spa.h revision eb633035c80613ec93d62f90482837adaaf21a0a
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, 2018 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
45#ifdef	__cplusplus
46extern "C" {
47#endif
48
49/*
50 * Forward references that lots of things need.
51 */
52typedef struct spa spa_t;
53typedef struct vdev vdev_t;
54typedef struct metaslab metaslab_t;
55typedef struct metaslab_group metaslab_group_t;
56typedef struct metaslab_class metaslab_class_t;
57typedef struct zio zio_t;
58typedef struct zilog zilog_t;
59typedef struct spa_aux_vdev spa_aux_vdev_t;
60typedef struct ddt ddt_t;
61typedef struct ddt_entry ddt_entry_t;
62struct dsl_pool;
63struct dsl_dataset;
64struct dsl_crypto_params;
65
66/*
67 * General-purpose 32-bit and 64-bit bitfield encodings.
68 */
69#define	BF32_DECODE(x, low, len)	P2PHASE((x) >> (low), 1U << (len))
70#define	BF64_DECODE(x, low, len)	P2PHASE((x) >> (low), 1ULL << (len))
71#define	BF32_ENCODE(x, low, len)	(P2PHASE((x), 1U << (len)) << (low))
72#define	BF64_ENCODE(x, low, len)	(P2PHASE((x), 1ULL << (len)) << (low))
73
74#define	BF32_GET(x, low, len)		BF32_DECODE(x, low, len)
75#define	BF64_GET(x, low, len)		BF64_DECODE(x, low, len)
76
77#define	BF32_SET(x, low, len, val) do { \
78	ASSERT3U(val, <, 1U << (len)); \
79	ASSERT3U(low + len, <=, 32); \
80	(x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \
81_NOTE(CONSTCOND) } while (0)
82
83#define	BF64_SET(x, low, len, val) do { \
84	ASSERT3U(val, <, 1ULL << (len)); \
85	ASSERT3U(low + len, <=, 64); \
86	((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \
87_NOTE(CONSTCOND) } while (0)
88
89#define	BF32_GET_SB(x, low, len, shift, bias)	\
90	((BF32_GET(x, low, len) + (bias)) << (shift))
91#define	BF64_GET_SB(x, low, len, shift, bias)	\
92	((BF64_GET(x, low, len) + (bias)) << (shift))
93
94#define	BF32_SET_SB(x, low, len, shift, bias, val) do { \
95	ASSERT(IS_P2ALIGNED(val, 1U << shift)); \
96	ASSERT3S((val) >> (shift), >=, bias); \
97	BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \
98_NOTE(CONSTCOND) } while (0)
99#define	BF64_SET_SB(x, low, len, shift, bias, val) do { \
100	ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \
101	ASSERT3S((val) >> (shift), >=, bias); \
102	BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \
103_NOTE(CONSTCOND) } while (0)
104
105/*
106 * We currently support block sizes from 512 bytes to 16MB.
107 * The benefits of larger blocks, and thus larger IO, need to be weighed
108 * against the cost of COWing a giant block to modify one byte, and the
109 * large latency of reading or writing a large block.
110 *
111 * Note that although blocks up to 16MB are supported, the recordsize
112 * property can not be set larger than zfs_max_recordsize (default 1MB).
113 * See the comment near zfs_max_recordsize in dsl_dataset.c for details.
114 *
115 * Note that although the LSIZE field of the blkptr_t can store sizes up
116 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to
117 * 32MB - 512 bytes.  Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
118 */
119#define	SPA_MINBLOCKSHIFT	9
120#define	SPA_OLD_MAXBLOCKSHIFT	17
121#define	SPA_MAXBLOCKSHIFT	24
122#define	SPA_MINBLOCKSIZE	(1ULL << SPA_MINBLOCKSHIFT)
123#define	SPA_OLD_MAXBLOCKSIZE	(1ULL << SPA_OLD_MAXBLOCKSHIFT)
124#define	SPA_MAXBLOCKSIZE	(1ULL << SPA_MAXBLOCKSHIFT)
125
126/*
127 * Size of block to hold the configuration data (a packed nvlist)
128 */
129#define	SPA_CONFIG_BLOCKSIZE	(1ULL << 14)
130
131/*
132 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
133 * The ASIZE encoding should be at least 64 times larger (6 more bits)
134 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
135 * overhead, three DVAs per bp, plus one more bit in case we do anything
136 * else that expands the ASIZE.
137 */
138#define	SPA_LSIZEBITS		16	/* LSIZE up to 32M (2^16 * 512)	*/
139#define	SPA_PSIZEBITS		16	/* PSIZE up to 32M (2^16 * 512)	*/
140#define	SPA_ASIZEBITS		24	/* ASIZE up to 64 times larger	*/
141
142#define	SPA_COMPRESSBITS	7
143#define	SPA_VDEVBITS		24
144
145/*
146 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
147 * The members of the dva_t should be considered opaque outside the SPA.
148 */
149typedef struct dva {
150	uint64_t	dva_word[2];
151} dva_t;
152
153/*
154 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
155 */
156typedef struct zio_cksum {
157	uint64_t	zc_word[4];
158} zio_cksum_t;
159
160/*
161 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept
162 * secret and is suitable for use in MAC algorithms as the key.
163 */
164typedef struct zio_cksum_salt {
165	uint8_t		zcs_bytes[32];
166} zio_cksum_salt_t;
167
168/*
169 * Each block is described by its DVAs, time of birth, checksum, etc.
