SAFE public projects git trees. - safe/jmp/linux-2.6/blob - fs/ext4/ext4.h

   1 /*
   2  *  ext4.h
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card (card@masi.ibp.fr)
   6  * Laboratoire MASI - Institut Blaise Pascal
   7  * Universite Pierre et Marie Curie (Paris VI)
   8  *
   9  *  from
  10  *
  11  *  linux/include/linux/minix_fs.h
  12  *
  13  *  Copyright (C) 1991, 1992  Linus Torvalds
  14  */
  15
  16 #ifndef _EXT4_H
  17 #define _EXT4_H
  18
  19 #include <linux/types.h>
  20 #include <linux/blkdev.h>
  21 #include <linux/magic.h>
  22 #include <linux/jbd2.h>
  23 #include <linux/quota.h>
  24 #include <linux/rwsem.h>
  25 #include <linux/rbtree.h>
  26 #include <linux/seqlock.h>
  27 #include <linux/mutex.h>
  28 #include <linux/timer.h>
  29 #include <linux/wait.h>
  30 #include <linux/blockgroup_lock.h>
  31 #include <linux/percpu_counter.h>
  32
  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              0x0001
  71 /* blocks already reserved */
  72 #define EXT4_MB_HINT_RESERVED           0x0002
  73 /* metadata is being allocated */
  74 #define EXT4_MB_HINT_METADATA           0x0004
  75 /* first blocks in the file */
  76 #define EXT4_MB_HINT_FIRST              0x0008
  77 /* search for the best chunk */
  78 #define EXT4_MB_HINT_BEST               0x0010
  79 /* data is being allocated */
  80 #define EXT4_MB_HINT_DATA               0x0020
  81 /* don't preallocate (for tails) */
  82 #define EXT4_MB_HINT_NOPREALLOC         0x0040
  83 /* allocate for locality group */
  84 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
  85 /* allocate goal blocks or none */
  86 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
  87 /* goal is meaningful */
  88 #define EXT4_MB_HINT_TRY_GOAL           0x0200
  89 /* blocks already pre-reserved by delayed allocation */
  90 #define EXT4_MB_DELALLOC_RESERVED       0x0400
  91 /* We are doing stream allocation */
  92 #define EXT4_MB_STREAM_ALLOC            0x0800
  93
  94
  95 struct ext4_allocation_request {
  96         /* target inode for block we're allocating */
  97         struct inode *inode;
  98         /* how many blocks we want to allocate */
  99         unsigned int len;
 100         /* logical block in target inode */
 101         ext4_lblk_t logical;
 102         /* the closest logical allocated block to the left */
 103         ext4_lblk_t lleft;
 104         /* the closest logical allocated block to the right */
 105         ext4_lblk_t lright;
 106         /* phys. target (a hint) */
 107         ext4_fsblk_t goal;
 108         /* phys. block for the closest logical allocated block to the left */
 109         ext4_fsblk_t pleft;
 110         /* phys. block for the closest logical allocated block to the right */
 111         ext4_fsblk_t pright;
 112         /* flags. see above EXT4_MB_HINT_* */
 113         unsigned int flags;
 114 };
 115
 116 /*
 117  * For delayed allocation tracking
 118  */
 119 struct mpage_da_data {
 120         struct inode *inode;
 121         sector_t b_blocknr;             /* start block number of extent */
 122         size_t b_size;                  /* size of extent */
 123         unsigned long b_state;          /* state of the extent */
 124         unsigned long first_page, next_page;    /* extent of pages */
 125         struct writeback_control *wbc;
 126         int io_done;
 127         int pages_written;
 128         int retval;
 129 };
 130
 131 typedef struct ext4_io_end {
 132         struct inode            *inode;         /* file being written to */
 133         unsigned int            flag;           /* sync IO or AIO */
 134         int                     error;          /* I/O error code */
 135         ext4_lblk_t             offset;         /* offset in the file */
 136         size_t                  size;           /* size of the extent */
 137         struct work_struct      work;           /* data work queue */
 138 } ext4_io_end_t;
 139
 140 /*
 141  * Special inodes numbers
 142  */
 143 #define EXT4_BAD_INO             1      /* Bad blocks inode */
 144 #define EXT4_ROOT_INO            2      /* Root inode */
 145 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
 146 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
 147 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
 148 #define EXT4_JOURNAL_INO         8      /* Journal inode */
 149
 150 /* First non-reserved inode for old ext4 filesystems */
 151 #define EXT4_GOOD_OLD_FIRST_INO 11
 152
 153 /*
 154  * Maximal count of links to a file
 155  */
 156 #define EXT4_LINK_MAX           65000
 157
 158 /*
 159  * Macro-instructions used to manage several block sizes
 160  */
 161 #define EXT4_MIN_BLOCK_SIZE             1024
 162 #define EXT4_MAX_BLOCK_SIZE             65536
 163 #define EXT4_MIN_BLOCK_LOG_SIZE         10
 164 #ifdef __KERNEL__
 165 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
 166 #else
 167 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
 168 #endif
 169 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
 170 #ifdef __KERNEL__
 171 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
 172 #else
 173 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
 174 #endif
 175 #ifdef __KERNEL__
 176 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
 177 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
 178 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
 179 #else
 180 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 181                                  EXT4_GOOD_OLD_INODE_SIZE : \
 182                                  (s)->s_inode_size)
 183 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 184                                  EXT4_GOOD_OLD_FIRST_INO : \
 185                                  (s)->s_first_ino)
 186 #endif
 187 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
 188
 189 /*
 190  * Structure of a blocks group descriptor
 191  */
 192 struct ext4_group_desc
 193 {
 194         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
 195         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
 196         __le32  bg_inode_table_lo;      /* Inodes table block */
 197         __le16  bg_free_blocks_count_lo;/* Free blocks count */
 198         __le16  bg_free_inodes_count_lo;/* Free inodes count */
 199         __le16  bg_used_dirs_count_lo;  /* Directories count */
 200         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
 201         __u32   bg_reserved[2];         /* Likely block/inode bitmap checksum */
 202         __le16  bg_itable_unused_lo;    /* Unused inodes count */
 203         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
 204         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
 205         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
 206         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
 207         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
 208         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
 209         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
 210         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
 211         __u32   bg_reserved2[3];
 212 };
 213
 214 /*
 215  * Structure of a flex block group info
 216  */
 217
 218 struct flex_groups {
 219         atomic_t free_inodes;
 220         atomic_t free_blocks;
 221         atomic_t used_dirs;
 222 };
 223
 224 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
 225 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
 226 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
 227
 228 /*
 229  * Macro-instructions used to manage group descriptors
 230  */
 231 #define EXT4_MIN_DESC_SIZE              32
 232 #define EXT4_MIN_DESC_SIZE_64BIT        64
 233 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
 234 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
 235 #ifdef __KERNEL__
 236 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
 237 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
 238 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
 239 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
 240 #else
 241 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
 242 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
 243 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
 244 #endif
 245
 246 /*
 247  * Constants relative to the data blocks
 248  */
 249 #define EXT4_NDIR_BLOCKS                12
 250 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
 251 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
 252 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
 253 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
 254
 255 /*
 256  * Inode flags
 257  */
 258 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
 259 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
 260 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
 261 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
 262 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
 263 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
 264 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
 265 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
 266 /* Reserved for compression usage... */
 267 #define EXT4_DIRTY_FL                   0x00000100
 268 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
 269 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
 270 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
 271 /* End compression flags --- maybe not all used */
 272 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
 273 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
 274 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
 275 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
 276 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
 277 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
 278 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
 279 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
 280 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
 281
 282 #define EXT4_FL_USER_VISIBLE            0x000BDFFF /* User visible flags */
 283 #define EXT4_FL_USER_MODIFIABLE         0x000B80FF /* User modifiable flags */
 284
 285 /* Flags that should be inherited by new inodes from their parent. */
 286 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
 287                            EXT4_SYNC_FL | EXT4_IMMUTABLE_FL | EXT4_APPEND_FL |\
