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 /* Mask out the low bits to get the starting block of the cluster */
 282 #define EXT4_PBLK_CMASK(s, pblk) ((pblk) &                              \
 283                                   ~((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
 284 #define EXT4_LBLK_CMASK(s, lblk) ((lblk) &                              \
 285                                   ~((ext4_lblk_t) (s)->s_cluster_ratio - 1))
 286 /* Get the cluster offset */
 287 #define EXT4_PBLK_COFF(s, pblk) ((pblk) &                               \
 288                                  ((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
 289 #define EXT4_LBLK_COFF(s, lblk) ((lblk) &                               \
 290                                  ((ext4_lblk_t) (s)->s_cluster_ratio - 1))
 291
 292 /*
 293  * Structure of a blocks group descriptor
 294  */
 295 struct ext4_group_desc
 296 {
 297         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
 298         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
 299         __le32  bg_inode_table_lo;      /* Inodes table block */
 300         __le16  bg_free_blocks_count_lo;/* Free blocks count */
 301         __le16  bg_free_inodes_count_lo;/* Free inodes count */
 302         __le16  bg_used_dirs_count_lo;  /* Directories count */
 303         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
 304         __u32   bg_reserved[2];         /* Likely block/inode bitmap checksum */
 305         __le16  bg_itable_unused_lo;    /* Unused inodes count */
 306         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
 307         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
 308         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
 309         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
 310         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
 311         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
 312         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
 313         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
 314         __u32   bg_reserved2[3];
 315 };
 316
 317 /*
 318  * Structure of a flex block group info
 319  */
 320
 321 struct flex_groups {
 322         atomic64_t      free_clusters;
 323         atomic_t        free_inodes;
 324         atomic_t        used_dirs;
 325 };
 326
 327 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
 328 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
 329 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
 330
 331 /*
 332  * Macro-instructions used to manage group descriptors
 333  */
 334 #define EXT4_MIN_DESC_SIZE              32
 335 #define EXT4_MIN_DESC_SIZE_64BIT        64
 336 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
 337 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
 338 #ifdef __KERNEL__
 339 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
 340 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
 341 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
 342 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
 343 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
 344 #else
 345 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
 346 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
 347 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
 348 #endif
 349
 350 /*
 351  * Constants relative to the data blocks
 352  */
 353 #define EXT4_NDIR_BLOCKS                12
 354 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
 355 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
 356 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
 357 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
 358
 359 /*
 360  * Inode flags
 361  */
 362 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
 363 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
 364 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
 365 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
 366 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
 367 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
 368 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
 369 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
 370 /* Reserved for compression usage... */
 371 #define EXT4_DIRTY_FL                   0x00000100
 372 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
 373 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
 374 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
 375 /* End compression flags --- maybe not all used */
 376 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
 377 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
 378 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
 379 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
 380 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
 381 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
 382 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
 383 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
 384 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
 385 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
 386 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
 387
 388 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
 389 #define EXT4_FL_USER_MODIFIABLE         0x004B80FF /* User modifiable flags */
 390
 391 /* Flags that should be inherited by new inodes from their parent. */
 392 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
 393                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
 394                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
 395                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
 396
 397 /* Flags that are appropriate for regular files (all but dir-specific ones). */
 398 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
 399
 400 /* Flags that are appropriate for non-directories/regular files. */
 401 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
 402
 403 /* Mask out flags that are inappropriate for the given type of inode. */
 404 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
 405 {
 406         if (S_ISDIR(mode))
 407                 return flags;
 408         else if (S_ISREG(mode))
 409                 return flags & EXT4_REG_FLMASK;
 410         else
 411                 return flags & EXT4_OTHER_FLMASK;
 412 }
 413
 414 /*
 415  * Inode flags used for atomic set/get
 416  */
 417 enum {
 418         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
 419         EXT4_INODE_UNRM         = 1,    /* Undelete */
 420         EXT4_INODE_COMPR        = 2,    /* Compress file */
 421         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
 422         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
 423         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
 424         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
 425         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
 426 /* Reserved for compression usage... */
 427         EXT4_INODE_DIRTY        = 8,
 428         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
 429         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
 430         EXT4_INODE_ECOMPR       = 11,   /* Compression error */
 431 /* End compression flags --- maybe not all used */
 432         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
 433         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
 434         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
 435         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
 436         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
 437         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
 438         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
 439         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
 440         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
 441         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
 442         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
 443 };
 444
 445 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
 446 #define CHECK_FLAG_VALUE(FLAG) if (!TEST_FLAG_VALUE(FLAG)) { \
 447         printk(KERN_EMERG "EXT4 flag fail: " #FLAG ": %d %d\n", \
 448                 EXT4_##FLAG##_FL, EXT4_INODE_##FLAG); BUG_ON(1); }
 449
 450 /*
 451  * Since it's pretty easy to mix up bit numbers and hex values, and we
 452  * can't do a compile-time test for ENUM values, we use a run-time
 453  * test to make sure that EXT4_XXX_FL is consistent with respect to
 454  * EXT4_INODE_XXX.  If all is well the printk and BUG_ON will all drop
 455  * out so it won't cost any extra space in the compiled kernel image.
 456  * But it's important that these values are the same, since we are
 457  * using EXT4_INODE_XXX to test for the flag values, but EXT4_XX_FL
 458  * must be consistent with the values of FS_XXX_FL defined in
 459  * include/linux/fs.h and the on-disk values found in ext2, ext3, and
 460  * ext4 filesystems, and of course the values defined in e2fsprogs.
 461  *
 462  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
 463  */
 464 static inline void ext4_check_flag_values(void)
 465 {
 466         CHECK_FLAG_VALUE(SECRM);
 467         CHECK_FLAG_VALUE(UNRM);
 468         CHECK_FLAG_VALUE(COMPR);
 469         CHECK_FLAG_VALUE(SYNC);
 470         CHECK_FLAG_VALUE(IMMUTABLE);
 471         CHECK_FLAG_VALUE(APPEND);
 472         CHECK_FLAG_VALUE(NODUMP);
 473         CHECK_FLAG_VALUE(NOATIME);
 474         CHECK_FLAG_VALUE(DIRTY);
 475         CHECK_FLAG_VALUE(COMPRBLK);
 476         CHECK_FLAG_VALUE(NOCOMPR);
 477         CHECK_FLAG_VALUE(ECOMPR);
 478         CHECK_FLAG_VALUE(INDEX);
 479         CHECK_FLAG_VALUE(IMAGIC);
 480         CHECK_FLAG_VALUE(JOURNAL_DATA);
 481         CHECK_FLAG_VALUE(NOTAIL);
 482         CHECK_FLAG_VALUE(DIRSYNC);
 483         CHECK_FLAG_VALUE(TOPDIR);
 484         CHECK_FLAG_VALUE(HUGE_FILE);
 485         CHECK_FLAG_VALUE(EXTENTS);
 486         CHECK_FLAG_VALUE(EA_INODE);
 487         CHECK_FLAG_VALUE(EOFBLOCKS);
 488         CHECK_FLAG_VALUE(RESERVED);
 489 }
 490
 491 /* Used to pass group descriptor data when online resize is done */
 492 struct ext4_new_group_input {
 493         __u32 group;            /* Group number for this data */
 494         __u64 block_bitmap;     /* Absolute block number of block bitmap */
 495         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
 496         __u64 inode_table;      /* Absolute block number of inode table start */
 497         __u32 blocks_count;     /* Total number of blocks in this group */
 498         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
 499         __u16 unused;
 500 };
 501
 502 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 503 struct compat_ext4_new_group_input {
 504         u32 group;
 505         compat_u64 block_bitmap;
 506         compat_u64 inode_bitmap;
 507         compat_u64 inode_table;
 508         u32 blocks_count;
 509         u16 reserved_blocks;
 510         u16 unused;
 511 };
 512 #endif
 513
 514 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
 515 struct ext4_new_group_data {
 516         __u32 group;
 517         __u64 block_bitmap;
 518         __u64 inode_bitmap;
 519         __u64 inode_table;
 520         __u32 blocks_count;
 521         __u16 reserved_blocks;
 522         __u16 unused;
 523         __u32 free_blocks_count;
 524 };
 525
 526 /*
 527  * Flags used by ext4_map_blocks()
 528  */
 529         /* Allocate any needed blocks and/or convert an unitialized
 530            extent to be an initialized ext4 */
 531 #define EXT4_GET_BLOCKS_CREATE                  0x0001
 532         /* Request the creation of an unitialized extent */
 533 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
 534 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
 535                                                  EXT4_GET_BLOCKS_CREATE)
 536         /* Caller is from the delayed allocation writeout path,
 537            so set the magic i_delalloc_reserve_flag after taking the
 538            inode allocation semaphore for */
 539 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