170 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
171 *
172 *	64	56	48	40	32	24	16	8	0
173 *	+-------+-------+-------+-------+-------+-------+-------+-------+
174 * 0	|  pad  |	  vdev1         | GRID  |	  ASIZE		|
175 *	+-------+-------+-------+-------+-------+-------+-------+-------+
176 * 1	|G|			 offset1				|
177 *	+-------+-------+-------+-------+-------+-------+-------+-------+
178 * 2	|  pad  |	  vdev2         | GRID  |	  ASIZE		|
179 *	+-------+-------+-------+-------+-------+-------+-------+-------+
180 * 3	|G|			 offset2				|
181 *	+-------+-------+-------+-------+-------+-------+-------+-------+
182 * 4	|  pad  |	  vdev3         | GRID  |	  ASIZE		|
183 *	+-------+-------+-------+-------+-------+-------+-------+-------+
184 * 5	|G|			 offset3				|
185 *	+-------+-------+-------+-------+-------+-------+-------+-------+
186 * 6	|BDX|lvl| type	| cksum |E| comp|    PSIZE	|     LSIZE	|
187 *	+-------+-------+-------+-------+-------+-------+-------+-------+
188 * 7	|			padding					|
189 *	+-------+-------+-------+-------+-------+-------+-------+-------+
190 * 8	|			padding					|
191 *	+-------+-------+-------+-------+-------+-------+-------+-------+
192 * 9	|			physical birth txg			|
193 *	+-------+-------+-------+-------+-------+-------+-------+-------+
194 * a	|			logical birth txg			|
195 *	+-------+-------+-------+-------+-------+-------+-------+-------+
196 * b	|			fill count				|
197 *	+-------+-------+-------+-------+-------+-------+-------+-------+
198 * c	|			checksum[0]				|
199 *	+-------+-------+-------+-------+-------+-------+-------+-------+
200 * d	|			checksum[1]				|
201 *	+-------+-------+-------+-------+-------+-------+-------+-------+
202 * e	|			checksum[2]				|
203 *	+-------+-------+-------+-------+-------+-------+-------+-------+
204 * f	|			checksum[3]				|
205 *	+-------+-------+-------+-------+-------+-------+-------+-------+
206 *
207 * Legend:
208 *
209 * vdev		virtual device ID
210 * offset	offset into virtual device
211 * LSIZE	logical size
212 * PSIZE	physical size (after compression)
213 * ASIZE	allocated size (including RAID-Z parity and gang block headers)
214 * GRID		RAID-Z layout information (reserved for future use)
215 * cksum	checksum function
216 * comp		compression function
217 * G		gang block indicator
218 * B		byteorder (endianness)
219 * D		dedup
220 * X		encryption
221 * E		blkptr_t contains embedded data (see below)
222 * lvl		level of indirection
223 * type		DMU object type
224 * phys birth	txg when dva[0] was written; zero if same as logical birth txg
225 *              note that typically all the dva's would be written in this
226 *              txg, but they could be different if they were moved by
227 *              device removal.
228 * log. birth	transaction group in which the block was logically born
229 * fill count	number of non-zero blocks under this bp
230 * checksum[4]	256-bit checksum of the data this bp describes
231 */
232
233/*
234 * The blkptr_t's of encrypted blocks also need to store the encryption
235 * parameters so that the block can be decrypted. This layout is as follows:
236 *
237 *	64	56	48	40	32	24	16	8	0
238 *	+-------+-------+-------+-------+-------+-------+-------+-------+
239 * 0	|		vdev1		| GRID  |	  ASIZE		|
240 *	+-------+-------+-------+-------+-------+-------+-------+-------+
241 * 1	|G|			 offset1				|
242 *	+-------+-------+-------+-------+-------+-------+-------+-------+
243 * 2	|		vdev2		| GRID  |	  ASIZE		|
244 *	+-------+-------+-------+-------+-------+-------+-------+-------+
245 * 3	|G|			 offset2				|
246 *	+-------+-------+-------+-------+-------+-------+-------+-------+
247 * 4	|			salt					|
248 *	+-------+-------+-------+-------+-------+-------+-------+-------+
249 * 5	|			IV1					|
250 *	+-------+-------+-------+-------+-------+-------+-------+-------+
251 * 6	|BDX|lvl| type	| cksum |E| comp|    PSIZE	|     LSIZE	|
252 *	+-------+-------+-------+-------+-------+-------+-------+-------+
253 * 7	|			padding					|
254 *	+-------+-------+-------+-------+-------+-------+-------+-------+
255 * 8	|			padding					|
256 *	+-------+-------+-------+-------+-------+-------+-------+-------+
257 * 9	|			physical birth txg			|
258 *	+-------+-------+-------+-------+-------+-------+-------+-------+
259 * a	|			logical birth txg			|
260 *	+-------+-------+-------+-------+-------+-------+-------+-------+
261 * b	|		IV2		|	    fill count		|
262 *	+-------+-------+-------+-------+-------+-------+-------+-------+
263 * c	|			checksum[0]				|
264 *	+-------+-------+-------+-------+-------+-------+-------+-------+
265 * d	|			checksum[1]				|
266 *	+-------+-------+-------+-------+-------+-------+-------+-------+
267 * e	|			MAC[0]					|
268 *	+-------+-------+-------+-------+-------+-------+-------+-------+
269 * f	|			MAC[1]					|
270 *	+-------+-------+-------+-------+-------+-------+-------+-------+
271 *
272 * Legend:
273 *
274 * salt		Salt for generating encryption keys
275 * IV1		First 64 bits of encryption IV
276 * X		Block requires encryption handling (set to 1)
277 * E		blkptr_t contains embedded data (set to 0, see below)
278 * fill count	number of non-zero blocks under this bp (truncated to 32 bits)
279 * IV2		Last 32 bits of encryption IV
280 * checksum[2]	128-bit checksum of the data this bp describes
281 * MAC[2]	128-bit message authentication code for this data
282 *
283 * The X bit being set indicates that this block is one of 3 types. If this is
284 * a level 0 block with an encrypted object type, the block is encrypted
285 * (see BP_IS_ENCRYPTED()). If this is a level 0 block with an unencrypted
286 * object type, this block is authenticated with an HMAC (see
287 * BP_IS_AUTHENTICATED()). Otherwise (if level > 0), this bp will use the MAC
288 * words to store a checksum-of-MACs from the level below (see
289 * BP_HAS_INDIRECT_MAC_CKSUM()). For convenience in the code, BP_IS_PROTECTED()
290 * refers to both encrypted and authenticated blocks and BP_USES_CRYPT()
291 * refers to any of these 3 kinds of blocks.
292 *
293 * The additional encryption parameters are the salt, IV, and MAC which are
294 * explained in greater detail in the block comment at the top of zio_crypt.c.
