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.
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.
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]
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 * Copyright 2017 Joyent, Inc.
29 * Copyright (c) 2017 Datto Inc.
36 #include <sys/zfs_context.h>
37 #include <sys/kstat.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/spa_checksum.h>
50 * Forward references that lots of things need.
52 typedef struct spa spa_t
;
53 typedef struct vdev vdev_t
;
54 typedef struct metaslab metaslab_t
;
55 typedef struct metaslab_group metaslab_group_t
;
56 typedef struct metaslab_class metaslab_class_t
;
57 typedef struct zio zio_t
;
58 typedef struct zilog zilog_t
;
59 typedef struct spa_aux_vdev spa_aux_vdev_t
;
60 typedef struct ddt ddt_t
;
61 typedef struct ddt_entry ddt_entry_t
;
62 typedef struct zbookmark_phys zbookmark_phys_t
;
68 * General-purpose 32-bit and 64-bit bitfield encodings.
70 #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len))
71 #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len))
72 #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low))
73 #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low))
75 #define BF32_GET(x, low, len) BF32_DECODE(x, low, len)
76 #define BF64_GET(x, low, len) BF64_DECODE(x, low, len)
78 #define BF32_SET(x, low, len, val) do { \
79 ASSERT3U(val, <, 1U << (len)); \
80 ASSERT3U(low + len, <=, 32); \
81 (x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \
82 _NOTE(CONSTCOND) } while (0)
84 #define BF64_SET(x, low, len, val) do { \
85 ASSERT3U(val, <, 1ULL << (len)); \
86 ASSERT3U(low + len, <=, 64); \
87 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \
88 _NOTE(CONSTCOND) } while (0)
90 #define BF32_GET_SB(x, low, len, shift, bias) \
91 ((BF32_GET(x, low, len) + (bias)) << (shift))
92 #define BF64_GET_SB(x, low, len, shift, bias) \
93 ((BF64_GET(x, low, len) + (bias)) << (shift))
95 #define BF32_SET_SB(x, low, len, shift, bias, val) do { \
96 ASSERT(IS_P2ALIGNED(val, 1U << shift)); \
97 ASSERT3S((val) >> (shift), >=, bias); \
98 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \
99 _NOTE(CONSTCOND) } while (0)
100 #define BF64_SET_SB(x, low, len, shift, bias, val) do { \
101 ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \
102 ASSERT3S((val) >> (shift), >=, bias); \
103 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \
104 _NOTE(CONSTCOND) } while (0)
107 * We currently support block sizes from 512 bytes to 16MB.
108 * The benefits of larger blocks, and thus larger IO, need to be weighed
109 * against the cost of COWing a giant block to modify one byte, and the
110 * large latency of reading or writing a large block.
112 * Note that although blocks up to 16MB are supported, the recordsize
113 * property can not be set larger than zfs_max_recordsize (default 1MB).
114 * See the comment near zfs_max_recordsize in dsl_dataset.c for details.
116 * Note that although the LSIZE field of the blkptr_t can store sizes up
117 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to
118 * 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
120 #define SPA_MINBLOCKSHIFT 9
121 #define SPA_OLD_MAXBLOCKSHIFT 17
122 #define SPA_MAXBLOCKSHIFT 24
123 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT)
124 #define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT)
125 #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT)
128 * Alignment Shift (ashift) is an immutable, internal top-level vdev property
129 * which can only be set at vdev creation time. Physical writes are always done
130 * according to it, which makes 2^ashift the smallest possible IO on a vdev.
132 * We currently allow values ranging from 512 bytes (2^9 = 512) to 64 KiB
136 #define ASHIFT_MAX 16
139 * Size of block to hold the configuration data (a packed nvlist)
141 #define SPA_CONFIG_BLOCKSIZE (1ULL << 14)
144 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
145 * The ASIZE encoding should be at least 64 times larger (6 more bits)
146 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
147 * overhead, three DVAs per bp, plus one more bit in case we do anything
148 * else that expands the ASIZE.
150 #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */
151 #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */
152 #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */
154 #define SPA_COMPRESSBITS 7
157 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
158 * The members of the dva_t should be considered opaque outside the SPA.
161 uint64_t dva_word
[2];
166 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept
167 * secret and is suitable for use in MAC algorithms as the key.
169 typedef struct zio_cksum_salt
{
170 uint8_t zcs_bytes
[32];
174 * Each block is described by its DVAs, time of birth, checksum, etc.
