spa.h revision c3d26abc9ee97b4f60233556aadeb57e0bd30bb9
1/*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21/*
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
25 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
26 * Copyright 2013 Saso Kiselkov. All rights reserved.
27 * Copyright (c) 2014 Integros [integros.com]
28 */
29
30#ifndef _SYS_SPA_H
31#define	_SYS_SPA_H
32
33#include <sys/avl.h>
34#include <sys/zfs_context.h>
35#include <sys/nvpair.h>
36#include <sys/sysmacros.h>
37#include <sys/types.h>
38#include <sys/fs/zfs.h>
39
40#ifdef	__cplusplus
41extern "C" {
42#endif
43
44/*
45 * Forward references that lots of things need.
46 */
47typedef struct spa spa_t;
48typedef struct vdev vdev_t;
49typedef struct metaslab metaslab_t;
50typedef struct metaslab_group metaslab_group_t;
51typedef struct metaslab_class metaslab_class_t;
52typedef struct zio zio_t;
53typedef struct zilog zilog_t;
54typedef struct spa_aux_vdev spa_aux_vdev_t;
55typedef struct ddt ddt_t;
56typedef struct ddt_entry ddt_entry_t;
57struct dsl_pool;
58struct dsl_dataset;
59
60/*
61 * General-purpose 32-bit and 64-bit bitfield encodings.
62 */
63#define	BF32_DECODE(x, low, len)	P2PHASE((x) >> (low), 1U << (len))
64#define	BF64_DECODE(x, low, len)	P2PHASE((x) >> (low), 1ULL << (len))
65#define	BF32_ENCODE(x, low, len)	(P2PHASE((x), 1U << (len)) << (low))
66#define	BF64_ENCODE(x, low, len)	(P2PHASE((x), 1ULL << (len)) << (low))
67
68#define	BF32_GET(x, low, len)		BF32_DECODE(x, low, len)
69#define	BF64_GET(x, low, len)		BF64_DECODE(x, low, len)
70
71#define	BF32_SET(x, low, len, val) do { \
72	ASSERT3U(val, <, 1U << (len)); \
73	ASSERT3U(low + len, <=, 32); \
74	(x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \
75_NOTE(CONSTCOND) } while (0)
76
77#define	BF64_SET(x, low, len, val) do { \
78	ASSERT3U(val, <, 1ULL << (len)); \
79	ASSERT3U(low + len, <=, 64); \
80	((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \
81_NOTE(CONSTCOND) } while (0)
82
83#define	BF32_GET_SB(x, low, len, shift, bias)	\
84	((BF32_GET(x, low, len) + (bias)) << (shift))
85#define	BF64_GET_SB(x, low, len, shift, bias)	\
86	((BF64_GET(x, low, len) + (bias)) << (shift))
87
88#define	BF32_SET_SB(x, low, len, shift, bias, val) do { \
89	ASSERT(IS_P2ALIGNED(val, 1U << shift)); \
90	ASSERT3S((val) >> (shift), >=, bias); \
91	BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \
92_NOTE(CONSTCOND) } while (0)
93#define	BF64_SET_SB(x, low, len, shift, bias, val) do { \
94	ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \
95	ASSERT3S((val) >> (shift), >=, bias); \
96	BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \
97_NOTE(CONSTCOND) } while (0)
98
99/*
100 * We currently support block sizes from 512 bytes to 16MB.
101 * The benefits of larger blocks, and thus larger IO, need to be weighed
102 * against the cost of COWing a giant block to modify one byte, and the
103 * large latency of reading or writing a large block.
104 *
105 * Note that although blocks up to 16MB are supported, the recordsize
106 * property can not be set larger than zfs_max_recordsize (default 1MB).
107 * See the comment near zfs_max_recordsize in dsl_dataset.c for details.
108 *
109 * Note that although the LSIZE field of the blkptr_t can store sizes up
110 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to
111 * 32MB - 512 bytes.  Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
112 */
113#define	SPA_MINBLOCKSHIFT	9
114#define	SPA_OLD_MAXBLOCKSHIFT	17
115#define	SPA_MAXBLOCKSHIFT	24
116#define	SPA_MINBLOCKSIZE	(1ULL << SPA_MINBLOCKSHIFT)
117#define	SPA_OLD_MAXBLOCKSIZE	(1ULL << SPA_OLD_MAXBLOCKSHIFT)
118#define	SPA_MAXBLOCKSIZE	(1ULL << SPA_MAXBLOCKSHIFT)
119
120/*
121 * Size of block to hold the configuration data (a packed nvlist)
122 */
123#define	SPA_CONFIG_BLOCKSIZE	(1ULL << 14)
124
125/*
126 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
127 * The ASIZE encoding should be at least 64 times larger (6 more bits)
128 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
129 * overhead, three DVAs per bp, plus one more bit in case we do anything
130 * else that expands the ASIZE.
