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) 2013 by Delphix. All rights reserved.
62 #define MAXNAMELEN 256
67 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */
70 * Macros for various sorts of alignment and rounding when the alignment
71 * is known to be a power of 2.
73 #define P2ALIGN(x, align) ((x) & -(align))
74 #define P2PHASE(x, align) ((x) & ((align) - 1))
75 #define P2NPHASE(x, align) (-(x) & ((align) - 1))
76 #define P2ROUNDUP(x, align) (-(-(x) & -(align)))
77 #define P2END(x, align) (-(~(x) & -(align)))
78 #define P2PHASEUP(x, align, phase) ((phase) - (((phase) - (x)) & -(align)))
79 #define P2BOUNDARY(off, len, align) (((off) ^ ((off) + (len) - 1)) > (align) - 1)
82 * General-purpose 32-bit and 64-bit bitfield encodings.
84 #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len))
85 #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len))
86 #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low))
87 #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low))
89 #define BF32_GET(x, low, len) BF32_DECODE(x, low, len)
90 #define BF64_GET(x, low, len) BF64_DECODE(x, low, len)
92 #define BF32_SET(x, low, len, val) \
93 ((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len))
94 #define BF64_SET(x, low, len, val) \
95 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len))
97 #define BF32_GET_SB(x, low, len, shift, bias) \
98 ((BF32_GET(x, low, len) + (bias)) << (shift))
99 #define BF64_GET_SB(x, low, len, shift, bias) \
100 ((BF64_GET(x, low, len) + (bias)) << (shift))
102 #define BF32_SET_SB(x, low, len, shift, bias, val) \
103 BF32_SET(x, low, len, ((val) >> (shift)) - (bias))
104 #define BF64_SET_SB(x, low, len, shift, bias, val) \
105 BF64_SET(x, low, len, ((val) >> (shift)) - (bias))
108 * Macros to reverse byte order
110 #define BSWAP_8(x) ((x) & 0xff)
111 #define BSWAP_16(x) ((BSWAP_8(x) << 8) | BSWAP_8((x) >> 8))
112 #define BSWAP_32(x) ((BSWAP_16(x) << 16) | BSWAP_16((x) >> 16))
113 #define BSWAP_64(x) ((BSWAP_32(x) << 32) | BSWAP_32((x) >> 32))
116 * Note: the boot loader can't actually read blocks larger than 128KB,
117 * due to lack of memory. Therefore its SPA_MAXBLOCKSIZE is still 128KB.
119 #define SPA_MINBLOCKSHIFT 9
120 #define SPA_MAXBLOCKSHIFT 17
121 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT)
122 #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT)
125 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
126 * The ASIZE encoding should be at least 64 times larger (6 more bits)
127 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
128 * overhead, three DVAs per bp, plus one more bit in case we do anything
129 * else that expands the ASIZE.
131 #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */
132 #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */
133 #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */
136 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
137 * The members of the dva_t should be considered opaque outside the SPA.
140 uint64_t dva_word[2];
144 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
146 typedef struct zio_cksum {
151 * Each block is described by its DVAs, time of birth, checksum, etc.
152 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
154 * 64 56 48 40 32 24 16 8 0
155 * +-------+-------+-------+-------+-------+-------+-------+-------+
156 * 0 | vdev1 | GRID | ASIZE |
157 * +-------+-------+-------+-------+-------+-------+-------+-------+
159 * +-------+-------+-------+-------+-------+-------+-------+-------+
160 * 2 | vdev2 | GRID | ASIZE |
161 * +-------+-------+-------+-------+-------+-------+-------+-------+
163 * +-------+-------+-------+-------+-------+-------+-------+-------+
164 * 4 | vdev3 | GRID | ASIZE |
165 * +-------+-------+-------+-------+-------+-------+-------+-------+
167 * +-------+-------+-------+-------+-------+-------+-------+-------+
168 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE |
169 * +-------+-------+-------+-------+-------+-------+-------+-------+
171 * +-------+-------+-------+-------+-------+-------+-------+-------+
173 * +-------+-------+-------+-------+-------+-------+-------+-------+
174 * 9 | physical birth txg |
175 * +-------+-------+-------+-------+-------+-------+-------+-------+
176 * a | logical birth txg |
177 * +-------+-------+-------+-------+-------+-------+-------+-------+
179 * +-------+-------+-------+-------+-------+-------+-------+-------+
181 * +-------+-------+-------+-------+-------+-------+-------+-------+
183 * +-------+-------+-------+-------+-------+-------+-------+-------+
185 * +-------+-------+-------+-------+-------+-------+-------+-------+
187 * +-------+-------+-------+-------+-------+-------+-------+-------+
191 * vdev virtual device ID
192 * offset offset into virtual device
194 * PSIZE physical size (after compression)
195 * ASIZE allocated size (including RAID-Z parity and gang block headers)
196 * GRID RAID-Z layout information (reserved for future use)
197 * cksum checksum function
198 * comp compression function
199 * G gang block indicator
200 * B byteorder (endianness)
202 * X encryption (on version 30, which is not supported)
203 * E blkptr_t contains embedded data (see below)
204 * lvl level of indirection
205 * type DMU object type
206 * phys birth txg of block allocation; zero if same as logical birth txg
207 * log. birth transaction group in which the block was logically born
208 * fill count number of non-zero blocks under this bp
209 * checksum[4] 256-bit checksum of the data this bp describes
213 * "Embedded" blkptr_t's don't actually point to a block, instead they
214 * have a data payload embedded in the blkptr_t itself. See the comment
215 * in blkptr.c for more details.
