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 2007 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 #pragma ident "%Z%%M% %I% %E% SMI" 30 31 #include <sys/types.h> 32 #include <sys/spa.h> 33 #include <sys/zio.h> 34 #include <sys/dmu.h> 35 36 #ifdef __cplusplus 37 extern "C" { 38 #endif 39 40 /* 41 * Intent log format: 42 * 43 * Each objset has its own intent log. The log header (zil_header_t) 44 * for objset N's intent log is kept in the Nth object of the SPA's 45 * intent_log objset. The log header points to a chain of log blocks, 46 * each of which contains log records (i.e., transactions) followed by 47 * a log block trailer (zil_trailer_t). The format of a log record 48 * depends on the record (or transaction) type, but all records begin 49 * with a common structure that defines the type, length, and txg. 50 */ 51 52 /* 53 * Intent log header - this on disk structure holds fields to manage 54 * the log. All fields are 64 bit to easily handle cross architectures. 55 */ 56 typedef struct zil_header { 57 uint64_t zh_claim_txg; /* txg in which log blocks were claimed */ 58 uint64_t zh_replay_seq; /* highest replayed sequence number */ 59 blkptr_t zh_log; /* log chain */ 60 uint64_t zh_claim_seq; /* highest claimed sequence number */ 61 uint64_t zh_pad[5]; 62 } zil_header_t; 63 64 /* 65 * Log block trailer - structure at the end of the header and each log block 66 * 67 * The zit_bt contains a zbt_cksum which for the intent log is 68 * the sequence number of this log block. A seq of 0 is invalid. 69 * The zbt_cksum is checked by the SPA against the sequence 70 * number passed in the blk_cksum field of the blkptr_t 71 */ 72 typedef struct zil_trailer { 73 uint64_t zit_pad; 74 blkptr_t zit_next_blk; /* next block in chain */ 75 uint64_t zit_nused; /* bytes in log block used */ 76 zio_block_tail_t zit_bt; /* block trailer */ 77 } zil_trailer_t; 78 79 #define ZIL_MIN_BLKSZ 4096ULL 80 #define ZIL_MAX_BLKSZ SPA_MAXBLOCKSIZE 81 #define ZIL_BLK_DATA_SZ(lwb) ((lwb)->lwb_sz - sizeof (zil_trailer_t)) 82 83 /* 84 * The words of a log block checksum. 85 */ 86 #define ZIL_ZC_GUID_0 0 87 #define ZIL_ZC_GUID_1 1 88 #define ZIL_ZC_OBJSET 2 89 #define ZIL_ZC_SEQ 3 90 91 typedef enum zil_create { 92 Z_FILE, 93 Z_DIR, 94 Z_XATTRDIR, 95 } zil_create_t; 96 97 /* 98 * size of xvattr log section. 99 * its composed of lr_attr_t + xvattr bitmap + 2 64 bit timestamps 100 * for create time and a single 64 bit integer for all of the attributes, 101 * and 4 64 bit integers (32 bytes) for the scanstamp. 102 * 103 */ 104 105 #define ZIL_XVAT_SIZE(mapsize) \ 106 sizeof (lr_attr_t) + (sizeof (uint32_t) * (mapsize - 1)) + \ 107 (sizeof (uint64_t) * 7) 108 /* 109 * Intent log transaction types and record structures 110 */ 111 #define TX_CREATE 1 /* Create file */ 112 #define TX_MKDIR 2 /* Make directory */ 113 #define TX_MKXATTR 3 /* Make XATTR directory */ 114 #define TX_SYMLINK 4 /* Create symbolic link to a file */ 115 #define TX_REMOVE 5 /* Remove file */ 116 #define TX_RMDIR 6 /* Remove directory */ 117 #define TX_LINK 7 /* Create hard link to a file */ 118 #define TX_RENAME 8 /* Rename a file */ 119 #define TX_WRITE 9 /* File write */ 120 #define TX_TRUNCATE 10 /* Truncate a file */ 121 #define TX_SETATTR 11 /* Set file attributes */ 122 #define