Pileus Git - ~andy/linux/blob - 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 <crypto/hash.h>
  33 #ifdef __KERNEL__
  34 #include <linux/compat.h>
  35 #endif
  36
  37 /*
  38  * The fourth extended filesystem constants/structures
  39  */
  40
  41 /*
  42  * Define EXT4FS_DEBUG to produce debug messages
  43  */
  44 #undef EXT4FS_DEBUG
  45
  46 /*
  47  * Debug code
  48  */
  49 #ifdef EXT4FS_DEBUG
  50 #define ext4_debug(f, a...)                                             \
  51         do {                                                            \
  52                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
  53                         __FILE__, __LINE__, __func__);                  \
  54                 printk(KERN_DEBUG f, ## a);                             \
  55         } while (0)
  56 #else
  57 #define ext4_debug(fmt, ...)    no_printk(fmt, ##__VA_ARGS__)
  58 #endif
  59
  60 /*
  61  * Turn on EXT_DEBUG to get lots of info about extents operations.
  62  */
  63 #define EXT_DEBUG__
  64 #ifdef EXT_DEBUG
  65 #define ext_debug(fmt, ...)     printk(fmt, ##__VA_ARGS__)
  66 #else
  67 #define ext_debug(fmt, ...)     no_printk(fmt, ##__VA_ARGS__)
  68 #endif
  69
  70 #define EXT4_ERROR_INODE(inode, fmt, a...) \
  71         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
  72
  73 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...)                 \
  74         ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
  75
  76 #define EXT4_ERROR_FILE(file, block, fmt, a...)                         \
  77         ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
  78
  79 /* data type for block offset of block group */
  80 typedef int ext4_grpblk_t;
  81
  82 /* data type for filesystem-wide blocks number */
  83 typedef unsigned long long ext4_fsblk_t;
  84
  85 /* data type for file logical block number */
  86 typedef __u32 ext4_lblk_t;
  87
  88 /* data type for block group number */
  89 typedef unsigned int ext4_group_t;
  90
  91 /*
  92  * Flags used in mballoc's allocation_context flags field.
  93  *
  94  * Also used to show what's going on for debugging purposes when the
  95  * flag field is exported via the traceport interface
  96  */
  97
  98 /* prefer goal again. length */
  99 #define EXT4_MB_HINT_MERGE              0x0001
 100 /* blocks already reserved */
 101 #define EXT4_MB_HINT_RESERVED           0x0002
 102 /* metadata is being allocated */
 103 #define EXT4_MB_HINT_METADATA           0x0004
 104 /* first blocks in the file */
 105 #define EXT4_MB_HINT_FIRST              0x0008
 106 /* search for the best chunk */
 107 #define EXT4_MB_HINT_BEST               0x0010
 108 /* data is being allocated */
 109 #define EXT4_MB_HINT_DATA               0x0020
 110 /* don't preallocate (for tails) */
 111 #define EXT4_MB_HINT_NOPREALLOC         0x0040
 112 /* allocate for locality group */
 113 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
 114 /* allocate goal blocks or none */
 115 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
 116 /* goal is meaningful */
 117 #define EXT4_MB_HINT_TRY_GOAL           0x0200
 118 /* blocks already pre-reserved by delayed allocation */
 119 #define EXT4_MB_DELALLOC_RESERVED       0x0400
 120 /* We are doing stream allocation */
 121 #define EXT4_MB_STREAM_ALLOC            0x0800
 122 /* Use reserved root blocks if needed */
 123 #define EXT4_MB_USE_ROOT_BLOCKS         0x1000
 124 /* Use blocks from reserved pool */
 125 #define EXT4_MB_USE_RESERVED            0x2000
 126
 127 struct ext4_allocation_request {
 128         /* target inode for block we're allocating */
 129         struct inode *inode;
 130         /* how many blocks we want to allocate */
 131         unsigned int len;
 132         /* logical block in target inode */
 133         ext4_lblk_t logical;
 134         /* the closest logical allocated block to the left */
 135         ext4_lblk_t lleft;
 136         /* the closest logical allocated block to the right */
 137         ext4_lblk_t lright;
 138         /* phys. target (a hint) */
 139         ext4_fsblk_t goal;
 140         /* phys. block for the closest logical allocated block to the left */
 141         ext4_fsblk_t pleft;
 142         /* phys. block for the closest logical allocated block to the right */
 143         ext4_fsblk_t pright;
 144         /* flags. see above EXT4_MB_HINT_* */
 145         unsigned int flags;
 146 };
 147
 148 /*
 149  * Logical to physical block mapping, used by ext4_map_blocks()
 150  *
 151  * This structure is used to pass requests into ext4_map_blocks() as
 152  * well as to store the information returned by ext4_map_blocks().  It
 153  * takes less room on the stack than a struct buffer_head.
 154  */
 155 #define EXT4_MAP_NEW            (1 << BH_New)
 156 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
 157 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
 158 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
 159 #define EXT4_MAP_UNINIT         (1 << BH_Uninit)
 160 /* Sometimes (in the bigalloc case, from ext4_da_get_block_prep) the caller of
 161  * ext4_map_blocks wants to know whether or not the underlying cluster has
 162  * already been accounted for. EXT4_MAP_FROM_CLUSTER conveys to the caller that
 163  * the requested mapping was from previously mapped (or delayed allocated)
 164  * cluster. We use BH_AllocFromCluster only for this flag. BH_AllocFromCluster
 165  * should never appear on buffer_head's state flags.
 166  */
 167 #define EXT4_MAP_FROM_CLUSTER   (1 << BH_AllocFromCluster)
 168 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
 169                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
 170                                  EXT4_MAP_UNINIT | EXT4_MAP_FROM_CLUSTER)
 171
 172 struct ext4_map_blocks {
 173         ext4_fsblk_t m_pblk;
 174         ext4_lblk_t m_lblk;
 175         unsigned int m_len;
 176         unsigned int m_flags;
 177 };
 178
 179 /*
 180  * Flags for ext4_io_end->flags
 181  */
 182 #define EXT4_IO_END_UNWRITTEN   0x0001
 183 #define EXT4_IO_END_DIRECT      0x0002
 184
 185 /*
 186  * For converting uninitialized extents on a work queue. 'handle' is used for
 187  * buffered writeback.
 188  */
 189 typedef struct ext4_io_end {
 190         struct list_head        list;           /* per-file finished IO list */
 191         handle_t                *handle;        /* handle reserved for extent
 192                                                  * conversion */
 193         struct inode            *inode;         /* file being written to */
 194         struct bio              *bio;           /* Linked list of completed
 195                                                  * bios covering the extent */
 196         unsigned int            flag;           /* unwritten or not */
 197         loff_t                  offset;         /* offset in the file */
 198         ssize_t                 size;           /* size of the extent */
 199         struct kiocb            *iocb;          /* iocb struct for AIO */
 200         int                     result;         /* error value for AIO */
 201         atomic_t                count;          /* reference counter */
 202 } ext4_io_end_t;
 203
 204 struct ext4_io_submit {
 205         int                     io_op;
 206         struct bio              *io_bio;
 207         ext4_io_end_t           *io_end;
 208         sector_t                io_next_block;
 209 };
 210
 211 /*
 212  * Special inodes numbers
 213  */
 214 #define EXT4_BAD_INO             1      /* Bad blocks inode */
 215 #define EXT4_ROOT_INO            2      /* Root inode */
 216 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
 217 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
 218 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
 219 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
 220 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
 221 #define EXT4_JOURNAL_INO         8      /* Journal inode */
 222
 223 /* First non-reserved inode for old ext4 filesystems */
 224 #define EXT4_GOOD_OLD_FIRST_INO 11
 225
 226 /*
 227  * Maximal count of links to a file
 228  */
 229 #define EXT4_LINK_MAX           65000
 230
 231 /*
 232  * Macro-instructions used to manage several block sizes
 233  */
 234 #define EXT4_MIN_BLOCK_SIZE             1024
 235 #define EXT4_MAX_BLOCK_SIZE             65536
 236 #define EXT4_MIN_BLOCK_LOG_SIZE         10
 237 #define EXT4_MAX_BLOCK_LOG_SIZE         16
 238 #ifdef __KERNEL__
 239 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
 240 #else
 241 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
 242 #endif
 243 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
 244 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
 245                                          EXT4_SB(s)->s_cluster_bits)
 246 #ifdef __KERNEL__
 247 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
 248 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
 249 #else
 250 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
 251 #endif
 252 #ifdef __KERNEL__
 253 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
 254 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
 255 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
 256 #else
 257 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 258                                  EXT4_GOOD_OLD_INODE_SIZE : \
 259                                  (s)->s_inode_size)
 260 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 261                                  EXT4_GOOD_OLD_FIRST_INO : \
 262                                  (s)->s_first_ino)
 263 #endif
 264 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
 265
 266 /* Translate a block number to a cluster number */
 267 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
 268 /* Translate a cluster number to a block number */
 269 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
 270 /* Translate # of blks to # of clusters */
 271 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
 272                                  (sbi)->s_cluster_bits)
 273
 274 /*
 275  * Structure of a blocks group descriptor
 276  */
 277 struct ext4_group_desc
 278 {
 279         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
 280         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
 281         __le32  bg_inode_table_lo;      /* Inodes table block */
 282         __le16  bg_free_blocks_count_lo;/* Free blocks count */
 283         __le16  bg_free_inodes_count_lo;/* Free inodes count */
 284         __le16  bg_used_dirs_count_lo;  /* Directories count */
 285         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
 286         __le32  bg_exclude_bitmap_lo;   /* Exclude bitmap for snapshots */
 287         __le16  bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
 288         __le16  bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
 289         __le16  bg_itable_unused_lo;    /* Unused inodes count */
 290         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
 291         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
 292         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
 293         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
 294         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
 295         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
 296         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
 297         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
 298         __le32  bg_exclude_bitmap_hi;   /* Exclude bitmap block MSB */
 299         __le16  bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
 300         __le16  bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
 301         __u32   bg_reserved;
 302 };
 303
 304 #define EXT4_BG_INODE_BITMAP_CSUM_HI_END        \
 305         (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
 306          sizeof(__le16))
 307 #define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END        \
 308         (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
 309          sizeof(__le16))
 310
 311 /*
 312  * Structure of a flex block group info
 313  */
 314
 315 struct flex_groups {
 316         atomic64_t      free_clusters;
 317         atomic_t        free_inodes;
 318         atomic_t        used_dirs;
 319 };
 320
 321 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
 322 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
 323 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
 324
 325 /*
 326  * Macro-instructions used to manage group descriptors
 327  */
 328 #define EXT4_MIN_DESC_SIZE              32
 329 #define EXT4_MIN_DESC_SIZE_64BIT        64
 330 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
 331 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
 332 #ifdef __KERNEL__
 333 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
 334 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
 335 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
 336 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
 337 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
 338 #else
 339 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
 340 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
 341 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
 342 #endif
 343
 344 /*
 345  * Constants relative to the data blocks
 346  */
 347 #define EXT4_NDIR_BLOCKS                12
 348 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
 349 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
 350 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
 351 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
 352
 353 /*
 354  * Inode flags
 355  */
 356 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
 357 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
 358 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
 359 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
 360 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
 361 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
 362 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
 363 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
 364 /* Reserved for compression usage... */
 365 #define EXT4_DIRTY_FL                   0x00000100
 366 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
 367 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
 368 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
 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_ECOMPR       = 11,   /* Compression error */
 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(ECOMPR);
 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 unitialized
 532            extent to be an initialized ext4 */
 533 #define EXT4_GET_BLOCKS_CREATE                  0x0001
