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 2006 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 #ifndef _SYS_DBUF_H 27 #define _SYS_DBUF_H 28 29 #pragma ident "%Z%%M% %I% %E% SMI" 30 31 #include <sys/dmu.h> 32 #include <sys/spa.h> 33 #include <sys/txg.h> 34 #include <sys/zio.h> 35 #include <sys/arc.h> 36 #include <sys/zfs_context.h> 37 #include <sys/refcount.h> 38 39 #ifdef __cplusplus 40 extern "C" { 41 #endif 42 43 #define DB_BONUS_BLKID (-1ULL) 44 #define IN_DMU_SYNC ((blkptr_t *)-1) 45 46 /* 47 * define flags for dbuf_read 48 */ 49 50 #define DB_RF_MUST_SUCCEED 0 51 #define DB_RF_CANFAIL (1 << 1) 52 #define DB_RF_HAVESTRUCT (1 << 2) 53 #define DB_RF_NOPREFETCH (1 << 3) 54 #define DB_RF_NEVERWAIT (1 << 4) 55 56 /* 57 * The state transition diagram for dbufs looks like: 58 * 59 * +----> READ ----+ 60 * | | 61 * | V 62 * (alloc)-->UNCACHED CACHED-->EVICTING-->(free) 63 * | ^ 64 * | | 65 * +----> FILL ----+ 66 */ 67 typedef enum dbuf_states { 68 DB_UNCACHED, 69 DB_FILL, 70 DB_READ, 71 DB_CACHED, 72 DB_EVICTING 73 } dbuf_states_t; 74 75 struct objset_impl; 76 struct dnode; 77 struct dmu_tx; 78 79 /* 80 * level = 0 means the user data 81 * level = 1 means the single indirect block 82 * etc. 83 */ 84 85 #define LIST_LINK_INACTIVE(link) \ 86 ((link)->list_next == NULL && (link)->list_prev == NULL) 87 88 typedef struct dmu_buf_impl { 89 /* 90 * The following members are immutable, with the exception of 91 * db.db_data, which is protected by db_mtx. 92 */ 93 94 /* the publicly visible structure */ 95 dmu_buf_t db; 96 97 /* the objset we belong to */ 98 struct objset_impl *db_objset; 99 100 /* 101 * the dnode we belong to (NULL when evicted) 102 */ 103 struct dnode *db_dnode; 104 105 /* 106 * our parent buffer; if the dnode points to us directly, 107 * db_parent == db_dnode->dn_dbuf 108 * only accessed by sync thread ??? 109 * (NULL when evicted) 110 */ 111 struct dmu_buf_impl *db_parent; 112 113 /* 114 * link for hash table of all dmu_buf_impl_t's 115 */ 116 struct dmu_buf_impl *db_hash_next; 117 118 /* our block number */ 119 uint64_t db_blkid; 120 121 /* 122 * Pointer to the blkptr_t which points to us. May be NULL if we 123 * don't have one yet. (NULL when evicted) 124 */ 125 blkptr_t *db_blkptr; 126 127 /* 128 * Our indirection level. Data buffers have db_level==0. 129 * Indirect buffers which point to data buffers have 130 * db_level==1. etc. Buffers which contain dnodes have 131 * db_level==0, since the dnodes are stored in a file. 132 */ 133 uint8_t db_level; 134 135 /* db_mtx protects the members below */ 136 kmutex_t db_mtx; 137 138 /* 139 * Current state of the buffer 140 */ 141 dbuf_states_t db_state; 142 143 /* 144 * Refcount accessed by dmu_buf_{hold,rele}. 145 * If nonzero, the buffer can't be destroyed. 146 * Protected by db_mtx. 147 */ 148 refcount_t db_holds; 149 150 /* buffer holding our data */ 151 arc_buf_t *db_buf; 152 153 kcondvar_t db_changed; 154 arc_buf_t *db_data_pending; 155 156 /* 157 * Last time (transaction group) this buffer was dirtied. 158 */ 159 uint64_t db_dirtied; 160 161 /* 162 * If db_dnode != NULL, our link on the owner dnodes's dn_dbufs list. 163 * Protected by its dn_dbufs_mtx. 164 */ 165 list_node_t db_link; 166 167 /* Our link on dn_dirty_dbufs[txg] */ 168 list_node_t db_dirty_node[TXG_SIZE]; 169 uint8_t db_dirtycnt; 170 171 /* 172 * Data which is unique to data (leaf) blocks: 173 */ 174 struct { 175 /* stuff we store for the user (see dmu_buf_set_user) */ 176 void *db_user_ptr; 177 void **db_user_data_ptr_ptr; 178 dmu_buf_evict_func_t *db_evict_func; 179 uint8_t db_immediate_evict; 180 uint8_t db_freed_in_flight; 181 182 /* 183 * db_data_old[txg&TXG_MASK] is set when we 184 * dirty the buffer, so that we can retain the 185 * pointer even if it gets COW'd in a subsequent 186 * transaction group. 