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 * We currently support nine block sizes, from 512 bytes to 128K.
117 * We could go higher, but the benefits are near-zero and the cost
118 * of COWing a giant block to modify one byte would become excessive.
120 #define SPA_MINBLOCKSHIFT 9
121 #define SPA_MAXBLOCKSHIFT 17
122 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT)
123 #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT)
125 #define SPA_BLOCKSIZES (SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1)
128 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
129 * The ASIZE encoding should be at least 64 times larger (6 more bits)
130 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
131 * overhead, three DVAs per bp, plus one more bit in case we do anything
132 * else that expands the ASIZE.
134 #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */
135 #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */
136 #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */
139 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
140 * The members of the dva_t should be considered opaque outside the SPA.
143 uint64_t dva_word[2];
147 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
149 typedef struct zio_cksum {
154 * Each block is described by its DVAs, time of birth, checksum, etc.
155 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
157 * 64 56 48 40 32 24 16 8 0
158 * +-------+-------+-------+-------+-------+-------+-------+-------+
159 * 0 | vdev1 | GRID | ASIZE |
160 * +-------+-------+-------+-------+-------+-------+-------+-------+
162 * +-------+-------+-------+-------+-------+-------+-------+-------+
163 * 2 | vdev2 | GRID | ASIZE |
164 * +-------+-------+-------+-------+-------+-------+-------+-------+
166 * +-------+-------+-------+-------+-------+-------+-------+-------+
167 * 4 | vdev3 | GRID | ASIZE |
168 * +-------+-------+-------+-------+-------+-------+-------+-------+
170 * +-------+-------+-------+-------+-------+-------+-------+-------+
171 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE |
172 * +-------+-------+-------+-------+-------+-------+-------+-------+
174 * +-------+-------+-------+-------+-------+-------+-------+-------+
176 * +-------+-------+-------+-------+-------+-------+-------+-------+
177 * 9 | physical birth txg |
178 * +-------+-------+-------+-------+-------+-------+-------+-------+
179 * a | logical birth txg |
180 * +-------+-------+-------+-------+-------+-------+-------+-------+
182 * +-------+-------+-------+-------+-------+-------+-------+-------+
184 * +-------+-------+-------+-------+-------+-------+-------+-------+
186 * +-------+-------+-------+-------+-------+-------+-------+-------+
188 * +-------+-------+-------+-------+-------+-------+-------+-------+
190 * +-------+-------+-------+-------+-------+-------+-------+-------+
194 * vdev virtual device ID
195 * offset offset into virtual device
197 * PSIZE physical size (after compression)
198 * ASIZE allocated size (including RAID-Z parity and gang block headers)
199 * GRID RAID-Z layout information (reserved for future use)
200 * cksum checksum function
201 * comp compression function
202 * G gang block indicator
203 * B byteorder (endianness)
205 * X encryption (on version 30, which is not supported)
206 * E blkptr_t contains embedded data (see below)
207 * lvl level of indirection
208 * type DMU object type
209 * phys birth txg of block allocation; zero if same as logical birth txg
210 * log. birth transaction group in which the block was logically born
211 * fill count number of non-zero blocks under this bp
212 * checksum[4] 256-bit checksum of the data this bp describes
216 * "Embedded" blkptr_t's don't actually point to a block, instead they
217 * have a data payload embedded in the blkptr_t itself. See the comment
218 * in blkptr.c for more details.
220 * The blkptr_t is laid out as follows:
222 * 64 56 48 40 32 24 16 8 0
223 * +-------+-------+-------+-------+-------+-------+-------+-------+
230 * +-------+-------+-------+-------+-------+-------+-------+-------+
231 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE |
232 * +-------+-------+-------+-------+-------+-------+-------+-------+
236 * +-------+-------+-------+-------+-------+-------+-------+-------+
237 * a | logical birth txg |
238 * +-------+-------+-------+-------+-------+-------+-------+-------+
244 * +-------+-------+-------+-------+-------+-------+-------+-------+
248 * payload contains the embedded data
249 * B (byteorder) byteorder (endianness)
250 * D (dedup) padding (set to zero)
251 * X encryption (set to zero; see above)
252 * E (embedded) set to one
253 * lvl indirection level
254 * type DMU object type
255 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*)
256 * comp compression function of payload
257 * PSIZE size of payload after compression, in bytes
258 * LSIZE logical size of payload, in bytes
259 * note that 25 bits is enough to store the largest
260 * "normal" BP's LSIZE (2^16 * 2^9) in bytes
261 * log. birth transaction group in which the block was logically born
263 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
264 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
265 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
266 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
267 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must
268 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before
269 * other macros, as they assert that they are only used on BP's of the correct
273 #define BPE_GET_ETYPE(bp) \
274 (ASSERT(BP_IS_EMBEDDED(bp)), \
275 BF64_GET((bp)->blk_prop, 40, 8))
276 #define BPE_SET_ETYPE(bp, t) do { \
277 ASSERT(BP_IS_EMBEDDED(bp)); \
278 BF64_SET((bp)->blk_prop, 40, 8, t); \
279 _NOTE(CONSTCOND) } while (0)
281 #define BPE_GET_LSIZE(bp) \
282 (ASSERT(BP_IS_EMBEDDED(bp)), \
283 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
284 #define BPE_SET_LSIZE(bp, x) do { \
285 ASSERT(BP_IS_EMBEDDED(bp)); \
286 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
287 _NOTE(CONSTCOND) } while (0)
289 #define BPE_GET_PSIZE(bp) \
290 (ASSERT(BP_IS_EMBEDDED(bp)), \
291 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
292 #define BPE_SET_PSIZE(bp, x) do { \
293 ASSERT(BP_IS_EMBEDDED(bp)); \
294 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
295 _NOTE(CONSTCOND) } while (0)
297 typedef enum bp_embedded_type {
298 BP_EMBEDDED_TYPE_DATA,
299 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */
300 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED
301 } bp_embedded_type_t;
303 #define BPE_NUM_WORDS 14
304 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
305 #define BPE_IS_PAYLOADWORD(bp, wp) \
306 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
308 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */
309 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */
311 typedef struct blkptr {
312 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
313 uint64_t blk_prop; /* size, compression, type, etc */
314 uint64_t blk_pad[2]; /* Extra space for the future */
315 uint64_t blk_phys_birth; /* txg when block was allocated */
316 uint64_t blk_birth; /* transaction group at birth */
317 uint64_t blk_fill; /* fill count */
318 zio_cksum_t blk_cksum; /* 256-bit checksum */
322 * Macros to get and set fields in a bp or DVA.
