spa.h revision 98110f08fa182032082d98be2ddb9391fcd62bf1
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, 2014 by Delphix. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
25 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
26 */
27
28#ifndef _SYS_SPA_H
29#define	_SYS_SPA_H
30
31#include <sys/avl.h>
32#include <sys/zfs_context.h>
33#include <sys/nvpair.h>
34#include <sys/sysmacros.h>
35#include <sys/types.h>
36#include <sys/fs/zfs.h>
37
38#ifdef	__cplusplus
39extern "C" {
40#endif
41
42/*
43 * Forward references that lots of things need.
44 */
45typedef struct spa spa_t;
46typedef struct vdev vdev_t;
47typedef struct metaslab metaslab_t;
48typedef struct metaslab_group metaslab_group_t;
49typedef struct metaslab_class metaslab_class_t;
50typedef struct zio zio_t;
51typedef struct zilog zilog_t;
52typedef struct spa_aux_vdev spa_aux_vdev_t;
53typedef struct ddt ddt_t;
54typedef struct ddt_entry ddt_entry_t;
55struct dsl_pool;
56struct dsl_dataset;
57
58/*
59 * General-purpose 32-bit and 64-bit bitfield encodings.
60 */
61#define	BF32_DECODE(x, low, len)	P2PHASE((x) >> (low), 1U << (len))
62#define	BF64_DECODE(x, low, len)	P2PHASE((x) >> (low), 1ULL << (len))
63#define	BF32_ENCODE(x, low, len)	(P2PHASE((x), 1U << (len)) << (low))
64#define	BF64_ENCODE(x, low, len)	(P2PHASE((x), 1ULL << (len)) << (low))
65
66#define	BF32_GET(x, low, len)		BF32_DECODE(x, low, len)
67#define	BF64_GET(x, low, len)		BF64_DECODE(x, low, len)
68
69#define	BF32_SET(x, low, len, val) do { \
70	ASSERT3U(val, <, 1U << (len)); \
71	ASSERT3U(low + len, <=, 32); \
72	(x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \
73_NOTE(CONSTCOND) } while (0)
74
75#define	BF64_SET(x, low, len, val) do { \
76	ASSERT3U(val, <, 1ULL << (len)); \
77	ASSERT3U(low + len, <=, 64); \
78	((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \
79_NOTE(CONSTCOND) } while (0)
80
81#define	BF32_GET_SB(x, low, len, shift, bias)	\
82	((BF32_GET(x, low, len) + (bias)) << (shift))
83#define	BF64_GET_SB(x, low, len, shift, bias)	\
84	((BF64_GET(x, low, len) + (bias)) << (shift))
85
86#define	BF32_SET_SB(x, low, len, shift, bias, val) do { \
87	ASSERT(IS_P2ALIGNED(val, 1U << shift)); \
88	ASSERT3S((val) >> (shift), >=, bias); \
89	BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \
90_NOTE(CONSTCOND) } while (0)
91#define	BF64_SET_SB(x, low, len, shift, bias, val) do { \
92	ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \
93	ASSERT3S((val) >> (shift), >=, bias); \
94	BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \
95_NOTE(CONSTCOND) } while (0)
96
97/*
98 * We currently support block sizes from 512 bytes to 16MB.
99 * The benefits of larger blocks, and thus larger IO, need to be weighed
100 * against the cost of COWing a giant block to modify one byte, and the
101 * large latency of reading or writing a large block.
102 *
103 * Note that although blocks up to 16MB are supported, the recordsize
104 * property can not be set larger than zfs_max_recordsize (default 1MB).
105 * See the comment near zfs_max_recordsize in dsl_dataset.c for details.
106 *
107 * Note that although the LSIZE field of the blkptr_t can store sizes up
108 * to 32MB, the dnode's dn_datablkszsec can only store sizes up to
109 * 32MB - 512 bytes.  Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
110 */
111#define	SPA_MINBLOCKSHIFT	9
112#define	SPA_OLD_MAXBLOCKSHIFT	17
113#define	SPA_MAXBLOCKSHIFT	24
114#define	SPA_MINBLOCKSIZE	(1ULL << SPA_MINBLOCKSHIFT)
115#define	SPA_OLD_MAXBLOCKSIZE	(1ULL << SPA_OLD_MAXBLOCKSHIFT)
116#define	SPA_MAXBLOCKSIZE	(1ULL << SPA_MAXBLOCKSHIFT)
117
118/*
119 * Size of block to hold the configuration data (a packed nvlist)
120 */
121#define	SPA_CONFIG_BLOCKSIZE	(1ULL << 14)
122
123/*
124 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
125 * The ASIZE encoding should be at least 64 times larger (6 more bits)
126 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
127 * overhead, three DVAs per bp, plus one more bit in case we do anything
128 * else that expands the ASIZE.
