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, 2019 by Delphix. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
26 * Copyright 2013 Saso Kiselkov. All rights reserved.
27 * Copyright (c) 2014 Integros [integros.com]
28 * Copyright 2017 Joyent, Inc.
29 * Copyright (c) 2017 Datto Inc.
30 * Copyright (c) 2017, Intel Corporation.
37 #include <sys/zfs_context.h>
38 #include <sys/kstat.h>
39 #include <sys/nvpair.h>
40 #include <sys/sysmacros.h>
41 #include <sys/types.h>
42 #include <sys/fs/zfs.h>
43 #include <sys/spa_checksum.h>
45 #include <sys/space_map.h>
52 * Forward references that lots of things need.
54 typedef struct spa spa_t
;
55 typedef struct vdev vdev_t
;
56 typedef struct metaslab metaslab_t
;
57 typedef struct metaslab_group metaslab_group_t
;
58 typedef struct metaslab_class metaslab_class_t
;
59 typedef struct zio zio_t
;
60 typedef struct zilog zilog_t
;
61 typedef struct spa_aux_vdev spa_aux_vdev_t
;
62 typedef struct ddt ddt_t
;
63 typedef struct ddt_entry ddt_entry_t
;
64 typedef struct zbookmark_phys zbookmark_phys_t
;
70 struct dsl_crypto_params
;
73 * General-purpose 32-bit and 64-bit bitfield encodings.
75 #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len))
76 #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len))
77 #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low))
78 #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low))
80 #define BF32_GET(x, low, len) BF32_DECODE(x, low, len)
81 #define BF64_GET(x, low, len) BF64_DECODE(x, low, len)
83 #define BF32_SET(x, low, len, val) do { \
84 ASSERT3U(val, <, 1U << (len)); \
85 ASSERT3U(low + len, <=, 32); \
86 (x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \
87 _NOTE(CONSTCOND) } while (0)
89 #define BF64_SET(x, low, len, val) do { \
90 ASSERT3U(val, <, 1ULL << (len)); \
91 ASSERT3U(low + len, <=, 64); \
92 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \
93 _NOTE(CONSTCOND) } while (0)
95 #define BF32_GET_SB(x, low, len, shift, bias) \
96 ((BF32_GET(x, low, len) + (bias)) << (shift))
97 #define BF64_GET_SB(x, low, len, shift, bias) \
98 ((BF64_GET(x, low, len) + (bias)) << (shift))
101 * We use ASSERT3U instead of ASSERT in these macros to prevent a lint error in
102 * the case where val is a constant. We can't fix ASSERT because it's used as
103 * an expression in several places in the kernel; as a result, changing it to
104 * the do{} while() syntax to allow us to _NOTE the CONSTCOND is not an option.
106 #define BF32_SET_SB(x, low, len, shift, bias, val) do { \
107 ASSERT3U(IS_P2ALIGNED(val, 1U << shift), !=, B_FALSE); \
108 ASSERT3S((val) >> (shift), >=, bias); \
109 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \
110 _NOTE(CONSTCOND) } while (0)
111 #define BF64_SET_SB(x, low, len, shift, bias, val) do { \
112 ASSERT3U(IS_P2ALIGNED(val, 1ULL << shift), !=, B_FALSE); \
113 ASSERT3S((val) >> (shift), >=, bias); \
114 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \
115 _NOTE(CONSTCOND) } while (0)
118 * We currently support block sizes from 512 bytes to 16MB.
119 * The benefits of larger blocks, and thus larger IO, need to be weighed
120 * against the cost of COWing a giant block to modify one byte, and the
121 * large latency of reading or writing a large block.
123 * Note that although blocks up to 16MB are supported, the recordsize
124 * property can not be set larger than zfs_max_recordsize (default 1MB).
125 * See the comment near zfs_max_recordsize in dsl_dataset.c for details.
127 * Note that although the LSIZE field of the blkptr_t can store sizes up
128 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to
129 * 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
131 #define SPA_MINBLOCKSHIFT 9
132 #define SPA_OLD_MAXBLOCKSHIFT 17
133 #define SPA_MAXBLOCKSHIFT 24
134 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT)
135 #define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT)
136 #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT)
139 * Alignment Shift (ashift) is an immutable, internal top-level vdev property
140 * which can only be set at vdev creation time. Physical writes are always done
141 * according to it, which makes 2^ashift the smallest possible IO on a vdev.
143 * We currently allow values ranging from 512 bytes (2^9 = 512) to 64 KiB
147 #define ASHIFT_MAX 16
150 * Size of block to hold the configuration data (a packed nvlist)
152 #define SPA_CONFIG_BLOCKSIZE (1ULL << 14)
155 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
156 * The ASIZE encoding should be at least 64 times larger (6 more bits)
157 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
158 * overhead, three DVAs per bp, plus one more bit in case we do anything
159 * else that expands the ASIZE.
161 #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */
162 #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */
163 #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */
165 #define SPA_COMPRESSBITS 7
166 #define SPA_VDEVBITS 24
169 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
170 * The members of the dva_t should be considered opaque outside the SPA.
173 uint64_t dva_word
[2];
178 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept
179 * secret and is suitable for use in MAC algorithms as the key.
181 typedef struct zio_cksum_salt
{
182 uint8_t zcs_bytes
[32];
186 * Each block is described by its DVAs, time of birth, checksum, etc.
