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) 2013 by Delphix. All rights reserved.
60 */
61
62#ifndef _ZFSIMPL_H
63#define	_ZFSIMPL_H
64
65#define	MAXNAMELEN	256
66
67#define _NOTE(s)
68
69/* CRC64 table */
70#define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
71
72/*
73 * Macros for various sorts of alignment and rounding when the alignment
74 * is known to be a power of 2.
75 */
76#define	P2ALIGN(x, align)		((x) & -(align))
77#define	P2PHASE(x, align)		((x) & ((align) - 1))
78#define	P2NPHASE(x, align)		(-(x) & ((align) - 1))
79#define	P2ROUNDUP(x, align)		(-(-(x) & -(align)))
80#define	P2END(x, align)			(-(~(x) & -(align)))
81#define	P2PHASEUP(x, align, phase)	((phase) - (((phase) - (x)) & -(align)))
82#define	P2BOUNDARY(off, len, align)	\
83	(((off) ^ ((off) + (len) - 1)) > (align) - 1)
84
85/*
86 * General-purpose 32-bit and 64-bit bitfield encodings.
87 */
88#define	BF32_DECODE(x, low, len)	P2PHASE((x) >> (low), 1U << (len))
89#define	BF64_DECODE(x, low, len)	P2PHASE((x) >> (low), 1ULL << (len))
90#define	BF32_ENCODE(x, low, len)	(P2PHASE((x), 1U << (len)) << (low))
91#define	BF64_ENCODE(x, low, len)	(P2PHASE((x), 1ULL << (len)) << (low))
92
93#define	BF32_GET(x, low, len)		BF32_DECODE(x, low, len)
94#define	BF64_GET(x, low, len)		BF64_DECODE(x, low, len)
95
96#define	BF32_SET(x, low, len, val)	\
97	((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len))
98#define	BF64_SET(x, low, len, val)	\
99	((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len))
100
101#define	BF32_GET_SB(x, low, len, shift, bias)	\
102	((BF32_GET(x, low, len) + (bias)) << (shift))
103#define	BF64_GET_SB(x, low, len, shift, bias)	\
104	((BF64_GET(x, low, len) + (bias)) << (shift))
105
106#define	BF32_SET_SB(x, low, len, shift, bias, val)	\
107	BF32_SET(x, low, len, ((val) >> (shift)) - (bias))
108#define	BF64_SET_SB(x, low, len, shift, bias, val)	\
109	BF64_SET(x, low, len, ((val) >> (shift)) - (bias))
110
111/*
112 * Macros to reverse byte order
113 */
114#define	BSWAP_8(x)	((x) & 0xff)
115#define	BSWAP_16(x)	((BSWAP_8(x) << 8) | BSWAP_8((x) >> 8))
116#define	BSWAP_32(x)	((BSWAP_16(x) << 16) | BSWAP_16((x) >> 16))
117#define	BSWAP_64(x)	((BSWAP_32(x) << 32) | BSWAP_32((x) >> 32))
118
119#define	SPA_MINBLOCKSHIFT	9
120#define	SPA_OLDMAXBLOCKSHIFT	17
121#define	SPA_MAXBLOCKSHIFT	24
122#define	SPA_MINBLOCKSIZE	(1ULL << SPA_MINBLOCKSHIFT)
123#define	SPA_OLDMAXBLOCKSIZE	(1ULL << SPA_OLDMAXBLOCKSHIFT)
124#define	SPA_MAXBLOCKSIZE	(1ULL << SPA_MAXBLOCKSHIFT)
125
126/*
127 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
128 * The ASIZE encoding should be at least 64 times larger (6 more bits)
129 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
130 * overhead, three DVAs per bp, plus one more bit in case we do anything
131 * else that expands the ASIZE.
132 */
133#define	SPA_LSIZEBITS		16	/* LSIZE up to 32M (2^16 * 512)	*/
134#define	SPA_PSIZEBITS		16	/* PSIZE up to 32M (2^16 * 512)	*/
135#define	SPA_ASIZEBITS		24	/* ASIZE up to 64 times larger	*/
136
137/*
138 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
139 * The members of the dva_t should be considered opaque outside the SPA.
140 */
141typedef struct dva {
142	uint64_t	dva_word[2];
143} dva_t;
144
145/*
146 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
147 */
148typedef struct zio_cksum {
149	uint64_t	zc_word[4];
150} zio_cksum_t;
151
152/*
153 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept
154 * secret and is suitable for use in MAC algorithms as the key.
155 */
156typedef struct zio_cksum_salt {
157	uint8_t		zcs_bytes[32];
158} zio_cksum_salt_t;
159
160/*
161 * Each block is described by its DVAs, time of birth, checksum, etc.
162 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
163 *
164 *	64	56	48	40	32	24	16	8	0
165 *	+-------+-------+-------+-------+-------+-------+-------+-------+
166 * 0	|		vdev1		| GRID  |	  ASIZE		|
167 *	+-------+-------+-------+-------+-------+-------+-------+-------+
168 * 1	|G|			 offset1				|
169 *	+-------+-------+-------+-------+-------+-------+-------+-------+
170 * 2	|		vdev2		| GRID  |	  ASIZE		|
171 *	+-------+-------+-------+-------+-------+-------+-------+-------+
172 * 3	|G|			 offset2				|
173 *	+-------+-------+-------+-------+-------+-------+-------+-------+
174 * 4	|		vdev3		| GRID  |	  ASIZE		|
175 *	+-------+-------+-------+-------+-------+-------+-------+-------+
176 * 5	|G|			 offset3				|
177 *	+-------+-------+-------+-------+-------+-------+-------+-------+
178 * 6	|BDX|lvl| type	| cksum |E| comp|    PSIZE	|     LSIZE	|
179 *	+-------+-------+-------+-------+-------+-------+-------+-------+
180 * 7	|			padding					|
181 *	+-------+-------+-------+-------+-------+-------+-------+-------+
182 * 8	|			padding					|
183 *	+-------+-------+-------+-------+-------+-------+-------+-------+
184 * 9	|			physical birth txg			|
185 *	+-------+-------+-------+-------+-------+-------+-------+-------+
186 * a	|			logical birth txg			|
187 *	+-------+-------+-------+-------+-------+-------+-------+-------+
188 * b	|			fill count				|
189 *	+-------+-------+-------+-------+-------+-------+-------+-------+
190 * c	|			checksum[0]				|
191 *	+-------+-------+-------+-------+-------+-------+-------+-------+
192 * d	|			checksum[1]				|
193 *	+-------+-------+-------+-------+-------+-------+-------+-------+
194 * e	|			checksum[2]				|
195 *	+-------+-------+-------+-------+-------+-------+-------+-------+
196 * f	|			checksum[3]				|
197 *	+-------+-------+-------+-------+-------+-------+-------+-------+
198 *
199 * Legend:
200 *
201 * vdev		virtual device ID
202 * offset	offset into virtual device
203 * LSIZE	logical size
204 * PSIZE	physical size (after compression)
205 * ASIZE	allocated size (including RAID-Z parity and gang block headers)
206 * GRID		RAID-Z layout information (reserved for future use)
207 * cksum	checksum function
208 * comp		compression function
209 * G		gang block indicator
210 * B		byteorder (endianness)
211 * D		dedup
212 * X		encryption (on version 30, which is not supported)
213 * E		blkptr_t contains embedded data (see below)
214 * lvl		level of indirection
215 * type		DMU object type
216 * phys birth	txg of block allocation; zero if same as logical birth txg
217 * log. birth	transaction group in which the block was logically born
218 * fill count	number of non-zero blocks under this bp
219 * checksum[4]	256-bit checksum of the data this bp describes
220 */
221
222/*
223 * "Embedded" blkptr_t's don't actually point to a block, instead they
224 * have a data payload embedded in the blkptr_t itself.  See the comment
225 * in blkptr.c for more details.
226 *
227 * The blkptr_t is laid out as follows:
228 *
229 *	64	56	48	40	32	24	16	8	0
230 *	+-------+-------+-------+-------+-------+-------+-------+-------+
231 * 0	|      payload                                                  |
232 * 1	|      payload                                                  |
233 * 2	|      payload                                                  |
234 * 3	|      payload                                                  |
235 * 4	|      payload                                                  |
236 * 5	|      payload                                                  |
237 *	+-------+-------+-------+-------+-------+-------+-------+-------+
238 * 6	|BDX|lvl| type	| etype |E| comp| PSIZE|              LSIZE	|
239 *	+-------+-------+-------+-------+-------+-------+-------+-------+
240 * 7	|      payload                                                  |
241 * 8	|      payload                                                  |
242 * 9	|      payload                                                  |
243 *	+-------+-------+-------+-------+-------+-------+-------+-------+
244 * a	|			logical birth txg			|
245 *	+-------+-------+-------+-------+-------+-------+-------+-------+
246 * b	|      payload                                                  |
247 * c	|      payload                                                  |
248 * d	|      payload                                                  |
249 * e	|      payload                                                  |
250 * f	|      payload                                                  |
251 *	+-------+-------+-------+-------+-------+-------+-------+-------+
252 *
253 * Legend:
254 *
255 * payload		contains the embedded data
256 * B (byteorder)	byteorder (endianness)
257 * D (dedup)		padding (set to zero)
258 * X			encryption (set to zero; see above)
259 * E (embedded)		set to one
260 * lvl			indirection level
261 * type			DMU object type
262 * etype		how to interpret embedded data (BP_EMBEDDED_TYPE_*)
263 * comp			compression function of payload
264 * PSIZE		size of payload after compression, in bytes
265 * LSIZE		logical size of payload, in bytes
266 *			note that 25 bits is enough to store the largest
267 *			"normal" BP's LSIZE (2^16 * 2^9) in bytes
268 * log. birth		transaction group in which the block was logically born
269 *
270 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
271 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
272 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
273 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
274 * BP's so the BP_SET_* macros can be used with them.  etype, PSIZE, LSIZE must
275 * be set with the BPE_SET_* macros.  BP_SET_EMBEDDED() should be called before
276 * other macros, as they assert that they are only used on BP's of the correct
277 * "embedded-ness".