129 */
130#define	SPA_LSIZEBITS		16	/* LSIZE up to 32M (2^16 * 512)	*/
131#define	SPA_PSIZEBITS		16	/* PSIZE up to 32M (2^16 * 512)	*/
132#define	SPA_ASIZEBITS		24	/* ASIZE up to 64 times larger	*/
133
134/*
135 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
136 * The members of the dva_t should be considered opaque outside the SPA.
137 */
138typedef struct dva {
139	uint64_t	dva_word[2];
140} dva_t;
141
142/*
143 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
144 */
145typedef struct zio_cksum {
146	uint64_t	zc_word[4];
147} zio_cksum_t;
148
149/*
150 * Each block is described by its DVAs, time of birth, checksum, etc.
151 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
152 *
153 *	64	56	48	40	32	24	16	8	0
154 *	+-------+-------+-------+-------+-------+-------+-------+-------+
155 * 0	|		vdev1		| GRID  |	  ASIZE		|
156 *	+-------+-------+-------+-------+-------+-------+-------+-------+
157 * 1	|G|			 offset1				|
158 *	+-------+-------+-------+-------+-------+-------+-------+-------+
159 * 2	|		vdev2		| GRID  |	  ASIZE		|
160 *	+-------+-------+-------+-------+-------+-------+-------+-------+
161 * 3	|G|			 offset2				|
162 *	+-------+-------+-------+-------+-------+-------+-------+-------+
163 * 4	|		vdev3		| GRID  |	  ASIZE		|
164 *	+-------+-------+-------+-------+-------+-------+-------+-------+
165 * 5	|G|			 offset3				|
166 *	+-------+-------+-------+-------+-------+-------+-------+-------+
167 * 6	|BDX|lvl| type	| cksum |E| comp|    PSIZE	|     LSIZE	|
168 *	+-------+-------+-------+-------+-------+-------+-------+-------+
169 * 7	|			padding					|
170 *	+-------+-------+-------+-------+-------+-------+-------+-------+
171 * 8	|			padding					|
172 *	+-------+-------+-------+-------+-------+-------+-------+-------+
173 * 9	|			physical birth txg			|
174 *	+-------+-------+-------+-------+-------+-------+-------+-------+
175 * a	|			logical birth txg			|
176 *	+-------+-------+-------+-------+-------+-------+-------+-------+
177 * b	|			fill count				|
178 *	+-------+-------+-------+-------+-------+-------+-------+-------+
179 * c	|			checksum[0]				|
180 *	+-------+-------+-------+-------+-------+-------+-------+-------+
181 * d	|			checksum[1]				|
182 *	+-------+-------+-------+-------+-------+-------+-------+-------+
183 * e	|			checksum[2]				|
184 *	+-------+-------+-------+-------+-------+-------+-------+-------+
185 * f	|			checksum[3]				|
186 *	+-------+-------+-------+-------+-------+-------+-------+-------+
187 *
188 * Legend:
189 *
190 * vdev		virtual device ID
191 * offset	offset into virtual device
192 * LSIZE	logical size
193 * PSIZE	physical size (after compression)
194 * ASIZE	allocated size (including RAID-Z parity and gang block headers)
195 * GRID		RAID-Z layout information (reserved for future use)
196 * cksum	checksum function
197 * comp		compression function
198 * G		gang block indicator
199 * B		byteorder (endianness)
200 * D		dedup
201 * X		encryption (on version 30, which is not supported)
202 * E		blkptr_t contains embedded data (see below)
203 * lvl		level of indirection
204 * type		DMU object type
205 * phys birth	txg of block allocation; zero if same as logical birth txg
206 * log. birth	transaction group in which the block was logically born
207 * fill count	number of non-zero blocks under this bp
208 * checksum[4]	256-bit checksum of the data this bp describes
209 */
210
211/*
212 * "Embedded" blkptr_t's don't actually point to a block, instead they
213 * have a data payload embedded in the blkptr_t itself.  See the comment
214 * in blkptr.c for more details.
