fs/ext4/ext4.h

   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 */
  82 typedef int ext4_grpblk_t;
  83
  84 /* data type for filesystem-wide blocks number */
  85 typedef unsigned long long ext4_fsblk_t;
  86
  87 /* data type for file logical block number */
  88 typedef __u32 ext4_lblk_t;
  89
  90 /* data type for block group number */
  91 typedef 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
 129 struct 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 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
 162                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY)
 163
 164 struct ext4_map_blocks {
 165         ext4_fsblk_t m_pblk;
 166         ext4_lblk_t m_lblk;
 167         unsigned int m_len;
 168         unsigned int m_flags;
 169 };
 170
 171 /*
 172  * Flags for ext4_io_end->flags
 173  */
 174 #define EXT4_IO_END_UNWRITTEN   0x0001
 175
 176 /*
 177  * For converting unwritten extents on a work queue. 'handle' is used for
 178  * buffered writeback.
 179  */
 180 typedef struct ext4_io_end {
 181         struct list_head        list;           /* per-file finished IO list */
 182         handle_t                *handle;        /* handle reserved for extent
 183                                                  * conversion */
 184         struct inode            *inode;         /* file being written to */
 185         struct bio              *bio;           /* Linked list of completed
 186                                                  * bios covering the extent */
 187         unsigned int            flag;           /* unwritten or not */
 188         loff_t                  offset;         /* offset in the file */
 189         ssize_t                 size;           /* size of the extent */
 190         atomic_t                count;          /* reference counter */
 191 } ext4_io_end_t;
 192
 193 struct ext4_io_submit {
 194         int                     io_op;
 195         struct bio              *io_bio;
 196         ext4_io_end_t           *io_end;
 197         sector_t                io_next_block;
 198 };
 199
 200 /*
 201  * Special inodes numbers
 202  */
 203 #define EXT4_BAD_INO             1      /* Bad blocks inode */
 204 #define EXT4_ROOT_INO            2      /* Root inode */
 205 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
 206 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
 207 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
 208 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
 209 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
 210 #define EXT4_JOURNAL_INO         8      /* Journal inode */
 211
 212 /* First non-reserved inode for old ext4 filesystems */
 213 #define EXT4_GOOD_OLD_FIRST_INO 11
 214
 215 /*
 216  * Maximal count of links to a file
 217  */
 218 #define EXT4_LINK_MAX           65000
 219
 220 /*
 221  * Macro-instructions used to manage several block sizes
 222  */
 223 #define EXT4_MIN_BLOCK_SIZE             1024
 224 #define EXT4_MAX_BLOCK_SIZE             65536
 225 #define EXT4_MIN_BLOCK_LOG_SIZE         10
 226 #define EXT4_MAX_BLOCK_LOG_SIZE         16
 227 #ifdef __KERNEL__
 228 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
 229 #else
 230 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
 231 #endif
 232 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
 233 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
 234                                          EXT4_SB(s)->s_cluster_bits)
 235 #ifdef __KERNEL__
 236 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
 237 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
 238 #else
 239 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
 240 #endif
 241 #ifdef __KERNEL__
 242 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
 243 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
 244 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
 245 #else
 246 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 247                                  EXT4_GOOD_OLD_INODE_SIZE : \
 248                                  (s)->s_inode_size)
 249 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 250                                  EXT4_GOOD_OLD_FIRST_INO : \
 251                                  (s)->s_first_ino)
 252 #endif
 253 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
 254
 255 /* Translate a block number to a cluster number */
 256 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
 257 /* Translate a cluster number to a block number */
 258 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
 259 /* Translate # of blks to # of clusters */
 260 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
 261                                  (sbi)->s_cluster_bits)
 262 /* Mask out the low bits to get the starting block of the cluster */
 263 #define EXT4_PBLK_CMASK(s, pblk) ((pblk) &                              \
 264                                   ~((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
 265 #define EXT4_LBLK_CMASK(s, lblk) ((lblk) &                              \
 266                                   ~((ext4_lblk_t) (s)->s_cluster_ratio - 1))
 267 /* Get the cluster offset */
 268 #define EXT4_PBLK_COFF(s, pblk) ((pblk) &                               \
 269                                  ((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
 270 #define EXT4_LBLK_COFF(s, lblk) ((lblk) &                               \
 271                                  ((ext4_lblk_t) (s)->s_cluster_ratio - 1))
 272
 273 /*
 274  * Structure of a blocks group descriptor
 275  */
 276 struct ext4_group_desc
 277 {
 278         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
 279         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
 280         __le32  bg_inode_table_lo;      /* Inodes table block */
 281         __le16  bg_free_blocks_count_lo;/* Free blocks count */
 282         __le16  bg_free_inodes_count_lo;/* Free inodes count */
 283         __le16  bg_used_dirs_count_lo;  /* Directories count */
 284         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
 285         __le32  bg_exclude_bitmap_lo;   /* Exclude bitmap for snapshots */
 286         __le16  bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
 287         __le16  bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
 288         __le16  bg_itable_unused_lo;    /* Unused inodes count */
 289         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
 290         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
 291         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
 292         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
 293         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
 294         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
 295         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
 296         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
 297         __le32  bg_exclude_bitmap_hi;   /* Exclude bitmap block MSB */
 298         __le16  bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
 299         __le16  bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
 300         __u32   bg_reserved;
 301 };
 302
 303 #define EXT4_BG_INODE_BITMAP_CSUM_HI_END        \
 304         (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
 305          sizeof(__le16))
 306 #define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END        \
 307         (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
 308          sizeof(__le16))
 309
 310 /*
 311  * Structure of a flex block group info
 312  */
 313
 314 struct flex_groups {
 315         atomic64_t      free_clusters;
 316         atomic_t        free_inodes;
 317         atomic_t        used_dirs;
 318 };
 319
 320 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
 321 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
 322 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
 323
 324 /*
 325  * Macro-instructions used to manage group descriptors
 326  */
 327 #define EXT4_MIN_DESC_SIZE              32
 328 #define EXT4_MIN_DESC_SIZE_64BIT        64
 329 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
 330 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
 331 #ifdef __KERNEL__
 332 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
 333 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
 334 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
 335 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
 336 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
 337 #else
 338 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
 339 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
 340 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
 341 #endif
 342
 343 /*
 344  * Constants relative to the data blocks
 345  */
 346 #define EXT4_NDIR_BLOCKS                12
 347 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
 348 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
 349 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
 350 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
 351
 352 /*
 353  * Inode flags
 354  */
 355 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
 356 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
 357 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
 358 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
 359 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
 360 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
 361 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
 362 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
 363 /* Reserved for compression usage... */
 364 #define EXT4_DIRTY_FL                   0x00000100
 365 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
 366 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
 367         /* nb: was previously EXT2_ECOMPR_FL */
 368 #define EXT4_ENCRYPT_FL                 0x00000800 /* encrypted file */
 369 /* End compression flags --- maybe not all used */
 370 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
 371 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
 372 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
 373 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
 374 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
 375 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
 376 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
 377 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
 378 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
 379 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
 380 #define EXT4_INLINE_DATA_FL             0x10000000 /* Inode has inline data. */
 381 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
 382
 383 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
 384 #define EXT4_FL_USER_MODIFIABLE         0x004380FF /* User modifiable flags */
 385
 386 /* Flags that should be inherited by new inodes from their parent. */
 387 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
 388                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
 389                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
 390                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
 391
 392 /* Flags that are appropriate for regular files (all but dir-specific ones). */
 393 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
 394
 395 /* Flags that are appropriate for non-directories/regular files. */
 396 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
 397
 398 /* Mask out flags that are inappropriate for the given type of inode. */
 399 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
 400 {
 401         if (S_ISDIR(mode))
 402                 return flags;
 403         else if (S_ISREG(mode))
 404                 return flags & EXT4_REG_FLMASK;
 405         else
 406                 return flags & EXT4_OTHER_FLMASK;
 407 }
 408
 409 /*
 410  * Inode flags used for atomic set/get
 411  */
 412 enum {
 413         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
 414         EXT4_INODE_UNRM         = 1,    /* Undelete */
 415         EXT4_INODE_COMPR        = 2,    /* Compress file */
 416         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
 417         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
 418         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
 419         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
 420         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
 421 /* Reserved for compression usage... */
 422         EXT4_INODE_DIRTY        = 8,
 423         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
 424         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
 425         EXT4_INODE_ENCRYPT      = 11,   /* Encrypted file */
 426 /* End compression flags --- maybe not all used */
 427         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
 428         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
 429         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
 430         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
 431         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
 432         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
 433         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
 434         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
 435         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
 436         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
 437         EXT4_INODE_INLINE_DATA  = 28,   /* Data in inode. */
 438         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
 439 };
 440
 441 /*
 442  * Since it's pretty easy to mix up bit numbers and hex values, we use a
 443  * build-time check to make sure that EXT4_XXX_FL is consistent with respect to
 444  * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost
 445  * any extra space in the compiled kernel image, otherwise, the build will fail.
 446  * It's important that these values are the same, since we are using
 447  * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent
 448  * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk
 449  * values found in ext2, ext3 and ext4 filesystems, and of course the values
 450  * defined in e2fsprogs.
 451  *
 452  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
 453  */
 454 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
 455 #define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG))
 456
 457 static inline void ext4_check_flag_values(void)
 458 {
 459         CHECK_FLAG_VALUE(SECRM);
 460         CHECK_FLAG_VALUE(UNRM);
 461         CHECK_FLAG_VALUE(COMPR);
 462         CHECK_FLAG_VALUE(SYNC);
 463         CHECK_FLAG_VALUE(IMMUTABLE);
 464         CHECK_FLAG_VALUE(APPEND);
 465         CHECK_FLAG_VALUE(NODUMP);
 466         CHECK_FLAG_VALUE(NOATIME);
 467         CHECK_FLAG_VALUE(DIRTY);
 468         CHECK_FLAG_VALUE(COMPRBLK);
 469         CHECK_FLAG_VALUE(NOCOMPR);
 470         CHECK_FLAG_VALUE(ENCRYPT);
 471         CHECK_FLAG_VALUE(INDEX);
 472         CHECK_FLAG_VALUE(IMAGIC);
 473         CHECK_FLAG_VALUE(JOURNAL_DATA);
 474         CHECK_FLAG_VALUE(NOTAIL);
 475         CHECK_FLAG_VALUE(DIRSYNC);
 476         CHECK_FLAG_VALUE(TOPDIR);
 477         CHECK_FLAG_VALUE(HUGE_FILE);
 478         CHECK_FLAG_VALUE(EXTENTS);
 479         CHECK_FLAG_VALUE(EA_INODE);
 480         CHECK_FLAG_VALUE(EOFBLOCKS);
 481         CHECK_FLAG_VALUE(INLINE_DATA);
 482         CHECK_FLAG_VALUE(RESERVED);
 483 }
 484
 485 /* Used to pass group descriptor data when online resize is done */
 486 struct ext4_new_group_input {
 487         __u32 group;            /* Group number for this data */
 488         __u64 block_bitmap;     /* Absolute block number of block bitmap */
 489         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
 490         __u64 inode_table;      /* Absolute block number of inode table start */
 491         __u32 blocks_count;     /* Total number of blocks in this group */
 492         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
 493         __u16 unused;
 494 };
 495
 496 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 497 struct compat_ext4_new_group_input {
 498         u32 group;
 499         compat_u64 block_bitmap;
 500         compat_u64 inode_bitmap;
 501         compat_u64 inode_table;
 502         u32 blocks_count;
 503         u16 reserved_blocks;
 504         u16 unused;
 505 };
 506 #endif
 507
 508 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
 509 struct ext4_new_group_data {
 510         __u32 group;
 511         __u64 block_bitmap;
 512         __u64 inode_bitmap;
 513         __u64 inode_table;
 514         __u32 blocks_count;
 515         __u16 reserved_blocks;
 516         __u16 unused;
 517         __u32 free_blocks_count;
 518 };
 519
 520 /* Indexes used to index group tables in ext4_new_group_data */
 521 enum {
 522         BLOCK_BITMAP = 0,       /* block bitmap */
 523         INODE_BITMAP,           /* inode bitmap */
 524         INODE_TABLE,            /* inode tables */
 525         GROUP_TABLE_COUNT,
 526 };
 527
 528 /*
 529  * Flags used by ext4_map_blocks()
 530  */
 531         /* Allocate any needed blocks and/or convert an unwritten
 532            extent to be an initialized ext4 */
 533 #define EXT4_GET_BLOCKS_CREATE                  0x0001
 534         /* Request the creation of an unwritten extent */
 535 #define EXT4_GET_BLOCKS_UNWRIT_EXT              0x0002
 536 #define EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT       (EXT4_GET_BLOCKS_UNWRIT_EXT|\
 537                                                  EXT4_GET_BLOCKS_CREATE)
 538         /* Caller is from the delayed allocation writeout path
 539          * finally doing the actual allocation of delayed blocks */
 540 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
 541         /* caller is from the direct IO path, request to creation of an
 542         unwritten extents if not allocated, split the unwritten
 543         extent if blocks has been preallocated already*/
 544 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
 545 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
 546 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
 547                                          EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
 548         /* Convert extent to initialized after IO complete */
 549 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
 550                                          EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
 551         /* Eventual metadata allocation (due to growing extent tree)
 552          * should not fail, so try to use reserved blocks for that.*/
 553 #define EXT4_GET_BLOCKS_METADATA_NOFAIL         0x0020
 554         /* Don't normalize allocation size (used for fallocate) */
 555 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
 556         /* Request will not result in inode size update (user for fallocate) */
 557 #define EXT4_GET_BLOCKS_KEEP_SIZE               0x0080
 558         /* Do not take i_data_sem locking in ext4_map_blocks */
 559 #define EXT4_GET_BLOCKS_NO_LOCK                 0x0100
 560         /* Convert written extents to unwritten */
 561 #define EXT4_GET_BLOCKS_CONVERT_UNWRITTEN       0x0200
 562
 563 /*
 564  * The bit position of these flags must not overlap with any of the
 565  * EXT4_GET_BLOCKS_*.  They are used by ext4_find_extent(),
 566  * read_extent_tree_block(), ext4_split_extent_at(),
 567  * ext4_ext_insert_extent(), and ext4_ext_create_new_leaf().
