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 #ifdef __KERNEL__
  33 #include <linux/compat.h>
  34 #endif
  35
  36 /*
  37  * The fourth extended filesystem constants/structures
  38  */
  39
  40 /*
  41  * Define EXT4FS_DEBUG to produce debug messages
  42  */
  43 #undef EXT4FS_DEBUG
  44
  45 /*
  46  * Debug code
  47  */
  48 #ifdef EXT4FS_DEBUG
  49 #define ext4_debug(f, a...)                                             \
  50         do {                                                            \
  51                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
  52                         __FILE__, __LINE__, __func__);                  \
  53                 printk(KERN_DEBUG f, ## a);                             \
  54         } while (0)
  55 #else
  56 #define ext4_debug(f, a...)     do {} while (0)
  57 #endif
  58
  59 #define EXT4_ERROR_INODE(inode, fmt, a...) \
  60         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
  61
  62 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...)                 \
  63         ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
  64
  65 #define EXT4_ERROR_FILE(file, block, fmt, a...)                         \
  66         ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
  67
  68 /* data type for block offset of block group */
  69 typedef int ext4_grpblk_t;
  70
  71 /* data type for filesystem-wide blocks number */
  72 typedef unsigned long long ext4_fsblk_t;
  73
  74 /* data type for file logical block number */
  75 typedef __u32 ext4_lblk_t;
  76
  77 /* data type for block group number */
  78 typedef unsigned int ext4_group_t;
  79
  80 /*
  81  * Flags used in mballoc's allocation_context flags field.
  82  *
  83  * Also used to show what's going on for debugging purposes when the
  84  * flag field is exported via the traceport interface
  85  */
  86
  87 /* prefer goal again. length */
  88 #define EXT4_MB_HINT_MERGE              0x0001
  89 /* blocks already reserved */
  90 #define EXT4_MB_HINT_RESERVED           0x0002
  91 /* metadata is being allocated */
  92 #define EXT4_MB_HINT_METADATA           0x0004
  93 /* first blocks in the file */
  94 #define EXT4_MB_HINT_FIRST              0x0008
  95 /* search for the best chunk */
  96 #define EXT4_MB_HINT_BEST               0x0010
  97 /* data is being allocated */
  98 #define EXT4_MB_HINT_DATA               0x0020
  99 /* don't preallocate (for tails) */
 100 #define EXT4_MB_HINT_NOPREALLOC         0x0040
 101 /* allocate for locality group */
 102 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
 103 /* allocate goal blocks or none */
 104 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
 105 /* goal is meaningful */
 106 #define EXT4_MB_HINT_TRY_GOAL           0x0200
 107 /* blocks already pre-reserved by delayed allocation */
 108 #define EXT4_MB_DELALLOC_RESERVED       0x0400
 109 /* We are doing stream allocation */
 110 #define EXT4_MB_STREAM_ALLOC            0x0800
 111 /* Use reserved root blocks if needed */
 112 #define EXT4_MB_USE_ROOT_BLOCKS         0x1000
 113
 114 struct ext4_allocation_request {
 115         /* target inode for block we're allocating */
 116         struct inode *inode;
 117         /* how many blocks we want to allocate */
 118         unsigned int len;
 119         /* logical block in target inode */
 120         ext4_lblk_t logical;
 121         /* the closest logical allocated block to the left */
 122         ext4_lblk_t lleft;
 123         /* the closest logical allocated block to the right */
 124         ext4_lblk_t lright;
 125         /* phys. target (a hint) */
 126         ext4_fsblk_t goal;
 127         /* phys. block for the closest logical allocated block to the left */
 128         ext4_fsblk_t pleft;
 129         /* phys. block for the closest logical allocated block to the right */
 130         ext4_fsblk_t pright;
 131         /* flags. see above EXT4_MB_HINT_* */
 132         unsigned int flags;
 133 };
 134
 135 /*
 136  * Logical to physical block mapping, used by ext4_map_blocks()
 137  *
 138  * This structure is used to pass requests into ext4_map_blocks() as
 139  * well as to store the information returned by ext4_map_blocks().  It
 140  * takes less room on the stack than a struct buffer_head.
 141  */
 142 #define EXT4_MAP_NEW            (1 << BH_New)
 143 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
 144 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
 145 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
 146 #define EXT4_MAP_UNINIT         (1 << BH_Uninit)
 147 /* Sometimes (in the bigalloc case, from ext4_da_get_block_prep) the caller of
 148  * ext4_map_blocks wants to know whether or not the underlying cluster has
 149  * already been accounted for. EXT4_MAP_FROM_CLUSTER conveys to the caller that
 150  * the requested mapping was from previously mapped (or delayed allocated)
 151  * cluster. We use BH_AllocFromCluster only for this flag. BH_AllocFromCluster
 152  * should never appear on buffer_head's state flags.
 153  */
 154 #define EXT4_MAP_FROM_CLUSTER   (1 << BH_AllocFromCluster)
 155 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
 156                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
 157                                  EXT4_MAP_UNINIT | EXT4_MAP_FROM_CLUSTER)
 158
 159 struct ext4_map_blocks {
 160         ext4_fsblk_t m_pblk;
 161         ext4_lblk_t m_lblk;
 162         unsigned int m_len;
 163         unsigned int m_flags;
 164 };
 165
 166 /*
 167  * For delayed allocation tracking
 168  */
 169 struct mpage_da_data {
 170         struct inode *inode;
 171         sector_t b_blocknr;             /* start block number of extent */
 172         size_t b_size;                  /* size of extent */
 173         unsigned long b_state;          /* state of the extent */
 174         unsigned long first_page, next_page;    /* extent of pages */
 175         struct writeback_control *wbc;
 176         int io_done;
 177         int pages_written;
 178         int retval;
 179 };
 180
 181 /*
 182  * Flags for ext4_io_end->flags
 183  */
 184 #define EXT4_IO_END_UNWRITTEN   0x0001
 185 #define EXT4_IO_END_ERROR       0x0002
 186 #define EXT4_IO_END_QUEUED      0x0004
 187 #define EXT4_IO_END_DIRECT      0x0008
 188 #define EXT4_IO_END_IN_FSYNC    0x0010
 189
 190 struct ext4_io_page {
 191         struct page     *p_page;
 192         atomic_t        p_count;
 193 };
 194
 195 #define MAX_IO_PAGES 128
 196
 197 typedef struct ext4_io_end {
 198         struct list_head        list;           /* per-file finished IO list */
 199         struct inode            *inode;         /* file being written to */
 200         unsigned int            flag;           /* unwritten or not */
 201         struct page             *page;          /* page struct for buffer write */
 202         loff_t                  offset;         /* offset in the file */
 203         ssize_t                 size;           /* size of the extent */
 204         struct work_struct      work;           /* data work queue */
 205         struct kiocb            *iocb;          /* iocb struct for AIO */
 206         int                     result;         /* error value for AIO */
 207         int                     num_io_pages;
 208         struct ext4_io_page     *pages[MAX_IO_PAGES];
 209 } ext4_io_end_t;
 210
 211 struct ext4_io_submit {
 212         int                     io_op;
 213         struct bio              *io_bio;
 214         ext4_io_end_t           *io_end;
 215         struct ext4_io_page     *io_page;
 216         sector_t                io_next_block;
 217 };
 218
 219 /*
 220  * Special inodes numbers
 221  */
 222 #define EXT4_BAD_INO             1      /* Bad blocks inode */
 223 #define EXT4_ROOT_INO            2      /* Root inode */
 224 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
 225 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
 226 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
 227 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
 228 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
 229 #define EXT4_JOURNAL_INO         8      /* Journal inode */
 230
 231 /* First non-reserved inode for old ext4 filesystems */
 232 #define EXT4_GOOD_OLD_FIRST_INO 11
 233
 234 /*
 235  * Maximal count of links to a file
 236  */
 237 #define EXT4_LINK_MAX           65000
 238
 239 /*
 240  * Macro-instructions used to manage several block sizes
 241  */
 242 #define EXT4_MIN_BLOCK_SIZE             1024
 243 #define EXT4_MAX_BLOCK_SIZE             65536
 244 #define EXT4_MIN_BLOCK_LOG_SIZE         10
 245 #define EXT4_MAX_BLOCK_LOG_SIZE         16
 246 #ifdef __KERNEL__
 247 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
 248 #else
 249 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
 250 #endif
 251 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
 252 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
 253                                          EXT4_SB(s)->s_cluster_bits)
 254 #ifdef __KERNEL__
 255 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
 256 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
 257 #else
 258 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
 259 #endif
 260 #ifdef __KERNEL__
 261 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
 262 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
 263 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
 264 #else
 265 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 266                                  EXT4_GOOD_OLD_INODE_SIZE : \
 267                                  (s)->s_inode_size)
 268 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 269                                  EXT4_GOOD_OLD_FIRST_INO : \
 270                                  (s)->s_first_ino)
 271 #endif
 272 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
 273
 274 /* Translate a block number to a cluster number */
 275 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
 276 /* Translate a cluster number to a block number */
 277 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
 278 /* Translate # of blks to # of clusters */
 279 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
 280                                  (sbi)->s_cluster_bits)
 281
 282 /*
 283  * Structure of a blocks group descriptor
 284  */
 285 struct ext4_group_desc
 286 {
 287         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
 288         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
 289         __le32  bg_inode_table_lo;      /* Inodes table block */
 290         __le16  bg_free_blocks_count_lo;/* Free blocks count */
 291         __le16  bg_free_inodes_count_lo;/* Free inodes count */
 292         __le16  bg_used_dirs_count_lo;  /* Directories count */
 293         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
 294         __u32   bg_reserved[2];         /* Likely block/inode bitmap checksum */
 295         __le16  bg_itable_unused_lo;    /* Unused inodes count */
 296         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
 297         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
 298         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
 299         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
 300         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
 301         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
 302         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
 303         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
 304         __u32   bg_reserved2[3];
 305 };
 306
 307 /*
 308  * Structure of a flex block group info
 309  */
 310
 311 struct flex_groups {
 312         atomic_t free_inodes;
 313         atomic_t free_clusters;
 314         atomic_t used_dirs;
 315 };
 316
 317 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
 318 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
 319 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
 320
 321 /*
 322  * Macro-instructions used to manage group descriptors
 323  */
 324 #define EXT4_MIN_DESC_SIZE              32
 325 #define EXT4_MIN_DESC_SIZE_64BIT        64
 326 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
 327 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
 328 #ifdef __KERNEL__
 329 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
 330 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