324 #define DVA_GET_ASIZE(dva) \
325 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
326 #define DVA_SET_ASIZE(dva, x) \
327 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
328 SPA_MINBLOCKSHIFT, 0, x)
330 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8)
331 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x)
333 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32)
334 #define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x)
336 #define DVA_GET_OFFSET(dva) \
337 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
338 #define DVA_SET_OFFSET(dva, x) \
339 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
341 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1)
342 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x)
344 #define BP_GET_LSIZE(bp) \
345 (BP_IS_EMBEDDED(bp) ? \
346 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
347 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
348 #define BP_SET_LSIZE(bp, x) do { \
349 ASSERT(!BP_IS_EMBEDDED(bp)); \
350 BF64_SET_SB((bp)->blk_prop, \
351 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
352 _NOTE(CONSTCOND) } while (0)
354 #define BP_GET_PSIZE(bp) \
355 BF64_GET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)
356 #define BP_SET_PSIZE(bp, x) \
357 BF64_SET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x)
359 #define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7)
360 #define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x)
362 #define BP_GET_CHECKSUM(bp) BF64_GET((bp)->blk_prop, 40, 8)
363 #define BP_SET_CHECKSUM(bp, x) BF64_SET((bp)->blk_prop, 40, 8, x)
365 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8)
366 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x)
368 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5)
369 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x)
371 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1)
373 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1)
374 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x)
376 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1)
377 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x)
379 #define BP_PHYSICAL_BIRTH(bp) \
380 ((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
382 #define BP_GET_ASIZE(bp) \
383 (DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
384 DVA_GET_ASIZE(&(bp)->blk_dva[2]))
386 #define BP_GET_UCSIZE(bp) \
387 ((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \
388 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp));
390 #define BP_GET_NDVAS(bp) \
391 (!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
392 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
393 !!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
395 #define DVA_EQUAL(dva1, dva2) \
396 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
397 (dva1)->dva_word[0] == (dva2)->dva_word[0])
399 #define ZIO_CHECKSUM_EQUAL(zc1, zc2) \
400 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
401 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \
402 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \
403 ((zc1).zc_word[3] - (zc2).zc_word[3])))
406 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0)
408 #define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \
410 (zcp)->zc_word[0] = w0; \
411 (zcp)->zc_word[1] = w1; \
412 (zcp)->zc_word[2] = w2; \
413 (zcp)->zc_word[3] = w3; \
416 #define BP_IDENTITY(bp) (&(bp)->blk_dva[0])
417 #define BP_IS_GANG(bp) DVA_GET_GANG(BP_IDENTITY(bp))
418 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \
419 (dva)->dva_word[1] == 0ULL)
420 #define BP_IS_HOLE(bp) DVA_IS_EMPTY(BP_IDENTITY(bp))
421 #define BP_IS_OLDER(bp, txg) (!BP_IS_HOLE(bp) && (bp)->blk_birth < (txg))
423 #define BP_ZERO(bp) \
425 (bp)->blk_dva[0].dva_word[0] = 0; \
426 (bp)->blk_dva[0].dva_word[1] = 0; \
427 (bp)->blk_dva[1].dva_word[0] = 0; \
428 (bp)->blk_dva[1].dva_word[1] = 0; \
429 (bp)->blk_dva[2].dva_word[0] = 0; \
430 (bp)->blk_dva[2].dva_word[1] = 0; \
431 (bp)->blk_prop = 0; \
432 (bp)->blk_pad[0] = 0; \
433 (bp)->blk_pad[1] = 0; \
434 (bp)->blk_phys_birth = 0; \
435 (bp)->blk_birth = 0; \
436 (bp)->blk_fill = 0; \
437 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \
440 #define BPE_NUM_WORDS 14
441 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
442 #define BPE_IS_PAYLOADWORD(bp, wp) \
443 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
448 #define ZEC_MAGIC 0x210da7ab10c7a11ULL
450 typedef struct zio_eck {
451 uint64_t zec_magic; /* for validation, endianness */
452 zio_cksum_t zec_cksum; /* 256-bit checksum */
456 * Gang block headers are self-checksumming and contain an array
459 #define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE
460 #define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \
461 sizeof (zio_eck_t)) / sizeof (blkptr_t))
462 #define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \
463 sizeof (zio_eck_t) - \
464 (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\
467 typedef struct zio_gbh {
468 blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS];
469 uint64_t zg_filler[SPA_GBH_FILLER];
473 #define VDEV_RAIDZ_MAXPARITY 3
475 #define VDEV_PAD_SIZE (8 << 10)
476 /* 2 padding areas (vl_pad1 and vl_pad2) to skip */
477 #define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2
478 #define VDEV_PHYS_SIZE (112 << 10)
479 #define VDEV_UBERBLOCK_RING (128 << 10)
481 #define VDEV_UBERBLOCK_SHIFT(vd) \
482 MAX((vd)->v_top->v_ashift, UBERBLOCK_SHIFT)
483 #define VDEV_UBERBLOCK_COUNT(vd) \
484 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
485 #define VDEV_UBERBLOCK_OFFSET(vd, n) \
486 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
487 #define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd))
489 typedef struct vdev_phys {
490 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)];
494 typedef struct vdev_label {
495 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */
496 char vl_pad2[VDEV_PAD_SIZE]; /* 8K */
497 vdev_phys_t vl_vdev_phys; /* 112K */
498 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */
499 } vdev_label_t; /* 256K total */
504 #define VDD_METASLAB 0x01
508 * Size and offset of embedded boot loader region on each label.
509 * The total size of the first two labels plus the boot area is 4MB.
511 #define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t))
512 #define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */
515 * Size of label regions at the start and end of each leaf device.
