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, 2019 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 * Copyright (c) 2017, Intel Corporation.
31 */
32
33#ifndef _SYS_SPA_H
34#define	_SYS_SPA_H
35
36#include <sys/avl.h>
37#include <sys/zfs_context.h>
38#include <sys/nvpair.h>
39#include <sys/sysevent.h>
40#include <sys/sysmacros.h>
41#include <sys/types.h>
42#include <sys/fs/zfs.h>
43#include <sys/dmu.h>
44#include <sys/space_map.h>
45#include <sys/bitops.h>
46
47#ifdef	__cplusplus
48extern "C" {
49#endif
50
51/*
52 * Forward references that lots of things need.
53 */
54typedef struct spa spa_t;
55typedef struct vdev vdev_t;
56typedef struct metaslab metaslab_t;
57typedef struct metaslab_group metaslab_group_t;
58typedef struct metaslab_class metaslab_class_t;
59typedef struct zio zio_t;
60typedef struct zilog zilog_t;
61typedef struct spa_aux_vdev spa_aux_vdev_t;
62typedef struct ddt ddt_t;
63typedef struct ddt_entry ddt_entry_t;
64struct dsl_pool;
65struct dsl_dataset;
66struct dsl_crypto_params;
67
68/*
69 * We currently support block sizes from 512 bytes to 16MB.
70 * The benefits of larger blocks, and thus larger IO, need to be weighed
71 * against the cost of COWing a giant block to modify one byte, and the
72 * large latency of reading or writing a large block.
73 *
74 * Note that although blocks up to 16MB are supported, the recordsize
75 * property can not be set larger than zfs_max_recordsize (default 1MB).
76 * See the comment near zfs_max_recordsize in dsl_dataset.c for details.
77 *
78 * Note that although the LSIZE field of the blkptr_t can store sizes up
79 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to
80 * 32MB - 512 bytes.  Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
81 */
82#define	SPA_MINBLOCKSHIFT	9
83#define	SPA_OLD_MAXBLOCKSHIFT	17
84#define	SPA_MAXBLOCKSHIFT	24
85#define	SPA_MINBLOCKSIZE	(1ULL << SPA_MINBLOCKSHIFT)
86#define	SPA_OLD_MAXBLOCKSIZE	(1ULL << SPA_OLD_MAXBLOCKSHIFT)
87#define	SPA_MAXBLOCKSIZE	(1ULL << SPA_MAXBLOCKSHIFT)
88
89/*
90 * Alignment Shift (ashift) is an immutable, internal top-level vdev property
91 * which can only be set at vdev creation time. Physical writes are always done
92 * according to it, which makes 2^ashift the smallest possible IO on a vdev.
93 *
94 * We currently allow values ranging from 512 bytes (2^9 = 512) to 64 KiB
95 * (2^16 = 65,536).
96 */
97#define	ASHIFT_MIN		9
98#define	ASHIFT_MAX		16
99
100/*
101 * Size of block to hold the configuration data (a packed nvlist)
102 */
103#define	SPA_CONFIG_BLOCKSIZE	(1ULL << 14)
104
105/*
106 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
107 * The ASIZE encoding should be at least 64 times larger (6 more bits)
108 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
109 * overhead, three DVAs per bp, plus one more bit in case we do anything
110 * else that expands the ASIZE.
111 */
112#define	SPA_LSIZEBITS		16	/* LSIZE up to 32M (2^16 * 512)	*/
113#define	SPA_PSIZEBITS		16	/* PSIZE up to 32M (2^16 * 512)	*/
114#define	SPA_ASIZEBITS		24	/* ASIZE up to 64 times larger	*/
115
116#define	SPA_COMPRESSBITS	7
117#define	SPA_VDEVBITS		24
118
119/*
120 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
121 * The members of the dva_t should be considered opaque outside the SPA.
122 */
123typedef struct dva {
124	uint64_t	dva_word[2];
125} dva_t;
126
127/*
128 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
129 */
130typedef struct zio_cksum {
131	uint64_t	zc_word[4];
132} zio_cksum_t;
133
134/*
135 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept
136 * secret and is suitable for use in MAC algorithms as the key.
137 */
138typedef struct zio_cksum_salt {
139	uint8_t		zcs_bytes[32];
140} zio_cksum_salt_t;
141
142/*
143 * Each block is described by its DVAs, time of birth, checksum, etc.
144 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
145 *
146 *	64	56	48	40	32	24	16	8	0
147 *	+-------+-------+-------+-------+-------+-------+-------+-------+
148 * 0	|  pad  |	  vdev1         | GRID  |	  ASIZE		|
149 *	+-------+-------+-------+-------+-------+-------+-------+-------+
150 * 1	|G|			 offset1				|
151 *	+-------+-------+-------+-------+-------+-------+-------+-------+
152 * 2	|  pad  |	  vdev2         | GRID  |	  ASIZE		|
153 *	+-------+-------+-------+-------+-------+-------+-------+-------+
154 * 3	|G|			 offset2				|
155 *	+-------+-------+-------+-------+-------+-------+-------+-------+
156 * 4	|  pad  |	  vdev3         | GRID  |	  ASIZE		|
157 *	+-------+-------+-------+-------+-------+-------+-------+-------+
158 * 5	|G|			 offset3				|
159 *	+-------+-------+-------+-------+-------+-------+-------+-------+
160 * 6	|BDX|lvl| type	| cksum |E| comp|    PSIZE	|     LSIZE	|
161 *	+-------+-------+-------+-------+-------+-------+-------+-------+
162 * 7	|			padding					|
163 *	+-------+-------+-------+-------+-------+-------+-------+-------+
164 * 8	|			padding					|
165 *	+-------+-------+-------+-------+-------+-------+-------+-------+
166 * 9	|			physical birth txg			|
167 *	+-------+-------+-------+-------+-------+-------+-------+-------+
168 * a	|			logical birth txg			|
169 *	+-------+-------+-------+-------+-------+-------+-------+-------+
170 * b	|			fill count				|
171 *	+-------+-------+-------+-------+-------+-------+-------+-------+
172 * c	|			checksum[0]				|
173 *	+-------+-------+-------+-------+-------+-------+-------+-------+
174 * d	|			checksum[1]				|
175 *	+-------+-------+-------+-------+-------+-------+-------+-------+
176 * e	|			checksum[2]				|
177 *	+-------+-------+-------+-------+-------+-------+-------+-------+
178 * f	|			checksum[3]				|
179 *	+-------+-------+-------+-------+-------+-------+-------+-------+
180 *
181 * Legend:
182 *
183 * vdev		virtual device ID
184 * offset	offset into virtual device
185 * LSIZE	logical size
186 * PSIZE	physical size (after compression)
187 * ASIZE	allocated size (including RAID-Z parity and gang block headers)
188 * GRID		RAID-Z layout information (reserved for future use)
189 * cksum	checksum function
190 * comp		compression function
191 * G		gang block indicator
192 * B		byteorder (endianness)
193 * D		dedup
194 * X		encryption
195 * E		blkptr_t contains embedded data (see below)
196 * lvl		level of indirection
197 * type		DMU object type
198 * phys birth	txg when dva[0] was written; zero if same as logical birth txg
199 *              note that typically all the dva's would be written in this
200 *              txg, but they could be different if they were moved by
201 *              device removal.
