xref: /freebsd/sys/cddl/boot/zfs/zfsimpl.h (revision eb6d21b4ca6d668cf89afd99eef7baeafa712197)
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 2007 Sun Microsystems, Inc.  All rights reserved.
53  * Use is subject to license terms.
54  */
55 
56 /* CRC64 table */
57 #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
58 
59 /*
60  * Macros for various sorts of alignment and rounding when the alignment
61  * is known to be a power of 2.
62  */
63 #define	P2ALIGN(x, align)		((x) & -(align))
64 #define	P2PHASE(x, align)		((x) & ((align) - 1))
65 #define	P2NPHASE(x, align)		(-(x) & ((align) - 1))
66 #define	P2ROUNDUP(x, align)		(-(-(x) & -(align)))
67 #define	P2END(x, align)			(-(~(x) & -(align)))
68 #define	P2PHASEUP(x, align, phase)	((phase) - (((phase) - (x)) & -(align)))
69 #define	P2CROSS(x, y, align)		(((x) ^ (y)) > (align) - 1)
70 
71 /*
72  * General-purpose 32-bit and 64-bit bitfield encodings.
73  */
74 #define	BF32_DECODE(x, low, len)	P2PHASE((x) >> (low), 1U << (len))
75 #define	BF64_DECODE(x, low, len)	P2PHASE((x) >> (low), 1ULL << (len))
76 #define	BF32_ENCODE(x, low, len)	(P2PHASE((x), 1U << (len)) << (low))
77 #define	BF64_ENCODE(x, low, len)	(P2PHASE((x), 1ULL << (len)) << (low))
78 
79 #define	BF32_GET(x, low, len)		BF32_DECODE(x, low, len)
80 #define	BF64_GET(x, low, len)		BF64_DECODE(x, low, len)
81 
82 #define	BF32_SET(x, low, len, val)	\
83 	((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len))
84 #define	BF64_SET(x, low, len, val)	\
85 	((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len))
86 
87 #define	BF32_GET_SB(x, low, len, shift, bias)	\
88 	((BF32_GET(x, low, len) + (bias)) << (shift))
89 #define	BF64_GET_SB(x, low, len, shift, bias)	\
90 	((BF64_GET(x, low, len) + (bias)) << (shift))
91 
92 #define	BF32_SET_SB(x, low, len, shift, bias, val)	\
93 	BF32_SET(x, low, len, ((val) >> (shift)) - (bias))
94 #define	BF64_SET_SB(x, low, len, shift, bias, val)	\
95 	BF64_SET(x, low, len, ((val) >> (shift)) - (bias))
96 
97 /*
98  * We currently support nine block sizes, from 512 bytes to 128K.
99  * We could go higher, but the benefits are near-zero and the cost
100  * of COWing a giant block to modify one byte would become excessive.
101  */
102 #define	SPA_MINBLOCKSHIFT	9
103 #define	SPA_MAXBLOCKSHIFT	17
104 #define	SPA_MINBLOCKSIZE	(1ULL << SPA_MINBLOCKSHIFT)
105 #define	SPA_MAXBLOCKSIZE	(1ULL << SPA_MAXBLOCKSHIFT)
106 
107 #define	SPA_BLOCKSIZES		(SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1)
108 
109 /*
110  * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
111  * The ASIZE encoding should be at least 64 times larger (6 more bits)
112  * to support up to 4-way RAID-Z mirror mode with worst-case gang block
113  * overhead, three DVAs per bp, plus one more bit in case we do anything
114  * else that expands the ASIZE.
115  */
116 #define	SPA_LSIZEBITS		16	/* LSIZE up to 32M (2^16 * 512)	*/
117 #define	SPA_PSIZEBITS		16	/* PSIZE up to 32M (2^16 * 512)	*/
118 #define	SPA_ASIZEBITS		24	/* ASIZE up to 64 times larger	*/
119 
120 /*
121  * All SPA data is represented by 128-bit data virtual addresses (DVAs).
122  * The members of the dva_t should be considered opaque outside the SPA.
123  */
124 typedef struct dva {
125 	uint64_t	dva_word[2];
126 } dva_t;
127 
128 /*
129  * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
130  */
131 typedef struct zio_cksum {
132 	uint64_t	zc_word[4];
133 } zio_cksum_t;
134 
135 /*
136  * Each block is described by its DVAs, time of birth, checksum, etc.
137  * The word-by-word, bit-by-bit layout of the blkptr is as follows:
138  *
139  *	64	56	48	40	32	24	16	8	0
140  *	+-------+-------+-------+-------+-------+-------+-------+-------+
141  * 0	|		vdev1		| GRID  |	  ASIZE		|
142  *	+-------+-------+-------+-------+-------+-------+-------+-------+
143  * 1	|G|			 offset1				|
144  *	+-------+-------+-------+-------+-------+-------+-------+-------+
145  * 2	|		vdev2		| GRID  |	  ASIZE		|
146  *	+-------+-------+-------+-------+-------+-------+-------+-------+
147  * 3	|G|			 offset2				|
148  *	+-------+-------+-------+-------+-------+-------+-------+-------+
149  * 4	|		vdev3		| GRID  |	  ASIZE		|
150  *	+-------+-------+-------+-------+-------+-------+-------+-------+
151  * 5	|G|			 offset3				|
152  *	+-------+-------+-------+-------+-------+-------+-------+-------+
153  * 6	|E| lvl | type	| cksum | comp	|     PSIZE	|     LSIZE	|
154  *	+-------+-------+-------+-------+-------+-------+-------+-------+
155  * 7	|			padding					|
156  *	+-------+-------+-------+-------+-------+-------+-------+-------+
157  * 8	|			padding					|
158  *	+-------+-------+-------+-------+-------+-------+-------+-------+
159  * 9	|			padding					|
160  *	+-------+-------+-------+-------+-------+-------+-------+-------+
161  * a	|			birth txg				|
162  *	+-------+-------+-------+-------+-------+-------+-------+-------+
163  * b	|			fill count				|
164  *	+-------+-------+-------+-------+-------+-------+-------+-------+
165  * c	|			checksum[0]				|
166  *	+-------+-------+-------+-------+-------+-------+-------+-------+
167  * d	|			checksum[1]				|
168  *	+-------+-------+-------+-------+-------+-------+-------+-------+
169  * e	|			checksum[2]				|
170  *	+-------+-------+-------+-------+-------+-------+-------+-------+
171  * f	|			checksum[3]				|
172  *	+-------+-------+-------+-------+-------+-------+-------+-------+
173  *
174  * Legend:
175  *
176  * vdev		virtual device ID
177  * offset	offset into virtual device
178  * LSIZE	logical size
179  * PSIZE	physical size (after compression)
180  * ASIZE	allocated size (including RAID-Z parity and gang block headers)
181  * GRID		RAID-Z layout information (reserved for future use)
182  * cksum	checksum function
183  * comp		compression function
184  * G		gang block indicator
185  * E		endianness
186  * type		DMU object type
187  * lvl		level of indirection
188  * birth txg	transaction group in which the block was born
189  * fill count	number of non-zero blocks under this bp
190  * checksum[4]	256-bit checksum of the data this bp describes
191  */
192 typedef struct blkptr {
193 	dva_t		blk_dva[3];	/* 128-bit Data Virtual Address	*/
194 	uint64_t	blk_prop;	/* size, compression, type, etc	*/
195 	uint64_t	blk_pad[3];	/* Extra space for the future	*/
196 	uint64_t	blk_birth;	/* transaction group at birth	*/
197 	uint64_t	blk_fill;	/* fill count			*/
198 	zio_cksum_t	blk_cksum;	/* 256-bit checksum		*/
199 } blkptr_t;
200 
201 #define	SPA_BLKPTRSHIFT	7		/* blkptr_t is 128 bytes	*/
202 #define	SPA_DVAS_PER_BP	3		/* Number of DVAs in a bp	*/
203 
204 /*
205  * Macros to get and set fields in a bp or DVA.
