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 #include <crypto/hash.h>
  33 #ifdef __KERNEL__
  34 #include <linux/compat.h>
  35 #endif
  36
  37 /*
  38  * The fourth extended filesystem constants/structures
  39  */
  40
  41 /*
  42  * Define EXT4FS_DEBUG to produce debug messages
  43  */
  44 #undef EXT4FS_DEBUG
  45
  46 /*
  47  * Debug code
  48  */
  49 #ifdef EXT4FS_DEBUG
  50 #define ext4_debug(f, a...)                                             \
  51         do {                                                            \
  52                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
  53                         __FILE__, __LINE__, __func__);                  \
  54                 printk(KERN_DEBUG f, ## a);                             \
  55         } while (0)
  56 #else
  57 #define ext4_debug(fmt, ...)    no_printk(fmt, ##__VA_ARGS__)
  58 #endif
  59
  60 /*
  61  * Turn on EXT_DEBUG to get lots of info about extents operations.
  62  */
  63 #define EXT_DEBUG__
  64 #ifdef EXT_DEBUG
  65 #define ext_debug(fmt, ...)     printk(fmt, ##__VA_ARGS__)
  66 #else
  67 #define ext_debug(fmt, ...)     no_printk(fmt, ##__VA_ARGS__)
  68 #endif
  69
  70 #define EXT4_ERROR_INODE(inode, fmt, a...) \
  71         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
  72
  73 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...)                 \
  74         ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
  75
  76 #define EXT4_ERROR_FILE(file, block, fmt, a...)                         \
  77         ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
  78
  79 /* data type for block offset of block group */
  80 typedef int ext4_grpblk_t;
  81
  82 /* data type for filesystem-wide blocks number */
  83 typedef unsigned long long ext4_fsblk_t;
  84
  85 /* data type for file logical block number */
  86 typedef __u32 ext4_lblk_t;
  87
  88 /* data type for block group number */
  89 typedef unsigned int ext4_group_t;
  90
  91 /*
  92  * Flags used in mballoc's allocation_context flags field.
  93  *
  94  * Also used to show what's going on for debugging purposes when the
  95  * flag field is exported via the traceport interface
  96  */
  97
  98 /* prefer goal again. length */
  99 #define EXT4_MB_HINT_MERGE              0x0001
 100 /* blocks already reserved */
 101 #define EXT4_MB_HINT_RESERVED           0x0002
 102 /* metadata is being allocated */
 103 #define EXT4_MB_HINT_METADATA           0x0004
 104 /* first blocks in the file */
 105 #define EXT4_MB_HINT_FIRST              0x0008
 106 /* search for the best chunk */
 107 #define EXT4_MB_HINT_BEST               0x0010
 108 /* data is being allocated */
 109 #define EXT4_MB_HINT_DATA               0x0020
 110 /* don't preallocate (for tails) */
 111 #define EXT4_MB_HINT_NOPREALLOC         0x0040
 112 /* allocate for locality group */
 113 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
 114 /* allocate goal blocks or none */
 115 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
 116 /* goal is meaningful */
 117 #define EXT4_MB_HINT_TRY_GOAL           0x0200
 118 /* blocks already pre-reserved by delayed allocation */
 119 #define EXT4_MB_DELALLOC_RESERVED       0x0400
 120 /* We are doing stream allocation */
 121 #define EXT4_MB_STREAM_ALLOC            0x0800
 122 /* Use reserved root blocks if needed */
 123 #define EXT4_MB_USE_ROOT_BLOCKS         0x1000
 124 /* Use blocks from reserved pool */
 125 #define EXT4_MB_USE_RESERVED            0x2000
 126
 127 struct ext4_allocation_request {
 128         /* target inode for block we're allocating */
 129         struct inode *inode;
 130         /* how many blocks we want to allocate */
 131         unsigned int len;
 132         /* logical block in target inode */
 133         ext4_lblk_t logical;
 134         /* the closest logical allocated block to the left */
 135         ext4_lblk_t lleft;
 136         /* the closest logical allocated block to the right */
 137         ext4_lblk_t lright;
 138         /* phys. target (a hint) */
 139         ext4_fsblk_t goal;
 140         /* phys. block for the closest logical allocated block to the left */
 141         ext4_fsblk_t pleft;
 142         /* phys. block for the closest logical allocated block to the right */
 143         ext4_fsblk_t pright;
 144         /* flags. see above EXT4_MB_HINT_* */
 145         unsigned int flags;
 146 };
 147
 148 /*
 149  * Logical to physical block mapping, used by ext4_map_blocks()
 150  *
 151  * This structure is used to pass requests into ext4_map_blocks() as
 152  * well as to store the information returned by ext4_map_blocks().  It
 153  * takes less room on the stack than a struct buffer_head.
 154  */
 155 #define EXT4_MAP_NEW            (1 << BH_New)
 156 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
 157 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
 158 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
 159 #define EXT4_MAP_UNINIT         (1 << BH_Uninit)
 160 /* Sometimes (in the bigalloc case, from ext4_da_get_block_prep) the caller of
 161  * ext4_map_blocks wants to know whether or not the underlying cluster has
 162  * already been accounted for. EXT4_MAP_FROM_CLUSTER conveys to the caller that
 163  * the requested mapping was from previously mapped (or delayed allocated)
 164  * cluster. We use BH_AllocFromCluster only for this flag. BH_AllocFromCluster
 165  * should never appear on buffer_head's state flags.
 166  */
 167 #define EXT4_MAP_FROM_CLUSTER   (1 << BH_AllocFromCluster)
 168 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
 169                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
 170                                  EXT4_MAP_UNINIT | EXT4_MAP_FROM_CLUSTER)
 171
 172 struct ext4_map_blocks {
 173         ext4_fsblk_t m_pblk;
 174         ext4_lblk_t m_lblk;
 175         unsigned int m_len;
 176         unsigned int m_flags;
 177 };
 178
 179 /*
 180  * For delayed allocation tracking
 181  */
 182 struct mpage_da_data {
 183         struct inode *inode;
 184         sector_t b_blocknr;             /* start block number of extent */
 185         size_t b_size;                  /* size of extent */
 186         unsigned long b_state;          /* state of the extent */
 187         unsigned long first_page, next_page;    /* extent of pages */
 188         struct writeback_control *wbc;
 189         int io_done;
 190         int pages_written;
 191         int retval;
 192 };
 193
 194 /*
 195  * Flags for ext4_io_end->flags
 196  */
 197 #define EXT4_IO_END_UNWRITTEN   0x0001
 198 #define EXT4_IO_END_ERROR       0x0002
 199 #define EXT4_IO_END_DIRECT      0x0004
 200
 201 /*
 202  * For converting uninitialized extents on a work queue.
 203  */
 204 typedef struct ext4_io_end {
 205         struct list_head        list;           /* per-file finished IO list */
 206         struct inode            *inode;         /* file being written to */
 207         unsigned int            flag;           /* unwritten or not */
 208         loff_t                  offset;         /* offset in the file */
 209         ssize_t                 size;           /* size of the extent */
 210         struct kiocb            *iocb;          /* iocb struct for AIO */
 211         int                     result;         /* error value for AIO */
 212 } ext4_io_end_t;
 213
 214 struct ext4_io_submit {
 215         int                     io_op;
 216         struct bio              *io_bio;
 217         ext4_io_end_t           *io_end;
 218         sector_t                io_next_block;
 219 };
 220
 221 /*
 222  * Special inodes numbers
 223  */
 224 #define EXT4_BAD_INO             1      /* Bad blocks inode */
 225 #define EXT4_ROOT_INO            2      /* Root inode */
 226 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
 227 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
 228 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
 229 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
 230 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
 231 #define EXT4_JOURNAL_INO         8      /* Journal inode */
 232
 233 /* First non-reserved inode for old ext4 filesystems */
 234 #define EXT4_GOOD_OLD_FIRST_INO 11
 235
 236 /*
 237  * Maximal count of links to a file
 238  */
 239 #define EXT4_LINK_MAX           65000
 240
 241 /*
 242  * Macro-instructions used to manage several block sizes
 243  */
 244 #define EXT4_MIN_BLOCK_SIZE             1024
 245 #define EXT4_MAX_BLOCK_SIZE             65536
 246 #define EXT4_MIN_BLOCK_LOG_SIZE         10
 247 #define EXT4_MAX_BLOCK_LOG_SIZE         16
 248 #ifdef __KERNEL__
 249 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
 250 #else
 251 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
 252 #endif
 253 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
 254 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
 255                                          EXT4_SB(s)->s_cluster_bits)
 256 #ifdef __KERNEL__
 257 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
 258 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
 259 #else
 260 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
 261 #endif
 262 #ifdef __KERNEL__
 263 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
 264 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
 265 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
 266 #else
 267 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 268                                  EXT4_GOOD_OLD_INODE_SIZE : \
 269                                  (s)->s_inode_size)
 270 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 271                                  EXT4_GOOD_OLD_FIRST_INO : \
 272                                  (s)->s_first_ino)
 273 #endif
 274 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
 275
 276 /* Translate a block number to a cluster number */
 277 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
 278 /* Translate a cluster number to a block number */
 279 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
 280 /* Translate # of blks to # of clusters */
 281 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
 282                                  (sbi)->s_cluster_bits)
 283 /* Mask out the low bits to get the starting block of the cluster */
 284 #define EXT4_PBLK_CMASK(s, pblk) ((pblk) &                              \
 285                                   ~((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
 286 #define EXT4_LBLK_CMASK(s, lblk) ((lblk) &                              \
 287                                   ~((ext4_lblk_t) (s)->s_cluster_ratio - 1))
 288 /* Get the cluster offset */
 289 #define EXT4_PBLK_COFF(s, pblk) ((pblk) &                               \
 290                                  ((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
 291 #define EXT4_LBLK_COFF(s, lblk) ((lblk) &                               \
 292                                  ((ext4_lblk_t) (s)->s_cluster_ratio - 1))
 293
 294 /*
 295  * Structure of a blocks group descriptor
 296  */
 297 struct ext4_group_desc
 298 {
 299         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
 300         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
 301         __le32  bg_inode_table_lo;      /* Inodes table block */
 302         __le16  bg_free_blocks_count_lo;/* Free blocks count */
 303         __le16  bg_free_inodes_count_lo;/* Free inodes count */
 304         __le16  bg_used_dirs_count_lo;  /* Directories count */
 305         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
 306         __le32  bg_exclude_bitmap_lo;   /* Exclude bitmap for snapshots */
 307         __le16  bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
 308         __le16  bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
 309         __le16  bg_itable_unused_lo;    /* Unused inodes count */
 310         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
 311         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
 312         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
 313         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
 314         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
 315         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
 316         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
 317         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
 318         __le32  bg_exclude_bitmap_hi;   /* Exclude bitmap block MSB */
 319         __le16  bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
 320         __le16  bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
 321         __u32   bg_reserved;
 322 };
 323
 324 #define EXT4_BG_INODE_BITMAP_CSUM_HI_END        \
 325         (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
 326          sizeof(__le16))
 327 #define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END        \
 328         (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
 329          sizeof(__le16))
 330
 331 /*
 332  * Structure of a flex block group info
 333  */
 334
 335 struct flex_groups {
 336         atomic64_t      free_clusters;
 337         atomic_t        free_inodes;
 338         atomic_t        used_dirs;
 339 };
 340
 341 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
 342 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
 343 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
 344
 345 /*
 346  * Macro-instructions used to manage group descriptors
 347  */
 348 #define EXT4_MIN_DESC_SIZE              32
 349 #define EXT4_MIN_DESC_SIZE_64BIT        64
 350 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
 351 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
 352 #ifdef __KERNEL__
 353 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
 354 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
 355 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
 356 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
 357 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
 358 #else
 359 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
 360 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
 361 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
 362 #endif
 363
 364 /*
 365  * Constants relative to the data blocks
 366  */
 367 #define EXT4_NDIR_BLOCKS                12
 368 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
 369 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
 370 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
 371 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
 372
 373 /*
 374  * Inode flags
 375  */
 376 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
 377 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
 378 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
 379 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
 380 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
 381 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
 382 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
 383 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
 384 /* Reserved for compression usage... */
 385 #define EXT4_DIRTY_FL                   0x00000100
 386 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
 387 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
 388 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
 389 /* End compression flags --- maybe not all used */
 390 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
 391 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
 392 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
 393 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
 394 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
 395 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
 396 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
 397 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
 398 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
 399 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
 400 #define EXT4_INLINE_DATA_FL             0x10000000 /* Inode has inline data. */
 401 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
 402
 403 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
 404 #define EXT4_FL_USER_MODIFIABLE         0x004380FF /* User modifiable flags */
 405
 406 /* Flags that should be inherited by new inodes from their parent. */
 407 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
 408                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
 409                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
 410                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
 411
 412 /* Flags that are appropriate for regular files (all but dir-specific ones). */
 413 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
 414
 415 /* Flags that are appropriate for non-directories/regular files. */
 416 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
 417
 418 /* Mask out flags that are inappropriate for the given type of inode. */
 419 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
 420 {
 421         if (S_ISDIR(mode))
 422                 return flags;
 423         else if (S_ISREG(mode))
 424                 return flags & EXT4_REG_FLMASK;
 425         else
 426                 return flags & EXT4_OTHER_FLMASK;
 427 }
 428
 429 /*
 430  * Inode flags used for atomic set/get
 431  */
 432 enum {
 433         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
 434         EXT4_INODE_UNRM         = 1,    /* Undelete */
 435         EXT4_INODE_COMPR        = 2,    /* Compress file */
 436         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
 437         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
 438         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
 439         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
 440         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
 441 /* Reserved for compression usage... */
 442         EXT4_INODE_DIRTY        = 8,
 443         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
 444         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
 445         EXT4_INODE_ECOMPR       = 11,   /* Compression error */
 446 /* End compression flags --- maybe not all used */
 447         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
 448         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
 449         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
 450         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
 451         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
 452         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
 453         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
 454         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
 455         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
 456         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
 457         EXT4_INODE_INLINE_DATA  = 28,   /* Data in inode. */
 458         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
 459 };
 460
 461 /*
 462  * Since it's pretty easy to mix up bit numbers and hex values, we use a
 463  * build-time check to make sure that EXT4_XXX_FL is consistent with respect to
 464  * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost
 465  * any extra space in the compiled kernel image, otherwise, the build will fail.