 331 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
 332 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
 333 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
 334 #else
 335 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
 336 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
 337 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
 338 #endif
 339
 340 /*
 341  * Constants relative to the data blocks
 342  */
 343 #define EXT4_NDIR_BLOCKS                12
 344 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
 345 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
 346 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
 347 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
 348
 349 /*
 350  * Inode flags
 351  */
 352 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
 353 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
 354 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
 355 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
 356 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
 357 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
 358 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
 359 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
 360 /* Reserved for compression usage... */
 361 #define EXT4_DIRTY_FL                   0x00000100
 362 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
 363 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
 364 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
 365 /* End compression flags --- maybe not all used */
 366 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
 367 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
 368 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
 369 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
 370 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
 371 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
 372 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
 373 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
 374 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
 375 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
 376 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
 377
 378 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
 379 #define EXT4_FL_USER_MODIFIABLE         0x004B80FF /* User modifiable flags */
 380
 381 /* Flags that should be inherited by new inodes from their parent. */
 382 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
 383                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
 384                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
 385                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
 386
 387 /* Flags that are appropriate for regular files (all but dir-specific ones). */
 388 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
 389
 390 /* Flags that are appropriate for non-directories/regular files. */
 391 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
 392
 393 /* Mask out flags that are inappropriate for the given type of inode. */
 394 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
 395 {
 396         if (S_ISDIR(mode))
 397                 return flags;
 398         else if (S_ISREG(mode))
 399                 return flags & EXT4_REG_FLMASK;
 400         else
 401                 return flags & EXT4_OTHER_FLMASK;
 402 }
 403
 404 /*
 405  * Inode flags used for atomic set/get
 406  */
 407 enum {
 408         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
 409         EXT4_INODE_UNRM         = 1,    /* Undelete */
 410         EXT4_INODE_COMPR        = 2,    /* Compress file */
 411         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
 412         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
 413         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
 414         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
 415         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
 416 /* Reserved for compression usage... */
 417         EXT4_INODE_DIRTY        = 8,
 418         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
 419         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
 420         EXT4_INODE_ECOMPR       = 11,   /* Compression error */
 421 /* End compression flags --- maybe not all used */
 422         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
 423         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
 424         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
 425         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
 426         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
 427         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
 428         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
 429         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
 430         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
 431         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
 432         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
 433 };
 434
 435 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
 436 #define CHECK_FLAG_VALUE(FLAG) if (!TEST_FLAG_VALUE(FLAG)) { \
 437         printk(KERN_EMERG "EXT4 flag fail: " #FLAG ": %d %d\n", \
 438                 EXT4_##FLAG##_FL, EXT4_INODE_##FLAG); BUG_ON(1); }
 439
 440 /*
 441  * Since it's pretty easy to mix up bit numbers and hex values, and we
 442  * can't do a compile-time test for ENUM values, we use a run-time
 443  * test to make sure that EXT4_XXX_FL is consistent with respect to
 444  * EXT4_INODE_XXX.  If all is well the printk and BUG_ON will all drop
 445  * out so it won't cost any extra space in the compiled kernel image.
 446  * But it's important that these values are the same, since we are
 447  * using EXT4_INODE_XXX to test for the flag values, but EXT4_XX_FL
 448  * must be consistent with the values of FS_XXX_FL defined in
 449  * include/linux/fs.h and the on-disk values found in ext2, ext3, and
 450  * ext4 filesystems, and of course the values defined in e2fsprogs.
 451  *
 452  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
 453  */
 454 static inline void ext4_check_flag_values(void)
 455 {
 456         CHECK_FLAG_VALUE(SECRM);
 457         CHECK_FLAG_VALUE(UNRM);
 458         CHECK_FLAG_VALUE(COMPR);
 459         CHECK_FLAG_VALUE(SYNC);
 460         CHECK_FLAG_VALUE(IMMUTABLE);
 461         CHECK_FLAG_VALUE(APPEND);
 462         CHECK_FLAG_VALUE(NODUMP);
 463         CHECK_FLAG_VALUE(NOATIME);
 464         CHECK_FLAG_VALUE(DIRTY);
 465         CHECK_FLAG_VALUE(COMPRBLK);
 466         CHECK_FLAG_VALUE(NOCOMPR);
 467         CHECK_FLAG_VALUE(ECOMPR);
 468         CHECK_FLAG_VALUE(INDEX);
 469         CHECK_FLAG_VALUE(IMAGIC);
 470         CHECK_FLAG_VALUE(JOURNAL_DATA);
 471         CHECK_FLAG_VALUE(NOTAIL);
 472         CHECK_FLAG_VALUE(DIRSYNC);
 473         CHECK_FLAG_VALUE(TOPDIR);
 474         CHECK_FLAG_VALUE(HUGE_FILE);
 475         CHECK_FLAG_VALUE(EXTENTS);
 476         CHECK_FLAG_VALUE(EA_INODE);
 477         CHECK_FLAG_VALUE(EOFBLOCKS);
 478         CHECK_FLAG_VALUE(RESERVED);
 479 }
 480
 481 /* Used to pass group descriptor data when online resize is done */
 482 struct ext4_new_group_input {
 483         __u32 group;            /* Group number for this data */
 484         __u64 block_bitmap;     /* Absolute block number of block bitmap */
 485         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
 486         __u64 inode_table;      /* Absolute block number of inode table start */
 487         __u32 blocks_count;     /* Total number of blocks in this group */
 488         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
 489         __u16 unused;
 490 };
 491
 492 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 493 struct compat_ext4_new_group_input {
 494         u32 group;
 495         compat_u64 block_bitmap;
 496         compat_u64 inode_bitmap;
 497         compat_u64 inode_table;
 498         u32 blocks_count;
 499         u16 reserved_blocks;
 500         u16 unused;
 501 };
 502 #endif
 503
 504 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
 505 struct ext4_new_group_data {
 506         __u32 group;
 507         __u64 block_bitmap;
 508         __u64 inode_bitmap;
 509         __u64 inode_table;
 510         __u32 blocks_count;
 511         __u16 reserved_blocks;
 512         __u16 unused;
 513         __u32 free_blocks_count;
 514 };
 515
 516 /*
 517  * Flags used by ext4_map_blocks()
 518  */
 519         /* Allocate any needed blocks and/or convert an unitialized
 520            extent to be an initialized ext4 */
 521 #define EXT4_GET_BLOCKS_CREATE                  0x0001
 522         /* Request the creation of an unitialized extent */
 523 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
 524 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
 525                                                  EXT4_GET_BLOCKS_CREATE)
 526         /* Caller is from the delayed allocation writeout path,
 527            so set the magic i_delalloc_reserve_flag after taking the
 528            inode allocation semaphore for */
 529 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
 530         /* caller is from the direct IO path, request to creation of an
 531         unitialized extents if not allocated, split the uninitialized
 532         extent if blocks has been preallocated already*/
 533 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
 534 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
 535 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
 536                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
 537         /* Convert extent to initialized after IO complete */
 538 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
 539                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
 540         /* Punch out blocks of an extent */
 541 #define EXT4_GET_BLOCKS_PUNCH_OUT_EXT           0x0020
 542         /* Don't normalize allocation size (used for fallocate) */
 543 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
 544         /* Request will not result in inode size update (user for fallocate) */
 545 #define EXT4_GET_BLOCKS_KEEP_SIZE               0x0080
 546
 547 /*
 548  * Flags used by ext4_free_blocks
 549  */
 550 #define EXT4_FREE_BLOCKS_METADATA       0x0001
 551 #define EXT4_FREE_BLOCKS_FORGET         0x0002
 552 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
 553 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
 554 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
 555 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
 556
 557 /*
 558  * Flags used by ext4_discard_partial_page_buffers
 559  */
 560 #define EXT4_DISCARD_PARTIAL_PG_ZERO_UNMAPPED   0x0001
 561
 562 /*
 563  * ioctl commands
 564  */
 565 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
 566 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
 567 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
 568 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
 569 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
 570 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
 571 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
 572 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
 573 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
 574 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
 575 #define EXT4_IOC_MIGRATE                _IO('f', 9)
 576  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
 577  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
 578 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
 579 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
 580
 581 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 582 /*
 583  * ioctl commands in 32 bit emulation
 584  */
 585 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
 586 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
 587 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
 588 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
 589 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
 590 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
 591 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
 592 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
 593 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
 594 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
 595 #endif
 596
