1/*-
2 * Copyright (c) 2002 McAfee, Inc.
3 * All rights reserved.
4 *
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
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
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.
18 *
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
29 * SUCH DAMAGE.
30 */
31/*
32 * CDDL HEADER START
33 *
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.
37 *
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.
42 *
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]
48 *
49 * CDDL HEADER END
50 */
51/*
52 * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
53 * Use is subject to license terms.
54 */
55/*
56 * Copyright 2013 by Saso Kiselkov. All rights reserved.
57 */
58/*
59 * Copyright (c) 2020 by Delphix. All rights reserved.
60 */
61
62#include <sys/queue.h>
63#include <sys/list.h>
64#include <bootstrap.h>
65
66#ifndef _ZFSIMPL_H_
67#define	_ZFSIMPL_H_
68
69#define	MAXNAMELEN	256
70
71#define _NOTE(s)
72
73/*
74 * AVL comparator helpers
75 */
76#define	AVL_ISIGN(a)	(((a) > 0) - ((a) < 0))
77#define	AVL_CMP(a, b)	(((a) > (b)) - ((a) < (b)))
78#define	AVL_PCMP(a, b)	\
79	(((uintptr_t)(a) > (uintptr_t)(b)) - ((uintptr_t)(a) < (uintptr_t)(b)))
80
81typedef enum { B_FALSE, B_TRUE } boolean_t;
82
83/* CRC64 table */
84#define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
85
86/*
87 * Macros for various sorts of alignment and rounding when the alignment
88 * is known to be a power of 2.
89 */
90#define	P2ALIGN(x, align)		((x) & -(align))
91#define	P2PHASE(x, align)		((x) & ((align) - 1))
92#define	P2NPHASE(x, align)		(-(x) & ((align) - 1))
93#define	P2ROUNDUP(x, align)		(-(-(x) & -(align)))
94#define	P2END(x, align)			(-(~(x) & -(align)))
95#define	P2PHASEUP(x, align, phase)	((phase) - (((phase) - (x)) & -(align)))
96#define	P2BOUNDARY(off, len, align)	(((off) ^ ((off) + (len) - 1)) > (align) - 1)
97
98/*
99 * General-purpose 32-bit and 64-bit bitfield encodings.
100 */
101#define	BF32_DECODE(x, low, len)	P2PHASE((x) >> (low), 1U << (len))
102#define	BF64_DECODE(x, low, len)	P2PHASE((x) >> (low), 1ULL << (len))
103#define	BF32_ENCODE(x, low, len)	(P2PHASE((x), 1U << (len)) << (low))
104#define	BF64_ENCODE(x, low, len)	(P2PHASE((x), 1ULL << (len)) << (low))
105
106#define	BF32_GET(x, low, len)		BF32_DECODE(x, low, len)
107#define	BF64_GET(x, low, len)		BF64_DECODE(x, low, len)
108
109#define	BF32_SET(x, low, len, val)	\
110	((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len))
111#define	BF64_SET(x, low, len, val)	\
112	((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len))
113
114#define	BF32_GET_SB(x, low, len, shift, bias)	\
115	((BF32_GET(x, low, len) + (bias)) << (shift))
116#define	BF64_GET_SB(x, low, len, shift, bias)	\
117	((BF64_GET(x, low, len) + (bias)) << (shift))
118
119#define	BF32_SET_SB(x, low, len, shift, bias, val)	\
120	BF32_SET(x, low, len, ((val) >> (shift)) - (bias))
121#define	BF64_SET_SB(x, low, len, shift, bias, val)	\
122	BF64_SET(x, low, len, ((val) >> (shift)) - (bias))
123
124/*
125 * Macros to reverse byte order
126 */
127#define	BSWAP_8(x)	((x) & 0xff)
128#define	BSWAP_16(x)	((BSWAP_8(x) << 8) | BSWAP_8((x) >> 8))
129#define	BSWAP_32(x)	((BSWAP_16(x) << 16) | BSWAP_16((x) >> 16))
130#define	BSWAP_64(x)	((BSWAP_32(x) << 32) | BSWAP_32((x) >> 32))
131
132#define	SPA_MINBLOCKSHIFT	9
133#define	SPA_OLDMAXBLOCKSHIFT	17
134#define	SPA_MAXBLOCKSHIFT	24
135#define	SPA_MINBLOCKSIZE	(1ULL << SPA_MINBLOCKSHIFT)
136#define	SPA_OLDMAXBLOCKSIZE	(1ULL << SPA_OLDMAXBLOCKSHIFT)
137#define	SPA_MAXBLOCKSIZE	(1ULL << SPA_MAXBLOCKSHIFT)
138
139/*
140 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
141 * The ASIZE encoding should be at least 64 times larger (6 more bits)
142 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
143 * overhead, three DVAs per bp, plus one more bit in case we do anything
144 * else that expands the ASIZE.
145 */
146#define	SPA_LSIZEBITS		16	/* LSIZE up to 32M (2^16 * 512)	*/
147#define	SPA_PSIZEBITS		16	/* PSIZE up to 32M (2^16 * 512)	*/
148#define	SPA_ASIZEBITS		24	/* ASIZE up to 64 times larger	*/
149
150/*
151 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
152 * The members of the dva_t should be considered opaque outside the SPA.
153 */
154typedef struct dva {
155	uint64_t	dva_word[2];
156} dva_t;
157
158/*
159 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
160 */
161typedef struct zio_cksum {
162	uint64_t	zc_word[4];
163} zio_cksum_t;
164
165/*
166 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept
167 * secret and is suitable for use in MAC algorithms as the key.
168 */
169typedef struct zio_cksum_salt {
170	uint8_t		zcs_bytes[32];
171} zio_cksum_salt_t;
172
173/*
174 * Each block is described by its DVAs, time of birth, checksum, etc.
175 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
176 *
177 *	64	56	48	40	32	24	16	8	0
178 *	+-------+-------+-------+-------+-------+-------+-------+-------+
179 * 0	|		vdev1		| GRID  |	  ASIZE		|
180 *	+-------+-------+-------+-------+-------+-------+-------+-------+
181 * 1	|G|			 offset1				|
182 *	+-------+-------+-------+-------+-------+-------+-------+-------+
183 * 2	|		vdev2		| GRID  |	  ASIZE		|
184 *	+-------+-------+-------+-------+-------+-------+-------+-------+
185 * 3	|G|			 offset2				|
186 *	+-------+-------+-------+-------+-------+-------+-------+-------+
187 * 4	|		vdev3		| GRID  |	  ASIZE		|
188 *	+-------+-------+-------+-------+-------+-------+-------+-------+
189 * 5	|G|			 offset3				|
190 *	+-------+-------+-------+-------+-------+-------+-------+-------+
191 * 6	|BDX|lvl| type	| cksum |E| comp|    PSIZE	|     LSIZE	|
192 *	+-------+-------+-------+-------+-------+-------+-------+-------+
193 * 7	|			padding					|
194 *	+-------+-------+-------+-------+-------+-------+-------+-------+
195 * 8	|			padding					|
196 *	+-------+-------+-------+-------+-------+-------+-------+-------+
197 * 9	|			physical birth txg			|
198 *	+-------+-------+-------+-------+-------+-------+-------+-------+
199 * a	|			logical birth txg			|
200 *	+-------+-------+-------+-------+-------+-------+-------+-------+
201 * b	|			fill count				|
202 *	+-------+-------+-------+-------+-------+-------+-------+-------+
203 * c	|			checksum[0]				|
204 *	+-------+-------+-------+-------+-------+-------+-------+-------+
205 * d	|			checksum[1]				|
206 *	+-------+-------+-------+-------+-------+-------+-------+-------+
207 * e	|			checksum[2]				|
208 *	+-------+-------+-------+-------+-------+-------+-------+-------+
209 * f	|			checksum[3]				|
210 *	+-------+-------+-------+-------+-------+-------+-------+-------+
211 *
212 * Legend:
213 *
214 * vdev		virtual device ID
215 * offset	offset into virtual device
216 * LSIZE	logical size
217 * PSIZE	physical size (after compression)
218 * ASIZE	allocated size (including RAID-Z parity and gang block headers)
219 * GRID		RAID-Z layout information (reserved for future use)
220 * cksum	checksum function
221 * comp		compression function
222 * G		gang block indicator
223 * B		byteorder (endianness)
224 * D		dedup
225 * X		encryption (on version 30, which is not supported)
226 * E		blkptr_t contains embedded data (see below)
227 * lvl		level of indirection
228 * type		DMU object type
229 * phys birth	txg of block allocation; zero if same as logical birth txg
230 * log. birth	transaction group in which the block was logically born
231 * fill count	number of non-zero blocks under this bp
232 * checksum[4]	256-bit checksum of the data this bp describes
233 */
234
235/*
236 * "Embedded" blkptr_t's don't actually point to a block, instead they
237 * have a data payload embedded in the blkptr_t itself.  See the comment
238 * in blkptr.c for more details.
239 *
240 * The blkptr_t is laid out as follows:
241 *
242 *	64	56	48	40	32	24	16	8	0
243 *	+-------+-------+-------+-------+-------+-------+-------+-------+
244 * 0	|      payload                                                  |
245 * 1	|      payload                                                  |
246 * 2	|      payload                                                  |
247 * 3	|      payload                                                  |
248 * 4	|      payload                                                  |
249 * 5	|      payload                                                  |
250 *	+-------+-------+-------+-------+-------+-------+-------+-------+
251 * 6	|BDX|lvl| type	| etype |E| comp| PSIZE|              LSIZE	|
252 *	+-------+-------+-------+-------+-------+-------+-------+-------+
253 * 7	|      payload                                                  |
254 * 8	|      payload                                                  |
255 * 9	|      payload                                                  |
256 *	+-------+-------+-------+-------+-------+-------+-------+-------+
257 * a	|			logical birth txg			|
258 *	+-------+-------+-------+-------+-------+-------+-------+-------+
259 * b	|      payload                                                  |
260 * c	|      payload                                                  |
261 * d	|      payload                                                  |
262 * e	|      payload                                                  |
263 * f	|      payload                                                  |
264 *	+-------+-------+-------+-------+-------+-------+-------+-------+
265 *
266 * Legend:
267 *
268 * payload		contains the embedded data
269 * B (byteorder)	byteorder (endianness)
270 * D (dedup)		padding (set to zero)
271 * X			encryption (set to zero; see above)
272 * E (embedded)		set to one
273 * lvl			indirection level
274 * type			DMU object type
275 * etype		how to interpret embedded data (BP_EMBEDDED_TYPE_*)
276 * comp			compression function of payload
277 * PSIZE		size of payload after compression, in bytes
278 * LSIZE		logical size of payload, in bytes
279 *			note that 25 bits is enough to store the largest
280 *			"normal" BP's LSIZE (2^16 * 2^9) in bytes
281 * log. birth		transaction group in which the block was logically born
282 *
283 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
284 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
285 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
286 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
287 * BP's so the BP_SET_* macros can be used with them.  etype, PSIZE, LSIZE must
288 * be set with the BPE_SET_* macros.  BP_SET_EMBEDDED() should be called before
289 * other macros, as they assert that they are only used on BP's of the correct
290 * "embedded-ness".
