4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
32 #include <sys/zfs_context.h>
33 #include <sys/kstat.h>
34 #include <sys/nvpair.h>
35 #include <sys/sysmacros.h>
36 #include <sys/types.h>
37 #include <sys/fs/zfs.h>
38 #include <sys/spa_checksum.h>
45 * Forward references that lots of things need.
47 typedef struct spa spa_t;
48 typedef struct vdev vdev_t;
49 typedef struct metaslab metaslab_t;
50 typedef struct metaslab_group metaslab_group_t;
51 typedef struct metaslab_class metaslab_class_t;
52 typedef struct zio zio_t;
53 typedef struct zilog zilog_t;
54 typedef struct spa_aux_vdev spa_aux_vdev_t;
55 typedef struct ddt ddt_t;
56 typedef struct ddt_entry ddt_entry_t;
57 typedef struct zbookmark_phys zbookmark_phys_t;
63 * General-purpose 32-bit and 64-bit bitfield encodings.
65 #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len))
66 #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len))
67 #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low))
68 #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low))
70 #define BF32_GET(x, low, len) BF32_DECODE(x, low, len)
71 #define BF64_GET(x, low, len) BF64_DECODE(x, low, len)
73 #define BF32_SET(x, low, len, val) do { \
74 ASSERT3U(val, <, 1U << (len)); \
75 ASSERT3U(low + len, <=, 32); \
76 (x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \
77 _NOTE(CONSTCOND) } while (0)
79 #define BF64_SET(x, low, len, val) do { \
80 ASSERT3U(val, <, 1ULL << (len)); \
81 ASSERT3U(low + len, <=, 64); \
82 ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \
83 _NOTE(CONSTCOND) } while (0)
85 #define BF32_GET_SB(x, low, len, shift, bias) \
86 ((BF32_GET(x, low, len) + (bias)) << (shift))
87 #define BF64_GET_SB(x, low, len, shift, bias) \
88 ((BF64_GET(x, low, len) + (bias)) << (shift))
90 #define BF32_SET_SB(x, low, len, shift, bias, val) do { \
91 ASSERT(IS_P2ALIGNED(val, 1U << shift)); \
92 ASSERT3S((val) >> (shift), >=, bias); \
93 BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \
94 _NOTE(CONSTCOND) } while (0)
95 #define BF64_SET_SB(x, low, len, shift, bias, val) do { \
96 ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \
97 ASSERT3S((val) >> (shift), >=, bias); \
98 BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \
99 _NOTE(CONSTCOND) } while (0)
102 * We currently support block sizes from 512 bytes to 16MB.
103 * The benefits of larger blocks, and thus larger IO, need to be weighed
104 * against the cost of COWing a giant block to modify one byte, and the
105 * large latency of reading or writing a large block.
107 * Note that although blocks up to 16MB are supported, the recordsize
108 * property can not be set larger than zfs_max_recordsize (default 1MB).
109 * See the comment near zfs_max_recordsize in dsl_dataset.c for details.
111 * Note that although the LSIZE field of the blkptr_t can store sizes up
112 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to
113 * 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
115 #define SPA_MINBLOCKSHIFT 9
116 #define SPA_OLD_MAXBLOCKSHIFT 17
117 #define SPA_MAXBLOCKSHIFT 24
118 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT)
119 #define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT)
120 #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT)
123 * Size of block to hold the configuration data (a packed nvlist)
125 #define SPA_CONFIG_BLOCKSIZE (1ULL << 14)
128 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
129 * The ASIZE encoding should be at least 64 times larger (6 more bits)
130 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
131 * overhead, three DVAs per bp, plus one more bit in case we do anything
132 * else that expands the ASIZE.
134 #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */
135 #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */
136 #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */
139 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
140 * The members of the dva_t should be considered opaque outside the SPA.
143 uint64_t dva_word[2];
148 * Each block is described by its DVAs, time of birth, checksum, etc.
