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  * Flags used in mballoc's allocation_context flags field.
  70  *
  71  * Also used to show what's going on for debugging purposes when the
  72  * flag field is exported via the traceport interface
  73  */
  74
  75 /* prefer goal again. length */
  76 #define EXT4_MB_HINT_MERGE              0x0001
  77 /* blocks already reserved */
  78 #define EXT4_MB_HINT_RESERVED           0x0002
  79 /* metadata is being allocated */
  80 #define EXT4_MB_HINT_METADATA           0x0004
  81 /* first blocks in the file */
  82 #define EXT4_MB_HINT_FIRST              0x0008
  83 /* search for the best chunk */
  84 #define EXT4_MB_HINT_BEST               0x0010
  85 /* data is being allocated */
  86 #define EXT4_MB_HINT_DATA               0x0020
  87 /* don't preallocate (for tails) */
  88 #define EXT4_MB_HINT_NOPREALLOC         0x0040
  89 /* allocate for locality group */
  90 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
  91 /* allocate goal blocks or none */
  92 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
  93 /* goal is meaningful */
  94 #define EXT4_MB_HINT_TRY_GOAL           0x0200
  95 /* blocks already pre-reserved by delayed allocation */
  96 #define EXT4_MB_DELALLOC_RESERVED       0x0400
  97 /* We are doing stream allocation */
  98 #define EXT4_MB_STREAM_ALLOC            0x0800
  99
 100
 101 struct ext4_allocation_request {
 102         /* target inode for block we're allocating */
 103         struct inode *inode;
 104         /* how many blocks we want to allocate */
 105         unsigned int len;
 106         /* logical block in target inode */
 107         ext4_lblk_t logical;
 108         /* the closest logical allocated block to the left */
 109         ext4_lblk_t lleft;
 110         /* the closest logical allocated block to the right */
 111         ext4_lblk_t lright;
 112         /* phys. target (a hint) */
 113         ext4_fsblk_t goal;
 114         /* phys. block for the closest logical allocated block to the left */
 115         ext4_fsblk_t pleft;
 116         /* phys. block for the closest logical allocated block to the right */
 117         ext4_fsblk_t pright;
 118         /* flags. see above EXT4_MB_HINT_* */
 119         unsigned int flags;
 120 };
 121
 122 /*
 123  * For delayed allocation tracking
 124  */
 125 struct mpage_da_data {
 126         struct inode *inode;
 127         sector_t b_blocknr;             /* start block number of extent */
 128         size_t b_size;                  /* size of extent */
 129         unsigned long b_state;          /* state of the extent */
 130         unsigned long first_page, next_page;    /* extent of pages */
 131         struct writeback_control *wbc;
 132         int io_done;
 133         int pages_written;
 134         int retval;
 135 };
 136 #define DIO_AIO_UNWRITTEN       0x1
 137 typedef struct ext4_io_end {
 138         struct list_head        list;           /* per-file finished AIO list */
 139         struct inode            *inode;         /* file being written to */
 140         unsigned int            flag;           /* unwritten or not */
 141         int                     error;          /* I/O error code */
 142         loff_t                  offset;         /* offset in the file */
 143         ssize_t                 size;           /* size of the extent */
 144         struct work_struct      work;           /* data work queue */
 145 } ext4_io_end_t;
 146
 147 /*
 148  * Special inodes numbers
 149  */
 150 #define EXT4_BAD_INO             1      /* Bad blocks inode */
 151 #define EXT4_ROOT_INO            2      /* Root inode */
 152 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
 153 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
 154 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
 155 #define EXT4_JOURNAL_INO         8      /* Journal inode */
 156
 157 /* First non-reserved inode for old ext4 filesystems */
 158 #define EXT4_GOOD_OLD_FIRST_INO 11
 159
 160 /*
 161  * Maximal count of links to a file
 162  */
 163 #define EXT4_LINK_MAX           65000
 164
 165 /*
 166  * Macro-instructions used to manage several block sizes
 167  */
 168 #define EXT4_MIN_BLOCK_SIZE             1024
 169 #define EXT4_MAX_BLOCK_SIZE             65536
 170 #define EXT4_MIN_BLOCK_LOG_SIZE         10
 171 #ifdef __KERNEL__
 172 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
 173 #else
 174 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
 175 #endif
 176 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
 177 #ifdef __KERNEL__
 178 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
 179 #else
 180 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
 181 #endif
 182 #ifdef __KERNEL__
 183 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
 184 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
 185 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
 186 #else
 187 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 188                                  EXT4_GOOD_OLD_INODE_SIZE : \
 189                                  (s)->s_inode_size)
 190 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 191                                  EXT4_GOOD_OLD_FIRST_INO : \
 192                                  (s)->s_first_ino)
 193 #endif
 194 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
 195
 196 /*
 197  * Structure of a blocks group descriptor
 198  */
 199 struct ext4_group_desc
 200 {
 201         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
 202         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
 203         __le32  bg_inode_table_lo;      /* Inodes table block */
 204         __le16  bg_free_blocks_count_lo;/* Free blocks count */
 205         __le16  bg_free_inodes_count_lo;/* Free inodes count */
 206         __le16  bg_used_dirs_count_lo;  /* Directories count */
 207         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
 208         __u32   bg_reserved[2];         /* Likely block/inode bitmap checksum */
 209         __le16  bg_itable_unused_lo;    /* Unused inodes count */
 210         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
 211         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
 212         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
 213         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
 214         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
 215         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
 216         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
 217         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
 218         __u32   bg_reserved2[3];
 219 };
 220
 221 /*
 222  * Structure of a flex block group info
 223  */
 224
 225 struct flex_groups {
 226         atomic_t free_inodes;
 227         atomic_t free_blocks;
 228         atomic_t used_dirs;
 229 };
 230
 231 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
 232 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
 233 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
 234
 235 /*
 236  * Macro-instructions used to manage group descriptors
 237  */
 238 #define EXT4_MIN_DESC_SIZE              32
 239 #define EXT4_MIN_DESC_SIZE_64BIT        64
 240 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
 241 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
 242 #ifdef __KERNEL__
 243 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
 244 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
 245 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
 246 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
 247 #else
 248 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
 249 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
 250 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
 251 #endif
 252
 253 /*
 254  * Constants relative to the data blocks
 255  */
 256 #define EXT4_NDIR_BLOCKS                12
 257 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
 258 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
 259 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
 260 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
 261
 262 /*
 263  * Inode flags
 264  */
 265 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
 266 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
 267 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
 268 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
 269 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
 270 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
 271 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
 272 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
 273 /* Reserved for compression usage... */
 274 #define EXT4_DIRTY_FL                   0x00000100
 275 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
 276 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
 277 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
 278 /* End compression flags --- maybe not all used */
 279 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
 280 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
 281 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
 282 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
 283 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
 284 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
 285 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
 286 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
 287 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
 288
 289 #define EXT4_FL_USER_VISIBLE            0x000BDFFF /* User visible flags */
 290 #define EXT4_FL_USER_MODIFIABLE         0x000B80FF /* User modifiable flags */
 291
 292 /* Flags that should be inherited by new inodes from their parent. */
 293 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
 294                            EXT4_SYNC_FL | EXT4_IMMUTABLE_FL | EXT4_APPEND_FL |\
