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
66 typedef enum { B_FALSE, B_TRUE } boolean_t;
69 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */
72 * Macros for various sorts of alignment and rounding when the alignment
73 * is known to be a power of 2.
75 #define P2ALIGN(x, align) ((x) & -(align))
76 #define P2PHASE(x, align) ((x) & ((align) - 1))
77 #define P2NPHASE(x, align) (-(x) & ((align) - 1))
78 #define P2ROUNDUP(x, align) (-(-(x) & -(align)))
79 #define P2END(x, align) (-(~(x) & -(align)))
80 #define P2PHASEUP(x, align, phase) ((phase) - (((phase) - (x)) & -(align)))
81 #define P2BOUNDARY(off, len, align) (((off) ^ ((off) + (len) - 1)) > (align) - 1)
84 * General-purpose 32-bit and 64-bit bitfield encodings.
86 #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len))
87 #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len))
88 #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low))
89 #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low))
91 #define BF32_GET(x, low, len) BF32_DECODE(x, low, len)
92 #define BF64_GET(x, low, len) BF64_DECODE(x, low, len)
94 #define BF32_SET(x, low, len, val) \
95 ((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len))
96 #define BF64_SET(x, low, len, val) \
97 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len))
99 #define BF32_GET_SB(x, low, len, shift, bias) \
100 ((BF32_GET(x, low, len) + (bias)) << (shift))
101 #define BF64_GET_SB(x, low, len, shift, bias) \
102 ((BF64_GET(x, low, len) + (bias)) << (shift))
104 #define BF32_SET_SB(x, low, len, shift, bias, val) \
105 BF32_SET(x, low, len, ((val) >> (shift)) - (bias))
106 #define BF64_SET_SB(x, low, len, shift, bias, val) \
107 BF64_SET(x, low, len, ((val) >> (shift)) - (bias))
110 * Macros to reverse byte order
112 #define BSWAP_8(x) ((x) & 0xff)
113 #define BSWAP_16(x) ((BSWAP_8(x) << 8) | BSWAP_8((x) >> 8))
114 #define BSWAP_32(x) ((BSWAP_16(x) << 16) | BSWAP_16((x) >> 16))
115 #define BSWAP_64(x) ((BSWAP_32(x) << 32) | BSWAP_32((x) >> 32))
117 #define SPA_MINBLOCKSHIFT 9
118 #define SPA_OLDMAXBLOCKSHIFT 17
119 #define SPA_MAXBLOCKSHIFT 24
120 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT)
121 #define SPA_OLDMAXBLOCKSIZE (1ULL << SPA_OLDMAXBLOCKSHIFT)
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 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept
152 * secret and is suitable for use in MAC algorithms as the key.
154 typedef struct zio_cksum_salt {
155 uint8_t zcs_bytes[32];
159 * Each block is described by its DVAs, time of birth, checksum, etc.
160 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
162 * 64 56 48 40 32 24 16 8 0
163 * +-------+-------+-------+-------+-------+-------+-------+-------+
164 * 0 | vdev1 | GRID | ASIZE |
165 * +-------+-------+-------+-------+-------+-------+-------+-------+
167 * +-------+-------+-------+-------+-------+-------+-------+-------+
168 * 2 | vdev2 | GRID | ASIZE |
169 * +-------+-------+-------+-------+-------+-------+-------+-------+
171 * +-------+-------+-------+-------+-------+-------+-------+-------+
172 * 4 | vdev3 | GRID | ASIZE |
173 * +-------+-------+-------+-------+-------+-------+-------+-------+
175 * +-------+-------+-------+-------+-------+-------+-------+-------+
176 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE |
177 * +-------+-------+-------+-------+-------+-------+-------+-------+
179 * +-------+-------+-------+-------+-------+-------+-------+-------+
181 * +-------+-------+-------+-------+-------+-------+-------+-------+
182 * 9 | physical birth txg |
183 * +-------+-------+-------+-------+-------+-------+-------+-------+
184 * a | logical birth txg |
185 * +-------+-------+-------+-------+-------+-------+-------+-------+
187 * +-------+-------+-------+-------+-------+-------+-------+-------+
189 * +-------+-------+-------+-------+-------+-------+-------+-------+
191 * +-------+-------+-------+-------+-------+-------+-------+-------+
193 * +-------+-------+-------+-------+-------+-------+-------+-------+
195 * +-------+-------+-------+-------+-------+-------+-------+-------+
199 * vdev virtual device ID
200 * offset offset into virtual device
202 * PSIZE physical size (after compression)
203 * ASIZE allocated size (including RAID-Z parity and gang block headers)
204 * GRID RAID-Z layout information (reserved for future use)
205 * cksum checksum function
206 * comp compression function
207 * G gang block indicator
208 * B byteorder (endianness)
210 * X encryption (on version 30, which is not supported)
211 * E blkptr_t contains embedded data (see below)
212 * lvl level of indirection
213 * type DMU object type
214 * phys birth txg of block allocation; zero if same as logical birth txg
215 * log. birth transaction group in which the block was logically born
216 * fill count number of non-zero blocks under this bp
217 * checksum[4] 256-bit checksum of the data this bp describes
221 * "Embedded" blkptr_t's don't actually point to a block, instead they
222 * have a data payload embedded in the blkptr_t itself. See the comment
223 * in blkptr.c for more details.
