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
  33 /*
  34  * The fourth extended filesystem constants/structures
  35  */
  36
  37 /*
  38  * Define EXT4FS_DEBUG to produce debug messages
  39  */
  40 #undef EXT4FS_DEBUG
  41
  42 /*
  43  * Debug code
  44  */
  45 #ifdef EXT4FS_DEBUG
  46 #define ext4_debug(f, a...)                                             \
  47         do {                                                            \
  48                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
  49                         __FILE__, __LINE__, __func__);                  \
  50                 printk(KERN_DEBUG f, ## a);                             \
  51         } while (0)
  52 #else
  53 #define ext4_debug(f, a...)     do {} while (0)
  54 #endif
  55
  56 /* data type for block offset of block group */
  57 typedef int ext4_grpblk_t;
  58
  59 /* data type for filesystem-wide blocks number */
  60 typedef unsigned long long ext4_fsblk_t;
  61
  62 /* data type for file logical block number */
  63 typedef __u32 ext4_lblk_t;
  64
  65 /* data type for block group number */
  66 typedef unsigned int ext4_group_t;
  67
  68
  69 /* prefer goal again. length */
  70 #define EXT4_MB_HINT_MERGE              1
  71 /* blocks already reserved */
  72 #define EXT4_MB_HINT_RESERVED           2
  73 /* metadata is being allocated */
  74 #define EXT4_MB_HINT_METADATA           4
  75 /* first blocks in the file */
  76 #define EXT4_MB_HINT_FIRST              8
  77 /* search for the best chunk */
  78 #define EXT4_MB_HINT_BEST               16
  79 /* data is being allocated */
  80 #define EXT4_MB_HINT_DATA               32
  81 /* don't preallocate (for tails) */
  82 #define EXT4_MB_HINT_NOPREALLOC         64
  83 /* allocate for locality group */
  84 #define EXT4_MB_HINT_GROUP_ALLOC        128
  85 /* allocate goal blocks or none */
  86 #define EXT4_MB_HINT_GOAL_ONLY          256
  87 /* goal is meaningful */
  88 #define EXT4_MB_HINT_TRY_GOAL           512
  89 /* blocks already pre-reserved by delayed allocation */
  90 #define EXT4_MB_DELALLOC_RESERVED      1024
  91
  92
  93 struct ext4_allocation_request {
  94         /* target inode for block we're allocating */
  95         struct inode *inode;
  96         /* logical block in target inode */
  97         ext4_lblk_t logical;
  98         /* phys. target (a hint) */
  99         ext4_fsblk_t goal;
 100         /* the closest logical allocated block to the left */
 101         ext4_lblk_t lleft;
 102         /* phys. block for ^^^ */
 103         ext4_fsblk_t pleft;
 104         /* the closest logical allocated block to the right */
 105         ext4_lblk_t lright;
 106         /* phys. block for ^^^ */
 107         ext4_fsblk_t pright;
 108         /* how many blocks we want to allocate */
 109         unsigned int len;
 110         /* flags. see above EXT4_MB_HINT_* */
 111         unsigned int flags;
 112 };
 113
 114 /*
 115  * Special inodes numbers
 116  */
 117 #define EXT4_BAD_INO             1      /* Bad blocks inode */
 118 #define EXT4_ROOT_INO            2      /* Root inode */
 119 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
 120 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
 121 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
 122 #define EXT4_JOURNAL_INO         8      /* Journal inode */
 123
 124 /* First non-reserved inode for old ext4 filesystems */
 125 #define EXT4_GOOD_OLD_FIRST_INO 11
 126
 127 /*
 128  * Maximal count of links to a file
 129  */
 130 #define EXT4_LINK_MAX           65000
 131
 132 /*
 133  * Macro-instructions used to manage several block sizes
 134  */
 135 #define EXT4_MIN_BLOCK_SIZE             1024
 136 #define EXT4_MAX_BLOCK_SIZE             65536
 137 #define EXT4_MIN_BLOCK_LOG_SIZE         10
 138 #ifdef __KERNEL__
 139 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
 140 #else
 141 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
 142 #endif
 143 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
 144 #ifdef __KERNEL__
 145 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
 146 #else
 147 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
 148 #endif
 149 #ifdef __KERNEL__
 150 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
 151 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
 152 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
 153 #else
 154 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 155                                  EXT4_GOOD_OLD_INODE_SIZE : \
 156                                  (s)->s_inode_size)
 157 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 158                                  EXT4_GOOD_OLD_FIRST_INO : \
 159                                  (s)->s_first_ino)
 160 #endif
 161 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
 162
 163 /*
 164  * Structure of a blocks group descriptor
 165  */
 166 struct ext4_group_desc
 167 {
 168         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
 169         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
 170         __le32  bg_inode_table_lo;      /* Inodes table block */
 171         __le16  bg_free_blocks_count_lo;/* Free blocks count */
 172         __le16  bg_free_inodes_count_lo;/* Free inodes count */
 173         __le16  bg_used_dirs_count_lo;  /* Directories count */
 174         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
 175         __u32   bg_reserved[2];         /* Likely block/inode bitmap checksum */
 176         __le16  bg_itable_unused_lo;    /* Unused inodes count */
 177         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
 178         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
 179         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
 180         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
 181         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
 182         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
 183         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
 184         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
 185         __u32   bg_reserved2[3];
 186 };
 187
 188 /*
 189  * Structure of a flex block group info
 190  */
 191
 192 struct flex_groups {
 193         atomic_t free_inodes;
 194         atomic_t free_blocks;
 195         atomic_t used_dirs;
 196 };
 197
 198 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
 199 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
 200 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
 201
 202 /*
 203  * Macro-instructions used to manage group descriptors
 204  */
 205 #define EXT4_MIN_DESC_SIZE              32
 206 #define EXT4_MIN_DESC_SIZE_64BIT        64
 207 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
 208 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
 209 #ifdef __KERNEL__
 210 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
 211 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
 212 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
 213 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
 214 #else
 215 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
 216 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
 217 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
 218 #endif
 219
 220 /*
 221  * Constants relative to the data blocks
 222  */
 223 #define EXT4_NDIR_BLOCKS                12
 224 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
 225 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
 226 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
 227 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
 228
 229 /*
 230  * Inode flags
 231  */
 232 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
 233 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
 234 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
 235 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
 236 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
 237 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
 238 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
 239 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
 240 /* Reserved for compression usage... */
 241 #define EXT4_DIRTY_FL                   0x00000100
 242 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
 243 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
 244 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
 245 /* End compression flags --- maybe not all used */
 246 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
 247 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
 248 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
 249 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
 250 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
 251 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
 252 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
 253 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
 254 #define EXT4_EXT_MIGRATE                0x00100000 /* Inode is migrating */
 255 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
 256
 257 #define EXT4_FL_USER_VISIBLE            0x000BDFFF /* User visible flags */