149 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
151 * 64 56 48 40 32 24 16 8 0
152 * +-------+-------+-------+-------+-------+-------+-------+-------+
153 * 0 | vdev1 | GRID | ASIZE |
154 * +-------+-------+-------+-------+-------+-------+-------+-------+
156 * +-------+-------+-------+-------+-------+-------+-------+-------+
157 * 2 | vdev2 | GRID | ASIZE |
158 * +-------+-------+-------+-------+-------+-------+-------+-------+
160 * +-------+-------+-------+-------+-------+-------+-------+-------+
161 * 4 | vdev3 | GRID | ASIZE |
162 * +-------+-------+-------+-------+-------+-------+-------+-------+
164 * +-------+-------+-------+-------+-------+-------+-------+-------+
165 * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE |
166 * +-------+-------+-------+-------+-------+-------+-------+-------+
168 * +-------+-------+-------+-------+-------+-------+-------+-------+
170 * +-------+-------+-------+-------+-------+-------+-------+-------+
171 * 9 | physical birth txg |
172 * +-------+-------+-------+-------+-------+-------+-------+-------+
173 * a | logical birth txg |
174 * +-------+-------+-------+-------+-------+-------+-------+-------+
176 * +-------+-------+-------+-------+-------+-------+-------+-------+
178 * +-------+-------+-------+-------+-------+-------+-------+-------+
180 * +-------+-------+-------+-------+-------+-------+-------+-------+
182 * +-------+-------+-------+-------+-------+-------+-------+-------+
184 * +-------+-------+-------+-------+-------+-------+-------+-------+
188 * vdev virtual device ID
189 * offset offset into virtual device
191 * PSIZE physical size (after compression)
192 * ASIZE allocated size (including RAID-Z parity and gang block headers)
193 * GRID RAID-Z layout information (reserved for future use)
194 * cksum checksum function
195 * comp compression function
196 * G gang block indicator
197 * B byteorder (endianness)
199 * X encryption (on version 30, which is not supported)
200 * E blkptr_t contains embedded data (see below)
201 * lvl level of indirection
202 * type DMU object type
203 * phys birth txg of block allocation; zero if same as logical birth txg
204 * log. birth transaction group in which the block was logically born
205 * fill count number of non-zero blocks under this bp
206 * checksum[4] 256-bit checksum of the data this bp describes
210 * "Embedded" blkptr_t's don't actually point to a block, instead they
211 * have a data payload embedded in the blkptr_t itself. See the comment
212 * in blkptr.c for more details.
214 * The blkptr_t is laid out as follows:
216 * 64 56 48 40 32 24 16 8 0
217 * +-------+-------+-------+-------+-------+-------+-------+-------+
224 * +-------+-------+-------+-------+-------+-------+-------+-------+
225 * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE |
226 * +-------+-------+-------+-------+-------+-------+-------+-------+
230 * +-------+-------+-------+-------+-------+-------+-------+-------+
231 * a | logical birth txg |
232 * +-------+-------+-------+-------+-------+-------+-------+-------+
238 * +-------+-------+-------+-------+-------+-------+-------+-------+
242 * payload contains the embedded data
243 * B (byteorder) byteorder (endianness)
244 * D (dedup) padding (set to zero)
245 * X encryption (set to zero; see above)
246 * E (embedded) set to one
247 * lvl indirection level
248 * type DMU object type
249 * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*)
250 * comp compression function of payload
251 * PSIZE size of payload after compression, in bytes
252 * LSIZE logical size of payload, in bytes
253 * note that 25 bits is enough to store the largest
254 * "normal" BP's LSIZE (2^16 * 2^9) in bytes
255 * log. birth transaction group in which the block was logically born
257 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
258 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
259 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
260 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
