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 2013 Saso Kiselkov. All rights reserved.
24 * Copyright (c) 2011, 2016 by Delphix. All rights reserved.
25 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
26 * Copyright 2016 Nexenta Systems, Inc. All rights reserved.
27 * Copyright 2013 Saso Kiselkov. All rights reserved.
28 * Copyright (c) 2014 Integros [integros.com]
29 * Copyright 2017 Joyent, Inc.
30 * Copyright (c) 2017 Datto Inc.
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/sysmacros.h>
40 #include <sys/types.h>
41 #include <sys/fs/zfs.h>
42 #include <sys/dmu.h>
43
44 #ifdef __cplusplus
45 extern "C" {
46 #endif
47
48 /*
49 * Forward references that lots of things need.
50 */
51 typedef struct spa spa_t;
52 typedef struct vdev vdev_t;
53 typedef struct metaslab metaslab_t;
54 typedef struct metaslab_group metaslab_group_t;
55 typedef struct metaslab_class metaslab_class_t;
56 typedef struct zio zio_t;
57 typedef struct zilog zilog_t;
58 typedef struct spa_aux_vdev spa_aux_vdev_t;
59 typedef struct ddt ddt_t;
60 typedef struct ddt_entry ddt_entry_t;
61 struct dsl_pool;
62 struct dsl_dataset;
63
64 struct zfs_autosnap;
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
144 /*
145 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
146 * The members of the dva_t should be considered opaque outside the SPA.
147 */
148 typedef struct dva {
149 uint64_t dva_word[2];
150 } dva_t;
151
152 /*
153 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
154 */
155 typedef struct zio_cksum {
156 uint64_t zc_word[4];
157 } zio_cksum_t;
158
159 /*
160 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept
161 * secret and is suitable for use in MAC algorithms as the key.
162 */
163 typedef struct zio_cksum_salt {
164 uint8_t zcs_bytes[32];
165 } zio_cksum_salt_t;
166
167 /*
168 * Each block is described by its DVAs, time of birth, checksum, etc.
169 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
170 *
171 * 64 56 48 40 32 24 16 8 0
172 * +-------+-------+-------+-------+-------+-------+-------+-------+
173 * 0 | vdev1 | GRID | ASIZE |
174 * +-------+-------+-------+-------+-------+-------+-------+-------+
175 * 1 |G| offset1 |
176 * +-------+-------+-------+-------+-------+-------+-------+-------+
177 * 2 | vdev2 | GRID | ASIZE |
178 * +-------+-------+-------+-------+-------+-------+-------+-------+
179 * 3 |G| offset2 |
180 * +-------+-------+-------+-------+-------+-------+-------+-------+
181 * 4 | vdev3 | GRID | ASIZE |
182 * +-------+-------+-------+-------+-------+-------+-------+-------+
183 * 5 |G| offset3 |
184 * +-------+-------+-------+-------+-------+-------+-------+-------+
185 * 6 |BDS|lvl| type | cksum |E| comp| PSIZE | LSIZE |
186 * +-------+-------+-------+-------+-------+-------+-------+-------+
187 * 7 | padding |
188 * +-------+-------+-------+-------+-------+-------+-------+-------+
189 * 8 | padding |
190 * +-------+-------+-------+-------+-------+-------+-------+-------+
191 * 9 | physical birth txg |
192 * +-------+-------+-------+-------+-------+-------+-------+-------+
193 * a | logical birth txg |
194 * +-------+-------+-------+-------+-------+-------+-------+-------+
195 * b | fill count |
196 * +-------+-------+-------+-------+-------+-------+-------+-------+
197 * c | checksum[0] |
198 * +-------+-------+-------+-------+-------+-------+-------+-------+
199 * d | checksum[1] |
200 * +-------+-------+-------+-------+-------+-------+-------+-------+
201 * e | checksum[2] |
202 * +-------+-------+-------+-------+-------+-------+-------+-------+
203 * f | checksum[3] |
204 * +-------+-------+-------+-------+-------+-------+-------+-------+
205 *
206 * Legend:
207 *
208 * vdev virtual device ID
209 * offset offset into virtual device
210 * LSIZE logical size
211 * PSIZE physical size (after compression)
212 * ASIZE allocated size (including RAID-Z parity and gang block headers)
213 * GRID RAID-Z layout information (reserved for future use)
214 * cksum checksum function
215 * comp compression function
216 * G gang block indicator
217 * B byteorder (endianness)
218 * D dedup
219 * S special WBC block (unused for embedded blocks)
220 * E blkptr_t contains embedded data (see below)
221 * lvl level of indirection
222 * type DMU object type
223 * phys birth txg of block allocation; zero if same as logical birth txg
224 * log. birth transaction group in which the block was logically born
225 * fill count number of non-zero blocks under this bp
226 * checksum[4] 256-bit checksum of the data this bp describes
227 */
228
229 /*
230 * "Embedded" blkptr_t's don't actually point to a block, instead they
231 * have a data payload embedded in the blkptr_t itself. See the comment
232 * in blkptr.c for more details.