 466  * It's important that these values are the same, since we are using
 467  * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent
 468  * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk
 469  * values found in ext2, ext3 and ext4 filesystems, and of course the values
 470  * defined in e2fsprogs.
 471  *
 472  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
 473  */
 474 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
 475 #define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG))
 476
 477 static inline void ext4_check_flag_values(void)
 478 {
 479         CHECK_FLAG_VALUE(SECRM);
 480         CHECK_FLAG_VALUE(UNRM);
 481         CHECK_FLAG_VALUE(COMPR);
 482         CHECK_FLAG_VALUE(SYNC);
 483         CHECK_FLAG_VALUE(IMMUTABLE);
 484         CHECK_FLAG_VALUE(APPEND);
 485         CHECK_FLAG_VALUE(NODUMP);
 486         CHECK_FLAG_VALUE(NOATIME);
 487         CHECK_FLAG_VALUE(DIRTY);
 488         CHECK_FLAG_VALUE(COMPRBLK);
 489         CHECK_FLAG_VALUE(NOCOMPR);
 490         CHECK_FLAG_VALUE(ECOMPR);
 491         CHECK_FLAG_VALUE(INDEX);
 492         CHECK_FLAG_VALUE(IMAGIC);
 493         CHECK_FLAG_VALUE(JOURNAL_DATA);
 494         CHECK_FLAG_VALUE(NOTAIL);
 495         CHECK_FLAG_VALUE(DIRSYNC);
 496         CHECK_FLAG_VALUE(TOPDIR);
 497         CHECK_FLAG_VALUE(HUGE_FILE);
 498         CHECK_FLAG_VALUE(EXTENTS);
 499         CHECK_FLAG_VALUE(EA_INODE);
 500         CHECK_FLAG_VALUE(EOFBLOCKS);
 501         CHECK_FLAG_VALUE(INLINE_DATA);
 502         CHECK_FLAG_VALUE(RESERVED);
 503 }
 504
 505 /* Used to pass group descriptor data when online resize is done */
 506 struct ext4_new_group_input {
 507         __u32 group;            /* Group number for this data */
 508         __u64 block_bitmap;     /* Absolute block number of block bitmap */
 509         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
 510         __u64 inode_table;      /* Absolute block number of inode table start */
 511         __u32 blocks_count;     /* Total number of blocks in this group */
 512         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
 513         __u16 unused;
 514 };
 515
 516 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 517 struct compat_ext4_new_group_input {
 518         u32 group;
 519         compat_u64 block_bitmap;
 520         compat_u64 inode_bitmap;
 521         compat_u64 inode_table;
 522         u32 blocks_count;
 523         u16 reserved_blocks;
 524         u16 unused;
 525 };
 526 #endif
 527
 528 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
 529 struct ext4_new_group_data {
 530         __u32 group;
 531         __u64 block_bitmap;
 532         __u64 inode_bitmap;
 533         __u64 inode_table;
 534         __u32 blocks_count;
 535         __u16 reserved_blocks;
 536         __u16 unused;
 537         __u32 free_blocks_count;
 538 };
 539
 540 /* Indexes used to index group tables in ext4_new_group_data */
 541 enum {
 542         BLOCK_BITMAP = 0,       /* block bitmap */
 543         INODE_BITMAP,           /* inode bitmap */
 544         INODE_TABLE,            /* inode tables */
 545         GROUP_TABLE_COUNT,
 546 };
 547
 548 /*
 549  * Flags used by ext4_map_blocks()
 550  */
 551         /* Allocate any needed blocks and/or convert an unitialized
 552            extent to be an initialized ext4 */
 553 #define EXT4_GET_BLOCKS_CREATE                  0x0001
 554         /* Request the creation of an unitialized extent */
 555 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
 556 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
 557                                                  EXT4_GET_BLOCKS_CREATE)
 558         /* Caller is from the delayed allocation writeout path
 559          * finally doing the actual allocation of delayed blocks */
 560 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
 561         /* caller is from the direct IO path, request to creation of an
 562         unitialized extents if not allocated, split the uninitialized
 563         extent if blocks has been preallocated already*/
 564 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
 565 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
 566 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
 567                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
 568         /* Convert extent to initialized after IO complete */
 569 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
 570                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
 571         /* Eventual metadata allocation (due to growing extent tree)
 572          * should not fail, so try to use reserved blocks for that.*/
 573 #define EXT4_GET_BLOCKS_METADATA_NOFAIL         0x0020
 574         /* Don't normalize allocation size (used for fallocate) */
 575 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
 576         /* Request will not result in inode size update (user for fallocate) */
 577 #define EXT4_GET_BLOCKS_KEEP_SIZE               0x0080
 578         /* Do not take i_data_sem locking in ext4_map_blocks */
 579 #define EXT4_GET_BLOCKS_NO_LOCK                 0x0100
 580         /* Do not put hole in extent cache */
 581 #define EXT4_GET_BLOCKS_NO_PUT_HOLE             0x0200
 582
 583 /*
 584  * Flags used by ext4_free_blocks
 585  */
 586 #define EXT4_FREE_BLOCKS_METADATA       0x0001
 587 #define EXT4_FREE_BLOCKS_FORGET         0x0002
 588 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
 589 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
 590 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
 591 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
 592
 593 /*
 594  * Flags used by ext4_discard_partial_page_buffers
 595  */
 596 #define EXT4_DISCARD_PARTIAL_PG_ZERO_UNMAPPED   0x0001
 597
 598 /*
 599  * ioctl commands
 600  */
 601 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
 602 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
 603 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
 604 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
 605 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
 606 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
 607 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
 608 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
 609 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
 610 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
 611 #define EXT4_IOC_MIGRATE                _IO('f', 9)
 612  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
 613  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
 614 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
 615 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
 616 #define EXT4_IOC_RESIZE_FS              _IOW('f', 16, __u64)
 617 #define EXT4_IOC_SWAP_BOOT              _IO('f', 17)
 618
 619 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 620 /*
 621  * ioctl commands in 32 bit emulation
 622  */
 623 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
 624 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
 625 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
 626 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
 627 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
 628 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
 629 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
 630 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
 631 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
 632 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
 633 #endif
 634
 635 /* Max physical block we can address w/o extents */
 636 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
 637
 638 /*
 639  * Structure of an inode on the disk
 640  */
 641 struct ext4_inode {
 642         __le16  i_mode;         /* File mode */
 643         __le16  i_uid;          /* Low 16 bits of Owner Uid */
 644         __le32  i_size_lo;      /* Size in bytes */
 645         __le32  i_atime;        /* Access time */
 646         __le32  i_ctime;        /* Inode Change time */
 647         __le32  i_mtime;        /* Modification time */
 648         __le32  i_dtime;        /* Deletion Time */
 649         __le16  i_gid;          /* Low 16 bits of Group Id */
 650         __le16  i_links_count;  /* Links count */
 651         __le32  i_blocks_lo;    /* Blocks count */
 652         __le32  i_flags;        /* File flags */
 653         union {
 654                 struct {
 655                         __le32  l_i_version;
 656                 } linux1;
 657                 struct {
 658                         __u32  h_i_translator;
 659                 } hurd1;
 660                 struct {
 661                         __u32  m_i_reserved1;
 662                 } masix1;
 663         } osd1;                         /* OS dependent 1 */
 664         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
 665         __le32  i_generation;   /* File version (for NFS) */
 666         __le32  i_file_acl_lo;  /* File ACL */
 667         __le32  i_size_high;
 668         __le32  i_obso_faddr;   /* Obsoleted fragment address */
 669         union {
 670                 struct {
 671                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
 672                         __le16  l_i_file_acl_high;
 673                         __le16  l_i_uid_high;   /* these 2 fields */
 674                         __le16  l_i_gid_high;   /* were reserved2[0] */
 675                         __le16  l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
 676                         __le16  l_i_reserved;
 677                 } linux2;
 678                 struct {
 679                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 680                         __u16   h_i_mode_high;
 681                         __u16   h_i_uid_high;
 682                         __u16   h_i_gid_high;
 683                         __u32   h_i_author;
 684                 } hurd2;
 685                 struct {
 686                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 687                         __le16  m_i_file_acl_high;
 688                         __u32   m_i_reserved2[2];
 689                 } masix2;
 690         } osd2;                         /* OS dependent 2 */
 691         __le16  i_extra_isize;
 692         __le16  i_checksum_hi;  /* crc32c(uuid+inum+inode) BE */
 693         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
 694         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
 695         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
 696         __le32  i_crtime;       /* File Creation time */
 697         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
 698         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
 699 };
 700
 701 struct move_extent {
 702         __u32 reserved;         /* should be zero */
 703         __u32 donor_fd;         /* donor file descriptor */
 704         __u64 orig_start;       /* logical start offset in block for orig */
 705         __u64 donor_start;      /* logical start offset in block for donor */
 706         __u64 len;              /* block length to be moved */
 707         __u64 moved_len;        /* moved block length */
 708 };
 709
 710 #define EXT4_EPOCH_BITS 2
 711 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
 712 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
 713
 714 /*
 715  * Extended fields will fit into an inode if the filesystem was formatted
 716  * with large inodes (-I 256 or larger) and there are not currently any EAs
 717  * consuming all of the available space. For new inodes we always reserve
 718  * enough space for the kernel's known extended fields, but for inodes
 719  * created with an old kernel this might not have been the case. None of
 720  * the extended inode fields is critical for correct filesystem operation.
