02b764b4e6354b42dfb5108d0dde55fdd5c51531
[linux-drm-fsl-dcu.git] / fs / ext4 / ext4.h
1 /*
2  *  ext4.h
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/include/linux/minix_fs.h
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  */
15
16 #ifndef _EXT4_H
17 #define _EXT4_H
18
19 #include <linux/types.h>
20 #include <linux/blkdev.h>
21 #include <linux/magic.h>
22 #include <linux/jbd2.h>
23 #include <linux/quota.h>
24 #include <linux/rwsem.h>
25 #include <linux/rbtree.h>
26 #include <linux/seqlock.h>
27 #include <linux/mutex.h>
28 #include <linux/timer.h>
29 #include <linux/wait.h>
30 #include <linux/blockgroup_lock.h>
31 #include <linux/percpu_counter.h>
32 #include <crypto/hash.h>
33 #ifdef __KERNEL__
34 #include <linux/compat.h>
35 #endif
36
37 /*
38  * The fourth extended filesystem constants/structures
39  */
40
41 /*
42  * Define EXT4FS_DEBUG to produce debug messages
43  */
44 #undef EXT4FS_DEBUG
45
46 /*
47  * Debug code
48  */
49 #ifdef EXT4FS_DEBUG
50 #define ext4_debug(f, a...)                                             \
51         do {                                                            \
52                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
53                         __FILE__, __LINE__, __func__);                  \
54                 printk(KERN_DEBUG f, ## a);                             \
55         } while (0)
56 #else
57 #define ext4_debug(fmt, ...)    no_printk(fmt, ##__VA_ARGS__)
58 #endif
59
60 /*
61  * Turn on EXT_DEBUG to get lots of info about extents operations.
62  */
63 #define EXT_DEBUG__
64 #ifdef EXT_DEBUG
65 #define ext_debug(fmt, ...)     printk(fmt, ##__VA_ARGS__)
66 #else
67 #define ext_debug(fmt, ...)     no_printk(fmt, ##__VA_ARGS__)
68 #endif
69
70 #define EXT4_ERROR_INODE(inode, fmt, a...) \
71         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
72
73 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...)                 \
74         ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
75
76 #define EXT4_ERROR_FILE(file, block, fmt, a...)                         \
77         ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
78
79 /* data type for block offset of block group */
80 typedef int ext4_grpblk_t;
81
82 /* data type for filesystem-wide blocks number */
83 typedef unsigned long long ext4_fsblk_t;
84
85 /* data type for file logical block number */
86 typedef __u32 ext4_lblk_t;
87
88 /* data type for block group number */
89 typedef unsigned int ext4_group_t;
90
91 /*
92  * Flags used in mballoc's allocation_context flags field.
93  *
94  * Also used to show what's going on for debugging purposes when the
95  * flag field is exported via the traceport interface
96  */
97
98 /* prefer goal again. length */
99 #define EXT4_MB_HINT_MERGE              0x0001
100 /* blocks already reserved */
101 #define EXT4_MB_HINT_RESERVED           0x0002
102 /* metadata is being allocated */
103 #define EXT4_MB_HINT_METADATA           0x0004
104 /* first blocks in the file */
105 #define EXT4_MB_HINT_FIRST              0x0008
106 /* search for the best chunk */
107 #define EXT4_MB_HINT_BEST               0x0010
108 /* data is being allocated */
109 #define EXT4_MB_HINT_DATA               0x0020
110 /* don't preallocate (for tails) */
111 #define EXT4_MB_HINT_NOPREALLOC         0x0040
112 /* allocate for locality group */
113 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
114 /* allocate goal blocks or none */
115 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
116 /* goal is meaningful */
117 #define EXT4_MB_HINT_TRY_GOAL           0x0200
118 /* blocks already pre-reserved by delayed allocation */
119 #define EXT4_MB_DELALLOC_RESERVED       0x0400
120 /* We are doing stream allocation */
121 #define EXT4_MB_STREAM_ALLOC            0x0800
122 /* Use reserved root blocks if needed */
123 #define EXT4_MB_USE_ROOT_BLOCKS         0x1000
124 /* Use blocks from reserved pool */
125 #define EXT4_MB_USE_RESERVED            0x2000
126
127 struct ext4_allocation_request {
128         /* target inode for block we're allocating */
129         struct inode *inode;
130         /* how many blocks we want to allocate */
131         unsigned int len;
132         /* logical block in target inode */
133         ext4_lblk_t logical;
134         /* the closest logical allocated block to the left */
135         ext4_lblk_t lleft;
136         /* the closest logical allocated block to the right */
137         ext4_lblk_t lright;
138         /* phys. target (a hint) */
139         ext4_fsblk_t goal;
140         /* phys. block for the closest logical allocated block to the left */
141         ext4_fsblk_t pleft;
142         /* phys. block for the closest logical allocated block to the right */
143         ext4_fsblk_t pright;
144         /* flags. see above EXT4_MB_HINT_* */
145         unsigned int flags;
146 };
147
148 /*
149  * Logical to physical block mapping, used by ext4_map_blocks()
150  *
151  * This structure is used to pass requests into ext4_map_blocks() as
152  * well as to store the information returned by ext4_map_blocks().  It
153  * takes less room on the stack than a struct buffer_head.
154  */
155 #define EXT4_MAP_NEW            (1 << BH_New)
156 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
157 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
158 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
159 #define EXT4_MAP_UNINIT         (1 << BH_Uninit)
160 /* Sometimes (in the bigalloc case, from ext4_da_get_block_prep) the caller of
161  * ext4_map_blocks wants to know whether or not the underlying cluster has
162  * already been accounted for. EXT4_MAP_FROM_CLUSTER conveys to the caller that
163  * the requested mapping was from previously mapped (or delayed allocated)
164  * cluster. We use BH_AllocFromCluster only for this flag. BH_AllocFromCluster
165  * should never appear on buffer_head's state flags.
166  */
167 #define EXT4_MAP_FROM_CLUSTER   (1 << BH_AllocFromCluster)
168 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
169                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
170                                  EXT4_MAP_UNINIT | EXT4_MAP_FROM_CLUSTER)
171
172 struct ext4_map_blocks {
173         ext4_fsblk_t m_pblk;
174         ext4_lblk_t m_lblk;
175         unsigned int m_len;
176         unsigned int m_flags;
177 };
178
179 /*
180  * Flags for ext4_io_end->flags
181  */
182 #define EXT4_IO_END_UNWRITTEN   0x0001
183 #define EXT4_IO_END_DIRECT      0x0002
184
185 /*
186  * For converting uninitialized extents on a work queue. 'handle' is used for
187  * buffered writeback.
188  */
189 typedef struct ext4_io_end {
190         struct list_head        list;           /* per-file finished IO list */
191         handle_t                *handle;        /* handle reserved for extent
192                                                  * conversion */
193         struct inode            *inode;         /* file being written to */
194         struct bio              *bio;           /* Linked list of completed
195                                                  * bios covering the extent */
196         unsigned int            flag;           /* unwritten or not */
197         loff_t                  offset;         /* offset in the file */
198         ssize_t                 size;           /* size of the extent */
199         struct kiocb            *iocb;          /* iocb struct for AIO */
200         int                     result;         /* error value for AIO */
201         atomic_t                count;          /* reference counter */
202 } ext4_io_end_t;
203
204 struct ext4_io_submit {
205         int                     io_op;
206         struct bio              *io_bio;
207         ext4_io_end_t           *io_end;
208         sector_t                io_next_block;
209 };
210
211 /*
212  * Special inodes numbers
213  */
214 #define EXT4_BAD_INO             1      /* Bad blocks inode */
215 #define EXT4_ROOT_INO            2      /* Root inode */
216 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
217 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
218 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
219 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
220 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
221 #define EXT4_JOURNAL_INO         8      /* Journal inode */
222
223 /* First non-reserved inode for old ext4 filesystems */
224 #define EXT4_GOOD_OLD_FIRST_INO 11
225
226 /*
227  * Maximal count of links to a file
228  */
229 #define EXT4_LINK_MAX           65000
230
231 /*
232  * Macro-instructions used to manage several block sizes
233  */
234 #define EXT4_MIN_BLOCK_SIZE             1024
235 #define EXT4_MAX_BLOCK_SIZE             65536
236 #define EXT4_MIN_BLOCK_LOG_SIZE         10
237 #define EXT4_MAX_BLOCK_LOG_SIZE         16
238 #ifdef __KERNEL__
239 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
240 #else
241 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
242 #endif
243 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
244 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
245                                          EXT4_SB(s)->s_cluster_bits)
246 #ifdef __KERNEL__
247 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
248 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
249 #else
250 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
251 #endif
252 #ifdef __KERNEL__
253 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
254 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
255 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
256 #else
257 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
258                                  EXT4_GOOD_OLD_INODE_SIZE : \
259                                  (s)->s_inode_size)
260 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
261                                  EXT4_GOOD_OLD_FIRST_INO : \
262                                  (s)->s_first_ino)
263 #endif
264 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
265
266 /* Translate a block number to a cluster number */
267 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
268 /* Translate a cluster number to a block number */
269 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
270 /* Translate # of blks to # of clusters */
271 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
272                                  (sbi)->s_cluster_bits)
273
274 /*
275  * Structure of a blocks group descriptor
276  */
277 struct ext4_group_desc
278 {
279         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
280         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
281         __le32  bg_inode_table_lo;      /* Inodes table block */
282         __le16  bg_free_blocks_count_lo;/* Free blocks count */
283         __le16  bg_free_inodes_count_lo;/* Free inodes count */
284         __le16  bg_used_dirs_count_lo;  /* Directories count */
285         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
286         __le32  bg_exclude_bitmap_lo;   /* Exclude bitmap for snapshots */
287         __le16  bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
288         __le16  bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
289         __le16  bg_itable_unused_lo;    /* Unused inodes count */
290         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
291         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
292         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
293         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
294         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
295         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
296         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
297         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
298         __le32  bg_exclude_bitmap_hi;   /* Exclude bitmap block MSB */
299         __le16  bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
300         __le16  bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
301         __u32   bg_reserved;
302 };
303
304 #define EXT4_BG_INODE_BITMAP_CSUM_HI_END        \
305         (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
306          sizeof(__le16))
307 #define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END        \
308         (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
309          sizeof(__le16))
310
311 /*
312  * Structure of a flex block group info
313  */
314
315 struct flex_groups {
316         atomic64_t      free_clusters;
317         atomic_t        free_inodes;
318         atomic_t        used_dirs;
319 };
320
321 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
322 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
323 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
324
325 /*
326  * Macro-instructions used to manage group descriptors
327  */
328 #define EXT4_MIN_DESC_SIZE              32
329 #define EXT4_MIN_DESC_SIZE_64BIT        64
330 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
331 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
332 #ifdef __KERNEL__
333 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
334 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
335 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
336 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
337 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
338 #else
339 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
340 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
341 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
342 #endif
343
344 /*
345  * Constants relative to the data blocks
346  */
347 #define EXT4_NDIR_BLOCKS                12
348 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
349 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
350 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
351 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
352
353 /*
354  * Inode flags
355  */
356 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
357 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
358 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
359 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
360 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
361 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
362 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
363 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
364 /* Reserved for compression usage... */
365 #define EXT4_DIRTY_FL                   0x00000100
366 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
367 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
368 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
369 /* End compression flags --- maybe not all used */
370 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
371 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
372 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
373 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
374 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
375 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
376 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
377 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
378 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
379 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
380 #define EXT4_INLINE_DATA_FL             0x10000000 /* Inode has inline data. */
381 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
382
383 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
384 #define EXT4_FL_USER_MODIFIABLE         0x004380FF /* User modifiable flags */
385
386 /* Flags that should be inherited by new inodes from their parent. */
387 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
388                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
389                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
390                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
391
392 /* Flags that are appropriate for regular files (all but dir-specific ones). */
393 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
394
395 /* Flags that are appropriate for non-directories/regular files. */
396 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
397
398 /* Mask out flags that are inappropriate for the given type of inode. */
399 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
400 {
401         if (S_ISDIR(mode))
402                 return flags;
403         else if (S_ISREG(mode))
404                 return flags & EXT4_REG_FLMASK;
405         else
406                 return flags & EXT4_OTHER_FLMASK;
407 }
408
409 /*
410  * Inode flags used for atomic set/get
411  */
412 enum {
413         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
414         EXT4_INODE_UNRM         = 1,    /* Undelete */
415         EXT4_INODE_COMPR        = 2,    /* Compress file */
416         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
417         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
418         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
419         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
420         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
421 /* Reserved for compression usage... */
422         EXT4_INODE_DIRTY        = 8,
423         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
424         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
425         EXT4_INODE_ECOMPR       = 11,   /* Compression error */
426 /* End compression flags --- maybe not all used */
427         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
428         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
429         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
430         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
431         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
432         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
433         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
434         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
435         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
436         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
437         EXT4_INODE_INLINE_DATA  = 28,   /* Data in inode. */
438         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
439 };
440
441 /*
442  * Since it's pretty easy to mix up bit numbers and hex values, we use a
443  * build-time check to make sure that EXT4_XXX_FL is consistent with respect to
444  * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost
445  * any extra space in the compiled kernel image, otherwise, the build will fail.
