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) 2013 by Delphix. All rights reserved.
  24  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
  25  */
  26 
  27 #ifndef _SYS_SPA_H
  28 #define _SYS_SPA_H
  29 
  30 #include <sys/avl.h>
  31 #include <sys/zfs_context.h>
  32 #include <sys/nvpair.h>
  33 #include <sys/sysmacros.h>
  34 #include <sys/types.h>
  35 #include <sys/fs/zfs.h>
  36 
  37 #ifdef  __cplusplus
  38 extern "C" {
  39 #endif
  40 
  41 /*
  42  * Forward references that lots of things need.
  43  */
  44 typedef struct spa spa_t;
  45 typedef struct vdev vdev_t;
  46 typedef struct metaslab metaslab_t;
  47 typedef struct metaslab_group metaslab_group_t;
  48 typedef struct metaslab_class metaslab_class_t;
  49 typedef struct zio zio_t;
  50 typedef struct zilog zilog_t;
  51 typedef struct spa_aux_vdev spa_aux_vdev_t;
  52 typedef struct ddt ddt_t;
  53 typedef struct ddt_entry ddt_entry_t;
  54 struct dsl_pool;
  55 struct dsl_dataset;
  56 
  57 /*
  58  * General-purpose 32-bit and 64-bit bitfield encodings.
  59  */
  60 #define BF32_DECODE(x, low, len)        P2PHASE((x) >> (low), 1U << (len))
  61 #define BF64_DECODE(x, low, len)        P2PHASE((x) >> (low), 1ULL << (len))
  62 #define BF32_ENCODE(x, low, len)        (P2PHASE((x), 1U << (len)) << (low))
  63 #define BF64_ENCODE(x, low, len)        (P2PHASE((x), 1ULL << (len)) << (low))
  64 
  65 #define BF32_GET(x, low, len)           BF32_DECODE(x, low, len)
  66 #define BF64_GET(x, low, len)           BF64_DECODE(x, low, len)
  67 
  68 #define BF32_SET(x, low, len, val) do { \
  69         ASSERT3U(val, <, 1U << (len)); \
  70         ASSERT3U(low + len, <=, 32); \
  71         (x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \
  72 _NOTE(CONSTCOND) } while (0)
  73 
  74 #define BF64_SET(x, low, len, val) do { \
  75         ASSERT3U(val, <, 1ULL << (len)); \
  76         ASSERT3U(low + len, <=, 64); \
  77         ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \
  78 _NOTE(CONSTCOND) } while (0)
  79 
  80 #define BF32_GET_SB(x, low, len, shift, bias)   \
  81         ((BF32_GET(x, low, len) + (bias)) << (shift))
  82 #define BF64_GET_SB(x, low, len, shift, bias)   \
  83         ((BF64_GET(x, low, len) + (bias)) << (shift))
  84 
  85 #define BF32_SET_SB(x, low, len, shift, bias, val) do { \
  86         ASSERT(IS_P2ALIGNED(val, 1U << shift)); \
  87         ASSERT3S((val) >> (shift), >=, bias); \
  88         BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \
  89 _NOTE(CONSTCOND) } while (0)
  90 #define BF64_SET_SB(x, low, len, shift, bias, val) do { \
  91         ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \
  92         ASSERT3S((val) >> (shift), >=, bias); \
  93         BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \
  94 _NOTE(CONSTCOND) } while (0)
  95 
  96 /*
  97  * We currently support nine block sizes, from 512 bytes to 128K.
  98  * We could go higher, but the benefits are near-zero and the cost
  99  * of COWing a giant block to modify one byte would become excessive.
 100  */
 101 #define SPA_MINBLOCKSHIFT       9
 102 #define SPA_MAXBLOCKSHIFT       17
 103 #define SPA_MINBLOCKSIZE        (1ULL << SPA_MINBLOCKSHIFT)
 104 #define SPA_MAXBLOCKSIZE        (1ULL << SPA_MAXBLOCKSHIFT)
 105 
 106 #define SPA_BLOCKSIZES          (SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1)
 107 
 108 /*
 109  * Size of block to hold the configuration data (a packed nvlist)
 110  */
 111 #define SPA_CONFIG_BLOCKSIZE    (1ULL << 14)
 112 
 113 /*
 114  * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
 115  * The ASIZE encoding should be at least 64 times larger (6 more bits)
 116  * to support up to 4-way RAID-Z mirror mode with worst-case gang block
 117  * overhead, three DVAs per bp, plus one more bit in case we do anything
 118  * else that expands the ASIZE.
 119  */
 120 #define SPA_LSIZEBITS           16      /* LSIZE up to 32M (2^16 * 512) */
 121 #define SPA_PSIZEBITS           16      /* PSIZE up to 32M (2^16 * 512) */
 122 #define SPA_ASIZEBITS           24      /* ASIZE up to 64 times larger  */
 123 
 124 /*
 125  * All SPA data is represented by 128-bit data virtual addresses (DVAs).
