spa.h revision 266122 
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) 2013 by Delphix. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
25 */
26
27#ifndef _SYS_SPA_H
28#define	_SYS_SPA_H
29
30#include <sys/avl.h>
31#include <sys/zfs_context.h>
32#include <sys/nvpair.h>
33#include <sys/sysmacros.h>
34#include <sys/types.h>
35#include <sys/fs/zfs.h>
36
37#ifdef	__cplusplus
38extern "C" {
39#endif
40
41/*
42 * Forward references that lots of things need.
43 */
44typedef struct spa spa_t;
45typedef struct vdev vdev_t;
46typedef struct metaslab metaslab_t;
47typedef struct metaslab_group metaslab_group_t;
48typedef struct metaslab_class metaslab_class_t;
49typedef struct zio zio_t;
50typedef struct zilog zilog_t;
51typedef struct spa_aux_vdev spa_aux_vdev_t;
52typedef struct ddt ddt_t;
53typedef struct ddt_entry ddt_entry_t;
54struct dsl_pool;
55struct dsl_dataset;
56
57/*
58 * General-purpose 32-bit and 64-bit bitfield encodings.
59 */
60#define	BF32_DECODE(x, low, len)	P2PHASE((x) >> (low), 1U << (len))
61#define	BF64_DECODE(x, low, len)	P2PHASE((x) >> (low), 1ULL << (len))
62#define	BF32_ENCODE(x, low, len)	(P2PHASE((x), 1U << (len)) << (low))
63#define	BF64_ENCODE(x, low, len)	(P2PHASE((x), 1ULL << (len)) << (low))
64
65#define	BF32_GET(x, low, len)		BF32_DECODE(x, low, len)
66#define	BF64_GET(x, low, len)		BF64_DECODE(x, low, len)
67
68#define	BF32_SET(x, low, len, val) do { \
69	ASSERT3U(val, <, 1U << (len)); \
70	ASSERT3U(low + len, <=, 32); \
71	(x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \
72_NOTE(CONSTCOND) } while (0)
73
74#define	BF64_SET(x, low, len, val) do { \
75	ASSERT3U(val, <, 1ULL << (len)); \
76	ASSERT3U(low + len, <=, 64); \
77	((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \
78_NOTE(CONSTCOND) } while (0)
79
80#define	BF32_GET_SB(x, low, len, shift, bias)	\
81	((BF32_GET(x, low, len) + (bias)) << (shift))
82#define	BF64_GET_SB(x, low, len, shift, bias)	\
83	((BF64_GET(x, low, len) + (bias)) << (shift))
84
85#define	BF32_SET_SB(x, low, len, shift, bias, val) do { \
86	ASSERT(IS_P2ALIGNED(val, 1U << shift)); \
87	ASSERT3S((val) >> (shift), >=, bias); \
88	BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \
89_NOTE(CONSTCOND) } while (0)
90#define	BF64_SET_SB(x, low, len, shift, bias, val) do { \
91	ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \
92	ASSERT3S((val) >> (shift), >=, bias); \
93	BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \
94_NOTE(CONSTCOND) } while (0)
95
96/*
97 * We currently support nine block sizes, from 512 bytes to 128K.
98 * We could go higher, but the benefits are near-zero and the cost
99 * of COWing a giant block to modify one byte would become excessive.
100 */
101#define	SPA_MINBLOCKSHIFT	9
102#define	SPA_MAXBLOCKSHIFT	17
103#define	SPA_MINBLOCKSIZE	(1ULL << SPA_MINBLOCKSHIFT)
104#define	SPA_MAXBLOCKSIZE	(1ULL << SPA_MAXBLOCKSHIFT)
105
106#define	SPA_BLOCKSIZES		(SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1)
107
108/*
109 * Default maximum supported logical ashift.
110 *
111 * The current 8k allocation block size limit is due to the 8k
112 * aligned/sized operations performed by vdev_probe() on
113 * vdev_label->vl_pad2.  Using another "safe region" for these tests
114 * would allow the limit to be raised to 16k, at the expense of
115 * only having 8 available uberblocks in the label area.
