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 2009 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 /*
27  * Copyright (c) 2012, 2019 by Delphix. All rights reserved.
28  */
29 
30 #ifndef _SYS_SPACE_MAP_H
31 #define	_SYS_SPACE_MAP_H
32 
33 #include <sys/avl.h>
34 #include <sys/range_tree.h>
35 #include <sys/dmu.h>
36 
37 #ifdef	__cplusplus
38 extern "C" {
39 #endif
40 
41 /*
42  * The size of the space map object has increased to include a histogram.
43  * The SPACE_MAP_SIZE_V0 designates the original size and is used to
44  * maintain backward compatibility.
45  */
46 #define	SPACE_MAP_SIZE_V0	(3 * sizeof (uint64_t))
47 #define	SPACE_MAP_HISTOGRAM_SIZE	32
48 
49 /*
50  * The space_map_phys is the on-disk representation of the space map.
51  * Consumers of space maps should never reference any of the members of this
52  * structure directly. These members may only be updated in syncing context.
53  *
54  * Note the smp_object is no longer used but remains in the structure
55  * for backward compatibility.
56  */
57 typedef struct space_map_phys {
58 	/* object number: not needed but kept for backwards compatibility */
59 	uint64_t	smp_object;
60 
61 	/* length of the object in bytes */
62 	uint64_t	smp_length;
63 
64 	/* space allocated from the map */
65 	int64_t		smp_alloc;
66 
67 	/* reserved */
68 	uint64_t	smp_pad[5];
69 
70 	/*
71 	 * The smp_histogram maintains a histogram of free regions. Each
72 	 * bucket, smp_histogram[i], contains the number of free regions
73 	 * whose size is:
74 	 * 2^(i+sm_shift) <= size of free region in bytes < 2^(i+sm_shift+1)
75 	 *
76 	 * Note that, if log space map feature is enabled, histograms of
77 	 * space maps that belong to metaslabs will take into account any
78 	 * unflushed changes for their metaslabs, even though the actual
79 	 * space map doesn't have entries for these changes.
80 	 */
81 	uint64_t	smp_histogram[SPACE_MAP_HISTOGRAM_SIZE];
82 } space_map_phys_t;
83 
84 /*
85  * The space map object defines a region of space, its size, how much is
86  * allocated, and the on-disk object that stores this information.
87  * Consumers of space maps may only access the members of this structure.
88  *
89  * Note: the space_map may not be accessed concurrently; consumers
90  * must provide external locking if required.
91  */
92 typedef struct space_map {
93 	uint64_t	sm_start;	/* start of map */
94 	uint64_t	sm_size;	/* size of map */
95 	uint8_t		sm_shift;	/* unit shift */
96 	objset_t	*sm_os;		/* objset for this map */
97 	uint64_t	sm_object;	/* object id for this map */
98 	uint32_t	sm_blksz;	/* block size for space map */
99 	dmu_buf_t	*sm_dbuf;	/* space_map_phys_t dbuf */
100 	space_map_phys_t *sm_phys;	/* on-disk space map */
101 } space_map_t;
102 
103 /*
104  * debug entry
105  *
106  *     2     2        10                     50
107  *  +-----+-----+------------+----------------------------------+
108  *  | 1 0 | act |  syncpass  |        txg (lower bits)          |
109  *  +-----+-----+------------+----------------------------------+
110  *   63 62 61 60 59        50 49                                0
111  *
112  *
113  * one-word entry
114  *
115  *    1               47                   1           15
116  *  +-----------------------------------------------------------+
117  *  | 0 |   offset (sm_shift units)    | type |       run       |
118  *  +-----------------------------------------------------------+
119  *   63  62                          16   15   14               0
120  *
121  *
122  * two-word entry
123  *
124  *     2     2               36                      24
125  *  +-----+-----+---------------------------+-------------------+
126  *  | 1 1 | pad |            run            |       vdev        |
127  *  +-----+-----+---------------------------+-------------------+
128  *   63 62 61 60 59                       24 23                 0
129  *
130  *     1                            63
131  *  +------+----------------------------------------------------+
132  *  | type |                      offset                        |
133  *  +------+----------------------------------------------------+
134  *     63   62                                                  0
135  *
136  * Note that a two-word entry will not strandle a block boundary.
137  * If necessary, the last word of a block will be padded with a
138  * debug entry (with act = syncpass = txg = 0).
