include/sys/dbuf.h

   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, 2018 by Delphix. All rights reserved.
  24  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
  25  * Copyright (c) 2014 Spectra Logic Corporation, 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 #include <sys/multilist.h>
  40
  41 #ifdef  __cplusplus
  42 extern "C" {
  43 #endif
  44
  45 #define IN_DMU_SYNC 2
  46
  47 /*
  48  * define flags for dbuf_read
  49  */
  50
  51 #define DB_RF_MUST_SUCCEED      (1 << 0)
  52 #define DB_RF_CANFAIL           (1 << 1)
  53 #define DB_RF_HAVESTRUCT        (1 << 2)
  54 #define DB_RF_NOPREFETCH        (1 << 3)
  55 #define DB_RF_NEVERWAIT         (1 << 4)
  56 #define DB_RF_CACHED            (1 << 5)
  57 #define DB_RF_NO_DECRYPT        (1 << 6)
  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 typedef enum dbuf_cached_state {
  88         DB_NO_CACHE = -1,
  89         DB_DBUF_CACHE,
  90         DB_DBUF_METADATA_CACHE,
  91         DB_CACHE_MAX
  92 } dbuf_cached_state_t;
  93
  94 struct dnode;
  95 struct dmu_tx;
  96
  97 /*
  98  * level = 0 means the user data
  99  * level = 1 means the single indirect block
 100  * etc.
 101  */
 102
 103 struct dmu_buf_impl;
 104
 105 typedef enum override_states {
 106         DR_NOT_OVERRIDDEN,
 107         DR_IN_DMU_SYNC,
 108         DR_OVERRIDDEN
 109 } override_states_t;
 110
 111 typedef struct dbuf_dirty_record {
 112         /* link on our parents dirty list */
 113         list_node_t dr_dirty_node;
 114
 115         /* transaction group this data will sync in */
 116         uint64_t dr_txg;
 117
 118         /* zio of outstanding write IO */
 119         zio_t *dr_zio;
 120
 121         /* pointer back to our dbuf */
 122         struct dmu_buf_impl *dr_dbuf;
 123
 124         /* pointer to next dirty record */
 125         struct dbuf_dirty_record *dr_next;
 126
 127         /* pointer to parent dirty record */
 128         struct dbuf_dirty_record *dr_parent;
 129
 130         /* How much space was changed to dsl_pool_dirty_space() for this? */
 131         unsigned int dr_accounted;
 132
 133         /* A copy of the bp that points to us */
 134         blkptr_t dr_bp_copy;
 135
 136         union dirty_types {
 137                 struct dirty_indirect {
 138
 139                         /* protect access to list */
 140                         kmutex_t dr_mtx;
 141
 142                         /* Our list of dirty children */
 143                         list_t dr_children;
 144                 } di;
 145                 struct dirty_leaf {
 146
 147                         /*
 148                          * dr_data is set when we dirty the buffer
 149                          * so that we can retain the pointer even if it
 150                          * gets COW'd in a subsequent transaction group.
 151                          */
 152                         arc_buf_t *dr_data;
 153                         blkptr_t dr_overridden_by;
 154                         override_states_t dr_override_state;
 155                         uint8_t dr_copies;
 156                         boolean_t dr_nopwrite;
 157                         boolean_t dr_has_raw_params;
 158
 159                         /*
 160                          * If dr_has_raw_params is set, the following crypt
 161                          * params will be set on the BP that's written.
 162                          */
 163                         boolean_t dr_byteorder;
 164                         uint8_t dr_salt[ZIO_DATA_SALT_LEN];
 165                         uint8_t dr_iv[ZIO_DATA_IV_LEN];
 166                         uint8_t dr_mac[ZIO_DATA_MAC_LEN];
 167                 } dl;
 168         } dt;
 169 } dbuf_dirty_record_t;
 170
 171 typedef struct dmu_buf_impl {
 172         /*
 173          * The following members are immutable, with the exception of
 174          * db.db_data, which is protected by db_mtx.