 126  * The members of the dva_t should be considered opaque outside the SPA.
 127  */
 128 typedef struct dva {
 129         uint64_t        dva_word[2];
 130 } dva_t;
 131 
 132 /*
 133  * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
 134  */
 135 typedef struct zio_cksum {
 136         uint64_t        zc_word[4];
 137 } zio_cksum_t;
 138 
 139 /*
 140  * Each block is described by its DVAs, time of birth, checksum, etc.
 141  * The word-by-word, bit-by-bit layout of the blkptr is as follows:
 142  *
 143  *      64      56      48      40      32      24      16      8       0
 144  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 145  * 0    |               vdev1           | GRID  |         ASIZE         |
 146  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 147  * 1    |G|                      offset1                                |
 148  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 149  * 2    |               vdev2           | GRID  |         ASIZE         |
 150  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 151  * 3    |G|                      offset2                                |
 152  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 153  * 4    |               vdev3           | GRID  |         ASIZE         |
 154  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 155  * 5    |G|                      offset3                                |
 156  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 157  * 6    |BDX|lvl| type  | cksum |E| comp|    PSIZE      |     LSIZE     |
 158  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 159  * 7    |                       padding                                 |
 160  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 161  * 8    |                       padding                                 |
 162  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 163  * 9    |                       physical birth txg                      |
 164  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 165  * a    |                       logical birth txg                       |
 166  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 167  * b    |                       fill count                              |
 168  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 169  * c    |                       checksum[0]                             |
 170  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 171  * d    |                       checksum[1]                             |
 172  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 173  * e    |                       checksum[2]                             |
 174  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 175  * f    |                       checksum[3]                             |
 176  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 177  *
 178  * Legend:
 179  *
 180  * vdev         virtual device ID
 181  * offset       offset into virtual device
 182  * LSIZE        logical size
 183  * PSIZE        physical size (after compression)
 184  * ASIZE        allocated size (including RAID-Z parity and gang block headers)
 185  * GRID         RAID-Z layout information (reserved for future use)
 186  * cksum        checksum function
 187  * comp         compression function
 188  * G            gang block indicator
 189  * B            byteorder (endianness)
 190  * D            dedup
 191  * X            encryption (on version 30, which is not supported)
 192  * E            blkptr_t contains embedded data (see below)
 193  * lvl          level of indirection
 194  * type         DMU object type
 195  * phys birth   txg of block allocation; zero if same as logical birth txg
 196  * log. birth   transaction group in which the block was logically born
 197  * fill count   number of non-zero blocks under this bp
 198  * checksum[4]  256-bit checksum of the data this bp describes
 199  */
 200 
 201 /*
 202  * "Embedded" blkptr_t's don't actually point to a block, instead they
 203  * have a data payload embedded in the blkptr_t itself.  See the comment
 204  * in blkptr.c for more details.
 205  *
 206  * The blkptr_t is laid out as follows:
 207  *
 208  *      64      56      48      40      32      24      16      8       0
 209  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 210  * 0    |      payload                                                  |
 211  * 1    |      payload                                                  |
 212  * 2    |      payload                                                  |
 213  * 3    |      payload                                                  |
 214  * 4    |      payload                                                  |
 215  * 5    |      payload                                                  |
 216  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 217  * 6    |BDX|lvl| type  | etype |E| comp| PSIZE|              LSIZE     |
 218  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 219  * 7    |      payload                                                  |
 220  * 8    |      payload                                                  |
 221  * 9    |      payload                                                  |
 222  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 223  * a    |                       logical birth txg                       |
 224  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 225  * b    |      payload                                                  |
 226  * c    |      payload                                                  |
 227  * d    |      payload                                                  |
 228  * e    |      payload                                                  |
 229  * f    |      payload                                                  |
 230  *      +-------+-------+-------+-------+-------+-------+-------+-------+
 231  *
 232  * Legend:
 233  *
 234  * payload              contains the embedded data
 235  * B (byteorder)        byteorder (endianness)
 236  * D (dedup)            padding (set to zero)
 237  * X                    encryption (set to zero; see above)
 238  * E (embedded)         set to one
 239  * lvl                  indirection level
 240  * type                 DMU object type
 241  * etype                how to interpret embedded data (BP_EMBEDDED_TYPE_*)
 242  * comp                 compression function of payload
 243  * PSIZE                size of payload after compression, in bytes
 244  * LSIZE                logical size of payload, in bytes
 245  *                      note that 25 bits is enough to store the largest
 246  *                      "normal" BP's LSIZE (2^16 * 2^9) in bytes
 247  * log. birth           transaction group in which the block was logically born
 248  *
 249  * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
 250  * bp's they are stored in units of SPA_MINBLOCKSHIFT.