278 */
279
280#define	BPE_GET_ETYPE(bp)	\
281	(ASSERT(BP_IS_EMBEDDED(bp)), \
282	BF64_GET((bp)->blk_prop, 40, 8))
283#define	BPE_SET_ETYPE(bp, t)	do { \
284	ASSERT(BP_IS_EMBEDDED(bp)); \
285	BF64_SET((bp)->blk_prop, 40, 8, t); \
286_NOTE(CONSTCOND) } while (0)
287
288#define	BPE_GET_LSIZE(bp)	\
289	(ASSERT(BP_IS_EMBEDDED(bp)), \
290	BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
291#define	BPE_SET_LSIZE(bp, x)	do { \
292	ASSERT(BP_IS_EMBEDDED(bp)); \
293	BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
294_NOTE(CONSTCOND) } while (0)
295
296#define	BPE_GET_PSIZE(bp)	\
297	(ASSERT(BP_IS_EMBEDDED(bp)), \
298	BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
299#define	BPE_SET_PSIZE(bp, x)	do { \
300	ASSERT(BP_IS_EMBEDDED(bp)); \
301	BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
302_NOTE(CONSTCOND) } while (0)
303
304typedef enum bp_embedded_type {
305	BP_EMBEDDED_TYPE_DATA,
306	BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */
307	NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED
308} bp_embedded_type_t;
309
310#define	BPE_NUM_WORDS 14
311#define	BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
312#define	BPE_IS_PAYLOADWORD(bp, wp) \
313	((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
314
315#define	SPA_BLKPTRSHIFT	7		/* blkptr_t is 128 bytes	*/
316#define	SPA_DVAS_PER_BP	3		/* Number of DVAs in a bp	*/
317
318typedef struct blkptr {
319	dva_t		blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
320	uint64_t	blk_prop;	/* size, compression, type, etc	    */
321	uint64_t	blk_pad[2];	/* Extra space for the future	    */
322	uint64_t	blk_phys_birth;	/* txg when block was allocated	    */
323	uint64_t	blk_birth;	/* transaction group at birth	    */
324	uint64_t	blk_fill;	/* fill count			    */
325	zio_cksum_t	blk_cksum;	/* 256-bit checksum		    */
326} blkptr_t;
327
328/*
329 * Macros to get and set fields in a bp or DVA.
330 */
331#define	DVA_GET_ASIZE(dva)	\
332	BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
333#define	DVA_SET_ASIZE(dva, x)	\
334	BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
335	SPA_MINBLOCKSHIFT, 0, x)
336
337#define	DVA_GET_GRID(dva)	BF64_GET((dva)->dva_word[0], 24, 8)
338#define	DVA_SET_GRID(dva, x)	BF64_SET((dva)->dva_word[0], 24, 8, x)
339
340#define	DVA_GET_VDEV(dva)	BF64_GET((dva)->dva_word[0], 32, 32)
341#define	DVA_SET_VDEV(dva, x)	BF64_SET((dva)->dva_word[0], 32, 32, x)
342
343#define	DVA_GET_OFFSET(dva)	\
344	BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
345#define	DVA_SET_OFFSET(dva, x)	\
346	BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
347
348#define	DVA_GET_GANG(dva)	BF64_GET((dva)->dva_word[1], 63, 1)
349#define	DVA_SET_GANG(dva, x)	BF64_SET((dva)->dva_word[1], 63, 1, x)
350
351#define	BP_GET_LSIZE(bp)	\
352	(BP_IS_EMBEDDED(bp) ?	\
353	(BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
354	BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
355#define	BP_SET_LSIZE(bp, x)	do { \
356	ASSERT(!BP_IS_EMBEDDED(bp)); \
357	BF64_SET_SB((bp)->blk_prop, \
358	    0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
359_NOTE(CONSTCOND) } while (0)
360
361#define	BP_GET_PSIZE(bp)	\
362	BF64_GET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)
363#define	BP_SET_PSIZE(bp, x)	\
364	BF64_SET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x)
365
366#define	BP_GET_COMPRESS(bp)	BF64_GET((bp)->blk_prop, 32, 7)
367#define	BP_SET_COMPRESS(bp, x)	BF64_SET((bp)->blk_prop, 32, 7, x)
368
369#define	BP_GET_CHECKSUM(bp)	BF64_GET((bp)->blk_prop, 40, 8)
370#define	BP_SET_CHECKSUM(bp, x)	BF64_SET((bp)->blk_prop, 40, 8, x)
371
372#define	BP_GET_TYPE(bp)		BF64_GET((bp)->blk_prop, 48, 8)
373#define	BP_SET_TYPE(bp, x)	BF64_SET((bp)->blk_prop, 48, 8, x)
374
375#define	BP_GET_LEVEL(bp)	BF64_GET((bp)->blk_prop, 56, 5)
376#define	BP_SET_LEVEL(bp, x)	BF64_SET((bp)->blk_prop, 56, 5, x)
377
378#define	BP_IS_EMBEDDED(bp)	BF64_GET((bp)->blk_prop, 39, 1)
379
380#define	BP_GET_DEDUP(bp)	BF64_GET((bp)->blk_prop, 62, 1)
381#define	BP_SET_DEDUP(bp, x)	BF64_SET((bp)->blk_prop, 62, 1, x)
382
383#define	BP_GET_BYTEORDER(bp)	BF64_GET((bp)->blk_prop, 63, 1)
384#define	BP_SET_BYTEORDER(bp, x)	BF64_SET((bp)->blk_prop, 63, 1, x)
385
386#define	BP_PHYSICAL_BIRTH(bp)		\
387	((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
388
389#define	BP_GET_ASIZE(bp)	\
390	(DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
391		DVA_GET_ASIZE(&(bp)->blk_dva[2]))
392
393#define	BP_GET_UCSIZE(bp) \
394	((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \
395	BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp));
396
397#define	BP_GET_NDVAS(bp)	\
398	(!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
399	!!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
400	!!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
401
402#define	DVA_EQUAL(dva1, dva2)	\
403	((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
404	(dva1)->dva_word[0] == (dva2)->dva_word[0])
405
406#define	ZIO_CHECKSUM_EQUAL(zc1, zc2) \
407	(0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
408	((zc1).zc_word[1] - (zc2).zc_word[1]) | \
409	((zc1).zc_word[2] - (zc2).zc_word[2]) | \
410	((zc1).zc_word[3] - (zc2).zc_word[3])))
411
412
413#define	DVA_IS_VALID(dva)	(DVA_GET_ASIZE(dva) != 0)
414
415#define	ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3)	\
416{						\
417	(zcp)->zc_word[0] = w0;			\
418	(zcp)->zc_word[1] = w1;			\
419	(zcp)->zc_word[2] = w2;			\
420	(zcp)->zc_word[3] = w3;			\
421}
422
423#define	BP_IDENTITY(bp)		(&(bp)->blk_dva[0])
424#define	BP_IS_GANG(bp)		DVA_GET_GANG(BP_IDENTITY(bp))
425#define	DVA_IS_EMPTY(dva)	((dva)->dva_word[0] == 0ULL &&  \
426	(dva)->dva_word[1] == 0ULL)
427#define	BP_IS_HOLE(bp)		DVA_IS_EMPTY(BP_IDENTITY(bp))
428#define	BP_IS_OLDER(bp, txg)	(!BP_IS_HOLE(bp) && (bp)->blk_birth < (txg))
429
430#define	BP_ZERO(bp)				\
431{						\
432	(bp)->blk_dva[0].dva_word[0] = 0;	\
433	(bp)->blk_dva[0].dva_word[1] = 0;	\
434	(bp)->blk_dva[1].dva_word[0] = 0;	\
435	(bp)->blk_dva[1].dva_word[1] = 0;	\
436	(bp)->blk_dva[2].dva_word[0] = 0;	\
437	(bp)->blk_dva[2].dva_word[1] = 0;	\
438	(bp)->blk_prop = 0;			\
439	(bp)->blk_pad[0] = 0;			\
440	(bp)->blk_pad[1] = 0;			\
441	(bp)->blk_phys_birth = 0;		\
442	(bp)->blk_birth = 0;			\
443	(bp)->blk_fill = 0;			\
444	ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0);	\
445}
446
447#define	BPE_NUM_WORDS 14
448#define	BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
449#define	BPE_IS_PAYLOADWORD(bp, wp) \
450	((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
451
452/*
453 * Embedded checksum
454 */
455#define	ZEC_MAGIC	0x210da7ab10c7a11ULL
456
457typedef struct zio_eck {
458	uint64_t	zec_magic;	/* for validation, endianness	*/
459	zio_cksum_t	zec_cksum;	/* 256-bit checksum		*/
460} zio_eck_t;
461
462/*
463 * Gang block headers are self-checksumming and contain an array
464 * of block pointers.
