]> 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
Log Spacemap Project
[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, 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.
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 #include <sys/space_map.h>
46
47 #ifdef __cplusplus
48 extern "C" {
49 #endif
50
51 /*
52 * Forward references that lots of things need.
53 */
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;
65
66 struct dsl_pool;
67 struct dsl_dataset;
68 struct dsl_crypto_params;
69
70 /*
71 * General-purpose 32-bit and 64-bit bitfield encodings.
72 */
73 #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len))
74 #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len))
75 #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low))
76 #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low))
77
78 #define BF32_GET(x, low, len) BF32_DECODE(x, low, len)
79 #define BF64_GET(x, low, len) BF64_DECODE(x, low, len)
80
81 #define BF32_SET(x, low, len, val) do { \
82 ASSERT3U(val, <, 1U << (len)); \
83 ASSERT3U(low + len, <=, 32); \
84 (x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \
85 _NOTE(CONSTCOND) } while (0)
86
87 #define BF64_SET(x, low, len, val) do { \
88 ASSERT3U(val, <, 1ULL << (len)); \
89 ASSERT3U(low + len, <=, 64); \
90 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \
91 _NOTE(CONSTCOND) } while (0)
92
93 #define BF32_GET_SB(x, low, len, shift, bias) \
94 ((BF32_GET(x, low, len) + (bias)) << (shift))
95 #define BF64_GET_SB(x, low, len, shift, bias) \
96 ((BF64_GET(x, low, len) + (bias)) << (shift))
97
98 /*
99 * We use ASSERT3U instead of ASSERT in these macros to prevent a lint error in
100 * the case where val is a constant. We can't fix ASSERT because it's used as
101 * an expression in several places in the kernel; as a result, changing it to
102 * the do{} while() syntax to allow us to _NOTE the CONSTCOND is not an option.
103 */
104 #define BF32_SET_SB(x, low, len, shift, bias, val) do { \
105 ASSERT3U(IS_P2ALIGNED(val, 1U << shift), !=, B_FALSE); \
106 ASSERT3S((val) >> (shift), >=, bias); \
107 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \
108 _NOTE(CONSTCOND) } while (0)
109 #define BF64_SET_SB(x, low, len, shift, bias, val) do { \
110 ASSERT3U(IS_P2ALIGNED(val, 1ULL << shift), !=, B_FALSE); \
111 ASSERT3S((val) >> (shift), >=, bias); \
112 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \
113 _NOTE(CONSTCOND) } while (0)
114
115 /*
116 * We currently support block sizes from 512 bytes to 16MB.
117 * The benefits of larger blocks, and thus larger IO, need to be weighed
118 * against the cost of COWing a giant block to modify one byte, and the
119 * large latency of reading or writing a large block.
120 *
121 * Note that although blocks up to 16MB are supported, the recordsize
122 * property can not be set larger than zfs_max_recordsize (default 1MB).
123 * See the comment near zfs_max_recordsize in dsl_dataset.c for details.
124 *
125 * Note that although the LSIZE field of the blkptr_t can store sizes up
126 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to
127 * 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
128 */
129 #define SPA_MINBLOCKSHIFT 9
130 #define SPA_OLD_MAXBLOCKSHIFT 17
131 #define SPA_MAXBLOCKSHIFT 24
132 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT)
133 #define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT)
134 #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT)
135
136 /*
137 * Alignment Shift (ashift) is an immutable, internal top-level vdev property
138 * which can only be set at vdev creation time. Physical writes are always done
139 * according to it, which makes 2^ashift the smallest possible IO on a vdev.
140 *
141 * We currently allow values ranging from 512 bytes (2^9 = 512) to 64 KiB
142 * (2^16 = 65,536).
143 */
144 #define ASHIFT_MIN 9
145 #define ASHIFT_MAX 16
146
147 /*
148 * Size of block to hold the configuration data (a packed nvlist)
149 */
150 #define SPA_CONFIG_BLOCKSIZE (1ULL << 14)
151
152 /*
153 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
154 * The ASIZE encoding should be at least 64 times larger (6 more bits)
155 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
156 * overhead, three DVAs per bp, plus one more bit in case we do anything
157 * else that expands the ASIZE.
158 */
159 #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */
160 #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */
161 #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */
162
163 #define SPA_COMPRESSBITS 7
164 #define SPA_VDEVBITS 24
165
166 /*
167 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
168 * The members of the dva_t should be considered opaque outside the SPA.
169 */
170 typedef struct dva {
171 uint64_t dva_word[2];
172 } dva_t;
173
174
175 /*
176 * Some checksums/hashes need a 256-bit initialization salt. This salt is kept
177 * secret and is suitable for use in MAC algorithms as the key.
178 */
179 typedef struct zio_cksum_salt {
180 uint8_t zcs_bytes[32];
181 } zio_cksum_salt_t;
182
183 /*
184 * Each block is described by its DVAs, time of birth, checksum, etc.
185 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
186 *
187 * 64 56 48 40 32 24 16 8 0
188 * +-------+-------+-------+-------+-------+-------+-------+-------+
189 * 0 | pad | vdev1 | GRID | ASIZE |
190 * +-------+-------+-------+-------+-------+-------+-------+-------+
191 * 1 |G| offset1 |
192 * +-------+-------+-------+-------+-------+-------+-------+-------+
193 * 2 | pad | vdev2 | GRID | ASIZE |
194 * +-------+-------+-------+-------+-------+-------+-------+-------+
195 * 3 |G| offset2 |
196 * +-------+-------+-------+-------+-------+-------+-------+-------+
197 * 4 | pad | vdev3 | GRID | ASIZE |
198 * +-------+-------+-------+-------+-------+-------+-------+-------+
199 * 5 |G| offset3 |
200 * +-------+-------+-------+-------+-------+-------+-------+-------+
201 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE |
202 * +-------+-------+-------+-------+-------+-------+-------+-------+
203 * 7 | padding |
204 * +-------+-------+-------+-------+-------+-------+-------+-------+
205 * 8 | padding |
206 * +-------+-------+-------+-------+-------+-------+-------+-------+
207 * 9 | physical birth txg |
208 * +-------+-------+-------+-------+-------+-------+-------+-------+
209 * a | logical birth txg |
210 * +-------+-------+-------+-------+-------+-------+-------+-------+
211 * b | fill count |
212 * +-------+-------+-------+-------+-------+-------+-------+-------+
213 * c | checksum[0] |
214 * +-------+-------+-------+-------+-------+-------+-------+-------+
215 * d | checksum[1] |
216 * +-------+-------+-------+-------+-------+-------+-------+-------+
217 * e | checksum[2] |
218 * +-------+-------+-------+-------+-------+-------+-------+-------+
219 * f | checksum[3] |
220 * +-------+-------+-------+-------+-------+-------+-------+-------+
221 *
222 * Legend:
223 *
224 * vdev virtual device ID
225 * offset offset into virtual device
226 * LSIZE logical size
227 * PSIZE physical size (after compression)
228 * ASIZE allocated size (including RAID-Z parity and gang block headers)
229 * GRID RAID-Z layout information (reserved for future use)
230 * cksum checksum function
231 * comp compression function
232 * G gang block indicator
233 * B byteorder (endianness)
234 * D dedup
235 * X encryption
236 * E blkptr_t contains embedded data (see below)
237 * lvl level of indirection
238 * type DMU object type
239 * phys birth txg when dva[0] was written; zero if same as logical birth txg
240 * note that typically all the dva's would be written in this
241 * txg, but they could be different if they were moved by
242 * device removal.