291 */
292
293#define	BPE_GET_ETYPE(bp)	\
294	(ASSERT(BP_IS_EMBEDDED(bp)), \
295	BF64_GET((bp)->blk_prop, 40, 8))
296#define	BPE_SET_ETYPE(bp, t)	do { \
297	ASSERT(BP_IS_EMBEDDED(bp)); \
298	BF64_SET((bp)->blk_prop, 40, 8, t); \
299_NOTE(CONSTCOND) } while (0)
300
301#define	BPE_GET_LSIZE(bp)	\
302	(ASSERT(BP_IS_EMBEDDED(bp)), \
303	BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
304#define	BPE_SET_LSIZE(bp, x)	do { \
305	ASSERT(BP_IS_EMBEDDED(bp)); \
306	BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
307_NOTE(CONSTCOND) } while (0)
308
309#define	BPE_GET_PSIZE(bp)	\
310	(ASSERT(BP_IS_EMBEDDED(bp)), \
311	BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
312#define	BPE_SET_PSIZE(bp, x)	do { \
313	ASSERT(BP_IS_EMBEDDED(bp)); \
314	BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
315_NOTE(CONSTCOND) } while (0)
316
317typedef enum bp_embedded_type {
318	BP_EMBEDDED_TYPE_DATA,
319	BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */
320	NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED
321} bp_embedded_type_t;
322
323#define	BPE_NUM_WORDS 14
324#define	BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
325#define	BPE_IS_PAYLOADWORD(bp, wp) \
326	((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
327
328#define	SPA_BLKPTRSHIFT	7		/* blkptr_t is 128 bytes	*/
329#define	SPA_DVAS_PER_BP	3		/* Number of DVAs in a bp	*/
330
331typedef struct blkptr {
332	dva_t		blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
333	uint64_t	blk_prop;	/* size, compression, type, etc	    */
334	uint64_t	blk_pad[2];	/* Extra space for the future	    */
335	uint64_t	blk_phys_birth;	/* txg when block was allocated	    */
336	uint64_t	blk_birth;	/* transaction group at birth	    */
337	uint64_t	blk_fill;	/* fill count			    */
338	zio_cksum_t	blk_cksum;	/* 256-bit checksum		    */
339} blkptr_t;
340
341/*
342 * Macros to get and set fields in a bp or DVA.
343 */
344#define	DVA_GET_ASIZE(dva)	\
345	BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
346#define	DVA_SET_ASIZE(dva, x)	\
347	BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
348	SPA_MINBLOCKSHIFT, 0, x)
349
350#define	DVA_GET_GRID(dva)	BF64_GET((dva)->dva_word[0], 24, 8)
351#define	DVA_SET_GRID(dva, x)	BF64_SET((dva)->dva_word[0], 24, 8, x)
352
353#define	DVA_GET_VDEV(dva)	BF64_GET((dva)->dva_word[0], 32, 32)
354#define	DVA_SET_VDEV(dva, x)	BF64_SET((dva)->dva_word[0], 32, 32, x)
355
356#define	DVA_GET_OFFSET(dva)	\
357	BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
358#define	DVA_SET_OFFSET(dva, x)	\
359	BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
360
361#define	DVA_GET_GANG(dva)	BF64_GET((dva)->dva_word[1], 63, 1)
362#define	DVA_SET_GANG(dva, x)	BF64_SET((dva)->dva_word[1], 63, 1, x)
363
364#define	BP_GET_LSIZE(bp)	\
365	(BP_IS_EMBEDDED(bp) ?	\
366	(BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
367	BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
368#define	BP_SET_LSIZE(bp, x)	do { \
369	ASSERT(!BP_IS_EMBEDDED(bp)); \
370	BF64_SET_SB((bp)->blk_prop, \
371	    0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
372_NOTE(CONSTCOND) } while (0)
373
374#define	BP_GET_PSIZE(bp)	\
375	BF64_GET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)
376#define	BP_SET_PSIZE(bp, x)	\
377	BF64_SET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x)
378
379#define	BP_GET_COMPRESS(bp)	BF64_GET((bp)->blk_prop, 32, 7)
380#define	BP_SET_COMPRESS(bp, x)	BF64_SET((bp)->blk_prop, 32, 7, x)
381
382#define	BP_GET_CHECKSUM(bp)	BF64_GET((bp)->blk_prop, 40, 8)
383#define	BP_SET_CHECKSUM(bp, x)	BF64_SET((bp)->blk_prop, 40, 8, x)
384
385#define	BP_GET_TYPE(bp)		BF64_GET((bp)->blk_prop, 48, 8)
386#define	BP_SET_TYPE(bp, x)	BF64_SET((bp)->blk_prop, 48, 8, x)
387
388#define	BP_GET_LEVEL(bp)	BF64_GET((bp)->blk_prop, 56, 5)
389#define	BP_SET_LEVEL(bp, x)	BF64_SET((bp)->blk_prop, 56, 5, x)
390
391#define	BP_IS_EMBEDDED(bp)	BF64_GET((bp)->blk_prop, 39, 1)
392
393#define	BP_GET_DEDUP(bp)	BF64_GET((bp)->blk_prop, 62, 1)
394#define	BP_SET_DEDUP(bp, x)	BF64_SET((bp)->blk_prop, 62, 1, x)
395
396#define	BP_GET_BYTEORDER(bp)	BF64_GET((bp)->blk_prop, 63, 1)
397#define	BP_SET_BYTEORDER(bp, x)	BF64_SET((bp)->blk_prop, 63, 1, x)
398
399#define	BP_PHYSICAL_BIRTH(bp)		\
400	((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
401
402#define	BP_GET_ASIZE(bp)	\
403	(DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
404		DVA_GET_ASIZE(&(bp)->blk_dva[2]))
405
406#define	BP_GET_UCSIZE(bp) \
407	((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \
408	BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp));
409
410#define	BP_GET_NDVAS(bp)	\
411	(!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
412	!!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
413	!!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
414
415#define	DVA_EQUAL(dva1, dva2)	\
416	((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
417	(dva1)->dva_word[0] == (dva2)->dva_word[0])
418
419#define	ZIO_CHECKSUM_EQUAL(zc1, zc2) \
420	(0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
421	((zc1).zc_word[1] - (zc2).zc_word[1]) | \
422	((zc1).zc_word[2] - (zc2).zc_word[2]) | \
423	((zc1).zc_word[3] - (zc2).zc_word[3])))
424
425
426#define	DVA_IS_VALID(dva)	(DVA_GET_ASIZE(dva) != 0)
427
428#define	ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3)	\
429{						\
430	(zcp)->zc_word[0] = w0;			\
431	(zcp)->zc_word[1] = w1;			\
432	(zcp)->zc_word[2] = w2;			\
433	(zcp)->zc_word[3] = w3;			\
434}
435
436#define	BP_IDENTITY(bp)		(&(bp)->blk_dva[0])
437#define	BP_IS_GANG(bp)		DVA_GET_GANG(BP_IDENTITY(bp))
438#define	DVA_IS_EMPTY(dva)	((dva)->dva_word[0] == 0ULL &&  \
439	(dva)->dva_word[1] == 0ULL)
440#define	BP_IS_HOLE(bp)		DVA_IS_EMPTY(BP_IDENTITY(bp))
441#define	BP_IS_OLDER(bp, txg)	(!BP_IS_HOLE(bp) && (bp)->blk_birth < (txg))
442
443#define	BP_ZERO(bp)				\
444{						\
445	(bp)->blk_dva[0].dva_word[0] = 0;	\
446	(bp)->blk_dva[0].dva_word[1] = 0;	\
447	(bp)->blk_dva[1].dva_word[0] = 0;	\
448	(bp)->blk_dva[1].dva_word[1] = 0;	\
449	(bp)->blk_dva[2].dva_word[0] = 0;	\
450	(bp)->blk_dva[2].dva_word[1] = 0;	\
451	(bp)->blk_prop = 0;			\
452	(bp)->blk_pad[0] = 0;			\
453	(bp)->blk_pad[1] = 0;			\
454	(bp)->blk_phys_birth = 0;		\
455	(bp)->blk_birth = 0;			\
456	(bp)->blk_fill = 0;			\
457	ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0);	\
458}
459
460#if BYTE_ORDER == _BIG_ENDIAN
461#define	ZFS_HOST_BYTEORDER	(0ULL)
462#else
463#define	ZFS_HOST_BYTEORDER	(1ULL)
464#endif
465
466#define	BP_SHOULD_BYTESWAP(bp)	(BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
467#define	BPE_NUM_WORDS 14
468#define	BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
469#define	BPE_IS_PAYLOADWORD(bp, wp) \
470	((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
471
472/*
473 * Embedded checksum
474 */
475#define	ZEC_MAGIC	0x210da7ab10c7a11ULL
476
477typedef struct zio_eck {
478	uint64_t	zec_magic;	/* for validation, endianness	*/
479	zio_cksum_t	zec_cksum;	/* 256-bit checksum		*/
480} zio_eck_t;
481
482/*
483 * Gang block headers are self-checksumming and contain an array
484 * of block pointers.
485 */
486#define	SPA_GANGBLOCKSIZE	SPA_MINBLOCKSIZE
487#define	SPA_GBH_NBLKPTRS	((SPA_GANGBLOCKSIZE - \
488	sizeof (zio_eck_t)) / sizeof (blkptr_t))
489#define	SPA_GBH_FILLER		((SPA_GANGBLOCKSIZE - \
490	sizeof (zio_eck_t) - \
491	(SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\
492	sizeof (uint64_t))
493
494typedef struct zio_gbh {
495	blkptr_t		zg_blkptr[SPA_GBH_NBLKPTRS];
496	uint64_t		zg_filler[SPA_GBH_FILLER];
497	zio_eck_t		zg_tail;
498} zio_gbh_phys_t;
499
500#define	VDEV_RAIDZ_MAXPARITY	3
501
502#define	VDEV_PAD_SIZE		(8 << 10)
503/* 2 padding areas (vl_pad1 and vl_be) to skip */
504#define	VDEV_SKIP_SIZE		VDEV_PAD_SIZE * 2
505#define	VDEV_PHYS_SIZE		(112 << 10)
506#define	VDEV_UBERBLOCK_RING	(128 << 10)
507
508/*
509 * MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock
510 * ring when MMP is enabled.