465 */
466#define	SPA_GANGBLOCKSIZE	SPA_MINBLOCKSIZE
467#define	SPA_GBH_NBLKPTRS	((SPA_GANGBLOCKSIZE - \
468	sizeof (zio_eck_t)) / sizeof (blkptr_t))
469#define	SPA_GBH_FILLER		((SPA_GANGBLOCKSIZE - \
470	sizeof (zio_eck_t) - \
471	(SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\
472	sizeof (uint64_t))
473
474typedef struct zio_gbh {
475	blkptr_t		zg_blkptr[SPA_GBH_NBLKPTRS];
476	uint64_t		zg_filler[SPA_GBH_FILLER];
477	zio_eck_t		zg_tail;
478} zio_gbh_phys_t;
479
480#define	VDEV_RAIDZ_MAXPARITY	3
481
482#define	VDEV_PAD_SIZE		(8 << 10)
483/* 2 padding areas (vl_pad1 and vl_pad2) to skip */
484#define	VDEV_SKIP_SIZE		VDEV_PAD_SIZE * 2
485#define	VDEV_PHYS_SIZE		(112 << 10)
486#define	VDEV_UBERBLOCK_RING	(128 << 10)
487
488#define	VDEV_UBERBLOCK_SHIFT(vd)	\
489	MAX((vd)->v_top->v_ashift, UBERBLOCK_SHIFT)
490#define	VDEV_UBERBLOCK_COUNT(vd)	\
491	(VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
492#define	VDEV_UBERBLOCK_OFFSET(vd, n)	\
493	offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
494#define	VDEV_UBERBLOCK_SIZE(vd)		(1ULL << VDEV_UBERBLOCK_SHIFT(vd))
495
496typedef struct vdev_phys {
497	char		vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)];
498	zio_eck_t	vp_zbt;
499} vdev_phys_t;
500
501typedef struct vdev_label {
502	char		vl_pad1[VDEV_PAD_SIZE];			/*  8K  */
503	char		vl_pad2[VDEV_PAD_SIZE];			/*  8K  */
504	vdev_phys_t	vl_vdev_phys;				/* 112K	*/
505	char		vl_uberblock[VDEV_UBERBLOCK_RING];	/* 128K	*/
506} vdev_label_t;							/* 256K total */
507
508/*
509 * vdev_dirty() flags
510 */
511#define	VDD_METASLAB	0x01
512#define	VDD_DTL		0x02
513
514/*
515 * Size and offset of embedded boot loader region on each label.
516 * The total size of the first two labels plus the boot area is 4MB.
517 */
518#define	VDEV_BOOT_OFFSET	(2 * sizeof (vdev_label_t))
519#define	VDEV_BOOT_SIZE		(7ULL << 19)			/* 3.5M	*/
520
521/*
522 * Size of label regions at the start and end of each leaf device.
523 */
524#define	VDEV_LABEL_START_SIZE	(2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
525#define	VDEV_LABEL_END_SIZE	(2 * sizeof (vdev_label_t))
526#define	VDEV_LABELS		4
527
528enum zio_checksum {
529	ZIO_CHECKSUM_INHERIT = 0,
530	ZIO_CHECKSUM_ON,
531	ZIO_CHECKSUM_OFF,
532	ZIO_CHECKSUM_LABEL,
533	ZIO_CHECKSUM_GANG_HEADER,
534	ZIO_CHECKSUM_ZILOG,
535	ZIO_CHECKSUM_FLETCHER_2,
536	ZIO_CHECKSUM_FLETCHER_4,
537	ZIO_CHECKSUM_SHA256,
538	ZIO_CHECKSUM_ZILOG2,
539	ZIO_CHECKSUM_NOPARITY,
540	ZIO_CHECKSUM_SHA512,
541	ZIO_CHECKSUM_SKEIN,
542	ZIO_CHECKSUM_EDONR,
543	ZIO_CHECKSUM_FUNCTIONS
544};
545
546#define	ZIO_CHECKSUM_ON_VALUE	ZIO_CHECKSUM_FLETCHER_4
547#define	ZIO_CHECKSUM_DEFAULT	ZIO_CHECKSUM_ON
548
549enum zio_compress {
550	ZIO_COMPRESS_INHERIT = 0,
551	ZIO_COMPRESS_ON,
552	ZIO_COMPRESS_OFF,
553	ZIO_COMPRESS_LZJB,
554	ZIO_COMPRESS_EMPTY,
555	ZIO_COMPRESS_GZIP_1,
556	ZIO_COMPRESS_GZIP_2,
557	ZIO_COMPRESS_GZIP_3,
558	ZIO_COMPRESS_GZIP_4,
559	ZIO_COMPRESS_GZIP_5,
560	ZIO_COMPRESS_GZIP_6,
561	ZIO_COMPRESS_GZIP_7,
562	ZIO_COMPRESS_GZIP_8,
563	ZIO_COMPRESS_GZIP_9,
564	ZIO_COMPRESS_ZLE,
565	ZIO_COMPRESS_LZ4,
566	ZIO_COMPRESS_FUNCTIONS
567};
568
569#define	ZIO_COMPRESS_ON_VALUE	ZIO_COMPRESS_LZJB
570#define	ZIO_COMPRESS_DEFAULT	ZIO_COMPRESS_OFF
571
572/* nvlist pack encoding */
573#define	NV_ENCODE_NATIVE	0
574#define	NV_ENCODE_XDR		1
575
576typedef enum {
577	DATA_TYPE_UNKNOWN = 0,
578	DATA_TYPE_BOOLEAN,
579	DATA_TYPE_BYTE,
580	DATA_TYPE_INT16,
581	DATA_TYPE_UINT16,
582	DATA_TYPE_INT32,
583	DATA_TYPE_UINT32,
584	DATA_TYPE_INT64,
585	DATA_TYPE_UINT64,
586	DATA_TYPE_STRING,
587	DATA_TYPE_BYTE_ARRAY,
588	DATA_TYPE_INT16_ARRAY,
589	DATA_TYPE_UINT16_ARRAY,
590	DATA_TYPE_INT32_ARRAY,
591	DATA_TYPE_UINT32_ARRAY,
592	DATA_TYPE_INT64_ARRAY,
593	DATA_TYPE_UINT64_ARRAY,
594	DATA_TYPE_STRING_ARRAY,
595	DATA_TYPE_HRTIME,
596	DATA_TYPE_NVLIST,
597	DATA_TYPE_NVLIST_ARRAY,
598	DATA_TYPE_BOOLEAN_VALUE,
599	DATA_TYPE_INT8,
600	DATA_TYPE_UINT8,
601	DATA_TYPE_BOOLEAN_ARRAY,
602	DATA_TYPE_INT8_ARRAY,
603	DATA_TYPE_UINT8_ARRAY
604} data_type_t;
605
606/*
607 * On-disk version number.
608 */
609#define	SPA_VERSION_1			1ULL
610#define	SPA_VERSION_2			2ULL
611#define	SPA_VERSION_3			3ULL
612#define	SPA_VERSION_4			4ULL
613#define	SPA_VERSION_5			5ULL
614#define	SPA_VERSION_6			6ULL
615#define	SPA_VERSION_7			7ULL
616#define	SPA_VERSION_8			8ULL
617#define	SPA_VERSION_9			9ULL
618#define	SPA_VERSION_10			10ULL
619#define	SPA_VERSION_11			11ULL
620#define	SPA_VERSION_12			12ULL
621#define	SPA_VERSION_13			13ULL
622#define	SPA_VERSION_14			14ULL
623#define	SPA_VERSION_15			15ULL
624#define	SPA_VERSION_16			16ULL
625#define	SPA_VERSION_17			17ULL
626#define	SPA_VERSION_18			18ULL
627#define	SPA_VERSION_19			19ULL
628#define	SPA_VERSION_20			20ULL
629#define	SPA_VERSION_21			21ULL
630#define	SPA_VERSION_22			22ULL
631#define	SPA_VERSION_23			23ULL
632#define	SPA_VERSION_24			24ULL
633#define	SPA_VERSION_25			25ULL
634#define	SPA_VERSION_26			26ULL
635#define	SPA_VERSION_27			27ULL
636#define	SPA_VERSION_28			28ULL
637#define	SPA_VERSION_5000		5000ULL
638
639/*
640 * When bumping up SPA_VERSION, make sure GRUB ZFS understands the on-disk
641 * format change. Go to usr/src/grub/grub-0.97/stage2/{zfs-include/, fsys_zfs*},
642 * and do the appropriate changes.  Also bump the version number in
643 * usr/src/grub/capability.
644 */
645#define	SPA_VERSION			SPA_VERSION_5000
646#define	SPA_VERSION_STRING		"5000"
647
648/*
649 * Symbolic names for the changes that caused a SPA_VERSION switch.
650 * Used in the code when checking for presence or absence of a feature.
651 * Feel free to define multiple symbolic names for each version if there
652 * were multiple changes to on-disk structures during that version.
653 *
654 * NOTE: When checking the current SPA_VERSION in your code, be sure
655 *       to use spa_version() since it reports the version of the
656 *       last synced uberblock.  Checking the in-flight version can
657 *       be dangerous in some cases.
658 */
659#define	SPA_VERSION_INITIAL		SPA_VERSION_1
660#define	SPA_VERSION_DITTO_BLOCKS	SPA_VERSION_2
661#define	SPA_VERSION_SPARES		SPA_VERSION_3
662#define	SPA_VERSION_RAID6		SPA_VERSION_3
663#define	SPA_VERSION_BPLIST_ACCOUNT	SPA_VERSION_3
664#define	SPA_VERSION_RAIDZ_DEFLATE	SPA_VERSION_3
665#define	SPA_VERSION_DNODE_BYTES		SPA_VERSION_3
666#define	SPA_VERSION_ZPOOL_HISTORY	SPA_VERSION_4
667#define	SPA_VERSION_GZIP_COMPRESSION	SPA_VERSION_5
668#define	SPA_VERSION_BOOTFS		SPA_VERSION_6
669#define	SPA_VERSION_SLOGS		SPA_VERSION_7
670#define	SPA_VERSION_DELEGATED_PERMS	SPA_VERSION_8
671#define	SPA_VERSION_FUID		SPA_VERSION_9
672#define	SPA_VERSION_REFRESERVATION	SPA_VERSION_9
673#define	SPA_VERSION_REFQUOTA		SPA_VERSION_9
674#define	SPA_VERSION_UNIQUE_ACCURATE	SPA_VERSION_9
675#define	SPA_VERSION_L2CACHE		SPA_VERSION_10
676#define	SPA_VERSION_NEXT_CLONES		SPA_VERSION_11
677#define	SPA_VERSION_ORIGIN		SPA_VERSION_11
678#define	SPA_VERSION_DSL_SCRUB		SPA_VERSION_11
679#define	SPA_VERSION_SNAP_PROPS		SPA_VERSION_12
680#define	SPA_VERSION_USED_BREAKDOWN	SPA_VERSION_13
681#define	SPA_VERSION_PASSTHROUGH_X	SPA_VERSION_14
682#define	SPA_VERSION_USERSPACE		SPA_VERSION_15
683#define	SPA_VERSION_STMF_PROP		SPA_VERSION_16
684#define	SPA_VERSION_RAIDZ3		SPA_VERSION_17
685#define	SPA_VERSION_USERREFS		SPA_VERSION_18
686#define	SPA_VERSION_HOLES		SPA_VERSION_19
687#define	SPA_VERSION_ZLE_COMPRESSION	SPA_VERSION_20
688#define	SPA_VERSION_DEDUP		SPA_VERSION_21
689#define	SPA_VERSION_RECVD_PROPS		SPA_VERSION_22
690#define	SPA_VERSION_SLIM_ZIL		SPA_VERSION_23
691#define	SPA_VERSION_SA			SPA_VERSION_24
692#define	SPA_VERSION_SCAN		SPA_VERSION_25
693#define	SPA_VERSION_DIR_CLONES		SPA_VERSION_26
694#define	SPA_VERSION_DEADLISTS		SPA_VERSION_26
695#define	SPA_VERSION_FAST_SNAP		SPA_VERSION_27
696#define	SPA_VERSION_MULTI_REPLACE	SPA_VERSION_28
697#define	SPA_VERSION_BEFORE_FEATURES	SPA_VERSION_28
698#define	SPA_VERSION_FEATURES		SPA_VERSION_5000
699
700#define	SPA_VERSION_IS_SUPPORTED(v) \
701	(((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \
702	((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION))
703
704/*
705 * The following are configuration names used in the nvlist describing a pool's
706 * configuration.
