spa.h revision 1702cce751c5cb7ead878d0205a6c90b027e3de8
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, 2016 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 * Copyright 2017 Joyent, Inc.
29 * Copyright (c) 2017 Datto Inc.
30 */
31
32#ifndef _SYS_SPA_H
33#define	_SYS_SPA_H
34
35#include <sys/avl.h>
36#include <sys/zfs_context.h>
37#include <sys/nvpair.h>
38#include <sys/sysmacros.h>
39#include <sys/types.h>
40#include <sys/fs/zfs.h>
41#include <sys/dmu.h>
42
43#ifdef	__cplusplus
44extern "C" {
45#endif
46
47/*
48 * Forward references that lots of things need.
49 */
50typedef struct spa spa_t;
51typedef struct vdev vdev_t;
52typedef struct metaslab metaslab_t;
53typedef struct metaslab_group metaslab_group_t;
54typedef struct metaslab_class metaslab_class_t;
55typedef struct zio zio_t;
56typedef struct zilog zilog_t;
57typedef struct spa_aux_vdev spa_aux_vdev_t;
58typedef struct ddt ddt_t;
59typedef struct ddt_entry ddt_entry_t;
60struct dsl_pool;
61struct dsl_dataset;
62
63/*
64 * General-purpose 32-bit and 64-bit bitfield encodings.
65 */
66#define	BF32_DECODE(x, low, len)	P2PHASE((x) >> (low), 1U << (len))
67#define	BF64_DECODE(x, low, len)	P2PHASE((x) >> (low), 1ULL << (len))
68#define	BF32_ENCODE(x, low, len)	(P2PHASE((x), 1U << (len)) << (low))
69#define	BF64_ENCODE(x, low, len)	(P2PHASE((x), 1ULL << (len)) << (low))
70
71#define	BF32_GET(x, low, len)		BF32_DECODE(x, low, len)
72#define	BF64_GET(x, low, len)		BF64_DECODE(x, low, len)
73
74#define	BF32_SET(x, low, len, val) do { \
75	ASSERT3U(val, <, 1U << (len)); \
76	ASSERT3U(low + len, <=, 32); \
77	(x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \
78_NOTE(CONSTCOND) } while (0)
79
80#define	BF64_SET(x, low, len, val) do { \
81	ASSERT3U(val, <, 1ULL << (len)); \
82	ASSERT3U(low + len, <=, 64); \
83	((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \
84_NOTE(CONSTCOND) } while (0)
85
86#define	BF32_GET_SB(x, low, len, shift, bias)	\
87	((BF32_GET(x, low, len) + (bias)) << (shift))
88#define	BF64_GET_SB(x, low, len, shift, bias)	\
89	((BF64_GET(x, low, len) + (bias)) << (shift))
90
91#define	BF32_SET_SB(x, low, len, shift, bias, val) do { \
92	ASSERT(IS_P2ALIGNED(val, 1U << shift)); \
93	ASSERT3S((val) >> (shift), >=, bias); \
94	BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \
95_NOTE(CONSTCOND) } while (0)
96#define	BF64_SET_SB(x, low, len, shift, bias, val) do { \
97	ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \
98	ASSERT3S((val) >> (shift), >=, bias); \
99	BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \
100_NOTE(CONSTCOND) } while (0)
101
102/*
103 * We currently support block sizes from 512 bytes to 16MB.
104 * The benefits of larger blocks, and thus larger IO, need to be weighed
105 * against the cost of COWing a giant block to modify one byte, and the
106 * large latency of reading or writing a large block.
107 *
108 * Note that although blocks up to 16MB are supported, the recordsize
109 * property can not be set larger than zfs_max_recordsize (default 1MB).
110 * See the comment near zfs_max_recordsize in dsl_dataset.c for details.
111 *
112 * Note that although the LSIZE field of the blkptr_t can store sizes up
113 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to
114 * 32MB - 512 bytes.  Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
115 */
116#define	SPA_MINBLOCKSHIFT	9
117#define	SPA_OLD_MAXBLOCKSHIFT	17
118#define	SPA_MAXBLOCKSHIFT	24
119#define	SPA_MINBLOCKSIZE	(1ULL << SPA_MINBLOCKSHIFT)
120#define	SPA_OLD_MAXBLOCKSIZE	(1ULL << SPA_OLD_MAXBLOCKSHIFT)
121#define	SPA_MAXBLOCKSIZE	(1ULL << SPA_MAXBLOCKSHIFT)
122
123/*
124 * Size of block to hold the configuration data (a packed nvlist)
125 */
126#define	SPA_CONFIG_BLOCKSIZE	(1ULL << 14)
127
128/*
129 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
130 * The ASIZE encoding should be at least 64 times larger (6 more bits)
131 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
132 * overhead, three DVAs per bp, plus one more bit in case we do anything
133 * else that expands the ASIZE.
