spa.h revision 663207adb1669640c01c5ec6949ce78fd806efae
1fa9e406ahrens/*
2fa9e406ahrens * CDDL HEADER START
3fa9e406ahrens *
4fa9e406ahrens * The contents of this file are subject to the terms of the
5ea8dc4beschrock * Common Development and Distribution License (the "License").
6ea8dc4beschrock * You may not use this file except in compliance with the License.
7fa9e406ahrens *
8fa9e406ahrens * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9fa9e406ahrens * or http://www.opensolaris.org/os/licensing.
10fa9e406ahrens * See the License for the specific language governing permissions
11fa9e406ahrens * and limitations under the License.
12fa9e406ahrens *
13fa9e406ahrens * When distributing Covered Code, include this CDDL HEADER in each
14fa9e406ahrens * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15fa9e406ahrens * If applicable, add the following below this CDDL HEADER, with the
16fa9e406ahrens * fields enclosed by brackets "[]" replaced with your own identifying
17fa9e406ahrens * information: Portions Copyright [yyyy] [name of copyright owner]
18fa9e406ahrens *
19fa9e406ahrens * CDDL HEADER END
20fa9e406ahrens */
21fa9e406ahrens/*
223f9d6adLin Ling * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
235cabbc6Prashanth Sreenivasa * Copyright (c) 2011, 2018 by Delphix. All rights reserved.
24e9103aaGarrett D'Amore * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
25bc9014eJustin Gibbs * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
2645818eeMatthew Ahrens * Copyright 2013 Saso Kiselkov. All rights reserved.
27c3d26abMatthew Ahrens * Copyright (c) 2014 Integros [integros.com]
28ce1577bDave Eddy * Copyright 2017 Joyent, Inc.
291702cceAlek Pinchuk * Copyright (c) 2017 Datto Inc.
30663207aDon Brady * Copyright (c) 2017, Intel Corporation.
31fa9e406ahrens */
32fa9e406ahrens
33fa9e406ahrens#ifndef _SYS_SPA_H
34fa9e406ahrens#define	_SYS_SPA_H
35fa9e406ahrens
36fa9e406ahrens#include <sys/avl.h>
37fa9e406ahrens#include <sys/zfs_context.h>
38fa9e406ahrens#include <sys/nvpair.h>
395cabbc6Prashanth Sreenivasa#include <sys/sysevent.h>
40fa9e406ahrens#include <sys/sysmacros.h>
41fa9e406ahrens#include <sys/types.h>
42fa9e406ahrens#include <sys/fs/zfs.h>
434ba5b96Dan Kimmel#include <sys/dmu.h>
44fa9e406ahrens
45fa9e406ahrens#ifdef	__cplusplus
46fa9e406ahrensextern "C" {
47fa9e406ahrens#endif
48fa9e406ahrens
49fa9e406ahrens/*
50fa9e406ahrens * Forward references that lots of things need.
51fa9e406ahrens */
52fa9e406ahrenstypedef struct spa spa_t;
53fa9e406ahrenstypedef struct vdev vdev_t;
54fa9e406ahrenstypedef struct metaslab metaslab_t;
55b24ab67Jeff Bonwicktypedef struct metaslab_group metaslab_group_t;
56b24ab67Jeff Bonwicktypedef struct metaslab_class metaslab_class_t;
57b24ab67Jeff Bonwicktypedef struct zio zio_t;
58fa9e406ahrenstypedef struct zilog zilog_t;
59fa94a07brendantypedef struct spa_aux_vdev spa_aux_vdev_t;
60b24ab67Jeff Bonwicktypedef struct ddt ddt_t;
61b24ab67Jeff Bonwicktypedef struct ddt_entry ddt_entry_t;
62fa9e406ahrensstruct dsl_pool;
634445fffMatthew Ahrensstruct dsl_dataset;
64fa9e406ahrens
65fa9e406ahrens/*
66fa9e406ahrens * General-purpose 32-bit and 64-bit bitfield encodings.
67fa9e406ahrens */
68fa9e406ahrens#define	BF32_DECODE(x, low, len)	P2PHASE((x) >> (low), 1U << (len))
69fa9e406ahrens#define	BF64_DECODE(x, low, len)	P2PHASE((x) >> (low), 1ULL << (len))
70fa9e406ahrens#define	BF32_ENCODE(x, low, len)	(P2PHASE((x), 1U << (len)) << (low))
71fa9e406ahrens#define	BF64_ENCODE(x, low, len)	(P2PHASE((x), 1ULL << (len)) << (low))
72fa9e406ahrens
73fa9e406ahrens#define	BF32_GET(x, low, len)		BF32_DECODE(x, low, len)
74fa9e406ahrens#define	BF64_GET(x, low, len)		BF64_DECODE(x, low, len)
75fa9e406ahrens
7643466aaMax Grossman#define	BF32_SET(x, low, len, val) do { \
7743466aaMax Grossman	ASSERT3U(val, <, 1U << (len)); \
7843466aaMax Grossman	ASSERT3U(low + len, <=, 32); \
7943466aaMax Grossman	(x) ^= BF32_ENCODE((x >> low) ^ (val), low, len); \
8043466aaMax Grossman_NOTE(CONSTCOND) } while (0)
8143466aaMax Grossman
8243466aaMax Grossman#define	BF64_SET(x, low, len, val) do { \
8343466aaMax Grossman	ASSERT3U(val, <, 1ULL << (len)); \
8443466aaMax Grossman	ASSERT3U(low + len, <=, 64); \
8543466aaMax Grossman	((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)); \
8643466aaMax Grossman_NOTE(CONSTCOND) } while (0)
87fa9e406ahrens
88fa9e406ahrens#define	BF32_GET_SB(x, low, len, shift, bias)	\
89fa9e406ahrens	((BF32_GET(x, low, len) + (bias)) << (shift))
90fa9e406ahrens#define	BF64_GET_SB(x, low, len, shift, bias)	\
91fa9e406ahrens	((BF64_GET(x, low, len) + (bias)) << (shift))
92fa9e406ahrens
9343466aaMax Grossman#define	BF32_SET_SB(x, low, len, shift, bias, val) do { \
9443466aaMax Grossman	ASSERT(IS_P2ALIGNED(val, 1U << shift)); \
9543466aaMax Grossman	ASSERT3S((val) >> (shift), >=, bias); \
9643466aaMax Grossman	BF32_SET(x, low, len, ((val) >> (shift)) - (bias)); \
9743466aaMax Grossman_NOTE(CONSTCOND) } while (0)
9843466aaMax Grossman#define	BF64_SET_SB(x, low, len, shift, bias, val) do { \
9943466aaMax Grossman	ASSERT(IS_P2ALIGNED(val, 1ULL << shift)); \
10043466aaMax Grossman	ASSERT3S((val) >> (shift), >=, bias); \
10143466aaMax Grossman	BF64_SET(x, low, len, ((val) >> (shift)) - (bias)); \
10243466aaMax Grossman_NOTE(CONSTCOND) } while (0)
103fa9e406ahrens
104fa9e406ahrens/*
105b515258Matthew Ahrens * We currently support block sizes from 512 bytes to 16MB.
106b515258Matthew Ahrens * The benefits of larger blocks, and thus larger IO, need to be weighed
107b515258Matthew Ahrens * against the cost of COWing a giant block to modify one byte, and the
108b515258Matthew Ahrens * large latency of reading or writing a large block.
