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 2010 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26#ifndef _SYS_DMU_IMPL_H 27#define _SYS_DMU_IMPL_H 28 29#include <sys/txg_impl.h> 30#include <sys/zio.h> 31#include <sys/dnode.h> 32#include <sys/zfs_context.h> 33 34#ifdef __cplusplus 35extern "C" { 36#endif 37 38/* 39 * This is the locking strategy for the DMU. Numbers in parenthesis are 40 * cases that use that lock order, referenced below: 41 * 42 * ARC is self-contained 43 * bplist is self-contained 44 * refcount is self-contained 45 * txg is self-contained (hopefully!) 46 * zst_lock 47 * zf_rwlock 48 * 49 * XXX try to improve evicting path? 50 * 51 * dp_config_rwlock > os_obj_lock > dn_struct_rwlock > 52 * dn_dbufs_mtx > hash_mutexes > db_mtx > dd_lock > leafs 53 * 54 * dp_config_rwlock 55 * must be held before: everything 56 * protects dd namespace changes 57 * protects property changes globally 58 * held from: 59 * dsl_dir_open/r: 60 * dsl_dir_create_sync/w: 61 * dsl_dir_sync_destroy/w: 62 * dsl_dir_rename_sync/w: 63 * dsl_prop_changed_notify/r: 64 * 65 * os_obj_lock 66 * must be held before: 67 * everything except dp_config_rwlock 68 * protects os_obj_next 69 * held from: 70 * dmu_object_alloc: dn_dbufs_mtx, db_mtx, hash_mutexes, dn_struct_rwlock 71 * 72 * dn_struct_rwlock 73 * must be held before: 74 * everything except dp_config_rwlock and os_obj_lock 75 * protects structure of dnode (eg. nlevels) 76 * db_blkptr can change when syncing out change to nlevels 77 * dn_maxblkid 78 * dn_nlevels 79 * dn_*blksz* 80 * phys nlevels, maxblkid, physical blkptr_t's (?) 81 * held from: 82 * callers of dbuf_read_impl, dbuf_hold[_impl], dbuf_prefetch 83 * dmu_object_info_from_dnode: dn_dirty_mtx (dn_datablksz) 84 * dmu_tx_count_free: 85 * dbuf_read_impl: db_mtx, dmu_zfetch() 86 * dmu_zfetch: zf_rwlock/r, zst_lock, dbuf_prefetch() 87 * dbuf_new_size: db_mtx 88 * dbuf_dirty: db_mtx 89 * dbuf_findbp: (callers, phys? - the real need) 90 * dbuf_create: dn_dbufs_mtx, hash_mutexes, db_mtx (phys?) 91 * dbuf_prefetch: dn_dirty_mtx, hash_mutexes, db_mtx, dn_dbufs_mtx 92 * dbuf_hold_impl: hash_mutexes, db_mtx, dn_dbufs_mtx, dbuf_findbp() 93 * dnode_sync/w (increase_indirection): db_mtx (phys) 94 * dnode_set_blksz/w: dn_dbufs_mtx (dn_*blksz*) 95 * dnode_new_blkid/w: (dn_maxblkid) 96 * dnode_free_range/w: dn_dirty_mtx (dn_maxblkid) 97 * dnode_next_offset: (phys) 98 * 99 * dn_dbufs_mtx 100 * must be held before: 101 * db_mtx, hash_mutexes 102 * protects: 103 * dn_dbufs 104 * dn_evicted 105 * held from: 106 * dmu_evict_user: db_mtx (dn_dbufs) 107 * dbuf_free_range: db_mtx (dn_dbufs) 108 * dbuf_remove_ref: db_mtx, callees: 109 * dbuf_hash_remove: hash_mutexes, db_mtx 110 * dbuf_create: hash_mutexes, db_mtx (dn_dbufs) 111 * dnode_set_blksz: (dn_dbufs) 112 * 113 * hash_mutexes (global) 114 * must be held before: 115 * db_mtx 116 * protects dbuf_hash_table (global) and db_hash_next 117 * held from: 118 * dbuf_find: db_mtx 119 * dbuf_hash_insert: db_mtx 120 * dbuf_hash_remove: db_mtx 121 * 122 * db_mtx (meta-leaf) 123 * must be held before: 124 * dn_mtx, dn_dirty_mtx, dd_lock (leaf mutexes) 125 * protects: 126 * db_state 127 * db_holds 128 * db_buf 129 * db_changed 130 * db_data_pending 131 * db_dirtied 132 * db_link 133 * db_dirty_node (??) 134 * db_dirtycnt 135 * db_d.* 136 * db.