225 * The blkptr_t is laid out as follows:
227 * 64 56 48 40 32 24 16 8 0
228 * +-------+-------+-------+-------+-------+-------+-------+-------+
235 * +-------+-------+-------+-------+-------+-------+-------+-------+
236 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE |
237 * +-------+-------+-------+-------+-------+-------+-------+-------+
241 * +-------+-------+-------+-------+-------+-------+-------+-------+
242 * a | logical birth txg |
243 * +-------+-------+-------+-------+-------+-------+-------+-------+
249 * +-------+-------+-------+-------+-------+-------+-------+-------+
253 * payload contains the embedded data
254 * B (byteorder) byteorder (endianness)
255 * D (dedup) padding (set to zero)
256 * X encryption (set to zero; see above)
257 * E (embedded) set to one
258 * lvl indirection level
259 * type DMU object type
260 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*)
261 * comp compression function of payload
262 * PSIZE size of payload after compression, in bytes
263 * LSIZE logical size of payload, in bytes
264 * note that 25 bits is enough to store the largest
265 * "normal" BP's LSIZE (2^16 * 2^9) in bytes
266 * log. birth transaction group in which the block was logically born
268 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
269 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
270 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
271 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
272 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must
273 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before
274 * other macros, as they assert that they are only used on BP's of the correct
278 #define BPE_GET_ETYPE(bp) \
279 (ASSERT(BP_IS_EMBEDDED(bp)), \
280 BF64_GET((bp)->blk_prop, 40, 8))
281 #define BPE_SET_ETYPE(bp, t) do { \
282 ASSERT(BP_IS_EMBEDDED(bp)); \
283 BF64_SET((bp)->blk_prop, 40, 8, t); \
284 _NOTE(CONSTCOND) } while (0)
286 #define BPE_GET_LSIZE(bp) \
287 (ASSERT(BP_IS_EMBEDDED(bp)), \
288 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
289 #define BPE_SET_LSIZE(bp, x) do { \
290 ASSERT(BP_IS_EMBEDDED(bp)); \
291 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
292 _NOTE(CONSTCOND) } while (0)
294 #define BPE_GET_PSIZE(bp) \
295 (ASSERT(BP_IS_EMBEDDED(bp)), \
296 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
297 #define BPE_SET_PSIZE(bp, x) do { \
298 ASSERT(BP_IS_EMBEDDED(bp)); \
299 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
300 _NOTE(CONSTCOND) } while (0)
302 typedef enum bp_embedded_type {
303 BP_EMBEDDED_TYPE_DATA,
304 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */
305 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED
306 } bp_embedded_type_t;
308 #define BPE_NUM_WORDS 14
309 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
310 #define BPE_IS_PAYLOADWORD(bp, wp) \
311 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
313 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */
314 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */
316 typedef struct blkptr {
317 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
318 uint64_t blk_prop; /* size, compression, type, etc */
319 uint64_t blk_pad[2]; /* Extra space for the future */
320 uint64_t blk_phys_birth; /* txg when block was allocated */
321 uint64_t blk_birth; /* transaction group at birth */
322 uint64_t blk_fill; /* fill count */
323 zio_cksum_t blk_cksum; /* 256-bit checksum */
327 * Macros to get and set fields in a bp or DVA.
329 #define DVA_GET_ASIZE(dva) \
330 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
331 #define DVA_SET_ASIZE(dva, x) \
332 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
333 SPA_MINBLOCKSHIFT, 0, x)
335 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8)
336 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x)
338 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32)
339 #define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x)
341 #define DVA_GET_OFFSET(dva) \
342 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
343 #define DVA_SET_OFFSET(dva, x) \
344 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
346 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1)
347 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x)
349 #define BP_GET_LSIZE(bp) \
350 (BP_IS_EMBEDDED(bp) ? \
351 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
352 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
353 #define BP_SET_LSIZE(bp, x) do { \
354 ASSERT(!BP_IS_EMBEDDED(bp)); \
355 BF64_SET_SB((bp)->blk_prop, \
356 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
357 _NOTE(CONSTCOND) } while (0)
359 #define BP_GET_PSIZE(bp) \
360 BF64_GET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)
361 #define BP_SET_PSIZE(bp, x) \
362 BF64_SET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x)
364 #define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7)
365 #define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x)
367 #define BP_GET_CHECKSUM(bp) BF64_GET((bp)->blk_prop, 40, 8)
368 #define BP_SET_CHECKSUM(bp, x) BF64_SET((bp)->blk_prop, 40, 8, x)
370 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8)
371 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x)
373 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5)
374 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x)
376 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1)
378 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1)
379 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x)
381 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1)
382 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x)
384 #define BP_PHYSICAL_BIRTH(bp) \
385 ((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
387 #define BP_GET_ASIZE(bp) \
388 (DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
389 DVA_GET_ASIZE(&(bp)->blk_dva[2]))
391 #define BP_GET_UCSIZE(bp) \
392 ((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \
393 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp));
395 #define BP_GET_NDVAS(bp) \
396 (!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
397 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
398 !!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
400 #define DVA_EQUAL(dva1, dva2) \
401 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
402 (dva1)->dva_word[0] == (dva2)->dva_word[0])
404 #define ZIO_CHECKSUM_EQUAL(zc1, zc2) \
405 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
406 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \
407 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \
408 ((zc1).zc_word[3] - (zc2).zc_word[3])))
411 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0)
413 #define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \
415 (zcp)->zc_word[0] = w0; \
416 (zcp)->zc_word[1] = w1; \
417 (zcp)->zc_word[2] = w2; \
418 (zcp)->zc_word[3] = w3; \
421 #define BP_IDENTITY(bp) (&(bp)->blk_dva[0])
422 #define BP_IS_GANG(bp) DVA_GET_GANG(BP_IDENTITY(bp))
423 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \
424 (dva)->dva_word[1] == 0ULL)
425 #define BP_IS_HOLE(bp) DVA_IS_EMPTY(BP_IDENTITY(bp))
426 #define BP_IS_OLDER(bp, txg) (!BP_IS_HOLE(bp) && (bp)->blk_birth < (txg))
428 #define BP_ZERO(bp) \
430 (bp)->blk_dva[0].dva_word[0] = 0; \
431 (bp)->blk_dva[0].dva_word[1] = 0; \
432 (bp)->blk_dva[1].dva_word[0] = 0; \
433 (bp)->blk_dva[1].dva_word[1] = 0; \
434 (bp)->blk_dva[2].dva_word[0] = 0; \
435 (bp)->blk_dva[2].dva_word[1] = 0; \
436 (bp)->blk_prop = 0; \
437 (bp)->blk_pad[0] = 0; \
438 (bp)->blk_pad[1] = 0; \
439 (bp)->blk_phys_birth = 0; \
440 (bp)->blk_birth = 0; \
441 (bp)->blk_fill = 0; \
442 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \
445 #define BPE_NUM_WORDS 14
446 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
447 #define BPE_IS_PAYLOADWORD(bp, wp) \
448 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
453 #define ZEC_MAGIC 0x210da7ab10c7a11ULL
455 typedef struct zio_eck {
456 uint64_t zec_magic; /* for validation, endianness */
457 zio_cksum_t zec_cksum; /* 256-bit checksum */
461 * Gang block headers are self-checksumming and contain an array
464 #define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE
465 #define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \
466 sizeof (zio_eck_t)) / sizeof (blkptr_t))
467 #define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \
468 sizeof (zio_eck_t) - \
469 (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\
472 typedef struct zio_gbh {
473 blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS];
474 uint64_t zg_filler[SPA_GBH_FILLER];
478 #define VDEV_RAIDZ_MAXPARITY 3
480 #define VDEV_PAD_SIZE (8 << 10)
481 /* 2 padding areas (vl_pad1 and vl_pad2) to skip */
482 #define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2
483 #define VDEV_PHYS_SIZE (112 << 10)
484 #define VDEV_UBERBLOCK_RING (128 << 10)
486 #define VDEV_UBERBLOCK_SHIFT(vd) \
487 MAX((vd)->v_top->v_ashift, UBERBLOCK_SHIFT)
488 #define VDEV_UBERBLOCK_COUNT(vd) \
489 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
490 #define VDEV_UBERBLOCK_OFFSET(vd, n) \
491 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
492 #define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd))
494 typedef struct vdev_phys {
495 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)];
499 typedef struct vdev_label {
500 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */
501 char vl_pad2[VDEV_PAD_SIZE]; /* 8K */
502 vdev_phys_t vl_vdev_phys; /* 112K */
503 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */
504 } vdev_label_t; /* 256K total */
509 #define VDD_METASLAB 0x01
513 * Size and offset of embedded boot loader region on each label.