 295                            EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
 296                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
 297                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
 298
 299 /* Flags that are appropriate for regular files (all but dir-specific ones). */
 300 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
 301
 302 /* Flags that are appropriate for non-directories/regular files. */
 303 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
 304
 305 /* Mask out flags that are inappropriate for the given type of inode. */
 306 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
 307 {
 308         if (S_ISDIR(mode))
 309                 return flags;
 310         else if (S_ISREG(mode))
 311                 return flags & EXT4_REG_FLMASK;
 312         else
 313                 return flags & EXT4_OTHER_FLMASK;
 314 }
 315
 316 /* Used to pass group descriptor data when online resize is done */
 317 struct ext4_new_group_input {
 318         __u32 group;            /* Group number for this data */
 319         __u64 block_bitmap;     /* Absolute block number of block bitmap */
 320         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
 321         __u64 inode_table;      /* Absolute block number of inode table start */
 322         __u32 blocks_count;     /* Total number of blocks in this group */
 323         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
 324         __u16 unused;
 325 };
 326
 327 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
 328 struct ext4_new_group_data {
 329         __u32 group;
 330         __u64 block_bitmap;
 331         __u64 inode_bitmap;
 332         __u64 inode_table;
 333         __u32 blocks_count;
 334         __u16 reserved_blocks;
 335         __u16 unused;
 336         __u32 free_blocks_count;
 337 };
 338
 339 /*
 340  * Flags used by ext4_get_blocks()
 341  */
 342         /* Allocate any needed blocks and/or convert an unitialized
 343            extent to be an initialized ext4 */
 344 #define EXT4_GET_BLOCKS_CREATE                  0x0001
 345         /* Request the creation of an unitialized extent */
 346 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
 347 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
 348                                                  EXT4_GET_BLOCKS_CREATE)
 349         /* Caller is from the delayed allocation writeout path,
 350            so set the magic i_delalloc_reserve_flag after taking the
 351            inode allocation semaphore for */
 352 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
 353         /* caller is from the direct IO path, request to creation of an
 354         unitialized extents if not allocated, split the uninitialized
 355         extent if blocks has been preallocated already*/
 356 #define EXT4_GET_BLOCKS_DIO                     0x0008
 357 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
 358 #define EXT4_GET_BLOCKS_DIO_CREATE_EXT          (EXT4_GET_BLOCKS_DIO|\
 359                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
 360         /* Convert extent to initialized after direct IO complete */
 361 #define EXT4_GET_BLOCKS_DIO_CONVERT_EXT         (EXT4_GET_BLOCKS_CONVERT|\
 362                                          EXT4_GET_BLOCKS_DIO_CREATE_EXT)
 363
 364 /*
 365  * Flags used by ext4_free_blocks
 366  */
 367 #define EXT4_FREE_BLOCKS_METADATA       0x0001
 368 #define EXT4_FREE_BLOCKS_FORGET         0x0002
 369 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
 370
 371 /*
 372  * ioctl commands
 373  */
 374 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
 375 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
 376 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
 377 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
 378 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
 379 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
 380 #ifdef CONFIG_JBD2_DEBUG
 381 #define EXT4_IOC_WAIT_FOR_READONLY      _IOR('f', 99, long)
 382 #endif
 383 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
 384 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
 385 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
 386 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
 387 #define EXT4_IOC_MIGRATE                _IO('f', 9)
 388  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
 389  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
 390 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
 391 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
 392
 393 /*
 394  * ioctl commands in 32 bit emulation
 395  */
 396 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
 397 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
 398 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
 399 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
 400 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
 401 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
 402 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
 403 #ifdef CONFIG_JBD2_DEBUG
 404 #define EXT4_IOC32_WAIT_FOR_READONLY    _IOR('f', 99, int)
 405 #endif
 406 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
 407 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
 408
 409
 410 /*
 411  *  Mount options
 412  */
 413 struct ext4_mount_options {
 414         unsigned long s_mount_opt;
 415         uid_t s_resuid;
 416         gid_t s_resgid;
 417         unsigned long s_commit_interval;
 418         u32 s_min_batch_time, s_max_batch_time;
 419 #ifdef CONFIG_QUOTA
 420         int s_jquota_fmt;
 421         char *s_qf_names[MAXQUOTAS];
 422 #endif
 423 };
 424
 425 /* Max physical block we can addres w/o extents */
 426 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
 427
 428 /*
 429  * Structure of an inode on the disk
 430  */
 431 struct ext4_inode {
 432         __le16  i_mode;         /* File mode */
 433         __le16  i_uid;          /* Low 16 bits of Owner Uid */
 434         __le32  i_size_lo;      /* Size in bytes */
 435         __le32  i_atime;        /* Access time */
 436         __le32  i_ctime;        /* Inode Change time */
 437         __le32  i_mtime;        /* Modification time */
 438         __le32  i_dtime;        /* Deletion Time */
 439         __le16  i_gid;          /* Low 16 bits of Group Id */
 440         __le16  i_links_count;  /* Links count */
 441         __le32  i_blocks_lo;    /* Blocks count */
 442         __le32  i_flags;        /* File flags */
 443         union {
 444                 struct {
 445                         __le32  l_i_version;
 446                 } linux1;
 447                 struct {
 448                         __u32  h_i_translator;
 449                 } hurd1;
 450                 struct {
 451                         __u32  m_i_reserved1;
 452                 } masix1;
 453         } osd1;                         /* OS dependent 1 */
 454         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
 455         __le32  i_generation;   /* File version (for NFS) */
 456         __le32  i_file_acl_lo;  /* File ACL */
 457         __le32  i_size_high;
 458         __le32  i_obso_faddr;   /* Obsoleted fragment address */
 459         union {
 460                 struct {
 461                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
 462                         __le16  l_i_file_acl_high;
 463                         __le16  l_i_uid_high;   /* these 2 fields */
 464                         __le16  l_i_gid_high;   /* were reserved2[0] */
 465                         __u32   l_i_reserved2;
 466                 } linux2;
 467                 struct {
 468                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 469                         __u16   h_i_mode_high;
 470                         __u16   h_i_uid_high;
 471                         __u16   h_i_gid_high;
 472                         __u32   h_i_author;
 473                 } hurd2;
 474                 struct {
 475                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 476                         __le16  m_i_file_acl_high;
 477                         __u32   m_i_reserved2[2];
 478                 } masix2;
 479         } osd2;                         /* OS dependent 2 */
 480         __le16  i_extra_isize;
 481         __le16  i_pad1;
 482         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
 483         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
 484         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
 485         __le32  i_crtime;       /* File Creation time */
 486         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
 487         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
 488 };
 489
 490 struct move_extent {
 491         __u32 reserved;         /* should be zero */
 492         __u32 donor_fd;         /* donor file descriptor */
 493         __u64 orig_start;       /* logical start offset in block for orig */
 494         __u64 donor_start;      /* logical start offset in block for donor */
 495         __u64 len;              /* block length to be moved */
 496         __u64 moved_len;        /* moved block length */
 497 };
 498
 499 #define EXT4_EPOCH_BITS 2
 500 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
 501 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
 502
 503 /*
 504  * Extended fields will fit into an inode if the filesystem was formatted
 505  * with large inodes (-I 256 or larger) and there are not currently any EAs
 506  * consuming all of the available space. For new inodes we always reserve
 507  * enough space for the kernel's known extended fields, but for inodes
 508  * created with an old kernel this might not have been the case. None of
 509  * the extended inode fields is critical for correct filesystem operation.
