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 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
 356
 357 /*
 358  * ioctl commands in 32 bit emulation
 359  */
 360 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
 361 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
 362 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
 363 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
 364 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
 365 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
 366 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
 367 #ifdef CONFIG_JBD2_DEBUG
 368 #define EXT4_IOC32_WAIT_FOR_READONLY    _IOR('f', 99, int)
 369 #endif
 370 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
 371 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
 372
 373
 374 /*
 375  *  Mount options
 376  */
 377 struct ext4_mount_options {
 378         unsigned long s_mount_opt;
 379         uid_t s_resuid;
 380         gid_t s_resgid;
 381         unsigned long s_commit_interval;
 382         u32 s_min_batch_time, s_max_batch_time;
 383 #ifdef CONFIG_QUOTA
 384         int s_jquota_fmt;
 385         char *s_qf_names[MAXQUOTAS];
 386 #endif
 387 };
 388
 389 /*
 390  * Structure of an inode on the disk
 391  */
 392 struct ext4_inode {
 393         __le16  i_mode;         /* File mode */
 394         __le16  i_uid;          /* Low 16 bits of Owner Uid */
 395         __le32  i_size_lo;      /* Size in bytes */
 396         __le32  i_atime;        /* Access time */
 397         __le32  i_ctime;        /* Inode Change time */
 398         __le32  i_mtime;        /* Modification time */
 399         __le32  i_dtime;        /* Deletion Time */
 400         __le16  i_gid;          /* Low 16 bits of Group Id */
 401         __le16  i_links_count;  /* Links count */
 402         __le32  i_blocks_lo;    /* Blocks count */
 403         __le32  i_flags;        /* File flags */
 404         union {
 405                 struct {
 406                         __le32  l_i_version;
 407                 } linux1;
 408                 struct {
 409                         __u32  h_i_translator;
 410                 } hurd1;
 411                 struct {
 412                         __u32  m_i_reserved1;
 413                 } masix1;
 414         } osd1;                         /* OS dependent 1 */
 415         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
 416         __le32  i_generation;   /* File version (for NFS) */
 417         __le32  i_file_acl_lo;  /* File ACL */
 418         __le32  i_size_high;
 419         __le32  i_obso_faddr;   /* Obsoleted fragment address */
 420         union {
 421                 struct {
 422                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
 423                         __le16  l_i_file_acl_high;
 424                         __le16  l_i_uid_high;   /* these 2 fields */
 425                         __le16  l_i_gid_high;   /* were reserved2[0] */
 426                         __u32   l_i_reserved2;
 427                 } linux2;
 428                 struct {
 429                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 430                         __u16   h_i_mode_high;
 431                         __u16   h_i_uid_high;
 432                         __u16   h_i_gid_high;
 433                         __u32   h_i_author;
 434                 } hurd2;
 435                 struct {
 436                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 437                         __le16  m_i_file_acl_high;
 438                         __u32   m_i_reserved2[2];
 439                 } masix2;
 440         } osd2;                         /* OS dependent 2 */
 441         __le16  i_extra_isize;
 442         __le16  i_pad1;
 443         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
 444         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
 445         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
 446         __le32  i_crtime;       /* File Creation time */
 447         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
 448         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
 449 };
 450
 451 struct move_extent {
 452         __u32 reserved;         /* should be zero */
 453         __u32 donor_fd;         /* donor file descriptor */
 454         __u64 orig_start;       /* logical start offset in block for orig */
 455         __u64 donor_start;      /* logical start offset in block for donor */
 456         __u64 len;              /* block length to be moved */
 457         __u64 moved_len;        /* moved block length */
 458 };
 459 #define MAX_DEFRAG_SIZE         ((1UL<<31) - 1)
 460
 461 #define EXT4_EPOCH_BITS 2
 462 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
 463 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
 464
 465 /*
 466  * Extended fields will fit into an inode if the filesystem was formatted
 467  * with large inodes (-I 256 or larger) and there are not currently any EAs
 468  * consuming all of the available space. For new inodes we always reserve
 469  * enough space for the kernel's known extended fields, but for inodes
 470  * created with an old kernel this might not have been the case. None of
 471  * the extended inode fields is critical for correct filesystem operation.