707 */
708#define	ZPOOL_CONFIG_VERSION		"version"
709#define	ZPOOL_CONFIG_POOL_NAME		"name"
710#define	ZPOOL_CONFIG_POOL_STATE		"state"
711#define	ZPOOL_CONFIG_POOL_TXG		"txg"
712#define	ZPOOL_CONFIG_POOL_GUID		"pool_guid"
713#define	ZPOOL_CONFIG_CREATE_TXG		"create_txg"
714#define	ZPOOL_CONFIG_TOP_GUID		"top_guid"
715#define	ZPOOL_CONFIG_VDEV_TREE		"vdev_tree"
716#define	ZPOOL_CONFIG_TYPE		"type"
717#define	ZPOOL_CONFIG_CHILDREN		"children"
718#define	ZPOOL_CONFIG_ID			"id"
719#define	ZPOOL_CONFIG_GUID		"guid"
720#define	ZPOOL_CONFIG_INDIRECT_OBJECT	"com.delphix:indirect_object"
721#define	ZPOOL_CONFIG_INDIRECT_BIRTHS	"com.delphix:indirect_births"
722#define	ZPOOL_CONFIG_PREV_INDIRECT_VDEV	"com.delphix:prev_indirect_vdev"
723#define	ZPOOL_CONFIG_PATH		"path"
724#define	ZPOOL_CONFIG_DEVID		"devid"
725#define	ZPOOL_CONFIG_PHYS_PATH		"phys_path"
726#define	ZPOOL_CONFIG_METASLAB_ARRAY	"metaslab_array"
727#define	ZPOOL_CONFIG_METASLAB_SHIFT	"metaslab_shift"
728#define	ZPOOL_CONFIG_ASHIFT		"ashift"
729#define	ZPOOL_CONFIG_ASIZE		"asize"
730#define	ZPOOL_CONFIG_DTL		"DTL"
731#define	ZPOOL_CONFIG_STATS		"stats"
732#define	ZPOOL_CONFIG_WHOLE_DISK		"whole_disk"
733#define	ZPOOL_CONFIG_ERRCOUNT		"error_count"
734#define	ZPOOL_CONFIG_NOT_PRESENT	"not_present"
735#define	ZPOOL_CONFIG_SPARES		"spares"
736#define	ZPOOL_CONFIG_IS_SPARE		"is_spare"
737#define	ZPOOL_CONFIG_NPARITY		"nparity"
738#define	ZPOOL_CONFIG_HOSTID		"hostid"
739#define	ZPOOL_CONFIG_HOSTNAME		"hostname"
740#define	ZPOOL_CONFIG_IS_LOG		"is_log"
741#define	ZPOOL_CONFIG_TIMESTAMP		"timestamp" /* not stored on disk */
742#define	ZPOOL_CONFIG_FEATURES_FOR_READ	"features_for_read"
743
744/*
745 * The persistent vdev state is stored as separate values rather than a single
746 * 'vdev_state' entry.  This is because a device can be in multiple states, such
747 * as offline and degraded.
748 */
749#define	ZPOOL_CONFIG_OFFLINE		"offline"
750#define	ZPOOL_CONFIG_FAULTED		"faulted"
751#define	ZPOOL_CONFIG_DEGRADED		"degraded"
752#define	ZPOOL_CONFIG_REMOVED		"removed"
753#define	ZPOOL_CONFIG_FRU		"fru"
754#define	ZPOOL_CONFIG_AUX_STATE		"aux_state"
755
756#define	VDEV_TYPE_ROOT			"root"
757#define	VDEV_TYPE_MIRROR		"mirror"
758#define	VDEV_TYPE_REPLACING		"replacing"
759#define	VDEV_TYPE_RAIDZ			"raidz"
760#define	VDEV_TYPE_DISK			"disk"
761#define	VDEV_TYPE_FILE			"file"
762#define	VDEV_TYPE_MISSING		"missing"
763#define	VDEV_TYPE_HOLE			"hole"
764#define	VDEV_TYPE_SPARE			"spare"
765#define	VDEV_TYPE_LOG			"log"
766#define	VDEV_TYPE_L2CACHE		"l2cache"
767#define	VDEV_TYPE_INDIRECT		"indirect"
768
769/*
770 * This is needed in userland to report the minimum necessary device size.
771 */
772#define	SPA_MINDEVSIZE		(64ULL << 20)
773
774/*
775 * The location of the pool configuration repository, shared between kernel and
776 * userland.
777 */
778#define	ZPOOL_CACHE		"/boot/zfs/zpool.cache"
779
780/*
781 * vdev states are ordered from least to most healthy.
782 * A vdev that's CANT_OPEN or below is considered unusable.
783 */
784typedef enum vdev_state {
785	VDEV_STATE_UNKNOWN = 0,	/* Uninitialized vdev			*/
786	VDEV_STATE_CLOSED,	/* Not currently open			*/
787	VDEV_STATE_OFFLINE,	/* Not allowed to open			*/
788	VDEV_STATE_REMOVED,	/* Explicitly removed from system	*/
789	VDEV_STATE_CANT_OPEN,	/* Tried to open, but failed		*/
790	VDEV_STATE_FAULTED,	/* External request to fault device	*/
791	VDEV_STATE_DEGRADED,	/* Replicated vdev with unhealthy kids	*/
792	VDEV_STATE_HEALTHY	/* Presumed good			*/
793} vdev_state_t;
794
795/*
796 * vdev aux states.  When a vdev is in the CANT_OPEN state, the aux field
797 * of the vdev stats structure uses these constants to distinguish why.
798 */
799typedef enum vdev_aux {
800	VDEV_AUX_NONE,		/* no error				*/
801	VDEV_AUX_OPEN_FAILED,	/* ldi_open_*() or vn_open() failed	*/
802	VDEV_AUX_CORRUPT_DATA,	/* bad label or disk contents		*/
803	VDEV_AUX_NO_REPLICAS,	/* insufficient number of replicas	*/
804	VDEV_AUX_BAD_GUID_SUM,	/* vdev guid sum doesn't match		*/
805	VDEV_AUX_TOO_SMALL,	/* vdev size is too small		*/
806	VDEV_AUX_BAD_LABEL,	/* the label is OK but invalid		*/
807	VDEV_AUX_VERSION_NEWER,	/* on-disk version is too new		*/
808	VDEV_AUX_VERSION_OLDER,	/* on-disk version is too old		*/
809	VDEV_AUX_SPARED		/* hot spare used in another pool	*/
810} vdev_aux_t;
811
812/*
813 * pool state.  The following states are written to disk as part of the normal
814 * SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE.  The remaining states are
815 * software abstractions used at various levels to communicate pool state.
816 */
817typedef enum pool_state {
818	POOL_STATE_ACTIVE = 0,		/* In active use		*/
819	POOL_STATE_EXPORTED,		/* Explicitly exported		*/
820	POOL_STATE_DESTROYED,		/* Explicitly destroyed		*/
821	POOL_STATE_SPARE,		/* Reserved for hot spare use	*/
822	POOL_STATE_UNINITIALIZED,	/* Internal spa_t state		*/
823	POOL_STATE_UNAVAIL,		/* Internal libzfs state	*/
824	POOL_STATE_POTENTIALLY_ACTIVE	/* Internal libzfs state	*/
825} pool_state_t;
826
827/*
828 * The uberblock version is incremented whenever an incompatible on-disk
829 * format change is made to the SPA, DMU, or ZAP.
830 *
831 * Note: the first two fields should never be moved.  When a storage pool
832 * is opened, the uberblock must be read off the disk before the version
833 * can be checked.  If the ub_version field is moved, we may not detect
834 * version mismatch.  If the ub_magic field is moved, applications that
835 * expect the magic number in the first word won't work.
836 */
837#define	UBERBLOCK_MAGIC		0x00bab10c		/* oo-ba-bloc!	*/
838#define	UBERBLOCK_SHIFT		10			/* up to 1K	*/
839
840struct uberblock {
841	uint64_t	ub_magic;	/* UBERBLOCK_MAGIC		*/
842	uint64_t	ub_version;	/* SPA_VERSION			*/
843	uint64_t	ub_txg;		/* txg of last sync		*/
844	uint64_t	ub_guid_sum;	/* sum of all vdev guids	*/
845	uint64_t	ub_timestamp;	/* UTC time of last sync	*/
846	blkptr_t	ub_rootbp;	/* MOS objset_phys_t		*/
847};
848
849/*
850 * Flags.