514 * The total size of the first two labels plus the boot area is 4MB.
516 #define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t))
517 #define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */
520 * Size of label regions at the start and end of each leaf device.
522 #define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
523 #define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t))
524 #define VDEV_LABELS 4
527 ZIO_CHECKSUM_INHERIT = 0,
531 ZIO_CHECKSUM_GANG_HEADER,
533 ZIO_CHECKSUM_FLETCHER_2,
534 ZIO_CHECKSUM_FLETCHER_4,
537 ZIO_CHECKSUM_NOPARITY,
541 ZIO_CHECKSUM_FUNCTIONS
544 #define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4
545 #define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON
548 ZIO_COMPRESS_INHERIT = 0,
564 ZIO_COMPRESS_FUNCTIONS
567 #define ZIO_COMPRESS_ON_VALUE ZIO_COMPRESS_LZJB
568 #define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF
570 /* nvlist pack encoding */
571 #define NV_ENCODE_NATIVE 0
572 #define NV_ENCODE_XDR 1
575 DATA_TYPE_UNKNOWN = 0,
585 DATA_TYPE_BYTE_ARRAY,
586 DATA_TYPE_INT16_ARRAY,
587 DATA_TYPE_UINT16_ARRAY,
588 DATA_TYPE_INT32_ARRAY,
589 DATA_TYPE_UINT32_ARRAY,
590 DATA_TYPE_INT64_ARRAY,
591 DATA_TYPE_UINT64_ARRAY,
592 DATA_TYPE_STRING_ARRAY,
595 DATA_TYPE_NVLIST_ARRAY,
596 DATA_TYPE_BOOLEAN_VALUE,
599 DATA_TYPE_BOOLEAN_ARRAY,
600 DATA_TYPE_INT8_ARRAY,
601 DATA_TYPE_UINT8_ARRAY
605 * On-disk version number.
607 #define SPA_VERSION_1 1ULL
608 #define SPA_VERSION_2 2ULL
609 #define SPA_VERSION_3 3ULL
610 #define SPA_VERSION_4 4ULL
611 #define SPA_VERSION_5 5ULL
612 #define SPA_VERSION_6 6ULL
613 #define SPA_VERSION_7 7ULL
614 #define SPA_VERSION_8 8ULL
615 #define SPA_VERSION_9 9ULL
616 #define SPA_VERSION_10 10ULL
617 #define SPA_VERSION_11 11ULL
618 #define SPA_VERSION_12 12ULL
619 #define SPA_VERSION_13 13ULL
620 #define SPA_VERSION_14 14ULL
621 #define SPA_VERSION_15 15ULL
622 #define SPA_VERSION_16 16ULL
623 #define SPA_VERSION_17 17ULL
624 #define SPA_VERSION_18 18ULL
625 #define SPA_VERSION_19 19ULL
626 #define SPA_VERSION_20 20ULL
627 #define SPA_VERSION_21 21ULL
628 #define SPA_VERSION_22 22ULL
629 #define SPA_VERSION_23 23ULL
630 #define SPA_VERSION_24 24ULL
631 #define SPA_VERSION_25 25ULL
632 #define SPA_VERSION_26 26ULL
633 #define SPA_VERSION_27 27ULL
634 #define SPA_VERSION_28 28ULL
635 #define SPA_VERSION_5000 5000ULL
638 * When bumping up SPA_VERSION, make sure GRUB ZFS understands the on-disk
639 * format change. Go to usr/src/grub/grub-0.97/stage2/{zfs-include/, fsys_zfs*},
640 * and do the appropriate changes. Also bump the version number in
641 * usr/src/grub/capability.
643 #define SPA_VERSION SPA_VERSION_5000
644 #define SPA_VERSION_STRING "5000"
647 * Symbolic names for the changes that caused a SPA_VERSION switch.
648 * Used in the code when checking for presence or absence of a feature.
649 * Feel free to define multiple symbolic names for each version if there
650 * were multiple changes to on-disk structures during that version.
652 * NOTE: When checking the current SPA_VERSION in your code, be sure
653 * to use spa_version() since it reports the version of the
654 * last synced uberblock. Checking the in-flight version can
655 * be dangerous in some cases.
657 #define SPA_VERSION_INITIAL SPA_VERSION_1
658 #define SPA_VERSION_DITTO_BLOCKS SPA_VERSION_2
659 #define SPA_VERSION_SPARES SPA_VERSION_3
660 #define SPA_VERSION_RAID6 SPA_VERSION_3
661 #define SPA_VERSION_BPLIST_ACCOUNT SPA_VERSION_3
662 #define SPA_VERSION_RAIDZ_DEFLATE SPA_VERSION_3
663 #define SPA_VERSION_DNODE_BYTES SPA_VERSION_3
664 #define SPA_VERSION_ZPOOL_HISTORY SPA_VERSION_4
665 #define SPA_VERSION_GZIP_COMPRESSION SPA_VERSION_5
666 #define SPA_VERSION_BOOTFS SPA_VERSION_6
667 #define SPA_VERSION_SLOGS SPA_VERSION_7
668 #define SPA_VERSION_DELEGATED_PERMS SPA_VERSION_8
669 #define SPA_VERSION_FUID SPA_VERSION_9
670 #define SPA_VERSION_REFRESERVATION SPA_VERSION_9
671 #define SPA_VERSION_REFQUOTA SPA_VERSION_9
672 #define SPA_VERSION_UNIQUE_ACCURATE SPA_VERSION_9
673 #define SPA_VERSION_L2CACHE SPA_VERSION_10
674 #define SPA_VERSION_NEXT_CLONES SPA_VERSION_11
675 #define SPA_VERSION_ORIGIN SPA_VERSION_11
676 #define SPA_VERSION_DSL_SCRUB SPA_VERSION_11
677 #define SPA_VERSION_SNAP_PROPS SPA_VERSION_12
678 #define SPA_VERSION_USED_BREAKDOWN SPA_VERSION_13
679 #define SPA_VERSION_PASSTHROUGH_X SPA_VERSION_14
680 #define SPA_VERSION_USERSPACE SPA_VERSION_15
681 #define SPA_VERSION_STMF_PROP SPA_VERSION_16
682 #define SPA_VERSION_RAIDZ3 SPA_VERSION_17
683 #define SPA_VERSION_USERREFS SPA_VERSION_18
684 #define SPA_VERSION_HOLES SPA_VERSION_19
685 #define SPA_VERSION_ZLE_COMPRESSION SPA_VERSION_20
686 #define SPA_VERSION_DEDUP SPA_VERSION_21
687 #define SPA_VERSION_RECVD_PROPS SPA_VERSION_22