295 * The MAC occupies half of the checksum space since it serves a very similar
296 * purpose: to prevent data corruption on disk. The only functional difference
297 * is that the checksum is used to detect on-disk corruption whether or not the
298 * encryption key is loaded and the MAC provides additional protection against
299 * malicious disk tampering. We use the 3rd DVA to store the salt and first
300 * 64 bits of the IV. As a result encrypted blocks can only have 2 copies
301 * maximum instead of the normal 3. The last 32 bits of the IV are stored in
302 * the upper bits of what is usually the fill count. Note that only blocks at
303 * level 0 or -2 are ever encrypted, which allows us to guarantee that these
304 * 32 bits are not trampled over by other code (see zio_crypt.c for details).
305 * The salt and IV are not used for authenticated bps or bps with an indirect
306 * MAC checksum, so these blocks can utilize all 3 DVAs and the full 64 bits
307 * for the fill count.
308 */
309
310/*
311 * "Embedded" blkptr_t's don't actually point to a block, instead they
312 * have a data payload embedded in the blkptr_t itself.  See the comment
313 * in blkptr.c for more details.
314 *
315 * The blkptr_t is laid out as follows:
316 *
317 *	64	56	48	40	32	24	16	8	0
318 *	+-------+-------+-------+-------+-------+-------+-------+-------+
319 * 0	|      payload                                                  |
320 * 1	|      payload                                                  |
321 * 2	|      payload                                                  |
322 * 3	|      payload                                                  |
323 * 4	|      payload                                                  |
324 * 5	|      payload                                                  |
325 *	+-------+-------+-------+-------+-------+-------+-------+-------+
326 * 6	|BDX|lvl| type	| etype |E| comp| PSIZE|              LSIZE	|
327 *	+-------+-------+-------+-------+-------+-------+-------+-------+
328 * 7	|      payload                                                  |
329 * 8	|      payload                                                  |
330 * 9	|      payload                                                  |
331 *	+-------+-------+-------+-------+-------+-------+-------+-------+
332 * a	|			logical birth txg			|
333 *	+-------+-------+-------+-------+-------+-------+-------+-------+
334 * b	|      payload                                                  |
335 * c	|      payload                                                  |
336 * d	|      payload                                                  |
337 * e	|      payload                                                  |
338 * f	|      payload                                                  |
339 *	+-------+-------+-------+-------+-------+-------+-------+-------+
340 *
341 * Legend:
342 *
343 * payload		contains the embedded data
344 * B (byteorder)	byteorder (endianness)
345 * D (dedup)		padding (set to zero)
346 * X			encryption (set to zero; see above)
347 * E (embedded)		set to one
348 * lvl			indirection level
349 * type			DMU object type
350 * etype		how to interpret embedded data (BP_EMBEDDED_TYPE_*)
351 * comp			compression function of payload
352 * PSIZE		size of payload after compression, in bytes
353 * LSIZE		logical size of payload, in bytes
354 *			note that 25 bits is enough to store the largest
355 *			"normal" BP's LSIZE (2^16 * 2^9) in bytes
356 * log. birth		transaction group in which the block was logically born
357 *
358 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
359 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
360 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
361 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
362 * BP's so the BP_SET_* macros can be used with them.  etype, PSIZE, LSIZE must
363 * be set with the BPE_SET_* macros.  BP_SET_EMBEDDED() should be called before
364 * other macros, as they assert that they are only used on BP's of the correct
365 * "embedded-ness". Encrypted blkptr_t's cannot be embedded because they use
366 * the payload space for encryption parameters (see the comment above on
367 * how encryption parameters are stored).
368 */
369
370#define	BPE_GET_ETYPE(bp)	\
371	(ASSERT(BP_IS_EMBEDDED(bp)), \
372	BF64_GET((bp)->blk_prop, 40, 8))
373#define	BPE_SET_ETYPE(bp, t)	do { \
374	ASSERT(BP_IS_EMBEDDED(bp)); \
375	BF64_SET((bp)->blk_prop, 40, 8, t); \
376_NOTE(CONSTCOND) } while (0)
377
378#define	BPE_GET_LSIZE(bp)	\
379	(ASSERT(BP_IS_EMBEDDED(bp)), \
380	BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
381#define	BPE_SET_LSIZE(bp, x)	do { \
382	ASSERT(BP_IS_EMBEDDED(bp)); \
383	BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
384_NOTE(CONSTCOND) } while (0)
385
386#define	BPE_GET_PSIZE(bp)	\
387	(ASSERT(BP_IS_EMBEDDED(bp)), \
388	BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
389#define	BPE_SET_PSIZE(bp, x)	do { \
390	ASSERT(BP_IS_EMBEDDED(bp)); \
391	BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x);	\
392_NOTE(CONSTCOND) } while (0)
393
394typedef enum bp_embedded_type {
395	BP_EMBEDDED_TYPE_DATA,
396	BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */
397	NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED
398} bp_embedded_type_t;
399
400#define	BPE_NUM_WORDS 14
401#define	BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
402#define	BPE_IS_PAYLOADWORD(bp, wp) \
403	((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
404
405#define	SPA_BLKPTRSHIFT	7		/* blkptr_t is 128 bytes	*/
406#define	SPA_DVAS_PER_BP	3		/* Number of DVAs in a bp	*/
407#define	SPA_SYNC_MIN_VDEVS 3		/* min vdevs to update during sync */
408
409/*
410 * A block is a hole when it has either 1) never been written to, or
411 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
412 * without physically allocating disk space. Holes are represented in the
413 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is
414 * done through the BP_IS_HOLE macro. For holes, the logical size, level,
415 * DMU object type, and birth times are all also stored for holes that
416 * were written to at some point (i.e. were punched after having been filled).
417 */
418typedef struct blkptr {
419	dva_t		blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
420	uint64_t	blk_prop;	/* size, compression, type, etc	    */
421	uint64_t	blk_pad[2];	/* Extra space for the future	    */
422	uint64_t	blk_phys_birth;	/* txg when block was allocated	    */
423	uint64_t	blk_birth;	/* transaction group at birth	    */
424	uint64_t	blk_fill;	/* fill count			    */
425	zio_cksum_t	blk_cksum;	/* 256-bit checksum		    */
426} blkptr_t;
427
428/*
429 * Macros to get and set fields in a bp or DVA.