 258 #define EXT4_FL_USER_MODIFIABLE         0x000B80FF /* User modifiable flags */
 259
 260 /* Flags that should be inherited by new inodes from their parent. */
 261 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
 262                            EXT4_SYNC_FL | EXT4_IMMUTABLE_FL | EXT4_APPEND_FL |\
 263                            EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
 264                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
 265                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
 266
 267 /* Flags that are appropriate for regular files (all but dir-specific ones). */
 268 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
 269
 270 /* Flags that are appropriate for non-directories/regular files. */
 271 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
 272
 273 /* Mask out flags that are inappropriate for the given type of inode. */
 274 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
 275 {
 276         if (S_ISDIR(mode))
 277                 return flags;
 278         else if (S_ISREG(mode))
 279                 return flags & EXT4_REG_FLMASK;
 280         else
 281                 return flags & EXT4_OTHER_FLMASK;
 282 }
 283
 284 /*
 285  * Inode dynamic state flags
 286  */
 287 #define EXT4_STATE_JDATA                0x00000001 /* journaled data exists */
 288 #define EXT4_STATE_NEW                  0x00000002 /* inode is newly created */
 289 #define EXT4_STATE_XATTR                0x00000004 /* has in-inode xattrs */
 290 #define EXT4_STATE_NO_EXPAND            0x00000008 /* No space for expansion */
 291 #define EXT4_STATE_DA_ALLOC_CLOSE       0x00000010 /* Alloc DA blks on close */
 292
 293 /* Used to pass group descriptor data when online resize is done */
 294 struct ext4_new_group_input {
 295         __u32 group;            /* Group number for this data */
 296         __u64 block_bitmap;     /* Absolute block number of block bitmap */
 297         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
 298         __u64 inode_table;      /* Absolute block number of inode table start */
 299         __u32 blocks_count;     /* Total number of blocks in this group */
 300         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
 301         __u16 unused;
 302 };
 303
 304 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
 305 struct ext4_new_group_data {
 306         __u32 group;
 307         __u64 block_bitmap;
 308         __u64 inode_bitmap;
 309         __u64 inode_table;
 310         __u32 blocks_count;
 311         __u16 reserved_blocks;
 312         __u16 unused;
 313         __u32 free_blocks_count;
 314 };
 315
 316 /*
 317  * Flags used by ext4_get_blocks()
 318  */
 319         /* Allocate any needed blocks and/or convert an unitialized
 320            extent to be an initialized ext4 */
 321 #define EXT4_GET_BLOCKS_CREATE                  0x0001
 322         /* Request the creation of an unitialized extent */
 323 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
 324 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
 325                                                  EXT4_GET_BLOCKS_CREATE)
 326         /* Caller is from the delayed allocation writeout path,
 327            so set the magic i_delalloc_reserve_flag after taking the
 328            inode allocation semaphore for */
 329 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
 330         /* Call ext4_da_update_reserve_space() after successfully
 331            allocating the blocks */
 332 #define EXT4_GET_BLOCKS_UPDATE_RESERVE_SPACE    0x0008
 333
 334
 335 /*
 336  * ioctl commands
 337  */
 338 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
 339 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
 340 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
 341 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
 342 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
 343 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
 344 #ifdef CONFIG_JBD2_DEBUG
 345 #define EXT4_IOC_WAIT_FOR_READONLY      _IOR('f', 99, long)
 346 #endif
 347 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
 348 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
 349 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
 350 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
 351 #define EXT4_IOC_MIGRATE                _IO('f', 9)
 352  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
 353  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
 354 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
 355
 356 /*
 357  * ioctl commands in 32 bit emulation
 358  */
 359 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
 360 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
 361 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
 362 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
 363 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
 364 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
 365 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
 366 #ifdef CONFIG_JBD2_DEBUG
 367 #define EXT4_IOC32_WAIT_FOR_READONLY    _IOR('f', 99, int)
 368 #endif
 369 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
 370 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
 371
 372
 373 /*
 374  *  Mount options
 375  */
 376 struct ext4_mount_options {
 377         unsigned long s_mount_opt;
 378         uid_t s_resuid;
 379         gid_t s_resgid;
 380         unsigned long s_commit_interval;
 381         u32 s_min_batch_time, s_max_batch_time;
 382 #ifdef CONFIG_QUOTA
 383         int s_jquota_fmt;
 384         char *s_qf_names[MAXQUOTAS];
 385 #endif
 386 };
 387
 388 /*
 389  * Structure of an inode on the disk
 390  */
 391 struct ext4_inode {
 392         __le16  i_mode;         /* File mode */
 393         __le16  i_uid;          /* Low 16 bits of Owner Uid */
 394         __le32  i_size_lo;      /* Size in bytes */
 395         __le32  i_atime;        /* Access time */
 396         __le32  i_ctime;        /* Inode Change time */
 397         __le32  i_mtime;        /* Modification time */
 398         __le32  i_dtime;        /* Deletion Time */
 399         __le16  i_gid;          /* Low 16 bits of Group Id */
 400         __le16  i_links_count;  /* Links count */
 401         __le32  i_blocks_lo;    /* Blocks count */
 402         __le32  i_flags;        /* File flags */
 403         union {
 404                 struct {
 405                         __le32  l_i_version;
 406                 } linux1;
 407                 struct {
 408                         __u32  h_i_translator;
 409                 } hurd1;
 410                 struct {
 411                         __u32  m_i_reserved1;
 412                 } masix1;
 413         } osd1;                         /* OS dependent 1 */
 414         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
 415         __le32  i_generation;   /* File version (for NFS) */
 416         __le32  i_file_acl_lo;  /* File ACL */
 417         __le32  i_size_high;
 418         __le32  i_obso_faddr;   /* Obsoleted fragment address */
 419         union {
 420                 struct {
 421                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
 422                         __le16  l_i_file_acl_high;
 423                         __le16  l_i_uid_high;   /* these 2 fields */
 424                         __le16  l_i_gid_high;   /* were reserved2[0] */
 425                         __u32   l_i_reserved2;
 426                 } linux2;
 427                 struct {
 428                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 429                         __u16   h_i_mode_high;
 430                         __u16   h_i_uid_high;
 431                         __u16   h_i_gid_high;
 432                         __u32   h_i_author;
 433                 } hurd2;
 434                 struct {
 435                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 436                         __le16  m_i_file_acl_high;
 437                         __u32   m_i_reserved2[2];
 438                 } masix2;
 439         } osd2;                         /* OS dependent 2 */
 440         __le16  i_extra_isize;
 441         __le16  i_pad1;
 442         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
 443         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
 444         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
 445         __le32  i_crtime;       /* File Creation time */
 446         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
 447         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
 448 };
 449
 450
 451 #define EXT4_EPOCH_BITS 2
 452 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
 453 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
 454
 455 /*
 456  * Extended fields will fit into an inode if the filesystem was formatted
 457  * with large inodes (-I 256 or larger) and there are not currently any EAs
 458  * consuming all of the available space. For new inodes we always reserve
 459  * enough space for the kernel's known extended fields, but for inodes
 460  * created with an old kernel this might not have been the case. None of
 461  * the extended inode fields is critical for correct filesystem operation.