202 * log. birth	transaction group in which the block was logically born
203 * fill count	number of non-zero blocks under this bp
204 * checksum[4]	256-bit checksum of the data this bp describes
205 */
206
207/*
208 * The blkptr_t's of encrypted blocks also need to store the encryption
209 * parameters so that the block can be decrypted. This layout is as follows:
210 *
211 *	64	56	48	40	32	24	16	8	0
212 *	+-------+-------+-------+-------+-------+-------+-------+-------+
213 * 0	|		vdev1		| GRID  |	  ASIZE		|
214 *	+-------+-------+-------+-------+-------+-------+-------+-------+
215 * 1	|G|			 offset1				|
216 *	+-------+-------+-------+-------+-------+-------+-------+-------+
217 * 2	|		vdev2		| GRID  |	  ASIZE		|
218 *	+-------+-------+-------+-------+-------+-------+-------+-------+
219 * 3	|G|			 offset2				|
220 *	+-------+-------+-------+-------+-------+-------+-------+-------+
221 * 4	|			salt					|
222 *	+-------+-------+-------+-------+-------+-------+-------+-------+
223 * 5	|			IV1					|
224 *	+-------+-------+-------+-------+-------+-------+-------+-------+
225 * 6	|BDX|lvl| type	| cksum |E| comp|    PSIZE	|     LSIZE	|
226 *	+-------+-------+-------+-------+-------+-------+-------+-------+
227 * 7	|			padding					|
228 *	+-------+-------+-------+-------+-------+-------+-------+-------+
229 * 8	|			padding					|
230 *	+-------+-------+-------+-------+-------+-------+-------+-------+
231 * 9	|			physical birth txg			|
232 *	+-------+-------+-------+-------+-------+-------+-------+-------+
233 * a	|			logical birth txg			|
234 *	+-------+-------+-------+-------+-------+-------+-------+-------+
235 * b	|		IV2		|	    fill count		|
236 *	+-------+-------+-------+-------+-------+-------+-------+-------+
237 * c	|			checksum[0]				|
238 *	+-------+-------+-------+-------+-------+-------+-------+-------+
239 * d	|			checksum[1]				|
240 *	+-------+-------+-------+-------+-------+-------+-------+-------+
241 * e	|			MAC[0]					|
242 *	+-------+-------+-------+-------+-------+-------+-------+-------+
243 * f	|			MAC[1]					|
244 *	+-------+-------+-------+-------+-------+-------+-------+-------+
245 *
246 * Legend:
247 *
248 * salt		Salt for generating encryption keys
249 * IV1		First 64 bits of encryption IV
250 * X		Block requires encryption handling (set to 1)
251 * E		blkptr_t contains embedded data (set to 0, see below)
252 * fill count	number of non-zero blocks under this bp (truncated to 32 bits)
253 * IV2		Last 32 bits of encryption IV
254 * checksum[2]	128-bit checksum of the data this bp describes
255 * MAC[2]	128-bit message authentication code for this data
256 *
257 * The X bit being set indicates that this block is one of 3 types. If this is
258 * a level 0 block with an encrypted object type, the block is encrypted
259 * (see BP_IS_ENCRYPTED()). If this is a level 0 block with an unencrypted
260 * object type, this block is authenticated with an HMAC (see
261 * BP_IS_AUTHENTICATED()). Otherwise (if level > 0), this bp will use the MAC
262 * words to store a checksum-of-MACs from the level below (see
263 * BP_HAS_INDIRECT_MAC_CKSUM()). For convenience in the code, BP_IS_PROTECTED()
264 * refers to both encrypted and authenticated blocks and BP_USES_CRYPT()
265 * refers to any of these 3 kinds of blocks.
266 *
267 * The additional encryption parameters are the salt, IV, and MAC which are
268 * explained in greater detail in the block comment at the top of zio_crypt.c.
269 * The MAC occupies half of the checksum space since it serves a very similar
270 * purpose: to prevent data corruption on disk. The only functional difference
271 * is that the checksum is used to detect on-disk corruption whether or not the
272 * encryption key is loaded and the MAC provides additional protection against
273 * malicious disk tampering. We use the 3rd DVA to store the salt and first
274 * 64 bits of the IV. As a result encrypted blocks can only have 2 copies
275 * maximum instead of the normal 3. The last 32 bits of the IV are stored in
276 * the upper bits of what is usually the fill count. Note that only blocks at
277 * level 0 or -2 are ever encrypted, which allows us to guarantee that these
278 * 32 bits are not trampled over by other code (see zio_crypt.c for details).
279 * The salt and IV are not used for authenticated bps or bps with an indirect
280 * MAC checksum, so these blocks can utilize all 3 DVAs and the full 64 bits
281 * for the fill count.
282 */
283
284/*
285 * "Embedded" blkptr_t's don't actually point to a block, instead they
286 * have a data payload embedded in the blkptr_t itself.  See the comment
287 * in blkptr.c for more details.
