spa.h revision b24ab6762772a3f6a89393947930c7fa61306783
1/*
2 *  GRUB  --  GRand Unified Bootloader
3 *  Copyright (C) 1999,2000,2001,2002,2003,2004  Free Software Foundation, Inc.
4 *
5 *  This program is free software; you can redistribute it and/or modify
6 *  it under the terms of the GNU General Public License as published by
7 *  the Free Software Foundation; either version 2 of the License, or
8 *  (at your option) any later version.
9 *
10 *  This program is distributed in the hope that it will be useful,
11 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
12 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
13 *  GNU General Public License for more details.
14 *
15 *  You should have received a copy of the GNU General Public License
16 *  along with this program; if not, write to the Free Software
17 *  Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
18 */
19/*
20 * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
21 * Use is subject to license terms.
22 */
23
24#ifndef _SYS_SPA_H
25#define	_SYS_SPA_H
26
27/*
28 * General-purpose 32-bit and 64-bit bitfield encodings.
29 */
30#define	BF32_DECODE(x, low, len)	P2PHASE((x) >> (low), 1U << (len))
31#define	BF64_DECODE(x, low, len)	P2PHASE((x) >> (low), 1ULL << (len))
32#define	BF32_ENCODE(x, low, len)	(P2PHASE((x), 1U << (len)) << (low))
33#define	BF64_ENCODE(x, low, len)	(P2PHASE((x), 1ULL << (len)) << (low))
34
35#define	BF32_GET(x, low, len)		BF32_DECODE(x, low, len)
36#define	BF64_GET(x, low, len)		BF64_DECODE(x, low, len)
37
38#define	BF32_SET(x, low, len, val)	\
39	((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len))
40#define	BF64_SET(x, low, len, val)	\
41	((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len))
42
43#define	BF32_GET_SB(x, low, len, shift, bias)	\
44	((BF32_GET(x, low, len) + (bias)) << (shift))
45#define	BF64_GET_SB(x, low, len, shift, bias)	\
46	((BF64_GET(x, low, len) + (bias)) << (shift))
47
48#define	BF32_SET_SB(x, low, len, shift, bias, val)	\
49	BF32_SET(x, low, len, ((val) >> (shift)) - (bias))
50#define	BF64_SET_SB(x, low, len, shift, bias, val)	\
51	BF64_SET(x, low, len, ((val) >> (shift)) - (bias))
52
53/*
54 * We currently support nine block sizes, from 512 bytes to 128K.
55 * We could go higher, but the benefits are near-zero and the cost
56 * of COWing a giant block to modify one byte would become excessive.
57 */
58#define	SPA_MINBLOCKSHIFT	9
59#define	SPA_MAXBLOCKSHIFT	17
60#define	SPA_MINBLOCKSIZE	(1ULL << SPA_MINBLOCKSHIFT)
61#define	SPA_MAXBLOCKSIZE	(1ULL << SPA_MAXBLOCKSHIFT)
62
63#define	SPA_BLOCKSIZES		(SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1)
64
65/*
66 * Size of block to hold the configuration data (a packed nvlist)
67 */
68#define	SPA_CONFIG_BLOCKSIZE	(1 << 14)
69
70/*
71 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
72 * The ASIZE encoding should be at least 64 times larger (6 more bits)
73 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
74 * overhead, three DVAs per bp, plus one more bit in case we do anything
75 * else that expands the ASIZE.
76 */
77#define	SPA_LSIZEBITS		16	/* LSIZE up to 32M (2^16 * 512)	*/
78#define	SPA_PSIZEBITS		16	/* PSIZE up to 32M (2^16 * 512)	*/
79#define	SPA_ASIZEBITS		24	/* ASIZE up to 64 times larger	*/
80
81/*
82 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
83 * The members of the dva_t should be considered opaque outside the SPA.
84 */
85typedef struct dva {
86	uint64_t	dva_word[2];
87} dva_t;
88
89/*
90 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
91 */
92typedef struct zio_cksum {
93	uint64_t	zc_word[4];
94} zio_cksum_t;
95
96/*
97 * Each block is described by its DVAs, time of birth, checksum, etc.
98 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
99 *
100 *	64	56	48	40	32	24	16	8	0
101 *	+-------+-------+-------+-------+-------+-------+-------+-------+
102 * 0	|		vdev1		| GRID  |	  ASIZE		|
103 *	+-------+-------+-------+-------+-------+-------+-------+-------+
104 * 1	|G|			 offset1				|
105 *	+-------+-------+-------+-------+-------+-------+-------+-------+
106 * 2	|		vdev2		| GRID  |	  ASIZE		|
107 *	+-------+-------+-------+-------+-------+-------+-------+-------+
108 * 3	|G|			 offset2				|
109 *	+-------+-------+-------+-------+-------+-------+-------+-------+
110 * 4	|		vdev3		| GRID  |	  ASIZE		|
111 *	+-------+-------+-------+-------+-------+-------+-------+-------+
112 * 5	|G|			 offset3				|
113 *	+-------+-------+-------+-------+-------+-------+-------+-------+
114 * 6	|BDX|lvl| type	| cksum | comp	|     PSIZE	|     LSIZE	|
115 *	+-------+-------+-------+-------+-------+-------+-------+-------+
116 * 7	|			padding					|
117 *	+-------+-------+-------+-------+-------+-------+-------+-------+
118 * 8	|			padding					|
119 *	+-------+-------+-------+-------+-------+-------+-------+-------+
120 * 9	|			physical birth txg			|
121 *	+-------+-------+-------+-------+-------+-------+-------+-------+
122 * a	|			logical birth txg			|
123 *	+-------+-------+-------+-------+-------+-------+-------+-------+
124 * b	|			fill count				|
125 *	+-------+-------+-------+-------+-------+-------+-------+-------+
126 * c	|			checksum[0]				|
127 *	+-------+-------+-------+-------+-------+-------+-------+-------+
128 * d	|			checksum[1]				|
129 *	+-------+-------+-------+-------+-------+-------+-------+-------+
130 * e	|			checksum[2]				|
131 *	+-------+-------+-------+-------+-------+-------+-------+-------+
132 * f	|			checksum[3]				|
133 *	+-------+-------+-------+-------+-------+-------+-------+-------+
134 *
135 * Legend:
136 *
137 * vdev		virtual device ID
138 * offset	offset into virtual device
139 * LSIZE	logical size
140 * PSIZE	physical size (after compression)
141 * ASIZE	allocated size (including RAID-Z parity and gang block headers)
142 * GRID		RAID-Z layout information (reserved for future use)
143 * cksum	checksum function
144 * comp		compression function
145 * G		gang block indicator
146 * B		byteorder (endianness)
147 * D		dedup
148 * X		unused
149 * lvl		level of indirection
150 * type		DMU object type
151 * phys birth	txg of block allocation; zero if same as logical birth txg
152 * log. birth	transaction group in which the block was logically born
153 * fill count	number of non-zero blocks under this bp
154 * checksum[4]	256-bit checksum of the data this bp describes
155 */
156#define	SPA_BLKPTRSHIFT	7		/* blkptr_t is 128 bytes	*/
157#define	SPA_DVAS_PER_BP	3		/* Number of DVAs in a bp	*/
158
159typedef struct blkptr {
160	dva_t		blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
161	uint64_t	blk_prop;	/* size, compression, type, etc	    */
162	uint64_t	blk_pad[2];	/* Extra space for the future	    */
163	uint64_t	blk_phys_birth;	/* txg when block was allocated	    */
164	uint64_t	blk_birth;	/* transaction group at birth	    */
165	uint64_t	blk_fill;	/* fill count			    */
166	zio_cksum_t	blk_cksum;	/* 256-bit checksum		    */
167} blkptr_t;
168
169/*
170 * Macros to get and set fields in a bp or DVA.
171 */
172#define	DVA_GET_ASIZE(dva)	\
173	BF64_GET_SB((dva)->dva_word[0], 0, 24, SPA_MINBLOCKSHIFT, 0)
174#define	DVA_SET_ASIZE(dva, x)	\
175	BF64_SET_SB((dva)->dva_word[0], 0, 24, SPA_MINBLOCKSHIFT, 0, x)
176
177#define	DVA_GET_GRID(dva)	BF64_GET((dva)->dva_word[0], 24, 8)
178#define	DVA_SET_GRID(dva, x)	BF64_SET((dva)->dva_word[0], 24, 8, x)
179
180#define	DVA_GET_VDEV(dva)	BF64_GET((dva)->dva_word[0], 32, 32)
181#define	DVA_SET_VDEV(dva, x)	BF64_SET((dva)->dva_word[0], 32, 32, x)
182
183#define	DVA_GET_OFFSET(dva)	\
184	BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
185#define	DVA_SET_OFFSET(dva, x)	\
186	BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
187
188#define	DVA_GET_GANG(dva)	BF64_GET((dva)->dva_word[1], 63, 1)
189#define	DVA_SET_GANG(dva, x)	BF64_SET((dva)->dva_word[1], 63, 1, x)
190
191#define	BP_GET_LSIZE(bp)	\
192	BF64_GET_SB((bp)->blk_prop, 0, 16, SPA_MINBLOCKSHIFT, 1)
193#define	BP_SET_LSIZE(bp, x)	\