217 * The blkptr_t is laid out as follows:
219 * 64 56 48 40 32 24 16 8 0
220 * +-------+-------+-------+-------+-------+-------+-------+-------+
227 * +-------+-------+-------+-------+-------+-------+-------+-------+
228 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE |
229 * +-------+-------+-------+-------+-------+-------+-------+-------+
233 * +-------+-------+-------+-------+-------+-------+-------+-------+
234 * a | logical birth txg |
235 * +-------+-------+-------+-------+-------+-------+-------+-------+
241 * +-------+-------+-------+-------+-------+-------+-------+-------+
245 * payload contains the embedded data
246 * B (byteorder) byteorder (endianness)
247 * D (dedup) padding (set to zero)
248 * X encryption (set to zero; see above)
249 * E (embedded) set to one
250 * lvl indirection level
251 * type DMU object type
252 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*)
253 * comp compression function of payload
254 * PSIZE size of payload after compression, in bytes
255 * LSIZE logical size of payload, in bytes
256 * note that 25 bits is enough to store the largest
257 * "normal" BP's LSIZE (2^16 * 2^9) in bytes
258 * log. birth transaction group in which the block was logically born
260 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
261 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
262 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
263 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
264 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must
265 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before
266 * other macros, as they assert that they are only used on BP's of the correct
270 #define BPE_GET_ETYPE(bp) \
271 (ASSERT(BP_IS_EMBEDDED(bp)), \
272 BF64_GET((bp)->blk_prop, 40, 8))
273 #define BPE_SET_ETYPE(bp, t) do { \
274 ASSERT(BP_IS_EMBEDDED(bp)); \
275 BF64_SET((bp)->blk_prop, 40, 8, t); \
276 _NOTE(CONSTCOND) } while (0)
278 #define BPE_GET_LSIZE(bp) \
279 (ASSERT(BP_IS_EMBEDDED(bp)), \
280 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
281 #define BPE_SET_LSIZE(bp, x) do { \
282 ASSERT(BP_IS_EMBEDDED(bp)); \
283 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
284 _NOTE(CONSTCOND) } while (0)
286 #define BPE_GET_PSIZE(bp) \
287 (ASSERT(BP_IS_EMBEDDED(bp)), \
288 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
289 #define BPE_SET_PSIZE(bp, x) do { \
290 ASSERT(BP_IS_EMBEDDED(bp)); \
291 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
292 _NOTE(CONSTCOND) } while (0)
294 typedef enum bp_embedded_type {
295 BP_EMBEDDED_TYPE_DATA,
296 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */
297 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED
298 } bp_embedded_type_t;
300 #define BPE_NUM_WORDS 14
301 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
302 #define BPE_IS_PAYLOADWORD(bp, wp) \
303 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
305 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */
306 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */
308 typedef struct blkptr {
309 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
310 uint64_t blk_prop; /* size, compression, type, etc */
311 uint64_t blk_pad[2]; /* Extra space for the future */
312 uint64_t blk_phys_birth; /* txg when block was allocated */
313 uint64_t blk_birth; /* transaction group at birth */
314 uint64_t blk_fill; /* fill count */
315 zio_cksum_t blk_cksum; /* 256-bit checksum */
319 * Macros to get and set fields in a bp or DVA.
321 #define DVA_GET_ASIZE(dva) \
322 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
323 #define DVA_SET_ASIZE(dva, x) \
324 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
325 SPA_MINBLOCKSHIFT, 0, x)
327 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8)
328 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x)
330 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32)
331 #define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x)
333 #define DVA_GET_OFFSET(dva) \
334 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
335 #define DVA_SET_OFFSET(dva, x) \
336 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
338 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1)
339 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x)
341 #define BP_GET_LSIZE(bp) \
342 (BP_IS_EMBEDDED(bp) ? \
343 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
344 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
345 #define BP_SET_LSIZE(bp, x) do { \
346 ASSERT(!BP_IS_EMBEDDED(bp)); \
347 BF64_SET_SB((bp)->blk_prop, \
348 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
349 _NOTE(CONSTCOND) } while (0)
351 #define BP_GET_PSIZE(bp) \
352 BF64_GET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)
353 #define BP_SET_PSIZE(bp, x) \
354 BF64_SET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x)
356 #define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7)
357 #define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x)
359 #define BP_GET_CHECKSUM(bp) BF64_GET((bp)->blk_prop, 40, 8)
360 #define BP_SET_CHECKSUM(bp, x) BF64_SET((bp)->blk_prop, 40, 8, x)
362 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8)
363 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x)
365 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5)
366 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x)
368 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1)
370 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1)
371 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x)
373 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1)
374 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x)
376 #define BP_PHYSICAL_BIRTH(bp) \
377 ((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
379 #define BP_GET_ASIZE(bp) \
380 (DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
381 DVA_GET_ASIZE(&(bp)->blk_dva[2]))
383 #define BP_GET_UCSIZE(bp) \
384 ((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \
385 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp));
387 #define BP_GET_NDVAS(bp) \
388 (!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
389 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
390 !!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
392 #define DVA_EQUAL(dva1, dva2) \
393 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
394 (dva1)->dva_word[0] == (dva2)->dva_word[0])
396 #define ZIO_CHECKSUM_EQUAL(zc1, zc2) \
397 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
398 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \
399 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \
400 ((zc1).zc_word[3] - (zc2).zc_word[3])))
403 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0)
405 #define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \
407 (zcp)->zc_word[0] = w0; \
408 (zcp)->zc_word[1] = w1; \
409 (zcp)->zc_word[2] = w2; \
410 (zcp)->zc_word[3] = w3; \
413 #define BP_IDENTITY(bp) (&(bp)->blk_dva[0])
414 #define BP_IS_GANG(bp) DVA_GET_GANG(BP_IDENTITY(bp))
415 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \
416 (dva)->dva_word[1] == 0ULL)
417 #define BP_IS_HOLE(bp) DVA_IS_EMPTY(BP_IDENTITY(bp))
418 #define BP_IS_OLDER(bp, txg) (!BP_IS_HOLE(bp) && (bp)->blk_birth < (txg))
420 #define BP_ZERO(bp) \
422 (bp)->blk_dva[0].dva_word[0] = 0; \
423 (bp)->blk_dva[0].dva_word[1] = 0; \
424 (bp)->blk_dva[1].dva_word[0] = 0; \
425 (bp)->blk_dva[1].dva_word[1] = 0; \
426 (bp)->blk_dva[2].dva_word[0] = 0; \
427 (bp)->blk_dva[2].dva_word[1] = 0; \
428 (bp)->blk_prop = 0; \
429 (bp)->blk_pad[0] = 0; \
430 (bp)->blk_pad[1] = 0; \
431 (bp)->blk_phys_birth = 0; \
432 (bp)->blk_birth = 0; \
433 (bp)->blk_fill = 0; \
434 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \
437 #define BPE_NUM_WORDS 14
438 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
439 #define BPE_IS_PAYLOADWORD(bp, wp) \
440 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
445 #define ZEC_MAGIC 0x210da7ab10c7a11ULL
447 typedef struct zio_eck {
448 uint64_t zec_magic; /* for validation, endianness */
449 zio_cksum_t zec_cksum; /* 256-bit checksum */
453 * Gang block headers are self-checksumming and contain an array
456 #define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE
457 #define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \
458 sizeof (zio_eck_t)) / sizeof (blkptr_t))
459 #define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \
460 sizeof (zio_eck_t) - \
461 (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\
464 typedef struct zio_gbh {
465 blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS];
466 uint64_t zg_filler[SPA_GBH_FILLER];
470 #define VDEV_RAIDZ_MAXPARITY 3
472 #define VDEV_PAD_SIZE (8 << 10)
473 /* 2 padding areas (vl_pad1 and vl_pad2) to skip */
474 #define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2
475 #define VDEV_PHYS_SIZE (112 << 10)
476 #define VDEV_UBERBLOCK_RING (128 << 10)
478 #define VDEV_UBERBLOCK_SHIFT(vd) \
479 MAX((vd)->v_top->v_ashift, UBERBLOCK_SHIFT)
480 #define VDEV_UBERBLOCK_COUNT(vd) \
481 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
482 #define VDEV_UBERBLOCK_OFFSET(vd, n) \
483 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
484 #define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd))
486 typedef struct vdev_phys {
487 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)];
491 typedef struct vdev_label {
492 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */
493 char vl_pad2[VDEV_PAD_SIZE]; /* 8K */
494 vdev_phys_t vl_vdev_phys; /* 112K */
495 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */
496 } vdev_label_t; /* 256K total */
501 #define VDD_METASLAB 0x01
505 * Size and offset of embedded boot loader region on each label.
