1 /* 2 * ext4.h 3 * 4 * Copyright (C) 1992, 1993, 1994, 1995 5 * Remy Card (card@masi.ibp.fr) 6 * Laboratoire MASI - Institut Blaise Pascal 7 * Universite Pierre et Marie Curie (Paris VI) 8 * 9 * from 10 * 11 * linux/include/linux/minix_fs.h 12 * 13 * Copyright (C) 1991, 1992 Linus Torvalds 14 */ 15 16 #ifndef _EXT4_H 17 #define _EXT4_H 18 19 #include <linux/types.h> 20 #include <linux/blkdev.h> 21 #include <linux/magic.h> 22 #include <linux/jbd2.h> 23 #include <linux/quota.h> 24 #include <linux/rwsem.h> 25 #include <linux/rbtree.h> 26 #include <linux/seqlock.h> 27 #include <linux/mutex.h> 28 #include <linux/timer.h> 29 #include <linux/wait.h> 30 #include <linux/blockgroup_lock.h> 31 #include <linux/percpu_counter.h> 32 #ifdef __KERNEL__ 33 #include <linux/compat.h> 34 #endif 35 36 /* 37 * The fourth extended filesystem constants/structures 38 */ 39 40 /* 41 * Define EXT4FS_DEBUG to produce debug messages 42 */ 43 #undef EXT4FS_DEBUG 44 45 /* 46 * Debug code 47 */ 48 #ifdef EXT4FS_DEBUG 49 #define ext4_debug(f, a...) \ 50 do { \ 51 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:", \ 52 __FILE__, __LINE__, __func__); \ 53 printk(KERN_DEBUG f, ## a); \ 54 } while (0) 55 #else 56 #define ext4_debug(fmt, ...) no_printk(fmt, ##__VA_ARGS__) 57 #endif 58 59 #define EXT4_ERROR_INODE(inode, fmt, a...) \ 60 ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a) 61 62 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...) \ 63 ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a) 64 65 #define EXT4_ERROR_FILE(file, block, fmt, a...) \ 66 ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a) 67 68 /* data type for block offset of block group */ 69 typedef int ext4_grpblk_t; 70 71 /* data type for filesystem-wide blocks number */ 72 typedef unsigned long long ext4_fsblk_t; 73 74 /* data type for file logical block number */ 75 typedef __u32 ext4_lblk_t; 76 77 /* data type for block group number */ 78 typedef unsigned int ext4_group_t; 79 80 /* 81 * Flags used in mballoc's allocation_context flags field. 82 * 83 * Also used to show what's going on for debugging purposes when the 84 * flag field is exported via the traceport interface 85 */ 86 87 /* prefer goal again. length */ 88 #define EXT4_MB_HINT_MERGE 0x0001 89 /* blocks already reserved */ 90 #define EXT4_MB_HINT_RESERVED 0x0002 91 /* metadata is being allocated */ 92 #define EXT4_MB_HINT_METADATA 0x0004 93 /* first blocks in the file */ 94 #define EXT4_MB_HINT_FIRST 0x0008 95 /* search for the best chunk */ 96 #define EXT4_MB_HINT_BEST 0x0010 97 /* data is being allocated */ 98 #define EXT4_MB_HINT_DATA 0x0020 99 /* don't preallocate (for tails) */ 100 #define EXT4_MB_HINT_NOPREALLOC 0x0040 101 /* allocate for locality group */ 102 #define EXT4_MB_HINT_GROUP_ALLOC 0x0080 103 /* allocate goal blocks or none */ 104 #define EXT4_MB_HINT_GOAL_ONLY 0x0100 105 /* goal is meaningful */ 106 #define EXT4_MB_HINT_TRY_GOAL 0x0200 107 /* blocks already pre-reserved by delayed allocation */ 108 #define EXT4_MB_DELALLOC_RESERVED 0x0400 109 /* We are doing stream allocation */ 110 #define EXT4_MB_STREAM_ALLOC 0x0800 111 /* Use reserved root blocks if needed */ 112 #define EXT4_MB_USE_ROOT_BLOCKS 0x1000 113 114 struct ext4_allocation_request { 115 /* target inode for block we're allocating */ 116 struct inode *inode; 117 /* how many blocks we want to allocate */ 118 unsigned int len; 119 /* logical block in target inode */ 120 ext4_lblk_t logical; 121 /* the closest logical allocated block to the left */ 122 ext4_lblk_t lleft; 123 /* the closest logical allocated block to the right */ 124 ext4_lblk_t lright; 125 /* phys. target (a hint) */ 126 ext4_fsblk_t goal; 127 /* phys. block for the closest logical allocated block to the left */ 128 ext4_fsblk_t pleft; 129 /* phys. block for the closest logical allocated block to the right */ 130 ext4_fsblk_t pright; 131 /* flags. see above EXT4_MB_HINT_* */ 132 unsigned int flags; 133 }; 134 135 /* 136 * Logical to physical block mapping, used by ext4_map_blocks() 137 * 138 * This structure is used to pass requests into ext4_map_blocks() as 139 * well as to store the information returned by ext4_map_blocks(). It 140 * takes less room on the stack than a struct buffer_head. 141 */ 142 #define EXT4_MAP_NEW (1 << BH_New) 143 #define EXT4_MAP_MAPPED (1 << BH_Mapped) 144 #define EXT4_MAP_UNWRITTEN (1 << BH_Unwritten) 145 #define EXT4_MAP_BOUNDARY (1 << BH_Boundary) 146 #define EXT4_MAP_UNINIT (1 << BH_Uninit) 147 /* Sometimes (in the bigalloc case, from ext4_da_get_block_prep) the caller of 148 * ext4_map_blocks wants to know whether or not the underlying cluster has 149 * already been accounted for. EXT4_MAP_FROM_CLUSTER conveys to the caller that 150 * the requested mapping was from previously mapped (or delayed allocated) 151 * cluster. We use BH_AllocFromCluster only for this flag. BH_AllocFromCluster 152 * should never appear on buffer_head's state flags. 153 */ 154 #define EXT4_MAP_FROM_CLUSTER (1 << BH_AllocFromCluster) 155 #define EXT4_MAP_FLAGS (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\ 156 EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\ 157 EXT4_MAP_UNINIT | EXT4_MAP_FROM_CLUSTER) 158 159 struct ext4_map_blocks { 160 ext4_fsblk_t m_pblk; 161 ext4_lblk_t m_lblk; 162 unsigned int m_len; 163 unsigned int m_flags; 164 }; 165 166 /* 167 * For delayed allocation tracking 168 */ 169 struct mpage_da_data { 170 struct inode *inode; 171 sector_t b_blocknr; /* start block number of extent */ 172 size_t b_size; /* size of extent */ 173 unsigned long b_state; /* state of the extent */ 174 unsigned long first_page, next_page; /* extent of pages */ 175 struct writeback_control *wbc; 176 int io_done; 177 int pages_written; 178 int retval; 179 }; 180 181 /* 182 * Flags for ext4_io_end->flags 183 */ 184 #define EXT4_IO_END_UNWRITTEN 0x0001 185 #define EXT4_IO_END_ERROR 0x0002 186 #define EXT4_IO_END_QUEUED 0x0004 187 #define EXT4_IO_END_DIRECT 0x0008 188 #define EXT4_IO_END_IN_FSYNC 0x0010 189 190 struct ext4_io_page { 191 struct page *p_page; 192 atomic_t p_count; 193 }; 194 195 #define MAX_IO_PAGES 128 196 197 /* 198 * For converting uninitialized extents on a work queue. 199 * 200 * 'page' is only used from the writepage() path; 'pages' is only used for 201 * buffered writes; they are used to keep page references until conversion 202 * takes place. For AIO/DIO, neither field is filled in. 203 */ 204 typedef struct ext4_io_end { 205 struct list_head list; /* per-file finished IO list */ 206 struct inode *inode; /* file being written to */ 207 unsigned int flag; /* unwritten or not */ 208 struct page *page; /* for writepage() path */ 209 loff_t offset; /* offset in the file */ 210 ssize_t size; /* size of the extent */ 211 struct work_struct work; /* data work queue */ 212 struct kiocb *iocb; /* iocb struct for AIO */ 213 int result; /* error value for AIO */ 214 int num_io_pages; /* for writepages() */ 215 struct ext4_io_page *pages[MAX_IO_PAGES]; /* for writepages() */ 216 } ext4_io_end_t; 217 218 struct ext4_io_submit { 219 int io_op; 220 struct bio *io_bio; 221 ext4_io_end_t *io_end; 222 struct ext4_io_page *io_page; 223 sector_t io_next_block; 224 }; 225 226 /* 227 * Special inodes numbers 228 */ 229 #define EXT4_BAD_INO 1 /* Bad blocks inode */ 230 #define EXT4_ROOT_INO 2 /* Root inode */ 231 #define EXT4_USR_QUOTA_INO 3 /* User quota inode */ 232 #define EXT4_GRP_QUOTA_INO 4 /* Group quota inode */ 233 #define EXT4_BOOT_LOADER_INO 5 /* Boot loader inode */ 234 #define EXT4_UNDEL_DIR_INO 6 /* Undelete directory inode */ 235 #define EXT4_RESIZE_INO 7 /* Reserved group descriptors inode */ 236 #define EXT4_JOURNAL_INO 8 /* Journal inode */ 237 238 /* First non-reserved inode for old ext4 filesystems */ 239 #define EXT4_GOOD_OLD_FIRST_INO 11 240 241 /* 242 * Maximal count of links to a file 243 */ 244 #define EXT4_LINK_MAX 65000 245 246 /* 247 * Macro-instructions used to manage several block sizes 248 */ 249 #define EXT4_MIN_BLOCK_SIZE 1024 250 #define EXT4_MAX_BLOCK_SIZE 65536 251 #define EXT4_MIN_BLOCK_LOG_SIZE 10 252 #define EXT4_MAX_BLOCK_LOG_SIZE 16 253 #ifdef __KERNEL__ 254 # define EXT4_BLOCK_SIZE(s) ((s)->s_blocksize) 255 #else 256 # define EXT4_BLOCK_SIZE(s) (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size) 257 #endif 258 #define EXT4_ADDR_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / sizeof(__u32)) 259 #define EXT4_CLUSTER_SIZE(s) (EXT4_BLOCK_SIZE(s) << \ 260 EXT4_SB(s)->s_cluster_bits) 261 #ifdef __KERNEL__ 262 # define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits) 263 # define EXT4_CLUSTER_BITS(s) (EXT4_SB(s)->s_cluster_bits) 264 #else 265 # define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_log_block_size + 10) 266 #endif 267 #ifdef __KERNEL__ 268 #define EXT4_ADDR_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_addr_per_block_bits) 269 #define EXT4_INODE_SIZE(s) (EXT4_SB(s)->s_inode_size) 270 #define EXT4_FIRST_INO(s) (EXT4_SB(s)->s_first_ino) 271 #else 272 #define EXT4_INODE_SIZE(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \ 273 EXT4_GOOD_OLD_INODE_SIZE : \ 274 (s)->s_inode_size) 275 #define EXT4_FIRST_INO(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \ 276 EXT4_GOOD_OLD_FIRST_INO : \ 277 (s)->s_first_ino) 278 #endif 279 #define EXT4_BLOCK_ALIGN(size, blkbits) ALIGN((size), (1 << (blkbits))) 280 281 /* Translate a block number to a cluster number */ 282 #define EXT4_B2C(sbi, blk) ((blk) >> (sbi)->s_cluster_bits) 283 /* Translate a cluster number to a block number */ 284 #define EXT4_C2B(sbi, cluster) ((cluster) << (sbi)->s_cluster_bits) 285 /* Translate # of blks to # of clusters */ 286 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \ 287 (sbi)->s_cluster_bits) 288 289 /* 290 * Structure of a blocks group descriptor 291 */ 292 struct ext4_group_desc 293 { 294 __le32 bg_block_bitmap_lo; /* Blocks bitmap block */ 295 __le32 bg_inode_bitmap_lo; /* Inodes bitmap block */ 296 __le32 bg_inode_table_lo; /* Inodes table block */ 297 __le16 bg_free_blocks_count_lo;/* Free blocks count */ 298 __le16 bg_free_inodes_count_lo;/* Free inodes count */ 299 __le16 bg_used_dirs_count_lo; /* Directories count */ 300 __le16 bg_flags; /* EXT4_BG_flags (INODE_UNINIT, etc) */ 301 __le32 bg_exclude_bitmap_lo; /* Exclude bitmap for snapshots */ 302 __le16 bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */ 303 __le16 bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */ 304 __le16 bg_itable_unused_lo; /* Unused inodes count */ 305 __le16 bg_checksum; /* crc16(sb_uuid+group+desc) */ 306 __le32 bg_block_bitmap_hi; /* Blocks bitmap block MSB */ 307 __le32 bg_inode_bitmap_hi; /* Inodes bitmap block MSB */ 308 __le32 bg_inode_table_hi; /* Inodes table block MSB */ 309 __le16 bg_free_blocks_count_hi;/* Free blocks count MSB */ 310 __le16 bg_free_inodes_count_hi;/* Free inodes count MSB */ 311 __le16 bg_used_dirs_count_hi; /* Directories count MSB */ 312 __le16 bg_itable_unused_hi; /* Unused inodes count MSB */ 313 __le32 bg_exclude_bitmap_hi; /* Exclude bitmap block MSB */ 314 __le16 bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */ 315 __le16 bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */ 316 __u32 bg_reserved; 317 }; 318 319 /* 320 * Structure of a flex block group info 321 */ 322 323 struct flex_groups { 324 atomic_t free_inodes; 325 atomic_t free_clusters; 326 atomic_t used_dirs; 327 }; 328 329 #define EXT4_BG_INODE_UNINIT 0x0001 /* Inode table/bitmap not in use */ 330 #define EXT4_BG_BLOCK_UNINIT 0x0002 /* Block bitmap not in use */ 331 #define EXT4_BG_INODE_ZEROED 0x0004 /* On-disk itable initialized to zero */ 332 