139  */
140 
141 typedef enum {
142 	SM_ALLOC,
143 	SM_FREE
144 } maptype_t;
145 
146 typedef struct space_map_entry {
147 	maptype_t sme_type;
148 	uint32_t sme_vdev;	/* max is 2^24-1; SM_NO_VDEVID if not present */
149 	uint64_t sme_offset;	/* max is 2^63-1; units of sm_shift */
150 	uint64_t sme_run;	/* max is 2^36; units of sm_shift */
151 } space_map_entry_t;
152 
153 #define	SM_NO_VDEVID	(1 << SPA_VDEVBITS)
154 
155 /* one-word entry constants */
156 #define	SM_DEBUG_PREFIX	2
157 #define	SM_OFFSET_BITS	47
158 #define	SM_RUN_BITS	15
159 
160 /* two-word entry constants */
161 #define	SM2_PREFIX	3
162 #define	SM2_OFFSET_BITS	63
163 #define	SM2_RUN_BITS	36
164 
165 #define	SM_PREFIX_DECODE(x)	BF64_DECODE(x, 62, 2)
166 #define	SM_PREFIX_ENCODE(x)	BF64_ENCODE(x, 62, 2)
167 
168 #define	SM_DEBUG_ACTION_DECODE(x)	BF64_DECODE(x, 60, 2)
169 #define	SM_DEBUG_ACTION_ENCODE(x)	BF64_ENCODE(x, 60, 2)
170 #define	SM_DEBUG_SYNCPASS_DECODE(x)	BF64_DECODE(x, 50, 10)
171 #define	SM_DEBUG_SYNCPASS_ENCODE(x)	BF64_ENCODE(x, 50, 10)
172 #define	SM_DEBUG_TXG_DECODE(x)		BF64_DECODE(x, 0, 50)
173 #define	SM_DEBUG_TXG_ENCODE(x)		BF64_ENCODE(x, 0, 50)
174 
175 #define	SM_OFFSET_DECODE(x)	BF64_DECODE(x, 16, SM_OFFSET_BITS)
176 #define	SM_OFFSET_ENCODE(x)	BF64_ENCODE(x, 16, SM_OFFSET_BITS)
177 #define	SM_TYPE_DECODE(x)	BF64_DECODE(x, 15, 1)
178 #define	SM_TYPE_ENCODE(x)	BF64_ENCODE(x, 15, 1)
179 #define	SM_RUN_DECODE(x)	(BF64_DECODE(x, 0, SM_RUN_BITS) + 1)
180 #define	SM_RUN_ENCODE(x)	BF64_ENCODE((x) - 1, 0, SM_RUN_BITS)
181 #define	SM_RUN_MAX		SM_RUN_DECODE(~0ULL)
182 #define	SM_OFFSET_MAX		SM_OFFSET_DECODE(~0ULL)
183 
184 #define	SM2_RUN_DECODE(x)	(BF64_DECODE(x, SPA_VDEVBITS, SM2_RUN_BITS) + 1)
185 #define	SM2_RUN_ENCODE(x)	BF64_ENCODE((x) - 1, SPA_VDEVBITS, SM2_RUN_BITS)
186 #define	SM2_VDEV_DECODE(x)	BF64_DECODE(x, 0, SPA_VDEVBITS)
187 #define	SM2_VDEV_ENCODE(x)	BF64_ENCODE(x, 0, SPA_VDEVBITS)
188 #define	SM2_TYPE_DECODE(x)	BF64_DECODE(x, SM2_OFFSET_BITS, 1)
189 #define	SM2_TYPE_ENCODE(x)	BF64_ENCODE(x, SM2_OFFSET_BITS, 1)
190 #define	SM2_OFFSET_DECODE(x)	BF64_DECODE(x, 0, SM2_OFFSET_BITS)
191 #define	SM2_OFFSET_ENCODE(x)	BF64_ENCODE(x, 0, SM2_OFFSET_BITS)
192 #define	SM2_RUN_MAX		SM2_RUN_DECODE(~0ULL)
193 #define	SM2_OFFSET_MAX		SM2_OFFSET_DECODE(~0ULL)
194 
195 boolean_t sm_entry_is_debug(uint64_t e);
196 boolean_t sm_entry_is_single_word(uint64_t e);
197 boolean_t sm_entry_is_double_word(uint64_t e);
198 
199 typedef int (*sm_cb_t)(space_map_entry_t *sme, void *arg);
200 
201 int space_map_load(space_map_t *sm, range_tree_t *rt, maptype_t maptype);
202 int space_map_load_length(space_map_t *sm, range_tree_t *rt, maptype_t maptype,
203     uint64_t length);
204 int space_map_iterate(space_map_t *sm, uint64_t length,
205     sm_cb_t callback, void *arg);
206 int space_map_incremental_destroy(space_map_t *sm, sm_cb_t callback, void *arg,
207     dmu_tx_t *tx);
208 
209 boolean_t space_map_histogram_verify(space_map_t *sm, range_tree_t *rt);
210 void space_map_histogram_clear(space_map_t *sm);
211 void space_map_histogram_add(space_map_t *sm, range_tree_t *rt,
212     dmu_tx_t *tx);
213 
214 uint64_t space_map_object(space_map_t *sm);
215 int64_t space_map_allocated(space_map_t *sm);
216 uint64_t space_map_length(space_map_t *sm);
217 uint64_t space_map_entries(space_map_t *sm, range_tree_t *rt);
218 uint64_t space_map_nblocks(space_map_t *sm);
219 
220 void space_map_write(space_map_t *sm, range_tree_t *rt, maptype_t maptype,
221     uint64_t vdev_id, dmu_tx_t *tx);
222 uint64_t space_map_estimate_optimal_size(space_map_t *sm, range_tree_t *rt,
223     uint64_t vdev_id);
224 void space_map_truncate(space_map_t *sm, int blocksize, dmu_tx_t *tx);
225 uint64_t space_map_alloc(objset_t *os, int blocksize, dmu_tx_t *tx);
226 void space_map_free(space_map_t *sm, dmu_tx_t *tx);
227 void space_map_free_obj(objset_t *os, uint64_t smobj, dmu_tx_t *tx);
228 
229 int space_map_open(space_map_t **smp, objset_t *os, uint64_t object,
230     uint64_t start, uint64_t size, uint8_t shift);
231 void space_map_close(space_map_t *sm);
232 
233 #ifdef	__cplusplus
234 }
235 #endif
236 
237 #endif	/* _SYS_SPACE_MAP_H */
238