 721  * This macro checks if a certain field fits in the inode. Note that
 722  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
 723  */
 724 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
 725         ((offsetof(typeof(*ext4_inode), field) +        \
 726           sizeof((ext4_inode)->field))                  \
 727         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
 728             (einode)->i_extra_isize))                   \
 729
 730 static inline __le32 ext4_encode_extra_time(struct timespec *time)
 731 {
 732        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
 733                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
 734                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
 735 }
 736
 737 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
 738 {
 739        if (sizeof(time->tv_sec) > 4)
 740                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
 741                                << 32;
 742        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
 743 }
 744
 745 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
 746 do {                                                                           \
 747         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
 748         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 749                 (raw_inode)->xtime ## _extra =                                 \
 750                                 ext4_encode_extra_time(&(inode)->xtime);       \
 751 } while (0)
 752
 753 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
 754 do {                                                                           \
 755         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 756                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
 757         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 758                 (raw_inode)->xtime ## _extra =                                 \
 759                                 ext4_encode_extra_time(&(einode)->xtime);      \
 760 } while (0)
 761
 762 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
 763 do {                                                                           \
 764         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
 765         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 766                 ext4_decode_extra_time(&(inode)->xtime,                        \
 767                                        raw_inode->xtime ## _extra);            \
 768         else                                                                   \
 769                 (inode)->xtime.tv_nsec = 0;                                    \
 770 } while (0)
 771
 772 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
 773 do {                                                                           \
 774         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 775                 (einode)->xtime.tv_sec =                                       \
 776                         (signed)le32_to_cpu((raw_inode)->xtime);               \
 777         else                                                                   \
 778                 (einode)->xtime.tv_sec = 0;                                    \
 779         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 780                 ext4_decode_extra_time(&(einode)->xtime,                       \
 781                                        raw_inode->xtime ## _extra);            \
 782         else                                                                   \
 783                 (einode)->xtime.tv_nsec = 0;                                   \
 784 } while (0)
 785
 786 #define i_disk_version osd1.linux1.l_i_version
 787
 788 #if defined(__KERNEL__) || defined(__linux__)
 789 #define i_reserved1     osd1.linux1.l_i_reserved1
 790 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
 791 #define i_blocks_high   osd2.linux2.l_i_blocks_high
 792 #define i_uid_low       i_uid
 793 #define i_gid_low       i_gid
 794 #define i_uid_high      osd2.linux2.l_i_uid_high
 795 #define i_gid_high      osd2.linux2.l_i_gid_high
 796 #define i_checksum_lo   osd2.linux2.l_i_checksum_lo
 797
 798 #elif defined(__GNU__)
 799
 800 #define i_translator    osd1.hurd1.h_i_translator
 801 #define i_uid_high      osd2.hurd2.h_i_uid_high
 802 #define i_gid_high      osd2.hurd2.h_i_gid_high
 803 #define i_author        osd2.hurd2.h_i_author
 804
 805 #elif defined(__masix__)
 806
 807 #define i_reserved1     osd1.masix1.m_i_reserved1
 808 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
 809 #define i_reserved2     osd2.masix2.m_i_reserved2
 810
 811 #endif /* defined(__KERNEL__) || defined(__linux__) */
 812
 813 #include "extents_status.h"
 814
 815 /*
 816  * fourth extended file system inode data in memory
 817  */
 818 struct ext4_inode_info {
 819         __le32  i_data[15];     /* unconverted */
 820         __u32   i_dtime;
 821         ext4_fsblk_t    i_file_acl;
 822
 823         /*
 824          * i_block_group is the number of the block group which contains
 825          * this file's inode.  Constant across the lifetime of the inode,
 826          * it is ued for making block allocation decisions - we try to
 827          * place a file's data blocks near its inode block, and new inodes
 828          * near to their parent directory's inode.
 829          */
 830         ext4_group_t    i_block_group;
 831         ext4_lblk_t     i_dir_start_lookup;
 832 #if (BITS_PER_LONG < 64)
 833         unsigned long   i_state_flags;          /* Dynamic state flags */
 834 #endif
 835         unsigned long   i_flags;
 836
 837         /*
 838          * Extended attributes can be read independently of the main file
 839          * data. Taking i_mutex even when reading would cause contention
 840          * between readers of EAs and writers of regular file data, so
 841          * instead we synchronize on xattr_sem when reading or changing
 842          * EAs.
 843          */
 844         struct rw_semaphore xattr_sem;
 845
 846         struct list_head i_orphan;      /* unlinked but open inodes */
 847
 848         /*
 849          * i_disksize keeps track of what the inode size is ON DISK, not
 850          * in memory.  During truncate, i_size is set to the new size by
 851          * the VFS prior to calling ext4_truncate(), but the filesystem won't
 852          * set i_disksize to 0 until the truncate is actually under way.
 853          *
 854          * The intent is that i_disksize always represents the blocks which
 855          * are used by this file.  This allows recovery to restart truncate
 856          * on orphans if we crash during truncate.  We actually write i_disksize
 857          * into the on-disk inode when writing inodes out, instead of i_size.
 858          *
 859          * The only time when i_disksize and i_size may be different is when
 860          * a truncate is in progress.  The only things which change i_disksize
 861          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
 862          */
 863         loff_t  i_disksize;
 864
 865         /*
 866          * i_data_sem is for serialising ext4_truncate() against
 867          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
 868          * data tree are chopped off during truncate. We can't do that in
 869          * ext4 because whenever we perform intermediate commits during
 870          * truncate, the inode and all the metadata blocks *must* be in a
 871          * consistent state which allows truncation of the orphans to restart
 872          * during recovery.  Hence we must fix the get_block-vs-truncate race
 873          * by other means, so we have i_data_sem.
 874          */
 875         struct rw_semaphore i_data_sem;
 876         struct inode vfs_inode;
 877         struct jbd2_inode *jinode;
 878
 879         /*
 880          * File creation time. Its function is same as that of
 881          * struct timespec i_{a,c,m}time in the generic inode.
 882          */
 883         struct timespec i_crtime;
 884
 885         /* mballoc */
 886         struct list_head i_prealloc_list;
 887         spinlock_t i_prealloc_lock;
 888
 889         /* extents status tree */
 890         struct ext4_es_tree i_es_tree;
 891         rwlock_t i_es_lock;
 892         struct list_head i_es_lru;
 893         unsigned int i_es_lru_nr;       /* protected by i_es_lock */
 894
 895         /* ialloc */
 896         ext4_group_t    i_last_alloc_group;
 897
 898         /* allocation reservation info for delalloc */
 899         /* In case of bigalloc, these refer to clusters rather than blocks */
 900         unsigned int i_reserved_data_blocks;
 901         unsigned int i_reserved_meta_blocks;
 902         unsigned int i_allocated_meta_blocks;
 903         ext4_lblk_t i_da_metadata_calc_last_lblock;
 904         int i_da_metadata_calc_len;
 905
 906         /* on-disk additional length */
 907         __u16 i_extra_isize;
 908
 909         /* Indicate the inline data space. */
 910         u16 i_inline_off;
 911         u16 i_inline_size;
 912
 913 #ifdef CONFIG_QUOTA
 914         /* quota space reservation, managed internally by quota code */
 915         qsize_t i_reserved_quota;
 916 #endif
 917
 918         /* completed IOs that might need unwritten extents handling */
 919         struct list_head i_completed_io_list;
 920         spinlock_t i_completed_io_lock;
 921         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
 922         atomic_t i_unwritten; /* Nr. of inflight conversions pending */
 923         struct work_struct i_unwritten_work;    /* deferred extent conversion */
 924
 925         spinlock_t i_block_reservation_lock;
 926
 927         /*
 928          * Transactions that contain inode's metadata needed to complete
 929          * fsync and fdatasync, respectively.
 930          */
 931         tid_t i_sync_tid;
 932         tid_t i_datasync_tid;
 933
 934         /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
 935         __u32 i_csum_seed;
 936 };
 937
 938 /*
 939  * File system states
 940  */
 941 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
 942 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
 943 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
 944
 945 /*
 946  * Misc. filesystem flags
 947  */
 948 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
 949 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
 950 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
 951
 952 /*
 953  * Mount flags set via mount options or defaults
 954  */
 955 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
 956 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
 957 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
 958 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
 959 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
 960 #define EXT4_MOUNT_ERRORS_MASK          0x00070
 961 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
 962 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
 963 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
 964 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
 965 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
 966 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
 967 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
 968 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
 969 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
 970 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
 971 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
 972 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
 973 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
 974 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
 975 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
 976 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
 977 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
 978 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
 979 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
 980 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
 981 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
 982 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
 983 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
 984
 985 /*
 986  * Mount flags set either automatically (could not be set by mount option)
 987  * based on per file system feature or property or in special cases such as
 988  * distinguishing between explicit mount option definition and default.
 989  */
 990 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
 991                                                       specified delalloc */
 992 #define EXT4_MOUNT2_STD_GROUP_SIZE      0x00000002 /* We have standard group
 993                                                       size of blocksize * 8
 994                                                       blocks */
 995
 996 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
 997                                                 ~EXT4_MOUNT_##opt
 998 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
 999                                                 EXT4_MOUNT_##opt
1000 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
1001                                          EXT4_MOUNT_##opt)
1002
1003 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
1004                                                 ~EXT4_MOUNT2_##opt
1005 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
1006                                                 EXT4_MOUNT2_##opt
1007 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
1008                                          EXT4_MOUNT2_##opt)
1009
1010 #define ext4_test_and_set_bit           __test_and_set_bit_le
1011 #define ext4_set_bit                    __set_bit_le
1012 #define ext4_set_bit_atomic             ext2_set_bit_atomic
1013 #define ext4_test_and_clear_bit         __test_and_clear_bit_le
1014 #define ext4_clear_bit                  __clear_bit_le
1015 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
1016 #define ext4_test_bit                   test_bit_le
1017 #define ext4_find_next_zero_bit         find_next_zero_bit_le
1018 #define ext4_find_next_bit              find_next_bit_le
1019
1020 extern void ext4_set_bits(void *bm, int cur, int len);
1021
1022 /*
1023  * Maximal mount counts between two filesystem checks
1024  */
1025 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
1026 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
1027
1028 /*
1029  * Behaviour when detecting errors
1030  */
1031 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
1032 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
1033 #define EXT4_ERRORS_PANIC               3       /* Panic */
1034 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
1035
1036 /* Metadata checksum algorithm codes */
1037 #define EXT4_CRC32C_CHKSUM              1
1038
1039 /*
1040  * Structure of the super block
1041  */
1042 struct ext4_super_block {
1043 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1044         __le32  s_blocks_count_lo;      /* Blocks count */
1045         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1046         __le32  s_free_blocks_count_lo; /* Free blocks count */
1047 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1048         __le32  s_first_data_block;     /* First Data Block */
1049         __le32  s_log_block_size;       /* Block size */
1050         __le32  s_log_cluster_size;     /* Allocation cluster size */
1051 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1052         __le32  s_clusters_per_group;   /* # Clusters per group */
1053         __le32  s_inodes_per_group;     /* # Inodes per group */
1054         __le32  s_mtime;                /* Mount time */
1055 /*30*/  __le32  s_wtime;                /* Write time */
1056         __le16  s_mnt_count;            /* Mount count */
1057         __le16  s_max_mnt_count;        /* Maximal mount count */
1058         __le16  s_magic;                /* Magic signature */
1059         __le16  s_state;                /* File system state */
1060         __le16  s_errors;               /* Behaviour when detecting errors */
1061         __le16  s_minor_rev_level;      /* minor revision level */
1062 /*40*/  __le32  s_lastcheck;            /* time of last check */
1063         __le32  s_checkinterval;        /* max. time between checks */
1064         __le32  s_creator_os;           /* OS */
1065         __le32  s_rev_level;            /* Revision level */
1066 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1067         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1068         /*
1069          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1070          *
1071          * Note: the difference between the compatible feature set and
1072          * the incompatible feature set is that if there is a bit set
1073          * in the incompatible feature set that the kernel doesn't
1074          * know about, it should refuse to mount the filesystem.