109b515258Matthew Ahrens *
110b515258Matthew Ahrens * Note that although blocks up to 16MB are supported, the recordsize
111b515258Matthew Ahrens * property can not be set larger than zfs_max_recordsize (default 1MB).
112b515258Matthew Ahrens * See the comment near zfs_max_recordsize in dsl_dataset.c for details.
113b515258Matthew Ahrens *
114b515258Matthew Ahrens * Note that although the LSIZE field of the blkptr_t can store sizes up
115b515258Matthew Ahrens * to 32MB, the dnode's dn_datablkszsec can only store sizes up to
116b515258Matthew Ahrens * 32MB - 512 bytes.  Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
117fa9e406ahrens */
118fa9e406ahrens#define	SPA_MINBLOCKSHIFT	9
119b515258Matthew Ahrens#define	SPA_OLD_MAXBLOCKSHIFT	17
120b515258Matthew Ahrens#define	SPA_MAXBLOCKSHIFT	24
121fa9e406ahrens#define	SPA_MINBLOCKSIZE	(1ULL << SPA_MINBLOCKSHIFT)
122b515258Matthew Ahrens#define	SPA_OLD_MAXBLOCKSIZE	(1ULL << SPA_OLD_MAXBLOCKSHIFT)
123fa9e406ahrens#define	SPA_MAXBLOCKSIZE	(1ULL << SPA_MAXBLOCKSHIFT)
124fa9e406ahrens
125fa9e406ahrens/*
126f7991baTim Haley * Size of block to hold the configuration data (a packed nvlist)
127f7991baTim Haley */
128ad135b5Christopher Siden#define	SPA_CONFIG_BLOCKSIZE	(1ULL << 14)
129f7991baTim Haley
130f7991baTim Haley/*
131fa9e406ahrens * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
132fa9e406ahrens * The ASIZE encoding should be at least 64 times larger (6 more bits)
133fa9e406ahrens * to support up to 4-way RAID-Z mirror mode with worst-case gang block
134fa9e406ahrens * overhead, three DVAs per bp, plus one more bit in case we do anything
135fa9e406ahrens * else that expands the ASIZE.
136fa9e406ahrens */
137fa9e406ahrens#define	SPA_LSIZEBITS		16	/* LSIZE up to 32M (2^16 * 512)	*/
138fa9e406ahrens#define	SPA_PSIZEBITS		16	/* PSIZE up to 32M (2^16 * 512)	*/
139fa9e406ahrens#define	SPA_ASIZEBITS		24	/* ASIZE up to 64 times larger	*/
140fa9e406ahrens
141dcbf3bdGeorge Wilson#define	SPA_COMPRESSBITS	7
14217f1128Serapheim Dimitropoulos#define	SPA_VDEVBITS		24
143dcbf3bdGeorge Wilson
144fa9e406ahrens/*
145fa9e406ahrens * All SPA data is represented by 128-bit data virtual addresses (DVAs).
146fa9e406ahrens * The members of the dva_t should be considered opaque outside the SPA.
147fa9e406ahrens */
148fa9e406ahrenstypedef struct dva {
149fa9e406ahrens	uint64_t	dva_word[2];
150fa9e406ahrens} dva_t;
151fa9e406ahrens
152fa9e406ahrens/*
153fa9e406ahrens * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
154fa9e406ahrens */
155fa9e406ahrenstypedef struct zio_cksum {
156fa9e406ahrens	uint64_t	zc_word[4];
157fa9e406ahrens} zio_cksum_t;
158fa9e406ahrens
159fa9e406ahrens/*
16045818eeMatthew Ahrens * Some checksums/hashes need a 256-bit initialization salt. This salt is kept
16145818eeMatthew Ahrens * secret and is suitable for use in MAC algorithms as the key.
16245818eeMatthew Ahrens */
16345818eeMatthew Ahrenstypedef struct zio_cksum_salt {
16445818eeMatthew Ahrens	uint8_t		zcs_bytes[32];
16545818eeMatthew Ahrens} zio_cksum_salt_t;
16645818eeMatthew Ahrens
16745818eeMatthew Ahrens/*
168fa9e406ahrens * Each block is described by its DVAs, time of birth, checksum, etc.
169fa9e406ahrens * The word-by-word, bit-by-bit layout of the blkptr is as follows:
170fa9e406ahrens *
171fa9e406ahrens *	64	56	48	40	32	24	16	8	0
172fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
17317f1128Serapheim Dimitropoulos * 0	|  pad  |	  vdev1         | GRID  |	  ASIZE		|
174fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
175fa9e406ahrens * 1	|G|			 offset1				|
176fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
17717f1128Serapheim Dimitropoulos * 2	|  pad  |	  vdev2         | GRID  |	  ASIZE		|
178fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
179fa9e406ahrens * 3	|G|			 offset2				|
180fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
18117f1128Serapheim Dimitropoulos * 4	|  pad  |	  vdev3         | GRID  |	  ASIZE		|
182fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
183fa9e406ahrens * 5	|G|			 offset3				|
184fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
1855d7b4d4Matthew Ahrens * 6	|BDX|lvl| type	| cksum |E| comp|    PSIZE	|     LSIZE	|
186fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
187fa9e406ahrens * 7	|			padding					|
188fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
189fa9e406ahrens * 8	|			padding					|
190fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
191b24ab67Jeff Bonwick * 9	|			physical birth txg			|
192fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
193b24ab67Jeff Bonwick * a	|			logical birth txg			|
194fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
195fa9e406ahrens * b	|			fill count				|
196fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
197fa9e406ahrens * c	|			checksum[0]				|
198fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
199fa9e406ahrens * d	|			checksum[1]				|
200fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
201fa9e406ahrens * e	|			checksum[2]				|
202fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
203fa9e406ahrens * f	|			checksum[3]				|
204fa9e406ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
205fa9e406ahrens *
206fa9e406ahrens * Legend:
207fa9e406ahrens *
208fa9e406ahrens * vdev		virtual device ID
209fa9e406ahrens * offset	offset into virtual device
210fa9e406ahrens * LSIZE	logical size
211fa9e406ahrens * PSIZE	physical size (after compression)
212fa9e406ahrens * ASIZE	allocated size (including RAID-Z parity and gang block headers)
213fa9e406ahrens * GRID		RAID-Z layout information (reserved for future use)
214fa9e406ahrens * cksum	checksum function
215fa9e406ahrens * comp		compression function
216fa9e406ahrens * G		gang block indicator
217b24ab67Jeff Bonwick * B		byteorder (endianness)
218b24ab67Jeff Bonwick * D		dedup
2195d7b4d4Matthew Ahrens * X		encryption (on version 30, which is not supported)
2205d7b4d4Matthew Ahrens * E		blkptr_t contains embedded data (see below)
221fa9e406ahrens * lvl		level of indirection
222b24ab67Jeff Bonwick * type		DMU object type
2235cabbc6Prashanth Sreenivasa * phys birth	txg when dva[0] was written; zero if same as logical birth txg
2245cabbc6Prashanth Sreenivasa *              note that typically all the dva's would be written in this
2255cabbc6Prashanth Sreenivasa *              txg, but they could be different if they were moved by
2265cabbc6Prashanth Sreenivasa *              device removal.
227b24ab67Jeff Bonwick * log. birth	transaction group in which the block was logically born
228fa9e406ahrens * fill count	number of non-zero blocks under this bp
229fa9e406ahrens * checksum[4]	256-bit checksum of the data this bp describes
230fa9e406ahrens */
2315d7b4d4Matthew Ahrens
2325d7b4d4Matthew Ahrens/*
2335d7b4d4Matthew Ahrens * "Embedded" blkptr_t's don't actually point to a block, instead they
2345d7b4d4Matthew Ahrens * have a data payload embedded in the blkptr_t itself.  See the comment
2355d7b4d4Matthew Ahrens * in blkptr.c for more details.