333 /* 334 * Macro-instructions used to manage group descriptors 335 */ 336 #define EXT4_MIN_DESC_SIZE 32 337 #define EXT4_MIN_DESC_SIZE_64BIT 64 338 #define EXT4_MAX_DESC_SIZE EXT4_MIN_BLOCK_SIZE 339 #define EXT4_DESC_SIZE(s) (EXT4_SB(s)->s_desc_size) 340 #ifdef __KERNEL__ 341 # define EXT4_BLOCKS_PER_GROUP(s) (EXT4_SB(s)->s_blocks_per_group) 342 # define EXT4_CLUSTERS_PER_GROUP(s) (EXT4_SB(s)->s_clusters_per_group) 343 # define EXT4_DESC_PER_BLOCK(s) (EXT4_SB(s)->s_desc_per_block) 344 # define EXT4_INODES_PER_GROUP(s) (EXT4_SB(s)->s_inodes_per_group) 345 # define EXT4_DESC_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_desc_per_block_bits) 346 #else 347 # define EXT4_BLOCKS_PER_GROUP(s) ((s)->s_blocks_per_group) 348 # define EXT4_DESC_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s)) 349 # define EXT4_INODES_PER_GROUP(s) ((s)->s_inodes_per_group) 350 #endif 351 352 /* 353 * Constants relative to the data blocks 354 */ 355 #define EXT4_NDIR_BLOCKS 12 356 #define EXT4_IND_BLOCK EXT4_NDIR_BLOCKS 357 #define EXT4_DIND_BLOCK (EXT4_IND_BLOCK + 1) 358 #define EXT4_TIND_BLOCK (EXT4_DIND_BLOCK + 1) 359 #define EXT4_N_BLOCKS (EXT4_TIND_BLOCK + 1) 360 361 /* 362 * Inode flags 363 */ 364 #define EXT4_SECRM_FL 0x00000001 /* Secure deletion */ 365 #define EXT4_UNRM_FL 0x00000002 /* Undelete */ 366 #define EXT4_COMPR_FL 0x00000004 /* Compress file */ 367 #define EXT4_SYNC_FL 0x00000008 /* Synchronous updates */ 368 #define EXT4_IMMUTABLE_FL 0x00000010 /* Immutable file */ 369 #define EXT4_APPEND_FL 0x00000020 /* writes to file may only append */ 370 #define EXT4_NODUMP_FL 0x00000040 /* do not dump file */ 371 #define EXT4_NOATIME_FL 0x00000080 /* do not update atime */ 372 /* Reserved for compression usage... */ 373 #define EXT4_DIRTY_FL 0x00000100 374 #define EXT4_COMPRBLK_FL 0x00000200 /* One or more compressed clusters */ 375 #define EXT4_NOCOMPR_FL 0x00000400 /* Don't compress */ 376 #define EXT4_ECOMPR_FL 0x00000800 /* Compression error */ 377 /* End compression flags --- maybe not all used */ 378 #define EXT4_INDEX_FL 0x00001000 /* hash-indexed directory */ 379 #define EXT4_IMAGIC_FL 0x00002000 /* AFS directory */ 380 #define EXT4_JOURNAL_DATA_FL 0x00004000 /* file data should be journaled */ 381 #define EXT4_NOTAIL_FL 0x00008000 /* file tail should not be merged */ 382 #define EXT4_DIRSYNC_FL 0x00010000 /* dirsync behaviour (directories only) */ 383 #define EXT4_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/ 384 #define EXT4_HUGE_FILE_FL 0x00040000 /* Set to each huge file */ 385 #define EXT4_EXTENTS_FL 0x00080000 /* Inode uses extents */ 386 #define EXT4_EA_INODE_FL 0x00200000 /* Inode used for large EA */ 387 #define EXT4_EOFBLOCKS_FL 0x00400000 /* Blocks allocated beyond EOF */ 388 #define EXT4_RESERVED_FL 0x80000000 /* reserved for ext4 lib */ 389 390 #define EXT4_FL_USER_VISIBLE 0x004BDFFF /* User visible flags */ 391 #define EXT4_FL_USER_MODIFIABLE 0x004B80FF /* User modifiable flags */ 392 393 /* Flags that should be inherited by new inodes from their parent. */ 394 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\ 395 EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\ 396 EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\ 397 EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL) 398 399 /* Flags that are appropriate for regular files (all but dir-specific ones). */ 400 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL)) 401 402 /* Flags that are appropriate for non-directories/regular files. */ 403 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL) 404 405 /* Mask out flags that are inappropriate for the given type of inode. */ 406 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags) 407 { 408 if (S_ISDIR(mode)) 409 return flags; 410 else if (S_ISREG(mode)) 411 return flags & EXT4_REG_FLMASK; 412 else 413 return flags & EXT4_OTHER_FLMASK; 414 } 415 416 /* 417 * Inode flags used for atomic set/get 418 */ 419 enum { 420 EXT4_INODE_SECRM = 0, /* Secure deletion */ 421 EXT4_INODE_UNRM = 1, /* Undelete */ 422 EXT4_INODE_COMPR = 2, /* Compress file */ 423 EXT4_INODE_SYNC = 3, /* Synchronous updates */ 424 EXT4_INODE_IMMUTABLE = 4, /* Immutable file */ 425 EXT4_INODE_APPEND = 5, /* writes to file may only append */ 426 EXT4_INODE_NODUMP = 6, /* do not dump file */ 427 EXT4_INODE_NOATIME = 7, /* do not update atime */ 428 /* Reserved for compression usage... */ 429 EXT4_INODE_DIRTY = 8, 430 EXT4_INODE_COMPRBLK = 9, /* One or more compressed clusters */ 431 EXT4_INODE_NOCOMPR = 10, /* Don't compress */ 432 EXT4_INODE_ECOMPR = 11, /* Compression error */ 433 /* End compression flags --- maybe not all used */ 434 EXT4_INODE_INDEX = 12, /* hash-indexed directory */ 435 EXT4_INODE_IMAGIC = 13, /* AFS directory */ 436 EXT4_INODE_JOURNAL_DATA = 14, /* file data should be journaled */ 437 EXT4_INODE_NOTAIL = 15, /* file tail should not be merged */ 438 EXT4_INODE_DIRSYNC = 16, /* dirsync behaviour (directories only) */ 439 EXT4_INODE_TOPDIR = 17, /* Top of directory hierarchies*/ 440 EXT4_INODE_HUGE_FILE = 18, /* Set to each huge file */ 441 EXT4_INODE_EXTENTS = 19, /* Inode uses extents */ 442 EXT4_INODE_EA_INODE = 21, /* Inode used for large EA */ 443 EXT4_INODE_EOFBLOCKS = 22, /* Blocks allocated beyond EOF */ 444 EXT4_INODE_RESERVED = 31, /* reserved for ext4 lib */ 445 }; 446 447 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG)) 448 #define CHECK_FLAG_VALUE(FLAG) if (!TEST_FLAG_VALUE(FLAG)) { \ 449 printk(KERN_EMERG "EXT4 flag fail: " #FLAG ": %d %d\n", \ 450 EXT4_##FLAG##_FL, EXT4_INODE_##FLAG); BUG_ON(1); } 451 452 /* 453 * Since it's pretty easy to mix up bit numbers and hex values, and we 454 * can't do a compile-time test for ENUM values, we use a run-time 455 * test to make sure that EXT4_XXX_FL is consistent with respect to 456 * EXT4_INODE_XXX. If all is well the printk and BUG_ON will all drop 457 * out so it won't cost any extra space in the compiled kernel image. 458 * But it's important that these values are the same, since we are 459 * using EXT4_INODE_XXX to test for the flag values, but EXT4_XX_FL 460 * must be consistent with the values of FS_XXX_FL defined in 461 * include/linux/fs.h and the on-disk values found in ext2, ext3, and 462 * ext4 filesystems, and of course the values defined in e2fsprogs. 463 * 464 * It's not paranoia if the Murphy's Law really *is* out to get you. :-) 465 */ 466 static inline void ext4_check_flag_values(void) 467 { 468 CHECK_FLAG_VALUE(SECRM); 469 CHECK_FLAG_VALUE(UNRM); 470 CHECK_FLAG_VALUE(COMPR); 471 CHECK_FLAG_VALUE(SYNC); 472 CHECK_FLAG_VALUE(IMMUTABLE); 473 CHECK_FLAG_VALUE(APPEND); 474 CHECK_FLAG_VALUE(NODUMP); 475 CHECK_FLAG_VALUE(NOATIME); 476 CHECK_FLAG_VALUE(DIRTY); 477 CHECK_FLAG_VALUE(COMPRBLK); 478 CHECK_FLAG_VALUE(NOCOMPR); 479 CHECK_FLAG_VALUE(ECOMPR); 480 CHECK_FLAG_VALUE(INDEX); 481 CHECK_FLAG_VALUE(IMAGIC); 482 CHECK_FLAG_VALUE(JOURNAL_DATA); 483 CHECK_FLAG_VALUE(NOTAIL); 484 CHECK_FLAG_VALUE(DIRSYNC); 485 CHECK_FLAG_VALUE(TOPDIR); 486 CHECK_FLAG_VALUE(HUGE_FILE); 487 CHECK_FLAG_VALUE(EXTENTS); 488 CHECK_FLAG_VALUE(EA_INODE); 489 CHECK_FLAG_VALUE(EOFBLOCKS); 490 CHECK_FLAG_VALUE(RESERVED); 491 } 492 493 /* Used to pass group descriptor data when online resize is done */ 494 struct ext4_new_group_input { 495 __u32 group; /* Group number for this data */ 496 __u64 block_bitmap; /* Absolute block number of block bitmap */ 497 __u64 inode_bitmap; /* Absolute block number of inode bitmap */ 498 __u64 inode_table; /* Absolute block number of inode table start */ 499 __u32 blocks_count; /* Total number of blocks in this group */ 500 __u16 reserved_blocks; /* Number of reserved blocks in this group */ 501 __u16 unused; 502 }; 503 504 #if defined(__KERNEL__) && defined(CONFIG_COMPAT) 505 struct compat_ext4_new_group_input { 506 u32 group; 507 compat_u64 block_bitmap; 508 compat_u64 inode_bitmap; 509 compat_u64 inode_table; 510 u32 blocks_count; 511 u16 reserved_blocks; 512 u16 unused; 513 }; 514 #endif 515 516 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */ 517 struct ext4_new_group_data { 518 __u32 group; 519 __u64 block_bitmap; 520 __u64 inode_bitmap; 521 __u64 inode_table; 522 __u32 blocks_count; 523 __u16 reserved_blocks; 524 __u16 unused; 525 __u32 free_blocks_count; 526 }; 527 528 /* Indexes used to index group tables in ext4_new_group_data */ 529 enum { 530 BLOCK_BITMAP = 0, /* block bitmap */ 531 INODE_BITMAP, /* inode bitmap */ 532 INODE_TABLE, /* inode tables */ 533 GROUP_TABLE_COUNT, 534 }; 535 536 /* 537 * Flags used by ext4_map_blocks() 538 */ 539 /* Allocate any needed blocks and/or convert an unitialized 540 extent to be an initialized ext4 */ 541 #define EXT4_GET_BLOCKS_CREATE 0x0001 542 /* Request the creation of an unitialized extent */ 543 #define EXT4_GET_BLOCKS_UNINIT_EXT 0x0002 544 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT (EXT4_GET_BLOCKS_UNINIT_EXT|\ 545 EXT4_GET_BLOCKS_CREATE) 546 /* Caller is from the delayed allocation writeout path, 547 so set the magic i_delalloc_reserve_flag after taking the 548 inode allocation semaphore for */ 549 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE 0x0004 550 /* caller is from the direct IO path, request to creation of an 551 unitialized extents if not allocated, split the uninitialized 552 extent if blocks has been preallocated already*/ 553 #define EXT4_GET_BLOCKS_PRE_IO 0x0008 554 #define EXT4_GET_BLOCKS_CONVERT 0x0010 555 #define EXT4_GET_BLOCKS_IO_CREATE_EXT (EXT4_GET_BLOCKS_PRE_IO|\ 556 EXT4_GET_BLOCKS_CREATE_UNINIT_EXT) 557 /* Convert extent to initialized after IO complete */ 558 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT (EXT4_GET_BLOCKS_CONVERT|\ 559 EXT4_GET_BLOCKS_CREATE_UNINIT_EXT) 560 /* Punch out blocks of an extent */ 561 #define EXT4_GET_BLOCKS_PUNCH_OUT_EXT 0x0020 562 /* Don't normalize allocation size (used for fallocate) */ 563 #define EXT4_GET_BLOCKS_NO_NORMALIZE 0x0040 564 /* Request will not result in inode size update (user for fallocate) */ 565 #define EXT4_GET_BLOCKS_KEEP_SIZE 0x0080 566 567 /* 568 * Flags used by ext4_free_blocks 569 */ 570 #define EXT4_FREE_BLOCKS_METADATA 0x0001 571 #define EXT4_FREE_BLOCKS_FORGET 0x0002 572 #define EXT4_FREE_BLOCKS_VALIDATED 0x0004 573 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008 574 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER 0x0010 575 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER 0x0020 576 577 /* 578 * Flags used by ext4_discard_partial_page_buffers 579 */ 580 #define EXT4_DISCARD_PARTIAL_PG_ZERO_UNMAPPED 0x0001 581 582 /* 583 * ioctl commands 584 */ 585 #define EXT4_IOC_GETFLAGS FS_IOC_GETFLAGS 586 #define EXT4_IOC_SETFLAGS FS_IOC_SETFLAGS 587 #define EXT4_IOC_GETVERSION _IOR('f', 3, long) 588 #define EXT4_IOC_SETVERSION _IOW('f', 4, long) 589 #define EXT4_IOC_GETVERSION_OLD FS_IOC_GETVERSION 590 #define EXT4_IOC_SETVERSION_OLD FS_IOC_SETVERSION 591 #define EXT4_IOC_GETRSVSZ _IOR('f', 5, long) 592 #define EXT4_IOC_SETRSVSZ _IOW('f', 6, long) 593 #define EXT4_IOC_GROUP_EXTEND _IOW('f', 7, unsigned long) 594 #define EXT4_IOC_GROUP_ADD _IOW('f', 8, struct ext4_new_group_input) 595 #define EXT4_IOC_MIGRATE _IO('f', 9) 596 /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */ 597 /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */ 598 #define EXT4_IOC_ALLOC_DA_BLKS _IO('f', 12) 599 #define EXT4_IOC_MOVE_EXT _IOWR('f', 15, struct move_extent) 600 #define EXT4_IOC_RESIZE_FS _IOW('f', 16, __u64) 601 602 #if defined(__KERNEL__) && defined(CONFIG_COMPAT) 603 /* 604 * ioctl commands in 32 bit emulation 605 */ 606 #define EXT4_IOC32_GETFLAGS FS_IOC32_GETFLAGS 607 #define EXT4_IOC32_SETFLAGS