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/version.h>
  30 #include <linux/wait.h>
  31 #include <linux/blockgroup_lock.h>
  32 #include <linux/percpu_counter.h>
  33 #include <linux/ratelimit.h>
  34 #include <crypto/hash.h>
  35 #ifdef CONFIG_EXT4_FS_ENCRYPTION
  36 #include <linux/fscrypt_supp.h>
  37 #else
  38 #include <linux/fscrypt_notsupp.h>
  39 #endif
  40 #include <linux/falloc.h>
  41 #include <linux/percpu-rwsem.h>
  42 #ifdef __KERNEL__
  43 #include <linux/compat.h>
  44 #endif
  45
  46 /*
  47  * The fourth extended filesystem constants/structures
  48  */
  49
  50 /*
  51  * with AGGRESSIVE_CHECK allocator runs consistency checks over
  52  * structures. these checks slow things down a lot
  53  */
  54 #define AGGRESSIVE_CHECK__
  55
  56 /*
  57  * with DOUBLE_CHECK defined mballoc creates persistent in-core
  58  * bitmaps, maintains and uses them to check for double allocations
  59  */
  60 #define DOUBLE_CHECK__
  61
  62 /*
  63  * Define EXT4FS_DEBUG to produce debug messages
  64  */
  65 #undef EXT4FS_DEBUG
  66
  67 /*
  68  * Debug code
  69  */
  70 #ifdef EXT4FS_DEBUG
  71 #define ext4_debug(f, a...)                                             \
  72         do {                                                            \
  73                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
  74                         __FILE__, __LINE__, __func__);                  \
  75                 printk(KERN_DEBUG f, ## a);                             \
  76         } while (0)
  77 #else
  78 #define ext4_debug(fmt, ...)    no_printk(fmt, ##__VA_ARGS__)
  79 #endif
  80
  81 /*
  82  * Turn on EXT_DEBUG to get lots of info about extents operations.
  83  */
  84 #define EXT_DEBUG__
  85 #ifdef EXT_DEBUG
  86 #define ext_debug(fmt, ...)     printk(fmt, ##__VA_ARGS__)
  87 #else
  88 #define ext_debug(fmt, ...)     no_printk(fmt, ##__VA_ARGS__)
  89 #endif
  90
  91 /* data type for block offset of block group */
  92 typedef int ext4_grpblk_t;
  93
  94 /* data type for filesystem-wide blocks number */
  95 typedef unsigned long long ext4_fsblk_t;
  96
  97 /* data type for file logical block number */
  98 typedef __u32 ext4_lblk_t;
  99
 100 /* data type for block group number */
 101 typedef unsigned int ext4_group_t;
 102
 103 enum SHIFT_DIRECTION {
 104         SHIFT_LEFT = 0,
 105         SHIFT_RIGHT,
 106 };
 107
 108 /*
 109  * Flags used in mballoc's allocation_context flags field.
 110  *
 111  * Also used to show what's going on for debugging purposes when the
 112  * flag field is exported via the traceport interface
 113  */
 114
 115 /* prefer goal again. length */
 116 #define EXT4_MB_HINT_MERGE              0x0001
 117 /* blocks already reserved */
 118 #define EXT4_MB_HINT_RESERVED           0x0002
 119 /* metadata is being allocated */
 120 #define EXT4_MB_HINT_METADATA           0x0004
 121 /* first blocks in the file */
 122 #define EXT4_MB_HINT_FIRST              0x0008
 123 /* search for the best chunk */
 124 #define EXT4_MB_HINT_BEST               0x0010
 125 /* data is being allocated */
 126 #define EXT4_MB_HINT_DATA               0x0020
 127 /* don't preallocate (for tails) */
 128 #define EXT4_MB_HINT_NOPREALLOC         0x0040
 129 /* allocate for locality group */
 130 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
 131 /* allocate goal blocks or none */
 132 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
 133 /* goal is meaningful */
 134 #define EXT4_MB_HINT_TRY_GOAL           0x0200
 135 /* blocks already pre-reserved by delayed allocation */
 136 #define EXT4_MB_DELALLOC_RESERVED       0x0400
 137 /* We are doing stream allocation */
 138 #define EXT4_MB_STREAM_ALLOC            0x0800
 139 /* Use reserved root blocks if needed */
 140 #define EXT4_MB_USE_ROOT_BLOCKS         0x1000
 141 /* Use blocks from reserved pool */
 142 #define EXT4_MB_USE_RESERVED            0x2000
 143
 144 struct ext4_allocation_request {
 145         /* target inode for block we're allocating */
 146         struct inode *inode;
 147         /* how many blocks we want to allocate */
 148         unsigned int len;
 149         /* logical block in target inode */
 150         ext4_lblk_t logical;
 151         /* the closest logical allocated block to the left */
 152         ext4_lblk_t lleft;
 153         /* the closest logical allocated block to the right */
 154         ext4_lblk_t lright;
 155         /* phys. target (a hint) */
 156         ext4_fsblk_t goal;
 157         /* phys. block for the closest logical allocated block to the left */
 158         ext4_fsblk_t pleft;
 159         /* phys. block for the closest logical allocated block to the right */
 160         ext4_fsblk_t pright;
 161         /* flags. see above EXT4_MB_HINT_* */
 162         unsigned int flags;
 163 };
 164
 165 /*
 166  * Logical to physical block mapping, used by ext4_map_blocks()
 167  *
 168  * This structure is used to pass requests into ext4_map_blocks() as
 169  * well as to store the information returned by ext4_map_blocks().  It
 170  * takes less room on the stack than a struct buffer_head.
 171  */
 172 #define EXT4_MAP_NEW            (1 << BH_New)
 173 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
 174 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
 175 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
 176 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
 177                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY)
 178
 179 struct ext4_map_blocks {
 180         ext4_fsblk_t m_pblk;
 181         ext4_lblk_t m_lblk;
 182         unsigned int m_len;
 183         unsigned int m_flags;
 184 };
 185
 186 /*
 187  * Flags for ext4_io_end->flags
 188  */
 189 #define EXT4_IO_END_UNWRITTEN   0x0001
 190
 191 /*
 192  * For converting unwritten extents on a work queue. 'handle' is used for
 193  * buffered writeback.
 194  */
 195 typedef struct ext4_io_end {
 196         struct list_head        list;           /* per-file finished IO list */
 197         handle_t                *handle;        /* handle reserved for extent
 198                                                  * conversion */
 199         struct inode            *inode;         /* file being written to */
 200         struct bio              *bio;           /* Linked list of completed
 201                                                  * bios covering the extent */
 202         unsigned int            flag;           /* unwritten or not */
 203         atomic_t                count;          /* reference counter */
 204         loff_t                  offset;         /* offset in the file */
 205         ssize_t                 size;           /* size of the extent */
 206 } ext4_io_end_t;
 207
 208 struct ext4_io_submit {
 209         struct writeback_control *io_wbc;
 210         struct bio              *io_bio;
 211         ext4_io_end_t           *io_end;
 212         sector_t                io_next_block;
 213 };
 214
 215 /*
 216  * Special inodes numbers
 217  */
 218 #define EXT4_BAD_INO             1      /* Bad blocks inode */
 219 #define EXT4_ROOT_INO            2      /* Root inode */
 220 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
 221 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
 222 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
 223 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
 224 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
 225 #define EXT4_JOURNAL_INO         8      /* Journal inode */
 226
 227 /* First non-reserved inode for old ext4 filesystems */
 228 #define EXT4_GOOD_OLD_FIRST_INO 11
 229
 230 /*
 231  * Maximal count of links to a file
 232  */
 233 #define EXT4_LINK_MAX           65000
 234
 235 /*
 236  * Macro-instructions used to manage several block sizes
 237  */
 238 #define EXT4_MIN_BLOCK_SIZE             1024
 239 #define EXT4_MAX_BLOCK_SIZE             65536
 240 #define EXT4_MIN_BLOCK_LOG_SIZE         10
 241 #define EXT4_MAX_BLOCK_LOG_SIZE         16
 242 #define EXT4_MAX_CLUSTER_LOG_SIZE       30
 243 #ifdef __KERNEL__
 244 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
 245 #else
 246 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
 247 #endif
 248 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
 249 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
 250                                          EXT4_SB(s)->s_cluster_bits)
 251 #ifdef __KERNEL__
 252 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
 253 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
 254 #else
 255 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
 256 #endif
 257 #ifdef __KERNEL__
 258 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
 259 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
 260 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
 261 #else
 262 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 263                                  EXT4_GOOD_OLD_INODE_SIZE : \
 264                                  (s)->s_inode_size)
 265 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 266                                  EXT4_GOOD_OLD_FIRST_INO : \
 267                                  (s)->s_first_ino)
 268 #endif
 269 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
 270 #define EXT4_MAX_BLOCKS(size, offset, blkbits) \
 271         ((EXT4_BLOCK_ALIGN(size + offset, blkbits) >> blkbits) - (offset >> \
 272                                                                   blkbits))
 273
 274 /* Translate a block number to a cluster number */
 275 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
 276 /* Translate a cluster number to a block number */
 277 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
 278 /* Translate # of blks to # of clusters */
 279 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
 280                                  (sbi)->s_cluster_bits)
 281 /* Mask out the low bits to get the starting block of the cluster */
 282 #define EXT4_PBLK_CMASK(s, pblk) ((pblk) &                              \
 283                                   ~((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
 284 #define EXT4_LBLK_CMASK(s, lblk) ((lblk) &                              \
 285                                   ~((ext4_lblk_t) (s)->s_cluster_ratio - 1))
 286 /* Get the cluster offset */
 287 #define EXT4_PBLK_COFF(s, pblk) ((pblk) &                               \
 288                                  ((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
 289 #define EXT4_LBLK_COFF(s, lblk) ((lblk) &                               \
 290                                  ((ext4_lblk_t) (s)->s_cluster_ratio - 1))
 291
 292 /*
 293  * Structure of a blocks group descriptor
 294  */
 295 struct ext4_group_desc
 296 {
 297         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
 298         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
 299         __le32  bg_inode_table_lo;      /* Inodes table block */
 300         __le16  bg_free_blocks_count_lo;/* Free blocks count */
 301         __le16  bg_free_inodes_count_lo;/* Free inodes count */
 302         __le16  bg_used_dirs_count_lo;  /* Directories count */
 303         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
 304         __le32  bg_exclude_bitmap_lo;   /* Exclude bitmap for snapshots */
 305         __le16  bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
 306         __le16  bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
 307         __le16  bg_itable_unused_lo;    /* Unused inodes count */
 308         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
 309         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
 310         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
 311         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
 312         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
 313         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
 314         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
 315         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
 316         __le32  bg_exclude_bitmap_hi;   /* Exclude bitmap block MSB */
 317         __le16  bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
 318         __le16  bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
 319         __u32   bg_reserved;
 320 };
 321
 322 #define EXT4_BG_INODE_BITMAP_CSUM_HI_END        \
 323         (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
 324          sizeof(__le16))
 325 #define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END        \
 326         (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
 327          sizeof(__le16))
 328
 329 /*
 330  * Structure of a flex block group info
 331  */
 332
 333 struct flex_groups {
 334         atomic64_t      free_clusters;
 335         atomic_t        free_inodes;
 336         atomic_t        used_dirs;
 337 };
 338
 339 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
 340 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
 341 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
 342
 343 /*
 344  * Macro-instructions used to manage group descriptors
 345  */
 346 #define EXT4_MIN_DESC_SIZE              32
 347 #define EXT4_MIN_DESC_SIZE_64BIT        64
 348 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
 349 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
 350 #ifdef __KERNEL__
 351 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
 352 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
 353 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
 354 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
 355 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
 356 #else
 357 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
 358 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
 359 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
 360 #endif
 361
 362 /*
 363  * Constants relative to the data blocks
 364  */
 365 #define EXT4_NDIR_BLOCKS                12
 366 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
 367 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
 368 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
 369 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
 370
 371 /*
 372  * Inode flags
 373  */
 374 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
 375 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
 376 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
 377 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
 378 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
 379 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
 380 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
 381 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
 382 /* Reserved for compression usage... */
 383 #define EXT4_DIRTY_FL                   0x00000100
 384 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
 385 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
 386         /* nb: was previously EXT2_ECOMPR_FL */
 387 #define EXT4_ENCRYPT_FL                 0x00000800 /* encrypted file */
 388 /* End compression flags --- maybe not all used */
 389 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
 390 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
 391 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
 392 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
 393 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
 394 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
 395 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
 396 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
 397 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
 398 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
 399 #define EXT4_INLINE_DATA_FL             0x10000000 /* Inode has inline data. */
 400 #define EXT4_PROJINHERIT_FL             0x20000000 /* Create with parents projid */
 401 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
 402
 403 #define EXT4_FL_USER_VISIBLE            0x304BDFFF /* User visible flags */
 404 #define EXT4_FL_USER_MODIFIABLE         0x204BC0FF /* User modifiable flags */
 405
 406 /* Flags we can manipulate with through EXT4_IOC_FSSETXATTR */
 407 #define EXT4_FL_XFLAG_VISIBLE           (EXT4_SYNC_FL | \
 408                                          EXT4_IMMUTABLE_FL | \
 409                                          EXT4_APPEND_FL | \
 410                                          EXT4_NODUMP_FL | \
 411                                          EXT4_NOATIME_FL | \
 412                                          EXT4_PROJINHERIT_FL)
 413
 414 /* Flags that should be inherited by new inodes from their parent. */
 415 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
 416                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
 417                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
 418                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL |\
 419                            EXT4_PROJINHERIT_FL)
 420
 421 /* Flags that are appropriate for regular files (all but dir-specific ones). */
 422 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
 423
 424 /* Flags that are appropriate for non-directories/regular files. */
 425 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
 426
 427 /* Mask out flags that are inappropriate for the given type of inode. */
 428 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
 429 {
 430         if (S_ISDIR(mode))
 431                 return flags;
 432         else if (S_ISREG(mode))
 433                 return flags & EXT4_REG_FLMASK;
 434         else
 435                 return flags & EXT4_OTHER_FLMASK;
 436 }
 437
 438 /*
 439  * Inode flags used for atomic set/get
 440  */
 441 enum {
 442         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
 443         EXT4_INODE_UNRM         = 1,    /* Undelete */
 444         EXT4_INODE_COMPR        = 2,    /* Compress file */
 445         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
 446         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
 447         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
 448         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
 449         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
 450 /* Reserved for compression usage... */
 451         EXT4_INODE_DIRTY        = 8,
 452         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
 453         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
 454         EXT4_INODE_ENCRYPT      = 11,   /* Encrypted file */
 455 /* End compression flags --- maybe not all used */
 456         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
 457         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
 458         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
 459         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
 460         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
 461         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
 462         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
 463         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
 464         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
 465         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
 466         EXT4_INODE_INLINE_DATA  = 28,   /* Data in inode. */
 467         EXT4_INODE_PROJINHERIT  = 29,   /* Create with parents projid */
 468         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
 469 };
 470
 471 /*
 472  * Since it's pretty easy to mix up bit numbers and hex values, we use a
 473  * build-time check to make sure that EXT4_XXX_FL is consistent with respect to
 474  * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost
 475  * any extra space in the compiled kernel image, otherwise, the build will fail.
 476  * It's important that these values are the same, since we are using
 477  * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent
 478  * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk
 479  * values found in ext2, ext3 and ext4 filesystems, and of course the values
 480  * defined in e2fsprogs.
