fs/ext4/ext4.h

   1 /*
   2  *  ext4.h
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card (card@masi.ibp.fr)
   6  * Laboratoire MASI - Institut Blaise Pascal
   7  * Universite Pierre et Marie Curie (Paris VI)
   8  *
   9  *  from
  10  *
  11  *  linux/include/linux/minix_fs.h
  12  *
  13  *  Copyright (C) 1991, 1992  Linus Torvalds
  14  */
  15
  16 #ifndef _EXT4_H
  17 #define _EXT4_H
  18
  19 #include <linux/types.h>
  20 #include <linux/blkdev.h>
  21 #include <linux/magic.h>
  22 #include <linux/jbd2.h>
  23 #include <linux/quota.h>
  24 #include <linux/rwsem.h>
  25 #include <linux/rbtree.h>
  26 #include <linux/seqlock.h>
  27 #include <linux/mutex.h>
  28 #include <linux/timer.h>
  29 #include <linux/wait.h>
  30 #include <linux/blockgroup_lock.h>
  31 #include <linux/percpu_counter.h>
  32
  33 /*
  34  * The fourth extended filesystem constants/structures
  35  */
  36
  37 /*
  38  * Define EXT4FS_DEBUG to produce debug messages
  39  */
  40 #undef EXT4FS_DEBUG
  41
  42 /*
  43  * Debug code
  44  */
  45 #ifdef EXT4FS_DEBUG
  46 #define ext4_debug(f, a...)                                             \
  47         do {                                                            \
  48                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
  49                         __FILE__, __LINE__, __func__);                  \
  50                 printk(KERN_DEBUG f, ## a);                             \
  51         } while (0)
  52 #else
  53 #define ext4_debug(f, a...)     do {} while (0)
  54 #endif
  55
  56 /* data type for block offset of block group */
  57 typedef int ext4_grpblk_t;
  58
  59 /* data type for filesystem-wide blocks number */
  60 typedef unsigned long long ext4_fsblk_t;
  61
  62 /* data type for file logical block number */
  63 typedef __u32 ext4_lblk_t;
  64
  65 /* data type for block group number */
  66 typedef unsigned int ext4_group_t;
  67
  68 /*
  69  * Flags used in mballoc's allocation_context flags field.
  70  *
  71  * Also used to show what's going on for debugging purposes when the
  72  * flag field is exported via the traceport interface
  73  */
  74
  75 /* prefer goal again. length */
  76 #define EXT4_MB_HINT_MERGE              0x0001
  77 /* blocks already reserved */
  78 #define EXT4_MB_HINT_RESERVED           0x0002
  79 /* metadata is being allocated */
  80 #define EXT4_MB_HINT_METADATA           0x0004
  81 /* first blocks in the file */
  82 #define EXT4_MB_HINT_FIRST              0x0008
  83 /* search for the best chunk */
  84 #define EXT4_MB_HINT_BEST               0x0010
  85 /* data is being allocated */
  86 #define EXT4_MB_HINT_DATA               0x0020
  87 /* don't preallocate (for tails) */
  88 #define EXT4_MB_HINT_NOPREALLOC         0x0040
  89 /* allocate for locality group */
  90 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
  91 /* allocate goal blocks or none */
  92 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
  93 /* goal is meaningful */
  94 #define EXT4_MB_HINT_TRY_GOAL           0x0200
  95 /* blocks already pre-reserved by delayed allocation */
  96 #define EXT4_MB_DELALLOC_RESERVED       0x0400
  97 /* We are doing stream allocation */
  98 #define EXT4_MB_STREAM_ALLOC            0x0800
  99
 100
 101 struct ext4_allocation_request {
 102         /* target inode for block we're allocating */
 103         struct inode *inode;
 104         /* how many blocks we want to allocate */
 105         unsigned int len;
 106         /* logical block in target inode */
 107         ext4_lblk_t logical;
 108         /* the closest logical allocated block to the left */
 109         ext4_lblk_t lleft;
 110         /* the closest logical allocated block to the right */
 111         ext4_lblk_t lright;
 112         /* phys. target (a hint) */
 113         ext4_fsblk_t goal;
 114         /* phys. block for the closest logical allocated block to the left */
 115         ext4_fsblk_t pleft;
 116         /* phys. block for the closest logical allocated block to the right */
 117         ext4_fsblk_t pright;
 118         /* flags. see above EXT4_MB_HINT_* */
 119         unsigned int flags;
 120 };
 121
 122 /*
 123  * For delayed allocation tracking
 124  */
 125 struct mpage_da_data {
 126         struct inode *inode;
 127         sector_t b_blocknr;             /* start block number of extent */
 128         size_t b_size;                  /* size of extent */
 129         unsigned long b_state;          /* state of the extent */
 130         unsigned long first_page, next_page;    /* extent of pages */
 131         struct writeback_control *wbc;
 132         int io_done;
 133         int pages_written;
 134         int retval;
 135 };
 136 #define DIO_AIO_UNWRITTEN       0x1
 137 typedef struct ext4_io_end {
 138         struct list_head        list;           /* per-file finished AIO list */
 139         struct inode            *inode;         /* file being written to */
 140         unsigned int            flag;           /* unwritten or not */
 141         int                     error;          /* I/O error code */
 142         loff_t                  offset;         /* offset in the file */
 143         ssize_t                 size;           /* size of the extent */
 144         struct work_struct      work;           /* data work queue */
 145 } ext4_io_end_t;
 146
 147 /*
 148  * Special inodes numbers
 149  */
 150 #define EXT4_BAD_INO             1      /* Bad blocks inode */
 151 #define EXT4_ROOT_INO            2      /* Root inode */
 152 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
 153 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
 154 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
 155 #define EXT4_JOURNAL_INO         8      /* Journal inode */
 156
 157 /* First non-reserved inode for old ext4 filesystems */
 158 #define EXT4_GOOD_OLD_FIRST_INO 11
 159
 160 /*
 161  * Maximal count of links to a file
 162  */
 163 #define EXT4_LINK_MAX           65000
 164
 165 /*
 166  * Macro-instructions used to manage several block sizes
 167  */
 168 #define EXT4_MIN_BLOCK_SIZE             1024
 169 #define EXT4_MAX_BLOCK_SIZE             65536
 170 #define EXT4_MIN_BLOCK_LOG_SIZE         10
 171 #ifdef __KERNEL__
 172 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
 173 #else
 174 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
 175 #endif
 176 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
 177 #ifdef __KERNEL__
 178 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
 179 #else
 180 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
 181 #endif
 182 #ifdef __KERNEL__
 183 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
 184 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
 185 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
 186 #else
 187 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 188                                  EXT4_GOOD_OLD_INODE_SIZE : \
 189                                  (s)->s_inode_size)
 190 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 191                                  EXT4_GOOD_OLD_FIRST_INO : \
 192                                  (s)->s_first_ino)
 193 #endif
 194 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
 195
 196 /*
 197  * Structure of a blocks group descriptor
 198  */
 199 struct ext4_group_desc
 200 {
 201         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
 202         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
 203         __le32  bg_inode_table_lo;      /* Inodes table block */
 204         __le16  bg_free_blocks_count_lo;/* Free blocks count */
 205         __le16  bg_free_inodes_count_lo;/* Free inodes count */
 206         __le16  bg_used_dirs_count_lo;  /* Directories count */
 207         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
 208         __u32   bg_reserved[2];         /* Likely block/inode bitmap checksum */
 209         __le16  bg_itable_unused_lo;    /* Unused inodes count */
 210         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
 211         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
 212         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
 213         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
 214         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
 215         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
 216         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
 217         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
 218         __u32   bg_reserved2[3];
 219 };
 220
 221 /*
 222  * Structure of a flex block group info
 223  */
 224
 225 struct flex_groups {
 226         atomic_t free_inodes;
 227         atomic_t free_blocks;
 228         atomic_t used_dirs;
 229 };
 230
 231 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
 232 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
 233 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
 234
 235 /*
 236  * Macro-instructions used to manage group descriptors
 237  */
 238 #define EXT4_MIN_DESC_SIZE              32
 239 #define EXT4_MIN_DESC_SIZE_64BIT        64
 240 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
 241 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
 242 #ifdef __KERNEL__
 243 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
 244 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
 245 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
 246 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
 247 #else
 248 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
 249 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
 250 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
 251 #endif
 252
 253 /*
 254  * Constants relative to the data blocks
 255  */
 256 #define EXT4_NDIR_BLOCKS                12
 257 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
 258 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
 259 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
 260 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
 261
 262 /*
 263  * Inode flags
 264  */
 265 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
 266 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
 267 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
 268 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
 269 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
