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 <linux/ratelimit.h> 33#include <crypto/hash.h> 34#include <linux/falloc.h> 35#ifdef __KERNEL__ 36#include <linux/compat.h> 37#endif 38 39/* 40 * The fourth extended filesystem constants/structures 41 */ 42 43/* 44 * Define EXT4FS_DEBUG to produce debug messages 45 */ 46#undef EXT4FS_DEBUG 47 48/* 49 * Debug code 50 */ 51#ifdef EXT4FS_DEBUG 52#define ext4_debug(f, a...) \ 53 do { \ 54 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:", \ 55 __FILE__, __LINE__, __func__); \ 56 printk(KERN_DEBUG f, ## a); \ 57 } while (0) 58#else 59#define ext4_debug(fmt, ...) no_printk(fmt, ##__VA_ARGS__) 60#endif 61 62/* 63 * Turn on EXT_DEBUG to get lots of info about extents operations. 64 */ 65#define EXT_DEBUG__ 66#ifdef EXT_DEBUG 67#define ext_debug(fmt, ...) printk(fmt, ##__VA_ARGS__) 68#else 69#define ext_debug(fmt, ...) no_printk(fmt, ##__VA_ARGS__) 70#endif 71 72#define EXT4_ERROR_INODE(inode, fmt, a...) \ 73 ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a) 74 75#define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...) \ 76 ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a) 77 78#define EXT4_ERROR_FILE(file, block, fmt, a...) \ 79 ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a) 80 81/* data type for block offset of block group */ 82typedef int ext4_grpblk_t; 83 84/* data type for filesystem-wide blocks number */ 85typedef unsigned long long ext4_fsblk_t; 86 87/* data type for file logical block number */ 88typedef __u32 ext4_lblk_t; 89 90/* data type for block group number */ 91typedef unsigned int ext4_group_t; 92 93/* 94 * Flags used in mballoc's allocation_context flags field. 95 * 96 * Also used to show what's going on for debugging purposes when the 97 * flag field is exported via the traceport interface 98 */ 99 100/* prefer goal again. length */ 101#define EXT4_MB_HINT_MERGE 0x0001 102/* blocks already reserved */ 103#define EXT4_MB_HINT_RESERVED 0x0002 104/* metadata is being allocated */ 105#define EXT4_MB_HINT_METADATA 0x0004 106/* first blocks in the file */ 107#define EXT4_MB_HINT_FIRST 0x0008 108/* search for the best chunk */ 109#define EXT4_MB_HINT_BEST 0x0010 110/* data is being allocated */ 111#define EXT4_MB_HINT_DATA 0x0020 112/* don't preallocate (for tails) */ 113#define EXT4_MB_HINT_NOPREALLOC 0x0040 114/* allocate for locality group */ 115#define EXT4_MB_HINT_GROUP_ALLOC 0x0080 116/* allocate goal blocks or none */ 117#define EXT4_MB_HINT_GOAL_ONLY 0x0100 118/* goal is meaningful */ 119#define EXT4_MB_HINT_TRY_GOAL 0x0200 120/* blocks already pre-reserved by delayed allocation */ 121#define EXT4_MB_DELALLOC_RESERVED 0x0400 122/* We are doing stream allocation */ 123#define EXT4_MB_STREAM_ALLOC 0x0800 124/* Use reserved root blocks if needed */ 125#define EXT4_MB_USE_ROOT_BLOCKS 0x1000 126/* Use blocks from reserved pool */ 127#define EXT4_MB_USE_RESERVED 0x2000 128 129struct ext4_allocation_request { 130 /* target inode for block we're allocating */ 131 struct inode *inode; 132 /* how many blocks we want to allocate */ 133 unsigned int len; 134 /* logical block in target inode */ 135 ext4_lblk_t logical; 136 /* the closest logical allocated block to the left */ 137 ext4_lblk_t lleft; 138 /* the closest logical allocated block to the right */ 139 ext4_lblk_t lright; 140 /* phys. target (a hint) */ 141 ext4_fsblk_t goal; 142 /* phys. block for the closest logical allocated block to the left */ 143 ext4_fsblk_t pleft; 144 /* phys. block for the closest logical allocated block to the right */ 145 ext4_fsblk_t pright; 146 /* flags. see above EXT4_MB_HINT_* */ 147 unsigned int flags; 148}; 149 150/* 151 * Logical to physical block mapping, used by ext4_map_blocks() 152 * 153 * This structure is used to pass requests into ext4_map_blocks() as 154 * well as to store the information returned by ext4_map_blocks(). It 155 * takes less room on the stack than a struct buffer_head. 156 */ 157#define EXT4_MAP_NEW (1 << BH_New) 158#define EXT4_MAP_MAPPED (1 << BH_Mapped) 159#define EXT4_MAP_UNWRITTEN (1 << BH_Unwritten) 160#define EXT4_MAP_BOUNDARY (1 << BH_Boundary) 161/* Sometimes (in the bigalloc case, from ext4_da_get_block_prep) the caller of 162 * ext4_map_blocks wants to know whether or not the underlying cluster has 163 * already been accounted for. EXT4_MAP_FROM_CLUSTER conveys to the caller that 164 * the requested mapping was from previously mapped (or delayed allocated) 165 * cluster. We use BH_AllocFromCluster only for this flag. BH_AllocFromCluster 166 * should never appear on buffer_head's state flags. 167 */ 168#define EXT4_MAP_FROM_CLUSTER (1 << BH_AllocFromCluster) 169#define EXT4_MAP_FLAGS (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\ 170 EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\ 171 EXT4_MAP_FROM_CLUSTER) 172 173struct ext4_map_blocks { 174 ext4_fsblk_t m_pblk; 175 ext4_lblk_t m_lblk; 176 unsigned int m_len; 177 unsigned int m_flags; 178}; 179 180/* 181 * Flags for ext4_io_end->flags 182 */ 183#define EXT4_IO_END_UNWRITTEN 0x0001 184 185/* 186 * For converting unwritten extents on a work queue. 'handle' is used for 187 * buffered writeback. 188 */ 189typedef struct ext4_io_end { 190 struct list_head list; /* per-file finished IO list */ 191 handle_t *handle; /* handle reserved for extent 192 * conversion */ 193 struct inode *inode; /* file being written to */ 194 struct bio *bio; /* Linked list of completed 195 * bios covering the extent */ 196 unsigned int flag; /* unwritten or not */ 197 loff_t offset; /* offset in the file */ 198 ssize_t size; /* size of the extent */ 199 atomic_t count; /* reference counter */ 200} ext4_io_end_t; 201 202struct ext4_io_submit { 203 int io_op; 204 struct bio *io_bio; 205 ext4_io_end_t *io_end; 206 sector_t io_next_block; 207}; 208 209/* 210 * Special inodes numbers 211 */ 212#define EXT4_BAD_INO 1 /* Bad blocks inode */ 213#define EXT4_ROOT_INO 2 /* Root inode */ 214#define EXT4_USR_QUOTA_INO 3 /* User quota inode */ 215#define EXT4_GRP_QUOTA_INO 4 /* Group quota inode */ 216#define EXT4_BOOT_LOADER_INO 5 /* Boot loader inode */ 217#define EXT4_UNDEL_DIR_INO 6 /* Undelete directory inode */ 218#define EXT4_RESIZE_INO 7 /* Reserved group descriptors inode */ 219#define EXT4_JOURNAL_INO 8 /* Journal inode */ 220 221/* First non-reserved inode for old ext4 filesystems */ 222#define EXT4_GOOD_OLD_FIRST_INO 11 223 224/* 225 * Maximal count of links to a file 226 */ 227#define EXT4_LINK_MAX 65000 228 229/* 230 * Macro-instructions used to manage several block sizes 231 */ 232#define EXT4_MIN_BLOCK_SIZE 1024 233#define EXT4_MAX_BLOCK_SIZE 65536 234#define EXT4_MIN_BLOCK_LOG_SIZE 10 235#define EXT4_MAX_BLOCK_LOG_SIZE 16 236#ifdef __KERNEL__ 237# define EXT4_BLOCK_SIZE(s) ((s)->s_blocksize) 238#else 239# define EXT4_BLOCK_SIZE(s) (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size) 240#endif 241#define EXT4_ADDR_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / sizeof(__u32)) 242#define EXT4_CLUSTER_SIZE(s) (EXT4_BLOCK_SIZE(s) << \ 243 EXT4_SB(s)->s_cluster_bits) 244#ifdef __KERNEL__ 245# define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits) 246# define EXT4_CLUSTER_BITS(s) (EXT4_SB(s)->s_cluster_bits) 247#else 248# define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_log_block_size + 10) 249#endif 250#ifdef __KERNEL__ 251#define EXT4_ADDR_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_addr_per_block_bits) 252#define EXT4_INODE_SIZE(s) (EXT4_SB(s)->s_inode_size) 253#define EXT4_FIRST_INO(s) (EXT4_SB(s)->s_first_ino) 254#else 255#define EXT4_INODE_SIZE(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \ 256 EXT4_GOOD_OLD_INODE_SIZE : \ 257 (s)->s_inode_size) 258#define EXT4_FIRST_INO(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \ 259 EXT4_GOOD_OLD_FIRST_INO : \ 260 (s)->s_first_ino) 261#endif 262#define EXT4_BLOCK_ALIGN(size, blkbits) ALIGN((size), (1 << (blkbits))) 263 264/* Translate a block number to a cluster number */ 265#define EXT4_B2C(sbi, blk) ((blk) >> (sbi)->s_cluster_bits) 266/* Translate a cluster number to a block number */ 267#define EXT4_C2B(sbi, cluster) ((cluster) << (sbi)->s_cluster_bits) 268/* Translate # of blks to # of clusters */ 269#define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \ 270 (sbi)->s_cluster_bits) 271/* Mask out the low bits to get the starting block of the cluster */ 272#define EXT4_PBLK_CMASK(s, pblk) ((pblk) & \ 273 ~((ext4_fsblk_t) (s)->s_cluster_ratio - 1)) 274#define EXT4_LBLK_CMASK(s, lblk) ((lblk) & \ 275 ~((ext4_lblk_t) (s)->s_cluster_ratio - 1)) 276/* Get the cluster offset */ 277#define EXT4_PBLK_COFF(s, pblk) ((pblk) & \ 278 ((ext4_fsblk_t) (s)->s_cluster_ratio - 1)) 279#define EXT4_LBLK_COFF(s, lblk) ((lblk) & \ 280 ((ext4_lblk_t) (s)->s_cluster_ratio - 1)) 281 282/* 283 * Structure of a blocks group descriptor 284 */ 285struct ext4_group_desc 286{ 287 __le32 bg_block_bitmap_lo; /* Blocks bitmap block */ 288 __le32 bg_inode_bitmap_lo; /* Inodes bitmap block */ 289 __le32 bg_inode_table_lo; /* Inodes table block */ 290 __le16 bg_free_blocks_count_lo;/* Free blocks count */ 291 __le16 bg_free_inodes_count_lo;/* Free inodes count */ 292 __le16 bg_used_dirs_count_lo; /* Directories count */ 293 __le16 bg_flags; /* EXT4_BG_flags (INODE_UNINIT, etc) */ 294 __le32 bg_exclude_bitmap_lo; /* Exclude bitmap for snapshots */ 295 __le16 bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */ 296 __le16 bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */ 297 __le16 bg_itable_unused_lo; /* Unused inodes count */ 298 __le16 bg_checksum; /* crc16(sb_uuid+group+desc) */ 299 __le32 bg_block_bitmap_hi; /* Blocks bitmap block MSB */ 300 __le32 bg_inode_bitmap_hi; /* Inodes bitmap block MSB */ 301 __le32 bg_inode_table_hi; /* Inodes table block MSB */ 302 __le16 bg_free_blocks_count_hi;/* Free blocks count MSB */ 303 __le16 bg_free_inodes_count_hi;/* Free inodes count MSB */ 304 __le16 bg_used_dirs_count_hi; /* Directories count MSB */ 305 __le16 bg_itable_unused_hi; /* Unused inodes count MSB */ 306 __le32 bg_exclude_bitmap_hi; /* Exclude bitmap block MSB */ 307 __le16 bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */ 308 __le16 bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */ 309 __u32 bg_reserved; 310}; 311 312#define EXT4_BG_INODE_BITMAP_CSUM_HI_END \ 313 (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \ 314 sizeof(__le16)) 315#define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END \ 316 (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \ 317 sizeof(__le16)) 318 319/* 320 * Structure of a flex block group info 321 */ 322 323struct flex_groups { 324 atomic64_t free_clusters; 325 atomic_t free_inodes; 326 atomic_t used_dirs; 327}; 328 329#define EXT4_BG_INODE_UNINIT 0x0001 /* Inode table/bitmap not in use */ 330#define EXT4_BG_BLOCK_UNINIT 0x0002 /* Block bitmap not in use */ 331#define EXT4_BG_INODE_ZEROED 0x0004 /* On-disk itable initialized to zero */ 332 333/* 334 * Macro-instructions used to manage group descriptors 335 */ 336#define EXT4_MIN_DESC_SIZE 32 337#define EXT4_MIN_DESC_SIZE_64BIT 64 338#define EXT4_MAX_DESC_SIZE EXT4_MIN_BLOCK_SIZE 339#define EXT4_DESC_SIZE(s) (EXT4_SB(s)->s_desc_size) 340#ifdef __KERNEL__ 341# define EXT4_BLOCKS_PER_GROUP(s) (EXT4_SB(s)->s_blocks_per_group) 342# define EXT4_CLUSTERS_PER_GROUP(s) (EXT4_SB(s)->s_clusters_per_group) 343# define EXT4_DESC_PER_BLOCK(s) (EXT4_SB(s)->s_desc_per_block) 344# define EXT4_INODES_PER_GROUP(s) (EXT4_SB(s)->s_inodes_per_group) 345# define EXT4_DESC_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_desc_per_block_bits) 346#else 347# define EXT4_BLOCKS_PER_GROUP(s) ((s)->s_blocks_per_group) 348# define EXT4_DESC_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s)) 349# define EXT4_INODES_PER_GROUP(s) ((s)->s_inodes_per_group) 350#endif 351 352/* 353 * Constants relative to the data blocks 354 */ 355#define EXT4_NDIR_BLOCKS 12 356#define EXT4_IND_BLOCK EXT4_NDIR_BLOCKS 357#define EXT4_DIND_BLOCK (EXT4_IND_BLOCK + 1) 358#define EXT4_TIND_BLOCK (EXT4_DIND_BLOCK + 1) 359#define EXT4_N_BLOCKS (EXT4_TIND_BLOCK + 1) 360 361/* 362 * Inode flags 363 */ 364#define EXT4_SECRM_FL 0x00000001 /* Secure deletion */ 365#define EXT4_UNRM_FL 0x00000002 /* Undelete */ 366#define EXT4_COMPR_FL 0x00000004 /* Compress file */ 367#define EXT4_SYNC_FL 0x00000008 /* Synchronous updates */ 368#define EXT4_IMMUTABLE_FL 0x00000010 /* Immutable file */ 369#define EXT4_APPEND_FL 0x00000020 /* writes to file may only append */ 370#define EXT4_NODUMP_FL 0x00000040 /* do not dump file */ 371#define EXT4_NOATIME_FL 0x00000080 /* do not update atime */ 372/* Reserved for compression usage... */ 373#define EXT4_DIRTY_FL 0x00000100 374#define EXT4_COMPRBLK_FL 0x00000200 /* One or more compressed clusters */ 375#define EXT4_NOCOMPR_FL 0x00000400 /* Don't compress */ 376#define EXT4_ECOMPR_FL 0x00000800 /* Compression error */ 377/* End compression flags --- maybe not all used */ 378#define EXT4_INDEX_FL 0x00001000 /* hash-indexed directory */ 379#define EXT4_IMAGIC_FL 0x00002000 /* AFS directory */ 380#define EXT4_JOURNAL_DATA_FL 0x00004000 /* file data should be journaled */ 381#define EXT4_NOTAIL_FL 0x00008000 /* file tail should not be merged */ 382#define EXT4_DIRSYNC_FL 0x00010000 /* dirsync behaviour (directories only) */ 383#define EXT4_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/ 384#define EXT4_HUGE_FILE_FL 0x00040000 /* Set to each huge file */ 385#define EXT4_EXTENTS_FL 0x00080000 /* Inode uses extents */ 386#define EXT4_EA_INODE_FL 0x00200000 /* Inode used for large EA */ 387#define EXT4_EOFBLOCKS_FL 0x00400000 /* Blocks allocated beyond EOF */ 388#define EXT4_INLINE_DATA_FL 0x10000000 /* Inode has inline data. */ 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 0x004380FF /* 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. */ 407static 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 */ 420enum { 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_INLINE_DATA = 28, /* Data in inode. */ 446 EXT4_INODE_RESERVED = 31, /* reserved for ext4 lib */ 447}; 448 449/* 450 * Since it's pretty easy to mix up bit numbers and hex values, we use a 451 * build-time check to make sure that EXT4_XXX_FL is consistent with respect to 452 * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost 453 * any extra space in the compiled kernel image, otherwise, the build will fail. 454 * It's important that these values are the same, since we are using 455 * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent 456 * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk 457 * values found in ext2, ext3 and ext4 filesystems, and of course the values 458 * defined in e2fsprogs. 