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.
29 #include <sys/types.h>
41 * Each objset has its own intent log. The log header (zil_header_t)
42 * for objset N's intent log is kept in the Nth object of the SPA's
43 * intent_log objset. The log header points to a chain of log blocks,
44 * each of which contains log records (i.e., transactions) followed by
45 * a log block trailer (zil_trailer_t). The format of a log record
46 * depends on the record (or transaction) type, but all records begin
47 * with a common structure that defines the type, length, and txg.
51 * Intent log header - this on disk structure holds fields to manage
52 * the log. All fields are 64 bit to easily handle cross architectures.
54 typedef struct zil_header {
55 uint64_t zh_claim_txg; /* txg in which log blocks were claimed */
56 uint64_t zh_replay_seq; /* highest replayed sequence number */
57 blkptr_t zh_log; /* log chain */
58 uint64_t zh_claim_seq; /* highest claimed sequence number */
59 uint64_t zh_flags; /* header flags */
64 * zh_flags bit settings
66 #define ZIL_REPLAY_NEEDED 0x1 /* replay needed - internal only */
69 * Log block trailer - structure at the end of the header and each log block
71 * The zit_bt contains a zbt_cksum which for the intent log is
72 * the sequence number of this log block. A seq of 0 is invalid.
73 * The zbt_cksum is checked by the SPA against the sequence
74 * number passed in the blk_cksum field of the blkptr_t
76 typedef struct zil_trailer {
78 blkptr_t zit_next_blk; /* next block in chain */
79 uint64_t zit_nused; /* bytes in log block used */
80 zio_block_tail_t zit_bt; /* block trailer */
83 #define ZIL_MIN_BLKSZ 4096ULL
84 #define ZIL_MAX_BLKSZ SPA_MAXBLOCKSIZE
85 #define ZIL_BLK_DATA_SZ(lwb) ((lwb)->lwb_sz - sizeof (zil_trailer_t))
88 * The words of a log block checksum.
90 #define ZIL_ZC_GUID_0 0
91 #define ZIL_ZC_GUID_1 1
92 #define ZIL_ZC_OBJSET 2
95 typedef enum zil_create {
102 * size of xvattr log section.
103 * its composed of lr_attr_t + xvattr bitmap + 2 64 bit timestamps
104 * for create time and a single 64 bit integer for all of the attributes,
105 * and 4 64 bit integers (32 bytes) for the scanstamp.
109 #define ZIL_XVAT_SIZE(mapsize) \
110 sizeof (lr_attr_t) + (sizeof (uint32_t) * (mapsize - 1)) + \
111 (sizeof (uint64_t) * 7)
114 * Size of ACL in log. The ACE data is padded out to properly align
115 * on 8 byte boundary.
118 #define ZIL_ACE_LENGTH(x) (roundup(x, sizeof (uint64_t)))
121 * Intent log transaction types and record structures
123 #define TX_CREATE 1 /* Create file */
124 #define TX_MKDIR 2 /* Make directory */
125 #define TX_MKXATTR 3 /* Make XATTR directory */
126 #define TX_SYMLINK 4 /* Create symbolic link to a file */
127 #define TX_REMOVE 5 /* Remove file */
128 #define TX_RMDIR 6 /* Remove directory */
129 #define TX_LINK 7 /* Create hard link to a file */
130 #define TX_RENAME 8 /* Rename a file */
131 #define TX_WRITE 9 /* File write */
132 #define TX_TRUNCATE 10 /* Truncate a file */
133 #define TX_SETATTR 11 /* Set file attributes */
134 #define TX_ACL_V0 12 /* Set old formatted ACL */
135 #define TX_ACL 13 /* Set ACL */
136 #define TX_CREATE_ACL 14 /* create with ACL */
137 #define TX_CREATE_ATTR 15 /* create + attrs */
138 #define TX_CREATE_ACL_ATTR 16 /* create with ACL + attrs */
139 #define TX_MKDIR_ACL 17 /* mkdir with ACL */
140 #define TX_MKDIR_ATTR 18 /* mkdir with attr */
141 #define TX_MKDIR_ACL_ATTR 19 /* mkdir with ACL + attrs */
142 #define TX_MAX_TYPE 20 /* Max transaction type */
145 * The transactions for mkdir, symlink, remove, rmdir, link, and rename
146 * may have the following bit set, indicating the original request
147 * specified case-insensitive handling of names.
149 #define TX_CI ((uint64_t)0x1 << 63) /* case-insensitive behavior requested */
152 * Format of log records.
153 * The fields are carefully defined to allow them to be aligned
154 * and sized the same on sparc & intel architectures.
155 * Each log record has a common structure at the beginning.
157 * Note, lrc_seq holds two different sequence numbers. Whilst in memory
158 * it contains the transaction sequence number. The log record on
159 * disk holds the sequence number of all log records which is used to
160 * ensure we don't replay the same record. The two sequence numbers are
161 * different because the transactions can now be pushed out of order.