187 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
189 * 64 56 48 40 32 24 16 8 0
190 * +-------+-------+-------+-------+-------+-------+-------+-------+
191 * 0 | pad | vdev1 | GRID | ASIZE |
192 * +-------+-------+-------+-------+-------+-------+-------+-------+
194 * +-------+-------+-------+-------+-------+-------+-------+-------+
195 * 2 | pad | vdev2 | GRID | ASIZE |
196 * +-------+-------+-------+-------+-------+-------+-------+-------+
198 * +-------+-------+-------+-------+-------+-------+-------+-------+
199 * 4 | pad | vdev3 | GRID | ASIZE |
200 * +-------+-------+-------+-------+-------+-------+-------+-------+
202 * +-------+-------+-------+-------+-------+-------+-------+-------+
203 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE |
204 * +-------+-------+-------+-------+-------+-------+-------+-------+
206 * +-------+-------+-------+-------+-------+-------+-------+-------+
208 * +-------+-------+-------+-------+-------+-------+-------+-------+
209 * 9 | physical birth txg |
210 * +-------+-------+-------+-------+-------+-------+-------+-------+
211 * a | logical birth txg |
212 * +-------+-------+-------+-------+-------+-------+-------+-------+
214 * +-------+-------+-------+-------+-------+-------+-------+-------+
216 * +-------+-------+-------+-------+-------+-------+-------+-------+
218 * +-------+-------+-------+-------+-------+-------+-------+-------+
220 * +-------+-------+-------+-------+-------+-------+-------+-------+
222 * +-------+-------+-------+-------+-------+-------+-------+-------+
226 * vdev virtual device ID
227 * offset offset into virtual device
229 * PSIZE physical size (after compression)
230 * ASIZE allocated size (including RAID-Z parity and gang block headers)
231 * GRID RAID-Z layout information (reserved for future use)
232 * cksum checksum function
233 * comp compression function
234 * G gang block indicator
235 * B byteorder (endianness)
238 * E blkptr_t contains embedded data (see below)
239 * lvl level of indirection
240 * type DMU object type
241 * phys birth txg when dva[0] was written; zero if same as logical birth txg
242 * note that typically all the dva's would be written in this
243 * txg, but they could be different if they were moved by
245 * log. birth transaction group in which the block was logically born
246 * fill count number of non-zero blocks under this bp
247 * checksum[4] 256-bit checksum of the data this bp describes
251 * The blkptr_t's of encrypted blocks also need to store the encryption
252 * parameters so that the block can be decrypted. This layout is as follows:
254 * 64 56 48 40 32 24 16 8 0
255 * +-------+-------+-------+-------+-------+-------+-------+-------+
256 * 0 | vdev1 | GRID | ASIZE |
257 * +-------+-------+-------+-------+-------+-------+-------+-------+
259 * +-------+-------+-------+-------+-------+-------+-------+-------+
260 * 2 | vdev2 | GRID | ASIZE |
261 * +-------+-------+-------+-------+-------+-------+-------+-------+
263 * +-------+-------+-------+-------+-------+-------+-------+-------+
265 * +-------+-------+-------+-------+-------+-------+-------+-------+
267 * +-------+-------+-------+-------+-------+-------+-------+-------+
268 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE |
269 * +-------+-------+-------+-------+-------+-------+-------+-------+
271 * +-------+-------+-------+-------+-------+-------+-------+-------+
273 * +-------+-------+-------+-------+-------+-------+-------+-------+
274 * 9 | physical birth txg |
275 * +-------+-------+-------+-------+-------+-------+-------+-------+
276 * a | logical birth txg |
277 * +-------+-------+-------+-------+-------+-------+-------+-------+
278 * b | IV2 | fill count |
279 * +-------+-------+-------+-------+-------+-------+-------+-------+
281 * +-------+-------+-------+-------+-------+-------+-------+-------+
283 * +-------+-------+-------+-------+-------+-------+-------+-------+
285 * +-------+-------+-------+-------+-------+-------+-------+-------+
287 * +-------+-------+-------+-------+-------+-------+-------+-------+
291 * salt Salt for generating encryption keys
292 * IV1 First 64 bits of encryption IV
293 * X Block requires encryption handling (set to 1)
294 * E blkptr_t contains embedded data (set to 0, see below)
295 * fill count number of non-zero blocks under this bp (truncated to 32 bits)
296 * IV2 Last 32 bits of encryption IV
297 * checksum[2] 128-bit checksum of the data this bp describes
298 * MAC[2] 128-bit message authentication code for this data
300 * The X bit being set indicates that this block is one of 3 types. If this is
301 * a level 0 block with an encrypted object type, the block is encrypted
302 * (see BP_IS_ENCRYPTED()). If this is a level 0 block with an unencrypted
303 * object type, this block is authenticated with an HMAC (see
304 * BP_IS_AUTHENTICATED()). Otherwise (if level > 0), this bp will use the MAC
305 * words to store a checksum-of-MACs from the level below (see
306 * BP_HAS_INDIRECT_MAC_CKSUM()). For convenience in the code, BP_IS_PROTECTED()
307 * refers to both encrypted and authenticated blocks and BP_USES_CRYPT()
308 * refers to any of these 3 kinds of blocks.
310 * The additional encryption parameters are the salt, IV, and MAC which are
311 * explained in greater detail in the block comment at the top of zio_crypt.c.
312 * The MAC occupies half of the checksum space since it serves a very similar
313 * purpose: to prevent data corruption on disk. The only functional difference
314 * is that the checksum is used to detect on-disk corruption whether or not the
315 * encryption key is loaded and the MAC provides additional protection against
316 * malicious disk tampering. We use the 3rd DVA to store the salt and first
317 * 64 bits of the IV. As a result encrypted blocks can only have 2 copies
318 * maximum instead of the normal 3. The last 32 bits of the IV are stored in
319 * the upper bits of what is usually the fill count. Note that only blocks at
320 * level 0 or -2 are ever encrypted, which allows us to guarantee that these
321 * 32 bits are not trampled over by other code (see zio_crypt.c for details).
322 * The salt and IV are not used for authenticated bps or bps with an indirect
323 * MAC checksum, so these blocks can utilize all 3 DVAs and the full 64 bits
324 * for the fill count.
328 * "Embedded" blkptr_t's don't actually point to a block, instead they
329 * have a data payload embedded in the blkptr_t itself. See the comment
330 * in blkptr.c for more details.
332 * The blkptr_t is laid out as follows:
334 * 64 56 48 40 32 24 16 8 0
335 * +-------+-------+-------+-------+-------+-------+-------+-------+
342 * +-------+-------+-------+-------+-------+-------+-------+-------+
343 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE |
344 * +-------+-------+-------+-------+-------+-------+-------+-------+
348 * +-------+-------+-------+-------+-------+-------+-------+-------+
349 * a | logical birth txg |
350 * +-------+-------+-------+-------+-------+-------+-------+-------+
356 * +-------+-------+-------+-------+-------+-------+-------+-------+
360 * payload contains the embedded data
361 * B (byteorder) byteorder (endianness)
362 * D (dedup) padding (set to zero)
363 * X encryption (set to zero)
364 * E (embedded) set to one
365 * lvl indirection level
366 * type DMU object type
367 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*)
368 * comp compression function of payload
369 * PSIZE size of payload after compression, in bytes
370 * LSIZE logical size of payload, in bytes
371 * note that 25 bits is enough to store the largest
372 * "normal" BP's LSIZE (2^16 * 2^9) in bytes
373 * log. birth transaction group in which the block was logically born
375 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
376 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
377 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
378 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
379 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must
380 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before
381 * other macros, as they assert that they are only used on BP's of the correct
382 * "embedded-ness". Encrypted blkptr_t's cannot be embedded because they use
383 * the payload space for encryption parameters (see the comment above on
384 * how encryption parameters are stored).
