, the logical size, level, 385 * DMU object type, and birth times are all also stored for holes that 386 * were written to at some point (i.e. were punched after having been filled). 387 */ 388 typedef struct blkptr { 389 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ 390 uint64_t blk_prop; /* size, compression, type, etc */ 391 uint64_t blk_pad[2]; /* Extra space for the future */ 392 uint64_t blk_phys_birth; /* txg when block was allocated */ 393 uint64_t blk_birth; /* transaction group at birth */ 394 uint64_t blk_fill; /* fill count */ 395 zio_cksum_t blk_cksum; /* 256-bit checksum */ 396 } blkptr_t; 397 398 /* 399 * Macros to get and set fields in a bp or DVA. 400 */ 401 #define DVA_GET_ASIZE(dva) \ 402 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) 403 #define DVA_SET_ASIZE(dva, x) \ 404 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ 405 SPA_MINBLOCKSHIFT, 0, x) 406 407 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) 408 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) 409 410 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, SPA_VDEVBITS) 411 #define DVA_SET_VDEV(dva, x) \ 412 BF64_SET((dva)->dva_word[0], 32, SPA_VDEVBITS, x) 413 414 #define DVA_GET_OFFSET(dva) \ 415 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) 416 #define DVA_SET_OFFSET(dva, x) \ 417 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) 418 419 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) 420 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) 421 422 #define BP_GET_LSIZE(bp) \ 423 (BP_IS_EMBEDDED(bp) ? \ 424 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ 425 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 426 #define BP_SET_LSIZE(bp, x) do { \ 427 ASSERT(!BP_IS_EMBEDDED(bp)); \ 428 BF64_SET_SB((bp)->blk_prop, \ 429 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 430 _NOTE(CONSTCOND) } while (0) 431 432 #define BP_GET_PSIZE(bp) \ 433 (BP_IS_EMBEDDED(bp) ? 0 : \ 434 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1)) 435 #define BP_SET_PSIZE(bp, x) do { \ 436 ASSERT(!BP_IS_EMBEDDED(bp)); \ 437 BF64_SET_SB((bp)->blk_prop, \ 438 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ 439 _NOTE(CONSTCOND) } while (0) 440 441 #define BP_GET_COMPRESS(bp) \ 442 BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS) 443 #define BP_SET_COMPRESS(bp, x) \ 444 BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x) 445 446 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) 447 #define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x) 448 449 #define BP_GET_CHECKSUM(bp) \ 450 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \ 451 BF64_GET((bp)->blk_prop, 40, 8)) 452 #define BP_SET_CHECKSUM(bp, x) do { \ 453 ASSERT(!BP_IS_EMBEDDED(bp)); \ 454 BF64_SET((bp)->blk_prop, 40, 8, x); \ 455 _NOTE(CONSTCOND) } while (0) 456 457 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) 458 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) 459 460 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) 461 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) 462 463 /* encrypted, authenticated, and MAC cksum bps use the same bit */ 464 #define BP_USES_CRYPT(bp) BF64_GET((bp)->blk_prop, 61, 1) 465 #define BP_SET_CRYPT(bp, x) BF64_SET((bp)->blk_prop, 61, 1, x) 466 467 #define BP_IS_ENCRYPTED(bp) \ 468 (BP_USES_CRYPT(bp) && \ 469 BP_GET_LEVEL(bp) == 0 && \ 470 DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp))) 471 472 #define BP_IS_AUTHENTICATED(bp) \ 