2365d7b4d4Matthew Ahrens *
2375d7b4d4Matthew Ahrens * The blkptr_t is laid out as follows:
2385d7b4d4Matthew Ahrens *
2395d7b4d4Matthew Ahrens *	64	56	48	40	32	24	16	8	0
2405d7b4d4Matthew Ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
2415d7b4d4Matthew Ahrens * 0	|      payload                                                  |
2425d7b4d4Matthew Ahrens * 1	|      payload                                                  |
2435d7b4d4Matthew Ahrens * 2	|      payload                                                  |
2445d7b4d4Matthew Ahrens * 3	|      payload                                                  |
2455d7b4d4Matthew Ahrens * 4	|      payload                                                  |
2465d7b4d4Matthew Ahrens * 5	|      payload                                                  |
2475d7b4d4Matthew Ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
2485d7b4d4Matthew Ahrens * 6	|BDX|lvl| type	| etype |E| comp| PSIZE|              LSIZE	|
2495d7b4d4Matthew Ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
2505d7b4d4Matthew Ahrens * 7	|      payload                                                  |
2515d7b4d4Matthew Ahrens * 8	|      payload                                                  |
2525d7b4d4Matthew Ahrens * 9	|      payload                                                  |
2535d7b4d4Matthew Ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
2545d7b4d4Matthew Ahrens * a	|			logical birth txg			|
2555d7b4d4Matthew Ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
2565d7b4d4Matthew Ahrens * b	|      payload                                                  |
2575d7b4d4Matthew Ahrens * c	|      payload                                                  |
2585d7b4d4Matthew Ahrens * d	|      payload                                                  |
2595d7b4d4Matthew Ahrens * e	|      payload                                                  |
2605d7b4d4Matthew Ahrens * f	|      payload                                                  |
2615d7b4d4Matthew Ahrens *	+-------+-------+-------+-------+-------+-------+-------+-------+
2625d7b4d4Matthew Ahrens *
2635d7b4d4Matthew Ahrens * Legend:
2645d7b4d4Matthew Ahrens *
2655d7b4d4Matthew Ahrens * payload		contains the embedded data
2665d7b4d4Matthew Ahrens * B (byteorder)	byteorder (endianness)
2675d7b4d4Matthew Ahrens * D (dedup)		padding (set to zero)
2685d7b4d4Matthew Ahrens * X			encryption (set to zero; see above)
2695d7b4d4Matthew Ahrens * E (embedded)		set to one
2705d7b4d4Matthew Ahrens * lvl			indirection level
2715d7b4d4Matthew Ahrens * type			DMU object type
2725d7b4d4Matthew Ahrens * etype		how to interpret embedded data (BP_EMBEDDED_TYPE_*)
2735d7b4d4Matthew Ahrens * comp			compression function of payload
2745d7b4d4Matthew Ahrens * PSIZE		size of payload after compression, in bytes
2755d7b4d4Matthew Ahrens * LSIZE		logical size of payload, in bytes
2765d7b4d4Matthew Ahrens *			note that 25 bits is enough to store the largest
2775d7b4d4Matthew Ahrens *			"normal" BP's LSIZE (2^16 * 2^9) in bytes
2785d7b4d4Matthew Ahrens * log. birth		transaction group in which the block was logically born
2795d7b4d4Matthew Ahrens *
2805d7b4d4Matthew Ahrens * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
2815d7b4d4Matthew Ahrens * bp's they are stored in units of SPA_MINBLOCKSHIFT.
2825d7b4d4Matthew Ahrens * Generally, the generic BP_GET_*() macros can be used on embedded BP's.
2835d7b4d4Matthew Ahrens * The B, D, X, lvl, type, and comp fields are stored the same as with normal
2845d7b4d4Matthew Ahrens * BP's so the BP_SET_* macros can be used with them.  etype, PSIZE, LSIZE must
2855d7b4d4Matthew Ahrens * be set with the BPE_SET_* macros.  BP_SET_EMBEDDED() should be called before
2865d7b4d4Matthew Ahrens * other macros, as they assert that they are only used on BP's of the correct
2875d7b4d4Matthew Ahrens * "embedded-ness".
2885d7b4d4Matthew Ahrens */
2895d7b4d4Matthew Ahrens
2905d7b4d4Matthew Ahrens#define	BPE_GET_ETYPE(bp)	\
2915d7b4d4Matthew Ahrens	(ASSERT(BP_IS_EMBEDDED(bp)), \
2925d7b4d4Matthew Ahrens	BF64_GET((bp)->blk_prop, 40, 8))
2935d7b4d4Matthew Ahrens#define	BPE_SET_ETYPE(bp, t)	do { \
2945d7b4d4Matthew Ahrens	ASSERT(BP_IS_EMBEDDED(bp)); \
2955d7b4d4Matthew Ahrens	BF64_SET((bp)->blk_prop, 40, 8, t); \
2965d7b4d4Matthew Ahrens_NOTE(CONSTCOND) } while (0)
2975d7b4d4Matthew Ahrens
2985d7b4d4Matthew Ahrens#define	BPE_GET_LSIZE(bp)	\
2995d7b4d4Matthew Ahrens	(ASSERT(BP_IS_EMBEDDED(bp)), \
3005d7b4d4Matthew Ahrens	BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
3015d7b4d4Matthew Ahrens#define	BPE_SET_LSIZE(bp, x)	do { \
3025d7b4d4Matthew Ahrens	ASSERT(BP_IS_EMBEDDED(bp)); \
3035d7b4d4Matthew Ahrens	BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
3045d7b4d4Matthew Ahrens_NOTE(CONSTCOND) } while (0)
3055d7b4d4Matthew Ahrens
3065d7b4d4Matthew Ahrens#define	BPE_GET_PSIZE(bp)	\
3075d7b4d4Matthew Ahrens	(ASSERT(BP_IS_EMBEDDED(bp)), \
3085d7b4d4Matthew Ahrens	BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
3095d7b4d4Matthew Ahrens#define	BPE_SET_PSIZE(bp, x)	do { \
3105d7b4d4Matthew Ahrens	ASSERT(BP_IS_EMBEDDED(bp)); \
3115d7b4d4Matthew Ahrens	BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
3125d7b4d4Matthew Ahrens_NOTE(CONSTCOND) } while (0)
3135d7b4d4Matthew Ahrens
3145d7b4d4Matthew Ahrenstypedef enum bp_embedded_type {
3155d7b4d4Matthew Ahrens	BP_EMBEDDED_TYPE_DATA,
3165d7b4d4Matthew Ahrens	BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */
3175d7b4d4Matthew Ahrens	NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED
3185d7b4d4Matthew Ahrens} bp_embedded_type_t;
3195d7b4d4Matthew Ahrens
3205d7b4d4Matthew Ahrens#define	BPE_NUM_WORDS 14
3215d7b4d4Matthew Ahrens#define	BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
3225d7b4d4Matthew Ahrens#define	BPE_IS_PAYLOADWORD(bp, wp) \
3235d7b4d4Matthew Ahrens	((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
3245d7b4d4Matthew Ahrens
325fa9e406ahrens#define	SPA_BLKPTRSHIFT	7		/* blkptr_t is 128 bytes	*/
326fa9e406ahrens#define	SPA_DVAS_PER_BP	3		/* Number of DVAs in a bp	*/
3276f79381Pavel Zakharov#define	SPA_SYNC_MIN_VDEVS 3		/* min vdevs to update during sync */
328fa9e406ahrens
32943466aaMax Grossman/*
33043466aaMax Grossman * A block is a hole when it has either 1) never been written to, or
33143466aaMax Grossman * 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
33243466aaMax Grossman * without physically allocating disk space. Holes are represented in the
33343466aaMax Grossman * blkptr_t structure by zeroed blk_dva. Correct checking for holes is
33443466aaMax Grossman * done through the BP_IS_HOLE macro. For holes, the logical size, level,
33543466aaMax Grossman * DMU object type, and birth times are all also stored for holes that
33643466aaMax Grossman * were written to at some point (i.e. were punched after having been filled).