288 *
289 * The blkptr_t is laid out as follows:
290 *
291 *	64	56	48	40	32	24	16	8	0
292 *	+-------+-------+-------+-------+-------+-------+-------+-------+
293 * 0	|      payload                                                  |
294 * 1	|      payload                                                  |
295 * 2	|      payload                                                  |
296 * 3	|      payload                                                  |
297 * 4	|      payload                                                  |
298 * 5	|      payload                                                  |
299 *	+-------+-------+-------+-------+-------+-------+-------+-------+
300 * 6	|BDX|lvl| type	| etype |E| comp| PSIZE|              LSIZE	|
301 *	+-------+-------+-------+-------+-------+-------+-------+-------+
302 * 7	|      payload                                                  |
303 * 8	|      payload                                                  |
304 * 9	|      payload                                                  |
305 *	+-------+-------+-------+-------+-------+-------+-------+-------+
306 * a	|			logical birth txg			|
307 *	+-------+-------+-------+-------+-------+-------+-------+-------+
308 * b	|      payload                                                  |
309 * c	|      payload                                                  |
310 * d	|      payload                                                  |
311 * e	|      payload                                                  |
312 * f	|      payload                                                  |
313 *	+-------+-------+-------+-------+-------+-------+-------+-------+
314 *
315 * Legend:
316 *
317 * payload		contains the embedded data
318 * B (byteorder)	byteorder (endianness)
319 * D (dedup)		padding (set to zero)
320 * X			encryption (set to zero; see above)
321 * E (embedded)		set to one
322 * lvl			indirection level
323 * type			DMU object type
324 * etype		how to interpret embedded data (BP_EMBEDDED_TYPE_*)
325 * comp			compression function of payload
326 * PSIZE		size of payload after compression, in bytes
327 * LSIZE		logical size of payload, in bytes
328 *			note that 25 bits is enough to store the largest
329 *			"normal" BP's LSIZE (2^16 * 2^9) in bytes
330 * log. birth		transaction group in which the block was logically born
331 *
332 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
333 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
334 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
335 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
336 * BP's so the BP_SET_* macros can be used with them.  etype, PSIZE, LSIZE must
337 * be set with the BPE_SET_* macros.  BP_SET_EMBEDDED() should be called before
338 * other macros, as they assert that they are only used on BP's of the correct
339 * "embedded-ness". Encrypted blkptr_t's cannot be embedded because they use
340 * the payload space for encryption parameters (see the comment above on
341 * how encryption parameters are stored).
342 */
343
344#define	BPE_GET_ETYPE(bp)	\
345	(ASSERT(BP_IS_EMBEDDED(bp)), \
346	BF64_GET((bp)->blk_prop, 40, 8))
347#define	BPE_SET_ETYPE(bp, t)	do { \
348	ASSERT(BP_IS_EMBEDDED(bp)); \
349	BF64_SET((bp)->blk_prop, 40, 8, t); \
350_NOTE(CONSTCOND) } while (0)
351
352#define	BPE_GET_LSIZE(bp)	\
353	(ASSERT(BP_IS_EMBEDDED(bp)), \
354	BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
355#define	BPE_SET_LSIZE(bp, x)	do { \
356	ASSERT(BP_IS_EMBEDDED(bp)); \
357	BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
358_NOTE(CONSTCOND) } while (0)
359
360#define	BPE_GET_PSIZE(bp)	\
361	(ASSERT(BP_IS_EMBEDDED(bp)), \
362	BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
363#define	BPE_SET_PSIZE(bp, x)	do { \
364	ASSERT(BP_IS_EMBEDDED(bp)); \
365	BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x);	\
366_NOTE(CONSTCOND) } while (0)
367
368typedef enum bp_embedded_type {
369	BP_EMBEDDED_TYPE_DATA,
370	BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */
371	NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED
372} bp_embedded_type_t;
373
374#define	BPE_NUM_WORDS 14
375#define	BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
376#define	BPE_IS_PAYLOADWORD(bp, wp) \
377	((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
378
379#define	SPA_BLKPTRSHIFT	7		/* blkptr_t is 128 bytes	*/
380#define	SPA_DVAS_PER_BP	3		/* Number of DVAs in a bp	*/
381#define	SPA_SYNC_MIN_VDEVS 3		/* min vdevs to update during sync */
382
383/*
384 * A block is a hole when it has either 1) never been written to, or
385 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
386 * without physically allocating disk space. Holes are represented in the
387 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is
388 * done through the BP_IS_HOLE macro. For holes, the logical size, level,
389 * DMU object type, and birth times are all also stored for holes that
390 * were written to at some point (i.e. were punched after having been filled).
391 */
392typedef struct blkptr {
393	dva_t		blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
394	uint64_t	blk_prop;	/* size, compression, type, etc	    */
395	uint64_t	blk_pad[2];	/* Extra space for the future	    */
396	uint64_t	blk_phys_birth;	/* txg when block was allocated	    */
397	uint64_t	blk_birth;	/* transaction group at birth	    */
398	uint64_t	blk_fill;	/* fill count			    */
399	zio_cksum_t	blk_cksum;	/* 256-bit checksum		    */
400} blkptr_t;
401
402/*
403 * Macros to get and set fields in a bp or DVA.