517 #define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
518 #define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t))
519 #define VDEV_LABELS 4
522 ZIO_CHECKSUM_INHERIT = 0,
526 ZIO_CHECKSUM_GANG_HEADER,
528 ZIO_CHECKSUM_FLETCHER_2,
529 ZIO_CHECKSUM_FLETCHER_4,
532 ZIO_CHECKSUM_FUNCTIONS
535 #define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4
536 #define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON
539 ZIO_COMPRESS_INHERIT = 0,
555 ZIO_COMPRESS_FUNCTIONS
558 #define ZIO_COMPRESS_ON_VALUE ZIO_COMPRESS_LZJB
559 #define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF
561 /* nvlist pack encoding */
562 #define NV_ENCODE_NATIVE 0
563 #define NV_ENCODE_XDR 1
566 DATA_TYPE_UNKNOWN = 0,
576 DATA_TYPE_BYTE_ARRAY,
577 DATA_TYPE_INT16_ARRAY,
578 DATA_TYPE_UINT16_ARRAY,
579 DATA_TYPE_INT32_ARRAY,
580 DATA_TYPE_UINT32_ARRAY,
581 DATA_TYPE_INT64_ARRAY,
582 DATA_TYPE_UINT64_ARRAY,
583 DATA_TYPE_STRING_ARRAY,
586 DATA_TYPE_NVLIST_ARRAY,
587 DATA_TYPE_BOOLEAN_VALUE,
590 DATA_TYPE_BOOLEAN_ARRAY,
591 DATA_TYPE_INT8_ARRAY,
592 DATA_TYPE_UINT8_ARRAY
596 * On-disk version number.
598 #define SPA_VERSION_1 1ULL
599 #define SPA_VERSION_2 2ULL
600 #define SPA_VERSION_3 3ULL
601 #define SPA_VERSION_4 4ULL
602 #define SPA_VERSION_5 5ULL
603 #define SPA_VERSION_6 6ULL
604 #define SPA_VERSION_7 7ULL
605 #define SPA_VERSION_8 8ULL
606 #define SPA_VERSION_9 9ULL
607 #define SPA_VERSION_10 10ULL
608 #define SPA_VERSION_11 11ULL
609 #define SPA_VERSION_12 12ULL
610 #define SPA_VERSION_13 13ULL
611 #define SPA_VERSION_14 14ULL
612 #define SPA_VERSION_15 15ULL
613 #define SPA_VERSION_16 16ULL
614 #define SPA_VERSION_17 17ULL
615 #define SPA_VERSION_18 18ULL
616 #define SPA_VERSION_19 19ULL
617 #define SPA_VERSION_20 20ULL
618 #define SPA_VERSION_21 21ULL
619 #define SPA_VERSION_22 22ULL
620 #define SPA_VERSION_23 23ULL
621 #define SPA_VERSION_24 24ULL
622 #define SPA_VERSION_25 25ULL
623 #define SPA_VERSION_26 26ULL
624 #define SPA_VERSION_27 27ULL
625 #define SPA_VERSION_28 28ULL
626 #define SPA_VERSION_5000 5000ULL
629 * When bumping up SPA_VERSION, make sure GRUB ZFS understands the on-disk
630 * format change. Go to usr/src/grub/grub-0.97/stage2/{zfs-include/, fsys_zfs*},
631 * and do the appropriate changes. Also bump the version number in
632 * usr/src/grub/capability.
634 #define SPA_VERSION SPA_VERSION_5000
635 #define SPA_VERSION_STRING "5000"
638 * Symbolic names for the changes that caused a SPA_VERSION switch.
639 * Used in the code when checking for presence or absence of a feature.
640 * Feel free to define multiple symbolic names for each version if there
641 * were multiple changes to on-disk structures during that version.
643 * NOTE: When checking the current SPA_VERSION in your code, be sure
644 * to use spa_version() since it reports the version of the
645 * last synced uberblock. Checking the in-flight version can
646 * be dangerous in some cases.
648 #define SPA_VERSION_INITIAL SPA_VERSION_1
649 #define SPA_VERSION_DITTO_BLOCKS SPA_VERSION_2
650 #define SPA_VERSION_SPARES SPA_VERSION_3
651 #define SPA_VERSION_RAID6 SPA_VERSION_3
652 #define SPA_VERSION_BPLIST_ACCOUNT SPA_VERSION_3
653 #define SPA_VERSION_RAIDZ_DEFLATE SPA_VERSION_3
654 #define SPA_VERSION_DNODE_BYTES SPA_VERSION_3
655 #define SPA_VERSION_ZPOOL_HISTORY SPA_VERSION_4
656 #define SPA_VERSION_GZIP_COMPRESSION SPA_VERSION_5
657 #define SPA_VERSION_BOOTFS SPA_VERSION_6
658 #define SPA_VERSION_SLOGS SPA_VERSION_7
659 #define SPA_VERSION_DELEGATED_PERMS SPA_VERSION_8
660 #define SPA_VERSION_FUID SPA_VERSION_9
661 #define SPA_VERSION_REFRESERVATION SPA_VERSION_9
662 #define SPA_VERSION_REFQUOTA SPA_VERSION_9
663 #define SPA_VERSION_UNIQUE_ACCURATE SPA_VERSION_9
664 #define SPA_VERSION_L2CACHE SPA_VERSION_10
665 #define SPA_VERSION_NEXT_CLONES SPA_VERSION_11
666 #define SPA_VERSION_ORIGIN SPA_VERSION_11
667 #define SPA_VERSION_DSL_SCRUB SPA_VERSION_11
668 #define SPA_VERSION_SNAP_PROPS SPA_VERSION_12
669 #define SPA_VERSION_USED_BREAKDOWN SPA_VERSION_13
670 #define SPA_VERSION_PASSTHROUGH_X SPA_VERSION_14
671 #define SPA_VERSION_USERSPACE SPA_VERSION_15
672 #define SPA_VERSION_STMF_PROP SPA_VERSION_16
673 #define SPA_VERSION_RAIDZ3 SPA_VERSION_17
674 #define SPA_VERSION_USERREFS SPA_VERSION_18
675 #define SPA_VERSION_HOLES SPA_VERSION_19
676 #define SPA_VERSION_ZLE_COMPRESSION SPA_VERSION_20
677 #define SPA_VERSION_DEDUP SPA_VERSION_21
678 #define SPA_VERSION_RECVD_PROPS SPA_VERSION_22
679 #define SPA_VERSION_SLIM_ZIL SPA_VERSION_23
680 #define SPA_VERSION_SA SPA_VERSION_24
681 #define SPA_VERSION_SCAN SPA_VERSION_25
682 #define SPA_VERSION_DIR_CLONES SPA_VERSION_26