506 * The total size of the first two labels plus the boot area is 4MB.
508 #define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t))
509 #define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */
512 * Size of label regions at the start and end of each leaf device.
514 #define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
515 #define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t))
516 #define VDEV_LABELS 4
519 ZIO_CHECKSUM_INHERIT = 0,
523 ZIO_CHECKSUM_GANG_HEADER,
525 ZIO_CHECKSUM_FLETCHER_2,
526 ZIO_CHECKSUM_FLETCHER_4,
529 ZIO_CHECKSUM_FUNCTIONS
532 #define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4
533 #define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON
536 ZIO_COMPRESS_INHERIT = 0,
552 ZIO_COMPRESS_FUNCTIONS
555 #define ZIO_COMPRESS_ON_VALUE ZIO_COMPRESS_LZJB
556 #define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF
558 /* nvlist pack encoding */
559 #define NV_ENCODE_NATIVE 0
560 #define NV_ENCODE_XDR 1
563 DATA_TYPE_UNKNOWN = 0,
573 DATA_TYPE_BYTE_ARRAY,
574 DATA_TYPE_INT16_ARRAY,
575 DATA_TYPE_UINT16_ARRAY,
576 DATA_TYPE_INT32_ARRAY,
577 DATA_TYPE_UINT32_ARRAY,
578 DATA_TYPE_INT64_ARRAY,
579 DATA_TYPE_UINT64_ARRAY,
580 DATA_TYPE_STRING_ARRAY,
583 DATA_TYPE_NVLIST_ARRAY,
584 DATA_TYPE_BOOLEAN_VALUE,
587 DATA_TYPE_BOOLEAN_ARRAY,
588 DATA_TYPE_INT8_ARRAY,
589 DATA_TYPE_UINT8_ARRAY
593 * On-disk version number.
595 #define SPA_VERSION_1 1ULL
596 #define SPA_VERSION_2 2ULL
597 #define SPA_VERSION_3 3ULL
598 #define SPA_VERSION_4 4ULL
599 #define SPA_VERSION_5 5ULL
600 #define SPA_VERSION_6 6ULL
601 #define SPA_VERSION_7 7ULL
602 #define SPA_VERSION_8 8ULL
603 #define SPA_VERSION_9 9ULL
604 #define SPA_VERSION_10 10ULL
605 #define SPA_VERSION_11 11ULL
606 #define SPA_VERSION_12 12ULL
607 #define SPA_VERSION_13 13ULL
608 #define SPA_VERSION_14 14ULL
609 #define SPA_VERSION_15 15ULL
610 #define SPA_VERSION_16 16ULL
611 #define SPA_VERSION_17 17ULL
612 #define SPA_VERSION_18 18ULL
613 #define SPA_VERSION_19 19ULL
614 #define SPA_VERSION_20 20ULL
615 #define SPA_VERSION_21 21ULL
616 #define SPA_VERSION_22 22ULL
617 #define SPA_VERSION_23 23ULL
618 #define SPA_VERSION_24 24ULL
619 #define SPA_VERSION_25 25ULL
620 #define SPA_VERSION_26 26ULL
621 #define SPA_VERSION_27 27ULL
622 #define SPA_VERSION_28 28ULL
623 #define SPA_VERSION_5000 5000ULL
626 * When bumping up SPA_VERSION, make sure GRUB ZFS understands the on-disk
627 * format change. Go to usr/src/grub/grub-0.97/stage2/{zfs-include/, fsys_zfs*},
628 * and do the appropriate changes. Also bump the version number in
629 * usr/src/grub/capability.
631 #define SPA_VERSION SPA_VERSION_5000
632 #define SPA_VERSION_STRING "5000"
635 * Symbolic names for the changes that caused a SPA_VERSION switch.
636 * Used in the code when checking for presence or absence of a feature.
637 * Feel free to define multiple symbolic names for each version if there
638 * were multiple changes to on-disk structures during that version.
640 * NOTE: When checking the current SPA_VERSION in your code, be sure
641 * to use spa_version() since it reports the version of the
642 * last synced uberblock. Checking the in-flight version can
643 * be dangerous in some cases.
