1/*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21/*
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
26 * Copyright 2013 Saso Kiselkov. All rights reserved.
27 * Copyright (c) 2014 Integros [integros.com]
28 */
29
30#ifndef _SYS_SPA_H
31#define _SYS_SPA_H
32
33#include <sys/avl.h>
34#include <sys/zfs_context.h>
35#include <sys/nvpair.h>
36#include <sys/sysmacros.h>
37#include <sys/types.h>
38#include <sys/fs/zfs.h>
39
40#ifdef __cplusplus
41extern "C" {
42#endif
43
44/*
45 * Forward references that lots of things need.
46 */
47typedef struct spa spa_t;
48typedef struct vdev vdev_t;
49typedef struct metaslab metaslab_t;
50typedef struct metaslab_group metaslab_group_t;
51typedef struct metaslab_class metaslab_class_t;
52typedef struct zio zio_t;
53typedef struct zilog zilog_t;
54typedef struct spa_aux_vdev spa_aux_vdev_t;
55typedef struct ddt ddt_t;
56typedef struct ddt_entry ddt_entry_t;
57struct dsl_pool;
58struct dsl_dataset;
59
60/*
61 * General-purpose 32-bit and 64-bit bitfield encodings.
62 */
63#define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len))
64#define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len))
65#define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low))
66#define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low))
67
68#define BF32_GET(x, low, len) BF32_DECODE(x, low, len)
69#define BF64_GET(x, low, len) BF64_DECODE(x, low, len)
70
71#define BF32_SET(x, low, len, val) do { \
72 ASSERT3U(val, <, 1U << (len)); \
73 ASSERT3U(low + len, <=, 32); \
74 (x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \
75_NOTE(CONSTCOND) } while (0)
76
77#define BF64_SET(x, low, len, val) do { \
78 ASSERT3U(val, <, 1ULL << (len)); \
79 ASSERT3U(low + len, <=, 64); \
80 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \
81_NOTE(CONSTCOND) } while (0)
82
83#define BF32_GET_SB(x, low, len, shift, bias) \
84 ((BF32_GET(x, low, len) + (bias)) << (shift))
85#define BF64_GET_SB(x, low, len, shift, bias) \
86 ((BF64_GET(x, low, len) + (bias)) << (shift))
87
88#define BF32_SET_SB(x, low, len, shift, bias, val) do { \
89 ASSERT(IS_P2ALIGNED(val, 1U << shift)); \
90 ASSERT3S((val) >> (shift), >=, bias); \
91 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \
92_NOTE(CONSTCOND) } while (0)
93#define BF64_SET_SB(x, low, len, shift, bias, val) do { \
94 ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \
95 ASSERT3S((val) >> (shift), >=, bias); \
96 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \
97_NOTE(CONSTCOND) } while (0)
98
99/*
100 * We currently support block sizes from 512 bytes to 16MB.
101 * The benefits of larger blocks, and thus larger IO, need to be weighed
102 * against the cost of COWing a giant block to modify one byte, and the
103 * large latency of reading or writing a large block.
104 *
105 * Note that although blocks up to 16MB are supported, the recordsize
106 * property can not be set larger than zfs_max_recordsize (default 1MB).
107 * See the comment near zfs_max_recordsize in dsl_dataset.c for details.
108 *
109 * Note that although the LSIZE field of the blkptr_t can store sizes up
110 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to
111 * 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
112 */
113#define SPA_MINBLOCKSHIFT 9
114#define SPA_OLD_MAXBLOCKSHIFT 17
115#define SPA_MAXBLOCKSHIFT 24
116#define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT)
117#define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT)
118#define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT)
119
120/*
121 * Default maximum supported logical ashift.
122 *
123 * The current 8k allocation block size limit is due to the 8k
124 * aligned/sized operations performed by vdev_probe() on
125 * vdev_label->vl_pad2. Using another "safe region" for these tests
126 * would allow the limit to be raised to 16k, at the expense of
127 * only having 8 available uberblocks in the label area.
128 */
129#define SPA_MAXASHIFT 13
130
131/*
132 * Default minimum supported logical ashift.