131 */
132#define	SPA_LSIZEBITS		16	/* LSIZE up to 32M (2^16 * 512)	*/
133#define	SPA_PSIZEBITS		16	/* PSIZE up to 32M (2^16 * 512)	*/
134#define	SPA_ASIZEBITS		24	/* ASIZE up to 64 times larger	*/
135
136/*
137 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
138 * The members of the dva_t should be considered opaque outside the SPA.
139 */
140typedef struct dva {
141	uint64_t	dva_word[2];
142} dva_t;
143
144/*
145 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
146 */
147typedef struct zio_cksum {
148	uint64_t	zc_word[4];
149} zio_cksum_t;
150
151/*
152 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept
153 * secret and is suitable for use in MAC algorithms as the key.
154 */
155typedef struct zio_cksum_salt {
156	uint8_t		zcs_bytes[32];
157} zio_cksum_salt_t;
158
159/*
160 * Each block is described by its DVAs, time of birth, checksum, etc.
161 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
162 *
163 *	64	56	48	40	32	24	16	8	0
164 *	+-------+-------+-------+-------+-------+-------+-------+-------+
165 * 0	|		vdev1		| GRID  |	  ASIZE		|
166 *	+-------+-------+-------+-------+-------+-------+-------+-------+
167 * 1	|G|			 offset1				|
168 *	+-------+-------+-------+-------+-------+-------+-------+-------+
169 * 2	|		vdev2		| GRID  |	  ASIZE		|
170 *	+-------+-------+-------+-------+-------+-------+-------+-------+
171 * 3	|G|			 offset2				|
172 *	+-------+-------+-------+-------+-------+-------+-------+-------+
173 * 4	|		vdev3		| GRID  |	  ASIZE		|
174 *	+-------+-------+-------+-------+-------+-------+-------+-------+
175 * 5	|G|			 offset3				|
176 *	+-------+-------+-------+-------+-------+-------+-------+-------+
177 * 6	|BDX|lvl| type	| cksum |E| comp|    PSIZE	|     LSIZE	|
178 *	+-------+-------+-------+-------+-------+-------+-------+-------+
179 * 7	|			padding					|
180 *	+-------+-------+-------+-------+-------+-------+-------+-------+
181 * 8	|			padding					|
182 *	+-------+-------+-------+-------+-------+-------+-------+-------+
183 * 9	|			physical birth txg			|
184 *	+-------+-------+-------+-------+-------+-------+-------+-------+
185 * a	|			logical birth txg			|
186 *	+-------+-------+-------+-------+-------+-------+-------+-------+
187 * b	|			fill count				|
188 *	+-------+-------+-------+-------+-------+-------+-------+-------+
189 * c	|			checksum[0]				|
190 *	+-------+-------+-------+-------+-------+-------+-------+-------+
191 * d	|			checksum[1]				|
192 *	+-------+-------+-------+-------+-------+-------+-------+-------+
193 * e	|			checksum[2]				|
194 *	+-------+-------+-------+-------+-------+-------+-------+-------+
195 * f	|			checksum[3]				|
196 *	+-------+-------+-------+-------+-------+-------+-------+-------+
197 *
198 * Legend:
199 *
200 * vdev		virtual device ID
201 * offset	offset into virtual device
202 * LSIZE	logical size
203 * PSIZE	physical size (after compression)
204 * ASIZE	allocated size (including RAID-Z parity and gang block headers)
205 * GRID		RAID-Z layout information (reserved for future use)
206 * cksum	checksum function
207 * comp		compression function
208 * G		gang block indicator
209 * B		byteorder (endianness)
210 * D		dedup
211 * X		encryption (on version 30, which is not supported)
212 * E		blkptr_t contains embedded data (see below)
213 * lvl		level of indirection
214 * type		DMU object type
215 * phys birth	txg of block allocation; zero if same as logical birth txg
216 * log. birth	transaction group in which the block was logically born
217 * fill count	number of non-zero blocks under this bp
218 * checksum[4]	256-bit checksum of the data this bp describes
219 */
220
221/*
222 * "Embedded" blkptr_t's don't actually point to a block, instead they
223 * have a data payload embedded in the blkptr_t itself.  See the comment
224 * in blkptr.c for more details.
