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 #include <crypto/hash.h> 33 #ifdef __KERNEL__ 34 #include <linux/compat.h> 35 #endif 36 37 /* 38 * The fourth extended filesystem constants/structures 39 */ 40 41 /* 42 * Define EXT4FS_DEBUG to produce debug messages 43 */ 44 #undef EXT4FS_DEBUG 45 46 /* 47 * Debug code 48 */ 49 #ifdef EXT4FS_DEBUG 50 #define ext4_debug(f, a...) \ 51 do { \ 52 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:", \ 53 __FILE__, __LINE__, __func__); \ 54 printk(KERN_DEBUG f, ## a); \ 55 } while (0) 56 #else 57 #define ext4_debug(fmt, ...) no_printk(fmt, ##__VA_ARGS__) 58 #endif 59 60 #define EXT4_ERROR_INODE(inode, fmt, a...) \ 61 ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a) 62 63 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...) \ 64 ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a) 65 66 #define EXT4_ERROR_FILE(file, block, fmt, a...) \ 67 ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a) 68 69 /* data type for block offset of block group */ 70 typedef int ext4_grpblk_t; 71 72 /* data type for filesystem-wide blocks number */ 73 typedef unsigned long long ext4_fsblk_t; 74 75 /* data type for file logical block number */ 76 typedef __u32 ext4_lblk_t; 77 78 /* data type for block group number */ 79 typedef unsigned int ext4_group_t; 80 81 /* 82 * Flags used in mballoc's allocation_context flags field. 83 * 84 * Also used to show what's going on for debugging purposes when the 85 * flag field is exported via the traceport interface 86 */ 87 88 /* prefer goal again. length */ 89 #define EXT4_MB_HINT_MERGE 0x0001 90 /* blocks already reserved */ 91 #define EXT4_MB_HINT_RESERVED 0x0002 92 /* metadata is being allocated */ 93 #define EXT4_MB_HINT_METADATA 0x0004 94 /* first blocks in the file */ 95 #define EXT4_MB_HINT_FIRST 0x0008 96 /* search for the best chunk */ 97 #define EXT4_MB_HINT_BEST 0x0010 98 /* data is being allocated */ 99 #define EXT4_MB_HINT_DATA 0x0020 100 /* don't preallocate (for tails) */ 101 #define EXT4_MB_HINT_NOPREALLOC 0x0040 102 /* allocate for locality group */ 103 #define EXT4_MB_HINT_GROUP_ALLOC 0x0080 104 /* allocate goal blocks or none */ 105 #define EXT4_MB_HINT_GOAL_ONLY 0x0100 106 /* goal is meaningful */ 107 #define EXT4_MB_HINT_TRY_GOAL 0x0200 108 /* blocks already pre-reserved by delayed allocation */ 109 #define EXT4_MB_DELALLOC_RESERVED 0x0400 110 /* We are doing stream allocation */ 111 #define EXT4_MB_STREAM_ALLOC 0x0800 112 /* Use reserved root blocks if needed */ 113 #define EXT4_MB_USE_ROOT_BLOCKS 0x1000 114 115 struct ext4_allocation_request { 116 /* target inode for block we're allocating */ 117 struct inode *inode; 118 /* how many blocks we want to allocate */ 119 unsigned int len; 120 /* logical block in target inode */ 121 ext4_lblk_t logical; 122 /* the closest logical allocated block to the left */ 123 ext4_lblk_t lleft; 124 /* the closest logical allocated block to the right */ 125 ext4_lblk_t lright; 126 /* phys. target (a hint) */ 127 ext4_fsblk_t goal; 128 /* phys. block for the closest logical allocated block to the left */ 129 ext4_fsblk_t pleft; 130 /* phys. block for the closest logical allocated block to the right */ 131 ext4_fsblk_t pright; 132 /* flags. see above EXT4_MB_HINT_* */ 133 unsigned int flags; 134 }; 135 136 /* 137 * Logical to physical block mapping, used by ext4_map_blocks() 138 * 139 * This structure is used to pass requests into ext4_map_blocks() as 140 * well as to store the information returned by ext4_map_blocks(). It 141 * takes less room on the stack than a struct buffer_head. 142 */ 143 #define EXT4_MAP_NEW (1 << BH_New) 144 #define EXT4_MAP_MAPPED (1 << BH_Mapped) 145 #define EXT4_MAP_UNWRITTEN (1 << BH_Unwritten) 146 #define EXT4_MAP_BOUNDARY (1 << BH_Boundary) 147 #define EXT4_MAP_UNINIT (1 << BH_Uninit) 148 /* Sometimes (in the bigalloc case, from ext4_da_get_block_prep) the caller of 149 * ext4_map_blocks wants to know whether or not the underlying cluster has 150 * already been accounted for. EXT4_MAP_FROM_CLUSTER conveys to the caller that 151 * the requested mapping was from previously mapped (or delayed allocated) 152 * cluster. We use BH_AllocFromCluster only for this flag. BH_AllocFromCluster 153 * should never appear on buffer_head's state flags. 154 */ 155 #define EXT4_MAP_FROM_CLUSTER (1 << BH_AllocFromCluster) 156 #define EXT4_MAP_FLAGS (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\ 157 EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\ 158 EXT4_MAP_UNINIT | EXT4_MAP_FROM_CLUSTER) 159 160 struct ext4_map_blocks { 161 ext4_fsblk_t m_pblk; 162 ext4_lblk_t m_lblk; 163 unsigned int m_len; 164 unsigned int m_flags; 165 }; 166 167 /* 168 * For delayed allocation tracking 169 */ 170 struct mpage_da_data { 171 struct inode *inode; 172 sector_t b_blocknr; /* start block number of extent */ 173 size_t b_size; /* size of extent */ 174 unsigned long b_state; /* state of the extent */ 175 unsigned long first_page, next_page; /* extent of pages */ 176 struct writeback_control *wbc; 177 int io_done; 178 int pages_written; 179 int retval; 180 }; 181 182 /* 183 * Flags for ext4_io_end->flags 184 */ 185 #define EXT4_IO_END_UNWRITTEN 0x0001 186 #define EXT4_IO_END_ERROR 0x0002 187 #define EXT4_IO_END_QUEUED 0x0004 188 #define EXT4_IO_END_DIRECT 0x0008 189 #define EXT4_IO_END_IN_FSYNC 0x0010 190 191 struct ext4_io_page { 192 struct page *p_page; 193 atomic_t p_count; 194 }; 195 196 #define MAX_IO_PAGES 128 197 198 /* 199 * For converting uninitialized extents on a work queue. 200 * 201 * 'page' is only used from the writepage() path; 'pages' is only used for 202 * buffered writes; they are used to keep page references until conversion 203 * takes place. For AIO/DIO, neither field is filled in. 204 */ 205 typedef struct ext4_io_end { 206 struct list_head list; /* per-file finished IO list */ 207 struct inode *inode; /* file being written to */ 208 unsigned int flag; /* unwritten or not */ 209 struct page *page; /* for writepage() path */ 210 loff_t offset; /* offset in the file */ 211 ssize_t size; /* size of the extent */ 212 struct work_struct work; /* data work queue */ 213 struct kiocb *iocb; /* iocb struct for AIO */ 214 int result; /* error value for AIO */ 215 int num_io_pages; /* for writepages() */ 216 struct ext4_io_page *pages[MAX_IO_PAGES]; /* for writepages() */ 217 } ext4_io_end_t; 218 219 struct ext4_io_submit { 220 int io_op; 221 struct bio *io_bio; 222 ext4_io_end_t *io_end; 223 struct ext4_io_page *io_page; 224 sector_t io_next_block; 225 }; 226 227 /* 228 * Special inodes numbers 229 */ 230 #define EXT4_BAD_INO 1 /* Bad blocks inode */ 231 #define EXT4_ROOT_INO 2 /* Root inode */ 232 #define EXT4_USR_QUOTA_INO 3 /* User quota inode */ 233 #define EXT4_GRP_QUOTA_INO 4 /* Group quota inode */ 234 #define EXT4_BOOT_LOADER_INO 5 /* Boot loader inode */ 235 #define EXT4_UNDEL_DIR_INO 6 /* Undelete directory inode */ 236 #define EXT4_RESIZE_INO 7 /* Reserved group descriptors inode */ 237 #define EXT4_JOURNAL_INO 8 /* Journal inode */ 238 239 /* First non-reserved inode for old ext4 filesystems */ 240 #define EXT4_GOOD_OLD_FIRST_INO 11 241 242 /* 243 * Maximal count of links to a file 244 */ 245 #define EXT4_LINK_MAX 65000 246 247 /* 248 * Macro-instructions used to manage several block sizes 249 */ 250 #define EXT4_MIN_BLOCK_SIZE 1024 251 #define EXT4_MAX_BLOCK_SIZE 65536 252 #define EXT4_MIN_BLOCK_LOG_SIZE 10 253 #define EXT4_MAX_BLOCK_LOG_SIZE 16 254 #ifdef __KERNEL__ 255 # define EXT4_BLOCK_SIZE(s) ((s)->s_blocksize) 256 #else 257 # define EXT4_BLOCK_SIZE(s) (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size) 258 #endif 259 #define EXT4_ADDR_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / sizeof(__u32)) 260 #define EXT4_CLUSTER_SIZE(s) (EXT4_BLOCK_SIZE(s) << \ 261 EXT4_SB(s)->s_cluster_bits) 262 #ifdef __KERNEL__ 263 # define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits) 264 # define EXT4_CLUSTER_BITS(s) (EXT4_SB(s)->s_cluster_bits) 265 #else 266 # define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_log_block_size + 10) 267 #endif 268 #ifdef __KERNEL__ 269 #define EXT4_ADDR_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_addr_per_block_bits) 270 #define EXT4_INODE_SIZE(s) (EXT4_SB(s)->s_inode_size) 271 #define EXT4_FIRST_INO(s) (EXT4_SB(s)->s_first_ino) 272 #else 273 #define EXT4_INODE_SIZE(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \ 274 EXT4_GOOD_OLD_INODE_SIZE : \ 275 (s)->s_inode_size) 276 #define EXT4_FIRST_INO(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \ 277 EXT4_GOOD_OLD_FIRST_INO : \ 278 (s)->s_first_ino) 279 #endif 280 #define EXT4_BLOCK_ALIGN(size, blkbits) ALIGN((size), (1 << (blkbits))) 281 282 /* Translate a block number to a cluster number */ 283 #define EXT4_B2C(sbi, blk) ((blk) >> (sbi)->s_cluster_bits) 284 /* Translate a cluster number to a block number */ 285 #define EXT4_C2B(sbi, cluster) ((cluster) << (sbi)->s_cluster_bits) 286 /* Translate # of blks to # of clusters */ 287 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \ 288 (sbi)->s_cluster_bits) 289 290 /* 291 * Structure of a blocks group descriptor 292 */ 293 struct ext4_group_desc 294 { 295 __le32 bg_block_bitmap_lo; /* Blocks bitmap block */ 296 __le32 bg_inode_bitmap_lo; /* Inodes bitmap block */ 297 __le32 bg_inode_table_lo; /* Inodes table block */ 298 __le16 bg_free_blocks_count_lo;/* Free blocks count */ 299 __le16 bg_free_inodes_count_lo;/* Free inodes count */ 300 __le16 bg_used_dirs_count_lo; /* Directories count */ 301 __le16 bg_flags; /* EXT4_BG_flags (INODE_UNINIT, etc) */ 302 __le32 bg_exclude_bitmap_lo; /* Exclude bitmap for snapshots */ 303 __le16 bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */ 304 __le16 bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */ 305 __le16 bg_itable_unused_lo; /* Unused inodes count */ 306 __le16 bg_checksum; /* crc16(sb_uuid+group+desc) */ 307 __le32 bg_block_bitmap_hi; /* Blocks bitmap block MSB */ 308 __le32 bg_inode_bitmap_hi; /* Inodes bitmap block MSB */ 309 __le32 bg_inode_table_hi; /* Inodes table block MSB */ 310 __le16 bg_free_blocks_count_hi;/* Free blocks count MSB */ 311 __le16 bg_free_inodes_count_hi;/* Free inodes count MSB */ 312 __le16 bg_used_dirs_count_hi; /* Directories count MSB */ 313 __le16 bg_itable_unused_hi; /* Unused inodes count MSB */ 314 __le32 bg_exclude_bitmap_hi; /* Exclude bitmap block MSB */ 315 __le16 bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */ 316 __le16 bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */ 317 __u32 bg_reserved; 318 }; 319 320 /* 321 * Structure of a flex block group info 322 */ 323 324 struct flex_groups { 325 atomic_t free_inodes; 326 atomic_t free_clusters; 327 atomic_t used_dirs; 328 }; 329 330 #define EXT4_BG_INODE_UNINIT 0x0001 /* Inode table/bitmap not in use */ 331 #define EXT4_BG_BLOCK_UNINIT 0x0002 /* Block bitmap not in use */ 332 #define EXT4_BG_INODE_ZEROED 0x0004 /* On-disk itable initialized to zero */ 333 334 /* 335 * Macro-instructions used to manage group descriptors 336 */ 337 #define EXT4_MIN_DESC_SIZE 32 338 #define EXT4_MIN_DESC_SIZE_64BIT 64 339 #define EXT4_MAX_DESC_SIZE EXT4_MIN_BLOCK_SIZE 340 #define EXT4_DESC_SIZE(s) (EXT4_SB(s)->s_desc_size) 341 #ifdef __KERNEL__ 342 # define EXT4_BLOCKS_PER_GROUP(s) (EXT4_SB(s)->s_blocks_per_group) 343 # define EXT4_CLUSTERS_PER_GROUP(s) (EXT4_SB(s)->s_clusters_per_group) 344 # define EXT4_DESC_PER_BLOCK(s) (EXT4_SB(s)->s_desc_per_block) 345 # define EXT4_INODES_PER_GROUP(s) (EXT4_SB(s)->s_inodes_per_group) 346 # define EXT4_DESC_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_desc_per_block_bits) 