206  */
207 #define	DVA_GET_ASIZE(dva)	\
208 	BF64_GET_SB((dva)->dva_word[0], 0, 24, SPA_MINBLOCKSHIFT, 0)
209 #define	DVA_SET_ASIZE(dva, x)	\
210 	BF64_SET_SB((dva)->dva_word[0], 0, 24, SPA_MINBLOCKSHIFT, 0, x)
211 
212 #define	DVA_GET_GRID(dva)	BF64_GET((dva)->dva_word[0], 24, 8)
213 #define	DVA_SET_GRID(dva, x)	BF64_SET((dva)->dva_word[0], 24, 8, x)
214 
215 #define	DVA_GET_VDEV(dva)	BF64_GET((dva)->dva_word[0], 32, 32)
216 #define	DVA_SET_VDEV(dva, x)	BF64_SET((dva)->dva_word[0], 32, 32, x)
217 
218 #define	DVA_GET_OFFSET(dva)	\
219 	BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
220 #define	DVA_SET_OFFSET(dva, x)	\
221 	BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
222 
223 #define	DVA_GET_GANG(dva)	BF64_GET((dva)->dva_word[1], 63, 1)
224 #define	DVA_SET_GANG(dva, x)	BF64_SET((dva)->dva_word[1], 63, 1, x)
225 
226 #define	BP_GET_LSIZE(bp)	\
227 	(BP_IS_HOLE(bp) ? 0 : \
228 	BF64_GET_SB((bp)->blk_prop, 0, 16, SPA_MINBLOCKSHIFT, 1))
229 #define	BP_SET_LSIZE(bp, x)	\
230 	BF64_SET_SB((bp)->blk_prop, 0, 16, SPA_MINBLOCKSHIFT, 1, x)
231 
232 #define	BP_GET_PSIZE(bp)	\
233 	BF64_GET_SB((bp)->blk_prop, 16, 16, SPA_MINBLOCKSHIFT, 1)
234 #define	BP_SET_PSIZE(bp, x)	\
235 	BF64_SET_SB((bp)->blk_prop, 16, 16, SPA_MINBLOCKSHIFT, 1, x)
236 
237 #define	BP_GET_COMPRESS(bp)	BF64_GET((bp)->blk_prop, 32, 8)
238 #define	BP_SET_COMPRESS(bp, x)	BF64_SET((bp)->blk_prop, 32, 8, x)
239 
240 #define	BP_GET_CHECKSUM(bp)	BF64_GET((bp)->blk_prop, 40, 8)
241 #define	BP_SET_CHECKSUM(bp, x)	BF64_SET((bp)->blk_prop, 40, 8, x)
242 
243 #define	BP_GET_TYPE(bp)		BF64_GET((bp)->blk_prop, 48, 8)
244 #define	BP_SET_TYPE(bp, x)	BF64_SET((bp)->blk_prop, 48, 8, x)
245 
246 #define	BP_GET_LEVEL(bp)	BF64_GET((bp)->blk_prop, 56, 5)
247 #define	BP_SET_LEVEL(bp, x)	BF64_SET((bp)->blk_prop, 56, 5, x)
248 
249 #define	BP_GET_BYTEORDER(bp)	(0 - BF64_GET((bp)->blk_prop, 63, 1))
250 #define	BP_SET_BYTEORDER(bp, x)	BF64_SET((bp)->blk_prop, 63, 1, x)
251 
252 #define	BP_GET_ASIZE(bp)	\
253 	(DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
254 		DVA_GET_ASIZE(&(bp)->blk_dva[2]))
255 
256 #define	BP_GET_UCSIZE(bp) \
257 	((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \
258 	BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp));
259 
260 #define	BP_GET_NDVAS(bp)	\
261 	(!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
262 	!!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
263 	!!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
264 
265 #define	BP_COUNT_GANG(bp)	\
266 	(DVA_GET_GANG(&(bp)->blk_dva[0]) + \
267 	DVA_GET_GANG(&(bp)->blk_dva[1]) + \
268 	DVA_GET_GANG(&(bp)->blk_dva[2]))
269 
270 #define	DVA_EQUAL(dva1, dva2)	\
271 	((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
272 	(dva1)->dva_word[0] == (dva2)->dva_word[0])
273 
274 #define	ZIO_CHECKSUM_EQUAL(zc1, zc2) \
275 	(0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
276 	((zc1).zc_word[1] - (zc2).zc_word[1]) | \
277 	((zc1).zc_word[2] - (zc2).zc_word[2]) | \
278 	((zc1).zc_word[3] - (zc2).zc_word[3])))
279 
280 
281 #define	DVA_IS_VALID(dva)	(DVA_GET_ASIZE(dva) != 0)
282 
283 #define	ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3)	\
284 {						\
285 	(zcp)->zc_word[0] = w0;			\
286 	(zcp)->zc_word[1] = w1;			\
287 	(zcp)->zc_word[2] = w2;			\
288 	(zcp)->zc_word[3] = w3;			\
289 }
290 
291 #define	BP_IDENTITY(bp)		(&(bp)->blk_dva[0])
292 #define	BP_IS_GANG(bp)		DVA_GET_GANG(BP_IDENTITY(bp))
293 #define	BP_IS_HOLE(bp)		((bp)->blk_birth == 0)
294 #define	BP_IS_OLDER(bp, txg)	(!BP_IS_HOLE(bp) && (bp)->blk_birth < (txg))
295 
296 #define	BP_ZERO(bp)				\
297 {						\
298 	(bp)->blk_dva[0].dva_word[0] = 0;	\
299 	(bp)->blk_dva[0].dva_word[1] = 0;	\
300 	(bp)->blk_dva[1].dva_word[0] = 0;	\
301 	(bp)->blk_dva[1].dva_word[1] = 0;	\
302 	(bp)->blk_dva[2].dva_word[0] = 0;	\
303 	(bp)->blk_dva[2].dva_word[1] = 0;	\
304 	(bp)->blk_prop = 0;			\
305 	(bp)->blk_pad[0] = 0;			\
306 	(bp)->blk_pad[1] = 0;			\
307 	(bp)->blk_pad[2] = 0;			\
308 	(bp)->blk_birth = 0;			\
309 	(bp)->blk_fill = 0;			\
310 	ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0);	\
311 }
312 
313 #define	ZBT_MAGIC	0x210da7ab10c7a11ULL	/* zio data bloc tail */
314 
315 typedef struct zio_block_tail {
316 	uint64_t	zbt_magic;	/* for validation, endianness	*/
317 	zio_cksum_t	zbt_cksum;	/* 256-bit checksum		*/
318 } zio_block_tail_t;
319 
320 #define	VDEV_SKIP_SIZE		(8 << 10)
321 #define	VDEV_BOOT_HEADER_SIZE	(8 << 10)