261 * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must
262 * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before
263 * other macros, as they assert that they are only used on BP's of the correct
267 #define BPE_GET_ETYPE(bp) \
268 (ASSERT(BP_IS_EMBEDDED(bp)), \
269 BF64_GET((bp)->blk_prop, 40, 8))
270 #define BPE_SET_ETYPE(bp, t) do { \
271 ASSERT(BP_IS_EMBEDDED(bp)); \
272 BF64_SET((bp)->blk_prop, 40, 8, t); \
273 _NOTE(CONSTCOND) } while (0)
275 #define BPE_GET_LSIZE(bp) \
276 (ASSERT(BP_IS_EMBEDDED(bp)), \
277 BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
278 #define BPE_SET_LSIZE(bp, x) do { \
279 ASSERT(BP_IS_EMBEDDED(bp)); \
280 BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
281 _NOTE(CONSTCOND) } while (0)
283 #define BPE_GET_PSIZE(bp) \
284 (ASSERT(BP_IS_EMBEDDED(bp)), \
285 BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
286 #define BPE_SET_PSIZE(bp, x) do { \
287 ASSERT(BP_IS_EMBEDDED(bp)); \
288 BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
289 _NOTE(CONSTCOND) } while (0)
291 typedef enum bp_embedded_type {
292 BP_EMBEDDED_TYPE_DATA,
293 BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */
294 NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED
295 } bp_embedded_type_t;
297 #define BPE_NUM_WORDS 14
298 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
299 #define BPE_IS_PAYLOADWORD(bp, wp) \
300 ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
302 #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */
303 #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */
306 * A block is a hole when it has either 1) never been written to, or
307 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
308 * without physically allocating disk space. Holes are represented in the
309 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is
310 * done through the BP_IS_HOLE macro. For holes, the logical size, level,
311 * DMU object type, and birth times are all also stored for holes that
312 * were written to at some point (i.e. were punched after having been filled).
314 typedef struct blkptr {
315 dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
316 uint64_t blk_prop; /* size, compression, type, etc */
317 uint64_t blk_pad[2]; /* Extra space for the future */
318 uint64_t blk_phys_birth; /* txg when block was allocated */
319 uint64_t blk_birth; /* transaction group at birth */
320 uint64_t blk_fill; /* fill count */
321 zio_cksum_t blk_cksum; /* 256-bit checksum */
325 * Macros to get and set fields in a bp or DVA.
327 #define DVA_GET_ASIZE(dva) \
328 BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
329 #define DVA_SET_ASIZE(dva, x) \
330 BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
331 SPA_MINBLOCKSHIFT, 0, x)
333 #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8)
334 #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x)
336 #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32)
337 #define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x)
339 #define DVA_GET_OFFSET(dva) \
340 BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
341 #define DVA_SET_OFFSET(dva, x) \
342 BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
344 #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1)
345 #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x)
347 #define BP_GET_LSIZE(bp) \
348 (BP_IS_EMBEDDED(bp) ? \
349 (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
350 BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
351 #define BP_SET_LSIZE(bp, x) do { \
352 ASSERT(!BP_IS_EMBEDDED(bp)); \
353 BF64_SET_SB((bp)->blk_prop, \
354 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
355 _NOTE(CONSTCOND) } while (0)
357 #define BP_GET_PSIZE(bp) \
358 (BP_IS_EMBEDDED(bp) ? 0 : \
359 BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1))
360 #define BP_SET_PSIZE(bp, x) do { \
361 ASSERT(!BP_IS_EMBEDDED(bp)); \
362 BF64_SET_SB((bp)->blk_prop, \
363 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