194	BF64_SET_SB((bp)->blk_prop, 0, 16, SPA_MINBLOCKSHIFT, 1, x)
195
196#define	BP_GET_PSIZE(bp)	\
197	BF64_GET_SB((bp)->blk_prop, 16, 16, SPA_MINBLOCKSHIFT, 1)
198#define	BP_SET_PSIZE(bp, x)	\
199	BF64_SET_SB((bp)->blk_prop, 16, 16, SPA_MINBLOCKSHIFT, 1, x)
200
201#define	BP_GET_COMPRESS(bp)		BF64_GET((bp)->blk_prop, 32, 8)
202#define	BP_SET_COMPRESS(bp, x)		BF64_SET((bp)->blk_prop, 32, 8, x)
203
204#define	BP_GET_CHECKSUM(bp)		BF64_GET((bp)->blk_prop, 40, 8)
205#define	BP_SET_CHECKSUM(bp, x)		BF64_SET((bp)->blk_prop, 40, 8, x)
206
207#define	BP_GET_TYPE(bp)			BF64_GET((bp)->blk_prop, 48, 8)
208#define	BP_SET_TYPE(bp, x)		BF64_SET((bp)->blk_prop, 48, 8, x)
209
210#define	BP_GET_LEVEL(bp)		BF64_GET((bp)->blk_prop, 56, 5)
211#define	BP_SET_LEVEL(bp, x)		BF64_SET((bp)->blk_prop, 56, 5, x)
212
213#define	BP_GET_PROP_BIT_61(bp)		BF64_GET((bp)->blk_prop, 61, 1)
214#define	BP_SET_PROP_BIT_61(bp, x)	BF64_SET((bp)->blk_prop, 61, 1, x)
215
216#define	BP_GET_DEDUP(bp)		BF64_GET((bp)->blk_prop, 62, 1)
217#define	BP_SET_DEDUP(bp, x)		BF64_SET((bp)->blk_prop, 62, 1, x)
218
219#define	BP_GET_BYTEORDER(bp)		(0 - BF64_GET((bp)->blk_prop, 63, 1))
220#define	BP_SET_BYTEORDER(bp, x)		BF64_SET((bp)->blk_prop, 63, 1, x)
221
222#define	BP_PHYSICAL_BIRTH(bp)		\
223	((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
224
225#define	BP_SET_BIRTH(bp, logical, physical)	\
226{						\
227	(bp)->blk_birth = (logical);		\
228	(bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
229}
230
231#define	BP_GET_ASIZE(bp)	\
232	(DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
233		DVA_GET_ASIZE(&(bp)->blk_dva[2]))
234
235#define	BP_GET_UCSIZE(bp) \
236	((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \
237	BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp));
238
239#define	BP_GET_NDVAS(bp)	\
240	(!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
241	!!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
242	!!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
243
244#define	BP_COUNT_GANG(bp)	\
245	(DVA_GET_GANG(&(bp)->blk_dva[0]) + \
246	DVA_GET_GANG(&(bp)->blk_dva[1]) + \
247	DVA_GET_GANG(&(bp)->blk_dva[2]))
248
249#define	DVA_EQUAL(dva1, dva2)	\
250	((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
251	(dva1)->dva_word[0] == (dva2)->dva_word[0])
252
253#define	BP_EQUAL(bp1, bp2)	\
254	(BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) &&	\
255	DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) &&	\
256	DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) &&	\
257	DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
258
259#define	ZIO_CHECKSUM_EQUAL(zc1, zc2) \
260	(0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
261	((zc1).zc_word[1] - (zc2).zc_word[1]) | \
262	((zc1).zc_word[2] - (zc2).zc_word[2]) | \
263	((zc1).zc_word[3] - (zc2).zc_word[3])))
264
265#define	DVA_IS_VALID(dva)	(DVA_GET_ASIZE(dva) != 0)
266
267#define	ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3)	\
268{						\
269	(zcp)->zc_word[0] = w0;			\
270	(zcp)->zc_word[1] = w1;			\
271	(zcp)->zc_word[2] = w2;			\
272	(zcp)->zc_word[3] = w3;			\
273}
274
275#define	BP_IDENTITY(bp)		(&(bp)->blk_dva[0])
276#define	BP_IS_GANG(bp)		DVA_GET_GANG(BP_IDENTITY(bp))
277#define	BP_IS_HOLE(bp)		((bp)->blk_birth == 0)
278
279#define	BP_ZERO(bp)				\
280{						\
281	(bp)->blk_dva[0].dva_word[0] = 0;	\
282	(bp)->blk_dva[0].dva_word[1] = 0;	\
283	(bp)->blk_dva[1].dva_word[0] = 0;	\
284	(bp)->blk_dva[1].dva_word[1] = 0;	\
285	(bp)->blk_dva[2].dva_word[0] = 0;	\
286	(bp)->blk_dva[2].dva_word[1] = 0;	\
287	(bp)->blk_prop = 0;			\
288	(bp)->blk_pad[0] = 0;			\
289	(bp)->blk_pad[1] = 0;			\
290	(bp)->blk_phys_birth = 0;		\
291	(bp)->blk_birth = 0;			\
292	(bp)->blk_fill = 0;			\
293	ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0);	\
294}
295
296/*
297 * Note: the byteorder is either 0 or -1, both of which are palindromes.
298 * This simplifies the endianness handling a bit.
299 */
300#ifdef _BIG_ENDIAN
301#define	ZFS_HOST_BYTEORDER	(0ULL)
302#else
303#define	ZFS_HOST_BYTEORDER	(-1ULL)
304#endif
305
306#define	BP_SHOULD_BYTESWAP(bp)	(BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
307
308#define	BP_SPRINTF_LEN	320
309
310#endif	/* _SYS_SPA_H */
311