 568  * EXT4_EX_NOCACHE is used to indicate that the we shouldn't be
 569  * caching the extents when reading from the extent tree while a
 570  * truncate or punch hole operation is in progress.
 571  */
 572 #define EXT4_EX_NOCACHE                         0x40000000
 573 #define EXT4_EX_FORCE_CACHE                     0x20000000
 574
 575 /*
 576  * Flags used by ext4_free_blocks
 577  */
 578 #define EXT4_FREE_BLOCKS_METADATA       0x0001
 579 #define EXT4_FREE_BLOCKS_FORGET         0x0002
 580 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
 581 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
 582 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
 583 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
 584
 585 /* Encryption algorithms */
 586 #define EXT4_ENCRYPTION_MODE_INVALID            0
 587 #define EXT4_ENCRYPTION_MODE_AES_256_XTS        1
 588 #define EXT4_ENCRYPTION_MODE_AES_256_GCM        2
 589 #define EXT4_ENCRYPTION_MODE_AES_256_CBC        3
 590 #define EXT4_ENCRYPTION_MODE_AES_256_CTS        4
 591
 592 #include "ext4_crypto.h"
 593
 594 /*
 595  * ioctl commands
 596  */
 597 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
 598 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
 599 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
 600 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
 601 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
 602 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
 603 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
 604 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
 605 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
 606 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
 607 #define EXT4_IOC_MIGRATE                _IO('f', 9)
 608  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
 609  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
 610 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
 611 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
 612 #define EXT4_IOC_RESIZE_FS              _IOW('f', 16, __u64)
 613 #define EXT4_IOC_SWAP_BOOT              _IO('f', 17)
 614 #define EXT4_IOC_PRECACHE_EXTENTS       _IO('f', 18)
 615 #define EXT4_IOC_SET_ENCRYPTION_POLICY  _IOR('f', 19, struct ext4_encryption_policy)
 616 #define EXT4_IOC_GET_ENCRYPTION_PWSALT  _IOW('f', 20, __u8[16])
 617 #define EXT4_IOC_GET_ENCRYPTION_POLICY  _IOW('f', 21, struct ext4_encryption_policy)
 618
 619 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 620 /*
 621  * ioctl commands in 32 bit emulation
 622  */
 623 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
 624 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
 625 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
 626 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
 627 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
 628 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
 629 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
 630 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
 631 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
 632 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
 633 #endif
 634
 635 /* Max physical block we can address w/o extents */
 636 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
 637
 638 /*
 639  * Structure of an inode on the disk
 640  */
 641 struct ext4_inode {
 642         __le16  i_mode;         /* File mode */
 643         __le16  i_uid;          /* Low 16 bits of Owner Uid */
 644         __le32  i_size_lo;      /* Size in bytes */
 645         __le32  i_atime;        /* Access time */
 646         __le32  i_ctime;        /* Inode Change time */
 647         __le32  i_mtime;        /* Modification time */
 648         __le32  i_dtime;        /* Deletion Time */
 649         __le16  i_gid;          /* Low 16 bits of Group Id */
 650         __le16  i_links_count;  /* Links count */
 651         __le32  i_blocks_lo;    /* Blocks count */
 652         __le32  i_flags;        /* File flags */
 653         union {
 654                 struct {
 655                         __le32  l_i_version;
 656                 } linux1;
 657                 struct {
 658                         __u32  h_i_translator;
 659                 } hurd1;
 660                 struct {
 661                         __u32  m_i_reserved1;
 662                 } masix1;
 663         } osd1;                         /* OS dependent 1 */
 664         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
 665         __le32  i_generation;   /* File version (for NFS) */
 666         __le32  i_file_acl_lo;  /* File ACL */
 667         __le32  i_size_high;
 668         __le32  i_obso_faddr;   /* Obsoleted fragment address */
 669         union {
 670                 struct {
 671                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
 672                         __le16  l_i_file_acl_high;
 673                         __le16  l_i_uid_high;   /* these 2 fields */
 674                         __le16  l_i_gid_high;   /* were reserved2[0] */
 675                         __le16  l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
 676                         __le16  l_i_reserved;
 677                 } linux2;
 678                 struct {
 679                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 680                         __u16   h_i_mode_high;
 681                         __u16   h_i_uid_high;
 682                         __u16   h_i_gid_high;
 683                         __u32   h_i_author;
 684                 } hurd2;
 685                 struct {
 686                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 687                         __le16  m_i_file_acl_high;
 688                         __u32   m_i_reserved2[2];
 689                 } masix2;
 690         } osd2;                         /* OS dependent 2 */
 691         __le16  i_extra_isize;
 692         __le16  i_checksum_hi;  /* crc32c(uuid+inum+inode) BE */
 693         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
 694         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
 695         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
 696         __le32  i_crtime;       /* File Creation time */
 697         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
 698         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
 699 };
 700
 701 struct move_extent {
 702         __u32 reserved;         /* should be zero */
 703         __u32 donor_fd;         /* donor file descriptor */
 704         __u64 orig_start;       /* logical start offset in block for orig */
 705         __u64 donor_start;      /* logical start offset in block for donor */
 706         __u64 len;              /* block length to be moved */
 707         __u64 moved_len;        /* moved block length */
 708 };
 709
 710 #define EXT4_EPOCH_BITS 2
 711 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
 712 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
 713
 714 /*
 715  * Extended fields will fit into an inode if the filesystem was formatted
 716  * with large inodes (-I 256 or larger) and there are not currently any EAs
 717  * consuming all of the available space. For new inodes we always reserve
 718  * enough space for the kernel's known extended fields, but for inodes
 719  * created with an old kernel this might not have been the case. None of
 720  * the extended inode fields is critical for correct filesystem operation.
 721  * This macro checks if a certain field fits in the inode. Note that
 722  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
 723  */
 724 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
 725         ((offsetof(typeof(*ext4_inode), field) +        \
 726           sizeof((ext4_inode)->field))                  \
 727         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
 728             (einode)->i_extra_isize))                   \
 729
 730 static inline __le32 ext4_encode_extra_time(struct timespec *time)
 731 {
 732        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
 733                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
 734                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
 735 }
 736
 737 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
 738 {
 739        if (sizeof(time->tv_sec) > 4)
 740                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
 741                                << 32;
 742        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
 743 }
 744
 745 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
 746 do {                                                                           \
 747         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
 748         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 749                 (raw_inode)->xtime ## _extra =                                 \
 750                                 ext4_encode_extra_time(&(inode)->xtime);       \
 751 } while (0)
 752
 753 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
 754 do {                                                                           \
 755         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 756                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
 757         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 758                 (raw_inode)->xtime ## _extra =                                 \
 759                                 ext4_encode_extra_time(&(einode)->xtime);      \
 760 } while (0)
 761
 762 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
 763 do {                                                                           \
 764         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
 765         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 766                 ext4_decode_extra_time(&(inode)->xtime,                        \
 767                                        raw_inode->xtime ## _extra);            \
 768         else                                                                   \
 769                 (inode)->xtime.tv_nsec = 0;                                    \
 770 } while (0)
 771
 772 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
 773 do {                                                                           \
 774         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 775                 (einode)->xtime.tv_sec =                                       \
 776                         (signed)le32_to_cpu((raw_inode)->xtime);               \
 777         else                                                                   \
 778                 (einode)->xtime.tv_sec = 0;                                    \
 779         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 780                 ext4_decode_extra_time(&(einode)->xtime,                       \
 781                                        raw_inode->xtime ## _extra);            \
 782         else                                                                   \
 783                 (einode)->xtime.tv_nsec = 0;                                   \
 784 } while (0)
 785
 786 #define i_disk_version osd1.linux1.l_i_version
 787
 788 #if defined(__KERNEL__) || defined(__linux__)
 789 #define i_reserved1     osd1.linux1.l_i_reserved1
 790 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
 791 #define i_blocks_high   osd2.linux2.l_i_blocks_high
 792 #define i_uid_low       i_uid
 793 #define i_gid_low       i_gid
 794 #define i_uid_high      osd2.linux2.l_i_uid_high
 795 #define i_gid_high      osd2.linux2.l_i_gid_high
 796 #define i_checksum_lo   osd2.linux2.l_i_checksum_lo
 797
 798 #elif defined(__GNU__)
 799
 800 #define i_translator    osd1.hurd1.h_i_translator
 801 #define i_uid_high      osd2.hurd2.h_i_uid_high
 802 #define i_gid_high      osd2.hurd2.h_i_gid_high
 803 #define i_author        osd2.hurd2.h_i_author
 804
 805 #elif defined(__masix__)
 806
 807 #define i_reserved1     osd1.masix1.m_i_reserved1
 808 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
 809 #define i_reserved2     osd2.masix2.m_i_reserved2
 810
 811 #endif /* defined(__KERNEL__) || defined(__linux__) */
 812
 813 #include "extents_status.h"
 814
 815 /*
 816  * fourth extended file system inode data in memory
 817  */
 818 struct ext4_inode_info {
 819         __le32  i_data[15];     /* unconverted */
 820         __u32   i_dtime;
 821         ext4_fsblk_t    i_file_acl;
 822
 823         /*
 824          * i_block_group is the number of the block group which contains
 825          * this file's inode.  Constant across the lifetime of the inode,
 826          * it is ued for making block allocation decisions - we try to
 827          * place a file's data blocks near its inode block, and new inodes
 828          * near to their parent directory's inode.
 829          */
 830         ext4_group_t    i_block_group;
 831         ext4_lblk_t     i_dir_start_lookup;
 832 #if (BITS_PER_LONG < 64)
 833         unsigned long   i_state_flags;          /* Dynamic state flags */
 834 #endif
 835         unsigned long   i_flags;
 836
 837         /*
 838          * Extended attributes can be read independently of the main file
 839          * data. Taking i_mutex even when reading would cause contention
 840          * between readers of EAs and writers of regular file data, so
 841          * instead we synchronize on xattr_sem when reading or changing
 842          * EAs.
 843          */
 844         struct rw_semaphore xattr_sem;
 845
 846         struct list_head i_orphan;      /* unlinked but open inodes */
 847
 848         /*
 849          * i_disksize keeps track of what the inode size is ON DISK, not
 850          * in memory.  During truncate, i_size is set to the new size by
 851          * the VFS prior to calling ext4_truncate(), but the filesystem won't
 852          * set i_disksize to 0 until the truncate is actually under way.
 853          *
 854          * The intent is that i_disksize always represents the blocks which
 855          * are used by this file.  This allows recovery to restart truncate
 856          * on orphans if we crash during truncate.  We actually write i_disksize
 857          * into the on-disk inode when writing inodes out, instead of i_size.
 858          *
 859          * The only time when i_disksize and i_size may be different is when
 860          * a truncate is in progress.  The only things which change i_disksize
 861          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
 862          */
 863         loff_t  i_disksize;
 864
 865         /*
 866          * i_data_sem is for serialising ext4_truncate() against
 867          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
 868          * data tree are chopped off during truncate. We can't do that in
 869          * ext4 because whenever we perform intermediate commits during
 870          * truncate, the inode and all the metadata blocks *must* be in a
 871          * consistent state which allows truncation of the orphans to restart
 872          * during recovery.  Hence we must fix the get_block-vs-truncate race
 873          * by other means, so we have i_data_sem.
 874          */
 875         struct rw_semaphore i_data_sem;
 876         struct inode vfs_inode;
 877         struct jbd2_inode *jinode;
 878
 879         spinlock_t i_raw_lock;  /* protects updates to the raw inode */
 880
 881         /*
 882          * File creation time. Its function is same as that of
 883          * struct timespec i_{a,c,m}time in the generic inode.
 884          */
 885         struct timespec i_crtime;
 886
 887         /* mballoc */
 888         struct list_head i_prealloc_list;
 889         spinlock_t i_prealloc_lock;
 890
 891         /* extents status tree */
 892         struct ext4_es_tree i_es_tree;
 893         rwlock_t i_es_lock;
 894         struct list_head i_es_list;
 895         unsigned int i_es_all_nr;       /* protected by i_es_lock */
 896         unsigned int i_es_shk_nr;       /* protected by i_es_lock */
 897         ext4_lblk_t i_es_shrink_lblk;   /* Offset where we start searching for
 898                                            extents to shrink. Protected by
 899                                            i_es_lock  */
 900
 901         /* ialloc */
 902         ext4_group_t    i_last_alloc_group;
 903
 904         /* allocation reservation info for delalloc */
 905         /* In case of bigalloc, these refer to clusters rather than blocks */
 906         unsigned int i_reserved_data_blocks;
 907         unsigned int i_reserved_meta_blocks;
 908         unsigned int i_allocated_meta_blocks;
 909         ext4_lblk_t i_da_metadata_calc_last_lblock;
 910         int i_da_metadata_calc_len;
 911
 912         /* on-disk additional length */
 913         __u16 i_extra_isize;
 914
 915         /* Indicate the inline data space. */
 916         u16 i_inline_off;
 917         u16 i_inline_size;
 918
 919 #ifdef CONFIG_QUOTA
 920         /* quota space reservation, managed internally by quota code */
 921         qsize_t i_reserved_quota;
 922 #endif
 923
 924         /* Lock protecting lists below */
 925         spinlock_t i_completed_io_lock;
 926         /*
 927          * Completed IOs that need unwritten extents handling and have
 928          * transaction reserved
 929          */
 930         struct list_head i_rsv_conversion_list;
 931         /*
 932          * Completed IOs that need unwritten extents handling and don't have
 933          * transaction reserved
 934          */
 935         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
 936         atomic_t i_unwritten; /* Nr. of inflight conversions pending */
 937         struct work_struct i_rsv_conversion_work;
 938
 939         spinlock_t i_block_reservation_lock;
 940
 941         /*
 942          * Transactions that contain inode's metadata needed to complete
 943          * fsync and fdatasync, respectively.
 944          */
 945         tid_t i_sync_tid;
 946         tid_t i_datasync_tid;
 947
 948 #ifdef CONFIG_QUOTA
 949         struct dquot *i_dquot[MAXQUOTAS];
 950 #endif
 951
 952         /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
 953         __u32 i_csum_seed;
 954
 955 #ifdef CONFIG_EXT4_FS_ENCRYPTION
 956         /* Encryption params */
 957         struct ext4_encryption_key i_encryption_key;
 958 #endif
 959 };
 960
 961 /*
 962  * File system states
 963  */
 964 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
 965 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
 966 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
 967
 968 /*
 969  * Misc. filesystem flags
 970  */
 971 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
 972 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
 973 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
 974
 975 /*
 976  * Mount flags set via mount options or defaults
 977  */
 978 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
 979 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
 980 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
 981 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
 982 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
 983 #define EXT4_MOUNT_ERRORS_MASK          0x00070
 984 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
 985 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
 986 #ifdef CONFIG_FS_DAX
 987 #define EXT4_MOUNT_DAX                  0x00200 /* Direct Access */
 988 #else
 989 #define EXT4_MOUNT_DAX                  0
 990 #endif
 991 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
 992 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
 993 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
 994 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
 995 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
 996 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
 997 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
 998 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
 999 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
1000 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
1001 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
1002 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
1003 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
1004 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
1005 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
1006 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
1007 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
1008 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
1009 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
1010 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
1011 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
1012
1013 /*
1014  * Mount flags set either automatically (could not be set by mount option)
1015  * based on per file system feature or property or in special cases such as
1016  * distinguishing between explicit mount option definition and default.
1017  */
1018 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
1019                                                       specified delalloc */
1020 #define EXT4_MOUNT2_STD_GROUP_SIZE      0x00000002 /* We have standard group
1021                                                       size of blocksize * 8
1022                                                       blocks */
1023 #define EXT4_MOUNT2_HURD_COMPAT         0x00000004 /* Support HURD-castrated
1024                                                       file systems */
1025
1026 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
1027                                                 ~EXT4_MOUNT_##opt
1028 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
1029                                                 EXT4_MOUNT_##opt
1030 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
1031                                          EXT4_MOUNT_##opt)
1032
1033 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
1034                                                 ~EXT4_MOUNT2_##opt
1035 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
1036                                                 EXT4_MOUNT2_##opt
1037 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
1038                                          EXT4_MOUNT2_##opt)
1039
1040 #define ext4_test_and_set_bit           __test_and_set_bit_le
1041 #define ext4_set_bit                    __set_bit_le
1042 #define ext4_set_bit_atomic             ext2_set_bit_atomic
1043 #define ext4_test_and_clear_bit         __test_and_clear_bit_le
1044 #define ext4_clear_bit                  __clear_bit_le
1045 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
1046 #define ext4_test_bit                   test_bit_le
1047 #define ext4_find_next_zero_bit         find_next_zero_bit_le
1048 #define ext4_find_next_bit              find_next_bit_le
1049
1050 extern void ext4_set_bits(void *bm, int cur, int len);
1051
1052 /*
1053  * Maximal mount counts between two filesystem checks
1054  */
1055 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
1056 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
1057
1058 /*
1059  * Behaviour when detecting errors
1060  */
1061 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
1062 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
1063 #define EXT4_ERRORS_PANIC               3       /* Panic */
1064 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
1065
1066 /* Metadata checksum algorithm codes */
1067 #define EXT4_CRC32C_CHKSUM              1
1068
1069 /* Encryption algorithms */
1070 #define EXT4_ENCRYPTION_MODE_INVALID            0
1071 #define EXT4_ENCRYPTION_MODE_AES_256_XTS        1
1072 #define EXT4_ENCRYPTION_MODE_AES_256_GCM        2
1073 #define EXT4_ENCRYPTION_MODE_AES_256_CBC        3
1074
1075 /*
1076  * Structure of the super block
1077  */
1078 struct ext4_super_block {
1079 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1080         __le32  s_blocks_count_lo;      /* Blocks count */
1081         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1082         __le32  s_free_blocks_count_lo; /* Free blocks count */
1083 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1084         __le32  s_first_data_block;     /* First Data Block */
1085         __le32  s_log_block_size;       /* Block size */
1086         __le32  s_log_cluster_size;     /* Allocation cluster size */
1087 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1088         __le32  s_clusters_per_group;   /* # Clusters per group */
1089         __le32  s_inodes_per_group;     /* # Inodes per group */
1090         __le32  s_mtime;                /* Mount time */
1091 /*30*/  __le32  s_wtime;                /* Write time */
1092         __le16  s_mnt_count;            /* Mount count */
1093         __le16  s_max_mnt_count;        /* Maximal mount count */
1094         __le16  s_magic;                /* Magic signature */
1095         __le16  s_state;                /* File system state */
1096         __le16  s_errors;               /* Behaviour when detecting errors */
1097         __le16  s_minor_rev_level;      /* minor revision level */
1098 /*40*/  __le32  s_lastcheck;            /* time of last check */
1099         __le32  s_checkinterval;        /* max. time between checks */
1100         __le32  s_creator_os;           /* OS */
1101         __le32  s_rev_level;            /* Revision level */
1102 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1103         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1104         /*
1105          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1106          *
1107          * Note: the difference between the compatible feature set and
1108          * the incompatible feature set is that if there is a bit set
1109          * in the incompatible feature set that the kernel doesn't
1110          * know about, it should refuse to mount the filesystem.
1111          *
1112          * e2fsck's requirements are more strict; if it doesn't know
1113          * about a feature in either the compatible or incompatible
1114          * feature set, it must abort and not try to meddle with
1115          * things it doesn't understand...
1116          */
1117         __le32  s_first_ino;            /* First non-reserved inode */
1118         __le16  s_inode_size;           /* size of inode structure */
1119         __le16  s_block_group_nr;       /* block group # of this superblock */
1120         __le32  s_feature_compat;       /* compatible feature set */
1121 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1122         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1123 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1124 /*78*/  char    s_volume_name[16];      /* volume name */
1125 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1126 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1127         /*
1128          * Performance hints.  Directory preallocation should only
1129          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1130          */
1131         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1132         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1133         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1134         /*
1135          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1136          */
1137 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1138 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1139         __le32  s_journal_dev;          /* device number of journal file */
1140         __le32  s_last_orphan;          /* start of list of inodes to delete */
1141         __le32  s_hash_seed[4];         /* HTREE hash seed */
1142         __u8    s_def_hash_version;     /* Default hash version to use */
1143         __u8    s_jnl_backup_type;
1144         __le16  s_desc_size;            /* size of group descriptor */
1145 /*100*/ __le32  s_default_mount_opts;
1146         __le32  s_first_meta_bg;        /* First metablock block group */
1147         __le32  s_mkfs_time;            /* When the filesystem was created */
1148         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1149         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1150 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1151         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1152         __le32  s_free_blocks_count_hi; /* Free blocks count */
1153         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1154         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1155         __le32  s_flags;                /* Miscellaneous flags */
1156         __le16  s_raid_stride;          /* RAID stride */
1157         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1158         __le64  s_mmp_block;            /* Block for multi-mount protection */
1159         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1160         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1161         __u8    s_checksum_type;        /* metadata checksum algorithm used */
1162         __u8    s_encryption_level;     /* versioning level for encryption */
1163         __u8    s_reserved_pad;         /* Padding to next 32bits */
1164         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1165         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1166         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1167         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1168                                               snapshot's future use */
1169         __le32  s_snapshot_list;        /* inode number of the head of the
1170                                            on-disk snapshot list */
1171 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1172         __le32  s_error_count;          /* number of fs errors */
1173         __le32  s_first_error_time;     /* first time an error happened */
1174         __le32  s_first_error_ino;      /* inode involved in first error */
1175         __le64  s_first_error_block;    /* block involved of first error */
1176         __u8    s_first_error_func[32]; /* function where the error happened */
1177         __le32  s_first_error_line;     /* line number where error happened */
1178         __le32  s_last_error_time;      /* most recent time of an error */
1179         __le32  s_last_error_ino;       /* inode involved in last error */
1180         __le32  s_last_error_line;      /* line number where error happened */
1181         __le64  s_last_error_block;     /* block involved of last error */
1182         __u8    s_last_error_func[32];  /* function where the error happened */
1183 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1184         __u8    s_mount_opts[64];
1185         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1186         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1187         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1188         __le32  s_backup_bgs[2];        /* groups with sparse_super2 SBs */
1189         __u8    s_encrypt_algos[4];     /* Encryption algorithms in use  */
1190         __u8    s_encrypt_pw_salt[16];  /* Salt used for string2key algorithm */
1191         __le32  s_lpf_ino;              /* Location of the lost+found inode */
1192         __le32  s_reserved[100];        /* Padding to the end of the block */
1193         __le32  s_checksum;             /* crc32c(superblock) */
1194 };
1195
1196 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1197
1198 #ifdef __KERNEL__
1199
1200 /*
1201  * run-time mount flags
1202  */
1203 #define EXT4_MF_MNTDIR_SAMPLED          0x0001
1204 #define EXT4_MF_FS_ABORTED              0x0002  /* Fatal error detected */
1205 #define EXT4_MF_TEST_DUMMY_ENCRYPTION   0x0004
1206
1207 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1208 #define DUMMY_ENCRYPTION_ENABLED(sbi) (unlikely((sbi)->s_mount_flags & \
1209                                                 EXT4_MF_TEST_DUMMY_ENCRYPTION))
1210 #else
1211 #define DUMMY_ENCRYPTION_ENABLED(sbi) (0)
1212 #endif
1213
1214 /* Number of quota types we support */
1215 #define EXT4_MAXQUOTAS 2
1216
1217 /*
1218  * fourth extended-fs super-block data in memory
1219  */
1220 struct ext4_sb_info {
1221         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1222         unsigned long s_inodes_per_block;/* Number of inodes per block */
1223         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1224         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1225         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1226         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1227         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1228         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1229         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1230         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1231         unsigned long s_overhead;  /* # of fs overhead clusters */
1232         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1233         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1234         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1235         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1236         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1237         struct buffer_head **s_group_desc;
1238         unsigned int s_mount_opt;
1239         unsigned int s_mount_opt2;
1240         unsigned int s_mount_flags;
1241         unsigned int s_def_mount_opt;
1242         ext4_fsblk_t s_sb_block;
1243         atomic64_t s_resv_clusters;
1244         kuid_t s_resuid;
1245         kgid_t s_resgid;
1246         unsigned short s_mount_state;
1247         unsigned short s_pad;
1248         int s_addr_per_block_bits;
1249         int s_desc_per_block_bits;
1250         int s_inode_size;
1251         int s_first_ino;
1252         unsigned int s_inode_readahead_blks;
1253         unsigned int s_inode_goal;
1254         spinlock_t s_next_gen_lock;
1255         u32 s_next_generation;
1256         u32 s_hash_seed[4];
1257         int s_def_hash_version;
1258         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1259         struct percpu_counter s_freeclusters_counter;
1260         struct percpu_counter s_freeinodes_counter;
1261         struct percpu_counter s_dirs_counter;
1262         struct percpu_counter s_dirtyclusters_counter;
1263         struct blockgroup_lock *s_blockgroup_lock;
1264         struct proc_dir_entry *s_proc;
1265         struct kobject s_kobj;
1266         struct completion s_kobj_unregister;
1267         struct super_block *s_sb;
1268
1269         /* Journaling */
1270         struct journal_s *s_journal;
1271         struct list_head s_orphan;
1272         struct mutex s_orphan_lock;
1273         unsigned long s_resize_flags;           /* Flags indicating if there
1274                                                    is a resizer */
1275         unsigned long s_commit_interval;
1276         u32 s_max_batch_time;
1277         u32 s_min_batch_time;
1278         struct block_device *journal_bdev;
1279 #ifdef CONFIG_QUOTA
1280         char *s_qf_names[EXT4_MAXQUOTAS];       /* Names of quota files with journalled quota */
1281         int s_jquota_fmt;                       /* Format of quota to use */
1282 #endif
1283         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1284         struct rb_root system_blks;
1285
1286 #ifdef EXTENTS_STATS
1287         /* ext4 extents stats */
1288         unsigned long s_ext_min;
1289         unsigned long s_ext_max;
1290         unsigned long s_depth_max;
1291         spinlock_t s_ext_stats_lock;
1292         unsigned long s_ext_blocks;
1293         unsigned long s_ext_extents;
1294 #endif
1295
1296         /* for buddy allocator */
1297         struct ext4_group_info ***s_group_info;
1298         struct inode *s_buddy_cache;
1299         spinlock_t s_md_lock;
1300         unsigned short *s_mb_offsets;
1301         unsigned int *s_mb_maxs;
1302         unsigned int s_group_info_size;
1303
1304         /* tunables */
1305         unsigned long s_stripe;
1306         unsigned int s_mb_stream_request;
1307         unsigned int s_mb_max_to_scan;
1308         unsigned int s_mb_min_to_scan;
1309         unsigned int s_mb_stats;
1310         unsigned int s_mb_order2_reqs;
1311         unsigned int s_mb_group_prealloc;
1312         unsigned int s_max_dir_size_kb;
1313         /* where last allocation was done - for stream allocation */
1314         unsigned long s_mb_last_group;
1315         unsigned long s_mb_last_start;
1316
1317         /* stats for buddy allocator */
1318         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1319         atomic_t s_bal_success; /* we found long enough chunks */
1320         atomic_t s_bal_allocated;       /* in blocks */
1321         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1322         atomic_t s_bal_goals;   /* goal hits */
1323         atomic_t s_bal_breaks;  /* too long searches */
1324         atomic_t s_bal_2orders; /* 2^order hits */
1325         spinlock_t s_bal_lock;
1326         unsigned long s_mb_buddies_generated;
1327         unsigned long long s_mb_generation_time;
1328         atomic_t s_mb_lost_chunks;
1329         atomic_t s_mb_preallocated;
1330         atomic_t s_mb_discarded;
1331         atomic_t s_lock_busy;
1332
1333         /* locality groups */
1334         struct ext4_locality_group __percpu *s_locality_groups;
1335
1336         /* for write statistics */
1337         unsigned long s_sectors_written_start;
1338         u64 s_kbytes_written;
1339
1340         /* the size of zero-out chunk */
1341         unsigned int s_extent_max_zeroout_kb;
1342
1343         unsigned int s_log_groups_per_flex;
1344         struct flex_groups *s_flex_groups;
1345         ext4_group_t s_flex_groups_allocated;
1346
1347         /* workqueue for reserved extent conversions (buffered io) */
1348         struct workqueue_struct *rsv_conversion_wq;
1349
1350         /* timer for periodic error stats printing */
1351         struct timer_list s_err_report;
1352
1353         /* Lazy inode table initialization info */
1354         struct ext4_li_request *s_li_request;
1355         /* Wait multiplier for lazy initialization thread */
1356         unsigned int s_li_wait_mult;
1357
1358         /* Kernel thread for multiple mount protection */
1359         struct task_struct *s_mmp_tsk;
1360
1361         /* record the last minlen when FITRIM is called. */
1362         atomic_t s_last_trim_minblks;
1363
1364         /* Reference to checksum algorithm driver via cryptoapi */
1365         struct crypto_shash *s_chksum_driver;
1366
1367         /* Precomputed FS UUID checksum for seeding other checksums */
1368         __u32 s_csum_seed;
1369
1370         /* Reclaim extents from extent status tree */
1371         struct shrinker s_es_shrinker;
1372         struct list_head s_es_list;     /* List of inodes with reclaimable extents */
1373         long s_es_nr_inode;
1374         struct ext4_es_stats s_es_stats;
1375         struct mb_cache *s_mb_cache;
1376         spinlock_t s_es_lock ____cacheline_aligned_in_smp;
1377
1378         /* Ratelimit ext4 messages. */
1379         struct ratelimit_state s_err_ratelimit_state;
1380         struct ratelimit_state s_warning_ratelimit_state;
1381         struct ratelimit_state s_msg_ratelimit_state;
1382
1383 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1384         /* Encryption */
1385         uint32_t s_file_encryption_mode;
1386         uint32_t s_dir_encryption_mode;
1387 #endif
1388 };
1389
1390 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1391 {
1392         return sb->s_fs_info;
1393 }
1394 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1395 {
1396         return container_of(inode, struct ext4_inode_info, vfs_inode);
1397 }
1398
1399 static inline struct timespec ext4_current_time(struct inode *inode)
1400 {
1401         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1402                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1403 }
1404
1405 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1406 {
1407         return ino == EXT4_ROOT_INO ||
1408                 ino == EXT4_USR_QUOTA_INO ||
1409                 ino == EXT4_GRP_QUOTA_INO ||
1410                 ino == EXT4_BOOT_LOADER_INO ||
1411                 ino == EXT4_JOURNAL_INO ||
1412                 ino == EXT4_RESIZE_INO ||
1413                 (ino >= EXT4_FIRST_INO(sb) &&
1414                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1415 }
1416
1417 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
1418                                               struct ext4_io_end *io_end)
1419 {
1420         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
1421                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
1422                 atomic_inc(&EXT4_I(inode)->i_unwritten);
1423         }
1424 }
1425
1426 static inline ext4_io_end_t *ext4_inode_aio(struct inode *inode)
1427 {
1428         return inode->i_private;
1429 }
1430
1431 static inline void ext4_inode_aio_set(struct inode *inode, ext4_io_end_t *io)
1432 {
1433         inode->i_private = io;
1434 }
1435
1436 /*
1437  * Inode dynamic state flags
1438  */
1439 enum {
1440         EXT4_STATE_JDATA,               /* journaled data exists */
1441         EXT4_STATE_NEW,                 /* inode is newly created */
1442         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1443         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1444         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1445         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1446         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1447         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1448         EXT4_STATE_DIOREAD_LOCK,        /* Disable support for dio read
1449                                            nolocking */
1450         EXT4_STATE_MAY_INLINE_DATA,     /* may have in-inode data */
1451         EXT4_STATE_ORDERED_MODE,        /* data=ordered mode */
1452         EXT4_STATE_EXT_PRECACHED,       /* extents have been precached */
1453 };
1454
1455 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1456 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1457 {                                                                       \
1458         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1459 }                                                                       \
1460 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1461 {                                                                       \
1462         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1463 }                                                                       \
1464 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1465 {                                                                       \
1466         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1467 }
1468
1469 /* Add these declarations here only so that these functions can be
1470  * found by name.  Otherwise, they are very hard to locate. */
1471 static inline int ext4_test_inode_flag(struct inode *inode, int bit);
1472 static inline void ext4_set_inode_flag(struct inode *inode, int bit);
1473 static inline void ext4_clear_inode_flag(struct inode *inode, int bit);
1474 EXT4_INODE_BIT_FNS(flag, flags, 0)
1475
1476 /* Add these declarations here only so that these functions can be
1477  * found by name.  Otherwise, they are very hard to locate. */
1478 static inline int ext4_test_inode_state(struct inode *inode, int bit);
1479 static inline void ext4_set_inode_state(struct inode *inode, int bit);
1480 static inline void ext4_clear_inode_state(struct inode *inode, int bit);
1481 #if (BITS_PER_LONG < 64)
1482 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1483
1484 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1485 {
1486         (ei)->i_state_flags = 0;
1487 }
1488 #else
1489 EXT4_INODE_BIT_FNS(state, flags, 32)
1490
1491 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1492 {
1493         /* We depend on the fact that callers will set i_flags */
1494 }
1495 #endif
1496 #else
1497 /* Assume that user mode programs are passing in an ext4fs superblock, not
1498  * a kernel struct super_block.  This will allow us to call the feature-test
1499  * macros from user land. */
1500 #define EXT4_SB(sb)     (sb)
1501 #endif
1502
1503 /*
1504  * Returns true if the inode is inode is encrypted
1505  */
1506 static inline int ext4_encrypted_inode(struct inode *inode)
1507 {
1508 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1509         return ext4_test_inode_flag(inode, EXT4_INODE_ENCRYPT);
1510 #else
1511         return 0;
1512 #endif
1513 }
1514
1515 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1516
1517 /*
1518  * Codes for operating systems
1519  */
1520 #define EXT4_OS_LINUX           0
1521 #define EXT4_OS_HURD            1
1522 #define EXT4_OS_MASIX           2
1523 #define EXT4_OS_FREEBSD         3
1524 #define EXT4_OS_LITES           4
1525
1526 /*
1527  * Revision levels
1528  */
1529 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1530 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1531
1532 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1533 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1534
1535 #define EXT4_GOOD_OLD_INODE_SIZE 128
1536
1537 /*
1538  * Feature set definitions
1539  */
1540
1541 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1542         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1543 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1544         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1545 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1546         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1547 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1548         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1549 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1550         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1551 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1552         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1553 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1554         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1555 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1556         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1557 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1558         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1559
1560 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1561 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1562 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1563 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1564 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1565 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1566 #define EXT4_FEATURE_COMPAT_SPARSE_SUPER2       0x0200
1567
1568 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1569 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1570 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1571 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1572 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1573 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1574 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1575 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1576 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1577 /*
1578  * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM).  When
1579  * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
1580  * all other data structures' checksums.  However, the METADATA_CSUM and
1581  * GDT_CSUM bits are mutually exclusive.