446  * It's important that these values are the same, since we are using
447  * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent
448  * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk
449  * values found in ext2, ext3 and ext4 filesystems, and of course the values
450  * defined in e2fsprogs.
451  *
452  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
453  */
454 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
455 #define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG))
456
457 static inline void ext4_check_flag_values(void)
458 {
459         CHECK_FLAG_VALUE(SECRM);
460         CHECK_FLAG_VALUE(UNRM);
461         CHECK_FLAG_VALUE(COMPR);
462         CHECK_FLAG_VALUE(SYNC);
463         CHECK_FLAG_VALUE(IMMUTABLE);
464         CHECK_FLAG_VALUE(APPEND);
465         CHECK_FLAG_VALUE(NODUMP);
466         CHECK_FLAG_VALUE(NOATIME);
467         CHECK_FLAG_VALUE(DIRTY);
468         CHECK_FLAG_VALUE(COMPRBLK);
469         CHECK_FLAG_VALUE(NOCOMPR);
470         CHECK_FLAG_VALUE(ECOMPR);
471         CHECK_FLAG_VALUE(INDEX);
472         CHECK_FLAG_VALUE(IMAGIC);
473         CHECK_FLAG_VALUE(JOURNAL_DATA);
474         CHECK_FLAG_VALUE(NOTAIL);
475         CHECK_FLAG_VALUE(DIRSYNC);
476         CHECK_FLAG_VALUE(TOPDIR);
477         CHECK_FLAG_VALUE(HUGE_FILE);
478         CHECK_FLAG_VALUE(EXTENTS);
479         CHECK_FLAG_VALUE(EA_INODE);
480         CHECK_FLAG_VALUE(EOFBLOCKS);
481         CHECK_FLAG_VALUE(INLINE_DATA);
482         CHECK_FLAG_VALUE(RESERVED);
483 }
484
485 /* Used to pass group descriptor data when online resize is done */
486 struct ext4_new_group_input {
487         __u32 group;            /* Group number for this data */
488         __u64 block_bitmap;     /* Absolute block number of block bitmap */
489         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
490         __u64 inode_table;      /* Absolute block number of inode table start */
491         __u32 blocks_count;     /* Total number of blocks in this group */
492         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
493         __u16 unused;
494 };
495
496 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
497 struct compat_ext4_new_group_input {
498         u32 group;
499         compat_u64 block_bitmap;
500         compat_u64 inode_bitmap;
501         compat_u64 inode_table;
502         u32 blocks_count;
503         u16 reserved_blocks;
504         u16 unused;
505 };
506 #endif
507
508 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
509 struct ext4_new_group_data {
510         __u32 group;
511         __u64 block_bitmap;
512         __u64 inode_bitmap;
513         __u64 inode_table;
514         __u32 blocks_count;
515         __u16 reserved_blocks;
516         __u16 unused;
517         __u32 free_blocks_count;
518 };
519
520 /* Indexes used to index group tables in ext4_new_group_data */
521 enum {
522         BLOCK_BITMAP = 0,       /* block bitmap */
523         INODE_BITMAP,           /* inode bitmap */
524         INODE_TABLE,            /* inode tables */
525         GROUP_TABLE_COUNT,
526 };
527
528 /*
529  * Flags used by ext4_map_blocks()
530  */
531         /* Allocate any needed blocks and/or convert an unitialized
532            extent to be an initialized ext4 */
533 #define EXT4_GET_BLOCKS_CREATE                  0x0001
534         /* Request the creation of an unitialized extent */
535 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
536 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
537                                                  EXT4_GET_BLOCKS_CREATE)
538         /* Caller is from the delayed allocation writeout path
539          * finally doing the actual allocation of delayed blocks */
540 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
541         /* caller is from the direct IO path, request to creation of an
542         unitialized extents if not allocated, split the uninitialized
543         extent if blocks has been preallocated already*/
544 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
545 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
546 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
547                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
548         /* Convert extent to initialized after IO complete */
549 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
550                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
551         /* Eventual metadata allocation (due to growing extent tree)
552          * should not fail, so try to use reserved blocks for that.*/
553 #define EXT4_GET_BLOCKS_METADATA_NOFAIL         0x0020
554         /* Don't normalize allocation size (used for fallocate) */
555 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
556         /* Request will not result in inode size update (user for fallocate) */
557 #define EXT4_GET_BLOCKS_KEEP_SIZE               0x0080
558         /* Do not take i_data_sem locking in ext4_map_blocks */
559 #define EXT4_GET_BLOCKS_NO_LOCK                 0x0100
560         /* Do not put hole in extent cache */
561 #define EXT4_GET_BLOCKS_NO_PUT_HOLE             0x0200
562
563 /*
564  * The bit position of these flags must not overlap with any of the
565  * EXT4_GET_BLOCKS_*.  They are used by ext4_ext_find_extent(),
566  * read_extent_tree_block(), ext4_split_extent_at(),
567  * ext4_ext_insert_extent(), and ext4_ext_create_new_leaf().
568  * EXT4_EX_NOCACHE is used to indicate that the we shouldn't be
569  * caching the extents when reading from the extent tree while a
570  * truncate or punch hole operation is in progress.
571  */
572 #define EXT4_EX_NOCACHE                         0x0400
573 #define EXT4_EX_FORCE_CACHE                     0x0800
574
575 /*
576  * Flags used by ext4_free_blocks
577  */
578 #define EXT4_FREE_BLOCKS_METADATA       0x0001
579 #define EXT4_FREE_BLOCKS_FORGET         0x0002
580 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
581 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
582 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
583 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
584 #define EXT4_FREE_BLOCKS_RESERVE                0x0040
585
586 /*
587  * ioctl commands
588  */
589 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
590 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
591 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
592 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
593 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
594 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
595 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
596 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
597 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
598 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
599 #define EXT4_IOC_MIGRATE                _IO('f', 9)
600  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
601  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
602 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
603 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
604 #define EXT4_IOC_RESIZE_FS              _IOW('f', 16, __u64)
605 #define EXT4_IOC_SWAP_BOOT              _IO('f', 17)
606 #define EXT4_IOC_PRECACHE_EXTENTS       _IO('f', 18)
607
608 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
609 /*
610  * ioctl commands in 32 bit emulation
611  */
612 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
613 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
614 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
615 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
616 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
617 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
618 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
619 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
620 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
621 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
622 #endif
623
624 /* Max physical block we can address w/o extents */
625 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
626
627 /*
628  * Structure of an inode on the disk
629  */
630 struct ext4_inode {
631         __le16  i_mode;         /* File mode */
632         __le16  i_uid;          /* Low 16 bits of Owner Uid */
633         __le32  i_size_lo;      /* Size in bytes */
634         __le32  i_atime;        /* Access time */
635         __le32  i_ctime;        /* Inode Change time */
636         __le32  i_mtime;        /* Modification time */
637         __le32  i_dtime;        /* Deletion Time */
638         __le16  i_gid;          /* Low 16 bits of Group Id */
639         __le16  i_links_count;  /* Links count */
640         __le32  i_blocks_lo;    /* Blocks count */
641         __le32  i_flags;        /* File flags */
642         union {
643                 struct {
644                         __le32  l_i_version;
645                 } linux1;
646                 struct {
647                         __u32  h_i_translator;
648                 } hurd1;
649                 struct {
650                         __u32  m_i_reserved1;
651                 } masix1;
652         } osd1;                         /* OS dependent 1 */
653         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
654         __le32  i_generation;   /* File version (for NFS) */
655         __le32  i_file_acl_lo;  /* File ACL */
656         __le32  i_size_high;
657         __le32  i_obso_faddr;   /* Obsoleted fragment address */
658         union {
659                 struct {
660                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
661                         __le16  l_i_file_acl_high;
662                         __le16  l_i_uid_high;   /* these 2 fields */
663                         __le16  l_i_gid_high;   /* were reserved2[0] */
664                         __le16  l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
665                         __le16  l_i_reserved;
666                 } linux2;
667                 struct {
668                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
669                         __u16   h_i_mode_high;
670                         __u16   h_i_uid_high;
671                         __u16   h_i_gid_high;
672                         __u32   h_i_author;
673                 } hurd2;
674                 struct {
675                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
676                         __le16  m_i_file_acl_high;
677                         __u32   m_i_reserved2[2];
678                 } masix2;
679         } osd2;                         /* OS dependent 2 */
680         __le16  i_extra_isize;
681         __le16  i_checksum_hi;  /* crc32c(uuid+inum+inode) BE */
682         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
683         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
684         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
685         __le32  i_crtime;       /* File Creation time */
686         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
687         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
688 };
689
690 struct move_extent {
691         __u32 reserved;         /* should be zero */
692         __u32 donor_fd;         /* donor file descriptor */
693         __u64 orig_start;       /* logical start offset in block for orig */
694         __u64 donor_start;      /* logical start offset in block for donor */
695         __u64 len;              /* block length to be moved */
696         __u64 moved_len;        /* moved block length */
697 };
698
699 #define EXT4_EPOCH_BITS 2
700 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
701 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
702
703 /*
704  * Extended fields will fit into an inode if the filesystem was formatted
705  * with large inodes (-I 256 or larger) and there are not currently any EAs
706  * consuming all of the available space. For new inodes we always reserve
707  * enough space for the kernel's known extended fields, but for inodes
708  * created with an old kernel this might not have been the case. None of
709  * the extended inode fields is critical for correct filesystem operation.