233 *
234 * The blkptr_t is laid out as follows:
235 *
236 * 64 56 48 40 32 24 16 8 0
237 * +-------+-------+-------+-------+-------+-------+-------+-------+
238 * 0 | payload |
239 * 1 | payload |
240 * 2 | payload |
241 * 3 | payload |
242 * 4 | payload |
243 * 5 | payload |
244 * +-------+-------+-------+-------+-------+-------+-------+-------+
245 * 6 |BDS|lvl| type | etype |E| comp| PSIZE| LSIZE |
246 * +-------+-------+-------+-------+-------+-------+-------+-------+
247 * 7 | payload |
248 * 8 | payload |
249 * 9 | payload |
250 * +-------+-------+-------+-------+-------+-------+-------+-------+
251 * a | logical birth txg |
252 * +-------+-------+-------+-------+-------+-------+-------+-------+
253 * b | payload |
254 * c | payload |
255 * d | payload |
256 * e | payload |
257 * f | payload |
258 * +-------+-------+-------+-------+-------+-------+-------+-------+
259 *
260 * Legend:
261 *
262 * payload contains the embedded data
263 * B (byteorder) byteorder (endianness)
264 * D (dedup) padding (set to zero)
265 * S special WBC block (unused for embedded blocks)
266 * E (embedded) set to one
267 * lvl indirection level
268 * type DMU object type
269 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*)
270 * comp compression function of payload
271 * PSIZE size of payload after compression, in bytes
272 * LSIZE logical size of payload, in bytes
273 * note that 25 bits is enough to store the largest
274 * "normal" BP's LSIZE (2^16 * 2^9) in bytes
275 * log. birth transaction group in which the block was logically born
276 *
277 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
278 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
279 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
280 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
281 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must
282 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before
283 * other macros, as they assert that they are only used on BP's of the correct
284 * "embedded-ness".
285 */
286
287 #define BPE_GET_ETYPE(bp) \
288 (ASSERT(BP_IS_EMBEDDED(bp)), \
289 BF64_GET((bp)->blk_prop, 40, 8))
290 #define BPE_SET_ETYPE(bp, t) do { \
291 ASSERT(BP_IS_EMBEDDED(bp)); \
292 BF64_SET((bp)->blk_prop, 40, 8, t); \
293 _NOTE(CONSTCOND) } while (0)
294
295 #define BPE_GET_LSIZE(bp) \
296 (ASSERT(BP_IS_EMBEDDED(bp)), \
297 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
298 #define BPE_SET_LSIZE(bp, x) do { \
299 ASSERT(BP_IS_EMBEDDED(bp)); \
300 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
301 _NOTE(CONSTCOND) } while (0)
302
303 #define BPE_GET_PSIZE(bp) \
304 (ASSERT(BP_IS_EMBEDDED(bp)), \
305 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
306 #define BPE_SET_PSIZE(bp, x) do { \
307 ASSERT(BP_IS_EMBEDDED(bp)); \
308 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
309 _NOTE(CONSTCOND) } while (0)
310
311 typedef enum bp_embedded_type {
312 BP_EMBEDDED_TYPE_DATA,
313 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */
314 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED
315 } bp_embedded_type_t;
316
317 #define BPE_NUM_WORDS 14
318 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
319 #define BPE_IS_PAYLOADWORD(bp, wp) \
320 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
321
322 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */
323 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */
324
325 /*
326 * A block is a hole when it has either 1) never been written to, or
327 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
328 * without physically allocating disk space. Holes are represented in the
329 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is
330 * done through the BP_IS_HOLE macro. For holes, the logical size, level,
331 * DMU object type, and birth times are all also stored for holes that
332 * were written to at some point (i.e. were punched after having been filled).
333 */
334 typedef struct blkptr {
335 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
336 uint64_t blk_prop; /* size, compression, type, etc */
337 uint64_t blk_pad[2]; /* Extra space for the future */
338 uint64_t blk_phys_birth; /* txg when block was allocated */
339 uint64_t blk_birth; /* transaction group at birth */
340 uint64_t blk_fill; /* fill count */
341 zio_cksum_t blk_cksum; /* 256-bit checksum */
342 } blkptr_t;
343
344 /*
345 * Macros to get and set fields in a bp or DVA.