 288                            EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
 289                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
 290                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
 291
 292 /* Flags that are appropriate for regular files (all but dir-specific ones). */
 293 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
 294
 295 /* Flags that are appropriate for non-directories/regular files. */
 296 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
 297
 298 /* Mask out flags that are inappropriate for the given type of inode. */
 299 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
 300 {
 301         if (S_ISDIR(mode))
 302                 return flags;
 303         else if (S_ISREG(mode))
 304                 return flags & EXT4_REG_FLMASK;
 305         else
 306                 return flags & EXT4_OTHER_FLMASK;
 307 }
 308
 309 /*
 310  * Inode dynamic state flags
 311  */
 312 #define EXT4_STATE_JDATA                0x00000001 /* journaled data exists */
 313 #define EXT4_STATE_NEW                  0x00000002 /* inode is newly created */
 314 #define EXT4_STATE_XATTR                0x00000004 /* has in-inode xattrs */
 315 #define EXT4_STATE_NO_EXPAND            0x00000008 /* No space for expansion */
 316 #define EXT4_STATE_DA_ALLOC_CLOSE       0x00000010 /* Alloc DA blks on close */
 317 #define EXT4_STATE_EXT_MIGRATE          0x00000020 /* Inode is migrating */
 318
 319 /* Used to pass group descriptor data when online resize is done */
 320 struct ext4_new_group_input {
 321         __u32 group;            /* Group number for this data */
 322         __u64 block_bitmap;     /* Absolute block number of block bitmap */
 323         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
 324         __u64 inode_table;      /* Absolute block number of inode table start */
 325         __u32 blocks_count;     /* Total number of blocks in this group */
 326         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
 327         __u16 unused;
 328 };
 329
 330 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
 331 struct ext4_new_group_data {
 332         __u32 group;
 333         __u64 block_bitmap;
 334         __u64 inode_bitmap;
 335         __u64 inode_table;
 336         __u32 blocks_count;
 337         __u16 reserved_blocks;
 338         __u16 unused;
 339         __u32 free_blocks_count;
 340 };
 341
 342 /*
 343  * Flags used by ext4_get_blocks()
 344  */
 345         /* Allocate any needed blocks and/or convert an unitialized
 346            extent to be an initialized ext4 */
 347 #define EXT4_GET_BLOCKS_CREATE                  0x0001
 348         /* Request the creation of an unitialized extent */
 349 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
 350 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
 351                                                  EXT4_GET_BLOCKS_CREATE)
 352         /* Caller is from the delayed allocation writeout path,
 353            so set the magic i_delalloc_reserve_flag after taking the
 354            inode allocation semaphore for */
 355 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
 356         /* Call ext4_da_update_reserve_space() after successfully
 357            allocating the blocks */
 358 #define EXT4_GET_BLOCKS_UPDATE_RESERVE_SPACE    0x0008
 359         /* caller is from the direct IO path, request to creation of an
 360         unitialized extents if not allocated, split the uninitialized
 361         extent if blocks has been preallocated already*/
 362 #define EXT4_GET_BLOCKS_DIO                     0x0010
 363 #define EXT4_GET_BLOCKS_CONVERT                 0x0020
 364 #define EXT4_GET_BLOCKS_DIO_CREATE_EXT          (EXT4_GET_BLOCKS_DIO|\
 365                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
 366         /* Convert extent to initialized after direct IO complete */
 367 #define EXT4_GET_BLOCKS_DIO_CONVERT_EXT         (EXT4_GET_BLOCKS_CONVERT|\
 368                                          EXT4_GET_BLOCKS_DIO_CREATE_EXT)
 369
 370 /*
 371  * ioctl commands
 372  */
 373 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
 374 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
 375 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
 376 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
 377 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
 378 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
 379 #ifdef CONFIG_JBD2_DEBUG
 380 #define EXT4_IOC_WAIT_FOR_READONLY      _IOR('f', 99, long)
 381 #endif
 382 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
 383 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
 384 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
 385 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
 386 #define EXT4_IOC_MIGRATE                _IO('f', 9)
 387  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
 388  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
 389 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
 390 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
 391
 392 /*
 393  * ioctl commands in 32 bit emulation
 394  */
 395 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
 396 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
 397 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
 398 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
 399 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
 400 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
 401 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
 402 #ifdef CONFIG_JBD2_DEBUG
 403 #define EXT4_IOC32_WAIT_FOR_READONLY    _IOR('f', 99, int)
 404 #endif
 405 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
 406 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
 407
 408
 409 /*
 410  *  Mount options
 411  */
 412 struct ext4_mount_options {
 413         unsigned long s_mount_opt;
 414         uid_t s_resuid;
 415         gid_t s_resgid;
 416         unsigned long s_commit_interval;
 417         u32 s_min_batch_time, s_max_batch_time;
 418 #ifdef CONFIG_QUOTA
 419         int s_jquota_fmt;
 420         char *s_qf_names[MAXQUOTAS];
 421 #endif
 422 };
 423
 424 /* Max physical block we can addres w/o extents */
 425 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
 426
 427 /*
 428  * Structure of an inode on the disk
 429  */
 430 struct ext4_inode {
 431         __le16  i_mode;         /* File mode */
 432         __le16  i_uid;          /* Low 16 bits of Owner Uid */
 433         __le32  i_size_lo;      /* Size in bytes */
 434         __le32  i_atime;        /* Access time */
 435         __le32  i_ctime;        /* Inode Change time */
 436         __le32  i_mtime;        /* Modification time */
 437         __le32  i_dtime;        /* Deletion Time */
 438         __le16  i_gid;          /* Low 16 bits of Group Id */
 439         __le16  i_links_count;  /* Links count */
 440         __le32  i_blocks_lo;    /* Blocks count */
 441         __le32  i_flags;        /* File flags */
 442         union {
 443                 struct {
 444                         __le32  l_i_version;
 445                 } linux1;
 446                 struct {
 447                         __u32  h_i_translator;
 448                 } hurd1;
 449                 struct {
 450                         __u32  m_i_reserved1;
 451                 } masix1;
 452         } osd1;                         /* OS dependent 1 */
 453         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
 454         __le32  i_generation;   /* File version (for NFS) */
 455         __le32  i_file_acl_lo;  /* File ACL */
 456         __le32  i_size_high;
 457         __le32  i_obso_faddr;   /* Obsoleted fragment address */
 458         union {
 459                 struct {
 460                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
 461                         __le16  l_i_file_acl_high;
 462                         __le16  l_i_uid_high;   /* these 2 fields */
 463                         __le16  l_i_gid_high;   /* were reserved2[0] */
 464                         __u32   l_i_reserved2;
 465                 } linux2;
 466                 struct {
 467                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 468                         __u16   h_i_mode_high;
 469                         __u16   h_i_uid_high;
 470                         __u16   h_i_gid_high;
 471                         __u32   h_i_author;
 472                 } hurd2;
 473                 struct {
 474                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 475                         __le16  m_i_file_acl_high;
 476                         __u32   m_i_reserved2[2];
 477                 } masix2;
 478         } osd2;                         /* OS dependent 2 */
 479         __le16  i_extra_isize;
 480         __le16  i_pad1;
 481         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
 482         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
 483         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
 484         __le32  i_crtime;       /* File Creation time */
 485         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
 486         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
 487 };
 488
 489 struct move_extent {
 490         __u32 reserved;         /* should be zero */
 491         __u32 donor_fd;         /* donor file descriptor */
 492         __u64 orig_start;       /* logical start offset in block for orig */
 493         __u64 donor_start;      /* logical start offset in block for donor */
 494         __u64 len;              /* block length to be moved */
 495         __u64 moved_len;        /* moved block length */
 496 };
 497
 498 #define EXT4_EPOCH_BITS 2
 499 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
 500 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
 501
 502 /*
 503  * Extended fields will fit into an inode if the filesystem was formatted
 504  * with large inodes (-I 256 or larger) and there are not currently any EAs
 505  * consuming all of the available space. For new inodes we always reserve
 506  * enough space for the kernel's known extended fields, but for inodes
 507  * created with an old kernel this might not have been the case. None of
 508  * the extended inode fields is critical for correct filesystem operation.