33743466aaMax Grossman */
338b24ab67Jeff Bonwicktypedef struct blkptr {
339b24ab67Jeff Bonwick	dva_t		blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
340b24ab67Jeff Bonwick	uint64_t	blk_prop;	/* size, compression, type, etc	    */
341b24ab67Jeff Bonwick	uint64_t	blk_pad[2];	/* Extra space for the future	    */
342b24ab67Jeff Bonwick	uint64_t	blk_phys_birth;	/* txg when block was allocated	    */
343b24ab67Jeff Bonwick	uint64_t	blk_birth;	/* transaction group at birth	    */
344b24ab67Jeff Bonwick	uint64_t	blk_fill;	/* fill count			    */
345b24ab67Jeff Bonwick	zio_cksum_t	blk_cksum;	/* 256-bit checksum		    */
346b24ab67Jeff Bonwick} blkptr_t;
347b24ab67Jeff Bonwick
348fa9e406ahrens/*
349fa9e406ahrens * Macros to get and set fields in a bp or DVA.
350fa9e406ahrens */
351fa9e406ahrens#define	DVA_GET_ASIZE(dva)	\
35243466aaMax Grossman	BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
353fa9e406ahrens#define	DVA_SET_ASIZE(dva, x)	\
35443466aaMax Grossman	BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
35543466aaMax Grossman	SPA_MINBLOCKSHIFT, 0, x)
356fa9e406ahrens
357fa9e406ahrens#define	DVA_GET_GRID(dva)	BF64_GET((dva)->dva_word[0], 24, 8)
358fa9e406ahrens#define	DVA_SET_GRID(dva, x)	BF64_SET((dva)->dva_word[0], 24, 8, x)
359fa9e406ahrens
36017f1128Serapheim Dimitropoulos#define	DVA_GET_VDEV(dva)	BF64_GET((dva)->dva_word[0], 32, SPA_VDEVBITS)
36117f1128Serapheim Dimitropoulos#define	DVA_SET_VDEV(dva, x)	\
36217f1128Serapheim Dimitropoulos	BF64_SET((dva)->dva_word[0], 32, SPA_VDEVBITS, x)
363fa9e406ahrens
364fa9e406ahrens#define	DVA_GET_OFFSET(dva)	\
365fa9e406ahrens	BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
366fa9e406ahrens#define	DVA_SET_OFFSET(dva, x)	\
367fa9e406ahrens	BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
368fa9e406ahrens
369fa9e406ahrens#define	DVA_GET_GANG(dva)	BF64_GET((dva)->dva_word[1], 63, 1)
370fa9e406ahrens#define	DVA_SET_GANG(dva, x)	BF64_SET((dva)->dva_word[1], 63, 1, x)
371fa9e406ahrens
372fa9e406ahrens#define	BP_GET_LSIZE(bp)	\
3735d7b4d4Matthew Ahrens	(BP_IS_EMBEDDED(bp) ?	\
3745d7b4d4Matthew Ahrens	(BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
3755d7b4d4Matthew Ahrens	BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
3765d7b4d4Matthew Ahrens#define	BP_SET_LSIZE(bp, x)	do { \
3775d7b4d4Matthew Ahrens	ASSERT(!BP_IS_EMBEDDED(bp)); \
3785d7b4d4Matthew Ahrens	BF64_SET_SB((bp)->blk_prop, \
3795d7b4d4Matthew Ahrens	    0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
3805d7b4d4Matthew Ahrens_NOTE(CONSTCOND) } while (0)
381fa9e406ahrens
382fa9e406ahrens#define	BP_GET_PSIZE(bp)	\
3835d7b4d4Matthew Ahrens	(BP_IS_EMBEDDED(bp) ? 0 : \
3845d7b4d4Matthew Ahrens	BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1))
3855d7b4d4Matthew Ahrens#define	BP_SET_PSIZE(bp, x)	do { \
3865d7b4d4Matthew Ahrens	ASSERT(!BP_IS_EMBEDDED(bp)); \
3875d7b4d4Matthew Ahrens	BF64_SET_SB((bp)->blk_prop, \
3885d7b4d4Matthew Ahrens	    16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
3895d7b4d4Matthew Ahrens_NOTE(CONSTCOND) } while (0)
3905d7b4d4Matthew Ahrens
391dcbf3bdGeorge Wilson#define	BP_GET_COMPRESS(bp)		\
392dcbf3bdGeorge Wilson	BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS)
393dcbf3bdGeorge Wilson#define	BP_SET_COMPRESS(bp, x)		\
394dcbf3bdGeorge Wilson	BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x)
395fa9e406ahrens
3965d7b4d4Matthew Ahrens#define	BP_IS_EMBEDDED(bp)		BF64_GET((bp)->blk_prop, 39, 1)
3975d7b4d4Matthew Ahrens#define	BP_SET_EMBEDDED(bp, x)		BF64_SET((bp)->blk_prop, 39, 1, x)
398b24ab67Jeff Bonwick
3995d7b4d4Matthew Ahrens#define	BP_GET_CHECKSUM(bp)		\
4005d7b4d4Matthew Ahrens	(BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \
4015d7b4d4Matthew Ahrens	BF64_GET((bp)->blk_prop, 40, 8))
4025d7b4d4Matthew Ahrens#define	BP_SET_CHECKSUM(bp, x)		do { \
4035d7b4d4Matthew Ahrens	ASSERT(!BP_IS_EMBEDDED(bp)); \
4045d7b4d4Matthew Ahrens	BF64_SET((bp)->blk_prop, 40, 8, x); \
4055d7b4d4Matthew Ahrens_NOTE(CONSTCOND) } while (0)
406b24ab67Jeff Bonwick
407b24ab67Jeff Bonwick#define	BP_GET_TYPE(bp)			BF64_GET((bp)->blk_prop, 48, 8)
408b24ab67Jeff Bonwick#define	BP_SET_TYPE(bp, x)		BF64_SET((bp)->blk_prop, 48, 8, x)
409fa9e406ahrens
410b24ab67Jeff Bonwick#define	BP_GET_LEVEL(bp)		BF64_GET((bp)->blk_prop, 56, 5)
411b24ab67Jeff Bonwick#define	BP_SET_LEVEL(bp, x)		BF64_SET((bp)->blk_prop, 56, 5, x)
412fa9e406ahrens
413b24ab67Jeff Bonwick#define	BP_GET_DEDUP(bp)		BF64_GET((bp)->blk_prop, 62, 1)
414b24ab67Jeff Bonwick#define	BP_SET_DEDUP(bp, x)		BF64_SET((bp)->blk_prop, 62, 1, x)
415fa9e406ahrens
41643466aaMax Grossman#define	BP_GET_BYTEORDER(bp)		BF64_GET((bp)->blk_prop, 63, 1)
417b24ab67Jeff Bonwick#define	BP_SET_BYTEORDER(bp, x)		BF64_SET((bp)->blk_prop, 63, 1, x)
418b24ab67Jeff Bonwick
419b24ab67Jeff Bonwick#define	BP_PHYSICAL_BIRTH(bp)		\
4205d7b4d4Matthew Ahrens	(BP_IS_EMBEDDED(bp) ? 0 : \
4215d7b4d4Matthew Ahrens	(bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
422b24ab67Jeff Bonwick
423b24ab67Jeff Bonwick#define	BP_SET_BIRTH(bp, logical, physical)	\
424b24ab67Jeff Bonwick{						\
4255d7b4d4Matthew Ahrens	ASSERT(!BP_IS_EMBEDDED(bp));		\
426b24ab67Jeff Bonwick	(bp)->blk_birth = (logical);		\
427b24ab67Jeff Bonwick	(bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
428b24ab67Jeff Bonwick}
429fa9e406ahrens
4305d7b4d4Matthew Ahrens#define	BP_GET_FILL(bp) (BP_IS_EMBEDDED(bp) ? 1 : (bp)->blk_fill)
4315d7b4d4Matthew Ahrens