134 */
135#define	SPA_LSIZEBITS		16	/* LSIZE up to 32M (2^16 * 512)	*/
136#define	SPA_PSIZEBITS		16	/* PSIZE up to 32M (2^16 * 512)	*/
137#define	SPA_ASIZEBITS		24	/* ASIZE up to 64 times larger	*/
138
139#define	SPA_COMPRESSBITS	7
140
141/*
142 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
143 * The members of the dva_t should be considered opaque outside the SPA.
144 */
145typedef struct dva {
146	uint64_t	dva_word[2];
147} dva_t;
148
149/*
150 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
151 */
152typedef struct zio_cksum {
153	uint64_t	zc_word[4];
154} zio_cksum_t;
155
156/*
157 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept
158 * secret and is suitable for use in MAC algorithms as the key.
159 */
160typedef struct zio_cksum_salt {
161	uint8_t		zcs_bytes[32];
162} zio_cksum_salt_t;
163
164/*
165 * Each block is described by its DVAs, time of birth, checksum, etc.
166 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
167 *
168 *	64	56	48	40	32	24	16	8	0
169 *	+-------+-------+-------+-------+-------+-------+-------+-------+
170 * 0	|		vdev1		| GRID  |	  ASIZE		|
171 *	+-------+-------+-------+-------+-------+-------+-------+-------+
172 * 1	|G|			 offset1				|
173 *	+-------+-------+-------+-------+-------+-------+-------+-------+
174 * 2	|		vdev2		| GRID  |	  ASIZE		|
175 *	+-------+-------+-------+-------+-------+-------+-------+-------+
176 * 3	|G|			 offset2				|
177 *	+-------+-------+-------+-------+-------+-------+-------+-------+
178 * 4	|		vdev3		| GRID  |	  ASIZE		|
179 *	+-------+-------+-------+-------+-------+-------+-------+-------+
180 * 5	|G|			 offset3				|
181 *	+-------+-------+-------+-------+-------+-------+-------+-------+
182 * 6	|BDX|lvl| type	| cksum |E| comp|    PSIZE	|     LSIZE	|
183 *	+-------+-------+-------+-------+-------+-------+-------+-------+
184 * 7	|			padding					|
185 *	+-------+-------+-------+-------+-------+-------+-------+-------+
186 * 8	|			padding					|
187 *	+-------+-------+-------+-------+-------+-------+-------+-------+
188 * 9	|			physical birth txg			|
189 *	+-------+-------+-------+-------+-------+-------+-------+-------+
190 * a	|			logical birth txg			|
191 *	+-------+-------+-------+-------+-------+-------+-------+-------+
192 * b	|			fill count				|
193 *	+-------+-------+-------+-------+-------+-------+-------+-------+
194 * c	|			checksum[0]				|
195 *	+-------+-------+-------+-------+-------+-------+-------+-------+
196 * d	|			checksum[1]				|
197 *	+-------+-------+-------+-------+-------+-------+-------+-------+
198 * e	|			checksum[2]				|
199 *	+-------+-------+-------+-------+-------+-------+-------+-------+
200 * f	|			checksum[3]				|
201 *	+-------+-------+-------+-------+-------+-------+-------+-------+
202 *
203 * Legend:
204 *
205 * vdev		virtual device ID
206 * offset	offset into virtual device
207 * LSIZE	logical size
208 * PSIZE	physical size (after compression)
209 * ASIZE	allocated size (including RAID-Z parity and gang block headers)
210 * GRID		RAID-Z layout information (reserved for future use)
211 * cksum	checksum function
212 * comp		compression function
213 * G		gang block indicator
214 * B		byteorder (endianness)
215 * D		dedup
216 * X		encryption (on version 30, which is not supported)
217 * E		blkptr_t contains embedded data (see below)
218 * lvl		level of indirection
219 * type		DMU object type
220 * phys birth	txg of block allocation; zero if same as logical birth txg
221 * log. birth	transaction group in which the block was logically born
222 * fill count	number of non-zero blocks under this bp
223 * checksum[4]	256-bit checksum of the data this bp describes
224 */
225
226/*
227 * "Embedded" blkptr_t's don't actually point to a block, instead they
228 * have a data payload embedded in the blkptr_t itself.  See the comment
229 * in blkptr.c for more details.