 510  * This macro checks if a certain field fits in the inode. Note that
 511  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
 512  */
 513 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
 514         ((offsetof(typeof(*ext4_inode), field) +        \
 515           sizeof((ext4_inode)->field))                  \
 516         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
 517             (einode)->i_extra_isize))                   \
 518
 519 static inline __le32 ext4_encode_extra_time(struct timespec *time)
 520 {
 521        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
 522                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
 523                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
 524 }
 525
 526 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
 527 {
 528        if (sizeof(time->tv_sec) > 4)
 529                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
 530                                << 32;
 531        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
 532 }
 533
 534 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
 535 do {                                                                           \
 536         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
 537         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 538                 (raw_inode)->xtime ## _extra =                                 \
 539                                 ext4_encode_extra_time(&(inode)->xtime);       \
 540 } while (0)
 541
 542 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
 543 do {                                                                           \
 544         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 545                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
 546         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 547                 (raw_inode)->xtime ## _extra =                                 \
 548                                 ext4_encode_extra_time(&(einode)->xtime);      \
 549 } while (0)
 550
 551 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
 552 do {                                                                           \
 553         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
 554         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 555                 ext4_decode_extra_time(&(inode)->xtime,                        \
 556                                        raw_inode->xtime ## _extra);            \
 557 } while (0)
 558
 559 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
 560 do {                                                                           \
 561         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 562                 (einode)->xtime.tv_sec =                                       \
 563                         (signed)le32_to_cpu((raw_inode)->xtime);               \
 564         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 565                 ext4_decode_extra_time(&(einode)->xtime,                       \
 566                                        raw_inode->xtime ## _extra);            \
 567 } while (0)
 568
 569 #define i_disk_version osd1.linux1.l_i_version
 570
 571 #if defined(__KERNEL__) || defined(__linux__)
 572 #define i_reserved1     osd1.linux1.l_i_reserved1
 573 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
 574 #define i_blocks_high   osd2.linux2.l_i_blocks_high
 575 #define i_uid_low       i_uid
 576 #define i_gid_low       i_gid
 577 #define i_uid_high      osd2.linux2.l_i_uid_high
 578 #define i_gid_high      osd2.linux2.l_i_gid_high
 579 #define i_reserved2     osd2.linux2.l_i_reserved2
 580
 581 #elif defined(__GNU__)
 582
 583 #define i_translator    osd1.hurd1.h_i_translator
 584 #define i_uid_high      osd2.hurd2.h_i_uid_high
 585 #define i_gid_high      osd2.hurd2.h_i_gid_high
 586 #define i_author        osd2.hurd2.h_i_author
 587
 588 #elif defined(__masix__)
 589
 590 #define i_reserved1     osd1.masix1.m_i_reserved1
 591 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
 592 #define i_reserved2     osd2.masix2.m_i_reserved2
 593
 594 #endif /* defined(__KERNEL__) || defined(__linux__) */
 595
 596 /*
 597  * storage for cached extent
 598  */
 599 struct ext4_ext_cache {
 600         ext4_fsblk_t    ec_start;
 601         ext4_lblk_t     ec_block;
 602         __u32           ec_len; /* must be 32bit to return holes */
 603         __u32           ec_type;
 604 };
 605
 606 /*
 607  * fourth extended file system inode data in memory
 608  */
 609 struct ext4_inode_info {
 610         __le32  i_data[15];     /* unconverted */
 611         __u32   i_flags;
 612         ext4_fsblk_t    i_file_acl;
 613         __u32   i_dtime;
 614
 615         /*
 616          * i_block_group is the number of the block group which contains
 617          * this file's inode.  Constant across the lifetime of the inode,
 618          * it is ued for making block allocation decisions - we try to
 619          * place a file's data blocks near its inode block, and new inodes
 620          * near to their parent directory's inode.
 621          */
 622         ext4_group_t    i_block_group;
 623         unsigned long   i_state_flags;          /* Dynamic state flags */
 624
 625         ext4_lblk_t             i_dir_start_lookup;
 626 #ifdef CONFIG_EXT4_FS_XATTR
 627         /*
 628          * Extended attributes can be read independently of the main file
 629          * data. Taking i_mutex even when reading would cause contention
 630          * between readers of EAs and writers of regular file data, so
 631          * instead we synchronize on xattr_sem when reading or changing
 632          * EAs.
 633          */
 634         struct rw_semaphore xattr_sem;
 635 #endif
 636
 637         struct list_head i_orphan;      /* unlinked but open inodes */
 638
 639         /*
 640          * i_disksize keeps track of what the inode size is ON DISK, not
 641          * in memory.  During truncate, i_size is set to the new size by
 642          * the VFS prior to calling ext4_truncate(), but the filesystem won't
 643          * set i_disksize to 0 until the truncate is actually under way.
 644          *
 645          * The intent is that i_disksize always represents the blocks which
 646          * are used by this file.  This allows recovery to restart truncate
 647          * on orphans if we crash during truncate.  We actually write i_disksize
 648          * into the on-disk inode when writing inodes out, instead of i_size.
 649          *
 650          * The only time when i_disksize and i_size may be different is when
 651          * a truncate is in progress.  The only things which change i_disksize
 652          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
 653          */
 654         loff_t  i_disksize;
 655
 656         /*
 657          * i_data_sem is for serialising ext4_truncate() against
 658          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
 659          * data tree are chopped off during truncate. We can't do that in
 660          * ext4 because whenever we perform intermediate commits during
 661          * truncate, the inode and all the metadata blocks *must* be in a
 662          * consistent state which allows truncation of the orphans to restart
 663          * during recovery.  Hence we must fix the get_block-vs-truncate race
 664          * by other means, so we have i_data_sem.
 665          */
 666         struct rw_semaphore i_data_sem;
 667         struct inode vfs_inode;
 668         struct jbd2_inode jinode;
 669
 670         struct ext4_ext_cache i_cached_extent;
 671         /*
 672          * File creation time. Its function is same as that of
 673          * struct timespec i_{a,c,m}time in the generic inode.
 674          */
 675         struct timespec i_crtime;
 676
 677         /* mballoc */
 678         struct list_head i_prealloc_list;
 679         spinlock_t i_prealloc_lock;
 680
 681         /* ialloc */
 682         ext4_group_t    i_last_alloc_group;
 683
 684         /* allocation reservation info for delalloc */
 685         unsigned int i_reserved_data_blocks;
 686         unsigned int i_reserved_meta_blocks;
 687         unsigned int i_allocated_meta_blocks;
 688         unsigned short i_delalloc_reserved_flag;
 689         sector_t i_da_metadata_calc_last_lblock;
 690         int i_da_metadata_calc_len;
 691
 692         /* on-disk additional length */
 693         __u16 i_extra_isize;
 694
 695         spinlock_t i_block_reservation_lock;
 696 #ifdef CONFIG_QUOTA
 697         /* quota space reservation, managed internally by quota code */
 698         qsize_t i_reserved_quota;
 699 #endif
 700
 701         /* completed async DIOs that might need unwritten extents handling */
 702         struct list_head i_aio_dio_complete_list;
 703         /* current io_end structure for async DIO write*/
 704         ext4_io_end_t *cur_aio_dio;
 705
 706         /*
 707          * Transactions that contain inode's metadata needed to complete
 708          * fsync and fdatasync, respectively.
