direct_IO: use iov_iter_rw() instead of rw everywhere
[linux-2.6-block.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 <linux/ratelimit.h>
33 #include <crypto/hash.h>
34 #include <linux/falloc.h>
35 #ifdef __KERNEL__
36 #include <linux/compat.h>
37 #endif
38
39 /*
40  * The fourth extended filesystem constants/structures
41  */
42
43 /*
44  * Define EXT4FS_DEBUG to produce debug messages
45  */
46 #undef EXT4FS_DEBUG
47
48 /*
49  * Debug code
50  */
51 #ifdef EXT4FS_DEBUG
52 #define ext4_debug(f, a...)                                             \
53         do {                                                            \
54                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
55                         __FILE__, __LINE__, __func__);                  \
56                 printk(KERN_DEBUG f, ## a);                             \
57         } while (0)
58 #else
59 #define ext4_debug(fmt, ...)    no_printk(fmt, ##__VA_ARGS__)
60 #endif
61
62 /*
63  * Turn on EXT_DEBUG to get lots of info about extents operations.
64  */
65 #define EXT_DEBUG__
66 #ifdef EXT_DEBUG
67 #define ext_debug(fmt, ...)     printk(fmt, ##__VA_ARGS__)
68 #else
69 #define ext_debug(fmt, ...)     no_printk(fmt, ##__VA_ARGS__)
70 #endif
71
72 #define EXT4_ERROR_INODE(inode, fmt, a...) \
73         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
74
75 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...)                 \
76         ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
77
78 #define EXT4_ERROR_FILE(file, block, fmt, a...)                         \
79         ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
80
81 /* data type for block offset of block group */
82 typedef int ext4_grpblk_t;
83
84 /* data type for filesystem-wide blocks number */
85 typedef unsigned long long ext4_fsblk_t;
86
87 /* data type for file logical block number */
88 typedef __u32 ext4_lblk_t;
89
90 /* data type for block group number */
91 typedef unsigned int ext4_group_t;
92
93 /*
94  * Flags used in mballoc's allocation_context flags field.
95  *
96  * Also used to show what's going on for debugging purposes when the
97  * flag field is exported via the traceport interface
98  */
99
100 /* prefer goal again. length */
101 #define EXT4_MB_HINT_MERGE              0x0001
102 /* blocks already reserved */
103 #define EXT4_MB_HINT_RESERVED           0x0002
104 /* metadata is being allocated */
105 #define EXT4_MB_HINT_METADATA           0x0004
106 /* first blocks in the file */
107 #define EXT4_MB_HINT_FIRST              0x0008
108 /* search for the best chunk */
109 #define EXT4_MB_HINT_BEST               0x0010
110 /* data is being allocated */
111 #define EXT4_MB_HINT_DATA               0x0020
112 /* don't preallocate (for tails) */
113 #define EXT4_MB_HINT_NOPREALLOC         0x0040
114 /* allocate for locality group */
115 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
116 /* allocate goal blocks or none */
117 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
118 /* goal is meaningful */
119 #define EXT4_MB_HINT_TRY_GOAL           0x0200
120 /* blocks already pre-reserved by delayed allocation */
121 #define EXT4_MB_DELALLOC_RESERVED       0x0400
122 /* We are doing stream allocation */
123 #define EXT4_MB_STREAM_ALLOC            0x0800
124 /* Use reserved root blocks if needed */
125 #define EXT4_MB_USE_ROOT_BLOCKS         0x1000
126 /* Use blocks from reserved pool */
127 #define EXT4_MB_USE_RESERVED            0x2000
128
129 struct ext4_allocation_request {
130         /* target inode for block we're allocating */
131         struct inode *inode;
132         /* how many blocks we want to allocate */
133         unsigned int len;
134         /* logical block in target inode */
135         ext4_lblk_t logical;
136         /* the closest logical allocated block to the left */
137         ext4_lblk_t lleft;
138         /* the closest logical allocated block to the right */
139         ext4_lblk_t lright;
140         /* phys. target (a hint) */
141         ext4_fsblk_t goal;
142         /* phys. block for the closest logical allocated block to the left */
143         ext4_fsblk_t pleft;
144         /* phys. block for the closest logical allocated block to the right */
145         ext4_fsblk_t pright;
146         /* flags. see above EXT4_MB_HINT_* */
147         unsigned int flags;
148 };
149
150 /*
151  * Logical to physical block mapping, used by ext4_map_blocks()
152  *
153  * This structure is used to pass requests into ext4_map_blocks() as
154  * well as to store the information returned by ext4_map_blocks().  It
155  * takes less room on the stack than a struct buffer_head.
156  */
157 #define EXT4_MAP_NEW            (1 << BH_New)
158 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
159 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
160 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
161 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
162                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY)
163
164 struct ext4_map_blocks {
165         ext4_fsblk_t m_pblk;
166         ext4_lblk_t m_lblk;
167         unsigned int m_len;
168         unsigned int m_flags;
169 };
170
171 /*
172  * Flags for ext4_io_end->flags
173  */
174 #define EXT4_IO_END_UNWRITTEN   0x0001
175
176 /*
177  * For converting unwritten extents on a work queue. 'handle' is used for
178  * buffered writeback.
179  */
180 typedef struct ext4_io_end {
181         struct list_head        list;           /* per-file finished IO list */
182         handle_t                *handle;        /* handle reserved for extent
183                                                  * conversion */
184         struct inode            *inode;         /* file being written to */
185         struct bio              *bio;           /* Linked list of completed
186                                                  * bios covering the extent */
187         unsigned int            flag;           /* unwritten or not */
188         loff_t                  offset;         /* offset in the file */
189         ssize_t                 size;           /* size of the extent */
190         atomic_t                count;          /* reference counter */
191 } ext4_io_end_t;
192
193 struct ext4_io_submit {
194         int                     io_op;
195         struct bio              *io_bio;
196         ext4_io_end_t           *io_end;
197         sector_t                io_next_block;
198 };
199
200 /*
201  * Special inodes numbers
202  */
203 #define EXT4_BAD_INO             1      /* Bad blocks inode */
204 #define EXT4_ROOT_INO            2      /* Root inode */
205 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
206 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
207 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
208 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
209 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
210 #define EXT4_JOURNAL_INO         8      /* Journal inode */
211
212 /* First non-reserved inode for old ext4 filesystems */
213 #define EXT4_GOOD_OLD_FIRST_INO 11
214
215 /*
216  * Maximal count of links to a file
217  */
218 #define EXT4_LINK_MAX           65000
219
220 /*
221  * Macro-instructions used to manage several block sizes
222  */
223 #define EXT4_MIN_BLOCK_SIZE             1024
224 #define EXT4_MAX_BLOCK_SIZE             65536
225 #define EXT4_MIN_BLOCK_LOG_SIZE         10
226 #define EXT4_MAX_BLOCK_LOG_SIZE         16
227 #ifdef __KERNEL__
228 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
229 #else
230 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
231 #endif
232 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
233 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
234                                          EXT4_SB(s)->s_cluster_bits)
235 #ifdef __KERNEL__
236 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
237 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
238 #else
239 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
240 #endif
241 #ifdef __KERNEL__
242 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
243 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
244 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
245 #else
246 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
247                                  EXT4_GOOD_OLD_INODE_SIZE : \
248                                  (s)->s_inode_size)
249 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
250                                  EXT4_GOOD_OLD_FIRST_INO : \
251                                  (s)->s_first_ino)
252 #endif
253 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
254
255 /* Translate a block number to a cluster number */
256 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
257 /* Translate a cluster number to a block number */
258 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
259 /* Translate # of blks to # of clusters */
260 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
261                                  (sbi)->s_cluster_bits)
262 /* Mask out the low bits to get the starting block of the cluster */
263 #define EXT4_PBLK_CMASK(s, pblk) ((pblk) &                              \
264                                   ~((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
265 #define EXT4_LBLK_CMASK(s, lblk) ((lblk) &                              \
266                                   ~((ext4_lblk_t) (s)->s_cluster_ratio - 1))
267 /* Get the cluster offset */
268 #define EXT4_PBLK_COFF(s, pblk) ((pblk) &                               \
269                                  ((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
270 #define EXT4_LBLK_COFF(s, lblk) ((lblk) &                               \
271                                  ((ext4_lblk_t) (s)->s_cluster_ratio - 1))
272
273 /*
274  * Structure of a blocks group descriptor
275  */
276 struct ext4_group_desc
277 {
278         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
279         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
280         __le32  bg_inode_table_lo;      /* Inodes table block */
281         __le16  bg_free_blocks_count_lo;/* Free blocks count */
282         __le16  bg_free_inodes_count_lo;/* Free inodes count */
283         __le16  bg_used_dirs_count_lo;  /* Directories count */
284         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
285         __le32  bg_exclude_bitmap_lo;   /* Exclude bitmap for snapshots */
286         __le16  bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
287         __le16  bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
288         __le16  bg_itable_unused_lo;    /* Unused inodes count */
289         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
290         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
291         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
292         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
293         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
294         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
295         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
296         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
297         __le32  bg_exclude_bitmap_hi;   /* Exclude bitmap block MSB */
298         __le16  bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
299         __le16  bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
300         __u32   bg_reserved;
301 };
302
303 #define EXT4_BG_INODE_BITMAP_CSUM_HI_END        \
304         (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
305          sizeof(__le16))
306 #define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END        \
307         (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
308          sizeof(__le16))
309
310 /*
311  * Structure of a flex block group info
312  */
313
314 struct flex_groups {
315         atomic64_t      free_clusters;
316         atomic_t        free_inodes;
317         atomic_t        used_dirs;
318 };
319
320 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
321 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
322 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
323
324 /*
325  * Macro-instructions used to manage group descriptors
326  */
327 #define EXT4_MIN_DESC_SIZE              32
328 #define EXT4_MIN_DESC_SIZE_64BIT        64
329 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
330 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
331 #ifdef __KERNEL__
332 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
333 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
334 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
335 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
336 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
337 #else
338 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
339 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
340 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
341 #endif
342
343 /*
344  * Constants relative to the data blocks
345  */
346 #define EXT4_NDIR_BLOCKS                12
347 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
348 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
349 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
350 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
351
352 /*
353  * Inode flags
354  */
355 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
356 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
357 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
358 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
359 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
360 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
361 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
362 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
363 /* Reserved for compression usage... */
364 #define EXT4_DIRTY_FL                   0x00000100
365 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
366 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
367         /* nb: was previously EXT2_ECOMPR_FL */
368 #define EXT4_ENCRYPT_FL                 0x00000800 /* encrypted file */
369 /* End compression flags --- maybe not all used */
370 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
371 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
372 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
373 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
374 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
375 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
376 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
377 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
378 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
379 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
380 #define EXT4_INLINE_DATA_FL             0x10000000 /* Inode has inline data. */
381 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
382
383 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
384 #define EXT4_FL_USER_MODIFIABLE         0x004380FF /* User modifiable flags */
385
386 /* Flags that should be inherited by new inodes from their parent. */
387 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
388                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
389                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
390                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
391
392 /* Flags that are appropriate for regular files (all but dir-specific ones). */
393 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
394
395 /* Flags that are appropriate for non-directories/regular files. */
396 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
397
398 /* Mask out flags that are inappropriate for the given type of inode. */
399 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
400 {
401         if (S_ISDIR(mode))
402                 return flags;
403         else if (S_ISREG(mode))
404                 return flags & EXT4_REG_FLMASK;
405         else
406                 return flags & EXT4_OTHER_FLMASK;
407 }
408
409 /*
410  * Inode flags used for atomic set/get
411  */
412 enum {
413         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
414         EXT4_INODE_UNRM         = 1,    /* Undelete */
415         EXT4_INODE_COMPR        = 2,    /* Compress file */
416         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
417         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
418         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
419         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
420         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
421 /* Reserved for compression usage... */
422         EXT4_INODE_DIRTY        = 8,