511 */
512#define	MMP_BLOCKS_PER_LABEL	1
513
514/* The largest uberblock we support is 8k. */
515#define	MAX_UBERBLOCK_SHIFT	(13)
516#define	VDEV_UBERBLOCK_SHIFT(vd)	\
517	MIN(MAX((vd)->v_top->v_ashift, UBERBLOCK_SHIFT), MAX_UBERBLOCK_SHIFT)
518#define	VDEV_UBERBLOCK_COUNT(vd)	\
519	(VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
520#define	VDEV_UBERBLOCK_OFFSET(vd, n)	\
521	offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
522#define	VDEV_UBERBLOCK_SIZE(vd)		(1ULL << VDEV_UBERBLOCK_SHIFT(vd))
523
524typedef struct vdev_phys {
525	char		vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)];
526	zio_eck_t	vp_zbt;
527} vdev_phys_t;
528
529typedef enum vbe_vers {
530	/* The bootenv file is stored as ascii text in the envblock */
531	VB_RAW = 0,
532
533	/*
534	 * The bootenv file is converted to an nvlist and then packed into the
535	 * envblock.
536	 */
537	VB_NVLIST = 1
538} vbe_vers_t;
539
540typedef struct vdev_boot_envblock {
541	uint64_t	vbe_version;
542	char		vbe_bootenv[VDEV_PAD_SIZE - sizeof (uint64_t) -
543			sizeof (zio_eck_t)];
544 	zio_eck_t	vbe_zbt;
545} vdev_boot_envblock_t;
546
547CTASSERT(sizeof (vdev_boot_envblock_t) == VDEV_PAD_SIZE);
548
549typedef struct vdev_label {
550	char		vl_pad1[VDEV_PAD_SIZE];			/*  8K  */
551	vdev_boot_envblock_t	vl_be;				/*  8K  */
552	vdev_phys_t	vl_vdev_phys;				/* 112K	*/
553	char		vl_uberblock[VDEV_UBERBLOCK_RING];	/* 128K	*/
554} vdev_label_t;							/* 256K total */
555
556/*
557 * vdev_dirty() flags
558 */
559#define	VDD_METASLAB	0x01
560#define	VDD_DTL		0x02
561
562/*
563 * Size and offset of embedded boot loader region on each label.
564 * The total size of the first two labels plus the boot area is 4MB.
565 */
566#define	VDEV_BOOT_OFFSET	(2 * sizeof (vdev_label_t))
567#define	VDEV_BOOT_SIZE		(7ULL << 19)			/* 3.5M	*/
568
569/*
570 * Size of label regions at the start and end of each leaf device.
571 */
572#define	VDEV_LABEL_START_SIZE	(2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
573#define	VDEV_LABEL_END_SIZE	(2 * sizeof (vdev_label_t))
574#define	VDEV_LABELS		4
575
576enum zio_checksum {
577	ZIO_CHECKSUM_INHERIT = 0,
578	ZIO_CHECKSUM_ON,
579	ZIO_CHECKSUM_OFF,
580	ZIO_CHECKSUM_LABEL,
581	ZIO_CHECKSUM_GANG_HEADER,
582	ZIO_CHECKSUM_ZILOG,
583	ZIO_CHECKSUM_FLETCHER_2,
584	ZIO_CHECKSUM_FLETCHER_4,
585	ZIO_CHECKSUM_SHA256,
586	ZIO_CHECKSUM_ZILOG2,
587	ZIO_CHECKSUM_NOPARITY,
588	ZIO_CHECKSUM_SHA512,
589	ZIO_CHECKSUM_SKEIN,
590	ZIO_CHECKSUM_EDONR,
591	ZIO_CHECKSUM_FUNCTIONS
592};
593
594#define	ZIO_CHECKSUM_ON_VALUE	ZIO_CHECKSUM_FLETCHER_4
595#define	ZIO_CHECKSUM_DEFAULT	ZIO_CHECKSUM_ON
596
597enum zio_compress {
598	ZIO_COMPRESS_INHERIT = 0,
599	ZIO_COMPRESS_ON,
600	ZIO_COMPRESS_OFF,
601	ZIO_COMPRESS_LZJB,
602	ZIO_COMPRESS_EMPTY,
603	ZIO_COMPRESS_GZIP_1,
604	ZIO_COMPRESS_GZIP_2,
605	ZIO_COMPRESS_GZIP_3,
606	ZIO_COMPRESS_GZIP_4,
607	ZIO_COMPRESS_GZIP_5,
608	ZIO_COMPRESS_GZIP_6,
609	ZIO_COMPRESS_GZIP_7,
610	ZIO_COMPRESS_GZIP_8,
611	ZIO_COMPRESS_GZIP_9,
612	ZIO_COMPRESS_ZLE,
613	ZIO_COMPRESS_LZ4,
614	ZIO_COMPRESS_FUNCTIONS
615};
616
617#define	ZIO_COMPRESS_ON_VALUE	ZIO_COMPRESS_LZJB
618#define	ZIO_COMPRESS_DEFAULT	ZIO_COMPRESS_OFF
619
620/* nvlist pack encoding */
621#define	NV_ENCODE_NATIVE	0
622#define	NV_ENCODE_XDR		1
623
624typedef enum {
625	DATA_TYPE_UNKNOWN = 0,
626	DATA_TYPE_BOOLEAN,
627	DATA_TYPE_BYTE,
628	DATA_TYPE_INT16,
629	DATA_TYPE_UINT16,
630	DATA_TYPE_INT32,
631	DATA_TYPE_UINT32,
632	DATA_TYPE_INT64,
633	DATA_TYPE_UINT64,
634	DATA_TYPE_STRING,
635	DATA_TYPE_BYTE_ARRAY,
636	DATA_TYPE_INT16_ARRAY,
637	DATA_TYPE_UINT16_ARRAY,
638	DATA_TYPE_INT32_ARRAY,
639	DATA_TYPE_UINT32_ARRAY,
640	DATA_TYPE_INT64_ARRAY,
641	DATA_TYPE_UINT64_ARRAY,
642	DATA_TYPE_STRING_ARRAY,
643	DATA_TYPE_HRTIME,
644	DATA_TYPE_NVLIST,
645	DATA_TYPE_NVLIST_ARRAY,
646	DATA_TYPE_BOOLEAN_VALUE,
647	DATA_TYPE_INT8,
648	DATA_TYPE_UINT8,
649	DATA_TYPE_BOOLEAN_ARRAY,
650	DATA_TYPE_INT8_ARRAY,
651	DATA_TYPE_UINT8_ARRAY
652} data_type_t;
653
654/*
655 * On-disk version number.
656 */
657#define	SPA_VERSION_1			1ULL
658#define	SPA_VERSION_2			2ULL
659#define	SPA_VERSION_3			3ULL
660#define	SPA_VERSION_4			4ULL
661#define	SPA_VERSION_5			5ULL
662#define	SPA_VERSION_6			6ULL
663#define	SPA_VERSION_7			7ULL
664#define	SPA_VERSION_8			8ULL
665#define	SPA_VERSION_9			9ULL
666#define	SPA_VERSION_10			10ULL
667#define	SPA_VERSION_11			11ULL
668#define	SPA_VERSION_12			12ULL
669#define	SPA_VERSION_13			13ULL
670#define	SPA_VERSION_14			14ULL
671#define	SPA_VERSION_15			15ULL
672#define	SPA_VERSION_16			16ULL
673#define	SPA_VERSION_17			17ULL
674#define	SPA_VERSION_18			18ULL
675#define	SPA_VERSION_19			19ULL
676#define	SPA_VERSION_20			20ULL
677#define	SPA_VERSION_21			21ULL
678#define	SPA_VERSION_22			22ULL
679#define	SPA_VERSION_23			23ULL
680#define	SPA_VERSION_24			24ULL
681#define	SPA_VERSION_25			25ULL
682#define	SPA_VERSION_26			26ULL
683#define	SPA_VERSION_27			27ULL
684#define	SPA_VERSION_28			28ULL
685#define	SPA_VERSION_5000		5000ULL
686
687/*
688 * When bumping up SPA_VERSION, make sure GRUB ZFS understands the on-disk
689 * format change. Go to usr/src/grub/grub-0.97/stage2/{zfs-include/, fsys_zfs*},
690 * and do the appropriate changes.  Also bump the version number in
691 * usr/src/grub/capability.
692 */
693#define	SPA_VERSION			SPA_VERSION_5000
694#define	SPA_VERSION_STRING		"5000"
695
696/*
697 * Symbolic names for the changes that caused a SPA_VERSION switch.
698 * Used in the code when checking for presence or absence of a feature.
699 * Feel free to define multiple symbolic names for each version if there
700 * were multiple changes to on-disk structures during that version.
701 *
702 * NOTE: When checking the current SPA_VERSION in your code, be sure
703 *       to use spa_version() since it reports the version of the
704 *       last synced uberblock.  Checking the in-flight version can
705 *       be dangerous in some cases.
706 */
707#define	SPA_VERSION_INITIAL		SPA_VERSION_1
708#define	SPA_VERSION_DITTO_BLOCKS	SPA_VERSION_2
709#define	SPA_VERSION_SPARES		SPA_VERSION_3
710#define	SPA_VERSION_RAID6		SPA_VERSION_3
711#define	SPA_VERSION_BPLIST_ACCOUNT	SPA_VERSION_3
712#define	SPA_VERSION_RAIDZ_DEFLATE	SPA_VERSION_3
713#define	SPA_VERSION_DNODE_BYTES		SPA_VERSION_3
714#define	SPA_VERSION_ZPOOL_HISTORY	SPA_VERSION_4
715#define	SPA_VERSION_GZIP_COMPRESSION	SPA_VERSION_5
716#define	SPA_VERSION_BOOTFS		SPA_VERSION_6
717#define	SPA_VERSION_SLOGS		SPA_VERSION_7
718#define	SPA_VERSION_DELEGATED_PERMS	SPA_VERSION_8
719#define	SPA_VERSION_FUID		SPA_VERSION_9
720#define	SPA_VERSION_REFRESERVATION	SPA_VERSION_9
721#define	SPA_VERSION_REFQUOTA		SPA_VERSION_9
722#define	SPA_VERSION_UNIQUE_ACCURATE	SPA_VERSION_9
723#define	SPA_VERSION_L2CACHE		SPA_VERSION_10
724#define	SPA_VERSION_NEXT_CLONES		SPA_VERSION_11
725#define	SPA_VERSION_ORIGIN		SPA_VERSION_11
726#define	SPA_VERSION_DSL_SCRUB		SPA_VERSION_11
727#define	SPA_VERSION_SNAP_PROPS		SPA_VERSION_12
728#define	SPA_VERSION_USED_BREAKDOWN	SPA_VERSION_13
729#define	SPA_VERSION_PASSTHROUGH_X	SPA_VERSION_14
730#define SPA_VERSION_USERSPACE		SPA_VERSION_15
731#define	SPA_VERSION_STMF_PROP		SPA_VERSION_16
732#define	SPA_VERSION_RAIDZ3		SPA_VERSION_17
733#define	SPA_VERSION_USERREFS		SPA_VERSION_18
734#define	SPA_VERSION_HOLES		SPA_VERSION_19
735#define	SPA_VERSION_ZLE_COMPRESSION	SPA_VERSION_20
736#define	SPA_VERSION_DEDUP		SPA_VERSION_21
737#define	SPA_VERSION_RECVD_PROPS		SPA_VERSION_22
738#define	SPA_VERSION_SLIM_ZIL		SPA_VERSION_23
739#define	SPA_VERSION_SA			SPA_VERSION_24
740#define	SPA_VERSION_SCAN		SPA_VERSION_25
741#define	SPA_VERSION_DIR_CLONES		SPA_VERSION_26
742#define	SPA_VERSION_DEADLISTS		SPA_VERSION_26
743#define	SPA_VERSION_FAST_SNAP		SPA_VERSION_27
744#define	SPA_VERSION_MULTI_REPLACE	SPA_VERSION_28
745#define	SPA_VERSION_BEFORE_FEATURES	SPA_VERSION_28
746#define	SPA_VERSION_FEATURES		SPA_VERSION_5000
747
748#define	SPA_VERSION_IS_SUPPORTED(v) \
749	(((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \
750	((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION))
751
752/*
753 * The following are configuration names used in the nvlist describing a pool's
754 * configuration.
755 */
756#define	ZPOOL_CONFIG_VERSION		"version"
757#define	ZPOOL_CONFIG_POOL_NAME		"name"
758#define	ZPOOL_CONFIG_POOL_STATE		"state"
759#define	ZPOOL_CONFIG_POOL_TXG		"txg"
760#define	ZPOOL_CONFIG_POOL_GUID		"pool_guid"
761#define	ZPOOL_CONFIG_CREATE_TXG		"create_txg"
762#define	ZPOOL_CONFIG_TOP_GUID		"top_guid"
763#define	ZPOOL_CONFIG_VDEV_TREE		"vdev_tree"
764#define	ZPOOL_CONFIG_TYPE		"type"
765#define	ZPOOL_CONFIG_CHILDREN		"children"
766#define	ZPOOL_CONFIG_ID			"id"
767#define	ZPOOL_CONFIG_GUID		"guid"
768#define	ZPOOL_CONFIG_INDIRECT_OBJECT	"com.delphix:indirect_object"
769#define	ZPOOL_CONFIG_INDIRECT_BIRTHS	"com.delphix:indirect_births"
770#define	ZPOOL_CONFIG_PREV_INDIRECT_VDEV	"com.delphix:prev_indirect_vdev"
771#define	ZPOOL_CONFIG_PATH		"path"
772#define	ZPOOL_CONFIG_DEVID		"devid"
773#define	ZPOOL_CONFIG_METASLAB_ARRAY	"metaslab_array"
774#define	ZPOOL_CONFIG_METASLAB_SHIFT	"metaslab_shift"
775#define	ZPOOL_CONFIG_ASHIFT		"ashift"
776#define	ZPOOL_CONFIG_ASIZE		"asize"
777#define	ZPOOL_CONFIG_DTL		"DTL"
778#define	ZPOOL_CONFIG_STATS		"stats"
779#define	ZPOOL_CONFIG_WHOLE_DISK		"whole_disk"
780#define	ZPOOL_CONFIG_ERRCOUNT		"error_count"
781#define	ZPOOL_CONFIG_NOT_PRESENT	"not_present"
782#define	ZPOOL_CONFIG_SPARES		"spares"
783#define	ZPOOL_CONFIG_IS_SPARE		"is_spare"
784#define	ZPOOL_CONFIG_NPARITY		"nparity"
785#define	ZPOOL_CONFIG_HOSTID		"hostid"
786#define	ZPOOL_CONFIG_HOSTNAME		"hostname"
787#define	ZPOOL_CONFIG_IS_LOG		"is_log"
788#define	ZPOOL_CONFIG_TIMESTAMP		"timestamp" /* not stored on disk */
789#define	ZPOOL_CONFIG_FEATURES_FOR_READ	"features_for_read"
790#define	ZPOOL_CONFIG_VDEV_CHILDREN	"vdev_children"
791
792/*
793 * The persistent vdev state is stored as separate values rather than a single
794 * 'vdev_state' entry.  This is because a device can be in multiple states, such
795 * as offline and degraded.
796 */
797#define	ZPOOL_CONFIG_OFFLINE            "offline"
798#define	ZPOOL_CONFIG_FAULTED            "faulted"
799#define	ZPOOL_CONFIG_DEGRADED           "degraded"
800#define	ZPOOL_CONFIG_REMOVED            "removed"
801#define	ZPOOL_CONFIG_FRU		"fru"
802#define	ZPOOL_CONFIG_AUX_STATE		"aux_state"
803
804#define	VDEV_TYPE_ROOT			"root"
805#define	VDEV_TYPE_MIRROR		"mirror"
806#define	VDEV_TYPE_REPLACING		"replacing"
807#define	VDEV_TYPE_RAIDZ			"raidz"
808#define	VDEV_TYPE_DISK			"disk"
809#define	VDEV_TYPE_FILE			"file"
810#define	VDEV_TYPE_MISSING		"missing"
811#define	VDEV_TYPE_HOLE			"hole"
812#define	VDEV_TYPE_SPARE			"spare"
813#define	VDEV_TYPE_LOG			"log"
814#define	VDEV_TYPE_L2CACHE		"l2cache"
815#define	VDEV_TYPE_INDIRECT		"indirect"
816
817/*
818 * This is needed in userland to report the minimum necessary device size.
819 */
820#define	SPA_MINDEVSIZE		(64ULL << 20)
821
822/*
823 * The location of the pool configuration repository, shared between kernel and
824 * userland.
825 */
826#define	ZPOOL_CACHE		"/boot/zfs/zpool.cache"
827
828/*
829 * vdev states are ordered from least to most healthy.
830 * A vdev that's CANT_OPEN or below is considered unusable.
831 */
832typedef enum vdev_state {
833	VDEV_STATE_UNKNOWN = 0,	/* Uninitialized vdev			*/
834	VDEV_STATE_CLOSED,	/* Not currently open			*/
835	VDEV_STATE_OFFLINE,	/* Not allowed to open			*/
836	VDEV_STATE_REMOVED,	/* Explicitly removed from system	*/
837	VDEV_STATE_CANT_OPEN,	/* Tried to open, but failed		*/
838	VDEV_STATE_FAULTED,	/* External request to fault device	*/
839	VDEV_STATE_DEGRADED,	/* Replicated vdev with unhealthy kids	*/
840	VDEV_STATE_HEALTHY	/* Presumed good			*/
841} vdev_state_t;
842
843/*
844 * vdev aux states.  When a vdev is in the CANT_OPEN state, the aux field
845 * of the vdev stats structure uses these constants to distinguish why.
846 */
847typedef enum vdev_aux {
848	VDEV_AUX_NONE,		/* no error				*/
849	VDEV_AUX_OPEN_FAILED,	/* ldi_open_*() or vn_open() failed	*/
850	VDEV_AUX_CORRUPT_DATA,	/* bad label or disk contents		*/
851	VDEV_AUX_NO_REPLICAS,	/* insufficient number of replicas	*/
852	VDEV_AUX_BAD_GUID_SUM,	/* vdev guid sum doesn't match		*/
853	VDEV_AUX_TOO_SMALL,	/* vdev size is too small		*/
854	VDEV_AUX_BAD_LABEL,	/* the label is OK but invalid		*/
855	VDEV_AUX_VERSION_NEWER,	/* on-disk version is too new		*/
856	VDEV_AUX_VERSION_OLDER,	/* on-disk version is too old		*/
857	VDEV_AUX_SPARED		/* hot spare used in another pool	*/
858} vdev_aux_t;
859
860/*
861 * pool state.  The following states are written to disk as part of the normal
862 * SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE.  The remaining states are
863 * software abstractions used at various levels to communicate pool state.
864 */
865typedef enum pool_state {
866	POOL_STATE_ACTIVE = 0,		/* In active use		*/
867	POOL_STATE_EXPORTED,		/* Explicitly exported		*/
868	POOL_STATE_DESTROYED,		/* Explicitly destroyed		*/
869	POOL_STATE_SPARE,		/* Reserved for hot spare use	*/
870	POOL_STATE_UNINITIALIZED,	/* Internal spa_t state		*/
871	POOL_STATE_UNAVAIL,		/* Internal libzfs state	*/
872	POOL_STATE_POTENTIALLY_ACTIVE	/* Internal libzfs state	*/
873} pool_state_t;
874
875/*
876 * The uberblock version is incremented whenever an incompatible on-disk
877 * format change is made to the SPA, DMU, or ZAP.
878 *
879 * Note: the first two fields should never be moved.  When a storage pool
880 * is opened, the uberblock must be read off the disk before the version
881 * can be checked.  If the ub_version field is moved, we may not detect
882 * version mismatch.  If the ub_magic field is moved, applications that
883 * expect the magic number in the first word won't work.
884 */
885#define	UBERBLOCK_MAGIC		0x00bab10c		/* oo-ba-bloc!	*/
886#define	UBERBLOCK_SHIFT		10			/* up to 1K	*/
887
888#define	MMP_MAGIC		0xa11cea11		/* all-see-all  */
889
890#define	MMP_INTERVAL_VALID_BIT	0x01
891#define	MMP_SEQ_VALID_BIT	0x02
892#define	MMP_FAIL_INT_VALID_BIT	0x04
893
894#define	MMP_VALID(ubp)		(ubp->ub_magic == UBERBLOCK_MAGIC && \
895				    ubp->ub_mmp_magic == MMP_MAGIC)
896#define	MMP_INTERVAL_VALID(ubp)	(MMP_VALID(ubp) && (ubp->ub_mmp_config & \
897				    MMP_INTERVAL_VALID_BIT))
898#define	MMP_SEQ_VALID(ubp)	(MMP_VALID(ubp) && (ubp->ub_mmp_config & \
899				    MMP_SEQ_VALID_BIT))
900#define	MMP_FAIL_INT_VALID(ubp)	(MMP_VALID(ubp) && (ubp->ub_mmp_config & \
901				    MMP_FAIL_INT_VALID_BIT))
902
903#define	MMP_INTERVAL(ubp)	((ubp->ub_mmp_config & 0x00000000FFFFFF00) \
904				    >> 8)
905#define	MMP_SEQ(ubp)		((ubp->ub_mmp_config & 0x0000FFFF00000000) \
906				    >> 32)
907#define	MMP_FAIL_INT(ubp)	((ubp->ub_mmp_config & 0xFFFF000000000000) \
908				    >> 48)
909
910typedef struct uberblock {
911	uint64_t	ub_magic;	/* UBERBLOCK_MAGIC		*/
912	uint64_t	ub_version;	/* SPA_VERSION			*/
913	uint64_t	ub_txg;		/* txg of last sync		*/
914	uint64_t	ub_guid_sum;	/* sum of all vdev guids	*/
915	uint64_t	ub_timestamp;	/* UTC time of last sync	*/
916	blkptr_t	ub_rootbp;	/* MOS objset_phys_t		*/
917	/* highest SPA_VERSION supported by software that wrote this txg */
918	uint64_t	ub_software_version;
919	/* Maybe missing in uberblocks we read, but always written */
920	uint64_t	ub_mmp_magic;
921	/*
922	 * If ub_mmp_delay == 0 and ub_mmp_magic is valid, MMP is off.
923	 * Otherwise, nanosec since last MMP write.
924	 */
925	uint64_t	ub_mmp_delay;
926
927	/*
928	 * The ub_mmp_config contains the multihost write interval, multihost
929	 * fail intervals, sequence number for sub-second granularity, and
930	 * valid bit mask.  This layout is as follows:
931	 *
932	 *   64      56      48      40      32      24      16      8       0
933	 *   +-------+-------+-------+-------+-------+-------+-------+-------+
934	 * 0 | Fail Intervals|      Seq      |   Write Interval (ms) | VALID |
935	 *   +-------+-------+-------+-------+-------+-------+-------+-------+
936	 *
937	 * This allows a write_interval of (2^24/1000)s, over 4.5 hours
938	 *
939	 * VALID Bits:
940	 * - 0x01 - Write Interval (ms)
941	 * - 0x02 - Sequence number exists
942	 * - 0x04 - Fail Intervals
943	 * - 0xf8 - Reserved
944	 */
945	uint64_t	ub_mmp_config;
946
947	/*
948	 * ub_checkpoint_txg indicates two things about the current uberblock:
949	 *
950	 * 1] If it is not zero then this uberblock is a checkpoint. If it is
951	 *    zero, then this uberblock is not a checkpoint.
952	 *
953	 * 2] On checkpointed uberblocks, the value of ub_checkpoint_txg is
954	 *    the ub_txg that the uberblock had at the time we moved it to
955	 *    the MOS config.
956	 *
957	 * The field is set when we checkpoint the uberblock and continues to
958	 * hold that value even after we've rewound (unlike the ub_txg that
959	 * is reset to a higher value).
960	 *
961	 * Besides checks used to determine whether we are reopening the
962	 * pool from a checkpointed uberblock [see spa_ld_select_uberblock()],
963	 * the value of the field is used to determine which ZIL blocks have
964	 * been allocated according to the ms_sm when we are rewinding to a
965	 * checkpoint. Specifically, if blk_birth > ub_checkpoint_txg, then
966	 * the ZIL block is not allocated [see uses of spa_min_claim_txg()].
967	 */
968	uint64_t	ub_checkpoint_txg;
969} uberblock_t;
970
971/*
972 * Flags.
973 */
974#define	DNODE_MUST_BE_ALLOCATED	1
975#define	DNODE_MUST_BE_FREE	2
976
977/*
978 * Fixed constants.
979 */
980#define	DNODE_SHIFT		9	/* 512 bytes */
981#define	DN_MIN_INDBLKSHIFT	12	/* 4k */
982#define	DN_MAX_INDBLKSHIFT	17	/* 128k */
983#define	DNODE_BLOCK_SHIFT	14	/* 16k */
984#define	DNODE_CORE_SIZE		64	/* 64 bytes for dnode sans blkptrs */
985#define	DN_MAX_OBJECT_SHIFT	48	/* 256 trillion (zfs_fid_t limit) */
986#define	DN_MAX_OFFSET_SHIFT	64	/* 2^64 bytes in a dnode */
987
988/*
989 * Derived constants.
990 */
991#define	DNODE_MIN_SIZE		(1 << DNODE_SHIFT)
992#define	DNODE_MAX_SIZE		(1 << DNODE_BLOCK_SHIFT)
993#define	DNODE_BLOCK_SIZE	(1 << DNODE_BLOCK_SHIFT)
994#define	DNODE_MIN_SLOTS		(DNODE_MIN_SIZE >> DNODE_SHIFT)
995#define	DNODE_MAX_SLOTS		(DNODE_MAX_SIZE >> DNODE_SHIFT)
996#define	DN_BONUS_SIZE(dnsize)	((dnsize) - DNODE_CORE_SIZE - \
997	(1 << SPA_BLKPTRSHIFT))
998#define	DN_SLOTS_TO_BONUSLEN(slots)	DN_BONUS_SIZE((slots) << DNODE_SHIFT)
999#define	DN_OLD_MAX_BONUSLEN		(DN_BONUS_SIZE(DNODE_MIN_SIZE))
1000#define	DN_MAX_NBLKPTR		((DNODE_MIN_SIZE - DNODE_CORE_SIZE) >> \
1001	SPA_BLKPTRSHIFT)
1002#define	DN_MAX_OBJECT		(1ULL << DN_MAX_OBJECT_SHIFT)
1003#define	DN_ZERO_BONUSLEN	(DN_BONUS_SIZE(DNODE_MAX_SIZE) + 1)
1004
1005#define	DNODES_PER_BLOCK_SHIFT	(DNODE_BLOCK_SHIFT - DNODE_SHIFT)
1006#define	DNODES_PER_BLOCK	(1ULL << DNODES_PER_BLOCK_SHIFT)
1007#define	DNODES_PER_LEVEL_SHIFT	(DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT)
1008
1009/* The +2 here is a cheesy way to round up */
1010#define	DN_MAX_LEVELS	(2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \
1011	(DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT)))
1012
1013#define	DN_BONUS(dnp)	((void*)((dnp)->dn_bonus + \
1014	(((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t))))
1015
1016#define	DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \
1017	(dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT)
1018
1019#define	EPB(blkshift, typeshift)	(1 << (blkshift - typeshift))
1020
1021/* Is dn_used in bytes?  if not, it's in multiples of SPA_MINBLOCKSIZE */
1022#define	DNODE_FLAG_USED_BYTES		(1<<0)
1023#define	DNODE_FLAG_USERUSED_ACCOUNTED	(1<<1)
1024
1025/* Does dnode have a SA spill blkptr in bonus? */
1026#define	DNODE_FLAG_SPILL_BLKPTR	(1<<2)
1027
1028typedef struct dnode_phys {
1029	uint8_t dn_type;		/* dmu_object_type_t */
1030	uint8_t dn_indblkshift;		/* ln2(indirect block size) */
1031	uint8_t dn_nlevels;		/* 1=dn_blkptr->data blocks */
1032	uint8_t dn_nblkptr;		/* length of dn_blkptr */
1033	uint8_t dn_bonustype;		/* type of data in bonus buffer */
1034	uint8_t	dn_checksum;		/* ZIO_CHECKSUM type */
1035	uint8_t	dn_compress;		/* ZIO_COMPRESS type */
1036	uint8_t dn_flags;		/* DNODE_FLAG_* */
1037	uint16_t dn_datablkszsec;	/* data block size in 512b sectors */
1038	uint16_t dn_bonuslen;		/* length of dn_bonus */
1039	uint8_t dn_extra_slots;		/* # of subsequent slots consumed */
1040	uint8_t dn_pad2[3];
1041
1042	/* accounting is protected by dn_dirty_mtx */
1043	uint64_t dn_maxblkid;		/* largest allocated block ID */
1044	uint64_t dn_used;		/* bytes (or sectors) of disk space */
1045
1046	uint64_t dn_pad3[4];
1047
1048	/*
1049	 * The tail region is 448 bytes for a 512 byte dnode, and
1050	 * correspondingly larger for larger dnode sizes. The spill
1051	 * block pointer, when present, is always at the end of the tail
1052	 * region. There are three ways this space may be used, using
1053	 * a 512 byte dnode for this diagram:
1054	 *
1055	 * 0       64      128     192     256     320     384     448 (offset)
1056	 * +---------------+---------------+---------------+-------+
1057	 * | dn_blkptr[0]  | dn_blkptr[1]  | dn_blkptr[2]  | /     |
1058	 * +---------------+---------------+---------------+-------+
1059	 * | dn_blkptr[0]  | dn_bonus[0..319]                      |
1060	 * +---------------+-----------------------+---------------+
1061	 * | dn_blkptr[0]  | dn_bonus[0..191]      | dn_spill      |
1062	 * +---------------+-----------------------+---------------+
1063	 */
1064	union {
1065		blkptr_t dn_blkptr[1+DN_OLD_MAX_BONUSLEN/sizeof (blkptr_t)];
1066		struct {
1067			blkptr_t __dn_ignore1;
1068			uint8_t dn_bonus[DN_OLD_MAX_BONUSLEN];
1069		};
1070		struct {
1071			blkptr_t __dn_ignore2;
1072			uint8_t __dn_ignore3[DN_OLD_MAX_BONUSLEN -
1073			    sizeof (blkptr_t)];
1074			blkptr_t dn_spill;
1075		};
1076	};
1077} dnode_phys_t;
1078
1079#define	DN_SPILL_BLKPTR(dnp)	(blkptr_t *)((char *)(dnp) + \
1080	(((dnp)->dn_extra_slots + 1) << DNODE_SHIFT) - (1 << SPA_BLKPTRSHIFT))
1081
1082typedef enum dmu_object_byteswap {
1083	DMU_BSWAP_UINT8,
1084	DMU_BSWAP_UINT16,
1085	DMU_BSWAP_UINT32,
1086	DMU_BSWAP_UINT64,
1087	DMU_BSWAP_ZAP,
1088	DMU_BSWAP_DNODE,
1089	DMU_BSWAP_OBJSET,
1090	DMU_BSWAP_ZNODE,
1091	DMU_BSWAP_OLDACL,
1092	DMU_BSWAP_ACL,
1093	/*
1094	 * Allocating a new byteswap type number makes the on-disk format
1095	 * incompatible with any other format that uses the same number.
1096	 *
1097	 * Data can usually be structured to work with one of the
1098	 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types.
1099	 */
1100	DMU_BSWAP_NUMFUNCS
1101} dmu_object_byteswap_t;
1102
1103#define	DMU_OT_NEWTYPE 0x80
1104#define	DMU_OT_METADATA 0x40
1105#define	DMU_OT_BYTESWAP_MASK 0x3f
1106
1107/*
1108 * Defines a uint8_t object type. Object types specify if the data
1109 * in the object is metadata (boolean) and how to byteswap the data
1110 * (dmu_object_byteswap_t).