225 *
226 * The blkptr_t is laid out as follows:
227 *
228 *	64	56	48	40	32	24	16	8	0
229 *	+-------+-------+-------+-------+-------+-------+-------+-------+
230 * 0	|      payload                                                  |
231 * 1	|      payload                                                  |
232 * 2	|      payload                                                  |
233 * 3	|      payload                                                  |
234 * 4	|      payload                                                  |
235 * 5	|      payload                                                  |
236 *	+-------+-------+-------+-------+-------+-------+-------+-------+
237 * 6	|BDX|lvl| type	| etype |E| comp| PSIZE|              LSIZE	|
238 *	+-------+-------+-------+-------+-------+-------+-------+-------+
239 * 7	|      payload                                                  |
240 * 8	|      payload                                                  |
241 * 9	|      payload                                                  |
242 *	+-------+-------+-------+-------+-------+-------+-------+-------+
243 * a	|			logical birth txg			|
244 *	+-------+-------+-------+-------+-------+-------+-------+-------+
245 * b	|      payload                                                  |
246 * c	|      payload                                                  |
247 * d	|      payload                                                  |
248 * e	|      payload                                                  |
249 * f	|      payload                                                  |
250 *	+-------+-------+-------+-------+-------+-------+-------+-------+
251 *
252 * Legend:
253 *
254 * payload		contains the embedded data
255 * B (byteorder)	byteorder (endianness)
256 * D (dedup)		padding (set to zero)
257 * X			encryption (set to zero; see above)
258 * E (embedded)		set to one
259 * lvl			indirection level
260 * type			DMU object type
261 * etype		how to interpret embedded data (BP_EMBEDDED_TYPE_*)
262 * comp			compression function of payload
263 * PSIZE		size of payload after compression, in bytes
264 * LSIZE		logical size of payload, in bytes
265 *			note that 25 bits is enough to store the largest
266 *			"normal" BP's LSIZE (2^16 * 2^9) in bytes
267 * log. birth		transaction group in which the block was logically born
268 *
269 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
270 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
271 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
272 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
273 * BP's so the BP_SET_* macros can be used with them.  etype, PSIZE, LSIZE must
274 * be set with the BPE_SET_* macros.  BP_SET_EMBEDDED() should be called before
275 * other macros, as they assert that they are only used on BP's of the correct
276 * "embedded-ness".
277 */
278
279#define	BPE_GET_ETYPE(bp)	\
280	(ASSERT(BP_IS_EMBEDDED(bp)), \
281	BF64_GET((bp)->blk_prop, 40, 8))
282#define	BPE_SET_ETYPE(bp, t)	do { \
283	ASSERT(BP_IS_EMBEDDED(bp)); \
284	BF64_SET((bp)->blk_prop, 40, 8, t); \
285_NOTE(CONSTCOND) } while (0)
286
287#define	BPE_GET_LSIZE(bp)	\
288	(ASSERT(BP_IS_EMBEDDED(bp)), \
289	BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
290#define	BPE_SET_LSIZE(bp, x)	do { \
291	ASSERT(BP_IS_EMBEDDED(bp)); \
292	BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
293_NOTE(CONSTCOND) } while (0)
294
295#define	BPE_GET_PSIZE(bp)	\
296	(ASSERT(BP_IS_EMBEDDED(bp)), \
297	BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
298#define	BPE_SET_PSIZE(bp, x)	do { \
299	ASSERT(BP_IS_EMBEDDED(bp)); \
300	BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
301_NOTE(CONSTCOND) } while (0)
302
303typedef enum bp_embedded_type {
304	BP_EMBEDDED_TYPE_DATA,
305	BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */
306	NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED
307} bp_embedded_type_t;
308
309#define	BPE_NUM_WORDS 14
310#define	BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
311#define	BPE_IS_PAYLOADWORD(bp, wp) \
312	((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
313
314#define	SPA_BLKPTRSHIFT	7		/* blkptr_t is 128 bytes	*/
315#define	SPA_DVAS_PER_BP	3		/* Number of DVAs in a bp	*/
316
317/*
318 * A block is a hole when it has either 1) never been written to, or
319 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
320 * without physically allocating disk space. Holes are represented in the
321 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is
322 * done through the BP_IS_HOLE macro. For holes, the logical size, level,
323 * DMU object type, and birth times are all also stored for holes that
324 * were written to at some point (i.e. were punched after having been filled).
325 */
326typedef struct blkptr {
327	dva_t		blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
328	uint64_t	blk_prop;	/* size, compression, type, etc	    */
329	uint64_t	blk_pad[2];	/* Extra space for the future	    */
330	uint64_t	blk_phys_birth;	/* txg when block was allocated	    */
331	uint64_t	blk_birth;	/* transaction group at birth	    */
332	uint64_t	blk_fill;	/* fill count			    */
333	zio_cksum_t	blk_cksum;	/* 256-bit checksum		    */
334} blkptr_t;
335
336/*
337 * Macros to get and set fields in a bp or DVA.