 270 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
 271 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
 272 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
 273 /* Reserved for compression usage... */
 274 #define EXT4_DIRTY_FL                   0x00000100
 275 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
 276 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
 277 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
 278 /* End compression flags --- maybe not all used */
 279 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
 280 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
 281 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
 282 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
 283 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
 284 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
 285 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
 286 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
 287 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
 288 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
 289 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
 290
 291 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
 292 #define EXT4_FL_USER_MODIFIABLE         0x004B80FF /* User modifiable flags */
 293
 294 /* Flags that should be inherited by new inodes from their parent. */
 295 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
 296                            EXT4_SYNC_FL | EXT4_IMMUTABLE_FL | EXT4_APPEND_FL |\
 297                            EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
 298                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
 299                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
 300
 301 /* Flags that are appropriate for regular files (all but dir-specific ones). */
 302 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
 303
 304 /* Flags that are appropriate for non-directories/regular files. */
 305 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
 306
 307 /* Mask out flags that are inappropriate for the given type of inode. */
 308 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
 309 {
 310         if (S_ISDIR(mode))
 311                 return flags;
 312         else if (S_ISREG(mode))
 313                 return flags & EXT4_REG_FLMASK;
 314         else
 315                 return flags & EXT4_OTHER_FLMASK;
 316 }
 317
 318 /*
 319  * Inode flags used for atomic set/get
 320  */
 321 enum {
 322         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
 323         EXT4_INODE_UNRM         = 1,    /* Undelete */
 324         EXT4_INODE_COMPR        = 2,    /* Compress file */
 325         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
 326         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
 327         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
 328         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
 329         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
 330 /* Reserved for compression usage... */
 331         EXT4_INODE_DIRTY        = 8,
 332         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
 333         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
 334         EXT4_INODE_ECOMPR       = 11,   /* Compression error */
 335 /* End compression flags --- maybe not all used */
 336         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
 337         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
 338         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
 339         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
 340         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
 341         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
 342         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
 343         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
 344         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
 345         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
 346         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
 347 };
 348
 349 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
 350 #define CHECK_FLAG_VALUE(FLAG) if (!TEST_FLAG_VALUE(FLAG)) { \
 351         printk(KERN_EMERG "EXT4 flag fail: " #FLAG ": %d %d\n", \
 352                 EXT4_##FLAG##_FL, EXT4_INODE_##FLAG); BUG_ON(1); }
 353
 354 /*
 355  * Since it's pretty easy to mix up bit numbers and hex values, and we
 356  * can't do a compile-time test for ENUM values, we use a run-time
 357  * test to make sure that EXT4_XXX_FL is consistent with respect to
 358  * EXT4_INODE_XXX.  If all is well the printk and BUG_ON will all drop
 359  * out so it won't cost any extra space in the compiled kernel image.
 360  * But it's important that these values are the same, since we are
 361  * using EXT4_INODE_XXX to test for the flag values, but EXT4_XX_FL
 362  * must be consistent with the values of FS_XXX_FL defined in
 363  * include/linux/fs.h and the on-disk values found in ext2, ext3, and
 364  * ext4 filesystems, and of course the values defined in e2fsprogs.
 365  *
 366  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
 367  */
 368 static inline void ext4_check_flag_values(void)
 369 {
 370         CHECK_FLAG_VALUE(SECRM);
 371         CHECK_FLAG_VALUE(UNRM);
 372         CHECK_FLAG_VALUE(COMPR);
 373         CHECK_FLAG_VALUE(SYNC);
 374         CHECK_FLAG_VALUE(IMMUTABLE);
 375         CHECK_FLAG_VALUE(APPEND);
 376         CHECK_FLAG_VALUE(NODUMP);
 377         CHECK_FLAG_VALUE(NOATIME);
 378         CHECK_FLAG_VALUE(DIRTY);
 379         CHECK_FLAG_VALUE(COMPRBLK);
 380         CHECK_FLAG_VALUE(NOCOMPR);
 381         CHECK_FLAG_VALUE(ECOMPR);
 382         CHECK_FLAG_VALUE(INDEX);
 383         CHECK_FLAG_VALUE(IMAGIC);
 384         CHECK_FLAG_VALUE(JOURNAL_DATA);
 385         CHECK_FLAG_VALUE(NOTAIL);
 386         CHECK_FLAG_VALUE(DIRSYNC);
 387         CHECK_FLAG_VALUE(TOPDIR);
 388         CHECK_FLAG_VALUE(HUGE_FILE);
 389         CHECK_FLAG_VALUE(EXTENTS);
 390         CHECK_FLAG_VALUE(EA_INODE);
 391         CHECK_FLAG_VALUE(EOFBLOCKS);
 392         CHECK_FLAG_VALUE(RESERVED);
 393 }
 394
 395 /* Used to pass group descriptor data when online resize is done */
 396 struct ext4_new_group_input {
 397         __u32 group;            /* Group number for this data */
 398         __u64 block_bitmap;     /* Absolute block number of block bitmap */
 399         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
 400         __u64 inode_table;      /* Absolute block number of inode table start */
 401         __u32 blocks_count;     /* Total number of blocks in this group */
 402         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
 403         __u16 unused;
 404 };
 405
 406 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
 407 struct ext4_new_group_data {
 408         __u32 group;
 409         __u64 block_bitmap;
 410         __u64 inode_bitmap;
 411         __u64 inode_table;
 412         __u32 blocks_count;
 413         __u16 reserved_blocks;
 414         __u16 unused;
 415         __u32 free_blocks_count;
 416 };
 417
 418 /*
 419  * Flags used by ext4_get_blocks()
 420  */
 421         /* Allocate any needed blocks and/or convert an unitialized
 422            extent to be an initialized ext4 */
 423 #define EXT4_GET_BLOCKS_CREATE                  0x0001
 424         /* Request the creation of an unitialized extent */
 425 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
 426 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
 427                                                  EXT4_GET_BLOCKS_CREATE)
 428         /* Caller is from the delayed allocation writeout path,
 429            so set the magic i_delalloc_reserve_flag after taking the
 430            inode allocation semaphore for */
 431 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
 432         /* caller is from the direct IO path, request to creation of an
 433         unitialized extents if not allocated, split the uninitialized
 434         extent if blocks has been preallocated already*/
 435 #define EXT4_GET_BLOCKS_DIO                     0x0008
 436 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
 437 #define EXT4_GET_BLOCKS_DIO_CREATE_EXT          (EXT4_GET_BLOCKS_DIO|\
 438                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
 439         /* Convert extent to initialized after direct IO complete */
 440 #define EXT4_GET_BLOCKS_DIO_CONVERT_EXT         (EXT4_GET_BLOCKS_CONVERT|\
 441                                          EXT4_GET_BLOCKS_DIO_CREATE_EXT)
 442
 443 /*
 444  * ioctl commands
 445  */
 446 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
 447 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
 448 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
 449 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
 450 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
 451 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
 452 #ifdef CONFIG_JBD2_DEBUG
 453 #define EXT4_IOC_WAIT_FOR_READONLY      _IOR('f', 99, long)
 454 #endif
 455 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
 456 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
 457 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
 458 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
 459 #define EXT4_IOC_MIGRATE                _IO('f', 9)
 460  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
 461  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
 462 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
 463 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
 464
 465 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 466 /*
 467  * ioctl commands in 32 bit emulation
 468  */
 469 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
 470 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
 471 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
 472 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
 473 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
 474 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
 475 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
 476 #ifdef CONFIG_JBD2_DEBUG
 477 #define EXT4_IOC32_WAIT_FOR_READONLY    _IOR('f', 99, int)
 478 #endif
 479 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