387 #define BPE_GET_ETYPE(bp) \
388 (ASSERT(BP_IS_EMBEDDED(bp)), \
389 BF64_GET((bp)->blk_prop, 40, 8))
390 #define BPE_SET_ETYPE(bp, t) do { \
391 ASSERT(BP_IS_EMBEDDED(bp)); \
392 BF64_SET((bp)->blk_prop, 40, 8, t); \
393 _NOTE(CONSTCOND) } while (0)
395 #define BPE_GET_LSIZE(bp) \
396 (ASSERT(BP_IS_EMBEDDED(bp)), \
397 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
398 #define BPE_SET_LSIZE(bp, x) do { \
399 ASSERT(BP_IS_EMBEDDED(bp)); \
400 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
401 _NOTE(CONSTCOND) } while (0)
403 #define BPE_GET_PSIZE(bp) \
404 (ASSERT(BP_IS_EMBEDDED(bp)), \
405 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
406 #define BPE_SET_PSIZE(bp, x) do { \
407 ASSERT(BP_IS_EMBEDDED(bp)); \
408 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
409 _NOTE(CONSTCOND) } while (0)
411 typedef enum bp_embedded_type
{
412 BP_EMBEDDED_TYPE_DATA
,
413 BP_EMBEDDED_TYPE_RESERVED
, /* Reserved for Delphix byteswap feature. */
414 BP_EMBEDDED_TYPE_REDACTED
,
415 NUM_BP_EMBEDDED_TYPES
416 } bp_embedded_type_t
;
418 #define BPE_NUM_WORDS 14
419 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
420 #define BPE_IS_PAYLOADWORD(bp, wp) \
421 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
423 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */
424 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */
425 #define SPA_SYNC_MIN_VDEVS 3 /* min vdevs to update during sync */
428 * A block is a hole when it has either 1) never been written to, or
429 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
430 * without physically allocating disk space. Holes are represented in the
431 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is
432 * done through the BP_IS_HOLE macro. For holes, the logical size, level,
433 * DMU object type, and birth times are all also stored for holes that
434 * were written to at some point (i.e. were punched after having been filled).
436 typedef struct blkptr
{
437 dva_t blk_dva
[SPA_DVAS_PER_BP
]; /* Data Virtual Addresses */
438 uint64_t blk_prop
; /* size, compression, type, etc */
439 uint64_t blk_pad
[2]; /* Extra space for the future */
440 uint64_t blk_phys_birth
; /* txg when block was allocated */
441 uint64_t blk_birth
; /* transaction group at birth */
442 uint64_t blk_fill
; /* fill count */
443 zio_cksum_t blk_cksum
; /* 256-bit checksum */
447 * Macros to get and set fields in a bp or DVA.
449 #define DVA_GET_ASIZE(dva) \
450 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
451 #define DVA_SET_ASIZE(dva, x) \
452 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
453 SPA_MINBLOCKSHIFT, 0, x)
455 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8)
456 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x)
458 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, SPA_VDEVBITS)
459 #define DVA_SET_VDEV(dva, x) \
460 BF64_SET((dva)->dva_word[0], 32, SPA_VDEVBITS, x)
462 #define DVA_GET_OFFSET(dva) \
463 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
464 #define DVA_SET_OFFSET(dva, x) \
465 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
467 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1)
468 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x)
470 #define BP_GET_LSIZE(bp) \
471 (BP_IS_EMBEDDED(bp) ? \
472 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
473 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
474 #define BP_SET_LSIZE(bp, x) do { \
475 ASSERT(!BP_IS_EMBEDDED(bp)); \
476 BF64_SET_SB((bp)->blk_prop, \
477 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
478 _NOTE(CONSTCOND) } while (0)
480 #define BP_GET_PSIZE(bp) \
481 (BP_IS_EMBEDDED(bp) ? 0 : \
482 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1))
483 #define BP_SET_PSIZE(bp, x) do { \
484 ASSERT(!BP_IS_EMBEDDED(bp)); \
485 BF64_SET_SB((bp)->blk_prop, \
486 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
487 _NOTE(CONSTCOND) } while (0)
489 #define BP_GET_COMPRESS(bp) \
490 BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS)
491 #define BP_SET_COMPRESS(bp, x) \
492 BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x)
494 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1)
495 #define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x)
497 #define BP_GET_CHECKSUM(bp) \
498 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \
499 BF64_GET((bp)->blk_prop, 40, 8))
500 #define BP_SET_CHECKSUM(bp, x) do { \
501 ASSERT(!BP_IS_EMBEDDED(bp)); \
502 BF64_SET((bp)->blk_prop, 40, 8, x); \
503 _NOTE(CONSTCOND) } while (0)
505 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8)
506 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x)
508 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5)
509 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x)
511 /* encrypted, authenticated, and MAC cksum bps use the same bit */
512 #define BP_USES_CRYPT(bp) BF64_GET((bp)->blk_prop, 61, 1)
513 #define BP_SET_CRYPT(bp, x) BF64_SET((bp)->blk_prop, 61, 1, x)
515 #define BP_IS_ENCRYPTED(bp) \
516 (BP_USES_CRYPT(bp) && \
517 BP_GET_LEVEL(bp) <= 0 && \
518 DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp)))
520 #define BP_IS_AUTHENTICATED(bp) \
521 (BP_USES_CRYPT(bp) && \
522 BP_GET_LEVEL(bp) <= 0 && \
523 !DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp)))
525 #define BP_HAS_INDIRECT_MAC_CKSUM(bp) \
526 (BP_USES_CRYPT(bp) && BP_GET_LEVEL(bp) > 0)
528 #define BP_IS_PROTECTED(bp) \
529 (BP_IS_ENCRYPTED(bp) || BP_IS_AUTHENTICATED(bp))
531 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1)
532 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x)
534 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1)
535 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x)
537 #define BP_GET_FREE(bp) BF64_GET((bp)->blk_fill, 0, 1)
538 #define BP_SET_FREE(bp, x) BF64_SET((bp)->blk_fill, 0, 1, x)
540 #define BP_PHYSICAL_BIRTH(bp) \
541 (BP_IS_EMBEDDED(bp) ? 0 : \
542 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
544 #define BP_SET_BIRTH(bp, logical, physical) \
546 ASSERT(!BP_IS_EMBEDDED(bp)); \
547 (bp)->blk_birth = (logical); \
548 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
551 #define BP_GET_FILL(bp) \
552 ((BP_IS_ENCRYPTED(bp)) ? BF64_GET((bp)->blk_fill, 0, 32) : \
553 ((BP_IS_EMBEDDED(bp)) ? 1 : (bp)->blk_fill))
555 #define BP_SET_FILL(bp, fill) \
557 if (BP_IS_ENCRYPTED(bp)) \
