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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  23  * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  24  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
  25  * Copyright 2014 Nexenta Systems, Inc. All rights reserved.
  26  */
  27 
  28 #ifndef _SYS_DBUF_H
  29 #define _SYS_DBUF_H
  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 #include <sys/zrlock.h>
  39 
  40 #ifdef  __cplusplus
  41 extern "C" {
  42 #endif
  43 
  44 #define IN_DMU_SYNC 2
  45 
  46 /*
  47  * define flags for dbuf_read
  48  */
  49 
  50 #define DB_RF_MUST_SUCCEED      (1 << 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 #define DB_RF_CACHED            (1 << 5)
  56 
  57 #define DBUF_EVICT_ALL          -1
  58 
  59 /*
  60  * The simplified state transition diagram for dbufs looks like:
  61  *
  62  *              +----> READ ----+
  63  *              |               |
  64  *              |               V
  65  *  (alloc)-->UNCACHED            CACHED-->EVICTING-->(free)
  66  *              |               ^        ^
  67  *              |               |        |
  68  *              +----> FILL ----+     |
  69  *              |                        |
  70  *              |                        |
  71  *              +--------> NOFILL -------+
  72  *
  73  * DB_SEARCH is an invalid state for a dbuf. It is used by dbuf_free_range
  74  * to find all dbufs in a range of a dnode and must be less than any other
  75  * dbuf_states_t (see comment on dn_dbufs in dnode.h).
  76  */
  77 typedef enum dbuf_states {
  78         DB_SEARCH = -1,
  79         DB_UNCACHED,
  80         DB_FILL,
  81         DB_NOFILL,
  82         DB_READ,
  83         DB_CACHED,
  84         DB_EVICTING
  85 } dbuf_states_t;
  86 
  87 struct dnode;
  88 struct dmu_tx;
  89 
  90 /*
  91  * level = 0 means the user data
  92  * level = 1 means the single indirect block
  93  * etc.
  94  */
  95 
  96 struct dmu_buf_impl;
  97 
  98 typedef enum override_states {
  99         DR_NOT_OVERRIDDEN,
 100         DR_IN_DMU_SYNC,
 101         DR_OVERRIDDEN
 102 } override_states_t;
 103 
 104 typedef struct dbuf_dirty_record {
 105         /* link on our parents dirty list */
 106         list_node_t dr_dirty_node;
 107 
 108         /* transaction group this data will sync in */
 109         uint64_t dr_txg;
 110 
 111         /* zio of outstanding write IO */
 112         zio_t *dr_zio;
 113 
 114         /* pointer back to our dbuf */
 115         struct dmu_buf_impl *dr_dbuf;
 116 
 117         /* pointer to next dirty record */
 118         struct dbuf_dirty_record *dr_next;
 119 
 120         /* pointer to parent dirty record */
 121         struct dbuf_dirty_record *dr_parent;
 122 
 123         /* How much space was changed to dsl_pool_dirty_space() for this? */
 124         unsigned int dr_accounted;
 125 
 126         /* use special class of dirty entry */
 127         boolean_t dr_usesc;
 128 
 129         union dirty_types {
 130                 struct dirty_indirect {
 131 
 132                         /* protect access to list */
 133                         kmutex_t dr_mtx;
 134 
 135                         /* Our list of dirty children */
 136                         list_t dr_children;
 137                 } di;
 138                 struct dirty_leaf {
 139 
 140                         /*
 141                          * dr_data is set when we dirty the buffer
 142                          * so that we can retain the pointer even if it
 143                          * gets COW'd in a subsequent transaction group.
 144                          */
 145                         arc_buf_t *dr_data;
 146                         blkptr_t dr_overridden_by;
 147                         override_states_t dr_override_state;
 148                         uint8_t dr_copies;
 149                         boolean_t dr_nopwrite;
 150                 } dl;
 151         } dt;
 152 } dbuf_dirty_record_t;
 153 
 154 typedef struct dmu_buf_impl {
 155         /*
 156          * The following members are immutable, with the exception of
 157          * db.db_data, which is protected by db_mtx.
