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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2012, 2015 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.
36 #include <sys/zfs_context.h>
37 #include <sys/refcount.h>
38 #include <sys/zrlock.h>
47 * define flags for dbuf_read
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)
58 * The simplified state transition diagram for dbufs looks like:
63 * (alloc)-->UNCACHED CACHED-->EVICTING-->(free)
69 * +--------> NOFILL -------+
71 * DB_SEARCH is an invalid state for a dbuf. It is used by dbuf_free_range
72 * to find all dbufs in a range of a dnode and must be less than any other
73 * dbuf_states_t (see comment on dn_dbufs in dnode.h).
75 typedef enum dbuf_states
{
89 * level = 0 means the user data
90 * level = 1 means the single indirect block
96 typedef enum override_states
{
102 typedef struct dbuf_dirty_record
{
103 /* link on our parents dirty list */
104 list_node_t dr_dirty_node
;
106 /* transaction group this data will sync in */
109 /* zio of outstanding write IO */
112 /* pointer back to our dbuf */
113 struct dmu_buf_impl
*dr_dbuf
;
115 /* pointer to next dirty record */
116 struct dbuf_dirty_record
*dr_next
;
118 /* pointer to parent dirty record */
119 struct dbuf_dirty_record
*dr_parent
;
121 /* How much space was changed to dsl_pool_dirty_space() for this? */
122 unsigned int dr_accounted
;
125 struct dirty_indirect
{
127 /* protect access to list */
130 /* Our list of dirty children */
136 * dr_data is set when we dirty the buffer
137 * so that we can retain the pointer even if it
138 * gets COW'd in a subsequent transaction group.
141 blkptr_t dr_overridden_by
;
142 override_states_t dr_override_state
;
144 boolean_t dr_nopwrite
;
147 } dbuf_dirty_record_t
;
149 typedef struct dmu_buf_impl
{
151 * The following members are immutable, with the exception of
152 * db.db_data, which is protected by db_mtx.
155 /* the publicly visible structure */
158 /* the objset we belong to */
159 struct objset
*db_objset
;
162 * handle to safely access the dnode we belong to (NULL when evicted)
164 struct dnode_handle
*db_dnode_handle
;
167 * our parent buffer; if the dnode points to us directly,
168 * db_parent == db_dnode_handle->dnh_dnode->dn_dbuf
169 * only accessed by sync thread ???
170 * (NULL when evicted)
171 * May change from NULL to non-NULL under the protection of db_mtx
172 * (see dbuf_check_blkptr())
174 struct dmu_buf_impl
*db_parent
;
177 * link for hash table of all dmu_buf_impl_t's
179 struct dmu_buf_impl
*db_hash_next
;
181 /* our block number */
185 * Pointer to the blkptr_t which points to us. May be NULL if we
186 * don't have one yet. (NULL when evicted)
191 * Our indirection level. Data buffers have db_level==0.
192 * Indirect buffers which point to data buffers have
193 * db_level==1. etc. Buffers which contain dnodes have
194 * db_level==0, since the dnodes are stored in a file.
198 /* db_mtx protects the members below */
202 * Current state of the buffer
204 dbuf_states_t db_state
;
207 * Refcount accessed by dmu_buf_{hold,rele}.
208 * If nonzero, the buffer can't be destroyed.
209 * Protected by db_mtx.
213 /* buffer holding our data */
216 kcondvar_t db_changed
;
217 dbuf_dirty_record_t
*db_data_pending
;
219 /* pointer to most recent dirty record for this buffer */
220 dbuf_dirty_record_t
*db_last_dirty
;
223 * Our link on the owner dnodes's dn_dbufs list.
224 * Protected by its dn_dbufs_mtx.
228 /* Data which is unique to data (leaf) blocks: */
230 /* User callback information. */
231 dmu_buf_user_t
*db_user
;
233 uint8_t db_immediate_evict
;
234 uint8_t db_freed_in_flight
;
239 /* Note: the dbuf hash table is exposed only for the mdb module */
240 #define DBUF_MUTEXES 256
241 #define DBUF_HASH_MUTEX(h, idx) (&(h)->hash_mutexes[(idx) & (DBUF_MUTEXES-1)])
242 typedef struct dbuf_hash_table
{
243 uint64_t hash_table_mask
;
244 dmu_buf_impl_t
**hash_table
;
245 kmutex_t hash_mutexes
[DBUF_MUTEXES
];
248 uint64_t dbuf_whichblock(struct dnode
*di
, int64_t level
, uint64_t offset
);
250 dmu_buf_impl_t
*dbuf_create_tlib(struct dnode
*dn
, char *data
);
251 void dbuf_create_bonus(struct dnode
*dn
);
252 int dbuf_spill_set_blksz(dmu_buf_t
*db
, uint64_t blksz
, dmu_tx_t
*tx
);
253 void dbuf_spill_hold(struct dnode
*dn
, dmu_buf_impl_t
**dbp
, void *tag
);
255 void dbuf_rm_spill(struct dnode
*dn
, dmu_tx_t
*tx
);
257 dmu_buf_impl_t
*dbuf_hold(struct dnode
*dn
, uint64_t blkid
, void *tag
);
258 dmu_buf_impl_t
*dbuf_hold_level(struct dnode
*dn
, int level
, uint64_t blkid
,
260 int dbuf_hold_impl(struct dnode
*dn
, uint8_t level
, uint64_t blkid
,
261 boolean_t fail_sparse
, boolean_t fail_uncached
,
262 void *tag
, dmu_buf_impl_t
**dbp
);
264 void dbuf_prefetch(struct dnode
*dn
, int64_t level
, uint64_t blkid
,
265 zio_priority_t prio
, arc_flags_t aflags
);
267 void dbuf_add_ref(dmu_buf_impl_t
*db
, void *tag
);
268 boolean_t
dbuf_try_add_ref(dmu_buf_t
*db
, objset_t
*os
, uint64_t obj
,
269 uint64_t blkid
, void *tag
);
270 uint64_t dbuf_refcount(dmu_buf_impl_t
*db
);
272 void dbuf_rele(dmu_buf_impl_t
*db
, void *tag
);
273 void dbuf_rele_and_unlock(dmu_buf_impl_t
*db
, void *tag
);
275 dmu_buf_impl_t
*dbuf_find(struct objset
*os
, uint64_t object
, uint8_t level
,
278 int dbuf_read(dmu_buf_impl_t
*db
, zio_t
*zio
, uint32_t flags
);
279 void dmu_buf_will_not_fill(dmu_buf_t
*db
, dmu_tx_t
*tx
);
280 void dmu_buf_will_fill(dmu_buf_t
*db
, dmu_tx_t
*tx
);
281 void dmu_buf_fill_done(dmu_buf_t
*db
, dmu_tx_t
*tx
);
282 void dbuf_assign_arcbuf(dmu_buf_impl_t
*db
, arc_buf_t
*buf
, dmu_tx_t
*tx
);
283 dbuf_dirty_record_t
*dbuf_dirty(dmu_buf_impl_t
*db
, dmu_tx_t
*tx
);
284 arc_buf_t
*dbuf_loan_arcbuf(dmu_buf_impl_t
*db
);
285 void dmu_buf_write_embedded(dmu_buf_t
*dbuf
, void *data
,
286 bp_embedded_type_t etype
, enum zio_compress comp
,
287 int uncompressed_size
, int compressed_size
, int byteorder
, dmu_tx_t
*tx
);
289 void dbuf_clear(dmu_buf_impl_t
*db
);
290 void dbuf_evict(dmu_buf_impl_t
*db
);
292 void dbuf_setdirty(dmu_buf_impl_t
*db
, dmu_tx_t
*tx
);
293 void dbuf_unoverride(dbuf_dirty_record_t