1582  */
1583 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM    0x0400
1584 #define EXT4_FEATURE_RO_COMPAT_READONLY         0x1000
1585
1586 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1587 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1588 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1589 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1590 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1591 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1592 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1593 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1594 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1595 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1596 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1597 #define EXT4_FEATURE_INCOMPAT_BG_USE_META_CSUM  0x2000 /* use crc32c for bg */
1598 #define EXT4_FEATURE_INCOMPAT_LARGEDIR          0x4000 /* >2GB or 3-lvl htree */
1599 #define EXT4_FEATURE_INCOMPAT_INLINE_DATA       0x8000 /* data in inode */
1600 #define EXT4_FEATURE_INCOMPAT_ENCRYPT           0x10000
1601
1602 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1603 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1604                                          EXT4_FEATURE_INCOMPAT_META_BG)
1605 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1606                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1607                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1608
1609 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1610 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1611                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1612                                          EXT4_FEATURE_INCOMPAT_META_BG)
1613 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1614                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1615                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1616
1617 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1618 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1619                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1620                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1621                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1622                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1623                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1624                                          EXT4_FEATURE_INCOMPAT_MMP | \
1625                                          EXT4_FEATURE_INCOMPAT_INLINE_DATA | \
1626                                          EXT4_FEATURE_INCOMPAT_ENCRYPT)
1627 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1628                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1629                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1630                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1631                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1632                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1633                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1634                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
1635                                          EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
1636                                          EXT4_FEATURE_RO_COMPAT_QUOTA)
1637
1638 /*
1639  * Default values for user and/or group using reserved blocks
1640  */
1641 #define EXT4_DEF_RESUID         0
1642 #define EXT4_DEF_RESGID         0
1643
1644 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1645
1646 /*
1647  * Default mount options
1648  */
1649 #define EXT4_DEFM_DEBUG         0x0001
1650 #define EXT4_DEFM_BSDGROUPS     0x0002
1651 #define EXT4_DEFM_XATTR_USER    0x0004
1652 #define EXT4_DEFM_ACL           0x0008
1653 #define EXT4_DEFM_UID16         0x0010
1654 #define EXT4_DEFM_JMODE         0x0060
1655 #define EXT4_DEFM_JMODE_DATA    0x0020
1656 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1657 #define EXT4_DEFM_JMODE_WBACK   0x0060
1658 #define EXT4_DEFM_NOBARRIER     0x0100
1659 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1660 #define EXT4_DEFM_DISCARD       0x0400
1661 #define EXT4_DEFM_NODELALLOC    0x0800
1662
1663 /*
1664  * Default journal batch times
1665  */
1666 #define EXT4_DEF_MIN_BATCH_TIME 0
1667 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1668
1669 /*
1670  * Minimum number of groups in a flexgroup before we separate out
1671  * directories into the first block group of a flexgroup
1672  */
1673 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1674
1675 /*
1676  * Structure of a directory entry
1677  */
1678 #define EXT4_NAME_LEN 255
1679
1680 struct ext4_dir_entry {
1681         __le32  inode;                  /* Inode number */
1682         __le16  rec_len;                /* Directory entry length */
1683         __le16  name_len;               /* Name length */
1684         char    name[EXT4_NAME_LEN];    /* File name */
1685 };
1686
1687 /*
1688  * The new version of the directory entry.  Since EXT4 structures are
1689  * stored in intel byte order, and the name_len field could never be
1690  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1691  * file_type field.
1692  */
1693 struct ext4_dir_entry_2 {
1694         __le32  inode;                  /* Inode number */
1695         __le16  rec_len;                /* Directory entry length */
1696         __u8    name_len;               /* Name length */
1697         __u8    file_type;
1698         char    name[EXT4_NAME_LEN];    /* File name */
1699 };
1700
1701 /*
1702  * This is a bogus directory entry at the end of each leaf block that
1703  * records checksums.
1704  */
1705 struct ext4_dir_entry_tail {
1706         __le32  det_reserved_zero1;     /* Pretend to be unused */
1707         __le16  det_rec_len;            /* 12 */
1708         __u8    det_reserved_zero2;     /* Zero name length */
1709         __u8    det_reserved_ft;        /* 0xDE, fake file type */
1710         __le32  det_checksum;           /* crc32c(uuid+inum+dirblock) */
1711 };
1712
1713 #define EXT4_DIRENT_TAIL(block, blocksize) \
1714         ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
1715                                         ((blocksize) - \
1716                                          sizeof(struct ext4_dir_entry_tail))))
1717
1718 /*
1719  * Ext4 directory file types.  Only the low 3 bits are used.  The
1720  * other bits are reserved for now.
1721  */
1722 #define EXT4_FT_UNKNOWN         0
1723 #define EXT4_FT_REG_FILE        1
1724 #define EXT4_FT_DIR             2
1725 #define EXT4_FT_CHRDEV          3
1726 #define EXT4_FT_BLKDEV          4
1727 #define EXT4_FT_FIFO            5
1728 #define EXT4_FT_SOCK            6
1729 #define EXT4_FT_SYMLINK         7
1730
1731 #define EXT4_FT_MAX             8
1732
1733 #define EXT4_FT_DIR_CSUM        0xDE
1734
1735 /*
1736  * EXT4_DIR_PAD defines the directory entries boundaries
1737  *
1738  * NOTE: It must be a multiple of 4
1739  */
1740 #define EXT4_DIR_PAD                    4
1741 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1742 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1743                                          ~EXT4_DIR_ROUND)
1744 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1745
1746 /*
1747  * If we ever get support for fs block sizes > page_size, we'll need
1748  * to remove the #if statements in the next two functions...
1749  */
1750 static inline unsigned int
1751 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1752 {
1753         unsigned len = le16_to_cpu(dlen);
1754
1755 #if (PAGE_CACHE_SIZE >= 65536)
1756         if (len == EXT4_MAX_REC_LEN || len == 0)
1757                 return blocksize;
1758         return (len & 65532) | ((len & 3) << 16);
1759 #else
1760         return len;
1761 #endif
1762 }
1763
1764 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1765 {
1766         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1767                 BUG();
1768 #if (PAGE_CACHE_SIZE >= 65536)
1769         if (len < 65536)
1770                 return cpu_to_le16(len);
1771         if (len == blocksize) {
1772                 if (blocksize == 65536)
1773                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1774                 else
1775                         return cpu_to_le16(0);
1776         }
1777         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1778 #else
1779         return cpu_to_le16(len);
1780 #endif
1781 }
1782
1783 /*
1784  * Hash Tree Directory indexing
1785  * (c) Daniel Phillips, 2001
1786  */
1787
1788 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1789                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1790                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1791 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1792 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1793
1794 /* Legal values for the dx_root hash_version field: */
1795
1796 #define DX_HASH_LEGACY          0
1797 #define DX_HASH_HALF_MD4        1
1798 #define DX_HASH_TEA             2
1799 #define DX_HASH_LEGACY_UNSIGNED 3
1800 #define DX_HASH_HALF_MD4_UNSIGNED       4
1801 #define DX_HASH_TEA_UNSIGNED            5
1802
1803 static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
1804                               const void *address, unsigned int length)
1805 {
1806         struct {
1807                 struct shash_desc shash;
1808                 char ctx[4];
1809         } desc;
1810         int err;
1811
1812         BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
1813
1814         desc.shash.tfm = sbi->s_chksum_driver;
1815         desc.shash.flags = 0;
1816         *(u32 *)desc.ctx = crc;
1817
1818         err = crypto_shash_update(&desc.shash, address, length);
1819         BUG_ON(err);
1820
1821         return *(u32 *)desc.ctx;
1822 }
1823
1824 #ifdef __KERNEL__
1825
1826 /* hash info structure used by the directory hash */
1827 struct dx_hash_info
1828 {
1829         u32             hash;
1830         u32             minor_hash;
1831         int             hash_version;
1832         u32             *seed;
1833 };
1834
1835
1836 /* 32 and 64 bit signed EOF for dx directories */
1837 #define EXT4_HTREE_EOF_32BIT   ((1UL  << (32 - 1)) - 1)
1838 #define EXT4_HTREE_EOF_64BIT   ((1ULL << (64 - 1)) - 1)
1839
1840
1841 /*
1842  * Control parameters used by ext4_htree_next_block
1843  */
1844 #define HASH_NB_ALWAYS          1
1845
1846
1847 /*
1848  * Describe an inode's exact location on disk and in memory
1849  */
1850 struct ext4_iloc
1851 {
1852         struct buffer_head *bh;
1853         unsigned long offset;
1854         ext4_group_t block_group;
1855 };
1856
1857 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1858 {
1859         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1860 }
1861
1862 /*
1863  * This structure is stuffed into the struct file's private_data field
1864  * for directories.  It is where we put information so that we can do
1865  * readdir operations in hash tree order.
1866  */
1867 struct dir_private_info {
1868         struct rb_root  root;
1869         struct rb_node  *curr_node;
1870         struct fname    *extra_fname;
1871         loff_t          last_pos;
1872         __u32           curr_hash;
1873         __u32           curr_minor_hash;
1874         __u32           next_hash;
1875 };
1876
1877 /* calculate the first block number of the group */
1878 static inline ext4_fsblk_t
1879 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1880 {
1881         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1882                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1883 }
1884
1885 /*
1886  * Special error return code only used by dx_probe() and its callers.
1887  */
1888 #define ERR_BAD_DX_DIR  (-(MAX_ERRNO - 1))
1889
1890 /*
1891  * Timeout and state flag for lazy initialization inode thread.
1892  */
1893 #define EXT4_DEF_LI_WAIT_MULT                   10
1894 #define EXT4_DEF_LI_MAX_START_DELAY             5
1895 #define EXT4_LAZYINIT_QUIT                      0x0001
1896 #define EXT4_LAZYINIT_RUNNING                   0x0002
1897
1898 /*
1899  * Lazy inode table initialization info
1900  */
1901 struct ext4_lazy_init {
1902         unsigned long           li_state;
1903         struct list_head        li_request_list;
1904         struct mutex            li_list_mtx;
1905 };
1906
1907 struct ext4_li_request {
1908         struct super_block      *lr_super;
1909         struct ext4_sb_info     *lr_sbi;
1910         ext4_group_t            lr_next_group;
1911         struct list_head        lr_request;
1912         unsigned long           lr_next_sched;
1913         unsigned long           lr_timeout;
1914 };
1915
1916 struct ext4_features {
1917         struct kobject f_kobj;
1918         struct completion f_kobj_unregister;
1919 };
1920
1921 /*
1922  * This structure will be used for multiple mount protection. It will be
1923  * written into the block number saved in the s_mmp_block field in the
1924  * superblock. Programs that check MMP should assume that if
1925  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1926  * to use the filesystem, regardless of how old the timestamp is.
1927  */
1928 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
1929 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1930 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
1931 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
1932
1933 struct mmp_struct {
1934         __le32  mmp_magic;              /* Magic number for MMP */
1935         __le32  mmp_seq;                /* Sequence no. updated periodically */
1936
1937         /*
1938          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1939          * purposes and do not affect the correctness of the algorithm
1940          */
1941         __le64  mmp_time;               /* Time last updated */
1942         char    mmp_nodename[64];       /* Node which last updated MMP block */
1943         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
1944
1945         /*
1946          * mmp_check_interval is used to verify if the MMP block has been
1947          * updated on the block device. The value is updated based on the
1948          * maximum time to write the MMP block during an update cycle.
1949          */
1950         __le16  mmp_check_interval;
1951
1952         __le16  mmp_pad1;
1953         __le32  mmp_pad2[226];
1954         __le32  mmp_checksum;           /* crc32c(uuid+mmp_block) */
1955 };
1956
1957 /* arguments passed to the mmp thread */
1958 struct mmpd_data {
1959         struct buffer_head *bh; /* bh from initial read_mmp_block() */
1960         struct super_block *sb;  /* super block of the fs */
1961 };
1962
1963 /*
1964  * Check interval multiplier
1965  * The MMP block is written every update interval and initially checked every
1966  * update interval x the multiplier (the value is then adapted based on the
1967  * write latency). The reason is that writes can be delayed under load and we
1968  * don't want readers to incorrectly assume that the filesystem is no longer
1969  * in use.
1970  */
1971 #define EXT4_MMP_CHECK_MULT             2UL
1972
1973 /*
1974  * Minimum interval for MMP checking in seconds.
1975  */
1976 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
1977
1978 /*
1979  * Maximum interval for MMP checking in seconds.
1980  */
1981 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
1982
1983 /*
1984  * Function prototypes
1985  */
1986
1987 /*
1988  * Ok, these declarations are also in <linux/kernel.h> but none of the
1989  * ext4 source programs needs to include it so they are duplicated here.