 709          */
 710         tid_t i_sync_tid;
 711         tid_t i_datasync_tid;
 712 };
 713
 714 /*
 715  * File system states
 716  */
 717 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
 718 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
 719 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
 720
 721 /*
 722  * Misc. filesystem flags
 723  */
 724 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
 725 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
 726 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
 727
 728 /*
 729  * Mount flags
 730  */
 731 #define EXT4_MOUNT_OLDALLOC             0x00002  /* Don't use the new Orlov allocator */
 732 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
 733 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
 734 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
 735 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
 736 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
 737 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
 738 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
 739 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
 740 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
 741 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
 742 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
 743 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
 744 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
 745 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
 746 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
 747 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
 748 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
 749 #define EXT4_MOUNT_NOBH                 0x40000 /* No bufferheads */
 750 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
 751 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
 752 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
 753 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
 754 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
 755 #define EXT4_MOUNT_I_VERSION            0x2000000 /* i_version support */
 756 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
 757 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
 758 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
 759 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
 760
 761 #define clear_opt(o, opt)               o &= ~EXT4_MOUNT_##opt
 762 #define set_opt(o, opt)                 o |= EXT4_MOUNT_##opt
 763 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
 764                                          EXT4_MOUNT_##opt)
 765
 766 #define ext4_set_bit                    ext2_set_bit
 767 #define ext4_set_bit_atomic             ext2_set_bit_atomic
 768 #define ext4_clear_bit                  ext2_clear_bit
 769 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
 770 #define ext4_test_bit                   ext2_test_bit
 771 #define ext4_find_first_zero_bit        ext2_find_first_zero_bit
 772 #define ext4_find_next_zero_bit         ext2_find_next_zero_bit
 773 #define ext4_find_next_bit              ext2_find_next_bit
 774
 775 /*
 776  * Maximal mount counts between two filesystem checks
 777  */
 778 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
 779 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
 780
 781 /*
 782  * Behaviour when detecting errors
 783  */
 784 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
 785 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
 786 #define EXT4_ERRORS_PANIC               3       /* Panic */
 787 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
 788
 789 /*
 790  * Structure of the super block
 791  */
 792 struct ext4_super_block {
 793 /*00*/  __le32  s_inodes_count;         /* Inodes count */
 794         __le32  s_blocks_count_lo;      /* Blocks count */
 795         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
 796         __le32  s_free_blocks_count_lo; /* Free blocks count */
 797 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
 798         __le32  s_first_data_block;     /* First Data Block */
 799         __le32  s_log_block_size;       /* Block size */
 800         __le32  s_obso_log_frag_size;   /* Obsoleted fragment size */
 801 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
 802         __le32  s_obso_frags_per_group; /* Obsoleted fragments per group */
 803         __le32  s_inodes_per_group;     /* # Inodes per group */
 804         __le32  s_mtime;                /* Mount time */
 805 /*30*/  __le32  s_wtime;                /* Write time */
 806         __le16  s_mnt_count;            /* Mount count */
 807         __le16  s_max_mnt_count;        /* Maximal mount count */
 808         __le16  s_magic;                /* Magic signature */
 809         __le16  s_state;                /* File system state */
 810         __le16  s_errors;               /* Behaviour when detecting errors */
 811         __le16  s_minor_rev_level;      /* minor revision level */
 812 /*40*/  __le32  s_lastcheck;            /* time of last check */
 813         __le32  s_checkinterval;        /* max. time between checks */
 814         __le32  s_creator_os;           /* OS */
 815         __le32  s_rev_level;            /* Revision level */
 816 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
 817         __le16  s_def_resgid;           /* Default gid for reserved blocks */
 818         /*
 819          * These fields are for EXT4_DYNAMIC_REV superblocks only.
 820          *
 821          * Note: the difference between the compatible feature set and
 822          * the incompatible feature set is that if there is a bit set
 823          * in the incompatible feature set that the kernel doesn't
 824          * know about, it should refuse to mount the filesystem.
 825          *
 826          * e2fsck's requirements are more strict; if it doesn't know
 827          * about a feature in either the compatible or incompatible
 828          * feature set, it must abort and not try to meddle with
 829          * things it doesn't understand...
 830          */
 831         __le32  s_first_ino;            /* First non-reserved inode */
 832         __le16  s_inode_size;           /* size of inode structure */
 833         __le16  s_block_group_nr;       /* block group # of this superblock */
 834         __le32  s_feature_compat;       /* compatible feature set */
 835 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
 836         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
 837 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
 838 /*78*/  char    s_volume_name[16];      /* volume name */
 839 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
 840 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
 841         /*
 842          * Performance hints.  Directory preallocation should only
 843          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
 844          */
 845         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
 846         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
 847         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
 848         /*
 849          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
 850          */
 851 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
 852 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
 853         __le32  s_journal_dev;          /* device number of journal file */
 854         __le32  s_last_orphan;          /* start of list of inodes to delete */
 855         __le32  s_hash_seed[4];         /* HTREE hash seed */
 856         __u8    s_def_hash_version;     /* Default hash version to use */
 857         __u8    s_reserved_char_pad;
 858         __le16  s_desc_size;            /* size of group descriptor */
 859 /*100*/ __le32  s_default_mount_opts;
 860         __le32  s_first_meta_bg;        /* First metablock block group */
 861         __le32  s_mkfs_time;            /* When the filesystem was created */
 862         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
 863         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
 864 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
 865         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
 866         __le32  s_free_blocks_count_hi; /* Free blocks count */
 867         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
 868         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
 869         __le32  s_flags;                /* Miscellaneous flags */
 870         __le16  s_raid_stride;          /* RAID stride */
 871         __le16  s_mmp_interval;         /* # seconds to wait in MMP checking */
 872         __le64  s_mmp_block;            /* Block for multi-mount protection */
 873         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
 874         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
 875         __u8    s_reserved_char_pad2;
 876         __le16  s_reserved_pad;
 877         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
 878         __u32   s_reserved[160];        /* Padding to the end of the block */
 879 };
 880
 881 #ifdef __KERNEL__
 882
 883 /*
 884  * run-time mount flags
 885  */
 886 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
 887 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
 888
 889 /*
 890  * fourth extended-fs super-block data in memory
 891  */
 892 struct ext4_sb_info {
 893         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
 894         unsigned long s_inodes_per_block;/* Number of inodes per block */
 895         unsigned long s_blocks_per_group;/* Number of blocks in a group */
 896         unsigned long s_inodes_per_group;/* Number of inodes in a group */
 897         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
 898         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
 899         unsigned long s_desc_per_block; /* Number of group descriptors per block */
 900         ext4_group_t s_groups_count;    /* Number of groups in the fs */
 901         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
 902         unsigned long s_overhead_last;  /* Last calculated overhead */
 903         unsigned long s_blocks_last;    /* Last seen block count */
 904         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
 905         struct buffer_head * s_sbh;     /* Buffer containing the super block */
 906         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
 907         struct buffer_head **s_group_desc;
 908         unsigned int s_mount_opt;
 909         unsigned int s_mount_flags;
 910         ext4_fsblk_t s_sb_block;
 911         uid_t s_resuid;
 912         gid_t s_resgid;
 913         unsigned short s_mount_state;
 914         unsigned short s_pad;
 915         int s_addr_per_block_bits;
 916         int s_desc_per_block_bits;
 917         int s_inode_size;
 918         int s_first_ino;
 919         unsigned int s_inode_readahead_blks;
 920         unsigned int s_inode_goal;
 921         spinlock_t s_next_gen_lock;
 922         u32 s_next_generation;
 923         u32 s_hash_seed[4];
 924         int s_def_hash_version;
 925         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
 926         struct percpu_counter s_freeblocks_counter;
 927         struct percpu_counter s_freeinodes_counter;
 928         struct percpu_counter s_dirs_counter;
 929         struct percpu_counter s_dirtyblocks_counter;
 930         struct blockgroup_lock *s_blockgroup_lock;