 158          */
 159 
 160         /* the publicly visible structure */
 161         dmu_buf_t db;
 162 
 163         /* the objset we belong to */
 164         struct objset *db_objset;
 165 
 166         /*
 167          * handle to safely access the dnode we belong to (NULL when evicted)
 168          */
 169         struct dnode_handle *db_dnode_handle;
 170 
 171         /*
 172          * our parent buffer; if the dnode points to us directly,
 173          * db_parent == db_dnode_handle->dnh_dnode->dn_dbuf
 174          * only accessed by sync thread ???
 175          * (NULL when evicted)
 176          * May change from NULL to non-NULL under the protection of db_mtx
 177          * (see dbuf_check_blkptr())
 178          */
 179         struct dmu_buf_impl *db_parent;
 180 
 181         /*
 182          * link for hash table of all dmu_buf_impl_t's
 183          */
 184         struct dmu_buf_impl *db_hash_next;
 185 
 186         /* our block number */
 187         uint64_t db_blkid;
 188 
 189         /*
 190          * Pointer to the blkptr_t which points to us. May be NULL if we
 191          * don't have one yet. (NULL when evicted)
 192          */
 193         blkptr_t *db_blkptr;
 194 
 195         /*
 196          * Our indirection level.  Data buffers have db_level==0.
 197          * Indirect buffers which point to data buffers have
 198          * db_level==1. etc.  Buffers which contain dnodes have
 199          * db_level==0, since the dnodes are stored in a file.
 200          */
 201         uint8_t db_level;
 202 
 203         /* db_mtx protects the members below */
 204         kmutex_t db_mtx;
 205 
 206         /*
 207          * Current state of the buffer
 208          */
 209         dbuf_states_t db_state;
 210 
 211         /*
 212          * Refcount accessed by dmu_buf_{hold,rele}.
 213          * If nonzero, the buffer can't be destroyed.
 214          * Protected by db_mtx.
 215          */
 216         refcount_t db_holds;
 217 
 218         /* buffer holding our data */
 219         arc_buf_t *db_buf;
 220 
 221         kcondvar_t db_changed;
 222         dbuf_dirty_record_t *db_data_pending;
 223 
 224         /* pointer to most recent dirty record for this buffer */
 225         dbuf_dirty_record_t *db_last_dirty;
 226 
 227         /*
 228          * Our link on the owner dnodes's dn_dbufs list.
 229          * Protected by its dn_dbufs_mtx.
 230          */
 231         avl_node_t db_link;
 232 
 233         /* Data which is unique to data (leaf) blocks: */
 234 
 235         /* stuff we store for the user (see dmu_buf_set_user) */
 236         void *db_user_ptr;
 237         void **db_user_data_ptr_ptr;
 238         dmu_buf_evict_func_t *db_evict_func;
 239 
 240         uint8_t db_immediate_evict;
 241         uint8_t db_freed_in_flight;
 242 
 243         uint8_t db_dirtycnt;
 244 } dmu_buf_impl_t;
 245 
 246 /* Note: the dbuf hash table is exposed only for the mdb module */
 247 #define DBUF_MUTEXES    256
 248 #define DBUF_LOCK_PAD   64
 249 typedef struct {
 250         kmutex_t mtx;
 251 #ifdef _KERNEL
 252         unsigned char pad[(DBUF_LOCK_PAD - sizeof (kmutex_t))];
 253 #endif
 254 } dbuf_mutex_t;
 255 #define DBUF_HASH_MUTEX(h, idx) \
 256         (&((h)->hash_mutexes[(idx) & (DBUF_MUTEXES-1)].mtx))
 257 typedef struct dbuf_hash_table {
 258         uint64_t hash_table_mask;
 259         dmu_buf_impl_t **hash_table;
 260         dbuf_mutex_t hash_mutexes[DBUF_MUTEXES];
 261 } dbuf_hash_table_t;
