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.
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.
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]
22 * Copyright 2009 Sun Microsystems, Inc. All rights reserved.
23 * Use is subject to license terms.
34 #include <sys/zfs_context.h>
35 #include <sys/refcount.h>
41 #define DB_BONUS_BLKID (-1ULL)
45 * define flags for dbuf_read
48 #define DB_RF_MUST_SUCCEED (1 << 0)
49 #define DB_RF_CANFAIL (1 << 1)
50 #define DB_RF_HAVESTRUCT (1 << 2)
51 #define DB_RF_NOPREFETCH (1 << 3)
52 #define DB_RF_NEVERWAIT (1 << 4)
53 #define DB_RF_CACHED (1 << 5)
56 * The simplified state transition diagram for dbufs looks like:
61 * (alloc)-->UNCACHED CACHED-->EVICTING-->(free)
67 * +--------> NOFILL -------+
69 typedef enum dbuf_states
{
83 * level = 0 means the user data
84 * level = 1 means the single indirect block
88 #define LIST_LINK_INACTIVE(link) \
89 ((link)->list_next == NULL && (link)->list_prev == NULL)
93 typedef enum override_states
{
99 typedef struct dbuf_dirty_record
{
100 /* link on our parents dirty list */
101 list_node_t dr_dirty_node
;
103 /* transaction group this data will sync in */
106 /* zio of outstanding write IO */
109 /* pointer back to our dbuf */
110 struct dmu_buf_impl
*dr_dbuf
;
112 /* pointer to next dirty record */
113 struct dbuf_dirty_record
*dr_next
;
115 /* pointer to parent dirty record */
116 struct dbuf_dirty_record
*dr_parent
;
119 struct dirty_indirect
{
121 /* protect access to list */
124 /* Our list of dirty children */
130 * dr_data is set when we dirty the buffer
131 * so that we can retain the pointer even if it
132 * gets COW'd in a subsequent transaction group.
135 blkptr_t dr_overridden_by
;
136 override_states_t dr_override_state
;
139 } dbuf_dirty_record_t
;
141 typedef struct dmu_buf_impl
{
143 * The following members are immutable, with the exception of
144 * db.db_data, which is protected by db_mtx.
147 /* the publicly visible structure */
150 /* the objset we belong to */
151 struct objset_impl
*db_objset
;
154 * the dnode we belong to (NULL when evicted)
156 struct dnode
*db_dnode
;
159 * our parent buffer; if the dnode points to us directly,
160 * db_parent == db_dnode->dn_dbuf
161 * only accessed by sync thread ???
162 * (NULL when evicted)
164 struct dmu_buf_impl
*db_parent
;
167 * link for hash table of all dmu_buf_impl_t's
169 struct dmu_buf_impl
*db_hash_next
;
171 /* our block number */
175 * Pointer to the blkptr_t which points to us. May be NULL if we
176 * don't have one yet. (NULL when evicted)
181 * Our indirection level. Data buffers have db_level==0.
182 * Indirect buffers which point to data buffers have
183 * db_level==1. etc. Buffers which contain dnodes have
184 * db_level==0, since the dnodes are stored in a file.
188 /* db_mtx protects the members below */
192 * Current state of the buffer
194 dbuf_states_t db_state
;
197 * Refcount accessed by dmu_buf_{hold,rele}.
198 * If nonzero, the buffer can't be destroyed.
199 * Protected by db_mtx.
203 /* buffer holding our data */
206 kcondvar_t db_changed
;
207 dbuf_dirty_record_t
*db_data_pending
;
209 /* pointer to most recent dirty record for this buffer */
210 dbuf_dirty_record_t
*db_last_dirty
;
213 * Our link on the owner dnodes's dn_dbufs list.
214 * Protected by its dn_dbufs_mtx.
