2 * Copyright (c) 2002 McAfee, Inc.
5 * This software was developed for the FreeBSD Project by Marshall
6 * Kirk McKusick and McAfee Research,, the Security Research Division of
7 * McAfee, Inc. under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as
8 * part of the DARPA CHATS research program
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * The contents of this file are subject to the terms of the
35 * Common Development and Distribution License (the "License").
36 * You may not use this file except in compliance with the License.
38 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
39 * or http://www.opensolaris.org/os/licensing.
40 * See the License for the specific language governing permissions
41 * and limitations under the License.
43 * When distributing Covered Code, include this CDDL HEADER in each
44 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
45 * If applicable, add the following below this CDDL HEADER, with the
46 * fields enclosed by brackets "[]" replaced with your own identifying
47 * information: Portions Copyright [yyyy] [name of copyright owner]
52 * Copyright 2009 Sun Microsystems, Inc. All rights reserved.
53 * Use is subject to license terms.
56 * Copyright 2013 by Saso Kiselkov. All rights reserved.
59 * Copyright (c) 2020 by Delphix. All rights reserved.
62 #include <sys/queue.h>
67 #define MAXNAMELEN 256
72 * AVL comparator helpers
74 #define AVL_ISIGN(a) (((a) > 0) - ((a) < 0))
75 #define AVL_CMP(a, b) (((a) > (b)) - ((a) < (b)))
76 #define AVL_PCMP(a, b) \
77 (((uintptr_t)(a) > (uintptr_t)(b)) - ((uintptr_t)(a) < (uintptr_t)(b)))
79 typedef enum { B_FALSE, B_TRUE } boolean_t;
82 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */
85 * Macros for various sorts of alignment and rounding when the alignment
86 * is known to be a power of 2.
88 #define P2ALIGN(x, align) ((x) & -(align))
89 #define P2PHASE(x, align) ((x) & ((align) - 1))
90 #define P2NPHASE(x, align) (-(x) & ((align) - 1))
91 #define P2ROUNDUP(x, align) (-(-(x) & -(align)))
92 #define P2END(x, align) (-(~(x) & -(align)))
93 #define P2PHASEUP(x, align, phase) ((phase) - (((phase) - (x)) & -(align)))
94 #define P2BOUNDARY(off, len, align) (((off) ^ ((off) + (len) - 1)) > (align) - 1)
97 * General-purpose 32-bit and 64-bit bitfield encodings.
99 #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len))
100 #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len))
101 #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low))
102 #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low))
104 #define BF32_GET(x, low, len) BF32_DECODE(x, low, len)
105 #define BF64_GET(x, low, len) BF64_DECODE(x, low, len)
107 #define BF32_SET(x, low, len, val) \
108 ((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len))
109 #define BF64_SET(x, low, len, val) \
110 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len))
112 #define BF32_GET_SB(x, low, len, shift, bias) \
113 ((BF32_GET(x, low, len) + (bias)) << (shift))
114 #define BF64_GET_SB(x, low, len, shift, bias) \
115 ((BF64_GET(x, low, len) + (bias)) << (shift))
117 #define BF32_SET_SB(x, low, len, shift, bias, val) \
118 BF32_SET(x, low, len, ((val) >> (shift)) - (bias))
119 #define BF64_SET_SB(x, low, len, shift, bias, val) \
120 BF64_SET(x, low, len, ((val) >> (shift)) - (bias))
123 * Macros to reverse byte order
125 #define BSWAP_8(x) ((x) & 0xff)
126 #define BSWAP_16(x) ((BSWAP_8(x) << 8) | BSWAP_8((x) >> 8))
127 #define BSWAP_32(x) ((BSWAP_16(x) << 16) | BSWAP_16((x) >> 16))
128 #define BSWAP_64(x) ((BSWAP_32(x) << 32) | BSWAP_32((x) >> 32))
130 #define SPA_MINBLOCKSHIFT 9
131 #define SPA_OLDMAXBLOCKSHIFT 17
132 #define SPA_MAXBLOCKSHIFT 24
133 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT)
134 #define SPA_OLDMAXBLOCKSIZE (1ULL << SPA_OLDMAXBLOCKSHIFT)
135 #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT)
138 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
139 * The ASIZE encoding should be at least 64 times larger (6 more bits)
140 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
141 * overhead, three DVAs per bp, plus one more bit in case we do anything
142 * else that expands the ASIZE.
144 #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */
145 #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */
146 #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */
149 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
150 * The members of the dva_t should be considered opaque outside the SPA.
153 uint64_t dva_word[2];
157 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
159 typedef struct zio_cksum {
164 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept
165 * secret and is suitable for use in MAC algorithms as the key.
167 typedef struct zio_cksum_salt {
168 uint8_t zcs_bytes[32];
172 * Each block is described by its DVAs, time of birth, checksum, etc.
173 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
175 * 64 56 48 40 32 24 16 8 0
176 * +-------+-------+-------+-------+-------+-------+-------+-------+
177 * 0 | vdev1 | GRID | ASIZE |
178 * +-------+-------+-------+-------+-------+-------+-------+-------+
180 * +-------+-------+-------+-------+-------+-------+-------+-------+
181 * 2 | vdev2 | GRID | ASIZE |
182 * +-------+-------+-------+-------+-------+-------+-------+-------+
184 * +-------+-------+-------+-------+-------+-------+-------+-------+
185 * 4 | vdev3 | GRID | ASIZE |
186 * +-------+-------+-------+-------+-------+-------+-------+-------+
188 * +-------+-------+-------+-------+-------+-------+-------+-------+
189 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE |
190 * +-------+-------+-------+-------+-------+-------+-------+-------+
192 * +-------+-------+-------+-------+-------+-------+-------+-------+
194 * +-------+-------+-------+-------+-------+-------+-------+-------+
195 * 9 | physical birth txg |
196 * +-------+-------+-------+-------+-------+-------+-------+-------+
197 * a | logical birth txg |
198 * +-------+-------+-------+-------+-------+-------+-------+-------+
200 * +-------+-------+-------+-------+-------+-------+-------+-------+
202 * +-------+-------+-------+-------+-------+-------+-------+-------+
204 * +-------+-------+-------+-------+-------+-------+-------+-------+
206 * +-------+-------+-------+-------+-------+-------+-------+-------+
208 * +-------+-------+-------+-------+-------+-------+-------+-------+
212 * vdev virtual device ID
213 * offset offset into virtual device
215 * PSIZE physical size (after compression)
216 * ASIZE allocated size (including RAID-Z parity and gang block headers)
217 * GRID RAID-Z layout information (reserved for future use)
218 * cksum checksum function
219 * comp compression function
220 * G gang block indicator
221 * B byteorder (endianness)
223 * X encryption (on version 30, which is not supported)
224 * E blkptr_t contains embedded data (see below)
225 * lvl level of indirection
226 * type DMU object type
227 * phys birth txg of block allocation; zero if same as logical birth txg
228 * log. birth transaction group in which the block was logically born
229 * fill count number of non-zero blocks under this bp
230 * checksum[4] 256-bit checksum of the data this bp describes
234 * "Embedded" blkptr_t's don't actually point to a block, instead they
235 * have a data payload embedded in the blkptr_t itself. See the comment
236 * in blkptr.c for more details.
238 * The blkptr_t is laid out as follows:
240 * 64 56 48 40 32 24 16 8 0
241 * +-------+-------+-------+-------+-------+-------+-------+-------+
248 * +-------+-------+-------+-------+-------+-------+-------+-------+
249 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE |
250 * +-------+-------+-------+-------+-------+-------+-------+-------+
254 * +-------+-------+-------+-------+-------+-------+-------+-------+
255 * a | logical birth txg |
256 * +-------+-------+-------+-------+-------+-------+-------+-------+
262 * +-------+-------+-------+-------+-------+-------+-------+-------+
266 * payload contains the embedded data
267 * B (byteorder) byteorder (endianness)
268 * D (dedup) padding (set to zero)
269 * X encryption (set to zero; see above)
270 * E (embedded) set to one
271 * lvl indirection level
272 * type DMU object type
273 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*)
274 * comp compression function of payload
275 * PSIZE size of payload after compression, in bytes
276 * LSIZE logical size of payload, in bytes
277 * note that 25 bits is enough to store the largest
278 * "normal" BP's LSIZE (2^16 * 2^9) in bytes
279 * log. birth transaction group in which the block was logically born
281 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
282 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
283 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
284 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
285 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must
286 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before
287 * other macros, as they assert that they are only used on BP's of the correct
291 #define BPE_GET_ETYPE(bp) \
292 (ASSERT(BP_IS_EMBEDDED(bp)), \
293 BF64_GET((bp)->blk_prop, 40, 8))
294 #define BPE_SET_ETYPE(bp, t) do { \
295 ASSERT(BP_IS_EMBEDDED(bp)); \
296 BF64_SET((bp)->blk_prop, 40, 8, t); \
297 _NOTE(CONSTCOND) } while (0)
299 #define BPE_GET_LSIZE(bp) \
300 (ASSERT(BP_IS_EMBEDDED(bp)), \
301 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
302 #define BPE_SET_LSIZE(bp, x) do { \
303 ASSERT(BP_IS_EMBEDDED(bp)); \
304 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
305 _NOTE(CONSTCOND) } while (0)
307 #define BPE_GET_PSIZE(bp) \
308 (ASSERT(BP_IS_EMBEDDED(bp)), \
309 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
310 #define BPE_SET_PSIZE(bp, x) do { \
311 ASSERT(BP_IS_EMBEDDED(bp)); \
312 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
313 _NOTE(CONSTCOND) } while (0)
315 typedef enum bp_embedded_type {
316 BP_EMBEDDED_TYPE_DATA,
317 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */
318 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED
319 } bp_embedded_type_t;
321 #define BPE_NUM_WORDS 14
322 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
323 #define BPE_IS_PAYLOADWORD(bp, wp) \
324 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
326 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */
327 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */
329 typedef struct blkptr {
330 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
331 uint64_t blk_prop; /* size, compression, type, etc */
332 uint64_t blk_pad[2]; /* Extra space for the future */
333 uint64_t blk_phys_birth; /* txg when block was allocated */
334 uint64_t blk_birth; /* transaction group at birth */
335 uint64_t blk_fill; /* fill count */
336 zio_cksum_t blk_cksum; /* 256-bit checksum */
340 * Macros to get and set fields in a bp or DVA.
342 #define DVA_GET_ASIZE(dva) \
343 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
344 #define DVA_SET_ASIZE(dva, x) \
345 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
346 SPA_MINBLOCKSHIFT, 0, x)
348 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8)
349 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x)
351 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32)
352 #define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x)
354 #define DVA_GET_OFFSET(dva) \
355 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
356 #define DVA_SET_OFFSET(dva, x) \
357 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
359 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1)
360 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x)
362 #define BP_GET_LSIZE(bp) \
363 (BP_IS_EMBEDDED(bp) ? \
364 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
365 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
366 #define BP_SET_LSIZE(bp, x) do { \
367 ASSERT(!BP_IS_EMBEDDED(bp)); \
368 BF64_SET_SB((bp)->blk_prop, \
369 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
370 _NOTE(CONSTCOND) } while (0)
372 #define BP_GET_PSIZE(bp) \
373 BF64_GET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)
374 #define BP_SET_PSIZE(bp, x) \
375 BF64_SET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x)
377 #define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7)
378 #define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x)
380 #define BP_GET_CHECKSUM(bp) BF64_GET((bp)->blk_prop, 40, 8)
381 #define BP_SET_CHECKSUM(bp, x) BF64_SET((bp)->blk_prop, 40, 8, x)
383 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8)
384 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x)
386 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5)
387 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x)
389 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1)
391 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1)
392 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x)
394 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1)
395 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x)
397 #define BP_PHYSICAL_BIRTH(bp) \
398 ((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
400 #define BP_SET_BIRTH(bp, logical, physical) \
402 ASSERT(!BP_IS_EMBEDDED(bp)); \
403 (bp)->blk_birth = (logical); \
404 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
407 #define BP_GET_FILL(bp) \
408 ((BP_IS_EMBEDDED(bp)) ? 1 : (bp)->blk_fill)
410 #define BP_SET_FILL(bp, fill) \
412 (bp)->blk_fill = fill; \
415 #define BP_GET_ASIZE(bp) \
416 (DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
417 DVA_GET_ASIZE(&(bp)->blk_dva[2]))
419 #define BP_GET_UCSIZE(bp) \
420 ((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \
421 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp));
423 #define BP_GET_NDVAS(bp) \
424 (!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
425 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
426 !!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
428 #define DVA_EQUAL(dva1, dva2) \
429 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
430 (dva1)->dva_word[0] == (dva2)->dva_word[0])
432 #define ZIO_CHECKSUM_EQUAL(zc1, zc2) \
433 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
434 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \
435 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \
436 ((zc1).zc_word[3] - (zc2).zc_word[3])))
439 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0)
441 #define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \
443 (zcp)->zc_word[0] = w0; \
444 (zcp)->zc_word[1] = w1; \
445 (zcp)->zc_word[2] = w2; \
446 (zcp)->zc_word[3] = w3; \
449 #define BP_IDENTITY(bp) (&(bp)->blk_dva[0])
450 #define BP_IS_GANG(bp) DVA_GET_GANG(BP_IDENTITY(bp))
451 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \
452 (dva)->dva_word[1] == 0ULL)
453 #define BP_IS_HOLE(bp) DVA_IS_EMPTY(BP_IDENTITY(bp))
454 #define BP_IS_OLDER(bp, txg) (!BP_IS_HOLE(bp) && (bp)->blk_birth < (txg))
456 #define BP_ZERO(bp) \
458 (bp)->blk_dva[0].dva_word[0] = 0; \
459 (bp)->blk_dva[0].dva_word[1] = 0; \
460 (bp)->blk_dva[1].dva_word[0] = 0; \
461 (bp)->blk_dva[1].dva_word[1] = 0; \
462 (bp)->blk_dva[2].dva_word[0] = 0; \
463 (bp)->blk_dva[2].dva_word[1] = 0; \
464 (bp)->blk_prop = 0; \
465 (bp)->blk_pad[0] = 0; \
466 (bp)->blk_pad[1] = 0; \
467 (bp)->blk_phys_birth = 0; \
468 (bp)->blk_birth = 0; \
469 (bp)->blk_fill = 0; \
470 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \
473 #if BYTE_ORDER == _BIG_ENDIAN
474 #define ZFS_HOST_BYTEORDER (0ULL)
476 #define ZFS_HOST_BYTEORDER (1ULL)
479 #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
480 #define BPE_NUM_WORDS 14
481 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
482 #define BPE_IS_PAYLOADWORD(bp, wp) \
483 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
488 #define ZEC_MAGIC 0x210da7ab10c7a11ULL
490 typedef struct zio_eck {
491 uint64_t zec_magic; /* for validation, endianness */
492 zio_cksum_t zec_cksum; /* 256-bit checksum */
496 * Gang block headers are self-checksumming and contain an array
499 #define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE
500 #define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \
501 sizeof (zio_eck_t)) / sizeof (blkptr_t))
502 #define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \
503 sizeof (zio_eck_t) - \
504 (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\
507 typedef struct zio_gbh {
508 blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS];
509 uint64_t zg_filler[SPA_GBH_FILLER];
513 #define VDEV_RAIDZ_MAXPARITY 3
515 #define VDEV_PAD_SIZE (8 << 10)
516 /* 2 padding areas (vl_pad1 and vl_be) to skip */
517 #define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2
518 #define VDEV_PHYS_SIZE (112 << 10)
519 #define VDEV_UBERBLOCK_RING (128 << 10)
522 * MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock
523 * ring when MMP is enabled.
525 #define MMP_BLOCKS_PER_LABEL 1
527 /* The largest uberblock we support is 8k. */
528 #define MAX_UBERBLOCK_SHIFT (13)
529 #define VDEV_UBERBLOCK_SHIFT(vd) \
530 MIN(MAX((vd)->v_top->v_ashift, UBERBLOCK_SHIFT), MAX_UBERBLOCK_SHIFT)
531 #define VDEV_UBERBLOCK_COUNT(vd) \
532 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
533 #define VDEV_UBERBLOCK_OFFSET(vd, n) \
534 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
535 #define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd))
537 typedef struct vdev_phys {
538 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)];
542 typedef enum vbe_vers {
543 /* The bootenv file is stored as ascii text in the envblock */
547 * The bootenv file is converted to an nvlist and then packed into the
553 typedef struct vdev_boot_envblock {
554 uint64_t vbe_version;
555 char vbe_bootenv[VDEV_PAD_SIZE - sizeof (uint64_t) -
558 } vdev_boot_envblock_t;
560 _Static_assert(sizeof (vdev_boot_envblock_t) == VDEV_PAD_SIZE,
561 "bad size for vdev_boot_envblock_t");
563 typedef struct vdev_label {
564 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */
565 vdev_boot_envblock_t vl_be; /* 8K */
566 vdev_phys_t vl_vdev_phys; /* 112K */
567 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */
568 } vdev_label_t; /* 256K total */
573 #define VDD_METASLAB 0x01
577 * Size and offset of embedded boot loader region on each label.
578 * The total size of the first two labels plus the boot area is 4MB.
580 #define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t))
581 #define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */
584 * Size of label regions at the start and end of each leaf device.
586 #define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
587 #define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t))
588 #define VDEV_LABELS 4
591 ZIO_CHECKSUM_INHERIT = 0,
595 ZIO_CHECKSUM_GANG_HEADER,
597 ZIO_CHECKSUM_FLETCHER_2,
598 ZIO_CHECKSUM_FLETCHER_4,
601 ZIO_CHECKSUM_NOPARITY,
605 ZIO_CHECKSUM_FUNCTIONS
608 #define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4
609 #define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON
612 ZIO_COMPRESS_INHERIT = 0,
629 ZIO_COMPRESS_FUNCTIONS
632 enum zio_zstd_levels {
633 ZIO_ZSTD_LEVEL_INHERIT = 0,
635 #define ZIO_ZSTD_LEVEL_MIN ZIO_ZSTD_LEVEL_1
638 #define ZIO_ZSTD_LEVEL_DEFAULT ZIO_ZSTD_LEVEL_3
655 #define ZIO_ZSTD_LEVEL_MAX ZIO_ZSTD_LEVEL_19
656 ZIO_ZSTD_LEVEL_RESERVE = 101, /* Leave room for new positive levels */
657 ZIO_ZSTD_LEVEL_FAST, /* Fast levels are negative */
658 ZIO_ZSTD_LEVEL_FAST_1,
659 #define ZIO_ZSTD_LEVEL_FAST_DEFAULT ZIO_ZSTD_LEVEL_FAST_1
660 ZIO_ZSTD_LEVEL_FAST_2,
661 ZIO_ZSTD_LEVEL_FAST_3,
662 ZIO_ZSTD_LEVEL_FAST_4,
663 ZIO_ZSTD_LEVEL_FAST_5,
664 ZIO_ZSTD_LEVEL_FAST_6,
665 ZIO_ZSTD_LEVEL_FAST_7,
666 ZIO_ZSTD_LEVEL_FAST_8,
667 ZIO_ZSTD_LEVEL_FAST_9,
668 ZIO_ZSTD_LEVEL_FAST_10,
669 ZIO_ZSTD_LEVEL_FAST_20,
670 ZIO_ZSTD_LEVEL_FAST_30,
671 ZIO_ZSTD_LEVEL_FAST_40,
672 ZIO_ZSTD_LEVEL_FAST_50,
673 ZIO_ZSTD_LEVEL_FAST_60,
674 ZIO_ZSTD_LEVEL_FAST_70,
675 ZIO_ZSTD_LEVEL_FAST_80,
676 ZIO_ZSTD_LEVEL_FAST_90,
677 ZIO_ZSTD_LEVEL_FAST_100,
678 ZIO_ZSTD_LEVEL_FAST_500,
679 ZIO_ZSTD_LEVEL_FAST_1000,
680 #define ZIO_ZSTD_LEVEL_FAST_MAX ZIO_ZSTD_LEVEL_FAST_1000
681 ZIO_ZSTD_LEVEL_AUTO = 251, /* Reserved for future use */
682 ZIO_ZSTD_LEVEL_LEVELS
685 #define ZIO_COMPRESS_ON_VALUE ZIO_COMPRESS_LZJB
686 #define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF
689 * On-disk version number.