187 * 188 * If the buffer is dirty in any txg, it can't 189 * be destroyed. 190 */ 191 /* 192 * XXX Protected by db_mtx and dn_dirty_mtx. 193 * db_mtx must be held to read db_dirty[], and 194 * both db_mtx and dn_dirty_mtx must be held to 195 * modify (dirty or clean). db_mtx must be held 196 * before dn_dirty_mtx. 197 */ 198 void *db_data_old[TXG_SIZE]; 199 blkptr_t *db_overridden_by[TXG_SIZE]; 200 } db_d; 201 } dmu_buf_impl_t; 202 203 /* Note: the dbuf hash table is exposed only for the mdb module */ 204 #define DBUF_MUTEXES 256 205 #define DBUF_HASH_MUTEX(h, idx) (&(h)->hash_mutexes[(idx) & (DBUF_MUTEXES-1)]) 206 typedef struct dbuf_hash_table { 207 uint64_t hash_table_mask; 208 dmu_buf_impl_t **hash_table; 209 kmutex_t hash_mutexes[DBUF_MUTEXES]; 210 } dbuf_hash_table_t; 211 212 213 uint64_t dbuf_whichblock(struct dnode *di, uint64_t offset); 214 215 dmu_buf_impl_t *dbuf_create_tlib(struct dnode *dn, char *data); 216 dmu_buf_impl_t *dbuf_create_bonus(struct dnode *dn); 217 218 dmu_buf_impl_t *dbuf_hold(struct dnode *dn, uint64_t blkid, void *tag); 219 dmu_buf_impl_t *dbuf_hold_level(struct dnode *dn, int level, uint64_t blkid, 220 void *tag); 221 int dbuf_hold_impl(struct dnode *dn, uint8_t level, uint64_t blkid, int create, 222 void *tag, dmu_buf_impl_t **dbp); 223 224 void dbuf_prefetch(struct dnode *dn, uint64_t blkid); 225 226 void dbuf_add_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, void *tag); 230 231 dmu_buf_impl_t *dbuf_find(struct dnode *dn, uint8_t level, uint64_t blkid); 232 233 int dbuf_read(dmu_buf_impl_t *db, zio_t *zio, uint32_t flags); 234 void dbuf_will_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx); 235 void dmu_buf_will_fill(dmu_buf_t *db, dmu_tx_t *tx); 236 void dbuf_fill_done(dmu_buf_impl_t *db, dmu_tx_t *tx); 237 void dmu_buf_will_fill(dmu_buf_t *db, dmu_tx_t *tx); 238 void dmu_buf_fill_done(dmu_buf_t *db, dmu_tx_t *tx); 239 void dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx); 240 241 void dbuf_clear(dmu_buf_impl_t *db); 242 void dbuf_evict(dmu_buf_impl_t *db); 243 244 void dbuf_setdirty(dmu_buf_impl_t *db, dmu_tx_t *tx); 245 void dbuf_sync(dmu_buf_impl_t *db, zio_t *zio, dmu_tx_t *tx); 246 void dbuf_unoverride(dmu_buf_impl_t *db, uint64_t txg); 247 248 void dbuf_free_range(struct dnode *dn, uint64_t blkid, uint64_t nblks, 249 struct dmu_tx *); 250 251 void dbuf_new_size(dmu_buf_impl_t *db, int size, dmu_tx_t *tx); 252 253 void dbuf_init(void); 254 void dbuf_fini(void); 255 256 #ifdef ZFS_DEBUG 257 258 /* 259 * There should be a ## between the string literal and fmt, to make it 260 * clear that we're joining two strings together, but gcc does not 261 * support that preprocessor token. 262 */ 263 #define dprintf_dbuf(dbuf, fmt, ...) do { \ 264 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 265 char __db_buf[32]; \ 266 uint64_t __db_obj = (dbuf)->db.db_object; \ 267 if (__db_obj == DMU_META_DNODE_OBJECT) \ 268 (void) strcpy(__db_buf, "mdn"); \ 269 else \ 270 (void) snprintf(__db_buf, sizeof (__db_buf), "%lld", \ 271 (u_longlong_t)__db_obj); \ 272 dprintf_ds((dbuf)->db_objset->os_dsl_dataset, \ 273 "obj=%s lvl=%u blkid=%lld " fmt, \ 274 __db_buf, (dbuf)->db_level, \ 275 (u_longlong_t)(dbuf)->db_blkid, __VA_ARGS__); \ 276 } \ 277 _NOTE(CONSTCOND) } while (0) 278 279 #define dprintf_dbuf_bp(db, bp, fmt, ...) do { \ 280 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 281 char __blkbuf[BP_SPRINTF_LEN]; \ 282 sprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, bp); \ 283 dprintf_dbuf(db, fmt " %s\n", __VA_ARGS__, __blkbuf); \ 284 } \ 285 _NOTE(CONSTCOND) } while (0) 286 287 #define DBUF_VERIFY(db) dbuf_verify(db) 288 289 #else 290 291 #define dprintf_dbuf(db, fmt, ...) 292 #define dprintf_dbuf_bp(db, bp, fmt, ...) 293 #define DBUF_VERIFY(db) 294 295 #endif 296 297 298 #ifdef __cplusplus 299 } 300 #endif 301 302 #endif /* _SYS_DBUF_H */ 303