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, Version 1.0 only 6 * (the "License"). You may not use this file except in compliance 7 * with the License. 8 * 9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 10 * or http://www.opensolaris.org/os/licensing. 11 * See the License for the specific language governing permissions 12 * and limitations under the License. 13 * 14 * When distributing Covered Code, include this CDDL HEADER in each 15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 16 * If applicable, add the following below this CDDL HEADER, with the 17 * fields enclosed by brackets "[]" replaced with your own identifying 18 * information: Portions Copyright [yyyy] [name of copyright owner] 19 * 20 * CDDL HEADER END 21 */ 22 /* 23 * Copyright 2005 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 27 #ifndef _SYS_DBUF_H 28 #define _SYS_DBUF_H 29 30 #pragma ident "%Z%%M% %I% %E% SMI" 31 32 #include <sys/dmu.h> 33 #include <sys/spa.h> 34 #include <sys/txg.h> 35 #include <sys/zio.h> 36 #include <sys/arc.h> 37 #include <sys/zfs_context.h> 38 #include <sys/refcount.h> 39 40 #ifdef __cplusplus 41 extern "C" { 42 #endif 43 44 #define DB_BONUS_BLKID (-1ULL) 45 #define IN_DMU_SYNC ((blkptr_t *)-1) 46 47 /* 48 * define flags for dbuf_read and friends 49 */ 50 51 #define DB_RF_MUST_SUCCEED 0 52 #define DB_RF_CANFAIL (1 << 1) 53 #define DB_RF_HAVESTRUCT (1 << 2) 54 #define DB_RF_NOPREFETCH (1 << 3) 55 56 /* 57 * The state transition diagram for dbufs looks like: 58 * 59 * +----> READ ----+ 60 * | | 61 * | V 62 * (alloc)-->UNCACHED CACHED-->(free) 63 * | ^ 64 * | | 65 * +----> FILL ----+ 66 */ 67 typedef enum dbuf_states { 68 DB_UNCACHED, 69 DB_FILL, 70 DB_READ, 71 DB_CACHED 72 } dbuf_states_t; 73 74 struct objset_impl; 75 struct dnode; 76 struct dmu_tx; 77 78 /* 79 * level = 0 means the user data 80 * level = 1 means the single indirect block 81 * etc. 82 */ 83 84 #define LIST_LINK_INACTIVE(link) \ 85 ((link)->list_next == NULL && (link)->list_prev == NULL) 86 87 typedef struct dmu_buf_impl { 88 /* 89 * The following members are immutable, with the exception of 90 * db.db_data, which is protected by db_mtx. 91 */ 92 93 /* the publicly visible structure */ 94 dmu_buf_t db; 95 96 /* the objset we belong to */ 97 struct objset_impl *db_objset; 98 99 /* 100 * the dnode we belong to (NULL when evicted) 101 */ 102 struct dnode *db_dnode; 103 104 /* 105 * our parent buffer; if the dnode points to us directly, 106 * db_parent == db_dnode->dn_dbuf 107 * only accessed by sync thread ??? 108 * (NULL when evicted) 109 */ 110 struct dmu_buf_impl *db_parent; 111 112 /* 113 * link for hash table of all dmu_buf_impl_t's 114 */ 115 struct dmu_buf_impl *db_hash_next; 116 117 /* our block number */ 118 uint64_t db_blkid; 119 120 /* 121 * Pointer to the blkptr_t which points to us. May be NULL if we 122 * don't have one yet. (NULL when evicted) 123 */ 124 blkptr_t *db_blkptr; 125 126 /* 127 * Our indirection level. Data buffers have db_level==0. 128 * Indirect buffers which point to data buffers have 129 * db_level==1. etc. Buffers which contain dnodes have 130 * db_level==0, since the dnodes are stored in a file. 131 */ 132 uint8_t db_level; 133 134 /* db_mtx protects the members below */ 135 kmutex_t db_mtx; 136 137 /* 138 * Current state of the buffer 139 */ 140 dbuf_states_t db_state; 141 142 /* 143 * Refcount accessed by dmu_buf_{hold,rele}. 144 * If nonzero, the buffer can't be destroyed. 145 * Protected by db_mtx. 146 */ 147 refcount_t db_holds; 148 149 /* buffer holding our data */ 150 arc_buf_t *db_buf; 151 152 kcondvar_t db_changed; 153 arc_buf_t *db_data_pending; 154 155 /* 156 * Last time (transaction group) this buffer was dirtied. 157 */ 158 uint64_t db_dirtied; 159 160 /* 161 * If dd_dnode != NULL, our link on the owner dnodes's dn_dbufs list. 162 * Protected by its dn_mtx. 163 */ 164 list_node_t db_link; 165 166 /* Our link on dn_dirty_dbufs[txg] */ 167 list_node_t db_dirty_node[TXG_SIZE]; 168 uint8_t db_dirtycnt; 169 170 /* 171 * Data which is unique to data (leaf) blocks: 172 */ 173 struct { 174 /* stuff we store for the user (see dmu_buf_set_user) */ 175 void *db_user_ptr; 176 void **db_user_data_ptr_ptr; 177 dmu_buf_evict_func_t *db_evict_func; 178 uint8_t db_immediate_evict; 179 uint8_t db_freed_in_flight; 180 181 /* 182 * db_data_old[txg&TXG_MASK] is set when we 183 * dirty the buffer, so that we can retain the 184 * pointer even if it gets COW'd in a subsequent 185 * transaction group. 