 472  * This macro checks if a certain field fits in the inode. Note that
 473  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
 474  */
 475 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
 476         ((offsetof(typeof(*ext4_inode), field) +        \
 477           sizeof((ext4_inode)->field))                  \
 478         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
 479             (einode)->i_extra_isize))                   \
 480
 481 static inline __le32 ext4_encode_extra_time(struct timespec *time)
 482 {
 483        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
 484                            time->tv_sec >> 32 : 0) |
 485                            ((time->tv_nsec << 2) & EXT4_NSEC_MASK));
 486 }
 487
 488 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
 489 {
 490        if (sizeof(time->tv_sec) > 4)
 491                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
 492                                << 32;
 493        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> 2;
 494 }
 495
 496 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
 497 do {                                                                           \
 498         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
 499         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 500                 (raw_inode)->xtime ## _extra =                                 \
 501                                 ext4_encode_extra_time(&(inode)->xtime);       \
 502 } while (0)
 503
 504 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
 505 do {                                                                           \
 506         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 507                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
 508         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 509                 (raw_inode)->xtime ## _extra =                                 \
 510                                 ext4_encode_extra_time(&(einode)->xtime);      \
 511 } while (0)
 512
 513 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
 514 do {                                                                           \
 515         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
 516         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 517                 ext4_decode_extra_time(&(inode)->xtime,                        \
 518                                        raw_inode->xtime ## _extra);            \
 519 } while (0)
 520
 521 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
 522 do {                                                                           \
 523         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 524                 (einode)->xtime.tv_sec =                                       \
 525                         (signed)le32_to_cpu((raw_inode)->xtime);               \
 526         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 527                 ext4_decode_extra_time(&(einode)->xtime,                       \
 528                                        raw_inode->xtime ## _extra);            \
 529 } while (0)
 530
 531 #define i_disk_version osd1.linux1.l_i_version
 532
 533 #if defined(__KERNEL__) || defined(__linux__)
 534 #define i_reserved1     osd1.linux1.l_i_reserved1
 535 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
 536 #define i_blocks_high   osd2.linux2.l_i_blocks_high
 537 #define i_uid_low       i_uid
 538 #define i_gid_low       i_gid
 539 #define i_uid_high      osd2.linux2.l_i_uid_high
 540 #define i_gid_high      osd2.linux2.l_i_gid_high
 541 #define i_reserved2     osd2.linux2.l_i_reserved2
 542
 543 #elif defined(__GNU__)
 544
 545 #define i_translator    osd1.hurd1.h_i_translator
 546 #define i_uid_high      osd2.hurd2.h_i_uid_high
 547 #define i_gid_high      osd2.hurd2.h_i_gid_high
 548 #define i_author        osd2.hurd2.h_i_author
 549
 550 #elif defined(__masix__)
 551
 552 #define i_reserved1     osd1.masix1.m_i_reserved1
 553 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
 554 #define i_reserved2     osd2.masix2.m_i_reserved2
 555
 556 #endif /* defined(__KERNEL__) || defined(__linux__) */
 557
 558 /*
 559  * storage for cached extent
 560  */
 561 struct ext4_ext_cache {
 562         ext4_fsblk_t    ec_start;
 563         ext4_lblk_t     ec_block;
 564         __u32           ec_len; /* must be 32bit to return holes */
 565         __u32           ec_type;
 566 };
 567
 568 /*
 569  * fourth extended file system inode data in memory
 570  */
 571 struct ext4_inode_info {
 572         __le32  i_data[15];     /* unconverted */
 573         __u32   i_flags;
 574         ext4_fsblk_t    i_file_acl;
 575         __u32   i_dtime;
 576
 577         /*
 578          * i_block_group is the number of the block group which contains
 579          * this file's inode.  Constant across the lifetime of the inode,
 580          * it is ued for making block allocation decisions - we try to
 581          * place a file's data blocks near its inode block, and new inodes
 582          * near to their parent directory's inode.
 583          */
 584         ext4_group_t    i_block_group;
 585         __u32   i_state;                /* Dynamic state flags for ext4 */
 586
 587         ext4_lblk_t             i_dir_start_lookup;
 588 #ifdef CONFIG_EXT4_FS_XATTR
 589         /*
 590          * Extended attributes can be read independently of the main file
 591          * data. Taking i_mutex even when reading would cause contention
 592          * between readers of EAs and writers of regular file data, so
 593          * instead we synchronize on xattr_sem when reading or changing
 594          * EAs.
 595          */
 596         struct rw_semaphore xattr_sem;
 597 #endif
 598 #ifdef CONFIG_EXT4_FS_POSIX_ACL
 599         struct posix_acl        *i_acl;
 600         struct posix_acl        *i_default_acl;
 601 #endif
 602
 603         struct list_head i_orphan;      /* unlinked but open inodes */
 604
 605         /*
 606          * i_disksize keeps track of what the inode size is ON DISK, not
 607          * in memory.  During truncate, i_size is set to the new size by
 608          * the VFS prior to calling ext4_truncate(), but the filesystem won't
 609          * set i_disksize to 0 until the truncate is actually under way.
 610          *
 611          * The intent is that i_disksize always represents the blocks which
 612          * are used by this file.  This allows recovery to restart truncate
 613          * on orphans if we crash during truncate.  We actually write i_disksize
 614          * into the on-disk inode when writing inodes out, instead of i_size.
 615          *
 616          * The only time when i_disksize and i_size may be different is when
 617          * a truncate is in progress.  The only things which change i_disksize
 618          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
 619          */
 620         loff_t  i_disksize;
 621
 622         /*
 623          * i_data_sem is for serialising ext4_truncate() against
 624          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
 625          * data tree are chopped off during truncate. We can't do that in
 626          * ext4 because whenever we perform intermediate commits during
 627          * truncate, the inode and all the metadata blocks *must* be in a
 628          * consistent state which allows truncation of the orphans to restart
 629          * during recovery.  Hence we must fix the get_block-vs-truncate race
 630          * by other means, so we have i_data_sem.
 631          */
 632         struct rw_semaphore i_data_sem;
 633         struct inode vfs_inode;
 634         struct jbd2_inode jinode;
 635
 636         struct ext4_ext_cache i_cached_extent;
 637         /*
 638          * File creation time. Its function is same as that of
 639          * struct timespec i_{a,c,m}time in the generic inode.