364 _NOTE(CONSTCOND) } while (0)
366 #define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7)
367 #define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x)
369 #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1)
370 #define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x)
372 #define BP_GET_CHECKSUM(bp) \
373 (BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \
374 BF64_GET((bp)->blk_prop, 40, 8))
375 #define BP_SET_CHECKSUM(bp, x) do { \
376 ASSERT(!BP_IS_EMBEDDED(bp)); \
377 BF64_SET((bp)->blk_prop, 40, 8, x); \
378 _NOTE(CONSTCOND) } while (0)
380 #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8)
381 #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x)
383 #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5)
384 #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x)
386 #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1)
387 #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x)
389 #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1)
390 #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x)
392 #define BP_PHYSICAL_BIRTH(bp) \
393 (BP_IS_EMBEDDED(bp) ? 0 : \
394 (bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
396 #define BP_SET_BIRTH(bp, logical, physical) \
398 ASSERT(!BP_IS_EMBEDDED(bp)); \
399 (bp)->blk_birth = (logical); \
400 (bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
403 #define BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill)
405 #define BP_GET_ASIZE(bp) \
406 (BP_IS_EMBEDDED(bp) ? 0 : \
407 DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
408 DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
409 DVA_GET_ASIZE(&(bp)->blk_dva[2]))
411 #define BP_GET_UCSIZE(bp) \
412 ((BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) ? \
413 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
415 #define BP_GET_NDVAS(bp) \
416 (BP_IS_EMBEDDED(bp) ? 0 : \
417 !!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
418 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
419 !!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
421 #define BP_COUNT_GANG(bp) \
422 (BP_IS_EMBEDDED(bp) ? 0 : \
423 (DVA_GET_GANG(&(bp)->blk_dva[0]) + \
424 DVA_GET_GANG(&(bp)->blk_dva[1]) + \
425 DVA_GET_GANG(&(bp)->blk_dva[2])))
427 #define DVA_EQUAL(dva1, dva2) \
428 ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
429 (dva1)->dva_word[0] == (dva2)->dva_word[0])
431 #define BP_EQUAL(bp1, bp2) \
432 (BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \
433 (bp1)->blk_birth == (bp2)->blk_birth && \
434 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \
435 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \
436 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
439 #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0)
441 #define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0])
442 #define BP_IS_GANG(bp) \
443 (BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp)))
444 #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \
445 (dva)->dva_word[1] == 0ULL)
446 #define BP_IS_HOLE(bp) \
447 (!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp)))
449 /* BP_IS_RAIDZ(bp) assumes no block compression */
450 #define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
453 #define BP_ZERO(bp) \
455 (bp)->blk_dva[0].dva_word[0] = 0; \
456 (bp)->blk_dva[0].dva_word[1] = 0; \
457 (bp)->blk_dva[1].dva_word[0] = 0; \
458 (bp)->blk_dva[1].dva_word[1] = 0; \
459 (bp)->blk_dva[2].dva_word[0] = 0; \
460 (bp)->blk_dva[2].dva_word[1] = 0; \
461 (bp)->blk_prop = 0; \
462 (bp)->blk_pad[0] = 0; \
463 (bp)->blk_pad[1] = 0; \
464 (bp)->blk_phys_birth = 0; \
465 (bp)->blk_birth = 0; \
466 (bp)->blk_fill = 0; \
467 ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \
471 #define ZFS_HOST_BYTEORDER (0ULL)
473 #define ZFS_HOST_BYTEORDER (1ULL)
476 #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
478 #define BP_SPRINTF_LEN 320
481 * This macro allows code sharing between zfs, libzpool, and mdb.
482 * 'func' is either snprintf() or mdb_snprintf().