432770499eDan Kimmel#define	BP_IS_METADATA(bp)	\
433770499eDan Kimmel	(BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))
434770499eDan Kimmel
435fa9e406ahrens#define	BP_GET_ASIZE(bp)	\
4365d7b4d4Matthew Ahrens	(BP_IS_EMBEDDED(bp) ? 0 : \
4375d7b4d4Matthew Ahrens	DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
4385d7b4d4Matthew Ahrens	DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
4395d7b4d4Matthew Ahrens	DVA_GET_ASIZE(&(bp)->blk_dva[2]))
44099653d4eschrock
441770499eDan Kimmel#define	BP_GET_UCSIZE(bp)	\
442770499eDan Kimmel	(BP_IS_METADATA(bp) ? BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
443fa9e406ahrens
44444cd46cbillm#define	BP_GET_NDVAS(bp)	\
4455d7b4d4Matthew Ahrens	(BP_IS_EMBEDDED(bp) ? 0 : \
4465d7b4d4Matthew Ahrens	!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
44744cd46cbillm	!!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
44844cd46cbillm	!!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
44944cd46cbillm
45044cd46cbillm#define	BP_COUNT_GANG(bp)	\
4515d7b4d4Matthew Ahrens	(BP_IS_EMBEDDED(bp) ? 0 : \
45244cd46cbillm	(DVA_GET_GANG(&(bp)->blk_dva[0]) + \
45344cd46cbillm	DVA_GET_GANG(&(bp)->blk_dva[1]) + \
4545d7b4d4Matthew Ahrens	DVA_GET_GANG(&(bp)->blk_dva[2])))
45544cd46cbillm
456fa9e406ahrens#define	DVA_EQUAL(dva1, dva2)	\
457fa9e406ahrens	((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
458fa9e406ahrens	(dva1)->dva_word[0] == (dva2)->dva_word[0])
459fa9e406ahrens
460b24ab67Jeff Bonwick#define	BP_EQUAL(bp1, bp2)	\
461b24ab67Jeff Bonwick	(BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) &&	\
4625d7b4d4Matthew Ahrens	(bp1)->blk_birth == (bp2)->blk_birth &&			\
463b24ab67Jeff Bonwick	DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) &&	\
464b24ab67Jeff Bonwick	DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) &&	\
465b24ab67Jeff Bonwick	DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
466b24ab67Jeff Bonwick
4676b4acc8ahrens#define	ZIO_CHECKSUM_EQUAL(zc1, zc2) \
4686b4acc8ahrens	(0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
4696b4acc8ahrens	((zc1).zc_word[1] - (zc2).zc_word[1]) | \
4706b4acc8ahrens	((zc1).zc_word[2] - (zc2).zc_word[2]) | \
4716b4acc8ahrens	((zc1).zc_word[3] - (zc2).zc_word[3])))
4726b4acc8ahrens
47398110f0Matthew Ahrens#define	ZIO_CHECKSUM_IS_ZERO(zc) \
47498110f0Matthew Ahrens	(0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \
47598110f0Matthew Ahrens	(zc)->zc_word[2] | (zc)->zc_word[3]))
47698110f0Matthew Ahrens
47798110f0Matthew Ahrens#define	ZIO_CHECKSUM_BSWAP(zcp)					\
47898110f0Matthew Ahrens{								\
47998110f0Matthew Ahrens	(zcp)->zc_word[0] = BSWAP_64((zcp)->zc_word[0]);	\
48098110f0Matthew Ahrens	(zcp)->zc_word[1] = BSWAP_64((zcp)->zc_word[1]);	\
48198110f0Matthew Ahrens	(zcp)->zc_word[2] = BSWAP_64((zcp)->zc_word[2]);	\
48298110f0Matthew Ahrens	(zcp)->zc_word[3] = BSWAP_64((zcp)->zc_word[3]);	\
48398110f0Matthew Ahrens}
48498110f0Matthew Ahrens
48598110f0Matthew Ahrens
486fa9e406ahrens#define	DVA_IS_VALID(dva)	(DVA_GET_ASIZE(dva) != 0)
487fa9e406ahrens
488fa9e406ahrens#define	ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3)	\
489fa9e406ahrens{						\
490fa9e406ahrens	(zcp)->zc_word[0] = w0;			\
491fa9e406ahrens	(zcp)->zc_word[1] = w1;			\
492fa9e406ahrens	(zcp)->zc_word[2] = w2;			\
493fa9e406ahrens	(zcp)->zc_word[3] = w3;			\
494fa9e406ahrens}
495fa9e406ahrens
4965d7b4d4Matthew Ahrens#define	BP_IDENTITY(bp)		(ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0])
4975d7b4d4Matthew Ahrens#define	BP_IS_GANG(bp)		\
4985d7b4d4Matthew Ahrens	(BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp)))
49943466aaMax Grossman#define	DVA_IS_EMPTY(dva)	((dva)->dva_word[0] == 0ULL &&	\
50043466aaMax Grossman				(dva)->dva_word[1] == 0ULL)
5015d7b4d4Matthew Ahrens#define	BP_IS_HOLE(bp) \
5025d7b4d4Matthew Ahrens	(!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp)))
503fa9e406ahrens
5046e1f5caNeil Perrin/* BP_IS_RAIDZ(bp) assumes no block compression */
5056e1f5caNeil Perrin#define	BP_IS_RAIDZ(bp)		(DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
5066e1f5caNeil Perrin				BP_GET_PSIZE(bp))
5076e1f5caNeil Perrin
508e14bb32Jeff Bonwick#define	BP_ZERO(bp)				\
509fa9e406ahrens{						\
510fa9e406ahrens	(bp)->blk_dva[0].dva_word[0] = 0;	\
511fa9e406ahrens	(bp)->blk_dva[0].dva_word[1] = 0;	\
512fa9e406ahrens	(bp)->blk_dva[1].dva_word[0] = 0;	\
513fa9e406ahrens	(bp)->blk_dva[1].dva_word[1] = 0;	\
514fa9e406ahrens	(bp)->blk_dva[2].dva_word[0] = 0;	\
515fa9e406ahrens	(bp)->blk_dva[2].dva_word[1] = 0;	\
516fa9e406ahrens	(bp)->blk_prop = 0;			\
517fa9e406ahrens	(bp)->blk_pad[0] = 0;			\
518fa9e406ahrens	(bp)->blk_pad[1] = 0;			\
519b24ab67Jeff Bonwick	(bp)->blk_phys_birth = 0;		\
520e14bb32Jeff Bonwick	(bp)->blk_birth = 0;			\
521fa9e406ahrens	(bp)->blk_fill = 0;			\
522fa9e406ahrens	ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0);	\
523fa9e406ahrens}
524fa9e406ahrens
525fa9e406ahrens#ifdef _BIG_ENDIAN
526fa9e406ahrens#define	ZFS_HOST_BYTEORDER	(0ULL)
527fa9e406ahrens#else
52843466aaMax Grossman#define	ZFS_HOST_BYTEORDER	(1ULL)
529fa9e406ahrens#endif
530fa9e406ahrens
531fa9e406ahrens#define	BP_SHOULD_BYTESWAP(bp)	(BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
532fa9e406ahrens
53344cd46cbillm#define	BP_SPRINTF_LEN	320
534fbabab8maybee
535b24ab67Jeff Bonwick/*
536b24ab67Jeff Bonwick * This macro allows code sharing between zfs, libzpool, and mdb.