 509  * This macro checks if a certain field fits in the inode. Note that
 510  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
 511  */
 512 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
 513         ((offsetof(typeof(*ext4_inode), field) +        \
 514           sizeof((ext4_inode)->field))                  \
 515         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
 516             (einode)->i_extra_isize))                   \
 517
 518 static inline __le32 ext4_encode_extra_time(struct timespec *time)
 519 {
 520        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
 521                            time->tv_sec >> 32 : 0) |
 522                            ((time->tv_nsec << 2) & EXT4_NSEC_MASK));
 523 }
 524
 525 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
 526 {
 527        if (sizeof(time->tv_sec) > 4)
 528                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
 529                                << 32;
 530        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> 2;
 531 }
 532
 533 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
 534 do {                                                                           \
 535         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
 536         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 537                 (raw_inode)->xtime ## _extra =                                 \
 538                                 ext4_encode_extra_time(&(inode)->xtime);       \
 539 } while (0)
 540
 541 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
 542 do {                                                                           \
 543         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 544                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
 545         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 546                 (raw_inode)->xtime ## _extra =                                 \
 547                                 ext4_encode_extra_time(&(einode)->xtime);      \
 548 } while (0)
 549
 550 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
 551 do {                                                                           \
 552         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
 553         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 554                 ext4_decode_extra_time(&(inode)->xtime,                        \
 555                                        raw_inode->xtime ## _extra);            \
 556 } while (0)
 557
 558 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
 559 do {                                                                           \
 560         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 561                 (einode)->xtime.tv_sec =                                       \
 562                         (signed)le32_to_cpu((raw_inode)->xtime);               \
 563         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 564                 ext4_decode_extra_time(&(einode)->xtime,                       \
 565                                        raw_inode->xtime ## _extra);            \
 566 } while (0)
 567
 568 #define i_disk_version osd1.linux1.l_i_version
 569
 570 #if defined(__KERNEL__) || defined(__linux__)
 571 #define i_reserved1     osd1.linux1.l_i_reserved1
 572 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
 573 #define i_blocks_high   osd2.linux2.l_i_blocks_high
 574 #define i_uid_low       i_uid
 575 #define i_gid_low       i_gid
 576 #define i_uid_high      osd2.linux2.l_i_uid_high
 577 #define i_gid_high      osd2.linux2.l_i_gid_high
 578 #define i_reserved2     osd2.linux2.l_i_reserved2
 579
 580 #elif defined(__GNU__)
 581
 582 #define i_translator    osd1.hurd1.h_i_translator
 583 #define i_uid_high      osd2.hurd2.h_i_uid_high
 584 #define i_gid_high      osd2.hurd2.h_i_gid_high
 585 #define i_author        osd2.hurd2.h_i_author
 586
 587 #elif defined(__masix__)
 588
 589 #define i_reserved1     osd1.masix1.m_i_reserved1
 590 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
 591 #define i_reserved2     osd2.masix2.m_i_reserved2
 592
 593 #endif /* defined(__KERNEL__) || defined(__linux__) */
 594
 595 /*
 596  * storage for cached extent
 597  */
 598 struct ext4_ext_cache {
 599         ext4_fsblk_t    ec_start;
 600         ext4_lblk_t     ec_block;
 601         __u32           ec_len; /* must be 32bit to return holes */
 602         __u32           ec_type;
 603 };
 604
 605 /*
 606  * fourth extended file system inode data in memory
 607  */
 608 struct ext4_inode_info {
 609         __le32  i_data[15];     /* unconverted */
 610         __u32   i_flags;
 611         ext4_fsblk_t    i_file_acl;
 612         __u32   i_dtime;
 613
 614         /*
 615          * i_block_group is the number of the block group which contains
 616          * this file's inode.  Constant across the lifetime of the inode,
 617          * it is ued for making block allocation decisions - we try to
 618          * place a file's data blocks near its inode block, and new inodes
 619          * near to their parent directory's inode.
 620          */
 621         ext4_group_t    i_block_group;
 622         __u32   i_state;                /* Dynamic state flags for ext4 */
 623
 624         ext4_lblk_t             i_dir_start_lookup;
 625 #ifdef CONFIG_EXT4_FS_XATTR
 626         /*
 627          * Extended attributes can be read independently of the main file
 628          * data. Taking i_mutex even when reading would cause contention
 629          * between readers of EAs and writers of regular file data, so
 630          * instead we synchronize on xattr_sem when reading or changing
 631          * EAs.
 632          */
 633         struct rw_semaphore xattr_sem;
 634 #endif
 635
 636         struct list_head i_orphan;      /* unlinked but open inodes */
 637
 638         /*
 639          * i_disksize keeps track of what the inode size is ON DISK, not
 640          * in memory.  During truncate, i_size is set to the new size by
 641          * the VFS prior to calling ext4_truncate(), but the filesystem won't
 642          * set i_disksize to 0 until the truncate is actually under way.
 643          *
 644          * The intent is that i_disksize always represents the blocks which
 645          * are used by this file.  This allows recovery to restart truncate
 646          * on orphans if we crash during truncate.  We actually write i_disksize
 647          * into the on-disk inode when writing inodes out, instead of i_size.
 648          *
 649          * The only time when i_disksize and i_size may be different is when
 650          * a truncate is in progress.  The only things which change i_disksize
 651          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
 652          */
 653         loff_t  i_disksize;
 654
 655         /*
 656          * i_data_sem is for serialising ext4_truncate() against
 657          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
 658          * data tree are chopped off during truncate. We can't do that in
 659          * ext4 because whenever we perform intermediate commits during
 660          * truncate, the inode and all the metadata blocks *must* be in a
 661          * consistent state which allows truncation of the orphans to restart
 662          * during recovery.  Hence we must fix the get_block-vs-truncate race
 663          * by other means, so we have i_data_sem.
 664          */
 665         struct rw_semaphore i_data_sem;
 666         struct inode vfs_inode;
 667         struct jbd2_inode jinode;
 668
 669         struct ext4_ext_cache i_cached_extent;
 670         /*
 671          * File creation time. Its function is same as that of
 672          * struct timespec i_{a,c,m}time in the generic inode.