430 */
431#define	DVA_GET_ASIZE(dva)	\
432	BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
433#define	DVA_SET_ASIZE(dva, x)	\
434	BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
435	SPA_MINBLOCKSHIFT, 0, x)
436
437#define	DVA_GET_GRID(dva)	BF64_GET((dva)->dva_word[0], 24, 8)
438#define	DVA_SET_GRID(dva, x)	BF64_SET((dva)->dva_word[0], 24, 8, x)
439
440#define	DVA_GET_VDEV(dva)	BF64_GET((dva)->dva_word[0], 32, SPA_VDEVBITS)
441#define	DVA_SET_VDEV(dva, x)	\
442	BF64_SET((dva)->dva_word[0], 32, SPA_VDEVBITS, x)
443
444#define	DVA_GET_OFFSET(dva)	\
445	BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
446#define	DVA_SET_OFFSET(dva, x)	\
447	BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
448
449#define	DVA_GET_GANG(dva)	BF64_GET((dva)->dva_word[1], 63, 1)
450#define	DVA_SET_GANG(dva, x)	BF64_SET((dva)->dva_word[1], 63, 1, x)
451
452#define	BP_GET_LSIZE(bp)	\
453	(BP_IS_EMBEDDED(bp) ?	\
454	(BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
455	BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
456#define	BP_SET_LSIZE(bp, x)	do { \
457	ASSERT(!BP_IS_EMBEDDED(bp)); \
458	BF64_SET_SB((bp)->blk_prop, \
459	    0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
460_NOTE(CONSTCOND) } while (0)
461
462#define	BP_GET_PSIZE(bp)	\
463	(BP_IS_EMBEDDED(bp) ? 0 : \
464	BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1))
465#define	BP_SET_PSIZE(bp, x)	do { \
466	ASSERT(!BP_IS_EMBEDDED(bp)); \
467	BF64_SET_SB((bp)->blk_prop, \
468	    16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
469_NOTE(CONSTCOND) } while (0)
470
471#define	BP_GET_COMPRESS(bp)		\
472	BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS)
473#define	BP_SET_COMPRESS(bp, x)		\
474	BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x)
475
476#define	BP_IS_EMBEDDED(bp)		BF64_GET((bp)->blk_prop, 39, 1)
477#define	BP_SET_EMBEDDED(bp, x)		BF64_SET((bp)->blk_prop, 39, 1, x)
478
479#define	BP_GET_CHECKSUM(bp)		\
480	(BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \
481	BF64_GET((bp)->blk_prop, 40, 8))
482#define	BP_SET_CHECKSUM(bp, x)		do { \
483	ASSERT(!BP_IS_EMBEDDED(bp)); \
484	BF64_SET((bp)->blk_prop, 40, 8, x); \
485_NOTE(CONSTCOND) } while (0)
486
487#define	BP_GET_TYPE(bp)			BF64_GET((bp)->blk_prop, 48, 8)
488#define	BP_SET_TYPE(bp, x)		BF64_SET((bp)->blk_prop, 48, 8, x)
489
490#define	BP_GET_LEVEL(bp)		BF64_GET((bp)->blk_prop, 56, 5)
491#define	BP_SET_LEVEL(bp, x)		BF64_SET((bp)->blk_prop, 56, 5, x)
492
493/* encrypted, authenticated, and MAC cksum bps use the same bit */
494#define	BP_USES_CRYPT(bp)		BF64_GET((bp)->blk_prop, 61, 1)
495#define	BP_SET_CRYPT(bp, x)		BF64_SET((bp)->blk_prop, 61, 1, x)
496
497#define	BP_IS_ENCRYPTED(bp)			\
498	(BP_USES_CRYPT(bp) &&			\
499	BP_GET_LEVEL(bp) == 0 &&		\
500	DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp)))
501
502#define	BP_IS_AUTHENTICATED(bp)			\
503	(BP_USES_CRYPT(bp) &&			\
504	BP_GET_LEVEL(bp) == 0 &&		\
505	!DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp)))
506
507#define	BP_HAS_INDIRECT_MAC_CKSUM(bp)		\
508	(BP_USES_CRYPT(bp) && BP_GET_LEVEL(bp) > 0)
509
510#define	BP_IS_PROTECTED(bp)			\
511	(BP_IS_ENCRYPTED(bp) || BP_IS_AUTHENTICATED(bp))
512
513#define	BP_GET_DEDUP(bp)		BF64_GET((bp)->blk_prop, 62, 1)
514#define	BP_SET_DEDUP(bp, x)		BF64_SET((bp)->blk_prop, 62, 1, x)
515
516#define	BP_GET_BYTEORDER(bp)		BF64_GET((bp)->blk_prop, 63, 1)
517#define	BP_SET_BYTEORDER(bp, x)		BF64_SET((bp)->blk_prop, 63, 1, x)
518
519#define	BP_PHYSICAL_BIRTH(bp)		\
520	(BP_IS_EMBEDDED(bp) ? 0 : \
521	(bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
522
523#define	BP_SET_BIRTH(bp, logical, physical)	\
524{						\
525	ASSERT(!BP_IS_EMBEDDED(bp));		\
526	(bp)->blk_birth = (logical);		\
527	(bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
528}
529
530#define	BP_GET_FILL(bp)				\
531	((BP_IS_ENCRYPTED(bp)) ? BF64_GET((bp)->blk_fill, 0, 32) : \
532	((BP_IS_EMBEDDED(bp)) ? 1 : (bp)->blk_fill))
533
534#define	BP_SET_FILL(bp, fill)			\
535{						\
536	if (BP_IS_ENCRYPTED(bp))			\
537		BF64_SET((bp)->blk_fill, 0, 32, fill); \
538	else					\
539		(bp)->blk_fill = fill;		\
540}
541
542#define	BP_GET_IV2(bp)				\