TX_ACL_V0 12 /* Set old formatted ACL */ 123 #define TX_ACL 13 /* Set ACL */ 124 #define TX_CREATE_ACL 14 /* create with ACL */ 125 #define TX_CREATE_ATTR 15 /* create + attrs */ 126 #define TX_CREATE_ACL_ATTR 16 /* create with ACL + attrs */ 127 #define TX_MKDIR_ACL 17 /* mkdir with ACL */ 128 #define TX_MKDIR_ATTR 18 /* mkdir with attr */ 129 #define TX_MKDIR_ACL_ATTR 19 /* mkdir with ACL + attrs */ 130 #define TX_MAX_TYPE 20 /* Max transaction type */ 131 132 /* 133 * The transactions for mkdir, symlink, remove, rmdir, link, and rename 134 * may have the following bit set, indicating the original request 135 * specified case-insensitive handling of names. 136 */ 137 #define TX_CI ((uint64_t)0x1 << 63) /* case-insensitive behavior requested */ 138 139 /* 140 * Format of log records. 141 * The fields are carefully defined to allow them to be aligned 142 * and sized the same on sparc & intel architectures. 143 * Each log record has a common structure at the beginning. 144 * 145 * Note, lrc_seq holds two different sequence numbers. Whilst in memory 146 * it contains the transaction sequence number. The log record on 147 * disk holds the sequence number of all log records which is used to 148 * ensure we don't replay the same record. The two sequence numbers are 149 * different because the transactions can now be pushed out of order. 150 */ 151 typedef struct { /* common log record header */ 152 uint64_t lrc_txtype; /* intent log transaction type */ 153 uint64_t lrc_reclen; /* transaction record length */ 154 uint64_t lrc_txg; /* dmu transaction group number */ 155 uint64_t lrc_seq; /* see comment above */ 156 } lr_t; 157 158 /* 159 * Handle option extended vattr attributes. 160 * 161 * Whenever new attributes are added the version number 162 * will need to be updated as will code in 163 * zfs_log.c and zfs_replay.c 164 */ 165 typedef struct { 166 uint32_t lr_attr_masksize; /* number of elements in array */ 167 uint32_t lr_attr_bitmap; /* First entry of array */ 168 /* remainder of array and any additional fields */ 169 } lr_attr_t; 170 171 /* 172 * log record for creates without optional ACL. 173 * This log record does support optional xvattr_t attributes. 174 */ 175 typedef struct { 176 lr_t lr_common; /* common portion of log record */ 177 uint64_t lr_doid; /* object id of directory */ 178 uint64_t lr_foid; /* object id of created file object */ 179 uint64_t lr_mode; /* mode of object */ 180 uint64_t lr_uid; /* uid of object */ 181 uint64_t lr_gid; /* gid of object */ 182 uint64_t lr_gen; /* generation (txg of creation) */ 183 uint64_t lr_crtime[2]; /* creation time */ 184 uint64_t lr_rdev; /* rdev of object to create */ 185 /* name of object to create follows this */ 186 /* for symlinks, link content follows name */ 187 /* for creates with xvattr data, the name follows the xvattr info */ 188 } lr_create_t; 189 190 /* 191 * FUID ACL record will be an array of ACEs from the original ACL. 192 * If this array includes ephemeral IDs, the record will also include 193 * an array of log-specific FUIDs to replace the ephemeral IDs. 194 * Only one copy of each unique domain will be present, so the log-specific 195 * FUIDs will use an index into a compressed domain table. On replay this 196 * information will be used to construct real FUIDs (and bypass idmap, 197 * since it may not be available). 