175 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
177 * 64 56 48 40 32 24 16 8 0
178 * +-------+-------+-------+-------+-------+-------+-------+-------+
179 * 0 | vdev1 | GRID | ASIZE |
180 * +-------+-------+-------+-------+-------+-------+-------+-------+
182 * +-------+-------+-------+-------+-------+-------+-------+-------+
183 * 2 | vdev2 | GRID | ASIZE |
184 * +-------+-------+-------+-------+-------+-------+-------+-------+
186 * +-------+-------+-------+-------+-------+-------+-------+-------+
187 * 4 | vdev3 | GRID | ASIZE |
188 * +-------+-------+-------+-------+-------+-------+-------+-------+
190 * +-------+-------+-------+-------+-------+-------+-------+-------+
191 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE |
192 * +-------+-------+-------+-------+-------+-------+-------+-------+
194 * +-------+-------+-------+-------+-------+-------+-------+-------+
196 * +-------+-------+-------+-------+-------+-------+-------+-------+
197 * 9 | physical birth txg |
198 * +-------+-------+-------+-------+-------+-------+-------+-------+
199 * a | logical birth txg |
200 * +-------+-------+-------+-------+-------+-------+-------+-------+
202 * +-------+-------+-------+-------+-------+-------+-------+-------+
204 * +-------+-------+-------+-------+-------+-------+-------+-------+
206 * +-------+-------+-------+-------+-------+-------+-------+-------+
208 * +-------+-------+-------+-------+-------+-------+-------+-------+
210 * +-------+-------+-------+-------+-------+-------+-------+-------+
214 * vdev virtual device ID
215 * offset offset into virtual device
217 * PSIZE physical size (after compression)
218 * ASIZE allocated size (including RAID-Z parity and gang block headers)
219 * GRID RAID-Z layout information (reserved for future use)
220 * cksum checksum function
221 * comp compression function
222 * G gang block indicator
223 * B byteorder (endianness)
225 * X encryption (on version 30, which is not supported)
226 * E blkptr_t contains embedded data (see below)
227 * lvl level of indirection
228 * type DMU object type
229 * phys birth txg of block allocation; zero if same as logical birth txg
230 * log. birth transaction group in which the block was logically born
231 * fill count number of non-zero blocks under this bp
232 * checksum[4] 256-bit checksum of the data this bp describes
236 * "Embedded" blkptr_t's don't actually point to a block, instead they
237 * have a data payload embedded in the blkptr_t itself. See the comment
238 * in blkptr.c for more details.
240 * The blkptr_t is laid out as follows:
242 * 64 56 48 40 32 24 16 8 0
243 * +-------+-------+-------+-------+-------+-------+-------+-------+
250 * +-------+-------+-------+-------+-------+-------+-------+-------+
251 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE |
252 * +-------+-------+-------+-------+-------+-------+-------+-------+
256 * +-------+-------+-------+-------+-------+-------+-------+-------+
257 * a | logical birth txg |
258 * +-------+-------+-------+-------+-------+-------+-------+-------+
264 * +-------+-------+-------+-------+-------+-------+-------+-------+
268 * payload contains the embedded data
269 * B (byteorder) byteorder (endianness)
270 * D (dedup) padding (set to zero)
271 * X encryption (set to zero; see above)
272 * E (embedded) set to one
273 * lvl indirection level
274 * type DMU object type
275 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*)
276 * comp compression function of payload
277 * PSIZE size of payload after compression, in bytes
278 * LSIZE logical size of payload, in bytes
279 * note that 25 bits is enough to store the largest
280 * "normal" BP's LSIZE (2^16 * 2^9) in bytes
281 * log. birth transaction group in which the block was logically born
283 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
284 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
285 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
286 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
287 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must
288 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before
289 * other macros, as they assert that they are only used on BP's of the correct
293 #define BPE_GET_ETYPE(bp) \
294 (ASSERT(BP_IS_EMBEDDED(bp)), \
295 BF64_GET((bp)->blk_prop, 40, 8))
296 #define BPE_SET_ETYPE(bp, t) do { \
297 ASSERT(BP_IS_EMBEDDED(bp)); \
298 BF64_SET((bp)->blk_prop, 40, 8, t); \
299 _NOTE(CONSTCOND) } while (0)
301 #define BPE_GET_LSIZE(bp) \
302 (ASSERT(BP_IS_EMBEDDED(bp)), \
303 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
304 #define BPE_SET_LSIZE(bp, x) do { \
305 ASSERT(BP_IS_EMBEDDED(bp)); \
306 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
307 _NOTE(CONSTCOND) } while (0)
309 #define BPE_GET_PSIZE(bp) \
310 (ASSERT(BP_IS_EMBEDDED(bp)), \
311 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
312 #define BPE_SET_PSIZE(bp, x) do { \
313 ASSERT(BP_IS_EMBEDDED(bp)); \
314 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
315 _NOTE(CONSTCOND) } while (0)
317 typedef enum bp_embedded_type
{
318 BP_EMBEDDED_TYPE_DATA
,
319 BP_EMBEDDED_TYPE_RESERVED
, /* Reserved for an unintegrated feature. */
320 NUM_BP_EMBEDDED_TYPES
= BP_EMBEDDED_TYPE_RESERVED
321 } bp_embedded_type_t
;
323 #define BPE_NUM_WORDS 14
324 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
325 #define BPE_IS_PAYLOADWORD(bp, wp) \
326 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
328 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */
329 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */
332 * A block is a hole when it has either 1) never been written to, or
333 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
334 * without physically allocating disk space. Holes are represented in the
335 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is
336 * done through the BP_IS_HOLE macro. For holes, the logical size, level,
337 * DMU object type, and birth times are all also stored for holes that
338 * were written to at some point (i.e. were punched after having been filled).
