]> git.proxmox.com Git - mirror_zfs.git/blob - include/sys/spa.h
projects / mirror_zfs.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Fix bp_embedded_type enum definition
[mirror_zfs.git] / include / sys / spa.h
1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21 /*
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2011, 2018 by Delphix. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
26 * Copyright 2013 Saso Kiselkov. All rights reserved.
27 * Copyright (c) 2014 Integros [integros.com]
28 * Copyright 2017 Joyent, Inc.
29 * Copyright (c) 2017 Datto Inc.
30 * Copyright (c) 2017, Intel Corporation.
31 */
32
33 #ifndef _SYS_SPA_H
34 #define _SYS_SPA_H
35
36 #include <sys/avl.h>
37 #include <sys/zfs_context.h>
38 #include <sys/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>
44 #include <sys/dmu.h>
45
46 #ifdef __cplusplus
47 extern "C" {
48 #endif
49
50 /*
51 * Forward references that lots of things need.
52 */
53 typedef struct spa spa_t;
54 typedef struct vdev vdev_t;
55 typedef struct metaslab metaslab_t;
56 typedef struct metaslab_group metaslab_group_t;
57 typedef struct metaslab_class metaslab_class_t;
58 typedef struct zio zio_t;
59 typedef struct zilog zilog_t;
60 typedef struct spa_aux_vdev spa_aux_vdev_t;
61 typedef struct ddt ddt_t;
62 typedef struct ddt_entry ddt_entry_t;
63 typedef struct zbookmark_phys zbookmark_phys_t;
64
65 struct dsl_pool;
66 struct dsl_dataset;
67 struct dsl_crypto_params;
68
69 /*
70 * General-purpose 32-bit and 64-bit bitfield encodings.
71 */
72 #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len))
73 #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len))
74 #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low))
75 #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low))
76
77 #define BF32_GET(x, low, len) BF32_DECODE(x, low, len)
78 #define BF64_GET(x, low, len) BF64_DECODE(x, low, len)
79
80 #define BF32_SET(x, low, len, val) do { \
81 ASSERT3U(val, <, 1U << (len)); \
82 ASSERT3U(low + len, <=, 32); \
83 (x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \
84 _NOTE(CONSTCOND) } while (0)
85
86 #define BF64_SET(x, low, len, val) do { \
87 ASSERT3U(val, <, 1ULL << (len)); \
88 ASSERT3U(low + len, <=, 64); \
89 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \
90 _NOTE(CONSTCOND) } while (0)
91
92 #define BF32_GET_SB(x, low, len, shift, bias) \
93 ((BF32_GET(x, low, len) + (bias)) << (shift))
94 #define BF64_GET_SB(x, low, len, shift, bias) \
95 ((BF64_GET(x, low, len) + (bias)) << (shift))
96
97 /*
98 * We use ASSERT3U instead of ASSERT in these macros to prevent a lint error in
99 * the case where val is a constant. We can't fix ASSERT because it's used as
100 * an expression in several places in the kernel; as a result, changing it to
101 * the do{} while() syntax to allow us to _NOTE the CONSTCOND is not an option.
102 */
103 #define BF32_SET_SB(x, low, len, shift, bias, val) do { \
104 ASSERT3U(IS_P2ALIGNED(val, 1U << shift), !=, B_FALSE); \
105 ASSERT3S((val) >> (shift), >=, bias); \
106 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \
107 _NOTE(CONSTCOND) } while (0)
108 #define BF64_SET_SB(x, low, len, shift, bias, val) do { \
109 ASSERT3U(IS_P2ALIGNED(val, 1ULL << shift), !=, B_FALSE); \
110 ASSERT3S((val) >> (shift), >=, bias); \
111 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \
112 _NOTE(CONSTCOND) } while (0)
113
114 /*
115 * We currently support block sizes from 512 bytes to 16MB.
116 * The benefits of larger blocks, and thus larger IO, need to be weighed
117 * against the cost of COWing a giant block to modify one byte, and the
118 * large latency of reading or writing a large block.
119 *
120 * Note that although blocks up to 16MB are supported, the recordsize
121 * property can not be set larger than zfs_max_recordsize (default 1MB).
122 * See the comment near zfs_max_recordsize in dsl_dataset.c for details.
123 *
124 * Note that although the LSIZE field of the blkptr_t can store sizes up
125 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to
126 * 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
127 */
128 #define SPA_MINBLOCKSHIFT 9
129 #define SPA_OLD_MAXBLOCKSHIFT 17
130 #define SPA_MAXBLOCKSHIFT 24
131 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT)
132 #define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT)
133 #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT)
134
135 /*
136 * Alignment Shift (ashift) is an immutable, internal top-level vdev property
137 * which can only be set at vdev creation time. Physical writes are always done
138 * according to it, which makes 2^ashift the smallest possible IO on a vdev.
139 *
140 * We currently allow values ranging from 512 bytes (2^9 = 512) to 64 KiB
141 * (2^16 = 65,536).
142 */
143 #define ASHIFT_MIN 9
144 #define ASHIFT_MAX 16
145
146 /*
147 * Size of block to hold the configuration data (a packed nvlist)
148 */
149 #define SPA_CONFIG_BLOCKSIZE (1ULL << 14)
150
151 /*
152 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
153 * The ASIZE encoding should be at least 64 times larger (6 more bits)
154 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
155 * overhead, three DVAs per bp, plus one more bit in case we do anything
156 * else that expands the ASIZE.
157 */
158 #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */
159 #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */
160 #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */
161
162 #define SPA_COMPRESSBITS 7
163 #define SPA_VDEVBITS 24
164
165 /*
166 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
167 * The members of the dva_t should be considered opaque outside the SPA.
168 */
169 typedef struct dva {
170 uint64_t dva_word[2];
171 } dva_t;
172
173
174 /*
175 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept
176 * secret and is suitable for use in MAC algorithms as the key.
177 */
178 typedef struct zio_cksum_salt {
179 uint8_t zcs_bytes[32];
180 } zio_cksum_salt_t;
181
182 /*
183 * Each block is described by its DVAs, time of birth, checksum, etc.