1075          *
1076          * e2fsck's requirements are more strict; if it doesn't know
1077          * about a feature in either the compatible or incompatible
1078          * feature set, it must abort and not try to meddle with
1079          * things it doesn't understand...
1080          */
1081         __le32  s_first_ino;            /* First non-reserved inode */
1082         __le16  s_inode_size;           /* size of inode structure */
1083         __le16  s_block_group_nr;       /* block group # of this superblock */
1084         __le32  s_feature_compat;       /* compatible feature set */
1085 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1086         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1087 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1088 /*78*/  char    s_volume_name[16];      /* volume name */
1089 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1090 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1091         /*
1092          * Performance hints.  Directory preallocation should only
1093          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1094          */
1095         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1096         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1097         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1098         /*
1099          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1100          */
1101 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1102 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1103         __le32  s_journal_dev;          /* device number of journal file */
1104         __le32  s_last_orphan;          /* start of list of inodes to delete */
1105         __le32  s_hash_seed[4];         /* HTREE hash seed */
1106         __u8    s_def_hash_version;     /* Default hash version to use */
1107         __u8    s_jnl_backup_type;
1108         __le16  s_desc_size;            /* size of group descriptor */
1109 /*100*/ __le32  s_default_mount_opts;
1110         __le32  s_first_meta_bg;        /* First metablock block group */
1111         __le32  s_mkfs_time;            /* When the filesystem was created */
1112         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1113         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1114 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1115         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1116         __le32  s_free_blocks_count_hi; /* Free blocks count */
1117         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1118         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1119         __le32  s_flags;                /* Miscellaneous flags */
1120         __le16  s_raid_stride;          /* RAID stride */
1121         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1122         __le64  s_mmp_block;            /* Block for multi-mount protection */
1123         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1124         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1125         __u8    s_checksum_type;        /* metadata checksum algorithm used */
1126         __le16  s_reserved_pad;
1127         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1128         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1129         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1130         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1131                                               snapshot's future use */
1132         __le32  s_snapshot_list;        /* inode number of the head of the
1133                                            on-disk snapshot list */
1134 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1135         __le32  s_error_count;          /* number of fs errors */
1136         __le32  s_first_error_time;     /* first time an error happened */
1137         __le32  s_first_error_ino;      /* inode involved in first error */
1138         __le64  s_first_error_block;    /* block involved of first error */
1139         __u8    s_first_error_func[32]; /* function where the error happened */
1140         __le32  s_first_error_line;     /* line number where error happened */
1141         __le32  s_last_error_time;      /* most recent time of an error */
1142         __le32  s_last_error_ino;       /* inode involved in last error */
1143         __le32  s_last_error_line;      /* line number where error happened */
1144         __le64  s_last_error_block;     /* block involved of last error */
1145         __u8    s_last_error_func[32];  /* function where the error happened */
1146 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1147         __u8    s_mount_opts[64];
1148         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1149         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1150         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1151         __le32  s_reserved[108];        /* Padding to the end of the block */
1152         __le32  s_checksum;             /* crc32c(superblock) */
1153 };
1154
1155 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1156
1157 #ifdef __KERNEL__
1158
1159 /*
1160  * run-time mount flags
1161  */
1162 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
1163 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
1164
1165 /*
1166  * fourth extended-fs super-block data in memory
1167  */
1168 struct ext4_sb_info {
1169         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1170         unsigned long s_inodes_per_block;/* Number of inodes per block */
1171         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1172         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1173         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1174         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1175         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1176         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1177         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1178         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1179         unsigned long s_overhead;  /* # of fs overhead clusters */
1180         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1181         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1182         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1183         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1184         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1185         struct buffer_head **s_group_desc;
1186         unsigned int s_mount_opt;
1187         unsigned int s_mount_opt2;
1188         unsigned int s_mount_flags;
1189         unsigned int s_def_mount_opt;
1190         ext4_fsblk_t s_sb_block;
1191         atomic64_t s_resv_clusters;
1192         kuid_t s_resuid;
1193         kgid_t s_resgid;
1194         unsigned short s_mount_state;
1195         unsigned short s_pad;
1196         int s_addr_per_block_bits;
1197         int s_desc_per_block_bits;
1198         int s_inode_size;
1199         int s_first_ino;
1200         unsigned int s_inode_readahead_blks;
1201         unsigned int s_inode_goal;
1202         spinlock_t s_next_gen_lock;
1203         u32 s_next_generation;
1204         u32 s_hash_seed[4];
1205         int s_def_hash_version;
1206         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1207         struct percpu_counter s_freeclusters_counter;
1208         struct percpu_counter s_freeinodes_counter;
1209         struct percpu_counter s_dirs_counter;
1210         struct percpu_counter s_dirtyclusters_counter;
1211         struct blockgroup_lock *s_blockgroup_lock;
1212         struct proc_dir_entry *s_proc;
1213         struct kobject s_kobj;
1214         struct completion s_kobj_unregister;
1215         struct super_block *s_sb;
1216
1217         /* Journaling */
1218         struct journal_s *s_journal;
1219         struct list_head s_orphan;
1220         struct mutex s_orphan_lock;
1221         unsigned long s_resize_flags;           /* Flags indicating if there
1222                                                    is a resizer */
1223         unsigned long s_commit_interval;
1224         u32 s_max_batch_time;
1225         u32 s_min_batch_time;
1226         struct block_device *journal_bdev;
1227 #ifdef CONFIG_QUOTA
1228         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
1229         int s_jquota_fmt;                       /* Format of quota to use */
1230 #endif
1231         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1232         struct rb_root system_blks;
1233
1234 #ifdef EXTENTS_STATS
1235         /* ext4 extents stats */
1236         unsigned long s_ext_min;
1237         unsigned long s_ext_max;
1238         unsigned long s_depth_max;
1239         spinlock_t s_ext_stats_lock;
1240         unsigned long s_ext_blocks;
1241         unsigned long s_ext_extents;
1242 #endif
1243
1244         /* for buddy allocator */
1245         struct ext4_group_info ***s_group_info;
1246         struct inode *s_buddy_cache;
1247         spinlock_t s_md_lock;
1248         unsigned short *s_mb_offsets;
1249         unsigned int *s_mb_maxs;
1250         unsigned int s_group_info_size;
1251
1252         /* tunables */
1253         unsigned long s_stripe;
1254         unsigned int s_mb_stream_request;
1255         unsigned int s_mb_max_to_scan;
1256         unsigned int s_mb_min_to_scan;
1257         unsigned int s_mb_stats;
1258         unsigned int s_mb_order2_reqs;
1259         unsigned int s_mb_group_prealloc;
1260         unsigned int s_max_writeback_mb_bump;
1261         unsigned int s_max_dir_size_kb;
1262         /* where last allocation was done - for stream allocation */
1263         unsigned long s_mb_last_group;
1264         unsigned long s_mb_last_start;
1265
1266         /* stats for buddy allocator */
1267         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1268         atomic_t s_bal_success; /* we found long enough chunks */
1269         atomic_t s_bal_allocated;       /* in blocks */
1270         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1271         atomic_t s_bal_goals;   /* goal hits */
1272         atomic_t s_bal_breaks;  /* too long searches */
1273         atomic_t s_bal_2orders; /* 2^order hits */
1274         spinlock_t s_bal_lock;
1275         unsigned long s_mb_buddies_generated;
1276         unsigned long long s_mb_generation_time;
1277         atomic_t s_mb_lost_chunks;
1278         atomic_t s_mb_preallocated;
1279         atomic_t s_mb_discarded;
1280         atomic_t s_lock_busy;
1281
1282         /* locality groups */
1283         struct ext4_locality_group __percpu *s_locality_groups;
1284
1285         /* for write statistics */
1286         unsigned long s_sectors_written_start;
1287         u64 s_kbytes_written;
1288
1289         /* the size of zero-out chunk */
1290         unsigned int s_extent_max_zeroout_kb;
1291
1292         unsigned int s_log_groups_per_flex;
1293         struct flex_groups *s_flex_groups;
1294         ext4_group_t s_flex_groups_allocated;
1295
1296         /* workqueue for dio unwritten */
1297         struct workqueue_struct *dio_unwritten_wq;
1298
1299         /* timer for periodic error stats printing */
1300         struct timer_list s_err_report;
1301
1302         /* Lazy inode table initialization info */
1303         struct ext4_li_request *s_li_request;
1304         /* Wait multiplier for lazy initialization thread */
1305         unsigned int s_li_wait_mult;
1306
1307         /* Kernel thread for multiple mount protection */
1308         struct task_struct *s_mmp_tsk;
1309
1310         /* record the last minlen when FITRIM is called. */
1311         atomic_t s_last_trim_minblks;
1312
1313         /* Reference to checksum algorithm driver via cryptoapi */
1314         struct crypto_shash *s_chksum_driver;
1315
1316         /* Precomputed FS UUID checksum for seeding other checksums */
1317         __u32 s_csum_seed;
1318
1319         /* Reclaim extents from extent status tree */
1320         struct shrinker s_es_shrinker;
1321         struct list_head s_es_lru;
1322         struct percpu_counter s_extent_cache_cnt;
1323         spinlock_t s_es_lru_lock ____cacheline_aligned_in_smp;
1324 };
1325
1326 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1327 {
1328         return sb->s_fs_info;
1329 }
1330 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1331 {
1332         return container_of(inode, struct ext4_inode_info, vfs_inode);
1333 }
1334
1335 static inline struct timespec ext4_current_time(struct inode *inode)
1336 {
1337         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1338                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1339 }
1340
1341 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1342 {
1343         return ino == EXT4_ROOT_INO ||
1344                 ino == EXT4_USR_QUOTA_INO ||
1345                 ino == EXT4_GRP_QUOTA_INO ||
1346                 ino == EXT4_BOOT_LOADER_INO ||
1347                 ino == EXT4_JOURNAL_INO ||
1348                 ino == EXT4_RESIZE_INO ||
1349                 (ino >= EXT4_FIRST_INO(sb) &&
1350                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1351 }
1352
1353 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
1354                                               struct ext4_io_end *io_end)
1355 {
1356         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
1357                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
1358                 atomic_inc(&EXT4_I(inode)->i_unwritten);
1359         }
1360 }
1361
1362 static inline ext4_io_end_t *ext4_inode_aio(struct inode *inode)
1363 {
1364         return inode->i_private;
1365 }
1366
1367 static inline void ext4_inode_aio_set(struct inode *inode, ext4_io_end_t *io)
1368 {
1369         inode->i_private = io;
1370 }
1371
1372 /*
1373  * Inode dynamic state flags
1374  */
1375 enum {
1376         EXT4_STATE_JDATA,               /* journaled data exists */
1377         EXT4_STATE_NEW,                 /* inode is newly created */
1378         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1379         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1380         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1381         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1382         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1383         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1384         EXT4_STATE_DELALLOC_RESERVED,   /* blks already reserved for delalloc */
1385         EXT4_STATE_DIOREAD_LOCK,        /* Disable support for dio read
1386                                            nolocking */
1387         EXT4_STATE_MAY_INLINE_DATA,     /* may have in-inode data */
1388         EXT4_STATE_ORDERED_MODE,        /* data=ordered mode */
1389 };
1390
1391 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1392 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1393 {                                                                       \
1394         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1395 }                                                                       \
1396 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1397 {                                                                       \
1398         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1399 }                                                                       \
1400 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1401 {                                                                       \
1402         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1403 }
1404
1405 EXT4_INODE_BIT_FNS(flag, flags, 0)
1406 #if (BITS_PER_LONG < 64)
1407 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1408
1409 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1410 {
1411         (ei)->i_state_flags = 0;
1412 }
1413 #else
1414 EXT4_INODE_BIT_FNS(state, flags, 32)
1415
1416 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1417 {
1418         /* We depend on the fact that callers will set i_flags */
1419 }
1420 #endif
1421 #else
1422 /* Assume that user mode programs are passing in an ext4fs superblock, not
1423  * a kernel struct super_block.  This will allow us to call the feature-test
1424  * macros from user land. */
1425 #define EXT4_SB(sb)     (sb)
1426 #endif
1427
1428 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1429
1430 /*
1431  * Codes for operating systems
1432  */
1433 #define EXT4_OS_LINUX           0
1434 #define EXT4_OS_HURD            1
1435 #define EXT4_OS_MASIX           2
1436 #define EXT4_OS_FREEBSD         3
1437 #define EXT4_OS_LITES           4
1438
1439 /*
1440  * Revision levels
1441  */
1442 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1443 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1444
1445 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1446 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1447
1448 #define EXT4_GOOD_OLD_INODE_SIZE 128
1449
1450 /*
1451  * Feature set definitions
1452  */
1453
1454 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1455         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1456 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1457         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1458 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1459         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1460 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1461         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1462 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1463         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1464 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1465         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1466 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1467         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1468 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1469         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1470 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1471         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1472
1473 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1474 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1475 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1476 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1477 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1478 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1479
1480 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1481 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1482 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1483 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1484 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1485 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1486 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1487 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1488 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1489 /*
1490  * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM).  When
1491  * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
1492  * all other data structures' checksums.  However, the METADATA_CSUM and
1493  * GDT_CSUM bits are mutually exclusive.