230 *
231 * The blkptr_t is laid out as follows:
232 *
233 *	64	56	48	40	32	24	16	8	0
234 *	+-------+-------+-------+-------+-------+-------+-------+-------+
235 * 0	|      payload                                                  |
236 * 1	|      payload                                                  |
237 * 2	|      payload                                                  |
238 * 3	|      payload                                                  |
239 * 4	|      payload                                                  |
240 * 5	|      payload                                                  |
241 *	+-------+-------+-------+-------+-------+-------+-------+-------+
242 * 6	|BDX|lvl| type	| etype |E| comp| PSIZE|              LSIZE	|
243 *	+-------+-------+-------+-------+-------+-------+-------+-------+
244 * 7	|      payload                                                  |
245 * 8	|      payload                                                  |
246 * 9	|      payload                                                  |
247 *	+-------+-------+-------+-------+-------+-------+-------+-------+
248 * a	|			logical birth txg			|
249 *	+-------+-------+-------+-------+-------+-------+-------+-------+
250 * b	|      payload                                                  |
251 * c	|      payload                                                  |
252 * d	|      payload                                                  |
253 * e	|      payload                                                  |
254 * f	|      payload                                                  |
255 *	+-------+-------+-------+-------+-------+-------+-------+-------+
256 *
257 * Legend:
258 *
259 * payload		contains the embedded data
260 * B (byteorder)	byteorder (endianness)
261 * D (dedup)		padding (set to zero)
262 * X			encryption (set to zero; see above)
263 * E (embedded)		set to one
264 * lvl			indirection level
265 * type			DMU object type
266 * etype		how to interpret embedded data (BP_EMBEDDED_TYPE_*)
267 * comp			compression function of payload
268 * PSIZE		size of payload after compression, in bytes
269 * LSIZE		logical size of payload, in bytes
270 *			note that 25 bits is enough to store the largest
271 *			"normal" BP's LSIZE (2^16 * 2^9) in bytes
272 * log. birth		transaction group in which the block was logically born
273 *
274 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
275 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
276 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
277 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
278 * BP's so the BP_SET_* macros can be used with them.  etype, PSIZE, LSIZE must
279 * be set with the BPE_SET_* macros.  BP_SET_EMBEDDED() should be called before
280 * other macros, as they assert that they are only used on BP's of the correct
281 * "embedded-ness".
282 */
283
284#define	BPE_GET_ETYPE(bp)	\
285	(ASSERT(BP_IS_EMBEDDED(bp)), \
286	BF64_GET((bp)->blk_prop, 40, 8))
287#define	BPE_SET_ETYPE(bp, t)	do { \
288	ASSERT(BP_IS_EMBEDDED(bp)); \
289	BF64_SET((bp)->blk_prop, 40, 8, t); \
290_NOTE(CONSTCOND) } while (0)
291
292#define	BPE_GET_LSIZE(bp)	\
293	(ASSERT(BP_IS_EMBEDDED(bp)), \
294	BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
295#define	BPE_SET_LSIZE(bp, x)	do { \
296	ASSERT(BP_IS_EMBEDDED(bp)); \
297	BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
298_NOTE(CONSTCOND) } while (0)
299
300#define	BPE_GET_PSIZE(bp)	\
301	(ASSERT(BP_IS_EMBEDDED(bp)), \
302	BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
303#define	BPE_SET_PSIZE(bp, x)	do { \
304	ASSERT(BP_IS_EMBEDDED(bp)); \
305	BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
306_NOTE(CONSTCOND) } while (0)
307
308typedef enum bp_embedded_type {
309	BP_EMBEDDED_TYPE_DATA,
310	BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */
311	NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED
312} bp_embedded_type_t;
313
314#define	BPE_NUM_WORDS 14
315#define	BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
316#define	BPE_IS_PAYLOADWORD(bp, wp) \
317	((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
318
319#define	SPA_BLKPTRSHIFT	7		/* blkptr_t is 128 bytes	*/
320#define	SPA_DVAS_PER_BP	3		/* Number of DVAs in a bp	*/
321
322/*
323 * A block is a hole when it has either 1) never been written to, or
324 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
325 * without physically allocating disk space. Holes are represented in the
326 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is
327 * done through the BP_IS_HOLE macro. For holes, the logical size, level,
328 * DMU object type, and birth times are all also stored for holes that
329 * were written to at some point (i.e. were punched after having been filled).
330 */
331typedef struct blkptr {
332	dva_t		blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
333	uint64_t	blk_prop;	/* size, compression, type, etc	    */
334	uint64_t	blk_pad[2];	/* Extra space for the future	    */
335	uint64_t	blk_phys_birth;	/* txg when block was allocated	    */
336	uint64_t	blk_birth;	/* transaction group at birth	    */
337	uint64_t	blk_fill;	/* fill count			    */
338	zio_cksum_t	blk_cksum;	/* 256-bit checksum		    */
339} blkptr_t;
340
341/*
342 * Macros to get and set fields in a bp or DVA.