184 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
185 *
186 * 64 56 48 40 32 24 16 8 0
187 * +-------+-------+-------+-------+-------+-------+-------+-------+
188 * 0 | pad | vdev1 | GRID | ASIZE |
189 * +-------+-------+-------+-------+-------+-------+-------+-------+
190 * 1 |G| offset1 |
191 * +-------+-------+-------+-------+-------+-------+-------+-------+
192 * 2 | pad | vdev2 | GRID | ASIZE |
193 * +-------+-------+-------+-------+-------+-------+-------+-------+
194 * 3 |G| offset2 |
195 * +-------+-------+-------+-------+-------+-------+-------+-------+
196 * 4 | pad | vdev3 | GRID | ASIZE |
197 * +-------+-------+-------+-------+-------+-------+-------+-------+
198 * 5 |G| offset3 |
199 * +-------+-------+-------+-------+-------+-------+-------+-------+
200 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE |
201 * +-------+-------+-------+-------+-------+-------+-------+-------+
202 * 7 | padding |
203 * +-------+-------+-------+-------+-------+-------+-------+-------+
204 * 8 | padding |
205 * +-------+-------+-------+-------+-------+-------+-------+-------+
206 * 9 | physical birth txg |
207 * +-------+-------+-------+-------+-------+-------+-------+-------+
208 * a | logical birth txg |
209 * +-------+-------+-------+-------+-------+-------+-------+-------+
210 * b | fill count |
211 * +-------+-------+-------+-------+-------+-------+-------+-------+
212 * c | checksum[0] |
213 * +-------+-------+-------+-------+-------+-------+-------+-------+
214 * d | checksum[1] |
215 * +-------+-------+-------+-------+-------+-------+-------+-------+
216 * e | checksum[2] |
217 * +-------+-------+-------+-------+-------+-------+-------+-------+
218 * f | checksum[3] |
219 * +-------+-------+-------+-------+-------+-------+-------+-------+
220 *
221 * Legend:
222 *
223 * vdev virtual device ID
224 * offset offset into virtual device
225 * LSIZE logical size
226 * PSIZE physical size (after compression)
227 * ASIZE allocated size (including RAID-Z parity and gang block headers)
228 * GRID RAID-Z layout information (reserved for future use)
229 * cksum checksum function
230 * comp compression function
231 * G gang block indicator
232 * B byteorder (endianness)
233 * D dedup
234 * X encryption
235 * E blkptr_t contains embedded data (see below)
236 * lvl level of indirection
237 * type DMU object type
238 * phys birth txg when dva[0] was written; zero if same as logical birth txg
239 * note that typically all the dva's would be written in this
240 * txg, but they could be different if they were moved by
241 * device removal.
242 * log. birth transaction group in which the block was logically born
243 * fill count number of non-zero blocks under this bp
244 * checksum[4] 256-bit checksum of the data this bp describes
245 */
246
247 /*
248 * The blkptr_t's of encrypted blocks also need to store the encryption
249 * parameters so that the block can be decrypted. This layout is as follows:
250 *
251 * 64 56 48 40 32 24 16 8 0
252 * +-------+-------+-------+-------+-------+-------+-------+-------+
253 * 0 | vdev1 | GRID | ASIZE |
254 * +-------+-------+-------+-------+-------+-------+-------+-------+
255 * 1 |G| offset1 |
256 * +-------+-------+-------+-------+-------+-------+-------+-------+
257 * 2 | vdev2 | GRID | ASIZE |
258 * +-------+-------+-------+-------+-------+-------+-------+-------+
259 * 3 |G| offset2 |
260 * +-------+-------+-------+-------+-------+-------+-------+-------+
261 * 4 | salt |
262 * +-------+-------+-------+-------+-------+-------+-------+-------+
263 * 5 | IV1 |
264 * +-------+-------+-------+-------+-------+-------+-------+-------+
265 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE |
266 * +-------+-------+-------+-------+-------+-------+-------+-------+
267 * 7 | padding |
268 * +-------+-------+-------+-------+-------+-------+-------+-------+
269 * 8 | padding |
270 * +-------+-------+-------+-------+-------+-------+-------+-------+
271 * 9 | physical birth txg |
272 * +-------+-------+-------+-------+-------+-------+-------+-------+
273 * a | logical birth txg |
274 * +-------+-------+-------+-------+-------+-------+-------+-------+
275 * b | IV2 | fill count |
276 * +-------+-------+-------+-------+-------+-------+-------+-------+
277 * c | checksum[0] |
278 * +-------+-------+-------+-------+-------+-------+-------+-------+
279 * d | checksum[1] |
280 * +-------+-------+-------+-------+-------+-------+-------+-------+
281 * e | MAC[0] |
282 * +-------+-------+-------+-------+-------+-------+-------+-------+
283 * f | MAC[1] |
284 * +-------+-------+-------+-------+-------+-------+-------+-------+
285 *
286 * Legend:
287 *
288 * salt Salt for generating encryption keys
289 * IV1 First 64 bits of encryption IV
290 * X Block requires encryption handling (set to 1)
291 * E blkptr_t contains embedded data (set to 0, see below)
292 * fill count number of non-zero blocks under this bp (truncated to 32 bits)
293 * IV2 Last 32 bits of encryption IV
294 * checksum[2] 128-bit checksum of the data this bp describes
295 * MAC[2] 128-bit message authentication code for this data
296 *
297 * The X bit being set indicates that this block is one of 3 types. If this is
298 * a level 0 block with an encrypted object type, the block is encrypted
299 * (see BP_IS_ENCRYPTED()). If this is a level 0 block with an unencrypted
300 * object type, this block is authenticated with an HMAC (see
301 * BP_IS_AUTHENTICATED()). Otherwise (if level > 0), this bp will use the MAC
302 * words to store a checksum-of-MACs from the level below (see
303 * BP_HAS_INDIRECT_MAC_CKSUM()). For convenience in the code, BP_IS_PROTECTED()
304 * refers to both encrypted and authenticated blocks and BP_USES_CRYPT()
305 * refers to any of these 3 kinds of blocks.
306 *
307 * The additional encryption parameters are the salt, IV, and MAC which are
308 * explained in greater detail in the block comment at the top of zio_crypt.c.
309 * The MAC occupies half of the checksum space since it serves a very similar
310 * purpose: to prevent data corruption on disk. The only functional difference
311 * is that the checksum is used to detect on-disk corruption whether or not the
312 * encryption key is loaded and the MAC provides additional protection against
313 * malicious disk tampering. We use the 3rd DVA to store the salt and first
314 * 64 bits of the IV. As a result encrypted blocks can only have 2 copies
315 * maximum instead of the normal 3. The last 32 bits of the IV are stored in
316 * the upper bits of what is usually the fill count. Note that only blocks at
317 * level 0 or -2 are ever encrypted, which allows us to guarantee that these
318 * 32 bits are not trampled over by other code (see zio_crypt.c for details).
