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 <linux/ratelimit.h>
  33 #include <crypto/hash.h>
  34 #include <linux/falloc.h>
  35 #ifdef __KERNEL__
  36 #include <linux/compat.h>
  37 #endif
  38
  39 /*
  40  * The fourth extended filesystem constants/structures
  41  */
  42
  43 /*
  44  * Define EXT4FS_DEBUG to produce debug messages
  45  */
  46 #undef EXT4FS_DEBUG
  47
  48 /*
  49  * Debug code
  50  */
  51 #ifdef EXT4FS_DEBUG
  52 #define ext4_debug(f, a...)                                             \
  53         do {                                                            \
  54                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
  55                         __FILE__, __LINE__, __func__);                  \
  56                 printk(KERN_DEBUG f, ## a);                             \
  57         } while (0)
  58 #else
  59 #define ext4_debug(fmt, ...)    no_printk(fmt, ##__VA_ARGS__)
  60 #endif
  61
  62 /*
  63  * Turn on EXT_DEBUG to get lots of info about extents operations.
  64  */
  65 #define EXT_DEBUG__
  66 #ifdef EXT_DEBUG
  67 #define ext_debug(fmt, ...)     printk(fmt, ##__VA_ARGS__)
  68 #else
  69 #define ext_debug(fmt, ...)     no_printk(fmt, ##__VA_ARGS__)
  70 #endif
  71
  72 #define EXT4_ERROR_INODE(inode, fmt, a...) \
  73         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
  74
  75 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...)                 \
  76         ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
  77
  78 #define EXT4_ERROR_FILE(file, block, fmt, a...)                         \
  79         ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
  80
  81 /* data type for block offset of block group */
  82 typedef int ext4_grpblk_t;
  83
  84 /* data type for filesystem-wide blocks number */
  85 typedef unsigned long long ext4_fsblk_t;
  86
  87 /* data type for file logical block number */
  88 typedef __u32 ext4_lblk_t;
  89
  90 /* data type for block group number */
  91 typedef unsigned int ext4_group_t;
  92
  93 /*
  94  * Flags used in mballoc's allocation_context flags field.
  95  *
  96  * Also used to show what's going on for debugging purposes when the
  97  * flag field is exported via the traceport interface
  98  */
  99
 100 /* prefer goal again. length */
 101 #define EXT4_MB_HINT_MERGE              0x0001
 102 /* blocks already reserved */
 103 #define EXT4_MB_HINT_RESERVED           0x0002
 104 /* metadata is being allocated */
 105 #define EXT4_MB_HINT_METADATA           0x0004
 106 /* first blocks in the file */
 107 #define EXT4_MB_HINT_FIRST              0x0008
 108 /* search for the best chunk */
 109 #define EXT4_MB_HINT_BEST               0x0010
 110 /* data is being allocated */
 111 #define EXT4_MB_HINT_DATA               0x0020
 112 /* don't preallocate (for tails) */
 113 #define EXT4_MB_HINT_NOPREALLOC         0x0040
 114 /* allocate for locality group */
 115 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
 116 /* allocate goal blocks or none */
 117 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
 118 /* goal is meaningful */
 119 #define EXT4_MB_HINT_TRY_GOAL           0x0200
 120 /* blocks already pre-reserved by delayed allocation */
 121 #define EXT4_MB_DELALLOC_RESERVED       0x0400
 122 /* We are doing stream allocation */
 123 #define EXT4_MB_STREAM_ALLOC            0x0800
 124 /* Use reserved root blocks if needed */
 125 #define EXT4_MB_USE_ROOT_BLOCKS         0x1000
 126 /* Use blocks from reserved pool */
 127 #define EXT4_MB_USE_RESERVED            0x2000
 128
 129 struct ext4_allocation_request {
 130         /* target inode for block we're allocating */
 131         struct inode *inode;
 132         /* how many blocks we want to allocate */
 133         unsigned int len;
 134         /* logical block in target inode */
 135         ext4_lblk_t logical;
 136         /* the closest logical allocated block to the left */
 137         ext4_lblk_t lleft;
 138         /* the closest logical allocated block to the right */
 139         ext4_lblk_t lright;
 140         /* phys. target (a hint) */
 141         ext4_fsblk_t goal;
 142         /* phys. block for the closest logical allocated block to the left */
 143         ext4_fsblk_t pleft;
 144         /* phys. block for the closest logical allocated block to the right */
 145         ext4_fsblk_t pright;
 146         /* flags. see above EXT4_MB_HINT_* */
 147         unsigned int flags;
 148 };
 149
 150 /*
 151  * Logical to physical block mapping, used by ext4_map_blocks()
 152  *
 153  * This structure is used to pass requests into ext4_map_blocks() as
 154  * well as to store the information returned by ext4_map_blocks().  It
 155  * takes less room on the stack than a struct buffer_head.
 156  */
 157 #define EXT4_MAP_NEW            (1 << BH_New)
 158 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
 159 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
 160 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
 161 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
 162                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY)
 163
 164 struct ext4_map_blocks {
 165         ext4_fsblk_t m_pblk;
 166         ext4_lblk_t m_lblk;
 167         unsigned int m_len;
 168         unsigned int m_flags;
 169 };
 170
 171 /*
 172  * Flags for ext4_io_end->flags
 173  */
 174 #define EXT4_IO_END_UNWRITTEN   0x0001
 175
 176 /*
 177  * For converting unwritten extents on a work queue. 'handle' is used for
 178  * buffered writeback.
 179  */
 180 typedef struct ext4_io_end {
 181         struct list_head        list;           /* per-file finished IO list */
 182         handle_t                *handle;        /* handle reserved for extent
 183                                                  * conversion */
 184         struct inode            *inode;         /* file being written to */
 185         struct bio              *bio;           /* Linked list of completed
 186                                                  * bios covering the extent */
 187         unsigned int            flag;           /* unwritten or not */
 188         loff_t                  offset;         /* offset in the file */
 189         ssize_t                 size;           /* size of the extent */
 190         atomic_t                count;          /* reference counter */
 191 } ext4_io_end_t;
 192
 193 struct ext4_io_submit {
 194         int                     io_op;
 195         struct bio              *io_bio;
 196         ext4_io_end_t           *io_end;
 197         sector_t                io_next_block;
 198 };
 199
 200 /*
 201  * Special inodes numbers
 202  */
 203 #define EXT4_BAD_INO             1      /* Bad blocks inode */
 204 #define EXT4_ROOT_INO            2      /* Root inode */
 205 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
 206 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
 207 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
 208 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
 209 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
 210 #define EXT4_JOURNAL_INO         8      /* Journal inode */
 211
 212 /* First non-reserved inode for old ext4 filesystems */
 213 #define EXT4_GOOD_OLD_FIRST_INO 11
 214
 215 /*
 216  * Maximal count of links to a file
 217  */
 218 #define EXT4_LINK_MAX           65000
 219
 220 /*
 221  * Macro-instructions used to manage several block sizes
 222  */
 223 #define EXT4_MIN_BLOCK_SIZE             1024
 224 #define EXT4_MAX_BLOCK_SIZE             65536
 225 #define EXT4_MIN_BLOCK_LOG_SIZE         10
 226 #define EXT4_MAX_BLOCK_LOG_SIZE         16
 227 #ifdef __KERNEL__
 228 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
 229 #else
 230 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
 231 #endif
 232 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
 233 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
 234                                          EXT4_SB(s)->s_cluster_bits)
 235 #ifdef __KERNEL__
 236 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
 237 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
 238 #else
 239 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
 240 #endif
 241 #ifdef __KERNEL__
 242 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
 243 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
 244 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
 245 #else
 246 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 247                                  EXT4_GOOD_OLD_INODE_SIZE : \
 248                                  (s)->s_inode_size)
 249 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 250                                  EXT4_GOOD_OLD_FIRST_INO : \
 251                                  (s)->s_first_ino)
 252 #endif
 253 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
 254
 255 /* Translate a block number to a cluster number */
 256 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
 257 /* Translate a cluster number to a block number */
 258 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
 259 /* Translate # of blks to # of clusters */
 260 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
 261                                  (sbi)->s_cluster_bits)
 262 /* Mask out the low bits to get the starting block of the cluster */
 263 #define EXT4_PBLK_CMASK(s, pblk) ((pblk) &                              \
 264                                   ~((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
 265 #define EXT4_LBLK_CMASK(s, lblk) ((lblk) &                              \
 266                                   ~((ext4_lblk_t) (s)->s_cluster_ratio - 1))
 267 /* Get the cluster offset */
 268 #define EXT4_PBLK_COFF(s, pblk) ((pblk) &                               \
 269                                  ((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
 270 #define EXT4_LBLK_COFF(s, lblk) ((lblk) &                               \
 271                                  ((ext4_lblk_t) (s)->s_cluster_ratio - 1))
 272
 273 /*
 274  * Structure of a blocks group descriptor
 275  */
 276 struct ext4_group_desc
 277 {
 278         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
 279         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
 280         __le32  bg_inode_table_lo;      /* Inodes table block */
 281         __le16  bg_free_blocks_count_lo;/* Free blocks count */
 282         __le16  bg_free_inodes_count_lo;/* Free inodes count */
 283         __le16  bg_used_dirs_count_lo;  /* Directories count */
 284         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
 285         __le32  bg_exclude_bitmap_lo;   /* Exclude bitmap for snapshots */
 286         __le16  bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
 287         __le16  bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
 288         __le16  bg_itable_unused_lo;    /* Unused inodes count */
 289         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
 290         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
 291         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
 292         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
 293         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
 294         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
 295         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
 296         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
 297         __le32  bg_exclude_bitmap_hi;   /* Exclude bitmap block MSB */
 298         __le16  bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
 299         __le16  bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
 300         __u32   bg_reserved;
 301 };
 302
 303 #define EXT4_BG_INODE_BITMAP_CSUM_HI_END        \
 304         (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
 305          sizeof(__le16))
 306 #define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END        \
 307         (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
 308          sizeof(__le16))
 309
 310 /*
 311  * Structure of a flex block group info
 312  */
 313
 314 struct flex_groups {
 315         atomic64_t      free_clusters;
 316         atomic_t        free_inodes;
 317         atomic_t        used_dirs;
 318 };
 319
 320 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
 321 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
 322 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
 323
 324 /*
 325  * Macro-instructions used to manage group descriptors
 326  */
 327 #define EXT4_MIN_DESC_SIZE              32
 328 #define EXT4_MIN_DESC_SIZE_64BIT        64
 329 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
 330 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
 331 #ifdef __KERNEL__
 332 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
 333 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
 334 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
 335 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
 336 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
 337 #else
 338 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
 339 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
 340 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
 341 #endif
 342
 343 /*
 344  * Constants relative to the data blocks
 345  */
 346 #define EXT4_NDIR_BLOCKS                12
 347 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
 348 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
 349 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
 350 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
 351
 352 /*
 353  * Inode flags
 354  */
 355 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
 356 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
 357 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
 358 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
 359 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
 360 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
 361 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
 362 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
 363 /* Reserved for compression usage... */
 364 #define EXT4_DIRTY_FL                   0x00000100
 365 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
 366 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
 367         /* nb: was previously EXT2_ECOMPR_FL */
 368 #define EXT4_ENCRYPT_FL                 0x00000800 /* encrypted file */
 369 /* End compression flags --- maybe not all used */
 370 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
 371 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
 372 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
 373 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
 374 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
 375 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
 376 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
 377 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
 378 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
 379 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
 380 #define EXT4_INLINE_DATA_FL             0x10000000 /* Inode has inline data. */
 381 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
 382
 383 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
 384 #define EXT4_FL_USER_MODIFIABLE         0x004380FF /* User modifiable flags */
 385
 386 /* Flags that should be inherited by new inodes from their parent. */
 387 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
 388                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
 389                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
 390                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
 391
 392 /* Flags that are appropriate for regular files (all but dir-specific ones). */
 393 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
 394
 395 /* Flags that are appropriate for non-directories/regular files. */
 396 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
 397
 398 /* Mask out flags that are inappropriate for the given type of inode. */
 399 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
 400 {
 401         if (S_ISDIR(mode))
 402                 return flags;
 403         else if (S_ISREG(mode))
 404                 return flags & EXT4_REG_FLMASK;
 405         else
 406                 return flags & EXT4_OTHER_FLMASK;
 407 }
 408
 409 /*
 410  * Inode flags used for atomic set/get
 411  */
 412 enum {
 413         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
 414         EXT4_INODE_UNRM         = 1,    /* Undelete */
 415         EXT4_INODE_COMPR        = 2,    /* Compress file */
 416         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
 417         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
 418         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
 419         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
 420         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
 421 /* Reserved for compression usage... */
 422         EXT4_INODE_DIRTY        = 8,
 423         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
 424         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
 425         EXT4_INODE_ENCRYPT      = 11,   /* Encrypted file */
 426 /* End compression flags --- maybe not all used */
 427         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
 428         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
 429         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
 430         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
 431         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
 432         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
 433         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
 434         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
 435         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
 436         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
 437         EXT4_INODE_INLINE_DATA  = 28,   /* Data in inode. */
 438         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
 439 };
 440
 441 /*
 442  * Since it's pretty easy to mix up bit numbers and hex values, we use a
 443  * build-time check to make sure that EXT4_XXX_FL is consistent with respect to
 444  * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost
 445  * any extra space in the compiled kernel image, otherwise, the build will fail.