343 */
344#define	DVA_GET_ASIZE(dva)	\
345	BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
346#define	DVA_SET_ASIZE(dva, x)	\
347	BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
348	SPA_MINBLOCKSHIFT, 0, x)
349
350#define	DVA_GET_GRID(dva)	BF64_GET((dva)->dva_word[0], 24, 8)
351#define	DVA_SET_GRID(dva, x)	BF64_SET((dva)->dva_word[0], 24, 8, x)
352
353#define	DVA_GET_VDEV(dva)	BF64_GET((dva)->dva_word[0], 32, 32)
354#define	DVA_SET_VDEV(dva, x)	BF64_SET((dva)->dva_word[0], 32, 32, x)
355
356#define	DVA_GET_OFFSET(dva)	\
357	BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
358#define	DVA_SET_OFFSET(dva, x)	\
359	BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
360
361#define	DVA_GET_GANG(dva)	BF64_GET((dva)->dva_word[1], 63, 1)
362#define	DVA_SET_GANG(dva, x)	BF64_SET((dva)->dva_word[1], 63, 1, x)
363
364#define	BP_GET_LSIZE(bp)	\
365	(BP_IS_EMBEDDED(bp) ?	\
366	(BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
367	BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
368#define	BP_SET_LSIZE(bp, x)	do { \
369	ASSERT(!BP_IS_EMBEDDED(bp)); \
370	BF64_SET_SB((bp)->blk_prop, \
371	    0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
372_NOTE(CONSTCOND) } while (0)
373
374#define	BP_GET_PSIZE(bp)	\
375	(BP_IS_EMBEDDED(bp) ? 0 : \
376	BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1))
377#define	BP_SET_PSIZE(bp, x)	do { \
378	ASSERT(!BP_IS_EMBEDDED(bp)); \
379	BF64_SET_SB((bp)->blk_prop, \
380	    16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
381_NOTE(CONSTCOND) } while (0)
382
383#define	BP_GET_COMPRESS(bp)		\
384	BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS)
385#define	BP_SET_COMPRESS(bp, x)		\
386	BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x)
387
388#define	BP_IS_EMBEDDED(bp)		BF64_GET((bp)->blk_prop, 39, 1)
389#define	BP_SET_EMBEDDED(bp, x)		BF64_SET((bp)->blk_prop, 39, 1, x)
390
391#define	BP_GET_CHECKSUM(bp)		\
392	(BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \
393	BF64_GET((bp)->blk_prop, 40, 8))
394#define	BP_SET_CHECKSUM(bp, x)		do { \
395	ASSERT(!BP_IS_EMBEDDED(bp)); \
396	BF64_SET((bp)->blk_prop, 40, 8, x); \
397_NOTE(CONSTCOND) } while (0)
398
399#define	BP_GET_TYPE(bp)			BF64_GET((bp)->blk_prop, 48, 8)
400#define	BP_SET_TYPE(bp, x)		BF64_SET((bp)->blk_prop, 48, 8, x)
401
402#define	BP_GET_LEVEL(bp)		BF64_GET((bp)->blk_prop, 56, 5)
403#define	BP_SET_LEVEL(bp, x)		BF64_SET((bp)->blk_prop, 56, 5, x)
404
405#define	BP_GET_DEDUP(bp)		BF64_GET((bp)->blk_prop, 62, 1)
406#define	BP_SET_DEDUP(bp, x)		BF64_SET((bp)->blk_prop, 62, 1, x)
407
408#define	BP_GET_BYTEORDER(bp)		BF64_GET((bp)->blk_prop, 63, 1)
409#define	BP_SET_BYTEORDER(bp, x)		BF64_SET((bp)->blk_prop, 63, 1, x)
410
411#define	BP_PHYSICAL_BIRTH(bp)		\
412	(BP_IS_EMBEDDED(bp) ? 0 : \
413	(bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
414
415#define	BP_SET_BIRTH(bp, logical, physical)	\
416{						\
417	ASSERT(!BP_IS_EMBEDDED(bp));		\
418	(bp)->blk_birth = (logical);		\
419	(bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
420}
421
422#define	BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill)
423
424#define	BP_IS_METADATA(bp)	\
425	(BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))
426
427#define	BP_GET_ASIZE(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_GET_UCSIZE(bp)	\
434	(BP_IS_METADATA(bp) ? BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
435
436#define	BP_GET_NDVAS(bp)	\
437	(BP_IS_EMBEDDED(bp) ? 0 : \
438	!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
439	!!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
440	!!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
441
442#define	BP_COUNT_GANG(bp)	\
443	(BP_IS_EMBEDDED(bp) ? 0 : \
444	(DVA_GET_GANG(&(bp)->blk_dva[0]) + \
445	DVA_GET_GANG(&(bp)->blk_dva[1]) + \
446	DVA_GET_GANG(&(bp)->blk_dva[2])))
447
448#define	DVA_EQUAL(dva1, dva2)	\
449	((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
450	(dva1)->dva_word[0] == (dva2)->dva_word[0])
451
452#define	BP_EQUAL(bp1, bp2)	\
453	(BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) &&	\
454	(bp1)->blk_birth == (bp2)->blk_birth &&			\
455	DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) &&	\
456	DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) &&	\
457	DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
458
459#define	ZIO_CHECKSUM_EQUAL(zc1, zc2) \
460	(0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
461	((zc1).zc_word[1] - (zc2).zc_word[1]) | \
462	((zc1).zc_word[2] - (zc2).zc_word[2]) | \
463	((zc1).zc_word[3] - (zc2).zc_word[3])))
464
465#define	ZIO_CHECKSUM_IS_ZERO(zc) \
466	(0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \
467	(zc)->zc_word[2] | (zc)->zc_word[3]))
468
469#define	ZIO_CHECKSUM_BSWAP(zcp)					\
470{								\
471	(zcp)->zc_word[0] = BSWAP_64((zcp)->zc_word[0]);	\
472	(zcp)->zc_word[1] = BSWAP_64((zcp)->zc_word[1]);	\
473	(zcp)->zc_word[2] = BSWAP_64((zcp)->zc_word[2]);	\
474	(zcp)->zc_word[3] = BSWAP_64((zcp)->zc_word[3]);	\
475}
476
477
478#define	DVA_IS_VALID(dva)	(DVA_GET_ASIZE(dva) != 0)
479
480#define	ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3)	\
481{						\
482	(zcp)->zc_word[0] = w0;			\
483	(zcp)->zc_word[1] = w1;			\
484	(zcp)->zc_word[2] = w2;			\
485	(zcp)->zc_word[3] = w3;			\
486}
487
488#define	BP_IDENTITY(bp)		(ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0])
489#define	BP_IS_GANG(bp)		\
490	(BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp)))
491#define	DVA_IS_EMPTY(dva)	((dva)->dva_word[0] == 0ULL &&	\
492				(dva)->dva_word[1] == 0ULL)
493#define	BP_IS_HOLE(bp) \
494	(!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp)))
495
496/* BP_IS_RAIDZ(bp) assumes no block compression */
497#define	BP_IS_RAIDZ(bp)		(DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
498				BP_GET_PSIZE(bp))
499
500#define	BP_ZERO(bp)				\
501{						\
502	(bp)->blk_dva[0].dva_word[0] = 0;	\
503	(bp)->blk_dva[0].dva_word[1] = 0;	\
504	(bp)->blk_dva[1].dva_word[0] = 0;	\
505	(bp)->blk_dva[1].dva_word[1] = 0;	\
506	(bp)->blk_dva[2].dva_word[0] = 0;	\
507	(bp)->blk_dva[2].dva_word[1] = 0;	\
508	(bp)->blk_prop = 0;			\
509	(bp)->blk_pad[0] = 0;			\
510	(bp)->blk_pad[1] = 0;			\
511	(bp)->blk_phys_birth = 0;		\
512	(bp)->blk_birth = 0;			\
513	(bp)->blk_fill = 0;			\
514	ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0);	\
515}
516
517#ifdef _BIG_ENDIAN
518#define	ZFS_HOST_BYTEORDER	(0ULL)
519#else
520#define	ZFS_HOST_BYTEORDER	(1ULL)
521#endif
522
523#define	BP_SHOULD_BYTESWAP(bp)	(BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
524
525#define	BP_SPRINTF_LEN	320
526
527/*
528 * This macro allows code sharing between zfs, libzpool, and mdb.
529 * 'func' is either snprintf() or mdb_snprintf().
530 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
531 */
532#define	SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
533{									\
534	static const char *copyname[] =					\
535	    { "zero", "single", "double", "triple" };			\
536	int len = 0;							\
537	int copies = 0;							\
538									\
539	if (bp == NULL) {						\
540		len += func(buf + len, size - len, "<NULL>");		\
541	} else if (BP_IS_HOLE(bp)) {					\
542		len += func(buf + len, size - len,			\
543		    "HOLE [L%llu %s] "					\
544		    "size=%llxL birth=%lluL",				\
545		    (u_longlong_t)BP_GET_LEVEL(bp),			\
546		    type,						\
547		    (u_longlong_t)BP_GET_LSIZE(bp),			\
548		    (u_longlong_t)bp->blk_birth);			\
549	} else if (BP_IS_EMBEDDED(bp)) {				\
550		len = func(buf + len, size - len,			\