1990  */
1991 # define NORET_TYPE     /**/
1992 # define ATTRIB_NORET   __attribute__((noreturn))
1993 # define NORET_AND      noreturn,
1994
1995 /* bitmap.c */
1996 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
1997 void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1998                                 struct ext4_group_desc *gdp,
1999                                 struct buffer_head *bh, int sz);
2000 int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2001                                   struct ext4_group_desc *gdp,
2002                                   struct buffer_head *bh, int sz);
2003 void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
2004                                 struct ext4_group_desc *gdp,
2005                                 struct buffer_head *bh);
2006 int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2007                                   struct ext4_group_desc *gdp,
2008                                   struct buffer_head *bh);
2009
2010 /* balloc.c */
2011 extern void ext4_get_group_no_and_offset(struct super_block *sb,
2012                                          ext4_fsblk_t blocknr,
2013                                          ext4_group_t *blockgrpp,
2014                                          ext4_grpblk_t *offsetp);
2015 extern ext4_group_t ext4_get_group_number(struct super_block *sb,
2016                                           ext4_fsblk_t block);
2017
2018 extern unsigned int ext4_block_group(struct super_block *sb,
2019                         ext4_fsblk_t blocknr);
2020 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
2021                         ext4_fsblk_t blocknr);
2022 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
2023 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
2024                         ext4_group_t group);
2025 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
2026                                          ext4_fsblk_t goal,
2027                                          unsigned int flags,
2028                                          unsigned long *count,
2029                                          int *errp);
2030 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
2031                                     s64 nclusters, unsigned int flags);
2032 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
2033 extern void ext4_check_blocks_bitmap(struct super_block *);
2034 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
2035                                                     ext4_group_t block_group,
2036                                                     struct buffer_head ** bh);
2037 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
2038
2039 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
2040                                                 ext4_group_t block_group);
2041 extern int ext4_wait_block_bitmap(struct super_block *sb,
2042                                   ext4_group_t block_group,
2043                                   struct buffer_head *bh);
2044 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
2045                                                   ext4_group_t block_group);
2046 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
2047                                               ext4_group_t block_group,
2048                                               struct ext4_group_desc *gdp);
2049 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
2050
2051 /* crypto_policy.c */
2052 int ext4_is_child_context_consistent_with_parent(struct inode *parent,
2053                                                  struct inode *child);
2054 int ext4_inherit_context(struct inode *parent, struct inode *child);
2055 void ext4_to_hex(char *dst, char *src, size_t src_size);
2056 int ext4_process_policy(const struct ext4_encryption_policy *policy,
2057                         struct inode *inode);
2058 int ext4_get_policy(struct inode *inode,
2059                     struct ext4_encryption_policy *policy);
2060
2061 /* crypto.c */
2062 bool ext4_valid_contents_enc_mode(uint32_t mode);
2063 uint32_t ext4_validate_encryption_key_size(uint32_t mode, uint32_t size);
2064 extern struct workqueue_struct *ext4_read_workqueue;
2065 struct ext4_crypto_ctx *ext4_get_crypto_ctx(struct inode *inode);
2066 void ext4_release_crypto_ctx(struct ext4_crypto_ctx *ctx);
2067 void ext4_restore_control_page(struct page *data_page);
2068 struct page *ext4_encrypt(struct inode *inode,
2069                           struct page *plaintext_page);
2070 int ext4_decrypt(struct ext4_crypto_ctx *ctx, struct page *page);
2071 int ext4_decrypt_one(struct inode *inode, struct page *page);
2072 int ext4_encrypted_zeroout(struct inode *inode, struct ext4_extent *ex);
2073
2074 #ifdef CONFIG_EXT4_FS_ENCRYPTION
2075 int ext4_init_crypto(void);
2076 void ext4_exit_crypto(void);
2077 static inline int ext4_sb_has_crypto(struct super_block *sb)
2078 {
2079         return EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
2080 }
2081 #else
2082 static inline int ext4_init_crypto(void) { return 0; }
2083 static inline void ext4_exit_crypto(void) { }
2084 static inline int ext4_sb_has_crypto(struct super_block *sb)
2085 {
2086         return 0;
2087 }
2088 #endif
2089
2090 /* crypto_fname.c */
2091 bool ext4_valid_filenames_enc_mode(uint32_t mode);
2092 u32 ext4_fname_crypto_round_up(u32 size, u32 blksize);
2093 int ext4_fname_crypto_alloc_buffer(struct ext4_fname_crypto_ctx *ctx,
2094                                    u32 ilen, struct ext4_str *crypto_str);
2095 int _ext4_fname_disk_to_usr(struct ext4_fname_crypto_ctx *ctx,
2096                             const struct ext4_str *iname,
2097                             struct ext4_str *oname);
2098 int ext4_fname_disk_to_usr(struct ext4_fname_crypto_ctx *ctx,
2099                            const struct ext4_dir_entry_2 *de,
2100                            struct ext4_str *oname);
2101 int ext4_fname_usr_to_disk(struct ext4_fname_crypto_ctx *ctx,
2102                            const struct qstr *iname,
2103                            struct ext4_str *oname);
2104 int ext4_fname_usr_to_hash(struct ext4_fname_crypto_ctx *ctx,
2105                            const struct qstr *iname,
2106                            struct dx_hash_info *hinfo);
2107 int ext4_fname_disk_to_hash(struct ext4_fname_crypto_ctx *ctx,
2108                             const struct ext4_dir_entry_2 *de,
2109                             struct dx_hash_info *hinfo);
2110 int ext4_fname_crypto_namelen_on_disk(struct ext4_fname_crypto_ctx *ctx,
2111                                       u32 namelen);
2112
2113 #ifdef CONFIG_EXT4_FS_ENCRYPTION
2114 void ext4_put_fname_crypto_ctx(struct ext4_fname_crypto_ctx **ctx);
2115 struct ext4_fname_crypto_ctx *ext4_get_fname_crypto_ctx(struct inode *inode,
2116                                                         u32 max_len);
2117 void ext4_fname_crypto_free_buffer(struct ext4_str *crypto_str);
2118 #else
2119 static inline
2120 void ext4_put_fname_crypto_ctx(struct ext4_fname_crypto_ctx **ctx) { }
2121 static inline
2122 struct ext4_fname_crypto_ctx *ext4_get_fname_crypto_ctx(struct inode *inode,
2123                                                         u32 max_len)
2124 {
2125         return NULL;
2126 }
2127 static inline void ext4_fname_crypto_free_buffer(struct ext4_str *p) { }
2128 #endif
2129
2130
2131 /* crypto_key.c */
2132 int ext4_generate_encryption_key(struct inode *inode);
2133
2134 #ifdef CONFIG_EXT4_FS_ENCRYPTION
2135 int ext4_has_encryption_key(struct inode *inode);
2136 #else
2137 static inline int ext4_has_encryption_key(struct inode *inode)
2138 {
2139         return 0;
2140 }
2141 #endif
2142
2143
2144 /* dir.c */
2145 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
2146                                   struct file *,
2147                                   struct ext4_dir_entry_2 *,
2148                                   struct buffer_head *, char *, int,
2149                                   unsigned int);
2150 #define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset)      \
2151         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
2152                                         (de), (bh), (buf), (size), (offset)))
2153 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
2154                                 __u32 minor_hash,
2155                                 struct ext4_dir_entry_2 *dirent,
2156                                 struct ext4_str *ent_name);
2157 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
2158 extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
2159                              struct buffer_head *bh,
2160                              void *buf, int buf_size,
2161                              const char *name, int namelen,
2162                              struct ext4_dir_entry_2 **dest_de);
2163 int ext4_insert_dentry(struct inode *dir,
2164                         struct inode *inode,
2165                         struct ext4_dir_entry_2 *de,
2166                         int buf_size,
2167                        const struct qstr *iname,
2168                         const char *name, int namelen);
2169 static inline void ext4_update_dx_flag(struct inode *inode)
2170 {
2171         if (!EXT4_HAS_COMPAT_FEATURE(inode->i_sb,
2172                                      EXT4_FEATURE_COMPAT_DIR_INDEX))
2173                 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
2174 }
2175 static unsigned char ext4_filetype_table[] = {
2176         DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
2177 };
2178
2179 static inline  unsigned char get_dtype(struct super_block *sb, int filetype)
2180 {
2181         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE) ||
2182             (filetype >= EXT4_FT_MAX))
2183                 return DT_UNKNOWN;
2184
2185         return ext4_filetype_table[filetype];
2186 }
2187 extern int ext4_check_all_de(struct inode *dir, struct buffer_head *bh,
2188                              void *buf, int buf_size);
2189
2190 /* fsync.c */
2191 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2192
2193 /* hash.c */
2194 extern int ext4fs_dirhash(const char *name, int len, struct
2195                           dx_hash_info *hinfo);
2196
2197 /* ialloc.c */
2198 extern struct inode *__ext4_new_inode(handle_t *, struct inode *, umode_t,
2199                                       const struct qstr *qstr, __u32 goal,
2200                                       uid_t *owner, int handle_type,
2201                                       unsigned int line_no, int nblocks);
2202
2203 #define ext4_new_inode(handle, dir, mode, qstr, goal, owner) \
2204         __ext4_new_inode((handle), (dir), (mode), (qstr), (goal), (owner), \
2205                          0, 0, 0)
2206 #define ext4_new_inode_start_handle(dir, mode, qstr, goal, owner, \
2207                                     type, nblocks)                  \
2208         __ext4_new_inode(NULL, (dir), (mode), (qstr), (goal), (owner), \
2209                          (type), __LINE__, (nblocks))
2210
2211
2212 extern void ext4_free_inode(handle_t *, struct inode *);
2213 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2214 extern unsigned long ext4_count_free_inodes(struct super_block *);
2215 extern unsigned long ext4_count_dirs(struct super_block *);
2216 extern void ext4_check_inodes_bitmap(struct super_block *);
2217 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2218 extern int ext4_init_inode_table(struct super_block *sb,
2219                                  ext4_group_t group, int barrier);
2220 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2221
2222 /* mballoc.c */
2223 extern long ext4_mb_stats;
2224 extern long ext4_mb_max_to_scan;
2225 extern int ext4_mb_init(struct super_block *);
2226 extern int ext4_mb_release(struct super_block *);
2227 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2228                                 struct ext4_allocation_request *, int *);
2229 extern int ext4_mb_reserve_blocks(struct super_block *, int);
2230 extern void ext4_discard_preallocations(struct inode *);
2231 extern int __init ext4_init_mballoc(void);
2232 extern void ext4_exit_mballoc(void);
2233 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2234                              struct buffer_head *bh, ext4_fsblk_t block,
2235                              unsigned long count, int flags);
2236 extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2237                                    ext4_group_t ngroups);
2238 extern int ext4_mb_add_groupinfo(struct super_block *sb,
2239                 ext4_group_t i, struct ext4_group_desc *desc);
2240 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2241                                 ext4_fsblk_t block, unsigned long count);
2242 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2243
2244 /* inode.c */
2245 int ext4_inode_is_fast_symlink(struct inode *inode);
2246 struct buffer_head *ext4_getblk(handle_t *, struct inode *, ext4_lblk_t, int);
2247 struct buffer_head *ext4_bread(handle_t *, struct inode *, ext4_lblk_t, int);
2248 int ext4_get_block_write(struct inode *inode, sector_t iblock,
2249                          struct buffer_head *bh_result, int create);
2250 int ext4_get_block(struct inode *inode, sector_t iblock,
2251                                 struct buffer_head *bh_result, int create);
2252 int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2253                            struct buffer_head *bh, int create);
2254 int ext4_walk_page_buffers(handle_t *handle,
2255                            struct buffer_head *head,
2256                            unsigned from,
2257                            unsigned to,
2258                            int *partial,
2259                            int (*fn)(handle_t *handle,
2260                                      struct buffer_head *bh));
2261 int do_journal_get_write_access(handle_t *handle,
2262                                 struct buffer_head *bh);
2263 #define FALL_BACK_TO_NONDELALLOC 1
2264 #define CONVERT_INLINE_DATA      2
2265
2266 extern struct inode *ext4_iget(struct super_block *, unsigned long);
2267 extern struct inode *ext4_iget_normal(struct super_block *, unsigned long);
2268 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
2269 extern int  ext4_setattr(struct dentry *, struct iattr *);
2270 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
2271                                 struct kstat *stat);
2272 extern void ext4_evict_inode(struct inode *);
2273 extern void ext4_clear_inode(struct inode *);
2274 extern int  ext4_sync_inode(handle_t *, struct inode *);
2275 extern void ext4_dirty_inode(struct inode *, int);
2276 extern int ext4_change_inode_journal_flag(struct inode *, int);
2277 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
2278 extern int ext4_inode_attach_jinode(struct inode *inode);
2279 extern int ext4_can_truncate(struct inode *inode);
2280 extern void ext4_truncate(struct inode *);
2281 extern int ext4_punch_hole(struct inode *inode, loff_t offset, loff_t length);
2282 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
2283 extern void ext4_set_inode_flags(struct inode *);
2284 extern void ext4_get_inode_flags(struct ext4_inode_info *);
2285 extern int ext4_alloc_da_blocks(struct inode *inode);
2286 extern void ext4_set_aops(struct inode *inode);
2287 extern int ext4_writepage_trans_blocks(struct inode *);
2288 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
2289 extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
2290                              loff_t lstart, loff_t lend);
2291 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2292 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
2293 extern void ext4_da_update_reserve_space(struct inode *inode,
2294                                         int used, int quota_claim);
2295
2296 /* indirect.c */
2297 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
2298                                 struct ext4_map_blocks *map, int flags);
2299 extern ssize_t ext4_ind_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
2300                                   loff_t offset);
2301 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
2302 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
2303 extern void ext4_ind_truncate(handle_t *, struct inode *inode);
2304 extern int ext4_ind_remove_space(handle_t *handle, struct inode *inode,
2305                                  ext4_lblk_t start, ext4_lblk_t end);
2306
2307 /* ioctl.c */
2308 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
2309 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
2310
2311 /* migrate.c */
2312 extern int ext4_ext_migrate(struct inode *);
2313 extern int ext4_ind_migrate(struct inode *inode);
2314
2315 /* namei.c */
2316 extern int ext4_dirent_csum_verify(struct inode *inode,
2317                                    struct ext4_dir_entry *dirent);
2318 extern int ext4_orphan_add(handle_t *, struct inode *);
2319 extern int ext4_orphan_del(handle_t *, struct inode *);
2320 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
2321                                 __u32 start_minor_hash, __u32 *next_hash);
2322 extern int search_dir(struct buffer_head *bh,
2323                       char *search_buf,
2324                       int buf_size,
2325                       struct inode *dir,
2326                       const struct qstr *d_name,
2327                       unsigned int offset,
2328                       struct ext4_dir_entry_2 **res_dir);
2329 extern int ext4_generic_delete_entry(handle_t *handle,
2330                                      struct inode *dir,
2331                                      struct ext4_dir_entry_2 *de_del,
2332                                      struct buffer_head *bh,
2333                                      void *entry_buf,
2334                                      int buf_size,
2335                                      int csum_size);
2336 extern int ext4_empty_dir(struct inode *inode);
2337
2338 /* resize.c */
2339 extern int ext4_group_add(struct super_block *sb,
2340                                 struct ext4_new_group_data *input);
2341 extern int ext4_group_extend(struct super_block *sb,
2342                                 struct ext4_super_block *es,
2343                                 ext4_fsblk_t n_blocks_count);
2344 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
2345
2346 /* super.c */
2347 extern int ext4_calculate_overhead(struct super_block *sb);
2348 extern void ext4_superblock_csum_set(struct super_block *sb);
2349 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
2350 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
2351 extern int ext4_alloc_flex_bg_array(struct super_block *sb,
2352                                     ext4_group_t ngroup);
2353 extern const char *ext4_decode_error(struct super_block *sb, int errno,
2354                                      char nbuf[16]);
2355
2356 extern __printf(4, 5)
2357 void __ext4_error(struct super_block *, const char *, unsigned int,
2358                   const char *, ...);
2359 extern __printf(5, 6)
2360 void __ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
2361                       const char *, ...);
2362 extern __printf(5, 6)
2363 void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
2364                      const char *, ...);
2365 extern void __ext4_std_error(struct super_block *, const char *,
2366                              unsigned int, int);
2367 extern __printf(4, 5)
2368 void __ext4_abort(struct super_block *, const char *, unsigned int,
2369                   const char *, ...);
2370 extern __printf(4, 5)
2371 void __ext4_warning(struct super_block *, const char *, unsigned int,
2372                     const char *, ...);
2373 extern __printf(3, 4)
2374 void __ext4_msg(struct super_block *, const char *, const char *, ...);
2375 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
2376                            const char *, unsigned int, const char *);
2377 extern __printf(7, 8)
2378 void __ext4_grp_locked_error(const char *, unsigned int,
2379                              struct super_block *, ext4_group_t,
2380                              unsigned long, ext4_fsblk_t,
2381                              const char *, ...);
2382
2383 #ifdef CONFIG_PRINTK
2384
2385 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2386         __ext4_error_inode(inode, func, line, block, fmt, ##__VA_ARGS__)
2387 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2388         __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__)
2389 #define ext4_error(sb, fmt, ...)                                        \
2390         __ext4_error(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2391 #define ext4_abort(sb, fmt, ...)                                        \
2392         __ext4_abort(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2393 #define ext4_warning(sb, fmt, ...)                                      \
2394         __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2395 #define ext4_msg(sb, level, fmt, ...)                           \
2396         __ext4_msg(sb, level, fmt, ##__VA_ARGS__)
2397 #define dump_mmp_msg(sb, mmp, msg)                                      \
2398         __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg)
2399 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2400         __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \
2401                                 fmt, ##__VA_ARGS__)
2402
2403 #else
2404
2405 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2406 do {                                                                    \
2407         no_printk(fmt, ##__VA_ARGS__);                                  \
2408         __ext4_error_inode(inode, "", 0, block, " ");                   \
2409 } while (0)
2410 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2411 do {                                                                    \
2412         no_printk(fmt, ##__VA_ARGS__);                                  \
2413         __ext4_error_file(file, "", 0, block, " ");                     \
2414 } while (0)
2415 #define ext4_error(sb, fmt, ...)                                        \
2416 do {                                                                    \
2417         no_printk(fmt, ##__VA_ARGS__);                                  \
2418         __ext4_error(sb, "", 0, " ");                                   \
2419 } while (0)
2420 #define ext4_abort(sb, fmt, ...)                                        \
2421 do {                                                                    \
2422         no_printk(fmt, ##__VA_ARGS__);                                  \
2423         __ext4_abort(sb, "", 0, " ");                                   \
2424 } while (0)
2425 #define ext4_warning(sb, fmt, ...)                                      \
2426 do {                                                                    \
2427         no_printk(fmt, ##__VA_ARGS__);                                  \
2428         __ext4_warning(sb, "", 0, " ");                                 \
2429 } while (0)
2430 #define ext4_msg(sb, level, fmt, ...)                                   \
2431 do {                                                                    \
2432         no_printk(fmt, ##__VA_ARGS__);                                  \
2433         __ext4_msg(sb, "", " ");                                        \
2434 } while (0)
2435 #define dump_mmp_msg(sb, mmp, msg)                                      \
2436         __dump_mmp_msg(sb, mmp, "", 0, "")
2437 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2438 do {                                                                    \
2439         no_printk(fmt, ##__VA_ARGS__);                          \
2440         __ext4_grp_locked_error("", 0, sb, grp, ino, block, " ");       \
2441 } while (0)
2442
2443 #endif
2444
2445 extern void ext4_update_dynamic_rev(struct super_block *sb);
2446 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
2447                                         __u32 compat);
2448 extern int ext4_update_rocompat_feature(handle_t *handle,
2449                                         struct super_block *sb, __u32 rocompat);
2450 extern int ext4_update_incompat_feature(handle_t *handle,
2451                                         struct super_block *sb, __u32 incompat);
2452 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
2453                                       struct ext4_group_desc *bg);
2454 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
2455                                       struct ext4_group_desc *bg);
2456 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
2457                                      struct ext4_group_desc *bg);
2458 extern __u32 ext4_free_group_clusters(struct super_block *sb,
2459                                       struct ext4_group_desc *bg);
2460 extern __u32 ext4_free_inodes_count(struct super_block *sb,
2461                                  struct ext4_group_desc *bg);
2462 extern __u32 ext4_used_dirs_count(struct super_block *sb,
2463                                 struct ext4_group_desc *bg);
2464 extern __u32 ext4_itable_unused_count(struct super_block *sb,
2465                                    struct ext4_group_desc *bg);
2466 extern void ext4_block_bitmap_set(struct super_block *sb,
2467                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2468 extern void ext4_inode_bitmap_set(struct super_block *sb,
2469                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2470 extern void ext4_inode_table_set(struct super_block *sb,
2471                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
2472 extern void ext4_free_group_clusters_set(struct super_block *sb,
2473                                          struct ext4_group_desc *bg,
2474                                          __u32 count);
2475 extern void ext4_free_inodes_set(struct super_block *sb,
2476                                 struct ext4_group_desc *bg, __u32 count);
2477 extern void ext4_used_dirs_set(struct super_block *sb,
2478                                 struct ext4_group_desc *bg, __u32 count);
2479 extern void ext4_itable_unused_set(struct super_block *sb,
2480                                    struct ext4_group_desc *bg, __u32 count);
2481 extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
2482                                        struct ext4_group_desc *gdp);
2483 extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
2484                                      struct ext4_group_desc *gdp);
2485 extern int ext4_register_li_request(struct super_block *sb,
2486                                     ext4_group_t first_not_zeroed);
2487
2488 static inline int ext4_has_group_desc_csum(struct super_block *sb)
2489 {
2490         return EXT4_HAS_RO_COMPAT_FEATURE(sb,
2491                                           EXT4_FEATURE_RO_COMPAT_GDT_CSUM) ||
2492                (EXT4_SB(sb)->s_chksum_driver != NULL);
2493 }
2494
2495 static inline int ext4_has_metadata_csum(struct super_block *sb)
2496 {
2497         WARN_ON_ONCE(EXT4_HAS_RO_COMPAT_FEATURE(sb,
2498                         EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
2499                      !EXT4_SB(sb)->s_chksum_driver);
2500
2501         return (EXT4_SB(sb)->s_chksum_driver != NULL);
2502 }
2503 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2504 {
2505         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2506                 le32_to_cpu(es->s_blocks_count_lo);
2507 }
2508
2509 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2510 {
2511         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2512                 le32_to_cpu(es->s_r_blocks_count_lo);
2513 }
2514
2515 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2516 {
2517         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2518                 le32_to_cpu(es->s_free_blocks_count_lo);
2519 }
2520
2521 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2522                                          ext4_fsblk_t blk)
2523 {
2524         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2525         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2526 }
2527
2528 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2529                                               ext4_fsblk_t blk)
2530 {
2531         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2532         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2533 }
2534
2535 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2536                                            ext4_fsblk_t blk)
2537 {
2538         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2539         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2540 }
2541
2542 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2543 {
2544         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2545                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2546                         le32_to_cpu(raw_inode->i_size_lo);
2547         else
2548                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2549 }
2550
2551 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2552 {
2553         raw_inode->i_size_lo = cpu_to_le32(i_size);
2554         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2555 }
2556
2557 static inline
2558 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2559                                             ext4_group_t group)
2560 {
2561          struct ext4_group_info ***grp_info;
2562          long indexv, indexh;
2563          BUG_ON(group >= EXT4_SB(sb)->s_groups_count);
2564          grp_info = EXT4_SB(sb)->s_group_info;
2565          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2566          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2567          return grp_info[indexv][indexh];
2568 }
2569
2570 /*
2571  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2572  * the locking protocol documented in the comments of ext4_group_add()
2573  * in resize.c
2574  */
2575 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2576 {
2577         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2578
2579         smp_rmb();
2580         return ngroups;
2581 }
2582
2583 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2584                                              ext4_group_t block_group)
2585 {
2586         return block_group >> sbi->s_log_groups_per_flex;
2587 }
2588
2589 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2590 {
2591         return 1 << sbi->s_log_groups_per_flex;
2592 }
2593
2594 #define ext4_std_error(sb, errno)                               \
2595 do {                                                            \
2596         if ((errno))                                            \
2597                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2598 } while (0)
2599
2600 #ifdef CONFIG_SMP
2601 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2602  * counters. So we need to make sure we have free clusters more
2603  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2604  */
2605 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2606 #else
2607 #define EXT4_FREECLUSTERS_WATERMARK 0
2608 #endif
2609
2610 /* Update i_disksize. Requires i_mutex to avoid races with truncate */
2611 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2612 {
2613         WARN_ON_ONCE(S_ISREG(inode->i_mode) &&
2614                      !mutex_is_locked(&inode->i_mutex));
2615         down_write(&EXT4_I(inode)->i_data_sem);
2616         if (newsize > EXT4_I(inode)->i_disksize)
2617                 EXT4_I(inode)->i_disksize = newsize;
2618         up_write(&EXT4_I(inode)->i_data_sem);
2619 }
2620
2621 /* Update i_size, i_disksize. Requires i_mutex to avoid races with truncate */
2622 static inline int ext4_update_inode_size(struct inode *inode, loff_t newsize)
2623 {
2624         int changed = 0;
2625
2626         if (newsize > inode->i_size) {
2627                 i_size_write(inode, newsize);
2628                 changed = 1;
2629         }
2630         if (newsize > EXT4_I(inode)->i_disksize) {
2631                 ext4_update_i_disksize(inode, newsize);
2632                 changed |= 2;
2633         }
2634         return changed;
2635 }
2636
2637 struct ext4_group_info {
2638         unsigned long   bb_state;
2639         struct rb_root  bb_free_root;
2640         ext4_grpblk_t   bb_first_free;  /* first free block */
2641         ext4_grpblk_t   bb_free;        /* total free blocks */
2642         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2643         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2644         struct          list_head bb_prealloc_list;
2645 #ifdef DOUBLE_CHECK
2646         void            *bb_bitmap;
2647 #endif
2648         struct rw_semaphore alloc_sem;
2649         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2650                                          * regions, index is order.
2651                                          * bb_counters[3] = 5 means
2652                                          * 5 free 8-block regions. */
2653 };
2654
2655 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2656 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2657 #define EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT     2
2658 #define EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT     3
2659
2660 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2661         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2662 #define EXT4_MB_GRP_BBITMAP_CORRUPT(grp)        \
2663         (test_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
2664 #define EXT4_MB_GRP_IBITMAP_CORRUPT(grp)        \
2665         (test_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
2666
2667 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2668         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2669 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2670         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2671 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2672         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2673
2674 #define EXT4_MAX_CONTENTION             8
2675 #define EXT4_CONTENTION_THRESHOLD       2
2676
2677 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2678                                               ext4_group_t group)
2679 {
2680         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2681 }
2682
2683 /*
2684  * Returns true if the filesystem is busy enough that attempts to
2685  * access the block group locks has run into contention.
2686  */
2687 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2688 {
2689         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2690 }
2691
2692 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2693 {
2694         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2695         if (spin_trylock(lock))
2696                 /*
2697                  * We're able to grab the lock right away, so drop the
2698                  * lock contention counter.
2699                  */
2700                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2701         else {
2702                 /*
2703                  * The lock is busy, so bump the contention counter,
2704                  * and then wait on the spin lock.
2705                  */
2706                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2707                                   EXT4_MAX_CONTENTION);
2708                 spin_lock(lock);
2709         }
2710 }
2711
2712 static inline void ext4_unlock_group(struct super_block *sb,
2713                                         ext4_group_t group)
2714 {
2715         spin_unlock(ext4_group_lock_ptr(sb, group));
2716 }
2717
2718 /*
2719  * Block validity checking
2720  */
2721 #define ext4_check_indirect_blockref(inode, bh)                         \
2722         ext4_check_blockref(__func__, __LINE__, inode,                  \
2723                             (__le32 *)(bh)->b_data,                     \
2724                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2725
2726 #define ext4_ind_check_inode(inode)                                     \
2727         ext4_check_blockref(__func__, __LINE__, inode,                  \
2728                             EXT4_I(inode)->i_data,                      \
2729                             EXT4_NDIR_BLOCKS)
2730
2731 /*
2732  * Inodes and files operations
2733  */
2734
2735 /* dir.c */
2736 extern const struct file_operations ext4_dir_operations;
2737
2738 /* file.c */
2739 extern const struct inode_operations ext4_file_inode_operations;
2740 extern const struct file_operations ext4_file_operations;
2741 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2742
2743 /* inline.c */
2744 extern int ext4_get_max_inline_size(struct inode *inode);
2745 extern int ext4_find_inline_data_nolock(struct inode *inode);
2746 extern int ext4_init_inline_data(handle_t *handle, struct inode *inode,
2747                                  unsigned int len);
2748 extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
2749
2750 extern int ext4_readpage_inline(struct inode *inode, struct page *page);
2751 extern int ext4_try_to_write_inline_data(struct address_space *mapping,
2752                                          struct inode *inode,
2753                                          loff_t pos, unsigned len,
2754                                          unsigned flags,
2755                                          struct page **pagep);
2756 extern int ext4_write_inline_data_end(struct inode *inode,
2757                                       loff_t pos, unsigned len,
2758                                       unsigned copied,
2759                                       struct page *page);
2760 extern struct buffer_head *
2761 ext4_journalled_write_inline_data(struct inode *inode,
2762                                   unsigned len,
2763                                   struct page *page);
2764 extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
2765                                            struct inode *inode,
2766                                            loff_t pos, unsigned len,
2767                                            unsigned flags,
2768                                            struct page **pagep,
2769                                            void **fsdata);
2770 extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
2771                                          unsigned len, unsigned copied,
2772                                          struct page *page);
2773 extern int ext4_try_add_inline_entry(handle_t *handle, struct dentry *dentry,
2774                                      struct inode *inode);
2775 extern int ext4_try_create_inline_dir(handle_t *handle,
2776                                       struct inode *parent,
2777                                       struct inode *inode);
2778 extern int ext4_read_inline_dir(struct file *filp,
2779                                 struct dir_context *ctx,
2780                                 int *has_inline_data);
2781 extern int htree_inlinedir_to_tree(struct file *dir_file,
2782                                    struct inode *dir, ext4_lblk_t block,
2783                                    struct dx_hash_info *hinfo,
2784                                    __u32 start_hash, __u32 start_minor_hash,
2785                                    int *has_inline_data);
2786 extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
2787                                         const struct qstr *d_name,
2788                                         struct ext4_dir_entry_2 **res_dir,
2789                                         int *has_inline_data);
2790 extern int ext4_delete_inline_entry(handle_t *handle,
2791                                     struct inode *dir,
2792                                     struct ext4_dir_entry_2 *de_del,
2793                                     struct buffer_head *bh,
2794                                     int *has_inline_data);
2795 extern int empty_inline_dir(struct inode *dir, int *has_inline_data);
2796 extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
2797                                         struct ext4_dir_entry_2 **parent_de,
2798                                         int *retval);
2799 extern int ext4_inline_data_fiemap(struct inode *inode,
2800                                    struct fiemap_extent_info *fieinfo,
2801                                    int *has_inline, __u64 start, __u64 len);
2802 extern int ext4_try_to_evict_inline_data(handle_t *handle,
2803                                          struct inode *inode,
2804                                          int needed);
2805 extern void ext4_inline_data_truncate(struct inode *inode, int *has_inline);
2806
2807 extern int ext4_convert_inline_data(struct inode *inode);
2808
2809 static inline int ext4_has_inline_data(struct inode *inode)
2810 {
2811         return ext4_test_inode_flag(inode, EXT4_INODE_INLINE_DATA) &&
2812                EXT4_I(inode)->i_inline_off;
2813 }
2814
2815 /* namei.c */
2816 extern const struct inode_operations ext4_dir_inode_operations;
2817 extern const struct inode_operations ext4_special_inode_operations;
2818 extern struct dentry *ext4_get_parent(struct dentry *child);
2819 extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
2820                                  struct ext4_dir_entry_2 *de,
2821                                  int blocksize, int csum_size,
2822                                  unsigned int parent_ino, int dotdot_real_len);
2823 extern void initialize_dirent_tail(struct ext4_dir_entry_tail *t,
2824                                    unsigned int blocksize);
2825 extern int ext4_handle_dirty_dirent_node(handle_t *handle,
2826                                          struct inode *inode,
2827                                          struct buffer_head *bh);
2828 #define S_SHIFT 12
2829 static unsigned char ext4_type_by_mode[S_IFMT >> S_SHIFT] = {
2830         [S_IFREG >> S_SHIFT]    = EXT4_FT_REG_FILE,
2831         [S_IFDIR >> S_SHIFT]    = EXT4_FT_DIR,
2832         [S_IFCHR >> S_SHIFT]    = EXT4_FT_CHRDEV,
2833         [S_IFBLK >> S_SHIFT]    = EXT4_FT_BLKDEV,
2834         [S_IFIFO >> S_SHIFT]    = EXT4_FT_FIFO,
2835         [S_IFSOCK >> S_SHIFT]   = EXT4_FT_SOCK,
2836         [S_IFLNK >> S_SHIFT]    = EXT4_FT_SYMLINK,
2837 };
2838
2839 static inline void ext4_set_de_type(struct super_block *sb,
2840                                 struct ext4_dir_entry_2 *de,
2841                                 umode_t mode) {
2842         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE))
2843                 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
2844 }
2845
2846 /* readpages.c */
2847 extern int ext4_mpage_readpages(struct address_space *mapping,
2848                                 struct list_head *pages, struct page *page,
2849                                 unsigned nr_pages);
2850
2851 /* symlink.c */
2852 extern const struct inode_operations ext4_symlink_inode_operations;
2853 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2854
2855 /* block_validity */
2856 extern void ext4_release_system_zone(struct super_block *sb);
2857 extern int ext4_setup_system_zone(struct super_block *sb);
2858 extern int __init ext4_init_system_zone(void);
2859 extern void ext4_exit_system_zone(void);
2860 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2861                                  ext4_fsblk_t start_blk,
2862                                  unsigned int count);
2863 extern int ext4_check_blockref(const char *, unsigned int,
2864                                struct inode *, __le32 *, unsigned int);
2865
2866 /* extents.c */
2867 struct ext4_ext_path;
2868 struct ext4_extent;
2869
2870 /*
2871  * Maximum number of logical blocks in a file; ext4_extent's ee_block is
2872  * __le32.
2873  */
2874 #define EXT_MAX_BLOCKS  0xffffffff
2875
2876 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2877 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2878 extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
2879 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2880                                struct ext4_map_blocks *map, int flags);
2881 extern void ext4_ext_truncate(handle_t *, struct inode *);
2882 extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
2883                                  ext4_lblk_t end);
2884 extern void ext4_ext_init(struct super_block *);
2885 extern void ext4_ext_release(struct super_block *);
2886 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2887                           loff_t len);
2888 extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
2889                                           loff_t offset, ssize_t len);
2890 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2891                            struct ext4_map_blocks *map, int flags);
2892 extern int ext4_ext_calc_metadata_amount(struct inode *inode,
2893                                          ext4_lblk_t lblocks);
2894 extern int ext4_extent_tree_init(handle_t *, struct inode *);
2895 extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
2896                                                    int num,
2897                                                    struct ext4_ext_path *path);
2898 extern int ext4_can_extents_be_merged(struct inode *inode,
2899                                       struct ext4_extent *ex1,
2900                                       struct ext4_extent *ex2);
2901 extern int ext4_ext_insert_extent(handle_t *, struct inode *,
2902                                   struct ext4_ext_path **,
2903                                   struct ext4_extent *, int);
2904 extern struct ext4_ext_path *ext4_find_extent(struct inode *, ext4_lblk_t,
2905                                               struct ext4_ext_path **,
2906                                               int flags);
2907 extern void ext4_ext_drop_refs(struct ext4_ext_path *);
2908 extern int ext4_ext_check_inode(struct inode *inode);
2909 extern int ext4_find_delalloc_range(struct inode *inode,
2910                                     ext4_lblk_t lblk_start,
2911                                     ext4_lblk_t lblk_end);
2912 extern int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk);
2913 extern ext4_lblk_t ext4_ext_next_allocated_block(struct ext4_ext_path *path);
2914 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2915                         __u64 start, __u64 len);
2916 extern int ext4_ext_precache(struct inode *inode);
2917 extern int ext4_collapse_range(struct inode *inode, loff_t offset, loff_t len);
2918 extern int ext4_swap_extents(handle_t *handle, struct inode *inode1,
2919                                 struct inode *inode2, ext4_lblk_t lblk1,
2920                              ext4_lblk_t lblk2,  ext4_lblk_t count,
2921                              int mark_unwritten,int *err);
2922
2923 /* move_extent.c */
2924 extern void ext4_double_down_write_data_sem(struct inode *first,
2925                                             struct inode *second);
2926 extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
2927                                           struct inode *donor_inode);
2928 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2929                              __u64 start_orig, __u64 start_donor,
2930                              __u64 len, __u64 *moved_len);
2931
2932 /* page-io.c */
2933 extern int __init ext4_init_pageio(void);
2934 extern void ext4_exit_pageio(void);
2935 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2936 extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
2937 extern int ext4_put_io_end(ext4_io_end_t *io_end);
2938 extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
2939 extern void ext4_io_submit_init(struct ext4_io_submit *io,
2940                                 struct writeback_control *wbc);
2941 extern void ext4_end_io_rsv_work(struct work_struct *work);
2942 extern void ext4_io_submit(struct ext4_io_submit *io);
2943 extern int ext4_bio_write_page(struct ext4_io_submit *io,
2944                                struct page *page,
2945                                int len,
2946                                struct writeback_control *wbc,
2947                                bool keep_towrite);
2948
2949 /* mmp.c */
2950 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2951
2952 /*
2953  * Add new method to test whether block and inode bitmaps are properly
2954  * initialized. With uninit_bg reading the block from disk is not enough
2955  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2956  */
2957 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2958
2959 static inline int bitmap_uptodate(struct buffer_head *bh)
2960 {
2961         return (buffer_uptodate(bh) &&
2962                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2963 }
2964 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2965 {
2966         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2967 }
2968
2969 /*
2970  * Disable DIO read nolock optimization, so new dioreaders will be forced
2971  * to grab i_mutex
2972  */
2973 static inline void ext4_inode_block_unlocked_dio(struct inode *inode)
2974 {
2975         ext4_set_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2976         smp_mb();
2977 }
2978 static inline void ext4_inode_resume_unlocked_dio(struct inode *inode)
2979 {
2980         smp_mb();
2981         ext4_clear_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2982 }
2983
2984 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2985
2986 /* For ioend & aio unwritten conversion wait queues */
2987 #define EXT4_WQ_HASH_SZ         37
2988 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
2989                                             EXT4_WQ_HASH_SZ])
2990 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
2991                                              EXT4_WQ_HASH_SZ])
2992 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2993 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2994
2995 #define EXT4_RESIZING   0
2996 extern int ext4_resize_begin(struct super_block *sb);
2997 extern void ext4_resize_end(struct super_block *sb);
2998
2999 #endif  /* __KERNEL__ */
3000
3001 #endif  /* _EXT4_H */