 540         /* caller is from the direct IO path, request to creation of an
 541         unitialized extents if not allocated, split the uninitialized
 542         extent if blocks has been preallocated already*/
 543 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
 544 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
 545 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
 546                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
 547         /* Convert extent to initialized after IO complete */
 548 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
 549                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
 550         /* Punch out blocks of an extent */
 551 #define EXT4_GET_BLOCKS_PUNCH_OUT_EXT           0x0020
 552         /* Don't normalize allocation size (used for fallocate) */
 553 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
 554         /* Request will not result in inode size update (user for fallocate) */
 555 #define EXT4_GET_BLOCKS_KEEP_SIZE               0x0080
 556
 557 /*
 558  * Flags used by ext4_free_blocks
 559  */
 560 #define EXT4_FREE_BLOCKS_METADATA       0x0001
 561 #define EXT4_FREE_BLOCKS_FORGET         0x0002
 562 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
 563 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
 564 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
 565 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
 566
 567 /*
 568  * Flags used by ext4_discard_partial_page_buffers
 569  */
 570 #define EXT4_DISCARD_PARTIAL_PG_ZERO_UNMAPPED   0x0001
 571
 572 /*
 573  * ioctl commands
 574  */
 575 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
 576 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
 577 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
 578 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
 579 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
 580 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
 581 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
 582 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
 583 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
 584 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
 585 #define EXT4_IOC_MIGRATE                _IO('f', 9)
 586  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
 587  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
 588 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
 589 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
 590
 591 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 592 /*
 593  * ioctl commands in 32 bit emulation
 594  */
 595 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
 596 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
 597 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
 598 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
 599 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
 600 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
 601 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
 602 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
 603 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
 604 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
 605 #endif
 606
 607 /* Max physical block we can address w/o extents */
 608 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
 609
 610 /*
 611  * Structure of an inode on the disk
 612  */
 613 struct ext4_inode {
 614         __le16  i_mode;         /* File mode */
 615         __le16  i_uid;          /* Low 16 bits of Owner Uid */
 616         __le32  i_size_lo;      /* Size in bytes */
 617         __le32  i_atime;        /* Access time */
 618         __le32  i_ctime;        /* Inode Change time */
 619         __le32  i_mtime;        /* Modification time */
 620         __le32  i_dtime;        /* Deletion Time */
 621         __le16  i_gid;          /* Low 16 bits of Group Id */
 622         __le16  i_links_count;  /* Links count */
 623         __le32  i_blocks_lo;    /* Blocks count */
 624         __le32  i_flags;        /* File flags */
 625         union {
 626                 struct {
 627                         __le32  l_i_version;
 628                 } linux1;
 629                 struct {
 630                         __u32  h_i_translator;
 631                 } hurd1;
 632                 struct {
 633                         __u32  m_i_reserved1;
 634                 } masix1;
 635         } osd1;                         /* OS dependent 1 */
 636         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
 637         __le32  i_generation;   /* File version (for NFS) */
 638         __le32  i_file_acl_lo;  /* File ACL */
 639         __le32  i_size_high;
 640         __le32  i_obso_faddr;   /* Obsoleted fragment address */
 641         union {
 642                 struct {
 643                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
 644                         __le16  l_i_file_acl_high;
 645                         __le16  l_i_uid_high;   /* these 2 fields */
 646                         __le16  l_i_gid_high;   /* were reserved2[0] */
 647                         __u32   l_i_reserved2;
 648                 } linux2;
 649                 struct {
 650                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 651                         __u16   h_i_mode_high;
 652                         __u16   h_i_uid_high;
 653                         __u16   h_i_gid_high;
 654                         __u32   h_i_author;
 655                 } hurd2;
 656                 struct {
 657                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 658                         __le16  m_i_file_acl_high;
 659                         __u32   m_i_reserved2[2];
 660                 } masix2;
 661         } osd2;                         /* OS dependent 2 */
 662         __le16  i_extra_isize;
 663         __le16  i_pad1;
 664         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
 665         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
 666         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
 667         __le32  i_crtime;       /* File Creation time */
 668         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
 669         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
 670 };
 671
 672 struct move_extent {
 673         __u32 reserved;         /* should be zero */
 674         __u32 donor_fd;         /* donor file descriptor */
 675         __u64 orig_start;       /* logical start offset in block for orig */
 676         __u64 donor_start;      /* logical start offset in block for donor */
 677         __u64 len;              /* block length to be moved */
 678         __u64 moved_len;        /* moved block length */
 679 };
 680
 681 #define EXT4_EPOCH_BITS 2
 682 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
 683 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
 684
 685 /*
 686  * Extended fields will fit into an inode if the filesystem was formatted
 687  * with large inodes (-I 256 or larger) and there are not currently any EAs
 688  * consuming all of the available space. For new inodes we always reserve
 689  * enough space for the kernel's known extended fields, but for inodes
 690  * created with an old kernel this might not have been the case. None of
 691  * the extended inode fields is critical for correct filesystem operation.
 692  * This macro checks if a certain field fits in the inode. Note that
 693  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
 694  */
 695 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
 696         ((offsetof(typeof(*ext4_inode), field) +        \
 697           sizeof((ext4_inode)->field))                  \
 698         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
 699             (einode)->i_extra_isize))                   \
 700
 701 static inline __le32 ext4_encode_extra_time(struct timespec *time)
 702 {
 703        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
 704                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
 705                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
 706 }
 707
 708 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
 709 {
 710        if (sizeof(time->tv_sec) > 4)
 711                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
 712                                << 32;
 713        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
 714 }
 715
 716 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
 717 do {                                                                           \
 718         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
 719         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 720                 (raw_inode)->xtime ## _extra =                                 \
 721                                 ext4_encode_extra_time(&(inode)->xtime);       \
 722 } while (0)
 723
 724 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
 725 do {                                                                           \
 726         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 727                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
 728         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 729                 (raw_inode)->xtime ## _extra =                                 \
 730                                 ext4_encode_extra_time(&(einode)->xtime);      \
 731 } while (0)
 732
 733 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
 734 do {                                                                           \
 735         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
 736         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 737                 ext4_decode_extra_time(&(inode)->xtime,                        \
 738                                        raw_inode->xtime ## _extra);            \
 739         else                                                                   \
 740                 (inode)->xtime.tv_nsec = 0;                                    \
 741 } while (0)
 742
 743 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
 744 do {                                                                           \
 745         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 746                 (einode)->xtime.tv_sec =                                       \
 747                         (signed)le32_to_cpu((raw_inode)->xtime);               \
 748         else                                                                   \
 749                 (einode)->xtime.tv_sec = 0;                                    \
 750         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 751                 ext4_decode_extra_time(&(einode)->xtime,                       \
 752                                        raw_inode->xtime ## _extra);            \
 753         else                                                                   \
 754                 (einode)->xtime.tv_nsec = 0;                                   \
 755 } while (0)
 756
 757 #define i_disk_version osd1.linux1.l_i_version
 758
 759 #if defined(__KERNEL__) || defined(__linux__)
 760 #define i_reserved1     osd1.linux1.l_i_reserved1
 761 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
 762 #define i_blocks_high   osd2.linux2.l_i_blocks_high
 763 #define i_uid_low       i_uid
 764 #define i_gid_low       i_gid
 765 #define i_uid_high      osd2.linux2.l_i_uid_high
 766 #define i_gid_high      osd2.linux2.l_i_gid_high
 767 #define i_reserved2     osd2.linux2.l_i_reserved2
 768
 769 #elif defined(__GNU__)
 770
 771 #define i_translator    osd1.hurd1.h_i_translator
 772 #define i_uid_high      osd2.hurd2.h_i_uid_high
 773 #define i_gid_high      osd2.hurd2.h_i_gid_high
 774 #define i_author        osd2.hurd2.h_i_author
 775
 776 #elif defined(__masix__)
 777
 778 #define i_reserved1     osd1.masix1.m_i_reserved1
 779 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
 780 #define i_reserved2     osd2.masix2.m_i_reserved2
 781
 782 #endif /* defined(__KERNEL__) || defined(__linux__) */
 783
 784 /*
 785  * storage for cached extent
 786  * If ec_len == 0, then the cache is invalid.