 262 
 263 
 264 uint64_t dbuf_whichblock(struct dnode *di, uint64_t offset);
 265 
 266 dmu_buf_impl_t *dbuf_create_tlib(struct dnode *dn, char *data);
 267 void dbuf_create_bonus(struct dnode *dn);
 268 int dbuf_spill_set_blksz(dmu_buf_t *db, uint64_t blksz, dmu_tx_t *tx);
 269 void dbuf_spill_hold(struct dnode *dn, dmu_buf_impl_t **dbp, void *tag);
 270 
 271 void dbuf_rm_spill(struct dnode *dn, dmu_tx_t *tx);
 272 
 273 dmu_buf_impl_t *dbuf_hold(struct dnode *dn, uint64_t blkid, void *tag);
 274 dmu_buf_impl_t *dbuf_hold_level(struct dnode *dn, int level, uint64_t blkid,
 275     void *tag);
 276 int dbuf_hold_impl(struct dnode *dn, uint8_t level, uint64_t blkid, int create,
 277     void *tag, dmu_buf_impl_t **dbp);
 278 
 279 void dbuf_prefetch(struct dnode *dn, uint64_t blkid, zio_priority_t prio);
 280 
 281 void dbuf_add_ref(dmu_buf_impl_t *db, void *tag);
 282 uint64_t dbuf_refcount(dmu_buf_impl_t *db);
 283 
 284 void dbuf_rele(dmu_buf_impl_t *db, void *tag);
 285 void dbuf_rele_and_unlock(dmu_buf_impl_t *db, void *tag);
 286 
 287 dmu_buf_impl_t *dbuf_find(struct dnode *dn, uint8_t level, uint64_t blkid);
 288 
 289 int dbuf_read(dmu_buf_impl_t *db, zio_t *zio, uint32_t flags);
 290 void dmu_buf_will_not_fill(dmu_buf_t *db, dmu_tx_t *tx);
 291 void dmu_buf_will_fill(dmu_buf_t *db, dmu_tx_t *tx);
 292 void dmu_buf_fill_done(dmu_buf_t *db, dmu_tx_t *tx);
 293 void dbuf_assign_arcbuf(dmu_buf_impl_t *db, arc_buf_t *buf, dmu_tx_t *tx);
 294 dbuf_dirty_record_t *dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx);
 295 dbuf_dirty_record_t *dbuf_dirty_sc(dmu_buf_impl_t *db, dmu_tx_t *tx,
 296     boolean_t usesc);
 297 arc_buf_t *dbuf_loan_arcbuf(dmu_buf_impl_t *db);
 298 void dmu_buf_write_embedded(dmu_buf_t *dbuf, void *data,
 299     bp_embedded_type_t etype, enum zio_compress comp,
 300     int uncompressed_size, int compressed_size, int byteorder, dmu_tx_t *tx);
 301 
 302 void dbuf_clear(dmu_buf_impl_t *db);
 303 void dbuf_evict(dmu_buf_impl_t *db);
 304 
 305 void dbuf_setdirty(dmu_buf_impl_t *db, dmu_tx_t *tx);
 306 void dbuf_unoverride(dbuf_dirty_record_t *dr);
 307 void dbuf_sync_list(list_t *list, dmu_tx_t *tx);
 308 void dbuf_release_bp(dmu_buf_impl_t *db);
 309 
 310 void dbuf_free_range(struct dnode *dn, uint64_t start, uint64_t end,
 311     struct dmu_tx *);
 312 
 313 void dbuf_new_size(dmu_buf_impl_t *db, int size, dmu_tx_t *tx);
 314 
 315 #define DB_DNODE(_db)           ((_db)->db_dnode_handle->dnh_dnode)
 316 #define DB_DNODE_LOCK(_db)      ((_db)->db_dnode_handle->dnh_zrlock)
 317 #define DB_DNODE_ENTER(_db)     (zrl_add(&DB_DNODE_LOCK(_db)))
 318 #define DB_DNODE_EXIT(_db)      (zrl_remove(&DB_DNODE_LOCK(_db)))
 319 #define DB_DNODE_HELD(_db)      (!zrl_is_zero(&DB_DNODE_LOCK(_db)))
 320 
 321 void dbuf_init(void);
 322 void dbuf_fini(void);
 323 
 324 boolean_t dbuf_is_metadata(dmu_buf_impl_t *db);
 325 boolean_t dbuf_ddt_is_l2cacheable(dmu_buf_impl_t *db);
 326 boolean_t dbuf_meta_is_l2cacheable(dmu_buf_impl_t *db);
 327 
 328 #define DBUF_GET_BUFC_TYPE(_db) \
 329         (dbuf_is_metadata(_db) ? ARC_BUFC_METADATA : ARC_BUFC_DATA)
 330 
 331 #define DBUF_IS_CACHEABLE(_db)                                          \
 332         ((_db)->db_objset->os_primary_cache == ZFS_CACHE_ALL ||           \
 333         (dbuf_is_metadata(_db) &&                                       \
 334         ((_db)->db_objset->os_primary_cache == ZFS_CACHE_METADATA)))
 335 
 336 
 337 /*
 338  * Checks whether we need to cache dbuf in l2arc.