346 */
347 #define DVA_GET_ASIZE(dva) \
348 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
349 #define DVA_SET_ASIZE(dva, x) \
350 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
351 SPA_MINBLOCKSHIFT, 0, x)
352
353 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8)
354 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x)
355
356 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32)
357 #define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x)
358
359 #define DVA_GET_OFFSET(dva) \
360 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
361 #define DVA_SET_OFFSET(dva, x) \
362 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
363
364 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1)
365 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x)
366
367 #define BP_GET_LSIZE(bp) \
368 (BP_IS_EMBEDDED(bp) ? \
369 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
370 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
371 #define BP_SET_LSIZE(bp, x) do { \
372 ASSERT(!BP_IS_EMBEDDED(bp)); \
373 BF64_SET_SB((bp)->blk_prop, \
374 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
375 _NOTE(CONSTCOND) } while (0)
376
377 #define BP_GET_PSIZE(bp) \
378 (BP_IS_EMBEDDED(bp) ? 0 : \
379 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1))
380 #define BP_SET_PSIZE(bp, x) do { \
381 ASSERT(!BP_IS_EMBEDDED(bp)); \
382 BF64_SET_SB((bp)->blk_prop, \
383 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
384 _NOTE(CONSTCOND) } while (0)
385
386 #define BP_GET_COMPRESS(bp) \
387 BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS)
388 #define BP_SET_COMPRESS(bp, x) \
389 BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x)
390
391 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1)
392 #define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x)
393
394 #define BP_GET_CHECKSUM(bp) \
395 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \
396 BF64_GET((bp)->blk_prop, 40, 8))
397 #define BP_SET_CHECKSUM(bp, x) do { \
398 ASSERT(!BP_IS_EMBEDDED(bp)); \
399 BF64_SET((bp)->blk_prop, 40, 8, x); \
400 _NOTE(CONSTCOND) } while (0)
401
402 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8)
403 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x)
404
405 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5)
406 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x)
407
408 #define BP_IS_SPECIAL(bp) BF64_GET((bp)->blk_prop, 61, 1)
409 #define BP_SET_SPECIAL(bp, x) BF64_SET((bp)->blk_prop, 61, 1, x)
410
411 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1)
412 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x)
413
414 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1)
415 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x)
416
417 #define BP_PHYSICAL_BIRTH(bp) \
418 (BP_IS_EMBEDDED(bp) ? 0 : \
419 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
420
421 #define BP_SET_BIRTH(bp, logical, physical) \
422 { \
423 ASSERT(!BP_IS_EMBEDDED(bp)); \
424 (bp)->blk_birth = (logical); \
425 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
426 }
427
428 #define BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill)
429
430 #define BP_IS_METADATA(bp) \
431 (BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))
432
433 #define BP_GET_ASIZE(bp) \
434 (BP_IS_EMBEDDED(bp) ? 0 : \
435 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
436 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
437 DVA_GET_ASIZE(&(bp)->blk_dva[2]))
438
439 #define BP_GET_UCSIZE(bp) \
440 (BP_IS_METADATA(bp) ? BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
441
442 #define BP_GET_NDVAS(bp) \
443 (BP_IS_EMBEDDED(bp) ? 0 : \
444 !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
445 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
446 !!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
447
448 #define BP_COUNT_GANG(bp) \
449 (BP_IS_EMBEDDED(bp) ? 0 : \
450 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \
451 DVA_GET_GANG(&(bp)->blk_dva[1]) + \
452 DVA_GET_GANG(&(bp)->blk_dva[2])))
453
454 #define DVA_EQUAL(dva1, dva2) \
455 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
456 (dva1)->dva_word[0] == (dva2)->dva_word[0])
457
458 #define BP_EQUAL(bp1, bp2) \
459 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \
460 (bp1)->blk_birth == (bp2)->blk_birth && \
461 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \
462 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \
463 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
464
465 #define ZIO_CHECKSUM_EQUAL(zc1, zc2) \
466 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
467 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \
468 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \
469 ((zc1).zc_word[3] - (zc2).zc_word[3])))
470
471 #define ZIO_CHECKSUM_BSWAP(_zc) \
472 do { \
473 zio_cksum_t *zc = (_zc); \
474 zc->zc_word[0] = BSWAP_64(zc->zc_word[0]); \
475 zc->zc_word[1] = BSWAP_64(zc->zc_word[1]); \
476 zc->zc_word[2] = BSWAP_64(zc->zc_word[2]); \
477 zc->zc_word[3] = BSWAP_64(zc->zc_word[3]); \
478 _NOTE(NOTREACHED) \
479 _NOTE(CONSTCOND) \
480 } while (0)
481
482 #define ZIO_CHECKSUM_IS_ZERO(zc) \
483 (0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \
484 (zc)->zc_word[2] | (zc)->zc_word[3]))
485
486 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0)
487
488 #define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \
489 { \
490 (zcp)->zc_word[0] = w0; \
491 (zcp)->zc_word[1] = w1; \
492 (zcp)->zc_word[2] = w2; \
493 (zcp)->zc_word[3] = w3; \
494 }
495
496 #define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0])
497 #define BP_IS_GANG(bp) \
498 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp)))
499 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \
500 (dva)->dva_word[1] == 0ULL)
501 #define BP_IS_HOLE(bp) \
502 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp)))
503
504 /* BP_IS_RAIDZ(bp) assumes no block compression */
505 #define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
506 BP_GET_PSIZE(bp))
507
508 #define BP_ZERO(bp) \
509 { \
510 (bp)->blk_dva[0].dva_word[0] = 0; \
511 (bp)->blk_dva[0].dva_word[1] = 0; \
512 (bp)->blk_dva[1].dva_word[0] = 0; \
513 (bp)->blk_dva[1].dva_word[1] = 0; \
514 (bp)->blk_dva[2].dva_word[0] = 0; \
515 (bp)->blk_dva[2].dva_word[1] = 0; \
516 (bp)->blk_prop = 0; \
517 (bp)->blk_pad[0] = 0; \
518 (bp)->blk_pad[1] = 0; \
519 (bp)->blk_phys_birth = 0; \
520 (bp)->blk_birth = 0; \
521 (bp)->blk_fill = 0; \
522 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \
523 }
524
525 #ifdef _BIG_ENDIAN
526 #define ZFS_HOST_BYTEORDER (0ULL)
527 #else
528 #define ZFS_HOST_BYTEORDER (1ULL)
529 #endif
530
531 #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
532
533 #define BP_SPRINTF_LEN 320
534
535 /*
536 * This macro allows code sharing between zfs, libzpool, and mdb.
537 * 'func' is either snprintf() or mdb_snprintf().
538 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
539 */
540 #define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
541 { \
542 static const char *copyname[] = \
543 { "zero", "single", "double", "triple" }; \
544 int len = 0; \
545 int copies = 0; \
546 \
547 if (bp == NULL) { \
548 len += func(buf + len, size - len, "<NULL>"); \
549 } else if (BP_IS_HOLE(bp)) { \
550 len += func(buf + len, size - len, \
551 "HOLE [L%llu %s] " \
552 "size=%llxL birth=%lluL", \
553 (u_longlong_t)BP_GET_LEVEL(bp), \
554 type, \
555 (u_longlong_t)BP_GET_LSIZE(bp), \
556 (u_longlong_t)bp->blk_birth); \
557 } else if (BP_IS_EMBEDDED(bp)) { \
558 len = func(buf + len, size - len, \
559 "EMBEDDED [L%llu %s] et=%u %s " \
560 "size=%llxL/%llxP birth=%lluL", \
561 (u_longlong_t)BP_GET_LEVEL(bp), \
562 type, \
563 (int)BPE_GET_ETYPE(bp), \
564 compress, \
565 (u_longlong_t)BPE_GET_LSIZE(bp), \
566 (u_longlong_t)BPE_GET_PSIZE(bp), \
567 (u_longlong_t)bp->blk_birth); \
568 } else { \
569 for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \
570 const dva_t *dva = &bp->blk_dva[d]; \
571 if (DVA_IS_VALID(dva)) \
572 copies++; \
573 len += func(buf + len, size - len, \
574 "DVA[%d]=<%llu:%llx:%llx>%c", d, \
575 (u_longlong_t)DVA_GET_VDEV(dva), \
576 (u_longlong_t)DVA_GET_OFFSET(dva), \
577 (u_longlong_t)DVA_GET_ASIZE(dva), \
578 ws); \
579 } \
580 if (BP_IS_GANG(bp) && \
581 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \
582 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \
583 copies--; \
584 len += func(buf + len, size - len, \
585 "[L%llu %s] %s %s %s %s %s %s%c" \
586 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \
587 "cksum=%llx:%llx:%llx:%llx", \
588 (u_longlong_t)BP_GET_LEVEL(bp), \
589 type, \
590 checksum, \
591 compress, \
592 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \
593 BP_IS_GANG(bp) ? "gang" : "contiguous", \
594 BP_GET_DEDUP(bp) ? "dedup" : "unique", \
595 copyname[copies], \
596 ws, \
597 (u_longlong_t)BP_GET_LSIZE(bp), \
598 (u_longlong_t)BP_GET_PSIZE(bp), \
599 (u_longlong_t)bp->blk_birth, \
600 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \
601 (u_longlong_t)BP_GET_FILL(bp), \
602 ws, \
603 (u_longlong_t)bp->blk_cksum.zc_word[0], \
604 (u_longlong_t)bp->blk_cksum.zc_word[1], \
605 (u_longlong_t)bp->blk_cksum.zc_word[2], \
606 (u_longlong_t)bp->blk_cksum.zc_word[3]); \
607 } \
608 ASSERT(len < size); \
609 }
610
611 #define BP_GET_BUFC_TYPE(bp) \
612 ((BP_GET_TYPE(bp) == DMU_OT_DDT_ZAP || \
613 BP_GET_TYPE(bp) == DMU_OT_DDT_STATS) ? ARC_BUFC_DDT : \
614 (BP_IS_METADATA(bp) ? ARC_BUFC_METADATA : ARC_BUFC_DATA))
615
616 typedef enum spa_import_type {
617 SPA_IMPORT_EXISTING,
618 SPA_IMPORT_ASSEMBLE
619 } spa_import_type_t;
620
621 /*
622 * Should we force sending TRIM commands even to devices which evidently
623 * don't support it?
624 * OFF: no, only send to devices which indicated support
625 * ON: yes, force send to everybody
626 */
627 typedef enum {
628 SPA_FORCE_TRIM_OFF = 0, /* default */
629 SPA_FORCE_TRIM_ON
630 } spa_force_trim_t;
631
632 /*
633 * Should we send TRIM commands in-line during normal pool operation while
634 * deleting stuff?
635 * OFF: no
636 * ON: yes
637 */
638 typedef enum {
639 SPA_AUTO_TRIM_OFF = 0, /* default */
640 SPA_AUTO_TRIM_ON
641 } spa_auto_trim_t;
642
643 /* state manipulation functions */
644 extern int spa_open(const char *pool, spa_t **, void *tag);
645 extern int spa_open_rewind(const char *pool, spa_t **, void *tag,
646 nvlist_t *policy, nvlist_t **config);
647 extern int spa_get_stats(const char *name, nvlist_t **config,
648 char *altroot, size_t buflen);
649 extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props,
650 nvlist_t *zplprops);
651 extern int spa_import_rootpool(char *devpath, char *devid);
652 extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props,
653 uint64_t flags);
654 extern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
655 extern int spa_destroy(char *pool);
656 extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force,
657 boolean_t hardforce, boolean_t saveconfig);
658 extern int spa_reset(char *pool);
659 extern void spa_async_request(spa_t *spa, int flag);
660 extern void spa_async_unrequest(spa_t *spa, int flag);
661 extern void spa_async_suspend(spa_t *spa);
662 extern void spa_async_resume(spa_t *spa);
663 extern spa_t *spa_inject_addref(char *pool);
664 extern void spa_inject_delref(spa_t *spa);
665 extern void spa_scan_stat_init(spa_t *spa);
666 extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps);
667
668 #define SPA_ASYNC_CONFIG_UPDATE 0x01
669 #define SPA_ASYNC_REMOVE 0x02
670 #define SPA_ASYNC_PROBE 0x04
671 #define SPA_ASYNC_RESILVER_DONE 0x08
672 #define SPA_ASYNC_RESILVER 0x10
673 #define SPA_ASYNC_AUTOEXPAND 0x20
674 #define SPA_ASYNC_REMOVE_DONE 0x40
675 #define SPA_ASYNC_REMOVE_STOP 0x80
676 #define SPA_ASYNC_L2CACHE_REBUILD 0x100
677 #define SPA_ASYNC_MAN_TRIM_TASKQ_DESTROY 0x200
678
679 /*
680 * Controls the behavior of spa_vdev_remove().