186 * 187 * If the buffer is dirty in any txg, it can't 188 * be destroyed. 189 */ 190 /* 191 * XXX Protected by db_mtx and dn_dirty_mtx. 192 * db_mtx must be held to read db_dirty[], and 193 * both db_mtx and dn_dirty_mtx must be held to 194 * modify (dirty or clean). db_mtx must be held 195 * before dn_dirty_mtx. 196 */ 197 arc_buf_t *db_data_old[TXG_SIZE]; 198 blkptr_t *db_overridden_by[TXG_SIZE]; 199 } db_d; 200 } dmu_buf_impl_t; 201 202 /* Note: the dbuf hash table is exposed only for the mdb module */ 203 #define DBUF_MUTEXES 256 204 #define DBUF_HASH_MUTEX(h, idx) (&(h)->hash_mutexes[(idx) & (DBUF_MUTEXES-1)]) 205 typedef struct dbuf_hash_table { 206 uint64_t hash_table_mask; 207 dmu_buf_impl_t **hash_table; 208 kmutex_t hash_mutexes[DBUF_MUTEXES]; 209 } dbuf_hash_table_t; 210 211 212 uint64_t dbuf_whichblock(struct dnode *di, uint64_t offset); 213 214 dmu_buf_impl_t *dbuf_create_tlib(struct dnode *dn, char *data); 215 216 dmu_buf_impl_t *dbuf_hold(struct dnode *dn, uint64_t blkid); 217 dmu_buf_impl_t *dbuf_hold_level(struct dnode *dn, int level, uint64_t blkid, 218 void *tag); 219 dmu_buf_impl_t *dbuf_hold_bonus(struct dnode *dn, void *tag); 220 int dbuf_hold_impl(struct dnode *dn, uint8_t level, uint64_t blkid, int create, 221 void *tag, dmu_buf_impl_t **dbp); 222 223 void dbuf_prefetch(struct dnode *dn, uint64_t blkid); 224 225 void dbuf_add_ref(dmu_buf_impl_t *db, void *tag); 226 void dbuf_remove_ref(dmu_buf_impl_t *db, void *tag); 227 uint64_t dbuf_refcount(dmu_buf_impl_t *db); 228 229 void dbuf_rele(dmu_buf_impl_t *db); 230 231 dmu_buf_impl_t *dbuf_find(struct dnode *dn, uint8_t level, uint64_t blkid); 232 233 void dbuf_read(dmu_buf_impl_t *db); 234 int dbuf_read_canfail(dmu_buf_impl_t *db); 235 void dbuf_read_havestruct(dmu_buf_impl_t *db); 236 void dbuf_read_impl(dmu_buf_impl_t *db, zio_t *zio, uint32_t flags); 237 void dbuf_will_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx); 238 void dbuf_will_fill(dmu_buf_impl_t *db, dmu_tx_t *tx); 239 void dbuf_fill_done(dmu_buf_impl_t *db, dmu_tx_t *tx); 240 void dmu_buf_will_fill(dmu_buf_t *db, dmu_tx_t *tx); 241 void dmu_buf_fill_done(dmu_buf_t *db, dmu_tx_t *tx); 242 void dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx); 243 244 void dbuf_evict(dmu_buf_impl_t *db); 245 246 void dbuf_setdirty(dmu_buf_impl_t *db, dmu_tx_t *tx); 247 void dbuf_sync(dmu_buf_impl_t *db, zio_t *zio, dmu_tx_t *tx); 248 void dbuf_unoverride(dmu_buf_impl_t *db, uint64_t txg); 249 250 void dbuf_free_range(struct dnode *dn, uint64_t blkid, uint64_t nblks, 251 struct dmu_tx *); 252 253 void dbuf_downgrade(dmu_buf_impl_t *db, int evicting); 254 void dbuf_new_size(dmu_buf_impl_t *db, int size, dmu_tx_t *tx); 255 256 void dbuf_init(void); 257 void dbuf_fini(void); 258 259 #ifdef ZFS_DEBUG 260 261 /* 262 * There should be a ## between the string literal and fmt, to make it 263 * clear that we're joining two strings together, but that piece of shit 264 * gcc doesn't support that preprocessor token. 265 */ 266 #define dprintf_dbuf(dbuf, fmt, ...) do { \ 267 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 268 char __db_buf[32]; \ 269 uint64_t __db_obj = (dbuf)->db.db_object; \ 270 if (__db_obj == DMU_META_DNODE_OBJECT) \ 271 (void) strcpy(__db_buf, "mdn"); \ 272 else \ 273 (void) snprintf(__db_buf, sizeof (__db_buf), "%lld", \ 274 (u_longlong_t)__db_obj); \ 275 dprintf_ds((dbuf)->db_objset->os_dsl_dataset, \ 276 "obj=%s lvl=%u blkid=%lld " fmt, \ 277 __db_buf, (dbuf)->db_level, \ 278 (u_longlong_t)(dbuf)->db_blkid, __VA_ARGS__); \ 279 } \ 280 _NOTE(CONSTCOND) } while (0) 281 282 #define dprintf_dbuf_bp(db, bp, fmt, ...) do { \ 283 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 284 char __blkbuf[200]; \ 285 sprintf_blkptr(__blkbuf, bp); \ 286 dprintf_dbuf(db, fmt " %s\n", __VA_ARGS__, __blkbuf); \ 287 } \ 288 _NOTE(CONSTCOND) } while (0) 289 290 #else 291 292 #define dprintf_dbuf(db, fmt, ...) 293 #define dprintf_dbuf_bp(db, bp, fmt, ...) 294 295 #endif 296 297 298 #ifdef __cplusplus 299 } 300 #endif 301 302 #endif /* _SYS_DBUF_H */ 303