1111 */
1112#define	DMU_OT(byteswap, metadata) \
1113	(DMU_OT_NEWTYPE | \
1114	((metadata) ? DMU_OT_METADATA : 0) | \
1115	((byteswap) & DMU_OT_BYTESWAP_MASK))
1116
1117typedef enum dmu_object_type {
1118	DMU_OT_NONE,
1119	/* general: */
1120	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
1121	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
1122	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
1123	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
1124	DMU_OT_BPLIST,			/* UINT64 */
1125	DMU_OT_BPLIST_HDR,		/* UINT64 */
1126	/* spa: */
1127	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
1128	DMU_OT_SPACE_MAP,		/* UINT64 */
1129	/* zil: */
1130	DMU_OT_INTENT_LOG,		/* UINT64 */
1131	/* dmu: */
1132	DMU_OT_DNODE,			/* DNODE */
1133	DMU_OT_OBJSET,			/* OBJSET */
1134	/* dsl: */
1135	DMU_OT_DSL_DIR,			/* UINT64 */
1136	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
1137	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
1138	DMU_OT_DSL_PROPS,		/* ZAP */
1139	DMU_OT_DSL_DATASET,		/* UINT64 */
1140	/* zpl: */
1141	DMU_OT_ZNODE,			/* ZNODE */
1142	DMU_OT_OLDACL,			/* Old ACL */
1143	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
1144	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
1145	DMU_OT_MASTER_NODE,		/* ZAP */
1146	DMU_OT_UNLINKED_SET,		/* ZAP */
1147	/* zvol: */
1148	DMU_OT_ZVOL,			/* UINT8 */
1149	DMU_OT_ZVOL_PROP,		/* ZAP */
1150	/* other; for testing only! */
1151	DMU_OT_PLAIN_OTHER,		/* UINT8 */
1152	DMU_OT_UINT64_OTHER,		/* UINT64 */
1153	DMU_OT_ZAP_OTHER,		/* ZAP */
1154	/* new object types: */
1155	DMU_OT_ERROR_LOG,		/* ZAP */
1156	DMU_OT_SPA_HISTORY,		/* UINT8 */
1157	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
1158	DMU_OT_POOL_PROPS,		/* ZAP */
1159	DMU_OT_DSL_PERMS,		/* ZAP */
1160	DMU_OT_ACL,			/* ACL */
1161	DMU_OT_SYSACL,			/* SYSACL */
1162	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
1163	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
1164	DMU_OT_NEXT_CLONES,		/* ZAP */
1165	DMU_OT_SCAN_QUEUE,		/* ZAP */
1166	DMU_OT_USERGROUP_USED,		/* ZAP */
1167	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
1168	DMU_OT_USERREFS,		/* ZAP */
1169	DMU_OT_DDT_ZAP,			/* ZAP */
1170	DMU_OT_DDT_STATS,		/* ZAP */
1171	DMU_OT_SA,			/* System attr */
1172	DMU_OT_SA_MASTER_NODE,		/* ZAP */
1173	DMU_OT_SA_ATTR_REGISTRATION,	/* ZAP */
1174	DMU_OT_SA_ATTR_LAYOUTS,		/* ZAP */
1175	DMU_OT_SCAN_XLATE,		/* ZAP */
1176	DMU_OT_DEDUP,			/* fake dedup BP from ddt_bp_create() */
1177	DMU_OT_NUMTYPES,
1178
1179	/*
1180	 * Names for valid types declared with DMU_OT().
1181	 */
1182	DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE),
1183	DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE),
1184	DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE),
1185	DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE),
1186	DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE),
1187	DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE),
1188	DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE),
1189	DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE),
1190	DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE),
1191	DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE)
1192} dmu_object_type_t;
1193
1194typedef enum dmu_objset_type {
1195	DMU_OST_NONE,
1196	DMU_OST_META,
1197	DMU_OST_ZFS,
1198	DMU_OST_ZVOL,
1199	DMU_OST_OTHER,			/* For testing only! */
1200	DMU_OST_ANY,			/* Be careful! */
1201	DMU_OST_NUMTYPES
1202} dmu_objset_type_t;
1203
1204#define	ZAP_MAXVALUELEN	(1024 * 8)
1205
1206/*
1207 * header for all bonus and spill buffers.
1208 * The header has a fixed portion with a variable number
1209 * of "lengths" depending on the number of variable sized
1210 * attribues which are determined by the "layout number"
1211 */
1212
1213#define	SA_MAGIC	0x2F505A  /* ZFS SA */
1214typedef struct sa_hdr_phys {
1215	uint32_t sa_magic;
1216	uint16_t sa_layout_info;  /* Encoded with hdrsize and layout number */
1217	uint16_t sa_lengths[1];	/* optional sizes for variable length attrs */
1218	/* ... Data follows the lengths.  */
1219} sa_hdr_phys_t;
1220
1221/*
1222 * sa_hdr_phys -> sa_layout_info
1223 *
1224 * 16      10       0
1225 * +--------+-------+
1226 * | hdrsz  |layout |
1227 * +--------+-------+
1228 *
1229 * Bits 0-10 are the layout number
1230 * Bits 11-16 are the size of the header.
1231 * The hdrsize is the number * 8
1232 *
1233 * For example.
1234 * hdrsz of 1 ==> 8 byte header
1235 *          2 ==> 16 byte header
1236 *
1237 */
1238
1239#define	SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10)
1240#define	SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 16, 3, 0)
1241#define	SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \
1242{ \
1243	BF32_SET_SB(x, 10, 6, 3, 0, size); \
1244	BF32_SET(x, 0, 10, num); \
1245}
1246
1247#define	SA_MODE_OFFSET		0
1248#define	SA_SIZE_OFFSET		8
1249#define	SA_GEN_OFFSET		16
1250#define	SA_UID_OFFSET		24
1251#define	SA_GID_OFFSET		32
1252#define	SA_PARENT_OFFSET	40
1253#define	SA_SYMLINK_OFFSET	160
1254
1255#define	ZIO_OBJSET_MAC_LEN		32
1256
1257/*
1258 * Intent log header - this on disk structure holds fields to manage
1259 * the log.  All fields are 64 bit to easily handle cross architectures.
1260 */
1261typedef struct zil_header {
1262	uint64_t zh_claim_txg;	/* txg in which log blocks were claimed */
1263	uint64_t zh_replay_seq;	/* highest replayed sequence number */
1264	blkptr_t zh_log;	/* log chain */
1265	uint64_t zh_claim_seq;	/* highest claimed sequence number */
1266	uint64_t zh_pad[5];
1267} zil_header_t;
1268
1269#define	OBJSET_PHYS_SIZE_V2 2048
1270#define	OBJSET_PHYS_SIZE_V3 4096
1271
1272typedef struct objset_phys {
1273	dnode_phys_t os_meta_dnode;
1274	zil_header_t os_zil_header;
1275	uint64_t os_type;
1276	uint64_t os_flags;
1277	uint8_t os_portable_mac[ZIO_OBJSET_MAC_LEN];
1278	uint8_t os_local_mac[ZIO_OBJSET_MAC_LEN];
1279	char os_pad0[OBJSET_PHYS_SIZE_V2 - sizeof (dnode_phys_t)*3 -
1280		sizeof (zil_header_t) - sizeof (uint64_t)*2 -
1281		2*ZIO_OBJSET_MAC_LEN];
1282	dnode_phys_t os_userused_dnode;
1283	dnode_phys_t os_groupused_dnode;
1284	dnode_phys_t os_projectused_dnode;
1285	char os_pad1[OBJSET_PHYS_SIZE_V3 - OBJSET_PHYS_SIZE_V2 -
1286	    sizeof (dnode_phys_t)];
1287} objset_phys_t;
1288
1289typedef struct dsl_dir_phys {
1290	uint64_t dd_creation_time; /* not actually used */
1291	uint64_t dd_head_dataset_obj;
1292	uint64_t dd_parent_obj;
1293	uint64_t dd_clone_parent_obj;
1294	uint64_t dd_child_dir_zapobj;
1295	/*
1296	 * how much space our children are accounting for; for leaf
1297	 * datasets, == physical space used by fs + snaps
1298	 */
1299	uint64_t dd_used_bytes;
1300	uint64_t dd_compressed_bytes;
1301	uint64_t dd_uncompressed_bytes;
1302	/* Administrative quota setting */
1303	uint64_t dd_quota;
1304	/* Administrative reservation setting */
1305	uint64_t dd_reserved;
1306	uint64_t dd_props_zapobj;
1307	uint64_t dd_pad[21]; /* pad out to 256 bytes for good measure */
1308} dsl_dir_phys_t;
1309
1310typedef struct dsl_dataset_phys {
1311	uint64_t ds_dir_obj;
1312	uint64_t ds_prev_snap_obj;
1313	uint64_t ds_prev_snap_txg;
1314	uint64_t ds_next_snap_obj;
1315	uint64_t ds_snapnames_zapobj;	/* zap obj of snaps; ==0 for snaps */
1316	uint64_t ds_num_children;	/* clone/snap children; ==0 for head */
1317	uint64_t ds_creation_time;	/* seconds since 1970 */
1318	uint64_t ds_creation_txg;
1319	uint64_t ds_deadlist_obj;
1320	uint64_t ds_used_bytes;
1321	uint64_t ds_compressed_bytes;
1322	uint64_t ds_uncompressed_bytes;
1323	uint64_t ds_unique_bytes;	/* only relevant to snapshots */
1324	/*
1325	 * The ds_fsid_guid is a 56-bit ID that can change to avoid
1326	 * collisions.  The ds_guid is a 64-bit ID that will never
1327	 * change, so there is a small probability that it will collide.