473 (BP_USES_CRYPT(bp) && \ 474 BP_GET_LEVEL(bp) == 0 && \ 475 !DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp))) 476 477 #define BP_HAS_INDIRECT_MAC_CKSUM(bp) \ 478 (BP_USES_CRYPT(bp) && BP_GET_LEVEL(bp) > 0) 479 480 #define BP_IS_PROTECTED(bp) \ 481 (BP_IS_ENCRYPTED(bp) || BP_IS_AUTHENTICATED(bp)) 482 483 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) 484 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) 485 486 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) 487 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) 488 489 #define BP_PHYSICAL_BIRTH(bp) \ 490 (BP_IS_EMBEDDED(bp) ? 0 : \ 491 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) 492 493 #define BP_SET_BIRTH(bp, logical, physical) \ 494 { \ 495 ASSERT(!BP_IS_EMBEDDED(bp)); \ 496 (bp)->blk_birth = (logical); \ 497 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \ 498 } 499 500 #define BP_GET_FILL(bp) \ 501 ((BP_IS_ENCRYPTED(bp)) ? BF64_GET((bp)->blk_fill, 0, 32) : \ 502 ((BP_IS_EMBEDDED(bp)) ? 1 : (bp)->blk_fill)) 503 504 #define BP_SET_FILL(bp, fill) \ 505 { \ 506 if (BP_IS_ENCRYPTED(bp)) \ 507 BF64_SET((bp)->blk_fill, 0, 32, fill); \ 508 else \ 509 (bp)->blk_fill = fill; \ 510 } 511 512 #define BP_GET_IV2(bp) \ 513 (ASSERT(BP_IS_ENCRYPTED(bp)), \ 514 BF64_GET((bp)->blk_fill, 32, 32)) 515 #define BP_SET_IV2(bp, iv2) \ 516 { \ 517 ASSERT(BP_IS_ENCRYPTED(bp)); \ 518 BF64_SET((bp)->blk_fill, 32, 32, iv2); \ 519 } 520 521 #define BP_IS_METADATA(bp) \ 522 (BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) 523 524 #define BP_GET_ASIZE(bp) \ 525 (BP_IS_EMBEDDED(bp) ? 0 : \ 526 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 527 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 528 (DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp))) 529 530 #define BP_GET_UCSIZE(bp) \ 531 (BP_IS_METADATA(bp) ? BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)) 532 533 #define BP_GET_NDVAS(bp) \ 534 (BP_IS_EMBEDDED(bp) ? 0 : \ 535 !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ 536 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ 537 (!!DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp))) 538 539 #define BP_COUNT_GANG(bp) \ 540 (BP_IS_EMBEDDED(bp) ? 0 : \ 541 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \ 542 DVA_GET_GANG(&(bp)->blk_dva[1]) + \ 543 (DVA_GET_GANG(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))) 544 545 #define DVA_EQUAL(dva1, dva2) \ 546 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ 547 (dva1)->dva_word[0] == (dva2)->dva_word[0]) 548 549 #define BP_EQUAL(bp1, bp2) \ 550 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \ 551 (bp1)->blk_birth == (bp2)->blk_birth && \ 552 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \ 553 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \ 554 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2])) 555 556 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) 557 558 #define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0]) 559 #define BP_IS_GANG(bp) \ 560 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp))) 561 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ 562 (dva)->dva_word[1] == 0ULL) 563 #define BP_IS_HOLE(bp) \ 564 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp))) 565 566 /* BP_IS_RAIDZ(bp) assumes no block compression */ 567 #define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \ 568 BP_GET_PSIZE(bp)) 569 570 #define BP_ZERO(bp) \ 571 { \ 572 (bp)->blk_dva[0].dva_word[0] = 0; \ 573 (bp)->blk_dva[0].dva_word[1] = 0; \ 574 (bp)->blk_dva[1].dva_word[0] = 0; \ 575 (bp)->blk_dva[1].dva_word[1] = 0; \ 576 (bp)->blk_dva[2].dva_word[0] = 0; \ 577 (bp)->blk_dva[2].dva_word[1] = 0; \ 578 (bp)->blk_prop = 0; \ 579 (bp)->blk_pad[0] = 0; \ 580 (bp)->blk_pad[1] = 0; \ 581 (bp)->blk_phys_birth = 0; \ 582 (bp)->blk_birth = 0; \ 583 (bp)->blk_fill = 0; \ 584 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ 585 } 586 587 #ifdef _BIG_ENDIAN 588 #define ZFS_HOST_BYTEORDER (0ULL) 589 #else 590 #define ZFS_HOST_BYTEORDER (1ULL) 591 #endif 592 593 #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) 594 595 #define BP_SPRINTF_LEN 400 596 597 /* 598 * This macro allows code sharing between zfs, libzpool, and mdb. 599 * 'func' is either snprintf() or mdb_snprintf(). 600 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line. 601 */ 602 #define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \ 603 { \ 604 static const char *copyname[] = \ 605 { "zero", "single", "double", "triple" }; \ 606 int len = 0; \ 607 int copies = 0; \ 608 const char *crypt_type; \ 609 if (bp != NULL) { \ 610 if (BP_IS_ENCRYPTED(bp)) { \ 611 crypt_type = "encrypted"; \ 612 } else if (BP_IS_AUTHENTICATED(bp)) { \ 613 crypt_type = "authenticated"; \ 614 } else if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) { \ 615 crypt_type = "indirect-MAC"; \ 616 } else { \ 617 crypt_type = "unencrypted"; \ 618 } \ 619 } \ 620 if (bp == NULL) { \ 621 len += func(buf + len, size - len, "<NULL>"); \ 622 } else if (BP_IS_HOLE(bp)) { \ 623 len += func(buf + len, size - len, \ 624 "HOLE [L%llu %s] " \ 625 "size=%llxL birth=%lluL", \ 626 (u_longlong_t)BP_GET_LEVEL(bp), \ 627 type, \ 628 (u_longlong_t)BP_GET_LSIZE(bp), \ 629 (u_longlong_t)bp->blk_birth); \ 630 } else if (BP_IS_EMBEDDED(bp)) { \ 631 len = func(buf + len, size - len, \ 632 "EMBEDDED [L%llu %s] et=%u %s " \ 633 "size=%llxL/%llxP birth=%lluL", \ 634 (u_longlong_t)BP_GET_LEVEL(bp), \ 635 type, \ 636 (int)BPE_GET_ETYPE(bp), \ 637 compress, \ 638 (u_longlong_t)BPE_GET_LSIZE(bp), \ 639 (u_longlong_t)BPE_GET_PSIZE(bp), \ 640 (u_longlong_t)bp->blk_birth); \ 641 } else { \ 642 for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \ 643 const dva_t *dva = &bp->blk_dva[d]; \ 644 if (DVA_IS_VALID(dva)) \ 645 copies++; \ 646 len += func(buf + len, size - len, \ 647 "DVA[%d]=<%llu:%llx:%llx>%c", d, \ 648 (u_longlong_t)DVA_GET_VDEV(dva), \ 649 (u_longlong_t)DVA_GET_OFFSET(dva), \ 650 (u_longlong_t)DVA_GET_ASIZE(dva), \ 651 ws); \ 652 } \ 653 if (BP_IS_ENCRYPTED(bp)) { \ 654 len += func(buf + len, size - len, \ 655 "salt=%llx iv=%llx:%llx%c", \ 656 (u_longlong_t)bp->blk_dva[2].dva_word[0], \ 657 (u_longlong_t)bp->blk_dva[2].dva_word[1], \ 658 (u_longlong_t)BP_GET_IV2(bp), \ 659 ws); \ 660 } \ 661 if (BP_IS_GANG(bp) && \ 662 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \ 663 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \ 664 copies--; \ 665 len += func(buf + len, size - len, \ 666 "[L%llu %s] %s %s %s %s %s %s %s%c" \ 667 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \ 668 "cksum=%llx:%llx:%llx:%llx", \ 669 (u_longlong_t)BP_GET_LEVEL(bp), \ 670 type, \ 671 checksum, \ 672 compress, \ 673 crypt_type, \ 674 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \ 675 BP_IS_GANG(bp) ? "gang" : "contiguous", \ 676 BP_GET_DEDUP(bp) ? "dedup" : "unique", \ 677 copyname[copies], \ 678 ws, \ 679 (u_longlong_t)BP_GET_LSIZE(bp), \ 