 480 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
 481 #endif
 482
 483
 484 /*
 485  *  Mount options
 486  */
 487 struct ext4_mount_options {
 488         unsigned long s_mount_opt;
 489         uid_t s_resuid;
 490         gid_t s_resgid;
 491         unsigned long s_commit_interval;
 492         u32 s_min_batch_time, s_max_batch_time;
 493 #ifdef CONFIG_QUOTA
 494         int s_jquota_fmt;
 495         char *s_qf_names[MAXQUOTAS];
 496 #endif
 497 };
 498
 499 /* Max physical block we can addres w/o extents */
 500 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
 501
 502 /*
 503  * Structure of an inode on the disk
 504  */
 505 struct ext4_inode {
 506         __le16  i_mode;         /* File mode */
 507         __le16  i_uid;          /* Low 16 bits of Owner Uid */
 508         __le32  i_size_lo;      /* Size in bytes */
 509         __le32  i_atime;        /* Access time */
 510         __le32  i_ctime;        /* Inode Change time */
 511         __le32  i_mtime;        /* Modification time */
 512         __le32  i_dtime;        /* Deletion Time */
 513         __le16  i_gid;          /* Low 16 bits of Group Id */
 514         __le16  i_links_count;  /* Links count */
 515         __le32  i_blocks_lo;    /* Blocks count */
 516         __le32  i_flags;        /* File flags */
 517         union {
 518                 struct {
 519                         __le32  l_i_version;
 520                 } linux1;
 521                 struct {
 522                         __u32  h_i_translator;
 523                 } hurd1;
 524                 struct {
 525                         __u32  m_i_reserved1;
 526                 } masix1;
 527         } osd1;                         /* OS dependent 1 */
 528         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
 529         __le32  i_generation;   /* File version (for NFS) */
 530         __le32  i_file_acl_lo;  /* File ACL */
 531         __le32  i_size_high;
 532         __le32  i_obso_faddr;   /* Obsoleted fragment address */
 533         union {
 534                 struct {
 535                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
 536                         __le16  l_i_file_acl_high;
 537                         __le16  l_i_uid_high;   /* these 2 fields */
 538                         __le16  l_i_gid_high;   /* were reserved2[0] */
 539                         __u32   l_i_reserved2;
 540                 } linux2;
 541                 struct {
 542                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 543                         __u16   h_i_mode_high;
 544                         __u16   h_i_uid_high;
 545                         __u16   h_i_gid_high;
 546                         __u32   h_i_author;
 547                 } hurd2;
 548                 struct {
 549                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 550                         __le16  m_i_file_acl_high;
 551                         __u32   m_i_reserved2[2];
 552                 } masix2;
 553         } osd2;                         /* OS dependent 2 */
 554         __le16  i_extra_isize;
 555         __le16  i_pad1;
 556         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
 557         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
 558         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
 559         __le32  i_crtime;       /* File Creation time */
 560         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
 561         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
 562 };
 563
 564 struct move_extent {
 565         __u32 reserved;         /* should be zero */
 566         __u32 donor_fd;         /* donor file descriptor */
 567         __u64 orig_start;       /* logical start offset in block for orig */
 568         __u64 donor_start;      /* logical start offset in block for donor */
 569         __u64 len;              /* block length to be moved */
 570         __u64 moved_len;        /* moved block length */
 571 };
 572
 573 #define EXT4_EPOCH_BITS 2
 574 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
 575 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
 576
 577 /*
 578  * Extended fields will fit into an inode if the filesystem was formatted
 579  * with large inodes (-I 256 or larger) and there are not currently any EAs
 580  * consuming all of the available space. For new inodes we always reserve
 581  * enough space for the kernel's known extended fields, but for inodes
 582  * created with an old kernel this might not have been the case. None of
 583  * the extended inode fields is critical for correct filesystem operation.
 584  * This macro checks if a certain field fits in the inode. Note that
 585  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
 586  */
 587 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
 588         ((offsetof(typeof(*ext4_inode), field) +        \
 589           sizeof((ext4_inode)->field))                  \
 590         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
 591             (einode)->i_extra_isize))                   \
 592
 593 static inline __le32 ext4_encode_extra_time(struct timespec *time)
 594 {
 595        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
 596                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
 597                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
 598 }
 599
 600 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
 601 {
 602        if (sizeof(time->tv_sec) > 4)
 603                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
 604                                << 32;
 605        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
 606 }
 607
 608 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
 609 do {                                                                           \
 610         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
 611         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 612                 (raw_inode)->xtime ## _extra =                                 \
 613                                 ext4_encode_extra_time(&(inode)->xtime);       \
 614 } while (0)
 615
 616 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
 617 do {                                                                           \
 618         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 619                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
 620         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 621                 (raw_inode)->xtime ## _extra =                                 \
 622                                 ext4_encode_extra_time(&(einode)->xtime);      \
 623 } while (0)
 624
 625 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
 626 do {                                                                           \
 627         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
 628         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 629                 ext4_decode_extra_time(&(inode)->xtime,                        \
 630                                        raw_inode->xtime ## _extra);            \
 631 } while (0)
 632
 633 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
 634 do {                                                                           \
 635         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 636                 (einode)->xtime.tv_sec =                                       \
 637                         (signed)le32_to_cpu((raw_inode)->xtime);               \
 638         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 639                 ext4_decode_extra_time(&(einode)->xtime,                       \
 640                                        raw_inode->xtime ## _extra);            \
 641 } while (0)
 642
 643 #define i_disk_version osd1.linux1.l_i_version
 644
 645 #if defined(__KERNEL__) || defined(__linux__)
 646 #define i_reserved1     osd1.linux1.l_i_reserved1
 647 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
 648 #define i_blocks_high   osd2.linux2.l_i_blocks_high
 649 #define i_uid_low       i_uid
 650 #define i_gid_low       i_gid
 651 #define i_uid_high      osd2.linux2.l_i_uid_high
 652 #define i_gid_high      osd2.linux2.l_i_gid_high
 653 #define i_reserved2     osd2.linux2.l_i_reserved2
 654
 655 #elif defined(__GNU__)
 656
 657 #define i_translator    osd1.hurd1.h_i_translator
 658 #define i_uid_high      osd2.hurd2.h_i_uid_high
 659 #define i_gid_high      osd2.hurd2.h_i_gid_high
 660 #define i_author        osd2.hurd2.h_i_author
 661
 662 #elif defined(__masix__)
 663
 664 #define i_reserved1     osd1.masix1.m_i_reserved1
 665 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
 666 #define i_reserved2     osd2.masix2.m_i_reserved2
 667
 668 #endif /* defined(__KERNEL__) || defined(__linux__) */
 669
 670 /*
 671  * storage for cached extent
 672  */
 673 struct ext4_ext_cache {
 674         ext4_fsblk_t    ec_start;
 675         ext4_lblk_t     ec_block;
 676         __u32           ec_len; /* must be 32bit to return holes */
 677         __u32           ec_type;
 678 };
 679
 680 /*
 681  * fourth extended file system inode data in memory
 682  */
 683 struct ext4_inode_info {
 684         __le32  i_data[15];     /* unconverted */
 685         __u32   i_dtime;
 686         ext4_fsblk_t    i_file_acl;
 687
 688         /*
 689          * i_block_group is the number of the block group which contains
 690          * this file's inode.  Constant across the lifetime of the inode,
 691          * it is ued for making block allocation decisions - we try to
 692          * place a file's data blocks near its inode block, and new inodes
 693          * near to their parent directory's inode.
 694          */
 695         ext4_group_t    i_block_group;
 696         unsigned long   i_state_flags;          /* Dynamic state flags */
 697         unsigned long   i_flags;
 698
 699         ext4_lblk_t             i_dir_start_lookup;
 700 #ifdef CONFIG_EXT4_FS_XATTR
 701         /*
 702          * Extended attributes can be read independently of the main file
 703          * data. Taking i_mutex even when reading would cause contention
 704          * between readers of EAs and writers of regular file data, so
 705          * instead we synchronize on xattr_sem when reading or changing
 706          * EAs.