710  * This macro checks if a certain field fits in the inode. Note that
711  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
712  */
713 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
714         ((offsetof(typeof(*ext4_inode), field) +        \
715           sizeof((ext4_inode)->field))                  \
716         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
717             (einode)->i_extra_isize))                   \
718
719 static inline __le32 ext4_encode_extra_time(struct timespec *time)
720 {
721        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
722                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
723                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
724 }
725
726 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
727 {
728        if (sizeof(time->tv_sec) > 4)
729                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
730                                << 32;
731        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
732 }
733
734 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
735 do {                                                                           \
736         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
737         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
738                 (raw_inode)->xtime ## _extra =                                 \
739                                 ext4_encode_extra_time(&(inode)->xtime);       \
740 } while (0)
741
742 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
743 do {                                                                           \
744         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
745                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
746         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
747                 (raw_inode)->xtime ## _extra =                                 \
748                                 ext4_encode_extra_time(&(einode)->xtime);      \
749 } while (0)
750
751 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
752 do {                                                                           \
753         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
754         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
755                 ext4_decode_extra_time(&(inode)->xtime,                        \
756                                        raw_inode->xtime ## _extra);            \
757         else                                                                   \
758                 (inode)->xtime.tv_nsec = 0;                                    \
759 } while (0)
760
761 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
762 do {                                                                           \
763         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
764                 (einode)->xtime.tv_sec =                                       \
765                         (signed)le32_to_cpu((raw_inode)->xtime);               \
766         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
767                 ext4_decode_extra_time(&(einode)->xtime,                       \
768                                        raw_inode->xtime ## _extra);            \
769         else                                                                   \
770                 (einode)->xtime.tv_nsec = 0;                                   \
771 } while (0)
772
773 #define i_disk_version osd1.linux1.l_i_version
774
775 #if defined(__KERNEL__) || defined(__linux__)
776 #define i_reserved1     osd1.linux1.l_i_reserved1
777 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
778 #define i_blocks_high   osd2.linux2.l_i_blocks_high
779 #define i_uid_low       i_uid
780 #define i_gid_low       i_gid
781 #define i_uid_high      osd2.linux2.l_i_uid_high
782 #define i_gid_high      osd2.linux2.l_i_gid_high
783 #define i_checksum_lo   osd2.linux2.l_i_checksum_lo
784
785 #elif defined(__GNU__)
786
787 #define i_translator    osd1.hurd1.h_i_translator
788 #define i_uid_high      osd2.hurd2.h_i_uid_high
789 #define i_gid_high      osd2.hurd2.h_i_gid_high
790 #define i_author        osd2.hurd2.h_i_author
791
792 #elif defined(__masix__)
793
794 #define i_reserved1     osd1.masix1.m_i_reserved1
795 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
796 #define i_reserved2     osd2.masix2.m_i_reserved2
797
798 #endif /* defined(__KERNEL__) || defined(__linux__) */
799
800 #include "extents_status.h"
801
802 /*
803  * fourth extended file system inode data in memory
804  */
805 struct ext4_inode_info {
806         __le32  i_data[15];     /* unconverted */
807         __u32   i_dtime;
808         ext4_fsblk_t    i_file_acl;
809
810         /*
811          * i_block_group is the number of the block group which contains
812          * this file's inode.  Constant across the lifetime of the inode,
813          * it is ued for making block allocation decisions - we try to
814          * place a file's data blocks near its inode block, and new inodes
815          * near to their parent directory's inode.
816          */
817         ext4_group_t    i_block_group;
818         ext4_lblk_t     i_dir_start_lookup;
819 #if (BITS_PER_LONG < 64)
820         unsigned long   i_state_flags;          /* Dynamic state flags */
821 #endif
822         unsigned long   i_flags;
823
824         /*
825          * Extended attributes can be read independently of the main file
826          * data. Taking i_mutex even when reading would cause contention
827          * between readers of EAs and writers of regular file data, so
828          * instead we synchronize on xattr_sem when reading or changing
829          * EAs.
830          */
831         struct rw_semaphore xattr_sem;
832
833         struct list_head i_orphan;      /* unlinked but open inodes */
834
835         /*
836          * i_disksize keeps track of what the inode size is ON DISK, not
837          * in memory.  During truncate, i_size is set to the new size by
838          * the VFS prior to calling ext4_truncate(), but the filesystem won't
839          * set i_disksize to 0 until the truncate is actually under way.
840          *
841          * The intent is that i_disksize always represents the blocks which
842          * are used by this file.  This allows recovery to restart truncate
843          * on orphans if we crash during truncate.  We actually write i_disksize
844          * into the on-disk inode when writing inodes out, instead of i_size.
845          *
846          * The only time when i_disksize and i_size may be different is when
847          * a truncate is in progress.  The only things which change i_disksize
848          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
849          */
850         loff_t  i_disksize;
851
852         /*
853          * i_data_sem is for serialising ext4_truncate() against
854          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
855          * data tree are chopped off during truncate. We can't do that in
856          * ext4 because whenever we perform intermediate commits during
857          * truncate, the inode and all the metadata blocks *must* be in a
858          * consistent state which allows truncation of the orphans to restart
859          * during recovery.  Hence we must fix the get_block-vs-truncate race
860          * by other means, so we have i_data_sem.
861          */
862         struct rw_semaphore i_data_sem;
863         struct inode vfs_inode;
864         struct jbd2_inode *jinode;
865
866         /*
867          * File creation time. Its function is same as that of
868          * struct timespec i_{a,c,m}time in the generic inode.
869          */
870         struct timespec i_crtime;
871
872         /* mballoc */
873         struct list_head i_prealloc_list;
874         spinlock_t i_prealloc_lock;
875
876         /* extents status tree */
877         struct ext4_es_tree i_es_tree;
878         rwlock_t i_es_lock;
879         struct list_head i_es_lru;
880         unsigned int i_es_lru_nr;       /* protected by i_es_lock */
881         unsigned long i_touch_when;     /* jiffies of last accessing */
882
883         /* ialloc */
884         ext4_group_t    i_last_alloc_group;
885
886         /* allocation reservation info for delalloc */
887         /* In case of bigalloc, these refer to clusters rather than blocks */
888         unsigned int i_reserved_data_blocks;
889         unsigned int i_reserved_meta_blocks;
890         unsigned int i_allocated_meta_blocks;
891         ext4_lblk_t i_da_metadata_calc_last_lblock;
892         int i_da_metadata_calc_len;
893
894         /* on-disk additional length */
895         __u16 i_extra_isize;
896
897         /* Indicate the inline data space. */
898         u16 i_inline_off;
899         u16 i_inline_size;
900
901 #ifdef CONFIG_QUOTA
902         /* quota space reservation, managed internally by quota code */
903         qsize_t i_reserved_quota;
904 #endif
905
906         /* Lock protecting lists below */
907         spinlock_t i_completed_io_lock;
908         /*
909          * Completed IOs that need unwritten extents handling and have
910          * transaction reserved
911          */
912         struct list_head i_rsv_conversion_list;
913         /*
914          * Completed IOs that need unwritten extents handling and don't have
915          * transaction reserved
916          */
917         struct list_head i_unrsv_conversion_list;
918         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
919         atomic_t i_unwritten; /* Nr. of inflight conversions pending */
920         struct work_struct i_rsv_conversion_work;
921         struct work_struct i_unrsv_conversion_work;
922
923         spinlock_t i_block_reservation_lock;
924
925         /*
926          * Transactions that contain inode's metadata needed to complete
927          * fsync and fdatasync, respectively.
928          */
929         tid_t i_sync_tid;
930         tid_t i_datasync_tid;
931
932         /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
933         __u32 i_csum_seed;
934 };
935
936 /*
937  * File system states
938  */
939 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
940 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
941 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
942
943 /*
944  * Misc. filesystem flags
945  */
946 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
947 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
948 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
949
950 /*
951  * Mount flags set via mount options or defaults
952  */
953 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
954 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
955 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
956 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
957 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
958 #define EXT4_MOUNT_ERRORS_MASK          0x00070
959 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
960 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
961 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
962 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
963 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
964 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
965 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
966 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
967 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
968 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
969 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
970 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
971 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
972 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
973 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
974 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
975 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
976 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
977 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
978 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
979 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
980 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
981 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
982
983 /*
984  * Mount flags set either automatically (could not be set by mount option)
985  * based on per file system feature or property or in special cases such as
986  * distinguishing between explicit mount option definition and default.
987  */
988 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
989                                                       specified delalloc */
990 #define EXT4_MOUNT2_STD_GROUP_SIZE      0x00000002 /* We have standard group
991                                                       size of blocksize * 8
992                                                       blocks */
993
994 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
995                                                 ~EXT4_MOUNT_##opt
996 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
997                                                 EXT4_MOUNT_##opt
998 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
999                                          EXT4_MOUNT_##opt)
1000
1001 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
1002                                                 ~EXT4_MOUNT2_##opt
1003 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
1004                                                 EXT4_MOUNT2_##opt
1005 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
1006                                          EXT4_MOUNT2_##opt)
1007
1008 #define ext4_test_and_set_bit           __test_and_set_bit_le
1009 #define ext4_set_bit                    __set_bit_le
1010 #define ext4_set_bit_atomic             ext2_set_bit_atomic
1011 #define ext4_test_and_clear_bit         __test_and_clear_bit_le
1012 #define ext4_clear_bit                  __clear_bit_le
1013 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
1014 #define ext4_test_bit                   test_bit_le
1015 #define ext4_find_next_zero_bit         find_next_zero_bit_le
1016 #define ext4_find_next_bit              find_next_bit_le
1017
1018 extern void ext4_set_bits(void *bm, int cur, int len);
1019
1020 /*
1021  * Maximal mount counts between two filesystem checks
1022  */
1023 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
1024 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
1025
1026 /*
1027  * Behaviour when detecting errors
1028  */
1029 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
1030 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
1031 #define EXT4_ERRORS_PANIC               3       /* Panic */
1032 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
1033
1034 /* Metadata checksum algorithm codes */
1035 #define EXT4_CRC32C_CHKSUM              1
1036
1037 /*
1038  * Structure of the super block
1039  */
1040 struct ext4_super_block {
1041 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1042         __le32  s_blocks_count_lo;      /* Blocks count */
1043         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1044         __le32  s_free_blocks_count_lo; /* Free blocks count */
1045 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1046         __le32  s_first_data_block;     /* First Data Block */
1047         __le32  s_log_block_size;       /* Block size */
1048         __le32  s_log_cluster_size;     /* Allocation cluster size */
1049 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1050         __le32  s_clusters_per_group;   /* # Clusters per group */
1051         __le32  s_inodes_per_group;     /* # Inodes per group */
1052         __le32  s_mtime;                /* Mount time */
1053 /*30*/  __le32  s_wtime;                /* Write time */
1054         __le16  s_mnt_count;            /* Mount count */
1055         __le16  s_max_mnt_count;        /* Maximal mount count */
1056         __le16  s_magic;                /* Magic signature */
1057         __le16  s_state;                /* File system state */
1058         __le16  s_errors;               /* Behaviour when detecting errors */
1059         __le16  s_minor_rev_level;      /* minor revision level */
1060 /*40*/  __le32  s_lastcheck;            /* time of last check */
1061         __le32  s_checkinterval;        /* max. time between checks */
1062         __le32  s_creator_os;           /* OS */
1063         __le32  s_rev_level;            /* Revision level */
1064 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1065         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1066         /*
1067          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1068          *
1069          * Note: the difference between the compatible feature set and
1070          * the incompatible feature set is that if there is a bit set
1071          * in the incompatible feature set that the kernel doesn't
1072          * know about, it should refuse to mount the filesystem.
1073          *
1074          * e2fsck's requirements are more strict; if it doesn't know
1075          * about a feature in either the compatible or incompatible
1076          * feature set, it must abort and not try to meddle with
1077          * things it doesn't understand...