 462  * This macro checks if a certain field fits in the inode. Note that
 463  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
 464  */
 465 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
 466         ((offsetof(typeof(*ext4_inode), field) +        \
 467           sizeof((ext4_inode)->field))                  \
 468         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
 469             (einode)->i_extra_isize))                   \
 470
 471 static inline __le32 ext4_encode_extra_time(struct timespec *time)
 472 {
 473        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
 474                            time->tv_sec >> 32 : 0) |
 475                            ((time->tv_nsec << 2) & EXT4_NSEC_MASK));
 476 }
 477
 478 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
 479 {
 480        if (sizeof(time->tv_sec) > 4)
 481                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
 482                                << 32;
 483        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> 2;
 484 }
 485
 486 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
 487 do {                                                                           \
 488         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
 489         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 490                 (raw_inode)->xtime ## _extra =                                 \
 491                                 ext4_encode_extra_time(&(inode)->xtime);       \
 492 } while (0)
 493
 494 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
 495 do {                                                                           \
 496         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 497                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
 498         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 499                 (raw_inode)->xtime ## _extra =                                 \
 500                                 ext4_encode_extra_time(&(einode)->xtime);      \
 501 } while (0)
 502
 503 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
 504 do {                                                                           \
 505         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
 506         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 507                 ext4_decode_extra_time(&(inode)->xtime,                        \
 508                                        raw_inode->xtime ## _extra);            \
 509 } while (0)
 510
 511 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
 512 do {                                                                           \
 513         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 514                 (einode)->xtime.tv_sec =                                       \
 515                         (signed)le32_to_cpu((raw_inode)->xtime);               \
 516         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 517                 ext4_decode_extra_time(&(einode)->xtime,                       \
 518                                        raw_inode->xtime ## _extra);            \
 519 } while (0)
 520
 521 #define i_disk_version osd1.linux1.l_i_version
 522
 523 #if defined(__KERNEL__) || defined(__linux__)
 524 #define i_reserved1     osd1.linux1.l_i_reserved1
 525 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
 526 #define i_blocks_high   osd2.linux2.l_i_blocks_high
 527 #define i_uid_low       i_uid
 528 #define i_gid_low       i_gid
 529 #define i_uid_high      osd2.linux2.l_i_uid_high
 530 #define i_gid_high      osd2.linux2.l_i_gid_high
 531 #define i_reserved2     osd2.linux2.l_i_reserved2
 532
 533 #elif defined(__GNU__)
 534
 535 #define i_translator    osd1.hurd1.h_i_translator
 536 #define i_uid_high      osd2.hurd2.h_i_uid_high
 537 #define i_gid_high      osd2.hurd2.h_i_gid_high
 538 #define i_author        osd2.hurd2.h_i_author
 539
 540 #elif defined(__masix__)
 541
 542 #define i_reserved1     osd1.masix1.m_i_reserved1
 543 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
 544 #define i_reserved2     osd2.masix2.m_i_reserved2
 545
 546 #endif /* defined(__KERNEL__) || defined(__linux__) */
 547
 548 /*
 549  * storage for cached extent
 550  */
 551 struct ext4_ext_cache {
 552         ext4_fsblk_t    ec_start;
 553         ext4_lblk_t     ec_block;
 554         __u32           ec_len; /* must be 32bit to return holes */
 555         __u32           ec_type;
 556 };
 557
 558 /*
 559  * fourth extended file system inode data in memory
 560  */
 561 struct ext4_inode_info {
 562         __le32  i_data[15];     /* unconverted */
 563         __u32   i_flags;
 564         ext4_fsblk_t    i_file_acl;
 565         __u32   i_dtime;
 566
 567         /*
 568          * i_block_group is the number of the block group which contains
 569          * this file's inode.  Constant across the lifetime of the inode,
 570          * it is ued for making block allocation decisions - we try to
 571          * place a file's data blocks near its inode block, and new inodes
 572          * near to their parent directory's inode.
 573          */
 574         ext4_group_t    i_block_group;
 575         __u32   i_state;                /* Dynamic state flags for ext4 */
 576
 577         ext4_lblk_t             i_dir_start_lookup;
 578 #ifdef CONFIG_EXT4_FS_XATTR
 579         /*
 580          * Extended attributes can be read independently of the main file
 581          * data. Taking i_mutex even when reading would cause contention
 582          * between readers of EAs and writers of regular file data, so
 583          * instead we synchronize on xattr_sem when reading or changing
 584          * EAs.
 585          */
 586         struct rw_semaphore xattr_sem;
 587 #endif
 588 #ifdef CONFIG_EXT4_FS_POSIX_ACL
 589         struct posix_acl        *i_acl;
 590         struct posix_acl        *i_default_acl;
 591 #endif
 592
 593         struct list_head i_orphan;      /* unlinked but open inodes */
 594
 595         /*
 596          * i_disksize keeps track of what the inode size is ON DISK, not
 597          * in memory.  During truncate, i_size is set to the new size by
 598          * the VFS prior to calling ext4_truncate(), but the filesystem won't
 599          * set i_disksize to 0 until the truncate is actually under way.
 600          *
 601          * The intent is that i_disksize always represents the blocks which
 602          * are used by this file.  This allows recovery to restart truncate
 603          * on orphans if we crash during truncate.  We actually write i_disksize
 604          * into the on-disk inode when writing inodes out, instead of i_size.
 605          *
 606          * The only time when i_disksize and i_size may be different is when
 607          * a truncate is in progress.  The only things which change i_disksize
 608          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
 609          */
 610         loff_t  i_disksize;
 611
 612         /*
 613          * i_data_sem is for serialising ext4_truncate() against
 614          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
 615          * data tree are chopped off during truncate. We can't do that in
 616          * ext4 because whenever we perform intermediate commits during
 617          * truncate, the inode and all the metadata blocks *must* be in a
 618          * consistent state which allows truncation of the orphans to restart
 619          * during recovery.  Hence we must fix the get_block-vs-truncate race
 620          * by other means, so we have i_data_sem.
 621          */
 622         struct rw_semaphore i_data_sem;
 623         struct inode vfs_inode;
 624         struct jbd2_inode jinode;
 625
 626         struct ext4_ext_cache i_cached_extent;
 627         /*
 628          * File creation time. Its function is same as that of
 629          * struct timespec i_{a,c,m}time in the generic inode.