319 * The salt and IV are not used for authenticated bps or bps with an indirect
320 * MAC checksum, so these blocks can utilize all 3 DVAs and the full 64 bits
321 * for the fill count.
322 */
323
324 /*
325 * "Embedded" blkptr_t's don't actually point to a block, instead they
326 * have a data payload embedded in the blkptr_t itself. See the comment
327 * in blkptr.c for more details.
328 *
329 * The blkptr_t is laid out as follows:
330 *
331 * 64 56 48 40 32 24 16 8 0
332 * +-------+-------+-------+-------+-------+-------+-------+-------+
333 * 0 | payload |
334 * 1 | payload |
335 * 2 | payload |
336 * 3 | payload |
337 * 4 | payload |
338 * 5 | payload |
339 * +-------+-------+-------+-------+-------+-------+-------+-------+
340 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE |
341 * +-------+-------+-------+-------+-------+-------+-------+-------+
342 * 7 | payload |
343 * 8 | payload |
344 * 9 | payload |
345 * +-------+-------+-------+-------+-------+-------+-------+-------+
346 * a | logical birth txg |
347 * +-------+-------+-------+-------+-------+-------+-------+-------+
348 * b | payload |
349 * c | payload |
350 * d | payload |
351 * e | payload |
352 * f | payload |
353 * +-------+-------+-------+-------+-------+-------+-------+-------+
354 *
355 * Legend:
356 *
357 * payload contains the embedded data
358 * B (byteorder) byteorder (endianness)
359 * D (dedup) padding (set to zero)
360 * X encryption (set to zero)
361 * E (embedded) set to one
362 * lvl indirection level
363 * type DMU object type
364 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*)
365 * comp compression function of payload
366 * PSIZE size of payload after compression, in bytes
367 * LSIZE logical size of payload, in bytes
368 * note that 25 bits is enough to store the largest
369 * "normal" BP's LSIZE (2^16 * 2^9) in bytes
370 * log. birth transaction group in which the block was logically born
371 *
372 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
373 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
374 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
375 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
376 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must
377 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before
378 * other macros, as they assert that they are only used on BP's of the correct
379 * "embedded-ness". Encrypted blkptr_t's cannot be embedded because they use
380 * the payload space for encryption parameters (see the comment above on
381 * how encryption parameters are stored).
382 */
383
384 #define BPE_GET_ETYPE(bp) \
385 (ASSERT(BP_IS_EMBEDDED(bp)), \
386 BF64_GET((bp)->blk_prop, 40, 8))
387 #define BPE_SET_ETYPE(bp, t) do { \
388 ASSERT(BP_IS_EMBEDDED(bp)); \
389 BF64_SET((bp)->blk_prop, 40, 8, t); \
390 _NOTE(CONSTCOND) } while (0)
391
392 #define BPE_GET_LSIZE(bp) \
393 (ASSERT(BP_IS_EMBEDDED(bp)), \
394 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
395 #define BPE_SET_LSIZE(bp, x) do { \
396 ASSERT(BP_IS_EMBEDDED(bp)); \
397 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
398 _NOTE(CONSTCOND) } while (0)
399
400 #define BPE_GET_PSIZE(bp) \
401 (ASSERT(BP_IS_EMBEDDED(bp)), \
402 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
403 #define BPE_SET_PSIZE(bp, x) do { \
404 ASSERT(BP_IS_EMBEDDED(bp)); \
405 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
406 _NOTE(CONSTCOND) } while (0)
407
408 typedef enum bp_embedded_type {
409 BP_EMBEDDED_TYPE_DATA,
410 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for Delphix byteswap feature. */
411 BP_EMBEDDED_TYPE_REDACTED,
412 NUM_BP_EMBEDDED_TYPES
413 } bp_embedded_type_t;
414
415 #define BPE_NUM_WORDS 14
416 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
417 #define BPE_IS_PAYLOADWORD(bp, wp) \
418 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
419
420 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */
421 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */
422 #define SPA_SYNC_MIN_VDEVS 3 /* min vdevs to update during sync */
423
424 /*
425 * A block is a hole when it has either 1) never been written to, or
426 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
427 * without physically allocating disk space. Holes are represented in the
428 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is
429 * done through the BP_IS_HOLE macro. For holes, the logical size, level,
430 * DMU object type, and birth times are all also stored for holes that
431 * were written to at some point (i.e. were punched after having been filled).
432 */
433 typedef struct blkptr {
434 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
435 uint64_t blk_prop; /* size, compression, type, etc */
436 uint64_t blk_pad[2]; /* Extra space for the future */
437 uint64_t blk_phys_birth; /* txg when block was allocated */
438 uint64_t blk_birth; /* transaction group at birth */
439 uint64_t blk_fill; /* fill count */
440 zio_cksum_t blk_cksum; /* 256-bit checksum */
441 } blkptr_t;
442
443 /*
444 * Macros to get and set fields in a bp or DVA.