133 */
134#define SPA_MINASHIFT SPA_MINBLOCKSHIFT
135
136/*
137 * Size of block to hold the configuration data (a packed nvlist)
138 */
139#define SPA_CONFIG_BLOCKSIZE (1ULL << 14)
140
141/*
142 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
143 * The ASIZE encoding should be at least 64 times larger (6 more bits)
144 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
145 * overhead, three DVAs per bp, plus one more bit in case we do anything
146 * else that expands the ASIZE.
147 */
148#define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */
149#define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */
150#define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */
151
152#define SPA_COMPRESSBITS 7
153
154/*
155 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
156 * The members of the dva_t should be considered opaque outside the SPA.
157 */
158typedef struct dva {
159 uint64_t dva_word[2];
160} dva_t;
161
162/*
163 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
164 */
165typedef struct zio_cksum {
166 uint64_t zc_word[4];
167} zio_cksum_t;
168
169/*
170 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept
171 * secret and is suitable for use in MAC algorithms as the key.
172 */
173typedef struct zio_cksum_salt {
174 uint8_t zcs_bytes[32];
175} zio_cksum_salt_t;
176
177/*
178 * Each block is described by its DVAs, time of birth, checksum, etc.
179 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
180 *
181 * 64 56 48 40 32 24 16 8 0
182 * +-------+-------+-------+-------+-------+-------+-------+-------+
183 * 0 | vdev1 | GRID | ASIZE |
184 * +-------+-------+-------+-------+-------+-------+-------+-------+
185 * 1 |G| offset1 |
186 * +-------+-------+-------+-------+-------+-------+-------+-------+
187 * 2 | vdev2 | GRID | ASIZE |
188 * +-------+-------+-------+-------+-------+-------+-------+-------+
189 * 3 |G| offset2 |
190 * +-------+-------+-------+-------+-------+-------+-------+-------+
191 * 4 | vdev3 | GRID | ASIZE |
192 * +-------+-------+-------+-------+-------+-------+-------+-------+
193 * 5 |G| offset3 |
194 * +-------+-------+-------+-------+-------+-------+-------+-------+
195 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE |
196 * +-------+-------+-------+-------+-------+-------+-------+-------+
197 * 7 | padding |
198 * +-------+-------+-------+-------+-------+-------+-------+-------+
199 * 8 | padding |
200 * +-------+-------+-------+-------+-------+-------+-------+-------+
201 * 9 | physical birth txg |
202 * +-------+-------+-------+-------+-------+-------+-------+-------+
203 * a | logical birth txg |
204 * +-------+-------+-------+-------+-------+-------+-------+-------+
205 * b | fill count |
206 * +-------+-------+-------+-------+-------+-------+-------+-------+
207 * c | checksum[0] |
208 * +-------+-------+-------+-------+-------+-------+-------+-------+
209 * d | checksum[1] |
210 * +-------+-------+-------+-------+-------+-------+-------+-------+
211 * e | checksum[2] |
212 * +-------+-------+-------+-------+-------+-------+-------+-------+
213 * f | checksum[3] |
214 * +-------+-------+-------+-------+-------+-------+-------+-------+
215 *
216 * Legend:
217 *
218 * vdev virtual device ID
219 * offset offset into virtual device
220 * LSIZE logical size
221 * PSIZE physical size (after compression)
222 * ASIZE allocated size (including RAID-Z parity and gang block headers)
223 * GRID RAID-Z layout information (reserved for future use)
224 * cksum checksum function
225 * comp compression function
226 * G gang block indicator
227 * B byteorder (endianness)
228 * D dedup
229 * X encryption (on version 30, which is not supported)
230 * E blkptr_t contains embedded data (see below)
231 * lvl level of indirection
232 * type DMU object type
233 * phys birth txg of block allocation; zero if same as logical birth txg
234 * log. birth transaction group in which the block was logically born
235 * fill count number of non-zero blocks under this bp
236 * checksum[4] 256-bit checksum of the data this bp describes
237 */
238
239/*
240 * "Embedded" blkptr_t's don't actually point to a block, instead they
241 * have a data payload embedded in the blkptr_t itself. See the comment
242 * in blkptr.c for more details.