338 */
339#define	DVA_GET_ASIZE(dva)	\
340	BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
341#define	DVA_SET_ASIZE(dva, x)	\
342	BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
343	SPA_MINBLOCKSHIFT, 0, x)
344
345#define	DVA_GET_GRID(dva)	BF64_GET((dva)->dva_word[0], 24, 8)
346#define	DVA_SET_GRID(dva, x)	BF64_SET((dva)->dva_word[0], 24, 8, x)
347
348#define	DVA_GET_VDEV(dva)	BF64_GET((dva)->dva_word[0], 32, 32)
349#define	DVA_SET_VDEV(dva, x)	BF64_SET((dva)->dva_word[0], 32, 32, x)
350
351#define	DVA_GET_OFFSET(dva)	\
352	BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
353#define	DVA_SET_OFFSET(dva, x)	\
354	BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
355
356#define	DVA_GET_GANG(dva)	BF64_GET((dva)->dva_word[1], 63, 1)
357#define	DVA_SET_GANG(dva, x)	BF64_SET((dva)->dva_word[1], 63, 1, x)
358
359#define	BP_GET_LSIZE(bp)	\
360	(BP_IS_EMBEDDED(bp) ?	\
361	(BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
362	BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
363#define	BP_SET_LSIZE(bp, x)	do { \
364	ASSERT(!BP_IS_EMBEDDED(bp)); \
365	BF64_SET_SB((bp)->blk_prop, \
366	    0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
367_NOTE(CONSTCOND) } while (0)
368
369#define	BP_GET_PSIZE(bp)	\
370	(BP_IS_EMBEDDED(bp) ? 0 : \
371	BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1))
372#define	BP_SET_PSIZE(bp, x)	do { \
373	ASSERT(!BP_IS_EMBEDDED(bp)); \
374	BF64_SET_SB((bp)->blk_prop, \
375	    16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
376_NOTE(CONSTCOND) } while (0)
377
378#define	BP_GET_COMPRESS(bp)		BF64_GET((bp)->blk_prop, 32, 7)
379#define	BP_SET_COMPRESS(bp, x)		BF64_SET((bp)->blk_prop, 32, 7, x)
380
381#define	BP_IS_EMBEDDED(bp)		BF64_GET((bp)->blk_prop, 39, 1)
382#define	BP_SET_EMBEDDED(bp, x)		BF64_SET((bp)->blk_prop, 39, 1, x)
383
384#define	BP_GET_CHECKSUM(bp)		\
385	(BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \
386	BF64_GET((bp)->blk_prop, 40, 8))
387#define	BP_SET_CHECKSUM(bp, x)		do { \
388	ASSERT(!BP_IS_EMBEDDED(bp)); \
389	BF64_SET((bp)->blk_prop, 40, 8, x); \
390_NOTE(CONSTCOND) } while (0)
391
392#define	BP_GET_TYPE(bp)			BF64_GET((bp)->blk_prop, 48, 8)
393#define	BP_SET_TYPE(bp, x)		BF64_SET((bp)->blk_prop, 48, 8, x)
394
395#define	BP_GET_LEVEL(bp)		BF64_GET((bp)->blk_prop, 56, 5)
396#define	BP_SET_LEVEL(bp, x)		BF64_SET((bp)->blk_prop, 56, 5, x)
397
398#define	BP_GET_DEDUP(bp)		BF64_GET((bp)->blk_prop, 62, 1)
399#define	BP_SET_DEDUP(bp, x)		BF64_SET((bp)->blk_prop, 62, 1, x)
400
401#define	BP_GET_BYTEORDER(bp)		BF64_GET((bp)->blk_prop, 63, 1)
402#define	BP_SET_BYTEORDER(bp, x)		BF64_SET((bp)->blk_prop, 63, 1, x)
403
404#define	BP_PHYSICAL_BIRTH(bp)		\
405	(BP_IS_EMBEDDED(bp) ? 0 : \
406	(bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
407
408#define	BP_SET_BIRTH(bp, logical, physical)	\
409{						\
410	ASSERT(!BP_IS_EMBEDDED(bp));		\
411	(bp)->blk_birth = (logical);		\
412	(bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
413}
414
415#define	BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill)
416
417#define	BP_GET_ASIZE(bp)	\
418	(BP_IS_EMBEDDED(bp) ? 0 : \
419	DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
420	DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
421	DVA_GET_ASIZE(&(bp)->blk_dva[2]))
422
423#define	BP_GET_UCSIZE(bp) \
424	((BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) ? \
425	BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
426
427#define	BP_GET_NDVAS(bp)	\
428	(BP_IS_EMBEDDED(bp) ? 0 : \
429	!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
430	!!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
431	!!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
432
433#define	BP_COUNT_GANG(bp)	\
434	(BP_IS_EMBEDDED(bp) ? 0 : \
435	(DVA_GET_GANG(&(bp)->blk_dva[0]) + \