645 #define SPA_VERSION_INITIAL SPA_VERSION_1
646 #define SPA_VERSION_DITTO_BLOCKS SPA_VERSION_2
647 #define SPA_VERSION_SPARES SPA_VERSION_3
648 #define SPA_VERSION_RAID6 SPA_VERSION_3
649 #define SPA_VERSION_BPLIST_ACCOUNT SPA_VERSION_3
650 #define SPA_VERSION_RAIDZ_DEFLATE SPA_VERSION_3
651 #define SPA_VERSION_DNODE_BYTES SPA_VERSION_3
652 #define SPA_VERSION_ZPOOL_HISTORY SPA_VERSION_4
653 #define SPA_VERSION_GZIP_COMPRESSION SPA_VERSION_5
654 #define SPA_VERSION_BOOTFS SPA_VERSION_6
655 #define SPA_VERSION_SLOGS SPA_VERSION_7
656 #define SPA_VERSION_DELEGATED_PERMS SPA_VERSION_8
657 #define SPA_VERSION_FUID SPA_VERSION_9
658 #define SPA_VERSION_REFRESERVATION SPA_VERSION_9
659 #define SPA_VERSION_REFQUOTA SPA_VERSION_9
660 #define SPA_VERSION_UNIQUE_ACCURATE SPA_VERSION_9
661 #define SPA_VERSION_L2CACHE SPA_VERSION_10
662 #define SPA_VERSION_NEXT_CLONES SPA_VERSION_11
663 #define SPA_VERSION_ORIGIN SPA_VERSION_11
664 #define SPA_VERSION_DSL_SCRUB SPA_VERSION_11
665 #define SPA_VERSION_SNAP_PROPS SPA_VERSION_12
666 #define SPA_VERSION_USED_BREAKDOWN SPA_VERSION_13
667 #define SPA_VERSION_PASSTHROUGH_X SPA_VERSION_14
668 #define SPA_VERSION_USERSPACE SPA_VERSION_15
669 #define SPA_VERSION_STMF_PROP SPA_VERSION_16
670 #define SPA_VERSION_RAIDZ3 SPA_VERSION_17
671 #define SPA_VERSION_USERREFS SPA_VERSION_18
672 #define SPA_VERSION_HOLES SPA_VERSION_19
673 #define SPA_VERSION_ZLE_COMPRESSION SPA_VERSION_20
674 #define SPA_VERSION_DEDUP SPA_VERSION_21
675 #define SPA_VERSION_RECVD_PROPS SPA_VERSION_22
676 #define SPA_VERSION_SLIM_ZIL SPA_VERSION_23
677 #define SPA_VERSION_SA SPA_VERSION_24
678 #define SPA_VERSION_SCAN SPA_VERSION_25
679 #define SPA_VERSION_DIR_CLONES SPA_VERSION_26
680 #define SPA_VERSION_DEADLISTS SPA_VERSION_26
681 #define SPA_VERSION_FAST_SNAP SPA_VERSION_27
682 #define SPA_VERSION_MULTI_REPLACE SPA_VERSION_28
683 #define SPA_VERSION_BEFORE_FEATURES SPA_VERSION_28
684 #define SPA_VERSION_FEATURES SPA_VERSION_5000
686 #define SPA_VERSION_IS_SUPPORTED(v) \
687 (((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \
688 ((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION))
691 * The following are configuration names used in the nvlist describing a pool's
694 #define ZPOOL_CONFIG_VERSION "version"
695 #define ZPOOL_CONFIG_POOL_NAME "name"
696 #define ZPOOL_CONFIG_POOL_STATE "state"
697 #define ZPOOL_CONFIG_POOL_TXG "txg"
698 #define ZPOOL_CONFIG_POOL_GUID "pool_guid"
699 #define ZPOOL_CONFIG_CREATE_TXG "create_txg"
700 #define ZPOOL_CONFIG_TOP_GUID "top_guid"
701 #define ZPOOL_CONFIG_VDEV_TREE "vdev_tree"
702 #define ZPOOL_CONFIG_TYPE "type"
703 #define ZPOOL_CONFIG_CHILDREN "children"
704 #define ZPOOL_CONFIG_ID "id"
705 #define ZPOOL_CONFIG_GUID "guid"
706 #define ZPOOL_CONFIG_PATH "path"
707 #define ZPOOL_CONFIG_DEVID "devid"
708 #define ZPOOL_CONFIG_METASLAB_ARRAY "metaslab_array"
709 #define ZPOOL_CONFIG_METASLAB_SHIFT "metaslab_shift"
710 #define ZPOOL_CONFIG_ASHIFT "ashift"
711 #define ZPOOL_CONFIG_ASIZE "asize"
712 #define ZPOOL_CONFIG_DTL "DTL"
713 #define ZPOOL_CONFIG_STATS "stats"
714 #define ZPOOL_CONFIG_WHOLE_DISK "whole_disk"
715 #define ZPOOL_CONFIG_ERRCOUNT "error_count"
716 #define ZPOOL_CONFIG_NOT_PRESENT "not_present"
717 #define ZPOOL_CONFIG_SPARES "spares"
718 #define ZPOOL_CONFIG_IS_SPARE "is_spare"
719 #define ZPOOL_CONFIG_NPARITY "nparity"
720 #define ZPOOL_CONFIG_HOSTID "hostid"
721 #define ZPOOL_CONFIG_HOSTNAME "hostname"
722 #define ZPOOL_CONFIG_IS_LOG "is_log"
723 #define ZPOOL_CONFIG_TIMESTAMP "timestamp" /* not stored on disk */
724 #define ZPOOL_CONFIG_FEATURES_FOR_READ "features_for_read"
727 * The persistent vdev state is stored as separate values rather than a single
728 * 'vdev_state' entry. This is because a device can be in multiple states, such
729 * as offline and degraded.
731 #define ZPOOL_CONFIG_OFFLINE "offline"
732 #define ZPOOL_CONFIG_FAULTED "faulted"
733 #define ZPOOL_CONFIG_DEGRADED "degraded"
734 #define ZPOOL_CONFIG_REMOVED "removed"
735 #define ZPOOL_CONFIG_FRU "fru"
736 #define ZPOOL_CONFIG_AUX_STATE "aux_state"
738 #define VDEV_TYPE_ROOT "root"
739 #define VDEV_TYPE_MIRROR "mirror"
740 #define VDEV_TYPE_REPLACING "replacing"
741 #define VDEV_TYPE_RAIDZ "raidz"
742 #define VDEV_TYPE_DISK "disk"
743 #define VDEV_TYPE_FILE "file"
744 #define VDEV_TYPE_MISSING "missing"
745 #define VDEV_TYPE_HOLE "hole"
746 #define VDEV_TYPE_SPARE "spare"
747 #define VDEV_TYPE_LOG "log"
748 #define VDEV_TYPE_L2CACHE "l2cache"
751 * This is needed in userland to report the minimum necessary device size.
753 #define SPA_MINDEVSIZE (64ULL << 20)
756 * The location of the pool configuration repository, shared between kernel and
759 #define ZPOOL_CACHE "/boot/zfs/zpool.cache"
762 * vdev states are ordered from least to most healthy.
763 * A vdev that's CANT_OPEN or below is considered unusable.
765 typedef enum vdev_state {
766 VDEV_STATE_UNKNOWN = 0, /* Uninitialized vdev */
767 VDEV_STATE_CLOSED, /* Not currently open */
768 VDEV_STATE_OFFLINE, /* Not allowed to open */
769 VDEV_STATE_REMOVED, /* Explicitly removed from system */
770 VDEV_STATE_CANT_OPEN, /* Tried to open, but failed */
771 VDEV_STATE_FAULTED, /* External request to fault device */
772 VDEV_STATE_DEGRADED, /* Replicated vdev with unhealthy kids */
773 VDEV_STATE_HEALTHY /* Presumed good */
777 * vdev aux states. When a vdev is in the CANT_OPEN state, the aux field
778 * of the vdev stats structure uses these constants to distinguish why.
