kernel/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 #include <linux/falloc.h>
  36 #ifdef __KERNEL__
  37 #include <linux/compat.h>
  38 #endif
  39
  40 /*
  41  * The fourth extended filesystem constants/structures
  42  */
  43
  44 /*
  45  * Define EXT4FS_DEBUG to produce debug messages
  46  */
  47 #undef EXT4FS_DEBUG
  48
  49 /*
  50  * Debug code
  51  */
  52 #ifdef EXT4FS_DEBUG
  53 #define ext4_debug(f, a...)                                             \
  54         do {                                                            \
  55                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
  56                         __FILE__, __LINE__, __func__);                  \
  57                 printk(KERN_DEBUG f, ## a);                             \
  58         } while (0)
  59 #else
  60 #define ext4_debug(fmt, ...)    no_printk(fmt, ##__VA_ARGS__)
  61 #endif
  62
  63 /*
  64  * Turn on EXT_DEBUG to get lots of info about extents operations.
  65  */
  66 #define EXT_DEBUG__
  67 #ifdef EXT_DEBUG
  68 #define ext_debug(fmt, ...)     printk(fmt, ##__VA_ARGS__)
  69 #else
  70 #define ext_debug(fmt, ...)     no_printk(fmt, ##__VA_ARGS__)
  71 #endif
  72
  73 /* data type for block offset of block group */
  74 typedef int ext4_grpblk_t;
  75
  76 /* data type for filesystem-wide blocks number */
  77 typedef unsigned long long ext4_fsblk_t;
  78
  79 /* data type for file logical block number */
  80 typedef __u32 ext4_lblk_t;
  81
  82 /* data type for block group number */
  83 typedef unsigned int ext4_group_t;
  84
  85 enum SHIFT_DIRECTION {
  86         SHIFT_LEFT = 0,
  87         SHIFT_RIGHT,
  88 };
  89
  90 /*
  91  * Flags used in mballoc's allocation_context flags field.
  92  *
  93  * Also used to show what's going on for debugging purposes when the
  94  * flag field is exported via the traceport interface
  95  */
  96
  97 /* prefer goal again. length */
  98 #define EXT4_MB_HINT_MERGE              0x0001
  99 /* blocks already reserved */
 100 #define EXT4_MB_HINT_RESERVED           0x0002
 101 /* metadata is being allocated */
 102 #define EXT4_MB_HINT_METADATA           0x0004
 103 /* first blocks in the file */
 104 #define EXT4_MB_HINT_FIRST              0x0008
 105 /* search for the best chunk */
 106 #define EXT4_MB_HINT_BEST               0x0010
 107 /* data is being allocated */
 108 #define EXT4_MB_HINT_DATA               0x0020
 109 /* don't preallocate (for tails) */
 110 #define EXT4_MB_HINT_NOPREALLOC         0x0040
 111 /* allocate for locality group */
 112 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
 113 /* allocate goal blocks or none */
 114 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
 115 /* goal is meaningful */
 116 #define EXT4_MB_HINT_TRY_GOAL           0x0200
 117 /* blocks already pre-reserved by delayed allocation */
 118 #define EXT4_MB_DELALLOC_RESERVED       0x0400
 119 /* We are doing stream allocation */
 120 #define EXT4_MB_STREAM_ALLOC            0x0800
 121 /* Use reserved root blocks if needed */
 122 #define EXT4_MB_USE_ROOT_BLOCKS         0x1000
 123 /* Use blocks from reserved pool */
 124 #define EXT4_MB_USE_RESERVED            0x2000
 125
 126 struct ext4_allocation_request {
 127         /* target inode for block we're allocating */
 128         struct inode *inode;
 129         /* how many blocks we want to allocate */
 130         unsigned int len;
 131         /* logical block in target inode */
 132         ext4_lblk_t logical;
 133         /* the closest logical allocated block to the left */
 134         ext4_lblk_t lleft;
 135         /* the closest logical allocated block to the right */
 136         ext4_lblk_t lright;
 137         /* phys. target (a hint) */
 138         ext4_fsblk_t goal;
 139         /* phys. block for the closest logical allocated block to the left */
 140         ext4_fsblk_t pleft;
 141         /* phys. block for the closest logical allocated block to the right */
 142         ext4_fsblk_t pright;
 143         /* flags. see above EXT4_MB_HINT_* */
 144         unsigned int flags;
 145 };
 146
 147 /*
 148  * Logical to physical block mapping, used by ext4_map_blocks()
 149  *
 150  * This structure is used to pass requests into ext4_map_blocks() as
 151  * well as to store the information returned by ext4_map_blocks().  It
 152  * takes less room on the stack than a struct buffer_head.
 153  */
 154 #define EXT4_MAP_NEW            (1 << BH_New)
 155 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
 156 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
 157 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
 158 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
 159                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY)
 160
 161 struct ext4_map_blocks {
 162         ext4_fsblk_t m_pblk;
 163         ext4_lblk_t m_lblk;
 164         unsigned int m_len;
 165         unsigned int m_flags;
 166 };
 167
 168 /*
 169  * Flags for ext4_io_end->flags
 170  */
 171 #define EXT4_IO_END_UNWRITTEN   0x0001
 172
 173 /*
 174  * For converting unwritten extents on a work queue. 'handle' is used for
 175  * buffered writeback.
 176  */
 177 typedef struct ext4_io_end {
 178         struct list_head        list;           /* per-file finished IO list */
 179         handle_t                *handle;        /* handle reserved for extent
 180                                                  * conversion */
 181         struct inode            *inode;         /* file being written to */
 182         struct bio              *bio;           /* Linked list of completed
 183                                                  * bios covering the extent */
 184         unsigned int            flag;           /* unwritten or not */
 185         loff_t                  offset;         /* offset in the file */
 186         ssize_t                 size;           /* size of the extent */
 187         atomic_t                count;          /* reference counter */
 188 } ext4_io_end_t;
 189
 190 struct ext4_io_submit {
 191         struct writeback_control *io_wbc;
 192         struct bio              *io_bio;
 193         ext4_io_end_t           *io_end;
 194         sector_t                io_next_block;
 195 };
 196
 197 /*
 198  * Special inodes numbers
 199  */
 200 #define EXT4_BAD_INO             1      /* Bad blocks inode */
 201 #define EXT4_ROOT_INO            2      /* Root inode */
 202 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
 203 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
 204 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
 205 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
 206 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
 207 #define EXT4_JOURNAL_INO         8      /* Journal inode */
 208
 209 /* First non-reserved inode for old ext4 filesystems */
 210 #define EXT4_GOOD_OLD_FIRST_INO 11
 211
 212 /*
 213  * Maximal count of links to a file
 214  */
 215 #define EXT4_LINK_MAX           65000
 216
 217 /*
 218  * Macro-instructions used to manage several block sizes
 219  */
 220 #define EXT4_MIN_BLOCK_SIZE             1024
 221 #define EXT4_MAX_BLOCK_SIZE             65536
 222 #define EXT4_MIN_BLOCK_LOG_SIZE         10
 223 #define EXT4_MAX_BLOCK_LOG_SIZE         16
 224 #define EXT4_MAX_CLUSTER_LOG_SIZE       30
 225 #ifdef __KERNEL__
 226 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
 227 #else
 228 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
 229 #endif
 230 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
 231 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
 232                                          EXT4_SB(s)->s_cluster_bits)
 233 #ifdef __KERNEL__
 234 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
 235 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
 236 #else
 237 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
 238 #endif
 239 #ifdef __KERNEL__
 240 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
 241 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
 242 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
 243 #else
 244 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 245                                  EXT4_GOOD_OLD_INODE_SIZE : \
 246                                  (s)->s_inode_size)
 247 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
 248                                  EXT4_GOOD_OLD_FIRST_INO : \
 249                                  (s)->s_first_ino)
 250 #endif
 251 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
 252
 253 /* Translate a block number to a cluster number */
 254 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
 255 /* Translate a cluster number to a block number */
 256 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
 257 /* Translate # of blks to # of clusters */
 258 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
 259                                  (sbi)->s_cluster_bits)
 260 /* Mask out the low bits to get the starting block of the cluster */
 261 #define EXT4_PBLK_CMASK(s, pblk) ((pblk) &                              \
 262                                   ~((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
 263 #define EXT4_LBLK_CMASK(s, lblk) ((lblk) &                              \
 264                                   ~((ext4_lblk_t) (s)->s_cluster_ratio - 1))
 265 /* Get the cluster offset */
 266 #define EXT4_PBLK_COFF(s, pblk) ((pblk) &                               \
 267                                  ((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
 268 #define EXT4_LBLK_COFF(s, lblk) ((lblk) &                               \
 269                                  ((ext4_lblk_t) (s)->s_cluster_ratio - 1))
 270
 271 /*
 272  * Structure of a blocks group descriptor
 273  */
 274 struct ext4_group_desc
 275 {
 276         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
 277         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
 278         __le32  bg_inode_table_lo;      /* Inodes table block */
 279         __le16  bg_free_blocks_count_lo;/* Free blocks count */
 280         __le16  bg_free_inodes_count_lo;/* Free inodes count */
 281         __le16  bg_used_dirs_count_lo;  /* Directories count */
 282         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
 283         __le32  bg_exclude_bitmap_lo;   /* Exclude bitmap for snapshots */
 284         __le16  bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
 285         __le16  bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
 286         __le16  bg_itable_unused_lo;    /* Unused inodes count */
 287         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
 288         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
 289         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
 290         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
 291         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
 292         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
 293         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
 294         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
 295         __le32  bg_exclude_bitmap_hi;   /* Exclude bitmap block MSB */
 296         __le16  bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
 297         __le16  bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
 298         __u32   bg_reserved;
 299 };
 300
 301 #define EXT4_BG_INODE_BITMAP_CSUM_HI_END        \
 302         (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
 303          sizeof(__le16))
 304 #define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END        \
 305         (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
 306          sizeof(__le16))
 307
 308 /*
 309  * Structure of a flex block group info
 310  */
 311
 312 struct flex_groups {
 313         atomic64_t      free_clusters;
 314         atomic_t        free_inodes;
 315         atomic_t        used_dirs;
 316 };
 317
 318 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
 319 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
 320 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
 321
 322 /*
 323  * Macro-instructions used to manage group descriptors
 324  */
 325 #define EXT4_MIN_DESC_SIZE              32
 326 #define EXT4_MIN_DESC_SIZE_64BIT        64
 327 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
 328 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
 329 #ifdef __KERNEL__
 330 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
 331 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
 332 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
 333 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
 334 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
 335 #else
 336 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
 337 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
 338 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
 339 #endif
 340
 341 /*
 342  * Constants relative to the data blocks
 343  */
 344 #define EXT4_NDIR_BLOCKS                12
 345 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
 346 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
 347 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
 348 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
 349
 350 /*
 351  * Inode flags
 352  */
 353 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
 354 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
 355 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
 356 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
 357 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
 358 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
 359 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
 360 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
 361 /* Reserved for compression usage... */
 362 #define EXT4_DIRTY_FL                   0x00000100
 363 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
 364 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
 365         /* nb: was previously EXT2_ECOMPR_FL */
 366 #define EXT4_ENCRYPT_FL                 0x00000800 /* encrypted file */
 367 /* End compression flags --- maybe not all used */
 368 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
 369 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
 370 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
 371 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
 372 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
 373 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
 374 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
 375 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
 376 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
 377 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
 378 #define EXT4_INLINE_DATA_FL             0x10000000 /* Inode has inline data. */
 379 #define EXT4_PROJINHERIT_FL             0x20000000 /* Create with parents projid */
 380 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
 381
 382 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
 383 #define EXT4_FL_USER_MODIFIABLE         0x004380FF /* User modifiable flags */
 384
 385 /* Flags that should be inherited by new inodes from their parent. */
 386 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
 387                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
 388                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
 389                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
 390
 391 /* Flags that are appropriate for regular files (all but dir-specific ones). */
 392 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
 393
 394 /* Flags that are appropriate for non-directories/regular files. */
 395 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
 396
 397 /* Mask out flags that are inappropriate for the given type of inode. */
 398 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
 399 {
 400         if (S_ISDIR(mode))
 401                 return flags;
 402         else if (S_ISREG(mode))
 403                 return flags & EXT4_REG_FLMASK;
 404         else
 405                 return flags & EXT4_OTHER_FLMASK;
 406 }
 407
 408 /*
 409  * Inode flags used for atomic set/get
 410  */
 411 enum {
 412         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
 413         EXT4_INODE_UNRM         = 1,    /* Undelete */
 414         EXT4_INODE_COMPR        = 2,    /* Compress file */
 415         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
 416         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
 417         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
 418         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
 419         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
 420 /* Reserved for compression usage... */
 421         EXT4_INODE_DIRTY        = 8,
 422         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
 423         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
 424         EXT4_INODE_ENCRYPT      = 11,   /* Encrypted file */
 425 /* End compression flags --- maybe not all used */
 426         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
 427         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
 428         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
 429         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
 430         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
 431         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
 432         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
 433         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
 434         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
 435         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
 436         EXT4_INODE_INLINE_DATA  = 28,   /* Data in inode. */
 437         EXT4_INODE_PROJINHERIT  = 29,   /* Create with parents projid */
 438         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
 439 };
 440
 441 /*
 442  * Since it's pretty easy to mix up bit numbers and hex values, we use a
 443  * build-time check to make sure that EXT4_XXX_FL is consistent with respect to
 444  * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost
 445  * any extra space in the compiled kernel image, otherwise, the build will fail.
 446  * It's important that these values are the same, since we are using
 447  * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent
 448  * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk
 449  * values found in ext2, ext3 and ext4 filesystems, and of course the values
 450  * defined in e2fsprogs.