FS_IOC32_SETFLAGS 608 #define EXT4_IOC32_GETVERSION _IOR('f', 3, int) 609 #define EXT4_IOC32_SETVERSION _IOW('f', 4, int) 610 #define EXT4_IOC32_GETRSVSZ _IOR('f', 5, int) 611 #define EXT4_IOC32_SETRSVSZ _IOW('f', 6, int) 612 #define EXT4_IOC32_GROUP_EXTEND _IOW('f', 7, unsigned int) 613 #define EXT4_IOC32_GROUP_ADD _IOW('f', 8, struct compat_ext4_new_group_input) 614 #define EXT4_IOC32_GETVERSION_OLD FS_IOC32_GETVERSION 615 #define EXT4_IOC32_SETVERSION_OLD FS_IOC32_SETVERSION 616 #endif 617 618 /* Max physical block we can address w/o extents */ 619 #define EXT4_MAX_BLOCK_FILE_PHYS 0xFFFFFFFF 620 621 /* 622 * Structure of an inode on the disk 623 */ 624 struct ext4_inode { 625 __le16 i_mode; /* File mode */ 626 __le16 i_uid; /* Low 16 bits of Owner Uid */ 627 __le32 i_size_lo; /* Size in bytes */ 628 __le32 i_atime; /* Access time */ 629 __le32 i_ctime; /* Inode Change time */ 630 __le32 i_mtime; /* Modification time */ 631 __le32 i_dtime; /* Deletion Time */ 632 __le16 i_gid; /* Low 16 bits of Group Id */ 633 __le16 i_links_count; /* Links count */ 634 __le32 i_blocks_lo; /* Blocks count */ 635 __le32 i_flags; /* File flags */ 636 union { 637 struct { 638 __le32 l_i_version; 639 } linux1; 640 struct { 641 __u32 h_i_translator; 642 } hurd1; 643 struct { 644 __u32 m_i_reserved1; 645 } masix1; 646 } osd1; /* OS dependent 1 */ 647 __le32 i_block[EXT4_N_BLOCKS];/* Pointers to blocks */ 648 __le32 i_generation; /* File version (for NFS) */ 649 __le32 i_file_acl_lo; /* File ACL */ 650 __le32 i_size_high; 651 __le32 i_obso_faddr; /* Obsoleted fragment address */ 652 union { 653 struct { 654 __le16 l_i_blocks_high; /* were l_i_reserved1 */ 655 __le16 l_i_file_acl_high; 656 __le16 l_i_uid_high; /* these 2 fields */ 657 __le16 l_i_gid_high; /* were reserved2[0] */ 658 __le16 l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */ 659 __le16 l_i_reserved; 660 } linux2; 661 struct { 662 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */ 663 __u16 h_i_mode_high; 664 __u16 h_i_uid_high; 665 __u16 h_i_gid_high; 666 __u32 h_i_author; 667 } hurd2; 668 struct { 669 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */ 670 __le16 m_i_file_acl_high; 671 __u32 m_i_reserved2[2]; 672 } masix2; 673 } osd2; /* OS dependent 2 */ 674 __le16 i_extra_isize; 675 __le16 i_checksum_hi; /* crc32c(uuid+inum+inode) BE */ 676 __le32 i_ctime_extra; /* extra Change time (nsec << 2 | epoch) */ 677 __le32 i_mtime_extra; /* extra Modification time(nsec << 2 | epoch) */ 678 __le32 i_atime_extra; /* extra Access time (nsec << 2 | epoch) */ 679 __le32 i_crtime; /* File Creation time */ 680 __le32 i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */ 681 __le32 i_version_hi; /* high 32 bits for 64-bit version */ 682 }; 683 684 struct move_extent { 685 __u32 reserved; /* should be zero */ 686 __u32 donor_fd; /* donor file descriptor */ 687 __u64 orig_start; /* logical start offset in block for orig */ 688 __u64 donor_start; /* logical start offset in block for donor */ 689 __u64 len; /* block length to be moved */ 690 __u64 moved_len; /* moved block length */ 691 }; 692 693 #define EXT4_EPOCH_BITS 2 694 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1) 695 #define EXT4_NSEC_MASK (~0UL << EXT4_EPOCH_BITS) 696 697 /* 698 * Extended fields will fit into an inode if the filesystem was formatted 699 * with large inodes (-I 256 or larger) and there are not currently any EAs 700 * consuming all of the available space. For new inodes we always reserve 701 * enough space for the kernel's known extended fields, but for inodes 702 * created with an old kernel this might not have been the case. None of 703 * the extended inode fields is critical for correct filesystem operation. 704 * This macro checks if a certain field fits in the inode. Note that 705 * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize 706 */ 707 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field) \ 708 ((offsetof(typeof(*ext4_inode), field) + \ 709 sizeof((ext4_inode)->field)) \ 710 <= (EXT4_GOOD_OLD_INODE_SIZE + \ 711 (einode)->i_extra_isize)) \ 712 713 static inline __le32 ext4_encode_extra_time(struct timespec *time) 714 { 715 return cpu_to_le32((sizeof(time->tv_sec) > 4 ? 716 (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) | 717 ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK)); 718 } 719 720 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra) 721 { 722 if (sizeof(time->tv_sec) > 4) 723 time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK) 724 << 32; 725 time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS; 726 } 727 728 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode) \ 729 do { \ 730 (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec); \ 731 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \ 732 (raw_inode)->xtime ## _extra = \ 733 ext4_encode_extra_time(&(inode)->xtime); \ 734 } while (0) 735 736 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode) \ 737 do { \ 738 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \ 739 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec); \ 740 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \ 741 (raw_inode)->xtime ## _extra = \ 742 ext4_encode_extra_time(&(einode)->xtime); \ 743 } while (0) 744 745 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode) \ 746 do { \ 747 (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime); \ 748 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \ 749 ext4_decode_extra_time(&(inode)->xtime, \ 750 raw_inode->xtime ## _extra); \ 751 else \ 752 (inode)->xtime.tv_nsec = 0; \ 753 } while (0) 754 755 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode) \ 756 do { \ 757 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \ 758 (einode)->xtime.tv_sec = \ 759 (signed)le32_to_cpu((raw_inode)->xtime); \ 760 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \ 761 ext4_decode_extra_time(&(einode)->xtime, \ 762 raw_inode->xtime ## _extra); \ 763 else \ 764 (einode)->xtime.tv_nsec = 0; \ 765 } while (0) 766 767 #define i_disk_version osd1.linux1.l_i_version 768 769 #if defined(__KERNEL__) || defined(__linux__) 770 #define i_reserved1 osd1.linux1.l_i_reserved1 771 #define i_file_acl_high osd2.linux2.l_i_file_acl_high 772 #define i_blocks_high osd2.linux2.l_i_blocks_high 773 #define i_uid_low i_uid 774 #define i_gid_low i_gid 775 #define i_uid_high osd2.linux2.l_i_uid_high 776 #define i_gid_high osd2.linux2.l_i_gid_high 777 #define i_checksum_lo osd2.linux2.l_i_checksum_lo 778 779 #elif defined(__GNU__) 780 781 #define i_translator osd1.hurd1.h_i_translator 782 #define i_uid_high osd2.hurd2.h_i_uid_high 783 #define i_gid_high osd2.hurd2.h_i_gid_high 784 #define i_author osd2.hurd2.h_i_author 785 786 #elif defined(__masix__) 787 788 #define i_reserved1 osd1.masix1.m_i_reserved1 789 #define i_file_acl_high osd2.masix2.m_i_file_acl_high 790 #define i_reserved2 osd2.masix2.m_i_reserved2 791 792 #endif /* defined(__KERNEL__) || defined(__linux__) */ 793 794 /* 795 * storage for cached extent 796 * If ec_len == 0, then the cache is invalid. 797 * If ec_start == 0, then the cache represents a gap (null mapping) 798 */ 799 struct ext4_ext_cache { 800 ext4_fsblk_t ec_start; 801 ext4_lblk_t ec_block; 802 __u32 ec_len; /* must be 32bit to return holes */ 803 }; 804 805 /* 806 * fourth extended file system inode data in memory 807 */ 808 struct ext4_inode_info { 809 __le32 i_data[15]; /* unconverted */ 810 __u32 i_dtime; 811 ext4_fsblk_t i_file_acl; 812 813 /* 814 * i_block_group is the number of the block group which contains 815 * this file's inode. Constant across the lifetime of the inode, 816 * it is ued for making block allocation decisions - we try to 817 * place a file's data blocks near its inode block, and new inodes 818 * near to their parent directory's inode. 819 */ 820 ext4_group_t i_block_group; 821 ext4_lblk_t i_dir_start_lookup; 822 #if (BITS_PER_LONG < 64) 823 unsigned long i_state_flags; /* Dynamic state flags */ 824 #endif 825 unsigned long i_flags; 826 827 #ifdef CONFIG_EXT4_FS_XATTR 828 /* 829 * Extended attributes can be read independently of the main file 830 * data. Taking i_mutex even when reading would cause contention 831 * between readers of EAs and writers of regular file data, so 832 * instead we synchronize on xattr_sem when reading or changing 833 * EAs. 834 */ 835 struct rw_semaphore xattr_sem; 836 #endif 837 838 struct list_head i_orphan; /* unlinked but open inodes */ 839 840 /* 841 * i_disksize keeps track of what the inode size is ON DISK, not 842 * in memory. During truncate, i_size is set to the new size by 843 * the VFS prior to calling ext4_truncate(), but the filesystem won't 844 * set i_disksize to 0 until the truncate is actually under way. 845 * 846 * The intent is that i_disksize always represents the blocks which 847 * are used by this file. This allows recovery to restart truncate 848 * on orphans if we crash during truncate. We actually write i_disksize 849 * into the on-disk inode when writing inodes out, instead of i_size. 850 * 851 * The only time when i_disksize and i_size may be different is when 852 * a truncate is in progress. The only things which change i_disksize 853 * are ext4_get_block (growth) and ext4_truncate (shrinkth). 854 */ 855 loff_t i_disksize; 856 857 /* 858 * i_data_sem is for serialising ext4_truncate() against 859 * ext4_getblock(). In the 2.4 ext2 design, great chunks of inode's 860 * data tree are chopped off during truncate. We can't do that in 861 * ext4 because whenever we perform intermediate commits during 862 * truncate, the inode and all the metadata blocks *must* be in a 863 * consistent state which allows truncation of the orphans to restart 864 * during recovery. Hence we must fix the get_block-vs-truncate race 865 * by other means, so we have i_data_sem. 866 */ 867 struct rw_semaphore i_data_sem; 868 struct inode vfs_inode; 869 struct jbd2_inode *jinode; 870 871 struct ext4_ext_cache i_cached_extent; 872 /* 873 * File creation time. Its function is same as that of 874 * struct timespec i_{a,c,m}time in the generic inode. 875 */ 876 struct timespec i_crtime; 877 878 /* mballoc */ 879 struct list_head i_prealloc_list; 880 spinlock_t i_prealloc_lock; 881 882 /* ialloc */ 883 ext4_group_t i_last_alloc_group; 884 885 /* allocation reservation info for delalloc */ 886 /* In case of bigalloc, these refer to clusters rather than blocks */ 887 unsigned int i_reserved_data_blocks; 888 unsigned int i_reserved_meta_blocks; 889 unsigned int i_allocated_meta_blocks; 890 ext4_lblk_t i_da_metadata_calc_last_lblock; 891 int i_da_metadata_calc_len; 892 893 /* on-disk additional length */ 894 __u16 i_extra_isize; 895 896 #ifdef CONFIG_QUOTA 897 /* quota space reservation, managed internally by quota code */ 898 qsize_t i_reserved_quota; 899 #endif 900 901 /* completed IOs that might need unwritten extents handling */ 902 struct list_head i_completed_io_list; 903 spinlock_t i_completed_io_lock; 904 atomic_t i_ioend_count; /* Number of outstanding io_end structs */ 905 /* current io_end structure for async DIO write*/ 906 ext4_io_end_t *cur_aio_dio; 907 atomic_t i_aiodio_unwritten; /* Nr. of inflight conversions pending */ 908 909 spinlock_t i_block_reservation_lock; 910 911 /* 912 * Transactions that contain inode's metadata needed to complete 913 * fsync and fdatasync, respectively. 