243 * log. birth transaction group in which the block was logically born
244 * fill count number of non-zero blocks under this bp
245 * checksum[4] 256-bit checksum of the data this bp describes
246 */
247
248 /*
249 * The blkptr_t's of encrypted blocks also need to store the encryption
250 * parameters so that the block can be decrypted. This layout is as follows:
251 *
252 * 64 56 48 40 32 24 16 8 0
253 * +-------+-------+-------+-------+-------+-------+-------+-------+
254 * 0 | vdev1 | GRID | ASIZE |
255 * +-------+-------+-------+-------+-------+-------+-------+-------+
256 * 1 |G| offset1 |
257 * +-------+-------+-------+-------+-------+-------+-------+-------+
258 * 2 | vdev2 | GRID | ASIZE |
259 * +-------+-------+-------+-------+-------+-------+-------+-------+
260 * 3 |G| offset2 |
261 * +-------+-------+-------+-------+-------+-------+-------+-------+
262 * 4 | salt |
263 * +-------+-------+-------+-------+-------+-------+-------+-------+
264 * 5 | IV1 |
265 * +-------+-------+-------+-------+-------+-------+-------+-------+
266 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE |
267 * +-------+-------+-------+-------+-------+-------+-------+-------+
268 * 7 | padding |
269 * +-------+-------+-------+-------+-------+-------+-------+-------+
270 * 8 | padding |
271 * +-------+-------+-------+-------+-------+-------+-------+-------+
272 * 9 | physical birth txg |
273 * +-------+-------+-------+-------+-------+-------+-------+-------+
274 * a | logical birth txg |
275 * +-------+-------+-------+-------+-------+-------+-------+-------+
276 * b | IV2 | fill count |
277 * +-------+-------+-------+-------+-------+-------+-------+-------+
278 * c | checksum[0] |
279 * +-------+-------+-------+-------+-------+-------+-------+-------+
280 * d | checksum[1] |
281 * +-------+-------+-------+-------+-------+-------+-------+-------+
282 * e | MAC[0] |
283 * +-------+-------+-------+-------+-------+-------+-------+-------+
284 * f | MAC[1] |
285 * +-------+-------+-------+-------+-------+-------+-------+-------+
286 *
287 * Legend:
288 *
289 * salt Salt for generating encryption keys
290 * IV1 First 64 bits of encryption IV
291 * X Block requires encryption handling (set to 1)
292 * E blkptr_t contains embedded data (set to 0, see below)
293 * fill count number of non-zero blocks under this bp (truncated to 32 bits)
294 * IV2 Last 32 bits of encryption IV
295 * checksum[2] 128-bit checksum of the data this bp describes
296 * MAC[2] 128-bit message authentication code for this data
297 *
298 * The X bit being set indicates that this block is one of 3 types. If this is
299 * a level 0 block with an encrypted object type, the block is encrypted
300 * (see BP_IS_ENCRYPTED()). If this is a level 0 block with an unencrypted
301 * object type, this block is authenticated with an HMAC (see
302 * BP_IS_AUTHENTICATED()). Otherwise (if level > 0), this bp will use the MAC
303 * words to store a checksum-of-MACs from the level below (see
304 * BP_HAS_INDIRECT_MAC_CKSUM()). For convenience in the code, BP_IS_PROTECTED()
305 * refers to both encrypted and authenticated blocks and BP_USES_CRYPT()
306 * refers to any of these 3 kinds of blocks.
307 *
308 * The additional encryption parameters are the salt, IV, and MAC which are
309 * explained in greater detail in the block comment at the top of zio_crypt.c.
310 * The MAC occupies half of the checksum space since it serves a very similar
311 * purpose: to prevent data corruption on disk. The only functional difference
312 * is that the checksum is used to detect on-disk corruption whether or not the
313 * encryption key is loaded and the MAC provides additional protection against
314 * malicious disk tampering. We use the 3rd DVA to store the salt and first
315 * 64 bits of the IV. As a result encrypted blocks can only have 2 copies
316 * maximum instead of the normal 3. The last 32 bits of the IV are stored in
317 * the upper bits of what is usually the fill count. Note that only blocks at
318 * level 0 or -2 are ever encrypted, which allows us to guarantee that these
319 * 32 bits are not trampled over by other code (see zio_crypt.c for details).
320 * The salt and IV are not used for authenticated bps or bps with an indirect
321 * MAC checksum, so these blocks can utilize all 3 DVAs and the full 64 bits
322 * for the fill count.
323 */
324
325 /*
326 * "Embedded" blkptr_t's don't actually point to a block, instead they
327 * have a data payload embedded in the blkptr_t itself. See the comment
328 * in blkptr.c for more details.