 451  *
 452  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
 453  */
 454 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
 455 #define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG))
 456
 457 static inline void ext4_check_flag_values(void)
 458 {
 459         CHECK_FLAG_VALUE(SECRM);
 460         CHECK_FLAG_VALUE(UNRM);
 461         CHECK_FLAG_VALUE(COMPR);
 462         CHECK_FLAG_VALUE(SYNC);
 463         CHECK_FLAG_VALUE(IMMUTABLE);
 464         CHECK_FLAG_VALUE(APPEND);
 465         CHECK_FLAG_VALUE(NODUMP);
 466         CHECK_FLAG_VALUE(NOATIME);
 467         CHECK_FLAG_VALUE(DIRTY);
 468         CHECK_FLAG_VALUE(COMPRBLK);
 469         CHECK_FLAG_VALUE(NOCOMPR);
 470         CHECK_FLAG_VALUE(ENCRYPT);
 471         CHECK_FLAG_VALUE(INDEX);
 472         CHECK_FLAG_VALUE(IMAGIC);
 473         CHECK_FLAG_VALUE(JOURNAL_DATA);
 474         CHECK_FLAG_VALUE(NOTAIL);
 475         CHECK_FLAG_VALUE(DIRSYNC);
 476         CHECK_FLAG_VALUE(TOPDIR);
 477         CHECK_FLAG_VALUE(HUGE_FILE);
 478         CHECK_FLAG_VALUE(EXTENTS);
 479         CHECK_FLAG_VALUE(EA_INODE);
 480         CHECK_FLAG_VALUE(EOFBLOCKS);
 481         CHECK_FLAG_VALUE(INLINE_DATA);
 482         CHECK_FLAG_VALUE(PROJINHERIT);
 483         CHECK_FLAG_VALUE(RESERVED);
 484 }
 485
 486 /* Used to pass group descriptor data when online resize is done */
 487 struct ext4_new_group_input {
 488         __u32 group;            /* Group number for this data */
 489         __u64 block_bitmap;     /* Absolute block number of block bitmap */
 490         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
 491         __u64 inode_table;      /* Absolute block number of inode table start */
 492         __u32 blocks_count;     /* Total number of blocks in this group */
 493         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
 494         __u16 unused;
 495 };
 496
 497 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 498 struct compat_ext4_new_group_input {
 499         u32 group;
 500         compat_u64 block_bitmap;
 501         compat_u64 inode_bitmap;
 502         compat_u64 inode_table;
 503         u32 blocks_count;
 504         u16 reserved_blocks;
 505         u16 unused;
 506 };
 507 #endif
 508
 509 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
 510 struct ext4_new_group_data {
 511         __u32 group;
 512         __u64 block_bitmap;
 513         __u64 inode_bitmap;
 514         __u64 inode_table;
 515         __u32 blocks_count;
 516         __u16 reserved_blocks;
 517         __u16 unused;
 518         __u32 free_blocks_count;
 519 };
 520
 521 /* Indexes used to index group tables in ext4_new_group_data */
 522 enum {
 523         BLOCK_BITMAP = 0,       /* block bitmap */
 524         INODE_BITMAP,           /* inode bitmap */
 525         INODE_TABLE,            /* inode tables */
 526         GROUP_TABLE_COUNT,
 527 };
 528
 529 /*
 530  * Flags used by ext4_map_blocks()
 531  */
 532         /* Allocate any needed blocks and/or convert an unwritten
 533            extent to be an initialized ext4 */
 534 #define EXT4_GET_BLOCKS_CREATE                  0x0001
 535         /* Request the creation of an unwritten extent */
 536 #define EXT4_GET_BLOCKS_UNWRIT_EXT              0x0002
 537 #define EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT       (EXT4_GET_BLOCKS_UNWRIT_EXT|\
 538                                                  EXT4_GET_BLOCKS_CREATE)
 539         /* Caller is from the delayed allocation writeout path
 540          * finally doing the actual allocation of delayed blocks */
 541 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
 542         /* caller is from the direct IO path, request to creation of an
 543         unwritten extents if not allocated, split the unwritten
 544         extent if blocks has been preallocated already*/
 545 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
 546 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
 547 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
 548                                          EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
 549         /* Convert extent to initialized after IO complete */
 550 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
 551                                          EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
 552         /* Eventual metadata allocation (due to growing extent tree)
 553          * should not fail, so try to use reserved blocks for that.*/
 554 #define EXT4_GET_BLOCKS_METADATA_NOFAIL         0x0020
 555         /* Don't normalize allocation size (used for fallocate) */
 556 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
 557         /* Request will not result in inode size update (user for fallocate) */
 558 #define EXT4_GET_BLOCKS_KEEP_SIZE               0x0080
 559         /* Do not take i_data_sem locking in ext4_map_blocks */
 560 #define EXT4_GET_BLOCKS_NO_LOCK                 0x0100
 561         /* Convert written extents to unwritten */
 562 #define EXT4_GET_BLOCKS_CONVERT_UNWRITTEN       0x0200
 563
 564 /*
 565  * The bit position of these flags must not overlap with any of the
 566  * EXT4_GET_BLOCKS_*.  They are used by ext4_find_extent(),
 567  * read_extent_tree_block(), ext4_split_extent_at(),
 568  * ext4_ext_insert_extent(), and ext4_ext_create_new_leaf().
 569  * EXT4_EX_NOCACHE is used to indicate that the we shouldn't be
 570  * caching the extents when reading from the extent tree while a
 571  * truncate or punch hole operation is in progress.
 572  */
 573 #define EXT4_EX_NOCACHE                         0x40000000
 574 #define EXT4_EX_FORCE_CACHE                     0x20000000
 575
 576 /*
 577  * Flags used by ext4_free_blocks
 578  */
 579 #define EXT4_FREE_BLOCKS_METADATA       0x0001
 580 #define EXT4_FREE_BLOCKS_FORGET         0x0002
 581 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
 582 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
 583 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
 584 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
 585
 586 /* Encryption algorithms */
 587 #define EXT4_ENCRYPTION_MODE_INVALID            0
 588 #define EXT4_ENCRYPTION_MODE_AES_256_XTS        1
 589 #define EXT4_ENCRYPTION_MODE_AES_256_GCM        2
 590 #define EXT4_ENCRYPTION_MODE_AES_256_CBC        3
 591 #define EXT4_ENCRYPTION_MODE_AES_256_CTS        4
 592
 593 #include "ext4_crypto.h"
 594
 595 /*
 596  * ioctl commands
 597  */
 598 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
 599 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
 600 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
 601 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
 602 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
 603 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
 604 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
 605 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
 606 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
 607 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
 608 #define EXT4_IOC_MIGRATE                _IO('f', 9)
 609  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
 610  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
 611 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
 612 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
 613 #define EXT4_IOC_RESIZE_FS              _IOW('f', 16, __u64)
 614 #define EXT4_IOC_SWAP_BOOT              _IO('f', 17)
 615 #define EXT4_IOC_PRECACHE_EXTENTS       _IO('f', 18)
 616 #define EXT4_IOC_SET_ENCRYPTION_POLICY  _IOR('f', 19, struct ext4_encryption_policy)
 617 #define EXT4_IOC_GET_ENCRYPTION_PWSALT  _IOW('f', 20, __u8[16])
 618 #define EXT4_IOC_GET_ENCRYPTION_POLICY  _IOW('f', 21, struct ext4_encryption_policy)
 619
 620 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 621 /*
 622  * ioctl commands in 32 bit emulation
 623  */
 624 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
 625 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
 626 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
 627 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
 628 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
 629 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
 630 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
 631 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
 632 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
 633 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
 634 #endif
 635
 636 /* Max physical block we can address w/o extents */
 637 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
 638
 639 /*
 640  * Structure of an inode on the disk
 641  */
 642 struct ext4_inode {
 643         __le16  i_mode;         /* File mode */
 644         __le16  i_uid;          /* Low 16 bits of Owner Uid */
 645         __le32  i_size_lo;      /* Size in bytes */
 646         __le32  i_atime;        /* Access time */
 647         __le32  i_ctime;        /* Inode Change time */
 648         __le32  i_mtime;        /* Modification time */
 649         __le32  i_dtime;        /* Deletion Time */
 650         __le16  i_gid;          /* Low 16 bits of Group Id */
 651         __le16  i_links_count;  /* Links count */
 652         __le32  i_blocks_lo;    /* Blocks count */
 653         __le32  i_flags;        /* File flags */
 654         union {
 655                 struct {
 656                         __le32  l_i_version;
 657                 } linux1;
 658                 struct {
 659                         __u32  h_i_translator;
 660                 } hurd1;
 661                 struct {
 662                         __u32  m_i_reserved1;
 663                 } masix1;
 664         } osd1;                         /* OS dependent 1 */
 665         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
 666         __le32  i_generation;   /* File version (for NFS) */
 667         __le32  i_file_acl_lo;  /* File ACL */
 668         __le32  i_size_high;
 669         __le32  i_obso_faddr;   /* Obsoleted fragment address */
 670         union {
 671                 struct {
 672                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
 673                         __le16  l_i_file_acl_high;
 674                         __le16  l_i_uid_high;   /* these 2 fields */
 675                         __le16  l_i_gid_high;   /* were reserved2[0] */
 676                         __le16  l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
 677                         __le16  l_i_reserved;
 678                 } linux2;
 679                 struct {
 680                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 681                         __u16   h_i_mode_high;
 682                         __u16   h_i_uid_high;
 683                         __u16   h_i_gid_high;
 684                         __u32   h_i_author;
 685                 } hurd2;
 686                 struct {
 687                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
 688                         __le16  m_i_file_acl_high;
 689                         __u32   m_i_reserved2[2];
 690                 } masix2;
 691         } osd2;                         /* OS dependent 2 */
 692         __le16  i_extra_isize;
 693         __le16  i_checksum_hi;  /* crc32c(uuid+inum+inode) BE */
 694         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
 695         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
 696         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
 697         __le32  i_crtime;       /* File Creation time */
 698         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
 699         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
 700         __le32  i_projid;       /* Project ID */
 701 };
 702
 703 struct move_extent {
 704         __u32 reserved;         /* should be zero */
 705         __u32 donor_fd;         /* donor file descriptor */
 706         __u64 orig_start;       /* logical start offset in block for orig */
 707         __u64 donor_start;      /* logical start offset in block for donor */
 708         __u64 len;              /* block length to be moved */
 709         __u64 moved_len;        /* moved block length */
 710 };
 711
 712 #define EXT4_EPOCH_BITS 2
 713 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
 714 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
 715
 716 /*
 717  * Extended fields will fit into an inode if the filesystem was formatted
 718  * with large inodes (-I 256 or larger) and there are not currently any EAs
 719  * consuming all of the available space. For new inodes we always reserve
 720  * enough space for the kernel's known extended fields, but for inodes
 721  * created with an old kernel this might not have been the case. None of
 722  * the extended inode fields is critical for correct filesystem operation.
 723  * This macro checks if a certain field fits in the inode. Note that
 724  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
 725  */
 726 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
 727         ((offsetof(typeof(*ext4_inode), field) +        \
 728           sizeof((ext4_inode)->field))                  \
 729         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
 730             (einode)->i_extra_isize))                   \
 731
 732 /*
 733  * We use an encoding that preserves the times for extra epoch "00":
 734  *
 735  * extra  msb of                         adjust for signed
 736  * epoch  32-bit                         32-bit tv_sec to
 737  * bits   time    decoded 64-bit tv_sec  64-bit tv_sec      valid time range
 738  * 0 0    1    -0x80000000..-0x00000001  0x000000000 1901-12-13..1969-12-31
 739  * 0 0    0    0x000000000..0x07fffffff  0x000000000 1970-01-01..2038-01-19
 740  * 0 1    1    0x080000000..0x0ffffffff  0x100000000 2038-01-19..2106-02-07
 741  * 0 1    0    0x100000000..0x17fffffff  0x100000000 2106-02-07..2174-02-25
 742  * 1 0    1    0x180000000..0x1ffffffff  0x200000000 2174-02-25..2242-03-16
 743  * 1 0    0    0x200000000..0x27fffffff  0x200000000 2242-03-16..2310-04-04
 744  * 1 1    1    0x280000000..0x2ffffffff  0x300000000 2310-04-04..2378-04-22
 745  * 1 1    0    0x300000000..0x37fffffff  0x300000000 2378-04-22..2446-05-10
 746  *
 747  * Note that previous versions of the kernel on 64-bit systems would
 748  * incorrectly use extra epoch bits 1,1 for dates between 1901 and
 749  * 1970.  e2fsck will correct this, assuming that it is run on the
 750  * affected filesystem before 2242.
 751  */
 752
 753 static inline __le32 ext4_encode_extra_time(struct timespec *time)
 754 {
 755         u32 extra = sizeof(time->tv_sec) > 4 ?
 756                 ((time->tv_sec - (s32)time->tv_sec) >> 32) & EXT4_EPOCH_MASK : 0;
 757         return cpu_to_le32(extra | (time->tv_nsec << EXT4_EPOCH_BITS));
 758 }
 759
 760 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
 761 {
 762         if (unlikely(sizeof(time->tv_sec) > 4 &&
 763                         (extra & cpu_to_le32(EXT4_EPOCH_MASK)))) {
 764 #if LINUX_VERSION_CODE < KERNEL_VERSION(4,20,0)
 765                 /* Handle legacy encoding of pre-1970 dates with epoch
 766                  * bits 1,1.  We assume that by kernel version 4.20,
 767                  * everyone will have run fsck over the affected
 768                  * filesystems to correct the problem.  (This
 769                  * backwards compatibility may be removed before this
 770                  * time, at the discretion of the ext4 developers.)
 771                  */
 772                 u64 extra_bits = le32_to_cpu(extra) & EXT4_EPOCH_MASK;
 773                 if (extra_bits == 3 && ((time->tv_sec) & 0x80000000) != 0)
 774                         extra_bits = 0;
 775                 time->tv_sec += extra_bits << 32;
 776 #else
 777                 time->tv_sec += (u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK) << 32;
 778 #endif
 779         }
 780         time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
 781 }
 782
 783 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
 784 do {                                                                           \
 785         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
 786         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 787                 (raw_inode)->xtime ## _extra =                                 \
 788                                 ext4_encode_extra_time(&(inode)->xtime);       \
 789 } while (0)
 790
 791 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
 792 do {                                                                           \
 793         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 794                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
 795         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 796                 (raw_inode)->xtime ## _extra =                                 \
 797                                 ext4_encode_extra_time(&(einode)->xtime);      \
 798 } while (0)
 799
 800 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
 801 do {                                                                           \
 802         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
 803         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
 804                 ext4_decode_extra_time(&(inode)->xtime,                        \
 805                                        raw_inode->xtime ## _extra);            \
 806         else                                                                   \
 807                 (inode)->xtime.tv_nsec = 0;                                    \
 808 } while (0)
 809
 810 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
 811 do {                                                                           \
 812         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
 813                 (einode)->xtime.tv_sec =                                       \
 814                         (signed)le32_to_cpu((raw_inode)->xtime);               \
 815         else                                                                   \
 816                 (einode)->xtime.tv_sec = 0;                                    \
 817         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
 818                 ext4_decode_extra_time(&(einode)->xtime,                       \
 819                                        raw_inode->xtime ## _extra);            \
 820         else                                                                   \
 821                 (einode)->xtime.tv_nsec = 0;                                   \
 822 } while (0)
 823
 824 #define i_disk_version osd1.linux1.l_i_version
 825
 826 #if defined(__KERNEL__) || defined(__linux__)
 827 #define i_reserved1     osd1.linux1.l_i_reserved1
 828 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
 829 #define i_blocks_high   osd2.linux2.l_i_blocks_high
 830 #define i_uid_low       i_uid
 831 #define i_gid_low       i_gid
 832 #define i_uid_high      osd2.linux2.l_i_uid_high
 833 #define i_gid_high      osd2.linux2.l_i_gid_high
 834 #define i_checksum_lo   osd2.linux2.l_i_checksum_lo
 835
 836 #elif defined(__GNU__)
 837
 838 #define i_translator    osd1.hurd1.h_i_translator
 839 #define i_uid_high      osd2.hurd2.h_i_uid_high
 840 #define i_gid_high      osd2.hurd2.h_i_gid_high
 841 #define i_author        osd2.hurd2.h_i_author
 842
 843 #elif defined(__masix__)
 844
 845 #define i_reserved1     osd1.masix1.m_i_reserved1
 846 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
 847 #define i_reserved2     osd2.masix2.m_i_reserved2
 848
 849 #endif /* defined(__KERNEL__) || defined(__linux__) */
 850
 851 #include "extents_status.h"
 852
 853 /*
 854  * Lock subclasses for i_data_sem in the ext4_inode_info structure.