340 typedef struct blkptr
{
341 dva_t blk_dva
[SPA_DVAS_PER_BP
]; /* Data Virtual Addresses */
342 uint64_t blk_prop
; /* size, compression, type, etc */
343 uint64_t blk_pad
[2]; /* Extra space for the future */
344 uint64_t blk_phys_birth
; /* txg when block was allocated */
345 uint64_t blk_birth
; /* transaction group at birth */
346 uint64_t blk_fill
; /* fill count */
347 zio_cksum_t blk_cksum
; /* 256-bit checksum */
351 * Macros to get and set fields in a bp or DVA.
353 #define DVA_GET_ASIZE(dva) \
354 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
355 #define DVA_SET_ASIZE(dva, x) \
356 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
357 SPA_MINBLOCKSHIFT, 0, x)
359 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8)
360 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x)
362 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32)
363 #define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x)
365 #define DVA_GET_OFFSET(dva) \
366 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
367 #define DVA_SET_OFFSET(dva, x) \
368 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
370 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1)
371 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x)
373 #define BP_GET_LSIZE(bp) \
374 (BP_IS_EMBEDDED(bp) ? \
375 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
376 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
377 #define BP_SET_LSIZE(bp, x) do { \
378 ASSERT(!BP_IS_EMBEDDED(bp)); \
379 BF64_SET_SB((bp)->blk_prop, \
380 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
381 _NOTE(CONSTCOND) } while (0)
383 #define BP_GET_PSIZE(bp) \
384 (BP_IS_EMBEDDED(bp) ? 0 : \
385 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1))
386 #define BP_SET_PSIZE(bp, x) do { \
387 ASSERT(!BP_IS_EMBEDDED(bp)); \
388 BF64_SET_SB((bp)->blk_prop, \
389 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
390 _NOTE(CONSTCOND) } while (0)
392 #define BP_GET_COMPRESS(bp) \
393 BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS)
394 #define BP_SET_COMPRESS(bp, x) \
395 BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x)
397 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1)
398 #define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x)
400 #define BP_GET_CHECKSUM(bp) \
401 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \
402 BF64_GET((bp)->blk_prop, 40, 8))
403 #define BP_SET_CHECKSUM(bp, x) do { \
404 ASSERT(!BP_IS_EMBEDDED(bp)); \
405 BF64_SET((bp)->blk_prop, 40, 8, x); \
406 _NOTE(CONSTCOND) } while (0)
408 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8)
409 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x)
411 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5)
412 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x)
414 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1)
415 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x)
417 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1)
418 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x)
420 #define BP_PHYSICAL_BIRTH(bp) \
421 (BP_IS_EMBEDDED(bp) ? 0 : \
422 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
424 #define BP_SET_BIRTH(bp, logical, physical) \
426 ASSERT(!BP_IS_EMBEDDED(bp)); \
427 (bp)->blk_birth = (logical); \
428 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
431 #define BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill)
433 #define BP_IS_METADATA(bp) \
434 (BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))
436 #define BP_GET_ASIZE(bp) \
437 (BP_IS_EMBEDDED(bp) ? 0 : \
438 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
439 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