436	DVA_GET_GANG(&(bp)->blk_dva[1]) + \
437	DVA_GET_GANG(&(bp)->blk_dva[2])))
438
439#define	DVA_EQUAL(dva1, dva2)	\
440	((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
441	(dva1)->dva_word[0] == (dva2)->dva_word[0])
442
443#define	BP_EQUAL(bp1, bp2)	\
444	(BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) &&	\
445	(bp1)->blk_birth == (bp2)->blk_birth &&			\
446	DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) &&	\
447	DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) &&	\
448	DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
449
450#define	ZIO_CHECKSUM_EQUAL(zc1, zc2) \
451	(0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
452	((zc1).zc_word[1] - (zc2).zc_word[1]) | \
453	((zc1).zc_word[2] - (zc2).zc_word[2]) | \
454	((zc1).zc_word[3] - (zc2).zc_word[3])))
455
456#define	ZIO_CHECKSUM_IS_ZERO(zc) \
457	(0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \
458	(zc)->zc_word[2] | (zc)->zc_word[3]))
459
460#define	ZIO_CHECKSUM_BSWAP(zcp)					\
461{								\
462	(zcp)->zc_word[0] = BSWAP_64((zcp)->zc_word[0]);	\
463	(zcp)->zc_word[1] = BSWAP_64((zcp)->zc_word[1]);	\
464	(zcp)->zc_word[2] = BSWAP_64((zcp)->zc_word[2]);	\
465	(zcp)->zc_word[3] = BSWAP_64((zcp)->zc_word[3]);	\
466}
467
468
469#define	DVA_IS_VALID(dva)	(DVA_GET_ASIZE(dva) != 0)
470
471#define	ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3)	\
472{						\
473	(zcp)->zc_word[0] = w0;			\
474	(zcp)->zc_word[1] = w1;			\
475	(zcp)->zc_word[2] = w2;			\
476	(zcp)->zc_word[3] = w3;			\
477}
478
479#define	BP_IDENTITY(bp)		(ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0])
480#define	BP_IS_GANG(bp)		\
481	(BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp)))
482#define	DVA_IS_EMPTY(dva)	((dva)->dva_word[0] == 0ULL &&	\
483				(dva)->dva_word[1] == 0ULL)
484#define	BP_IS_HOLE(bp) \
485	(!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp)))
486
487/* BP_IS_RAIDZ(bp) assumes no block compression */
488#define	BP_IS_RAIDZ(bp)		(DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
489				BP_GET_PSIZE(bp))
490
491#define	BP_ZERO(bp)				\
492{						\
493	(bp)->blk_dva[0].dva_word[0] = 0;	\
494	(bp)->blk_dva[0].dva_word[1] = 0;	\
495	(bp)->blk_dva[1].dva_word[0] = 0;	\
496	(bp)->blk_dva[1].dva_word[1] = 0;	\
497	(bp)->blk_dva[2].dva_word[0] = 0;	\
498	(bp)->blk_dva[2].dva_word[1] = 0;	\
499	(bp)->blk_prop = 0;			\
500	(bp)->blk_pad[0] = 0;			\
501	(bp)->blk_pad[1] = 0;			\
502	(bp)->blk_phys_birth = 0;		\
503	(bp)->blk_birth = 0;			\
504	(bp)->blk_fill = 0;			\
505	ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0);	\
506}
507
508#ifdef _BIG_ENDIAN
509#define	ZFS_HOST_BYTEORDER	(0ULL)
510#else
511#define	ZFS_HOST_BYTEORDER	(1ULL)
512#endif
513
514#define	BP_SHOULD_BYTESWAP(bp)	(BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
515
516#define	BP_SPRINTF_LEN	320
517
518/*
519 * This macro allows code sharing between zfs, libzpool, and mdb.
520 * 'func' is either snprintf() or mdb_snprintf().
521 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
522 */
523#define	SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
524{									\
525	static const char *copyname[] =					\
526	    { "zero", "single", "double", "triple" };			\
527	int len = 0;							\
528	int copies = 0;							\
529									\
530	if (bp == NULL) {						\
531		len += func(buf + len, size - len, "<NULL>");		\
532	} else if (BP_IS_HOLE(bp)) {					\
533		len += func(buf + len, size - len,			\
534		    "HOLE [L%llu %s] "					\
535		    "size=%llxL birth=%lluL",				\
536		    (u_longlong_t)BP_GET_LEVEL(bp),			\
537		    type,						\
538		    (u_longlong_t)BP_GET_LSIZE(bp),			\
539		    (u_longlong_t)bp->blk_birth);			\
540	} else if (BP_IS_EMBEDDED(bp)) {				\
541		len = func(buf + len, size - len,			\
542		    "EMBEDDED [L%llu %s] et=%u %s "			\