1078          */
1079         __le32  s_first_ino;            /* First non-reserved inode */
1080         __le16  s_inode_size;           /* size of inode structure */
1081         __le16  s_block_group_nr;       /* block group # of this superblock */
1082         __le32  s_feature_compat;       /* compatible feature set */
1083 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1084         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1085 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1086 /*78*/  char    s_volume_name[16];      /* volume name */
1087 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1088 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1089         /*
1090          * Performance hints.  Directory preallocation should only
1091          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1092          */
1093         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1094         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1095         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1096         /*
1097          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1098          */
1099 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1100 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1101         __le32  s_journal_dev;          /* device number of journal file */
1102         __le32  s_last_orphan;          /* start of list of inodes to delete */
1103         __le32  s_hash_seed[4];         /* HTREE hash seed */
1104         __u8    s_def_hash_version;     /* Default hash version to use */
1105         __u8    s_jnl_backup_type;
1106         __le16  s_desc_size;            /* size of group descriptor */
1107 /*100*/ __le32  s_default_mount_opts;
1108         __le32  s_first_meta_bg;        /* First metablock block group */
1109         __le32  s_mkfs_time;            /* When the filesystem was created */
1110         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1111         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1112 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1113         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1114         __le32  s_free_blocks_count_hi; /* Free blocks count */
1115         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1116         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1117         __le32  s_flags;                /* Miscellaneous flags */
1118         __le16  s_raid_stride;          /* RAID stride */
1119         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1120         __le64  s_mmp_block;            /* Block for multi-mount protection */
1121         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1122         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1123         __u8    s_checksum_type;        /* metadata checksum algorithm used */
1124         __le16  s_reserved_pad;
1125         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1126         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1127         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1128         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1129                                               snapshot's future use */
1130         __le32  s_snapshot_list;        /* inode number of the head of the
1131                                            on-disk snapshot list */
1132 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1133         __le32  s_error_count;          /* number of fs errors */
1134         __le32  s_first_error_time;     /* first time an error happened */
1135         __le32  s_first_error_ino;      /* inode involved in first error */
1136         __le64  s_first_error_block;    /* block involved of first error */
1137         __u8    s_first_error_func[32]; /* function where the error happened */
1138         __le32  s_first_error_line;     /* line number where error happened */
1139         __le32  s_last_error_time;      /* most recent time of an error */
1140         __le32  s_last_error_ino;       /* inode involved in last error */
1141         __le32  s_last_error_line;      /* line number where error happened */
1142         __le64  s_last_error_block;     /* block involved of last error */
1143         __u8    s_last_error_func[32];  /* function where the error happened */
1144 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1145         __u8    s_mount_opts[64];
1146         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1147         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1148         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1149         __le32  s_reserved[108];        /* Padding to the end of the block */
1150         __le32  s_checksum;             /* crc32c(superblock) */
1151 };
1152
1153 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1154
1155 #ifdef __KERNEL__
1156
1157 /*
1158  * run-time mount flags
1159  */
1160 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
1161 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
1162
1163 /*
1164  * fourth extended-fs super-block data in memory
1165  */
1166 struct ext4_sb_info {
1167         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1168         unsigned long s_inodes_per_block;/* Number of inodes per block */
1169         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1170         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1171         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1172         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1173         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1174         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1175         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1176         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1177         unsigned long s_overhead;  /* # of fs overhead clusters */
1178         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1179         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1180         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1181         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1182         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1183         struct buffer_head **s_group_desc;
1184         unsigned int s_mount_opt;
1185         unsigned int s_mount_opt2;
1186         unsigned int s_mount_flags;
1187         unsigned int s_def_mount_opt;
1188         ext4_fsblk_t s_sb_block;
1189         atomic64_t s_resv_clusters;
1190         kuid_t s_resuid;
1191         kgid_t s_resgid;
1192         unsigned short s_mount_state;
1193         unsigned short s_pad;
1194         int s_addr_per_block_bits;
1195         int s_desc_per_block_bits;
1196         int s_inode_size;
1197         int s_first_ino;
1198         unsigned int s_inode_readahead_blks;
1199         unsigned int s_inode_goal;
1200         spinlock_t s_next_gen_lock;
1201         u32 s_next_generation;
1202         u32 s_hash_seed[4];
1203         int s_def_hash_version;
1204         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1205         struct percpu_counter s_freeclusters_counter;
1206         struct percpu_counter s_freeinodes_counter;
1207         struct percpu_counter s_dirs_counter;
1208         struct percpu_counter s_dirtyclusters_counter;
1209         struct blockgroup_lock *s_blockgroup_lock;
1210         struct proc_dir_entry *s_proc;
1211         struct kobject s_kobj;
1212         struct completion s_kobj_unregister;
1213         struct super_block *s_sb;
1214
1215         /* Journaling */
1216         struct journal_s *s_journal;
1217         struct list_head s_orphan;
1218         struct mutex s_orphan_lock;
1219         unsigned long s_resize_flags;           /* Flags indicating if there
1220                                                    is a resizer */
1221         unsigned long s_commit_interval;
1222         u32 s_max_batch_time;
1223         u32 s_min_batch_time;
1224         struct block_device *journal_bdev;
1225 #ifdef CONFIG_QUOTA
1226         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
1227         int s_jquota_fmt;                       /* Format of quota to use */
1228 #endif
1229         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1230         struct rb_root system_blks;
1231
1232 #ifdef EXTENTS_STATS
1233         /* ext4 extents stats */
1234         unsigned long s_ext_min;
1235         unsigned long s_ext_max;
1236         unsigned long s_depth_max;
1237         spinlock_t s_ext_stats_lock;
1238         unsigned long s_ext_blocks;
1239         unsigned long s_ext_extents;
1240 #endif
1241
1242         /* for buddy allocator */
1243         struct ext4_group_info ***s_group_info;
1244         struct inode *s_buddy_cache;
1245         spinlock_t s_md_lock;
1246         unsigned short *s_mb_offsets;
1247         unsigned int *s_mb_maxs;
1248         unsigned int s_group_info_size;
1249
1250         /* tunables */
1251         unsigned long s_stripe;
1252         unsigned int s_mb_stream_request;
1253         unsigned int s_mb_max_to_scan;
1254         unsigned int s_mb_min_to_scan;
1255         unsigned int s_mb_stats;
1256         unsigned int s_mb_order2_reqs;
1257         unsigned int s_mb_group_prealloc;
1258         unsigned int s_max_dir_size_kb;
1259         /* where last allocation was done - for stream allocation */
1260         unsigned long s_mb_last_group;
1261         unsigned long s_mb_last_start;
1262
1263         /* stats for buddy allocator */
1264         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1265         atomic_t s_bal_success; /* we found long enough chunks */
1266         atomic_t s_bal_allocated;       /* in blocks */
1267         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1268         atomic_t s_bal_goals;   /* goal hits */
1269         atomic_t s_bal_breaks;  /* too long searches */
1270         atomic_t s_bal_2orders; /* 2^order hits */
1271         spinlock_t s_bal_lock;
1272         unsigned long s_mb_buddies_generated;
1273         unsigned long long s_mb_generation_time;
1274         atomic_t s_mb_lost_chunks;
1275         atomic_t s_mb_preallocated;
1276         atomic_t s_mb_discarded;
1277         atomic_t s_lock_busy;
1278
1279         /* locality groups */
1280         struct ext4_locality_group __percpu *s_locality_groups;
1281
1282         /* for write statistics */
1283         unsigned long s_sectors_written_start;
1284         u64 s_kbytes_written;
1285
1286         /* the size of zero-out chunk */
1287         unsigned int s_extent_max_zeroout_kb;
1288
1289         unsigned int s_log_groups_per_flex;
1290         struct flex_groups *s_flex_groups;
1291         ext4_group_t s_flex_groups_allocated;
1292
1293         /* workqueue for unreserved extent convertions (dio) */
1294         struct workqueue_struct *unrsv_conversion_wq;
1295         /* workqueue for reserved extent conversions (buffered io) */
1296         struct workqueue_struct *rsv_conversion_wq;
1297
1298         /* timer for periodic error stats printing */
1299         struct timer_list s_err_report;
1300
1301         /* Lazy inode table initialization info */
1302         struct ext4_li_request *s_li_request;
1303         /* Wait multiplier for lazy initialization thread */
1304         unsigned int s_li_wait_mult;
1305
1306         /* Kernel thread for multiple mount protection */
1307         struct task_struct *s_mmp_tsk;
1308
1309         /* record the last minlen when FITRIM is called. */
1310         atomic_t s_last_trim_minblks;
1311
1312         /* Reference to checksum algorithm driver via cryptoapi */
1313         struct crypto_shash *s_chksum_driver;
1314
1315         /* Precomputed FS UUID checksum for seeding other checksums */
1316         __u32 s_csum_seed;
1317
1318         /* Reclaim extents from extent status tree */
1319         struct shrinker s_es_shrinker;
1320         struct list_head s_es_lru;
1321         unsigned long s_es_last_sorted;
1322         struct percpu_counter s_extent_cache_cnt;
1323         spinlock_t s_es_lru_lock ____cacheline_aligned_in_smp;
1324 };
1325
1326 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1327 {
1328         return sb->s_fs_info;
1329 }
1330 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1331 {
1332         return container_of(inode, struct ext4_inode_info, vfs_inode);
1333 }
1334
1335 static inline struct timespec ext4_current_time(struct inode *inode)
1336 {
1337         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1338                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1339 }
1340
1341 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1342 {
1343         return ino == EXT4_ROOT_INO ||
1344                 ino == EXT4_USR_QUOTA_INO ||
1345                 ino == EXT4_GRP_QUOTA_INO ||
1346                 ino == EXT4_BOOT_LOADER_INO ||
1347                 ino == EXT4_JOURNAL_INO ||
1348                 ino == EXT4_RESIZE_INO ||
1349                 (ino >= EXT4_FIRST_INO(sb) &&
1350                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1351 }
1352
1353 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
1354                                               struct ext4_io_end *io_end)
1355 {
1356         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
1357                 /* Writeback has to have coversion transaction reserved */
1358                 WARN_ON(EXT4_SB(inode->i_sb)->s_journal && !io_end->handle &&
1359                         !(io_end->flag & EXT4_IO_END_DIRECT));
1360                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
1361                 atomic_inc(&EXT4_I(inode)->i_unwritten);
1362         }
1363 }
1364
1365 static inline ext4_io_end_t *ext4_inode_aio(struct inode *inode)
1366 {
1367         return inode->i_private;
1368 }
1369
1370 static inline void ext4_inode_aio_set(struct inode *inode, ext4_io_end_t *io)
1371 {
1372         inode->i_private = io;
1373 }
1374
1375 /*
1376  * Inode dynamic state flags
1377  */
1378 enum {
1379         EXT4_STATE_JDATA,               /* journaled data exists */
1380         EXT4_STATE_NEW,                 /* inode is newly created */
1381         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1382         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1383         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1384         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1385         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1386         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1387         EXT4_STATE_DELALLOC_RESERVED,   /* blks already reserved for delalloc */
1388         EXT4_STATE_DIOREAD_LOCK,        /* Disable support for dio read
1389                                            nolocking */
1390         EXT4_STATE_MAY_INLINE_DATA,     /* may have in-inode data */
1391         EXT4_STATE_ORDERED_MODE,        /* data=ordered mode */
1392         EXT4_STATE_EXT_PRECACHED,       /* extents have been precached */
1393 };
1394
1395 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1396 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1397 {                                                                       \
1398         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1399 }                                                                       \
1400 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1401 {                                                                       \
1402         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1403 }                                                                       \
1404 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1405 {                                                                       \
1406         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1407 }
1408
1409 EXT4_INODE_BIT_FNS(flag, flags, 0)
1410 #if (BITS_PER_LONG < 64)
1411 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1412
1413 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1414 {
1415         (ei)->i_state_flags = 0;
1416 }
1417 #else
1418 EXT4_INODE_BIT_FNS(state, flags, 32)
1419
1420 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1421 {
1422         /* We depend on the fact that callers will set i_flags */
1423 }
1424 #endif
1425 #else
1426 /* Assume that user mode programs are passing in an ext4fs superblock, not
1427  * a kernel struct super_block.  This will allow us to call the feature-test
1428  * macros from user land. */
1429 #define EXT4_SB(sb)     (sb)
1430 #endif
1431
1432 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1433
1434 /*
1435  * Codes for operating systems
1436  */
1437 #define EXT4_OS_LINUX           0
1438 #define EXT4_OS_HURD            1
1439 #define EXT4_OS_MASIX           2
1440 #define EXT4_OS_FREEBSD         3
1441 #define EXT4_OS_LITES           4
1442
1443 /*
1444  * Revision levels
1445  */
1446 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1447 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1448
1449 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1450 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1451
1452 #define EXT4_GOOD_OLD_INODE_SIZE 128
1453
1454 /*
1455  * Feature set definitions
1456  */
1457
1458 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1459         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1460 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1461         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1462 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1463         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1464 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1465         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1466 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1467         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1468 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1469         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1470 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1471         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1472 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1473         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1474 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1475         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1476
1477 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1478 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1479 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1480 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1481 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1482 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1483
1484 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1485 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1486 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1487 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1488 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1489 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1490 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1491 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1492 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1493 /*
1494  * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM).  When
1495  * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
1496  * all other data structures' checksums.  However, the METADATA_CSUM and
1497  * GDT_CSUM bits are mutually exclusive.