329 *
330 * The blkptr_t is laid out as follows:
331 *
332 * 64 56 48 40 32 24 16 8 0
333 * +-------+-------+-------+-------+-------+-------+-------+-------+
334 * 0 | payload |
335 * 1 | payload |
336 * 2 | payload |
337 * 3 | payload |
338 * 4 | payload |
339 * 5 | payload |
340 * +-------+-------+-------+-------+-------+-------+-------+-------+
341 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE |
342 * +-------+-------+-------+-------+-------+-------+-------+-------+
343 * 7 | payload |
344 * 8 | payload |
345 * 9 | payload |
346 * +-------+-------+-------+-------+-------+-------+-------+-------+
347 * a | logical birth txg |
348 * +-------+-------+-------+-------+-------+-------+-------+-------+
349 * b | payload |
350 * c | payload |
351 * d | payload |
352 * e | payload |
353 * f | payload |
354 * +-------+-------+-------+-------+-------+-------+-------+-------+
355 *
356 * Legend:
357 *
358 * payload contains the embedded data
359 * B (byteorder) byteorder (endianness)
360 * D (dedup) padding (set to zero)
361 * X encryption (set to zero)
362 * E (embedded) set to one
363 * lvl indirection level
364 * type DMU object type
365 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*)
366 * comp compression function of payload
367 * PSIZE size of payload after compression, in bytes
368 * LSIZE logical size of payload, in bytes
369 * note that 25 bits is enough to store the largest
370 * "normal" BP's LSIZE (2^16 * 2^9) in bytes
371 * log. birth transaction group in which the block was logically born
372 *
373 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
374 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
375 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
376 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
377 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must
378 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before
379 * other macros, as they assert that they are only used on BP's of the correct
380 * "embedded-ness". Encrypted blkptr_t's cannot be embedded because they use
381 * the payload space for encryption parameters (see the comment above on
382 * how encryption parameters are stored).
383 */
384
385 #define BPE_GET_ETYPE(bp) \
386 (ASSERT(BP_IS_EMBEDDED(bp)), \
387 BF64_GET((bp)->blk_prop, 40, 8))
388 #define BPE_SET_ETYPE(bp, t) do { \
389 ASSERT(BP_IS_EMBEDDED(bp)); \
390 BF64_SET((bp)->blk_prop, 40, 8, t); \
391 _NOTE(CONSTCOND) } while (0)
392
393 #define BPE_GET_LSIZE(bp) \
394 (ASSERT(BP_IS_EMBEDDED(bp)), \
395 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
396 #define BPE_SET_LSIZE(bp, x) do { \
397 ASSERT(BP_IS_EMBEDDED(bp)); \
398 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
399 _NOTE(CONSTCOND) } while (0)
400
401 #define BPE_GET_PSIZE(bp) \
402 (ASSERT(BP_IS_EMBEDDED(bp)), \
403 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
404 #define BPE_SET_PSIZE(bp, x) do { \
405 ASSERT(BP_IS_EMBEDDED(bp)); \
406 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
407 _NOTE(CONSTCOND) } while (0)
408
409 typedef enum bp_embedded_type {
410 BP_EMBEDDED_TYPE_DATA,
411 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for Delphix byteswap feature. */
412 BP_EMBEDDED_TYPE_REDACTED,
413 NUM_BP_EMBEDDED_TYPES
414 } bp_embedded_type_t;
415
416 #define BPE_NUM_WORDS 14
417 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
418 #define BPE_IS_PAYLOADWORD(bp, wp) \
419 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
420
421 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */
422 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */
423 #define SPA_SYNC_MIN_VDEVS 3 /* min vdevs to update during sync */
424
425 /*
426 * A block is a hole when it has either 1) never been written to, or
427 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
428 * without physically allocating disk space. Holes are represented in the
429 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is
430 * done through the BP_IS_HOLE macro. For holes, the logical size, level,
431 * DMU object type, and birth times are all also stored for holes that
432 * were written to at some point (i.e. were punched after having been filled).
433 */
434 typedef struct blkptr {
435 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
436 uint64_t blk_prop; /* size, compression, type, etc */
437 uint64_t blk_pad[2]; /* Extra space for the future */
438 uint64_t blk_phys_birth; /* txg when block was allocated */
439 uint64_t blk_birth; /* transaction group at birth */
440 uint64_t blk_fill; /* fill count */
441 zio_cksum_t blk_cksum; /* 256-bit checksum */
442 } blkptr_t;
443
444 /*
445 * Macros to get and set fields in a bp or DVA.
446 */
447 #define DVA_GET_ASIZE(dva) \
448 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
449 #define DVA_SET_ASIZE(dva, x) \
450 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
451 SPA_MINBLOCKSHIFT, 0, x)
452
453 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8)
454 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x)
455
456 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, SPA_VDEVBITS)
457 #define DVA_SET_VDEV(dva, x) \
458 BF64_SET((dva)->dva_word[0], 32, SPA_VDEVBITS, x)
459
460 #define DVA_GET_OFFSET(dva) \
461 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
462 #define DVA_SET_OFFSET(dva, x) \
463 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
464
465 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1)
466 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x)
467
468 #define BP_GET_LSIZE(bp) \
469 (BP_IS_EMBEDDED(bp) ? \
470 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
471 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
472 #define BP_SET_LSIZE(bp, x) do { \
473 ASSERT(!BP_IS_EMBEDDED(bp)); \
474 BF64_SET_SB((bp)->blk_prop, \
475 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
476 _NOTE(CONSTCOND) } while (0)
477
478 #define BP_GET_PSIZE(bp) \
479 (BP_IS_EMBEDDED(bp) ? 0 : \
480 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1))