683 #define SPA_VERSION_DEADLISTS SPA_VERSION_26
684 #define SPA_VERSION_FAST_SNAP SPA_VERSION_27
685 #define SPA_VERSION_MULTI_REPLACE SPA_VERSION_28
686 #define SPA_VERSION_BEFORE_FEATURES SPA_VERSION_28
687 #define SPA_VERSION_FEATURES SPA_VERSION_5000
689 #define SPA_VERSION_IS_SUPPORTED(v) \
690 (((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \
691 ((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION))
694 * The following are configuration names used in the nvlist describing a pool's
697 #define ZPOOL_CONFIG_VERSION "version"
698 #define ZPOOL_CONFIG_POOL_NAME "name"
699 #define ZPOOL_CONFIG_POOL_STATE "state"
700 #define ZPOOL_CONFIG_POOL_TXG "txg"
701 #define ZPOOL_CONFIG_POOL_GUID "pool_guid"
702 #define ZPOOL_CONFIG_CREATE_TXG "create_txg"
703 #define ZPOOL_CONFIG_TOP_GUID "top_guid"
704 #define ZPOOL_CONFIG_VDEV_TREE "vdev_tree"
705 #define ZPOOL_CONFIG_TYPE "type"
706 #define ZPOOL_CONFIG_CHILDREN "children"
707 #define ZPOOL_CONFIG_ID "id"
708 #define ZPOOL_CONFIG_GUID "guid"
709 #define ZPOOL_CONFIG_PATH "path"
710 #define ZPOOL_CONFIG_DEVID "devid"
711 #define ZPOOL_CONFIG_METASLAB_ARRAY "metaslab_array"
712 #define ZPOOL_CONFIG_METASLAB_SHIFT "metaslab_shift"
713 #define ZPOOL_CONFIG_ASHIFT "ashift"
714 #define ZPOOL_CONFIG_ASIZE "asize"
715 #define ZPOOL_CONFIG_DTL "DTL"
716 #define ZPOOL_CONFIG_STATS "stats"
717 #define ZPOOL_CONFIG_WHOLE_DISK "whole_disk"
718 #define ZPOOL_CONFIG_ERRCOUNT "error_count"
719 #define ZPOOL_CONFIG_NOT_PRESENT "not_present"
720 #define ZPOOL_CONFIG_SPARES "spares"
721 #define ZPOOL_CONFIG_IS_SPARE "is_spare"
722 #define ZPOOL_CONFIG_NPARITY "nparity"
723 #define ZPOOL_CONFIG_HOSTID "hostid"
724 #define ZPOOL_CONFIG_HOSTNAME "hostname"
725 #define ZPOOL_CONFIG_IS_LOG "is_log"
726 #define ZPOOL_CONFIG_TIMESTAMP "timestamp" /* not stored on disk */
727 #define ZPOOL_CONFIG_FEATURES_FOR_READ "features_for_read"
730 * The persistent vdev state is stored as separate values rather than a single
731 * 'vdev_state' entry. This is because a device can be in multiple states, such
732 * as offline and degraded.
734 #define ZPOOL_CONFIG_OFFLINE "offline"
735 #define ZPOOL_CONFIG_FAULTED "faulted"
736 #define ZPOOL_CONFIG_DEGRADED "degraded"
737 #define ZPOOL_CONFIG_REMOVED "removed"
738 #define ZPOOL_CONFIG_FRU "fru"
739 #define ZPOOL_CONFIG_AUX_STATE "aux_state"
741 #define VDEV_TYPE_ROOT "root"
742 #define VDEV_TYPE_MIRROR "mirror"
743 #define VDEV_TYPE_REPLACING "replacing"
744 #define VDEV_TYPE_RAIDZ "raidz"
745 #define VDEV_TYPE_DISK "disk"
746 #define VDEV_TYPE_FILE "file"
747 #define VDEV_TYPE_MISSING "missing"
748 #define VDEV_TYPE_HOLE "hole"
749 #define VDEV_TYPE_SPARE "spare"
750 #define VDEV_TYPE_LOG "log"
751 #define VDEV_TYPE_L2CACHE "l2cache"
754 * This is needed in userland to report the minimum necessary device size.
756 #define SPA_MINDEVSIZE (64ULL << 20)
759 * The location of the pool configuration repository, shared between kernel and
762 #define ZPOOL_CACHE "/boot/zfs/zpool.cache"
765 * vdev states are ordered from least to most healthy.
766 * A vdev that's CANT_OPEN or below is considered unusable.
768 typedef enum vdev_state {
769 VDEV_STATE_UNKNOWN = 0, /* Uninitialized vdev */
770 VDEV_STATE_CLOSED, /* Not currently open */
771 VDEV_STATE_OFFLINE, /* Not allowed to open */
772 VDEV_STATE_REMOVED, /* Explicitly removed from system */
773 VDEV_STATE_CANT_OPEN, /* Tried to open, but failed */
774 VDEV_STATE_FAULTED, /* External request to fault device */
775 VDEV_STATE_DEGRADED, /* Replicated vdev with unhealthy kids */
776 VDEV_STATE_HEALTHY /* Presumed good */
780 * vdev aux states. When a vdev is in the CANT_OPEN state, the aux field
781 * of the vdev stats structure uses these constants to distinguish why.
783 typedef enum vdev_aux {
784 VDEV_AUX_NONE, /* no error */
785 VDEV_AUX_OPEN_FAILED, /* ldi_open_*() or vn_open() failed */
786 VDEV_AUX_CORRUPT_DATA, /* bad label or disk contents */
787 VDEV_AUX_NO_REPLICAS, /* insufficient number of replicas */
788 VDEV_AUX_BAD_GUID_SUM, /* vdev guid sum doesn't match */
789 VDEV_AUX_TOO_SMALL, /* vdev size is too small */
790 VDEV_AUX_BAD_LABEL, /* the label is OK but invalid */
791 VDEV_AUX_VERSION_NEWER, /* on-disk version is too new */
792 VDEV_AUX_VERSION_OLDER, /* on-disk version is too old */
793 VDEV_AUX_SPARED /* hot spare used in another pool */
797 * pool state. The following states are written to disk as part of the normal
798 * SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE. The remaining states are
799 * software abstractions used at various levels to communicate pool state.