558 BF64_SET((bp)->blk_fill, 0, 32, fill); \
560 (bp)->blk_fill = fill; \
563 #define BP_GET_IV2(bp) \
564 (ASSERT(BP_IS_ENCRYPTED(bp)), \
565 BF64_GET((bp)->blk_fill, 32, 32))
566 #define BP_SET_IV2(bp, iv2) \
568 ASSERT(BP_IS_ENCRYPTED(bp)); \
569 BF64_SET((bp)->blk_fill, 32, 32, iv2); \
572 #define BP_IS_METADATA(bp) \
573 (BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))
575 #define BP_GET_ASIZE(bp) \
576 (BP_IS_EMBEDDED(bp) ? 0 : \
577 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
578 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
579 (DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))
581 #define BP_GET_UCSIZE(bp) \
582 (BP_IS_METADATA(bp) ? BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
584 #define BP_GET_NDVAS(bp) \
585 (BP_IS_EMBEDDED(bp) ? 0 : \
586 !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
587 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
588 (!!DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))
590 #define BP_COUNT_GANG(bp) \
591 (BP_IS_EMBEDDED(bp) ? 0 : \
592 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \
593 DVA_GET_GANG(&(bp)->blk_dva[1]) + \
594 (DVA_GET_GANG(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp))))
596 #define DVA_EQUAL(dva1, dva2) \
597 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
598 (dva1)->dva_word[0] == (dva2)->dva_word[0])
600 #define BP_EQUAL(bp1, bp2) \
601 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \
602 (bp1)->blk_birth == (bp2)->blk_birth && \
603 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \
604 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \
605 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
608 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0)
610 #define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0])
611 #define BP_IS_GANG(bp) \
612 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp)))
613 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \
614 (dva)->dva_word[1] == 0ULL)
615 #define BP_IS_HOLE(bp) \
616 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp)))
618 #define BP_SET_REDACTED(bp) \
620 BP_SET_EMBEDDED(bp, B_TRUE); \
621 BPE_SET_ETYPE(bp, BP_EMBEDDED_TYPE_REDACTED); \
623 #define BP_IS_REDACTED(bp) \
624 (BP_IS_EMBEDDED(bp) && BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_REDACTED)
626 /* BP_IS_RAIDZ(bp) assumes no block compression */
627 #define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
630 #define BP_ZERO(bp) \
632 (bp)->blk_dva[0].dva_word[0] = 0; \
633 (bp)->blk_dva[0].dva_word[1] = 0; \
634 (bp)->blk_dva[1].dva_word[0] = 0; \
635 (bp)->blk_dva[1].dva_word[1] = 0; \
636 (bp)->blk_dva[2].dva_word[0] = 0; \
637 (bp)->blk_dva[2].dva_word[1] = 0; \
638 (bp)->blk_prop = 0; \
639 (bp)->blk_pad[0] = 0; \
640 (bp)->blk_pad[1] = 0; \
641 (bp)->blk_phys_birth = 0; \
642 (bp)->blk_birth = 0; \
643 (bp)->blk_fill = 0; \
644 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \
648 #define ZFS_HOST_BYTEORDER (0ULL)
650 #define ZFS_HOST_BYTEORDER (1ULL)
653 #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
655 #define BP_SPRINTF_LEN 400
658 * This macro allows code sharing between zfs, libzpool, and mdb.
659 * 'func' is either snprintf() or mdb_snprintf().
660 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
663 #define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
665 static const char *copyname[] = \
666 { "zero", "single", "double", "triple" }; \
669 const char *crypt_type; \
671 if (BP_IS_ENCRYPTED(bp)) { \
672 crypt_type = "encrypted"; \
673 /* LINTED E_SUSPICIOUS_COMPARISON */ \
674 } else if (BP_IS_AUTHENTICATED(bp)) { \
675 crypt_type = "authenticated"; \
676 } else if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) { \
677 crypt_type = "indirect-MAC"; \
679 crypt_type = "unencrypted"; \
683 len += func(buf + len, size - len, "<NULL>"); \
684 } else if (BP_IS_HOLE(bp)) { \
685 len += func(buf + len, size - len, \
687 "size=%llxL birth=%lluL", \
688 (u_longlong_t)BP_GET_LEVEL(bp), \
690 (u_longlong_t)BP_GET_LSIZE(bp), \
691 (u_longlong_t)bp->blk_birth); \
692 } else if (BP_IS_EMBEDDED(bp)) { \
693 len = func(buf + len, size - len, \
694 "EMBEDDED [L%llu %s] et=%u %s " \
695 "size=%llxL/%llxP birth=%lluL", \
696 (u_longlong_t)BP_GET_LEVEL(bp), \
698 (int)BPE_GET_ETYPE(bp), \
700 (u_longlong_t)BPE_GET_LSIZE(bp), \
701 (u_longlong_t)BPE_GET_PSIZE(bp), \
702 (u_longlong_t)bp->blk_birth); \
703 } else if (BP_IS_REDACTED(bp)) { \
704 len += func(buf + len, size - len, \
705 "REDACTED [L%llu %s] size=%llxL birth=%lluL", \
706 (u_longlong_t)BP_GET_LEVEL(bp), \
708 (u_longlong_t)BP_GET_LSIZE(bp), \
709 (u_longlong_t)bp->blk_birth); \
711 for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \
712 const dva_t *dva = &bp->blk_dva[d]; \
713 if (DVA_IS_VALID(dva)) \
715 len += func(buf + len, size - len, \
716 "DVA[%d]=<%llu:%llx:%llx>%c", d, \
717 (u_longlong_t)DVA_GET_VDEV(dva), \
718 (u_longlong_t)DVA_GET_OFFSET(dva), \
719 (u_longlong_t)DVA_GET_ASIZE(dva), \
722 if (BP_IS_ENCRYPTED(bp)) { \
723 len += func(buf + len, size - len, \
724 "salt=%llx iv=%llx:%llx%c", \
725 (u_longlong_t)bp->blk_dva[2].dva_word[0], \
726 (u_longlong_t)bp->blk_dva[2].dva_word[1], \
727 (u_longlong_t)BP_GET_IV2(bp), \
730 if (BP_IS_GANG(bp) && \
731 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \
732 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \
734 len += func(buf + len, size - len, \
735 "[L%llu %s] %s %s %s %s %s %s %s%c" \
736 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \
737 "cksum=%llx:%llx:%llx:%llx", \
738 (u_longlong_t)BP_GET_LEVEL(bp), \
743 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \
744 BP_IS_GANG(bp) ? "gang" : "contiguous", \
745 BP_GET_DEDUP(bp) ? "dedup" : "unique", \
748 (u_longlong_t)BP_GET_LSIZE(bp), \
749 (u_longlong_t)BP_GET_PSIZE(bp), \
750 (u_longlong_t)bp->blk_birth, \
751 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \
752 (u_longlong_t)BP_GET_FILL(bp), \
754 (u_longlong_t)bp->blk_cksum.zc_word[0], \
755 (u_longlong_t)bp->blk_cksum.zc_word[1], \
756 (u_longlong_t)bp->blk_cksum.zc_word[2], \
757 (u_longlong_t)bp->blk_cksum.zc_word[3]); \
759 ASSERT(len < size); \
762 #define BP_GET_BUFC_TYPE(bp) \
763 (BP_IS_METADATA(bp) ? ARC_BUFC_METADATA : ARC_BUFC_DATA)
765 typedef enum spa_import_type
{
771 * Send TRIM commands in-line during normal pool operation while deleting.