551		    "EMBEDDED [L%llu %s] et=%u %s "			\
552		    "size=%llxL/%llxP birth=%lluL",			\
553		    (u_longlong_t)BP_GET_LEVEL(bp),			\
554		    type,						\
555		    (int)BPE_GET_ETYPE(bp),				\
556		    compress,						\
557		    (u_longlong_t)BPE_GET_LSIZE(bp),			\
558		    (u_longlong_t)BPE_GET_PSIZE(bp),			\
559		    (u_longlong_t)bp->blk_birth);			\
560	} else {							\
561		for (int d = 0; d < BP_GET_NDVAS(bp); d++) {		\
562			const dva_t *dva = &bp->blk_dva[d];		\
563			if (DVA_IS_VALID(dva))				\
564				copies++;				\
565			len += func(buf + len, size - len,		\
566			    "DVA[%d]=<%llu:%llx:%llx>%c", d,		\
567			    (u_longlong_t)DVA_GET_VDEV(dva),		\
568			    (u_longlong_t)DVA_GET_OFFSET(dva),		\
569			    (u_longlong_t)DVA_GET_ASIZE(dva),		\
570			    ws);					\
571		}							\
572		if (BP_IS_GANG(bp) &&					\
573		    DVA_GET_ASIZE(&bp->blk_dva[2]) <=			\
574		    DVA_GET_ASIZE(&bp->blk_dva[1]) / 2)			\
575			copies--;					\
576		len += func(buf + len, size - len,			\
577		    "[L%llu %s] %s %s %s %s %s %s%c"			\
578		    "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c"	\
579		    "cksum=%llx:%llx:%llx:%llx",			\
580		    (u_longlong_t)BP_GET_LEVEL(bp),			\
581		    type,						\
582		    checksum,						\
583		    compress,						\
584		    BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE",		\
585		    BP_IS_GANG(bp) ? "gang" : "contiguous",		\
586		    BP_GET_DEDUP(bp) ? "dedup" : "unique",		\
587		    copyname[copies],					\
588		    ws,							\
589		    (u_longlong_t)BP_GET_LSIZE(bp),			\
590		    (u_longlong_t)BP_GET_PSIZE(bp),			\
591		    (u_longlong_t)bp->blk_birth,			\
592		    (u_longlong_t)BP_PHYSICAL_BIRTH(bp),		\
593		    (u_longlong_t)BP_GET_FILL(bp),			\
594		    ws,							\
595		    (u_longlong_t)bp->blk_cksum.zc_word[0],		\
596		    (u_longlong_t)bp->blk_cksum.zc_word[1],		\
597		    (u_longlong_t)bp->blk_cksum.zc_word[2],		\
598		    (u_longlong_t)bp->blk_cksum.zc_word[3]);		\
599	}								\
600	ASSERT(len < size);						\
601}
602
603#define	BP_GET_BUFC_TYPE(bp)						\
604	(BP_IS_METADATA(bp) ? ARC_BUFC_METADATA : ARC_BUFC_DATA)
605
606typedef enum spa_import_type {
607	SPA_IMPORT_EXISTING,
608	SPA_IMPORT_ASSEMBLE
609} spa_import_type_t;
610
611/* state manipulation functions */
612extern int spa_open(const char *pool, spa_t **, void *tag);
613extern int spa_open_rewind(const char *pool, spa_t **, void *tag,
614    nvlist_t *policy, nvlist_t **config);
615extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot,
616    size_t buflen);
617extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props,
618    nvlist_t *zplprops);
619extern int spa_import_rootpool(char *devpath, char *devid);
620extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props,
621    uint64_t flags);
622extern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
623extern int spa_destroy(char *pool);
624extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force,
625    boolean_t hardforce);
626extern int spa_reset(char *pool);
627extern void spa_async_request(spa_t *spa, int flag);
628extern void spa_async_unrequest(spa_t *spa, int flag);
629extern void spa_async_suspend(spa_t *spa);
630extern void spa_async_resume(spa_t *spa);
631extern spa_t *spa_inject_addref(char *pool);
632extern void spa_inject_delref(spa_t *spa);
633extern void spa_scan_stat_init(spa_t *spa);
634extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps);
635
636#define	SPA_ASYNC_CONFIG_UPDATE	0x01
637#define	SPA_ASYNC_REMOVE	0x02
638#define	SPA_ASYNC_PROBE		0x04
639#define	SPA_ASYNC_RESILVER_DONE	0x08
640#define	SPA_ASYNC_RESILVER	0x10
641#define	SPA_ASYNC_AUTOEXPAND	0x20
642#define	SPA_ASYNC_REMOVE_DONE	0x40
643#define	SPA_ASYNC_REMOVE_STOP	0x80
644
645/*
646 * Controls the behavior of spa_vdev_remove().