445 */
446 #define DVA_GET_ASIZE(dva) \
447 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
448 #define DVA_SET_ASIZE(dva, x) \
449 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
450 SPA_MINBLOCKSHIFT, 0, x)
451
452 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8)
453 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x)
454
455 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, SPA_VDEVBITS)
456 #define DVA_SET_VDEV(dva, x) \
457 BF64_SET((dva)->dva_word[0], 32, SPA_VDEVBITS, x)
458
459 #define DVA_GET_OFFSET(dva) \
460 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
461 #define DVA_SET_OFFSET(dva, x) \
462 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
463
464 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1)
465 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x)
466
467 #define BP_GET_LSIZE(bp) \
468 (BP_IS_EMBEDDED(bp) ? \
469 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
470 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
471 #define BP_SET_LSIZE(bp, x) do { \
472 ASSERT(!BP_IS_EMBEDDED(bp)); \
473 BF64_SET_SB((bp)->blk_prop, \
474 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
475 _NOTE(CONSTCOND) } while (0)
476
477 #define BP_GET_PSIZE(bp) \
478 (BP_IS_EMBEDDED(bp) ? 0 : \
479 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1))
480 #define BP_SET_PSIZE(bp, x) do { \
481 ASSERT(!BP_IS_EMBEDDED(bp)); \
482 BF64_SET_SB((bp)->blk_prop, \
483 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
484 _NOTE(CONSTCOND) } while (0)
485
486 #define BP_GET_COMPRESS(bp) \
487 BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS)
488 #define BP_SET_COMPRESS(bp, x) \
489 BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x)
490
491 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1)
492 #define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x)
493
494 #define BP_GET_CHECKSUM(bp) \
495 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \
496 BF64_GET((bp)->blk_prop, 40, 8))
497 #define BP_SET_CHECKSUM(bp, x) do { \
498 ASSERT(!BP_IS_EMBEDDED(bp)); \
499 BF64_SET((bp)->blk_prop, 40, 8, x); \
500 _NOTE(CONSTCOND) } while (0)
501
502 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8)
503 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x)
504
505 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5)
506 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x)
507
508 /* encrypted, authenticated, and MAC cksum bps use the same bit */
509 #define BP_USES_CRYPT(bp) BF64_GET((bp)->blk_prop, 61, 1)
510 #define BP_SET_CRYPT(bp, x) BF64_SET((bp)->blk_prop, 61, 1, x)
511
512 #define BP_IS_ENCRYPTED(bp) \
513 (BP_USES_CRYPT(bp) && \
514 BP_GET_LEVEL(bp) <= 0 && \
515 DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp)))
516
517 #define BP_IS_AUTHENTICATED(bp) \
518 (BP_USES_CRYPT(bp) && \
519 BP_GET_LEVEL(bp) <= 0 && \
520 !DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp)))
521
522 #define BP_HAS_INDIRECT_MAC_CKSUM(bp) \
523 (BP_USES_CRYPT(bp) && BP_GET_LEVEL(bp) > 0)
524
525 #define BP_IS_PROTECTED(bp) \
526 (BP_IS_ENCRYPTED(bp) || BP_IS_AUTHENTICATED(bp))
527
528 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1)
529 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x)
530
531 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1)
532 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x)
533
534 #define BP_PHYSICAL_BIRTH(bp) \
535 (BP_IS_EMBEDDED(bp) ? 0 : \
536 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
537
538 #define BP_SET_BIRTH(bp, logical, physical) \
539 { \
540 ASSERT(!BP_IS_EMBEDDED(bp)); \
541 (bp)->blk_birth = (logical); \
542 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
543 }
544
545 #define BP_GET_FILL(bp) \
546 ((BP_IS_ENCRYPTED(bp)) ? BF64_GET((bp)->blk_fill, 0, 32) : \
547 ((BP_IS_EMBEDDED(bp)) ? 1 : (bp)->blk_fill))
548
549 #define BP_SET_FILL(bp, fill) \
550 { \
551 if (BP_IS_ENCRYPTED(bp)) \
552 BF64_SET((bp)->blk_fill, 0, 32, fill); \
553 else \
554 (bp)->blk_fill = fill; \
555 }
556
557 #define BP_GET_IV2(bp) \
558 (ASSERT(BP_IS_ENCRYPTED(bp)), \
559 BF64_GET((bp)->blk_fill, 32, 32))
560 #define BP_SET_IV2(bp, iv2) \
561 { \
562 ASSERT(BP_IS_ENCRYPTED(bp)); \
563 BF64_SET((bp)->blk_fill, 32, 32, iv2); \
564 }
565
566 #define BP_IS_METADATA(bp) \
567 (BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))
568
569 #define BP_GET_ASIZE(bp) \
570 (BP_IS_EMBEDDED(bp) ? 0 : \
571 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
572 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
573 (DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))
574
575 #define BP_GET_UCSIZE(bp) \
576 (BP_IS_METADATA(bp) ? BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
577
578 #define BP_GET_NDVAS(bp) \
579 (BP_IS_EMBEDDED(bp) ? 0 : \
580 !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
581 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
582 (!!DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))
583
584 #define BP_COUNT_GANG(bp) \
585 (BP_IS_EMBEDDED(bp) ? 0 : \
586 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \
587 DVA_GET_GANG(&(bp)->blk_dva[1]) + \
588 (DVA_GET_GANG(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp))))
589
590 #define DVA_EQUAL(dva1, dva2) \
591 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
592 (dva1)->dva_word[0] == (dva2)->dva_word[0])
593
594 #define BP_EQUAL(bp1, bp2) \
595 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \
596 (bp1)->blk_birth == (bp2)->blk_birth && \
597 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \
598 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \
599 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
600
601
602 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0)
603
604 #define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0])
605 #define BP_IS_GANG(bp) \
606 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp)))
607 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \
608 (dva)->dva_word[1] == 0ULL)
609 #define BP_IS_HOLE(bp) \
610 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp)))
611
612 #define BP_SET_REDACTED(bp) \
613 { \
614 BP_SET_EMBEDDED(bp, B_TRUE); \
615 BPE_SET_ETYPE(bp, BP_EMBEDDED_TYPE_REDACTED); \
616 }
617 #define BP_IS_REDACTED(bp) \
618 (BP_IS_EMBEDDED(bp) && BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_REDACTED)
619
620 /* BP_IS_RAIDZ(bp) assumes no block compression */
621 #define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
622 BP_GET_PSIZE(bp))
623
624 #define BP_ZERO(bp) \
625 { \
626 (bp)->blk_dva[0].dva_word[0] = 0; \
627 (bp)->blk_dva[0].dva_word[1] = 0; \
628 (bp)->blk_dva[1].dva_word[0] = 0; \
629 (bp)->blk_dva[1].dva_word[1] = 0; \
630 (bp)->blk_dva[2].dva_word[0] = 0; \
631 (bp)->blk_dva[2].dva_word[1] = 0; \
632 (bp)->blk_prop = 0; \
633 (bp)->blk_pad[0] = 0; \
634 (bp)->blk_pad[1] = 0; \
635 (bp)->blk_phys_birth = 0; \
636 (bp)->blk_birth = 0; \
637 (bp)->blk_fill = 0; \
638 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \
639 }
640
641 #ifdef _BIG_ENDIAN
642 #define ZFS_HOST_BYTEORDER (0ULL)
643 #else
644 #define ZFS_HOST_BYTEORDER (1ULL)
645 #endif
646
647 #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
648
649 #define BP_SPRINTF_LEN 400
650
651 /*
652 * This macro allows code sharing between zfs, libzpool, and mdb.
653 * 'func' is either snprintf() or mdb_snprintf().