780 typedef enum vdev_aux {
781 VDEV_AUX_NONE, /* no error */
782 VDEV_AUX_OPEN_FAILED, /* ldi_open_*() or vn_open() failed */
783 VDEV_AUX_CORRUPT_DATA, /* bad label or disk contents */
784 VDEV_AUX_NO_REPLICAS, /* insufficient number of replicas */
785 VDEV_AUX_BAD_GUID_SUM, /* vdev guid sum doesn't match */
786 VDEV_AUX_TOO_SMALL, /* vdev size is too small */
787 VDEV_AUX_BAD_LABEL, /* the label is OK but invalid */
788 VDEV_AUX_VERSION_NEWER, /* on-disk version is too new */
789 VDEV_AUX_VERSION_OLDER, /* on-disk version is too old */
790 VDEV_AUX_SPARED /* hot spare used in another pool */
794 * pool state. The following states are written to disk as part of the normal
795 * SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE. The remaining states are
796 * software abstractions used at various levels to communicate pool state.
798 typedef enum pool_state {
799 POOL_STATE_ACTIVE = 0, /* In active use */
800 POOL_STATE_EXPORTED, /* Explicitly exported */
801 POOL_STATE_DESTROYED, /* Explicitly destroyed */
802 POOL_STATE_SPARE, /* Reserved for hot spare use */
803 POOL_STATE_UNINITIALIZED, /* Internal spa_t state */
804 POOL_STATE_UNAVAIL, /* Internal libzfs state */
805 POOL_STATE_POTENTIALLY_ACTIVE /* Internal libzfs state */
809 * The uberblock version is incremented whenever an incompatible on-disk
810 * format change is made to the SPA, DMU, or ZAP.
812 * Note: the first two fields should never be moved. When a storage pool
813 * is opened, the uberblock must be read off the disk before the version
814 * can be checked. If the ub_version field is moved, we may not detect
815 * version mismatch. If the ub_magic field is moved, applications that
816 * expect the magic number in the first word won't work.
818 #define UBERBLOCK_MAGIC 0x00bab10c /* oo-ba-bloc! */
819 #define UBERBLOCK_SHIFT 10 /* up to 1K */
822 uint64_t ub_magic; /* UBERBLOCK_MAGIC */
823 uint64_t ub_version; /* SPA_VERSION */
824 uint64_t ub_txg; /* txg of last sync */
825 uint64_t ub_guid_sum; /* sum of all vdev guids */
826 uint64_t ub_timestamp; /* UTC time of last sync */
827 blkptr_t ub_rootbp; /* MOS objset_phys_t */
833 #define DNODE_MUST_BE_ALLOCATED 1
834 #define DNODE_MUST_BE_FREE 2
839 #define DNODE_SHIFT 9 /* 512 bytes */
840 #define DN_MIN_INDBLKSHIFT 12 /* 4k */
841 #define DN_MAX_INDBLKSHIFT 14 /* 16k */
842 #define DNODE_BLOCK_SHIFT 14 /* 16k */
843 #define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */
844 #define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */
845 #define DN_MAX_OFFSET_SHIFT 64 /* 2^64 bytes in a dnode */
850 #define DNODE_SIZE (1 << DNODE_SHIFT)
851 #define DN_MAX_NBLKPTR ((DNODE_SIZE - DNODE_CORE_SIZE) >> SPA_BLKPTRSHIFT)
852 #define DN_MAX_BONUSLEN (DNODE_SIZE - DNODE_CORE_SIZE - (1 << SPA_BLKPTRSHIFT))
853 #define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT)
855 #define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT)
856 #define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT)
857 #define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT)
859 /* The +2 here is a cheesy way to round up */
860 #define DN_MAX_LEVELS (2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \
861 (DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT)))
863 #define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \
864 (((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t))))
866 #define DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \
867 (dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT)
869 #define EPB(blkshift, typeshift) (1 << (blkshift - typeshift))
871 /* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */
872 #define DNODE_FLAG_USED_BYTES (1<<0)
873 #define DNODE_FLAG_USERUSED_ACCOUNTED (1<<1)
875 /* Does dnode have a SA spill blkptr in bonus? */
876 #define DNODE_FLAG_SPILL_BLKPTR (1<<2)
878 typedef struct dnode_phys {
879 uint8_t dn_type; /* dmu_object_type_t */
880 uint8_t dn_indblkshift; /* ln2(indirect block size) */
881 uint8_t dn_nlevels; /* 1=dn_blkptr->data blocks */
882 uint8_t dn_nblkptr; /* length of dn_blkptr */
883 uint8_t dn_bonustype; /* type of data in bonus buffer */
884 uint8_t dn_checksum; /* ZIO_CHECKSUM type */
885 uint8_t dn_compress; /* ZIO_COMPRESS type */
886 uint8_t dn_flags; /* DNODE_FLAG_* */
887 uint16_t dn_datablkszsec; /* data block size in 512b sectors */
888 uint16_t dn_bonuslen; /* length of dn_bonus */
891 /* accounting is protected by dn_dirty_mtx */
892 uint64_t dn_maxblkid; /* largest allocated block ID */
893 uint64_t dn_used; /* bytes (or sectors) of disk space */
897 blkptr_t dn_blkptr[1];
898 uint8_t dn_bonus[DN_MAX_BONUSLEN - sizeof (blkptr_t)];
902 typedef enum dmu_object_type {
905 DMU_OT_OBJECT_DIRECTORY, /* ZAP */
906 DMU_OT_OBJECT_ARRAY, /* UINT64 */
907 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */
908 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */
909 DMU_OT_BPLIST, /* UINT64 */
910 DMU_OT_BPLIST_HDR, /* UINT64 */
912 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */
913 DMU_OT_SPACE_MAP, /* UINT64 */
915 DMU_OT_INTENT_LOG, /* UINT64 */
917 DMU_OT_DNODE, /* DNODE */
918 DMU_OT_OBJSET, /* OBJSET */
920 DMU_OT_DSL_DIR, /* UINT64 */
921 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */
922 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */
923 DMU_OT_DSL_PROPS, /* ZAP */
924 DMU_OT_DSL_DATASET, /* UINT64 */
926 DMU_OT_ZNODE, /* ZNODE */
927 DMU_OT_OLDACL, /* Old ACL */
928 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */
929 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */
930 DMU_OT_MASTER_NODE, /* ZAP */
931 DMU_OT_UNLINKED_SET, /* ZAP */
933 DMU_OT_ZVOL, /* UINT8 */
934 DMU_OT_ZVOL_PROP, /* ZAP */
935 /* other; for testing only! */
936 DMU_OT_PLAIN_OTHER, /* UINT8 */
937 DMU_OT_UINT64_OTHER, /* UINT64 */
938 DMU_OT_ZAP_OTHER, /* ZAP */
939 /* new object types: */
940 DMU_OT_ERROR_LOG, /* ZAP */
941 DMU_OT_SPA_HISTORY, /* UINT8 */
942 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */
943 DMU_OT_POOL_PROPS, /* ZAP */
944 DMU_OT_DSL_PERMS, /* ZAP */
945 DMU_OT_ACL, /* ACL */
946 DMU_OT_SYSACL, /* SYSACL */
947 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */
948 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */
949 DMU_OT_NEXT_CLONES, /* ZAP */
950 DMU_OT_SCAN_QUEUE, /* ZAP */
951 DMU_OT_USERGROUP_USED, /* ZAP */
952 DMU_OT_USERGROUP_QUOTA, /* ZAP */
953 DMU_OT_USERREFS, /* ZAP */
954 DMU_OT_DDT_ZAP, /* ZAP */
955 DMU_OT_DDT_STATS, /* ZAP */
956 DMU_OT_SA, /* System attr */
957 DMU_OT_SA_MASTER_NODE, /* ZAP */
958 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */
959 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */
960 DMU_OT_SCAN_XLATE, /* ZAP */
961 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */
965 typedef enum dmu_objset_type {
970 DMU_OST_OTHER, /* For testing only! */
971 DMU_OST_ANY, /* Be careful! */
976 * header for all bonus and spill buffers.