688 #define SPA_VERSION_SLIM_ZIL SPA_VERSION_23
689 #define SPA_VERSION_SA SPA_VERSION_24
690 #define SPA_VERSION_SCAN SPA_VERSION_25
691 #define SPA_VERSION_DIR_CLONES SPA_VERSION_26
692 #define SPA_VERSION_DEADLISTS SPA_VERSION_26
693 #define SPA_VERSION_FAST_SNAP SPA_VERSION_27
694 #define SPA_VERSION_MULTI_REPLACE SPA_VERSION_28
695 #define SPA_VERSION_BEFORE_FEATURES SPA_VERSION_28
696 #define SPA_VERSION_FEATURES SPA_VERSION_5000
698 #define SPA_VERSION_IS_SUPPORTED(v) \
699 (((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \
700 ((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION))
703 * The following are configuration names used in the nvlist describing a pool's
706 #define ZPOOL_CONFIG_VERSION "version"
707 #define ZPOOL_CONFIG_POOL_NAME "name"
708 #define ZPOOL_CONFIG_POOL_STATE "state"
709 #define ZPOOL_CONFIG_POOL_TXG "txg"
710 #define ZPOOL_CONFIG_POOL_GUID "pool_guid"
711 #define ZPOOL_CONFIG_CREATE_TXG "create_txg"
712 #define ZPOOL_CONFIG_TOP_GUID "top_guid"
713 #define ZPOOL_CONFIG_VDEV_TREE "vdev_tree"
714 #define ZPOOL_CONFIG_TYPE "type"
715 #define ZPOOL_CONFIG_CHILDREN "children"
716 #define ZPOOL_CONFIG_ID "id"
717 #define ZPOOL_CONFIG_GUID "guid"
718 #define ZPOOL_CONFIG_PATH "path"
719 #define ZPOOL_CONFIG_DEVID "devid"
720 #define ZPOOL_CONFIG_METASLAB_ARRAY "metaslab_array"
721 #define ZPOOL_CONFIG_METASLAB_SHIFT "metaslab_shift"
722 #define ZPOOL_CONFIG_ASHIFT "ashift"
723 #define ZPOOL_CONFIG_ASIZE "asize"
724 #define ZPOOL_CONFIG_DTL "DTL"
725 #define ZPOOL_CONFIG_STATS "stats"
726 #define ZPOOL_CONFIG_WHOLE_DISK "whole_disk"
727 #define ZPOOL_CONFIG_ERRCOUNT "error_count"
728 #define ZPOOL_CONFIG_NOT_PRESENT "not_present"
729 #define ZPOOL_CONFIG_SPARES "spares"
730 #define ZPOOL_CONFIG_IS_SPARE "is_spare"
731 #define ZPOOL_CONFIG_NPARITY "nparity"
732 #define ZPOOL_CONFIG_HOSTID "hostid"
733 #define ZPOOL_CONFIG_HOSTNAME "hostname"
734 #define ZPOOL_CONFIG_IS_LOG "is_log"
735 #define ZPOOL_CONFIG_TIMESTAMP "timestamp" /* not stored on disk */
736 #define ZPOOL_CONFIG_FEATURES_FOR_READ "features_for_read"
739 * The persistent vdev state is stored as separate values rather than a single
740 * 'vdev_state' entry. This is because a device can be in multiple states, such
741 * as offline and degraded.
743 #define ZPOOL_CONFIG_OFFLINE "offline"
744 #define ZPOOL_CONFIG_FAULTED "faulted"
745 #define ZPOOL_CONFIG_DEGRADED "degraded"
746 #define ZPOOL_CONFIG_REMOVED "removed"
747 #define ZPOOL_CONFIG_FRU "fru"
748 #define ZPOOL_CONFIG_AUX_STATE "aux_state"
750 #define VDEV_TYPE_ROOT "root"
751 #define VDEV_TYPE_MIRROR "mirror"
752 #define VDEV_TYPE_REPLACING "replacing"
753 #define VDEV_TYPE_RAIDZ "raidz"
754 #define VDEV_TYPE_DISK "disk"
755 #define VDEV_TYPE_FILE "file"
756 #define VDEV_TYPE_MISSING "missing"
757 #define VDEV_TYPE_HOLE "hole"
758 #define VDEV_TYPE_SPARE "spare"
759 #define VDEV_TYPE_LOG "log"
760 #define VDEV_TYPE_L2CACHE "l2cache"
763 * This is needed in userland to report the minimum necessary device size.
765 #define SPA_MINDEVSIZE (64ULL << 20)
768 * The location of the pool configuration repository, shared between kernel and
771 #define ZPOOL_CACHE "/boot/zfs/zpool.cache"
774 * vdev states are ordered from least to most healthy.
775 * A vdev that's CANT_OPEN or below is considered unusable.
777 typedef enum vdev_state {
778 VDEV_STATE_UNKNOWN = 0, /* Uninitialized vdev */
779 VDEV_STATE_CLOSED, /* Not currently open */
780 VDEV_STATE_OFFLINE, /* Not allowed to open */
781 VDEV_STATE_REMOVED, /* Explicitly removed from system */
782 VDEV_STATE_CANT_OPEN, /* Tried to open, but failed */
783 VDEV_STATE_FAULTED, /* External request to fault device */
784 VDEV_STATE_DEGRADED, /* Replicated vdev with unhealthy kids */
785 VDEV_STATE_HEALTHY /* Presumed good */
789 * vdev aux states. When a vdev is in the CANT_OPEN state, the aux field
790 * of the vdev stats structure uses these constants to distinguish why.
792 typedef enum vdev_aux {
793 VDEV_AUX_NONE, /* no error */
794 VDEV_AUX_OPEN_FAILED, /* ldi_open_*() or vn_open() failed */
795 VDEV_AUX_CORRUPT_DATA, /* bad label or disk contents */
796 VDEV_AUX_NO_REPLICAS, /* insufficient number of replicas */
797 VDEV_AUX_BAD_GUID_SUM, /* vdev guid sum doesn't match */
798 VDEV_AUX_TOO_SMALL, /* vdev size is too small */
799 VDEV_AUX_BAD_LABEL, /* the label is OK but invalid */
800 VDEV_AUX_VERSION_NEWER, /* on-disk version is too new */
801 VDEV_AUX_VERSION_OLDER, /* on-disk version is too old */
802 VDEV_AUX_SPARED /* hot spare used in another pool */
806 * pool state. The following states are written to disk as part of the normal
807 * SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE. The remaining states are
808 * software abstractions used at various levels to communicate pool state.
810 typedef enum pool_state {
811 POOL_STATE_ACTIVE = 0, /* In active use */
812 POOL_STATE_EXPORTED, /* Explicitly exported */
813 POOL_STATE_DESTROYED, /* Explicitly destroyed */
814 POOL_STATE_SPARE, /* Reserved for hot spare use */
815 POOL_STATE_UNINITIALIZED, /* Internal spa_t state */
816 POOL_STATE_UNAVAIL, /* Internal libzfs state */
817 POOL_STATE_POTENTIALLY_ACTIVE /* Internal libzfs state */
821 * The uberblock version is incremented whenever an incompatible on-disk
822 * format change is made to the SPA, DMU, or ZAP.
824 * Note: the first two fields should never be moved. When a storage pool
825 * is opened, the uberblock must be read off the disk before the version
826 * can be checked. If the ub_version field is moved, we may not detect
827 * version mismatch. If the ub_magic field is moved, applications that
828 * expect the magic number in the first word won't work.
830 #define UBERBLOCK_MAGIC 0x00bab10c /* oo-ba-bloc! */
831 #define UBERBLOCK_SHIFT 10 /* up to 1K */
834 uint64_t ub_magic; /* UBERBLOCK_MAGIC */
835 uint64_t ub_version; /* SPA_VERSION */
836 uint64_t ub_txg; /* txg of last sync */
837 uint64_t ub_guid_sum; /* sum of all vdev guids */
838 uint64_t ub_timestamp; /* UTC time of last sync */
839 blkptr_t ub_rootbp; /* MOS objset_phys_t */
845 #define DNODE_MUST_BE_ALLOCATED 1
846 #define DNODE_MUST_BE_FREE 2
851 #define DNODE_SHIFT 9 /* 512 bytes */
852 #define DN_MIN_INDBLKSHIFT 12 /* 4k */
853 #define DN_MAX_INDBLKSHIFT 14 /* 16k */
854 #define DNODE_BLOCK_SHIFT 14 /* 16k */
855 #define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */
856 #define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */
857 #define DN_MAX_OFFSET_SHIFT 64 /* 2^64 bytes in a dnode */
862 #define DNODE_SIZE (1 << DNODE_SHIFT)
863 #define DN_MAX_NBLKPTR ((DNODE_SIZE - DNODE_CORE_SIZE) >> SPA_BLKPTRSHIFT)
864 #define DN_MAX_BONUSLEN (DNODE_SIZE - DNODE_CORE_SIZE - (1 << SPA_BLKPTRSHIFT))
865 #define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT)
867 #define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT)
868 #define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT)
869 #define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT)
871 /* The +2 here is a cheesy way to round up */
872 #define DN_MAX_LEVELS (2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \
873 (DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT)))
875 #define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \
876 (((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t))))
878 #define DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \
879 (dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT)
881 #define EPB(blkshift, typeshift) (1 << (blkshift - typeshift))
883 /* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */
884 #define DNODE_FLAG_USED_BYTES (1<<0)
885 #define DNODE_FLAG_USERUSED_ACCOUNTED (1<<1)
887 /* Does dnode have a SA spill blkptr in bonus? */
888 #define DNODE_FLAG_SPILL_BLKPTR (1<<2)
890 typedef struct dnode_phys {
891 uint8_t dn_type; /* dmu_object_type_t */
892 uint8_t dn_indblkshift; /* ln2(indirect block size) */
893 uint8_t dn_nlevels; /* 1=dn_blkptr->data blocks */
894 uint8_t dn_nblkptr; /* length of dn_blkptr */
895 uint8_t dn_bonustype; /* type of data in bonus buffer */
896 uint8_t dn_checksum; /* ZIO_CHECKSUM type */
897 uint8_t dn_compress; /* ZIO_COMPRESS type */
898 uint8_t dn_flags; /* DNODE_FLAG_* */
899 uint16_t dn_datablkszsec; /* data block size in 512b sectors */
900 uint16_t dn_bonuslen; /* length of dn_bonus */
903 /* accounting is protected by dn_dirty_mtx */
904 uint64_t dn_maxblkid; /* largest allocated block ID */
905 uint64_t dn_used; /* bytes (or sectors) of disk space */
909 blkptr_t dn_blkptr[1];
910 uint8_t dn_bonus[DN_MAX_BONUSLEN - sizeof (blkptr_t)];
914 typedef enum dmu_object_byteswap {
926 * Allocating a new byteswap type number makes the on-disk format
927 * incompatible with any other format that uses the same number.
929 * Data can usually be structured to work with one of the
930 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types.
933 } dmu_object_byteswap_t;
935 #define DMU_OT_NEWTYPE 0x80
936 #define DMU_OT_METADATA 0x40
937 #define DMU_OT_BYTESWAP_MASK 0x3f
940 * Defines a uint8_t object type. Object types specify if the data
941 * in the object is metadata (boolean) and how to byteswap the data
942 * (dmu_object_byteswap_t).
944 #define DMU_OT(byteswap, metadata) \
946 ((metadata) ? DMU_OT_METADATA : 0) | \
947 ((byteswap) & DMU_OT_BYTESWAP_MASK))
949 typedef enum dmu_object_type {
952 DMU_OT_OBJECT_DIRECTORY, /* ZAP */
953 DMU_OT_OBJECT_ARRAY, /* UINT64 */
954 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */
955 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */
956 DMU_OT_BPLIST, /* UINT64 */
957 DMU_OT_BPLIST_HDR, /* UINT64 */
959 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */
960 DMU_OT_SPACE_MAP, /* UINT64 */
962 DMU_OT_INTENT_LOG, /* UINT64 */
964 DMU_OT_DNODE, /* DNODE */
965 DMU_OT_OBJSET, /* OBJSET */
967 DMU_OT_DSL_DIR, /* UINT64 */
968 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */
969 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */
970 DMU_OT_DSL_PROPS, /* ZAP */
971 DMU_OT_DSL_DATASET, /* UINT64 */
973 DMU_OT_ZNODE, /* ZNODE */
974 DMU_OT_OLDACL, /* Old ACL */
975 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */
976 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */
977 DMU_OT_MASTER_NODE, /* ZAP */
978 DMU_OT_UNLINKED_SET, /* ZAP */
980 DMU_OT_ZVOL, /* UINT8 */
981 DMU_OT_ZVOL_PROP, /* ZAP */
982 /* other; for testing only! */
983 DMU_OT_PLAIN_OTHER, /* UINT8 */
984 DMU_OT_UINT64_OTHER, /* UINT64 */
985 DMU_OT_ZAP_OTHER, /* ZAP */
986 /* new object types: */
987 DMU_OT_ERROR_LOG, /* ZAP */
988 DMU_OT_SPA_HISTORY, /* UINT8 */
989 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */
990 DMU_OT_POOL_PROPS, /* ZAP */
991 DMU_OT_DSL_PERMS, /* ZAP */
992 DMU_OT_ACL, /* ACL */
993 DMU_OT_SYSACL, /* SYSACL */
994 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */
995 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */
996 DMU_OT_NEXT_CLONES, /* ZAP */
997 DMU_OT_SCAN_QUEUE, /* ZAP */
998 DMU_OT_USERGROUP_USED, /* ZAP */
999 DMU_OT_USERGROUP_QUOTA, /* ZAP */
1000 DMU_OT_USERREFS, /* ZAP */
1001 DMU_OT_DDT_ZAP, /* ZAP */
1002 DMU_OT_DDT_STATS, /* ZAP */
1003 DMU_OT_SA, /* System attr */
1004 DMU_OT_SA_MASTER_NODE, /* ZAP */
1005 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */
1006 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */
1007 DMU_OT_SCAN_XLATE, /* ZAP */
1008 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */
1012 * Names for valid types declared with DMU_OT().