691 #define SPA_VERSION_1 1ULL
692 #define SPA_VERSION_2 2ULL
693 #define SPA_VERSION_3 3ULL
694 #define SPA_VERSION_4 4ULL
695 #define SPA_VERSION_5 5ULL
696 #define SPA_VERSION_6 6ULL
697 #define SPA_VERSION_7 7ULL
698 #define SPA_VERSION_8 8ULL
699 #define SPA_VERSION_9 9ULL
700 #define SPA_VERSION_10 10ULL
701 #define SPA_VERSION_11 11ULL
702 #define SPA_VERSION_12 12ULL
703 #define SPA_VERSION_13 13ULL
704 #define SPA_VERSION_14 14ULL
705 #define SPA_VERSION_15 15ULL
706 #define SPA_VERSION_16 16ULL
707 #define SPA_VERSION_17 17ULL
708 #define SPA_VERSION_18 18ULL
709 #define SPA_VERSION_19 19ULL
710 #define SPA_VERSION_20 20ULL
711 #define SPA_VERSION_21 21ULL
712 #define SPA_VERSION_22 22ULL
713 #define SPA_VERSION_23 23ULL
714 #define SPA_VERSION_24 24ULL
715 #define SPA_VERSION_25 25ULL
716 #define SPA_VERSION_26 26ULL
717 #define SPA_VERSION_27 27ULL
718 #define SPA_VERSION_28 28ULL
719 #define SPA_VERSION_5000 5000ULL
722 * When bumping up SPA_VERSION, make sure GRUB ZFS understands the on-disk
723 * format change. Go to usr/src/grub/grub-0.97/stage2/{zfs-include/, fsys_zfs*},
724 * and do the appropriate changes. Also bump the version number in
725 * usr/src/grub/capability.
727 #define SPA_VERSION SPA_VERSION_5000
728 #define SPA_VERSION_STRING "5000"
731 * Symbolic names for the changes that caused a SPA_VERSION switch.
732 * Used in the code when checking for presence or absence of a feature.
733 * Feel free to define multiple symbolic names for each version if there
734 * were multiple changes to on-disk structures during that version.
736 * NOTE: When checking the current SPA_VERSION in your code, be sure
737 * to use spa_version() since it reports the version of the
738 * last synced uberblock. Checking the in-flight version can
739 * be dangerous in some cases.
741 #define SPA_VERSION_INITIAL SPA_VERSION_1
742 #define SPA_VERSION_DITTO_BLOCKS SPA_VERSION_2
743 #define SPA_VERSION_SPARES SPA_VERSION_3
744 #define SPA_VERSION_RAID6 SPA_VERSION_3
745 #define SPA_VERSION_BPLIST_ACCOUNT SPA_VERSION_3
746 #define SPA_VERSION_RAIDZ_DEFLATE SPA_VERSION_3
747 #define SPA_VERSION_DNODE_BYTES SPA_VERSION_3
748 #define SPA_VERSION_ZPOOL_HISTORY SPA_VERSION_4
749 #define SPA_VERSION_GZIP_COMPRESSION SPA_VERSION_5
750 #define SPA_VERSION_BOOTFS SPA_VERSION_6
751 #define SPA_VERSION_SLOGS SPA_VERSION_7
752 #define SPA_VERSION_DELEGATED_PERMS SPA_VERSION_8
753 #define SPA_VERSION_FUID SPA_VERSION_9
754 #define SPA_VERSION_REFRESERVATION SPA_VERSION_9
755 #define SPA_VERSION_REFQUOTA SPA_VERSION_9
756 #define SPA_VERSION_UNIQUE_ACCURATE SPA_VERSION_9
757 #define SPA_VERSION_L2CACHE SPA_VERSION_10
758 #define SPA_VERSION_NEXT_CLONES SPA_VERSION_11
759 #define SPA_VERSION_ORIGIN SPA_VERSION_11
760 #define SPA_VERSION_DSL_SCRUB SPA_VERSION_11
761 #define SPA_VERSION_SNAP_PROPS SPA_VERSION_12
762 #define SPA_VERSION_USED_BREAKDOWN SPA_VERSION_13
763 #define SPA_VERSION_PASSTHROUGH_X SPA_VERSION_14
764 #define SPA_VERSION_USERSPACE SPA_VERSION_15
765 #define SPA_VERSION_STMF_PROP SPA_VERSION_16
766 #define SPA_VERSION_RAIDZ3 SPA_VERSION_17
767 #define SPA_VERSION_USERREFS SPA_VERSION_18
768 #define SPA_VERSION_HOLES SPA_VERSION_19
769 #define SPA_VERSION_ZLE_COMPRESSION SPA_VERSION_20
770 #define SPA_VERSION_DEDUP SPA_VERSION_21
771 #define SPA_VERSION_RECVD_PROPS SPA_VERSION_22
772 #define SPA_VERSION_SLIM_ZIL SPA_VERSION_23
773 #define SPA_VERSION_SA SPA_VERSION_24
774 #define SPA_VERSION_SCAN SPA_VERSION_25
775 #define SPA_VERSION_DIR_CLONES SPA_VERSION_26
776 #define SPA_VERSION_DEADLISTS SPA_VERSION_26
777 #define SPA_VERSION_FAST_SNAP SPA_VERSION_27
778 #define SPA_VERSION_MULTI_REPLACE SPA_VERSION_28
779 #define SPA_VERSION_BEFORE_FEATURES SPA_VERSION_28
780 #define SPA_VERSION_FEATURES SPA_VERSION_5000
782 #define SPA_VERSION_IS_SUPPORTED(v) \
783 (((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \
784 ((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION))
787 * The following are configuration names used in the nvlist describing a pool's
790 #define ZPOOL_CONFIG_VERSION "version"
791 #define ZPOOL_CONFIG_POOL_NAME "name"
792 #define ZPOOL_CONFIG_POOL_STATE "state"
793 #define ZPOOL_CONFIG_POOL_TXG "txg"
794 #define ZPOOL_CONFIG_POOL_GUID "pool_guid"
795 #define ZPOOL_CONFIG_CREATE_TXG "create_txg"
796 #define ZPOOL_CONFIG_TOP_GUID "top_guid"
797 #define ZPOOL_CONFIG_VDEV_TREE "vdev_tree"
798 #define ZPOOL_CONFIG_TYPE "type"
799 #define ZPOOL_CONFIG_CHILDREN "children"
800 #define ZPOOL_CONFIG_ID "id"
801 #define ZPOOL_CONFIG_GUID "guid"
802 #define ZPOOL_CONFIG_INDIRECT_OBJECT "com.delphix:indirect_object"
803 #define ZPOOL_CONFIG_INDIRECT_BIRTHS "com.delphix:indirect_births"
804 #define ZPOOL_CONFIG_PREV_INDIRECT_VDEV "com.delphix:prev_indirect_vdev"
805 #define ZPOOL_CONFIG_PATH "path"
806 #define ZPOOL_CONFIG_DEVID "devid"
807 #define ZPOOL_CONFIG_METASLAB_ARRAY "metaslab_array"
808 #define ZPOOL_CONFIG_METASLAB_SHIFT "metaslab_shift"
809 #define ZPOOL_CONFIG_ASHIFT "ashift"
810 #define ZPOOL_CONFIG_ASIZE "asize"
811 #define ZPOOL_CONFIG_DTL "DTL"
812 #define ZPOOL_CONFIG_STATS "stats"
813 #define ZPOOL_CONFIG_WHOLE_DISK "whole_disk"
814 #define ZPOOL_CONFIG_ERRCOUNT "error_count"
815 #define ZPOOL_CONFIG_NOT_PRESENT "not_present"
816 #define ZPOOL_CONFIG_SPARES "spares"
817 #define ZPOOL_CONFIG_IS_SPARE "is_spare"
818 #define ZPOOL_CONFIG_NPARITY "nparity"
819 #define ZPOOL_CONFIG_HOSTID "hostid"
820 #define ZPOOL_CONFIG_HOSTNAME "hostname"
821 #define ZPOOL_CONFIG_IS_LOG "is_log"
822 #define ZPOOL_CONFIG_TIMESTAMP "timestamp" /* not stored on disk */
823 #define ZPOOL_CONFIG_FEATURES_FOR_READ "features_for_read"
824 #define ZPOOL_CONFIG_VDEV_CHILDREN "vdev_children"
827 * The persistent vdev state is stored as separate values rather than a single
828 * 'vdev_state' entry. This is because a device can be in multiple states, such
829 * as offline and degraded.
831 #define ZPOOL_CONFIG_OFFLINE "offline"
832 #define ZPOOL_CONFIG_FAULTED "faulted"
833 #define ZPOOL_CONFIG_DEGRADED "degraded"
834 #define ZPOOL_CONFIG_REMOVED "removed"
835 #define ZPOOL_CONFIG_FRU "fru"
836 #define ZPOOL_CONFIG_AUX_STATE "aux_state"
838 #define VDEV_TYPE_ROOT "root"
839 #define VDEV_TYPE_MIRROR "mirror"
840 #define VDEV_TYPE_REPLACING "replacing"
841 #define VDEV_TYPE_RAIDZ "raidz"
842 #define VDEV_TYPE_DISK "disk"
843 #define VDEV_TYPE_FILE "file"
844 #define VDEV_TYPE_MISSING "missing"
845 #define VDEV_TYPE_HOLE "hole"
846 #define VDEV_TYPE_SPARE "spare"
847 #define VDEV_TYPE_LOG "log"
848 #define VDEV_TYPE_L2CACHE "l2cache"
849 #define VDEV_TYPE_INDIRECT "indirect"
852 * This is needed in userland to report the minimum necessary device size.
854 #define SPA_MINDEVSIZE (64ULL << 20)
857 * The location of the pool configuration repository, shared between kernel and
860 #define ZPOOL_CACHE "/boot/zfs/zpool.cache"
863 * vdev states are ordered from least to most healthy.
