spa.h revision 43466aae47bfcd2ad9bf501faec8e75c08095e4f
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 * Size of block to hold the configuration data (a packed nvlist)
110 */
111#define	SPA_CONFIG_BLOCKSIZE	(1ULL << 14)
112
113/*
114 * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
115 * The ASIZE encoding should be at least 64 times larger (6 more bits)
116 * to support up to 4-way RAID-Z mirror mode with worst-case gang block
117 * overhead, three DVAs per bp, plus one more bit in case we do anything
118 * else that expands the ASIZE.
119 */
120#define	SPA_LSIZEBITS		16	/* LSIZE up to 32M (2^16 * 512)	*/
121#define	SPA_PSIZEBITS		16	/* PSIZE up to 32M (2^16 * 512)	*/
122#define	SPA_ASIZEBITS		24	/* ASIZE up to 64 times larger	*/
123
124/*
125 * All SPA data is represented by 128-bit data virtual addresses (DVAs).
126 * The members of the dva_t should be considered opaque outside the SPA.
127 */
128typedef struct dva {
129	uint64_t	dva_word[2];
130} dva_t;
131
132/*
133 * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
134 */
135typedef struct zio_cksum {
136	uint64_t	zc_word[4];
137} zio_cksum_t;
138
139/*
140 * Each block is described by its DVAs, time of birth, checksum, etc.
141 * The word-by-word, bit-by-bit layout of the blkptr is as follows:
142 *
143 *	64	56	48	40	32	24	16	8	0
144 *	+-------+-------+-------+-------+-------+-------+-------+-------+
145 * 0	|		vdev1		| GRID  |	  ASIZE		|
146 *	+-------+-------+-------+-------+-------+-------+-------+-------+
147 * 1	|G|			 offset1				|
148 *	+-------+-------+-------+-------+-------+-------+-------+-------+
149 * 2	|		vdev2		| GRID  |	  ASIZE		|
150 *	+-------+-------+-------+-------+-------+-------+-------+-------+
151 * 3	|G|			 offset2				|
152 *	+-------+-------+-------+-------+-------+-------+-------+-------+
153 * 4	|		vdev3		| GRID  |	  ASIZE		|
154 *	+-------+-------+-------+-------+-------+-------+-------+-------+
155 * 5	|G|			 offset3				|
156 *	+-------+-------+-------+-------+-------+-------+-------+-------+
157 * 6	|BDX|lvl| type	| cksum | comp	|     PSIZE	|     LSIZE	|
158 *	+-------+-------+-------+-------+-------+-------+-------+-------+
159 * 7	|			padding					|
160 *	+-------+-------+-------+-------+-------+-------+-------+-------+
161 * 8	|			padding					|
162 *	+-------+-------+-------+-------+-------+-------+-------+-------+
163 * 9	|			physical birth txg			|
164 *	+-------+-------+-------+-------+-------+-------+-------+-------+
165 * a	|			logical birth txg			|
166 *	+-------+-------+-------+-------+-------+-------+-------+-------+
167 * b	|			fill count				|
168 *	+-------+-------+-------+-------+-------+-------+-------+-------+
169 * c	|			checksum[0]				|
170 *	+-------+-------+-------+-------+-------+-------+-------+-------+
171 * d	|			checksum[1]				|
172 *	+-------+-------+-------+-------+-------+-------+-------+-------+
173 * e	|			checksum[2]				|
174 *	+-------+-------+-------+-------+-------+-------+-------+-------+
175 * f	|			checksum[3]				|
176 *	+-------+-------+-------+-------+-------+-------+-------+-------+
177 *
178 * Legend:
179 *
180 * vdev		virtual device ID
181 * offset	offset into virtual device
182 * LSIZE	logical size
183 * PSIZE	physical size (after compression)
184 * ASIZE	allocated size (including RAID-Z parity and gang block headers)
185 * GRID		RAID-Z layout information (reserved for future use)
186 * cksum	checksum function
187 * comp		compression function
188 * G		gang block indicator
189 * B		byteorder (endianness)
190 * D		dedup
191 * X		unused
192 * lvl		level of indirection
193 * type		DMU object type
194 * phys birth	txg of block allocation; zero if same as logical birth txg
195 * log. birth	transaction group in which the block was logically born
196 * fill count	number of non-zero blocks under this bp
197 * checksum[4]	256-bit checksum of the data this bp describes
198 */
199#define	SPA_BLKPTRSHIFT	7		/* blkptr_t is 128 bytes	*/
200#define	SPA_DVAS_PER_BP	3		/* Number of DVAs in a bp	*/
201
202/*
203 * A block is a hole when it has either 1) never been written to, or
204 * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
205 * without physically allocating disk space. Holes are represented in the
206 * blkptr_t structure by zeroed blk_dva. Correct checking for holes is
207 * done through the BP_IS_HOLE macro. For holes, the logical size, level,
208 * DMU object type, and birth times are all also stored for holes that
209 * were written to at some point (i.e. were punched after having been filled).