 931         struct proc_dir_entry *s_proc;
 932         struct kobject s_kobj;
 933         struct completion s_kobj_unregister;
 934
 935         /* Journaling */
 936         struct inode *s_journal_inode;
 937         struct journal_s *s_journal;
 938         struct list_head s_orphan;
 939         struct mutex s_orphan_lock;
 940         struct mutex s_resize_lock;
 941         unsigned long s_commit_interval;
 942         u32 s_max_batch_time;
 943         u32 s_min_batch_time;
 944         struct block_device *journal_bdev;
 945 #ifdef CONFIG_JBD2_DEBUG
 946         struct timer_list turn_ro_timer;        /* For turning read-only (crash simulation) */
 947         wait_queue_head_t ro_wait_queue;        /* For people waiting for the fs to go read-only */
 948 #endif
 949 #ifdef CONFIG_QUOTA
 950         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
 951         int s_jquota_fmt;                       /* Format of quota to use */
 952 #endif
 953         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
 954         struct rb_root system_blks;
 955
 956 #ifdef EXTENTS_STATS
 957         /* ext4 extents stats */
 958         unsigned long s_ext_min;
 959         unsigned long s_ext_max;
 960         unsigned long s_depth_max;
 961         spinlock_t s_ext_stats_lock;
 962         unsigned long s_ext_blocks;
 963         unsigned long s_ext_extents;
 964 #endif
 965
 966         /* for buddy allocator */
 967         struct ext4_group_info ***s_group_info;
 968         struct inode *s_buddy_cache;
 969         long s_blocks_reserved;
 970         spinlock_t s_reserve_lock;
 971         spinlock_t s_md_lock;
 972         tid_t s_last_transaction;
 973         unsigned short *s_mb_offsets;
 974         unsigned int *s_mb_maxs;
 975
 976         /* tunables */
 977         unsigned long s_stripe;
 978         unsigned int s_mb_stream_request;
 979         unsigned int s_mb_max_to_scan;
 980         unsigned int s_mb_min_to_scan;
 981         unsigned int s_mb_stats;
 982         unsigned int s_mb_order2_reqs;
 983         unsigned int s_mb_group_prealloc;
 984         unsigned int s_max_writeback_mb_bump;
 985         /* where last allocation was done - for stream allocation */
 986         unsigned long s_mb_last_group;
 987         unsigned long s_mb_last_start;
 988
 989         /* stats for buddy allocator */
 990         spinlock_t s_mb_pa_lock;
 991         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
 992         atomic_t s_bal_success; /* we found long enough chunks */
 993         atomic_t s_bal_allocated;       /* in blocks */
 994         atomic_t s_bal_ex_scanned;      /* total extents scanned */
 995         atomic_t s_bal_goals;   /* goal hits */
 996         atomic_t s_bal_breaks;  /* too long searches */
 997         atomic_t s_bal_2orders; /* 2^order hits */
 998         spinlock_t s_bal_lock;
 999         unsigned long s_mb_buddies_generated;
1000         unsigned long long s_mb_generation_time;
1001         atomic_t s_mb_lost_chunks;
1002         atomic_t s_mb_preallocated;
1003         atomic_t s_mb_discarded;
1004         atomic_t s_lock_busy;
1005
1006         /* locality groups */
1007         struct ext4_locality_group *s_locality_groups;
1008
1009         /* for write statistics */
1010         unsigned long s_sectors_written_start;
1011         u64 s_kbytes_written;
1012
1013         unsigned int s_log_groups_per_flex;
1014         struct flex_groups *s_flex_groups;
1015
1016         /* workqueue for dio unwritten */
1017         struct workqueue_struct *dio_unwritten_wq;
1018 };
1019
1020 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1021 {
1022         return sb->s_fs_info;
1023 }
1024 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1025 {
1026         return container_of(inode, struct ext4_inode_info, vfs_inode);
1027 }
1028
1029 static inline struct timespec ext4_current_time(struct inode *inode)
1030 {
1031         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1032                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1033 }
1034
1035 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1036 {
1037         return ino == EXT4_ROOT_INO ||
1038                 ino == EXT4_JOURNAL_INO ||
1039                 ino == EXT4_RESIZE_INO ||
1040                 (ino >= EXT4_FIRST_INO(sb) &&
1041                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1042 }
1043
1044 /*
1045  * Inode dynamic state flags
1046  */
1047 enum {
1048         EXT4_STATE_JDATA,               /* journaled data exists */
1049         EXT4_STATE_NEW,                 /* inode is newly created */
1050         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1051         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1052         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1053         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1054         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1055 };
1056
1057 static inline int ext4_test_inode_state(struct inode *inode, int bit)
1058 {
1059         return test_bit(bit, &EXT4_I(inode)->i_state_flags);
1060 }
1061
1062 static inline void ext4_set_inode_state(struct inode *inode, int bit)
1063 {
1064         set_bit(bit, &EXT4_I(inode)->i_state_flags);
1065 }
1066
1067 static inline void ext4_clear_inode_state(struct inode *inode, int bit)
1068 {
1069         clear_bit(bit, &EXT4_I(inode)->i_state_flags);
1070 }
1071 #else
1072 /* Assume that user mode programs are passing in an ext4fs superblock, not
1073  * a kernel struct super_block.  This will allow us to call the feature-test
1074  * macros from user land. */
1075 #define EXT4_SB(sb)     (sb)
1076 #endif
1077
1078 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1079
1080 /*
1081  * Codes for operating systems
1082  */
1083 #define EXT4_OS_LINUX           0
1084 #define EXT4_OS_HURD            1
1085 #define EXT4_OS_MASIX           2
1086 #define EXT4_OS_FREEBSD         3
1087 #define EXT4_OS_LITES           4
1088
1089 /*
1090  * Revision levels
1091  */
1092 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1093 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1094
1095 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1096 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1097
1098 #define EXT4_GOOD_OLD_INODE_SIZE 128
1099
1100 /*
1101  * Feature set definitions
1102  */
1103
1104 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1105         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1106 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1107         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1108 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1109         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1110 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1111         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1112 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1113         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1114 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1115         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1116 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1117         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1118 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1119         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1120 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1121         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1122
1123 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1124 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1125 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1126 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1127 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1128 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1129
1130 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1131 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1132 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1133 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1134 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1135 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1136 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1137
1138 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1139 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1140 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1141 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1142 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1143 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1144 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1145 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1146 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1147
1148 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1149 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1150                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1151                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1152                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1153                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1154                                          EXT4_FEATURE_INCOMPAT_FLEX_BG)
1155 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1156                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1157                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1158                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1159                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1160                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1161                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE)
1162
1163 /*
1164  * Default values for user and/or group using reserved blocks
1165  */
1166 #define EXT4_DEF_RESUID         0
1167 #define EXT4_DEF_RESGID         0
1168
1169 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1170
1171 /*
1172  * Default mount options
1173  */
1174 #define EXT4_DEFM_DEBUG         0x0001
1175 #define EXT4_DEFM_BSDGROUPS     0x0002
1176 #define EXT4_DEFM_XATTR_USER    0x0004
1177 #define EXT4_DEFM_ACL           0x0008
1178 #define EXT4_DEFM_UID16         0x0010
1179 #define EXT4_DEFM_JMODE         0x0060
1180 #define EXT4_DEFM_JMODE_DATA    0x0020
1181 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1182 #define EXT4_DEFM_JMODE_WBACK   0x0060
1183
1184 /*
1185  * Default journal batch times
1186  */
1187 #define EXT4_DEF_MIN_BATCH_TIME 0
1188 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1189
1190 /*
1191  * Minimum number of groups in a flexgroup before we separate out
1192  * directories into the first block group of a flexgroup
1193  */
1194 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1195
1196 /*
1197  * Structure of a directory entry
1198  */
1199 #define EXT4_NAME_LEN 255
1200
1201 struct ext4_dir_entry {
1202         __le32  inode;                  /* Inode number */
1203         __le16  rec_len;                /* Directory entry length */
1204         __le16  name_len;               /* Name length */
1205         char    name[EXT4_NAME_LEN];    /* File name */
1206 };
1207
1208 /*
1209  * The new version of the directory entry.  Since EXT4 structures are
1210  * stored in intel byte order, and the name_len field could never be
1211  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1212  * file_type field.