481 #define BP_SET_PSIZE(bp, x) do { \
482 ASSERT(!BP_IS_EMBEDDED(bp)); \
483 BF64_SET_SB((bp)->blk_prop, \
484 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
485 _NOTE(CONSTCOND) } while (0)
486
487 #define BP_GET_COMPRESS(bp) \
488 BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS)
489 #define BP_SET_COMPRESS(bp, x) \
490 BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x)
491
492 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1)
493 #define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x)
494
495 #define BP_GET_CHECKSUM(bp) \
496 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \
497 BF64_GET((bp)->blk_prop, 40, 8))
498 #define BP_SET_CHECKSUM(bp, x) do { \
499 ASSERT(!BP_IS_EMBEDDED(bp)); \
500 BF64_SET((bp)->blk_prop, 40, 8, x); \
501 _NOTE(CONSTCOND) } while (0)
502
503 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8)
504 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x)
505
506 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5)
507 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x)
508
509 /* encrypted, authenticated, and MAC cksum bps use the same bit */
510 #define BP_USES_CRYPT(bp) BF64_GET((bp)->blk_prop, 61, 1)
511 #define BP_SET_CRYPT(bp, x) BF64_SET((bp)->blk_prop, 61, 1, x)
512
513 #define BP_IS_ENCRYPTED(bp) \
514 (BP_USES_CRYPT(bp) && \
515 BP_GET_LEVEL(bp) <= 0 && \
516 DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp)))
517
518 #define BP_IS_AUTHENTICATED(bp) \
519 (BP_USES_CRYPT(bp) && \
520 BP_GET_LEVEL(bp) <= 0 && \
521 !DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp)))
522
523 #define BP_HAS_INDIRECT_MAC_CKSUM(bp) \
524 (BP_USES_CRYPT(bp) && BP_GET_LEVEL(bp) > 0)
525
526 #define BP_IS_PROTECTED(bp) \
527 (BP_IS_ENCRYPTED(bp) || BP_IS_AUTHENTICATED(bp))
528
529 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1)
530 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x)
531
532 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1)
533 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x)
534
535 #define BP_PHYSICAL_BIRTH(bp) \
536 (BP_IS_EMBEDDED(bp) ? 0 : \
537 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
538
539 #define BP_SET_BIRTH(bp, logical, physical) \
540 { \
541 ASSERT(!BP_IS_EMBEDDED(bp)); \
542 (bp)->blk_birth = (logical); \
543 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
544 }
545
546 #define BP_GET_FILL(bp) \
547 ((BP_IS_ENCRYPTED(bp)) ? BF64_GET((bp)->blk_fill, 0, 32) : \
548 ((BP_IS_EMBEDDED(bp)) ? 1 : (bp)->blk_fill))
549
550 #define BP_SET_FILL(bp, fill) \
551 { \
552 if (BP_IS_ENCRYPTED(bp)) \
553 BF64_SET((bp)->blk_fill, 0, 32, fill); \
554 else \
555 (bp)->blk_fill = fill; \
556 }
557
558 #define BP_GET_IV2(bp) \
559 (ASSERT(BP_IS_ENCRYPTED(bp)), \
560 BF64_GET((bp)->blk_fill, 32, 32))
561 #define BP_SET_IV2(bp, iv2) \
562 { \
563 ASSERT(BP_IS_ENCRYPTED(bp)); \
564 BF64_SET((bp)->blk_fill, 32, 32, iv2); \
565 }
566
567 #define BP_IS_METADATA(bp) \
568 (BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))
569
570 #define BP_GET_ASIZE(bp) \
571 (BP_IS_EMBEDDED(bp) ? 0 : \
572 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
573 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
574 (DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))
575
576 #define BP_GET_UCSIZE(bp) \
577 (BP_IS_METADATA(bp) ? BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
578
579 #define BP_GET_NDVAS(bp) \
580 (BP_IS_EMBEDDED(bp) ? 0 : \
581 !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
582 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
583 (!!DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))
584
585 #define BP_COUNT_GANG(bp) \
586 (BP_IS_EMBEDDED(bp) ? 0 : \
587 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \
588 DVA_GET_GANG(&(bp)->blk_dva[1]) + \
589 (DVA_GET_GANG(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp))))
590
591 #define DVA_EQUAL(dva1, dva2) \
592 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
593 (dva1)->dva_word[0] == (dva2)->dva_word[0])
594
595 #define BP_EQUAL(bp1, bp2) \
596 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \
597 (bp1)->blk_birth == (bp2)->blk_birth && \
598 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \
599 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \
600 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
601
602
603 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0)
604
605 #define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0])
606 #define BP_IS_GANG(bp) \
607 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp)))
608 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \
609 (dva)->dva_word[1] == 0ULL)
610 #define BP_IS_HOLE(bp) \
611 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp)))
612
613 #define BP_SET_REDACTED(bp) \
614 { \
615 BP_SET_EMBEDDED(bp, B_TRUE); \
616 BPE_SET_ETYPE(bp, BP_EMBEDDED_TYPE_REDACTED); \
617 }
618 #define BP_IS_REDACTED(bp) \
619 (BP_IS_EMBEDDED(bp) && BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_REDACTED)
620
621 /* BP_IS_RAIDZ(bp) assumes no block compression */
622 #define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
623 BP_GET_PSIZE(bp))
624
625 #define BP_ZERO(bp) \
626 { \
627 (bp)->blk_dva[0].dva_word[0] = 0; \
628 (bp)->blk_dva[0].dva_word[1] = 0; \
629 (bp)->blk_dva[1].dva_word[0] = 0; \
630 (bp)->blk_dva[1].dva_word[1] = 0; \
631 (bp)->blk_dva[2].dva_word[0] = 0; \
632 (bp)->blk_dva[2].dva_word[1] = 0; \
633 (bp)->blk_prop = 0; \
634 (bp)->blk_pad[0] = 0; \
635 (bp)->blk_pad[1] = 0; \
636 (bp)->blk_phys_birth = 0; \
637 (bp)->blk_birth = 0; \
638 (bp)->blk_fill = 0; \
639 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \
640 }
641
642 #ifdef _BIG_ENDIAN
643 #define ZFS_HOST_BYTEORDER (0ULL)
644 #else
645 #define ZFS_HOST_BYTEORDER (1ULL)
646 #endif
647
648 #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
649
650 #define BP_SPRINTF_LEN 400
651
652 /*
653 * This macro allows code sharing between zfs, libzpool, and mdb.
654 * 'func' is either snprintf() or mdb_snprintf().