 855  *
 856  * These are needed to avoid lockdep false positives when we need to
 857  * allocate blocks to the quota inode during ext4_map_blocks(), while
 858  * holding i_data_sem for a normal (non-quota) inode.  Since we don't
 859  * do quota tracking for the quota inode, this avoids deadlock (as
 860  * well as infinite recursion, since it isn't turtles all the way
 861  * down...)
 862  *
 863  *  I_DATA_SEM_NORMAL - Used for most inodes
 864  *  I_DATA_SEM_OTHER  - Used by move_inode.c for the second normal inode
 865  *                        where the second inode has larger inode number
 866  *                        than the first
 867  *  I_DATA_SEM_QUOTA  - Used for quota inodes only
 868  */
 869 enum {
 870         I_DATA_SEM_NORMAL = 0,
 871         I_DATA_SEM_OTHER,
 872         I_DATA_SEM_QUOTA,
 873 };
 874
 875
 876 /*
 877  * fourth extended file system inode data in memory
 878  */
 879 struct ext4_inode_info {
 880         __le32  i_data[15];     /* unconverted */
 881         __u32   i_dtime;
 882         ext4_fsblk_t    i_file_acl;
 883
 884         /*
 885          * i_block_group is the number of the block group which contains
 886          * this file's inode.  Constant across the lifetime of the inode,
 887          * it is ued for making block allocation decisions - we try to
 888          * place a file's data blocks near its inode block, and new inodes
 889          * near to their parent directory's inode.
 890          */
 891         ext4_group_t    i_block_group;
 892         ext4_lblk_t     i_dir_start_lookup;
 893 #if (BITS_PER_LONG < 64)
 894         unsigned long   i_state_flags;          /* Dynamic state flags */
 895 #endif
 896         unsigned long   i_flags;
 897
 898         /*
 899          * Extended attributes can be read independently of the main file
 900          * data. Taking i_mutex even when reading would cause contention
 901          * between readers of EAs and writers of regular file data, so
 902          * instead we synchronize on xattr_sem when reading or changing
 903          * EAs.
 904          */
 905         struct rw_semaphore xattr_sem;
 906
 907         struct list_head i_orphan;      /* unlinked but open inodes */
 908
 909         /*
 910          * i_disksize keeps track of what the inode size is ON DISK, not
 911          * in memory.  During truncate, i_size is set to the new size by
 912          * the VFS prior to calling ext4_truncate(), but the filesystem won't
 913          * set i_disksize to 0 until the truncate is actually under way.
 914          *
 915          * The intent is that i_disksize always represents the blocks which
 916          * are used by this file.  This allows recovery to restart truncate
 917          * on orphans if we crash during truncate.  We actually write i_disksize
 918          * into the on-disk inode when writing inodes out, instead of i_size.
 919          *
 920          * The only time when i_disksize and i_size may be different is when
 921          * a truncate is in progress.  The only things which change i_disksize
 922          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
 923          */
 924         loff_t  i_disksize;
 925
 926         /*
 927          * i_data_sem is for serialising ext4_truncate() against
 928          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
 929          * data tree are chopped off during truncate. We can't do that in
 930          * ext4 because whenever we perform intermediate commits during
 931          * truncate, the inode and all the metadata blocks *must* be in a
 932          * consistent state which allows truncation of the orphans to restart
 933          * during recovery.  Hence we must fix the get_block-vs-truncate race
 934          * by other means, so we have i_data_sem.
 935          */
 936         struct rw_semaphore i_data_sem;
 937         /*
 938          * i_mmap_sem is for serializing page faults with truncate / punch hole
 939          * operations. We have to make sure that new page cannot be faulted in
 940          * a section of the inode that is being punched. We cannot easily use
 941          * i_data_sem for this since we need protection for the whole punch
 942          * operation and i_data_sem ranks below transaction start so we have
 943          * to occasionally drop it.
 944          */
 945         struct rw_semaphore i_mmap_sem;
 946         struct inode vfs_inode;
 947         struct jbd2_inode *jinode;
 948
 949         spinlock_t i_raw_lock;  /* protects updates to the raw inode */
 950
 951         /*
 952          * File creation time. Its function is same as that of
 953          * struct timespec i_{a,c,m}time in the generic inode.
 954          */
 955         struct timespec i_crtime;
 956
 957         /* mballoc */
 958         struct list_head i_prealloc_list;
 959         spinlock_t i_prealloc_lock;
 960
 961         /* extents status tree */
 962         struct ext4_es_tree i_es_tree;
 963         rwlock_t i_es_lock;
 964         struct list_head i_es_list;
 965         unsigned int i_es_all_nr;       /* protected by i_es_lock */
 966         unsigned int i_es_shk_nr;       /* protected by i_es_lock */
 967         ext4_lblk_t i_es_shrink_lblk;   /* Offset where we start searching for
 968                                            extents to shrink. Protected by
 969                                            i_es_lock  */
 970
 971         /* ialloc */
 972         ext4_group_t    i_last_alloc_group;
 973
 974         /* allocation reservation info for delalloc */
 975         /* In case of bigalloc, these refer to clusters rather than blocks */
 976         unsigned int i_reserved_data_blocks;
 977         unsigned int i_reserved_meta_blocks;
 978         unsigned int i_allocated_meta_blocks;
 979         ext4_lblk_t i_da_metadata_calc_last_lblock;
 980         int i_da_metadata_calc_len;
 981
 982         /* on-disk additional length */
 983         __u16 i_extra_isize;
 984
 985         /* Indicate the inline data space. */
 986         u16 i_inline_off;
 987         u16 i_inline_size;
 988
 989 #ifdef CONFIG_QUOTA
 990         /* quota space reservation, managed internally by quota code */
 991         qsize_t i_reserved_quota;
 992 #endif
 993
 994         /* Lock protecting lists below */
 995         spinlock_t i_completed_io_lock;
 996         /*
 997          * Completed IOs that need unwritten extents handling and have
 998          * transaction reserved
 999          */
1000         struct list_head i_rsv_conversion_list;
1001         /*
1002          * Completed IOs that need unwritten extents handling and don't have
1003          * transaction reserved
1004          */
1005         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
1006         atomic_t i_unwritten; /* Nr. of inflight conversions pending */
1007         struct work_struct i_rsv_conversion_work;
1008
1009         spinlock_t i_block_reservation_lock;
1010
1011         /*
1012          * Transactions that contain inode's metadata needed to complete
1013          * fsync and fdatasync, respectively.
1014          */
1015         tid_t i_sync_tid;
1016         tid_t i_datasync_tid;
1017
1018 #ifdef CONFIG_QUOTA
1019         struct dquot *i_dquot[MAXQUOTAS];
1020 #endif
1021
1022         /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
1023         __u32 i_csum_seed;
1024
1025 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1026         /* Encryption params */
1027         struct ext4_crypt_info *i_crypt_info;
1028 #endif
1029 };
1030
1031 /*
1032  * File system states
1033  */
1034 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
1035 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
1036 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
1037
1038 /*
1039  * Misc. filesystem flags
1040  */
1041 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
1042 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
1043 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
1044
1045 /*
1046  * Mount flags set via mount options or defaults
1047  */
1048 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
1049 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
1050 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
1051 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
1052 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
1053 #define EXT4_MOUNT_ERRORS_MASK          0x00070
1054 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
1055 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
1056 #ifdef CONFIG_FS_DAX
1057 #define EXT4_MOUNT_DAX                  0x00200 /* Direct Access */
1058 #else
1059 #define EXT4_MOUNT_DAX                  0
1060 #endif
1061 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
1062 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
1063 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
1064 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
1065 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
1066 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
1067 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
1068 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
1069 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
1070 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
1071 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
1072 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
1073 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
1074 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
1075 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
1076 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
1077 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
1078 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
1079 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
1080 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
1081 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
1082
1083 /*
1084  * Mount flags set either automatically (could not be set by mount option)
1085  * based on per file system feature or property or in special cases such as
1086  * distinguishing between explicit mount option definition and default.
1087  */
1088 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
1089                                                       specified delalloc */
1090 #define EXT4_MOUNT2_STD_GROUP_SIZE      0x00000002 /* We have standard group
1091                                                       size of blocksize * 8
1092                                                       blocks */
1093 #define EXT4_MOUNT2_HURD_COMPAT         0x00000004 /* Support HURD-castrated
1094                                                       file systems */
1095
1096 #define EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM   0x00000008 /* User explicitly
1097                                                 specified journal checksum */
1098
1099 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
1100                                                 ~EXT4_MOUNT_##opt
1101 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
1102                                                 EXT4_MOUNT_##opt
1103 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
1104                                          EXT4_MOUNT_##opt)
1105
1106 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
1107                                                 ~EXT4_MOUNT2_##opt
1108 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
1109                                                 EXT4_MOUNT2_##opt
1110 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
1111                                          EXT4_MOUNT2_##opt)
1112
1113 #define ext4_test_and_set_bit           __test_and_set_bit_le
1114 #define ext4_set_bit                    __set_bit_le
1115 #define ext4_set_bit_atomic             ext2_set_bit_atomic
1116 #define ext4_test_and_clear_bit         __test_and_clear_bit_le
1117 #define ext4_clear_bit                  __clear_bit_le
1118 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
1119 #define ext4_test_bit                   test_bit_le
1120 #define ext4_find_next_zero_bit         find_next_zero_bit_le
1121 #define ext4_find_next_bit              find_next_bit_le
1122
1123 extern void ext4_set_bits(void *bm, int cur, int len);
1124
1125 /*
1126  * Maximal mount counts between two filesystem checks
1127  */
1128 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
1129 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
1130
1131 /*
1132  * Behaviour when detecting errors
1133  */
1134 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
1135 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
1136 #define EXT4_ERRORS_PANIC               3       /* Panic */
1137 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
1138
1139 /* Metadata checksum algorithm codes */
1140 #define EXT4_CRC32C_CHKSUM              1
1141
1142 /*
1143  * Structure of the super block
1144  */
1145 struct ext4_super_block {
1146 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1147         __le32  s_blocks_count_lo;      /* Blocks count */
1148         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1149         __le32  s_free_blocks_count_lo; /* Free blocks count */
1150 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1151         __le32  s_first_data_block;     /* First Data Block */
1152         __le32  s_log_block_size;       /* Block size */
1153         __le32  s_log_cluster_size;     /* Allocation cluster size */
1154 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1155         __le32  s_clusters_per_group;   /* # Clusters per group */
1156         __le32  s_inodes_per_group;     /* # Inodes per group */
1157         __le32  s_mtime;                /* Mount time */
1158 /*30*/  __le32  s_wtime;                /* Write time */
1159         __le16  s_mnt_count;            /* Mount count */
1160         __le16  s_max_mnt_count;        /* Maximal mount count */
1161         __le16  s_magic;                /* Magic signature */
1162         __le16  s_state;                /* File system state */
1163         __le16  s_errors;               /* Behaviour when detecting errors */
1164         __le16  s_minor_rev_level;      /* minor revision level */
1165 /*40*/  __le32  s_lastcheck;            /* time of last check */
1166         __le32  s_checkinterval;        /* max. time between checks */
1167         __le32  s_creator_os;           /* OS */
1168         __le32  s_rev_level;            /* Revision level */
1169 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1170         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1171         /*
1172          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1173          *
1174          * Note: the difference between the compatible feature set and
1175          * the incompatible feature set is that if there is a bit set
1176          * in the incompatible feature set that the kernel doesn't
1177          * know about, it should refuse to mount the filesystem.
1178          *
1179          * e2fsck's requirements are more strict; if it doesn't know
1180          * about a feature in either the compatible or incompatible
1181          * feature set, it must abort and not try to meddle with
1182          * things it doesn't understand...