851 */
852#define	DNODE_MUST_BE_ALLOCATED	1
853#define	DNODE_MUST_BE_FREE	2
854
855/*
856 * Fixed constants.
857 */
858#define	DNODE_SHIFT		9	/* 512 bytes */
859#define	DN_MIN_INDBLKSHIFT	12	/* 4k */
860#define	DN_MAX_INDBLKSHIFT	17	/* 128k */
861#define	DNODE_BLOCK_SHIFT	14	/* 16k */
862#define	DNODE_CORE_SIZE		64	/* 64 bytes for dnode sans blkptrs */
863#define	DN_MAX_OBJECT_SHIFT	48	/* 256 trillion (zfs_fid_t limit) */
864#define	DN_MAX_OFFSET_SHIFT	64	/* 2^64 bytes in a dnode */
865
866/*
867 * Derived constants.
868 */
869#define	DNODE_MIN_SIZE		(1 << DNODE_SHIFT)
870#define	DNODE_MAX_SIZE		(1 << DNODE_BLOCK_SHIFT)
871#define	DNODE_BLOCK_SIZE	(1 << DNODE_BLOCK_SHIFT)
872#define	DNODE_MIN_SLOTS		(DNODE_MIN_SIZE >> DNODE_SHIFT)
873#define	DNODE_MAX_SLOTS		(DNODE_MAX_SIZE >> DNODE_SHIFT)
874#define	DN_BONUS_SIZE(dnsize)	((dnsize) - DNODE_CORE_SIZE - \
875	(1 << SPA_BLKPTRSHIFT))
876#define	DN_SLOTS_TO_BONUSLEN(slots)	DN_BONUS_SIZE((slots) << DNODE_SHIFT)
877#define	DN_OLD_MAX_BONUSLEN		(DN_BONUS_SIZE(DNODE_MIN_SIZE))
878#define	DN_MAX_NBLKPTR		((DNODE_MIN_SIZE - DNODE_CORE_SIZE) >> \
879	SPA_BLKPTRSHIFT)
880#define	DN_MAX_OBJECT		(1ULL << DN_MAX_OBJECT_SHIFT)
881#define	DN_ZERO_BONUSLEN	(DN_BONUS_SIZE(DNODE_MAX_SIZE) + 1)
882
883#define	DNODES_PER_BLOCK_SHIFT	(DNODE_BLOCK_SHIFT - DNODE_SHIFT)
884#define	DNODES_PER_BLOCK	(1ULL << DNODES_PER_BLOCK_SHIFT)
885#define	DNODES_PER_LEVEL_SHIFT	(DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT)
886
887/* The +2 here is a cheesy way to round up */
888#define	DN_MAX_LEVELS	(2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \
889	(DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT)))
890
891#define	DN_BONUS(dnp)	((void*)((dnp)->dn_bonus + \
892	(((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t))))
893
894#define	DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \
895	(dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT)
896
897#define	EPB(blkshift, typeshift)	(1 << (blkshift - typeshift))
898
899/* Is dn_used in bytes?  if not, it's in multiples of SPA_MINBLOCKSIZE */
900#define	DNODE_FLAG_USED_BYTES		(1<<0)
901#define	DNODE_FLAG_USERUSED_ACCOUNTED	(1<<1)
902
903/* Does dnode have a SA spill blkptr in bonus? */
904#define	DNODE_FLAG_SPILL_BLKPTR	(1<<2)
905
906typedef struct dnode_phys {
907	uint8_t dn_type;		/* dmu_object_type_t */
908	uint8_t dn_indblkshift;		/* ln2(indirect block size) */
909	uint8_t dn_nlevels;		/* 1=dn_blkptr->data blocks */
910	uint8_t dn_nblkptr;		/* length of dn_blkptr */
911	uint8_t dn_bonustype;		/* type of data in bonus buffer */
912	uint8_t	dn_checksum;		/* ZIO_CHECKSUM type */
913	uint8_t	dn_compress;		/* ZIO_COMPRESS type */
914	uint8_t dn_flags;		/* DNODE_FLAG_* */
915	uint16_t dn_datablkszsec;	/* data block size in 512b sectors */
916	uint16_t dn_bonuslen;		/* length of dn_bonus */
917	uint8_t dn_extra_slots;		/* # of subsequent slots consumed */
918	uint8_t dn_pad2[3];
919
920	/* accounting is protected by dn_dirty_mtx */
921	uint64_t dn_maxblkid;		/* largest allocated block ID */
922	uint64_t dn_used;		/* bytes (or sectors) of disk space */
923
924	uint64_t dn_pad3[4];
925
926	/*
927	 * The tail region is 448 bytes for a 512 byte dnode, and
928	 * correspondingly larger for larger dnode sizes. The spill
929	 * block pointer, when present, is always at the end of the tail
930	 * region. There are three ways this space may be used, using
931	 * a 512 byte dnode for this diagram:
932	 *
933	 * 0       64      128     192     256     320     384     448 (offset)
934	 * +---------------+---------------+---------------+-------+
935	 * | dn_blkptr[0]  | dn_blkptr[1]  | dn_blkptr[2]  | /     |
936	 * +---------------+---------------+---------------+-------+
937	 * | dn_blkptr[0]  | dn_bonus[0..319]                      |
938	 * +---------------+-----------------------+---------------+
939	 * | dn_blkptr[0]  | dn_bonus[0..191]      | dn_spill      |
940	 * +---------------+-----------------------+---------------+
941	 */
942	union {
943		blkptr_t dn_blkptr[1+DN_OLD_MAX_BONUSLEN/sizeof (blkptr_t)];
944		struct {
945			blkptr_t __dn_ignore1;
946			uint8_t dn_bonus[DN_OLD_MAX_BONUSLEN];
947		};
948		struct {
949			blkptr_t __dn_ignore2;
950			uint8_t __dn_ignore3[DN_OLD_MAX_BONUSLEN -
951			    sizeof (blkptr_t)];
952			blkptr_t dn_spill;
953		};
954	};
955} dnode_phys_t;
956
957#define	DN_SPILL_BLKPTR(dnp)	(blkptr_t *)((char *)(dnp) + \
958	(((dnp)->dn_extra_slots + 1) << DNODE_SHIFT) - (1 << SPA_BLKPTRSHIFT))
959
960typedef enum dmu_object_byteswap {
961	DMU_BSWAP_UINT8,
962	DMU_BSWAP_UINT16,
963	DMU_BSWAP_UINT32,
964	DMU_BSWAP_UINT64,
965	DMU_BSWAP_ZAP,
966	DMU_BSWAP_DNODE,
967	DMU_BSWAP_OBJSET,
968	DMU_BSWAP_ZNODE,
969	DMU_BSWAP_OLDACL,
970	DMU_BSWAP_ACL,
971	/*
972	 * Allocating a new byteswap type number makes the on-disk format
973	 * incompatible with any other format that uses the same number.
974	 *
975	 * Data can usually be structured to work with one of the
976	 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types.
977	 */
978	DMU_BSWAP_NUMFUNCS
979} dmu_object_byteswap_t;
980
981#define	DMU_OT_NEWTYPE 0x80
982#define	DMU_OT_METADATA 0x40
983#define	DMU_OT_BYTESWAP_MASK 0x3f
984
985/*
986 * Defines a uint8_t object type. Object types specify if the data
987 * in the object is metadata (boolean) and how to byteswap the data
988 * (dmu_object_byteswap_t).
989 */
990#define	DMU_OT(byteswap, metadata) \
991	(DMU_OT_NEWTYPE | \
992	((metadata) ? DMU_OT_METADATA : 0) | \
993	((byteswap) & DMU_OT_BYTESWAP_MASK))
994
995typedef enum dmu_object_type {
996	DMU_OT_NONE,
997	/* general: */
998	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
999	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
1000	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
1001	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
1002	DMU_OT_BPLIST,			/* UINT64 */
1003	DMU_OT_BPLIST_HDR,		/* UINT64 */
1004	/* spa: */
1005	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
1006	DMU_OT_SPACE_MAP,		/* UINT64 */
1007	/* zil: */
1008	DMU_OT_INTENT_LOG,		/* UINT64 */
1009	/* dmu: */
1010	DMU_OT_DNODE,			/* DNODE */
1011	DMU_OT_OBJSET,			/* OBJSET */
1012	/* dsl: */
1013	DMU_OT_DSL_DIR,			/* UINT64 */
1014	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
1015	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
1016	DMU_OT_DSL_PROPS,		/* ZAP */
1017	DMU_OT_DSL_DATASET,		/* UINT64 */
1018	/* zpl: */
1019	DMU_OT_ZNODE,			/* ZNODE */
1020	DMU_OT_OLDACL,			/* Old ACL */
1021	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
1022	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
1023	DMU_OT_MASTER_NODE,		/* ZAP */
1024	DMU_OT_UNLINKED_SET,		/* ZAP */
1025	/* zvol: */
1026	DMU_OT_ZVOL,			/* UINT8 */
1027	DMU_OT_ZVOL_PROP,		/* ZAP */
1028	/* other; for testing only! */
1029	DMU_OT_PLAIN_OTHER,		/* UINT8 */
1030	DMU_OT_UINT64_OTHER,		/* UINT64 */
1031	DMU_OT_ZAP_OTHER,		/* ZAP */
1032	/* new object types: */
1033	DMU_OT_ERROR_LOG,		/* ZAP */
1034	DMU_OT_SPA_HISTORY,		/* UINT8 */
1035	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
1036	DMU_OT_POOL_PROPS,		/* ZAP */
1037	DMU_OT_DSL_PERMS,		/* ZAP */
1038	DMU_OT_ACL,			/* ACL */
1039	DMU_OT_SYSACL,			/* SYSACL */
1040	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
1041	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
1042	DMU_OT_NEXT_CLONES,		/* ZAP */
1043	DMU_OT_SCAN_QUEUE,		/* ZAP */
1044	DMU_OT_USERGROUP_USED,		/* ZAP */
1045	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
1046	DMU_OT_USERREFS,		/* ZAP */
1047	DMU_OT_DDT_ZAP,			/* ZAP */
1048	DMU_OT_DDT_STATS,		/* ZAP */
1049	DMU_OT_SA,			/* System attr */
1050	DMU_OT_SA_MASTER_NODE,		/* ZAP */
1051	DMU_OT_SA_ATTR_REGISTRATION,	/* ZAP */
1052	DMU_OT_SA_ATTR_LAYOUTS,		/* ZAP */
1053	DMU_OT_SCAN_XLATE,		/* ZAP */
1054	DMU_OT_DEDUP,			/* fake dedup BP from ddt_bp_create() */
1055	DMU_OT_NUMTYPES,
1056
1057	/*
1058	 * Names for valid types declared with DMU_OT().