1498  */
1499 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM    0x0400
1500
1501 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1502 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1503 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1504 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1505 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1506 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1507 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1508 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1509 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1510 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1511 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1512 #define EXT4_FEATURE_INCOMPAT_BG_USE_META_CSUM  0x2000 /* use crc32c for bg */
1513 #define EXT4_FEATURE_INCOMPAT_LARGEDIR          0x4000 /* >2GB or 3-lvl htree */
1514 #define EXT4_FEATURE_INCOMPAT_INLINE_DATA       0x8000 /* data in inode */
1515
1516 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1517 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1518                                          EXT4_FEATURE_INCOMPAT_META_BG)
1519 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1520                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1521                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1522
1523 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1524 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1525                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1526                                          EXT4_FEATURE_INCOMPAT_META_BG)
1527 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1528                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1529                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1530
1531 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1532 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1533                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1534                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1535                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1536                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1537                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1538                                          EXT4_FEATURE_INCOMPAT_MMP |    \
1539                                          EXT4_FEATURE_INCOMPAT_INLINE_DATA)
1540 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1541                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1542                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1543                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1544                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1545                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1546                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1547                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
1548                                          EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
1549                                          EXT4_FEATURE_RO_COMPAT_QUOTA)
1550
1551 /*
1552  * Default values for user and/or group using reserved blocks
1553  */
1554 #define EXT4_DEF_RESUID         0
1555 #define EXT4_DEF_RESGID         0
1556
1557 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1558
1559 /*
1560  * Default mount options
1561  */
1562 #define EXT4_DEFM_DEBUG         0x0001
1563 #define EXT4_DEFM_BSDGROUPS     0x0002
1564 #define EXT4_DEFM_XATTR_USER    0x0004
1565 #define EXT4_DEFM_ACL           0x0008
1566 #define EXT4_DEFM_UID16         0x0010
1567 #define EXT4_DEFM_JMODE         0x0060
1568 #define EXT4_DEFM_JMODE_DATA    0x0020
1569 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1570 #define EXT4_DEFM_JMODE_WBACK   0x0060
1571 #define EXT4_DEFM_NOBARRIER     0x0100
1572 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1573 #define EXT4_DEFM_DISCARD       0x0400
1574 #define EXT4_DEFM_NODELALLOC    0x0800
1575
1576 /*
1577  * Default journal batch times
1578  */
1579 #define EXT4_DEF_MIN_BATCH_TIME 0
1580 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1581
1582 /*
1583  * Minimum number of groups in a flexgroup before we separate out
1584  * directories into the first block group of a flexgroup
1585  */
1586 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1587
1588 /*
1589  * Structure of a directory entry
1590  */
1591 #define EXT4_NAME_LEN 255
1592
1593 struct ext4_dir_entry {
1594         __le32  inode;                  /* Inode number */
1595         __le16  rec_len;                /* Directory entry length */
1596         __le16  name_len;               /* Name length */
1597         char    name[EXT4_NAME_LEN];    /* File name */
1598 };
1599
1600 /*
1601  * The new version of the directory entry.  Since EXT4 structures are
1602  * stored in intel byte order, and the name_len field could never be
1603  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1604  * file_type field.
1605  */
1606 struct ext4_dir_entry_2 {
1607         __le32  inode;                  /* Inode number */
1608         __le16  rec_len;                /* Directory entry length */
1609         __u8    name_len;               /* Name length */
1610         __u8    file_type;
1611         char    name[EXT4_NAME_LEN];    /* File name */
1612 };
1613
1614 /*
1615  * This is a bogus directory entry at the end of each leaf block that
1616  * records checksums.
1617  */
1618 struct ext4_dir_entry_tail {
1619         __le32  det_reserved_zero1;     /* Pretend to be unused */
1620         __le16  det_rec_len;            /* 12 */
1621         __u8    det_reserved_zero2;     /* Zero name length */
1622         __u8    det_reserved_ft;        /* 0xDE, fake file type */
1623         __le32  det_checksum;           /* crc32c(uuid+inum+dirblock) */
1624 };
1625
1626 #define EXT4_DIRENT_TAIL(block, blocksize) \
1627         ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
1628                                         ((blocksize) - \
1629                                          sizeof(struct ext4_dir_entry_tail))))
1630
1631 /*
1632  * Ext4 directory file types.  Only the low 3 bits are used.  The
1633  * other bits are reserved for now.
1634  */
1635 #define EXT4_FT_UNKNOWN         0
1636 #define EXT4_FT_REG_FILE        1
1637 #define EXT4_FT_DIR             2
1638 #define EXT4_FT_CHRDEV          3
1639 #define EXT4_FT_BLKDEV          4
1640 #define EXT4_FT_FIFO            5
1641 #define EXT4_FT_SOCK            6
1642 #define EXT4_FT_SYMLINK         7
1643
1644 #define EXT4_FT_MAX             8
1645
1646 #define EXT4_FT_DIR_CSUM        0xDE
1647
1648 /*
1649  * EXT4_DIR_PAD defines the directory entries boundaries
1650  *
1651  * NOTE: It must be a multiple of 4
1652  */
1653 #define EXT4_DIR_PAD                    4
1654 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1655 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1656                                          ~EXT4_DIR_ROUND)
1657 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1658
1659 /*
1660  * If we ever get support for fs block sizes > page_size, we'll need
1661  * to remove the #if statements in the next two functions...
1662  */
1663 static inline unsigned int
1664 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1665 {
1666         unsigned len = le16_to_cpu(dlen);
1667
1668 #if (PAGE_CACHE_SIZE >= 65536)
1669         if (len == EXT4_MAX_REC_LEN || len == 0)
1670                 return blocksize;
1671         return (len & 65532) | ((len & 3) << 16);
1672 #else
1673         return len;
1674 #endif
1675 }
1676
1677 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1678 {
1679         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1680                 BUG();
1681 #if (PAGE_CACHE_SIZE >= 65536)
1682         if (len < 65536)
1683                 return cpu_to_le16(len);
1684         if (len == blocksize) {
1685                 if (blocksize == 65536)
1686                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1687                 else
1688                         return cpu_to_le16(0);
1689         }
1690         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1691 #else
1692         return cpu_to_le16(len);
1693 #endif
1694 }
1695
1696 /*
1697  * Hash Tree Directory indexing
1698  * (c) Daniel Phillips, 2001
1699  */
1700
1701 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1702                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1703                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1704 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1705 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1706
1707 /* Legal values for the dx_root hash_version field: */
1708
1709 #define DX_HASH_LEGACY          0
1710 #define DX_HASH_HALF_MD4        1
1711 #define DX_HASH_TEA             2
1712 #define DX_HASH_LEGACY_UNSIGNED 3
1713 #define DX_HASH_HALF_MD4_UNSIGNED       4
1714 #define DX_HASH_TEA_UNSIGNED            5
1715
1716 static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
1717                               const void *address, unsigned int length)
1718 {
1719         struct {
1720                 struct shash_desc shash;
1721                 char ctx[4];
1722         } desc;
1723         int err;
1724
1725         BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
1726
1727         desc.shash.tfm = sbi->s_chksum_driver;
1728         desc.shash.flags = 0;
1729         *(u32 *)desc.ctx = crc;
1730
1731         err = crypto_shash_update(&desc.shash, address, length);
1732         BUG_ON(err);
1733
1734         return *(u32 *)desc.ctx;
1735 }
1736
1737 #ifdef __KERNEL__
1738
1739 /* hash info structure used by the directory hash */
1740 struct dx_hash_info
1741 {
1742         u32             hash;
1743         u32             minor_hash;
1744         int             hash_version;
1745         u32             *seed;
1746 };
1747
1748
1749 /* 32 and 64 bit signed EOF for dx directories */
1750 #define EXT4_HTREE_EOF_32BIT   ((1UL  << (32 - 1)) - 1)
1751 #define EXT4_HTREE_EOF_64BIT   ((1ULL << (64 - 1)) - 1)
1752
1753
1754 /*
1755  * Control parameters used by ext4_htree_next_block
1756  */
1757 #define HASH_NB_ALWAYS          1
1758
1759
1760 /*
1761  * Describe an inode's exact location on disk and in memory
1762  */
1763 struct ext4_iloc
1764 {
1765         struct buffer_head *bh;
1766         unsigned long offset;
1767         ext4_group_t block_group;
1768 };
1769
1770 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1771 {
1772         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1773 }
1774
1775 /*
1776  * This structure is stuffed into the struct file's private_data field
1777  * for directories.  It is where we put information so that we can do
1778  * readdir operations in hash tree order.
1779  */
1780 struct dir_private_info {
1781         struct rb_root  root;
1782         struct rb_node  *curr_node;
1783         struct fname    *extra_fname;
1784         loff_t          last_pos;
1785         __u32           curr_hash;
1786         __u32           curr_minor_hash;
1787         __u32           next_hash;
1788 };
1789
1790 /* calculate the first block number of the group */
1791 static inline ext4_fsblk_t
1792 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1793 {
1794         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1795                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1796 }
1797
1798 /*
1799  * Special error return code only used by dx_probe() and its callers.
1800  */
1801 #define ERR_BAD_DX_DIR  -75000
1802
1803 /*
1804  * Timeout and state flag for lazy initialization inode thread.
1805  */
1806 #define EXT4_DEF_LI_WAIT_MULT                   10
1807 #define EXT4_DEF_LI_MAX_START_DELAY             5
1808 #define EXT4_LAZYINIT_QUIT                      0x0001
1809 #define EXT4_LAZYINIT_RUNNING                   0x0002
1810
1811 /*
1812  * Lazy inode table initialization info
1813  */
1814 struct ext4_lazy_init {
1815         unsigned long           li_state;
1816         struct list_head        li_request_list;
1817         struct mutex            li_list_mtx;
1818 };
1819
1820 struct ext4_li_request {
1821         struct super_block      *lr_super;
1822         struct ext4_sb_info     *lr_sbi;
1823         ext4_group_t            lr_next_group;
1824         struct list_head        lr_request;
1825         unsigned long           lr_next_sched;
1826         unsigned long           lr_timeout;
1827 };
1828
1829 struct ext4_features {
1830         struct kobject f_kobj;
1831         struct completion f_kobj_unregister;
1832 };
1833
1834 /*
1835  * This structure will be used for multiple mount protection. It will be
1836  * written into the block number saved in the s_mmp_block field in the
1837  * superblock. Programs that check MMP should assume that if
1838  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1839  * to use the filesystem, regardless of how old the timestamp is.
1840  */
1841 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
1842 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1843 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
1844 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
1845
1846 struct mmp_struct {
1847         __le32  mmp_magic;              /* Magic number for MMP */
1848         __le32  mmp_seq;                /* Sequence no. updated periodically */
1849
1850         /*
1851          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1852          * purposes and do not affect the correctness of the algorithm
1853          */
1854         __le64  mmp_time;               /* Time last updated */
1855         char    mmp_nodename[64];       /* Node which last updated MMP block */
1856         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
1857
1858         /*
1859          * mmp_check_interval is used to verify if the MMP block has been
1860          * updated on the block device. The value is updated based on the
1861          * maximum time to write the MMP block during an update cycle.
1862          */
1863         __le16  mmp_check_interval;
1864
1865         __le16  mmp_pad1;
1866         __le32  mmp_pad2[226];
1867         __le32  mmp_checksum;           /* crc32c(uuid+mmp_block) */
1868 };
1869
1870 /* arguments passed to the mmp thread */
1871 struct mmpd_data {
1872         struct buffer_head *bh; /* bh from initial read_mmp_block() */
1873         struct super_block *sb;  /* super block of the fs */
1874 };
1875
1876 /*
1877  * Check interval multiplier
1878  * The MMP block is written every update interval and initially checked every
1879  * update interval x the multiplier (the value is then adapted based on the
1880  * write latency). The reason is that writes can be delayed under load and we
1881  * don't want readers to incorrectly assume that the filesystem is no longer
1882  * in use.
1883  */
1884 #define EXT4_MMP_CHECK_MULT             2UL
1885
1886 /*
1887  * Minimum interval for MMP checking in seconds.