914 */ 915 tid_t i_sync_tid; 916 tid_t i_datasync_tid; 917 }; 918 919 /* 920 * File system states 921 */ 922 #define EXT4_VALID_FS 0x0001 /* Unmounted cleanly */ 923 #define EXT4_ERROR_FS 0x0002 /* Errors detected */ 924 #define EXT4_ORPHAN_FS 0x0004 /* Orphans being recovered */ 925 926 /* 927 * Misc. filesystem flags 928 */ 929 #define EXT2_FLAGS_SIGNED_HASH 0x0001 /* Signed dirhash in use */ 930 #define EXT2_FLAGS_UNSIGNED_HASH 0x0002 /* Unsigned dirhash in use */ 931 #define EXT2_FLAGS_TEST_FILESYS 0x0004 /* to test development code */ 932 933 /* 934 * Mount flags 935 */ 936 #define EXT4_MOUNT_GRPID 0x00004 /* Create files with directory's group */ 937 #define EXT4_MOUNT_DEBUG 0x00008 /* Some debugging messages */ 938 #define EXT4_MOUNT_ERRORS_CONT 0x00010 /* Continue on errors */ 939 #define EXT4_MOUNT_ERRORS_RO 0x00020 /* Remount fs ro on errors */ 940 #define EXT4_MOUNT_ERRORS_PANIC 0x00040 /* Panic on errors */ 941 #define EXT4_MOUNT_ERRORS_MASK 0x00070 942 #define EXT4_MOUNT_MINIX_DF 0x00080 /* Mimics the Minix statfs */ 943 #define EXT4_MOUNT_NOLOAD 0x00100 /* Don't use existing journal*/ 944 #define EXT4_MOUNT_DATA_FLAGS 0x00C00 /* Mode for data writes: */ 945 #define EXT4_MOUNT_JOURNAL_DATA 0x00400 /* Write data to journal */ 946 #define EXT4_MOUNT_ORDERED_DATA 0x00800 /* Flush data before commit */ 947 #define EXT4_MOUNT_WRITEBACK_DATA 0x00C00 /* No data ordering */ 948 #define EXT4_MOUNT_UPDATE_JOURNAL 0x01000 /* Update the journal format */ 949 #define EXT4_MOUNT_NO_UID32 0x02000 /* Disable 32-bit UIDs */ 950 #define EXT4_MOUNT_XATTR_USER 0x04000 /* Extended user attributes */ 951 #define EXT4_MOUNT_POSIX_ACL 0x08000 /* POSIX Access Control Lists */ 952 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC 0x10000 /* No auto delalloc mapping */ 953 #define EXT4_MOUNT_BARRIER 0x20000 /* Use block barriers */ 954 #define EXT4_MOUNT_QUOTA 0x80000 /* Some quota option set */ 955 #define EXT4_MOUNT_USRQUOTA 0x100000 /* "old" user quota */ 956 #define EXT4_MOUNT_GRPQUOTA 0x200000 /* "old" group quota */ 957 #define EXT4_MOUNT_DIOREAD_NOLOCK 0x400000 /* Enable support for dio read nolocking */ 958 #define EXT4_MOUNT_JOURNAL_CHECKSUM 0x800000 /* Journal checksums */ 959 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */ 960 #define EXT4_MOUNT_MBLK_IO_SUBMIT 0x4000000 /* multi-block io submits */ 961 #define EXT4_MOUNT_DELALLOC 0x8000000 /* Delalloc support */ 962 #define EXT4_MOUNT_DATA_ERR_ABORT 0x10000000 /* Abort on file data write */ 963 #define EXT4_MOUNT_BLOCK_VALIDITY 0x20000000 /* Block validity checking */ 964 #define EXT4_MOUNT_DISCARD 0x40000000 /* Issue DISCARD requests */ 965 #define EXT4_MOUNT_INIT_INODE_TABLE 0x80000000 /* Initialize uninitialized itables */ 966 967 #define EXT4_MOUNT2_EXPLICIT_DELALLOC 0x00000001 /* User explicitly 968 specified delalloc */ 969 970 #define clear_opt(sb, opt) EXT4_SB(sb)->s_mount_opt &= \ 971 ~EXT4_MOUNT_##opt 972 #define set_opt(sb, opt) EXT4_SB(sb)->s_mount_opt |= \ 973 EXT4_MOUNT_##opt 974 #define test_opt(sb, opt) (EXT4_SB(sb)->s_mount_opt & \ 975 EXT4_MOUNT_##opt) 976 977 #define clear_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 &= \ 978 ~EXT4_MOUNT2_##opt 979 #define set_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 |= \ 980 EXT4_MOUNT2_##opt 981 #define test_opt2(sb, opt) (EXT4_SB(sb)->s_mount_opt2 & \ 982 EXT4_MOUNT2_##opt) 983 984 #define ext4_test_and_set_bit __test_and_set_bit_le 985 #define ext4_set_bit __set_bit_le 986 #define ext4_set_bit_atomic ext2_set_bit_atomic 987 #define ext4_test_and_clear_bit __test_and_clear_bit_le 988 #define ext4_clear_bit __clear_bit_le 989 #define ext4_clear_bit_atomic ext2_clear_bit_atomic 990 #define ext4_test_bit test_bit_le 991 #define ext4_find_next_zero_bit find_next_zero_bit_le 992 #define ext4_find_next_bit find_next_bit_le 993 994 extern void ext4_set_bits(void *bm, int cur, int len); 995 996 /* 997 * Maximal mount counts between two filesystem checks 998 */ 999 #define EXT4_DFL_MAX_MNT_COUNT 20 /* Allow 20 mounts */ 1000 #define EXT4_DFL_CHECKINTERVAL 0 /* Don't use interval check */ 1001 1002 /* 1003 * Behaviour when detecting errors 1004 */ 1005 #define EXT4_ERRORS_CONTINUE 1 /* Continue execution */ 1006 #define EXT4_ERRORS_RO 2 /* Remount fs read-only */ 1007 #define EXT4_ERRORS_PANIC 3 /* Panic */ 1008 #define EXT4_ERRORS_DEFAULT EXT4_ERRORS_CONTINUE 1009 1010 /* Metadata checksum algorithm codes */ 1011 #define EXT4_CRC32C_CHKSUM 1 1012 1013 /* 1014 * Structure of the super block 1015 */ 1016 struct ext4_super_block { 1017 /*00*/ __le32 s_inodes_count; /* Inodes count */ 1018 __le32 s_blocks_count_lo; /* Blocks count */ 1019 __le32 s_r_blocks_count_lo; /* Reserved blocks count */ 1020 __le32 s_free_blocks_count_lo; /* Free blocks count */ 1021 /*10*/ __le32 s_free_inodes_count; /* Free inodes count */ 1022 __le32 s_first_data_block; /* First Data Block */ 1023 __le32 s_log_block_size; /* Block size */ 1024 __le32 s_log_cluster_size; /* Allocation cluster size */ 1025 /*20*/ __le32 s_blocks_per_group; /* # Blocks per group */ 1026 __le32 s_clusters_per_group; /* # Clusters per group */ 1027 __le32 s_inodes_per_group; /* # Inodes per group */ 1028 __le32 s_mtime; /* Mount time */ 1029 /*30*/ __le32 s_wtime; /* Write time */ 1030 __le16 s_mnt_count; /* Mount count */ 1031 __le16 s_max_mnt_count; /* Maximal mount count */ 1032 __le16 s_magic; /* Magic signature */ 1033 __le16 s_state; /* File system state */ 1034 __le16 s_errors; /* Behaviour when detecting errors */ 1035 __le16 s_minor_rev_level; /* minor revision level */ 1036 /*40*/ __le32 s_lastcheck; /* time of last check */ 1037 __le32 s_checkinterval; /* max. time between checks */ 1038 __le32 s_creator_os; /* OS */ 1039 __le32 s_rev_level; /* Revision level */ 1040 /*50*/ __le16 s_def_resuid; /* Default uid for reserved blocks */ 1041 __le16 s_def_resgid; /* Default gid for reserved blocks */ 1042 /* 1043 * These fields are for EXT4_DYNAMIC_REV superblocks only. 1044 * 1045 * Note: the difference between the compatible feature set and 1046 * the incompatible feature set is that if there is a bit set 1047 * in the incompatible feature set that the kernel doesn't 1048 * know about, it should refuse to mount the filesystem. 1049 * 1050 * e2fsck's requirements are more strict; if it doesn't know 1051 * about a feature in either the compatible or incompatible 1052 * feature set, it must abort and not try to meddle with 1053 * things it doesn't understand... 1054 */ 1055 __le32 s_first_ino; /* First non-reserved inode */ 1056 __le16 s_inode_size; /* size of inode structure */ 1057 __le16 s_block_group_nr; /* block group # of this superblock */ 1058 __le32 s_feature_compat; /* compatible feature set */ 1059 /*60*/ __le32 s_feature_incompat; /* incompatible feature set */ 1060 __le32 s_feature_ro_compat; /* readonly-compatible feature set */ 1061 /*68*/ __u8 s_uuid[16]; /* 128-bit uuid for volume */ 1062 /*78*/ char s_volume_name[16]; /* volume name */ 1063 /*88*/ char s_last_mounted[64]; /* directory where last mounted */ 1064 /*C8*/ __le32 s_algorithm_usage_bitmap; /* For compression */ 1065 /* 1066 * Performance hints. Directory preallocation should only 1067 * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on. 1068 */ 1069 __u8 s_prealloc_blocks; /* Nr of blocks to try to preallocate*/ 1070 __u8 s_prealloc_dir_blocks; /* Nr to preallocate for dirs */ 1071 __le16 s_reserved_gdt_blocks; /* Per group desc for online growth */ 1072 /* 1073 * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set. 1074 */ 1075 /*D0*/ __u8 s_journal_uuid[16]; /* uuid of journal superblock */ 1076 /*E0*/ __le32 s_journal_inum; /* inode number of journal file */ 1077 __le32 s_journal_dev; /* device number of journal file */ 1078 __le32 s_last_orphan; /* start of list of inodes to delete */ 1079 __le32 s_hash_seed[4]; /* HTREE hash seed */ 1080 __u8 s_def_hash_version; /* Default hash version to use */ 1081 __u8 s_jnl_backup_type; 1082 __le16 s_desc_size; /* size of group descriptor */ 1083 /*100*/ __le32 s_default_mount_opts; 1084 __le32 s_first_meta_bg; /* First metablock block group */ 1085 __le32 s_mkfs_time; /* When the filesystem was created */ 1086 __le32 s_jnl_blocks[17]; /* Backup of the journal inode */ 1087 /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */ 1088 /*150*/ __le32 s_blocks_count_hi; /* Blocks count */ 1089 __le32 s_r_blocks_count_hi; /* Reserved blocks count */ 1090 __le32 s_free_blocks_count_hi; /* Free blocks count */ 1091 __le16 s_min_extra_isize; /* All inodes have at least # bytes */ 1092 __le16 s_want_extra_isize; /* New inodes should reserve # bytes */ 1093 __le32 s_flags; /* Miscellaneous flags */ 1094 __le16 s_raid_stride; /* RAID stride */ 1095 __le16 s_mmp_update_interval; /* # seconds to wait in MMP checking */ 1096 __le64 s_mmp_block; /* Block for multi-mount protection */ 1097 __le32 s_raid_stripe_width; /* blocks on all data disks (N*stride)*/ 1098 __u8 s_log_groups_per_flex; /* FLEX_BG group size */ 1099 __u8 s_checksum_type; /* metadata checksum algorithm used */ 1100 __le16 s_reserved_pad; 1101 __le64 s_kbytes_written; /* nr of lifetime kilobytes written */ 1102 __le32 s_snapshot_inum; /* Inode number of active snapshot */ 1103 __le32 s_snapshot_id; /* sequential ID of active snapshot */ 1104 __le64 s_snapshot_r_blocks_count; /* reserved blocks for active 1105 snapshot's future use */ 1106 __le32 s_snapshot_list; /* inode number of the head of the 1107 on-disk snapshot list */ 1108 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count) 1109 __le32 s_error_count; /* number of fs errors */ 1110 __le32 s_first_error_time; /* first time an error happened */ 1111 __le32 s_first_error_ino; /* inode involved in first error */ 1112 __le64 s_first_error_block; /* block involved of first error */ 1113 __u8 s_first_error_func[32]; /* function where the error happened */ 1114 __le32 s_first_error_line; /* line number where error happened */ 1115 __le32 s_last_error_time; /* most recent time of an error */ 1116 __le32 s_last_error_ino; /* inode involved in last error */ 1117 __le32 s_last_error_line; /* line number where error happened */ 1118 __le64 s_last_error_block; /* block involved of last error */ 1119 __u8 s_last_error_func[32]; /* function where the error happened */ 1120 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts) 1121 __u8 s_mount_opts[64]; 1122 __le32 s_usr_quota_inum; /* inode for tracking user quota */ 1123 __le32 s_grp_quota_inum; /* inode for tracking group quota */ 1124 __le32 s_overhead_clusters; /* overhead blocks/clusters in fs */ 1125 __le32 s_reserved[108]; /* Padding to the end of the block */ 1126 __le32 s_checksum; /* crc32c(superblock) */ 1127 }; 1128 1129 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START) 1130 1131 #ifdef __KERNEL__ 1132 1133 /* 1134 * run-time mount flags 1135 */ 1136 #define EXT4_MF_MNTDIR_SAMPLED 0x0001 1137 #define EXT4_MF_FS_ABORTED 0x0002 /* Fatal error detected */ 1138 1139 /* 1140 * fourth extended-fs super-block data in memory 1141 */ 1142 struct ext4_sb_info { 1143 unsigned long s_desc_size; /* Size of a group descriptor in bytes */ 1144 unsigned long s_inodes_per_block;/* Number of inodes per block */ 1145 unsigned long s_blocks_per_group;/* Number of blocks in a group */ 1146 unsigned long s_clusters_per_group; /* Number of clusters in a group */ 1147 unsigned long s_inodes_per_group;/* Number of inodes in a group */ 1148 unsigned long s_itb_per_group; /* Number of inode table blocks per group */ 1149 unsigned long s_gdb_count; /* Number of group descriptor blocks */ 1150 unsigned long s_desc_per_block; /* Number of group descriptors per block */ 1151 ext4_group_t s_groups_count; /* Number of groups in the fs */ 1152 ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */ 1153 unsigned long s_overhead_last; /* Last calculated overhead */ 1154 unsigned long s_blocks_last; /* Last