977 * The header has a fixed portion with a variable number
978 * of "lengths" depending on the number of variable sized
979 * attribues which are determined by the "layout number"
982 #define SA_MAGIC 0x2F505A /* ZFS SA */
983 typedef struct sa_hdr_phys {
985 uint16_t sa_layout_info; /* Encoded with hdrsize and layout number */
986 uint16_t sa_lengths[1]; /* optional sizes for variable length attrs */
987 /* ... Data follows the lengths. */
991 * sa_hdr_phys -> sa_layout_info
998 * Bits 0-10 are the layout number
999 * Bits 11-16 are the size of the header.
1000 * The hdrsize is the number * 8
1003 * hdrsz of 1 ==> 8 byte header
1004 * 2 ==> 16 byte header
1008 #define SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10)
1009 #define SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 16, 3, 0)
1010 #define SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \
1012 BF32_SET_SB(x, 10, 6, 3, 0, size); \
1013 BF32_SET(x, 0, 10, num); \
1016 #define SA_MODE_OFFSET 0
1017 #define SA_SIZE_OFFSET 8
1018 #define SA_GEN_OFFSET 16
1019 #define SA_UID_OFFSET 24
1020 #define SA_GID_OFFSET 32
1021 #define SA_PARENT_OFFSET 40
1024 * Intent log header - this on disk structure holds fields to manage
1025 * the log. All fields are 64 bit to easily handle cross architectures.
1027 typedef struct zil_header {
1028 uint64_t zh_claim_txg; /* txg in which log blocks were claimed */
1029 uint64_t zh_replay_seq; /* highest replayed sequence number */
1030 blkptr_t zh_log; /* log chain */
1031 uint64_t zh_claim_seq; /* highest claimed sequence number */
1035 #define OBJSET_PHYS_SIZE 2048
1037 typedef struct objset_phys {
1038 dnode_phys_t os_meta_dnode;
1039 zil_header_t os_zil_header;
1042 char os_pad[OBJSET_PHYS_SIZE - sizeof (dnode_phys_t)*3 -
1043 sizeof (zil_header_t) - sizeof (uint64_t)*2];
1044 dnode_phys_t os_userused_dnode;
1045 dnode_phys_t os_groupused_dnode;
1048 typedef struct dsl_dir_phys {
1049 uint64_t dd_creation_time; /* not actually used */
1050 uint64_t dd_head_dataset_obj;
1051 uint64_t dd_parent_obj;
1052 uint64_t dd_clone_parent_obj;
1053 uint64_t dd_child_dir_zapobj;
1055 * how much space our children are accounting for; for leaf
1056 * datasets, == physical space used by fs + snaps
1058 uint64_t dd_used_bytes;
1059 uint64_t dd_compressed_bytes;
1060 uint64_t dd_uncompressed_bytes;
1061 /* Administrative quota setting */
1063 /* Administrative reservation setting */
1064 uint64_t dd_reserved;
1065 uint64_t dd_props_zapobj;
1066 uint64_t dd_pad[21]; /* pad out to 256 bytes for good measure */
1069 typedef struct dsl_dataset_phys {
1070 uint64_t ds_dir_obj;
1071 uint64_t ds_prev_snap_obj;
1072 uint64_t ds_prev_snap_txg;
1073 uint64_t ds_next_snap_obj;
1074 uint64_t ds_snapnames_zapobj; /* zap obj of snaps; ==0 for snaps */
1075 uint64_t ds_num_children; /* clone/snap children; ==0 for head */
1076 uint64_t ds_creation_time; /* seconds since 1970 */
1077 uint64_t ds_creation_txg;
1078 uint64_t ds_deadlist_obj;
1079 uint64_t ds_used_bytes;
1080 uint64_t ds_compressed_bytes;
1081 uint64_t ds_uncompressed_bytes;
1082 uint64_t ds_unique_bytes; /* only relevant to snapshots */
1084 * The ds_fsid_guid is a 56-bit ID that can change to avoid
1085 * collisions. The ds_guid is a 64-bit ID that will never
1086 * change, so there is a small probability that it will collide.