 630          */
 631         struct timespec i_crtime;
 632
 633         /* mballoc */
 634         struct list_head i_prealloc_list;
 635         spinlock_t i_prealloc_lock;
 636
 637         /* ialloc */
 638         ext4_group_t    i_last_alloc_group;
 639
 640         /* allocation reservation info for delalloc */
 641         unsigned int i_reserved_data_blocks;
 642         unsigned int i_reserved_meta_blocks;
 643         unsigned int i_allocated_meta_blocks;
 644         unsigned short i_delalloc_reserved_flag;
 645
 646         /* on-disk additional length */
 647         __u16 i_extra_isize;
 648
 649         spinlock_t i_block_reservation_lock;
 650 };
 651
 652 /*
 653  * File system states
 654  */
 655 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
 656 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
 657 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
 658
 659 /*
 660  * Misc. filesystem flags
 661  */
 662 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
 663 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
 664 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
 665
 666 /*
 667  * Mount flags
 668  */
 669 #define EXT4_MOUNT_OLDALLOC             0x00002  /* Don't use the new Orlov allocator */
 670 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
 671 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
 672 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
 673 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
 674 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
 675 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
 676 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
 677 #define EXT4_MOUNT_ABORT                0x00200 /* Fatal error detected */
 678 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
 679 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
 680 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
 681 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
 682 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
 683 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
 684 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
 685 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
 686 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
 687 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
 688 #define EXT4_MOUNT_NOBH                 0x40000 /* No bufferheads */
 689 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
 690 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
 691 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
 692 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
 693 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
 694 #define EXT4_MOUNT_I_VERSION            0x2000000 /* i_version support */
 695 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
 696 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
 697 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
 698
 699 /* Compatibility, for having both ext2_fs.h and ext4_fs.h included at once */
 700 #ifndef _LINUX_EXT2_FS_H
 701 #define clear_opt(o, opt)               o &= ~EXT4_MOUNT_##opt
 702 #define set_opt(o, opt)                 o |= EXT4_MOUNT_##opt
 703 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
 704                                          EXT4_MOUNT_##opt)
 705 #else
 706 #define EXT2_MOUNT_NOLOAD               EXT4_MOUNT_NOLOAD
 707 #define EXT2_MOUNT_ABORT                EXT4_MOUNT_ABORT
 708 #define EXT2_MOUNT_DATA_FLAGS           EXT4_MOUNT_DATA_FLAGS
 709 #endif
 710
 711 #define ext4_set_bit                    ext2_set_bit
 712 #define ext4_set_bit_atomic             ext2_set_bit_atomic
 713 #define ext4_clear_bit                  ext2_clear_bit
 714 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
 715 #define ext4_test_bit                   ext2_test_bit
 716 #define ext4_find_first_zero_bit        ext2_find_first_zero_bit
 717 #define ext4_find_next_zero_bit         ext2_find_next_zero_bit
 718 #define ext4_find_next_bit              ext2_find_next_bit
 719
 720 /*
 721  * Maximal mount counts between two filesystem checks
 722  */
 723 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
 724 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
 725
 726 /*
 727  * Behaviour when detecting errors
 728  */
 729 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
 730 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
 731 #define EXT4_ERRORS_PANIC               3       /* Panic */
 732 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
 733
 734 /*
 735  * Structure of the super block
 736  */
 737 struct ext4_super_block {
 738 /*00*/  __le32  s_inodes_count;         /* Inodes count */
 739         __le32  s_blocks_count_lo;      /* Blocks count */
 740         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
 741         __le32  s_free_blocks_count_lo; /* Free blocks count */
 742 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
 743         __le32  s_first_data_block;     /* First Data Block */
 744         __le32  s_log_block_size;       /* Block size */
 745         __le32  s_obso_log_frag_size;   /* Obsoleted fragment size */
 746 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
 747         __le32  s_obso_frags_per_group; /* Obsoleted fragments per group */
 748         __le32  s_inodes_per_group;     /* # Inodes per group */
 749         __le32  s_mtime;                /* Mount time */
 750 /*30*/  __le32  s_wtime;                /* Write time */
 751         __le16  s_mnt_count;            /* Mount count */
 752         __le16  s_max_mnt_count;        /* Maximal mount count */
 753         __le16  s_magic;                /* Magic signature */
 754         __le16  s_state;                /* File system state */
 755         __le16  s_errors;               /* Behaviour when detecting errors */
 756         __le16  s_minor_rev_level;      /* minor revision level */
 757 /*40*/  __le32  s_lastcheck;            /* time of last check */
 758         __le32  s_checkinterval;        /* max. time between checks */
 759         __le32  s_creator_os;           /* OS */
 760         __le32  s_rev_level;            /* Revision level */
 761 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
 762         __le16  s_def_resgid;           /* Default gid for reserved blocks */
 763         /*
 764          * These fields are for EXT4_DYNAMIC_REV superblocks only.
 765          *
 766          * Note: the difference between the compatible feature set and
 767          * the incompatible feature set is that if there is a bit set
 768          * in the incompatible feature set that the kernel doesn't
 769          * know about, it should refuse to mount the filesystem.
 770          *
 771          * e2fsck's requirements are more strict; if it doesn't know
 772          * about a feature in either the compatible or incompatible
 773          * feature set, it must abort and not try to meddle with
 774          * things it doesn't understand...
 775          */
 776         __le32  s_first_ino;            /* First non-reserved inode */
 777         __le16  s_inode_size;           /* size of inode structure */
 778         __le16  s_block_group_nr;       /* block group # of this superblock */
 779         __le32  s_feature_compat;       /* compatible feature set */
 780 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
 781         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
 782 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
 783 /*78*/  char    s_volume_name[16];      /* volume name */
 784 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
 785 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
 786         /*
 787          * Performance hints.  Directory preallocation should only
 788          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
 789          */
 790         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
 791         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
 792         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
 793         /*
 794          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
 795          */
 796 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
 797 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
 798         __le32  s_journal_dev;          /* device number of journal file */
 799         __le32  s_last_orphan;          /* start of list of inodes to delete */
 800         __le32  s_hash_seed[4];         /* HTREE hash seed */
 801         __u8    s_def_hash_version;     /* Default hash version to use */
 802         __u8    s_reserved_char_pad;
 803         __le16  s_desc_size;            /* size of group descriptor */
 804 /*100*/ __le32  s_default_mount_opts;
 805         __le32  s_first_meta_bg;        /* First metablock block group */
 806         __le32  s_mkfs_time;            /* When the filesystem was created */
 807         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
 808         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
 809 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
 810         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
 811         __le32  s_free_blocks_count_hi; /* Free blocks count */
 812         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
 813         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
 814         __le32  s_flags;                /* Miscellaneous flags */
 815         __le16  s_raid_stride;          /* RAID stride */
 816         __le16  s_mmp_interval;         /* # seconds to wait in MMP checking */
 817         __le64  s_mmp_block;            /* Block for multi-mount protection */
 818         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
 819         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
 820         __u8    s_reserved_char_pad2;
 821         __le16  s_reserved_pad;
 822         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
 823         __u32   s_reserved[160];        /* Padding to the end of the block */
 824 };
 825
 826 #ifdef __KERNEL__
 827 /*
 828  * fourth extended-fs super-block data in memory
 829  */
 830 struct ext4_sb_info {
 831         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
 832         unsigned long s_inodes_per_block;/* Number of inodes per block */
 833         unsigned long s_blocks_per_group;/* Number of blocks in a group */
 834         unsigned long s_inodes_per_group;/* Number of inodes in a group */
 835         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
 836         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
 837         unsigned long s_desc_per_block; /* Number of group descriptors per block */
 838         ext4_group_t s_groups_count;    /* Number of groups in the fs */
 839         unsigned long s_overhead_last;  /* Last calculated overhead */
 840         unsigned long s_blocks_last;    /* Last seen block count */
 841         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
 842         struct buffer_head * s_sbh;     /* Buffer containing the super block */
 843         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
 844         struct buffer_head **s_group_desc;
 845         unsigned long  s_mount_opt;
 846         ext4_fsblk_t s_sb_block;
 847         uid_t s_resuid;