537b24ab67Jeff Bonwick * 'func' is either snprintf() or mdb_snprintf().
538b24ab67Jeff Bonwick * 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
539b24ab67Jeff Bonwick */
54043466aaMax Grossman#define	SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
541b24ab67Jeff Bonwick{									\
542b24ab67Jeff Bonwick	static const char *copyname[] =					\
543b24ab67Jeff Bonwick	    { "zero", "single", "double", "triple" };			\
544b24ab67Jeff Bonwick	int len = 0;							\
545b24ab67Jeff Bonwick	int copies = 0;							\
546b24ab67Jeff Bonwick									\
547b24ab67Jeff Bonwick	if (bp == NULL) {						\
54843466aaMax Grossman		len += func(buf + len, size - len, "<NULL>");		\
549b24ab67Jeff Bonwick	} else if (BP_IS_HOLE(bp)) {					\
55070163acPrakash Surya		len += func(buf + len, size - len,			\
55170163acPrakash Surya		    "HOLE [L%llu %s] "					\
55270163acPrakash Surya		    "size=%llxL birth=%lluL",				\
55370163acPrakash Surya		    (u_longlong_t)BP_GET_LEVEL(bp),			\
55470163acPrakash Surya		    type,						\
55570163acPrakash Surya		    (u_longlong_t)BP_GET_LSIZE(bp),			\
55670163acPrakash Surya		    (u_longlong_t)bp->blk_birth);			\
5575d7b4d4Matthew Ahrens	} else if (BP_IS_EMBEDDED(bp)) {				\
5585d7b4d4Matthew Ahrens		len = func(buf + len, size - len,			\
5595d7b4d4Matthew Ahrens		    "EMBEDDED [L%llu %s] et=%u %s "			\
5605d7b4d4Matthew Ahrens		    "size=%llxL/%llxP birth=%lluL",			\
5615d7b4d4Matthew Ahrens		    (u_longlong_t)BP_GET_LEVEL(bp),			\
5625d7b4d4Matthew Ahrens		    type,						\
5635d7b4d4Matthew Ahrens		    (int)BPE_GET_ETYPE(bp),				\
5645d7b4d4Matthew Ahrens		    compress,						\
5655d7b4d4Matthew Ahrens		    (u_longlong_t)BPE_GET_LSIZE(bp),			\
5665d7b4d4Matthew Ahrens		    (u_longlong_t)BPE_GET_PSIZE(bp),			\
5675d7b4d4Matthew Ahrens		    (u_longlong_t)bp->blk_birth);			\
568b24ab67Jeff Bonwick	} else {							\
569b24ab67Jeff Bonwick		for (int d = 0; d < BP_GET_NDVAS(bp); d++) {		\
570b24ab67Jeff Bonwick			const dva_t *dva = &bp->blk_dva[d];		\
571b24ab67Jeff Bonwick			if (DVA_IS_VALID(dva))				\
572b24ab67Jeff Bonwick				copies++;				\
573b24ab67Jeff Bonwick			len += func(buf + len, size - len,		\
574b24ab67Jeff Bonwick			    "DVA[%d]=<%llu:%llx:%llx>%c", d,		\
575b24ab67Jeff Bonwick			    (u_longlong_t)DVA_GET_VDEV(dva),		\
576b24ab67Jeff Bonwick			    (u_longlong_t)DVA_GET_OFFSET(dva),		\
577b24ab67Jeff Bonwick			    (u_longlong_t)DVA_GET_ASIZE(dva),		\
578b24ab67Jeff Bonwick			    ws);					\
579b24ab67Jeff Bonwick		}							\
580b24ab67Jeff Bonwick		if (BP_IS_GANG(bp) &&					\
581b24ab67Jeff Bonwick		    DVA_GET_ASIZE(&bp->blk_dva[2]) <=			\
582b24ab67Jeff Bonwick		    DVA_GET_ASIZE(&bp->blk_dva[1]) / 2)			\
583b24ab67Jeff Bonwick			copies--;					\
584b24ab67Jeff Bonwick		len += func(buf + len, size - len,			\
585b24ab67Jeff Bonwick		    "[L%llu %s] %s %s %s %s %s %s%c"			\
586b24ab67Jeff Bonwick		    "size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c"	\
587b24ab67Jeff Bonwick		    "cksum=%llx:%llx:%llx:%llx",			\
588b24ab67Jeff Bonwick		    (u_longlong_t)BP_GET_LEVEL(bp),			\
589b24ab67Jeff Bonwick		    type,						\
590b24ab67Jeff Bonwick		    checksum,						\
591b24ab67Jeff Bonwick		    compress,						\
592b24ab67Jeff Bonwick		    BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE",		\
593b24ab67Jeff Bonwick		    BP_IS_GANG(bp) ? "gang" : "contiguous",		\
594b24ab67Jeff Bonwick		    BP_GET_DEDUP(bp) ? "dedup" : "unique",		\
595b24ab67Jeff Bonwick		    copyname[copies],					\
596b24ab67Jeff Bonwick		    ws,							\
597b24ab67Jeff Bonwick		    (u_longlong_t)BP_GET_LSIZE(bp),			\
598b24ab67Jeff Bonwick		    (u_longlong_t)BP_GET_PSIZE(bp),			\
599b24ab67Jeff Bonwick		    (u_longlong_t)bp->blk_birth,			\
600b24ab67Jeff Bonwick		    (u_longlong_t)BP_PHYSICAL_BIRTH(bp),		\
6015d7b4d4Matthew Ahrens		    (u_longlong_t)BP_GET_FILL(bp),			\
602b24ab67Jeff Bonwick		    ws,							\
603b24ab67Jeff Bonwick		    (u_longlong_t)bp->blk_cksum.zc_word[0],		\
604b24ab67Jeff Bonwick		    (u_longlong_t)bp->blk_cksum.zc_word[1],		\
605b24ab67Jeff Bonwick		    (u_longlong_t)bp->blk_cksum.zc_word[2],		\
606b24ab67Jeff Bonwick		    (u_longlong_t)bp->blk_cksum.zc_word[3]);		\
607b24ab67Jeff Bonwick	}								\
608b24ab67Jeff Bonwick	ASSERT(len < size);						\
609b24ab67Jeff Bonwick}
610b24ab67Jeff Bonwick
611ad23a2djohansen#define	BP_GET_BUFC_TYPE(bp)						\
612770499eDan Kimmel	(BP_IS_METADATA(bp) ? ARC_BUFC_METADATA : ARC_BUFC_DATA)
613fa9e406ahrens
6141195e68Mark J Musantetypedef enum spa_import_type {
6151195e68Mark J Musante	SPA_IMPORT_EXISTING,
6161195e68Mark J Musante	SPA_IMPORT_ASSEMBLE
6171195e68Mark J Musante} spa_import_type_t;
6181195e68Mark J Musante
619fa9e406ahrens/* state manipulation functions */
620fa9e406ahrensextern int spa_open(const char *pool, spa_t **, void *tag);
621468c413Tim Haleyextern int spa_open_rewind(const char *pool, spa_t **, void *tag,
622468c413Tim Haley    nvlist_t *policy, nvlist_t **config);
623ad135b5Christopher Sidenextern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot,
624ad135b5Christopher Siden    size_t buflen);
625990b485llingextern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props,
6264445fffMatthew Ahrens    nvlist_t *zplprops);
627051aabetaylorextern int spa_import_rootpool(char *devpath, char *devid);
6284b964adGeorge Wilsonextern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props,
6294b964adGeorge Wilson    uint64_t flags);
630fa9e406ahrensextern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