1888  */
1889 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
1890
1891 /*
1892  * Maximum interval for MMP checking in seconds.
1893  */
1894 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
1895
1896 /*
1897  * Function prototypes
1898  */
1899
1900 /*
1901  * Ok, these declarations are also in <linux/kernel.h> but none of the
1902  * ext4 source programs needs to include it so they are duplicated here.
1903  */
1904 # define NORET_TYPE     /**/
1905 # define ATTRIB_NORET   __attribute__((noreturn))
1906 # define NORET_AND      noreturn,
1907
1908 /* bitmap.c */
1909 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
1910 void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1911                                 struct ext4_group_desc *gdp,
1912                                 struct buffer_head *bh, int sz);
1913 int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1914                                   struct ext4_group_desc *gdp,
1915                                   struct buffer_head *bh, int sz);
1916 void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1917                                 struct ext4_group_desc *gdp,
1918                                 struct buffer_head *bh);
1919 int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1920                                   struct ext4_group_desc *gdp,
1921                                   struct buffer_head *bh);
1922
1923 /* balloc.c */
1924 extern void ext4_get_group_no_and_offset(struct super_block *sb,
1925                                          ext4_fsblk_t blocknr,
1926                                          ext4_group_t *blockgrpp,
1927                                          ext4_grpblk_t *offsetp);
1928 extern ext4_group_t ext4_get_group_number(struct super_block *sb,
1929                                           ext4_fsblk_t block);
1930
1931 extern void ext4_validate_block_bitmap(struct super_block *sb,
1932                                        struct ext4_group_desc *desc,
1933                                        ext4_group_t block_group,
1934                                        struct buffer_head *bh);
1935 extern unsigned int ext4_block_group(struct super_block *sb,
1936                         ext4_fsblk_t blocknr);
1937 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1938                         ext4_fsblk_t blocknr);
1939 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1940 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1941                         ext4_group_t group);
1942 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1943                                          ext4_fsblk_t goal,
1944                                          unsigned int flags,
1945                                          unsigned long *count,
1946                                          int *errp);
1947 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
1948                                     s64 nclusters, unsigned int flags);
1949 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
1950 extern void ext4_check_blocks_bitmap(struct super_block *);
1951 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1952                                                     ext4_group_t block_group,
1953                                                     struct buffer_head ** bh);
1954 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1955
1956 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
1957                                                 ext4_group_t block_group);
1958 extern int ext4_wait_block_bitmap(struct super_block *sb,
1959                                   ext4_group_t block_group,
1960                                   struct buffer_head *bh);
1961 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1962                                                   ext4_group_t block_group);
1963 extern void ext4_init_block_bitmap(struct super_block *sb,
1964                                    struct buffer_head *bh,
1965                                    ext4_group_t group,
1966                                    struct ext4_group_desc *desc);
1967 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
1968                                               ext4_group_t block_group,
1969                                               struct ext4_group_desc *gdp);
1970 extern unsigned ext4_num_overhead_clusters(struct super_block *sb,
1971                                            ext4_group_t block_group,
1972                                            struct ext4_group_desc *gdp);
1973 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
1974
1975 /* dir.c */
1976 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1977                                   struct file *,
1978                                   struct ext4_dir_entry_2 *,
1979                                   struct buffer_head *, char *, int,
1980                                   unsigned int);
1981 #define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset)      \
1982         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
1983                                         (de), (bh), (buf), (size), (offset)))
1984 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1985                                     __u32 minor_hash,
1986                                     struct ext4_dir_entry_2 *dirent);
1987 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1988 extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
1989                              struct buffer_head *bh,
1990                              void *buf, int buf_size,
1991                              const char *name, int namelen,
1992                              struct ext4_dir_entry_2 **dest_de);
1993 void ext4_insert_dentry(struct inode *inode,
1994                         struct ext4_dir_entry_2 *de,
1995                         int buf_size,
1996                         const char *name, int namelen);
1997 static inline void ext4_update_dx_flag(struct inode *inode)
1998 {
1999         if (!EXT4_HAS_COMPAT_FEATURE(inode->i_sb,
2000                                      EXT4_FEATURE_COMPAT_DIR_INDEX))
2001                 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
2002 }
2003 static unsigned char ext4_filetype_table[] = {
2004         DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
2005 };
2006
2007 static inline  unsigned char get_dtype(struct super_block *sb, int filetype)
2008 {
2009         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE) ||
2010             (filetype >= EXT4_FT_MAX))
2011                 return DT_UNKNOWN;
2012
2013         return ext4_filetype_table[filetype];
2014 }
2015
2016 /* fsync.c */
2017 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2018
2019 /* hash.c */
2020 extern int ext4fs_dirhash(const char *name, int len, struct
2021                           dx_hash_info *hinfo);
2022
2023 /* ialloc.c */
2024 extern struct inode *__ext4_new_inode(handle_t *, struct inode *, umode_t,
2025                                       const struct qstr *qstr, __u32 goal,
2026                                       uid_t *owner, int handle_type,
2027                                       unsigned int line_no, int nblocks);
2028
2029 #define ext4_new_inode(handle, dir, mode, qstr, goal, owner) \
2030         __ext4_new_inode((handle), (dir), (mode), (qstr), (goal), (owner), \
2031                          0, 0, 0)
2032 #define ext4_new_inode_start_handle(dir, mode, qstr, goal, owner, \
2033                                     type, nblocks)                  \
2034         __ext4_new_inode(NULL, (dir), (mode), (qstr), (goal), (owner), \
2035                          (type), __LINE__, (nblocks))
2036
2037
2038 extern void ext4_free_inode(handle_t *, struct inode *);
2039 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2040 extern unsigned long ext4_count_free_inodes(struct super_block *);
2041 extern unsigned long ext4_count_dirs(struct super_block *);
2042 extern void ext4_check_inodes_bitmap(struct super_block *);
2043 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2044 extern int ext4_init_inode_table(struct super_block *sb,
2045                                  ext4_group_t group, int barrier);
2046 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2047
2048 /* mballoc.c */
2049 extern long ext4_mb_stats;
2050 extern long ext4_mb_max_to_scan;
2051 extern int ext4_mb_init(struct super_block *);
2052 extern int ext4_mb_release(struct super_block *);
2053 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2054                                 struct ext4_allocation_request *, int *);
2055 extern int ext4_mb_reserve_blocks(struct super_block *, int);
2056 extern void ext4_discard_preallocations(struct inode *);
2057 extern int __init ext4_init_mballoc(void);
2058 extern void ext4_exit_mballoc(void);
2059 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2060                              struct buffer_head *bh, ext4_fsblk_t block,
2061                              unsigned long count, int flags);
2062 extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2063                                    ext4_group_t ngroups);
2064 extern int ext4_mb_add_groupinfo(struct super_block *sb,
2065                 ext4_group_t i, struct ext4_group_desc *desc);
2066 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2067                                 ext4_fsblk_t block, unsigned long count);
2068 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2069
2070 /* inode.c */
2071 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
2072                                                 ext4_lblk_t, int, int *);
2073 struct buffer_head *ext4_bread(handle_t *, struct inode *,
2074                                                 ext4_lblk_t, int, int *);
2075 int ext4_get_block_write(struct inode *inode, sector_t iblock,
2076                          struct buffer_head *bh_result, int create);
2077 int ext4_get_block(struct inode *inode, sector_t iblock,
2078                                 struct buffer_head *bh_result, int create);
2079 int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2080                            struct buffer_head *bh, int create);
2081 int ext4_walk_page_buffers(handle_t *handle,
2082                            struct buffer_head *head,
2083                            unsigned from,
2084                            unsigned to,
2085                            int *partial,
2086                            int (*fn)(handle_t *handle,
2087                                      struct buffer_head *bh));
2088 int do_journal_get_write_access(handle_t *handle,
2089                                 struct buffer_head *bh);
2090 #define FALL_BACK_TO_NONDELALLOC 1
2091 #define CONVERT_INLINE_DATA      2
2092
2093 extern struct inode *ext4_iget(struct super_block *, unsigned long);
2094 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
2095 extern int  ext4_setattr(struct dentry *, struct iattr *);
2096 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
2097                                 struct kstat *stat);
2098 extern void ext4_evict_inode(struct inode *);
2099 extern void ext4_clear_inode(struct inode *);
2100 extern int  ext4_sync_inode(handle_t *, struct inode *);
2101 extern void ext4_dirty_inode(struct inode *, int);
2102 extern int ext4_change_inode_journal_flag(struct inode *, int);
2103 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
2104 extern int ext4_inode_attach_jinode(struct inode *inode);
2105 extern int ext4_can_truncate(struct inode *inode);
2106 extern void ext4_truncate(struct inode *);
2107 extern int ext4_punch_hole(struct inode *inode, loff_t offset, loff_t length);
2108 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
2109 extern void ext4_set_inode_flags(struct inode *);
2110 extern void ext4_get_inode_flags(struct ext4_inode_info *);
2111 extern int ext4_alloc_da_blocks(struct inode *inode);
2112 extern void ext4_set_aops(struct inode *inode);
2113 extern int ext4_writepage_trans_blocks(struct inode *);
2114 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
2115 extern int ext4_block_truncate_page(handle_t *handle,
2116                 struct address_space *mapping, loff_t from);
2117 extern int ext4_block_zero_page_range(handle_t *handle,
2118                 struct address_space *mapping, loff_t from, loff_t length);
2119 extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
2120                              loff_t lstart, loff_t lend);
2121 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2122 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
2123 extern void ext4_da_update_reserve_space(struct inode *inode,
2124                                         int used, int quota_claim);
2125
2126 /* indirect.c */
2127 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
2128                                 struct ext4_map_blocks *map, int flags);
2129 extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
2130                                 const struct iovec *iov, loff_t offset,
2131                                 unsigned long nr_segs);
2132 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
2133 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
2134 extern void ext4_ind_truncate(handle_t *, struct inode *inode);
2135 extern int ext4_free_hole_blocks(handle_t *handle, struct inode *inode,
2136                                  ext4_lblk_t first, ext4_lblk_t stop);
2137
2138 /* ioctl.c */
2139 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
2140 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
2141
2142 /* migrate.c */
2143 extern int ext4_ext_migrate(struct inode *);
2144 extern int ext4_ind_migrate(struct inode *inode);
2145
2146 /* namei.c */
2147 extern int ext4_dirent_csum_verify(struct inode *inode,
2148                                    struct ext4_dir_entry *dirent);
2149 extern int ext4_orphan_add(handle_t *, struct inode *);
2150 extern int ext4_orphan_del(handle_t *, struct inode *);
2151 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
2152                                 __u32 start_minor_hash, __u32 *next_hash);
2153 extern int search_dir(struct buffer_head *bh,
2154                       char *search_buf,
2155                       int buf_size,
2156                       struct inode *dir,
2157                       const struct qstr *d_name,
2158                       unsigned int offset,
2159                       struct ext4_dir_entry_2 **res_dir);
2160 extern int ext4_generic_delete_entry(handle_t *handle,
2161                                      struct inode *dir,
2162                                      struct ext4_dir_entry_2 *de_del,
2163                                      struct buffer_head *bh,
2164                                      void *entry_buf,
2165                                      int buf_size,
2166                                      int csum_size);
2167
2168 /* resize.c */
2169 extern int ext4_group_add(struct super_block *sb,
2170                                 struct ext4_new_group_data *input);
2171 extern int ext4_group_extend(struct super_block *sb,
2172                                 struct ext4_super_block *es,
2173                                 ext4_fsblk_t n_blocks_count);
2174 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
2175
2176 /* super.c */
2177 extern int ext4_calculate_overhead(struct super_block *sb);
2178 extern int ext4_superblock_csum_verify(struct super_block *sb,
2179                                        struct ext4_super_block *es);
2180 extern void ext4_superblock_csum_set(struct super_block *sb);
2181 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
2182 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
2183 extern void ext4_kvfree(void *ptr);
2184 extern int ext4_alloc_flex_bg_array(struct super_block *sb,
2185                                     ext4_group_t ngroup);
2186 extern const char *ext4_decode_error(struct super_block *sb, int errno,
2187                                      char nbuf[16]);