 481  *
 482  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
 483  */
 484 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
 485 #define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG))
 486
 487 static inline void ext4_check_flag_values(void)
 488 {
 489         CHECK_FLAG_VALUE(SECRM);
 490         CHECK_FLAG_VALUE(UNRM);
 491         CHECK_FLAG_VALUE(COMPR);
 492         CHECK_FLAG_VALUE(SYNC);
 493         CHECK_FLAG_VALUE(IMMUTABLE);
 494         CHECK_FLAG_VALUE(APPEND);
 495         CHECK_FLAG_VALUE(NODUMP);
 496         CHECK_FLAG_VALUE(NOATIME);
 497         CHECK_FLAG_VALUE(DIRTY);
 498         CHECK_FLAG_VALUE(COMPRBLK);
 499         CHECK_FLAG_VALUE(NOCOMPR);
 500         CHECK_FLAG_VALUE(ENCRYPT);
 501         CHECK_FLAG_VALUE(INDEX);
 502         CHECK_FLAG_VALUE(IMAGIC);
 503         CHECK_FLAG_VALUE(JOURNAL_DATA);
 504         CHECK_FLAG_VALUE(NOTAIL);
 505         CHECK_FLAG_VALUE(DIRSYNC);
 506         CHECK_FLAG_VALUE(TOPDIR);
 507         CHECK_FLAG_VALUE(HUGE_FILE);
 508         CHECK_FLAG_VALUE(EXTENTS);
 509         CHECK_FLAG_VALUE(EA_INODE);
 510         CHECK_FLAG_VALUE(EOFBLOCKS);
 511         CHECK_FLAG_VALUE(INLINE_DATA);
 512         CHECK_FLAG_VALUE(PROJINHERIT);
 513         CHECK_FLAG_VALUE(RESERVED);
 514 }
 515
 516 /* Used to pass group descriptor data when online resize is done */
 517 struct ext4_new_group_input {
 518         __u32 group;            /* Group number for this data */
 519         __u64 block_bitmap;     /* Absolute block number of block bitmap */
 520         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
 521         __u64 inode_table;      /* Absolute block number of inode table start */
 522         __u32 blocks_count;     /* Total number of blocks in this group */
 523         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
 524         __u16 unused;
 525 };
 526
 527 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 528 struct compat_ext4_new_group_input {
 529         u32 group;
 530         compat_u64 block_bitmap;
 531         compat_u64 inode_bitmap;
 532         compat_u64 inode_table;
 533         u32 blocks_count;
 534         u16 reserved_blocks;
 535         u16 unused;
 536 };
 537 #endif
 538
 539 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
 540 struct ext4_new_group_data {
 541         __u32 group;
 542         __u64 block_bitmap;
 543         __u64 inode_bitmap;
 544         __u64 inode_table;
 545         __u32 blocks_count;
 546         __u16 reserved_blocks;
 547         __u16 unused;
 548         __u32 free_blocks_count;
 549 };
 550
 551 /* Indexes used to index group tables in ext4_new_group_data */
 552 enum {
 553         BLOCK_BITMAP = 0,       /* block bitmap */
 554         INODE_BITMAP,           /* inode bitmap */
 555         INODE_TABLE,            /* inode tables */
 556         GROUP_TABLE_COUNT,
 557 };
 558
 559 /*
 560  * Flags used by ext4_map_blocks()
 561  */
 562         /* Allocate any needed blocks and/or convert an unwritten
 563            extent to be an initialized ext4 */
 564 #define EXT4_GET_BLOCKS_CREATE                  0x0001
 565         /* Request the creation of an unwritten extent */
 566 #define EXT4_GET_BLOCKS_UNWRIT_EXT              0x0002
 567 #define EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT       (EXT4_GET_BLOCKS_UNWRIT_EXT|\
 568                                                  EXT4_GET_BLOCKS_CREATE)
 569         /* Caller is from the delayed allocation writeout path
 570          * finally doing the actual allocation of delayed blocks */
 571 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
 572         /* caller is from the direct IO path, request to creation of an
 573         unwritten extents if not allocated, split the unwritten
 574         extent if blocks has been preallocated already*/
 575 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
 576 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
 577 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
 578                                          EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
 579         /* Convert extent to initialized after IO complete */
 580 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
 581                                          EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
 582         /* Eventual metadata allocation (due to growing extent tree)
 583          * should not fail, so try to use reserved blocks for that.*/
 584 #define EXT4_GET_BLOCKS_METADATA_NOFAIL         0x0020
 585         /* Don't normalize allocation size (used for fallocate) */
 586 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
 587         /* Request will not result in inode size update (user for fallocate) */
 588 #define EXT4_GET_BLOCKS_KEEP_SIZE               0x0080
 589         /* Convert written extents to unwritten */
 590 #define EXT4_GET_BLOCKS_CONVERT_UNWRITTEN       0x0100
 591         /* Write zeros to newly created written extents */
 592 #define EXT4_GET_BLOCKS_ZERO                    0x0200
 593 #define EXT4_GET_BLOCKS_CREATE_ZERO             (EXT4_GET_BLOCKS_CREATE |\
 594                                         EXT4_GET_BLOCKS_ZERO)
 595         /* Caller will submit data before dropping transaction handle. This
 596          * allows jbd2 to avoid submitting data before commit. */
 597 #define EXT4_GET_BLOCKS_IO_SUBMIT               0x0400
 598
 599 /*
 600  * The bit position of these flags must not overlap with any of the
 601  * EXT4_GET_BLOCKS_*.  They are used by ext4_find_extent(),
 602  * read_extent_tree_block(), ext4_split_extent_at(),
 603  * ext4_ext_insert_extent(), and ext4_ext_create_new_leaf().
 604  * EXT4_EX_NOCACHE is used to indicate that the we shouldn't be
 605  * caching the extents when reading from the extent tree while a
 606  * truncate or punch hole operation is in progress.
 607  */
 608 #define EXT4_EX_NOCACHE                         0x40000000
 609 #define EXT4_EX_FORCE_CACHE                     0x20000000
 610
 611 /*
 612  * Flags used by ext4_free_blocks
 613  */
 614 #define EXT4_FREE_BLOCKS_METADATA       0x0001
 615 #define EXT4_FREE_BLOCKS_FORGET         0x0002
 616 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
 617 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
 618 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
 619 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
 620
 621 /*
 622  * ioctl commands
 623  */
 624 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
 625 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
 626 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
 627 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
 628 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
 629 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
 630 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
 631 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
 632 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
 633 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
 634 #define EXT4_IOC_MIGRATE                _IO('f', 9)
 635  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
 636  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
 637 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
 638 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
 639 #define EXT4_IOC_RESIZE_FS              _IOW('f', 16, __u64)
 640 #define EXT4_IOC_SWAP_BOOT              _IO('f', 17)
 641 #define EXT4_IOC_PRECACHE_EXTENTS       _IO('f', 18)
 642 #define EXT4_IOC_SET_ENCRYPTION_POLICY  FS_IOC_SET_ENCRYPTION_POLICY
 643 #define EXT4_IOC_GET_ENCRYPTION_PWSALT  FS_IOC_GET_ENCRYPTION_PWSALT
 644 #define EXT4_IOC_GET_ENCRYPTION_POLICY  FS_IOC_GET_ENCRYPTION_POLICY
 645
 646 #ifndef FS_IOC_FSGETXATTR
 647 /* Until the uapi changes get merged for project quota... */
 648
 649 #define FS_IOC_FSGETXATTR               _IOR('X', 31, struct fsxattr)
 650 #define FS_IOC_FSSETXATTR               _IOW('X', 32, struct fsxattr)
 651
 652 /*
 653  * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
 654  */
 655 struct fsxattr {
 656         __u32           fsx_xflags;     /* xflags field value (get/set) */
 657         __u32           fsx_extsize;    /* extsize field value (get/set)*/
 658         __u32           fsx_nextents;   /* nextents field value (get)   */
 659         __u32           fsx_projid;     /* project identifier (get/set) */
 660         unsigned char   fsx_pad[12];
 661 };
 662
 663 /*
 664  * Flags for the fsx_xflags field
 665  */
 666 #define FS_XFLAG_REALTIME       0x00000001      /* data in realtime volume */
 667 #define FS_XFLAG_PREALLOC       0x00000002      /* preallocated file extents */
 668 #define FS_XFLAG_IMMUTABLE      0x00000008      /* file cannot be modified */
 669 #define FS_XFLAG_APPEND         0x00000010      /* all writes append */
 670 #define FS_XFLAG_SYNC           0x00000020      /* all writes synchronous */
 671 #define FS_XFLAG_NOATIME        0x00000040      /* do not update access time */
 672 #define FS_XFLAG_NODUMP         0x00000080      /* do not include in backups */
 673 #define FS_XFLAG_RTINHERIT      0x00000100      /* create with rt bit set */
 674 #define FS_XFLAG_PROJINHERIT    0x00000200      /* create with parents projid */
 675 #define FS_XFLAG_NOSYMLINKS     0x00000400      /* disallow symlink creation */
 676 #define FS_XFLAG_EXTSIZE        0x00000800      /* extent size allocator hint */
 677 #define FS_XFLAG_EXTSZINHERIT   0x00001000      /* inherit inode extent size */
 678 #define FS_XFLAG_NODEFRAG       0x00002000      /* do not defragment */
 679 #define FS_XFLAG_FILESTREAM     0x00004000      /* use filestream allocator */
 680 #define FS_XFLAG_HASATTR        0x80000000      /* no DIFLAG for this */
 681 #endif /* !defined(FS_IOC_FSGETXATTR) */
 682
 683 #define EXT4_IOC_FSGETXATTR             FS_IOC_FSGETXATTR
 684 #define EXT4_IOC_FSSETXATTR             FS_IOC_FSSETXATTR
 685
 686 #define EXT4_IOC_SHUTDOWN _IOR ('X', 125, __u32)
 687
 688 /*
 689  * Flags for going down operation
 690  */
 691 #define EXT4_GOING_FLAGS_DEFAULT                0x0     /* going down */
 692 #define EXT4_GOING_FLAGS_LOGFLUSH               0x1     /* flush log but not data */
 693 #define EXT4_GOING_FLAGS_NOLOGFLUSH             0x2     /* don't flush log nor data */
 694
 695
 696 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 697 /*
 698  * ioctl commands in 32 bit emulation
 699  */
 700 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
 701 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
 702 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
 703 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
 704 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
 705 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
 706 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
 707 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
 708 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
 709 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
 710 #endif
 711
 712 /* Max physical block we can address w/o extents */
 713 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
 714
 715 /*
 716  * Structure of an inode on the disk
 717  */
 718 struct ext4_inode {
 719         __le16  i_mode;         /* File mode */
 720         __le16  i_uid;          /* Low 16 bits of Owner Uid */
 721         __le32  i_size_lo;      /* Size in bytes */
 722         __le32  i_atime;        /* Access time */
 723         __le32  i_ctime;        /* Inode Change time */
 724         __le32  i_mtime;        /* Modification time */
 725         __le32  i_dtime;        /* Deletion Time */
 726         __le16  i_gid;          /* Low 16 bits of Group Id */
 727         __le16  i_links_count;  /* Links count */
 728         __le32  i_blocks_lo;    /* Blocks count */
 729         __le32  i_flags;        /* File flags */
 730         union {
 731                 struct {
 732                         __le32  l_i_version;
 733                 } linux1;
 734                 struct {
 735                         __u32  h_i_translator;
 736                 } hurd1;
 737                 struct {
 738                         __u32  m_i_reserved1;
 739                 } masix1;
 740         } osd1;                         /* OS dependent 1 */
 741         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
 742         __le32  i_generation;   /* File version (for NFS) */
 743         __le32  i_file_acl_lo;  /* File ACL */
 744         __le32  i_size_high;
 745         __le32  i_obso_faddr;   /* Obsoleted fragment address */
 746         union {
 747                 struct {
 748                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
 749                         __le16  l_i_file_acl_high;
 750                         __le16  l_i_uid_high;   /* these 2 fields */
 751                         __le16  l_i_gid_high;   /* were reserved2[0] */
 752                         __le16  l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
 753                         __le16  l_i_reserved;
 754                 } linux2;
 755                 struct {
 756                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 757                         __u16   h_i_mode_high;
 758                         __u16   h_i_uid_high;
 759                         __u16   h_i_gid_high;
 760                         __u32   h_i_author;
 761                 } hurd2;
 762                 struct {
 763                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 764                         __le16  m_i_file_acl_high;
 765                         __u32   m_i_reserved2[2];
 766                 } masix2;
 767         } osd2;                         /* OS dependent 2 */
 768         __le16  i_extra_isize;
 769         __le16  i_checksum_hi;  /* crc32c(uuid+inum+inode) BE */
 770         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
 771         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
 772         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
 773         __le32  i_crtime;       /* File Creation time */
 774         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
 775         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
 776         __le32  i_projid;       /* Project ID */
 777 };
 778
 779 struct move_extent {
 780         __u32 reserved;         /* should be zero */
 781         __u32 donor_fd;         /* donor file descriptor */
 782         __u64 orig_start;       /* logical start offset in block for orig */
 783         __u64 donor_start;      /* logical start offset in block for donor */
 784         __u64 len;              /* block length to be moved */
 785         __u64 moved_len;        /* moved block length */
 786 };
 787
 788 #define EXT4_EPOCH_BITS 2
 789 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
 790 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
 791
 792 /*
 793  * Extended fields will fit into an inode if the filesystem was formatted
 794  * with large inodes (-I 256 or larger) and there are not currently any EAs
 795  * consuming all of the available space. For new inodes we always reserve
 796  * enough space for the kernel's known extended fields, but for inodes
 797  * created with an old kernel this might not have been the case. None of
 798  * the extended inode fields is critical for correct filesystem operation.
 799  * This macro checks if a certain field fits in the inode. Note that
 800  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
 801  */
 802 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
 803         ((offsetof(typeof(*ext4_inode), field) +        \
 804           sizeof((ext4_inode)->field))                  \
 805         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
 806             (einode)->i_extra_isize))                   \
 807
 808 /*
 809  * We use an encoding that preserves the times for extra epoch "00":
 810  *
 811  * extra  msb of                         adjust for signed
 812  * epoch  32-bit                         32-bit tv_sec to
 813  * bits   time    decoded 64-bit tv_sec  64-bit tv_sec      valid time range
 814  * 0 0    1    -0x80000000..-0x00000001  0x000000000 1901-12-13..1969-12-31
 815  * 0 0    0    0x000000000..0x07fffffff  0x000000000 1970-01-01..2038-01-19
 816  * 0 1    1    0x080000000..0x0ffffffff  0x100000000 2038-01-19..2106-02-07
 817  * 0 1    0    0x100000000..0x17fffffff  0x100000000 2106-02-07..2174-02-25
 818  * 1 0    1    0x180000000..0x1ffffffff  0x200000000 2174-02-25..2242-03-16
 819  * 1 0    0    0x200000000..0x27fffffff  0x200000000 2242-03-16..2310-04-04
 820  * 1 1    1    0x280000000..0x2ffffffff  0x300000000 2310-04-04..2378-04-22
 821  * 1 1    0    0x300000000..0x37fffffff  0x300000000 2378-04-22..2446-05-10
 822  *
 823  * Note that previous versions of the kernel on 64-bit systems would
 824  * incorrectly use extra epoch bits 1,1 for dates between 1901 and
 825  * 1970.  e2fsck will correct this, assuming that it is run on the
 826  * affected filesystem before 2242.
