]> Git Repo - linux.git/blob - fs/ext4/ext4.h
direct-io: Handle O_(D)SYNC AIO
[linux.git] / fs / ext4 / ext4.h
1 /*
2  *  ext4.h
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card ([email protected])
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/include/linux/minix_fs.h
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  */
15
16 #ifndef _EXT4_H
17 #define _EXT4_H
18
19 #include <linux/types.h>
20 #include <linux/blkdev.h>
21 #include <linux/magic.h>
22 #include <linux/jbd2.h>
23 #include <linux/quota.h>
24 #include <linux/rwsem.h>
25 #include <linux/rbtree.h>
26 #include <linux/seqlock.h>
27 #include <linux/mutex.h>
28 #include <linux/timer.h>
29 #include <linux/wait.h>
30 #include <linux/blockgroup_lock.h>
31 #include <linux/percpu_counter.h>
32 #include <crypto/hash.h>
33 #ifdef __KERNEL__
34 #include <linux/compat.h>
35 #endif
36
37 /*
38  * The fourth extended filesystem constants/structures
39  */
40
41 /*
42  * Define EXT4FS_DEBUG to produce debug messages
43  */
44 #undef EXT4FS_DEBUG
45
46 /*
47  * Debug code
48  */
49 #ifdef EXT4FS_DEBUG
50 #define ext4_debug(f, a...)                                             \
51         do {                                                            \
52                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
53                         __FILE__, __LINE__, __func__);                  \
54                 printk(KERN_DEBUG f, ## a);                             \
55         } while (0)
56 #else
57 #define ext4_debug(fmt, ...)    no_printk(fmt, ##__VA_ARGS__)
58 #endif
59
60 /*
61  * Turn on EXT_DEBUG to get lots of info about extents operations.
62  */
63 #define EXT_DEBUG__
64 #ifdef EXT_DEBUG
65 #define ext_debug(fmt, ...)     printk(fmt, ##__VA_ARGS__)
66 #else
67 #define ext_debug(fmt, ...)     no_printk(fmt, ##__VA_ARGS__)
68 #endif
69
70 #define EXT4_ERROR_INODE(inode, fmt, a...) \
71         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
72
73 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...)                 \
74         ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
75
76 #define EXT4_ERROR_FILE(file, block, fmt, a...)                         \
77         ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
78
79 /* data type for block offset of block group */
80 typedef int ext4_grpblk_t;
81
82 /* data type for filesystem-wide blocks number */
83 typedef unsigned long long ext4_fsblk_t;
84
85 /* data type for file logical block number */
86 typedef __u32 ext4_lblk_t;
87
88 /* data type for block group number */
89 typedef unsigned int ext4_group_t;
90
91 /*
92  * Flags used in mballoc's allocation_context flags field.
93  *
94  * Also used to show what's going on for debugging purposes when the
95  * flag field is exported via the traceport interface
96  */
97
98 /* prefer goal again. length */
99 #define EXT4_MB_HINT_MERGE              0x0001
100 /* blocks already reserved */
101 #define EXT4_MB_HINT_RESERVED           0x0002
102 /* metadata is being allocated */
103 #define EXT4_MB_HINT_METADATA           0x0004
104 /* first blocks in the file */
105 #define EXT4_MB_HINT_FIRST              0x0008
106 /* search for the best chunk */
107 #define EXT4_MB_HINT_BEST               0x0010
108 /* data is being allocated */
109 #define EXT4_MB_HINT_DATA               0x0020
110 /* don't preallocate (for tails) */
111 #define EXT4_MB_HINT_NOPREALLOC         0x0040
112 /* allocate for locality group */
113 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
114 /* allocate goal blocks or none */
115 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
116 /* goal is meaningful */
117 #define EXT4_MB_HINT_TRY_GOAL           0x0200
118 /* blocks already pre-reserved by delayed allocation */
119 #define EXT4_MB_DELALLOC_RESERVED       0x0400
120 /* We are doing stream allocation */
121 #define EXT4_MB_STREAM_ALLOC            0x0800
122 /* Use reserved root blocks if needed */
123 #define EXT4_MB_USE_ROOT_BLOCKS         0x1000
124 /* Use blocks from reserved pool */
125 #define EXT4_MB_USE_RESERVED            0x2000
126
127 struct ext4_allocation_request {
128         /* target inode for block we're allocating */
129         struct inode *inode;
130         /* how many blocks we want to allocate */
131         unsigned int len;
132         /* logical block in target inode */
133         ext4_lblk_t logical;
134         /* the closest logical allocated block to the left */
135         ext4_lblk_t lleft;
136         /* the closest logical allocated block to the right */
137         ext4_lblk_t lright;
138         /* phys. target (a hint) */
139         ext4_fsblk_t goal;
140         /* phys. block for the closest logical allocated block to the left */
141         ext4_fsblk_t pleft;
142         /* phys. block for the closest logical allocated block to the right */
143         ext4_fsblk_t pright;
144         /* flags. see above EXT4_MB_HINT_* */
145         unsigned int flags;
146 };
147
148 /*
149  * Logical to physical block mapping, used by ext4_map_blocks()
150  *
151  * This structure is used to pass requests into ext4_map_blocks() as
152  * well as to store the information returned by ext4_map_blocks().  It
153  * takes less room on the stack than a struct buffer_head.