483 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
485 #define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
487 static const char *copyname[] = \
488 { "zero", "single", "double", "triple" }; \
494 len += func(buf + len, size - len, "<NULL>"); \
495 } else if (BP_IS_HOLE(bp)) { \
496 len += func(buf + len, size - len, \
498 "size=%llxL birth=%lluL", \
499 (u_longlong_t)BP_GET_LEVEL(bp), \
501 (u_longlong_t)BP_GET_LSIZE(bp), \
502 (u_longlong_t)bp->blk_birth); \
503 } else if (BP_IS_EMBEDDED(bp)) { \
504 len = func(buf + len, size - len, \
505 "EMBEDDED [L%llu %s] et=%u %s " \
506 "size=%llxL/%llxP birth=%lluL", \
507 (u_longlong_t)BP_GET_LEVEL(bp), \
509 (int)BPE_GET_ETYPE(bp), \
511 (u_longlong_t)BPE_GET_LSIZE(bp), \
512 (u_longlong_t)BPE_GET_PSIZE(bp), \
513 (u_longlong_t)bp->blk_birth); \
515 for (d = 0; d < BP_GET_NDVAS(bp); d++) { \
516 const dva_t *dva = &bp->blk_dva[d]; \
517 if (DVA_IS_VALID(dva)) \
519 len += func(buf + len, size - len, \
520 "DVA[%d]=<%llu:%llx:%llx>%c", d, \
521 (u_longlong_t)DVA_GET_VDEV(dva), \
522 (u_longlong_t)DVA_GET_OFFSET(dva), \
523 (u_longlong_t)DVA_GET_ASIZE(dva), \
526 if (BP_IS_GANG(bp) && \
527 DVA_GET_ASIZE(&bp->blk_dva[2]) <= \
528 DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \
530 len += func(buf + len, size - len, \
531 "[L%llu %s] %s %s %s %s %s %s%c" \
532 "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \
533 "cksum=%llx:%llx:%llx:%llx", \
534 (u_longlong_t)BP_GET_LEVEL(bp), \
538 BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \
539 BP_IS_GANG(bp) ? "gang" : "contiguous", \
540 BP_GET_DEDUP(bp) ? "dedup" : "unique", \
543 (u_longlong_t)BP_GET_LSIZE(bp), \
544 (u_longlong_t)BP_GET_PSIZE(bp), \
545 (u_longlong_t)bp->blk_birth, \
546 (u_longlong_t)BP_PHYSICAL_BIRTH(bp), \
547 (u_longlong_t)BP_GET_FILL(bp), \
549 (u_longlong_t)bp->blk_cksum.zc_word[0], \
550 (u_longlong_t)bp->blk_cksum.zc_word[1], \
551 (u_longlong_t)bp->blk_cksum.zc_word[2], \
552 (u_longlong_t)bp->blk_cksum.zc_word[3]); \
554 ASSERT(len < size); \
559 #define BP_GET_BUFC_TYPE(bp) \
560 (((BP_GET_LEVEL(bp) > 0) || (DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))) ? \
561 ARC_BUFC_METADATA : ARC_BUFC_DATA)
563 typedef enum spa_import_type {
568 /* state manipulation functions */
569 extern int spa_open(const char *pool, spa_t **, void *tag);
570 extern int spa_open_rewind(const char *pool, spa_t **, void *tag,
571 nvlist_t *policy, nvlist_t **config);
572 extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot,
574 extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props,
576 extern int spa_import_rootpool(char *devpath, char *devid);
577 extern int spa_import(char *pool, nvlist_t *config, nvlist_t *props,
579 extern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
580 extern int spa_destroy(char *pool);
581 extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force,
582 boolean_t hardforce);
583 extern int spa_reset(char *pool);
584 extern void spa_async_request(spa_t *spa, int flag);
585 extern void spa_async_unrequest(spa_t *spa, int flag);
586 extern void spa_async_suspend(spa_t *spa);
587 extern void spa_async_resume(spa_t *spa);
588 extern spa_t *spa_inject_addref(char *pool);
589 extern void spa_inject_delref(spa_t *spa);
590 extern void spa_scan_stat_init(spa_t *spa);
591 extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps);
593 #define SPA_ASYNC_CONFIG_UPDATE 0x01
594 #define SPA_ASYNC_REMOVE 0x02
595 #define SPA_ASYNC_PROBE 0x04
596 #define SPA_ASYNC_RESILVER_DONE 0x08
597 #define SPA_ASYNC_RESILVER 0x10
598 #define SPA_ASYNC_AUTOEXPAND 0x20
599 #define SPA_ASYNC_REMOVE_DONE 0x40
600 #define SPA_ASYNC_REMOVE_STOP 0x80
603 * Controls the behavior of spa_vdev_remove().