116 */
117#define	SPA_MAXASHIFT		13
118
119/*
120 * Default minimum supported logical ashift.
121 */
122#define SPA_MINASHIFT		SPA_MINBLOCKSHIFT
123
124/*
125 * Size of block to hold the configuration data (a packed nvlist)
126 */
127#define	SPA_CONFIG_BLOCKSIZE	(1ULL << 14)
128
129/*
130 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
131 * The ASIZE encoding should be at least 64 times larger (6 more bits)
132 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
133 * overhead, three DVAs per bp, plus one more bit in case we do anything
134 * else that expands the ASIZE.
135 */
136#define	SPA_LSIZEBITS		16	/* LSIZE up to 32M (2^16 * 512)	*/
137#define	SPA_PSIZEBITS		16	/* PSIZE up to 32M (2^16 * 512)	*/
138#define	SPA_ASIZEBITS		24	/* ASIZE up to 64 times larger	*/
139
140/*
141 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
142 * The members of the dva_t should be considered opaque outside the SPA.
143 */
144typedef struct dva {
145	uint64_t	dva_word[2];
146} dva_t;
147
148/*
149 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
150 */
151typedef struct zio_cksum {
152	uint64_t	zc_word[4];
153} zio_cksum_t;
154
155/*
156 * Each block is described by its DVAs, time of birth, checksum, etc.
157 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
158 *
159 *	64	56	48	40	32	24	16	8	0
160 *	+-------+-------+-------+-------+-------+-------+-------+-------+
161 * 0	|		vdev1		| GRID  |	  ASIZE		|
162 *	+-------+-------+-------+-------+-------+-------+-------+-------+
163 * 1	|G|			 offset1				|
164 *	+-------+-------+-------+-------+-------+-------+-------+-------+
165 * 2	|		vdev2		| GRID  |	  ASIZE		|
166 *	+-------+-------+-------+-------+-------+-------+-------+-------+
167 * 3	|G|			 offset2				|
168 *	+-------+-------+-------+-------+-------+-------+-------+-------+
169 * 4	|		vdev3		| GRID  |	  ASIZE		|
170 *	+-------+-------+-------+-------+-------+-------+-------+-------+
171 * 5	|G|			 offset3				|
172 *	+-------+-------+-------+-------+-------+-------+-------+-------+
173 * 6	|BDX|lvl| type	| cksum | comp	|     PSIZE	|     LSIZE	|
174 *	+-------+-------+-------+-------+-------+-------+-------+-------+
175 * 7	|			padding					|
176 *	+-------+-------+-------+-------+-------+-------+-------+-------+
177 * 8	|			padding					|
178 *	+-------+-------+-------+-------+-------+-------+-------+-------+
179 * 9	|			physical birth txg			|
180 *	+-------+-------+-------+-------+-------+-------+-------+-------+
181 * a	|			logical birth txg			|
182 *	+-------+-------+-------+-------+-------+-------+-------+-------+
183 * b	|			fill count				|
184 *	+-------+-------+-------+-------+-------+-------+-------+-------+
185 * c	|			checksum[0]				|
186 *	+-------+-------+-------+-------+-------+-------+-------+-------+
187 * d	|			checksum[1]				|
188 *	+-------+-------+-------+-------+-------+-------+-------+-------+
189 * e	|			checksum[2]				|
190 *	+-------+-------+-------+-------+-------+-------+-------+-------+
191 * f	|			checksum[3]				|
192 *	+-------+-------+-------+-------+-------+-------+-------+-------+
193 *
194 * Legend:
195 *
196 * vdev		virtual device ID
197 * offset	offset into virtual device
198 * LSIZE	logical size
199 * PSIZE	physical size (after compression)
200 * ASIZE	allocated size (including RAID-Z parity and gang block headers)
201 * GRID		RAID-Z layout information (reserved for future use)
202 * cksum	checksum function
203 * comp		compression function
204 * G		gang block indicator
205 * B		byteorder (endianness)
206 * D		dedup
207 * X		unused
208 * lvl		level of indirection
209 * type		DMU object type
210 * phys birth	txg of block allocation; zero if same as logical birth txg
211 * log. birth	transaction group in which the block was logically born
212 * fill count	number of non-zero blocks under this bp
213 * checksum[4]	256-bit checksum of the data this bp describes
214 */
215#define	SPA_BLKPTRSHIFT	7		/* blkptr_t is 128 bytes	*/
216#define	SPA_DVAS_PER_BP	3		/* Number of DVAs in a bp	*/
217
218/*
219 * A block is a hole when it has either 1) never been written to, or
220 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
221 * without physically allocating disk space. Holes are represented in the
222 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is
223 * done through the BP_IS_HOLE macro. For holes, the logical size, level,
224 * DMU object type, and birth times are all also stored for holes that
225 * were written to at some point (i.e. were punched after having been filled).