423         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
424         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
425         EXT4_INODE_ENCRYPT      = 11,   /* 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(ENCRYPT);
471         CHECK_FLAG_VALUE(INDEX);
472         CHECK_FLAG_VALUE(IMAGIC);
473         CHECK_FLAG_VALUE(JOURNAL_DATA);
474         CHECK_FLAG_VALUE(NOTAIL);
475         CHECK_FLAG_VALUE(DIRSYNC);
476         CHECK_FLAG_VALUE(TOPDIR);
477         CHECK_FLAG_VALUE(HUGE_FILE);
478         CHECK_FLAG_VALUE(EXTENTS);
479         CHECK_FLAG_VALUE(EA_INODE);
480         CHECK_FLAG_VALUE(EOFBLOCKS);
481         CHECK_FLAG_VALUE(INLINE_DATA);
482         CHECK_FLAG_VALUE(RESERVED);
483 }
484
485 /* Used to pass group descriptor data when online resize is done */
486 struct ext4_new_group_input {
487         __u32 group;            /* Group number for this data */
488         __u64 block_bitmap;     /* Absolute block number of block bitmap */
489         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
490         __u64 inode_table;      /* Absolute block number of inode table start */
491         __u32 blocks_count;     /* Total number of blocks in this group */
492         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
493         __u16 unused;
494 };
495
496 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
497 struct compat_ext4_new_group_input {
498         u32 group;
499         compat_u64 block_bitmap;
500         compat_u64 inode_bitmap;
501         compat_u64 inode_table;
502         u32 blocks_count;
503         u16 reserved_blocks;
504         u16 unused;
505 };
506 #endif
507
508 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
509 struct ext4_new_group_data {
510         __u32 group;
511         __u64 block_bitmap;
512         __u64 inode_bitmap;
513         __u64 inode_table;
514         __u32 blocks_count;
515         __u16 reserved_blocks;
516         __u16 unused;
517         __u32 free_blocks_count;
518 };
519
520 /* Indexes used to index group tables in ext4_new_group_data */
521 enum {
522         BLOCK_BITMAP = 0,       /* block bitmap */
523         INODE_BITMAP,           /* inode bitmap */
524         INODE_TABLE,            /* inode tables */
525         GROUP_TABLE_COUNT,
526 };
527
528 /*
529  * Flags used by ext4_map_blocks()
530  */
531         /* Allocate any needed blocks and/or convert an unwritten
532            extent to be an initialized ext4 */
533 #define EXT4_GET_BLOCKS_CREATE                  0x0001
534         /* Request the creation of an unwritten extent */
535 #define EXT4_GET_BLOCKS_UNWRIT_EXT              0x0002
536 #define EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT       (EXT4_GET_BLOCKS_UNWRIT_EXT|\
537                                                  EXT4_GET_BLOCKS_CREATE)
538         /* Caller is from the delayed allocation writeout path
539          * finally doing the actual allocation of delayed blocks */
540 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
541         /* caller is from the direct IO path, request to creation of an
542         unwritten extents if not allocated, split the unwritten
543         extent if blocks has been preallocated already*/
544 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
545 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
546 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
547                                          EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
548         /* Convert extent to initialized after IO complete */
549 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
550                                          EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
551         /* Eventual metadata allocation (due to growing extent tree)
552          * should not fail, so try to use reserved blocks for that.*/
553 #define EXT4_GET_BLOCKS_METADATA_NOFAIL         0x0020
554         /* Don't normalize allocation size (used for fallocate) */
555 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
556         /* Request will not result in inode size update (user for fallocate) */
557 #define EXT4_GET_BLOCKS_KEEP_SIZE               0x0080
558         /* Do not take i_data_sem locking in ext4_map_blocks */
559 #define EXT4_GET_BLOCKS_NO_LOCK                 0x0100
560         /* Convert written extents to unwritten */
561 #define EXT4_GET_BLOCKS_CONVERT_UNWRITTEN       0x0200
562
563 /*
564  * The bit position of these flags must not overlap with any of the
565  * EXT4_GET_BLOCKS_*.  They are used by ext4_find_extent(),
566  * read_extent_tree_block(), ext4_split_extent_at(),
567  * ext4_ext_insert_extent(), and ext4_ext_create_new_leaf().
568  * EXT4_EX_NOCACHE is used to indicate that the we shouldn't be
569  * caching the extents when reading from the extent tree while a
570  * truncate or punch hole operation is in progress.
571  */
572 #define EXT4_EX_NOCACHE                         0x40000000
573 #define EXT4_EX_FORCE_CACHE                     0x20000000
574
575 /*
576  * Flags used by ext4_free_blocks
577  */
578 #define EXT4_FREE_BLOCKS_METADATA       0x0001
579 #define EXT4_FREE_BLOCKS_FORGET         0x0002
580 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
581 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
582 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
583 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
584
585 /*
586  * ioctl commands
587  */
588 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
589 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
590 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
591 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
592 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
593 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
594 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
595 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
596 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
597 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
598 #define EXT4_IOC_MIGRATE                _IO('f', 9)
599  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
600  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
601 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
602 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
603 #define EXT4_IOC_RESIZE_FS              _IOW('f', 16, __u64)
604 #define EXT4_IOC_SWAP_BOOT              _IO('f', 17)
605 #define EXT4_IOC_PRECACHE_EXTENTS       _IO('f', 18)
606
607 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
608 /*
609  * ioctl commands in 32 bit emulation
610  */
611 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
612 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
613 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
614 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
615 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
616 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
617 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
618 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
619 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
620 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
621 #endif
622
623 /* Max physical block we can address w/o extents */
624 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
625
626 /*
627  * Structure of an inode on the disk
628  */
629 struct ext4_inode {
630         __le16  i_mode;         /* File mode */
631         __le16  i_uid;          /* Low 16 bits of Owner Uid */
632         __le32  i_size_lo;      /* Size in bytes */
633         __le32  i_atime;        /* Access time */
634         __le32  i_ctime;        /* Inode Change time */
635         __le32  i_mtime;        /* Modification time */
636         __le32  i_dtime;        /* Deletion Time */
637         __le16  i_gid;          /* Low 16 bits of Group Id */
638         __le16  i_links_count;  /* Links count */
639         __le32  i_blocks_lo;    /* Blocks count */
640         __le32  i_flags;        /* File flags */
641         union {
642                 struct {
643                         __le32  l_i_version;
644                 } linux1;
645                 struct {
646                         __u32  h_i_translator;
647                 } hurd1;
648                 struct {
649                         __u32  m_i_reserved1;
650                 } masix1;
651         } osd1;                         /* OS dependent 1 */
652         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
653         __le32  i_generation;   /* File version (for NFS) */
654         __le32  i_file_acl_lo;  /* File ACL */
655         __le32  i_size_high;
656         __le32  i_obso_faddr;   /* Obsoleted fragment address */
657         union {
658                 struct {
659                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
660                         __le16  l_i_file_acl_high;
661                         __le16  l_i_uid_high;   /* these 2 fields */
662                         __le16  l_i_gid_high;   /* were reserved2[0] */
663                         __le16  l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
664                         __le16  l_i_reserved;
665                 } linux2;
666                 struct {
667                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
668                         __u16   h_i_mode_high;
669                         __u16   h_i_uid_high;
670                         __u16   h_i_gid_high;
671                         __u32   h_i_author;
672                 } hurd2;
673                 struct {
674                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
675                         __le16  m_i_file_acl_high;
676                         __u32   m_i_reserved2[2];
677                 } masix2;
678         } osd2;                         /* OS dependent 2 */
679         __le16  i_extra_isize;
680         __le16  i_checksum_hi;  /* crc32c(uuid+inum+inode) BE */
681         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
682         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
683         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
684         __le32  i_crtime;       /* File Creation time */
685         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
686         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
687 };
688
689 struct move_extent {
690         __u32 reserved;         /* should be zero */
691         __u32 donor_fd;         /* donor file descriptor */
692         __u64 orig_start;       /* logical start offset in block for orig */
693         __u64 donor_start;      /* logical start offset in block for donor */
694         __u64 len;              /* block length to be moved */
695         __u64 moved_len;        /* moved block length */
696 };
697
698 #define EXT4_EPOCH_BITS 2
699 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
700 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
701
702 /*
703  * Extended fields will fit into an inode if the filesystem was formatted
704  * with large inodes (-I 256 or larger) and there are not currently any EAs
705  * consuming all of the available space. For new inodes we always reserve
706  * enough space for the kernel's known extended fields, but for inodes
707  * created with an old kernel this might not have been the case. None of
708  * the extended inode fields is critical for correct filesystem operation.
709  * This macro checks if a certain field fits in the inode. Note that
710  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
711  */
712 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
713         ((offsetof(typeof(*ext4_inode), field) +        \
714           sizeof((ext4_inode)->field))                  \
715         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
716             (einode)->i_extra_isize))                   \
717
718 static inline __le32 ext4_encode_extra_time(struct timespec *time)
719 {
720        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
721                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
722                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
723 }
724
725 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
726 {
727        if (sizeof(time->tv_sec) > 4)
728                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
729                                << 32;
730        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
731 }
732
733 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
734 do {                                                                           \
735         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
736         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
737                 (raw_inode)->xtime ## _extra =                                 \
738                                 ext4_encode_extra_time(&(inode)->xtime);       \
739 } while (0)
740
741 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
742 do {                                                                           \
743         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
744                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
745         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
746                 (raw_inode)->xtime ## _extra =                                 \
747                                 ext4_encode_extra_time(&(einode)->xtime);      \
748 } while (0)
749
750 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
751 do {                                                                           \
752         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
753         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
754                 ext4_decode_extra_time(&(inode)->xtime,                        \
755                                        raw_inode->xtime ## _extra);            \
756         else                                                                   \
757                 (inode)->xtime.tv_nsec = 0;                                    \
758 } while (0)
759
760 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
761 do {                                                                           \
762         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
763                 (einode)->xtime.tv_sec =                                       \
764                         (signed)le32_to_cpu((raw_inode)->xtime);               \
765         else                                                                   \
766                 (einode)->xtime.tv_sec = 0;                                    \
767         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
768                 ext4_decode_extra_time(&(einode)->xtime,                       \
769                                        raw_inode->xtime ## _extra);            \
770         else                                                                   \
771                 (einode)->xtime.tv_nsec = 0;                                   \
772 } while (0)
773
774 #define i_disk_version osd1.linux1.l_i_version
775
776 #if defined(__KERNEL__) || defined(__linux__)
777 #define i_reserved1     osd1.linux1.l_i_reserved1
778 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
779 #define i_blocks_high   osd2.linux2.l_i_blocks_high
780 #define i_uid_low       i_uid
781 #define i_gid_low       i_gid
782 #define i_uid_high      osd2.linux2.l_i_uid_high
783 #define i_gid_high      osd2.linux2.l_i_gid_high
784 #define i_checksum_lo   osd2.linux2.l_i_checksum_lo
785
786 #elif defined(__GNU__)
787
788 #define i_translator    osd1.hurd1.h_i_translator
789 #define i_uid_high      osd2.hurd2.h_i_uid_high
790 #define i_gid_high      osd2.hurd2.h_i_gid_high
791 #define i_author        osd2.hurd2.h_i_author
792
793 #elif defined(__masix__)
794
795 #define i_reserved1     osd1.masix1.m_i_reserved1
796 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
797 #define i_reserved2     osd2.masix2.m_i_reserved2
798
799 #endif /* defined(__KERNEL__) || defined(__linux__) */
800
801 #include "extents_status.h"
802
803 /*
804  * fourth extended file system inode data in memory
805  */
806 struct ext4_inode_info {
807         __le32  i_data[15];     /* unconverted */
808         __u32   i_dtime;
809         ext4_fsblk_t    i_file_acl;
810
811         /*
812          * i_block_group is the number of the block group which contains
813          * this file's inode.  Constant across the lifetime of the inode,
814          * it is ued for making block allocation decisions - we try to
815          * place a file's data blocks near its inode block, and new inodes
816          * near to their parent directory's inode.