459 * 460 * It's not paranoia if the Murphy's Law really *is* out to get you. :-) 461 */ 462#define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG)) 463#define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG)) 464 465static inline void ext4_check_flag_values(void) 466{ 467 CHECK_FLAG_VALUE(SECRM); 468 CHECK_FLAG_VALUE(UNRM); 469 CHECK_FLAG_VALUE(COMPR); 470 CHECK_FLAG_VALUE(SYNC); 471 CHECK_FLAG_VALUE(IMMUTABLE); 472 CHECK_FLAG_VALUE(APPEND); 473 CHECK_FLAG_VALUE(NODUMP); 474 CHECK_FLAG_VALUE(NOATIME); 475 CHECK_FLAG_VALUE(DIRTY); 476 CHECK_FLAG_VALUE(COMPRBLK); 477 CHECK_FLAG_VALUE(NOCOMPR); 478 CHECK_FLAG_VALUE(ECOMPR); 479 CHECK_FLAG_VALUE(INDEX); 480 CHECK_FLAG_VALUE(IMAGIC); 481 CHECK_FLAG_VALUE(JOURNAL_DATA); 482 CHECK_FLAG_VALUE(NOTAIL); 483 CHECK_FLAG_VALUE(DIRSYNC); 484 CHECK_FLAG_VALUE(TOPDIR); 485 CHECK_FLAG_VALUE(HUGE_FILE); 486 CHECK_FLAG_VALUE(EXTENTS); 487 CHECK_FLAG_VALUE(EA_INODE); 488 CHECK_FLAG_VALUE(EOFBLOCKS); 489 CHECK_FLAG_VALUE(INLINE_DATA); 490 CHECK_FLAG_VALUE(RESERVED); 491} 492 493/* Used to pass group descriptor data when online resize is done */ 494struct ext4_new_group_input { 495 __u32 group; /* Group number for this data */ 496 __u64 block_bitmap; /* Absolute block number of block bitmap */ 497 __u64 inode_bitmap; /* Absolute block number of inode bitmap */ 498 __u64 inode_table; /* Absolute block number of inode table start */ 499 __u32 blocks_count; /* Total number of blocks in this group */ 500 __u16 reserved_blocks; /* Number of reserved blocks in this group */ 501 __u16 unused; 502}; 503 504#if defined(__KERNEL__) && defined(CONFIG_COMPAT) 505struct compat_ext4_new_group_input { 506 u32 group; 507 compat_u64 block_bitmap; 508 compat_u64 inode_bitmap; 509 compat_u64 inode_table; 510 u32 blocks_count; 511 u16 reserved_blocks; 512 u16 unused; 513}; 514#endif 515 516/* The struct ext4_new_group_input in kernel space, with free_blocks_count */ 517struct ext4_new_group_data { 518 __u32 group; 519 __u64 block_bitmap; 520 __u64 inode_bitmap; 521 __u64 inode_table; 522 __u32 blocks_count; 523 __u16 reserved_blocks; 524 __u16 unused; 525 __u32 free_blocks_count; 526}; 527 528/* Indexes used to index group tables in ext4_new_group_data */ 529enum { 530 BLOCK_BITMAP = 0, /* block bitmap */ 531 INODE_BITMAP, /* inode bitmap */ 532 INODE_TABLE, /* inode tables */ 533 GROUP_TABLE_COUNT, 534}; 535 536/* 537 * Flags used by ext4_map_blocks() 538 */ 539 /* Allocate any needed blocks and/or convert an unwritten 540 extent to be an initialized ext4 */ 541#define EXT4_GET_BLOCKS_CREATE 0x0001 542 /* Request the creation of an unwritten extent */ 543#define EXT4_GET_BLOCKS_UNWRIT_EXT 0x0002 544#define EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT (EXT4_GET_BLOCKS_UNWRIT_EXT|\ 545 EXT4_GET_BLOCKS_CREATE) 546 /* Caller is from the delayed allocation writeout path 547 * finally doing the actual allocation of delayed blocks */ 548#define EXT4_GET_BLOCKS_DELALLOC_RESERVE 0x0004 549 /* caller is from the direct IO path, request to creation of an 550 unwritten extents if not allocated, split the unwritten 551 extent if blocks has been preallocated already*/ 552#define EXT4_GET_BLOCKS_PRE_IO 0x0008 553#define EXT4_GET_BLOCKS_CONVERT 0x0010 554#define EXT4_GET_BLOCKS_IO_CREATE_EXT (EXT4_GET_BLOCKS_PRE_IO|\ 555 EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT) 556 /* Convert extent to initialized after IO complete */ 557#define EXT4_GET_BLOCKS_IO_CONVERT_EXT (EXT4_GET_BLOCKS_CONVERT|\ 558 EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT) 559 /* Eventual metadata allocation (due to growing extent tree) 560 * should not fail, so try to use reserved blocks for that.*/ 561#define EXT4_GET_BLOCKS_METADATA_NOFAIL 0x0020 562 /* Don't normalize allocation size (used for fallocate) */ 563#define EXT4_GET_BLOCKS_NO_NORMALIZE 0x0040 564 /* Request will not result in inode size update (user for fallocate) */ 565#define EXT4_GET_BLOCKS_KEEP_SIZE 0x0080 566 /* Do not take i_data_sem locking in ext4_map_blocks */ 567#define EXT4_GET_BLOCKS_NO_LOCK 0x0100 568 /* Do not put hole in extent cache */ 569#define EXT4_GET_BLOCKS_NO_PUT_HOLE 0x0200 570 /* Convert written extents to unwritten */ 571#define EXT4_GET_BLOCKS_CONVERT_UNWRITTEN 0x0400 572 573/* 574 * The bit position of these flags must not overlap with any of the 575 * EXT4_GET_BLOCKS_*. They are used by ext4_find_extent(), 576 * read_extent_tree_block(), ext4_split_extent_at(), 577 * ext4_ext_insert_extent(), and ext4_ext_create_new_leaf(). 578 * EXT4_EX_NOCACHE is used to indicate that the we shouldn't be 579 * caching the extents when reading from the extent tree while a 580 * truncate or punch hole operation is in progress. 581 */ 582#define EXT4_EX_NOCACHE 0x40000000 583#define EXT4_EX_FORCE_CACHE 0x20000000 584 585/* 586 * Flags used by ext4_free_blocks 587 */ 588#define EXT4_FREE_BLOCKS_METADATA 0x0001 589#define EXT4_FREE_BLOCKS_FORGET 0x0002 590#define EXT4_FREE_BLOCKS_VALIDATED 0x0004 591#define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008 592#define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER 0x0010 593#define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER 0x0020 594 595/* 596 * ioctl commands 597 */ 598#define EXT4_IOC_GETFLAGS FS_IOC_GETFLAGS 599#define EXT4_IOC_SETFLAGS FS_IOC_SETFLAGS 600#define EXT4_IOC_GETVERSION _IOR('f', 3, long) 601#define EXT4_IOC_SETVERSION _IOW('f', 4, long) 602#define EXT4_IOC_GETVERSION_OLD FS_IOC_GETVERSION 603#define EXT4_IOC_SETVERSION_OLD FS_IOC_SETVERSION 604#define EXT4_IOC_GETRSVSZ _IOR('f', 5, long) 605#define EXT4_IOC_SETRSVSZ _IOW('f', 6, long) 606#define EXT4_IOC_GROUP_EXTEND _IOW('f', 7, unsigned long) 607#define EXT4_IOC_GROUP_ADD _IOW('f', 8, struct ext4_new_group_input) 608#define EXT4_IOC_MIGRATE _IO('f', 9) 609 /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */ 610 /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */ 611#define EXT4_IOC_ALLOC_DA_BLKS _IO('f', 12) 612#define EXT4_IOC_MOVE_EXT _IOWR('f', 15, struct move_extent) 613#define EXT4_IOC_RESIZE_FS _IOW('f', 16, __u64) 614#define EXT4_IOC_SWAP_BOOT _IO('f', 17) 615#define EXT4_IOC_PRECACHE_EXTENTS _IO('f', 18) 616 617#if defined(__KERNEL__) && defined(CONFIG_COMPAT) 618/* 619 * ioctl commands in 32 bit emulation 620 */ 621#define EXT4_IOC32_GETFLAGS FS_IOC32_GETFLAGS 622#define EXT4_IOC32_SETFLAGS FS_IOC32_SETFLAGS 623#define EXT4_IOC32_GETVERSION _IOR('f', 3, int) 624#define EXT4_IOC32_SETVERSION _IOW('f', 4, int) 625#define EXT4_IOC32_GETRSVSZ _IOR('f', 5, int) 626#define EXT4_IOC32_SETRSVSZ _IOW('f', 6, int) 627#define EXT4_IOC32_GROUP_EXTEND _IOW('f', 7, unsigned int) 628#define EXT4_IOC32_GROUP_ADD _IOW('f', 8, struct compat_ext4_new_group_input) 629#define EXT4_IOC32_GETVERSION_OLD FS_IOC32_GETVERSION 630#define EXT4_IOC32_SETVERSION_OLD FS_IOC32_SETVERSION 631#endif 632 633/* Max physical block we can address w/o extents */ 634#define EXT4_MAX_BLOCK_FILE_PHYS 0xFFFFFFFF 635 636/* 637 * Structure of an inode on the disk 638 */ 639struct ext4_inode { 640 __le16 i_mode; /* File mode */ 641 __le16 i_uid; /* Low 16 bits of Owner Uid */ 642 __le32 i_size_lo; /* Size in bytes */ 643 __le32 i_atime; /* Access time */ 644 __le32 i_ctime; /* Inode Change time */ 645 __le32 i_mtime; /* Modification time */ 646 __le32 i_dtime; /* Deletion Time */ 647 __le16 i_gid; /* Low 16 bits of Group Id */ 648 __le16 i_links_count; /* Links count */ 649 __le32 i_blocks_lo; /* Blocks count */ 650 __le32 i_flags; /* File flags */ 651 union { 652 struct { 653 __le32 l_i_version; 654 } linux1; 655 struct { 656 __u32 h_i_translator; 657 } hurd1; 658 struct { 659 __u32 m_i_reserved1; 660 } masix1; 661 } osd1; /* OS dependent 1 */ 662 __le32 i_block[EXT4_N_BLOCKS];/* Pointers to blocks */ 663 __le32 i_generation; /* File version (for NFS) */ 664 __le32 i_file_acl_lo; /* File ACL */ 665 __le32 i_size_high; 666 __le32 i_obso_faddr; /* Obsoleted fragment address */ 667 union { 668 struct { 669 __le16 l_i_blocks_high; /* were l_i_reserved1 */ 670 __le16 l_i_file_acl_high; 671 __le16 l_i_uid_high; /* these 2 fields */ 672 __le16 l_i_gid_high; /* were reserved2[0] */ 673 __le16 l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */ 674 __le16 l_i_reserved; 675 } linux2; 676 struct { 677 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */ 678 __u16 h_i_mode_high; 679 __u16 h_i_uid_high; 680 __u16 h_i_gid_high; 681 __u32 h_i_author; 682 } hurd2; 683 struct { 684 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */ 685 __le16 m_i_file_acl_high; 686 __u32 m_i_reserved2[2]; 687 } masix2; 688 } osd2; /* OS dependent 2 */ 689 __le16 i_extra_isize; 690 __le16 i_checksum_hi; /* crc32c(uuid+inum+inode) BE */ 691 __le32 i_ctime_extra; /* extra Change time (nsec << 2 | epoch) */ 692 __le32 i_mtime_extra; /* extra Modification time(nsec << 2 | epoch) */ 693 __le32 i_atime_extra; /* extra Access time (nsec << 2 | epoch) */ 694 __le32 i_crtime; /* File Creation time */ 695 __le32 i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */ 696 __le32 i_version_hi; /* high 32 bits for 64-bit version */ 697}; 698 699struct move_extent { 700 __u32 reserved; /* should be zero */ 701 __u32 donor_fd; /* donor file descriptor */ 702 __u64 orig_start; /* logical start offset in block for orig */ 703 __u64 donor_start; /* logical start offset in block for donor */ 704 __u64 len; /* block length to be moved */ 705 __u64 moved_len; /* moved block length */ 706}; 707 708#define EXT4_EPOCH_BITS 2 709#define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1) 710#define EXT4_NSEC_MASK (~0UL << EXT4_EPOCH_BITS) 711 712/* 713 * Extended fields will fit into an inode if the filesystem was formatted 714 * with large inodes (-I 256 or larger) and there are not currently any EAs 715 * consuming all of the available space. For new inodes we always reserve 716 * enough space for the kernel's known extended fields, but for inodes 717 * created with an old kernel this might not have been the case. None of 718 * the extended inode fields is critical for correct filesystem operation. 719 * This macro checks if a certain field fits in the inode. Note that 720 * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize 721 */ 722#define EXT4_FITS_IN_INODE(ext4_inode, einode, field) \ 723 ((offsetof(typeof(*ext4_inode), field) + \ 724 sizeof((ext4_inode)->field)) \ 725 <= (EXT4_GOOD_OLD_INODE_SIZE + \ 726 (einode)->i_extra_isize)) \ 727 728static inline __le32 ext4_encode_extra_time(struct timespec *time) 729{ 730 return cpu_to_le32((sizeof(time->tv_sec) > 4 ? 