864 * A vdev that's CANT_OPEN or below is considered unusable.
866 typedef enum vdev_state {
867 VDEV_STATE_UNKNOWN = 0, /* Uninitialized vdev */
868 VDEV_STATE_CLOSED, /* Not currently open */
869 VDEV_STATE_OFFLINE, /* Not allowed to open */
870 VDEV_STATE_REMOVED, /* Explicitly removed from system */
871 VDEV_STATE_CANT_OPEN, /* Tried to open, but failed */
872 VDEV_STATE_FAULTED, /* External request to fault device */
873 VDEV_STATE_DEGRADED, /* Replicated vdev with unhealthy kids */
874 VDEV_STATE_HEALTHY /* Presumed good */
878 * vdev aux states. When a vdev is in the CANT_OPEN state, the aux field
879 * of the vdev stats structure uses these constants to distinguish why.
881 typedef enum vdev_aux {
882 VDEV_AUX_NONE, /* no error */
883 VDEV_AUX_OPEN_FAILED, /* ldi_open_*() or vn_open() failed */
884 VDEV_AUX_CORRUPT_DATA, /* bad label or disk contents */
885 VDEV_AUX_NO_REPLICAS, /* insufficient number of replicas */
886 VDEV_AUX_BAD_GUID_SUM, /* vdev guid sum doesn't match */
887 VDEV_AUX_TOO_SMALL, /* vdev size is too small */
888 VDEV_AUX_BAD_LABEL, /* the label is OK but invalid */
889 VDEV_AUX_VERSION_NEWER, /* on-disk version is too new */
890 VDEV_AUX_VERSION_OLDER, /* on-disk version is too old */
891 VDEV_AUX_SPARED /* hot spare used in another pool */
895 * pool state. The following states are written to disk as part of the normal
896 * SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE. The remaining states are
897 * software abstractions used at various levels to communicate pool state.
899 typedef enum pool_state {
900 POOL_STATE_ACTIVE = 0, /* In active use */
901 POOL_STATE_EXPORTED, /* Explicitly exported */
902 POOL_STATE_DESTROYED, /* Explicitly destroyed */
903 POOL_STATE_SPARE, /* Reserved for hot spare use */
904 POOL_STATE_UNINITIALIZED, /* Internal spa_t state */
905 POOL_STATE_UNAVAIL, /* Internal libzfs state */
906 POOL_STATE_POTENTIALLY_ACTIVE /* Internal libzfs state */
910 * The uberblock version is incremented whenever an incompatible on-disk
911 * format change is made to the SPA, DMU, or ZAP.
913 * Note: the first two fields should never be moved. When a storage pool
914 * is opened, the uberblock must be read off the disk before the version
915 * can be checked. If the ub_version field is moved, we may not detect
916 * version mismatch. If the ub_magic field is moved, applications that
917 * expect the magic number in the first word won't work.
919 #define UBERBLOCK_MAGIC 0x00bab10c /* oo-ba-bloc! */
920 #define UBERBLOCK_SHIFT 10 /* up to 1K */
922 #define MMP_MAGIC 0xa11cea11 /* all-see-all */
924 #define MMP_INTERVAL_VALID_BIT 0x01
925 #define MMP_SEQ_VALID_BIT 0x02
926 #define MMP_FAIL_INT_VALID_BIT 0x04
928 #define MMP_VALID(ubp) (ubp->ub_magic == UBERBLOCK_MAGIC && \
929 ubp->ub_mmp_magic == MMP_MAGIC)
930 #define MMP_INTERVAL_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \
931 MMP_INTERVAL_VALID_BIT))
932 #define MMP_SEQ_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \
934 #define MMP_FAIL_INT_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \
935 MMP_FAIL_INT_VALID_BIT))
937 #define MMP_INTERVAL(ubp) ((ubp->ub_mmp_config & 0x00000000FFFFFF00) \
939 #define MMP_SEQ(ubp) ((ubp->ub_mmp_config & 0x0000FFFF00000000) \
941 #define MMP_FAIL_INT(ubp) ((ubp->ub_mmp_config & 0xFFFF000000000000) \
944 typedef struct uberblock {
945 uint64_t ub_magic; /* UBERBLOCK_MAGIC */
946 uint64_t ub_version; /* SPA_VERSION */
947 uint64_t ub_txg; /* txg of last sync */
948 uint64_t ub_guid_sum; /* sum of all vdev guids */
949 uint64_t ub_timestamp; /* UTC time of last sync */
950 blkptr_t ub_rootbp; /* MOS objset_phys_t */
951 /* highest SPA_VERSION supported by software that wrote this txg */
952 uint64_t ub_software_version;
953 /* Maybe missing in uberblocks we read, but always written */
954 uint64_t ub_mmp_magic;
956 * If ub_mmp_delay == 0 and ub_mmp_magic is valid, MMP is off.
957 * Otherwise, nanosec since last MMP write.
959 uint64_t ub_mmp_delay;
962 * The ub_mmp_config contains the multihost write interval, multihost
963 * fail intervals, sequence number for sub-second granularity, and
964 * valid bit mask. This layout is as follows:
966 * 64 56 48 40 32 24 16 8 0
967 * +-------+-------+-------+-------+-------+-------+-------+-------+
968 * 0 | Fail Intervals| Seq | Write Interval (ms) | VALID |
969 * +-------+-------+-------+-------+-------+-------+-------+-------+
971 * This allows a write_interval of (2^24/1000)s, over 4.5 hours
974 * - 0x01 - Write Interval (ms)
975 * - 0x02 - Sequence number exists
976 * - 0x04 - Fail Intervals
979 uint64_t ub_mmp_config;
982 * ub_checkpoint_txg indicates two things about the current uberblock:
984 * 1] If it is not zero then this uberblock is a checkpoint. If it is
985 * zero, then this uberblock is not a checkpoint.
987 * 2] On checkpointed uberblocks, the value of ub_checkpoint_txg is
988 * the ub_txg that the uberblock had at the time we moved it to
991 * The field is set when we checkpoint the uberblock and continues to
992 * hold that value even after we've rewound (unlike the ub_txg that
993 * is reset to a higher value).
995 * Besides checks used to determine whether we are reopening the
996 * pool from a checkpointed uberblock [see spa_ld_select_uberblock()],
997 * the value of the field is used to determine which ZIL blocks have
998 * been allocated according to the ms_sm when we are rewinding to a
999 * checkpoint. Specifically, if blk_birth > ub_checkpoint_txg, then
1000 * the ZIL block is not allocated [see uses of spa_min_claim_txg()].
1002 uint64_t ub_checkpoint_txg;
1008 #define DNODE_MUST_BE_ALLOCATED 1
1009 #define DNODE_MUST_BE_FREE 2
1014 #define DNODE_SHIFT 9 /* 512 bytes */
1015 #define DN_MIN_INDBLKSHIFT 12 /* 4k */
1016 #define DN_MAX_INDBLKSHIFT 17 /* 128k */
1017 #define DNODE_BLOCK_SHIFT 14 /* 16k */
1018 #define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */
1019 #define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */
1020 #define DN_MAX_OFFSET_SHIFT 64 /* 2^64 bytes in a dnode */
1023 * Derived constants.
1025 #define DNODE_MIN_SIZE (1 << DNODE_SHIFT)
1026 #define DNODE_MAX_SIZE (1 << DNODE_BLOCK_SHIFT)
1027 #define DNODE_BLOCK_SIZE (1 << DNODE_BLOCK_SHIFT)
1028 #define DNODE_MIN_SLOTS (DNODE_MIN_SIZE >> DNODE_SHIFT)
1029 #define DNODE_MAX_SLOTS (DNODE_MAX_SIZE >> DNODE_SHIFT)
1030 #define DN_BONUS_SIZE(dnsize) ((dnsize) - DNODE_CORE_SIZE - \
1031 (1 << SPA_BLKPTRSHIFT))
1032 #define DN_SLOTS_TO_BONUSLEN(slots) DN_BONUS_SIZE((slots) << DNODE_SHIFT)
1033 #define DN_OLD_MAX_BONUSLEN (DN_BONUS_SIZE(DNODE_MIN_SIZE))
1034 #define DN_MAX_NBLKPTR ((DNODE_MIN_SIZE - DNODE_CORE_SIZE) >> \
1036 #define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT)
1037 #define DN_ZERO_BONUSLEN (DN_BONUS_SIZE(DNODE_MAX_SIZE) + 1)
1039 #define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT)
1040 #define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT)
1041 #define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT)
1043 /* The +2 here is a cheesy way to round up */
1044 #define DN_MAX_LEVELS (2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \
1045 (DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT)))
1047 #define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \
1048 (((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t))))
1050 #define DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \
1051 (dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT)
1053 #define EPB(blkshift, typeshift) (1 << (blkshift - typeshift))
1055 /* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */
1056 #define DNODE_FLAG_USED_BYTES (1<<0)
1057 #define DNODE_FLAG_USERUSED_ACCOUNTED (1<<1)
1059 /* Does dnode have a SA spill blkptr in bonus? */
1060 #define DNODE_FLAG_SPILL_BLKPTR (1<<2)
1062 typedef struct dnode_phys {
1063 uint8_t dn_type; /* dmu_object_type_t */
1064 uint8_t dn_indblkshift; /* ln2(indirect block size) */
1065 uint8_t dn_nlevels; /* 1=dn_blkptr->data blocks */
1066 uint8_t dn_nblkptr; /* length of dn_blkptr */
1067 uint8_t dn_bonustype; /* type of data in bonus buffer */
1068 uint8_t dn_checksum; /* ZIO_CHECKSUM type */
1069 uint8_t dn_compress; /* ZIO_COMPRESS type */
1070 uint8_t dn_flags; /* DNODE_FLAG_* */
1071 uint16_t dn_datablkszsec; /* data block size in 512b sectors */
1072 uint16_t dn_bonuslen; /* length of dn_bonus */
1073 uint8_t dn_extra_slots; /* # of subsequent slots consumed */
1076 /* accounting is protected by dn_dirty_mtx */
1077 uint64_t dn_maxblkid; /* largest allocated block ID */
1078 uint64_t dn_used; /* bytes (or sectors) of disk space */
1080 uint64_t dn_pad3[4];
1083 * The tail region is 448 bytes for a 512 byte dnode, and
1084 * correspondingly larger for larger dnode sizes. The spill
1085 * block pointer, when present, is always at the end of the tail
1086 * region. There are three ways this space may be used, using
1087 * a 512 byte dnode for this diagram:
1089 * 0 64 128 192 256 320 384 448 (offset)
1090 * +---------------+---------------+---------------+-------+
1091 * | dn_blkptr[0] | dn_blkptr[1] | dn_blkptr[2] | / |
1092 * +---------------+---------------+---------------+-------+
1093 * | dn_blkptr[0] | dn_bonus[0..319] |
1094 * +---------------+-----------------------+---------------+
1095 * | dn_blkptr[0] | dn_bonus[0..191] | dn_spill |
1096 * +---------------+-----------------------+---------------+
1099 blkptr_t dn_blkptr[1+DN_OLD_MAX_BONUSLEN/sizeof (blkptr_t)];
1101 blkptr_t __dn_ignore1;
1102 uint8_t dn_bonus[DN_OLD_MAX_BONUSLEN];
1105 blkptr_t __dn_ignore2;
1106 uint8_t __dn_ignore3[DN_OLD_MAX_BONUSLEN -
1113 #define DN_SPILL_BLKPTR(dnp) (blkptr_t *)((char *)(dnp) + \
1114 (((dnp)->dn_extra_slots + 1) << DNODE_SHIFT) - (1 << SPA_BLKPTRSHIFT))
1116 typedef enum dmu_object_byteswap {
1128 * Allocating a new byteswap type number makes the on-disk format
1129 * incompatible with any other format that uses the same number.