680 (u_longlong_t)BP_GET_PSIZE(bp), \ 681 (u_longlong_t)bp->blk_birth, \ 682 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \ 683 (u_longlong_t)BP_GET_FILL(bp), \ 684 ws, \ 685 (u_longlong_t)bp->blk_cksum.zc_word[0], \ 686 (u_longlong_t)bp->blk_cksum.zc_word[1], \ 687 (u_longlong_t)bp->blk_cksum.zc_word[2], \ 688 (u_longlong_t)bp->blk_cksum.zc_word[3]); \ 689 } \ 690 ASSERT(len < size); \ 691 } 692 693 #define BP_GET_BUFC_TYPE(bp) \ 694 (BP_IS_METADATA(bp) ? ARC_BUFC_METADATA : ARC_BUFC_DATA) 695 696 typedef enum spa_import_type { 697 SPA_IMPORT_EXISTING, 698 SPA_IMPORT_ASSEMBLE 699 } spa_import_type_t; 700 701 /* 702 * Send TRIM commands in-line during normal pool operation while deleting. 703 * OFF: no 704 * ON: yes 705 */ 706 typedef enum { 707 SPA_AUTOTRIM_OFF = 0, /* default */ 708 SPA_AUTOTRIM_ON 709 } spa_autotrim_t; 710 711 /* 712 * Reason TRIM command was issued, used internally for accounting purposes. 713 */ 714 typedef enum trim_type { 715 TRIM_TYPE_MANUAL = 0, 716 TRIM_TYPE_AUTO = 1, 717 } trim_type_t; 718 719 /* state manipulation functions */ 720 extern int spa_open(const char *pool, spa_t **, void *tag); 721 extern int spa_open_rewind(const char *pool, spa_t **, void *tag, 722 nvlist_t *policy, nvlist_t **config); 723 extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot, 724 size_t buflen); 725 extern int spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props, 726 nvlist_t *zplprops, struct dsl_crypto_params *dcp); 727 extern int spa_import_rootpool(char *devpath, char *devid, uint64_t pool_guid, 728 uint64_t vdev_guid); 729 extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props, 730 uint64_t flags); 731 extern nvlist_t *spa_tryimport(nvlist_t *tryconfig); 732 extern int spa_destroy(char *pool); 733 extern int spa_checkpoint(const char *pool); 734 extern int spa_checkpoint_discard(const char *pool); 735 extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force, 736 boolean_t hardforce); 737 extern int spa_reset(char *pool); 738 extern void spa_async_request(spa_t *spa, int flag); 739 extern void spa_async_unrequest(spa_t *spa, int flag); 740 extern void spa_async_suspend(spa_t *spa); 741 extern void spa_async_resume(spa_t *spa); 742 extern int spa_async_tasks(spa_t *spa); 743 extern spa_t *spa_inject_addref(char *pool); 744 extern void spa_inject_delref(spa_t *spa); 745 extern void spa_scan_stat_init(spa_t *spa); 746 extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps); 747 748 #define SPA_ASYNC_CONFIG_UPDATE 0x01 749 #define SPA_ASYNC_REMOVE 0x02 750 #define SPA_ASYNC_PROBE 0x04 751 #define SPA_ASYNC_RESILVER_DONE 0x08 752 #define SPA_ASYNC_RESILVER 0x10 753 #define SPA_ASYNC_AUTOEXPAND 0x20 754 #define SPA_ASYNC_REMOVE_DONE 0x40 755 #define SPA_ASYNC_REMOVE_STOP 0x80 756 #define SPA_ASYNC_INITIALIZE_RESTART 0x100 757 #define SPA_ASYNC_TRIM_RESTART 0x200 758 #define SPA_ASYNC_AUTOTRIM_RESTART 0x400 759 #define SPA_ASYNC_L2CACHE_REBUILD 0x800 760 761 /* 762 * Controls the behavior of spa_vdev_remove(). 