322 #define	VDEV_PHYS_SIZE		(112 << 10)
323 #define	VDEV_UBERBLOCK_RING	(128 << 10)
324 
325 #define	VDEV_UBERBLOCK_SHIFT(vd)	\
326 	MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT)
327 #define	VDEV_UBERBLOCK_COUNT(vd)	\
328 	(VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
329 #define	VDEV_UBERBLOCK_OFFSET(vd, n)	\
330 	offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
331 #define	VDEV_UBERBLOCK_SIZE(vd)		(1ULL << VDEV_UBERBLOCK_SHIFT(vd))
332 
333 /* ZFS boot block */
334 #define	VDEV_BOOT_MAGIC		0x2f5b007b10cULL
335 #define	VDEV_BOOT_VERSION	1		/* version number	*/
336 
337 typedef struct vdev_boot_header {
338 	uint64_t	vb_magic;		/* VDEV_BOOT_MAGIC	*/
339 	uint64_t	vb_version;		/* VDEV_BOOT_VERSION	*/
340 	uint64_t	vb_offset;		/* start offset	(bytes) */
341 	uint64_t	vb_size;		/* size (bytes)		*/
342 	char		vb_pad[VDEV_BOOT_HEADER_SIZE - 4 * sizeof (uint64_t)];
343 } vdev_boot_header_t;
344 
345 typedef struct vdev_phys {
346 	char		vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_block_tail_t)];
347 	zio_block_tail_t vp_zbt;
348 } vdev_phys_t;
349 
350 typedef struct vdev_label {
351 	char		vl_pad[VDEV_SKIP_SIZE];			/*   8K	*/
352 	vdev_boot_header_t vl_boot_header;			/*   8K	*/
353 	vdev_phys_t	vl_vdev_phys;				/* 112K	*/
354 	char		vl_uberblock[VDEV_UBERBLOCK_RING];	/* 128K	*/
355 } vdev_label_t;							/* 256K total */
356 
357 /*
358  * vdev_dirty() flags
359  */
360 #define	VDD_METASLAB	0x01
361 #define	VDD_DTL		0x02
362 
363 /*
364  * Size and offset of embedded boot loader region on each label.
365  * The total size of the first two labels plus the boot area is 4MB.
366  */
367 #define	VDEV_BOOT_OFFSET	(2 * sizeof (vdev_label_t))
368 #define	VDEV_BOOT_SIZE		(7ULL << 19)			/* 3.5M	*/
369 
370 /*
371  * Size of label regions at the start and end of each leaf device.
372  */
373 #define	VDEV_LABEL_START_SIZE	(2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
374 #define	VDEV_LABEL_END_SIZE	(2 * sizeof (vdev_label_t))
375 #define	VDEV_LABELS		4
376 
377 /*
378  * Gang block headers are self-checksumming and contain an array
379  * of block pointers.
380  */
381 #define SPA_GANGBLOCKSIZE	SPA_MINBLOCKSIZE
382 #define SPA_GBH_NBLKPTRS	((SPA_GANGBLOCKSIZE - \
383 	sizeof (zio_block_tail_t)) / sizeof (blkptr_t))
384 #define SPA_GBH_FILLER		((SPA_GANGBLOCKSIZE - \
385 	sizeof (zio_block_tail_t) - \
386 	(SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\
387 	sizeof (uint64_t))
388 
389 typedef struct zio_gbh {
390 	blkptr_t		zg_blkptr[SPA_GBH_NBLKPTRS];
391 	uint64_t		zg_filler[SPA_GBH_FILLER];
392 	zio_block_tail_t	zg_tail;
393 } zio_gbh_phys_t;
394 
395 enum zio_checksum {
396 	ZIO_CHECKSUM_INHERIT = 0,
397 	ZIO_CHECKSUM_ON,
398 	ZIO_CHECKSUM_OFF,
399 	ZIO_CHECKSUM_LABEL,
400 	ZIO_CHECKSUM_GANG_HEADER,
401 	ZIO_CHECKSUM_ZILOG,
402 	ZIO_CHECKSUM_FLETCHER_2,
403 	ZIO_CHECKSUM_FLETCHER_4,
404 	ZIO_CHECKSUM_SHA256,
405 	ZIO_CHECKSUM_FUNCTIONS
406 };
407 
408 #define	ZIO_CHECKSUM_ON_VALUE	ZIO_CHECKSUM_FLETCHER_2
409 #define	ZIO_CHECKSUM_DEFAULT	ZIO_CHECKSUM_ON
410 
411 enum zio_compress {
412 	ZIO_COMPRESS_INHERIT = 0,
413 	ZIO_COMPRESS_ON,
414 	ZIO_COMPRESS_OFF,
415 	ZIO_COMPRESS_LZJB,
416 	ZIO_COMPRESS_EMPTY,
417 	ZIO_COMPRESS_GZIP_1,
418 	ZIO_COMPRESS_GZIP_2,
419 	ZIO_COMPRESS_GZIP_3,
420 	ZIO_COMPRESS_GZIP_4,
421 	ZIO_COMPRESS_GZIP_5,
422 	ZIO_COMPRESS_GZIP_6,
423 	ZIO_COMPRESS_GZIP_7,
424 	ZIO_COMPRESS_GZIP_8,
425 	ZIO_COMPRESS_GZIP_9,
426 	ZIO_COMPRESS_FUNCTIONS
427 };
428 
429 #define	ZIO_COMPRESS_ON_VALUE	ZIO_COMPRESS_LZJB
430 #define	ZIO_COMPRESS_DEFAULT	ZIO_COMPRESS_OFF
431 
432 /* nvlist pack encoding */
433 #define	NV_ENCODE_NATIVE	0
434 #define	NV_ENCODE_XDR		1
435 
436 typedef enum {
437 	DATA_TYPE_UNKNOWN = 0,
438 	DATA_TYPE_BOOLEAN,
439 	DATA_TYPE_BYTE,
440 	DATA_TYPE_INT16,
441 	DATA_TYPE_UINT16,
442 	DATA_TYPE_INT32,
443 	DATA_TYPE_UINT32,
444 	DATA_TYPE_INT64,
445 	DATA_TYPE_UINT64,
446 	DATA_TYPE_STRING,
447 	DATA_TYPE_BYTE_ARRAY,
448 	DATA_TYPE_INT16_ARRAY,
449 	DATA_TYPE_UINT16_ARRAY,
450 	DATA_TYPE_INT32_ARRAY,
451 	DATA_TYPE_UINT32_ARRAY,
452 	DATA_TYPE_INT64_ARRAY,
453 	DATA_TYPE_UINT64_ARRAY,
454 	DATA_TYPE_STRING_ARRAY,
455 	DATA_TYPE_HRTIME,
456 	DATA_TYPE_NVLIST,
457 	DATA_TYPE_NVLIST_ARRAY,
458 	DATA_TYPE_BOOLEAN_VALUE,
459 	DATA_TYPE_INT8,
460 	DATA_TYPE_UINT8,
461 	DATA_TYPE_BOOLEAN_ARRAY,
462 	DATA_TYPE_INT8_ARRAY,
463 	DATA_TYPE_UINT8_ARRAY
464 } data_type_t;
465 
466 /*
467  * On-disk version number.