 597 /* Max physical block we can address w/o extents */
 598 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
 599
 600 /*
 601  * Structure of an inode on the disk
 602  */
 603 struct ext4_inode {
 604         __le16  i_mode;         /* File mode */
 605         __le16  i_uid;          /* Low 16 bits of Owner Uid */
 606         __le32  i_size_lo;      /* Size in bytes */
 607         __le32  i_atime;        /* Access time */
 608         __le32  i_ctime;        /* Inode Change time */
 609         __le32  i_mtime;        /* Modification time */
 610         __le32  i_dtime;        /* Deletion Time */
 611         __le16  i_gid;          /* Low 16 bits of Group Id */
 612         __le16  i_links_count;  /* Links count */
 613         __le32  i_blocks_lo;    /* Blocks count */
 614         __le32  i_flags;        /* File flags */
 615         union {
 616                 struct {
 617                         __le32  l_i_version;
 618                 } linux1;
 619                 struct {
 620                         __u32  h_i_translator;
 621                 } hurd1;
 622                 struct {
 623                         __u32  m_i_reserved1;
 624                 } masix1;
 625         } osd1;                         /* OS dependent 1 */
 626         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
 627         __le32  i_generation;   /* File version (for NFS) */
 628         __le32  i_file_acl_lo;  /* File ACL */
 629         __le32  i_size_high;
 630         __le32  i_obso_faddr;   /* Obsoleted fragment address */
 631         union {
 632                 struct {
 633                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
 634                         __le16  l_i_file_acl_high;
 635                         __le16  l_i_uid_high;   /* these 2 fields */
 636                         __le16  l_i_gid_high;   /* were reserved2[0] */
 637                         __u32   l_i_reserved2;
 638                 } linux2;
 639                 struct {
 640                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 641                         __u16   h_i_mode_high;
 642                         __u16   h_i_uid_high;
 643                         __u16   h_i_gid_high;
 644                         __u32   h_i_author;
 645                 } hurd2;
 646                 struct {
 647                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 648                         __le16  m_i_file_acl_high;
 649                         __u32   m_i_reserved2[2];
 650                 } masix2;
 651         } osd2;                         /* OS dependent 2 */
 652         __le16  i_extra_isize;
 653         __le16  i_pad1;
 654         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
 655         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
 656         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
 657         __le32  i_crtime;       /* File Creation time */
 658         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
 659         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
 660 };
 661
 662 struct move_extent {
 663         __u32 reserved;         /* should be zero */
 664         __u32 donor_fd;         /* donor file descriptor */
 665         __u64 orig_start;       /* logical start offset in block for orig */
 666         __u64 donor_start;      /* logical start offset in block for donor */
 667         __u64 len;              /* block length to be moved */
 668         __u64 moved_len;        /* moved block length */
 669 };
 670
 671 #define EXT4_EPOCH_BITS 2
 672 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
 673 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
 674
 675 /*
 676  * Extended fields will fit into an inode if the filesystem was formatted
 677  * with large inodes (-I 256 or larger) and there are not currently any EAs
 678  * consuming all of the available space. For new inodes we always reserve
 679  * enough space for the kernel's known extended fields, but for inodes
 680  * created with an old kernel this might not have been the case. None of
 681  * the extended inode fields is critical for correct filesystem operation.
 682  * This macro checks if a certain field fits in the inode. Note that
 683  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
 684  */
 685 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
 686         ((offsetof(typeof(*ext4_inode), field) +        \
 687           sizeof((ext4_inode)->field))                  \
 688         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
 689             (einode)->i_extra_isize))                   \
 690
 691 static inline __le32 ext4_encode_extra_time(struct timespec *time)
 692 {
 693        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
 694                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
 695                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
 696 }
 697
 698 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
 699 {
 700        if (sizeof(time->tv_sec) > 4)
 701                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
 702                                << 32;
 703        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
 704 }
 705
 706 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
 707 do {                                                                           \
 708         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
 709         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 710                 (raw_inode)->xtime ## _extra =                                 \
 711                                 ext4_encode_extra_time(&(inode)->xtime);       \
 712 } while (0)
 713
 714 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
 715 do {                                                                           \
 716         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 717                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
 718         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 719                 (raw_inode)->xtime ## _extra =                                 \
 720                                 ext4_encode_extra_time(&(einode)->xtime);      \
 721 } while (0)
 722
 723 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
 724 do {                                                                           \
 725         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
 726         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 727                 ext4_decode_extra_time(&(inode)->xtime,                        \
 728                                        raw_inode->xtime ## _extra);            \
 729         else                                                                   \
 730                 (inode)->xtime.tv_nsec = 0;                                    \
 731 } while (0)
 732
 733 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
 734 do {                                                                           \
 735         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 736                 (einode)->xtime.tv_sec =                                       \
 737                         (signed)le32_to_cpu((raw_inode)->xtime);               \
 738         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 739                 ext4_decode_extra_time(&(einode)->xtime,                       \
 740                                        raw_inode->xtime ## _extra);            \
 741         else                                                                   \
 742                 (einode)->xtime.tv_nsec = 0;                                   \
 743 } while (0)
 744
 745 #define i_disk_version osd1.linux1.l_i_version
 746
 747 #if defined(__KERNEL__) || defined(__linux__)
 748 #define i_reserved1     osd1.linux1.l_i_reserved1
 749 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
 750 #define i_blocks_high   osd2.linux2.l_i_blocks_high
 751 #define i_uid_low       i_uid
 752 #define i_gid_low       i_gid
 753 #define i_uid_high      osd2.linux2.l_i_uid_high
 754 #define i_gid_high      osd2.linux2.l_i_gid_high
 755 #define i_reserved2     osd2.linux2.l_i_reserved2
 756
 757 #elif defined(__GNU__)
 758
 759 #define i_translator    osd1.hurd1.h_i_translator
 760 #define i_uid_high      osd2.hurd2.h_i_uid_high
 761 #define i_gid_high      osd2.hurd2.h_i_gid_high
 762 #define i_author        osd2.hurd2.h_i_author
 763
 764 #elif defined(__masix__)
 765
 766 #define i_reserved1     osd1.masix1.m_i_reserved1
 767 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
 768 #define i_reserved2     osd2.masix2.m_i_reserved2
 769
 770 #endif /* defined(__KERNEL__) || defined(__linux__) */
 771
 772 /*
 773  * storage for cached extent
 774  * If ec_len == 0, then the cache is invalid.
 775  * If ec_start == 0, then the cache represents a gap (null mapping)
 776  */
 777 struct ext4_ext_cache {
 778         ext4_fsblk_t    ec_start;
 779         ext4_lblk_t     ec_block;
 780         __u32           ec_len; /* must be 32bit to return holes */
 781 };
 782
 783 /*
 784  * fourth extended file system inode data in memory
 785  */
 786 struct ext4_inode_info {
 787         __le32  i_data[15];     /* unconverted */
 788         __u32   i_dtime;
 789         ext4_fsblk_t    i_file_acl;
 790
 791         /*
 792          * i_block_group is the number of the block group which contains
 793          * this file's inode.  Constant across the lifetime of the inode,
 794          * it is ued for making block allocation decisions - we try to
 795          * place a file's data blocks near its inode block, and new inodes
 796          * near to their parent directory's inode.
 797          */
 798         ext4_group_t    i_block_group;
 799         ext4_lblk_t     i_dir_start_lookup;
 800 #if (BITS_PER_LONG < 64)
 801         unsigned long   i_state_flags;          /* Dynamic state flags */
 802 #endif
 803         unsigned long   i_flags;
 804
 805 #ifdef CONFIG_EXT4_FS_XATTR
 806         /*
 807          * Extended attributes can be read independently of the main file
 808          * data. Taking i_mutex even when reading would cause contention
 809          * between readers of EAs and writers of regular file data, so
 810          * instead we synchronize on xattr_sem when reading or changing
 811          * EAs.
 812          */
 813         struct rw_semaphore xattr_sem;
 814 #endif
 815
 816         struct list_head i_orphan;      /* unlinked but open inodes */
 817
 818         /*
 819          * i_disksize keeps track of what the inode size is ON DISK, not
 820          * in memory.  During truncate, i_size is set to the new size by
 821          * the VFS prior to calling ext4_truncate(), but the filesystem won't
 822          * set i_disksize to 0 until the truncate is actually under way.
 823          *
 824          * The intent is that i_disksize always represents the blocks which
 825          * are used by this file.  This allows recovery to restart truncate
 826          * on orphans if we crash during truncate.  We actually write i_disksize
 827          * into the on-disk inode when writing inodes out, instead of i_size.
 828          *
 829          * The only time when i_disksize and i_size may be different is when
 830          * a truncate is in progress.  The only things which change i_disksize
 831          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
 832          */
 833         loff_t  i_disksize;
 834
 835         /*
 836          * i_data_sem is for serialising ext4_truncate() against
 837          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
 838          * data tree are chopped off during truncate. We can't do that in
 839          * ext4 because whenever we perform intermediate commits during
 840          * truncate, the inode and all the metadata blocks *must* be in a
 841          * consistent state which allows truncation of the orphans to restart
 842          * during recovery.  Hence we must fix the get_block-vs-truncate race
 843          * by other means, so we have i_data_sem.
 844          */
 845         struct rw_semaphore i_data_sem;
 846         struct inode vfs_inode;
 847         struct jbd2_inode *jinode;
 848
 849         struct ext4_ext_cache i_cached_extent;
 850         /*
 851          * File creation time. Its function is same as that of
 852          * struct timespec i_{a,c,m}time in the generic inode.