198 */ 199 200 /* 201 * Log record for creates with optional ACL 202 * This log record is also used for recording any FUID 203 * information needed for replaying the create. If the 204 * file doesn't have any actual ACEs then the lr_aclcnt 205 * would be zero. 206 */ 207 typedef struct { 208 lr_create_t lr_create; /* common create portion */ 209 uint64_t lr_aclcnt; /* number of ACEs in ACL */ 210 uint64_t lr_domcnt; /* number of unique domains */ 211 uint64_t lr_fuidcnt; /* number of real fuids */ 212 uint64_t lr_acl_bytes; /* number of bytes in ACL */ 213 uint64_t lr_acl_flags; /* ACL flags */ 214 /* lr_acl_bytes number of variable sized ace's follows */ 215 /* if create is also setting xvattr's, then acl data follows xvattr */ 216 /* if ACE FUIDs are needed then they will follow the xvattr_t */ 217 /* Following the FUIDs will be the domain table information. */ 218 /* The FUIDs for the owner and group will be in the lr_create */ 219 /* portion of the record. */ 220 /* name follows ACL data */ 221 } lr_acl_create_t; 222 223 typedef struct { 224 lr_t lr_common; /* common portion of log record */ 225 uint64_t lr_doid; /* obj id of directory */ 226 /* name of object to remove follows this */ 227 } lr_remove_t; 228 229 typedef struct { 230 lr_t lr_common; /* common portion of log record */ 231 uint64_t lr_doid; /* obj id of directory */ 232 uint64_t lr_link_obj; /* obj id of link */ 233 /* name of object to link follows this */ 234 } lr_link_t; 235 236 typedef struct { 237 lr_t lr_common; /* common portion of log record */ 238 uint64_t lr_sdoid; /* obj id of source directory */ 239 uint64_t lr_tdoid; /* obj id of target directory */ 240 /* 2 strings: names of source and destination follow this */ 241 } lr_rename_t; 242 243 typedef struct { 244 lr_t lr_common; /* common portion of log record */ 245 uint64_t lr_foid; /* file object to write */ 246 uint64_t lr_offset; /* offset to write to */ 247 uint64_t lr_length; /* user data length to write */ 248 uint64_t lr_blkoff; /* offset represented by lr_blkptr */ 249 blkptr_t lr_blkptr; /* spa block pointer for replay */ 250 /* write data will follow for small writes */ 251 } lr_write_t; 252 253 typedef struct { 254 lr_t lr_common; /* common portion of log record */ 255 uint64_t lr_foid; /* object id of file to truncate */ 256 uint64_t lr_offset; /* offset to truncate from */ 257 uint64_t lr_length; /* length to truncate */ 258 } lr_truncate_t; 259 260 typedef struct { 261 lr_t lr_common; /* common portion of log record */ 262 uint64_t lr_foid; /* file object to change attributes */ 263 uint64_t lr_mask; /* mask of attributes to set */ 264 uint64_t lr_mode; /* mode to set */ 265 uint64_t lr_uid; /* uid to set */ 266 uint64_t lr_gid; /* gid to set */ 267 uint64_t lr_size; /* size to set */ 268 uint64_t lr_atime[2]; /* access time */ 269 uint64_t lr_mtime[2]; /* modification time */ 270 /* optional attribute lr_attr_t may be here */ 271 } lr_setattr_t; 272 273 typedef struct { 274 lr_t lr_common; /* common portion of log record */ 275 uint64_t lr_foid; /* obj id of file */ 276 uint64_t lr_aclcnt; /* number of acl entries */ 277 /* lr_aclcnt number of ace_t entries follow this */ 278 } lr_acl_v0_t; 279 280 typedef struct { 281 lr_t