543	(ASSERT(BP_IS_ENCRYPTED(bp)),		\
544	BF64_GET((bp)->blk_fill, 32, 32))
545#define	BP_SET_IV2(bp, iv2)			\
546{						\
547	ASSERT(BP_IS_ENCRYPTED(bp));		\
548	BF64_SET((bp)->blk_fill, 32, 32, iv2);	\
549}
550
551#define	BP_IS_METADATA(bp)	\
552	(BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))
553
554#define	BP_GET_ASIZE(bp)	\
555	(BP_IS_EMBEDDED(bp) ? 0 : \
556	DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
557	DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
558	(DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))
559
560#define	BP_GET_UCSIZE(bp)	\
561	(BP_IS_METADATA(bp) ? BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
562
563#define	BP_GET_NDVAS(bp)	\
564	(BP_IS_EMBEDDED(bp) ? 0 : \
565	!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
566	!!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
567	(!!DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))
568
569#define	BP_COUNT_GANG(bp)	\
570	(BP_IS_EMBEDDED(bp) ? 0 : \
571	(DVA_GET_GANG(&(bp)->blk_dva[0]) + \
572	DVA_GET_GANG(&(bp)->blk_dva[1]) + \
573	(DVA_GET_GANG(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp))))
574
575#define	DVA_EQUAL(dva1, dva2)	\
576	((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
577	(dva1)->dva_word[0] == (dva2)->dva_word[0])
578
579#define	BP_EQUAL(bp1, bp2)	\
580	(BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) &&	\
581	(bp1)->blk_birth == (bp2)->blk_birth &&			\
582	DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) &&	\
583	DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) &&	\
584	DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
585
586#define	ZIO_CHECKSUM_EQUAL(zc1, zc2) \
587	(0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
588	((zc1).zc_word[1] - (zc2).zc_word[1]) | \
589	((zc1).zc_word[2] - (zc2).zc_word[2]) | \
590	((zc1).zc_word[3] - (zc2).zc_word[3])))
591
592#define	ZIO_CHECKSUM_MAC_EQUAL(zc1, zc2) \
593	(0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
594	((zc1).zc_word[1] - (zc2).zc_word[1])))
595
596#define	ZIO_CHECKSUM_IS_ZERO(zc) \
597	(0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \
598	(zc)->zc_word[2] | (zc)->zc_word[3]))
599
600#define	ZIO_CHECKSUM_BSWAP(zcp)					\
601{								\
602	(zcp)->zc_word[0] = BSWAP_64((zcp)->zc_word[0]);	\
603	(zcp)->zc_word[1] = BSWAP_64((zcp)->zc_word[1]);	\
604	(zcp)->zc_word[2] = BSWAP_64((zcp)->zc_word[2]);	\
605	(zcp)->zc_word[3] = BSWAP_64((zcp)->zc_word[3]);	\
606}
607
608
609#define	DVA_IS_VALID(dva)	(DVA_GET_ASIZE(dva) != 0)
610
611#define	ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3)	\
612{						\
613	(zcp)->zc_word[0] = w0;			\
614	(zcp)->zc_word[1] = w1;			\
615	(zcp)->zc_word[2] = w2;			\
616	(zcp)->zc_word[3] = w3;			\
617}
618
619#define	BP_IDENTITY(bp)		(ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0])
620#define	BP_IS_GANG(bp)		\
621	(BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp)))
622#define	DVA_IS_EMPTY(dva)	((dva)->dva_word[0] == 0ULL &&	\
623				(dva)->dva_word[1] == 0ULL)
624#define	BP_IS_HOLE(bp) \
625	(!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp)))
626
627/* BP_IS_RAIDZ(bp) assumes no block compression */
628#define	BP_IS_RAIDZ(bp)		(DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
629				BP_GET_PSIZE(bp))
630
631#define	BP_ZERO(bp)				\
632{						\
633	(bp)->blk_dva[0].dva_word[0] = 0;	\
634	(bp)->blk_dva[0].dva_word[1] = 0;	\
635	(bp)->blk_dva[1].dva_word[0] = 0;	\
636	(bp)->blk_dva[1].dva_word[1] = 0;	\
637	(bp)->blk_dva[2].dva_word[0] = 0;	\
638	(bp)->blk_dva[2].dva_word[1] = 0;	\
639	(bp)->blk_prop = 0;			\
640	(bp)->blk_pad[0] = 0;			\
641	(bp)->blk_pad[1] = 0;			\
642	(bp)->blk_phys_birth = 0;		\
643	(bp)->blk_birth = 0;			\
644	(bp)->blk_fill = 0;			\
645	ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0);	\
646}
647
648#ifdef _BIG_ENDIAN
649#define	ZFS_HOST_BYTEORDER	(0ULL)
650#else
651#define	ZFS_HOST_BYTEORDER	(1ULL)
652#endif
653
654#define	BP_SHOULD_BYTESWAP(bp)	(BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
655
656#define	BP_SPRINTF_LEN	400
657
658/*
659 * This macro allows code sharing between zfs, libzpool, and mdb.
660 * 'func' is either snprintf() or mdb_snprintf().