440 DVA_GET_ASIZE(&(bp)->blk_dva[2]))
442 #define BP_GET_UCSIZE(bp) \
443 (BP_IS_METADATA(bp) ? BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
445 #define BP_GET_NDVAS(bp) \
446 (BP_IS_EMBEDDED(bp) ? 0 : \
447 !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
448 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
449 !!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
451 #define BP_COUNT_GANG(bp) \
452 (BP_IS_EMBEDDED(bp) ? 0 : \
453 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \
454 DVA_GET_GANG(&(bp)->blk_dva[1]) + \
455 DVA_GET_GANG(&(bp)->blk_dva[2])))
457 #define DVA_EQUAL(dva1, dva2) \
458 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
459 (dva1)->dva_word[0] == (dva2)->dva_word[0])
461 #define BP_EQUAL(bp1, bp2) \
462 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \
463 (bp1)->blk_birth == (bp2)->blk_birth && \
464 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \
465 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \
466 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
469 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0)
471 #define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0])
472 #define BP_IS_GANG(bp) \
473 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp)))
474 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \
475 (dva)->dva_word[1] == 0ULL)
476 #define BP_IS_HOLE(bp) \
477 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp)))
479 /* BP_IS_RAIDZ(bp) assumes no block compression */
480 #define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
483 #define BP_ZERO(bp) \
485 (bp)->blk_dva[0].dva_word[0] = 0; \
486 (bp)->blk_dva[0].dva_word[1] = 0; \
487 (bp)->blk_dva[1].dva_word[0] = 0; \
488 (bp)->blk_dva[1].dva_word[1] = 0; \
489 (bp)->blk_dva[2].dva_word[0] = 0; \
490 (bp)->blk_dva[2].dva_word[1] = 0; \
491 (bp)->blk_prop = 0; \
492 (bp)->blk_pad[0] = 0; \
493 (bp)->blk_pad[1] = 0; \
494 (bp)->blk_phys_birth = 0; \
495 (bp)->blk_birth = 0; \
496 (bp)->blk_fill = 0; \
497 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \
501 #define ZFS_HOST_BYTEORDER (0ULL)
503 #define ZFS_HOST_BYTEORDER (1ULL)
506 #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
508 #define BP_SPRINTF_LEN 320
511 * This macro allows code sharing between zfs, libzpool, and mdb.
512 * 'func' is either snprintf() or mdb_snprintf().
513 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
515 #define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
517 static const char *copyname[] = \
518 { "zero", "single", "double", "triple" }; \
524 len += func(buf + len, size - len, "<NULL>"); \
525 } else if (BP_IS_HOLE(bp)) { \
526 len += func(buf + len, size - len, \
528 "size=%llxL birth=%lluL", \
529 (u_longlong_t)BP_GET_LEVEL(bp), \
531 (u_longlong_t)BP_GET_LSIZE(bp), \
532 (u_longlong_t)bp->blk_birth); \
533 } else if (BP_IS_EMBEDDED(bp)) { \
534 len = func(buf + len, size - len, \
535 "EMBEDDED [L%llu %s] et=%u %s " \
536 "size=%llxL/%llxP birth=%lluL", \
537 (u_longlong_t)BP_GET_LEVEL(bp), \
539 (int)BPE_GET_ETYPE(bp), \
541 (u_longlong_t)BPE_GET_LSIZE(bp), \
542 (u_longlong_t)BPE_GET_PSIZE(bp), \
543 (u_longlong_t)bp->blk_birth); \
545 for (d = 0; d < BP_GET_NDVAS(bp); d++) { \
546 const dva_t *dva = &bp->blk_dva[d]; \
547 if (DVA_IS_VALID(dva)) \
549 len += func(buf + len, size - len, \
550 "DVA[%d]=<%llu:%llx:%llx>%c", d, \
551 (u_longlong_t)DVA_GET_VDEV(dva), \
552 (u_longlong_t)DVA_GET_OFFSET(dva), \
553 (u_longlong_t)DVA_GET_ASIZE(dva), \
556 if (BP_IS_GANG(bp) && \
557 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \
558 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \
560 len += func(buf + len, size - len, \
561 "[L%llu %s] %s %s %s %s %s %s%c" \