 534         /* Request the creation of an unitialized extent */
 535 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
 536 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_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         unitialized extents if not allocated, split the uninitialized
 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_UNINIT_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_UNINIT_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         /* Do not put hole in extent cache */
 561 #define EXT4_GET_BLOCKS_NO_PUT_HOLE             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_ext_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                         0x0400
 573 #define EXT4_EX_FORCE_CACHE                     0x0800
 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 /*
 586  * ioctl commands
 587  */
 588 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
 589 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
 590 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
 591 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
 592 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
 593 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
 594 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
 595 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
 596 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
 597 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
 598 #define EXT4_IOC_MIGRATE                _IO('f', 9)
 599  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
 600  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
 601 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
 602 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
 603 #define EXT4_IOC_RESIZE_FS              _IOW('f', 16, __u64)
 604 #define EXT4_IOC_SWAP_BOOT              _IO('f', 17)
 605 #define EXT4_IOC_PRECACHE_EXTENTS       _IO('f', 18)
 606
 607 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 608 /*
 609  * ioctl commands in 32 bit emulation
 610  */
 611 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
 612 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
 613 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
 614 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
 615 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
 616 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
 617 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
 618 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
 619 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
 620 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
 621 #endif
 622
 623 /* Max physical block we can address w/o extents */
 624 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
 625
 626 /*
 627  * Structure of an inode on the disk
 628  */
 629 struct ext4_inode {
 630         __le16  i_mode;         /* File mode */
 631         __le16  i_uid;          /* Low 16 bits of Owner Uid */
 632         __le32  i_size_lo;      /* Size in bytes */
 633         __le32  i_atime;        /* Access time */
 634         __le32  i_ctime;        /* Inode Change time */
 635         __le32  i_mtime;        /* Modification time */
 636         __le32  i_dtime;        /* Deletion Time */
 637         __le16  i_gid;          /* Low 16 bits of Group Id */
 638         __le16  i_links_count;  /* Links count */
 639         __le32  i_blocks_lo;    /* Blocks count */
 640         __le32  i_flags;        /* File flags */
 641         union {
 642                 struct {
 643                         __le32  l_i_version;
 644                 } linux1;
 645                 struct {
 646                         __u32  h_i_translator;
 647                 } hurd1;
 648                 struct {
 649                         __u32  m_i_reserved1;
 650                 } masix1;
 651         } osd1;                         /* OS dependent 1 */
 652         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
 653         __le32  i_generation;   /* File version (for NFS) */
 654         __le32  i_file_acl_lo;  /* File ACL */
 655         __le32  i_size_high;
 656         __le32  i_obso_faddr;   /* Obsoleted fragment address */
 657         union {
 658                 struct {
 659                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
 660                         __le16  l_i_file_acl_high;
 661                         __le16  l_i_uid_high;   /* these 2 fields */
 662                         __le16  l_i_gid_high;   /* were reserved2[0] */
 663                         __le16  l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
 664                         __le16  l_i_reserved;
 665                 } linux2;
 666                 struct {
 667                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 668                         __u16   h_i_mode_high;
 669                         __u16   h_i_uid_high;
 670                         __u16   h_i_gid_high;
 671                         __u32   h_i_author;
 672                 } hurd2;
 673                 struct {
 674                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 675                         __le16  m_i_file_acl_high;
 676                         __u32   m_i_reserved2[2];
 677                 } masix2;
 678         } osd2;                         /* OS dependent 2 */
 679         __le16  i_extra_isize;
 680         __le16  i_checksum_hi;  /* crc32c(uuid+inum+inode) BE */
 681         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
 682         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
 683         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
 684         __le32  i_crtime;       /* File Creation time */
 685         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
 686         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
 687 };
 688
 689 struct move_extent {
 690         __u32 reserved;         /* should be zero */
 691         __u32 donor_fd;         /* donor file descriptor */
 692         __u64 orig_start;       /* logical start offset in block for orig */
 693         __u64 donor_start;      /* logical start offset in block for donor */
 694         __u64 len;              /* block length to be moved */
 695         __u64 moved_len;        /* moved block length */
 696 };
 697
 698 #define EXT4_EPOCH_BITS 2
 699 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
 700 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
 701
 702 /*
 703  * Extended fields will fit into an inode if the filesystem was formatted
 704  * with large inodes (-I 256 or larger) and there are not currently any EAs
 705  * consuming all of the available space. For new inodes we always reserve
 706  * enough space for the kernel's known extended fields, but for inodes
 707  * created with an old kernel this might not have been the case. None of
 708  * the extended inode fields is critical for correct filesystem operation.
 709  * This macro checks if a certain field fits in the inode. Note that
 710  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
 711  */
 712 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
 713         ((offsetof(typeof(*ext4_inode), field) +        \
 714           sizeof((ext4_inode)->field))                  \
 715         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
 716             (einode)->i_extra_isize))                   \
 717
 718 static inline __le32 ext4_encode_extra_time(struct timespec *time)
 719 {
 720        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
 721                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
 722                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
 723 }
 724
 725 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
 726 {
 727        if (sizeof(time->tv_sec) > 4)
 728                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
 729                                << 32;
 730        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
 731 }
 732
 733 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
 734 do {                                                                           \
 735         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
 736         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 737                 (raw_inode)->xtime ## _extra =                                 \
 738                                 ext4_encode_extra_time(&(inode)->xtime);       \
 739 } while (0)
 740
 741 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
 742 do {                                                                           \
 743         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 744                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
 745         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 746                 (raw_inode)->xtime ## _extra =                                 \
 747                                 ext4_encode_extra_time(&(einode)->xtime);      \
 748 } while (0)
 749
 750 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
 751 do {                                                                           \
 752         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
 753         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 754                 ext4_decode_extra_time(&(inode)->xtime,                        \
 755                                        raw_inode->xtime ## _extra);            \
 756         else                                                                   \
 757                 (inode)->xtime.tv_nsec = 0;                                    \
 758 } while (0)
 759
 760 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
 761 do {                                                                           \
 762         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 763                 (einode)->xtime.tv_sec =                                       \
 764                         (signed)le32_to_cpu((raw_inode)->xtime);               \
 765         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 766                 ext4_decode_extra_time(&(einode)->xtime,                       \
 767                                        raw_inode->xtime ## _extra);            \
 768         else                                                                   \
 769                 (einode)->xtime.tv_nsec = 0;                                   \
 770 } while (0)
 771
 772 #define i_disk_version osd1.linux1.l_i_version
 773
 774 #if defined(__KERNEL__) || defined(__linux__)
 775 #define i_reserved1     osd1.linux1.l_i_reserved1
 776 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
 777 #define i_blocks_high   osd2.linux2.l_i_blocks_high
 778 #define i_uid_low       i_uid
 779 #define i_gid_low       i_gid
 780 #define i_uid_high      osd2.linux2.l_i_uid_high
 781 #define i_gid_high      osd2.linux2.l_i_gid_high
 782 #define i_checksum_lo   osd2.linux2.l_i_checksum_lo
 783
 784 #elif defined(__GNU__)
 785
 786 #define i_translator    osd1.hurd1.h_i_translator
 787 #define i_uid_high      osd2.hurd2.h_i_uid_high
 788 #define i_gid_high      osd2.hurd2.h_i_gid_high
 789 #define i_author        osd2.hurd2.h_i_author
 790
 791 #elif defined(__masix__)
 792
 793 #define i_reserved1     osd1.masix1.m_i_reserved1
 794 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
 795 #define i_reserved2     osd2.masix2.m_i_reserved2
 796
 797 #endif /* defined(__KERNEL__) || defined(__linux__) */
 798
 799 #include "extents_status.h"
 800
 801 /*
 802  * fourth extended file system inode data in memory
 803  */
 804 struct ext4_inode_info {
 805         __le32  i_data[15];     /* unconverted */
 806         __u32   i_dtime;
 807         ext4_fsblk_t    i_file_acl;
 808
 809         /*
 810          * i_block_group is the number of the block group which contains
 811          * this file's inode.  Constant across the lifetime of the inode,
 812          * it is ued for making block allocation decisions - we try to
 813          * place a file's data blocks near its inode block, and new inodes
 814          * near to their parent directory's inode.
 815          */
 816         ext4_group_t    i_block_group;
 817         ext4_lblk_t     i_dir_start_lookup;
 818 #if (BITS_PER_LONG < 64)
 819         unsigned long   i_state_flags;          /* Dynamic state flags */
 820 #endif
 821         unsigned long   i_flags;
 822
 823         /*
 824          * Extended attributes can be read independently of the main file
 825          * data. Taking i_mutex even when reading would cause contention
 826          * between readers of EAs and writers of regular file data, so
 827          * instead we synchronize on xattr_sem when reading or changing
 828          * EAs.
 829          */
 830         struct rw_semaphore xattr_sem;
 831
 832         struct list_head i_orphan;      /* unlinked but open inodes */
 833
 834         /*
 835          * i_disksize keeps track of what the inode size is ON DISK, not
 836          * in memory.  During truncate, i_size is set to the new size by
 837          * the VFS prior to calling ext4_truncate(), but the filesystem won't
 838          * set i_disksize to 0 until the truncate is actually under way.
 839          *
 840          * The intent is that i_disksize always represents the blocks which
 841          * are used by this file.  This allows recovery to restart truncate
 842          * on orphans if we crash during truncate.  We actually write i_disksize
 843          * into the on-disk inode when writing inodes out, instead of i_size.
 844          *
 845          * The only time when i_disksize and i_size may be different is when
 846          * a truncate is in progress.  The only things which change i_disksize
 847          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
 848          */
 849         loff_t  i_disksize;
 850
 851         /*
 852          * i_data_sem is for serialising ext4_truncate() against
 853          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
 854          * data tree are chopped off during truncate. We can't do that in
 855          * ext4 because whenever we perform intermediate commits during
 856          * truncate, the inode and all the metadata blocks *must* be in a
 857          * consistent state which allows truncation of the orphans to restart
 858          * during recovery.  Hence we must fix the get_block-vs-truncate race
 859          * by other means, so we have i_data_sem.
 860          */
 861         struct rw_semaphore i_data_sem;
 862         struct inode vfs_inode;
 863         struct jbd2_inode *jinode;
 864
 865         /*
 866          * File creation time. Its function is same as that of
 867          * struct timespec i_{a,c,m}time in the generic inode.