 707          */
 708         struct rw_semaphore xattr_sem;
 709 #endif
 710
 711         struct list_head i_orphan;      /* unlinked but open inodes */
 712
 713         /*
 714          * i_disksize keeps track of what the inode size is ON DISK, not
 715          * in memory.  During truncate, i_size is set to the new size by
 716          * the VFS prior to calling ext4_truncate(), but the filesystem won't
 717          * set i_disksize to 0 until the truncate is actually under way.
 718          *
 719          * The intent is that i_disksize always represents the blocks which
 720          * are used by this file.  This allows recovery to restart truncate
 721          * on orphans if we crash during truncate.  We actually write i_disksize
 722          * into the on-disk inode when writing inodes out, instead of i_size.
 723          *
 724          * The only time when i_disksize and i_size may be different is when
 725          * a truncate is in progress.  The only things which change i_disksize
 726          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
 727          */
 728         loff_t  i_disksize;
 729
 730         /*
 731          * i_data_sem is for serialising ext4_truncate() against
 732          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
 733          * data tree are chopped off during truncate. We can't do that in
 734          * ext4 because whenever we perform intermediate commits during
 735          * truncate, the inode and all the metadata blocks *must* be in a
 736          * consistent state which allows truncation of the orphans to restart
 737          * during recovery.  Hence we must fix the get_block-vs-truncate race
 738          * by other means, so we have i_data_sem.
 739          */
 740         struct rw_semaphore i_data_sem;
 741         struct inode vfs_inode;
 742         struct jbd2_inode jinode;
 743
 744         struct ext4_ext_cache i_cached_extent;
 745         /*
 746          * File creation time. Its function is same as that of
 747          * struct timespec i_{a,c,m}time in the generic inode.
 748          */
 749         struct timespec i_crtime;
 750
 751         /* mballoc */
 752         struct list_head i_prealloc_list;
 753         spinlock_t i_prealloc_lock;
 754
 755         /* ialloc */
 756         ext4_group_t    i_last_alloc_group;
 757
 758         /* allocation reservation info for delalloc */
 759         unsigned int i_reserved_data_blocks;
 760         unsigned int i_reserved_meta_blocks;
 761         unsigned int i_allocated_meta_blocks;
 762         unsigned short i_delalloc_reserved_flag;
 763         sector_t i_da_metadata_calc_last_lblock;
 764         int i_da_metadata_calc_len;
 765
 766         /* on-disk additional length */
 767         __u16 i_extra_isize;
 768
 769         spinlock_t i_block_reservation_lock;
 770 #ifdef CONFIG_QUOTA
 771         /* quota space reservation, managed internally by quota code */
 772         qsize_t i_reserved_quota;
 773 #endif
 774
 775         /* completed async DIOs that might need unwritten extents handling */
 776         struct list_head i_aio_dio_complete_list;
 777         /* current io_end structure for async DIO write*/
 778         ext4_io_end_t *cur_aio_dio;
 779
 780         /*
 781          * Transactions that contain inode's metadata needed to complete
 782          * fsync and fdatasync, respectively.
 783          */
 784         tid_t i_sync_tid;
 785         tid_t i_datasync_tid;
 786 };
 787
 788 /*
 789  * File system states
 790  */
 791 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
 792 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
 793 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
 794
 795 /*
 796  * Misc. filesystem flags
 797  */
 798 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
 799 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
 800 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
 801
 802 /*
 803  * Mount flags
 804  */
 805 #define EXT4_MOUNT_OLDALLOC             0x00002  /* Don't use the new Orlov allocator */
 806 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
 807 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
 808 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
 809 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
 810 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
 811 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
 812 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
 813 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
 814 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
 815 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
 816 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
 817 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
 818 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
 819 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
 820 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
 821 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
 822 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
 823 #define EXT4_MOUNT_NOBH                 0x40000 /* No bufferheads */
 824 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
 825 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
 826 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
 827 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
 828 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
 829 #define EXT4_MOUNT_I_VERSION            0x2000000 /* i_version support */
 830 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
 831 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
 832 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
 833 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
 834
 835 #define clear_opt(o, opt)               o &= ~EXT4_MOUNT_##opt
 836 #define set_opt(o, opt)                 o |= EXT4_MOUNT_##opt
 837 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
 838                                          EXT4_MOUNT_##opt)
 839
 840 #define ext4_set_bit                    ext2_set_bit
 841 #define ext4_set_bit_atomic             ext2_set_bit_atomic
 842 #define ext4_clear_bit                  ext2_clear_bit
 843 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
 844 #define ext4_test_bit                   ext2_test_bit
 845 #define ext4_find_first_zero_bit        ext2_find_first_zero_bit
 846 #define ext4_find_next_zero_bit         ext2_find_next_zero_bit
 847 #define ext4_find_next_bit              ext2_find_next_bit
 848
 849 /*
 850  * Maximal mount counts between two filesystem checks
 851  */
 852 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
 853 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
 854
 855 /*
 856  * Behaviour when detecting errors
 857  */
 858 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
 859 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
 860 #define EXT4_ERRORS_PANIC               3       /* Panic */
 861 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
 862
 863 /*
 864  * Structure of the super block
 865  */
 866 struct ext4_super_block {
 867 /*00*/  __le32  s_inodes_count;         /* Inodes count */
 868         __le32  s_blocks_count_lo;      /* Blocks count */
 869         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
 870         __le32  s_free_blocks_count_lo; /* Free blocks count */
 871 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
 872         __le32  s_first_data_block;     /* First Data Block */
 873         __le32  s_log_block_size;       /* Block size */
 874         __le32  s_obso_log_frag_size;   /* Obsoleted fragment size */
 875 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
 876         __le32  s_obso_frags_per_group; /* Obsoleted fragments per group */
 877         __le32  s_inodes_per_group;     /* # Inodes per group */
 878         __le32  s_mtime;                /* Mount time */
 879 /*30*/  __le32  s_wtime;                /* Write time */
 880         __le16  s_mnt_count;            /* Mount count */
 881         __le16  s_max_mnt_count;        /* Maximal mount count */
 882         __le16  s_magic;                /* Magic signature */
 883         __le16  s_state;                /* File system state */
 884         __le16  s_errors;               /* Behaviour when detecting errors */
 885         __le16  s_minor_rev_level;      /* minor revision level */
 886 /*40*/  __le32  s_lastcheck;            /* time of last check */
 887         __le32  s_checkinterval;        /* max. time between checks */
 888         __le32  s_creator_os;           /* OS */
 889         __le32  s_rev_level;            /* Revision level */
 890 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
 891         __le16  s_def_resgid;           /* Default gid for reserved blocks */
 892         /*
 893          * These fields are for EXT4_DYNAMIC_REV superblocks only.
 894          *
 895          * Note: the difference between the compatible feature set and
 896          * the incompatible feature set is that if there is a bit set
 897          * in the incompatible feature set that the kernel doesn't
 898          * know about, it should refuse to mount the filesystem.
 899          *
 900          * e2fsck's requirements are more strict; if it doesn't know
 901          * about a feature in either the compatible or incompatible
 902          * feature set, it must abort and not try to meddle with
 903          * things it doesn't understand...
 904          */
 905         __le32  s_first_ino;            /* First non-reserved inode */
 906         __le16  s_inode_size;           /* size of inode structure */
 907         __le16  s_block_group_nr;       /* block group # of this superblock */
 908         __le32  s_feature_compat;       /* compatible feature set */
 909 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
 910         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
 911 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
 912 /*78*/  char    s_volume_name[16];      /* volume name */
 913 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
 914 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
 915         /*
 916          * Performance hints.  Directory preallocation should only
 917          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
 918          */
 919         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
 920         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
 921         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
 922         /*
 923          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
 924          */
 925 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
 926 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
 927         __le32  s_journal_dev;          /* device number of journal file */
 928         __le32  s_last_orphan;          /* start of list of inodes to delete */
 929         __le32  s_hash_seed[4];         /* HTREE hash seed */
 930         __u8    s_def_hash_version;     /* Default hash version to use */
 931         __u8    s_reserved_char_pad;
 932         __le16  s_desc_size;            /* size of group descriptor */
 933 /*100*/ __le32  s_default_mount_opts;
 934         __le32  s_first_meta_bg;        /* First metablock block group */
 935         __le32  s_mkfs_time;            /* When the filesystem was created */
 936         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
 937         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
 938 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
 939         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
 940         __le32  s_free_blocks_count_hi; /* Free blocks count */
 941         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
 942         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
 943         __le32  s_flags;                /* Miscellaneous flags */
 944         __le16  s_raid_stride;          /* RAID stride */
 945         __le16  s_mmp_interval;         /* # seconds to wait in MMP checking */
 946         __le64  s_mmp_block;            /* Block for multi-mount protection */
 947         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
 948         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
 949         __u8    s_reserved_char_pad2;
 950         __le16  s_reserved_pad;
 951         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
 952         __u32   s_reserved[160];        /* Padding to the end of the block */
 953 };
 954
 955 #ifdef __KERNEL__
 956
 957 /*
 958  * run-time mount flags
 959  */
 960 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
 961 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
 962
 963 /*
 964  * fourth extended-fs super-block data in memory
 965  */
 966 struct ext4_sb_info {
 967         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
 968         unsigned long s_inodes_per_block;/* Number of inodes per block */
 969         unsigned long s_blocks_per_group;/* Number of blocks in a group */
 970         unsigned long s_inodes_per_group;/* Number of inodes in a group */
 971         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
 972         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
 973         unsigned long s_desc_per_block; /* Number of group descriptors per block */
 974         ext4_group_t s_groups_count;    /* Number of groups in the fs */
 975         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
 976         unsigned long s_overhead_last;  /* Last calculated overhead */
 977         unsigned long s_blocks_last;    /* Last seen block count */
 978         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
 979         struct buffer_head * s_sbh;     /* Buffer containing the super block */
 980         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
 981         struct buffer_head **s_group_desc;
 982         unsigned int s_mount_opt;
 983         unsigned int s_mount_flags;
 984         ext4_fsblk_t s_sb_block;
 985         uid_t s_resuid;
 986         gid_t s_resgid;
 987         unsigned short s_mount_state;
 988         unsigned short s_pad;
 989         int s_addr_per_block_bits;
 990         int s_desc_per_block_bits;
 991         int s_inode_size;
 992         int s_first_ino;
 993         unsigned int s_inode_readahead_blks;
 994         unsigned int s_inode_goal;
 995         spinlock_t s_next_gen_lock;
 996         u32 s_next_generation;
 997         u32 s_hash_seed[4];
 998         int s_def_hash_version;
 999         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1000         struct percpu_counter s_freeblocks_counter;
1001         struct percpu_counter s_freeinodes_counter;
1002         struct percpu_counter s_dirs_counter;
1003         struct percpu_counter s_dirtyblocks_counter;
1004         struct blockgroup_lock *s_blockgroup_lock;
1005         struct proc_dir_entry *s_proc;
1006         struct kobject s_kobj;
1007         struct completion s_kobj_unregister;
1008
1009         /* Journaling */
1010         struct inode *s_journal_inode;
1011         struct journal_s *s_journal;
1012         struct list_head s_orphan;
1013         struct mutex s_orphan_lock;
1014         struct mutex s_resize_lock;
1015         unsigned long s_commit_interval;
1016         u32 s_max_batch_time;
1017         u32 s_min_batch_time;
1018         struct block_device *journal_bdev;
1019 #ifdef CONFIG_JBD2_DEBUG
1020         struct timer_list turn_ro_timer;        /* For turning read-only (crash simulation) */
1021         wait_queue_head_t ro_wait_queue;        /* For people waiting for the fs to go read-only */
1022 #endif
1023 #ifdef CONFIG_QUOTA
1024         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
1025         int s_jquota_fmt;                       /* Format of quota to use */
1026 #endif
1027         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1028         struct rb_root system_blks;
1029
1030 #ifdef EXTENTS_STATS
1031         /* ext4 extents stats */
1032         unsigned long s_ext_min;
1033         unsigned long s_ext_max;
1034         unsigned long s_depth_max;
1035         spinlock_t s_ext_stats_lock;
1036         unsigned long s_ext_blocks;
1037         unsigned long s_ext_extents;
1038 #endif
1039
1040         /* for buddy allocator */
1041         struct ext4_group_info ***s_group_info;
1042         struct inode *s_buddy_cache;
1043         long s_blocks_reserved;
1044         spinlock_t s_reserve_lock;
1045         spinlock_t s_md_lock;
1046         tid_t s_last_transaction;
1047         unsigned short *s_mb_offsets;
1048         unsigned int *s_mb_maxs;
1049
1050         /* tunables */
1051         unsigned long s_stripe;
1052         unsigned int s_mb_stream_request;
1053         unsigned int s_mb_max_to_scan;
1054         unsigned int s_mb_min_to_scan;
1055         unsigned int s_mb_stats;
1056         unsigned int s_mb_order2_reqs;
1057         unsigned int s_mb_group_prealloc;
1058         unsigned int s_max_writeback_mb_bump;
1059         /* where last allocation was done - for stream allocation */
1060         unsigned long s_mb_last_group;
1061         unsigned long s_mb_last_start;
1062
1063         /* stats for buddy allocator */
1064         spinlock_t s_mb_pa_lock;
1065         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1066         atomic_t s_bal_success; /* we found long enough chunks */
1067         atomic_t s_bal_allocated;       /* in blocks */
1068         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1069         atomic_t s_bal_goals;   /* goal hits */
1070         atomic_t s_bal_breaks;  /* too long searches */
1071         atomic_t s_bal_2orders; /* 2^order hits */
1072         spinlock_t s_bal_lock;
1073         unsigned long s_mb_buddies_generated;
1074         unsigned long long s_mb_generation_time;
1075         atomic_t s_mb_lost_chunks;
1076         atomic_t s_mb_preallocated;
1077         atomic_t s_mb_discarded;
1078         atomic_t s_lock_busy;
1079
1080         /* locality groups */
1081         struct ext4_locality_group *s_locality_groups;
1082
1083         /* for write statistics */
1084         unsigned long s_sectors_written_start;
1085         u64 s_kbytes_written;
1086
1087         unsigned int s_log_groups_per_flex;
1088         struct flex_groups *s_flex_groups;
1089
1090         /* workqueue for dio unwritten */
1091         struct workqueue_struct *dio_unwritten_wq;
1092 };
1093
1094 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1095 {
1096         return sb->s_fs_info;
1097 }
1098 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1099 {
1100         return container_of(inode, struct ext4_inode_info, vfs_inode);
1101 }
1102
1103 static inline struct timespec ext4_current_time(struct inode *inode)
1104 {
1105         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1106                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1107 }
1108
1109 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1110 {
1111         return ino == EXT4_ROOT_INO ||
1112                 ino == EXT4_JOURNAL_INO ||
1113                 ino == EXT4_RESIZE_INO ||
1114                 (ino >= EXT4_FIRST_INO(sb) &&
1115                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1116 }
1117
1118 /*
1119  * Inode dynamic state flags
1120  */
1121 enum {
1122         EXT4_STATE_JDATA,               /* journaled data exists */
1123         EXT4_STATE_NEW,                 /* inode is newly created */
1124         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1125         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1126         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1127         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1128         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1129 };
1130
1131 #define EXT4_INODE_BIT_FNS(name, field)                                 \
1132 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1133 {                                                                       \
1134         return test_bit(bit, &EXT4_I(inode)->i_##field);                \
1135 }                                                                       \