 446  * It's important that these values are the same, since we are using
 447  * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent
 448  * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk
 449  * values found in ext2, ext3 and ext4 filesystems, and of course the values
 450  * defined in e2fsprogs.
 451  *
 452  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
 453  */
 454 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
 455 #define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG))
 456
 457 static inline void ext4_check_flag_values(void)
 458 {
 459         CHECK_FLAG_VALUE(SECRM);
 460         CHECK_FLAG_VALUE(UNRM);
 461         CHECK_FLAG_VALUE(COMPR);
 462         CHECK_FLAG_VALUE(SYNC);
 463         CHECK_FLAG_VALUE(IMMUTABLE);
 464         CHECK_FLAG_VALUE(APPEND);
 465         CHECK_FLAG_VALUE(NODUMP);
 466         CHECK_FLAG_VALUE(NOATIME);
 467         CHECK_FLAG_VALUE(DIRTY);
 468         CHECK_FLAG_VALUE(COMPRBLK);
 469         CHECK_FLAG_VALUE(NOCOMPR);
 470         CHECK_FLAG_VALUE(ENCRYPT);
 471         CHECK_FLAG_VALUE(INDEX);
 472         CHECK_FLAG_VALUE(IMAGIC);
 473         CHECK_FLAG_VALUE(JOURNAL_DATA);
 474         CHECK_FLAG_VALUE(NOTAIL);
 475         CHECK_FLAG_VALUE(DIRSYNC);
 476         CHECK_FLAG_VALUE(TOPDIR);
 477         CHECK_FLAG_VALUE(HUGE_FILE);
 478         CHECK_FLAG_VALUE(EXTENTS);
 479         CHECK_FLAG_VALUE(EA_INODE);
 480         CHECK_FLAG_VALUE(EOFBLOCKS);
 481         CHECK_FLAG_VALUE(INLINE_DATA);
 482         CHECK_FLAG_VALUE(RESERVED);
 483 }
 484
 485 /* Used to pass group descriptor data when online resize is done */
 486 struct ext4_new_group_input {
 487         __u32 group;            /* Group number for this data */
 488         __u64 block_bitmap;     /* Absolute block number of block bitmap */
 489         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
 490         __u64 inode_table;      /* Absolute block number of inode table start */
 491         __u32 blocks_count;     /* Total number of blocks in this group */
 492         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
 493         __u16 unused;
 494 };
 495
 496 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 497 struct compat_ext4_new_group_input {
 498         u32 group;
 499         compat_u64 block_bitmap;
 500         compat_u64 inode_bitmap;
 501         compat_u64 inode_table;
 502         u32 blocks_count;
 503         u16 reserved_blocks;
 504         u16 unused;
 505 };
 506 #endif
 507
 508 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
 509 struct ext4_new_group_data {
 510         __u32 group;
 511         __u64 block_bitmap;
 512         __u64 inode_bitmap;
 513         __u64 inode_table;
 514         __u32 blocks_count;
 515         __u16 reserved_blocks;
 516         __u16 unused;
 517         __u32 free_blocks_count;
 518 };
 519
 520 /* Indexes used to index group tables in ext4_new_group_data */
 521 enum {
 522         BLOCK_BITMAP = 0,       /* block bitmap */
 523         INODE_BITMAP,           /* inode bitmap */
 524         INODE_TABLE,            /* inode tables */
 525         GROUP_TABLE_COUNT,
 526 };
 527
 528 /*
 529  * Flags used by ext4_map_blocks()
 530  */
 531         /* Allocate any needed blocks and/or convert an unwritten
 532            extent to be an initialized ext4 */
 533 #define EXT4_GET_BLOCKS_CREATE                  0x0001
 534         /* Request the creation of an unwritten extent */
 535 #define EXT4_GET_BLOCKS_UNWRIT_EXT              0x0002
 536 #define EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT       (EXT4_GET_BLOCKS_UNWRIT_EXT|\
 537                                                  EXT4_GET_BLOCKS_CREATE)
 538         /* Caller is from the delayed allocation writeout path
 539          * finally doing the actual allocation of delayed blocks */
 540 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
 541         /* caller is from the direct IO path, request to creation of an
 542         unwritten extents if not allocated, split the unwritten
 543         extent if blocks has been preallocated already*/
 544 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
 545 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
 546 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
 547                                          EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
 548         /* Convert extent to initialized after IO complete */
 549 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
 550                                          EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
 551         /* Eventual metadata allocation (due to growing extent tree)
 552          * should not fail, so try to use reserved blocks for that.*/
 553 #define EXT4_GET_BLOCKS_METADATA_NOFAIL         0x0020
 554         /* Don't normalize allocation size (used for fallocate) */
 555 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
 556         /* Request will not result in inode size update (user for fallocate) */
 557 #define EXT4_GET_BLOCKS_KEEP_SIZE               0x0080
 558         /* Do not take i_data_sem locking in ext4_map_blocks */
 559 #define EXT4_GET_BLOCKS_NO_LOCK                 0x0100
 560         /* Convert written extents to unwritten */
 561 #define EXT4_GET_BLOCKS_CONVERT_UNWRITTEN       0x0200
 562
 563 /*
 564  * The bit position of these flags must not overlap with any of the
 565  * EXT4_GET_BLOCKS_*.  They are used by ext4_find_extent(),
 566  * read_extent_tree_block(), ext4_split_extent_at(),
 567  * ext4_ext_insert_extent(), and ext4_ext_create_new_leaf().
 568  * EXT4_EX_NOCACHE is used to indicate that the we shouldn't be
 569  * caching the extents when reading from the extent tree while a
 570  * truncate or punch hole operation is in progress.
 571  */
 572 #define EXT4_EX_NOCACHE                         0x40000000
 573 #define EXT4_EX_FORCE_CACHE                     0x20000000
 574
 575 /*
 576  * Flags used by ext4_free_blocks
 577  */
 578 #define EXT4_FREE_BLOCKS_METADATA       0x0001
 579 #define EXT4_FREE_BLOCKS_FORGET         0x0002
 580 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
 581 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
 582 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
 583 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
 584
 585 /* Encryption algorithms */
 586 #define EXT4_ENCRYPTION_MODE_INVALID            0
 587 #define EXT4_ENCRYPTION_MODE_AES_256_XTS        1
 588 #define EXT4_ENCRYPTION_MODE_AES_256_GCM        2
 589 #define EXT4_ENCRYPTION_MODE_AES_256_CBC        3
 590 #define EXT4_ENCRYPTION_MODE_AES_256_CTS        4
 591
 592 #include "ext4_crypto.h"
 593
 594 /*
 595  * ioctl commands
 596  */
 597 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
 598 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
 599 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
 600 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
 601 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
 602 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
 603 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
 604 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
 605 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
 606 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
 607 #define EXT4_IOC_MIGRATE                _IO('f', 9)
 608  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
 609  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
 610 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
 611 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
 612 #define EXT4_IOC_RESIZE_FS              _IOW('f', 16, __u64)
 613 #define EXT4_IOC_SWAP_BOOT              _IO('f', 17)
 614 #define EXT4_IOC_PRECACHE_EXTENTS       _IO('f', 18)
 615 #define EXT4_IOC_SET_ENCRYPTION_POLICY  _IOR('f', 19, struct ext4_encryption_policy)
 616 #define EXT4_IOC_GET_ENCRYPTION_PWSALT  _IOW('f', 20, __u8[16])
 617 #define EXT4_IOC_GET_ENCRYPTION_POLICY  _IOW('f', 21, struct ext4_encryption_policy)
 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         spinlock_t i_raw_lock;  /* protects updates to the raw inode */
 880
 881         /*
 882          * File creation time. Its function is same as that of
 883          * struct timespec i_{a,c,m}time in the generic inode.
 884          */
 885         struct timespec i_crtime;
 886
 887         /* mballoc */
 888         struct list_head i_prealloc_list;
 889         spinlock_t i_prealloc_lock;
 890
 891         /* extents status tree */
 892         struct ext4_es_tree i_es_tree;
 893         rwlock_t i_es_lock;
 894         struct list_head i_es_list;
 895         unsigned int i_es_all_nr;       /* protected by i_es_lock */
 896         unsigned int i_es_shk_nr;       /* protected by i_es_lock */
 897         ext4_lblk_t i_es_shrink_lblk;   /* Offset where we start searching for
 898                                            extents to shrink. Protected by
 899                                            i_es_lock  */
 900
 901         /* ialloc */
 902         ext4_group_t    i_last_alloc_group;
 903
 904         /* allocation reservation info for delalloc */
 905         /* In case of bigalloc, these refer to clusters rather than blocks */
 906         unsigned int i_reserved_data_blocks;
 907         unsigned int i_reserved_meta_blocks;
 908         unsigned int i_allocated_meta_blocks;
 909         ext4_lblk_t i_da_metadata_calc_last_lblock;
 910         int i_da_metadata_calc_len;
 911
 912         /* on-disk additional length */
 913         __u16 i_extra_isize;
 914         char i_crypt_policy_flags;
 915
 916         /* Indicate the inline data space. */
 917         u16 i_inline_off;
 918         u16 i_inline_size;
 919
 920 #ifdef CONFIG_QUOTA
 921         /* quota space reservation, managed internally by quota code */
 922         qsize_t i_reserved_quota;
 923 #endif
 924
 925         /* Lock protecting lists below */
 926         spinlock_t i_completed_io_lock;
 927         /*
 928          * Completed IOs that need unwritten extents handling and have
 929          * transaction reserved
 930          */
 931         struct list_head i_rsv_conversion_list;
 932         /*
 933          * Completed IOs that need unwritten extents handling and don't have
 934          * transaction reserved
 935          */
 936         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
 937         atomic_t i_unwritten; /* Nr. of inflight conversions pending */
 938         struct work_struct i_rsv_conversion_work;
 939
 940         spinlock_t i_block_reservation_lock;
 941
 942         /*
 943          * Transactions that contain inode's metadata needed to complete
 944          * fsync and fdatasync, respectively.
 945          */
 946         tid_t i_sync_tid;
 947         tid_t i_datasync_tid;
 948
 949 #ifdef CONFIG_QUOTA
 950         struct dquot *i_dquot[MAXQUOTAS];
 951 #endif
 952
 953         /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
 954         __u32 i_csum_seed;
 955
 956 #ifdef CONFIG_EXT4_FS_ENCRYPTION
 957         /* Encryption params */
 958         struct ext4_encryption_key i_encryption_key;
 959 #endif
 960 };
 961
 962 /*
 963  * File system states
 964  */
 965 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
 966 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
 967 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
 968
 969 /*
 970  * Misc. filesystem flags
 971  */
 972 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
 973 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
 974 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
 975
 976 /*
 977  * Mount flags set via mount options or defaults
 978  */
 979 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
 980 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
 981 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
 982 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
 983 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
 984 #define EXT4_MOUNT_ERRORS_MASK          0x00070
 985 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
 986 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
 987 #ifdef CONFIG_FS_DAX
 988 #define EXT4_MOUNT_DAX                  0x00200 /* Direct Access */
 989 #else
 990 #define EXT4_MOUNT_DAX                  0
 991 #endif
 992 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
 993 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
 994 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
 995 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
 996 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
 997 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
 998 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
 999 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
1000 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
1001 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
1002 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
1003 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
1004 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
1005 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
1006 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
1007 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
1008 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
1009 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
1010 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
1011 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
1012 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
1013
1014 /*
1015  * Mount flags set either automatically (could not be set by mount option)
1016  * based on per file system feature or property or in special cases such as
1017  * distinguishing between explicit mount option definition and default.