1131 * Data can usually be structured to work with one of the
1132 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types.
1135 } dmu_object_byteswap_t;
1137 #define DMU_OT_NEWTYPE 0x80
1138 #define DMU_OT_METADATA 0x40
1139 #define DMU_OT_BYTESWAP_MASK 0x3f
1142 * Defines a uint8_t object type. Object types specify if the data
1143 * in the object is metadata (boolean) and how to byteswap the data
1144 * (dmu_object_byteswap_t).
1146 #define DMU_OT(byteswap, metadata) \
1148 ((metadata) ? DMU_OT_METADATA : 0) | \
1149 ((byteswap) & DMU_OT_BYTESWAP_MASK))
1151 typedef enum dmu_object_type {
1154 DMU_OT_OBJECT_DIRECTORY, /* ZAP */
1155 DMU_OT_OBJECT_ARRAY, /* UINT64 */
1156 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */
1157 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */
1158 DMU_OT_BPOBJ, /* UINT64 */
1159 DMU_OT_BPOBJ_HDR, /* UINT64 */
1161 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */
1162 DMU_OT_SPACE_MAP, /* UINT64 */
1164 DMU_OT_INTENT_LOG, /* UINT64 */
1166 DMU_OT_DNODE, /* DNODE */
1167 DMU_OT_OBJSET, /* OBJSET */
1169 DMU_OT_DSL_DIR, /* UINT64 */
1170 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */
1171 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */
1172 DMU_OT_DSL_PROPS, /* ZAP */
1173 DMU_OT_DSL_DATASET, /* UINT64 */
1175 DMU_OT_ZNODE, /* ZNODE */
1176 DMU_OT_OLDACL, /* Old ACL */
1177 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */
1178 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */
1179 DMU_OT_MASTER_NODE, /* ZAP */
1180 DMU_OT_UNLINKED_SET, /* ZAP */
1182 DMU_OT_ZVOL, /* UINT8 */
1183 DMU_OT_ZVOL_PROP, /* ZAP */
1184 /* other; for testing only! */
1185 DMU_OT_PLAIN_OTHER, /* UINT8 */
1186 DMU_OT_UINT64_OTHER, /* UINT64 */
1187 DMU_OT_ZAP_OTHER, /* ZAP */
1188 /* new object types: */
1189 DMU_OT_ERROR_LOG, /* ZAP */
1190 DMU_OT_SPA_HISTORY, /* UINT8 */
1191 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */
1192 DMU_OT_POOL_PROPS, /* ZAP */
1193 DMU_OT_DSL_PERMS, /* ZAP */
1194 DMU_OT_ACL, /* ACL */
1195 DMU_OT_SYSACL, /* SYSACL */
1196 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */
1197 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */
1198 DMU_OT_NEXT_CLONES, /* ZAP */
1199 DMU_OT_SCAN_QUEUE, /* ZAP */
1200 DMU_OT_USERGROUP_USED, /* ZAP */
1201 DMU_OT_USERGROUP_QUOTA, /* ZAP */
1202 DMU_OT_USERREFS, /* ZAP */
1203 DMU_OT_DDT_ZAP, /* ZAP */
1204 DMU_OT_DDT_STATS, /* ZAP */
1205 DMU_OT_SA, /* System attr */
1206 DMU_OT_SA_MASTER_NODE, /* ZAP */
1207 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */
1208 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */
1209 DMU_OT_SCAN_XLATE, /* ZAP */
1210 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */
1211 DMU_OT_DEADLIST, /* ZAP */
1212 DMU_OT_DEADLIST_HDR, /* UINT64 */
1213 DMU_OT_DSL_CLONES, /* ZAP */
1214 DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */
1218 * Names for valid types declared with DMU_OT().
1220 DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE),
1221 DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE),
1222 DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE),
1223 DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE),
1224 DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE),
1225 DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE),
1226 DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE),
1227 DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE),
1228 DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE),
1229 DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE)
1230 } dmu_object_type_t;
1232 typedef enum dmu_objset_type {
1237 DMU_OST_OTHER, /* For testing only! */
1238 DMU_OST_ANY, /* Be careful! */
1240 } dmu_objset_type_t;
1242 #define ZAP_MAXVALUELEN (1024 * 8)
1245 * header for all bonus and spill buffers.
1246 * The header has a fixed portion with a variable number
1247 * of "lengths" depending on the number of variable sized
1248 * attribues which are determined by the "layout number"
1251 #define SA_MAGIC 0x2F505A /* ZFS SA */
1252 typedef struct sa_hdr_phys {
1254 uint16_t sa_layout_info; /* Encoded with hdrsize and layout number */
1255 uint16_t sa_lengths[1]; /* optional sizes for variable length attrs */
1256 /* ... Data follows the lengths. */
1260 * sa_hdr_phys -> sa_layout_info
1263 * +--------+-------+
1265 * +--------+-------+
1267 * Bits 0-10 are the layout number
1268 * Bits 11-16 are the size of the header.
1269 * The hdrsize is the number * 8
1272 * hdrsz of 1 ==> 8 byte header
1273 * 2 ==> 16 byte header
1277 #define SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10)
1278 #define SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 16, 3, 0)
1279 #define SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \
1281 BF32_SET_SB(x, 10, 6, 3, 0, size); \
1282 BF32_SET(x, 0, 10, num); \
1285 #define SA_ATTR_BSWAP(x) BF32_GET(x, 16, 8)
1286 #define SA_ATTR_LENGTH(x) BF32_GET(x, 24, 16)
1287 #define SA_ATTR_NUM(x) BF32_GET(x, 0, 16)
1288 #define SA_ATTR_ENCODE(x, attr, length, bswap) \
1290 BF64_SET(x, 24, 16, length); \
1291 BF64_SET(x, 16, 8, bswap); \
1292 BF64_SET(x, 0, 16, attr); \
1295 #define SA_MODE_OFFSET 0
1296 #define SA_SIZE_OFFSET 8
1297 #define SA_GEN_OFFSET 16
1298 #define SA_UID_OFFSET 24
1299 #define SA_GID_OFFSET 32
1300 #define SA_PARENT_OFFSET 40
1301 #define SA_SYMLINK_OFFSET 160
1303 #define SA_REGISTRY "REGISTRY"
1304 #define SA_LAYOUTS "LAYOUTS"
1306 typedef enum sa_bswap_type {
1314 typedef uint16_t sa_attr_type_t;
1316 #define ZIO_OBJSET_MAC_LEN 32
1319 * Intent log header - this on disk structure holds fields to manage
1320 * the log. All fields are 64 bit to easily handle cross architectures.