681 */
682 #define SPA_REMOVE_UNSPARE 0x01
683 #define SPA_REMOVE_DONE 0x02
684
685 /* device manipulation */
686 extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
687 extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot,
688 int replacing);
689 extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
690 int replace_done);
691 extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare);
692 extern boolean_t spa_vdev_remove_active(spa_t *spa);
693 extern int spa_vdev_setl2adddt(spa_t *spa, uint64_t guid, const char *newval);
694 extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
695 extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru);
696 extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
697 nvlist_t *props, boolean_t exp);
698
699 extern int spa_load_vdev_props(spa_t *spa);
700
701 extern int spa_vdev_prop_validate(spa_t *spa, uint64_t vdev_guid,
702 nvlist_t *nvp);
703 extern int spa_vdev_prop_set(spa_t *spa, uint64_t vdev_guid, nvlist_t *nvp);
704 extern int spa_vdev_prop_get(spa_t *spa, uint64_t vdev_guid, nvlist_t **nvp);
705 extern int spa_vdev_props_sync_task_do(spa_t *spa);
706
707 /* spare state (which is global across all pools) */
708 extern void spa_spare_add(vdev_t *vd);
709 extern void spa_spare_remove(vdev_t *vd);
710 extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt);
711 extern void spa_spare_activate(vdev_t *vd);
712
713 /* L2ARC state (which is global across all pools) */
714 extern void spa_l2cache_add(vdev_t *vd);
715 extern void spa_l2cache_remove(vdev_t *vd);
716 extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
717 extern void spa_l2cache_activate(vdev_t *vd);
718 extern void spa_l2cache_drop(spa_t *spa);
719
720 /* scanning */
721 extern int spa_scan(spa_t *spa, pool_scan_func_t func);
722 extern int spa_scan_stop(spa_t *spa);
723 extern int spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t flag);
724
725 /* trimming */
726 extern void spa_man_trim(spa_t *spa, uint64_t rate);
727 extern void spa_man_trim_stop(spa_t *spa);
728 extern void spa_get_trim_prog(spa_t *spa, uint64_t *prog, uint64_t *rate,
729 uint64_t *start_time, uint64_t *stop_time);
730 extern void spa_trim_stop_wait(spa_t *spa);
731
732 /* spa syncing */
733 extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
734 extern void spa_sync_allpools(void);
735
736 /* spa namespace global mutex */
737 extern kmutex_t spa_namespace_lock;
738
739 /*
740 * SPA configuration functions in spa_config.c
741 */
742
743 #define SPA_CONFIG_UPDATE_POOL 0
744 #define SPA_CONFIG_UPDATE_VDEVS 1
745
746 extern void spa_config_sync(spa_t *, boolean_t, boolean_t);
747 extern void spa_config_load(void);
748 extern nvlist_t *spa_all_configs(uint64_t *);
749 extern void spa_config_set(spa_t *spa, nvlist_t *config);
750 extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
751 int getstats);
752 extern void spa_config_update(spa_t *spa, int what);
753
754 /*
755 * Miscellaneous SPA routines in spa_misc.c
756 */
757
758 /* dedup ceiling helper functions */
759 extern uint64_t spa_get_ddts_size(spa_t *spa, boolean_t phys);
760 extern int spa_get_l2arc_ddt_utilization(spa_t *spa);
761 extern boolean_t spa_enable_dedup_cap(spa_t *spa);
762
763 /* Namespace manipulation */
764 extern spa_t *spa_lookup(const char *name);
765 extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot);
766 extern void spa_remove(spa_t *spa);
767 extern spa_t *spa_next(spa_t *prev);
768
769 /* Refcount functions */
770 extern void spa_open_ref(spa_t *spa, void *tag);
771 extern void spa_close(spa_t *spa, void *tag);
772 extern void spa_async_close(spa_t *spa, void *tag);
773 extern boolean_t spa_refcount_zero(spa_t *spa);
774
775 #define SCL_NONE 0x00
776 #define SCL_CONFIG 0x01
777 #define SCL_STATE 0x02
778 #define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */
779 #define SCL_ALLOC 0x08
780 #define SCL_ZIO 0x10
781 #define SCL_FREE 0x20
782 #define SCL_VDEV 0x40
783 #define SCL_LOCKS 7
784 #define SCL_ALL ((1 << SCL_LOCKS) - 1)
785 #define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO)
786
787 /* Pool configuration locks */
788 extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw);
789 extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw);
790 extern void spa_config_exit(spa_t *spa, int locks, void *tag);
791 extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
792
793 /* Pool vdev add/remove lock */