817          */
818         ext4_group_t    i_block_group;
819         ext4_lblk_t     i_dir_start_lookup;
820 #if (BITS_PER_LONG < 64)
821         unsigned long   i_state_flags;          /* Dynamic state flags */
822 #endif
823         unsigned long   i_flags;
824
825         /*
826          * Extended attributes can be read independently of the main file
827          * data. Taking i_mutex even when reading would cause contention
828          * between readers of EAs and writers of regular file data, so
829          * instead we synchronize on xattr_sem when reading or changing
830          * EAs.
831          */
832         struct rw_semaphore xattr_sem;
833
834         struct list_head i_orphan;      /* unlinked but open inodes */
835
836         /*
837          * i_disksize keeps track of what the inode size is ON DISK, not
838          * in memory.  During truncate, i_size is set to the new size by
839          * the VFS prior to calling ext4_truncate(), but the filesystem won't
840          * set i_disksize to 0 until the truncate is actually under way.
841          *
842          * The intent is that i_disksize always represents the blocks which
843          * are used by this file.  This allows recovery to restart truncate
844          * on orphans if we crash during truncate.  We actually write i_disksize
845          * into the on-disk inode when writing inodes out, instead of i_size.
846          *
847          * The only time when i_disksize and i_size may be different is when
848          * a truncate is in progress.  The only things which change i_disksize
849          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
850          */
851         loff_t  i_disksize;
852
853         /*
854          * i_data_sem is for serialising ext4_truncate() against
855          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
856          * data tree are chopped off during truncate. We can't do that in
857          * ext4 because whenever we perform intermediate commits during
858          * truncate, the inode and all the metadata blocks *must* be in a
859          * consistent state which allows truncation of the orphans to restart
860          * during recovery.  Hence we must fix the get_block-vs-truncate race
861          * by other means, so we have i_data_sem.
862          */
863         struct rw_semaphore i_data_sem;
864         struct inode vfs_inode;
865         struct jbd2_inode *jinode;
866
867         spinlock_t i_raw_lock;  /* protects updates to the raw inode */
868
869         /*
870          * File creation time. Its function is same as that of
871          * struct timespec i_{a,c,m}time in the generic inode.
872          */
873         struct timespec i_crtime;
874
875         /* mballoc */
876         struct list_head i_prealloc_list;
877         spinlock_t i_prealloc_lock;
878
879         /* extents status tree */
880         struct ext4_es_tree i_es_tree;
881         rwlock_t i_es_lock;
882         struct list_head i_es_list;
883         unsigned int i_es_all_nr;       /* protected by i_es_lock */
884         unsigned int i_es_shk_nr;       /* protected by i_es_lock */
885         ext4_lblk_t i_es_shrink_lblk;   /* Offset where we start searching for
886                                            extents to shrink. Protected by
887                                            i_es_lock  */
888
889         /* ialloc */
890         ext4_group_t    i_last_alloc_group;
891
892         /* allocation reservation info for delalloc */
893         /* In case of bigalloc, these refer to clusters rather than blocks */
894         unsigned int i_reserved_data_blocks;
895         unsigned int i_reserved_meta_blocks;
896         unsigned int i_allocated_meta_blocks;
897         ext4_lblk_t i_da_metadata_calc_last_lblock;
898         int i_da_metadata_calc_len;
899
900         /* on-disk additional length */
901         __u16 i_extra_isize;
902
903         /* Indicate the inline data space. */
904         u16 i_inline_off;
905         u16 i_inline_size;
906
907 #ifdef CONFIG_QUOTA
908         /* quota space reservation, managed internally by quota code */
909         qsize_t i_reserved_quota;
910 #endif
911
912         /* Lock protecting lists below */
913         spinlock_t i_completed_io_lock;
914         /*
915          * Completed IOs that need unwritten extents handling and have
916          * transaction reserved
917          */
918         struct list_head i_rsv_conversion_list;
919         /*
920          * Completed IOs that need unwritten extents handling and don't have
921          * transaction reserved
922          */
923         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
924         atomic_t i_unwritten; /* Nr. of inflight conversions pending */
925         struct work_struct i_rsv_conversion_work;
926
927         spinlock_t i_block_reservation_lock;
928
929         /*
930          * Transactions that contain inode's metadata needed to complete
931          * fsync and fdatasync, respectively.
932          */
933         tid_t i_sync_tid;
934         tid_t i_datasync_tid;
935
936 #ifdef CONFIG_QUOTA
937         struct dquot *i_dquot[MAXQUOTAS];
938 #endif
939
940         /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
941         __u32 i_csum_seed;
942 };
943
944 /*
945  * File system states
946  */
947 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
948 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
949 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
950
951 /*
952  * Misc. filesystem flags
953  */
954 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
955 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
956 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
957
958 /*
959  * Mount flags set via mount options or defaults
960  */
961 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
962 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
963 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
964 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
965 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
966 #define EXT4_MOUNT_ERRORS_MASK          0x00070
967 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
968 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
969 #ifdef CONFIG_FS_DAX
970 #define EXT4_MOUNT_DAX                  0x00200 /* Direct Access */
971 #else
972 #define EXT4_MOUNT_DAX                  0
973 #endif
974 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
975 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
976 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
977 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
978 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
979 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
980 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
981 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
982 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
983 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
984 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
985 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
986 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
987 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
988 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
989 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
990 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
991 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
992 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
993 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
994 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
995
996 /*
997  * Mount flags set either automatically (could not be set by mount option)
998  * based on per file system feature or property or in special cases such as
999  * distinguishing between explicit mount option definition and default.
1000  */
1001 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
1002                                                       specified delalloc */
1003 #define EXT4_MOUNT2_STD_GROUP_SIZE      0x00000002 /* We have standard group
1004                                                       size of blocksize * 8
1005                                                       blocks */
1006 #define EXT4_MOUNT2_HURD_COMPAT         0x00000004 /* Support HURD-castrated
1007                                                       file systems */
1008
1009 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
1010                                                 ~EXT4_MOUNT_##opt
1011 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
1012                                                 EXT4_MOUNT_##opt
1013 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
1014                                          EXT4_MOUNT_##opt)
1015
1016 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
1017                                                 ~EXT4_MOUNT2_##opt
1018 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
1019                                                 EXT4_MOUNT2_##opt
1020 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
1021                                          EXT4_MOUNT2_##opt)
1022
1023 #define ext4_test_and_set_bit           __test_and_set_bit_le
1024 #define ext4_set_bit                    __set_bit_le
1025 #define ext4_set_bit_atomic             ext2_set_bit_atomic
1026 #define ext4_test_and_clear_bit         __test_and_clear_bit_le
1027 #define ext4_clear_bit                  __clear_bit_le
1028 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
1029 #define ext4_test_bit                   test_bit_le
1030 #define ext4_find_next_zero_bit         find_next_zero_bit_le
1031 #define ext4_find_next_bit              find_next_bit_le
1032
1033 extern void ext4_set_bits(void *bm, int cur, int len);
1034
1035 /*
1036  * Maximal mount counts between two filesystem checks
1037  */
1038 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
1039 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
1040
1041 /*
1042  * Behaviour when detecting errors
1043  */
1044 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
1045 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
1046 #define EXT4_ERRORS_PANIC               3       /* Panic */
1047 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
1048
1049 /* Metadata checksum algorithm codes */
1050 #define EXT4_CRC32C_CHKSUM              1
1051
1052 /* Encryption algorithms */
1053 #define EXT4_ENCRYPTION_MODE_INVALID            0
1054 #define EXT4_ENCRYPTION_MODE_AES_256_XTS        1
1055 #define EXT4_ENCRYPTION_MODE_AES_256_GCM        2
1056 #define EXT4_ENCRYPTION_MODE_AES_256_CBC        3
1057
1058 /*
1059  * Structure of the super block
1060  */
1061 struct ext4_super_block {
1062 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1063         __le32  s_blocks_count_lo;      /* Blocks count */
1064         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1065         __le32  s_free_blocks_count_lo; /* Free blocks count */
1066 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1067         __le32  s_first_data_block;     /* First Data Block */
1068         __le32  s_log_block_size;       /* Block size */
1069         __le32  s_log_cluster_size;     /* Allocation cluster size */
1070 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1071         __le32  s_clusters_per_group;   /* # Clusters per group */
1072         __le32  s_inodes_per_group;     /* # Inodes per group */
1073         __le32  s_mtime;                /* Mount time */
1074 /*30*/  __le32  s_wtime;                /* Write time */
1075         __le16  s_mnt_count;            /* Mount count */
1076         __le16  s_max_mnt_count;        /* Maximal mount count */
1077         __le16  s_magic;                /* Magic signature */
1078         __le16  s_state;                /* File system state */
1079         __le16  s_errors;               /* Behaviour when detecting errors */
1080         __le16  s_minor_rev_level;      /* minor revision level */
1081 /*40*/  __le32  s_lastcheck;            /* time of last check */
1082         __le32  s_checkinterval;        /* max. time between checks */
1083         __le32  s_creator_os;           /* OS */
1084         __le32  s_rev_level;            /* Revision level */
1085 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1086         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1087         /*
1088          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1089          *
1090          * Note: the difference between the compatible feature set and
1091          * the incompatible feature set is that if there is a bit set
1092          * in the incompatible feature set that the kernel doesn't
1093          * know about, it should refuse to mount the filesystem.
1094          *
1095          * e2fsck's requirements are more strict; if it doesn't know
1096          * about a feature in either the compatible or incompatible
1097          * feature set, it must abort and not try to meddle with
1098          * things it doesn't understand...