731 (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) | 732 ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK)); 733} 734 735static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra) 736{ 737 if (sizeof(time->tv_sec) > 4) 738 time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK) 739 << 32; 740 time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS; 741} 742 743#define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode) \ 744do { \ 745 (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec); \ 746 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \ 747 (raw_inode)->xtime ## _extra = \ 748 ext4_encode_extra_time(&(inode)->xtime); \ 749} while (0) 750 751#define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode) \ 752do { \ 753 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \ 754 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec); \ 755 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \ 756 (raw_inode)->xtime ## _extra = \ 757 ext4_encode_extra_time(&(einode)->xtime); \ 758} while (0) 759 760#define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode) \ 761do { \ 762 (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime); \ 763 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \ 764 ext4_decode_extra_time(&(inode)->xtime, \ 765 raw_inode->xtime ## _extra); \ 766 else \ 767 (inode)->xtime.tv_nsec = 0; \ 768} while (0) 769 770#define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode) \ 771do { \ 772 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \ 773 (einode)->xtime.tv_sec = \ 774 (signed)le32_to_cpu((raw_inode)->xtime); \ 775 else \ 776 (einode)->xtime.tv_sec = 0; \ 777 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \ 778 ext4_decode_extra_time(&(einode)->xtime, \ 779 raw_inode->xtime ## _extra); \ 780 else \ 781 (einode)->xtime.tv_nsec = 0; \ 782} while (0) 783 784#define i_disk_version osd1.linux1.l_i_version 785 786#if defined(__KERNEL__) || defined(__linux__) 787#define i_reserved1 osd1.linux1.l_i_reserved1 788#define i_file_acl_high osd2.linux2.l_i_file_acl_high 789#define i_blocks_high osd2.linux2.l_i_blocks_high 790#define i_uid_low i_uid 791#define i_gid_low i_gid 792#define i_uid_high osd2.linux2.l_i_uid_high 793#define i_gid_high osd2.linux2.l_i_gid_high 794#define i_checksum_lo osd2.linux2.l_i_checksum_lo 795 796#elif defined(__GNU__) 797 798#define i_translator osd1.hurd1.h_i_translator 799#define i_uid_high osd2.hurd2.h_i_uid_high 800#define i_gid_high osd2.hurd2.h_i_gid_high 801#define i_author osd2.hurd2.h_i_author 802 803#elif defined(__masix__) 804 805#define i_reserved1 osd1.masix1.m_i_reserved1 806#define i_file_acl_high osd2.masix2.m_i_file_acl_high 807#define i_reserved2 osd2.masix2.m_i_reserved2 808 809#endif /* defined(__KERNEL__) || defined(__linux__) */ 810 811#include "extents_status.h" 812 813/* 814 * fourth extended file system inode data in memory 815 */ 816struct ext4_inode_info { 817 __le32 i_data[15]; /* unconverted */ 818 __u32 i_dtime; 819 ext4_fsblk_t i_file_acl; 820 821 /* 822 * i_block_group is the number of the block group which contains 823 * this file's inode. Constant across the lifetime of the inode, 824 * it is ued for making block allocation decisions - we try to 825 * place a file's data blocks near its inode block, and new inodes 826 * near to their parent directory's inode. 827 */ 828 ext4_group_t i_block_group; 829 ext4_lblk_t i_dir_start_lookup; 830#if (BITS_PER_LONG < 64) 831 unsigned long i_state_flags; /* Dynamic state flags */ 832#endif 833 unsigned long i_flags; 834 835 /* 836 * Extended attributes can be read independently of the main file 837 * data. Taking i_mutex even when reading would cause contention 838 * between readers of EAs and writers of regular file data, so 839 * instead we synchronize on xattr_sem when reading or changing 840 * EAs. 841 */ 842 struct rw_semaphore xattr_sem; 843 844 struct list_head i_orphan; /* unlinked but open inodes */ 845 846 /* 847 * i_disksize keeps track of what the inode size is ON DISK, not 848 * in memory. During truncate, i_size is set to the new size by 849 * the VFS prior to calling ext4_truncate(), but the filesystem won't 850 * set i_disksize to 0 until the truncate is actually under way. 851 * 852 * The intent is that i_disksize always represents the blocks which 853 * are used by this file. This allows recovery to restart truncate 854 * on orphans if we crash during truncate. We actually write i_disksize 855 * into the on-disk inode when writing inodes out, instead of i_size. 856 * 857 * The only time when i_disksize and i_size may be different is when 858 * a truncate is in progress. The only things which change i_disksize 859 * are ext4_get_block (growth) and ext4_truncate (shrinkth). 860 */ 861 loff_t i_disksize; 862 863 /* 864 * i_data_sem is for serialising ext4_truncate() against 865 * ext4_getblock(). In the 2.4 ext2 design, great chunks of inode's 866 * data tree are chopped off during truncate. We can't do that in 867 * ext4 because whenever we perform intermediate commits during 868 * truncate, the inode and all the metadata blocks *must* be in a 869 * consistent state which allows truncation of the orphans to restart 870 * during recovery. Hence we must fix the get_block-vs-truncate race 871 * by other means, so we have i_data_sem. 872 */ 873 struct rw_semaphore i_data_sem; 874 struct inode vfs_inode; 875 struct jbd2_inode *jinode; 876 877 spinlock_t i_raw_lock; /* protects updates to the raw inode */ 878 879 /* 880 * File creation time. Its function is same as that of 881 * struct timespec i_{a,c,m}time in the generic inode. 882 */ 883 struct timespec i_crtime; 884 885 /* mballoc */ 886 struct list_head i_prealloc_list; 887 spinlock_t i_prealloc_lock; 888 889 /* extents status tree */ 890 struct ext4_es_tree i_es_tree; 891 rwlock_t i_es_lock; 892 struct list_head i_es_lru; 893 unsigned int i_es_all_nr; /* protected by i_es_lock */ 894 unsigned int i_es_lru_nr; /* protected by i_es_lock */ 895 unsigned long i_touch_when; /* jiffies of last accessing */ 896 897 /* ialloc */ 898 ext4_group_t i_last_alloc_group; 899 900 /* allocation reservation info for delalloc */ 901 /* In case of bigalloc, these refer to clusters rather than blocks */ 902 unsigned int i_reserved_data_blocks; 903 unsigned int i_reserved_meta_blocks; 904 unsigned int i_allocated_meta_blocks; 905 ext4_lblk_t i_da_metadata_calc_last_lblock; 906 int i_da_metadata_calc_len; 907 908 /* on-disk additional length */ 909 __u16 i_extra_isize; 910 911 /* Indicate the inline data space. */ 912 u16 i_inline_off; 913 u16 i_inline_size; 914 915#ifdef CONFIG_QUOTA 916 /* quota space reservation, managed internally by quota code */ 917 qsize_t i_reserved_quota; 918#endif 919 920 /* Lock protecting lists below */ 921 spinlock_t i_completed_io_lock; 922 /* 923 * Completed IOs that need unwritten extents handling and have 924 * transaction reserved 925 */ 926 struct list_head i_rsv_conversion_list; 927 /* 928 * Completed IOs that need unwritten extents handling and don't have 929 * transaction reserved 930 */ 931 atomic_t i_ioend_count; /* Number of outstanding io_end structs */ 932 atomic_t i_unwritten; /* Nr. of inflight conversions pending */ 933 struct work_struct i_rsv_conversion_work; 934 935 spinlock_t i_block_reservation_lock; 936 937 /* 938 * Transactions that contain inode's metadata needed to complete 939 * fsync and fdatasync, respectively. 940 */ 941 tid_t i_sync_tid; 942 tid_t i_datasync_tid; 943 944 /* Precomputed uuid+inum+igen checksum for seeding inode checksums */ 945 __u32 i_csum_seed; 946}; 947 948/* 949 * File system states 950 */ 951#define EXT4_VALID_FS 0x0001 /* Unmounted cleanly */ 952#define EXT4_ERROR_FS 0x0002 /* Errors detected */ 953#define EXT4_ORPHAN_FS 0x0004 /* Orphans being recovered */ 954 955/* 956 * Misc. filesystem flags 957 */ 958#define EXT2_FLAGS_SIGNED_HASH 0x0001 /* Signed dirhash in use */ 959#define EXT2_FLAGS_UNSIGNED_HASH 0x0002 /* Unsigned dirhash in use */ 960#define EXT2_FLAGS_TEST_FILESYS 0x0004 /* to test development code */ 961 962/* 963 * Mount flags set via mount options or defaults 964 */ 965#define EXT4_MOUNT_GRPID 0x00004 /* Create files with directory's group */ 966#define EXT4_MOUNT_DEBUG 0x00008 /* Some debugging messages */ 967#define EXT4_MOUNT_ERRORS_CONT 0x00010 /* Continue on errors */ 968#define EXT4_MOUNT_ERRORS_RO 0x00020 /* Remount fs ro on errors */ 969#define EXT4_MOUNT_ERRORS_PANIC 0x00040 /* Panic on errors */ 970#define EXT4_MOUNT_ERRORS_MASK 0x00070 971#define EXT4_MOUNT_MINIX_DF 0x00080 /* Mimics the Minix statfs */ 972#define EXT4_MOUNT_NOLOAD 0x00100 /* Don't use existing journal*/ 973#define EXT4_MOUNT_DATA_FLAGS 0x00C00 /* Mode for data writes: */ 974#define EXT4_MOUNT_JOURNAL_DATA 0x00400 /* Write data to journal */ 975#define EXT4_MOUNT_ORDERED_DATA 0x00800 /* Flush data before commit */ 976#define EXT4_MOUNT_WRITEBACK_DATA 0x00C00 /* No data ordering */ 977#define EXT4_MOUNT_UPDATE_JOURNAL 0x01000 /* Update the journal format */ 978#define EXT4_MOUNT_NO_UID32 0x02000 /* Disable 32-bit UIDs */ 979#define EXT4_MOUNT_XATTR_USER 0x04000 /* Extended user attributes */ 980#define EXT4_MOUNT_POSIX_ACL 0x08000 /* POSIX Access Control Lists */ 981#define EXT4_MOUNT_NO_AUTO_DA_ALLOC 0x10000 /* No auto delalloc mapping */ 982#define EXT4_MOUNT_BARRIER 0x20000 /* Use block barriers */ 983#define EXT4_MOUNT_QUOTA 0x80000 /* Some quota option set */ 984#define EXT4_MOUNT_USRQUOTA 0x100000 /* "old" user quota */ 985#define EXT4_MOUNT_GRPQUOTA 0x200000 /* "old" group quota */ 986#define EXT4_MOUNT_DIOREAD_NOLOCK 0x400000 /* Enable support for dio read nolocking */ 987#define EXT4_MOUNT_JOURNAL_CHECKSUM 0x800000 /* Journal checksums */ 988#define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */ 989#define EXT4_MOUNT_DELALLOC 0x8000000 /* Delalloc support */ 990#define EXT4_MOUNT_DATA_ERR_ABORT 0x10000000 /* Abort on file data write */ 991#define EXT4_MOUNT_BLOCK_VALIDITY 0x20000000 /* Block validity checking */ 992#define EXT4_MOUNT_DISCARD 0x40000000 /* Issue DISCARD requests */ 993#define EXT4_MOUNT_INIT_INODE_TABLE 0x80000000 /* Initialize uninitialized itables */ 994 995/* 996 * Mount flags set either automatically (could not be set by mount option) 997 * based on per file system feature or property or in special cases such as 998 * distinguishing between explicit mount option definition and default. 999 */ 1000#define EXT4_MOUNT2_EXPLICIT_DELALLOC 0x00000001 /* User explicitly 1001 specified delalloc */ 1002#define EXT4_MOUNT2_STD_GROUP_SIZE 0x00000002 /* We have standard group 1003 size of blocksize * 8 1004 blocks */ 1005#define EXT4_MOUNT2_HURD_COMPAT 0x00000004 /* Support HURD-castrated 1006 file systems */ 1007 1008#define clear_opt(sb, opt) EXT4_SB(sb)->s_mount_opt &= \ 1009 ~EXT4_MOUNT_##opt 1010#define set_opt(sb, opt) EXT4_SB(sb)->s_mount_opt |= \ 1011 EXT4_MOUNT_##opt 1012#define test_opt(sb, opt) (EXT4_SB(sb)->s_mount_opt & \ 1013 EXT4_MOUNT_##opt) 1014 1015#define clear_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 &= \ 1016 ~EXT4_MOUNT2_##opt 1017#define set_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 |= \ 1018 EXT4_MOUNT2_##opt 1019#define test_opt2(sb, opt) (EXT4_SB(sb)->s_mount_opt2 & \ 1020 EXT4_MOUNT2_##opt) 1021 1022#define ext4_test_and_set_bit __test_and_set_bit_le 1023#define ext4_set_bit __set_bit_le 1024#define ext4_set_bit_atomic ext2_set_bit_atomic 1025#define ext4_test_and_clear_bit __test_and_clear_bit_le 1026#define ext4_clear_bit __clear_bit_le 1027#define ext4_clear_bit_atomic ext2_clear_bit_atomic 1028#define ext4_test_bit test_bit_le 1029#define ext4_find_next_zero_bit find_next_zero_bit_le 1030#define ext4_find_next_bit find_next_bit_le 1031 1032extern void ext4_set_bits(void *bm, int cur, int len); 1033 1034/* 1035 * Maximal mount counts between two filesystem checks 1036 */ 1037#define EXT4_DFL_MAX_MNT_COUNT 20 /* Allow 20 mounts */ 1038#define EXT4_DFL_CHECKINTERVAL 0 /* Don't use interval check */ 1039 1040/* 1041 * Behaviour when detecting errors 1042 */ 1043#define EXT4_ERRORS_CONTINUE 1 /* Continue execution */ 1044#define EXT4_ERRORS_RO 2 /* Remount fs read-only */ 1045#define EXT4_ERRORS_PANIC 3 /* Panic */ 1046#define EXT4_ERRORS_DEFAULT EXT4_ERRORS_CONTINUE 1047 1048/* Metadata checksum algorithm codes */ 1049#define EXT4_CRC32C_CHKSUM 1 1050 1051/* 1052 * Structure of the super block 1053 */ 1054struct ext4_super_block { 1055/*00*/ __le32 s_inodes_count; /* Inodes count */ 1056 __le32 s_blocks_count_lo; /* Blocks count */ 1057 __le32 s_r_blocks_count_lo; /* Reserved blocks count */ 1058 __le32 s_free_blocks_count_lo; /* Free blocks count */ 1059/*10*/ __le32 s_free_inodes_count; /* Free inodes count */ 1060 __le32 s_first_data_block; /* First Data Block */ 1061 __le32 s_log_block_size; /* Block size */ 1062 __le32 s_log_cluster_size; /* Allocation cluster size */ 1063/*20*/ __le32 s_blocks_per_group; /* # Blocks per group */ 1064 __le32 s_clusters_per_group; /* # Clusters per group */ 1065 __le32 s_inodes_per_group; /* # Inodes per group */ 1066 __le32 s_mtime; /* Mount time */ 1067/*30*/ __le32 s_wtime; /* Write time */ 1068 __le16 s_mnt_count; /* Mount count */ 1069 __le16 s_max_mnt_count; /* Maximal mount count */ 1070 __le16 s_magic; /* Magic signature */ 1071 __le16 s_state; /* File system state */ 1072 __le16 s_errors; /* Behaviour when detecting errors */ 1073 __le16 s_minor_rev_level; /* minor revision level */ 1074/*40*/ __le32 s_lastcheck; /* time of last check */ 1075 __le32 s_checkinterval; /* max. time between checks */ 1076 __le32 s_creator_os; /* OS */ 1077 __le32 s_rev_level; /* Revision level */ 1078/*50*/ __le16 s_def_resuid; /* Default uid for reserved blocks */ 1079 __le16 s_def_resgid; /* Default gid for reserved blocks */ 1080 /* 1081 * These fields are for EXT4_DYNAMIC_REV superblocks only. 1082 * 1083 * Note: the difference between the compatible feature set and 1084 * the incompatible feature set is that if there is a bit set 1085 * in the incompatible feature set that the kernel doesn't 1086 * know about, it should refuse to mount the filesystem. 1087 * 1088 * e2fsck's requirements are more strict; if it doesn't know 1089 * about a feature in either the compatible or incompatible 1090 * feature set, it must abort and not try to meddle with 1091 * things it doesn't understand... 