654 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
655 */
656 #define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
657 { \
658 static const char *copyname[] = \
659 { "zero", "single", "double", "triple" }; \
660 int len = 0; \
661 int copies = 0; \
662 const char *crypt_type; \
663 if (bp != NULL) { \
664 if (BP_IS_ENCRYPTED(bp)) { \
665 crypt_type = "encrypted"; \
666 /* LINTED E_SUSPICIOUS_COMPARISON */ \
667 } else if (BP_IS_AUTHENTICATED(bp)) { \
668 crypt_type = "authenticated"; \
669 } else if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) { \
670 crypt_type = "indirect-MAC"; \
671 } else { \
672 crypt_type = "unencrypted"; \
673 } \
674 } \
675 if (bp == NULL) { \
676 len += func(buf + len, size - len, "<NULL>"); \
677 } else if (BP_IS_HOLE(bp)) { \
678 len += func(buf + len, size - len, \
679 "HOLE [L%llu %s] " \
680 "size=%llxL birth=%lluL", \
681 (u_longlong_t)BP_GET_LEVEL(bp), \
682 type, \
683 (u_longlong_t)BP_GET_LSIZE(bp), \
684 (u_longlong_t)bp->blk_birth); \
685 } else if (BP_IS_EMBEDDED(bp)) { \
686 len = func(buf + len, size - len, \
687 "EMBEDDED [L%llu %s] et=%u %s " \
688 "size=%llxL/%llxP birth=%lluL", \
689 (u_longlong_t)BP_GET_LEVEL(bp), \
690 type, \
691 (int)BPE_GET_ETYPE(bp), \
692 compress, \
693 (u_longlong_t)BPE_GET_LSIZE(bp), \
694 (u_longlong_t)BPE_GET_PSIZE(bp), \
695 (u_longlong_t)bp->blk_birth); \
696 } else if (BP_IS_REDACTED(bp)) { \
697 len += func(buf + len, size - len, \
698 "REDACTED [L%llu %s] size=%llxL birth=%lluL", \
699 (u_longlong_t)BP_GET_LEVEL(bp), \
700 type, \
701 (u_longlong_t)BP_GET_LSIZE(bp), \
702 (u_longlong_t)bp->blk_birth); \
703 } else { \
704 for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \
705 const dva_t *dva = &bp->blk_dva[d]; \
706 if (DVA_IS_VALID(dva)) \
707 copies++; \
708 len += func(buf + len, size - len, \
709 "DVA[%d]=<%llu:%llx:%llx>%c", d, \
710 (u_longlong_t)DVA_GET_VDEV(dva), \
711 (u_longlong_t)DVA_GET_OFFSET(dva), \
712 (u_longlong_t)DVA_GET_ASIZE(dva), \
713 ws); \
714 } \
715 if (BP_IS_ENCRYPTED(bp)) { \
716 len += func(buf + len, size - len, \
717 "salt=%llx iv=%llx:%llx%c", \
718 (u_longlong_t)bp->blk_dva[2].dva_word[0], \
719 (u_longlong_t)bp->blk_dva[2].dva_word[1], \
720 (u_longlong_t)BP_GET_IV2(bp), \
721 ws); \
722 } \
723 if (BP_IS_GANG(bp) && \
724 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \
725 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \
726 copies--; \
727 len += func(buf + len, size - len, \
728 "[L%llu %s] %s %s %s %s %s %s %s%c" \
729 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \
730 "cksum=%llx:%llx:%llx:%llx", \
731 (u_longlong_t)BP_GET_LEVEL(bp), \
732 type, \
733 checksum, \
734 compress, \
735 crypt_type, \
736 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \
737 BP_IS_GANG(bp) ? "gang" : "contiguous", \
738 BP_GET_DEDUP(bp) ? "dedup" : "unique", \
739 copyname[copies], \
740 ws, \
741 (u_longlong_t)BP_GET_LSIZE(bp), \
742 (u_longlong_t)BP_GET_PSIZE(bp), \
743 (u_longlong_t)bp->blk_birth, \
744 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \
745 (u_longlong_t)BP_GET_FILL(bp), \
746 ws, \
747 (u_longlong_t)bp->blk_cksum.zc_word[0], \
748 (u_longlong_t)bp->blk_cksum.zc_word[1], \
749 (u_longlong_t)bp->blk_cksum.zc_word[2], \
750 (u_longlong_t)bp->blk_cksum.zc_word[3]); \
751 } \
752 ASSERT(len < size); \
753 }
754
755 #define BP_GET_BUFC_TYPE(bp) \
756 (BP_IS_METADATA(bp) ? ARC_BUFC_METADATA : ARC_BUFC_DATA)
757
758 typedef enum spa_import_type {
759 SPA_IMPORT_EXISTING,
760 SPA_IMPORT_ASSEMBLE
761 } spa_import_type_t;
762
763 /*
764 * Send TRIM commands in-line during normal pool operation while deleting.
765 * OFF: no
766 * ON: yes
767 */
768 typedef enum {
769 SPA_AUTOTRIM_OFF = 0, /* default */
770 SPA_AUTOTRIM_ON
771 } spa_autotrim_t;
772
773 /*
774 * Reason TRIM command was issued, used internally for accounting purposes.
775 */
776 typedef enum trim_type {
777 TRIM_TYPE_MANUAL = 0,
778 TRIM_TYPE_AUTO = 1,
779 } trim_type_t;
780
781 /* state manipulation functions */
782 extern int spa_open(const char *pool, spa_t **, void *tag);
783 extern int spa_open_rewind(const char *pool, spa_t **, void *tag,
784 nvlist_t *policy, nvlist_t **config);
785 extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot,
786 size_t buflen);
787 extern int spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props,
788 nvlist_t *zplprops, struct dsl_crypto_params *dcp);
789 extern int spa_import(char *pool, nvlist_t *config, nvlist_t *props,
790 uint64_t flags);
791 extern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
792 extern int spa_destroy(char *pool);
793 extern int spa_checkpoint(const char *pool);
794 extern int spa_checkpoint_discard(const char *pool);
795 extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force,
796 boolean_t hardforce);
797 extern int spa_reset(char *pool);
798 extern void spa_async_request(spa_t *spa, int flag);
799 extern void spa_async_unrequest(spa_t *spa, int flag);
800 extern void spa_async_suspend(spa_t *spa);
801 extern void spa_async_resume(spa_t *spa);
802 extern spa_t *spa_inject_addref(char *pool);
803 extern void spa_inject_delref(spa_t *spa);
804 extern void spa_scan_stat_init(spa_t *spa);
805 extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps);
806
807 #define SPA_ASYNC_CONFIG_UPDATE 0x01
808 #define SPA_ASYNC_REMOVE 0x02
809 #define SPA_ASYNC_PROBE 0x04
810 #define SPA_ASYNC_RESILVER_DONE 0x08
811 #define SPA_ASYNC_RESILVER 0x10
812 #define SPA_ASYNC_AUTOEXPAND 0x20
813 #define SPA_ASYNC_REMOVE_DONE 0x40
814 #define SPA_ASYNC_REMOVE_STOP 0x80
815 #define SPA_ASYNC_INITIALIZE_RESTART 0x100
816 #define SPA_ASYNC_TRIM_RESTART 0x200
817 #define SPA_ASYNC_AUTOTRIM_RESTART 0x400
818
819 /*
820 * Controls the behavior of spa_vdev_remove().