776 SPA_AUTOTRIM_OFF
= 0, /* default */
781 * Reason TRIM command was issued, used internally for accounting purposes.
783 typedef enum trim_type
{
784 TRIM_TYPE_MANUAL
= 0,
788 /* state manipulation functions */
789 extern int spa_open(const char *pool
, spa_t
**, void *tag
);
790 extern int spa_open_rewind(const char *pool
, spa_t
**, void *tag
,
791 nvlist_t
*policy
, nvlist_t
**config
);
792 extern int spa_get_stats(const char *pool
, nvlist_t
**config
, char *altroot
,
794 extern int spa_create(const char *pool
, nvlist_t
*nvroot
, nvlist_t
*props
,
795 nvlist_t
*zplprops
, struct dsl_crypto_params
*dcp
);
796 extern int spa_import(char *pool
, nvlist_t
*config
, nvlist_t
*props
,
798 extern nvlist_t
*spa_tryimport(nvlist_t
*tryconfig
);
799 extern int spa_destroy(char *pool
);
800 extern int spa_checkpoint(const char *pool
);
801 extern int spa_checkpoint_discard(const char *pool
);
802 extern int spa_export(char *pool
, nvlist_t
**oldconfig
, boolean_t force
,
803 boolean_t hardforce
);
804 extern int spa_reset(char *pool
);
805 extern void spa_async_request(spa_t
*spa
, int flag
);
806 extern void spa_async_unrequest(spa_t
*spa
, int flag
);
807 extern void spa_async_suspend(spa_t
*spa
);
808 extern void spa_async_resume(spa_t
*spa
);
809 extern spa_t
*spa_inject_addref(char *pool
);
810 extern void spa_inject_delref(spa_t
*spa
);
811 extern void spa_scan_stat_init(spa_t
*spa
);
812 extern int spa_scan_get_stats(spa_t
*spa
, pool_scan_stat_t
*ps
);
813 extern int bpobj_enqueue_alloc_cb(void *arg
, const blkptr_t
*bp
, dmu_tx_t
*tx
);
814 extern int bpobj_enqueue_free_cb(void *arg
, const blkptr_t
*bp
, dmu_tx_t
*tx
);
816 #define SPA_ASYNC_CONFIG_UPDATE 0x01
817 #define SPA_ASYNC_REMOVE 0x02
818 #define SPA_ASYNC_PROBE 0x04
819 #define SPA_ASYNC_RESILVER_DONE 0x08
820 #define SPA_ASYNC_RESILVER 0x10
821 #define SPA_ASYNC_AUTOEXPAND 0x20
822 #define SPA_ASYNC_REMOVE_DONE 0x40
823 #define SPA_ASYNC_REMOVE_STOP 0x80
824 #define SPA_ASYNC_INITIALIZE_RESTART 0x100
825 #define SPA_ASYNC_TRIM_RESTART 0x200
826 #define SPA_ASYNC_AUTOTRIM_RESTART 0x400
829 * Controls the behavior of spa_vdev_remove().
831 #define SPA_REMOVE_UNSPARE 0x01
832 #define SPA_REMOVE_DONE 0x02
834 /* device manipulation */
835 extern int spa_vdev_add(spa_t
*spa
, nvlist_t
*nvroot
);
836 extern int spa_vdev_attach(spa_t
*spa
, uint64_t guid
, nvlist_t
*nvroot
,
838 extern int spa_vdev_detach(spa_t
*spa
, uint64_t guid
, uint64_t pguid
,
840 extern int spa_vdev_remove(spa_t
*spa
, uint64_t guid
, boolean_t unspare
);
841 extern boolean_t
spa_vdev_remove_active(spa_t
*spa
);
842 extern int spa_vdev_initialize(spa_t
*spa
, nvlist_t
*nv
, uint64_t cmd_type
,
843 nvlist_t
*vdev_errlist
);
844 extern int spa_vdev_trim(spa_t
*spa
, nvlist_t
*nv
, uint64_t cmd_type
,
845 uint64_t rate
, boolean_t partial
, boolean_t secure
, nvlist_t
*vdev_errlist
);
846 extern int spa_vdev_setpath(spa_t
*spa
, uint64_t guid
, const char *newpath
);
847 extern int spa_vdev_setfru(spa_t
*spa
, uint64_t guid
, const char *newfru
);
848 extern int spa_vdev_split_mirror(spa_t
*spa
, char *newname
, nvlist_t
*config
,
849 nvlist_t
*props
, boolean_t exp
);
851 /* spare state (which is global across all pools) */
852 extern void spa_spare_add(vdev_t
*vd
);
853 extern void spa_spare_remove(vdev_t
*vd
);
854 extern boolean_t
spa_spare_exists(uint64_t guid
, uint64_t *pool
, int *refcnt
);
855 extern void spa_spare_activate(vdev_t
*vd
);
857 /* L2ARC state (which is global across all pools) */
858 extern void spa_l2cache_add(vdev_t
*vd
);
859 extern void spa_l2cache_remove(vdev_t
*vd
);
860 extern boolean_t
spa_l2cache_exists(uint64_t guid
, uint64_t *pool
);
861 extern void spa_l2cache_activate(vdev_t
*vd
);
862 extern void spa_l2cache_drop(spa_t
*spa
);
865 extern int spa_scan(spa_t
*spa
, pool_scan_func_t func
);
866 extern int spa_scan_stop(spa_t
*spa
);
867 extern int spa_scrub_pause_resume(spa_t
*spa
, pool_scrub_cmd_t flag
);
870 extern void spa_sync(spa_t
*spa
, uint64_t txg
); /* only for DMU use */
871 extern void spa_sync_allpools(void);
873 extern int zfs_sync_pass_deferred_free
;
875 /* spa namespace global mutex */
876 extern kmutex_t spa_namespace_lock
;
879 * SPA configuration functions in spa_config.c
882 #define SPA_CONFIG_UPDATE_POOL 0
883 #define SPA_CONFIG_UPDATE_VDEVS 1
885 extern void spa_write_cachefile(spa_t
*, boolean_t
, boolean_t
);
886 extern void spa_config_load(void);
887 extern nvlist_t
*spa_all_configs(uint64_t *);
888 extern void spa_config_set(spa_t
*spa
, nvlist_t
*config
);
889 extern nvlist_t
*spa_config_generate(spa_t
*spa
, vdev_t
*vd
, uint64_t txg
,
891 extern void spa_config_update(spa_t
*spa
, int what
);