 251  * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
 252  * The B, D, X, lvl, type, and comp fields are stored the same as with normal
 253  * BP's so the BP_SET_* macros can be used with them.  etype, PSIZE, LSIZE must
 254  * be set with the BPE_SET_* macros.  BP_SET_EMBEDDED() should be called before
 255  * other macros, as they assert that they are only used on BP's of the correct
 256  * "embedded-ness".
 257  */
 258 
 259 #define BPE_GET_ETYPE(bp)       \
 260         (ASSERT(BP_IS_EMBEDDED(bp)), \
 261         BF64_GET((bp)->blk_prop, 40, 8))
 262 #define BPE_SET_ETYPE(bp, t)    do { \
 263         ASSERT(BP_IS_EMBEDDED(bp)); \
 264         BF64_SET((bp)->blk_prop, 40, 8, t); \
 265 _NOTE(CONSTCOND) } while (0)
 266 
 267 #define BPE_GET_LSIZE(bp)       \
 268         (ASSERT(BP_IS_EMBEDDED(bp)), \
 269         BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
 270 #define BPE_SET_LSIZE(bp, x)    do { \
 271         ASSERT(BP_IS_EMBEDDED(bp)); \
 272         BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
 273 _NOTE(CONSTCOND) } while (0)
 274 
 275 #define BPE_GET_PSIZE(bp)       \
 276         (ASSERT(BP_IS_EMBEDDED(bp)), \
 277         BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
 278 #define BPE_SET_PSIZE(bp, x)    do { \
 279         ASSERT(BP_IS_EMBEDDED(bp)); \
 280         BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
 281 _NOTE(CONSTCOND) } while (0)
 282 
 283 typedef enum bp_embedded_type {
 284         BP_EMBEDDED_TYPE_DATA,
 285         BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */
 286         NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED
 287 } bp_embedded_type_t;
 288 
 289 #define BPE_NUM_WORDS 14
 290 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
 291 #define BPE_IS_PAYLOADWORD(bp, wp) \
 292         ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
 293 
 294 #define SPA_BLKPTRSHIFT 7               /* blkptr_t is 128 bytes        */
 295 #define SPA_DVAS_PER_BP 3               /* Number of DVAs in a bp       */
 296 
 297 /*
 298  * A block is a hole when it has either 1) never been written to, or
 299  * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
 300  * without physically allocating disk space. Holes are represented in the
 301  * blkptr_t structure by zeroed blk_dva. Correct checking for holes is
 302  * done through the BP_IS_HOLE macro. For holes, the logical size, level,
 303  * DMU object type, and birth times are all also stored for holes that
 304  * were written to at some point (i.e. were punched after having been filled).
 305  */
 306 typedef struct blkptr {
 307         dva_t           blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
 308         uint64_t        blk_prop;       /* size, compression, type, etc     */
 309         uint64_t        blk_pad[2];     /* Extra space for the future       */
 310         uint64_t        blk_phys_birth; /* txg when block was allocated     */
 311         uint64_t        blk_birth;      /* transaction group at birth       */
 312         uint64_t        blk_fill;       /* fill count                       */
 313         zio_cksum_t     blk_cksum;      /* 256-bit checksum                 */
 314 } blkptr_t;
 315 
 316 /*
 317  * Macros to get and set fields in a bp or DVA.
 318  */
 319 #define DVA_GET_ASIZE(dva)      \
 320         BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
 321 #define DVA_SET_ASIZE(dva, x)   \
 322         BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
 323         SPA_MINBLOCKSHIFT, 0, x)
 324 
 325 #define DVA_GET_GRID(dva)       BF64_GET((dva)->dva_word[0], 24, 8)
 326 #define DVA_SET_GRID(dva, x)    BF64_SET((dva)->dva_word[0], 24, 8, x)
 327 
 328 #define DVA_GET_VDEV(dva)       BF64_GET((dva)->dva_word[0], 32, 32)
 329 #define DVA_SET_VDEV(dva, x)    BF64_SET((dva)->dva_word[0], 32, 32, x)
 330 
 331 #define DVA_GET_OFFSET(dva)     \
 332         BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
 333 #define DVA_SET_OFFSET(dva, x)  \
 334         BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
 335 
 336 #define DVA_GET_GANG(dva)       BF64_GET((dva)->dva_word[1], 63, 1)
 337 #define DVA_SET_GANG(dva, x)    BF64_SET((dva)->dva_word[1], 63, 1, x)
 338 
 339 #define BP_GET_LSIZE(bp)        \
 340         (BP_IS_EMBEDDED(bp) ?   \
 341         (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
 342         BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
 343 #define BP_SET_LSIZE(bp, x)     do { \
 344         ASSERT(!BP_IS_EMBEDDED(bp)); \
 345         BF64_SET_SB((bp)->blk_prop, \
 346             0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