 827  */
 828
 829 static inline __le32 ext4_encode_extra_time(struct timespec *time)
 830 {
 831         u32 extra = sizeof(time->tv_sec) > 4 ?
 832                 ((time->tv_sec - (s32)time->tv_sec) >> 32) & EXT4_EPOCH_MASK : 0;
 833         return cpu_to_le32(extra | (time->tv_nsec << EXT4_EPOCH_BITS));
 834 }
 835
 836 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
 837 {
 838         if (unlikely(sizeof(time->tv_sec) > 4 &&
 839                         (extra & cpu_to_le32(EXT4_EPOCH_MASK)))) {
 840 #if LINUX_VERSION_CODE < KERNEL_VERSION(4,20,0)
 841                 /* Handle legacy encoding of pre-1970 dates with epoch
 842                  * bits 1,1.  We assume that by kernel version 4.20,
 843                  * everyone will have run fsck over the affected
 844                  * filesystems to correct the problem.  (This
 845                  * backwards compatibility may be removed before this
 846                  * time, at the discretion of the ext4 developers.)
 847                  */
 848                 u64 extra_bits = le32_to_cpu(extra) & EXT4_EPOCH_MASK;
 849                 if (extra_bits == 3 && ((time->tv_sec) & 0x80000000) != 0)
 850                         extra_bits = 0;
 851                 time->tv_sec += extra_bits << 32;
 852 #else
 853                 time->tv_sec += (u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK) << 32;
 854 #endif
 855         }
 856         time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
 857 }
 858
 859 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
 860 do {                                                                           \
 861         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
 862         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 863                 (raw_inode)->xtime ## _extra =                                 \
 864                                 ext4_encode_extra_time(&(inode)->xtime);       \
 865 } while (0)
 866
 867 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
 868 do {                                                                           \
 869         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 870                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
 871         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 872                 (raw_inode)->xtime ## _extra =                                 \
 873                                 ext4_encode_extra_time(&(einode)->xtime);      \
 874 } while (0)
 875
 876 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
 877 do {                                                                           \
 878         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
 879         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 880                 ext4_decode_extra_time(&(inode)->xtime,                        \
 881                                        raw_inode->xtime ## _extra);            \
 882         else                                                                   \
 883                 (inode)->xtime.tv_nsec = 0;                                    \
 884 } while (0)
 885
 886 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
 887 do {                                                                           \
 888         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 889                 (einode)->xtime.tv_sec =                                       \
 890                         (signed)le32_to_cpu((raw_inode)->xtime);               \
 891         else                                                                   \
 892                 (einode)->xtime.tv_sec = 0;                                    \
 893         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 894                 ext4_decode_extra_time(&(einode)->xtime,                       \
 895                                        raw_inode->xtime ## _extra);            \
 896         else                                                                   \
 897                 (einode)->xtime.tv_nsec = 0;                                   \
 898 } while (0)
 899
 900 #define i_disk_version osd1.linux1.l_i_version
 901
 902 #if defined(__KERNEL__) || defined(__linux__)
 903 #define i_reserved1     osd1.linux1.l_i_reserved1
 904 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
 905 #define i_blocks_high   osd2.linux2.l_i_blocks_high
 906 #define i_uid_low       i_uid
 907 #define i_gid_low       i_gid
 908 #define i_uid_high      osd2.linux2.l_i_uid_high
 909 #define i_gid_high      osd2.linux2.l_i_gid_high
 910 #define i_checksum_lo   osd2.linux2.l_i_checksum_lo
 911
 912 #elif defined(__GNU__)
 913
 914 #define i_translator    osd1.hurd1.h_i_translator
 915 #define i_uid_high      osd2.hurd2.h_i_uid_high
 916 #define i_gid_high      osd2.hurd2.h_i_gid_high
 917 #define i_author        osd2.hurd2.h_i_author
 918
 919 #elif defined(__masix__)
 920
 921 #define i_reserved1     osd1.masix1.m_i_reserved1
 922 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
 923 #define i_reserved2     osd2.masix2.m_i_reserved2
 924
 925 #endif /* defined(__KERNEL__) || defined(__linux__) */
 926
 927 #include "extents_status.h"
 928
 929 /*
 930  * Lock subclasses for i_data_sem in the ext4_inode_info structure.
 931  *
 932  * These are needed to avoid lockdep false positives when we need to
 933  * allocate blocks to the quota inode during ext4_map_blocks(), while
 934  * holding i_data_sem for a normal (non-quota) inode.  Since we don't
 935  * do quota tracking for the quota inode, this avoids deadlock (as
 936  * well as infinite recursion, since it isn't turtles all the way
 937  * down...)
 938  *
 939  *  I_DATA_SEM_NORMAL - Used for most inodes
 940  *  I_DATA_SEM_OTHER  - Used by move_inode.c for the second normal inode
 941  *                        where the second inode has larger inode number
 942  *                        than the first
 943  *  I_DATA_SEM_QUOTA  - Used for quota inodes only
 944  */
 945 enum {
 946         I_DATA_SEM_NORMAL = 0,
 947         I_DATA_SEM_OTHER,
 948         I_DATA_SEM_QUOTA,
 949 };
 950
 951
 952 /*
 953  * fourth extended file system inode data in memory
 954  */
 955 struct ext4_inode_info {
 956         __le32  i_data[15];     /* unconverted */
 957         __u32   i_dtime;
 958         ext4_fsblk_t    i_file_acl;
 959
 960         /*
 961          * i_block_group is the number of the block group which contains
 962          * this file's inode.  Constant across the lifetime of the inode,
 963          * it is ued for making block allocation decisions - we try to
 964          * place a file's data blocks near its inode block, and new inodes
 965          * near to their parent directory's inode.
 966          */
 967         ext4_group_t    i_block_group;
 968         ext4_lblk_t     i_dir_start_lookup;
 969 #if (BITS_PER_LONG < 64)
 970         unsigned long   i_state_flags;          /* Dynamic state flags */
 971 #endif
 972         unsigned long   i_flags;
 973
 974         /*
 975          * Extended attributes can be read independently of the main file
 976          * data. Taking i_mutex even when reading would cause contention
 977          * between readers of EAs and writers of regular file data, so
 978          * instead we synchronize on xattr_sem when reading or changing
 979          * EAs.
 980          */
 981         struct rw_semaphore xattr_sem;
 982
 983         struct list_head i_orphan;      /* unlinked but open inodes */
 984
 985         /*
 986          * i_disksize keeps track of what the inode size is ON DISK, not
 987          * in memory.  During truncate, i_size is set to the new size by
 988          * the VFS prior to calling ext4_truncate(), but the filesystem won't
 989          * set i_disksize to 0 until the truncate is actually under way.
 990          *
 991          * The intent is that i_disksize always represents the blocks which
 992          * are used by this file.  This allows recovery to restart truncate
 993          * on orphans if we crash during truncate.  We actually write i_disksize
 994          * into the on-disk inode when writing inodes out, instead of i_size.
 995          *
 996          * The only time when i_disksize and i_size may be different is when
 997          * a truncate is in progress.  The only things which change i_disksize
 998          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
 999          */
1000         loff_t  i_disksize;
1001
1002         /*
1003          * i_data_sem is for serialising ext4_truncate() against
1004          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
1005          * data tree are chopped off during truncate. We can't do that in
1006          * ext4 because whenever we perform intermediate commits during
1007          * truncate, the inode and all the metadata blocks *must* be in a
1008          * consistent state which allows truncation of the orphans to restart
1009          * during recovery.  Hence we must fix the get_block-vs-truncate race
1010          * by other means, so we have i_data_sem.
1011          */
1012         struct rw_semaphore i_data_sem;
1013         /*
1014          * i_mmap_sem is for serializing page faults with truncate / punch hole
1015          * operations. We have to make sure that new page cannot be faulted in
1016          * a section of the inode that is being punched. We cannot easily use
1017          * i_data_sem for this since we need protection for the whole punch
1018          * operation and i_data_sem ranks below transaction start so we have
1019          * to occasionally drop it.
1020          */
1021         struct rw_semaphore i_mmap_sem;
1022         struct inode vfs_inode;
1023         struct jbd2_inode *jinode;
1024
1025         spinlock_t i_raw_lock;  /* protects updates to the raw inode */
1026
1027         /*
1028          * File creation time. Its function is same as that of
1029          * struct timespec i_{a,c,m}time in the generic inode.
1030          */
1031         struct timespec i_crtime;
1032
1033         /* mballoc */
1034         struct list_head i_prealloc_list;
1035         spinlock_t i_prealloc_lock;
1036
1037         /* extents status tree */
1038         struct ext4_es_tree i_es_tree;
1039         rwlock_t i_es_lock;
1040         struct list_head i_es_list;
1041         unsigned int i_es_all_nr;       /* protected by i_es_lock */
1042         unsigned int i_es_shk_nr;       /* protected by i_es_lock */
1043         ext4_lblk_t i_es_shrink_lblk;   /* Offset where we start searching for
1044                                            extents to shrink. Protected by
1045                                            i_es_lock  */
1046
1047         /* ialloc */
1048         ext4_group_t    i_last_alloc_group;
1049
1050         /* allocation reservation info for delalloc */
1051         /* In case of bigalloc, these refer to clusters rather than blocks */
1052         unsigned int i_reserved_data_blocks;
1053         unsigned int i_reserved_meta_blocks;
1054         unsigned int i_allocated_meta_blocks;
1055         ext4_lblk_t i_da_metadata_calc_last_lblock;
1056         int i_da_metadata_calc_len;
1057
1058         /* on-disk additional length */
1059         __u16 i_extra_isize;
1060
1061         /* Indicate the inline data space. */
1062         u16 i_inline_off;
1063         u16 i_inline_size;
1064
1065 #ifdef CONFIG_QUOTA
1066         /* quota space reservation, managed internally by quota code */
1067         qsize_t i_reserved_quota;
1068 #endif
1069
1070         /* Lock protecting lists below */
1071         spinlock_t i_completed_io_lock;
1072         /*
1073          * Completed IOs that need unwritten extents handling and have
1074          * transaction reserved
1075          */
1076         struct list_head i_rsv_conversion_list;
1077         struct work_struct i_rsv_conversion_work;
1078         atomic_t i_unwritten; /* Nr. of inflight conversions pending */
1079
1080         spinlock_t i_block_reservation_lock;
1081
1082         /*
1083          * Transactions that contain inode's metadata needed to complete
1084          * fsync and fdatasync, respectively.
1085          */
1086         tid_t i_sync_tid;
1087         tid_t i_datasync_tid;
1088
1089 #ifdef CONFIG_QUOTA
1090         struct dquot *i_dquot[MAXQUOTAS];
1091 #endif
1092
1093         /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
1094         __u32 i_csum_seed;
1095
1096         kprojid_t i_projid;
1097 };
1098
1099 /*
1100  * File system states
1101  */
1102 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
1103 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
1104 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
1105
1106 /*
1107  * Misc. filesystem flags
1108  */
1109 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
1110 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
1111 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
1112
1113 /*
1114  * Mount flags set via mount options or defaults
1115  */
1116 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
1117 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
1118 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
1119 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
1120 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
1121 #define EXT4_MOUNT_ERRORS_MASK          0x00070
1122 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
1123 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
1124 #ifdef CONFIG_FS_DAX
1125 #define EXT4_MOUNT_DAX                  0x00200 /* Direct Access */
1126 #else
1127 #define EXT4_MOUNT_DAX                  0
1128 #endif
1129 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
1130 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
1131 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
1132 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
1133 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
1134 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
1135 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
1136 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
1137 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
1138 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
1139 #define EXT4_MOUNT_QUOTA                0x40000 /* Some quota option set */
1140 #define EXT4_MOUNT_USRQUOTA             0x80000 /* "old" user quota,
1141                                                  * enable enforcement for hidden
1142                                                  * quota files */
1143 #define EXT4_MOUNT_GRPQUOTA             0x100000 /* "old" group quota, enable
1144                                                   * enforcement for hidden quota
1145                                                   * files */
1146 #define EXT4_MOUNT_PRJQUOTA             0x200000 /* Enable project quota
1147                                                   * enforcement */
1148 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
1149 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
1150 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
1151 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
1152 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
1153 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
1154 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
1155 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
1156
1157 /*
1158  * Mount flags set either automatically (could not be set by mount option)
1159  * based on per file system feature or property or in special cases such as
1160  * distinguishing between explicit mount option definition and default.
1161  */
1162 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
1163                                                       specified delalloc */
1164 #define EXT4_MOUNT2_STD_GROUP_SIZE      0x00000002 /* We have standard group
1165                                                       size of blocksize * 8
1166                                                       blocks */
1167 #define EXT4_MOUNT2_HURD_COMPAT         0x00000004 /* Support HURD-castrated
1168                                                       file systems */
1169
1170 #define EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM   0x00000008 /* User explicitly
1171                                                 specified journal checksum */
1172
1173 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
1174                                                 ~EXT4_MOUNT_##opt
1175 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
1176                                                 EXT4_MOUNT_##opt
1177 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
1178                                          EXT4_MOUNT_##opt)
1179
1180 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
1181                                                 ~EXT4_MOUNT2_##opt
1182 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
1183                                                 EXT4_MOUNT2_##opt
1184 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
1185                                          EXT4_MOUNT2_##opt)
1186
1187 #define ext4_test_and_set_bit           __test_and_set_bit_le
1188 #define ext4_set_bit                    __set_bit_le
1189 #define ext4_set_bit_atomic             ext2_set_bit_atomic
1190 #define ext4_test_and_clear_bit         __test_and_clear_bit_le
1191 #define ext4_clear_bit                  __clear_bit_le
1192 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
1193 #define ext4_test_bit                   test_bit_le
1194 #define ext4_find_next_zero_bit         find_next_zero_bit_le
1195 #define ext4_find_next_bit              find_next_bit_le
1196
1197 extern void ext4_set_bits(void *bm, int cur, int len);
1198
1199 /*
1200  * Maximal mount counts between two filesystem checks
1201  */
1202 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
1203 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
1204
1205 /*
1206  * Behaviour when detecting errors
1207  */
1208 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
1209 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
1210 #define EXT4_ERRORS_PANIC               3       /* Panic */
1211 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
1212
1213 /* Metadata checksum algorithm codes */
1214 #define EXT4_CRC32C_CHKSUM              1
1215
1216 /*
1217  * Structure of the super block
1218  */
1219 struct ext4_super_block {
1220 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1221         __le32  s_blocks_count_lo;      /* Blocks count */
1222         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1223         __le32  s_free_blocks_count_lo; /* Free blocks count */
1224 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1225         __le32  s_first_data_block;     /* First Data Block */
1226         __le32  s_log_block_size;       /* Block size */
1227         __le32  s_log_cluster_size;     /* Allocation cluster size */
1228 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1229         __le32  s_clusters_per_group;   /* # Clusters per group */
1230         __le32  s_inodes_per_group;     /* # Inodes per group */
1231         __le32  s_mtime;                /* Mount time */
1232 /*30*/  __le32  s_wtime;                /* Write time */
1233         __le16  s_mnt_count;            /* Mount count */
1234         __le16  s_max_mnt_count;        /* Maximal mount count */
1235         __le16  s_magic;                /* Magic signature */
1236         __le16  s_state;                /* File system state */
1237         __le16  s_errors;               /* Behaviour when detecting errors */
1238         __le16  s_minor_rev_level;      /* minor revision level */
1239 /*40*/  __le32  s_lastcheck;            /* time of last check */
1240         __le32  s_checkinterval;        /* max. time between checks */
1241         __le32  s_creator_os;           /* OS */
1242         __le32  s_rev_level;            /* Revision level */
1243 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1244         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1245         /*
1246          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1247          *
1248          * Note: the difference between the compatible feature set and
1249          * the incompatible feature set is that if there is a bit set
1250          * in the incompatible feature set that the kernel doesn't
1251          * know about, it should refuse to mount the filesystem.