763 */ 764 #define SPA_REMOVE_UNSPARE 0x01 765 #define SPA_REMOVE_DONE 0x02 766 767 /* device manipulation */ 768 extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot); 769 extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, 770 int replacing); 771 extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, 772 int replace_done); 773 extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare); 774 extern boolean_t spa_vdev_remove_active(spa_t *spa); 775 extern int spa_vdev_initialize(spa_t *spa, nvlist_t *nv, uint64_t cmd_type, 776 nvlist_t *vdev_errlist); 777 extern int spa_vdev_trim(spa_t *spa, nvlist_t *nv, uint64_t cmd_type, 778 uint64_t rate, boolean_t partial, boolean_t secure, nvlist_t *vdev_errlist); 779 extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath); 780 extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru); 781 extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config, 782 nvlist_t *props, boolean_t exp); 783 784 /* spare state (which is global across all pools) */ 785 extern void spa_spare_add(vdev_t *vd); 786 extern void spa_spare_remove(vdev_t *vd); 787 extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt); 788 extern void spa_spare_activate(vdev_t *vd); 789 790 /* spare polling */ 791 extern void spa_spare_poll(spa_t *spa); 792 793 /* L2ARC state (which is global across all pools) */ 794 extern void spa_l2cache_add(vdev_t *vd); 795 extern void spa_l2cache_remove(vdev_t *vd); 796 extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool); 797 extern void spa_l2cache_activate(vdev_t *vd); 798 extern void spa_l2cache_drop(spa_t *spa); 799 800 /* scanning */ 801 extern int spa_scan(spa_t *spa, pool_scan_func_t func); 802 extern int spa_scan_stop(spa_t *spa); 803 extern int spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t flag); 804 805 /* spa syncing */ 806 extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */ 807 extern void spa_sync_allpools(void); 808 809 /* spa namespace global mutex */ 810 extern kmutex_t spa_namespace_lock; 811 812 /* 813 * SPA configuration functions in spa_config.c 814 */ 815 816 #define SPA_CONFIG_UPDATE_POOL 0 817 #define SPA_CONFIG_UPDATE_VDEVS 1 818 819 extern void spa_write_cachefile(spa_t *, boolean_t, boolean_t); 820 extern void spa_config_load(void); 821 extern nvlist_t *spa_all_configs(uint64_t *); 822 extern void spa_config_set(spa_t *spa, nvlist_t *config); 823 extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, 824 int getstats); 825 extern void spa_config_update(spa_t *spa, int what); 826 827 /* 828 * Miscellaneous SPA routines in spa_misc.c 829 */ 830 831 /* Namespace manipulation */ 832 extern spa_t *spa_lookup(const char *name); 833 extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot); 834 extern void spa_remove(spa_t *spa); 835 extern spa_t *spa_next(spa_t *prev); 836 837 /* Refcount functions */ 838 extern void spa_open_ref(spa_t *spa, void *tag); 839 extern void spa_close(spa_t *spa, void *tag); 840 extern void spa_async_close(spa_t *spa, void *tag); 841 extern boolean_t spa_refcount_zero(spa_t *spa); 842 843 #define SCL_NONE 0x00 844 #define SCL_CONFIG 0x01 845 #define SCL_STATE 0x02 846 #define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */ 847 #define SCL_ALLOC 0x08 848 #define SCL_ZIO 0x10 849 #define SCL_FREE 0x20 850 #define SCL_VDEV 0x40 851 #define SCL_LOCKS 7 852 #define SCL_ALL ((1 << SCL_LOCKS) - 1) 853 #define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO) 854 855 /* Assorted pool IO kstats */ 856 typedef struct spa_iostats { 857 kstat_named_t trim_extents_written; 858 kstat_named_t trim_bytes_written; 859 kstat_named_t trim_extents_skipped; 860 kstat_named_t trim_bytes_skipped; 861 kstat_named_t trim_extents_failed; 862 kstat_named_t trim_bytes_failed; 863 kstat_named_t autotrim_extents_written; 864 kstat_named_t autotrim_bytes_written; 865 kstat_named_t autotrim_extents_skipped; 866 kstat_named_t