468  */
469 #define	SPA_VERSION_1			1ULL
470 #define	SPA_VERSION_2			2ULL
471 #define	SPA_VERSION_3			3ULL
472 #define	SPA_VERSION_4			4ULL
473 #define	SPA_VERSION_5			5ULL
474 #define	SPA_VERSION_6			6ULL
475 #define	SPA_VERSION_7			7ULL
476 #define	SPA_VERSION_8			8ULL
477 #define	SPA_VERSION_9			9ULL
478 #define	SPA_VERSION_10			10ULL
479 #define	SPA_VERSION_11			11ULL
480 #define	SPA_VERSION_12			12ULL
481 #define	SPA_VERSION_13			13ULL
482 /*
483  * When bumping up SPA_VERSION, make sure GRUB ZFS understand the on-disk
484  * format change. Go to usr/src/grub/grub-0.95/stage2/{zfs-include/, fsys_zfs*},
485  * and do the appropriate changes.
486  */
487 #define	SPA_VERSION			SPA_VERSION_13
488 #define	SPA_VERSION_STRING		"13"
489 
490 /*
491  * Symbolic names for the changes that caused a SPA_VERSION switch.
492  * Used in the code when checking for presence or absence of a feature.
493  * Feel free to define multiple symbolic names for each version if there
494  * were multiple changes to on-disk structures during that version.
495  *
496  * NOTE: When checking the current SPA_VERSION in your code, be sure
497  *       to use spa_version() since it reports the version of the
498  *       last synced uberblock.  Checking the in-flight version can
499  *       be dangerous in some cases.
500  */
501 #define	SPA_VERSION_INITIAL		SPA_VERSION_1
502 #define	SPA_VERSION_DITTO_BLOCKS	SPA_VERSION_2
503 #define	SPA_VERSION_SPARES		SPA_VERSION_3
504 #define	SPA_VERSION_RAID6		SPA_VERSION_3
505 #define	SPA_VERSION_BPLIST_ACCOUNT	SPA_VERSION_3
506 #define	SPA_VERSION_RAIDZ_DEFLATE	SPA_VERSION_3
507 #define	SPA_VERSION_DNODE_BYTES		SPA_VERSION_3
508 #define	SPA_VERSION_ZPOOL_HISTORY	SPA_VERSION_4
509 #define	SPA_VERSION_GZIP_COMPRESSION	SPA_VERSION_5
510 #define	SPA_VERSION_BOOTFS		SPA_VERSION_6
511 #define	SPA_VERSION_SLOGS		SPA_VERSION_7
512 #define	SPA_VERSION_DELEGATED_PERMS	SPA_VERSION_8
513 #define	SPA_VERSION_FUID		SPA_VERSION_9
514 #define	SPA_VERSION_REFRESERVATION	SPA_VERSION_9
515 #define	SPA_VERSION_REFQUOTA		SPA_VERSION_9
516 #define	SPA_VERSION_UNIQUE_ACCURATE	SPA_VERSION_9
517 #define	SPA_VERSION_L2CACHE		SPA_VERSION_10
518 #define	SPA_VERSION_NEXT_CLONES		SPA_VERSION_11
519 #define	SPA_VERSION_ORIGIN		SPA_VERSION_11
520 #define	SPA_VERSION_DSL_SCRUB		SPA_VERSION_11
521 #define	SPA_VERSION_SNAP_PROPS		SPA_VERSION_12
522 #define	SPA_VERSION_USED_BREAKDOWN	SPA_VERSION_13
523 
524 /*
525  * The following are configuration names used in the nvlist describing a pool's
526  * configuration.
527  */
528 #define	ZPOOL_CONFIG_VERSION		"version"
529 #define	ZPOOL_CONFIG_POOL_NAME		"name"
530 #define	ZPOOL_CONFIG_POOL_STATE		"state"
531 #define	ZPOOL_CONFIG_POOL_TXG		"txg"
532 #define	ZPOOL_CONFIG_POOL_GUID		"pool_guid"
533 #define	ZPOOL_CONFIG_CREATE_TXG		"create_txg"
534 #define	ZPOOL_CONFIG_TOP_GUID		"top_guid"
535 #define	ZPOOL_CONFIG_VDEV_TREE		"vdev_tree"
536 #define	ZPOOL_CONFIG_TYPE		"type"
537 #define	ZPOOL_CONFIG_CHILDREN		"children"
538 #define	ZPOOL_CONFIG_ID			"id"
539 #define	ZPOOL_CONFIG_GUID		"guid"
540 #define	ZPOOL_CONFIG_PATH		"path"
541 #define	ZPOOL_CONFIG_DEVID		"devid"
542 #define	ZPOOL_CONFIG_METASLAB_ARRAY	"metaslab_array"
543 #define	ZPOOL_CONFIG_METASLAB_SHIFT	"metaslab_shift"
544 #define	ZPOOL_CONFIG_ASHIFT		"ashift"
545 #define	ZPOOL_CONFIG_ASIZE		"asize"
546 #define	ZPOOL_CONFIG_DTL		"DTL"
547 #define	ZPOOL_CONFIG_STATS		"stats"
548 #define	ZPOOL_CONFIG_WHOLE_DISK		"whole_disk"
549 #define	ZPOOL_CONFIG_OFFLINE		"offline"
550 #define	ZPOOL_CONFIG_ERRCOUNT		"error_count"
551 #define	ZPOOL_CONFIG_NOT_PRESENT	"not_present"
552 #define	ZPOOL_CONFIG_SPARES		"spares"
553 #define	ZPOOL_CONFIG_IS_SPARE		"is_spare"
554 #define	ZPOOL_CONFIG_NPARITY		"nparity"
555 #define	ZPOOL_CONFIG_HOSTID		"hostid"
556 #define	ZPOOL_CONFIG_HOSTNAME		"hostname"
557 #define	ZPOOL_CONFIG_TIMESTAMP		"timestamp" /* not stored on disk */
558 
559 #define	VDEV_TYPE_ROOT			"root"
560 #define	VDEV_TYPE_MIRROR		"mirror"
561 #define	VDEV_TYPE_REPLACING		"replacing"
562 #define	VDEV_TYPE_RAIDZ			"raidz"
563 #define	VDEV_TYPE_DISK			"disk"
564 #define	VDEV_TYPE_FILE			"file"
565 #define	VDEV_TYPE_MISSING		"missing"
566 #define	VDEV_TYPE_SPARE			"spare"
567 
568 /*
569  * This is needed in userland to report the minimum necessary device size.
570  */
571 #define	SPA_MINDEVSIZE		(64ULL << 20)
572 
573 /*
574  * The location of the pool configuration repository, shared between kernel and
575  * userland.
576  */
577 #define	ZPOOL_CACHE_DIR		"/boot/zfs"
578 #define	ZPOOL_CACHE_FILE	"zpool.cache"
579 #define	ZPOOL_CACHE_TMP		".zpool.cache"
580 
581 #define	ZPOOL_CACHE		ZPOOL_CACHE_DIR "/" ZPOOL_CACHE_FILE
582 
583 /*
584  * vdev states are ordered from least to most healthy.
585  * A vdev that's CANT_OPEN or below is considered unusable.