801 typedef enum pool_state {
802 POOL_STATE_ACTIVE = 0, /* In active use */
803 POOL_STATE_EXPORTED, /* Explicitly exported */
804 POOL_STATE_DESTROYED, /* Explicitly destroyed */
805 POOL_STATE_SPARE, /* Reserved for hot spare use */
806 POOL_STATE_UNINITIALIZED, /* Internal spa_t state */
807 POOL_STATE_UNAVAIL, /* Internal libzfs state */
808 POOL_STATE_POTENTIALLY_ACTIVE /* Internal libzfs state */
812 * The uberblock version is incremented whenever an incompatible on-disk
813 * format change is made to the SPA, DMU, or ZAP.
815 * Note: the first two fields should never be moved. When a storage pool
816 * is opened, the uberblock must be read off the disk before the version
817 * can be checked. If the ub_version field is moved, we may not detect
818 * version mismatch. If the ub_magic field is moved, applications that
819 * expect the magic number in the first word won't work.
821 #define UBERBLOCK_MAGIC 0x00bab10c /* oo-ba-bloc! */
822 #define UBERBLOCK_SHIFT 10 /* up to 1K */
825 uint64_t ub_magic; /* UBERBLOCK_MAGIC */
826 uint64_t ub_version; /* SPA_VERSION */
827 uint64_t ub_txg; /* txg of last sync */
828 uint64_t ub_guid_sum; /* sum of all vdev guids */
829 uint64_t ub_timestamp; /* UTC time of last sync */
830 blkptr_t ub_rootbp; /* MOS objset_phys_t */
836 #define DNODE_MUST_BE_ALLOCATED 1
837 #define DNODE_MUST_BE_FREE 2
842 #define DNODE_SHIFT 9 /* 512 bytes */
843 #define DN_MIN_INDBLKSHIFT 10 /* 1k */
844 #define DN_MAX_INDBLKSHIFT 14 /* 16k */
845 #define DNODE_BLOCK_SHIFT 14 /* 16k */
846 #define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */
847 #define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */
848 #define DN_MAX_OFFSET_SHIFT 64 /* 2^64 bytes in a dnode */
853 #define DNODE_SIZE (1 << DNODE_SHIFT)
854 #define DN_MAX_NBLKPTR ((DNODE_SIZE - DNODE_CORE_SIZE) >> SPA_BLKPTRSHIFT)
855 #define DN_MAX_BONUSLEN (DNODE_SIZE - DNODE_CORE_SIZE - (1 << SPA_BLKPTRSHIFT))
856 #define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT)
858 #define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT)
859 #define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT)
860 #define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT)
862 /* The +2 here is a cheesy way to round up */
863 #define DN_MAX_LEVELS (2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \
864 (DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT)))
866 #define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \
867 (((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t))))
869 #define DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \
870 (dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT)
872 #define EPB(blkshift, typeshift) (1 << (blkshift - typeshift))
874 /* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */
875 #define DNODE_FLAG_USED_BYTES (1<<0)
876 #define DNODE_FLAG_USERUSED_ACCOUNTED (1<<1)
878 /* Does dnode have a SA spill blkptr in bonus? */
879 #define DNODE_FLAG_SPILL_BLKPTR (1<<2)
881 typedef struct dnode_phys {
882 uint8_t dn_type; /* dmu_object_type_t */
883 uint8_t dn_indblkshift; /* ln2(indirect block size) */
884 uint8_t dn_nlevels; /* 1=dn_blkptr->data blocks */
885 uint8_t dn_nblkptr; /* length of dn_blkptr */
886 uint8_t dn_bonustype; /* type of data in bonus buffer */
887 uint8_t dn_checksum; /* ZIO_CHECKSUM type */
888 uint8_t dn_compress; /* ZIO_COMPRESS type */
889 uint8_t dn_flags; /* DNODE_FLAG_* */
890 uint16_t dn_datablkszsec; /* data block size in 512b sectors */
891 uint16_t dn_bonuslen; /* length of dn_bonus */
894 /* accounting is protected by dn_dirty_mtx */
895 uint64_t dn_maxblkid; /* largest allocated block ID */
896 uint64_t dn_used; /* bytes (or sectors) of disk space */
900 blkptr_t dn_blkptr[1];
901 uint8_t dn_bonus[DN_MAX_BONUSLEN - sizeof (blkptr_t)];
905 typedef enum dmu_object_type {
908 DMU_OT_OBJECT_DIRECTORY, /* ZAP */
909 DMU_OT_OBJECT_ARRAY, /* UINT64 */
910 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */
911 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */
912 DMU_OT_BPLIST, /* UINT64 */
913 DMU_OT_BPLIST_HDR, /* UINT64 */
915 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */
916 DMU_OT_SPACE_MAP, /* UINT64 */
918 DMU_OT_INTENT_LOG, /* UINT64 */
920 DMU_OT_DNODE, /* DNODE */
921 DMU_OT_OBJSET, /* OBJSET */
923 DMU_OT_DSL_DIR, /* UINT64 */
924 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */
925 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */
926 DMU_OT_DSL_PROPS, /* ZAP */
927 DMU_OT_DSL_DATASET, /* UINT64 */
929 DMU_OT_ZNODE, /* ZNODE */
930 DMU_OT_OLDACL, /* Old ACL */
931 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */
932 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */
933 DMU_OT_MASTER_NODE, /* ZAP */
934 DMU_OT_UNLINKED_SET, /* ZAP */
936 DMU_OT_ZVOL, /* UINT8 */
937 DMU_OT_ZVOL_PROP, /* ZAP */
938 /* other; for testing only! */
939 DMU_OT_PLAIN_OTHER, /* UINT8 */
940 DMU_OT_UINT64_OTHER, /* UINT64 */
941 DMU_OT_ZAP_OTHER, /* ZAP */
942 /* new object types: */
943 DMU_OT_ERROR_LOG, /* ZAP */
944 DMU_OT_SPA_HISTORY, /* UINT8 */
945 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */
946 DMU_OT_POOL_PROPS, /* ZAP */
947 DMU_OT_DSL_PERMS, /* ZAP */
948 DMU_OT_ACL, /* ACL */
949 DMU_OT_SYSACL, /* SYSACL */
950 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */
951 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */
952 DMU_OT_NEXT_CLONES, /* ZAP */
953 DMU_OT_SCAN_QUEUE, /* ZAP */
954 DMU_OT_USERGROUP_USED, /* ZAP */
955 DMU_OT_USERGROUP_QUOTA, /* ZAP */
956 DMU_OT_USERREFS, /* ZAP */
957 DMU_OT_DDT_ZAP, /* ZAP */
958 DMU_OT_DDT_STATS, /* ZAP */
959 DMU_OT_SA, /* System attr */
960 DMU_OT_SA_MASTER_NODE, /* ZAP */
961 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */
962 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */
963 DMU_OT_SCAN_XLATE, /* ZAP */
964 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */
968 typedef enum dmu_objset_type {
973 DMU_OST_OTHER, /* For testing only! */
974 DMU_OST_ANY, /* Be careful! */
979 * header for all bonus and spill buffers.