1252          *
1253          * e2fsck's requirements are more strict; if it doesn't know
1254          * about a feature in either the compatible or incompatible
1255          * feature set, it must abort and not try to meddle with
1256          * things it doesn't understand...
1257          */
1258         __le32  s_first_ino;            /* First non-reserved inode */
1259         __le16  s_inode_size;           /* size of inode structure */
1260         __le16  s_block_group_nr;       /* block group # of this superblock */
1261         __le32  s_feature_compat;       /* compatible feature set */
1262 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1263         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1264 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1265 /*78*/  char    s_volume_name[16];      /* volume name */
1266 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1267 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1268         /*
1269          * Performance hints.  Directory preallocation should only
1270          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1271          */
1272         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1273         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1274         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1275         /*
1276          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1277          */
1278 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1279 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1280         __le32  s_journal_dev;          /* device number of journal file */
1281         __le32  s_last_orphan;          /* start of list of inodes to delete */
1282         __le32  s_hash_seed[4];         /* HTREE hash seed */
1283         __u8    s_def_hash_version;     /* Default hash version to use */
1284         __u8    s_jnl_backup_type;
1285         __le16  s_desc_size;            /* size of group descriptor */
1286 /*100*/ __le32  s_default_mount_opts;
1287         __le32  s_first_meta_bg;        /* First metablock block group */
1288         __le32  s_mkfs_time;            /* When the filesystem was created */
1289         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1290         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1291 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1292         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1293         __le32  s_free_blocks_count_hi; /* Free blocks count */
1294         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1295         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1296         __le32  s_flags;                /* Miscellaneous flags */
1297         __le16  s_raid_stride;          /* RAID stride */
1298         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1299         __le64  s_mmp_block;            /* Block for multi-mount protection */
1300         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1301         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1302         __u8    s_checksum_type;        /* metadata checksum algorithm used */
1303         __u8    s_encryption_level;     /* versioning level for encryption */
1304         __u8    s_reserved_pad;         /* Padding to next 32bits */
1305         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1306         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1307         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1308         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1309                                               snapshot's future use */
1310         __le32  s_snapshot_list;        /* inode number of the head of the
1311                                            on-disk snapshot list */
1312 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1313         __le32  s_error_count;          /* number of fs errors */
1314         __le32  s_first_error_time;     /* first time an error happened */
1315         __le32  s_first_error_ino;      /* inode involved in first error */
1316         __le64  s_first_error_block;    /* block involved of first error */
1317         __u8    s_first_error_func[32]; /* function where the error happened */
1318         __le32  s_first_error_line;     /* line number where error happened */
1319         __le32  s_last_error_time;      /* most recent time of an error */
1320         __le32  s_last_error_ino;       /* inode involved in last error */
1321         __le32  s_last_error_line;      /* line number where error happened */
1322         __le64  s_last_error_block;     /* block involved of last error */
1323         __u8    s_last_error_func[32];  /* function where the error happened */
1324 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1325         __u8    s_mount_opts[64];
1326         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1327         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1328         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1329         __le32  s_backup_bgs[2];        /* groups with sparse_super2 SBs */
1330         __u8    s_encrypt_algos[4];     /* Encryption algorithms in use  */
1331         __u8    s_encrypt_pw_salt[16];  /* Salt used for string2key algorithm */
1332         __le32  s_lpf_ino;              /* Location of the lost+found inode */
1333         __le32  s_prj_quota_inum;       /* inode for tracking project quota */
1334         __le32  s_checksum_seed;        /* crc32c(uuid) if csum_seed set */
1335         __le32  s_reserved[98];         /* Padding to the end of the block */
1336         __le32  s_checksum;             /* crc32c(superblock) */
1337 };
1338
1339 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1340
1341 #ifdef __KERNEL__
1342
1343 /*
1344  * run-time mount flags
1345  */
1346 #define EXT4_MF_MNTDIR_SAMPLED          0x0001
1347 #define EXT4_MF_FS_ABORTED              0x0002  /* Fatal error detected */
1348 #define EXT4_MF_TEST_DUMMY_ENCRYPTION   0x0004
1349
1350 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1351 #define DUMMY_ENCRYPTION_ENABLED(sbi) (unlikely((sbi)->s_mount_flags & \
1352                                                 EXT4_MF_TEST_DUMMY_ENCRYPTION))
1353 #else
1354 #define DUMMY_ENCRYPTION_ENABLED(sbi) (0)
1355 #endif
1356
1357 /* Number of quota types we support */
1358 #define EXT4_MAXQUOTAS 3
1359
1360 /*
1361  * fourth extended-fs super-block data in memory
1362  */
1363 struct ext4_sb_info {
1364         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1365         unsigned long s_inodes_per_block;/* Number of inodes per block */
1366         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1367         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1368         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1369         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1370         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1371         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1372         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1373         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1374         unsigned long s_overhead;  /* # of fs overhead clusters */
1375         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1376         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1377         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1378         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1379         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1380         struct buffer_head **s_group_desc;
1381         unsigned int s_mount_opt;
1382         unsigned int s_mount_opt2;
1383         unsigned int s_mount_flags;
1384         unsigned int s_def_mount_opt;
1385         ext4_fsblk_t s_sb_block;
1386         atomic64_t s_resv_clusters;
1387         kuid_t s_resuid;
1388         kgid_t s_resgid;
1389         unsigned short s_mount_state;
1390         unsigned short s_pad;
1391         int s_addr_per_block_bits;
1392         int s_desc_per_block_bits;
1393         int s_inode_size;
1394         int s_first_ino;
1395         unsigned int s_inode_readahead_blks;
1396         unsigned int s_inode_goal;
1397         spinlock_t s_next_gen_lock;
1398         u32 s_next_generation;
1399         u32 s_hash_seed[4];
1400         int s_def_hash_version;
1401         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1402         struct percpu_counter s_freeclusters_counter;
1403         struct percpu_counter s_freeinodes_counter;
1404         struct percpu_counter s_dirs_counter;
1405         struct percpu_counter s_dirtyclusters_counter;
1406         struct blockgroup_lock *s_blockgroup_lock;
1407         struct proc_dir_entry *s_proc;
1408         struct kobject s_kobj;
1409         struct completion s_kobj_unregister;
1410         struct super_block *s_sb;
1411
1412         /* Journaling */
1413         struct journal_s *s_journal;
1414         struct list_head s_orphan;
1415         struct mutex s_orphan_lock;
1416         unsigned long s_ext4_flags;             /* Ext4 superblock flags */
1417         unsigned long s_commit_interval;
1418         u32 s_max_batch_time;
1419         u32 s_min_batch_time;
1420         struct block_device *journal_bdev;
1421 #ifdef CONFIG_QUOTA
1422         char *s_qf_names[EXT4_MAXQUOTAS];       /* Names of quota files with journalled quota */
1423         int s_jquota_fmt;                       /* Format of quota to use */
1424 #endif
1425         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1426         struct rb_root system_blks;
1427
1428 #ifdef EXTENTS_STATS
1429         /* ext4 extents stats */
1430         unsigned long s_ext_min;
1431         unsigned long s_ext_max;
1432         unsigned long s_depth_max;
1433         spinlock_t s_ext_stats_lock;
1434         unsigned long s_ext_blocks;
1435         unsigned long s_ext_extents;
1436 #endif
1437
1438         /* for buddy allocator */
1439         struct ext4_group_info ***s_group_info;
1440         struct inode *s_buddy_cache;
1441         spinlock_t s_md_lock;
1442         unsigned short *s_mb_offsets;
1443         unsigned int *s_mb_maxs;
1444         unsigned int s_group_info_size;
1445         unsigned int s_mb_free_pending;
1446
1447         /* tunables */
1448         unsigned long s_stripe;
1449         unsigned int s_mb_stream_request;
1450         unsigned int s_mb_max_to_scan;
1451         unsigned int s_mb_min_to_scan;
1452         unsigned int s_mb_stats;
1453         unsigned int s_mb_order2_reqs;
1454         unsigned int s_mb_group_prealloc;
1455         unsigned int s_max_dir_size_kb;
1456         /* where last allocation was done - for stream allocation */
1457         unsigned long s_mb_last_group;
1458         unsigned long s_mb_last_start;
1459
1460         /* stats for buddy allocator */
1461         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1462         atomic_t s_bal_success; /* we found long enough chunks */
1463         atomic_t s_bal_allocated;       /* in blocks */
1464         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1465         atomic_t s_bal_goals;   /* goal hits */
1466         atomic_t s_bal_breaks;  /* too long searches */
1467         atomic_t s_bal_2orders; /* 2^order hits */
1468         spinlock_t s_bal_lock;
1469         unsigned long s_mb_buddies_generated;
1470         unsigned long long s_mb_generation_time;
1471         atomic_t s_mb_lost_chunks;
1472         atomic_t s_mb_preallocated;
1473         atomic_t s_mb_discarded;
1474         atomic_t s_lock_busy;
1475
1476         /* locality groups */
1477         struct ext4_locality_group __percpu *s_locality_groups;
1478
1479         /* for write statistics */
1480         unsigned long s_sectors_written_start;
1481         u64 s_kbytes_written;
1482
1483         /* the size of zero-out chunk */
1484         unsigned int s_extent_max_zeroout_kb;
1485
1486         unsigned int s_log_groups_per_flex;
1487         struct flex_groups *s_flex_groups;
1488         ext4_group_t s_flex_groups_allocated;
1489
1490         /* workqueue for reserved extent conversions (buffered io) */
1491         struct workqueue_struct *rsv_conversion_wq;
1492
1493         /* timer for periodic error stats printing */
1494         struct timer_list s_err_report;
1495
1496         /* Lazy inode table initialization info */
1497         struct ext4_li_request *s_li_request;
1498         /* Wait multiplier for lazy initialization thread */
1499         unsigned int s_li_wait_mult;
1500
1501         /* Kernel thread for multiple mount protection */
1502         struct task_struct *s_mmp_tsk;
1503
1504         /* record the last minlen when FITRIM is called. */
1505         atomic_t s_last_trim_minblks;
1506
1507         /* Reference to checksum algorithm driver via cryptoapi */
1508         struct crypto_shash *s_chksum_driver;
1509
1510         /* Precomputed FS UUID checksum for seeding other checksums */
1511         __u32 s_csum_seed;
1512
1513         /* Reclaim extents from extent status tree */
1514         struct shrinker s_es_shrinker;
1515         struct list_head s_es_list;     /* List of inodes with reclaimable extents */
1516         long s_es_nr_inode;
1517         struct ext4_es_stats s_es_stats;
1518         struct mb_cache *s_mb_cache;
1519         spinlock_t s_es_lock ____cacheline_aligned_in_smp;
1520
1521         /* Ratelimit ext4 messages. */
1522         struct ratelimit_state s_err_ratelimit_state;
1523         struct ratelimit_state s_warning_ratelimit_state;
1524         struct ratelimit_state s_msg_ratelimit_state;
1525
1526         /* Barrier between changing inodes' journal flags and writepages ops. */
1527         struct percpu_rw_semaphore s_journal_flag_rwsem;
1528 };
1529
1530 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1531 {
1532         return sb->s_fs_info;
1533 }
1534 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1535 {
1536         return container_of(inode, struct ext4_inode_info, vfs_inode);
1537 }
1538
1539 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1540 {
1541         return ino == EXT4_ROOT_INO ||
1542                 ino == EXT4_USR_QUOTA_INO ||
1543                 ino == EXT4_GRP_QUOTA_INO ||
1544                 ino == EXT4_BOOT_LOADER_INO ||
1545                 ino == EXT4_JOURNAL_INO ||
1546                 ino == EXT4_RESIZE_INO ||
1547                 (ino >= EXT4_FIRST_INO(sb) &&
1548                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1549 }
1550
1551 /*
1552  * Inode dynamic state flags
1553  */
1554 enum {
1555         EXT4_STATE_JDATA,               /* journaled data exists */
1556         EXT4_STATE_NEW,                 /* inode is newly created */
1557         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1558         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1559         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1560         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1561         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1562         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1563         EXT4_STATE_DIOREAD_LOCK,        /* Disable support for dio read
1564                                            nolocking */
1565         EXT4_STATE_MAY_INLINE_DATA,     /* may have in-inode data */
1566         EXT4_STATE_EXT_PRECACHED,       /* extents have been precached */
1567 };
1568
1569 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1570 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1571 {                                                                       \
1572         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1573 }                                                                       \
1574 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1575 {                                                                       \
1576         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1577 }                                                                       \
1578 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1579 {                                                                       \
1580         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1581 }
1582
1583 /* Add these declarations here only so that these functions can be
1584  * found by name.  Otherwise, they are very hard to locate. */
1585 static inline int ext4_test_inode_flag(struct inode *inode, int bit);
1586 static inline void ext4_set_inode_flag(struct inode *inode, int bit);
1587 static inline void ext4_clear_inode_flag(struct inode *inode, int bit);
1588 EXT4_INODE_BIT_FNS(flag, flags, 0)
1589
1590 /* Add these declarations here only so that these functions can be
1591  * found by name.  Otherwise, they are very hard to locate. */
1592 static inline int ext4_test_inode_state(struct inode *inode, int bit);
1593 static inline void ext4_set_inode_state(struct inode *inode, int bit);
1594 static inline void ext4_clear_inode_state(struct inode *inode, int bit);
1595 #if (BITS_PER_LONG < 64)
1596 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1597
1598 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1599 {
1600         (ei)->i_state_flags = 0;
1601 }
1602 #else
1603 EXT4_INODE_BIT_FNS(state, flags, 32)
1604
1605 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1606 {
1607         /* We depend on the fact that callers will set i_flags */
1608 }
1609 #endif
1610 #else
1611 /* Assume that user mode programs are passing in an ext4fs superblock, not
1612  * a kernel struct super_block.  This will allow us to call the feature-test
1613  * macros from user land. */
1614 #define EXT4_SB(sb)     (sb)
1615 #endif
1616
1617 /*
1618  * Returns true if the inode is inode is encrypted
1619  */
1620 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1621
1622 /*
1623  * Codes for operating systems
1624  */
1625 #define EXT4_OS_LINUX           0
1626 #define EXT4_OS_HURD            1
1627 #define EXT4_OS_MASIX           2
1628 #define EXT4_OS_FREEBSD         3
1629 #define EXT4_OS_LITES           4
1630
1631 /*
1632  * Revision levels
1633  */
1634 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1635 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1636
1637 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1638 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1639
1640 #define EXT4_GOOD_OLD_INODE_SIZE 128
1641
1642 /*
1643  * Feature set definitions
1644  */
1645
1646 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1647 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1648 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1649 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1650 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1651 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1652 #define EXT4_FEATURE_COMPAT_SPARSE_SUPER2       0x0200
1653
1654 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1655 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1656 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1657 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1658 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1659 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1660 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1661 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1662 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1663 /*
1664  * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM).  When
1665  * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
1666  * all other data structures' checksums.  However, the METADATA_CSUM and
1667  * GDT_CSUM bits are mutually exclusive.