1088 uint64_t ds_fsid_guid;
1092 uint64_t ds_pad[8]; /* pad out to 320 bytes for good measure */
1093 } dsl_dataset_phys_t;
1096 * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
1098 #define DMU_POOL_DIRECTORY_OBJECT 1
1099 #define DMU_POOL_CONFIG "config"
1100 #define DMU_POOL_ROOT_DATASET "root_dataset"
1101 #define DMU_POOL_SYNC_BPLIST "sync_bplist"
1102 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub"
1103 #define DMU_POOL_ERRLOG_LAST "errlog_last"
1104 #define DMU_POOL_SPARES "spares"
1105 #define DMU_POOL_DEFLATE "deflate"
1106 #define DMU_POOL_HISTORY "history"
1107 #define DMU_POOL_PROPS "pool_props"
1109 #define ZAP_MAGIC 0x2F52AB2ABULL
1111 #define FZAP_BLOCK_SHIFT(zap) ((zap)->zap_block_shift)
1113 #define ZAP_MAXCD (uint32_t)(-1)
1114 #define ZAP_HASHBITS 28
1115 #define MZAP_ENT_LEN 64
1116 #define MZAP_NAME_LEN (MZAP_ENT_LEN - 8 - 4 - 2)
1117 #define MZAP_MAX_BLKSHIFT SPA_MAXBLOCKSHIFT
1118 #define MZAP_MAX_BLKSZ (1 << MZAP_MAX_BLKSHIFT)
1120 typedef struct mzap_ent_phys {
1123 uint16_t mze_pad; /* in case we want to chain them someday */
1124 char mze_name[MZAP_NAME_LEN];
1127 typedef struct mzap_phys {
1128 uint64_t mz_block_type; /* ZBT_MICRO */
1131 mzap_ent_phys_t mz_chunk[1];
1132 /* actually variable size depending on block size */
1136 * The (fat) zap is stored in one object. It is an array of
1137 * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of:
1139 * ptrtbl fits in first block:
1140 * [zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ...
1142 * ptrtbl too big for first block:
1143 * [zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ...
1147 #define ZBT_LEAF ((1ULL << 63) + 0)
1148 #define ZBT_HEADER ((1ULL << 63) + 1)
1149 #define ZBT_MICRO ((1ULL << 63) + 3)
1150 /* any other values are ptrtbl blocks */
1153 * the embedded pointer table takes up half a block:
1154 * block size / entry size (2^3) / 2
1156 #define ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1)
1159 * The embedded pointer table starts half-way through the block. Since
1160 * the pointer table itself is half the block, it starts at (64-bit)
1161 * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)).
1163 #define ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \
1164 ((uint64_t *)(zap)->zap_phys) \
1165 [(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))]
1169 * If zap_phys_t is modified, zap_byteswap() must be modified.
1171 typedef struct zap_phys {
1172 uint64_t zap_block_type; /* ZBT_HEADER */
1173 uint64_t zap_magic; /* ZAP_MAGIC */
1175 struct zap_table_phys {
1176 uint64_t zt_blk; /* starting block number */
1177 uint64_t zt_numblks; /* number of blocks */
1178 uint64_t zt_shift; /* bits to index it */
1179 uint64_t zt_nextblk; /* next (larger) copy start block */
1180 uint64_t zt_blks_copied; /* number source blocks copied */
1183 uint64_t zap_freeblk; /* the next free block */
1184 uint64_t zap_num_leafs; /* number of leafs */
1185 uint64_t zap_num_entries; /* number of entries */
1186 uint64_t zap_salt; /* salt to stir into hash function */
1188 * This structure is followed by padding, and then the embedded
1189 * pointer table. The embedded pointer table takes up second
1190 * half of the block. It is accessed using the
1191 * ZAP_EMBEDDED_PTRTBL_ENT() macro.
1195 typedef struct zap_table_phys zap_table_phys_t;
1197 typedef struct fat_zap {
1198 int zap_block_shift; /* block size shift */
1199 zap_phys_t *zap_phys;
1202 #define ZAP_LEAF_MAGIC 0x2AB1EAF
1204 /* chunk size = 24 bytes */
1205 #define ZAP_LEAF_CHUNKSIZE 24
1208 * The amount of space available for chunks is:
1209 * block size (1<<l->l_bs) - hash entry size (2) * number of hash
1210 * entries - header space (2*chunksize)
1212 #define ZAP_LEAF_NUMCHUNKS(l) \
1213 (((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \
1214 ZAP_LEAF_CHUNKSIZE - 2)
1217 * The amount of space within the chunk available for the array is:
1218 * chunk size - space for type (1) - space for next pointer (2)
1220 #define ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3)
1222 #define ZAP_LEAF_ARRAY_NCHUNKS(bytes) \
1223 (((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES)
1226 * Low water mark: when there are only this many chunks free, start
1227 * growing the ptrtbl. Ideally, this should be larger than a
1228 * "reasonably-sized" entry. 20 chunks is more than enough for the
1229 * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value),
1230 * while still being only around 3% for 16k blocks.
1232 #define ZAP_LEAF_LOW_WATER (20)
1235 * The leaf hash table has block size / 2^5 (32) number of entries,
1236 * which should be more than enough for the maximum number of entries,
1237 * which is less than block size / CHUNKSIZE (24) / minimum number of
1238 * chunks per entry (3).
1240 #define ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5)
1241 #define ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l))
1244 * The chunks start immediately after the hash table. The end of the
1245 * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a
1248 #define ZAP_LEAF_CHUNK(l, idx) \
1249 ((zap_leaf_chunk_t *) \
1250 ((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx]
1251 #define ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry)
1253 typedef enum zap_chunk_type {
1254 ZAP_CHUNK_FREE = 253,
1255 ZAP_CHUNK_ENTRY = 252,
1256 ZAP_CHUNK_ARRAY = 251,
1257 ZAP_CHUNK_TYPE_MAX = 250
1262 * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified.
1264 typedef struct zap_leaf_phys {
1265 struct zap_leaf_header {
1266 uint64_t lh_block_type; /* ZBT_LEAF */
1268 uint64_t lh_prefix; /* hash prefix of this leaf */
1269 uint32_t lh_magic; /* ZAP_LEAF_MAGIC */
1270 uint16_t lh_nfree; /* number free chunks */
1271 uint16_t lh_nentries; /* number of entries */
1272 uint16_t lh_prefix_len; /* num bits used to id this */
1274 /* above is accessable to zap, below is zap_leaf private */
1276 uint16_t lh_freelist; /* chunk head of free list */
1277 uint8_t lh_pad2[12];
1278 } l_hdr; /* 2 24-byte chunks */
1281 * The header is followed by a hash table with
1282 * ZAP_LEAF_HASH_NUMENTRIES(zap) entries. The hash table is
1283 * followed by an array of ZAP_LEAF_NUMCHUNKS(zap)
1284 * zap_leaf_chunk structures. These structures are accessed
1285 * with the ZAP_LEAF_CHUNK() macro.