1059	 */
1060	DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE),
1061	DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE),
1062	DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE),
1063	DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE),
1064	DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE),
1065	DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE),
1066	DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE),
1067	DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE),
1068	DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE),
1069	DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE)
1070} dmu_object_type_t;
1071
1072typedef enum dmu_objset_type {
1073	DMU_OST_NONE,
1074	DMU_OST_META,
1075	DMU_OST_ZFS,
1076	DMU_OST_ZVOL,
1077	DMU_OST_OTHER,			/* For testing only! */
1078	DMU_OST_ANY,			/* Be careful! */
1079	DMU_OST_NUMTYPES
1080} dmu_objset_type_t;
1081
1082#define	ZAP_MAXVALUELEN	(1024 * 8)
1083
1084/*
1085 * header for all bonus and spill buffers.
1086 * The header has a fixed portion with a variable number
1087 * of "lengths" depending on the number of variable sized
1088 * attribues which are determined by the "layout number"
1089 */
1090
1091#define	SA_MAGIC	0x2F505A  /* ZFS SA */
1092typedef struct sa_hdr_phys {
1093	uint32_t sa_magic;
1094	uint16_t sa_layout_info;  /* Encoded with hdrsize and layout number */
1095	uint16_t sa_lengths[1];	/* optional sizes for variable length attrs */
1096	/* ... Data follows the lengths.  */
1097} sa_hdr_phys_t;
1098
1099/*
1100 * sa_hdr_phys -> sa_layout_info
1101 *
1102 * 16      10       0
1103 * +--------+-------+
1104 * | hdrsz  |layout |
1105 * +--------+-------+
1106 *
1107 * Bits 0-10 are the layout number
1108 * Bits 11-16 are the size of the header.
1109 * The hdrsize is the number * 8
1110 *
1111 * For example.
1112 * hdrsz of 1 ==> 8 byte header
1113 *          2 ==> 16 byte header
1114 *
1115 */
1116
1117#define	SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10)
1118#define	SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 16, 3, 0)
1119#define	SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \
1120{ \
1121	BF32_SET_SB(x, 10, 6, 3, 0, size); \
1122	BF32_SET(x, 0, 10, num); \
1123}
1124
1125#define	SA_MODE_OFFSET		0
1126#define	SA_SIZE_OFFSET		8
1127#define	SA_GEN_OFFSET		16
1128#define	SA_UID_OFFSET		24
1129#define	SA_GID_OFFSET		32
1130#define	SA_PARENT_OFFSET	40
1131#define	SA_SYMLINK_OFFSET	160
1132
1133#define	ZIO_OBJSET_MAC_LEN	32
1134
1135/*
1136 * Intent log header - this on disk structure holds fields to manage
1137 * the log.  All fields are 64 bit to easily handle cross architectures.
1138 */
1139typedef struct zil_header {
1140	uint64_t zh_claim_txg;	/* txg in which log blocks were claimed */
1141	uint64_t zh_replay_seq;	/* highest replayed sequence number */
1142	blkptr_t zh_log;	/* log chain */
1143	uint64_t zh_claim_seq;	/* highest claimed sequence number */
1144	uint64_t zh_pad[5];
1145} zil_header_t;
1146
1147#define	OBJSET_PHYS_SIZE_V2 2048
1148#define	OBJSET_PHYS_SIZE_V3 4096
1149
1150#define	OBJSET_PHYS_PAD0_SIZE	\
1151	(OBJSET_PHYS_SIZE_V2 - sizeof (dnode_phys_t) * 3 -	\
1152	    sizeof (zil_header_t) - sizeof (uint64_t) * 2 -	\
1153	    2 * ZIO_OBJSET_MAC_LEN)
1154#define	OBJSET_PHYS_PAD1_SIZE	\
1155	(OBJSET_PHYS_SIZE_V3 - OBJSET_PHYS_SIZE_V2 - sizeof (dnode_phys_t))
1156
1157typedef struct objset_phys {
1158	dnode_phys_t os_meta_dnode;
1159	zil_header_t os_zil_header;
1160	uint64_t os_type;
1161	uint64_t os_flags;
1162	uint8_t os_portable_mac[ZIO_OBJSET_MAC_LEN];
1163	uint8_t os_local_mac[ZIO_OBJSET_MAC_LEN];
1164	char os_pad0[OBJSET_PHYS_PAD0_SIZE];
1165	dnode_phys_t os_userused_dnode;
1166	dnode_phys_t os_groupused_dnode;
1167	dnode_phys_t os_projectused_dnode;
1168	char os_pad1[OBJSET_PHYS_PAD1_SIZE];
1169} objset_phys_t;
1170
1171typedef struct dsl_dir_phys {
1172	uint64_t dd_creation_time; /* not actually used */
1173	uint64_t dd_head_dataset_obj;
1174	uint64_t dd_parent_obj;
1175	uint64_t dd_clone_parent_obj;
1176	uint64_t dd_child_dir_zapobj;
1177	/*
1178	 * how much space our children are accounting for; for leaf
1179	 * datasets, == physical space used by fs + snaps
1180	 */
1181	uint64_t dd_used_bytes;
1182	uint64_t dd_compressed_bytes;
1183	uint64_t dd_uncompressed_bytes;
1184	/* Administrative quota setting */
1185	uint64_t dd_quota;
1186	/* Administrative reservation setting */
1187	uint64_t dd_reserved;
1188	uint64_t dd_props_zapobj;
1189	uint64_t dd_pad[21]; /* pad out to 256 bytes for good measure */
1190} dsl_dir_phys_t;
1191
1192typedef struct dsl_dataset_phys {
1193	uint64_t ds_dir_obj;
1194	uint64_t ds_prev_snap_obj;
1195	uint64_t ds_prev_snap_txg;
1196	uint64_t ds_next_snap_obj;
1197	uint64_t ds_snapnames_zapobj;	/* zap obj of snaps; ==0 for snaps */
1198	uint64_t ds_num_children;	/* clone/snap children; ==0 for head */
1199	uint64_t ds_creation_time;	/* seconds since 1970 */
1200	uint64_t ds_creation_txg;
1201	uint64_t ds_deadlist_obj;
1202	uint64_t ds_used_bytes;
1203	uint64_t ds_compressed_bytes;
1204	uint64_t ds_uncompressed_bytes;
1205	uint64_t ds_unique_bytes;	/* only relevant to snapshots */
1206	/*
1207	 * The ds_fsid_guid is a 56-bit ID that can change to avoid
1208	 * collisions.  The ds_guid is a 64-bit ID that will never
1209	 * change, so there is a small probability that it will collide.
1210	 */
1211	uint64_t ds_fsid_guid;
1212	uint64_t ds_guid;
1213	uint64_t ds_flags;
1214	blkptr_t ds_bp;
1215	uint64_t ds_pad[8]; /* pad out to 320 bytes for good measure */
1216} dsl_dataset_phys_t;
1217
1218/*
1219 * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
1220 */
1221#define	DMU_POOL_DIRECTORY_OBJECT	1
1222#define	DMU_POOL_CONFIG			"config"
1223#define	DMU_POOL_FEATURES_FOR_READ	"features_for_read"
1224#define	DMU_POOL_ROOT_DATASET		"root_dataset"
1225#define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
1226#define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
1227#define	DMU_POOL_ERRLOG_LAST		"errlog_last"
1228#define	DMU_POOL_SPARES			"spares"
1229#define	DMU_POOL_DEFLATE		"deflate"
1230#define	DMU_POOL_HISTORY		"history"
1231#define	DMU_POOL_PROPS			"pool_props"
1232#define	DMU_POOL_CHECKSUM_SALT		"org.illumos:checksum_salt"
1233#define	DMU_POOL_REMOVING		"com.delphix:removing"
1234#define	DMU_POOL_OBSOLETE_BPOBJ		"com.delphix:obsolete_bpobj"
1235#define	DMU_POOL_CONDENSING_INDIRECT	"com.delphix:condensing_indirect"
1236
1237#define	ZAP_MAGIC 0x2F52AB2ABULL
1238
1239#define	FZAP_BLOCK_SHIFT(zap)	((zap)->zap_block_shift)
1240
1241#define	ZAP_MAXCD		(uint32_t)(-1)
1242#define	ZAP_HASHBITS		28
1243#define	MZAP_ENT_LEN		64
1244#define	MZAP_NAME_LEN		(MZAP_ENT_LEN - 8 - 4 - 2)
1245#define	MZAP_MAX_BLKSHIFT	SPA_MAXBLOCKSHIFT
1246#define	MZAP_MAX_BLKSZ		(1 << MZAP_MAX_BLKSHIFT)
1247
1248typedef struct mzap_ent_phys {
1249	uint64_t mze_value;
1250	uint32_t mze_cd;
1251	uint16_t mze_pad;	/* in case we want to chain them someday */
1252	char mze_name[MZAP_NAME_LEN];
1253} mzap_ent_phys_t;
1254
1255typedef struct mzap_phys {
1256	uint64_t mz_block_type;	/* ZBT_MICRO */
1257	uint64_t mz_salt;
1258	uint64_t mz_pad[6];
1259	mzap_ent_phys_t mz_chunk[1];
1260	/* actually variable size depending on block size */
1261} mzap_phys_t;
1262
1263/*
1264 * The (fat) zap is stored in one object. It is an array of
1265 * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of:
1266 *
1267 * ptrtbl fits in first block:
1268 *	[zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ...