 673          */
 674         struct timespec i_crtime;
 675
 676         /* mballoc */
 677         struct list_head i_prealloc_list;
 678         spinlock_t i_prealloc_lock;
 679
 680         /* ialloc */
 681         ext4_group_t    i_last_alloc_group;
 682
 683         /* allocation reservation info for delalloc */
 684         unsigned int i_reserved_data_blocks;
 685         unsigned int i_reserved_meta_blocks;
 686         unsigned int i_allocated_meta_blocks;
 687         unsigned short i_delalloc_reserved_flag;
 688
 689         /* on-disk additional length */
 690         __u16 i_extra_isize;
 691
 692         spinlock_t i_block_reservation_lock;
 693 };
 694
 695 /*
 696  * File system states
 697  */
 698 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
 699 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
 700 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
 701
 702 /*
 703  * Misc. filesystem flags
 704  */
 705 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
 706 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
 707 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
 708
 709 /*
 710  * Mount flags
 711  */
 712 #define EXT4_MOUNT_OLDALLOC             0x00002  /* Don't use the new Orlov allocator */
 713 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
 714 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
 715 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
 716 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
 717 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
 718 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
 719 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
 720 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
 721 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
 722 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
 723 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
 724 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
 725 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
 726 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
 727 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
 728 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
 729 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
 730 #define EXT4_MOUNT_NOBH                 0x40000 /* No bufferheads */
 731 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
 732 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
 733 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
 734 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
 735 #define EXT4_MOUNT_I_VERSION            0x2000000 /* i_version support */
 736 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
 737 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
 738 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
 739
 740 #define clear_opt(o, opt)               o &= ~EXT4_MOUNT_##opt
 741 #define set_opt(o, opt)                 o |= EXT4_MOUNT_##opt
 742 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
 743                                          EXT4_MOUNT_##opt)
 744
 745 #define ext4_set_bit                    ext2_set_bit
 746 #define ext4_set_bit_atomic             ext2_set_bit_atomic
 747 #define ext4_clear_bit                  ext2_clear_bit
 748 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
 749 #define ext4_test_bit                   ext2_test_bit
 750 #define ext4_find_first_zero_bit        ext2_find_first_zero_bit
 751 #define ext4_find_next_zero_bit         ext2_find_next_zero_bit
 752 #define ext4_find_next_bit              ext2_find_next_bit
 753
 754 /*
 755  * Maximal mount counts between two filesystem checks
 756  */
 757 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
 758 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
 759
 760 /*
 761  * Behaviour when detecting errors
 762  */
 763 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
 764 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
 765 #define EXT4_ERRORS_PANIC               3       /* Panic */
 766 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
 767
 768 /*
 769  * Structure of the super block
 770  */
 771 struct ext4_super_block {
 772 /*00*/  __le32  s_inodes_count;         /* Inodes count */
 773         __le32  s_blocks_count_lo;      /* Blocks count */
 774         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
 775         __le32  s_free_blocks_count_lo; /* Free blocks count */
 776 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
 777         __le32  s_first_data_block;     /* First Data Block */
 778         __le32  s_log_block_size;       /* Block size */
 779         __le32  s_obso_log_frag_size;   /* Obsoleted fragment size */
 780 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
 781         __le32  s_obso_frags_per_group; /* Obsoleted fragments per group */
 782         __le32  s_inodes_per_group;     /* # Inodes per group */
 783         __le32  s_mtime;                /* Mount time */
 784 /*30*/  __le32  s_wtime;                /* Write time */
 785         __le16  s_mnt_count;            /* Mount count */
 786         __le16  s_max_mnt_count;        /* Maximal mount count */
 787         __le16  s_magic;                /* Magic signature */
 788         __le16  s_state;                /* File system state */
 789         __le16  s_errors;               /* Behaviour when detecting errors */
 790         __le16  s_minor_rev_level;      /* minor revision level */
 791 /*40*/  __le32  s_lastcheck;            /* time of last check */
 792         __le32  s_checkinterval;        /* max. time between checks */
 793         __le32  s_creator_os;           /* OS */
 794         __le32  s_rev_level;            /* Revision level */
 795 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
 796         __le16  s_def_resgid;           /* Default gid for reserved blocks */
 797         /*
 798          * These fields are for EXT4_DYNAMIC_REV superblocks only.
 799          *
 800          * Note: the difference between the compatible feature set and
 801          * the incompatible feature set is that if there is a bit set
 802          * in the incompatible feature set that the kernel doesn't
 803          * know about, it should refuse to mount the filesystem.
 804          *
 805          * e2fsck's requirements are more strict; if it doesn't know
 806          * about a feature in either the compatible or incompatible
 807          * feature set, it must abort and not try to meddle with
 808          * things it doesn't understand...
 809          */
 810         __le32  s_first_ino;            /* First non-reserved inode */
 811         __le16  s_inode_size;           /* size of inode structure */
 812         __le16  s_block_group_nr;       /* block group # of this superblock */
 813         __le32  s_feature_compat;       /* compatible feature set */
 814 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
 815         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
 816 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
 817 /*78*/  char    s_volume_name[16];      /* volume name */
 818 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
 819 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
 820         /*
 821          * Performance hints.  Directory preallocation should only
 822          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
 823          */
 824         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
 825         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
 826         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
 827         /*
 828          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
 829          */
 830 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
 831 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
 832         __le32  s_journal_dev;          /* device number of journal file */
 833         __le32  s_last_orphan;          /* start of list of inodes to delete */
 834         __le32  s_hash_seed[4];         /* HTREE hash seed */
 835         __u8    s_def_hash_version;     /* Default hash version to use */
 836         __u8    s_reserved_char_pad;
 837         __le16  s_desc_size;            /* size of group descriptor */
 838 /*100*/ __le32  s_default_mount_opts;
 839         __le32  s_first_meta_bg;        /* First metablock block group */
 840         __le32  s_mkfs_time;            /* When the filesystem was created */
 841         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
 842         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
 843 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
 844         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
 845         __le32  s_free_blocks_count_hi; /* Free blocks count */
 846         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
 847         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
 848         __le32  s_flags;                /* Miscellaneous flags */
 849         __le16  s_raid_stride;          /* RAID stride */
 850         __le16  s_mmp_interval;         /* # seconds to wait in MMP checking */
 851         __le64  s_mmp_block;            /* Block for multi-mount protection */
 852         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
 853         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
 854         __u8    s_reserved_char_pad2;
 855         __le16  s_reserved_pad;
 856         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
 857         __u32   s_reserved[160];        /* Padding to the end of the block */
 858 };
 859
 860 #ifdef __KERNEL__
 861
 862 /*
 863  * run-time mount flags
 864  */
 865 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
 866 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
 867
 868 /*
 869  * fourth extended-fs super-block data in memory
 870  */
 871 struct ext4_sb_info {
 872         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
 873         unsigned long s_inodes_per_block;/* Number of inodes per block */
 874         unsigned long s_blocks_per_group;/* Number of blocks in a group */
 875         unsigned long s_inodes_per_group;/* Number of inodes in a group */
 876         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
 877         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
 878         unsigned long s_desc_per_block; /* Number of group descriptors per block */
 879         ext4_group_t s_groups_count;    /* Number of groups in the fs */
 880         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
 881         unsigned long s_overhead_last;  /* Last calculated overhead */
 882         unsigned long s_blocks_last;    /* Last seen block count */
 883         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
 884         struct buffer_head * s_sbh;     /* Buffer containing the super block */
 885         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
 886         struct buffer_head **s_group_desc;
 887         unsigned int s_mount_opt;
 888         unsigned int s_mount_flags;
 889         ext4_fsblk_t s_sb_block;
 890         uid_t s_resuid;
 891         gid_t s_resgid;
 892         unsigned short s_mount_state;
 893         unsigned short s_pad;
 894         int s_addr_per_block_bits;