 640          */
 641         struct timespec i_crtime;
 642
 643         /* mballoc */
 644         struct list_head i_prealloc_list;
 645         spinlock_t i_prealloc_lock;
 646
 647         /* ialloc */
 648         ext4_group_t    i_last_alloc_group;
 649
 650         /* allocation reservation info for delalloc */
 651         unsigned int i_reserved_data_blocks;
 652         unsigned int i_reserved_meta_blocks;
 653         unsigned int i_allocated_meta_blocks;
 654         unsigned short i_delalloc_reserved_flag;
 655
 656         /* on-disk additional length */
 657         __u16 i_extra_isize;
 658
 659         spinlock_t i_block_reservation_lock;
 660 };
 661
 662 /*
 663  * File system states
 664  */
 665 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
 666 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
 667 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
 668
 669 /*
 670  * Misc. filesystem flags
 671  */
 672 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
 673 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
 674 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
 675
 676 /*
 677  * Mount flags
 678  */
 679 #define EXT4_MOUNT_OLDALLOC             0x00002  /* Don't use the new Orlov allocator */
 680 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
 681 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
 682 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
 683 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
 684 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
 685 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
 686 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
 687 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
 688 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
 689 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
 690 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
 691 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
 692 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
 693 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
 694 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
 695 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
 696 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
 697 #define EXT4_MOUNT_NOBH                 0x40000 /* No bufferheads */
 698 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
 699 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
 700 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
 701 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
 702 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
 703 #define EXT4_MOUNT_I_VERSION            0x2000000 /* i_version support */
 704 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
 705 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
 706 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
 707
 708 #define clear_opt(o, opt)               o &= ~EXT4_MOUNT_##opt
 709 #define set_opt(o, opt)                 o |= EXT4_MOUNT_##opt
 710 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
 711                                          EXT4_MOUNT_##opt)
 712
 713 #define ext4_set_bit                    ext2_set_bit
 714 #define ext4_set_bit_atomic             ext2_set_bit_atomic
 715 #define ext4_clear_bit                  ext2_clear_bit
 716 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
 717 #define ext4_test_bit                   ext2_test_bit
 718 #define ext4_find_first_zero_bit        ext2_find_first_zero_bit
 719 #define ext4_find_next_zero_bit         ext2_find_next_zero_bit
 720 #define ext4_find_next_bit              ext2_find_next_bit
 721
 722 /*
 723  * Maximal mount counts between two filesystem checks
 724  */
 725 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
 726 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
 727
 728 /*
 729  * Behaviour when detecting errors
 730  */
 731 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
 732 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
 733 #define EXT4_ERRORS_PANIC               3       /* Panic */
 734 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
 735
 736 /*
 737  * Structure of the super block
 738  */
 739 struct ext4_super_block {
 740 /*00*/  __le32  s_inodes_count;         /* Inodes count */
 741         __le32  s_blocks_count_lo;      /* Blocks count */
 742         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
 743         __le32  s_free_blocks_count_lo; /* Free blocks count */
 744 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
 745         __le32  s_first_data_block;     /* First Data Block */
 746         __le32  s_log_block_size;       /* Block size */
 747         __le32  s_obso_log_frag_size;   /* Obsoleted fragment size */
 748 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
 749         __le32  s_obso_frags_per_group; /* Obsoleted fragments per group */
 750         __le32  s_inodes_per_group;     /* # Inodes per group */
 751         __le32  s_mtime;                /* Mount time */
 752 /*30*/  __le32  s_wtime;                /* Write time */
 753         __le16  s_mnt_count;            /* Mount count */
 754         __le16  s_max_mnt_count;        /* Maximal mount count */
 755         __le16  s_magic;                /* Magic signature */
 756         __le16  s_state;                /* File system state */
 757         __le16  s_errors;               /* Behaviour when detecting errors */
 758         __le16  s_minor_rev_level;      /* minor revision level */
 759 /*40*/  __le32  s_lastcheck;            /* time of last check */
 760         __le32  s_checkinterval;        /* max. time between checks */
 761         __le32  s_creator_os;           /* OS */
 762         __le32  s_rev_level;            /* Revision level */
 763 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
 764         __le16  s_def_resgid;           /* Default gid for reserved blocks */
 765         /*
 766          * These fields are for EXT4_DYNAMIC_REV superblocks only.
 767          *
 768          * Note: the difference between the compatible feature set and
 769          * the incompatible feature set is that if there is a bit set
 770          * in the incompatible feature set that the kernel doesn't
 771          * know about, it should refuse to mount the filesystem.
 772          *
 773          * e2fsck's requirements are more strict; if it doesn't know
 774          * about a feature in either the compatible or incompatible
 775          * feature set, it must abort and not try to meddle with
 776          * things it doesn't understand...
 777          */
 778         __le32  s_first_ino;            /* First non-reserved inode */
 779         __le16  s_inode_size;           /* size of inode structure */
 780         __le16  s_block_group_nr;       /* block group # of this superblock */
 781         __le32  s_feature_compat;       /* compatible feature set */
 782 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
 783         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
 784 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
 785 /*78*/  char    s_volume_name[16];      /* volume name */
 786 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
 787 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
 788         /*
 789          * Performance hints.  Directory preallocation should only
 790          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
 791          */
 792         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
 793         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
 794         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
 795         /*
 796          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
 797          */
 798 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
 799 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
 800         __le32  s_journal_dev;          /* device number of journal file */
 801         __le32  s_last_orphan;          /* start of list of inodes to delete */
 802         __le32  s_hash_seed[4];         /* HTREE hash seed */
 803         __u8    s_def_hash_version;     /* Default hash version to use */
 804         __u8    s_reserved_char_pad;
 805         __le16  s_desc_size;            /* size of group descriptor */
 806 /*100*/ __le32  s_default_mount_opts;
 807         __le32  s_first_meta_bg;        /* First metablock block group */
 808         __le32  s_mkfs_time;            /* When the filesystem was created */