1322 typedef struct zil_header {
1323 uint64_t zh_claim_txg; /* txg in which log blocks were claimed */
1324 uint64_t zh_replay_seq; /* highest replayed sequence number */
1325 blkptr_t zh_log; /* log chain */
1326 uint64_t zh_claim_seq; /* highest claimed sequence number */
1330 #define OBJSET_PHYS_SIZE_V2 2048
1331 #define OBJSET_PHYS_SIZE_V3 4096
1333 typedef struct objset_phys {
1334 dnode_phys_t os_meta_dnode;
1335 zil_header_t os_zil_header;
1338 uint8_t os_portable_mac[ZIO_OBJSET_MAC_LEN];
1339 uint8_t os_local_mac[ZIO_OBJSET_MAC_LEN];
1340 char os_pad0[OBJSET_PHYS_SIZE_V2 - sizeof (dnode_phys_t)*3 -
1341 sizeof (zil_header_t) - sizeof (uint64_t)*2 -
1342 2*ZIO_OBJSET_MAC_LEN];
1343 dnode_phys_t os_userused_dnode;
1344 dnode_phys_t os_groupused_dnode;
1345 dnode_phys_t os_projectused_dnode;
1346 char os_pad1[OBJSET_PHYS_SIZE_V3 - OBJSET_PHYS_SIZE_V2 -
1347 sizeof (dnode_phys_t)];
1350 typedef struct space_map_phys {
1351 /* object number: not needed but kept for backwards compatibility */
1352 uint64_t smp_object;
1354 /* length of the object in bytes */
1355 uint64_t smp_length;
1357 /* space allocated from the map */
1366 /* one-word entry constants */
1367 #define SM_DEBUG_PREFIX 2
1368 #define SM_OFFSET_BITS 47
1369 #define SM_RUN_BITS 15
1371 /* two-word entry constants */
1372 #define SM2_PREFIX 3
1373 #define SM2_OFFSET_BITS 63
1374 #define SM2_RUN_BITS 36
1376 #define SM_PREFIX_DECODE(x) BF64_DECODE(x, 62, 2)
1377 #define SM_PREFIX_ENCODE(x) BF64_ENCODE(x, 62, 2)
1379 #define SM_DEBUG_ACTION_DECODE(x) BF64_DECODE(x, 60, 2)
1380 #define SM_DEBUG_ACTION_ENCODE(x) BF64_ENCODE(x, 60, 2)
1381 #define SM_DEBUG_SYNCPASS_DECODE(x) BF64_DECODE(x, 50, 10)
1382 #define SM_DEBUG_SYNCPASS_ENCODE(x) BF64_ENCODE(x, 50, 10)
1383 #define SM_DEBUG_TXG_DECODE(x) BF64_DECODE(x, 0, 50)
1384 #define SM_DEBUG_TXG_ENCODE(x) BF64_ENCODE(x, 0, 50)
1386 #define SM_OFFSET_DECODE(x) BF64_DECODE(x, 16, SM_OFFSET_BITS)
1387 #define SM_OFFSET_ENCODE(x) BF64_ENCODE(x, 16, SM_OFFSET_BITS)
1388 #define SM_TYPE_DECODE(x) BF64_DECODE(x, 15, 1)
1389 #define SM_TYPE_ENCODE(x) BF64_ENCODE(x, 15, 1)
1390 #define SM_RUN_DECODE(x) (BF64_DECODE(x, 0, SM_RUN_BITS) + 1)
1391 #define SM_RUN_ENCODE(x) BF64_ENCODE((x) - 1, 0, SM_RUN_BITS)
1392 #define SM_RUN_MAX SM_RUN_DECODE(~0ULL)
1393 #define SM_OFFSET_MAX SM_OFFSET_DECODE(~0ULL)
1395 #define SM2_RUN_DECODE(x) (BF64_DECODE(x, 24, SM2_RUN_BITS) + 1)
1396 #define SM2_RUN_ENCODE(x) BF64_ENCODE((x) - 1, 24, SM2_RUN_BITS)
1397 #define SM2_VDEV_DECODE(x) BF64_DECODE(x, 0, 24)
1398 #define SM2_VDEV_ENCODE(x) BF64_ENCODE(x, 0, 24)
1399 #define SM2_TYPE_DECODE(x) BF64_DECODE(x, SM2_OFFSET_BITS, 1)
1400 #define SM2_TYPE_ENCODE(x) BF64_ENCODE(x, SM2_OFFSET_BITS, 1)
1401 #define SM2_OFFSET_DECODE(x) BF64_DECODE(x, 0, SM2_OFFSET_BITS)
1402 #define SM2_OFFSET_ENCODE(x) BF64_ENCODE(x, 0, SM2_OFFSET_BITS)
1403 #define SM2_RUN_MAX SM2_RUN_DECODE(~0ULL)
1404 #define SM2_OFFSET_MAX SM2_OFFSET_DECODE(~0ULL)
1406 typedef struct dsl_dir_phys {
1407 uint64_t dd_creation_time; /* not actually used */
1408 uint64_t dd_head_dataset_obj;
1409 uint64_t dd_parent_obj;
1410 uint64_t dd_clone_parent_obj;
1411 uint64_t dd_child_dir_zapobj;
1413 * how much space our children are accounting for; for leaf
1414 * datasets, == physical space used by fs + snaps
1416 uint64_t dd_used_bytes;
1417 uint64_t dd_compressed_bytes;
1418 uint64_t dd_uncompressed_bytes;
1419 /* Administrative quota setting */
1421 /* Administrative reservation setting */
1422 uint64_t dd_reserved;
1423 uint64_t dd_props_zapobj;
1426 uint64_t dd_pad1[13]; /* pad out to 256 bytes for good measure */
1429 typedef struct dsl_dataset_phys {
1430 uint64_t ds_dir_obj;
1431 uint64_t ds_prev_snap_obj;
1432 uint64_t ds_prev_snap_txg;
1433 uint64_t ds_next_snap_obj;
1434 uint64_t ds_snapnames_zapobj; /* zap obj of snaps; ==0 for snaps */
1435 uint64_t ds_num_children; /* clone/snap children; ==0 for head */
1436 uint64_t ds_creation_time; /* seconds since 1970 */
1437 uint64_t ds_creation_txg;
1438 uint64_t ds_deadlist_obj;
1439 uint64_t ds_used_bytes;
1440 uint64_t ds_compressed_bytes;
1441 uint64_t ds_uncompressed_bytes;
1442 uint64_t ds_unique_bytes; /* only relevant to snapshots */
1444 * The ds_fsid_guid is a 56-bit ID that can change to avoid
1445 * collisions. The ds_guid is a 64-bit ID that will never
1446 * change, so there is a small probability that it will collide.
1448 uint64_t ds_fsid_guid;
1452 uint64_t ds_pad[8]; /* pad out to 320 bytes for good measure */
1453 } dsl_dataset_phys_t;
1455 typedef struct dsl_deadlist_phys {
1459 uint64_t dl_pad[37]; /* pad out to 320b for future expansion */
1460 } dsl_deadlist_phys_t;
1462 #define BPOBJ_SIZE_V2 (6 * sizeof (uint64_t))
1464 typedef struct bpobj_phys {
1465 uint64_t bpo_num_blkptrs;
1468 uint64_t bpo_uncomp;
1469 uint64_t bpo_subobjs;
1470 uint64_t bpo_num_subobjs;
1471 uint64_t bpo_num_freed;
1475 * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
1477 #define DMU_POOL_DIRECTORY_OBJECT 1
1478 #define DMU_POOL_CONFIG "config"
1479 #define DMU_POOL_FEATURES_FOR_READ "features_for_read"
1480 #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write"
1481 #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions"
1482 #define DMU_POOL_ROOT_DATASET "root_dataset"
1483 #define DMU_POOL_SYNC_BPLIST "sync_bplist"
1484 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub"
1485 #define DMU_POOL_ERRLOG_LAST "errlog_last"
1486 #define DMU_POOL_SPARES "spares"
1487 #define DMU_POOL_DEFLATE "deflate"
1488 #define DMU_POOL_HISTORY "history"
1489 #define DMU_POOL_PROPS "pool_props"
1490 #define DMU_POOL_FREE_BPOBJ "free_bpobj"
1491 #define DMU_POOL_BPTREE_OBJ "bptree_obj"
1492 #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj"
1493 #define DMU_POOL_TMP_USERREFS "tmp_userrefs"
1494 #define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt"
1495 #define DMU_POOL_REMOVING "com.delphix:removing"
1496 #define DMU_POOL_OBSOLETE_BPOBJ "com.delphix:obsolete_bpobj"
1497 #define DMU_POOL_CONDENSING_INDIRECT "com.delphix:condensing_indirect"
1498 #define DMU_POOL_ZPOOL_CHECKPOINT "com.delphix:zpool_checkpoint"
1500 #define ZAP_MAGIC 0x2F52AB2ABULL
1502 #define FZAP_BLOCK_SHIFT(zap) ((zap)->zap_block_shift)
1504 #define ZAP_MAXCD (uint32_t)(-1)
1505 #define ZAP_HASHBITS 28
1506 #define MZAP_ENT_LEN 64
1507 #define MZAP_ENT_MAX \
1508 ((MZAP_MAX_BLKSZ - sizeof(mzap_phys_t)) / sizeof(mzap_ent_phys_t) + 1)
1509 #define MZAP_NAME_LEN (MZAP_ENT_LEN - 8 - 4 - 2)
1510 #define MZAP_MAX_BLKSZ SPA_OLDMAXBLOCKSIZE
1512 typedef struct mzap_ent_phys {
1515 uint16_t mze_pad; /* in case we want to chain them someday */
1516 char mze_name[MZAP_NAME_LEN];
1519 typedef struct mzap_phys {
1520 uint64_t mz_block_type; /* ZBT_MICRO */
1522 uint64_t mz_normflags;
1524 mzap_ent_phys_t mz_chunk[1];
1525 /* actually variable size depending on block size */
1529 * The (fat) zap is stored in one object. It is an array of
1530 * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of:
1532 * ptrtbl fits in first block:
1533 * [zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ...
1535 * ptrtbl too big for first block:
1536 * [zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ...
1540 #define ZBT_LEAF ((1ULL << 63) + 0)
1541 #define ZBT_HEADER ((1ULL << 63) + 1)
1542 #define ZBT_MICRO ((1ULL << 63) + 3)
1543 /* any other values are ptrtbl blocks */
1546 * the embedded pointer table takes up half a block:
1547 * block size / entry size (2^3) / 2
1549 #define ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1)
1552 * The embedded pointer table starts half-way through the block. Since
1553 * the pointer table itself is half the block, it starts at (64-bit)
1554 * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)).
1556 #define ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \
1557 ((uint64_t *)(zap)->zap_phys) \
1558 [(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))]
1560 #define ZAP_HASH_IDX(hash, n) (((n) == 0) ? 0 : ((hash) >> (64 - (n))))
1564 * If zap_phys_t is modified, zap_byteswap() must be modified.
1566 typedef struct zap_phys {
1567 uint64_t zap_block_type; /* ZBT_HEADER */
1568 uint64_t zap_magic; /* ZAP_MAGIC */
1570 struct zap_table_phys {
1571 uint64_t zt_blk; /* starting block number */
1572 uint64_t zt_numblks; /* number of blocks */
1573 uint64_t zt_shift; /* bits to index it */
1574 uint64_t zt_nextblk; /* next (larger) copy start block */
1575 uint64_t zt_blks_copied; /* number source blocks copied */
1578 uint64_t zap_freeblk; /* the next free block */
1579 uint64_t zap_num_leafs; /* number of leafs */
1580 uint64_t zap_num_entries; /* number of entries */
1581 uint64_t zap_salt; /* salt to stir into hash function */
1582 uint64_t zap_normflags; /* flags for u8_textprep_str() */
1583 uint64_t zap_flags; /* zap_flags_t */
1585 * This structure is followed by padding, and then the embedded
1586 * pointer table. The embedded pointer table takes up second
1587 * half of the block. It is accessed using the
1588 * ZAP_EMBEDDED_PTRTBL_ENT() macro.