autotrim_bytes_skipped; 867 kstat_named_t autotrim_extents_failed; 868 kstat_named_t autotrim_bytes_failed; 869 } spa_iostats_t; 870 871 extern int spa_import_progress_set_state(spa_t *, spa_load_state_t); 872 extern int spa_import_progress_set_max_txg(spa_t *, uint64_t); 873 extern int spa_import_progress_set_mmp_check(spa_t *, uint64_t); 874 extern void spa_import_progress_add(spa_t *); 875 extern void spa_import_progress_remove(spa_t *); 876 877 /* Pool configuration locks */ 878 extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw); 879 extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw); 880 extern void spa_config_exit(spa_t *spa, int locks, void *tag); 881 extern int spa_config_held(spa_t *spa, int locks, krw_t rw); 882 883 /* Pool vdev add/remove lock */ 884 extern uint64_t spa_vdev_enter(spa_t *spa); 885 extern uint64_t spa_vdev_config_enter(spa_t *spa); 886 extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, 887 int error, char *tag); 888 extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error); 889 890 /* Pool vdev state change lock */ 891 extern void spa_vdev_state_enter(spa_t *spa, int oplock); 892 extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error); 893 894 /* Log state */ 895 typedef enum spa_log_state { 896 SPA_LOG_UNKNOWN = 0, /* unknown log state */ 897 SPA_LOG_MISSING, /* missing log(s) */ 898 SPA_LOG_CLEAR, /* clear the log(s) */ 899 SPA_LOG_GOOD, /* log(s) are good */ 900 } spa_log_state_t; 901 902 extern spa_log_state_t spa_get_log_state(spa_t *spa); 903 extern void spa_set_log_state(spa_t *spa, spa_log_state_t state); 904 extern int spa_reset_logs(spa_t *spa); 905 906 /* Log claim callback */ 907 extern void spa_claim_notify(zio_t *zio); 908 909 /* Accessor functions */ 910 extern boolean_t spa_shutting_down(spa_t *spa); 911 extern struct dsl_pool *spa_get_dsl(spa_t *spa); 912 extern boolean_t spa_is_initializing(spa_t *spa); 913 extern boolean_t spa_indirect_vdevs_loaded(spa_t *spa); 914 extern blkptr_t *spa_get_rootblkptr(spa_t *spa); 915 extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp); 916 extern void spa_altroot(spa_t *, char *, size_t); 917 extern int spa_sync_pass(spa_t *spa); 918 extern char *spa_name(spa_t *spa); 919 extern uint64_t spa_guid(spa_t *spa); 920 extern uint64_t spa_load_guid(spa_t *spa); 921 extern uint64_t spa_last_synced_txg(spa_t *spa); 922 extern uint64_t spa_first_txg(spa_t *spa); 923 extern uint64_t spa_syncing_txg(spa_t *spa); 924 extern uint64_t spa_final_dirty_txg(spa_t *spa); 925 extern uint64_t spa_version(spa_t *spa); 926 extern pool_state_t spa_state(spa_t *spa); 927 extern spa_load_state_t spa_load_state(spa_t *spa); 928 extern uint64_t spa_freeze_txg(spa_t *spa); 929 extern uint64_t spa_get_worst_case_asize(spa_t *spa, uint64_t lsize); 930 extern uint64_t spa_get_dspace(spa_t *spa); 931 extern uint64_t spa_get_checkpoint_space(spa_t *spa); 932 extern uint64_t spa_get_slop_space(spa_t *spa); 933 extern void spa_update_dspace(spa_t *spa); 934 extern uint64_t spa_version(spa_t *spa); 935 extern boolean_t spa_deflate(spa_t *spa); 936 extern metaslab_class_t *spa_normal_class(spa_t *spa); 937 extern metaslab_class_t *spa_log_class(spa_t *spa); 938 extern metaslab_class_t *spa_special_class(spa_t *spa); 939 extern metaslab_class_t *spa_dedup_class(spa_t *spa); 940 extern metaslab_class_t *spa_preferred_class(spa_t *spa, uint64_t size, 941 dmu_object_type_t objtype, uint_t level, uint_t special_smallblk); 