980 * The header has a fixed portion with a variable number
981 * of "lengths" depending on the number of variable sized
982 * attribues which are determined by the "layout number"
985 #define SA_MAGIC 0x2F505A /* ZFS SA */
986 typedef struct sa_hdr_phys {
988 uint16_t sa_layout_info; /* Encoded with hdrsize and layout number */
989 uint16_t sa_lengths[1]; /* optional sizes for variable length attrs */
990 /* ... Data follows the lengths. */
994 * sa_hdr_phys -> sa_layout_info
1001 * Bits 0-10 are the layout number
1002 * Bits 11-16 are the size of the header.
1003 * The hdrsize is the number * 8
1006 * hdrsz of 1 ==> 8 byte header
1007 * 2 ==> 16 byte header
1011 #define SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10)
1012 #define SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 16, 3, 0)
1013 #define SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \
1015 BF32_SET_SB(x, 10, 6, 3, 0, size); \
1016 BF32_SET(x, 0, 10, num); \
1019 #define SA_MODE_OFFSET 0
1020 #define SA_SIZE_OFFSET 8
1021 #define SA_GEN_OFFSET 16
1022 #define SA_UID_OFFSET 24
1023 #define SA_GID_OFFSET 32
1024 #define SA_PARENT_OFFSET 40
1027 * Intent log header - this on disk structure holds fields to manage
1028 * the log. All fields are 64 bit to easily handle cross architectures.
1030 typedef struct zil_header {
1031 uint64_t zh_claim_txg; /* txg in which log blocks were claimed */
1032 uint64_t zh_replay_seq; /* highest replayed sequence number */
1033 blkptr_t zh_log; /* log chain */
1034 uint64_t zh_claim_seq; /* highest claimed sequence number */
1038 #define OBJSET_PHYS_SIZE 2048
1040 typedef struct objset_phys {
1041 dnode_phys_t os_meta_dnode;
1042 zil_header_t os_zil_header;
1045 char os_pad[OBJSET_PHYS_SIZE - sizeof (dnode_phys_t)*3 -
1046 sizeof (zil_header_t) - sizeof (uint64_t)*2];
1047 dnode_phys_t os_userused_dnode;
1048 dnode_phys_t os_groupused_dnode;
1051 typedef struct dsl_dir_phys {
1052 uint64_t dd_creation_time; /* not actually used */
1053 uint64_t dd_head_dataset_obj;
1054 uint64_t dd_parent_obj;
1055 uint64_t dd_clone_parent_obj;
1056 uint64_t dd_child_dir_zapobj;
1058 * how much space our children are accounting for; for leaf
1059 * datasets, == physical space used by fs + snaps
1061 uint64_t dd_used_bytes;
1062 uint64_t dd_compressed_bytes;
1063 uint64_t dd_uncompressed_bytes;
1064 /* Administrative quota setting */
1066 /* Administrative reservation setting */
1067 uint64_t dd_reserved;
1068 uint64_t dd_props_zapobj;
1069 uint64_t dd_pad[21]; /* pad out to 256 bytes for good measure */
1072 typedef struct dsl_dataset_phys {
1073 uint64_t ds_dir_obj;
1074 uint64_t ds_prev_snap_obj;
1075 uint64_t ds_prev_snap_txg;
1076 uint64_t ds_next_snap_obj;
1077 uint64_t ds_snapnames_zapobj; /* zap obj of snaps; ==0 for snaps */
1078 uint64_t ds_num_children; /* clone/snap children; ==0 for head */
1079 uint64_t ds_creation_time; /* seconds since 1970 */
1080 uint64_t ds_creation_txg;
1081 uint64_t ds_deadlist_obj;
1082 uint64_t ds_used_bytes;
1083 uint64_t ds_compressed_bytes;
1084 uint64_t ds_uncompressed_bytes;
1085 uint64_t ds_unique_bytes; /* only relevant to snapshots */
1087 * The ds_fsid_guid is a 56-bit ID that can change to avoid
1088 * collisions. The ds_guid is a 64-bit ID that will never
1089 * change, so there is a small probability that it will collide.
1091 uint64_t ds_fsid_guid;
1095 uint64_t ds_pad[8]; /* pad out to 320 bytes for good measure */
1096 } dsl_dataset_phys_t;
1099 * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
1101 #define DMU_POOL_DIRECTORY_OBJECT 1
1102 #define DMU_POOL_CONFIG "config"
1103 #define DMU_POOL_ROOT_DATASET "root_dataset"
1104 #define DMU_POOL_SYNC_BPLIST "sync_bplist"
1105 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub"
1106 #define DMU_POOL_ERRLOG_LAST "errlog_last"
1107 #define DMU_POOL_SPARES "spares"
1108 #define DMU_POOL_DEFLATE "deflate"
1109 #define DMU_POOL_HISTORY "history"
1110 #define DMU_POOL_PROPS "pool_props"
1112 #define ZAP_MAGIC 0x2F52AB2ABULL
1114 #define FZAP_BLOCK_SHIFT(zap) ((zap)->zap_block_shift)
1116 #define ZAP_MAXCD (uint32_t)(-1)
1117 #define ZAP_HASHBITS 28
1118 #define MZAP_ENT_LEN 64
1119 #define MZAP_NAME_LEN (MZAP_ENT_LEN - 8 - 4 - 2)
1120 #define MZAP_MAX_BLKSHIFT SPA_MAXBLOCKSHIFT
1121 #define MZAP_MAX_BLKSZ (1 << MZAP_MAX_BLKSHIFT)
1123 typedef struct mzap_ent_phys {
1126 uint16_t mze_pad; /* in case we want to chain them someday */
1127 char mze_name[MZAP_NAME_LEN];
1130 typedef struct mzap_phys {
1131 uint64_t mz_block_type; /* ZBT_MICRO */
1134 mzap_ent_phys_t mz_chunk[1];
1135 /* actually variable size depending on block size */
1139 * The (fat) zap is stored in one object. It is an array of
1140 * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of:
1142 * ptrtbl fits in first block:
1143 * [zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ...