347 #else 348 # define EXT4_BLOCKS_PER_GROUP(s) ((s)->s_blocks_per_group) 349 # define EXT4_DESC_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s)) 350 # define EXT4_INODES_PER_GROUP(s) ((s)->s_inodes_per_group) 351 #endif 352 353 /* 354 * Constants relative to the data blocks 355 */ 356 #define EXT4_NDIR_BLOCKS 12 357 #define EXT4_IND_BLOCK EXT4_NDIR_BLOCKS 358 #define EXT4_DIND_BLOCK (EXT4_IND_BLOCK + 1) 359 #define EXT4_TIND_BLOCK (EXT4_DIND_BLOCK + 1) 360 #define EXT4_N_BLOCKS (EXT4_TIND_BLOCK + 1) 361 362 /* 363 * Inode flags 364 */ 365 #define EXT4_SECRM_FL 0x00000001 /* Secure deletion */ 366 #define EXT4_UNRM_FL 0x00000002 /* Undelete */ 367 #define EXT4_COMPR_FL 0x00000004 /* Compress file */ 368 #define EXT4_SYNC_FL 0x00000008 /* Synchronous updates */ 369 #define EXT4_IMMUTABLE_FL 0x00000010 /* Immutable file */ 370 #define EXT4_APPEND_FL 0x00000020 /* writes to file may only append */ 371 #define EXT4_NODUMP_FL 0x00000040 /* do not dump file */ 372 #define EXT4_NOATIME_FL 0x00000080 /* do not update atime */ 373 /* Reserved for compression usage... */ 374 #define EXT4_DIRTY_FL 0x00000100 375 #define EXT4_COMPRBLK_FL 0x00000200 /* One or more compressed clusters */ 376 #define EXT4_NOCOMPR_FL 0x00000400 /* Don't compress */ 377 #define EXT4_ECOMPR_FL 0x00000800 /* Compression error */ 378 /* End compression flags --- maybe not all used */ 379 #define EXT4_INDEX_FL 0x00001000 /* hash-indexed directory */ 380 #define EXT4_IMAGIC_FL 0x00002000 /* AFS directory */ 381 #define EXT4_JOURNAL_DATA_FL 0x00004000 /* file data should be journaled */ 382 #define EXT4_NOTAIL_FL 0x00008000 /* file tail should not be merged */ 383 #define EXT4_DIRSYNC_FL 0x00010000 /* dirsync behaviour (directories only) */ 384 #define EXT4_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/ 385 #define EXT4_HUGE_FILE_FL 0x00040000 /* Set to each huge file */ 386 #define EXT4_EXTENTS_FL 0x00080000 /* Inode uses extents */ 387 #define EXT4_EA_INODE_FL 0x00200000 /* Inode used for large EA */ 388 #define EXT4_EOFBLOCKS_FL 0x00400000 /* Blocks allocated beyond EOF */ 389 #define EXT4_RESERVED_FL 0x80000000 /* reserved for ext4 lib */ 390 391 #define EXT4_FL_USER_VISIBLE 0x004BDFFF /* User visible flags */ 392 #define EXT4_FL_USER_MODIFIABLE 0x004B80FF /* User modifiable flags */ 393 394 /* Flags that should be inherited by new inodes from their parent. */ 395 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\ 396 EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\ 397 EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\ 398 EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL) 399 400 /* Flags that are appropriate for regular files (all but dir-specific ones). */ 401 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL)) 402 403 /* Flags that are appropriate for non-directories/regular files. */ 404 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL) 405 406 /* Mask out flags that are inappropriate for the given type of inode. */ 407 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags) 408 { 409 if (S_ISDIR(mode)) 410 return flags; 411 else if (S_ISREG(mode)) 412 return flags & EXT4_REG_FLMASK; 413 else 414 return flags & EXT4_OTHER_FLMASK; 415 } 416 417 /* 418 * Inode flags used for atomic set/get 419 */ 420 enum { 421 EXT4_INODE_SECRM = 0, /* Secure deletion */ 422 EXT4_INODE_UNRM = 1, /* Undelete */ 423 EXT4_INODE_COMPR = 2, /* Compress file */ 424 EXT4_INODE_SYNC = 3, /* Synchronous updates */ 425 EXT4_INODE_IMMUTABLE = 4, /* Immutable file */ 426 EXT4_INODE_APPEND = 5, /* writes to file may only append */ 427 EXT4_INODE_NODUMP = 6, /* do not dump file */ 428 EXT4_INODE_NOATIME = 7, /* do not update atime */ 429 /* Reserved for compression usage... */ 430 EXT4_INODE_DIRTY = 8, 431 EXT4_INODE_COMPRBLK = 9, /* One or more compressed clusters */ 432 EXT4_INODE_NOCOMPR = 10, /* Don't compress */ 433 EXT4_INODE_ECOMPR = 11, /* Compression error */ 434 /* End compression flags --- maybe not all used */ 435 EXT4_INODE_INDEX = 12, /* hash-indexed directory */ 436 EXT4_INODE_IMAGIC = 13, /* AFS directory */ 437 EXT4_INODE_JOURNAL_DATA = 14, /* file data should be journaled */ 438 EXT4_INODE_NOTAIL = 15, /* file tail should not be merged */ 439 EXT4_INODE_DIRSYNC = 16, /* dirsync behaviour (directories only) */ 440 EXT4_INODE_TOPDIR = 17, /* Top of directory hierarchies*/ 441 EXT4_INODE_HUGE_FILE = 18, /* Set to each huge file */ 442 EXT4_INODE_EXTENTS = 19, /* Inode uses extents */ 443 EXT4_INODE_EA_INODE = 21, /* Inode used for large EA */ 444 EXT4_INODE_EOFBLOCKS = 22, /* Blocks allocated beyond EOF */ 445 EXT4_INODE_RESERVED = 31, /* reserved for ext4 lib */ 446 }; 447 448 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG)) 449 #define CHECK_FLAG_VALUE(FLAG) if (!TEST_FLAG_VALUE(FLAG)) { \ 450 printk(KERN_EMERG "EXT4 flag fail: " #FLAG ": %d %d\n", \ 451 EXT4_##FLAG##_FL, EXT4_INODE_##FLAG); BUG_ON(1); } 452 453 /* 454 * Since it's pretty easy to mix up bit numbers and hex values, and we 455 * can't do a compile-time test for ENUM values, we use a run-time 456 * test to make sure that EXT4_XXX_FL is consistent with respect to 457 * EXT4_INODE_XXX. If all is well the printk and BUG_ON will all drop 458 * out so it won't cost any extra space in the compiled kernel image. 459 * But it's important that these values are the same, since we are 460 * using EXT4_INODE_XXX to test for the flag values, but EXT4_XX_FL 461 * must be consistent with the values of FS_XXX_FL defined in 462 * include/linux/fs.h and the on-disk values found in ext2, ext3, and 463 * ext4 filesystems, and of course the values defined in e2fsprogs. 464 * 465 * It's not paranoia if the Murphy's Law really *is* out to get you. :-) 466 */ 467 static inline void ext4_check_flag_values(void) 468 { 469 CHECK_FLAG_VALUE(SECRM); 470 CHECK_FLAG_VALUE(UNRM); 471 CHECK_FLAG_VALUE(COMPR); 472 CHECK_FLAG_VALUE(SYNC); 473 CHECK_FLAG_VALUE(IMMUTABLE); 474 CHECK_FLAG_VALUE(APPEND); 475 CHECK_FLAG_VALUE(NODUMP); 476 CHECK_FLAG_VALUE(NOATIME); 477 CHECK_FLAG_VALUE(DIRTY); 478 CHECK_FLAG_VALUE(COMPRBLK); 479 CHECK_FLAG_VALUE(NOCOMPR); 480 CHECK_FLAG_VALUE(ECOMPR); 481 CHECK_FLAG_VALUE(INDEX); 482 CHECK_FLAG_VALUE(IMAGIC); 483 CHECK_FLAG_VALUE(JOURNAL_DATA); 484 CHECK_FLAG_VALUE(NOTAIL); 485 CHECK_FLAG_VALUE(DIRSYNC); 486 CHECK_FLAG_VALUE(TOPDIR); 487 CHECK_FLAG_VALUE(HUGE_FILE); 488 CHECK_FLAG_VALUE(EXTENTS); 489 CHECK_FLAG_VALUE(EA_INODE); 490 CHECK_FLAG_VALUE(EOFBLOCKS); 491 CHECK_FLAG_VALUE(RESERVED); 492 } 493 494 /* Used to pass group descriptor data when online resize is done */ 495 struct ext4_new_group_input { 496 __u32 group; /* Group number for this data */ 497 __u64 block_bitmap; /* Absolute block number of block bitmap */ 498 __u64 inode_bitmap; /* Absolute block number of inode bitmap */ 499 __u64 inode_table; /* Absolute block number of inode table start */ 500 __u32 blocks_count; /* Total number of blocks in this group */ 501 __u16 reserved_blocks; /* Number of reserved blocks in this group */ 502 __u16 unused; 503 }; 504 505 #if defined(__KERNEL__) && defined(CONFIG_COMPAT) 506 struct compat_ext4_new_group_input { 507 u32 group; 508 compat_u64 block_bitmap; 509 compat_u64 inode_bitmap; 510 compat_u64 inode_table; 511 u32 blocks_count; 512 u16 reserved_blocks; 513 u16 unused; 514 }; 515 #endif 516 517 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */ 518 struct ext4_new_group_data { 519 __u32 group; 520 __u64 block_bitmap; 521 __u64 inode_bitmap; 522 __u64 inode_table; 523 __u32 blocks_count; 524 __u16 reserved_blocks; 525 __u16 unused; 526 __u32 free_blocks_count; 527 }; 528 529 /* Indexes used to index group tables in ext4_new_group_data */ 530 enum { 531 BLOCK_BITMAP = 0, /* block bitmap */ 532 INODE_BITMAP, /* inode bitmap */ 533 INODE_TABLE, /* inode tables */ 534 GROUP_TABLE_COUNT, 535 }; 536 537 /* 538 * Flags used by ext4_map_blocks() 539 */ 540 /* Allocate any needed blocks and/or convert an unitialized 541 extent to be an initialized ext4 */ 542 #define EXT4_GET_BLOCKS_CREATE 0x0001 543 /* Request the creation of an unitialized extent */ 544 #define EXT4_GET_BLOCKS_UNINIT_EXT 0x0002 545 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT (EXT4_GET_BLOCKS_UNINIT_EXT|\ 546 EXT4_GET_BLOCKS_CREATE) 547 /* Caller is from the delayed allocation writeout path, 548 so set the magic i_delalloc_reserve_flag after taking the 549 inode allocation semaphore for */ 550 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE 0x0004 551 /* caller is from the direct IO path, request to creation of an 552 unitialized extents if not allocated, split the uninitialized 553 extent if blocks has been preallocated already*/ 554 #define EXT4_GET_BLOCKS_PRE_IO 0x0008 555 #define EXT4_GET_BLOCKS_CONVERT 0x0010 556 #define EXT4_GET_BLOCKS_IO_CREATE_EXT (EXT4_GET_BLOCKS_PRE_IO|\ 557 EXT4_GET_BLOCKS_CREATE_UNINIT_EXT) 558 /* Convert extent to initialized after IO complete */ 559 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT (EXT4_GET_BLOCKS_CONVERT|\ 560 EXT4_GET_BLOCKS_CREATE_UNINIT_EXT) 561 /* Punch out blocks of an extent */ 562 #define EXT4_GET_BLOCKS_PUNCH_OUT_EXT 0x0020 563 /* Don't normalize allocation size (used for fallocate) */ 564 #define EXT4_GET_BLOCKS_NO_NORMALIZE 0x0040 565 /* Request will not result in inode size update (user for fallocate) */ 566 #define EXT4_GET_BLOCKS_KEEP_SIZE 0x0080 567 568 /* 569 * Flags used by ext4_free_blocks 570 */ 571 #define EXT4_FREE_BLOCKS_METADATA 0x0001 572 #define EXT4_FREE_BLOCKS_FORGET 0x0002 573 #define EXT4_FREE_BLOCKS_VALIDATED 0x0004 574 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008 575 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER 0x0010 576 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER 0x0020 577 578 /* 579 * Flags used by ext4_discard_partial_page_buffers 580 */ 581 #define EXT4_DISCARD_PARTIAL_PG_ZERO_UNMAPPED 0x0001 582 583 /* 584 * ioctl commands 585 */ 586 #define EXT4_IOC_GETFLAGS FS_IOC_GETFLAGS 587 #define EXT4_IOC_SETFLAGS FS_IOC_SETFLAGS 588 #define EXT4_IOC_GETVERSION _IOR('f', 3, long) 589 #define EXT4_IOC_SETVERSION _IOW('f', 4, long) 590 #define EXT4_IOC_GETVERSION_OLD FS_IOC_GETVERSION 591 #define EXT4_IOC_SETVERSION_OLD FS_IOC_SETVERSION 592 #define EXT4_IOC_GETRSVSZ _IOR('f', 5, long) 593 #define EXT4_IOC_SETRSVSZ _IOW('f', 6, long) 594 #define EXT4_IOC_GROUP_EXTEND _IOW('f', 7, unsigned long) 595 #define EXT4_IOC_GROUP_ADD _IOW('f', 8, struct ext4_new_group_input) 596 #define EXT4_IOC_MIGRATE _IO('f', 9) 597 /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */ 598 /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */ 599 #define EXT4_IOC_ALLOC_DA_BLKS _IO('f', 12) 600 #define EXT4_IOC_MOVE_EXT _IOWR('f', 15, struct move_extent) 601 #define EXT4_IOC_RESIZE_FS _IOW('f', 16, __u64) 602 603 #if defined(__KERNEL__) && defined(CONFIG_COMPAT) 604 /* 605 * ioctl commands in 32 bit emulation 606 */ 607 #define EXT4_IOC32_GETFLAGS FS_IOC32_GETFLAGS 608 #define EXT4_IOC32_SETFLAGS FS_IOC32_SETFLAGS 609 #define EXT4_IOC32_GETVERSION _IOR('f', 3, int) 610 #define EXT4_IOC32_SETVERSION _IOW('f', 4, int) 611 #define EXT4_IOC32_GETRSVSZ _IOR('f', 5, int) 612 #define EXT4_IOC32_SETRSVSZ _IOW('f', 6, int) 613 #define EXT4_IOC32_GROUP_EXTEND _IOW('f', 7, unsigned int) 614 #define EXT4_IOC32_GROUP_ADD _IOW('f', 8, struct compat_ext4_new_group_input) 615 #define EXT4_IOC32_GETVERSION_OLD FS_IOC32_GETVERSION 616 #define EXT4_IOC32_SETVERSION_OLD FS_IOC32_SETVERSION 617 #endif 618 619 /* Max physical block we can address w/o extents */ 620 #define EXT4_MAX_BLOCK_FILE_PHYS 0xFFFFFFFF 621 622 /* 623 * Structure of an inode on the disk 624 */ 625 struct ext4_inode { 626 __le16 i_mode; /* File mode */ 627 __le16 i_uid; /* Low 16 bits of Owner Uid */ 628 __le32 i_size_lo; /* Size in bytes */ 629 __le32 i_atime; /* Access time */ 630 __le32 i_ctime; /* Inode Change time */ 631 __le32 i_mtime; /* Modification time */ 632 __le32 i_dtime; /* Deletion Time */ 633 __le16 i_gid; /* Low 16 bits of Group Id */ 634 __le16 i_links_count; /* Links count */ 635 __le32 i_blocks_lo; /* Blocks count */ 636 __le32 i_flags; /* File flags */ 637 union { 638 struct { 639 __le32 l_i_version; 640 } linux1; 641 struct { 642 __u32 h_i_translator; 643 } hurd1; 644 struct { 645 __u32 m_i_reserved1; 646 } masix1; 647 } osd1; /* OS dependent 1 */ 648 __le32 i_block[EXT4_N_BLOCKS];/* Pointers to blocks */ 649 __le32 i_generation; /* File version (for NFS) */ 650 __le32 i_file_acl_lo; /* File ACL */ 651 __le32 i_size_high; 652 __le32 i_obso_faddr; /* Obsoleted fragment address */ 653 union { 654 struct { 655 __le16 l_i_blocks_high; /* were l_i_reserved1 */ 656 __le16 l_i_file_acl_high; 657 __le16 l_i_uid_high; /* these 2 fields */ 658 __le16 l_i_gid_high; /* were reserved2[0] */ 659 __le16 l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */ 660 __le16 l_i_reserved; 661 } linux2; 662 struct { 663 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */ 664 __u16 h_i_mode_high; 665 __u16 h_i_uid_high; 666 __u16 h_i_gid_high; 667 __u32 h_i_author; 668 } hurd2; 669 struct { 670 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */ 671 __le16 m_i_file_acl_high; 672 __u32 m_i_reserved2[2]; 673 } masix2; 674 } osd2; /* OS dependent 2 */ 675 __le16 i_extra_isize; 676 __le16 i_checksum_hi; /* crc32c(uuid+inum+inode) BE */ 677 __le32 i_ctime_extra; /* extra Change time (nsec << 2 | epoch) */ 678 __le32 i_mtime_extra; /* extra Modification time(nsec << 2 | epoch) */ 679 __le32 i_atime_extra; /* extra Access time (nsec << 2 | epoch) */ 680 __le32 i_crtime; /* File Creation time */ 681 __le32 i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */ 682 __le32 i_version_hi; /* high 32 bits for 64-bit version */ 683 }; 684 685 struct move_extent { 686 __u32 reserved; /* should be zero */ 687 __u32 donor_fd; /* donor file descriptor */ 688 __u64 orig_start; /* logical start offset in block for orig */ 689 __u64 donor_start; /* logical start offset in block for donor */ 690 __u64 len; /* block length to be moved */ 691 __u64 moved_len; /* moved block length */ 692 }; 693 694 #define EXT4_EPOCH_BITS 2 695 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1) 696 #define EXT4_NSEC_MASK (~0UL << EXT4_EPOCH_BITS) 697 698 /* 699 * Extended fields will fit into an inode if the filesystem was formatted 700 * with large inodes (-I 256 or larger) and there are not currently any EAs 701 * consuming all of the available space. For new inodes we always reserve 702 * enough space for the kernel's known extended fields, but for inodes 703 * created with an old kernel this might not have been the case. None of 704 * the extended inode fields is critical for correct filesystem operation. 705 * This macro checks if a certain field fits in the inode. Note that 706 * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize 707 */ 708 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field) \ 709 ((offsetof(typeof(*ext4_inode), field) + \ 710 sizeof((ext4_inode)->field)) \ 711 <= (EXT4_GOOD_OLD_INODE_SIZE + \ 712 (einode)->i_extra_isize)) \ 713 714 static inline __le32 ext4_encode_extra_time(struct timespec *time) 715 { 716 return cpu_to_le32((sizeof(time->tv_sec) > 4 ? 717 (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) | 718 ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK)); 719 } 720 721 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra) 722 { 723 if (sizeof(time->tv_sec) > 4) 724 time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK) 725 << 32; 726 time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS; 727 } 728 729 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode) \ 730 do { \ 731 (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec); \ 732 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \ 733 (raw_inode)->xtime ## _extra = \ 734 ext4_encode_extra_time(&(inode)->xtime); \ 735 } while (0) 736 737 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode) \ 738 do { \ 739 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \ 740 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec); \ 741 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \ 742 (raw_inode)->xtime ## _extra = \ 743 ext4_encode_extra_time(&(einode)->xtime); \ 744 } while (0) 745 746 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode) \ 747 do { \ 748 (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime); \ 749 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \ 750 ext4_decode_extra_time(&(inode)->xtime, \ 751 raw_inode->xtime ## _extra); \ 752 else \ 753 (inode)->xtime.tv_nsec = 0; \ 754 } while (0) 755 756 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode) \ 757 do { \ 758 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \ 759 (einode)->xtime.tv_sec = \ 760 (signed)le32_to_cpu((raw_inode)->xtime); \ 761 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \ 762 ext4_decode_extra_time(&(einode)->xtime, \ 763 raw_inode->xtime ## _extra); \ 764 else \ 765 (einode)->xtime.tv_nsec = 0; \ 766 } while (0) 767 768 #define i_disk_version osd1.linux1.l_i_version 769 770 #if defined(__KERNEL__) || defined(__linux__) 771 #define i_reserved1 osd1.linux1.l_i_reserved1 772 #define i_file_acl_high osd2.linux2.l_i_file_acl_high 773 #define i_blocks_high osd2.linux2.l_i_blocks_high 774 #define i_uid_low i_uid 775 #define i_gid_low i_gid 776 #define i_uid_high osd2.linux2.l_i_uid_high 777 #define i_gid_high osd2.linux2.l_i_gid_high 778 #define i_checksum_lo osd2.linux2.l_i_checksum_lo 779 780 #elif defined(__GNU__) 781 782 #define i_translator osd1.hurd1.h_i_translator 783 #define i_uid_high osd2.hurd2.h_i_uid_high 784 #define i_gid_high osd2.hurd2.h_i_gid_high 785 #define i_author osd2.hurd2.h_i_author 786 787 #elif defined(__masix__) 788 789 #define i_reserved1 osd1.masix1.m_i_reserved1 790 #define i_file_acl_high osd2.masix2.m_i_file_acl_high 791 #define i_reserved2 osd2.masix2.m_i_reserved2 792 793 #endif /* defined(__KERNEL__) || defined(__linux__) */ 794 795 /* 796 * storage for cached extent 797 * If ec_len == 0, then the cache is invalid. 798 * If ec_start == 0, then the cache represents a gap (null mapping) 799 */ 800 struct ext4_ext_cache { 801 ext4_fsblk_t ec_start; 802 ext4_lblk_t ec_block; 803 __u32 ec_len; /* must be 32bit to return holes */ 804 }; 805 806 /* 807 * fourth extended file system inode data in memory 808 */ 809 struct ext4_inode_info { 810 __le32 i_data[15]; /* unconverted */ 811 __u32 i_dtime; 812 ext4_fsblk_t i_file_acl; 813 814 /* 815 * i_block_group is the number of the block group which contains 816 * this file's inode. Constant across the lifetime of the inode, 817 * it is ued for making block allocation decisions - we try to 818 * place a file's data blocks near its inode block, and new inodes 819 * near to their parent directory's inode. 820 */ 821 ext4_group_t i_block_group; 822 ext4_lblk_t i_dir_start_lookup; 823 #if (BITS_PER_LONG < 64) 824 unsigned long i_state_flags; /* Dynamic state flags */ 825 #endif 826 unsigned long i_flags; 827 828 #ifdef CONFIG_EXT4_FS_XATTR 829 /* 830 * Extended attributes can be read independently of the main file 831 * data. Taking i_mutex even when reading would cause contention 832 * between readers of EAs and writers of regular file data, so 833 * instead we synchronize on xattr_sem when reading or changing 834 * EAs. 835 */ 836 struct rw_semaphore xattr_sem; 837 #endif 838 839 struct list_head i_orphan; /* unlinked but open inodes */ 840 841 /* 842 * i_disksize keeps track of what the inode size is ON DISK, not 843 * in memory. During truncate, i_size is set to the new size by 844 * the VFS prior to calling ext4_truncate(), but the filesystem won't 845 * set i_disksize to 0 until the truncate is actually under way. 846 * 847 * The intent is that i_disksize always represents the blocks which 848 * are used by this file. This allows recovery to restart truncate 849 * on orphans if we crash during truncate. We actually write i_disksize 850 * into the on-disk inode when writing inodes out, instead of i_size. 851 * 852 * The only time when i_disksize and i_size may be different is when 853 * a truncate is in progress. The only things which change i_disksize 854 * are ext4_get_block (growth) and ext4_truncate (shrinkth). 855 */ 856 loff_t i_disksize; 857 858 /* 859 * i_data_sem is for serialising ext4_truncate() against 860 * ext4_getblock(). In the 2.4 ext2 design, great chunks of inode's 861 * data tree are chopped off during truncate. We can't do that in 862 * ext4 because whenever we perform intermediate commits during 863 * truncate, the inode and all the metadata blocks *must* be in a 864 * consistent state which allows truncation of the orphans to restart 865 * during recovery. Hence we must fix the get_block-vs-truncate race 866 * by other means, so we have i_data_sem. 867 */ 868 struct rw_semaphore i_data_sem; 869 struct inode vfs_inode; 870 struct jbd2_inode *jinode; 871 872 struct ext4_ext_cache i_cached_extent; 873 /* 874 * File creation time. Its function is same as that of 875 * struct timespec i_{a,c,m}time in the generic inode. 876 */ 877 struct timespec i_crtime; 878 879 /* mballoc */ 880 struct list_head i_prealloc_list; 881 spinlock_t i_prealloc_lock; 882 883 /* ialloc */ 884 ext4_group_t i_last_alloc_group; 885 886 /* allocation reservation info for delalloc */ 887 /* In case of bigalloc, these refer to clusters rather than blocks */ 888 unsigned int i_reserved_data_blocks; 889 unsigned int i_reserved_meta_blocks; 890 unsigned int i_allocated_meta_blocks; 891 ext4_lblk_t i_da_metadata_calc_last_lblock; 892 int i_da_metadata_calc_len; 893 894 /* on-disk additional length */ 895 __u16 i_extra_isize; 896 897 #ifdef CONFIG_QUOTA 898 /* quota space reservation, managed internally by quota code */ 899 qsize_t i_reserved_quota; 900 #endif 901 902 /* completed IOs that might need unwritten extents handling */ 903 struct list_head i_completed_io_list; 904 spinlock_t i_completed_io_lock; 905 atomic_t i_ioend_count; /* Number of outstanding io_end structs */ 906 /* current io_end structure for async DIO write*/ 907 ext4_io_end_t *cur_aio_dio; 908 atomic_t i_aiodio_unwritten; /* Nr. of inflight conversions pending */ 909 910 spinlock_t i_block_reservation_lock; 911 912 /* 913 * Transactions that contain inode's metadata needed to complete 914 * fsync and fdatasync, respectively. 