* 137 * held from: 138 * dbuf_dirty: dn_mtx, dn_dirty_mtx 139 * dbuf_dirty->dsl_dir_willuse_space: dd_lock 140 * dbuf_dirty->dbuf_new_block->dsl_dataset_block_freeable: dd_lock 141 * dbuf_undirty: dn_dirty_mtx (db_d) 142 * dbuf_write_done: dn_dirty_mtx (db_state) 143 * dbuf_* 144 * dmu_buf_update_user: none (db_d) 145 * dmu_evict_user: none (db_d) (maybe can eliminate) 146 * dbuf_find: none (db_holds) 147 * dbuf_hash_insert: none (db_holds) 148 * dmu_buf_read_array_impl: none (db_state, db_changed) 149 * dmu_sync: none (db_dirty_node, db_d) 150 * dnode_reallocate: none (db) 151 * 152 * dn_mtx (leaf) 153 * protects: 154 * dn_dirty_dbufs 155 * dn_ranges 156 * phys accounting 157 * dn_allocated_txg 158 * dn_free_txg 159 * dn_assigned_txg 160 * dd_assigned_tx 161 * dn_notxholds 162 * dn_dirtyctx 163 * dn_dirtyctx_firstset 164 * (dn_phys copy fields?) 165 * (dn_phys contents?) 166 * held from: 167 * dnode_* 168 * dbuf_dirty: none 169 * dbuf_sync: none (phys accounting) 170 * dbuf_undirty: none (dn_ranges, dn_dirty_dbufs) 171 * dbuf_write_done: none (phys accounting) 172 * dmu_object_info_from_dnode: none (accounting) 173 * dmu_tx_commit: none 174 * dmu_tx_hold_object_impl: none 175 * dmu_tx_try_assign: dn_notxholds(cv) 176 * dmu_tx_unassign: none 177 * 178 * dd_lock 179 * must be held before: 180 * ds_lock 181 * ancestors' dd_lock 182 * protects: 183 * dd_prop_cbs 184 * dd_sync_* 185 * dd_used_bytes 186 * dd_tempreserved 187 * dd_space_towrite 188 * dd_myname 189 * dd_phys accounting? 190 * held from: 191 * dsl_dir_* 192 * dsl_prop_changed_notify: none (dd_prop_cbs) 193 * dsl_prop_register: none (dd_prop_cbs) 194 * dsl_prop_unregister: none (dd_prop_cbs) 195 * dsl_dataset_block_freeable: none (dd_sync_*) 196 * 197 * os_lock (leaf) 198 * protects: 199 * os_dirty_dnodes 200 * os_free_dnodes 201 * os_dnodes 202 * os_downgraded_dbufs 203 * dn_dirtyblksz 204 * dn_dirty_link 205 * held from: 206 * dnode_create: none (os_dnodes) 207 * dnode_destroy: none (os_dnodes) 208 * dnode_setdirty: none (dn_dirtyblksz, os_*_dnodes) 209 * dnode_free: none (dn_dirtyblksz, os_*_dnodes) 210 * 211 * ds_lock 212 * protects: 213 * ds_objset 214 * ds_open_refcount 215 * ds_snapname 216 * ds_phys accounting 217 * ds_phys userrefs zapobj 218 * ds_reserved 219 * held from: 220 * dsl_dataset_* 221 * 222 * dr_mtx (leaf) 223 * protects: 224 * dr_children 225 * held from: 226 * dbuf_dirty 227 * dbuf_undirty 228 * dbuf_sync_indirect 229 * dnode_new_blkid 230 */ 231 232struct objset; 233struct dmu_pool; 234 235typedef struct dmu_xuio { 236 int next; 237 int cnt; 238 struct arc_buf **bufs; 239 iovec_t *iovp; 240} dmu_xuio_t; 241 242typedef struct xuio_stats { 243 /* loaned yet not returned arc_buf */ 244 kstat_named_t xuiostat_onloan_rbuf; 245 kstat_named_t xuiostat_onloan_wbuf; 246 /* whether a copy is made when loaning out a read buffer */ 247 kstat_named_t xuiostat_rbuf_copied; 248 kstat_named_t xuiostat_rbuf_nocopy; 249 /* whether a copy is made when assigning a write buffer */ 250 kstat_named_t xuiostat_wbuf_copied; 251 kstat_named_t xuiostat_wbuf_nocopy; 252} xuio_stats_t; 253 254static xuio_stats_t xuio_stats = { 255 { "onloan_read_buf", KSTAT_DATA_UINT64 }, 256 { "onloan_write_buf", KSTAT_DATA_UINT64 }, 257 { "read_buf_copied", KSTAT_DATA_UINT64 }, 258 { "read_buf_nocopy", KSTAT_DATA_UINT64 }, 259 { "write_buf_copied", KSTAT_DATA_UINT64 }, 260 { "write_buf_nocopy", KSTAT_DATA_UINT64 } 261}; 262 263#define XUIOSTAT_INCR(stat, val) \ 264 atomic_add_64(&xuio_stats.stat.value.ui64, (val)) 265#define XUIOSTAT_BUMP(stat) XUIOSTAT_INCR(stat, 1) 266 267 268#ifdef __cplusplus 269} 270#endif 271 272#endif /* _SYS_DMU_IMPL_H */ 273