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 2010 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 #ifndef _SYS_ZIL_H 27 #define _SYS_ZIL_H 28 29 #include <sys/types.h> 30 #include <sys/spa.h> 31 #include <sys/zio.h> 32 #include <sys/dmu.h> 33 34 #ifdef __cplusplus 35 extern "C" { 36 #endif 37 38 /* 39 * Intent log format: 40 * 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. 48 */ 49 50 /* 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. 53 */ 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_blk_seq; /* highest claimed block sequence number */ 59 uint64_t zh_flags; /* header flags */ 60 uint64_t zh_claim_lr_seq; /* highest claimed lr sequence number */ 61 uint64_t zh_pad[3]; 62 } zil_header_t; 63 64 /* 65 * zh_flags bit settings 66 */ 67 #define ZIL_REPLAY_NEEDED 0x1 /* replay needed - internal only */ 68 #define ZIL_CLAIM_LR_SEQ_VALID 0x2 /* zh_claim_lr_seq field is valid */ 69 70 /* 71 * Log block chaining. 72 * 73 * Log blocks are chained together. Originally they were chained at the 74 * end of the block. For performance reasons the chain was moved to the 75 * beginning of the block which allows writes for only the data being used. 76 * The older position is supported for backwards compatability. 77 * 78 * The zio_eck_t contains a zec_cksum which for the intent log is 79 * the sequence number of this log block. A seq of 0 is invalid. 80 * The zec_cksum is checked by the SPA against the sequence 81 * number passed in the blk_cksum field of the blkptr_t 82 */ 83 typedef struct zil_chain { 84 uint64_t zc_pad; 85 blkptr_t zc_next_blk; /* next block in chain */ 86 uint64_t zc_nused; /* bytes in log block used */ 87 zio_eck_t zc_eck; /* block trailer */ 88 } zil_chain_t; 89 90 #define ZIL_MIN_BLKSZ 4096ULL 91 #define ZIL_MAX_BLKSZ SPA_MAXBLOCKSIZE 92 93 /* 94 * The words of a log block checksum. 95 */ 96 #define ZIL_ZC_GUID_0 0 97 #define ZIL_ZC_GUID_1 1 98 #define ZIL_ZC_OBJSET 2 99 #define ZIL_ZC_SEQ 3 100 101 typedef enum zil_create { 102 Z_FILE, 103 Z_DIR, 104 Z_XATTRDIR, 105 } zil_create_t; 106 107 /* 108 * size of xvattr log section. 109 * its composed of lr_attr_t + xvattr bitmap + 2 64 bit timestamps 110 * for create time and a single 64 bit integer for all of the attributes, 111 * and 4 64 bit integers (32 bytes) for the scanstamp. 112 * 113 */ 114 115 #define ZIL_XVAT_SIZE(mapsize) \ 116 sizeof (lr_attr_t) + (sizeof (uint32_t) * (mapsize - 1)) + \ 117 (sizeof (uint64_t) * 7) 118 119 /* 120 * Size of ACL in log. The ACE data is padded out to properly align 121 * on 8 byte boundary. 122 */ 123 124 #define ZIL_ACE_LENGTH(x) (roundup(x, sizeof (uint64_t))) 125 126 /* 127 * Intent log transaction types and record structures 128 */ 129 #define TX_CREATE 1 /* Create file */ 130 #define TX_MKDIR 2 /* Make directory */ 131 #define TX_MKXATTR 3 /* Make XATTR directory */ 132 #define TX_SYMLINK 4 /* Create symbolic link to a file */ 133 #define TX_REMOVE 5 /* Remove file */ 134 #define TX_RMDIR 6 /* Remove directory */ 135 #define TX_LINK 7 /* Create hard link to a file */ 136 #define TX_RENAME 8 /* Rename a file */ 137 #define TX_WRITE 9 /* File write */ 138 #define TX_TRUNCATE 10 /* Truncate a file */ 139 #define TX_SETATTR 11 /* Set file attributes */ 140 #define TX_ACL_V0 12 /* Set old formatted ACL */ 141 #define TX_ACL 13 /* Set ACL */ 142 #define TX_CREATE_ACL 14 /* create with ACL */ 143 #define TX_CREATE_ATTR 15 /* create + attrs */ 144 #define TX_CREATE_ACL_ATTR 16 /* create with ACL + attrs */ 145 #define TX_MKDIR_ACL 17 /* mkdir with ACL */ 146 #define TX_MKDIR_ATTR 18 /* mkdir with attr */ 147 #define TX_MKDIR_ACL_ATTR 19 /* mkdir with ACL + attrs */ 148 #define TX_WRITE2 20 /* dmu_sync EALREADY write */ 149 #define TX_MAX_TYPE 21 /* Max transaction type */ 150 151 /* 152 * The transactions for mkdir, symlink, remove, rmdir, link, and rename 153 * may have the following bit set, indicating the original request 154 * specified case-insensitive handling of names. 