1099          */
1100         __le32  s_first_ino;            /* First non-reserved inode */
1101         __le16  s_inode_size;           /* size of inode structure */
1102         __le16  s_block_group_nr;       /* block group # of this superblock */
1103         __le32  s_feature_compat;       /* compatible feature set */
1104 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1105         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1106 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1107 /*78*/  char    s_volume_name[16];      /* volume name */
1108 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1109 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1110         /*
1111          * Performance hints.  Directory preallocation should only
1112          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1113          */
1114         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1115         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1116         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1117         /*
1118          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1119          */
1120 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1121 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1122         __le32  s_journal_dev;          /* device number of journal file */
1123         __le32  s_last_orphan;          /* start of list of inodes to delete */
1124         __le32  s_hash_seed[4];         /* HTREE hash seed */
1125         __u8    s_def_hash_version;     /* Default hash version to use */
1126         __u8    s_jnl_backup_type;
1127         __le16  s_desc_size;            /* size of group descriptor */
1128 /*100*/ __le32  s_default_mount_opts;
1129         __le32  s_first_meta_bg;        /* First metablock block group */
1130         __le32  s_mkfs_time;            /* When the filesystem was created */
1131         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1132         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1133 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1134         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1135         __le32  s_free_blocks_count_hi; /* Free blocks count */
1136         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1137         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1138         __le32  s_flags;                /* Miscellaneous flags */
1139         __le16  s_raid_stride;          /* RAID stride */
1140         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1141         __le64  s_mmp_block;            /* Block for multi-mount protection */
1142         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1143         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1144         __u8    s_checksum_type;        /* metadata checksum algorithm used */
1145         __le16  s_reserved_pad;
1146         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1147         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1148         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1149         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1150                                               snapshot's future use */
1151         __le32  s_snapshot_list;        /* inode number of the head of the
1152                                            on-disk snapshot list */
1153 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1154         __le32  s_error_count;          /* number of fs errors */
1155         __le32  s_first_error_time;     /* first time an error happened */
1156         __le32  s_first_error_ino;      /* inode involved in first error */
1157         __le64  s_first_error_block;    /* block involved of first error */
1158         __u8    s_first_error_func[32]; /* function where the error happened */
1159         __le32  s_first_error_line;     /* line number where error happened */
1160         __le32  s_last_error_time;      /* most recent time of an error */
1161         __le32  s_last_error_ino;       /* inode involved in last error */
1162         __le32  s_last_error_line;      /* line number where error happened */
1163         __le64  s_last_error_block;     /* block involved of last error */
1164         __u8    s_last_error_func[32];  /* function where the error happened */
1165 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1166         __u8    s_mount_opts[64];
1167         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1168         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1169         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1170         __le32  s_backup_bgs[2];        /* groups with sparse_super2 SBs */
1171         __u8    s_encrypt_algos[4];     /* Encryption algorithms in use  */
1172         __le32  s_reserved[105];        /* Padding to the end of the block */
1173         __le32  s_checksum;             /* crc32c(superblock) */
1174 };
1175
1176 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1177
1178 #ifdef __KERNEL__
1179
1180 /*
1181  * run-time mount flags
1182  */
1183 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
1184 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
1185
1186 /* Number of quota types we support */
1187 #define EXT4_MAXQUOTAS 2
1188
1189 /*
1190  * fourth extended-fs super-block data in memory
1191  */
1192 struct ext4_sb_info {
1193         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1194         unsigned long s_inodes_per_block;/* Number of inodes per block */
1195         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1196         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1197         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1198         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1199         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1200         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1201         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1202         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1203         unsigned long s_overhead;  /* # of fs overhead clusters */
1204         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1205         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1206         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1207         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1208         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1209         struct buffer_head **s_group_desc;
1210         unsigned int s_mount_opt;
1211         unsigned int s_mount_opt2;
1212         unsigned int s_mount_flags;
1213         unsigned int s_def_mount_opt;
1214         ext4_fsblk_t s_sb_block;
1215         atomic64_t s_resv_clusters;
1216         kuid_t s_resuid;
1217         kgid_t s_resgid;
1218         unsigned short s_mount_state;
1219         unsigned short s_pad;
1220         int s_addr_per_block_bits;
1221         int s_desc_per_block_bits;
1222         int s_inode_size;
1223         int s_first_ino;
1224         unsigned int s_inode_readahead_blks;
1225         unsigned int s_inode_goal;
1226         spinlock_t s_next_gen_lock;
1227         u32 s_next_generation;
1228         u32 s_hash_seed[4];
1229         int s_def_hash_version;
1230         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1231         struct percpu_counter s_freeclusters_counter;
1232         struct percpu_counter s_freeinodes_counter;
1233         struct percpu_counter s_dirs_counter;
1234         struct percpu_counter s_dirtyclusters_counter;
1235         struct blockgroup_lock *s_blockgroup_lock;
1236         struct proc_dir_entry *s_proc;
1237         struct kobject s_kobj;
1238         struct completion s_kobj_unregister;
1239         struct super_block *s_sb;
1240
1241         /* Journaling */
1242         struct journal_s *s_journal;
1243         struct list_head s_orphan;
1244         struct mutex s_orphan_lock;
1245         unsigned long s_resize_flags;           /* Flags indicating if there
1246                                                    is a resizer */
1247         unsigned long s_commit_interval;
1248         u32 s_max_batch_time;
1249         u32 s_min_batch_time;
1250         struct block_device *journal_bdev;
1251 #ifdef CONFIG_QUOTA
1252         char *s_qf_names[EXT4_MAXQUOTAS];       /* Names of quota files with journalled quota */
1253         int s_jquota_fmt;                       /* Format of quota to use */
1254 #endif
1255         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1256         struct rb_root system_blks;
1257
1258 #ifdef EXTENTS_STATS
1259         /* ext4 extents stats */
1260         unsigned long s_ext_min;
1261         unsigned long s_ext_max;
1262         unsigned long s_depth_max;
1263         spinlock_t s_ext_stats_lock;
1264         unsigned long s_ext_blocks;
1265         unsigned long s_ext_extents;
1266 #endif
1267
1268         /* for buddy allocator */
1269         struct ext4_group_info ***s_group_info;
1270         struct inode *s_buddy_cache;
1271         spinlock_t s_md_lock;
1272         unsigned short *s_mb_offsets;
1273         unsigned int *s_mb_maxs;
1274         unsigned int s_group_info_size;
1275
1276         /* tunables */
1277         unsigned long s_stripe;
1278         unsigned int s_mb_stream_request;
1279         unsigned int s_mb_max_to_scan;
1280         unsigned int s_mb_min_to_scan;
1281         unsigned int s_mb_stats;
1282         unsigned int s_mb_order2_reqs;
1283         unsigned int s_mb_group_prealloc;
1284         unsigned int s_max_dir_size_kb;
1285         /* where last allocation was done - for stream allocation */
1286         unsigned long s_mb_last_group;
1287         unsigned long s_mb_last_start;
1288
1289         /* stats for buddy allocator */
1290         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1291         atomic_t s_bal_success; /* we found long enough chunks */
1292         atomic_t s_bal_allocated;       /* in blocks */
1293         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1294         atomic_t s_bal_goals;   /* goal hits */
1295         atomic_t s_bal_breaks;  /* too long searches */
1296         atomic_t s_bal_2orders; /* 2^order hits */
1297         spinlock_t s_bal_lock;
1298         unsigned long s_mb_buddies_generated;
1299         unsigned long long s_mb_generation_time;
1300         atomic_t s_mb_lost_chunks;
1301         atomic_t s_mb_preallocated;
1302         atomic_t s_mb_discarded;
1303         atomic_t s_lock_busy;
1304
1305         /* locality groups */
1306         struct ext4_locality_group __percpu *s_locality_groups;
1307
1308         /* for write statistics */
1309         unsigned long s_sectors_written_start;
1310         u64 s_kbytes_written;
1311
1312         /* the size of zero-out chunk */
1313         unsigned int s_extent_max_zeroout_kb;
1314
1315         unsigned int s_log_groups_per_flex;
1316         struct flex_groups *s_flex_groups;
1317         ext4_group_t s_flex_groups_allocated;
1318
1319         /* workqueue for reserved extent conversions (buffered io) */
1320         struct workqueue_struct *rsv_conversion_wq;
1321
1322         /* timer for periodic error stats printing */
1323         struct timer_list s_err_report;
1324
1325         /* Lazy inode table initialization info */
1326         struct ext4_li_request *s_li_request;
1327         /* Wait multiplier for lazy initialization thread */
1328         unsigned int s_li_wait_mult;
1329
1330         /* Kernel thread for multiple mount protection */
1331         struct task_struct *s_mmp_tsk;
1332
1333         /* record the last minlen when FITRIM is called. */
1334         atomic_t s_last_trim_minblks;
1335
1336         /* Reference to checksum algorithm driver via cryptoapi */
1337         struct crypto_shash *s_chksum_driver;
1338
1339         /* Precomputed FS UUID checksum for seeding other checksums */
1340         __u32 s_csum_seed;
1341
1342         /* Reclaim extents from extent status tree */
1343         struct shrinker s_es_shrinker;
1344         struct list_head s_es_list;     /* List of inodes with reclaimable extents */
1345         long s_es_nr_inode;
1346         struct ext4_es_stats s_es_stats;
1347         struct mb_cache *s_mb_cache;
1348         spinlock_t s_es_lock ____cacheline_aligned_in_smp;
1349
1350         /* Ratelimit ext4 messages. */
1351         struct ratelimit_state s_err_ratelimit_state;
1352         struct ratelimit_state s_warning_ratelimit_state;
1353         struct ratelimit_state s_msg_ratelimit_state;
1354 };
1355
1356 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1357 {
1358         return sb->s_fs_info;
1359 }
1360 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1361 {
1362         return container_of(inode, struct ext4_inode_info, vfs_inode);
1363 }
1364
1365 static inline struct timespec ext4_current_time(struct inode *inode)
1366 {
1367         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1368                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1369 }
1370
1371 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1372 {
1373         return ino == EXT4_ROOT_INO ||
1374                 ino == EXT4_USR_QUOTA_INO ||
1375                 ino == EXT4_GRP_QUOTA_INO ||
1376                 ino == EXT4_BOOT_LOADER_INO ||
1377                 ino == EXT4_JOURNAL_INO ||
1378                 ino == EXT4_RESIZE_INO ||
1379                 (ino >= EXT4_FIRST_INO(sb) &&
1380                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1381 }
1382
1383 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
1384                                               struct ext4_io_end *io_end)
1385 {
1386         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
1387                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
1388                 atomic_inc(&EXT4_I(inode)->i_unwritten);
1389         }
1390 }
1391
1392 static inline ext4_io_end_t *ext4_inode_aio(struct inode *inode)
1393 {
1394         return inode->i_private;
1395 }
1396
1397 static inline void ext4_inode_aio_set(struct inode *inode, ext4_io_end_t *io)
1398 {
1399         inode->i_private = io;
1400 }
1401
1402 /*
1403  * Inode dynamic state flags
1404  */
1405 enum {
1406         EXT4_STATE_JDATA,               /* journaled data exists */
1407         EXT4_STATE_NEW,                 /* inode is newly created */
1408         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1409         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1410         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1411         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1412         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1413         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1414         EXT4_STATE_DIOREAD_LOCK,        /* Disable support for dio read
1415                                            nolocking */
1416         EXT4_STATE_MAY_INLINE_DATA,     /* may have in-inode data */
1417         EXT4_STATE_ORDERED_MODE,        /* data=ordered mode */
1418         EXT4_STATE_EXT_PRECACHED,       /* extents have been precached */
1419 };
1420
1421 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1422 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1423 {                                                                       \
1424         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1425 }                                                                       \
1426 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1427 {                                                                       \
1428         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1429 }                                                                       \
1430 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1431 {                                                                       \
1432         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1433 }
1434
1435 /* Add these declarations here only so that these functions can be
1436  * found by name.  Otherwise, they are very hard to locate. */
1437 static inline int ext4_test_inode_flag(struct inode *inode, int bit);
1438 static inline void ext4_set_inode_flag(struct inode *inode, int bit);
1439 static inline void ext4_clear_inode_flag(struct inode *inode, int bit);
1440 EXT4_INODE_BIT_FNS(flag, flags, 0)
1441
1442 /* Add these declarations here only so that these functions can be
1443  * found by name.  Otherwise, they are very hard to locate. */
1444 static inline int ext4_test_inode_state(struct inode *inode, int bit);
1445 static inline void ext4_set_inode_state(struct inode *inode, int bit);
1446 static inline void ext4_clear_inode_state(struct inode *inode, int bit);
1447 #if (BITS_PER_LONG < 64)
1448 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1449
1450 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1451 {
1452         (ei)->i_state_flags = 0;
1453 }
1454 #else
1455 EXT4_INODE_BIT_FNS(state, flags, 32)
1456
1457 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1458 {
1459         /* We depend on the fact that callers will set i_flags */
1460 }
1461 #endif
1462 #else
1463 /* Assume that user mode programs are passing in an ext4fs superblock, not
1464  * a kernel struct super_block.  This will allow us to call the feature-test
1465  * macros from user land. */
1466 #define EXT4_SB(sb)     (sb)
1467 #endif
1468
1469 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1470
1471 /*
1472  * Codes for operating systems
1473  */
1474 #define EXT4_OS_LINUX           0
1475 #define EXT4_OS_HURD            1
1476 #define EXT4_OS_MASIX           2
1477 #define EXT4_OS_FREEBSD         3
1478 #define EXT4_OS_LITES           4
1479
1480 /*
1481  * Revision levels
1482  */
1483 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1484 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1485
1486 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1487 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1488
1489 #define EXT4_GOOD_OLD_INODE_SIZE 128
1490
1491 /*
1492  * Feature set definitions
1493  */
1494
1495 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1496         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1497 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1498         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1499 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1500         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1501 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1502         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1503 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1504         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1505 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1506         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1507 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1508         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1509 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1510         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1511 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1512         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1513
1514 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1515 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1516 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1517 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1518 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1519 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1520 #define EXT4_FEATURE_COMPAT_SPARSE_SUPER2       0x0200
1521
1522 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1523 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1524 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1525 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1526 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1527 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1528 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1529 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1530 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1531 /*
1532  * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM).  When
1533  * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
1534  * all other data structures' checksums.  However, the METADATA_CSUM and
1535  * GDT_CSUM bits are mutually exclusive.
1536  */
1537 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM    0x0400
1538 #define EXT4_FEATURE_RO_COMPAT_READONLY         0x1000
1539
1540 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1541 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1542 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1543 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1544 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1545 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1546 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1547 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1548 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1549 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1550 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1551 #define EXT4_FEATURE_INCOMPAT_BG_USE_META_CSUM  0x2000 /* use crc32c for bg */
1552 #define EXT4_FEATURE_INCOMPAT_LARGEDIR          0x4000 /* >2GB or 3-lvl htree */
1553 #define EXT4_FEATURE_INCOMPAT_INLINE_DATA       0x8000 /* data in inode */
1554 #define EXT4_FEATURE_INCOMPAT_ENCRYPT           0x10000
1555
1556 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1557 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1558                                          EXT4_FEATURE_INCOMPAT_META_BG)
1559 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1560                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1561                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1562
1563 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1564 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1565                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1566                                          EXT4_FEATURE_INCOMPAT_META_BG)
1567 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1568                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1569                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1570
1571 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1572 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1573                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1574                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1575                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1576                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1577                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1578                                          EXT4_FEATURE_INCOMPAT_MMP |    \
1579                                          EXT4_FEATURE_INCOMPAT_INLINE_DATA)
1580 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1581                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1582                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1583                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1584                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1585                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1586                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1587                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
1588                                          EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
1589                                          EXT4_FEATURE_RO_COMPAT_QUOTA)
1590
1591 /*
1592  * Default values for user and/or group using reserved blocks
1593  */
1594 #define EXT4_DEF_RESUID         0
1595 #define EXT4_DEF_RESGID         0
1596
1597 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1598
1599 /*
1600  * Default mount options
1601  */
1602 #define EXT4_DEFM_DEBUG         0x0001
1603 #define EXT4_DEFM_BSDGROUPS     0x0002
1604 #define EXT4_DEFM_XATTR_USER    0x0004
1605 #define EXT4_DEFM_ACL           0x0008
1606 #define EXT4_DEFM_UID16         0x0010
1607 #define EXT4_DEFM_JMODE         0x0060
1608 #define EXT4_DEFM_JMODE_DATA    0x0020
1609 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1610 #define EXT4_DEFM_JMODE_WBACK   0x0060
1611 #define EXT4_DEFM_NOBARRIER     0x0100
1612 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1613 #define EXT4_DEFM_DISCARD       0x0400
1614 #define EXT4_DEFM_NODELALLOC    0x0800
1615
1616 /*
1617  * Default journal batch times
1618  */
1619 #define EXT4_DEF_MIN_BATCH_TIME 0
1620 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1621
1622 /*
1623  * Minimum number of groups in a flexgroup before we separate out
1624  * directories into the first block group of a flexgroup
1625  */
1626 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1627
1628 /*
1629  * Structure of a directory entry
1630  */
1631 #define EXT4_NAME_LEN 255
1632
1633 struct ext4_dir_entry {
1634         __le32  inode;                  /* Inode number */
1635         __le16  rec_len;                /* Directory entry length */
1636         __le16  name_len;               /* Name length */
1637         char    name[EXT4_NAME_LEN];    /* File name */
1638 };
1639
1640 /*
1641  * The new version of the directory entry.  Since EXT4 structures are
1642  * stored in intel byte order, and the name_len field could never be
1643  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1644  * file_type field.
1645  */
1646 struct ext4_dir_entry_2 {
1647         __le32  inode;                  /* Inode number */
1648         __le16  rec_len;                /* Directory entry length */
1649         __u8    name_len;               /* Name length */
1650         __u8    file_type;
1651         char    name[EXT4_NAME_LEN];    /* File name */
1652 };
1653
1654 /*
1655  * This is a bogus directory entry at the end of each leaf block that
1656  * records checksums.
1657  */
1658 struct ext4_dir_entry_tail {
1659         __le32  det_reserved_zero1;     /* Pretend to be unused */
1660         __le16  det_rec_len;            /* 12 */
1661         __u8    det_reserved_zero2;     /* Zero name length */
1662         __u8    det_reserved_ft;        /* 0xDE, fake file type */
1663         __le32  det_checksum;           /* crc32c(uuid+inum+dirblock) */
1664 };
1665
1666 #define EXT4_DIRENT_TAIL(block, blocksize) \
1667         ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
1668                                         ((blocksize) - \
1669                                          sizeof(struct ext4_dir_entry_tail))))
1670
1671 /*
1672  * Ext4 directory file types.  Only the low 3 bits are used.  The
1673  * other bits are reserved for now.
1674  */
1675 #define EXT4_FT_UNKNOWN         0
1676 #define EXT4_FT_REG_FILE        1
1677 #define EXT4_FT_DIR             2
1678 #define EXT4_FT_CHRDEV          3
1679 #define EXT4_FT_BLKDEV          4
1680 #define EXT4_FT_FIFO            5
1681 #define EXT4_FT_SOCK            6
1682 #define EXT4_FT_SYMLINK         7
1683
1684 #define EXT4_FT_MAX             8
1685
1686 #define EXT4_FT_DIR_CSUM        0xDE
1687
1688 /*
1689  * EXT4_DIR_PAD defines the directory entries boundaries
1690  *
1691  * NOTE: It must be a multiple of 4
1692  */
1693 #define EXT4_DIR_PAD                    4
1694 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1695 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1696                                          ~EXT4_DIR_ROUND)
1697 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1698
1699 /*
1700  * If we ever get support for fs block sizes > page_size, we'll need
1701  * to remove the #if statements in the next two functions...
1702  */
1703 static inline unsigned int
1704 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1705 {
1706         unsigned len = le16_to_cpu(dlen);
1707
1708 #if (PAGE_CACHE_SIZE >= 65536)
1709         if (len == EXT4_MAX_REC_LEN || len == 0)
1710                 return blocksize;
1711         return (len & 65532) | ((len & 3) << 16);
1712 #else
1713         return len;
1714 #endif
1715 }
1716
1717 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1718 {
1719         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1720                 BUG();
1721 #if (PAGE_CACHE_SIZE >= 65536)
1722         if (len < 65536)
1723                 return cpu_to_le16(len);
1724         if (len == blocksize) {
1725                 if (blocksize == 65536)
1726                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1727                 else
1728                         return cpu_to_le16(0);
1729         }
1730         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1731 #else
1732         return cpu_to_le16(len);
1733 #endif
1734 }
1735
1736 /*
1737  * Hash Tree Directory indexing
1738  * (c) Daniel Phillips, 2001
1739  */
1740
1741 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1742                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1743                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1744 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1745 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1746
1747 /* Legal values for the dx_root hash_version field: */
1748
1749 #define DX_HASH_LEGACY          0
1750 #define DX_HASH_HALF_MD4        1
1751 #define DX_HASH_TEA             2
1752 #define DX_HASH_LEGACY_UNSIGNED 3
1753 #define DX_HASH_HALF_MD4_UNSIGNED       4
1754 #define DX_HASH_TEA_UNSIGNED            5
1755
1756 static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
1757                               const void *address, unsigned int length)
1758 {
1759         struct {
1760                 struct shash_desc shash;
1761                 char ctx[4];
1762         } desc;
1763         int err;
1764
1765         BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
1766
1767         desc.shash.tfm = sbi->s_chksum_driver;
1768         desc.shash.flags = 0;
1769         *(u32 *)desc.ctx = crc;
1770
1771         err = crypto_shash_update(&desc.shash, address, length);
1772         BUG_ON(err);
1773
1774         return *(u32 *)desc.ctx;
1775 }
1776
1777 #ifdef __KERNEL__
1778
1779 /* hash info structure used by the directory hash */
1780 struct dx_hash_info
1781 {
1782         u32             hash;
1783         u32             minor_hash;
1784         int             hash_version;
1785         u32             *seed;
1786 };
1787
1788
1789 /* 32 and 64 bit signed EOF for dx directories */
1790 #define EXT4_HTREE_EOF_32BIT   ((1UL  << (32 - 1)) - 1)
1791 #define EXT4_HTREE_EOF_64BIT   ((1ULL << (64 - 1)) - 1)
1792
1793
1794 /*
1795  * Control parameters used by ext4_htree_next_block
1796  */
1797 #define HASH_NB_ALWAYS          1
1798
1799
1800 /*
1801  * Describe an inode's exact location on disk and in memory
1802  */
1803 struct ext4_iloc
1804 {
1805         struct buffer_head *bh;
1806         unsigned long offset;
1807         ext4_group_t block_group;
1808 };
1809
1810 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1811 {
1812         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1813 }
1814
1815 /*
1816  * This structure is stuffed into the struct file's private_data field
1817  * for directories.  It is where we put information so that we can do
1818  * readdir operations in hash tree order.
1819  */
1820 struct dir_private_info {
1821         struct rb_root  root;
1822         struct rb_node  *curr_node;
1823         struct fname    *extra_fname;
1824         loff_t          last_pos;
1825         __u32           curr_hash;
1826         __u32           curr_minor_hash;
1827         __u32           next_hash;
1828 };
1829
1830 /* calculate the first block number of the group */
1831 static inline ext4_fsblk_t
1832 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1833 {
1834         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1835                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1836 }
1837
1838 /*
1839  * Special error return code only used by dx_probe() and its callers.
1840  */
1841 #define ERR_BAD_DX_DIR  (-(MAX_ERRNO - 1))
1842
1843 /*
1844  * Timeout and state flag for lazy initialization inode thread.
1845  */
1846 #define EXT4_DEF_LI_WAIT_MULT                   10
1847 #define EXT4_DEF_LI_MAX_START_DELAY             5
1848 #define EXT4_LAZYINIT_QUIT                      0x0001
1849 #define EXT4_LAZYINIT_RUNNING                   0x0002
1850
1851 /*
1852  * Lazy inode table initialization info
1853  */
1854 struct ext4_lazy_init {
1855         unsigned long           li_state;
1856         struct list_head        li_request_list;
1857         struct mutex            li_list_mtx;
1858 };
1859
1860 struct ext4_li_request {
1861         struct super_block      *lr_super;
1862         struct ext4_sb_info     *lr_sbi;
1863         ext4_group_t            lr_next_group;
1864         struct list_head        lr_request;
1865         unsigned long           lr_next_sched;
1866         unsigned long           lr_timeout;
1867 };
1868
1869 struct ext4_features {
1870         struct kobject f_kobj;
1871         struct completion f_kobj_unregister;
1872 };
1873
1874 /*
1875  * This structure will be used for multiple mount protection. It will be
1876  * written into the block number saved in the s_mmp_block field in the
1877  * superblock. Programs that check MMP should assume that if
1878  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1879  * to use the filesystem, regardless of how old the timestamp is.
1880  */
1881 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
1882 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1883 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
1884 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
1885
1886 struct mmp_struct {
1887         __le32  mmp_magic;              /* Magic number for MMP */
1888         __le32  mmp_seq;                /* Sequence no. updated periodically */
1889
1890         /*
1891          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1892          * purposes and do not affect the correctness of the algorithm
1893          */
1894         __le64  mmp_time;               /* Time last updated */
1895         char    mmp_nodename[64];       /* Node which last updated MMP block */
1896         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
1897
1898         /*
1899          * mmp_check_interval is used to verify if the MMP block has been
1900          * updated on the block device. The value is updated based on the
1901          * maximum time to write the MMP block during an update cycle.