163 typedef struct { /* common log record header */
164 uint64_t lrc_txtype; /* intent log transaction type */
165 uint64_t lrc_reclen; /* transaction record length */
166 uint64_t lrc_txg; /* dmu transaction group number */
167 uint64_t lrc_seq; /* see comment above */
171 * Handle option extended vattr attributes.
173 * Whenever new attributes are added the version number
174 * will need to be updated as will code in
175 * zfs_log.c and zfs_replay.c
178 uint32_t lr_attr_masksize; /* number of elements in array */
179 uint32_t lr_attr_bitmap; /* First entry of array */
180 /* remainder of array and any additional fields */
184 * log record for creates without optional ACL.
185 * This log record does support optional xvattr_t attributes.
188 lr_t lr_common; /* common portion of log record */
189 uint64_t lr_doid; /* object id of directory */
190 uint64_t lr_foid; /* object id of created file object */
191 uint64_t lr_mode; /* mode of object */
192 uint64_t lr_uid; /* uid of object */
193 uint64_t lr_gid; /* gid of object */
194 uint64_t lr_gen; /* generation (txg of creation) */
195 uint64_t lr_crtime[2]; /* creation time */
196 uint64_t lr_rdev; /* rdev of object to create */
197 /* name of object to create follows this */
198 /* for symlinks, link content follows name */
199 /* for creates with xvattr data, the name follows the xvattr info */
203 * FUID ACL record will be an array of ACEs from the original ACL.
204 * If this array includes ephemeral IDs, the record will also include
205 * an array of log-specific FUIDs to replace the ephemeral IDs.
206 * Only one copy of each unique domain will be present, so the log-specific
207 * FUIDs will use an index into a compressed domain table. On replay this
208 * information will be used to construct real FUIDs (and bypass idmap,
209 * since it may not be available).
213 * Log record for creates with optional ACL
214 * This log record is also used for recording any FUID
215 * information needed for replaying the create. If the
216 * file doesn't have any actual ACEs then the lr_aclcnt
220 lr_create_t lr_create; /* common create portion */
221 uint64_t lr_aclcnt; /* number of ACEs in ACL */
222 uint64_t lr_domcnt; /* number of unique domains */
223 uint64_t lr_fuidcnt; /* number of real fuids */
224 uint64_t lr_acl_bytes; /* number of bytes in ACL */
225 uint64_t lr_acl_flags; /* ACL flags */
226 /* lr_acl_bytes number of variable sized ace's follows */
227 /* if create is also setting xvattr's, then acl data follows xvattr */
228 /* if ACE FUIDs are needed then they will follow the xvattr_t */
229 /* Following the FUIDs will be the domain table information. */
230 /* The FUIDs for the owner and group will be in the lr_create */
231 /* portion of the record. */
232 /* name follows ACL data */
236 lr_t lr_common; /* common portion of log record */
237 uint64_t lr_doid; /* obj id of directory */
238 /* name of object to remove follows this */
242 lr_t lr_common; /* common portion of log record */
243 uint64_t lr_doid; /* obj id of directory */
244 uint64_t lr_link_obj; /* obj id of link */
245 /* name of object to link follows this */
249 lr_t lr_common; /* common portion of log record */
250 uint64_t lr_sdoid; /* obj id of source directory */
251 uint64_t lr_tdoid; /* obj id of target directory */
252 /* 2 strings: names of source and destination follow this */
256 lr_t lr_common; /* common portion of log record */
257 uint64_t lr_foid; /* file object to write */
258 uint64_t lr_offset; /* offset to write to */
259 uint64_t lr_length; /* user data length to write */
260 uint64_t lr_blkoff; /* offset represented by lr_blkptr */
261 blkptr_t lr_blkptr; /* spa block pointer for replay */
262 /* write data will follow for small writes */
266 lr_t lr_common; /* common portion of log record */
267 uint64_t lr_foid; /* object id of file to truncate */
268 uint64_t lr_offset; /* offset to truncate from */
269 uint64_t lr_length; /* length to truncate */
273 lr_t lr_common; /* common portion of log record */
274 uint64_t lr_foid; /* file object to change attributes */
275 uint64_t lr_mask; /* mask of attributes to set */
276 uint64_t lr_mode; /* mode to set */
277 uint64_t lr_uid; /* uid to set */
278 uint64_t lr_gid; /* gid to set */
279 uint64_t lr_size; /* size to set */
280 uint64_t lr_atime[2]; /* access time */
281 uint64_t lr_mtime[2]; /* modification time */
282 /* optional attribute lr_attr_t may be here */
286 lr_t lr_common; /* common portion of log record */
287 uint64_t lr_foid; /* obj id of file */
288 uint64_t lr_aclcnt; /* number of acl entries */
289 /* lr_aclcnt number of ace_t entries follow this */
293 lr_t lr_common; /* common portion of log record */
294 uint64_t lr_foid; /* obj id of file */
295 uint64_t lr_aclcnt; /* number of ACEs in ACL */
296 uint64_t lr_domcnt; /* number of unique domains */
297 uint64_t lr_fuidcnt; /* number of real fuids */
298 uint64_t lr_acl_bytes; /* number of bytes in ACL */
299 uint64_t lr_acl_flags; /* ACL flags */
300 /* lr_acl_bytes number of variable sized ace's follows */
304 * ZIL structure definitions, interface function prototype and globals.