154  */
155 #define EXT4_MAP_NEW            (1 << BH_New)
156 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
157 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
158 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
159 #define EXT4_MAP_UNINIT         (1 << BH_Uninit)
160 /* Sometimes (in the bigalloc case, from ext4_da_get_block_prep) the caller of
161  * ext4_map_blocks wants to know whether or not the underlying cluster has
162  * already been accounted for. EXT4_MAP_FROM_CLUSTER conveys to the caller that
163  * the requested mapping was from previously mapped (or delayed allocated)
164  * cluster. We use BH_AllocFromCluster only for this flag. BH_AllocFromCluster
165  * should never appear on buffer_head's state flags.
166  */
167 #define EXT4_MAP_FROM_CLUSTER   (1 << BH_AllocFromCluster)
168 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
169                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
170                                  EXT4_MAP_UNINIT | EXT4_MAP_FROM_CLUSTER)
171
172 struct ext4_map_blocks {
173         ext4_fsblk_t m_pblk;
174         ext4_lblk_t m_lblk;
175         unsigned int m_len;
176         unsigned int m_flags;
177 };
178
179 /*
180  * Flags for ext4_io_end->flags
181  */
182 #define EXT4_IO_END_UNWRITTEN   0x0001
183
184 /*
185  * For converting uninitialized extents on a work queue. 'handle' is used for
186  * buffered writeback.
187  */
188 typedef struct ext4_io_end {
189         struct list_head        list;           /* per-file finished IO list */
190         handle_t                *handle;        /* handle reserved for extent
191                                                  * conversion */
192         struct inode            *inode;         /* file being written to */
193         struct bio              *bio;           /* Linked list of completed
194                                                  * bios covering the extent */
195         unsigned int            flag;           /* unwritten or not */
196         loff_t                  offset;         /* offset in the file */
197         ssize_t                 size;           /* size of the extent */
198         atomic_t                count;          /* reference counter */
199 } ext4_io_end_t;
200
201 struct ext4_io_submit {
202         int                     io_op;
203         struct bio              *io_bio;
204         ext4_io_end_t           *io_end;
205         sector_t                io_next_block;
206 };
207
208 /*
209  * Special inodes numbers
210  */
211 #define EXT4_BAD_INO             1      /* Bad blocks inode */
212 #define EXT4_ROOT_INO            2      /* Root inode */
213 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
214 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
215 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
216 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
217 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
218 #define EXT4_JOURNAL_INO         8      /* Journal inode */
219
220 /* First non-reserved inode for old ext4 filesystems */
221 #define EXT4_GOOD_OLD_FIRST_INO 11
222
223 /*
224  * Maximal count of links to a file
225  */
226 #define EXT4_LINK_MAX           65000
227
228 /*
229  * Macro-instructions used to manage several block sizes
230  */
231 #define EXT4_MIN_BLOCK_SIZE             1024
232 #define EXT4_MAX_BLOCK_SIZE             65536
233 #define EXT4_MIN_BLOCK_LOG_SIZE         10
234 #define EXT4_MAX_BLOCK_LOG_SIZE         16
235 #ifdef __KERNEL__
236 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
237 #else
238 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
239 #endif
240 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
241 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
242                                          EXT4_SB(s)->s_cluster_bits)
243 #ifdef __KERNEL__
244 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
245 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
246 #else
247 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
248 #endif
249 #ifdef __KERNEL__
250 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
251 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
252 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
253 #else
254 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
255                                  EXT4_GOOD_OLD_INODE_SIZE : \
256                                  (s)->s_inode_size)
257 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
258                                  EXT4_GOOD_OLD_FIRST_INO : \
259                                  (s)->s_first_ino)
260 #endif
261 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
262
263 /* Translate a block number to a cluster number */
264 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
265 /* Translate a cluster number to a block number */
266 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
267 /* Translate # of blks to # of clusters */
268 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
269                                  (sbi)->s_cluster_bits)
270
271 /*
272  * Structure of a blocks group descriptor
273  */
274 struct ext4_group_desc
275 {
276         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
277         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
278         __le32  bg_inode_table_lo;      /* Inodes table block */
279         __le16  bg_free_blocks_count_lo;/* Free blocks count */
280         __le16  bg_free_inodes_count_lo;/* Free inodes count */
281         __le16  bg_used_dirs_count_lo;  /* Directories count */
282         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
283         __le32  bg_exclude_bitmap_lo;   /* Exclude bitmap for snapshots */
284         __le16  bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
285         __le16  bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
286         __le16  bg_itable_unused_lo;    /* Unused inodes count */
287         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
288         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
289         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
290         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
291         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
292         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
293         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
294         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
295         __le32  bg_exclude_bitmap_hi;   /* Exclude bitmap block MSB */
296         __le16  bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
297         __le16  bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
298         __u32   bg_reserved;
299 };
300
301 #define EXT4_BG_INODE_BITMAP_CSUM_HI_END        \
302         (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
303          sizeof(__le16))
304 #define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END        \
305         (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
306          sizeof(__le16))
307
308 /*
309  * Structure of a flex block group info
310  */
311
312 struct flex_groups {
313         atomic64_t      free_clusters;
314         atomic_t        free_inodes;
315         atomic_t        used_dirs;
316 };
317
318 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
319 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
320 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
321
322 /*
323  * Macro-instructions used to manage group descriptors
324  */
325 #define EXT4_MIN_DESC_SIZE              32
326 #define EXT4_MIN_DESC_SIZE_64BIT        64
327 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
328 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
329 #ifdef __KERNEL__
330 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
331 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
332 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
333 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
334 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
335 #else
336 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
337 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
338 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
339 #endif
340
341 /*