 848         gid_t s_resgid;
 849         unsigned short s_mount_state;
 850         unsigned short s_pad;
 851         int s_addr_per_block_bits;
 852         int s_desc_per_block_bits;
 853         int s_inode_size;
 854         int s_first_ino;
 855         unsigned int s_inode_readahead_blks;
 856         spinlock_t s_next_gen_lock;
 857         u32 s_next_generation;
 858         u32 s_hash_seed[4];
 859         int s_def_hash_version;
 860         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
 861         struct percpu_counter s_freeblocks_counter;
 862         struct percpu_counter s_freeinodes_counter;
 863         struct percpu_counter s_dirs_counter;
 864         struct percpu_counter s_dirtyblocks_counter;
 865         struct blockgroup_lock *s_blockgroup_lock;
 866         struct proc_dir_entry *s_proc;
 867         struct kobject s_kobj;
 868         struct completion s_kobj_unregister;
 869
 870         /* Journaling */
 871         struct inode *s_journal_inode;
 872         struct journal_s *s_journal;
 873         struct list_head s_orphan;
 874         struct mutex s_orphan_lock;
 875         struct mutex s_resize_lock;
 876         unsigned long s_commit_interval;
 877         u32 s_max_batch_time;
 878         u32 s_min_batch_time;
 879         struct block_device *journal_bdev;
 880 #ifdef CONFIG_JBD2_DEBUG
 881         struct timer_list turn_ro_timer;        /* For turning read-only (crash simulation) */
 882         wait_queue_head_t ro_wait_queue;        /* For people waiting for the fs to go read-only */
 883 #endif
 884 #ifdef CONFIG_QUOTA
 885         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
 886         int s_jquota_fmt;                       /* Format of quota to use */
 887 #endif
 888         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
 889         struct rb_root system_blks;
 890
 891 #ifdef EXTENTS_STATS
 892         /* ext4 extents stats */
 893         unsigned long s_ext_min;
 894         unsigned long s_ext_max;
 895         unsigned long s_depth_max;
 896         spinlock_t s_ext_stats_lock;
 897         unsigned long s_ext_blocks;
 898         unsigned long s_ext_extents;
 899 #endif
 900
 901         /* for buddy allocator */
 902         struct ext4_group_info ***s_group_info;
 903         struct inode *s_buddy_cache;
 904         long s_blocks_reserved;
 905         spinlock_t s_reserve_lock;
 906         spinlock_t s_md_lock;
 907         tid_t s_last_transaction;
 908         unsigned short *s_mb_offsets;
 909         unsigned int *s_mb_maxs;
 910
 911         /* tunables */
 912         unsigned long s_stripe;
 913         unsigned int s_mb_stream_request;
 914         unsigned int s_mb_max_to_scan;
 915         unsigned int s_mb_min_to_scan;
 916         unsigned int s_mb_stats;
 917         unsigned int s_mb_order2_reqs;
 918         unsigned int s_mb_group_prealloc;
 919         /* where last allocation was done - for stream allocation */
 920         unsigned long s_mb_last_group;
 921         unsigned long s_mb_last_start;
 922
 923         /* history to debug policy */
 924         struct ext4_mb_history *s_mb_history;
 925         int s_mb_history_cur;
 926         int s_mb_history_max;
 927         int s_mb_history_num;
 928         spinlock_t s_mb_history_lock;
 929         int s_mb_history_filter;
 930
 931         /* stats for buddy allocator */
 932         spinlock_t s_mb_pa_lock;
 933         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
 934         atomic_t s_bal_success; /* we found long enough chunks */
 935         atomic_t s_bal_allocated;       /* in blocks */
 936         atomic_t s_bal_ex_scanned;      /* total extents scanned */
 937         atomic_t s_bal_goals;   /* goal hits */
 938         atomic_t s_bal_breaks;  /* too long searches */
 939         atomic_t s_bal_2orders; /* 2^order hits */
 940         spinlock_t s_bal_lock;
 941         unsigned long s_mb_buddies_generated;
 942         unsigned long long s_mb_generation_time;
 943         atomic_t s_mb_lost_chunks;
 944         atomic_t s_mb_preallocated;
 945         atomic_t s_mb_discarded;
 946
 947         /* locality groups */
 948         struct ext4_locality_group *s_locality_groups;
 949
 950         /* for write statistics */
 951         unsigned long s_sectors_written_start;
 952         u64 s_kbytes_written;
 953
 954         unsigned int s_log_groups_per_flex;
 955         struct flex_groups *s_flex_groups;
 956 };
 957
 958 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
 959 {
 960         return sb->s_fs_info;
 961 }
 962 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
 963 {
 964         return container_of(inode, struct ext4_inode_info, vfs_inode);
 965 }
 966
 967 static inline struct timespec ext4_current_time(struct inode *inode)
 968 {
 969         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
 970                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
 971 }
 972
 973 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
 974 {
 975         return ino == EXT4_ROOT_INO ||
 976                 ino == EXT4_JOURNAL_INO ||
 977                 ino == EXT4_RESIZE_INO ||
 978                 (ino >= EXT4_FIRST_INO(sb) &&
 979                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
 980 }
 981 #else
 982 /* Assume that user mode programs are passing in an ext4fs superblock, not
 983  * a kernel struct super_block.  This will allow us to call the feature-test
 984  * macros from user land. */
 985 #define EXT4_SB(sb)     (sb)
 986 #endif
 987
 988 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
 989
 990 /*
 991  * Codes for operating systems
 992  */
 993 #define EXT4_OS_LINUX           0
 994 #define EXT4_OS_HURD            1
 995 #define EXT4_OS_MASIX           2
 996 #define EXT4_OS_FREEBSD         3
 997 #define EXT4_OS_LITES           4
 998
 999 /*
1000  * Revision levels
1001  */
1002 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1003 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1004
1005 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1006 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1007
1008 #define EXT4_GOOD_OLD_INODE_SIZE 128
1009
1010 /*
1011  * Feature set definitions
1012  */
1013
1014 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1015         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1016 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1017         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1018 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1019         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1020 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1021         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1022 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1023         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1024 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1025         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1026 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1027         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1028 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1029         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1030 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1031         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1032
1033 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1034 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1035 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1036 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1037 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1038 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1039
1040 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1041 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1042 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1043 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1044 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1045 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1046 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1047
1048 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1049 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1050 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1051 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1052 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1053 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1054 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1055 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1056 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1057
1058 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1059 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1060                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1061                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1062                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1063                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1064                                          EXT4_FEATURE_INCOMPAT_FLEX_BG)
1065 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1066                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1067                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1068                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1069                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1070                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1071                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE)
1072
1073 /*
1074  * Default values for user and/or group using reserved blocks
1075  */
1076 #define EXT4_DEF_RESUID         0
1077 #define EXT4_DEF_RESGID         0
1078
1079 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1080
1081 /*
1082  * Default mount options
1083  */
1084 #define EXT4_DEFM_DEBUG         0x0001
1085 #define EXT4_DEFM_BSDGROUPS     0x0002
1086 #define EXT4_DEFM_XATTR_USER    0x0004
1087 #define EXT4_DEFM_ACL           0x0008
1088 #define EXT4_DEFM_UID16         0x0010
1089 #define EXT4_DEFM_JMODE         0x0060
1090 #define EXT4_DEFM_JMODE_DATA    0x0020
1091 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1092 #define EXT4_DEFM_JMODE_WBACK   0x0060
1093
1094 /*
1095  * Default journal batch times
1096  */
1097 #define EXT4_DEF_MIN_BATCH_TIME 0
1098 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1099
1100 /*
1101  * Minimum number of groups in a flexgroup before we separate out
1102  * directories into the first block group of a flexgroup
1103  */
1104 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1105
1106 /*
1107  * Structure of a directory entry
1108  */
1109 #define EXT4_NAME_LEN 255
1110
1111 struct ext4_dir_entry {
1112         __le32  inode;                  /* Inode number */
1113         __le16  rec_len;                /* Directory entry length */
1114         __le16  name_len;               /* Name length */
1115         char    name[EXT4_NAME_LEN];    /* File name */
1116 };
1117
1118 /*
1119  * The new version of the directory entry.  Since EXT4 structures are
1120  * stored in intel byte order, and the name_len field could never be
1121  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1122  * file_type field.