655 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
656 */
657 #define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
658 { \
659 static const char *copyname[] = \
660 { "zero", "single", "double", "triple" }; \
661 int len = 0; \
662 int copies = 0; \
663 const char *crypt_type; \
664 if (bp != NULL) { \
665 if (BP_IS_ENCRYPTED(bp)) { \
666 crypt_type = "encrypted"; \
667 /* LINTED E_SUSPICIOUS_COMPARISON */ \
668 } else if (BP_IS_AUTHENTICATED(bp)) { \
669 crypt_type = "authenticated"; \
670 } else if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) { \
671 crypt_type = "indirect-MAC"; \
672 } else { \
673 crypt_type = "unencrypted"; \
674 } \
675 } \
676 if (bp == NULL) { \
677 len += func(buf + len, size - len, "<NULL>"); \
678 } else if (BP_IS_HOLE(bp)) { \
679 len += func(buf + len, size - len, \
680 "HOLE [L%llu %s] " \
681 "size=%llxL birth=%lluL", \
682 (u_longlong_t)BP_GET_LEVEL(bp), \
683 type, \
684 (u_longlong_t)BP_GET_LSIZE(bp), \
685 (u_longlong_t)bp->blk_birth); \
686 } else if (BP_IS_EMBEDDED(bp)) { \
687 len = func(buf + len, size - len, \
688 "EMBEDDED [L%llu %s] et=%u %s " \
689 "size=%llxL/%llxP birth=%lluL", \
690 (u_longlong_t)BP_GET_LEVEL(bp), \
691 type, \
692 (int)BPE_GET_ETYPE(bp), \
693 compress, \
694 (u_longlong_t)BPE_GET_LSIZE(bp), \
695 (u_longlong_t)BPE_GET_PSIZE(bp), \
696 (u_longlong_t)bp->blk_birth); \
697 } else if (BP_IS_REDACTED(bp)) { \
698 len += func(buf + len, size - len, \
699 "REDACTED [L%llu %s] size=%llxL birth=%lluL", \
700 (u_longlong_t)BP_GET_LEVEL(bp), \
701 type, \
702 (u_longlong_t)BP_GET_LSIZE(bp), \
703 (u_longlong_t)bp->blk_birth); \
704 } else { \
705 for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \
706 const dva_t *dva = &bp->blk_dva[d]; \
707 if (DVA_IS_VALID(dva)) \
708 copies++; \
709 len += func(buf + len, size - len, \
710 "DVA[%d]=<%llu:%llx:%llx>%c", d, \
711 (u_longlong_t)DVA_GET_VDEV(dva), \
712 (u_longlong_t)DVA_GET_OFFSET(dva), \
713 (u_longlong_t)DVA_GET_ASIZE(dva), \
714 ws); \
715 } \
716 if (BP_IS_ENCRYPTED(bp)) { \
717 len += func(buf + len, size - len, \
718 "salt=%llx iv=%llx:%llx%c", \
719 (u_longlong_t)bp->blk_dva[2].dva_word[0], \
720 (u_longlong_t)bp->blk_dva[2].dva_word[1], \
721 (u_longlong_t)BP_GET_IV2(bp), \
722 ws); \
723 } \
724 if (BP_IS_GANG(bp) && \
725 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \
726 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \
727 copies--; \
728 len += func(buf + len, size - len, \
729 "[L%llu %s] %s %s %s %s %s %s %s%c" \
730 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \
731 "cksum=%llx:%llx:%llx:%llx", \
732 (u_longlong_t)BP_GET_LEVEL(bp), \
733 type, \
734 checksum, \
735 compress, \
736 crypt_type, \
737 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \
738 BP_IS_GANG(bp) ? "gang" : "contiguous", \
739 BP_GET_DEDUP(bp) ? "dedup" : "unique", \
740 copyname[copies], \
741 ws, \
742 (u_longlong_t)BP_GET_LSIZE(bp), \
743 (u_longlong_t)BP_GET_PSIZE(bp), \
744 (u_longlong_t)bp->blk_birth, \
745 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \
746 (u_longlong_t)BP_GET_FILL(bp), \
747 ws, \
748 (u_longlong_t)bp->blk_cksum.zc_word[0], \
749 (u_longlong_t)bp->blk_cksum.zc_word[1], \
750 (u_longlong_t)bp->blk_cksum.zc_word[2], \
751 (u_longlong_t)bp->blk_cksum.zc_word[3]); \
752 } \
753 ASSERT(len < size); \
754 }
755
756 #define BP_GET_BUFC_TYPE(bp) \
757 (BP_IS_METADATA(bp) ? ARC_BUFC_METADATA : ARC_BUFC_DATA)
758
759 typedef enum spa_import_type {
760 SPA_IMPORT_EXISTING,
761 SPA_IMPORT_ASSEMBLE
762 } spa_import_type_t;
763
764 /*
765 * Send TRIM commands in-line during normal pool operation while deleting.
766 * OFF: no
767 * ON: yes
768 */
769 typedef enum {
770 SPA_AUTOTRIM_OFF = 0, /* default */
771 SPA_AUTOTRIM_ON
772 } spa_autotrim_t;
773
774 /*
775 * Reason TRIM command was issued, used internally for accounting purposes.
776 */
777 typedef enum trim_type {
778 TRIM_TYPE_MANUAL = 0,
779 TRIM_TYPE_AUTO = 1,
780 } trim_type_t;
781
782 /* state manipulation functions */
783 extern int spa_open(const char *pool, spa_t **, void *tag);
784 extern int spa_open_rewind(const char *pool, spa_t **, void *tag,
785 nvlist_t *policy, nvlist_t **config);
786 extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot,
787 size_t buflen);
788 extern int spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props,
789 nvlist_t *zplprops, struct dsl_crypto_params *dcp);
790 extern int spa_import(char *pool, nvlist_t *config, nvlist_t *props,
791 uint64_t flags);
792 extern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
793 extern int spa_destroy(char *pool);
794 extern int spa_checkpoint(const char *pool);
795 extern int spa_checkpoint_discard(const char *pool);
796 extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force,
797 boolean_t hardforce);
798 extern int spa_reset(char *pool);
799 extern void spa_async_request(spa_t *spa, int flag);
800 extern void spa_async_unrequest(spa_t *spa, int flag);
801 extern void spa_async_suspend(spa_t *spa);
802 extern void spa_async_resume(spa_t *spa);
803 extern spa_t *spa_inject_addref(char *pool);
804 extern void spa_inject_delref(spa_t *spa);
805 extern void spa_scan_stat_init(spa_t *spa);
806 extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps);
807
808 #define SPA_ASYNC_CONFIG_UPDATE 0x01
809 #define SPA_ASYNC_REMOVE 0x02
810 #define SPA_ASYNC_PROBE 0x04
811 #define SPA_ASYNC_RESILVER_DONE 0x08
812 #define SPA_ASYNC_RESILVER 0x10
813 #define SPA_ASYNC_AUTOEXPAND 0x20
814 #define SPA_ASYNC_REMOVE_DONE 0x40
815 #define SPA_ASYNC_REMOVE_STOP 0x80
816 #define SPA_ASYNC_INITIALIZE_RESTART 0x100
817 #define SPA_ASYNC_TRIM_RESTART 0x200
818 #define SPA_ASYNC_AUTOTRIM_RESTART 0x400
819
820 /*
821 * Controls the behavior of spa_vdev_remove().