 787  * If ec_start == 0, then the cache represents a gap (null mapping)
 788  */
 789 struct ext4_ext_cache {
 790         ext4_fsblk_t    ec_start;
 791         ext4_lblk_t     ec_block;
 792         __u32           ec_len; /* must be 32bit to return holes */
 793 };
 794
 795 /*
 796  * fourth extended file system inode data in memory
 797  */
 798 struct ext4_inode_info {
 799         __le32  i_data[15];     /* unconverted */
 800         __u32   i_dtime;
 801         ext4_fsblk_t    i_file_acl;
 802
 803         /*
 804          * i_block_group is the number of the block group which contains
 805          * this file's inode.  Constant across the lifetime of the inode,
 806          * it is ued for making block allocation decisions - we try to
 807          * place a file's data blocks near its inode block, and new inodes
 808          * near to their parent directory's inode.
 809          */
 810         ext4_group_t    i_block_group;
 811         ext4_lblk_t     i_dir_start_lookup;
 812 #if (BITS_PER_LONG < 64)
 813         unsigned long   i_state_flags;          /* Dynamic state flags */
 814 #endif
 815         unsigned long   i_flags;
 816
 817 #ifdef CONFIG_EXT4_FS_XATTR
 818         /*
 819          * Extended attributes can be read independently of the main file
 820          * data. Taking i_mutex even when reading would cause contention
 821          * between readers of EAs and writers of regular file data, so
 822          * instead we synchronize on xattr_sem when reading or changing
 823          * EAs.
 824          */
 825         struct rw_semaphore xattr_sem;
 826 #endif
 827
 828         struct list_head i_orphan;      /* unlinked but open inodes */
 829
 830         /*
 831          * i_disksize keeps track of what the inode size is ON DISK, not
 832          * in memory.  During truncate, i_size is set to the new size by
 833          * the VFS prior to calling ext4_truncate(), but the filesystem won't
 834          * set i_disksize to 0 until the truncate is actually under way.
 835          *
 836          * The intent is that i_disksize always represents the blocks which
 837          * are used by this file.  This allows recovery to restart truncate
 838          * on orphans if we crash during truncate.  We actually write i_disksize
 839          * into the on-disk inode when writing inodes out, instead of i_size.
 840          *
 841          * The only time when i_disksize and i_size may be different is when
 842          * a truncate is in progress.  The only things which change i_disksize
 843          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
 844          */
 845         loff_t  i_disksize;
 846
 847         /*
 848          * i_data_sem is for serialising ext4_truncate() against
 849          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
 850          * data tree are chopped off during truncate. We can't do that in
 851          * ext4 because whenever we perform intermediate commits during
 852          * truncate, the inode and all the metadata blocks *must* be in a
 853          * consistent state which allows truncation of the orphans to restart
 854          * during recovery.  Hence we must fix the get_block-vs-truncate race
 855          * by other means, so we have i_data_sem.
 856          */
 857         struct rw_semaphore i_data_sem;
 858         struct inode vfs_inode;
 859         struct jbd2_inode *jinode;
 860
 861         struct ext4_ext_cache i_cached_extent;
 862         /*
 863          * File creation time. Its function is same as that of
 864          * struct timespec i_{a,c,m}time in the generic inode.
 865          */
 866         struct timespec i_crtime;
 867
 868         /* mballoc */
 869         struct list_head i_prealloc_list;
 870         spinlock_t i_prealloc_lock;
 871
 872         /* ialloc */
 873         ext4_group_t    i_last_alloc_group;
 874
 875         /* allocation reservation info for delalloc */
 876         /* In case of bigalloc, these refer to clusters rather than blocks */
 877         unsigned int i_reserved_data_blocks;
 878         unsigned int i_reserved_meta_blocks;
 879         unsigned int i_allocated_meta_blocks;
 880         ext4_lblk_t i_da_metadata_calc_last_lblock;
 881         int i_da_metadata_calc_len;
 882
 883         /* on-disk additional length */
 884         __u16 i_extra_isize;
 885
 886 #ifdef CONFIG_QUOTA
 887         /* quota space reservation, managed internally by quota code */
 888         qsize_t i_reserved_quota;
 889 #endif
 890
 891         /* completed IOs that might need unwritten extents handling */
 892         struct list_head i_completed_io_list;
 893         spinlock_t i_completed_io_lock;
 894         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
 895         /* current io_end structure for async DIO write*/
 896         ext4_io_end_t *cur_aio_dio;
 897         atomic_t i_aiodio_unwritten; /* Nr. of inflight conversions pending */
 898
 899         spinlock_t i_block_reservation_lock;
 900
 901         /*
 902          * Transactions that contain inode's metadata needed to complete
 903          * fsync and fdatasync, respectively.
 904          */
 905         tid_t i_sync_tid;
 906         tid_t i_datasync_tid;
 907 };
 908
 909 /*
 910  * File system states
 911  */
 912 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
 913 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
 914 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
 915
 916 /*
 917  * Misc. filesystem flags
 918  */
 919 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
 920 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
 921 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
 922
 923 /*
 924  * Mount flags
 925  */
 926 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
 927 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
 928 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
 929 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
 930 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
 931 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
 932 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
 933 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
 934 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
 935 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
 936 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
 937 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
 938 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
 939 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
 940 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
 941 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
 942 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
 943 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
 944 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
 945 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
 946 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
 947 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
 948 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
 949 #define EXT4_MOUNT_I_VERSION            0x2000000 /* i_version support */
 950 #define EXT4_MOUNT_MBLK_IO_SUBMIT       0x4000000 /* multi-block io submits */
 951 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
 952 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
 953 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
 954 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
 955 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
 956
 957 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
 958                                                       specified delalloc */
 959
 960 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
 961                                                 ~EXT4_MOUNT_##opt
 962 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
 963                                                 EXT4_MOUNT_##opt
 964 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
 965                                          EXT4_MOUNT_##opt)
 966
 967 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
 968                                                 ~EXT4_MOUNT2_##opt
 969 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
 970                                                 EXT4_MOUNT2_##opt
 971 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
 972                                          EXT4_MOUNT2_##opt)
 973
 974 #define ext4_set_bit                    __test_and_set_bit_le
 975 #define ext4_set_bit_atomic             ext2_set_bit_atomic
 976 #define ext4_clear_bit                  __test_and_clear_bit_le
 977 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
 978 #define ext4_test_bit                   test_bit_le
 979 #define ext4_find_first_zero_bit        find_first_zero_bit_le
 980 #define ext4_find_next_zero_bit         find_next_zero_bit_le
 981 #define ext4_find_next_bit              find_next_bit_le
 982
 983 extern void ext4_set_bits(void *bm, int cur, int len);
 984
 985 /*
 986  * Maximal mount counts between two filesystem checks
 987  */
 988 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
 989 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
 990
 991 /*
 992  * Behaviour when detecting errors
 993  */
 994 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
 995 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
 996 #define EXT4_ERRORS_PANIC               3       /* Panic */
 997 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
 998
 999 /*
1000  * Structure of the super block
1001  */
1002 struct ext4_super_block {
1003 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1004         __le32  s_blocks_count_lo;      /* Blocks count */
1005         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1006         __le32  s_free_blocks_count_lo; /* Free blocks count */
1007 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1008         __le32  s_first_data_block;     /* First Data Block */
1009         __le32  s_log_block_size;       /* Block size */
1010         __le32  s_log_cluster_size;     /* Allocation cluster size */
1011 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1012         __le32  s_clusters_per_group;   /* # Clusters per group */
1013         __le32  s_inodes_per_group;     /* # Inodes per group */
1014         __le32  s_mtime;                /* Mount time */
1015 /*30*/  __le32  s_wtime;                /* Write time */
1016         __le16  s_mnt_count;            /* Mount count */
1017         __le16  s_max_mnt_count;        /* Maximal mount count */
1018         __le16  s_magic;                /* Magic signature */
1019         __le16  s_state;                /* File system state */
1020         __le16  s_errors;               /* Behaviour when detecting errors */
1021         __le16  s_minor_rev_level;      /* minor revision level */
1022 /*40*/  __le32  s_lastcheck;            /* time of last check */
1023         __le32  s_checkinterval;        /* max. time between checks */
1024         __le32  s_creator_os;           /* OS */
1025         __le32  s_rev_level;            /* Revision level */
1026 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1027         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1028         /*
1029          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1030          *
1031          * Note: the difference between the compatible feature set and
1032          * the incompatible feature set is that if there is a bit set
1033          * in the incompatible feature set that the kernel doesn't
1034          * know about, it should refuse to mount the filesystem.