2188
2189 extern __printf(4, 5)
2190 void __ext4_error(struct super_block *, const char *, unsigned int,
2191                   const char *, ...);
2192 extern __printf(5, 6)
2193 void __ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
2194                       const char *, ...);
2195 extern __printf(5, 6)
2196 void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
2197                      const char *, ...);
2198 extern void __ext4_std_error(struct super_block *, const char *,
2199                              unsigned int, int);
2200 extern __printf(4, 5)
2201 void __ext4_abort(struct super_block *, const char *, unsigned int,
2202                   const char *, ...);
2203 extern __printf(4, 5)
2204 void __ext4_warning(struct super_block *, const char *, unsigned int,
2205                     const char *, ...);
2206 extern __printf(3, 4)
2207 void __ext4_msg(struct super_block *, const char *, const char *, ...);
2208 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
2209                            const char *, unsigned int, const char *);
2210 extern __printf(7, 8)
2211 void __ext4_grp_locked_error(const char *, unsigned int,
2212                              struct super_block *, ext4_group_t,
2213                              unsigned long, ext4_fsblk_t,
2214                              const char *, ...);
2215
2216 #ifdef CONFIG_PRINTK
2217
2218 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2219         __ext4_error_inode(inode, func, line, block, fmt, ##__VA_ARGS__)
2220 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2221         __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__)
2222 #define ext4_error(sb, fmt, ...)                                        \
2223         __ext4_error(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2224 #define ext4_abort(sb, fmt, ...)                                        \
2225         __ext4_abort(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2226 #define ext4_warning(sb, fmt, ...)                                      \
2227         __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2228 #define ext4_msg(sb, level, fmt, ...)                           \
2229         __ext4_msg(sb, level, fmt, ##__VA_ARGS__)
2230 #define dump_mmp_msg(sb, mmp, msg)                                      \
2231         __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg)
2232 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2233         __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \
2234                                 fmt, ##__VA_ARGS__)
2235
2236 #else
2237
2238 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2239 do {                                                                    \
2240         no_printk(fmt, ##__VA_ARGS__);                                  \
2241         __ext4_error_inode(inode, "", 0, block, " ");                   \
2242 } while (0)
2243 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2244 do {                                                                    \
2245         no_printk(fmt, ##__VA_ARGS__);                                  \
2246         __ext4_error_file(file, "", 0, block, " ");                     \
2247 } while (0)
2248 #define ext4_error(sb, fmt, ...)                                        \
2249 do {                                                                    \
2250         no_printk(fmt, ##__VA_ARGS__);                                  \
2251         __ext4_error(sb, "", 0, " ");                                   \
2252 } while (0)
2253 #define ext4_abort(sb, fmt, ...)                                        \
2254 do {                                                                    \
2255         no_printk(fmt, ##__VA_ARGS__);                                  \
2256         __ext4_abort(sb, "", 0, " ");                                   \
2257 } while (0)
2258 #define ext4_warning(sb, fmt, ...)                                      \
2259 do {                                                                    \
2260         no_printk(fmt, ##__VA_ARGS__);                                  \
2261         __ext4_warning(sb, "", 0, " ");                                 \
2262 } while (0)
2263 #define ext4_msg(sb, level, fmt, ...)                                   \
2264 do {                                                                    \
2265         no_printk(fmt, ##__VA_ARGS__);                                  \
2266         __ext4_msg(sb, "", " ");                                        \
2267 } while (0)
2268 #define dump_mmp_msg(sb, mmp, msg)                                      \
2269         __dump_mmp_msg(sb, mmp, "", 0, "")
2270 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2271 do {                                                                    \
2272         no_printk(fmt, ##__VA_ARGS__);                          \
2273         __ext4_grp_locked_error("", 0, sb, grp, ino, block, " ");       \
2274 } while (0)
2275
2276 #endif
2277
2278 extern void ext4_update_dynamic_rev(struct super_block *sb);
2279 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
2280                                         __u32 compat);
2281 extern int ext4_update_rocompat_feature(handle_t *handle,
2282                                         struct super_block *sb, __u32 rocompat);
2283 extern int ext4_update_incompat_feature(handle_t *handle,
2284                                         struct super_block *sb, __u32 incompat);
2285 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
2286                                       struct ext4_group_desc *bg);
2287 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
2288                                       struct ext4_group_desc *bg);
2289 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
2290                                      struct ext4_group_desc *bg);
2291 extern __u32 ext4_free_group_clusters(struct super_block *sb,
2292                                       struct ext4_group_desc *bg);
2293 extern __u32 ext4_free_inodes_count(struct super_block *sb,
2294                                  struct ext4_group_desc *bg);
2295 extern __u32 ext4_used_dirs_count(struct super_block *sb,
2296                                 struct ext4_group_desc *bg);
2297 extern __u32 ext4_itable_unused_count(struct super_block *sb,
2298                                    struct ext4_group_desc *bg);
2299 extern void ext4_block_bitmap_set(struct super_block *sb,
2300                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2301 extern void ext4_inode_bitmap_set(struct super_block *sb,
2302                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2303 extern void ext4_inode_table_set(struct super_block *sb,
2304                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
2305 extern void ext4_free_group_clusters_set(struct super_block *sb,
2306                                          struct ext4_group_desc *bg,
2307                                          __u32 count);
2308 extern void ext4_free_inodes_set(struct super_block *sb,
2309                                 struct ext4_group_desc *bg, __u32 count);
2310 extern void ext4_used_dirs_set(struct super_block *sb,
2311                                 struct ext4_group_desc *bg, __u32 count);
2312 extern void ext4_itable_unused_set(struct super_block *sb,
2313                                    struct ext4_group_desc *bg, __u32 count);
2314 extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
2315                                        struct ext4_group_desc *gdp);
2316 extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
2317                                      struct ext4_group_desc *gdp);
2318 extern int ext4_register_li_request(struct super_block *sb,
2319                                     ext4_group_t first_not_zeroed);
2320
2321 static inline int ext4_has_group_desc_csum(struct super_block *sb)
2322 {
2323         return EXT4_HAS_RO_COMPAT_FEATURE(sb,
2324                                           EXT4_FEATURE_RO_COMPAT_GDT_CSUM |
2325                                           EXT4_FEATURE_RO_COMPAT_METADATA_CSUM);
2326 }
2327
2328 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2329 {
2330         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2331                 le32_to_cpu(es->s_blocks_count_lo);
2332 }
2333
2334 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2335 {
2336         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2337                 le32_to_cpu(es->s_r_blocks_count_lo);
2338 }
2339
2340 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2341 {
2342         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2343                 le32_to_cpu(es->s_free_blocks_count_lo);
2344 }
2345
2346 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2347                                          ext4_fsblk_t blk)
2348 {
2349         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2350         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2351 }
2352
2353 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2354                                               ext4_fsblk_t blk)
2355 {
2356         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2357         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2358 }
2359
2360 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2361                                            ext4_fsblk_t blk)
2362 {
2363         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2364         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2365 }
2366
2367 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2368 {
2369         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2370                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2371                         le32_to_cpu(raw_inode->i_size_lo);
2372         else
2373                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2374 }
2375
2376 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2377 {
2378         raw_inode->i_size_lo = cpu_to_le32(i_size);
2379         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2380 }
2381
2382 static inline
2383 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2384                                             ext4_group_t group)
2385 {
2386          struct ext4_group_info ***grp_info;
2387          long indexv, indexh;
2388          BUG_ON(group >= EXT4_SB(sb)->s_groups_count);
2389          grp_info = EXT4_SB(sb)->s_group_info;
2390          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2391          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2392          return grp_info[indexv][indexh];
2393 }
2394
2395 /*
2396  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2397  * the locking protocol documented in the comments of ext4_group_add()
2398  * in resize.c
2399  */
2400 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2401 {
2402         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2403
2404         smp_rmb();
2405         return ngroups;
2406 }
2407
2408 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2409                                              ext4_group_t block_group)
2410 {
2411         return block_group >> sbi->s_log_groups_per_flex;
2412 }
2413
2414 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2415 {
2416         return 1 << sbi->s_log_groups_per_flex;
2417 }
2418
2419 #define ext4_std_error(sb, errno)                               \
2420 do {                                                            \
2421         if ((errno))                                            \
2422                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2423 } while (0)
2424
2425 #ifdef CONFIG_SMP
2426 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2427  * counters. So we need to make sure we have free clusters more
2428  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2429  */
2430 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2431 #else
2432 #define EXT4_FREECLUSTERS_WATERMARK 0
2433 #endif
2434
2435 /* Update i_disksize. Requires i_mutex to avoid races with truncate */
2436 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2437 {
2438         WARN_ON_ONCE(S_ISREG(inode->i_mode) &&
2439                      !mutex_is_locked(&inode->i_mutex));
2440         down_write(&EXT4_I(inode)->i_data_sem);
2441         if (newsize > EXT4_I(inode)->i_disksize)
2442                 EXT4_I(inode)->i_disksize = newsize;
2443         up_write(&EXT4_I(inode)->i_data_sem);
2444 }
2445
2446 /*
2447  * Update i_disksize after writeback has been started. Races with truncate
2448  * are avoided by checking i_size under i_data_sem.
2449  */
2450 static inline void ext4_wb_update_i_disksize(struct inode *inode, loff_t newsize)
2451 {
2452         loff_t i_size;
2453
2454         down_write(&EXT4_I(inode)->i_data_sem);
2455         i_size = i_size_read(inode);
2456         if (newsize > i_size)
2457                 newsize = i_size;
2458         if (newsize > EXT4_I(inode)->i_disksize)
2459                 EXT4_I(inode)->i_disksize = newsize;
2460         up_write(&EXT4_I(inode)->i_data_sem);
2461 }
2462
2463 struct ext4_group_info {
2464         unsigned long   bb_state;
2465         struct rb_root  bb_free_root;
2466         ext4_grpblk_t   bb_first_free;  /* first free block */
2467         ext4_grpblk_t   bb_free;        /* total free blocks */
2468         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2469         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2470         struct          list_head bb_prealloc_list;
2471 #ifdef DOUBLE_CHECK
2472         void            *bb_bitmap;
2473 #endif
2474         struct rw_semaphore alloc_sem;
2475         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2476                                          * regions, index is order.
2477                                          * bb_counters[3] = 5 means
2478                                          * 5 free 8-block regions. */
2479 };
2480
2481 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2482 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2483 #define EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT     2
2484
2485 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2486         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2487 #define EXT4_MB_GRP_BBITMAP_CORRUPT(grp)        \
2488         (test_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
2489
2490 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2491         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2492 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2493         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2494 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2495         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2496
2497 #define EXT4_MAX_CONTENTION             8
2498 #define EXT4_CONTENTION_THRESHOLD       2
2499
2500 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2501                                               ext4_group_t group)
2502 {
2503         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2504 }
2505
2506 /*
2507  * Returns true if the filesystem is busy enough that attempts to
2508  * access the block group locks has run into contention.
2509  */
2510 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2511 {
2512         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2513 }
2514
2515 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2516 {
2517         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2518         if (spin_trylock(lock))
2519                 /*
2520                  * We're able to grab the lock right away, so drop the
2521                  * lock contention counter.
2522                  */
2523                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2524         else {
2525                 /*
2526                  * The lock is busy, so bump the contention counter,
2527                  * and then wait on the spin lock.