1018  */
1019 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
1020                                                       specified delalloc */
1021 #define EXT4_MOUNT2_STD_GROUP_SIZE      0x00000002 /* We have standard group
1022                                                       size of blocksize * 8
1023                                                       blocks */
1024 #define EXT4_MOUNT2_HURD_COMPAT         0x00000004 /* Support HURD-castrated
1025                                                       file systems */
1026
1027 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
1028                                                 ~EXT4_MOUNT_##opt
1029 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
1030                                                 EXT4_MOUNT_##opt
1031 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
1032                                          EXT4_MOUNT_##opt)
1033
1034 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
1035                                                 ~EXT4_MOUNT2_##opt
1036 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
1037                                                 EXT4_MOUNT2_##opt
1038 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
1039                                          EXT4_MOUNT2_##opt)
1040
1041 #define ext4_test_and_set_bit           __test_and_set_bit_le
1042 #define ext4_set_bit                    __set_bit_le
1043 #define ext4_set_bit_atomic             ext2_set_bit_atomic
1044 #define ext4_test_and_clear_bit         __test_and_clear_bit_le
1045 #define ext4_clear_bit                  __clear_bit_le
1046 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
1047 #define ext4_test_bit                   test_bit_le
1048 #define ext4_find_next_zero_bit         find_next_zero_bit_le
1049 #define ext4_find_next_bit              find_next_bit_le
1050
1051 extern void ext4_set_bits(void *bm, int cur, int len);
1052
1053 /*
1054  * Maximal mount counts between two filesystem checks
1055  */
1056 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
1057 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
1058
1059 /*
1060  * Behaviour when detecting errors
1061  */
1062 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
1063 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
1064 #define EXT4_ERRORS_PANIC               3       /* Panic */
1065 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
1066
1067 /* Metadata checksum algorithm codes */
1068 #define EXT4_CRC32C_CHKSUM              1
1069
1070 /*
1071  * Structure of the super block
1072  */
1073 struct ext4_super_block {
1074 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1075         __le32  s_blocks_count_lo;      /* Blocks count */
1076         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1077         __le32  s_free_blocks_count_lo; /* Free blocks count */
1078 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1079         __le32  s_first_data_block;     /* First Data Block */
1080         __le32  s_log_block_size;       /* Block size */
1081         __le32  s_log_cluster_size;     /* Allocation cluster size */
1082 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1083         __le32  s_clusters_per_group;   /* # Clusters per group */
1084         __le32  s_inodes_per_group;     /* # Inodes per group */
1085         __le32  s_mtime;                /* Mount time */
1086 /*30*/  __le32  s_wtime;                /* Write time */
1087         __le16  s_mnt_count;            /* Mount count */
1088         __le16  s_max_mnt_count;        /* Maximal mount count */
1089         __le16  s_magic;                /* Magic signature */
1090         __le16  s_state;                /* File system state */
1091         __le16  s_errors;               /* Behaviour when detecting errors */
1092         __le16  s_minor_rev_level;      /* minor revision level */
1093 /*40*/  __le32  s_lastcheck;            /* time of last check */
1094         __le32  s_checkinterval;        /* max. time between checks */
1095         __le32  s_creator_os;           /* OS */
1096         __le32  s_rev_level;            /* Revision level */
1097 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1098         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1099         /*
1100          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1101          *
1102          * Note: the difference between the compatible feature set and
1103          * the incompatible feature set is that if there is a bit set
1104          * in the incompatible feature set that the kernel doesn't
1105          * know about, it should refuse to mount the filesystem.
1106          *
1107          * e2fsck's requirements are more strict; if it doesn't know
1108          * about a feature in either the compatible or incompatible
1109          * feature set, it must abort and not try to meddle with
1110          * things it doesn't understand...
1111          */
1112         __le32  s_first_ino;            /* First non-reserved inode */
1113         __le16  s_inode_size;           /* size of inode structure */
1114         __le16  s_block_group_nr;       /* block group # of this superblock */
1115         __le32  s_feature_compat;       /* compatible feature set */
1116 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1117         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1118 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1119 /*78*/  char    s_volume_name[16];      /* volume name */
1120 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1121 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1122         /*
1123          * Performance hints.  Directory preallocation should only
1124          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1125          */
1126         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1127         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1128         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1129         /*
1130          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1131          */
1132 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1133 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1134         __le32  s_journal_dev;          /* device number of journal file */
1135         __le32  s_last_orphan;          /* start of list of inodes to delete */
1136         __le32  s_hash_seed[4];         /* HTREE hash seed */
1137         __u8    s_def_hash_version;     /* Default hash version to use */
1138         __u8    s_jnl_backup_type;
1139         __le16  s_desc_size;            /* size of group descriptor */
1140 /*100*/ __le32  s_default_mount_opts;
1141         __le32  s_first_meta_bg;        /* First metablock block group */
1142         __le32  s_mkfs_time;            /* When the filesystem was created */
1143         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1144         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1145 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1146         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1147         __le32  s_free_blocks_count_hi; /* Free blocks count */
1148         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1149         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1150         __le32  s_flags;                /* Miscellaneous flags */
1151         __le16  s_raid_stride;          /* RAID stride */
1152         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1153         __le64  s_mmp_block;            /* Block for multi-mount protection */
1154         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1155         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1156         __u8    s_checksum_type;        /* metadata checksum algorithm used */
1157         __u8    s_encryption_level;     /* versioning level for encryption */
1158         __u8    s_reserved_pad;         /* Padding to next 32bits */
1159         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1160         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1161         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1162         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1163                                               snapshot's future use */
1164         __le32  s_snapshot_list;        /* inode number of the head of the
1165                                            on-disk snapshot list */
1166 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1167         __le32  s_error_count;          /* number of fs errors */
1168         __le32  s_first_error_time;     /* first time an error happened */
1169         __le32  s_first_error_ino;      /* inode involved in first error */
1170         __le64  s_first_error_block;    /* block involved of first error */
1171         __u8    s_first_error_func[32]; /* function where the error happened */
1172         __le32  s_first_error_line;     /* line number where error happened */
1173         __le32  s_last_error_time;      /* most recent time of an error */
1174         __le32  s_last_error_ino;       /* inode involved in last error */
1175         __le32  s_last_error_line;      /* line number where error happened */
1176         __le64  s_last_error_block;     /* block involved of last error */
1177         __u8    s_last_error_func[32];  /* function where the error happened */
1178 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1179         __u8    s_mount_opts[64];
1180         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1181         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1182         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1183         __le32  s_backup_bgs[2];        /* groups with sparse_super2 SBs */
1184         __u8    s_encrypt_algos[4];     /* Encryption algorithms in use  */
1185         __u8    s_encrypt_pw_salt[16];  /* Salt used for string2key algorithm */
1186         __le32  s_lpf_ino;              /* Location of the lost+found inode */
1187         __le32  s_reserved[100];        /* Padding to the end of the block */
1188         __le32  s_checksum;             /* crc32c(superblock) */
1189 };
1190
1191 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1192
1193 #ifdef __KERNEL__
1194
1195 /*
1196  * run-time mount flags
1197  */
1198 #define EXT4_MF_MNTDIR_SAMPLED          0x0001
1199 #define EXT4_MF_FS_ABORTED              0x0002  /* Fatal error detected */
1200 #define EXT4_MF_TEST_DUMMY_ENCRYPTION   0x0004
1201
1202 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1203 #define DUMMY_ENCRYPTION_ENABLED(sbi) (unlikely((sbi)->s_mount_flags & \
1204                                                 EXT4_MF_TEST_DUMMY_ENCRYPTION))
1205 #else
1206 #define DUMMY_ENCRYPTION_ENABLED(sbi) (0)
1207 #endif
1208
1209 /* Number of quota types we support */
1210 #define EXT4_MAXQUOTAS 2
1211
1212 /*
1213  * fourth extended-fs super-block data in memory
1214  */
1215 struct ext4_sb_info {
1216         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1217         unsigned long s_inodes_per_block;/* Number of inodes per block */
1218         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1219         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1220         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1221         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1222         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1223         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1224         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1225         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1226         unsigned long s_overhead;  /* # of fs overhead clusters */
1227         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1228         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1229         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1230         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1231         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1232         struct buffer_head **s_group_desc;
1233         unsigned int s_mount_opt;
1234         unsigned int s_mount_opt2;
1235         unsigned int s_mount_flags;
1236         unsigned int s_def_mount_opt;
1237         ext4_fsblk_t s_sb_block;
1238         atomic64_t s_resv_clusters;
1239         kuid_t s_resuid;
1240         kgid_t s_resgid;
1241         unsigned short s_mount_state;
1242         unsigned short s_pad;
1243         int s_addr_per_block_bits;
1244         int s_desc_per_block_bits;
1245         int s_inode_size;
1246         int s_first_ino;
1247         unsigned int s_inode_readahead_blks;
1248         unsigned int s_inode_goal;
1249         spinlock_t s_next_gen_lock;
1250         u32 s_next_generation;
1251         u32 s_hash_seed[4];
1252         int s_def_hash_version;
1253         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1254         struct percpu_counter s_freeclusters_counter;
1255         struct percpu_counter s_freeinodes_counter;
1256         struct percpu_counter s_dirs_counter;
1257         struct percpu_counter s_dirtyclusters_counter;
1258         struct blockgroup_lock *s_blockgroup_lock;
1259         struct proc_dir_entry *s_proc;
1260         struct kobject s_kobj;
1261         struct completion s_kobj_unregister;
1262         struct super_block *s_sb;
1263
1264         /* Journaling */
1265         struct journal_s *s_journal;
1266         struct list_head s_orphan;
1267         struct mutex s_orphan_lock;
1268         unsigned long s_resize_flags;           /* Flags indicating if there
1269                                                    is a resizer */
1270         unsigned long s_commit_interval;
1271         u32 s_max_batch_time;
1272         u32 s_min_batch_time;
1273         struct block_device *journal_bdev;
1274 #ifdef CONFIG_QUOTA
1275         char *s_qf_names[EXT4_MAXQUOTAS];       /* Names of quota files with journalled quota */
1276         int s_jquota_fmt;                       /* Format of quota to use */
1277 #endif
1278         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1279         struct rb_root system_blks;
1280
1281 #ifdef EXTENTS_STATS
1282         /* ext4 extents stats */
1283         unsigned long s_ext_min;
1284         unsigned long s_ext_max;
1285         unsigned long s_depth_max;
1286         spinlock_t s_ext_stats_lock;
1287         unsigned long s_ext_blocks;
1288         unsigned long s_ext_extents;
1289 #endif
1290
1291         /* for buddy allocator */
1292         struct ext4_group_info ***s_group_info;
1293         struct inode *s_buddy_cache;
1294         spinlock_t s_md_lock;
1295         unsigned short *s_mb_offsets;
1296         unsigned int *s_mb_maxs;
1297         unsigned int s_group_info_size;
1298
1299         /* tunables */
1300         unsigned long s_stripe;
1301         unsigned int s_mb_stream_request;
1302         unsigned int s_mb_max_to_scan;
1303         unsigned int s_mb_min_to_scan;
1304         unsigned int s_mb_stats;
1305         unsigned int s_mb_order2_reqs;
1306         unsigned int s_mb_group_prealloc;
1307         unsigned int s_max_dir_size_kb;
1308         /* where last allocation was done - for stream allocation */
1309         unsigned long s_mb_last_group;
1310         unsigned long s_mb_last_start;
1311
1312         /* stats for buddy allocator */
1313         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1314         atomic_t s_bal_success; /* we found long enough chunks */
1315         atomic_t s_bal_allocated;       /* in blocks */
1316         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1317         atomic_t s_bal_goals;   /* goal hits */
1318         atomic_t s_bal_breaks;  /* too long searches */
1319         atomic_t s_bal_2orders; /* 2^order hits */
1320         spinlock_t s_bal_lock;
1321         unsigned long s_mb_buddies_generated;
1322         unsigned long long s_mb_generation_time;
1323         atomic_t s_mb_lost_chunks;
1324         atomic_t s_mb_preallocated;
1325         atomic_t s_mb_discarded;
1326         atomic_t s_lock_busy;
1327
1328         /* locality groups */