1328	 */
1329	uint64_t ds_fsid_guid;
1330	uint64_t ds_guid;
1331	uint64_t ds_flags;
1332	blkptr_t ds_bp;
1333	uint64_t ds_pad[8]; /* pad out to 320 bytes for good measure */
1334} dsl_dataset_phys_t;
1335
1336/*
1337 * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
1338 */
1339#define	DMU_POOL_DIRECTORY_OBJECT	1
1340#define	DMU_POOL_CONFIG			"config"
1341#define	DMU_POOL_FEATURES_FOR_READ	"features_for_read"
1342#define	DMU_POOL_ROOT_DATASET		"root_dataset"
1343#define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
1344#define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
1345#define	DMU_POOL_ERRLOG_LAST		"errlog_last"
1346#define	DMU_POOL_SPARES			"spares"
1347#define	DMU_POOL_DEFLATE		"deflate"
1348#define	DMU_POOL_HISTORY		"history"
1349#define	DMU_POOL_PROPS			"pool_props"
1350#define	DMU_POOL_CHECKSUM_SALT		"org.illumos:checksum_salt"
1351#define	DMU_POOL_REMOVING		"com.delphix:removing"
1352#define	DMU_POOL_OBSOLETE_BPOBJ		"com.delphix:obsolete_bpobj"
1353#define	DMU_POOL_CONDENSING_INDIRECT	"com.delphix:condensing_indirect"
1354#define	DMU_POOL_ZPOOL_CHECKPOINT       "com.delphix:zpool_checkpoint"
1355
1356#define	ZAP_MAGIC 0x2F52AB2ABULL
1357
1358#define	FZAP_BLOCK_SHIFT(zap)	((zap)->zap_block_shift)
1359
1360#define	ZAP_MAXCD		(uint32_t)(-1)
1361#define	ZAP_HASHBITS		28
1362#define	MZAP_ENT_LEN		64
1363#define	MZAP_NAME_LEN		(MZAP_ENT_LEN - 8 - 4 - 2)
1364#define	MZAP_MAX_BLKSZ		SPA_OLD_MAXBLOCKSIZE
1365
1366typedef struct mzap_ent_phys {
1367	uint64_t mze_value;
1368	uint32_t mze_cd;
1369	uint16_t mze_pad;	/* in case we want to chain them someday */
1370	char mze_name[MZAP_NAME_LEN];
1371} mzap_ent_phys_t;
1372
1373typedef struct mzap_phys {
1374	uint64_t mz_block_type;	/* ZBT_MICRO */
1375	uint64_t mz_salt;
1376	uint64_t mz_normflags;
1377	uint64_t mz_pad[5];
1378	mzap_ent_phys_t mz_chunk[1];
1379	/* actually variable size depending on block size */
1380} mzap_phys_t;
1381
1382/*
1383 * The (fat) zap is stored in one object. It is an array of
1384 * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of:
1385 *
1386 * ptrtbl fits in first block:
1387 * 	[zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ...
1388 *
1389 * ptrtbl too big for first block:
1390 * 	[zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ...
1391 *
1392 */
1393
1394#define	ZBT_LEAF		((1ULL << 63) + 0)
1395#define	ZBT_HEADER		((1ULL << 63) + 1)
1396#define	ZBT_MICRO		((1ULL << 63) + 3)
1397/* any other values are ptrtbl blocks */
1398
1399/*
1400 * the embedded pointer table takes up half a block:
1401 * block size / entry size (2^3) / 2
1402 */
1403#define	ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1)
1404
1405/*
1406 * The embedded pointer table starts half-way through the block.  Since
1407 * the pointer table itself is half the block, it starts at (64-bit)
1408 * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)).
1409 */
1410#define	ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \
1411	((uint64_t *)(zap)->zap_phys) \
1412	[(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))]
1413
1414/*
1415 * TAKE NOTE:
1416 * If zap_phys_t is modified, zap_byteswap() must be modified.
1417 */
1418typedef struct zap_phys {
1419	uint64_t zap_block_type;	/* ZBT_HEADER */
1420	uint64_t zap_magic;		/* ZAP_MAGIC */
1421
1422	struct zap_table_phys {
1423		uint64_t zt_blk;	/* starting block number */
1424		uint64_t zt_numblks;	/* number of blocks */
1425		uint64_t zt_shift;	/* bits to index it */
1426		uint64_t zt_nextblk;	/* next (larger) copy start block */
1427		uint64_t zt_blks_copied; /* number source blocks copied */
1428	} zap_ptrtbl;
1429
1430	uint64_t zap_freeblk;		/* the next free block */
1431	uint64_t zap_num_leafs;		/* number of leafs */
1432	uint64_t zap_num_entries;	/* number of entries */
1433	uint64_t zap_salt;		/* salt to stir into hash function */
1434	uint64_t zap_normflags;		/* flags for u8_textprep_str() */
1435	uint64_t zap_flags;		/* zap_flags_t */
1436	/*
1437	 * This structure is followed by padding, and then the embedded
1438	 * pointer table.  The embedded pointer table takes up second
1439	 * half of the block.  It is accessed using the
1440	 * ZAP_EMBEDDED_PTRTBL_ENT() macro.
1441	 */
1442} zap_phys_t;
1443
1444typedef struct zap_table_phys zap_table_phys_t;
1445
1446struct spa;
1447typedef struct fat_zap {
1448	int zap_block_shift;			/* block size shift */
1449	zap_phys_t *zap_phys;
1450	const struct spa *zap_spa;
1451	const dnode_phys_t *zap_dnode;
1452} fat_zap_t;
1453
1454#define	ZAP_LEAF_MAGIC 0x2AB1EAF
1455
1456/* chunk size = 24 bytes */
1457#define	ZAP_LEAF_CHUNKSIZE 24
1458
1459/*
1460 * The amount of space available for chunks is:
1461 * block size (1<<l->l_bs) - hash entry size (2) * number of hash
1462 * entries - header space (2*chunksize)
1463 */
1464#define	ZAP_LEAF_NUMCHUNKS(l) \
1465	(((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \
1466	ZAP_LEAF_CHUNKSIZE - 2)
1467
1468/*
1469 * The amount of space within the chunk available for the array is:
1470 * chunk size - space for type (1) - space for next pointer (2)
1471 */
1472#define	ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3)
1473
1474#define	ZAP_LEAF_ARRAY_NCHUNKS(bytes) \
1475	(((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES)
1476
1477/*
1478 * Low water mark:  when there are only this many chunks free, start
1479 * growing the ptrtbl.  Ideally, this should be larger than a
1480 * "reasonably-sized" entry.  20 chunks is more than enough for the
1481 * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value),
1482 * while still being only around 3% for 16k blocks.
1483 */
1484#define	ZAP_LEAF_LOW_WATER (20)
1485
1486/*
1487 * The leaf hash table has block size / 2^5 (32) number of entries,
1488 * which should be more than enough for the maximum number of entries,
1489 * which is less than block size / CHUNKSIZE (24) / minimum number of
1490 * chunks per entry (3).
1491 */
1492#define	ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5)
1493#define	ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l))
1494
1495/*
1496 * The chunks start immediately after the hash table.  The end of the
1497 * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a
1498 * chunk_t.
1499 */
1500#define	ZAP_LEAF_CHUNK(l, idx) \
1501	((zap_leaf_chunk_t *) \
1502	((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx]
1503#define	ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry)
1504
1505typedef enum zap_chunk_type {
1506	ZAP_CHUNK_FREE = 253,
1507	ZAP_CHUNK_ENTRY = 252,
1508	ZAP_CHUNK_ARRAY = 251,
1509	ZAP_CHUNK_TYPE_MAX = 250
1510} zap_chunk_type_t;
1511
1512/*
1513 * TAKE NOTE:
1514 * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified.
1515 */
1516typedef struct zap_leaf_phys {
1517	struct zap_leaf_header {
1518		uint64_t lh_block_type;		/* ZBT_LEAF */
1519		uint64_t lh_pad1;
1520		uint64_t lh_prefix;		/* hash prefix of this leaf */
1521		uint32_t lh_magic;		/* ZAP_LEAF_MAGIC */
1522		uint16_t lh_nfree;		/* number free chunks */
1523		uint16_t lh_nentries;		/* number of entries */
1524		uint16_t lh_prefix_len;		/* num bits used to id this */
1525
1526/* above is accessable to zap, below is zap_leaf private */
1527
1528		uint16_t lh_freelist;		/* chunk head of free list */
1529		uint8_t lh_pad2[12];
1530	} l_hdr; /* 2 24-byte chunks */
1531
1532	/*
1533	 * The header is followed by a hash table with
1534	 * ZAP_LEAF_HASH_NUMENTRIES(zap) entries.  The hash table is
1535	 * followed by an array of ZAP_LEAF_NUMCHUNKS(zap)
1536	 * zap_leaf_chunk structures.  These structures are accessed
1537	 * with the ZAP_LEAF_CHUNK() macro.
1538	 */
1539
1540	uint16_t l_hash[1];
1541} zap_leaf_phys_t;
1542
1543typedef union zap_leaf_chunk {
1544	struct zap_leaf_entry {
1545		uint8_t le_type; 		/* always ZAP_CHUNK_ENTRY */
1546		uint8_t le_value_intlen;	/* size of ints */
1547		uint16_t le_next;		/* next entry in hash chain */
1548		uint16_t le_name_chunk;		/* first chunk of the name */
1549		uint16_t le_name_numints;	/* bytes in name, incl null */
1550		uint16_t le_value_chunk;	/* first chunk of the value */
1551		uint16_t le_value_numints;	/* value length in ints */
1552		uint32_t le_cd;			/* collision differentiator */
1553		uint64_t le_hash;		/* hash value of the name */
1554	} l_entry;
1555	struct zap_leaf_array {
1556		uint8_t la_type;		/* always ZAP_CHUNK_ARRAY */
1557		uint8_t la_array[ZAP_LEAF_ARRAY_BYTES];
1558		uint16_t la_next;		/* next blk or CHAIN_END */
1559	} l_array;
1560	struct zap_leaf_free {
1561		uint8_t lf_type;		/* always ZAP_CHUNK_FREE */
1562		uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES];
1563		uint16_t lf_next;	/* next in free list, or CHAIN_END */
1564	} l_free;
1565} zap_leaf_chunk_t;
1566
1567typedef struct zap_leaf {
1568	int l_bs;			/* block size shift */
1569	zap_leaf_phys_t *l_phys;
1570} zap_leaf_t;
1571
1572/*
1573 * Define special zfs pflags
1574 */
1575#define	ZFS_XATTR	0x1		/* is an extended attribute */
1576#define	ZFS_INHERIT_ACE	0x2		/* ace has inheritable ACEs */
1577#define	ZFS_ACL_TRIVIAL 0x4		/* files ACL is trivial */
1578
1579#define	MASTER_NODE_OBJ	1
1580
1581/*
1582 * special attributes for master node.
1583 */
1584
1585#define	ZFS_FSID		"FSID"
1586#define	ZFS_UNLINKED_SET	"DELETE_QUEUE"
1587#define	ZFS_ROOT_OBJ		"ROOT"
1588#define	ZPL_VERSION_OBJ		"VERSION"
1589#define	ZFS_PROP_BLOCKPERPAGE	"BLOCKPERPAGE"
1590#define	ZFS_PROP_NOGROWBLOCKS	"NOGROWBLOCKS"
1591
1592#define	ZFS_FLAG_BLOCKPERPAGE	0x1
1593#define	ZFS_FLAG_NOGROWBLOCKS	0x2
1594
1595/*
1596 * ZPL version - rev'd whenever an incompatible on-disk format change
1597 * occurs.  Independent of SPA/DMU/ZAP versioning.
1598 */
1599
1600#define	ZPL_VERSION		1ULL
1601
1602/*
1603 * The directory entry has the type (currently unused on Solaris) in the
1604 * top 4 bits, and the object number in the low 48 bits.  The "middle"
1605 * 12 bits are unused.