210 */
211typedef struct blkptr {
212	dva_t		blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
213	uint64_t	blk_prop;	/* size, compression, type, etc	    */
214	uint64_t	blk_pad[2];	/* Extra space for the future	    */
215	uint64_t	blk_phys_birth;	/* txg when block was allocated	    */
216	uint64_t	blk_birth;	/* transaction group at birth	    */
217	uint64_t	blk_fill;	/* fill count			    */
218	zio_cksum_t	blk_cksum;	/* 256-bit checksum		    */
219} blkptr_t;
220
221/*
222 * Macros to get and set fields in a bp or DVA.
223 */
224#define	DVA_GET_ASIZE(dva)	\
225	BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
226#define	DVA_SET_ASIZE(dva, x)	\
227	BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
228	SPA_MINBLOCKSHIFT, 0, x)
229
230#define	DVA_GET_GRID(dva)	BF64_GET((dva)->dva_word[0], 24, 8)
231#define	DVA_SET_GRID(dva, x)	BF64_SET((dva)->dva_word[0], 24, 8, x)
232
233#define	DVA_GET_VDEV(dva)	BF64_GET((dva)->dva_word[0], 32, 32)
234#define	DVA_SET_VDEV(dva, x)	BF64_SET((dva)->dva_word[0], 32, 32, x)
235
236#define	DVA_GET_OFFSET(dva)	\
237	BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
238#define	DVA_SET_OFFSET(dva, x)	\
239	BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
240
241#define	DVA_GET_GANG(dva)	BF64_GET((dva)->dva_word[1], 63, 1)
242#define	DVA_SET_GANG(dva, x)	BF64_SET((dva)->dva_word[1], 63, 1, x)
243
244#define	BP_GET_LSIZE(bp)	\
245	BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)
246#define	BP_SET_LSIZE(bp, x)	\
247	BF64_SET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x)
248
249#define	BP_GET_PSIZE(bp)	\
250	BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1)
251#define	BP_SET_PSIZE(bp, x)	\
252	BF64_SET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x)
253
254#define	BP_GET_COMPRESS(bp)		BF64_GET((bp)->blk_prop, 32, 8)
255#define	BP_SET_COMPRESS(bp, x)		BF64_SET((bp)->blk_prop, 32, 8, x)
256
257#define	BP_GET_CHECKSUM(bp)		BF64_GET((bp)->blk_prop, 40, 8)
258#define	BP_SET_CHECKSUM(bp, x)		BF64_SET((bp)->blk_prop, 40, 8, x)
259
260#define	BP_GET_TYPE(bp)			BF64_GET((bp)->blk_prop, 48, 8)
261#define	BP_SET_TYPE(bp, x)		BF64_SET((bp)->blk_prop, 48, 8, x)
262
263#define	BP_GET_LEVEL(bp)		BF64_GET((bp)->blk_prop, 56, 5)
264#define	BP_SET_LEVEL(bp, x)		BF64_SET((bp)->blk_prop, 56, 5, x)
265
266#define	BP_GET_PROP_BIT_61(bp)		BF64_GET((bp)->blk_prop, 61, 1)
267#define	BP_SET_PROP_BIT_61(bp, x)	BF64_SET((bp)->blk_prop, 61, 1, x)
268
269#define	BP_GET_DEDUP(bp)		BF64_GET((bp)->blk_prop, 62, 1)
270#define	BP_SET_DEDUP(bp, x)		BF64_SET((bp)->blk_prop, 62, 1, x)
271
272#define	BP_GET_BYTEORDER(bp)		BF64_GET((bp)->blk_prop, 63, 1)
273#define	BP_SET_BYTEORDER(bp, x)		BF64_SET((bp)->blk_prop, 63, 1, x)
274
275#define	BP_PHYSICAL_BIRTH(bp)		\
276	((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
277
278#define	BP_SET_BIRTH(bp, logical, physical)	\
279{						\
280	(bp)->blk_birth = (logical);		\