1123  */
1124 struct ext4_dir_entry_2 {
1125         __le32  inode;                  /* Inode number */
1126         __le16  rec_len;                /* Directory entry length */
1127         __u8    name_len;               /* Name length */
1128         __u8    file_type;
1129         char    name[EXT4_NAME_LEN];    /* File name */
1130 };
1131
1132 /*
1133  * Ext4 directory file types.  Only the low 3 bits are used.  The
1134  * other bits are reserved for now.
1135  */
1136 #define EXT4_FT_UNKNOWN         0
1137 #define EXT4_FT_REG_FILE        1
1138 #define EXT4_FT_DIR             2
1139 #define EXT4_FT_CHRDEV          3
1140 #define EXT4_FT_BLKDEV          4
1141 #define EXT4_FT_FIFO            5
1142 #define EXT4_FT_SOCK            6
1143 #define EXT4_FT_SYMLINK         7
1144
1145 #define EXT4_FT_MAX             8
1146
1147 /*
1148  * EXT4_DIR_PAD defines the directory entries boundaries
1149  *
1150  * NOTE: It must be a multiple of 4
1151  */
1152 #define EXT4_DIR_PAD                    4
1153 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1154 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1155                                          ~EXT4_DIR_ROUND)
1156 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1157
1158 /*
1159  * Hash Tree Directory indexing
1160  * (c) Daniel Phillips, 2001
1161  */
1162
1163 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1164                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1165                       (EXT4_I(dir)->i_flags & EXT4_INDEX_FL))
1166 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1167 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1168
1169 /* Legal values for the dx_root hash_version field: */
1170
1171 #define DX_HASH_LEGACY          0
1172 #define DX_HASH_HALF_MD4        1
1173 #define DX_HASH_TEA             2
1174 #define DX_HASH_LEGACY_UNSIGNED 3
1175 #define DX_HASH_HALF_MD4_UNSIGNED       4
1176 #define DX_HASH_TEA_UNSIGNED            5
1177
1178 #ifdef __KERNEL__
1179
1180 /* hash info structure used by the directory hash */
1181 struct dx_hash_info
1182 {
1183         u32             hash;
1184         u32             minor_hash;
1185         int             hash_version;
1186         u32             *seed;
1187 };
1188
1189 #define EXT4_HTREE_EOF  0x7fffffff
1190
1191 /*
1192  * Control parameters used by ext4_htree_next_block
1193  */
1194 #define HASH_NB_ALWAYS          1
1195
1196
1197 /*
1198  * Describe an inode's exact location on disk and in memory
1199  */
1200 struct ext4_iloc
1201 {
1202         struct buffer_head *bh;
1203         unsigned long offset;
1204         ext4_group_t block_group;
1205 };
1206
1207 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1208 {
1209         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1210 }
1211
1212 /*
1213  * This structure is stuffed into the struct file's private_data field
1214  * for directories.  It is where we put information so that we can do
1215  * readdir operations in hash tree order.
1216  */
1217 struct dir_private_info {
1218         struct rb_root  root;
1219         struct rb_node  *curr_node;
1220         struct fname    *extra_fname;
1221         loff_t          last_pos;
1222         __u32           curr_hash;
1223         __u32           curr_minor_hash;
1224         __u32           next_hash;
1225 };
1226
1227 /* calculate the first block number of the group */
1228 static inline ext4_fsblk_t
1229 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1230 {
1231         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1232                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1233 }
1234
1235 /*
1236  * Special error return code only used by dx_probe() and its callers.
1237  */
1238 #define ERR_BAD_DX_DIR  -75000
1239
1240 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1241                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1242
1243 extern struct proc_dir_entry *ext4_proc_root;
1244
1245 /*
1246  * Function prototypes
1247  */
1248
1249 /*
1250  * Ok, these declarations are also in <linux/kernel.h> but none of the
1251  * ext4 source programs needs to include it so they are duplicated here.