 809         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
 810         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
 811 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
 812         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
 813         __le32  s_free_blocks_count_hi; /* Free blocks count */
 814         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
 815         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
 816         __le32  s_flags;                /* Miscellaneous flags */
 817         __le16  s_raid_stride;          /* RAID stride */
 818         __le16  s_mmp_interval;         /* # seconds to wait in MMP checking */
 819         __le64  s_mmp_block;            /* Block for multi-mount protection */
 820         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
 821         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
 822         __u8    s_reserved_char_pad2;
 823         __le16  s_reserved_pad;
 824         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
 825         __u32   s_reserved[160];        /* Padding to the end of the block */
 826 };
 827
 828 #ifdef __KERNEL__
 829
 830 /*
 831  * run-time mount flags
 832  */
 833 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
 834 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
 835
 836 /*
 837  * fourth extended-fs super-block data in memory
 838  */
 839 struct ext4_sb_info {
 840         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
 841         unsigned long s_inodes_per_block;/* Number of inodes per block */
 842         unsigned long s_blocks_per_group;/* Number of blocks in a group */
 843         unsigned long s_inodes_per_group;/* Number of inodes in a group */
 844         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
 845         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
 846         unsigned long s_desc_per_block; /* Number of group descriptors per block */
 847         ext4_group_t s_groups_count;    /* Number of groups in the fs */
 848         unsigned long s_overhead_last;  /* Last calculated overhead */
 849         unsigned long s_blocks_last;    /* Last seen block count */
 850         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
 851         struct buffer_head * s_sbh;     /* Buffer containing the super block */
 852         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
 853         struct buffer_head **s_group_desc;
 854         unsigned int s_mount_opt;
 855         unsigned int s_mount_flags;
 856         ext4_fsblk_t s_sb_block;
 857         uid_t s_resuid;
 858         gid_t s_resgid;
 859         unsigned short s_mount_state;
 860         unsigned short s_pad;
 861         int s_addr_per_block_bits;
 862         int s_desc_per_block_bits;
 863         int s_inode_size;
 864         int s_first_ino;
 865         unsigned int s_inode_readahead_blks;
 866         unsigned int s_inode_goal;
 867         spinlock_t s_next_gen_lock;
 868         u32 s_next_generation;
 869         u32 s_hash_seed[4];
 870         int s_def_hash_version;
 871         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
 872         struct percpu_counter s_freeblocks_counter;
 873         struct percpu_counter s_freeinodes_counter;
 874         struct percpu_counter s_dirs_counter;
 875         struct percpu_counter s_dirtyblocks_counter;
 876         struct blockgroup_lock *s_blockgroup_lock;
 877         struct proc_dir_entry *s_proc;
 878         struct kobject s_kobj;
 879         struct completion s_kobj_unregister;
 880
 881         /* Journaling */
 882         struct inode *s_journal_inode;
 883         struct journal_s *s_journal;
 884         struct list_head s_orphan;
 885         struct mutex s_orphan_lock;
 886         struct mutex s_resize_lock;
 887         unsigned long s_commit_interval;
 888         u32 s_max_batch_time;
 889         u32 s_min_batch_time;
 890         struct block_device *journal_bdev;
 891 #ifdef CONFIG_JBD2_DEBUG
 892         struct timer_list turn_ro_timer;        /* For turning read-only (crash simulation) */
 893         wait_queue_head_t ro_wait_queue;        /* For people waiting for the fs to go read-only */
 894 #endif
 895 #ifdef CONFIG_QUOTA
 896         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
 897         int s_jquota_fmt;                       /* Format of quota to use */
 898 #endif
 899         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
 900         struct rb_root system_blks;
 901
 902 #ifdef EXTENTS_STATS
 903         /* ext4 extents stats */
 904         unsigned long s_ext_min;
 905         unsigned long s_ext_max;
 906         unsigned long s_depth_max;
 907         spinlock_t s_ext_stats_lock;
 908         unsigned long s_ext_blocks;
 909         unsigned long s_ext_extents;
 910 #endif
 911
 912         /* for buddy allocator */
 913         struct ext4_group_info ***s_group_info;
 914         struct inode *s_buddy_cache;
 915         long s_blocks_reserved;
 916         spinlock_t s_reserve_lock;
 917         spinlock_t s_md_lock;
 918         tid_t s_last_transaction;
 919         unsigned short *s_mb_offsets;
 920         unsigned int *s_mb_maxs;
 921
 922         /* tunables */
 923         unsigned long s_stripe;
 924         unsigned int s_mb_stream_request;
 925         unsigned int s_mb_max_to_scan;
 926         unsigned int s_mb_min_to_scan;
 927         unsigned int s_mb_stats;
 928         unsigned int s_mb_order2_reqs;
 929         unsigned int s_mb_group_prealloc;
 930         /* where last allocation was done - for stream allocation */
 931         unsigned long s_mb_last_group;
 932         unsigned long s_mb_last_start;
 933
 934         /* history to debug policy */
 935         struct ext4_mb_history *s_mb_history;
 936         int s_mb_history_cur;
 937         int s_mb_history_max;
 938         int s_mb_history_num;
 939         spinlock_t s_mb_history_lock;
 940         int s_mb_history_filter;
 941
 942         /* stats for buddy allocator */
 943         spinlock_t s_mb_pa_lock;
 944         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
 945         atomic_t s_bal_success; /* we found long enough chunks */
 946         atomic_t s_bal_allocated;       /* in blocks */
 947         atomic_t s_bal_ex_scanned;      /* total extents scanned */
 948         atomic_t s_bal_goals;   /* goal hits */
 949         atomic_t s_bal_breaks;  /* too long searches */
 950         atomic_t s_bal_2orders; /* 2^order hits */
 951         spinlock_t s_bal_lock;
 952         unsigned long s_mb_buddies_generated;
 953         unsigned long long s_mb_generation_time;
 954         atomic_t s_mb_lost_chunks;
 955         atomic_t s_mb_preallocated;
 956         atomic_t s_mb_discarded;
 957
 958         /* locality groups */
 959         struct ext4_locality_group *s_locality_groups;
 960
 961         /* for write statistics */
 962         unsigned long s_sectors_written_start;
 963         u64 s_kbytes_written;
 964
 965         unsigned int s_log_groups_per_flex;
 966         struct flex_groups *s_flex_groups;
 967 };
 968
 969 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
 970 {
 971         return sb->s_fs_info;
 972 }
 973 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
 974 {
 975         return container_of(inode, struct ext4_inode_info, vfs_inode);
 976 }
 977
 978 static inline struct timespec ext4_current_time(struct inode *inode)
 979 {
 980         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
 981                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
 982 }
 983
 984 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
 985 {
 986         return ino == EXT4_ROOT_INO ||
 987                 ino == EXT4_JOURNAL_INO ||
 988                 ino == EXT4_RESIZE_INO ||
 989                 (ino >= EXT4_FIRST_INO(sb) &&
 990                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
 991 }
 992 #else
 993 /* Assume that user mode programs are passing in an ext4fs superblock, not
 994  * a kernel struct super_block.  This will allow us to call the feature-test
 995  * macros from user land. */
 996 #define EXT4_SB(sb)     (sb)
 997 #endif
 998
 999 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1000
1001 /*
1002  * Codes for operating systems
1003  */
1004 #define EXT4_OS_LINUX           0
1005 #define EXT4_OS_HURD            1
1006 #define EXT4_OS_MASIX           2
1007 #define EXT4_OS_FREEBSD         3