218 /* Data which is unique to data (leaf) blocks: */
220 /* stuff we store for the user (see dmu_buf_set_user) */
222 void **db_user_data_ptr_ptr
;
223 dmu_buf_evict_func_t
*db_evict_func
;
225 uint8_t db_immediate_evict
;
226 uint8_t db_freed_in_flight
;
231 /* Note: the dbuf hash table is exposed only for the mdb module */
232 #define DBUF_MUTEXES 256
233 #define DBUF_HASH_MUTEX(h, idx) (&(h)->hash_mutexes[(idx) & (DBUF_MUTEXES-1)])
234 typedef struct dbuf_hash_table
{
235 uint64_t hash_table_mask
;
236 dmu_buf_impl_t
**hash_table
;
237 kmutex_t hash_mutexes
[DBUF_MUTEXES
];
241 uint64_t dbuf_whichblock(struct dnode
*di
, uint64_t offset
);
243 dmu_buf_impl_t
*dbuf_create_tlib(struct dnode
*dn
, char *data
);
244 void dbuf_create_bonus(struct dnode
*dn
);
246 dmu_buf_impl_t
*dbuf_hold(struct dnode
*dn
, uint64_t blkid
, void *tag
);
247 dmu_buf_impl_t
*dbuf_hold_level(struct dnode
*dn
, int level
, uint64_t blkid
,
249 int dbuf_hold_impl(struct dnode
*dn
, uint8_t level
, uint64_t blkid
, int create
,
250 void *tag
, dmu_buf_impl_t
**dbp
);
252 void dbuf_prefetch(struct dnode
*dn
, uint64_t blkid
);
254 void dbuf_add_ref(dmu_buf_impl_t
*db
, void *tag
);
255 uint64_t dbuf_refcount(dmu_buf_impl_t
*db
);
257 void dbuf_rele(dmu_buf_impl_t
*db
, void *tag
);
259 dmu_buf_impl_t
*dbuf_find(struct dnode
*dn
, uint8_t level
, uint64_t blkid
);
261 int dbuf_read(dmu_buf_impl_t
*db
, zio_t
*zio
, uint32_t flags
);
262 void dbuf_will_dirty(dmu_buf_impl_t
*db
, dmu_tx_t
*tx
);
263 void dbuf_fill_done(dmu_buf_impl_t
*db
, dmu_tx_t
*tx
);
264 void dmu_buf_will_not_fill(dmu_buf_t
*db
, dmu_tx_t
*tx
);
265 void dmu_buf_will_fill(dmu_buf_t
*db
, dmu_tx_t
*tx
);
266 void dmu_buf_fill_done(dmu_buf_t
*db
, dmu_tx_t
*tx
);
267 void dbuf_assign_arcbuf(dmu_buf_impl_t
*db
, arc_buf_t
*buf
, dmu_tx_t
*tx
);
268 dbuf_dirty_record_t
*dbuf_dirty(dmu_buf_impl_t
*db
, dmu_tx_t
*tx
);
270 void dbuf_clear(dmu_buf_impl_t
*db
);
271 void dbuf_evict(dmu_buf_impl_t
*db
);
273 void dbuf_setdirty(dmu_buf_impl_t
*db
, dmu_tx_t
*tx
);
274 void dbuf_unoverride(dbuf_dirty_record_t
*dr
);
275 void dbuf_sync_list(list_t
*list
, dmu_tx_t
*tx
);
277 void dbuf_free_range(struct dnode
*dn
, uint64_t start
, uint64_t end
,
280 void dbuf_new_size(dmu_buf_impl_t
*db
, int size
, dmu_tx_t
*tx
);
282 void dbuf_init(void);
283 void dbuf_fini(void);
285 #define DBUF_IS_METADATA(db) \
286 ((db)->db_level > 0 || dmu_ot[(db)->db_dnode->dn_type].ot_metadata)
288 #define DBUF_GET_BUFC_TYPE(db) \
289 (DBUF_IS_METADATA(db) ? ARC_BUFC_METADATA : ARC_BUFC_DATA)
291 #define DBUF_IS_CACHEABLE(db) \
292 ((db)->db_objset->os_primary_cache == ZFS_CACHE_ALL || \
293 (DBUF_IS_METADATA(db) && \
294 ((db)->db_objset->os_primary_cache == ZFS_CACHE_METADATA)))
296 #define DBUF_IS_L2CACHEABLE(db) \
297 ((db)->db_objset->os_secondary_cache == ZFS_CACHE_ALL || \
298 (DBUF_IS_METADATA(db) && \
299 ((db)->db_objset->os_secondary_cache == ZFS_CACHE_METADATA)))
304 * There should be a ## between the string literal and fmt, to make it
305 * clear that we're joining two strings together, but gcc does not
306 * support that preprocessor token.
308 #define dprintf_dbuf(dbuf, fmt, ...) do { \
309 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
311 uint64_t __db_obj = (dbuf)->db.db_object; \
312 if (__db_obj == DMU_META_DNODE_OBJECT) \
313 (void) strcpy(__db_buf, "mdn"); \
315 (void) snprintf(__db_buf, sizeof (__db_buf), "%lld", \
316 (u_longlong_t)__db_obj); \
317 dprintf_ds((dbuf)->db_objset->os_dsl_dataset, \
318 "obj=%s lvl=%u blkid=%lld " fmt, \
319 __db_buf, (dbuf)->db_level, \
320 (u_longlong_t)(dbuf)->db_blkid, __VA_ARGS__); \
322 _NOTE(CONSTCOND) } while (0)
324 #define dprintf_dbuf_bp(db, bp, fmt, ...) do { \
325 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
326 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \
327 sprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, bp); \
328 dprintf_dbuf(db, fmt " %s\n", __VA_ARGS__, __blkbuf); \
329 kmem_free(__blkbuf, BP_SPRINTF_LEN); \
331 _NOTE(CONSTCOND) } while (0)
333 #define DBUF_VERIFY(db) dbuf_verify(db)
337 #define dprintf_dbuf(db, fmt, ...)
338 #define dprintf_dbuf_bp(db, bp, fmt, ...)
339 #define DBUF_VERIFY(db)
348 #endif /* _SYS_DBUF_H */