seen block count */ 1155 unsigned int s_cluster_ratio; /* Number of blocks per cluster */ 1156 unsigned int s_cluster_bits; /* log2 of s_cluster_ratio */ 1157 loff_t s_bitmap_maxbytes; /* max bytes for bitmap files */ 1158 struct buffer_head * s_sbh; /* Buffer containing the super block */ 1159 struct ext4_super_block *s_es; /* Pointer to the super block in the buffer */ 1160 struct buffer_head **s_group_desc; 1161 unsigned int s_mount_opt; 1162 unsigned int s_mount_opt2; 1163 unsigned int s_mount_flags; 1164 unsigned int s_def_mount_opt; 1165 ext4_fsblk_t s_sb_block; 1166 uid_t s_resuid; 1167 gid_t s_resgid; 1168 unsigned short s_mount_state; 1169 unsigned short s_pad; 1170 int s_addr_per_block_bits; 1171 int s_desc_per_block_bits; 1172 int s_inode_size; 1173 int s_first_ino; 1174 unsigned int s_inode_readahead_blks; 1175 unsigned int s_inode_goal; 1176 spinlock_t s_next_gen_lock; 1177 u32 s_next_generation; 1178 u32 s_hash_seed[4]; 1179 int s_def_hash_version; 1180 int s_hash_unsigned; /* 3 if hash should be signed, 0 if not */ 1181 struct percpu_counter s_freeclusters_counter; 1182 struct percpu_counter s_freeinodes_counter; 1183 struct percpu_counter s_dirs_counter; 1184 struct percpu_counter s_dirtyclusters_counter; 1185 struct blockgroup_lock *s_blockgroup_lock; 1186 struct proc_dir_entry *s_proc; 1187 struct kobject s_kobj; 1188 struct completion s_kobj_unregister; 1189 1190 /* Journaling */ 1191 struct journal_s *s_journal; 1192 struct list_head s_orphan; 1193 struct mutex s_orphan_lock; 1194 unsigned long s_resize_flags; /* Flags indicating if there 1195 is a resizer */ 1196 unsigned long s_commit_interval; 1197 u32 s_max_batch_time; 1198 u32 s_min_batch_time; 1199 struct block_device *journal_bdev; 1200 #ifdef CONFIG_QUOTA 1201 char *s_qf_names[MAXQUOTAS]; /* Names of quota files with journalled quota */ 1202 int s_jquota_fmt; /* Format of quota to use */ 1203 #endif 1204 unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */ 1205 struct rb_root system_blks; 1206 1207 #ifdef EXTENTS_STATS 1208 /* ext4 extents stats */ 1209 unsigned long s_ext_min; 1210 unsigned long s_ext_max; 1211 unsigned long s_depth_max; 1212 spinlock_t s_ext_stats_lock; 1213 unsigned long s_ext_blocks; 1214 unsigned long s_ext_extents; 1215 #endif 1216 1217 /* for buddy allocator */ 1218 struct ext4_group_info ***s_group_info; 1219 struct inode *s_buddy_cache; 1220 spinlock_t s_md_lock; 1221 unsigned short *s_mb_offsets; 1222 unsigned int *s_mb_maxs; 1223 1224 /* tunables */ 1225 unsigned long s_stripe; 1226 unsigned int s_mb_stream_request; 1227 unsigned int s_mb_max_to_scan; 1228 unsigned int s_mb_min_to_scan; 1229 unsigned int s_mb_stats; 1230 unsigned int s_mb_order2_reqs; 1231 unsigned int s_mb_group_prealloc; 1232 unsigned int s_max_writeback_mb_bump; 1233 /* where last allocation was done - for stream allocation */ 1234 unsigned long s_mb_last_group; 1235 unsigned long s_mb_last_start; 1236 1237 /* stats for buddy allocator */ 1238 atomic_t s_bal_reqs; /* number of reqs with len > 1 */ 1239 atomic_t s_bal_success; /* we found long enough chunks */ 1240 atomic_t s_bal_allocated; /* in blocks */ 1241 atomic_t s_bal_ex_scanned; /* total extents scanned */ 1242 atomic_t s_bal_goals; /* goal hits */ 1243 atomic_t s_bal_breaks; /* too long searches */ 1244 atomic_t s_bal_2orders; /* 2^order hits */ 1245 spinlock_t s_bal_lock; 1246 unsigned long s_mb_buddies_generated; 1247 unsigned long long s_mb_generation_time; 1248 atomic_t s_mb_lost_chunks; 1249 atomic_t s_mb_preallocated; 1250 atomic_t s_mb_discarded; 1251 atomic_t s_lock_busy; 1252 1253 /* locality groups */ 1254 struct ext4_locality_group __percpu *s_locality_groups; 1255 1256 /* for write statistics */ 1257 unsigned long s_sectors_written_start; 1258 u64 s_kbytes_written; 1259 1260 unsigned int s_log_groups_per_flex; 1261 struct flex_groups *s_flex_groups; 1262 1263 /* workqueue for dio unwritten */ 1264 struct workqueue_struct *dio_unwritten_wq; 1265 1266 /* timer for periodic error stats printing */ 1267 struct timer_list s_err_report; 1268 1269 /* Lazy inode table initialization info */ 1270 struct ext4_li_request *s_li_request; 1271 /* Wait multiplier for lazy initialization thread */ 1272 unsigned int s_li_wait_mult; 1273 1274 /* Kernel thread for multiple mount protection */ 1275 struct task_struct *s_mmp_tsk; 1276 1277 /* record the last minlen when FITRIM is called. */ 1278 atomic_t s_last_trim_minblks; 1279 }; 1280 1281 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb) 1282 { 1283 return sb->s_fs_info; 1284 } 1285 static inline struct ext4_inode_info *EXT4_I(struct inode *inode) 1286 { 1287 return container_of(inode, struct ext4_inode_info, vfs_inode); 1288 } 1289 1290 static inline struct timespec ext4_current_time(struct inode *inode) 1291 { 1292 return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ? 1293 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC; 1294 } 1295 1296 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino) 1297 { 1298 return ino == EXT4_ROOT_INO || 1299 ino == EXT4_JOURNAL_INO || 1300 ino == EXT4_RESIZE_INO || 1301 (ino >= EXT4_FIRST_INO(sb) && 1302 ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)); 1303 } 1304 1305 static inline void ext4_set_io_unwritten_flag(struct inode *inode, 1306 struct ext4_io_end *io_end) 1307 { 1308 if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) { 1309 io_end->flag |= EXT4_IO_END_UNWRITTEN; 1310 atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten); 1311 } 1312 } 1313 1314 /* 1315 * Inode dynamic state flags 1316 */ 1317 enum { 1318 EXT4_STATE_JDATA, /* journaled data exists */ 1319 EXT4_STATE_NEW, /* inode is newly created */ 1320 EXT4_STATE_XATTR, /* has in-inode xattrs */ 1321 EXT4_STATE_NO_EXPAND, /* No space for expansion */ 1322 EXT4_STATE_DA_ALLOC_CLOSE, /* Alloc DA blks on close */ 1323 EXT4_STATE_EXT_MIGRATE, /* Inode is migrating */ 1324 EXT4_STATE_DIO_UNWRITTEN, /* need convert on dio done*/ 1325 EXT4_STATE_NEWENTRY, /* File just added to dir */ 1326 EXT4_STATE_DELALLOC_RESERVED, /* blks already reserved for delalloc */ 1327 }; 1328 1329 #define EXT4_INODE_BIT_FNS(name, field, offset) \ 1330 static inline int ext4_test_inode_##name(struct inode *inode, int bit) \ 1331 { \ 1332 return test_bit(bit + (offset), &EXT4_I(inode)->i_##field); \ 1333 } \ 1334 static inline void ext4_set_inode_##name(struct inode *inode, int bit) \ 1335 { \ 1336 set_bit(bit + (offset), &EXT4_I(inode)->i_##field); \ 1337 } \ 1338 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \ 1339 { \ 1340 clear_bit(bit + (offset), &EXT4_I(inode)->i_##field); \ 1341 } 1342 1343 EXT4_INODE_BIT_FNS(flag, flags, 0) 1344 #if (BITS_PER_LONG < 64) 1345 EXT4_INODE_BIT_FNS(state, state_flags, 0) 1346 1347 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei) 1348 { 1349 (ei)->i_state_flags = 0; 1350 } 1351 #else 1352 EXT4_INODE_BIT_FNS(state, flags, 32) 1353 1354 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei) 1355 { 1356 /* We depend on the fact that callers will set i_flags */ 1357 } 1358 #endif 1359 #else 1360 /* Assume that user mode programs are passing in an ext4fs superblock, not 1361 * a kernel struct super_block. This will allow us to call the feature-test 1362 * macros from user land. */ 1363 #define EXT4_SB(sb) (sb) 1364 #endif 1365 1366 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime 1367 1368 /* 1369 * Codes for operating systems 1370 */ 1371 #define EXT4_OS_LINUX 0 1372 #define EXT4_OS_HURD 1 1373 #define EXT4_OS_MASIX 2 1374 #define EXT4_OS_FREEBSD 3 1375 #define EXT4_OS_LITES 4 1376 1377 /* 1378 * Revision levels 1379 */ 1380 #define EXT4_GOOD_OLD_REV 0 /* The good old (original) format */ 1381 #define EXT4_DYNAMIC_REV 1 /* V2 format w/ dynamic inode sizes */ 1382 1383 #define EXT4_CURRENT_REV EXT4_GOOD_OLD_REV 1384 #define EXT4_MAX_SUPP_REV EXT4_DYNAMIC_REV 1385 1386 #define EXT4_GOOD_OLD_INODE_SIZE 128 1387 1388 /* 1389 * Feature set definitions 1390 */ 1391 1392 #define EXT4_HAS_COMPAT_FEATURE(sb,mask) \ 1393 ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0) 1394 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask) \ 1395 ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0) 1396 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask) \ 1397 ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0) 1398 #define EXT4_SET_COMPAT_FEATURE(sb,mask) \ 1399 EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask) 1400 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask) \ 1401 EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask) 1402 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask) \ 1403 EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask) 1404 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask) \ 1405 EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask) 1406 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask) \ 1407 EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask) 1408 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask) \ 1409 EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask) 1410 1411 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC 0x0001 1412 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES 0x0002 1413 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL 0x0004 1414 #define EXT4_FEATURE_COMPAT_EXT_ATTR 0x0008 1415 #define EXT4_FEATURE_COMPAT_RESIZE_INODE 0x0010 1416 #define EXT4_FEATURE_COMPAT_DIR_INDEX 0x0020 1417 1418 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER 0x0001 1419 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE 0x0002 1420 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR 0x0004 1421 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE 0x0008 1422 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM 0x0010 1423 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK 0x0020 1424 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE 0x0040 1425 #define EXT4_FEATURE_RO_COMPAT_QUOTA 0x0100 1426 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC 0x0200 1427 /* 1428 * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM). When 1429 * METADATA_CSUM is set, group descriptor checksums use the same algorithm as 1430 * all other data structures' checksums. However, the METADATA_CSUM and 1431 * GDT_CSUM bits are mutually exclusive. 