1092 */ 1093 __le32 s_first_ino; /* First non-reserved inode */ 1094 __le16 s_inode_size; /* size of inode structure */ 1095 __le16 s_block_group_nr; /* block group # of this superblock */ 1096 __le32 s_feature_compat; /* compatible feature set */ 1097/*60*/ __le32 s_feature_incompat; /* incompatible feature set */ 1098 __le32 s_feature_ro_compat; /* readonly-compatible feature set */ 1099/*68*/ __u8 s_uuid[16]; /* 128-bit uuid for volume */ 1100/*78*/ char s_volume_name[16]; /* volume name */ 1101/*88*/ char s_last_mounted[64]; /* directory where last mounted */ 1102/*C8*/ __le32 s_algorithm_usage_bitmap; /* For compression */ 1103 /* 1104 * Performance hints. Directory preallocation should only 1105 * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on. 1106 */ 1107 __u8 s_prealloc_blocks; /* Nr of blocks to try to preallocate*/ 1108 __u8 s_prealloc_dir_blocks; /* Nr to preallocate for dirs */ 1109 __le16 s_reserved_gdt_blocks; /* Per group desc for online growth */ 1110 /* 1111 * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set. 1112 */ 1113/*D0*/ __u8 s_journal_uuid[16]; /* uuid of journal superblock */ 1114/*E0*/ __le32 s_journal_inum; /* inode number of journal file */ 1115 __le32 s_journal_dev; /* device number of journal file */ 1116 __le32 s_last_orphan; /* start of list of inodes to delete */ 1117 __le32 s_hash_seed[4]; /* HTREE hash seed */ 1118 __u8 s_def_hash_version; /* Default hash version to use */ 1119 __u8 s_jnl_backup_type; 1120 __le16 s_desc_size; /* size of group descriptor */ 1121/*100*/ __le32 s_default_mount_opts; 1122 __le32 s_first_meta_bg; /* First metablock block group */ 1123 __le32 s_mkfs_time; /* When the filesystem was created */ 1124 __le32 s_jnl_blocks[17]; /* Backup of the journal inode */ 1125 /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */ 1126/*150*/ __le32 s_blocks_count_hi; /* Blocks count */ 1127 __le32 s_r_blocks_count_hi; /* Reserved blocks count */ 1128 __le32 s_free_blocks_count_hi; /* Free blocks count */ 1129 __le16 s_min_extra_isize; /* All inodes have at least # bytes */ 1130 __le16 s_want_extra_isize; /* New inodes should reserve # bytes */ 1131 __le32 s_flags; /* Miscellaneous flags */ 1132 __le16 s_raid_stride; /* RAID stride */ 1133 __le16 s_mmp_update_interval; /* # seconds to wait in MMP checking */ 1134 __le64 s_mmp_block; /* Block for multi-mount protection */ 1135 __le32 s_raid_stripe_width; /* blocks on all data disks (N*stride)*/ 1136 __u8 s_log_groups_per_flex; /* FLEX_BG group size */ 1137 __u8 s_checksum_type; /* metadata checksum algorithm used */ 1138 __le16 s_reserved_pad; 1139 __le64 s_kbytes_written; /* nr of lifetime kilobytes written */ 1140 __le32 s_snapshot_inum; /* Inode number of active snapshot */ 1141 __le32 s_snapshot_id; /* sequential ID of active snapshot */ 1142 __le64 s_snapshot_r_blocks_count; /* reserved blocks for active 1143 snapshot's future use */ 1144 __le32 s_snapshot_list; /* inode number of the head of the 1145 on-disk snapshot list */ 1146#define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count) 1147 __le32 s_error_count; /* number of fs errors */ 1148 __le32 s_first_error_time; /* first time an error happened */ 1149 __le32 s_first_error_ino; /* inode involved in first error */ 1150 __le64 s_first_error_block; /* block involved of first error */ 1151 __u8 s_first_error_func[32]; /* function where the error happened */ 1152 __le32 s_first_error_line; /* line number where error happened */ 1153 __le32 s_last_error_time; /* most recent time of an error */ 1154 __le32 s_last_error_ino; /* inode involved in last error */ 1155 __le32 s_last_error_line; /* line number where error happened */ 1156 __le64 s_last_error_block; /* block involved of last error */ 1157 __u8 s_last_error_func[32]; /* function where the error happened */ 1158#define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts) 1159 __u8 s_mount_opts[64]; 1160 __le32 s_usr_quota_inum; /* inode for tracking user quota */ 1161 __le32 s_grp_quota_inum; /* inode for tracking group quota */ 1162 __le32 s_overhead_clusters; /* overhead blocks/clusters in fs */ 1163 __le32 s_backup_bgs[2]; /* groups with sparse_super2 SBs */ 1164 __le32 s_reserved[106]; /* Padding to the end of the block */ 1165 __le32 s_checksum; /* crc32c(superblock) */ 1166}; 1167 1168#define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START) 1169 1170#ifdef __KERNEL__ 1171 1172/* 1173 * run-time mount flags 1174 */ 1175#define EXT4_MF_MNTDIR_SAMPLED 0x0001 1176#define EXT4_MF_FS_ABORTED 0x0002 /* Fatal error detected */ 1177 1178/* Number of quota types we support */ 1179#define EXT4_MAXQUOTAS 2 1180 1181/* 1182 * fourth extended-fs super-block data in memory 1183 */ 1184struct ext4_sb_info { 1185 unsigned long s_desc_size; /* Size of a group descriptor in bytes */ 1186 unsigned long s_inodes_per_block;/* Number of inodes per block */ 1187 unsigned long s_blocks_per_group;/* Number of blocks in a group */ 1188 unsigned long s_clusters_per_group; /* Number of clusters in a group */ 1189 unsigned long s_inodes_per_group;/* Number of inodes in a group */ 1190 unsigned long s_itb_per_group; /* Number of inode table blocks per group */ 1191 unsigned long s_gdb_count; /* Number of group descriptor blocks */ 1192 unsigned long s_desc_per_block; /* Number of group descriptors per block */ 1193 ext4_group_t s_groups_count; /* Number of groups in the fs */ 1194 ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */ 1195 unsigned long s_overhead; /* # of fs overhead clusters */ 1196 unsigned int s_cluster_ratio; /* Number of blocks per cluster */ 1197 unsigned int s_cluster_bits; /* log2 of s_cluster_ratio */ 1198 loff_t s_bitmap_maxbytes; /* max bytes for bitmap files */ 1199 struct buffer_head * s_sbh; /* Buffer containing the super block */ 1200 struct ext4_super_block *s_es; /* Pointer to the super block in the buffer */ 1201 struct buffer_head **s_group_desc; 1202 unsigned int s_mount_opt; 1203 unsigned int s_mount_opt2; 1204 unsigned int s_mount_flags; 1205 unsigned int s_def_mount_opt; 1206 ext4_fsblk_t s_sb_block; 1207 atomic64_t s_resv_clusters; 1208 kuid_t s_resuid; 1209 kgid_t s_resgid; 1210 unsigned short s_mount_state; 1211 unsigned short s_pad; 1212 int s_addr_per_block_bits; 1213 int s_desc_per_block_bits; 1214 int s_inode_size; 1215 int s_first_ino; 1216 unsigned int s_inode_readahead_blks; 1217 unsigned int s_inode_goal; 1218 spinlock_t s_next_gen_lock; 1219 u32 s_next_generation; 1220 u32 s_hash_seed[4]; 1221 int s_def_hash_version; 1222 int s_hash_unsigned; /* 3 if hash should be signed, 0 if not */ 1223 struct percpu_counter s_freeclusters_counter; 1224 struct percpu_counter s_freeinodes_counter; 1225 struct percpu_counter s_dirs_counter; 1226 struct percpu_counter s_dirtyclusters_counter; 1227 struct blockgroup_lock *s_blockgroup_lock; 1228 struct proc_dir_entry *s_proc; 1229 struct kobject s_kobj; 1230 struct completion s_kobj_unregister; 1231 struct super_block *s_sb; 1232 1233 /* Journaling */ 1234 struct journal_s *s_journal; 1235 struct list_head s_orphan; 1236 struct mutex s_orphan_lock; 1237 unsigned long s_resize_flags; /* Flags indicating if there 1238 is a resizer */ 1239 unsigned long s_commit_interval; 1240 u32 s_max_batch_time; 1241 u32 s_min_batch_time; 1242 struct block_device *journal_bdev; 1243#ifdef CONFIG_QUOTA 1244 char *s_qf_names[EXT4_MAXQUOTAS]; /* Names of quota files with journalled quota */ 1245 int s_jquota_fmt; /* Format of quota to use */ 1246#endif 1247 unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */ 1248 struct rb_root system_blks; 1249 1250#ifdef EXTENTS_STATS 1251 /* ext4 extents stats */ 1252 unsigned long s_ext_min; 1253 unsigned long s_ext_max; 1254 unsigned long s_depth_max; 1255 spinlock_t s_ext_stats_lock; 1256 unsigned long s_ext_blocks; 1257 unsigned long s_ext_extents; 1258#endif 1259 1260 /* for buddy allocator */ 1261 struct ext4_group_info ***s_group_info; 1262 struct inode *s_buddy_cache; 1263 spinlock_t s_md_lock; 1264 unsigned short *s_mb_offsets; 1265 unsigned int *s_mb_maxs; 1266 unsigned int s_group_info_size; 1267 1268 /* tunables */ 1269 unsigned long s_stripe; 1270 unsigned int s_mb_stream_request; 1271 unsigned int s_mb_max_to_scan; 1272 unsigned int s_mb_min_to_scan; 1273 unsigned int s_mb_stats; 1274 unsigned int s_mb_order2_reqs; 1275 unsigned int s_mb_group_prealloc; 1276 unsigned int s_max_dir_size_kb; 1277 /* where last allocation was done - for stream allocation */ 1278 unsigned long s_mb_last_group; 1279 unsigned long s_mb_last_start; 1280 1281 /* stats for buddy allocator */ 1282 atomic_t s_bal_reqs; /* number of reqs with len > 1 */ 1283 atomic_t s_bal_success; /* we found long enough chunks */ 1284 atomic_t s_bal_allocated; /* in blocks */ 1285 atomic_t s_bal_ex_scanned; /* total extents scanned */ 1286 atomic_t s_bal_goals; /* goal hits */ 1287 atomic_t s_bal_breaks; /* too long searches */ 1288 atomic_t s_bal_2orders; /* 2^order hits */ 1289 spinlock_t s_bal_lock; 1290 unsigned long s_mb_buddies_generated; 1291 unsigned long long s_mb_generation_time; 1292 atomic_t s_mb_lost_chunks; 1293 atomic_t s_mb_preallocated; 1294 atomic_t s_mb_discarded; 1295 atomic_t s_lock_busy; 1296 1297 /* locality groups */ 1298 struct ext4_locality_group __percpu *s_locality_groups; 1299 1300 /* for write statistics */ 1301 unsigned long s_sectors_written_start; 1302 u64 s_kbytes_written; 1303 1304 /* the size of zero-out chunk */ 1305 unsigned int s_extent_max_zeroout_kb; 1306 1307 unsigned int s_log_groups_per_flex; 1308 struct flex_groups *s_flex_groups; 1309 ext4_group_t s_flex_groups_allocated; 1310 1311 /* workqueue for reserved extent conversions (buffered io) */ 1312 struct workqueue_struct *rsv_conversion_wq; 1313 1314 /* timer for periodic error stats printing */ 1315 struct timer_list s_err_report; 1316 1317 /* Lazy inode table initialization info */ 1318 struct ext4_li_request *s_li_request; 1319 /* Wait multiplier for lazy initialization thread */ 1320 unsigned int s_li_wait_mult; 1321 1322 /* Kernel thread for multiple mount protection */ 1323 struct task_struct *s_mmp_tsk; 1324 1325 /* record the last minlen when FITRIM is called. */ 1326 atomic_t s_last_trim_minblks; 1327 1328 /* Reference to checksum algorithm driver via cryptoapi */ 1329 struct crypto_shash *s_chksum_driver; 1330 1331 /* Precomputed FS UUID checksum for seeding other checksums */ 1332 __u32 s_csum_seed; 1333 1334 /* Reclaim extents from extent status tree */ 1335 struct shrinker s_es_shrinker; 1336 struct list_head s_es_lru; 1337 struct ext4_es_stats s_es_stats; 1338 struct mb_cache *s_mb_cache; 1339 spinlock_t s_es_lru_lock ____cacheline_aligned_in_smp; 1340 1341 /* Ratelimit ext4 messages. */ 1342 struct ratelimit_state s_err_ratelimit_state; 1343 struct ratelimit_state s_warning_ratelimit_state; 1344 struct ratelimit_state s_msg_ratelimit_state; 1345}; 1346 1347static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb) 1348{ 1349 return sb->s_fs_info; 1350} 1351static inline struct ext4_inode_info *EXT4_I(struct inode *inode) 1352{ 1353 return container_of(inode, struct ext4_inode_info, vfs_inode); 1354} 1355 1356static inline struct timespec ext4_current_time(struct inode *inode) 1357{ 1358 return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ? 1359 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC; 1360} 1361 1362static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino) 1363{ 1364 return ino == EXT4_ROOT_INO || 1365 ino == EXT4_USR_QUOTA_INO || 1366 ino == EXT4_GRP_QUOTA_INO || 1367 ino == EXT4_BOOT_LOADER_INO || 1368 ino == EXT4_JOURNAL_INO || 1369 ino == EXT4_RESIZE_INO || 1370 (ino >= EXT4_FIRST_INO(sb) && 1371 ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)); 1372} 1373 1374static inline void ext4_set_io_unwritten_flag(struct inode *inode, 1375 struct ext4_io_end *io_end) 1376{ 1377 if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) { 1378 io_end->flag |= EXT4_IO_END_UNWRITTEN; 1379 atomic_inc(&EXT4_I(inode)->i_unwritten); 1380 } 1381} 1382 1383static inline ext4_io_end_t *ext4_inode_aio(struct inode *inode) 1384{ 1385 return inode->i_private; 1386} 1387 1388static inline void ext4_inode_aio_set(struct inode *inode, ext4_io_end_t *io) 1389{ 1390 inode->i_private = io; 1391} 1392 1393/* 1394 * Inode dynamic state flags 1395 */ 1396enum { 1397 EXT4_STATE_JDATA, /* journaled data exists */ 1398 EXT4_STATE_NEW, /* inode is newly created */ 1399 EXT4_STATE_XATTR, /* has in-inode xattrs */ 1400 EXT4_STATE_NO_EXPAND, /* No space for expansion */ 1401 EXT4_STATE_DA_ALLOC_CLOSE, /* Alloc DA blks on close */ 1402 EXT4_STATE_EXT_MIGRATE, /* Inode is migrating */ 1403 EXT4_STATE_DIO_UNWRITTEN, /* need convert on dio done*/ 1404 EXT4_STATE_NEWENTRY, /* File just added to dir */ 1405 EXT4_STATE_DIOREAD_LOCK, /* Disable support for dio read 1406 nolocking */ 1407 EXT4_STATE_MAY_INLINE_DATA, /* may have in-inode data */ 1408 EXT4_STATE_ORDERED_MODE, /* data=ordered mode */ 1409 EXT4_STATE_EXT_PRECACHED, /* extents have been precached */ 1410}; 1411 1412#define EXT4_INODE_BIT_FNS(name, field, offset) \ 1413static inline int ext4_test_inode_##name(struct inode *inode, int bit) \ 1414{ \ 1415 return test_bit(bit + (offset), &EXT4_I(inode)->i_##field); \ 1416} \ 1417static inline void ext4_set_inode_##name(struct inode *inode, int bit) \ 1418{ \ 1419 set_bit(bit + (offset), &EXT4_I(inode)->i_##field); \ 1420} \ 1421static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \ 1422{ \ 1423 clear_bit(bit + (offset), &EXT4_I(inode)->i_##field); \ 1424} 1425 1426/* Add these declarations here only so that these functions can be 1427 * found by name. Otherwise, they are very hard to locate. */ 1428static inline int ext4_test_inode_flag(struct inode *inode, int bit); 1429static inline void ext4_set_inode_flag(struct inode *inode, int bit); 1430static inline void ext4_clear_inode_flag(struct