562 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \
563 "cksum=%llx:%llx:%llx:%llx", \
564 (u_longlong_t)BP_GET_LEVEL(bp), \
568 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \
569 BP_IS_GANG(bp) ? "gang" : "contiguous", \
570 BP_GET_DEDUP(bp) ? "dedup" : "unique", \
573 (u_longlong_t)BP_GET_LSIZE(bp), \
574 (u_longlong_t)BP_GET_PSIZE(bp), \
575 (u_longlong_t)bp->blk_birth, \
576 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \
577 (u_longlong_t)BP_GET_FILL(bp), \
579 (u_longlong_t)bp->blk_cksum.zc_word[0], \
580 (u_longlong_t)bp->blk_cksum.zc_word[1], \
581 (u_longlong_t)bp->blk_cksum.zc_word[2], \
582 (u_longlong_t)bp->blk_cksum.zc_word[3]); \
584 ASSERT(len < size); \
587 #define BP_GET_BUFC_TYPE(bp) \
588 (BP_IS_METADATA(bp) ? ARC_BUFC_METADATA : ARC_BUFC_DATA)
590 typedef enum spa_import_type
{
595 /* state manipulation functions */
596 extern int spa_open(const char *pool
, spa_t
**, void *tag
);
597 extern int spa_open_rewind(const char *pool
, spa_t
**, void *tag
,
598 nvlist_t
*policy
, nvlist_t
**config
);
599 extern int spa_get_stats(const char *pool
, nvlist_t
**config
, char *altroot
,
601 extern int spa_create(const char *pool
, nvlist_t
*config
, nvlist_t
*props
,
603 extern int spa_import(char *pool
, nvlist_t
*config
, nvlist_t
*props
,
605 extern nvlist_t
*spa_tryimport(nvlist_t
*tryconfig
);
606 extern int spa_destroy(char *pool
);
607 extern int spa_export(char *pool
, nvlist_t
**oldconfig
, boolean_t force
,
608 boolean_t hardforce
);
609 extern int spa_reset(char *pool
);
610 extern void spa_async_request(spa_t
*spa
, int flag
);
611 extern void spa_async_unrequest(spa_t
*spa
, int flag
);
612 extern void spa_async_suspend(spa_t
*spa
);
613 extern void spa_async_resume(spa_t
*spa
);
614 extern spa_t
*spa_inject_addref(char *pool
);
615 extern void spa_inject_delref(spa_t
*spa
);
616 extern void spa_scan_stat_init(spa_t
*spa
);
617 extern int spa_scan_get_stats(spa_t
*spa
, pool_scan_stat_t
*ps
);
619 #define SPA_ASYNC_CONFIG_UPDATE 0x01
620 #define SPA_ASYNC_REMOVE 0x02
621 #define SPA_ASYNC_PROBE 0x04
622 #define SPA_ASYNC_RESILVER_DONE 0x08
623 #define SPA_ASYNC_RESILVER 0x10
624 #define SPA_ASYNC_AUTOEXPAND 0x20
625 #define SPA_ASYNC_REMOVE_DONE 0x40
626 #define SPA_ASYNC_REMOVE_STOP 0x80
629 * Controls the behavior of spa_vdev_remove().
631 #define SPA_REMOVE_UNSPARE 0x01
632 #define SPA_REMOVE_DONE 0x02
634 /* device manipulation */
635 extern int spa_vdev_add(spa_t
*spa
, nvlist_t
*nvroot
);
636 extern int spa_vdev_attach(spa_t
*spa
, uint64_t guid
, nvlist_t
*nvroot
,
638 extern int spa_vdev_detach(spa_t
*spa
, uint64_t guid
, uint64_t pguid
,
640 extern int spa_vdev_remove(spa_t
*spa
, uint64_t guid
, boolean_t unspare
);
641 extern boolean_t
spa_vdev_remove_active(spa_t
*spa
);
642 extern int spa_vdev_setpath(spa_t
*spa
, uint64_t guid
, const char *newpath
);
643 extern int spa_vdev_setfru(spa_t
*spa
, uint64_t guid
, const char *newfru
);
644 extern int spa_vdev_split_mirror(spa_t
*spa
, char *newname
, nvlist_t
*config
,
645 nvlist_t
*props
, boolean_t exp
);
647 /* spare state (which is global across all pools) */
648 extern void spa_spare_add(vdev_t
*vd
);
649 extern void spa_spare_remove(vdev_t
*vd
);
650 extern boolean_t
spa_spare_exists(uint64_t guid
, uint64_t *pool
, int *refcnt
);
651 extern void spa_spare_activate(vdev_t
*vd
);
653 /* L2ARC state (which is global across all pools) */
654 extern void spa_l2cache_add(vdev_t
*vd
);
655 extern void spa_l2cache_remove(vdev_t
*vd
);
656 extern boolean_t
spa_l2cache_exists(uint64_t guid
, uint64_t *pool
);
657 extern void spa_l2cache_activate(vdev_t
*vd
);
658 extern void spa_l2cache_drop(spa_t
*spa
);