215 *
216 * The blkptr_t is laid out as follows:
217 *
218 *	64	56	48	40	32	24	16	8	0
219 *	+-------+-------+-------+-------+-------+-------+-------+-------+
220 * 0	|      payload                                                  |
221 * 1	|      payload                                                  |
222 * 2	|      payload                                                  |
223 * 3	|      payload                                                  |
224 * 4	|      payload                                                  |
225 * 5	|      payload                                                  |
226 *	+-------+-------+-------+-------+-------+-------+-------+-------+
227 * 6	|BDX|lvl| type	| etype |E| comp| PSIZE|              LSIZE	|
228 *	+-------+-------+-------+-------+-------+-------+-------+-------+
229 * 7	|      payload                                                  |
230 * 8	|      payload                                                  |
231 * 9	|      payload                                                  |
232 *	+-------+-------+-------+-------+-------+-------+-------+-------+
233 * a	|			logical birth txg			|
234 *	+-------+-------+-------+-------+-------+-------+-------+-------+
235 * b	|      payload                                                  |
236 * c	|      payload                                                  |
237 * d	|      payload                                                  |
238 * e	|      payload                                                  |
239 * f	|      payload                                                  |
240 *	+-------+-------+-------+-------+-------+-------+-------+-------+
241 *
242 * Legend:
243 *
244 * payload		contains the embedded data
245 * B (byteorder)	byteorder (endianness)
246 * D (dedup)		padding (set to zero)
247 * X			encryption (set to zero; see above)
248 * E (embedded)		set to one
249 * lvl			indirection level
250 * type			DMU object type
251 * etype		how to interpret embedded data (BP_EMBEDDED_TYPE_*)
252 * comp			compression function of payload
253 * PSIZE		size of payload after compression, in bytes
254 * LSIZE		logical size of payload, in bytes
255 *			note that 25 bits is enough to store the largest
256 *			"normal" BP's LSIZE (2^16 * 2^9) in bytes
257 * log. birth		transaction group in which the block was logically born
258 *
259 * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
260 * bp's they are stored in units of SPA_MINBLOCKSHIFT.
261 * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
262 * The B, D, X, lvl, type, and comp fields are stored the same as with normal
263 * BP's so the BP_SET_* macros can be used with them.  etype, PSIZE, LSIZE must
264 * be set with the BPE_SET_* macros.  BP_SET_EMBEDDED() should be called before
265 * other macros, as they assert that they are only used on BP's of the correct
266 * "embedded-ness".
267 */
268
269#define	BPE_GET_ETYPE(bp)	\
270	(ASSERT(BP_IS_EMBEDDED(bp)), \
271	BF64_GET((bp)->blk_prop, 40, 8))
272#define	BPE_SET_ETYPE(bp, t)	do { \
273	ASSERT(BP_IS_EMBEDDED(bp)); \
274	BF64_SET((bp)->blk_prop, 40, 8, t); \
275_NOTE(CONSTCOND) } while (0)
276
277#define	BPE_GET_LSIZE(bp)	\
278	(ASSERT(BP_IS_EMBEDDED(bp)), \
279	BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
280#define	BPE_SET_LSIZE(bp, x)	do { \
281	ASSERT(BP_IS_EMBEDDED(bp)); \
282	BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
283_NOTE(CONSTCOND) } while (0)
284
285#define	BPE_GET_PSIZE(bp)	\
286	(ASSERT(BP_IS_EMBEDDED(bp)), \
287	BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
288#define	BPE_SET_PSIZE(bp, x)	do { \
289	ASSERT(BP_IS_EMBEDDED(bp)); \
290	BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
291_NOTE(CONSTCOND) } while (0)
292
293typedef enum bp_embedded_type {
294	BP_EMBEDDED_TYPE_DATA,
295	BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */
296	NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED
297} bp_embedded_type_t;
298
299#define	BPE_NUM_WORDS 14
300#define	BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
301#define	BPE_IS_PAYLOADWORD(bp, wp) \
302	((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
303
304#define	SPA_BLKPTRSHIFT	7		/* blkptr_t is 128 bytes	*/
305#define	SPA_DVAS_PER_BP	3		/* Number of DVAs in a bp	*/
306
307/*
308 * A block is a hole when it has either 1) never been written to, or
309 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
310 * without physically allocating disk space. Holes are represented in the
311 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is
312 * done through the BP_IS_HOLE macro. For holes, the logical size, level,
313 * DMU object type, and birth times are all also stored for holes that
314 * were written to at some point (i.e. were punched after having been filled).
315 */
316typedef struct blkptr {
317	dva_t		blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
318	uint64_t	blk_prop;	/* size, compression, type, etc	    */
319	uint64_t	blk_pad[2];	/* Extra space for the future	    */
320	uint64_t	blk_phys_birth;	/* txg when block was allocated	    */
321	uint64_t	blk_birth;	/* transaction group at birth	    */
322	uint64_t	blk_fill;	/* fill count			    */
323	zio_cksum_t	blk_cksum;	/* 256-bit checksum		    */
324} blkptr_t;
325
326/*
327 * Macros to get and set fields in a bp or DVA.