342  * Constants relative to the data blocks
343  */
344 #define EXT4_NDIR_BLOCKS                12
345 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
346 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
347 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
348 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
349
350 /*
351  * Inode flags
352  */
353 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
354 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
355 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
356 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
357 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
358 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
359 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
360 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
361 /* Reserved for compression usage... */
362 #define EXT4_DIRTY_FL                   0x00000100
363 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
364 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
365 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
366 /* End compression flags --- maybe not all used */
367 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
368 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
369 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
370 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
371 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
372 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
373 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
374 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
375 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
376 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
377 #define EXT4_INLINE_DATA_FL             0x10000000 /* Inode has inline data. */
378 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
379
380 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
381 #define EXT4_FL_USER_MODIFIABLE         0x004380FF /* User modifiable flags */
382
383 /* Flags that should be inherited by new inodes from their parent. */
384 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
385                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
386                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
387                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
388
389 /* Flags that are appropriate for regular files (all but dir-specific ones). */
390 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
391
392 /* Flags that are appropriate for non-directories/regular files. */
393 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
394
395 /* Mask out flags that are inappropriate for the given type of inode. */
396 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
397 {
398         if (S_ISDIR(mode))
399                 return flags;
400         else if (S_ISREG(mode))
401                 return flags & EXT4_REG_FLMASK;
402         else
403                 return flags & EXT4_OTHER_FLMASK;
404 }
405
406 /*
407  * Inode flags used for atomic set/get
408  */
409 enum {
410         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
411         EXT4_INODE_UNRM         = 1,    /* Undelete */
412         EXT4_INODE_COMPR        = 2,    /* Compress file */
413         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
414         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
415         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
416         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
417         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
418 /* Reserved for compression usage... */
419         EXT4_INODE_DIRTY        = 8,
420         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
421         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
422         EXT4_INODE_ECOMPR       = 11,   /* Compression error */
423 /* End compression flags --- maybe not all used */
424         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
425         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
426         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
427         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
428         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
429         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
430         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
431         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
432         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
433         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
434         EXT4_INODE_INLINE_DATA  = 28,   /* Data in inode. */
435         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
436 };
437
438 /*
439  * Since it's pretty easy to mix up bit numbers and hex values, we use a
440  * build-time check to make sure that EXT4_XXX_FL is consistent with respect to
441  * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost
442  * any extra space in the compiled kernel image, otherwise, the build will fail.
443  * It's important that these values are the same, since we are using
444  * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent
445  * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk
446  * values found in ext2, ext3 and ext4 filesystems, and of course the values
447  * defined in e2fsprogs.
448  *
449  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
450  */
451 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
452 #define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG))
453
454 static inline void ext4_check_flag_values(void)
455 {
456         CHECK_FLAG_VALUE(SECRM);
457         CHECK_FLAG_VALUE(UNRM);
458         CHECK_FLAG_VALUE(COMPR);
459         CHECK_FLAG_VALUE(SYNC);
460         CHECK_FLAG_VALUE(IMMUTABLE);
461         CHECK_FLAG_VALUE(APPEND);
462         CHECK_FLAG_VALUE(NODUMP);
463         CHECK_FLAG_VALUE(NOATIME);
464         CHECK_FLAG_VALUE(DIRTY);
465         CHECK_FLAG_VALUE(COMPRBLK);
466         CHECK_FLAG_VALUE(NOCOMPR);
467         CHECK_FLAG_VALUE(ECOMPR);
468         CHECK_FLAG_VALUE(INDEX);
469         CHECK_FLAG_VALUE(IMAGIC);
470         CHECK_FLAG_VALUE(JOURNAL_DATA);
471         CHECK_FLAG_VALUE(NOTAIL);
472         CHECK_FLAG_VALUE(DIRSYNC);
473         CHECK_FLAG_VALUE(TOPDIR);
474         CHECK_FLAG_VALUE(HUGE_FILE);
475         CHECK_FLAG_VALUE(EXTENTS);
476         CHECK_FLAG_VALUE(EA_INODE);
477         CHECK_FLAG_VALUE(EOFBLOCKS);
478         CHECK_FLAG_VALUE(INLINE_DATA);
479         CHECK_FLAG_VALUE(RESERVED);
480 }
481
482 /* Used to pass group descriptor data when online resize is done */
483 struct ext4_new_group_input {
484         __u32 group;            /* Group number for this data */
485         __u64 block_bitmap;     /* Absolute block number of block bitmap */
486         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
487         __u64 inode_table;      /* Absolute block number of inode table start */
488         __u32 blocks_count;     /* Total number of blocks in this group */
489         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
490         __u16 unused;
491 };
492
493 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
494 struct compat_ext4_new_group_input {
495         u32 group;
496         compat_u64 block_bitmap;
497         compat_u64 inode_bitmap;
498         compat_u64 inode_table;
499         u32 blocks_count;
500         u16 reserved_blocks;
501         u16 unused;
502 };
503 #endif
504
505 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
506 struct ext4_new_group_data {
507         __u32 group;
508         __u64 block_bitmap;
509         __u64 inode_bitmap;
510         __u64 inode_table;
511         __u32 blocks_count;
512         __u16 reserved_blocks;
513         __u16 unused;
514         __u32 free_blocks_count;
515 };
516
517 /* Indexes used to index group tables in ext4_new_group_data */
518 enum {
519         BLOCK_BITMAP = 0,       /* block bitmap */
520         INODE_BITMAP,           /* inode bitmap */
521         INODE_TABLE,            /* inode tables */
522         GROUP_TABLE_COUNT,
523 };
524
525 /*
526  * Flags used by ext4_map_blocks()
527  */
528         /* Allocate any needed blocks and/or convert an unitialized