1329         struct ext4_locality_group __percpu *s_locality_groups;
1330
1331         /* for write statistics */
1332         unsigned long s_sectors_written_start;
1333         u64 s_kbytes_written;
1334
1335         /* the size of zero-out chunk */
1336         unsigned int s_extent_max_zeroout_kb;
1337
1338         unsigned int s_log_groups_per_flex;
1339         struct flex_groups *s_flex_groups;
1340         ext4_group_t s_flex_groups_allocated;
1341
1342         /* workqueue for reserved extent conversions (buffered io) */
1343         struct workqueue_struct *rsv_conversion_wq;
1344
1345         /* timer for periodic error stats printing */
1346         struct timer_list s_err_report;
1347
1348         /* Lazy inode table initialization info */
1349         struct ext4_li_request *s_li_request;
1350         /* Wait multiplier for lazy initialization thread */
1351         unsigned int s_li_wait_mult;
1352
1353         /* Kernel thread for multiple mount protection */
1354         struct task_struct *s_mmp_tsk;
1355
1356         /* record the last minlen when FITRIM is called. */
1357         atomic_t s_last_trim_minblks;
1358
1359         /* Reference to checksum algorithm driver via cryptoapi */
1360         struct crypto_shash *s_chksum_driver;
1361
1362         /* Precomputed FS UUID checksum for seeding other checksums */
1363         __u32 s_csum_seed;
1364
1365         /* Reclaim extents from extent status tree */
1366         struct shrinker s_es_shrinker;
1367         struct list_head s_es_list;     /* List of inodes with reclaimable extents */
1368         long s_es_nr_inode;
1369         struct ext4_es_stats s_es_stats;
1370         struct mb_cache *s_mb_cache;
1371         spinlock_t s_es_lock ____cacheline_aligned_in_smp;
1372
1373         /* Ratelimit ext4 messages. */
1374         struct ratelimit_state s_err_ratelimit_state;
1375         struct ratelimit_state s_warning_ratelimit_state;
1376         struct ratelimit_state s_msg_ratelimit_state;
1377
1378 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1379         /* Encryption */
1380         uint32_t s_file_encryption_mode;
1381         uint32_t s_dir_encryption_mode;
1382 #endif
1383 };
1384
1385 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1386 {
1387         return sb->s_fs_info;
1388 }
1389 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1390 {
1391         return container_of(inode, struct ext4_inode_info, vfs_inode);
1392 }
1393
1394 static inline struct timespec ext4_current_time(struct inode *inode)
1395 {
1396         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1397                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1398 }
1399
1400 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1401 {
1402         return ino == EXT4_ROOT_INO ||
1403                 ino == EXT4_USR_QUOTA_INO ||
1404                 ino == EXT4_GRP_QUOTA_INO ||
1405                 ino == EXT4_BOOT_LOADER_INO ||
1406                 ino == EXT4_JOURNAL_INO ||
1407                 ino == EXT4_RESIZE_INO ||
1408                 (ino >= EXT4_FIRST_INO(sb) &&
1409                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1410 }
1411
1412 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
1413                                               struct ext4_io_end *io_end)
1414 {
1415         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
1416                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
1417                 atomic_inc(&EXT4_I(inode)->i_unwritten);
1418         }
1419 }
1420
1421 static inline ext4_io_end_t *ext4_inode_aio(struct inode *inode)
1422 {
1423         return inode->i_private;
1424 }
1425
1426 static inline void ext4_inode_aio_set(struct inode *inode, ext4_io_end_t *io)
1427 {
1428         inode->i_private = io;
1429 }
1430
1431 /*
1432  * Inode dynamic state flags
1433  */
1434 enum {
1435         EXT4_STATE_JDATA,               /* journaled data exists */
1436         EXT4_STATE_NEW,                 /* inode is newly created */
1437         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1438         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1439         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1440         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1441         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1442         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1443         EXT4_STATE_DIOREAD_LOCK,        /* Disable support for dio read
1444                                            nolocking */
1445         EXT4_STATE_MAY_INLINE_DATA,     /* may have in-inode data */
1446         EXT4_STATE_ORDERED_MODE,        /* data=ordered mode */
1447         EXT4_STATE_EXT_PRECACHED,       /* extents have been precached */
1448 };
1449
1450 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1451 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1452 {                                                                       \
1453         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1454 }                                                                       \
1455 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1456 {                                                                       \
1457         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1458 }                                                                       \
1459 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1460 {                                                                       \
1461         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1462 }
1463
1464 /* Add these declarations here only so that these functions can be
1465  * found by name.  Otherwise, they are very hard to locate. */
1466 static inline int ext4_test_inode_flag(struct inode *inode, int bit);
1467 static inline void ext4_set_inode_flag(struct inode *inode, int bit);
1468 static inline void ext4_clear_inode_flag(struct inode *inode, int bit);
1469 EXT4_INODE_BIT_FNS(flag, flags, 0)
1470
1471 /* Add these declarations here only so that these functions can be
1472  * found by name.  Otherwise, they are very hard to locate. */
1473 static inline int ext4_test_inode_state(struct inode *inode, int bit);
1474 static inline void ext4_set_inode_state(struct inode *inode, int bit);
1475 static inline void ext4_clear_inode_state(struct inode *inode, int bit);
1476 #if (BITS_PER_LONG < 64)
1477 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1478
1479 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1480 {
1481         (ei)->i_state_flags = 0;
1482 }
1483 #else
1484 EXT4_INODE_BIT_FNS(state, flags, 32)
1485
1486 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1487 {
1488         /* We depend on the fact that callers will set i_flags */
1489 }
1490 #endif
1491 #else
1492 /* Assume that user mode programs are passing in an ext4fs superblock, not
1493  * a kernel struct super_block.  This will allow us to call the feature-test
1494  * macros from user land. */
1495 #define EXT4_SB(sb)     (sb)
1496 #endif
1497
1498 /*
1499  * Returns true if the inode is inode is encrypted
1500  */
1501 static inline int ext4_encrypted_inode(struct inode *inode)
1502 {
1503 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1504         return ext4_test_inode_flag(inode, EXT4_INODE_ENCRYPT);
1505 #else
1506         return 0;
1507 #endif
1508 }
1509
1510 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1511
1512 /*
1513  * Codes for operating systems
1514  */
1515 #define EXT4_OS_LINUX           0
1516 #define EXT4_OS_HURD            1
1517 #define EXT4_OS_MASIX           2
1518 #define EXT4_OS_FREEBSD         3
1519 #define EXT4_OS_LITES           4
1520
1521 /*
1522  * Revision levels
1523  */
1524 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1525 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1526
1527 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1528 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1529
1530 #define EXT4_GOOD_OLD_INODE_SIZE 128
1531
1532 /*
1533  * Feature set definitions
1534  */
1535
1536 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1537         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1538 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1539         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1540 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1541         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1542 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1543         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1544 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1545         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1546 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1547         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1548 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1549         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1550 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1551         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1552 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1553         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1554
1555 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1556 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1557 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1558 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1559 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1560 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1561 #define EXT4_FEATURE_COMPAT_SPARSE_SUPER2       0x0200
1562
1563 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1564 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1565 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1566 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1567 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1568 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1569 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1570 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1571 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1572 /*
1573  * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM).  When
1574  * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
1575  * all other data structures' checksums.  However, the METADATA_CSUM and
1576  * GDT_CSUM bits are mutually exclusive.
1577  */
1578 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM    0x0400
1579 #define EXT4_FEATURE_RO_COMPAT_READONLY         0x1000
1580
1581 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1582 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1583 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1584 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1585 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1586 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1587 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1588 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1589 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1590 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1591 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1592 #define EXT4_FEATURE_INCOMPAT_BG_USE_META_CSUM  0x2000 /* use crc32c for bg */
1593 #define EXT4_FEATURE_INCOMPAT_LARGEDIR          0x4000 /* >2GB or 3-lvl htree */
1594 #define EXT4_FEATURE_INCOMPAT_INLINE_DATA       0x8000 /* data in inode */
1595 #define EXT4_FEATURE_INCOMPAT_ENCRYPT           0x10000
1596
1597 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1598 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1599                                          EXT4_FEATURE_INCOMPAT_META_BG)
1600 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1601                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1602                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1603
1604 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1605 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1606                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1607                                          EXT4_FEATURE_INCOMPAT_META_BG)
1608 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1609                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1610                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1611
1612 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1613 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1614                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1615                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1616                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1617                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1618                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1619                                          EXT4_FEATURE_INCOMPAT_MMP | \
1620                                          EXT4_FEATURE_INCOMPAT_INLINE_DATA | \
1621                                          EXT4_FEATURE_INCOMPAT_ENCRYPT)
1622 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1623                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1624                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1625                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1626                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1627                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1628                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1629                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
1630                                          EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
1631                                          EXT4_FEATURE_RO_COMPAT_QUOTA)
1632
1633 /*
1634  * Default values for user and/or group using reserved blocks
1635  */
1636 #define EXT4_DEF_RESUID         0
1637 #define EXT4_DEF_RESGID         0
1638
1639 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1640
1641 /*
1642  * Default mount options
1643  */
1644 #define EXT4_DEFM_DEBUG         0x0001
1645 #define EXT4_DEFM_BSDGROUPS     0x0002
1646 #define EXT4_DEFM_XATTR_USER    0x0004
1647 #define EXT4_DEFM_ACL           0x0008
1648 #define EXT4_DEFM_UID16         0x0010
1649 #define EXT4_DEFM_JMODE         0x0060
1650 #define EXT4_DEFM_JMODE_DATA    0x0020
1651 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1652 #define EXT4_DEFM_JMODE_WBACK   0x0060
1653 #define EXT4_DEFM_NOBARRIER     0x0100
1654 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1655 #define EXT4_DEFM_DISCARD       0x0400
1656 #define EXT4_DEFM_NODELALLOC    0x0800
1657
1658 /*
1659  * Default journal batch times
1660  */
1661 #define EXT4_DEF_MIN_BATCH_TIME 0
1662 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1663
1664 /*
1665  * Minimum number of groups in a flexgroup before we separate out
1666  * directories into the first block group of a flexgroup
1667  */
1668 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1669
1670 /*
1671  * Structure of a directory entry
1672  */
1673 #define EXT4_NAME_LEN 255
1674
1675 struct ext4_dir_entry {
1676         __le32  inode;                  /* Inode number */
1677         __le16  rec_len;                /* Directory entry length */
1678         __le16  name_len;               /* Name length */
1679         char    name[EXT4_NAME_LEN];    /* File name */
1680 };
1681
1682 /*
1683  * The new version of the directory entry.  Since EXT4 structures are
1684  * stored in intel byte order, and the name_len field could never be
1685  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1686  * file_type field.
1687  */
1688 struct ext4_dir_entry_2 {
1689         __le32  inode;                  /* Inode number */
1690         __le16  rec_len;                /* Directory entry length */
1691         __u8    name_len;               /* Name length */
1692         __u8    file_type;
1693         char    name[EXT4_NAME_LEN];    /* File name */
1694 };
1695
1696 /*
1697  * This is a bogus directory entry at the end of each leaf block that
1698  * records checksums.