 175          */
 176
 177         /* the publicly visible structure */
 178         dmu_buf_t db;
 179
 180         /* the objset we belong to */
 181         struct objset *db_objset;
 182
 183         /*
 184          * handle to safely access the dnode we belong to (NULL when evicted)
 185          */
 186         struct dnode_handle *db_dnode_handle;
 187
 188         /*
 189          * our parent buffer; if the dnode points to us directly,
 190          * db_parent == db_dnode_handle->dnh_dnode->dn_dbuf
 191          * only accessed by sync thread ???
 192          * (NULL when evicted)
 193          * May change from NULL to non-NULL under the protection of db_mtx
 194          * (see dbuf_check_blkptr())
 195          */
 196         struct dmu_buf_impl *db_parent;
 197
 198         /*
 199          * link for hash table of all dmu_buf_impl_t's
 200          */
 201         struct dmu_buf_impl *db_hash_next;
 202
 203         /* our block number */
 204         uint64_t db_blkid;
 205
 206         /*
 207          * Pointer to the blkptr_t which points to us. May be NULL if we
 208          * don't have one yet. (NULL when evicted)
 209          */
 210         blkptr_t *db_blkptr;
 211
 212         /*
 213          * Our indirection level.  Data buffers have db_level==0.
 214          * Indirect buffers which point to data buffers have
 215          * db_level==1. etc.  Buffers which contain dnodes have
 216          * db_level==0, since the dnodes are stored in a file.
 217          */
 218         uint8_t db_level;
 219
 220         /* db_mtx protects the members below */
 221         kmutex_t db_mtx;
 222
 223         /*
 224          * Current state of the buffer
 225          */
 226         dbuf_states_t db_state;
 227
 228         /*
 229          * Refcount accessed by dmu_buf_{hold,rele}.
 230          * If nonzero, the buffer can't be destroyed.
 231          * Protected by db_mtx.
 232          */
 233         zfs_refcount_t db_holds;
 234
 235         /* buffer holding our data */
 236         arc_buf_t *db_buf;
 237
 238         kcondvar_t db_changed;
 239         dbuf_dirty_record_t *db_data_pending;
 240
 241         /* pointer to most recent dirty record for this buffer */
 242         dbuf_dirty_record_t *db_last_dirty;
 243
 244         /*
 245          * Our link on the owner dnodes's dn_dbufs list.
 246          * Protected by its dn_dbufs_mtx.
 247          */
 248         avl_node_t db_link;
 249
 250         /* Link in dbuf_cache or dbuf_metadata_cache */
 251         multilist_node_t db_cache_link;
 252
 253         /* Tells us which dbuf cache this dbuf is in, if any */
 254         dbuf_cached_state_t db_caching_status;
 255
 256         /* Data which is unique to data (leaf) blocks: */
 257
 258         /* User callback information. */
 259         dmu_buf_user_t *db_user;
 260
 261         /*
 262          * Evict user data as soon as the dirty and reference
 263          * counts are equal.
 264          */
 265         uint8_t db_user_immediate_evict;
 266
 267         /*
 268          * This block was freed while a read or write was
 269          * active.
 270          */
 271         uint8_t db_freed_in_flight;
 272
 273         /*
 274          * dnode_evict_dbufs() or dnode_evict_bonus() tried to
 275          * evict this dbuf, but couldn't due to outstanding
 276          * references.  Evict once the refcount drops to 0.