308 * Writes are handled in three different ways:
311 * In this mode, if we need to commit the write later, then the block
312 * is immediately written into the file system (using dmu_sync),
313 * and a pointer to the block is put into the log record.
314 * When the txg commits the block is linked in.
315 * This saves additionally writing the data into the log record.
316 * There are a few requirements for this to occur:
317 * - write is greater than zfs/zvol_immediate_write_sz
318 * - not using slogs (as slogs are assumed to always be faster
319 * than writing into the main pool)
320 * - the write occupies only one block
322 * If we know we'll immediately be committing the
323 * transaction (FSYNC or FDSYNC), the we allocate a larger
324 * log record here for the data and copy the data in.
326 * Otherwise we don't allocate a buffer, and *if* we need to
327 * flush the write later then a buffer is allocated and
328 * we retrieve the data using the dmu.
331 WR_INDIRECT, /* indirect - a large write (dmu_sync() data */
332 /* and put blkptr in log, rather than actual data) */
333 WR_COPIED, /* immediate - data is copied into lr_write_t */
334 WR_NEED_COPY, /* immediate - data needs to be copied if pushed */
338 list_node_t itx_node; /* linkage on zl_itx_list */
339 void *itx_private; /* type-specific opaque data */
340 itx_wr_state_t itx_wr_state; /* write state */
341 uint8_t itx_sync; /* synchronous transaction */
342 uint64_t itx_sod; /* record size on disk */
343 lr_t itx_lr; /* common part of log record */
344 /* followed by type-specific part of lr_xx_t and its immediate data */
349 * zgd_t is passed through dmu_sync() to the callback routine zfs_get_done()
350 * to handle the cleanup of the dmu_sync() buffer write
353 zilog_t *zgd_zilog; /* zilog */
354 blkptr_t *zgd_bp; /* block pointer */
355 struct rl *zgd_rl; /* range lock */
359 typedef void zil_parse_blk_func_t(zilog_t *zilog, blkptr_t *bp, void *arg,
361 typedef void zil_parse_lr_func_t(zilog_t *zilog, lr_t *lr, void *arg,
363 typedef int zil_replay_func_t();
364 typedef int zil_get_data_t(void *arg, lr_write_t *lr, char *dbuf, zio_t *zio);
366 extern uint64_t zil_parse(zilog_t *zilog, zil_parse_blk_func_t *parse_blk_func,
367 zil_parse_lr_func_t *parse_lr_func, void *arg, uint64_t txg);
369 extern void zil_init(void);
370 extern void zil_fini(void);
372 extern zilog_t *zil_alloc(objset_t *os, zil_header_t *zh_phys);
373 extern void zil_free(zilog_t *zilog);
375 extern zilog_t *zil_open(objset_t *os, zil_get_data_t *get_data);
376 extern void zil_close(zilog_t *zilog);
378 extern void zil_replay(objset_t *os, void *arg,
379 zil_replay_func_t *replay_func[TX_MAX_TYPE]);
380 extern void zil_destroy(zilog_t *zilog, boolean_t keep_first);
381 extern void zil_rollback_destroy(zilog_t *zilog, dmu_tx_t *tx);
383 extern itx_t *zil_itx_create(uint64_t txtype, size_t lrsize);
384 extern uint64_t zil_itx_assign(zilog_t *zilog, itx_t *itx, dmu_tx_t *tx);
386 extern void zil_commit(zilog_t *zilog, uint64_t seq, uint64_t oid);
388 extern int zil_vdev_offline(char *osname, void *txarg);
389 extern int zil_claim(char *osname, void *txarg);
390 extern int zil_check_log_chain(char *osname, void *txarg);
391 extern void zil_sync(zilog_t *zilog, dmu_tx_t *tx);
392 extern void zil_clean(zilog_t *zilog);
393 extern int zil_is_committed(zilog_t *zilog);
395 extern int zil_suspend(zilog_t *zilog);
396 extern void zil_resume(zilog_t *zilog);
398 extern void zil_add_block(zilog_t *zilog, blkptr_t *bp);
400 extern int zil_disable;
406 #endif /* _SYS_ZIL_H */