821 */
822 #define SPA_REMOVE_UNSPARE 0x01
823 #define SPA_REMOVE_DONE 0x02
824
825 /* device manipulation */
826 extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
827 extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot,
828 int replacing);
829 extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
830 int replace_done);
831 extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare);
832 extern boolean_t spa_vdev_remove_active(spa_t *spa);
833 extern int spa_vdev_initialize(spa_t *spa, nvlist_t *nv, uint64_t cmd_type,
834 nvlist_t *vdev_errlist);
835 extern int spa_vdev_trim(spa_t *spa, nvlist_t *nv, uint64_t cmd_type,
836 uint64_t rate, boolean_t partial, boolean_t secure, nvlist_t *vdev_errlist);
837 extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
838 extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru);
839 extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
840 nvlist_t *props, boolean_t exp);
841
842 /* spare state (which is global across all pools) */
843 extern void spa_spare_add(vdev_t *vd);
844 extern void spa_spare_remove(vdev_t *vd);
845 extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt);
846 extern void spa_spare_activate(vdev_t *vd);
847
848 /* L2ARC state (which is global across all pools) */
849 extern void spa_l2cache_add(vdev_t *vd);
850 extern void spa_l2cache_remove(vdev_t *vd);
851 extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
852 extern void spa_l2cache_activate(vdev_t *vd);
853 extern void spa_l2cache_drop(spa_t *spa);
854
855 /* scanning */
856 extern int spa_scan(spa_t *spa, pool_scan_func_t func);
857 extern int spa_scan_stop(spa_t *spa);
858 extern int spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t flag);
859
860 /* spa syncing */
861 extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
862 extern void spa_sync_allpools(void);
863
864 extern int zfs_sync_pass_deferred_free;
865
866 /* spa namespace global mutex */
867 extern kmutex_t spa_namespace_lock;
868
869 /*
870 * SPA configuration functions in spa_config.c
871 */
872
873 #define SPA_CONFIG_UPDATE_POOL 0
874 #define SPA_CONFIG_UPDATE_VDEVS 1
875
876 extern void spa_write_cachefile(spa_t *, boolean_t, boolean_t);
877 extern void spa_config_load(void);
878 extern nvlist_t *spa_all_configs(uint64_t *);
879 extern void spa_config_set(spa_t *spa, nvlist_t *config);
880 extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
881 int getstats);
882 extern void spa_config_update(spa_t *spa, int what);
883
884 /*
885 * Miscellaneous SPA routines in spa_misc.c
886 */
887
888 /* Namespace manipulation */
889 extern spa_t *spa_lookup(const char *name);
890 extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot);
891 extern void spa_remove(spa_t *spa);
892 extern spa_t *spa_next(spa_t *prev);
893
894 /* Refcount functions */
895 extern void spa_open_ref(spa_t *spa, void *tag);
896 extern void spa_close(spa_t *spa, void *tag);
897 extern void spa_async_close(spa_t *spa, void *tag);
898 extern boolean_t spa_refcount_zero(spa_t *spa);
899
900 #define SCL_NONE 0x00
901 #define SCL_CONFIG 0x01
902 #define SCL_STATE 0x02
903 #define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */
904 #define SCL_ALLOC 0x08
905 #define SCL_ZIO 0x10
906 #define SCL_FREE 0x20
907 #define SCL_VDEV 0x40
908 #define SCL_LOCKS 7
909 #define SCL_ALL ((1 << SCL_LOCKS) - 1)
910 #define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO)
911
912 /* Historical pool statistics */
913 typedef struct spa_history_kstat {
914 kmutex_t lock;
915 uint64_t count;
916 uint64_t size;
917 kstat_t *kstat;
918 void *private;
919 list_t list;
920 } spa_history_kstat_t;
921
922 typedef struct spa_history_list {
923 uint64_t size;
924 procfs_list_t procfs_list;
925 } spa_history_list_t;
926
927 typedef struct spa_stats {
928 spa_history_list_t read_history;
929 spa_history_list_t txg_history;
930 spa_history_kstat_t tx_assign_histogram;
931 spa_history_kstat_t io_history;
932 spa_history_list_t mmp_history;
933 spa_history_kstat_t state; /* pool state */
934 spa_history_kstat_t iostats;
935 } spa_stats_t;
936
937 typedef enum txg_state {
938 TXG_STATE_BIRTH = 0,
939 TXG_STATE_OPEN = 1,
940 TXG_STATE_QUIESCED = 2,
941 TXG_STATE_WAIT_FOR_SYNC = 3,
942 TXG_STATE_SYNCED = 4,
943 TXG_STATE_COMMITTED = 5,
944 } txg_state_t;
945
946 typedef struct txg_stat {
947 vdev_stat_t vs1;
948 vdev_stat_t vs2;
949 uint64_t txg;
950 uint64_t ndirty;
951 } txg_stat_t;
952
953 /* Assorted pool IO kstats */
954 typedef struct spa_iostats {
955 kstat_named_t trim_extents_written;
956 kstat_named_t trim_bytes_written;
957 kstat_named_t trim_extents_skipped;
958 kstat_named_t trim_bytes_skipped;
959 kstat_named_t trim_extents_failed;
960 kstat_named_t trim_bytes_failed;
961 kstat_named_t autotrim_extents_written;
962 kstat_named_t autotrim_bytes_written;
963 kstat_named_t autotrim_extents_skipped;
964 kstat_named_t autotrim_bytes_skipped;
965 kstat_named_t autotrim_extents_failed;
966 kstat_named_t autotrim_bytes_failed;
967 } spa_iostats_t;
968
969 extern void spa_stats_init(spa_t *spa);
970 extern void spa_stats_destroy(spa_t *spa);
971 extern void spa_read_history_add(spa_t *spa, const zbookmark_phys_t *zb,
972 uint32_t aflags);
973 extern void spa_txg_history_add(spa_t *spa, uint64_t txg, hrtime_t birth_time);
974 extern int spa_txg_history_set(spa_t *spa, uint64_t txg,
975 txg_state_t completed_state, hrtime_t completed_time);
976 extern txg_stat_t *spa_txg_history_init_io(spa_t *, uint64_t,
977 struct dsl_pool *);
978 extern void spa_txg_history_fini_io(spa_t *, txg_stat_t *);
979 extern void spa_tx_assign_add_nsecs(spa_t *spa, uint64_t nsecs);