 277          */
 278         uint8_t db_pending_evict;
 279
 280         uint8_t db_dirtycnt;
 281 } dmu_buf_impl_t;
 282
 283 /* Note: the dbuf hash table is exposed only for the mdb module */
 284 #define DBUF_MUTEXES 8192
 285 #define DBUF_HASH_MUTEX(h, idx) (&(h)->hash_mutexes[(idx) & (DBUF_MUTEXES-1)])
 286 typedef struct dbuf_hash_table {
 287         uint64_t hash_table_mask;
 288         dmu_buf_impl_t **hash_table;
 289         kmutex_t hash_mutexes[DBUF_MUTEXES];
 290 } dbuf_hash_table_t;
 291
 292 uint64_t dbuf_whichblock(const struct dnode *di, const int64_t level,
 293     const uint64_t offset);
 294
 295 void dbuf_create_bonus(struct dnode *dn);
 296 int dbuf_spill_set_blksz(dmu_buf_t *db, uint64_t blksz, dmu_tx_t *tx);
 297
 298 void dbuf_rm_spill(struct dnode *dn, dmu_tx_t *tx);
 299
 300 dmu_buf_impl_t *dbuf_hold(struct dnode *dn, uint64_t blkid, void *tag);
 301 dmu_buf_impl_t *dbuf_hold_level(struct dnode *dn, int level, uint64_t blkid,
 302     void *tag);
 303 int dbuf_hold_impl(struct dnode *dn, uint8_t level, uint64_t blkid,
 304     boolean_t fail_sparse, boolean_t fail_uncached,
 305     void *tag, dmu_buf_impl_t **dbp);
 306
 307 void dbuf_prefetch(struct dnode *dn, int64_t level, uint64_t blkid,
 308     zio_priority_t prio, arc_flags_t aflags);
 309
 310 void dbuf_add_ref(dmu_buf_impl_t *db, void *tag);
 311 boolean_t dbuf_try_add_ref(dmu_buf_t *db, objset_t *os, uint64_t obj,
 312     uint64_t blkid, void *tag);
 313 uint64_t dbuf_refcount(dmu_buf_impl_t *db);
 314
 315 void dbuf_rele(dmu_buf_impl_t *db, void *tag);
 316 void dbuf_rele_and_unlock(dmu_buf_impl_t *db, void *tag, boolean_t evicting);
 317
 318 dmu_buf_impl_t *dbuf_find(struct objset *os, uint64_t object, uint8_t level,
 319     uint64_t blkid);
 320
 321 int dbuf_read(dmu_buf_impl_t *db, zio_t *zio, uint32_t flags);
 322 void dmu_buf_will_not_fill(dmu_buf_t *db, dmu_tx_t *tx);
 323 void dmu_buf_will_fill(dmu_buf_t *db, dmu_tx_t *tx);
 324 void dmu_buf_fill_done(dmu_buf_t *db, dmu_tx_t *tx);
 325 void dbuf_assign_arcbuf(dmu_buf_impl_t *db, arc_buf_t *buf, dmu_tx_t *tx);
 326 dbuf_dirty_record_t *dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx);
 327 arc_buf_t *dbuf_loan_arcbuf(dmu_buf_impl_t *db);
 328 void dmu_buf_write_embedded(dmu_buf_t *dbuf, void *data,
 329     bp_embedded_type_t etype, enum zio_compress comp,
 330     int uncompressed_size, int compressed_size, int byteorder, dmu_tx_t *tx);
 331
 332 void dbuf_destroy(dmu_buf_impl_t *db);
 333
 334 void dbuf_unoverride(dbuf_dirty_record_t *dr);
 335 void dbuf_sync_list(list_t *list, int level, dmu_tx_t *tx);
 336 void dbuf_release_bp(dmu_buf_impl_t *db);
 337
 338 boolean_t dbuf_can_remap(const dmu_buf_impl_t *buf);
 339
 340 void dbuf_free_range(struct dnode *dn, uint64_t start, uint64_t end,
 341     struct dmu_tx *);
 342
 343 void dbuf_new_size(dmu_buf_impl_t *db, int size, dmu_tx_t *tx);
 344
 345 void dbuf_stats_init(dbuf_hash_table_t *hash);
 346 void dbuf_stats_destroy(void);
 347
 348 #define DB_DNODE(_db)           ((_db)->db_dnode_handle->dnh_dnode)
 349 #define DB_DNODE_LOCK(_db)      ((_db)->db_dnode_handle->dnh_zrlock)
 350 #define DB_DNODE_ENTER(_db)     (zrl_add(&DB_DNODE_LOCK(_db)))