1432 */ 1433 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM 0x0400 1434 1435 #define EXT4_FEATURE_INCOMPAT_COMPRESSION 0x0001 1436 #define EXT4_FEATURE_INCOMPAT_FILETYPE 0x0002 1437 #define EXT4_FEATURE_INCOMPAT_RECOVER 0x0004 /* Needs recovery */ 1438 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 /* Journal device */ 1439 #define EXT4_FEATURE_INCOMPAT_META_BG 0x0010 1440 #define EXT4_FEATURE_INCOMPAT_EXTENTS 0x0040 /* extents support */ 1441 #define EXT4_FEATURE_INCOMPAT_64BIT 0x0080 1442 #define EXT4_FEATURE_INCOMPAT_MMP 0x0100 1443 #define EXT4_FEATURE_INCOMPAT_FLEX_BG 0x0200 1444 #define EXT4_FEATURE_INCOMPAT_EA_INODE 0x0400 /* EA in inode */ 1445 #define EXT4_FEATURE_INCOMPAT_DIRDATA 0x1000 /* data in dirent */ 1446 #define EXT4_FEATURE_INCOMPAT_BG_USE_META_CSUM 0x2000 /* use crc32c for bg */ 1447 #define EXT4_FEATURE_INCOMPAT_LARGEDIR 0x4000 /* >2GB or 3-lvl htree */ 1448 #define EXT4_FEATURE_INCOMPAT_INLINEDATA 0x8000 /* data in inode */ 1449 1450 #define EXT2_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR 1451 #define EXT2_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \ 1452 EXT4_FEATURE_INCOMPAT_META_BG) 1453 #define EXT2_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \ 1454 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \ 1455 EXT4_FEATURE_RO_COMPAT_BTREE_DIR) 1456 1457 #define EXT3_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR 1458 #define EXT3_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \ 1459 EXT4_FEATURE_INCOMPAT_RECOVER| \ 1460 EXT4_FEATURE_INCOMPAT_META_BG) 1461 #define EXT3_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \ 1462 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \ 1463 EXT4_FEATURE_RO_COMPAT_BTREE_DIR) 1464 1465 #define EXT4_FEATURE_COMPAT_SUPP EXT2_FEATURE_COMPAT_EXT_ATTR 1466 #define EXT4_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \ 1467 EXT4_FEATURE_INCOMPAT_RECOVER| \ 1468 EXT4_FEATURE_INCOMPAT_META_BG| \ 1469 EXT4_FEATURE_INCOMPAT_EXTENTS| \ 1470 EXT4_FEATURE_INCOMPAT_64BIT| \ 1471 EXT4_FEATURE_INCOMPAT_FLEX_BG| \ 1472 EXT4_FEATURE_INCOMPAT_MMP) 1473 #define EXT4_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \ 1474 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \ 1475 EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \ 1476 EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \ 1477 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \ 1478 EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\ 1479 EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\ 1480 EXT4_FEATURE_RO_COMPAT_BIGALLOC) 1481 1482 /* 1483 * Default values for user and/or group using reserved blocks 1484 */ 1485 #define EXT4_DEF_RESUID 0 1486 #define EXT4_DEF_RESGID 0 1487 1488 #define EXT4_DEF_INODE_READAHEAD_BLKS 32 1489 1490 /* 1491 * Default mount options 1492 */ 1493 #define EXT4_DEFM_DEBUG 0x0001 1494 #define EXT4_DEFM_BSDGROUPS 0x0002 1495 #define EXT4_DEFM_XATTR_USER 0x0004 1496 #define EXT4_DEFM_ACL 0x0008 1497 #define EXT4_DEFM_UID16 0x0010 1498 #define EXT4_DEFM_JMODE 0x0060 1499 #define EXT4_DEFM_JMODE_DATA 0x0020 1500 #define EXT4_DEFM_JMODE_ORDERED 0x0040 1501 #define EXT4_DEFM_JMODE_WBACK 0x0060 1502 #define EXT4_DEFM_NOBARRIER 0x0100 1503 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200 1504 #define EXT4_DEFM_DISCARD 0x0400 1505 #define EXT4_DEFM_NODELALLOC 0x0800 1506 1507 /* 1508 * Default journal batch times 1509 */ 1510 #define EXT4_DEF_MIN_BATCH_TIME 0 1511 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */ 1512 1513 /* 1514 * Minimum number of groups in a flexgroup before we separate out 1515 * directories into the first block group of a flexgroup 1516 */ 1517 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4 1518 1519 /* 1520 * Structure of a directory entry 1521 */ 1522 #define EXT4_NAME_LEN 255 1523 1524 struct ext4_dir_entry { 1525 __le32 inode; /* Inode number */ 1526 __le16 rec_len; /* Directory entry length */ 1527 __le16 name_len; /* Name length */ 1528 char name[EXT4_NAME_LEN]; /* File name */ 1529 }; 1530 1531 /* 1532 * The new version of the directory entry. Since EXT4 structures are 1533 * stored in intel byte order, and the name_len field could never be 1534 * bigger than 255 chars, it's safe to reclaim the extra byte for the 1535 * file_type field. 1536 */ 1537 struct ext4_dir_entry_2 { 1538 __le32 inode; /* Inode number */ 1539 __le16 rec_len; /* Directory entry length */ 1540 __u8 name_len; /* Name length */ 1541 __u8 file_type; 1542 char name[EXT4_NAME_LEN]; /* File name */ 1543 }; 1544 1545 /* 1546 * This is a bogus directory entry at the end of each leaf block that 1547 * records checksums. 1548 */ 1549 struct ext4_dir_entry_tail { 1550 __le32 det_reserved_zero1; /* Pretend to be unused */ 1551 __le16 det_rec_len; /* 12 */ 1552 __u8 det_reserved_zero2; /* Zero name length */ 1553 __u8 det_reserved_ft; /* 0xDE, fake file type */ 1554 __le32 det_checksum; /* crc32c(uuid+inum+dirblock) */ 1555 }; 1556 1557 /* 1558 * Ext4 directory file types. Only the low 3 bits are used. The 1559 * other bits are reserved for now. 1560 */ 1561 #define EXT4_FT_UNKNOWN 0 1562 #define EXT4_FT_REG_FILE 1 1563 #define EXT4_FT_DIR 2 1564 #define EXT4_FT_CHRDEV 3 1565 #define EXT4_FT_BLKDEV 4 1566 #define EXT4_FT_FIFO 5 1567 #define EXT4_FT_SOCK 6 1568 #define EXT4_FT_SYMLINK 7 1569 1570 #define EXT4_FT_MAX 8 1571 1572 #define EXT4_FT_DIR_CSUM 0xDE 1573 1574 /* 1575 * EXT4_DIR_PAD defines the directory entries boundaries 1576 * 1577 * NOTE: It must be a multiple of 4 1578 */ 1579 #define EXT4_DIR_PAD 4 1580 #define EXT4_DIR_ROUND (EXT4_DIR_PAD - 1) 1581 #define EXT4_DIR_REC_LEN(name_len) (((name_len) + 8 + EXT4_DIR_ROUND) & \ 1582 ~EXT4_DIR_ROUND) 1583 #define EXT4_MAX_REC_LEN ((1<<16)-1) 1584 1585 /* 1586 * If we ever get support for fs block sizes > page_size, we'll need 1587 * to remove the #if statements in the next two functions... 1588 */ 1589 static inline unsigned int 1590 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize) 1591 { 1592 unsigned len = le16_to_cpu(dlen); 1593 1594 #if (PAGE_CACHE_SIZE >= 65536) 1595 if (len == EXT4_MAX_REC_LEN || len == 0) 1596 return blocksize; 1597 return (len & 65532) | ((len & 3) << 16); 1598 #else 1599 return len; 1600 #endif 1601 } 1602 1603 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize) 1604 { 1605 if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3)) 1606 BUG(); 1607 #if (PAGE_CACHE_SIZE >= 65536) 1608 if (len < 65536) 1609 return cpu_to_le16(len); 1610 if (len == blocksize) { 1611 if (blocksize == 65536) 1612 return cpu_to_le16(EXT4_MAX_REC_LEN); 1613 else 1614 return cpu_to_le16(0); 1615 } 1616 return cpu_to_le16((len & 65532) | ((len >> 16) & 3)); 1617 #else 1618 return cpu_to_le16(len); 1619 #endif 1620 } 1621 1622 /* 1623 * Hash Tree Directory indexing 1624 * (c) Daniel Phillips, 2001 1625 */ 1626 1627 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \ 1628 EXT4_FEATURE_COMPAT_DIR_INDEX) && \ 1629 ext4_test_inode_flag((dir), EXT4_INODE_INDEX)) 1630 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX) 1631 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1) 1632 1633 /* Legal values for the dx_root hash_version field: */ 1634 1635 #define DX_HASH_LEGACY 0 1636 #define DX_HASH_HALF_MD4 1 1637 #define DX_HASH_TEA 2 1638 #define DX_HASH_LEGACY_UNSIGNED 3 1639 #define DX_HASH_HALF_MD4_UNSIGNED 4 1640 #define DX_HASH_TEA_UNSIGNED 5 1641 1642 #ifdef __KERNEL__ 1643 1644 /* hash info structure used by the directory hash */ 1645 struct dx_hash_info 1646 { 1647 u32 hash; 1648 u32 minor_hash; 1649 int hash_version; 1650 u32 *seed; 1651 }; 1652 1653 1654 /* 32 and 64 bit signed EOF for dx directories */ 1655 #define EXT4_HTREE_EOF_32BIT ((1UL << (32 - 1)) - 1) 1656 #define EXT4_HTREE_EOF_64BIT ((1ULL << (64 - 1)) - 1) 1657 1658 1659 /* 1660 * Control parameters used by ext4_htree_next_block 1661 */ 1662 #define HASH_NB_ALWAYS 1 1663 1664 1665 /* 1666 * Describe an inode's exact location on disk and in memory 1667 */ 1668 struct ext4_iloc 1669 { 1670 struct buffer_head *bh; 1671 unsigned long offset; 1672 ext4_group_t block_group; 1673 }; 1674 1675 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc) 1676 { 1677 return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset); 1678 } 1679 1680 /* 1681 * This structure is stuffed into the struct file's private_data field 1682 * for directories. It is where we put information so that we can do 1683 * readdir operations in hash tree order. 1684 */ 1685 struct dir_private_info { 1686 struct rb_root root; 1687 struct rb_node *curr_node; 1688 struct fname *extra_fname; 1689 loff_t last_pos; 1690 __u32 curr_hash; 1691 __u32 curr_minor_hash; 1692 __u32 next_hash; 1693 }; 1694 1695 /* calculate the first block number of the group */ 1696 static inline ext4_fsblk_t 1697 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no) 1698 { 1699 return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) + 1700 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block); 1701 } 1702 1703 /* 1704 * Special error return code only used by dx_probe() and its callers. 1705 */ 1706 #define ERR_BAD_DX_DIR -75000 1707 1708 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr, 1709 ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp); 1710 1711 /* 1712 * Timeout and state flag for lazy initialization inode thread. 1713 */ 1714 #define EXT4_DEF_LI_WAIT_MULT 10 1715 #define EXT4_DEF_LI_MAX_START_DELAY 5 1716 #define EXT4_LAZYINIT_QUIT 0x0001 1717 #define EXT4_LAZYINIT_RUNNING 0x0002 1718 1719 /* 1720 * Lazy inode table initialization info 1721 */ 1722 struct ext4_lazy_init { 1723 unsigned long li_state; 1724 struct list_head li_request_list; 1725 struct mutex li_list_mtx; 1726 }; 1727 1728 struct ext4_li_request { 1729 struct super_block *lr_super; 1730 struct ext4_sb_info *lr_sbi; 1731 ext4_group_t lr_next_group; 1732 struct list_head lr_request; 1733 unsigned long lr_next_sched; 1734 unsigned long lr_timeout; 1735 }; 1736 1737 struct ext4_features { 1738 struct kobject f_kobj; 1739 struct completion f_kobj_unregister; 1740 }; 1741 1742 /* 1743 * This structure will be used for multiple mount protection. It will be 1744 * written into the block number saved in the s_mmp_block field in the 1745 * superblock. Programs that check MMP should assume that if 1746 * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe 1747 * to use the filesystem, regardless of how old the timestamp is. 1748 */ 1749 #define EXT4_MMP_MAGIC 0x004D4D50U /* ASCII for MMP */ 1750 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */ 1751 #define EXT4_MMP_SEQ_FSCK 0xE24D4D50U /* mmp_seq value when being fscked */ 1752 #define EXT4_MMP_SEQ_MAX 0xE24D4D4FU /* maximum valid mmp_seq value */ 1753 1754 struct mmp_struct { 1755 __le32 mmp_magic; /* Magic number for MMP */ 1756 __le32 mmp_seq; /* Sequence no. updated periodically */ 1757 1758 /* 1759 * mmp_time, mmp_nodename & mmp_bdevname are only used for information 1760 * purposes and do not affect the correctness of the algorithm 1761 */ 1762 __le64 mmp_time; /* Time last updated */ 1763 char mmp_nodename[64]; /* Node which last updated MMP block */ 1764 char mmp_bdevname[32]; /* Bdev which last updated MMP block */ 1765 1766 /* 1767 * mmp_check_interval is used to verify if the MMP block has been 1768 * updated on the block device. The value is updated based on the 1769 * maximum time to write the MMP block during an update cycle. 1770 */ 1771 __le16 mmp_check_interval; 1772 1773 __le16 mmp_pad1; 1774 __le32 mmp_pad2[226]; 1775 __le32 mmp_checksum; /* crc32c(uuid+mmp_block) */ 1776 }; 1777 1778 /* arguments passed to the mmp thread */ 1779 struct mmpd_data { 1780 struct buffer_head *bh; /* bh from initial read_mmp_block() */ 1781 struct super_block *sb; /* super block of the fs */ 1782 }; 1783 1784 /* 1785 * Check interval multiplier 1786 * The MMP block is written every update interval and initially checked every 1787 * update interval x the multiplier (the value is then adapted based on the 1788 * write latency). The reason is that writes can be delayed under load and we 1789 * don't want readers to incorrectly assume that the filesystem is no longer 1790 * in use. 1791 */ 1792 #define EXT4_MMP_CHECK_MULT 2UL 1793 1794 /* 1795 * Minimum interval for MMP checking in seconds. 1796 */ 1797 #define EXT4_MMP_MIN_CHECK_INTERVAL 5UL 1798 1799 /* 1800 * Maximum interval for MMP checking in seconds. 