1213  */
1214 struct ext4_dir_entry_2 {
1215         __le32  inode;                  /* Inode number */
1216         __le16  rec_len;                /* Directory entry length */
1217         __u8    name_len;               /* Name length */
1218         __u8    file_type;
1219         char    name[EXT4_NAME_LEN];    /* File name */
1220 };
1221
1222 /*
1223  * Ext4 directory file types.  Only the low 3 bits are used.  The
1224  * other bits are reserved for now.
1225  */
1226 #define EXT4_FT_UNKNOWN         0
1227 #define EXT4_FT_REG_FILE        1
1228 #define EXT4_FT_DIR             2
1229 #define EXT4_FT_CHRDEV          3
1230 #define EXT4_FT_BLKDEV          4
1231 #define EXT4_FT_FIFO            5
1232 #define EXT4_FT_SOCK            6
1233 #define EXT4_FT_SYMLINK         7
1234
1235 #define EXT4_FT_MAX             8
1236
1237 /*
1238  * EXT4_DIR_PAD defines the directory entries boundaries
1239  *
1240  * NOTE: It must be a multiple of 4
1241  */
1242 #define EXT4_DIR_PAD                    4
1243 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1244 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1245                                          ~EXT4_DIR_ROUND)
1246 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1247
1248 /*
1249  * Hash Tree Directory indexing
1250  * (c) Daniel Phillips, 2001
1251  */
1252
1253 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1254                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1255                       (EXT4_I(dir)->i_flags & EXT4_INDEX_FL))
1256 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1257 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1258
1259 /* Legal values for the dx_root hash_version field: */
1260
1261 #define DX_HASH_LEGACY          0
1262 #define DX_HASH_HALF_MD4        1
1263 #define DX_HASH_TEA             2
1264 #define DX_HASH_LEGACY_UNSIGNED 3
1265 #define DX_HASH_HALF_MD4_UNSIGNED       4
1266 #define DX_HASH_TEA_UNSIGNED            5
1267
1268 #ifdef __KERNEL__
1269
1270 /* hash info structure used by the directory hash */
1271 struct dx_hash_info
1272 {
1273         u32             hash;
1274         u32             minor_hash;
1275         int             hash_version;
1276         u32             *seed;
1277 };
1278
1279 #define EXT4_HTREE_EOF  0x7fffffff
1280
1281 /*
1282  * Control parameters used by ext4_htree_next_block
1283  */
1284 #define HASH_NB_ALWAYS          1
1285
1286
1287 /*
1288  * Describe an inode's exact location on disk and in memory
1289  */
1290 struct ext4_iloc
1291 {
1292         struct buffer_head *bh;
1293         unsigned long offset;
1294         ext4_group_t block_group;
1295 };
1296
1297 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1298 {
1299         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1300 }
1301
1302 /*
1303  * This structure is stuffed into the struct file's private_data field
1304  * for directories.  It is where we put information so that we can do
1305  * readdir operations in hash tree order.
1306  */
1307 struct dir_private_info {
1308         struct rb_root  root;
1309         struct rb_node  *curr_node;
1310         struct fname    *extra_fname;
1311         loff_t          last_pos;
1312         __u32           curr_hash;
1313         __u32           curr_minor_hash;
1314         __u32           next_hash;
1315 };
1316
1317 /* calculate the first block number of the group */
1318 static inline ext4_fsblk_t
1319 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1320 {
1321         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1322                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1323 }
1324
1325 /*
1326  * Special error return code only used by dx_probe() and its callers.
1327  */
1328 #define ERR_BAD_DX_DIR  -75000
1329
1330 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1331                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1332
1333 extern struct proc_dir_entry *ext4_proc_root;
1334
1335 /*
1336  * Function prototypes
1337  */
1338
1339 /*
1340  * Ok, these declarations are also in <linux/kernel.h> but none of the
1341  * ext4 source programs needs to include it so they are duplicated here.