 853          */
 854         struct timespec i_crtime;
 855
 856         /* mballoc */
 857         struct list_head i_prealloc_list;
 858         spinlock_t i_prealloc_lock;
 859
 860         /* ialloc */
 861         ext4_group_t    i_last_alloc_group;
 862
 863         /* allocation reservation info for delalloc */
 864         /* In case of bigalloc, these refer to clusters rather than blocks */
 865         unsigned int i_reserved_data_blocks;
 866         unsigned int i_reserved_meta_blocks;
 867         unsigned int i_allocated_meta_blocks;
 868         ext4_lblk_t i_da_metadata_calc_last_lblock;
 869         int i_da_metadata_calc_len;
 870
 871         /* on-disk additional length */
 872         __u16 i_extra_isize;
 873
 874 #ifdef CONFIG_QUOTA
 875         /* quota space reservation, managed internally by quota code */
 876         qsize_t i_reserved_quota;
 877 #endif
 878
 879         /* completed IOs that might need unwritten extents handling */
 880         struct list_head i_completed_io_list;
 881         spinlock_t i_completed_io_lock;
 882         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
 883         /* current io_end structure for async DIO write*/
 884         ext4_io_end_t *cur_aio_dio;
 885         atomic_t i_aiodio_unwritten; /* Nr. of inflight conversions pending */
 886
 887         spinlock_t i_block_reservation_lock;
 888
 889         /*
 890          * Transactions that contain inode's metadata needed to complete
 891          * fsync and fdatasync, respectively.
 892          */
 893         tid_t i_sync_tid;
 894         tid_t i_datasync_tid;
 895 };
 896
 897 /*
 898  * File system states
 899  */
 900 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
 901 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
 902 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
 903
 904 /*
 905  * Misc. filesystem flags
 906  */
 907 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
 908 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
 909 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
 910
 911 /*
 912  * Mount flags
 913  */
 914 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
 915 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
 916 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
 917 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
 918 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
 919 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
 920 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
 921 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
 922 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
 923 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
 924 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
 925 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
 926 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
 927 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
 928 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
 929 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
 930 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
 931 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
 932 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
 933 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
 934 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
 935 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
 936 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
 937 #define EXT4_MOUNT_I_VERSION            0x2000000 /* i_version support */
 938 #define EXT4_MOUNT_MBLK_IO_SUBMIT       0x4000000 /* multi-block io submits */
 939 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
 940 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
 941 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
 942 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
 943 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
 944
 945 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
 946                                                       specified delalloc */
 947
 948 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
 949                                                 ~EXT4_MOUNT_##opt
 950 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
 951                                                 EXT4_MOUNT_##opt
 952 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
 953                                          EXT4_MOUNT_##opt)
 954
 955 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
 956                                                 ~EXT4_MOUNT2_##opt
 957 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
 958                                                 EXT4_MOUNT2_##opt
 959 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
 960                                          EXT4_MOUNT2_##opt)
 961
 962 #define ext4_set_bit                    __test_and_set_bit_le
 963 #define ext4_set_bit_atomic             ext2_set_bit_atomic
 964 #define ext4_clear_bit                  __test_and_clear_bit_le
 965 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
 966 #define ext4_test_bit                   test_bit_le
 967 #define ext4_find_first_zero_bit        find_first_zero_bit_le
 968 #define ext4_find_next_zero_bit         find_next_zero_bit_le
 969 #define ext4_find_next_bit              find_next_bit_le
 970
 971 extern void ext4_set_bits(void *bm, int cur, int len);
 972
 973 /*
 974  * Maximal mount counts between two filesystem checks
 975  */
 976 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
 977 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
 978
 979 /*
 980  * Behaviour when detecting errors
 981  */
 982 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
 983 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
 984 #define EXT4_ERRORS_PANIC               3       /* Panic */
 985 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
 986
 987 /*
 988  * Structure of the super block
 989  */
 990 struct ext4_super_block {
 991 /*00*/  __le32  s_inodes_count;         /* Inodes count */
 992         __le32  s_blocks_count_lo;      /* Blocks count */
 993         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
 994         __le32  s_free_blocks_count_lo; /* Free blocks count */
 995 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
 996         __le32  s_first_data_block;     /* First Data Block */
 997         __le32  s_log_block_size;       /* Block size */
 998         __le32  s_log_cluster_size;     /* Allocation cluster size */
 999 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1000         __le32  s_clusters_per_group;   /* # Clusters per group */
1001         __le32  s_inodes_per_group;     /* # Inodes per group */
1002         __le32  s_mtime;                /* Mount time */
1003 /*30*/  __le32  s_wtime;                /* Write time */
1004         __le16  s_mnt_count;            /* Mount count */
1005         __le16  s_max_mnt_count;        /* Maximal mount count */
1006         __le16  s_magic;                /* Magic signature */
1007         __le16  s_state;                /* File system state */
1008         __le16  s_errors;               /* Behaviour when detecting errors */
1009         __le16  s_minor_rev_level;      /* minor revision level */
1010 /*40*/  __le32  s_lastcheck;            /* time of last check */
1011         __le32  s_checkinterval;        /* max. time between checks */
1012         __le32  s_creator_os;           /* OS */
1013         __le32  s_rev_level;            /* Revision level */
1014 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1015         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1016         /*
1017          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1018          *
1019          * Note: the difference between the compatible feature set and
1020          * the incompatible feature set is that if there is a bit set
1021          * in the incompatible feature set that the kernel doesn't
1022          * know about, it should refuse to mount the filesystem.
1023          *
1024          * e2fsck's requirements are more strict; if it doesn't know
1025          * about a feature in either the compatible or incompatible
1026          * feature set, it must abort and not try to meddle with
1027          * things it doesn't understand...
1028          */
1029         __le32  s_first_ino;            /* First non-reserved inode */
1030         __le16  s_inode_size;           /* size of inode structure */
1031         __le16  s_block_group_nr;       /* block group # of this superblock */
1032         __le32  s_feature_compat;       /* compatible feature set */
1033 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1034         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1035 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1036 /*78*/  char    s_volume_name[16];      /* volume name */
1037 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1038 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1039         /*
1040          * Performance hints.  Directory preallocation should only
1041          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1042          */
1043         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1044         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1045         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1046         /*
1047          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1048          */
1049 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1050 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1051         __le32  s_journal_dev;          /* device number of journal file */
1052         __le32  s_last_orphan;          /* start of list of inodes to delete */
1053         __le32  s_hash_seed[4];         /* HTREE hash seed */
1054         __u8    s_def_hash_version;     /* Default hash version to use */
1055         __u8    s_jnl_backup_type;
1056         __le16  s_desc_size;            /* size of group descriptor */
1057 /*100*/ __le32  s_default_mount_opts;
1058         __le32  s_first_meta_bg;        /* First metablock block group */
1059         __le32  s_mkfs_time;            /* When the filesystem was created */
1060         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1061         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1062 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1063         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1064         __le32  s_free_blocks_count_hi; /* Free blocks count */
1065         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1066         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1067         __le32  s_flags;                /* Miscellaneous flags */
1068         __le16  s_raid_stride;          /* RAID stride */
1069         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1070         __le64  s_mmp_block;            /* Block for multi-mount protection */
1071         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1072         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1073         __u8    s_reserved_char_pad;
1074         __le16  s_reserved_pad;
1075         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1076         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1077         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1078         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1079                                               snapshot's future use */
1080         __le32  s_snapshot_list;        /* inode number of the head of the
1081                                            on-disk snapshot list */
1082 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1083         __le32  s_error_count;          /* number of fs errors */
1084         __le32  s_first_error_time;     /* first time an error happened */
1085         __le32  s_first_error_ino;      /* inode involved in first error */
1086         __le64  s_first_error_block;    /* block involved of first error */
1087         __u8    s_first_error_func[32]; /* function where the error happened */
1088         __le32  s_first_error_line;     /* line number where error happened */
1089         __le32  s_last_error_time;      /* most recent time of an error */
1090         __le32  s_last_error_ino;       /* inode involved in last error */
1091         __le32  s_last_error_line;      /* line number where error happened */
1092         __le64  s_last_error_block;     /* block involved of last error */
1093         __u8    s_last_error_func[32];  /* function where the error happened */
1094 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1095         __u8    s_mount_opts[64];
1096         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1097         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1098         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1099         __le32  s_reserved[109];        /* Padding to the end of the block */
1100 };
1101
1102 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1103
1104 #ifdef __KERNEL__
1105
1106 /*
1107  * run-time mount flags
1108  */
1109 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
1110 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
1111
1112 /*
1113  * fourth extended-fs super-block data in memory
1114  */
1115 struct ext4_sb_info {
1116         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1117         unsigned long s_inodes_per_block;/* Number of inodes per block */
1118         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1119         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1120         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1121         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1122         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1123         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1124         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1125         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1126         unsigned long s_overhead_last;  /* Last calculated overhead */
1127         unsigned long s_blocks_last;    /* Last seen block count */
1128         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1129         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1130         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1131         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1132         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1133         struct buffer_head **s_group_desc;
1134         unsigned int s_mount_opt;
1135         unsigned int s_mount_opt2;
1136         unsigned int s_mount_flags;
1137         ext4_fsblk_t s_sb_block;
1138         uid_t s_resuid;
1139         gid_t s_resgid;
1140         unsigned short s_mount_state;
1141         unsigned short s_pad;
1142         int s_addr_per_block_bits;
1143         int s_desc_per_block_bits;
1144         int s_inode_size;
1145         int s_first_ino;
1146         unsigned int s_inode_readahead_blks;
1147         unsigned int s_inode_goal;