1668  */
1669 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM    0x0400
1670 #define EXT4_FEATURE_RO_COMPAT_READONLY         0x1000
1671 #define EXT4_FEATURE_RO_COMPAT_PROJECT          0x2000
1672
1673 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1674 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1675 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1676 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1677 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1678 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1679 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1680 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1681 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1682 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1683 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1684 #define EXT4_FEATURE_INCOMPAT_CSUM_SEED         0x2000
1685 #define EXT4_FEATURE_INCOMPAT_LARGEDIR          0x4000 /* >2GB or 3-lvl htree */
1686 #define EXT4_FEATURE_INCOMPAT_INLINE_DATA       0x8000 /* data in inode */
1687 #define EXT4_FEATURE_INCOMPAT_ENCRYPT           0x10000
1688
1689 #define EXT4_FEATURE_COMPAT_FUNCS(name, flagname) \
1690 static inline bool ext4_has_feature_##name(struct super_block *sb) \
1691 { \
1692         return ((EXT4_SB(sb)->s_es->s_feature_compat & \
1693                 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname)) != 0); \
1694 } \
1695 static inline void ext4_set_feature_##name(struct super_block *sb) \
1696 { \
1697         EXT4_SB(sb)->s_es->s_feature_compat |= \
1698                 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
1699 } \
1700 static inline void ext4_clear_feature_##name(struct super_block *sb) \
1701 { \
1702         EXT4_SB(sb)->s_es->s_feature_compat &= \
1703                 ~cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
1704 }
1705
1706 #define EXT4_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
1707 static inline bool ext4_has_feature_##name(struct super_block *sb) \
1708 { \
1709         return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
1710                 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname)) != 0); \
1711 } \
1712 static inline void ext4_set_feature_##name(struct super_block *sb) \
1713 { \
1714         EXT4_SB(sb)->s_es->s_feature_ro_compat |= \
1715                 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
1716 } \
1717 static inline void ext4_clear_feature_##name(struct super_block *sb) \
1718 { \
1719         EXT4_SB(sb)->s_es->s_feature_ro_compat &= \
1720                 ~cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
1721 }
1722
1723 #define EXT4_FEATURE_INCOMPAT_FUNCS(name, flagname) \
1724 static inline bool ext4_has_feature_##name(struct super_block *sb) \
1725 { \
1726         return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
1727                 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname)) != 0); \
1728 } \
1729 static inline void ext4_set_feature_##name(struct super_block *sb) \
1730 { \
1731         EXT4_SB(sb)->s_es->s_feature_incompat |= \
1732                 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
1733 } \
1734 static inline void ext4_clear_feature_##name(struct super_block *sb) \
1735 { \
1736         EXT4_SB(sb)->s_es->s_feature_incompat &= \
1737                 ~cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
1738 }
1739
1740 EXT4_FEATURE_COMPAT_FUNCS(dir_prealloc,         DIR_PREALLOC)
1741 EXT4_FEATURE_COMPAT_FUNCS(imagic_inodes,        IMAGIC_INODES)
1742 EXT4_FEATURE_COMPAT_FUNCS(journal,              HAS_JOURNAL)
1743 EXT4_FEATURE_COMPAT_FUNCS(xattr,                EXT_ATTR)
1744 EXT4_FEATURE_COMPAT_FUNCS(resize_inode,         RESIZE_INODE)
1745 EXT4_FEATURE_COMPAT_FUNCS(dir_index,            DIR_INDEX)
1746 EXT4_FEATURE_COMPAT_FUNCS(sparse_super2,        SPARSE_SUPER2)
1747
1748 EXT4_FEATURE_RO_COMPAT_FUNCS(sparse_super,      SPARSE_SUPER)
1749 EXT4_FEATURE_RO_COMPAT_FUNCS(large_file,        LARGE_FILE)
1750 EXT4_FEATURE_RO_COMPAT_FUNCS(btree_dir,         BTREE_DIR)
1751 EXT4_FEATURE_RO_COMPAT_FUNCS(huge_file,         HUGE_FILE)
1752 EXT4_FEATURE_RO_COMPAT_FUNCS(gdt_csum,          GDT_CSUM)
1753 EXT4_FEATURE_RO_COMPAT_FUNCS(dir_nlink,         DIR_NLINK)
1754 EXT4_FEATURE_RO_COMPAT_FUNCS(extra_isize,       EXTRA_ISIZE)
1755 EXT4_FEATURE_RO_COMPAT_FUNCS(quota,             QUOTA)
1756 EXT4_FEATURE_RO_COMPAT_FUNCS(bigalloc,          BIGALLOC)
1757 EXT4_FEATURE_RO_COMPAT_FUNCS(metadata_csum,     METADATA_CSUM)
1758 EXT4_FEATURE_RO_COMPAT_FUNCS(readonly,          READONLY)
1759 EXT4_FEATURE_RO_COMPAT_FUNCS(project,           PROJECT)
1760
1761 EXT4_FEATURE_INCOMPAT_FUNCS(compression,        COMPRESSION)
1762 EXT4_FEATURE_INCOMPAT_FUNCS(filetype,           FILETYPE)
1763 EXT4_FEATURE_INCOMPAT_FUNCS(journal_needs_recovery,     RECOVER)
1764 EXT4_FEATURE_INCOMPAT_FUNCS(journal_dev,        JOURNAL_DEV)
1765 EXT4_FEATURE_INCOMPAT_FUNCS(meta_bg,            META_BG)
1766 EXT4_FEATURE_INCOMPAT_FUNCS(extents,            EXTENTS)
1767 EXT4_FEATURE_INCOMPAT_FUNCS(64bit,              64BIT)
1768 EXT4_FEATURE_INCOMPAT_FUNCS(mmp,                MMP)
1769 EXT4_FEATURE_INCOMPAT_FUNCS(flex_bg,            FLEX_BG)
1770 EXT4_FEATURE_INCOMPAT_FUNCS(ea_inode,           EA_INODE)
1771 EXT4_FEATURE_INCOMPAT_FUNCS(dirdata,            DIRDATA)
1772 EXT4_FEATURE_INCOMPAT_FUNCS(csum_seed,          CSUM_SEED)
1773 EXT4_FEATURE_INCOMPAT_FUNCS(largedir,           LARGEDIR)
1774 EXT4_FEATURE_INCOMPAT_FUNCS(inline_data,        INLINE_DATA)
1775 EXT4_FEATURE_INCOMPAT_FUNCS(encrypt,            ENCRYPT)
1776
1777 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1778 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1779                                          EXT4_FEATURE_INCOMPAT_META_BG)
1780 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1781                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1782                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1783
1784 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1785 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1786                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1787                                          EXT4_FEATURE_INCOMPAT_META_BG)
1788 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1789                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1790                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1791
1792 #define EXT4_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1793 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1794                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1795                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1796                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1797                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1798                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1799                                          EXT4_FEATURE_INCOMPAT_MMP | \
1800                                          EXT4_FEATURE_INCOMPAT_INLINE_DATA | \
1801                                          EXT4_FEATURE_INCOMPAT_ENCRYPT | \
1802                                          EXT4_FEATURE_INCOMPAT_CSUM_SEED)
1803 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1804                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1805                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1806                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1807                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1808                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1809                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1810                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
1811                                          EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
1812                                          EXT4_FEATURE_RO_COMPAT_QUOTA |\
1813                                          EXT4_FEATURE_RO_COMPAT_PROJECT)
1814
1815 #define EXTN_FEATURE_FUNCS(ver) \
1816 static inline bool ext4_has_unknown_ext##ver##_compat_features(struct super_block *sb) \
1817 { \
1818         return ((EXT4_SB(sb)->s_es->s_feature_compat & \
1819                 cpu_to_le32(~EXT##ver##_FEATURE_COMPAT_SUPP)) != 0); \
1820 } \
1821 static inline bool ext4_has_unknown_ext##ver##_ro_compat_features(struct super_block *sb) \
1822 { \
1823         return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
1824                 cpu_to_le32(~EXT##ver##_FEATURE_RO_COMPAT_SUPP)) != 0); \
1825 } \
1826 static inline bool ext4_has_unknown_ext##ver##_incompat_features(struct super_block *sb) \
1827 { \
1828         return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
1829                 cpu_to_le32(~EXT##ver##_FEATURE_INCOMPAT_SUPP)) != 0); \
1830 }
1831
1832 EXTN_FEATURE_FUNCS(2)
1833 EXTN_FEATURE_FUNCS(3)
1834 EXTN_FEATURE_FUNCS(4)
1835
1836 static inline bool ext4_has_compat_features(struct super_block *sb)
1837 {
1838         return (EXT4_SB(sb)->s_es->s_feature_compat != 0);
1839 }
1840 static inline bool ext4_has_ro_compat_features(struct super_block *sb)
1841 {
1842         return (EXT4_SB(sb)->s_es->s_feature_ro_compat != 0);
1843 }
1844 static inline bool ext4_has_incompat_features(struct super_block *sb)
1845 {
1846         return (EXT4_SB(sb)->s_es->s_feature_incompat != 0);
1847 }
1848
1849 /*
1850  * Superblock flags
1851  */
1852 #define EXT4_FLAGS_RESIZING     0
1853 #define EXT4_FLAGS_SHUTDOWN     1
1854
1855 static inline int ext4_forced_shutdown(struct ext4_sb_info *sbi)
1856 {
1857         return test_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
1858 }
1859
1860
1861 /*
1862  * Default values for user and/or group using reserved blocks
1863  */
1864 #define EXT4_DEF_RESUID         0
1865 #define EXT4_DEF_RESGID         0
1866
1867 /*
1868  * Default project ID
1869  */
1870 #define EXT4_DEF_PROJID         0
1871
1872 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1873
1874 /*
1875  * Default mount options
1876  */
1877 #define EXT4_DEFM_DEBUG         0x0001
1878 #define EXT4_DEFM_BSDGROUPS     0x0002
1879 #define EXT4_DEFM_XATTR_USER    0x0004
1880 #define EXT4_DEFM_ACL           0x0008
1881 #define EXT4_DEFM_UID16         0x0010
1882 #define EXT4_DEFM_JMODE         0x0060
1883 #define EXT4_DEFM_JMODE_DATA    0x0020
1884 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1885 #define EXT4_DEFM_JMODE_WBACK   0x0060
1886 #define EXT4_DEFM_NOBARRIER     0x0100
1887 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1888 #define EXT4_DEFM_DISCARD       0x0400
1889 #define EXT4_DEFM_NODELALLOC    0x0800
1890
1891 /*
1892  * Default journal batch times
1893  */
1894 #define EXT4_DEF_MIN_BATCH_TIME 0
1895 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1896
1897 /*
1898  * Minimum number of groups in a flexgroup before we separate out
1899  * directories into the first block group of a flexgroup
1900  */
1901 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1902
1903 /*
1904  * Structure of a directory entry
1905  */
1906 #define EXT4_NAME_LEN 255
1907
1908 struct ext4_dir_entry {
1909         __le32  inode;                  /* Inode number */
1910         __le16  rec_len;                /* Directory entry length */
1911         __le16  name_len;               /* Name length */
1912         char    name[EXT4_NAME_LEN];    /* File name */
1913 };
1914
1915 /*
1916  * The new version of the directory entry.  Since EXT4 structures are
1917  * stored in intel byte order, and the name_len field could never be
1918  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1919  * file_type field.
1920  */
1921 struct ext4_dir_entry_2 {
1922         __le32  inode;                  /* Inode number */
1923         __le16  rec_len;                /* Directory entry length */
1924         __u8    name_len;               /* Name length */
1925         __u8    file_type;
1926         char    name[EXT4_NAME_LEN];    /* File name */
1927 };
1928
1929 /*
1930  * This is a bogus directory entry at the end of each leaf block that
1931  * records checksums.
1932  */
1933 struct ext4_dir_entry_tail {
1934         __le32  det_reserved_zero1;     /* Pretend to be unused */
1935         __le16  det_rec_len;            /* 12 */
1936         __u8    det_reserved_zero2;     /* Zero name length */
1937         __u8    det_reserved_ft;        /* 0xDE, fake file type */
1938         __le32  det_checksum;           /* crc32c(uuid+inum+dirblock) */
1939 };
1940
1941 #define EXT4_DIRENT_TAIL(block, blocksize) \
1942         ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
1943                                         ((blocksize) - \
1944                                          sizeof(struct ext4_dir_entry_tail))))
1945
1946 /*
1947  * Ext4 directory file types.  Only the low 3 bits are used.  The
1948  * other bits are reserved for now.
1949  */
1950 #define EXT4_FT_UNKNOWN         0
1951 #define EXT4_FT_REG_FILE        1
1952 #define EXT4_FT_DIR             2
1953 #define EXT4_FT_CHRDEV          3
1954 #define EXT4_FT_BLKDEV          4
1955 #define EXT4_FT_FIFO            5
1956 #define EXT4_FT_SOCK            6
1957 #define EXT4_FT_SYMLINK         7
1958
1959 #define EXT4_FT_MAX             8
1960
1961 #define EXT4_FT_DIR_CSUM        0xDE
1962
1963 /*
1964  * EXT4_DIR_PAD defines the directory entries boundaries
1965  *
1966  * NOTE: It must be a multiple of 4
1967  */
1968 #define EXT4_DIR_PAD                    4
1969 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1970 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1971                                          ~EXT4_DIR_ROUND)
1972 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1973
1974 /*
1975  * If we ever get support for fs block sizes > page_size, we'll need
1976  * to remove the #if statements in the next two functions...