794 extern uint64_t spa_vdev_enter(spa_t *spa);
795 extern uint64_t spa_vdev_config_enter(spa_t *spa);
796 extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg,
797 int error, char *tag);
798 extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error);
799
800 /* Pool vdev state change lock */
801 extern void spa_vdev_state_enter(spa_t *spa, int oplock);
802 extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error);
803
804 /* Log state */
805 typedef enum spa_log_state {
806 SPA_LOG_UNKNOWN = 0, /* unknown log state */
807 SPA_LOG_MISSING, /* missing log(s) */
808 SPA_LOG_CLEAR, /* clear the log(s) */
809 SPA_LOG_GOOD, /* log(s) are good */
810 } spa_log_state_t;
811
812 extern spa_log_state_t spa_get_log_state(spa_t *spa);
813 extern void spa_set_log_state(spa_t *spa, spa_log_state_t state);
814 extern int spa_offline_log(spa_t *spa);
815
816 /* Log claim callback */
817 extern void spa_claim_notify(zio_t *zio);
818
819 /* Accessor functions */
820 extern boolean_t spa_shutting_down(spa_t *spa);
821 extern struct dsl_pool *spa_get_dsl(spa_t *spa);
822 extern boolean_t spa_is_initializing(spa_t *spa);
823 extern blkptr_t *spa_get_rootblkptr(spa_t *spa);
824 extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
825 extern void spa_altroot(spa_t *, char *, size_t);
826 extern int spa_sync_pass(spa_t *spa);
827 extern char *spa_name(spa_t *spa);
828 extern uint64_t spa_guid(spa_t *spa);
829 extern uint64_t spa_load_guid(spa_t *spa);
830 extern uint64_t spa_last_synced_txg(spa_t *spa);
831 extern uint64_t spa_first_txg(spa_t *spa);
832 extern uint64_t spa_syncing_txg(spa_t *spa);
833 extern uint64_t spa_final_dirty_txg(spa_t *spa);
834 extern uint64_t spa_version(spa_t *spa);
835 extern int spa_get_obj_mtx_sz(spa_t *spa);
836 extern pool_state_t spa_state(spa_t *spa);
837 extern spa_load_state_t spa_load_state(spa_t *spa);
838 extern uint64_t spa_freeze_txg(spa_t *spa);
839 extern uint64_t spa_get_worst_case_asize(spa_t *spa, uint64_t lsize);
840 extern uint64_t spa_get_dspace(spa_t *spa);
841 extern uint64_t spa_get_slop_space(spa_t *spa);
842 extern void spa_update_dspace(spa_t *spa);
843 extern void spa_update_latency(spa_t *spa);
844 extern uint64_t spa_version(spa_t *spa);
845 extern boolean_t spa_deflate(spa_t *spa);
846 extern metaslab_class_t *spa_normal_class(spa_t *spa);
847 extern metaslab_class_t *spa_log_class(spa_t *spa);
848 extern metaslab_class_t *spa_special_class(spa_t *spa);
849 extern void spa_evicting_os_register(spa_t *, objset_t *os);
850 extern void spa_evicting_os_deregister(spa_t *, objset_t *os);
851 extern void spa_evicting_os_wait(spa_t *spa);
852 extern uint64_t spa_class_alloc_percentage(metaslab_class_t *mc);
853 extern int spa_max_replication(spa_t *spa);
854 extern int spa_prev_software_version(spa_t *spa);
855 extern int spa_busy(void);
856 extern uint8_t spa_get_failmode(spa_t *spa);
857 extern boolean_t spa_suspended(spa_t *spa);
858 extern uint64_t spa_bootfs(spa_t *spa);
859 extern uint64_t spa_delegation(spa_t *spa);
860 extern objset_t *spa_meta_objset(spa_t *spa);
861 extern uint64_t spa_deadman_synctime(spa_t *spa);
862 extern spa_force_trim_t spa_get_force_trim(spa_t *spa);
863 extern spa_auto_trim_t spa_get_auto_trim(spa_t *spa);
864
865 /* Miscellaneous support routines */
866 extern void spa_activate_mos_feature(spa_t *spa, const char *feature,
867 dmu_tx_t *tx);
868 extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature);
869 extern int spa_rename(const char *oldname, const char *newname);
870 extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid);
871 extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid);
872 extern boolean_t spa_config_guid_exists(uint64_t pool_guid);
873 extern char *spa_strdup(const char *);
874 extern void spa_strfree(char *);
875 extern uint64_t spa_get_random(uint64_t range);
876 extern uint64_t spa_generate_guid(spa_t *spa);
877 extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp);
878 extern void spa_freeze(spa_t *spa);
879 extern int spa_change_guid(spa_t *spa);
880 extern void spa_upgrade(spa_t *spa, uint64_t version);
881 extern void spa_evict_all(void);
882 extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid,
883 boolean_t l2cache);
884 extern boolean_t spa_has_spare(spa_t *, uint64_t guid);
885 extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva);
886 extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp);