1148         spinlock_t s_next_gen_lock;
1149         u32 s_next_generation;
1150         u32 s_hash_seed[4];
1151         int s_def_hash_version;
1152         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1153         struct percpu_counter s_freeclusters_counter;
1154         struct percpu_counter s_freeinodes_counter;
1155         struct percpu_counter s_dirs_counter;
1156         struct percpu_counter s_dirtyclusters_counter;
1157         struct blockgroup_lock *s_blockgroup_lock;
1158         struct proc_dir_entry *s_proc;
1159         struct kobject s_kobj;
1160         struct completion s_kobj_unregister;
1161
1162         /* Journaling */
1163         struct journal_s *s_journal;
1164         struct list_head s_orphan;
1165         struct mutex s_orphan_lock;
1166         unsigned long s_resize_flags;           /* Flags indicating if there
1167                                                    is a resizer */
1168         unsigned long s_commit_interval;
1169         u32 s_max_batch_time;
1170         u32 s_min_batch_time;
1171         struct block_device *journal_bdev;
1172 #ifdef CONFIG_QUOTA
1173         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
1174         int s_jquota_fmt;                       /* Format of quota to use */
1175 #endif
1176         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1177         struct rb_root system_blks;
1178
1179 #ifdef EXTENTS_STATS
1180         /* ext4 extents stats */
1181         unsigned long s_ext_min;
1182         unsigned long s_ext_max;
1183         unsigned long s_depth_max;
1184         spinlock_t s_ext_stats_lock;
1185         unsigned long s_ext_blocks;
1186         unsigned long s_ext_extents;
1187 #endif
1188         /* ext4 extent cache stats */
1189         unsigned long extent_cache_hits;
1190         unsigned long extent_cache_misses;
1191
1192         /* for buddy allocator */
1193         struct ext4_group_info ***s_group_info;
1194         struct inode *s_buddy_cache;
1195         spinlock_t s_md_lock;
1196         unsigned short *s_mb_offsets;
1197         unsigned int *s_mb_maxs;
1198
1199         /* tunables */
1200         unsigned long s_stripe;
1201         unsigned int s_mb_stream_request;
1202         unsigned int s_mb_max_to_scan;
1203         unsigned int s_mb_min_to_scan;
1204         unsigned int s_mb_stats;
1205         unsigned int s_mb_order2_reqs;
1206         unsigned int s_mb_group_prealloc;
1207         unsigned int s_max_writeback_mb_bump;
1208         /* where last allocation was done - for stream allocation */
1209         unsigned long s_mb_last_group;
1210         unsigned long s_mb_last_start;
1211
1212         /* stats for buddy allocator */
1213         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1214         atomic_t s_bal_success; /* we found long enough chunks */
1215         atomic_t s_bal_allocated;       /* in blocks */
1216         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1217         atomic_t s_bal_goals;   /* goal hits */
1218         atomic_t s_bal_breaks;  /* too long searches */
1219         atomic_t s_bal_2orders; /* 2^order hits */
1220         spinlock_t s_bal_lock;
1221         unsigned long s_mb_buddies_generated;
1222         unsigned long long s_mb_generation_time;
1223         atomic_t s_mb_lost_chunks;
1224         atomic_t s_mb_preallocated;
1225         atomic_t s_mb_discarded;
1226         atomic_t s_lock_busy;
1227
1228         /* locality groups */
1229         struct ext4_locality_group __percpu *s_locality_groups;
1230
1231         /* for write statistics */
1232         unsigned long s_sectors_written_start;
1233         u64 s_kbytes_written;
1234
1235         unsigned int s_log_groups_per_flex;
1236         struct flex_groups *s_flex_groups;
1237
1238         /* workqueue for dio unwritten */
1239         struct workqueue_struct *dio_unwritten_wq;
1240
1241         /* timer for periodic error stats printing */
1242         struct timer_list s_err_report;
1243
1244         /* Lazy inode table initialization info */
1245         struct ext4_li_request *s_li_request;
1246         /* Wait multiplier for lazy initialization thread */
1247         unsigned int s_li_wait_mult;
1248
1249         /* Kernel thread for multiple mount protection */
1250         struct task_struct *s_mmp_tsk;
1251
1252         /* record the last minlen when FITRIM is called. */
1253         atomic_t s_last_trim_minblks;
1254 };
1255
1256 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1257 {
1258         return sb->s_fs_info;
1259 }
1260 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1261 {
1262         return container_of(inode, struct ext4_inode_info, vfs_inode);
1263 }
1264
1265 static inline struct timespec ext4_current_time(struct inode *inode)
1266 {
1267         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1268                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1269 }
1270
1271 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1272 {
1273         return ino == EXT4_ROOT_INO ||
1274                 ino == EXT4_JOURNAL_INO ||
1275                 ino == EXT4_RESIZE_INO ||
1276                 (ino >= EXT4_FIRST_INO(sb) &&
1277                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1278 }
1279
1280 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
1281                                               struct ext4_io_end *io_end)
1282 {
1283         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
1284                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
1285                 atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten);
1286         }
1287 }
1288
1289 /*
1290  * Inode dynamic state flags
1291  */
1292 enum {
1293         EXT4_STATE_JDATA,               /* journaled data exists */
1294         EXT4_STATE_NEW,                 /* inode is newly created */
1295         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1296         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1297         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1298         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1299         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1300         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1301         EXT4_STATE_DELALLOC_RESERVED,   /* blks already reserved for delalloc */
1302 };
1303
1304 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1305 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1306 {                                                                       \
1307         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1308 }                                                                       \
1309 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1310 {                                                                       \
1311         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1312 }                                                                       \
1313 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1314 {                                                                       \
1315         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1316 }
1317
1318 EXT4_INODE_BIT_FNS(flag, flags, 0)
1319 #if (BITS_PER_LONG < 64)
1320 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1321
1322 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1323 {
1324         (ei)->i_state_flags = 0;
1325 }
1326 #else
1327 EXT4_INODE_BIT_FNS(state, flags, 32)
1328
1329 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1330 {
1331         /* We depend on the fact that callers will set i_flags */
1332 }
1333 #endif
1334 #else
1335 /* Assume that user mode programs are passing in an ext4fs superblock, not
1336  * a kernel struct super_block.  This will allow us to call the feature-test
1337  * macros from user land. */
1338 #define EXT4_SB(sb)     (sb)
1339 #endif
1340
1341 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1342
1343 /*
1344  * Codes for operating systems
1345  */
1346 #define EXT4_OS_LINUX           0
1347 #define EXT4_OS_HURD            1
1348 #define EXT4_OS_MASIX           2
1349 #define EXT4_OS_FREEBSD         3
1350 #define EXT4_OS_LITES           4
1351
1352 /*
1353  * Revision levels
1354  */
1355 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1356 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1357
1358 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1359 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1360
1361 #define EXT4_GOOD_OLD_INODE_SIZE 128
1362
1363 /*
1364  * Feature set definitions
1365  */
1366
1367 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1368         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1369 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1370         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1371 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1372         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1373 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1374         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1375 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1376         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1377 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1378         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1379 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1380         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1381 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1382         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1383 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1384         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1385
1386 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1387 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1388 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1389 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1390 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1391 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1392
1393 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1394 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1395 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1396 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1397 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1398 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1399 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1400 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1401 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1402
1403 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1404 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1405 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1406 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1407 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1408 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1409 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1410 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1411 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1412 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1413 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1414
1415 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1416 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1417                                          EXT4_FEATURE_INCOMPAT_META_BG)
1418 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1419                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1420                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1421
1422 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1423 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1424                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1425                                          EXT4_FEATURE_INCOMPAT_META_BG)
1426 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1427                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1428                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1429
1430 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1431 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1432                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1433                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1434                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1435                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1436                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1437                                          EXT4_FEATURE_INCOMPAT_MMP)
1438 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1439                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1440                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1441                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1442                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1443                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1444                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1445                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC)
1446
1447 /*
1448  * Default values for user and/or group using reserved blocks
1449  */
1450 #define EXT4_DEF_RESUID         0
1451 #define EXT4_DEF_RESGID         0
1452
1453 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1454
1455 /*
1456  * Default mount options
1457  */
1458 #define EXT4_DEFM_DEBUG         0x0001
1459 #define EXT4_DEFM_BSDGROUPS     0x0002
1460 #define EXT4_DEFM_XATTR_USER    0x0004
1461 #define EXT4_DEFM_ACL           0x0008
1462 #define EXT4_DEFM_UID16         0x0010
1463 #define EXT4_DEFM_JMODE         0x0060
1464 #define EXT4_DEFM_JMODE_DATA    0x0020
1465 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1466 #define EXT4_DEFM_JMODE_WBACK   0x0060
1467 #define EXT4_DEFM_NOBARRIER     0x0100
1468 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1469 #define EXT4_DEFM_DISCARD       0x0400
1470 #define EXT4_DEFM_NODELALLOC    0x0800
1471
1472 /*
1473  * Default journal batch times
1474  */
1475 #define EXT4_DEF_MIN_BATCH_TIME 0
1476 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1477
1478 /*
1479  * Minimum number of groups in a flexgroup before we separate out
1480  * directories into the first block group of a flexgroup
1481  */
1482 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1483
1484 /*
1485  * Structure of a directory entry
1486  */
1487 #define EXT4_NAME_LEN 255
1488
1489 struct ext4_dir_entry {
1490         __le32  inode;                  /* Inode number */
1491         __le16  rec_len;                /* Directory entry length */
1492         __le16  name_len;               /* Name length */
1493         char    name[EXT4_NAME_LEN];    /* File name */
1494 };
1495
1496 /*
1497  * The new version of the directory entry.  Since EXT4 structures are
1498  * stored in intel byte order, and the name_len field could never be
1499  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1500  * file_type field.