 347 _NOTE(CONSTCOND) } while (0)
 348 
 349 #define BP_GET_PSIZE(bp)        \
 350         (BP_IS_EMBEDDED(bp) ? 0 : \
 351         BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1))
 352 #define BP_SET_PSIZE(bp, x)     do { \
 353         ASSERT(!BP_IS_EMBEDDED(bp)); \
 354         BF64_SET_SB((bp)->blk_prop, \
 355             16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
 356 _NOTE(CONSTCOND) } while (0)
 357 
 358 #define BP_GET_COMPRESS(bp)             BF64_GET((bp)->blk_prop, 32, 7)
 359 #define BP_SET_COMPRESS(bp, x)          BF64_SET((bp)->blk_prop, 32, 7, x)
 360 
 361 #define BP_IS_EMBEDDED(bp)              BF64_GET((bp)->blk_prop, 39, 1)
 362 #define BP_SET_EMBEDDED(bp, x)          BF64_SET((bp)->blk_prop, 39, 1, x)
 363 
 364 #define BP_GET_CHECKSUM(bp)             \
 365         (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \
 366         BF64_GET((bp)->blk_prop, 40, 8))
 367 #define BP_SET_CHECKSUM(bp, x)          do { \
 368         ASSERT(!BP_IS_EMBEDDED(bp)); \
 369         BF64_SET((bp)->blk_prop, 40, 8, x); \
 370 _NOTE(CONSTCOND) } while (0)
 371 
 372 #define BP_GET_TYPE(bp)                 BF64_GET((bp)->blk_prop, 48, 8)
 373 #define BP_SET_TYPE(bp, x)              BF64_SET((bp)->blk_prop, 48, 8, x)
 374 
 375 #define BP_GET_LEVEL(bp)                BF64_GET((bp)->blk_prop, 56, 5)
 376 #define BP_SET_LEVEL(bp, x)             BF64_SET((bp)->blk_prop, 56, 5, x)
 377 
 378 #define BP_GET_DEDUP(bp)                BF64_GET((bp)->blk_prop, 62, 1)
 379 #define BP_SET_DEDUP(bp, x)             BF64_SET((bp)->blk_prop, 62, 1, x)
 380 
 381 #define BP_GET_BYTEORDER(bp)            BF64_GET((bp)->blk_prop, 63, 1)
 382 #define BP_SET_BYTEORDER(bp, x)         BF64_SET((bp)->blk_prop, 63, 1, x)
 383 
 384 #define BP_PHYSICAL_BIRTH(bp)           \
 385         (BP_IS_EMBEDDED(bp) ? 0 : \
 386         (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
 387 
 388 #define BP_SET_BIRTH(bp, logical, physical)     \
 389 {                                               \
 390         ASSERT(!BP_IS_EMBEDDED(bp));            \
 391         (bp)->blk_birth = (logical);         \
 392         (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
 393 }
 394 
 395 #define BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill)
 396 
 397 #define BP_GET_ASIZE(bp)        \
 398         (BP_IS_EMBEDDED(bp) ? 0 : \
 399         DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
 400         DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
 401         DVA_GET_ASIZE(&(bp)->blk_dva[2]))
 402 
 403 #define BP_GET_UCSIZE(bp) \
 404         ((BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) ? \
 405         BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
 406 
 407 #define BP_GET_NDVAS(bp)        \
 408         (BP_IS_EMBEDDED(bp) ? 0 : \
 409         !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
 410         !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
 411         !!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
 412 
 413 #define BP_COUNT_GANG(bp)       \
 414         (BP_IS_EMBEDDED(bp) ? 0 : \
 415         (DVA_GET_GANG(&(bp)->blk_dva[0]) + \
 416         DVA_GET_GANG(&(bp)->blk_dva[1]) + \
 417         DVA_GET_GANG(&(bp)->blk_dva[2])))
 418 
 419 #define DVA_EQUAL(dva1, dva2)   \
 420         ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
 421         (dva1)->dva_word[0] == (dva2)->dva_word[0])
 422 
 423 #define BP_EQUAL(bp1, bp2)      \
 424         (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) &&    \
 425         (bp1)->blk_birth == (bp2)->blk_birth &&                   \
 426         DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) &&      \
 427         DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) &&      \
 428         DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
 429 
 430 #define ZIO_CHECKSUM_EQUAL(zc1, zc2) \
 431         (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
 432         ((zc1).zc_word[1] - (zc2).zc_word[1]) | \
 433         ((zc1).zc_word[2] - (zc2).zc_word[2]) | \
 434         ((zc1).zc_word[3] - (zc2).zc_word[3])))
 435 
 436 #define DVA_IS_VALID(dva)       (DVA_GET_ASIZE(dva) != 0)
 437 
 438 #define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3)   \
 439 {                                               \
 440         (zcp)->zc_word[0] = w0;                      \
 441         (zcp)->zc_word[1] = w1;                      \
 442         (zcp)->zc_word[2] = w2;                      \
 443         (zcp)->zc_word[3] = w3;                      \
 444 }
 445 
 446 #define BP_IDENTITY(bp)         (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0])