661 extern int spa_scan(spa_t
*spa
, pool_scan_func_t func
);
662 extern int spa_scan_stop(spa_t
*spa
);
663 extern int spa_scrub_pause_resume(spa_t
*spa
, pool_scrub_cmd_t flag
);
666 extern void spa_sync(spa_t
*spa
, uint64_t txg
); /* only for DMU use */
667 extern void spa_sync_allpools(void);
669 extern int zfs_sync_pass_deferred_free
;
671 /* spa namespace global mutex */
672 extern kmutex_t spa_namespace_lock
;
675 * SPA configuration functions in spa_config.c
678 #define SPA_CONFIG_UPDATE_POOL 0
679 #define SPA_CONFIG_UPDATE_VDEVS 1
681 extern void spa_config_sync(spa_t
*, boolean_t
, boolean_t
);
682 extern void spa_config_load(void);
683 extern nvlist_t
*spa_all_configs(uint64_t *);
684 extern void spa_config_set(spa_t
*spa
, nvlist_t
*config
);
685 extern nvlist_t
*spa_config_generate(spa_t
*spa
, vdev_t
*vd
, uint64_t txg
,
687 extern void spa_config_update(spa_t
*spa
, int what
);
690 * Miscellaneous SPA routines in spa_misc.c
693 /* Namespace manipulation */
694 extern spa_t
*spa_lookup(const char *name
);
695 extern spa_t
*spa_add(const char *name
, nvlist_t
*config
, const char *altroot
);
696 extern void spa_remove(spa_t
*spa
);
697 extern spa_t
*spa_next(spa_t
*prev
);
699 /* Refcount functions */
700 extern void spa_open_ref(spa_t
*spa
, void *tag
);
701 extern void spa_close(spa_t
*spa
, void *tag
);
702 extern void spa_async_close(spa_t
*spa
, void *tag
);
703 extern boolean_t
spa_refcount_zero(spa_t
*spa
);
705 #define SCL_NONE 0x00
706 #define SCL_CONFIG 0x01
707 #define SCL_STATE 0x02
708 #define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */
709 #define SCL_ALLOC 0x08
711 #define SCL_FREE 0x20
712 #define SCL_VDEV 0x40
714 #define SCL_ALL ((1 << SCL_LOCKS) - 1)
715 #define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO)
717 /* Historical pool statistics */
718 typedef struct spa_stats_history
{
725 } spa_stats_history_t
;
727 typedef struct spa_stats
{
728 spa_stats_history_t read_history
;
729 spa_stats_history_t txg_history
;
730 spa_stats_history_t tx_assign_histogram
;
731 spa_stats_history_t io_history
;
732 spa_stats_history_t mmp_history
;
735 typedef enum txg_state
{
738 TXG_STATE_QUIESCED
= 2,
739 TXG_STATE_WAIT_FOR_SYNC
= 3,
740 TXG_STATE_SYNCED
= 4,
741 TXG_STATE_COMMITTED
= 5,
744 typedef struct txg_stat
{
751 extern void spa_stats_init(spa_t
*spa
);
752 extern void spa_stats_destroy(spa_t
*spa
);
753 extern void spa_read_history_add(spa_t
*spa
, const zbookmark_phys_t
*zb
,
755 extern void spa_txg_history_add(spa_t
*spa
, uint64_t txg
, hrtime_t birth_time
);
756 extern int spa_txg_history_set(spa_t
*spa
, uint64_t txg
,
757 txg_state_t completed_state
, hrtime_t completed_time
);
758 extern txg_stat_t
*spa_txg_history_init_io(spa_t
*, uint64_t,
760 extern void spa_txg_history_fini_io(spa_t
*, txg_stat_t
*);
761 extern void spa_tx_assign_add_nsecs(spa_t
*spa
, uint64_t nsecs
);
762 extern void spa_mmp_history_add(uint64_t txg
, uint64_t timestamp
,
763 uint64_t mmp_delay
, vdev_t
*vd
, int label
);
765 /* Pool configuration locks */
766 extern int spa_config_tryenter(spa_t
*spa
, int locks
, void *tag
, krw_t rw
);
767 extern void spa_config_enter(spa_t
*spa
, int locks
, void *tag
, krw_t rw
);
768 extern void spa_config_exit(spa_t
*spa
, int locks
, void *tag
);
769 extern int spa_config_held(spa_t
*spa
, int locks
, krw_t rw
);
771 /* Pool vdev add/remove lock */
772 extern uint64_t spa_vdev_enter(spa_t
*spa
);
773 extern uint64_t spa_vdev_config_enter(spa_t
*spa
);
774 extern void spa_vdev_config_exit(spa_t
*spa
, vdev_t
*vd
, uint64_t txg
,
775 int error
, char *tag
);
776 extern int spa_vdev_exit(spa_t
*spa
, vdev_t
*vd
, uint64_t txg
, int error
);
778 /* Pool vdev state change lock */