915 */ 916 tid_t i_sync_tid; 917 tid_t i_datasync_tid; 918 }; 919 920 /* 921 * File system states 922 */ 923 #define EXT4_VALID_FS 0x0001 /* Unmounted cleanly */ 924 #define EXT4_ERROR_FS 0x0002 /* Errors detected */ 925 #define EXT4_ORPHAN_FS 0x0004 /* Orphans being recovered */ 926 927 /* 928 * Misc. filesystem flags 929 */ 930 #define EXT2_FLAGS_SIGNED_HASH 0x0001 /* Signed dirhash in use */ 931 #define EXT2_FLAGS_UNSIGNED_HASH 0x0002 /* Unsigned dirhash in use */ 932 #define EXT2_FLAGS_TEST_FILESYS 0x0004 /* to test development code */ 933 934 /* 935 * Mount flags 936 */ 937 #define EXT4_MOUNT_GRPID 0x00004 /* Create files with directory's group */ 938 #define EXT4_MOUNT_DEBUG 0x00008 /* Some debugging messages */ 939 #define EXT4_MOUNT_ERRORS_CONT 0x00010 /* Continue on errors */ 940 #define EXT4_MOUNT_ERRORS_RO 0x00020 /* Remount fs ro on errors */ 941 #define EXT4_MOUNT_ERRORS_PANIC 0x00040 /* Panic on errors */ 942 #define EXT4_MOUNT_ERRORS_MASK 0x00070 943 #define EXT4_MOUNT_MINIX_DF 0x00080 /* Mimics the Minix statfs */ 944 #define EXT4_MOUNT_NOLOAD 0x00100 /* Don't use existing journal*/ 945 #define EXT4_MOUNT_DATA_FLAGS 0x00C00 /* Mode for data writes: */ 946 #define EXT4_MOUNT_JOURNAL_DATA 0x00400 /* Write data to journal */ 947 #define EXT4_MOUNT_ORDERED_DATA 0x00800 /* Flush data before commit */ 948 #define EXT4_MOUNT_WRITEBACK_DATA 0x00C00 /* No data ordering */ 949 #define EXT4_MOUNT_UPDATE_JOURNAL 0x01000 /* Update the journal format */ 950 #define EXT4_MOUNT_NO_UID32 0x02000 /* Disable 32-bit UIDs */ 951 #define EXT4_MOUNT_XATTR_USER 0x04000 /* Extended user attributes */ 952 #define EXT4_MOUNT_POSIX_ACL 0x08000 /* POSIX Access Control Lists */ 953 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC 0x10000 /* No auto delalloc mapping */ 954 #define EXT4_MOUNT_BARRIER 0x20000 /* Use block barriers */ 955 #define EXT4_MOUNT_QUOTA 0x80000 /* Some quota option set */ 956 #define EXT4_MOUNT_USRQUOTA 0x100000 /* "old" user quota */ 957 #define EXT4_MOUNT_GRPQUOTA 0x200000 /* "old" group quota */ 958 #define EXT4_MOUNT_DIOREAD_NOLOCK 0x400000 /* Enable support for dio read nolocking */ 959 #define EXT4_MOUNT_JOURNAL_CHECKSUM 0x800000 /* Journal checksums */ 960 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */ 961 #define EXT4_MOUNT_MBLK_IO_SUBMIT 0x4000000 /* multi-block io submits */ 962 #define EXT4_MOUNT_DELALLOC 0x8000000 /* Delalloc support */ 963 #define EXT4_MOUNT_DATA_ERR_ABORT 0x10000000 /* Abort on file data write */ 964 #define EXT4_MOUNT_BLOCK_VALIDITY 0x20000000 /* Block validity checking */ 965 #define EXT4_MOUNT_DISCARD 0x40000000 /* Issue DISCARD requests */ 966 #define EXT4_MOUNT_INIT_INODE_TABLE 0x80000000 /* Initialize uninitialized itables */ 967 968 #define EXT4_MOUNT2_EXPLICIT_DELALLOC 0x00000001 /* User explicitly 969 specified delalloc */ 970 971 #define clear_opt(sb, opt) EXT4_SB(sb)->s_mount_opt &= \ 972 ~EXT4_MOUNT_##opt 973 #define set_opt(sb, opt) EXT4_SB(sb)->s_mount_opt |= \ 974 EXT4_MOUNT_##opt 975 #define test_opt(sb, opt) (EXT4_SB(sb)->s_mount_opt & \ 976 EXT4_MOUNT_##opt) 977 978 #define clear_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 &= \ 979 ~EXT4_MOUNT2_##opt 980 #define set_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 |= \ 981 EXT4_MOUNT2_##opt 982 #define test_opt2(sb, opt) (EXT4_SB(sb)->s_mount_opt2 & \ 983 EXT4_MOUNT2_##opt) 984 985 #define ext4_test_and_set_bit __test_and_set_bit_le 986 #define ext4_set_bit __set_bit_le 987 #define ext4_set_bit_atomic ext2_set_bit_atomic 988 #define ext4_test_and_clear_bit __test_and_clear_bit_le 989 #define ext4_clear_bit __clear_bit_le 990 #define ext4_clear_bit_atomic ext2_clear_bit_atomic 991 #define ext4_test_bit test_bit_le 992 #define ext4_find_next_zero_bit find_next_zero_bit_le 993 #define ext4_find_next_bit find_next_bit_le 994 995 extern void ext4_set_bits(void *bm, int cur, int len); 996 997 /* 998 * Maximal mount counts between two filesystem checks 999 */ 1000 #define EXT4_DFL_MAX_MNT_COUNT 20 /* Allow 20 mounts */ 1001 #define EXT4_DFL_CHECKINTERVAL 0 /* Don't use interval check */ 1002 1003 /* 1004 * Behaviour when detecting errors 1005 */ 1006 #define EXT4_ERRORS_CONTINUE 1 /* Continue execution */ 1007 #define EXT4_ERRORS_RO 2 /* Remount fs read-only */ 1008 #define EXT4_ERRORS_PANIC 3 /* Panic */ 1009 #define EXT4_ERRORS_DEFAULT EXT4_ERRORS_CONTINUE 1010 1011 /* Metadata checksum algorithm codes */ 1012 #define EXT4_CRC32C_CHKSUM 1 1013 1014 /* 1015 * Structure of the super block 1016 */ 1017 struct ext4_super_block { 1018 /*00*/ __le32 s_inodes_count; /* Inodes count */ 1019 __le32 s_blocks_count_lo; /* Blocks count */ 1020 __le32 s_r_blocks_count_lo; /* Reserved blocks count */ 1021 __le32 s_free_blocks_count_lo; /* Free blocks count */ 1022 /*10*/ __le32 s_free_inodes_count; /* Free inodes count */ 1023 __le32 s_first_data_block; /* First Data Block */ 1024 __le32 s_log_block_size; /* Block size */ 1025 __le32 s_log_cluster_size; /* Allocation cluster size */ 1026 /*20*/ __le32 s_blocks_per_group; /* # Blocks per group */ 1027 __le32 s_clusters_per_group; /* # Clusters per group */ 1028 __le32 s_inodes_per_group; /* # Inodes per group */ 1029 __le32 s_mtime; /* Mount time */ 1030 /*30*/ __le32 s_wtime; /* Write time */ 1031 __le16 s_mnt_count; /* Mount count */ 1032 __le16 s_max_mnt_count; /* Maximal mount count */ 1033 __le16 s_magic; /* Magic signature */ 1034 __le16 s_state; /* File system state */ 1035 __le16 s_errors; /* Behaviour when detecting errors */ 1036 __le16 s_minor_rev_level; /* minor revision level */ 1037 /*40*/ __le32 s_lastcheck; /* time of last check */ 1038 __le32 s_checkinterval; /* max. time between checks */ 1039 __le32 s_creator_os; /* OS */ 1040 __le32 s_rev_level; /* Revision level */ 1041 /*50*/ __le16 s_def_resuid; /* Default uid for reserved blocks */ 1042 __le16 s_def_resgid; /* Default gid for reserved blocks */ 1043 /* 1044 * These fields are for EXT4_DYNAMIC_REV superblocks only. 1045 * 1046 * Note: the difference between the compatible feature set and 1047 * the incompatible feature set is that if there is a bit set 1048 * in the incompatible feature set that the kernel doesn't 1049 * know about, it should refuse to mount the filesystem. 1050 * 1051 * e2fsck's requirements are more strict; if it doesn't know 1052 * about a feature in either the compatible or incompatible 1053 * feature set, it must abort and not try to meddle with 1054 * things it doesn't understand... 1055 */ 1056 __le32 s_first_ino; /* First non-reserved inode */ 1057 __le16 s_inode_size; /* size of inode structure */ 1058 __le16 s_block_group_nr; /* block group # of this superblock */ 1059 __le32 s_feature_compat; /* compatible feature set */ 1060 /*60*/ __le32 s_feature_incompat; /* incompatible feature set */ 1061 __le32 s_feature_ro_compat; /* readonly-compatible feature set */ 1062 /*68*/ __u8 s_uuid[16]; /* 128-bit uuid for volume */ 1063 /*78*/ char s_volume_name[16]; /* volume name */ 1064 /*88*/ char s_last_mounted[64]; /* directory where last mounted */ 1065 /*C8*/ __le32 s_algorithm_usage_bitmap; /* For compression */ 1066 /* 1067 * Performance hints. Directory preallocation should only 1068 * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on. 1069 */ 1070 __u8 s_prealloc_blocks; /* Nr of blocks to try to preallocate*/ 1071 __u8 s_prealloc_dir_blocks; /* Nr to preallocate for dirs */ 1072 __le16 s_reserved_gdt_blocks; /* Per group desc for online growth */ 1073 /* 1074 * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set. 1075 */ 1076 /*D0*/ __u8 s_journal_uuid[16]; /* uuid of journal superblock */ 1077 /*E0*/ __le32 s_journal_inum; /* inode number of journal file */ 1078 __le32 s_journal_dev; /* device number of journal file */ 1079 __le32 s_last_orphan; /* start of list of inodes to delete */ 1080 __le32 s_hash_seed[4]; /* HTREE hash seed */ 1081 __u8 s_def_hash_version; /* Default hash version to use */ 1082 __u8 s_jnl_backup_type; 1083 __le16 s_desc_size; /* size of group descriptor */ 1084 /*100*/ __le32 s_default_mount_opts; 1085 __le32 s_first_meta_bg; /* First metablock block group */ 1086 __le32 s_mkfs_time; /* When the filesystem was created */ 1087 __le32 s_jnl_blocks[17]; /* Backup of the journal inode */ 1088 /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */ 1089 /*150*/ __le32 s_blocks_count_hi; /* Blocks count */ 1090 __le32 s_r_blocks_count_hi; /* Reserved blocks count */ 1091 __le32 s_free_blocks_count_hi; /* Free blocks count */ 1092 __le16 s_min_extra_isize; /* All inodes have at least # bytes */ 1093 __le16 s_want_extra_isize; /* New inodes should reserve # bytes */ 1094 __le32 s_flags; /* Miscellaneous flags */ 1095 __le16 s_raid_stride; /* RAID stride */ 1096 __le16 s_mmp_update_interval; /* # seconds to wait in MMP checking */ 1097 __le64 s_mmp_block; /* Block for multi-mount protection */ 1098 __le32 s_raid_stripe_width; /* blocks on all data disks (N*stride)*/ 1099 __u8 s_log_groups_per_flex; /* FLEX_BG group size */ 1100 __u8 s_checksum_type; /* metadata checksum algorithm used */ 1101 __le16 s_reserved_pad; 1102 __le64 s_kbytes_written; /* nr of lifetime kilobytes written */ 1103 __le32 s_snapshot_inum; /* Inode number of active snapshot */ 1104 __le32 s_snapshot_id; /* sequential ID of active snapshot */ 1105 __le64 s_snapshot_r_blocks_count; /* reserved blocks for active 1106 snapshot's future use */ 1107 __le32 s_snapshot_list; /* inode number of the head of the 1108 on-disk snapshot list */ 1109 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count) 1110 __le32 s_error_count; /* number of fs errors */ 1111 __le32 s_first_error_time; /* first time an error happened */ 1112 __le32 s_first_error_ino; /* inode involved in first error */ 1113 __le64 s_first_error_block; /* block involved of first error */ 1114 __u8 s_first_error_func[32]; /* function where the error happened */ 1115 __le32 s_first_error_line; /* line number where error happened */ 1116 __le32 s_last_error_time; /* most recent time of an error */ 1117 __le32 s_last_error_ino; /* inode involved in last error */ 1118 __le32 s_last_error_line; /* line number where error happened */ 1119 __le64 s_last_error_block; /* block involved of last error */ 1120 __u8 s_last_error_func[32]; /* function where the error happened */ 1121 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts) 1122 __u8 s_mount_opts[64]; 1123 __le32 s_usr_quota_inum; /* inode for tracking user quota */ 1124 __le32 s_grp_quota_inum; /* inode for tracking group quota */ 1125 __le32 s_overhead_clusters; /* overhead blocks/clusters in fs */ 1126 __le32 s_reserved[108]; /* Padding to the end of the block */ 1127 __le32 s_checksum; /* crc32c(superblock) */ 1128 }; 1129 1130 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START) 1131 1132 #ifdef __KERNEL__ 1133 1134 /* 1135 * run-time mount flags 1136 */ 1137 #define EXT4_MF_MNTDIR_SAMPLED 0x0001 1138 #define EXT4_MF_FS_ABORTED 0x0002 /* Fatal error detected */ 1139 1140 /* 1141 * fourth extended-fs super-block data in memory 1142 */ 1143 struct ext4_sb_info { 1144 unsigned long s_desc_size; /* Size of a group descriptor in bytes */ 1145 unsigned long s_inodes_per_block;/* Number of inodes per block */ 1146 unsigned long s_blocks_per_group;/* Number of blocks in a group */ 1147 unsigned long s_clusters_per_group; /* Number of clusters in a group */ 1148 unsigned long s_inodes_per_group;/* Number of inodes in a group */ 1149 unsigned long s_itb_per_group; /* Number of inode table blocks per group */ 1150 unsigned long s_gdb_count; /* Number of group descriptor blocks */ 1151 unsigned long s_desc_per_block; /* Number of group descriptors per block */ 1152 ext4_group_t s_groups_count; /* Number of groups in the fs */ 1153 ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */ 1154 unsigned long s_overhead_last; /* Last calculated overhead */ 1155 unsigned long s_blocks_last; /* Last seen block count */ 1156 unsigned int s_cluster_ratio; /* Number of blocks per cluster */ 1157 unsigned int s_cluster_bits; /* log2 of s_cluster_ratio */ 1158 loff_t s_bitmap_maxbytes; /* max bytes for bitmap files */ 1159 struct buffer_head * s_sbh; /* Buffer containing the super block */ 1160 struct ext4_super_block *s_es; /* Pointer to the super block in the buffer */ 1161 struct buffer_head **s_group_desc; 1162 unsigned int s_mount_opt; 1163 unsigned int s_mount_opt2; 1164 unsigned int s_mount_flags; 1165 unsigned int s_def_mount_opt; 1166 ext4_fsblk_t s_sb_block; 1167 uid_t s_resuid; 1168 gid_t s_resgid; 