822 */
823 #define SPA_REMOVE_UNSPARE 0x01
824 #define SPA_REMOVE_DONE 0x02
825
826 /* device manipulation */
827 extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
828 extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot,
829 int replacing);
830 extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
831 int replace_done);
832 extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare);
833 extern boolean_t spa_vdev_remove_active(spa_t *spa);
834 extern int spa_vdev_initialize(spa_t *spa, nvlist_t *nv, uint64_t cmd_type,
835 nvlist_t *vdev_errlist);
836 extern int spa_vdev_trim(spa_t *spa, nvlist_t *nv, uint64_t cmd_type,
837 uint64_t rate, boolean_t partial, boolean_t secure, nvlist_t *vdev_errlist);
838 extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
839 extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru);
840 extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
841 nvlist_t *props, boolean_t exp);
842
843 /* spare state (which is global across all pools) */
844 extern void spa_spare_add(vdev_t *vd);
845 extern void spa_spare_remove(vdev_t *vd);
846 extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt);
847 extern void spa_spare_activate(vdev_t *vd);
848
849 /* L2ARC state (which is global across all pools) */
850 extern void spa_l2cache_add(vdev_t *vd);
851 extern void spa_l2cache_remove(vdev_t *vd);
852 extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
853 extern void spa_l2cache_activate(vdev_t *vd);
854 extern void spa_l2cache_drop(spa_t *spa);
855
856 /* scanning */
857 extern int spa_scan(spa_t *spa, pool_scan_func_t func);
858 extern int spa_scan_stop(spa_t *spa);
859 extern int spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t flag);
860
861 /* spa syncing */
862 extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
863 extern void spa_sync_allpools(void);
864
865 extern int zfs_sync_pass_deferred_free;
866
867 /* spa namespace global mutex */
868 extern kmutex_t spa_namespace_lock;
869
870 /*
871 * SPA configuration functions in spa_config.c
872 */
873
874 #define SPA_CONFIG_UPDATE_POOL 0
875 #define SPA_CONFIG_UPDATE_VDEVS 1
876
877 extern void spa_write_cachefile(spa_t *, boolean_t, boolean_t);
878 extern void spa_config_load(void);
879 extern nvlist_t *spa_all_configs(uint64_t *);
880 extern void spa_config_set(spa_t *spa, nvlist_t *config);
881 extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
882 int getstats);
883 extern void spa_config_update(spa_t *spa, int what);
884
885 /*
886 * Miscellaneous SPA routines in spa_misc.c
887 */
888
889 /* Namespace manipulation */
890 extern spa_t *spa_lookup(const char *name);
891 extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot);
892 extern void spa_remove(spa_t *spa);
893 extern spa_t *spa_next(spa_t *prev);
894
895 /* Refcount functions */
896 extern void spa_open_ref(spa_t *spa, void *tag);
897 extern void spa_close(spa_t *spa, void *tag);
898 extern void spa_async_close(spa_t *spa, void *tag);
899 extern boolean_t spa_refcount_zero(spa_t *spa);
900
901 #define SCL_NONE 0x00
902 #define SCL_CONFIG 0x01
903 #define SCL_STATE 0x02
904 #define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */
905 #define SCL_ALLOC 0x08
906 #define SCL_ZIO 0x10
907 #define SCL_FREE 0x20
908 #define SCL_VDEV 0x40
909 #define SCL_LOCKS 7
910 #define SCL_ALL ((1 << SCL_LOCKS) - 1)
911 #define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO)
912
913 /* Historical pool statistics */
914 typedef struct spa_history_kstat {
915 kmutex_t lock;
916 uint64_t count;
917 uint64_t size;
918 kstat_t *kstat;
919 void *private;
920 list_t list;
921 } spa_history_kstat_t;
922
923 typedef struct spa_history_list {
924 uint64_t size;
925 procfs_list_t procfs_list;
926 } spa_history_list_t;
927
928 typedef struct spa_stats {
929 spa_history_list_t read_history;
930 spa_history_list_t txg_history;
931 spa_history_kstat_t tx_assign_histogram;
932 spa_history_kstat_t io_history;
933 spa_history_list_t mmp_history;
934 spa_history_kstat_t state; /* pool state */
935 spa_history_kstat_t iostats;
936 } spa_stats_t;
937
938 typedef enum txg_state {
939 TXG_STATE_BIRTH = 0,
940 TXG_STATE_OPEN = 1,
941 TXG_STATE_QUIESCED = 2,
942 TXG_STATE_WAIT_FOR_SYNC = 3,
943 TXG_STATE_SYNCED = 4,
944 TXG_STATE_COMMITTED = 5,
945 } txg_state_t;
946
947 typedef struct txg_stat {
948 vdev_stat_t vs1;
949 vdev_stat_t vs2;
950 uint64_t txg;
951 uint64_t ndirty;
952 } txg_stat_t;
953
954 /* Assorted pool IO kstats */
955 typedef struct spa_iostats {
956 kstat_named_t trim_extents_written;
957 kstat_named_t trim_bytes_written;
958 kstat_named_t trim_extents_skipped;
959 kstat_named_t trim_bytes_skipped;
960 kstat_named_t trim_extents_failed;
961 kstat_named_t trim_bytes_failed;
962 kstat_named_t autotrim_extents_written;
963 kstat_named_t autotrim_bytes_written;
964 kstat_named_t autotrim_extents_skipped;
965 kstat_named_t autotrim_bytes_skipped;
966 kstat_named_t autotrim_extents_failed;
967 kstat_named_t autotrim_bytes_failed;
968 } spa_iostats_t;
969
970 extern void spa_stats_init(spa_t *spa);
971 extern void spa_stats_destroy(spa_t *spa);
972 extern void spa_read_history_add(spa_t *spa, const zbookmark_phys_t *zb,
973 uint32_t aflags);
974 extern void spa_txg_history_add(spa_t *spa, uint64_t txg, hrtime_t birth_time);
975 extern int spa_txg_history_set(spa_t *spa, uint64_t txg,
976 txg_state_t completed_state, hrtime_t completed_time);
977 extern txg_stat_t *spa_txg_history_init_io(spa_t *, uint64_t,
978 struct dsl_pool *);
979 extern void spa_txg_history_fini_io(spa_t *, txg_stat_t *);
980 extern void spa_tx_assign_add_nsecs(spa_t *spa, uint64_t nsecs);