1592 typedef struct zap_table_phys zap_table_phys_t;
1595 typedef struct fat_zap {
1596 int zap_block_shift; /* block size shift */
1597 zap_phys_t *zap_phys;
1598 const struct spa *zap_spa;
1599 const dnode_phys_t *zap_dnode;
1602 #define ZAP_LEAF_MAGIC 0x2AB1EAF
1604 /* chunk size = 24 bytes */
1605 #define ZAP_LEAF_CHUNKSIZE 24
1608 * The amount of space available for chunks is:
1609 * block size (1<<l->l_bs) - hash entry size (2) * number of hash
1610 * entries - header space (2*chunksize)
1612 #define ZAP_LEAF_NUMCHUNKS(l) \
1613 (((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \
1614 ZAP_LEAF_CHUNKSIZE - 2)
1617 * The amount of space within the chunk available for the array is:
1618 * chunk size - space for type (1) - space for next pointer (2)
1620 #define ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3)
1622 #define ZAP_LEAF_ARRAY_NCHUNKS(bytes) \
1623 (((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES)
1626 * Low water mark: when there are only this many chunks free, start
1627 * growing the ptrtbl. Ideally, this should be larger than a
1628 * "reasonably-sized" entry. 20 chunks is more than enough for the
1629 * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value),
1630 * while still being only around 3% for 16k blocks.
1632 #define ZAP_LEAF_LOW_WATER (20)
1635 * The leaf hash table has block size / 2^5 (32) number of entries,
1636 * which should be more than enough for the maximum number of entries,
1637 * which is less than block size / CHUNKSIZE (24) / minimum number of
1638 * chunks per entry (3).
1640 #define ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5)
1641 #define ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l))
1644 * The chunks start immediately after the hash table. The end of the
1645 * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a
1648 #define ZAP_LEAF_CHUNK(l, idx) \
1649 ((zap_leaf_chunk_t *)(void *) \
1650 ((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx]
1651 #define ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry)
1653 #define ZAP_LEAF_HASH(l, h) \
1654 ((ZAP_LEAF_HASH_NUMENTRIES(l)-1) & \
1656 (64 - ZAP_LEAF_HASH_SHIFT(l) - (l)->l_phys->l_hdr.lh_prefix_len)))
1657 #define ZAP_LEAF_HASH_ENTPTR(l, h) (&(l)->l_phys->l_hash[ZAP_LEAF_HASH(l, h)])
1659 typedef enum zap_chunk_type {
1660 ZAP_CHUNK_FREE = 253,
1661 ZAP_CHUNK_ENTRY = 252,
1662 ZAP_CHUNK_ARRAY = 251,
1663 ZAP_CHUNK_TYPE_MAX = 250
1668 * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified.
1670 typedef struct zap_leaf_phys {
1671 struct zap_leaf_header {
1672 uint64_t lh_block_type; /* ZBT_LEAF */
1674 uint64_t lh_prefix; /* hash prefix of this leaf */
1675 uint32_t lh_magic; /* ZAP_LEAF_MAGIC */
1676 uint16_t lh_nfree; /* number free chunks */
1677 uint16_t lh_nentries; /* number of entries */
1678 uint16_t lh_prefix_len; /* num bits used to id this */
1680 /* above is accessable to zap, below is zap_leaf private */
1682 uint16_t lh_freelist; /* chunk head of free list */
1683 uint8_t lh_pad2[12];
1684 } l_hdr; /* 2 24-byte chunks */
1687 * The header is followed by a hash table with
1688 * ZAP_LEAF_HASH_NUMENTRIES(zap) entries. The hash table is
1689 * followed by an array of ZAP_LEAF_NUMCHUNKS(zap)
1690 * zap_leaf_chunk structures. These structures are accessed
1691 * with the ZAP_LEAF_CHUNK() macro.
1697 typedef union zap_leaf_chunk {
1698 struct zap_leaf_entry {
1699 uint8_t le_type; /* always ZAP_CHUNK_ENTRY */
1700 uint8_t le_value_intlen; /* size of ints */
1701 uint16_t le_next; /* next entry in hash chain */
1702 uint16_t le_name_chunk; /* first chunk of the name */
1703 uint16_t le_name_numints; /* bytes in name, incl null */
1704 uint16_t le_value_chunk; /* first chunk of the value */
1705 uint16_t le_value_numints; /* value length in ints */
1706 uint32_t le_cd; /* collision differentiator */
1707 uint64_t le_hash; /* hash value of the name */
1709 struct zap_leaf_array {
1710 uint8_t la_type; /* always ZAP_CHUNK_ARRAY */
1711 uint8_t la_array[ZAP_LEAF_ARRAY_BYTES];
1712 uint16_t la_next; /* next blk or CHAIN_END */
1714 struct zap_leaf_free {
1715 uint8_t lf_type; /* always ZAP_CHUNK_FREE */
1716 uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES];
1717 uint16_t lf_next; /* next in free list, or CHAIN_END */
1721 typedef struct zap_leaf {
1722 int l_bs; /* block size shift */
1723 zap_leaf_phys_t *l_phys;
1726 #define ZAP_MAXNAMELEN 256
1727 #define ZAP_MAXVALUELEN (1024 * 8)
1729 #define ACE_READ_DATA 0x00000001 /* file: read data */
1730 #define ACE_LIST_DIRECTORY 0x00000001 /* dir: list files */
1731 #define ACE_WRITE_DATA 0x00000002 /* file: write data */
1732 #define ACE_ADD_FILE 0x00000002 /* dir: create file */
1733 #define ACE_APPEND_DATA 0x00000004 /* file: append data */
1734 #define ACE_ADD_SUBDIRECTORY 0x00000004 /* dir: create subdir */
1735 #define ACE_READ_NAMED_ATTRS 0x00000008 /* FILE_READ_EA */
1736 #define ACE_WRITE_NAMED_ATTRS 0x00000010 /* FILE_WRITE_EA */
1737 #define ACE_EXECUTE 0x00000020 /* file: execute */
1738 #define ACE_TRAVERSE 0x00000020 /* dir: lookup name */
1739 #define ACE_DELETE_CHILD 0x00000040 /* dir: unlink child */
1740 #define ACE_READ_ATTRIBUTES 0x00000080 /* (all) stat, etc. */
1741 #define ACE_WRITE_ATTRIBUTES 0x00000100 /* (all) utimes, etc. */
1742 #define ACE_DELETE 0x00010000 /* (all) unlink self */
1743 #define ACE_READ_ACL 0x00020000 /* (all) getsecattr */
1744 #define ACE_WRITE_ACL 0x00040000 /* (all) setsecattr */
1745 #define ACE_WRITE_OWNER 0x00080000 /* (all) chown */
1746 #define ACE_SYNCHRONIZE 0x00100000 /* (all) */
1748 #define ACE_FILE_INHERIT_ACE 0x0001
1749 #define ACE_DIRECTORY_INHERIT_ACE 0x0002
1750 #define ACE_NO_PROPAGATE_INHERIT_ACE 0x0004
1751 #define ACE_INHERIT_ONLY_ACE 0x0008
1752 #define ACE_SUCCESSFUL_ACCESS_ACE_FLAG 0x0010
1753 #define ACE_FAILED_ACCESS_ACE_FLAG 0x0020
1754 #define ACE_IDENTIFIER_GROUP 0x0040
1755 #define ACE_INHERITED_ACE 0x0080
1756 #define ACE_OWNER 0x1000
1757 #define ACE_GROUP 0x2000
1758 #define ACE_EVERYONE 0x4000
1760 #define ACE_ACCESS_ALLOWED_ACE_TYPE 0x0000
1761 #define ACE_ACCESS_DENIED_ACE_TYPE 0x0001
1762 #define ACE_SYSTEM_AUDIT_ACE_TYPE 0x0002
1763 #define ACE_SYSTEM_ALARM_ACE_TYPE 0x0003
1765 typedef struct zfs_ace_hdr {
1768 uint32_t z_access_mask;
1772 * Define special zfs pflags
1774 #define ZFS_XATTR 0x1 /* is an extended attribute */
1775 #define ZFS_INHERIT_ACE 0x2 /* ace has inheritable ACEs */
1776 #define ZFS_ACL_TRIVIAL 0x4 /* files ACL is trivial */
1777 #define ZFS_ACL_OBJ_ACE 0x8 /* ACL has CMPLX Object ACE */
1778 #define ZFS_ACL_PROTECTED 0x10 /* ACL protected */
1779 #define ZFS_ACL_DEFAULTED 0x20 /* ACL should be defaulted */
1780 #define ZFS_ACL_AUTO_INHERIT 0x40 /* ACL should be inherited */
1781 #define ZFS_BONUS_SCANSTAMP 0x80 /* Scanstamp in bonus area */
1782 #define ZFS_NO_EXECS_DENIED 0x100 /* exec was given to everyone */
1784 #define ZFS_READONLY 0x0000000100000000ull
1785 #define ZFS_HIDDEN 0x0000000200000000ull
1786 #define ZFS_SYSTEM 0x0000000400000000ull
1787 #define ZFS_ARCHIVE 0x0000000800000000ull
1788 #define ZFS_IMMUTABLE 0x0000001000000000ull
1789 #define ZFS_NOUNLINK 0x0000002000000000ull
1790 #define ZFS_APPENDONLY 0x0000004000000000ull
1791 #define ZFS_NODUMP 0x0000008000000000ull
1792 #define ZFS_OPAQUE 0x0000010000000000ull
1793 #define ZFS_AV_QUARANTINED 0x0000020000000000ull
1794 #define ZFS_AV_MODIFIED 0x0000040000000000ull
1795 #define ZFS_REPARSE 0x0000080000000000ull
1796 #define ZFS_OFFLINE 0x0000100000000000ull
1797 #define ZFS_SPARSE 0x0000200000000000ull
1799 #define MASTER_NODE_OBJ 1
1802 * special attributes for master node.