243 *
244 * The blkptr_t is laid out as follows:
245 *
246 * 64 56 48 40 32 24 16 8 0
247 * +-------+-------+-------+-------+-------+-------+-------+-------+
248 * 0 | payload |
249 * 1 | payload |
250 * 2 | payload |
251 * 3 | payload |
252 * 4 | payload |
253 * 5 | payload |
254 * +-------+-------+-------+-------+-------+-------+-------+-------+
255 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE |
256 * +-------+-------+-------+-------+-------+-------+-------+-------+
257 * 7 | payload |
258 * 8 | payload |
259 * 9 | payload |
260 * +-------+-------+-------+-------+-------+-------+-------+-------+
261 * a | logical birth txg |
262 * +-------+-------+-------+-------+-------+-------+-------+-------+
263 * b | payload |
264 * c | payload |
265 * d | payload |
266 * e | payload |
267 * f | payload |
268 * +-------+-------+-------+-------+-------+-------+-------+-------+
269 *
270 * Legend:
271 *
272 * payload contains the embedded data
273 * B (byteorder) byteorder (endianness)
274 * D (dedup) padding (set to zero)
275 * X encryption (set to zero; see above)
276 * E (embedded) set to one
277 * lvl indirection level
278 * type DMU object type
279 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*)
280 * comp compression function of payload
281 * PSIZE size of payload after compression, in bytes
282 * LSIZE logical size of payload, in bytes
283 * note that 25 bits is enough to store the largest
284 * "normal" BP's LSIZE (2^16 * 2^9) in bytes
285 * log. birth transaction group in which the block was logically born
286 *
287 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
288 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
289 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
290 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
291 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must
292 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before
293 * other macros, as they assert that they are only used on BP's of the correct
294 * "embedded-ness".
295 */
296
297#define BPE_GET_ETYPE(bp) \
298 (ASSERT(BP_IS_EMBEDDED(bp)), \
299 BF64_GET((bp)->blk_prop, 40, 8))
300#define BPE_SET_ETYPE(bp, t) do { \
301 ASSERT(BP_IS_EMBEDDED(bp)); \
302 BF64_SET((bp)->blk_prop, 40, 8, t); \
303_NOTE(CONSTCOND) } while (0)
304
305#define BPE_GET_LSIZE(bp) \
306 (ASSERT(BP_IS_EMBEDDED(bp)), \
307 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
308#define BPE_SET_LSIZE(bp, x) do { \
309 ASSERT(BP_IS_EMBEDDED(bp)); \
310 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
311_NOTE(CONSTCOND) } while (0)
312
313#define BPE_GET_PSIZE(bp) \
314 (ASSERT(BP_IS_EMBEDDED(bp)), \
315 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
316#define BPE_SET_PSIZE(bp, x) do { \
317 ASSERT(BP_IS_EMBEDDED(bp)); \
318 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
319_NOTE(CONSTCOND) } while (0)
320
321typedef enum bp_embedded_type {
322 BP_EMBEDDED_TYPE_DATA,
323 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */
324 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED
325} bp_embedded_type_t;
326
327#define BPE_NUM_WORDS 14
328#define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
329#define BPE_IS_PAYLOADWORD(bp, wp) \
330 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
331
332#define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */
333#define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */
334
335/*
336 * A block is a hole when it has either 1) never been written to, or
337 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
338 * without physically allocating disk space. Holes are represented in the
339 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is
340 * done through the BP_IS_HOLE macro. For holes, the logical size, level,
341 * DMU object type, and birth times are all also stored for holes that
342 * were written to at some point (i.e. were punched after having been filled).
343 */
344typedef struct blkptr {
345 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
346 uint64_t blk_prop; /* size, compression, type, etc */
347 uint64_t blk_pad[2]; /* Extra space for the future */
348 uint64_t blk_phys_birth; /* txg when block was allocated */
349 uint64_t blk_birth; /* transaction group at birth */
350 uint64_t blk_fill; /* fill count */
351 zio_cksum_t blk_cksum; /* 256-bit checksum */
352} blkptr_t;
353
354/*
355 * Macros to get and set fields in a bp or DVA.