980 extern int spa_mmp_history_set_skip(spa_t *spa, uint64_t mmp_kstat_id);
981 extern int spa_mmp_history_set(spa_t *spa, uint64_t mmp_kstat_id, int io_error,
982 hrtime_t duration);
983 extern void spa_mmp_history_add(spa_t *spa, uint64_t txg, uint64_t timestamp,
984 uint64_t mmp_delay, vdev_t *vd, int label, uint64_t mmp_kstat_id,
985 int error);
986 extern void spa_iostats_trim_add(spa_t *spa, trim_type_t type,
987 uint64_t extents_written, uint64_t bytes_written,
988 uint64_t extents_skipped, uint64_t bytes_skipped,
989 uint64_t extents_failed, uint64_t bytes_failed);
990 extern void spa_import_progress_add(spa_t *spa);
991 extern void spa_import_progress_remove(uint64_t spa_guid);
992 extern int spa_import_progress_set_mmp_check(uint64_t pool_guid,
993 uint64_t mmp_sec_remaining);
994 extern int spa_import_progress_set_max_txg(uint64_t pool_guid,
995 uint64_t max_txg);
996 extern int spa_import_progress_set_state(uint64_t pool_guid,
997 spa_load_state_t spa_load_state);
998
999 /* Pool configuration locks */
1000 extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw);
1001 extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw);
1002 extern void spa_config_exit(spa_t *spa, int locks, void *tag);
1003 extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
1004
1005 /* Pool vdev add/remove lock */
1006 extern uint64_t spa_vdev_enter(spa_t *spa);
1007 extern uint64_t spa_vdev_config_enter(spa_t *spa);
1008 extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg,
1009 int error, char *tag);
1010 extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error);
1011
1012 /* Pool vdev state change lock */
1013 extern void spa_vdev_state_enter(spa_t *spa, int oplock);
1014 extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error);
1015
1016 /* Log state */
1017 typedef enum spa_log_state {
1018 SPA_LOG_UNKNOWN = 0, /* unknown log state */
1019 SPA_LOG_MISSING, /* missing log(s) */
1020 SPA_LOG_CLEAR, /* clear the log(s) */
1021 SPA_LOG_GOOD, /* log(s) are good */
1022 } spa_log_state_t;
1023
1024 extern spa_log_state_t spa_get_log_state(spa_t *spa);
1025 extern void spa_set_log_state(spa_t *spa, spa_log_state_t state);
1026 extern int spa_reset_logs(spa_t *spa);
1027
1028 /* Log claim callback */
1029 extern void spa_claim_notify(zio_t *zio);
1030 extern void spa_deadman(void *);
1031
1032 /* Accessor functions */
1033 extern boolean_t spa_shutting_down(spa_t *spa);
1034 extern struct dsl_pool *spa_get_dsl(spa_t *spa);
1035 extern boolean_t spa_is_initializing(spa_t *spa);
1036 extern boolean_t spa_indirect_vdevs_loaded(spa_t *spa);
1037 extern blkptr_t *spa_get_rootblkptr(spa_t *spa);
1038 extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
1039 extern void spa_altroot(spa_t *, char *, size_t);
1040 extern int spa_sync_pass(spa_t *spa);
1041 extern char *spa_name(spa_t *spa);
1042 extern uint64_t spa_guid(spa_t *spa);
1043 extern uint64_t spa_load_guid(spa_t *spa);
1044 extern uint64_t spa_last_synced_txg(spa_t *spa);
1045 extern uint64_t spa_first_txg(spa_t *spa);
1046 extern uint64_t spa_syncing_txg(spa_t *spa);
1047 extern uint64_t spa_final_dirty_txg(spa_t *spa);
1048 extern uint64_t spa_version(spa_t *spa);
1049 extern pool_state_t spa_state(spa_t *spa);
1050 extern spa_load_state_t spa_load_state(spa_t *spa);
1051 extern uint64_t spa_freeze_txg(spa_t *spa);
1052 extern uint64_t spa_get_worst_case_asize(spa_t *spa, uint64_t lsize);
1053 extern uint64_t spa_get_dspace(spa_t *spa);
1054 extern uint64_t spa_get_checkpoint_space(spa_t *spa);
1055 extern uint64_t spa_get_slop_space(spa_t *spa);
1056 extern void spa_update_dspace(spa_t *spa);
1057 extern uint64_t spa_version(spa_t *spa);
1058 extern boolean_t spa_deflate(spa_t *spa);
1059 extern metaslab_class_t *spa_normal_class(spa_t *spa);
1060 extern metaslab_class_t *spa_log_class(spa_t *spa);
1061 extern metaslab_class_t *spa_special_class(spa_t *spa);
1062 extern metaslab_class_t *spa_dedup_class(spa_t *spa);
1063 extern metaslab_class_t *spa_preferred_class(spa_t *spa, uint64_t size,
1064 dmu_object_type_t objtype, uint_t level, uint_t special_smallblk);
1065
1066 extern void spa_evicting_os_register(spa_t *, objset_t *os);
1067 extern void spa_evicting_os_deregister(spa_t *, objset_t *os);
1068 extern void spa_evicting_os_wait(spa_t *spa);
1069 extern int spa_max_replication(spa_t *spa);
1070 extern int spa_prev_software_version(spa_t *spa);
1071 extern uint64_t spa_get_failmode(spa_t *spa);
1072 extern uint64_t spa_get_deadman_failmode(spa_t *spa);
1073 extern void spa_set_deadman_failmode(spa_t *spa, const char *failmode);
1074 extern boolean_t spa_suspended(spa_t *spa);
1075 extern uint64_t spa_bootfs(spa_t *spa);
1076 extern uint64_t spa_delegation(spa_t *spa);
1077 extern objset_t *spa_meta_objset(spa_t *spa);
1078 extern uint64_t spa_deadman_synctime(spa_t *spa);
1079 extern uint64_t spa_deadman_ziotime(spa_t *spa);
1080 extern uint64_t spa_dirty_data(spa_t *spa);
1081 extern spa_autotrim_t spa_get_autotrim(spa_t *spa);
1082
1083 /* Miscellaneous support routines */
1084 extern void spa_load_failed(spa_t *spa, const char *fmt, ...);
1085 extern void spa_load_note(spa_t *spa, const char *fmt, ...);
1086 extern void spa_activate_mos_feature(spa_t *spa, const char *feature,
1087 dmu_tx_t *tx);
1088 extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature);
1089 extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid);
1090 extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid);
1091 extern char *spa_strdup(const char *);
1092 extern void spa_strfree(char *);
1093 extern uint64_t spa_get_random(uint64_t range);