226 */
227typedef struct blkptr {
228	dva_t		blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
229	uint64_t	blk_prop;	/* size, compression, type, etc	    */
230	uint64_t	blk_pad[2];	/* Extra space for the future	    */
231	uint64_t	blk_phys_birth;	/* txg when block was allocated	    */
232	uint64_t	blk_birth;	/* transaction group at birth	    */
233	uint64_t	blk_fill;	/* fill count			    */
234	zio_cksum_t	blk_cksum;	/* 256-bit checksum		    */
235} blkptr_t;
236
237/*
238 * Macros to get and set fields in a bp or DVA.
239 */
240#define	DVA_GET_ASIZE(dva)	\
241	BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
242#define	DVA_SET_ASIZE(dva, x)	\
243	BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
244	SPA_MINBLOCKSHIFT, 0, x)
245
246#define	DVA_GET_GRID(dva)	BF64_GET((dva)->dva_word[0], 24, 8)
247#define	DVA_SET_GRID(dva, x)	BF64_SET((dva)->dva_word[0], 24, 8, x)
248
249#define	DVA_GET_VDEV(dva)	BF64_GET((dva)->dva_word[0], 32, 32)
250#define	DVA_SET_VDEV(dva, x)	BF64_SET((dva)->dva_word[0], 32, 32, x)
251
252#define	DVA_GET_OFFSET(dva)	\
253	BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
254#define	DVA_SET_OFFSET(dva, x)	\
255	BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
256
257#define	DVA_GET_GANG(dva)	BF64_GET((dva)->dva_word[1], 63, 1)
258#define	DVA_SET_GANG(dva, x)	BF64_SET((dva)->dva_word[1], 63, 1, x)
259
260#define	BP_GET_LSIZE(bp)	\
261	BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)
262#define	BP_SET_LSIZE(bp, x)	\
263	BF64_SET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x)
264
265#define	BP_GET_PSIZE(bp)	\
266	BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1)
267#define	BP_SET_PSIZE(bp, x)	\
268	BF64_SET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x)
269
270#define	BP_GET_COMPRESS(bp)		BF64_GET((bp)->blk_prop, 32, 8)
271#define	BP_SET_COMPRESS(bp, x)		BF64_SET((bp)->blk_prop, 32, 8, x)
272
273#define	BP_GET_CHECKSUM(bp)		BF64_GET((bp)->blk_prop, 40, 8)
274#define	BP_SET_CHECKSUM(bp, x)		BF64_SET((bp)->blk_prop, 40, 8, x)
275
276#define	BP_GET_TYPE(bp)			BF64_GET((bp)->blk_prop, 48, 8)
277#define	BP_SET_TYPE(bp, x)		BF64_SET((bp)->blk_prop, 48, 8, x)
278
279#define	BP_GET_LEVEL(bp)		BF64_GET((bp)->blk_prop, 56, 5)
280#define	BP_SET_LEVEL(bp, x)		BF64_SET((bp)->blk_prop, 56, 5, x)
281
282#define	BP_GET_PROP_BIT_61(bp)		BF64_GET((bp)->blk_prop, 61, 1)
283#define	BP_SET_PROP_BIT_61(bp, x)	BF64_SET((bp)->blk_prop, 61, 1, x)
284
285#define	BP_GET_DEDUP(bp)		BF64_GET((bp)->blk_prop, 62, 1)
286#define	BP_SET_DEDUP(bp, x)		BF64_SET((bp)->blk_prop, 62, 1, x)
287
288#define	BP_GET_BYTEORDER(bp)		BF64_GET((bp)->blk_prop, 63, 1)
289#define	BP_SET_BYTEORDER(bp, x)		BF64_SET((bp)->blk_prop, 63, 1, x)
290
291#define	BP_PHYSICAL_BIRTH(bp)		\
292	((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
293
294#define	BP_SET_BIRTH(bp, logical, physical)	\
295{						\
296	(bp)->blk_birth = (logical);		\
297	(bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