942 943 extern void spa_evicting_os_register(spa_t *, objset_t *os); 944 extern void spa_evicting_os_deregister(spa_t *, objset_t *os); 945 extern void spa_evicting_os_wait(spa_t *spa); 946 extern int spa_max_replication(spa_t *spa); 947 extern int spa_prev_software_version(spa_t *spa); 948 extern int spa_busy(void); 949 extern uint8_t spa_get_failmode(spa_t *spa); 950 extern boolean_t spa_suspended(spa_t *spa); 951 extern uint64_t spa_bootfs(spa_t *spa); 952 extern uint64_t spa_delegation(spa_t *spa); 953 extern objset_t *spa_meta_objset(spa_t *spa); 954 extern space_map_t *spa_syncing_log_sm(spa_t *spa); 955 extern uint64_t spa_deadman_synctime(spa_t *spa); 956 extern uint64_t spa_dirty_data(spa_t *spa); 957 extern spa_autotrim_t spa_get_autotrim(spa_t *spa); 958 959 /* Miscellaneous support routines */ 960 extern void spa_load_failed(spa_t *spa, const char *fmt, ...); 961 extern void spa_load_note(spa_t *spa, const char *fmt, ...); 962 extern void spa_activate_mos_feature(spa_t *spa, const char *feature, 963 dmu_tx_t *tx); 964 extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature); 965 extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid); 966 extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid); 967 extern char *spa_strdup(const char *); 968 extern void spa_strfree(char *); 969 extern uint64_t spa_get_random(uint64_t range); 970 extern uint64_t spa_generate_guid(spa_t *spa); 971 extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp); 972 extern void spa_freeze(spa_t *spa); 973 extern int spa_change_guid(spa_t *spa); 974 extern void spa_upgrade(spa_t *spa, uint64_t version); 975 extern void spa_evict_all(void); 976 extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid, 977 boolean_t l2cache); 978 extern boolean_t spa_has_spare(spa_t *, uint64_t guid); 979 extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva); 980 extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp); 981 extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp); 982 extern boolean_t spa_has_slogs(spa_t *spa); 983 extern boolean_t spa_is_root(spa_t *spa); 984 extern boolean_t spa_writeable(spa_t *spa); 985 extern boolean_t spa_has_pending_synctask(spa_t *spa); 986 extern int spa_maxblocksize(spa_t *spa); 987 extern int spa_maxdnodesize(spa_t *spa); 988 extern boolean_t spa_multihost(spa_t *spa); 989 extern unsigned long spa_get_hostid(void); 990 extern boolean_t spa_has_checkpoint(spa_t *spa); 991 extern boolean_t spa_importing_readonly_checkpoint(spa_t *spa); 992 extern boolean_t spa_suspend_async_destroy(spa_t *spa); 993 extern uint64_t spa_min_claim_txg(spa_t *spa); 994 extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp); 995 extern boolean_t zfs_dva_valid(spa_t *spa, const dva_t *dva, 996 const blkptr_t *bp); 997 typedef void (*spa_remap_cb_t)(uint64_t vdev, uint64_t offset, uint64_t size, 998 void *arg); 999 extern boolean_t spa_remap_blkptr(spa_t *spa, blkptr_t *bp, 1000 spa_remap_cb_t callback, void *arg); 1001 extern uint64_t spa_get_last_removal_txg(spa_t *spa); 1002 extern boolean_t spa_trust_config(spa_t *spa); 1003 extern uint64_t spa_missing_tvds_allowed(spa_t *spa); 1004 extern void spa_set_missing_tvds(spa_t *spa, uint64_t missing); 1005 extern boolean_t spa_top_vdevs_spacemap_addressable(spa_t *spa); 1006 extern uint64_t spa_total_metaslabs(spa_t *spa); 1007 extern void spa_activate_allocation_classes(spa_t *, dmu_tx_t *); 1008 1009 extern int spa_mode(spa_t *spa); 1010 