1094 extern uint64_t spa_generate_guid(spa_t *spa);
1095 extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp);
1096 extern void spa_freeze(spa_t *spa);
1097 extern int spa_change_guid(spa_t *spa);
1098 extern void spa_upgrade(spa_t *spa, uint64_t version);
1099 extern void spa_evict_all(void);
1100 extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid,
1101 boolean_t l2cache);
1102 extern boolean_t spa_has_spare(spa_t *, uint64_t guid);
1103 extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva);
1104 extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp);
1105 extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp);
1106 extern boolean_t spa_has_slogs(spa_t *spa);
1107 extern boolean_t spa_is_root(spa_t *spa);
1108 extern boolean_t spa_writeable(spa_t *spa);
1109 extern boolean_t spa_has_pending_synctask(spa_t *spa);
1110 extern int spa_maxblocksize(spa_t *spa);
1111 extern int spa_maxdnodesize(spa_t *spa);
1112 extern boolean_t spa_has_checkpoint(spa_t *spa);
1113 extern boolean_t spa_importing_readonly_checkpoint(spa_t *spa);
1114 extern boolean_t spa_suspend_async_destroy(spa_t *spa);
1115 extern uint64_t spa_min_claim_txg(spa_t *spa);
1116 extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp);
1117 extern boolean_t zfs_dva_valid(spa_t *spa, const dva_t *dva,
1118 const blkptr_t *bp);
1119 typedef void (*spa_remap_cb_t)(uint64_t vdev, uint64_t offset, uint64_t size,
1120 void *arg);
1121 extern boolean_t spa_remap_blkptr(spa_t *spa, blkptr_t *bp,
1122 spa_remap_cb_t callback, void *arg);
1123 extern uint64_t spa_get_last_removal_txg(spa_t *spa);
1124 extern boolean_t spa_trust_config(spa_t *spa);
1125 extern uint64_t spa_missing_tvds_allowed(spa_t *spa);
1126 extern void spa_set_missing_tvds(spa_t *spa, uint64_t missing);
1127 extern boolean_t spa_top_vdevs_spacemap_addressable(spa_t *spa);
1128 extern boolean_t spa_multihost(spa_t *spa);
1129 extern unsigned long spa_get_hostid(void);
1130 extern void spa_activate_allocation_classes(spa_t *, dmu_tx_t *);
1131
1132 extern int spa_mode(spa_t *spa);
1133 extern uint64_t zfs_strtonum(const char *str, char **nptr);
1134
1135 extern char *spa_his_ievent_table[];
1136
1137 extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx);
1138 extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read,
1139 char *his_buf);
1140 extern int spa_history_log(spa_t *spa, const char *his_buf);
1141 extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl);
1142 extern void spa_history_log_version(spa_t *spa, const char *operation,
1143 dmu_tx_t *tx);
1144 extern void spa_history_log_internal(spa_t *spa, const char *operation,
1145 dmu_tx_t *tx, const char *fmt, ...);
1146 extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op,
1147 dmu_tx_t *tx, const char *fmt, ...);
1148 extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
1149 dmu_tx_t *tx, const char *fmt, ...);
1150
1151 extern const char *spa_state_to_name(spa_t *spa);
1152
1153 /* error handling */
1154 struct zbookmark_phys;
1155 extern void spa_log_error(spa_t *spa, const zbookmark_phys_t *zb);
1156 extern int zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd,
1157 const zbookmark_phys_t *zb, zio_t *zio, uint64_t stateoroffset,
1158 uint64_t length);
1159 extern boolean_t zfs_ereport_is_valid(const char *class, spa_t *spa, vdev_t *vd,
1160 zio_t *zio);
1161 extern nvlist_t *zfs_event_create(spa_t *spa, vdev_t *vd, const char *type,
1162 const char *name, nvlist_t *aux);
1163 extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
1164 extern void zfs_post_state_change(spa_t *spa, vdev_t *vd, uint64_t laststate);
1165 extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
1166 extern uint64_t spa_get_errlog_size(spa_t *spa);
1167 extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
1168 extern void spa_errlog_rotate(spa_t *spa);
1169 extern void spa_errlog_drain(spa_t *spa);
1170 extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
1171 extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
1172
1173 /* vdev cache */
1174 extern void vdev_cache_stat_init(void);
1175 extern void vdev_cache_stat_fini(void);
1176
1177 /* vdev mirror */
1178 extern void vdev_mirror_stat_init(void);
1179 extern void vdev_mirror_stat_fini(void);
1180
1181 /* Initialization and termination */
1182 extern void spa_init(int flags);
1183 extern void spa_fini(void);
1184 extern void spa_boot_init(void);
1185
1186 /* properties */
1187 extern int spa_prop_set(spa_t *spa, nvlist_t *nvp);
1188 extern int spa_prop_get(spa_t *spa, nvlist_t **nvp);
1189 extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx);
1190 extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t);
1191
1192 /* asynchronous event notification */
1193 extern void spa_event_notify(spa_t *spa, vdev_t *vdev, nvlist_t *hist_nvl,
1194 const char *name);
1195
1196 #ifdef ZFS_DEBUG
1197 #define dprintf_bp(bp, fmt, ...) do { \
1198 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
1199 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \
1200 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \
1201 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \
1202 kmem_free(__blkbuf, BP_SPRINTF_LEN); \
1203 } \
1204 _NOTE(CONSTCOND) } while (0)
1205 #else
1206 #define dprintf_bp(bp, fmt, ...)
1207 #endif
1208
1209 extern int spa_mode_global; /* mode, e.g. FREAD | FWRITE */
1210 extern int zfs_deadman_enabled;
1211 extern unsigned long zfs_deadman_synctime_ms;
1212 extern unsigned long zfs_deadman_ziotime_ms;
1213 extern unsigned long zfs_deadman_checktime_ms;
1214
1215 #ifdef __cplusplus
1216 }
1217 #endif
1218
1219 #endif /* _SYS_SPA_H */
ZFS On Linux mirror for PVE