2528                  */
2529                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2530                                   EXT4_MAX_CONTENTION);
2531                 spin_lock(lock);
2532         }
2533 }
2534
2535 static inline void ext4_unlock_group(struct super_block *sb,
2536                                         ext4_group_t group)
2537 {
2538         spin_unlock(ext4_group_lock_ptr(sb, group));
2539 }
2540
2541 /*
2542  * Block validity checking
2543  */
2544 #define ext4_check_indirect_blockref(inode, bh)                         \
2545         ext4_check_blockref(__func__, __LINE__, inode,                  \
2546                             (__le32 *)(bh)->b_data,                     \
2547                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2548
2549 #define ext4_ind_check_inode(inode)                                     \
2550         ext4_check_blockref(__func__, __LINE__, inode,                  \
2551                             EXT4_I(inode)->i_data,                      \
2552                             EXT4_NDIR_BLOCKS)
2553
2554 /*
2555  * Inodes and files operations
2556  */
2557
2558 /* dir.c */
2559 extern const struct file_operations ext4_dir_operations;
2560
2561 /* file.c */
2562 extern const struct inode_operations ext4_file_inode_operations;
2563 extern const struct file_operations ext4_file_operations;
2564 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2565 extern void ext4_unwritten_wait(struct inode *inode);
2566
2567 /* inline.c */
2568 extern int ext4_has_inline_data(struct inode *inode);
2569 extern int ext4_get_inline_size(struct inode *inode);
2570 extern int ext4_get_max_inline_size(struct inode *inode);
2571 extern int ext4_find_inline_data_nolock(struct inode *inode);
2572 extern void ext4_write_inline_data(struct inode *inode,
2573                                    struct ext4_iloc *iloc,
2574                                    void *buffer, loff_t pos,
2575                                    unsigned int len);
2576 extern int ext4_prepare_inline_data(handle_t *handle, struct inode *inode,
2577                                     unsigned int len);
2578 extern int ext4_init_inline_data(handle_t *handle, struct inode *inode,
2579                                  unsigned int len);
2580 extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
2581
2582 extern int ext4_readpage_inline(struct inode *inode, struct page *page);
2583 extern int ext4_try_to_write_inline_data(struct address_space *mapping,
2584                                          struct inode *inode,
2585                                          loff_t pos, unsigned len,
2586                                          unsigned flags,
2587                                          struct page **pagep);
2588 extern int ext4_write_inline_data_end(struct inode *inode,
2589                                       loff_t pos, unsigned len,
2590                                       unsigned copied,
2591                                       struct page *page);
2592 extern struct buffer_head *
2593 ext4_journalled_write_inline_data(struct inode *inode,
2594                                   unsigned len,
2595                                   struct page *page);
2596 extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
2597                                            struct inode *inode,
2598                                            loff_t pos, unsigned len,
2599                                            unsigned flags,
2600                                            struct page **pagep,
2601                                            void **fsdata);
2602 extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
2603                                          unsigned len, unsigned copied,
2604                                          struct page *page);
2605 extern int ext4_try_add_inline_entry(handle_t *handle, struct dentry *dentry,
2606                                      struct inode *inode);
2607 extern int ext4_try_create_inline_dir(handle_t *handle,
2608                                       struct inode *parent,
2609                                       struct inode *inode);
2610 extern int ext4_read_inline_dir(struct file *filp,
2611                                 struct dir_context *ctx,
2612                                 int *has_inline_data);
2613 extern int htree_inlinedir_to_tree(struct file *dir_file,
2614                                    struct inode *dir, ext4_lblk_t block,
2615                                    struct dx_hash_info *hinfo,
2616                                    __u32 start_hash, __u32 start_minor_hash,
2617                                    int *has_inline_data);
2618 extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
2619                                         const struct qstr *d_name,
2620                                         struct ext4_dir_entry_2 **res_dir,
2621                                         int *has_inline_data);
2622 extern int ext4_delete_inline_entry(handle_t *handle,
2623                                     struct inode *dir,
2624                                     struct ext4_dir_entry_2 *de_del,
2625                                     struct buffer_head *bh,
2626                                     int *has_inline_data);
2627 extern int empty_inline_dir(struct inode *dir, int *has_inline_data);
2628 extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
2629                                         struct ext4_dir_entry_2 **parent_de,
2630                                         int *retval);
2631 extern int ext4_inline_data_fiemap(struct inode *inode,
2632                                    struct fiemap_extent_info *fieinfo,
2633                                    int *has_inline);
2634 extern int ext4_try_to_evict_inline_data(handle_t *handle,
2635                                          struct inode *inode,
2636                                          int needed);
2637 extern void ext4_inline_data_truncate(struct inode *inode, int *has_inline);
2638
2639 extern int ext4_convert_inline_data(struct inode *inode);
2640
2641 /* namei.c */
2642 extern const struct inode_operations ext4_dir_inode_operations;
2643 extern const struct inode_operations ext4_special_inode_operations;
2644 extern struct dentry *ext4_get_parent(struct dentry *child);
2645 extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
2646                                  struct ext4_dir_entry_2 *de,
2647                                  int blocksize, int csum_size,
2648                                  unsigned int parent_ino, int dotdot_real_len);
2649 extern void initialize_dirent_tail(struct ext4_dir_entry_tail *t,
2650                                    unsigned int blocksize);
2651 extern int ext4_handle_dirty_dirent_node(handle_t *handle,
2652                                          struct inode *inode,
2653                                          struct buffer_head *bh);
2654 #define S_SHIFT 12
2655 static unsigned char ext4_type_by_mode[S_IFMT >> S_SHIFT] = {
2656         [S_IFREG >> S_SHIFT]    = EXT4_FT_REG_FILE,
2657         [S_IFDIR >> S_SHIFT]    = EXT4_FT_DIR,
2658         [S_IFCHR >> S_SHIFT]    = EXT4_FT_CHRDEV,
2659         [S_IFBLK >> S_SHIFT]    = EXT4_FT_BLKDEV,
2660         [S_IFIFO >> S_SHIFT]    = EXT4_FT_FIFO,
2661         [S_IFSOCK >> S_SHIFT]   = EXT4_FT_SOCK,
2662         [S_IFLNK >> S_SHIFT]    = EXT4_FT_SYMLINK,
2663 };
2664
2665 static inline void ext4_set_de_type(struct super_block *sb,
2666                                 struct ext4_dir_entry_2 *de,
2667                                 umode_t mode) {
2668         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE))
2669                 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
2670 }
2671
2672
2673 /* symlink.c */
2674 extern const struct inode_operations ext4_symlink_inode_operations;
2675 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2676
2677 /* block_validity */
2678 extern void ext4_release_system_zone(struct super_block *sb);
2679 extern int ext4_setup_system_zone(struct super_block *sb);
2680 extern int __init ext4_init_system_zone(void);
2681 extern void ext4_exit_system_zone(void);
2682 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2683                                  ext4_fsblk_t start_blk,
2684                                  unsigned int count);
2685 extern int ext4_check_blockref(const char *, unsigned int,
2686                                struct inode *, __le32 *, unsigned int);
2687
2688 /* extents.c */
2689 struct ext4_ext_path;
2690 struct ext4_extent;
2691
2692 /*
2693  * Maximum number of logical blocks in a file; ext4_extent's ee_block is
2694  * __le32.
2695  */
2696 #define EXT_MAX_BLOCKS  0xffffffff
2697
2698 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2699 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2700 extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
2701 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2702                                struct ext4_map_blocks *map, int flags);
2703 extern void ext4_ext_truncate(handle_t *, struct inode *);
2704 extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
2705                                  ext4_lblk_t end);
2706 extern void ext4_ext_init(struct super_block *);
2707 extern void ext4_ext_release(struct super_block *);
2708 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2709                           loff_t len);
2710 extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
2711                                           loff_t offset, ssize_t len);
2712 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2713                            struct ext4_map_blocks *map, int flags);
2714 extern int ext4_ext_calc_metadata_amount(struct inode *inode,
2715                                          ext4_lblk_t lblocks);
2716 extern int ext4_extent_tree_init(handle_t *, struct inode *);
2717 extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
2718                                                    int num,
2719                                                    struct ext4_ext_path *path);
2720 extern int ext4_can_extents_be_merged(struct inode *inode,
2721                                       struct ext4_extent *ex1,
2722                                       struct ext4_extent *ex2);
2723 extern int ext4_ext_insert_extent(handle_t *, struct inode *,
2724                                   struct ext4_ext_path *,
2725                                   struct ext4_extent *, int);
2726 extern struct ext4_ext_path *ext4_ext_find_extent(struct inode *, ext4_lblk_t,
2727                                                   struct ext4_ext_path *,
2728                                                   int flags);
2729 extern void ext4_ext_drop_refs(struct ext4_ext_path *);
2730 extern int ext4_ext_check_inode(struct inode *inode);
2731 extern int ext4_find_delalloc_range(struct inode *inode,
2732                                     ext4_lblk_t lblk_start,
2733                                     ext4_lblk_t lblk_end);
2734 extern int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk);
2735 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2736                         __u64 start, __u64 len);
2737 extern int ext4_ext_precache(struct inode *inode);
2738
2739 /* move_extent.c */
2740 extern void ext4_double_down_write_data_sem(struct inode *first,
2741                                             struct inode *second);
2742 extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
2743                                           struct inode *donor_inode);
2744 void ext4_inode_double_lock(struct inode *inode1, struct inode *inode2);
2745 void ext4_inode_double_unlock(struct inode *inode1, struct inode *inode2);
2746 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2747                              __u64 start_orig, __u64 start_donor,
2748                              __u64 len, __u64 *moved_len);
2749
2750 /* page-io.c */
2751 extern int __init ext4_init_pageio(void);
2752 extern void ext4_exit_pageio(void);
2753 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2754 extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
2755 extern int ext4_put_io_end(ext4_io_end_t *io_end);
2756 extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
2757 extern void ext4_io_submit_init(struct ext4_io_submit *io,
2758                                 struct writeback_control *wbc);
2759 extern void ext4_end_io_rsv_work(struct work_struct *work);
2760 extern void ext4_end_io_unrsv_work(struct work_struct *work);
2761 extern void ext4_io_submit(struct ext4_io_submit *io);
2762 extern int ext4_bio_write_page(struct ext4_io_submit *io,
2763                                struct page *page,
2764                                int len,
2765                                struct writeback_control *wbc);
2766
2767 /* mmp.c */
2768 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2769 extern void ext4_mmp_csum_set(struct super_block *sb, struct mmp_struct *mmp);
2770 extern int ext4_mmp_csum_verify(struct super_block *sb,
2771                                 struct mmp_struct *mmp);
2772
2773 /*
2774  * Note that these flags will never ever appear in a buffer_head's state flag.
2775  * See EXT4_MAP_... to see where this is used.
2776  */
2777 enum ext4_state_bits {
2778         BH_Uninit       /* blocks are allocated but uninitialized on disk */
2779          = BH_JBDPrivateStart,
2780         BH_AllocFromCluster,    /* allocated blocks were part of already
2781                                  * allocated cluster. */
2782 };
2783
2784 /*
2785  * Add new method to test whether block and inode bitmaps are properly
2786  * initialized. With uninit_bg reading the block from disk is not enough
2787  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2788  */
2789 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2790
2791 static inline int bitmap_uptodate(struct buffer_head *bh)
2792 {
2793         return (buffer_uptodate(bh) &&
2794                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2795 }
2796 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2797 {
2798         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2799 }
2800
2801 /*
2802  * Disable DIO read nolock optimization, so new dioreaders will be forced
2803  * to grab i_mutex
2804  */
2805 static inline void ext4_inode_block_unlocked_dio(struct inode *inode)
2806 {
2807         ext4_set_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2808         smp_mb();
2809 }
2810 static inline void ext4_inode_resume_unlocked_dio(struct inode *inode)
2811 {
2812         smp_mb();
2813         ext4_clear_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2814 }
2815
2816 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2817
2818 /* For ioend & aio unwritten conversion wait queues */
2819 #define EXT4_WQ_HASH_SZ         37
2820 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
2821                                             EXT4_WQ_HASH_SZ])
2822 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
2823                                              EXT4_WQ_HASH_SZ])
2824 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2825 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2826
2827 #define EXT4_RESIZING   0
2828 extern int ext4_resize_begin(struct super_block *sb);
2829 extern void ext4_resize_end(struct super_block *sb);
2830
2831 #endif  /* __KERNEL__ */
2832
2833 #endif  /* _EXT4_H */