328 */
329#define	DVA_GET_ASIZE(dva)	\
330	BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
331#define	DVA_SET_ASIZE(dva, x)	\
332	BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
333	SPA_MINBLOCKSHIFT, 0, x)
334
335#define	DVA_GET_GRID(dva)	BF64_GET((dva)->dva_word[0], 24, 8)
336#define	DVA_SET_GRID(dva, x)	BF64_SET((dva)->dva_word[0], 24, 8, x)
337
338#define	DVA_GET_VDEV(dva)	BF64_GET((dva)->dva_word[0], 32, 32)
339#define	DVA_SET_VDEV(dva, x)	BF64_SET((dva)->dva_word[0], 32, 32, x)
340
341#define	DVA_GET_OFFSET(dva)	\
342	BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
343#define	DVA_SET_OFFSET(dva, x)	\
344	BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
345
346#define	DVA_GET_GANG(dva)	BF64_GET((dva)->dva_word[1], 63, 1)
347#define	DVA_SET_GANG(dva, x)	BF64_SET((dva)->dva_word[1], 63, 1, x)
348
349#define	BP_GET_LSIZE(bp)	\
350	(BP_IS_EMBEDDED(bp) ?	\
351	(BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
352	BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
353#define	BP_SET_LSIZE(bp, x)	do { \
354	ASSERT(!BP_IS_EMBEDDED(bp)); \
355	BF64_SET_SB((bp)->blk_prop, \
356	    0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
357_NOTE(CONSTCOND) } while (0)
358
359#define	BP_GET_PSIZE(bp)	\
360	(BP_IS_EMBEDDED(bp) ? 0 : \
361	BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1))
362#define	BP_SET_PSIZE(bp, x)	do { \
363	ASSERT(!BP_IS_EMBEDDED(bp)); \
364	BF64_SET_SB((bp)->blk_prop, \
365	    16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
366_NOTE(CONSTCOND) } while (0)
367
368#define	BP_GET_COMPRESS(bp)		BF64_GET((bp)->blk_prop, 32, 7)
369#define	BP_SET_COMPRESS(bp, x)		BF64_SET((bp)->blk_prop, 32, 7, x)
370
371#define	BP_IS_EMBEDDED(bp)		BF64_GET((bp)->blk_prop, 39, 1)
372#define	BP_SET_EMBEDDED(bp, x)		BF64_SET((bp)->blk_prop, 39, 1, x)
373
374#define	BP_GET_CHECKSUM(bp)		\
375	(BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \
376	BF64_GET((bp)->blk_prop, 40, 8))
377#define	BP_SET_CHECKSUM(bp, x)		do { \
378	ASSERT(!BP_IS_EMBEDDED(bp)); \
379	BF64_SET((bp)->blk_prop, 40, 8, x); \
380_NOTE(CONSTCOND) } while (0)
381
382#define	BP_GET_TYPE(bp)			BF64_GET((bp)->blk_prop, 48, 8)
383#define	BP_SET_TYPE(bp, x)		BF64_SET((bp)->blk_prop, 48, 8, x)
384
385#define	BP_GET_LEVEL(bp)		BF64_GET((bp)->blk_prop, 56, 5)
386#define	BP_SET_LEVEL(bp, x)		BF64_SET((bp)->blk_prop, 56, 5, x)
387
388#define	BP_GET_DEDUP(bp)		BF64_GET((bp)->blk_prop, 62, 1)
389#define	BP_SET_DEDUP(bp, x)		BF64_SET((bp)->blk_prop, 62, 1, x)
390
391#define	BP_GET_BYTEORDER(bp)		BF64_GET((bp)->blk_prop, 63, 1)
392#define	BP_SET_BYTEORDER(bp, x)		BF64_SET((bp)->blk_prop, 63, 1, x)
393
394#define	BP_PHYSICAL_BIRTH(bp)		\
395	(BP_IS_EMBEDDED(bp) ? 0 : \
396	(bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
397
398#define	BP_SET_BIRTH(bp, logical, physical)	\
399{						\
400	ASSERT(!BP_IS_EMBEDDED(bp));		\
401	(bp)->blk_birth = (logical);		\
402	(bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
403}
404
405#define	BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill)
406
407#define	BP_GET_ASIZE(bp)	\
408	(BP_IS_EMBEDDED(bp) ? 0 : \
409	DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
410	DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
411	DVA_GET_ASIZE(&(bp)->blk_dva[2]))
412
413#define	BP_GET_UCSIZE(bp) \
414	((BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) ? \
415	BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
416
417#define	BP_GET_NDVAS(bp)	\
418	(BP_IS_EMBEDDED(bp) ? 0 : \
419	!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
420	!!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
421	!!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
422
423#define	BP_COUNT_GANG(bp)	\
424	(BP_IS_EMBEDDED(bp) ? 0 : \
425	(DVA_GET_GANG(&(bp)->blk_dva[0]) + \