1494  */
1495 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM    0x0400
1496
1497 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1498 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1499 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1500 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1501 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1502 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1503 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1504 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1505 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1506 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1507 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1508 #define EXT4_FEATURE_INCOMPAT_BG_USE_META_CSUM  0x2000 /* use crc32c for bg */
1509 #define EXT4_FEATURE_INCOMPAT_LARGEDIR          0x4000 /* >2GB or 3-lvl htree */
1510 #define EXT4_FEATURE_INCOMPAT_INLINE_DATA       0x8000 /* data in inode */
1511
1512 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1513 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1514                                          EXT4_FEATURE_INCOMPAT_META_BG)
1515 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1516                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1517                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1518
1519 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1520 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1521                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1522                                          EXT4_FEATURE_INCOMPAT_META_BG)
1523 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1524                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1525                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1526
1527 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1528 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1529                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1530                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1531                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1532                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1533                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1534                                          EXT4_FEATURE_INCOMPAT_MMP |    \
1535                                          EXT4_FEATURE_INCOMPAT_INLINE_DATA)
1536 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1537                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1538                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1539                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1540                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1541                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1542                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1543                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
1544                                          EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
1545                                          EXT4_FEATURE_RO_COMPAT_QUOTA)
1546
1547 /*
1548  * Default values for user and/or group using reserved blocks
1549  */
1550 #define EXT4_DEF_RESUID         0
1551 #define EXT4_DEF_RESGID         0
1552
1553 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1554
1555 /*
1556  * Default mount options
1557  */
1558 #define EXT4_DEFM_DEBUG         0x0001
1559 #define EXT4_DEFM_BSDGROUPS     0x0002
1560 #define EXT4_DEFM_XATTR_USER    0x0004
1561 #define EXT4_DEFM_ACL           0x0008
1562 #define EXT4_DEFM_UID16         0x0010
1563 #define EXT4_DEFM_JMODE         0x0060
1564 #define EXT4_DEFM_JMODE_DATA    0x0020
1565 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1566 #define EXT4_DEFM_JMODE_WBACK   0x0060
1567 #define EXT4_DEFM_NOBARRIER     0x0100
1568 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1569 #define EXT4_DEFM_DISCARD       0x0400
1570 #define EXT4_DEFM_NODELALLOC    0x0800
1571
1572 /*
1573  * Default journal batch times
1574  */
1575 #define EXT4_DEF_MIN_BATCH_TIME 0
1576 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1577
1578 /*
1579  * Minimum number of groups in a flexgroup before we separate out
1580  * directories into the first block group of a flexgroup
1581  */
1582 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1583
1584 /*
1585  * Structure of a directory entry
1586  */
1587 #define EXT4_NAME_LEN 255
1588
1589 struct ext4_dir_entry {
1590         __le32  inode;                  /* Inode number */
1591         __le16  rec_len;                /* Directory entry length */
1592         __le16  name_len;               /* Name length */
1593         char    name[EXT4_NAME_LEN];    /* File name */
1594 };
1595
1596 /*
1597  * The new version of the directory entry.  Since EXT4 structures are
1598  * stored in intel byte order, and the name_len field could never be
1599  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1600  * file_type field.
1601  */
1602 struct ext4_dir_entry_2 {
1603         __le32  inode;                  /* Inode number */
1604         __le16  rec_len;                /* Directory entry length */
1605         __u8    name_len;               /* Name length */
1606         __u8    file_type;
1607         char    name[EXT4_NAME_LEN];    /* File name */
1608 };
1609
1610 /*
1611  * This is a bogus directory entry at the end of each leaf block that
1612  * records checksums.
1613  */
1614 struct ext4_dir_entry_tail {
1615         __le32  det_reserved_zero1;     /* Pretend to be unused */
1616         __le16  det_rec_len;            /* 12 */
1617         __u8    det_reserved_zero2;     /* Zero name length */
1618         __u8    det_reserved_ft;        /* 0xDE, fake file type */
1619         __le32  det_checksum;           /* crc32c(uuid+inum+dirblock) */
1620 };
1621
1622 #define EXT4_DIRENT_TAIL(block, blocksize) \
1623         ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
1624                                         ((blocksize) - \
1625                                          sizeof(struct ext4_dir_entry_tail))))
1626
1627 /*
1628  * Ext4 directory file types.  Only the low 3 bits are used.  The
1629  * other bits are reserved for now.
1630  */
1631 #define EXT4_FT_UNKNOWN         0
1632 #define EXT4_FT_REG_FILE        1
1633 #define EXT4_FT_DIR             2
1634 #define EXT4_FT_CHRDEV          3
1635 #define EXT4_FT_BLKDEV          4
1636 #define EXT4_FT_FIFO            5
1637 #define EXT4_FT_SOCK            6
1638 #define EXT4_FT_SYMLINK         7
1639
1640 #define EXT4_FT_MAX             8
1641
1642 #define EXT4_FT_DIR_CSUM        0xDE
1643
1644 /*
1645  * EXT4_DIR_PAD defines the directory entries boundaries
1646  *
1647  * NOTE: It must be a multiple of 4
1648  */
1649 #define EXT4_DIR_PAD                    4
1650 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1651 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1652                                          ~EXT4_DIR_ROUND)
1653 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1654
1655 /*
1656  * If we ever get support for fs block sizes > page_size, we'll need
1657  * to remove the #if statements in the next two functions...
1658  */
1659 static inline unsigned int
1660 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1661 {
1662         unsigned len = le16_to_cpu(dlen);
1663
1664 #if (PAGE_CACHE_SIZE >= 65536)
1665         if (len == EXT4_MAX_REC_LEN || len == 0)
1666                 return blocksize;
1667         return (len & 65532) | ((len & 3) << 16);
1668 #else
1669         return len;
1670 #endif
1671 }
1672
1673 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1674 {
1675         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1676                 BUG();
1677 #if (PAGE_CACHE_SIZE >= 65536)
1678         if (len < 65536)
1679                 return cpu_to_le16(len);
1680         if (len == blocksize) {
1681                 if (blocksize == 65536)
1682                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1683                 else
1684                         return cpu_to_le16(0);
1685         }
1686         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1687 #else
1688         return cpu_to_le16(len);
1689 #endif
1690 }
1691
1692 /*
1693  * Hash Tree Directory indexing
1694  * (c) Daniel Phillips, 2001
1695  */
1696
1697 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1698                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1699                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1700 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1701 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1702
1703 /* Legal values for the dx_root hash_version field: */
1704
1705 #define DX_HASH_LEGACY          0
1706 #define DX_HASH_HALF_MD4        1
1707 #define DX_HASH_TEA             2
1708 #define DX_HASH_LEGACY_UNSIGNED 3
1709 #define DX_HASH_HALF_MD4_UNSIGNED       4
1710 #define DX_HASH_TEA_UNSIGNED            5
1711
1712 static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
1713                               const void *address, unsigned int length)
1714 {
1715         struct {
1716                 struct shash_desc shash;
1717                 char ctx[4];
1718         } desc;
1719         int err;
1720
1721         BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
1722
1723         desc.shash.tfm = sbi->s_chksum_driver;
1724         desc.shash.flags = 0;
1725         *(u32 *)desc.ctx = crc;
1726
1727         err = crypto_shash_update(&desc.shash, address, length);
1728         BUG_ON(err);
1729
1730         return *(u32 *)desc.ctx;
1731 }
1732
1733 #ifdef __KERNEL__
1734
1735 /* hash info structure used by the directory hash */
1736 struct dx_hash_info
1737 {
1738         u32             hash;
1739         u32             minor_hash;
1740         int             hash_version;
1741         u32             *seed;
1742 };
1743
1744
1745 /* 32 and 64 bit signed EOF for dx directories */
1746 #define EXT4_HTREE_EOF_32BIT   ((1UL  << (32 - 1)) - 1)
1747 #define EXT4_HTREE_EOF_64BIT   ((1ULL << (64 - 1)) - 1)
1748
1749
1750 /*
1751  * Control parameters used by ext4_htree_next_block
1752  */
1753 #define HASH_NB_ALWAYS          1
1754
1755
1756 /*
1757  * Describe an inode's exact location on disk and in memory
1758  */
1759 struct ext4_iloc
1760 {
1761         struct buffer_head *bh;
1762         unsigned long offset;
1763         ext4_group_t block_group;
1764 };
1765
1766 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1767 {
1768         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1769 }
1770
1771 /*
1772  * This structure is stuffed into the struct file's private_data field
1773  * for directories.  It is where we put information so that we can do
1774  * readdir operations in hash tree order.
1775  */
1776 struct dir_private_info {
1777         struct rb_root  root;
1778         struct rb_node  *curr_node;
1779         struct fname    *extra_fname;
1780         loff_t          last_pos;
1781         __u32           curr_hash;
1782         __u32           curr_minor_hash;
1783         __u32           next_hash;
1784 };
1785
1786 /* calculate the first block number of the group */
1787 static inline ext4_fsblk_t
1788 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1789 {
1790         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1791                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1792 }
1793
1794 /*
1795  * Special error return code only used by dx_probe() and its callers.
1796  */
1797 #define ERR_BAD_DX_DIR  -75000
1798
1799 /*
1800  * Timeout and state flag for lazy initialization inode thread.