 868          */
 869         struct timespec i_crtime;
 870
 871         /* mballoc */
 872         struct list_head i_prealloc_list;
 873         spinlock_t i_prealloc_lock;
 874
 875         /* extents status tree */
 876         struct ext4_es_tree i_es_tree;
 877         rwlock_t i_es_lock;
 878         struct list_head i_es_lru;
 879         unsigned int i_es_lru_nr;       /* protected by i_es_lock */
 880         unsigned long i_touch_when;     /* jiffies of last accessing */
 881
 882         /* ialloc */
 883         ext4_group_t    i_last_alloc_group;
 884
 885         /* allocation reservation info for delalloc */
 886         /* In case of bigalloc, these refer to clusters rather than blocks */
 887         unsigned int i_reserved_data_blocks;
 888         unsigned int i_reserved_meta_blocks;
 889         unsigned int i_allocated_meta_blocks;
 890         ext4_lblk_t i_da_metadata_calc_last_lblock;
 891         int i_da_metadata_calc_len;
 892
 893         /* on-disk additional length */
 894         __u16 i_extra_isize;
 895
 896         /* Indicate the inline data space. */
 897         u16 i_inline_off;
 898         u16 i_inline_size;
 899
 900 #ifdef CONFIG_QUOTA
 901         /* quota space reservation, managed internally by quota code */
 902         qsize_t i_reserved_quota;
 903 #endif
 904
 905         /* Lock protecting lists below */
 906         spinlock_t i_completed_io_lock;
 907         /*
 908          * Completed IOs that need unwritten extents handling and have
 909          * transaction reserved
 910          */
 911         struct list_head i_rsv_conversion_list;
 912         /*
 913          * Completed IOs that need unwritten extents handling and don't have
 914          * transaction reserved
 915          */
 916         struct list_head i_unrsv_conversion_list;
 917         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
 918         atomic_t i_unwritten; /* Nr. of inflight conversions pending */
 919         struct work_struct i_rsv_conversion_work;
 920         struct work_struct i_unrsv_conversion_work;
 921
 922         spinlock_t i_block_reservation_lock;
 923
 924         /*
 925          * Transactions that contain inode's metadata needed to complete
 926          * fsync and fdatasync, respectively.
 927          */
 928         tid_t i_sync_tid;
 929         tid_t i_datasync_tid;
 930
 931         /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
 932         __u32 i_csum_seed;
 933 };
 934
 935 /*
 936  * File system states
 937  */
 938 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
 939 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
 940 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
 941
 942 /*
 943  * Misc. filesystem flags
 944  */
 945 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
 946 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
 947 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
 948
 949 /*
 950  * Mount flags set via mount options or defaults
 951  */
 952 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
 953 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
 954 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
 955 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
 956 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
 957 #define EXT4_MOUNT_ERRORS_MASK          0x00070
 958 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
 959 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
 960 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
 961 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
 962 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
 963 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
 964 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
 965 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
 966 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
 967 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
 968 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
 969 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
 970 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
 971 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
 972 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
 973 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
 974 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
 975 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
 976 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
 977 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
 978 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
 979 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
 980 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
 981
 982 /*
 983  * Mount flags set either automatically (could not be set by mount option)
 984  * based on per file system feature or property or in special cases such as
 985  * distinguishing between explicit mount option definition and default.
 986  */
 987 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
 988                                                       specified delalloc */
 989 #define EXT4_MOUNT2_STD_GROUP_SIZE      0x00000002 /* We have standard group
 990                                                       size of blocksize * 8
 991                                                       blocks */
 992
 993 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
 994                                                 ~EXT4_MOUNT_##opt
 995 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
 996                                                 EXT4_MOUNT_##opt
 997 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
 998                                          EXT4_MOUNT_##opt)
 999
1000 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
1001                                                 ~EXT4_MOUNT2_##opt
1002 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
1003                                                 EXT4_MOUNT2_##opt
1004 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
1005                                          EXT4_MOUNT2_##opt)
1006
1007 #define ext4_test_and_set_bit           __test_and_set_bit_le
1008 #define ext4_set_bit                    __set_bit_le
1009 #define ext4_set_bit_atomic             ext2_set_bit_atomic
1010 #define ext4_test_and_clear_bit         __test_and_clear_bit_le
1011 #define ext4_clear_bit                  __clear_bit_le
1012 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
1013 #define ext4_test_bit                   test_bit_le
1014 #define ext4_find_next_zero_bit         find_next_zero_bit_le
1015 #define ext4_find_next_bit              find_next_bit_le
1016
1017 extern void ext4_set_bits(void *bm, int cur, int len);
1018
1019 /*
1020  * Maximal mount counts between two filesystem checks
1021  */
1022 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
1023 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
1024
1025 /*
1026  * Behaviour when detecting errors
1027  */
1028 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
1029 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
1030 #define EXT4_ERRORS_PANIC               3       /* Panic */
1031 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
1032
1033 /* Metadata checksum algorithm codes */
1034 #define EXT4_CRC32C_CHKSUM              1
1035
1036 /*
1037  * Structure of the super block
1038  */
1039 struct ext4_super_block {
1040 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1041         __le32  s_blocks_count_lo;      /* Blocks count */
1042         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1043         __le32  s_free_blocks_count_lo; /* Free blocks count */
1044 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1045         __le32  s_first_data_block;     /* First Data Block */
1046         __le32  s_log_block_size;       /* Block size */
1047         __le32  s_log_cluster_size;     /* Allocation cluster size */
1048 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1049         __le32  s_clusters_per_group;   /* # Clusters per group */
1050         __le32  s_inodes_per_group;     /* # Inodes per group */
1051         __le32  s_mtime;                /* Mount time */
1052 /*30*/  __le32  s_wtime;                /* Write time */
1053         __le16  s_mnt_count;            /* Mount count */
1054         __le16  s_max_mnt_count;        /* Maximal mount count */
1055         __le16  s_magic;                /* Magic signature */
1056         __le16  s_state;                /* File system state */
1057         __le16  s_errors;               /* Behaviour when detecting errors */
1058         __le16  s_minor_rev_level;      /* minor revision level */
1059 /*40*/  __le32  s_lastcheck;            /* time of last check */
1060         __le32  s_checkinterval;        /* max. time between checks */
1061         __le32  s_creator_os;           /* OS */
1062         __le32  s_rev_level;            /* Revision level */
1063 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1064         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1065         /*
1066          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1067          *
1068          * Note: the difference between the compatible feature set and
1069          * the incompatible feature set is that if there is a bit set
1070          * in the incompatible feature set that the kernel doesn't
1071          * know about, it should refuse to mount the filesystem.
1072          *
1073          * e2fsck's requirements are more strict; if it doesn't know
1074          * about a feature in either the compatible or incompatible
1075          * feature set, it must abort and not try to meddle with
1076          * things it doesn't understand...
1077          */
1078         __le32  s_first_ino;            /* First non-reserved inode */
1079         __le16  s_inode_size;           /* size of inode structure */
1080         __le16  s_block_group_nr;       /* block group # of this superblock */
1081         __le32  s_feature_compat;       /* compatible feature set */
1082 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1083         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1084 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1085 /*78*/  char    s_volume_name[16];      /* volume name */
1086 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1087 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1088         /*
1089          * Performance hints.  Directory preallocation should only
1090          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1091          */
1092         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1093         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1094         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1095         /*
1096          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1097          */
1098 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1099 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1100         __le32  s_journal_dev;          /* device number of journal file */
1101         __le32  s_last_orphan;          /* start of list of inodes to delete */
1102         __le32  s_hash_seed[4];         /* HTREE hash seed */
1103         __u8    s_def_hash_version;     /* Default hash version to use */
1104         __u8    s_jnl_backup_type;
1105         __le16  s_desc_size;            /* size of group descriptor */
1106 /*100*/ __le32  s_default_mount_opts;
1107         __le32  s_first_meta_bg;        /* First metablock block group */
1108         __le32  s_mkfs_time;            /* When the filesystem was created */
1109         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1110         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1111 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1112         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1113         __le32  s_free_blocks_count_hi; /* Free blocks count */
1114         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1115         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1116         __le32  s_flags;                /* Miscellaneous flags */
1117         __le16  s_raid_stride;          /* RAID stride */
1118         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1119         __le64  s_mmp_block;            /* Block for multi-mount protection */
1120         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1121         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1122         __u8    s_checksum_type;        /* metadata checksum algorithm used */
1123         __le16  s_reserved_pad;
1124         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1125         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1126         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1127         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1128                                               snapshot's future use */
1129         __le32  s_snapshot_list;        /* inode number of the head of the
1130                                            on-disk snapshot list */
1131 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1132         __le32  s_error_count;          /* number of fs errors */
1133         __le32  s_first_error_time;     /* first time an error happened */
1134         __le32  s_first_error_ino;      /* inode involved in first error */
1135         __le64  s_first_error_block;    /* block involved of first error */
1136         __u8    s_first_error_func[32]; /* function where the error happened */
1137         __le32  s_first_error_line;     /* line number where error happened */
1138         __le32  s_last_error_time;      /* most recent time of an error */
1139         __le32  s_last_error_ino;       /* inode involved in last error */
1140         __le32  s_last_error_line;      /* line number where error happened */
1141         __le64  s_last_error_block;     /* block involved of last error */
1142         __u8    s_last_error_func[32];  /* function where the error happened */
1143 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1144         __u8    s_mount_opts[64];
1145         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1146         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1147         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1148         __le32  s_reserved[108];        /* Padding to the end of the block */
1149         __le32  s_checksum;             /* crc32c(superblock) */
1150 };
1151
1152 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1153
1154 #ifdef __KERNEL__
1155
1156 /*
1157  * run-time mount flags
1158  */
1159 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
1160 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
1161
1162 /*
1163  * fourth extended-fs super-block data in memory
1164  */
1165 struct ext4_sb_info {
1166         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1167         unsigned long s_inodes_per_block;/* Number of inodes per block */
1168         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1169         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1170         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1171         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1172         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1173         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1174         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1175         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1176         unsigned long s_overhead;  /* # of fs overhead clusters */
1177         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1178         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1179         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1180         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1181         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1182         struct buffer_head **s_group_desc;
1183         unsigned int s_mount_opt;
1184         unsigned int s_mount_opt2;
1185         unsigned int s_mount_flags;
1186         unsigned int s_def_mount_opt;
1187         ext4_fsblk_t s_sb_block;
1188         atomic64_t s_resv_clusters;
1189         kuid_t s_resuid;
1190         kgid_t s_resgid;
1191         unsigned short s_mount_state;
1192         unsigned short s_pad;
1193         int s_addr_per_block_bits;
1194         int s_desc_per_block_bits;
1195         int s_inode_size;
1196         int s_first_ino;
1197         unsigned int s_inode_readahead_blks;
1198         unsigned int s_inode_goal;
1199         spinlock_t s_next_gen_lock;
1200         u32 s_next_generation;
1201         u32 s_hash_seed[4];
1202         int s_def_hash_version;
1203         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1204         struct percpu_counter s_freeclusters_counter;
1205         struct percpu_counter s_freeinodes_counter;
1206         struct percpu_counter s_dirs_counter;
1207         struct percpu_counter s_dirtyclusters_counter;
1208         struct blockgroup_lock *s_blockgroup_lock;
1209         struct proc_dir_entry *s_proc;
1210         struct kobject s_kobj;
1211         struct completion s_kobj_unregister;
1212         struct super_block *s_sb;
1213
1214         /* Journaling */
1215         struct journal_s *s_journal;
1216         struct list_head s_orphan;
1217         struct mutex s_orphan_lock;
1218         unsigned long s_resize_flags;           /* Flags indicating if there
1219                                                    is a resizer */
1220         unsigned long s_commit_interval;
1221         u32 s_max_batch_time;
1222         u32 s_min_batch_time;
1223         struct block_device *journal_bdev;
1224 #ifdef CONFIG_QUOTA
1225         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
1226         int s_jquota_fmt;                       /* Format of quota to use */
1227 #endif
1228         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1229         struct rb_root system_blks;
1230
1231 #ifdef EXTENTS_STATS
1232         /* ext4 extents stats */
1233         unsigned long s_ext_min;
1234         unsigned long s_ext_max;
1235         unsigned long s_depth_max;
1236         spinlock_t s_ext_stats_lock;
1237         unsigned long s_ext_blocks;
1238         unsigned long s_ext_extents;
1239 #endif
1240
1241         /* for buddy allocator */
1242         struct ext4_group_info ***s_group_info;
1243         struct inode *s_buddy_cache;
1244         spinlock_t s_md_lock;
1245         unsigned short *s_mb_offsets;
1246         unsigned int *s_mb_maxs;
1247         unsigned int s_group_info_size;
1248
1249         /* tunables */
1250         unsigned long s_stripe;
1251         unsigned int s_mb_stream_request;
1252         unsigned int s_mb_max_to_scan;
1253         unsigned int s_mb_min_to_scan;
1254         unsigned int s_mb_stats;
1255         unsigned int s_mb_order2_reqs;
1256         unsigned int s_mb_group_prealloc;
1257         unsigned int s_max_dir_size_kb;
1258         /* where last allocation was done - for stream allocation */
1259         unsigned long s_mb_last_group;
1260         unsigned long s_mb_last_start;
1261
1262         /* stats for buddy allocator */
1263         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1264         atomic_t s_bal_success; /* we found long enough chunks */
1265         atomic_t s_bal_allocated;       /* in blocks */
1266         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1267         atomic_t s_bal_goals;   /* goal hits */
1268         atomic_t s_bal_breaks;  /* too long searches */
1269         atomic_t s_bal_2orders; /* 2^order hits */
1270         spinlock_t s_bal_lock;
1271         unsigned long s_mb_buddies_generated;
1272         unsigned long long s_mb_generation_time;
1273         atomic_t s_mb_lost_chunks;
1274         atomic_t s_mb_preallocated;
1275         atomic_t s_mb_discarded;
1276         atomic_t s_lock_busy;
1277
1278         /* locality groups */
1279         struct ext4_locality_group __percpu *s_locality_groups;
1280
1281         /* for write statistics */
1282         unsigned long s_sectors_written_start;
1283         u64 s_kbytes_written;
1284
1285         /* the size of zero-out chunk */
1286         unsigned int s_extent_max_zeroout_kb;
1287
1288         unsigned int s_log_groups_per_flex;
1289         struct flex_groups *s_flex_groups;
1290         ext4_group_t s_flex_groups_allocated;
1291
1292         /* workqueue for unreserved extent convertions (dio) */
1293         struct workqueue_struct *unrsv_conversion_wq;
1294         /* workqueue for reserved extent conversions (buffered io) */
1295         struct workqueue_struct *rsv_conversion_wq;
1296
1297         /* timer for periodic error stats printing */
1298         struct timer_list s_err_report;
1299
1300         /* Lazy inode table initialization info */
1301         struct ext4_li_request *s_li_request;
1302         /* Wait multiplier for lazy initialization thread */
1303         unsigned int s_li_wait_mult;
1304
1305         /* Kernel thread for multiple mount protection */
1306         struct task_struct *s_mmp_tsk;
1307
1308         /* record the last minlen when FITRIM is called. */
1309         atomic_t s_last_trim_minblks;
1310
1311         /* Reference to checksum algorithm driver via cryptoapi */
1312         struct crypto_shash *s_chksum_driver;
1313
1314         /* Precomputed FS UUID checksum for seeding other checksums */
1315         __u32 s_csum_seed;
1316
1317         /* Reclaim extents from extent status tree */
1318         struct shrinker s_es_shrinker;
1319         struct list_head s_es_lru;
1320         unsigned long s_es_last_sorted;
1321         struct percpu_counter s_extent_cache_cnt;
1322         spinlock_t s_es_lru_lock ____cacheline_aligned_in_smp;
1323 };
1324
1325 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1326 {
1327         return sb->s_fs_info;
1328 }
1329 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1330 {
1331         return container_of(inode, struct ext4_inode_info, vfs_inode);
1332 }
1333
1334 static inline struct timespec ext4_current_time(struct inode *inode)
1335 {
1336         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1337                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1338 }
1339
1340 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1341 {
1342         return ino == EXT4_ROOT_INO ||
1343                 ino == EXT4_USR_QUOTA_INO ||
1344                 ino == EXT4_GRP_QUOTA_INO ||
1345                 ino == EXT4_BOOT_LOADER_INO ||
1346                 ino == EXT4_JOURNAL_INO ||
1347                 ino == EXT4_RESIZE_INO ||
1348                 (ino >= EXT4_FIRST_INO(sb) &&
1349                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1350 }
1351
1352 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
1353                                               struct ext4_io_end *io_end)
1354 {
1355         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
1356                 /* Writeback has to have coversion transaction reserved */
1357                 WARN_ON(EXT4_SB(inode->i_sb)->s_journal && !io_end->handle &&
1358                         !(io_end->flag & EXT4_IO_END_DIRECT));
1359                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
1360                 atomic_inc(&EXT4_I(inode)->i_unwritten);
1361         }
1362 }
1363
1364 static inline ext4_io_end_t *ext4_inode_aio(struct inode *inode)
1365 {
1366         return inode->i_private;
1367 }
1368
1369 static inline void ext4_inode_aio_set(struct inode *inode, ext4_io_end_t *io)
1370 {
1371         inode->i_private = io;
1372 }
1373
1374 /*
1375  * Inode dynamic state flags
1376  */
1377 enum {
1378         EXT4_STATE_JDATA,               /* journaled data exists */
1379         EXT4_STATE_NEW,                 /* inode is newly created */
1380         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1381         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1382         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1383         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1384         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1385         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1386         EXT4_STATE_DELALLOC_RESERVED,   /* blks already reserved for delalloc */
1387         EXT4_STATE_DIOREAD_LOCK,        /* Disable support for dio read
1388                                            nolocking */
1389         EXT4_STATE_MAY_INLINE_DATA,     /* may have in-inode data */
1390         EXT4_STATE_ORDERED_MODE,        /* data=ordered mode */
1391         EXT4_STATE_EXT_PRECACHED,       /* extents have been precached */
1392 };
1393
1394 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1395 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1396 {                                                                       \
1397         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1398 }                                                                       \
1399 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1400 {                                                                       \
1401         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1402 }                                                                       \
1403 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1404 {                                                                       \
1405         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1406 }
1407
1408 EXT4_INODE_BIT_FNS(flag, flags, 0)
1409 #if (BITS_PER_LONG < 64)
1410 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1411
1412 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1413 {
1414         (ei)->i_state_flags = 0;
1415 }
1416 #else
1417 EXT4_INODE_BIT_FNS(state, flags, 32)
1418
1419 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1420 {
1421         /* We depend on the fact that callers will set i_flags */
1422 }
1423 #endif
1424 #else
1425 /* Assume that user mode programs are passing in an ext4fs superblock, not
1426  * a kernel struct super_block.  This will allow us to call the feature-test
1427  * macros from user land. */
1428 #define EXT4_SB(sb)     (sb)
1429 #endif
1430
1431 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1432
1433 /*
1434  * Codes for operating systems
1435  */
1436 #define EXT4_OS_LINUX           0
1437 #define EXT4_OS_HURD            1
1438 #define EXT4_OS_MASIX           2
1439 #define EXT4_OS_FREEBSD         3
1440 #define EXT4_OS_LITES           4
1441
1442 /*
1443  * Revision levels
1444  */
1445 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1446 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1447
1448 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1449 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1450
1451 #define EXT4_GOOD_OLD_INODE_SIZE 128
1452
1453 /*
1454  * Feature set definitions
1455  */
1456
1457 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1458         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1459 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1460         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1461 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1462         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1463 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1464         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1465 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1466         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1467 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1468         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1469 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1470         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1471 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1472         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1473 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1474         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1475
1476 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1477 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1478 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1479 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1480 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1481 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1482
1483 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1484 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1485 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1486 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1487 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1488 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1489 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1490 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1491 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1492 /*
1493  * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM).  When
1494  * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
1495  * all other data structures' checksums.  However, the METADATA_CSUM and
1496  * GDT_CSUM bits are mutually exclusive.
1497  */
1498 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM    0x0400
1499
1500 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1501 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1502 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1503 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1504 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1505 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1506 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1507 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1508 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1509 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1510 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1511 #define EXT4_FEATURE_INCOMPAT_BG_USE_META_CSUM  0x2000 /* use crc32c for bg */
1512 #define EXT4_FEATURE_INCOMPAT_LARGEDIR          0x4000 /* >2GB or 3-lvl htree */
1513 #define EXT4_FEATURE_INCOMPAT_INLINE_DATA       0x8000 /* data in inode */
1514
1515 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1516 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1517                                          EXT4_FEATURE_INCOMPAT_META_BG)
1518 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1519                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1520                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1521
1522 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1523 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1524                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1525                                          EXT4_FEATURE_INCOMPAT_META_BG)
1526 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1527                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1528                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1529
1530 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1531 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1532                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1533                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1534                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1535                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1536                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1537                                          EXT4_FEATURE_INCOMPAT_MMP |    \
1538                                          EXT4_FEATURE_INCOMPAT_INLINE_DATA)
1539 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1540                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1541                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1542                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1543                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1544                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1545                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1546                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
1547                                          EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
1548                                          EXT4_FEATURE_RO_COMPAT_QUOTA)
1549
1550 /*
1551  * Default values for user and/or group using reserved blocks
1552  */
1553 #define EXT4_DEF_RESUID         0
1554 #define EXT4_DEF_RESGID         0
1555
1556 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1557
1558 /*
1559  * Default mount options
1560  */
1561 #define EXT4_DEFM_DEBUG         0x0001
1562 #define EXT4_DEFM_BSDGROUPS     0x0002
1563 #define EXT4_DEFM_XATTR_USER    0x0004
1564 #define EXT4_DEFM_ACL           0x0008
1565 #define EXT4_DEFM_UID16         0x0010
1566 #define EXT4_DEFM_JMODE         0x0060
1567 #define EXT4_DEFM_JMODE_DATA    0x0020
1568 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1569 #define EXT4_DEFM_JMODE_WBACK   0x0060
1570 #define EXT4_DEFM_NOBARRIER     0x0100
1571 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1572 #define EXT4_DEFM_DISCARD       0x0400
1573 #define EXT4_DEFM_NODELALLOC    0x0800
1574
1575 /*
1576  * Default journal batch times
1577  */
1578 #define EXT4_DEF_MIN_BATCH_TIME 0
1579 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1580
1581 /*
1582  * Minimum number of groups in a flexgroup before we separate out
1583  * directories into the first block group of a flexgroup
1584  */
1585 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1586
1587 /*
1588  * Structure of a directory entry
1589  */
1590 #define EXT4_NAME_LEN 255
1591
1592 struct ext4_dir_entry {
1593         __le32  inode;                  /* Inode number */
1594         __le16  rec_len;                /* Directory entry length */
1595         __le16  name_len;               /* Name length */
1596         char    name[EXT4_NAME_LEN];    /* File name */
1597 };
1598
1599 /*
1600  * The new version of the directory entry.  Since EXT4 structures are
1601  * stored in intel byte order, and the name_len field could never be
1602  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1603  * file_type field.
1604  */
1605 struct ext4_dir_entry_2 {
1606         __le32  inode;                  /* Inode number */
1607         __le16  rec_len;                /* Directory entry length */
1608         __u8    name_len;               /* Name length */
1609         __u8    file_type;
1610         char    name[EXT4_NAME_LEN];    /* File name */
1611 };
1612
1613 /*
1614  * This is a bogus directory entry at the end of each leaf block that
1615  * records checksums.
1616  */
1617 struct ext4_dir_entry_tail {
1618         __le32  det_reserved_zero1;     /* Pretend to be unused */
1619         __le16  det_rec_len;            /* 12 */
1620         __u8    det_reserved_zero2;     /* Zero name length */
1621         __u8    det_reserved_ft;        /* 0xDE, fake file type */
1622         __le32  det_checksum;           /* crc32c(uuid+inum+dirblock) */
1623 };
1624
1625 #define EXT4_DIRENT_TAIL(block, blocksize) \
1626         ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
1627                                         ((blocksize) - \
1628                                          sizeof(struct ext4_dir_entry_tail))))
1629
1630 /*
1631  * Ext4 directory file types.  Only the low 3 bits are used.  The
1632  * other bits are reserved for now.
1633  */
1634 #define EXT4_FT_UNKNOWN         0
1635 #define EXT4_FT_REG_FILE        1
1636 #define EXT4_FT_DIR             2
1637 #define EXT4_FT_CHRDEV          3
1638 #define EXT4_FT_BLKDEV          4
1639 #define EXT4_FT_FIFO            5
1640 #define EXT4_FT_SOCK            6
1641 #define EXT4_FT_SYMLINK         7
1642
1643 #define EXT4_FT_MAX             8
1644
1645 #define EXT4_FT_DIR_CSUM        0xDE
1646
1647 /*
1648  * EXT4_DIR_PAD defines the directory entries boundaries
1649  *
1650  * NOTE: It must be a multiple of 4
1651  */
1652 #define EXT4_DIR_PAD                    4
1653 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1654 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1655                                          ~EXT4_DIR_ROUND)
1656 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1657
1658 /*
1659  * If we ever get support for fs block sizes > page_size, we'll need
1660  * to remove the #if statements in the next two functions...
1661  */
1662 static inline unsigned int
1663 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1664 {
1665         unsigned len = le16_to_cpu(dlen);
1666
1667 #if (PAGE_CACHE_SIZE >= 65536)
1668         if (len == EXT4_MAX_REC_LEN || len == 0)
1669                 return blocksize;
1670         return (len & 65532) | ((len & 3) << 16);
1671 #else
1672         return len;
1673 #endif
1674 }
1675
1676 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1677 {
1678         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1679                 BUG();
1680 #if (PAGE_CACHE_SIZE >= 65536)
1681         if (len < 65536)
1682                 return cpu_to_le16(len);
1683         if (len == blocksize) {
1684                 if (blocksize == 65536)
1685                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1686                 else
1687                         return cpu_to_le16(0);
1688         }
1689         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1690 #else
1691         return cpu_to_le16(len);
1692 #endif
1693 }
1694
1695 /*
1696  * Hash Tree Directory indexing
1697  * (c) Daniel Phillips, 2001
1698  */
1699
1700 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1701                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1702                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1703 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1704 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1705
1706 /* Legal values for the dx_root hash_version field: */
1707
1708 #define DX_HASH_LEGACY          0
1709 #define DX_HASH_HALF_MD4        1
1710 #define DX_HASH_TEA             2
1711 #define DX_HASH_LEGACY_UNSIGNED 3
1712 #define DX_HASH_HALF_MD4_UNSIGNED       4
1713 #define DX_HASH_TEA_UNSIGNED            5
1714
1715 static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
1716                               const void *address, unsigned int length)
1717 {
1718         struct {
1719                 struct shash_desc shash;
1720                 char ctx[4];
1721         } desc;
1722         int err;
1723
1724         BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
1725
1726         desc.shash.tfm = sbi->s_chksum_driver;
1727         desc.shash.flags = 0;
1728         *(u32 *)desc.ctx = crc;
1729
1730         err = crypto_shash_update(&desc.shash, address, length);
1731         BUG_ON(err);
1732
1733         return *(u32 *)desc.ctx;
1734 }
1735
1736 #ifdef __KERNEL__
1737
1738 /* hash info structure used by the directory hash */
1739 struct dx_hash_info
1740 {
1741         u32             hash;
1742         u32             minor_hash;
1743         int             hash_version;
1744         u32             *seed;
1745 };
1746
1747
1748 /* 32 and 64 bit signed EOF for dx directories */
1749 #define EXT4_HTREE_EOF_32BIT   ((1UL  << (32 - 1)) - 1)
1750 #define EXT4_HTREE_EOF_64BIT   ((1ULL << (64 - 1)) - 1)
1751
1752
1753 /*
1754  * Control parameters used by ext4_htree_next_block
1755  */
1756 #define HASH_NB_ALWAYS          1
1757
1758
1759 /*
1760  * Describe an inode's exact location on disk and in memory
1761  */
1762 struct ext4_iloc
1763 {
1764         struct buffer_head *bh;
1765         unsigned long offset;
1766         ext4_group_t block_group;
1767 };
1768
1769 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1770 {
1771         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1772 }
1773
1774 /*
1775  * This structure is stuffed into the struct file's private_data field
1776  * for directories.  It is where we put information so that we can do
1777  * readdir operations in hash tree order.
1778  */
1779 struct dir_private_info {
1780         struct rb_root  root;
1781         struct rb_node  *curr_node;
1782         struct fname    *extra_fname;
1783         loff_t          last_pos;
1784         __u32           curr_hash;
1785         __u32           curr_minor_hash;
1786         __u32           next_hash;
1787 };
1788
1789 /* calculate the first block number of the group */
1790 static inline ext4_fsblk_t
1791 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1792 {
1793         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1794                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1795 }
1796
1797 /*
1798  * Special error return code only used by dx_probe() and its callers.
1799  */
1800 #define ERR_BAD_DX_DIR  -75000
1801
1802 /*
1803  * Timeout and state flag for lazy initialization inode thread.
1804  */
1805 #define EXT4_DEF_LI_WAIT_MULT                   10
1806 #define EXT4_DEF_LI_MAX_START_DELAY             5
1807 #define EXT4_LAZYINIT_QUIT                      0x0001
1808 #define EXT4_LAZYINIT_RUNNING                   0x0002
1809
1810 /*
1811  * Lazy inode table initialization info
1812  */
1813 struct ext4_lazy_init {
1814         unsigned long           li_state;
1815         struct list_head        li_request_list;
1816         struct mutex            li_list_mtx;
1817 };
1818
1819 struct ext4_li_request {
1820         struct super_block      *lr_super;
1821         struct ext4_sb_info     *lr_sbi;
1822         ext4_group_t            lr_next_group;
1823         struct list_head        lr_request;
1824         unsigned long           lr_next_sched;
1825         unsigned long           lr_timeout;
1826 };
1827
1828 struct ext4_features {
1829         struct kobject f_kobj;
1830         struct completion f_kobj_unregister;
1831 };
1832
1833 /*
1834  * This structure will be used for multiple mount protection. It will be
1835  * written into the block number saved in the s_mmp_block field in the
1836  * superblock. Programs that check MMP should assume that if
1837  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1838  * to use the filesystem, regardless of how old the timestamp is.
1839  */
1840 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
1841 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1842 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
1843 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
1844
1845 struct mmp_struct {
1846         __le32  mmp_magic;              /* Magic number for MMP */
1847         __le32  mmp_seq;                /* Sequence no. updated periodically */
1848
1849         /*
1850          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1851          * purposes and do not affect the correctness of the algorithm
1852          */
1853         __le64  mmp_time;               /* Time last updated */
1854         char    mmp_nodename[64];       /* Node which last updated MMP block */
1855         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
1856
1857         /*
1858          * mmp_check_interval is used to verify if the MMP block has been
1859          * updated on the block device. The value is updated based on the
1860          * maximum time to write the MMP block during an update cycle.
1861          */
1862         __le16  mmp_check_interval;
1863
1864         __le16  mmp_pad1;
1865         __le32  mmp_pad2[226];
1866         __le32  mmp_checksum;           /* crc32c(uuid+mmp_block) */
1867 };
1868
1869 /* arguments passed to the mmp thread */
1870 struct mmpd_data {
1871         struct buffer_head *bh; /* bh from initial read_mmp_block() */
1872         struct super_block *sb;  /* super block of the fs */
1873 };
1874
1875 /*
1876  * Check interval multiplier
1877  * The MMP block is written every update interval and initially checked every
1878  * update interval x the multiplier (the value is then adapted based on the
1879  * write latency). The reason is that writes can be delayed under load and we
1880  * don't want readers to incorrectly assume that the filesystem is no longer
1881  * in use.
1882  */
1883 #define EXT4_MMP_CHECK_MULT             2UL
1884
1885 /*
1886  * Minimum interval for MMP checking in seconds.
1887  */
1888 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
1889
1890 /*
1891  * Maximum interval for MMP checking in seconds.
1892  */
1893 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
1894
1895 /*
1896  * Function prototypes
1897  */
1898
1899 /*
1900  * Ok, these declarations are also in <linux/kernel.h> but none of the
1901  * ext4 source programs needs to include it so they are duplicated here.
1902  */
1903 # define NORET_TYPE     /**/
1904 # define ATTRIB_NORET   __attribute__((noreturn))
1905 # define NORET_AND      noreturn,
1906
1907 /* bitmap.c */
1908 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
1909 void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1910                                 struct ext4_group_desc *gdp,
1911                                 struct buffer_head *bh, int sz);
1912 int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1913                                   struct ext4_group_desc *gdp,
1914                                   struct buffer_head *bh, int sz);
1915 void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1916                                 struct ext4_group_desc *gdp,
1917                                 struct buffer_head *bh);
1918 int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1919                                   struct ext4_group_desc *gdp,
1920                                   struct buffer_head *bh);
1921
1922 /* balloc.c */
1923 extern void ext4_get_group_no_and_offset(struct super_block *sb,
1924                                          ext4_fsblk_t blocknr,
1925                                          ext4_group_t *blockgrpp,
1926                                          ext4_grpblk_t *offsetp);
1927 extern ext4_group_t ext4_get_group_number(struct super_block *sb,
1928                                           ext4_fsblk_t block);
1929
1930 extern void ext4_validate_block_bitmap(struct super_block *sb,
1931                                        struct ext4_group_desc *desc,
1932                                        unsigned int block_group,
1933                                        struct buffer_head *bh);
1934 extern unsigned int ext4_block_group(struct super_block *sb,
1935                         ext4_fsblk_t blocknr);
1936 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1937                         ext4_fsblk_t blocknr);
1938 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1939 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1940                         ext4_group_t group);
1941 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1942                                          ext4_fsblk_t goal,
1943                                          unsigned int flags,
1944                                          unsigned long *count,
1945                                          int *errp);
1946 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
1947                                     s64 nclusters, unsigned int flags);
1948 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
1949 extern void ext4_check_blocks_bitmap(struct super_block *);
1950 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1951                                                     ext4_group_t block_group,
1952                                                     struct buffer_head ** bh);
1953 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1954
1955 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
1956                                                 ext4_group_t block_group);
1957 extern int ext4_wait_block_bitmap(struct super_block *sb,
1958                                   ext4_group_t block_group,
1959                                   struct buffer_head *bh);
1960 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1961                                                   ext4_group_t block_group);
1962 extern void ext4_init_block_bitmap(struct super_block *sb,
1963                                    struct buffer_head *bh,
1964                                    ext4_group_t group,
1965                                    struct ext4_group_desc *desc);
1966 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
1967                                               ext4_group_t block_group,
1968                                               struct ext4_group_desc *gdp);
1969 extern unsigned ext4_num_overhead_clusters(struct super_block *sb,
1970                                            ext4_group_t block_group,
1971                                            struct ext4_group_desc *gdp);
1972 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
1973
1974 /* dir.c */
1975 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1976                                   struct file *,
1977                                   struct ext4_dir_entry_2 *,
1978                                   struct buffer_head *, char *, int,
1979                                   unsigned int);
1980 #define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset)      \
1981         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
1982                                         (de), (bh), (buf), (size), (offset)))
1983 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1984                                     __u32 minor_hash,
1985                                     struct ext4_dir_entry_2 *dirent);
1986 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1987 extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
1988                              struct buffer_head *bh,
1989                              void *buf, int buf_size,
1990                              const char *name, int namelen,
1991                              struct ext4_dir_entry_2 **dest_de);
1992 void ext4_insert_dentry(struct inode *inode,
1993                         struct ext4_dir_entry_2 *de,
1994                         int buf_size,
1995                         const char *name, int namelen);
1996 static inline void ext4_update_dx_flag(struct inode *inode)
1997 {
1998         if (!EXT4_HAS_COMPAT_FEATURE(inode->i_sb,
1999                                      EXT4_FEATURE_COMPAT_DIR_INDEX))
2000                 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
2001 }
2002 static unsigned char ext4_filetype_table[] = {
2003         DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
2004 };
2005
2006 static inline  unsigned char get_dtype(struct super_block *sb, int filetype)
2007 {
2008         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE) ||
2009             (filetype >= EXT4_FT_MAX))
2010                 return DT_UNKNOWN;
2011
2012         return ext4_filetype_table[filetype];
2013 }
2014
2015 /* fsync.c */
2016 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2017
2018 /* hash.c */
2019 extern int ext4fs_dirhash(const char *name, int len, struct
2020                           dx_hash_info *hinfo);
2021
2022 /* ialloc.c */
2023 extern struct inode *__ext4_new_inode(handle_t *, struct inode *, umode_t,
2024                                       const struct qstr *qstr, __u32 goal,
2025                                       uid_t *owner, int handle_type,
2026                                       unsigned int line_no, int nblocks);
2027
2028 #define ext4_new_inode(handle, dir, mode, qstr, goal, owner) \
2029         __ext4_new_inode((handle), (dir), (mode), (qstr), (goal), (owner), \
2030                          0, 0, 0)
2031 #define ext4_new_inode_start_handle(dir, mode, qstr, goal, owner, \
2032                                     type, nblocks)                  \
2033         __ext4_new_inode(NULL, (dir), (mode), (qstr), (goal), (owner), \
2034                          (type), __LINE__, (nblocks))
2035
2036
2037 extern void ext4_free_inode(handle_t *, struct inode *);
2038 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2039 extern unsigned long ext4_count_free_inodes(struct super_block *);
2040 extern unsigned long ext4_count_dirs(struct super_block *);
2041 extern void ext4_check_inodes_bitmap(struct super_block *);
2042 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2043 extern int ext4_init_inode_table(struct super_block *sb,
2044                                  ext4_group_t group, int barrier);
2045 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2046
2047 /* mballoc.c */
2048 extern long ext4_mb_stats;
2049 extern long ext4_mb_max_to_scan;
2050 extern int ext4_mb_init(struct super_block *);
2051 extern int ext4_mb_release(struct super_block *);
2052 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2053                                 struct ext4_allocation_request *, int *);
2054 extern int ext4_mb_reserve_blocks(struct super_block *, int);
2055 extern void ext4_discard_preallocations(struct inode *);
2056 extern int __init ext4_init_mballoc(void);
2057 extern void ext4_exit_mballoc(void);
2058 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2059                              struct buffer_head *bh, ext4_fsblk_t block,
2060                              unsigned long count, int flags);
2061 extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2062                                    ext4_group_t ngroups);
2063 extern int ext4_mb_add_groupinfo(struct super_block *sb,
2064                 ext4_group_t i, struct ext4_group_desc *desc);
2065 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2066                                 ext4_fsblk_t block, unsigned long count);
2067 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2068
2069 /* inode.c */
2070 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
2071                                                 ext4_lblk_t, int, int *);
2072 struct buffer_head *ext4_bread(handle_t *, struct inode *,
2073                                                 ext4_lblk_t, int, int *);
2074 int ext4_get_block_write(struct inode *inode, sector_t iblock,
2075                          struct buffer_head *bh_result, int create);
2076 int ext4_get_block(struct inode *inode, sector_t iblock,
2077                                 struct buffer_head *bh_result, int create);
2078 int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2079                            struct buffer_head *bh, int create);
2080 int ext4_walk_page_buffers(handle_t *handle,
2081                            struct buffer_head *head,
2082                            unsigned from,
2083                            unsigned to,
2084                            int *partial,
2085                            int (*fn)(handle_t *handle,
2086                                      struct buffer_head *bh));
2087 int do_journal_get_write_access(handle_t *handle,
2088                                 struct buffer_head *bh);
2089 #define FALL_BACK_TO_NONDELALLOC 1
2090 #define CONVERT_INLINE_DATA      2
2091
2092 extern struct inode *ext4_iget(struct super_block *, unsigned long);
2093 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
2094 extern int  ext4_setattr(struct dentry *, struct iattr *);
2095 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
2096                                 struct kstat *stat);
2097 extern void ext4_evict_inode(struct inode *);
2098 extern void ext4_clear_inode(struct inode *);
2099 extern int  ext4_sync_inode(handle_t *, struct inode *);
2100 extern void ext4_dirty_inode(struct inode *, int);
2101 extern int ext4_change_inode_journal_flag(struct inode *, int);
2102 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
2103 extern int ext4_inode_attach_jinode(struct inode *inode);
2104 extern int ext4_can_truncate(struct inode *inode);
2105 extern void ext4_truncate(struct inode *);
2106 extern int ext4_punch_hole(struct inode *inode, loff_t offset, loff_t length);
2107 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
2108 extern void ext4_set_inode_flags(struct inode *);
2109 extern void ext4_get_inode_flags(struct ext4_inode_info *);
2110 extern int ext4_alloc_da_blocks(struct inode *inode);
2111 extern void ext4_set_aops(struct inode *inode);
2112 extern int ext4_writepage_trans_blocks(struct inode *);
2113 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
2114 extern int ext4_block_truncate_page(handle_t *handle,
2115                 struct address_space *mapping, loff_t from);
2116 extern int ext4_block_zero_page_range(handle_t *handle,
2117                 struct address_space *mapping, loff_t from, loff_t length);
2118 extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
2119                              loff_t lstart, loff_t lend);
2120 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2121 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
2122 extern void ext4_da_update_reserve_space(struct inode *inode,
2123                                         int used, int quota_claim);
2124
2125 /* indirect.c */
2126 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
2127                                 struct ext4_map_blocks *map, int flags);
2128 extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
2129                                 const struct iovec *iov, loff_t offset,
2130                                 unsigned long nr_segs);
2131 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
2132 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
2133 extern void ext4_ind_truncate(handle_t *, struct inode *inode);
2134 extern int ext4_free_hole_blocks(handle_t *handle, struct inode *inode,
2135                                  ext4_lblk_t first, ext4_lblk_t stop);
2136
2137 /* ioctl.c */
2138 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
2139 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
2140
2141 /* migrate.c */
2142 extern int ext4_ext_migrate(struct inode *);
2143 extern int ext4_ind_migrate(struct inode *inode);
2144
2145 /* namei.c */
2146 extern int ext4_dirent_csum_verify(struct inode *inode,
2147                                    struct ext4_dir_entry *dirent);
2148 extern int ext4_orphan_add(handle_t *, struct inode *);
2149 extern int ext4_orphan_del(handle_t *, struct inode *);
2150 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
2151                                 __u32 start_minor_hash, __u32 *next_hash);
2152 extern int search_dir(struct buffer_head *bh,
2153                       char *search_buf,
2154                       int buf_size,
2155                       struct inode *dir,
2156                       const struct qstr *d_name,
2157                       unsigned int offset,
2158                       struct ext4_dir_entry_2 **res_dir);
2159 extern int ext4_generic_delete_entry(handle_t *handle,
2160                                      struct inode *dir,
2161                                      struct ext4_dir_entry_2 *de_del,
2162                                      struct buffer_head *bh,
2163                                      void *entry_buf,
2164                                      int buf_size,
2165                                      int csum_size);
2166
2167 /* resize.c */
2168 extern int ext4_group_add(struct super_block *sb,
2169                                 struct ext4_new_group_data *input);
2170 extern int ext4_group_extend(struct super_block *sb,
2171                                 struct ext4_super_block *es,
2172                                 ext4_fsblk_t n_blocks_count);
2173 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
2174
2175 /* super.c */
2176 extern int ext4_calculate_overhead(struct super_block *sb);
2177 extern int ext4_superblock_csum_verify(struct super_block *sb,
2178                                        struct ext4_super_block *es);
2179 extern void ext4_superblock_csum_set(struct super_block *sb);
2180 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
2181 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
2182 extern void ext4_kvfree(void *ptr);
2183 extern int ext4_alloc_flex_bg_array(struct super_block *sb,
2184                                     ext4_group_t ngroup);
2185 extern const char *ext4_decode_error(struct super_block *sb, int errno,
2186                                      char nbuf[16]);
2187
2188 extern __printf(4, 5)
2189 void __ext4_error(struct super_block *, const char *, unsigned int,
2190                   const char *, ...);
2191 extern __printf(5, 6)
2192 void __ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
2193                       const char *, ...);
2194 extern __printf(5, 6)
2195 void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
2196                      const char *, ...);
2197 extern void __ext4_std_error(struct super_block *, const char *,
2198                              unsigned int, int);
2199 extern __printf(4, 5)
2200 void __ext4_abort(struct super_block *, const char *, unsigned int,
2201                   const char *, ...);
2202 extern __printf(4, 5)
2203 void __ext4_warning(struct super_block *, const char *, unsigned int,
2204                     const char *, ...);
2205 extern __printf(3, 4)
2206 void __ext4_msg(struct super_block *, const char *, const char *, ...);
2207 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
2208                            const char *, unsigned int, const char *);
2209 extern __printf(7, 8)
2210 void __ext4_grp_locked_error(const char *, unsigned int,
2211                              struct super_block *, ext4_group_t,
2212                              unsigned long, ext4_fsblk_t,
2213                              const char *, ...);
2214
2215 #ifdef CONFIG_PRINTK
2216
2217 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2218         __ext4_error_inode(inode, func, line, block, fmt, ##__VA_ARGS__)
2219 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2220         __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__)
2221 #define ext4_error(sb, fmt, ...)                                        \
2222         __ext4_error(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2223 #define ext4_abort(sb, fmt, ...)                                        \
2224         __ext4_abort(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2225 #define ext4_warning(sb, fmt, ...)                                      \
2226         __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2227 #define ext4_msg(sb, level, fmt, ...)                           \
2228         __ext4_msg(sb, level, fmt, ##__VA_ARGS__)
2229 #define dump_mmp_msg(sb, mmp, msg)                                      \
2230         __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg)
2231 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2232         __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \
2233                                 fmt, ##__VA_ARGS__)
2234
2235 #else
2236
2237 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2238 do {                                                                    \
2239         no_printk(fmt, ##__VA_ARGS__);                                  \
2240         __ext4_error_inode(inode, "", 0, block, " ");                   \
2241 } while (0)
2242 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2243 do {                                                                    \
2244         no_printk(fmt, ##__VA_ARGS__);                                  \
2245         __ext4_error_file(file, "", 0, block, " ");                     \
2246 } while (0)
2247 #define ext4_error(sb, fmt, ...)                                        \
2248 do {                                                                    \
2249         no_printk(fmt, ##__VA_ARGS__);                                  \
2250         __ext4_error(sb, "", 0, " ");                                   \
2251 } while (0)
2252 #define ext4_abort(sb, fmt, ...)                                        \
2253 do {                                                                    \
2254         no_printk(fmt, ##__VA_ARGS__);                                  \
2255         __ext4_abort(sb, "", 0, " ");                                   \
2256 } while (0)
2257 #define ext4_warning(sb, fmt, ...)                                      \
2258 do {                                                                    \
2259         no_printk(fmt, ##__VA_ARGS__);                                  \
2260         __ext4_warning(sb, "", 0, " ");                                 \
2261 } while (0)
2262 #define ext4_msg(sb, level, fmt, ...)                                   \
2263 do {                                                                    \
2264         no_printk(fmt, ##__VA_ARGS__);                                  \
2265         __ext4_msg(sb, "", " ");                                        \
2266 } while (0)
2267 #define dump_mmp_msg(sb, mmp, msg)                                      \
2268         __dump_mmp_msg(sb, mmp, "", 0, "")
2269 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2270 do {                                                                    \
2271         no_printk(fmt, ##__VA_ARGS__);                          \
2272         __ext4_grp_locked_error("", 0, sb, grp, ino, block, " ");       \
2273 } while (0)
2274
2275 #endif
2276
2277 extern void ext4_update_dynamic_rev(struct super_block *sb);
2278 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
2279                                         __u32 compat);
2280 extern int ext4_update_rocompat_feature(handle_t *handle,
2281                                         struct super_block *sb, __u32 rocompat);
2282 extern int ext4_update_incompat_feature(handle_t *handle,
2283                                         struct super_block *sb, __u32 incompat);
2284 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
2285                                       struct ext4_group_desc *bg);
2286 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
2287                                       struct ext4_group_desc *bg);
2288 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
2289                                      struct ext4_group_desc *bg);
2290 extern __u32 ext4_free_group_clusters(struct super_block *sb,
2291                                       struct ext4_group_desc *bg);
2292 extern __u32 ext4_free_inodes_count(struct super_block *sb,
2293                                  struct ext4_group_desc *bg);
2294 extern __u32 ext4_used_dirs_count(struct super_block *sb,
2295                                 struct ext4_group_desc *bg);
2296 extern __u32 ext4_itable_unused_count(struct super_block *sb,
2297                                    struct ext4_group_desc *bg);
2298 extern void ext4_block_bitmap_set(struct super_block *sb,
2299                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2300 extern void ext4_inode_bitmap_set(struct super_block *sb,
2301                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2302 extern void ext4_inode_table_set(struct super_block *sb,
2303                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
2304 extern void ext4_free_group_clusters_set(struct super_block *sb,
2305                                          struct ext4_group_desc *bg,
2306                                          __u32 count);
2307 extern void ext4_free_inodes_set(struct super_block *sb,
2308                                 struct ext4_group_desc *bg, __u32 count);
2309 extern void ext4_used_dirs_set(struct super_block *sb,
2310                                 struct ext4_group_desc *bg, __u32 count);
2311 extern void ext4_itable_unused_set(struct super_block *sb,
2312                                    struct ext4_group_desc *bg, __u32 count);
2313 extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
2314                                        struct ext4_group_desc *gdp);
2315 extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
2316                                      struct ext4_group_desc *gdp);
2317 extern int ext4_register_li_request(struct super_block *sb,
2318                                     ext4_group_t first_not_zeroed);
2319
2320 static inline int ext4_has_group_desc_csum(struct super_block *sb)
2321 {
2322         return EXT4_HAS_RO_COMPAT_FEATURE(sb,
2323                                           EXT4_FEATURE_RO_COMPAT_GDT_CSUM |
2324                                           EXT4_FEATURE_RO_COMPAT_METADATA_CSUM);
2325 }
2326
2327 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2328 {
2329         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2330                 le32_to_cpu(es->s_blocks_count_lo);
2331 }
2332
2333 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2334 {
2335         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2336                 le32_to_cpu(es->s_r_blocks_count_lo);
2337 }
2338
2339 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2340 {
2341         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2342                 le32_to_cpu(es->s_free_blocks_count_lo);
2343 }
2344
2345 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2346                                          ext4_fsblk_t blk)
2347 {
2348         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2349         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2350 }
2351
2352 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2353                                               ext4_fsblk_t blk)
2354 {
2355         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2356         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2357 }
2358
2359 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2360                                            ext4_fsblk_t blk)
2361 {
2362         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2363         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2364 }
2365
2366 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2367 {
2368         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2369                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2370                         le32_to_cpu(raw_inode->i_size_lo);
2371         else
2372                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2373 }
2374
2375 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2376 {
2377         raw_inode->i_size_lo = cpu_to_le32(i_size);
2378         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2379 }
2380
2381 static inline
2382 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2383                                             ext4_group_t group)
2384 {
2385          struct ext4_group_info ***grp_info;
2386          long indexv, indexh;
2387          BUG_ON(group >= EXT4_SB(sb)->s_groups_count);
2388          grp_info = EXT4_SB(sb)->s_group_info;
2389          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2390          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2391          return grp_info[indexv][indexh];
2392 }
2393
2394 /*
2395  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2396  * the locking protocol documented in the comments of ext4_group_add()
2397  * in resize.c
2398  */
2399 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2400 {
2401         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2402
2403         smp_rmb();
2404         return ngroups;
2405 }
2406
2407 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2408                                              ext4_group_t block_group)
2409 {
2410         return block_group >> sbi->s_log_groups_per_flex;
2411 }
2412
2413 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2414 {
2415         return 1 << sbi->s_log_groups_per_flex;
2416 }
2417
2418 #define ext4_std_error(sb, errno)                               \
2419 do {                                                            \
2420         if ((errno))                                            \
2421                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2422 } while (0)
2423
2424 #ifdef CONFIG_SMP
2425 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2426  * counters. So we need to make sure we have free clusters more
2427  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2428  */
2429 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2430 #else
2431 #define EXT4_FREECLUSTERS_WATERMARK 0
2432 #endif
2433
2434 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2435 {
2436         /*
2437          * XXX: replace with spinlock if seen contended -bzzz
2438          */
2439         down_write(&EXT4_I(inode)->i_data_sem);
2440         if (newsize > EXT4_I(inode)->i_disksize)
2441                 EXT4_I(inode)->i_disksize = newsize;
2442         up_write(&EXT4_I(inode)->i_data_sem);
2443         return ;
2444 }
2445
2446 struct ext4_group_info {
2447         unsigned long   bb_state;
2448         struct rb_root  bb_free_root;
2449         ext4_grpblk_t   bb_first_free;  /* first free block */
2450         ext4_grpblk_t   bb_free;        /* total free blocks */
2451         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2452         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2453         struct          list_head bb_prealloc_list;
2454 #ifdef DOUBLE_CHECK
2455         void            *bb_bitmap;
2456 #endif
2457         struct rw_semaphore alloc_sem;
2458         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2459                                          * regions, index is order.
2460                                          * bb_counters[3] = 5 means
2461                                          * 5 free 8-block regions. */
2462 };
2463
2464 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2465 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2466
2467 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2468         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2469
2470 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2471         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2472 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2473         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2474 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2475         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2476
2477 #define EXT4_MAX_CONTENTION             8
2478 #define EXT4_CONTENTION_THRESHOLD       2
2479
2480 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2481                                               ext4_group_t group)
2482 {
2483         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2484 }
2485
2486 /*
2487  * Returns true if the filesystem is busy enough that attempts to
2488  * access the block group locks has run into contention.
2489  */
2490 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2491 {
2492         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2493 }
2494
2495 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2496 {
2497         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2498         if (spin_trylock(lock))
2499                 /*
2500                  * We're able to grab the lock right away, so drop the
2501                  * lock contention counter.
2502                  */
2503                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2504         else {
2505                 /*
2506                  * The lock is busy, so bump the contention counter,
2507                  * and then wait on the spin lock.
2508                  */
2509                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2510                                   EXT4_MAX_CONTENTION);
2511                 spin_lock(lock);
2512         }
2513 }
2514
2515 static inline void ext4_unlock_group(struct super_block *sb,
2516                                         ext4_group_t group)
2517 {
2518         spin_unlock(ext4_group_lock_ptr(sb, group));
2519 }
2520
2521 /*
2522  * Block validity checking
2523  */
2524 #define ext4_check_indirect_blockref(inode, bh)                         \
2525         ext4_check_blockref(__func__, __LINE__, inode,                  \
2526                             (__le32 *)(bh)->b_data,                     \
2527                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2528
2529 #define ext4_ind_check_inode(inode)                                     \
2530         ext4_check_blockref(__func__, __LINE__, inode,                  \
2531                             EXT4_I(inode)->i_data,                      \
2532                             EXT4_NDIR_BLOCKS)
2533
2534 /*
2535  * Inodes and files operations
2536  */
2537
2538 /* dir.c */
2539 extern const struct file_operations ext4_dir_operations;
2540
2541 /* file.c */
2542 extern const struct inode_operations ext4_file_inode_operations;
2543 extern const struct file_operations ext4_file_operations;
2544 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2545 extern void ext4_unwritten_wait(struct inode *inode);
2546
2547 /* inline.c */
2548 extern int ext4_has_inline_data(struct inode *inode);
2549 extern int ext4_get_inline_size(struct inode *inode);
2550 extern int ext4_get_max_inline_size(struct inode *inode);
2551 extern int ext4_find_inline_data_nolock(struct inode *inode);
2552 extern void ext4_write_inline_data(struct inode *inode,
2553                                    struct ext4_iloc *iloc,
2554                                    void *buffer, loff_t pos,
2555                                    unsigned int len);
2556 extern int ext4_prepare_inline_data(handle_t *handle, struct inode *inode,
2557                                     unsigned int len);
2558 extern int ext4_init_inline_data(handle_t *handle, struct inode *inode,
2559                                  unsigned int len);
2560 extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
2561
2562 extern int ext4_readpage_inline(struct inode *inode, struct page *page);
2563 extern int ext4_try_to_write_inline_data(struct address_space *mapping,
2564                                          struct inode *inode,
2565                                          loff_t pos, unsigned len,
2566                                          unsigned flags,
2567                                          struct page **pagep);
2568 extern int ext4_write_inline_data_end(struct inode *inode,
2569                                       loff_t pos, unsigned len,
2570                                       unsigned copied,
2571                                       struct page *page);
2572 extern struct buffer_head *
2573 ext4_journalled_write_inline_data(struct inode *inode,
2574                                   unsigned len,
2575                                   struct page *page);
2576 extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
2577                                            struct inode *inode,
2578                                            loff_t pos, unsigned len,
2579                                            unsigned flags,
2580                                            struct page **pagep,
2581                                            void **fsdata);
2582 extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
2583                                          unsigned len, unsigned copied,
2584                                          struct page *page);
2585 extern int ext4_try_add_inline_entry(handle_t *handle, struct dentry *dentry,
2586                                      struct inode *inode);
2587 extern int ext4_try_create_inline_dir(handle_t *handle,
2588                                       struct inode *parent,
2589                                       struct inode *inode);
2590 extern int ext4_read_inline_dir(struct file *filp,
2591                                 struct dir_context *ctx,
2592                                 int *has_inline_data);
2593 extern int htree_inlinedir_to_tree(struct file *dir_file,
2594                                    struct inode *dir, ext4_lblk_t block,
2595                                    struct dx_hash_info *hinfo,
2596                                    __u32 start_hash, __u32 start_minor_hash,
2597                                    int *has_inline_data);
2598 extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
2599                                         const struct qstr *d_name,
2600                                         struct ext4_dir_entry_2 **res_dir,
2601                                         int *has_inline_data);
2602 extern int ext4_delete_inline_entry(handle_t *handle,
2603                                     struct inode *dir,
2604                                     struct ext4_dir_entry_2 *de_del,
2605                                     struct buffer_head *bh,
2606                                     int *has_inline_data);
2607 extern int empty_inline_dir(struct inode *dir, int *has_inline_data);
2608 extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
2609                                         struct ext4_dir_entry_2 **parent_de,
2610                                         int *retval);
2611 extern int ext4_inline_data_fiemap(struct inode *inode,
2612                                    struct fiemap_extent_info *fieinfo,
2613                                    int *has_inline);
2614 extern int ext4_try_to_evict_inline_data(handle_t *handle,
2615                                          struct inode *inode,
2616                                          int needed);
2617 extern void ext4_inline_data_truncate(struct inode *inode, int *has_inline);
2618
2619 extern int ext4_convert_inline_data(struct inode *inode);
2620
2621 /* namei.c */
2622 extern const struct inode_operations ext4_dir_inode_operations;
2623 extern const struct inode_operations ext4_special_inode_operations;
2624 extern struct dentry *ext4_get_parent(struct dentry *child);
2625 extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
2626                                  struct ext4_dir_entry_2 *de,
2627                                  int blocksize, int csum_size,
2628                                  unsigned int parent_ino, int dotdot_real_len);
2629 extern void initialize_dirent_tail(struct ext4_dir_entry_tail *t,
2630                                    unsigned int blocksize);
2631 extern int ext4_handle_dirty_dirent_node(handle_t *handle,
2632                                          struct inode *inode,
2633                                          struct buffer_head *bh);
2634 #define S_SHIFT 12
2635 static unsigned char ext4_type_by_mode[S_IFMT >> S_SHIFT] = {
2636         [S_IFREG >> S_SHIFT]    = EXT4_FT_REG_FILE,
2637         [S_IFDIR >> S_SHIFT]    = EXT4_FT_DIR,
2638         [S_IFCHR >> S_SHIFT]    = EXT4_FT_CHRDEV,
2639         [S_IFBLK >> S_SHIFT]    = EXT4_FT_BLKDEV,
2640         [S_IFIFO >> S_SHIFT]    = EXT4_FT_FIFO,
2641         [S_IFSOCK >> S_SHIFT]   = EXT4_FT_SOCK,
2642         [S_IFLNK >> S_SHIFT]    = EXT4_FT_SYMLINK,
2643 };
2644
2645 static inline void ext4_set_de_type(struct super_block *sb,
2646                                 struct ext4_dir_entry_2 *de,
2647                                 umode_t mode) {
2648         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE))
2649                 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
2650 }
2651
2652
2653 /* symlink.c */
2654 extern const struct inode_operations ext4_symlink_inode_operations;
2655 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2656
2657 /* block_validity */
2658 extern void ext4_release_system_zone(struct super_block *sb);
2659 extern int ext4_setup_system_zone(struct super_block *sb);
2660 extern int __init ext4_init_system_zone(void);
2661 extern void ext4_exit_system_zone(void);
2662 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2663                                  ext4_fsblk_t start_blk,
2664                                  unsigned int count);
2665 extern int ext4_check_blockref(const char *, unsigned int,
2666                                struct inode *, __le32 *, unsigned int);
2667
2668 /* extents.c */
2669 struct ext4_ext_path;
2670 struct ext4_extent;
2671
2672 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2673 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2674 extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
2675 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2676                                struct ext4_map_blocks *map, int flags);
2677 extern void ext4_ext_truncate(handle_t *, struct inode *);
2678 extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
2679                                  ext4_lblk_t end);
2680 extern void ext4_ext_init(struct super_block *);
2681 extern void ext4_ext_release(struct super_block *);
2682 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2683                           loff_t len);
2684 extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
2685                                           loff_t offset, ssize_t len);
2686 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2687                            struct ext4_map_blocks *map, int flags);
2688 extern int ext4_ext_calc_metadata_amount(struct inode *inode,
2689                                          ext4_lblk_t lblocks);
2690 extern int ext4_extent_tree_init(handle_t *, struct inode *);
2691 extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
2692                                                    int num,
2693                                                    struct ext4_ext_path *path);
2694 extern int ext4_can_extents_be_merged(struct inode *inode,
2695                                       struct ext4_extent *ex1,
2696                                       struct ext4_extent *ex2);
2697 extern int ext4_ext_insert_extent(handle_t *, struct inode *,
2698                                   struct ext4_ext_path *,
2699                                   struct ext4_extent *, int);
2700 extern struct ext4_ext_path *ext4_ext_find_extent(struct inode *, ext4_lblk_t,
2701                                                   struct ext4_ext_path *,
2702                                                   int flags);
2703 extern void ext4_ext_drop_refs(struct ext4_ext_path *);
2704 extern int ext4_ext_check_inode(struct inode *inode);
2705 extern int ext4_find_delalloc_range(struct inode *inode,
2706                                     ext4_lblk_t lblk_start,
2707                                     ext4_lblk_t lblk_end);
2708 extern int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk);
2709 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2710                         __u64 start, __u64 len);
2711 extern int ext4_ext_precache(struct inode *inode);
2712
2713 /* move_extent.c */
2714 extern void ext4_double_down_write_data_sem(struct inode *first,
2715                                             struct inode *second);
2716 extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
2717                                           struct inode *donor_inode);
2718 void ext4_inode_double_lock(struct inode *inode1, struct inode *inode2);
2719 void ext4_inode_double_unlock(struct inode *inode1, struct inode *inode2);
2720 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2721                              __u64 start_orig, __u64 start_donor,
2722                              __u64 len, __u64 *moved_len);
2723
2724 /* page-io.c */
2725 extern int __init ext4_init_pageio(void);
2726 extern void ext4_exit_pageio(void);
2727 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2728 extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
2729 extern int ext4_put_io_end(ext4_io_end_t *io_end);
2730 extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
2731 extern void ext4_io_submit_init(struct ext4_io_submit *io,
2732                                 struct writeback_control *wbc);
2733 extern void ext4_end_io_rsv_work(struct work_struct *work);
2734 extern void ext4_end_io_unrsv_work(struct work_struct *work);
2735 extern void ext4_io_submit(struct ext4_io_submit *io);
2736 extern int ext4_bio_write_page(struct ext4_io_submit *io,
2737                                struct page *page,
2738                                int len,
2739                                struct writeback_control *wbc);
2740
2741 /* mmp.c */
2742 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2743 extern void ext4_mmp_csum_set(struct super_block *sb, struct mmp_struct *mmp);
2744 extern int ext4_mmp_csum_verify(struct super_block *sb,
2745                                 struct mmp_struct *mmp);
2746
2747 /*
2748  * Note that these flags will never ever appear in a buffer_head's state flag.
2749  * See EXT4_MAP_... to see where this is used.
2750  */
2751 enum ext4_state_bits {
2752         BH_Uninit       /* blocks are allocated but uninitialized on disk */
2753          = BH_JBDPrivateStart,
2754         BH_AllocFromCluster,    /* allocated blocks were part of already
2755                                  * allocated cluster. */
2756 };
2757
2758 /*
2759  * Add new method to test whether block and inode bitmaps are properly
2760  * initialized. With uninit_bg reading the block from disk is not enough
2761  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2762  */
2763 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2764
2765 static inline int bitmap_uptodate(struct buffer_head *bh)
2766 {
2767         return (buffer_uptodate(bh) &&
2768                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2769 }
2770 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2771 {
2772         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2773 }
2774
2775 /*
2776  * Disable DIO read nolock optimization, so new dioreaders will be forced
2777  * to grab i_mutex
2778  */
2779 static inline void ext4_inode_block_unlocked_dio(struct inode *inode)
2780 {
2781         ext4_set_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2782         smp_mb();
2783 }
2784 static inline void ext4_inode_resume_unlocked_dio(struct inode *inode)
2785 {
2786         smp_mb();
2787         ext4_clear_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2788 }
2789
2790 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2791
2792 /* For ioend & aio unwritten conversion wait queues */
2793 #define EXT4_WQ_HASH_SZ         37
2794 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
2795                                             EXT4_WQ_HASH_SZ])
2796 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
2797                                              EXT4_WQ_HASH_SZ])
2798 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2799 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2800
2801 #define EXT4_RESIZING   0
2802 extern int ext4_resize_begin(struct super_block *sb);
2803 extern void ext4_resize_end(struct super_block *sb);
2804
2805 #endif  /* __KERNEL__ */
2806
2807 #endif  /* _EXT4_H */