298}
299
300#define	BP_GET_ASIZE(bp)	\
301	(DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
302		DVA_GET_ASIZE(&(bp)->blk_dva[2]))
303
304#define	BP_GET_UCSIZE(bp) \
305	((BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) ? \
306	BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
307
308#define	BP_GET_NDVAS(bp)	\
309	(!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
310	!!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
311	!!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
312
313#define	BP_COUNT_GANG(bp)	\
314	(DVA_GET_GANG(&(bp)->blk_dva[0]) + \
315	DVA_GET_GANG(&(bp)->blk_dva[1]) + \
316	DVA_GET_GANG(&(bp)->blk_dva[2]))
317
318#define	DVA_EQUAL(dva1, dva2)	\
319	((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
320	(dva1)->dva_word[0] == (dva2)->dva_word[0])
321
322#define	BP_EQUAL(bp1, bp2)	\
323	(BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) &&	\
324	DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) &&	\
325	DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) &&	\
326	DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
327
328#define	ZIO_CHECKSUM_EQUAL(zc1, zc2) \
329	(0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
330	((zc1).zc_word[1] - (zc2).zc_word[1]) | \
331	((zc1).zc_word[2] - (zc2).zc_word[2]) | \
332	((zc1).zc_word[3] - (zc2).zc_word[3])))
333
334#define	DVA_IS_VALID(dva)	(DVA_GET_ASIZE(dva) != 0)
335
336#define	ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3)	\
337{						\
338	(zcp)->zc_word[0] = w0;			\
339	(zcp)->zc_word[1] = w1;			\
340	(zcp)->zc_word[2] = w2;			\
341	(zcp)->zc_word[3] = w3;			\
342}
343
344#define	BP_IDENTITY(bp)		(&(bp)->blk_dva[0])
345#define	BP_IS_GANG(bp)		DVA_GET_GANG(BP_IDENTITY(bp))
346#define	DVA_IS_EMPTY(dva)	((dva)->dva_word[0] == 0ULL &&	\
347				(dva)->dva_word[1] == 0ULL)
348#define	BP_IS_HOLE(bp)		DVA_IS_EMPTY(BP_IDENTITY(bp))
349
350/* BP_IS_RAIDZ(bp) assumes no block compression */
351#define	BP_IS_RAIDZ(bp)		(DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
352				BP_GET_PSIZE(bp))
353
354#define	BP_ZERO(bp)				\
355{						\
356	(bp)->blk_dva[0].dva_word[0] = 0;	\
357	(bp)->blk_dva[0].dva_word[1] = 0;	\
358	(bp)->blk_dva[1].dva_word[0] = 0;	\
359	(bp)->blk_dva[1].dva_word[1] = 0;	\
360	(bp)->blk_dva[2].dva_word[0] = 0;	\
361	(bp)->blk_dva[2].dva_word[1] = 0;	\
362	(bp)->blk_prop = 0;			\
363	(bp)->blk_pad[0] = 0;			\
364	(bp)->blk_pad[1] = 0;			\
365	(bp)->blk_phys_birth = 0;		\
366	(bp)->blk_birth = 0;			\
367	(bp)->blk_fill = 0;			\
368	ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0);	\
369}
370
371#if BYTE_ORDER == _BIG_ENDIAN
372#define	ZFS_HOST_BYTEORDER	(0ULL)
373#else
374#define	ZFS_HOST_BYTEORDER	(1ULL)
375#endif
376
377#define	BP_SHOULD_BYTESWAP(bp)	(BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
378
379#define	BP_SPRINTF_LEN	320
380
381/*
382 * This macro allows code sharing between zfs, libzpool, and mdb.