529            extent to be an initialized ext4 */
530 #define EXT4_GET_BLOCKS_CREATE                  0x0001
531         /* Request the creation of an unitialized extent */
532 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
533 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
534                                                  EXT4_GET_BLOCKS_CREATE)
535         /* Caller is from the delayed allocation writeout path
536          * finally doing the actual allocation of delayed blocks */
537 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
538         /* caller is from the direct IO path, request to creation of an
539         unitialized extents if not allocated, split the uninitialized
540         extent if blocks has been preallocated already*/
541 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
542 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
543 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
544                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
545         /* Convert extent to initialized after IO complete */
546 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
547                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
548         /* Eventual metadata allocation (due to growing extent tree)
549          * should not fail, so try to use reserved blocks for that.*/
550 #define EXT4_GET_BLOCKS_METADATA_NOFAIL         0x0020
551         /* Don't normalize allocation size (used for fallocate) */
552 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
553         /* Request will not result in inode size update (user for fallocate) */
554 #define EXT4_GET_BLOCKS_KEEP_SIZE               0x0080
555         /* Do not take i_data_sem locking in ext4_map_blocks */
556 #define EXT4_GET_BLOCKS_NO_LOCK                 0x0100
557         /* Do not put hole in extent cache */
558 #define EXT4_GET_BLOCKS_NO_PUT_HOLE             0x0200
559
560 /*
561  * Flags used by ext4_free_blocks
562  */
563 #define EXT4_FREE_BLOCKS_METADATA       0x0001
564 #define EXT4_FREE_BLOCKS_FORGET         0x0002
565 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
566 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
567 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
568 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
569
570 /*
571  * ioctl commands
572  */
573 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
574 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
575 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
576 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
577 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
578 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
579 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
580 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
581 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
582 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
583 #define EXT4_IOC_MIGRATE                _IO('f', 9)
584  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
585  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
586 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
587 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
588 #define EXT4_IOC_RESIZE_FS              _IOW('f', 16, __u64)
589 #define EXT4_IOC_SWAP_BOOT              _IO('f', 17)
590
591 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
592 /*
593  * ioctl commands in 32 bit emulation
594  */
595 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
596 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
597 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
598 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
599 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
600 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
601 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
602 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
603 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
604 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
605 #endif
606
607 /* Max physical block we can address w/o extents */
608 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
609
610 /*
611  * Structure of an inode on the disk
612  */
613 struct ext4_inode {
614         __le16  i_mode;         /* File mode */
615         __le16  i_uid;          /* Low 16 bits of Owner Uid */
616         __le32  i_size_lo;      /* Size in bytes */
617         __le32  i_atime;        /* Access time */
618         __le32  i_ctime;        /* Inode Change time */
619         __le32  i_mtime;        /* Modification time */
620         __le32  i_dtime;        /* Deletion Time */
621         __le16  i_gid;          /* Low 16 bits of Group Id */
622         __le16  i_links_count;  /* Links count */
623         __le32  i_blocks_lo;    /* Blocks count */
624         __le32  i_flags;        /* File flags */
625         union {
626                 struct {
627                         __le32  l_i_version;
628                 } linux1;
629                 struct {
630                         __u32  h_i_translator;
631                 } hurd1;
632                 struct {
633                         __u32  m_i_reserved1;
634                 } masix1;
635         } osd1;                         /* OS dependent 1 */
636         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
637         __le32  i_generation;   /* File version (for NFS) */
638         __le32  i_file_acl_lo;  /* File ACL */
639         __le32  i_size_high;
640         __le32  i_obso_faddr;   /* Obsoleted fragment address */
641         union {
642                 struct {
643                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
644                         __le16  l_i_file_acl_high;
645                         __le16  l_i_uid_high;   /* these 2 fields */
646                         __le16  l_i_gid_high;   /* were reserved2[0] */
647                         __le16  l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
648                         __le16  l_i_reserved;
649                 } linux2;
650                 struct {
651                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
652                         __u16   h_i_mode_high;
653                         __u16   h_i_uid_high;
654                         __u16   h_i_gid_high;
655                         __u32   h_i_author;
656                 } hurd2;
657                 struct {
658                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
659                         __le16  m_i_file_acl_high;
660                         __u32   m_i_reserved2[2];
661                 } masix2;
662         } osd2;                         /* OS dependent 2 */
663         __le16  i_extra_isize;
664         __le16  i_checksum_hi;  /* crc32c(uuid+inum+inode) BE */
665         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
666         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
667         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
668         __le32  i_crtime;       /* File Creation time */
669         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
670         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
671 };
672
673 struct move_extent {
674         __u32 reserved;         /* should be zero */
675         __u32 donor_fd;         /* donor file descriptor */
676         __u64 orig_start;       /* logical start offset in block for orig */
677         __u64 donor_start;      /* logical start offset in block for donor */
678         __u64 len;              /* block length to be moved */
679         __u64 moved_len;        /* moved block length */
680 };
681
682 #define EXT4_EPOCH_BITS 2
683 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
684 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
685
686 /*
687  * Extended fields will fit into an inode if the filesystem was formatted
688  * with large inodes (-I 256 or larger) and there are not currently any EAs
689  * consuming all of the available space. For new inodes we always reserve
690  * enough space for the kernel's known extended fields, but for inodes
691  * created with an old kernel this might not have been the case. None of
692  * the extended inode fields is critical for correct filesystem operation.