1977  */
1978 static inline unsigned int
1979 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1980 {
1981         unsigned len = le16_to_cpu(dlen);
1982
1983 #if (PAGE_SIZE >= 65536)
1984         if (len == EXT4_MAX_REC_LEN || len == 0)
1985                 return blocksize;
1986         return (len & 65532) | ((len & 3) << 16);
1987 #else
1988         return len;
1989 #endif
1990 }
1991
1992 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1993 {
1994         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1995                 BUG();
1996 #if (PAGE_SIZE >= 65536)
1997         if (len < 65536)
1998                 return cpu_to_le16(len);
1999         if (len == blocksize) {
2000                 if (blocksize == 65536)
2001                         return cpu_to_le16(EXT4_MAX_REC_LEN);
2002                 else
2003                         return cpu_to_le16(0);
2004         }
2005         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
2006 #else
2007         return cpu_to_le16(len);
2008 #endif
2009 }
2010
2011 /*
2012  * Hash Tree Directory indexing
2013  * (c) Daniel Phillips, 2001
2014  */
2015
2016 #define is_dx(dir) (ext4_has_feature_dir_index((dir)->i_sb) && \
2017                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
2018 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
2019 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
2020
2021 /* Legal values for the dx_root hash_version field: */
2022
2023 #define DX_HASH_LEGACY          0
2024 #define DX_HASH_HALF_MD4        1
2025 #define DX_HASH_TEA             2
2026 #define DX_HASH_LEGACY_UNSIGNED 3
2027 #define DX_HASH_HALF_MD4_UNSIGNED       4
2028 #define DX_HASH_TEA_UNSIGNED            5
2029
2030 static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
2031                               const void *address, unsigned int length)
2032 {
2033         struct {
2034                 struct shash_desc shash;
2035                 char ctx[4];
2036         } desc;
2037         int err;
2038
2039         BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
2040
2041         desc.shash.tfm = sbi->s_chksum_driver;
2042         desc.shash.flags = 0;
2043         *(u32 *)desc.ctx = crc;
2044
2045         err = crypto_shash_update(&desc.shash, address, length);
2046         BUG_ON(err);
2047
2048         return *(u32 *)desc.ctx;
2049 }
2050
2051 #ifdef __KERNEL__
2052
2053 /* hash info structure used by the directory hash */
2054 struct dx_hash_info
2055 {
2056         u32             hash;
2057         u32             minor_hash;
2058         int             hash_version;
2059         u32             *seed;
2060 };
2061
2062
2063 /* 32 and 64 bit signed EOF for dx directories */
2064 #define EXT4_HTREE_EOF_32BIT   ((1UL  << (32 - 1)) - 1)
2065 #define EXT4_HTREE_EOF_64BIT   ((1ULL << (64 - 1)) - 1)
2066
2067
2068 /*
2069  * Control parameters used by ext4_htree_next_block
2070  */
2071 #define HASH_NB_ALWAYS          1
2072
2073 struct ext4_filename {
2074         const struct qstr *usr_fname;
2075         struct fscrypt_str disk_name;
2076         struct dx_hash_info hinfo;
2077 #ifdef CONFIG_EXT4_FS_ENCRYPTION
2078         struct fscrypt_str crypto_buf;
2079 #endif
2080 };
2081
2082 #define fname_name(p) ((p)->disk_name.name)
2083 #define fname_len(p)  ((p)->disk_name.len)
2084
2085 /*
2086  * Describe an inode's exact location on disk and in memory
2087  */
2088 struct ext4_iloc
2089 {
2090         struct buffer_head *bh;
2091         unsigned long offset;
2092         ext4_group_t block_group;
2093 };
2094
2095 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
2096 {
2097         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
2098 }
2099
2100 /*
2101  * This structure is stuffed into the struct file's private_data field
2102  * for directories.  It is where we put information so that we can do
2103  * readdir operations in hash tree order.
2104  */
2105 struct dir_private_info {
2106         struct rb_root  root;
2107         struct rb_node  *curr_node;
2108         struct fname    *extra_fname;
2109         loff_t          last_pos;
2110         __u32           curr_hash;
2111         __u32           curr_minor_hash;
2112         __u32           next_hash;
2113 };
2114
2115 /* calculate the first block number of the group */
2116 static inline ext4_fsblk_t
2117 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
2118 {
2119         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
2120                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
2121 }
2122
2123 /*
2124  * Special error return code only used by dx_probe() and its callers.
2125  */
2126 #define ERR_BAD_DX_DIR  (-(MAX_ERRNO - 1))
2127
2128 /*
2129  * Timeout and state flag for lazy initialization inode thread.
2130  */
2131 #define EXT4_DEF_LI_WAIT_MULT                   10
2132 #define EXT4_DEF_LI_MAX_START_DELAY             5
2133 #define EXT4_LAZYINIT_QUIT                      0x0001
2134 #define EXT4_LAZYINIT_RUNNING                   0x0002
2135
2136 /*
2137  * Lazy inode table initialization info
2138  */
2139 struct ext4_lazy_init {
2140         unsigned long           li_state;
2141         struct list_head        li_request_list;
2142         struct mutex            li_list_mtx;
2143 };
2144
2145 struct ext4_li_request {
2146         struct super_block      *lr_super;
2147         struct ext4_sb_info     *lr_sbi;
2148         ext4_group_t            lr_next_group;
2149         struct list_head        lr_request;
2150         unsigned long           lr_next_sched;
2151         unsigned long           lr_timeout;
2152 };
2153
2154 struct ext4_features {
2155         struct kobject f_kobj;
2156         struct completion f_kobj_unregister;
2157 };
2158
2159 /*
2160  * This structure will be used for multiple mount protection. It will be
2161  * written into the block number saved in the s_mmp_block field in the
2162  * superblock. Programs that check MMP should assume that if
2163  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
2164  * to use the filesystem, regardless of how old the timestamp is.
2165  */
2166 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
2167 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
2168 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
2169 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
2170
2171 struct mmp_struct {
2172         __le32  mmp_magic;              /* Magic number for MMP */
2173         __le32  mmp_seq;                /* Sequence no. updated periodically */
2174
2175         /*
2176          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
2177          * purposes and do not affect the correctness of the algorithm
2178          */
2179         __le64  mmp_time;               /* Time last updated */
2180         char    mmp_nodename[64];       /* Node which last updated MMP block */
2181         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
2182
2183         /*
2184          * mmp_check_interval is used to verify if the MMP block has been
2185          * updated on the block device. The value is updated based on the
2186          * maximum time to write the MMP block during an update cycle.
2187          */
2188         __le16  mmp_check_interval;
2189
2190         __le16  mmp_pad1;
2191         __le32  mmp_pad2[226];
2192         __le32  mmp_checksum;           /* crc32c(uuid+mmp_block) */
2193 };
2194
2195 /* arguments passed to the mmp thread */
2196 struct mmpd_data {
2197         struct buffer_head *bh; /* bh from initial read_mmp_block() */
2198         struct super_block *sb;  /* super block of the fs */
2199 };
2200
2201 /*
2202  * Check interval multiplier
2203  * The MMP block is written every update interval and initially checked every
2204  * update interval x the multiplier (the value is then adapted based on the
2205  * write latency). The reason is that writes can be delayed under load and we
2206  * don't want readers to incorrectly assume that the filesystem is no longer
2207  * in use.
2208  */
2209 #define EXT4_MMP_CHECK_MULT             2UL
2210
2211 /*
2212  * Minimum interval for MMP checking in seconds.
2213  */
2214 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
2215
2216 /*
2217  * Maximum interval for MMP checking in seconds.
2218  */
2219 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
2220
2221 /*
2222  * Function prototypes
2223  */
2224
2225 /*
2226  * Ok, these declarations are also in <linux/kernel.h> but none of the
2227  * ext4 source programs needs to include it so they are duplicated here.
2228  */
2229 # define NORET_TYPE     /**/
2230 # define ATTRIB_NORET   __attribute__((noreturn))
2231 # define NORET_AND      noreturn,
2232
2233 /* bitmap.c */
2234 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
2235 void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
2236                                 struct ext4_group_desc *gdp,
2237                                 struct buffer_head *bh, int sz);
2238 int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2239                                   struct ext4_group_desc *gdp,
2240                                   struct buffer_head *bh, int sz);
2241 void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
2242                                 struct ext4_group_desc *gdp,
2243                                 struct buffer_head *bh);
2244 int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2245                                   struct ext4_group_desc *gdp,
2246                                   struct buffer_head *bh);
2247
2248 /* balloc.c */
2249 extern void ext4_get_group_no_and_offset(struct super_block *sb,
2250                                          ext4_fsblk_t blocknr,
2251                                          ext4_group_t *blockgrpp,
2252                                          ext4_grpblk_t *offsetp);
2253 extern ext4_group_t ext4_get_group_number(struct super_block *sb,
2254                                           ext4_fsblk_t block);
2255
2256 extern unsigned int ext4_block_group(struct super_block *sb,
2257                         ext4_fsblk_t blocknr);
2258 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
2259                         ext4_fsblk_t blocknr);
2260 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
2261 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
2262                         ext4_group_t group);
2263 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
2264                                          ext4_fsblk_t goal,
2265                                          unsigned int flags,
2266                                          unsigned long *count,
2267                                          int *errp);
2268 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
2269                                     s64 nclusters, unsigned int flags);
2270 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
2271 extern void ext4_check_blocks_bitmap(struct super_block *);
2272 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
2273                                                     ext4_group_t block_group,
2274                                                     struct buffer_head ** bh);
2275 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
2276
2277 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
2278                                                 ext4_group_t block_group);
2279 extern int ext4_wait_block_bitmap(struct super_block *sb,
2280                                   ext4_group_t block_group,
2281                                   struct buffer_head *bh);
2282 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
2283                                                   ext4_group_t block_group);
2284 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
2285                                               ext4_group_t block_group,
2286                                               struct ext4_group_desc *gdp);
2287 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
2288
2289 static inline bool ext4_encrypted_inode(struct inode *inode)
2290 {
2291         return ext4_test_inode_flag(inode, EXT4_INODE_ENCRYPT);
2292 }
2293
2294 #ifdef CONFIG_EXT4_FS_ENCRYPTION
2295 static inline int ext4_fname_setup_filename(struct inode *dir,
2296                         const struct qstr *iname,
2297                         int lookup, struct ext4_filename *fname)
2298 {
2299         struct fscrypt_name name;
2300         int err;
2301
2302         memset(fname, 0, sizeof(struct ext4_filename));
2303
2304         err = fscrypt_setup_filename(dir, iname, lookup, &name);
2305
2306         fname->usr_fname = name.usr_fname;
2307         fname->disk_name = name.disk_name;
2308         fname->hinfo.hash = name.hash;
2309         fname->hinfo.minor_hash = name.minor_hash;
2310         fname->crypto_buf = name.crypto_buf;
2311         return err;
2312 }
2313
2314 static inline void ext4_fname_free_filename(struct ext4_filename *fname)
2315 {
2316         struct fscrypt_name name;
2317
2318         name.crypto_buf = fname->crypto_buf;
2319         fscrypt_free_filename(&name);
2320
2321         fname->crypto_buf.name = NULL;
2322         fname->usr_fname = NULL;
2323         fname->disk_name.name = NULL;
2324 }
2325 #else
2326 static inline int ext4_fname_setup_filename(struct inode *dir,
2327                 const struct qstr *iname,
2328                 int lookup, struct ext4_filename *fname)
2329 {
2330         fname->usr_fname = iname;
2331         fname->disk_name.name = (unsigned char *) iname->name;
2332         fname->disk_name.len = iname->len;
2333         return 0;
2334 }
2335 static inline void ext4_fname_free_filename(struct ext4_filename *fname) { }
2336
2337 #endif
2338
2339 /* dir.c */
2340 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
2341                                   struct file *,
2342                                   struct ext4_dir_entry_2 *,
2343                                   struct buffer_head *, char *, int,
2344                                   unsigned int);
2345 #define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset)      \
2346         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
2347                                         (de), (bh), (buf), (size), (offset)))
2348 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
2349                                 __u32 minor_hash,
2350                                 struct ext4_dir_entry_2 *dirent,
2351                                 struct fscrypt_str *ent_name);
2352 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
2353 extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
2354                              struct buffer_head *bh,
2355                              void *buf, int buf_size,
2356                              struct ext4_filename *fname,
2357                              struct ext4_dir_entry_2 **dest_de);
2358 int ext4_insert_dentry(struct inode *dir,
2359                        struct inode *inode,
2360                        struct ext4_dir_entry_2 *de,
2361                        int buf_size,
2362                        struct ext4_filename *fname);
2363 static inline void ext4_update_dx_flag(struct inode *inode)
2364 {
2365         if (!ext4_has_feature_dir_index(inode->i_sb))
2366                 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
2367 }
2368 static unsigned char ext4_filetype_table[] = {
2369         DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
2370 };
2371
2372 static inline  unsigned char get_dtype(struct super_block *sb, int filetype)
2373 {
2374         if (!ext4_has_feature_filetype(sb) || filetype >= EXT4_FT_MAX)
2375                 return DT_UNKNOWN;
2376
2377         return ext4_filetype_table[filetype];
2378 }
2379 extern int ext4_check_all_de(struct inode *dir, struct buffer_head *bh,
2380                              void *buf, int buf_size);
2381
2382 /* fsync.c */
2383 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2384
2385 /* hash.c */
2386 extern int ext4fs_dirhash(const char *name, int len, struct
2387                           dx_hash_info *hinfo);
2388
2389 /* ialloc.c */
2390 extern struct inode *__ext4_new_inode(handle_t *, struct inode *, umode_t,
2391                                       const struct qstr *qstr, __u32 goal,
2392                                       uid_t *owner, int handle_type,
2393                                       unsigned int line_no, int nblocks);
2394
2395 #define ext4_new_inode(handle, dir, mode, qstr, goal, owner) \
2396         __ext4_new_inode((handle), (dir), (mode), (qstr), (goal), (owner), \
2397                          0, 0, 0)
2398 #define ext4_new_inode_start_handle(dir, mode, qstr, goal, owner, \
2399                                     type, nblocks)                  \
2400         __ext4_new_inode(NULL, (dir), (mode), (qstr), (goal), (owner), \
2401                          (type), __LINE__, (nblocks))
2402
2403
2404 extern void ext4_free_inode(handle_t *, struct inode *);
2405 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2406 extern unsigned long ext4_count_free_inodes(struct super_block *);
2407 extern unsigned long ext4_count_dirs(struct super_block *);
2408 extern void ext4_check_inodes_bitmap(struct super_block *);
2409 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2410 extern int ext4_init_inode_table(struct super_block *sb,
2411                                  ext4_group_t group, int barrier);
2412 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2413
2414 /* mballoc.c */
2415 extern const struct file_operations ext4_seq_mb_groups_fops;
2416 extern long ext4_mb_stats;
2417 extern long ext4_mb_max_to_scan;
2418 extern int ext4_mb_init(struct super_block *);
2419 extern int ext4_mb_release(struct super_block *);
2420 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2421                                 struct ext4_allocation_request *, int *);
2422 extern int ext4_mb_reserve_blocks(struct super_block *, int);
2423 extern void ext4_discard_preallocations(struct inode *);
2424 extern int __init ext4_init_mballoc(void);
2425 extern void ext4_exit_mballoc(void);
2426 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2427                              struct buffer_head *bh, ext4_fsblk_t block,
2428                              unsigned long count, int flags);
2429 extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2430                                    ext4_group_t ngroups);
2431 extern int ext4_mb_add_groupinfo(struct super_block *sb,
2432                 ext4_group_t i, struct ext4_group_desc *desc);
2433 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2434                                 ext4_fsblk_t block, unsigned long count);
2435 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2436
2437 /* inode.c */
2438 int ext4_inode_is_fast_symlink(struct inode *inode);
2439 struct buffer_head *ext4_getblk(handle_t *, struct inode *, ext4_lblk_t, int);
2440 struct buffer_head *ext4_bread(handle_t *, struct inode *, ext4_lblk_t, int);
2441 int ext4_get_block_unwritten(struct inode *inode, sector_t iblock,
2442                              struct buffer_head *bh_result, int create);
2443 int ext4_get_block(struct inode *inode, sector_t iblock,
2444                    struct buffer_head *bh_result, int create);
2445 int ext4_dio_get_block(struct inode *inode, sector_t iblock,
2446                        struct buffer_head *bh_result, int create);
2447 int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2448                            struct buffer_head *bh, int create);
2449 int ext4_walk_page_buffers(handle_t *handle,
2450                            struct buffer_head *head,
2451                            unsigned from,
2452                            unsigned to,
2453                            int *partial,
2454                            int (*fn)(handle_t *handle,
2455                                      struct buffer_head *bh));
2456 int do_journal_get_write_access(handle_t *handle,
2457                                 struct buffer_head *bh);
2458 #define FALL_BACK_TO_NONDELALLOC 1
2459 #define CONVERT_INLINE_DATA      2
2460
2461 extern struct inode *ext4_iget(struct super_block *, unsigned long);
2462 extern struct inode *ext4_iget_normal(struct super_block *, unsigned long);
2463 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
