Print this page
4101 metaslab_debug should allow for fine-grained control
4102 space_maps should store more information about themselves
4103 space map object blocksize should be increased
4104 ::spa_space no longer works
4105 removing a mirrored log device results in a leaked object
4106 asynchronously load metaslab
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Sebastien Roy <seb@delphix.com>
Split |
Close |
Expand all |
Collapse all |
--- old/usr/src/uts/common/fs/zfs/sys/metaslab_impl.h
+++ new/usr/src/uts/common/fs/zfs/sys/metaslab_impl.h
1 1 /*
2 2 * CDDL HEADER START
3 3 *
4 4 * The contents of this file are subject to the terms of the
5 5 * Common Development and Distribution License (the "License").
6 6 * You may not use this file except in compliance with the License.
7 7 *
8 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 9 * or http://www.opensolaris.org/os/licensing.
10 10 * See the License for the specific language governing permissions
11 11 * and limitations under the License.
12 12 *
13 13 * When distributing Covered Code, include this CDDL HEADER in each
14 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 15 * If applicable, add the following below this CDDL HEADER, with the
16 16 * fields enclosed by brackets "[]" replaced with your own identifying
17 17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 18 *
19 19 * CDDL HEADER END
20 20 */
21 21 /*
22 22 * Copyright 2009 Sun Microsystems, Inc. All rights reserved.
23 23 * Use is subject to license terms.
24 24 */
↓ open down ↓ |
24 lines elided |
↑ open up ↑ |
25 25
26 26 /*
27 27 * Copyright (c) 2013 by Delphix. All rights reserved.
28 28 */
29 29
30 30 #ifndef _SYS_METASLAB_IMPL_H
31 31 #define _SYS_METASLAB_IMPL_H
32 32
33 33 #include <sys/metaslab.h>
34 34 #include <sys/space_map.h>
35 +#include <sys/range_tree.h>
35 36 #include <sys/vdev.h>
36 37 #include <sys/txg.h>
37 38 #include <sys/avl.h>
38 39
39 40 #ifdef __cplusplus
40 41 extern "C" {
41 42 #endif
42 43
43 44 struct metaslab_class {
44 45 spa_t *mc_spa;
45 46 metaslab_group_t *mc_rotor;
46 - space_map_ops_t *mc_ops;
47 + metaslab_ops_t *mc_ops;
47 48 uint64_t mc_aliquot;
48 49 uint64_t mc_alloc_groups; /* # of allocatable groups */
49 50 uint64_t mc_alloc; /* total allocated space */
50 51 uint64_t mc_deferred; /* total deferred frees */
51 52 uint64_t mc_space; /* total space (alloc + free) */
52 53 uint64_t mc_dspace; /* total deflated space */
53 54 };
54 55
55 56 struct metaslab_group {
56 57 kmutex_t mg_lock;
57 58 avl_tree_t mg_metaslab_tree;
58 59 uint64_t mg_aliquot;
59 - uint64_t mg_bonus_area;
60 60 uint64_t mg_alloc_failures;
61 61 boolean_t mg_allocatable; /* can we allocate? */
62 62 uint64_t mg_free_capacity; /* percentage free */
63 63 int64_t mg_bias;
64 64 int64_t mg_activation_count;
65 65 metaslab_class_t *mg_class;
66 66 vdev_t *mg_vd;
67 + taskq_t *mg_taskq;
67 68 metaslab_group_t *mg_prev;
68 69 metaslab_group_t *mg_next;
69 70 };
70 71
71 72 /*
72 - * Each metaslab maintains an in-core free map (ms_map) that contains the
73 - * current list of free segments. As blocks are allocated, the allocated
74 - * segment is removed from the ms_map and added to a per txg allocation map.
75 - * As blocks are freed, they are added to the per txg free map. These per
76 - * txg maps allow us to process all allocations and frees in syncing context
77 - * where it is safe to update the on-disk space maps.
73 + * This value defines the number of elements in the ms_lbas array. The value
74 + * of 64 was chosen as it covers to cover all power of 2 buckets up to
75 + * UINT64_MAX. This is the equivalent of highbit(UINT64_MAX).
76 + */
77 +#define MAX_LBAS 64
78 +
79 +/*
80 + * Each metaslab maintains a set of in-core trees to track metaslab operations.
81 + * The in-core free tree (ms_tree) contains the current list of free segments.
82 + * As blocks are allocated, the allocated segment are removed from the ms_tree
83 + * and added to a per txg allocation tree (ms_alloctree). As blocks are freed,
84 + * they are added to the per txg free tree (ms_freetree). These per txg
85 + * trees allow us to process all allocations and frees in syncing context
86 + * where it is safe to update the on-disk space maps. One additional in-core
87 + * tree is maintained to track deferred frees (ms_defertree). Once a block
88 + * is freed it will move from the ms_freetree to the ms_defertree. A deferred
89 + * free means that a block has been freed but cannot be used by the pool
90 + * until TXG_DEFER_SIZE transactions groups later. For example, a block
91 + * that is freed in txg 50 will not be available for reallocation until
92 + * txg 52 (50 + TXG_DEFER_SIZE). This provides a safety net for uberblock
93 + * rollback. A pool could be safely rolled back TXG_DEFERS_SIZE
94 + * transactions groups and ensure that no block has been reallocated.