693  * This macro checks if a certain field fits in the inode. Note that
694  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
695  */
696 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
697         ((offsetof(typeof(*ext4_inode), field) +        \
698           sizeof((ext4_inode)->field))                  \
699         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
700             (einode)->i_extra_isize))                   \
701
702 static inline __le32 ext4_encode_extra_time(struct timespec *time)
703 {
704        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
705                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
706                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
707 }
708
709 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
710 {
711        if (sizeof(time->tv_sec) > 4)
712                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
713                                << 32;
714        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
715 }
716
717 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
718 do {                                                                           \
719         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
720         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
721                 (raw_inode)->xtime ## _extra =                                 \
722                                 ext4_encode_extra_time(&(inode)->xtime);       \
723 } while (0)
724
725 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
726 do {                                                                           \
727         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
728                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
729         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
730                 (raw_inode)->xtime ## _extra =                                 \
731                                 ext4_encode_extra_time(&(einode)->xtime);      \
732 } while (0)
733
734 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
735 do {                                                                           \
736         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
737         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
738                 ext4_decode_extra_time(&(inode)->xtime,                        \
739                                        raw_inode->xtime ## _extra);            \
740         else                                                                   \
741                 (inode)->xtime.tv_nsec = 0;                                    \
742 } while (0)
743
744 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
745 do {                                                                           \
746         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
747                 (einode)->xtime.tv_sec =                                       \
748                         (signed)le32_to_cpu((raw_inode)->xtime);               \
749         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
750                 ext4_decode_extra_time(&(einode)->xtime,                       \
751                                        raw_inode->xtime ## _extra);            \
752         else                                                                   \
753                 (einode)->xtime.tv_nsec = 0;                                   \
754 } while (0)
755
756 #define i_disk_version osd1.linux1.l_i_version
757
758 #if defined(__KERNEL__) || defined(__linux__)
759 #define i_reserved1     osd1.linux1.l_i_reserved1
760 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
761 #define i_blocks_high   osd2.linux2.l_i_blocks_high
762 #define i_uid_low       i_uid
763 #define i_gid_low       i_gid
764 #define i_uid_high      osd2.linux2.l_i_uid_high
765 #define i_gid_high      osd2.linux2.l_i_gid_high
766 #define i_checksum_lo   osd2.linux2.l_i_checksum_lo
767
768 #elif defined(__GNU__)
769
770 #define i_translator    osd1.hurd1.h_i_translator
771 #define i_uid_high      osd2.hurd2.h_i_uid_high
772 #define i_gid_high      osd2.hurd2.h_i_gid_high
773 #define i_author        osd2.hurd2.h_i_author
774
775 #elif defined(__masix__)
776
777 #define i_reserved1     osd1.masix1.m_i_reserved1
778 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
779 #define i_reserved2     osd2.masix2.m_i_reserved2
780
781 #endif /* defined(__KERNEL__) || defined(__linux__) */
782
783 #include "extents_status.h"
784
785 /*
786  * fourth extended file system inode data in memory
787  */
788 struct ext4_inode_info {
789         __le32  i_data[15];     /* unconverted */
790         __u32   i_dtime;
791         ext4_fsblk_t    i_file_acl;
792
793         /*
794          * i_block_group is the number of the block group which contains
795          * this file's inode.  Constant across the lifetime of the inode,
796          * it is ued for making block allocation decisions - we try to
797          * place a file's data blocks near its inode block, and new inodes
798          * near to their parent directory's inode.
799          */
800         ext4_group_t    i_block_group;
801         ext4_lblk_t     i_dir_start_lookup;
802 #if (BITS_PER_LONG < 64)
803         unsigned long   i_state_flags;          /* Dynamic state flags */
804 #endif
805         unsigned long   i_flags;
806
807         /*
808          * Extended attributes can be read independently of the main file
809          * data. Taking i_mutex even when reading would cause contention
810          * between readers of EAs and writers of regular file data, so
811          * instead we synchronize on xattr_sem when reading or changing
812          * EAs.
813          */
814         struct rw_semaphore xattr_sem;
815
816         struct list_head i_orphan;      /* unlinked but open inodes */
817
818         /*
819          * i_disksize keeps track of what the inode size is ON DISK, not
820          * in memory.  During truncate, i_size is set to the new size by
821          * the VFS prior to calling ext4_truncate(), but the filesystem won't
822          * set i_disksize to 0 until the truncate is actually under way.
823          *
824          * The intent is that i_disksize always represents the blocks which
825          * are used by this file.  This allows recovery to restart truncate
826          * on orphans if we crash during truncate.  We actually write i_disksize
827          * into the on-disk inode when writing inodes out, instead of i_size.