383 * 'func' is either snprintf() or mdb_snprintf().
384 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
385 */
386#define	SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
387{									\
388	static const char *copyname[] =					\
389	    { "zero", "single", "double", "triple" };			\
390	int len = 0;							\
391	int copies = 0;							\
392									\
393	if (bp == NULL) {						\
394		len += func(buf + len, size - len, "<NULL>");		\
395	} else if (BP_IS_HOLE(bp)) {					\
396		len += func(buf + len, size - len, "<hole>");		\
397		if (bp->blk_birth > 0) {				\
398			len += func(buf + len, size - len,		\
399			    " birth=%lluL",				\
400			    (u_longlong_t)bp->blk_birth);		\
401		}							\
402	} else {							\
403		for (int d = 0; d < BP_GET_NDVAS(bp); d++) {		\
404			const dva_t *dva = &bp->blk_dva[d];		\
405			if (DVA_IS_VALID(dva))				\
406				copies++;				\
407			len += func(buf + len, size - len,		\
408			    "DVA[%d]=<%llu:%llx:%llx>%c", d,		\
409			    (u_longlong_t)DVA_GET_VDEV(dva),		\
410			    (u_longlong_t)DVA_GET_OFFSET(dva),		\
411			    (u_longlong_t)DVA_GET_ASIZE(dva),		\
412			    ws);					\
413		}							\
414		if (BP_IS_GANG(bp) &&					\
415		    DVA_GET_ASIZE(&bp->blk_dva[2]) <=			\
416		    DVA_GET_ASIZE(&bp->blk_dva[1]) / 2)			\
417			copies--;					\
418		len += func(buf + len, size - len,			\
419		    "[L%llu %s] %s %s %s %s %s %s%c"			\
420		    "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c"	\
421		    "cksum=%llx:%llx:%llx:%llx",			\
422		    (u_longlong_t)BP_GET_LEVEL(bp),			\
423		    type,						\
424		    checksum,						\
425		    compress,						\
426		    BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE",		\
427		    BP_IS_GANG(bp) ? "gang" : "contiguous",		\
428		    BP_GET_DEDUP(bp) ? "dedup" : "unique",		\
429		    copyname[copies],					\
430		    ws,							\
431		    (u_longlong_t)BP_GET_LSIZE(bp),			\
432		    (u_longlong_t)BP_GET_PSIZE(bp),			\
433		    (u_longlong_t)bp->blk_birth,			\
434		    (u_longlong_t)BP_PHYSICAL_BIRTH(bp),		\
435		    (u_longlong_t)bp->blk_fill,				\
436		    ws,							\
437		    (u_longlong_t)bp->blk_cksum.zc_word[0],		\
438		    (u_longlong_t)bp->blk_cksum.zc_word[1],		\
439		    (u_longlong_t)bp->blk_cksum.zc_word[2],		\
440		    (u_longlong_t)bp->blk_cksum.zc_word[3]);		\
441	}								\
442	ASSERT(len < size);						\
443}
444
445#include <sys/dmu.h>
446
447#define	BP_GET_BUFC_TYPE(bp)						\
448	(((BP_GET_LEVEL(bp) > 0) || (DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))) ? \
449	ARC_BUFC_METADATA : ARC_BUFC_DATA)
450
451typedef enum spa_import_type {
452	SPA_IMPORT_EXISTING,
453	SPA_IMPORT_ASSEMBLE
454} spa_import_type_t;
455
456/* state manipulation functions */
457extern int spa_open(const char *pool, spa_t **, void *tag);
458extern int spa_open_rewind(const char *pool, spa_t **, void *tag,
459    nvlist_t *policy, nvlist_t **config);
460extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot,
461    size_t buflen);
462extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props,
463    nvlist_t *zplprops);
464#if defined(sun)
465extern int spa_import_rootpool(char *devpath, char *devid);
466#else
467extern int spa_import_rootpool(const char *name);
468#endif
469extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props,
470    uint64_t flags);
471extern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
472extern int spa_destroy(char *pool);
473extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force,
474    boolean_t hardforce);
475extern int spa_reset(char *pool);
476extern void spa_async_request(spa_t *spa, int flag);