894 * Miscellaneous SPA routines in spa_misc.c
897 /* Namespace manipulation */
898 extern spa_t
*spa_lookup(const char *name
);
899 extern spa_t
*spa_add(const char *name
, nvlist_t
*config
, const char *altroot
);
900 extern void spa_remove(spa_t
*spa
);
901 extern spa_t
*spa_next(spa_t
*prev
);
903 /* Refcount functions */
904 extern void spa_open_ref(spa_t
*spa
, void *tag
);
905 extern void spa_close(spa_t
*spa
, void *tag
);
906 extern void spa_async_close(spa_t
*spa
, void *tag
);
907 extern boolean_t
spa_refcount_zero(spa_t
*spa
);
909 #define SCL_NONE 0x00
910 #define SCL_CONFIG 0x01
911 #define SCL_STATE 0x02
912 #define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */
913 #define SCL_ALLOC 0x08
915 #define SCL_FREE 0x20
916 #define SCL_VDEV 0x40
918 #define SCL_ALL ((1 << SCL_LOCKS) - 1)
919 #define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO)
921 /* Historical pool statistics */
922 typedef struct spa_history_kstat
{
929 } spa_history_kstat_t
;
931 typedef struct spa_history_list
{
933 procfs_list_t procfs_list
;
934 } spa_history_list_t
;
936 typedef struct spa_stats
{
937 spa_history_list_t read_history
;
938 spa_history_list_t txg_history
;
939 spa_history_kstat_t tx_assign_histogram
;
940 spa_history_kstat_t io_history
;
941 spa_history_list_t mmp_history
;
942 spa_history_kstat_t state
; /* pool state */
943 spa_history_kstat_t iostats
;
946 typedef enum txg_state
{
949 TXG_STATE_QUIESCED
= 2,
950 TXG_STATE_WAIT_FOR_SYNC
= 3,
951 TXG_STATE_SYNCED
= 4,
952 TXG_STATE_COMMITTED
= 5,
955 typedef struct txg_stat
{
962 /* Assorted pool IO kstats */
963 typedef struct spa_iostats
{
964 kstat_named_t trim_extents_written
;
965 kstat_named_t trim_bytes_written
;
966 kstat_named_t trim_extents_skipped
;
967 kstat_named_t trim_bytes_skipped
;
968 kstat_named_t trim_extents_failed
;
969 kstat_named_t trim_bytes_failed
;
970 kstat_named_t autotrim_extents_written
;
971 kstat_named_t autotrim_bytes_written
;
972 kstat_named_t autotrim_extents_skipped
;
973 kstat_named_t autotrim_bytes_skipped
;
974 kstat_named_t autotrim_extents_failed
;
975 kstat_named_t autotrim_bytes_failed
;
978 extern void spa_stats_init(spa_t
*spa
);
979 extern void spa_stats_destroy(spa_t
*spa
);
980 extern void spa_read_history_add(spa_t
*spa
, const zbookmark_phys_t
*zb
,
982 extern void spa_txg_history_add(spa_t
*spa
, uint64_t txg
, hrtime_t birth_time
);
983 extern int spa_txg_history_set(spa_t
*spa
, uint64_t txg
,
984 txg_state_t completed_state
, hrtime_t completed_time
);
985 extern txg_stat_t
*spa_txg_history_init_io(spa_t
*, uint64_t,
987 extern void spa_txg_history_fini_io(spa_t
*, txg_stat_t
*);
988 extern void spa_tx_assign_add_nsecs(spa_t
*spa
, uint64_t nsecs
);
989 extern int spa_mmp_history_set_skip(spa_t
*spa
, uint64_t mmp_kstat_id
);
990 extern int spa_mmp_history_set(spa_t
*spa
, uint64_t mmp_kstat_id
, int io_error
,
992 extern void spa_mmp_history_add(spa_t
*spa
, uint64_t txg
, uint64_t timestamp
,
993 uint64_t mmp_delay
, vdev_t
*vd
, int label
, uint64_t mmp_kstat_id
,
995 extern void spa_iostats_trim_add(spa_t
*spa
, trim_type_t type
,
996 uint64_t extents_written
, uint64_t bytes_written
,
997 uint64_t extents_skipped
, uint64_t bytes_skipped
,
998 uint64_t extents_failed
, uint64_t bytes_failed
);
999 extern void spa_import_progress_add(spa_t
*spa
);
1000 extern void spa_import_progress_remove(uint64_t spa_guid
);
1001 extern int spa_import_progress_set_mmp_check(uint64_t pool_guid
,
1002 uint64_t mmp_sec_remaining
);
1003 extern int spa_import_progress_set_max_txg(uint64_t pool_guid
,
1005 extern int spa_import_progress_set_state(uint64_t pool_guid
,
1006 spa_load_state_t spa_load_state
);
1008 /* Pool configuration locks */
1009 extern int spa_config_tryenter(spa_t
*spa
, int locks
, void *tag
, krw_t rw
);
1010 extern void spa_config_enter(spa_t
*spa
, int locks
, const void *tag
, krw_t rw
);
1011 extern void spa_config_exit(spa_t
*spa
, int locks
, const void *tag
);
1012 extern int spa_config_held(spa_t
*spa
, int locks
, krw_t rw
);
1014 /* Pool vdev add/remove lock */
1015 extern uint64_t spa_vdev_enter(spa_t
*spa
);
1016 extern uint64_t spa_vdev_config_enter(spa_t
*spa
);
1017 extern void spa_vdev_config_exit(spa_t
*spa
, vdev_t
*vd
, uint64_t txg
,
1018 int error
, char *tag
);
1019 extern int spa_vdev_exit(spa_t
*spa
, vdev_t
*vd
, uint64_t txg
, int error
);
1021 /* Pool vdev state change lock */
1022 extern void spa_vdev_state_enter(spa_t
*spa
, int oplock
);
1023 extern int spa_vdev_state_exit(spa_t
*spa
, vdev_t
*vd
, int error
);
1026 typedef enum spa_log_state
{
1027 SPA_LOG_UNKNOWN
= 0, /* unknown log state */
1028 SPA_LOG_MISSING
, /* missing log(s) */
1029 SPA_LOG_CLEAR