605 #define SPA_REMOVE_UNSPARE 0x01
606 #define SPA_REMOVE_DONE 0x02
608 /* device manipulation */
609 extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
610 extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot,
612 extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
614 extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare);
615 extern boolean_t spa_vdev_remove_active(spa_t *spa);
616 extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
617 extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru);
618 extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
619 nvlist_t *props, boolean_t exp);
621 /* spare state (which is global across all pools) */
622 extern void spa_spare_add(vdev_t *vd);
623 extern void spa_spare_remove(vdev_t *vd);
624 extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt);
625 extern void spa_spare_activate(vdev_t *vd);
627 /* L2ARC state (which is global across all pools) */
628 extern void spa_l2cache_add(vdev_t *vd);
629 extern void spa_l2cache_remove(vdev_t *vd);
630 extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
631 extern void spa_l2cache_activate(vdev_t *vd);
632 extern void spa_l2cache_drop(spa_t *spa);
635 extern int spa_scan(spa_t *spa, pool_scan_func_t func);
636 extern int spa_scan_stop(spa_t *spa);
639 extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
640 extern void spa_sync_allpools(void);
642 extern int zfs_sync_pass_deferred_free;
644 /* spa namespace global mutex */
645 extern kmutex_t spa_namespace_lock;
648 * SPA configuration functions in spa_config.c
651 #define SPA_CONFIG_UPDATE_POOL 0
652 #define SPA_CONFIG_UPDATE_VDEVS 1
654 extern void spa_config_sync(spa_t *, boolean_t, boolean_t);
655 extern void spa_config_load(void);
656 extern nvlist_t *spa_all_configs(uint64_t *);
657 extern void spa_config_set(spa_t *spa, nvlist_t *config);
658 extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
660 extern void spa_config_update(spa_t *spa, int what);
663 * Miscellaneous SPA routines in spa_misc.c
666 /* Namespace manipulation */
667 extern spa_t *spa_lookup(const char *name);
668 extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot);
669 extern void spa_remove(spa_t *spa);
670 extern spa_t *spa_next(spa_t *prev);
672 /* Refcount functions */
673 extern void spa_open_ref(spa_t *spa, void *tag);
674 extern void spa_close(spa_t *spa, void *tag);
675 extern void spa_async_close(spa_t *spa, void *tag);
676 extern boolean_t spa_refcount_zero(spa_t *spa);
678 #define SCL_NONE 0x00
679 #define SCL_CONFIG 0x01
680 #define SCL_STATE 0x02
681 #define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */
682 #define SCL_ALLOC 0x08
684 #define SCL_FREE 0x20
685 #define SCL_VDEV 0x40
687 #define SCL_ALL ((1 << SCL_LOCKS) - 1)
688 #define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO)
690 /* Historical pool statistics */
691 typedef struct spa_stats_history {
698 } spa_stats_history_t;
700 typedef struct spa_stats {
701 spa_stats_history_t read_history;
702 spa_stats_history_t txg_history;
703 spa_stats_history_t tx_assign_histogram;
704 spa_stats_history_t io_history;
707 typedef enum txg_state {
710 TXG_STATE_QUIESCED = 2,
711 TXG_STATE_WAIT_FOR_SYNC = 3,
712 TXG_STATE_SYNCED = 4,
713 TXG_STATE_COMMITTED = 5,
716 extern void spa_stats_init(spa_t *spa);
717 extern void spa_stats_destroy(spa_t *spa);
718 extern void spa_read_history_add(spa_t *spa, const zbookmark_phys_t *zb,
720 extern void spa_txg_history_add(spa_t *spa, uint64_t txg, hrtime_t birth_time);
721 extern int spa_txg_history_set(spa_t *spa, uint64_t txg,
722 txg_state_t completed_state, hrtime_t completed_time);
723 extern int spa_txg_history_set_io(spa_t *spa, uint64_t txg, uint64_t nread,
724 uint64_t nwritten, uint64_t reads, uint64_t writes, uint64_t ndirty);
725 extern void spa_tx_assign_add_nsecs(spa_t *spa, uint64_t nsecs);
727 /* Pool configuration locks */
728 extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw);
729 extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw);
730 extern void spa_config_exit(spa_t *spa, int locks, void *tag);
731 extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
733 /* Pool vdev add/remove lock */
734 extern uint64_t spa_vdev_enter(spa_t *spa);
735 extern uint64_t spa_vdev_config_enter(spa_t *spa);
736 extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg,
737 int error, char *tag);
738 extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error);
740 /* Pool vdev state change lock */
741 extern void spa_vdev_state_enter(spa_t *spa, int oplock);
742 extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error);
745 typedef enum spa_log_state {
746 SPA_LOG_UNKNOWN = 0, /* unknown log state */
747 SPA_LOG_MISSING, /* missing log(s) */
748 SPA_LOG_CLEAR, /* clear the log(s) */
749 SPA_LOG_GOOD, /* log(s) are good */
752 extern spa_log_state_t spa_get_log_state(spa_t *spa);
753 extern void spa_set_log_state(spa_t *spa, spa_log_state_t state);
754 extern int spa_offline_log(spa_t *spa);
756 /* Log claim callback */
757 extern void spa_claim_notify(zio_t *zio);
758 extern void spa_deadman(void *);
760 /* Accessor functions */
761 extern boolean_t spa_shutting_down(spa_t *spa);
762 extern struct dsl_pool *spa_get_dsl(spa_t *spa);
763 extern boolean_t spa_is_initializing(spa_t *spa);
764 extern blkptr_t *spa_get_rootblkptr(spa_t *spa);
765 extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
766 extern void spa_altroot(spa_t *, char *, size_t);
767 extern int spa_sync_pass(spa_t *spa);
768 extern char *spa_name(spa_t *spa);
769 extern uint64_t spa_guid(spa_t *spa);
770 extern uint64_t spa_load_guid(spa_t *spa);
771 extern uint64_t spa_last_synced_txg(spa_t *spa);
772 extern uint64_t spa_first_txg(spa_t *spa);
773 extern uint64_t spa_syncing_txg(spa_t *spa);
774 extern uint64_t spa_version(spa_t *spa);
775 extern pool_state_t spa_state(spa_t *spa);
776 extern spa_load_state_t spa_load_state(spa_t *spa);
777 extern uint64_t spa_freeze_txg(spa_t *spa);
778 extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize);
779 extern uint64_t spa_get_dspace(spa_t *spa);
780 extern uint64_t spa_get_slop_space(spa_t *spa);
781 extern void spa_update_dspace(spa_t *spa);
782 extern uint64_t spa_version(spa_t *spa);
783 extern boolean_t spa_deflate(spa_t *spa);
784 extern metaslab_class_t *spa_normal_class(spa_t *spa);
785 extern metaslab_class_t *spa_log_class(spa_t *spa);
786 extern void spa_evicting_os_register(spa_t *, objset_t *os);
787 extern void spa_evicting_os_deregister(spa_t *, objset_t *os);
788 extern void spa_evicting_os_wait(spa_t *spa);
789 extern int spa_max_replication(spa_t *spa);
790 extern int spa_prev_software_version(spa_t *spa);
791 extern uint8_t spa_get_failmode(spa_t *spa);
792 extern boolean_t spa_suspended(spa_t *spa);
793 extern uint64_t spa_bootfs(spa_t *spa);
794 extern uint64_t spa_delegation(spa_t *spa);
795 extern objset_t *spa_meta_objset(spa_t *spa);
796 extern uint64_t spa_deadman_synctime(spa_t *spa);
798 /* Miscellaneous support routines */
799 extern void spa_activate_mos_feature(spa_t *spa, const char *feature,
801 extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature);
802 extern int spa_rename(const char *oldname, const char *newname);
803 extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid);
804 extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid);
805 extern char *spa_strdup(const char *);
806 extern void spa_strfree(char *);
807 extern uint64_t spa_get_random(uint64_t range);
808 extern uint64_t spa_generate_guid(spa_t *spa);