586  */
587 typedef enum vdev_state {
588 	VDEV_STATE_UNKNOWN = 0,	/* Uninitialized vdev			*/
589 	VDEV_STATE_CLOSED,	/* Not currently open			*/
590 	VDEV_STATE_OFFLINE,	/* Not allowed to open			*/
591 	VDEV_STATE_CANT_OPEN,	/* Tried to open, but failed		*/
592 	VDEV_STATE_DEGRADED,	/* Replicated vdev with unhealthy kids	*/
593 	VDEV_STATE_HEALTHY	/* Presumed good			*/
594 } vdev_state_t;
595 
596 /*
597  * vdev aux states.  When a vdev is in the CANT_OPEN state, the aux field
598  * of the vdev stats structure uses these constants to distinguish why.
599  */
600 typedef enum vdev_aux {
601 	VDEV_AUX_NONE,		/* no error				*/
602 	VDEV_AUX_OPEN_FAILED,	/* ldi_open_*() or vn_open() failed	*/
603 	VDEV_AUX_CORRUPT_DATA,	/* bad label or disk contents		*/
604 	VDEV_AUX_NO_REPLICAS,	/* insufficient number of replicas	*/
605 	VDEV_AUX_BAD_GUID_SUM,	/* vdev guid sum doesn't match		*/
606 	VDEV_AUX_TOO_SMALL,	/* vdev size is too small		*/
607 	VDEV_AUX_BAD_LABEL,	/* the label is OK but invalid		*/
608 	VDEV_AUX_VERSION_NEWER,	/* on-disk version is too new		*/
609 	VDEV_AUX_VERSION_OLDER,	/* on-disk version is too old		*/
610 	VDEV_AUX_SPARED		/* hot spare used in another pool	*/
611 } vdev_aux_t;
612 
613 /*
614  * pool state.  The following states are written to disk as part of the normal
615  * SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE.  The remaining states are
616  * software abstractions used at various levels to communicate pool state.
617  */
618 typedef enum pool_state {
619 	POOL_STATE_ACTIVE = 0,		/* In active use		*/
620 	POOL_STATE_EXPORTED,		/* Explicitly exported		*/
621 	POOL_STATE_DESTROYED,		/* Explicitly destroyed		*/
622 	POOL_STATE_SPARE,		/* Reserved for hot spare use	*/
623 	POOL_STATE_UNINITIALIZED,	/* Internal spa_t state		*/
624 	POOL_STATE_UNAVAIL,		/* Internal libzfs state	*/
625 	POOL_STATE_POTENTIALLY_ACTIVE	/* Internal libzfs state	*/
626 } pool_state_t;
627 
628 /*
629  * The uberblock version is incremented whenever an incompatible on-disk
630  * format change is made to the SPA, DMU, or ZAP.
631  *
632  * Note: the first two fields should never be moved.  When a storage pool
633  * is opened, the uberblock must be read off the disk before the version
634  * can be checked.  If the ub_version field is moved, we may not detect
635  * version mismatch.  If the ub_magic field is moved, applications that
636  * expect the magic number in the first word won't work.
637  */
638 #define	UBERBLOCK_MAGIC		0x00bab10c		/* oo-ba-bloc!	*/
639 #define	UBERBLOCK_SHIFT		10			/* up to 1K	*/
640 
641 struct uberblock {
642 	uint64_t	ub_magic;	/* UBERBLOCK_MAGIC		*/
643 	uint64_t	ub_version;	/* SPA_VERSION			*/
644 	uint64_t	ub_txg;		/* txg of last sync		*/
645 	uint64_t	ub_guid_sum;	/* sum of all vdev guids	*/
646 	uint64_t	ub_timestamp;	/* UTC time of last sync	*/
647 	blkptr_t	ub_rootbp;	/* MOS objset_phys_t		*/
648 };
649 
650 /*
651  * Flags.
652  */
653 #define	DNODE_MUST_BE_ALLOCATED	1
654 #define	DNODE_MUST_BE_FREE	2
655 
656 /*
657  * Fixed constants.
658  */
659 #define	DNODE_SHIFT		9	/* 512 bytes */
660 #define	DN_MIN_INDBLKSHIFT	10	/* 1k */
661 #define	DN_MAX_INDBLKSHIFT	14	/* 16k */
662 #define	DNODE_BLOCK_SHIFT	14	/* 16k */
663 #define	DNODE_CORE_SIZE		64	/* 64 bytes for dnode sans blkptrs */
664 #define	DN_MAX_OBJECT_SHIFT	48	/* 256 trillion (zfs_fid_t limit) */
665 #define	DN_MAX_OFFSET_SHIFT	64	/* 2^64 bytes in a dnode */
666 
667 /*
668  * Derived constants.
669  */
670 #define	DNODE_SIZE	(1 << DNODE_SHIFT)
671 #define	DN_MAX_NBLKPTR	((DNODE_SIZE - DNODE_CORE_SIZE) >> SPA_BLKPTRSHIFT)
672 #define	DN_MAX_BONUSLEN	(DNODE_SIZE - DNODE_CORE_SIZE - (1 << SPA_BLKPTRSHIFT))
673 #define	DN_MAX_OBJECT	(1ULL << DN_MAX_OBJECT_SHIFT)
674 
675 #define	DNODES_PER_BLOCK_SHIFT	(DNODE_BLOCK_SHIFT - DNODE_SHIFT)
676 #define	DNODES_PER_BLOCK	(1ULL << DNODES_PER_BLOCK_SHIFT)
677 #define	DNODES_PER_LEVEL_SHIFT	(DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT)
678 
679 /* The +2 here is a cheesy way to round up */
680 #define	DN_MAX_LEVELS	(2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \
681 	(DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT)))
682 
683 #define	DN_BONUS(dnp)	((void*)((dnp)->dn_bonus + \
684 	(((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t))))
685 
686 #define	DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \
687 	(dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT)
688 
689 #define	EPB(blkshift, typeshift)	(1 << (blkshift - typeshift))
690 
691 /* Is dn_used in bytes?  if not, it's in multiples of SPA_MINBLOCKSIZE */
692 #define	DNODE_FLAG_USED_BYTES	(1<<0)
693 
694 typedef struct dnode_phys {
695 	uint8_t dn_type;		/* dmu_object_type_t */
696 	uint8_t dn_indblkshift;		/* ln2(indirect block size) */
697 	uint8_t dn_nlevels;		/* 1=dn_blkptr->data blocks */
698 	uint8_t dn_nblkptr;		/* length of dn_blkptr */
699 	uint8_t dn_bonustype;		/* type of data in bonus buffer */
700 	uint8_t	dn_checksum;		/* ZIO_CHECKSUM type */
701 	uint8_t	dn_compress;		/* ZIO_COMPRESS type */
702 	uint8_t dn_flags;		/* DNODE_FLAG_* */
703 	uint16_t dn_datablkszsec;	/* data block size in 512b sectors */
704 	uint16_t dn_bonuslen;		/* length of dn_bonus */
705 	uint8_t dn_pad2[4];
706 
707 	/* accounting is protected by dn_dirty_mtx */
708 	uint64_t dn_maxblkid;		/* largest allocated block ID */
709 	uint64_t dn_used;		/* bytes (or sectors) of disk space */
710 
711 	uint64_t dn_pad3[4];
712 
713 	blkptr_t dn_blkptr[1];
714 	uint8_t dn_bonus[DN_MAX_BONUSLEN];
715 } dnode_phys_t;
716 
717 typedef enum dmu_object_type {
718 	DMU_OT_NONE,
719 	/* general: */
720 	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
721 	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