1269 *
1270 * ptrtbl too big for first block:
1271 *	[zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ...
1272 *
1273 */
1274
1275#define	ZBT_LEAF		((1ULL << 63) + 0)
1276#define	ZBT_HEADER		((1ULL << 63) + 1)
1277#define	ZBT_MICRO		((1ULL << 63) + 3)
1278/* any other values are ptrtbl blocks */
1279
1280/*
1281 * the embedded pointer table takes up half a block:
1282 * block size / entry size (2^3) / 2
1283 */
1284#define	ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1)
1285
1286/*
1287 * The embedded pointer table starts half-way through the block.  Since
1288 * the pointer table itself is half the block, it starts at (64-bit)
1289 * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)).
1290 */
1291#define	ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \
1292	((uint64_t *)(zap)->zap_phys) \
1293	[(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))]
1294
1295/*
1296 * TAKE NOTE:
1297 * If zap_phys_t is modified, zap_byteswap() must be modified.
1298 */
1299typedef struct zap_phys {
1300	uint64_t zap_block_type;	/* ZBT_HEADER */
1301	uint64_t zap_magic;		/* ZAP_MAGIC */
1302
1303	struct zap_table_phys {
1304		uint64_t zt_blk;	/* starting block number */
1305		uint64_t zt_numblks;	/* number of blocks */
1306		uint64_t zt_shift;	/* bits to index it */
1307		uint64_t zt_nextblk;	/* next (larger) copy start block */
1308		uint64_t zt_blks_copied; /* number source blocks copied */
1309	} zap_ptrtbl;
1310
1311	uint64_t zap_freeblk;		/* the next free block */
1312	uint64_t zap_num_leafs;		/* number of leafs */
1313	uint64_t zap_num_entries;	/* number of entries */
1314	uint64_t zap_salt;		/* salt to stir into hash function */
1315	/*
1316	 * This structure is followed by padding, and then the embedded
1317	 * pointer table.  The embedded pointer table takes up second
1318	 * half of the block.  It is accessed using the
1319	 * ZAP_EMBEDDED_PTRTBL_ENT() macro.
1320	 */
1321} zap_phys_t;
1322
1323typedef struct zap_table_phys zap_table_phys_t;
1324
1325typedef struct fat_zap {
1326	int zap_block_shift;			/* block size shift */
1327	zap_phys_t *zap_phys;
1328} fat_zap_t;
1329
1330#define	ZAP_LEAF_MAGIC 0x2AB1EAF
1331
1332/* chunk size = 24 bytes */
1333#define	ZAP_LEAF_CHUNKSIZE 24
1334
1335/*
1336 * The amount of space available for chunks is:
1337 * block size (1<<l->l_bs) - hash entry size (2) * number of hash
1338 * entries - header space (2*chunksize)
1339 */
1340#define	ZAP_LEAF_NUMCHUNKS(l) \
1341	(((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \
1342	ZAP_LEAF_CHUNKSIZE - 2)
1343
1344/*
1345 * The amount of space within the chunk available for the array is:
1346 * chunk size - space for type (1) - space for next pointer (2)
1347 */
1348#define	ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3)
1349
1350#define	ZAP_LEAF_ARRAY_NCHUNKS(bytes) \
1351	(((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES)
1352
1353/*
1354 * Low water mark:  when there are only this many chunks free, start
1355 * growing the ptrtbl.  Ideally, this should be larger than a
1356 * "reasonably-sized" entry.  20 chunks is more than enough for the
1357 * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value),
1358 * while still being only around 3% for 16k blocks.
1359 */
1360#define	ZAP_LEAF_LOW_WATER (20)
1361
1362/*
1363 * The leaf hash table has block size / 2^5 (32) number of entries,
1364 * which should be more than enough for the maximum number of entries,
1365 * which is less than block size / CHUNKSIZE (24) / minimum number of
1366 * chunks per entry (3).
1367 */
1368#define	ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5)
1369#define	ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l))
1370
1371/*
1372 * The chunks start immediately after the hash table.  The end of the
1373 * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a
1374 * chunk_t.
1375 */
1376#define	ZAP_LEAF_CHUNK(l, idx) \
1377	((zap_leaf_chunk_t *) \
1378	((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx]
1379#define	ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry)
1380
1381typedef enum zap_chunk_type {
1382	ZAP_CHUNK_FREE = 253,
1383	ZAP_CHUNK_ENTRY = 252,
1384	ZAP_CHUNK_ARRAY = 251,
1385	ZAP_CHUNK_TYPE_MAX = 250
1386} zap_chunk_type_t;
1387
1388/*
1389 * TAKE NOTE:
1390 * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified.
1391 */
1392typedef struct zap_leaf_phys {
1393	struct zap_leaf_header {
1394		uint64_t lh_block_type;		/* ZBT_LEAF */
1395		uint64_t lh_pad1;
1396		uint64_t lh_prefix;		/* hash prefix of this leaf */
1397		uint32_t lh_magic;		/* ZAP_LEAF_MAGIC */
1398		uint16_t lh_nfree;		/* number free chunks */
1399		uint16_t lh_nentries;		/* number of entries */
1400		uint16_t lh_prefix_len;		/* num bits used to id this */
1401
1402/* above is accessable to zap, below is zap_leaf private */
1403
1404		uint16_t lh_freelist;		/* chunk head of free list */
1405		uint8_t lh_pad2[12];
1406	} l_hdr; /* 2 24-byte chunks */
1407
1408	/*
1409	 * The header is followed by a hash table with
1410	 * ZAP_LEAF_HASH_NUMENTRIES(zap) entries.  The hash table is
1411	 * followed by an array of ZAP_LEAF_NUMCHUNKS(zap)
1412	 * zap_leaf_chunk structures.  These structures are accessed
1413	 * with the ZAP_LEAF_CHUNK() macro.
1414	 */
1415
1416	uint16_t l_hash[1];
1417} zap_leaf_phys_t;
1418
1419typedef union zap_leaf_chunk {
1420	struct zap_leaf_entry {
1421		uint8_t le_type;		/* always ZAP_CHUNK_ENTRY */
1422		uint8_t le_value_intlen;	/* size of ints */
1423		uint16_t le_next;		/* next entry in hash chain */
1424		uint16_t le_name_chunk;		/* first chunk of the name */
1425		uint16_t le_name_numints;	/* bytes in name, incl null */
1426		uint16_t le_value_chunk;	/* first chunk of the value */
1427		uint16_t le_value_numints;	/* value length in ints */
1428		uint32_t le_cd;			/* collision differentiator */
1429		uint64_t le_hash;		/* hash value of the name */
1430	} l_entry;
1431	struct zap_leaf_array {
1432		uint8_t la_type;		/* always ZAP_CHUNK_ARRAY */
1433		uint8_t la_array[ZAP_LEAF_ARRAY_BYTES];
1434		uint16_t la_next;		/* next blk or CHAIN_END */
1435	} l_array;
1436	struct zap_leaf_free {
1437		uint8_t lf_type;		/* always ZAP_CHUNK_FREE */
1438		uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES];
1439		uint16_t lf_next;	/* next in free list, or CHAIN_END */
1440	} l_free;
1441} zap_leaf_chunk_t;
1442
1443typedef struct zap_leaf {
1444	int l_bs;			/* block size shift */
1445	zap_leaf_phys_t *l_phys;
1446} zap_leaf_t;
1447
1448/*
1449 * Define special zfs pflags
1450 */
1451#define	ZFS_XATTR	0x1		/* is an extended attribute */
1452#define	ZFS_INHERIT_ACE	0x2		/* ace has inheritable ACEs */
1453#define	ZFS_ACL_TRIVIAL 0x4		/* files ACL is trivial */
1454
1455#define	MASTER_NODE_OBJ	1
1456
1457/*
1458 * special attributes for master node.
1459 */
1460
1461#define	ZFS_FSID		"FSID"
1462#define	ZFS_UNLINKED_SET	"DELETE_QUEUE"
1463#define	ZFS_ROOT_OBJ		"ROOT"
1464#define	ZPL_VERSION_OBJ		"VERSION"
1465#define	ZFS_PROP_BLOCKPERPAGE	"BLOCKPERPAGE"
1466#define	ZFS_PROP_NOGROWBLOCKS	"NOGROWBLOCKS"
1467
1468#define	ZFS_FLAG_BLOCKPERPAGE	0x1
1469#define	ZFS_FLAG_NOGROWBLOCKS	0x2
1470
1471/*
1472 * ZPL version - rev'd whenever an incompatible on-disk format change
1473 * occurs.  Independent of SPA/DMU/ZAP versioning.
1474 */
1475
1476#define	ZPL_VERSION		1ULL
1477
1478/*
1479 * The directory entry has the type (currently unused on Solaris) in the
1480 * top 4 bits, and the object number in the low 48 bits.  The "middle"
1481 * 12 bits are unused.
1482 */
1483#define	ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4)
1484#define	ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48)
1485#define	ZFS_DIRENT_MAKE(type, obj) (((uint64_t)type << 60) | obj)
1486
1487typedef struct ace {
1488	uid_t		a_who;		/* uid or gid */
1489	uint32_t	a_access_mask;	/* read,write,... */
1490	uint16_t	a_flags;	/* see below */
1491	uint16_t	a_type;		/* allow or deny */
1492} ace_t;
1493
1494#define	ACE_SLOT_CNT	6
1495
1496typedef struct zfs_znode_acl {
1497	uint64_t	z_acl_extern_obj;	  /* ext acl pieces */
1498	uint32_t	z_acl_count;		  /* Number of ACEs */
1499	uint16_t	z_acl_version;		  /* acl version */
1500	uint16_t	z_acl_pad;		  /* pad */
1501	ace_t		z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */
1502} zfs_znode_acl_t;
1503
1504/*
1505 * This is the persistent portion of the znode.  It is stored
1506 * in the "bonus buffer" of the file.  Short symbolic links
1507 * are also stored in the bonus buffer.