1035          *
1036          * e2fsck's requirements are more strict; if it doesn't know
1037          * about a feature in either the compatible or incompatible
1038          * feature set, it must abort and not try to meddle with
1039          * things it doesn't understand...
1040          */
1041         __le32  s_first_ino;            /* First non-reserved inode */
1042         __le16  s_inode_size;           /* size of inode structure */
1043         __le16  s_block_group_nr;       /* block group # of this superblock */
1044         __le32  s_feature_compat;       /* compatible feature set */
1045 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1046         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1047 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1048 /*78*/  char    s_volume_name[16];      /* volume name */
1049 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1050 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1051         /*
1052          * Performance hints.  Directory preallocation should only
1053          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1054          */
1055         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1056         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1057         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1058         /*
1059          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1060          */
1061 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1062 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1063         __le32  s_journal_dev;          /* device number of journal file */
1064         __le32  s_last_orphan;          /* start of list of inodes to delete */
1065         __le32  s_hash_seed[4];         /* HTREE hash seed */
1066         __u8    s_def_hash_version;     /* Default hash version to use */
1067         __u8    s_jnl_backup_type;
1068         __le16  s_desc_size;            /* size of group descriptor */
1069 /*100*/ __le32  s_default_mount_opts;
1070         __le32  s_first_meta_bg;        /* First metablock block group */
1071         __le32  s_mkfs_time;            /* When the filesystem was created */
1072         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1073         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1074 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1075         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1076         __le32  s_free_blocks_count_hi; /* Free blocks count */
1077         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1078         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1079         __le32  s_flags;                /* Miscellaneous flags */
1080         __le16  s_raid_stride;          /* RAID stride */
1081         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1082         __le64  s_mmp_block;            /* Block for multi-mount protection */
1083         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1084         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1085         __u8    s_reserved_char_pad;
1086         __le16  s_reserved_pad;
1087         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1088         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1089         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1090         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1091                                               snapshot's future use */
1092         __le32  s_snapshot_list;        /* inode number of the head of the
1093                                            on-disk snapshot list */
1094 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1095         __le32  s_error_count;          /* number of fs errors */
1096         __le32  s_first_error_time;     /* first time an error happened */
1097         __le32  s_first_error_ino;      /* inode involved in first error */
1098         __le64  s_first_error_block;    /* block involved of first error */
1099         __u8    s_first_error_func[32]; /* function where the error happened */
1100         __le32  s_first_error_line;     /* line number where error happened */
1101         __le32  s_last_error_time;      /* most recent time of an error */
1102         __le32  s_last_error_ino;       /* inode involved in last error */
1103         __le32  s_last_error_line;      /* line number where error happened */
1104         __le64  s_last_error_block;     /* block involved of last error */
1105         __u8    s_last_error_func[32];  /* function where the error happened */
1106 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1107         __u8    s_mount_opts[64];
1108         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1109         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1110         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1111         __le32  s_reserved[109];        /* Padding to the end of the block */
1112 };
1113
1114 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1115
1116 #ifdef __KERNEL__
1117
1118 /*
1119  * run-time mount flags
1120  */
1121 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
1122 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
1123
1124 /*
1125  * fourth extended-fs super-block data in memory
1126  */
1127 struct ext4_sb_info {
1128         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1129         unsigned long s_inodes_per_block;/* Number of inodes per block */
1130         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1131         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1132         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1133         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1134         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1135         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1136         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1137         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1138         unsigned long s_overhead;  /* # of fs overhead clusters */
1139         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1140         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1141         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1142         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1143         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1144         struct buffer_head **s_group_desc;
1145         unsigned int s_mount_opt;
1146         unsigned int s_mount_opt2;
1147         unsigned int s_mount_flags;
1148         ext4_fsblk_t s_sb_block;
1149         uid_t s_resuid;
1150         gid_t s_resgid;
1151         unsigned short s_mount_state;
1152         unsigned short s_pad;
1153         int s_addr_per_block_bits;
1154         int s_desc_per_block_bits;
1155         int s_inode_size;
1156         int s_first_ino;
1157         unsigned int s_inode_readahead_blks;
1158         unsigned int s_inode_goal;
1159         spinlock_t s_next_gen_lock;
1160         u32 s_next_generation;
1161         u32 s_hash_seed[4];
1162         int s_def_hash_version;
1163         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1164         struct percpu_counter s_freeclusters_counter;
1165         struct percpu_counter s_freeinodes_counter;
1166         struct percpu_counter s_dirs_counter;
1167         struct percpu_counter s_dirtyclusters_counter;
1168         struct blockgroup_lock *s_blockgroup_lock;
1169         struct proc_dir_entry *s_proc;
1170         struct kobject s_kobj;
1171         struct completion s_kobj_unregister;
1172
1173         /* Journaling */
1174         struct journal_s *s_journal;
1175         struct list_head s_orphan;
1176         struct mutex s_orphan_lock;
1177         unsigned long s_resize_flags;           /* Flags indicating if there
1178                                                    is a resizer */
1179         unsigned long s_commit_interval;
1180         u32 s_max_batch_time;
1181         u32 s_min_batch_time;
1182         struct block_device *journal_bdev;
1183 #ifdef CONFIG_QUOTA
1184         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
1185         int s_jquota_fmt;                       /* Format of quota to use */
1186 #endif
1187         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1188         struct rb_root system_blks;
1189
1190 #ifdef EXTENTS_STATS
1191         /* ext4 extents stats */
1192         unsigned long s_ext_min;
1193         unsigned long s_ext_max;
1194         unsigned long s_depth_max;
1195         spinlock_t s_ext_stats_lock;
1196         unsigned long s_ext_blocks;
1197         unsigned long s_ext_extents;
1198 #endif
1199
1200         /* for buddy allocator */
1201         struct ext4_group_info ***s_group_info;
1202         struct inode *s_buddy_cache;
1203         spinlock_t s_md_lock;
1204         unsigned short *s_mb_offsets;
1205         unsigned int *s_mb_maxs;
1206
1207         /* tunables */
1208         unsigned long s_stripe;
1209         unsigned int s_mb_stream_request;
1210         unsigned int s_mb_max_to_scan;
1211         unsigned int s_mb_min_to_scan;
1212         unsigned int s_mb_stats;
1213         unsigned int s_mb_order2_reqs;
1214         unsigned int s_mb_group_prealloc;
1215         unsigned int s_max_writeback_mb_bump;
1216         /* where last allocation was done - for stream allocation */
1217         unsigned long s_mb_last_group;
1218         unsigned long s_mb_last_start;
1219
1220         /* stats for buddy allocator */
1221         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1222         atomic_t s_bal_success; /* we found long enough chunks */
1223         atomic_t s_bal_allocated;       /* in blocks */
1224         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1225         atomic_t s_bal_goals;   /* goal hits */
1226         atomic_t s_bal_breaks;  /* too long searches */
1227         atomic_t s_bal_2orders; /* 2^order hits */
1228         spinlock_t s_bal_lock;
1229         unsigned long s_mb_buddies_generated;
1230         unsigned long long s_mb_generation_time;
1231         atomic_t s_mb_lost_chunks;
1232         atomic_t s_mb_preallocated;
1233         atomic_t s_mb_discarded;
1234         atomic_t s_lock_busy;
1235
1236         /* locality groups */
1237         struct ext4_locality_group __percpu *s_locality_groups;
1238
1239         /* for write statistics */
1240         unsigned long s_sectors_written_start;
1241         u64 s_kbytes_written;
1242
1243         unsigned int s_log_groups_per_flex;
1244         struct flex_groups *s_flex_groups;
1245
1246         /* workqueue for dio unwritten */
1247         struct workqueue_struct *dio_unwritten_wq;
1248
1249         /* timer for periodic error stats printing */
1250         struct timer_list s_err_report;
1251
1252         /* Lazy inode table initialization info */
1253         struct ext4_li_request *s_li_request;
1254         /* Wait multiplier for lazy initialization thread */
1255         unsigned int s_li_wait_mult;
1256
1257         /* Kernel thread for multiple mount protection */
1258         struct task_struct *s_mmp_tsk;
1259
1260         /* record the last minlen when FITRIM is called. */
1261         atomic_t s_last_trim_minblks;
1262 };
1263
1264 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1265 {
1266         return sb->s_fs_info;
1267 }
1268 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1269 {
1270         return container_of(inode, struct ext4_inode_info, vfs_inode);
1271 }
1272
1273 static inline struct timespec ext4_current_time(struct inode *inode)
1274 {
1275         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1276                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1277 }
1278
1279 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1280 {
1281         return ino == EXT4_ROOT_INO ||
1282                 ino == EXT4_JOURNAL_INO ||
1283                 ino == EXT4_RESIZE_INO ||
1284                 (ino >= EXT4_FIRST_INO(sb) &&
1285                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1286 }
1287
1288 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
1289                                               struct ext4_io_end *io_end)
1290 {
1291         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
1292                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
1293                 atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten);
1294         }
1295 }
1296
1297 /*
1298  * Inode dynamic state flags
1299  */
1300 enum {
1301         EXT4_STATE_JDATA,               /* journaled data exists */
1302         EXT4_STATE_NEW,                 /* inode is newly created */
1303         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1304         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1305         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1306         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1307         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1308         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1309         EXT4_STATE_DELALLOC_RESERVED,   /* blks already reserved for delalloc */