722 	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
723 	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
724 	DMU_OT_BPLIST,			/* UINT64 */
725 	DMU_OT_BPLIST_HDR,		/* UINT64 */
726 	/* spa: */
727 	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
728 	DMU_OT_SPACE_MAP,		/* UINT64 */
729 	/* zil: */
730 	DMU_OT_INTENT_LOG,		/* UINT64 */
731 	/* dmu: */
732 	DMU_OT_DNODE,			/* DNODE */
733 	DMU_OT_OBJSET,			/* OBJSET */
734 	/* dsl: */
735 	DMU_OT_DSL_DIR,			/* UINT64 */
736 	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
737 	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
738 	DMU_OT_DSL_PROPS,		/* ZAP */
739 	DMU_OT_DSL_DATASET,		/* UINT64 */
740 	/* zpl: */
741 	DMU_OT_ZNODE,			/* ZNODE */
742 	DMU_OT_ACL,			/* ACL */
743 	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
744 	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
745 	DMU_OT_MASTER_NODE,		/* ZAP */
746 	DMU_OT_UNLINKED_SET,		/* ZAP */
747 	/* zvol: */
748 	DMU_OT_ZVOL,			/* UINT8 */
749 	DMU_OT_ZVOL_PROP,		/* ZAP */
750 	/* other; for testing only! */
751 	DMU_OT_PLAIN_OTHER,		/* UINT8 */
752 	DMU_OT_UINT64_OTHER,		/* UINT64 */
753 	DMU_OT_ZAP_OTHER,		/* ZAP */
754 	/* new object types: */
755 	DMU_OT_ERROR_LOG,		/* ZAP */
756 	DMU_OT_SPA_HISTORY,		/* UINT8 */
757 	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
758 	DMU_OT_POOL_PROPS,		/* ZAP */
759 
760 	DMU_OT_NUMTYPES
761 } dmu_object_type_t;
762 
763 typedef enum dmu_objset_type {
764 	DMU_OST_NONE,
765 	DMU_OST_META,
766 	DMU_OST_ZFS,
767 	DMU_OST_ZVOL,
768 	DMU_OST_OTHER,			/* For testing only! */
769 	DMU_OST_ANY,			/* Be careful! */
770 	DMU_OST_NUMTYPES
771 } dmu_objset_type_t;
772 
773 /*
774  * Intent log header - this on disk structure holds fields to manage
775  * the log.  All fields are 64 bit to easily handle cross architectures.
776  */
777 typedef struct zil_header {
778 	uint64_t zh_claim_txg;	/* txg in which log blocks were claimed */
779 	uint64_t zh_replay_seq;	/* highest replayed sequence number */
780 	blkptr_t zh_log;	/* log chain */
781 	uint64_t zh_claim_seq;	/* highest claimed sequence number */
782 	uint64_t zh_pad[5];
783 } zil_header_t;
784 
785 typedef struct objset_phys {
786 	dnode_phys_t os_meta_dnode;
787 	zil_header_t os_zil_header;
788 	uint64_t os_type;
789 	char os_pad[1024 - sizeof (dnode_phys_t) - sizeof (zil_header_t) -
790 	    sizeof (uint64_t)];
791 } objset_phys_t;
792 
793 typedef struct dsl_dir_phys {
794 	uint64_t dd_creation_time; /* not actually used */
795 	uint64_t dd_head_dataset_obj;
796 	uint64_t dd_parent_obj;
797 	uint64_t dd_clone_parent_obj;
798 	uint64_t dd_child_dir_zapobj;
799 	/*
800 	 * how much space our children are accounting for; for leaf
801 	 * datasets, == physical space used by fs + snaps
802 	 */
803 	uint64_t dd_used_bytes;
804 	uint64_t dd_compressed_bytes;
805 	uint64_t dd_uncompressed_bytes;
806 	/* Administrative quota setting */
807 	uint64_t dd_quota;
808 	/* Administrative reservation setting */
809 	uint64_t dd_reserved;
810 	uint64_t dd_props_zapobj;
811 	uint64_t dd_pad[21]; /* pad out to 256 bytes for good measure */
812 } dsl_dir_phys_t;
813 
814 typedef struct dsl_dataset_phys {
815 	uint64_t ds_dir_obj;
816 	uint64_t ds_prev_snap_obj;
817 	uint64_t ds_prev_snap_txg;
818 	uint64_t ds_next_snap_obj;
819 	uint64_t ds_snapnames_zapobj;	/* zap obj of snaps; ==0 for snaps */
820 	uint64_t ds_num_children;	/* clone/snap children; ==0 for head */
821 	uint64_t ds_creation_time;	/* seconds since 1970 */
822 	uint64_t ds_creation_txg;
823 	uint64_t ds_deadlist_obj;
824 	uint64_t ds_used_bytes;
825 	uint64_t ds_compressed_bytes;
826 	uint64_t ds_uncompressed_bytes;
827 	uint64_t ds_unique_bytes;	/* only relevant to snapshots */
828 	/*
829 	 * The ds_fsid_guid is a 56-bit ID that can change to avoid
830 	 * collisions.  The ds_guid is a 64-bit ID that will never
831 	 * change, so there is a small probability that it will collide.
832 	 */
833 	uint64_t ds_fsid_guid;
834 	uint64_t ds_guid;
835 	uint64_t ds_flags;
836 	blkptr_t ds_bp;
837 	uint64_t ds_pad[8]; /* pad out to 320 bytes for good measure */
838 } dsl_dataset_phys_t;
839 
840 /*
841  * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
842  */
843 #define	DMU_POOL_DIRECTORY_OBJECT	1
844 #define	DMU_POOL_CONFIG			"config"
845 #define	DMU_POOL_ROOT_DATASET		"root_dataset"
846 #define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
847 #define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
848 #define	DMU_POOL_ERRLOG_LAST		"errlog_last"
849 #define	DMU_POOL_SPARES			"spares"
850 #define	DMU_POOL_DEFLATE		"deflate"
851 #define	DMU_POOL_HISTORY		"history"
852 #define	DMU_POOL_PROPS			"pool_props"
853 
854 #define	ZAP_MAGIC 0x2F52AB2ABULL
855 
856 #define	FZAP_BLOCK_SHIFT(zap)	((zap)->zap_block_shift)
857 
858 #define	ZAP_MAXCD		(uint32_t)(-1)
859 #define	ZAP_HASHBITS		28
860 #define	MZAP_ENT_LEN		64
861 #define	MZAP_NAME_LEN		(MZAP_ENT_LEN - 8 - 4 - 2)
862 #define	MZAP_MAX_BLKSHIFT	SPA_MAXBLOCKSHIFT
863 #define	MZAP_MAX_BLKSZ		(1 << MZAP_MAX_BLKSHIFT)
864 
865 typedef struct mzap_ent_phys {
866 	uint64_t mze_value;
867 	uint32_t mze_cd;
868 	uint16_t mze_pad;	/* in case we want to chain them someday */
869 	char mze_name[MZAP_NAME_LEN];
870 } mzap_ent_phys_t;
871 
872 typedef struct mzap_phys {
873 	uint64_t mz_block_type;	/* ZBT_MICRO */
874 	uint64_t mz_salt;
875 	uint64_t mz_pad[6];
876 	mzap_ent_phys_t mz_chunk[1];
877 	/* actually variable size depending on block size */
878 } mzap_phys_t;
879 
880 /*
881  * The (fat) zap is stored in one object. It is an array of
882  * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of:
883  *
884  * ptrtbl fits in first block:
885  * 	[zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ...