1801  */
1802 #define EXT4_DEF_LI_WAIT_MULT                   10
1803 #define EXT4_DEF_LI_MAX_START_DELAY             5
1804 #define EXT4_LAZYINIT_QUIT                      0x0001
1805 #define EXT4_LAZYINIT_RUNNING                   0x0002
1806
1807 /*
1808  * Lazy inode table initialization info
1809  */
1810 struct ext4_lazy_init {
1811         unsigned long           li_state;
1812         struct list_head        li_request_list;
1813         struct mutex            li_list_mtx;
1814 };
1815
1816 struct ext4_li_request {
1817         struct super_block      *lr_super;
1818         struct ext4_sb_info     *lr_sbi;
1819         ext4_group_t            lr_next_group;
1820         struct list_head        lr_request;
1821         unsigned long           lr_next_sched;
1822         unsigned long           lr_timeout;
1823 };
1824
1825 struct ext4_features {
1826         struct kobject f_kobj;
1827         struct completion f_kobj_unregister;
1828 };
1829
1830 /*
1831  * This structure will be used for multiple mount protection. It will be
1832  * written into the block number saved in the s_mmp_block field in the
1833  * superblock. Programs that check MMP should assume that if
1834  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1835  * to use the filesystem, regardless of how old the timestamp is.
1836  */
1837 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
1838 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1839 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
1840 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
1841
1842 struct mmp_struct {
1843         __le32  mmp_magic;              /* Magic number for MMP */
1844         __le32  mmp_seq;                /* Sequence no. updated periodically */
1845
1846         /*
1847          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1848          * purposes and do not affect the correctness of the algorithm
1849          */
1850         __le64  mmp_time;               /* Time last updated */
1851         char    mmp_nodename[64];       /* Node which last updated MMP block */
1852         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
1853
1854         /*
1855          * mmp_check_interval is used to verify if the MMP block has been
1856          * updated on the block device. The value is updated based on the
1857          * maximum time to write the MMP block during an update cycle.
1858          */
1859         __le16  mmp_check_interval;
1860
1861         __le16  mmp_pad1;
1862         __le32  mmp_pad2[226];
1863         __le32  mmp_checksum;           /* crc32c(uuid+mmp_block) */
1864 };
1865
1866 /* arguments passed to the mmp thread */
1867 struct mmpd_data {
1868         struct buffer_head *bh; /* bh from initial read_mmp_block() */
1869         struct super_block *sb;  /* super block of the fs */
1870 };
1871
1872 /*
1873  * Check interval multiplier
1874  * The MMP block is written every update interval and initially checked every
1875  * update interval x the multiplier (the value is then adapted based on the
1876  * write latency). The reason is that writes can be delayed under load and we
1877  * don't want readers to incorrectly assume that the filesystem is no longer
1878  * in use.
1879  */
1880 #define EXT4_MMP_CHECK_MULT             2UL
1881
1882 /*
1883  * Minimum interval for MMP checking in seconds.
1884  */
1885 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
1886
1887 /*
1888  * Maximum interval for MMP checking in seconds.
1889  */
1890 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
1891
1892 /*
1893  * Function prototypes
1894  */
1895
1896 /*
1897  * Ok, these declarations are also in <linux/kernel.h> but none of the
1898  * ext4 source programs needs to include it so they are duplicated here.
1899  */
1900 # define NORET_TYPE     /**/
1901 # define ATTRIB_NORET   __attribute__((noreturn))
1902 # define NORET_AND      noreturn,
1903
1904 /* bitmap.c */
1905 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
1906 void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1907                                 struct ext4_group_desc *gdp,
1908                                 struct buffer_head *bh, int sz);
1909 int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1910                                   struct ext4_group_desc *gdp,
1911                                   struct buffer_head *bh, int sz);
1912 void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1913                                 struct ext4_group_desc *gdp,
1914                                 struct buffer_head *bh);
1915 int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1916                                   struct ext4_group_desc *gdp,
1917                                   struct buffer_head *bh);
1918
1919 /* balloc.c */
1920 extern void ext4_get_group_no_and_offset(struct super_block *sb,
1921                                          ext4_fsblk_t blocknr,
1922                                          ext4_group_t *blockgrpp,
1923                                          ext4_grpblk_t *offsetp);
1924 extern ext4_group_t ext4_get_group_number(struct super_block *sb,
1925                                           ext4_fsblk_t block);
1926
1927 extern void ext4_validate_block_bitmap(struct super_block *sb,
1928                                        struct ext4_group_desc *desc,
1929                                        unsigned int block_group,
1930                                        struct buffer_head *bh);
1931 extern unsigned int ext4_block_group(struct super_block *sb,
1932                         ext4_fsblk_t blocknr);
1933 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1934                         ext4_fsblk_t blocknr);
1935 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1936 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1937                         ext4_group_t group);
1938 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1939                                          ext4_fsblk_t goal,
1940                                          unsigned int flags,
1941                                          unsigned long *count,
1942                                          int *errp);
1943 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
1944                                     s64 nclusters, unsigned int flags);
1945 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
1946 extern void ext4_check_blocks_bitmap(struct super_block *);
1947 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1948                                                     ext4_group_t block_group,
1949                                                     struct buffer_head ** bh);
1950 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1951
1952 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
1953                                                 ext4_group_t block_group);
1954 extern int ext4_wait_block_bitmap(struct super_block *sb,
1955                                   ext4_group_t block_group,
1956                                   struct buffer_head *bh);
1957 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1958                                                   ext4_group_t block_group);
1959 extern void ext4_init_block_bitmap(struct super_block *sb,
1960                                    struct buffer_head *bh,
1961                                    ext4_group_t group,
1962                                    struct ext4_group_desc *desc);
1963 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
1964                                               ext4_group_t block_group,
1965                                               struct ext4_group_desc *gdp);
1966 extern unsigned ext4_num_overhead_clusters(struct super_block *sb,
1967                                            ext4_group_t block_group,
1968                                            struct ext4_group_desc *gdp);
1969 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
1970
1971 /* dir.c */
1972 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1973                                   struct file *,
1974                                   struct ext4_dir_entry_2 *,
1975                                   struct buffer_head *, char *, int,
1976                                   unsigned int);
1977 #define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset)      \
1978         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
1979                                         (de), (bh), (buf), (size), (offset)))
1980 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1981                                     __u32 minor_hash,
1982                                     struct ext4_dir_entry_2 *dirent);
1983 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1984 extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
1985                              struct buffer_head *bh,
1986                              void *buf, int buf_size,
1987                              const char *name, int namelen,
1988                              struct ext4_dir_entry_2 **dest_de);
1989 void ext4_insert_dentry(struct inode *inode,
1990                         struct ext4_dir_entry_2 *de,
1991                         int buf_size,
1992                         const char *name, int namelen);
1993 static inline void ext4_update_dx_flag(struct inode *inode)
1994 {
1995         if (!EXT4_HAS_COMPAT_FEATURE(inode->i_sb,
1996                                      EXT4_FEATURE_COMPAT_DIR_INDEX))
1997                 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
1998 }
1999 static unsigned char ext4_filetype_table[] = {
2000         DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
2001 };
2002
2003 static inline  unsigned char get_dtype(struct super_block *sb, int filetype)
2004 {
2005         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE) ||
2006             (filetype >= EXT4_FT_MAX))
2007                 return DT_UNKNOWN;
2008
2009         return ext4_filetype_table[filetype];
2010 }
2011
2012 /* fsync.c */
2013 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2014 extern int ext4_flush_unwritten_io(struct inode *);
2015
2016 /* hash.c */
2017 extern int ext4fs_dirhash(const char *name, int len, struct
2018                           dx_hash_info *hinfo);
2019
2020 /* ialloc.c */
2021 extern struct inode *__ext4_new_inode(handle_t *, struct inode *, umode_t,
2022                                       const struct qstr *qstr, __u32 goal,
2023                                       uid_t *owner, int handle_type,
2024                                       unsigned int line_no, int nblocks);
2025
2026 #define ext4_new_inode(handle, dir, mode, qstr, goal, owner) \
2027         __ext4_new_inode((handle), (dir), (mode), (qstr), (goal), (owner), \
2028                          0, 0, 0)
2029 #define ext4_new_inode_start_handle(dir, mode, qstr, goal, owner, \
2030                                     type, nblocks)                  \
2031         __ext4_new_inode(NULL, (dir), (mode), (qstr), (goal), (owner), \
2032                          (type), __LINE__, (nblocks))
2033
2034
2035 extern void ext4_free_inode(handle_t *, struct inode *);
2036 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2037 extern unsigned long ext4_count_free_inodes(struct super_block *);
2038 extern unsigned long ext4_count_dirs(struct super_block *);
2039 extern void ext4_check_inodes_bitmap(struct super_block *);
2040 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2041 extern int ext4_init_inode_table(struct super_block *sb,
2042                                  ext4_group_t group, int barrier);
2043 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2044
2045 /* mballoc.c */
2046 extern long ext4_mb_stats;
2047 extern long ext4_mb_max_to_scan;
2048 extern int ext4_mb_init(struct super_block *);
2049 extern int ext4_mb_release(struct super_block *);
2050 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2051                                 struct ext4_allocation_request *, int *);
2052 extern int ext4_mb_reserve_blocks(struct super_block *, int);
2053 extern void ext4_discard_preallocations(struct inode *);
2054 extern int __init ext4_init_mballoc(void);
2055 extern void ext4_exit_mballoc(void);
2056 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2057                              struct buffer_head *bh, ext4_fsblk_t block,
2058                              unsigned long count, int flags);
2059 extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2060                                    ext4_group_t ngroups);
2061 extern int ext4_mb_add_groupinfo(struct super_block *sb,
2062                 ext4_group_t i, struct ext4_group_desc *desc);
2063 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2064                                 ext4_fsblk_t block, unsigned long count);
2065 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *,
2066                                 unsigned long blkdev_flags);
2067
2068 /* inode.c */
2069 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
2070                                                 ext4_lblk_t, int, int *);
2071 struct buffer_head *ext4_bread(handle_t *, struct inode *,
2072                                                 ext4_lblk_t, int, int *);
2073 int ext4_get_block_write(struct inode *inode, sector_t iblock,
2074                          struct buffer_head *bh_result, int create);
2075 int ext4_get_block(struct inode *inode, sector_t iblock,
2076                                 struct buffer_head *bh_result, int create);
2077 int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2078                            struct buffer_head *bh, int create);
2079 int ext4_walk_page_buffers(handle_t *handle,
2080                            struct buffer_head *head,
2081                            unsigned from,
2082                            unsigned to,
2083                            int *partial,
2084                            int (*fn)(handle_t *handle,
2085                                      struct buffer_head *bh));
2086 int do_journal_get_write_access(handle_t *handle,
2087                                 struct buffer_head *bh);
2088 #define FALL_BACK_TO_NONDELALLOC 1
2089 #define CONVERT_INLINE_DATA      2
2090
2091 extern struct inode *ext4_iget(struct super_block *, unsigned long);
2092 extern struct inode *ext4_iget_normal(struct super_block *, unsigned long);
2093 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
2094 extern int  ext4_setattr(struct dentry *, struct iattr *);
2095 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
2096                                 struct kstat *stat);
2097 extern void ext4_evict_inode(struct inode *);
2098 extern void ext4_clear_inode(struct inode *);
2099 extern int  ext4_sync_inode(handle_t *, struct inode *);
2100 extern void ext4_dirty_inode(struct inode *, int);
2101 extern int ext4_change_inode_journal_flag(struct inode *, int);
2102 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
2103 extern int ext4_can_truncate(struct inode *inode);
2104 extern void ext4_truncate(struct inode *);
2105 extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length);
2106 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
2107 extern void ext4_set_inode_flags(struct inode *);
2108 extern void ext4_get_inode_flags(struct ext4_inode_info *);
2109 extern int ext4_alloc_da_blocks(struct inode *inode);
2110 extern void ext4_set_aops(struct inode *inode);
2111 extern int ext4_writepage_trans_blocks(struct inode *);
2112 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
2113 extern int ext4_discard_partial_page_buffers(handle_t *handle,
2114                 struct address_space *mapping, loff_t from,
2115                 loff_t length, int flags);