 447 #define BP_IS_GANG(bp)          \
 448         (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp)))
 449 #define DVA_IS_EMPTY(dva)       ((dva)->dva_word[0] == 0ULL &&       \
 450                                 (dva)->dva_word[1] == 0ULL)
 451 #define BP_IS_HOLE(bp) \
 452         (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp)))
 453 
 454 /* BP_IS_RAIDZ(bp) assumes no block compression */
 455 #define BP_IS_RAIDZ(bp)         (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
 456                                 BP_GET_PSIZE(bp))
 457 
 458 #define BP_ZERO(bp)                             \
 459 {                                               \
 460         (bp)->blk_dva[0].dva_word[0] = 0;    \
 461         (bp)->blk_dva[0].dva_word[1] = 0;    \
 462         (bp)->blk_dva[1].dva_word[0] = 0;    \
 463         (bp)->blk_dva[1].dva_word[1] = 0;    \
 464         (bp)->blk_dva[2].dva_word[0] = 0;    \
 465         (bp)->blk_dva[2].dva_word[1] = 0;    \
 466         (bp)->blk_prop = 0;                  \
 467         (bp)->blk_pad[0] = 0;                        \
 468         (bp)->blk_pad[1] = 0;                        \
 469         (bp)->blk_phys_birth = 0;            \
 470         (bp)->blk_birth = 0;                 \
 471         (bp)->blk_fill = 0;                  \
 472         ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0);  \
 473 }
 474 
 475 #ifdef _BIG_ENDIAN
 476 #define ZFS_HOST_BYTEORDER      (0ULL)
 477 #else
 478 #define ZFS_HOST_BYTEORDER      (1ULL)
 479 #endif
 480 
 481 #define BP_SHOULD_BYTESWAP(bp)  (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
 482 
 483 #define BP_SPRINTF_LEN  320
 484 
 485 /*
 486  * This macro allows code sharing between zfs, libzpool, and mdb.
 487  * 'func' is either snprintf() or mdb_snprintf().
 488  * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
 489  */
 490 #define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
 491 {                                                                       \
 492         static const char *copyname[] =                                 \
 493             { "zero", "single", "double", "triple" };                   \
 494         int len = 0;                                                    \
 495         int copies = 0;                                                 \
 496                                                                         \
 497         if (bp == NULL) {                                               \
 498                 len += func(buf + len, size - len, "<NULL>");             \
 499         } else if (BP_IS_HOLE(bp)) {                                    \
 500                 len += func(buf + len, size - len, "<hole>");             \
 501                 if (bp->blk_birth > 0) {                          \
 502                         len += func(buf + len, size - len,              \
 503                             " birth=%lluL",                             \
 504                             (u_longlong_t)bp->blk_birth);            \
 505                 }                                                       \
 506         } else if (BP_IS_EMBEDDED(bp)) {                                \
 507                 len = func(buf + len, size - len,                       \
 508                     "EMBEDDED [L%llu %s] et=%u %s "                     \
 509                     "size=%llxL/%llxP birth=%lluL",                     \
 510                     (u_longlong_t)BP_GET_LEVEL(bp),                     \
 511                     type,                                               \
 512                     (int)BPE_GET_ETYPE(bp),                             \
 513                     compress,                                           \
 514                     (u_longlong_t)BPE_GET_LSIZE(bp),                    \
 515                     (u_longlong_t)BPE_GET_PSIZE(bp),                    \
 516                     (u_longlong_t)bp->blk_birth);                    \
 517         } else {                                                        \
 518                 for (int d = 0; d < BP_GET_NDVAS(bp); d++) {         \
 519                         const dva_t *dva = &bp->blk_dva[d];              \
 520                         if (DVA_IS_VALID(dva))                          \
 521                                 copies++;                               \
 522                         len += func(buf + len, size - len,              \
 523                             "DVA[%d]=<%llu:%llx:%llx>%c", d,              \
 524                             (u_longlong_t)DVA_GET_VDEV(dva),            \
 525                             (u_longlong_t)DVA_GET_OFFSET(dva),          \