1183          */
1184         __le32  s_first_ino;            /* First non-reserved inode */
1185         __le16  s_inode_size;           /* size of inode structure */
1186         __le16  s_block_group_nr;       /* block group # of this superblock */
1187         __le32  s_feature_compat;       /* compatible feature set */
1188 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1189         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1190 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1191 /*78*/  char    s_volume_name[16];      /* volume name */
1192 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1193 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1194         /*
1195          * Performance hints.  Directory preallocation should only
1196          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1197          */
1198         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1199         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1200         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1201         /*
1202          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1203          */
1204 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1205 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1206         __le32  s_journal_dev;          /* device number of journal file */
1207         __le32  s_last_orphan;          /* start of list of inodes to delete */
1208         __le32  s_hash_seed[4];         /* HTREE hash seed */
1209         __u8    s_def_hash_version;     /* Default hash version to use */
1210         __u8    s_jnl_backup_type;
1211         __le16  s_desc_size;            /* size of group descriptor */
1212 /*100*/ __le32  s_default_mount_opts;
1213         __le32  s_first_meta_bg;        /* First metablock block group */
1214         __le32  s_mkfs_time;            /* When the filesystem was created */
1215         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1216         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1217 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1218         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1219         __le32  s_free_blocks_count_hi; /* Free blocks count */
1220         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1221         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1222         __le32  s_flags;                /* Miscellaneous flags */
1223         __le16  s_raid_stride;          /* RAID stride */
1224         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1225         __le64  s_mmp_block;            /* Block for multi-mount protection */
1226         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1227         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1228         __u8    s_checksum_type;        /* metadata checksum algorithm used */
1229         __u8    s_encryption_level;     /* versioning level for encryption */
1230         __u8    s_reserved_pad;         /* Padding to next 32bits */
1231         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1232         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1233         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1234         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1235                                               snapshot's future use */
1236         __le32  s_snapshot_list;        /* inode number of the head of the
1237                                            on-disk snapshot list */
1238 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1239         __le32  s_error_count;          /* number of fs errors */
1240         __le32  s_first_error_time;     /* first time an error happened */
1241         __le32  s_first_error_ino;      /* inode involved in first error */
1242         __le64  s_first_error_block;    /* block involved of first error */
1243         __u8    s_first_error_func[32]; /* function where the error happened */
1244         __le32  s_first_error_line;     /* line number where error happened */
1245         __le32  s_last_error_time;      /* most recent time of an error */
1246         __le32  s_last_error_ino;       /* inode involved in last error */
1247         __le32  s_last_error_line;      /* line number where error happened */
1248         __le64  s_last_error_block;     /* block involved of last error */
1249         __u8    s_last_error_func[32];  /* function where the error happened */
1250 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1251         __u8    s_mount_opts[64];
1252         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1253         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1254         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1255         __le32  s_backup_bgs[2];        /* groups with sparse_super2 SBs */
1256         __u8    s_encrypt_algos[4];     /* Encryption algorithms in use  */
1257         __u8    s_encrypt_pw_salt[16];  /* Salt used for string2key algorithm */
1258         __le32  s_lpf_ino;              /* Location of the lost+found inode */
1259         __le32  s_prj_quota_inum;       /* inode for tracking project quota */
1260         __le32  s_checksum_seed;        /* crc32c(uuid) if csum_seed set */
1261         __le32  s_reserved[98];         /* Padding to the end of the block */
1262         __le32  s_checksum;             /* crc32c(superblock) */
1263 };
1264
1265 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1266
1267 #ifdef __KERNEL__
1268
1269 /*
1270  * run-time mount flags
1271  */
1272 #define EXT4_MF_MNTDIR_SAMPLED          0x0001
1273 #define EXT4_MF_FS_ABORTED              0x0002  /* Fatal error detected */
1274 #define EXT4_MF_TEST_DUMMY_ENCRYPTION   0x0004
1275
1276 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1277 #define DUMMY_ENCRYPTION_ENABLED(sbi) (unlikely((sbi)->s_mount_flags & \
1278                                                 EXT4_MF_TEST_DUMMY_ENCRYPTION))
1279 #else
1280 #define DUMMY_ENCRYPTION_ENABLED(sbi) (0)
1281 #endif
1282
1283 /* Number of quota types we support */
1284 #define EXT4_MAXQUOTAS 2
1285
1286 /*
1287  * fourth extended-fs super-block data in memory
1288  */
1289 struct ext4_sb_info {
1290         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1291         unsigned long s_inodes_per_block;/* Number of inodes per block */
1292         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1293         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1294         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1295         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1296         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1297         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1298         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1299         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1300         unsigned long s_overhead;  /* # of fs overhead clusters */
1301         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1302         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1303         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1304         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1305         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1306         struct buffer_head **s_group_desc;
1307         unsigned int s_mount_opt;
1308         unsigned int s_mount_opt2;
1309         unsigned int s_mount_flags;
1310         unsigned int s_def_mount_opt;
1311         ext4_fsblk_t s_sb_block;
1312         atomic64_t s_resv_clusters;
1313         kuid_t s_resuid;
1314         kgid_t s_resgid;
1315         unsigned short s_mount_state;
1316         unsigned short s_pad;
1317         int s_addr_per_block_bits;
1318         int s_desc_per_block_bits;
1319         int s_inode_size;
1320         int s_first_ino;
1321         unsigned int s_inode_readahead_blks;
1322         unsigned int s_inode_goal;
1323         spinlock_t s_next_gen_lock;
1324         u32 s_next_generation;
1325         u32 s_hash_seed[4];
1326         int s_def_hash_version;
1327         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1328         struct percpu_counter s_freeclusters_counter;
1329         struct percpu_counter s_freeinodes_counter;
1330         struct percpu_counter s_dirs_counter;
1331         struct percpu_counter s_dirtyclusters_counter;
1332         struct blockgroup_lock *s_blockgroup_lock;
1333         struct proc_dir_entry *s_proc;
1334         struct kobject s_kobj;
1335         struct completion s_kobj_unregister;
1336         struct super_block *s_sb;
1337
1338         /* Journaling */
1339         struct journal_s *s_journal;
1340         struct list_head s_orphan;
1341         struct mutex s_orphan_lock;
1342         unsigned long s_resize_flags;           /* Flags indicating if there
1343                                                    is a resizer */
1344         unsigned long s_commit_interval;
1345         u32 s_max_batch_time;
1346         u32 s_min_batch_time;
1347         struct block_device *journal_bdev;
1348 #ifdef CONFIG_QUOTA
1349         char *s_qf_names[EXT4_MAXQUOTAS];       /* Names of quota files with journalled quota */
1350         int s_jquota_fmt;                       /* Format of quota to use */
1351 #endif
1352         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1353         struct rb_root system_blks;
1354
1355 #ifdef EXTENTS_STATS
1356         /* ext4 extents stats */
1357         unsigned long s_ext_min;
1358         unsigned long s_ext_max;
1359         unsigned long s_depth_max;
1360         spinlock_t s_ext_stats_lock;
1361         unsigned long s_ext_blocks;
1362         unsigned long s_ext_extents;
1363 #endif
1364
1365         /* for buddy allocator */
1366         struct ext4_group_info ***s_group_info;
1367         struct inode *s_buddy_cache;
1368         spinlock_t s_md_lock;
1369         unsigned short *s_mb_offsets;
1370         unsigned int *s_mb_maxs;
1371         unsigned int s_group_info_size;
1372
1373         /* tunables */
1374         unsigned long s_stripe;
1375         unsigned int s_mb_stream_request;
1376         unsigned int s_mb_max_to_scan;
1377         unsigned int s_mb_min_to_scan;
1378         unsigned int s_mb_stats;
1379         unsigned int s_mb_order2_reqs;
1380         unsigned int s_mb_group_prealloc;
1381         unsigned int s_max_dir_size_kb;
1382         /* where last allocation was done - for stream allocation */
1383         unsigned long s_mb_last_group;
1384         unsigned long s_mb_last_start;
1385
1386         /* stats for buddy allocator */
1387         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1388         atomic_t s_bal_success; /* we found long enough chunks */
1389         atomic_t s_bal_allocated;       /* in blocks */
1390         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1391         atomic_t s_bal_goals;   /* goal hits */
1392         atomic_t s_bal_breaks;  /* too long searches */
1393         atomic_t s_bal_2orders; /* 2^order hits */
1394         spinlock_t s_bal_lock;
1395         unsigned long s_mb_buddies_generated;
1396         unsigned long long s_mb_generation_time;
1397         atomic_t s_mb_lost_chunks;
1398         atomic_t s_mb_preallocated;
1399         atomic_t s_mb_discarded;
1400         atomic_t s_lock_busy;
1401
1402         /* locality groups */
1403         struct ext4_locality_group __percpu *s_locality_groups;
1404
1405         /* for write statistics */
1406         unsigned long s_sectors_written_start;
1407         u64 s_kbytes_written;
1408
1409         /* the size of zero-out chunk */
1410         unsigned int s_extent_max_zeroout_kb;
1411
1412         unsigned int s_log_groups_per_flex;
1413         struct flex_groups *s_flex_groups;
1414         ext4_group_t s_flex_groups_allocated;
1415
1416         /* workqueue for reserved extent conversions (buffered io) */
1417         struct workqueue_struct *rsv_conversion_wq;
1418
1419         /* timer for periodic error stats printing */
1420         struct timer_list s_err_report;
1421
1422         /* Lazy inode table initialization info */
1423         struct ext4_li_request *s_li_request;
1424         /* Wait multiplier for lazy initialization thread */
1425         unsigned int s_li_wait_mult;
1426
1427         /* Kernel thread for multiple mount protection */
1428         struct task_struct *s_mmp_tsk;
1429
1430         /* record the last minlen when FITRIM is called. */
1431         atomic_t s_last_trim_minblks;
1432
1433         /* Reference to checksum algorithm driver via cryptoapi */
1434         struct crypto_shash *s_chksum_driver;
1435
1436         /* Precomputed FS UUID checksum for seeding other checksums */
1437         __u32 s_csum_seed;
1438
1439         /* Reclaim extents from extent status tree */
1440         struct shrinker s_es_shrinker;
1441         struct list_head s_es_list;     /* List of inodes with reclaimable extents */
1442         long s_es_nr_inode;
1443         struct ext4_es_stats s_es_stats;
1444         struct mb_cache *s_mb_cache;
1445         spinlock_t s_es_lock ____cacheline_aligned_in_smp;
1446
1447         /* Ratelimit ext4 messages. */
1448         struct ratelimit_state s_err_ratelimit_state;
1449         struct ratelimit_state s_warning_ratelimit_state;
1450         struct ratelimit_state s_msg_ratelimit_state;
1451 };
1452
1453 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1454 {
1455         return sb->s_fs_info;
1456 }
1457 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1458 {
1459         return container_of(inode, struct ext4_inode_info, vfs_inode);
1460 }
1461
1462 static inline struct timespec ext4_current_time(struct inode *inode)
1463 {
1464         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1465                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1466 }
1467
1468 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1469 {
1470         return ino == EXT4_ROOT_INO ||
1471                 ino == EXT4_USR_QUOTA_INO ||
1472                 ino == EXT4_GRP_QUOTA_INO ||
1473                 ino == EXT4_BOOT_LOADER_INO ||
1474                 ino == EXT4_JOURNAL_INO ||
1475                 ino == EXT4_RESIZE_INO ||
1476                 (ino >= EXT4_FIRST_INO(sb) &&
1477                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1478 }
1479
1480 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
1481                                               struct ext4_io_end *io_end)
1482 {
1483         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
1484                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
1485                 atomic_inc(&EXT4_I(inode)->i_unwritten);
1486         }
1487 }
1488
1489 static inline ext4_io_end_t *ext4_inode_aio(struct inode *inode)
1490 {
1491         return inode->i_private;
1492 }
1493
1494 static inline void ext4_inode_aio_set(struct inode *inode, ext4_io_end_t *io)
1495 {
1496         inode->i_private = io;
1497 }
1498
1499 /*
1500  * Inode dynamic state flags
1501  */
1502 enum {
1503         EXT4_STATE_JDATA,               /* journaled data exists */
1504         EXT4_STATE_NEW,                 /* inode is newly created */
1505         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1506         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1507         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1508         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1509         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1510         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1511         EXT4_STATE_DIOREAD_LOCK,        /* Disable support for dio read
1512                                            nolocking */
1513         EXT4_STATE_MAY_INLINE_DATA,     /* may have in-inode data */
1514         EXT4_STATE_ORDERED_MODE,        /* data=ordered mode */
1515         EXT4_STATE_EXT_PRECACHED,       /* extents have been precached */
1516 };
1517
1518 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1519 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1520 {                                                                       \
1521         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1522 }                                                                       \
1523 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1524 {                                                                       \
1525         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1526 }                                                                       \
1527 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1528 {                                                                       \
1529         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1530 }
1531
1532 /* Add these declarations here only so that these functions can be
1533  * found by name.  Otherwise, they are very hard to locate. */
1534 static inline int ext4_test_inode_flag(struct inode *inode, int bit);
1535 static inline void ext4_set_inode_flag(struct inode *inode, int bit);
1536 static inline void ext4_clear_inode_flag(struct inode *inode, int bit);
1537 EXT4_INODE_BIT_FNS(flag, flags, 0)
1538
1539 /* Add these declarations here only so that these functions can be
1540  * found by name.  Otherwise, they are very hard to locate. */
1541 static inline int ext4_test_inode_state(struct inode *inode, int bit);
1542 static inline void ext4_set_inode_state(struct inode *inode, int bit);
1543 static inline void ext4_clear_inode_state(struct inode *inode, int bit);
1544 #if (BITS_PER_LONG < 64)
1545 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1546
1547 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1548 {
1549         (ei)->i_state_flags = 0;
1550 }
1551 #else
1552 EXT4_INODE_BIT_FNS(state, flags, 32)
1553
1554 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1555 {
1556         /* We depend on the fact that callers will set i_flags */
1557 }
1558 #endif
1559 #else
1560 /* Assume that user mode programs are passing in an ext4fs superblock, not
1561  * a kernel struct super_block.  This will allow us to call the feature-test
1562  * macros from user land. */
1563 #define EXT4_SB(sb)     (sb)
1564 #endif
1565
1566 /*
1567  * Returns true if the inode is inode is encrypted
1568  */
1569 static inline int ext4_encrypted_inode(struct inode *inode)
1570 {
1571 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1572         return ext4_test_inode_flag(inode, EXT4_INODE_ENCRYPT);
1573 #else
1574         return 0;
1575 #endif
1576 }
1577
1578 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1579
1580 /*
1581  * Codes for operating systems
1582  */
1583 #define EXT4_OS_LINUX           0
1584 #define EXT4_OS_HURD            1
1585 #define EXT4_OS_MASIX           2
1586 #define EXT4_OS_FREEBSD         3
1587 #define EXT4_OS_LITES           4
1588
1589 /*
1590  * Revision levels
1591  */
1592 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1593 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1594
1595 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1596 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1597
1598 #define EXT4_GOOD_OLD_INODE_SIZE 128
1599
1600 /*
1601  * Feature set definitions
1602  */
1603
1604 /* Use the ext4_{has,set,clear}_feature_* helpers; these will be removed */
1605 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1606         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1607 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1608         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1609 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1610         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1611 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1612         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1613 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1614         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1615 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1616         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1617 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1618         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1619 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1620         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1621 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1622         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1623
1624 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1625 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1626 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1627 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1628 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1629 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1630 #define EXT4_FEATURE_COMPAT_SPARSE_SUPER2       0x0200
1631
1632 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1633 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1634 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1635 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1636 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1637 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1638 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1639 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1640 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1641 /*
1642  * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM).  When
1643  * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
1644  * all other data structures' checksums.  However, the METADATA_CSUM and
1645  * GDT_CSUM bits are mutually exclusive.