1310 };
1311
1312 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1313 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1314 {                                                                       \
1315         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1316 }                                                                       \
1317 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1318 {                                                                       \
1319         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1320 }                                                                       \
1321 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1322 {                                                                       \
1323         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1324 }
1325
1326 EXT4_INODE_BIT_FNS(flag, flags, 0)
1327 #if (BITS_PER_LONG < 64)
1328 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1329
1330 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1331 {
1332         (ei)->i_state_flags = 0;
1333 }
1334 #else
1335 EXT4_INODE_BIT_FNS(state, flags, 32)
1336
1337 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1338 {
1339         /* We depend on the fact that callers will set i_flags */
1340 }
1341 #endif
1342 #else
1343 /* Assume that user mode programs are passing in an ext4fs superblock, not
1344  * a kernel struct super_block.  This will allow us to call the feature-test
1345  * macros from user land. */
1346 #define EXT4_SB(sb)     (sb)
1347 #endif
1348
1349 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1350
1351 /*
1352  * Codes for operating systems
1353  */
1354 #define EXT4_OS_LINUX           0
1355 #define EXT4_OS_HURD            1
1356 #define EXT4_OS_MASIX           2
1357 #define EXT4_OS_FREEBSD         3
1358 #define EXT4_OS_LITES           4
1359
1360 /*
1361  * Revision levels
1362  */
1363 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1364 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1365
1366 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1367 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1368
1369 #define EXT4_GOOD_OLD_INODE_SIZE 128
1370
1371 /*
1372  * Feature set definitions
1373  */
1374
1375 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1376         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1377 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1378         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1379 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1380         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1381 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1382         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1383 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1384         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1385 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1386         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1387 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1388         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1389 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1390         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1391 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1392         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1393
1394 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1395 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1396 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1397 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1398 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1399 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1400
1401 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1402 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1403 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1404 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1405 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1406 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1407 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1408 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1409 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1410
1411 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1412 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1413 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1414 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1415 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1416 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1417 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1418 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1419 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1420 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1421 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1422
1423 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1424 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1425                                          EXT4_FEATURE_INCOMPAT_META_BG)
1426 #define EXT2_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 EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1431 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1432                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1433                                          EXT4_FEATURE_INCOMPAT_META_BG)
1434 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1435                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1436                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1437
1438 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1439 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1440                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1441                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1442                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1443                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1444                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1445                                          EXT4_FEATURE_INCOMPAT_MMP)
1446 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1447                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1448                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1449                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1450                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1451                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1452                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1453                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC)
1454
1455 /*
1456  * Default values for user and/or group using reserved blocks
1457  */
1458 #define EXT4_DEF_RESUID         0
1459 #define EXT4_DEF_RESGID         0
1460
1461 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1462
1463 /*
1464  * Default mount options
1465  */
1466 #define EXT4_DEFM_DEBUG         0x0001
1467 #define EXT4_DEFM_BSDGROUPS     0x0002
1468 #define EXT4_DEFM_XATTR_USER    0x0004
1469 #define EXT4_DEFM_ACL           0x0008
1470 #define EXT4_DEFM_UID16         0x0010
1471 #define EXT4_DEFM_JMODE         0x0060
1472 #define EXT4_DEFM_JMODE_DATA    0x0020
1473 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1474 #define EXT4_DEFM_JMODE_WBACK   0x0060
1475 #define EXT4_DEFM_NOBARRIER     0x0100
1476 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1477 #define EXT4_DEFM_DISCARD       0x0400
1478 #define EXT4_DEFM_NODELALLOC    0x0800
1479
1480 /*
1481  * Default journal batch times
1482  */
1483 #define EXT4_DEF_MIN_BATCH_TIME 0
1484 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1485
1486 /*
1487  * Minimum number of groups in a flexgroup before we separate out
1488  * directories into the first block group of a flexgroup
1489  */
1490 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1491
1492 /*
1493  * Structure of a directory entry
1494  */
1495 #define EXT4_NAME_LEN 255
1496
1497 struct ext4_dir_entry {
1498         __le32  inode;                  /* Inode number */
1499         __le16  rec_len;                /* Directory entry length */
1500         __le16  name_len;               /* Name length */
1501         char    name[EXT4_NAME_LEN];    /* File name */
1502 };
1503
1504 /*
1505  * The new version of the directory entry.  Since EXT4 structures are
1506  * stored in intel byte order, and the name_len field could never be
1507  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1508  * file_type field.
1509  */
1510 struct ext4_dir_entry_2 {
1511         __le32  inode;                  /* Inode number */
1512         __le16  rec_len;                /* Directory entry length */
1513         __u8    name_len;               /* Name length */
1514         __u8    file_type;
1515         char    name[EXT4_NAME_LEN];    /* File name */
1516 };
1517
1518 /*
1519  * Ext4 directory file types.  Only the low 3 bits are used.  The
1520  * other bits are reserved for now.
1521  */
1522 #define EXT4_FT_UNKNOWN         0
1523 #define EXT4_FT_REG_FILE        1
1524 #define EXT4_FT_DIR             2
1525 #define EXT4_FT_CHRDEV          3
1526 #define EXT4_FT_BLKDEV          4
1527 #define EXT4_FT_FIFO            5
1528 #define EXT4_FT_SOCK            6
1529 #define EXT4_FT_SYMLINK         7
1530
1531 #define EXT4_FT_MAX             8
1532
1533 /*
1534  * EXT4_DIR_PAD defines the directory entries boundaries
1535  *
1536  * NOTE: It must be a multiple of 4
1537  */
1538 #define EXT4_DIR_PAD                    4
1539 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1540 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1541                                          ~EXT4_DIR_ROUND)
1542 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1543
1544 /*
1545  * If we ever get support for fs block sizes > page_size, we'll need
1546  * to remove the #if statements in the next two functions...
1547  */
1548 static inline unsigned int
1549 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1550 {
1551         unsigned len = le16_to_cpu(dlen);
1552
1553 #if (PAGE_CACHE_SIZE >= 65536)
1554         if (len == EXT4_MAX_REC_LEN || len == 0)
1555                 return blocksize;
1556         return (len & 65532) | ((len & 3) << 16);
1557 #else
1558         return len;
1559 #endif
1560 }
1561
1562 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1563 {
1564         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1565                 BUG();
1566 #if (PAGE_CACHE_SIZE >= 65536)
1567         if (len < 65536)
1568                 return cpu_to_le16(len);
1569         if (len == blocksize) {
1570                 if (blocksize == 65536)
1571                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1572                 else
1573                         return cpu_to_le16(0);
1574         }
1575         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1576 #else
1577         return cpu_to_le16(len);
1578 #endif
1579 }
1580
1581 /*
1582  * Hash Tree Directory indexing
1583  * (c) Daniel Phillips, 2001
1584  */
1585
1586 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1587                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1588                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1589 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1590 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1591
1592 /* Legal values for the dx_root hash_version field: */
1593
1594 #define DX_HASH_LEGACY          0
1595 #define DX_HASH_HALF_MD4        1
1596 #define DX_HASH_TEA             2
1597 #define DX_HASH_LEGACY_UNSIGNED 3
1598 #define DX_HASH_HALF_MD4_UNSIGNED       4
1599 #define DX_HASH_TEA_UNSIGNED            5
1600
1601 #ifdef __KERNEL__
1602
1603 /* hash info structure used by the directory hash */
1604 struct dx_hash_info
1605 {
1606         u32             hash;
1607         u32             minor_hash;
1608         int             hash_version;
1609         u32             *seed;
1610 };
1611
1612
1613 /* 32 and 64 bit signed EOF for dx directories */
1614 #define EXT4_HTREE_EOF_32BIT   ((1UL  << (32 - 1)) - 1)
1615 #define EXT4_HTREE_EOF_64BIT   ((1ULL << (64 - 1)) - 1)
1616
1617
1618 /*
1619  * Control parameters used by ext4_htree_next_block
1620  */
1621 #define HASH_NB_ALWAYS          1
1622
1623
1624 /*
1625  * Describe an inode's exact location on disk and in memory
1626  */
1627 struct ext4_iloc
1628 {
1629         struct buffer_head *bh;
1630         unsigned long offset;
1631         ext4_group_t block_group;
1632 };
1633
1634 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1635 {
1636         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1637 }
1638
1639 /*
1640  * This structure is stuffed into the struct file's private_data field
1641  * for directories.  It is where we put information so that we can do
1642  * readdir operations in hash tree order.