779 extern void spa_vdev_state_enter(spa_t
*spa
, int oplock
);
780 extern int spa_vdev_state_exit(spa_t
*spa
, vdev_t
*vd
, int error
);
783 typedef enum spa_log_state
{
784 SPA_LOG_UNKNOWN
= 0, /* unknown log state */
785 SPA_LOG_MISSING
, /* missing log(s) */
786 SPA_LOG_CLEAR
, /* clear the log(s) */
787 SPA_LOG_GOOD
, /* log(s) are good */
790 extern spa_log_state_t
spa_get_log_state(spa_t
*spa
);
791 extern void spa_set_log_state(spa_t
*spa
, spa_log_state_t state
);
792 extern int spa_offline_log(spa_t
*spa
);
794 /* Log claim callback */
795 extern void spa_claim_notify(zio_t
*zio
);
796 extern void spa_deadman(void *);
798 /* Accessor functions */
799 extern boolean_t
spa_shutting_down(spa_t
*spa
);
800 extern struct dsl_pool
*spa_get_dsl(spa_t
*spa
);
801 extern boolean_t
spa_is_initializing(spa_t
*spa
);
802 extern blkptr_t
*spa_get_rootblkptr(spa_t
*spa
);
803 extern void spa_set_rootblkptr(spa_t
*spa
, const blkptr_t
*bp
);
804 extern void spa_altroot(spa_t
*, char *, size_t);
805 extern int spa_sync_pass(spa_t
*spa
);
806 extern char *spa_name(spa_t
*spa
);
807 extern uint64_t spa_guid(spa_t
*spa
);
808 extern uint64_t spa_load_guid(spa_t
*spa
);
809 extern uint64_t spa_last_synced_txg(spa_t
*spa
);
810 extern uint64_t spa_first_txg(spa_t
*spa
);
811 extern uint64_t spa_syncing_txg(spa_t
*spa
);
812 extern uint64_t spa_final_dirty_txg(spa_t
*spa
);
813 extern uint64_t spa_version(spa_t
*spa
);
814 extern pool_state_t
spa_state(spa_t
*spa
);
815 extern spa_load_state_t
spa_load_state(spa_t
*spa
);
816 extern uint64_t spa_freeze_txg(spa_t
*spa
);
817 extern uint64_t spa_get_worst_case_asize(spa_t
*spa
, uint64_t lsize
);
818 extern uint64_t spa_get_dspace(spa_t
*spa
);
819 extern uint64_t spa_get_slop_space(spa_t
*spa
);
820 extern void spa_update_dspace(spa_t
*spa
);
821 extern uint64_t spa_version(spa_t
*spa
);
822 extern boolean_t
spa_deflate(spa_t
*spa
);
823 extern metaslab_class_t
*spa_normal_class(spa_t
*spa
);
824 extern metaslab_class_t
*spa_log_class(spa_t
*spa
);
825 extern void spa_evicting_os_register(spa_t
*, objset_t
*os
);
826 extern void spa_evicting_os_deregister(spa_t
*, objset_t
*os
);
827 extern void spa_evicting_os_wait(spa_t
*spa
);
828 extern int spa_max_replication(spa_t
*spa
);
829 extern int spa_prev_software_version(spa_t
*spa
);
830 extern uint8_t spa_get_failmode(spa_t
*spa
);
831 extern boolean_t
spa_suspended(spa_t
*spa
);
832 extern uint64_t spa_bootfs(spa_t
*spa
);
833 extern uint64_t spa_delegation(spa_t
*spa
);
834 extern objset_t
*spa_meta_objset(spa_t
*spa
);
835 extern uint64_t spa_deadman_synctime(spa_t
*spa
);
837 /* Miscellaneous support routines */
838 extern void spa_activate_mos_feature(spa_t
*spa
, const char *feature
,
840 extern void spa_deactivate_mos_feature(spa_t
*spa
, const char *feature
);
841 extern int spa_rename(const char *oldname
, const char *newname
);
842 extern spa_t
*spa_by_guid(uint64_t pool_guid
, uint64_t device_guid
);
843 extern boolean_t
spa_guid_exists(uint64_t pool_guid
, uint64_t device_guid
);
844 extern char *spa_strdup(const char *);
845 extern void spa_strfree(char *);
846 extern uint64_t spa_get_random(uint64_t range
);
847 extern uint64_t spa_generate_guid(spa_t
*spa
);
848 extern void snprintf_blkptr(char *buf
, size_t buflen
, const blkptr_t
*bp
);
849 extern void spa_freeze(spa_t
*spa
);
850 extern int spa_change_guid(spa_t
*spa
);
851 extern void spa_upgrade(spa_t
*spa
, uint64_t version
);
852 extern void spa_evict_all(void);
853 extern vdev_t
*spa_lookup_by_guid(spa_t
*spa
, uint64_t guid
,
855 extern boolean_t
spa_has_spare(spa_t
*, uint64_t guid
);
856 extern uint64_t dva_get_dsize_sync(spa_t
*spa
, const dva_t