1145 * ptrtbl too big for first block:
1146 * [zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ...
1150 #define ZBT_LEAF ((1ULL << 63) + 0)
1151 #define ZBT_HEADER ((1ULL << 63) + 1)
1152 #define ZBT_MICRO ((1ULL << 63) + 3)
1153 /* any other values are ptrtbl blocks */
1156 * the embedded pointer table takes up half a block:
1157 * block size / entry size (2^3) / 2
1159 #define ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1)
1162 * The embedded pointer table starts half-way through the block. Since
1163 * the pointer table itself is half the block, it starts at (64-bit)
1164 * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)).
1166 #define ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \
1167 ((uint64_t *)(zap)->zap_phys) \
1168 [(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))]
1172 * If zap_phys_t is modified, zap_byteswap() must be modified.
1174 typedef struct zap_phys {
1175 uint64_t zap_block_type; /* ZBT_HEADER */
1176 uint64_t zap_magic; /* ZAP_MAGIC */
1178 struct zap_table_phys {
1179 uint64_t zt_blk; /* starting block number */
1180 uint64_t zt_numblks; /* number of blocks */
1181 uint64_t zt_shift; /* bits to index it */
1182 uint64_t zt_nextblk; /* next (larger) copy start block */
1183 uint64_t zt_blks_copied; /* number source blocks copied */
1186 uint64_t zap_freeblk; /* the next free block */
1187 uint64_t zap_num_leafs; /* number of leafs */
1188 uint64_t zap_num_entries; /* number of entries */
1189 uint64_t zap_salt; /* salt to stir into hash function */
1191 * This structure is followed by padding, and then the embedded
1192 * pointer table. The embedded pointer table takes up second
1193 * half of the block. It is accessed using the
1194 * ZAP_EMBEDDED_PTRTBL_ENT() macro.
1198 typedef struct zap_table_phys zap_table_phys_t;
1200 typedef struct fat_zap {
1201 int zap_block_shift; /* block size shift */
1202 zap_phys_t *zap_phys;
1205 #define ZAP_LEAF_MAGIC 0x2AB1EAF
1207 /* chunk size = 24 bytes */
1208 #define ZAP_LEAF_CHUNKSIZE 24
1211 * The amount of space available for chunks is:
1212 * block size (1<<l->l_bs) - hash entry size (2) * number of hash
1213 * entries - header space (2*chunksize)
1215 #define ZAP_LEAF_NUMCHUNKS(l) \
1216 (((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \
1217 ZAP_LEAF_CHUNKSIZE - 2)
1220 * The amount of space within the chunk available for the array is:
1221 * chunk size - space for type (1) - space for next pointer (2)
1223 #define ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3)
1225 #define ZAP_LEAF_ARRAY_NCHUNKS(bytes) \
1226 (((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES)
1229 * Low water mark: when there are only this many chunks free, start
1230 * growing the ptrtbl. Ideally, this should be larger than a
1231 * "reasonably-sized" entry. 20 chunks is more than enough for the
1232 * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value),
1233 * while still being only around 3% for 16k blocks.
1235 #define ZAP_LEAF_LOW_WATER (20)
1238 * The leaf hash table has block size / 2^5 (32) number of entries,
1239 * which should be more than enough for the maximum number of entries,
1240 * which is less than block size / CHUNKSIZE (24) / minimum number of
1241 * chunks per entry (3).
1243 #define ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5)
1244 #define ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l))
1247 * The chunks start immediately after the hash table. The end of the
1248 * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a
1251 #define ZAP_LEAF_CHUNK(l, idx) \
1252 ((zap_leaf_chunk_t *) \
1253 ((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx]
1254 #define ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry)
1256 typedef enum zap_chunk_type {
1257 ZAP_CHUNK_FREE = 253,
1258 ZAP_CHUNK_ENTRY = 252,
1259 ZAP_CHUNK_ARRAY = 251,
1260 ZAP_CHUNK_TYPE_MAX = 250
1265 * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified.
1267 typedef struct zap_leaf_phys {
1268 struct zap_leaf_header {
1269 uint64_t lh_block_type; /* ZBT_LEAF */
1271 uint64_t lh_prefix; /* hash prefix of this leaf */
1272 uint32_t lh_magic; /* ZAP_LEAF_MAGIC */
1273 uint16_t lh_nfree; /* number free chunks */
1274 uint16_t lh_nentries; /* number of entries */
1275 uint16_t lh_prefix_len; /* num bits used to id this */
1277 /* above is accessable to zap, below is zap_leaf private */
1279 uint16_t lh_freelist; /* chunk head of free list */
1280 uint8_t lh_pad2[12];
1281 } l_hdr; /* 2 24-byte chunks */
1284 * The header is followed by a hash table with
1285 * ZAP_LEAF_HASH_NUMENTRIES(zap) entries. The hash table is
1286 * followed by an array of ZAP_LEAF_NUMCHUNKS(zap)
1287 * zap_leaf_chunk structures. These structures are accessed
1288 * with the ZAP_LEAF_CHUNK() macro.