2116 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2117 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
2118 extern void ext4_da_update_reserve_space(struct inode *inode,
2119                                         int used, int quota_claim);
2120
2121 /* indirect.c */
2122 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
2123                                 struct ext4_map_blocks *map, int flags);
2124 extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
2125                                 const struct iovec *iov, loff_t offset,
2126                                 unsigned long nr_segs);
2127 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
2128 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks, int chunk);
2129 extern void ext4_ind_truncate(handle_t *, struct inode *inode);
2130 extern int ext4_free_hole_blocks(handle_t *handle, struct inode *inode,
2131                                  ext4_lblk_t first, ext4_lblk_t stop);
2132
2133 /* ioctl.c */
2134 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
2135 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
2136
2137 /* migrate.c */
2138 extern int ext4_ext_migrate(struct inode *);
2139 extern int ext4_ind_migrate(struct inode *inode);
2140
2141 /* namei.c */
2142 extern int ext4_dirent_csum_verify(struct inode *inode,
2143                                    struct ext4_dir_entry *dirent);
2144 extern int ext4_orphan_add(handle_t *, struct inode *);
2145 extern int ext4_orphan_del(handle_t *, struct inode *);
2146 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
2147                                 __u32 start_minor_hash, __u32 *next_hash);
2148 extern int search_dir(struct buffer_head *bh,
2149                       char *search_buf,
2150                       int buf_size,
2151                       struct inode *dir,
2152                       const struct qstr *d_name,
2153                       unsigned int offset,
2154                       struct ext4_dir_entry_2 **res_dir);
2155 extern int ext4_generic_delete_entry(handle_t *handle,
2156                                      struct inode *dir,
2157                                      struct ext4_dir_entry_2 *de_del,
2158                                      struct buffer_head *bh,
2159                                      void *entry_buf,
2160                                      int buf_size,
2161                                      int csum_size);
2162
2163 /* resize.c */
2164 extern int ext4_group_add(struct super_block *sb,
2165                                 struct ext4_new_group_data *input);
2166 extern int ext4_group_extend(struct super_block *sb,
2167                                 struct ext4_super_block *es,
2168                                 ext4_fsblk_t n_blocks_count);
2169 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
2170
2171 /* super.c */
2172 extern int ext4_calculate_overhead(struct super_block *sb);
2173 extern int ext4_superblock_csum_verify(struct super_block *sb,
2174                                        struct ext4_super_block *es);
2175 extern void ext4_superblock_csum_set(struct super_block *sb);
2176 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
2177 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
2178 extern void ext4_kvfree(void *ptr);
2179 extern int ext4_alloc_flex_bg_array(struct super_block *sb,
2180                                     ext4_group_t ngroup);
2181 extern const char *ext4_decode_error(struct super_block *sb, int errno,
2182                                      char nbuf[16]);
2183 extern __printf(4, 5)
2184 void __ext4_error(struct super_block *, const char *, unsigned int,
2185                   const char *, ...);
2186 #define ext4_error(sb, message...)      __ext4_error(sb, __func__,      \
2187                                                      __LINE__, ## message)
2188 extern __printf(5, 6)
2189 void ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
2190                       const char *, ...);
2191 extern __printf(5, 6)
2192 void ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
2193                      const char *, ...);
2194 extern void __ext4_std_error(struct super_block *, const char *,
2195                              unsigned int, int);
2196 extern __printf(4, 5)
2197 void __ext4_abort(struct super_block *, const char *, unsigned int,
2198                   const char *, ...);
2199 #define ext4_abort(sb, message...)      __ext4_abort(sb, __func__, \
2200                                                        __LINE__, ## message)
2201 extern __printf(4, 5)
2202 void __ext4_warning(struct super_block *, const char *, unsigned int,
2203                     const char *, ...);
2204 #define ext4_warning(sb, message...)    __ext4_warning(sb, __func__, \
2205                                                        __LINE__, ## message)
2206 extern __printf(3, 4)
2207 void ext4_msg(struct super_block *, const char *, const char *, ...);
2208 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
2209                            const char *, unsigned int, const char *);
2210 #define dump_mmp_msg(sb, mmp, msg)      __dump_mmp_msg(sb, mmp, __func__, \
2211                                                        __LINE__, msg)
2212 extern __printf(7, 8)
2213 void __ext4_grp_locked_error(const char *, unsigned int,
2214                              struct super_block *, ext4_group_t,
2215                              unsigned long, ext4_fsblk_t,
2216                              const char *, ...);
2217 #define ext4_grp_locked_error(sb, grp, message...) \
2218         __ext4_grp_locked_error(__func__, __LINE__, (sb), (grp), ## message)
2219 extern void ext4_update_dynamic_rev(struct super_block *sb);
2220 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
2221                                         __u32 compat);
2222 extern int ext4_update_rocompat_feature(handle_t *handle,
2223                                         struct super_block *sb, __u32 rocompat);
2224 extern int ext4_update_incompat_feature(handle_t *handle,
2225                                         struct super_block *sb, __u32 incompat);
2226 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
2227                                       struct ext4_group_desc *bg);
2228 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
2229                                       struct ext4_group_desc *bg);
2230 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
2231                                      struct ext4_group_desc *bg);
2232 extern __u32 ext4_free_group_clusters(struct super_block *sb,
2233                                       struct ext4_group_desc *bg);
2234 extern __u32 ext4_free_inodes_count(struct super_block *sb,
2235                                  struct ext4_group_desc *bg);
2236 extern __u32 ext4_used_dirs_count(struct super_block *sb,
2237                                 struct ext4_group_desc *bg);
2238 extern __u32 ext4_itable_unused_count(struct super_block *sb,
2239                                    struct ext4_group_desc *bg);
2240 extern void ext4_block_bitmap_set(struct super_block *sb,
2241                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2242 extern void ext4_inode_bitmap_set(struct super_block *sb,
2243                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2244 extern void ext4_inode_table_set(struct super_block *sb,
2245                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
2246 extern void ext4_free_group_clusters_set(struct super_block *sb,
2247                                          struct ext4_group_desc *bg,
2248                                          __u32 count);
2249 extern void ext4_free_inodes_set(struct super_block *sb,
2250                                 struct ext4_group_desc *bg, __u32 count);
2251 extern void ext4_used_dirs_set(struct super_block *sb,
2252                                 struct ext4_group_desc *bg, __u32 count);
2253 extern void ext4_itable_unused_set(struct super_block *sb,
2254                                    struct ext4_group_desc *bg, __u32 count);
2255 extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
2256                                        struct ext4_group_desc *gdp);
2257 extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
2258                                      struct ext4_group_desc *gdp);
2259 extern int ext4_register_li_request(struct super_block *sb,
2260                                     ext4_group_t first_not_zeroed);
2261
2262 static inline int ext4_has_group_desc_csum(struct super_block *sb)
2263 {
2264         return EXT4_HAS_RO_COMPAT_FEATURE(sb,
2265                                           EXT4_FEATURE_RO_COMPAT_GDT_CSUM) ||
2266                (EXT4_SB(sb)->s_chksum_driver != NULL);
2267 }
2268
2269 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2270 {
2271         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2272                 le32_to_cpu(es->s_blocks_count_lo);
2273 }
2274
2275 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2276 {
2277         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2278                 le32_to_cpu(es->s_r_blocks_count_lo);
2279 }
2280
2281 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2282 {
2283         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2284                 le32_to_cpu(es->s_free_blocks_count_lo);
2285 }
2286
2287 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2288                                          ext4_fsblk_t blk)
2289 {
2290         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2291         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2292 }
2293
2294 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2295                                               ext4_fsblk_t blk)
2296 {
2297         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2298         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2299 }
2300
2301 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2302                                            ext4_fsblk_t blk)
2303 {
2304         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2305         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2306 }
2307
2308 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2309 {
2310         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2311                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2312                         le32_to_cpu(raw_inode->i_size_lo);
2313         else
2314                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2315 }
2316
2317 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2318 {
2319         raw_inode->i_size_lo = cpu_to_le32(i_size);
2320         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2321 }
2322
2323 static inline
2324 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2325                                             ext4_group_t group)
2326 {
2327          struct ext4_group_info ***grp_info;
2328          long indexv, indexh;
2329          grp_info = EXT4_SB(sb)->s_group_info;
2330          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2331          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2332          return grp_info[indexv][indexh];
2333 }
2334
2335 /*
2336  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2337  * the locking protocol documented in the comments of ext4_group_add()
2338  * in resize.c
2339  */
2340 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2341 {
2342         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2343
2344         smp_rmb();
2345         return ngroups;
2346 }
2347
2348 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2349                                              ext4_group_t block_group)
2350 {
2351         return block_group >> sbi->s_log_groups_per_flex;
2352 }
2353
2354 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2355 {
2356         return 1 << sbi->s_log_groups_per_flex;
2357 }
2358
2359 #define ext4_std_error(sb, errno)                               \
2360 do {                                                            \
2361         if ((errno))                                            \
2362                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2363 } while (0)
2364
2365 #ifdef CONFIG_SMP
2366 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2367  * counters. So we need to make sure we have free clusters more
2368  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2369  */
2370 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2371 #else
2372 #define EXT4_FREECLUSTERS_WATERMARK 0
2373 #endif
2374
2375 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2376 {
2377         /*
2378          * XXX: replace with spinlock if seen contended -bzzz
2379          */
2380         down_write(&EXT4_I(inode)->i_data_sem);
2381         if (newsize > EXT4_I(inode)->i_disksize)
2382                 EXT4_I(inode)->i_disksize = newsize;
2383         up_write(&EXT4_I(inode)->i_data_sem);
2384         return ;
2385 }
2386
2387 struct ext4_group_info {
2388         unsigned long   bb_state;
2389         struct rb_root  bb_free_root;
2390         ext4_grpblk_t   bb_first_free;  /* first free block */
2391         ext4_grpblk_t   bb_free;        /* total free blocks */
2392         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2393         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2394         struct          list_head bb_prealloc_list;
2395 #ifdef DOUBLE_CHECK
2396         void            *bb_bitmap;
2397 #endif
2398         struct rw_semaphore alloc_sem;
2399         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2400                                          * regions, index is order.
2401                                          * bb_counters[3] = 5 means
2402                                          * 5 free 8-block regions. */
2403 };
2404
2405 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2406 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2407
2408 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2409         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2410
2411 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2412         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2413 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2414         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2415 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2416         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2417
2418 #define EXT4_MAX_CONTENTION             8
2419 #define EXT4_CONTENTION_THRESHOLD       2
2420
2421 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2422                                               ext4_group_t group)
2423 {
2424         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2425 }
2426
2427 /*
2428  * Returns true if the filesystem is busy enough that attempts to
2429  * access the block group locks has run into contention.
2430  */
2431 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2432 {
2433         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2434 }
2435
2436 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2437 {
2438         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2439         if (spin_trylock(lock))
2440                 /*
2441                  * We're able to grab the lock right away, so drop the
2442                  * lock contention counter.
2443                  */
2444                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2445         else {
2446                 /*
2447                  * The lock is busy, so bump the contention counter,
2448                  * and then wait on the spin lock.