1643  */
1644 struct dir_private_info {
1645         struct rb_root  root;
1646         struct rb_node  *curr_node;
1647         struct fname    *extra_fname;
1648         loff_t          last_pos;
1649         __u32           curr_hash;
1650         __u32           curr_minor_hash;
1651         __u32           next_hash;
1652 };
1653
1654 /* calculate the first block number of the group */
1655 static inline ext4_fsblk_t
1656 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1657 {
1658         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1659                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1660 }
1661
1662 /*
1663  * Special error return code only used by dx_probe() and its callers.
1664  */
1665 #define ERR_BAD_DX_DIR  -75000
1666
1667 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1668                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1669
1670 /*
1671  * Timeout and state flag for lazy initialization inode thread.
1672  */
1673 #define EXT4_DEF_LI_WAIT_MULT                   10
1674 #define EXT4_DEF_LI_MAX_START_DELAY             5
1675 #define EXT4_LAZYINIT_QUIT                      0x0001
1676 #define EXT4_LAZYINIT_RUNNING                   0x0002
1677
1678 /*
1679  * Lazy inode table initialization info
1680  */
1681 struct ext4_lazy_init {
1682         unsigned long           li_state;
1683         struct list_head        li_request_list;
1684         struct mutex            li_list_mtx;
1685 };
1686
1687 struct ext4_li_request {
1688         struct super_block      *lr_super;
1689         struct ext4_sb_info     *lr_sbi;
1690         ext4_group_t            lr_next_group;
1691         struct list_head        lr_request;
1692         unsigned long           lr_next_sched;
1693         unsigned long           lr_timeout;
1694 };
1695
1696 struct ext4_features {
1697         struct kobject f_kobj;
1698         struct completion f_kobj_unregister;
1699 };
1700
1701 /*
1702  * This structure will be used for multiple mount protection. It will be
1703  * written into the block number saved in the s_mmp_block field in the
1704  * superblock. Programs that check MMP should assume that if
1705  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1706  * to use the filesystem, regardless of how old the timestamp is.
1707  */
1708 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
1709 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1710 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
1711 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
1712
1713 struct mmp_struct {
1714         __le32  mmp_magic;              /* Magic number for MMP */
1715         __le32  mmp_seq;                /* Sequence no. updated periodically */
1716
1717         /*
1718          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1719          * purposes and do not affect the correctness of the algorithm
1720          */
1721         __le64  mmp_time;               /* Time last updated */
1722         char    mmp_nodename[64];       /* Node which last updated MMP block */
1723         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
1724
1725         /*
1726          * mmp_check_interval is used to verify if the MMP block has been
1727          * updated on the block device. The value is updated based on the
1728          * maximum time to write the MMP block during an update cycle.
1729          */
1730         __le16  mmp_check_interval;
1731
1732         __le16  mmp_pad1;
1733         __le32  mmp_pad2[227];
1734 };
1735
1736 /* arguments passed to the mmp thread */
1737 struct mmpd_data {
1738         struct buffer_head *bh; /* bh from initial read_mmp_block() */
1739         struct super_block *sb;  /* super block of the fs */
1740 };
1741
1742 /*
1743  * Check interval multiplier
1744  * The MMP block is written every update interval and initially checked every
1745  * update interval x the multiplier (the value is then adapted based on the
1746  * write latency). The reason is that writes can be delayed under load and we
1747  * don't want readers to incorrectly assume that the filesystem is no longer
1748  * in use.
1749  */
1750 #define EXT4_MMP_CHECK_MULT             2UL
1751
1752 /*
1753  * Minimum interval for MMP checking in seconds.
1754  */
1755 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
1756
1757 /*
1758  * Maximum interval for MMP checking in seconds.
1759  */
1760 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
1761
1762 /*
1763  * Function prototypes
1764  */
1765
1766 /*
1767  * Ok, these declarations are also in <linux/kernel.h> but none of the
1768  * ext4 source programs needs to include it so they are duplicated here.
1769  */
1770 # define NORET_TYPE     /**/
1771 # define ATTRIB_NORET   __attribute__((noreturn))
1772 # define NORET_AND      noreturn,
1773
1774 /* bitmap.c */
1775 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
1776
1777 /* balloc.c */
1778 extern unsigned int ext4_block_group(struct super_block *sb,
1779                         ext4_fsblk_t blocknr);
1780 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1781                         ext4_fsblk_t blocknr);
1782 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1783 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1784                         ext4_group_t group);
1785 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1786                                          ext4_fsblk_t goal,
1787                                          unsigned int flags,
1788                                          unsigned long *count,
1789                                          int *errp);
1790 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
1791                                     s64 nclusters, unsigned int flags);
1792 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
1793 extern void ext4_check_blocks_bitmap(struct super_block *);
1794 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1795                                                     ext4_group_t block_group,
1796                                                     struct buffer_head ** bh);
1797 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1798 struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1799                                       ext4_group_t block_group);
1800 extern void ext4_init_block_bitmap(struct super_block *sb,
1801                                    struct buffer_head *bh,
1802                                    ext4_group_t group,
1803                                    struct ext4_group_desc *desc);
1804 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
1805                                               ext4_group_t block_group,
1806                                               struct ext4_group_desc *gdp);
1807 extern unsigned ext4_num_base_meta_clusters(struct super_block *sb,
1808                                             ext4_group_t block_group);
1809 extern unsigned ext4_num_overhead_clusters(struct super_block *sb,
1810                                            ext4_group_t block_group,
1811                                            struct ext4_group_desc *gdp);
1812 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
1813
1814 /* dir.c */
1815 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1816                                   struct file *,
1817                                   struct ext4_dir_entry_2 *,
1818                                   struct buffer_head *, unsigned int);
1819 #define ext4_check_dir_entry(dir, filp, de, bh, offset)                 \
1820         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
1821                                         (de), (bh), (offset)))
1822 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1823                                     __u32 minor_hash,
1824                                     struct ext4_dir_entry_2 *dirent);
1825 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1826
1827 /* fsync.c */
1828 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
1829 extern int ext4_flush_completed_IO(struct inode *);
1830
1831 /* hash.c */
1832 extern int ext4fs_dirhash(const char *name, int len, struct
1833                           dx_hash_info *hinfo);
1834
1835 /* ialloc.c */
1836 extern struct inode *ext4_new_inode(handle_t *, struct inode *, int,
1837                                     const struct qstr *qstr, __u32 goal,
1838                                     uid_t *owner);
1839 extern void ext4_free_inode(handle_t *, struct inode *);
1840 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1841 extern unsigned long ext4_count_free_inodes(struct super_block *);
1842 extern unsigned long ext4_count_dirs(struct super_block *);
1843 extern void ext4_check_inodes_bitmap(struct super_block *);
1844 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1845 extern int ext4_init_inode_table(struct super_block *sb,
1846                                  ext4_group_t group, int barrier);
1847
1848 /* mballoc.c */
1849 extern long ext4_mb_stats;
1850 extern long ext4_mb_max_to_scan;
1851 extern int ext4_mb_init(struct super_block *, int);
1852 extern int ext4_mb_release(struct super_block *);
1853 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1854                                 struct ext4_allocation_request *, int *);
1855 extern int ext4_mb_reserve_blocks(struct super_block *, int);
1856 extern void ext4_discard_preallocations(struct inode *);
1857 extern int __init ext4_init_mballoc(void);
1858 extern void ext4_exit_mballoc(void);
1859 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1860                              struct buffer_head *bh, ext4_fsblk_t block,
1861                              unsigned long count, int flags);
1862 extern int ext4_mb_add_groupinfo(struct super_block *sb,
1863                 ext4_group_t i, struct ext4_group_desc *desc);
1864 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
1865                                 ext4_fsblk_t block, unsigned long count);
1866 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
1867
1868 /* inode.c */
1869 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1870                                                 ext4_lblk_t, int, int *);
1871 struct buffer_head *ext4_bread(handle_t *, struct inode *,
1872                                                 ext4_lblk_t, int, int *);
1873 int ext4_get_block(struct inode *inode, sector_t iblock,
1874                                 struct buffer_head *bh_result, int create);
1875
1876 extern struct inode *ext4_iget(struct super_block *, unsigned long);
1877 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
1878 extern int  ext4_setattr(struct dentry *, struct iattr *);
1879 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1880                                 struct kstat *stat);
1881 extern void ext4_evict_inode(struct inode *);
1882 extern void ext4_clear_inode(struct inode *);
1883 extern int  ext4_sync_inode(handle_t *, struct inode *);
1884 extern void ext4_dirty_inode(struct inode *, int);
1885 extern int ext4_change_inode_journal_flag(struct inode *, int);
1886 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1887 extern int ext4_can_truncate(struct inode *inode);
1888 extern void ext4_truncate(struct inode *);
1889 extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length);
1890 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
1891 extern void ext4_set_inode_flags(struct inode *);
1892 extern void ext4_get_inode_flags(struct ext4_inode_info *);
1893 extern int ext4_alloc_da_blocks(struct inode *inode);
1894 extern void ext4_set_aops(struct inode *inode);
1895 extern int ext4_writepage_trans_blocks(struct inode *);
1896 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1897 extern int ext4_block_truncate_page(handle_t *handle,
1898                 struct address_space *mapping, loff_t from);
1899 extern int ext4_block_zero_page_range(handle_t *handle,
1900                 struct address_space *mapping, loff_t from, loff_t length);