 526                             (u_longlong_t)DVA_GET_ASIZE(dva),           \
 527                             ws);                                        \
 528                 }                                                       \
 529                 if (BP_IS_GANG(bp) &&                                   \
 530                     DVA_GET_ASIZE(&bp->blk_dva[2]) <=                 \
 531                     DVA_GET_ASIZE(&bp->blk_dva[1]) / 2)                  \
 532                         copies--;                                       \
 533                 len += func(buf + len, size - len,                      \
 534                     "[L%llu %s] %s %s %s %s %s %s%c"                    \
 535                     "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c"    \
 536                     "cksum=%llx:%llx:%llx:%llx",                        \
 537                     (u_longlong_t)BP_GET_LEVEL(bp),                     \
 538                     type,                                               \
 539                     checksum,                                           \
 540                     compress,                                           \
 541                     BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE",            \
 542                     BP_IS_GANG(bp) ? "gang" : "contiguous",             \
 543                     BP_GET_DEDUP(bp) ? "dedup" : "unique",              \
 544                     copyname[copies],                                   \
 545                     ws,                                                 \
 546                     (u_longlong_t)BP_GET_LSIZE(bp),                     \
 547                     (u_longlong_t)BP_GET_PSIZE(bp),                     \
 548                     (u_longlong_t)bp->blk_birth,                     \
 549                     (u_longlong_t)BP_PHYSICAL_BIRTH(bp),                \
 550                     (u_longlong_t)BP_GET_FILL(bp),                      \
 551                     ws,                                                 \
 552                     (u_longlong_t)bp->blk_cksum.zc_word[0],          \
 553                     (u_longlong_t)bp->blk_cksum.zc_word[1],          \
 554                     (u_longlong_t)bp->blk_cksum.zc_word[2],          \
 555                     (u_longlong_t)bp->blk_cksum.zc_word[3]);         \
 556         }                                                               \
 557         ASSERT(len < size);                                          \
 558 }
 559 
 560 #include <sys/dmu.h>
 561 
 562 #define BP_GET_BUFC_TYPE(bp)                                            \
 563         (((BP_GET_LEVEL(bp) > 0) || (DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))) ? \
 564         ARC_BUFC_METADATA : ARC_BUFC_DATA)
 565 
 566 typedef enum spa_import_type {
 567         SPA_IMPORT_EXISTING,
 568         SPA_IMPORT_ASSEMBLE
 569 } spa_import_type_t;
 570 
 571 /* state manipulation functions */
 572 extern int spa_open(const char *pool, spa_t **, void *tag);
 573 extern int spa_open_lock(const char *pool, spa_t **, void *tag, int lock);
 574 extern int spa_open_rewind(const char *pool, spa_t **, void *tag,
 575     nvlist_t *policy, nvlist_t **config);
 576 extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot,
 577     size_t buflen);
 578 extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props,
 579     nvlist_t *zplprops);
 580 extern int spa_import_rootpool(char *devpath, char *devid);
 581 extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props,
 582     uint64_t flags);
 583 extern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
 584 extern int spa_destroy(char *pool);
 585 extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force,
 586     boolean_t hardforce);
 587 extern int spa_reset(char *pool);
 588 extern void spa_async_request(spa_t *spa, int flag);
 589 extern void spa_async_unrequest(spa_t *spa, int flag);
 590 extern void spa_async_suspend(spa_t *spa);
 591 extern void spa_async_resume(spa_t *spa);
 592 extern spa_t *spa_inject_addref(char *pool);
 593 extern void spa_inject_delref(spa_t *spa);
 594 extern void spa_scan_stat_init(spa_t *spa);
 595 extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps);
 596 
 597 #define SPA_ASYNC_CONFIG_UPDATE 0x01
 598 #define SPA_ASYNC_REMOVE        0x02
 599 #define SPA_ASYNC_PROBE         0x04
 600 #define SPA_ASYNC_RESILVER_DONE 0x08
 601 #define SPA_ASYNC_RESILVER      0x10
 602 #define SPA_ASYNC_AUTOEXPAND    0x20
 603 #define SPA_ASYNC_REMOVE_DONE   0x40
 604 #define SPA_ASYNC_REMOVE_STOP   0x80
 605 
 606 /*
 607  * Controls the behavior of spa_vdev_remove().