lr_common; /* common portion of log record */ 282 uint64_t lr_foid; /* obj id of file */ 283 uint64_t lr_aclcnt; /* number of ACEs in ACL */ 284 uint64_t lr_domcnt; /* number of unique domains */ 285 uint64_t lr_fuidcnt; /* number of real fuids */ 286 uint64_t lr_acl_bytes; /* number of bytes in ACL */ 287 uint64_t lr_acl_flags; /* ACL flags */ 288 /* lr_acl_bytes number of variable sized ace's follows */ 289 } lr_acl_t; 290 291 /* 292 * ZIL structure definitions, interface function prototype and globals. 293 */ 294 295 /* 296 * ZFS intent log transaction structure 297 */ 298 typedef enum { 299 WR_INDIRECT, /* indirect - a large write (dmu_sync() data */ 300 /* and put blkptr in log, rather than actual data) */ 301 WR_COPIED, /* immediate - data is copied into lr_write_t */ 302 WR_NEED_COPY, /* immediate - data needs to be copied if pushed */ 303 } itx_wr_state_t; 304 305 typedef struct itx { 306 list_node_t itx_node; /* linkage on zl_itx_list */ 307 void *itx_private; /* type-specific opaque data */ 308 itx_wr_state_t itx_wr_state; /* write state */ 309 uint8_t itx_sync; /* synchronous transaction */ 310 lr_t itx_lr; /* common part of log record */ 311 /* followed by type-specific part of lr_xx_t and its immediate data */ 312 } itx_t; 313 314 315 /* 316 * zgd_t is passed through dmu_sync() to the callback routine zfs_get_done() 317 * to handle the cleanup of the dmu_sync() buffer write 318 */ 319 typedef struct { 320 zilog_t *zgd_zilog; /* zilog */ 321 blkptr_t *zgd_bp; /* block pointer */ 322 struct rl *zgd_rl; /* range lock */ 323 } zgd_t; 324 325 326 typedef void zil_parse_blk_func_t(zilog_t *zilog, blkptr_t *bp, void *arg, 327 uint64_t txg); 328 typedef void zil_parse_lr_func_t(zilog_t *zilog, lr_t *lr, void *arg, 329 uint64_t txg); 330 typedef int zil_replay_func_t(); 331 typedef int zil_get_data_t(void *arg, lr_write_t *lr, char *dbuf, zio_t *zio); 332 333 extern uint64_t zil_parse(zilog_t *zilog, zil_parse_blk_func_t *parse_blk_func, 334 zil_parse_lr_func_t *parse_lr_func, void *arg, uint64_t txg); 335 336 extern void zil_init(void); 337 extern void zil_fini(void); 338 339 extern zilog_t *zil_alloc(objset_t *os, zil_header_t *zh_phys); 340 extern void zil_free(zilog_t *zilog); 341 342 extern zilog_t *zil_open(objset_t *os, zil_get_data_t *get_data); 343 extern void zil_close(zilog_t *zilog); 344 345 extern void zil_replay(objset_t *os, void *arg, uint64_t *txgp, 346 zil_replay_func_t *replay_func[TX_MAX_TYPE]); 347 extern void zil_destroy(zilog_t *zilog, boolean_t keep_first); 348 extern void zil_rollback_destroy(zilog_t *zilog, dmu_tx_t *tx); 349 350 extern itx_t *zil_itx_create(uint64_t txtype, size_t lrsize); 351 extern uint64_t zil_itx_assign(zilog_t *zilog, itx_t *itx, dmu_tx_t *tx); 352 353 extern void zil_commit(zilog_t *zilog, uint64_t seq, uint64_t oid); 354 355 extern int zil_claim(char *osname, void *txarg); 356 extern void zil_sync(zilog_t *zilog, dmu_tx_t *tx); 357 extern void zil_clean(zilog_t *zilog); 358 extern int zil_is_committed(zilog_t *zilog); 359 360 extern int zil_suspend(zilog_t *zilog); 361 extern void zil_resume(zilog_t *zilog); 362 363 extern void zil_add_vdev(zilog_t *zilog, uint64_t vdev); 364 365 extern int zil_disable; 366 367 #ifdef __cplusplus 368 } 369 #endif 370 371 #endif /* _SYS_ZIL_H */ 372