inode *inode, int bit); 1431EXT4_INODE_BIT_FNS(flag, flags, 0) 1432 1433/* Add these declarations here only so that these functions can be 1434 * found by name. Otherwise, they are very hard to locate. */ 1435static inline int ext4_test_inode_state(struct inode *inode, int bit); 1436static inline void ext4_set_inode_state(struct inode *inode, int bit); 1437static inline void ext4_clear_inode_state(struct inode *inode, int bit); 1438#if (BITS_PER_LONG < 64) 1439EXT4_INODE_BIT_FNS(state, state_flags, 0) 1440 1441static inline void ext4_clear_state_flags(struct ext4_inode_info *ei) 1442{ 1443 (ei)->i_state_flags = 0; 1444} 1445#else 1446EXT4_INODE_BIT_FNS(state, flags, 32) 1447 1448static inline void ext4_clear_state_flags(struct ext4_inode_info *ei) 1449{ 1450 /* We depend on the fact that callers will set i_flags */ 1451} 1452#endif 1453#else 1454/* Assume that user mode programs are passing in an ext4fs superblock, not 1455 * a kernel struct super_block. This will allow us to call the feature-test 1456 * macros from user land. */ 1457#define EXT4_SB(sb) (sb) 1458#endif 1459 1460#define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime 1461 1462/* 1463 * Codes for operating systems 1464 */ 1465#define EXT4_OS_LINUX 0 1466#define EXT4_OS_HURD 1 1467#define EXT4_OS_MASIX 2 1468#define EXT4_OS_FREEBSD 3 1469#define EXT4_OS_LITES 4 1470 1471/* 1472 * Revision levels 1473 */ 1474#define EXT4_GOOD_OLD_REV 0 /* The good old (original) format */ 1475#define EXT4_DYNAMIC_REV 1 /* V2 format w/ dynamic inode sizes */ 1476 1477#define EXT4_CURRENT_REV EXT4_GOOD_OLD_REV 1478#define EXT4_MAX_SUPP_REV EXT4_DYNAMIC_REV 1479 1480#define EXT4_GOOD_OLD_INODE_SIZE 128 1481 1482/* 1483 * Feature set definitions 1484 */ 1485 1486#define EXT4_HAS_COMPAT_FEATURE(sb,mask) \ 1487 ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0) 1488#define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask) \ 1489 ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0) 1490#define EXT4_HAS_INCOMPAT_FEATURE(sb,mask) \ 1491 ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0) 1492#define EXT4_SET_COMPAT_FEATURE(sb,mask) \ 1493 EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask) 1494#define EXT4_SET_RO_COMPAT_FEATURE(sb,mask) \ 1495 EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask) 1496#define EXT4_SET_INCOMPAT_FEATURE(sb,mask) \ 1497 EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask) 1498#define EXT4_CLEAR_COMPAT_FEATURE(sb,mask) \ 1499 EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask) 1500#define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask) \ 1501 EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask) 1502#define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask) \ 1503 EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask) 1504 1505#define EXT4_FEATURE_COMPAT_DIR_PREALLOC 0x0001 1506#define EXT4_FEATURE_COMPAT_IMAGIC_INODES 0x0002 1507#define EXT4_FEATURE_COMPAT_HAS_JOURNAL 0x0004 1508#define EXT4_FEATURE_COMPAT_EXT_ATTR 0x0008 1509#define EXT4_FEATURE_COMPAT_RESIZE_INODE 0x0010 1510#define EXT4_FEATURE_COMPAT_DIR_INDEX 0x0020 1511#define EXT4_FEATURE_COMPAT_SPARSE_SUPER2 0x0200 1512 1513#define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER 0x0001 1514#define EXT4_FEATURE_RO_COMPAT_LARGE_FILE 0x0002 1515#define EXT4_FEATURE_RO_COMPAT_BTREE_DIR 0x0004 1516#define EXT4_FEATURE_RO_COMPAT_HUGE_FILE 0x0008 1517#define EXT4_FEATURE_RO_COMPAT_GDT_CSUM 0x0010 1518#define EXT4_FEATURE_RO_COMPAT_DIR_NLINK 0x0020 1519#define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE 0x0040 1520#define EXT4_FEATURE_RO_COMPAT_QUOTA 0x0100 1521#define EXT4_FEATURE_RO_COMPAT_BIGALLOC 0x0200 1522/* 1523 * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM). When 1524 * METADATA_CSUM is set, group descriptor checksums use the same algorithm as 1525 * all other data structures' checksums. However, the METADATA_CSUM and 1526 * GDT_CSUM bits are mutually exclusive. 1527 */ 1528#define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM 0x0400 1529 1530#define EXT4_FEATURE_INCOMPAT_COMPRESSION 0x0001 1531#define EXT4_FEATURE_INCOMPAT_FILETYPE 0x0002 1532#define EXT4_FEATURE_INCOMPAT_RECOVER 0x0004 /* Needs recovery */ 1533#define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 /* Journal device */ 1534#define EXT4_FEATURE_INCOMPAT_META_BG 0x0010 1535#define EXT4_FEATURE_INCOMPAT_EXTENTS 0x0040 /* extents support */ 1536#define EXT4_FEATURE_INCOMPAT_64BIT 0x0080 1537#define EXT4_FEATURE_INCOMPAT_MMP 0x0100 1538#define EXT4_FEATURE_INCOMPAT_FLEX_BG 0x0200 1539#define EXT4_FEATURE_INCOMPAT_EA_INODE 0x0400 /* EA in inode */ 1540#define EXT4_FEATURE_INCOMPAT_DIRDATA 0x1000 /* data in dirent */ 1541#define EXT4_FEATURE_INCOMPAT_BG_USE_META_CSUM 0x2000 /* use crc32c for bg */ 1542#define EXT4_FEATURE_INCOMPAT_LARGEDIR 0x4000 /* >2GB or 3-lvl htree */ 1543#define EXT4_FEATURE_INCOMPAT_INLINE_DATA 0x8000 /* data in inode */ 1544 1545#define EXT2_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR 1546#define EXT2_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \ 1547 EXT4_FEATURE_INCOMPAT_META_BG) 1548#define EXT2_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \ 1549 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \ 1550 EXT4_FEATURE_RO_COMPAT_BTREE_DIR) 1551 1552#define EXT3_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR 1553#define EXT3_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \ 1554 EXT4_FEATURE_INCOMPAT_RECOVER| \ 1555 EXT4_FEATURE_INCOMPAT_META_BG) 1556#define EXT3_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \ 1557 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \ 1558 EXT4_FEATURE_RO_COMPAT_BTREE_DIR) 1559 1560#define EXT4_FEATURE_COMPAT_SUPP EXT2_FEATURE_COMPAT_EXT_ATTR 1561#define EXT4_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \ 1562 EXT4_FEATURE_INCOMPAT_RECOVER| \ 1563 EXT4_FEATURE_INCOMPAT_META_BG| \ 1564 EXT4_FEATURE_INCOMPAT_EXTENTS| \ 1565 EXT4_FEATURE_INCOMPAT_64BIT| \ 1566 EXT4_FEATURE_INCOMPAT_FLEX_BG| \ 1567 EXT4_FEATURE_INCOMPAT_MMP | \ 1568 EXT4_FEATURE_INCOMPAT_INLINE_DATA) 1569#define EXT4_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \ 1570 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \ 1571 EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \ 1572 EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \ 1573 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \ 1574 EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\ 1575 EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\ 1576 EXT4_FEATURE_RO_COMPAT_BIGALLOC |\ 1577 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\ 1578 EXT4_FEATURE_RO_COMPAT_QUOTA) 1579 1580/* 1581 * Default values for user and/or group using reserved blocks 1582 */ 1583#define EXT4_DEF_RESUID 0 1584#define EXT4_DEF_RESGID 0 1585 1586#define EXT4_DEF_INODE_READAHEAD_BLKS 32 1587 1588/* 1589 * Default mount options 1590 */ 1591#define EXT4_DEFM_DEBUG 0x0001 1592#define EXT4_DEFM_BSDGROUPS 0x0002 1593#define EXT4_DEFM_XATTR_USER 0x0004 1594#define EXT4_DEFM_ACL 0x0008 1595#define EXT4_DEFM_UID16 0x0010 1596#define EXT4_DEFM_JMODE 0x0060 1597#define EXT4_DEFM_JMODE_DATA 0x0020 1598#define EXT4_DEFM_JMODE_ORDERED 0x0040 1599#define EXT4_DEFM_JMODE_WBACK 0x0060 1600#define EXT4_DEFM_NOBARRIER 0x0100 1601#define EXT4_DEFM_BLOCK_VALIDITY 0x0200 1602#define EXT4_DEFM_DISCARD 0x0400 1603#define EXT4_DEFM_NODELALLOC 0x0800 1604 1605/* 1606 * Default journal batch times 1607 */ 1608#define EXT4_DEF_MIN_BATCH_TIME 0 1609#define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */ 1610 1611/* 1612 * Minimum number of groups in a flexgroup before we separate out 1613 * directories into the first block group of a flexgroup 1614 */ 1615#define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4 1616 1617/* 1618 * Structure of a directory entry 1619 */ 1620#define EXT4_NAME_LEN 255 1621 1622struct ext4_dir_entry { 1623 __le32 inode; /* Inode number */ 1624 __le16 rec_len; /* Directory entry length */ 1625 __le16 name_len; /* Name length */ 1626 char name[EXT4_NAME_LEN]; /* File name */ 1627}; 1628 1629/* 1630 * The new version of the directory entry. Since EXT4 structures are 1631 * stored in intel byte order, and the name_len field could never be 1632 * bigger than 255 chars, it's safe to reclaim the extra byte for the 1633 * file_type field. 1634 */ 1635struct ext4_dir_entry_2 { 1636 __le32 inode; /* Inode number */ 1637 __le16 rec_len; /* Directory entry length */ 1638 __u8 name_len; /* Name length */ 1639 __u8 file_type; 1640 char name[EXT4_NAME_LEN]; /* File name */ 1641}; 1642 1643/* 1644 * This is a bogus directory entry at the end of each leaf block that 1645 * records checksums. 1646 */ 1647struct ext4_dir_entry_tail { 1648 __le32 det_reserved_zero1; /* Pretend to be unused */ 1649 __le16 det_rec_len; /* 12 */ 1650 __u8 det_reserved_zero2; /* Zero name length */ 1651 __u8 det_reserved_ft; /* 0xDE, fake file type */ 1652 __le32 det_checksum; /* crc32c(uuid+inum+dirblock) */ 1653}; 1654 1655#define EXT4_DIRENT_TAIL(block, blocksize) \ 1656 ((struct ext4_dir_entry_tail *)(((void *)(block)) + \ 1657 ((blocksize) - \ 1658 sizeof(struct ext4_dir_entry_tail)))) 1659 1660/* 1661 * Ext4 directory file types. Only the low 3 bits are used. The 1662 * other bits are reserved for now. 1663 */ 1664#define EXT4_FT_UNKNOWN 0 1665#define EXT4_FT_REG_FILE 1 1666#define EXT4_FT_DIR 2 1667#define EXT4_FT_CHRDEV 3 1668#define EXT4_FT_BLKDEV 4 1669#define EXT4_FT_FIFO 5 1670#define EXT4_FT_SOCK 6 1671#define EXT4_FT_SYMLINK 7 1672 1673#define EXT4_FT_MAX 8 1674 1675#define EXT4_FT_DIR_CSUM 0xDE 1676 1677/* 1678 * EXT4_DIR_PAD defines the directory entries boundaries 1679 * 1680 * NOTE: It must be a multiple of 4 1681 */ 1682#define EXT4_DIR_PAD 4 1683#define EXT4_DIR_ROUND (EXT4_DIR_PAD - 1) 1684#define EXT4_DIR_REC_LEN(name_len) (((name_len) + 8 + EXT4_DIR_ROUND) & \ 1685 ~EXT4_DIR_ROUND) 1686#define EXT4_MAX_REC_LEN ((1<<16)-1) 1687 1688/* 1689 * If we ever get support for fs block sizes > page_size, we'll need 1690 * to remove the #if statements in the next two functions... 1691 */ 1692static inline unsigned int 1693ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize) 1694{ 1695 unsigned len = le16_to_cpu(dlen); 1696 1697#if (PAGE_CACHE_SIZE >= 65536) 1698 if (len == EXT4_MAX_REC_LEN || len == 0) 1699 return blocksize; 1700 return (len & 65532) | ((len & 3) << 16); 1701#else 1702 return len; 1703#endif 1704} 1705 1706static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize) 1707{ 1708 if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3)) 1709 BUG(); 1710#if (PAGE_CACHE_SIZE >= 65536) 1711 if (len < 65536) 1712 return cpu_to_le16(len); 1713 if (len == blocksize) { 1714 if (blocksize == 65536) 1715 return cpu_to_le16(EXT4_MAX_REC_LEN); 1716 else 1717 return cpu_to_le16(0); 1718 } 1719 return cpu_to_le16((len & 65532) | ((len >> 16) & 3)); 1720#else 1721 return cpu_to_le16(len); 1722#endif 1723} 1724 1725/* 1726 * Hash Tree Directory indexing 1727 * (c) Daniel Phillips, 2001 1728 */ 1729 1730#define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \ 1731 EXT4_FEATURE_COMPAT_DIR_INDEX) && \ 1732 ext4_test_inode_flag((dir), EXT4_INODE_INDEX)) 1733#define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX) 1734#define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1) 1735 1736/* Legal values for the dx_root hash_version field: */ 1737 1738#define DX_HASH_LEGACY 0 1739#define DX_HASH_HALF_MD4 1 1740#define DX_HASH_TEA 2 1741#define DX_HASH_LEGACY_UNSIGNED 3 1742#define DX_HASH_HALF_MD4_UNSIGNED 4 1743#define DX_HASH_TEA_UNSIGNED 5 1744 1745static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc, 1746 const void *address, unsigned int length) 1747{ 1748 struct { 1749 struct shash_desc shash; 1750 char ctx[4]; 1751 } desc; 1752 int err; 1753 1754 BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx)); 1755 1756 desc.shash.tfm = sbi->s_chksum_driver; 1757 desc.shash.flags = 0; 1758 *(u32 *)desc.ctx = crc; 1759 1760 err = crypto_shash_update(&desc.shash, address, length); 1761 BUG_ON(err); 1762 1763 return *(u32 *)desc.ctx; 1764} 1765 1766#ifdef __KERNEL__ 1767 1768/* hash info structure used by the directory hash */ 1769struct dx_hash_info 1770{ 1771 u32 hash; 1772 u32 minor_hash; 1773 int hash_version; 1774 u32 *seed; 1775}; 1776 1777 1778/* 32 and 64 bit signed EOF for dx directories */ 1779#define EXT4_HTREE_EOF_32BIT ((1UL << (32 - 1)) - 1) 1780#define EXT4_HTREE_EOF_64BIT ((1ULL << (64 - 1)) - 1) 1781 1782 1783/* 1784 * Control parameters used by ext4_htree_next_block 1785 */ 1786#define HASH_NB_ALWAYS 1 1787 1788 1789/* 1790 * Describe an inode's exact location on disk and in memory 1791 */ 1792struct ext4_iloc 1793{ 1794 struct buffer_head *bh; 1795 unsigned long offset; 1796 ext4_group_t block_group; 1797}; 1798 1799static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc) 1800{ 1801 return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset); 1802} 1803 1804/* 1805 * This structure is stuffed into the struct file's private_data field 1806 * for directories. It is where we put information so that we can do 1807 * readdir operations in hash tree order. 