1014 DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE),
1015 DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE),
1016 DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE),
1017 DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE),
1018 DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE),
1019 DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE),
1020 DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE),
1021 DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE),
1022 DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE),
1023 DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE)
1024 } dmu_object_type_t;
1026 typedef enum dmu_objset_type {
1031 DMU_OST_OTHER, /* For testing only! */
1032 DMU_OST_ANY, /* Be careful! */
1034 } dmu_objset_type_t;
1037 * header for all bonus and spill buffers.
1038 * The header has a fixed portion with a variable number
1039 * of "lengths" depending on the number of variable sized
1040 * attribues which are determined by the "layout number"
1043 #define SA_MAGIC 0x2F505A /* ZFS SA */
1044 typedef struct sa_hdr_phys {
1046 uint16_t sa_layout_info; /* Encoded with hdrsize and layout number */
1047 uint16_t sa_lengths[1]; /* optional sizes for variable length attrs */
1048 /* ... Data follows the lengths. */
1052 * sa_hdr_phys -> sa_layout_info
1055 * +--------+-------+
1057 * +--------+-------+
1059 * Bits 0-10 are the layout number
1060 * Bits 11-16 are the size of the header.
1061 * The hdrsize is the number * 8
1064 * hdrsz of 1 ==> 8 byte header
1065 * 2 ==> 16 byte header
1069 #define SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10)
1070 #define SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 16, 3, 0)
1071 #define SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \
1073 BF32_SET_SB(x, 10, 6, 3, 0, size); \
1074 BF32_SET(x, 0, 10, num); \
1077 #define SA_MODE_OFFSET 0
1078 #define SA_SIZE_OFFSET 8
1079 #define SA_GEN_OFFSET 16
1080 #define SA_UID_OFFSET 24
1081 #define SA_GID_OFFSET 32
1082 #define SA_PARENT_OFFSET 40
1085 * Intent log header - this on disk structure holds fields to manage
1086 * the log. All fields are 64 bit to easily handle cross architectures.
1088 typedef struct zil_header {
1089 uint64_t zh_claim_txg; /* txg in which log blocks were claimed */
1090 uint64_t zh_replay_seq; /* highest replayed sequence number */
1091 blkptr_t zh_log; /* log chain */
1092 uint64_t zh_claim_seq; /* highest claimed sequence number */
1096 #define OBJSET_PHYS_SIZE 2048
1098 typedef struct objset_phys {
1099 dnode_phys_t os_meta_dnode;
1100 zil_header_t os_zil_header;
1103 char os_pad[OBJSET_PHYS_SIZE - sizeof (dnode_phys_t)*3 -
1104 sizeof (zil_header_t) - sizeof (uint64_t)*2];
1105 dnode_phys_t os_userused_dnode;
1106 dnode_phys_t os_groupused_dnode;
1109 typedef struct dsl_dir_phys {
1110 uint64_t dd_creation_time; /* not actually used */
1111 uint64_t dd_head_dataset_obj;
1112 uint64_t dd_parent_obj;
1113 uint64_t dd_clone_parent_obj;
1114 uint64_t dd_child_dir_zapobj;
1116 * how much space our children are accounting for; for leaf
1117 * datasets, == physical space used by fs + snaps
1119 uint64_t dd_used_bytes;
1120 uint64_t dd_compressed_bytes;
1121 uint64_t dd_uncompressed_bytes;
1122 /* Administrative quota setting */
1124 /* Administrative reservation setting */
1125 uint64_t dd_reserved;
1126 uint64_t dd_props_zapobj;
1127 uint64_t dd_pad[21]; /* pad out to 256 bytes for good measure */
1130 typedef struct dsl_dataset_phys {
1131 uint64_t ds_dir_obj;
1132 uint64_t ds_prev_snap_obj;
1133 uint64_t ds_prev_snap_txg;
1134 uint64_t ds_next_snap_obj;
1135 uint64_t ds_snapnames_zapobj; /* zap obj of snaps; ==0 for snaps */
1136 uint64_t ds_num_children; /* clone/snap children; ==0 for head */
1137 uint64_t ds_creation_time; /* seconds since 1970 */
1138 uint64_t ds_creation_txg;
1139 uint64_t ds_deadlist_obj;
1140 uint64_t ds_used_bytes;
1141 uint64_t ds_compressed_bytes;
1142 uint64_t ds_uncompressed_bytes;
1143 uint64_t ds_unique_bytes; /* only relevant to snapshots */
1145 * The ds_fsid_guid is a 56-bit ID that can change to avoid
1146 * collisions. The ds_guid is a 64-bit ID that will never
1147 * change, so there is a small probability that it will collide.