, /* clear the log(s) */
1030 SPA_LOG_GOOD
, /* log(s) are good */
1033 extern spa_log_state_t
spa_get_log_state(spa_t
*spa
);
1034 extern void spa_set_log_state(spa_t
*spa
, spa_log_state_t state
);
1035 extern int spa_reset_logs(spa_t
*spa
);
1037 /* Log claim callback */
1038 extern void spa_claim_notify(zio_t
*zio
);
1039 extern void spa_deadman(void *);
1041 /* Accessor functions */
1042 extern boolean_t
spa_shutting_down(spa_t
*spa
);
1043 extern struct dsl_pool
*spa_get_dsl(spa_t
*spa
);
1044 extern boolean_t
spa_is_initializing(spa_t
*spa
);
1045 extern boolean_t
spa_indirect_vdevs_loaded(spa_t
*spa
);
1046 extern blkptr_t
*spa_get_rootblkptr(spa_t
*spa
);
1047 extern void spa_set_rootblkptr(spa_t
*spa
, const blkptr_t
*bp
);
1048 extern void spa_altroot(spa_t
*, char *, size_t);
1049 extern int spa_sync_pass(spa_t
*spa
);
1050 extern char *spa_name(spa_t
*spa
);
1051 extern uint64_t spa_guid(spa_t
*spa
);
1052 extern uint64_t spa_load_guid(spa_t
*spa
);
1053 extern uint64_t spa_last_synced_txg(spa_t
*spa
);
1054 extern uint64_t spa_first_txg(spa_t
*spa
);
1055 extern uint64_t spa_syncing_txg(spa_t
*spa
);
1056 extern uint64_t spa_final_dirty_txg(spa_t
*spa
);
1057 extern uint64_t spa_version(spa_t
*spa
);
1058 extern pool_state_t
spa_state(spa_t
*spa
);
1059 extern spa_load_state_t
spa_load_state(spa_t
*spa
);
1060 extern uint64_t spa_freeze_txg(spa_t
*spa
);
1061 extern uint64_t spa_get_worst_case_asize(spa_t
*spa
, uint64_t lsize
);
1062 extern uint64_t spa_get_dspace(spa_t
*spa
);
1063 extern uint64_t spa_get_checkpoint_space(spa_t
*spa
);
1064 extern uint64_t spa_get_slop_space(spa_t
*spa
);
1065 extern void spa_update_dspace(spa_t
*spa
);
1066 extern uint64_t spa_version(spa_t
*spa
);
1067 extern boolean_t
spa_deflate(spa_t
*spa
);
1068 extern metaslab_class_t
*spa_normal_class(spa_t
*spa
);
1069 extern metaslab_class_t
*spa_log_class(spa_t
*spa
);
1070 extern metaslab_class_t
*spa_special_class(spa_t
*spa
);
1071 extern metaslab_class_t
*spa_dedup_class(spa_t
*spa
);
1072 extern metaslab_class_t
*spa_preferred_class(spa_t
*spa
, uint64_t size
,
1073 dmu_object_type_t objtype
, uint_t level
, uint_t special_smallblk
);
1075 extern void spa_evicting_os_register(spa_t
*, objset_t
*os
);
1076 extern void spa_evicting_os_deregister(spa_t
*, objset_t
*os
);
1077 extern void spa_evicting_os_wait(spa_t
*spa
);
1078 extern int spa_max_replication(spa_t
*spa
);
1079 extern int spa_prev_software_version(spa_t
*spa
);
1080 extern uint64_t spa_get_failmode(spa_t
*spa
);
1081 extern uint64_t spa_get_deadman_failmode(spa_t
*spa
);
1082 extern void spa_set_deadman_failmode(spa_t
*spa
, const char *failmode
);
1083 extern boolean_t
spa_suspended(spa_t
*spa
);
1084 extern uint64_t spa_bootfs(spa_t
*spa
);
1085 extern uint64_t spa_delegation(spa_t
*spa
);
1086 extern objset_t
*spa_meta_objset(spa_t
*spa
);
1087 extern space_map_t
*spa_syncing_log_sm(spa_t
*spa
);
1088 extern uint64_t spa_deadman_synctime(spa_t
*spa
);
1089 extern uint64_t spa_deadman_ziotime(spa_t
*spa
);
1090 extern uint64_t spa_dirty_data(spa_t
*spa
);
1091 extern spa_autotrim_t
spa_get_autotrim(spa_t
*spa
);
1093 /* Miscellaneous support routines */
1094 extern void spa_load_failed(spa_t
*spa
, const char *fmt
, ...);
1095 extern void spa_load_note(spa_t
*spa
, const char *fmt
, ...);
1096 extern void spa_activate_mos_feature(spa_t
*spa
, const char *feature
,
1098 extern void spa_deactivate_mos_feature(spa_t
*spa
, const char *feature
);
1099 extern spa_t
*spa_by_guid(uint64_t pool_guid
, uint64_t device_guid
);
1100 extern boolean_t
spa_guid_exists(uint64_t pool_guid
, uint64_t device_guid
);
1101 extern char *spa_strdup(const char *);
1102 extern void spa_strfree(char *);
1103 extern uint64_t spa_get_random(uint64_t range
);
1104 extern uint64_t spa_generate_guid(spa_t
*spa
);
1105 extern void snprintf_blkptr(char *buf
, size_t buflen
, const blkptr_t
*bp
);
1106 extern void spa_freeze(spa_t
*spa
);
1107 extern int spa_change_guid(spa_t
*spa
);
1108 extern void spa_upgrade(spa_t
*spa
, uint64_t version
);
1109 extern void spa_evict_all(void);
1110 extern vdev_t
*spa_lookup_by_guid(spa_t
*spa
, uint64_t guid
,
1112 extern boolean_t
spa_has_spare(spa_t
*, uint64_t guid
);
1113 extern uint64_t dva_get_dsize_sync(spa_t
*spa
, const dva_t
*dva
);
1114 extern uint64_t bp_get_dsize_sync(spa_t
*spa
, const blkptr_t
*bp
);
1115 extern uint64_t bp_get_dsize(spa_t
*spa
, const blkptr_t
*bp
);
1116 extern boolean_t
spa_has_slogs(spa_t
*spa
);
1117 extern boolean_t
spa_is_root(spa_t
*spa
);
1118 extern boolean_t
spa_writeable(spa_t
*spa
);
1119 extern boolean_t
spa_has_pending_synctask(spa_t
*spa
);
1120 extern int spa_maxblocksize(spa_t
*spa
);