1902          */
1903         __le16  mmp_check_interval;
1904
1905         __le16  mmp_pad1;
1906         __le32  mmp_pad2[226];
1907         __le32  mmp_checksum;           /* crc32c(uuid+mmp_block) */
1908 };
1909
1910 /* arguments passed to the mmp thread */
1911 struct mmpd_data {
1912         struct buffer_head *bh; /* bh from initial read_mmp_block() */
1913         struct super_block *sb;  /* super block of the fs */
1914 };
1915
1916 /*
1917  * Check interval multiplier
1918  * The MMP block is written every update interval and initially checked every
1919  * update interval x the multiplier (the value is then adapted based on the
1920  * write latency). The reason is that writes can be delayed under load and we
1921  * don't want readers to incorrectly assume that the filesystem is no longer
1922  * in use.
1923  */
1924 #define EXT4_MMP_CHECK_MULT             2UL
1925
1926 /*
1927  * Minimum interval for MMP checking in seconds.
1928  */
1929 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
1930
1931 /*
1932  * Maximum interval for MMP checking in seconds.
1933  */
1934 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
1935
1936 /*
1937  * Function prototypes
1938  */
1939
1940 /*
1941  * Ok, these declarations are also in <linux/kernel.h> but none of the
1942  * ext4 source programs needs to include it so they are duplicated here.
1943  */
1944 # define NORET_TYPE     /**/
1945 # define ATTRIB_NORET   __attribute__((noreturn))
1946 # define NORET_AND      noreturn,
1947
1948 /* bitmap.c */
1949 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
1950 void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1951                                 struct ext4_group_desc *gdp,
1952                                 struct buffer_head *bh, int sz);
1953 int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1954                                   struct ext4_group_desc *gdp,
1955                                   struct buffer_head *bh, int sz);
1956 void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1957                                 struct ext4_group_desc *gdp,
1958                                 struct buffer_head *bh);
1959 int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1960                                   struct ext4_group_desc *gdp,
1961                                   struct buffer_head *bh);
1962
1963 /* balloc.c */
1964 extern void ext4_get_group_no_and_offset(struct super_block *sb,
1965                                          ext4_fsblk_t blocknr,
1966                                          ext4_group_t *blockgrpp,
1967                                          ext4_grpblk_t *offsetp);
1968 extern ext4_group_t ext4_get_group_number(struct super_block *sb,
1969                                           ext4_fsblk_t block);
1970
1971 extern unsigned int ext4_block_group(struct super_block *sb,
1972                         ext4_fsblk_t blocknr);
1973 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1974                         ext4_fsblk_t blocknr);
1975 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1976 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1977                         ext4_group_t group);
1978 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1979                                          ext4_fsblk_t goal,
1980                                          unsigned int flags,
1981                                          unsigned long *count,
1982                                          int *errp);
1983 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
1984                                     s64 nclusters, unsigned int flags);
1985 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
1986 extern void ext4_check_blocks_bitmap(struct super_block *);
1987 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1988                                                     ext4_group_t block_group,
1989                                                     struct buffer_head ** bh);
1990 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1991
1992 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
1993                                                 ext4_group_t block_group);
1994 extern int ext4_wait_block_bitmap(struct super_block *sb,
1995                                   ext4_group_t block_group,
1996                                   struct buffer_head *bh);
1997 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1998                                                   ext4_group_t block_group);
1999 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
2000                                               ext4_group_t block_group,
2001                                               struct ext4_group_desc *gdp);
2002 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
2003
2004 /* dir.c */
2005 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
2006                                   struct file *,
2007                                   struct ext4_dir_entry_2 *,
2008                                   struct buffer_head *, char *, int,
2009                                   unsigned int);
2010 #define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset)      \
2011         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
2012                                         (de), (bh), (buf), (size), (offset)))
2013 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
2014                                     __u32 minor_hash,
2015                                     struct ext4_dir_entry_2 *dirent);
2016 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
2017 extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
2018                              struct buffer_head *bh,
2019                              void *buf, int buf_size,
2020                              const char *name, int namelen,
2021                              struct ext4_dir_entry_2 **dest_de);
2022 void ext4_insert_dentry(struct inode *inode,
2023                         struct ext4_dir_entry_2 *de,
2024                         int buf_size,
2025                         const char *name, int namelen);
2026 static inline void ext4_update_dx_flag(struct inode *inode)
2027 {
2028         if (!EXT4_HAS_COMPAT_FEATURE(inode->i_sb,
2029                                      EXT4_FEATURE_COMPAT_DIR_INDEX))
2030                 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
2031 }
2032 static unsigned char ext4_filetype_table[] = {
2033         DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
2034 };
2035
2036 static inline  unsigned char get_dtype(struct super_block *sb, int filetype)
2037 {
2038         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE) ||
2039             (filetype >= EXT4_FT_MAX))
2040                 return DT_UNKNOWN;
2041
2042         return ext4_filetype_table[filetype];
2043 }
2044 extern int ext4_check_all_de(struct inode *dir, struct buffer_head *bh,
2045                              void *buf, int buf_size);
2046
2047 /* fsync.c */
2048 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2049
2050 /* hash.c */
2051 extern int ext4fs_dirhash(const char *name, int len, struct
2052                           dx_hash_info *hinfo);
2053
2054 /* ialloc.c */
2055 extern struct inode *__ext4_new_inode(handle_t *, struct inode *, umode_t,
2056                                       const struct qstr *qstr, __u32 goal,
2057                                       uid_t *owner, int handle_type,
2058                                       unsigned int line_no, int nblocks);
2059
2060 #define ext4_new_inode(handle, dir, mode, qstr, goal, owner) \
2061         __ext4_new_inode((handle), (dir), (mode), (qstr), (goal), (owner), \
2062                          0, 0, 0)
2063 #define ext4_new_inode_start_handle(dir, mode, qstr, goal, owner, \
2064                                     type, nblocks)                  \
2065         __ext4_new_inode(NULL, (dir), (mode), (qstr), (goal), (owner), \
2066                          (type), __LINE__, (nblocks))
2067
2068
2069 extern void ext4_free_inode(handle_t *, struct inode *);
2070 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2071 extern unsigned long ext4_count_free_inodes(struct super_block *);
2072 extern unsigned long ext4_count_dirs(struct super_block *);
2073 extern void ext4_check_inodes_bitmap(struct super_block *);
2074 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2075 extern int ext4_init_inode_table(struct super_block *sb,
2076                                  ext4_group_t group, int barrier);
2077 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2078
2079 /* mballoc.c */
2080 extern long ext4_mb_stats;
2081 extern long ext4_mb_max_to_scan;
2082 extern int ext4_mb_init(struct super_block *);
2083 extern int ext4_mb_release(struct super_block *);
2084 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2085                                 struct ext4_allocation_request *, int *);
2086 extern int ext4_mb_reserve_blocks(struct super_block *, int);
2087 extern void ext4_discard_preallocations(struct inode *);
2088 extern int __init ext4_init_mballoc(void);
2089 extern void ext4_exit_mballoc(void);
2090 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2091                              struct buffer_head *bh, ext4_fsblk_t block,
2092                              unsigned long count, int flags);
2093 extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2094                                    ext4_group_t ngroups);
2095 extern int ext4_mb_add_groupinfo(struct super_block *sb,
2096                 ext4_group_t i, struct ext4_group_desc *desc);
2097 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2098                                 ext4_fsblk_t block, unsigned long count);
2099 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2100
2101 /* inode.c */
2102 struct buffer_head *ext4_getblk(handle_t *, struct inode *, ext4_lblk_t, int);
2103 struct buffer_head *ext4_bread(handle_t *, struct inode *, ext4_lblk_t, int);
2104 int ext4_get_block_write(struct inode *inode, sector_t iblock,
2105                          struct buffer_head *bh_result, int create);
2106 int ext4_get_block(struct inode *inode, sector_t iblock,
2107                                 struct buffer_head *bh_result, int create);
2108 int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2109                            struct buffer_head *bh, int create);
2110 int ext4_walk_page_buffers(handle_t *handle,
2111                            struct buffer_head *head,
2112                            unsigned from,
2113                            unsigned to,
2114                            int *partial,
2115                            int (*fn)(handle_t *handle,
2116                                      struct buffer_head *bh));
2117 int do_journal_get_write_access(handle_t *handle,
2118                                 struct buffer_head *bh);
2119 #define FALL_BACK_TO_NONDELALLOC 1
2120 #define CONVERT_INLINE_DATA      2
2121
2122 extern struct inode *ext4_iget(struct super_block *, unsigned long);
2123 extern struct inode *ext4_iget_normal(struct super_block *, unsigned long);
2124 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
2125 extern int  ext4_setattr(struct dentry *, struct iattr *);
2126 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
2127                                 struct kstat *stat);
2128 extern void ext4_evict_inode(struct inode *);
2129 extern void ext4_clear_inode(struct inode *);
2130 extern int  ext4_sync_inode(handle_t *, struct inode *);
2131 extern void ext4_dirty_inode(struct inode *, int);
2132 extern int ext4_change_inode_journal_flag(struct inode *, int);
2133 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
2134 extern int ext4_inode_attach_jinode(struct inode *inode);
2135 extern int ext4_can_truncate(struct inode *inode);
2136 extern void ext4_truncate(struct inode *);
2137 extern int ext4_punch_hole(struct inode *inode, loff_t offset, loff_t length);
2138 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
2139 extern void ext4_set_inode_flags(struct inode *);
2140 extern void ext4_get_inode_flags(struct ext4_inode_info *);
2141 extern int ext4_alloc_da_blocks(struct inode *inode);
2142 extern void ext4_set_aops(struct inode *inode);
2143 extern int ext4_writepage_trans_blocks(struct inode *);
2144 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
2145 extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
2146                              loff_t lstart, loff_t lend);
2147 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2148 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
2149 extern void ext4_da_update_reserve_space(struct inode *inode,
2150                                         int used, int quota_claim);
2151
2152 /* indirect.c */
2153 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
2154                                 struct ext4_map_blocks *map, int flags);
2155 extern ssize_t ext4_ind_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
2156                                   loff_t offset);
2157 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
2158 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
2159 extern void ext4_ind_truncate(handle_t *, struct inode *inode);
2160 extern int ext4_ind_remove_space(handle_t *handle, struct inode *inode,
2161                                  ext4_lblk_t start, ext4_lblk_t end);
2162
2163 /* ioctl.c */
2164 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
2165 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
2166
2167 /* migrate.c */
2168 extern int ext4_ext_migrate(struct inode *);
2169 extern int ext4_ind_migrate(struct inode *inode);
2170
2171 /* namei.c */
2172 extern int ext4_dirent_csum_verify(struct inode *inode,
2173                                    struct ext4_dir_entry *dirent);
2174 extern int ext4_orphan_add(handle_t *, struct inode *);
2175 extern int ext4_orphan_del(handle_t *, struct inode *);
2176 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
2177                                 __u32 start_minor_hash, __u32 *next_hash);
2178 extern int search_dir(struct buffer_head *bh,
2179                       char *search_buf,
2180                       int buf_size,
2181                       struct inode *dir,
2182                       const struct qstr *d_name,
2183                       unsigned int offset,
2184                       struct ext4_dir_entry_2 **res_dir);
2185 extern int ext4_generic_delete_entry(handle_t *handle,
2186                                      struct inode *dir,
2187                                      struct ext4_dir_entry_2 *de_del,
2188                                      struct buffer_head *bh,
2189                                      void *entry_buf,
2190                                      int buf_size,
2191                                      int csum_size);
2192
2193 /* resize.c */
2194 extern int ext4_group_add(struct super_block *sb,
2195                                 struct ext4_new_group_data *input);
2196 extern int ext4_group_extend(struct super_block *sb,
2197                                 struct ext4_super_block *es,
2198                                 ext4_fsblk_t n_blocks_count);
2199 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
2200
2201 /* super.c */
2202 extern int ext4_calculate_overhead(struct super_block *sb);
2203 extern void ext4_superblock_csum_set(struct super_block *sb);
2204 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
2205 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
2206 extern int ext4_alloc_flex_bg_array(struct super_block *sb,
2207                                     ext4_group_t ngroup);
2208 extern const char *ext4_decode_error(struct super_block *sb, int errno,
2209                                      char nbuf[16]);
2210
2211 extern __printf(4, 5)
2212 void __ext4_error(struct super_block *, const char *, unsigned int,
2213                   const char *, ...);
2214 extern __printf(5, 6)
2215 void __ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
2216                       const char *, ...);
2217 extern __printf(5, 6)