281	(bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
282}
283
284#define	BP_GET_ASIZE(bp)	\
285	(DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
286		DVA_GET_ASIZE(&(bp)->blk_dva[2]))
287
288#define	BP_GET_UCSIZE(bp) \
289	((BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp))) ? \
290	BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
291
292#define	BP_GET_NDVAS(bp)	\
293	(!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
294	!!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
295	!!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
296
297#define	BP_COUNT_GANG(bp)	\
298	(DVA_GET_GANG(&(bp)->blk_dva[0]) + \
299	DVA_GET_GANG(&(bp)->blk_dva[1]) + \
300	DVA_GET_GANG(&(bp)->blk_dva[2]))
301
302#define	DVA_EQUAL(dva1, dva2)	\
303	((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
304	(dva1)->dva_word[0] == (dva2)->dva_word[0])
305
306#define	BP_EQUAL(bp1, bp2)	\
307	(BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) &&	\
308	DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) &&	\
309	DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) &&	\
310	DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
311
312#define	ZIO_CHECKSUM_EQUAL(zc1, zc2) \
313	(0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
314	((zc1).zc_word[1] - (zc2).zc_word[1]) | \
315	((zc1).zc_word[2] - (zc2).zc_word[2]) | \
316	((zc1).zc_word[3] - (zc2).zc_word[3])))
317
318#define	DVA_IS_VALID(dva)	(DVA_GET_ASIZE(dva) != 0)
319
320#define	ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3)	\
321{						\
322	(zcp)->zc_word[0] = w0;			\
323	(zcp)->zc_word[1] = w1;			\
324	(zcp)->zc_word[2] = w2;			\
325	(zcp)->zc_word[3] = w3;			\
326}
327
328#define	BP_IDENTITY(bp)		(&(bp)->blk_dva[0])
329#define	BP_IS_GANG(bp)		DVA_GET_GANG(BP_IDENTITY(bp))
330#define	DVA_IS_EMPTY(dva)	((dva)->dva_word[0] == 0ULL &&	\
331				(dva)->dva_word[1] == 0ULL)
332#define	BP_IS_HOLE(bp)		DVA_IS_EMPTY(BP_IDENTITY(bp))
333
334/* BP_IS_RAIDZ(bp) assumes no block compression */
335#define	BP_IS_RAIDZ(bp)		(DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
336				BP_GET_PSIZE(bp))
337
338#define	BP_ZERO(bp)				\
339{						\
340	(bp)->blk_dva[0].dva_word[0] = 0;	\
341	(bp)->blk_dva[0].dva_word[1] = 0;	\
342	(bp)->blk_dva[1].dva_word[0] = 0;	\
343	(bp)->blk_dva[1].dva_word[1] = 0;	\
344	(bp)->blk_dva[2].dva_word[0] = 0;	\
345	(bp)->blk_dva[2].dva_word[1] = 0;	\
346	(bp)->blk_prop = 0;			\
347	(bp)->blk_pad[0] = 0;			\
348	(bp)->blk_pad[1] = 0;			\
349	(bp)->blk_phys_birth = 0;		\
350	(bp)->blk_birth = 0;			\
351	(bp)->blk_fill = 0;			\
352	ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0);	\
353}
354
355#ifdef _BIG_ENDIAN
356#define	ZFS_HOST_BYTEORDER	(0ULL)
357#else
358#define	ZFS_HOST_BYTEORDER	(1ULL)
359#endif
360
361#define	BP_SHOULD_BYTESWAP(bp)	(BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
362
363#define	BP_SPRINTF_LEN	320
364
365/*
366 * This macro allows code sharing between zfs, libzpool, and mdb.