1169 unsigned short s_mount_state; 1170 unsigned short s_pad; 1171 int s_addr_per_block_bits; 1172 int s_desc_per_block_bits; 1173 int s_inode_size; 1174 int s_first_ino; 1175 unsigned int s_inode_readahead_blks; 1176 unsigned int s_inode_goal; 1177 spinlock_t s_next_gen_lock; 1178 u32 s_next_generation; 1179 u32 s_hash_seed[4]; 1180 int s_def_hash_version; 1181 int s_hash_unsigned; /* 3 if hash should be signed, 0 if not */ 1182 struct percpu_counter s_freeclusters_counter; 1183 struct percpu_counter s_freeinodes_counter; 1184 struct percpu_counter s_dirs_counter; 1185 struct percpu_counter s_dirtyclusters_counter; 1186 struct blockgroup_lock *s_blockgroup_lock; 1187 struct proc_dir_entry *s_proc; 1188 struct kobject s_kobj; 1189 struct completion s_kobj_unregister; 1190 1191 /* Journaling */ 1192 struct journal_s *s_journal; 1193 struct list_head s_orphan; 1194 struct mutex s_orphan_lock; 1195 unsigned long s_resize_flags; /* Flags indicating if there 1196 is a resizer */ 1197 unsigned long s_commit_interval; 1198 u32 s_max_batch_time; 1199 u32 s_min_batch_time; 1200 struct block_device *journal_bdev; 1201 #ifdef CONFIG_QUOTA 1202 char *s_qf_names[MAXQUOTAS]; /* Names of quota files with journalled quota */ 1203 int s_jquota_fmt; /* Format of quota to use */ 1204 #endif 1205 unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */ 1206 struct rb_root system_blks; 1207 1208 #ifdef EXTENTS_STATS 1209 /* ext4 extents stats */ 1210 unsigned long s_ext_min; 1211 unsigned long s_ext_max; 1212 unsigned long s_depth_max; 1213 spinlock_t s_ext_stats_lock; 1214 unsigned long s_ext_blocks; 1215 unsigned long s_ext_extents; 1216 #endif 1217 1218 /* for buddy allocator */ 1219 struct ext4_group_info ***s_group_info; 1220 struct inode *s_buddy_cache; 1221 spinlock_t s_md_lock; 1222 unsigned short *s_mb_offsets; 1223 unsigned int *s_mb_maxs; 1224 1225 /* tunables */ 1226 unsigned long s_stripe; 1227 unsigned int s_mb_stream_request; 1228 unsigned int s_mb_max_to_scan; 1229 unsigned int s_mb_min_to_scan; 1230 unsigned int s_mb_stats; 1231 unsigned int s_mb_order2_reqs; 1232 unsigned int s_mb_group_prealloc; 1233 unsigned int s_max_writeback_mb_bump; 1234 /* where last allocation was done - for stream allocation */ 1235 unsigned long s_mb_last_group; 1236 unsigned long s_mb_last_start; 1237 1238 /* stats for buddy allocator */ 1239 atomic_t s_bal_reqs; /* number of reqs with len > 1 */ 1240 atomic_t s_bal_success; /* we found long enough chunks */ 1241 atomic_t s_bal_allocated; /* in blocks */ 1242 atomic_t s_bal_ex_scanned; /* total extents scanned */ 1243 atomic_t s_bal_goals; /* goal hits */ 1244 atomic_t s_bal_breaks; /* too long searches */ 1245 atomic_t s_bal_2orders; /* 2^order hits */ 1246 spinlock_t s_bal_lock; 1247 unsigned long s_mb_buddies_generated; 1248 unsigned long long s_mb_generation_time; 1249 atomic_t s_mb_lost_chunks; 1250 atomic_t s_mb_preallocated; 1251 atomic_t s_mb_discarded; 1252 atomic_t s_lock_busy; 1253 1254 /* locality groups */ 1255 struct ext4_locality_group __percpu *s_locality_groups; 1256 1257 /* for write statistics */ 1258 unsigned long s_sectors_written_start; 1259 u64 s_kbytes_written; 1260 1261 unsigned int s_log_groups_per_flex; 1262 struct flex_groups *s_flex_groups; 1263 1264 /* workqueue for dio unwritten */ 1265 struct workqueue_struct *dio_unwritten_wq; 1266 1267 /* timer for periodic error stats printing */ 1268 struct timer_list s_err_report; 1269 1270 /* Lazy inode table initialization info */ 1271 struct ext4_li_request *s_li_request; 1272 /* Wait multiplier for lazy initialization thread */ 1273 unsigned int s_li_wait_mult; 1274 1275 /* Kernel thread for multiple mount protection */ 1276 struct task_struct *s_mmp_tsk; 1277 1278 /* record the last minlen when FITRIM is called. */ 1279 atomic_t s_last_trim_minblks; 1280 1281 /* Reference to checksum algorithm driver via cryptoapi */ 1282 struct crypto_shash *s_chksum_driver; 1283 }; 1284 1285 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb) 1286 { 1287 return sb->s_fs_info; 1288 } 1289 static inline struct ext4_inode_info *EXT4_I(struct inode *inode) 1290 { 1291 return container_of(inode, struct ext4_inode_info, vfs_inode); 1292 } 1293 1294 static inline struct timespec ext4_current_time(struct inode *inode) 1295 { 1296 return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ? 1297 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC; 1298 } 1299 1300 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino) 1301 { 1302 return ino == EXT4_ROOT_INO || 1303 ino == EXT4_JOURNAL_INO || 1304 ino == EXT4_RESIZE_INO || 1305 (ino >= EXT4_FIRST_INO(sb) && 1306 ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)); 1307 } 1308 1309 static inline void ext4_set_io_unwritten_flag(struct inode *inode, 1310 struct ext4_io_end *io_end) 1311 { 1312 if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) { 1313 io_end->flag |= EXT4_IO_END_UNWRITTEN; 1314 atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten); 1315 } 1316 } 1317 1318 /* 1319 * Inode dynamic state flags 1320 */ 1321 enum { 1322 EXT4_STATE_JDATA, /* journaled data exists */ 1323 EXT4_STATE_NEW, /* inode is newly created */ 1324 EXT4_STATE_XATTR, /* has in-inode xattrs */ 1325 EXT4_STATE_NO_EXPAND, /* No space for expansion */ 1326 EXT4_STATE_DA_ALLOC_CLOSE, /* Alloc DA blks on close */ 1327 EXT4_STATE_EXT_MIGRATE, /* Inode is migrating */ 1328 EXT4_STATE_DIO_UNWRITTEN, /* need convert on dio done*/ 1329 EXT4_STATE_NEWENTRY, /* File just added to dir */ 1330 EXT4_STATE_DELALLOC_RESERVED, /* blks already reserved for delalloc */ 1331 }; 1332 1333 #define EXT4_INODE_BIT_FNS(name, field, offset) \ 1334 static inline int ext4_test_inode_##name(struct inode *inode, int bit) \ 1335 { \ 1336 return test_bit(bit + (offset), &EXT4_I(inode)->i_##field); \ 1337 } \ 1338 static inline void ext4_set_inode_##name(struct inode *inode, int bit) \ 1339 { \ 1340 set_bit(bit + (offset), &EXT4_I(inode)->i_##field); \ 1341 } \ 1342 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \ 1343 { \ 1344 clear_bit(bit + (offset), &EXT4_I(inode)->i_##field); \ 1345 } 1346 1347 EXT4_INODE_BIT_FNS(flag, flags, 0) 1348 #if (BITS_PER_LONG < 64) 1349 EXT4_INODE_BIT_FNS(state, state_flags, 0) 1350 1351 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei) 1352 { 1353 (ei)->i_state_flags = 0; 1354 } 1355 #else 1356 EXT4_INODE_BIT_FNS(state, flags, 32) 1357 1358 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei) 1359 { 1360 /* We depend on the fact that callers will set i_flags */ 1361 } 1362 #endif 1363 #else 1364 /* Assume that user mode programs are passing in an ext4fs superblock, not 1365 * a kernel struct super_block. This will allow us to call the feature-test 1366 * macros from user land. */ 1367 #define EXT4_SB(sb) (sb) 1368 #endif 1369 1370 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime 1371 1372 /* 1373 * Codes for operating systems 1374 */ 1375 #define EXT4_OS_LINUX 0 1376 #define EXT4_OS_HURD 1 1377 #define EXT4_OS_MASIX 2 1378 #define EXT4_OS_FREEBSD 3 1379 #define EXT4_OS_LITES 4 1380 1381 /* 1382 * Revision levels 1383 */ 1384 #define EXT4_GOOD_OLD_REV 0 /* The good old (original) format */ 1385 #define EXT4_DYNAMIC_REV 1 /* V2 format w/ dynamic inode sizes */ 1386 1387 #define EXT4_CURRENT_REV EXT4_GOOD_OLD_REV 1388 #define EXT4_MAX_SUPP_REV EXT4_DYNAMIC_REV 1389 1390 #define EXT4_GOOD_OLD_INODE_SIZE 128 1391 1392 /* 1393 * Feature set definitions 1394 */ 1395 1396 #define EXT4_HAS_COMPAT_FEATURE(sb,mask) \ 1397 ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0) 1398 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask) \ 1399 ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0) 1400 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask) \ 1401 ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0) 1402 #define EXT4_SET_COMPAT_FEATURE(sb,mask) \ 1403 EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask) 1404 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask) \ 1405 EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask) 1406 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask) \ 1407 EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask) 1408 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask) \ 1409 EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask) 1410 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask) \ 1411 EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask) 1412 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask) \ 1413 EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask) 1414 1415 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC 0x0001 1416 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES 0x0002 1417 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL 0x0004 1418 #define EXT4_FEATURE_COMPAT_EXT_ATTR 0x0008 1419 #define EXT4_FEATURE_COMPAT_RESIZE_INODE 0x0010 1420 #define EXT4_FEATURE_COMPAT_DIR_INDEX 0x0020 1421 1422 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER 0x0001 1423 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE 0x0002 1424 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR 0x0004 1425 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE 0x0008 1426 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM 0x0010 1427 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK 0x0020 1428 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE 0x0040 1429 #define EXT4_FEATURE_RO_COMPAT_QUOTA 0x0100 1430 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC 0x0200 1431 /* 1432 * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM). When 1433 * METADATA_CSUM is set, group descriptor checksums use the same algorithm as 1434 * all other data structures' checksums. However, the METADATA_CSUM and 1435 * GDT_CSUM bits are mutually exclusive. 1436 */ 1437 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM 0x0400 1438 1439 #define EXT4_FEATURE_INCOMPAT_COMPRESSION 0x0001 1440 #define EXT4_FEATURE_INCOMPAT_FILETYPE 0x0002 1441 #define EXT4_FEATURE_INCOMPAT_RECOVER 0x0004 /* Needs recovery */ 1442 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 /* Journal device */ 1443 #define EXT4_FEATURE_INCOMPAT_META_BG 0x0010 1444 #define EXT4_FEATURE_INCOMPAT_EXTENTS 0x0040 /* extents support */ 1445 #define EXT4_FEATURE_INCOMPAT_64BIT 0x0080 1446 #define EXT4_FEATURE_INCOMPAT_MMP 0x0100 1447 #define EXT4_FEATURE_INCOMPAT_FLEX_BG 0x0200 1448 #define EXT4_FEATURE_INCOMPAT_EA_INODE 0x0400 /* EA in inode */ 1449 #define EXT4_FEATURE_INCOMPAT_DIRDATA 0x1000 /* data in dirent */ 1450 #define EXT4_FEATURE_INCOMPAT_BG_USE_META_CSUM 0x2000 /* use crc32c for bg */ 1451 #define EXT4_FEATURE_INCOMPAT_LARGEDIR 0x4000 /* >2GB or 3-lvl htree */ 1452 #define EXT4_FEATURE_INCOMPAT_INLINEDATA 0x8000 /* data in inode */ 1453 1454 #define EXT2_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR 1455 #define EXT2_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \ 1456 EXT4_FEATURE_INCOMPAT_META_BG) 1457 #define EXT2_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \ 1458 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \ 1459 EXT4_FEATURE_RO_COMPAT_BTREE_DIR) 1460 1461 #define EXT3_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR 1462 #define EXT3_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \ 1463 EXT4_FEATURE_INCOMPAT_RECOVER| \ 1464 EXT4_FEATURE_INCOMPAT_META_BG) 1465 #define EXT3_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \ 1466 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \ 1467 EXT4_FEATURE_RO_COMPAT_BTREE_DIR) 1468 1469 #define EXT4_FEATURE_COMPAT_SUPP EXT2_FEATURE_COMPAT_EXT_ATTR 1470 #define EXT4_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \ 1471 EXT4_FEATURE_INCOMPAT_RECOVER| \ 1472 EXT4_FEATURE_INCOMPAT_META_BG| \ 1473 EXT4_FEATURE_INCOMPAT_EXTENTS| \ 1474 EXT4_FEATURE_INCOMPAT_64BIT| \ 1475 EXT4_FEATURE_INCOMPAT_FLEX_BG| \ 1476 EXT4_FEATURE_INCOMPAT_MMP) 1477 #define EXT4_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \ 1478 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \ 1479 EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \ 1480 EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \ 1481 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \ 1482 EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\ 1483 EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\ 1484 EXT4_FEATURE_RO_COMPAT_BIGALLOC) 1485 1486 /* 1487 * Default values for user and/or group using reserved blocks 1488 */ 1489 #define EXT4_DEF_RESUID 0 1490 #define EXT4_DEF_RESGID 0 1491 1492 #define EXT4_DEF_INODE_READAHEAD_BLKS 32 1493 1494 /* 1495 * Default mount options 1496 */ 1497 #define EXT4_DEFM_DEBUG 0x0001 1498 #define EXT4_DEFM_BSDGROUPS 0x0002 1499 #define EXT4_DEFM_XATTR_USER 0x0004 1500 #define EXT4_DEFM_ACL 0x0008 1501 #define EXT4_DEFM_UID16 0x0010 1502 #define EXT4_DEFM_JMODE 0x0060 1503 #define EXT4_DEFM_JMODE_DATA 0x0020 1504 #define EXT4_DEFM_JMODE_ORDERED 0x0040 1505 #define EXT4_DEFM_JMODE_WBACK 0x0060 1506 #define EXT4_DEFM_NOBARRIER 0x0100 1507 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200 1508 #define EXT4_DEFM_DISCARD 0x0400 1509 #define EXT4_DEFM_NODELALLOC 0x0800 1510 1511 /* 1512 * Default journal batch times 1513 */ 1514 #define EXT4_DEF_MIN_BATCH_TIME 0 1515 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */ 1516 1517 /* 1518 * Minimum number of groups in a flexgroup before we separate out 1519 * directories into the first block group of a flexgroup 1520 */ 1521 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4 1522 1523 /* 1524 * Structure of a directory entry 1525 */ 1526 #define EXT4_NAME_LEN 255 1527 1528 struct ext4_dir_entry { 1529 __le32 inode; /* Inode number */ 1530 __le16 rec_len; /* Directory entry length */ 1531 __le16 name_len; /* Name length */ 1532 char name[EXT4_NAME_LEN]; /* File name */ 1533 }; 1534 1535 /* 1536 * The new version of the directory entry. Since EXT4 structures are 1537 * stored in intel byte order, and the name_len field could never be 1538 * bigger than 255 chars, it's safe to reclaim the extra byte for the 1539 * file_type field. 1540 */ 1541 struct ext4_dir_entry_2 { 1542 __le32 inode; /* Inode number */ 1543 __le16 rec_len; /* Directory entry length */ 1544 __u8 name_len; /* Name length */ 1545 __u8 file_type; 1546 char name[EXT4_NAME_LEN]; /* File name */ 1547 }; 1548 1549 /* 1550 * This is a bogus directory entry at the end of each leaf block that 1551 * records checksums. 1552 */ 1553 struct ext4_dir_entry_tail { 1554 __le32 det_reserved_zero1; /* Pretend to be unused */ 1555 __le16 det_rec_len; /* 12 */ 1556 __u8 det_reserved_zero2; /* Zero name length */ 1557 __u8 det_reserved_ft; /* 0xDE, fake file type */ 1558 __le32 det_checksum; /* crc32c(uuid+inum+dirblock) */ 1559 }; 1560 1561 /* 1562 * Ext4 directory file types. Only the low 3 bits are used. The 1563 * other bits are reserved for now. 1564 */ 1565 #define EXT4_FT_UNKNOWN 0 1566 #define EXT4_FT_REG_FILE 1 1567 #define EXT4_FT_DIR 2 1568 #define EXT4_FT_CHRDEV 3 1569 #define EXT4_FT_BLKDEV 4 1570 #define EXT4_FT_FIFO 5 1571 #define EXT4_FT_SOCK 6 1572 #define EXT4_FT_SYMLINK 7 1573 1574 #define EXT4_FT_MAX 8 1575 1576 #define EXT4_FT_DIR_CSUM 0xDE 1577 1578 /* 1579 * EXT4_DIR_PAD defines the directory entries boundaries 1580 * 1581 * NOTE: It must be a multiple of 4 1582 */ 1583 #define EXT4_DIR_PAD 4 1584 #define EXT4_DIR_ROUND (EXT4_DIR_PAD - 1) 1585 #define EXT4_DIR_REC_LEN(name_len) (((name_len) + 8 + EXT4_DIR_ROUND) & \ 1586 ~EXT4_DIR_ROUND) 1587 #define EXT4_MAX_REC_LEN ((1<<16)-1) 1588 1589 /* 1590 * If we ever get support for fs block sizes > page_size, we'll need 1591 * to remove the #if statements in the next two functions... 1592 */ 1593 static inline unsigned int 1594 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize) 1595 { 1596 unsigned len = le16_to_cpu(dlen); 1597 1598 #if (PAGE_CACHE_SIZE >= 65536) 1599 if (len == EXT4_MAX_REC_LEN || len == 0) 1600 return blocksize; 1601 return (len & 65532) | ((len & 3) << 16); 1602 #else 1603 return len; 1604 #endif 1605 } 1606 1607 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize) 1608 { 1609 if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3)) 1610 BUG(); 1611 #if (PAGE_CACHE_SIZE >= 65536) 1612 if (len < 65536) 1613 return cpu_to_le16(len); 1614 if (len == blocksize) { 1615 if (blocksize == 65536) 1616 return cpu_to_le16(EXT4_MAX_REC_LEN); 1617 else 1618 return cpu_to_le16(0); 1619 } 1620 return cpu_to_le16((len & 65532) | ((len >> 16) & 3)); 1621 #else 1622 return cpu_to_le16(len); 1623 #endif 1624 } 1625 1626 /* 1627 * Hash Tree Directory indexing 1628 * (c) Daniel Phillips, 2001 1629 */ 1630 1631 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \ 1632 EXT4_FEATURE_COMPAT_DIR_INDEX) && \ 1633 ext4_test_inode_flag((dir), EXT4_INODE_INDEX)) 1634 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX) 1635 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1) 1636 1637 /* Legal values for the dx_root hash_version field: */ 1638 1639 #define DX_HASH_LEGACY 0 1640 #define DX_HASH_HALF_MD4 1 1641 #define DX_HASH_TEA 2 1642 #define DX_HASH_LEGACY_UNSIGNED 3 1643 #define DX_HASH_HALF_MD4_UNSIGNED 4 1644 #define DX_HASH_TEA_UNSIGNED 5 1645 1646 static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc, 1647 const void *address, unsigned int length) 1648 { 1649 struct { 1650 struct shash_desc shash; 1651 char ctx[crypto_shash_descsize(sbi->s_chksum_driver)]; 1652 } desc; 1653 int err; 1654 1655 desc.shash.tfm = sbi->s_chksum_driver; 1656 desc.shash.flags = 0; 1657 *(u32 *)desc.ctx = crc; 1658 1659 err = crypto_shash_update(&desc.shash, address, length); 1660 BUG_ON(err); 1661 1662 return *(u32 *)desc.ctx; 1663 } 1664 1665 #ifdef __KERNEL__ 1666 1667 /* hash info structure used by the directory hash */ 1668 struct dx_hash_info 1669 { 1670 u32 hash; 1671 u32 minor_hash; 1672 int hash_version; 1673 u32 *seed; 1674 }; 1675 1676 1677 /* 32 and 64 bit signed EOF for dx directories */ 1678 #define EXT4_HTREE_EOF_32BIT ((1UL << (32 - 1)) - 1) 1679 #define EXT4_HTREE_EOF_64BIT ((1ULL << (64 - 1)) - 1) 1680 1681 1682 /* 1683 * Control parameters used by ext4_htree_next_block 1684 */ 1685 #define HASH_NB_ALWAYS 1 1686 1687 1688 /* 1689 * Describe an inode's exact location on disk and in memory 1690 */ 1691 struct ext4_iloc 1692 { 1693 struct buffer_head *bh; 1694 unsigned long offset; 1695 ext4_group_t block_group; 1696 }; 1697 1698 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc) 1699 { 1700 return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset); 1701 } 1702 1703 /* 1704 * This structure is stuffed into the struct file's private_data field 1705 * for directories. It is where we put information so that we can do 1706 * readdir operations in hash tree order. 1707 */ 1708 struct dir_private_info { 1709 struct rb_root root; 1710 struct rb_node *curr_node; 1711 struct fname *extra_fname; 1712 loff_t last_pos; 1713 __u32 curr_hash; 1714 __u32 curr_minor_hash; 1715 __u32 next_hash; 1716 }; 1717 1718 /* calculate the first block number of the group */ 1719 static inline ext4_fsblk_t 1720 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no) 1721 { 1722 return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) + 1723 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block); 1724 } 1725 1726 /* 1727 * Special error return code only used by dx_probe() and its callers. 1728 */ 1729 #define ERR_BAD_DX_DIR -75000 1730 1731 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr, 1732 ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp); 1733 1734 /* 1735 * Timeout and state flag for lazy initialization inode thread. 1736 */ 1737 #define EXT4_DEF_LI_WAIT_MULT 10 1738 #define EXT4_DEF_LI_MAX_START_DELAY 5 1739 #define EXT4_LAZYINIT_QUIT 0x0001 1740 #define EXT4_LAZYINIT_RUNNING 0x0002 1741 1742 /* 1743 * Lazy inode table initialization info 1744 */ 1745 struct ext4_lazy_init { 1746 unsigned long li_state; 1747 struct list_head li_request_list; 1748 struct mutex li_list_mtx; 1749 }; 1750 1751 struct ext4_li_request { 1752 struct super_block *lr_super; 1753 struct ext4_sb_info *lr_sbi; 1754 ext4_group_t lr_next_group; 1755 struct list_head lr_request; 1756 unsigned long lr_next_sched; 1757 unsigned long lr_timeout; 1758 }; 1759 1760 struct ext4_features { 1761 struct kobject f_kobj; 1762 struct completion f_kobj_unregister; 1763 }; 1764 1765 /* 1766 * This structure will be used for multiple mount protection. It will be 1767 * written into the block number saved in the s_mmp_block field in the 1768 * superblock. Programs that check MMP should assume that if 1769 * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe 1770 * to use the filesystem, regardless of how old the timestamp is. 1771 */ 1772 #define EXT4_MMP_MAGIC 0x004D4D50U /* ASCII for MMP */ 1773 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */ 1774 #define EXT4_MMP_SEQ_FSCK 0xE24D4D50U /* mmp_seq value when being fscked */ 1775 #define EXT4_MMP_SEQ_MAX 0xE24D4D4FU /* maximum valid mmp_seq value */ 1776 1777 struct mmp_struct { 1778 __le32 mmp_magic; /* Magic number for MMP */ 1779 __le32 mmp_seq; /* Sequence no. updated periodically */ 1780 1781 /* 1782 * mmp_time, mmp_nodename & mmp_bdevname are only used for information 1783 * purposes and do not affect the correctness of the algorithm 1784 */ 1785 __le64 mmp_time; /* Time last updated */ 1786 char mmp_nodename[64]; /* Node which last updated MMP block */ 1787 char mmp_bdevname[32]; /* Bdev which last updated MMP block */ 1788 1789 /* 1790 * mmp_check_interval is used to verify if the MMP block has been 1791 * updated on the block device. The value is updated based on the 1792 * maximum time to write the MMP block during an update cycle. 1793 */ 1794 __le16 mmp_check_interval; 1795 1796 __le16 mmp_pad1; 1797 __le32 mmp_pad2[226]; 1798 __le32 mmp_checksum; /* crc32c(uuid+mmp_block) */ 1799 }; 1800 1801 /* arguments passed to the mmp thread */ 1802 struct mmpd_data { 1803 struct buffer_head *bh; /* bh from initial read_mmp_block() */ 1804 struct super_block *sb; /* super block of the fs */ 1805 }; 1806 1807 /* 1808 * Check interval multiplier 1809 * The MMP block is written every update interval and initially checked every 1810 * update interval x the multiplier (the value is then adapted based on the 1811 * write latency). The reason is that writes can be delayed under load and we 1812 * don't want readers to incorrectly assume that the filesystem is no longer 1813 * in use. 1814 */ 1815 #define EXT4_MMP_CHECK_MULT 2UL 1816 1817 /* 1818 * Minimum interval for MMP checking in seconds. 1819 */ 1820 #define EXT4_MMP_MIN_CHECK_INTERVAL 5UL 1821 1822 /* 1823 * Maximum interval for MMP checking in seconds. 