2449                  */
2450                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2451                                   EXT4_MAX_CONTENTION);
2452                 spin_lock(lock);
2453         }
2454 }
2455
2456 static inline void ext4_unlock_group(struct super_block *sb,
2457                                         ext4_group_t group)
2458 {
2459         spin_unlock(ext4_group_lock_ptr(sb, group));
2460 }
2461
2462 /*
2463  * Block validity checking
2464  */
2465 #define ext4_check_indirect_blockref(inode, bh)                         \
2466         ext4_check_blockref(__func__, __LINE__, inode,                  \
2467                             (__le32 *)(bh)->b_data,                     \
2468                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2469
2470 #define ext4_ind_check_inode(inode)                                     \
2471         ext4_check_blockref(__func__, __LINE__, inode,                  \
2472                             EXT4_I(inode)->i_data,                      \
2473                             EXT4_NDIR_BLOCKS)
2474
2475 /*
2476  * Inodes and files operations
2477  */
2478
2479 /* dir.c */
2480 extern const struct file_operations ext4_dir_operations;
2481
2482 /* file.c */
2483 extern const struct inode_operations ext4_file_inode_operations;
2484 extern const struct file_operations ext4_file_operations;
2485 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2486 extern void ext4_unwritten_wait(struct inode *inode);
2487
2488 /* inline.c */
2489 extern int ext4_has_inline_data(struct inode *inode);
2490 extern int ext4_get_inline_size(struct inode *inode);
2491 extern int ext4_get_max_inline_size(struct inode *inode);
2492 extern int ext4_find_inline_data_nolock(struct inode *inode);
2493 extern void ext4_write_inline_data(struct inode *inode,
2494                                    struct ext4_iloc *iloc,
2495                                    void *buffer, loff_t pos,
2496                                    unsigned int len);
2497 extern int ext4_prepare_inline_data(handle_t *handle, struct inode *inode,
2498                                     unsigned int len);
2499 extern int ext4_init_inline_data(handle_t *handle, struct inode *inode,
2500                                  unsigned int len);
2501 extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
2502
2503 extern int ext4_readpage_inline(struct inode *inode, struct page *page);
2504 extern int ext4_try_to_write_inline_data(struct address_space *mapping,
2505                                          struct inode *inode,
2506                                          loff_t pos, unsigned len,
2507                                          unsigned flags,
2508                                          struct page **pagep);
2509 extern int ext4_write_inline_data_end(struct inode *inode,
2510                                       loff_t pos, unsigned len,
2511                                       unsigned copied,
2512                                       struct page *page);
2513 extern struct buffer_head *
2514 ext4_journalled_write_inline_data(struct inode *inode,
2515                                   unsigned len,
2516                                   struct page *page);
2517 extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
2518                                            struct inode *inode,
2519                                            loff_t pos, unsigned len,
2520                                            unsigned flags,
2521                                            struct page **pagep,
2522                                            void **fsdata);
2523 extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
2524                                          unsigned len, unsigned copied,
2525                                          struct page *page);
2526 extern int ext4_try_add_inline_entry(handle_t *handle, struct dentry *dentry,
2527                                      struct inode *inode);
2528 extern int ext4_try_create_inline_dir(handle_t *handle,
2529                                       struct inode *parent,
2530                                       struct inode *inode);
2531 extern int ext4_read_inline_dir(struct file *filp,
2532                                 void *dirent, filldir_t filldir,
2533                                 int *has_inline_data);
2534 extern int htree_inlinedir_to_tree(struct file *dir_file,
2535                                    struct inode *dir, ext4_lblk_t block,
2536                                    struct dx_hash_info *hinfo,
2537                                    __u32 start_hash, __u32 start_minor_hash,
2538                                    int *has_inline_data);
2539 extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
2540                                         const struct qstr *d_name,
2541                                         struct ext4_dir_entry_2 **res_dir,
2542                                         int *has_inline_data);
2543 extern int ext4_delete_inline_entry(handle_t *handle,
2544                                     struct inode *dir,
2545                                     struct ext4_dir_entry_2 *de_del,
2546                                     struct buffer_head *bh,
2547                                     int *has_inline_data);
2548 extern int empty_inline_dir(struct inode *dir, int *has_inline_data);
2549 extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
2550                                         struct ext4_dir_entry_2 **parent_de,
2551                                         int *retval);
2552 extern int ext4_inline_data_fiemap(struct inode *inode,
2553                                    struct fiemap_extent_info *fieinfo,
2554                                    int *has_inline);
2555 extern int ext4_try_to_evict_inline_data(handle_t *handle,
2556                                          struct inode *inode,
2557                                          int needed);
2558 extern void ext4_inline_data_truncate(struct inode *inode, int *has_inline);
2559
2560 extern int ext4_convert_inline_data(struct inode *inode);
2561
2562 /* namei.c */
2563 extern const struct inode_operations ext4_dir_inode_operations;
2564 extern const struct inode_operations ext4_special_inode_operations;
2565 extern struct dentry *ext4_get_parent(struct dentry *child);
2566 extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
2567                                  struct ext4_dir_entry_2 *de,
2568                                  int blocksize, int csum_size,
2569                                  unsigned int parent_ino, int dotdot_real_len);
2570 extern void initialize_dirent_tail(struct ext4_dir_entry_tail *t,
2571                                    unsigned int blocksize);
2572 extern int ext4_handle_dirty_dirent_node(handle_t *handle,
2573                                          struct inode *inode,
2574                                          struct buffer_head *bh);
2575 #define S_SHIFT 12
2576 static unsigned char ext4_type_by_mode[S_IFMT >> S_SHIFT] = {
2577         [S_IFREG >> S_SHIFT]    = EXT4_FT_REG_FILE,
2578         [S_IFDIR >> S_SHIFT]    = EXT4_FT_DIR,
2579         [S_IFCHR >> S_SHIFT]    = EXT4_FT_CHRDEV,
2580         [S_IFBLK >> S_SHIFT]    = EXT4_FT_BLKDEV,
2581         [S_IFIFO >> S_SHIFT]    = EXT4_FT_FIFO,
2582         [S_IFSOCK >> S_SHIFT]   = EXT4_FT_SOCK,
2583         [S_IFLNK >> S_SHIFT]    = EXT4_FT_SYMLINK,
2584 };
2585
2586 static inline void ext4_set_de_type(struct super_block *sb,
2587                                 struct ext4_dir_entry_2 *de,
2588                                 umode_t mode) {
2589         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE))
2590                 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
2591 }
2592
2593
2594 /* symlink.c */
2595 extern const struct inode_operations ext4_symlink_inode_operations;
2596 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2597
2598 /* block_validity */
2599 extern void ext4_release_system_zone(struct super_block *sb);
2600 extern int ext4_setup_system_zone(struct super_block *sb);
2601 extern int __init ext4_init_system_zone(void);
2602 extern void ext4_exit_system_zone(void);
2603 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2604                                  ext4_fsblk_t start_blk,
2605                                  unsigned int count);
2606 extern int ext4_check_blockref(const char *, unsigned int,
2607                                struct inode *, __le32 *, unsigned int);
2608
2609 /* extents.c */
2610 struct ext4_ext_path;
2611 struct ext4_extent;
2612
2613 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2614 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2615 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
2616                                        int chunk);
2617 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2618                                struct ext4_map_blocks *map, int flags);
2619 extern void ext4_ext_truncate(handle_t *, struct inode *);
2620 extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
2621                                  ext4_lblk_t end);
2622 extern void ext4_ext_init(struct super_block *);
2623 extern void ext4_ext_release(struct super_block *);
2624 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2625                           loff_t len);
2626 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
2627                           ssize_t len);
2628 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2629                            struct ext4_map_blocks *map, int flags);
2630 extern int ext4_ext_calc_metadata_amount(struct inode *inode,
2631                                          ext4_lblk_t lblocks);
2632 extern int ext4_extent_tree_init(handle_t *, struct inode *);
2633 extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
2634                                                    int num,
2635                                                    struct ext4_ext_path *path);
2636 extern int ext4_can_extents_be_merged(struct inode *inode,
2637                                       struct ext4_extent *ex1,
2638                                       struct ext4_extent *ex2);
2639 extern int ext4_ext_insert_extent(handle_t *, struct inode *,
2640                                   struct ext4_ext_path *,
2641                                   struct ext4_extent *, int);
2642 extern struct ext4_ext_path *ext4_ext_find_extent(struct inode *, ext4_lblk_t,
2643                                                   struct ext4_ext_path *);
2644 extern void ext4_ext_drop_refs(struct ext4_ext_path *);
2645 extern int ext4_ext_check_inode(struct inode *inode);
2646 extern int ext4_find_delalloc_range(struct inode *inode,
2647                                     ext4_lblk_t lblk_start,
2648                                     ext4_lblk_t lblk_end);
2649 extern int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk);
2650 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2651                         __u64 start, __u64 len);
2652
2653
2654 /* move_extent.c */
2655 extern void ext4_double_down_write_data_sem(struct inode *first,
2656                                             struct inode *second);
2657 extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
2658                                           struct inode *donor_inode);
2659 void ext4_inode_double_lock(struct inode *inode1, struct inode *inode2);
2660 void ext4_inode_double_unlock(struct inode *inode1, struct inode *inode2);
2661 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2662                              __u64 start_orig, __u64 start_donor,
2663                              __u64 len, __u64 *moved_len);
2664
2665 /* page-io.c */
2666 extern int __init ext4_init_pageio(void);
2667 extern void ext4_add_complete_io(ext4_io_end_t *io_end);
2668 extern void ext4_exit_pageio(void);
2669 extern void ext4_ioend_shutdown(struct inode *);
2670 extern void ext4_free_io_end(ext4_io_end_t *io);
2671 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2672 extern void ext4_end_io_work(struct work_struct *work);
2673 extern void ext4_io_submit(struct ext4_io_submit *io);
2674 extern int ext4_bio_write_page(struct ext4_io_submit *io,
2675                                struct page *page,
2676                                int len,
2677                                struct writeback_control *wbc);
2678
2679 /* mmp.c */
2680 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2681 extern void ext4_mmp_csum_set(struct super_block *sb, struct mmp_struct *mmp);
2682 extern int ext4_mmp_csum_verify(struct super_block *sb,
2683                                 struct mmp_struct *mmp);
2684
2685 /* BH_Uninit flag: blocks are allocated but uninitialized on disk */
2686 enum ext4_state_bits {
2687         BH_Uninit       /* blocks are allocated but uninitialized on disk */
2688           = BH_JBDPrivateStart,
2689         BH_AllocFromCluster,    /* allocated blocks were part of already
2690                                  * allocated cluster. Note that this flag will
2691                                  * never, ever appear in a buffer_head's state
2692                                  * flag. See EXT4_MAP_FROM_CLUSTER to see where
2693                                  * this is used. */
2694 };
2695
2696 BUFFER_FNS(Uninit, uninit)
2697 TAS_BUFFER_FNS(Uninit, uninit)
2698
2699 /*
2700  * Add new method to test whether block and inode bitmaps are properly
2701  * initialized. With uninit_bg reading the block from disk is not enough
2702  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2703  */
2704 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2705
2706 static inline int bitmap_uptodate(struct buffer_head *bh)
2707 {
2708         return (buffer_uptodate(bh) &&
2709                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2710 }
2711 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2712 {
2713         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2714 }
2715
2716 /*
2717  * Disable DIO read nolock optimization, so new dioreaders will be forced
2718  * to grab i_mutex
2719  */
2720 static inline void ext4_inode_block_unlocked_dio(struct inode *inode)
2721 {
2722         ext4_set_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2723         smp_mb();
2724 }
2725 static inline void ext4_inode_resume_unlocked_dio(struct inode *inode)
2726 {
2727         smp_mb();
2728         ext4_clear_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2729 }
2730
2731 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2732
2733 /* For ioend & aio unwritten conversion wait queues */
2734 #define EXT4_WQ_HASH_SZ         37
2735 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
2736                                             EXT4_WQ_HASH_SZ])
2737 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
2738                                              EXT4_WQ_HASH_SZ])
2739 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2740 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2741
2742 #define EXT4_RESIZING   0
2743 extern int ext4_resize_begin(struct super_block *sb);
2744 extern void ext4_resize_end(struct super_block *sb);
2745
2746 #endif  /* __KERNEL__ */
2747
2748 #endif  /* _EXT4_H */