367 * 'func' is either snprintf() or mdb_snprintf().
368 * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
369 */
370#define	SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
371{									\
372	static const char *copyname[] =					\
373	    { "zero", "single", "double", "triple" };			\
374	int len = 0;							\
375	int copies = 0;							\
376									\
377	if (bp == NULL) {						\
378		len += func(buf + len, size - len, "<NULL>");		\
379	} else if (BP_IS_HOLE(bp)) {					\
380		len += func(buf + len, size - len, "<hole>");		\
381		if (bp->blk_birth > 0) {				\
382			len += func(buf + len, size - len,		\
383			    " birth=%lluL",				\
384			    (u_longlong_t)bp->blk_birth);		\
385		}							\
386	} else {							\
387		for (int d = 0; d < BP_GET_NDVAS(bp); d++) {		\
388			const dva_t *dva = &bp->blk_dva[d];		\
389			if (DVA_IS_VALID(dva))				\
390				copies++;				\
391			len += func(buf + len, size - len,		\
392			    "DVA[%d]=<%llu:%llx:%llx>%c", d,		\
393			    (u_longlong_t)DVA_GET_VDEV(dva),		\
394			    (u_longlong_t)DVA_GET_OFFSET(dva),		\
395			    (u_longlong_t)DVA_GET_ASIZE(dva),		\
396			    ws);					\
397		}							\
398		if (BP_IS_GANG(bp) &&					\
399		    DVA_GET_ASIZE(&bp->blk_dva[2]) <=			\
400		    DVA_GET_ASIZE(&bp->blk_dva[1]) / 2)			\
401			copies--;					\
402		len += func(buf + len, size - len,			\
403		    "[L%llu %s] %s %s %s %s %s %s%c"			\
404		    "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c"	\
405		    "cksum=%llx:%llx:%llx:%llx",			\
406		    (u_longlong_t)BP_GET_LEVEL(bp),			\
407		    type,						\
408		    checksum,						\
409		    compress,						\
410		    BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE",		\
411		    BP_IS_GANG(bp) ? "gang" : "contiguous",		\
412		    BP_GET_DEDUP(bp) ? "dedup" : "unique",		\
413		    copyname[copies],					\
414		    ws,							\
415		    (u_longlong_t)BP_GET_LSIZE(bp),			\
416		    (u_longlong_t)BP_GET_PSIZE(bp),			\
417		    (u_longlong_t)bp->blk_birth,			\
418		    (u_longlong_t)BP_PHYSICAL_BIRTH(bp),		\
419		    (u_longlong_t)bp->blk_fill,				\
420		    ws,							\
421		    (u_longlong_t)bp->blk_cksum.zc_word[0],		\
422		    (u_longlong_t)bp->blk_cksum.zc_word[1],		\
423		    (u_longlong_t)bp->blk_cksum.zc_word[2],		\
424		    (u_longlong_t)bp->blk_cksum.zc_word[3]);		\
425	}								\
426	ASSERT(len < size);						\
427}
428
429#include <sys/dmu.h>
430
431#define	BP_GET_BUFC_TYPE(bp)						\
432	(((BP_GET_LEVEL(bp) > 0) || (DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))) ? \
433	ARC_BUFC_METADATA : ARC_BUFC_DATA)
434
435typedef enum spa_import_type {
436	SPA_IMPORT_EXISTING,
437	SPA_IMPORT_ASSEMBLE
438} spa_import_type_t;
439
440/* state manipulation functions */
441extern int spa_open(const char *pool, spa_t **, void *tag);
442extern int spa_open_rewind(const char *pool, spa_t **, void *tag,
443    nvlist_t *policy, nvlist_t **config);
444extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot,
445    size_t buflen);
446extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props,
447    nvlist_t *zplprops);
448extern int spa_import_rootpool(char *devpath, char *devid);
449extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props,
450    uint64_t flags);
451extern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
452extern int spa_destroy(char *pool);
453extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force,
454    boolean_t hardforce);
455extern int spa_reset(char *pool);
456extern void spa_async_request(spa_t *spa, int flag);
457extern void spa_async_unrequest(spa_t *spa, int flag);