1824 */ 1825 #define EXT4_MMP_MAX_CHECK_INTERVAL 300UL 1826 1827 /* 1828 * Function prototypes 1829 */ 1830 1831 /* 1832 * Ok, these declarations are also in <linux/kernel.h> but none of the 1833 * ext4 source programs needs to include it so they are duplicated here. 1834 */ 1835 # define NORET_TYPE /**/ 1836 # define ATTRIB_NORET __attribute__((noreturn)) 1837 # define NORET_AND noreturn, 1838 1839 /* bitmap.c */ 1840 extern unsigned int ext4_count_free(struct buffer_head *, unsigned); 1841 1842 /* balloc.c */ 1843 extern unsigned int ext4_block_group(struct super_block *sb, 1844 ext4_fsblk_t blocknr); 1845 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb, 1846 ext4_fsblk_t blocknr); 1847 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group); 1848 extern unsigned long ext4_bg_num_gdb(struct super_block *sb, 1849 ext4_group_t group); 1850 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode, 1851 ext4_fsblk_t goal, 1852 unsigned int flags, 1853 unsigned long *count, 1854 int *errp); 1855 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi, 1856 s64 nclusters, unsigned int flags); 1857 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *); 1858 extern void ext4_check_blocks_bitmap(struct super_block *); 1859 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb, 1860 ext4_group_t block_group, 1861 struct buffer_head ** bh); 1862 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries); 1863 1864 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb, 1865 ext4_group_t block_group); 1866 extern int ext4_wait_block_bitmap(struct super_block *sb, 1867 ext4_group_t block_group, 1868 struct buffer_head *bh); 1869 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb, 1870 ext4_group_t block_group); 1871 extern void ext4_init_block_bitmap(struct super_block *sb, 1872 struct buffer_head *bh, 1873 ext4_group_t group, 1874 struct ext4_group_desc *desc); 1875 extern unsigned ext4_free_clusters_after_init(struct super_block *sb, 1876 ext4_group_t block_group, 1877 struct ext4_group_desc *gdp); 1878 extern unsigned ext4_num_overhead_clusters(struct super_block *sb, 1879 ext4_group_t block_group, 1880 struct ext4_group_desc *gdp); 1881 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *); 1882 1883 /* dir.c */ 1884 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *, 1885 struct file *, 1886 struct ext4_dir_entry_2 *, 1887 struct buffer_head *, unsigned int); 1888 #define ext4_check_dir_entry(dir, filp, de, bh, offset) \ 1889 unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \ 1890 (de), (bh), (offset))) 1891 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash, 1892 __u32 minor_hash, 1893 struct ext4_dir_entry_2 *dirent); 1894 extern void ext4_htree_free_dir_info(struct dir_private_info *p); 1895 1896 /* fsync.c */ 1897 extern int ext4_sync_file(struct file *, loff_t, loff_t, int); 1898 extern int ext4_flush_completed_IO(struct inode *); 1899 1900 /* hash.c */ 1901 extern int ext4fs_dirhash(const char *name, int len, struct 1902 dx_hash_info *hinfo); 1903 1904 /* ialloc.c */ 1905 extern struct inode *ext4_new_inode(handle_t *, struct inode *, umode_t, 1906 const struct qstr *qstr, __u32 goal, 1907 uid_t *owner); 1908 extern void ext4_free_inode(handle_t *, struct inode *); 1909 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long); 1910 extern unsigned long ext4_count_free_inodes(struct super_block *); 1911 extern unsigned long ext4_count_dirs(struct super_block *); 1912 extern void ext4_check_inodes_bitmap(struct super_block *); 1913 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap); 1914 extern int ext4_init_inode_table(struct super_block *sb, 1915 ext4_group_t group, int barrier); 1916 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate); 1917 1918 /* mballoc.c */ 1919 extern long ext4_mb_stats; 1920 extern long ext4_mb_max_to_scan; 1921 extern int ext4_mb_init(struct super_block *, int); 1922 extern int ext4_mb_release(struct super_block *); 1923 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *, 1924 struct ext4_allocation_request *, int *); 1925 extern int ext4_mb_reserve_blocks(struct super_block *, int); 1926 extern void ext4_discard_preallocations(struct inode *); 1927 extern int __init ext4_init_mballoc(void); 1928 extern void ext4_exit_mballoc(void); 1929 extern void ext4_free_blocks(handle_t *handle, struct inode *inode, 1930 struct buffer_head *bh, ext4_fsblk_t block, 1931 unsigned long count, int flags); 1932 extern int ext4_mb_add_groupinfo(struct super_block *sb, 1933 ext4_group_t i, struct ext4_group_desc *desc); 1934 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb, 1935 ext4_fsblk_t block, unsigned long count); 1936 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *); 1937 1938 /* inode.c */ 1939 struct buffer_head *ext4_getblk(handle_t *, struct inode *, 1940 ext4_lblk_t, int, int *); 1941 struct buffer_head *ext4_bread(handle_t *, struct inode *, 1942 ext4_lblk_t, int, int *); 1943 int ext4_get_block(struct inode *inode, sector_t iblock, 1944 struct buffer_head *bh_result, int create); 1945 1946 extern struct inode *ext4_iget(struct super_block *, unsigned long); 1947 extern int ext4_write_inode(struct inode *, struct writeback_control *); 1948 extern int ext4_setattr(struct dentry *, struct iattr *); 1949 extern int ext4_getattr(struct vfsmount *mnt, struct dentry *dentry, 1950 struct kstat *stat); 1951 extern void ext4_evict_inode(struct inode *); 1952 extern void ext4_clear_inode(struct inode *); 1953 extern int ext4_sync_inode(handle_t *, struct inode *); 1954 extern void ext4_dirty_inode(struct inode *, int); 1955 extern int ext4_change_inode_journal_flag(struct inode *, int); 1956 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *); 1957 extern int ext4_can_truncate(struct inode *inode); 1958 extern void ext4_truncate(struct inode *); 1959 extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length); 1960 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks); 1961 extern void ext4_set_inode_flags(struct inode *); 1962 extern void ext4_get_inode_flags(struct ext4_inode_info *); 1963 extern int ext4_alloc_da_blocks(struct inode *inode); 1964 extern void ext4_set_aops(struct inode *inode); 1965 extern int ext4_writepage_trans_blocks(struct inode *); 1966 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks); 1967 extern int ext4_discard_partial_page_buffers(handle_t *handle, 1968 struct address_space *mapping, loff_t from, 1969 loff_t length, int flags); 1970 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf); 1971 extern qsize_t *ext4_get_reserved_space(struct inode *inode); 1972 extern void ext4_da_update_reserve_space(struct inode *inode, 1973 int used, int quota_claim); 1974 1975 /* indirect.c */ 1976 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode, 1977 struct ext4_map_blocks *map, int flags); 1978 extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb, 1979 const struct iovec *iov, loff_t offset, 1980 unsigned long nr_segs); 1981 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock); 1982 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks, int chunk); 1983 extern void ext4_ind_truncate(struct inode *inode); 1984 1985 /* ioctl.c */ 1986 extern long ext4_ioctl(struct file *, unsigned int, unsigned long); 1987 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long); 1988 1989 /* migrate.c */ 1990 extern int ext4_ext_migrate(struct inode *); 1991 1992 /* namei.c */ 1993 extern int ext4_orphan_add(handle_t *, struct inode *); 1994 extern int ext4_orphan_del(handle_t *, struct inode *); 1995 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash, 1996 __u32 start_minor_hash, __u32 *next_hash); 1997 1998 /* resize.c */ 1999 extern int ext4_group_add(struct super_block *sb, 2000 struct ext4_new_group_data *input); 2001 extern int ext4_group_extend(struct super_block *sb, 2002 struct ext4_super_block *es, 2003 ext4_fsblk_t n_blocks_count); 2004 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count); 2005 2006 /* super.c */ 2007 extern void *ext4_kvmalloc(size_t size, gfp_t flags); 2008 extern void *ext4_kvzalloc(size_t size, gfp_t flags); 2009 extern void ext4_kvfree(void *ptr); 2010 extern __printf(4, 5) 2011 void __ext4_error(struct super_block *, const char *, unsigned int, 2012 const char *, ...); 2013 #define ext4_error(sb, message...) __ext4_error(sb, __func__, \ 2014 __LINE__, ## message) 2015 extern __printf(5, 6) 2016 void ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t, 2017 const char *, ...); 2018 extern __printf(5, 6) 2019 void ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t, 2020 const char *, ...); 2021 extern void __ext4_std_error(struct super_block *, const char *, 2022 unsigned int, int); 2023 extern __printf(4, 5) 2024 void __ext4_abort(struct super_block *, const char *, unsigned int, 2025 const char *, ...); 2026 #define ext4_abort(sb, message...) __ext4_abort(sb, __func__, \ 2027 __LINE__, ## message) 2028 extern __printf(4, 5) 2029 void __ext4_warning(struct super_block *, const char *, unsigned int, 2030 const char *, ...); 2031 #define ext4_warning(sb, message...) __ext4_warning(sb, __func__, \ 2032 __LINE__, ## message) 2033 extern __printf(3, 4) 2034 void ext4_msg(struct super_block *, const char *, const char *, ...); 2035 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp, 2036 const char *, unsigned int, const char *); 2037 #define dump_mmp_msg(sb, mmp, msg) __dump_mmp_msg(sb, mmp, __func__, \ 2038 __LINE__, msg) 2039 extern __printf(7, 8) 2040 void __ext4_grp_locked_error(const char *, unsigned int, 2041 struct super_block *, ext4_group_t, 2042 unsigned long, ext4_fsblk_t, 2043 const char *, ...); 2044 #define ext4_grp_locked_error(sb, grp, message...) \ 2045 __ext4_grp_locked_error(__func__, __LINE__, (sb), (grp), ## message) 2046 extern void ext4_update_dynamic_rev(struct super_block *sb); 2047 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb, 2048 __u32 compat); 2049 extern int ext4_update_rocompat_feature(handle_t *handle, 2050 struct super_block *sb, __u32 rocompat); 2051 extern int ext4_update_incompat_feature(handle_t *handle, 2052 struct super_block *sb, __u32 incompat); 2053 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 2054 struct ext4_group_desc *bg); 2055 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 2056 struct ext4_group_desc *bg); 2057 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb, 2058 struct ext4_group_desc *bg); 2059 extern __u32 ext4_free_group_clusters(struct super_block *sb, 2060 struct ext4_group_desc *bg); 2061 extern __u32 ext4_free_inodes_count(struct super_block *sb, 2062 struct ext4_group_desc *bg); 2063 extern __u32 ext4_used_dirs_count(struct super_block *sb, 2064 struct ext4_group_desc *bg); 2065 extern __u32 ext4_itable_unused_count(struct super_block *sb, 2066 struct ext4_group_desc *bg); 2067 extern void ext4_block_bitmap_set(struct super_block *sb, 2068 struct ext4_group_desc *bg, ext4_fsblk_t blk); 2069 extern void ext4_inode_bitmap_set(struct super_block *sb, 2070 struct ext4_group_desc *bg, ext4_fsblk_t blk); 2071 extern void ext4_inode_table_set(struct super_block *sb, 2072 struct ext4_group_desc *bg, ext4_fsblk_t blk); 2073 extern void ext4_free_group_clusters_set(struct super_block *sb, 2074 struct ext4_group_desc *bg, 2075 __u32 count); 2076 extern void ext4_free_inodes_set(struct super_block *sb, 2077 struct ext4_group_desc *bg, __u32 count); 2078 extern void ext4_used_dirs_set(struct super_block *sb, 2079 struct ext4_group_desc *bg, __u32 count); 2080 extern void ext4_itable_unused_set(struct super_block *sb, 2081 struct ext4_group_desc *bg, __u32 count); 2082 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group, 2083 struct ext4_group_desc *gdp); 2084 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group, 2085 struct ext4_group_desc *gdp); 2086 2087 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es) 2088 { 2089 return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) | 2090 le32_to_cpu(es->s_blocks_count_lo); 2091 } 2092 2093 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es) 2094 { 2095 return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) | 