631fa9e406ahrensextern int spa_destroy(char *pool);
6328671400Serapheim Dimitropoulosextern int spa_checkpoint(const char *pool);
6338671400Serapheim Dimitropoulosextern int spa_checkpoint_discard(const char *pool);
634394ab0cGeorge Wilsonextern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force,
635394ab0cGeorge Wilson    boolean_t hardforce);
636ea8dc4beschrockextern int spa_reset(char *pool);
637ea8dc4beschrockextern void spa_async_request(spa_t *spa, int flag);
638088f389ahrensextern void spa_async_unrequest(spa_t *spa, int flag);
639ea8dc4beschrockextern void spa_async_suspend(spa_t *spa);
640ea8dc4beschrockextern void spa_async_resume(spa_t *spa);
641ea8dc4beschrockextern spa_t *spa_inject_addref(char *pool);
642ea8dc4beschrockextern void spa_inject_delref(spa_t *spa);
6433f9d6adLin Lingextern void spa_scan_stat_init(spa_t *spa);
6443f9d6adLin Lingextern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps);
645ea8dc4beschrock
646e14bb32Jeff Bonwick#define	SPA_ASYNC_CONFIG_UPDATE	0x01
647e14bb32Jeff Bonwick#define	SPA_ASYNC_REMOVE	0x02
648e14bb32Jeff Bonwick#define	SPA_ASYNC_PROBE		0x04
649e14bb32Jeff Bonwick#define	SPA_ASYNC_RESILVER_DONE	0x08
650e14bb32Jeff Bonwick#define	SPA_ASYNC_RESILVER	0x10
651573ca77George Wilson#define	SPA_ASYNC_AUTOEXPAND	0x20
6523f9d6adLin Ling#define	SPA_ASYNC_REMOVE_DONE	0x40
6533f9d6adLin Ling#define	SPA_ASYNC_REMOVE_STOP	0x80
654094e47eGeorge Wilson#define	SPA_ASYNC_INITIALIZE_RESTART	0x100
6553f9d6adLin Ling
6563f9d6adLin Ling/*
6573f9d6adLin Ling * Controls the behavior of spa_vdev_remove().
6583f9d6adLin Ling */
6593f9d6adLin Ling#define	SPA_REMOVE_UNSPARE	0x01
6603f9d6adLin Ling#define	SPA_REMOVE_DONE		0x02
661fa9e406ahrens
662fa9e406ahrens/* device manipulation */
663fa9e406ahrensextern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
664ea8dc4beschrockextern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot,
665fa9e406ahrens    int replacing);
6668ad4d6dJeff Bonwickextern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
6678ad4d6dJeff Bonwick    int replace_done);
66899653d4eschrockextern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare);
6693f9d6adLin Lingextern boolean_t spa_vdev_remove_active(spa_t *spa);
670094e47eGeorge Wilsonextern int spa_vdev_initialize(spa_t *spa, uint64_t guid, uint64_t cmd_type);
671c67d967eschrockextern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
6726809eb4Eric Schrockextern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru);
6731195e68Mark J Musanteextern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
6741195e68Mark J Musante    nvlist_t *props, boolean_t exp);
675fa9e406ahrens
67699653d4eschrock/* spare state (which is global across all pools) */
67739c2341eschrockextern void spa_spare_add(vdev_t *vd);
67839c2341eschrockextern void spa_spare_remove(vdev_t *vd);
67989a89ebllingextern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt);
68039c2341eschrockextern void spa_spare_activate(vdev_t *vd);
68199653d4eschrock
682fa94a07brendan/* L2ARC state (which is global across all pools) */
683fa94a07brendanextern void spa_l2cache_add(vdev_t *vd);
684fa94a07brendanextern void spa_l2cache_remove(vdev_t *vd);
685fa94a07brendanextern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
686fa94a07brendanextern void spa_l2cache_activate(vdev_t *vd);
687fa94a07brendanextern void spa_l2cache_drop(spa_t *spa);
688fa94a07brendan
6893f9d6adLin Ling/* scanning */
6903f9d6adLin Lingextern int spa_scan(spa_t *spa, pool_scan_func_t func);
6913f9d6adLin Lingextern int spa_scan_stop(spa_t *spa);
6921702cceAlek Pinchukextern int spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t flag);
693fa9e406ahrens
694fa9e406ahrens/* spa syncing */
695fa9e406ahrensextern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
696fa9e406ahrensextern void spa_sync_allpools(void);
697fa9e406ahrens
6983a737e0brendan/* spa namespace global mutex */
6993a737e0brendanextern kmutex_t spa_namespace_lock;
7003a737e0brendan
701fa9e406ahrens/*
702fa9e406ahrens * SPA configuration functions in spa_config.c
703fa9e406ahrens */
7040373e76bonwick
7050373e76bonwick#define	SPA_CONFIG_UPDATE_POOL	0
7060373e76bonwick#define	SPA_CONFIG_UPDATE_VDEVS	1
7070373e76bonwick
7085cabbc6Prashanth Sreenivasaextern void spa_write_cachefile(spa_t *, boolean_t, boolean_t);
709fa9e406ahrensextern void spa_config_load(void);
710fa9e406ahrensextern nvlist_t *spa_all_configs(uint64_t *);
711fa9e406ahrensextern void spa_config_set(spa_t *spa, nvlist_t *config);
712fa9e406ahrensextern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
713fa9e406ahrens    int getstats);
7140373e76bonwickextern void spa_config_update(spa_t *spa, int what);
715fa9e406ahrens
716fa9e406ahrens/*
717fa9e406ahrens * Miscellaneous SPA routines in spa_misc.c
718fa9e406ahrens */
719fa9e406ahrens
720fa9e406ahrens/* Namespace manipulation */
721fa9e406ahrensextern spa_t *spa_lookup(const char *name);
722468c413Tim Haleyextern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot);
723fa9e406ahrensextern void spa_remove(spa_t *spa);
724fa9e406ahrensextern spa_t *spa_next(spa_t *prev);
725fa9e406ahrens
726fa9e406ahrens/* Refcount functions */
727fa9e406ahrensextern void spa_open_ref(spa_t *spa, void *tag);
728fa9e406ahrensextern void spa_close(spa_t *spa, void *tag);
729bc9014eJustin Gibbsextern void spa_async_close(spa_t *spa, void *tag);
730fa9e406ahrensextern boolean_t spa_refcount_zero(spa_t *spa);
731fa9e406ahrens