1342  */
1343 # define NORET_TYPE     /**/
1344 # define ATTRIB_NORET   __attribute__((noreturn))
1345 # define NORET_AND      noreturn,
1346
1347 /* bitmap.c */
1348 extern unsigned int ext4_count_free(struct buffer_head *, unsigned);
1349
1350 /* balloc.c */
1351 extern unsigned int ext4_block_group(struct super_block *sb,
1352                         ext4_fsblk_t blocknr);
1353 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1354                         ext4_fsblk_t blocknr);
1355 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1356 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1357                         ext4_group_t group);
1358 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1359                         ext4_fsblk_t goal, unsigned long *count, int *errp);
1360 extern int ext4_claim_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1361 extern int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1362 extern void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
1363                                 ext4_fsblk_t block, unsigned long count);
1364 extern ext4_fsblk_t ext4_count_free_blocks(struct super_block *);
1365 extern void ext4_check_blocks_bitmap(struct super_block *);
1366 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1367                                                     ext4_group_t block_group,
1368                                                     struct buffer_head ** bh);
1369 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1370 struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1371                                       ext4_group_t block_group);
1372 extern unsigned ext4_init_block_bitmap(struct super_block *sb,
1373                                        struct buffer_head *bh,
1374                                        ext4_group_t group,
1375                                        struct ext4_group_desc *desc);
1376 #define ext4_free_blocks_after_init(sb, group, desc)                    \
1377                 ext4_init_block_bitmap(sb, NULL, group, desc)
1378
1379 /* dir.c */
1380 extern int ext4_check_dir_entry(const char *, struct inode *,
1381                                 struct ext4_dir_entry_2 *,
1382                                 struct buffer_head *, unsigned int);
1383 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1384                                     __u32 minor_hash,
1385                                     struct ext4_dir_entry_2 *dirent);
1386 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1387
1388 /* fsync.c */
1389 extern int ext4_sync_file(struct file *, struct dentry *, int);
1390
1391 /* hash.c */
1392 extern int ext4fs_dirhash(const char *name, int len, struct
1393                           dx_hash_info *hinfo);
1394
1395 /* ialloc.c */
1396 extern struct inode *ext4_new_inode(handle_t *, struct inode *, int,
1397                                     const struct qstr *qstr, __u32 goal);
1398 extern void ext4_free_inode(handle_t *, struct inode *);
1399 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1400 extern unsigned long ext4_count_free_inodes(struct super_block *);
1401 extern unsigned long ext4_count_dirs(struct super_block *);
1402 extern void ext4_check_inodes_bitmap(struct super_block *);
1403 extern unsigned ext4_init_inode_bitmap(struct super_block *sb,
1404                                        struct buffer_head *bh,
1405                                        ext4_group_t group,
1406                                        struct ext4_group_desc *desc);
1407 extern void mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1408
1409 /* mballoc.c */
1410 extern long ext4_mb_stats;
1411 extern long ext4_mb_max_to_scan;
1412 extern int ext4_mb_init(struct super_block *, int);
1413 extern int ext4_mb_release(struct super_block *);
1414 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1415                                 struct ext4_allocation_request *, int *);
1416 extern int ext4_mb_reserve_blocks(struct super_block *, int);
1417 extern void ext4_discard_preallocations(struct inode *);
1418 extern int __init init_ext4_mballoc(void);
1419 extern void exit_ext4_mballoc(void);
1420 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1421                              struct buffer_head *bh, ext4_fsblk_t block,
1422                              unsigned long count, int flags);
1423 extern int ext4_mb_add_groupinfo(struct super_block *sb,
1424                 ext4_group_t i, struct ext4_group_desc *desc);
1425 extern int ext4_mb_get_buddy_cache_lock(struct super_block *, ext4_group_t);
1426 extern void ext4_mb_put_buddy_cache_lock(struct super_block *,
1427                                                 ext4_group_t, int);
1428 /* inode.c */
1429 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1430                                                 ext4_lblk_t, int, int *);
1431 struct buffer_head *ext4_bread(handle_t *, struct inode *,
1432                                                 ext4_lblk_t, int, int *);
1433 int ext4_get_block(struct inode *inode, sector_t iblock,
1434                                 struct buffer_head *bh_result, int create);
1435
1436 extern struct inode *ext4_iget(struct super_block *, unsigned long);
1437 extern int  ext4_write_inode(struct inode *, int);
1438 extern int  ext4_setattr(struct dentry *, struct iattr *);
1439 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1440                                 struct kstat *stat);
1441 extern void ext4_delete_inode(struct inode *);
1442 extern int  ext4_sync_inode(handle_t *, struct inode *);
1443 extern void ext4_dirty_inode(struct inode *);
1444 extern int ext4_change_inode_journal_flag(struct inode *, int);
1445 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1446 extern int ext4_can_truncate(struct inode *inode);
1447 extern void ext4_truncate(struct inode *);
1448 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
1449 extern void ext4_set_inode_flags(struct inode *);
1450 extern void ext4_get_inode_flags(struct ext4_inode_info *);
1451 extern int ext4_alloc_da_blocks(struct inode *inode);
1452 extern void ext4_set_aops(struct inode *inode);
1453 extern int ext4_writepage_trans_blocks(struct inode *);
1454 extern int ext4_meta_trans_blocks(struct inode *, int nrblocks, int idxblocks);
1455 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1456 extern int ext4_block_truncate_page(handle_t *handle,
1457                 struct address_space *mapping, loff_t from);
1458 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1459 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
1460 extern int flush_aio_dio_completed_IO(struct inode *inode);
1461 extern void ext4_da_update_reserve_space(struct inode *inode,
1462                                         int used, int quota_claim);
1463 /* ioctl.c */
1464 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1465 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1466
1467 /* migrate.c */
1468 extern int ext4_ext_migrate(struct inode *);
1469
1470 /* namei.c */
1471 extern unsigned int ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize);
1472 extern __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize);
1473 extern int ext4_orphan_add(handle_t *, struct inode *);
1474 extern int ext4_orphan_del(handle_t *, struct inode *);
1475 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1476                                 __u32 start_minor_hash, __u32 *next_hash);
1477
1478 /* resize.c */
1479 extern int ext4_group_add(struct super_block *sb,
1480                                 struct ext4_new_group_data *input);
1481 extern int ext4_group_extend(struct super_block *sb,
1482                                 struct ext4_super_block *es,
1483                                 ext4_fsblk_t n_blocks_count);
1484
1485 /* super.c */
1486 extern void ext4_error(struct super_block *, const char *, const char *, ...)
1487         __attribute__ ((format (printf, 3, 4)));
1488 extern void __ext4_std_error(struct super_block *, const char *, int);
1489 extern void ext4_abort(struct super_block *, const char *, const char *, ...)
1490         __attribute__ ((format (printf, 3, 4)));
1491 extern void ext4_warning(struct super_block *, const char *, const char *, ...)
1492         __attribute__ ((format (printf, 3, 4)));
1493 extern void ext4_msg(struct super_block *, const char *, const char *, ...)
1494         __attribute__ ((format (printf, 3, 4)));
1495 extern void ext4_grp_locked_error(struct super_block *, ext4_group_t,
1496                                 const char *, const char *, ...)
1497         __attribute__ ((format (printf, 4, 5)));
1498 extern void ext4_update_dynamic_rev(struct super_block *sb);
1499 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1500                                         __u32 compat);
1501 extern int ext4_update_rocompat_feature(handle_t *handle,
1502                                         struct super_block *sb, __u32 rocompat);
1503 extern int ext4_update_incompat_feature(handle_t *handle,
1504                                         struct super_block *sb, __u32 incompat);
1505 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1506                                       struct ext4_group_desc *bg);
1507 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1508                                       struct ext4_group_desc *bg);
1509 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1510                                      struct ext4_group_desc *bg);
1511 extern __u32 ext4_free_blks_count(struct super_block *sb,
1512                                 struct ext4_group_desc *bg);
1513 extern __u32 ext4_free_inodes_count(struct super_block *sb,
1514                                  struct ext4_group_desc *bg);
1515 extern __u32 ext4_used_dirs_count(struct super_block *sb,
1516                                 struct ext4_group_desc *bg);
1517 extern __u32 ext4_itable_unused_count(struct super_block *sb,
1518                                    struct ext4_group_desc *bg);
1519 extern void ext4_block_bitmap_set(struct super_block *sb,
1520                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1521 extern void ext4_inode_bitmap_set(struct super_block *sb,
1522                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1523 extern void ext4_inode_table_set(struct super_block *sb,
1524                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
1525 extern void ext4_free_blks_set(struct super_block *sb,
1526                                struct ext4_group_desc *bg, __u32 count);
1527 extern void ext4_free_inodes_set(struct super_block *sb,
1528                                 struct ext4_group_desc *bg, __u32 count);