1008 #define EXT4_OS_LITES           4
1009
1010 /*
1011  * Revision levels
1012  */
1013 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1014 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1015
1016 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1017 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1018
1019 #define EXT4_GOOD_OLD_INODE_SIZE 128
1020
1021 /*
1022  * Feature set definitions
1023  */
1024
1025 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1026         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1027 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1028         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1029 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1030         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1031 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1032         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1033 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1034         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1035 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1036         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1037 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1038         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1039 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1040         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1041 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1042         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1043
1044 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1045 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1046 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1047 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1048 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1049 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1050
1051 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1052 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1053 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1054 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1055 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1056 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1057 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1058
1059 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1060 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1061 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1062 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1063 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1064 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1065 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1066 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1067 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1068
1069 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1070 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1071                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1072                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1073                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1074                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1075                                          EXT4_FEATURE_INCOMPAT_FLEX_BG)
1076 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1077                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1078                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1079                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1080                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1081                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1082                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE)
1083
1084 /*
1085  * Default values for user and/or group using reserved blocks
1086  */
1087 #define EXT4_DEF_RESUID         0
1088 #define EXT4_DEF_RESGID         0
1089
1090 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1091
1092 /*
1093  * Default mount options
1094  */
1095 #define EXT4_DEFM_DEBUG         0x0001
1096 #define EXT4_DEFM_BSDGROUPS     0x0002
1097 #define EXT4_DEFM_XATTR_USER    0x0004
1098 #define EXT4_DEFM_ACL           0x0008
1099 #define EXT4_DEFM_UID16         0x0010
1100 #define EXT4_DEFM_JMODE         0x0060
1101 #define EXT4_DEFM_JMODE_DATA    0x0020
1102 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1103 #define EXT4_DEFM_JMODE_WBACK   0x0060
1104
1105 /*
1106  * Default journal batch times
1107  */
1108 #define EXT4_DEF_MIN_BATCH_TIME 0
1109 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1110
1111 /*
1112  * Minimum number of groups in a flexgroup before we separate out
1113  * directories into the first block group of a flexgroup
1114  */
1115 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1116
1117 /*
1118  * Structure of a directory entry
1119  */
1120 #define EXT4_NAME_LEN 255
1121
1122 struct ext4_dir_entry {
1123         __le32  inode;                  /* Inode number */
1124         __le16  rec_len;                /* Directory entry length */
1125         __le16  name_len;               /* Name length */
1126         char    name[EXT4_NAME_LEN];    /* File name */
1127 };
1128
1129 /*
1130  * The new version of the directory entry.  Since EXT4 structures are
1131  * stored in intel byte order, and the name_len field could never be
1132  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1133  * file_type field.
1134  */
1135 struct ext4_dir_entry_2 {
1136         __le32  inode;                  /* Inode number */
1137         __le16  rec_len;                /* Directory entry length */
1138         __u8    name_len;               /* Name length */
1139         __u8    file_type;
1140         char    name[EXT4_NAME_LEN];    /* File name */
1141 };
1142
1143 /*
1144  * Ext4 directory file types.  Only the low 3 bits are used.  The
1145  * other bits are reserved for now.
1146  */
1147 #define EXT4_FT_UNKNOWN         0
1148 #define EXT4_FT_REG_FILE        1
1149 #define EXT4_FT_DIR             2
1150 #define EXT4_FT_CHRDEV          3
1151 #define EXT4_FT_BLKDEV          4
1152 #define EXT4_FT_FIFO            5
1153 #define EXT4_FT_SOCK            6
1154 #define EXT4_FT_SYMLINK         7
1155
1156 #define EXT4_FT_MAX             8
1157
1158 /*
1159  * EXT4_DIR_PAD defines the directory entries boundaries
1160  *
1161  * NOTE: It must be a multiple of 4
1162  */
1163 #define EXT4_DIR_PAD                    4
1164 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1165 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1166                                          ~EXT4_DIR_ROUND)
1167 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1168
1169 /*
1170  * Hash Tree Directory indexing
1171  * (c) Daniel Phillips, 2001
1172  */
1173
1174 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1175                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1176                       (EXT4_I(dir)->i_flags & EXT4_INDEX_FL))
1177 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1178 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1179
1180 /* Legal values for the dx_root hash_version field: */
1181
1182 #define DX_HASH_LEGACY          0
1183 #define DX_HASH_HALF_MD4        1
1184 #define DX_HASH_TEA             2
1185 #define DX_HASH_LEGACY_UNSIGNED 3
1186 #define DX_HASH_HALF_MD4_UNSIGNED       4
1187 #define DX_HASH_TEA_UNSIGNED            5
1188
1189 #ifdef __KERNEL__
1190
1191 /* hash info structure used by the directory hash */
1192 struct dx_hash_info
1193 {
1194         u32             hash;
1195         u32             minor_hash;
1196         int             hash_version;
1197         u32             *seed;
1198 };
1199
1200 #define EXT4_HTREE_EOF  0x7fffffff
1201
1202 /*
1203  * Control parameters used by ext4_htree_next_block
1204  */
1205 #define HASH_NB_ALWAYS          1
1206
1207
1208 /*
1209  * Describe an inode's exact location on disk and in memory
1210  */
1211 struct ext4_iloc
1212 {
1213         struct buffer_head *bh;
1214         unsigned long offset;
1215         ext4_group_t block_group;
1216 };
1217
1218 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1219 {
1220         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1221 }
1222
1223 /*
1224  * This structure is stuffed into the struct file's private_data field
1225  * for directories.  It is where we put information so that we can do
1226  * readdir operations in hash tree order.