 339  * Metadata is l2cacheable if it is not placed on special device
 340  * or it is placed on special device in "dual" mode. We need to check
 341  * for ddt in ZFS_CACHE_ALL and ZFS_CACHE_METADATA because it is in MOS.
 342  * ZFS_CACHE_DATA mode actually means to cache both data and cacheable
 343  * metadata.
 344  */
 345 #define DBUF_IS_L2CACHEABLE(_db)                                        \
 346         (((_db)->db_objset->os_secondary_cache == ZFS_CACHE_ALL &&        \
 347         (dbuf_ddt_is_l2cacheable(_db) == B_TRUE)) ||                    \
 348         ((_db)->db_objset->os_secondary_cache == ZFS_CACHE_METADATA &&  \
 349         (dbuf_is_metadata(_db)) &&                                      \
 350         (dbuf_ddt_is_l2cacheable(_db) == B_TRUE)) ||                    \
 351         ((dbuf_meta_is_l2cacheable(_db) == B_TRUE) &&                   \
 352         ((_db)->db_objset->os_secondary_cache == ZFS_CACHE_DATA)))
 353 
 354 #define DBUF_IS_L2COMPRESSIBLE(_db)                                     \
 355         ((_db)->db_objset->os_compress != ZIO_COMPRESS_OFF ||             \
 356         (dbuf_is_metadata(_db) && zfs_mdcomp_disable == B_FALSE))
 357 
 358 #ifdef ZFS_DEBUG
 359 
 360 /*
 361  * There should be a ## between the string literal and fmt, to make it
 362  * clear that we're joining two strings together, but gcc does not
 363  * support that preprocessor token.
 364  */
 365 #define dprintf_dbuf(dbuf, fmt, ...) do { \
 366         if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
 367         char __db_buf[32]; \
 368         uint64_t __db_obj = (dbuf)->db.db_object; \
 369         if (__db_obj == DMU_META_DNODE_OBJECT) \
 370                 (void) strcpy(__db_buf, "mdn"); \
 371         else \
 372                 (void) snprintf(__db_buf, sizeof (__db_buf), "%lld", \
 373                     (u_longlong_t)__db_obj); \
 374         dprintf_ds((dbuf)->db_objset->os_dsl_dataset, \
 375             "obj=%s lvl=%u blkid=%lld " fmt, \
 376             __db_buf, (dbuf)->db_level, \
 377             (u_longlong_t)(dbuf)->db_blkid, __VA_ARGS__); \
 378         } \
 379 _NOTE(CONSTCOND) } while (0)
 380 
 381 #define dprintf_dbuf_bp(db, bp, fmt, ...) do {                  \
 382         if (zfs_flags & ZFS_DEBUG_DPRINTF) {                        \
 383         char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP);  \
 384         snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, bp);          \
 385         dprintf_dbuf(db, fmt " %s\n", __VA_ARGS__, __blkbuf);   \
 386         kmem_free(__blkbuf, BP_SPRINTF_LEN);                    \
 387         }                                                       \
 388 _NOTE(CONSTCOND) } while (0)
 389 
 390 #define DBUF_VERIFY(db) dbuf_verify(db)
 391 
 392 #else
 393 
 394 #define dprintf_dbuf(db, fmt, ...)
 395 #define dprintf_dbuf_bp(db, bp, fmt, ...)
 396 #define DBUF_VERIFY(db)
 397 
 398 #endif
 399 
 400 
 401 #ifdef  __cplusplus
 402 }
 403 #endif
 404 
 405 #endif /* _SYS_DBUF_H */