404 */
405#define	DVA_GET_ASIZE(dva)	\
406	BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
407#define	DVA_SET_ASIZE(dva, x)	\
408	BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
409	SPA_MINBLOCKSHIFT, 0, x)
410
411#define	DVA_GET_GRID(dva)	BF64_GET((dva)->dva_word[0], 24, 8)
412#define	DVA_SET_GRID(dva, x)	BF64_SET((dva)->dva_word[0], 24, 8, x)
413
414#define	DVA_GET_VDEV(dva)	BF64_GET((dva)->dva_word[0], 32, SPA_VDEVBITS)
415#define	DVA_SET_VDEV(dva, x)	\
416	BF64_SET((dva)->dva_word[0], 32, SPA_VDEVBITS, x)
417
418#define	DVA_GET_OFFSET(dva)	\
419	BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
420#define	DVA_SET_OFFSET(dva, x)	\
421	BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
422
423#define	DVA_GET_GANG(dva)	BF64_GET((dva)->dva_word[1], 63, 1)
424#define	DVA_SET_GANG(dva, x)	BF64_SET((dva)->dva_word[1], 63, 1, x)
425
426#define	BP_GET_LSIZE(bp)	\
427	(BP_IS_EMBEDDED(bp) ?	\
428	(BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
429	BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
430#define	BP_SET_LSIZE(bp, x)	do { \
431	ASSERT(!BP_IS_EMBEDDED(bp)); \
432	BF64_SET_SB((bp)->blk_prop, \
433	    0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
434_NOTE(CONSTCOND) } while (0)
435
436#define	BP_GET_PSIZE(bp)	\
437	(BP_IS_EMBEDDED(bp) ? 0 : \
438	BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1))
439#define	BP_SET_PSIZE(bp, x)	do { \
440	ASSERT(!BP_IS_EMBEDDED(bp)); \
441	BF64_SET_SB((bp)->blk_prop, \
442	    16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
443_NOTE(CONSTCOND) } while (0)
444
445#define	BP_GET_COMPRESS(bp)		\
446	BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS)
447#define	BP_SET_COMPRESS(bp, x)		\
448	BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x)
449
450#define	BP_IS_EMBEDDED(bp)		BF64_GET((bp)->blk_prop, 39, 1)
451#define	BP_SET_EMBEDDED(bp, x)		BF64_SET((bp)->blk_prop, 39, 1, x)
452
453#define	BP_GET_CHECKSUM(bp)		\
454	(BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \
455	BF64_GET((bp)->blk_prop, 40, 8))
456#define	BP_SET_CHECKSUM(bp, x)		do { \
457	ASSERT(!BP_IS_EMBEDDED(bp)); \
458	BF64_SET((bp)->blk_prop, 40, 8, x); \
459_NOTE(CONSTCOND) } while (0)
460
461#define	BP_GET_TYPE(bp)			BF64_GET((bp)->blk_prop, 48, 8)
462#define	BP_SET_TYPE(bp, x)		BF64_SET((bp)->blk_prop, 48, 8, x)
463
464#define	BP_GET_LEVEL(bp)		BF64_GET((bp)->blk_prop, 56, 5)
465#define	BP_SET_LEVEL(bp, x)		BF64_SET((bp)->blk_prop, 56, 5, x)
466
467/* encrypted, authenticated, and MAC cksum bps use the same bit */
468#define	BP_USES_CRYPT(bp)		BF64_GET((bp)->blk_prop, 61, 1)
469#define	BP_SET_CRYPT(bp, x)		BF64_SET((bp)->blk_prop, 61, 1, x)
470
471#define	BP_IS_ENCRYPTED(bp)			\
472	(BP_USES_CRYPT(bp) &&			\
473	BP_GET_LEVEL(bp) == 0 &&		\
474	DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp)))
475
476#define	BP_IS_AUTHENTICATED(bp)			\
477	(BP_USES_CRYPT(bp) &&			\
478	BP_GET_LEVEL(bp) == 0 &&		\
479	!DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp)))
480
481#define	BP_HAS_INDIRECT_MAC_CKSUM(bp)		\
482	(BP_USES_CRYPT(bp) && BP_GET_LEVEL(bp) > 0)
483
484#define	BP_IS_PROTECTED(bp)			\
485	(BP_IS_ENCRYPTED(bp) || BP_IS_AUTHENTICATED(bp))
486
487#define	BP_GET_DEDUP(bp)		BF64_GET((bp)->blk_prop, 62, 1)
488#define	BP_SET_DEDUP(bp, x)		BF64_SET((bp)->blk_prop, 62, 1, x)
489
490#define	BP_GET_BYTEORDER(bp)		BF64_GET((bp)->blk_prop, 63, 1)
491#define	BP_SET_BYTEORDER(bp, x)		BF64_SET((bp)->blk_prop, 63, 1, x)
492
493#define	BP_PHYSICAL_BIRTH(bp)		\
494	(BP_IS_EMBEDDED(bp) ? 0 : \
495	(bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
496
497#define	BP_SET_BIRTH(bp, logical, physical)	\
498{						\
499	ASSERT(!BP_IS_EMBEDDED(bp));		\
500	(bp)->blk_birth = (logical);		\
501	(bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
502}
503
504#define	BP_GET_FILL(bp)				\
505	((BP_IS_ENCRYPTED(bp)) ? BF64_GET((bp)->blk_fill, 0, 32) : \
506	((BP_IS_EMBEDDED(bp)) ? 1 : (bp)->blk_fill))
507
508#define	BP_SET_FILL(bp, fill)			\
509{						\
510	if (BP_IS_ENCRYPTED(bp))			\
511		BF64_SET((bp)->blk_fill, 0, 32, fill); \
512	else					\
513		(bp)->blk_fill = fill;		\
514}
515
516#define	BP_GET_IV2(bp)				\
517	(ASSERT(BP_IS_ENCRYPTED(bp)),		\
518	BF64_GET((bp)->blk_fill, 32, 32))
519#define	BP_SET_IV2(bp, iv2)			\
520{						\
521	ASSERT(BP_IS_ENCRYPTED(bp));		\
522	BF64_SET((bp)->blk_fill, 32, 32, iv2);	\
523}
524
525#define	BP_IS_METADATA(bp)	\
526	(BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))
527
528#define	BP_GET_ASIZE(bp)	\
529	(BP_IS_EMBEDDED(bp) ? 0 : \
530	DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
531	DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
532	(DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))
533
534#define	BP_GET_UCSIZE(bp)	\
535	(BP_IS_METADATA(bp) ? BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
536
537#define	BP_GET_NDVAS(bp)	\
538	(BP_IS_EMBEDDED(bp) ? 0 : \
539	!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
540	!!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
541	(!!DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))
542
543#define	BP_COUNT_GANG(bp)	\
544	(BP_IS_EMBEDDED(bp) ? 0 : \
545	(DVA_GET_GANG(&(bp)->blk_dva[0]) + \
546	DVA_GET_GANG(&(bp)->blk_dva[1]) + \
547	(DVA_GET_GANG(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp))))
548
549#define	DVA_EQUAL(dva1, dva2)	\
550	((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
551	(dva1)->dva_word[0] == (dva2)->dva_word[0])
552
553#define	BP_EQUAL(bp1, bp2)	\
554	(BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) &&	\
555	(bp1)->blk_birth == (bp2)->blk_birth &&			\