2464 extern int  ext4_setattr(struct dentry *, struct iattr *);
2465 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
2466                                 struct kstat *stat);
2467 extern void ext4_evict_inode(struct inode *);
2468 extern void ext4_clear_inode(struct inode *);
2469 extern int  ext4_sync_inode(handle_t *, struct inode *);
2470 extern void ext4_dirty_inode(struct inode *, int);
2471 extern int ext4_change_inode_journal_flag(struct inode *, int);
2472 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
2473 extern int ext4_inode_attach_jinode(struct inode *inode);
2474 extern int ext4_can_truncate(struct inode *inode);
2475 extern int ext4_truncate(struct inode *);
2476 extern int ext4_punch_hole(struct inode *inode, loff_t offset, loff_t length);
2477 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
2478 extern void ext4_set_inode_flags(struct inode *);
2479 extern void ext4_get_inode_flags(struct ext4_inode_info *);
2480 extern int ext4_alloc_da_blocks(struct inode *inode);
2481 extern void ext4_set_aops(struct inode *inode);
2482 extern int ext4_writepage_trans_blocks(struct inode *);
2483 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
2484 extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
2485                              loff_t lstart, loff_t lend);
2486 extern int ext4_page_mkwrite(struct vm_fault *vmf);
2487 extern int ext4_filemap_fault(struct vm_fault *vmf);
2488 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
2489 extern int ext4_get_projid(struct inode *inode, kprojid_t *projid);
2490 extern void ext4_da_update_reserve_space(struct inode *inode,
2491                                         int used, int quota_claim);
2492 extern int ext4_issue_zeroout(struct inode *inode, ext4_lblk_t lblk,
2493                               ext4_fsblk_t pblk, ext4_lblk_t len);
2494 extern int ext4_get_next_extent(struct inode *inode, ext4_lblk_t lblk,
2495                                 unsigned int map_len,
2496                                 struct extent_status *result);
2497
2498 /* indirect.c */
2499 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
2500                                 struct ext4_map_blocks *map, int flags);
2501 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
2502 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
2503 extern void ext4_ind_truncate(handle_t *, struct inode *inode);
2504 extern int ext4_ind_remove_space(handle_t *handle, struct inode *inode,
2505                                  ext4_lblk_t start, ext4_lblk_t end);
2506
2507 /* ioctl.c */
2508 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
2509 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
2510
2511 /* migrate.c */
2512 extern int ext4_ext_migrate(struct inode *);
2513 extern int ext4_ind_migrate(struct inode *inode);
2514
2515 /* namei.c */
2516 extern int ext4_dirent_csum_verify(struct inode *inode,
2517                                    struct ext4_dir_entry *dirent);
2518 extern int ext4_orphan_add(handle_t *, struct inode *);
2519 extern int ext4_orphan_del(handle_t *, struct inode *);
2520 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
2521                                 __u32 start_minor_hash, __u32 *next_hash);
2522 extern int ext4_search_dir(struct buffer_head *bh,
2523                            char *search_buf,
2524                            int buf_size,
2525                            struct inode *dir,
2526                            struct ext4_filename *fname,
2527                            const struct qstr *d_name,
2528                            unsigned int offset,
2529                            struct ext4_dir_entry_2 **res_dir);
2530 extern int ext4_generic_delete_entry(handle_t *handle,
2531                                      struct inode *dir,
2532                                      struct ext4_dir_entry_2 *de_del,
2533                                      struct buffer_head *bh,
2534                                      void *entry_buf,
2535                                      int buf_size,
2536                                      int csum_size);
2537 extern bool ext4_empty_dir(struct inode *inode);
2538
2539 /* resize.c */
2540 extern int ext4_group_add(struct super_block *sb,
2541                                 struct ext4_new_group_data *input);
2542 extern int ext4_group_extend(struct super_block *sb,
2543                                 struct ext4_super_block *es,
2544                                 ext4_fsblk_t n_blocks_count);
2545 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
2546
2547 /* super.c */
2548 extern int ext4_seq_options_show(struct seq_file *seq, void *offset);
2549 extern int ext4_calculate_overhead(struct super_block *sb);
2550 extern void ext4_superblock_csum_set(struct super_block *sb);
2551 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
2552 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
2553 extern int ext4_alloc_flex_bg_array(struct super_block *sb,
2554                                     ext4_group_t ngroup);
2555 extern const char *ext4_decode_error(struct super_block *sb, int errno,
2556                                      char nbuf[16]);
2557
2558 extern __printf(4, 5)
2559 void __ext4_error(struct super_block *, const char *, unsigned int,
2560                   const char *, ...);
2561 extern __printf(5, 6)
2562 void __ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
2563                       const char *, ...);
2564 extern __printf(5, 6)
2565 void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
2566                      const char *, ...);
2567 extern void __ext4_std_error(struct super_block *, const char *,
2568                              unsigned int, int);
2569 extern __printf(4, 5)
2570 void __ext4_abort(struct super_block *, const char *, unsigned int,
2571                   const char *, ...);
2572 extern __printf(4, 5)
2573 void __ext4_warning(struct super_block *, const char *, unsigned int,
2574                     const char *, ...);
2575 extern __printf(4, 5)
2576 void __ext4_warning_inode(const struct inode *inode, const char *function,
2577                           unsigned int line, const char *fmt, ...);
2578 extern __printf(3, 4)
2579 void __ext4_msg(struct super_block *, const char *, const char *, ...);
2580 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
2581                            const char *, unsigned int, const char *);
2582 extern __printf(7, 8)
2583 void __ext4_grp_locked_error(const char *, unsigned int,
2584                              struct super_block *, ext4_group_t,
2585                              unsigned long, ext4_fsblk_t,
2586                              const char *, ...);
2587
2588 #define EXT4_ERROR_INODE(inode, fmt, a...) \
2589         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
2590
2591 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...)                 \
2592         ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
2593
2594 #define EXT4_ERROR_FILE(file, block, fmt, a...)                         \
2595         ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
2596
2597 #ifdef CONFIG_PRINTK
2598
2599 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2600         __ext4_error_inode(inode, func, line, block, fmt, ##__VA_ARGS__)
2601 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2602         __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__)
2603 #define ext4_error(sb, fmt, ...)                                        \
2604         __ext4_error(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2605 #define ext4_abort(sb, fmt, ...)                                        \
2606         __ext4_abort(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2607 #define ext4_warning(sb, fmt, ...)                                      \
2608         __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2609 #define ext4_warning_inode(inode, fmt, ...)                             \
2610         __ext4_warning_inode(inode, __func__, __LINE__, fmt, ##__VA_ARGS__)
2611 #define ext4_msg(sb, level, fmt, ...)                           \
2612         __ext4_msg(sb, level, fmt, ##__VA_ARGS__)
2613 #define dump_mmp_msg(sb, mmp, msg)                                      \
2614         __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg)
2615 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2616         __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \
2617                                 fmt, ##__VA_ARGS__)
2618
2619 #else
2620
2621 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2622 do {                                                                    \
2623         no_printk(fmt, ##__VA_ARGS__);                                  \
2624         __ext4_error_inode(inode, "", 0, block, " ");                   \
2625 } while (0)
2626 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2627 do {                                                                    \
2628         no_printk(fmt, ##__VA_ARGS__);                                  \
2629         __ext4_error_file(file, "", 0, block, " ");                     \
2630 } while (0)
2631 #define ext4_error(sb, fmt, ...)                                        \
2632 do {                                                                    \
2633         no_printk(fmt, ##__VA_ARGS__);                                  \
2634         __ext4_error(sb, "", 0, " ");                                   \
2635 } while (0)
2636 #define ext4_abort(sb, fmt, ...)                                        \
2637 do {                                                                    \
2638         no_printk(fmt, ##__VA_ARGS__);                                  \
2639         __ext4_abort(sb, "", 0, " ");                                   \
2640 } while (0)
2641 #define ext4_warning(sb, fmt, ...)                                      \
2642 do {                                                                    \
2643         no_printk(fmt, ##__VA_ARGS__);                                  \
2644         __ext4_warning(sb, "", 0, " ");                                 \
2645 } while (0)
2646 #define ext4_warning_inode(inode, fmt, ...)                             \
2647 do {                                                                    \
2648         no_printk(fmt, ##__VA_ARGS__);                                  \
2649         __ext4_warning_inode(inode, "", 0, " ");                        \
2650 } while (0)
2651 #define ext4_msg(sb, level, fmt, ...)                                   \
2652 do {                                                                    \
2653         no_printk(fmt, ##__VA_ARGS__);                                  \
2654         __ext4_msg(sb, "", " ");                                        \
2655 } while (0)
2656 #define dump_mmp_msg(sb, mmp, msg)                                      \
2657         __dump_mmp_msg(sb, mmp, "", 0, "")
2658 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2659 do {                                                                    \
2660         no_printk(fmt, ##__VA_ARGS__);                          \
2661         __ext4_grp_locked_error("", 0, sb, grp, ino, block, " ");       \
2662 } while (0)
2663
2664 #endif
2665
2666 extern void ext4_update_dynamic_rev(struct super_block *sb);
2667 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
2668                                         __u32 compat);
2669 extern int ext4_update_rocompat_feature(handle_t *handle,
2670                                         struct super_block *sb, __u32 rocompat);
2671 extern int ext4_update_incompat_feature(handle_t *handle,
2672                                         struct super_block *sb, __u32 incompat);
2673 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
2674                                       struct ext4_group_desc *bg);
2675 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
2676                                       struct ext4_group_desc *bg);
2677 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
2678                                      struct ext4_group_desc *bg);
2679 extern __u32 ext4_free_group_clusters(struct super_block *sb,
2680                                       struct ext4_group_desc *bg);
2681 extern __u32 ext4_free_inodes_count(struct super_block *sb,
2682                                  struct ext4_group_desc *bg);
2683 extern __u32 ext4_used_dirs_count(struct super_block *sb,
2684                                 struct ext4_group_desc *bg);
2685 extern __u32 ext4_itable_unused_count(struct super_block *sb,
2686                                    struct ext4_group_desc *bg);
2687 extern void ext4_block_bitmap_set(struct super_block *sb,
2688                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2689 extern void ext4_inode_bitmap_set(struct super_block *sb,
2690                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2691 extern void ext4_inode_table_set(struct super_block *sb,
2692                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
2693 extern void ext4_free_group_clusters_set(struct super_block *sb,
2694                                          struct ext4_group_desc *bg,
2695                                          __u32 count);
2696 extern void ext4_free_inodes_set(struct super_block *sb,
2697                                 struct ext4_group_desc *bg, __u32 count);
2698 extern void ext4_used_dirs_set(struct super_block *sb,
2699                                 struct ext4_group_desc *bg, __u32 count);
2700 extern void ext4_itable_unused_set(struct super_block *sb,
2701                                    struct ext4_group_desc *bg, __u32 count);
2702 extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
2703                                        struct ext4_group_desc *gdp);
2704 extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
2705                                      struct ext4_group_desc *gdp);
2706 extern int ext4_register_li_request(struct super_block *sb,
2707                                     ext4_group_t first_not_zeroed);
2708
2709 static inline int ext4_has_group_desc_csum(struct super_block *sb)
2710 {
2711         return ext4_has_feature_gdt_csum(sb) ||
2712                EXT4_SB(sb)->s_chksum_driver != NULL;
2713 }
2714
2715 static inline int ext4_has_metadata_csum(struct super_block *sb)
2716 {
2717         WARN_ON_ONCE(ext4_has_feature_metadata_csum(sb) &&
2718                      !EXT4_SB(sb)->s_chksum_driver);
2719
2720         return (EXT4_SB(sb)->s_chksum_driver != NULL);
2721 }
2722 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2723 {
2724         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2725                 le32_to_cpu(es->s_blocks_count_lo);
2726 }
2727
2728 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2729 {
2730         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2731                 le32_to_cpu(es->s_r_blocks_count_lo);
2732 }
2733
2734 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2735 {
2736         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2737                 le32_to_cpu(es->s_free_blocks_count_lo);
2738 }
2739
2740 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2741                                          ext4_fsblk_t blk)
2742 {
2743         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2744         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2745 }
2746
2747 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2748                                               ext4_fsblk_t blk)
2749 {
2750         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2751         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2752 }
2753
2754 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2755                                            ext4_fsblk_t blk)
2756 {
2757         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2758         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2759 }
2760
2761 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2762 {
2763         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2764                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2765                         le32_to_cpu(raw_inode->i_size_lo);
2766         else
2767                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2768 }
2769
2770 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2771 {
2772         raw_inode->i_size_lo = cpu_to_le32(i_size);
2773         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2774 }
2775
2776 static inline
2777 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2778                                             ext4_group_t group)
2779 {
2780          struct ext4_group_info ***grp_info;
2781          long indexv, indexh;
2782          BUG_ON(group >= EXT4_SB(sb)->s_groups_count);
2783          grp_info = EXT4_SB(sb)->s_group_info;
2784          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2785          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2786          return grp_info[indexv][indexh];
2787 }
2788
2789 /*
2790  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2791  * the locking protocol documented in the comments of ext4_group_add()
2792  * in resize.c
2793  */
2794 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2795 {
2796         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2797
2798         smp_rmb();
2799         return ngroups;
2800 }
2801
2802 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2803                                              ext4_group_t block_group)
2804 {
2805         return block_group >> sbi->s_log_groups_per_flex;
2806 }
2807
2808 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2809 {
2810         return 1 << sbi->s_log_groups_per_flex;
2811 }
2812
2813 #define ext4_std_error(sb, errno)                               \
2814 do {                                                            \
2815         if ((errno))                                            \
2816                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2817 } while (0)
2818
2819 #ifdef CONFIG_SMP
2820 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2821  * counters. So we need to make sure we have free clusters more
2822  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2823  */
2824 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2825 #else
2826 #define EXT4_FREECLUSTERS_WATERMARK 0
2827 #endif
2828
2829 /* Update i_disksize. Requires i_mutex to avoid races with truncate */
2830 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2831 {
2832         WARN_ON_ONCE(S_ISREG(inode->i_mode) &&
2833                      !inode_is_locked(inode));
2834         down_write(&EXT4_I(inode)->i_data_sem);
2835         if (newsize > EXT4_I(inode)->i_disksize)
2836                 EXT4_I(inode)->i_disksize = newsize;
2837         up_write(&EXT4_I(inode)->i_data_sem);
2838 }
2839
2840 /* Update i_size, i_disksize. Requires i_mutex to avoid races with truncate */
2841 static inline int ext4_update_inode_size(struct inode *inode, loff_t newsize)
2842 {
2843         int changed = 0;
2844
2845         if (newsize > inode->i_size) {
2846                 i_size_write(inode, newsize);
2847                 changed = 1;
2848         }
2849         if (newsize > EXT4_I(inode)->i_disksize) {
2850                 ext4_update_i_disksize(inode, newsize);
2851                 changed |= 2;
2852         }
2853         return changed;
2854 }
2855
2856 int ext4_update_disksize_before_punch(struct inode *inode, loff_t offset,
2857                                       loff_t len);
2858
2859 struct ext4_group_info {
2860         unsigned long   bb_state;
2861         struct rb_root  bb_free_root;
2862         ext4_grpblk_t   bb_first_free;  /* first free block */
2863         ext4_grpblk_t   bb_free;        /* total free blocks */
2864         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2865         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2866         struct          list_head bb_prealloc_list;
2867 #ifdef DOUBLE_CHECK
2868         void            *bb_bitmap;
2869 #endif
2870         struct rw_semaphore alloc_sem;
2871         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2872                                          * regions, index is order.