78 95 *
79 - * Each metaslab's free space is tracked in a space map object in the MOS,
96 + * The simplified transition diagram looks like this:
97 + *
98 + *
99 + * ALLOCATE
100 + * |
101 + * V
102 + * free segment (ms_tree) --------> ms_alloctree ----> (write to space map)
103 + * ^
104 + * |
105 + * | ms_freetree <--- FREE
106 + * | |
107 + * | |
108 + * | |
109 + * +----------- ms_defertree <-------+---------> (write to space map)
110 + *
111 + *
112 + * Each metaslab's space is tracked in a single space map in the MOS,
80 113 * which is only updated in syncing context. Each time we sync a txg,
81 - * we append the allocs and frees from that txg to the space map object.
82 - * When the txg is done syncing, metaslab_sync_done() updates ms_smo
83 - * to ms_smo_syncing. Everything in ms_smo is always safe to allocate.
114 + * we append the allocs and frees from that txg to the space map.
115 + * The pool space is only updated once all metaslabs have finished syncing.
84 116 *
85 - * To load the in-core free map we read the space map object from disk.
117 + * To load the in-core free tree we read the space map from disk.
86 118 * This object contains a series of alloc and free records that are
87 119 * combined to make up the list of all free segments in this metaslab. These
88 - * segments are represented in-core by the ms_map and are stored in an
120 + * segments are represented in-core by the ms_tree and are stored in an
89 121 * AVL tree.
90 122 *
91 - * As the space map objects grows (as a result of the appends) it will
92 - * eventually become space-inefficient. When the space map object is
93 - * zfs_condense_pct/100 times the size of the minimal on-disk representation,
94 - * we rewrite it in its minimized form.
123 + * As the space map grows (as a result of the appends) it will
124 + * eventually become space-inefficient. When the metaslab's in-core free tree
125 + * is zfs_condense_pct/100 times the size of the minimal on-disk
126 + * representation, we rewrite it in its minimized form. If a metaslab
127 + * needs to condense then we must set the ms_condensing flag to ensure
128 + * that allocations are not performed on the metaslab that is being written.
95 129 */
96 130 struct metaslab {
97 - kmutex_t ms_lock; /* metaslab lock */
98 - space_map_obj_t ms_smo; /* synced space map object */
99 - space_map_obj_t ms_smo_syncing; /* syncing space map object */
100 - space_map_t *ms_allocmap[TXG_SIZE]; /* allocated this txg */
101 - space_map_t *ms_freemap[TXG_SIZE]; /* freed this txg */
102 - space_map_t *ms_defermap[TXG_DEFER_SIZE]; /* deferred frees */
103 - space_map_t *ms_map; /* in-core free space map */
131 + kmutex_t ms_lock;
132 + kcondvar_t ms_load_cv;
133 + space_map_t *ms_sm;
134 + metaslab_ops_t *ms_ops;
135 + uint64_t ms_id;
136 + uint64_t ms_start;
137 + uint64_t ms_size;
138 +
139 + range_tree_t *ms_alloctree[TXG_SIZE];
140 + range_tree_t *ms_freetree[TXG_SIZE];
141 + range_tree_t *ms_defertree[TXG_DEFER_SIZE];
142 + range_tree_t *ms_tree;
143 +
144 + boolean_t ms_condensing; /* condensing? */
145 + boolean_t ms_loaded;
146 + boolean_t ms_loading;
147 +
104 148 int64_t ms_deferspace; /* sum of ms_defermap[] space */
105 149 uint64_t ms_weight; /* weight vs. others in group */
150 + uint64_t ms_factor;
151 + uint64_t ms_access_txg;
152 +
153 + /*
154 + * The metaslab block allocators can optionally use a size-ordered
155 + * range tree and/or an array of LBAs. Not all allocators use
156 + * this functionality. The ms_size_tree should always contain the
157 + * same number of segments as the ms_tree. The only difference
158 + * is that the ms_size_tree is ordered by segment sizes.
159 + */
160 + avl_tree_t ms_size_tree;
161 + uint64_t ms_lbas[MAX_LBAS];
162 +
106 163 metaslab_group_t *ms_group; /* metaslab group */
107 164 avl_node_t ms_group_node; /* node in metaslab group tree */
108 165 txg_node_t ms_txg_node; /* per-txg dirty metaslab links */
109 166 };
110 167
111 168 #ifdef __cplusplus
112 169 }
113 170 #endif
114 171
115 172 #endif /* _SYS_METASLAB_IMPL_H */
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX