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
 125 struct ext4_allocation_request {
 126         /* target inode for block we're allocating */
 127         struct inode *inode;
 128         /* how many blocks we want to allocate */
 129         unsigned int len;
 130         /* logical block in target inode */
 131         ext4_lblk_t logical;
 132         /* the closest logical allocated block to the left */
 133         ext4_lblk_t lleft;
 134         /* the closest logical allocated block to the right */
 135         ext4_lblk_t lright;
 136         /* phys. target (a hint) */
 137         ext4_fsblk_t goal;
 138         /* phys. block for the closest logical allocated block to the left */
 139         ext4_fsblk_t pleft;
 140         /* phys. block for the closest logical allocated block to the right */
 141         ext4_fsblk_t pright;
 142         /* flags. see above EXT4_MB_HINT_* */
 143         unsigned int flags;
 144 };
 145
 146 /*
 147  * Logical to physical block mapping, used by ext4_map_blocks()
 148  *
 149  * This structure is used to pass requests into ext4_map_blocks() as
 150  * well as to store the information returned by ext4_map_blocks().  It
 151  * takes less room on the stack than a struct buffer_head.
 152  */
 153 #define EXT4_MAP_NEW            (1 << BH_New)
 154 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
 155 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
 156 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
 157 #define EXT4_MAP_UNINIT         (1 << BH_Uninit)
 158 /* Sometimes (in the bigalloc case, from ext4_da_get_block_prep) the caller of
 159  * ext4_map_blocks wants to know whether or not the underlying cluster has
 160  * already been accounted for. EXT4_MAP_FROM_CLUSTER conveys to the caller that
 161  * the requested mapping was from previously mapped (or delayed allocated)
 162  * cluster. We use BH_AllocFromCluster only for this flag. BH_AllocFromCluster
 163  * should never appear on buffer_head's state flags.
 164  */
 165 #define EXT4_MAP_FROM_CLUSTER   (1 << BH_AllocFromCluster)
 166 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
 167                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
 168                                  EXT4_MAP_UNINIT | EXT4_MAP_FROM_CLUSTER)
 169
 170 struct ext4_map_blocks {
 171         ext4_fsblk_t m_pblk;
 172         ext4_lblk_t m_lblk;
 173         unsigned int m_len;
 174         unsigned int m_flags;
 175 };
 176
 177 /*
 178  * For delayed allocation tracking
 179  */
 180 struct mpage_da_data {
 181         struct inode *inode;
 182         sector_t b_blocknr;             /* start block number of extent */
 183         size_t b_size;                  /* size of extent */
 184         unsigned long b_state;          /* state of the extent */
 185         unsigned long first_page, next_page;    /* extent of pages */
 186         struct writeback_control *wbc;
 187         int io_done;
 188         int pages_written;
 189         int retval;
 190 };
 191
 192 /*
 193  * Flags for ext4_io_end->flags
 194  */
 195 #define EXT4_IO_END_UNWRITTEN   0x0001
 196 #define EXT4_IO_END_ERROR       0x0002
 197 #define EXT4_IO_END_DIRECT      0x0004
 198
 199 struct ext4_io_page {
 200         struct page     *p_page;
 201         atomic_t        p_count;
 202 };
 203
 204 #define MAX_IO_PAGES 128
 205
 206 /*
 207  * For converting uninitialized extents on a work queue.
 208  *
 209  * 'page' is only used from the writepage() path; 'pages' is only used for
 210  * buffered writes; they are used to keep page references until conversion
 211  * takes place.  For AIO/DIO, neither field is filled in.
 212  */
 213 typedef struct ext4_io_end {
 214         struct list_head        list;           /* per-file finished IO list */
 215         struct inode            *inode;         /* file being written to */
 216         unsigned int            flag;           /* unwritten or not */
 217         loff_t                  offset;         /* offset in the file */
 218         ssize_t                 size;           /* size of the extent */
 219         struct kiocb            *iocb;          /* iocb struct for AIO */
 220         int                     result;         /* error value for AIO */
 221         int                     num_io_pages;   /* for writepages() */
 222         struct ext4_io_page     *pages[MAX_IO_PAGES]; /* for writepages() */
 223 } ext4_io_end_t;
 224
 225 struct ext4_io_submit {
 226         int                     io_op;
 227         struct bio              *io_bio;
 228         ext4_io_end_t           *io_end;
 229         struct ext4_io_page     *io_page;
 230         sector_t                io_next_block;
 231 };
 232
 233 /*
 234  * Special inodes numbers
 235  */
 236 #define EXT4_BAD_INO             1      /* Bad blocks inode */
 237 #define EXT4_ROOT_INO            2      /* Root inode */
 238 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
 239 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
 240 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
 241 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
 242 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
 243 #define EXT4_JOURNAL_INO         8      /* Journal inode */
 244
 245 /* First non-reserved inode for old ext4 filesystems */
 246 #define EXT4_GOOD_OLD_FIRST_INO 11
 247
 248 /*
 249  * Maximal count of links to a file
 250  */
 251 #define EXT4_LINK_MAX           65000
 252
 253 /*
 254  * Macro-instructions used to manage several block sizes
 255  */
 256 #define EXT4_MIN_BLOCK_SIZE             1024
 257 #define EXT4_MAX_BLOCK_SIZE             65536
 258 #define EXT4_MIN_BLOCK_LOG_SIZE         10
 259 #define EXT4_MAX_BLOCK_LOG_SIZE         16
 260 #ifdef __KERNEL__
 261 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
 262 #else
 263 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
 264 #endif
 265 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
 266 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
 267                                          EXT4_SB(s)->s_cluster_bits)
 268 #ifdef __KERNEL__
 269 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
 270 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
 271 #else
 272 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
 273 #endif
 274 #ifdef __KERNEL__
 275 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
 276 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
 277 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
 278 #else
 279 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 280                                  EXT4_GOOD_OLD_INODE_SIZE : \
 281                                  (s)->s_inode_size)
 282 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 283                                  EXT4_GOOD_OLD_FIRST_INO : \
 284                                  (s)->s_first_ino)
 285 #endif
 286 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
 287
 288 /* Translate a block number to a cluster number */
 289 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
 290 /* Translate a cluster number to a block number */
 291 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
 292 /* Translate # of blks to # of clusters */
 293 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
 294                                  (sbi)->s_cluster_bits)
 295
 296 /*
 297  * Structure of a blocks group descriptor
 298  */
 299 struct ext4_group_desc
 300 {
 301         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
 302         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
 303         __le32  bg_inode_table_lo;      /* Inodes table block */
 304         __le16  bg_free_blocks_count_lo;/* Free blocks count */
 305         __le16  bg_free_inodes_count_lo;/* Free inodes count */
 306         __le16  bg_used_dirs_count_lo;  /* Directories count */
 307         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
 308         __le32  bg_exclude_bitmap_lo;   /* Exclude bitmap for snapshots */
 309         __le16  bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
 310         __le16  bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
 311         __le16  bg_itable_unused_lo;    /* Unused inodes count */
 312         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
 313         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
 314         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
 315         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
 316         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
 317         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
 318         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
 319         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
 320         __le32  bg_exclude_bitmap_hi;   /* Exclude bitmap block MSB */
 321         __le16  bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
 322         __le16  bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
 323         __u32   bg_reserved;
 324 };
 325
 326 #define EXT4_BG_INODE_BITMAP_CSUM_HI_END        \
 327         (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
 328          sizeof(__le16))
 329 #define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END        \
 330         (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
 331          sizeof(__le16))
 332
 333 /*
 334  * Structure of a flex block group info
 335  */
 336
 337 struct flex_groups {
 338         atomic_t free_inodes;
 339         atomic_t free_clusters;
 340         atomic_t used_dirs;
 341 };
 342
 343 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
 344 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
 345 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
 346
 347 /*
 348  * Macro-instructions used to manage group descriptors
 349  */
 350 #define EXT4_MIN_DESC_SIZE              32
 351 #define EXT4_MIN_DESC_SIZE_64BIT        64
 352 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
 353 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
 354 #ifdef __KERNEL__
 355 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
 356 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
 357 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
 358 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
 359 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
 360 #else
 361 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
 362 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
 363 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
 364 #endif
 365
 366 /*
 367  * Constants relative to the data blocks
 368  */
 369 #define EXT4_NDIR_BLOCKS                12
 370 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
 371 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
 372 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
 373 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
 374
 375 /*
 376  * Inode flags
 377  */
 378 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
 379 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
 380 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
 381 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
 382 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
 383 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
 384 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
 385 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
 386 /* Reserved for compression usage... */
 387 #define EXT4_DIRTY_FL                   0x00000100
 388 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
 389 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
 390 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
 391 /* End compression flags --- maybe not all used */
 392 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
 393 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
 394 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
 395 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
 396 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
 397 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
 398 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
 399 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
 400 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
 401 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
 402 #define EXT4_INLINE_DATA_FL             0x10000000 /* Inode has inline data. */
 403 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
 404
 405 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
 406 #define EXT4_FL_USER_MODIFIABLE         0x004B80FF /* User modifiable flags */
 407
 408 /* Flags that should be inherited by new inodes from their parent. */
 409 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
 410                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
 411                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
 412                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
 413
 414 /* Flags that are appropriate for regular files (all but dir-specific ones). */
 415 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
 416
 417 /* Flags that are appropriate for non-directories/regular files. */
 418 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
 419
 420 /* Mask out flags that are inappropriate for the given type of inode. */
 421 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
 422 {
 423         if (S_ISDIR(mode))
 424                 return flags;
 425         else if (S_ISREG(mode))
 426                 return flags & EXT4_REG_FLMASK;
 427         else
 428                 return flags & EXT4_OTHER_FLMASK;
 429 }
 430
 431 /*
 432  * Inode flags used for atomic set/get
 433  */
 434 enum {
 435         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
 436         EXT4_INODE_UNRM         = 1,    /* Undelete */
 437         EXT4_INODE_COMPR        = 2,    /* Compress file */
 438         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
 439         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
 440         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
 441         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
 442         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
 443 /* Reserved for compression usage... */
 444         EXT4_INODE_DIRTY        = 8,
 445         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
 446         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
 447         EXT4_INODE_ECOMPR       = 11,   /* Compression error */
 448 /* End compression flags --- maybe not all used */
 449         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
 450         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
 451         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
 452         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
 453         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
 454         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
 455         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
 456         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
 457         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
 458         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
 459         EXT4_INODE_INLINE_DATA  = 28,   /* Data in inode. */
 460         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
 461 };
 462
 463 /*
 464  * Since it's pretty easy to mix up bit numbers and hex values, we use a
 465  * build-time check to make sure that EXT4_XXX_FL is consistent with respect to
 466  * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost
 467  * any extra space in the compiled kernel image, otherwise, the build will fail.
 468  * It's important that these values are the same, since we are using
 469  * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent
 470  * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk
 471  * values found in ext2, ext3 and ext4 filesystems, and of course the values
 472  * defined in e2fsprogs.
 473  *
 474  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
 475  */
 476 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
 477 #define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG))
 478
 479 static inline void ext4_check_flag_values(void)
 480 {
 481         CHECK_FLAG_VALUE(SECRM);
 482         CHECK_FLAG_VALUE(UNRM);
 483         CHECK_FLAG_VALUE(COMPR);
 484         CHECK_FLAG_VALUE(SYNC);
 485         CHECK_FLAG_VALUE(IMMUTABLE);
 486         CHECK_FLAG_VALUE(APPEND);
 487         CHECK_FLAG_VALUE(NODUMP);
 488         CHECK_FLAG_VALUE(NOATIME);
 489         CHECK_FLAG_VALUE(DIRTY);
 490         CHECK_FLAG_VALUE(COMPRBLK);
 491         CHECK_FLAG_VALUE(NOCOMPR);
 492         CHECK_FLAG_VALUE(ECOMPR);
 493         CHECK_FLAG_VALUE(INDEX);
 494         CHECK_FLAG_VALUE(IMAGIC);
 495         CHECK_FLAG_VALUE(JOURNAL_DATA);
 496         CHECK_FLAG_VALUE(NOTAIL);
 497         CHECK_FLAG_VALUE(DIRSYNC);
 498         CHECK_FLAG_VALUE(TOPDIR);
 499         CHECK_FLAG_VALUE(HUGE_FILE);
 500         CHECK_FLAG_VALUE(EXTENTS);
 501         CHECK_FLAG_VALUE(EA_INODE);
 502         CHECK_FLAG_VALUE(EOFBLOCKS);
 503         CHECK_FLAG_VALUE(INLINE_DATA);
 504         CHECK_FLAG_VALUE(RESERVED);
 505 }
 506
 507 /* Used to pass group descriptor data when online resize is done */
 508 struct ext4_new_group_input {
 509         __u32 group;            /* Group number for this data */
 510         __u64 block_bitmap;     /* Absolute block number of block bitmap */
 511         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
 512         __u64 inode_table;      /* Absolute block number of inode table start */
 513         __u32 blocks_count;     /* Total number of blocks in this group */
 514         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
 515         __u16 unused;
 516 };
 517
 518 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 519 struct compat_ext4_new_group_input {
 520         u32 group;
 521         compat_u64 block_bitmap;
 522         compat_u64 inode_bitmap;
 523         compat_u64 inode_table;
 524         u32 blocks_count;
 525         u16 reserved_blocks;
 526         u16 unused;
 527 };
 528 #endif
 529
 530 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
 531 struct ext4_new_group_data {
 532         __u32 group;
 533         __u64 block_bitmap;
 534         __u64 inode_bitmap;
 535         __u64 inode_table;
 536         __u32 blocks_count;
 537         __u16 reserved_blocks;
 538         __u16 unused;
 539         __u32 free_blocks_count;
 540 };
 541
 542 /* Indexes used to index group tables in ext4_new_group_data */
 543 enum {
 544         BLOCK_BITMAP = 0,       /* block bitmap */
 545         INODE_BITMAP,           /* inode bitmap */
 546         INODE_TABLE,            /* inode tables */
 547         GROUP_TABLE_COUNT,
 548 };
 549
 550 /*
 551  * Flags used by ext4_map_blocks()
 552  */
 553         /* Allocate any needed blocks and/or convert an unitialized
 554            extent to be an initialized ext4 */
 555 #define EXT4_GET_BLOCKS_CREATE                  0x0001
 556         /* Request the creation of an unitialized extent */
 557 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
 558 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
 559                                                  EXT4_GET_BLOCKS_CREATE)
 560         /* Caller is from the delayed allocation writeout path,
 561            so set the magic i_delalloc_reserve_flag after taking the
 562            inode allocation semaphore for */
 563 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
 564         /* caller is from the direct IO path, request to creation of an
 565         unitialized extents if not allocated, split the uninitialized
 566         extent if blocks has been preallocated already*/
 567 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
 568 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
 569 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
 570                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
 571         /* Convert extent to initialized after IO complete */
 572 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
 573                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
 574         /* Punch out blocks of an extent */
 575 #define EXT4_GET_BLOCKS_PUNCH_OUT_EXT           0x0020
 576         /* Don't normalize allocation size (used for fallocate) */
 577 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
 578         /* Request will not result in inode size update (user for fallocate) */
 579 #define EXT4_GET_BLOCKS_KEEP_SIZE               0x0080
 580         /* Do not take i_data_sem locking in ext4_map_blocks */
 581 #define EXT4_GET_BLOCKS_NO_LOCK                 0x0100
 582
 583 /*
 584  * Flags used by ext4_free_blocks
 585  */
 586 #define EXT4_FREE_BLOCKS_METADATA       0x0001
 587 #define EXT4_FREE_BLOCKS_FORGET         0x0002
 588 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
 589 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
 590 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
 591 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
 592
 593 /*
 594  * Flags used by ext4_discard_partial_page_buffers
 595  */
 596 #define EXT4_DISCARD_PARTIAL_PG_ZERO_UNMAPPED   0x0001
 597
 598 /*
 599  * ioctl commands
 600  */
 601 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
 602 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
 603 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
 604 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
 605 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
 606 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
 607 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
 608 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
 609 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
 610 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
 611 #define EXT4_IOC_MIGRATE                _IO('f', 9)
 612  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
 613  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
 614 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
 615 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
 616 #define EXT4_IOC_RESIZE_FS              _IOW('f', 16, __u64)
 617
 618 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 619 /*
 620  * ioctl commands in 32 bit emulation
 621  */
 622 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
 623 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
 624 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
 625 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
 626 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
 627 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
 628 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
 629 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
 630 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
 631 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
 632 #endif
 633
 634 /* Max physical block we can address w/o extents */
 635 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
 636
 637 /*
 638  * Structure of an inode on the disk
 639  */
 640 struct ext4_inode {
 641         __le16  i_mode;         /* File mode */
 642         __le16  i_uid;          /* Low 16 bits of Owner Uid */
 643         __le32  i_size_lo;      /* Size in bytes */
 644         __le32  i_atime;        /* Access time */
 645         __le32  i_ctime;        /* Inode Change time */
 646         __le32  i_mtime;        /* Modification time */
 647         __le32  i_dtime;        /* Deletion Time */
 648         __le16  i_gid;          /* Low 16 bits of Group Id */
 649         __le16  i_links_count;  /* Links count */
 650         __le32  i_blocks_lo;    /* Blocks count */
 651         __le32  i_flags;        /* File flags */
 652         union {
 653                 struct {
 654                         __le32  l_i_version;
 655                 } linux1;
 656                 struct {
 657                         __u32  h_i_translator;
 658                 } hurd1;
 659                 struct {
 660                         __u32  m_i_reserved1;
 661                 } masix1;
 662         } osd1;                         /* OS dependent 1 */
 663         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
 664         __le32  i_generation;   /* File version (for NFS) */
 665         __le32  i_file_acl_lo;  /* File ACL */
 666         __le32  i_size_high;
 667         __le32  i_obso_faddr;   /* Obsoleted fragment address */
 668         union {
 669                 struct {
 670                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
 671                         __le16  l_i_file_acl_high;
 672                         __le16  l_i_uid_high;   /* these 2 fields */
 673                         __le16  l_i_gid_high;   /* were reserved2[0] */
 674                         __le16  l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
 675                         __le16  l_i_reserved;
 676                 } linux2;
 677                 struct {
 678                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 679                         __u16   h_i_mode_high;
 680                         __u16   h_i_uid_high;
 681                         __u16   h_i_gid_high;
 682                         __u32   h_i_author;
 683                 } hurd2;
 684                 struct {
 685                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 686                         __le16  m_i_file_acl_high;
 687                         __u32   m_i_reserved2[2];
 688                 } masix2;
 689         } osd2;                         /* OS dependent 2 */
 690         __le16  i_extra_isize;
 691         __le16  i_checksum_hi;  /* crc32c(uuid+inum+inode) BE */
 692         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
 693         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
 694         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
 695         __le32  i_crtime;       /* File Creation time */
 696         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
 697         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
 698 };
 699
 700 struct move_extent {
 701         __u32 reserved;         /* should be zero */
 702         __u32 donor_fd;         /* donor file descriptor */
 703         __u64 orig_start;       /* logical start offset in block for orig */
 704         __u64 donor_start;      /* logical start offset in block for donor */
 705         __u64 len;              /* block length to be moved */
 706         __u64 moved_len;        /* moved block length */
 707 };
 708
 709 #define EXT4_EPOCH_BITS 2
 710 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
 711 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
 712
 713 /*
 714  * Extended fields will fit into an inode if the filesystem was formatted
 715  * with large inodes (-I 256 or larger) and there are not currently any EAs
 716  * consuming all of the available space. For new inodes we always reserve
 717  * enough space for the kernel's known extended fields, but for inodes
 718  * created with an old kernel this might not have been the case. None of
 719  * the extended inode fields is critical for correct filesystem operation.
 720  * This macro checks if a certain field fits in the inode. Note that
 721  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
 722  */
 723 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
 724         ((offsetof(typeof(*ext4_inode), field) +        \
 725           sizeof((ext4_inode)->field))                  \
 726         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
 727             (einode)->i_extra_isize))                   \
 728
 729 static inline __le32 ext4_encode_extra_time(struct timespec *time)
 730 {
 731        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
 732                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
 733                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
 734 }
 735
 736 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
 737 {
 738        if (sizeof(time->tv_sec) > 4)
 739                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
 740                                << 32;
 741        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
 742 }
 743
 744 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
 745 do {                                                                           \
 746         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
 747         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 748                 (raw_inode)->xtime ## _extra =                                 \
 749                                 ext4_encode_extra_time(&(inode)->xtime);       \
 750 } while (0)
 751
 752 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
 753 do {                                                                           \
 754         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 755                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
 756         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 757                 (raw_inode)->xtime ## _extra =                                 \
 758                                 ext4_encode_extra_time(&(einode)->xtime);      \
 759 } while (0)
 760
 761 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
 762 do {                                                                           \
 763         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
 764         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 765                 ext4_decode_extra_time(&(inode)->xtime,                        \
 766                                        raw_inode->xtime ## _extra);            \
 767         else                                                                   \
 768                 (inode)->xtime.tv_nsec = 0;                                    \
 769 } while (0)
 770
 771 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
 772 do {                                                                           \
 773         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 774                 (einode)->xtime.tv_sec =                                       \
 775                         (signed)le32_to_cpu((raw_inode)->xtime);               \
 776         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 777                 ext4_decode_extra_time(&(einode)->xtime,                       \
 778                                        raw_inode->xtime ## _extra);            \
 779         else                                                                   \
 780                 (einode)->xtime.tv_nsec = 0;                                   \
 781 } while (0)
 782
 783 #define i_disk_version osd1.linux1.l_i_version
 784
 785 #if defined(__KERNEL__) || defined(__linux__)
 786 #define i_reserved1     osd1.linux1.l_i_reserved1
 787 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
 788 #define i_blocks_high   osd2.linux2.l_i_blocks_high
 789 #define i_uid_low       i_uid
 790 #define i_gid_low       i_gid
 791 #define i_uid_high      osd2.linux2.l_i_uid_high
 792 #define i_gid_high      osd2.linux2.l_i_gid_high
 793 #define i_checksum_lo   osd2.linux2.l_i_checksum_lo
 794
 795 #elif defined(__GNU__)
 796
 797 #define i_translator    osd1.hurd1.h_i_translator
 798 #define i_uid_high      osd2.hurd2.h_i_uid_high
 799 #define i_gid_high      osd2.hurd2.h_i_gid_high
 800 #define i_author        osd2.hurd2.h_i_author
 801
 802 #elif defined(__masix__)
 803
 804 #define i_reserved1     osd1.masix1.m_i_reserved1
 805 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
 806 #define i_reserved2     osd2.masix2.m_i_reserved2
 807
 808 #endif /* defined(__KERNEL__) || defined(__linux__) */
 809
 810 /*
 811  * storage for cached extent
 812  * If ec_len == 0, then the cache is invalid.
 813  * If ec_start == 0, then the cache represents a gap (null mapping)
 814  */
 815 struct ext4_ext_cache {
 816         ext4_fsblk_t    ec_start;
 817         ext4_lblk_t     ec_block;
 818         __u32           ec_len; /* must be 32bit to return holes */
 819 };
 820
 821 #include "extents_status.h"
 822
 823 /*
 824  * fourth extended file system inode data in memory
 825  */
 826 struct ext4_inode_info {
 827         __le32  i_data[15];     /* unconverted */
 828         __u32   i_dtime;
 829         ext4_fsblk_t    i_file_acl;
 830
 831         /*
 832          * i_block_group is the number of the block group which contains
 833          * this file's inode.  Constant across the lifetime of the inode,
 834          * it is ued for making block allocation decisions - we try to
 835          * place a file's data blocks near its inode block, and new inodes
 836          * near to their parent directory's inode.
 837          */
 838         ext4_group_t    i_block_group;
 839         ext4_lblk_t     i_dir_start_lookup;
 840 #if (BITS_PER_LONG < 64)
 841         unsigned long   i_state_flags;          /* Dynamic state flags */
 842 #endif
 843         unsigned long   i_flags;
 844
 845         /*
 846          * Extended attributes can be read independently of the main file
 847          * data. Taking i_mutex even when reading would cause contention
 848          * between readers of EAs and writers of regular file data, so
 849          * instead we synchronize on xattr_sem when reading or changing
 850          * EAs.
 851          */
 852         struct rw_semaphore xattr_sem;
 853
 854         struct list_head i_orphan;      /* unlinked but open inodes */
 855
 856         /*
 857          * i_disksize keeps track of what the inode size is ON DISK, not
 858          * in memory.  During truncate, i_size is set to the new size by
 859          * the VFS prior to calling ext4_truncate(), but the filesystem won't
 860          * set i_disksize to 0 until the truncate is actually under way.
 861          *
 862          * The intent is that i_disksize always represents the blocks which
 863          * are used by this file.  This allows recovery to restart truncate
 864          * on orphans if we crash during truncate.  We actually write i_disksize
 865          * into the on-disk inode when writing inodes out, instead of i_size.
 866          *
 867          * The only time when i_disksize and i_size may be different is when
 868          * a truncate is in progress.  The only things which change i_disksize
 869          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
 870          */
 871         loff_t  i_disksize;
 872
 873         /*
 874          * i_data_sem is for serialising ext4_truncate() against
 875          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
 876          * data tree are chopped off during truncate. We can't do that in
 877          * ext4 because whenever we perform intermediate commits during
 878          * truncate, the inode and all the metadata blocks *must* be in a
 879          * consistent state which allows truncation of the orphans to restart
 880          * during recovery.  Hence we must fix the get_block-vs-truncate race
 881          * by other means, so we have i_data_sem.
 882          */
 883         struct rw_semaphore i_data_sem;
 884         struct inode vfs_inode;
 885         struct jbd2_inode *jinode;
 886
 887         struct ext4_ext_cache i_cached_extent;
 888         /*
 889          * File creation time. Its function is same as that of
 890          * struct timespec i_{a,c,m}time in the generic inode.
 891          */
 892         struct timespec i_crtime;
 893
 894         /* mballoc */
 895         struct list_head i_prealloc_list;
 896         spinlock_t i_prealloc_lock;
 897
 898         /* extents status tree */
 899         struct ext4_es_tree i_es_tree;
 900         rwlock_t i_es_lock;
 901
 902         /* ialloc */
 903         ext4_group_t    i_last_alloc_group;
 904
 905         /* allocation reservation info for delalloc */
 906         /* In case of bigalloc, these refer to clusters rather than blocks */
 907         unsigned int i_reserved_data_blocks;
 908         unsigned int i_reserved_meta_blocks;
 909         unsigned int i_allocated_meta_blocks;
 910         ext4_lblk_t i_da_metadata_calc_last_lblock;
 911         int i_da_metadata_calc_len;
 912
 913         /* on-disk additional length */
 914         __u16 i_extra_isize;
 915
 916         /* Indicate the inline data space. */
 917         u16 i_inline_off;
 918         u16 i_inline_size;
 919
 920 #ifdef CONFIG_QUOTA
 921         /* quota space reservation, managed internally by quota code */
 922         qsize_t i_reserved_quota;
 923 #endif
 924
 925         /* completed IOs that might need unwritten extents handling */
 926         struct list_head i_completed_io_list;
 927         spinlock_t i_completed_io_lock;
 928         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
 929         atomic_t i_unwritten; /* Nr. of inflight conversions pending */
 930         struct work_struct i_unwritten_work;    /* deferred extent conversion */
 931
 932         spinlock_t i_block_reservation_lock;
 933
 934         /*
 935          * Transactions that contain inode's metadata needed to complete
 936          * fsync and fdatasync, respectively.
 937          */
 938         tid_t i_sync_tid;
 939         tid_t i_datasync_tid;
 940
 941         /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
 942         __u32 i_csum_seed;
 943 };
 944
 945 /*
 946  * File system states
 947  */
 948 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
 949 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
 950 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
 951
 952 /*
 953  * Misc. filesystem flags
 954  */
 955 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
 956 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
 957 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
 958
 959 /*
 960  * Mount flags
 961  */
 962 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
 963 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
 964 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
 965 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
 966 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
 967 #define EXT4_MOUNT_ERRORS_MASK          0x00070
 968 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
 969 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
 970 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
 971 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
 972 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
 973 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
 974 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
 975 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
 976 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
 977 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
 978 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
 979 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
 980 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
 981 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
 982 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
 983 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
 984 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
 985 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
 986 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
 987 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
 988 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
 989 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
 990 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
 991
 992 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
 993                                                       specified delalloc */
 994
 995 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
 996                                                 ~EXT4_MOUNT_##opt
 997 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
 998                                                 EXT4_MOUNT_##opt
 999 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
1000                                          EXT4_MOUNT_##opt)
1001
1002 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
1003                                                 ~EXT4_MOUNT2_##opt
1004 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
1005                                                 EXT4_MOUNT2_##opt
1006 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
1007                                          EXT4_MOUNT2_##opt)
1008
1009 #define ext4_test_and_set_bit           __test_and_set_bit_le
1010 #define ext4_set_bit                    __set_bit_le
1011 #define ext4_set_bit_atomic             ext2_set_bit_atomic
1012 #define ext4_test_and_clear_bit         __test_and_clear_bit_le
1013 #define ext4_clear_bit                  __clear_bit_le
1014 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
1015 #define ext4_test_bit                   test_bit_le
1016 #define ext4_find_next_zero_bit         find_next_zero_bit_le
1017 #define ext4_find_next_bit              find_next_bit_le
1018
1019 extern void ext4_set_bits(void *bm, int cur, int len);
1020
1021 /*
1022  * Maximal mount counts between two filesystem checks
1023  */
1024 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
1025 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
1026
1027 /*
1028  * Behaviour when detecting errors
1029  */
1030 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
1031 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
1032 #define EXT4_ERRORS_PANIC               3       /* Panic */
1033 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
1034
1035 /* Metadata checksum algorithm codes */
1036 #define EXT4_CRC32C_CHKSUM              1
1037
1038 /*
1039  * Structure of the super block
1040  */
1041 struct ext4_super_block {
1042 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1043         __le32  s_blocks_count_lo;      /* Blocks count */
1044         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1045         __le32  s_free_blocks_count_lo; /* Free blocks count */
1046 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1047         __le32  s_first_data_block;     /* First Data Block */
1048         __le32  s_log_block_size;       /* Block size */
1049         __le32  s_log_cluster_size;     /* Allocation cluster size */
1050 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1051         __le32  s_clusters_per_group;   /* # Clusters per group */
1052         __le32  s_inodes_per_group;     /* # Inodes per group */
1053         __le32  s_mtime;                /* Mount time */
1054 /*30*/  __le32  s_wtime;                /* Write time */
1055         __le16  s_mnt_count;            /* Mount count */
1056         __le16  s_max_mnt_count;        /* Maximal mount count */
1057         __le16  s_magic;                /* Magic signature */
1058         __le16  s_state;                /* File system state */
1059         __le16  s_errors;               /* Behaviour when detecting errors */
1060         __le16  s_minor_rev_level;      /* minor revision level */
1061 /*40*/  __le32  s_lastcheck;            /* time of last check */
1062         __le32  s_checkinterval;        /* max. time between checks */
1063         __le32  s_creator_os;           /* OS */
1064         __le32  s_rev_level;            /* Revision level */
1065 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1066         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1067         /*
1068          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1069          *
1070          * Note: the difference between the compatible feature set and
1071          * the incompatible feature set is that if there is a bit set
1072          * in the incompatible feature set that the kernel doesn't
1073          * know about, it should refuse to mount the filesystem.
1074          *
1075          * e2fsck's requirements are more strict; if it doesn't know
1076          * about a feature in either the compatible or incompatible
1077          * feature set, it must abort and not try to meddle with
1078          * things it doesn't understand...
1079          */
1080         __le32  s_first_ino;            /* First non-reserved inode */
1081         __le16  s_inode_size;           /* size of inode structure */
1082         __le16  s_block_group_nr;       /* block group # of this superblock */
1083         __le32  s_feature_compat;       /* compatible feature set */
1084 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1085         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1086 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1087 /*78*/  char    s_volume_name[16];      /* volume name */
1088 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1089 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1090         /*
1091          * Performance hints.  Directory preallocation should only
1092          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1093          */
1094         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1095         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1096         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1097         /*
1098          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1099          */
1100 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1101 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1102         __le32  s_journal_dev;          /* device number of journal file */
1103         __le32  s_last_orphan;          /* start of list of inodes to delete */
1104         __le32  s_hash_seed[4];         /* HTREE hash seed */
1105         __u8    s_def_hash_version;     /* Default hash version to use */
1106         __u8    s_jnl_backup_type;
1107         __le16  s_desc_size;            /* size of group descriptor */
1108 /*100*/ __le32  s_default_mount_opts;
1109         __le32  s_first_meta_bg;        /* First metablock block group */
1110         __le32  s_mkfs_time;            /* When the filesystem was created */
1111         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1112         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1113 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1114         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1115         __le32  s_free_blocks_count_hi; /* Free blocks count */
1116         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1117         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1118         __le32  s_flags;                /* Miscellaneous flags */
1119         __le16  s_raid_stride;          /* RAID stride */
1120         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1121         __le64  s_mmp_block;            /* Block for multi-mount protection */
1122         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1123         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1124         __u8    s_checksum_type;        /* metadata checksum algorithm used */
1125         __le16  s_reserved_pad;
1126         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1127         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1128         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1129         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1130                                               snapshot's future use */
1131         __le32  s_snapshot_list;        /* inode number of the head of the
1132                                            on-disk snapshot list */
1133 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1134         __le32  s_error_count;          /* number of fs errors */
1135         __le32  s_first_error_time;     /* first time an error happened */
1136         __le32  s_first_error_ino;      /* inode involved in first error */
1137         __le64  s_first_error_block;    /* block involved of first error */
1138         __u8    s_first_error_func[32]; /* function where the error happened */
1139         __le32  s_first_error_line;     /* line number where error happened */
1140         __le32  s_last_error_time;      /* most recent time of an error */
1141         __le32  s_last_error_ino;       /* inode involved in last error */
1142         __le32  s_last_error_line;      /* line number where error happened */
1143         __le64  s_last_error_block;     /* block involved of last error */
1144         __u8    s_last_error_func[32];  /* function where the error happened */
1145 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1146         __u8    s_mount_opts[64];
1147         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1148         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1149         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1150         __le32  s_reserved[108];        /* Padding to the end of the block */
1151         __le32  s_checksum;             /* crc32c(superblock) */
1152 };
1153
1154 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1155
1156 #ifdef __KERNEL__
1157
1158 /*
1159  * run-time mount flags
1160  */
1161 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
1162 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
1163
1164 /*
1165  * fourth extended-fs super-block data in memory
1166  */
1167 struct ext4_sb_info {
1168         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1169         unsigned long s_inodes_per_block;/* Number of inodes per block */
1170         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1171         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1172         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1173         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1174         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1175         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1176         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1177         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1178         unsigned long s_overhead;  /* # of fs overhead clusters */
1179         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1180         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1181         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1182         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1183         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1184         struct buffer_head **s_group_desc;
1185         unsigned int s_mount_opt;
1186         unsigned int s_mount_opt2;
1187         unsigned int s_mount_flags;
1188         unsigned int s_def_mount_opt;
1189         ext4_fsblk_t s_sb_block;
1190         kuid_t s_resuid;
1191         kgid_t s_resgid;
1192         unsigned short s_mount_state;
1193         unsigned short s_pad;
1194         int s_addr_per_block_bits;
1195         int s_desc_per_block_bits;
1196         int s_inode_size;
1197         int s_first_ino;
1198         unsigned int s_inode_readahead_blks;
1199         unsigned int s_inode_goal;
1200         spinlock_t s_next_gen_lock;
1201         u32 s_next_generation;
1202         u32 s_hash_seed[4];
1203         int s_def_hash_version;
1204         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1205         struct percpu_counter s_freeclusters_counter;
1206         struct percpu_counter s_freeinodes_counter;
1207         struct percpu_counter s_dirs_counter;
1208         struct percpu_counter s_dirtyclusters_counter;
1209         struct blockgroup_lock *s_blockgroup_lock;
1210         struct proc_dir_entry *s_proc;
1211         struct kobject s_kobj;
1212         struct completion s_kobj_unregister;
1213         struct super_block *s_sb;
1214
1215         /* Journaling */
1216         struct journal_s *s_journal;
1217         struct list_head s_orphan;
1218         struct mutex s_orphan_lock;
1219         unsigned long s_resize_flags;           /* Flags indicating if there
1220                                                    is a resizer */
1221         unsigned long s_commit_interval;
1222         u32 s_max_batch_time;
1223         u32 s_min_batch_time;
1224         struct block_device *journal_bdev;
1225 #ifdef CONFIG_QUOTA
1226         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
1227         int s_jquota_fmt;                       /* Format of quota to use */
1228 #endif
1229         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1230         struct rb_root system_blks;
1231
1232 #ifdef EXTENTS_STATS
1233         /* ext4 extents stats */
1234         unsigned long s_ext_min;
1235         unsigned long s_ext_max;
1236         unsigned long s_depth_max;
1237         spinlock_t s_ext_stats_lock;
1238         unsigned long s_ext_blocks;
1239         unsigned long s_ext_extents;
1240 #endif
1241
1242         /* for buddy allocator */
1243         struct ext4_group_info ***s_group_info;
1244         struct inode *s_buddy_cache;
1245         spinlock_t s_md_lock;
1246         unsigned short *s_mb_offsets;
1247         unsigned int *s_mb_maxs;
1248         unsigned int s_group_info_size;
1249
1250         /* tunables */
1251         unsigned long s_stripe;
1252         unsigned int s_mb_stream_request;
1253         unsigned int s_mb_max_to_scan;
1254         unsigned int s_mb_min_to_scan;
1255         unsigned int s_mb_stats;
1256         unsigned int s_mb_order2_reqs;
1257         unsigned int s_mb_group_prealloc;
1258         unsigned int s_max_writeback_mb_bump;
1259         unsigned int s_max_dir_size_kb;
1260         /* where last allocation was done - for stream allocation */
1261         unsigned long s_mb_last_group;
1262         unsigned long s_mb_last_start;
1263
1264         /* stats for buddy allocator */
1265         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1266         atomic_t s_bal_success; /* we found long enough chunks */
1267         atomic_t s_bal_allocated;       /* in blocks */
1268         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1269         atomic_t s_bal_goals;   /* goal hits */
1270         atomic_t s_bal_breaks;  /* too long searches */
1271         atomic_t s_bal_2orders; /* 2^order hits */
1272         spinlock_t s_bal_lock;
1273         unsigned long s_mb_buddies_generated;
1274         unsigned long long s_mb_generation_time;
1275         atomic_t s_mb_lost_chunks;
1276         atomic_t s_mb_preallocated;
1277         atomic_t s_mb_discarded;
1278         atomic_t s_lock_busy;
1279
1280         /* locality groups */
1281         struct ext4_locality_group __percpu *s_locality_groups;
1282
1283         /* for write statistics */
1284         unsigned long s_sectors_written_start;
1285         u64 s_kbytes_written;
1286
1287         /* the size of zero-out chunk */
1288         unsigned int s_extent_max_zeroout_kb;
1289
1290         unsigned int s_log_groups_per_flex;
1291         struct flex_groups *s_flex_groups;
1292         ext4_group_t s_flex_groups_allocated;
1293
1294         /* workqueue for dio unwritten */
1295         struct workqueue_struct *dio_unwritten_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
1318 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1319 {
1320         return sb->s_fs_info;
1321 }
1322 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1323 {
1324         return container_of(inode, struct ext4_inode_info, vfs_inode);
1325 }
1326
1327 static inline struct timespec ext4_current_time(struct inode *inode)
1328 {
1329         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1330                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1331 }
1332
1333 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1334 {
1335         return ino == EXT4_ROOT_INO ||
1336                 ino == EXT4_USR_QUOTA_INO ||
1337                 ino == EXT4_GRP_QUOTA_INO ||
1338                 ino == EXT4_JOURNAL_INO ||
1339                 ino == EXT4_RESIZE_INO ||
1340                 (ino >= EXT4_FIRST_INO(sb) &&
1341                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1342 }
1343
1344 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
1345                                               struct ext4_io_end *io_end)
1346 {
1347         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
1348                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
1349                 atomic_inc(&EXT4_I(inode)->i_unwritten);
1350         }
1351 }
1352
1353 static inline ext4_io_end_t *ext4_inode_aio(struct inode *inode)
1354 {
1355         return inode->i_private;
1356 }
1357
1358 static inline void ext4_inode_aio_set(struct inode *inode, ext4_io_end_t *io)
1359 {
1360         inode->i_private = io;
1361 }
1362
1363 /*
1364  * Inode dynamic state flags
1365  */
1366 enum {
1367         EXT4_STATE_JDATA,               /* journaled data exists */
1368         EXT4_STATE_NEW,                 /* inode is newly created */
1369         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1370         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1371         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1372         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1373         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1374         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1375         EXT4_STATE_DELALLOC_RESERVED,   /* blks already reserved for delalloc */
1376         EXT4_STATE_DIOREAD_LOCK,        /* Disable support for dio read
1377                                            nolocking */
1378         EXT4_STATE_MAY_INLINE_DATA,     /* may have in-inode data */
1379 };
1380
1381 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1382 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1383 {                                                                       \
1384         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1385 }                                                                       \
1386 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1387 {                                                                       \
1388         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1389 }                                                                       \
1390 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1391 {                                                                       \
1392         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1393 }
1394
1395 EXT4_INODE_BIT_FNS(flag, flags, 0)
1396 #if (BITS_PER_LONG < 64)
1397 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1398
1399 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1400 {
1401         (ei)->i_state_flags = 0;
1402 }
1403 #else
1404 EXT4_INODE_BIT_FNS(state, flags, 32)
1405
1406 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1407 {
1408         /* We depend on the fact that callers will set i_flags */
1409 }
1410 #endif
1411 #else
1412 /* Assume that user mode programs are passing in an ext4fs superblock, not
1413  * a kernel struct super_block.  This will allow us to call the feature-test
1414  * macros from user land. */
1415 #define EXT4_SB(sb)     (sb)
1416 #endif
1417
1418 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1419
1420 /*
1421  * Codes for operating systems
1422  */
1423 #define EXT4_OS_LINUX           0
1424 #define EXT4_OS_HURD            1
1425 #define EXT4_OS_MASIX           2
1426 #define EXT4_OS_FREEBSD         3
1427 #define EXT4_OS_LITES           4
1428
1429 /*
1430  * Revision levels
1431  */
1432 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1433 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1434
1435 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1436 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1437
1438 #define EXT4_GOOD_OLD_INODE_SIZE 128
1439
1440 /*
1441  * Feature set definitions
1442  */
1443
1444 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1445         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1446 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1447         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1448 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1449         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1450 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1451         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1452 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1453         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1454 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1455         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1456 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1457         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1458 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1459         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1460 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1461         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1462
1463 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1464 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1465 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1466 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1467 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1468 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1469
1470 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1471 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1472 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1473 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1474 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1475 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1476 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1477 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1478 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1479 /*
1480  * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM).  When
1481  * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
1482  * all other data structures' checksums.  However, the METADATA_CSUM and
1483  * GDT_CSUM bits are mutually exclusive.
1484  */
1485 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM    0x0400
1486
1487 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1488 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1489 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1490 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1491 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1492 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1493 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1494 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1495 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1496 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1497 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1498 #define EXT4_FEATURE_INCOMPAT_BG_USE_META_CSUM  0x2000 /* use crc32c for bg */
1499 #define EXT4_FEATURE_INCOMPAT_LARGEDIR          0x4000 /* >2GB or 3-lvl htree */
1500 #define EXT4_FEATURE_INCOMPAT_INLINE_DATA       0x8000 /* data in inode */
1501
1502 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1503 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1504                                          EXT4_FEATURE_INCOMPAT_META_BG)
1505 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1506                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1507                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1508
1509 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1510 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1511                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1512                                          EXT4_FEATURE_INCOMPAT_META_BG)
1513 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1514                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1515                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1516
1517 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1518 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1519                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1520                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1521                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1522                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1523                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1524                                          EXT4_FEATURE_INCOMPAT_MMP |    \
1525                                          EXT4_FEATURE_INCOMPAT_INLINE_DATA)
1526 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1527                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1528                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1529                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1530                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1531                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1532                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1533                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
1534                                          EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
1535                                          EXT4_FEATURE_RO_COMPAT_QUOTA)
1536
1537 /*
1538  * Default values for user and/or group using reserved blocks
1539  */
1540 #define EXT4_DEF_RESUID         0
1541 #define EXT4_DEF_RESGID         0
1542
1543 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1544
1545 /*
1546  * Default mount options
1547  */
1548 #define EXT4_DEFM_DEBUG         0x0001
1549 #define EXT4_DEFM_BSDGROUPS     0x0002
1550 #define EXT4_DEFM_XATTR_USER    0x0004
1551 #define EXT4_DEFM_ACL           0x0008
1552 #define EXT4_DEFM_UID16         0x0010
1553 #define EXT4_DEFM_JMODE         0x0060
1554 #define EXT4_DEFM_JMODE_DATA    0x0020
1555 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1556 #define EXT4_DEFM_JMODE_WBACK   0x0060
1557 #define EXT4_DEFM_NOBARRIER     0x0100
1558 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1559 #define EXT4_DEFM_DISCARD       0x0400
1560 #define EXT4_DEFM_NODELALLOC    0x0800
1561
1562 /*
1563  * Default journal batch times
1564  */
1565 #define EXT4_DEF_MIN_BATCH_TIME 0
1566 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1567
1568 /*
1569  * Minimum number of groups in a flexgroup before we separate out
1570  * directories into the first block group of a flexgroup
1571  */
1572 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1573
1574 /*
1575  * Structure of a directory entry
1576  */
1577 #define EXT4_NAME_LEN 255
1578
1579 struct ext4_dir_entry {
1580         __le32  inode;                  /* Inode number */
1581         __le16  rec_len;                /* Directory entry length */
1582         __le16  name_len;               /* Name length */
1583         char    name[EXT4_NAME_LEN];    /* File name */
1584 };
1585
1586 /*
1587  * The new version of the directory entry.  Since EXT4 structures are
1588  * stored in intel byte order, and the name_len field could never be
1589  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1590  * file_type field.
1591  */
1592 struct ext4_dir_entry_2 {
1593         __le32  inode;                  /* Inode number */
1594         __le16  rec_len;                /* Directory entry length */
1595         __u8    name_len;               /* Name length */
1596         __u8    file_type;
1597         char    name[EXT4_NAME_LEN];    /* File name */
1598 };
1599
1600 /*
1601  * This is a bogus directory entry at the end of each leaf block that
1602  * records checksums.
1603  */
1604 struct ext4_dir_entry_tail {
1605         __le32  det_reserved_zero1;     /* Pretend to be unused */
1606         __le16  det_rec_len;            /* 12 */
1607         __u8    det_reserved_zero2;     /* Zero name length */
1608         __u8    det_reserved_ft;        /* 0xDE, fake file type */
1609         __le32  det_checksum;           /* crc32c(uuid+inum+dirblock) */
1610 };
1611
1612 #define EXT4_DIRENT_TAIL(block, blocksize) \
1613         ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
1614                                         ((blocksize) - \
1615                                          sizeof(struct ext4_dir_entry_tail))))
1616
1617 /*
1618  * Ext4 directory file types.  Only the low 3 bits are used.  The
1619  * other bits are reserved for now.
1620  */
1621 #define EXT4_FT_UNKNOWN         0
1622 #define EXT4_FT_REG_FILE        1
1623 #define EXT4_FT_DIR             2
1624 #define EXT4_FT_CHRDEV          3
1625 #define EXT4_FT_BLKDEV          4
1626 #define EXT4_FT_FIFO            5
1627 #define EXT4_FT_SOCK            6
1628 #define EXT4_FT_SYMLINK         7
1629
1630 #define EXT4_FT_MAX             8
1631
1632 #define EXT4_FT_DIR_CSUM        0xDE
1633
1634 /*
1635  * EXT4_DIR_PAD defines the directory entries boundaries
1636  *
1637  * NOTE: It must be a multiple of 4
1638  */
1639 #define EXT4_DIR_PAD                    4
1640 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1641 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1642                                          ~EXT4_DIR_ROUND)
1643 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1644
1645 /*
1646  * If we ever get support for fs block sizes > page_size, we'll need
1647  * to remove the #if statements in the next two functions...
1648  */
1649 static inline unsigned int
1650 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1651 {
1652         unsigned len = le16_to_cpu(dlen);
1653
1654 #if (PAGE_CACHE_SIZE >= 65536)
1655         if (len == EXT4_MAX_REC_LEN || len == 0)
1656                 return blocksize;
1657         return (len & 65532) | ((len & 3) << 16);
1658 #else
1659         return len;
1660 #endif
1661 }
1662
1663 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1664 {
1665         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1666                 BUG();
1667 #if (PAGE_CACHE_SIZE >= 65536)
1668         if (len < 65536)
1669                 return cpu_to_le16(len);
1670         if (len == blocksize) {
1671                 if (blocksize == 65536)
1672                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1673                 else
1674                         return cpu_to_le16(0);
1675         }
1676         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1677 #else
1678         return cpu_to_le16(len);
1679 #endif
1680 }
1681
1682 /*
1683  * Hash Tree Directory indexing
1684  * (c) Daniel Phillips, 2001
1685  */
1686
1687 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1688                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1689                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1690 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1691 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1692
1693 /* Legal values for the dx_root hash_version field: */
1694
1695 #define DX_HASH_LEGACY          0
1696 #define DX_HASH_HALF_MD4        1
1697 #define DX_HASH_TEA             2
1698 #define DX_HASH_LEGACY_UNSIGNED 3
1699 #define DX_HASH_HALF_MD4_UNSIGNED       4
1700 #define DX_HASH_TEA_UNSIGNED            5
1701
1702 static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
1703                               const void *address, unsigned int length)
1704 {
1705         struct {
1706                 struct shash_desc shash;
1707                 char ctx[4];
1708         } desc;
1709         int err;
1710
1711         BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
1712
1713         desc.shash.tfm = sbi->s_chksum_driver;
1714         desc.shash.flags = 0;
1715         *(u32 *)desc.ctx = crc;
1716
1717         err = crypto_shash_update(&desc.shash, address, length);
1718         BUG_ON(err);
1719
1720         return *(u32 *)desc.ctx;
1721 }
1722
1723 #ifdef __KERNEL__
1724
1725 /* hash info structure used by the directory hash */
1726 struct dx_hash_info
1727 {
1728         u32             hash;
1729         u32             minor_hash;
1730         int             hash_version;
1731         u32             *seed;
1732 };
1733
1734
1735 /* 32 and 64 bit signed EOF for dx directories */
1736 #define EXT4_HTREE_EOF_32BIT   ((1UL  << (32 - 1)) - 1)
1737 #define EXT4_HTREE_EOF_64BIT   ((1ULL << (64 - 1)) - 1)
1738
1739
1740 /*
1741  * Control parameters used by ext4_htree_next_block
1742  */
1743 #define HASH_NB_ALWAYS          1
1744
1745
1746 /*
1747  * Describe an inode's exact location on disk and in memory
1748  */
1749 struct ext4_iloc
1750 {
1751         struct buffer_head *bh;
1752         unsigned long offset;
1753         ext4_group_t block_group;
1754 };
1755
1756 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1757 {
1758         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1759 }
1760
1761 /*
1762  * This structure is stuffed into the struct file's private_data field
1763  * for directories.  It is where we put information so that we can do
1764  * readdir operations in hash tree order.
1765  */
1766 struct dir_private_info {
1767         struct rb_root  root;
1768         struct rb_node  *curr_node;
1769         struct fname    *extra_fname;
1770         loff_t          last_pos;
1771         __u32           curr_hash;
1772         __u32           curr_minor_hash;
1773         __u32           next_hash;
1774 };
1775
1776 /* calculate the first block number of the group */
1777 static inline ext4_fsblk_t
1778 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1779 {
1780         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1781                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1782 }
1783
1784 /*
1785  * Special error return code only used by dx_probe() and its callers.
1786  */
1787 #define ERR_BAD_DX_DIR  -75000
1788
1789 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1790                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1791
1792 /*
1793  * Timeout and state flag for lazy initialization inode thread.
1794  */
1795 #define EXT4_DEF_LI_WAIT_MULT                   10
1796 #define EXT4_DEF_LI_MAX_START_DELAY             5
1797 #define EXT4_LAZYINIT_QUIT                      0x0001
1798 #define EXT4_LAZYINIT_RUNNING                   0x0002
1799
1800 /*
1801  * Lazy inode table initialization info
1802  */
1803 struct ext4_lazy_init {
1804         unsigned long           li_state;
1805         struct list_head        li_request_list;
1806         struct mutex            li_list_mtx;
1807 };
1808
1809 struct ext4_li_request {
1810         struct super_block      *lr_super;
1811         struct ext4_sb_info     *lr_sbi;
1812         ext4_group_t            lr_next_group;
1813         struct list_head        lr_request;
1814         unsigned long           lr_next_sched;
1815         unsigned long           lr_timeout;
1816 };
1817
1818 struct ext4_features {
1819         struct kobject f_kobj;
1820         struct completion f_kobj_unregister;
1821 };
1822
1823 /*
1824  * This structure will be used for multiple mount protection. It will be
1825  * written into the block number saved in the s_mmp_block field in the
1826  * superblock. Programs that check MMP should assume that if
1827  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1828  * to use the filesystem, regardless of how old the timestamp is.
1829  */
1830 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
1831 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1832 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
1833 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
1834
1835 struct mmp_struct {
1836         __le32  mmp_magic;              /* Magic number for MMP */
1837         __le32  mmp_seq;                /* Sequence no. updated periodically */
1838
1839         /*
1840          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1841          * purposes and do not affect the correctness of the algorithm
1842          */
1843         __le64  mmp_time;               /* Time last updated */
1844         char    mmp_nodename[64];       /* Node which last updated MMP block */
1845         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
1846
1847         /*
1848          * mmp_check_interval is used to verify if the MMP block has been
1849          * updated on the block device. The value is updated based on the
1850          * maximum time to write the MMP block during an update cycle.
1851          */
1852         __le16  mmp_check_interval;
1853
1854         __le16  mmp_pad1;
1855         __le32  mmp_pad2[226];
1856         __le32  mmp_checksum;           /* crc32c(uuid+mmp_block) */
1857 };
1858
1859 /* arguments passed to the mmp thread */
1860 struct mmpd_data {
1861         struct buffer_head *bh; /* bh from initial read_mmp_block() */
1862         struct super_block *sb;  /* super block of the fs */
1863 };
1864
1865 /*
1866  * Check interval multiplier
1867  * The MMP block is written every update interval and initially checked every
1868  * update interval x the multiplier (the value is then adapted based on the
1869  * write latency). The reason is that writes can be delayed under load and we
1870  * don't want readers to incorrectly assume that the filesystem is no longer
1871  * in use.
1872  */
1873 #define EXT4_MMP_CHECK_MULT             2UL
1874
1875 /*
1876  * Minimum interval for MMP checking in seconds.
1877  */
1878 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
1879
1880 /*
1881  * Maximum interval for MMP checking in seconds.
1882  */
1883 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
1884
1885 /*
1886  * Function prototypes
1887  */
1888
1889 /*
1890  * Ok, these declarations are also in <linux/kernel.h> but none of the
1891  * ext4 source programs needs to include it so they are duplicated here.
1892  */
1893 # define NORET_TYPE     /**/
1894 # define ATTRIB_NORET   __attribute__((noreturn))
1895 # define NORET_AND      noreturn,
1896
1897 /* bitmap.c */
1898 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
1899 void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1900                                 struct ext4_group_desc *gdp,
1901                                 struct buffer_head *bh, int sz);
1902 int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1903                                   struct ext4_group_desc *gdp,
1904                                   struct buffer_head *bh, int sz);
1905 void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1906                                 struct ext4_group_desc *gdp,
1907                                 struct buffer_head *bh);
1908 int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1909                                   struct ext4_group_desc *gdp,
1910                                   struct buffer_head *bh);
1911
1912 /* balloc.c */
1913 extern void ext4_validate_block_bitmap(struct super_block *sb,
1914                                        struct ext4_group_desc *desc,
1915                                        unsigned int block_group,
1916                                        struct buffer_head *bh);
1917 extern unsigned int ext4_block_group(struct super_block *sb,
1918                         ext4_fsblk_t blocknr);
1919 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1920                         ext4_fsblk_t blocknr);
1921 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1922 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1923                         ext4_group_t group);
1924 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1925                                          ext4_fsblk_t goal,
1926                                          unsigned int flags,
1927                                          unsigned long *count,
1928                                          int *errp);
1929 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
1930                                     s64 nclusters, unsigned int flags);
1931 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
1932 extern void ext4_check_blocks_bitmap(struct super_block *);
1933 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1934                                                     ext4_group_t block_group,
1935                                                     struct buffer_head ** bh);
1936 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1937
1938 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
1939                                                 ext4_group_t block_group);
1940 extern int ext4_wait_block_bitmap(struct super_block *sb,
1941                                   ext4_group_t block_group,
1942                                   struct buffer_head *bh);
1943 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1944                                                   ext4_group_t block_group);
1945 extern void ext4_init_block_bitmap(struct super_block *sb,
1946                                    struct buffer_head *bh,
1947                                    ext4_group_t group,
1948                                    struct ext4_group_desc *desc);
1949 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
1950                                               ext4_group_t block_group,
1951                                               struct ext4_group_desc *gdp);
1952 extern unsigned ext4_num_overhead_clusters(struct super_block *sb,
1953                                            ext4_group_t block_group,
1954                                            struct ext4_group_desc *gdp);
1955 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
1956
1957 /* dir.c */
1958 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1959                                   struct file *,
1960                                   struct ext4_dir_entry_2 *,
1961                                   struct buffer_head *, char *, int,
1962                                   unsigned int);
1963 #define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset)      \
1964         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
1965                                         (de), (bh), (buf), (size), (offset)))
1966 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1967                                     __u32 minor_hash,
1968                                     struct ext4_dir_entry_2 *dirent);
1969 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1970 extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
1971                              struct buffer_head *bh,
1972                              void *buf, int buf_size,
1973                              const char *name, int namelen,
1974                              struct ext4_dir_entry_2 **dest_de);
1975 void ext4_insert_dentry(struct inode *inode,
1976                         struct ext4_dir_entry_2 *de,
1977                         int buf_size,
1978                         const char *name, int namelen);
1979 static inline void ext4_update_dx_flag(struct inode *inode)
1980 {
1981         if (!EXT4_HAS_COMPAT_FEATURE(inode->i_sb,
1982                                      EXT4_FEATURE_COMPAT_DIR_INDEX))
1983                 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
1984 }
1985 static unsigned char ext4_filetype_table[] = {
1986         DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1987 };
1988
1989 static inline  unsigned char get_dtype(struct super_block *sb, int filetype)
1990 {
1991         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE) ||
1992             (filetype >= EXT4_FT_MAX))
1993                 return DT_UNKNOWN;
1994
1995         return ext4_filetype_table[filetype];
1996 }
1997
1998 /* fsync.c */
1999 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2000 extern int ext4_flush_unwritten_io(struct inode *);
2001
2002 /* hash.c */
2003 extern int ext4fs_dirhash(const char *name, int len, struct
2004                           dx_hash_info *hinfo);
2005
2006 /* ialloc.c */
2007 extern struct inode *ext4_new_inode(handle_t *, struct inode *, umode_t,
2008                                     const struct qstr *qstr, __u32 goal,
2009                                     uid_t *owner);
2010 extern void ext4_free_inode(handle_t *, struct inode *);
2011 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2012 extern unsigned long ext4_count_free_inodes(struct super_block *);
2013 extern unsigned long ext4_count_dirs(struct super_block *);
2014 extern void ext4_check_inodes_bitmap(struct super_block *);
2015 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2016 extern int ext4_init_inode_table(struct super_block *sb,
2017                                  ext4_group_t group, int barrier);
2018 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2019
2020 /* mballoc.c */
2021 extern long ext4_mb_stats;
2022 extern long ext4_mb_max_to_scan;
2023 extern int ext4_mb_init(struct super_block *);
2024 extern int ext4_mb_release(struct super_block *);
2025 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2026                                 struct ext4_allocation_request *, int *);
2027 extern int ext4_mb_reserve_blocks(struct super_block *, int);
2028 extern void ext4_discard_preallocations(struct inode *);
2029 extern int __init ext4_init_mballoc(void);
2030 extern void ext4_exit_mballoc(void);
2031 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2032                              struct buffer_head *bh, ext4_fsblk_t block,
2033                              unsigned long count, int flags);
2034 extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2035                                    ext4_group_t ngroups);
2036 extern int ext4_mb_add_groupinfo(struct super_block *sb,
2037                 ext4_group_t i, struct ext4_group_desc *desc);
2038 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2039                                 ext4_fsblk_t block, unsigned long count);
2040 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2041
2042 /* inode.c */
2043 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
2044                                                 ext4_lblk_t, int, int *);
2045 struct buffer_head *ext4_bread(handle_t *, struct inode *,
2046                                                 ext4_lblk_t, int, int *);
2047 int ext4_get_block_write(struct inode *inode, sector_t iblock,
2048                          struct buffer_head *bh_result, int create);
2049 int ext4_get_block(struct inode *inode, sector_t iblock,
2050                                 struct buffer_head *bh_result, int create);
2051 int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2052                            struct buffer_head *bh, int create);
2053 int ext4_walk_page_buffers(handle_t *handle,
2054                            struct buffer_head *head,
2055                            unsigned from,
2056                            unsigned to,
2057                            int *partial,
2058                            int (*fn)(handle_t *handle,
2059                                      struct buffer_head *bh));
2060 int do_journal_get_write_access(handle_t *handle,
2061                                 struct buffer_head *bh);
2062 #define FALL_BACK_TO_NONDELALLOC 1
2063 #define CONVERT_INLINE_DATA      2
2064
2065 extern struct inode *ext4_iget(struct super_block *, unsigned long);
2066 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
2067 extern int  ext4_setattr(struct dentry *, struct iattr *);
2068 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
2069                                 struct kstat *stat);
2070 extern void ext4_evict_inode(struct inode *);
2071 extern void ext4_clear_inode(struct inode *);
2072 extern int  ext4_sync_inode(handle_t *, struct inode *);
2073 extern void ext4_dirty_inode(struct inode *, int);
2074 extern int ext4_change_inode_journal_flag(struct inode *, int);
2075 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
2076 extern int ext4_can_truncate(struct inode *inode);
2077 extern void ext4_truncate(struct inode *);
2078 extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length);
2079 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
2080 extern void ext4_set_inode_flags(struct inode *);
2081 extern void ext4_get_inode_flags(struct ext4_inode_info *);
2082 extern int ext4_alloc_da_blocks(struct inode *inode);
2083 extern void ext4_set_aops(struct inode *inode);
2084 extern int ext4_writepage_trans_blocks(struct inode *);
2085 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
2086 extern int ext4_discard_partial_page_buffers(handle_t *handle,
2087                 struct address_space *mapping, loff_t from,
2088                 loff_t length, int flags);
2089 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2090 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
2091 extern void ext4_da_update_reserve_space(struct inode *inode,
2092                                         int used, int quota_claim);
2093
2094 /* indirect.c */
2095 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
2096                                 struct ext4_map_blocks *map, int flags);
2097 extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
2098                                 const struct iovec *iov, loff_t offset,
2099                                 unsigned long nr_segs);
2100 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
2101 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks, int chunk);
2102 extern void ext4_ind_truncate(struct inode *inode);
2103 extern int ext4_ind_punch_hole(struct file *file, loff_t offset, loff_t length);
2104
2105 /* ioctl.c */
2106 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
2107 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
2108
2109 /* migrate.c */
2110 extern int ext4_ext_migrate(struct inode *);
2111
2112 /* namei.c */
2113 extern int ext4_dirent_csum_verify(struct inode *inode,
2114                                    struct ext4_dir_entry *dirent);
2115 extern int ext4_orphan_add(handle_t *, struct inode *);
2116 extern int ext4_orphan_del(handle_t *, struct inode *);
2117 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
2118                                 __u32 start_minor_hash, __u32 *next_hash);
2119 extern int search_dir(struct buffer_head *bh,
2120                       char *search_buf,
2121                       int buf_size,
2122                       struct inode *dir,
2123                       const struct qstr *d_name,
2124                       unsigned int offset,
2125                       struct ext4_dir_entry_2 **res_dir);
2126 extern int ext4_generic_delete_entry(handle_t *handle,
2127                                      struct inode *dir,
2128                                      struct ext4_dir_entry_2 *de_del,
2129                                      struct buffer_head *bh,
2130                                      void *entry_buf,
2131                                      int buf_size,
2132                                      int csum_size);
2133
2134 /* resize.c */
2135 extern int ext4_group_add(struct super_block *sb,
2136                                 struct ext4_new_group_data *input);
2137 extern int ext4_group_extend(struct super_block *sb,
2138                                 struct ext4_super_block *es,
2139                                 ext4_fsblk_t n_blocks_count);
2140 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
2141
2142 /* super.c */
2143 extern int ext4_calculate_overhead(struct super_block *sb);
2144 extern int ext4_superblock_csum_verify(struct super_block *sb,
2145                                        struct ext4_super_block *es);
2146 extern void ext4_superblock_csum_set(struct super_block *sb);
2147 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
2148 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
2149 extern void ext4_kvfree(void *ptr);
2150 extern int ext4_alloc_flex_bg_array(struct super_block *sb,
2151                                     ext4_group_t ngroup);
2152 extern __printf(4, 5)
2153 void __ext4_error(struct super_block *, const char *, unsigned int,
2154                   const char *, ...);
2155 #define ext4_error(sb, message...)      __ext4_error(sb, __func__,      \
2156                                                      __LINE__, ## message)
2157 extern __printf(5, 6)
2158 void ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
2159                       const char *, ...);
2160 extern __printf(5, 6)
2161 void ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
2162                      const char *, ...);
2163 extern void __ext4_std_error(struct super_block *, const char *,
2164                              unsigned int, int);
2165 extern __printf(4, 5)
2166 void __ext4_abort(struct super_block *, const char *, unsigned int,
2167                   const char *, ...);
2168 #define ext4_abort(sb, message...)      __ext4_abort(sb, __func__, \
2169                                                        __LINE__, ## message)
2170 extern __printf(4, 5)
2171 void __ext4_warning(struct super_block *, const char *, unsigned int,
2172                     const char *, ...);
2173 #define ext4_warning(sb, message...)    __ext4_warning(sb, __func__, \
2174                                                        __LINE__, ## message)
2175 extern __printf(3, 4)
2176 void ext4_msg(struct super_block *, const char *, const char *, ...);
2177 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
2178                            const char *, unsigned int, const char *);
2179 #define dump_mmp_msg(sb, mmp, msg)      __dump_mmp_msg(sb, mmp, __func__, \
2180                                                        __LINE__, msg)
2181 extern __printf(7, 8)
2182 void __ext4_grp_locked_error(const char *, unsigned int,
2183                              struct super_block *, ext4_group_t,
2184                              unsigned long, ext4_fsblk_t,
2185                              const char *, ...);
2186 #define ext4_grp_locked_error(sb, grp, message...) \
2187         __ext4_grp_locked_error(__func__, __LINE__, (sb), (grp), ## message)
2188 extern void ext4_update_dynamic_rev(struct super_block *sb);
2189 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
2190                                         __u32 compat);
2191 extern int ext4_update_rocompat_feature(handle_t *handle,
2192                                         struct super_block *sb, __u32 rocompat);
2193 extern int ext4_update_incompat_feature(handle_t *handle,
2194                                         struct super_block *sb, __u32 incompat);
2195 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
2196                                       struct ext4_group_desc *bg);
2197 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
2198                                       struct ext4_group_desc *bg);
2199 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
2200                                      struct ext4_group_desc *bg);
2201 extern __u32 ext4_free_group_clusters(struct super_block *sb,
2202                                       struct ext4_group_desc *bg);
2203 extern __u32 ext4_free_inodes_count(struct super_block *sb,
2204                                  struct ext4_group_desc *bg);
2205 extern __u32 ext4_used_dirs_count(struct super_block *sb,
2206                                 struct ext4_group_desc *bg);
2207 extern __u32 ext4_itable_unused_count(struct super_block *sb,
2208                                    struct ext4_group_desc *bg);
2209 extern void ext4_block_bitmap_set(struct super_block *sb,
2210                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2211 extern void ext4_inode_bitmap_set(struct super_block *sb,
2212                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2213 extern void ext4_inode_table_set(struct super_block *sb,
2214                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
2215 extern void ext4_free_group_clusters_set(struct super_block *sb,
2216                                          struct ext4_group_desc *bg,
2217                                          __u32 count);
2218 extern void ext4_free_inodes_set(struct super_block *sb,
2219                                 struct ext4_group_desc *bg, __u32 count);
2220 extern void ext4_used_dirs_set(struct super_block *sb,
2221                                 struct ext4_group_desc *bg, __u32 count);
2222 extern void ext4_itable_unused_set(struct super_block *sb,
2223                                    struct ext4_group_desc *bg, __u32 count);
2224 extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
2225                                        struct ext4_group_desc *gdp);
2226 extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
2227                                      struct ext4_group_desc *gdp);
2228 extern int ext4_register_li_request(struct super_block *sb,
2229                                     ext4_group_t first_not_zeroed);
2230
2231 static inline int ext4_has_group_desc_csum(struct super_block *sb)
2232 {
2233         return EXT4_HAS_RO_COMPAT_FEATURE(sb,
2234                                           EXT4_FEATURE_RO_COMPAT_GDT_CSUM |
2235                                           EXT4_FEATURE_RO_COMPAT_METADATA_CSUM);
2236 }
2237
2238 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2239 {
2240         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2241                 le32_to_cpu(es->s_blocks_count_lo);
2242 }
2243
2244 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2245 {
2246         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2247                 le32_to_cpu(es->s_r_blocks_count_lo);
2248 }
2249
2250 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2251 {
2252         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2253                 le32_to_cpu(es->s_free_blocks_count_lo);
2254 }
2255
2256 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2257                                          ext4_fsblk_t blk)
2258 {
2259         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2260         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2261 }
2262
2263 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2264                                               ext4_fsblk_t blk)
2265 {
2266         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2267         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2268 }
2269
2270 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2271                                            ext4_fsblk_t blk)
2272 {
2273         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2274         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2275 }
2276
2277 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2278 {
2279         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2280                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2281                         le32_to_cpu(raw_inode->i_size_lo);
2282         else
2283                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2284 }
2285
2286 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2287 {
2288         raw_inode->i_size_lo = cpu_to_le32(i_size);
2289         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2290 }
2291
2292 static inline
2293 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2294                                             ext4_group_t group)
2295 {
2296          struct ext4_group_info ***grp_info;
2297          long indexv, indexh;
2298          grp_info = EXT4_SB(sb)->s_group_info;
2299          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2300          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2301          return grp_info[indexv][indexh];
2302 }
2303
2304 /*
2305  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2306  * the locking protocol documented in the comments of ext4_group_add()
2307  * in resize.c
2308  */
2309 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2310 {
2311         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2312
2313         smp_rmb();
2314         return ngroups;
2315 }
2316
2317 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2318                                              ext4_group_t block_group)
2319 {
2320         return block_group >> sbi->s_log_groups_per_flex;
2321 }
2322
2323 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2324 {
2325         return 1 << sbi->s_log_groups_per_flex;
2326 }
2327
2328 #define ext4_std_error(sb, errno)                               \
2329 do {                                                            \
2330         if ((errno))                                            \
2331                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2332 } while (0)
2333
2334 #ifdef CONFIG_SMP
2335 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2336  * counters. So we need to make sure we have free clusters more
2337  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2338  */
2339 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2340 #else
2341 #define EXT4_FREECLUSTERS_WATERMARK 0
2342 #endif
2343
2344 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2345 {
2346         /*
2347          * XXX: replace with spinlock if seen contended -bzzz
2348          */
2349         down_write(&EXT4_I(inode)->i_data_sem);
2350         if (newsize > EXT4_I(inode)->i_disksize)
2351                 EXT4_I(inode)->i_disksize = newsize;
2352         up_write(&EXT4_I(inode)->i_data_sem);
2353         return ;
2354 }
2355
2356 struct ext4_group_info {
2357         unsigned long   bb_state;
2358         struct rb_root  bb_free_root;
2359         ext4_grpblk_t   bb_first_free;  /* first free block */
2360         ext4_grpblk_t   bb_free;        /* total free blocks */
2361         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2362         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2363         struct          list_head bb_prealloc_list;
2364 #ifdef DOUBLE_CHECK
2365         void            *bb_bitmap;
2366 #endif
2367         struct rw_semaphore alloc_sem;
2368         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2369                                          * regions, index is order.
2370                                          * bb_counters[3] = 5 means
2371                                          * 5 free 8-block regions. */
2372 };
2373
2374 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2375 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2376
2377 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2378         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2379
2380 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2381         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2382 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2383         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2384 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2385         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2386
2387 #define EXT4_MAX_CONTENTION             8
2388 #define EXT4_CONTENTION_THRESHOLD       2
2389
2390 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2391                                               ext4_group_t group)
2392 {
2393         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2394 }
2395
2396 /*
2397  * Returns true if the filesystem is busy enough that attempts to
2398  * access the block group locks has run into contention.
2399  */
2400 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2401 {
2402         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2403 }
2404
2405 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2406 {
2407         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2408         if (spin_trylock(lock))
2409                 /*
2410                  * We're able to grab the lock right away, so drop the
2411                  * lock contention counter.
2412                  */
2413                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2414         else {
2415                 /*
2416                  * The lock is busy, so bump the contention counter,
2417                  * and then wait on the spin lock.
2418                  */
2419                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2420                                   EXT4_MAX_CONTENTION);
2421                 spin_lock(lock);
2422         }
2423 }
2424
2425 static inline void ext4_unlock_group(struct super_block *sb,
2426                                         ext4_group_t group)
2427 {
2428         spin_unlock(ext4_group_lock_ptr(sb, group));
2429 }
2430
2431 /*
2432  * Block validity checking
2433  */
2434 #define ext4_check_indirect_blockref(inode, bh)                         \
2435         ext4_check_blockref(__func__, __LINE__, inode,                  \
2436                             (__le32 *)(bh)->b_data,                     \
2437                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2438
2439 #define ext4_ind_check_inode(inode)                                     \
2440         ext4_check_blockref(__func__, __LINE__, inode,                  \
2441                             EXT4_I(inode)->i_data,                      \
2442                             EXT4_NDIR_BLOCKS)
2443
2444 /*
2445  * Inodes and files operations
2446  */
2447
2448 /* dir.c */
2449 extern const struct file_operations ext4_dir_operations;
2450
2451 /* file.c */
2452 extern const struct inode_operations ext4_file_inode_operations;
2453 extern const struct file_operations ext4_file_operations;
2454 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2455 extern void ext4_unwritten_wait(struct inode *inode);
2456
2457 /* namei.c */
2458 extern const struct inode_operations ext4_dir_inode_operations;
2459 extern const struct inode_operations ext4_special_inode_operations;
2460 extern struct dentry *ext4_get_parent(struct dentry *child);
2461 extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
2462                                  struct ext4_dir_entry_2 *de,
2463                                  int blocksize, int csum_size,
2464                                  unsigned int parent_ino, int dotdot_real_len);
2465 extern void initialize_dirent_tail(struct ext4_dir_entry_tail *t,
2466                                    unsigned int blocksize);
2467 extern int ext4_handle_dirty_dirent_node(handle_t *handle,
2468                                          struct inode *inode,
2469                                          struct buffer_head *bh);
2470
2471 /* symlink.c */
2472 extern const struct inode_operations ext4_symlink_inode_operations;
2473 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2474
2475 /* block_validity */
2476 extern void ext4_release_system_zone(struct super_block *sb);
2477 extern int ext4_setup_system_zone(struct super_block *sb);
2478 extern int __init ext4_init_system_zone(void);
2479 extern void ext4_exit_system_zone(void);
2480 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2481                                  ext4_fsblk_t start_blk,
2482                                  unsigned int count);
2483 extern int ext4_check_blockref(const char *, unsigned int,
2484                                struct inode *, __le32 *, unsigned int);
2485
2486 /* extents.c */
2487 struct ext4_ext_path;
2488 struct ext4_extent;
2489
2490 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2491 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2492 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
2493                                        int chunk);
2494 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2495                                struct ext4_map_blocks *map, int flags);
2496 extern void ext4_ext_truncate(struct inode *);
2497 extern int ext4_ext_punch_hole(struct file *file, loff_t offset,
2498                                 loff_t length);
2499 extern void ext4_ext_init(struct super_block *);
2500 extern void ext4_ext_release(struct super_block *);
2501 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2502                           loff_t len);
2503 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
2504                           ssize_t len);
2505 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2506                            struct ext4_map_blocks *map, int flags);
2507 extern int ext4_ext_calc_metadata_amount(struct inode *inode,
2508                                          ext4_lblk_t lblocks);
2509 extern int ext4_extent_tree_init(handle_t *, struct inode *);
2510 extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
2511                                                    int num,
2512                                                    struct ext4_ext_path *path);
2513 extern int ext4_can_extents_be_merged(struct inode *inode,
2514                                       struct ext4_extent *ex1,
2515                                       struct ext4_extent *ex2);
2516 extern int ext4_ext_insert_extent(handle_t *, struct inode *,
2517                                   struct ext4_ext_path *,
2518                                   struct ext4_extent *, int);
2519 extern struct ext4_ext_path *ext4_ext_find_extent(struct inode *, ext4_lblk_t,
2520                                                   struct ext4_ext_path *);
2521 extern void ext4_ext_drop_refs(struct ext4_ext_path *);
2522 extern int ext4_ext_check_inode(struct inode *inode);
2523 extern int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk);
2524 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2525                         __u64 start, __u64 len);
2526
2527
2528 /* move_extent.c */
2529 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2530                              __u64 start_orig, __u64 start_donor,
2531                              __u64 len, __u64 *moved_len);
2532
2533 /* page-io.c */
2534 extern int __init ext4_init_pageio(void);
2535 extern void ext4_add_complete_io(ext4_io_end_t *io_end);
2536 extern void ext4_exit_pageio(void);
2537 extern void ext4_ioend_wait(struct inode *);
2538 extern void ext4_free_io_end(ext4_io_end_t *io);
2539 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2540 extern void ext4_end_io_work(struct work_struct *work);
2541 extern void ext4_io_submit(struct ext4_io_submit *io);
2542 extern int ext4_bio_write_page(struct ext4_io_submit *io,
2543                                struct page *page,
2544                                int len,
2545                                struct writeback_control *wbc);
2546
2547 /* mmp.c */
2548 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2549 extern void ext4_mmp_csum_set(struct super_block *sb, struct mmp_struct *mmp);
2550 extern int ext4_mmp_csum_verify(struct super_block *sb,
2551                                 struct mmp_struct *mmp);
2552
2553 /* BH_Uninit flag: blocks are allocated but uninitialized on disk */
2554 enum ext4_state_bits {
2555         BH_Uninit       /* blocks are allocated but uninitialized on disk */
2556           = BH_JBDPrivateStart,
2557         BH_AllocFromCluster,    /* allocated blocks were part of already
2558                                  * allocated cluster. Note that this flag will
2559                                  * never, ever appear in a buffer_head's state
2560                                  * flag. See EXT4_MAP_FROM_CLUSTER to see where
2561                                  * this is used. */
2562 };
2563
2564 BUFFER_FNS(Uninit, uninit)
2565 TAS_BUFFER_FNS(Uninit, uninit)
2566
2567 /*
2568  * Add new method to test whether block and inode bitmaps are properly
2569  * initialized. With uninit_bg reading the block from disk is not enough
2570  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2571  */
2572 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2573
2574 static inline int bitmap_uptodate(struct buffer_head *bh)
2575 {
2576         return (buffer_uptodate(bh) &&
2577                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2578 }
2579 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2580 {
2581         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2582 }
2583
2584 /*
2585  * Disable DIO read nolock optimization, so new dioreaders will be forced
2586  * to grab i_mutex
2587  */
2588 static inline void ext4_inode_block_unlocked_dio(struct inode *inode)
2589 {
2590         ext4_set_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2591         smp_mb();
2592 }
2593 static inline void ext4_inode_resume_unlocked_dio(struct inode *inode)
2594 {
2595         smp_mb();
2596         ext4_clear_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2597 }
2598
2599 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2600
2601 /* For ioend & aio unwritten conversion wait queues */
2602 #define EXT4_WQ_HASH_SZ         37
2603 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
2604                                             EXT4_WQ_HASH_SZ])
2605 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
2606                                              EXT4_WQ_HASH_SZ])
2607 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2608 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2609
2610 #define EXT4_RESIZING   0
2611 extern int ext4_resize_begin(struct super_block *sb);
2612 extern void ext4_resize_end(struct super_block *sb);
2613
2614 #endif  /* __KERNEL__ */
2615
2616 #endif  /* _EXT4_H */