887 extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp);
888 extern boolean_t spa_has_slogs(spa_t *spa);
889 extern boolean_t spa_is_root(spa_t *spa);
890 extern boolean_t spa_writeable(spa_t *spa);
891 extern boolean_t spa_has_pending_synctask(spa_t *spa);
892 extern boolean_t spa_has_special(spa_t *spa);
893 extern int spa_maxblocksize(spa_t *spa);
894 extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp);
895 extern boolean_t spa_wbc_present(spa_t *spa);
896 extern boolean_t spa_wbc_active(spa_t *spa);
897 extern struct zfs_autosnap *spa_get_autosnap(spa_t *spa);
898 extern void wbc_purge_window(spa_t *spa, dmu_tx_t *tx);
899
900 extern int spa_mode(spa_t *spa);
901 extern uint64_t zfs_strtonum(const char *str, char **nptr);
902
903 /* Selector for dynamic I/O balancing between special and regular vdevs */
904 extern boolean_t spa_use_special_class(spa_t *spa);
905
906 /* Pool perfmon thread management */
907 extern void spa_start_perfmon_thread(spa_t *spa);
908 extern boolean_t spa_stop_perfmon_thread(spa_t *spa);
909
910 extern char *spa_his_ievent_table[];
911
912 extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx);
913 extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read,
914 char *his_buf);
915 extern int spa_history_log(spa_t *spa, const char *his_buf);
916 extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl);
917 extern void spa_history_log_version(spa_t *spa, const char *operation);
918 extern void spa_history_log_internal(spa_t *spa, const char *operation,
919 dmu_tx_t *tx, const char *fmt, ...);
920 extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op,
921 dmu_tx_t *tx, const char *fmt, ...);
922 extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
923 dmu_tx_t *tx, const char *fmt, ...);
924
925 /* error handling */
926 struct zbookmark_phys;
927 extern void spa_log_error(spa_t *spa, zio_t *zio);
928 extern void zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd,
929 zio_t *zio, uint64_t stateoroffset, uint64_t length);
930 extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
931 extern void zfs_post_state_change(spa_t *spa, vdev_t *vd);
932 extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
933 extern uint64_t spa_get_errlog_size(spa_t *spa);
934 extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
935 extern void spa_errlog_rotate(spa_t *spa);
936 extern void spa_errlog_drain(spa_t *spa);
937 extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
938 extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
939
940 /* vdev cache */
941 extern void vdev_cache_stat_init(void);
942 extern void vdev_cache_stat_fini(void);
943
944 /* Initialization and termination */
945 extern void spa_init(int flags);
946 extern void spa_fini(void);
947 extern void spa_boot_init(void);
948
949 /* properties */
950 extern int spa_prop_set(spa_t *spa, nvlist_t *nvp);
951 extern int spa_prop_get(spa_t *spa, nvlist_t **nvp);
952 extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx);
953 extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t);
954
955 /* asynchronous event notification */
956 extern void spa_event_notify(spa_t *spa, vdev_t *vdev, nvlist_t *hist_nvl,
957 const char *name);
958
959 extern int spa_wbc_mode(const char *name);
960
961 typedef enum spa_wbc_mode {
962 WBC_MODE_OFF,
963 WBC_MODE_ACTIVE,
964 WBC_MODE_PASSIVE
965 } spa_wbc_mode_t;
966
967 /* TRIM/UNMAP kstat update */
968 extern void spa_trimstats_update(spa_t *spa, uint64_t extents, uint64_t bytes,
969 uint64_t extents_skipped, uint64_t bytes_skipped);
970 extern void spa_trimstats_auto_slow_incr(spa_t *spa);
971
972 #ifdef ZFS_DEBUG
973 #define dprintf_bp(bp, fmt, ...) do { \
974 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
975 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \
976 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \
977 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \
978 kmem_free(__blkbuf, BP_SPRINTF_LEN); \
979 } \
980 _NOTE(CONSTCOND) } while (0)
981 #else
982 #define dprintf_bp(bp, fmt, ...)
983 #endif
984
985 extern boolean_t spa_debug_enabled(spa_t *spa);
986 #define spa_dbgmsg(spa, ...) \
987 { \
988 if (spa_debug_enabled(spa)) \
989 zfs_dbgmsg(__VA_ARGS__); \
990 }
991
992 extern int spa_mode_global; /* mode, e.g. FREAD | FWRITE */
993
994 #ifdef __cplusplus
995 }
996 #endif
997
998 #endif /* _SYS_SPA_H */