356 */
357#define DVA_GET_ASIZE(dva) \
358 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
359#define DVA_SET_ASIZE(dva, x) \
360 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
361 SPA_MINBLOCKSHIFT, 0, x)
362
363#define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8)
364#define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x)
365
366#define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32)
367#define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x)
368
369#define DVA_GET_OFFSET(dva) \
370 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
371#define DVA_SET_OFFSET(dva, x) \
372 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
373
374#define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1)
375#define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x)
376
377#define BP_GET_LSIZE(bp) \
378 (BP_IS_EMBEDDED(bp) ? \
379 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
380 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
381#define BP_SET_LSIZE(bp, x) do { \
382 ASSERT(!BP_IS_EMBEDDED(bp)); \
383 BF64_SET_SB((bp)->blk_prop, \
384 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
385_NOTE(CONSTCOND) } while (0)
386
387#define BP_GET_PSIZE(bp) \
388 (BP_IS_EMBEDDED(bp) ? 0 : \
389 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1))
390#define BP_SET_PSIZE(bp, x) do { \
391 ASSERT(!BP_IS_EMBEDDED(bp)); \
392 BF64_SET_SB((bp)->blk_prop, \
393 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
394_NOTE(CONSTCOND) } while (0)
395
396#define BP_GET_COMPRESS(bp) \
397 BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS)
398#define BP_SET_COMPRESS(bp, x) \
399 BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x)
400
401#define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1)
402#define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x)
403
404#define BP_GET_CHECKSUM(bp) \
405 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \
406 BF64_GET((bp)->blk_prop, 40, 8))
407#define BP_SET_CHECKSUM(bp, x) do { \
408 ASSERT(!BP_IS_EMBEDDED(bp)); \
409 BF64_SET((bp)->blk_prop, 40, 8, x); \
410_NOTE(CONSTCOND) } while (0)
411
412#define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8)
413#define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x)
414
415#define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5)
416#define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x)
417
418#define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1)
419#define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x)
420
421#define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1)
422#define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x)
423
424#define BP_PHYSICAL_BIRTH(bp) \
425 (BP_IS_EMBEDDED(bp) ? 0 : \
426 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
427
428#define BP_SET_BIRTH(bp, logical, physical) \
429{ \
430 ASSERT(!BP_IS_EMBEDDED(bp)); \
431 (bp)->blk_birth = (logical); \
432 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
433}
434
435#define BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill)
436
437#define BP_GET_ASIZE(bp) \
438 (BP_IS_EMBEDDED(bp) ? 0 : \
439 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
440 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
441 DVA_GET_ASIZE(&(bp)->blk_dva[2]))
442
443#define BP_GET_UCSIZE(bp) \
444 ((BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) ? \
445 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
446
447#define BP_GET_NDVAS(bp) \
448 (BP_IS_EMBEDDED(bp) ? 0 : \
449 !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
450 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
451 !!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
452
453#define BP_COUNT_GANG(bp) \
454 (BP_IS_EMBEDDED(bp) ? 0 : \
455 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \
456 DVA_GET_GANG(&(bp)->blk_dva[1]) + \
457 DVA_GET_GANG(&(bp)->blk_dva[2])))
458
459#define DVA_EQUAL(dva1, dva2) \
460 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
461 (dva1)->dva_word[0] == (dva2)->dva_word[0])
462
463#define BP_EQUAL(bp1, bp2) \
464 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \
465 (bp1)->blk_birth == (bp2)->blk_birth && \
466 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \
467 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \
468 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
469
470#define ZIO_CHECKSUM_EQUAL(zc1, zc2) \
471 (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
472 ((zc1).zc_word[1] - (zc2).zc_word[1]) | \
473 ((zc1).zc_word[2] - (zc2).zc_word[2]) | \
474 ((zc1).zc_word[3] - (zc2).zc_word[3])))
475
476#define ZIO_CHECKSUM_IS_ZERO(zc) \
477 (0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \
478 (zc)->zc_word[2] | (zc)->zc_word[3]))
479
480#define ZIO_CHECKSUM_BSWAP(zcp) \
481{ \
482 (zcp)->zc_word[0] = BSWAP_64((zcp)->zc_word[0]); \
483 (zcp)->zc_word[1] = BSWAP_64((zcp)->zc_word[1]); \
484 (zcp)->zc_word[2] = BSWAP_64((zcp)->zc_word[2]); \
485 (zcp)->zc_word[3] = BSWAP_64((zcp)->zc_word[3]); \
486}
487
488
489#define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0)
490
491#define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \
492{ \
493 (zcp)->zc_word[0] = w0; \
494 (zcp)->zc_word[1] = w1; \
495 (zcp)->zc_word[2] = w2; \
496 (zcp)->zc_word[3] = w3; \
497}
498
499#define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0])
500#define BP_IS_GANG(bp) \
501 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp)))
502#define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \
503 (dva)->dva_word[1] == 0ULL)
504#define BP_IS_HOLE(bp) \
505 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp)))
506
507/* BP_IS_RAIDZ(bp) assumes no block compression */
508#define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
509 BP_GET_PSIZE(bp))
510
511#define BP_ZERO(bp) \
512{ \
513 (bp)->blk_dva[0].dva_word[0] = 0; \
514 (bp)->blk_dva[0].dva_word[1] = 0; \
515 (bp)->blk_dva[1].dva_word[0] = 0; \
516 (bp)->blk_dva[1].dva_word[1] = 0; \
517 (bp)->blk_dva[2].dva_word[0] = 0; \
518 (bp)->blk_dva[2].dva_word[1] = 0; \
519 (bp)->blk_prop = 0; \
520 (bp)->blk_pad[0] = 0; \
521 (bp)->blk_pad[1] = 0; \
522 (bp)->blk_phys_birth = 0; \
523 (bp)->blk_birth = 0; \
524 (bp)->blk_fill = 0; \
525 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \
526}
527
528#if BYTE_ORDER == _BIG_ENDIAN
529#define ZFS_HOST_BYTEORDER (0ULL)
530#else
531#define ZFS_HOST_BYTEORDER (1ULL)
532#endif
533
534#define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
535
536#define BP_SPRINTF_LEN 320
537
538/*
539 * This macro allows code sharing between zfs, libzpool, and mdb.
540 * 'func' is either snprintf() or mdb_snprintf().
541 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
542 */
543#define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
544{ \
545 static const char *copyname[] = \
546 { "zero", "single", "double", "triple" }; \
547 int len = 0; \
548 int copies = 0; \
549 \
550 if (bp == NULL) { \
551 len += func(buf + len, size - len, "<NULL>"); \
552 } else if (BP_IS_HOLE(bp)) { \
553 len += func(buf + len, size - len, \
554 "HOLE [L%llu %s] " \
555 "size=%llxL birth=%lluL", \
556 (u_longlong_t)BP_GET_LEVEL(bp), \
557 type, \
558 (u_longlong_t)BP_GET_LSIZE(bp), \
559 (u_longlong_t)bp->blk_birth); \
560 } else if (BP_IS_EMBEDDED(bp)) { \
561 len = func(buf + len, size - len, \
562 "EMBEDDED [L%llu %s] et=%u %s " \
563 "size=%llxL/%llxP birth=%lluL", \
564 (u_longlong_t)BP_GET_LEVEL(bp), \
565 type, \
566 (int)BPE_GET_ETYPE(bp), \
567 compress, \
568 (u_longlong_t)BPE_GET_LSIZE(bp), \
569 (u_longlong_t)BPE_GET_PSIZE(bp), \
570 (u_longlong_t)bp->blk_birth); \
571 } else { \
572 for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \
573 const dva_t *dva = &bp->blk_dva[d]; \
574 if (DVA_IS_VALID(dva)) \
575 copies++; \
576 len += func(buf + len, size - len, \
577 "DVA[%d]=<%llu:%llx:%llx>%c", d, \
578 (u_longlong_t)DVA_GET_VDEV(dva), \
579 (u_longlong_t)DVA_GET_OFFSET(dva), \
580 (u_longlong_t)DVA_GET_ASIZE(dva), \
581 ws); \
582 } \
583 if (BP_IS_GANG(bp) && \
584 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \
585 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \
586 copies--; \
587 len += func(buf + len, size - len, \
588 "[L%llu %s] %s %s %s %s %s %s%c" \
589 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \
590 "cksum=%llx:%llx:%llx:%llx", \
591 (u_longlong_t)BP_GET_LEVEL(bp), \
592 type, \
593 checksum, \
594 compress, \
595 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \
596 BP_IS_GANG(bp) ? "gang" : "contiguous", \
597 BP_GET_DEDUP(bp) ? "dedup" : "unique", \
598 copyname[copies], \
599 ws, \