1294 typedef union zap_leaf_chunk {
1295 struct zap_leaf_entry {
1296 uint8_t le_type; /* always ZAP_CHUNK_ENTRY */
1297 uint8_t le_value_intlen; /* size of ints */
1298 uint16_t le_next; /* next entry in hash chain */
1299 uint16_t le_name_chunk; /* first chunk of the name */
1300 uint16_t le_name_numints; /* bytes in name, incl null */
1301 uint16_t le_value_chunk; /* first chunk of the value */
1302 uint16_t le_value_numints; /* value length in ints */
1303 uint32_t le_cd; /* collision differentiator */
1304 uint64_t le_hash; /* hash value of the name */
1306 struct zap_leaf_array {
1307 uint8_t la_type; /* always ZAP_CHUNK_ARRAY */
1308 uint8_t la_array[ZAP_LEAF_ARRAY_BYTES];
1309 uint16_t la_next; /* next blk or CHAIN_END */
1311 struct zap_leaf_free {
1312 uint8_t lf_type; /* always ZAP_CHUNK_FREE */
1313 uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES];
1314 uint16_t lf_next; /* next in free list, or CHAIN_END */
1318 typedef struct zap_leaf {
1319 int l_bs; /* block size shift */
1320 zap_leaf_phys_t *l_phys;
1324 * Define special zfs pflags
1326 #define ZFS_XATTR 0x1 /* is an extended attribute */
1327 #define ZFS_INHERIT_ACE 0x2 /* ace has inheritable ACEs */
1328 #define ZFS_ACL_TRIVIAL 0x4 /* files ACL is trivial */
1330 #define MASTER_NODE_OBJ 1
1333 * special attributes for master node.
1336 #define ZFS_FSID "FSID"
1337 #define ZFS_UNLINKED_SET "DELETE_QUEUE"
1338 #define ZFS_ROOT_OBJ "ROOT"
1339 #define ZPL_VERSION_OBJ "VERSION"
1340 #define ZFS_PROP_BLOCKPERPAGE "BLOCKPERPAGE"
1341 #define ZFS_PROP_NOGROWBLOCKS "NOGROWBLOCKS"
1343 #define ZFS_FLAG_BLOCKPERPAGE 0x1
1344 #define ZFS_FLAG_NOGROWBLOCKS 0x2
1347 * ZPL version - rev'd whenever an incompatible on-disk format change
1348 * occurs. Independent of SPA/DMU/ZAP versioning.
1351 #define ZPL_VERSION 1ULL
1354 * The directory entry has the type (currently unused on Solaris) in the
1355 * top 4 bits, and the object number in the low 48 bits. The "middle"
1356 * 12 bits are unused.
1358 #define ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4)
1359 #define ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48)
1360 #define ZFS_DIRENT_MAKE(type, obj) (((uint64_t)type << 60) | obj)
1362 typedef struct ace {
1363 uid_t a_who; /* uid or gid */
1364 uint32_t a_access_mask; /* read,write,... */
1365 uint16_t a_flags; /* see below */
1366 uint16_t a_type; /* allow or deny */
1369 #define ACE_SLOT_CNT 6
1371 typedef struct zfs_znode_acl {
1372 uint64_t z_acl_extern_obj; /* ext acl pieces */
1373 uint32_t z_acl_count; /* Number of ACEs */
1374 uint16_t z_acl_version; /* acl version */
1375 uint16_t z_acl_pad; /* pad */
1376 ace_t z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */
1380 * This is the persistent portion of the znode. It is stored
1381 * in the "bonus buffer" of the file. Short symbolic links
1382 * are also stored in the bonus buffer.
1384 typedef struct znode_phys {
1385 uint64_t zp_atime[2]; /* 0 - last file access time */
1386 uint64_t zp_mtime[2]; /* 16 - last file modification time */
1387 uint64_t zp_ctime[2]; /* 32 - last file change time */
1388 uint64_t zp_crtime[2]; /* 48 - creation time */
1389 uint64_t zp_gen; /* 64 - generation (txg of creation) */
1390 uint64_t zp_mode; /* 72 - file mode bits */
1391 uint64_t zp_size; /* 80 - size of file */
1392 uint64_t zp_parent; /* 88 - directory parent (`..') */
1393 uint64_t zp_links; /* 96 - number of links to file */
1394 uint64_t zp_xattr; /* 104 - DMU object for xattrs */
1395 uint64_t zp_rdev; /* 112 - dev_t for VBLK & VCHR files */
1396 uint64_t zp_flags; /* 120 - persistent flags */
1397 uint64_t zp_uid; /* 128 - file owner */
1398 uint64_t zp_gid; /* 136 - owning group */
1399 uint64_t zp_pad[4]; /* 144 - future */
1400 zfs_znode_acl_t zp_acl; /* 176 - 263 ACL */
1402 * Data may pad out any remaining bytes in the znode buffer, eg:
1404 * |<---------------------- dnode_phys (512) ------------------------>|
1405 * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->|
1406 * |<---- znode (264) ---->|<---- data (56) ---->|
1408 * At present, we only use this space to store symbolic links.
1413 * In-core vdev representation.
1416 typedef int vdev_phys_read_t(struct vdev *vdev, void *priv,
1417 off_t offset, void *buf, size_t bytes);
1418 typedef int vdev_read_t(struct vdev *vdev, const blkptr_t *bp,
1419 void *buf, off_t offset, size_t bytes);
1421 typedef STAILQ_HEAD(vdev_list, vdev) vdev_list_t;
1423 typedef struct vdev {
1424 STAILQ_ENTRY(vdev) v_childlink; /* link in parent's child list */
1425 STAILQ_ENTRY(vdev) v_alllink; /* link in global vdev list */
1426 vdev_list_t v_children; /* children of this vdev */
1427 const char *v_name; /* vdev name */
1428 uint64_t v_guid; /* vdev guid */
1429 int v_id; /* index in parent */
1430 int v_ashift; /* offset to block shift */
1431 int v_nparity; /* # parity for raidz */
1432 struct vdev *v_top; /* parent vdev */
1433 int v_nchildren; /* # children */
1434 vdev_state_t v_state; /* current state */
1435 vdev_phys_read_t *v_phys_read; /* read from raw leaf vdev */
1436 vdev_read_t *v_read; /* read from vdev */
1437 void *v_read_priv; /* private data for read function */
1441 * In-core pool representation.
1443 typedef STAILQ_HEAD(spa_list, spa) spa_list_t;
1445 typedef struct spa {
1446 STAILQ_ENTRY(spa) spa_link; /* link in global pool list */
1447 char *spa_name; /* pool name */
1448 uint64_t spa_guid; /* pool guid */
1449 uint64_t spa_txg; /* most recent transaction */
1450 struct uberblock spa_uberblock; /* best uberblock so far */
1451 vdev_list_t spa_vdevs; /* list of all toplevel vdevs */
1452 objset_phys_t spa_mos; /* MOS for this pool */
1453 int spa_inited; /* initialized */
1456 static void decode_embedded_bp_compressed(const blkptr_t *, void *);