1808 */ 1809struct dir_private_info { 1810 struct rb_root root; 1811 struct rb_node *curr_node; 1812 struct fname *extra_fname; 1813 loff_t last_pos; 1814 __u32 curr_hash; 1815 __u32 curr_minor_hash; 1816 __u32 next_hash; 1817}; 1818 1819/* calculate the first block number of the group */ 1820static inline ext4_fsblk_t 1821ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no) 1822{ 1823 return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) + 1824 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block); 1825} 1826 1827/* 1828 * Special error return code only used by dx_probe() and its callers. 1829 */ 1830#define ERR_BAD_DX_DIR (-(MAX_ERRNO - 1)) 1831 1832/* 1833 * Timeout and state flag for lazy initialization inode thread. 1834 */ 1835#define EXT4_DEF_LI_WAIT_MULT 10 1836#define EXT4_DEF_LI_MAX_START_DELAY 5 1837#define EXT4_LAZYINIT_QUIT 0x0001 1838#define EXT4_LAZYINIT_RUNNING 0x0002 1839 1840/* 1841 * Lazy inode table initialization info 1842 */ 1843struct ext4_lazy_init { 1844 unsigned long li_state; 1845 struct list_head li_request_list; 1846 struct mutex li_list_mtx; 1847}; 1848 1849struct ext4_li_request { 1850 struct super_block *lr_super; 1851 struct ext4_sb_info *lr_sbi; 1852 ext4_group_t lr_next_group; 1853 struct list_head lr_request; 1854 unsigned long lr_next_sched; 1855 unsigned long lr_timeout; 1856}; 1857 1858struct ext4_features { 1859 struct kobject f_kobj; 1860 struct completion f_kobj_unregister; 1861}; 1862 1863/* 1864 * This structure will be used for multiple mount protection. It will be 1865 * written into the block number saved in the s_mmp_block field in the 1866 * superblock. Programs that check MMP should assume that if 1867 * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe 1868 * to use the filesystem, regardless of how old the timestamp is. 1869 */ 1870#define EXT4_MMP_MAGIC 0x004D4D50U /* ASCII for MMP */ 1871#define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */ 1872#define EXT4_MMP_SEQ_FSCK 0xE24D4D50U /* mmp_seq value when being fscked */ 1873#define EXT4_MMP_SEQ_MAX 0xE24D4D4FU /* maximum valid mmp_seq value */ 1874 1875struct mmp_struct { 1876 __le32 mmp_magic; /* Magic number for MMP */ 1877 __le32 mmp_seq; /* Sequence no. updated periodically */ 1878 1879 /* 1880 * mmp_time, mmp_nodename & mmp_bdevname are only used for information 1881 * purposes and do not affect the correctness of the algorithm 1882 */ 1883 __le64 mmp_time; /* Time last updated */ 1884 char mmp_nodename[64]; /* Node which last updated MMP block */ 1885 char mmp_bdevname[32]; /* Bdev which last updated MMP block */ 1886 1887 /* 1888 * mmp_check_interval is used to verify if the MMP block has been 1889 * updated on the block device. The value is updated based on the 1890 * maximum time to write the MMP block during an update cycle. 1891 */ 1892 __le16 mmp_check_interval; 1893 1894 __le16 mmp_pad1; 1895 __le32 mmp_pad2[226]; 1896 __le32 mmp_checksum; /* crc32c(uuid+mmp_block) */ 1897}; 1898 1899/* arguments passed to the mmp thread */ 1900struct mmpd_data { 1901 struct buffer_head *bh; /* bh from initial read_mmp_block() */ 1902 struct super_block *sb; /* super block of the fs */ 1903}; 1904 1905/* 1906 * Check interval multiplier 1907 * The MMP block is written every update interval and initially checked every 1908 * update interval x the multiplier (the value is then adapted based on the 1909 * write latency). The reason is that writes can be delayed under load and we 1910 * don't want readers to incorrectly assume that the filesystem is no longer 1911 * in use. 1912 */ 1913#define EXT4_MMP_CHECK_MULT 2UL 1914 1915/* 1916 * Minimum interval for MMP checking in seconds. 1917 */ 1918#define EXT4_MMP_MIN_CHECK_INTERVAL 5UL 1919 1920/* 1921 * Maximum interval for MMP checking in seconds. 1922 */ 1923#define EXT4_MMP_MAX_CHECK_INTERVAL 300UL 1924 1925/* 1926 * Function prototypes 1927 */ 1928 1929/* 1930 * Ok, these declarations are also in <linux/kernel.h> but none of the 1931 * ext4 source programs needs to include it so they are duplicated here. 1932 */ 1933# define NORET_TYPE /**/ 1934# define ATTRIB_NORET __attribute__((noreturn)) 1935# define NORET_AND noreturn, 1936 1937/* bitmap.c */ 1938extern unsigned int ext4_count_free(char *bitmap, unsigned numchars); 1939void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group, 1940 struct ext4_group_desc *gdp, 1941 struct buffer_head *bh, int sz); 1942int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group, 1943 struct ext4_group_desc *gdp, 1944 struct buffer_head *bh, int sz); 1945void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group, 1946 struct ext4_group_desc *gdp, 1947 struct buffer_head *bh); 1948int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group, 1949 struct ext4_group_desc *gdp, 1950 struct buffer_head *bh); 1951 1952/* balloc.c */ 1953extern void ext4_get_group_no_and_offset(struct super_block *sb, 1954 ext4_fsblk_t blocknr, 1955 ext4_group_t *blockgrpp, 1956 ext4_grpblk_t *offsetp); 1957extern ext4_group_t ext4_get_group_number(struct super_block *sb, 1958 ext4_fsblk_t block); 1959 1960extern unsigned int ext4_block_group(struct super_block *sb, 1961 ext4_fsblk_t blocknr); 1962extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb, 1963 ext4_fsblk_t blocknr); 1964extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group); 1965extern unsigned long ext4_bg_num_gdb(struct super_block *sb, 1966 ext4_group_t group); 1967extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode, 1968 ext4_fsblk_t goal, 1969 unsigned int flags, 1970 unsigned long *count, 1971 int *errp); 1972extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi, 1973 s64 nclusters, unsigned int flags); 1974extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *); 1975extern void ext4_check_blocks_bitmap(struct super_block *); 1976extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb, 1977 ext4_group_t block_group, 1978 struct buffer_head ** bh); 1979extern int ext4_should_retry_alloc(struct super_block *sb, int *retries); 1980 1981extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb, 1982 ext4_group_t block_group); 1983extern int ext4_wait_block_bitmap(struct super_block *sb, 1984 ext4_group_t block_group, 1985 struct buffer_head *bh); 1986extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb, 1987 ext4_group_t block_group); 1988extern unsigned ext4_free_clusters_after_init(struct super_block *sb, 1989 ext4_group_t block_group, 1990 struct ext4_group_desc *gdp); 1991ext4_fsblk_t ext4_inode_to_goal_block(struct inode *); 1992 1993/* dir.c */ 1994extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *, 1995 struct file *, 1996 struct ext4_dir_entry_2 *, 1997 struct buffer_head *, char *, int, 1998 unsigned int); 1999#define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset) \ 2000 unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \ 2001 (de), (bh), (buf), (size), (offset))) 2002extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash, 2003 __u32 minor_hash, 2004 struct ext4_dir_entry_2 *dirent); 2005extern void ext4_htree_free_dir_info(struct dir_private_info *p); 2006extern int ext4_find_dest_de(struct inode *dir, struct inode *inode, 2007 struct buffer_head *bh, 2008 void *buf, int buf_size, 2009 const char *name, int namelen, 2010 struct ext4_dir_entry_2 **dest_de); 2011void ext4_insert_dentry(struct inode *inode, 2012 struct ext4_dir_entry_2 *de, 2013 int buf_size, 2014 const char *name, int namelen); 2015static inline void ext4_update_dx_flag(struct inode *inode) 2016{ 2017 if (!EXT4_HAS_COMPAT_FEATURE(inode->i_sb, 2018 EXT4_FEATURE_COMPAT_DIR_INDEX)) 2019 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX); 2020} 2021static unsigned char ext4_filetype_table[] = { 2022 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK 2023}; 2024 2025static inline unsigned char get_dtype(struct super_block *sb, int filetype) 2026{ 2027 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE) || 2028 (filetype >= EXT4_FT_MAX)) 2029 return DT_UNKNOWN; 2030 2031 return ext4_filetype_table[filetype]; 2032} 2033extern int ext4_check_all_de(struct inode *dir, struct buffer_head *bh, 2034 void *buf, int buf_size); 2035 2036/* fsync.c */ 2037extern int ext4_sync_file(struct file *, loff_t, loff_t, int); 2038 2039/* hash.c */ 2040extern int ext4fs_dirhash(const char *name, int len, struct 2041 dx_hash_info *hinfo); 2042 2043/* ialloc.c */ 2044extern struct inode *__ext4_new_inode(handle_t *, struct inode *, umode_t, 2045 const struct qstr *qstr, __u32 goal, 2046 uid_t *owner, int handle_type, 2047 unsigned int line_no, int nblocks); 2048 2049#define ext4_new_inode(handle, dir, mode, qstr, goal, owner) \ 2050 __ext4_new_inode((handle), (dir), (mode), (qstr), (goal), (owner), \ 2051 0, 0, 0) 2052#define ext4_new_inode_start_handle(dir, mode, qstr, goal, owner, \ 2053 type, nblocks) \ 2054 __ext4_new_inode(NULL, (dir), (mode), (qstr), (goal), (owner), \ 2055 (type), __LINE__, (nblocks)) 2056 2057 2058extern void ext4_free_inode(handle_t *, struct inode *); 2059extern struct inode * ext4_orphan_get(struct super_block *, unsigned long); 2060extern unsigned long ext4_count_free_inodes(struct super_block *); 2061extern unsigned long ext4_count_dirs(struct super_block *); 2062extern void ext4_check_inodes_bitmap(struct super_block *); 2063extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap); 2064extern int ext4_init_inode_table(struct super_block *sb, 2065 ext4_group_t group, int barrier); 2066extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate); 2067 2068/* mballoc.c */ 2069extern long ext4_mb_stats; 2070extern long ext4_mb_max_to_scan; 2071extern int ext4_mb_init(struct super_block *); 2072extern int ext4_mb_release(struct super_block *); 2073extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *, 2074 struct ext4_allocation_request *, int *); 2075extern int ext4_mb_reserve_blocks(struct super_block *, int); 2076extern void ext4_discard_preallocations(struct inode *); 2077extern int __init ext4_init_mballoc(void); 2078extern void ext4_exit_mballoc(void); 2079extern void ext4_free_blocks(handle_t *handle, struct inode *inode, 2080 struct buffer_head *bh, ext4_fsblk_t block, 2081 unsigned long count, int flags); 2082extern int ext4_mb_alloc_groupinfo(struct super_block *sb, 2083 ext4_group_t ngroups); 2084extern int ext4_mb_add_groupinfo(struct super_block *sb, 2085 ext4_group_t i, struct ext4_group_desc *desc); 2086extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb, 2087 ext4_fsblk_t block, unsigned long count); 2088extern int ext4_trim_fs(struct super_block *, struct fstrim_range *, 2089 unsigned long blkdev_flags); 2090 2091/* inode.c */ 2092struct buffer_head *ext4_getblk(handle_t *, struct inode *, ext4_lblk_t, int); 2093struct buffer_head *ext4_bread(handle_t *, struct inode *, ext4_lblk_t, int); 2094int ext4_get_block_write(struct inode *inode, sector_t iblock, 2095 struct buffer_head *bh_result, int create); 2096int ext4_get_block(struct inode *inode, sector_t iblock, 2097 struct buffer_head *bh_result, int create); 2098int ext4_da_get_block_prep(struct inode *inode, sector_t iblock, 2099 struct buffer_head *bh, int create); 2100int ext4_walk_page_buffers(handle_t *handle, 2101 struct buffer_head *head, 2102 unsigned from, 2103 unsigned to, 2104 int *partial, 2105 int (*fn)(handle_t *handle, 2106 struct buffer_head *bh)); 2107int do_journal_get_write_access(handle_t *handle, 2108 struct buffer_head *bh); 2109#define FALL_BACK_TO_NONDELALLOC 1 2110#define CONVERT_INLINE_DATA 2 2111 2112extern struct inode *ext4_iget(struct super_block *, unsigned long); 2113extern struct inode *ext4_iget_normal(struct super_block *, unsigned long); 2114extern int ext4_write_inode(struct inode *, struct writeback_control *); 2115extern int ext4_setattr(struct dentry *, struct iattr *); 2116extern int ext4_getattr(struct vfsmount *mnt, struct dentry *dentry, 2117 struct kstat *stat); 2118extern void ext4_evict_inode(struct inode *); 2119extern void ext4_clear_inode(struct inode *); 2120extern int ext4_sync_inode(handle_t *, struct inode *); 2121extern void ext4_dirty_inode(struct inode *, int); 2122extern int ext4_change_inode_journal_flag(struct inode *, int); 2123extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *); 2124extern int ext4_inode_attach_jinode(struct inode *inode); 2125extern int ext4_can_truncate(struct inode *inode); 2126extern void ext4_truncate(struct inode *); 2127extern int ext4_punch_hole(struct inode *inode, loff_t offset, loff_t length); 2128extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks); 2129extern void ext4_set_inode_flags(struct inode *); 2130extern void ext4_get_inode_flags(struct ext4_inode_info *); 2131extern int ext4_alloc_da_blocks(struct inode *inode); 2132extern void ext4_set_aops(struct inode *inode); 2133extern int ext4_writepage_trans_blocks(struct inode *); 2134extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks); 2135extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode, 2136 loff_t lstart, loff_t lend); 2137extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf); 2138extern qsize_t *ext4_get_reserved_space(struct inode *inode); 2139extern void ext4_da_update_reserve_space(struct inode *inode, 2140 int used, int quota_claim); 2141 2142/* indirect.c */ 2143extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode, 2144 struct ext4_map_blocks *map, int flags); 2145extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb, 2146 struct iov_iter *iter, loff_t offset); 2147extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock); 