2218 void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
2219                      const char *, ...);
2220 extern void __ext4_std_error(struct super_block *, const char *,
2221                              unsigned int, int);
2222 extern __printf(4, 5)
2223 void __ext4_abort(struct super_block *, const char *, unsigned int,
2224                   const char *, ...);
2225 extern __printf(4, 5)
2226 void __ext4_warning(struct super_block *, const char *, unsigned int,
2227                     const char *, ...);
2228 extern __printf(3, 4)
2229 void __ext4_msg(struct super_block *, const char *, const char *, ...);
2230 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
2231                            const char *, unsigned int, const char *);
2232 extern __printf(7, 8)
2233 void __ext4_grp_locked_error(const char *, unsigned int,
2234                              struct super_block *, ext4_group_t,
2235                              unsigned long, ext4_fsblk_t,
2236                              const char *, ...);
2237
2238 #ifdef CONFIG_PRINTK
2239
2240 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2241         __ext4_error_inode(inode, func, line, block, fmt, ##__VA_ARGS__)
2242 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2243         __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__)
2244 #define ext4_error(sb, fmt, ...)                                        \
2245         __ext4_error(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2246 #define ext4_abort(sb, fmt, ...)                                        \
2247         __ext4_abort(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2248 #define ext4_warning(sb, fmt, ...)                                      \
2249         __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2250 #define ext4_msg(sb, level, fmt, ...)                           \
2251         __ext4_msg(sb, level, fmt, ##__VA_ARGS__)
2252 #define dump_mmp_msg(sb, mmp, msg)                                      \
2253         __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg)
2254 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2255         __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \
2256                                 fmt, ##__VA_ARGS__)
2257
2258 #else
2259
2260 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2261 do {                                                                    \
2262         no_printk(fmt, ##__VA_ARGS__);                                  \
2263         __ext4_error_inode(inode, "", 0, block, " ");                   \
2264 } while (0)
2265 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2266 do {                                                                    \
2267         no_printk(fmt, ##__VA_ARGS__);                                  \
2268         __ext4_error_file(file, "", 0, block, " ");                     \
2269 } while (0)
2270 #define ext4_error(sb, fmt, ...)                                        \
2271 do {                                                                    \
2272         no_printk(fmt, ##__VA_ARGS__);                                  \
2273         __ext4_error(sb, "", 0, " ");                                   \
2274 } while (0)
2275 #define ext4_abort(sb, fmt, ...)                                        \
2276 do {                                                                    \
2277         no_printk(fmt, ##__VA_ARGS__);                                  \
2278         __ext4_abort(sb, "", 0, " ");                                   \
2279 } while (0)
2280 #define ext4_warning(sb, fmt, ...)                                      \
2281 do {                                                                    \
2282         no_printk(fmt, ##__VA_ARGS__);                                  \
2283         __ext4_warning(sb, "", 0, " ");                                 \
2284 } while (0)
2285 #define ext4_msg(sb, level, fmt, ...)                                   \
2286 do {                                                                    \
2287         no_printk(fmt, ##__VA_ARGS__);                                  \
2288         __ext4_msg(sb, "", " ");                                        \
2289 } while (0)
2290 #define dump_mmp_msg(sb, mmp, msg)                                      \
2291         __dump_mmp_msg(sb, mmp, "", 0, "")
2292 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2293 do {                                                                    \
2294         no_printk(fmt, ##__VA_ARGS__);                          \
2295         __ext4_grp_locked_error("", 0, sb, grp, ino, block, " ");       \
2296 } while (0)
2297
2298 #endif
2299
2300 extern void ext4_update_dynamic_rev(struct super_block *sb);
2301 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
2302                                         __u32 compat);
2303 extern int ext4_update_rocompat_feature(handle_t *handle,
2304                                         struct super_block *sb, __u32 rocompat);
2305 extern int ext4_update_incompat_feature(handle_t *handle,
2306                                         struct super_block *sb, __u32 incompat);
2307 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
2308                                       struct ext4_group_desc *bg);
2309 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
2310                                       struct ext4_group_desc *bg);
2311 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
2312                                      struct ext4_group_desc *bg);
2313 extern __u32 ext4_free_group_clusters(struct super_block *sb,
2314                                       struct ext4_group_desc *bg);
2315 extern __u32 ext4_free_inodes_count(struct super_block *sb,
2316                                  struct ext4_group_desc *bg);
2317 extern __u32 ext4_used_dirs_count(struct super_block *sb,
2318                                 struct ext4_group_desc *bg);
2319 extern __u32 ext4_itable_unused_count(struct super_block *sb,
2320                                    struct ext4_group_desc *bg);
2321 extern void ext4_block_bitmap_set(struct super_block *sb,
2322                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2323 extern void ext4_inode_bitmap_set(struct super_block *sb,
2324                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2325 extern void ext4_inode_table_set(struct super_block *sb,
2326                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
2327 extern void ext4_free_group_clusters_set(struct super_block *sb,
2328                                          struct ext4_group_desc *bg,
2329                                          __u32 count);
2330 extern void ext4_free_inodes_set(struct super_block *sb,
2331                                 struct ext4_group_desc *bg, __u32 count);
2332 extern void ext4_used_dirs_set(struct super_block *sb,
2333                                 struct ext4_group_desc *bg, __u32 count);
2334 extern void ext4_itable_unused_set(struct super_block *sb,
2335                                    struct ext4_group_desc *bg, __u32 count);
2336 extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
2337                                        struct ext4_group_desc *gdp);
2338 extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
2339                                      struct ext4_group_desc *gdp);
2340 extern int ext4_register_li_request(struct super_block *sb,
2341                                     ext4_group_t first_not_zeroed);
2342
2343 static inline int ext4_has_group_desc_csum(struct super_block *sb)
2344 {
2345         return EXT4_HAS_RO_COMPAT_FEATURE(sb,
2346                                           EXT4_FEATURE_RO_COMPAT_GDT_CSUM) ||
2347                (EXT4_SB(sb)->s_chksum_driver != NULL);
2348 }
2349
2350 static inline int ext4_has_metadata_csum(struct super_block *sb)
2351 {
2352         WARN_ON_ONCE(EXT4_HAS_RO_COMPAT_FEATURE(sb,
2353                         EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
2354                      !EXT4_SB(sb)->s_chksum_driver);
2355
2356         return (EXT4_SB(sb)->s_chksum_driver != NULL);
2357 }
2358 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2359 {
2360         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2361                 le32_to_cpu(es->s_blocks_count_lo);
2362 }
2363
2364 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2365 {
2366         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2367                 le32_to_cpu(es->s_r_blocks_count_lo);
2368 }
2369
2370 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2371 {
2372         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2373                 le32_to_cpu(es->s_free_blocks_count_lo);
2374 }
2375
2376 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2377                                          ext4_fsblk_t blk)
2378 {
2379         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2380         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2381 }
2382
2383 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2384                                               ext4_fsblk_t blk)
2385 {
2386         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2387         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2388 }
2389
2390 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2391                                            ext4_fsblk_t blk)
2392 {
2393         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2394         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2395 }
2396
2397 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2398 {
2399         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2400                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2401                         le32_to_cpu(raw_inode->i_size_lo);
2402         else
2403                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2404 }
2405
2406 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2407 {
2408         raw_inode->i_size_lo = cpu_to_le32(i_size);
2409         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2410 }
2411
2412 static inline
2413 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2414                                             ext4_group_t group)
2415 {
2416          struct ext4_group_info ***grp_info;
2417          long indexv, indexh;
2418          BUG_ON(group >= EXT4_SB(sb)->s_groups_count);
2419          grp_info = EXT4_SB(sb)->s_group_info;
2420          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2421          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2422          return grp_info[indexv][indexh];
2423 }
2424
2425 /*
2426  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2427  * the locking protocol documented in the comments of ext4_group_add()
2428  * in resize.c
2429  */
2430 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2431 {
2432         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2433
2434         smp_rmb();
2435         return ngroups;
2436 }
2437
2438 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2439                                              ext4_group_t block_group)
2440 {
2441         return block_group >> sbi->s_log_groups_per_flex;
2442 }
2443
2444 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2445 {
2446         return 1 << sbi->s_log_groups_per_flex;
2447 }
2448
2449 #define ext4_std_error(sb, errno)                               \
2450 do {                                                            \
2451         if ((errno))                                            \
2452                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2453 } while (0)
2454
2455 #ifdef CONFIG_SMP
2456 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2457  * counters. So we need to make sure we have free clusters more
2458  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2459  */
2460 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2461 #else
2462 #define EXT4_FREECLUSTERS_WATERMARK 0
2463 #endif
2464
2465 /* Update i_disksize. Requires i_mutex to avoid races with truncate */
2466 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2467 {
2468         WARN_ON_ONCE(S_ISREG(inode->i_mode) &&
2469                      !mutex_is_locked(&inode->i_mutex));
2470         down_write(&EXT4_I(inode)->i_data_sem);
2471         if (newsize > EXT4_I(inode)->i_disksize)
2472                 EXT4_I(inode)->i_disksize = newsize;
2473         up_write(&EXT4_I(inode)->i_data_sem);
2474 }
2475
2476 /* Update i_size, i_disksize. Requires i_mutex to avoid races with truncate */
2477 static inline int ext4_update_inode_size(struct inode *inode, loff_t newsize)
2478 {
2479         int changed = 0;
2480
2481         if (newsize > inode->i_size) {
2482                 i_size_write(inode, newsize);
2483                 changed = 1;
2484         }
2485         if (newsize > EXT4_I(inode)->i_disksize) {
2486                 ext4_update_i_disksize(inode, newsize);
2487                 changed |= 2;
2488         }
2489         return changed;
2490 }
2491
2492 struct ext4_group_info {
2493         unsigned long   bb_state;
2494         struct rb_root  bb_free_root;
2495         ext4_grpblk_t   bb_first_free;  /* first free block */
2496         ext4_grpblk_t   bb_free;        /* total free blocks */
2497         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2498         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2499         struct          list_head bb_prealloc_list;
2500 #ifdef DOUBLE_CHECK
2501         void            *bb_bitmap;
2502 #endif
2503         struct rw_semaphore alloc_sem;
2504         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2505                                          * regions, index is order.
2506                                          * bb_counters[3] = 5 means
2507                                          * 5 free 8-block regions. */
2508 };
2509
2510 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2511 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2512 #define EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT     2
2513 #define EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT     3
2514
2515 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2516         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2517 #define EXT4_MB_GRP_BBITMAP_CORRUPT(grp)        \
2518         (test_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
2519 #define EXT4_MB_GRP_IBITMAP_CORRUPT(grp)        \
2520         (test_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
2521
2522 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2523         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2524 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2525         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2526 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2527         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2528
2529 #define EXT4_MAX_CONTENTION             8
2530 #define EXT4_CONTENTION_THRESHOLD       2
2531
2532 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2533                                               ext4_group_t group)
2534 {
2535         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2536 }
2537
2538 /*
2539  * Returns true if the filesystem is busy enough that attempts to
2540  * access the block group locks has run into contention.
2541  */
2542 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2543 {
2544         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2545 }
2546
2547 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2548 {
2549         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2550         if (spin_trylock(lock))
2551                 /*
2552                  * We're able to grab the lock right away, so drop the
2553                  * lock contention counter.
2554                  */
2555                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2556         else {
2557                 /*
2558                  * The lock is busy, so bump the contention counter,
2559                  * and then wait on the spin lock.