886  *
887  * ptrtbl too big for first block:
888  * 	[zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ...
889  *
890  */
891 
892 #define	ZBT_LEAF		((1ULL << 63) + 0)
893 #define	ZBT_HEADER		((1ULL << 63) + 1)
894 #define	ZBT_MICRO		((1ULL << 63) + 3)
895 /* any other values are ptrtbl blocks */
896 
897 /*
898  * the embedded pointer table takes up half a block:
899  * block size / entry size (2^3) / 2
900  */
901 #define	ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1)
902 
903 /*
904  * The embedded pointer table starts half-way through the block.  Since
905  * the pointer table itself is half the block, it starts at (64-bit)
906  * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)).
907  */
908 #define	ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \
909 	((uint64_t *)(zap)->zap_phys) \
910 	[(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))]
911 
912 /*
913  * TAKE NOTE:
914  * If zap_phys_t is modified, zap_byteswap() must be modified.
915  */
916 typedef struct zap_phys {
917 	uint64_t zap_block_type;	/* ZBT_HEADER */
918 	uint64_t zap_magic;		/* ZAP_MAGIC */
919 
920 	struct zap_table_phys {
921 		uint64_t zt_blk;	/* starting block number */
922 		uint64_t zt_numblks;	/* number of blocks */
923 		uint64_t zt_shift;	/* bits to index it */
924 		uint64_t zt_nextblk;	/* next (larger) copy start block */
925 		uint64_t zt_blks_copied; /* number source blocks copied */
926 	} zap_ptrtbl;
927 
928 	uint64_t zap_freeblk;		/* the next free block */
929 	uint64_t zap_num_leafs;		/* number of leafs */
930 	uint64_t zap_num_entries;	/* number of entries */
931 	uint64_t zap_salt;		/* salt to stir into hash function */
932 	/*
933 	 * This structure is followed by padding, and then the embedded
934 	 * pointer table.  The embedded pointer table takes up second
935 	 * half of the block.  It is accessed using the
936 	 * ZAP_EMBEDDED_PTRTBL_ENT() macro.
937 	 */
938 } zap_phys_t;
939 
940 typedef struct zap_table_phys zap_table_phys_t;
941 
942 typedef struct fat_zap {
943 	int zap_block_shift;			/* block size shift */
944 	zap_phys_t *zap_phys;
945 } fat_zap_t;
946 
947 #define	ZAP_LEAF_MAGIC 0x2AB1EAF
948 
949 /* chunk size = 24 bytes */
950 #define	ZAP_LEAF_CHUNKSIZE 24
951 
952 /*
953  * The amount of space available for chunks is:
954  * block size (1<<l->l_bs) - hash entry size (2) * number of hash
955  * entries - header space (2*chunksize)
956  */
957 #define	ZAP_LEAF_NUMCHUNKS(l) \
958 	(((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \
959 	ZAP_LEAF_CHUNKSIZE - 2)
960 
961 /*
962  * The amount of space within the chunk available for the array is:
963  * chunk size - space for type (1) - space for next pointer (2)
964  */
965 #define	ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3)
966 
967 #define	ZAP_LEAF_ARRAY_NCHUNKS(bytes) \
968 	(((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES)
969 
970 /*
971  * Low water mark:  when there are only this many chunks free, start
972  * growing the ptrtbl.  Ideally, this should be larger than a
973  * "reasonably-sized" entry.  20 chunks is more than enough for the
974  * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value),
975  * while still being only around 3% for 16k blocks.
976  */
977 #define	ZAP_LEAF_LOW_WATER (20)
978 
979 /*
980  * The leaf hash table has block size / 2^5 (32) number of entries,
981  * which should be more than enough for the maximum number of entries,
982  * which is less than block size / CHUNKSIZE (24) / minimum number of
983  * chunks per entry (3).
984  */
985 #define	ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5)
986 #define	ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l))
987 
988 /*
989  * The chunks start immediately after the hash table.  The end of the
990  * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a
991  * chunk_t.
992  */
993 #define	ZAP_LEAF_CHUNK(l, idx) \
994 	((zap_leaf_chunk_t *) \
995 	((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx]
996 #define	ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry)
997 
998 typedef enum zap_chunk_type {
999 	ZAP_CHUNK_FREE = 253,
1000 	ZAP_CHUNK_ENTRY = 252,
1001 	ZAP_CHUNK_ARRAY = 251,
1002 	ZAP_CHUNK_TYPE_MAX = 250
1003 } zap_chunk_type_t;
1004 
1005 /*
1006  * TAKE NOTE:
1007  * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified.
1008  */
1009 typedef struct zap_leaf_phys {
1010 	struct zap_leaf_header {
1011 		uint64_t lh_block_type;		/* ZBT_LEAF */
1012 		uint64_t lh_pad1;
1013 		uint64_t lh_prefix;		/* hash prefix of this leaf */
1014 		uint32_t lh_magic;		/* ZAP_LEAF_MAGIC */
1015 		uint16_t lh_nfree;		/* number free chunks */
1016 		uint16_t lh_nentries;		/* number of entries */
1017 		uint16_t lh_prefix_len;		/* num bits used to id this */
1018 
1019 /* above is accessable to zap, below is zap_leaf private */
1020 
1021 		uint16_t lh_freelist;		/* chunk head of free list */
1022 		uint8_t lh_pad2[12];
1023 	} l_hdr; /* 2 24-byte chunks */
1024 
1025 	/*
1026 	 * The header is followed by a hash table with
1027 	 * ZAP_LEAF_HASH_NUMENTRIES(zap) entries.  The hash table is
1028 	 * followed by an array of ZAP_LEAF_NUMCHUNKS(zap)
1029 	 * zap_leaf_chunk structures.  These structures are accessed
1030 	 * with the ZAP_LEAF_CHUNK() macro.