661 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
662 */
663#define	SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
664{									\
665	static const char *copyname[] =					\
666	    { "zero", "single", "double", "triple" };			\
667	int len = 0;							\
668	int copies = 0;							\
669	const char *crypt_type;						\
670	if (bp != NULL) {						\
671		if (BP_IS_ENCRYPTED(bp)) {				\
672			crypt_type = "encrypted";			\
673		} else if (BP_IS_AUTHENTICATED(bp)) {			\
674			crypt_type = "authenticated";			\
675		} else if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) {		\
676			crypt_type = "indirect-MAC";			\
677		} else {						\
678			crypt_type = "unencrypted";			\
679		}							\
680	}								\
681	if (bp == NULL) {						\
682		len += func(buf + len, size - len, "<NULL>");		\
683	} else if (BP_IS_HOLE(bp)) {					\
684		len += func(buf + len, size - len,			\
685		    "HOLE [L%llu %s] "					\
686		    "size=%llxL birth=%lluL",				\
687		    (u_longlong_t)BP_GET_LEVEL(bp),			\
688		    type,						\
689		    (u_longlong_t)BP_GET_LSIZE(bp),			\
690		    (u_longlong_t)bp->blk_birth);			\
691	} else if (BP_IS_EMBEDDED(bp)) {				\
692		len = func(buf + len, size - len,			\
693		    "EMBEDDED [L%llu %s] et=%u %s "			\
694		    "size=%llxL/%llxP birth=%lluL",			\
695		    (u_longlong_t)BP_GET_LEVEL(bp),			\
696		    type,						\
697		    (int)BPE_GET_ETYPE(bp),				\
698		    compress,						\
699		    (u_longlong_t)BPE_GET_LSIZE(bp),			\
700		    (u_longlong_t)BPE_GET_PSIZE(bp),			\
701		    (u_longlong_t)bp->blk_birth);			\
702	} else {							\
703		for (int d = 0; d < BP_GET_NDVAS(bp); d++) {		\
704			const dva_t *dva = &bp->blk_dva[d];		\
705			if (DVA_IS_VALID(dva))				\
706				copies++;				\
707			len += func(buf + len, size - len,		\
708			    "DVA[%d]=<%llu:%llx:%llx>%c", d,		\
709			    (u_longlong_t)DVA_GET_VDEV(dva),		\
710			    (u_longlong_t)DVA_GET_OFFSET(dva),		\
711			    (u_longlong_t)DVA_GET_ASIZE(dva),		\
712			    ws);					\
713		}							\
714		if (BP_IS_ENCRYPTED(bp)) {				\
715			len += func(buf + len, size - len,		\
716			    "salt=%llx iv=%llx:%llx%c",			\
717			    (u_longlong_t)bp->blk_dva[2].dva_word[0],	\
718			    (u_longlong_t)bp->blk_dva[2].dva_word[1],	\
719			    (u_longlong_t)BP_GET_IV2(bp),		\
720			    ws);					\
721		}							\
722		if (BP_IS_GANG(bp) &&					\
723		    DVA_GET_ASIZE(&bp->blk_dva[2]) <=			\
724		    DVA_GET_ASIZE(&bp->blk_dva[1]) / 2)			\
725			copies--;					\
726		len += func(buf + len, size - len,			\
727		    "[L%llu %s] %s %s %s %s %s %s %s%c"			\
728		    "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c"	\
729		    "cksum=%llx:%llx:%llx:%llx",			\
730		    (u_longlong_t)BP_GET_LEVEL(bp),			\
731		    type,						\
732		    checksum,						\
733		    compress,						\
734		    crypt_type,						\
735		    BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE",		\
736		    BP_IS_GANG(bp) ? "gang" : "contiguous",		\
737		    BP_GET_DEDUP(bp) ? "dedup" : "unique",		\
738		    copyname[copies],					\
739		    ws,							\
740		    (u_longlong_t)BP_GET_LSIZE(bp),			\
741		    (u_longlong_t)BP_GET_PSIZE(bp),			\
742		    (u_longlong_t)bp->blk_birth,			\
743		    (u_longlong_t)BP_PHYSICAL_BIRTH(bp),		\
744		    (u_longlong_t)BP_GET_FILL(bp),			\
745		    ws,							\
746		    (u_longlong_t)bp->blk_cksum.zc_word[0],		\
747		    (u_longlong_t)bp->blk_cksum.zc_word[1],		\
748		    (u_longlong_t)bp->blk_cksum.zc_word[2],		\
749		    (u_longlong_t)bp->blk_cksum.zc_word[3]);		\
750	}								\
751	ASSERT(len < size);						\
752}
753
754#define	BP_GET_BUFC_TYPE(bp)						\
755	(BP_IS_METADATA(bp) ? ARC_BUFC_METADATA : ARC_BUFC_DATA)
756
757typedef enum spa_import_type {
758	SPA_IMPORT_EXISTING,
759	SPA_IMPORT_ASSEMBLE
760} spa_import_type_t;
761
762/* state manipulation functions */
763extern int spa_open(const char *pool, spa_t **, void *tag);
764extern int spa_open_rewind(const char *pool, spa_t **, void *tag,
765    nvlist_t *policy, nvlist_t **config);
766extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot,
767    size_t buflen);
768extern int spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props,
769    nvlist_t *zplprops, struct dsl_crypto_params *dcp);
770extern int spa_import_rootpool(char *devpath, char *devid);
771extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props,
772    uint64_t flags);
773extern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
774extern int spa_destroy(char *pool);
775extern int spa_checkpoint(const char *pool);
776extern int spa_checkpoint_discard(const char *pool);
777extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force,
778    boolean_t hardforce);
779extern int spa_reset(char *pool);
780extern void spa_async_request(spa_t *spa, int flag);
781extern void spa_async_unrequest(spa_t *spa, int flag);
782extern void spa_async_suspend(spa_t *spa);
783extern void spa_async_resume(spa_t *spa);
784extern spa_t *spa_inject_addref(char *pool);
785extern void spa_inject_delref(spa_t *spa);
786extern void spa_scan_stat_init(spa_t *spa);
787extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps);
788
789#define	SPA_ASYNC_CONFIG_UPDATE	0x01
790#define	SPA_ASYNC_REMOVE	0x02
791#define	SPA_ASYNC_PROBE		0x04
792#define	SPA_ASYNC_RESILVER_DONE	0x08
793#define	SPA_ASYNC_RESILVER	0x10
794#define	SPA_ASYNC_AUTOEXPAND	0x20
795#define	SPA_ASYNC_REMOVE_DONE	0x40
796#define	SPA_ASYNC_REMOVE_STOP	0x80
797#define	SPA_ASYNC_INITIALIZE_RESTART	0x100
798
799/*
800 * Controls the behavior of spa_vdev_remove().