556	DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) &&	\
557	DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) &&	\
558	DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
559
560#define	ZIO_CHECKSUM_EQUAL(zc1, zc2) \
561	(0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
562	((zc1).zc_word[1] - (zc2).zc_word[1]) | \
563	((zc1).zc_word[2] - (zc2).zc_word[2]) | \
564	((zc1).zc_word[3] - (zc2).zc_word[3])))
565
566#define	ZIO_CHECKSUM_MAC_EQUAL(zc1, zc2) \
567	(0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
568	((zc1).zc_word[1] - (zc2).zc_word[1])))
569
570#define	ZIO_CHECKSUM_IS_ZERO(zc) \
571	(0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \
572	(zc)->zc_word[2] | (zc)->zc_word[3]))
573
574#define	ZIO_CHECKSUM_BSWAP(zcp)					\
575{								\
576	(zcp)->zc_word[0] = BSWAP_64((zcp)->zc_word[0]);	\
577	(zcp)->zc_word[1] = BSWAP_64((zcp)->zc_word[1]);	\
578	(zcp)->zc_word[2] = BSWAP_64((zcp)->zc_word[2]);	\
579	(zcp)->zc_word[3] = BSWAP_64((zcp)->zc_word[3]);	\
580}
581
582
583#define	DVA_IS_VALID(dva)	(DVA_GET_ASIZE(dva) != 0)
584
585#define	ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3)	\
586{						\
587	(zcp)->zc_word[0] = w0;			\
588	(zcp)->zc_word[1] = w1;			\
589	(zcp)->zc_word[2] = w2;			\
590	(zcp)->zc_word[3] = w3;			\
591}
592
593#define	BP_IDENTITY(bp)		(ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0])
594#define	BP_IS_GANG(bp)		\
595	(BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp)))
596#define	DVA_IS_EMPTY(dva)	((dva)->dva_word[0] == 0ULL &&	\
597				(dva)->dva_word[1] == 0ULL)
598#define	BP_IS_HOLE(bp) \
599	(!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp)))
600
601/* BP_IS_RAIDZ(bp) assumes no block compression */
602#define	BP_IS_RAIDZ(bp)		(DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
603				BP_GET_PSIZE(bp))
604
605#define	BP_ZERO(bp)				\
606{						\
607	(bp)->blk_dva[0].dva_word[0] = 0;	\
608	(bp)->blk_dva[0].dva_word[1] = 0;	\
609	(bp)->blk_dva[1].dva_word[0] = 0;	\
610	(bp)->blk_dva[1].dva_word[1] = 0;	\
611	(bp)->blk_dva[2].dva_word[0] = 0;	\
612	(bp)->blk_dva[2].dva_word[1] = 0;	\
613	(bp)->blk_prop = 0;			\
614	(bp)->blk_pad[0] = 0;			\
615	(bp)->blk_pad[1] = 0;			\
616	(bp)->blk_phys_birth = 0;		\
617	(bp)->blk_birth = 0;			\
618	(bp)->blk_fill = 0;			\
619	ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0);	\
620}
621
622#ifdef _BIG_ENDIAN
623#define	ZFS_HOST_BYTEORDER	(0ULL)
624#else
625#define	ZFS_HOST_BYTEORDER	(1ULL)
626#endif
627
628#define	BP_SHOULD_BYTESWAP(bp)	(BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
629
630#define	BP_SPRINTF_LEN	400
631
632/*
633 * This macro allows code sharing between zfs, libzpool, and mdb.
634 * 'func' is either snprintf() or mdb_snprintf().
635 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
636 */
637#define	SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
638{									\
639	static const char *copyname[] =					\
640	    { "zero", "single", "double", "triple" };			\
641	int len = 0;							\
642	int copies = 0;							\
643	const char *crypt_type;						\
644	if (bp != NULL) {						\
645		if (BP_IS_ENCRYPTED(bp)) {				\
646			crypt_type = "encrypted";			\
647		} else if (BP_IS_AUTHENTICATED(bp)) {			\
648			crypt_type = "authenticated";			\
649		} else if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) {		\
650			crypt_type = "indirect-MAC";			\
651		} else {						\
652			crypt_type = "unencrypted";			\
653		}							\
654	}								\
655	if (bp == NULL) {						\
656		len += func(buf + len, size - len, "<NULL>");		\
657	} else if (BP_IS_HOLE(bp)) {					\
658		len += func(buf + len, size - len,			\
659		    "HOLE [L%llu %s] "					\
660		    "size=%llxL birth=%lluL",				\
661		    (u_longlong_t)BP_GET_LEVEL(bp),			\
662		    type,						\
663		    (u_longlong_t)BP_GET_LSIZE(bp),			\
664		    (u_longlong_t)bp->blk_birth);			\
665	} else if (BP_IS_EMBEDDED(bp)) {				\
666		len = func(buf + len, size - len,			\
667		    "EMBEDDED [L%llu %s] et=%u %s "			\
668		    "size=%llxL/%llxP birth=%lluL",			\
669		    (u_longlong_t)BP_GET_LEVEL(bp),			\
670		    type,						\
671		    (int)BPE_GET_ETYPE(bp),				\
672		    compress,						\
673		    (u_longlong_t)BPE_GET_LSIZE(bp),			\
674		    (u_longlong_t)BPE_GET_PSIZE(bp),			\
675		    (u_longlong_t)bp->blk_birth);			\
676	} else {							\
677		for (int d = 0; d < BP_GET_NDVAS(bp); d++) {		\
678			const dva_t *dva = &bp->blk_dva[d];		\
679			if (DVA_IS_VALID(dva))				\
680				copies++;				\
681			len += func(buf + len, size - len,		\
682			    "DVA[%d]=<%llu:%llx:%llx>%c", d,		\
683			    (u_longlong_t)DVA_GET_VDEV(dva),		\
684			    (u_longlong_t)DVA_GET_OFFSET(dva),		\
685			    (u_longlong_t)DVA_GET_ASIZE(dva),		\
686			    ws);					\
687		}							\
688		if (BP_IS_ENCRYPTED(bp)) {				\
689			len += func(buf + len, size - len,		\
690			    "salt=%llx iv=%llx:%llx%c",			\
691			    (u_longlong_t)bp->blk_dva[2].dva_word[0],	\
692			    (u_longlong_t)bp->blk_dva[2].dva_word[1],	\
693			    (u_longlong_t)BP_GET_IV2(bp),		\
694			    ws);					\
695		}							\
696		if (BP_IS_GANG(bp) &&					\
697		    DVA_GET_ASIZE(&bp->blk_dva[2]) <=			\
698		    DVA_GET_ASIZE(&bp->blk_dva[1]) / 2)			\
699			copies--;					\
700		len += func(buf + len, size - len,			\
701		    "[L%llu %s] %s %s %s %s %s %s %s%c"			\
702		    "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c"	\
703		    "cksum=%llx:%llx:%llx:%llx",			\
704		    (u_longlong_t)BP_GET_LEVEL(bp),			\
705		    type,						\
706		    checksum,						\
707		    compress,						\
708		    crypt_type,						\