1699  */
1700 struct ext4_dir_entry_tail {
1701         __le32  det_reserved_zero1;     /* Pretend to be unused */
1702         __le16  det_rec_len;            /* 12 */
1703         __u8    det_reserved_zero2;     /* Zero name length */
1704         __u8    det_reserved_ft;        /* 0xDE, fake file type */
1705         __le32  det_checksum;           /* crc32c(uuid+inum+dirblock) */
1706 };
1707
1708 #define EXT4_DIRENT_TAIL(block, blocksize) \
1709         ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
1710                                         ((blocksize) - \
1711                                          sizeof(struct ext4_dir_entry_tail))))
1712
1713 /*
1714  * Ext4 directory file types.  Only the low 3 bits are used.  The
1715  * other bits are reserved for now.
1716  */
1717 #define EXT4_FT_UNKNOWN         0
1718 #define EXT4_FT_REG_FILE        1
1719 #define EXT4_FT_DIR             2
1720 #define EXT4_FT_CHRDEV          3
1721 #define EXT4_FT_BLKDEV          4
1722 #define EXT4_FT_FIFO            5
1723 #define EXT4_FT_SOCK            6
1724 #define EXT4_FT_SYMLINK         7
1725
1726 #define EXT4_FT_MAX             8
1727
1728 #define EXT4_FT_DIR_CSUM        0xDE
1729
1730 /*
1731  * EXT4_DIR_PAD defines the directory entries boundaries
1732  *
1733  * NOTE: It must be a multiple of 4
1734  */
1735 #define EXT4_DIR_PAD                    4
1736 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1737 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1738                                          ~EXT4_DIR_ROUND)
1739 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1740
1741 /*
1742  * If we ever get support for fs block sizes > page_size, we'll need
1743  * to remove the #if statements in the next two functions...
1744  */
1745 static inline unsigned int
1746 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1747 {
1748         unsigned len = le16_to_cpu(dlen);
1749
1750 #if (PAGE_CACHE_SIZE >= 65536)
1751         if (len == EXT4_MAX_REC_LEN || len == 0)
1752                 return blocksize;
1753         return (len & 65532) | ((len & 3) << 16);
1754 #else
1755         return len;
1756 #endif
1757 }
1758
1759 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1760 {
1761         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1762                 BUG();
1763 #if (PAGE_CACHE_SIZE >= 65536)
1764         if (len < 65536)
1765                 return cpu_to_le16(len);
1766         if (len == blocksize) {
1767                 if (blocksize == 65536)
1768                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1769                 else
1770                         return cpu_to_le16(0);
1771         }
1772         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1773 #else
1774         return cpu_to_le16(len);
1775 #endif
1776 }
1777
1778 /*
1779  * Hash Tree Directory indexing
1780  * (c) Daniel Phillips, 2001
1781  */
1782
1783 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1784                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1785                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1786 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1787 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1788
1789 /* Legal values for the dx_root hash_version field: */
1790
1791 #define DX_HASH_LEGACY          0
1792 #define DX_HASH_HALF_MD4        1
1793 #define DX_HASH_TEA             2
1794 #define DX_HASH_LEGACY_UNSIGNED 3
1795 #define DX_HASH_HALF_MD4_UNSIGNED       4
1796 #define DX_HASH_TEA_UNSIGNED            5
1797
1798 static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
1799                               const void *address, unsigned int length)
1800 {
1801         struct {
1802                 struct shash_desc shash;
1803                 char ctx[4];
1804         } desc;
1805         int err;
1806
1807         BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
1808
1809         desc.shash.tfm = sbi->s_chksum_driver;
1810         desc.shash.flags = 0;
1811         *(u32 *)desc.ctx = crc;
1812
1813         err = crypto_shash_update(&desc.shash, address, length);
1814         BUG_ON(err);
1815
1816         return *(u32 *)desc.ctx;
1817 }
1818
1819 #ifdef __KERNEL__
1820
1821 /* hash info structure used by the directory hash */
1822 struct dx_hash_info
1823 {
1824         u32             hash;
1825         u32             minor_hash;
1826         int             hash_version;
1827         u32             *seed;
1828 };
1829
1830
1831 /* 32 and 64 bit signed EOF for dx directories */
1832 #define EXT4_HTREE_EOF_32BIT   ((1UL  << (32 - 1)) - 1)
1833 #define EXT4_HTREE_EOF_64BIT   ((1ULL << (64 - 1)) - 1)
1834
1835
1836 /*
1837  * Control parameters used by ext4_htree_next_block
1838  */
1839 #define HASH_NB_ALWAYS          1
1840
1841
1842 /*
1843  * Describe an inode's exact location on disk and in memory
1844  */
1845 struct ext4_iloc
1846 {
1847         struct buffer_head *bh;
1848         unsigned long offset;
1849         ext4_group_t block_group;
1850 };
1851
1852 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1853 {
1854         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1855 }
1856
1857 /*
1858  * This structure is stuffed into the struct file's private_data field
1859  * for directories.  It is where we put information so that we can do
1860  * readdir operations in hash tree order.
1861  */
1862 struct dir_private_info {
1863         struct rb_root  root;
1864         struct rb_node  *curr_node;
1865         struct fname    *extra_fname;
1866         loff_t          last_pos;
1867         __u32           curr_hash;
1868         __u32           curr_minor_hash;
1869         __u32           next_hash;
1870 };
1871
1872 /* calculate the first block number of the group */
1873 static inline ext4_fsblk_t
1874 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1875 {
1876         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1877                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1878 }
1879
1880 /*
1881  * Special error return code only used by dx_probe() and its callers.
1882  */
1883 #define ERR_BAD_DX_DIR  (-(MAX_ERRNO - 1))
1884
1885 /*
1886  * Timeout and state flag for lazy initialization inode thread.
1887  */
1888 #define EXT4_DEF_LI_WAIT_MULT                   10
1889 #define EXT4_DEF_LI_MAX_START_DELAY             5
1890 #define EXT4_LAZYINIT_QUIT                      0x0001
1891 #define EXT4_LAZYINIT_RUNNING                   0x0002
1892
1893 /*
1894  * Lazy inode table initialization info
1895  */
1896 struct ext4_lazy_init {
1897         unsigned long           li_state;
1898         struct list_head        li_request_list;
1899         struct mutex            li_list_mtx;
1900 };
1901
1902 struct ext4_li_request {
1903         struct super_block      *lr_super;
1904         struct ext4_sb_info     *lr_sbi;
1905         ext4_group_t            lr_next_group;
1906         struct list_head        lr_request;
1907         unsigned long           lr_next_sched;
1908         unsigned long           lr_timeout;
1909 };
1910
1911 struct ext4_features {
1912         struct kobject f_kobj;
1913         struct completion f_kobj_unregister;
1914 };
1915
1916 /*
1917  * This structure will be used for multiple mount protection. It will be
1918  * written into the block number saved in the s_mmp_block field in the
1919  * superblock. Programs that check MMP should assume that if
1920  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1921  * to use the filesystem, regardless of how old the timestamp is.
1922  */
1923 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
1924 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1925 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
1926 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
1927
1928 struct mmp_struct {
1929         __le32  mmp_magic;              /* Magic number for MMP */
1930         __le32  mmp_seq;                /* Sequence no. updated periodically */
1931
1932         /*
1933          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1934          * purposes and do not affect the correctness of the algorithm
1935          */
1936         __le64  mmp_time;               /* Time last updated */
1937         char    mmp_nodename[64];       /* Node which last updated MMP block */
1938         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
1939
1940         /*
1941          * mmp_check_interval is used to verify if the MMP block has been
1942          * updated on the block device. The value is updated based on the
1943          * maximum time to write the MMP block during an update cycle.
1944          */
1945         __le16  mmp_check_interval;
1946
1947         __le16  mmp_pad1;
1948         __le32  mmp_pad2[226];
1949         __le32  mmp_checksum;           /* crc32c(uuid+mmp_block) */
1950 };
1951
1952 /* arguments passed to the mmp thread */
1953 struct mmpd_data {
1954         struct buffer_head *bh; /* bh from initial read_mmp_block() */
1955         struct super_block *sb;  /* super block of the fs */
1956 };
1957
1958 /*
1959  * Check interval multiplier
1960  * The MMP block is written every update interval and initially checked every
1961  * update interval x the multiplier (the value is then adapted based on the
1962  * write latency). The reason is that writes can be delayed under load and we
1963  * don't want readers to incorrectly assume that the filesystem is no longer
1964  * in use.
1965  */
1966 #define EXT4_MMP_CHECK_MULT             2UL
1967
1968 /*
1969  * Minimum interval for MMP checking in seconds.
1970  */
1971 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
1972
1973 /*
1974  * Maximum interval for MMP checking in seconds.
1975  */
1976 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
1977
1978 /*
1979  * Function prototypes
1980  */
1981
1982 /*
1983  * Ok, these declarations are also in <linux/kernel.h> but none of the
1984  * ext4 source programs needs to include it so they are duplicated here.