155 */ 156 #define TX_CI ((uint64_t)0x1 << 63) /* case-insensitive behavior requested */ 157 158 /* 159 * Transactions for write, truncate, setattr, acl_v0, and acl can be logged 160 * out of order. For convenience in the code, all such records must have 161 * lr_foid at the same offset. 162 */ 163 #define TX_OOO(txtype) \ 164 ((txtype) == TX_WRITE || \ 165 (txtype) == TX_TRUNCATE || \ 166 (txtype) == TX_SETATTR || \ 167 (txtype) == TX_ACL_V0 || \ 168 (txtype) == TX_ACL || \ 169 (txtype) == TX_WRITE2) 170 171 172 /* 173 * Format of log records. 174 * The fields are carefully defined to allow them to be aligned 175 * and sized the same on sparc & intel architectures. 176 * Each log record has a common structure at the beginning. 177 * 178 * Note, lrc_seq holds two different sequence numbers. Whilst in memory 179 * it contains the transaction sequence number. The log record on 180 * disk holds the sequence number of all log records which is used to 181 * ensure we don't replay the same record. The two sequence numbers are 182 * different because the transactions can now be pushed out of order. 183 */ 184 typedef struct { /* common log record header */ 185 uint64_t lrc_txtype; /* intent log transaction type */ 186 uint64_t lrc_reclen; /* transaction record length */ 187 uint64_t lrc_txg; /* dmu transaction group number */ 188 uint64_t lrc_seq; /* see comment above */ 189 } lr_t; 190 191 /* 192 * Common start of all out-of-order record types (TX_OOO() above). 193 */ 194 typedef struct { 195 lr_t lr_common; /* common portion of log record */ 196 uint64_t lr_foid; /* object id */ 197 } lr_ooo_t; 198 199 /* 200 * Handle option extended vattr attributes. 201 * 202 * Whenever new attributes are added the version number 203 * will need to be updated as will code in 204 * zfs_log.c and zfs_replay.c 205 */ 206 typedef struct { 207 uint32_t lr_attr_masksize; /* number of elements in array */ 208 uint32_t lr_attr_bitmap; /* First entry of array */ 209 /* remainder of array and any additional fields */ 210 } lr_attr_t; 211 212 /* 213 * log record for creates without optional ACL. 214 * This log record does support optional xvattr_t attributes. 215 */ 216 typedef struct { 217 lr_t lr_common; /* common portion of log record */ 218 uint64_t lr_doid; /* object id of directory */ 219 uint64_t lr_foid; /* object id of created file object */ 220 uint64_t lr_mode; /* mode of object */ 221 uint64_t lr_uid; /* uid of object */ 222 uint64_t lr_gid; /* gid of object */ 223 uint64_t lr_gen; /* generation (txg of creation) */ 224 uint64_t lr_crtime[2]; /* creation time */ 225 uint64_t lr_rdev; /* rdev of object to create */ 226 /* name of object to create follows this */ 227 /* for symlinks, link content follows name */ 228 /* for creates with xvattr data, the name follows the xvattr info */ 229 } lr_create_t; 230 231 /* 232 * FUID ACL record will be an array of ACEs from the original ACL. 233 * If this array includes ephemeral IDs, the record will also include 234 * an array of log-specific FUIDs to replace the ephemeral IDs. 235 * Only one copy of each unique domain will be present, so the log-specific 236 * FUIDs will use an index into a compressed domain table. On replay this 237 * information will be used to construct real FUIDs (and bypass idmap, 238 * since it may not be available). 