Print this page
3742 zfs comments need cleaner, more consistent style
Submitted by: Will Andrews <willa@spectralogic.com>
Submitted by: Alan Somers <alans@spectralogic.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Eric Schrock <eric.schrock@delphix.com>
| Split |
Close |
| Expand all |
| Collapse all |
--- old/usr/src/uts/common/fs/zfs/bptree.c
+++ new/usr/src/uts/common/fs/zfs/bptree.c
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 /*
23 23 * Copyright (c) 2012 by Delphix. All rights reserved.
24 24 */
25 25
26 26 #include <sys/arc.h>
27 27 #include <sys/bptree.h>
28 28 #include <sys/dmu.h>
29 29 #include <sys/dmu_objset.h>
30 30 #include <sys/dmu_tx.h>
31 31 #include <sys/dmu_traverse.h>
32 32 #include <sys/dsl_dataset.h>
33 33 #include <sys/dsl_dir.h>
34 34 #include <sys/dsl_pool.h>
35 35 #include <sys/dnode.h>
|
↓ open down ↓ |
35 lines elided |
↑ open up ↑ |
36 36 #include <sys/refcount.h>
37 37 #include <sys/spa.h>
38 38
39 39 /*
40 40 * A bptree is a queue of root block pointers from destroyed datasets. When a
41 41 * dataset is destroyed its root block pointer is put on the end of the pool's
42 42 * bptree queue so the dataset's blocks can be freed asynchronously by
43 43 * dsl_scan_sync. This allows the delete operation to finish without traversing
44 44 * all the dataset's blocks.
45 45 *
46 - * Note that while bt_begin and bt_end are only ever incremented in this code
46 + * Note that while bt_begin and bt_end are only ever incremented in this code,
47 47 * they are effectively reset to 0 every time the entire bptree is freed because
48 48 * the bptree's object is destroyed and re-created.
49 49 */
50 50
51 51 struct bptree_args {
52 52 bptree_phys_t *ba_phys; /* data in bonus buffer, dirtied if freeing */
53 53 boolean_t ba_free; /* true if freeing during traversal */
54 54
55 55 bptree_itor_t *ba_func; /* function to call for each blockpointer */
56 56 void *ba_arg; /* caller supplied argument to ba_func */
57 57 dmu_tx_t *ba_tx; /* caller supplied tx, NULL if not freeing */
58 58 } bptree_args_t;
59 59
60 60 uint64_t
61 61 bptree_alloc(objset_t *os, dmu_tx_t *tx)
62 62 {
63 63 uint64_t obj;
64 64 dmu_buf_t *db;
65 65 bptree_phys_t *bt;
66 66
67 67 obj = dmu_object_alloc(os, DMU_OTN_UINT64_METADATA,
68 68 SPA_MAXBLOCKSIZE, DMU_OTN_UINT64_METADATA,
69 69 sizeof (bptree_phys_t), tx);
70 70
71 71 /*
72 72 * Bonus buffer contents are already initialized to 0, but for
73 73 * readability we make it explicit.
74 74 */
75 75 VERIFY3U(0, ==, dmu_bonus_hold(os, obj, FTAG, &db));
76 76 dmu_buf_will_dirty(db, tx);
77 77 bt = db->db_data;
78 78 bt->bt_begin = 0;
79 79 bt->bt_end = 0;
80 80 bt->bt_bytes = 0;
81 81 bt->bt_comp = 0;
82 82 bt->bt_uncomp = 0;
83 83 dmu_buf_rele(db, FTAG);
84 84
85 85 return (obj);
86 86 }
87 87
88 88 int
89 89 bptree_free(objset_t *os, uint64_t obj, dmu_tx_t *tx)
90 90 {
91 91 dmu_buf_t *db;
92 92 bptree_phys_t *bt;
93 93
94 94 VERIFY3U(0, ==, dmu_bonus_hold(os, obj, FTAG, &db));
95 95 bt = db->db_data;
96 96 ASSERT3U(bt->bt_begin, ==, bt->bt_end);
97 97 ASSERT0(bt->bt_bytes);
98 98 ASSERT0(bt->bt_comp);
99 99 ASSERT0(bt->bt_uncomp);
100 100 dmu_buf_rele(db, FTAG);
101 101
102 102 return (dmu_object_free(os, obj, tx));
103 103 }
104 104
105 105 void
106 106 bptree_add(objset_t *os, uint64_t obj, blkptr_t *bp, uint64_t birth_txg,
107 107 uint64_t bytes, uint64_t comp, uint64_t uncomp, dmu_tx_t *tx)
108 108 {
109 109 dmu_buf_t *db;
110 110 bptree_phys_t *bt;
111 111 bptree_entry_phys_t bte;
112 112
113 113 /*
114 114 * bptree objects are in the pool mos, therefore they can only be
115 115 * modified in syncing context. Furthermore, this is only modified
116 116 * by the sync thread, so no locking is necessary.