426	DVA_GET_GANG(&(bp)->blk_dva[1]) + \
427	DVA_GET_GANG(&(bp)->blk_dva[2])))
428
429#define	DVA_EQUAL(dva1, dva2)	\
430	((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
431	(dva1)->dva_word[0] == (dva2)->dva_word[0])
432
433#define	BP_EQUAL(bp1, bp2)	\
434	(BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) &&	\
435	(bp1)->blk_birth == (bp2)->blk_birth &&			\
436	DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) &&	\
437	DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) &&	\
438	DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
439
440#define	ZIO_CHECKSUM_EQUAL(zc1, zc2) \
441	(0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
442	((zc1).zc_word[1] - (zc2).zc_word[1]) | \
443	((zc1).zc_word[2] - (zc2).zc_word[2]) | \
444	((zc1).zc_word[3] - (zc2).zc_word[3])))
445
446#define	ZIO_CHECKSUM_IS_ZERO(zc) \
447	(0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \
448	(zc)->zc_word[2] | (zc)->zc_word[3]))
449
450#define	ZIO_CHECKSUM_BSWAP(zcp)					\
451{								\
452	(zcp)->zc_word[0] = BSWAP_64((zcp)->zc_word[0]);	\
453	(zcp)->zc_word[1] = BSWAP_64((zcp)->zc_word[1]);	\
454	(zcp)->zc_word[2] = BSWAP_64((zcp)->zc_word[2]);	\
455	(zcp)->zc_word[3] = BSWAP_64((zcp)->zc_word[3]);	\
456}
457
458
459#define	DVA_IS_VALID(dva)	(DVA_GET_ASIZE(dva) != 0)
460
461#define	ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3)	\
462{						\
463	(zcp)->zc_word[0] = w0;			\
464	(zcp)->zc_word[1] = w1;			\
465	(zcp)->zc_word[2] = w2;			\
466	(zcp)->zc_word[3] = w3;			\
467}
468
469#define	BP_IDENTITY(bp)		(ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0])
470#define	BP_IS_GANG(bp)		\
471	(BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp)))
472#define	DVA_IS_EMPTY(dva)	((dva)->dva_word[0] == 0ULL &&	\
473				(dva)->dva_word[1] == 0ULL)
474#define	BP_IS_HOLE(bp) \
475	(!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp)))
476
477/* BP_IS_RAIDZ(bp) assumes no block compression */
478#define	BP_IS_RAIDZ(bp)		(DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
479				BP_GET_PSIZE(bp))
480
481#define	BP_ZERO(bp)				\
482{						\
483	(bp)->blk_dva[0].dva_word[0] = 0;	\
484	(bp)->blk_dva[0].dva_word[1] = 0;	\
485	(bp)->blk_dva[1].dva_word[0] = 0;	\
486	(bp)->blk_dva[1].dva_word[1] = 0;	\
487	(bp)->blk_dva[2].dva_word[0] = 0;	\
488	(bp)->blk_dva[2].dva_word[1] = 0;	\
489	(bp)->blk_prop = 0;			\
490	(bp)->blk_pad[0] = 0;			\
491	(bp)->blk_pad[1] = 0;			\
492	(bp)->blk_phys_birth = 0;		\
493	(bp)->blk_birth = 0;			\
494	(bp)->blk_fill = 0;			\
495	ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0);	\
496}
497
498#ifdef _BIG_ENDIAN
499#define	ZFS_HOST_BYTEORDER	(0ULL)
500#else
501#define	ZFS_HOST_BYTEORDER	(1ULL)
502#endif
503
504#define	BP_SHOULD_BYTESWAP(bp)	(BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
505
506#define	BP_SPRINTF_LEN	320
507
508/*
509 * This macro allows code sharing between zfs, libzpool, and mdb.
510 * 'func' is either snprintf() or mdb_snprintf().
511 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
512 */
513#define	SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
514{									\
515	static const char *copyname[] =					\
516	    { "zero", "single", "double", "triple" };			\
517	int len = 0;							\
518	int copies = 0;							\
519									\
520	if (bp == NULL) {						\
521		len += func(buf + len, size - len, "<NULL>");		\
522	} else if (BP_IS_HOLE(bp)) {					\
523		len += func(buf + len, size - len,			\
524		    "HOLE [L%llu %s] "					\
525		    "size=%llxL birth=%lluL",				\
526		    (u_longlong_t)BP_GET_LEVEL(bp),			\
527		    type,						\
528		    (u_longlong_t)BP_GET_LSIZE(bp),			\
529		    (u_longlong_t)bp->blk_birth);			\
530	} else if (BP_IS_EMBEDDED(bp)) {				\
531		len = func(buf + len, size - len,			\
532		    "EMBEDDED [L%llu %s] et=%u %s "			\
533		    "size=%llxL/%llxP birth=%lluL",			\
534		    (u_longlong_t)BP_GET_LEVEL(bp),			\
535		    type,						\
536		    (int)BPE_GET_ETYPE(bp),				\
537		    compress,						\
538		    (u_longlong_t)BPE_GET_LSIZE(bp),			\
539		    (u_longlong_t)BPE_GET_PSIZE(bp),			\
540		    (u_longlong_t)bp->blk_birth);			\
541	} else {							\
542		for (int d = 0; d < BP_GET_NDVAS(bp); d++) {		\
543			const dva_t *dva = &bp->blk_dva[d];		\
544			if (DVA_IS_VALID(dva))				\
545				copies++;				\
546			len += func(buf + len, size - len,		\
547			    "DVA[%d]=<%llu:%llx:%llx>%c", d,		\
548			    (u_longlong_t)DVA_GET_VDEV(dva),		\
549			    (u_longlong_t)DVA_GET_OFFSET(dva),		\
550			    (u_longlong_t)DVA_GET_ASIZE(dva),		\
551			    ws);					\
552		}							\
553		if (BP_IS_GANG(bp) &&					\
554		    DVA_GET_ASIZE(&bp->blk_dva[2]) <=			\
555		    DVA_GET_ASIZE(&bp->blk_dva[1]) / 2)			\
556			copies--;					\
557		len += func(buf + len, size - len,			\
558		    "[L%llu %s] %s %s %s %s %s %s%c"			\
559		    "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c"	\
560		    "cksum=%llx:%llx:%llx:%llx",			\
561		    (u_longlong_t)BP_GET_LEVEL(bp),			\
562		    type,						\
563		    checksum,						\
564		    compress,						\
565		    BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE",		\
566		    BP_IS_GANG(bp) ? "gang" : "contiguous",		\
567		    BP_GET_DEDUP(bp) ? "dedup" : "unique",		\
568		    copyname[copies],					\
569		    ws,							\
570		    (u_longlong_t)BP_GET_LSIZE(bp),			\
571		    (u_longlong_t)BP_GET_PSIZE(bp),			\
572		    (u_longlong_t)bp->blk_birth,			\
573		    (u_longlong_t)BP_PHYSICAL_BIRTH(bp),		\
574		    (u_longlong_t)BP_GET_FILL(bp),			\
575		    ws,							\
576		    (u_longlong_t)bp->blk_cksum.zc_word[0],		\
577		    (u_longlong_t)bp->blk_cksum.zc_word[1],		\
578		    (u_longlong_t)bp->blk_cksum.zc_word[2],		\
579		    (u_longlong_t)bp->blk_cksum.zc_word[3]);		\
580	}								\
581	ASSERT(len < size);						\
582}
583
584#include <sys/dmu.h>
585
586#define	BP_GET_BUFC_TYPE(bp)						\
587	(((BP_GET_LEVEL(bp) > 0) || (DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))) ? \
588	ARC_BUFC_METADATA : ARC_BUFC_DATA)
589
590typedef enum spa_import_type {
591	SPA_IMPORT_EXISTING,
592	SPA_IMPORT_ASSEMBLE
593} spa_import_type_t;
594
595/* state manipulation functions */
596extern int spa_open(const char *pool, spa_t **, void *tag);
597extern int spa_open_rewind(const char *pool, spa_t **, void *tag,
598    nvlist_t *policy, nvlist_t **config);
599extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot,
600    size_t buflen);
601extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props,
602    nvlist_t *zplprops);
603extern int spa_import_rootpool(char *devpath, char *devid);
604extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props,
605    uint64_t flags);
606extern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
607extern int spa_destroy(char *pool);
608extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force,
609    boolean_t hardforce);
610extern int spa_reset(char *pool);
611extern void spa_async_request(spa_t *spa, int flag);
612extern void spa_async_unrequest(spa_t *spa, int flag);
613extern void spa_async_suspend(spa_t *spa);
614extern void spa_async_resume(spa_t *spa);
615extern spa_t *spa_inject_addref(char *pool);
616extern void spa_inject_delref(spa_t *spa);
617extern void spa_scan_stat_init(spa_t *spa);
618extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps);
619
620#define	SPA_ASYNC_CONFIG_UPDATE	0x01
621#define	SPA_ASYNC_REMOVE	0x02
622#define	SPA_ASYNC_PROBE		0x04
623#define	SPA_ASYNC_RESILVER_DONE	0x08
624#define	SPA_ASYNC_RESILVER	0x10
625#define	SPA_ASYNC_AUTOEXPAND	0x20
626#define	SPA_ASYNC_REMOVE_DONE	0x40
627#define	SPA_ASYNC_REMOVE_STOP	0x80
628
629/*
630 * Controls the behavior of spa_vdev_remove().