981 extern int spa_mmp_history_set_skip(spa_t *spa, uint64_t mmp_kstat_id);
982 extern int spa_mmp_history_set(spa_t *spa, uint64_t mmp_kstat_id, int io_error,
983 hrtime_t duration);
984 extern void spa_mmp_history_add(spa_t *spa, uint64_t txg, uint64_t timestamp,
985 uint64_t mmp_delay, vdev_t *vd, int label, uint64_t mmp_kstat_id,
986 int error);
987 extern void spa_iostats_trim_add(spa_t *spa, trim_type_t type,
988 uint64_t extents_written, uint64_t bytes_written,
989 uint64_t extents_skipped, uint64_t bytes_skipped,
990 uint64_t extents_failed, uint64_t bytes_failed);
991 extern void spa_import_progress_add(spa_t *spa);
992 extern void spa_import_progress_remove(uint64_t spa_guid);
993 extern int spa_import_progress_set_mmp_check(uint64_t pool_guid,
994 uint64_t mmp_sec_remaining);
995 extern int spa_import_progress_set_max_txg(uint64_t pool_guid,
996 uint64_t max_txg);
997 extern int spa_import_progress_set_state(uint64_t pool_guid,
998 spa_load_state_t spa_load_state);
999
1000 /* Pool configuration locks */
1001 extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw);
1002 extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw);
1003 extern void spa_config_exit(spa_t *spa, int locks, void *tag);
1004 extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
1005
1006 /* Pool vdev add/remove lock */
1007 extern uint64_t spa_vdev_enter(spa_t *spa);
1008 extern uint64_t spa_vdev_config_enter(spa_t *spa);
1009 extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg,
1010 int error, char *tag);
1011 extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error);
1012
1013 /* Pool vdev state change lock */
1014 extern void spa_vdev_state_enter(spa_t *spa, int oplock);
1015 extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error);
1016
1017 /* Log state */
1018 typedef enum spa_log_state {
1019 SPA_LOG_UNKNOWN = 0, /* unknown log state */
1020 SPA_LOG_MISSING, /* missing log(s) */
1021 SPA_LOG_CLEAR, /* clear the log(s) */
1022 SPA_LOG_GOOD, /* log(s) are good */
1023 } spa_log_state_t;
1024
1025 extern spa_log_state_t spa_get_log_state(spa_t *spa);
1026 extern void spa_set_log_state(spa_t *spa, spa_log_state_t state);
1027 extern int spa_reset_logs(spa_t *spa);
1028
1029 /* Log claim callback */
1030 extern void spa_claim_notify(zio_t *zio);
1031 extern void spa_deadman(void *);
1032
1033 /* Accessor functions */
1034 extern boolean_t spa_shutting_down(spa_t *spa);
1035 extern struct dsl_pool *spa_get_dsl(spa_t *spa);
1036 extern boolean_t spa_is_initializing(spa_t *spa);
1037 extern boolean_t spa_indirect_vdevs_loaded(spa_t *spa);
1038 extern blkptr_t *spa_get_rootblkptr(spa_t *spa);
1039 extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
1040 extern void spa_altroot(spa_t *, char *, size_t);
1041 extern int spa_sync_pass(spa_t *spa);
1042 extern char *spa_name(spa_t *spa);
1043 extern uint64_t spa_guid(spa_t *spa);
1044 extern uint64_t spa_load_guid(spa_t *spa);
1045 extern uint64_t spa_last_synced_txg(spa_t *spa);
1046 extern uint64_t spa_first_txg(spa_t *spa);
1047 extern uint64_t spa_syncing_txg(spa_t *spa);
1048 extern uint64_t spa_final_dirty_txg(spa_t *spa);
1049 extern uint64_t spa_version(spa_t *spa);
1050 extern pool_state_t spa_state(spa_t *spa);
1051 extern spa_load_state_t spa_load_state(spa_t *spa);
1052 extern uint64_t spa_freeze_txg(spa_t *spa);
1053 extern uint64_t spa_get_worst_case_asize(spa_t *spa, uint64_t lsize);
1054 extern uint64_t spa_get_dspace(spa_t *spa);
1055 extern uint64_t spa_get_checkpoint_space(spa_t *spa);
1056 extern uint64_t spa_get_slop_space(spa_t *spa);
1057 extern void spa_update_dspace(spa_t *spa);
1058 extern uint64_t spa_version(spa_t *spa);
1059 extern boolean_t spa_deflate(spa_t *spa);
1060 extern metaslab_class_t *spa_normal_class(spa_t *spa);
1061 extern metaslab_class_t *spa_log_class(spa_t *spa);
1062 extern metaslab_class_t *spa_special_class(spa_t *spa);
1063 extern metaslab_class_t *spa_dedup_class(spa_t *spa);
1064 extern metaslab_class_t *spa_preferred_class(spa_t *spa, uint64_t size,
1065 dmu_object_type_t objtype, uint_t level, uint_t special_smallblk);
1066
1067 extern void spa_evicting_os_register(spa_t *, objset_t *os);
1068 extern void spa_evicting_os_deregister(spa_t *, objset_t *os);
1069 extern void spa_evicting_os_wait(spa_t *spa);
1070 extern int spa_max_replication(spa_t *spa);
1071 extern int spa_prev_software_version(spa_t *spa);
1072 extern uint64_t spa_get_failmode(spa_t *spa);
1073 extern uint64_t spa_get_deadman_failmode(spa_t *spa);
1074 extern void spa_set_deadman_failmode(spa_t *spa, const char *failmode);
1075 extern boolean_t spa_suspended(spa_t *spa);
1076 extern uint64_t spa_bootfs(spa_t *spa);
1077 extern uint64_t spa_delegation(spa_t *spa);
1078 extern objset_t *spa_meta_objset(spa_t *spa);
1079 extern space_map_t *spa_syncing_log_sm(spa_t *spa);
1080 extern uint64_t spa_deadman_synctime(spa_t *spa);
1081 extern uint64_t spa_deadman_ziotime(spa_t *spa);
1082 extern uint64_t spa_dirty_data(spa_t *spa);
1083 extern spa_autotrim_t spa_get_autotrim(spa_t *spa);
1084
1085 /* Miscellaneous support routines */
1086 extern void spa_load_failed(spa_t *spa, const char *fmt, ...);
1087 extern void spa_load_note(spa_t *spa, const char *fmt, ...);
1088 extern void spa_activate_mos_feature(spa_t *spa, const char *feature,
1089 dmu_tx_t *tx);
1090 extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature);
1091 extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid);
1092 extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid);
1093 extern char *spa_strdup(const char *);
1094 extern void spa_strfree(char *);
1095 extern uint64_t spa_get_random(uint64_t range);
1096 extern uint64_t spa_generate_guid(spa_t *spa);