 608  */
 609 #define SPA_REMOVE_UNSPARE      0x01
 610 #define SPA_REMOVE_DONE         0x02
 611 
 612 /* device manipulation */
 613 extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
 614 extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot,
 615     int replacing);
 616 extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
 617     int replace_done);
 618 extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare);
 619 extern boolean_t spa_vdev_remove_active(spa_t *spa);
 620 extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
 621 extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru);
 622 extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
 623     nvlist_t *props, boolean_t exp);
 624 
 625 /* spare state (which is global across all pools) */
 626 extern void spa_spare_add(vdev_t *vd);
 627 extern void spa_spare_remove(vdev_t *vd);
 628 extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt);
 629 extern void spa_spare_activate(vdev_t *vd);
 630 
 631 /* L2ARC state (which is global across all pools) */
 632 extern void spa_l2cache_add(vdev_t *vd);
 633 extern void spa_l2cache_remove(vdev_t *vd);
 634 extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
 635 extern void spa_l2cache_activate(vdev_t *vd);
 636 extern void spa_l2cache_drop(spa_t *spa);
 637 
 638 /* scanning */
 639 extern int spa_scan(spa_t *spa, pool_scan_func_t func);
 640 extern int spa_scan_stop(spa_t *spa);
 641 
 642 /* spa syncing */
 643 extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
 644 extern void spa_sync_allpools(void);
 645 
 646 /* spa namespace global mutex */
 647 extern kmutex_t spa_namespace_lock;
 648 
 649 /*
 650  * SPA configuration functions in spa_config.c
 651  */
 652 
 653 #define SPA_CONFIG_UPDATE_POOL  0
 654 #define SPA_CONFIG_UPDATE_VDEVS 1
 655 
 656 extern void spa_config_sync(spa_t *, boolean_t, boolean_t);
 657 extern void spa_config_load(void);
 658 extern nvlist_t *spa_all_configs(uint64_t *);
 659 extern void spa_config_set(spa_t *spa, nvlist_t *config);
 660 extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
 661     int getstats);
 662 extern void spa_config_update(spa_t *spa, int what);
 663 
 664 /*
 665  * Miscellaneous SPA routines in spa_misc.c
 666  */
 667 
 668 /* Namespace manipulation */
 669 extern spa_t *spa_lookup(const char *name);
 670 extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot);
 671 extern void spa_remove(spa_t *spa);
 672 extern spa_t *spa_next(spa_t *prev);
 673 
 674 /* Refcount functions */
 675 extern void spa_open_ref(spa_t *spa, void *tag);
 676 extern void spa_close(spa_t *spa, void *tag);
 677 extern boolean_t spa_refcount_zero(spa_t *spa);
 678 
 679 #define SCL_NONE        0x00
 680 #define SCL_CONFIG      0x01
 681 #define SCL_STATE       0x02
 682 #define SCL_L2ARC       0x04            /* hack until L2ARC 2.0 */
 683 #define SCL_ALLOC       0x08
 684 #define SCL_ZIO         0x10
 685 #define SCL_FREE        0x20
 686 #define SCL_VDEV        0x40
 687 #define SCL_LOCKS       7
 688 #define SCL_ALL         ((1 << SCL_LOCKS) - 1)
 689 #define SCL_STATE_ALL   (SCL_STATE | SCL_L2ARC | SCL_ZIO)
 690 
 691 /* Pool configuration locks */
 692 extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw);
 693 extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw);
 694 extern void spa_config_exit(spa_t *spa, int locks, void *tag);
 695 extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
 696 
 697 /* Pool vdev add/remove lock */
 698 extern uint64_t spa_vdev_enter(spa_t *spa);
 699 extern uint64_t spa_vdev_config_enter(spa_t *spa);
 700 extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg,
 701     int error, char *tag);
 702 extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error);
 703 
 704 /* Pool vdev state change lock */
 705 extern void spa_vdev_state_enter(spa_t *spa, int oplock);
 706 extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error);
 707 
 708 /* Log state */
 709 typedef enum spa_log_state {
 710         SPA_LOG_UNKNOWN = 0,    /* unknown log state */
 711         SPA_LOG_MISSING,        /* missing log(s) */
 712         SPA_LOG_CLEAR,          /* clear the log(s) */
 713         SPA_LOG_GOOD,           /* log(s) are good */
 714 } spa_log_state_t;
 715 
 716 extern spa_log_state_t spa_get_log_state(spa_t *spa);
 717 extern void spa_set_log_state(spa_t *spa, spa_log_state_t state);
 718 extern int spa_offline_log(spa_t *spa);
 719 
 720 /* Log claim callback */
 721 extern void spa_claim_notify(zio_t *zio);
 722 
 723 /* Accessor functions */
 724 extern boolean_t spa_shutting_down(spa_t *spa);
 725 extern struct dsl_pool *spa_get_dsl(spa_t *spa);
 726 extern boolean_t spa_is_initializing(spa_t *spa);
 727 extern blkptr_t *spa_get_rootblkptr(spa_t *spa);
 728 extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
 729 extern void spa_altroot(spa_t *, char *, size_t);
 730 extern int spa_sync_pass(spa_t *spa);
 731 extern char *spa_name(spa_t *spa);
 732 extern uint64_t spa_guid(spa_t *spa);
 733 extern uint64_t spa_load_guid(spa_t *spa);
 734 extern uint64_t spa_last_synced_txg(spa_t *spa);
 735 extern uint64_t spa_first_txg(spa_t *spa);
 736 extern uint64_t spa_syncing_txg(spa_t *spa);
 737 extern uint64_t spa_version(spa_t *spa);
 738 extern pool_state_t spa_state(spa_t *spa);
 739 extern spa_load_state_t spa_load_state(spa_t *spa);
 740 extern uint64_t spa_freeze_txg(spa_t *spa);
 741 extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize);