1149 uint64_t ds_fsid_guid;
1153 uint64_t ds_pad[8]; /* pad out to 320 bytes for good measure */
1154 } dsl_dataset_phys_t;
1157 * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
1159 #define DMU_POOL_DIRECTORY_OBJECT 1
1160 #define DMU_POOL_CONFIG "config"
1161 #define DMU_POOL_FEATURES_FOR_READ "features_for_read"
1162 #define DMU_POOL_ROOT_DATASET "root_dataset"
1163 #define DMU_POOL_SYNC_BPLIST "sync_bplist"
1164 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub"
1165 #define DMU_POOL_ERRLOG_LAST "errlog_last"
1166 #define DMU_POOL_SPARES "spares"
1167 #define DMU_POOL_DEFLATE "deflate"
1168 #define DMU_POOL_HISTORY "history"
1169 #define DMU_POOL_PROPS "pool_props"
1170 #define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt"
1172 #define ZAP_MAGIC 0x2F52AB2ABULL
1174 #define FZAP_BLOCK_SHIFT(zap) ((zap)->zap_block_shift)
1176 #define ZAP_MAXCD (uint32_t)(-1)
1177 #define ZAP_HASHBITS 28
1178 #define MZAP_ENT_LEN 64
1179 #define MZAP_NAME_LEN (MZAP_ENT_LEN - 8 - 4 - 2)
1180 #define MZAP_MAX_BLKSHIFT SPA_MAXBLOCKSHIFT
1181 #define MZAP_MAX_BLKSZ (1 << MZAP_MAX_BLKSHIFT)
1183 typedef struct mzap_ent_phys {
1186 uint16_t mze_pad; /* in case we want to chain them someday */
1187 char mze_name[MZAP_NAME_LEN];
1190 typedef struct mzap_phys {
1191 uint64_t mz_block_type; /* ZBT_MICRO */
1194 mzap_ent_phys_t mz_chunk[1];
1195 /* actually variable size depending on block size */
1199 * The (fat) zap is stored in one object. It is an array of
1200 * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of:
1202 * ptrtbl fits in first block:
1203 * [zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ...
1205 * ptrtbl too big for first block:
1206 * [zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ...
1210 #define ZBT_LEAF ((1ULL << 63) + 0)
1211 #define ZBT_HEADER ((1ULL << 63) + 1)
1212 #define ZBT_MICRO ((1ULL << 63) + 3)
1213 /* any other values are ptrtbl blocks */
1216 * the embedded pointer table takes up half a block:
1217 * block size / entry size (2^3) / 2
1219 #define ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1)
1222 * The embedded pointer table starts half-way through the block. Since
1223 * the pointer table itself is half the block, it starts at (64-bit)
1224 * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)).
1226 #define ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \
1227 ((uint64_t *)(zap)->zap_phys) \
1228 [(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))]
1232 * If zap_phys_t is modified, zap_byteswap() must be modified.
1234 typedef struct zap_phys {
1235 uint64_t zap_block_type; /* ZBT_HEADER */
1236 uint64_t zap_magic; /* ZAP_MAGIC */
1238 struct zap_table_phys {
1239 uint64_t zt_blk; /* starting block number */
1240 uint64_t zt_numblks; /* number of blocks */
1241 uint64_t zt_shift; /* bits to index it */
1242 uint64_t zt_nextblk; /* next (larger) copy start block */
1243 uint64_t zt_blks_copied; /* number source blocks copied */
1246 uint64_t zap_freeblk; /* the next free block */
1247 uint64_t zap_num_leafs; /* number of leafs */
1248 uint64_t zap_num_entries; /* number of entries */
1249 uint64_t zap_salt; /* salt to stir into hash function */
1251 * This structure is followed by padding, and then the embedded
1252 * pointer table. The embedded pointer table takes up second
1253 * half of the block. It is accessed using the
1254 * ZAP_EMBEDDED_PTRTBL_ENT() macro.
1258 typedef struct zap_table_phys zap_table_phys_t;
1260 typedef struct fat_zap {
1261 int zap_block_shift; /* block size shift */
1262 zap_phys_t *zap_phys;
1265 #define ZAP_LEAF_MAGIC 0x2AB1EAF
1267 /* chunk size = 24 bytes */
1268 #define ZAP_LEAF_CHUNKSIZE 24
1271 * The amount of space available for chunks is:
1272 * block size (1<<l->l_bs) - hash entry size (2) * number of hash
1273 * entries - header space (2*chunksize)
1275 #define ZAP_LEAF_NUMCHUNKS(l) \
1276 (((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \
1277 ZAP_LEAF_CHUNKSIZE - 2)
1280 * The amount of space within the chunk available for the array is:
1281 * chunk size - space for type (1) - space for next pointer (2)
1283 #define ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3)
1285 #define ZAP_LEAF_ARRAY_NCHUNKS(bytes) \
1286 (((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES)
1289 * Low water mark: when there are only this many chunks free, start
1290 * growing the ptrtbl. Ideally, this should be larger than a
1291 * "reasonably-sized" entry. 20 chunks is more than enough for the
1292 * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value),
1293 * while still being only around 3% for 16k blocks.
1295 #define ZAP_LEAF_LOW_WATER (20)
1298 * The leaf hash table has block size / 2^5 (32) number of entries,
1299 * which should be more than enough for the maximum number of entries,
1300 * which is less than block size / CHUNKSIZE (24) / minimum number of
1301 * chunks per entry (3).
1303 #define ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5)
1304 #define ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l))
1307 * The chunks start immediately after the hash table. The end of the
1308 * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a
1311 #define ZAP_LEAF_CHUNK(l, idx) \
1312 ((zap_leaf_chunk_t *) \
1313 ((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx]
1314 #define ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry)
1316 typedef enum zap_chunk_type {
1317 ZAP_CHUNK_FREE = 253,
1318 ZAP_CHUNK_ENTRY = 252,
1319 ZAP_CHUNK_ARRAY = 251,
1320 ZAP_CHUNK_TYPE_MAX = 250
1325 * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified.
1327 typedef struct zap_leaf_phys {
1328 struct zap_leaf_header {
1329 uint64_t lh_block_type; /* ZBT_LEAF */
1331 uint64_t lh_prefix; /* hash prefix of this leaf */
1332 uint32_t lh_magic; /* ZAP_LEAF_MAGIC */
1333 uint16_t lh_nfree; /* number free chunks */
1334 uint16_t lh_nentries; /* number of entries */
1335 uint16_t lh_prefix_len; /* num bits used to id this */
1337 /* above is accessable to zap, below is zap_leaf private */
1339 uint16_t lh_freelist; /* chunk head of free list */
1340 uint8_t lh_pad2[12];
1341 } l_hdr; /* 2 24-byte chunks */
1344 * The header is followed by a hash table with
1345 * ZAP_LEAF_HASH_NUMENTRIES(zap) entries. The hash table is
1346 * followed by an array of ZAP_LEAF_NUMCHUNKS(zap)
1347 * zap_leaf_chunk structures. These structures are accessed
1348 * with the ZAP_LEAF_CHUNK() macro.