1901 extern int ext4_discard_partial_page_buffers(handle_t *handle,
1902                 struct address_space *mapping, loff_t from,
1903                 loff_t length, int flags);
1904 extern int ext4_discard_partial_page_buffers_no_lock(handle_t *handle,
1905                 struct inode *inode, struct page *page, loff_t from,
1906                 loff_t length, int flags);
1907 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1908 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
1909 extern void ext4_da_update_reserve_space(struct inode *inode,
1910                                         int used, int quota_claim);
1911
1912 /* indirect.c */
1913 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
1914                                 struct ext4_map_blocks *map, int flags);
1915 extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
1916                                 const struct iovec *iov, loff_t offset,
1917                                 unsigned long nr_segs);
1918 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
1919 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks, int chunk);
1920 extern void ext4_ind_truncate(struct inode *inode);
1921
1922 /* ioctl.c */
1923 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1924 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1925
1926 /* migrate.c */
1927 extern int ext4_ext_migrate(struct inode *);
1928
1929 /* namei.c */
1930 extern int ext4_orphan_add(handle_t *, struct inode *);
1931 extern int ext4_orphan_del(handle_t *, struct inode *);
1932 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1933                                 __u32 start_minor_hash, __u32 *next_hash);
1934
1935 /* resize.c */
1936 extern int ext4_group_add(struct super_block *sb,
1937                                 struct ext4_new_group_data *input);
1938 extern int ext4_group_extend(struct super_block *sb,
1939                                 struct ext4_super_block *es,
1940                                 ext4_fsblk_t n_blocks_count);
1941
1942 /* super.c */
1943 extern int ext4_calculate_overhead(struct super_block *sb);
1944 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
1945 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
1946 extern void ext4_kvfree(void *ptr);
1947 extern __printf(4, 5)
1948 void __ext4_error(struct super_block *, const char *, unsigned int,
1949                   const char *, ...);
1950 #define ext4_error(sb, message...)      __ext4_error(sb, __func__,      \
1951                                                      __LINE__, ## message)
1952 extern __printf(5, 6)
1953 void ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
1954                       const char *, ...);
1955 extern __printf(5, 6)
1956 void ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
1957                      const char *, ...);
1958 extern void __ext4_std_error(struct super_block *, const char *,
1959                              unsigned int, int);
1960 extern __printf(4, 5)
1961 void __ext4_abort(struct super_block *, const char *, unsigned int,
1962                   const char *, ...);
1963 #define ext4_abort(sb, message...)      __ext4_abort(sb, __func__, \
1964                                                        __LINE__, ## message)
1965 extern __printf(4, 5)
1966 void __ext4_warning(struct super_block *, const char *, unsigned int,
1967                     const char *, ...);
1968 #define ext4_warning(sb, message...)    __ext4_warning(sb, __func__, \
1969                                                        __LINE__, ## message)
1970 extern __printf(3, 4)
1971 void ext4_msg(struct super_block *, const char *, const char *, ...);
1972 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
1973                            const char *, unsigned int, const char *);
1974 #define dump_mmp_msg(sb, mmp, msg)      __dump_mmp_msg(sb, mmp, __func__, \
1975                                                        __LINE__, msg)
1976 extern __printf(7, 8)
1977 void __ext4_grp_locked_error(const char *, unsigned int,
1978                              struct super_block *, ext4_group_t,
1979                              unsigned long, ext4_fsblk_t,
1980                              const char *, ...);
1981 #define ext4_grp_locked_error(sb, grp, message...) \
1982         __ext4_grp_locked_error(__func__, __LINE__, (sb), (grp), ## message)
1983 extern void ext4_update_dynamic_rev(struct super_block *sb);
1984 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1985                                         __u32 compat);
1986 extern int ext4_update_rocompat_feature(handle_t *handle,
1987                                         struct super_block *sb, __u32 rocompat);
1988 extern int ext4_update_incompat_feature(handle_t *handle,
1989                                         struct super_block *sb, __u32 incompat);
1990 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1991                                       struct ext4_group_desc *bg);
1992 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1993                                       struct ext4_group_desc *bg);
1994 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1995                                      struct ext4_group_desc *bg);
1996 extern __u32 ext4_free_group_clusters(struct super_block *sb,
1997                                       struct ext4_group_desc *bg);
1998 extern __u32 ext4_free_inodes_count(struct super_block *sb,
1999                                  struct ext4_group_desc *bg);
2000 extern __u32 ext4_used_dirs_count(struct super_block *sb,
2001                                 struct ext4_group_desc *bg);
2002 extern __u32 ext4_itable_unused_count(struct super_block *sb,
2003                                    struct ext4_group_desc *bg);
2004 extern void ext4_block_bitmap_set(struct super_block *sb,
2005                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2006 extern void ext4_inode_bitmap_set(struct super_block *sb,
2007                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2008 extern void ext4_inode_table_set(struct super_block *sb,
2009                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
2010 extern void ext4_free_group_clusters_set(struct super_block *sb,
2011                                          struct ext4_group_desc *bg,
2012                                          __u32 count);
2013 extern void ext4_free_inodes_set(struct super_block *sb,
2014                                 struct ext4_group_desc *bg, __u32 count);
2015 extern void ext4_used_dirs_set(struct super_block *sb,
2016                                 struct ext4_group_desc *bg, __u32 count);
2017 extern void ext4_itable_unused_set(struct super_block *sb,
2018                                    struct ext4_group_desc *bg, __u32 count);
2019 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
2020                                    struct ext4_group_desc *gdp);
2021 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
2022                                        struct ext4_group_desc *gdp);
2023
2024 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2025 {
2026         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2027                 le32_to_cpu(es->s_blocks_count_lo);
2028 }
2029
2030 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2031 {
2032         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2033                 le32_to_cpu(es->s_r_blocks_count_lo);
2034 }
2035
2036 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2037 {
2038         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2039                 le32_to_cpu(es->s_free_blocks_count_lo);
2040 }
2041
2042 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2043                                          ext4_fsblk_t blk)
2044 {
2045         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2046         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2047 }
2048
2049 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2050                                               ext4_fsblk_t blk)
2051 {
2052         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2053         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2054 }
2055
2056 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2057                                            ext4_fsblk_t blk)
2058 {
2059         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2060         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2061 }
2062
2063 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2064 {
2065         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2066                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2067                         le32_to_cpu(raw_inode->i_size_lo);
2068         else
2069                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2070 }
2071
2072 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2073 {
2074         raw_inode->i_size_lo = cpu_to_le32(i_size);
2075         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2076 }
2077
2078 static inline
2079 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2080                                             ext4_group_t group)
2081 {
2082          struct ext4_group_info ***grp_info;
2083          long indexv, indexh;
2084          grp_info = EXT4_SB(sb)->s_group_info;
2085          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2086          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2087          return grp_info[indexv][indexh];
2088 }
2089
2090 /*
2091  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2092  * the locking protocol documented in the comments of ext4_group_add()
2093  * in resize.c
2094  */
2095 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2096 {
2097         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2098
2099         smp_rmb();
2100         return ngroups;
2101 }
2102
2103 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2104                                              ext4_group_t block_group)
2105 {
2106         return block_group >> sbi->s_log_groups_per_flex;
2107 }
2108
2109 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2110 {
2111         return 1 << sbi->s_log_groups_per_flex;
2112 }
2113
2114 #define ext4_std_error(sb, errno)                               \
2115 do {                                                            \
2116         if ((errno))                                            \
2117                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2118 } while (0)
2119
2120 #ifdef CONFIG_SMP
2121 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2122  * counters. So we need to make sure we have free clusters more
2123  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2124  */
2125 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2126 #else
2127 #define EXT4_FREECLUSTERS_WATERMARK 0
2128 #endif
2129
2130 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2131 {
2132         /*
2133          * XXX: replace with spinlock if seen contended -bzzz
2134          */
2135         down_write(&EXT4_I(inode)->i_data_sem);
2136         if (newsize > EXT4_I(inode)->i_disksize)
2137                 EXT4_I(inode)->i_disksize = newsize;
2138         up_write(&EXT4_I(inode)->i_data_sem);
2139         return ;
2140 }
2141
2142 struct ext4_group_info {
2143         unsigned long   bb_state;
2144         struct rb_root  bb_free_root;
2145         ext4_grpblk_t   bb_first_free;  /* first free block */
2146         ext4_grpblk_t   bb_free;        /* total free blocks */
2147         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2148         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2149         struct          list_head bb_prealloc_list;
2150 #ifdef DOUBLE_CHECK
2151         void            *bb_bitmap;
2152 #endif
2153         struct rw_semaphore alloc_sem;
2154         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2155                                          * regions, index is order.