809 extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp);
810 extern void spa_freeze(spa_t *spa);
811 extern int spa_change_guid(spa_t *spa);
812 extern void spa_upgrade(spa_t *spa, uint64_t version);
813 extern void spa_evict_all(void);
814 extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid,
816 extern boolean_t spa_has_spare(spa_t *, uint64_t guid);
817 extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva);
818 extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp);
819 extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp);
820 extern boolean_t spa_has_slogs(spa_t *spa);
821 extern boolean_t spa_is_root(spa_t *spa);
822 extern boolean_t spa_writeable(spa_t *spa);
823 extern boolean_t spa_has_pending_synctask(spa_t *spa);
824 extern int spa_maxblocksize(spa_t *spa);
825 extern int spa_maxdnodesize(spa_t *spa);
826 extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp);
828 extern int spa_mode(spa_t *spa);
829 extern uint64_t strtonum(const char *str, char **nptr);
831 extern char *spa_his_ievent_table[];
833 extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx);
834 extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read,
836 extern int spa_history_log(spa_t *spa, const char *his_buf);
837 extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl);
838 extern void spa_history_log_version(spa_t *spa, const char *operation);
839 extern void spa_history_log_internal(spa_t *spa, const char *operation,
840 dmu_tx_t *tx, const char *fmt, ...);
841 extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op,
842 dmu_tx_t *tx, const char *fmt, ...);
843 extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
844 dmu_tx_t *tx, const char *fmt, ...);
847 struct zbookmark_phys;
848 extern void spa_log_error(spa_t *spa, zio_t *zio);
849 extern void zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd,
850 zio_t *zio, uint64_t stateoroffset, uint64_t length);
851 extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
852 extern void zfs_post_state_change(spa_t *spa, vdev_t *vd, uint64_t laststate);
853 extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
854 extern void zfs_post_sysevent(spa_t *spa, vdev_t *vd, const char *name);
855 extern uint64_t spa_get_errlog_size(spa_t *spa);
856 extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
857 extern void spa_errlog_rotate(spa_t *spa);
858 extern void spa_errlog_drain(spa_t *spa);
859 extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
860 extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
863 extern void vdev_cache_stat_init(void);
864 extern void vdev_cache_stat_fini(void);
866 /* Initialization and termination */
867 extern void spa_init(int flags);
868 extern void spa_fini(void);
869 extern void spa_boot_init(void);
872 extern int spa_prop_set(spa_t *spa, nvlist_t *nvp);
873 extern int spa_prop_get(spa_t *spa, nvlist_t **nvp);
874 extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx);
875 extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t);
877 /* asynchronous event notification */
878 extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name);
881 #define dprintf_bp(bp, fmt, ...) do { \
882 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
883 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \
884 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \
885 dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \
886 kmem_free(__blkbuf, BP_SPRINTF_LEN); \
888 _NOTE(CONSTCOND) } while (0)
890 #define dprintf_bp(bp, fmt, ...)
893 extern boolean_t spa_debug_enabled(spa_t *spa);
894 #define spa_dbgmsg(spa, ...) \
896 if (spa_debug_enabled(spa)) \
897 zfs_dbgmsg(__VA_ARGS__); \
900 extern int spa_mode_global; /* mode, e.g. FREAD | FWRITE */
906 #endif /* _SYS_SPA_H */