1354 typedef union zap_leaf_chunk {
1355 struct zap_leaf_entry {
1356 uint8_t le_type; /* always ZAP_CHUNK_ENTRY */
1357 uint8_t le_value_intlen; /* size of ints */
1358 uint16_t le_next; /* next entry in hash chain */
1359 uint16_t le_name_chunk; /* first chunk of the name */
1360 uint16_t le_name_numints; /* bytes in name, incl null */
1361 uint16_t le_value_chunk; /* first chunk of the value */
1362 uint16_t le_value_numints; /* value length in ints */
1363 uint32_t le_cd; /* collision differentiator */
1364 uint64_t le_hash; /* hash value of the name */
1366 struct zap_leaf_array {
1367 uint8_t la_type; /* always ZAP_CHUNK_ARRAY */
1368 uint8_t la_array[ZAP_LEAF_ARRAY_BYTES];
1369 uint16_t la_next; /* next blk or CHAIN_END */
1371 struct zap_leaf_free {
1372 uint8_t lf_type; /* always ZAP_CHUNK_FREE */
1373 uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES];
1374 uint16_t lf_next; /* next in free list, or CHAIN_END */
1378 typedef struct zap_leaf {
1379 int l_bs; /* block size shift */
1380 zap_leaf_phys_t *l_phys;
1384 * Define special zfs pflags
1386 #define ZFS_XATTR 0x1 /* is an extended attribute */
1387 #define ZFS_INHERIT_ACE 0x2 /* ace has inheritable ACEs */
1388 #define ZFS_ACL_TRIVIAL 0x4 /* files ACL is trivial */
1390 #define MASTER_NODE_OBJ 1
1393 * special attributes for master node.
1396 #define ZFS_FSID "FSID"
1397 #define ZFS_UNLINKED_SET "DELETE_QUEUE"
1398 #define ZFS_ROOT_OBJ "ROOT"
1399 #define ZPL_VERSION_OBJ "VERSION"
1400 #define ZFS_PROP_BLOCKPERPAGE "BLOCKPERPAGE"
1401 #define ZFS_PROP_NOGROWBLOCKS "NOGROWBLOCKS"
1403 #define ZFS_FLAG_BLOCKPERPAGE 0x1
1404 #define ZFS_FLAG_NOGROWBLOCKS 0x2
1407 * ZPL version - rev'd whenever an incompatible on-disk format change
1408 * occurs. Independent of SPA/DMU/ZAP versioning.
1411 #define ZPL_VERSION 1ULL
1414 * The directory entry has the type (currently unused on Solaris) in the
1415 * top 4 bits, and the object number in the low 48 bits. The "middle"
1416 * 12 bits are unused.
1418 #define ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4)
1419 #define ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48)
1420 #define ZFS_DIRENT_MAKE(type, obj) (((uint64_t)type << 60) | obj)
1422 typedef struct ace {
1423 uid_t a_who; /* uid or gid */
1424 uint32_t a_access_mask; /* read,write,... */
1425 uint16_t a_flags; /* see below */
1426 uint16_t a_type; /* allow or deny */
1429 #define ACE_SLOT_CNT 6
1431 typedef struct zfs_znode_acl {
1432 uint64_t z_acl_extern_obj; /* ext acl pieces */
1433 uint32_t z_acl_count; /* Number of ACEs */
1434 uint16_t z_acl_version; /* acl version */
1435 uint16_t z_acl_pad; /* pad */
1436 ace_t z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */
1440 * This is the persistent portion of the znode. It is stored
1441 * in the "bonus buffer" of the file. Short symbolic links
1442 * are also stored in the bonus buffer.
1444 typedef struct znode_phys {
1445 uint64_t zp_atime[2]; /* 0 - last file access time */
1446 uint64_t zp_mtime[2]; /* 16 - last file modification time */
1447 uint64_t zp_ctime[2]; /* 32 - last file change time */
1448 uint64_t zp_crtime[2]; /* 48 - creation time */
1449 uint64_t zp_gen; /* 64 - generation (txg of creation) */
1450 uint64_t zp_mode; /* 72 - file mode bits */
1451 uint64_t zp_size; /* 80 - size of file */
1452 uint64_t zp_parent; /* 88 - directory parent (`..') */
1453 uint64_t zp_links; /* 96 - number of links to file */
1454 uint64_t zp_xattr; /* 104 - DMU object for xattrs */
1455 uint64_t zp_rdev; /* 112 - dev_t for VBLK & VCHR files */
1456 uint64_t zp_flags; /* 120 - persistent flags */
1457 uint64_t zp_uid; /* 128 - file owner */
1458 uint64_t zp_gid; /* 136 - owning group */
1459 uint64_t zp_pad[4]; /* 144 - future */
1460 zfs_znode_acl_t zp_acl; /* 176 - 263 ACL */
1462 * Data may pad out any remaining bytes in the znode buffer, eg:
1464 * |<---------------------- dnode_phys (512) ------------------------>|
1465 * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->|
1466 * |<---- znode (264) ---->|<---- data (56) ---->|
1468 * At present, we only use this space to store symbolic links.
1473 * In-core vdev representation.
1477 typedef int vdev_phys_read_t(struct vdev *vdev, void *priv,
1478 off_t offset, void *buf, size_t bytes);
1479 typedef int vdev_read_t(struct vdev *vdev, const blkptr_t *bp,
1480 void *buf, off_t offset, size_t bytes);
1482 typedef STAILQ_HEAD(vdev_list, vdev) vdev_list_t;
1484 typedef struct vdev {
1485 STAILQ_ENTRY(vdev) v_childlink; /* link in parent's child list */
1486 STAILQ_ENTRY(vdev) v_alllink; /* link in global vdev list */
1487 vdev_list_t v_children; /* children of this vdev */
1488 const char *v_name; /* vdev name */
1489 uint64_t v_guid; /* vdev guid */
1490 int v_id; /* index in parent */
1491 int v_ashift; /* offset to block shift */
1492 int v_nparity; /* # parity for raidz */
1493 struct vdev *v_top; /* parent vdev */
1494 int v_nchildren; /* # children */
1495 vdev_state_t v_state; /* current state */
1496 vdev_phys_read_t *v_phys_read; /* read from raw leaf vdev */
1497 vdev_read_t *v_read; /* read from vdev */
1498 void *v_read_priv; /* private data for read function */
1499 struct spa *spa; /* link to spa */
1503 * In-core pool representation.
1505 typedef STAILQ_HEAD(spa_list, spa) spa_list_t;
1507 typedef struct spa {
1508 STAILQ_ENTRY(spa) spa_link; /* link in global pool list */
1509 char *spa_name; /* pool name */
1510 uint64_t spa_guid; /* pool guid */
1511 uint64_t spa_txg; /* most recent transaction */
1512 struct uberblock spa_uberblock; /* best uberblock so far */
1513 vdev_list_t spa_vdevs; /* list of all toplevel vdevs */
1514 objset_phys_t spa_mos; /* MOS for this pool */
1515 zio_cksum_salt_t spa_cksum_salt; /* secret salt for cksum */
1516 void *spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS];
1517 int spa_inited; /* initialized */
1520 static void decode_embedded_bp_compressed(const blkptr_t *, void *);