1227  */
1228 struct dir_private_info {
1229         struct rb_root  root;
1230         struct rb_node  *curr_node;
1231         struct fname    *extra_fname;
1232         loff_t          last_pos;
1233         __u32           curr_hash;
1234         __u32           curr_minor_hash;
1235         __u32           next_hash;
1236 };
1237
1238 /* calculate the first block number of the group */
1239 static inline ext4_fsblk_t
1240 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1241 {
1242         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1243                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1244 }
1245
1246 /*
1247  * Special error return code only used by dx_probe() and its callers.
1248  */
1249 #define ERR_BAD_DX_DIR  -75000
1250
1251 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1252                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1253
1254 extern struct proc_dir_entry *ext4_proc_root;
1255
1256 /*
1257  * Function prototypes
1258  */
1259
1260 /*
1261  * Ok, these declarations are also in <linux/kernel.h> but none of the
1262  * ext4 source programs needs to include it so they are duplicated here.
1263  */
1264 # define NORET_TYPE     /**/
1265 # define ATTRIB_NORET   __attribute__((noreturn))
1266 # define NORET_AND      noreturn,
1267
1268 /* bitmap.c */
1269 extern unsigned int ext4_count_free(struct buffer_head *, unsigned);
1270
1271 /* balloc.c */
1272 extern unsigned int ext4_block_group(struct super_block *sb,
1273                         ext4_fsblk_t blocknr);
1274 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1275                         ext4_fsblk_t blocknr);
1276 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1277 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1278                         ext4_group_t group);
1279 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1280                         ext4_fsblk_t goal, unsigned long *count, int *errp);
1281 extern int ext4_claim_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1282 extern int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1283 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1284                         ext4_fsblk_t block, unsigned long count, int metadata);
1285 extern void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
1286                                 ext4_fsblk_t block, unsigned long count);
1287 extern ext4_fsblk_t ext4_count_free_blocks(struct super_block *);
1288 extern void ext4_check_blocks_bitmap(struct super_block *);
1289 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1290                                                     ext4_group_t block_group,
1291                                                     struct buffer_head ** bh);
1292 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1293 struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1294                                       ext4_group_t block_group);
1295 extern unsigned ext4_init_block_bitmap(struct super_block *sb,
1296                                        struct buffer_head *bh,
1297                                        ext4_group_t group,
1298                                        struct ext4_group_desc *desc);
1299 #define ext4_free_blocks_after_init(sb, group, desc)                    \
1300                 ext4_init_block_bitmap(sb, NULL, group, desc)
1301
1302 /* dir.c */
1303 extern int ext4_check_dir_entry(const char *, struct inode *,
1304                                 struct ext4_dir_entry_2 *,
1305                                 struct buffer_head *, unsigned int);
1306 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1307                                     __u32 minor_hash,
1308                                     struct ext4_dir_entry_2 *dirent);
1309 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1310
1311 /* fsync.c */
1312 extern int ext4_sync_file(struct file *, struct dentry *, int);
1313
1314 /* hash.c */
1315 extern int ext4fs_dirhash(const char *name, int len, struct
1316                           dx_hash_info *hinfo);
1317
1318 /* ialloc.c */
1319 extern struct inode *ext4_new_inode(handle_t *, struct inode *, int,
1320                                     const struct qstr *qstr, __u32 goal);
1321 extern void ext4_free_inode(handle_t *, struct inode *);
1322 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1323 extern unsigned long ext4_count_free_inodes(struct super_block *);
1324 extern unsigned long ext4_count_dirs(struct super_block *);
1325 extern void ext4_check_inodes_bitmap(struct super_block *);
1326 extern unsigned ext4_init_inode_bitmap(struct super_block *sb,
1327                                        struct buffer_head *bh,
1328                                        ext4_group_t group,
1329                                        struct ext4_group_desc *desc);
1330 extern void mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1331
1332 /* mballoc.c */
1333 extern long ext4_mb_stats;
1334 extern long ext4_mb_max_to_scan;
1335 extern int ext4_mb_init(struct super_block *, int);
1336 extern int ext4_mb_release(struct super_block *);
1337 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1338                                 struct ext4_allocation_request *, int *);
1339 extern int ext4_mb_reserve_blocks(struct super_block *, int);
1340 extern void ext4_discard_preallocations(struct inode *);
1341 extern int __init init_ext4_mballoc(void);
1342 extern void exit_ext4_mballoc(void);
1343 extern void ext4_mb_free_blocks(handle_t *, struct inode *,
1344                 ext4_fsblk_t, unsigned long, int, unsigned long *);
1345 extern int ext4_mb_add_groupinfo(struct super_block *sb,
1346                 ext4_group_t i, struct ext4_group_desc *desc);
1347 extern void ext4_mb_update_group_info(struct ext4_group_info *grp,
1348                 ext4_grpblk_t add);
1349 extern int ext4_mb_get_buddy_cache_lock(struct super_block *, ext4_group_t);
1350 extern void ext4_mb_put_buddy_cache_lock(struct super_block *,
1351                                                 ext4_group_t, int);
1352 /* inode.c */
1353 int ext4_forget(handle_t *handle, int is_metadata, struct inode *inode,
1354                 struct buffer_head *bh, ext4_fsblk_t blocknr);
1355 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1356                                                 ext4_lblk_t, int, int *);
1357 struct buffer_head *ext4_bread(handle_t *, struct inode *,
1358                                                 ext4_lblk_t, int, int *);
1359 int ext4_get_block(struct inode *inode, sector_t iblock,
1360                                 struct buffer_head *bh_result, int create);
1361
1362 extern struct inode *ext4_iget(struct super_block *, unsigned long);
1363 extern int  ext4_write_inode(struct inode *, int);
1364 extern int  ext4_setattr(struct dentry *, struct iattr *);
1365 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1366                                 struct kstat *stat);
1367 extern void ext4_delete_inode(struct inode *);
1368 extern int  ext4_sync_inode(handle_t *, struct inode *);