458extern void spa_async_suspend(spa_t *spa);
459extern void spa_async_resume(spa_t *spa);
460extern spa_t *spa_inject_addref(char *pool);
461extern void spa_inject_delref(spa_t *spa);
462extern void spa_scan_stat_init(spa_t *spa);
463extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps);
464
465#define	SPA_ASYNC_CONFIG_UPDATE	0x01
466#define	SPA_ASYNC_REMOVE	0x02
467#define	SPA_ASYNC_PROBE		0x04
468#define	SPA_ASYNC_RESILVER_DONE	0x08
469#define	SPA_ASYNC_RESILVER	0x10
470#define	SPA_ASYNC_AUTOEXPAND	0x20
471#define	SPA_ASYNC_REMOVE_DONE	0x40
472#define	SPA_ASYNC_REMOVE_STOP	0x80
473
474/*
475 * Controls the behavior of spa_vdev_remove().
476 */
477#define	SPA_REMOVE_UNSPARE	0x01
478#define	SPA_REMOVE_DONE		0x02
479
480/* device manipulation */
481extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
482extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot,
483    int replacing);
484extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
485    int replace_done);
486extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare);
487extern boolean_t spa_vdev_remove_active(spa_t *spa);
488extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
489extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru);
490extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
491    nvlist_t *props, boolean_t exp);
492
493/* spare state (which is global across all pools) */
494extern void spa_spare_add(vdev_t *vd);
495extern void spa_spare_remove(vdev_t *vd);
496extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt);
497extern void spa_spare_activate(vdev_t *vd);
498
499/* L2ARC state (which is global across all pools) */
500extern void spa_l2cache_add(vdev_t *vd);
501extern void spa_l2cache_remove(vdev_t *vd);
502extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
503extern void spa_l2cache_activate(vdev_t *vd);
504extern void spa_l2cache_drop(spa_t *spa);
505
506/* scanning */
507extern int spa_scan(spa_t *spa, pool_scan_func_t func);
508extern int spa_scan_stop(spa_t *spa);
509
510/* spa syncing */
511extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
512extern void spa_sync_allpools(void);
513
514/* spa namespace global mutex */
515extern kmutex_t spa_namespace_lock;
516
517/*
518 * SPA configuration functions in spa_config.c
519 */
520
521#define	SPA_CONFIG_UPDATE_POOL	0
522#define	SPA_CONFIG_UPDATE_VDEVS	1
523
524extern void spa_config_sync(spa_t *, boolean_t, boolean_t);
525extern void spa_config_load(void);
526extern nvlist_t *spa_all_configs(uint64_t *);
527extern void spa_config_set(spa_t *spa, nvlist_t *config);
528extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
529    int getstats);
530extern void spa_config_update(spa_t *spa, int what);
531
532/*
533 * Miscellaneous SPA routines in spa_misc.c
534 */
535
536/* Namespace manipulation */
537extern spa_t *spa_lookup(const char *name);
538extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot);
539extern void spa_remove(spa_t *spa);
540extern spa_t *spa_next(spa_t *prev);
541
542/* Refcount functions */
543extern void spa_open_ref(spa_t *spa, void *tag);
544extern void spa_close(spa_t *spa, void *tag);
545extern boolean_t spa_refcount_zero(spa_t *spa);
546
547#define	SCL_NONE	0x00
548#define	SCL_CONFIG	0x01
549#define	SCL_STATE	0x02
550#define	SCL_L2ARC	0x04		/* hack until L2ARC 2.0 */
551#define	SCL_ALLOC	0x08
552#define	SCL_ZIO		0x10
553#define	SCL_FREE	0x20
554#define	SCL_VDEV	0x40
555#define	SCL_LOCKS	7
556#define	SCL_ALL		((1 << SCL_LOCKS) - 1)
557#define	SCL_STATE_ALL	(SCL_STATE | SCL_L2ARC | SCL_ZIO)
558