801 */
802#define	SPA_REMOVE_UNSPARE	0x01
803#define	SPA_REMOVE_DONE		0x02
804
805/* device manipulation */
806extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
807extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot,
808    int replacing);
809extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
810    int replace_done);
811extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare);
812extern boolean_t spa_vdev_remove_active(spa_t *spa);
813extern int spa_vdev_initialize(spa_t *spa, uint64_t guid, uint64_t cmd_type);
814extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
815extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru);
816extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
817    nvlist_t *props, boolean_t exp);
818
819/* spare state (which is global across all pools) */
820extern void spa_spare_add(vdev_t *vd);
821extern void spa_spare_remove(vdev_t *vd);
822extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt);
823extern void spa_spare_activate(vdev_t *vd);
824
825/* L2ARC state (which is global across all pools) */
826extern void spa_l2cache_add(vdev_t *vd);
827extern void spa_l2cache_remove(vdev_t *vd);
828extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
829extern void spa_l2cache_activate(vdev_t *vd);
830extern void spa_l2cache_drop(spa_t *spa);
831
832/* scanning */
833extern int spa_scan(spa_t *spa, pool_scan_func_t func);
834extern int spa_scan_stop(spa_t *spa);
835extern int spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t flag);
836
837/* spa syncing */
838extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
839extern void spa_sync_allpools(void);
840
841/* spa namespace global mutex */
842extern kmutex_t spa_namespace_lock;
843
844/*
845 * SPA configuration functions in spa_config.c
846 */
847
848#define	SPA_CONFIG_UPDATE_POOL	0
849#define	SPA_CONFIG_UPDATE_VDEVS	1
850
851extern void spa_write_cachefile(spa_t *, boolean_t, boolean_t);
852extern void spa_config_load(void);
853extern nvlist_t *spa_all_configs(uint64_t *);
854extern void spa_config_set(spa_t *spa, nvlist_t *config);
855extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
856    int getstats);
857extern void spa_config_update(spa_t *spa, int what);
858
859/*
860 * Miscellaneous SPA routines in spa_misc.c
861 */
862
863/* Namespace manipulation */
864extern spa_t *spa_lookup(const char *name);
865extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot);
866extern void spa_remove(spa_t *spa);
867extern spa_t *spa_next(spa_t *prev);
868
869/* Refcount functions */
870extern void spa_open_ref(spa_t *spa, void *tag);
871extern void spa_close(spa_t *spa, void *tag);
872extern void spa_async_close(spa_t *spa, void *tag);
873extern boolean_t spa_refcount_zero(spa_t *spa);
874
875#define	SCL_NONE	0x00
876#define	SCL_CONFIG	0x01
877#define	SCL_STATE	0x02
878#define	SCL_L2ARC	0x04		/* hack until L2ARC 2.0 */
879#define	SCL_ALLOC	0x08
880#define	SCL_ZIO		0x10
881#define	SCL_FREE	0x20
882#define	SCL_VDEV	0x40
883#define	SCL_LOCKS	7
884#define	SCL_ALL		((1 << SCL_LOCKS) - 1)
885#define	SCL_STATE_ALL	(SCL_STATE | SCL_L2ARC | SCL_ZIO)
886
887/* Pool configuration locks */
888extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw);
889extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw);
890extern void spa_config_exit(spa_t *spa, int locks, void *tag);
891extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
892
893/* Pool vdev add/remove lock */
894extern uint64_t spa_vdev_enter(spa_t *spa);
895extern uint64_t spa_vdev_config_enter(spa_t *spa);
896extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg,
897    int error, char *tag);
898extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error);
899
900/* Pool vdev state change lock */
901extern void spa_vdev_state_enter(spa_t *spa, int oplock);
902extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error);
903
904/* Log state */
905typedef enum spa_log_state {
906	SPA_LOG_UNKNOWN = 0,	/* unknown log state */
907	SPA_LOG_MISSING,	/* missing log(s) */
908	SPA_LOG_CLEAR,		/* clear the log(s) */
909	SPA_LOG_GOOD,		/* log(s) are good */
910} spa_log_state_t;
911
912extern spa_log_state_t spa_get_log_state(spa_t *spa);
913extern void spa_set_log_state(spa_t *spa, spa_log_state_t state);
914extern int spa_reset_logs(spa_t *spa);
915
916/* Log claim callback */
917extern void spa_claim_notify(zio_t *zio);
918
919/* Accessor functions */
920extern boolean_t spa_shutting_down(spa_t *spa);
921extern struct dsl_pool *spa_get_dsl(spa_t *spa);
922extern boolean_t spa_is_initializing(spa_t *spa);
923extern boolean_t spa_indirect_vdevs_loaded(spa_t *spa);
924extern blkptr_t *spa_get_rootblkptr(spa_t *spa);
925extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
926extern void spa_altroot(spa_t *, char *, size_t);
927extern int spa_sync_pass(spa_t *spa);
928extern char *spa_name(spa_t *spa);
929extern uint64_t spa_guid(spa_t *spa);
930extern uint64_t spa_load_guid(spa_t *spa);
931extern uint64_t spa_last_synced_txg(spa_t *spa);
932extern uint64_t spa_first_txg(spa_t *spa);
933extern uint64_t spa_syncing_txg(spa_t *spa);
934extern uint64_t spa_final_dirty_txg(spa_t *spa);
935extern uint64_t spa_version(spa_t *spa);
936extern pool_state_t spa_state(spa_t *spa);
937extern spa_load_state_t spa_load_state(spa_t *spa);
938extern uint64_t spa_freeze_txg(spa_t *spa);
939extern uint64_t spa_get_worst_case_asize(spa_t *spa, uint64_t lsize);
940extern uint64_t spa_get_dspace(spa_t *spa);
941extern uint64_t spa_get_checkpoint_space(spa_t *spa);
942extern uint64_t spa_get_slop_space(spa_t *spa);
943extern void spa_update_dspace(spa_t *spa);
944extern uint64_t spa_version(spa_t *spa);
945extern boolean_t spa_deflate(spa_t *spa);
946extern metaslab_class_t *spa_normal_class(spa_t *spa);
947extern metaslab_class_t *spa_log_class(spa_t *spa);
948extern metaslab_class_t *spa_special_class(spa_t *spa);
949extern metaslab_class_t *spa_dedup_class(spa_t *spa);
950extern metaslab_class_t *spa_preferred_class(spa_t *spa, uint64_t size,
951    dmu_object_type_t objtype, uint_t level, uint_t special_smallblk);
952
953extern void spa_evicting_os_register(spa_t *, objset_t *os);
954extern void spa_evicting_os_deregister(spa_t *, objset_t *os);
955extern void spa_evicting_os_wait(spa_t *spa);
956extern int spa_max_replication(spa_t *spa);