 895         int s_desc_per_block_bits;
 896         int s_inode_size;
 897         int s_first_ino;
 898         unsigned int s_inode_readahead_blks;
 899         unsigned int s_inode_goal;
 900         spinlock_t s_next_gen_lock;
 901         u32 s_next_generation;
 902         u32 s_hash_seed[4];
 903         int s_def_hash_version;
 904         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
 905         struct percpu_counter s_freeblocks_counter;
 906         struct percpu_counter s_freeinodes_counter;
 907         struct percpu_counter s_dirs_counter;
 908         struct percpu_counter s_dirtyblocks_counter;
 909         struct blockgroup_lock *s_blockgroup_lock;
 910         struct proc_dir_entry *s_proc;
 911         struct kobject s_kobj;
 912         struct completion s_kobj_unregister;
 913
 914         /* Journaling */
 915         struct inode *s_journal_inode;
 916         struct journal_s *s_journal;
 917         struct list_head s_orphan;
 918         struct mutex s_orphan_lock;
 919         struct mutex s_resize_lock;
 920         unsigned long s_commit_interval;
 921         u32 s_max_batch_time;
 922         u32 s_min_batch_time;
 923         struct block_device *journal_bdev;
 924 #ifdef CONFIG_JBD2_DEBUG
 925         struct timer_list turn_ro_timer;        /* For turning read-only (crash simulation) */
 926         wait_queue_head_t ro_wait_queue;        /* For people waiting for the fs to go read-only */
 927 #endif
 928 #ifdef CONFIG_QUOTA
 929         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
 930         int s_jquota_fmt;                       /* Format of quota to use */
 931 #endif
 932         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
 933         struct rb_root system_blks;
 934
 935 #ifdef EXTENTS_STATS
 936         /* ext4 extents stats */
 937         unsigned long s_ext_min;
 938         unsigned long s_ext_max;
 939         unsigned long s_depth_max;
 940         spinlock_t s_ext_stats_lock;
 941         unsigned long s_ext_blocks;
 942         unsigned long s_ext_extents;
 943 #endif
 944
 945         /* for buddy allocator */
 946         struct ext4_group_info ***s_group_info;
 947         struct inode *s_buddy_cache;
 948         long s_blocks_reserved;
 949         spinlock_t s_reserve_lock;
 950         spinlock_t s_md_lock;
 951         tid_t s_last_transaction;
 952         unsigned short *s_mb_offsets;
 953         unsigned int *s_mb_maxs;
 954
 955         /* tunables */
 956         unsigned long s_stripe;
 957         unsigned int s_mb_stream_request;
 958         unsigned int s_mb_max_to_scan;
 959         unsigned int s_mb_min_to_scan;
 960         unsigned int s_mb_stats;
 961         unsigned int s_mb_order2_reqs;
 962         unsigned int s_mb_group_prealloc;
 963         unsigned int s_max_writeback_mb_bump;
 964         /* where last allocation was done - for stream allocation */
 965         unsigned long s_mb_last_group;
 966         unsigned long s_mb_last_start;
 967
 968         /* history to debug policy */
 969         struct ext4_mb_history *s_mb_history;
 970         int s_mb_history_cur;
 971         int s_mb_history_max;
 972         int s_mb_history_num;
 973         spinlock_t s_mb_history_lock;
 974         int s_mb_history_filter;
 975
 976         /* stats for buddy allocator */
 977         spinlock_t s_mb_pa_lock;
 978         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
 979         atomic_t s_bal_success; /* we found long enough chunks */
 980         atomic_t s_bal_allocated;       /* in blocks */
 981         atomic_t s_bal_ex_scanned;      /* total extents scanned */
 982         atomic_t s_bal_goals;   /* goal hits */
 983         atomic_t s_bal_breaks;  /* too long searches */
 984         atomic_t s_bal_2orders; /* 2^order hits */
 985         spinlock_t s_bal_lock;
 986         unsigned long s_mb_buddies_generated;
 987         unsigned long long s_mb_generation_time;
 988         atomic_t s_mb_lost_chunks;
 989         atomic_t s_mb_preallocated;
 990         atomic_t s_mb_discarded;
 991         atomic_t s_lock_busy;
 992
 993         /* locality groups */
 994         struct ext4_locality_group *s_locality_groups;
 995
 996         /* for write statistics */
 997         unsigned long s_sectors_written_start;
 998         u64 s_kbytes_written;
 999
1000         unsigned int s_log_groups_per_flex;
1001         struct flex_groups *s_flex_groups;
1002 };
1003
1004 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1005 {
1006         return sb->s_fs_info;
1007 }
1008 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1009 {
1010         return container_of(inode, struct ext4_inode_info, vfs_inode);
1011 }
1012
1013 static inline struct timespec ext4_current_time(struct inode *inode)
1014 {
1015         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1016                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1017 }
1018
1019 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1020 {
1021         return ino == EXT4_ROOT_INO ||
1022                 ino == EXT4_JOURNAL_INO ||
1023                 ino == EXT4_RESIZE_INO ||
1024                 (ino >= EXT4_FIRST_INO(sb) &&
1025                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1026 }
1027 #else
1028 /* Assume that user mode programs are passing in an ext4fs superblock, not
1029  * a kernel struct super_block.  This will allow us to call the feature-test
1030  * macros from user land. */
1031 #define EXT4_SB(sb)     (sb)
1032 #endif
1033
1034 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1035
1036 /*
1037  * Codes for operating systems
1038  */
1039 #define EXT4_OS_LINUX           0
1040 #define EXT4_OS_HURD            1
1041 #define EXT4_OS_MASIX           2
1042 #define EXT4_OS_FREEBSD         3
1043 #define EXT4_OS_LITES           4
1044
1045 /*
1046  * Revision levels
1047  */
1048 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1049 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1050
1051 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1052 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1053
1054 #define EXT4_GOOD_OLD_INODE_SIZE 128
1055
1056 /*
1057  * Feature set definitions
1058  */
1059
1060 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1061         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1062 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1063         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1064 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1065         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1066 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1067         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1068 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1069         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1070 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1071         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1072 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1073         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1074 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1075         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1076 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1077         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1078
1079 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1080 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1081 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1082 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1083 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1084 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1085
1086 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1087 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1088 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1089 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1090 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1091 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1092 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1093
1094 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1095 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1096 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1097 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1098 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1099 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1100 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1101 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1102 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1103
1104 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1105 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1106                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1107                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1108                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1109                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1110                                          EXT4_FEATURE_INCOMPAT_FLEX_BG)
1111 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1112                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1113                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1114                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1115                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1116                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1117                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE)
1118
1119 /*
1120  * Default values for user and/or group using reserved blocks
1121  */
1122 #define EXT4_DEF_RESUID         0
1123 #define EXT4_DEF_RESGID         0
1124
1125 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1126
1127 /*
1128  * Default mount options
1129  */
1130 #define EXT4_DEFM_DEBUG         0x0001
1131 #define EXT4_DEFM_BSDGROUPS     0x0002
1132 #define EXT4_DEFM_XATTR_USER    0x0004
1133 #define EXT4_DEFM_ACL           0x0008
1134 #define EXT4_DEFM_UID16         0x0010
1135 #define EXT4_DEFM_JMODE         0x0060
1136 #define EXT4_DEFM_JMODE_DATA    0x0020
1137 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1138 #define EXT4_DEFM_JMODE_WBACK   0x0060
1139
1140 /*
1141  * Default journal batch times
1142  */
1143 #define EXT4_DEF_MIN_BATCH_TIME 0
1144 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1145
1146 /*
1147  * Minimum number of groups in a flexgroup before we separate out
1148  * directories into the first block group of a flexgroup
1149  */
1150 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1151
1152 /*
1153  * Structure of a directory entry
1154  */
1155 #define EXT4_NAME_LEN 255
1156
1157 struct ext4_dir_entry {
1158         __le32  inode;                  /* Inode number */
1159         __le16  rec_len;                /* Directory entry length */
1160         __le16  name_len;               /* Name length */
1161         char    name[EXT4_NAME_LEN];    /* File name */
1162 };
1163
1164 /*
1165  * The new version of the directory entry.  Since EXT4 structures are
1166  * stored in intel byte order, and the name_len field could never be
1167  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1168  * file_type field.