extern uint64_t zfs_strtonum(const char *str, char **nptr); 1011 1012 extern char *spa_his_ievent_table[]; 1013 1014 extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx); 1015 extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read, 1016 char *his_buf); 1017 extern int spa_history_log(spa_t *spa, const char *his_buf); 1018 extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl); 1019 extern void spa_history_log_version(spa_t *spa, const char *operation); 1020 extern void spa_history_log_internal(spa_t *spa, const char *operation, 1021 dmu_tx_t *tx, const char *fmt, ...); 1022 extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op, 1023 dmu_tx_t *tx, const char *fmt, ...); 1024 extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation, 1025 dmu_tx_t *tx, const char *fmt, ...); 1026 1027 /* error handling */ 1028 struct zbookmark_phys; 1029 extern void spa_log_error(spa_t *spa, const struct zbookmark_phys *zb); 1030 extern int zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd, 1031 const struct zbookmark_phys *zb, struct zio *zio, uint64_t stateoroffset, 1032 uint64_t length); 1033 extern boolean_t zfs_ereport_is_valid(const char *class, spa_t *spa, vdev_t *vd, 1034 zio_t *zio); 1035 extern void zfs_post_remove(spa_t *spa, vdev_t *vd); 1036 extern void zfs_post_state_change(spa_t *spa, vdev_t *vd); 1037 extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd); 1038 extern uint64_t spa_get_errlog_size(spa_t *spa); 1039 extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count); 1040 extern void spa_errlog_rotate(spa_t *spa); 1041 extern void spa_errlog_drain(spa_t *spa); 1042 extern void spa_errlog_sync(spa_t *spa, uint64_t txg); 1043 extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub); 1044 1045 /* vdev cache */ 1046 extern void vdev_cache_stat_init(void); 1047 extern void vdev_cache_stat_fini(void); 1048 1049 /* vdev mirror */ 1050 extern void vdev_mirror_stat_init(void); 1051 extern void vdev_mirror_stat_fini(void); 1052 1053 /* Initialization and termination */ 1054 extern void spa_init(int flags); 1055 extern void spa_fini(void); 1056 extern void spa_boot_init(void); 1057 1058 /* properties */ 1059 extern int spa_prop_set(spa_t *spa, nvlist_t *nvp); 1060 extern int spa_prop_get(spa_t *spa, nvlist_t **nvp); 1061 extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx); 1062 extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t); 1063 1064 /* asynchronous event notification */ 1065 extern void spa_event_notify(spa_t *spa, vdev_t *vdev, nvlist_t *hist_nvl, 1066 const char *name); 1067 extern sysevent_t *spa_event_create(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl, 1068 const char *name); 1069 extern void spa_event_post(sysevent_t *ev); 1070 extern void spa_event_discard(sysevent_t *ev); 1071 extern void zfs_post_dle_sysevent(const char *); 1072 1073 #ifdef ZFS_DEBUG 1074 #define dprintf_bp(bp, fmt, ...) do { \ 1075 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 1076 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \ 1077 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \ 1078 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \ 1079 kmem_free(__blkbuf, BP_SPRINTF_LEN); \ 1080 } \ 1081 _NOTE(CONSTCOND) } while (0) 1082 #else 1083 #define dprintf_bp(bp, fmt, ...) 1084 #endif 1085 1086 extern int spa_mode_global; /* mode, e.g. FREAD | FWRITE */ 1087 1088 #ifdef __cplusplus 1089 } 1090 #endif 1091 1092 #endif /* _SYS_SPA_H */ 1093