2560                  */
2561                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2562                                   EXT4_MAX_CONTENTION);
2563                 spin_lock(lock);
2564         }
2565 }
2566
2567 static inline void ext4_unlock_group(struct super_block *sb,
2568                                         ext4_group_t group)
2569 {
2570         spin_unlock(ext4_group_lock_ptr(sb, group));
2571 }
2572
2573 /*
2574  * Block validity checking
2575  */
2576 #define ext4_check_indirect_blockref(inode, bh)                         \
2577         ext4_check_blockref(__func__, __LINE__, inode,                  \
2578                             (__le32 *)(bh)->b_data,                     \
2579                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2580
2581 #define ext4_ind_check_inode(inode)                                     \
2582         ext4_check_blockref(__func__, __LINE__, inode,                  \
2583                             EXT4_I(inode)->i_data,                      \
2584                             EXT4_NDIR_BLOCKS)
2585
2586 /*
2587  * Inodes and files operations
2588  */
2589
2590 /* dir.c */
2591 extern const struct file_operations ext4_dir_operations;
2592
2593 /* file.c */
2594 extern const struct inode_operations ext4_file_inode_operations;
2595 extern const struct file_operations ext4_file_operations;
2596 extern const struct file_operations ext4_dax_file_operations;
2597 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2598
2599 /* inline.c */
2600 extern int ext4_get_max_inline_size(struct inode *inode);
2601 extern int ext4_find_inline_data_nolock(struct inode *inode);
2602 extern int ext4_init_inline_data(handle_t *handle, struct inode *inode,
2603                                  unsigned int len);
2604 extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
2605
2606 extern int ext4_readpage_inline(struct inode *inode, struct page *page);
2607 extern int ext4_try_to_write_inline_data(struct address_space *mapping,
2608                                          struct inode *inode,
2609                                          loff_t pos, unsigned len,
2610                                          unsigned flags,
2611                                          struct page **pagep);
2612 extern int ext4_write_inline_data_end(struct inode *inode,
2613                                       loff_t pos, unsigned len,
2614                                       unsigned copied,
2615                                       struct page *page);
2616 extern struct buffer_head *
2617 ext4_journalled_write_inline_data(struct inode *inode,
2618                                   unsigned len,
2619                                   struct page *page);
2620 extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
2621                                            struct inode *inode,
2622                                            loff_t pos, unsigned len,
2623                                            unsigned flags,
2624                                            struct page **pagep,
2625                                            void **fsdata);
2626 extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
2627                                          unsigned len, unsigned copied,
2628                                          struct page *page);
2629 extern int ext4_try_add_inline_entry(handle_t *handle, struct dentry *dentry,
2630                                      struct inode *inode);
2631 extern int ext4_try_create_inline_dir(handle_t *handle,
2632                                       struct inode *parent,
2633                                       struct inode *inode);
2634 extern int ext4_read_inline_dir(struct file *filp,
2635                                 struct dir_context *ctx,
2636                                 int *has_inline_data);
2637 extern int htree_inlinedir_to_tree(struct file *dir_file,
2638                                    struct inode *dir, ext4_lblk_t block,
2639                                    struct dx_hash_info *hinfo,
2640                                    __u32 start_hash, __u32 start_minor_hash,
2641                                    int *has_inline_data);
2642 extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
2643                                         const struct qstr *d_name,
2644                                         struct ext4_dir_entry_2 **res_dir,
2645                                         int *has_inline_data);
2646 extern int ext4_delete_inline_entry(handle_t *handle,
2647                                     struct inode *dir,
2648                                     struct ext4_dir_entry_2 *de_del,
2649                                     struct buffer_head *bh,
2650                                     int *has_inline_data);
2651 extern int empty_inline_dir(struct inode *dir, int *has_inline_data);
2652 extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
2653                                         struct ext4_dir_entry_2 **parent_de,
2654                                         int *retval);
2655 extern int ext4_inline_data_fiemap(struct inode *inode,
2656                                    struct fiemap_extent_info *fieinfo,
2657                                    int *has_inline, __u64 start, __u64 len);
2658 extern int ext4_try_to_evict_inline_data(handle_t *handle,
2659                                          struct inode *inode,
2660                                          int needed);
2661 extern void ext4_inline_data_truncate(struct inode *inode, int *has_inline);
2662
2663 extern int ext4_convert_inline_data(struct inode *inode);
2664
2665 static inline int ext4_has_inline_data(struct inode *inode)
2666 {
2667         return ext4_test_inode_flag(inode, EXT4_INODE_INLINE_DATA) &&
2668                EXT4_I(inode)->i_inline_off;
2669 }
2670
2671 /* namei.c */
2672 extern const struct inode_operations ext4_dir_inode_operations;
2673 extern const struct inode_operations ext4_special_inode_operations;
2674 extern struct dentry *ext4_get_parent(struct dentry *child);
2675 extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
2676                                  struct ext4_dir_entry_2 *de,
2677                                  int blocksize, int csum_size,
2678                                  unsigned int parent_ino, int dotdot_real_len);
2679 extern void initialize_dirent_tail(struct ext4_dir_entry_tail *t,
2680                                    unsigned int blocksize);
2681 extern int ext4_handle_dirty_dirent_node(handle_t *handle,
2682                                          struct inode *inode,
2683                                          struct buffer_head *bh);
2684 #define S_SHIFT 12
2685 static unsigned char ext4_type_by_mode[S_IFMT >> S_SHIFT] = {
2686         [S_IFREG >> S_SHIFT]    = EXT4_FT_REG_FILE,
2687         [S_IFDIR >> S_SHIFT]    = EXT4_FT_DIR,
2688         [S_IFCHR >> S_SHIFT]    = EXT4_FT_CHRDEV,
2689         [S_IFBLK >> S_SHIFT]    = EXT4_FT_BLKDEV,
2690         [S_IFIFO >> S_SHIFT]    = EXT4_FT_FIFO,
2691         [S_IFSOCK >> S_SHIFT]   = EXT4_FT_SOCK,
2692         [S_IFLNK >> S_SHIFT]    = EXT4_FT_SYMLINK,
2693 };
2694
2695 static inline void ext4_set_de_type(struct super_block *sb,
2696                                 struct ext4_dir_entry_2 *de,
2697                                 umode_t mode) {
2698         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE))
2699                 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
2700 }
2701
2702
2703 /* symlink.c */
2704 extern const struct inode_operations ext4_symlink_inode_operations;
2705 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2706
2707 /* block_validity */
2708 extern void ext4_release_system_zone(struct super_block *sb);
2709 extern int ext4_setup_system_zone(struct super_block *sb);
2710 extern int __init ext4_init_system_zone(void);
2711 extern void ext4_exit_system_zone(void);
2712 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2713                                  ext4_fsblk_t start_blk,
2714                                  unsigned int count);
2715 extern int ext4_check_blockref(const char *, unsigned int,
2716                                struct inode *, __le32 *, unsigned int);
2717
2718 /* extents.c */
2719 struct ext4_ext_path;
2720 struct ext4_extent;
2721
2722 /*
2723  * Maximum number of logical blocks in a file; ext4_extent's ee_block is
2724  * __le32.
2725  */
2726 #define EXT_MAX_BLOCKS  0xffffffff
2727
2728 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2729 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2730 extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
2731 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2732                                struct ext4_map_blocks *map, int flags);
2733 extern void ext4_ext_truncate(handle_t *, struct inode *);
2734 extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
2735                                  ext4_lblk_t end);
2736 extern void ext4_ext_init(struct super_block *);
2737 extern void ext4_ext_release(struct super_block *);
2738 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2739                           loff_t len);
2740 extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
2741                                           loff_t offset, ssize_t len);
2742 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2743                            struct ext4_map_blocks *map, int flags);
2744 extern int ext4_ext_calc_metadata_amount(struct inode *inode,
2745                                          ext4_lblk_t lblocks);
2746 extern int ext4_extent_tree_init(handle_t *, struct inode *);
2747 extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
2748                                                    int num,
2749                                                    struct ext4_ext_path *path);
2750 extern int ext4_can_extents_be_merged(struct inode *inode,
2751                                       struct ext4_extent *ex1,
2752                                       struct ext4_extent *ex2);
2753 extern int ext4_ext_insert_extent(handle_t *, struct inode *,
2754                                   struct ext4_ext_path **,
2755                                   struct ext4_extent *, int);
2756 extern struct ext4_ext_path *ext4_find_extent(struct inode *, ext4_lblk_t,
2757                                               struct ext4_ext_path **,
2758                                               int flags);
2759 extern void ext4_ext_drop_refs(struct ext4_ext_path *);
2760 extern int ext4_ext_check_inode(struct inode *inode);
2761 extern int ext4_find_delalloc_range(struct inode *inode,
2762                                     ext4_lblk_t lblk_start,
2763                                     ext4_lblk_t lblk_end);
2764 extern int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk);
2765 extern ext4_lblk_t ext4_ext_next_allocated_block(struct ext4_ext_path *path);
2766 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2767                         __u64 start, __u64 len);
2768 extern int ext4_ext_precache(struct inode *inode);
2769 extern int ext4_collapse_range(struct inode *inode, loff_t offset, loff_t len);
2770 extern int ext4_swap_extents(handle_t *handle, struct inode *inode1,
2771                                 struct inode *inode2, ext4_lblk_t lblk1,
2772                              ext4_lblk_t lblk2,  ext4_lblk_t count,
2773                              int mark_unwritten,int *err);
2774
2775 /* move_extent.c */
2776 extern void ext4_double_down_write_data_sem(struct inode *first,
2777                                             struct inode *second);
2778 extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
2779                                           struct inode *donor_inode);
2780 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2781                              __u64 start_orig, __u64 start_donor,
2782                              __u64 len, __u64 *moved_len);
2783
2784 /* page-io.c */
2785 extern int __init ext4_init_pageio(void);
2786 extern void ext4_exit_pageio(void);
2787 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2788 extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
2789 extern int ext4_put_io_end(ext4_io_end_t *io_end);
2790 extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
2791 extern void ext4_io_submit_init(struct ext4_io_submit *io,
2792                                 struct writeback_control *wbc);
2793 extern void ext4_end_io_rsv_work(struct work_struct *work);
2794 extern void ext4_io_submit(struct ext4_io_submit *io);
2795 extern int ext4_bio_write_page(struct ext4_io_submit *io,
2796                                struct page *page,
2797                                int len,
2798                                struct writeback_control *wbc,
2799                                bool keep_towrite);
2800
2801 /* mmp.c */
2802 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2803
2804 /*
2805  * Add new method to test whether block and inode bitmaps are properly
2806  * initialized. With uninit_bg reading the block from disk is not enough
2807  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2808  */
2809 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2810
2811 static inline int bitmap_uptodate(struct buffer_head *bh)
2812 {
2813         return (buffer_uptodate(bh) &&
2814                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2815 }
2816 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2817 {
2818         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2819 }
2820
2821 /*
2822  * Disable DIO read nolock optimization, so new dioreaders will be forced
2823  * to grab i_mutex
2824  */
2825 static inline void ext4_inode_block_unlocked_dio(struct inode *inode)
2826 {
2827         ext4_set_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2828         smp_mb();
2829 }
2830 static inline void ext4_inode_resume_unlocked_dio(struct inode *inode)
2831 {
2832         smp_mb();
2833         ext4_clear_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2834 }
2835
2836 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2837
2838 /* For ioend & aio unwritten conversion wait queues */
2839 #define EXT4_WQ_HASH_SZ         37
2840 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
2841                                             EXT4_WQ_HASH_SZ])
2842 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
2843                                              EXT4_WQ_HASH_SZ])
2844 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2845 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2846
2847 #define EXT4_RESIZING   0
2848 extern int ext4_resize_begin(struct super_block *sb);
2849 extern void ext4_resize_end(struct super_block *sb);
2850
2851 #endif  /* __KERNEL__ */
2852
2853 #endif  /* _EXT4_H */