2148extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks); 2149extern void ext4_ind_truncate(handle_t *, struct inode *inode); 2150extern int ext4_ind_remove_space(handle_t *handle, struct inode *inode, 2151 ext4_lblk_t start, ext4_lblk_t end); 2152 2153/* ioctl.c */ 2154extern long ext4_ioctl(struct file *, unsigned int, unsigned long); 2155extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long); 2156 2157/* migrate.c */ 2158extern int ext4_ext_migrate(struct inode *); 2159extern int ext4_ind_migrate(struct inode *inode); 2160 2161/* namei.c */ 2162extern int ext4_dirent_csum_verify(struct inode *inode, 2163 struct ext4_dir_entry *dirent); 2164extern int ext4_orphan_add(handle_t *, struct inode *); 2165extern int ext4_orphan_del(handle_t *, struct inode *); 2166extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash, 2167 __u32 start_minor_hash, __u32 *next_hash); 2168extern int search_dir(struct buffer_head *bh, 2169 char *search_buf, 2170 int buf_size, 2171 struct inode *dir, 2172 const struct qstr *d_name, 2173 unsigned int offset, 2174 struct ext4_dir_entry_2 **res_dir); 2175extern int ext4_generic_delete_entry(handle_t *handle, 2176 struct inode *dir, 2177 struct ext4_dir_entry_2 *de_del, 2178 struct buffer_head *bh, 2179 void *entry_buf, 2180 int buf_size, 2181 int csum_size); 2182 2183/* resize.c */ 2184extern int ext4_group_add(struct super_block *sb, 2185 struct ext4_new_group_data *input); 2186extern int ext4_group_extend(struct super_block *sb, 2187 struct ext4_super_block *es, 2188 ext4_fsblk_t n_blocks_count); 2189extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count); 2190 2191/* super.c */ 2192extern int ext4_calculate_overhead(struct super_block *sb); 2193extern void ext4_superblock_csum_set(struct super_block *sb); 2194extern void *ext4_kvmalloc(size_t size, gfp_t flags); 2195extern void *ext4_kvzalloc(size_t size, gfp_t flags); 2196extern void ext4_kvfree(void *ptr); 2197extern int ext4_alloc_flex_bg_array(struct super_block *sb, 2198 ext4_group_t ngroup); 2199extern const char *ext4_decode_error(struct super_block *sb, int errno, 2200 char nbuf[16]); 2201 2202extern __printf(4, 5) 2203void __ext4_error(struct super_block *, const char *, unsigned int, 2204 const char *, ...); 2205extern __printf(5, 6) 2206void __ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t, 2207 const char *, ...); 2208extern __printf(5, 6) 2209void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t, 2210 const char *, ...); 2211extern void __ext4_std_error(struct super_block *, const char *, 2212 unsigned int, int); 2213extern __printf(4, 5) 2214void __ext4_abort(struct super_block *, const char *, unsigned int, 2215 const char *, ...); 2216extern __printf(4, 5) 2217void __ext4_warning(struct super_block *, const char *, unsigned int, 2218 const char *, ...); 2219extern __printf(3, 4) 2220void __ext4_msg(struct super_block *, const char *, const char *, ...); 2221extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp, 2222 const char *, unsigned int, const char *); 2223extern __printf(7, 8) 2224void __ext4_grp_locked_error(const char *, unsigned int, 2225 struct super_block *, ext4_group_t, 2226 unsigned long, ext4_fsblk_t, 2227 const char *, ...); 2228 2229#ifdef CONFIG_PRINTK 2230 2231#define ext4_error_inode(inode, func, line, block, fmt, ...) \ 2232 __ext4_error_inode(inode, func, line, block, fmt, ##__VA_ARGS__) 2233#define ext4_error_file(file, func, line, block, fmt, ...) \ 2234 __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__) 2235#define ext4_error(sb, fmt, ...) \ 2236 __ext4_error(sb, __func__, __LINE__, fmt, ##__VA_ARGS__) 2237#define ext4_abort(sb, fmt, ...) \ 2238 __ext4_abort(sb, __func__, __LINE__, fmt, ##__VA_ARGS__) 2239#define ext4_warning(sb, fmt, ...) \ 2240 __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__) 2241#define ext4_msg(sb, level, fmt, ...) \ 2242 __ext4_msg(sb, level, fmt, ##__VA_ARGS__) 2243#define dump_mmp_msg(sb, mmp, msg) \ 2244 __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg) 2245#define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...) \ 2246 __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \ 2247 fmt, ##__VA_ARGS__) 2248 2249#else 2250 2251#define ext4_error_inode(inode, func, line, block, fmt, ...) \ 2252do { \ 2253 no_printk(fmt, ##__VA_ARGS__); \ 2254 __ext4_error_inode(inode, "", 0, block, " "); \ 2255} while (0) 2256#define ext4_error_file(file, func, line, block, fmt, ...) \ 2257do { \ 2258 no_printk(fmt, ##__VA_ARGS__); \ 2259 __ext4_error_file(file, "", 0, block, " "); \ 2260} while (0) 2261#define ext4_error(sb, fmt, ...) \ 2262do { \ 2263 no_printk(fmt, ##__VA_ARGS__); \ 2264 __ext4_error(sb, "", 0, " "); \ 2265} while (0) 2266#define ext4_abort(sb, fmt, ...) \ 2267do { \ 2268 no_printk(fmt, ##__VA_ARGS__); \ 2269 __ext4_abort(sb, "", 0, " "); \ 2270} while (0) 2271#define ext4_warning(sb, fmt, ...) \ 2272do { \ 2273 no_printk(fmt, ##__VA_ARGS__); \ 2274 __ext4_warning(sb, "", 0, " "); \ 2275} while (0) 2276#define ext4_msg(sb, level, fmt, ...) \ 2277do { \ 2278 no_printk(fmt, ##__VA_ARGS__); \ 2279 __ext4_msg(sb, "", " "); \ 2280} while (0) 2281#define dump_mmp_msg(sb, mmp, msg) \ 2282 __dump_mmp_msg(sb, mmp, "", 0, "") 2283#define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...) \ 2284do { \ 2285 no_printk(fmt, ##__VA_ARGS__); \ 2286 __ext4_grp_locked_error("", 0, sb, grp, ino, block, " "); \ 2287} while (0) 2288 2289#endif 2290 2291extern void ext4_update_dynamic_rev(struct super_block *sb); 2292extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb, 2293 __u32 compat); 2294extern int ext4_update_rocompat_feature(handle_t *handle, 2295 struct super_block *sb, __u32 rocompat); 2296extern int ext4_update_incompat_feature(handle_t *handle, 2297 struct super_block *sb, __u32 incompat); 2298extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 2299 struct ext4_group_desc *bg); 2300extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 2301 struct ext4_group_desc *bg); 2302extern ext4_fsblk_t ext4_inode_table(struct super_block *sb, 2303 struct ext4_group_desc *bg); 2304extern __u32 ext4_free_group_clusters(struct super_block *sb, 2305 struct ext4_group_desc *bg); 2306extern __u32 ext4_free_inodes_count(struct super_block *sb, 2307 struct ext4_group_desc *bg); 2308extern __u32 ext4_used_dirs_count(struct super_block *sb, 2309 struct ext4_group_desc *bg); 2310extern __u32 ext4_itable_unused_count(struct super_block *sb, 2311 struct ext4_group_desc *bg); 2312extern void ext4_block_bitmap_set(struct super_block *sb, 2313 struct ext4_group_desc *bg, ext4_fsblk_t blk); 2314extern void ext4_inode_bitmap_set(struct super_block *sb, 2315 struct ext4_group_desc *bg, ext4_fsblk_t blk); 2316extern void ext4_inode_table_set(struct super_block *sb, 2317 struct ext4_group_desc *bg, ext4_fsblk_t blk); 2318extern void ext4_free_group_clusters_set(struct super_block *sb, 2319 struct ext4_group_desc *bg, 2320 __u32 count); 2321extern void ext4_free_inodes_set(struct super_block *sb, 2322 struct ext4_group_desc *bg, __u32 count); 2323extern void ext4_used_dirs_set(struct super_block *sb, 2324 struct ext4_group_desc *bg, __u32 count); 2325extern void ext4_itable_unused_set(struct super_block *sb, 2326 struct ext4_group_desc *bg, __u32 count); 2327extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group, 2328 struct ext4_group_desc *gdp); 2329extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group, 2330 struct ext4_group_desc *gdp); 2331extern int ext4_register_li_request(struct super_block *sb, 2332 ext4_group_t first_not_zeroed); 2333 2334static inline int ext4_has_group_desc_csum(struct super_block *sb) 2335{ 2336 return EXT4_HAS_RO_COMPAT_FEATURE(sb, 2337 EXT4_FEATURE_RO_COMPAT_GDT_CSUM) || 2338 (EXT4_SB(sb)->s_chksum_driver != NULL); 2339} 2340 2341static inline int ext4_has_metadata_csum(struct super_block *sb) 2342{ 2343 WARN_ON_ONCE(EXT4_HAS_RO_COMPAT_FEATURE(sb, 2344 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 2345 !EXT4_SB(sb)->s_chksum_driver); 2346 2347 return (EXT4_SB(sb)->s_chksum_driver != NULL); 2348} 2349static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es) 2350{ 2351 return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) | 2352 le32_to_cpu(es->s_blocks_count_lo); 2353} 2354 2355static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es) 2356{ 2357 return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) | 2358 le32_to_cpu(es->s_r_blocks_count_lo); 2359} 2360 2361static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es) 2362{ 2363 return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) | 2364 le32_to_cpu(es->s_free_blocks_count_lo); 2365} 2366 2367static inline void ext4_blocks_count_set(struct ext4_super_block *es, 2368 ext4_fsblk_t blk) 2369{ 2370 es->s_blocks_count_lo = cpu_to_le32((u32)blk); 2371 es->s_blocks_count_hi = cpu_to_le32(blk >> 32); 2372} 2373 2374static inline void ext4_free_blocks_count_set(struct ext4_super_block *es, 2375 ext4_fsblk_t blk) 2376{ 2377 es->s_free_blocks_count_lo = cpu_to_le32((u32)blk); 2378 es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32); 2379} 2380 2381static inline void ext4_r_blocks_count_set(struct ext4_super_block *es, 2382 ext4_fsblk_t blk) 2383{ 2384 es->s_r_blocks_count_lo = cpu_to_le32((u32)blk); 2385 es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32); 2386} 2387 2388static inline loff_t ext4_isize(struct ext4_inode *raw_inode) 2389{ 2390 if (S_ISREG(le16_to_cpu(raw_inode->i_mode))) 2391 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) | 2392 le32_to_cpu(raw_inode->i_size_lo); 2393 else 2394 return (loff_t) le32_to_cpu(raw_inode->i_size_lo); 2395} 2396 2397static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size) 2398{ 2399 raw_inode->i_size_lo = cpu_to_le32(i_size); 2400 raw_inode->i_size_high = cpu_to_le32(i_size >> 32); 2401} 2402 2403static inline 2404struct ext4_group_info *ext4_get_group_info(struct super_block *sb, 2405 ext4_group_t group) 2406{ 2407 struct ext4_group_info ***grp_info; 2408 long indexv, indexh; 2409 BUG_ON(group >= EXT4_SB(sb)->s_groups_count); 2410 grp_info = EXT4_SB(sb)->s_group_info; 2411 indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb)); 2412 indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1); 2413 return grp_info[indexv][indexh]; 2414} 2415 2416/* 2417 * Reading s_groups_count requires using smp_rmb() afterwards. See 2418 * the locking protocol documented in the comments of ext4_group_add() 2419 * in resize.c 2420 */ 2421static inline ext4_group_t ext4_get_groups_count(struct super_block *sb) 2422{ 2423 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 2424 2425 smp_rmb(); 2426 return ngroups; 2427} 2428 2429static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi, 2430 ext4_group_t block_group) 2431{ 2432 return block_group >> sbi->s_log_groups_per_flex; 2433} 2434 2435static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi) 2436{ 2437 return 1 << sbi->s_log_groups_per_flex; 2438} 2439 2440#define ext4_std_error(sb, errno) \ 2441do { \ 2442 if ((errno)) \ 2443 __ext4_std_error((sb), __func__, __LINE__, (errno)); \ 2444} while (0) 2445 2446#ifdef CONFIG_SMP 2447/* Each CPU can accumulate percpu_counter_batch clusters in their local 2448 * counters. So we need to make sure we have free clusters more 2449 * than percpu_counter_batch * nr_cpu_ids. Also add a window of 4 times. 2450 */ 2451#define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids)) 2452#else 2453#define EXT4_FREECLUSTERS_WATERMARK 0 2454#endif 2455 2456/* Update i_disksize. Requires i_mutex to avoid races with truncate */ 2457static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize) 2458{ 2459 WARN_ON_ONCE(S_ISREG(inode->i_mode) && 2460 !mutex_is_locked(&inode->i_mutex)); 2461 down_write(&EXT4_I(inode)->i_data_sem); 2462 if (newsize > EXT4_I(inode)->i_disksize) 2463 EXT4_I(inode)->i_disksize = newsize; 2464 up_write(&EXT4_I(inode)->i_data_sem); 2465} 2466 2467/* Update i_size, i_disksize. Requires i_mutex to avoid races with truncate */ 2468static inline int ext4_update_inode_size(struct inode *inode, loff_t newsize) 2469{ 2470 int changed = 0; 2471 2472 if (newsize > inode->i_size) { 2473 i_size_write(inode, newsize); 2474 changed = 1; 2475 } 2476 if (newsize > EXT4_I(inode)->i_disksize) { 2477 ext4_update_i_disksize(inode, newsize); 2478 changed |= 2; 2479 } 2480 return changed; 2481} 2482 2483struct ext4_group_info { 2484 unsigned long bb_state; 2485 struct rb_root bb_free_root; 2486 ext4_grpblk_t bb_first_free; /* first free block */ 2487 ext4_grpblk_t bb_free; /* total free blocks */ 2488 ext4_grpblk_t bb_fragments; /* nr of freespace fragments */ 2489 ext4_grpblk_t bb_largest_free_order;/* order of largest frag in BG */ 2490 struct list_head bb_prealloc_list; 2491#ifdef DOUBLE_CHECK 2492 void *bb_bitmap; 2493#endif 2494 struct rw_semaphore alloc_sem; 2495 ext4_grpblk_t bb_counters[]; /* Nr of free power-of-two-block 2496 * regions, index is order. 