2873                                          * bb_counters[3] = 5 means
2874                                          * 5 free 8-block regions. */
2875 };
2876
2877 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2878 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2879 #define EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT     2
2880 #define EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT     3
2881
2882 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2883         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2884 #define EXT4_MB_GRP_BBITMAP_CORRUPT(grp)        \
2885         (test_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
2886 #define EXT4_MB_GRP_IBITMAP_CORRUPT(grp)        \
2887         (test_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
2888
2889 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2890         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2891 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2892         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2893 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2894         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2895
2896 #define EXT4_MAX_CONTENTION             8
2897 #define EXT4_CONTENTION_THRESHOLD       2
2898
2899 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2900                                               ext4_group_t group)
2901 {
2902         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2903 }
2904
2905 /*
2906  * Returns true if the filesystem is busy enough that attempts to
2907  * access the block group locks has run into contention.
2908  */
2909 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2910 {
2911         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2912 }
2913
2914 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2915 {
2916         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2917         if (spin_trylock(lock))
2918                 /*
2919                  * We're able to grab the lock right away, so drop the
2920                  * lock contention counter.
2921                  */
2922                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2923         else {
2924                 /*
2925                  * The lock is busy, so bump the contention counter,
2926                  * and then wait on the spin lock.
2927                  */
2928                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2929                                   EXT4_MAX_CONTENTION);
2930                 spin_lock(lock);
2931         }
2932 }
2933
2934 static inline void ext4_unlock_group(struct super_block *sb,
2935                                         ext4_group_t group)
2936 {
2937         spin_unlock(ext4_group_lock_ptr(sb, group));
2938 }
2939
2940 /*
2941  * Block validity checking
2942  */
2943 #define ext4_check_indirect_blockref(inode, bh)                         \
2944         ext4_check_blockref(__func__, __LINE__, inode,                  \
2945                             (__le32 *)(bh)->b_data,                     \
2946                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2947
2948 #define ext4_ind_check_inode(inode)                                     \
2949         ext4_check_blockref(__func__, __LINE__, inode,                  \
2950                             EXT4_I(inode)->i_data,                      \
2951                             EXT4_NDIR_BLOCKS)
2952
2953 /*
2954  * Inodes and files operations
2955  */
2956
2957 /* dir.c */
2958 extern const struct file_operations ext4_dir_operations;
2959
2960 /* file.c */
2961 extern const struct inode_operations ext4_file_inode_operations;
2962 extern const struct file_operations ext4_file_operations;
2963 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2964
2965 /* inline.c */
2966 extern int ext4_get_max_inline_size(struct inode *inode);
2967 extern int ext4_find_inline_data_nolock(struct inode *inode);
2968 extern int ext4_init_inline_data(handle_t *handle, struct inode *inode,
2969                                  unsigned int len);
2970 extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
2971
2972 extern int ext4_readpage_inline(struct inode *inode, struct page *page);
2973 extern int ext4_try_to_write_inline_data(struct address_space *mapping,
2974                                          struct inode *inode,
2975                                          loff_t pos, unsigned len,
2976                                          unsigned flags,
2977                                          struct page **pagep);
2978 extern int ext4_write_inline_data_end(struct inode *inode,
2979                                       loff_t pos, unsigned len,
2980                                       unsigned copied,
2981                                       struct page *page);
2982 extern struct buffer_head *
2983 ext4_journalled_write_inline_data(struct inode *inode,
2984                                   unsigned len,
2985                                   struct page *page);
2986 extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
2987                                            struct inode *inode,
2988                                            loff_t pos, unsigned len,
2989                                            unsigned flags,
2990                                            struct page **pagep,
2991                                            void **fsdata);
2992 extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
2993                                          unsigned len, unsigned copied,
2994                                          struct page *page);
2995 extern int ext4_try_add_inline_entry(handle_t *handle,
2996                                      struct ext4_filename *fname,
2997                                      struct inode *dir, struct inode *inode);
2998 extern int ext4_try_create_inline_dir(handle_t *handle,
2999                                       struct inode *parent,
3000                                       struct inode *inode);
3001 extern int ext4_read_inline_dir(struct file *filp,
3002                                 struct dir_context *ctx,
3003                                 int *has_inline_data);
3004 extern int htree_inlinedir_to_tree(struct file *dir_file,
3005                                    struct inode *dir, ext4_lblk_t block,
3006                                    struct dx_hash_info *hinfo,
3007                                    __u32 start_hash, __u32 start_minor_hash,
3008                                    int *has_inline_data);
3009 extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
3010                                         struct ext4_filename *fname,
3011                                         const struct qstr *d_name,
3012                                         struct ext4_dir_entry_2 **res_dir,
3013                                         int *has_inline_data);
3014 extern int ext4_delete_inline_entry(handle_t *handle,
3015                                     struct inode *dir,
3016                                     struct ext4_dir_entry_2 *de_del,
3017                                     struct buffer_head *bh,
3018                                     int *has_inline_data);
3019 extern bool empty_inline_dir(struct inode *dir, int *has_inline_data);
3020 extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
3021                                         struct ext4_dir_entry_2 **parent_de,
3022                                         int *retval);
3023 extern int ext4_inline_data_fiemap(struct inode *inode,
3024                                    struct fiemap_extent_info *fieinfo,
3025                                    int *has_inline, __u64 start, __u64 len);
3026 extern int ext4_try_to_evict_inline_data(handle_t *handle,
3027                                          struct inode *inode,
3028                                          int needed);
3029 extern int ext4_inline_data_truncate(struct inode *inode, int *has_inline);
3030
3031 extern int ext4_convert_inline_data(struct inode *inode);
3032
3033 static inline int ext4_has_inline_data(struct inode *inode)
3034 {
3035         return ext4_test_inode_flag(inode, EXT4_INODE_INLINE_DATA) &&
3036                EXT4_I(inode)->i_inline_off;
3037 }
3038
3039 /* namei.c */
3040 extern const struct inode_operations ext4_dir_inode_operations;
3041 extern const struct inode_operations ext4_special_inode_operations;
3042 extern struct dentry *ext4_get_parent(struct dentry *child);
3043 extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
3044                                  struct ext4_dir_entry_2 *de,
3045                                  int blocksize, int csum_size,
3046                                  unsigned int parent_ino, int dotdot_real_len);
3047 extern void initialize_dirent_tail(struct ext4_dir_entry_tail *t,
3048                                    unsigned int blocksize);
3049 extern int ext4_handle_dirty_dirent_node(handle_t *handle,
3050                                          struct inode *inode,
3051                                          struct buffer_head *bh);
3052 #define S_SHIFT 12
3053 static unsigned char ext4_type_by_mode[S_IFMT >> S_SHIFT] = {
3054         [S_IFREG >> S_SHIFT]    = EXT4_FT_REG_FILE,
3055         [S_IFDIR >> S_SHIFT]    = EXT4_FT_DIR,
3056         [S_IFCHR >> S_SHIFT]    = EXT4_FT_CHRDEV,
3057         [S_IFBLK >> S_SHIFT]    = EXT4_FT_BLKDEV,
3058         [S_IFIFO >> S_SHIFT]    = EXT4_FT_FIFO,
3059         [S_IFSOCK >> S_SHIFT]   = EXT4_FT_SOCK,
3060         [S_IFLNK >> S_SHIFT]    = EXT4_FT_SYMLINK,
3061 };
3062
3063 static inline void ext4_set_de_type(struct super_block *sb,
3064                                 struct ext4_dir_entry_2 *de,
3065                                 umode_t mode) {
3066         if (ext4_has_feature_filetype(sb))
3067                 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
3068 }
3069
3070 /* readpages.c */
3071 extern int ext4_mpage_readpages(struct address_space *mapping,
3072                                 struct list_head *pages, struct page *page,
3073                                 unsigned nr_pages);
3074
3075 /* symlink.c */
3076 extern const struct inode_operations ext4_encrypted_symlink_inode_operations;
3077 extern const struct inode_operations ext4_symlink_inode_operations;
3078 extern const struct inode_operations ext4_fast_symlink_inode_operations;
3079
3080 /* sysfs.c */
3081 extern int ext4_register_sysfs(struct super_block *sb);
3082 extern void ext4_unregister_sysfs(struct super_block *sb);
3083 extern int __init ext4_init_sysfs(void);
3084 extern void ext4_exit_sysfs(void);
3085
3086 /* block_validity */
3087 extern void ext4_release_system_zone(struct super_block *sb);
3088 extern int ext4_setup_system_zone(struct super_block *sb);
3089 extern int __init ext4_init_system_zone(void);
3090 extern void ext4_exit_system_zone(void);
3091 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
3092                                  ext4_fsblk_t start_blk,
3093                                  unsigned int count);
3094 extern int ext4_check_blockref(const char *, unsigned int,
3095                                struct inode *, __le32 *, unsigned int);
3096
3097 /* extents.c */
3098 struct ext4_ext_path;
3099 struct ext4_extent;
3100
3101 /*
3102  * Maximum number of logical blocks in a file; ext4_extent's ee_block is
3103  * __le32.
3104  */
3105 #define EXT_MAX_BLOCKS  0xffffffff
3106
3107 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
3108 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
3109 extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
3110 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
3111                                struct ext4_map_blocks *map, int flags);
3112 extern int ext4_ext_truncate(handle_t *, struct inode *);
3113 extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
3114                                  ext4_lblk_t end);
3115 extern void ext4_ext_init(struct super_block *);
3116 extern void ext4_ext_release(struct super_block *);
3117 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
3118                           loff_t len);
3119 extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
3120                                           loff_t offset, ssize_t len);
3121 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
3122                            struct ext4_map_blocks *map, int flags);
3123 extern int ext4_ext_calc_metadata_amount(struct inode *inode,
3124                                          ext4_lblk_t lblocks);
3125 extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
3126                                                    int num,
3127                                                    struct ext4_ext_path *path);
3128 extern int ext4_can_extents_be_merged(struct inode *inode,
3129                                       struct ext4_extent *ex1,
3130                                       struct ext4_extent *ex2);
3131 extern int ext4_ext_insert_extent(handle_t *, struct inode *,
3132                                   struct ext4_ext_path **,
3133                                   struct ext4_extent *, int);
3134 extern struct ext4_ext_path *ext4_find_extent(struct inode *, ext4_lblk_t,
3135                                               struct ext4_ext_path **,
3136                                               int flags);
3137 extern void ext4_ext_drop_refs(struct ext4_ext_path *);
3138 extern int ext4_ext_check_inode(struct inode *inode);
3139 extern int ext4_find_delalloc_range(struct inode *inode,
3140                                     ext4_lblk_t lblk_start,
3141                                     ext4_lblk_t lblk_end);
3142 extern int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk);
3143 extern ext4_lblk_t ext4_ext_next_allocated_block(struct ext4_ext_path *path);
3144 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
3145                         __u64 start, __u64 len);
3146 extern int ext4_ext_precache(struct inode *inode);
3147 extern int ext4_collapse_range(struct inode *inode, loff_t offset, loff_t len);
3148 extern int ext4_insert_range(struct inode *inode, loff_t offset, loff_t len);
3149 extern int ext4_swap_extents(handle_t *handle, struct inode *inode1,
3150                                 struct inode *inode2, ext4_lblk_t lblk1,
3151                              ext4_lblk_t lblk2,  ext4_lblk_t count,
3152                              int mark_unwritten,int *err);
3153
3154 /* move_extent.c */
3155 extern void ext4_double_down_write_data_sem(struct inode *first,
3156                                             struct inode *second);
3157 extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
3158                                           struct inode *donor_inode);
3159 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
3160                              __u64 start_orig, __u64 start_donor,
3161                              __u64 len, __u64 *moved_len);
3162
3163 /* page-io.c */
3164 extern int __init ext4_init_pageio(void);
3165 extern void ext4_exit_pageio(void);
3166 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
3167 extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
3168 extern int ext4_put_io_end(ext4_io_end_t *io_end);
3169 extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
3170 extern void ext4_io_submit_init(struct ext4_io_submit *io,
3171                                 struct writeback_control *wbc);
3172 extern void ext4_end_io_rsv_work(struct work_struct *work);
3173 extern void ext4_io_submit(struct ext4_io_submit *io);
3174 extern int ext4_bio_write_page(struct ext4_io_submit *io,
3175                                struct page *page,
3176                                int len,
3177                                struct writeback_control *wbc,
3178                                bool keep_towrite);
3179
3180 /* mmp.c */
3181 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
3182
3183 /*
3184  * Add new method to test whether block and inode bitmaps are properly
3185  * initialized. With uninit_bg reading the block from disk is not enough
3186  * to mark the bitmap uptodate. We need to also zero-out the bitmap
3187  */
3188 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
3189
3190 static inline int bitmap_uptodate(struct buffer_head *bh)
3191 {
3192         return (buffer_uptodate(bh) &&
3193                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
3194 }
3195 static inline void set_bitmap_uptodate(struct buffer_head *bh)
3196 {
3197         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
3198 }
3199
3200 /*
3201  * Disable DIO read nolock optimization, so new dioreaders will be forced
3202  * to grab i_mutex
3203  */
3204 static inline void ext4_inode_block_unlocked_dio(struct inode *inode)
3205 {
3206         ext4_set_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
3207         smp_mb();
3208 }
3209 static inline void ext4_inode_resume_unlocked_dio(struct inode *inode)
3210 {
3211         smp_mb();
3212         ext4_clear_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
3213 }
3214
3215 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
3216
3217 /* For ioend & aio unwritten conversion wait queues */
3218 #define EXT4_WQ_HASH_SZ         37
3219 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
3220                                             EXT4_WQ_HASH_SZ])
3221 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
3222
3223 extern int ext4_resize_begin(struct super_block *sb);
3224 extern void ext4_resize_end(struct super_block *sb);
3225
3226 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
3227                                               struct ext4_io_end *io_end)
3228 {
3229         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
3230                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
3231                 atomic_inc(&EXT4_I(inode)->i_unwritten);
3232         }
3233 }
3234
3235 static inline void ext4_clear_io_unwritten_flag(ext4_io_end_t *io_end)
3236 {
3237         struct inode *inode = io_end->inode;
3238
3239         if (io_end->flag & EXT4_IO_END_UNWRITTEN) {
3240                 io_end->flag &= ~EXT4_IO_END_UNWRITTEN;
3241                 /* Wake up anyone waiting on unwritten extent conversion */
3242                 if (atomic_dec_and_test(&EXT4_I(inode)->i_unwritten))
3243                         wake_up_all(ext4_ioend_wq(inode));
3244         }
3245 }
3246
3247 extern const struct iomap_ops ext4_iomap_ops;
3248
3249 #endif  /* __KERNEL__ */
3250
3251 #define EFSBADCRC       EBADMSG         /* Bad CRC detected */
3252 #define EFSCORRUPTED    EUCLEAN         /* Filesystem is corrupted */
3253
3254 #endif  /* _EXT4_H */