477extern void spa_async_unrequest(spa_t *spa, int flag);
478extern void spa_async_suspend(spa_t *spa);
479extern void spa_async_resume(spa_t *spa);
480extern spa_t *spa_inject_addref(char *pool);
481extern void spa_inject_delref(spa_t *spa);
482extern void spa_scan_stat_init(spa_t *spa);
483extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps);
484
485#define	SPA_ASYNC_CONFIG_UPDATE	0x01
486#define	SPA_ASYNC_REMOVE	0x02
487#define	SPA_ASYNC_PROBE		0x04
488#define	SPA_ASYNC_RESILVER_DONE	0x08
489#define	SPA_ASYNC_RESILVER	0x10
490#define	SPA_ASYNC_AUTOEXPAND	0x20
491#define	SPA_ASYNC_REMOVE_DONE	0x40
492#define	SPA_ASYNC_REMOVE_STOP	0x80
493
494/*
495 * Controls the behavior of spa_vdev_remove().
496 */
497#define	SPA_REMOVE_UNSPARE	0x01
498#define	SPA_REMOVE_DONE		0x02
499
500/* device manipulation */
501extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
502extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot,
503    int replacing);
504extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
505    int replace_done);
506extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare);
507extern boolean_t spa_vdev_remove_active(spa_t *spa);
508extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
509extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru);
510extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
511    nvlist_t *props, boolean_t exp);
512
513/* spare state (which is global across all pools) */
514extern void spa_spare_add(vdev_t *vd);
515extern void spa_spare_remove(vdev_t *vd);
516extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt);
517extern void spa_spare_activate(vdev_t *vd);
518
519/* L2ARC state (which is global across all pools) */
520extern void spa_l2cache_add(vdev_t *vd);
521extern void spa_l2cache_remove(vdev_t *vd);
522extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
523extern void spa_l2cache_activate(vdev_t *vd);
524extern void spa_l2cache_drop(spa_t *spa);
525
526/* scanning */
527extern int spa_scan(spa_t *spa, pool_scan_func_t func);
528extern int spa_scan_stop(spa_t *spa);
529
530/* spa syncing */
531extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
532extern void spa_sync_allpools(void);
533
534/* spa namespace global mutex */
535extern kmutex_t spa_namespace_lock;
536
537/*
538 * SPA configuration functions in spa_config.c
539 */
540
541#define	SPA_CONFIG_UPDATE_POOL	0
542#define	SPA_CONFIG_UPDATE_VDEVS	1
543
544extern void spa_config_sync(spa_t *, boolean_t, boolean_t);
545extern void spa_config_load(void);
546extern nvlist_t *spa_all_configs(uint64_t *);
547extern void spa_config_set(spa_t *spa, nvlist_t *config);
548extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
549    int getstats);
550extern void spa_config_update(spa_t *spa, int what);
551
552/*
553 * Miscellaneous SPA routines in spa_misc.c
554 */
555
556/* Namespace manipulation */
557extern spa_t *spa_lookup(const char *name);
558extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot);
559extern void spa_remove(spa_t *spa);
560extern spa_t *spa_next(spa_t *prev);
561
562/* Refcount functions */
563extern void spa_open_ref(spa_t *spa, void *tag);
564extern void spa_close(spa_t *spa, void *tag);
565extern boolean_t spa_refcount_zero(spa_t *spa);
566
567#define	SCL_NONE	0x00
568#define	SCL_CONFIG	0x01
569#define	SCL_STATE	0x02
570#define	SCL_L2ARC	0x04		/* hack until L2ARC 2.0 */
571#define	SCL_ALLOC	0x08
572#define	SCL_ZIO		0x10
573#define	SCL_FREE	0x20
574#define	SCL_VDEV	0x40
575#define	SCL_LOCKS	7
576#define	SCL_ALL		((1 << SCL_LOCKS) - 1)
577#define	SCL_STATE_ALL	(SCL_STATE | SCL_L2ARC | SCL_ZIO)
578