1985  */
1986 # define NORET_TYPE     /**/
1987 # define ATTRIB_NORET   __attribute__((noreturn))
1988 # define NORET_AND      noreturn,
1989
1990 /* bitmap.c */
1991 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
1992 void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1993                                 struct ext4_group_desc *gdp,
1994                                 struct buffer_head *bh, int sz);
1995 int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1996                                   struct ext4_group_desc *gdp,
1997                                   struct buffer_head *bh, int sz);
1998 void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1999                                 struct ext4_group_desc *gdp,
2000                                 struct buffer_head *bh);
2001 int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2002                                   struct ext4_group_desc *gdp,
2003                                   struct buffer_head *bh);
2004
2005 /* balloc.c */
2006 extern void ext4_get_group_no_and_offset(struct super_block *sb,
2007                                          ext4_fsblk_t blocknr,
2008                                          ext4_group_t *blockgrpp,
2009                                          ext4_grpblk_t *offsetp);
2010 extern ext4_group_t ext4_get_group_number(struct super_block *sb,
2011                                           ext4_fsblk_t block);
2012
2013 extern unsigned int ext4_block_group(struct super_block *sb,
2014                         ext4_fsblk_t blocknr);
2015 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
2016                         ext4_fsblk_t blocknr);
2017 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
2018 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
2019                         ext4_group_t group);
2020 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
2021                                          ext4_fsblk_t goal,
2022                                          unsigned int flags,
2023                                          unsigned long *count,
2024                                          int *errp);
2025 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
2026                                     s64 nclusters, unsigned int flags);
2027 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
2028 extern void ext4_check_blocks_bitmap(struct super_block *);
2029 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
2030                                                     ext4_group_t block_group,
2031                                                     struct buffer_head ** bh);
2032 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
2033
2034 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
2035                                                 ext4_group_t block_group);
2036 extern int ext4_wait_block_bitmap(struct super_block *sb,
2037                                   ext4_group_t block_group,
2038                                   struct buffer_head *bh);
2039 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
2040                                                   ext4_group_t block_group);
2041 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
2042                                               ext4_group_t block_group,
2043                                               struct ext4_group_desc *gdp);
2044 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
2045
2046 /* crypto_policy.c */
2047 int ext4_is_child_context_consistent_with_parent(struct inode *parent,
2048                                                  struct inode *child);
2049 int ext4_inherit_context(struct inode *parent, struct inode *child);
2050 void ext4_to_hex(char *dst, char *src, size_t src_size);
2051 int ext4_process_policy(const struct ext4_encryption_policy *policy,
2052                         struct inode *inode);
2053 int ext4_get_policy(struct inode *inode,
2054                     struct ext4_encryption_policy *policy);
2055
2056 /* crypto.c */
2057 bool ext4_valid_contents_enc_mode(uint32_t mode);
2058 uint32_t ext4_validate_encryption_key_size(uint32_t mode, uint32_t size);
2059 extern struct workqueue_struct *ext4_read_workqueue;
2060 struct ext4_crypto_ctx *ext4_get_crypto_ctx(struct inode *inode);
2061 void ext4_release_crypto_ctx(struct ext4_crypto_ctx *ctx);
2062 void ext4_restore_control_page(struct page *data_page);
2063 struct page *ext4_encrypt(struct inode *inode,
2064                           struct page *plaintext_page);
2065 int ext4_decrypt(struct ext4_crypto_ctx *ctx, struct page *page);
2066 int ext4_decrypt_one(struct inode *inode, struct page *page);
2067 int ext4_encrypted_zeroout(struct inode *inode, struct ext4_extent *ex);
2068
2069 #ifdef CONFIG_EXT4_FS_ENCRYPTION
2070 int ext4_init_crypto(void);
2071 void ext4_exit_crypto(void);
2072 static inline int ext4_sb_has_crypto(struct super_block *sb)
2073 {
2074         return EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
2075 }
2076 #else
2077 static inline int ext4_init_crypto(void) { return 0; }
2078 static inline void ext4_exit_crypto(void) { }
2079 static inline int ext4_sb_has_crypto(struct super_block *sb)
2080 {
2081         return 0;
2082 }
2083 #endif
2084
2085 /* crypto_fname.c */
2086 bool ext4_valid_filenames_enc_mode(uint32_t mode);
2087 u32 ext4_fname_crypto_round_up(u32 size, u32 blksize);
2088 int ext4_fname_crypto_alloc_buffer(struct ext4_fname_crypto_ctx *ctx,
2089                                    u32 ilen, struct ext4_str *crypto_str);
2090 int _ext4_fname_disk_to_usr(struct ext4_fname_crypto_ctx *ctx,
2091                             struct dx_hash_info *hinfo,
2092                             const struct ext4_str *iname,
2093                             struct ext4_str *oname);
2094 int ext4_fname_disk_to_usr(struct ext4_fname_crypto_ctx *ctx,
2095                            struct dx_hash_info *hinfo,
2096                            const struct ext4_dir_entry_2 *de,
2097                            struct ext4_str *oname);
2098 int ext4_fname_usr_to_disk(struct ext4_fname_crypto_ctx *ctx,
2099                            const struct qstr *iname,
2100                            struct ext4_str *oname);
2101 int ext4_fname_usr_to_hash(struct ext4_fname_crypto_ctx *ctx,
2102                            const struct qstr *iname,
2103                            struct dx_hash_info *hinfo);
2104 int ext4_fname_crypto_namelen_on_disk(struct ext4_fname_crypto_ctx *ctx,
2105                                       u32 namelen);
2106 int ext4_fname_match(struct ext4_fname_crypto_ctx *ctx, struct ext4_str *cstr,
2107                      int len, const char * const name,
2108                      struct ext4_dir_entry_2 *de);
2109
2110
2111 #ifdef CONFIG_EXT4_FS_ENCRYPTION
2112 void ext4_put_fname_crypto_ctx(struct ext4_fname_crypto_ctx **ctx);
2113 struct ext4_fname_crypto_ctx *ext4_get_fname_crypto_ctx(struct inode *inode,
2114                                                         u32 max_len);
2115 void ext4_fname_crypto_free_buffer(struct ext4_str *crypto_str);
2116 #else
2117 static inline
2118 void ext4_put_fname_crypto_ctx(struct ext4_fname_crypto_ctx **ctx) { }
2119 static inline
2120 struct ext4_fname_crypto_ctx *ext4_get_fname_crypto_ctx(struct inode *inode,
2121                                                         u32 max_len)
2122 {
2123         return NULL;
2124 }
2125 static inline void ext4_fname_crypto_free_buffer(struct ext4_str *p) { }
2126 #endif
2127
2128
2129 /* crypto_key.c */
2130 int ext4_generate_encryption_key(struct inode *inode);
2131
2132 #ifdef CONFIG_EXT4_FS_ENCRYPTION
2133 int ext4_has_encryption_key(struct inode *inode);
2134 #else
2135 static inline int ext4_has_encryption_key(struct inode *inode)
2136 {
2137         return 0;
2138 }
2139 #endif
2140
2141
2142 /* dir.c */
2143 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
2144                                   struct file *,
2145                                   struct ext4_dir_entry_2 *,
2146                                   struct buffer_head *, char *, int,
2147                                   unsigned int);
2148 #define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset)      \
2149         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
2150                                         (de), (bh), (buf), (size), (offset)))
2151 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
2152                                 __u32 minor_hash,
2153                                 struct ext4_dir_entry_2 *dirent,
2154                                 struct ext4_str *ent_name);
2155 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
2156 extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
2157                              struct buffer_head *bh,
2158                              void *buf, int buf_size,
2159                              const char *name, int namelen,
2160                              struct ext4_dir_entry_2 **dest_de);
2161 int ext4_insert_dentry(struct inode *dir,
2162                         struct inode *inode,
2163                         struct ext4_dir_entry_2 *de,
2164                         int buf_size,
2165                        const struct qstr *iname,
2166                         const char *name, int namelen);
2167 static inline void ext4_update_dx_flag(struct inode *inode)
2168 {
2169         if (!EXT4_HAS_COMPAT_FEATURE(inode->i_sb,
2170                                      EXT4_FEATURE_COMPAT_DIR_INDEX))
2171                 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
2172 }
2173 static unsigned char ext4_filetype_table[] = {
2174         DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
2175 };
2176
2177 static inline  unsigned char get_dtype(struct super_block *sb, int filetype)
2178 {
2179         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE) ||
2180             (filetype >= EXT4_FT_MAX))
2181                 return DT_UNKNOWN;
2182
2183         return ext4_filetype_table[filetype];
2184 }
2185 extern int ext4_check_all_de(struct inode *dir, struct buffer_head *bh,
2186                              void *buf, int buf_size);
2187
2188 /* fsync.c */
2189 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2190
2191 /* hash.c */
2192 extern int ext4fs_dirhash(const char *name, int len, struct
2193                           dx_hash_info *hinfo);
2194
2195 /* ialloc.c */
2196 extern struct inode *__ext4_new_inode(handle_t *, struct inode *, umode_t,
2197                                       const struct qstr *qstr, __u32 goal,
2198                                       uid_t *owner, int handle_type,
2199                                       unsigned int line_no, int nblocks);
2200
2201 #define ext4_new_inode(handle, dir, mode, qstr, goal, owner) \
2202         __ext4_new_inode((handle), (dir), (mode), (qstr), (goal), (owner), \
2203                          0, 0, 0)
2204 #define ext4_new_inode_start_handle(dir, mode, qstr, goal, owner, \
2205                                     type, nblocks)                  \
2206         __ext4_new_inode(NULL, (dir), (mode), (qstr), (goal), (owner), \
2207                          (type), __LINE__, (nblocks))
2208
2209
2210 extern void ext4_free_inode(handle_t *, struct inode *);
2211 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2212 extern unsigned long ext4_count_free_inodes(struct super_block *);
2213 extern unsigned long ext4_count_dirs(struct super_block *);
2214 extern void ext4_check_inodes_bitmap(struct super_block *);
2215 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2216 extern int ext4_init_inode_table(struct super_block *sb,
2217                                  ext4_group_t group, int barrier);
2218 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2219
2220 /* mballoc.c */
2221 extern long ext4_mb_stats;
2222 extern long ext4_mb_max_to_scan;
2223 extern int ext4_mb_init(struct super_block *);
2224 extern int ext4_mb_release(struct super_block *);
2225 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2226                                 struct ext4_allocation_request *, int *);
2227 extern int ext4_mb_reserve_blocks(struct super_block *, int);
2228 extern void ext4_discard_preallocations(struct inode *);
2229 extern int __init ext4_init_mballoc(void);
2230 extern void ext4_exit_mballoc(void);
2231 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2232                              struct buffer_head *bh, ext4_fsblk_t block,
2233                              unsigned long count, int flags);
2234 extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2235                                    ext4_group_t ngroups);
2236 extern int ext4_mb_add_groupinfo(struct super_block *sb,
2237                 ext4_group_t i, struct ext4_group_desc *desc);
2238 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2239                                 ext4_fsblk_t block, unsigned long count);
2240 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2241
2242 /* inode.c */
2243 int ext4_inode_is_fast_symlink(struct inode *inode);
2244 struct buffer_head *ext4_getblk(handle_t *, struct inode *, ext4_lblk_t, int);
2245 struct buffer_head *ext4_bread(handle_t *, struct inode *, ext4_lblk_t, int);
2246 int ext4_get_block_write(struct inode *inode, sector_t iblock,
2247                          struct buffer_head *bh_result, int create);
2248 int ext4_get_block(struct inode *inode, sector_t iblock,
2249                                 struct buffer_head *bh_result, int create);
2250 int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2251                            struct buffer_head *bh, int create);
2252 int ext4_walk_page_buffers(handle_t *handle,
2253                            struct buffer_head *head,
2254                            unsigned from,
2255                            unsigned to,
2256                            int *partial,
2257                            int (*fn)(handle_t *handle,
2258                                      struct buffer_head *bh));
2259 int do_journal_get_write_access(handle_t *handle,
2260                                 struct buffer_head *bh);
2261 #define FALL_BACK_TO_NONDELALLOC 1
2262 #define CONVERT_INLINE_DATA      2
2263
2264 extern struct inode *ext4_iget(struct super_block *, unsigned long);
2265 extern struct inode *ext4_iget_normal(struct super_block *, unsigned long);
2266 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
2267 extern int  ext4_setattr(struct dentry *, struct iattr *);
2268 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
2269                                 struct kstat *stat);
2270 extern void ext4_evict_inode(struct inode *);
2271 extern void ext4_clear_inode(struct inode *);
2272 extern int  ext4_sync_inode(handle_t *, struct inode *);
2273 extern void ext4_dirty_inode(struct inode *, int);
2274 extern int ext4_change_inode_journal_flag(struct inode *, int);
2275 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
2276 extern int ext4_inode_attach_jinode(struct inode *inode);
2277 extern int ext4_can_truncate(struct inode *inode);
2278 extern void ext4_truncate(struct inode *);
2279 extern int ext4_punch_hole(struct inode *inode, loff_t offset, loff_t length);
2280 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
2281 extern void ext4_set_inode_flags(struct inode *);
2282 extern void ext4_get_inode_flags(struct ext4_inode_info *);
2283 extern int ext4_alloc_da_blocks(struct inode *inode);
2284 extern void ext4_set_aops(struct inode *inode);
2285 extern int ext4_writepage_trans_blocks(struct inode *);
2286 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
2287 extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
2288                              loff_t lstart, loff_t lend);
2289 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2290 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
2291 extern void ext4_da_update_reserve_space(struct inode *inode,
2292                                         int used, int quota_claim);
2293
2294 /* indirect.c */
2295 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
2296                                 struct ext4_map_blocks *map, int flags);
2297 extern ssize_t ext4_ind_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
2298                                   loff_t offset);
2299 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
2300 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
2301 extern void ext4_ind_truncate(handle_t *, struct inode *inode);
2302 extern int ext4_ind_remove_space(handle_t *handle, struct inode *inode,
2303                                  ext4_lblk_t start, ext4_lblk_t end);
2304
2305 /* ioctl.c */
2306 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
2307 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
2308
2309 /* migrate.c */
2310 extern int ext4_ext_migrate(struct inode *);
2311 extern int ext4_ind_migrate(struct inode *inode);
2312
2313 /* namei.c */
2314 extern int ext4_dirent_csum_verify(struct inode *inode,
2315                                    struct ext4_dir_entry *dirent);
2316 extern int ext4_orphan_add(handle_t *, struct inode *);
2317 extern int ext4_orphan_del(handle_t *, struct inode *);
2318 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
2319                                 __u32 start_minor_hash, __u32 *next_hash);
2320 extern int search_dir(struct buffer_head *bh,
2321                       char *search_buf,
2322                       int buf_size,
2323                       struct inode *dir,
2324                       const struct qstr *d_name,
2325                       unsigned int offset,
2326                       struct ext4_dir_entry_2 **res_dir);
2327 extern int ext4_generic_delete_entry(handle_t *handle,
2328                                      struct inode *dir,