1169  */
1170 struct ext4_dir_entry_2 {
1171         __le32  inode;                  /* Inode number */
1172         __le16  rec_len;                /* Directory entry length */
1173         __u8    name_len;               /* Name length */
1174         __u8    file_type;
1175         char    name[EXT4_NAME_LEN];    /* File name */
1176 };
1177
1178 /*
1179  * Ext4 directory file types.  Only the low 3 bits are used.  The
1180  * other bits are reserved for now.
1181  */
1182 #define EXT4_FT_UNKNOWN         0
1183 #define EXT4_FT_REG_FILE        1
1184 #define EXT4_FT_DIR             2
1185 #define EXT4_FT_CHRDEV          3
1186 #define EXT4_FT_BLKDEV          4
1187 #define EXT4_FT_FIFO            5
1188 #define EXT4_FT_SOCK            6
1189 #define EXT4_FT_SYMLINK         7
1190
1191 #define EXT4_FT_MAX             8
1192
1193 /*
1194  * EXT4_DIR_PAD defines the directory entries boundaries
1195  *
1196  * NOTE: It must be a multiple of 4
1197  */
1198 #define EXT4_DIR_PAD                    4
1199 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1200 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1201                                          ~EXT4_DIR_ROUND)
1202 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1203
1204 /*
1205  * Hash Tree Directory indexing
1206  * (c) Daniel Phillips, 2001
1207  */
1208
1209 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1210                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1211                       (EXT4_I(dir)->i_flags & EXT4_INDEX_FL))
1212 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1213 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1214
1215 /* Legal values for the dx_root hash_version field: */
1216
1217 #define DX_HASH_LEGACY          0
1218 #define DX_HASH_HALF_MD4        1
1219 #define DX_HASH_TEA             2
1220 #define DX_HASH_LEGACY_UNSIGNED 3
1221 #define DX_HASH_HALF_MD4_UNSIGNED       4
1222 #define DX_HASH_TEA_UNSIGNED            5
1223
1224 #ifdef __KERNEL__
1225
1226 /* hash info structure used by the directory hash */
1227 struct dx_hash_info
1228 {
1229         u32             hash;
1230         u32             minor_hash;
1231         int             hash_version;
1232         u32             *seed;
1233 };
1234
1235 #define EXT4_HTREE_EOF  0x7fffffff
1236
1237 /*
1238  * Control parameters used by ext4_htree_next_block
1239  */
1240 #define HASH_NB_ALWAYS          1
1241
1242
1243 /*
1244  * Describe an inode's exact location on disk and in memory
1245  */
1246 struct ext4_iloc
1247 {
1248         struct buffer_head *bh;
1249         unsigned long offset;
1250         ext4_group_t block_group;
1251 };
1252
1253 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1254 {
1255         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1256 }
1257
1258 /*
1259  * This structure is stuffed into the struct file's private_data field
1260  * for directories.  It is where we put information so that we can do
1261  * readdir operations in hash tree order.
1262  */
1263 struct dir_private_info {
1264         struct rb_root  root;
1265         struct rb_node  *curr_node;
1266         struct fname    *extra_fname;
1267         loff_t          last_pos;
1268         __u32           curr_hash;
1269         __u32           curr_minor_hash;
1270         __u32           next_hash;
1271 };
1272
1273 /* calculate the first block number of the group */
1274 static inline ext4_fsblk_t
1275 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1276 {
1277         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1278                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1279 }
1280
1281 /*
1282  * Special error return code only used by dx_probe() and its callers.
1283  */
1284 #define ERR_BAD_DX_DIR  -75000
1285
1286 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1287                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1288
1289 extern struct proc_dir_entry *ext4_proc_root;
1290
1291 /*
1292  * Function prototypes
1293  */
1294
1295 /*
1296  * Ok, these declarations are also in <linux/kernel.h> but none of the
1297  * ext4 source programs needs to include it so they are duplicated here.