828          *
829          * The only time when i_disksize and i_size may be different is when
830          * a truncate is in progress.  The only things which change i_disksize
831          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
832          */
833         loff_t  i_disksize;
834
835         /*
836          * i_data_sem is for serialising ext4_truncate() against
837          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
838          * data tree are chopped off during truncate. We can't do that in
839          * ext4 because whenever we perform intermediate commits during
840          * truncate, the inode and all the metadata blocks *must* be in a
841          * consistent state which allows truncation of the orphans to restart
842          * during recovery.  Hence we must fix the get_block-vs-truncate race
843          * by other means, so we have i_data_sem.
844          */
845         struct rw_semaphore i_data_sem;
846         struct inode vfs_inode;
847         struct jbd2_inode *jinode;
848
849         /*
850          * File creation time. Its function is same as that of
851          * struct timespec i_{a,c,m}time in the generic inode.
852          */
853         struct timespec i_crtime;
854
855         /* mballoc */
856         struct list_head i_prealloc_list;
857         spinlock_t i_prealloc_lock;
858
859         /* extents status tree */
860         struct ext4_es_tree i_es_tree;
861         rwlock_t i_es_lock;
862         struct list_head i_es_lru;
863         unsigned int i_es_lru_nr;       /* protected by i_es_lock */
864         unsigned long i_touch_when;     /* jiffies of last accessing */
865
866         /* ialloc */
867         ext4_group_t    i_last_alloc_group;
868
869         /* allocation reservation info for delalloc */
870         /* In case of bigalloc, these refer to clusters rather than blocks */
871         unsigned int i_reserved_data_blocks;
872         unsigned int i_reserved_meta_blocks;
873         unsigned int i_allocated_meta_blocks;
874         ext4_lblk_t i_da_metadata_calc_last_lblock;
875         int i_da_metadata_calc_len;
876
877         /* on-disk additional length */
878         __u16 i_extra_isize;
879
880         /* Indicate the inline data space. */
881         u16 i_inline_off;
882         u16 i_inline_size;
883
884 #ifdef CONFIG_QUOTA
885         /* quota space reservation, managed internally by quota code */
886         qsize_t i_reserved_quota;
887 #endif
888
889         /* Lock protecting lists below */
890         spinlock_t i_completed_io_lock;
891         /*
892          * Completed IOs that need unwritten extents handling and have
893          * transaction reserved
894          */
895         struct list_head i_rsv_conversion_list;
896         /*
897          * Completed IOs that need unwritten extents handling and don't have
898          * transaction reserved
899          */
900         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
901         atomic_t i_unwritten; /* Nr. of inflight conversions pending */
902         struct work_struct i_rsv_conversion_work;
903
904         spinlock_t i_block_reservation_lock;
905
906         /*
907          * Transactions that contain inode's metadata needed to complete
908          * fsync and fdatasync, respectively.
909          */
910         tid_t i_sync_tid;
911         tid_t i_datasync_tid;
912
913         /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
914         __u32 i_csum_seed;
915 };
916
917 /*
918  * File system states
919  */
920 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
921 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
922 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
923
924 /*
925  * Misc. filesystem flags
926  */
927 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
928 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
929 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
930
931 /*
932  * Mount flags set via mount options or defaults
933  */
934 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
935 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
936 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
937 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
938 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
939 #define EXT4_MOUNT_ERRORS_MASK          0x00070
940 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
941 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
942 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
943 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
944 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
945 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
946 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
947 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
948 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
949 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
950 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
951 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
952 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
953 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
954 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
955 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
956 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
957 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
958 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
959 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
960 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
961 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
962 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
963
964 /*
965  * Mount flags set either automatically (could not be set by mount option)
966  * based on per file system feature or property or in special cases such as
967  * distinguishing between explicit mount option definition and default.