2329                                      struct ext4_dir_entry_2 *de_del,
2330                                      struct buffer_head *bh,
2331                                      void *entry_buf,
2332                                      int buf_size,
2333                                      int csum_size);
2334 extern int ext4_empty_dir(struct inode *inode);
2335
2336 /* resize.c */
2337 extern int ext4_group_add(struct super_block *sb,
2338                                 struct ext4_new_group_data *input);
2339 extern int ext4_group_extend(struct super_block *sb,
2340                                 struct ext4_super_block *es,
2341                                 ext4_fsblk_t n_blocks_count);
2342 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
2343
2344 /* super.c */
2345 extern int ext4_calculate_overhead(struct super_block *sb);
2346 extern void ext4_superblock_csum_set(struct super_block *sb);
2347 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
2348 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
2349 extern int ext4_alloc_flex_bg_array(struct super_block *sb,
2350                                     ext4_group_t ngroup);
2351 extern const char *ext4_decode_error(struct super_block *sb, int errno,
2352                                      char nbuf[16]);
2353
2354 extern __printf(4, 5)
2355 void __ext4_error(struct super_block *, const char *, unsigned int,
2356                   const char *, ...);
2357 extern __printf(5, 6)
2358 void __ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
2359                       const char *, ...);
2360 extern __printf(5, 6)
2361 void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
2362                      const char *, ...);
2363 extern void __ext4_std_error(struct super_block *, const char *,
2364                              unsigned int, int);
2365 extern __printf(4, 5)
2366 void __ext4_abort(struct super_block *, const char *, unsigned int,
2367                   const char *, ...);
2368 extern __printf(4, 5)
2369 void __ext4_warning(struct super_block *, const char *, unsigned int,
2370                     const char *, ...);
2371 extern __printf(3, 4)
2372 void __ext4_msg(struct super_block *, const char *, const char *, ...);
2373 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
2374                            const char *, unsigned int, const char *);
2375 extern __printf(7, 8)
2376 void __ext4_grp_locked_error(const char *, unsigned int,
2377                              struct super_block *, ext4_group_t,
2378                              unsigned long, ext4_fsblk_t,
2379                              const char *, ...);
2380
2381 #ifdef CONFIG_PRINTK
2382
2383 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2384         __ext4_error_inode(inode, func, line, block, fmt, ##__VA_ARGS__)
2385 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2386         __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__)
2387 #define ext4_error(sb, fmt, ...)                                        \
2388         __ext4_error(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2389 #define ext4_abort(sb, fmt, ...)                                        \
2390         __ext4_abort(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2391 #define ext4_warning(sb, fmt, ...)                                      \
2392         __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2393 #define ext4_msg(sb, level, fmt, ...)                           \
2394         __ext4_msg(sb, level, fmt, ##__VA_ARGS__)
2395 #define dump_mmp_msg(sb, mmp, msg)                                      \
2396         __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg)
2397 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2398         __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \
2399                                 fmt, ##__VA_ARGS__)
2400
2401 #else
2402
2403 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2404 do {                                                                    \
2405         no_printk(fmt, ##__VA_ARGS__);                                  \
2406         __ext4_error_inode(inode, "", 0, block, " ");                   \
2407 } while (0)
2408 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2409 do {                                                                    \
2410         no_printk(fmt, ##__VA_ARGS__);                                  \
2411         __ext4_error_file(file, "", 0, block, " ");                     \
2412 } while (0)
2413 #define ext4_error(sb, fmt, ...)                                        \
2414 do {                                                                    \
2415         no_printk(fmt, ##__VA_ARGS__);                                  \
2416         __ext4_error(sb, "", 0, " ");                                   \
2417 } while (0)
2418 #define ext4_abort(sb, fmt, ...)                                        \
2419 do {                                                                    \
2420         no_printk(fmt, ##__VA_ARGS__);                                  \
2421         __ext4_abort(sb, "", 0, " ");                                   \
2422 } while (0)
2423 #define ext4_warning(sb, fmt, ...)                                      \
2424 do {                                                                    \
2425         no_printk(fmt, ##__VA_ARGS__);                                  \
2426         __ext4_warning(sb, "", 0, " ");                                 \
2427 } while (0)
2428 #define ext4_msg(sb, level, fmt, ...)                                   \
2429 do {                                                                    \
2430         no_printk(fmt, ##__VA_ARGS__);                                  \
2431         __ext4_msg(sb, "", " ");                                        \
2432 } while (0)
2433 #define dump_mmp_msg(sb, mmp, msg)                                      \
2434         __dump_mmp_msg(sb, mmp, "", 0, "")
2435 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2436 do {                                                                    \
2437         no_printk(fmt, ##__VA_ARGS__);                          \
2438         __ext4_grp_locked_error("", 0, sb, grp, ino, block, " ");       \
2439 } while (0)
2440
2441 #endif
2442
2443 extern void ext4_update_dynamic_rev(struct super_block *sb);
2444 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
2445                                         __u32 compat);
2446 extern int ext4_update_rocompat_feature(handle_t *handle,
2447                                         struct super_block *sb, __u32 rocompat);
2448 extern int ext4_update_incompat_feature(handle_t *handle,
2449                                         struct super_block *sb, __u32 incompat);
2450 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
2451                                       struct ext4_group_desc *bg);
2452 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
2453                                       struct ext4_group_desc *bg);
2454 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
2455                                      struct ext4_group_desc *bg);
2456 extern __u32 ext4_free_group_clusters(struct super_block *sb,
2457                                       struct ext4_group_desc *bg);
2458 extern __u32 ext4_free_inodes_count(struct super_block *sb,
2459                                  struct ext4_group_desc *bg);
2460 extern __u32 ext4_used_dirs_count(struct super_block *sb,
2461                                 struct ext4_group_desc *bg);
2462 extern __u32 ext4_itable_unused_count(struct super_block *sb,
2463                                    struct ext4_group_desc *bg);
2464 extern void ext4_block_bitmap_set(struct super_block *sb,
2465                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2466 extern void ext4_inode_bitmap_set(struct super_block *sb,
2467                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2468 extern void ext4_inode_table_set(struct super_block *sb,
2469                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
2470 extern void ext4_free_group_clusters_set(struct super_block *sb,
2471                                          struct ext4_group_desc *bg,
2472                                          __u32 count);
2473 extern void ext4_free_inodes_set(struct super_block *sb,
2474                                 struct ext4_group_desc *bg, __u32 count);
2475 extern void ext4_used_dirs_set(struct super_block *sb,
2476                                 struct ext4_group_desc *bg, __u32 count);
2477 extern void ext4_itable_unused_set(struct super_block *sb,
2478                                    struct ext4_group_desc *bg, __u32 count);
2479 extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
2480                                        struct ext4_group_desc *gdp);
2481 extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
2482                                      struct ext4_group_desc *gdp);
2483 extern int ext4_register_li_request(struct super_block *sb,
2484                                     ext4_group_t first_not_zeroed);
2485
2486 static inline int ext4_has_group_desc_csum(struct super_block *sb)
2487 {
2488         return EXT4_HAS_RO_COMPAT_FEATURE(sb,
2489                                           EXT4_FEATURE_RO_COMPAT_GDT_CSUM) ||
2490                (EXT4_SB(sb)->s_chksum_driver != NULL);
2491 }
2492
2493 static inline int ext4_has_metadata_csum(struct super_block *sb)
2494 {
2495         WARN_ON_ONCE(EXT4_HAS_RO_COMPAT_FEATURE(sb,
2496                         EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
2497                      !EXT4_SB(sb)->s_chksum_driver);
2498
2499         return (EXT4_SB(sb)->s_chksum_driver != NULL);
2500 }
2501 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2502 {
2503         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2504                 le32_to_cpu(es->s_blocks_count_lo);
2505 }
2506
2507 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2508 {
2509         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2510                 le32_to_cpu(es->s_r_blocks_count_lo);
2511 }
2512
2513 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2514 {
2515         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2516                 le32_to_cpu(es->s_free_blocks_count_lo);
2517 }
2518
2519 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2520                                          ext4_fsblk_t blk)
2521 {
2522         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2523         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2524 }
2525
2526 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2527                                               ext4_fsblk_t blk)
2528 {
2529         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2530         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2531 }
2532
2533 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2534                                            ext4_fsblk_t blk)
2535 {
2536         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2537         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2538 }
2539
2540 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2541 {
2542         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2543                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2544                         le32_to_cpu(raw_inode->i_size_lo);
2545         else
2546                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2547 }
2548
2549 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2550 {
2551         raw_inode->i_size_lo = cpu_to_le32(i_size);
2552         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2553 }
2554
2555 static inline
2556 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2557                                             ext4_group_t group)
2558 {
2559          struct ext4_group_info ***grp_info;
2560          long indexv, indexh;
2561          BUG_ON(group >= EXT4_SB(sb)->s_groups_count);
2562          grp_info = EXT4_SB(sb)->s_group_info;
2563          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2564          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2565          return grp_info[indexv][indexh];
2566 }
2567
2568 /*
2569  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2570  * the locking protocol documented in the comments of ext4_group_add()
2571  * in resize.c
2572  */
2573 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2574 {
2575         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2576
2577         smp_rmb();
2578         return ngroups;
2579 }
2580
2581 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2582                                              ext4_group_t block_group)
2583 {
2584         return block_group >> sbi->s_log_groups_per_flex;
2585 }
2586
2587 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2588 {
2589         return 1 << sbi->s_log_groups_per_flex;
2590 }
2591
2592 #define ext4_std_error(sb, errno)                               \
2593 do {                                                            \
2594         if ((errno))                                            \
2595                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2596 } while (0)
2597
2598 #ifdef CONFIG_SMP
2599 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2600  * counters. So we need to make sure we have free clusters more
2601  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2602  */
2603 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2604 #else
2605 #define EXT4_FREECLUSTERS_WATERMARK 0
2606 #endif
2607
2608 /* Update i_disksize. Requires i_mutex to avoid races with truncate */
2609 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2610 {
2611         WARN_ON_ONCE(S_ISREG(inode->i_mode) &&
2612                      !mutex_is_locked(&inode->i_mutex));
2613         down_write(&EXT4_I(inode)->i_data_sem);
2614         if (newsize > EXT4_I(inode)->i_disksize)
2615                 EXT4_I(inode)->i_disksize = newsize;
2616         up_write(&EXT4_I(inode)->i_data_sem);
2617 }
2618
2619 /* Update i_size, i_disksize. Requires i_mutex to avoid races with truncate */
2620 static inline int ext4_update_inode_size(struct inode *inode, loff_t newsize)
2621 {
2622         int changed = 0;
2623
2624         if (newsize > inode->i_size) {
2625                 i_size_write(inode, newsize);
2626                 changed = 1;
2627         }
2628         if (newsize > EXT4_I(inode)->i_disksize) {
2629                 ext4_update_i_disksize(inode, newsize);
2630                 changed |= 2;
2631         }
2632         return changed;
2633 }
2634
2635 struct ext4_group_info {
2636         unsigned long   bb_state;
2637         struct rb_root  bb_free_root;
2638         ext4_grpblk_t   bb_first_free;  /* first free block */
2639         ext4_grpblk_t   bb_free;        /* total free blocks */
2640         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2641         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2642         struct          list_head bb_prealloc_list;
2643 #ifdef DOUBLE_CHECK
2644         void            *bb_bitmap;
2645 #endif
2646         struct rw_semaphore alloc_sem;
2647         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2648                                          * regions, index is order.
2649                                          * bb_counters[3] = 5 means
2650                                          * 5 free 8-block regions. */
2651 };
2652
2653 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2654 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2655 #define EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT     2
2656 #define EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT     3
2657
2658 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2659         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2660 #define EXT4_MB_GRP_BBITMAP_CORRUPT(grp)        \
2661         (test_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
2662 #define EXT4_MB_GRP_IBITMAP_CORRUPT(grp)        \
2663         (test_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
2664
2665 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2666         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2667 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2668         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2669 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2670         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2671
2672 #define EXT4_MAX_CONTENTION             8
2673 #define EXT4_CONTENTION_THRESHOLD       2
2674
2675 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2676                                               ext4_group_t group)
2677 {
2678         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2679 }
2680
2681 /*
2682  * Returns true if the filesystem is busy enough that attempts to
2683  * access the block group locks has run into contention.
2684  */
2685 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2686 {
2687         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2688 }
2689
2690 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2691 {
2692         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2693         if (spin_trylock(lock))
2694                 /*
2695                  * We're able to grab the lock right away, so drop the
2696                  * lock contention counter.
2697                  */
2698                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2699         else {
2700                 /*
2701                  * The lock is busy, so bump the contention counter,
2702                  * and then wait on the spin lock.