1529 extern void ext4_used_dirs_set(struct super_block *sb,
1530                                 struct ext4_group_desc *bg, __u32 count);
1531 extern void ext4_itable_unused_set(struct super_block *sb,
1532                                    struct ext4_group_desc *bg, __u32 count);
1533 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
1534                                    struct ext4_group_desc *gdp);
1535 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
1536                                        struct ext4_group_desc *gdp);
1537
1538 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
1539 {
1540         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
1541                 le32_to_cpu(es->s_blocks_count_lo);
1542 }
1543
1544 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
1545 {
1546         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
1547                 le32_to_cpu(es->s_r_blocks_count_lo);
1548 }
1549
1550 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
1551 {
1552         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
1553                 le32_to_cpu(es->s_free_blocks_count_lo);
1554 }
1555
1556 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
1557                                          ext4_fsblk_t blk)
1558 {
1559         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
1560         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
1561 }
1562
1563 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
1564                                               ext4_fsblk_t blk)
1565 {
1566         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
1567         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
1568 }
1569
1570 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
1571                                            ext4_fsblk_t blk)
1572 {
1573         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
1574         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
1575 }
1576
1577 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
1578 {
1579         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
1580                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
1581                         le32_to_cpu(raw_inode->i_size_lo);
1582         else
1583                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
1584 }
1585
1586 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
1587 {
1588         raw_inode->i_size_lo = cpu_to_le32(i_size);
1589         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
1590 }
1591
1592 static inline
1593 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
1594                                             ext4_group_t group)
1595 {
1596          struct ext4_group_info ***grp_info;
1597          long indexv, indexh;
1598          grp_info = EXT4_SB(sb)->s_group_info;
1599          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
1600          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
1601          return grp_info[indexv][indexh];
1602 }
1603
1604 /*
1605  * Reading s_groups_count requires using smp_rmb() afterwards.  See
1606  * the locking protocol documented in the comments of ext4_group_add()
1607  * in resize.c
1608  */
1609 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
1610 {
1611         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
1612
1613         smp_rmb();
1614         return ngroups;
1615 }
1616
1617 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
1618                                              ext4_group_t block_group)
1619 {
1620         return block_group >> sbi->s_log_groups_per_flex;
1621 }
1622
1623 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
1624 {
1625         return 1 << sbi->s_log_groups_per_flex;
1626 }
1627
1628 #define ext4_std_error(sb, errno)                               \
1629 do {                                                            \
1630         if ((errno))                                            \
1631                 __ext4_std_error((sb), __func__, (errno));      \
1632 } while (0)
1633
1634 #ifdef CONFIG_SMP
1635 /* Each CPU can accumulate percpu_counter_batch blocks in their local
1636  * counters. So we need to make sure we have free blocks more
1637  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
1638  */
1639 #define EXT4_FREEBLOCKS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
1640 #else
1641 #define EXT4_FREEBLOCKS_WATERMARK 0
1642 #endif
1643
1644 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
1645 {
1646         /*
1647          * XXX: replace with spinlock if seen contended -bzzz
1648          */
1649         down_write(&EXT4_I(inode)->i_data_sem);
1650         if (newsize > EXT4_I(inode)->i_disksize)
1651                 EXT4_I(inode)->i_disksize = newsize;
1652         up_write(&EXT4_I(inode)->i_data_sem);
1653         return ;
1654 }
1655
1656 struct ext4_group_info {
1657         unsigned long   bb_state;
1658         struct rb_root  bb_free_root;
1659         ext4_grpblk_t   bb_first_free;  /* first free block */
1660         ext4_grpblk_t   bb_free;        /* total free blocks */
1661         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
1662         struct          list_head bb_prealloc_list;
1663 #ifdef DOUBLE_CHECK
1664         void            *bb_bitmap;
1665 #endif
1666         struct rw_semaphore alloc_sem;
1667         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
1668                                          * regions, index is order.
1669                                          * bb_counters[3] = 5 means
1670                                          * 5 free 8-block regions. */
1671 };
1672
1673 #define EXT4_GROUP_INFO_NEED_INIT_BIT   0
1674
1675 #define EXT4_MB_GRP_NEED_INIT(grp)      \
1676         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
1677
1678 #define EXT4_MAX_CONTENTION             8
1679 #define EXT4_CONTENTION_THRESHOLD       2
1680
1681 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
1682                                               ext4_group_t group)
1683 {
1684         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
1685 }
1686
1687 /*
1688  * Returns true if the filesystem is busy enough that attempts to
1689  * access the block group locks has run into contention.
1690  */
1691 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
1692 {
1693         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
1694 }
1695
1696 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
1697 {
1698         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
1699         if (spin_trylock(lock))
1700                 /*
1701                  * We're able to grab the lock right away, so drop the
1702                  * lock contention counter.
1703                  */
1704                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
1705         else {
1706                 /*
1707                  * The lock is busy, so bump the contention counter,
1708                  * and then wait on the spin lock.
1709                  */
1710                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
1711                                   EXT4_MAX_CONTENTION);
1712                 spin_lock(lock);
1713         }
1714 }
1715
1716 static inline void ext4_unlock_group(struct super_block *sb,
1717                                         ext4_group_t group)
1718 {
1719         spin_unlock(ext4_group_lock_ptr(sb, group));
1720 }
1721
1722 /*
1723  * Inodes and files operations
1724  */
1725
1726 /* dir.c */
1727 extern const struct file_operations ext4_dir_operations;
1728
1729 /* file.c */
1730 extern const struct inode_operations ext4_file_inode_operations;
1731 extern const struct file_operations ext4_file_operations;
1732
1733 /* namei.c */
1734 extern const struct inode_operations ext4_dir_inode_operations;
1735 extern const struct inode_operations ext4_special_inode_operations;
1736 extern struct dentry *ext4_get_parent(struct dentry *child);
1737
1738 /* symlink.c */
1739 extern const struct inode_operations ext4_symlink_inode_operations;
1740 extern const struct inode_operations ext4_fast_symlink_inode_operations;
1741
1742 /* block_validity */
1743 extern void ext4_release_system_zone(struct super_block *sb);
1744 extern int ext4_setup_system_zone(struct super_block *sb);
1745 extern int __init init_ext4_system_zone(void);
1746 extern void exit_ext4_system_zone(void);
1747 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
1748                                  ext4_fsblk_t start_blk,
1749                                  unsigned int count);
1750
1751 /* extents.c */
1752 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
1753 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
1754 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
1755                                        int chunk);
1756 extern int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
1757                                ext4_lblk_t iblock, unsigned int max_blocks,
1758                                struct buffer_head *bh_result, int flags);
1759 extern void ext4_ext_truncate(struct inode *);
1760 extern void ext4_ext_init(struct super_block *);
1761 extern void ext4_ext_release(struct super_block *);
1762 extern long ext4_fallocate(struct inode *inode, int mode, loff_t offset,
1763                           loff_t len);
1764 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
1765                           ssize_t len);
1766 extern int ext4_get_blocks(handle_t *handle, struct inode *inode,
1767                            sector_t block, unsigned int max_blocks,
1768                            struct buffer_head *bh, int flags);
1769 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1770                         __u64 start, __u64 len);
1771 /* move_extent.c */
1772 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
1773                              __u64 start_orig, __u64 start_donor,
1774                              __u64 len, __u64 *moved_len);
1775
1776
1777 /*
1778  * Add new method to test wether block and inode bitmaps are properly
1779  * initialized. With uninit_bg reading the block from disk is not enough
1780  * to mark the bitmap uptodate. We need to also zero-out the bitmap
1781  */
1782 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
1783
1784 static inline int bitmap_uptodate(struct buffer_head *bh)
1785 {
1786         return (buffer_uptodate(bh) &&
1787                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
1788 }
1789 static inline void set_bitmap_uptodate(struct buffer_head *bh)
1790 {
1791         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
1792 }
1793
1794 #endif  /* __KERNEL__ */
1795
1796 #endif  /* _EXT4_H */