647 */
648#define	SPA_REMOVE_UNSPARE	0x01
649#define	SPA_REMOVE_DONE		0x02
650
651/* device manipulation */
652extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
653extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot,
654    int replacing);
655extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
656    int replace_done);
657extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare);
658extern boolean_t spa_vdev_remove_active(spa_t *spa);
659extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
660extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru);
661extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
662    nvlist_t *props, boolean_t exp);
663
664/* spare state (which is global across all pools) */
665extern void spa_spare_add(vdev_t *vd);
666extern void spa_spare_remove(vdev_t *vd);
667extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt);
668extern void spa_spare_activate(vdev_t *vd);
669
670/* L2ARC state (which is global across all pools) */
671extern void spa_l2cache_add(vdev_t *vd);
672extern void spa_l2cache_remove(vdev_t *vd);
673extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
674extern void spa_l2cache_activate(vdev_t *vd);
675extern void spa_l2cache_drop(spa_t *spa);
676
677/* scanning */
678extern int spa_scan(spa_t *spa, pool_scan_func_t func);
679extern int spa_scan_stop(spa_t *spa);
680extern int spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t flag);
681
682/* spa syncing */
683extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
684extern void spa_sync_allpools(void);
685
686/* spa namespace global mutex */
687extern kmutex_t spa_namespace_lock;
688
689/*
690 * SPA configuration functions in spa_config.c
691 */
692
693#define	SPA_CONFIG_UPDATE_POOL	0
694#define	SPA_CONFIG_UPDATE_VDEVS	1
695
696extern void spa_config_sync(spa_t *, boolean_t, boolean_t);
697extern void spa_config_load(void);
698extern nvlist_t *spa_all_configs(uint64_t *);
699extern void spa_config_set(spa_t *spa, nvlist_t *config);
700extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
701    int getstats);
702extern void spa_config_update(spa_t *spa, int what);
703
704/*
705 * Miscellaneous SPA routines in spa_misc.c
706 */
707
708/* Namespace manipulation */
709extern spa_t *spa_lookup(const char *name);
710extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot);
711extern void spa_remove(spa_t *spa);
712extern spa_t *spa_next(spa_t *prev);
713
714/* Refcount functions */
715extern void spa_open_ref(spa_t *spa, void *tag);
716extern void spa_close(spa_t *spa, void *tag);
717extern void spa_async_close(spa_t *spa, void *tag);
718extern boolean_t spa_refcount_zero(spa_t *spa);
719
720#define	SCL_NONE	0x00
721#define	SCL_CONFIG	0x01
722#define	SCL_STATE	0x02
723#define	SCL_L2ARC	0x04		/* hack until L2ARC 2.0 */
724#define	SCL_ALLOC	0x08
725#define	SCL_ZIO		0x10
726#define	SCL_FREE	0x20
727#define	SCL_VDEV	0x40
728#define	SCL_LOCKS	7
729#define	SCL_ALL		((1 << SCL_LOCKS) - 1)
730#define	SCL_STATE_ALL	(SCL_STATE | SCL_L2ARC | SCL_ZIO)
731
732/* Pool configuration locks */
733extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw);
734extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw);
735extern void spa_config_exit(spa_t *spa, int locks, void *tag);
736extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
737
738/* Pool vdev add/remove lock */
739extern uint64_t spa_vdev_enter(spa_t *spa);
740extern uint64_t spa_vdev_config_enter(spa_t *spa);
741extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg,
742    int error, char *tag);
743extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error);
744
745/* Pool vdev state change lock */
746extern void spa_vdev_state_enter(spa_t *spa, int oplock);
747extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error);
748
749/* Log state */
750typedef enum spa_log_state {
751	SPA_LOG_UNKNOWN = 0,	/* unknown log state */
752	SPA_LOG_MISSING,	/* missing log(s) */
753	SPA_LOG_CLEAR,		/* clear the log(s) */
754	SPA_LOG_GOOD,		/* log(s) are good */
755} spa_log_state_t;
756
757extern spa_log_state_t spa_get_log_state(spa_t *spa);
758extern void spa_set_log_state(spa_t *spa, spa_log_state_t state);
759extern int spa_offline_log(spa_t *spa);
760
761/* Log claim callback */
762extern void spa_claim_notify(zio_t *zio);
763
764/* Accessor functions */
765extern boolean_t spa_shutting_down(spa_t *spa);
766extern struct dsl_pool *spa_get_dsl(spa_t *spa);
767extern boolean_t spa_is_initializing(spa_t *spa);
768extern blkptr_t *spa_get_rootblkptr(spa_t *spa);
769extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
770extern void spa_altroot(spa_t *, char *, size_t);
771extern int spa_sync_pass(spa_t *spa);
772extern char *spa_name(spa_t *spa);
773extern uint64_t spa_guid(spa_t *spa);
774extern uint64_t spa_load_guid(spa_t *spa);
775extern uint64_t spa_last_synced_txg(spa_t *spa);
776extern uint64_t spa_first_txg(spa_t *spa);
777extern uint64_t spa_syncing_txg(spa_t *spa);
778extern uint64_t spa_final_dirty_txg(spa_t *spa);
779extern uint64_t spa_version(spa_t *spa);
780extern pool_state_t spa_state(spa_t *spa);
781extern spa_load_state_t spa_load_state(spa_t *spa);
782extern uint64_t spa_freeze_txg(spa_t *spa);
783extern uint64_t spa_get_worst_case_asize(spa_t *spa, uint64_t lsize);
784extern uint64_t spa_get_dspace(spa_t *spa);