543		    "size=%llxL/%llxP birth=%lluL",			\
544		    (u_longlong_t)BP_GET_LEVEL(bp),			\
545		    type,						\
546		    (int)BPE_GET_ETYPE(bp),				\
547		    compress,						\
548		    (u_longlong_t)BPE_GET_LSIZE(bp),			\
549		    (u_longlong_t)BPE_GET_PSIZE(bp),			\
550		    (u_longlong_t)bp->blk_birth);			\
551	} else {							\
552		for (int d = 0; d < BP_GET_NDVAS(bp); d++) {		\
553			const dva_t *dva = &bp->blk_dva[d];		\
554			if (DVA_IS_VALID(dva))				\
555				copies++;				\
556			len += func(buf + len, size - len,		\
557			    "DVA[%d]=<%llu:%llx:%llx>%c", d,		\
558			    (u_longlong_t)DVA_GET_VDEV(dva),		\
559			    (u_longlong_t)DVA_GET_OFFSET(dva),		\
560			    (u_longlong_t)DVA_GET_ASIZE(dva),		\
561			    ws);					\
562		}							\
563		if (BP_IS_GANG(bp) &&					\
564		    DVA_GET_ASIZE(&bp->blk_dva[2]) <=			\
565		    DVA_GET_ASIZE(&bp->blk_dva[1]) / 2)			\
566			copies--;					\
567		len += func(buf + len, size - len,			\
568		    "[L%llu %s] %s %s %s %s %s %s%c"			\
569		    "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c"	\
570		    "cksum=%llx:%llx:%llx:%llx",			\
571		    (u_longlong_t)BP_GET_LEVEL(bp),			\
572		    type,						\
573		    checksum,						\
574		    compress,						\
575		    BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE",		\
576		    BP_IS_GANG(bp) ? "gang" : "contiguous",		\
577		    BP_GET_DEDUP(bp) ? "dedup" : "unique",		\
578		    copyname[copies],					\
579		    ws,							\
580		    (u_longlong_t)BP_GET_LSIZE(bp),			\
581		    (u_longlong_t)BP_GET_PSIZE(bp),			\
582		    (u_longlong_t)bp->blk_birth,			\
583		    (u_longlong_t)BP_PHYSICAL_BIRTH(bp),		\
584		    (u_longlong_t)BP_GET_FILL(bp),			\
585		    ws,							\
586		    (u_longlong_t)bp->blk_cksum.zc_word[0],		\
587		    (u_longlong_t)bp->blk_cksum.zc_word[1],		\
588		    (u_longlong_t)bp->blk_cksum.zc_word[2],		\
589		    (u_longlong_t)bp->blk_cksum.zc_word[3]);		\
590	}								\
591	ASSERT(len < size);						\
592}
593
594#include <sys/dmu.h>
595
596#define	BP_GET_BUFC_TYPE(bp)						\
597	(((BP_GET_LEVEL(bp) > 0) || (DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))) ? \
598	ARC_BUFC_METADATA : ARC_BUFC_DATA)
599
600typedef enum spa_import_type {
601	SPA_IMPORT_EXISTING,
602	SPA_IMPORT_ASSEMBLE
603} spa_import_type_t;
604
605/* state manipulation functions */
606extern int spa_open(const char *pool, spa_t **, void *tag);
607extern int spa_open_rewind(const char *pool, spa_t **, void *tag,
608    nvlist_t *policy, nvlist_t **config);
609extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot,
610    size_t buflen);
611extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props,
612    nvlist_t *zplprops);
613extern int spa_import_rootpool(char *devpath, char *devid);
614extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props,
615    uint64_t flags);
616extern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
617extern int spa_destroy(char *pool);
618extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force,
619    boolean_t hardforce);
620extern int spa_reset(char *pool);
621extern void spa_async_request(spa_t *spa, int flag);
622extern void spa_async_unrequest(spa_t *spa, int flag);
623extern void spa_async_suspend(spa_t *spa);
624extern void spa_async_resume(spa_t *spa);
625extern spa_t *spa_inject_addref(char *pool);
626extern void spa_inject_delref(spa_t *spa);
627extern void spa_scan_stat_init(spa_t *spa);
628extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps);
629
630#define	SPA_ASYNC_CONFIG_UPDATE	0x01
631#define	SPA_ASYNC_REMOVE	0x02
632#define	SPA_ASYNC_PROBE		0x04
633#define	SPA_ASYNC_RESILVER_DONE	0x08
634#define	SPA_ASYNC_RESILVER	0x10
635#define	SPA_ASYNC_AUTOEXPAND	0x20
636#define	SPA_ASYNC_REMOVE_DONE	0x40
637#define	SPA_ASYNC_REMOVE_STOP	0x80
638
639/*
640 * Controls the behavior of spa_vdev_remove().