117 117 */
118 118 ASSERT(dmu_tx_is_syncing(tx));
119 119
120 120 VERIFY3U(0, ==, dmu_bonus_hold(os, obj, FTAG, &db));
121 121 bt = db->db_data;
122 122
123 123 bte.be_birth_txg = birth_txg;
124 124 bte.be_bp = *bp;
125 125 bzero(&bte.be_zb, sizeof (bte.be_zb));
126 126 dmu_write(os, obj, bt->bt_end * sizeof (bte), sizeof (bte), &bte, tx);
127 127
128 128 dmu_buf_will_dirty(db, tx);
129 129 bt->bt_end++;
130 130 bt->bt_bytes += bytes;
131 131 bt->bt_comp += comp;
132 132 bt->bt_uncomp += uncomp;
133 133 dmu_buf_rele(db, FTAG);
134 134 }
135 135
136 136 /* ARGSUSED */
137 137 static int
138 138 bptree_visit_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
139 139 const zbookmark_t *zb, const dnode_phys_t *dnp, void *arg)
140 140 {
141 141 int err;
142 142 struct bptree_args *ba = arg;
143 143
144 144 if (bp == NULL)
145 145 return (0);
146 146
147 147 err = ba->ba_func(ba->ba_arg, bp, ba->ba_tx);
148 148 if (err == 0 && ba->ba_free) {
149 149 ba->ba_phys->bt_bytes -= bp_get_dsize_sync(spa, bp);
150 150 ba->ba_phys->bt_comp -= BP_GET_PSIZE(bp);
151 151 ba->ba_phys->bt_uncomp -= BP_GET_UCSIZE(bp);
152 152 }
153 153 return (err);
154 154 }
155 155
156 156 int
157 157 bptree_iterate(objset_t *os, uint64_t obj, boolean_t free, bptree_itor_t func,
158 158 void *arg, dmu_tx_t *tx)
159 159 {
160 160 int err;
161 161 uint64_t i;
162 162 dmu_buf_t *db;
163 163 struct bptree_args ba;
164 164
165 165 ASSERT(!free || dmu_tx_is_syncing(tx));
166 166
167 167 err = dmu_bonus_hold(os, obj, FTAG, &db);
168 168 if (err != 0)
169 169 return (err);
170 170
171 171 if (free)
172 172 dmu_buf_will_dirty(db, tx);
173 173
174 174 ba.ba_phys = db->db_data;
175 175 ba.ba_free = free;
176 176 ba.ba_func = func;
177 177 ba.ba_arg = arg;
178 178 ba.ba_tx = tx;
179 179
180 180 err = 0;
181 181 for (i = ba.ba_phys->bt_begin; i < ba.ba_phys->bt_end; i++) {
182 182 bptree_entry_phys_t bte;
183 183
184 184 ASSERT(!free || i == ba.ba_phys->bt_begin);
185 185
186 186 err = dmu_read(os, obj, i * sizeof (bte), sizeof (bte),
187 187 &bte, DMU_READ_NO_PREFETCH);
188 188 if (err != 0)
189 189 break;
190 190
191 191 err = traverse_dataset_destroyed(os->os_spa, &bte.be_bp,
192 192 bte.be_birth_txg, &bte.be_zb,
193 193 TRAVERSE_PREFETCH_METADATA | TRAVERSE_POST,
194 194 bptree_visit_cb, &ba);
195 195 if (free) {
196 196 ASSERT(err == 0 || err == ERESTART);
197 197 if (err != 0) {
198 198 /* save bookmark for future resume */
199 199 ASSERT3U(bte.be_zb.zb_objset, ==,
200 200 ZB_DESTROYED_OBJSET);
201 201 ASSERT0(bte.be_zb.zb_level);
202 202 dmu_write(os, obj, i * sizeof (bte),
203 203 sizeof (bte), &bte, tx);
204 204 break;
205 205 } else {
206 206 ba.ba_phys->bt_begin++;
207 207 (void) dmu_free_range(os, obj,
208 208 i * sizeof (bte), sizeof (bte), tx);
209 209 }
210 210 }
211 211 }
212 212
213 213 ASSERT(!free || err != 0 || ba.ba_phys->bt_begin == ba.ba_phys->bt_end);
214 214
215 215 /* if all blocks are free there should be no used space */
216 216 if (ba.ba_phys->bt_begin == ba.ba_phys->bt_end) {
217 217 ASSERT0(ba.ba_phys->bt_bytes);
218 218 ASSERT0(ba.ba_phys->bt_comp);
219 219 ASSERT0(ba.ba_phys->bt_uncomp);
220 220 }
221 221
222 222 dmu_buf_rele(db, FTAG);
223 223
224 224 return (err);
225 225 }
|
↓ open down ↓ |
169 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX