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 #ifdef __KERNEL__
  33 #include <linux/compat.h>
  34 #endif
  35
  36 /*
  37  * The fourth extended filesystem constants/structures
  38  */
  39
  40 /*
  41  * Define EXT4FS_DEBUG to produce debug messages
  42  */
  43 #undef EXT4FS_DEBUG
  44
  45 /*
  46  * Debug code
  47  */
  48 #ifdef EXT4FS_DEBUG
  49 #define ext4_debug(f, a...)                                             \
  50         do {                                                            \
  51                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
  52                         __FILE__, __LINE__, __func__);                  \
  53                 printk(KERN_DEBUG f, ## a);                             \
  54         } while (0)
  55 #else
  56 #define ext4_debug(f, a...)     do {} while (0)
  57 #endif
  58
  59 #define EXT4_ERROR_INODE(inode, fmt, a...) \
  60         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
  61
  62 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...)                 \
  63         ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
  64
  65 #define EXT4_ERROR_FILE(file, block, fmt, a...)                         \
  66         ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
  67
  68 /* data type for block offset of block group */
  69 typedef int ext4_grpblk_t;
  70
  71 /* data type for filesystem-wide blocks number */
  72 typedef unsigned long long ext4_fsblk_t;
  73
  74 /* data type for file logical block number */
  75 typedef __u32 ext4_lblk_t;
  76
  77 /* data type for block group number */
  78 typedef unsigned int ext4_group_t;
  79
  80 /*
  81  * Flags used in mballoc's allocation_context flags field.
  82  *
  83  * Also used to show what's going on for debugging purposes when the
  84  * flag field is exported via the traceport interface
  85  */
  86
  87 /* prefer goal again. length */
  88 #define EXT4_MB_HINT_MERGE              0x0001
  89 /* blocks already reserved */
  90 #define EXT4_MB_HINT_RESERVED           0x0002
  91 /* metadata is being allocated */
  92 #define EXT4_MB_HINT_METADATA           0x0004
  93 /* first blocks in the file */
  94 #define EXT4_MB_HINT_FIRST              0x0008
  95 /* search for the best chunk */
  96 #define EXT4_MB_HINT_BEST               0x0010
  97 /* data is being allocated */
  98 #define EXT4_MB_HINT_DATA               0x0020
  99 /* don't preallocate (for tails) */
 100 #define EXT4_MB_HINT_NOPREALLOC         0x0040
 101 /* allocate for locality group */
 102 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
 103 /* allocate goal blocks or none */
 104 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
 105 /* goal is meaningful */
 106 #define EXT4_MB_HINT_TRY_GOAL           0x0200
 107 /* blocks already pre-reserved by delayed allocation */
 108 #define EXT4_MB_DELALLOC_RESERVED       0x0400
 109 /* We are doing stream allocation */
 110 #define EXT4_MB_STREAM_ALLOC            0x0800
 111 /* Use reserved root blocks if needed */
 112 #define EXT4_MB_USE_ROOT_BLOCKS         0x1000
 113
 114 struct ext4_allocation_request {
 115         /* target inode for block we're allocating */
 116         struct inode *inode;
 117         /* how many blocks we want to allocate */
 118         unsigned int len;
 119         /* logical block in target inode */
 120         ext4_lblk_t logical;
 121         /* the closest logical allocated block to the left */
 122         ext4_lblk_t lleft;
 123         /* the closest logical allocated block to the right */
 124         ext4_lblk_t lright;
 125         /* phys. target (a hint) */
 126         ext4_fsblk_t goal;
 127         /* phys. block for the closest logical allocated block to the left */
 128         ext4_fsblk_t pleft;
 129         /* phys. block for the closest logical allocated block to the right */
 130         ext4_fsblk_t pright;
 131         /* flags. see above EXT4_MB_HINT_* */
 132         unsigned int flags;
 133 };
 134
 135 /*
 136  * Logical to physical block mapping, used by ext4_map_blocks()
 137  *
 138  * This structure is used to pass requests into ext4_map_blocks() as
 139  * well as to store the information returned by ext4_map_blocks().  It
 140  * takes less room on the stack than a struct buffer_head.
 141  */
 142 #define EXT4_MAP_NEW            (1 << BH_New)
 143 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
 144 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
 145 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
 146 #define EXT4_MAP_UNINIT         (1 << BH_Uninit)
 147 /* Sometimes (in the bigalloc case, from ext4_da_get_block_prep) the caller of
 148  * ext4_map_blocks wants to know whether or not the underlying cluster has
 149  * already been accounted for. EXT4_MAP_FROM_CLUSTER conveys to the caller that
 150  * the requested mapping was from previously mapped (or delayed allocated)
 151  * cluster. We use BH_AllocFromCluster only for this flag. BH_AllocFromCluster
 152  * should never appear on buffer_head's state flags.
 153  */
 154 #define EXT4_MAP_FROM_CLUSTER   (1 << BH_AllocFromCluster)
 155 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
 156                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
 157                                  EXT4_MAP_UNINIT | EXT4_MAP_FROM_CLUSTER)
 158
 159 struct ext4_map_blocks {
 160         ext4_fsblk_t m_pblk;
 161         ext4_lblk_t m_lblk;
 162         unsigned int m_len;
 163         unsigned int m_flags;
 164 };
 165
 166 /*
 167  * For delayed allocation tracking
 168  */
 169 struct mpage_da_data {
 170         struct inode *inode;
 171         sector_t b_blocknr;             /* start block number of extent */
 172         size_t b_size;                  /* size of extent */
 173         unsigned long b_state;          /* state of the extent */
 174         unsigned long first_page, next_page;    /* extent of pages */
 175         struct writeback_control *wbc;
 176         int io_done;
 177         int pages_written;
 178         int retval;
 179 };
 180
 181 /*
 182  * Flags for ext4_io_end->flags
 183  */
 184 #define EXT4_IO_END_UNWRITTEN   0x0001
 185 #define EXT4_IO_END_ERROR       0x0002
 186 #define EXT4_IO_END_QUEUED      0x0004
 187
 188 struct ext4_io_page {
 189         struct page     *p_page;
 190         atomic_t        p_count;
 191 };
 192
 193 #define MAX_IO_PAGES 128
 194
 195 typedef struct ext4_io_end {
 196         struct list_head        list;           /* per-file finished IO list */
 197         struct inode            *inode;         /* file being written to */
 198         unsigned int            flag;           /* unwritten or not */
 199         struct page             *page;          /* page struct for buffer write */
 200         loff_t                  offset;         /* offset in the file */
 201         ssize_t                 size;           /* size of the extent */
 202         struct work_struct      work;           /* data work queue */
 203         struct kiocb            *iocb;          /* iocb struct for AIO */
 204         int                     result;         /* error value for AIO */
 205         int                     num_io_pages;
 206         struct ext4_io_page     *pages[MAX_IO_PAGES];
 207 } ext4_io_end_t;
 208
 209 struct ext4_io_submit {
 210         int                     io_op;
 211         struct bio              *io_bio;
 212         ext4_io_end_t           *io_end;
 213         struct ext4_io_page     *io_page;
 214         sector_t                io_next_block;
 215 };
 216
 217 /*
 218  * Special inodes numbers
 219  */
 220 #define EXT4_BAD_INO             1      /* Bad blocks inode */
 221 #define EXT4_ROOT_INO            2      /* Root inode */
 222 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
 223 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
 224 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
 225 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
 226 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
 227 #define EXT4_JOURNAL_INO         8      /* Journal inode */
 228
 229 /* First non-reserved inode for old ext4 filesystems */
 230 #define EXT4_GOOD_OLD_FIRST_INO 11
 231
 232 /*
 233  * Maximal count of links to a file
 234  */
 235 #define EXT4_LINK_MAX           65000
 236
 237 /*
 238  * Macro-instructions used to manage several block sizes
 239  */
 240 #define EXT4_MIN_BLOCK_SIZE             1024
 241 #define EXT4_MAX_BLOCK_SIZE             65536
 242 #define EXT4_MIN_BLOCK_LOG_SIZE         10
 243 #define EXT4_MAX_BLOCK_LOG_SIZE         16
 244 #ifdef __KERNEL__
 245 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
 246 #else
 247 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
 248 #endif
 249 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
 250 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
 251                                          EXT4_SB(s)->s_cluster_bits)
 252 #ifdef __KERNEL__
 253 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
 254 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
 255 #else
 256 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
 257 #endif
 258 #ifdef __KERNEL__
 259 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
 260 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
 261 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
 262 #else
 263 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 264                                  EXT4_GOOD_OLD_INODE_SIZE : \
 265                                  (s)->s_inode_size)
 266 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 267                                  EXT4_GOOD_OLD_FIRST_INO : \
 268                                  (s)->s_first_ino)
 269 #endif
 270 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
 271
 272 /* Translate a block number to a cluster number */
 273 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
 274 /* Translate a cluster number to a block number */
 275 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
 276 /* Translate # of blks to # of clusters */
 277 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
 278                                  (sbi)->s_cluster_bits)
 279
 280 /*
 281  * Structure of a blocks group descriptor
 282  */
 283 struct ext4_group_desc
 284 {
 285         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
 286         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
 287         __le32  bg_inode_table_lo;      /* Inodes table block */
 288         __le16  bg_free_blocks_count_lo;/* Free blocks count */
 289         __le16  bg_free_inodes_count_lo;/* Free inodes count */
 290         __le16  bg_used_dirs_count_lo;  /* Directories count */
 291         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
 292         __u32   bg_reserved[2];         /* Likely block/inode bitmap checksum */
 293         __le16  bg_itable_unused_lo;    /* Unused inodes count */
 294         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
 295         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
 296         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
 297         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
 298         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
 299         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
 300         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
 301         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
 302         __u32   bg_reserved2[3];
 303 };
 304
 305 /*
 306  * Structure of a flex block group info
 307  */
 308
 309 struct flex_groups {
 310         atomic_t free_inodes;
 311         atomic_t free_clusters;
 312         atomic_t used_dirs;
 313 };
 314
 315 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
 316 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
 317 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
 318
 319 /*
 320  * Macro-instructions used to manage group descriptors
 321  */
 322 #define EXT4_MIN_DESC_SIZE              32
 323 #define EXT4_MIN_DESC_SIZE_64BIT        64
 324 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
 325 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
 326 #ifdef __KERNEL__
 327 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
 328 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
 329 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
 330 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
 331 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
 332 #else
 333 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
 334 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
 335 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
 336 #endif
 337
 338 /*
 339  * Constants relative to the data blocks
 340  */
 341 #define EXT4_NDIR_BLOCKS                12
 342 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
 343 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
 344 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
 345 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
 346
 347 /*
 348  * Inode flags
 349  */
 350 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
 351 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
 352 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
 353 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
 354 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
 355 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
 356 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
 357 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
 358 /* Reserved for compression usage... */
 359 #define EXT4_DIRTY_FL                   0x00000100
 360 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
 361 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
 362 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
 363 /* End compression flags --- maybe not all used */
 364 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
 365 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
 366 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
 367 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
 368 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
 369 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
 370 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
 371 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
 372 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
 373 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
 374 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
 375
 376 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
 377 #define EXT4_FL_USER_MODIFIABLE         0x004B80FF /* User modifiable flags */
 378
 379 /* Flags that should be inherited by new inodes from their parent. */
 380 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
 381                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
 382                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
 383                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
 384
 385 /* Flags that are appropriate for regular files (all but dir-specific ones). */
 386 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
 387
 388 /* Flags that are appropriate for non-directories/regular files. */
 389 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
 390
 391 /* Mask out flags that are inappropriate for the given type of inode. */
 392 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
 393 {
 394         if (S_ISDIR(mode))
 395                 return flags;
 396         else if (S_ISREG(mode))
 397                 return flags & EXT4_REG_FLMASK;
 398         else
 399                 return flags & EXT4_OTHER_FLMASK;
 400 }
 401
 402 /*
 403  * Inode flags used for atomic set/get
 404  */
 405 enum {
 406         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
 407         EXT4_INODE_UNRM         = 1,    /* Undelete */
 408         EXT4_INODE_COMPR        = 2,    /* Compress file */
 409         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
 410         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
 411         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
 412         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
 413         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
 414 /* Reserved for compression usage... */
 415         EXT4_INODE_DIRTY        = 8,
 416         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
 417         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
 418         EXT4_INODE_ECOMPR       = 11,   /* Compression error */
 419 /* End compression flags --- maybe not all used */
 420         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
 421         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
 422         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
 423         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
 424         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
 425         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
 426         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
 427         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
 428         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
 429         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
 430         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
 431 };
 432
 433 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
 434 #define CHECK_FLAG_VALUE(FLAG) if (!TEST_FLAG_VALUE(FLAG)) { \
 435         printk(KERN_EMERG "EXT4 flag fail: " #FLAG ": %d %d\n", \
 436                 EXT4_##FLAG##_FL, EXT4_INODE_##FLAG); BUG_ON(1); }
 437
 438 /*
 439  * Since it's pretty easy to mix up bit numbers and hex values, and we
 440  * can't do a compile-time test for ENUM values, we use a run-time
 441  * test to make sure that EXT4_XXX_FL is consistent with respect to
 442  * EXT4_INODE_XXX.  If all is well the printk and BUG_ON will all drop
 443  * out so it won't cost any extra space in the compiled kernel image.
 444  * But it's important that these values are the same, since we are
 445  * using EXT4_INODE_XXX to test for the flag values, but EXT4_XX_FL
 446  * must be consistent with the values of FS_XXX_FL defined in
 447  * include/linux/fs.h and the on-disk values found in ext2, ext3, and
 448  * ext4 filesystems, and of course the values defined in e2fsprogs.
 449  *
 450  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
 451  */
 452 static inline void ext4_check_flag_values(void)
 453 {
 454         CHECK_FLAG_VALUE(SECRM);
 455         CHECK_FLAG_VALUE(UNRM);
 456         CHECK_FLAG_VALUE(COMPR);
 457         CHECK_FLAG_VALUE(SYNC);
 458         CHECK_FLAG_VALUE(IMMUTABLE);
 459         CHECK_FLAG_VALUE(APPEND);
 460         CHECK_FLAG_VALUE(NODUMP);
 461         CHECK_FLAG_VALUE(NOATIME);
 462         CHECK_FLAG_VALUE(DIRTY);
 463         CHECK_FLAG_VALUE(COMPRBLK);
 464         CHECK_FLAG_VALUE(NOCOMPR);
 465         CHECK_FLAG_VALUE(ECOMPR);
 466         CHECK_FLAG_VALUE(INDEX);
 467         CHECK_FLAG_VALUE(IMAGIC);
 468         CHECK_FLAG_VALUE(JOURNAL_DATA);
 469         CHECK_FLAG_VALUE(NOTAIL);
 470         CHECK_FLAG_VALUE(DIRSYNC);
 471         CHECK_FLAG_VALUE(TOPDIR);
 472         CHECK_FLAG_VALUE(HUGE_FILE);
 473         CHECK_FLAG_VALUE(EXTENTS);
 474         CHECK_FLAG_VALUE(EA_INODE);
 475         CHECK_FLAG_VALUE(EOFBLOCKS);
 476         CHECK_FLAG_VALUE(RESERVED);
 477 }
 478
 479 /* Used to pass group descriptor data when online resize is done */
 480 struct ext4_new_group_input {
 481         __u32 group;            /* Group number for this data */
 482         __u64 block_bitmap;     /* Absolute block number of block bitmap */
 483         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
 484         __u64 inode_table;      /* Absolute block number of inode table start */
 485         __u32 blocks_count;     /* Total number of blocks in this group */
 486         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
 487         __u16 unused;
 488 };
 489
 490 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 491 struct compat_ext4_new_group_input {
 492         u32 group;
 493         compat_u64 block_bitmap;
 494         compat_u64 inode_bitmap;
 495         compat_u64 inode_table;
 496         u32 blocks_count;
 497         u16 reserved_blocks;
 498         u16 unused;
 499 };
 500 #endif
 501
 502 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
 503 struct ext4_new_group_data {
 504         __u32 group;
 505         __u64 block_bitmap;
 506         __u64 inode_bitmap;
 507         __u64 inode_table;
 508         __u32 blocks_count;
 509         __u16 reserved_blocks;
 510         __u16 unused;
 511         __u32 free_blocks_count;
 512 };
 513
 514 /*
 515  * Flags used by ext4_map_blocks()
 516  */
 517         /* Allocate any needed blocks and/or convert an unitialized
 518            extent to be an initialized ext4 */
 519 #define EXT4_GET_BLOCKS_CREATE                  0x0001
 520         /* Request the creation of an unitialized extent */
 521 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
 522 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
 523                                                  EXT4_GET_BLOCKS_CREATE)
 524         /* Caller is from the delayed allocation writeout path,
 525            so set the magic i_delalloc_reserve_flag after taking the
 526            inode allocation semaphore for */
 527 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
 528         /* caller is from the direct IO path, request to creation of an
 529         unitialized extents if not allocated, split the uninitialized
 530         extent if blocks has been preallocated already*/
 531 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
 532 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
 533 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
 534                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
 535         /* Convert extent to initialized after IO complete */
 536 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
 537                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
 538         /* Punch out blocks of an extent */
 539 #define EXT4_GET_BLOCKS_PUNCH_OUT_EXT           0x0020
 540         /* Don't normalize allocation size (used for fallocate) */
 541 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
 542
 543 /*
 544  * Flags used by ext4_free_blocks
 545  */
 546 #define EXT4_FREE_BLOCKS_METADATA       0x0001
 547 #define EXT4_FREE_BLOCKS_FORGET         0x0002
 548 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
 549 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
 550 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
 551 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
 552
 553 /*
 554  * Flags used by ext4_discard_partial_page_buffers
 555  */
 556 #define EXT4_DISCARD_PARTIAL_PG_ZERO_UNMAPPED   0x0001
 557
 558 /*
 559  * ioctl commands
 560  */
 561 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
 562 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
 563 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
 564 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
 565 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
 566 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
 567 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
 568 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
 569 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
 570 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
 571 #define EXT4_IOC_MIGRATE                _IO('f', 9)
 572  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
 573  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
 574 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
 575 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
 576
 577 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 578 /*
 579  * ioctl commands in 32 bit emulation
 580  */
 581 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
 582 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
 583 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
 584 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
 585 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
 586 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
 587 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
 588 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
 589 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
 590 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
 591 #endif
 592
 593 /* Max physical block we can address w/o extents */
 594 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
 595
 596 /*
 597  * Structure of an inode on the disk
 598  */
 599 struct ext4_inode {
 600         __le16  i_mode;         /* File mode */
 601         __le16  i_uid;          /* Low 16 bits of Owner Uid */
 602         __le32  i_size_lo;      /* Size in bytes */
 603         __le32  i_atime;        /* Access time */
 604         __le32  i_ctime;        /* Inode Change time */
 605         __le32  i_mtime;        /* Modification time */
 606         __le32  i_dtime;        /* Deletion Time */
 607         __le16  i_gid;          /* Low 16 bits of Group Id */
 608         __le16  i_links_count;  /* Links count */
 609         __le32  i_blocks_lo;    /* Blocks count */
 610         __le32  i_flags;        /* File flags */
 611         union {
 612                 struct {
 613                         __le32  l_i_version;
 614                 } linux1;
 615                 struct {
 616                         __u32  h_i_translator;
 617                 } hurd1;
 618                 struct {
 619                         __u32  m_i_reserved1;
 620                 } masix1;
 621         } osd1;                         /* OS dependent 1 */
 622         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
 623         __le32  i_generation;   /* File version (for NFS) */
 624         __le32  i_file_acl_lo;  /* File ACL */
 625         __le32  i_size_high;
 626         __le32  i_obso_faddr;   /* Obsoleted fragment address */
 627         union {
 628                 struct {
 629                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
 630                         __le16  l_i_file_acl_high;
 631                         __le16  l_i_uid_high;   /* these 2 fields */
 632                         __le16  l_i_gid_high;   /* were reserved2[0] */
 633                         __u32   l_i_reserved2;
 634                 } linux2;
 635                 struct {
 636                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 637                         __u16   h_i_mode_high;
 638                         __u16   h_i_uid_high;
 639                         __u16   h_i_gid_high;
 640                         __u32   h_i_author;
 641                 } hurd2;
 642                 struct {
 643                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 644                         __le16  m_i_file_acl_high;
 645                         __u32   m_i_reserved2[2];
 646                 } masix2;
 647         } osd2;                         /* OS dependent 2 */
 648         __le16  i_extra_isize;
 649         __le16  i_pad1;
 650         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
 651         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
 652         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
 653         __le32  i_crtime;       /* File Creation time */
 654         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
 655         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
 656 };
 657
 658 struct move_extent {
 659         __u32 reserved;         /* should be zero */
 660         __u32 donor_fd;         /* donor file descriptor */
 661         __u64 orig_start;       /* logical start offset in block for orig */
 662         __u64 donor_start;      /* logical start offset in block for donor */
 663         __u64 len;              /* block length to be moved */
 664         __u64 moved_len;        /* moved block length */
 665 };
 666
 667 #define EXT4_EPOCH_BITS 2
 668 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
 669 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
 670
 671 /*
 672  * Extended fields will fit into an inode if the filesystem was formatted
 673  * with large inodes (-I 256 or larger) and there are not currently any EAs
 674  * consuming all of the available space. For new inodes we always reserve
 675  * enough space for the kernel's known extended fields, but for inodes
 676  * created with an old kernel this might not have been the case. None of
 677  * the extended inode fields is critical for correct filesystem operation.
 678  * This macro checks if a certain field fits in the inode. Note that
 679  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
 680  */
 681 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
 682         ((offsetof(typeof(*ext4_inode), field) +        \
 683           sizeof((ext4_inode)->field))                  \
 684         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
 685             (einode)->i_extra_isize))                   \
 686
 687 static inline __le32 ext4_encode_extra_time(struct timespec *time)
 688 {
 689        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
 690                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
 691                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
 692 }
 693
 694 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
 695 {
 696        if (sizeof(time->tv_sec) > 4)
 697                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
 698                                << 32;
 699        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
 700 }
 701
 702 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
 703 do {                                                                           \
 704         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
 705         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 706                 (raw_inode)->xtime ## _extra =                                 \
 707                                 ext4_encode_extra_time(&(inode)->xtime);       \
 708 } while (0)
 709
 710 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
 711 do {                                                                           \
 712         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 713                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
 714         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 715                 (raw_inode)->xtime ## _extra =                                 \
 716                                 ext4_encode_extra_time(&(einode)->xtime);      \
 717 } while (0)
 718
 719 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
 720 do {                                                                           \
 721         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
 722         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 723                 ext4_decode_extra_time(&(inode)->xtime,                        \
 724                                        raw_inode->xtime ## _extra);            \
 725         else                                                                   \
 726                 (inode)->xtime.tv_nsec = 0;                                    \
 727 } while (0)
 728
 729 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
 730 do {                                                                           \
 731         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 732                 (einode)->xtime.tv_sec =                                       \
 733                         (signed)le32_to_cpu((raw_inode)->xtime);               \
 734         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 735                 ext4_decode_extra_time(&(einode)->xtime,                       \
 736                                        raw_inode->xtime ## _extra);            \
 737         else                                                                   \
 738                 (einode)->xtime.tv_nsec = 0;                                   \
 739 } while (0)
 740
 741 #define i_disk_version osd1.linux1.l_i_version
 742
 743 #if defined(__KERNEL__) || defined(__linux__)
 744 #define i_reserved1     osd1.linux1.l_i_reserved1
 745 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
 746 #define i_blocks_high   osd2.linux2.l_i_blocks_high
 747 #define i_uid_low       i_uid
 748 #define i_gid_low       i_gid
 749 #define i_uid_high      osd2.linux2.l_i_uid_high
 750 #define i_gid_high      osd2.linux2.l_i_gid_high
 751 #define i_reserved2     osd2.linux2.l_i_reserved2
 752
 753 #elif defined(__GNU__)
 754
 755 #define i_translator    osd1.hurd1.h_i_translator
 756 #define i_uid_high      osd2.hurd2.h_i_uid_high
 757 #define i_gid_high      osd2.hurd2.h_i_gid_high
 758 #define i_author        osd2.hurd2.h_i_author
 759
 760 #elif defined(__masix__)
 761
 762 #define i_reserved1     osd1.masix1.m_i_reserved1
 763 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
 764 #define i_reserved2     osd2.masix2.m_i_reserved2
 765
 766 #endif /* defined(__KERNEL__) || defined(__linux__) */
 767
 768 /*
 769  * storage for cached extent
 770  * If ec_len == 0, then the cache is invalid.
 771  * If ec_start == 0, then the cache represents a gap (null mapping)
 772  */
 773 struct ext4_ext_cache {
 774         ext4_fsblk_t    ec_start;
 775         ext4_lblk_t     ec_block;
 776         __u32           ec_len; /* must be 32bit to return holes */
 777 };
 778
 779 /*
 780  * fourth extended file system inode data in memory
 781  */
 782 struct ext4_inode_info {
 783         __le32  i_data[15];     /* unconverted */
 784         __u32   i_dtime;
 785         ext4_fsblk_t    i_file_acl;
 786
 787         /*
 788          * i_block_group is the number of the block group which contains
 789          * this file's inode.  Constant across the lifetime of the inode,
 790          * it is ued for making block allocation decisions - we try to
 791          * place a file's data blocks near its inode block, and new inodes
 792          * near to their parent directory's inode.
 793          */
 794         ext4_group_t    i_block_group;
 795         ext4_lblk_t     i_dir_start_lookup;
 796 #if (BITS_PER_LONG < 64)
 797         unsigned long   i_state_flags;          /* Dynamic state flags */
 798 #endif
 799         unsigned long   i_flags;
 800
 801 #ifdef CONFIG_EXT4_FS_XATTR
 802         /*
 803          * Extended attributes can be read independently of the main file
 804          * data. Taking i_mutex even when reading would cause contention
 805          * between readers of EAs and writers of regular file data, so
 806          * instead we synchronize on xattr_sem when reading or changing
 807          * EAs.
 808          */
 809         struct rw_semaphore xattr_sem;
 810 #endif
 811
 812         struct list_head i_orphan;      /* unlinked but open inodes */
 813
 814         /*
 815          * i_disksize keeps track of what the inode size is ON DISK, not
 816          * in memory.  During truncate, i_size is set to the new size by
 817          * the VFS prior to calling ext4_truncate(), but the filesystem won't
 818          * set i_disksize to 0 until the truncate is actually under way.
 819          *
 820          * The intent is that i_disksize always represents the blocks which
 821          * are used by this file.  This allows recovery to restart truncate
 822          * on orphans if we crash during truncate.  We actually write i_disksize
 823          * into the on-disk inode when writing inodes out, instead of i_size.
 824          *
 825          * The only time when i_disksize and i_size may be different is when
 826          * a truncate is in progress.  The only things which change i_disksize
 827          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
 828          */
 829         loff_t  i_disksize;
 830
 831         /*
 832          * i_data_sem is for serialising ext4_truncate() against
 833          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
 834          * data tree are chopped off during truncate. We can't do that in
 835          * ext4 because whenever we perform intermediate commits during
 836          * truncate, the inode and all the metadata blocks *must* be in a
 837          * consistent state which allows truncation of the orphans to restart
 838          * during recovery.  Hence we must fix the get_block-vs-truncate race
 839          * by other means, so we have i_data_sem.
 840          */
 841         struct rw_semaphore i_data_sem;
 842         struct inode vfs_inode;
 843         struct jbd2_inode *jinode;
 844
 845         struct ext4_ext_cache i_cached_extent;
 846         /*
 847          * File creation time. Its function is same as that of
 848          * struct timespec i_{a,c,m}time in the generic inode.
 849          */
 850         struct timespec i_crtime;
 851
 852         /* mballoc */
 853         struct list_head i_prealloc_list;
 854         spinlock_t i_prealloc_lock;
 855
 856         /* ialloc */
 857         ext4_group_t    i_last_alloc_group;
 858
 859         /* allocation reservation info for delalloc */
 860         /* In case of bigalloc, these refer to clusters rather than blocks */
 861         unsigned int i_reserved_data_blocks;
 862         unsigned int i_reserved_meta_blocks;
 863         unsigned int i_allocated_meta_blocks;
 864         ext4_lblk_t i_da_metadata_calc_last_lblock;
 865         int i_da_metadata_calc_len;
 866
 867         /* on-disk additional length */
 868         __u16 i_extra_isize;
 869
 870 #ifdef CONFIG_QUOTA
 871         /* quota space reservation, managed internally by quota code */
 872         qsize_t i_reserved_quota;
 873 #endif
 874
 875         /* completed IOs that might need unwritten extents handling */
 876         struct list_head i_completed_io_list;
 877         spinlock_t i_completed_io_lock;
 878         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
 879         /* current io_end structure for async DIO write*/
 880         ext4_io_end_t *cur_aio_dio;
 881         atomic_t i_aiodio_unwritten; /* Nr. of inflight conversions pending */
 882
 883         spinlock_t i_block_reservation_lock;
 884
 885         /*
 886          * Transactions that contain inode's metadata needed to complete
 887          * fsync and fdatasync, respectively.
 888          */
 889         tid_t i_sync_tid;
 890         tid_t i_datasync_tid;
 891 };
 892
 893 /*
 894  * File system states
 895  */
 896 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
 897 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
 898 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
 899
 900 /*
 901  * Misc. filesystem flags
 902  */
 903 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
 904 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
 905 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
 906
 907 /*
 908  * Mount flags
 909  */
 910 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
 911 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
 912 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
 913 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
 914 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
 915 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
 916 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
 917 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
 918 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
 919 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
 920 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
 921 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
 922 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
 923 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
 924 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
 925 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
 926 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
 927 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
 928 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
 929 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
 930 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
 931 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
 932 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
 933 #define EXT4_MOUNT_I_VERSION            0x2000000 /* i_version support */
 934 #define EXT4_MOUNT_MBLK_IO_SUBMIT       0x4000000 /* multi-block io submits */
 935 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
 936 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
 937 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
 938 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
 939 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
 940
 941 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
 942                                                       specified delalloc */
 943
 944 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
 945                                                 ~EXT4_MOUNT_##opt
 946 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
 947                                                 EXT4_MOUNT_##opt
 948 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
 949                                          EXT4_MOUNT_##opt)
 950
 951 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
 952                                                 ~EXT4_MOUNT2_##opt
 953 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
 954                                                 EXT4_MOUNT2_##opt
 955 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
 956                                          EXT4_MOUNT2_##opt)
 957
 958 #define ext4_set_bit                    __test_and_set_bit_le
 959 #define ext4_set_bit_atomic             ext2_set_bit_atomic
 960 #define ext4_clear_bit                  __test_and_clear_bit_le
 961 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
 962 #define ext4_test_bit                   test_bit_le
 963 #define ext4_find_first_zero_bit        find_first_zero_bit_le
 964 #define ext4_find_next_zero_bit         find_next_zero_bit_le
 965 #define ext4_find_next_bit              find_next_bit_le
 966
 967 extern void ext4_set_bits(void *bm, int cur, int len);
 968
 969 /*
 970  * Maximal mount counts between two filesystem checks
 971  */
 972 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
 973 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
 974
 975 /*
 976  * Behaviour when detecting errors
 977  */
 978 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
 979 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
 980 #define EXT4_ERRORS_PANIC               3       /* Panic */
 981 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
 982
 983 /*
 984  * Structure of the super block
 985  */
 986 struct ext4_super_block {
 987 /*00*/  __le32  s_inodes_count;         /* Inodes count */
 988         __le32  s_blocks_count_lo;      /* Blocks count */
 989         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
 990         __le32  s_free_blocks_count_lo; /* Free blocks count */
 991 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
 992         __le32  s_first_data_block;     /* First Data Block */
 993         __le32  s_log_block_size;       /* Block size */
 994         __le32  s_log_cluster_size;     /* Allocation cluster size */
 995 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
 996         __le32  s_clusters_per_group;   /* # Clusters per group */
 997         __le32  s_inodes_per_group;     /* # Inodes per group */
 998         __le32  s_mtime;                /* Mount time */
 999 /*30*/  __le32  s_wtime;                /* Write time */
1000         __le16  s_mnt_count;            /* Mount count */
1001         __le16  s_max_mnt_count;        /* Maximal mount count */
1002         __le16  s_magic;                /* Magic signature */
1003         __le16  s_state;                /* File system state */
1004         __le16  s_errors;               /* Behaviour when detecting errors */
1005         __le16  s_minor_rev_level;      /* minor revision level */
1006 /*40*/  __le32  s_lastcheck;            /* time of last check */
1007         __le32  s_checkinterval;        /* max. time between checks */
1008         __le32  s_creator_os;           /* OS */
1009         __le32  s_rev_level;            /* Revision level */
1010 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1011         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1012         /*
1013          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1014          *
1015          * Note: the difference between the compatible feature set and
1016          * the incompatible feature set is that if there is a bit set
1017          * in the incompatible feature set that the kernel doesn't
1018          * know about, it should refuse to mount the filesystem.
1019          *
1020          * e2fsck's requirements are more strict; if it doesn't know
1021          * about a feature in either the compatible or incompatible
1022          * feature set, it must abort and not try to meddle with
1023          * things it doesn't understand...
1024          */
1025         __le32  s_first_ino;            /* First non-reserved inode */
1026         __le16  s_inode_size;           /* size of inode structure */
1027         __le16  s_block_group_nr;       /* block group # of this superblock */
1028         __le32  s_feature_compat;       /* compatible feature set */
1029 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1030         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1031 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1032 /*78*/  char    s_volume_name[16];      /* volume name */
1033 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1034 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1035         /*
1036          * Performance hints.  Directory preallocation should only
1037          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1038          */
1039         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1040         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1041         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1042         /*
1043          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1044          */
1045 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1046 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1047         __le32  s_journal_dev;          /* device number of journal file */
1048         __le32  s_last_orphan;          /* start of list of inodes to delete */
1049         __le32  s_hash_seed[4];         /* HTREE hash seed */
1050         __u8    s_def_hash_version;     /* Default hash version to use */
1051         __u8    s_jnl_backup_type;
1052         __le16  s_desc_size;            /* size of group descriptor */
1053 /*100*/ __le32  s_default_mount_opts;
1054         __le32  s_first_meta_bg;        /* First metablock block group */
1055         __le32  s_mkfs_time;            /* When the filesystem was created */
1056         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1057         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1058 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1059         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1060         __le32  s_free_blocks_count_hi; /* Free blocks count */
1061         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1062         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1063         __le32  s_flags;                /* Miscellaneous flags */
1064         __le16  s_raid_stride;          /* RAID stride */
1065         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1066         __le64  s_mmp_block;            /* Block for multi-mount protection */
1067         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1068         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1069         __u8    s_reserved_char_pad;
1070         __le16  s_reserved_pad;
1071         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1072         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1073         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1074         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1075                                               snapshot's future use */
1076         __le32  s_snapshot_list;        /* inode number of the head of the
1077                                            on-disk snapshot list */
1078 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1079         __le32  s_error_count;          /* number of fs errors */
1080         __le32  s_first_error_time;     /* first time an error happened */
1081         __le32  s_first_error_ino;      /* inode involved in first error */
1082         __le64  s_first_error_block;    /* block involved of first error */
1083         __u8    s_first_error_func[32]; /* function where the error happened */
1084         __le32  s_first_error_line;     /* line number where error happened */
1085         __le32  s_last_error_time;      /* most recent time of an error */
1086         __le32  s_last_error_ino;       /* inode involved in last error */
1087         __le32  s_last_error_line;      /* line number where error happened */
1088         __le64  s_last_error_block;     /* block involved of last error */
1089         __u8    s_last_error_func[32];  /* function where the error happened */
1090 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1091         __u8    s_mount_opts[64];
1092         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1093         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1094         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1095         __le32  s_reserved[109];        /* Padding to the end of the block */
1096 };
1097
1098 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1099
1100 #ifdef __KERNEL__
1101
1102 /*
1103  * run-time mount flags
1104  */
1105 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
1106 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
1107
1108 /*
1109  * fourth extended-fs super-block data in memory
1110  */
1111 struct ext4_sb_info {
1112         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1113         unsigned long s_inodes_per_block;/* Number of inodes per block */
1114         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1115         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1116         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1117         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1118         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1119         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1120         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1121         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1122         unsigned long s_overhead_last;  /* Last calculated overhead */
1123         unsigned long s_blocks_last;    /* Last seen block count */
1124         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1125         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1126         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1127         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1128         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1129         struct buffer_head **s_group_desc;
1130         unsigned int s_mount_opt;
1131         unsigned int s_mount_opt2;
1132         unsigned int s_mount_flags;
1133         ext4_fsblk_t s_sb_block;
1134         uid_t s_resuid;
1135         gid_t s_resgid;
1136         unsigned short s_mount_state;
1137         unsigned short s_pad;
1138         int s_addr_per_block_bits;
1139         int s_desc_per_block_bits;
1140         int s_inode_size;
1141         int s_first_ino;
1142         unsigned int s_inode_readahead_blks;
1143         unsigned int s_inode_goal;
1144         spinlock_t s_next_gen_lock;
1145         u32 s_next_generation;
1146         u32 s_hash_seed[4];
1147         int s_def_hash_version;
1148         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1149         struct percpu_counter s_freeclusters_counter;
1150         struct percpu_counter s_freeinodes_counter;
1151         struct percpu_counter s_dirs_counter;
1152         struct percpu_counter s_dirtyclusters_counter;
1153         struct blockgroup_lock *s_blockgroup_lock;
1154         struct proc_dir_entry *s_proc;
1155         struct kobject s_kobj;
1156         struct completion s_kobj_unregister;
1157
1158         /* Journaling */
1159         struct journal_s *s_journal;
1160         struct list_head s_orphan;
1161         struct mutex s_orphan_lock;
1162         unsigned long s_resize_flags;           /* Flags indicating if there
1163                                                    is a resizer */
1164         unsigned long s_commit_interval;
1165         u32 s_max_batch_time;
1166         u32 s_min_batch_time;
1167         struct block_device *journal_bdev;
1168 #ifdef CONFIG_QUOTA
1169         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
1170         int s_jquota_fmt;                       /* Format of quota to use */
1171 #endif
1172         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1173         struct rb_root system_blks;
1174
1175 #ifdef EXTENTS_STATS
1176         /* ext4 extents stats */
1177         unsigned long s_ext_min;
1178         unsigned long s_ext_max;
1179         unsigned long s_depth_max;
1180         spinlock_t s_ext_stats_lock;
1181         unsigned long s_ext_blocks;
1182         unsigned long s_ext_extents;
1183 #endif
1184         /* ext4 extent cache stats */
1185         unsigned long extent_cache_hits;
1186         unsigned long extent_cache_misses;
1187
1188         /* for buddy allocator */
1189         struct ext4_group_info ***s_group_info;
1190         struct inode *s_buddy_cache;
1191         spinlock_t s_md_lock;
1192         unsigned short *s_mb_offsets;
1193         unsigned int *s_mb_maxs;
1194
1195         /* tunables */
1196         unsigned long s_stripe;
1197         unsigned int s_mb_stream_request;
1198         unsigned int s_mb_max_to_scan;
1199         unsigned int s_mb_min_to_scan;
1200         unsigned int s_mb_stats;
1201         unsigned int s_mb_order2_reqs;
1202         unsigned int s_mb_group_prealloc;
1203         unsigned int s_max_writeback_mb_bump;
1204         /* where last allocation was done - for stream allocation */
1205         unsigned long s_mb_last_group;
1206         unsigned long s_mb_last_start;
1207
1208         /* stats for buddy allocator */
1209         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1210         atomic_t s_bal_success; /* we found long enough chunks */
1211         atomic_t s_bal_allocated;       /* in blocks */
1212         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1213         atomic_t s_bal_goals;   /* goal hits */
1214         atomic_t s_bal_breaks;  /* too long searches */
1215         atomic_t s_bal_2orders; /* 2^order hits */
1216         spinlock_t s_bal_lock;
1217         unsigned long s_mb_buddies_generated;
1218         unsigned long long s_mb_generation_time;
1219         atomic_t s_mb_lost_chunks;
1220         atomic_t s_mb_preallocated;
1221         atomic_t s_mb_discarded;
1222         atomic_t s_lock_busy;
1223
1224         /* locality groups */
1225         struct ext4_locality_group __percpu *s_locality_groups;
1226
1227         /* for write statistics */
1228         unsigned long s_sectors_written_start;
1229         u64 s_kbytes_written;
1230
1231         unsigned int s_log_groups_per_flex;
1232         struct flex_groups *s_flex_groups;
1233
1234         /* workqueue for dio unwritten */
1235         struct workqueue_struct *dio_unwritten_wq;
1236
1237         /* timer for periodic error stats printing */
1238         struct timer_list s_err_report;
1239
1240         /* Lazy inode table initialization info */
1241         struct ext4_li_request *s_li_request;
1242         /* Wait multiplier for lazy initialization thread */
1243         unsigned int s_li_wait_mult;
1244
1245         /* Kernel thread for multiple mount protection */
1246         struct task_struct *s_mmp_tsk;
1247
1248         /* record the last minlen when FITRIM is called. */
1249         atomic_t s_last_trim_minblks;
1250 };
1251
1252 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1253 {
1254         return sb->s_fs_info;
1255 }
1256 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1257 {
1258         return container_of(inode, struct ext4_inode_info, vfs_inode);
1259 }
1260
1261 static inline struct timespec ext4_current_time(struct inode *inode)
1262 {
1263         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1264                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1265 }
1266
1267 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1268 {
1269         return ino == EXT4_ROOT_INO ||
1270                 ino == EXT4_JOURNAL_INO ||
1271                 ino == EXT4_RESIZE_INO ||
1272                 (ino >= EXT4_FIRST_INO(sb) &&
1273                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1274 }
1275
1276 /*
1277  * Inode dynamic state flags
1278  */
1279 enum {
1280         EXT4_STATE_JDATA,               /* journaled data exists */
1281         EXT4_STATE_NEW,                 /* inode is newly created */
1282         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1283         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1284         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1285         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1286         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1287         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1288         EXT4_STATE_DELALLOC_RESERVED,   /* blks already reserved for delalloc */
1289 };
1290
1291 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1292 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1293 {                                                                       \
1294         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1295 }                                                                       \
1296 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1297 {                                                                       \
1298         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1299 }                                                                       \
1300 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1301 {                                                                       \
1302         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1303 }
1304
1305 EXT4_INODE_BIT_FNS(flag, flags, 0)
1306 #if (BITS_PER_LONG < 64)
1307 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1308
1309 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1310 {
1311         (ei)->i_state_flags = 0;
1312 }
1313 #else
1314 EXT4_INODE_BIT_FNS(state, flags, 32)
1315
1316 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1317 {
1318         /* We depend on the fact that callers will set i_flags */
1319 }
1320 #endif
1321 #else
1322 /* Assume that user mode programs are passing in an ext4fs superblock, not
1323  * a kernel struct super_block.  This will allow us to call the feature-test
1324  * macros from user land. */
1325 #define EXT4_SB(sb)     (sb)
1326 #endif
1327
1328 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1329
1330 /*
1331  * Codes for operating systems
1332  */
1333 #define EXT4_OS_LINUX           0
1334 #define EXT4_OS_HURD            1
1335 #define EXT4_OS_MASIX           2
1336 #define EXT4_OS_FREEBSD         3
1337 #define EXT4_OS_LITES           4
1338
1339 /*
1340  * Revision levels
1341  */
1342 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1343 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1344
1345 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1346 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1347
1348 #define EXT4_GOOD_OLD_INODE_SIZE 128
1349
1350 /*
1351  * Feature set definitions
1352  */
1353
1354 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1355         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1356 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1357         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1358 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1359         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1360 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1361         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1362 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1363         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1364 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1365         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1366 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1367         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1368 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1369         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1370 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1371         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1372
1373 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1374 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1375 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1376 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1377 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1378 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1379
1380 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1381 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1382 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1383 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1384 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1385 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1386 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1387 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1388 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1389
1390 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1391 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1392 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1393 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1394 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1395 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1396 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1397 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1398 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1399 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1400 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1401
1402 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1403 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1404                                          EXT4_FEATURE_INCOMPAT_META_BG)
1405 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1406                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1407                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1408
1409 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1410 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1411                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1412                                          EXT4_FEATURE_INCOMPAT_META_BG)
1413 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1414                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1415                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1416
1417 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1418 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1419                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1420                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1421                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1422                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1423                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1424                                          EXT4_FEATURE_INCOMPAT_MMP)
1425 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1426                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1427                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1428                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1429                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1430                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1431                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1432                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC)
1433
1434 /*
1435  * Default values for user and/or group using reserved blocks
1436  */
1437 #define EXT4_DEF_RESUID         0
1438 #define EXT4_DEF_RESGID         0
1439
1440 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1441
1442 /*
1443  * Default mount options
1444  */
1445 #define EXT4_DEFM_DEBUG         0x0001
1446 #define EXT4_DEFM_BSDGROUPS     0x0002
1447 #define EXT4_DEFM_XATTR_USER    0x0004
1448 #define EXT4_DEFM_ACL           0x0008
1449 #define EXT4_DEFM_UID16         0x0010
1450 #define EXT4_DEFM_JMODE         0x0060
1451 #define EXT4_DEFM_JMODE_DATA    0x0020
1452 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1453 #define EXT4_DEFM_JMODE_WBACK   0x0060
1454 #define EXT4_DEFM_NOBARRIER     0x0100
1455 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1456 #define EXT4_DEFM_DISCARD       0x0400
1457 #define EXT4_DEFM_NODELALLOC    0x0800
1458
1459 /*
1460  * Default journal batch times
1461  */
1462 #define EXT4_DEF_MIN_BATCH_TIME 0
1463 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1464
1465 /*
1466  * Minimum number of groups in a flexgroup before we separate out
1467  * directories into the first block group of a flexgroup
1468  */
1469 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1470
1471 /*
1472  * Structure of a directory entry
1473  */
1474 #define EXT4_NAME_LEN 255
1475
1476 struct ext4_dir_entry {
1477         __le32  inode;                  /* Inode number */
1478         __le16  rec_len;                /* Directory entry length */
1479         __le16  name_len;               /* Name length */
1480         char    name[EXT4_NAME_LEN];    /* File name */
1481 };
1482
1483 /*
1484  * The new version of the directory entry.  Since EXT4 structures are
1485  * stored in intel byte order, and the name_len field could never be
1486  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1487  * file_type field.
1488  */
1489 struct ext4_dir_entry_2 {
1490         __le32  inode;                  /* Inode number */
1491         __le16  rec_len;                /* Directory entry length */
1492         __u8    name_len;               /* Name length */
1493         __u8    file_type;
1494         char    name[EXT4_NAME_LEN];    /* File name */
1495 };
1496
1497 /*
1498  * Ext4 directory file types.  Only the low 3 bits are used.  The
1499  * other bits are reserved for now.
1500  */
1501 #define EXT4_FT_UNKNOWN         0
1502 #define EXT4_FT_REG_FILE        1
1503 #define EXT4_FT_DIR             2
1504 #define EXT4_FT_CHRDEV          3
1505 #define EXT4_FT_BLKDEV          4
1506 #define EXT4_FT_FIFO            5
1507 #define EXT4_FT_SOCK            6
1508 #define EXT4_FT_SYMLINK         7
1509
1510 #define EXT4_FT_MAX             8
1511
1512 /*
1513  * EXT4_DIR_PAD defines the directory entries boundaries
1514  *
1515  * NOTE: It must be a multiple of 4
1516  */
1517 #define EXT4_DIR_PAD                    4
1518 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1519 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1520                                          ~EXT4_DIR_ROUND)
1521 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1522
1523 /*
1524  * If we ever get support for fs block sizes > page_size, we'll need
1525  * to remove the #if statements in the next two functions...
1526  */
1527 static inline unsigned int
1528 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1529 {
1530         unsigned len = le16_to_cpu(dlen);
1531
1532 #if (PAGE_CACHE_SIZE >= 65536)
1533         if (len == EXT4_MAX_REC_LEN || len == 0)
1534                 return blocksize;
1535         return (len & 65532) | ((len & 3) << 16);
1536 #else
1537         return len;
1538 #endif
1539 }
1540
1541 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1542 {
1543         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1544                 BUG();
1545 #if (PAGE_CACHE_SIZE >= 65536)
1546         if (len < 65536)
1547                 return cpu_to_le16(len);
1548         if (len == blocksize) {
1549                 if (blocksize == 65536)
1550                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1551                 else
1552                         return cpu_to_le16(0);
1553         }
1554         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1555 #else
1556         return cpu_to_le16(len);
1557 #endif
1558 }
1559
1560 /*
1561  * Hash Tree Directory indexing
1562  * (c) Daniel Phillips, 2001
1563  */
1564
1565 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1566                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1567                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1568 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1569 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1570
1571 /* Legal values for the dx_root hash_version field: */
1572
1573 #define DX_HASH_LEGACY          0
1574 #define DX_HASH_HALF_MD4        1
1575 #define DX_HASH_TEA             2
1576 #define DX_HASH_LEGACY_UNSIGNED 3
1577 #define DX_HASH_HALF_MD4_UNSIGNED       4
1578 #define DX_HASH_TEA_UNSIGNED            5
1579
1580 #ifdef __KERNEL__
1581
1582 /* hash info structure used by the directory hash */
1583 struct dx_hash_info
1584 {
1585         u32             hash;
1586         u32             minor_hash;
1587         int             hash_version;
1588         u32             *seed;
1589 };
1590
1591 #define EXT4_HTREE_EOF  0x7fffffff
1592
1593 /*
1594  * Control parameters used by ext4_htree_next_block
1595  */
1596 #define HASH_NB_ALWAYS          1
1597
1598
1599 /*
1600  * Describe an inode's exact location on disk and in memory
1601  */
1602 struct ext4_iloc
1603 {
1604         struct buffer_head *bh;
1605         unsigned long offset;
1606         ext4_group_t block_group;
1607 };
1608
1609 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1610 {
1611         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1612 }
1613
1614 /*
1615  * This structure is stuffed into the struct file's private_data field
1616  * for directories.  It is where we put information so that we can do
1617  * readdir operations in hash tree order.
1618  */
1619 struct dir_private_info {
1620         struct rb_root  root;
1621         struct rb_node  *curr_node;
1622         struct fname    *extra_fname;
1623         loff_t          last_pos;
1624         __u32           curr_hash;
1625         __u32           curr_minor_hash;
1626         __u32           next_hash;
1627 };
1628
1629 /* calculate the first block number of the group */
1630 static inline ext4_fsblk_t
1631 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1632 {
1633         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1634                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1635 }
1636
1637 /*
1638  * Special error return code only used by dx_probe() and its callers.
1639  */
1640 #define ERR_BAD_DX_DIR  -75000
1641
1642 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1643                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1644
1645 /*
1646  * Timeout and state flag for lazy initialization inode thread.
1647  */
1648 #define EXT4_DEF_LI_WAIT_MULT                   10
1649 #define EXT4_DEF_LI_MAX_START_DELAY             5
1650 #define EXT4_LAZYINIT_QUIT                      0x0001
1651 #define EXT4_LAZYINIT_RUNNING                   0x0002
1652
1653 /*
1654  * Lazy inode table initialization info
1655  */
1656 struct ext4_lazy_init {
1657         unsigned long           li_state;
1658         struct list_head        li_request_list;
1659         struct mutex            li_list_mtx;
1660 };
1661
1662 struct ext4_li_request {
1663         struct super_block      *lr_super;
1664         struct ext4_sb_info     *lr_sbi;
1665         ext4_group_t            lr_next_group;
1666         struct list_head        lr_request;
1667         unsigned long           lr_next_sched;
1668         unsigned long           lr_timeout;
1669 };
1670
1671 struct ext4_features {
1672         struct kobject f_kobj;
1673         struct completion f_kobj_unregister;
1674 };
1675
1676 /*
1677  * This structure will be used for multiple mount protection. It will be
1678  * written into the block number saved in the s_mmp_block field in the
1679  * superblock. Programs that check MMP should assume that if
1680  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1681  * to use the filesystem, regardless of how old the timestamp is.
1682  */
1683 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
1684 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1685 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
1686 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
1687
1688 struct mmp_struct {
1689         __le32  mmp_magic;              /* Magic number for MMP */
1690         __le32  mmp_seq;                /* Sequence no. updated periodically */
1691
1692         /*
1693          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1694          * purposes and do not affect the correctness of the algorithm
1695          */
1696         __le64  mmp_time;               /* Time last updated */
1697         char    mmp_nodename[64];       /* Node which last updated MMP block */
1698         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
1699
1700         /*
1701          * mmp_check_interval is used to verify if the MMP block has been
1702          * updated on the block device. The value is updated based on the
1703          * maximum time to write the MMP block during an update cycle.
1704          */
1705         __le16  mmp_check_interval;
1706
1707         __le16  mmp_pad1;
1708         __le32  mmp_pad2[227];
1709 };
1710
1711 /* arguments passed to the mmp thread */
1712 struct mmpd_data {
1713         struct buffer_head *bh; /* bh from initial read_mmp_block() */
1714         struct super_block *sb;  /* super block of the fs */
1715 };
1716
1717 /*
1718  * Check interval multiplier
1719  * The MMP block is written every update interval and initially checked every
1720  * update interval x the multiplier (the value is then adapted based on the
1721  * write latency). The reason is that writes can be delayed under load and we
1722  * don't want readers to incorrectly assume that the filesystem is no longer
1723  * in use.
1724  */
1725 #define EXT4_MMP_CHECK_MULT             2UL
1726
1727 /*
1728  * Minimum interval for MMP checking in seconds.
1729  */
1730 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
1731
1732 /*
1733  * Maximum interval for MMP checking in seconds.
1734  */
1735 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
1736
1737 /*
1738  * Function prototypes
1739  */
1740
1741 /*
1742  * Ok, these declarations are also in <linux/kernel.h> but none of the
1743  * ext4 source programs needs to include it so they are duplicated here.
1744  */
1745 # define NORET_TYPE     /**/
1746 # define ATTRIB_NORET   __attribute__((noreturn))
1747 # define NORET_AND      noreturn,
1748
1749 /* bitmap.c */
1750 extern unsigned int ext4_count_free(struct buffer_head *, unsigned);
1751
1752 /* balloc.c */
1753 extern unsigned int ext4_block_group(struct super_block *sb,
1754                         ext4_fsblk_t blocknr);
1755 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1756                         ext4_fsblk_t blocknr);
1757 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1758 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1759                         ext4_group_t group);
1760 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1761                                          ext4_fsblk_t goal,
1762                                          unsigned int flags,
1763                                          unsigned long *count,
1764                                          int *errp);
1765 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
1766                                     s64 nclusters, unsigned int flags);
1767 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
1768 extern void ext4_check_blocks_bitmap(struct super_block *);
1769 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1770                                                     ext4_group_t block_group,
1771                                                     struct buffer_head ** bh);
1772 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1773 struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1774                                       ext4_group_t block_group);
1775 extern void ext4_init_block_bitmap(struct super_block *sb,
1776                                    struct buffer_head *bh,
1777                                    ext4_group_t group,
1778                                    struct ext4_group_desc *desc);
1779 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
1780                                               ext4_group_t block_group,
1781                                               struct ext4_group_desc *gdp);
1782 extern unsigned ext4_num_base_meta_clusters(struct super_block *sb,
1783                                             ext4_group_t block_group);
1784 extern unsigned ext4_num_overhead_clusters(struct super_block *sb,
1785                                            ext4_group_t block_group,
1786                                            struct ext4_group_desc *gdp);
1787 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
1788
1789 /* dir.c */
1790 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1791                                   struct file *,
1792                                   struct ext4_dir_entry_2 *,
1793                                   struct buffer_head *, unsigned int);
1794 #define ext4_check_dir_entry(dir, filp, de, bh, offset)                 \
1795         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
1796                                         (de), (bh), (offset)))
1797 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1798                                     __u32 minor_hash,
1799                                     struct ext4_dir_entry_2 *dirent);
1800 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1801
1802 /* fsync.c */
1803 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
1804 extern int ext4_flush_completed_IO(struct inode *);
1805
1806 /* hash.c */
1807 extern int ext4fs_dirhash(const char *name, int len, struct
1808                           dx_hash_info *hinfo);
1809
1810 /* ialloc.c */
1811 extern struct inode *ext4_new_inode(handle_t *, struct inode *, int,
1812                                     const struct qstr *qstr, __u32 goal);
1813 extern void ext4_free_inode(handle_t *, struct inode *);
1814 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1815 extern unsigned long ext4_count_free_inodes(struct super_block *);
1816 extern unsigned long ext4_count_dirs(struct super_block *);
1817 extern void ext4_check_inodes_bitmap(struct super_block *);
1818 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1819 extern int ext4_init_inode_table(struct super_block *sb,
1820                                  ext4_group_t group, int barrier);
1821
1822 /* mballoc.c */
1823 extern long ext4_mb_stats;
1824 extern long ext4_mb_max_to_scan;
1825 extern int ext4_mb_init(struct super_block *, int);
1826 extern int ext4_mb_release(struct super_block *);
1827 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1828                                 struct ext4_allocation_request *, int *);
1829 extern int ext4_mb_reserve_blocks(struct super_block *, int);
1830 extern void ext4_discard_preallocations(struct inode *);
1831 extern int __init ext4_init_mballoc(void);
1832 extern void ext4_exit_mballoc(void);
1833 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1834                              struct buffer_head *bh, ext4_fsblk_t block,
1835                              unsigned long count, int flags);
1836 extern int ext4_mb_add_groupinfo(struct super_block *sb,
1837                 ext4_group_t i, struct ext4_group_desc *desc);
1838 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
1839                                 ext4_fsblk_t block, unsigned long count);
1840 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
1841
1842 /* inode.c */
1843 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1844                                                 ext4_lblk_t, int, int *);
1845 struct buffer_head *ext4_bread(handle_t *, struct inode *,
1846                                                 ext4_lblk_t, int, int *);
1847 int ext4_get_block(struct inode *inode, sector_t iblock,
1848                                 struct buffer_head *bh_result, int create);
1849
1850 extern struct inode *ext4_iget(struct super_block *, unsigned long);
1851 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
1852 extern int  ext4_setattr(struct dentry *, struct iattr *);
1853 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1854                                 struct kstat *stat);
1855 extern void ext4_evict_inode(struct inode *);
1856 extern void ext4_clear_inode(struct inode *);
1857 extern int  ext4_sync_inode(handle_t *, struct inode *);
1858 extern void ext4_dirty_inode(struct inode *, int);
1859 extern int ext4_change_inode_journal_flag(struct inode *, int);
1860 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1861 extern int ext4_can_truncate(struct inode *inode);
1862 extern void ext4_truncate(struct inode *);
1863 extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length);
1864 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
1865 extern void ext4_set_inode_flags(struct inode *);
1866 extern void ext4_get_inode_flags(struct ext4_inode_info *);
1867 extern int ext4_alloc_da_blocks(struct inode *inode);
1868 extern void ext4_set_aops(struct inode *inode);
1869 extern int ext4_writepage_trans_blocks(struct inode *);
1870 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1871 extern int ext4_block_truncate_page(handle_t *handle,
1872                 struct address_space *mapping, loff_t from);
1873 extern int ext4_block_zero_page_range(handle_t *handle,
1874                 struct address_space *mapping, loff_t from, loff_t length);
1875 extern int ext4_discard_partial_page_buffers(handle_t *handle,
1876                 struct address_space *mapping, loff_t from,
1877                 loff_t length, int flags);
1878 extern int ext4_discard_partial_page_buffers_no_lock(handle_t *handle,
1879                 struct inode *inode, struct page *page, loff_t from,
1880                 loff_t length, int flags);
1881 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1882 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
1883 extern void ext4_da_update_reserve_space(struct inode *inode,
1884                                         int used, int quota_claim);
1885
1886 /* indirect.c */
1887 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
1888                                 struct ext4_map_blocks *map, int flags);
1889 extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
1890                                 const struct iovec *iov, loff_t offset,
1891                                 unsigned long nr_segs);
1892 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
1893 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks, int chunk);
1894 extern void ext4_ind_truncate(struct inode *inode);
1895
1896 /* ioctl.c */
1897 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1898 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1899
1900 /* migrate.c */
1901 extern int ext4_ext_migrate(struct inode *);
1902
1903 /* namei.c */
1904 extern int ext4_orphan_add(handle_t *, struct inode *);
1905 extern int ext4_orphan_del(handle_t *, struct inode *);
1906 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1907                                 __u32 start_minor_hash, __u32 *next_hash);
1908
1909 /* resize.c */
1910 extern int ext4_group_add(struct super_block *sb,
1911                                 struct ext4_new_group_data *input);
1912 extern int ext4_group_extend(struct super_block *sb,
1913                                 struct ext4_super_block *es,
1914                                 ext4_fsblk_t n_blocks_count);
1915
1916 /* super.c */
1917 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
1918 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
1919 extern void ext4_kvfree(void *ptr);
1920 extern void __ext4_error(struct super_block *, const char *, unsigned int,
1921                          const char *, ...)
1922         __attribute__ ((format (printf, 4, 5)));
1923 #define ext4_error(sb, message...)      __ext4_error(sb, __func__,      \
1924                                                      __LINE__, ## message)
1925 extern void ext4_error_inode(struct inode *, const char *, unsigned int,
1926                              ext4_fsblk_t, const char *, ...)
1927         __attribute__ ((format (printf, 5, 6)));
1928 extern void ext4_error_file(struct file *, const char *, unsigned int,
1929                             ext4_fsblk_t, const char *, ...)
1930         __attribute__ ((format (printf, 5, 6)));
1931 extern void __ext4_std_error(struct super_block *, const char *,
1932                              unsigned int, int);
1933 extern void __ext4_abort(struct super_block *, const char *, unsigned int,
1934                        const char *, ...)
1935         __attribute__ ((format (printf, 4, 5)));
1936 #define ext4_abort(sb, message...)      __ext4_abort(sb, __func__, \
1937                                                        __LINE__, ## message)
1938 extern void __ext4_warning(struct super_block *, const char *, unsigned int,
1939                           const char *, ...)
1940         __attribute__ ((format (printf, 4, 5)));
1941 #define ext4_warning(sb, message...)    __ext4_warning(sb, __func__, \
1942                                                        __LINE__, ## message)
1943 extern void ext4_msg(struct super_block *, const char *, const char *, ...)
1944         __attribute__ ((format (printf, 3, 4)));
1945 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
1946                            const char *, unsigned int, const char *);
1947 #define dump_mmp_msg(sb, mmp, msg)      __dump_mmp_msg(sb, mmp, __func__, \
1948                                                        __LINE__, msg)
1949 extern void __ext4_grp_locked_error(const char *, unsigned int, \
1950                                     struct super_block *, ext4_group_t, \
1951                                     unsigned long, ext4_fsblk_t, \
1952                                     const char *, ...)
1953         __attribute__ ((format (printf, 7, 8)));
1954 #define ext4_grp_locked_error(sb, grp, message...) \
1955         __ext4_grp_locked_error(__func__, __LINE__, (sb), (grp), ## message)
1956 extern void ext4_update_dynamic_rev(struct super_block *sb);
1957 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1958                                         __u32 compat);
1959 extern int ext4_update_rocompat_feature(handle_t *handle,
1960                                         struct super_block *sb, __u32 rocompat);
1961 extern int ext4_update_incompat_feature(handle_t *handle,
1962                                         struct super_block *sb, __u32 incompat);
1963 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1964                                       struct ext4_group_desc *bg);
1965 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1966                                       struct ext4_group_desc *bg);
1967 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1968                                      struct ext4_group_desc *bg);
1969 extern __u32 ext4_free_group_clusters(struct super_block *sb,
1970                                       struct ext4_group_desc *bg);
1971 extern __u32 ext4_free_inodes_count(struct super_block *sb,
1972                                  struct ext4_group_desc *bg);
1973 extern __u32 ext4_used_dirs_count(struct super_block *sb,
1974                                 struct ext4_group_desc *bg);
1975 extern __u32 ext4_itable_unused_count(struct super_block *sb,
1976                                    struct ext4_group_desc *bg);
1977 extern void ext4_block_bitmap_set(struct super_block *sb,
1978                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1979 extern void ext4_inode_bitmap_set(struct super_block *sb,
1980                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1981 extern void ext4_inode_table_set(struct super_block *sb,
1982                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
1983 extern void ext4_free_group_clusters_set(struct super_block *sb,
1984                                          struct ext4_group_desc *bg,
1985                                          __u32 count);
1986 extern void ext4_free_inodes_set(struct super_block *sb,
1987                                 struct ext4_group_desc *bg, __u32 count);
1988 extern void ext4_used_dirs_set(struct super_block *sb,
1989                                 struct ext4_group_desc *bg, __u32 count);
1990 extern void ext4_itable_unused_set(struct super_block *sb,
1991                                    struct ext4_group_desc *bg, __u32 count);
1992 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
1993                                    struct ext4_group_desc *gdp);
1994 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
1995                                        struct ext4_group_desc *gdp);
1996
1997 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
1998 {
1999         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2000                 le32_to_cpu(es->s_blocks_count_lo);
2001 }
2002
2003 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2004 {
2005         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2006                 le32_to_cpu(es->s_r_blocks_count_lo);
2007 }
2008
2009 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2010 {
2011         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2012                 le32_to_cpu(es->s_free_blocks_count_lo);
2013 }
2014
2015 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2016                                          ext4_fsblk_t blk)
2017 {
2018         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2019         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2020 }
2021
2022 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2023                                               ext4_fsblk_t blk)
2024 {
2025         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2026         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2027 }
2028
2029 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2030                                            ext4_fsblk_t blk)
2031 {
2032         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2033         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2034 }
2035
2036 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2037 {
2038         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2039                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2040                         le32_to_cpu(raw_inode->i_size_lo);
2041         else
2042                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2043 }
2044
2045 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2046 {
2047         raw_inode->i_size_lo = cpu_to_le32(i_size);
2048         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2049 }
2050
2051 static inline
2052 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2053                                             ext4_group_t group)
2054 {
2055          struct ext4_group_info ***grp_info;
2056          long indexv, indexh;
2057          grp_info = EXT4_SB(sb)->s_group_info;
2058          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2059          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2060          return grp_info[indexv][indexh];
2061 }
2062
2063 /*
2064  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2065  * the locking protocol documented in the comments of ext4_group_add()
2066  * in resize.c
2067  */
2068 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2069 {
2070         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2071
2072         smp_rmb();
2073         return ngroups;
2074 }
2075
2076 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2077                                              ext4_group_t block_group)
2078 {
2079         return block_group >> sbi->s_log_groups_per_flex;
2080 }
2081
2082 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2083 {
2084         return 1 << sbi->s_log_groups_per_flex;
2085 }
2086
2087 #define ext4_std_error(sb, errno)                               \
2088 do {                                                            \
2089         if ((errno))                                            \
2090                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2091 } while (0)
2092
2093 #ifdef CONFIG_SMP
2094 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2095  * counters. So we need to make sure we have free clusters more
2096  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2097  */
2098 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2099 #else
2100 #define EXT4_FREECLUSTERS_WATERMARK 0
2101 #endif
2102
2103 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2104 {
2105         /*
2106          * XXX: replace with spinlock if seen contended -bzzz
2107          */
2108         down_write(&EXT4_I(inode)->i_data_sem);
2109         if (newsize > EXT4_I(inode)->i_disksize)
2110                 EXT4_I(inode)->i_disksize = newsize;
2111         up_write(&EXT4_I(inode)->i_data_sem);
2112         return ;
2113 }
2114
2115 struct ext4_group_info {
2116         unsigned long   bb_state;
2117         struct rb_root  bb_free_root;
2118         ext4_grpblk_t   bb_first_free;  /* first free block */
2119         ext4_grpblk_t   bb_free;        /* total free blocks */
2120         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2121         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2122         struct          list_head bb_prealloc_list;
2123 #ifdef DOUBLE_CHECK
2124         void            *bb_bitmap;
2125 #endif
2126         struct rw_semaphore alloc_sem;
2127         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2128                                          * regions, index is order.
2129                                          * bb_counters[3] = 5 means
2130                                          * 5 free 8-block regions. */
2131 };
2132
2133 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2134 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2135
2136 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2137         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2138
2139 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2140         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2141 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2142         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2143 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2144         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2145
2146 #define EXT4_MAX_CONTENTION             8
2147 #define EXT4_CONTENTION_THRESHOLD       2
2148
2149 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2150                                               ext4_group_t group)
2151 {
2152         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2153 }
2154
2155 /*
2156  * Returns true if the filesystem is busy enough that attempts to
2157  * access the block group locks has run into contention.
2158  */
2159 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2160 {
2161         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2162 }
2163
2164 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2165 {
2166         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2167         if (spin_trylock(lock))
2168                 /*
2169                  * We're able to grab the lock right away, so drop the
2170                  * lock contention counter.
2171                  */
2172                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2173         else {
2174                 /*
2175                  * The lock is busy, so bump the contention counter,
2176                  * and then wait on the spin lock.
2177                  */
2178                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2179                                   EXT4_MAX_CONTENTION);
2180                 spin_lock(lock);
2181         }
2182 }
2183
2184 static inline void ext4_unlock_group(struct super_block *sb,
2185                                         ext4_group_t group)
2186 {
2187         spin_unlock(ext4_group_lock_ptr(sb, group));
2188 }
2189
2190 static inline void ext4_mark_super_dirty(struct super_block *sb)
2191 {
2192         if (EXT4_SB(sb)->s_journal == NULL)
2193                 sb->s_dirt =1;
2194 }
2195
2196 /*
2197  * Block validity checking
2198  */
2199 #define ext4_check_indirect_blockref(inode, bh)                         \
2200         ext4_check_blockref(__func__, __LINE__, inode,                  \
2201                             (__le32 *)(bh)->b_data,                     \
2202                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2203
2204 #define ext4_ind_check_inode(inode)                                     \
2205         ext4_check_blockref(__func__, __LINE__, inode,                  \
2206                             EXT4_I(inode)->i_data,                      \
2207                             EXT4_NDIR_BLOCKS)
2208
2209 /*
2210  * Inodes and files operations
2211  */
2212
2213 /* dir.c */
2214 extern const struct file_operations ext4_dir_operations;
2215
2216 /* file.c */
2217 extern const struct inode_operations ext4_file_inode_operations;
2218 extern const struct file_operations ext4_file_operations;
2219 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2220
2221 /* namei.c */
2222 extern const struct inode_operations ext4_dir_inode_operations;
2223 extern const struct inode_operations ext4_special_inode_operations;
2224 extern struct dentry *ext4_get_parent(struct dentry *child);
2225
2226 /* symlink.c */
2227 extern const struct inode_operations ext4_symlink_inode_operations;
2228 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2229
2230 /* block_validity */
2231 extern void ext4_release_system_zone(struct super_block *sb);
2232 extern int ext4_setup_system_zone(struct super_block *sb);
2233 extern int __init ext4_init_system_zone(void);
2234 extern void ext4_exit_system_zone(void);
2235 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2236                                  ext4_fsblk_t start_blk,
2237                                  unsigned int count);
2238 extern int ext4_check_blockref(const char *, unsigned int,
2239                                struct inode *, __le32 *, unsigned int);
2240
2241 /* extents.c */
2242 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2243 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2244 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
2245                                        int chunk);
2246 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2247                                struct ext4_map_blocks *map, int flags);
2248 extern void ext4_ext_truncate(struct inode *);
2249 extern int ext4_ext_punch_hole(struct file *file, loff_t offset,
2250                                 loff_t length);
2251 extern void ext4_ext_init(struct super_block *);
2252 extern void ext4_ext_release(struct super_block *);
2253 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2254                           loff_t len);
2255 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
2256                           ssize_t len);
2257 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2258                            struct ext4_map_blocks *map, int flags);
2259 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2260                         __u64 start, __u64 len);
2261 /* move_extent.c */
2262 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2263                              __u64 start_orig, __u64 start_donor,
2264                              __u64 len, __u64 *moved_len);
2265
2266 /* page-io.c */
2267 extern int __init ext4_init_pageio(void);
2268 extern void ext4_exit_pageio(void);
2269 extern void ext4_ioend_wait(struct inode *);
2270 extern void ext4_free_io_end(ext4_io_end_t *io);
2271 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2272 extern int ext4_end_io_nolock(ext4_io_end_t *io);
2273 extern void ext4_io_submit(struct ext4_io_submit *io);
2274 extern int ext4_bio_write_page(struct ext4_io_submit *io,
2275                                struct page *page,
2276                                int len,
2277                                struct writeback_control *wbc);
2278
2279 /* mmp.c */
2280 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2281
2282 /* BH_Uninit flag: blocks are allocated but uninitialized on disk */
2283 enum ext4_state_bits {
2284         BH_Uninit       /* blocks are allocated but uninitialized on disk */
2285           = BH_JBDPrivateStart,
2286         BH_AllocFromCluster,    /* allocated blocks were part of already
2287                                  * allocated cluster. Note that this flag will
2288                                  * never, ever appear in a buffer_head's state
2289                                  * flag. See EXT4_MAP_FROM_CLUSTER to see where
2290                                  * this is used. */
2291         BH_Da_Mapped,   /* Delayed allocated block that now has a mapping. This
2292                          * flag is set when ext4_map_blocks is called on a
2293                          * delayed allocated block to get its real mapping. */
2294 };
2295
2296 BUFFER_FNS(Uninit, uninit)
2297 TAS_BUFFER_FNS(Uninit, uninit)
2298 BUFFER_FNS(Da_Mapped, da_mapped)
2299
2300 /*
2301  * Add new method to test wether block and inode bitmaps are properly
2302  * initialized. With uninit_bg reading the block from disk is not enough
2303  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2304  */
2305 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2306
2307 static inline int bitmap_uptodate(struct buffer_head *bh)
2308 {
2309         return (buffer_uptodate(bh) &&
2310                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2311 }
2312 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2313 {
2314         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2315 }
2316
2317 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2318
2319 /* For ioend & aio unwritten conversion wait queues */
2320 #define EXT4_WQ_HASH_SZ         37
2321 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
2322                                             EXT4_WQ_HASH_SZ])
2323 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
2324                                              EXT4_WQ_HASH_SZ])
2325 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2326 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2327
2328 #define EXT4_RESIZING   0
2329 extern int ext4_resize_begin(struct super_block *sb);
2330 extern void ext4_resize_end(struct super_block *sb);
2331
2332 #endif  /* __KERNEL__ */
2333
2334 #include "ext4_extents.h"
2335
2336 #endif  /* _EXT4_H */