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