 742 extern uint64_t spa_get_dspace(spa_t *spa);
 743 extern void spa_update_dspace(spa_t *spa);
 744 extern uint64_t spa_version(spa_t *spa);
 745 extern boolean_t spa_deflate(spa_t *spa);
 746 extern metaslab_class_t *spa_normal_class(spa_t *spa);
 747 extern metaslab_class_t *spa_log_class(spa_t *spa);
 748 extern int spa_max_replication(spa_t *spa);
 749 extern int spa_prev_software_version(spa_t *spa);
 750 extern int spa_busy(void);
 751 extern uint8_t spa_get_failmode(spa_t *spa);
 752 extern boolean_t spa_suspended(spa_t *spa);
 753 extern uint64_t spa_bootfs(spa_t *spa);
 754 extern uint64_t spa_delegation(spa_t *spa);
 755 extern objset_t *spa_meta_objset(spa_t *spa);
 756 extern uint64_t spa_deadman_synctime(spa_t *spa);
 757 
 758 /* Miscellaneous support routines */
 759 extern void spa_activate_mos_feature(spa_t *spa, const char *feature,
 760     dmu_tx_t *tx);
 761 extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature);
 762 extern int spa_rename(const char *oldname, const char *newname);
 763 extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid);
 764 extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid);
 765 extern char *spa_strdup(const char *);
 766 extern void spa_strfree(char *);
 767 extern uint64_t spa_get_random(uint64_t range);
 768 extern uint64_t spa_generate_guid(spa_t *spa);
 769 extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp);
 770 extern void spa_freeze(spa_t *spa);
 771 extern int spa_change_guid(spa_t *spa);
 772 extern void spa_upgrade(spa_t *spa, uint64_t version);
 773 extern void spa_evict_all(void);
 774 extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid,
 775     boolean_t l2cache);
 776 extern boolean_t spa_has_spare(spa_t *, uint64_t guid);
 777 extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva);
 778 extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp);
 779 extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp);
 780 extern boolean_t spa_has_slogs(spa_t *spa);
 781 extern boolean_t spa_is_root(spa_t *spa);
 782 extern boolean_t spa_writeable(spa_t *spa);
 783 
 784 extern int spa_mode(spa_t *spa);
 785 extern uint64_t strtonum(const char *str, char **nptr);
 786 
 787 extern char *spa_his_ievent_table[];
 788 
 789 extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx);
 790 extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read,
 791     char *his_buf);
 792 extern int spa_history_log(spa_t *spa, const char *his_buf);
 793 extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl);
 794 extern void spa_history_log_version(spa_t *spa, const char *operation);
 795 extern void spa_history_log_internal(spa_t *spa, const char *operation,
 796     dmu_tx_t *tx, const char *fmt, ...);
 797 extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op,
 798     dmu_tx_t *tx, const char *fmt, ...);
 799 extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
 800     dmu_tx_t *tx, const char *fmt, ...);
 801 
 802 /* error handling */
 803 struct zbookmark;
 804 extern void spa_log_error(spa_t *spa, zio_t *zio);
 805 extern void zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd,
 806     zio_t *zio, uint64_t stateoroffset, uint64_t length);
 807 extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
 808 extern void zfs_post_state_change(spa_t *spa, vdev_t *vd);
 809 extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
 810 extern uint64_t spa_get_errlog_size(spa_t *spa);
 811 extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
 812 extern void spa_errlog_rotate(spa_t *spa);
 813 extern void spa_errlog_drain(spa_t *spa);
 814 extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
 815 extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
 816 
 817 /* vdev cache */
 818 extern void vdev_cache_stat_init(void);
 819 extern void vdev_cache_stat_fini(void);
 820 
 821 /* Initialization and termination */
 822 extern void spa_init(int flags);
 823 extern void spa_fini(void);
 824 extern void spa_boot_init();
 825 
 826 /* properties */
 827 extern int spa_prop_set(spa_t *spa, nvlist_t *nvp);
 828 extern int spa_prop_get(spa_t *spa, nvlist_t **nvp);
 829 extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx);
 830 extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t);
 831 
 832 /* asynchronous event notification */
 833 extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name);
 834 
 835 #ifdef ZFS_DEBUG
 836 #define dprintf_bp(bp, fmt, ...) do {                           \
 837         if (zfs_flags & ZFS_DEBUG_DPRINTF) {                        \
 838         char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP);  \
 839         snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp));        \
 840         dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf);            \
 841         kmem_free(__blkbuf, BP_SPRINTF_LEN);                    \
 842         } \
 843 _NOTE(CONSTCOND) } while (0)
 844 #else
 845 #define dprintf_bp(bp, fmt, ...)
 846 #endif
 847 
 848 extern boolean_t spa_debug_enabled(spa_t *spa);
 849 #define spa_dbgmsg(spa, ...)                    \
 850 {                                               \
 851         if (spa_debug_enabled(spa))             \
 852                 zfs_dbgmsg(__VA_ARGS__);        \
 853 }
 854 
 855 extern int spa_mode_global;                     /* mode, e.g. FREAD | FWRITE */
 856 
 857 #ifdef  __cplusplus
 858 }
 859 #endif
 860 
 861 #endif  /* _SYS_SPA_H */