1501  */
1502 struct ext4_dir_entry_2 {
1503         __le32  inode;                  /* Inode number */
1504         __le16  rec_len;                /* Directory entry length */
1505         __u8    name_len;               /* Name length */
1506         __u8    file_type;
1507         char    name[EXT4_NAME_LEN];    /* File name */
1508 };
1509
1510 /*
1511  * Ext4 directory file types.  Only the low 3 bits are used.  The
1512  * other bits are reserved for now.
1513  */
1514 #define EXT4_FT_UNKNOWN         0
1515 #define EXT4_FT_REG_FILE        1
1516 #define EXT4_FT_DIR             2
1517 #define EXT4_FT_CHRDEV          3
1518 #define EXT4_FT_BLKDEV          4
1519 #define EXT4_FT_FIFO            5
1520 #define EXT4_FT_SOCK            6
1521 #define EXT4_FT_SYMLINK         7
1522
1523 #define EXT4_FT_MAX             8
1524
1525 /*
1526  * EXT4_DIR_PAD defines the directory entries boundaries
1527  *
1528  * NOTE: It must be a multiple of 4
1529  */
1530 #define EXT4_DIR_PAD                    4
1531 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1532 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1533                                          ~EXT4_DIR_ROUND)
1534 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1535
1536 /*
1537  * If we ever get support for fs block sizes > page_size, we'll need
1538  * to remove the #if statements in the next two functions...
1539  */
1540 static inline unsigned int
1541 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1542 {
1543         unsigned len = le16_to_cpu(dlen);
1544
1545 #if (PAGE_CACHE_SIZE >= 65536)
1546         if (len == EXT4_MAX_REC_LEN || len == 0)
1547                 return blocksize;
1548         return (len & 65532) | ((len & 3) << 16);
1549 #else
1550         return len;
1551 #endif
1552 }
1553
1554 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1555 {
1556         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1557                 BUG();
1558 #if (PAGE_CACHE_SIZE >= 65536)
1559         if (len < 65536)
1560                 return cpu_to_le16(len);
1561         if (len == blocksize) {
1562                 if (blocksize == 65536)
1563                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1564                 else
1565                         return cpu_to_le16(0);
1566         }
1567         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1568 #else
1569         return cpu_to_le16(len);
1570 #endif
1571 }
1572
1573 /*
1574  * Hash Tree Directory indexing
1575  * (c) Daniel Phillips, 2001
1576  */
1577
1578 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1579                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1580                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1581 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1582 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1583
1584 /* Legal values for the dx_root hash_version field: */
1585
1586 #define DX_HASH_LEGACY          0
1587 #define DX_HASH_HALF_MD4        1
1588 #define DX_HASH_TEA             2
1589 #define DX_HASH_LEGACY_UNSIGNED 3
1590 #define DX_HASH_HALF_MD4_UNSIGNED       4
1591 #define DX_HASH_TEA_UNSIGNED            5
1592
1593 #ifdef __KERNEL__
1594
1595 /* hash info structure used by the directory hash */
1596 struct dx_hash_info
1597 {
1598         u32             hash;
1599         u32             minor_hash;
1600         int             hash_version;
1601         u32             *seed;
1602 };
1603
1604 #define EXT4_HTREE_EOF  0x7fffffff
1605
1606 /*
1607  * Control parameters used by ext4_htree_next_block
1608  */
1609 #define HASH_NB_ALWAYS          1
1610
1611
1612 /*
1613  * Describe an inode's exact location on disk and in memory
1614  */
1615 struct ext4_iloc
1616 {
1617         struct buffer_head *bh;
1618         unsigned long offset;
1619         ext4_group_t block_group;
1620 };
1621
1622 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1623 {
1624         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1625 }
1626
1627 /*
1628  * This structure is stuffed into the struct file's private_data field
1629  * for directories.  It is where we put information so that we can do
1630  * readdir operations in hash tree order.
1631  */
1632 struct dir_private_info {
1633         struct rb_root  root;
1634         struct rb_node  *curr_node;
1635         struct fname    *extra_fname;
1636         loff_t          last_pos;
1637         __u32           curr_hash;
1638         __u32           curr_minor_hash;
1639         __u32           next_hash;
1640 };
1641
1642 /* calculate the first block number of the group */
1643 static inline ext4_fsblk_t
1644 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1645 {
1646         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1647                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1648 }
1649
1650 /*
1651  * Special error return code only used by dx_probe() and its callers.
1652  */
1653 #define ERR_BAD_DX_DIR  -75000
1654
1655 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1656                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1657
1658 /*
1659  * Timeout and state flag for lazy initialization inode thread.
1660  */
1661 #define EXT4_DEF_LI_WAIT_MULT                   10
1662 #define EXT4_DEF_LI_MAX_START_DELAY             5
1663 #define EXT4_LAZYINIT_QUIT                      0x0001
1664 #define EXT4_LAZYINIT_RUNNING                   0x0002
1665
1666 /*
1667  * Lazy inode table initialization info
1668  */
1669 struct ext4_lazy_init {
1670         unsigned long           li_state;
1671         struct list_head        li_request_list;
1672         struct mutex            li_list_mtx;
1673 };
1674
1675 struct ext4_li_request {
1676         struct super_block      *lr_super;
1677         struct ext4_sb_info     *lr_sbi;
1678         ext4_group_t            lr_next_group;
1679         struct list_head        lr_request;
1680         unsigned long           lr_next_sched;
1681         unsigned long           lr_timeout;
1682 };
1683
1684 struct ext4_features {
1685         struct kobject f_kobj;
1686         struct completion f_kobj_unregister;
1687 };
1688
1689 /*
1690  * This structure will be used for multiple mount protection. It will be
1691  * written into the block number saved in the s_mmp_block field in the
1692  * superblock. Programs that check MMP should assume that if
1693  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1694  * to use the filesystem, regardless of how old the timestamp is.
1695  */
1696 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
1697 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1698 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
1699 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
1700
1701 struct mmp_struct {
1702         __le32  mmp_magic;              /* Magic number for MMP */
1703         __le32  mmp_seq;                /* Sequence no. updated periodically */
1704
1705         /*
1706          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1707          * purposes and do not affect the correctness of the algorithm
1708          */
1709         __le64  mmp_time;               /* Time last updated */
1710         char    mmp_nodename[64];       /* Node which last updated MMP block */
1711         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
1712
1713         /*
1714          * mmp_check_interval is used to verify if the MMP block has been
1715          * updated on the block device. The value is updated based on the
1716          * maximum time to write the MMP block during an update cycle.
1717          */
1718         __le16  mmp_check_interval;
1719
1720         __le16  mmp_pad1;
1721         __le32  mmp_pad2[227];
1722 };
1723
1724 /* arguments passed to the mmp thread */
1725 struct mmpd_data {
1726         struct buffer_head *bh; /* bh from initial read_mmp_block() */
1727         struct super_block *sb;  /* super block of the fs */
1728 };
1729
1730 /*
1731  * Check interval multiplier
1732  * The MMP block is written every update interval and initially checked every
1733  * update interval x the multiplier (the value is then adapted based on the
1734  * write latency). The reason is that writes can be delayed under load and we
1735  * don't want readers to incorrectly assume that the filesystem is no longer
1736  * in use.
1737  */
1738 #define EXT4_MMP_CHECK_MULT             2UL
1739
1740 /*
1741  * Minimum interval for MMP checking in seconds.
1742  */
1743 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
1744
1745 /*
1746  * Maximum interval for MMP checking in seconds.
1747  */
1748 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
1749
1750 /*
1751  * Function prototypes
1752  */
1753
1754 /*
1755  * Ok, these declarations are also in <linux/kernel.h> but none of the
1756  * ext4 source programs needs to include it so they are duplicated here.