1508 */
1509typedef struct znode_phys {
1510	uint64_t zp_atime[2];		/*  0 - last file access time */
1511	uint64_t zp_mtime[2];		/* 16 - last file modification time */
1512	uint64_t zp_ctime[2];		/* 32 - last file change time */
1513	uint64_t zp_crtime[2];		/* 48 - creation time */
1514	uint64_t zp_gen;		/* 64 - generation (txg of creation) */
1515	uint64_t zp_mode;		/* 72 - file mode bits */
1516	uint64_t zp_size;		/* 80 - size of file */
1517	uint64_t zp_parent;		/* 88 - directory parent (`..') */
1518	uint64_t zp_links;		/* 96 - number of links to file */
1519	uint64_t zp_xattr;		/* 104 - DMU object for xattrs */
1520	uint64_t zp_rdev;		/* 112 - dev_t for VBLK & VCHR files */
1521	uint64_t zp_flags;		/* 120 - persistent flags */
1522	uint64_t zp_uid;		/* 128 - file owner */
1523	uint64_t zp_gid;		/* 136 - owning group */
1524	uint64_t zp_pad[4];		/* 144 - future */
1525	zfs_znode_acl_t zp_acl;		/* 176 - 263 ACL */
1526	/*
1527	 * Data may pad out any remaining bytes in the znode buffer, eg:
1528	 *
1529	 * |<---------------------- dnode_phys (512) ------------------------>|
1530	 * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->|
1531	 *			|<---- znode (264) ---->|<---- data (56) ---->|
1532	 *
1533	 * At present, we only use this space to store symbolic links.
1534	 */
1535} znode_phys_t;
1536
1537/*
1538 * In-core vdev representation.
1539 */
1540struct vdev;
1541struct spa;
1542typedef int vdev_phys_read_t(struct vdev *vdev, void *priv,
1543    off_t offset, void *buf, size_t bytes);
1544typedef int vdev_read_t(struct vdev *vdev, const blkptr_t *bp,
1545    void *buf, off_t offset, size_t bytes);
1546
1547typedef STAILQ_HEAD(vdev_list, vdev) vdev_list_t;
1548
1549typedef struct vdev_indirect_mapping_entry_phys {
1550	/*
1551	 * Decode with DVA_MAPPING_* macros.
1552	 * Contains:
1553	 *   the source offset (low 63 bits)
1554	 *   the one-bit "mark", used for garbage collection (by zdb)
1555	 */
1556	uint64_t vimep_src;
1557
1558	/*
1559	 * Note: the DVA's asize is 24 bits, and can thus store ranges
1560	 * up to 8GB.
1561	 */
1562	dva_t	vimep_dst;
1563} vdev_indirect_mapping_entry_phys_t;
1564
1565#define	DVA_MAPPING_GET_SRC_OFFSET(vimep)	\
1566	BF64_GET_SB((vimep)->vimep_src, 0, 63, SPA_MINBLOCKSHIFT, 0)
1567#define	DVA_MAPPING_SET_SRC_OFFSET(vimep, x)	\
1568	BF64_SET_SB((vimep)->vimep_src, 0, 63, SPA_MINBLOCKSHIFT, 0, x)
1569
1570typedef struct vdev_indirect_mapping_entry {
1571	vdev_indirect_mapping_entry_phys_t	vime_mapping;
1572	uint32_t				vime_obsolete_count;
1573	list_node_t				vime_node;
1574} vdev_indirect_mapping_entry_t;
1575
1576/*
1577 * This is stored in the bonus buffer of the mapping object, see comment of
1578 * vdev_indirect_config for more details.
1579 */
1580typedef struct vdev_indirect_mapping_phys {
1581	uint64_t	vimp_max_offset;
1582	uint64_t	vimp_bytes_mapped;
1583	uint64_t	vimp_num_entries; /* number of v_i_m_entry_phys_t's */
1584
1585	/*
1586	 * For each entry in the mapping object, this object contains an
1587	 * entry representing the number of bytes of that mapping entry
1588	 * that were no longer in use by the pool at the time this indirect
1589	 * vdev was last condensed.
1590	 */
1591	uint64_t	vimp_counts_object;
1592} vdev_indirect_mapping_phys_t;
1593
1594#define	VDEV_INDIRECT_MAPPING_SIZE_V0	(3 * sizeof (uint64_t))
1595
1596typedef struct vdev_indirect_mapping {
1597	uint64_t	vim_object;
1598	boolean_t	vim_havecounts;
1599
1600	/* vim_entries segment offset currently in memory. */
1601	uint64_t	vim_entry_offset;
1602	/* vim_entries segment size. */
1603	size_t		vim_num_entries;
1604
1605	/* Needed by dnode_read() */
1606	const void	*vim_spa;
1607	dnode_phys_t	*vim_dn;
1608
1609	/*
1610	 * An ordered array of mapping entries, sorted by source offset.
1611	 * Note that vim_entries is needed during a removal (and contains
1612	 * mappings that have been synced to disk so far) to handle frees
1613	 * from the removing device.
1614	 */
1615	vdev_indirect_mapping_entry_phys_t *vim_entries;
1616	objset_phys_t	*vim_objset;
1617	vdev_indirect_mapping_phys_t	*vim_phys;
1618} vdev_indirect_mapping_t;
1619
1620/*
1621 * On-disk indirect vdev state.
1622 *
1623 * An indirect vdev is described exclusively in the MOS config of a pool.
1624 * The config for an indirect vdev includes several fields, which are
1625 * accessed in memory by a vdev_indirect_config_t.
1626 */
1627typedef struct vdev_indirect_config {
1628	/*
1629	 * Object (in MOS) which contains the indirect mapping. This object
1630	 * contains an array of vdev_indirect_mapping_entry_phys_t ordered by
1631	 * vimep_src. The bonus buffer for this object is a
1632	 * vdev_indirect_mapping_phys_t. This object is allocated when a vdev
1633	 * removal is initiated.
1634	 *
1635	 * Note that this object can be empty if none of the data on the vdev
1636	 * has been copied yet.
1637	 */
1638	uint64_t	vic_mapping_object;
1639
1640	/*
1641	 * Object (in MOS) which contains the birth times for the mapping
1642	 * entries. This object contains an array of
1643	 * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus
1644	 * buffer for this object is a vdev_indirect_birth_phys_t. This object
1645	 * is allocated when a vdev removal is initiated.
1646	 *
1647	 * Note that this object can be empty if none of the vdev has yet been
1648	 * copied.
1649	 */
1650	uint64_t	vic_births_object;
1651
1652/*
1653 * This is the vdev ID which was removed previous to this vdev, or
1654 * UINT64_MAX if there are no previously removed vdevs.
1655 */
1656	uint64_t	vic_prev_indirect_vdev;
1657} vdev_indirect_config_t;
1658
1659typedef struct vdev {
1660	STAILQ_ENTRY(vdev) v_childlink;	/* link in parent's child list */
1661	STAILQ_ENTRY(vdev) v_alllink;	/* link in global vdev list */
1662	vdev_list_t	v_children;	/* children of this vdev */
1663	const char	*v_name;	/* vdev name */
1664	const char	*v_phys_path;	/* vdev bootpath */
1665	const char	*v_devid;	/* vdev devid */
1666	uint64_t	v_guid;		/* vdev guid */
1667	int		v_id;		/* index in parent */
1668	int		v_ashift;	/* offset to block shift */
1669	int		v_nparity;	/* # parity for raidz */
1670	struct vdev	*v_top;		/* parent vdev */
1671	int		v_nchildren;	/* # children */
1672	vdev_state_t	v_state;	/* current state */
1673	vdev_phys_read_t *v_phys_read;	/* read from raw leaf vdev */
1674	vdev_read_t	*v_read;	/* read from vdev */
1675	void		*v_read_priv;	/* private data for read function */
1676	struct spa	*spa;		/* link to spa */
1677	/*
1678	 * Values stored in the config for an indirect or removing vdev.
1679	 */
1680	vdev_indirect_config_t vdev_indirect_config;
1681	vdev_indirect_mapping_t *v_mapping;
1682} vdev_t;
1683
1684/*
1685 * In-core pool representation.
1686 */
1687typedef STAILQ_HEAD(spa_list, spa) spa_list_t;
1688
1689typedef struct spa {
1690	STAILQ_ENTRY(spa) spa_link;	/* link in global pool list */
1691	char		*spa_name;	/* pool name */
1692	uint64_t	spa_guid;	/* pool guid */
1693	uint64_t	spa_txg;	/* most recent transaction */
1694	struct uberblock spa_uberblock;	/* best uberblock so far */
1695	vdev_list_t	spa_vdevs;	/* list of all toplevel vdevs */
1696	objset_phys_t	spa_mos;	/* MOS for this pool */
1697	zio_cksum_salt_t spa_cksum_salt;	/* secret salt for cksum */
1698	void		*spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS];
1699	int		spa_inited;	/* initialized */
1700	vdev_t		*spa_boot_vdev;	/* boot device for kernel */
1701} spa_t;
1702
1703/* IO related arguments. */
1704typedef struct zio {
1705	spa_t		*io_spa;
1706	blkptr_t	*io_bp;
1707	void		*io_data;
1708	uint64_t	io_size;
1709	uint64_t	io_offset;
1710
1711	/* Stuff for the vdev stack */
1712	vdev_t		*io_vd;
1713	void		*io_vsd;
1714
1715	int		io_error;
1716} zio_t;
1717
1718static void decode_embedded_bp_compressed(const blkptr_t *, void *);
1719
1720#endif	/* _ZFSIMPL_H */
1721