1646  */
1647 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM    0x0400
1648 #define EXT4_FEATURE_RO_COMPAT_READONLY         0x1000
1649 #define EXT4_FEATURE_RO_COMPAT_PROJECT          0x2000
1650
1651 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1652 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1653 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1654 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1655 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1656 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1657 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1658 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1659 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1660 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1661 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1662 #define EXT4_FEATURE_INCOMPAT_CSUM_SEED         0x2000
1663 #define EXT4_FEATURE_INCOMPAT_LARGEDIR          0x4000 /* >2GB or 3-lvl htree */
1664 #define EXT4_FEATURE_INCOMPAT_INLINE_DATA       0x8000 /* data in inode */
1665 #define EXT4_FEATURE_INCOMPAT_ENCRYPT           0x10000
1666
1667 #define EXT4_FEATURE_COMPAT_FUNCS(name, flagname) \
1668 static inline bool ext4_has_feature_##name(struct super_block *sb) \
1669 { \
1670         return ((EXT4_SB(sb)->s_es->s_feature_compat & \
1671                 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname)) != 0); \
1672 } \
1673 static inline void ext4_set_feature_##name(struct super_block *sb) \
1674 { \
1675         EXT4_SB(sb)->s_es->s_feature_compat |= \
1676                 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
1677 } \
1678 static inline void ext4_clear_feature_##name(struct super_block *sb) \
1679 { \
1680         EXT4_SB(sb)->s_es->s_feature_compat &= \
1681                 ~cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
1682 }
1683
1684 #define EXT4_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
1685 static inline bool ext4_has_feature_##name(struct super_block *sb) \
1686 { \
1687         return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
1688                 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname)) != 0); \
1689 } \
1690 static inline void ext4_set_feature_##name(struct super_block *sb) \
1691 { \
1692         EXT4_SB(sb)->s_es->s_feature_ro_compat |= \
1693                 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
1694 } \
1695 static inline void ext4_clear_feature_##name(struct super_block *sb) \
1696 { \
1697         EXT4_SB(sb)->s_es->s_feature_ro_compat &= \
1698                 ~cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
1699 }
1700
1701 #define EXT4_FEATURE_INCOMPAT_FUNCS(name, flagname) \
1702 static inline bool ext4_has_feature_##name(struct super_block *sb) \
1703 { \
1704         return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
1705                 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname)) != 0); \
1706 } \
1707 static inline void ext4_set_feature_##name(struct super_block *sb) \
1708 { \
1709         EXT4_SB(sb)->s_es->s_feature_incompat |= \
1710                 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
1711 } \
1712 static inline void ext4_clear_feature_##name(struct super_block *sb) \
1713 { \
1714         EXT4_SB(sb)->s_es->s_feature_incompat &= \
1715                 ~cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
1716 }
1717
1718 EXT4_FEATURE_COMPAT_FUNCS(dir_prealloc,         DIR_PREALLOC)
1719 EXT4_FEATURE_COMPAT_FUNCS(imagic_inodes,        IMAGIC_INODES)
1720 EXT4_FEATURE_COMPAT_FUNCS(journal,              HAS_JOURNAL)
1721 EXT4_FEATURE_COMPAT_FUNCS(xattr,                EXT_ATTR)
1722 EXT4_FEATURE_COMPAT_FUNCS(resize_inode,         RESIZE_INODE)
1723 EXT4_FEATURE_COMPAT_FUNCS(dir_index,            DIR_INDEX)
1724 EXT4_FEATURE_COMPAT_FUNCS(sparse_super2,        SPARSE_SUPER2)
1725
1726 EXT4_FEATURE_RO_COMPAT_FUNCS(sparse_super,      SPARSE_SUPER)
1727 EXT4_FEATURE_RO_COMPAT_FUNCS(large_file,        LARGE_FILE)
1728 EXT4_FEATURE_RO_COMPAT_FUNCS(btree_dir,         BTREE_DIR)
1729 EXT4_FEATURE_RO_COMPAT_FUNCS(huge_file,         HUGE_FILE)
1730 EXT4_FEATURE_RO_COMPAT_FUNCS(gdt_csum,          GDT_CSUM)
1731 EXT4_FEATURE_RO_COMPAT_FUNCS(dir_nlink,         DIR_NLINK)
1732 EXT4_FEATURE_RO_COMPAT_FUNCS(extra_isize,       EXTRA_ISIZE)
1733 EXT4_FEATURE_RO_COMPAT_FUNCS(quota,             QUOTA)
1734 EXT4_FEATURE_RO_COMPAT_FUNCS(bigalloc,          BIGALLOC)
1735 EXT4_FEATURE_RO_COMPAT_FUNCS(metadata_csum,     METADATA_CSUM)
1736 EXT4_FEATURE_RO_COMPAT_FUNCS(readonly,          READONLY)
1737 EXT4_FEATURE_RO_COMPAT_FUNCS(project,           PROJECT)
1738
1739 EXT4_FEATURE_INCOMPAT_FUNCS(compression,        COMPRESSION)
1740 EXT4_FEATURE_INCOMPAT_FUNCS(filetype,           FILETYPE)
1741 EXT4_FEATURE_INCOMPAT_FUNCS(journal_needs_recovery,     RECOVER)
1742 EXT4_FEATURE_INCOMPAT_FUNCS(journal_dev,        JOURNAL_DEV)
1743 EXT4_FEATURE_INCOMPAT_FUNCS(meta_bg,            META_BG)
1744 EXT4_FEATURE_INCOMPAT_FUNCS(extents,            EXTENTS)
1745 EXT4_FEATURE_INCOMPAT_FUNCS(64bit,              64BIT)
1746 EXT4_FEATURE_INCOMPAT_FUNCS(mmp,                MMP)
1747 EXT4_FEATURE_INCOMPAT_FUNCS(flex_bg,            FLEX_BG)
1748 EXT4_FEATURE_INCOMPAT_FUNCS(ea_inode,           EA_INODE)
1749 EXT4_FEATURE_INCOMPAT_FUNCS(dirdata,            DIRDATA)
1750 EXT4_FEATURE_INCOMPAT_FUNCS(csum_seed,          CSUM_SEED)
1751 EXT4_FEATURE_INCOMPAT_FUNCS(largedir,           LARGEDIR)
1752 EXT4_FEATURE_INCOMPAT_FUNCS(inline_data,        INLINE_DATA)
1753 EXT4_FEATURE_INCOMPAT_FUNCS(encrypt,            ENCRYPT)
1754
1755 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1756 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1757                                          EXT4_FEATURE_INCOMPAT_META_BG)
1758 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1759                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1760                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1761
1762 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1763 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1764                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1765                                          EXT4_FEATURE_INCOMPAT_META_BG)
1766 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1767                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1768                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1769
1770 #define EXT4_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1771 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1772                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1773                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1774                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1775                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1776                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1777                                          EXT4_FEATURE_INCOMPAT_MMP | \
1778                                          EXT4_FEATURE_INCOMPAT_INLINE_DATA | \
1779                                          EXT4_FEATURE_INCOMPAT_ENCRYPT | \
1780                                          EXT4_FEATURE_INCOMPAT_CSUM_SEED)
1781 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1782                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1783                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1784                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1785                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1786                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1787                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1788                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
1789                                          EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
1790                                          EXT4_FEATURE_RO_COMPAT_QUOTA)
1791
1792 #define EXTN_FEATURE_FUNCS(ver) \
1793 static inline bool ext4_has_unknown_ext##ver##_compat_features(struct super_block *sb) \
1794 { \
1795         return ((EXT4_SB(sb)->s_es->s_feature_compat & \
1796                 cpu_to_le32(~EXT##ver##_FEATURE_COMPAT_SUPP)) != 0); \
1797 } \
1798 static inline bool ext4_has_unknown_ext##ver##_ro_compat_features(struct super_block *sb) \
1799 { \
1800         return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
1801                 cpu_to_le32(~EXT##ver##_FEATURE_RO_COMPAT_SUPP)) != 0); \
1802 } \
1803 static inline bool ext4_has_unknown_ext##ver##_incompat_features(struct super_block *sb) \
1804 { \
1805         return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
1806                 cpu_to_le32(~EXT##ver##_FEATURE_INCOMPAT_SUPP)) != 0); \
1807 }
1808
1809 EXTN_FEATURE_FUNCS(2)
1810 EXTN_FEATURE_FUNCS(3)
1811 EXTN_FEATURE_FUNCS(4)
1812
1813 static inline bool ext4_has_compat_features(struct super_block *sb)
1814 {
1815         return (EXT4_SB(sb)->s_es->s_feature_compat != 0);
1816 }
1817 static inline bool ext4_has_ro_compat_features(struct super_block *sb)
1818 {
1819         return (EXT4_SB(sb)->s_es->s_feature_ro_compat != 0);
1820 }
1821 static inline bool ext4_has_incompat_features(struct super_block *sb)
1822 {
1823         return (EXT4_SB(sb)->s_es->s_feature_incompat != 0);
1824 }
1825
1826 /*
1827  * Default values for user and/or group using reserved blocks
1828  */
1829 #define EXT4_DEF_RESUID         0
1830 #define EXT4_DEF_RESGID         0
1831
1832 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1833
1834 /*
1835  * Default mount options
1836  */
1837 #define EXT4_DEFM_DEBUG         0x0001
1838 #define EXT4_DEFM_BSDGROUPS     0x0002
1839 #define EXT4_DEFM_XATTR_USER    0x0004
1840 #define EXT4_DEFM_ACL           0x0008
1841 #define EXT4_DEFM_UID16         0x0010
1842 #define EXT4_DEFM_JMODE         0x0060
1843 #define EXT4_DEFM_JMODE_DATA    0x0020
1844 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1845 #define EXT4_DEFM_JMODE_WBACK   0x0060
1846 #define EXT4_DEFM_NOBARRIER     0x0100
1847 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1848 #define EXT4_DEFM_DISCARD       0x0400
1849 #define EXT4_DEFM_NODELALLOC    0x0800
1850
1851 /*
1852  * Default journal batch times
1853  */
1854 #define EXT4_DEF_MIN_BATCH_TIME 0
1855 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1856
1857 /*
1858  * Minimum number of groups in a flexgroup before we separate out
1859  * directories into the first block group of a flexgroup
1860  */
1861 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1862
1863 /*
1864  * Structure of a directory entry
1865  */
1866 #define EXT4_NAME_LEN 255
1867
1868 struct ext4_dir_entry {
1869         __le32  inode;                  /* Inode number */
1870         __le16  rec_len;                /* Directory entry length */
1871         __le16  name_len;               /* Name length */
1872         char    name[EXT4_NAME_LEN];    /* File name */
1873 };
1874
1875 /*
1876  * The new version of the directory entry.  Since EXT4 structures are
1877  * stored in intel byte order, and the name_len field could never be
1878  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1879  * file_type field.
1880  */
1881 struct ext4_dir_entry_2 {
1882         __le32  inode;                  /* Inode number */
1883         __le16  rec_len;                /* Directory entry length */
1884         __u8    name_len;               /* Name length */
1885         __u8    file_type;
1886         char    name[EXT4_NAME_LEN];    /* File name */
1887 };
1888
1889 /*
1890  * This is a bogus directory entry at the end of each leaf block that
1891  * records checksums.
1892  */
1893 struct ext4_dir_entry_tail {
1894         __le32  det_reserved_zero1;     /* Pretend to be unused */
1895         __le16  det_rec_len;            /* 12 */
1896         __u8    det_reserved_zero2;     /* Zero name length */
1897         __u8    det_reserved_ft;        /* 0xDE, fake file type */
1898         __le32  det_checksum;           /* crc32c(uuid+inum+dirblock) */
1899 };
1900
1901 #define EXT4_DIRENT_TAIL(block, blocksize) \
1902         ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
1903                                         ((blocksize) - \
1904                                          sizeof(struct ext4_dir_entry_tail))))
1905
1906 /*
1907  * Ext4 directory file types.  Only the low 3 bits are used.  The
1908  * other bits are reserved for now.
1909  */
1910 #define EXT4_FT_UNKNOWN         0
1911 #define EXT4_FT_REG_FILE        1
1912 #define EXT4_FT_DIR             2
1913 #define EXT4_FT_CHRDEV          3
1914 #define EXT4_FT_BLKDEV          4
1915 #define EXT4_FT_FIFO            5
1916 #define EXT4_FT_SOCK            6
1917 #define EXT4_FT_SYMLINK         7
1918
1919 #define EXT4_FT_MAX             8
1920
1921 #define EXT4_FT_DIR_CSUM        0xDE
1922
1923 /*
1924  * EXT4_DIR_PAD defines the directory entries boundaries
1925  *
1926  * NOTE: It must be a multiple of 4
1927  */
1928 #define EXT4_DIR_PAD                    4
1929 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1930 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1931                                          ~EXT4_DIR_ROUND)
1932 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1933
1934 /*
1935  * If we ever get support for fs block sizes > page_size, we'll need
1936  * to remove the #if statements in the next two functions...
1937  */
1938 static inline unsigned int
1939 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1940 {
1941         unsigned len = le16_to_cpu(dlen);
1942
1943 #if (PAGE_CACHE_SIZE >= 65536)
1944         if (len == EXT4_MAX_REC_LEN || len == 0)
1945                 return blocksize;
1946         return (len & 65532) | ((len & 3) << 16);
1947 #else
1948         return len;
1949 #endif
1950 }
1951
1952 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1953 {
1954         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1955                 BUG();
1956 #if (PAGE_CACHE_SIZE >= 65536)
1957         if (len < 65536)
1958                 return cpu_to_le16(len);
1959         if (len == blocksize) {
1960                 if (blocksize == 65536)
1961                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1962                 else
1963                         return cpu_to_le16(0);
1964         }
1965         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1966 #else
1967         return cpu_to_le16(len);
1968 #endif
1969 }
1970
1971 /*
1972  * Hash Tree Directory indexing
1973  * (c) Daniel Phillips, 2001
1974  */
1975
1976 #define is_dx(dir) (ext4_has_feature_dir_index((dir)->i_sb) && \
1977                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1978 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1979 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1980
1981 /* Legal values for the dx_root hash_version field: */
1982
1983 #define DX_HASH_LEGACY          0
1984 #define DX_HASH_HALF_MD4        1
1985 #define DX_HASH_TEA             2
1986 #define DX_HASH_LEGACY_UNSIGNED 3
1987 #define DX_HASH_HALF_MD4_UNSIGNED       4
1988 #define DX_HASH_TEA_UNSIGNED            5
1989
1990 static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
1991                               const void *address, unsigned int length)
1992 {
1993         struct {
1994                 struct shash_desc shash;
1995                 char ctx[4];
1996         } desc;
1997         int err;
1998
1999         BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
2000
2001         desc.shash.tfm = sbi->s_chksum_driver;
2002         desc.shash.flags = 0;
2003         *(u32 *)desc.ctx = crc;
2004
2005         err = crypto_shash_update(&desc.shash, address, length);
2006         BUG_ON(err);
2007
2008         return *(u32 *)desc.ctx;
2009 }
2010
2011 #ifdef __KERNEL__
2012
2013 /* hash info structure used by the directory hash */
2014 struct dx_hash_info
2015 {
2016         u32             hash;
2017         u32             minor_hash;
2018         int             hash_version;
2019         u32             *seed;
2020 };
2021
2022
2023 /* 32 and 64 bit signed EOF for dx directories */
2024 #define EXT4_HTREE_EOF_32BIT   ((1UL  << (32 - 1)) - 1)
2025 #define EXT4_HTREE_EOF_64BIT   ((1ULL << (64 - 1)) - 1)
2026
2027
2028 /*
2029  * Control parameters used by ext4_htree_next_block
2030  */
2031 #define HASH_NB_ALWAYS          1
2032
2033 struct ext4_filename {
2034         const struct qstr *usr_fname;
2035         struct ext4_str disk_name;
2036         struct dx_hash_info hinfo;
2037 #ifdef CONFIG_EXT4_FS_ENCRYPTION
2038         struct ext4_str crypto_buf;
2039 #endif
2040 };
2041
2042 #define fname_name(p) ((p)->disk_name.name)
2043 #define fname_len(p)  ((p)->disk_name.len)
2044
2045 /*
2046  * Describe an inode's exact location on disk and in memory
2047  */
2048 struct ext4_iloc
2049 {
2050         struct buffer_head *bh;
2051         unsigned long offset;
2052         ext4_group_t block_group;
2053 };
2054
2055 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
2056 {
2057         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
2058 }
2059
2060 /*
2061  * This structure is stuffed into the struct file's private_data field
2062  * for directories.  It is where we put information so that we can do
2063  * readdir operations in hash tree order.
2064  */
2065 struct dir_private_info {
2066         struct rb_root  root;
2067         struct rb_node  *curr_node;
2068         struct fname    *extra_fname;
2069         loff_t          last_pos;
2070         __u32           curr_hash;
2071         __u32           curr_minor_hash;
2072         __u32           next_hash;
2073 };
2074
2075 /* calculate the first block number of the group */
2076 static inline ext4_fsblk_t
2077 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
2078 {
2079         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
2080                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
2081 }
2082
2083 /*
2084  * Special error return code only used by dx_probe() and its callers.
2085  */
2086 #define ERR_BAD_DX_DIR  (-(MAX_ERRNO - 1))
2087
2088 /*
2089  * Timeout and state flag for lazy initialization inode thread.