641 */
642#define	SPA_REMOVE_UNSPARE	0x01
643#define	SPA_REMOVE_DONE		0x02
644
645/* device manipulation */
646extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
647extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot,
648    int replacing);
649extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
650    int replace_done);
651extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare);
652extern boolean_t spa_vdev_remove_active(spa_t *spa);
653extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
654extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru);
655extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
656    nvlist_t *props, boolean_t exp);
657
658/* spare state (which is global across all pools) */
659extern void spa_spare_add(vdev_t *vd);
660extern void spa_spare_remove(vdev_t *vd);
661extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt);
662extern void spa_spare_activate(vdev_t *vd);
663
664/* L2ARC state (which is global across all pools) */
665extern void spa_l2cache_add(vdev_t *vd);
666extern void spa_l2cache_remove(vdev_t *vd);
667extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
668extern void spa_l2cache_activate(vdev_t *vd);
669extern void spa_l2cache_drop(spa_t *spa);
670
671/* scanning */
672extern int spa_scan(spa_t *spa, pool_scan_func_t func);
673extern int spa_scan_stop(spa_t *spa);
674
675/* spa syncing */
676extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
677extern void spa_sync_allpools(void);
678
679/* spa namespace global mutex */
680extern kmutex_t spa_namespace_lock;
681
682/*
683 * SPA configuration functions in spa_config.c
684 */
685
686#define	SPA_CONFIG_UPDATE_POOL	0
687#define	SPA_CONFIG_UPDATE_VDEVS	1
688
689extern void spa_config_sync(spa_t *, boolean_t, boolean_t);
690extern void spa_config_load(void);
691extern nvlist_t *spa_all_configs(uint64_t *);
692extern void spa_config_set(spa_t *spa, nvlist_t *config);
693extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
694    int getstats);
695extern void spa_config_update(spa_t *spa, int what);
696
697/*
698 * Miscellaneous SPA routines in spa_misc.c
699 */
700
701/* Namespace manipulation */
702extern spa_t *spa_lookup(const char *name);
703extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot);
704extern void spa_remove(spa_t *spa);
705extern spa_t *spa_next(spa_t *prev);
706
707/* Refcount functions */
708extern void spa_open_ref(spa_t *spa, void *tag);
709extern void spa_close(spa_t *spa, void *tag);
710extern void spa_async_close(spa_t *spa, void *tag);
711extern boolean_t spa_refcount_zero(spa_t *spa);
712
713#define	SCL_NONE	0x00
714#define	SCL_CONFIG	0x01
715#define	SCL_STATE	0x02
716#define	SCL_L2ARC	0x04		/* hack until L2ARC 2.0 */
717#define	SCL_ALLOC	0x08
718#define	SCL_ZIO		0x10
719#define	SCL_FREE	0x20
720#define	SCL_VDEV	0x40
721#define	SCL_LOCKS	7
722#define	SCL_ALL		((1 << SCL_LOCKS) - 1)
723#define	SCL_STATE_ALL	(SCL_STATE | SCL_L2ARC | SCL_ZIO)
724
725/* Pool configuration locks */
726extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw);
727extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw);
728extern void spa_config_exit(spa_t *spa, int locks, void *tag);
729extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
730
731/* Pool vdev add/remove lock */
732extern uint64_t spa_vdev_enter(spa_t *spa);
733extern uint64_t spa_vdev_config_enter(spa_t *spa);
734extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg,
735    int error, char *tag);
736extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error);
737
738/* Pool vdev state change lock */
739extern void spa_vdev_state_enter(spa_t *spa, int oplock);
740extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error);
741
742/* Log state */
743typedef enum spa_log_state {
744	SPA_LOG_UNKNOWN = 0,	/* unknown log state */
745	SPA_LOG_MISSING,	/* missing log(s) */
746	SPA_LOG_CLEAR,		/* clear the log(s) */
747	SPA_LOG_GOOD,		/* log(s) are good */
748} spa_log_state_t;
749
750extern spa_log_state_t spa_get_log_state(spa_t *spa);
751extern void spa_set_log_state(spa_t *spa, spa_log_state_t state);
752extern int spa_offline_log(spa_t *spa);
753
754/* Log claim callback */
755extern void spa_claim_notify(zio_t *zio);
756
757/* Accessor functions */
758extern boolean_t spa_shutting_down(spa_t *spa);
759extern struct dsl_pool *spa_get_dsl(spa_t *spa);
760extern boolean_t spa_is_initializing(spa_t *spa);
761extern blkptr_t *spa_get_rootblkptr(spa_t *spa);
762extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
763extern void spa_altroot(spa_t *, char *, size_t);
764extern int spa_sync_pass(spa_t *spa);
765extern char *spa_name(spa_t *spa);
766extern uint64_t spa_guid(spa_t *spa);
767extern uint64_t spa_load_guid(spa_t *spa);
768extern uint64_t spa_last_synced_txg(spa_t *spa);
769extern uint64_t spa_first_txg(spa_t *spa);
770extern uint64_t spa_syncing_txg(spa_t *spa);
771extern uint64_t spa_version(spa_t *spa);
772extern pool_state_t spa_state(spa_t *spa);
773extern spa_load_state_t spa_load_state(spa_t *spa);
774extern uint64_t spa_freeze_txg(spa_t *spa);
775extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize);
776extern uint64_t spa_get_dspace(spa_t *spa);
777extern uint64_t spa_get_slop_space(spa_t *spa);