2497 * bb_counters[3] = 5 means 2498 * 5 free 8-block regions. */ 2499}; 2500 2501#define EXT4_GROUP_INFO_NEED_INIT_BIT 0 2502#define EXT4_GROUP_INFO_WAS_TRIMMED_BIT 1 2503#define EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT 2 2504#define EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT 3 2505 2506#define EXT4_MB_GRP_NEED_INIT(grp) \ 2507 (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state))) 2508#define EXT4_MB_GRP_BBITMAP_CORRUPT(grp) \ 2509 (test_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &((grp)->bb_state))) 2510#define EXT4_MB_GRP_IBITMAP_CORRUPT(grp) \ 2511 (test_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &((grp)->bb_state))) 2512 2513#define EXT4_MB_GRP_WAS_TRIMMED(grp) \ 2514 (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state))) 2515#define EXT4_MB_GRP_SET_TRIMMED(grp) \ 2516 (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state))) 2517#define EXT4_MB_GRP_CLEAR_TRIMMED(grp) \ 2518 (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state))) 2519 2520#define EXT4_MAX_CONTENTION 8 2521#define EXT4_CONTENTION_THRESHOLD 2 2522 2523static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb, 2524 ext4_group_t group) 2525{ 2526 return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group); 2527} 2528 2529/* 2530 * Returns true if the filesystem is busy enough that attempts to 2531 * access the block group locks has run into contention. 2532 */ 2533static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi) 2534{ 2535 return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD); 2536} 2537 2538static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group) 2539{ 2540 spinlock_t *lock = ext4_group_lock_ptr(sb, group); 2541 if (spin_trylock(lock)) 2542 /* 2543 * We're able to grab the lock right away, so drop the 2544 * lock contention counter. 2545 */ 2546 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0); 2547 else { 2548 /* 2549 * The lock is busy, so bump the contention counter, 2550 * and then wait on the spin lock. 2551 */ 2552 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1, 2553 EXT4_MAX_CONTENTION); 2554 spin_lock(lock); 2555 } 2556} 2557 2558static inline void ext4_unlock_group(struct super_block *sb, 2559 ext4_group_t group) 2560{ 2561 spin_unlock(ext4_group_lock_ptr(sb, group)); 2562} 2563 2564/* 2565 * Block validity checking 2566 */ 2567#define ext4_check_indirect_blockref(inode, bh) \ 2568 ext4_check_blockref(__func__, __LINE__, inode, \ 2569 (__le32 *)(bh)->b_data, \ 2570 EXT4_ADDR_PER_BLOCK((inode)->i_sb)) 2571 2572#define ext4_ind_check_inode(inode) \ 2573 ext4_check_blockref(__func__, __LINE__, inode, \ 2574 EXT4_I(inode)->i_data, \ 2575 EXT4_NDIR_BLOCKS) 2576 2577/* 2578 * Inodes and files operations 2579 */ 2580 2581/* dir.c */ 2582extern const struct file_operations ext4_dir_operations; 2583 2584/* file.c */ 2585extern const struct inode_operations ext4_file_inode_operations; 2586extern const struct file_operations ext4_file_operations; 2587extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin); 2588 2589/* inline.c */ 2590extern int ext4_get_max_inline_size(struct inode *inode); 2591extern int ext4_find_inline_data_nolock(struct inode *inode); 2592extern int ext4_init_inline_data(handle_t *handle, struct inode *inode, 2593 unsigned int len); 2594extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode); 2595 2596extern int ext4_readpage_inline(struct inode *inode, struct page *page); 2597extern int ext4_try_to_write_inline_data(struct address_space *mapping, 2598 struct inode *inode, 2599 loff_t pos, unsigned len, 2600 unsigned flags, 2601 struct page **pagep); 2602extern int ext4_write_inline_data_end(struct inode *inode, 2603 loff_t pos, unsigned len, 2604 unsigned copied, 2605 struct page *page); 2606extern struct buffer_head * 2607ext4_journalled_write_inline_data(struct inode *inode, 2608 unsigned len, 2609 struct page *page); 2610extern int ext4_da_write_inline_data_begin(struct address_space *mapping, 2611 struct inode *inode, 2612 loff_t pos, unsigned len, 2613 unsigned flags, 2614 struct page **pagep, 2615 void **fsdata); 2616extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos, 2617 unsigned len, unsigned copied, 2618 struct page *page); 2619extern int ext4_try_add_inline_entry(handle_t *handle, struct dentry *dentry, 2620 struct inode *inode); 2621extern int ext4_try_create_inline_dir(handle_t *handle, 2622 struct inode *parent, 2623 struct inode *inode); 2624extern int ext4_read_inline_dir(struct file *filp, 2625 struct dir_context *ctx, 2626 int *has_inline_data); 2627extern int htree_inlinedir_to_tree(struct file *dir_file, 2628 struct inode *dir, ext4_lblk_t block, 2629 struct dx_hash_info *hinfo, 2630 __u32 start_hash, __u32 start_minor_hash, 2631 int *has_inline_data); 2632extern struct buffer_head *ext4_find_inline_entry(struct inode *dir, 2633 const struct qstr *d_name, 2634 struct ext4_dir_entry_2 **res_dir, 2635 int *has_inline_data); 2636extern int ext4_delete_inline_entry(handle_t *handle, 2637 struct inode *dir, 2638 struct ext4_dir_entry_2 *de_del, 2639 struct buffer_head *bh, 2640 int *has_inline_data); 2641extern int empty_inline_dir(struct inode *dir, int *has_inline_data); 2642extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode, 2643 struct ext4_dir_entry_2 **parent_de, 2644 int *retval); 2645extern int ext4_inline_data_fiemap(struct inode *inode, 2646 struct fiemap_extent_info *fieinfo, 2647 int *has_inline); 2648extern int ext4_try_to_evict_inline_data(handle_t *handle, 2649 struct inode *inode, 2650 int needed); 2651extern void ext4_inline_data_truncate(struct inode *inode, int *has_inline); 2652 2653extern int ext4_convert_inline_data(struct inode *inode); 2654 2655static inline int ext4_has_inline_data(struct inode *inode) 2656{ 2657 return ext4_test_inode_flag(inode, EXT4_INODE_INLINE_DATA) && 2658 EXT4_I(inode)->i_inline_off; 2659} 2660 2661/* namei.c */ 2662extern const struct inode_operations ext4_dir_inode_operations; 2663extern const struct inode_operations ext4_special_inode_operations; 2664extern struct dentry *ext4_get_parent(struct dentry *child); 2665extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode, 2666 struct ext4_dir_entry_2 *de, 2667 int blocksize, int csum_size, 2668 unsigned int parent_ino, int dotdot_real_len); 2669extern void initialize_dirent_tail(struct ext4_dir_entry_tail *t, 2670 unsigned int blocksize); 2671extern int ext4_handle_dirty_dirent_node(handle_t *handle, 2672 struct inode *inode, 2673 struct buffer_head *bh); 2674#define S_SHIFT 12 2675static unsigned char ext4_type_by_mode[S_IFMT >> S_SHIFT] = { 2676 [S_IFREG >> S_SHIFT] = EXT4_FT_REG_FILE, 2677 [S_IFDIR >> S_SHIFT] = EXT4_FT_DIR, 2678 [S_IFCHR >> S_SHIFT] = EXT4_FT_CHRDEV, 2679 [S_IFBLK >> S_SHIFT] = EXT4_FT_BLKDEV, 2680 [S_IFIFO >> S_SHIFT] = EXT4_FT_FIFO, 2681 [S_IFSOCK >> S_SHIFT] = EXT4_FT_SOCK, 2682 [S_IFLNK >> S_SHIFT] = EXT4_FT_SYMLINK, 2683}; 2684 2685static inline void ext4_set_de_type(struct super_block *sb, 2686 struct ext4_dir_entry_2 *de, 2687 umode_t mode) { 2688 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE)) 2689 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT]; 2690} 2691 2692 2693/* symlink.c */ 2694extern const struct inode_operations ext4_symlink_inode_operations; 2695extern const struct inode_operations ext4_fast_symlink_inode_operations; 2696 2697/* block_validity */ 2698extern void ext4_release_system_zone(struct super_block *sb); 2699extern int ext4_setup_system_zone(struct super_block *sb); 2700extern int __init ext4_init_system_zone(void); 2701extern void ext4_exit_system_zone(void); 2702extern int ext4_data_block_valid(struct ext4_sb_info *sbi, 2703 ext4_fsblk_t start_blk, 2704 unsigned int count); 2705extern int ext4_check_blockref(const char *, unsigned int, 2706 struct inode *, __le32 *, unsigned int); 2707 2708/* extents.c */ 2709struct ext4_ext_path; 2710struct ext4_extent; 2711 2712/* 2713 * Maximum number of logical blocks in a file; ext4_extent's ee_block is 2714 * __le32. 2715 */ 2716#define EXT_MAX_BLOCKS 0xffffffff 2717 2718extern int ext4_ext_tree_init(handle_t *handle, struct inode *); 2719extern int ext4_ext_writepage_trans_blocks(struct inode *, int); 2720extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents); 2721extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode, 2722 struct ext4_map_blocks *map, int flags); 2723extern void ext4_ext_truncate(handle_t *, struct inode *); 2724extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start, 2725 ext4_lblk_t end); 2726extern void ext4_ext_init(struct super_block *); 2727extern void ext4_ext_release(struct super_block *); 2728extern long ext4_fallocate(struct file *file, int mode, loff_t offset, 2729 loff_t len); 2730extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode, 2731 loff_t offset, ssize_t len); 2732extern int ext4_map_blocks(handle_t *handle, struct inode *inode, 2733 struct ext4_map_blocks *map, int flags); 2734extern int ext4_ext_calc_metadata_amount(struct inode *inode, 2735 ext4_lblk_t lblocks); 2736extern int ext4_extent_tree_init(handle_t *, struct inode *); 2737extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode, 2738 int num, 2739 struct ext4_ext_path *path); 2740extern int ext4_can_extents_be_merged(struct inode *inode, 2741 struct ext4_extent *ex1, 2742 struct ext4_extent *ex2); 2743extern int ext4_ext_insert_extent(handle_t *, struct inode *, 2744 struct ext4_ext_path **, 2745 struct ext4_extent *, int); 2746extern struct ext4_ext_path *ext4_find_extent(struct inode *, ext4_lblk_t, 2747 struct ext4_ext_path **, 2748 int flags); 2749extern void ext4_ext_drop_refs(struct ext4_ext_path *); 2750extern int ext4_ext_check_inode(struct inode *inode); 2751extern int ext4_find_delalloc_range(struct inode *inode, 2752 ext4_lblk_t lblk_start, 2753 ext4_lblk_t lblk_end); 2754extern int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk); 2755extern ext4_lblk_t ext4_ext_next_allocated_block(struct ext4_ext_path *path); 2756extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, 2757 __u64 start, __u64 len); 2758extern int ext4_ext_precache(struct inode *inode); 2759extern int ext4_collapse_range(struct inode *inode, loff_t offset, loff_t len); 2760extern int ext4_swap_extents(handle_t *handle, struct inode *inode1, 2761 struct inode *inode2, ext4_lblk_t lblk1, 2762 ext4_lblk_t lblk2, ext4_lblk_t count, 2763 int mark_unwritten,int *err); 2764 2765/* move_extent.c */ 2766extern void ext4_double_down_write_data_sem(struct inode *first, 2767 struct inode *second); 2768extern void ext4_double_up_write_data_sem(struct inode *orig_inode, 2769 struct inode *donor_inode); 2770extern int ext4_move_extents(struct file *o_filp, struct file *d_filp, 2771 __u64 start_orig, __u64 start_donor, 2772 __u64 len, __u64 *moved_len); 2773 2774/* page-io.c */ 2775extern int __init ext4_init_pageio(void); 2776extern void ext4_exit_pageio(void); 2777extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags); 2778extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end); 2779extern int ext4_put_io_end(ext4_io_end_t *io_end); 2780extern void ext4_put_io_end_defer(ext4_io_end_t *io_end); 2781extern void ext4_io_submit_init(struct ext4_io_submit *io, 2782 struct writeback_control *wbc); 2783extern void ext4_end_io_rsv_work(struct work_struct *work); 2784extern void ext4_io_submit(struct ext4_io_submit *io); 2785extern int ext4_bio_write_page(struct ext4_io_submit *io, 2786 struct page *page, 2787 int len, 2788 struct writeback_control *wbc, 2789 bool keep_towrite); 2790 2791/* mmp.c */ 2792extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t); 2793 2794/* 2795 * Note that these flags will never ever appear in a buffer_head's state flag. 2796 * See EXT4_MAP_... to see where this is used. 2797 */ 2798enum ext4_state_bits { 2799 BH_AllocFromCluster /* allocated blocks were part of already 2800 * allocated cluster. */ 2801 = BH_JBDPrivateStart 2802}; 2803 2804/* 2805 * Add new method to test whether block and inode bitmaps are properly 2806 * initialized. With uninit_bg reading the block from disk is not enough 2807 * to mark the bitmap uptodate. We need to also zero-out the bitmap 2808 */ 2809#define BH_BITMAP_UPTODATE BH_JBDPrivateStart 2810 2811static inline int bitmap_uptodate(struct buffer_head *bh) 2812{ 2813 return (buffer_uptodate(bh) && 2814 test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state)); 2815} 2816static inline void set_bitmap_uptodate(struct buffer_head *bh) 2817{ 2818 set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state); 2819} 2820 2821/* 2822 * Disable DIO read nolock optimization, so new dioreaders will be forced 2823 * to grab i_mutex 2824 */ 2825static inline void ext4_inode_block_unlocked_dio(struct inode *inode) 2826{ 2827 ext4_set_inode_state(inode, EXT4_STATE_DIOREAD_LOCK); 2828 smp_mb(); 2829} 2830static inline void ext4_inode_resume_unlocked_dio(struct inode *inode) 2831{ 2832 smp_mb(); 2833 ext4_clear_inode_state(inode, EXT4_STATE_DIOREAD_LOCK); 2834} 2835 2836#define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1) 2837 2838/* For ioend & aio unwritten conversion wait queues */ 2839#define EXT4_WQ_HASH_SZ 37 2840#define ext4_ioend_wq(v) (&ext4__ioend_wq[((unsigned long)(v)) %\ 2841 EXT4_WQ_HASH_SZ]) 2842#define ext4_aio_mutex(v) (&ext4__aio_mutex[((unsigned long)(v)) %\ 2843 EXT4_WQ_HASH_SZ]) 2844extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 2845extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ]; 2846 2847#define EXT4_RESIZING 0 2848extern int ext4_resize_begin(struct super_block *sb); 2849extern void ext4_resize_end(struct super_block *sb); 2850 2851#endif /* __KERNEL__ */ 2852 2853#endif /* _EXT4_H */ 2854