709		    BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE",		\
710		    BP_IS_GANG(bp) ? "gang" : "contiguous",		\
711		    BP_GET_DEDUP(bp) ? "dedup" : "unique",		\
712		    copyname[copies],					\
713		    ws,							\
714		    (u_longlong_t)BP_GET_LSIZE(bp),			\
715		    (u_longlong_t)BP_GET_PSIZE(bp),			\
716		    (u_longlong_t)bp->blk_birth,			\
717		    (u_longlong_t)BP_PHYSICAL_BIRTH(bp),		\
718		    (u_longlong_t)BP_GET_FILL(bp),			\
719		    ws,							\
720		    (u_longlong_t)bp->blk_cksum.zc_word[0],		\
721		    (u_longlong_t)bp->blk_cksum.zc_word[1],		\
722		    (u_longlong_t)bp->blk_cksum.zc_word[2],		\
723		    (u_longlong_t)bp->blk_cksum.zc_word[3]);		\
724	}								\
725	ASSERT(len < size);						\
726}
727
728#define	BP_GET_BUFC_TYPE(bp)						\
729	(BP_IS_METADATA(bp) ? ARC_BUFC_METADATA : ARC_BUFC_DATA)
730
731typedef enum spa_import_type {
732	SPA_IMPORT_EXISTING,
733	SPA_IMPORT_ASSEMBLE
734} spa_import_type_t;
735
736/*
737 * Send TRIM commands in-line during normal pool operation while deleting.
738 *	OFF: no
739 *	ON: yes
740 */
741typedef enum {
742	SPA_AUTOTRIM_OFF = 0,	/* default */
743	SPA_AUTOTRIM_ON
744} spa_autotrim_t;
745
746/*
747 * Reason TRIM command was issued, used internally for accounting purposes.
748 */
749typedef enum trim_type {
750	TRIM_TYPE_MANUAL = 0,
751	TRIM_TYPE_AUTO = 1,
752} trim_type_t;
753
754/* state manipulation functions */
755extern int spa_open(const char *pool, spa_t **, void *tag);
756extern int spa_open_rewind(const char *pool, spa_t **, void *tag,
757    nvlist_t *policy, nvlist_t **config);
758extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot,
759    size_t buflen);
760extern int spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props,
761    nvlist_t *zplprops, struct dsl_crypto_params *dcp);
762extern int spa_import_rootpool(char *devpath, char *devid);
763extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props,
764    uint64_t flags);
765extern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
766extern int spa_destroy(char *pool);
767extern int spa_checkpoint(const char *pool);
768extern int spa_checkpoint_discard(const char *pool);
769extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force,
770    boolean_t hardforce);
771extern int spa_reset(char *pool);
772extern void spa_async_request(spa_t *spa, int flag);
773extern void spa_async_unrequest(spa_t *spa, int flag);
774extern void spa_async_suspend(spa_t *spa);
775extern void spa_async_resume(spa_t *spa);
776extern spa_t *spa_inject_addref(char *pool);
777extern void spa_inject_delref(spa_t *spa);
778extern void spa_scan_stat_init(spa_t *spa);
779extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps);
780
781#define	SPA_ASYNC_CONFIG_UPDATE			0x01
782#define	SPA_ASYNC_REMOVE			0x02
783#define	SPA_ASYNC_PROBE				0x04
784#define	SPA_ASYNC_RESILVER_DONE			0x08
785#define	SPA_ASYNC_RESILVER			0x10
786#define	SPA_ASYNC_AUTOEXPAND			0x20
787#define	SPA_ASYNC_REMOVE_DONE			0x40
788#define	SPA_ASYNC_REMOVE_STOP			0x80
789#define	SPA_ASYNC_INITIALIZE_RESTART		0x100
790#define	SPA_ASYNC_TRIM_RESTART			0x200
791#define	SPA_ASYNC_AUTOTRIM_RESTART		0x400
792
793/*
794 * Controls the behavior of spa_vdev_remove().
795 */
796#define	SPA_REMOVE_UNSPARE	0x01
797#define	SPA_REMOVE_DONE		0x02
798
799/* device manipulation */
800extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
801extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot,
802    int replacing);
803extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
804    int replace_done);
805extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare);
806extern boolean_t spa_vdev_remove_active(spa_t *spa);
807extern int spa_vdev_initialize(spa_t *spa, nvlist_t *nv, uint64_t cmd_type,
808    nvlist_t *vdev_errlist);
809extern int spa_vdev_trim(spa_t *spa, nvlist_t *nv, uint64_t cmd_type,
810    uint64_t rate, boolean_t partial, boolean_t secure, nvlist_t *vdev_errlist);
811extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
812extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru);
813extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
814    nvlist_t *props, boolean_t exp);
815
816/* spare state (which is global across all pools) */
817extern void spa_spare_add(vdev_t *vd);
818extern void spa_spare_remove(vdev_t *vd);
819extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt);
820extern void spa_spare_activate(vdev_t *vd);
821
822/* L2ARC state (which is global across all pools) */
823extern void spa_l2cache_add(vdev_t *vd);
824extern void spa_l2cache_remove(vdev_t *vd);
825extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
826extern void spa_l2cache_activate(vdev_t *vd);
827extern void spa_l2cache_drop(spa_t *spa);
828
829/* scanning */
830extern int spa_scan(spa_t *spa, pool_scan_func_t func);
831extern int spa_scan_stop(spa_t *spa);
832extern int spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t flag);
833
834/* spa syncing */
835extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
836extern void spa_sync_allpools(void);
837
838/* spa namespace global mutex */
839extern kmutex_t spa_namespace_lock;
840
841/*
842 * SPA configuration functions in spa_config.c
843 */
844
845#define	SPA_CONFIG_UPDATE_POOL	0
846#define	SPA_CONFIG_UPDATE_VDEVS	1
847
848extern void spa_write_cachefile(spa_t *, boolean_t, boolean_t);
849extern void spa_config_load(void);
850extern nvlist_t *spa_all_configs(uint64_t *);
851extern void spa_config_set(spa_t *spa, nvlist_t *config);
852extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
853    int getstats);
854extern void spa_config_update(spa_t *spa, int what);