1298  */
1299 # define NORET_TYPE     /**/
1300 # define ATTRIB_NORET   __attribute__((noreturn))
1301 # define NORET_AND      noreturn,
1302
1303 /* bitmap.c */
1304 extern unsigned int ext4_count_free(struct buffer_head *, unsigned);
1305
1306 /* balloc.c */
1307 extern unsigned int ext4_block_group(struct super_block *sb,
1308                         ext4_fsblk_t blocknr);
1309 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1310                         ext4_fsblk_t blocknr);
1311 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1312 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1313                         ext4_group_t group);
1314 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1315                         ext4_fsblk_t goal, unsigned long *count, int *errp);
1316 extern int ext4_claim_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1317 extern int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1318 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1319                         ext4_fsblk_t block, unsigned long count, int metadata);
1320 extern void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
1321                                 ext4_fsblk_t block, unsigned long count);
1322 extern ext4_fsblk_t ext4_count_free_blocks(struct super_block *);
1323 extern void ext4_check_blocks_bitmap(struct super_block *);
1324 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1325                                                     ext4_group_t block_group,
1326                                                     struct buffer_head ** bh);
1327 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1328 struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1329                                       ext4_group_t block_group);
1330 extern unsigned ext4_init_block_bitmap(struct super_block *sb,
1331                                        struct buffer_head *bh,
1332                                        ext4_group_t group,
1333                                        struct ext4_group_desc *desc);
1334 #define ext4_free_blocks_after_init(sb, group, desc)                    \
1335                 ext4_init_block_bitmap(sb, NULL, group, desc)
1336
1337 /* dir.c */
1338 extern int ext4_check_dir_entry(const char *, struct inode *,
1339                                 struct ext4_dir_entry_2 *,
1340                                 struct buffer_head *, unsigned int);
1341 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1342                                     __u32 minor_hash,
1343                                     struct ext4_dir_entry_2 *dirent);
1344 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1345
1346 /* fsync.c */
1347 extern int ext4_sync_file(struct file *, struct dentry *, int);
1348
1349 /* hash.c */
1350 extern int ext4fs_dirhash(const char *name, int len, struct
1351                           dx_hash_info *hinfo);
1352
1353 /* ialloc.c */
1354 extern struct inode *ext4_new_inode(handle_t *, struct inode *, int,
1355                                     const struct qstr *qstr, __u32 goal);
1356 extern void ext4_free_inode(handle_t *, struct inode *);
1357 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1358 extern unsigned long ext4_count_free_inodes(struct super_block *);
1359 extern unsigned long ext4_count_dirs(struct super_block *);
1360 extern void ext4_check_inodes_bitmap(struct super_block *);
1361 extern unsigned ext4_init_inode_bitmap(struct super_block *sb,
1362                                        struct buffer_head *bh,
1363                                        ext4_group_t group,
1364                                        struct ext4_group_desc *desc);
1365 extern void mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1366
1367 /* mballoc.c */
1368 extern long ext4_mb_stats;
1369 extern long ext4_mb_max_to_scan;
1370 extern int ext4_mb_init(struct super_block *, int);
1371 extern int ext4_mb_release(struct super_block *);
1372 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1373                                 struct ext4_allocation_request *, int *);
1374 extern int ext4_mb_reserve_blocks(struct super_block *, int);
1375 extern void ext4_discard_preallocations(struct inode *);
1376 extern int __init init_ext4_mballoc(void);
1377 extern void exit_ext4_mballoc(void);
1378 extern void ext4_mb_free_blocks(handle_t *, struct inode *,
1379                 ext4_fsblk_t, unsigned long, int, unsigned long *);
1380 extern int ext4_mb_add_groupinfo(struct super_block *sb,
1381                 ext4_group_t i, struct ext4_group_desc *desc);
1382 extern int ext4_mb_get_buddy_cache_lock(struct super_block *, ext4_group_t);
1383 extern void ext4_mb_put_buddy_cache_lock(struct super_block *,
1384                                                 ext4_group_t, int);
1385 /* inode.c */
1386 int ext4_forget(handle_t *handle, int is_metadata, struct inode *inode,
1387                 struct buffer_head *bh, ext4_fsblk_t blocknr);
1388 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1389                                                 ext4_lblk_t, int, int *);
1390 struct buffer_head *ext4_bread(handle_t *, struct inode *,
1391                                                 ext4_lblk_t, int, int *);
1392 int ext4_get_block(struct inode *inode, sector_t iblock,
1393                                 struct buffer_head *bh_result, int create);
1394
1395 extern struct inode *ext4_iget(struct super_block *, unsigned long);
1396 extern int  ext4_write_inode(struct inode *, int);
1397 extern int  ext4_setattr(struct dentry *, struct iattr *);
1398 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1399                                 struct kstat *stat);
1400 extern void ext4_delete_inode(struct inode *);
1401 extern int  ext4_sync_inode(handle_t *, struct inode *);
1402 extern void ext4_dirty_inode(struct inode *);
1403 extern int ext4_change_inode_journal_flag(struct inode *, int);
1404 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1405 extern int ext4_can_truncate(struct inode *inode);
1406 extern void ext4_truncate(struct inode *);
1407 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
1408 extern void ext4_set_inode_flags(struct inode *);
1409 extern void ext4_get_inode_flags(struct ext4_inode_info *);
1410 extern int ext4_alloc_da_blocks(struct inode *inode);
1411 extern void ext4_set_aops(struct inode *inode);
1412 extern int ext4_writepage_trans_blocks(struct inode *);
1413 extern int ext4_meta_trans_blocks(struct inode *, int nrblocks, int idxblocks);
1414 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1415 extern int ext4_block_truncate_page(handle_t *handle,
1416                 struct address_space *mapping, loff_t from);
1417 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1418 extern qsize_t ext4_get_reserved_space(struct inode *inode);
1419
1420 /* ioctl.c */
1421 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1422 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1423
1424 /* migrate.c */
1425 extern int ext4_ext_migrate(struct inode *);
1426
1427 /* namei.c */
1428 extern unsigned int ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize);
1429 extern __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize);
1430 extern int ext4_orphan_add(handle_t *, struct inode *);
1431 extern int ext4_orphan_del(handle_t *, struct inode *);
1432 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1433                                 __u32 start_minor_hash, __u32 *next_hash);
1434
1435 /* resize.c */
1436 extern int ext4_group_add(struct super_block *sb,
1437                                 struct ext4_new_group_data *input);
1438 extern int ext4_group_extend(struct super_block *sb,
1439                                 struct ext4_super_block *es,
1440                                 ext4_fsblk_t n_blocks_count);
1441
1442 /* super.c */
1443 extern void ext4_error(struct super_block *, const char *, const char *, ...)
1444         __attribute__ ((format (printf, 3, 4)));
1445 extern void __ext4_std_error(struct super_block *, const char *, int);
1446 extern void ext4_abort(struct super_block *, const char *, const char *, ...)
1447         __attribute__ ((format (printf, 3, 4)));
1448 extern void ext4_warning(struct super_block *, const char *, const char *, ...)
1449         __attribute__ ((format (printf, 3, 4)));
1450 extern void ext4_msg(struct super_block *, const char *, const char *, ...)
1451         __attribute__ ((format (printf, 3, 4)));
1452 extern void ext4_grp_locked_error(struct super_block *, ext4_group_t,
1453                                 const char *, const char *, ...)