1031 	 */
1032 
1033 	uint16_t l_hash[1];
1034 } zap_leaf_phys_t;
1035 
1036 typedef union zap_leaf_chunk {
1037 	struct zap_leaf_entry {
1038 		uint8_t le_type; 		/* always ZAP_CHUNK_ENTRY */
1039 		uint8_t le_int_size;		/* size of ints */
1040 		uint16_t le_next;		/* next entry in hash chain */
1041 		uint16_t le_name_chunk;		/* first chunk of the name */
1042 		uint16_t le_name_length;	/* bytes in name, incl null */
1043 		uint16_t le_value_chunk;	/* first chunk of the value */
1044 		uint16_t le_value_length;	/* value length in ints */
1045 		uint32_t le_cd;			/* collision differentiator */
1046 		uint64_t le_hash;		/* hash value of the name */
1047 	} l_entry;
1048 	struct zap_leaf_array {
1049 		uint8_t la_type;		/* always ZAP_CHUNK_ARRAY */
1050 		uint8_t la_array[ZAP_LEAF_ARRAY_BYTES];
1051 		uint16_t la_next;		/* next blk or CHAIN_END */
1052 	} l_array;
1053 	struct zap_leaf_free {
1054 		uint8_t lf_type;		/* always ZAP_CHUNK_FREE */
1055 		uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES];
1056 		uint16_t lf_next;	/* next in free list, or CHAIN_END */
1057 	} l_free;
1058 } zap_leaf_chunk_t;
1059 
1060 typedef struct zap_leaf {
1061 	int l_bs;			/* block size shift */
1062 	zap_leaf_phys_t *l_phys;
1063 } zap_leaf_t;
1064 
1065 /*
1066  * Define special zfs pflags
1067  */
1068 #define	ZFS_XATTR	0x1		/* is an extended attribute */
1069 #define	ZFS_INHERIT_ACE	0x2		/* ace has inheritable ACEs */
1070 #define	ZFS_ACL_TRIVIAL 0x4		/* files ACL is trivial */
1071 
1072 #define	MASTER_NODE_OBJ	1
1073 
1074 /*
1075  * special attributes for master node.
1076  */
1077 
1078 #define	ZFS_FSID		"FSID"
1079 #define	ZFS_UNLINKED_SET	"DELETE_QUEUE"
1080 #define	ZFS_ROOT_OBJ		"ROOT"
1081 #define	ZPL_VERSION_OBJ		"VERSION"
1082 #define	ZFS_PROP_BLOCKPERPAGE	"BLOCKPERPAGE"
1083 #define	ZFS_PROP_NOGROWBLOCKS	"NOGROWBLOCKS"
1084 
1085 #define	ZFS_FLAG_BLOCKPERPAGE	0x1
1086 #define	ZFS_FLAG_NOGROWBLOCKS	0x2
1087 
1088 /*
1089  * ZPL version - rev'd whenever an incompatible on-disk format change
1090  * occurs.  Independent of SPA/DMU/ZAP versioning.
1091  */
1092 
1093 #define	ZPL_VERSION		1ULL
1094 
1095 /*
1096  * The directory entry has the type (currently unused on Solaris) in the
1097  * top 4 bits, and the object number in the low 48 bits.  The "middle"
1098  * 12 bits are unused.
1099  */
1100 #define	ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4)
1101 #define	ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48)
1102 #define	ZFS_DIRENT_MAKE(type, obj) (((uint64_t)type << 60) | obj)
1103 
1104 typedef struct ace {
1105 	uid_t		a_who;		/* uid or gid */
1106 	uint32_t	a_access_mask;	/* read,write,... */
1107 	uint16_t	a_flags;	/* see below */
1108 	uint16_t	a_type;		/* allow or deny */
1109 } ace_t;
1110 
1111 #define ACE_SLOT_CNT	6
1112 
1113 typedef struct zfs_znode_acl {
1114 	uint64_t	z_acl_extern_obj;	  /* ext acl pieces */
1115 	uint32_t	z_acl_count;		  /* Number of ACEs */
1116 	uint16_t	z_acl_version;		  /* acl version */
1117 	uint16_t	z_acl_pad;		  /* pad */
1118 	ace_t		z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */
1119 } zfs_znode_acl_t;
1120 
1121 /*
1122  * This is the persistent portion of the znode.  It is stored
1123  * in the "bonus buffer" of the file.  Short symbolic links
1124  * are also stored in the bonus buffer.
1125  */
1126 typedef struct znode_phys {
1127 	uint64_t zp_atime[2];		/*  0 - last file access time */
1128 	uint64_t zp_mtime[2];		/* 16 - last file modification time */
1129 	uint64_t zp_ctime[2];		/* 32 - last file change time */
1130 	uint64_t zp_crtime[2];		/* 48 - creation time */
1131 	uint64_t zp_gen;		/* 64 - generation (txg of creation) */
1132 	uint64_t zp_mode;		/* 72 - file mode bits */
1133 	uint64_t zp_size;		/* 80 - size of file */
1134 	uint64_t zp_parent;		/* 88 - directory parent (`..') */
1135 	uint64_t zp_links;		/* 96 - number of links to file */
1136 	uint64_t zp_xattr;		/* 104 - DMU object for xattrs */
1137 	uint64_t zp_rdev;		/* 112 - dev_t for VBLK & VCHR files */
1138 	uint64_t zp_flags;		/* 120 - persistent flags */
1139 	uint64_t zp_uid;		/* 128 - file owner */
1140 	uint64_t zp_gid;		/* 136 - owning group */
1141 	uint64_t zp_pad[4];		/* 144 - future */
1142 	zfs_znode_acl_t zp_acl;		/* 176 - 263 ACL */
1143 	/*
1144 	 * Data may pad out any remaining bytes in the znode buffer, eg:
1145 	 *
1146 	 * |<---------------------- dnode_phys (512) ------------------------>|
1147 	 * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->|
1148 	 *			|<---- znode (264) ---->|<---- data (56) ---->|
1149 	 *
1150 	 * At present, we only use this space to store symbolic links.
1151 	 */
1152 } znode_phys_t;
1153 
1154 /*
1155  * In-core vdev representation.
1156  */
1157 struct vdev;
1158 typedef int vdev_phys_read_t(struct vdev *vdev, void *priv,
1159     off_t offset, void *buf, size_t bytes);
1160 typedef int vdev_read_t(struct vdev *vdev, const blkptr_t *bp,
1161     void *buf, off_t offset, size_t bytes);
1162 
1163 typedef STAILQ_HEAD(vdev_list, vdev) vdev_list_t;
1164 
1165 typedef struct vdev {
1166 	STAILQ_ENTRY(vdev) v_childlink;	/* link in parent's child list */
1167 	STAILQ_ENTRY(vdev) v_alllink;	/* link in global vdev list */
1168 	vdev_list_t	v_children;	/* children of this vdev */
1169 	char		*v_name;	/* vdev name */
1170 	uint64_t	v_guid;		/* vdev guid */
1171 	int		v_id;		/* index in parent */
1172 	int		v_ashift;	/* offset to block shift */
1173 	int		v_nparity;	/* # parity for raidz */
1174 	int		v_nchildren;	/* # children */
1175 	vdev_state_t	v_state;	/* current state */
1176 	vdev_phys_read_t *v_phys_read;	/* read from raw leaf vdev */
1177 	vdev_read_t	*v_read;	/* read from vdev */
1178 	void		*v_read_priv;	/* private data for read function */
1179 } vdev_t;
1180 
1181 /*
1182  * In-core pool representation.
1183  */
1184 typedef STAILQ_HEAD(spa_list, spa) spa_list_t;
1185 
1186 typedef struct spa {
1187 	STAILQ_ENTRY(spa) spa_link;	/* link in global pool list */
1188 	char		*spa_name;	/* pool name */
1189 	uint64_t	spa_guid;	/* pool guid */
1190 	uint64_t	spa_txg;	/* most recent transaction */
1191 	struct uberblock spa_uberblock;	/* best uberblock so far */
1192 	vdev_list_t	spa_vdevs;	/* list of all toplevel vdevs */
1193 	objset_phys_t	spa_mos;	/* MOS for this pool */
1194 	objset_phys_t	spa_root_objset; /* current mounted ZPL objset */
1195 } spa_t;
1196