1805 #define ZFS_FSID "FSID"
1806 #define ZFS_UNLINKED_SET "DELETE_QUEUE"
1807 #define ZFS_ROOT_OBJ "ROOT"
1808 #define ZPL_VERSION_OBJ "VERSION"
1809 #define ZFS_PROP_BLOCKPERPAGE "BLOCKPERPAGE"
1810 #define ZFS_PROP_NOGROWBLOCKS "NOGROWBLOCKS"
1811 #define ZFS_SA_ATTRS "SA_ATTRS"
1813 #define ZFS_FLAG_BLOCKPERPAGE 0x1
1814 #define ZFS_FLAG_NOGROWBLOCKS 0x2
1817 * ZPL version - rev'd whenever an incompatible on-disk format change
1818 * occurs. Independent of SPA/DMU/ZAP versioning.
1821 #define ZPL_VERSION 1ULL
1824 * The directory entry has the type (currently unused on Solaris) in the
1825 * top 4 bits, and the object number in the low 48 bits. The "middle"
1826 * 12 bits are unused.
1828 #define ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4)
1829 #define ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48)
1830 #define ZFS_DIRENT_MAKE(type, obj) (((uint64_t)type << 60) | obj)
1832 typedef struct ace {
1833 uid_t a_who; /* uid or gid */
1834 uint32_t a_access_mask; /* read,write,... */
1835 uint16_t a_flags; /* see below */
1836 uint16_t a_type; /* allow or deny */
1839 #define ACE_SLOT_CNT 6
1841 typedef struct zfs_znode_acl {
1842 uint64_t z_acl_extern_obj; /* ext acl pieces */
1843 uint32_t z_acl_count; /* Number of ACEs */
1844 uint16_t z_acl_version; /* acl version */
1845 uint16_t z_acl_pad; /* pad */
1846 ace_t z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */
1850 * This is the persistent portion of the znode. It is stored
1851 * in the "bonus buffer" of the file. Short symbolic links
1852 * are also stored in the bonus buffer.
1854 typedef struct znode_phys {
1855 uint64_t zp_atime[2]; /* 0 - last file access time */
1856 uint64_t zp_mtime[2]; /* 16 - last file modification time */
1857 uint64_t zp_ctime[2]; /* 32 - last file change time */
1858 uint64_t zp_crtime[2]; /* 48 - creation time */
1859 uint64_t zp_gen; /* 64 - generation (txg of creation) */
1860 uint64_t zp_mode; /* 72 - file mode bits */
1861 uint64_t zp_size; /* 80 - size of file */
1862 uint64_t zp_parent; /* 88 - directory parent (`..') */
1863 uint64_t zp_links; /* 96 - number of links to file */
1864 uint64_t zp_xattr; /* 104 - DMU object for xattrs */
1865 uint64_t zp_rdev; /* 112 - dev_t for VBLK & VCHR files */
1866 uint64_t zp_flags; /* 120 - persistent flags */
1867 uint64_t zp_uid; /* 128 - file owner */
1868 uint64_t zp_gid; /* 136 - owning group */
1869 uint64_t zp_pad[4]; /* 144 - future */
1870 zfs_znode_acl_t zp_acl; /* 176 - 263 ACL */
1872 * Data may pad out any remaining bytes in the znode buffer, eg:
1874 * |<---------------------- dnode_phys (512) ------------------------>|
1875 * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->|
1876 * |<---- znode (264) ---->|<---- data (56) ---->|
1878 * At present, we only use this space to store symbolic links.
1883 * In-core vdev representation.
1887 typedef int vdev_phys_read_t(struct vdev *, void *, off_t, void *, size_t);
1888 typedef int vdev_phys_write_t(struct vdev *, off_t, void *, size_t);
1889 typedef int vdev_read_t(struct vdev *, const blkptr_t *, void *, off_t, size_t);
1891 typedef STAILQ_HEAD(vdev_list, vdev) vdev_list_t;
1893 typedef struct vdev_indirect_mapping_entry_phys {
1895 * Decode with DVA_MAPPING_* macros.
1897 * the source offset (low 63 bits)
1898 * the one-bit "mark", used for garbage collection (by zdb)
1903 * Note: the DVA's asize is 24 bits, and can thus store ranges
1907 } vdev_indirect_mapping_entry_phys_t;
1909 #define DVA_MAPPING_GET_SRC_OFFSET(vimep) \
1910 BF64_GET_SB((vimep)->vimep_src, 0, 63, SPA_MINBLOCKSHIFT, 0)
1911 #define DVA_MAPPING_SET_SRC_OFFSET(vimep, x) \
1912 BF64_SET_SB((vimep)->vimep_src, 0, 63, SPA_MINBLOCKSHIFT, 0, x)
1915 * This is stored in the bonus buffer of the mapping object, see comment of
1916 * vdev_indirect_config for more details.
1918 typedef struct vdev_indirect_mapping_phys {
1919 uint64_t vimp_max_offset;
1920 uint64_t vimp_bytes_mapped;
1921 uint64_t vimp_num_entries; /* number of v_i_m_entry_phys_t's */
1924 * For each entry in the mapping object, this object contains an
1925 * entry representing the number of bytes of that mapping entry
1926 * that were no longer in use by the pool at the time this indirect
1927 * vdev was last condensed.
1929 uint64_t vimp_counts_object;
1930 } vdev_indirect_mapping_phys_t;
1932 #define VDEV_INDIRECT_MAPPING_SIZE_V0 (3 * sizeof (uint64_t))
1934 typedef struct vdev_indirect_mapping {
1935 uint64_t vim_object;
1936 boolean_t vim_havecounts;
1938 /* vim_entries segment offset currently in memory. */
1939 uint64_t vim_entry_offset;
1940 /* vim_entries segment size. */
1941 size_t vim_num_entries;
1943 /* Needed by dnode_read() */
1944 const void *vim_spa;
1945 dnode_phys_t *vim_dn;
1948 * An ordered array of mapping entries, sorted by source offset.
1949 * Note that vim_entries is needed during a removal (and contains
1950 * mappings that have been synced to disk so far) to handle frees
1951 * from the removing device.
1953 vdev_indirect_mapping_entry_phys_t *vim_entries;
1954 objset_phys_t *vim_objset;
1955 vdev_indirect_mapping_phys_t *vim_phys;
1956 } vdev_indirect_mapping_t;
1959 * On-disk indirect vdev state.
1961 * An indirect vdev is described exclusively in the MOS config of a pool.
1962 * The config for an indirect vdev includes several fields, which are
1963 * accessed in memory by a vdev_indirect_config_t.
1965 typedef struct vdev_indirect_config {
1967 * Object (in MOS) which contains the indirect mapping. This object
1968 * contains an array of vdev_indirect_mapping_entry_phys_t ordered by
1969 * vimep_src. The bonus buffer for this object is a
1970 * vdev_indirect_mapping_phys_t. This object is allocated when a vdev
1971 * removal is initiated.
1973 * Note that this object can be empty if none of the data on the vdev
1974 * has been copied yet.
1976 uint64_t vic_mapping_object;
1979 * Object (in MOS) which contains the birth times for the mapping
1980 * entries. This object contains an array of
1981 * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus
1982 * buffer for this object is a vdev_indirect_birth_phys_t. This object
1983 * is allocated when a vdev removal is initiated.
1985 * Note that this object can be empty if none of the vdev has yet been
1988 uint64_t vic_births_object;
1991 * This is the vdev ID which was removed previous to this vdev, or
1992 * UINT64_MAX if there are no previously removed vdevs.
1994 uint64_t vic_prev_indirect_vdev;
1995 } vdev_indirect_config_t;
1997 typedef struct vdev {
1998 STAILQ_ENTRY(vdev) v_childlink; /* link in parent's child list */
1999 STAILQ_ENTRY(vdev) v_alllink; /* link in global vdev list */
2000 vdev_list_t v_children; /* children of this vdev */
2001 const char *v_name; /* vdev name */
2002 uint64_t v_guid; /* vdev guid */
2003 uint64_t v_id; /* index in parent */
2004 uint64_t v_psize; /* physical device capacity */
2005 int v_ashift; /* offset to block shift */
2006 int v_nparity; /* # parity for raidz */
2007 struct vdev *v_top; /* parent vdev */
2008 size_t v_nchildren; /* # children */
2009 vdev_state_t v_state; /* current state */
2010 vdev_phys_read_t *v_phys_read; /* read from raw leaf vdev */
2011 vdev_phys_write_t *v_phys_write; /* write to raw leaf vdev */
2012 vdev_read_t *v_read; /* read from vdev */
2013 void *v_priv; /* data for read/write function */
2015 struct spa *v_spa; /* link to spa */
2017 * Values stored in the config for an indirect or removing vdev.
2019 vdev_indirect_config_t vdev_indirect_config;
2020 vdev_indirect_mapping_t *v_mapping;
2024 * In-core pool representation.
2026 typedef STAILQ_HEAD(spa_list, spa) spa_list_t;
2028 typedef struct spa {
2029 STAILQ_ENTRY(spa) spa_link; /* link in global pool list */
2030 char *spa_name; /* pool name */
2031 uint64_t spa_guid; /* pool guid */
2032 uint64_t spa_txg; /* most recent transaction */
2033 struct uberblock *spa_uberblock; /* best uberblock so far */
2034 vdev_t *spa_root_vdev; /* toplevel vdev container */
2035 objset_phys_t *spa_mos; /* MOS for this pool */
2036 zio_cksum_salt_t spa_cksum_salt; /* secret salt for cksum */
2037 void *spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS];
2038 boolean_t spa_with_log; /* this pool has log */
2040 struct uberblock spa_uberblock_master; /* best uberblock so far */
2041 objset_phys_t spa_mos_master; /* MOS for this pool */
2042 struct uberblock spa_uberblock_checkpoint; /* checkpoint uberblock */
2043 objset_phys_t spa_mos_checkpoint; /* Checkpoint MOS */
2044 void *spa_bootenv; /* bootenv from pool label */
2047 /* IO related arguments. */
2048 typedef struct zio {
2055 /* Stuff for the vdev stack */
2062 extern void decode_embedded_bp_compressed(const blkptr_t *, void *);
2064 #endif /* _ZFSIMPL_H_ */