1136 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1137 {                                                                       \
1138         set_bit(bit, &EXT4_I(inode)->i_##field);                        \
1139 }                                                                       \
1140 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1141 {                                                                       \
1142         clear_bit(bit, &EXT4_I(inode)->i_##field);                      \
1143 }
1144
1145 EXT4_INODE_BIT_FNS(flag, flags)
1146 EXT4_INODE_BIT_FNS(state, state_flags)
1147 #else
1148 /* Assume that user mode programs are passing in an ext4fs superblock, not
1149  * a kernel struct super_block.  This will allow us to call the feature-test
1150  * macros from user land. */
1151 #define EXT4_SB(sb)     (sb)
1152 #endif
1153
1154 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1155
1156 /*
1157  * Codes for operating systems
1158  */
1159 #define EXT4_OS_LINUX           0
1160 #define EXT4_OS_HURD            1
1161 #define EXT4_OS_MASIX           2
1162 #define EXT4_OS_FREEBSD         3
1163 #define EXT4_OS_LITES           4
1164
1165 /*
1166  * Revision levels
1167  */
1168 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1169 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1170
1171 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1172 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1173
1174 #define EXT4_GOOD_OLD_INODE_SIZE 128
1175
1176 /*
1177  * Feature set definitions
1178  */
1179
1180 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1181         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1182 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1183         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1184 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1185         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1186 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1187         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1188 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1189         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1190 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1191         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1192 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1193         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1194 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1195         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1196 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1197         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1198
1199 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1200 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1201 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1202 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1203 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1204 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1205
1206 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1207 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1208 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1209 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1210 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1211 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1212 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1213
1214 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1215 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1216 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1217 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1218 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1219 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1220 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1221 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1222 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1223
1224 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1225 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1226                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1227                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1228                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1229                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1230                                          EXT4_FEATURE_INCOMPAT_FLEX_BG)
1231 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1232                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1233                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1234                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1235                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1236                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1237                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE)
1238
1239 /*
1240  * Default values for user and/or group using reserved blocks
1241  */
1242 #define EXT4_DEF_RESUID         0
1243 #define EXT4_DEF_RESGID         0
1244
1245 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1246
1247 /*
1248  * Default mount options
1249  */
1250 #define EXT4_DEFM_DEBUG         0x0001
1251 #define EXT4_DEFM_BSDGROUPS     0x0002
1252 #define EXT4_DEFM_XATTR_USER    0x0004
1253 #define EXT4_DEFM_ACL           0x0008
1254 #define EXT4_DEFM_UID16         0x0010
1255 #define EXT4_DEFM_JMODE         0x0060
1256 #define EXT4_DEFM_JMODE_DATA    0x0020
1257 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1258 #define EXT4_DEFM_JMODE_WBACK   0x0060
1259
1260 /*
1261  * Default journal batch times
1262  */
1263 #define EXT4_DEF_MIN_BATCH_TIME 0
1264 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1265
1266 /*
1267  * Minimum number of groups in a flexgroup before we separate out
1268  * directories into the first block group of a flexgroup
1269  */
1270 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1271
1272 /*
1273  * Structure of a directory entry
1274  */
1275 #define EXT4_NAME_LEN 255
1276
1277 struct ext4_dir_entry {
1278         __le32  inode;                  /* Inode number */
1279         __le16  rec_len;                /* Directory entry length */
1280         __le16  name_len;               /* Name length */
1281         char    name[EXT4_NAME_LEN];    /* File name */
1282 };
1283
1284 /*
1285  * The new version of the directory entry.  Since EXT4 structures are
1286  * stored in intel byte order, and the name_len field could never be
1287  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1288  * file_type field.
1289  */
1290 struct ext4_dir_entry_2 {
1291         __le32  inode;                  /* Inode number */
1292         __le16  rec_len;                /* Directory entry length */
1293         __u8    name_len;               /* Name length */
1294         __u8    file_type;
1295         char    name[EXT4_NAME_LEN];    /* File name */
1296 };
1297
1298 /*
1299  * Ext4 directory file types.  Only the low 3 bits are used.  The
1300  * other bits are reserved for now.
1301  */
1302 #define EXT4_FT_UNKNOWN         0
1303 #define EXT4_FT_REG_FILE        1
1304 #define EXT4_FT_DIR             2
1305 #define EXT4_FT_CHRDEV          3
1306 #define EXT4_FT_BLKDEV          4
1307 #define EXT4_FT_FIFO            5
1308 #define EXT4_FT_SOCK            6
1309 #define EXT4_FT_SYMLINK         7
1310
1311 #define EXT4_FT_MAX             8
1312
1313 /*
1314  * EXT4_DIR_PAD defines the directory entries boundaries
1315  *
1316  * NOTE: It must be a multiple of 4
1317  */
1318 #define EXT4_DIR_PAD                    4
1319 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1320 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1321                                          ~EXT4_DIR_ROUND)
1322 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1323
1324 /*
1325  * Hash Tree Directory indexing
1326  * (c) Daniel Phillips, 2001
1327  */
1328
1329 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1330                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1331                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1332 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1333 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1334
1335 /* Legal values for the dx_root hash_version field: */
1336
1337 #define DX_HASH_LEGACY          0
1338 #define DX_HASH_HALF_MD4        1
1339 #define DX_HASH_TEA             2
1340 #define DX_HASH_LEGACY_UNSIGNED 3
1341 #define DX_HASH_HALF_MD4_UNSIGNED       4
1342 #define DX_HASH_TEA_UNSIGNED            5
1343
1344 #ifdef __KERNEL__
1345
1346 /* hash info structure used by the directory hash */
1347 struct dx_hash_info
1348 {
1349         u32             hash;
1350         u32             minor_hash;
1351         int             hash_version;
1352         u32             *seed;
1353 };
1354
1355 #define EXT4_HTREE_EOF  0x7fffffff
1356
1357 /*
1358  * Control parameters used by ext4_htree_next_block
1359  */
1360 #define HASH_NB_ALWAYS          1
1361
1362
1363 /*
1364  * Describe an inode's exact location on disk and in memory
1365  */
1366 struct ext4_iloc
1367 {
1368         struct buffer_head *bh;
1369         unsigned long offset;
1370         ext4_group_t block_group;
1371 };
1372
1373 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1374 {
1375         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1376 }
1377
1378 /*
1379  * This structure is stuffed into the struct file's private_data field
1380  * for directories.  It is where we put information so that we can do
1381  * readdir operations in hash tree order.
1382  */
1383 struct dir_private_info {
1384         struct rb_root  root;
1385         struct rb_node  *curr_node;
1386         struct fname    *extra_fname;
1387         loff_t          last_pos;
1388         __u32           curr_hash;
1389         __u32           curr_minor_hash;
1390         __u32           next_hash;
1391 };
1392
1393 /* calculate the first block number of the group */
1394 static inline ext4_fsblk_t
1395 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1396 {
1397         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1398                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1399 }
1400
1401 /*
1402  * Special error return code only used by dx_probe() and its callers.
1403  */
1404 #define ERR_BAD_DX_DIR  -75000
1405
1406 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1407                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1408
1409 extern struct proc_dir_entry *ext4_proc_root;
1410
1411 /*
1412  * Function prototypes
1413  */
1414
1415 /*
1416  * Ok, these declarations are also in <linux/kernel.h> but none of the
1417  * ext4 source programs needs to include it so they are duplicated here.