968  */
969 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
970                                                       specified delalloc */
971 #define EXT4_MOUNT2_STD_GROUP_SIZE      0x00000002 /* We have standard group
972                                                       size of blocksize * 8
973                                                       blocks */
974
975 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
976                                                 ~EXT4_MOUNT_##opt
977 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
978                                                 EXT4_MOUNT_##opt
979 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
980                                          EXT4_MOUNT_##opt)
981
982 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
983                                                 ~EXT4_MOUNT2_##opt
984 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
985                                                 EXT4_MOUNT2_##opt
986 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
987                                          EXT4_MOUNT2_##opt)
988
989 #define ext4_test_and_set_bit           __test_and_set_bit_le
990 #define ext4_set_bit                    __set_bit_le
991 #define ext4_set_bit_atomic             ext2_set_bit_atomic
992 #define ext4_test_and_clear_bit         __test_and_clear_bit_le
993 #define ext4_clear_bit                  __clear_bit_le
994 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
995 #define ext4_test_bit                   test_bit_le
996 #define ext4_find_next_zero_bit         find_next_zero_bit_le
997 #define ext4_find_next_bit              find_next_bit_le
998
999 extern void ext4_set_bits(void *bm, int cur, int len);
1000
1001 /*
1002  * Maximal mount counts between two filesystem checks
1003  */
1004 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
1005 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
1006
1007 /*
1008  * Behaviour when detecting errors
1009  */
1010 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
1011 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
1012 #define EXT4_ERRORS_PANIC               3       /* Panic */
1013 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
1014
1015 /* Metadata checksum algorithm codes */
1016 #define EXT4_CRC32C_CHKSUM              1
1017
1018 /*
1019  * Structure of the super block
1020  */
1021 struct ext4_super_block {
1022 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1023         __le32  s_blocks_count_lo;      /* Blocks count */
1024         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1025         __le32  s_free_blocks_count_lo; /* Free blocks count */
1026 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1027         __le32  s_first_data_block;     /* First Data Block */
1028         __le32  s_log_block_size;       /* Block size */
1029         __le32  s_log_cluster_size;     /* Allocation cluster size */
1030 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1031         __le32  s_clusters_per_group;   /* # Clusters per group */
1032         __le32  s_inodes_per_group;     /* # Inodes per group */
1033         __le32  s_mtime;                /* Mount time */
1034 /*30*/  __le32  s_wtime;                /* Write time */
1035         __le16  s_mnt_count;            /* Mount count */
1036         __le16  s_max_mnt_count;        /* Maximal mount count */
1037         __le16  s_magic;                /* Magic signature */
1038         __le16  s_state;                /* File system state */
1039         __le16  s_errors;               /* Behaviour when detecting errors */
1040         __le16  s_minor_rev_level;      /* minor revision level */
1041 /*40*/  __le32  s_lastcheck;            /* time of last check */
1042         __le32  s_checkinterval;        /* max. time between checks */
1043         __le32  s_creator_os;           /* OS */
1044         __le32  s_rev_level;            /* Revision level */
1045 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1046         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1047         /*
1048          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1049          *
1050          * Note: the difference between the compatible feature set and
1051          * the incompatible feature set is that if there is a bit set
1052          * in the incompatible feature set that the kernel doesn't
1053          * know about, it should refuse to mount the filesystem.
1054          *
1055          * e2fsck's requirements are more strict; if it doesn't know
1056          * about a feature in either the compatible or incompatible
1057          * feature set, it must abort and not try to meddle with
1058          * things it doesn't understand...
1059          */
1060         __le32  s_first_ino;            /* First non-reserved inode */
1061         __le16  s_inode_size;           /* size of inode structure */
1062         __le16  s_block_group_nr;       /* block group # of this superblock */
1063         __le32  s_feature_compat;       /* compatible feature set */
1064 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1065         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1066 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1067 /*78*/  char    s_volume_name[16];      /* volume name */
1068 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1069 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1070         /*
1071          * Performance hints.  Directory preallocation should only
1072          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1073          */
1074         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1075         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1076         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1077         /*
1078          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1079          */
1080 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1081 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1082         __le32  s_journal_dev;          /* device number of journal file */
1083         __le32  s_last_orphan;          /* start of list of inodes to delete */
1084         __le32  s_hash_seed[4];         /* HTREE hash seed */
1085         __u8    s_def_hash_version;     /* Default hash version to use */
1086         __u8    s_jnl_backup_type;
1087         __le16  s_desc_size;            /* size of group descriptor */
1088 /*100*/ __le32  s_default_mount_opts;
1089         __le32  s_first_meta_bg;        /* First metablock block group */
1090         __le32  s_mkfs_time;            /* When the filesystem was created */
1091         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1092         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1093 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1094         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1095         __le32  s_free_blocks_count_hi; /* Free blocks count */
1096         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1097         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1098         __le32  s_flags;                /* Miscellaneous flags */
1099         __le16  s_raid_stride;          /* RAID stride */
1100         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1101         __le64  s_mmp_block;            /* Block for multi-mount protection */
1102         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1103         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1104         __u8    s_checksum_type;        /* metadata checksum algorithm used */
1105         __le16  s_reserved_pad;
1106         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1107         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1108         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1109         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1110                                               snapshot's future use */
1111         __le32  s_snapshot_list;        /* inode number of the head of the
1112                                            on-disk snapshot list */