1121 extern int spa_maxdnodesize(spa_t
*spa
);
1122 extern boolean_t
spa_has_checkpoint(spa_t
*spa
);
1123 extern boolean_t
spa_importing_readonly_checkpoint(spa_t
*spa
);
1124 extern boolean_t
spa_suspend_async_destroy(spa_t
*spa
);
1125 extern uint64_t spa_min_claim_txg(spa_t
*spa
);
1126 extern boolean_t
zfs_dva_valid(spa_t
*spa
, const dva_t
*dva
,
1127 const blkptr_t
*bp
);
1128 typedef void (*spa_remap_cb_t
)(uint64_t vdev
, uint64_t offset
, uint64_t size
,
1130 extern boolean_t
spa_remap_blkptr(spa_t
*spa
, blkptr_t
*bp
,
1131 spa_remap_cb_t callback
, void *arg
);
1132 extern uint64_t spa_get_last_removal_txg(spa_t
*spa
);
1133 extern boolean_t
spa_trust_config(spa_t
*spa
);
1134 extern uint64_t spa_missing_tvds_allowed(spa_t
*spa
);
1135 extern void spa_set_missing_tvds(spa_t
*spa
, uint64_t missing
);
1136 extern boolean_t
spa_top_vdevs_spacemap_addressable(spa_t
*spa
);
1137 extern uint64_t spa_total_metaslabs(spa_t
*spa
);
1138 extern boolean_t
spa_multihost(spa_t
*spa
);
1139 extern uint32_t spa_get_hostid(spa_t
*spa
);
1140 extern void spa_activate_allocation_classes(spa_t
*, dmu_tx_t
*);
1141 extern boolean_t
spa_livelist_delete_check(spa_t
*spa
);
1143 extern int spa_mode(spa_t
*spa
);
1144 extern uint64_t zfs_strtonum(const char *str
, char **nptr
);
1146 extern char *spa_his_ievent_table
[];
1148 extern void spa_history_create_obj(spa_t
*spa
, dmu_tx_t
*tx
);
1149 extern int spa_history_get(spa_t
*spa
, uint64_t *offset
, uint64_t *len_read
,
1151 extern int spa_history_log(spa_t
*spa
, const char *his_buf
);
1152 extern int spa_history_log_nvl(spa_t
*spa
, nvlist_t
*nvl
);
1153 extern void spa_history_log_version(spa_t
*spa
, const char *operation
,
1155 extern void spa_history_log_internal(spa_t
*spa
, const char *operation
,
1156 dmu_tx_t
*tx
, const char *fmt
, ...) __printflike(4, 5);
1157 extern void spa_history_log_internal_ds(struct dsl_dataset
*ds
, const char *op
,
1158 dmu_tx_t
*tx
, const char *fmt
, ...) __printflike(4, 5);
1159 extern void spa_history_log_internal_dd(dsl_dir_t
*dd
, const char *operation
,
1160 dmu_tx_t
*tx
, const char *fmt
, ...) __printflike(4, 5);
1162 extern const char *spa_state_to_name(spa_t
*spa
);
1164 /* error handling */
1165 struct zbookmark_phys
;
1166 extern void spa_log_error(spa_t
*spa
, const zbookmark_phys_t
*zb
);
1167 extern int zfs_ereport_post(const char *class, spa_t
*spa
, vdev_t
*vd
,
1168 const zbookmark_phys_t
*zb
, zio_t
*zio
, uint64_t stateoroffset
,
1170 extern boolean_t
zfs_ereport_is_valid(const char *class, spa_t
*spa
, vdev_t
*vd
,
1172 extern nvlist_t
*zfs_event_create(spa_t
*spa
, vdev_t
*vd
, const char *type
,
1173 const char *name
, nvlist_t
*aux
);
1174 extern void zfs_post_remove(spa_t
*spa
, vdev_t
*vd
);
1175 extern void zfs_post_state_change(spa_t
*spa
, vdev_t
*vd
, uint64_t laststate
);
1176 extern void zfs_post_autoreplace(spa_t
*spa
, vdev_t
*vd
);
1177 extern uint64_t spa_get_errlog_size(spa_t
*spa
);
1178 extern int spa_get_errlog(spa_t
*spa
, void *uaddr
, size_t *count
);
1179 extern void spa_errlog_rotate(spa_t
*spa
);
1180 extern void spa_errlog_drain(spa_t
*spa
);
1181 extern void spa_errlog_sync(spa_t
*spa
, uint64_t txg
);
1182 extern void spa_get_errlists(spa_t
*spa
, avl_tree_t
*last
, avl_tree_t
*scrub
);
1185 extern void vdev_cache_stat_init(void);
1186 extern void vdev_cache_stat_fini(void);
1189 extern void vdev_mirror_stat_init(void);
1190 extern void vdev_mirror_stat_fini(void);
1192 /* Initialization and termination */
1193 extern void spa_init(int flags
);
1194 extern void spa_fini(void);
1195 extern void spa_boot_init(void);
1198 extern int spa_prop_set(spa_t
*spa
, nvlist_t
*nvp
);
1199 extern int spa_prop_get(spa_t
*spa
, nvlist_t
**nvp
);
1200 extern void spa_prop_clear_bootfs(spa_t
*spa
, uint64_t obj
, dmu_tx_t
*tx
);
1201 extern void spa_configfile_set(spa_t
*, nvlist_t
*, boolean_t
);
1203 /* asynchronous event notification */
1204 extern void spa_event_notify(spa_t
*spa
, vdev_t
*vdev
, nvlist_t
*hist_nvl
,
1208 #define dprintf_bp(bp, fmt, ...) do { \
1209 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
1210 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \
1211 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \
1212 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \
1213 kmem_free(__blkbuf, BP_SPRINTF_LEN); \
1215 _NOTE(CONSTCOND) } while (0)
1217 #define dprintf_bp(bp, fmt, ...)
1220 extern int spa_mode_global
; /* mode, e.g. FREAD | FWRITE */
1221 extern int zfs_deadman_enabled
;
1222 extern unsigned long zfs_deadman_synctime_ms
;
1223 extern unsigned long zfs_deadman_ziotime_ms
;
1224 extern unsigned long zfs_deadman_checktime_ms
;
1230 #endif /* _SYS_SPA_H */