1418  */
1419 # define NORET_TYPE     /**/
1420 # define ATTRIB_NORET   __attribute__((noreturn))
1421 # define NORET_AND      noreturn,
1422
1423 /* bitmap.c */
1424 extern unsigned int ext4_count_free(struct buffer_head *, unsigned);
1425
1426 /* balloc.c */
1427 extern unsigned int ext4_block_group(struct super_block *sb,
1428                         ext4_fsblk_t blocknr);
1429 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1430                         ext4_fsblk_t blocknr);
1431 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1432 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1433                         ext4_group_t group);
1434 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1435                         ext4_fsblk_t goal, unsigned long *count, int *errp);
1436 extern int ext4_claim_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1437 extern int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1438 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1439                         ext4_fsblk_t block, unsigned long count, int metadata);
1440 extern void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
1441                                 ext4_fsblk_t block, unsigned long count);
1442 extern ext4_fsblk_t ext4_count_free_blocks(struct super_block *);
1443 extern void ext4_check_blocks_bitmap(struct super_block *);
1444 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1445                                                     ext4_group_t block_group,
1446                                                     struct buffer_head ** bh);
1447 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1448 struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1449                                       ext4_group_t block_group);
1450 extern unsigned ext4_init_block_bitmap(struct super_block *sb,
1451                                        struct buffer_head *bh,
1452                                        ext4_group_t group,
1453                                        struct ext4_group_desc *desc);
1454 #define ext4_free_blocks_after_init(sb, group, desc)                    \
1455                 ext4_init_block_bitmap(sb, NULL, group, desc)
1456
1457 /* dir.c */
1458 extern int ext4_check_dir_entry(const char *, struct inode *,
1459                                 struct ext4_dir_entry_2 *,
1460                                 struct buffer_head *, unsigned int);
1461 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1462                                     __u32 minor_hash,
1463                                     struct ext4_dir_entry_2 *dirent);
1464 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1465
1466 /* fsync.c */
1467 extern int ext4_sync_file(struct file *, struct dentry *, int);
1468
1469 /* hash.c */
1470 extern int ext4fs_dirhash(const char *name, int len, struct
1471                           dx_hash_info *hinfo);
1472
1473 /* ialloc.c */
1474 extern struct inode *ext4_new_inode(handle_t *, struct inode *, int,
1475                                     const struct qstr *qstr, __u32 goal);
1476 extern void ext4_free_inode(handle_t *, struct inode *);
1477 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1478 extern unsigned long ext4_count_free_inodes(struct super_block *);
1479 extern unsigned long ext4_count_dirs(struct super_block *);
1480 extern void ext4_check_inodes_bitmap(struct super_block *);
1481 extern unsigned ext4_init_inode_bitmap(struct super_block *sb,
1482                                        struct buffer_head *bh,
1483                                        ext4_group_t group,
1484                                        struct ext4_group_desc *desc);
1485 extern void mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1486
1487 /* mballoc.c */
1488 extern long ext4_mb_stats;
1489 extern long ext4_mb_max_to_scan;
1490 extern int ext4_mb_init(struct super_block *, int);
1491 extern int ext4_mb_release(struct super_block *);
1492 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1493                                 struct ext4_allocation_request *, int *);
1494 extern int ext4_mb_reserve_blocks(struct super_block *, int);
1495 extern void ext4_discard_preallocations(struct inode *);
1496 extern int __init init_ext4_mballoc(void);
1497 extern void exit_ext4_mballoc(void);
1498 extern void ext4_mb_free_blocks(handle_t *, struct inode *,
1499                 ext4_fsblk_t, unsigned long, int, unsigned long *);
1500 extern int ext4_mb_add_groupinfo(struct super_block *sb,
1501                 ext4_group_t i, struct ext4_group_desc *desc);
1502 extern int ext4_mb_get_buddy_cache_lock(struct super_block *, ext4_group_t);
1503 extern void ext4_mb_put_buddy_cache_lock(struct super_block *,
1504                                                 ext4_group_t, int);
1505 /* inode.c */
1506 int ext4_forget(handle_t *handle, int is_metadata, struct inode *inode,
1507                 struct buffer_head *bh, ext4_fsblk_t blocknr);
1508 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1509                                                 ext4_lblk_t, int, int *);
1510 struct buffer_head *ext4_bread(handle_t *, struct inode *,
1511                                                 ext4_lblk_t, int, int *);
1512 int ext4_get_block(struct inode *inode, sector_t iblock,
1513                                 struct buffer_head *bh_result, int create);
1514
1515 extern struct inode *ext4_iget(struct super_block *, unsigned long);
1516 extern int  ext4_write_inode(struct inode *, int);
1517 extern int  ext4_setattr(struct dentry *, struct iattr *);
1518 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1519                                 struct kstat *stat);
1520 extern void ext4_delete_inode(struct inode *);
1521 extern int  ext4_sync_inode(handle_t *, struct inode *);
1522 extern void ext4_dirty_inode(struct inode *);
1523 extern int ext4_change_inode_journal_flag(struct inode *, int);
1524 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1525 extern int ext4_can_truncate(struct inode *inode);
1526 extern void ext4_truncate(struct inode *);
1527 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
1528 extern void ext4_set_inode_flags(struct inode *);
1529 extern void ext4_get_inode_flags(struct ext4_inode_info *);
1530 extern int ext4_alloc_da_blocks(struct inode *inode);
1531 extern void ext4_set_aops(struct inode *inode);
1532 extern int ext4_writepage_trans_blocks(struct inode *);
1533 extern int ext4_meta_trans_blocks(struct inode *, int nrblocks, int idxblocks);
1534 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1535 extern int ext4_block_truncate_page(handle_t *handle,
1536                 struct address_space *mapping, loff_t from);
1537 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1538 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
1539 extern int flush_aio_dio_completed_IO(struct inode *inode);
1540 extern void ext4_da_update_reserve_space(struct inode *inode,
1541                                         int used, int quota_claim);
1542 /* ioctl.c */
1543 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1544 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1545
1546 /* migrate.c */
1547 extern int ext4_ext_migrate(struct inode *);
1548
1549 /* namei.c */
1550 extern unsigned int ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize);
1551 extern __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize);
1552 extern int ext4_orphan_add(handle_t *, struct inode *);
1553 extern int ext4_orphan_del(handle_t *, struct inode *);
1554 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1555                                 __u32 start_minor_hash, __u32 *next_hash);
1556
1557 /* resize.c */
1558 extern int ext4_group_add(struct super_block *sb,
1559                                 struct ext4_new_group_data *input);
1560 extern int ext4_group_extend(struct super_block *sb,
1561                                 struct ext4_super_block *es,
1562                                 ext4_fsblk_t n_blocks_count);
1563
1564 /* super.c */
1565 extern void ext4_error(struct super_block *, const char *, const char *, ...)
1566         __attribute__ ((format (printf, 3, 4)));
1567 extern void __ext4_std_error(struct super_block *, const char *, int);
1568 extern void ext4_abort(struct super_block *, const char *, const char *, ...)
1569         __attribute__ ((format (printf, 3, 4)));
1570 extern void ext4_warning(struct super_block *, const char *, const char *, ...)
1571         __attribute__ ((format (printf, 3, 4)));
1572 extern void ext4_msg(struct super_block *, const char *, const char *, ...)
1573         __attribute__ ((format (printf, 3, 4)));
1574 extern void ext4_grp_locked_error(struct super_block *, ext4_group_t,
1575                                 const char *, const char *, ...)