1801 */ 1802 #define EXT4_MMP_MAX_CHECK_INTERVAL 300UL 1803 1804 /* 1805 * Function prototypes 1806 */ 1807 1808 /* 1809 * Ok, these declarations are also in <linux/kernel.h> but none of the 1810 * ext4 source programs needs to include it so they are duplicated here. 1811 */ 1812 # define NORET_TYPE /**/ 1813 # define ATTRIB_NORET __attribute__((noreturn)) 1814 # define NORET_AND noreturn, 1815 1816 /* bitmap.c */ 1817 extern unsigned int ext4_count_free(struct buffer_head *, unsigned); 1818 1819 /* balloc.c */ 1820 extern unsigned int ext4_block_group(struct super_block *sb, 1821 ext4_fsblk_t blocknr); 1822 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb, 1823 ext4_fsblk_t blocknr); 1824 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group); 1825 extern unsigned long ext4_bg_num_gdb(struct super_block *sb, 1826 ext4_group_t group); 1827 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode, 1828 ext4_fsblk_t goal, 1829 unsigned int flags, 1830 unsigned long *count, 1831 int *errp); 1832 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi, 1833 s64 nclusters, unsigned int flags); 1834 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *); 1835 extern void ext4_check_blocks_bitmap(struct super_block *); 1836 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb, 1837 ext4_group_t block_group, 1838 struct buffer_head ** bh); 1839 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries); 1840 1841 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb, 1842 ext4_group_t block_group); 1843 extern int ext4_wait_block_bitmap(struct super_block *sb, 1844 ext4_group_t block_group, 1845 struct buffer_head *bh); 1846 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb, 1847 ext4_group_t block_group); 1848 extern void ext4_init_block_bitmap(struct super_block *sb, 1849 struct buffer_head *bh, 1850 ext4_group_t group, 1851 struct ext4_group_desc *desc); 1852 extern unsigned ext4_free_clusters_after_init(struct super_block *sb, 1853 ext4_group_t block_group, 1854 struct ext4_group_desc *gdp); 1855 extern unsigned ext4_num_overhead_clusters(struct super_block *sb, 1856 ext4_group_t block_group, 1857 struct ext4_group_desc *gdp); 1858 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *); 1859 1860 /* dir.c */ 1861 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *, 1862 struct file *, 1863 struct ext4_dir_entry_2 *, 1864 struct buffer_head *, unsigned int); 1865 #define ext4_check_dir_entry(dir, filp, de, bh, offset) \ 1866 unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \ 1867 (de), (bh), (offset))) 1868 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash, 1869 __u32 minor_hash, 1870 struct ext4_dir_entry_2 *dirent); 1871 extern void ext4_htree_free_dir_info(struct dir_private_info *p); 1872 1873 /* fsync.c */ 1874 extern int ext4_sync_file(struct file *, loff_t, loff_t, int); 1875 extern int ext4_flush_completed_IO(struct inode *); 1876 1877 /* hash.c */ 1878 extern int ext4fs_dirhash(const char *name, int len, struct 1879 dx_hash_info *hinfo); 1880 1881 /* ialloc.c */ 1882 extern struct inode *ext4_new_inode(handle_t *, struct inode *, umode_t, 1883 const struct qstr *qstr, __u32 goal, 1884 uid_t *owner); 1885 extern void ext4_free_inode(handle_t *, struct inode *); 1886 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long); 1887 extern unsigned long ext4_count_free_inodes(struct super_block *); 1888 extern unsigned long ext4_count_dirs(struct super_block *); 1889 extern void ext4_check_inodes_bitmap(struct super_block *); 1890 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap); 1891 extern int ext4_init_inode_table(struct super_block *sb, 1892 ext4_group_t group, int barrier); 1893 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate); 1894 1895 /* mballoc.c */ 1896 extern long ext4_mb_stats; 1897 extern long ext4_mb_max_to_scan; 1898 extern int ext4_mb_init(struct super_block *, int); 1899 extern int ext4_mb_release(struct super_block *); 1900 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *, 1901 struct ext4_allocation_request *, int *); 1902 extern int ext4_mb_reserve_blocks(struct super_block *, int); 1903 extern void ext4_discard_preallocations(struct inode *); 1904 extern int __init ext4_init_mballoc(void); 1905 extern void ext4_exit_mballoc(void); 1906 extern void ext4_free_blocks(handle_t *handle, struct inode *inode, 1907 struct buffer_head *bh, ext4_fsblk_t block, 1908 unsigned long count, int flags); 1909 extern int ext4_mb_add_groupinfo(struct super_block *sb, 1910 ext4_group_t i, struct ext4_group_desc *desc); 1911 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb, 1912 ext4_fsblk_t block, unsigned long count); 1913 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *); 1914 1915 /* inode.c */ 1916 struct buffer_head *ext4_getblk(handle_t *, struct inode *, 1917 ext4_lblk_t, int, int *); 1918 struct buffer_head *ext4_bread(handle_t *, struct inode *, 1919 ext4_lblk_t, int, int *); 1920 int ext4_get_block(struct inode *inode, sector_t iblock, 1921 struct buffer_head *bh_result, int create); 1922 1923 extern struct inode *ext4_iget(struct super_block *, unsigned long); 1924 extern int ext4_write_inode(struct inode *, struct writeback_control *); 1925 extern int ext4_setattr(struct dentry *, struct iattr *); 1926 extern int ext4_getattr(struct vfsmount *mnt, struct dentry *dentry, 1927 struct kstat *stat); 1928 extern void ext4_evict_inode(struct inode *); 1929 extern void ext4_clear_inode(struct inode *); 1930 extern int ext4_sync_inode(handle_t *, struct inode *); 1931 extern void ext4_dirty_inode(struct inode *, int); 1932 extern int ext4_change_inode_journal_flag(struct inode *, int); 1933 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *); 1934 extern int ext4_can_truncate(struct inode *inode); 1935 extern void ext4_truncate(struct inode *); 1936 extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length); 1937 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks); 1938 extern void ext4_set_inode_flags(struct inode *); 1939 extern void ext4_get_inode_flags(struct ext4_inode_info *); 1940 extern int ext4_alloc_da_blocks(struct inode *inode); 1941 extern void ext4_set_aops(struct inode *inode); 1942 extern int ext4_writepage_trans_blocks(struct inode *); 1943 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks); 1944 extern int ext4_discard_partial_page_buffers(handle_t *handle, 1945 struct address_space *mapping, loff_t from, 1946 loff_t length, int flags); 1947 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf); 1948 extern qsize_t *ext4_get_reserved_space(struct inode *inode); 1949 extern void ext4_da_update_reserve_space(struct inode *inode, 1950 int used, int quota_claim); 1951 1952 /* indirect.c */ 1953 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode, 1954 struct ext4_map_blocks *map, int flags); 1955 extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb, 1956 const struct iovec *iov, loff_t offset, 1957 unsigned long nr_segs); 1958 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock); 1959 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks, int chunk); 1960 extern void ext4_ind_truncate(struct inode *inode); 1961 1962 /* ioctl.c */ 1963 extern long ext4_ioctl(struct file *, unsigned int, unsigned long); 1964 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long); 1965 1966 /* migrate.c */ 1967 extern int ext4_ext_migrate(struct inode *); 1968 1969 /* namei.c */ 1970 extern int ext4_orphan_add(handle_t *, struct inode *); 1971 extern int ext4_orphan_del(handle_t *, struct inode *); 1972 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash, 1973 __u32 start_minor_hash, __u32 *next_hash); 1974 1975 /* resize.c */ 1976 extern int ext4_group_add(struct super_block *sb, 1977 struct ext4_new_group_data *input); 1978 extern int ext4_group_extend(struct super_block *sb, 1979 struct ext4_super_block *es, 1980 ext4_fsblk_t n_blocks_count); 1981 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count); 1982 1983 /* super.c */ 1984 extern void *ext4_kvmalloc(size_t size, gfp_t flags); 1985 extern void *ext4_kvzalloc(size_t size, gfp_t flags); 1986 extern void ext4_kvfree(void *ptr); 1987 extern __printf(4, 5) 1988 void __ext4_error(struct super_block *, const char *, unsigned int, 1989 const char *, ...); 1990 #define ext4_error(sb, message...) __ext4_error(sb, __func__, \ 1991 __LINE__, ## message) 1992 extern __printf(5, 6) 1993 void ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t, 1994 const char *, ...); 1995 extern __printf(5, 6) 1996 void ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t, 1997 const char *, ...); 1998 extern void __ext4_std_error(struct super_block *, const char *, 1999 unsigned int, int); 2000 extern __printf(4, 5) 2001 void __ext4_abort(struct super_block *, const char *, unsigned int, 2002 const char *, ...); 2003 #define ext4_abort(sb, message...) __ext4_abort(sb, __func__, \ 2004 __LINE__, ## message) 2005 extern __printf(4, 5) 2006 void __ext4_warning(struct super_block *, const char *, unsigned int, 2007 const char *, ...); 2008 #define ext4_warning(sb, message...) __ext4_warning(sb, __func__, \ 2009 __LINE__, ## message) 2010 extern __printf(3, 4) 2011 void ext4_msg(struct super_block *, const char *, const char *, ...); 2012 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp, 2013 const char *, unsigned int, const char *); 2014 #define dump_mmp_msg(sb, mmp, msg) __dump_mmp_msg(sb, mmp, __func__, \ 2015 __LINE__, msg) 2016 extern __printf(7, 8) 2017 void __ext4_grp_locked_error(const char *, unsigned int, 2018 struct super_block *, ext4_group_t, 2019 unsigned long, ext4_fsblk_t, 2020 const char *, ...); 2021 #define ext4_grp_locked_error(sb, grp, message...) \ 2022 __ext4_grp_locked_error(__func__, __LINE__, (sb), (grp), ## message) 2023 extern void ext4_update_dynamic_rev(struct super_block *sb); 2024 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb, 2025 __u32 compat); 2026 extern int ext4_update_rocompat_feature(handle_t *handle, 2027 struct super_block *sb, __u32 rocompat); 2028 extern int ext4_update_incompat_feature(handle_t *handle, 2029 struct super_block *sb, __u32 incompat); 2030 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 2031 struct ext4_group_desc *bg); 2032 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 2033 struct ext4_group_desc *bg); 2034 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb, 2035 struct ext4_group_desc *bg); 2036 extern __u32 ext4_free_group_clusters(struct super_block *sb, 2037 struct ext4_group_desc *bg); 2038 extern __u32 ext4_free_inodes_count(struct super_block *sb, 2039 struct ext4_group_desc *bg); 2040 extern __u32 ext4_used_dirs_count(struct super_block *sb, 2041 struct ext4_group_desc *bg); 2042 extern __u32 ext4_itable_unused_count(struct super_block *sb, 2043 struct ext4_group_desc *bg); 2044 extern void ext4_block_bitmap_set(struct super_block *sb, 2045 struct ext4_group_desc *bg, ext4_fsblk_t blk); 2046 extern void ext4_inode_bitmap_set(struct super_block *sb, 2047 struct ext4_group_desc *bg, ext4_fsblk_t blk); 2048 extern void ext4_inode_table_set(struct super_block *sb, 2049 struct ext4_group_desc *bg, ext4_fsblk_t blk); 2050 extern void ext4_free_group_clusters_set(struct super_block *sb, 2051 struct ext4_group_desc *bg, 2052 __u32 count); 2053 extern void ext4_free_inodes_set(struct super_block *sb, 2054 struct ext4_group_desc *bg, __u32 count); 2055 extern void ext4_used_dirs_set(struct super_block *sb, 2056 struct ext4_group_desc *bg, __u32 count); 2057 extern void ext4_itable_unused_set(struct super_block *sb, 2058 struct ext4_group_desc *bg, __u32 count); 2059 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group, 2060 struct ext4_group_desc *gdp); 2061 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group, 2062 struct ext4_group_desc *gdp); 2063 2064 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es) 2065 { 2066 return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) | 2067 le32_to_cpu(es->s_blocks_count_lo); 2068 } 2069 2070 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es) 2071 { 2072 return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) | 