855
856/*
857 * Miscellaneous SPA routines in spa_misc.c
858 */
859
860/* Namespace manipulation */
861extern spa_t *spa_lookup(const char *name);
862extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot);
863extern void spa_remove(spa_t *spa);
864extern spa_t *spa_next(spa_t *prev);
865
866/* Refcount functions */
867extern void spa_open_ref(spa_t *spa, void *tag);
868extern void spa_close(spa_t *spa, void *tag);
869extern void spa_async_close(spa_t *spa, void *tag);
870extern boolean_t spa_refcount_zero(spa_t *spa);
871
872#define	SCL_NONE	0x00
873#define	SCL_CONFIG	0x01
874#define	SCL_STATE	0x02
875#define	SCL_L2ARC	0x04		/* hack until L2ARC 2.0 */
876#define	SCL_ALLOC	0x08
877#define	SCL_ZIO		0x10
878#define	SCL_FREE	0x20
879#define	SCL_VDEV	0x40
880#define	SCL_LOCKS	7
881#define	SCL_ALL		((1 << SCL_LOCKS) - 1)
882#define	SCL_STATE_ALL	(SCL_STATE | SCL_L2ARC | SCL_ZIO)
883
884/* Assorted pool IO kstats */
885typedef struct spa_iostats {
886	kstat_named_t	trim_extents_written;
887	kstat_named_t	trim_bytes_written;
888	kstat_named_t	trim_extents_skipped;
889	kstat_named_t	trim_bytes_skipped;
890	kstat_named_t	trim_extents_failed;
891	kstat_named_t	trim_bytes_failed;
892	kstat_named_t	autotrim_extents_written;
893	kstat_named_t	autotrim_bytes_written;
894	kstat_named_t	autotrim_extents_skipped;
895	kstat_named_t	autotrim_bytes_skipped;
896	kstat_named_t	autotrim_extents_failed;
897	kstat_named_t	autotrim_bytes_failed;
898} spa_iostats_t;
899
900/* Pool configuration locks */
901extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw);
902extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw);
903extern void spa_config_exit(spa_t *spa, int locks, void *tag);
904extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
905
906/* Pool vdev add/remove lock */
907extern uint64_t spa_vdev_enter(spa_t *spa);
908extern uint64_t spa_vdev_config_enter(spa_t *spa);
909extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg,
910    int error, char *tag);
911extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error);
912
913/* Pool vdev state change lock */
914extern void spa_vdev_state_enter(spa_t *spa, int oplock);
915extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error);
916
917/* Log state */
918typedef enum spa_log_state {
919	SPA_LOG_UNKNOWN = 0,	/* unknown log state */
920	SPA_LOG_MISSING,	/* missing log(s) */
921	SPA_LOG_CLEAR,		/* clear the log(s) */
922	SPA_LOG_GOOD,		/* log(s) are good */
923} spa_log_state_t;
924
925extern spa_log_state_t spa_get_log_state(spa_t *spa);
926extern void spa_set_log_state(spa_t *spa, spa_log_state_t state);
927extern int spa_reset_logs(spa_t *spa);
928
929/* Log claim callback */
930extern void spa_claim_notify(zio_t *zio);
931
932/* Accessor functions */
933extern boolean_t spa_shutting_down(spa_t *spa);
934extern struct dsl_pool *spa_get_dsl(spa_t *spa);
935extern boolean_t spa_is_initializing(spa_t *spa);
936extern boolean_t spa_indirect_vdevs_loaded(spa_t *spa);
937extern blkptr_t *spa_get_rootblkptr(spa_t *spa);
938extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
939extern void spa_altroot(spa_t *, char *, size_t);
940extern int spa_sync_pass(spa_t *spa);
941extern char *spa_name(spa_t *spa);
942extern uint64_t spa_guid(spa_t *spa);
943extern uint64_t spa_load_guid(spa_t *spa);
944extern uint64_t spa_last_synced_txg(spa_t *spa);
945extern uint64_t spa_first_txg(spa_t *spa);
946extern uint64_t spa_syncing_txg(spa_t *spa);
947extern uint64_t spa_final_dirty_txg(spa_t *spa);
948extern uint64_t spa_version(spa_t *spa);
949extern pool_state_t spa_state(spa_t *spa);
950extern spa_load_state_t spa_load_state(spa_t *spa);
951extern uint64_t spa_freeze_txg(spa_t *spa);
952extern uint64_t spa_get_worst_case_asize(spa_t *spa, uint64_t lsize);
953extern uint64_t spa_get_dspace(spa_t *spa);
954extern uint64_t spa_get_checkpoint_space(spa_t *spa);
955extern uint64_t spa_get_slop_space(spa_t *spa);
956extern void spa_update_dspace(spa_t *spa);
957extern uint64_t spa_version(spa_t *spa);
958extern boolean_t spa_deflate(spa_t *spa);
959extern metaslab_class_t *spa_normal_class(spa_t *spa);
960extern metaslab_class_t *spa_log_class(spa_t *spa);
961extern metaslab_class_t *spa_special_class(spa_t *spa);
962extern metaslab_class_t *spa_dedup_class(spa_t *spa);
963extern metaslab_class_t *spa_preferred_class(spa_t *spa, uint64_t size,
964    dmu_object_type_t objtype, uint_t level, uint_t special_smallblk);
965
966extern void spa_evicting_os_register(spa_t *, objset_t *os);
967extern void spa_evicting_os_deregister(spa_t *, objset_t *os);
968extern void spa_evicting_os_wait(spa_t *spa);
969extern int spa_max_replication(spa_t *spa);
970extern int spa_prev_software_version(spa_t *spa);
971extern int spa_busy(void);
972extern uint8_t spa_get_failmode(spa_t *spa);
973extern boolean_t spa_suspended(spa_t *spa);
974extern uint64_t spa_bootfs(spa_t *spa);
975extern uint64_t spa_delegation(spa_t *spa);
976extern objset_t *spa_meta_objset(spa_t *spa);
977extern space_map_t *spa_syncing_log_sm(spa_t *spa);
978extern uint64_t spa_deadman_synctime(spa_t *spa);
979extern uint64_t spa_dirty_data(spa_t *spa);
980extern spa_autotrim_t spa_get_autotrim(spa_t *spa);
981
982/* Miscellaneous support routines */
983extern void spa_load_failed(spa_t *spa, const char *fmt, ...);
984extern void spa_load_note(spa_t *spa, const char *fmt, ...);
985extern void spa_activate_mos_feature(spa_t *spa, const char *feature,
986    dmu_tx_t *tx);
987extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature);
988extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid);
989extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid);
990extern char *spa_strdup(const char *);
991extern void spa_strfree(char *);
992extern uint64_t spa_get_random(uint64_t range);
993extern uint64_t spa_generate_guid(spa_t *spa);
994extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp);
995extern void spa_freeze(spa_t *spa);
996extern int spa_change_guid(spa_t *spa);
997extern void spa_upgrade(spa_t *spa, uint64_t version);
998extern void spa_evict_all(void);
999extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid,
1000    boolean_t l2cache);
1001extern boolean_t spa_has_spare(spa_t *, uint64_t guid);
1002extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva);
1003extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp);
1004extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp);
1005extern boolean_t spa_has_slogs(spa_t *spa);
1006extern boolean_t spa_is_root(spa_t *spa);
1007extern boolean_t spa_writeable(spa_t *spa);
1008extern boolean_t spa_has_pending_synctask(spa_t *spa);
1009extern int spa_maxblocksize(spa_t *spa);
1010extern int spa_maxdnodesize(spa_t *spa);
1011extern boolean_t spa_multihost(spa_t *spa);
1012extern unsigned long spa_get_hostid(void);
1013extern boolean_t spa_has_checkpoint(spa_t *spa);
1014extern boolean_t spa_importing_readonly_checkpoint(spa_t *spa);
1015extern boolean_t spa_suspend_async_destroy(spa_t *spa);
1016extern uint64_t spa_min_claim_txg(spa_t *spa);
1017extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp);
1018extern boolean_t zfs_dva_valid(spa_t *spa, const dva_t *dva,
1019    const blkptr_t *bp);
1020typedef void (*spa_remap_cb_t)(uint64_t vdev, uint64_t offset, uint64_t size,
1021    void *arg);
1022extern boolean_t spa_remap_blkptr(spa_t *spa, blkptr_t *bp,
1023    spa_remap_cb_t callback, void *arg);
1024extern uint64_t spa_get_last_removal_txg(spa_t *spa);
1025extern boolean_t spa_trust_config(spa_t *spa);
1026extern uint64_t spa_missing_tvds_allowed(spa_t *spa);
1027extern void spa_set_missing_tvds(spa_t *spa, uint64_t missing);
1028extern boolean_t spa_top_vdevs_spacemap_addressable(spa_t *spa);
1029extern uint64_t spa_total_metaslabs(spa_t *spa);
1030extern void spa_activate_allocation_classes(spa_t *, dmu_tx_t *);
1031
1032extern int spa_mode(spa_t *spa);
1033extern uint64_t zfs_strtonum(const char *str, char **nptr);
1034
1035extern char *spa_his_ievent_table[];
1036
1037extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx);
1038extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read,
1039    char *his_buf);
1040extern int spa_history_log(spa_t *spa, const char *his_buf);
1041extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl);
1042extern void spa_history_log_version(spa_t *spa, const char *operation);
1043extern void spa_history_log_internal(spa_t *spa, const char *operation,
1044    dmu_tx_t *tx, const char *fmt, ...);
1045extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op,
1046    dmu_tx_t *tx, const char *fmt, ...);
1047extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
1048    dmu_tx_t *tx, const char *fmt, ...);
1049
1050/* error handling */
1051struct zbookmark_phys;
1052extern void spa_log_error(spa_t *spa, const struct zbookmark_phys *zb);
1053extern void zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd,
1054    const struct zbookmark_phys *zb, struct zio *zio, uint64_t stateoroffset,
1055    uint64_t length);
1056extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
1057extern void zfs_post_state_change(spa_t *spa, vdev_t *vd);
1058extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
1059extern uint64_t spa_get_errlog_size(spa_t *spa);
1060extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
1061extern void spa_errlog_rotate(spa_t *spa);
1062extern void spa_errlog_drain(spa_t *spa);
1063extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
1064extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
1065
1066/* vdev cache */
1067extern void vdev_cache_stat_init(void);
1068extern void vdev_cache_stat_fini(void);
1069
1070/* vdev mirror */
1071extern void vdev_mirror_stat_init(void);
1072extern void vdev_mirror_stat_fini(void);
1073
1074/* Initialization and termination */
1075extern void spa_init(int flags);
1076extern void spa_fini(void);
1077extern void spa_boot_init(void);
1078
1079/* properties */
1080extern int spa_prop_set(spa_t *spa, nvlist_t *nvp);
1081extern int spa_prop_get(spa_t *spa, nvlist_t **nvp);
1082extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx);
1083extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t);
1084
1085/* asynchronous event notification */
1086extern void spa_event_notify(spa_t *spa, vdev_t *vdev, nvlist_t *hist_nvl,
1087    const char *name);
1088extern sysevent_t *spa_event_create(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl,
1089    const char *name);
1090extern void spa_event_post(sysevent_t *ev);
1091extern void spa_event_discard(sysevent_t *ev);
1092
1093#ifdef ZFS_DEBUG
1094#define	dprintf_bp(bp, fmt, ...) do {				\
1095	if (zfs_flags & ZFS_DEBUG_DPRINTF) {			\
1096	char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP);	\
1097	snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp));	\
1098	dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf);		\
1099	kmem_free(__blkbuf, BP_SPRINTF_LEN);			\
1100	} \
1101_NOTE(CONSTCOND) } while (0)
1102#else
1103#define	dprintf_bp(bp, fmt, ...)
1104#endif
1105
1106extern int spa_mode_global;			/* mode, e.g. FREAD | FWRITE */
1107
1108#ifdef	__cplusplus
1109}
1110#endif
1111
1112#endif	/* _SYS_SPA_H */
1113