2096 le32_to_cpu(es->s_r_blocks_count_lo); 2097 } 2098 2099 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es) 2100 { 2101 return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) | 2102 le32_to_cpu(es->s_free_blocks_count_lo); 2103 } 2104 2105 static inline void ext4_blocks_count_set(struct ext4_super_block *es, 2106 ext4_fsblk_t blk) 2107 { 2108 es->s_blocks_count_lo = cpu_to_le32((u32)blk); 2109 es->s_blocks_count_hi = cpu_to_le32(blk >> 32); 2110 } 2111 2112 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es, 2113 ext4_fsblk_t blk) 2114 { 2115 es->s_free_blocks_count_lo = cpu_to_le32((u32)blk); 2116 es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32); 2117 } 2118 2119 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es, 2120 ext4_fsblk_t blk) 2121 { 2122 es->s_r_blocks_count_lo = cpu_to_le32((u32)blk); 2123 es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32); 2124 } 2125 2126 static inline loff_t ext4_isize(struct ext4_inode *raw_inode) 2127 { 2128 if (S_ISREG(le16_to_cpu(raw_inode->i_mode))) 2129 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) | 2130 le32_to_cpu(raw_inode->i_size_lo); 2131 else 2132 return (loff_t) le32_to_cpu(raw_inode->i_size_lo); 2133 } 2134 2135 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size) 2136 { 2137 raw_inode->i_size_lo = cpu_to_le32(i_size); 2138 raw_inode->i_size_high = cpu_to_le32(i_size >> 32); 2139 } 2140 2141 static inline 2142 struct ext4_group_info *ext4_get_group_info(struct super_block *sb, 2143 ext4_group_t group) 2144 { 2145 struct ext4_group_info ***grp_info; 2146 long indexv, indexh; 2147 grp_info = EXT4_SB(sb)->s_group_info; 2148 indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb)); 2149 indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1); 2150 return grp_info[indexv][indexh]; 2151 } 2152 2153 /* 2154 * Reading s_groups_count requires using smp_rmb() afterwards. See 2155 * the locking protocol documented in the comments of ext4_group_add() 2156 * in resize.c 2157 */ 2158 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb) 2159 { 2160 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 2161 2162 smp_rmb(); 2163 return ngroups; 2164 } 2165 2166 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi, 2167 ext4_group_t block_group) 2168 { 2169 return block_group >> sbi->s_log_groups_per_flex; 2170 } 2171 2172 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi) 2173 { 2174 return 1 << sbi->s_log_groups_per_flex; 2175 } 2176 2177 #define ext4_std_error(sb, errno) \ 2178 do { \ 2179 if ((errno)) \ 2180 __ext4_std_error((sb), __func__, __LINE__, (errno)); \ 2181 } while (0) 2182 2183 #ifdef CONFIG_SMP 2184 /* Each CPU can accumulate percpu_counter_batch clusters in their local 2185 * counters. So we need to make sure we have free clusters more 2186 * than percpu_counter_batch * nr_cpu_ids. Also add a window of 4 times. 2187 */ 2188 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids)) 2189 #else 2190 #define EXT4_FREECLUSTERS_WATERMARK 0 2191 #endif 2192 2193 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize) 2194 { 2195 /* 2196 * XXX: replace with spinlock if seen contended -bzzz 2197 */ 2198 down_write(&EXT4_I(inode)->i_data_sem); 2199 if (newsize > EXT4_I(inode)->i_disksize) 2200 EXT4_I(inode)->i_disksize = newsize; 2201 up_write(&EXT4_I(inode)->i_data_sem); 2202 return ; 2203 } 2204 2205 struct ext4_group_info { 2206 unsigned long bb_state; 2207 struct rb_root bb_free_root; 2208 ext4_grpblk_t bb_first_free; /* first free block */ 2209 ext4_grpblk_t bb_free; /* total free blocks */ 2210 ext4_grpblk_t bb_fragments; /* nr of freespace fragments */ 2211 ext4_grpblk_t bb_largest_free_order;/* order of largest frag in BG */ 2212 struct list_head bb_prealloc_list; 2213 #ifdef DOUBLE_CHECK 2214 void *bb_bitmap; 2215 #endif 2216 struct rw_semaphore alloc_sem; 2217 ext4_grpblk_t bb_counters[]; /* Nr of free power-of-two-block 2218 * regions, index is order. 2219 * bb_counters[3] = 5 means 2220 * 5 free 8-block regions. */ 2221 }; 2222 2223 #define EXT4_GROUP_INFO_NEED_INIT_BIT 0 2224 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT 1 2225 2226 #define EXT4_MB_GRP_NEED_INIT(grp) \ 2227 (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state))) 2228 2229 #define EXT4_MB_GRP_WAS_TRIMMED(grp) \ 2230 (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state))) 2231 #define EXT4_MB_GRP_SET_TRIMMED(grp) \ 2232 (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state))) 2233 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp) \ 2234 (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state))) 2235 2236 #define EXT4_MAX_CONTENTION 8 2237 #define EXT4_CONTENTION_THRESHOLD 2 2238 2239 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb, 2240 ext4_group_t group) 2241 { 2242 return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group); 2243 } 2244 2245 /* 2246 * Returns true if the filesystem is busy enough that attempts to 2247 * access the block group locks has run into contention. 2248 */ 2249 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi) 2250 { 2251 return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD); 2252 } 2253 2254 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group) 2255 { 2256 spinlock_t *lock = ext4_group_lock_ptr(sb, group); 2257 if (spin_trylock(lock)) 2258 /* 2259 * We're able to grab the lock right away, so drop the 2260 * lock contention counter. 2261 */ 2262 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0); 2263 else { 2264 /* 2265 * The lock is busy, so bump the contention counter, 2266 * and then wait on the spin lock. 2267 */ 2268 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1, 2269 EXT4_MAX_CONTENTION); 2270 spin_lock(lock); 2271 } 2272 } 2273 2274 static inline void ext4_unlock_group(struct super_block *sb, 2275 ext4_group_t group) 2276 { 2277 spin_unlock(ext4_group_lock_ptr(sb, group)); 2278 } 2279 2280 static inline void ext4_mark_super_dirty(struct super_block *sb) 2281 { 2282 if (EXT4_SB(sb)->s_journal == NULL) 2283 sb->s_dirt =1; 2284 } 2285 2286 /* 2287 * Block validity checking 2288 */ 2289 #define ext4_check_indirect_blockref(inode, bh) \ 2290 ext4_check_blockref(__func__, __LINE__, inode, \ 2291 (__le32 *)(bh)->b_data, \ 2292 EXT4_ADDR_PER_BLOCK((inode)->i_sb)) 2293 2294 #define ext4_ind_check_inode(inode) \ 2295 ext4_check_blockref(__func__, __LINE__, inode, \ 2296 EXT4_I(inode)->i_data, \ 2297 EXT4_NDIR_BLOCKS) 2298 2299 /* 2300 * Inodes and files operations 2301 */ 2302 2303 /* dir.c */ 2304 extern const struct file_operations ext4_dir_operations; 2305 2306 /* file.c */ 2307 extern const struct inode_operations ext4_file_inode_operations; 2308 extern const struct file_operations ext4_file_operations; 2309 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin); 2310 2311 /* namei.c */ 2312 extern const struct inode_operations ext4_dir_inode_operations; 2313 extern const struct inode_operations ext4_special_inode_operations; 2314 extern struct dentry *ext4_get_parent(struct dentry *child); 2315 2316 /* symlink.c */ 2317 extern const struct inode_operations ext4_symlink_inode_operations; 2318 extern const struct inode_operations ext4_fast_symlink_inode_operations; 2319 2320 /* block_validity */ 2321 extern void ext4_release_system_zone(struct super_block *sb); 2322 extern int ext4_setup_system_zone(struct super_block *sb); 2323 extern int __init ext4_init_system_zone(void); 2324 extern void ext4_exit_system_zone(void); 2325 extern int ext4_data_block_valid(struct ext4_sb_info *sbi, 2326 ext4_fsblk_t start_blk, 2327 unsigned int count); 2328 extern int ext4_check_blockref(const char *, unsigned int, 2329 struct inode *, __le32 *, unsigned int); 2330 2331 /* extents.c */ 2332 extern int ext4_ext_tree_init(handle_t *handle, struct inode *); 2333 extern int ext4_ext_writepage_trans_blocks(struct inode *, int); 2334 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks, 2335 int chunk); 2336 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode, 2337 struct ext4_map_blocks *map, int flags); 2338 extern void ext4_ext_truncate(struct inode *); 2339 extern int ext4_ext_punch_hole(struct file *file, loff_t offset, 2340 loff_t length); 2341 extern void ext4_ext_init(struct super_block *); 2342 extern void ext4_ext_release(struct super_block *); 2343 extern long ext4_fallocate(struct file *file, int mode, loff_t offset, 2344 loff_t len); 2345 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset, 2346 ssize_t len); 2347 extern int ext4_map_blocks(handle_t *handle, struct inode *inode, 2348 struct ext4_map_blocks *map, int flags); 2349 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, 2350 __u64 start, __u64 len); 2351 /* move_extent.c */ 2352 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp, 2353 __u64 start_orig, __u64 start_donor, 2354 __u64 len, __u64 *moved_len); 2355 2356 /* page-io.c */ 2357 extern int __init ext4_init_pageio(void); 2358 extern void ext4_exit_pageio(void); 2359 extern void ext4_ioend_wait(struct inode *); 2360 extern void ext4_free_io_end(ext4_io_end_t *io); 2361 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags); 2362 extern int ext4_end_io_nolock(ext4_io_end_t *io); 2363 extern void ext4_io_submit(struct ext4_io_submit *io); 2364 extern int ext4_bio_write_page(struct ext4_io_submit *io, 2365 struct page *page, 2366 int len, 2367 struct writeback_control *wbc); 2368 2369 /* mmp.c */ 2370 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t); 2371 2372 /* BH_Uninit flag: blocks are allocated but uninitialized on disk */ 2373 enum ext4_state_bits { 2374 BH_Uninit /* blocks are allocated but uninitialized on disk */ 2375 = BH_JBDPrivateStart, 2376 BH_AllocFromCluster, /* allocated blocks were part of already 2377 * allocated cluster. Note that this flag will 2378 * never, ever appear in a buffer_head's state 2379 * flag. See EXT4_MAP_FROM_CLUSTER to see where 2380 * this is used. */ 2381 BH_Da_Mapped, /* Delayed allocated block that now has a mapping. This 2382 * flag is set when ext4_map_blocks is called on a 2383 * delayed allocated block to get its real mapping. */ 2384 }; 2385 2386 BUFFER_FNS(Uninit, uninit) 2387 TAS_BUFFER_FNS(Uninit, uninit) 2388 BUFFER_FNS(Da_Mapped, da_mapped) 2389 2390 /* 2391 * Add new method to test wether block and inode bitmaps are properly 2392 * initialized. With uninit_bg reading the block from disk is not enough 2393 * to mark the bitmap uptodate. We need to also zero-out the bitmap 2394 */ 2395 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart 2396 2397 static inline int bitmap_uptodate(struct buffer_head *bh) 2398 { 2399 return (buffer_uptodate(bh) && 2400 test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state)); 2401 } 2402 static inline void set_bitmap_uptodate(struct buffer_head *bh) 2403 { 2404 set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state); 2405 } 2406 2407 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1) 2408 2409 /* For ioend & aio unwritten conversion wait queues */ 2410 #define EXT4_WQ_HASH_SZ 37 2411 #define ext4_ioend_wq(v) (&ext4__ioend_wq[((unsigned long)(v)) %\ 2412 EXT4_WQ_HASH_SZ]) 2413 #define ext4_aio_mutex(v) (&ext4__aio_mutex[((unsigned long)(v)) %\ 2414 EXT4_WQ_HASH_SZ]) 2415 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 2416 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ]; 2417 2418 #define EXT4_RESIZING 0 2419 extern int ext4_resize_begin(struct super_block *sb); 2420 extern void ext4_resize_end(struct super_block *sb); 2421 2422 #endif /* __KERNEL__ */ 2423 2424 #include "ext4_extents.h" 2425 2426 #endif /* _EXT4_H */ 2427