2073 le32_to_cpu(es->s_r_blocks_count_lo); 2074 } 2075 2076 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es) 2077 { 2078 return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) | 2079 le32_to_cpu(es->s_free_blocks_count_lo); 2080 } 2081 2082 static inline void ext4_blocks_count_set(struct ext4_super_block *es, 2083 ext4_fsblk_t blk) 2084 { 2085 es->s_blocks_count_lo = cpu_to_le32((u32)blk); 2086 es->s_blocks_count_hi = cpu_to_le32(blk >> 32); 2087 } 2088 2089 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es, 2090 ext4_fsblk_t blk) 2091 { 2092 es->s_free_blocks_count_lo = cpu_to_le32((u32)blk); 2093 es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32); 2094 } 2095 2096 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es, 2097 ext4_fsblk_t blk) 2098 { 2099 es->s_r_blocks_count_lo = cpu_to_le32((u32)blk); 2100 es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32); 2101 } 2102 2103 static inline loff_t ext4_isize(struct ext4_inode *raw_inode) 2104 { 2105 if (S_ISREG(le16_to_cpu(raw_inode->i_mode))) 2106 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) | 2107 le32_to_cpu(raw_inode->i_size_lo); 2108 else 2109 return (loff_t) le32_to_cpu(raw_inode->i_size_lo); 2110 } 2111 2112 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size) 2113 { 2114 raw_inode->i_size_lo = cpu_to_le32(i_size); 2115 raw_inode->i_size_high = cpu_to_le32(i_size >> 32); 2116 } 2117 2118 static inline 2119 struct ext4_group_info *ext4_get_group_info(struct super_block *sb, 2120 ext4_group_t group) 2121 { 2122 struct ext4_group_info ***grp_info; 2123 long indexv, indexh; 2124 grp_info = EXT4_SB(sb)->s_group_info; 2125 indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb)); 2126 indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1); 2127 return grp_info[indexv][indexh]; 2128 } 2129 2130 /* 2131 * Reading s_groups_count requires using smp_rmb() afterwards. See 2132 * the locking protocol documented in the comments of ext4_group_add() 2133 * in resize.c 2134 */ 2135 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb) 2136 { 2137 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 2138 2139 smp_rmb(); 2140 return ngroups; 2141 } 2142 2143 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi, 2144 ext4_group_t block_group) 2145 { 2146 return block_group >> sbi->s_log_groups_per_flex; 2147 } 2148 2149 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi) 2150 { 2151 return 1 << sbi->s_log_groups_per_flex; 2152 } 2153 2154 #define ext4_std_error(sb, errno) \ 2155 do { \ 2156 if ((errno)) \ 2157 __ext4_std_error((sb), __func__, __LINE__, (errno)); \ 2158 } while (0) 2159 2160 #ifdef CONFIG_SMP 2161 /* Each CPU can accumulate percpu_counter_batch clusters in their local 2162 * counters. So we need to make sure we have free clusters more 2163 * than percpu_counter_batch * nr_cpu_ids. Also add a window of 4 times. 2164 */ 2165 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids)) 2166 #else 2167 #define EXT4_FREECLUSTERS_WATERMARK 0 2168 #endif 2169 2170 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize) 2171 { 2172 /* 2173 * XXX: replace with spinlock if seen contended -bzzz 2174 */ 2175 down_write(&EXT4_I(inode)->i_data_sem); 2176 if (newsize > EXT4_I(inode)->i_disksize) 2177 EXT4_I(inode)->i_disksize = newsize; 2178 up_write(&EXT4_I(inode)->i_data_sem); 2179 return ; 2180 } 2181 2182 struct ext4_group_info { 2183 unsigned long bb_state; 2184 struct rb_root bb_free_root; 2185 ext4_grpblk_t bb_first_free; /* first free block */ 2186 ext4_grpblk_t bb_free; /* total free blocks */ 2187 ext4_grpblk_t bb_fragments; /* nr of freespace fragments */ 2188 ext4_grpblk_t bb_largest_free_order;/* order of largest frag in BG */ 2189 struct list_head bb_prealloc_list; 2190 #ifdef DOUBLE_CHECK 2191 void *bb_bitmap; 2192 #endif 2193 struct rw_semaphore alloc_sem; 2194 ext4_grpblk_t bb_counters[]; /* Nr of free power-of-two-block 2195 * regions, index is order. 2196 * bb_counters[3] = 5 means 2197 * 5 free 8-block regions. */ 2198 }; 2199 2200 #define EXT4_GROUP_INFO_NEED_INIT_BIT 0 2201 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT 1 2202 2203 #define EXT4_MB_GRP_NEED_INIT(grp) \ 2204 (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state))) 2205 2206 #define EXT4_MB_GRP_WAS_TRIMMED(grp) \ 2207 (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state))) 2208 #define EXT4_MB_GRP_SET_TRIMMED(grp) \ 2209 (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state))) 2210 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp) \ 2211 (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state))) 2212 2213 #define EXT4_MAX_CONTENTION 8 2214 #define EXT4_CONTENTION_THRESHOLD 2 2215 2216 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb, 2217 ext4_group_t group) 2218 { 2219 return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group); 2220 } 2221 2222 /* 2223 * Returns true if the filesystem is busy enough that attempts to 2224 * access the block group locks has run into contention. 2225 */ 2226 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi) 2227 { 2228 return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD); 2229 } 2230 2231 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group) 2232 { 2233 spinlock_t *lock = ext4_group_lock_ptr(sb, group); 2234 if (spin_trylock(lock)) 2235 /* 2236 * We're able to grab the lock right away, so drop the 2237 * lock contention counter. 2238 */ 2239 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0); 2240 else { 2241 /* 2242 * The lock is busy, so bump the contention counter, 2243 * and then wait on the spin lock. 2244 */ 2245 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1, 2246 EXT4_MAX_CONTENTION); 2247 spin_lock(lock); 2248 } 2249 } 2250 2251 static inline void ext4_unlock_group(struct super_block *sb, 2252 ext4_group_t group) 2253 { 2254 spin_unlock(ext4_group_lock_ptr(sb, group)); 2255 } 2256 2257 static inline void ext4_mark_super_dirty(struct super_block *sb) 2258 { 2259 if (EXT4_SB(sb)->s_journal == NULL) 2260 sb->s_dirt =1; 2261 } 2262 2263 /* 2264 * Block validity checking 2265 */ 2266 #define ext4_check_indirect_blockref(inode, bh) \ 2267 ext4_check_blockref(__func__, __LINE__, inode, \ 2268 (__le32 *)(bh)->b_data, \ 2269 EXT4_ADDR_PER_BLOCK((inode)->i_sb)) 2270 2271 #define ext4_ind_check_inode(inode) \ 2272 ext4_check_blockref(__func__, __LINE__, inode, \ 2273 EXT4_I(inode)->i_data, \ 2274 EXT4_NDIR_BLOCKS) 2275 2276 /* 2277 * Inodes and files operations 2278 */ 2279 2280 /* dir.c */ 2281 extern const struct file_operations ext4_dir_operations; 2282 2283 /* file.c */ 2284 extern const struct inode_operations ext4_file_inode_operations; 2285 extern const struct file_operations ext4_file_operations; 2286 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin); 2287 2288 /* namei.c */ 2289 extern const struct inode_operations ext4_dir_inode_operations; 2290 extern const struct inode_operations ext4_special_inode_operations; 2291 extern struct dentry *ext4_get_parent(struct dentry *child); 2292 2293 /* symlink.c */ 2294 extern const struct inode_operations ext4_symlink_inode_operations; 2295 extern const struct inode_operations ext4_fast_symlink_inode_operations; 2296 2297 /* block_validity */ 2298 extern void ext4_release_system_zone(struct super_block *sb); 2299 extern int ext4_setup_system_zone(struct super_block *sb); 2300 extern int __init ext4_init_system_zone(void); 2301 extern void ext4_exit_system_zone(void); 2302 extern int ext4_data_block_valid(struct ext4_sb_info *sbi, 2303 ext4_fsblk_t start_blk, 2304 unsigned int count); 2305 extern int ext4_check_blockref(const char *, unsigned int, 2306 struct inode *, __le32 *, unsigned int); 2307 2308 /* extents.c */ 2309 extern int ext4_ext_tree_init(handle_t *handle, struct inode *); 2310 extern int ext4_ext_writepage_trans_blocks(struct inode *, int); 2311 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks, 2312 int chunk); 2313 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode, 2314 struct ext4_map_blocks *map, int flags); 2315 extern void ext4_ext_truncate(struct inode *); 2316 extern int ext4_ext_punch_hole(struct file *file, loff_t offset, 2317 loff_t length); 2318 extern void ext4_ext_init(struct super_block *); 2319 extern void ext4_ext_release(struct super_block *); 2320 extern long ext4_fallocate(struct file *file, int mode, loff_t offset, 2321 loff_t len); 2322 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset, 2323 ssize_t len); 2324 extern int ext4_map_blocks(handle_t *handle, struct inode *inode, 2325 struct ext4_map_blocks *map, int flags); 2326 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, 2327 __u64 start, __u64 len); 2328 /* move_extent.c */ 2329 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp, 2330 __u64 start_orig, __u64 start_donor, 2331 __u64 len, __u64 *moved_len); 2332 2333 /* page-io.c */ 2334 extern int __init ext4_init_pageio(void); 2335 extern void ext4_exit_pageio(void); 2336 extern void ext4_ioend_wait(struct inode *); 2337 extern void ext4_free_io_end(ext4_io_end_t *io); 2338 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags); 2339 extern int ext4_end_io_nolock(ext4_io_end_t *io); 2340 extern void ext4_io_submit(struct ext4_io_submit *io); 2341 extern int ext4_bio_write_page(struct ext4_io_submit *io, 2342 struct page *page, 2343 int len, 2344 struct writeback_control *wbc); 2345 2346 /* mmp.c */ 2347 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t); 2348 2349 /* BH_Uninit flag: blocks are allocated but uninitialized on disk */ 2350 enum ext4_state_bits { 2351 BH_Uninit /* blocks are allocated but uninitialized on disk */ 2352 = BH_JBDPrivateStart, 2353 BH_AllocFromCluster, /* allocated blocks were part of already 2354 * allocated cluster. Note that this flag will 2355 * never, ever appear in a buffer_head's state 2356 * flag. See EXT4_MAP_FROM_CLUSTER to see where 2357 * this is used. */ 2358 BH_Da_Mapped, /* Delayed allocated block that now has a mapping. This 2359 * flag is set when ext4_map_blocks is called on a 2360 * delayed allocated block to get its real mapping. */ 2361 }; 2362 2363 BUFFER_FNS(Uninit, uninit) 2364 TAS_BUFFER_FNS(Uninit, uninit) 2365 BUFFER_FNS(Da_Mapped, da_mapped) 2366 2367 /* 2368 * Add new method to test wether block and inode bitmaps are properly 2369 * initialized. With uninit_bg reading the block from disk is not enough 2370 * to mark the bitmap uptodate. We need to also zero-out the bitmap 2371 */ 2372 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart 2373 2374 static inline int bitmap_uptodate(struct buffer_head *bh) 2375 { 2376 return (buffer_uptodate(bh) && 2377 test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state)); 2378 } 2379 static inline void set_bitmap_uptodate(struct buffer_head *bh) 2380 { 2381 set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state); 2382 } 2383 2384 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1) 2385 2386 /* For ioend & aio unwritten conversion wait queues */ 2387 #define EXT4_WQ_HASH_SZ 37 2388 #define ext4_ioend_wq(v) (&ext4__ioend_wq[((unsigned long)(v)) %\ 2389 EXT4_WQ_HASH_SZ]) 2390 #define ext4_aio_mutex(v) (&ext4__aio_mutex[((unsigned long)(v)) %\ 2391 EXT4_WQ_HASH_SZ]) 2392 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 2393 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ]; 2394 2395 #define EXT4_RESIZING 0 2396 extern int ext4_resize_begin(struct super_block *sb); 2397 extern void ext4_resize_end(struct super_block *sb); 2398 2399 #endif /* __KERNEL__ */ 2400 2401 #include "ext4_extents.h" 2402 2403 #endif /* _EXT4_H */ 2404