631 */
632#define	SPA_REMOVE_UNSPARE	0x01
633#define	SPA_REMOVE_DONE		0x02
634
635/* device manipulation */
636extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
637extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot,
638    int replacing);
639extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
640    int replace_done);
641extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare);
642extern boolean_t spa_vdev_remove_active(spa_t *spa);
643extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
644extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru);
645extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
646    nvlist_t *props, boolean_t exp);
647
648/* spare state (which is global across all pools) */
649extern void spa_spare_add(vdev_t *vd);
650extern void spa_spare_remove(vdev_t *vd);
651extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt);
652extern void spa_spare_activate(vdev_t *vd);
653
654/* L2ARC state (which is global across all pools) */
655extern void spa_l2cache_add(vdev_t *vd);
656extern void spa_l2cache_remove(vdev_t *vd);
657extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
658extern void spa_l2cache_activate(vdev_t *vd);
659extern void spa_l2cache_drop(spa_t *spa);
660
661/* scanning */
662extern int spa_scan(spa_t *spa, pool_scan_func_t func);
663extern int spa_scan_stop(spa_t *spa);
664
665/* spa syncing */
666extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
667extern void spa_sync_allpools(void);
668
669/* spa namespace global mutex */
670extern kmutex_t spa_namespace_lock;
671
672/*
673 * SPA configuration functions in spa_config.c
674 */
675
676#define	SPA_CONFIG_UPDATE_POOL	0
677#define	SPA_CONFIG_UPDATE_VDEVS	1
678
679extern void spa_config_sync(spa_t *, boolean_t, boolean_t);
680extern void spa_config_load(void);
681extern nvlist_t *spa_all_configs(uint64_t *);
682extern void spa_config_set(spa_t *spa, nvlist_t *config);
683extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
684    int getstats);
685extern void spa_config_update(spa_t *spa, int what);
686
687/*
688 * Miscellaneous SPA routines in spa_misc.c
689 */
690
691/* Namespace manipulation */
692extern spa_t *spa_lookup(const char *name);
693extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot);
694extern void spa_remove(spa_t *spa);
695extern spa_t *spa_next(spa_t *prev);
696
697/* Refcount functions */
698extern void spa_open_ref(spa_t *spa, void *tag);
699extern void spa_close(spa_t *spa, void *tag);
700extern void spa_async_close(spa_t *spa, void *tag);
701extern boolean_t spa_refcount_zero(spa_t *spa);
702
703#define	SCL_NONE	0x00
704#define	SCL_CONFIG	0x01
705#define	SCL_STATE	0x02
706#define	SCL_L2ARC	0x04		/* hack until L2ARC 2.0 */
707#define	SCL_ALLOC	0x08
708#define	SCL_ZIO		0x10
709#define	SCL_FREE	0x20
710#define	SCL_VDEV	0x40
711#define	SCL_LOCKS	7
712#define	SCL_ALL		((1 << SCL_LOCKS) - 1)
713#define	SCL_STATE_ALL	(SCL_STATE | SCL_L2ARC | SCL_ZIO)
714
715/* Pool configuration locks */
716extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw);
717extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw);
718extern void spa_config_exit(spa_t *spa, int locks, void *tag);
719extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
720
721/* Pool vdev add/remove lock */
722extern uint64_t spa_vdev_enter(spa_t *spa);
723extern uint64_t spa_vdev_config_enter(spa_t *spa);
724extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg,
725    int error, char *tag);
726extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error);
727
728/* Pool vdev state change lock */
729extern void spa_vdev_state_enter(spa_t *spa, int oplock);
730extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error);
731
732/* Log state */
733typedef enum spa_log_state {
734	SPA_LOG_UNKNOWN = 0,	/* unknown log state */
735	SPA_LOG_MISSING,	/* missing log(s) */
736	SPA_LOG_CLEAR,		/* clear the log(s) */
737	SPA_LOG_GOOD,		/* log(s) are good */
738} spa_log_state_t;
739
740extern spa_log_state_t spa_get_log_state(spa_t *spa);
741extern void spa_set_log_state(spa_t *spa, spa_log_state_t state);
742extern int spa_offline_log(spa_t *spa);
743
744/* Log claim callback */
745extern void spa_claim_notify(zio_t *zio);
746
747/* Accessor functions */
748extern boolean_t spa_shutting_down(spa_t *spa);
749extern struct dsl_pool *spa_get_dsl(spa_t *spa);
750extern boolean_t spa_is_initializing(spa_t *spa);
751extern blkptr_t *spa_get_rootblkptr(spa_t *spa);
752extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
753extern void spa_altroot(spa_t *, char *, size_t);
754extern int spa_sync_pass(spa_t *spa);
755extern char *spa_name(spa_t *spa);
756extern uint64_t spa_guid(spa_t *spa);
757extern uint64_t spa_load_guid(spa_t *spa);
758extern uint64_t spa_last_synced_txg(spa_t *spa);
759extern uint64_t spa_first_txg(spa_t *spa);
760extern uint64_t spa_syncing_txg(spa_t *spa);
761extern uint64_t spa_version(spa_t *spa);
762extern pool_state_t spa_state(spa_t *spa);
763extern spa_load_state_t spa_load_state(spa_t *spa);
764extern uint64_t spa_freeze_txg(spa_t *spa);
765extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize);
766extern uint64_t spa_get_dspace(spa_t *spa);
767extern uint64_t spa_get_slop_space(spa_t *spa);