2090  */
2091 #define EXT4_DEF_LI_WAIT_MULT                   10
2092 #define EXT4_DEF_LI_MAX_START_DELAY             5
2093 #define EXT4_LAZYINIT_QUIT                      0x0001
2094 #define EXT4_LAZYINIT_RUNNING                   0x0002
2095
2096 /*
2097  * Lazy inode table initialization info
2098  */
2099 struct ext4_lazy_init {
2100         unsigned long           li_state;
2101         struct list_head        li_request_list;
2102         struct mutex            li_list_mtx;
2103 };
2104
2105 struct ext4_li_request {
2106         struct super_block      *lr_super;
2107         struct ext4_sb_info     *lr_sbi;
2108         ext4_group_t            lr_next_group;
2109         struct list_head        lr_request;
2110         unsigned long           lr_next_sched;
2111         unsigned long           lr_timeout;
2112 };
2113
2114 struct ext4_features {
2115         struct kobject f_kobj;
2116         struct completion f_kobj_unregister;
2117 };
2118
2119 /*
2120  * This structure will be used for multiple mount protection. It will be
2121  * written into the block number saved in the s_mmp_block field in the
2122  * superblock. Programs that check MMP should assume that if
2123  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
2124  * to use the filesystem, regardless of how old the timestamp is.
2125  */
2126 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
2127 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
2128 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
2129 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
2130
2131 struct mmp_struct {
2132         __le32  mmp_magic;              /* Magic number for MMP */
2133         __le32  mmp_seq;                /* Sequence no. updated periodically */
2134
2135         /*
2136          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
2137          * purposes and do not affect the correctness of the algorithm
2138          */
2139         __le64  mmp_time;               /* Time last updated */
2140         char    mmp_nodename[64];       /* Node which last updated MMP block */
2141         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
2142
2143         /*
2144          * mmp_check_interval is used to verify if the MMP block has been
2145          * updated on the block device. The value is updated based on the
2146          * maximum time to write the MMP block during an update cycle.
2147          */
2148         __le16  mmp_check_interval;
2149
2150         __le16  mmp_pad1;
2151         __le32  mmp_pad2[226];
2152         __le32  mmp_checksum;           /* crc32c(uuid+mmp_block) */
2153 };
2154
2155 /* arguments passed to the mmp thread */
2156 struct mmpd_data {
2157         struct buffer_head *bh; /* bh from initial read_mmp_block() */
2158         struct super_block *sb;  /* super block of the fs */
2159 };
2160
2161 /*
2162  * Check interval multiplier
2163  * The MMP block is written every update interval and initially checked every
2164  * update interval x the multiplier (the value is then adapted based on the
2165  * write latency). The reason is that writes can be delayed under load and we
2166  * don't want readers to incorrectly assume that the filesystem is no longer
2167  * in use.
2168  */
2169 #define EXT4_MMP_CHECK_MULT             2UL
2170
2171 /*
2172  * Minimum interval for MMP checking in seconds.
2173  */
2174 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
2175
2176 /*
2177  * Maximum interval for MMP checking in seconds.
2178  */
2179 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
2180
2181 /*
2182  * Function prototypes
2183  */
2184
2185 /*
2186  * Ok, these declarations are also in <linux/kernel.h> but none of the
2187  * ext4 source programs needs to include it so they are duplicated here.
2188  */
2189 # define NORET_TYPE     /**/
2190 # define ATTRIB_NORET   __attribute__((noreturn))
2191 # define NORET_AND      noreturn,
2192
2193 /* bitmap.c */
2194 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
2195 void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
2196                                 struct ext4_group_desc *gdp,
2197                                 struct buffer_head *bh, int sz);
2198 int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2199                                   struct ext4_group_desc *gdp,
2200                                   struct buffer_head *bh, int sz);
2201 void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
2202                                 struct ext4_group_desc *gdp,
2203                                 struct buffer_head *bh);
2204 int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2205                                   struct ext4_group_desc *gdp,
2206                                   struct buffer_head *bh);
2207
2208 /* balloc.c */
2209 extern void ext4_get_group_no_and_offset(struct super_block *sb,
2210                                          ext4_fsblk_t blocknr,
2211                                          ext4_group_t *blockgrpp,
2212                                          ext4_grpblk_t *offsetp);
2213 extern ext4_group_t ext4_get_group_number(struct super_block *sb,
2214                                           ext4_fsblk_t block);
2215
2216 extern unsigned int ext4_block_group(struct super_block *sb,
2217                         ext4_fsblk_t blocknr);
2218 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
2219                         ext4_fsblk_t blocknr);
2220 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
2221 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
2222                         ext4_group_t group);
2223 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
2224                                          ext4_fsblk_t goal,
2225                                          unsigned int flags,
2226                                          unsigned long *count,
2227                                          int *errp);
2228 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
2229                                     s64 nclusters, unsigned int flags);
2230 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
2231 extern void ext4_check_blocks_bitmap(struct super_block *);
2232 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
2233                                                     ext4_group_t block_group,
2234                                                     struct buffer_head ** bh);
2235 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
2236
2237 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
2238                                                 ext4_group_t block_group);
2239 extern int ext4_wait_block_bitmap(struct super_block *sb,
2240                                   ext4_group_t block_group,
2241                                   struct buffer_head *bh);
2242 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
2243                                                   ext4_group_t block_group);
2244 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
2245                                               ext4_group_t block_group,
2246                                               struct ext4_group_desc *gdp);
2247 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
2248
2249 /* crypto_policy.c */
2250 int ext4_is_child_context_consistent_with_parent(struct inode *parent,
2251                                                  struct inode *child);
2252 int ext4_inherit_context(struct inode *parent, struct inode *child);
2253 void ext4_to_hex(char *dst, char *src, size_t src_size);
2254 int ext4_process_policy(const struct ext4_encryption_policy *policy,
2255                         struct inode *inode);
2256 int ext4_get_policy(struct inode *inode,
2257                     struct ext4_encryption_policy *policy);
2258
2259 /* crypto.c */
2260 extern struct kmem_cache *ext4_crypt_info_cachep;
2261 bool ext4_valid_contents_enc_mode(uint32_t mode);
2262 uint32_t ext4_validate_encryption_key_size(uint32_t mode, uint32_t size);
2263 extern struct workqueue_struct *ext4_read_workqueue;
2264 struct ext4_crypto_ctx *ext4_get_crypto_ctx(struct inode *inode);
2265 void ext4_release_crypto_ctx(struct ext4_crypto_ctx *ctx);
2266 void ext4_restore_control_page(struct page *data_page);
2267 struct page *ext4_encrypt(struct inode *inode,
2268                           struct page *plaintext_page);
2269 int ext4_decrypt(struct page *page);
2270 int ext4_encrypted_zeroout(struct inode *inode, struct ext4_extent *ex);
2271
2272 #ifdef CONFIG_EXT4_FS_ENCRYPTION
2273 int ext4_init_crypto(void);
2274 void ext4_exit_crypto(void);
2275 static inline int ext4_sb_has_crypto(struct super_block *sb)
2276 {
2277         return ext4_has_feature_encrypt(sb);
2278 }
2279 #else
2280 static inline int ext4_init_crypto(void) { return 0; }
2281 static inline void ext4_exit_crypto(void) { }
2282 static inline int ext4_sb_has_crypto(struct super_block *sb)
2283 {
2284         return 0;
2285 }
2286 #endif
2287
2288 /* crypto_fname.c */
2289 bool ext4_valid_filenames_enc_mode(uint32_t mode);
2290 u32 ext4_fname_crypto_round_up(u32 size, u32 blksize);
2291 unsigned ext4_fname_encrypted_size(struct inode *inode, u32 ilen);
2292 int ext4_fname_crypto_alloc_buffer(struct inode *inode,
2293                                    u32 ilen, struct ext4_str *crypto_str);
2294 int _ext4_fname_disk_to_usr(struct inode *inode,
2295                             struct dx_hash_info *hinfo,
2296                             const struct ext4_str *iname,
2297                             struct ext4_str *oname);
2298 int ext4_fname_disk_to_usr(struct inode *inode,
2299                            struct dx_hash_info *hinfo,
2300                            const struct ext4_dir_entry_2 *de,
2301                            struct ext4_str *oname);
2302 int ext4_fname_usr_to_disk(struct inode *inode,
2303                            const struct qstr *iname,
2304                            struct ext4_str *oname);
2305 #ifdef CONFIG_EXT4_FS_ENCRYPTION
2306 void ext4_fname_crypto_free_buffer(struct ext4_str *crypto_str);
2307 int ext4_fname_setup_filename(struct inode *dir, const struct qstr *iname,
2308                               int lookup, struct ext4_filename *fname);
2309 void ext4_fname_free_filename(struct ext4_filename *fname);
2310 #else
2311 static inline
2312 int ext4_setup_fname_crypto(struct inode *inode)
2313 {
2314         return 0;
2315 }
2316 static inline void ext4_fname_crypto_free_buffer(struct ext4_str *p) { }
2317 static inline int ext4_fname_setup_filename(struct inode *dir,
2318                                      const struct qstr *iname,
2319                                      int lookup, struct ext4_filename *fname)
2320 {
2321         fname->usr_fname = iname;
2322         fname->disk_name.name = (unsigned char *) iname->name;
2323         fname->disk_name.len = iname->len;
2324         return 0;
2325 }
2326 static inline void ext4_fname_free_filename(struct ext4_filename *fname) { }
2327 #endif
2328
2329
2330 /* crypto_key.c */
2331 void ext4_free_crypt_info(struct ext4_crypt_info *ci);
2332 void ext4_free_encryption_info(struct inode *inode, struct ext4_crypt_info *ci);
2333 int _ext4_get_encryption_info(struct inode *inode);
2334
2335 #ifdef CONFIG_EXT4_FS_ENCRYPTION
2336 int ext4_has_encryption_key(struct inode *inode);
2337
2338 static inline int ext4_get_encryption_info(struct inode *inode)
2339 {
2340         struct ext4_crypt_info *ci = EXT4_I(inode)->i_crypt_info;
2341
2342         if (!ci ||
2343             (ci->ci_keyring_key &&
2344              (ci->ci_keyring_key->flags & ((1 << KEY_FLAG_INVALIDATED) |
2345                                            (1 << KEY_FLAG_REVOKED) |
2346                                            (1 << KEY_FLAG_DEAD)))))
2347                 return _ext4_get_encryption_info(inode);
2348         return 0;
2349 }
2350
2351 static inline struct ext4_crypt_info *ext4_encryption_info(struct inode *inode)
2352 {
2353         return EXT4_I(inode)->i_crypt_info;
2354 }
2355
2356 #else
2357 static inline int ext4_has_encryption_key(struct inode *inode)
2358 {
2359         return 0;
2360 }
2361 static inline int ext4_get_encryption_info(struct inode *inode)
2362 {
2363         return 0;
2364 }
2365 static inline struct ext4_crypt_info *ext4_encryption_info(struct inode *inode)
2366 {
2367         return NULL;
2368 }
2369 #endif
2370
2371
2372 /* dir.c */
2373 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
2374                                   struct file *,
2375                                   struct ext4_dir_entry_2 *,
2376                                   struct buffer_head *, char *, int,
2377                                   unsigned int);
2378 #define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset)      \
2379         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
2380                                         (de), (bh), (buf), (size), (offset)))
2381 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
2382                                 __u32 minor_hash,
2383                                 struct ext4_dir_entry_2 *dirent,
2384                                 struct ext4_str *ent_name);
2385 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
2386 extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
2387                              struct buffer_head *bh,
2388                              void *buf, int buf_size,
2389                              struct ext4_filename *fname,
2390                              struct ext4_dir_entry_2 **dest_de);
2391 int ext4_insert_dentry(struct inode *dir,
2392                        struct inode *inode,
2393                        struct ext4_dir_entry_2 *de,
2394                        int buf_size,
2395                        struct ext4_filename *fname);
2396 static inline void ext4_update_dx_flag(struct inode *inode)
2397 {
2398         if (!ext4_has_feature_dir_index(inode->i_sb))
2399                 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
2400 }
2401 static unsigned char ext4_filetype_table[] = {
2402         DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
2403 };
2404
2405 static inline  unsigned char get_dtype(struct super_block *sb, int filetype)
2406 {
2407         if (!ext4_has_feature_filetype(sb) || filetype >= EXT4_FT_MAX)
2408                 return DT_UNKNOWN;
2409
2410         return ext4_filetype_table[filetype];
2411 }
2412 extern int ext4_check_all_de(struct inode *dir, struct buffer_head *bh,
2413                              void *buf, int buf_size);
2414
2415 /* fsync.c */
2416 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2417
2418 /* hash.c */
2419 extern int ext4fs_dirhash(const char *name, int len, struct
2420                           dx_hash_info *hinfo);
2421
2422 /* ialloc.c */
2423 extern struct inode *__ext4_new_inode(handle_t *, struct inode *, umode_t,
2424                                       const struct qstr *qstr, __u32 goal,
2425                                       uid_t *owner, int handle_type,
2426                                       unsigned int line_no, int nblocks);
2427
2428 #define ext4_new_inode(handle, dir, mode, qstr, goal, owner) \
2429         __ext4_new_inode((handle), (dir), (mode), (qstr), (goal), (owner), \
2430                          0, 0, 0)
2431 #define ext4_new_inode_start_handle(dir, mode, qstr, goal, owner, \
2432                                     type, nblocks)                  \
2433         __ext4_new_inode(NULL, (dir), (mode), (qstr), (goal), (owner), \
2434                          (type), __LINE__, (nblocks))
2435
2436
2437 extern void ext4_free_inode(handle_t *, struct inode *);
2438 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2439 extern unsigned long ext4_count_free_inodes(struct super_block *);
2440 extern unsigned long ext4_count_dirs(struct super_block *);
2441 extern void ext4_check_inodes_bitmap(struct super_block *);
2442 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2443 extern int ext4_init_inode_table(struct super_block *sb,
2444                                  ext4_group_t group, int barrier);
2445 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2446
2447 /* mballoc.c */
2448 extern const struct file_operations ext4_seq_mb_groups_fops;
2449 extern long ext4_mb_stats;
2450 extern long ext4_mb_max_to_scan;
2451 extern int ext4_mb_init(struct super_block *);
2452 extern int ext4_mb_release(struct super_block *);
2453 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2454                                 struct ext4_allocation_request *, int *);
2455 extern int ext4_mb_reserve_blocks(struct super_block *, int);
2456 extern void ext4_discard_preallocations(struct inode *);
2457 extern int __init ext4_init_mballoc(void);
2458 extern void ext4_exit_mballoc(void);
2459 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2460                              struct buffer_head *bh, ext4_fsblk_t block,
2461                              unsigned long count, int flags);
2462 extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2463                                    ext4_group_t ngroups);
2464 extern int ext4_mb_add_groupinfo(struct super_block *sb,
2465                 ext4_group_t i, struct ext4_group_desc *desc);
2466 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2467                                 ext4_fsblk_t block, unsigned long count);
2468 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2469
2470 /* inode.c */
2471 int ext4_inode_is_fast_symlink(struct inode *inode);
2472 struct buffer_head *ext4_getblk(handle_t *, struct inode *, ext4_lblk_t, int);
2473 struct buffer_head *ext4_bread(handle_t *, struct inode *, ext4_lblk_t, int);
2474 int ext4_get_block_write(struct inode *inode, sector_t iblock,
2475                          struct buffer_head *bh_result, int create);
2476 int ext4_get_block_dax(struct inode *inode, sector_t iblock,
2477                          struct buffer_head *bh_result, int create);
2478 int ext4_get_block(struct inode *inode, sector_t iblock,
2479                                 struct buffer_head *bh_result, int create);
2480 int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2481                            struct buffer_head *bh, int create);
2482 int ext4_walk_page_buffers(handle_t *handle,
2483                            struct buffer_head *head,
2484                            unsigned from,
2485                            unsigned to,
2486                            int *partial,
2487                            int (*fn)(handle_t *handle,
2488                                      struct buffer_head *bh));
2489 int do_journal_get_write_access(handle_t *handle,
2490                                 struct buffer_head *bh);
2491 #define FALL_BACK_TO_NONDELALLOC 1
2492 #define CONVERT_INLINE_DATA      2
2493
2494 extern struct inode *ext4_iget(struct super_block *, unsigned long);
2495 extern struct inode *ext4_iget_normal(struct super_block *, unsigned long);
2496 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
2497 extern int  ext4_setattr(struct dentry *, struct iattr *);
2498 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
2499                                 struct kstat *stat);
2500 extern void ext4_evict_inode(struct inode *);
2501 extern void ext4_clear_inode(struct inode *);
2502 extern int  ext4_sync_inode(handle_t *, struct inode *);
2503 extern void ext4_dirty_inode(struct inode *, int);
2504 extern int ext4_change_inode_journal_flag(struct inode *, int);
2505 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