1369 extern void ext4_dirty_inode(struct inode *);
1370 extern int ext4_change_inode_journal_flag(struct inode *, int);
1371 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1372 extern int ext4_can_truncate(struct inode *inode);
1373 extern void ext4_truncate(struct inode *);
1374 extern void ext4_set_inode_flags(struct inode *);
1375 extern void ext4_get_inode_flags(struct ext4_inode_info *);
1376 extern int ext4_alloc_da_blocks(struct inode *inode);
1377 extern void ext4_set_aops(struct inode *inode);
1378 extern int ext4_writepage_trans_blocks(struct inode *);
1379 extern int ext4_meta_trans_blocks(struct inode *, int nrblocks, int idxblocks);
1380 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1381 extern int ext4_block_truncate_page(handle_t *handle,
1382                 struct address_space *mapping, loff_t from);
1383 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1384 extern qsize_t ext4_get_reserved_space(struct inode *inode);
1385
1386 /* ioctl.c */
1387 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1388 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1389
1390 /* migrate.c */
1391 extern int ext4_ext_migrate(struct inode *);
1392
1393 /* namei.c */
1394 extern unsigned int ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize);
1395 extern __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize);
1396 extern int ext4_orphan_add(handle_t *, struct inode *);
1397 extern int ext4_orphan_del(handle_t *, struct inode *);
1398 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1399                                 __u32 start_minor_hash, __u32 *next_hash);
1400
1401 /* resize.c */
1402 extern int ext4_group_add(struct super_block *sb,
1403                                 struct ext4_new_group_data *input);
1404 extern int ext4_group_extend(struct super_block *sb,
1405                                 struct ext4_super_block *es,
1406                                 ext4_fsblk_t n_blocks_count);
1407
1408 /* super.c */
1409 extern void ext4_error(struct super_block *, const char *, const char *, ...)
1410         __attribute__ ((format (printf, 3, 4)));
1411 extern void __ext4_std_error(struct super_block *, const char *, int);
1412 extern void ext4_abort(struct super_block *, const char *, const char *, ...)
1413         __attribute__ ((format (printf, 3, 4)));
1414 extern void ext4_warning(struct super_block *, const char *, const char *, ...)
1415         __attribute__ ((format (printf, 3, 4)));
1416 extern void ext4_msg(struct super_block *, const char *, const char *, ...)
1417         __attribute__ ((format (printf, 3, 4)));
1418 extern void ext4_grp_locked_error(struct super_block *, ext4_group_t,
1419                                 const char *, const char *, ...)
1420         __attribute__ ((format (printf, 4, 5)));
1421 extern void ext4_update_dynamic_rev(struct super_block *sb);
1422 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1423                                         __u32 compat);
1424 extern int ext4_update_rocompat_feature(handle_t *handle,
1425                                         struct super_block *sb, __u32 rocompat);
1426 extern int ext4_update_incompat_feature(handle_t *handle,
1427                                         struct super_block *sb, __u32 incompat);
1428 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1429                                       struct ext4_group_desc *bg);
1430 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1431                                       struct ext4_group_desc *bg);
1432 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1433                                      struct ext4_group_desc *bg);
1434 extern __u32 ext4_free_blks_count(struct super_block *sb,
1435                                 struct ext4_group_desc *bg);
1436 extern __u32 ext4_free_inodes_count(struct super_block *sb,
1437                                  struct ext4_group_desc *bg);
1438 extern __u32 ext4_used_dirs_count(struct super_block *sb,
1439                                 struct ext4_group_desc *bg);
1440 extern __u32 ext4_itable_unused_count(struct super_block *sb,
1441                                    struct ext4_group_desc *bg);
1442 extern void ext4_block_bitmap_set(struct super_block *sb,
1443                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1444 extern void ext4_inode_bitmap_set(struct super_block *sb,
1445                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1446 extern void ext4_inode_table_set(struct super_block *sb,
1447                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
1448 extern void ext4_free_blks_set(struct super_block *sb,
1449                                struct ext4_group_desc *bg, __u32 count);
1450 extern void ext4_free_inodes_set(struct super_block *sb,
1451                                 struct ext4_group_desc *bg, __u32 count);
1452 extern void ext4_used_dirs_set(struct super_block *sb,
1453                                 struct ext4_group_desc *bg, __u32 count);
1454 extern void ext4_itable_unused_set(struct super_block *sb,
1455                                    struct ext4_group_desc *bg, __u32 count);
1456 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
1457                                    struct ext4_group_desc *gdp);
1458 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
1459                                        struct ext4_group_desc *gdp);
1460
1461 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
1462 {
1463         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
1464                 le32_to_cpu(es->s_blocks_count_lo);
1465 }
1466
1467 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
1468 {
1469         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
1470                 le32_to_cpu(es->s_r_blocks_count_lo);
1471 }
1472
1473 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
1474 {
1475         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
1476                 le32_to_cpu(es->s_free_blocks_count_lo);
1477 }
1478
1479 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
1480                                          ext4_fsblk_t blk)
1481 {
1482         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
1483         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
1484 }
1485
1486 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
1487                                               ext4_fsblk_t blk)
1488 {
1489         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
1490         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
1491 }
1492
1493 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
1494                                            ext4_fsblk_t blk)
1495 {
1496         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
1497         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
1498 }
1499
1500 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
1501 {
1502         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
1503                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