768extern void spa_update_dspace(spa_t *spa);
769extern uint64_t spa_version(spa_t *spa);
770extern boolean_t spa_deflate(spa_t *spa);
771extern metaslab_class_t *spa_normal_class(spa_t *spa);
772extern metaslab_class_t *spa_log_class(spa_t *spa);
773extern void spa_evicting_os_register(spa_t *, objset_t *os);
774extern void spa_evicting_os_deregister(spa_t *, objset_t *os);
775extern void spa_evicting_os_wait(spa_t *spa);
776extern int spa_max_replication(spa_t *spa);
777extern int spa_prev_software_version(spa_t *spa);
778extern int spa_busy(void);
779extern uint8_t spa_get_failmode(spa_t *spa);
780extern boolean_t spa_suspended(spa_t *spa);
781extern uint64_t spa_bootfs(spa_t *spa);
782extern uint64_t spa_delegation(spa_t *spa);
783extern objset_t *spa_meta_objset(spa_t *spa);
784extern uint64_t spa_deadman_synctime(spa_t *spa);
785
786/* Miscellaneous support routines */
787extern void spa_activate_mos_feature(spa_t *spa, const char *feature,
788    dmu_tx_t *tx);
789extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature);
790extern int spa_rename(const char *oldname, const char *newname);
791extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid);
792extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid);
793extern char *spa_strdup(const char *);
794extern void spa_strfree(char *);
795extern uint64_t spa_get_random(uint64_t range);
796extern uint64_t spa_generate_guid(spa_t *spa);
797extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp);
798extern void spa_freeze(spa_t *spa);
799extern int spa_change_guid(spa_t *spa);
800extern void spa_upgrade(spa_t *spa, uint64_t version);
801extern void spa_evict_all(void);
802extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid,
803    boolean_t l2cache);
804extern boolean_t spa_has_spare(spa_t *, uint64_t guid);
805extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva);
806extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp);
807extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp);
808extern boolean_t spa_has_slogs(spa_t *spa);
809extern boolean_t spa_is_root(spa_t *spa);
810extern boolean_t spa_writeable(spa_t *spa);
811extern boolean_t spa_has_pending_synctask(spa_t *spa);
812extern int spa_maxblocksize(spa_t *spa);
813extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp);
814
815extern int spa_mode(spa_t *spa);
816extern uint64_t strtonum(const char *str, char **nptr);
817
818extern char *spa_his_ievent_table[];
819
820extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx);
821extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read,
822    char *his_buf);
823extern int spa_history_log(spa_t *spa, const char *his_buf);
824extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl);
825extern void spa_history_log_version(spa_t *spa, const char *operation);
826extern void spa_history_log_internal(spa_t *spa, const char *operation,
827    dmu_tx_t *tx, const char *fmt, ...);
828extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op,
829    dmu_tx_t *tx, const char *fmt, ...);
830extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
831    dmu_tx_t *tx, const char *fmt, ...);
832
833/* error handling */
834struct zbookmark_phys;
835extern void spa_log_error(spa_t *spa, zio_t *zio);
836extern void zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd,
837    zio_t *zio, uint64_t stateoroffset, uint64_t length);
838extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
839extern void zfs_post_state_change(spa_t *spa, vdev_t *vd);
840extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
841extern uint64_t spa_get_errlog_size(spa_t *spa);
842extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
843extern void spa_errlog_rotate(spa_t *spa);
844extern void spa_errlog_drain(spa_t *spa);
845extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
846extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
847
848/* vdev cache */
849extern void vdev_cache_stat_init(void);
850extern void vdev_cache_stat_fini(void);
851
852/* Initialization and termination */
853extern void spa_init(int flags);
854extern void spa_fini(void);
855extern void spa_boot_init();
856
857/* properties */
858extern int spa_prop_set(spa_t *spa, nvlist_t *nvp);
859extern int spa_prop_get(spa_t *spa, nvlist_t **nvp);
860extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx);
861extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t);
862
863/* asynchronous event notification */
864extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name);
865
866#ifdef ZFS_DEBUG
867#define	dprintf_bp(bp, fmt, ...) do {				\
868	if (zfs_flags & ZFS_DEBUG_DPRINTF) {			\
869	char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP);	\
870	snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp));	\
871	dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf);		\
872	kmem_free(__blkbuf, BP_SPRINTF_LEN);			\
873	} \
874_NOTE(CONSTCOND) } while (0)
875#else
876#define	dprintf_bp(bp, fmt, ...)
877#endif
878
879extern boolean_t spa_debug_enabled(spa_t *spa);
880#define	spa_dbgmsg(spa, ...)			\
881{						\
882	if (spa_debug_enabled(spa))		\
883		zfs_dbgmsg(__VA_ARGS__);	\
884}
885
886extern int spa_mode_global;			/* mode, e.g. FREAD | FWRITE */
887
888#ifdef	__cplusplus
889}
890#endif
891
892#endif	/* _SYS_SPA_H */
893