*dva
);
857 extern uint64_t bp_get_dsize_sync(spa_t
*spa
, const blkptr_t
*bp
);
858 extern uint64_t bp_get_dsize(spa_t
*spa
, const blkptr_t
*bp
);
859 extern boolean_t
spa_has_slogs(spa_t
*spa
);
860 extern boolean_t
spa_is_root(spa_t
*spa
);
861 extern boolean_t
spa_writeable(spa_t
*spa
);
862 extern boolean_t
spa_has_pending_synctask(spa_t
*spa
);
863 extern int spa_maxblocksize(spa_t
*spa
);
864 extern int spa_maxdnodesize(spa_t
*spa
);
865 extern void zfs_blkptr_verify(spa_t
*spa
, const blkptr_t
*bp
);
866 extern boolean_t
spa_multihost(spa_t
*spa
);
867 extern unsigned long spa_get_hostid(void);
869 extern int spa_mode(spa_t
*spa
);
870 extern uint64_t zfs_strtonum(const char *str
, char **nptr
);
872 extern char *spa_his_ievent_table
[];
874 extern void spa_history_create_obj(spa_t
*spa
, dmu_tx_t
*tx
);
875 extern int spa_history_get(spa_t
*spa
, uint64_t *offset
, uint64_t *len_read
,
877 extern int spa_history_log(spa_t
*spa
, const char *his_buf
);
878 extern int spa_history_log_nvl(spa_t
*spa
, nvlist_t
*nvl
);
879 extern void spa_history_log_version(spa_t
*spa
, const char *operation
,
881 extern void spa_history_log_internal(spa_t
*spa
, const char *operation
,
882 dmu_tx_t
*tx
, const char *fmt
, ...);
883 extern void spa_history_log_internal_ds(struct dsl_dataset
*ds
, const char *op
,
884 dmu_tx_t
*tx
, const char *fmt
, ...);
885 extern void spa_history_log_internal_dd(dsl_dir_t
*dd
, const char *operation
,
886 dmu_tx_t
*tx
, const char *fmt
, ...);
889 struct zbookmark_phys
;
890 extern void spa_log_error(spa_t
*spa
, zio_t
*zio
);
891 extern void zfs_ereport_post(const char *class, spa_t
*spa
, vdev_t
*vd
,
892 zio_t
*zio
, uint64_t stateoroffset
, uint64_t length
);
893 extern nvlist_t
*zfs_event_create(spa_t
*spa
, vdev_t
*vd
, const char *type
,
894 const char *name
, nvlist_t
*aux
);
895 extern void zfs_post_remove(spa_t
*spa
, vdev_t
*vd
);
896 extern void zfs_post_state_change(spa_t
*spa
, vdev_t
*vd
, uint64_t laststate
);
897 extern void zfs_post_autoreplace(spa_t
*spa
, vdev_t
*vd
);
898 extern uint64_t spa_get_errlog_size(spa_t
*spa
);
899 extern int spa_get_errlog(spa_t
*spa
, void *uaddr
, size_t *count
);
900 extern void spa_errlog_rotate(spa_t
*spa
);
901 extern void spa_errlog_drain(spa_t
*spa
);
902 extern void spa_errlog_sync(spa_t
*spa
, uint64_t txg
);
903 extern void spa_get_errlists(spa_t
*spa
, avl_tree_t
*last
, avl_tree_t
*scrub
);
906 extern void vdev_cache_stat_init(void);
907 extern void vdev_cache_stat_fini(void);
909 /* Initialization and termination */
910 extern void spa_init(int flags
);
911 extern void spa_fini(void);
912 extern void spa_boot_init(void);
915 extern int spa_prop_set(spa_t
*spa
, nvlist_t
*nvp
);
916 extern int spa_prop_get(spa_t
*spa
, nvlist_t
**nvp
);
917 extern void spa_prop_clear_bootfs(spa_t
*spa
, uint64_t obj
, dmu_tx_t
*tx
);
918 extern void spa_configfile_set(spa_t
*, nvlist_t
*, boolean_t
);
920 /* asynchronous event notification */
921 extern void spa_event_notify(spa_t
*spa
, vdev_t
*vdev
, nvlist_t
*hist_nvl
,
925 #define dprintf_bp(bp, fmt, ...) do { \
926 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
927 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \
928 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \
929 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \
930 kmem_free(__blkbuf, BP_SPRINTF_LEN); \
932 _NOTE(CONSTCOND) } while (0)
934 #define dprintf_bp(bp, fmt, ...)
937 extern boolean_t
spa_debug_enabled(spa_t
*spa
);
938 #define spa_dbgmsg(spa, ...) \
940 if (spa_debug_enabled(spa)) \
941 zfs_dbgmsg(__VA_ARGS__); \
944 extern int spa_mode_global
; /* mode, e.g. FREAD | FWRITE */
950 #endif /* _SYS_SPA_H */