2703                  */
2704                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2705                                   EXT4_MAX_CONTENTION);
2706                 spin_lock(lock);
2707         }
2708 }
2709
2710 static inline void ext4_unlock_group(struct super_block *sb,
2711                                         ext4_group_t group)
2712 {
2713         spin_unlock(ext4_group_lock_ptr(sb, group));
2714 }
2715
2716 /*
2717  * Block validity checking
2718  */
2719 #define ext4_check_indirect_blockref(inode, bh)                         \
2720         ext4_check_blockref(__func__, __LINE__, inode,                  \
2721                             (__le32 *)(bh)->b_data,                     \
2722                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2723
2724 #define ext4_ind_check_inode(inode)                                     \
2725         ext4_check_blockref(__func__, __LINE__, inode,                  \
2726                             EXT4_I(inode)->i_data,                      \
2727                             EXT4_NDIR_BLOCKS)
2728
2729 /*
2730  * Inodes and files operations
2731  */
2732
2733 /* dir.c */
2734 extern const struct file_operations ext4_dir_operations;
2735
2736 /* file.c */
2737 extern const struct inode_operations ext4_file_inode_operations;
2738 extern const struct file_operations ext4_file_operations;
2739 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2740
2741 /* inline.c */
2742 extern int ext4_get_max_inline_size(struct inode *inode);
2743 extern int ext4_find_inline_data_nolock(struct inode *inode);
2744 extern int ext4_init_inline_data(handle_t *handle, struct inode *inode,
2745                                  unsigned int len);
2746 extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
2747
2748 extern int ext4_readpage_inline(struct inode *inode, struct page *page);
2749 extern int ext4_try_to_write_inline_data(struct address_space *mapping,
2750                                          struct inode *inode,
2751                                          loff_t pos, unsigned len,
2752                                          unsigned flags,
2753                                          struct page **pagep);
2754 extern int ext4_write_inline_data_end(struct inode *inode,
2755                                       loff_t pos, unsigned len,
2756                                       unsigned copied,
2757                                       struct page *page);
2758 extern struct buffer_head *
2759 ext4_journalled_write_inline_data(struct inode *inode,
2760                                   unsigned len,
2761                                   struct page *page);
2762 extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
2763                                            struct inode *inode,
2764                                            loff_t pos, unsigned len,
2765                                            unsigned flags,
2766                                            struct page **pagep,
2767                                            void **fsdata);
2768 extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
2769                                          unsigned len, unsigned copied,
2770                                          struct page *page);
2771 extern int ext4_try_add_inline_entry(handle_t *handle, struct dentry *dentry,
2772                                      struct inode *inode);
2773 extern int ext4_try_create_inline_dir(handle_t *handle,
2774                                       struct inode *parent,
2775                                       struct inode *inode);
2776 extern int ext4_read_inline_dir(struct file *filp,
2777                                 struct dir_context *ctx,
2778                                 int *has_inline_data);
2779 extern int htree_inlinedir_to_tree(struct file *dir_file,
2780                                    struct inode *dir, ext4_lblk_t block,
2781                                    struct dx_hash_info *hinfo,
2782                                    __u32 start_hash, __u32 start_minor_hash,
2783                                    int *has_inline_data);
2784 extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
2785                                         const struct qstr *d_name,
2786                                         struct ext4_dir_entry_2 **res_dir,
2787                                         int *has_inline_data);
2788 extern int ext4_delete_inline_entry(handle_t *handle,
2789                                     struct inode *dir,
2790                                     struct ext4_dir_entry_2 *de_del,
2791                                     struct buffer_head *bh,
2792                                     int *has_inline_data);
2793 extern int empty_inline_dir(struct inode *dir, int *has_inline_data);
2794 extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
2795                                         struct ext4_dir_entry_2 **parent_de,
2796                                         int *retval);
2797 extern int ext4_inline_data_fiemap(struct inode *inode,
2798                                    struct fiemap_extent_info *fieinfo,
2799                                    int *has_inline, __u64 start, __u64 len);
2800 extern int ext4_try_to_evict_inline_data(handle_t *handle,
2801                                          struct inode *inode,
2802                                          int needed);
2803 extern void ext4_inline_data_truncate(struct inode *inode, int *has_inline);
2804
2805 extern int ext4_convert_inline_data(struct inode *inode);
2806
2807 static inline int ext4_has_inline_data(struct inode *inode)
2808 {
2809         return ext4_test_inode_flag(inode, EXT4_INODE_INLINE_DATA) &&
2810                EXT4_I(inode)->i_inline_off;
2811 }
2812
2813 /* namei.c */
2814 extern const struct inode_operations ext4_dir_inode_operations;
2815 extern const struct inode_operations ext4_special_inode_operations;
2816 extern struct dentry *ext4_get_parent(struct dentry *child);
2817 extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
2818                                  struct ext4_dir_entry_2 *de,
2819                                  int blocksize, int csum_size,
2820                                  unsigned int parent_ino, int dotdot_real_len);
2821 extern void initialize_dirent_tail(struct ext4_dir_entry_tail *t,
2822                                    unsigned int blocksize);
2823 extern int ext4_handle_dirty_dirent_node(handle_t *handle,
2824                                          struct inode *inode,
2825                                          struct buffer_head *bh);
2826 #define S_SHIFT 12
2827 static unsigned char ext4_type_by_mode[S_IFMT >> S_SHIFT] = {
2828         [S_IFREG >> S_SHIFT]    = EXT4_FT_REG_FILE,
2829         [S_IFDIR >> S_SHIFT]    = EXT4_FT_DIR,
2830         [S_IFCHR >> S_SHIFT]    = EXT4_FT_CHRDEV,
2831         [S_IFBLK >> S_SHIFT]    = EXT4_FT_BLKDEV,
2832         [S_IFIFO >> S_SHIFT]    = EXT4_FT_FIFO,
2833         [S_IFSOCK >> S_SHIFT]   = EXT4_FT_SOCK,
2834         [S_IFLNK >> S_SHIFT]    = EXT4_FT_SYMLINK,
2835 };
2836
2837 static inline void ext4_set_de_type(struct super_block *sb,
2838                                 struct ext4_dir_entry_2 *de,
2839                                 umode_t mode) {
2840         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE))
2841                 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
2842 }
2843
2844 /* readpages.c */
2845 extern int ext4_mpage_readpages(struct address_space *mapping,
2846                                 struct list_head *pages, struct page *page,
2847                                 unsigned nr_pages);
2848
2849 /* symlink.c */
2850 extern const struct inode_operations ext4_symlink_inode_operations;
2851 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2852
2853 /* block_validity */
2854 extern void ext4_release_system_zone(struct super_block *sb);
2855 extern int ext4_setup_system_zone(struct super_block *sb);
2856 extern int __init ext4_init_system_zone(void);
2857 extern void ext4_exit_system_zone(void);
2858 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2859                                  ext4_fsblk_t start_blk,
2860                                  unsigned int count);
2861 extern int ext4_check_blockref(const char *, unsigned int,
2862                                struct inode *, __le32 *, unsigned int);
2863
2864 /* extents.c */
2865 struct ext4_ext_path;
2866 struct ext4_extent;
2867
2868 /*
2869  * Maximum number of logical blocks in a file; ext4_extent's ee_block is
2870  * __le32.
2871  */
2872 #define EXT_MAX_BLOCKS  0xffffffff
2873
2874 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2875 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2876 extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
2877 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2878                                struct ext4_map_blocks *map, int flags);
2879 extern void ext4_ext_truncate(handle_t *, struct inode *);
2880 extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
2881                                  ext4_lblk_t end);
2882 extern void ext4_ext_init(struct super_block *);
2883 extern void ext4_ext_release(struct super_block *);
2884 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2885                           loff_t len);
2886 extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
2887                                           loff_t offset, ssize_t len);
2888 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2889                            struct ext4_map_blocks *map, int flags);
2890 extern int ext4_ext_calc_metadata_amount(struct inode *inode,
2891                                          ext4_lblk_t lblocks);
2892 extern int ext4_extent_tree_init(handle_t *, struct inode *);
2893 extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
2894                                                    int num,
2895                                                    struct ext4_ext_path *path);
2896 extern int ext4_can_extents_be_merged(struct inode *inode,
2897                                       struct ext4_extent *ex1,
2898                                       struct ext4_extent *ex2);
2899 extern int ext4_ext_insert_extent(handle_t *, struct inode *,
2900                                   struct ext4_ext_path **,
2901                                   struct ext4_extent *, int);
2902 extern struct ext4_ext_path *ext4_find_extent(struct inode *, ext4_lblk_t,
2903                                               struct ext4_ext_path **,
2904                                               int flags);
2905 extern void ext4_ext_drop_refs(struct ext4_ext_path *);
2906 extern int ext4_ext_check_inode(struct inode *inode);
2907 extern int ext4_find_delalloc_range(struct inode *inode,
2908                                     ext4_lblk_t lblk_start,
2909                                     ext4_lblk_t lblk_end);
2910 extern int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk);
2911 extern ext4_lblk_t ext4_ext_next_allocated_block(struct ext4_ext_path *path);
2912 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2913                         __u64 start, __u64 len);
2914 extern int ext4_ext_precache(struct inode *inode);
2915 extern int ext4_collapse_range(struct inode *inode, loff_t offset, loff_t len);
2916 extern int ext4_swap_extents(handle_t *handle, struct inode *inode1,
2917                                 struct inode *inode2, ext4_lblk_t lblk1,
2918                              ext4_lblk_t lblk2,  ext4_lblk_t count,
2919                              int mark_unwritten,int *err);
2920
2921 /* move_extent.c */
2922 extern void ext4_double_down_write_data_sem(struct inode *first,
2923                                             struct inode *second);
2924 extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
2925                                           struct inode *donor_inode);
2926 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2927                              __u64 start_orig, __u64 start_donor,
2928                              __u64 len, __u64 *moved_len);
2929
2930 /* page-io.c */
2931 extern int __init ext4_init_pageio(void);
2932 extern void ext4_exit_pageio(void);
2933 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2934 extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
2935 extern int ext4_put_io_end(ext4_io_end_t *io_end);
2936 extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
2937 extern void ext4_io_submit_init(struct ext4_io_submit *io,
2938                                 struct writeback_control *wbc);
2939 extern void ext4_end_io_rsv_work(struct work_struct *work);
2940 extern void ext4_io_submit(struct ext4_io_submit *io);
2941 extern int ext4_bio_write_page(struct ext4_io_submit *io,
2942                                struct page *page,
2943                                int len,
2944                                struct writeback_control *wbc,
2945                                bool keep_towrite);
2946
2947 /* mmp.c */
2948 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2949
2950 /*
2951  * Add new method to test whether block and inode bitmaps are properly
2952  * initialized. With uninit_bg reading the block from disk is not enough
2953  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2954  */
2955 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2956
2957 static inline int bitmap_uptodate(struct buffer_head *bh)
2958 {
2959         return (buffer_uptodate(bh) &&
2960                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2961 }
2962 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2963 {
2964         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2965 }
2966
2967 /*
2968  * Disable DIO read nolock optimization, so new dioreaders will be forced
2969  * to grab i_mutex
2970  */
2971 static inline void ext4_inode_block_unlocked_dio(struct inode *inode)
2972 {
2973         ext4_set_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2974         smp_mb();
2975 }
2976 static inline void ext4_inode_resume_unlocked_dio(struct inode *inode)
2977 {
2978         smp_mb();
2979         ext4_clear_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2980 }
2981
2982 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2983
2984 /* For ioend & aio unwritten conversion wait queues */
2985 #define EXT4_WQ_HASH_SZ         37
2986 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
2987                                             EXT4_WQ_HASH_SZ])
2988 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
2989                                              EXT4_WQ_HASH_SZ])
2990 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2991 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2992
2993 #define EXT4_RESIZING   0
2994 extern int ext4_resize_begin(struct super_block *sb);
2995 extern void ext4_resize_end(struct super_block *sb);
2996
2997 #endif  /* __KERNEL__ */
2998
2999 #endif  /* _EXT4_H */