600 (u_longlong_t)BP_GET_LSIZE(bp), \
601 (u_longlong_t)BP_GET_PSIZE(bp), \
602 (u_longlong_t)bp->blk_birth, \
603 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \
604 (u_longlong_t)BP_GET_FILL(bp), \
605 ws, \
606 (u_longlong_t)bp->blk_cksum.zc_word[0], \
607 (u_longlong_t)bp->blk_cksum.zc_word[1], \
608 (u_longlong_t)bp->blk_cksum.zc_word[2], \
609 (u_longlong_t)bp->blk_cksum.zc_word[3]); \
610 } \
611 ASSERT(len < size); \
612}
613
614#include <sys/dmu.h>
615
616#define BP_GET_BUFC_TYPE(bp) \
617 (((BP_GET_LEVEL(bp) > 0) || (DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))) ? \
618 ARC_BUFC_METADATA : ARC_BUFC_DATA)
619
620typedef enum spa_import_type {
621 SPA_IMPORT_EXISTING,
622 SPA_IMPORT_ASSEMBLE
623} spa_import_type_t;
624
625/* state manipulation functions */
626extern int spa_open(const char *pool, spa_t **, void *tag);
627extern int spa_open_rewind(const char *pool, spa_t **, void *tag,
628 nvlist_t *policy, nvlist_t **config);
629extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot,
630 size_t buflen);
631extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props,
632 nvlist_t *zplprops);
633#ifdef illumos
634extern int spa_import_rootpool(char *devpath, char *devid);
635#else
636extern int spa_import_rootpool(const char *name);
637#endif
638extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props,
639 uint64_t flags);
640extern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
641extern int spa_destroy(char *pool);
642extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force,
643 boolean_t hardforce);
644extern int spa_reset(char *pool);
645extern void spa_async_request(spa_t *spa, int flag);
646extern void spa_async_unrequest(spa_t *spa, int flag);
647extern void spa_async_suspend(spa_t *spa);
648extern void spa_async_resume(spa_t *spa);
649extern spa_t *spa_inject_addref(char *pool);
650extern void spa_inject_delref(spa_t *spa);
651extern void spa_scan_stat_init(spa_t *spa);
652extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps);
653
654#define SPA_ASYNC_CONFIG_UPDATE 0x01
655#define SPA_ASYNC_REMOVE 0x02
656#define SPA_ASYNC_PROBE 0x04
657#define SPA_ASYNC_RESILVER_DONE 0x08
658#define SPA_ASYNC_RESILVER 0x10
659#define SPA_ASYNC_AUTOEXPAND 0x20
660#define SPA_ASYNC_REMOVE_DONE 0x40
661#define SPA_ASYNC_REMOVE_STOP 0x80
662
663/*
664 * Controls the behavior of spa_vdev_remove().
665 */
666#define SPA_REMOVE_UNSPARE 0x01
667#define SPA_REMOVE_DONE 0x02
668
669/* device manipulation */
670extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
671extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot,
672 int replacing);
673extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
674 int replace_done);
675extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare);
676extern boolean_t spa_vdev_remove_active(spa_t *spa);
677extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
678extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru);
679extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
680 nvlist_t *props, boolean_t exp);
681
682/* spare state (which is global across all pools) */
683extern void spa_spare_add(vdev_t *vd);
684extern void spa_spare_remove(vdev_t *vd);
685extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt);
686extern void spa_spare_activate(vdev_t *vd);
687
688/* L2ARC state (which is global across all pools) */
689extern void spa_l2cache_add(vdev_t *vd);
690extern void spa_l2cache_remove(vdev_t *vd);
691extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
692extern void spa_l2cache_activate(vdev_t *vd);
693extern void spa_l2cache_drop(spa_t *spa);
694
695/* scanning */
696extern int spa_scan(spa_t *spa, pool_scan_func_t func);
697extern int spa_scan_stop(spa_t *spa);
698
699/* spa syncing */
700extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
701extern void spa_sync_allpools(void);
702
703/* spa namespace global mutex */
704extern kmutex_t spa_namespace_lock;
705
706/*
707 * SPA configuration functions in spa_config.c
708 */
709
710#define SPA_CONFIG_UPDATE_POOL 0
711#define SPA_CONFIG_UPDATE_VDEVS 1
712
713extern void spa_config_sync(spa_t *, boolean_t, boolean_t);
714extern void spa_config_load(void);
715extern nvlist_t *spa_all_configs(uint64_t *);
716extern void spa_config_set(spa_t *spa, nvlist_t *config);
717extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
718 int getstats);
719extern void spa_config_update(spa_t *spa, int what);
720
721/*
722 * Miscellaneous SPA routines in spa_misc.c
723 */
724
725/* Namespace manipulation */
726extern spa_t *spa_lookup(const char *name);
727extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot);
728extern void spa_remove(spa_t *spa);
729extern spa_t *spa_next(spa_t *prev);
730
731/* Refcount functions */
732extern void spa_open_ref(spa_t *spa, void *tag);
733extern void spa_close(spa_t *spa, void *tag);
734extern void spa_async_close(spa_t *spa, void *tag);
735extern boolean_t spa_refcount_zero(spa_t *spa);
736
737#define SCL_NONE 0x00
738#define SCL_CONFIG 0x01
739#define SCL_STATE 0x02
740#define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */
741#define SCL_ALLOC 0x08
742#define SCL_ZIO 0x10
743#define SCL_FREE 0x20
744#define SCL_VDEV 0x40
745#define SCL_LOCKS 7
746#define SCL_ALL ((1 << SCL_LOCKS) - 1)
747#define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO)
748
749/* Pool configuration locks */
750extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw);
751extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw);
752extern void spa_config_exit(spa_t *spa, int locks, void *tag);
753extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
754
755/* Pool vdev add/remove lock */
756extern uint64_t spa_vdev_enter(spa_t *spa);
757extern uint64_t spa_vdev_config_enter(spa_t *spa);
758extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg,
759 int error, char *tag);
760extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error);
761
762/* Pool vdev state change lock */
763extern void spa_vdev_state_enter(spa_t *spa, int oplock);
764extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error);
765
766/* Log state */
767typedef enum spa_log_state {
768 SPA_LOG_UNKNOWN = 0, /* unknown log state */
769 SPA_LOG_MISSING, /* missing log(s) */
770 SPA_LOG_CLEAR, /* clear the log(s) */
771 SPA_LOG_GOOD, /* log(s) are good */
772} spa_log_state_t;
773
774extern spa_log_state_t spa_get_log_state(spa_t *spa);
775extern void spa_set_log_state(spa_t *spa, spa_log_state_t state);
776extern int spa_offline_log(spa_t *spa);
777
778/* Log claim callback */
779extern void spa_claim_notify(zio_t *zio);
780
781/* Accessor functions */
782extern boolean_t spa_shutting_down(spa_t *spa);
783extern struct dsl_pool *spa_get_dsl(spa_t *spa);
784extern boolean_t spa_is_initializing(spa_t *spa);
785extern blkptr_t *spa_get_rootblkptr(spa_t *spa);
786extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
787extern void spa_altroot(spa_t *, char *, size_t);
788extern int spa_sync_pass(spa_t *spa);
789extern char *spa_name(spa_t *spa);
790extern uint64_t spa_guid(spa_t *spa);
791extern uint64_t spa_load_guid(spa_t *spa);
792extern uint64_t spa_last_synced_txg(spa_t *spa);
793extern uint64_t spa_first_txg(spa_t *spa);
794extern uint64_t spa_syncing_txg(spa_t *spa);
795extern uint64_t spa_version(spa_t *spa);
796extern pool_state_t spa_state(spa_t *spa);
797extern spa_load_state_t spa_load_state(spa_t *spa);
798extern uint64_t spa_freeze_txg(spa_t *spa);
799extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize);
800extern uint64_t spa_get_dspace(spa_t *spa);
801extern uint64_t spa_get_slop_space(spa_t *spa);
802extern void spa_update_dspace(spa_t *spa);
803extern uint64_t spa_version(spa_t *spa);
804extern boolean_t spa_deflate(spa_t *spa);
805extern metaslab_class_t *spa_normal_class(spa_t *spa);
806extern metaslab_class_t *spa_log_class(spa_t *spa);
807extern void spa_evicting_os_register(spa_t *, objset_t *os);
808extern void spa_evicting_os_deregister(spa_t *, objset_t *os);
809extern void spa_evicting_os_wait(spa_t *spa);
810extern int spa_max_replication(spa_t *spa);
811extern int spa_prev_software_version(spa_t *spa);
812extern int spa_busy(void);
813extern uint8_t spa_get_failmode(spa_t *spa);
814extern boolean_t spa_suspended(spa_t *spa);
815extern uint64_t spa_bootfs(spa_t *spa);
816extern uint64_t spa_delegation(spa_t *spa);
817extern objset_t *spa_meta_objset(spa_t *spa);
818extern uint64_t spa_deadman_synctime(spa_t *spa);
819
820/* Miscellaneous support routines */
821extern void spa_activate_mos_feature(spa_t *spa, const char *feature,
822 dmu_tx_t *tx);
823extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature);
824extern int spa_rename(const char *oldname, const char *newname);
825extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid);
826extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid);
827extern char *spa_strdup(const char *);
828extern void spa_strfree(char *);
829extern uint64_t spa_get_random(uint64_t range);
830extern uint64_t spa_generate_guid(spa_t *spa);
831extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp);
832extern void spa_freeze(spa_t *spa);
833extern int spa_change_guid(spa_t *spa);
834extern void spa_upgrade(spa_t *spa, uint64_t version);
835extern void spa_evict_all(void);
836extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid,
837 boolean_t l2cache);
838extern boolean_t spa_has_spare(spa_t *, uint64_t guid);
839extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva);
840extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp);
841extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp);
842extern boolean_t spa_has_slogs(spa_t *spa);
843extern boolean_t spa_is_root(spa_t *spa);
844extern boolean_t spa_writeable(spa_t *spa);
845extern boolean_t spa_has_pending_synctask(spa_t *spa);
846extern int spa_maxblocksize(spa_t *spa);
847extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp);
848
849extern int spa_mode(spa_t *spa);
850extern uint64_t zfs_strtonum(const char *str, char **nptr);
851#define strtonum(str, nptr) zfs_strtonum((str), (nptr))
852
853extern char *spa_his_ievent_table[];
854
855extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx);
856extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read,
857 char *his_buf);
858extern int spa_history_log(spa_t *spa, const char *his_buf);
859extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl);
860extern void spa_history_log_version(spa_t *spa, const char *operation);
861extern void spa_history_log_internal(spa_t *spa, const char *operation,
862 dmu_tx_t *tx, const char *fmt, ...);
863extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op,
864 dmu_tx_t *tx, const char *fmt, ...);
865extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
866 dmu_tx_t *tx, const char *fmt, ...);
867
868/* error handling */
869struct zbookmark_phys;
870extern void spa_log_error(spa_t *spa, zio_t *zio);
871extern void zfs_ereport_post(const char *cls, spa_t *spa, vdev_t *vd,
872 zio_t *zio, uint64_t stateoroffset, uint64_t length);
873extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
874extern void zfs_post_state_change(spa_t *spa, vdev_t *vd);
875extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
876extern uint64_t spa_get_errlog_size(spa_t *spa);
877extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
878extern void spa_errlog_rotate(spa_t *spa);
879extern void spa_errlog_drain(spa_t *spa);
880extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
881extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
882
883/* vdev cache */
884extern void vdev_cache_stat_init(void);
885extern void vdev_cache_stat_fini(void);
886
887/* Initialization and termination */
888extern void spa_init(int flags);
889extern void spa_fini(void);
890extern void spa_boot_init();
891
892/* properties */
893extern int spa_prop_set(spa_t *spa, nvlist_t *nvp);
894extern int spa_prop_get(spa_t *spa, nvlist_t **nvp);
895extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx);
896extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t);
897
898/* asynchronous event notification */
899extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name);
900
901#ifdef ZFS_DEBUG
902#define dprintf_bp(bp, fmt, ...) do { \
903 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
904 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \
905 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \
906 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \
907 kmem_free(__blkbuf, BP_SPRINTF_LEN); \
908 } \
909_NOTE(CONSTCOND) } while (0)
910#else
911#define dprintf_bp(bp, fmt, ...)
912#endif
913
914extern boolean_t spa_debug_enabled(spa_t *spa);
915#define spa_dbgmsg(spa, ...) \
916{ \
917 if (spa_debug_enabled(spa)) \
918 zfs_dbgmsg(__VA_ARGS__); \
919}
920
921extern int spa_mode_global; /* mode, e.g. FREAD | FWRITE */
922
923#ifdef __cplusplus
924}
925#endif
926
927#endif /* _SYS_SPA_H */
928