1252  */
1253 # define NORET_TYPE     /**/
1254 # define ATTRIB_NORET   __attribute__((noreturn))
1255 # define NORET_AND      noreturn,
1256
1257 /* bitmap.c */
1258 extern unsigned int ext4_count_free(struct buffer_head *, unsigned);
1259
1260 /* balloc.c */
1261 extern unsigned int ext4_block_group(struct super_block *sb,
1262                         ext4_fsblk_t blocknr);
1263 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1264                         ext4_fsblk_t blocknr);
1265 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1266 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1267                         ext4_group_t group);
1268 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1269                         ext4_fsblk_t goal, unsigned long *count, int *errp);
1270 extern int ext4_claim_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1271 extern int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1272 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1273                         ext4_fsblk_t block, unsigned long count, int metadata);
1274 extern void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
1275                                 ext4_fsblk_t block, unsigned long count);
1276 extern ext4_fsblk_t ext4_count_free_blocks(struct super_block *);
1277 extern void ext4_check_blocks_bitmap(struct super_block *);
1278 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1279                                                     ext4_group_t block_group,
1280                                                     struct buffer_head ** bh);
1281 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1282 struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1283                                       ext4_group_t block_group);
1284 extern unsigned ext4_init_block_bitmap(struct super_block *sb,
1285                                        struct buffer_head *bh,
1286                                        ext4_group_t group,
1287                                        struct ext4_group_desc *desc);
1288 #define ext4_free_blocks_after_init(sb, group, desc)                    \
1289                 ext4_init_block_bitmap(sb, NULL, group, desc)
1290
1291 /* dir.c */
1292 extern int ext4_check_dir_entry(const char *, struct inode *,
1293                                 struct ext4_dir_entry_2 *,
1294                                 struct buffer_head *, unsigned int);
1295 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1296                                     __u32 minor_hash,
1297                                     struct ext4_dir_entry_2 *dirent);
1298 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1299
1300 /* fsync.c */
1301 extern int ext4_sync_file(struct file *, struct dentry *, int);
1302
1303 /* hash.c */
1304 extern int ext4fs_dirhash(const char *name, int len, struct
1305                           dx_hash_info *hinfo);
1306
1307 /* ialloc.c */
1308 extern struct inode * ext4_new_inode(handle_t *, struct inode *, int);
1309 extern void ext4_free_inode(handle_t *, struct inode *);
1310 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1311 extern unsigned long ext4_count_free_inodes(struct super_block *);
1312 extern unsigned long ext4_count_dirs(struct super_block *);
1313 extern void ext4_check_inodes_bitmap(struct super_block *);
1314 extern unsigned ext4_init_inode_bitmap(struct super_block *sb,
1315                                        struct buffer_head *bh,
1316                                        ext4_group_t group,
1317                                        struct ext4_group_desc *desc);
1318 extern void mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1319
1320 /* mballoc.c */
1321 extern long ext4_mb_stats;
1322 extern long ext4_mb_max_to_scan;
1323 extern int ext4_mb_init(struct super_block *, int);
1324 extern int ext4_mb_release(struct super_block *);
1325 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1326                                 struct ext4_allocation_request *, int *);
1327 extern int ext4_mb_reserve_blocks(struct super_block *, int);
1328 extern void ext4_discard_preallocations(struct inode *);
1329 extern int __init init_ext4_mballoc(void);
1330 extern void exit_ext4_mballoc(void);
1331 extern void ext4_mb_free_blocks(handle_t *, struct inode *,
1332                 unsigned long, unsigned long, int, unsigned long *);
1333 extern int ext4_mb_add_groupinfo(struct super_block *sb,
1334                 ext4_group_t i, struct ext4_group_desc *desc);
1335 extern void ext4_mb_update_group_info(struct ext4_group_info *grp,
1336                 ext4_grpblk_t add);
1337 extern int ext4_mb_get_buddy_cache_lock(struct super_block *, ext4_group_t);
1338 extern void ext4_mb_put_buddy_cache_lock(struct super_block *,
1339                                                 ext4_group_t, int);
1340 /* inode.c */
1341 int ext4_forget(handle_t *handle, int is_metadata, struct inode *inode,
1342                 struct buffer_head *bh, ext4_fsblk_t blocknr);
1343 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1344                                                 ext4_lblk_t, int, int *);
1345 struct buffer_head *ext4_bread(handle_t *, struct inode *,
1346                                                 ext4_lblk_t, int, int *);
1347 int ext4_get_block(struct inode *inode, sector_t iblock,
1348                                 struct buffer_head *bh_result, int create);
1349
1350 extern struct inode *ext4_iget(struct super_block *, unsigned long);
1351 extern int  ext4_write_inode(struct inode *, int);
1352 extern int  ext4_setattr(struct dentry *, struct iattr *);
1353 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1354                                 struct kstat *stat);
1355 extern void ext4_delete_inode(struct inode *);
1356 extern int  ext4_sync_inode(handle_t *, struct inode *);
1357 extern void ext4_dirty_inode(struct inode *);
1358 extern int ext4_change_inode_journal_flag(struct inode *, int);
1359 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1360 extern int ext4_can_truncate(struct inode *inode);
1361 extern void ext4_truncate(struct inode *);
1362 extern void ext4_set_inode_flags(struct inode *);
1363 extern void ext4_get_inode_flags(struct ext4_inode_info *);
1364 extern int ext4_alloc_da_blocks(struct inode *inode);
1365 extern void ext4_set_aops(struct inode *inode);
1366 extern int ext4_writepage_trans_blocks(struct inode *);
1367 extern int ext4_meta_trans_blocks(struct inode *, int nrblocks, int idxblocks);
1368 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1369 extern int ext4_block_truncate_page(handle_t *handle,
1370                 struct address_space *mapping, loff_t from);
1371 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1372 extern qsize_t ext4_get_reserved_space(struct inode *inode);
1373
1374 /* ioctl.c */
1375 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1376 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1377
1378 /* migrate.c */
1379 extern int ext4_ext_migrate(struct inode *);
1380
1381 /* namei.c */
1382 extern unsigned int ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize);
1383 extern __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize);
1384 extern int ext4_orphan_add(handle_t *, struct inode *);
1385 extern int ext4_orphan_del(handle_t *, struct inode *);
1386 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1387                                 __u32 start_minor_hash, __u32 *next_hash);
1388
1389 /* resize.c */
1390 extern int ext4_group_add(struct super_block *sb,
1391                                 struct ext4_new_group_data *input);
1392 extern int ext4_group_extend(struct super_block *sb,
1393                                 struct ext4_super_block *es,
1394                                 ext4_fsblk_t n_blocks_count);
1395
1396 /* super.c */
1397 extern void ext4_error(struct super_block *, const char *, const char *, ...)
1398         __attribute__ ((format (printf, 3, 4)));
1399 extern void __ext4_std_error(struct super_block *, const char *, int);
1400 extern void ext4_abort(struct super_block *, const char *, const char *, ...)
1401         __attribute__ ((format (printf, 3, 4)));
1402 extern void ext4_warning(struct super_block *, const char *, const char *, ...)
1403         __attribute__ ((format (printf, 3, 4)));
1404 extern void ext4_msg(struct super_block *, const char *, const char *, ...)
1405         __attribute__ ((format (printf, 3, 4)));
1406 extern void ext4_grp_locked_error(struct super_block *, ext4_group_t,
1407                                 const char *, const char *, ...)