1504                         le32_to_cpu(raw_inode->i_size_lo);
1505         else
1506                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
1507 }
1508
1509 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
1510 {
1511         raw_inode->i_size_lo = cpu_to_le32(i_size);
1512         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
1513 }
1514
1515 static inline
1516 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
1517                                             ext4_group_t group)
1518 {
1519          struct ext4_group_info ***grp_info;
1520          long indexv, indexh;
1521          grp_info = EXT4_SB(sb)->s_group_info;
1522          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
1523          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
1524          return grp_info[indexv][indexh];
1525 }
1526
1527 /*
1528  * Reading s_groups_count requires using smp_rmb() afterwards.  See
1529  * the locking protocol documented in the comments of ext4_group_add()
1530  * in resize.c
1531  */
1532 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
1533 {
1534         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
1535
1536         smp_rmb();
1537         return ngroups;
1538 }
1539
1540 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
1541                                              ext4_group_t block_group)
1542 {
1543         return block_group >> sbi->s_log_groups_per_flex;
1544 }
1545
1546 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
1547 {
1548         return 1 << sbi->s_log_groups_per_flex;
1549 }
1550
1551 #define ext4_std_error(sb, errno)                               \
1552 do {                                                            \
1553         if ((errno))                                            \
1554                 __ext4_std_error((sb), __func__, (errno));      \
1555 } while (0)
1556
1557 #ifdef CONFIG_SMP
1558 /* Each CPU can accumulate percpu_counter_batch blocks in their local
1559  * counters. So we need to make sure we have free blocks more
1560  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
1561  */
1562 #define EXT4_FREEBLOCKS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
1563 #else
1564 #define EXT4_FREEBLOCKS_WATERMARK 0
1565 #endif
1566
1567 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
1568 {
1569         /*
1570          * XXX: replace with spinlock if seen contended -bzzz
1571          */
1572         down_write(&EXT4_I(inode)->i_data_sem);
1573         if (newsize > EXT4_I(inode)->i_disksize)
1574                 EXT4_I(inode)->i_disksize = newsize;
1575         up_write(&EXT4_I(inode)->i_data_sem);
1576         return ;
1577 }
1578
1579 struct ext4_group_info {
1580         unsigned long   bb_state;
1581         struct rb_root  bb_free_root;
1582         unsigned short  bb_first_free;
1583         unsigned short  bb_free;
1584         unsigned short  bb_fragments;
1585         struct          list_head bb_prealloc_list;
1586 #ifdef DOUBLE_CHECK
1587         void            *bb_bitmap;
1588 #endif
1589         struct rw_semaphore alloc_sem;
1590         unsigned short  bb_counters[];
1591 };
1592
1593 #define EXT4_GROUP_INFO_NEED_INIT_BIT   0
1594
1595 #define EXT4_MB_GRP_NEED_INIT(grp)      \
1596         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
1597
1598 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
1599                                               ext4_group_t group)
1600 {
1601         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
1602 }
1603
1604 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
1605 {
1606         spin_lock(ext4_group_lock_ptr(sb, group));
1607 }
1608
1609 static inline void ext4_unlock_group(struct super_block *sb,
1610                                         ext4_group_t group)
1611 {
1612         spin_unlock(ext4_group_lock_ptr(sb, group));
1613 }
1614
1615 /*
1616  * Inodes and files operations
1617  */
1618
1619 /* dir.c */
1620 extern const struct file_operations ext4_dir_operations;
1621
1622 /* file.c */
1623 extern const struct inode_operations ext4_file_inode_operations;
1624 extern const struct file_operations ext4_file_operations;
1625
1626 /* namei.c */
1627 extern const struct inode_operations ext4_dir_inode_operations;
1628 extern const struct inode_operations ext4_special_inode_operations;
1629 extern struct dentry *ext4_get_parent(struct dentry *child);
1630
1631 /* symlink.c */
1632 extern const struct inode_operations ext4_symlink_inode_operations;
1633 extern const struct inode_operations ext4_fast_symlink_inode_operations;
1634
1635 /* block_validity */
1636 extern void ext4_release_system_zone(struct super_block *sb);
1637 extern int ext4_setup_system_zone(struct super_block *sb);
1638 extern int __init init_ext4_system_zone(void);
1639 extern void exit_ext4_system_zone(void);
1640 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
1641                                  ext4_fsblk_t start_blk,
1642                                  unsigned int count);
1643
1644 /* extents.c */
1645 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
1646 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
1647 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
1648                                        int chunk);
1649 extern int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
1650                                ext4_lblk_t iblock, unsigned int max_blocks,
1651                                struct buffer_head *bh_result, int flags);
1652 extern void ext4_ext_truncate(struct inode *);
1653 extern void ext4_ext_init(struct super_block *);
1654 extern void ext4_ext_release(struct super_block *);
1655 extern long ext4_fallocate(struct inode *inode, int mode, loff_t offset,
1656                           loff_t len);
1657 extern int ext4_get_blocks(handle_t *handle, struct inode *inode,
1658                            sector_t block, unsigned int max_blocks,
1659                            struct buffer_head *bh, int flags);
1660 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1661                         __u64 start, __u64 len);
1662 /* move_extent.c */
1663 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
1664                              __u64 start_orig, __u64 start_donor,
1665                              __u64 len, __u64 *moved_len);
1666
1667
1668 /*
1669  * Add new method to test wether block and inode bitmaps are properly
1670  * initialized. With uninit_bg reading the block from disk is not enough
1671  * to mark the bitmap uptodate. We need to also zero-out the bitmap
1672  */
1673 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
1674
1675 static inline int bitmap_uptodate(struct buffer_head *bh)
1676 {
1677         return (buffer_uptodate(bh) &&
1678                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
1679 }
1680 static inline void set_bitmap_uptodate(struct buffer_head *bh)
1681 {
1682         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
1683 }
1684
1685 #endif  /* __KERNEL__ */
1686
1687 #endif  /* _EXT4_H */