559/* Pool configuration locks */
560extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw);
561extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw);
562extern void spa_config_exit(spa_t *spa, int locks, void *tag);
563extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
564
565/* Pool vdev add/remove lock */
566extern uint64_t spa_vdev_enter(spa_t *spa);
567extern uint64_t spa_vdev_config_enter(spa_t *spa);
568extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg,
569    int error, char *tag);
570extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error);
571
572/* Pool vdev state change lock */
573extern void spa_vdev_state_enter(spa_t *spa, int oplock);
574extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error);
575
576/* Log state */
577typedef enum spa_log_state {
578	SPA_LOG_UNKNOWN = 0,	/* unknown log state */
579	SPA_LOG_MISSING,	/* missing log(s) */
580	SPA_LOG_CLEAR,		/* clear the log(s) */
581	SPA_LOG_GOOD,		/* log(s) are good */
582} spa_log_state_t;
583
584extern spa_log_state_t spa_get_log_state(spa_t *spa);
585extern void spa_set_log_state(spa_t *spa, spa_log_state_t state);
586extern int spa_offline_log(spa_t *spa);
587
588/* Log claim callback */
589extern void spa_claim_notify(zio_t *zio);
590
591/* Accessor functions */
592extern boolean_t spa_shutting_down(spa_t *spa);
593extern struct dsl_pool *spa_get_dsl(spa_t *spa);
594extern boolean_t spa_is_initializing(spa_t *spa);
595extern blkptr_t *spa_get_rootblkptr(spa_t *spa);
596extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
597extern void spa_altroot(spa_t *, char *, size_t);
598extern int spa_sync_pass(spa_t *spa);
599extern char *spa_name(spa_t *spa);
600extern uint64_t spa_guid(spa_t *spa);
601extern uint64_t spa_load_guid(spa_t *spa);
602extern uint64_t spa_last_synced_txg(spa_t *spa);
603extern uint64_t spa_first_txg(spa_t *spa);
604extern uint64_t spa_syncing_txg(spa_t *spa);
605extern uint64_t spa_version(spa_t *spa);
606extern pool_state_t spa_state(spa_t *spa);
607extern spa_load_state_t spa_load_state(spa_t *spa);
608extern uint64_t spa_freeze_txg(spa_t *spa);
609extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize);
610extern uint64_t spa_get_dspace(spa_t *spa);
611extern void spa_update_dspace(spa_t *spa);
612extern uint64_t spa_version(spa_t *spa);
613extern boolean_t spa_deflate(spa_t *spa);
614extern metaslab_class_t *spa_normal_class(spa_t *spa);
615extern metaslab_class_t *spa_log_class(spa_t *spa);
616extern int spa_max_replication(spa_t *spa);
617extern int spa_prev_software_version(spa_t *spa);
618extern int spa_busy(void);
619extern uint8_t spa_get_failmode(spa_t *spa);
620extern boolean_t spa_suspended(spa_t *spa);
621extern uint64_t spa_bootfs(spa_t *spa);
622extern uint64_t spa_delegation(spa_t *spa);
623extern objset_t *spa_meta_objset(spa_t *spa);
624extern uint64_t spa_deadman_synctime(spa_t *spa);
625
626/* Miscellaneous support routines */
627extern void spa_activate_mos_feature(spa_t *spa, const char *feature,
628    dmu_tx_t *tx);
629extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature);
630extern int spa_rename(const char *oldname, const char *newname);
631extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid);
632extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid);
633extern char *spa_strdup(const char *);
634extern void spa_strfree(char *);
635extern uint64_t spa_get_random(uint64_t range);
636extern uint64_t spa_generate_guid(spa_t *spa);
637extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp);
638extern void spa_freeze(spa_t *spa);