1606 */
1607#define	ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4)
1608#define	ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48)
1609#define	ZFS_DIRENT_MAKE(type, obj) (((uint64_t)type << 60) | obj)
1610
1611typedef struct ace {
1612	uid_t		a_who;		/* uid or gid */
1613	uint32_t	a_access_mask;	/* read,write,... */
1614	uint16_t	a_flags;	/* see below */
1615	uint16_t	a_type;		/* allow or deny */
1616} ace_t;
1617
1618#define ACE_SLOT_CNT	6
1619
1620typedef struct zfs_znode_acl {
1621	uint64_t	z_acl_extern_obj;	  /* ext acl pieces */
1622	uint32_t	z_acl_count;		  /* Number of ACEs */
1623	uint16_t	z_acl_version;		  /* acl version */
1624	uint16_t	z_acl_pad;		  /* pad */
1625	ace_t		z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */
1626} zfs_znode_acl_t;
1627
1628/*
1629 * This is the persistent portion of the znode.  It is stored
1630 * in the "bonus buffer" of the file.  Short symbolic links
1631 * are also stored in the bonus buffer.
1632 */
1633typedef struct znode_phys {
1634	uint64_t zp_atime[2];		/*  0 - last file access time */
1635	uint64_t zp_mtime[2];		/* 16 - last file modification time */
1636	uint64_t zp_ctime[2];		/* 32 - last file change time */
1637	uint64_t zp_crtime[2];		/* 48 - creation time */
1638	uint64_t zp_gen;		/* 64 - generation (txg of creation) */
1639	uint64_t zp_mode;		/* 72 - file mode bits */
1640	uint64_t zp_size;		/* 80 - size of file */
1641	uint64_t zp_parent;		/* 88 - directory parent (`..') */
1642	uint64_t zp_links;		/* 96 - number of links to file */
1643	uint64_t zp_xattr;		/* 104 - DMU object for xattrs */
1644	uint64_t zp_rdev;		/* 112 - dev_t for VBLK & VCHR files */
1645	uint64_t zp_flags;		/* 120 - persistent flags */
1646	uint64_t zp_uid;		/* 128 - file owner */
1647	uint64_t zp_gid;		/* 136 - owning group */
1648	uint64_t zp_pad[4];		/* 144 - future */
1649	zfs_znode_acl_t zp_acl;		/* 176 - 263 ACL */
1650	/*
1651	 * Data may pad out any remaining bytes in the znode buffer, eg:
1652	 *
1653	 * |<---------------------- dnode_phys (512) ------------------------>|
1654	 * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->|
1655	 *			|<---- znode (264) ---->|<---- data (56) ---->|
1656	 *
1657	 * At present, we only use this space to store symbolic links.
1658	 */
1659} znode_phys_t;
1660
1661/*
1662 * In-core vdev representation.
1663 */
1664struct vdev;
1665struct spa;
1666typedef int vdev_phys_read_t(struct vdev *, void *, off_t, void *, size_t);
1667typedef int vdev_phys_write_t(struct vdev *, off_t, void *, size_t);
1668typedef int vdev_read_t(struct vdev *, const blkptr_t *, void *, off_t, size_t);
1669
1670typedef STAILQ_HEAD(vdev_list, vdev) vdev_list_t;
1671
1672typedef struct vdev_indirect_mapping_entry_phys {
1673	/*
1674	 * Decode with DVA_MAPPING_* macros.
1675	 * Contains:
1676	 *   the source offset (low 63 bits)
1677	 *   the one-bit "mark", used for garbage collection (by zdb)
1678	 */
1679	uint64_t vimep_src;
1680
1681	/*
1682	 * Note: the DVA's asize is 24 bits, and can thus store ranges
1683	 * up to 8GB.
1684	 */
1685	dva_t	vimep_dst;
1686} vdev_indirect_mapping_entry_phys_t;
1687
1688#define	DVA_MAPPING_GET_SRC_OFFSET(vimep)	\
1689	BF64_GET_SB((vimep)->vimep_src, 0, 63, SPA_MINBLOCKSHIFT, 0)
1690#define	DVA_MAPPING_SET_SRC_OFFSET(vimep, x)	\
1691	BF64_SET_SB((vimep)->vimep_src, 0, 63, SPA_MINBLOCKSHIFT, 0, x)
1692
1693typedef struct vdev_indirect_mapping_entry {
1694	vdev_indirect_mapping_entry_phys_t	vime_mapping;
1695	uint32_t				vime_obsolete_count;
1696	list_node_t				vime_node;
1697} vdev_indirect_mapping_entry_t;
1698
1699/*
1700 * This is stored in the bonus buffer of the mapping object, see comment of
1701 * vdev_indirect_config for more details.
1702 */
1703typedef struct vdev_indirect_mapping_phys {
1704	uint64_t	vimp_max_offset;
1705	uint64_t	vimp_bytes_mapped;
1706	uint64_t	vimp_num_entries; /* number of v_i_m_entry_phys_t's */
1707
1708	/*
1709	 * For each entry in the mapping object, this object contains an
1710	 * entry representing the number of bytes of that mapping entry
1711	 * that were no longer in use by the pool at the time this indirect
1712	 * vdev was last condensed.
1713	 */
1714	uint64_t	vimp_counts_object;
1715} vdev_indirect_mapping_phys_t;
1716
1717#define	VDEV_INDIRECT_MAPPING_SIZE_V0	(3 * sizeof (uint64_t))
1718
1719typedef struct vdev_indirect_mapping {
1720	uint64_t	vim_object;
1721	boolean_t	vim_havecounts;
1722
1723	/* vim_entries segment offset currently in memory. */
1724	uint64_t	vim_entry_offset;
1725	/* vim_entries segment size. */
1726	size_t		vim_num_entries;
1727
1728	/* Needed by dnode_read() */
1729	const void	*vim_spa;
1730	dnode_phys_t	*vim_dn;
1731
1732	/*
1733	 * An ordered array of mapping entries, sorted by source offset.
1734	 * Note that vim_entries is needed during a removal (and contains
1735	 * mappings that have been synced to disk so far) to handle frees
1736	 * from the removing device.
1737	 */
1738	vdev_indirect_mapping_entry_phys_t *vim_entries;
1739	objset_phys_t	*vim_objset;
1740	vdev_indirect_mapping_phys_t	*vim_phys;
1741} vdev_indirect_mapping_t;
1742
1743/*
1744 * On-disk indirect vdev state.
1745 *
1746 * An indirect vdev is described exclusively in the MOS config of a pool.
1747 * The config for an indirect vdev includes several fields, which are
1748 * accessed in memory by a vdev_indirect_config_t.
1749 */
1750typedef struct vdev_indirect_config {
1751	/*
1752	 * Object (in MOS) which contains the indirect mapping. This object
1753	 * contains an array of vdev_indirect_mapping_entry_phys_t ordered by
1754	 * vimep_src. The bonus buffer for this object is a
1755	 * vdev_indirect_mapping_phys_t. This object is allocated when a vdev
1756	 * removal is initiated.
1757	 *
1758	 * Note that this object can be empty if none of the data on the vdev
1759	 * has been copied yet.
1760	 */
1761	uint64_t	vic_mapping_object;
1762
1763	/*
1764	 * Object (in MOS) which contains the birth times for the mapping
1765	 * entries. This object contains an array of
1766	 * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus
1767	 * buffer for this object is a vdev_indirect_birth_phys_t. This object
1768	 * is allocated when a vdev removal is initiated.
1769	 *
1770	 * Note that this object can be empty if none of the vdev has yet been
1771	 * copied.
1772	 */
1773	uint64_t	vic_births_object;
1774
1775/*
1776 * This is the vdev ID which was removed previous to this vdev, or
1777 * UINT64_MAX if there are no previously removed vdevs.
1778 */
1779	uint64_t	vic_prev_indirect_vdev;
1780} vdev_indirect_config_t;
1781
1782typedef struct vdev {
1783	STAILQ_ENTRY(vdev) v_childlink;	/* link in parent's child list */
1784	STAILQ_ENTRY(vdev) v_alllink;	/* link in global vdev list */
1785	vdev_list_t	v_children;	/* children of this vdev */
1786	const char	*v_name;	/* vdev name */
1787	uint64_t	v_guid;		/* vdev guid */
1788	uint64_t	v_id;		/* index in parent */
1789	uint64_t	v_psize;	/* physical device capacity */
1790	int		v_ashift;	/* offset to block shift */
1791	int		v_nparity;	/* # parity for raidz */
1792	struct vdev	*v_top;		/* parent vdev */
1793	size_t		v_nchildren;	/* # children */
1794	vdev_state_t	v_state;	/* current state */
1795	vdev_phys_read_t *v_phys_read;	/* read from raw leaf vdev */
1796	vdev_phys_write_t *v_phys_write; /* write to raw leaf vdev */
1797	vdev_read_t	*v_read;	/* read from vdev */
1798	void		*v_priv;	/* data for read/write function */
1799	boolean_t	v_islog;
1800	struct spa	*v_spa;		/* link to spa */
1801	/*
1802	 * Values stored in the config for an indirect or removing vdev.
1803	 */
1804	vdev_indirect_config_t vdev_indirect_config;
1805	vdev_indirect_mapping_t *v_mapping;
1806} vdev_t;
1807
1808/*
1809 * In-core pool representation.
1810 */
1811typedef STAILQ_HEAD(spa_list, spa) spa_list_t;
1812
1813typedef struct spa {
1814	STAILQ_ENTRY(spa) spa_link;	/* link in global pool list */
1815	char		*spa_name;	/* pool name */
1816	uint64_t	spa_guid;	/* pool guid */
1817	uint64_t	spa_txg;	/* most recent transaction */
1818	struct uberblock *spa_uberblock;	/* best uberblock so far */
1819	vdev_t		*spa_root_vdev;	/* toplevel vdev container */
1820	objset_phys_t	*spa_mos;	/* MOS for this pool */
1821	zio_cksum_salt_t spa_cksum_salt;	/* secret salt for cksum */
1822	void		*spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS];
1823	boolean_t	spa_with_log;	/* this pool has log */
1824
1825	struct uberblock spa_uberblock_master;	/* best uberblock so far */
1826	objset_phys_t	spa_mos_master;		/* MOS for this pool */
1827	struct uberblock spa_uberblock_checkpoint; /* checkpoint uberblock */
1828	objset_phys_t	spa_mos_checkpoint;	/* Checkpoint MOS */
1829	void		*spa_bootenv;		/* bootenv from pool label */
1830} spa_t;
1831
1832/* IO related arguments. */
1833typedef struct zio {
1834	spa_t		*io_spa;
1835	blkptr_t	*io_bp;
1836	void		*io_data;
1837	uint64_t	io_size;
1838	uint64_t	io_offset;
1839
1840	/* Stuff for the vdev stack */
1841	vdev_t		*io_vd;
1842	void		*io_vsd;
1843
1844	int		io_error;
1845} zio_t;
1846
1847static void decode_embedded_bp_compressed(const blkptr_t *, void *);
1848
1849#endif /* _ZFSIMPL_H_ */
1850