1408         __attribute__ ((format (printf, 4, 5)));
1409 extern void ext4_update_dynamic_rev(struct super_block *sb);
1410 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1411                                         __u32 compat);
1412 extern int ext4_update_rocompat_feature(handle_t *handle,
1413                                         struct super_block *sb, __u32 rocompat);
1414 extern int ext4_update_incompat_feature(handle_t *handle,
1415                                         struct super_block *sb, __u32 incompat);
1416 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1417                                       struct ext4_group_desc *bg);
1418 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1419                                       struct ext4_group_desc *bg);
1420 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1421                                      struct ext4_group_desc *bg);
1422 extern __u32 ext4_free_blks_count(struct super_block *sb,
1423                                 struct ext4_group_desc *bg);
1424 extern __u32 ext4_free_inodes_count(struct super_block *sb,
1425                                  struct ext4_group_desc *bg);
1426 extern __u32 ext4_used_dirs_count(struct super_block *sb,
1427                                 struct ext4_group_desc *bg);
1428 extern __u32 ext4_itable_unused_count(struct super_block *sb,
1429                                    struct ext4_group_desc *bg);
1430 extern void ext4_block_bitmap_set(struct super_block *sb,
1431                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1432 extern void ext4_inode_bitmap_set(struct super_block *sb,
1433                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1434 extern void ext4_inode_table_set(struct super_block *sb,
1435                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
1436 extern void ext4_free_blks_set(struct super_block *sb,
1437                                struct ext4_group_desc *bg, __u32 count);
1438 extern void ext4_free_inodes_set(struct super_block *sb,
1439                                 struct ext4_group_desc *bg, __u32 count);
1440 extern void ext4_used_dirs_set(struct super_block *sb,
1441                                 struct ext4_group_desc *bg, __u32 count);
1442 extern void ext4_itable_unused_set(struct super_block *sb,
1443                                    struct ext4_group_desc *bg, __u32 count);
1444 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
1445                                    struct ext4_group_desc *gdp);
1446 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
1447                                        struct ext4_group_desc *gdp);
1448
1449 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
1450 {
1451         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
1452                 le32_to_cpu(es->s_blocks_count_lo);
1453 }
1454
1455 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
1456 {
1457         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
1458                 le32_to_cpu(es->s_r_blocks_count_lo);
1459 }
1460
1461 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
1462 {
1463         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
1464                 le32_to_cpu(es->s_free_blocks_count_lo);
1465 }
1466
1467 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
1468                                          ext4_fsblk_t blk)
1469 {
1470         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
1471         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
1472 }
1473
1474 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
1475                                               ext4_fsblk_t blk)
1476 {
1477         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
1478         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
1479 }
1480
1481 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
1482                                            ext4_fsblk_t blk)
1483 {
1484         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
1485         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
1486 }
1487
1488 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
1489 {
1490         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
1491                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
1492                         le32_to_cpu(raw_inode->i_size_lo);
1493         else
1494                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
1495 }
1496
1497 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
1498 {
1499         raw_inode->i_size_lo = cpu_to_le32(i_size);
1500         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
1501 }
1502
1503 static inline
1504 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
1505                                             ext4_group_t group)
1506 {
1507          struct ext4_group_info ***grp_info;
1508          long indexv, indexh;
1509          grp_info = EXT4_SB(sb)->s_group_info;
1510          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
1511          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
1512          return grp_info[indexv][indexh];
1513 }
1514
1515 /*
1516  * Reading s_groups_count requires using smp_rmb() afterwards.  See
1517  * the locking protocol documented in the comments of ext4_group_add()
1518  * in resize.c
1519  */
1520 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
1521 {
1522         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
1523
1524         smp_rmb();
1525         return ngroups;
1526 }
1527
1528 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
1529                                              ext4_group_t block_group)
1530 {
1531         return block_group >> sbi->s_log_groups_per_flex;
1532 }
1533
1534 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
1535 {
1536         return 1 << sbi->s_log_groups_per_flex;
1537 }
1538
1539 #define ext4_std_error(sb, errno)                               \
1540 do {                                                            \
1541         if ((errno))                                            \
1542                 __ext4_std_error((sb), __func__, (errno));      \
1543 } while (0)
1544
1545 #ifdef CONFIG_SMP
1546 /* Each CPU can accumulate percpu_counter_batch blocks in their local
1547  * counters. So we need to make sure we have free blocks more
1548  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
1549  */
1550 #define EXT4_FREEBLOCKS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
1551 #else
1552 #define EXT4_FREEBLOCKS_WATERMARK 0
1553 #endif
1554
1555 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
1556 {
1557         /*
1558          * XXX: replace with spinlock if seen contended -bzzz
1559          */
1560         down_write(&EXT4_I(inode)->i_data_sem);
1561         if (newsize > EXT4_I(inode)->i_disksize)
1562                 EXT4_I(inode)->i_disksize = newsize;
1563         up_write(&EXT4_I(inode)->i_data_sem);
1564         return ;
1565 }
1566
1567 struct ext4_group_info {
1568         unsigned long   bb_state;
1569         struct rb_root  bb_free_root;
1570         unsigned short  bb_first_free;
1571         unsigned short  bb_free;
1572         unsigned short  bb_fragments;
1573         struct          list_head bb_prealloc_list;
1574 #ifdef DOUBLE_CHECK
1575         void            *bb_bitmap;
1576 #endif
1577         struct rw_semaphore alloc_sem;
1578         unsigned short  bb_counters[];
1579 };
1580
1581 #define EXT4_GROUP_INFO_NEED_INIT_BIT   0
1582
1583 #define EXT4_MB_GRP_NEED_INIT(grp)      \
1584         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
1585
1586 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
1587                                               ext4_group_t group)
1588 {
1589         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
1590 }
1591
1592 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
1593 {
1594         spin_lock(ext4_group_lock_ptr(sb, group));
1595 }
1596
1597 static inline void ext4_unlock_group(struct super_block *sb,
1598                                         ext4_group_t group)
1599 {
1600         spin_unlock(ext4_group_lock_ptr(sb, group));
1601 }
1602
1603 /*
1604  * Inodes and files operations
1605  */
1606
1607 /* dir.c */
1608 extern const struct file_operations ext4_dir_operations;
1609
1610 /* file.c */
1611 extern const struct inode_operations ext4_file_inode_operations;
1612 extern const struct file_operations ext4_file_operations;
1613
1614 /* namei.c */
1615 extern const struct inode_operations ext4_dir_inode_operations;
1616 extern const struct inode_operations ext4_special_inode_operations;
1617 extern struct dentry *ext4_get_parent(struct dentry *child);
1618
1619 /* symlink.c */
1620 extern const struct inode_operations ext4_symlink_inode_operations;
1621 extern const struct inode_operations ext4_fast_symlink_inode_operations;
1622
1623 /* block_validity */
1624 extern void ext4_release_system_zone(struct super_block *sb);
1625 extern int ext4_setup_system_zone(struct super_block *sb);
1626 extern int __init init_ext4_system_zone(void);
1627 extern void exit_ext4_system_zone(void);
1628 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
1629                                  ext4_fsblk_t start_blk,
1630                                  unsigned int count);
1631
1632 /* extents.c */
1633 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
1634 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
1635 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
1636                                        int chunk);
1637 extern int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
1638                                ext4_lblk_t iblock, unsigned int max_blocks,
1639                                struct buffer_head *bh_result, int flags);
1640 extern void ext4_ext_truncate(struct inode *);
1641 extern void ext4_ext_init(struct super_block *);
1642 extern void ext4_ext_release(struct super_block *);
1643 extern long ext4_fallocate(struct inode *inode, int mode, loff_t offset,
1644                           loff_t len);
1645 extern int ext4_get_blocks(handle_t *handle, struct inode *inode,
1646                            sector_t block, unsigned int max_blocks,
1647                            struct buffer_head *bh, int flags);
1648 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1649                         __u64 start, __u64 len);
1650
1651 /*
1652  * Add new method to test wether block and inode bitmaps are properly
1653  * initialized. With uninit_bg reading the block from disk is not enough
1654  * to mark the bitmap uptodate. We need to also zero-out the bitmap
1655  */
1656 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
1657
1658 static inline int bitmap_uptodate(struct buffer_head *bh)
1659 {
1660         return (buffer_uptodate(bh) &&
1661                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
1662 }
1663 static inline void set_bitmap_uptodate(struct buffer_head *bh)
1664 {
1665         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
1666 }
1667
1668 #endif  /* __KERNEL__ */
1669
1670 #endif  /* _EXT4_H */