1757  */
1758 # define NORET_TYPE     /**/
1759 # define ATTRIB_NORET   __attribute__((noreturn))
1760 # define NORET_AND      noreturn,
1761
1762 /* bitmap.c */
1763 extern unsigned int ext4_count_free(struct buffer_head *, unsigned);
1764
1765 /* balloc.c */
1766 extern unsigned int ext4_block_group(struct super_block *sb,
1767                         ext4_fsblk_t blocknr);
1768 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1769                         ext4_fsblk_t blocknr);
1770 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1771 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1772                         ext4_group_t group);
1773 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1774                                          ext4_fsblk_t goal,
1775                                          unsigned int flags,
1776                                          unsigned long *count,
1777                                          int *errp);
1778 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
1779                                     s64 nclusters, unsigned int flags);
1780 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
1781 extern void ext4_check_blocks_bitmap(struct super_block *);
1782 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1783                                                     ext4_group_t block_group,
1784                                                     struct buffer_head ** bh);
1785 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1786 struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1787                                       ext4_group_t block_group);
1788 extern void ext4_init_block_bitmap(struct super_block *sb,
1789                                    struct buffer_head *bh,
1790                                    ext4_group_t group,
1791                                    struct ext4_group_desc *desc);
1792 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
1793                                               ext4_group_t block_group,
1794                                               struct ext4_group_desc *gdp);
1795 extern unsigned ext4_num_base_meta_clusters(struct super_block *sb,
1796                                             ext4_group_t block_group);
1797 extern unsigned ext4_num_overhead_clusters(struct super_block *sb,
1798                                            ext4_group_t block_group,
1799                                            struct ext4_group_desc *gdp);
1800 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
1801
1802 /* dir.c */
1803 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1804                                   struct file *,
1805                                   struct ext4_dir_entry_2 *,
1806                                   struct buffer_head *, unsigned int);
1807 #define ext4_check_dir_entry(dir, filp, de, bh, offset)                 \
1808         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
1809                                         (de), (bh), (offset)))
1810 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1811                                     __u32 minor_hash,
1812                                     struct ext4_dir_entry_2 *dirent);
1813 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1814
1815 /* fsync.c */
1816 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
1817 extern int ext4_flush_completed_IO(struct inode *);
1818
1819 /* hash.c */
1820 extern int ext4fs_dirhash(const char *name, int len, struct
1821                           dx_hash_info *hinfo);
1822
1823 /* ialloc.c */
1824 extern struct inode *ext4_new_inode(handle_t *, struct inode *, int,
1825                                     const struct qstr *qstr, __u32 goal,
1826                                     uid_t *owner);
1827 extern void ext4_free_inode(handle_t *, struct inode *);
1828 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1829 extern unsigned long ext4_count_free_inodes(struct super_block *);
1830 extern unsigned long ext4_count_dirs(struct super_block *);
1831 extern void ext4_check_inodes_bitmap(struct super_block *);
1832 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1833 extern int ext4_init_inode_table(struct super_block *sb,
1834                                  ext4_group_t group, int barrier);
1835
1836 /* mballoc.c */
1837 extern long ext4_mb_stats;
1838 extern long ext4_mb_max_to_scan;
1839 extern int ext4_mb_init(struct super_block *, int);
1840 extern int ext4_mb_release(struct super_block *);
1841 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1842                                 struct ext4_allocation_request *, int *);
1843 extern int ext4_mb_reserve_blocks(struct super_block *, int);
1844 extern void ext4_discard_preallocations(struct inode *);
1845 extern int __init ext4_init_mballoc(void);
1846 extern void ext4_exit_mballoc(void);
1847 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1848                              struct buffer_head *bh, ext4_fsblk_t block,
1849                              unsigned long count, int flags);
1850 extern int ext4_mb_add_groupinfo(struct super_block *sb,
1851                 ext4_group_t i, struct ext4_group_desc *desc);
1852 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
1853                                 ext4_fsblk_t block, unsigned long count);
1854 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
1855
1856 /* inode.c */
1857 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1858                                                 ext4_lblk_t, int, int *);
1859 struct buffer_head *ext4_bread(handle_t *, struct inode *,
1860                                                 ext4_lblk_t, int, int *);
1861 int ext4_get_block(struct inode *inode, sector_t iblock,
1862                                 struct buffer_head *bh_result, int create);
1863
1864 extern struct inode *ext4_iget(struct super_block *, unsigned long);
1865 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
1866 extern int  ext4_setattr(struct dentry *, struct iattr *);
1867 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1868                                 struct kstat *stat);
1869 extern void ext4_evict_inode(struct inode *);
1870 extern void ext4_clear_inode(struct inode *);
1871 extern int  ext4_sync_inode(handle_t *, struct inode *);
1872 extern void ext4_dirty_inode(struct inode *, int);
1873 extern int ext4_change_inode_journal_flag(struct inode *, int);
1874 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1875 extern int ext4_can_truncate(struct inode *inode);
1876 extern void ext4_truncate(struct inode *);
1877 extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length);
1878 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
1879 extern void ext4_set_inode_flags(struct inode *);
1880 extern void ext4_get_inode_flags(struct ext4_inode_info *);
1881 extern int ext4_alloc_da_blocks(struct inode *inode);
1882 extern void ext4_set_aops(struct inode *inode);
1883 extern int ext4_writepage_trans_blocks(struct inode *);
1884 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1885 extern int ext4_block_truncate_page(handle_t *handle,
1886                 struct address_space *mapping, loff_t from);
1887 extern int ext4_block_zero_page_range(handle_t *handle,
1888                 struct address_space *mapping, loff_t from, loff_t length);
1889 extern int ext4_discard_partial_page_buffers(handle_t *handle,
1890                 struct address_space *mapping, loff_t from,
1891                 loff_t length, int flags);
1892 extern int ext4_discard_partial_page_buffers_no_lock(handle_t *handle,
1893                 struct inode *inode, struct page *page, loff_t from,
1894                 loff_t length, int flags);
1895 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1896 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
1897 extern void ext4_da_update_reserve_space(struct inode *inode,
1898                                         int used, int quota_claim);
1899
1900 /* indirect.c */
1901 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
1902                                 struct ext4_map_blocks *map, int flags);
1903 extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
1904                                 const struct iovec *iov, loff_t offset,
1905                                 unsigned long nr_segs);
1906 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
1907 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks, int chunk);
1908 extern void ext4_ind_truncate(struct inode *inode);
1909
1910 /* ioctl.c */
1911 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1912 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1913
1914 /* migrate.c */
1915 extern int ext4_ext_migrate(struct inode *);
1916
1917 /* namei.c */
1918 extern int ext4_orphan_add(handle_t *, struct inode *);
1919 extern int ext4_orphan_del(handle_t *, struct inode *);
1920 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1921                                 __u32 start_minor_hash, __u32 *next_hash);
1922
1923 /* resize.c */
1924 extern int ext4_group_add(struct super_block *sb,
1925                                 struct ext4_new_group_data *input);
1926 extern int ext4_group_extend(struct super_block *sb,
1927                                 struct ext4_super_block *es,
1928                                 ext4_fsblk_t n_blocks_count);
1929
1930 /* super.c */
1931 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
1932 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
1933 extern void ext4_kvfree(void *ptr);
1934 extern __printf(4, 5)
1935 void __ext4_error(struct super_block *, const char *, unsigned int,
1936                   const char *, ...);
1937 #define ext4_error(sb, message...)      __ext4_error(sb, __func__,      \
1938                                                      __LINE__, ## message)
1939 extern __printf(5, 6)
1940 void ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
1941                       const char *, ...);
1942 extern __printf(5, 6)
1943 void ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
1944                      const char *, ...);
1945 extern void __ext4_std_error(struct super_block *, const char *,
1946                              unsigned int, int);
1947 extern __printf(4, 5)
1948 void __ext4_abort(struct super_block *, const char *, unsigned int,
1949                   const char *, ...);
1950 #define ext4_abort(sb, message...)      __ext4_abort(sb, __func__, \
1951                                                        __LINE__, ## message)
1952 extern __printf(4, 5)
1953 void __ext4_warning(struct super_block *, const char *, unsigned int,
1954                     const char *, ...);
1955 #define ext4_warning(sb, message...)    __ext4_warning(sb, __func__, \
1956                                                        __LINE__, ## message)
1957 extern __printf(3, 4)
1958 void ext4_msg(struct super_block *, const char *, const char *, ...);
1959 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
1960                            const char *, unsigned int, const char *);
1961 #define dump_mmp_msg(sb, mmp, msg)      __dump_mmp_msg(sb, mmp, __func__, \
1962                                                        __LINE__, msg)
1963 extern __printf(7, 8)
1964 void __ext4_grp_locked_error(const char *, unsigned int,
1965                              struct super_block *, ext4_group_t,
1966                              unsigned long, ext4_fsblk_t,
1967                              const char *, ...);
1968 #define ext4_grp_locked_error(sb, grp, message...) \
1969         __ext4_grp_locked_error(__func__, __LINE__, (sb), (grp), ## message)
1970 extern void ext4_update_dynamic_rev(struct super_block *sb);
1971 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1972                                         __u32 compat);
1973 extern int ext4_update_rocompat_feature(handle_t *handle,
1974                                         struct super_block *sb, __u32 rocompat);
1975 extern int ext4_update_incompat_feature(handle_t *handle,
1976                                         struct super_block *sb, __u32 incompat);
1977 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1978                                       struct ext4_group_desc *bg);
1979 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1980                                       struct ext4_group_desc *bg);
1981 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1982                                      struct ext4_group_desc *bg);
1983 extern __u32 ext4_free_group_clusters(struct super_block *sb,
1984                                       struct ext4_group_desc *bg);
1985 extern __u32 ext4_free_inodes_count(struct super_block *sb,
1986                                  struct ext4_group_desc *bg);
1987 extern __u32 ext4_used_dirs_count(struct super_block *sb,
1988                                 struct ext4_group_desc *bg);
1989 extern __u32 ext4_itable_unused_count(struct super_block *sb,
1990                                    struct ext4_group_desc *bg);
1991 extern void ext4_block_bitmap_set(struct super_block *sb,
1992                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1993 extern void ext4_inode_bitmap_set(struct super_block *sb,
1994                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1995 extern void ext4_inode_table_set(struct super_block *sb,
1996                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
1997 extern void ext4_free_group_clusters_set(struct super_block *sb,
1998                                          struct ext4_group_desc *bg,
1999                                          __u32 count);
2000 extern void ext4_free_inodes_set(struct super_block *sb,
2001                                 struct ext4_group_desc *bg, __u32 count);
2002 extern void ext4_used_dirs_set(struct super_block *sb,
2003                                 struct ext4_group_desc *bg, __u32 count);
2004 extern void ext4_itable_unused_set(struct super_block *sb,
2005                                    struct ext4_group_desc *bg, __u32 count);
2006 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