 351 #define DB_DNODE_EXIT(_db)      (zrl_remove(&DB_DNODE_LOCK(_db)))
 352 #define DB_DNODE_HELD(_db)      (!zrl_is_zero(&DB_DNODE_LOCK(_db)))
 353
 354 void dbuf_init(void);
 355 void dbuf_fini(void);
 356
 357 boolean_t dbuf_is_metadata(dmu_buf_impl_t *db);
 358
 359 #define DBUF_GET_BUFC_TYPE(_db) \
 360         (dbuf_is_metadata(_db) ? ARC_BUFC_METADATA : ARC_BUFC_DATA)
 361
 362 #define DBUF_IS_CACHEABLE(_db)                                          \
 363         ((_db)->db_objset->os_primary_cache == ZFS_CACHE_ALL ||         \
 364         (dbuf_is_metadata(_db) &&                                       \
 365         ((_db)->db_objset->os_primary_cache == ZFS_CACHE_METADATA)))
 366
 367 #define DBUF_IS_L2CACHEABLE(_db)                                        \
 368         ((_db)->db_objset->os_secondary_cache == ZFS_CACHE_ALL ||       \
 369         (dbuf_is_metadata(_db) &&                                       \
 370         ((_db)->db_objset->os_secondary_cache == ZFS_CACHE_METADATA)))
 371
 372 #define DNODE_LEVEL_IS_L2CACHEABLE(_dn, _level)                         \
 373         ((_dn)->dn_objset->os_secondary_cache == ZFS_CACHE_ALL ||       \
 374         (((_level) > 0 ||                                               \
 375         DMU_OT_IS_METADATA((_dn)->dn_handle->dnh_dnode->dn_type)) &&    \
 376         ((_dn)->dn_objset->os_secondary_cache == ZFS_CACHE_METADATA)))
 377
 378 #ifdef ZFS_DEBUG
 379
 380 /*
 381  * There should be a ## between the string literal and fmt, to make it
 382  * clear that we're joining two strings together, but gcc does not
 383  * support that preprocessor token.
 384  */
 385 #define dprintf_dbuf(dbuf, fmt, ...) do { \
 386         if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
 387         char __db_buf[32]; \
 388         uint64_t __db_obj = (dbuf)->db.db_object; \
 389         if (__db_obj == DMU_META_DNODE_OBJECT) \
 390                 (void) strcpy(__db_buf, "mdn"); \
 391         else \
 392                 (void) snprintf(__db_buf, sizeof (__db_buf), "%lld", \
 393                     (u_longlong_t)__db_obj); \
 394         dprintf_ds((dbuf)->db_objset->os_dsl_dataset, \
 395             "obj=%s lvl=%u blkid=%lld " fmt, \
 396             __db_buf, (dbuf)->db_level, \
 397             (u_longlong_t)(dbuf)->db_blkid, __VA_ARGS__); \
 398         } \
 399 _NOTE(CONSTCOND) } while (0)
 400
 401 #define dprintf_dbuf_bp(db, bp, fmt, ...) do {                  \
 402         if (zfs_flags & ZFS_DEBUG_DPRINTF) {                    \
 403         char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP);  \
 404         snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, bp);          \
 405         dprintf_dbuf(db, fmt " %s\n", __VA_ARGS__, __blkbuf);   \
 406         kmem_free(__blkbuf, BP_SPRINTF_LEN);                    \
 407         }                                                       \
 408 _NOTE(CONSTCOND) } while (0)
 409
 410 #define DBUF_VERIFY(db) dbuf_verify(db)
 411
 412 #else
 413
 414 #define dprintf_dbuf(db, fmt, ...)
 415 #define dprintf_dbuf_bp(db, bp, fmt, ...)
 416 #define DBUF_VERIFY(db)
 417
 418 #endif
 419
 420
 421 #ifdef  __cplusplus
 422 }
 423 #endif
 424
 425 #endif /* _SYS_DBUF_H */