579/* Pool configuration locks */
580extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw);
581extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw);
582extern void spa_config_exit(spa_t *spa, int locks, void *tag);
583extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
584
585/* Pool vdev add/remove lock */
586extern uint64_t spa_vdev_enter(spa_t *spa);
587extern uint64_t spa_vdev_config_enter(spa_t *spa);
588extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg,
589    int error, char *tag);
590extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error);
591
592/* Pool vdev state change lock */
593extern void spa_vdev_state_enter(spa_t *spa, int oplock);
594extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error);
595
596/* Log state */
597typedef enum spa_log_state {
598	SPA_LOG_UNKNOWN = 0,	/* unknown log state */
599	SPA_LOG_MISSING,	/* missing log(s) */
600	SPA_LOG_CLEAR,		/* clear the log(s) */
601	SPA_LOG_GOOD,		/* log(s) are good */
602} spa_log_state_t;
603
604extern spa_log_state_t spa_get_log_state(spa_t *spa);
605extern void spa_set_log_state(spa_t *spa, spa_log_state_t state);
606extern int spa_offline_log(spa_t *spa);
607
608/* Log claim callback */
609extern void spa_claim_notify(zio_t *zio);
610
611/* Accessor functions */
612extern boolean_t spa_shutting_down(spa_t *spa);
613extern struct dsl_pool *spa_get_dsl(spa_t *spa);
614extern boolean_t spa_is_initializing(spa_t *spa);
615extern blkptr_t *spa_get_rootblkptr(spa_t *spa);
616extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
617extern void spa_altroot(spa_t *, char *, size_t);
618extern int spa_sync_pass(spa_t *spa);
619extern char *spa_name(spa_t *spa);
620extern uint64_t spa_guid(spa_t *spa);
621extern uint64_t spa_load_guid(spa_t *spa);
622extern uint64_t spa_last_synced_txg(spa_t *spa);
623extern uint64_t spa_first_txg(spa_t *spa);
624extern uint64_t spa_syncing_txg(spa_t *spa);
625extern uint64_t spa_version(spa_t *spa);
626extern pool_state_t spa_state(spa_t *spa);
627extern spa_load_state_t spa_load_state(spa_t *spa);
628extern uint64_t spa_freeze_txg(spa_t *spa);
629extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize);
630extern uint64_t spa_get_dspace(spa_t *spa);
631extern void spa_update_dspace(spa_t *spa);
632extern uint64_t spa_version(spa_t *spa);
633extern boolean_t spa_deflate(spa_t *spa);
634extern metaslab_class_t *spa_normal_class(spa_t *spa);
635extern metaslab_class_t *spa_log_class(spa_t *spa);
636extern int spa_max_replication(spa_t *spa);
637extern int spa_prev_software_version(spa_t *spa);
638extern int spa_busy(void);
639extern uint8_t spa_get_failmode(spa_t *spa);
640extern boolean_t spa_suspended(spa_t *spa);
641extern uint64_t spa_bootfs(spa_t *spa);
642extern uint64_t spa_delegation(spa_t *spa);
643extern objset_t *spa_meta_objset(spa_t *spa);
644extern uint64_t spa_deadman_synctime(spa_t *spa);
645
646/* Miscellaneous support routines */
647extern void spa_activate_mos_feature(spa_t *spa, const char *feature,
648    dmu_tx_t *tx);
649extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature);
650extern int spa_rename(const char *oldname, const char *newname);
651extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid);
652extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid);
653extern char *spa_strdup(const char *);
654extern void spa_strfree(char *);
655extern uint64_t spa_get_random(uint64_t range);
656extern uint64_t spa_generate_guid(spa_t *spa);
657extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp);
658extern void spa_freeze(spa_t *spa);
659extern int spa_change_guid(spa_t *spa);