2007                                    struct ext4_group_desc *gdp);
2008 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
2009                                        struct ext4_group_desc *gdp);
2010
2011 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2012 {
2013         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2014                 le32_to_cpu(es->s_blocks_count_lo);
2015 }
2016
2017 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2018 {
2019         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2020                 le32_to_cpu(es->s_r_blocks_count_lo);
2021 }
2022
2023 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2024 {
2025         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2026                 le32_to_cpu(es->s_free_blocks_count_lo);
2027 }
2028
2029 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2030                                          ext4_fsblk_t blk)
2031 {
2032         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2033         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2034 }
2035
2036 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2037                                               ext4_fsblk_t blk)
2038 {
2039         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2040         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2041 }
2042
2043 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2044                                            ext4_fsblk_t blk)
2045 {
2046         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2047         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2048 }
2049
2050 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2051 {
2052         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2053                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2054                         le32_to_cpu(raw_inode->i_size_lo);
2055         else
2056                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2057 }
2058
2059 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2060 {
2061         raw_inode->i_size_lo = cpu_to_le32(i_size);
2062         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2063 }
2064
2065 static inline
2066 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2067                                             ext4_group_t group)
2068 {
2069          struct ext4_group_info ***grp_info;
2070          long indexv, indexh;
2071          grp_info = EXT4_SB(sb)->s_group_info;
2072          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2073          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2074          return grp_info[indexv][indexh];
2075 }
2076
2077 /*
2078  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2079  * the locking protocol documented in the comments of ext4_group_add()
2080  * in resize.c
2081  */
2082 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2083 {
2084         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2085
2086         smp_rmb();
2087         return ngroups;
2088 }
2089
2090 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2091                                              ext4_group_t block_group)
2092 {
2093         return block_group >> sbi->s_log_groups_per_flex;
2094 }
2095
2096 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2097 {
2098         return 1 << sbi->s_log_groups_per_flex;
2099 }
2100
2101 #define ext4_std_error(sb, errno)                               \
2102 do {                                                            \
2103         if ((errno))                                            \
2104                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2105 } while (0)
2106
2107 #ifdef CONFIG_SMP
2108 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2109  * counters. So we need to make sure we have free clusters more
2110  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2111  */
2112 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2113 #else
2114 #define EXT4_FREECLUSTERS_WATERMARK 0
2115 #endif
2116
2117 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2118 {
2119         /*
2120          * XXX: replace with spinlock if seen contended -bzzz
2121          */
2122         down_write(&EXT4_I(inode)->i_data_sem);
2123         if (newsize > EXT4_I(inode)->i_disksize)
2124                 EXT4_I(inode)->i_disksize = newsize;
2125         up_write(&EXT4_I(inode)->i_data_sem);
2126         return ;
2127 }
2128
2129 struct ext4_group_info {
2130         unsigned long   bb_state;
2131         struct rb_root  bb_free_root;
2132         ext4_grpblk_t   bb_first_free;  /* first free block */
2133         ext4_grpblk_t   bb_free;        /* total free blocks */
2134         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2135         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2136         struct          list_head bb_prealloc_list;
2137 #ifdef DOUBLE_CHECK
2138         void            *bb_bitmap;
2139 #endif
2140         struct rw_semaphore alloc_sem;
2141         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2142                                          * regions, index is order.
2143                                          * bb_counters[3] = 5 means
2144                                          * 5 free 8-block regions. */
2145 };
2146
2147 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2148 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2149
2150 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2151         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2152
2153 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2154         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2155 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2156         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2157 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2158         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2159
2160 #define EXT4_MAX_CONTENTION             8
2161 #define EXT4_CONTENTION_THRESHOLD       2
2162
2163 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2164                                               ext4_group_t group)
2165 {
2166         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2167 }
2168
2169 /*
2170  * Returns true if the filesystem is busy enough that attempts to
2171  * access the block group locks has run into contention.
2172  */
2173 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2174 {
2175         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2176 }
2177
2178 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2179 {
2180         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2181         if (spin_trylock(lock))
2182                 /*
2183                  * We're able to grab the lock right away, so drop the
2184                  * lock contention counter.
2185                  */
2186                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2187         else {
2188                 /*
2189                  * The lock is busy, so bump the contention counter,
2190                  * and then wait on the spin lock.
2191                  */
2192                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2193                                   EXT4_MAX_CONTENTION);
2194                 spin_lock(lock);
2195         }
2196 }
2197
2198 static inline void ext4_unlock_group(struct super_block *sb,
2199                                         ext4_group_t group)
2200 {
2201         spin_unlock(ext4_group_lock_ptr(sb, group));
2202 }
2203
2204 static inline void ext4_mark_super_dirty(struct super_block *sb)
2205 {
2206         if (EXT4_SB(sb)->s_journal == NULL)
2207                 sb->s_dirt =1;
2208 }
2209
2210 /*
2211  * Block validity checking
2212  */
2213 #define ext4_check_indirect_blockref(inode, bh)                         \
2214         ext4_check_blockref(__func__, __LINE__, inode,                  \
2215                             (__le32 *)(bh)->b_data,                     \
2216                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2217
2218 #define ext4_ind_check_inode(inode)                                     \
2219         ext4_check_blockref(__func__, __LINE__, inode,                  \
2220                             EXT4_I(inode)->i_data,                      \
2221                             EXT4_NDIR_BLOCKS)
2222
2223 /*
2224  * Inodes and files operations
2225  */
2226
2227 /* dir.c */
2228 extern const struct file_operations ext4_dir_operations;
2229
2230 /* file.c */
2231 extern const struct inode_operations ext4_file_inode_operations;
2232 extern const struct file_operations ext4_file_operations;
2233 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2234
2235 /* namei.c */
2236 extern const struct inode_operations ext4_dir_inode_operations;
2237 extern const struct inode_operations ext4_special_inode_operations;
2238 extern struct dentry *ext4_get_parent(struct dentry *child);
2239
2240 /* symlink.c */
2241 extern const struct inode_operations ext4_symlink_inode_operations;
2242 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2243
2244 /* block_validity */
2245 extern void ext4_release_system_zone(struct super_block *sb);
2246 extern int ext4_setup_system_zone(struct super_block *sb);
2247 extern int __init ext4_init_system_zone(void);
2248 extern void ext4_exit_system_zone(void);
2249 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2250                                  ext4_fsblk_t start_blk,
2251                                  unsigned int count);
2252 extern int ext4_check_blockref(const char *, unsigned int,
2253                                struct inode *, __le32 *, unsigned int);
2254
2255 /* extents.c */
2256 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2257 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2258 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
2259                                        int chunk);
2260 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2261                                struct ext4_map_blocks *map, int flags);
2262 extern void ext4_ext_truncate(struct inode *);
2263 extern int ext4_ext_punch_hole(struct file *file, loff_t offset,
2264                                 loff_t length);
2265 extern void ext4_ext_init(struct super_block *);
2266 extern void ext4_ext_release(struct super_block *);
2267 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2268                           loff_t len);
2269 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
2270                           ssize_t len);
2271 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2272                            struct ext4_map_blocks *map, int flags);
2273 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2274                         __u64 start, __u64 len);
2275 /* move_extent.c */
2276 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2277                              __u64 start_orig, __u64 start_donor,
2278                              __u64 len, __u64 *moved_len);
2279
2280 /* page-io.c */
2281 extern int __init ext4_init_pageio(void);
2282 extern void ext4_exit_pageio(void);
2283 extern void ext4_ioend_wait(struct inode *);
2284 extern void ext4_free_io_end(ext4_io_end_t *io);
2285 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2286 extern int ext4_end_io_nolock(ext4_io_end_t *io);
2287 extern void ext4_io_submit(struct ext4_io_submit *io);
2288 extern int ext4_bio_write_page(struct ext4_io_submit *io,
2289                                struct page *page,
2290                                int len,
2291                                struct writeback_control *wbc);
2292
2293 /* mmp.c */
2294 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2295
2296 /* BH_Uninit flag: blocks are allocated but uninitialized on disk */
2297 enum ext4_state_bits {
2298         BH_Uninit       /* blocks are allocated but uninitialized on disk */
2299           = BH_JBDPrivateStart,
2300         BH_AllocFromCluster,    /* allocated blocks were part of already
2301                                  * allocated cluster. Note that this flag will
2302                                  * never, ever appear in a buffer_head's state
2303                                  * flag. See EXT4_MAP_FROM_CLUSTER to see where
2304                                  * this is used. */
2305         BH_Da_Mapped,   /* Delayed allocated block that now has a mapping. This
2306                          * flag is set when ext4_map_blocks is called on a
2307                          * delayed allocated block to get its real mapping. */
2308 };
2309
2310 BUFFER_FNS(Uninit, uninit)
2311 TAS_BUFFER_FNS(Uninit, uninit)
2312 BUFFER_FNS(Da_Mapped, da_mapped)
2313
2314 /*
2315  * Add new method to test wether block and inode bitmaps are properly
2316  * initialized. With uninit_bg reading the block from disk is not enough
2317  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2318  */
2319 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2320
2321 static inline int bitmap_uptodate(struct buffer_head *bh)
2322 {
2323         return (buffer_uptodate(bh) &&
2324                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2325 }
2326 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2327 {
2328         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2329 }
2330
2331 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2332
2333 /* For ioend & aio unwritten conversion wait queues */
2334 #define EXT4_WQ_HASH_SZ         37
2335 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
2336                                             EXT4_WQ_HASH_SZ])
2337 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
2338                                              EXT4_WQ_HASH_SZ])
2339 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2340 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2341
2342 #define EXT4_RESIZING   0
2343 extern int ext4_resize_begin(struct super_block *sb);
2344 extern void ext4_resize_end(struct super_block *sb);
2345
2346 #endif  /* __KERNEL__ */
2347
2348 #include "ext4_extents.h"
2349
2350 #endif  /* _EXT4_H */