1454         __attribute__ ((format (printf, 4, 5)));
1455 extern void ext4_update_dynamic_rev(struct super_block *sb);
1456 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1457                                         __u32 compat);
1458 extern int ext4_update_rocompat_feature(handle_t *handle,
1459                                         struct super_block *sb, __u32 rocompat);
1460 extern int ext4_update_incompat_feature(handle_t *handle,
1461                                         struct super_block *sb, __u32 incompat);
1462 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1463                                       struct ext4_group_desc *bg);
1464 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1465                                       struct ext4_group_desc *bg);
1466 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1467                                      struct ext4_group_desc *bg);
1468 extern __u32 ext4_free_blks_count(struct super_block *sb,
1469                                 struct ext4_group_desc *bg);
1470 extern __u32 ext4_free_inodes_count(struct super_block *sb,
1471                                  struct ext4_group_desc *bg);
1472 extern __u32 ext4_used_dirs_count(struct super_block *sb,
1473                                 struct ext4_group_desc *bg);
1474 extern __u32 ext4_itable_unused_count(struct super_block *sb,
1475                                    struct ext4_group_desc *bg);
1476 extern void ext4_block_bitmap_set(struct super_block *sb,
1477                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1478 extern void ext4_inode_bitmap_set(struct super_block *sb,
1479                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1480 extern void ext4_inode_table_set(struct super_block *sb,
1481                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
1482 extern void ext4_free_blks_set(struct super_block *sb,
1483                                struct ext4_group_desc *bg, __u32 count);
1484 extern void ext4_free_inodes_set(struct super_block *sb,
1485                                 struct ext4_group_desc *bg, __u32 count);
1486 extern void ext4_used_dirs_set(struct super_block *sb,
1487                                 struct ext4_group_desc *bg, __u32 count);
1488 extern void ext4_itable_unused_set(struct super_block *sb,
1489                                    struct ext4_group_desc *bg, __u32 count);
1490 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
1491                                    struct ext4_group_desc *gdp);
1492 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
1493                                        struct ext4_group_desc *gdp);
1494
1495 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
1496 {
1497         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
1498                 le32_to_cpu(es->s_blocks_count_lo);
1499 }
1500
1501 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
1502 {
1503         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
1504                 le32_to_cpu(es->s_r_blocks_count_lo);
1505 }
1506
1507 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
1508 {
1509         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
1510                 le32_to_cpu(es->s_free_blocks_count_lo);
1511 }
1512
1513 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
1514                                          ext4_fsblk_t blk)
1515 {
1516         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
1517         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
1518 }
1519
1520 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
1521                                               ext4_fsblk_t blk)
1522 {
1523         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
1524         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
1525 }
1526
1527 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
1528                                            ext4_fsblk_t blk)
1529 {
1530         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
1531         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
1532 }
1533
1534 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
1535 {
1536         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
1537                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
1538                         le32_to_cpu(raw_inode->i_size_lo);
1539         else
1540                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
1541 }
1542
1543 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
1544 {
1545         raw_inode->i_size_lo = cpu_to_le32(i_size);
1546         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
1547 }
1548
1549 static inline
1550 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
1551                                             ext4_group_t group)
1552 {
1553          struct ext4_group_info ***grp_info;
1554          long indexv, indexh;
1555          grp_info = EXT4_SB(sb)->s_group_info;
1556          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
1557          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
1558          return grp_info[indexv][indexh];
1559 }
1560
1561 /*
1562  * Reading s_groups_count requires using smp_rmb() afterwards.  See
1563  * the locking protocol documented in the comments of ext4_group_add()
1564  * in resize.c
1565  */
1566 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
1567 {
1568         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
1569
1570         smp_rmb();
1571         return ngroups;
1572 }
1573
1574 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
1575                                              ext4_group_t block_group)
1576 {
1577         return block_group >> sbi->s_log_groups_per_flex;
1578 }
1579
1580 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
1581 {
1582         return 1 << sbi->s_log_groups_per_flex;
1583 }
1584
1585 #define ext4_std_error(sb, errno)                               \
1586 do {                                                            \
1587         if ((errno))                                            \
1588                 __ext4_std_error((sb), __func__, (errno));      \
1589 } while (0)
1590
1591 #ifdef CONFIG_SMP
1592 /* Each CPU can accumulate percpu_counter_batch blocks in their local
1593  * counters. So we need to make sure we have free blocks more
1594  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
1595  */
1596 #define EXT4_FREEBLOCKS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
1597 #else
1598 #define EXT4_FREEBLOCKS_WATERMARK 0
1599 #endif
1600
1601 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
1602 {
1603         /*
1604          * XXX: replace with spinlock if seen contended -bzzz
1605          */
1606         down_write(&EXT4_I(inode)->i_data_sem);
1607         if (newsize > EXT4_I(inode)->i_disksize)
1608                 EXT4_I(inode)->i_disksize = newsize;
1609         up_write(&EXT4_I(inode)->i_data_sem);
1610         return ;
1611 }
1612
1613 struct ext4_group_info {
1614         unsigned long   bb_state;
1615         struct rb_root  bb_free_root;
1616         ext4_grpblk_t   bb_first_free;  /* first free block */
1617         ext4_grpblk_t   bb_free;        /* total free blocks */
1618         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
1619         struct          list_head bb_prealloc_list;
1620 #ifdef DOUBLE_CHECK
1621         void            *bb_bitmap;
1622 #endif
1623         struct rw_semaphore alloc_sem;
1624         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
1625                                          * regions, index is order.
1626                                          * bb_counters[3] = 5 means
1627                                          * 5 free 8-block regions. */
1628 };
1629
1630 #define EXT4_GROUP_INFO_NEED_INIT_BIT   0
1631
1632 #define EXT4_MB_GRP_NEED_INIT(grp)      \
1633         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
1634
1635 #define EXT4_MAX_CONTENTION             8
1636 #define EXT4_CONTENTION_THRESHOLD       2
1637
1638 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
1639                                               ext4_group_t group)
1640 {
1641         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
1642 }
1643
1644 /*
1645  * Returns true if the filesystem is busy enough that attempts to
1646  * access the block group locks has run into contention.
1647  */
1648 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
1649 {
1650         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
1651 }
1652
1653 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
1654 {
1655         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
1656         if (spin_trylock(lock))
1657                 /*
1658                  * We're able to grab the lock right away, so drop the
1659                  * lock contention counter.
1660                  */
1661                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
1662         else {
1663                 /*
1664                  * The lock is busy, so bump the contention counter,
1665                  * and then wait on the spin lock.
1666                  */
1667                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
1668                                   EXT4_MAX_CONTENTION);
1669                 spin_lock(lock);
1670         }
1671 }
1672
1673 static inline void ext4_unlock_group(struct super_block *sb,
1674                                         ext4_group_t group)
1675 {
1676         spin_unlock(ext4_group_lock_ptr(sb, group));
1677 }
1678
1679 /*
1680  * Inodes and files operations
1681  */
1682
1683 /* dir.c */
1684 extern const struct file_operations ext4_dir_operations;
1685
1686 /* file.c */
1687 extern const struct inode_operations ext4_file_inode_operations;
1688 extern const struct file_operations ext4_file_operations;
1689
1690 /* namei.c */
1691 extern const struct inode_operations ext4_dir_inode_operations;
1692 extern const struct inode_operations ext4_special_inode_operations;
1693 extern struct dentry *ext4_get_parent(struct dentry *child);
1694
1695 /* symlink.c */
1696 extern const struct inode_operations ext4_symlink_inode_operations;
1697 extern const struct inode_operations ext4_fast_symlink_inode_operations;
1698
1699 /* block_validity */
1700 extern void ext4_release_system_zone(struct super_block *sb);
1701 extern int ext4_setup_system_zone(struct super_block *sb);
1702 extern int __init init_ext4_system_zone(void);
1703 extern void exit_ext4_system_zone(void);
1704 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
1705                                  ext4_fsblk_t start_blk,
1706                                  unsigned int count);
1707
1708 /* extents.c */
1709 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
1710 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
1711 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
1712                                        int chunk);
1713 extern int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
1714                                ext4_lblk_t iblock, unsigned int max_blocks,
1715                                struct buffer_head *bh_result, int flags);
1716 extern void ext4_ext_truncate(struct inode *);
1717 extern void ext4_ext_init(struct super_block *);
1718 extern void ext4_ext_release(struct super_block *);
1719 extern long ext4_fallocate(struct inode *inode, int mode, loff_t offset,
1720                           loff_t len);
1721 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
1722                           loff_t len);
1723 extern int ext4_get_blocks(handle_t *handle, struct inode *inode,
1724                            sector_t block, unsigned int max_blocks,
1725                            struct buffer_head *bh, int flags);
1726 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1727                         __u64 start, __u64 len);
1728 /* move_extent.c */
1729 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
1730                              __u64 start_orig, __u64 start_donor,
1731                              __u64 len, __u64 *moved_len);
1732
1733
1734 /*
1735  * Add new method to test wether block and inode bitmaps are properly
1736  * initialized. With uninit_bg reading the block from disk is not enough
1737  * to mark the bitmap uptodate. We need to also zero-out the bitmap
1738  */
1739 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
1740
1741 static inline int bitmap_uptodate(struct buffer_head *bh)
1742 {
1743         return (buffer_uptodate(bh) &&
1744                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
1745 }
1746 static inline void set_bitmap_uptodate(struct buffer_head *bh)
1747 {
1748         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
1749 }
1750
1751 #endif  /* __KERNEL__ */
1752
1753 #endif  /* _EXT4_H */