*dr
);
294 void dbuf_sync_list(list_t
*list
, int level
, dmu_tx_t
*tx
);
295 void dbuf_release_bp(dmu_buf_impl_t
*db
);
297 void dbuf_free_range(struct dnode
*dn
, uint64_t start
, uint64_t end
,
300 void dbuf_new_size(dmu_buf_impl_t
*db
, int size
, dmu_tx_t
*tx
);
302 #define DB_DNODE(_db) ((_db)->db_dnode_handle->dnh_dnode)
303 #define DB_DNODE_LOCK(_db) ((_db)->db_dnode_handle->dnh_zrlock)
304 #define DB_DNODE_ENTER(_db) (zrl_add(&DB_DNODE_LOCK(_db)))
305 #define DB_DNODE_EXIT(_db) (zrl_remove(&DB_DNODE_LOCK(_db)))
306 #define DB_DNODE_HELD(_db) (!zrl_is_zero(&DB_DNODE_LOCK(_db)))
308 void dbuf_init(void);
309 void dbuf_fini(void);
311 boolean_t
dbuf_is_metadata(dmu_buf_impl_t
*db
);
313 #define DBUF_GET_BUFC_TYPE(_db) \
314 (dbuf_is_metadata(_db) ? ARC_BUFC_METADATA : ARC_BUFC_DATA)
316 #define DBUF_IS_CACHEABLE(_db) \
317 ((_db)->db_objset->os_primary_cache == ZFS_CACHE_ALL || \
318 (dbuf_is_metadata(_db) && \
319 ((_db)->db_objset->os_primary_cache == ZFS_CACHE_METADATA)))
321 #define DBUF_IS_L2CACHEABLE(_db) \
322 ((_db)->db_objset->os_secondary_cache == ZFS_CACHE_ALL || \
323 (dbuf_is_metadata(_db) && \
324 ((_db)->db_objset->os_secondary_cache == ZFS_CACHE_METADATA)))
326 #define DBUF_IS_L2COMPRESSIBLE(_db) \
327 ((_db)->db_objset->os_compress != ZIO_COMPRESS_OFF || \
328 (dbuf_is_metadata(_db) && zfs_mdcomp_disable == B_FALSE))
333 * There should be a ## between the string literal and fmt, to make it
334 * clear that we're joining two strings together, but gcc does not
335 * support that preprocessor token.
337 #define dprintf_dbuf(dbuf, fmt, ...) do { \
338 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
340 uint64_t __db_obj = (dbuf)->db.db_object; \
341 if (__db_obj == DMU_META_DNODE_OBJECT) \
342 (void) strcpy(__db_buf, "mdn"); \
344 (void) snprintf(__db_buf, sizeof (__db_buf), "%lld", \
345 (u_longlong_t)__db_obj); \
346 dprintf_ds((dbuf)->db_objset->os_dsl_dataset, \
347 "obj=%s lvl=%u blkid=%lld " fmt, \
348 __db_buf, (dbuf)->db_level, \
349 (u_longlong_t)(dbuf)->db_blkid, __VA_ARGS__); \
351 _NOTE(CONSTCOND) } while (0)
353 #define dprintf_dbuf_bp(db, bp, fmt, ...) do { \
354 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
355 char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \
356 snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, bp); \
357 dprintf_dbuf(db, fmt " %s\n", __VA_ARGS__, __blkbuf); \
358 kmem_free(__blkbuf, BP_SPRINTF_LEN); \
360 _NOTE(CONSTCOND) } while (0)
362 #define DBUF_VERIFY(db) dbuf_verify(db)
366 #define dprintf_dbuf(db, fmt, ...)
367 #define dprintf_dbuf_bp(db, bp, fmt, ...)
368 #define DBUF_VERIFY(db)
377 #endif /* _SYS_DBUF_H */