2506 extern int ext4_inode_attach_jinode(struct inode *inode);
2507 extern int ext4_can_truncate(struct inode *inode);
2508 extern void ext4_truncate(struct inode *);
2509 extern int ext4_punch_hole(struct inode *inode, loff_t offset, loff_t length);
2510 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
2511 extern void ext4_set_inode_flags(struct inode *);
2512 extern void ext4_get_inode_flags(struct ext4_inode_info *);
2513 extern int ext4_alloc_da_blocks(struct inode *inode);
2514 extern void ext4_set_aops(struct inode *inode);
2515 extern int ext4_writepage_trans_blocks(struct inode *);
2516 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
2517 extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
2518                              loff_t lstart, loff_t lend);
2519 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2520 extern int ext4_filemap_fault(struct vm_area_struct *vma, struct vm_fault *vmf);
2521 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
2522 extern void ext4_da_update_reserve_space(struct inode *inode,
2523                                         int used, int quota_claim);
2524
2525 /* indirect.c */
2526 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
2527                                 struct ext4_map_blocks *map, int flags);
2528 extern ssize_t ext4_ind_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
2529                                   loff_t offset);
2530 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
2531 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
2532 extern void ext4_ind_truncate(handle_t *, struct inode *inode);
2533 extern int ext4_ind_remove_space(handle_t *handle, struct inode *inode,
2534                                  ext4_lblk_t start, ext4_lblk_t end);
2535
2536 /* ioctl.c */
2537 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
2538 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
2539
2540 /* migrate.c */
2541 extern int ext4_ext_migrate(struct inode *);
2542 extern int ext4_ind_migrate(struct inode *inode);
2543
2544 /* namei.c */
2545 extern int ext4_dirent_csum_verify(struct inode *inode,
2546                                    struct ext4_dir_entry *dirent);
2547 extern int ext4_orphan_add(handle_t *, struct inode *);
2548 extern int ext4_orphan_del(handle_t *, struct inode *);
2549 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
2550                                 __u32 start_minor_hash, __u32 *next_hash);
2551 extern int ext4_search_dir(struct buffer_head *bh,
2552                            char *search_buf,
2553                            int buf_size,
2554                            struct inode *dir,
2555                            struct ext4_filename *fname,
2556                            const struct qstr *d_name,
2557                            unsigned int offset,
2558                            struct ext4_dir_entry_2 **res_dir);
2559 extern int ext4_generic_delete_entry(handle_t *handle,
2560                                      struct inode *dir,
2561                                      struct ext4_dir_entry_2 *de_del,
2562                                      struct buffer_head *bh,
2563                                      void *entry_buf,
2564                                      int buf_size,
2565                                      int csum_size);
2566 extern int ext4_empty_dir(struct inode *inode);
2567
2568 /* resize.c */
2569 extern int ext4_group_add(struct super_block *sb,
2570                                 struct ext4_new_group_data *input);
2571 extern int ext4_group_extend(struct super_block *sb,
2572                                 struct ext4_super_block *es,
2573                                 ext4_fsblk_t n_blocks_count);
2574 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
2575
2576 /* super.c */
2577 extern int ext4_seq_options_show(struct seq_file *seq, void *offset);
2578 extern int ext4_calculate_overhead(struct super_block *sb);
2579 extern void ext4_superblock_csum_set(struct super_block *sb);
2580 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
2581 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
2582 extern int ext4_alloc_flex_bg_array(struct super_block *sb,
2583                                     ext4_group_t ngroup);
2584 extern const char *ext4_decode_error(struct super_block *sb, int errno,
2585                                      char nbuf[16]);
2586
2587 extern __printf(4, 5)
2588 void __ext4_error(struct super_block *, const char *, unsigned int,
2589                   const char *, ...);
2590 extern __printf(5, 6)
2591 void __ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
2592                       const char *, ...);
2593 extern __printf(5, 6)
2594 void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
2595                      const char *, ...);
2596 extern void __ext4_std_error(struct super_block *, const char *,
2597                              unsigned int, int);
2598 extern __printf(4, 5)
2599 void __ext4_abort(struct super_block *, const char *, unsigned int,
2600                   const char *, ...);
2601 extern __printf(4, 5)
2602 void __ext4_warning(struct super_block *, const char *, unsigned int,
2603                     const char *, ...);
2604 extern __printf(4, 5)
2605 void __ext4_warning_inode(const struct inode *inode, const char *function,
2606                           unsigned int line, const char *fmt, ...);
2607 extern __printf(3, 4)
2608 void __ext4_msg(struct super_block *, const char *, const char *, ...);
2609 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
2610                            const char *, unsigned int, const char *);
2611 extern __printf(7, 8)
2612 void __ext4_grp_locked_error(const char *, unsigned int,
2613                              struct super_block *, ext4_group_t,
2614                              unsigned long, ext4_fsblk_t,
2615                              const char *, ...);
2616
2617 #define EXT4_ERROR_INODE(inode, fmt, a...) \
2618         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
2619
2620 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...)                 \
2621         ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
2622
2623 #define EXT4_ERROR_FILE(file, block, fmt, a...)                         \
2624         ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
2625
2626 #ifdef CONFIG_PRINTK
2627
2628 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2629         __ext4_error_inode(inode, func, line, block, fmt, ##__VA_ARGS__)
2630 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2631         __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__)
2632 #define ext4_error(sb, fmt, ...)                                        \
2633         __ext4_error(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2634 #define ext4_abort(sb, fmt, ...)                                        \
2635         __ext4_abort(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2636 #define ext4_warning(sb, fmt, ...)                                      \
2637         __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2638 #define ext4_warning_inode(inode, fmt, ...)                             \
2639         __ext4_warning_inode(inode, __func__, __LINE__, fmt, ##__VA_ARGS__)
2640 #define ext4_msg(sb, level, fmt, ...)                           \
2641         __ext4_msg(sb, level, fmt, ##__VA_ARGS__)
2642 #define dump_mmp_msg(sb, mmp, msg)                                      \
2643         __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg)
2644 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2645         __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \
2646                                 fmt, ##__VA_ARGS__)
2647
2648 #else
2649
2650 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2651 do {                                                                    \
2652         no_printk(fmt, ##__VA_ARGS__);                                  \
2653         __ext4_error_inode(inode, "", 0, block, " ");                   \
2654 } while (0)
2655 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2656 do {                                                                    \
2657         no_printk(fmt, ##__VA_ARGS__);                                  \
2658         __ext4_error_file(file, "", 0, block, " ");                     \
2659 } while (0)
2660 #define ext4_error(sb, fmt, ...)                                        \
2661 do {                                                                    \
2662         no_printk(fmt, ##__VA_ARGS__);                                  \
2663         __ext4_error(sb, "", 0, " ");                                   \
2664 } while (0)
2665 #define ext4_abort(sb, fmt, ...)                                        \
2666 do {                                                                    \
2667         no_printk(fmt, ##__VA_ARGS__);                                  \
2668         __ext4_abort(sb, "", 0, " ");                                   \
2669 } while (0)
2670 #define ext4_warning(sb, fmt, ...)                                      \
2671 do {                                                                    \
2672         no_printk(fmt, ##__VA_ARGS__);                                  \
2673         __ext4_warning(sb, "", 0, " ");                                 \
2674 } while (0)
2675 #define ext4_warning_inode(inode, fmt, ...)                             \
2676 do {                                                                    \
2677         no_printk(fmt, ##__VA_ARGS__);                                  \
2678         __ext4_warning_inode(inode, "", 0, " ");                        \
2679 } while (0)
2680 #define ext4_msg(sb, level, fmt, ...)                                   \
2681 do {                                                                    \
2682         no_printk(fmt, ##__VA_ARGS__);                                  \
2683         __ext4_msg(sb, "", " ");                                        \
2684 } while (0)
2685 #define dump_mmp_msg(sb, mmp, msg)                                      \
2686         __dump_mmp_msg(sb, mmp, "", 0, "")
2687 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2688 do {                                                                    \
2689         no_printk(fmt, ##__VA_ARGS__);                          \
2690         __ext4_grp_locked_error("", 0, sb, grp, ino, block, " ");       \
2691 } while (0)
2692
2693 #endif
2694
2695 extern void ext4_update_dynamic_rev(struct super_block *sb);
2696 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
2697                                         __u32 compat);
2698 extern int ext4_update_rocompat_feature(handle_t *handle,
2699                                         struct super_block *sb, __u32 rocompat);
2700 extern int ext4_update_incompat_feature(handle_t *handle,
2701                                         struct super_block *sb, __u32 incompat);
2702 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
2703                                       struct ext4_group_desc *bg);
2704 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
2705                                       struct ext4_group_desc *bg);
2706 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
2707                                      struct ext4_group_desc *bg);
2708 extern __u32 ext4_free_group_clusters(struct super_block *sb,
2709                                       struct ext4_group_desc *bg);
2710 extern __u32 ext4_free_inodes_count(struct super_block *sb,
2711                                  struct ext4_group_desc *bg);
2712 extern __u32 ext4_used_dirs_count(struct super_block *sb,
2713                                 struct ext4_group_desc *bg);
2714 extern __u32 ext4_itable_unused_count(struct super_block *sb,
2715                                    struct ext4_group_desc *bg);
2716 extern void ext4_block_bitmap_set(struct super_block *sb,
2717                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2718 extern void ext4_inode_bitmap_set(struct super_block *sb,
2719                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2720 extern void ext4_inode_table_set(struct super_block *sb,
2721                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
2722 extern void ext4_free_group_clusters_set(struct super_block *sb,
2723                                          struct ext4_group_desc *bg,
2724                                          __u32 count);
2725 extern void ext4_free_inodes_set(struct super_block *sb,
2726                                 struct ext4_group_desc *bg, __u32 count);
2727 extern void ext4_used_dirs_set(struct super_block *sb,
2728                                 struct ext4_group_desc *bg, __u32 count);
2729 extern void ext4_itable_unused_set(struct super_block *sb,
2730                                    struct ext4_group_desc *bg, __u32 count);
2731 extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
2732                                        struct ext4_group_desc *gdp);
2733 extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
2734                                      struct ext4_group_desc *gdp);
2735 extern int ext4_register_li_request(struct super_block *sb,
2736                                     ext4_group_t first_not_zeroed);
2737
2738 static inline int ext4_has_group_desc_csum(struct super_block *sb)
2739 {
2740         return ext4_has_feature_gdt_csum(sb) ||
2741                EXT4_SB(sb)->s_chksum_driver != NULL;
2742 }
2743
2744 static inline int ext4_has_metadata_csum(struct super_block *sb)
2745 {
2746         WARN_ON_ONCE(ext4_has_feature_metadata_csum(sb) &&
2747                      !EXT4_SB(sb)->s_chksum_driver);
2748
2749         return (EXT4_SB(sb)->s_chksum_driver != NULL);
2750 }
2751 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2752 {
2753         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2754                 le32_to_cpu(es->s_blocks_count_lo);
2755 }
2756
2757 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2758 {
2759         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2760                 le32_to_cpu(es->s_r_blocks_count_lo);
2761 }
2762
2763 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2764 {
2765         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2766                 le32_to_cpu(es->s_free_blocks_count_lo);
2767 }
2768
2769 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2770                                          ext4_fsblk_t blk)
2771 {
2772         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2773         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2774 }
2775
2776 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2777                                               ext4_fsblk_t blk)
2778 {
2779         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2780         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2781 }
2782
2783 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2784                                            ext4_fsblk_t blk)
2785 {
2786         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2787         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2788 }
2789
2790 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2791 {
2792         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2793                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2794                         le32_to_cpu(raw_inode->i_size_lo);
2795         else
2796                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2797 }
2798
2799 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2800 {
2801         raw_inode->i_size_lo = cpu_to_le32(i_size);
2802         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2803 }
2804
2805 static inline
2806 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2807                                             ext4_group_t group)
2808 {
2809          struct ext4_group_info ***grp_info;
2810          long indexv, indexh;
2811          BUG_ON(group >= EXT4_SB(sb)->s_groups_count);
2812          grp_info = EXT4_SB(sb)->s_group_info;
2813          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2814          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2815          return grp_info[indexv][indexh];
2816 }
2817
2818 /*
2819  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2820  * the locking protocol documented in the comments of ext4_group_add()
2821  * in resize.c
2822  */
2823 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2824 {
2825         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2826
2827         smp_rmb();
2828         return ngroups;
2829 }
2830
2831 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2832                                              ext4_group_t block_group)
2833 {
2834         return block_group >> sbi->s_log_groups_per_flex;
2835 }
2836
2837 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2838 {
2839         return 1 << sbi->s_log_groups_per_flex;
2840 }
2841
2842 #define ext4_std_error(sb, errno)                               \
2843 do {                                                            \
2844         if ((errno))                                            \
2845                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2846 } while (0)
2847
2848 #ifdef CONFIG_SMP
2849 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2850  * counters. So we need to make sure we have free clusters more
2851  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2852  */
2853 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2854 #else
2855 #define EXT4_FREECLUSTERS_WATERMARK 0
2856 #endif
2857
2858 /* Update i_disksize. Requires i_mutex to avoid races with truncate */
2859 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2860 {
2861         WARN_ON_ONCE(S_ISREG(inode->i_mode) &&
2862                      !mutex_is_locked(&inode->i_mutex));
2863         down_write(&EXT4_I(inode)->i_data_sem);
2864         if (newsize > EXT4_I(inode)->i_disksize)
2865                 EXT4_I(inode)->i_disksize = newsize;
2866         up_write(&EXT4_I(inode)->i_data_sem);
2867 }
2868
2869 /* Update i_size, i_disksize. Requires i_mutex to avoid races with truncate */
2870 static inline int ext4_update_inode_size(struct inode *inode, loff_t newsize)
2871 {
2872         int changed = 0;
2873
2874         if (newsize > inode->i_size) {
2875                 i_size_write(inode, newsize);
2876                 changed = 1;
2877         }
2878         if (newsize > EXT4_I(inode)->i_disksize) {
2879                 ext4_update_i_disksize(inode, newsize);
2880                 changed |= 2;
2881         }
2882         return changed;
2883 }
2884
2885 int ext4_update_disksize_before_punch(struct inode *inode, loff_t offset,
2886                                       loff_t len);
2887
2888 struct ext4_group_info {
2889         unsigned long   bb_state;
2890         struct rb_root  bb_free_root;
2891         ext4_grpblk_t   bb_first_free;  /* first free block */
2892         ext4_grpblk_t   bb_free;        /* total free blocks */
2893         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2894         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2895         struct          list_head bb_prealloc_list;
2896 #ifdef DOUBLE_CHECK
2897         void            *bb_bitmap;
2898 #endif
2899         struct rw_semaphore alloc_sem;
2900         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2901                                          * regions, index is order.