660extern void spa_upgrade(spa_t *spa, uint64_t version);
661extern void spa_evict_all(void);
662extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid,
663    boolean_t l2cache);
664extern boolean_t spa_has_spare(spa_t *, uint64_t guid);
665extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva);
666extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp);
667extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp);
668extern boolean_t spa_has_slogs(spa_t *spa);
669extern boolean_t spa_is_root(spa_t *spa);
670extern boolean_t spa_writeable(spa_t *spa);
671
672extern int spa_mode(spa_t *spa);
673extern uint64_t zfs_strtonum(const char *str, char **nptr);
674#define	strtonum(str, nptr)	zfs_strtonum((str), (nptr))
675
676extern char *spa_his_ievent_table[];
677
678extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx);
679extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read,
680    char *his_buf);
681extern int spa_history_log(spa_t *spa, const char *his_buf);
682extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl);
683extern void spa_history_log_version(spa_t *spa, const char *operation);
684extern void spa_history_log_internal(spa_t *spa, const char *operation,
685    dmu_tx_t *tx, const char *fmt, ...);
686extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op,
687    dmu_tx_t *tx, const char *fmt, ...);
688extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
689    dmu_tx_t *tx, const char *fmt, ...);
690
691/* error handling */
692struct zbookmark;
693extern void spa_log_error(spa_t *spa, zio_t *zio);
694extern void zfs_ereport_post(const char *cls, spa_t *spa, vdev_t *vd,
695    zio_t *zio, uint64_t stateoroffset, uint64_t length);
696extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
697extern void zfs_post_state_change(spa_t *spa, vdev_t *vd);
698extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
699extern uint64_t spa_get_errlog_size(spa_t *spa);
700extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
701extern void spa_errlog_rotate(spa_t *spa);
702extern void spa_errlog_drain(spa_t *spa);
703extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
704extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
705
706/* vdev cache */
707extern void vdev_cache_stat_init(void);
708extern void vdev_cache_stat_fini(void);
709
710/* Initialization and termination */
711extern void spa_init(int flags);
712extern void spa_fini(void);
713extern void spa_boot_init();
714
715/* properties */
716extern int spa_prop_set(spa_t *spa, nvlist_t *nvp);
717extern int spa_prop_get(spa_t *spa, nvlist_t **nvp);
718extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx);
719extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t);
720
721/* asynchronous event notification */
722extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name);
723
724#ifdef ZFS_DEBUG
725#define	dprintf_bp(bp, fmt, ...) do {				\
726	if (zfs_flags & ZFS_DEBUG_DPRINTF) {			\
727	char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP);	\
728	snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp));	\
729	dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf);		\
730	kmem_free(__blkbuf, BP_SPRINTF_LEN);			\
731	} \
732_NOTE(CONSTCOND) } while (0)
733#else
734#define	dprintf_bp(bp, fmt, ...)
735#endif
736
737extern boolean_t spa_debug_enabled(spa_t *spa);
738#define	spa_dbgmsg(spa, ...)			\
739{						\
740	if (spa_debug_enabled(spa))		\
741		zfs_dbgmsg(__VA_ARGS__);	\
742}
743
744extern int spa_mode_global;			/* mode, e.g. FREAD | FWRITE */
745
746#ifdef	__cplusplus
747}
748#endif
749
750#endif	/* _SYS_SPA_H */
751