785extern uint64_t spa_get_slop_space(spa_t *spa);
786extern void spa_update_dspace(spa_t *spa);
787extern uint64_t spa_version(spa_t *spa);
788extern boolean_t spa_deflate(spa_t *spa);
789extern metaslab_class_t *spa_normal_class(spa_t *spa);
790extern metaslab_class_t *spa_log_class(spa_t *spa);
791extern void spa_evicting_os_register(spa_t *, objset_t *os);
792extern void spa_evicting_os_deregister(spa_t *, objset_t *os);
793extern void spa_evicting_os_wait(spa_t *spa);
794extern int spa_max_replication(spa_t *spa);
795extern int spa_prev_software_version(spa_t *spa);
796extern int spa_busy(void);
797extern uint8_t spa_get_failmode(spa_t *spa);
798extern boolean_t spa_suspended(spa_t *spa);
799extern uint64_t spa_bootfs(spa_t *spa);
800extern uint64_t spa_delegation(spa_t *spa);
801extern objset_t *spa_meta_objset(spa_t *spa);
802extern uint64_t spa_deadman_synctime(spa_t *spa);
803
804/* Miscellaneous support routines */
805extern void spa_activate_mos_feature(spa_t *spa, const char *feature,
806    dmu_tx_t *tx);
807extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature);
808extern int spa_rename(const char *oldname, const char *newname);
809extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid);
810extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid);
811extern char *spa_strdup(const char *);
812extern void spa_strfree(char *);
813extern uint64_t spa_get_random(uint64_t range);
814extern uint64_t spa_generate_guid(spa_t *spa);
815extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp);
816extern void spa_freeze(spa_t *spa);
817extern int spa_change_guid(spa_t *spa);
818extern void spa_upgrade(spa_t *spa, uint64_t version);
819extern void spa_evict_all(void);
820extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid,
821    boolean_t l2cache);
822extern boolean_t spa_has_spare(spa_t *, uint64_t guid);
823extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva);
824extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp);
825extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp);
826extern boolean_t spa_has_slogs(spa_t *spa);
827extern boolean_t spa_is_root(spa_t *spa);
828extern boolean_t spa_writeable(spa_t *spa);
829extern boolean_t spa_has_pending_synctask(spa_t *spa);
830extern int spa_maxblocksize(spa_t *spa);
831extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp);
832
833extern int spa_mode(spa_t *spa);
834extern uint64_t zfs_strtonum(const char *str, char **nptr);
835
836extern char *spa_his_ievent_table[];
837
838extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx);
839extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read,
840    char *his_buf);
841extern int spa_history_log(spa_t *spa, const char *his_buf);
842extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl);
843extern void spa_history_log_version(spa_t *spa, const char *operation);
844extern void spa_history_log_internal(spa_t *spa, const char *operation,
845    dmu_tx_t *tx, const char *fmt, ...);
846extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op,
847    dmu_tx_t *tx, const char *fmt, ...);
848extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
849    dmu_tx_t *tx, const char *fmt, ...);
850
851/* error handling */
852struct zbookmark_phys;
853extern void spa_log_error(spa_t *spa, zio_t *zio);
854extern void zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd,
855    zio_t *zio, uint64_t stateoroffset, uint64_t length);
856extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
857extern void zfs_post_state_change(spa_t *spa, vdev_t *vd);
858extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
859extern uint64_t spa_get_errlog_size(spa_t *spa);
860extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
861extern void spa_errlog_rotate(spa_t *spa);
862extern void spa_errlog_drain(spa_t *spa);
863extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
864extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
865
866/* vdev cache */
867extern void vdev_cache_stat_init(void);
868extern void vdev_cache_stat_fini(void);
869
870/* Initialization and termination */
871extern void spa_init(int flags);
872extern void spa_fini(void);
873extern void spa_boot_init();
874
875/* properties */
876extern int spa_prop_set(spa_t *spa, nvlist_t *nvp);
877extern int spa_prop_get(spa_t *spa, nvlist_t **nvp);
878extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx);
879extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t);
880
881/* asynchronous event notification */
882extern void spa_event_notify(spa_t *spa, vdev_t *vdev, nvlist_t *hist_nvl,
883    const char *name);
884
885#ifdef ZFS_DEBUG
886#define	dprintf_bp(bp, fmt, ...) do {				\
887	if (zfs_flags & ZFS_DEBUG_DPRINTF) {			\
888	char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP);	\
889	snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp));	\
890	dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf);		\
891	kmem_free(__blkbuf, BP_SPRINTF_LEN);			\
892	} \
893_NOTE(CONSTCOND) } while (0)
894#else
895#define	dprintf_bp(bp, fmt, ...)
896#endif
897
898extern boolean_t spa_debug_enabled(spa_t *spa);
899#define	spa_dbgmsg(spa, ...)			\
900{						\
901	if (spa_debug_enabled(spa))		\
902		zfs_dbgmsg(__VA_ARGS__);	\
903}
904
905extern int spa_mode_global;			/* mode, e.g. FREAD | FWRITE */
906
907#ifdef	__cplusplus
908}
909#endif
910
911#endif	/* _SYS_SPA_H */
912