239 */ 240 241 /* 242 * Log record for creates with optional ACL 243 * This log record is also used for recording any FUID 244 * information needed for replaying the create. If the 245 * file doesn't have any actual ACEs then the lr_aclcnt 246 * would be zero. 247 */ 248 typedef struct { 249 lr_create_t lr_create; /* common create portion */ 250 uint64_t lr_aclcnt; /* number of ACEs in ACL */ 251 uint64_t lr_domcnt; /* number of unique domains */ 252 uint64_t lr_fuidcnt; /* number of real fuids */ 253 uint64_t lr_acl_bytes; /* number of bytes in ACL */ 254 uint64_t lr_acl_flags; /* ACL flags */ 255 /* lr_acl_bytes number of variable sized ace's follows */ 256 /* if create is also setting xvattr's, then acl data follows xvattr */ 257 /* if ACE FUIDs are needed then they will follow the xvattr_t */ 258 /* Following the FUIDs will be the domain table information. */ 259 /* The FUIDs for the owner and group will be in the lr_create */ 260 /* portion of the record. */ 261 /* name follows ACL data */ 262 } lr_acl_create_t; 263 264 typedef struct { 265 lr_t lr_common; /* common portion of log record */ 266 uint64_t lr_doid; /* obj id of directory */ 267 /* name of object to remove follows this */ 268 } lr_remove_t; 269 270 typedef struct { 271 lr_t lr_common; /* common portion of log record */ 272 uint64_t lr_doid; /* obj id of directory */ 273 uint64_t lr_link_obj; /* obj id of link */ 274 /* name of object to link follows this */ 275 } lr_link_t; 276 277 typedef struct { 278 lr_t lr_common; /* common portion of log record */ 279 uint64_t lr_sdoid; /* obj id of source directory */ 280 uint64_t lr_tdoid; /* obj id of target directory */ 281 /* 2 strings: names of source and destination follow this */ 282 } lr_rename_t; 283 284 typedef struct { 285 lr_t lr_common; /* common portion of log record */ 286 uint64_t lr_foid; /* file object to write */ 287 uint64_t lr_offset; /* offset to write to */ 288 uint64_t lr_length; /* user data length to write */ 289 uint64_t lr_blkoff; /* no longer used */ 290 blkptr_t lr_blkptr; /* spa block pointer for replay */ 291 /* write data will follow for small writes */ 292 } lr_write_t; 293 294 typedef struct { 295 lr_t lr_common; /* common portion of log record */ 296 uint64_t lr_foid; /* object id of file to truncate */ 297 uint64_t lr_offset; /* offset to truncate from */ 298 uint64_t lr_length; /* length to truncate */ 299 } lr_truncate_t; 300 301 typedef struct { 302 lr_t lr_common; /* common portion of log record */ 303 uint64_t lr_foid; /* file object to change attributes */ 304 uint64_t lr_mask; /* mask of attributes to set */ 305 uint64_t lr_mode; /* mode to set */ 306 uint64_t lr_uid; /* uid to set */ 307 uint64_t lr_gid; /* gid to set */ 308 uint64_t lr_size; /* size to set */ 309 uint64_t lr_atime[2]; /* access time */ 310 uint64_t lr_mtime[2]; /* modification time */ 311 /* optional attribute lr_attr_t may be here */ 312 } lr_setattr_t; 313 314 typedef struct { 315 lr_t lr_common; /* common portion of log record */ 316 uint64_t lr_foid; /* obj id of file */ 317 uint64_t lr_aclcnt; /* number of acl entries */ 318 /* lr_aclcnt number of ace_t entries follow this */ 319 } lr_acl_v0_t; 320 321 typedef struct { 322 lr_t lr_common; /* common portion of log record */ 323 uint64_t lr_foid; /* obj id of file */ 324 uint64_t lr_aclcnt; /* number of ACEs in ACL */ 325 uint64_t lr_domcnt; /* number of unique domains */ 326 uint64_t lr_fuidcnt; /* number of real fuids */ 327 uint64_t lr_acl_bytes; /* number of bytes in ACL */ 328 uint64_t lr_acl_flags; /* ACL flags */ 329 /* lr_acl_bytes number of variable sized ace's follows */ 330 } lr_acl_t; 331 332 /* 333 * ZIL structure definitions, interface function prototype and globals. 