2902                                          * bb_counters[3] = 5 means
2903                                          * 5 free 8-block regions. */
2904 };
2905
2906 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2907 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2908 #define EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT     2
2909 #define EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT     3
2910
2911 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2912         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2913 #define EXT4_MB_GRP_BBITMAP_CORRUPT(grp)        \
2914         (test_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
2915 #define EXT4_MB_GRP_IBITMAP_CORRUPT(grp)        \
2916         (test_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
2917
2918 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2919         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2920 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2921         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2922 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2923         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2924
2925 #define EXT4_MAX_CONTENTION             8
2926 #define EXT4_CONTENTION_THRESHOLD       2
2927
2928 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2929                                               ext4_group_t group)
2930 {
2931         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2932 }
2933
2934 /*
2935  * Returns true if the filesystem is busy enough that attempts to
2936  * access the block group locks has run into contention.
2937  */
2938 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2939 {
2940         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2941 }
2942
2943 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2944 {
2945         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2946         if (spin_trylock(lock))
2947                 /*
2948                  * We're able to grab the lock right away, so drop the
2949                  * lock contention counter.
2950                  */
2951                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2952         else {
2953                 /*
2954                  * The lock is busy, so bump the contention counter,
2955                  * and then wait on the spin lock.
2956                  */
2957                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2958                                   EXT4_MAX_CONTENTION);
2959                 spin_lock(lock);
2960         }
2961 }
2962
2963 static inline void ext4_unlock_group(struct super_block *sb,
2964                                         ext4_group_t group)
2965 {
2966         spin_unlock(ext4_group_lock_ptr(sb, group));
2967 }
2968
2969 /*
2970  * Block validity checking
2971  */
2972 #define ext4_check_indirect_blockref(inode, bh)                         \
2973         ext4_check_blockref(__func__, __LINE__, inode,                  \
2974                             (__le32 *)(bh)->b_data,                     \
2975                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2976
2977 #define ext4_ind_check_inode(inode)                                     \
2978         ext4_check_blockref(__func__, __LINE__, inode,                  \
2979                             EXT4_I(inode)->i_data,                      \
2980                             EXT4_NDIR_BLOCKS)
2981
2982 /*
2983  * Inodes and files operations
2984  */
2985
2986 /* dir.c */
2987 extern const struct file_operations ext4_dir_operations;
2988
2989 /* file.c */
2990 extern const struct inode_operations ext4_file_inode_operations;
2991 extern const struct file_operations ext4_file_operations;
2992 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2993
2994 /* inline.c */
2995 extern int ext4_get_max_inline_size(struct inode *inode);
2996 extern int ext4_find_inline_data_nolock(struct inode *inode);
2997 extern int ext4_init_inline_data(handle_t *handle, struct inode *inode,
2998                                  unsigned int len);
2999 extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
3000
3001 extern int ext4_readpage_inline(struct inode *inode, struct page *page);
3002 extern int ext4_try_to_write_inline_data(struct address_space *mapping,
3003                                          struct inode *inode,
3004                                          loff_t pos, unsigned len,
3005                                          unsigned flags,
3006                                          struct page **pagep);
3007 extern int ext4_write_inline_data_end(struct inode *inode,
3008                                       loff_t pos, unsigned len,
3009                                       unsigned copied,
3010                                       struct page *page);
3011 extern struct buffer_head *
3012 ext4_journalled_write_inline_data(struct inode *inode,
3013                                   unsigned len,
3014                                   struct page *page);
3015 extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
3016                                            struct inode *inode,
3017                                            loff_t pos, unsigned len,
3018                                            unsigned flags,
3019                                            struct page **pagep,
3020                                            void **fsdata);
3021 extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
3022                                          unsigned len, unsigned copied,
3023                                          struct page *page);
3024 extern int ext4_try_add_inline_entry(handle_t *handle,
3025                                      struct ext4_filename *fname,
3026                                      struct dentry *dentry,
3027                                      struct inode *inode);
3028 extern int ext4_try_create_inline_dir(handle_t *handle,
3029                                       struct inode *parent,
3030                                       struct inode *inode);
3031 extern int ext4_read_inline_dir(struct file *filp,
3032                                 struct dir_context *ctx,
3033                                 int *has_inline_data);
3034 extern int htree_inlinedir_to_tree(struct file *dir_file,
3035                                    struct inode *dir, ext4_lblk_t block,
3036                                    struct dx_hash_info *hinfo,
3037                                    __u32 start_hash, __u32 start_minor_hash,
3038                                    int *has_inline_data);
3039 extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
3040                                         struct ext4_filename *fname,
3041                                         const struct qstr *d_name,
3042                                         struct ext4_dir_entry_2 **res_dir,
3043                                         int *has_inline_data);
3044 extern int ext4_delete_inline_entry(handle_t *handle,
3045                                     struct inode *dir,
3046                                     struct ext4_dir_entry_2 *de_del,
3047                                     struct buffer_head *bh,
3048                                     int *has_inline_data);
3049 extern int empty_inline_dir(struct inode *dir, int *has_inline_data);
3050 extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
3051                                         struct ext4_dir_entry_2 **parent_de,
3052                                         int *retval);
3053 extern int ext4_inline_data_fiemap(struct inode *inode,
3054                                    struct fiemap_extent_info *fieinfo,
3055                                    int *has_inline, __u64 start, __u64 len);
3056 extern int ext4_try_to_evict_inline_data(handle_t *handle,
3057                                          struct inode *inode,
3058                                          int needed);
3059 extern void ext4_inline_data_truncate(struct inode *inode, int *has_inline);
3060
3061 extern int ext4_convert_inline_data(struct inode *inode);
3062
3063 static inline int ext4_has_inline_data(struct inode *inode)
3064 {
3065         return ext4_test_inode_flag(inode, EXT4_INODE_INLINE_DATA) &&
3066                EXT4_I(inode)->i_inline_off;
3067 }
3068
3069 /* namei.c */
3070 extern const struct inode_operations ext4_dir_inode_operations;
3071 extern const struct inode_operations ext4_special_inode_operations;
3072 extern struct dentry *ext4_get_parent(struct dentry *child);
3073 extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
3074                                  struct ext4_dir_entry_2 *de,
3075                                  int blocksize, int csum_size,
3076                                  unsigned int parent_ino, int dotdot_real_len);
3077 extern void initialize_dirent_tail(struct ext4_dir_entry_tail *t,
3078                                    unsigned int blocksize);
3079 extern int ext4_handle_dirty_dirent_node(handle_t *handle,
3080                                          struct inode *inode,
3081                                          struct buffer_head *bh);
3082 #define S_SHIFT 12
3083 static unsigned char ext4_type_by_mode[S_IFMT >> S_SHIFT] = {
3084         [S_IFREG >> S_SHIFT]    = EXT4_FT_REG_FILE,
3085         [S_IFDIR >> S_SHIFT]    = EXT4_FT_DIR,
3086         [S_IFCHR >> S_SHIFT]    = EXT4_FT_CHRDEV,
3087         [S_IFBLK >> S_SHIFT]    = EXT4_FT_BLKDEV,
3088         [S_IFIFO >> S_SHIFT]    = EXT4_FT_FIFO,
3089         [S_IFSOCK >> S_SHIFT]   = EXT4_FT_SOCK,
3090         [S_IFLNK >> S_SHIFT]    = EXT4_FT_SYMLINK,
3091 };
3092
3093 static inline void ext4_set_de_type(struct super_block *sb,
3094                                 struct ext4_dir_entry_2 *de,
3095                                 umode_t mode) {
3096         if (ext4_has_feature_filetype(sb))
3097                 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
3098 }
3099
3100 /* readpages.c */
3101 extern int ext4_mpage_readpages(struct address_space *mapping,
3102                                 struct list_head *pages, struct page *page,
3103                                 unsigned nr_pages);
3104
3105 /* symlink.c */
3106 extern const struct inode_operations ext4_encrypted_symlink_inode_operations;
3107 extern const struct inode_operations ext4_symlink_inode_operations;
3108 extern const struct inode_operations ext4_fast_symlink_inode_operations;
3109
3110 /* sysfs.c */
3111 extern int ext4_register_sysfs(struct super_block *sb);
3112 extern void ext4_unregister_sysfs(struct super_block *sb);
3113 extern int __init ext4_init_sysfs(void);
3114 extern void ext4_exit_sysfs(void);
3115
3116 /* block_validity */
3117 extern void ext4_release_system_zone(struct super_block *sb);
3118 extern int ext4_setup_system_zone(struct super_block *sb);
3119 extern int __init ext4_init_system_zone(void);
3120 extern void ext4_exit_system_zone(void);
3121 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
3122                                  ext4_fsblk_t start_blk,
3123                                  unsigned int count);
3124 extern int ext4_check_blockref(const char *, unsigned int,
3125                                struct inode *, __le32 *, unsigned int);
3126
3127 /* extents.c */
3128 struct ext4_ext_path;
3129 struct ext4_extent;
3130
3131 /*
3132  * Maximum number of logical blocks in a file; ext4_extent's ee_block is
3133  * __le32.
3134  */
3135 #define EXT_MAX_BLOCKS  0xffffffff
3136
3137 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
3138 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
3139 extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
3140 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
3141                                struct ext4_map_blocks *map, int flags);
3142 extern void ext4_ext_truncate(handle_t *, struct inode *);
3143 extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
3144                                  ext4_lblk_t end);
3145 extern void ext4_ext_init(struct super_block *);
3146 extern void ext4_ext_release(struct super_block *);
3147 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
3148                           loff_t len);
3149 extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
3150                                           loff_t offset, ssize_t len);
3151 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
3152                            struct ext4_map_blocks *map, int flags);
3153 extern int ext4_ext_calc_metadata_amount(struct inode *inode,
3154                                          ext4_lblk_t lblocks);
3155 extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
3156                                                    int num,
3157                                                    struct ext4_ext_path *path);
3158 extern int ext4_can_extents_be_merged(struct inode *inode,
3159                                       struct ext4_extent *ex1,
3160                                       struct ext4_extent *ex2);
3161 extern int ext4_ext_insert_extent(handle_t *, struct inode *,
3162                                   struct ext4_ext_path **,
3163                                   struct ext4_extent *, int);
3164 extern struct ext4_ext_path *ext4_find_extent(struct inode *, ext4_lblk_t,
3165                                               struct ext4_ext_path **,
3166                                               int flags);
3167 extern void ext4_ext_drop_refs(struct ext4_ext_path *);
3168 extern int ext4_ext_check_inode(struct inode *inode);
3169 extern int ext4_find_delalloc_range(struct inode *inode,
3170                                     ext4_lblk_t lblk_start,
3171                                     ext4_lblk_t lblk_end);
3172 extern int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk);
3173 extern ext4_lblk_t ext4_ext_next_allocated_block(struct ext4_ext_path *path);
3174 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
3175                         __u64 start, __u64 len);
3176 extern int ext4_ext_precache(struct inode *inode);
3177 extern int ext4_collapse_range(struct inode *inode, loff_t offset, loff_t len);
3178 extern int ext4_insert_range(struct inode *inode, loff_t offset, loff_t len);
3179 extern int ext4_swap_extents(handle_t *handle, struct inode *inode1,
3180                                 struct inode *inode2, ext4_lblk_t lblk1,
3181                              ext4_lblk_t lblk2,  ext4_lblk_t count,
3182                              int mark_unwritten,int *err);
3183
3184 /* move_extent.c */
3185 extern void ext4_double_down_write_data_sem(struct inode *first,
3186                                             struct inode *second);
3187 extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
3188                                           struct inode *donor_inode);
3189 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
3190                              __u64 start_orig, __u64 start_donor,
3191                              __u64 len, __u64 *moved_len);
3192
3193 /* page-io.c */
3194 extern int __init ext4_init_pageio(void);
3195 extern void ext4_exit_pageio(void);
3196 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
3197 extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
3198 extern int ext4_put_io_end(ext4_io_end_t *io_end);
3199 extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
3200 extern void ext4_io_submit_init(struct ext4_io_submit *io,
3201                                 struct writeback_control *wbc);
3202 extern void ext4_end_io_rsv_work(struct work_struct *work);
3203 extern void ext4_io_submit(struct ext4_io_submit *io);
3204 extern int ext4_bio_write_page(struct ext4_io_submit *io,
3205                                struct page *page,
3206                                int len,
3207                                struct writeback_control *wbc,
3208                                bool keep_towrite);
3209
3210 /* mmp.c */
3211 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
3212
3213 /*
3214  * Add new method to test whether block and inode bitmaps are properly
3215  * initialized. With uninit_bg reading the block from disk is not enough
3216  * to mark the bitmap uptodate. We need to also zero-out the bitmap
3217  */
3218 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
3219
3220 static inline int bitmap_uptodate(struct buffer_head *bh)
3221 {
3222         return (buffer_uptodate(bh) &&
3223                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
3224 }
3225 static inline void set_bitmap_uptodate(struct buffer_head *bh)
3226 {
3227         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
3228 }
3229
3230 /*
3231  * Disable DIO read nolock optimization, so new dioreaders will be forced
3232  * to grab i_mutex
3233  */
3234 static inline void ext4_inode_block_unlocked_dio(struct inode *inode)
3235 {
3236         ext4_set_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
3237         smp_mb();
3238 }
3239 static inline void ext4_inode_resume_unlocked_dio(struct inode *inode)
3240 {
3241         smp_mb();
3242         ext4_clear_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
3243 }
3244
3245 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
3246
3247 /* For ioend & aio unwritten conversion wait queues */
3248 #define EXT4_WQ_HASH_SZ         37
3249 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
3250                                             EXT4_WQ_HASH_SZ])
3251 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
3252                                              EXT4_WQ_HASH_SZ])
3253 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
3254 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
3255
3256 #define EXT4_RESIZING   0
3257 extern int ext4_resize_begin(struct super_block *sb);
3258 extern void ext4_resize_end(struct super_block *sb);
3259
3260 #endif  /* __KERNEL__ */
3261
3262 #define EFSBADCRC       EBADMSG         /* Bad CRC detected */
3263 #define EFSCORRUPTED    EUCLEAN         /* Filesystem is corrupted */
3264
3265 #endif  /* _EXT4_H */