1097 extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp);
1098 extern void spa_freeze(spa_t *spa);
1099 extern int spa_change_guid(spa_t *spa);
1100 extern void spa_upgrade(spa_t *spa, uint64_t version);
1101 extern void spa_evict_all(void);
1102 extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid,
1103 boolean_t l2cache);
1104 extern boolean_t spa_has_spare(spa_t *, uint64_t guid);
1105 extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva);
1106 extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp);
1107 extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp);
1108 extern boolean_t spa_has_slogs(spa_t *spa);
1109 extern boolean_t spa_is_root(spa_t *spa);
1110 extern boolean_t spa_writeable(spa_t *spa);
1111 extern boolean_t spa_has_pending_synctask(spa_t *spa);
1112 extern int spa_maxblocksize(spa_t *spa);
1113 extern int spa_maxdnodesize(spa_t *spa);
1114 extern boolean_t spa_has_checkpoint(spa_t *spa);
1115 extern boolean_t spa_importing_readonly_checkpoint(spa_t *spa);
1116 extern boolean_t spa_suspend_async_destroy(spa_t *spa);
1117 extern uint64_t spa_min_claim_txg(spa_t *spa);
1118 extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp);
1119 extern boolean_t zfs_dva_valid(spa_t *spa, const dva_t *dva,
1120 const blkptr_t *bp);
1121 typedef void (*spa_remap_cb_t)(uint64_t vdev, uint64_t offset, uint64_t size,
1122 void *arg);
1123 extern boolean_t spa_remap_blkptr(spa_t *spa, blkptr_t *bp,
1124 spa_remap_cb_t callback, void *arg);
1125 extern uint64_t spa_get_last_removal_txg(spa_t *spa);
1126 extern boolean_t spa_trust_config(spa_t *spa);
1127 extern uint64_t spa_missing_tvds_allowed(spa_t *spa);
1128 extern void spa_set_missing_tvds(spa_t *spa, uint64_t missing);
1129 extern boolean_t spa_top_vdevs_spacemap_addressable(spa_t *spa);
1130 extern uint64_t spa_total_metaslabs(spa_t *spa);
1131 extern boolean_t spa_multihost(spa_t *spa);
1132 extern unsigned long spa_get_hostid(void);
1133 extern void spa_activate_allocation_classes(spa_t *, dmu_tx_t *);
1134
1135 extern int spa_mode(spa_t *spa);
1136 extern uint64_t zfs_strtonum(const char *str, char **nptr);
1137
1138 extern char *spa_his_ievent_table[];
1139
1140 extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx);
1141 extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read,
1142 char *his_buf);
1143 extern int spa_history_log(spa_t *spa, const char *his_buf);
1144 extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl);
1145 extern void spa_history_log_version(spa_t *spa, const char *operation,
1146 dmu_tx_t *tx);
1147 extern void spa_history_log_internal(spa_t *spa, const char *operation,
1148 dmu_tx_t *tx, const char *fmt, ...);
1149 extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op,
1150 dmu_tx_t *tx, const char *fmt, ...);
1151 extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
1152 dmu_tx_t *tx, const char *fmt, ...);
1153
1154 extern const char *spa_state_to_name(spa_t *spa);
1155
1156 /* error handling */
1157 struct zbookmark_phys;
1158 extern void spa_log_error(spa_t *spa, const zbookmark_phys_t *zb);
1159 extern int zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd,
1160 const zbookmark_phys_t *zb, zio_t *zio, uint64_t stateoroffset,
1161 uint64_t length);
1162 extern boolean_t zfs_ereport_is_valid(const char *class, spa_t *spa, vdev_t *vd,
1163 zio_t *zio);
1164 extern nvlist_t *zfs_event_create(spa_t *spa, vdev_t *vd, const char *type,
1165 const char *name, nvlist_t *aux);
1166 extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
1167 extern void zfs_post_state_change(spa_t *spa, vdev_t *vd, uint64_t laststate);
1168 extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
1169 extern uint64_t spa_get_errlog_size(spa_t *spa);
1170 extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
1171 extern void spa_errlog_rotate(spa_t *spa);
1172 extern void spa_errlog_drain(spa_t *spa);
1173 extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
1174 extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
1175
1176 /* vdev cache */
1177 extern void vdev_cache_stat_init(void);
1178 extern void vdev_cache_stat_fini(void);
1179
1180 /* vdev mirror */
1181 extern void vdev_mirror_stat_init(void);
1182 extern void vdev_mirror_stat_fini(void);
1183
1184 /* Initialization and termination */
1185 extern void spa_init(int flags);
1186 extern void spa_fini(void);
1187 extern void spa_boot_init(void);
1188
1189 /* properties */
1190 extern int spa_prop_set(spa_t *spa, nvlist_t *nvp);
1191 extern int spa_prop_get(spa_t *spa, nvlist_t **nvp);
1192 extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx);
1193 extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t);
1194
1195 /* asynchronous event notification */
1196 extern void spa_event_notify(spa_t *spa, vdev_t *vdev, nvlist_t *hist_nvl,
1197 const char *name);
1198
1199 #ifdef ZFS_DEBUG
1200 #define dprintf_bp(bp, fmt, ...) do { \
1201 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
1202 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \
1203 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \
1204 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \
1205 kmem_free(__blkbuf, BP_SPRINTF_LEN); \
1206 } \
1207 _NOTE(CONSTCOND) } while (0)
1208 #else
1209 #define dprintf_bp(bp, fmt, ...)
1210 #endif
1211
1212 extern int spa_mode_global; /* mode, e.g. FREAD | FWRITE */
1213 extern int zfs_deadman_enabled;
1214 extern unsigned long zfs_deadman_synctime_ms;
1215 extern unsigned long zfs_deadman_ziotime_ms;
1216 extern unsigned long zfs_deadman_checktime_ms;
1217
1218 #ifdef __cplusplus
1219 }
1220 #endif
1221
1222 #endif /* _SYS_SPA_H */
ZFS On Linux mirror for PVE