334 */ 335 336 /* 337 * Writes are handled in three different ways: 338 * 339 * WR_INDIRECT: 340 * In this mode, if we need to commit the write later, then the block 341 * is immediately written into the file system (using dmu_sync), 342 * and a pointer to the block is put into the log record. 343 * When the txg commits the block is linked in. 344 * This saves additionally writing the data into the log record. 345 * There are a few requirements for this to occur: 346 * - write is greater than zfs/zvol_immediate_write_sz 347 * - not using slogs (as slogs are assumed to always be faster 348 * than writing into the main pool) 349 * - the write occupies only one block 350 * WR_COPIED: 351 * If we know we'll immediately be committing the 352 * transaction (FSYNC or FDSYNC), the we allocate a larger 353 * log record here for the data and copy the data in. 354 * WR_NEED_COPY: 355 * Otherwise we don't allocate a buffer, and *if* we need to 356 * flush the write later then a buffer is allocated and 357 * we retrieve the data using the dmu. 358 */ 359 typedef enum { 360 WR_INDIRECT, /* indirect - a large write (dmu_sync() data */ 361 /* and put blkptr in log, rather than actual data) */ 362 WR_COPIED, /* immediate - data is copied into lr_write_t */ 363 WR_NEED_COPY, /* immediate - data needs to be copied if pushed */ 364 WR_NUM_STATES /* number of states */ 365 } itx_wr_state_t; 366 367 typedef struct itx { 368 list_node_t itx_node; /* linkage on zl_itx_list */ 369 void *itx_private; /* type-specific opaque data */ 370 itx_wr_state_t itx_wr_state; /* write state */ 371 uint8_t itx_sync; /* synchronous transaction */ 372 uint64_t itx_sod; /* record size on disk */ 373 lr_t itx_lr; /* common part of log record */ 374 /* followed by type-specific part of lr_xx_t and its immediate data */ 375 } itx_t; 376 377 typedef int zil_parse_blk_func_t(zilog_t *zilog, blkptr_t *bp, void *arg, 378 uint64_t txg); 379 typedef int zil_parse_lr_func_t(zilog_t *zilog, lr_t *lr, void *arg, 380 uint64_t txg); 381 typedef int zil_replay_func_t(); 382 typedef int zil_get_data_t(void *arg, lr_write_t *lr, char *dbuf, zio_t *zio); 383 384 extern int zil_parse(zilog_t *zilog, zil_parse_blk_func_t *parse_blk_func, 385 zil_parse_lr_func_t *parse_lr_func, void *arg, uint64_t txg); 386 387 extern void zil_init(void); 388 extern void zil_fini(void); 389 390 extern zilog_t *zil_alloc(objset_t *os, zil_header_t *zh_phys); 391 extern void zil_free(zilog_t *zilog); 392 393 extern zilog_t *zil_open(objset_t *os, zil_get_data_t *get_data); 394 extern void zil_close(zilog_t *zilog); 395 396 extern void zil_replay(objset_t *os, void *arg, 397 zil_replay_func_t *replay_func[TX_MAX_TYPE]); 398 extern boolean_t zil_replaying(zilog_t *zilog, dmu_tx_t *tx); 399 extern void zil_destroy(zilog_t *zilog, boolean_t keep_first); 400 extern void zil_rollback_destroy(zilog_t *zilog, dmu_tx_t *tx); 401 402 extern itx_t *zil_itx_create(uint64_t txtype, size_t lrsize); 403 extern void zil_itx_destroy(itx_t *itx); 404 extern uint64_t zil_itx_assign(zilog_t *zilog, itx_t *itx, dmu_tx_t *tx); 405 406 extern void zil_commit(zilog_t *zilog, uint64_t seq, uint64_t oid); 407 408 extern int zil_vdev_offline(const char *osname, void *txarg); 409 extern int zil_claim(const char *osname, void *txarg); 410 extern int zil_check_log_chain(const char *osname, void *txarg); 411 extern void zil_sync(zilog_t *zilog, dmu_tx_t *tx); 412 extern void zil_clean(zilog_t *zilog); 413 414 extern int zil_suspend(zilog_t *zilog); 415 extern void zil_resume(zilog_t *zilog); 416 417 extern void zil_add_block(zilog_t *zilog, const blkptr_t *bp); 418 extern int zil_bp_tree_add(zilog_t *zilog, const blkptr_t *bp); 419 420 extern void zil_set_logbias(zilog_t *zilog, uint64_t slogval); 421 422 extern int zil_disable; 423 424 #ifdef __cplusplus 425 } 426 #endif 427 428 #endif /* _SYS_ZIL_H */ 429