7328f18d1fGeorge Wilson#define	SCL_NONE	0x00
733e14bb32Jeff Bonwick#define	SCL_CONFIG	0x01
734e14bb32Jeff Bonwick#define	SCL_STATE	0x02
735e14bb32Jeff Bonwick#define	SCL_L2ARC	0x04		/* hack until L2ARC 2.0 */
736e14bb32Jeff Bonwick#define	SCL_ALLOC	0x08
737e14bb32Jeff Bonwick#define	SCL_ZIO		0x10
738e14bb32Jeff Bonwick#define	SCL_FREE	0x20
739e14bb32Jeff Bonwick#define	SCL_VDEV	0x40
740e14bb32Jeff Bonwick#define	SCL_LOCKS	7
741e14bb32Jeff Bonwick#define	SCL_ALL		((1 << SCL_LOCKS) - 1)
742e14bb32Jeff Bonwick#define	SCL_STATE_ALL	(SCL_STATE | SCL_L2ARC | SCL_ZIO)
743e14bb32Jeff Bonwick
744e14bb32Jeff Bonwick/* Pool configuration locks */
745e14bb32Jeff Bonwickextern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw);
746e14bb32Jeff Bonwickextern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw);
747e14bb32Jeff Bonwickextern void spa_config_exit(spa_t *spa, int locks, void *tag);
748e14bb32Jeff Bonwickextern int spa_config_held(spa_t *spa, int locks, krw_t rw);
749fa9e406ahrens
750fa9e406ahrens/* Pool vdev add/remove lock */
751fa9e406ahrensextern uint64_t spa_vdev_enter(spa_t *spa);
75288ecc94George Wilsonextern uint64_t spa_vdev_config_enter(spa_t *spa);
75388ecc94George Wilsonextern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg,
75488ecc94George Wilson    int error, char *tag);
755fa9e406ahrensextern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error);
756fa9e406ahrens
757e14bb32Jeff Bonwick/* Pool vdev state change lock */
7588f18d1fGeorge Wilsonextern void spa_vdev_state_enter(spa_t *spa, int oplock);
759e14bb32Jeff Bonwickextern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error);
760e14bb32Jeff Bonwick
761b24ab67Jeff Bonwick/* Log state */
762b24ab67Jeff Bonwicktypedef enum spa_log_state {
763b24ab67Jeff Bonwick	SPA_LOG_UNKNOWN = 0,	/* unknown log state */
764b24ab67Jeff Bonwick	SPA_LOG_MISSING,	/* missing log(s) */
765b24ab67Jeff Bonwick	SPA_LOG_CLEAR,		/* clear the log(s) */
766b24ab67Jeff Bonwick	SPA_LOG_GOOD,		/* log(s) are good */
767b24ab67Jeff Bonwick} spa_log_state_t;
768b24ab67Jeff Bonwick
769b24ab67Jeff Bonwickextern spa_log_state_t spa_get_log_state(spa_t *spa);
770b24ab67Jeff Bonwickextern void spa_set_log_state(spa_t *spa, spa_log_state_t state);
7715cabbc6Prashanth Sreenivasaextern int spa_reset_logs(spa_t *spa);
772b24ab67Jeff Bonwick
773b24ab67Jeff Bonwick/* Log claim callback */
774b24ab67Jeff Bonwickextern void spa_claim_notify(zio_t *zio);
775b24ab67Jeff Bonwick
776fa9e406ahrens/* Accessor functions */
77788b7b0fMatthew Ahrensextern boolean_t spa_shutting_down(spa_t *spa);
778fa9e406ahrensextern struct dsl_pool *spa_get_dsl(spa_t *spa);
779ad135b5Christopher Sidenextern boolean_t spa_is_initializing(spa_t *spa);
7805cabbc6Prashanth Sreenivasaextern boolean_t spa_indirect_vdevs_loaded(spa_t *spa);
781fa9e406ahrensextern blkptr_t *spa_get_rootblkptr(spa_t *spa);
782fa9e406ahrensextern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
783fa9e406ahrensextern void spa_altroot(spa_t *, char *, size_t);
784fa9e406ahrensextern int spa_sync_pass(spa_t *spa);
785fa9e406ahrensextern char *spa_name(spa_t *spa);
786fa9e406ahrensextern uint64_t spa_guid(spa_t *spa);
787e9103aaGarrett D'Amoreextern uint64_t spa_load_guid(spa_t *spa);
788fa9e406ahrensextern uint64_t spa_last_synced_txg(spa_t *spa);
789fa9e406ahrensextern uint64_t spa_first_txg(spa_t *spa);
790b24ab67Jeff Bonwickextern uint64_t spa_syncing_txg(spa_t *spa);
7913991b53George Wilsonextern uint64_t spa_final_dirty_txg(spa_t *spa);
79299653d4eschrockextern uint64_t spa_version(spa_t *spa);
79388b7b0fMatthew Ahrensextern pool_state_t spa_state(spa_t *spa);
794b16da2eGeorge Wilsonextern spa_load_state_t spa_load_state(spa_t *spa);
795fa9e406ahrensextern uint64_t spa_freeze_txg(spa_t *spa);
79661e255cMatthew Ahrensextern uint64_t spa_get_worst_case_asize(spa_t *spa, uint64_t lsize);
797485bbbfGeorge Wilsonextern uint64_t spa_get_dspace(spa_t *spa);
7988671400Serapheim Dimitropoulosextern uint64_t spa_get_checkpoint_space(spa_t *spa);
799c39f2c8Christopher Sidenextern uint64_t spa_get_slop_space(spa_t *spa);
800485bbbfGeorge Wilsonextern void spa_update_dspace(spa_t *spa);
80144cd46cbillmextern uint64_t spa_version(spa_t *spa);
802b24ab67Jeff Bonwickextern boolean_t spa_deflate(spa_t *spa);
803b24ab67Jeff Bonwickextern metaslab_class_t *spa_normal_class(spa_t *spa);
804b24ab67Jeff Bonwickextern metaslab_class_t *spa_log_class(spa_t *spa);
805663207aDon Bradyextern metaslab_class_t *spa_special_class(spa_t *spa);
806663207aDon Bradyextern metaslab_class_t *spa_dedup_class(spa_t *spa);
807663207aDon Bradyextern metaslab_class_t *spa_preferred_class(spa_t *spa, uint64_t size,
808663207aDon Brady    dmu_object_type_t objtype, uint_t level, uint_t special_smallblk);
809663207aDon Brady
810bc9014eJustin Gibbsextern void spa_evicting_os_register(spa_t *, objset_t *os);
811bc9014eJustin Gibbsextern void spa_evicting_os_deregister(spa_t *, objset_t *os);
812bc9014eJustin Gibbsextern void spa_evicting_os_wait(spa_t *spa);
81344cd46cbillmextern int spa_max_replication(spa_t *spa);
8143f9d6adLin Lingextern int spa_prev_software_version(spa_t *spa);
815fa9e406ahrensextern int spa_busy(void);
8160a4e951gwextern uint8_t spa_get_failmode(spa_t *spa);
817e14bb32Jeff Bonwickextern boolean_t spa_suspended(spa_t *spa);
818b24ab67Jeff Bonwickextern uint64_t spa_bootfs(spa_t *spa);
819b24ab67Jeff Bonwickextern uint64_t spa_delegation(spa_t *spa);
820b24ab67Jeff Bonwickextern