639extern int spa_change_guid(spa_t *spa);
640extern void spa_upgrade(spa_t *spa, uint64_t version);
641extern void spa_evict_all(void);
642extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid,
643    boolean_t l2cache);
644extern boolean_t spa_has_spare(spa_t *, uint64_t guid);
645extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva);
646extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp);
647extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp);
648extern boolean_t spa_has_slogs(spa_t *spa);
649extern boolean_t spa_is_root(spa_t *spa);
650extern boolean_t spa_writeable(spa_t *spa);
651
652extern int spa_mode(spa_t *spa);
653extern uint64_t strtonum(const char *str, char **nptr);
654
655extern char *spa_his_ievent_table[];
656
657extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx);
658extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read,
659    char *his_buf);
660extern int spa_history_log(spa_t *spa, const char *his_buf);
661extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl);
662extern void spa_history_log_version(spa_t *spa, const char *operation);
663extern void spa_history_log_internal(spa_t *spa, const char *operation,
664    dmu_tx_t *tx, const char *fmt, ...);
665extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op,
666    dmu_tx_t *tx, const char *fmt, ...);
667extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
668    dmu_tx_t *tx, const char *fmt, ...);
669
670/* error handling */
671struct zbookmark;
672extern void spa_log_error(spa_t *spa, zio_t *zio);
673extern void zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd,
674    zio_t *zio, uint64_t stateoroffset, uint64_t length);
675extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
676extern void zfs_post_state_change(spa_t *spa, vdev_t *vd);
677extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
678extern uint64_t spa_get_errlog_size(spa_t *spa);
679extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
680extern void spa_errlog_rotate(spa_t *spa);
681extern void spa_errlog_drain(spa_t *spa);
682extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
683extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
684
685/* vdev cache */
686extern void vdev_cache_stat_init(void);
687extern void vdev_cache_stat_fini(void);
688
689/* Initialization and termination */
690extern void spa_init(int flags);
691extern void spa_fini(void);
692extern void spa_boot_init();
693
694/* properties */
695extern int spa_prop_set(spa_t *spa, nvlist_t *nvp);
696extern int spa_prop_get(spa_t *spa, nvlist_t **nvp);
697extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx);
698extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t);
699
700/* asynchronous event notification */
701extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name);
702
703#ifdef ZFS_DEBUG
704#define	dprintf_bp(bp, fmt, ...) do {				\
705	if (zfs_flags & ZFS_DEBUG_DPRINTF) {			\
706	char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP);	\
707	snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp));	\
708	dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf);		\
709	kmem_free(__blkbuf, BP_SPRINTF_LEN);			\
710	} \
711_NOTE(CONSTCOND) } while (0)
712#else
713#define	dprintf_bp(bp, fmt, ...)
714#endif
715
716extern boolean_t spa_debug_enabled(spa_t *spa);
717#define	spa_dbgmsg(spa, ...)			\
718{						\
719	if (spa_debug_enabled(spa))		\
720		zfs_dbgmsg(__VA_ARGS__);	\
721}
722
723extern int spa_mode_global;			/* mode, e.g. FREAD | FWRITE */
724
725#ifdef	__cplusplus
726}
727#endif
728
729#endif	/* _SYS_SPA_H */
730