1576         __attribute__ ((format (printf, 4, 5)));
1577 extern void ext4_update_dynamic_rev(struct super_block *sb);
1578 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1579                                         __u32 compat);
1580 extern int ext4_update_rocompat_feature(handle_t *handle,
1581                                         struct super_block *sb, __u32 rocompat);
1582 extern int ext4_update_incompat_feature(handle_t *handle,
1583                                         struct super_block *sb, __u32 incompat);
1584 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1585                                       struct ext4_group_desc *bg);
1586 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1587                                       struct ext4_group_desc *bg);
1588 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1589                                      struct ext4_group_desc *bg);
1590 extern __u32 ext4_free_blks_count(struct super_block *sb,
1591                                 struct ext4_group_desc *bg);
1592 extern __u32 ext4_free_inodes_count(struct super_block *sb,
1593                                  struct ext4_group_desc *bg);
1594 extern __u32 ext4_used_dirs_count(struct super_block *sb,
1595                                 struct ext4_group_desc *bg);
1596 extern __u32 ext4_itable_unused_count(struct super_block *sb,
1597                                    struct ext4_group_desc *bg);
1598 extern void ext4_block_bitmap_set(struct super_block *sb,
1599                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1600 extern void ext4_inode_bitmap_set(struct super_block *sb,
1601                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1602 extern void ext4_inode_table_set(struct super_block *sb,
1603                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
1604 extern void ext4_free_blks_set(struct super_block *sb,
1605                                struct ext4_group_desc *bg, __u32 count);
1606 extern void ext4_free_inodes_set(struct super_block *sb,
1607                                 struct ext4_group_desc *bg, __u32 count);
1608 extern void ext4_used_dirs_set(struct super_block *sb,
1609                                 struct ext4_group_desc *bg, __u32 count);
1610 extern void ext4_itable_unused_set(struct super_block *sb,
1611                                    struct ext4_group_desc *bg, __u32 count);
1612 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
1613                                    struct ext4_group_desc *gdp);
1614 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
1615                                        struct ext4_group_desc *gdp);
1616
1617 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
1618 {
1619         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
1620                 le32_to_cpu(es->s_blocks_count_lo);
1621 }
1622
1623 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
1624 {
1625         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
1626                 le32_to_cpu(es->s_r_blocks_count_lo);
1627 }
1628
1629 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
1630 {
1631         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
1632                 le32_to_cpu(es->s_free_blocks_count_lo);
1633 }
1634
1635 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
1636                                          ext4_fsblk_t blk)
1637 {
1638         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
1639         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
1640 }
1641
1642 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
1643                                               ext4_fsblk_t blk)
1644 {
1645         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
1646         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
1647 }
1648
1649 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
1650                                            ext4_fsblk_t blk)
1651 {
1652         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
1653         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
1654 }
1655
1656 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
1657 {
1658         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
1659                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
1660                         le32_to_cpu(raw_inode->i_size_lo);
1661         else
1662                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
1663 }
1664
1665 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
1666 {
1667         raw_inode->i_size_lo = cpu_to_le32(i_size);
1668         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
1669 }
1670
1671 static inline
1672 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
1673                                             ext4_group_t group)
1674 {
1675          struct ext4_group_info ***grp_info;
1676          long indexv, indexh;
1677          grp_info = EXT4_SB(sb)->s_group_info;
1678          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
1679          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
1680          return grp_info[indexv][indexh];
1681 }
1682
1683 /*
1684  * Reading s_groups_count requires using smp_rmb() afterwards.  See
1685  * the locking protocol documented in the comments of ext4_group_add()
1686  * in resize.c
1687  */
1688 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
1689 {
1690         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
1691
1692         smp_rmb();
1693         return ngroups;
1694 }
1695
1696 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
1697                                              ext4_group_t block_group)
1698 {
1699         return block_group >> sbi->s_log_groups_per_flex;
1700 }
1701
1702 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
1703 {
1704         return 1 << sbi->s_log_groups_per_flex;
1705 }
1706
1707 #define ext4_std_error(sb, errno)                               \
1708 do {                                                            \
1709         if ((errno))                                            \
1710                 __ext4_std_error((sb), __func__, (errno));      \
1711 } while (0)
1712
1713 #ifdef CONFIG_SMP
1714 /* Each CPU can accumulate percpu_counter_batch blocks in their local
1715  * counters. So we need to make sure we have free blocks more
1716  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
1717  */
1718 #define EXT4_FREEBLOCKS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
1719 #else
1720 #define EXT4_FREEBLOCKS_WATERMARK 0
1721 #endif
1722
1723 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
1724 {
1725         /*
1726          * XXX: replace with spinlock if seen contended -bzzz
1727          */
1728         down_write(&EXT4_I(inode)->i_data_sem);
1729         if (newsize > EXT4_I(inode)->i_disksize)
1730                 EXT4_I(inode)->i_disksize = newsize;
1731         up_write(&EXT4_I(inode)->i_data_sem);
1732         return ;
1733 }
1734
1735 struct ext4_group_info {
1736         unsigned long   bb_state;
1737         struct rb_root  bb_free_root;
1738         ext4_grpblk_t   bb_first_free;  /* first free block */
1739         ext4_grpblk_t   bb_free;        /* total free blocks */
1740         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
1741         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
1742         struct          list_head bb_prealloc_list;
1743 #ifdef DOUBLE_CHECK
1744         void            *bb_bitmap;
1745 #endif
1746         struct rw_semaphore alloc_sem;
1747         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
1748                                          * regions, index is order.
1749                                          * bb_counters[3] = 5 means
1750                                          * 5 free 8-block regions. */
1751 };
1752
1753 #define EXT4_GROUP_INFO_NEED_INIT_BIT   0
1754
1755 #define EXT4_MB_GRP_NEED_INIT(grp)      \
1756         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
1757
1758 #define EXT4_MAX_CONTENTION             8
1759 #define EXT4_CONTENTION_THRESHOLD       2
1760
1761 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
1762                                               ext4_group_t group)
1763 {
1764         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
1765 }
1766
1767 /*
1768  * Returns true if the filesystem is busy enough that attempts to
1769  * access the block group locks has run into contention.
1770  */
1771 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
1772 {
1773         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
1774 }
1775
1776 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
1777 {
1778         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
1779         if (spin_trylock(lock))
1780                 /*
1781                  * We're able to grab the lock right away, so drop the
1782                  * lock contention counter.
1783                  */
1784                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
1785         else {
1786                 /*
1787                  * The lock is busy, so bump the contention counter,
1788                  * and then wait on the spin lock.
1789                  */
1790                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
1791                                   EXT4_MAX_CONTENTION);
1792                 spin_lock(lock);
1793         }
1794 }
1795
1796 static inline void ext4_unlock_group(struct super_block *sb,
1797                                         ext4_group_t group)
1798 {
1799         spin_unlock(ext4_group_lock_ptr(sb, group));
1800 }
1801
1802 /*
1803  * Inodes and files operations
1804  */
1805
1806 /* dir.c */
1807 extern const struct file_operations ext4_dir_operations;
1808
1809 /* file.c */
1810 extern const struct inode_operations ext4_file_inode_operations;
1811 extern const struct file_operations ext4_file_operations;
1812
1813 /* namei.c */
1814 extern const struct inode_operations ext4_dir_inode_operations;
1815 extern const struct inode_operations ext4_special_inode_operations;
1816 extern struct dentry *ext4_get_parent(struct dentry *child);
1817
1818 /* symlink.c */
1819 extern const struct inode_operations ext4_symlink_inode_operations;
1820 extern const struct inode_operations ext4_fast_symlink_inode_operations;
1821
1822 /* block_validity */
1823 extern void ext4_release_system_zone(struct super_block *sb);
1824 extern int ext4_setup_system_zone(struct super_block *sb);
1825 extern int __init init_ext4_system_zone(void);
1826 extern void exit_ext4_system_zone(void);
1827 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
1828                                  ext4_fsblk_t start_blk,
1829                                  unsigned int count);
1830
1831 /* extents.c */
1832 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
1833 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
1834 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
1835                                        int chunk);
1836 extern int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
1837                                ext4_lblk_t iblock, unsigned int max_blocks,
1838                                struct buffer_head *bh_result, int flags);
1839 extern void ext4_ext_truncate(struct inode *);
1840 extern void ext4_ext_init(struct super_block *);
1841 extern void ext4_ext_release(struct super_block *);
1842 extern long ext4_fallocate(struct inode *inode, int mode, loff_t offset,
1843                           loff_t len);
1844 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
1845                           ssize_t len);
1846 extern int ext4_get_blocks(handle_t *handle, struct inode *inode,
1847                            sector_t block, unsigned int max_blocks,
1848                            struct buffer_head *bh, int flags);
1849 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1850                         __u64 start, __u64 len);
1851 /* move_extent.c */
1852 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
1853                              __u64 start_orig, __u64 start_donor,
1854                              __u64 len, __u64 *moved_len);
1855
1856
1857 /*
1858  * Add new method to test wether block and inode bitmaps are properly
1859  * initialized. With uninit_bg reading the block from disk is not enough
1860  * to mark the bitmap uptodate. We need to also zero-out the bitmap
1861  */
1862 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
1863
1864 static inline int bitmap_uptodate(struct buffer_head *bh)
1865 {
1866         return (buffer_uptodate(bh) &&
1867                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
1868 }
1869 static inline void set_bitmap_uptodate(struct buffer_head *bh)
1870 {
1871         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
1872 }
1873
1874 #endif  /* __KERNEL__ */
1875
1876 #endif  /* _EXT4_H */