1291 typedef union zap_leaf_chunk {
1292 struct zap_leaf_entry {
1293 uint8_t le_type; /* always ZAP_CHUNK_ENTRY */
1294 uint8_t le_value_intlen; /* size of ints */
1295 uint16_t le_next; /* next entry in hash chain */
1296 uint16_t le_name_chunk; /* first chunk of the name */
1297 uint16_t le_name_numints; /* bytes in name, incl null */
1298 uint16_t le_value_chunk; /* first chunk of the value */
1299 uint16_t le_value_numints; /* value length in ints */
1300 uint32_t le_cd; /* collision differentiator */
1301 uint64_t le_hash; /* hash value of the name */
1303 struct zap_leaf_array {
1304 uint8_t la_type; /* always ZAP_CHUNK_ARRAY */
1305 uint8_t la_array[ZAP_LEAF_ARRAY_BYTES];
1306 uint16_t la_next; /* next blk or CHAIN_END */
1308 struct zap_leaf_free {
1309 uint8_t lf_type; /* always ZAP_CHUNK_FREE */
1310 uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES];
1311 uint16_t lf_next; /* next in free list, or CHAIN_END */
1315 typedef struct zap_leaf {
1316 int l_bs; /* block size shift */
1317 zap_leaf_phys_t *l_phys;
1321 * Define special zfs pflags
1323 #define ZFS_XATTR 0x1 /* is an extended attribute */
1324 #define ZFS_INHERIT_ACE 0x2 /* ace has inheritable ACEs */
1325 #define ZFS_ACL_TRIVIAL 0x4 /* files ACL is trivial */
1327 #define MASTER_NODE_OBJ 1
1330 * special attributes for master node.
1333 #define ZFS_FSID "FSID"
1334 #define ZFS_UNLINKED_SET "DELETE_QUEUE"
1335 #define ZFS_ROOT_OBJ "ROOT"
1336 #define ZPL_VERSION_OBJ "VERSION"
1337 #define ZFS_PROP_BLOCKPERPAGE "BLOCKPERPAGE"
1338 #define ZFS_PROP_NOGROWBLOCKS "NOGROWBLOCKS"
1340 #define ZFS_FLAG_BLOCKPERPAGE 0x1
1341 #define ZFS_FLAG_NOGROWBLOCKS 0x2
1344 * ZPL version - rev'd whenever an incompatible on-disk format change
1345 * occurs. Independent of SPA/DMU/ZAP versioning.
1348 #define ZPL_VERSION 1ULL
1351 * The directory entry has the type (currently unused on Solaris) in the
1352 * top 4 bits, and the object number in the low 48 bits. The "middle"
1353 * 12 bits are unused.
1355 #define ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4)
1356 #define ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48)
1357 #define ZFS_DIRENT_MAKE(type, obj) (((uint64_t)type << 60) | obj)
1359 typedef struct ace {
1360 uid_t a_who; /* uid or gid */
1361 uint32_t a_access_mask; /* read,write,... */
1362 uint16_t a_flags; /* see below */
1363 uint16_t a_type; /* allow or deny */
1366 #define ACE_SLOT_CNT 6
1368 typedef struct zfs_znode_acl {
1369 uint64_t z_acl_extern_obj; /* ext acl pieces */
1370 uint32_t z_acl_count; /* Number of ACEs */
1371 uint16_t z_acl_version; /* acl version */
1372 uint16_t z_acl_pad; /* pad */
1373 ace_t z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */
1377 * This is the persistent portion of the znode. It is stored
1378 * in the "bonus buffer" of the file. Short symbolic links
1379 * are also stored in the bonus buffer.
1381 typedef struct znode_phys {
1382 uint64_t zp_atime[2]; /* 0 - last file access time */
1383 uint64_t zp_mtime[2]; /* 16 - last file modification time */
1384 uint64_t zp_ctime[2]; /* 32 - last file change time */
1385 uint64_t zp_crtime[2]; /* 48 - creation time */
1386 uint64_t zp_gen; /* 64 - generation (txg of creation) */
1387 uint64_t zp_mode; /* 72 - file mode bits */
1388 uint64_t zp_size; /* 80 - size of file */
1389 uint64_t zp_parent; /* 88 - directory parent (`..') */
1390 uint64_t zp_links; /* 96 - number of links to file */
1391 uint64_t zp_xattr; /* 104 - DMU object for xattrs */
1392 uint64_t zp_rdev; /* 112 - dev_t for VBLK & VCHR files */
1393 uint64_t zp_flags; /* 120 - persistent flags */
1394 uint64_t zp_uid; /* 128 - file owner */
1395 uint64_t zp_gid; /* 136 - owning group */
1396 uint64_t zp_pad[4]; /* 144 - future */
1397 zfs_znode_acl_t zp_acl; /* 176 - 263 ACL */
1399 * Data may pad out any remaining bytes in the znode buffer, eg:
1401 * |<---------------------- dnode_phys (512) ------------------------>|
1402 * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->|
1403 * |<---- znode (264) ---->|<---- data (56) ---->|
1405 * At present, we only use this space to store symbolic links.
1410 * In-core vdev representation.
1413 typedef int vdev_phys_read_t(struct vdev *vdev, void *priv,
1414 off_t offset, void *buf, size_t bytes);
1415 typedef int vdev_read_t(struct vdev *vdev, const blkptr_t *bp,
1416 void *buf, off_t offset, size_t bytes);
1418 typedef STAILQ_HEAD(vdev_list, vdev) vdev_list_t;
1420 typedef struct vdev {
1421 STAILQ_ENTRY(vdev) v_childlink; /* link in parent's child list */
1422 STAILQ_ENTRY(vdev) v_alllink; /* link in global vdev list */
1423 vdev_list_t v_children; /* children of this vdev */
1424 const char *v_name; /* vdev name */
1425 uint64_t v_guid; /* vdev guid */
1426 int v_id; /* index in parent */
1427 int v_ashift; /* offset to block shift */
1428 int v_nparity; /* # parity for raidz */
1429 struct vdev *v_top; /* parent vdev */
1430 int v_nchildren; /* # children */
1431 vdev_state_t v_state; /* current state */
1432 vdev_phys_read_t *v_phys_read; /* read from raw leaf vdev */
1433 vdev_read_t *v_read; /* read from vdev */
1434 void *v_read_priv; /* private data for read function */
1438 * In-core pool representation.
1440 typedef STAILQ_HEAD(spa_list, spa) spa_list_t;
1442 typedef struct spa {
1443 STAILQ_ENTRY(spa) spa_link; /* link in global pool list */
1444 char *spa_name; /* pool name */
1445 uint64_t spa_guid; /* pool guid */
1446 uint64_t spa_txg; /* most recent transaction */
1447 struct uberblock spa_uberblock; /* best uberblock so far */
1448 vdev_list_t spa_vdevs; /* list of all toplevel vdevs */
1449 objset_phys_t spa_mos; /* MOS for this pool */
1450 int spa_inited; /* initialized */
1453 static void decode_embedded_bp_compressed(const blkptr_t *, void *);