957extern int spa_prev_software_version(spa_t *spa);
958extern int spa_busy(void);
959extern uint8_t spa_get_failmode(spa_t *spa);
960extern boolean_t spa_suspended(spa_t *spa);
961extern uint64_t spa_bootfs(spa_t *spa);
962extern uint64_t spa_delegation(spa_t *spa);
963extern objset_t *spa_meta_objset(spa_t *spa);
964extern uint64_t spa_deadman_synctime(spa_t *spa);
965extern uint64_t spa_dirty_data(spa_t *spa);
966
967/* Miscellaneous support routines */
968extern void spa_load_failed(spa_t *spa, const char *fmt, ...);
969extern void spa_load_note(spa_t *spa, const char *fmt, ...);
970extern void spa_activate_mos_feature(spa_t *spa, const char *feature,
971    dmu_tx_t *tx);
972extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature);
973extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid);
974extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid);
975extern char *spa_strdup(const char *);
976extern void spa_strfree(char *);
977extern uint64_t spa_get_random(uint64_t range);
978extern uint64_t spa_generate_guid(spa_t *spa);
979extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp);
980extern void spa_freeze(spa_t *spa);
981extern int spa_change_guid(spa_t *spa);
982extern void spa_upgrade(spa_t *spa, uint64_t version);
983extern void spa_evict_all(void);
984extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid,
985    boolean_t l2cache);
986extern boolean_t spa_has_spare(spa_t *, uint64_t guid);
987extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva);
988extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp);
989extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp);
990extern boolean_t spa_has_slogs(spa_t *spa);
991extern boolean_t spa_is_root(spa_t *spa);
992extern boolean_t spa_writeable(spa_t *spa);
993extern boolean_t spa_has_pending_synctask(spa_t *spa);
994extern int spa_maxblocksize(spa_t *spa);
995extern int spa_maxdnodesize(spa_t *spa);
996extern boolean_t spa_multihost(spa_t *spa);
997extern unsigned long spa_get_hostid(void);
998extern boolean_t spa_has_checkpoint(spa_t *spa);
999extern boolean_t spa_importing_readonly_checkpoint(spa_t *spa);
1000extern boolean_t spa_suspend_async_destroy(spa_t *spa);
1001extern uint64_t spa_min_claim_txg(spa_t *spa);
1002extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp);
1003extern boolean_t zfs_dva_valid(spa_t *spa, const dva_t *dva,
1004    const blkptr_t *bp);
1005typedef void (*spa_remap_cb_t)(uint64_t vdev, uint64_t offset, uint64_t size,
1006    void *arg);
1007extern boolean_t spa_remap_blkptr(spa_t *spa, blkptr_t *bp,
1008    spa_remap_cb_t callback, void *arg);
1009extern uint64_t spa_get_last_removal_txg(spa_t *spa);
1010extern boolean_t spa_trust_config(spa_t *spa);
1011extern uint64_t spa_missing_tvds_allowed(spa_t *spa);
1012extern void spa_set_missing_tvds(spa_t *spa, uint64_t missing);
1013extern boolean_t spa_top_vdevs_spacemap_addressable(spa_t *spa);
1014extern void spa_activate_allocation_classes(spa_t *, dmu_tx_t *);
1015
1016extern int spa_mode(spa_t *spa);
1017extern uint64_t zfs_strtonum(const char *str, char **nptr);
1018
1019extern char *spa_his_ievent_table[];
1020
1021extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx);
1022extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read,
1023    char *his_buf);
1024extern int spa_history_log(spa_t *spa, const char *his_buf);
1025extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl);
1026extern void spa_history_log_version(spa_t *spa, const char *operation);
1027extern void spa_history_log_internal(spa_t *spa, const char *operation,
1028    dmu_tx_t *tx, const char *fmt, ...);
1029extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op,
1030    dmu_tx_t *tx, const char *fmt, ...);
1031extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
1032    dmu_tx_t *tx, const char *fmt, ...);
1033
1034/* error handling */
1035struct zbookmark_phys;
1036extern void spa_log_error(spa_t *spa, const struct zbookmark_phys *zb);
1037extern void zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd,
1038    const struct zbookmark_phys *zb, struct zio *zio, uint64_t stateoroffset,
1039    uint64_t length);
1040extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
1041extern void zfs_post_state_change(spa_t *spa, vdev_t *vd);
1042extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
1043extern uint64_t spa_get_errlog_size(spa_t *spa);
1044extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
1045extern void spa_errlog_rotate(spa_t *spa);
1046extern void spa_errlog_drain(spa_t *spa);
1047extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
1048extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
1049
1050/* vdev cache */
1051extern void vdev_cache_stat_init(void);
1052extern void vdev_cache_stat_fini(void);
1053
1054/* vdev mirror */
1055extern void vdev_mirror_stat_init(void);
1056extern void vdev_mirror_stat_fini(void);
1057
1058/* Initialization and termination */
1059extern void spa_init(int flags);
1060extern void spa_fini(void);
1061extern void spa_boot_init(void);
1062
1063/* properties */
1064extern int spa_prop_set(spa_t *spa, nvlist_t *nvp);
1065extern int spa_prop_get(spa_t *spa, nvlist_t **nvp);
1066extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx);
1067extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t);
1068
1069/* asynchronous event notification */
1070extern void spa_event_notify(spa_t *spa, vdev_t *vdev, nvlist_t *hist_nvl,
1071    const char *name);
1072extern sysevent_t *spa_event_create(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl,
1073    const char *name);
1074extern void spa_event_post(sysevent_t *ev);
1075extern void spa_event_discard(sysevent_t *ev);
1076
1077#ifdef ZFS_DEBUG
1078#define	dprintf_bp(bp, fmt, ...) do {				\
1079	if (zfs_flags & ZFS_DEBUG_DPRINTF) {			\
1080	char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP);	\
1081	snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp));	\
1082	dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf);		\
1083	kmem_free(__blkbuf, BP_SPRINTF_LEN);			\
1084	} \
1085_NOTE(CONSTCOND) } while (0)
1086#else
1087#define	dprintf_bp(bp, fmt, ...)
1088#endif
1089
1090extern int spa_mode_global;			/* mode, e.g. FREAD | FWRITE */
1091
1092#ifdef	__cplusplus
1093}
1094#endif
1095
1096#endif	/* _SYS_SPA_H */
1097