778extern void spa_update_dspace(spa_t *spa);
779extern uint64_t spa_version(spa_t *spa);
780extern boolean_t spa_deflate(spa_t *spa);
781extern metaslab_class_t *spa_normal_class(spa_t *spa);
782extern metaslab_class_t *spa_log_class(spa_t *spa);
783extern void spa_evicting_os_register(spa_t *, objset_t *os);
784extern void spa_evicting_os_deregister(spa_t *, objset_t *os);
785extern void spa_evicting_os_wait(spa_t *spa);
786extern int spa_max_replication(spa_t *spa);
787extern int spa_prev_software_version(spa_t *spa);
788extern int spa_busy(void);
789extern uint8_t spa_get_failmode(spa_t *spa);
790extern boolean_t spa_suspended(spa_t *spa);
791extern uint64_t spa_bootfs(spa_t *spa);
792extern uint64_t spa_delegation(spa_t *spa);
793extern objset_t *spa_meta_objset(spa_t *spa);
794extern uint64_t spa_deadman_synctime(spa_t *spa);
795
796/* Miscellaneous support routines */
797extern void spa_activate_mos_feature(spa_t *spa, const char *feature,
798    dmu_tx_t *tx);
799extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature);
800extern int spa_rename(const char *oldname, const char *newname);
801extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid);
802extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid);
803extern char *spa_strdup(const char *);
804extern void spa_strfree(char *);
805extern uint64_t spa_get_random(uint64_t range);
806extern uint64_t spa_generate_guid(spa_t *spa);
807extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp);
808extern void spa_freeze(spa_t *spa);
809extern int spa_change_guid(spa_t *spa);
810extern void spa_upgrade(spa_t *spa, uint64_t version);
811extern void spa_evict_all(void);
812extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid,
813    boolean_t l2cache);
814extern boolean_t spa_has_spare(spa_t *, uint64_t guid);
815extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva);
816extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp);
817extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp);
818extern boolean_t spa_has_slogs(spa_t *spa);
819extern boolean_t spa_is_root(spa_t *spa);
820extern boolean_t spa_writeable(spa_t *spa);
821extern boolean_t spa_has_pending_synctask(spa_t *spa);
822extern int spa_maxblocksize(spa_t *spa);
823extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp);
824
825extern int spa_mode(spa_t *spa);
826extern uint64_t strtonum(const char *str, char **nptr);
827
828extern char *spa_his_ievent_table[];
829
830extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx);
831extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read,
832    char *his_buf);
833extern int spa_history_log(spa_t *spa, const char *his_buf);
834extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl);
835extern void spa_history_log_version(spa_t *spa, const char *operation);
836extern void spa_history_log_internal(spa_t *spa, const char *operation,
837    dmu_tx_t *tx, const char *fmt, ...);
838extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op,
839    dmu_tx_t *tx, const char *fmt, ...);
840extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
841    dmu_tx_t *tx, const char *fmt, ...);
842
843/* error handling */
844struct zbookmark_phys;
845extern void spa_log_error(spa_t *spa, zio_t *zio);
846extern void zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd,
847    zio_t *zio, uint64_t stateoroffset, uint64_t length);
848extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
849extern void zfs_post_state_change(spa_t *spa, vdev_t *vd);
850extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
851extern uint64_t spa_get_errlog_size(spa_t *spa);
852extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
853extern void spa_errlog_rotate(spa_t *spa);
854extern void spa_errlog_drain(spa_t *spa);
855extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
856extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
857
858/* vdev cache */
859extern void vdev_cache_stat_init(void);
860extern void vdev_cache_stat_fini(void);
861
862/* Initialization and termination */
863extern void spa_init(int flags);
864extern void spa_fini(void);
865extern void spa_boot_init();
866
867/* properties */
868extern int spa_prop_set(spa_t *spa, nvlist_t *nvp);
869extern int spa_prop_get(spa_t *spa, nvlist_t **nvp);
870extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx);
871extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t);
872
873/* asynchronous event notification */
874extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name);
875
876#ifdef ZFS_DEBUG
877#define	dprintf_bp(bp, fmt, ...) do {				\
878	if (zfs_flags & ZFS_DEBUG_DPRINTF) {			\
879	char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP);	\
880	snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp));	\
881	dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf);		\
882	kmem_free(__blkbuf, BP_SPRINTF_LEN);			\
883	} \
884_NOTE(CONSTCOND) } while (0)
885#else
886#define	dprintf_bp(bp, fmt, ...)
887#endif
888
889extern boolean_t spa_debug_enabled(spa_t *spa);
890#define	spa_dbgmsg(spa, ...)			\
891{						\
892	if (spa_debug_enabled(spa))		\
893		zfs_dbgmsg(__VA_ARGS__);	\
894}
895
896extern int spa_mode_global;			/* mode, e.g. FREAD | FWRITE */
897
898#ifdef	__cplusplus
899}
900#endif
901
902#endif	/* _SYS_SPA_H */
903