2156                                          * bb_counters[3] = 5 means
2157                                          * 5 free 8-block regions. */
2158 };
2159
2160 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2161 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2162
2163 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2164         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2165
2166 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2167         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2168 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2169         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2170 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2171         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2172
2173 #define EXT4_MAX_CONTENTION             8
2174 #define EXT4_CONTENTION_THRESHOLD       2
2175
2176 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2177                                               ext4_group_t group)
2178 {
2179         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2180 }
2181
2182 /*
2183  * Returns true if the filesystem is busy enough that attempts to
2184  * access the block group locks has run into contention.
2185  */
2186 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2187 {
2188         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2189 }
2190
2191 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2192 {
2193         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2194         if (spin_trylock(lock))
2195                 /*
2196                  * We're able to grab the lock right away, so drop the
2197                  * lock contention counter.
2198                  */
2199                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2200         else {
2201                 /*
2202                  * The lock is busy, so bump the contention counter,
2203                  * and then wait on the spin lock.
2204                  */
2205                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2206                                   EXT4_MAX_CONTENTION);
2207                 spin_lock(lock);
2208         }
2209 }
2210
2211 static inline void ext4_unlock_group(struct super_block *sb,
2212                                         ext4_group_t group)
2213 {
2214         spin_unlock(ext4_group_lock_ptr(sb, group));
2215 }
2216
2217 static inline void ext4_mark_super_dirty(struct super_block *sb)
2218 {
2219         if (EXT4_SB(sb)->s_journal == NULL)
2220                 sb->s_dirt =1;
2221 }
2222
2223 /*
2224  * Block validity checking
2225  */
2226 #define ext4_check_indirect_blockref(inode, bh)                         \
2227         ext4_check_blockref(__func__, __LINE__, inode,                  \
2228                             (__le32 *)(bh)->b_data,                     \
2229                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2230
2231 #define ext4_ind_check_inode(inode)                                     \
2232         ext4_check_blockref(__func__, __LINE__, inode,                  \
2233                             EXT4_I(inode)->i_data,                      \
2234                             EXT4_NDIR_BLOCKS)
2235
2236 /*
2237  * Inodes and files operations
2238  */
2239
2240 /* dir.c */
2241 extern const struct file_operations ext4_dir_operations;
2242
2243 /* file.c */
2244 extern const struct inode_operations ext4_file_inode_operations;
2245 extern const struct file_operations ext4_file_operations;
2246 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2247
2248 /* namei.c */
2249 extern const struct inode_operations ext4_dir_inode_operations;
2250 extern const struct inode_operations ext4_special_inode_operations;
2251 extern struct dentry *ext4_get_parent(struct dentry *child);
2252
2253 /* symlink.c */
2254 extern const struct inode_operations ext4_symlink_inode_operations;
2255 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2256
2257 /* block_validity */
2258 extern void ext4_release_system_zone(struct super_block *sb);
2259 extern int ext4_setup_system_zone(struct super_block *sb);
2260 extern int __init ext4_init_system_zone(void);
2261 extern void ext4_exit_system_zone(void);
2262 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2263                                  ext4_fsblk_t start_blk,
2264                                  unsigned int count);
2265 extern int ext4_check_blockref(const char *, unsigned int,
2266                                struct inode *, __le32 *, unsigned int);
2267
2268 /* extents.c */
2269 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2270 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2271 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
2272                                        int chunk);
2273 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2274                                struct ext4_map_blocks *map, int flags);
2275 extern void ext4_ext_truncate(struct inode *);
2276 extern int ext4_ext_punch_hole(struct file *file, loff_t offset,
2277                                 loff_t length);
2278 extern void ext4_ext_init(struct super_block *);
2279 extern void ext4_ext_release(struct super_block *);
2280 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2281                           loff_t len);
2282 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
2283                           ssize_t len);
2284 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2285                            struct ext4_map_blocks *map, int flags);
2286 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2287                         __u64 start, __u64 len);
2288 /* move_extent.c */
2289 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2290                              __u64 start_orig, __u64 start_donor,
2291                              __u64 len, __u64 *moved_len);
2292
2293 /* page-io.c */
2294 extern int __init ext4_init_pageio(void);
2295 extern void ext4_exit_pageio(void);
2296 extern void ext4_ioend_wait(struct inode *);
2297 extern void ext4_free_io_end(ext4_io_end_t *io);
2298 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2299 extern int ext4_end_io_nolock(ext4_io_end_t *io);
2300 extern void ext4_io_submit(struct ext4_io_submit *io);
2301 extern int ext4_bio_write_page(struct ext4_io_submit *io,
2302                                struct page *page,
2303                                int len,
2304                                struct writeback_control *wbc);
2305
2306 /* mmp.c */
2307 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2308
2309 /* BH_Uninit flag: blocks are allocated but uninitialized on disk */
2310 enum ext4_state_bits {
2311         BH_Uninit       /* blocks are allocated but uninitialized on disk */
2312           = BH_JBDPrivateStart,
2313         BH_AllocFromCluster,    /* allocated blocks were part of already
2314                                  * allocated cluster. Note that this flag will
2315                                  * never, ever appear in a buffer_head's state
2316                                  * flag. See EXT4_MAP_FROM_CLUSTER to see where
2317                                  * this is used. */
2318         BH_Da_Mapped,   /* Delayed allocated block that now has a mapping. This
2319                          * flag is set when ext4_map_blocks is called on a
2320                          * delayed allocated block to get its real mapping. */
2321 };
2322
2323 BUFFER_FNS(Uninit, uninit)
2324 TAS_BUFFER_FNS(Uninit, uninit)
2325 BUFFER_FNS(Da_Mapped, da_mapped)
2326
2327 /*
2328  * Add new method to test wether block and inode bitmaps are properly
2329  * initialized. With uninit_bg reading the block from disk is not enough
2330  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2331  */
2332 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2333
2334 static inline int bitmap_uptodate(struct buffer_head *bh)
2335 {
2336         return (buffer_uptodate(bh) &&
2337                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2338 }
2339 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2340 {
2341         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2342 }
2343
2344 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2345
2346 /* For ioend & aio unwritten conversion wait queues */
2347 #define EXT4_WQ_HASH_SZ         37
2348 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
2349                                             EXT4_WQ_HASH_SZ])
2350 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
2351                                              EXT4_WQ_HASH_SZ])
2352 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2353 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2354
2355 #define EXT4_RESIZING   0
2356 extern int ext4_resize_begin(struct super_block *sb);
2357 extern void ext4_resize_end(struct super_block *sb);
2358
2359 #endif  /* __KERNEL__ */
2360
2361 #include "ext4_extents.h"
2362
2363 #endif  /* _EXT4_H */