1113 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1114         __le32  s_error_count;          /* number of fs errors */
1115         __le32  s_first_error_time;     /* first time an error happened */
1116         __le32  s_first_error_ino;      /* inode involved in first error */
1117         __le64  s_first_error_block;    /* block involved of first error */
1118         __u8    s_first_error_func[32]; /* function where the error happened */
1119         __le32  s_first_error_line;     /* line number where error happened */
1120         __le32  s_last_error_time;      /* most recent time of an error */
1121         __le32  s_last_error_ino;       /* inode involved in last error */
1122         __le32  s_last_error_line;      /* line number where error happened */
1123         __le64  s_last_error_block;     /* block involved of last error */
1124         __u8    s_last_error_func[32];  /* function where the error happened */
1125 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1126         __u8    s_mount_opts[64];
1127         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1128         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1129         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1130         __le32  s_reserved[108];        /* Padding to the end of the block */
1131         __le32  s_checksum;             /* crc32c(superblock) */
1132 };
1133
1134 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1135
1136 #ifdef __KERNEL__
1137
1138 /*
1139  * run-time mount flags
1140  */
1141 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
1142 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
1143
1144 /*
1145  * fourth extended-fs super-block data in memory
1146  */
1147 struct ext4_sb_info {
1148         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1149         unsigned long s_inodes_per_block;/* Number of inodes per block */
1150         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1151         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1152         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1153         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1154         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1155         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1156         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1157         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1158         unsigned long s_overhead;  /* # of fs overhead clusters */
1159         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1160         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1161         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1162         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1163         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1164         struct buffer_head **s_group_desc;
1165         unsigned int s_mount_opt;
1166         unsigned int s_mount_opt2;
1167         unsigned int s_mount_flags;
1168         unsigned int s_def_mount_opt;
1169         ext4_fsblk_t s_sb_block;
1170         atomic64_t s_resv_clusters;
1171         kuid_t s_resuid;
1172         kgid_t s_resgid;
1173         unsigned short s_mount_state;
1174         unsigned short s_pad;
1175         int s_addr_per_block_bits;
1176         int s_desc_per_block_bits;
1177         int s_inode_size;
1178         int s_first_ino;
1179         unsigned int s_inode_readahead_blks;
1180         unsigned int s_inode_goal;
1181         spinlock_t s_next_gen_lock;
1182         u32 s_next_generation;
1183         u32 s_hash_seed[4];
1184         int s_def_hash_version;
1185         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1186         struct percpu_counter s_freeclusters_counter;
1187         struct percpu_counter s_freeinodes_counter;
1188         struct percpu_counter s_dirs_counter;
1189         struct percpu_counter s_dirtyclusters_counter;
1190         struct blockgroup_lock *s_blockgroup_lock;
1191         struct proc_dir_entry *s_proc;
1192         struct kobject s_kobj;
1193         struct completion s_kobj_unregister;
1194         struct super_block *s_sb;
1195
1196         /* Journaling */
1197         struct journal_s *s_journal;
1198         struct list_head s_orphan;
1199         struct mutex s_orphan_lock;
1200         unsigned long s_resize_flags;           /* Flags indicating if there
1201                                                    is a resizer */
1202         unsigned long s_commit_interval;
1203         u32 s_max_batch_time;
1204         u32 s_min_batch_time;
1205         struct block_device *journal_bdev;
1206 #ifdef CONFIG_QUOTA
1207         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
1208         int s_jquota_fmt;                       /* Format of quota to use */
1209 #endif
1210         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1211         struct rb_root system_blks;
1212
1213 #ifdef EXTENTS_STATS
1214         /* ext4 extents stats */
1215         unsigned long s_ext_min;
1216         unsigned long s_ext_max;
1217         unsigned long s_depth_max;
1218         spinlock_t s_ext_stats_lock;
1219         unsigned long s_ext_blocks;
1220         unsigned long s_ext_extents;
1221 #endif
1222
1223         /* for buddy allocator */
1224         struct ext4_group_info ***s_group_info;
1225         struct inode *s_buddy_cache;
1226         spinlock_t s_md_lock;
1227         unsigned short *s_mb_offsets;
1228         unsigned int *s_mb_maxs;
1229         unsigned int s_group_info_size;
1230
1231         /* tunables */
1232         unsigned long s_stripe;
1233         unsigned int s_mb_stream_request;
1234         unsigned int s_mb_max_to_scan;
1235         unsigned int s_mb_min_to_scan;
1236         unsigned int s_mb_stats;
1237         unsigned int s_mb_order2_reqs;
1238         unsigned int s_mb_group_prealloc;
1239         unsigned int s_max_dir_size_kb;
1240         /* where last allocation was done - for stream allocation */
1241         unsigned long s_mb_last_group;
1242         unsigned long s_mb_last_start;
1243
1244         /* stats for buddy allocator */
1245         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1246         atomic_t s_bal_success; /* we found long enough chunks */
1247         atomic_t s_bal_allocated;       /* in blocks */
1248         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1249         atomic_t s_bal_goals;   /* goal hits */
1250         atomic_t s_bal_breaks;  /* too long searches */
1251         atomic_t s_bal_2orders; /* 2^order hits */
1252         spinlock_t s_bal_lock;
1253         unsigned long s_mb_buddies_generated;
1254         unsigned long long s_mb_generation_time;
1255         atomic_t s_mb_lost_chunks;
1256         atomic_t s_mb_preallocated;
1257         atomic_t s_mb_discarded;
1258         atomic_t s_lock_busy;
1259
1260         /* locality groups */
1261         struct ext4_locality_group __percpu *s_locality_groups;
1262
1263         /* for write statistics */
1264         unsigned long s_sectors_written_start;
1265         u64 s_kbytes_written;
1266
1267         /* the size of zero-out chunk */
1268         unsigned int s_extent_max_zeroout_kb;
1269
1270         unsigned int s_log_groups_per_flex;
1271         struct flex_groups *s_flex_groups;
1272         ext4_group_t s_flex_groups_allocated;
1273
1274         /* workqueue for reserved extent conversions (buffered io) */
1275         struct workqueue_struct *rsv_conversion_wq;
1276
1277         /* timer for periodic error stats printing */
1278         struct timer_list s_err_report;
1279
1280         /* Lazy inode table initialization info */
1281         struct ext4_li_request *s_li_request;
1282         /* Wait multiplier for lazy initialization thread */
1283         unsigned int s_li_wait_mult;
1284
1285         /* Kernel thread for multiple mount protection */
1286         struct task_struct *s_mmp_tsk;
1287
1288         /* record the last minlen when FITRIM is called. */
1289         atomic_t s_last_trim_minblks;
1290
1291         /* Reference to checksum algorithm driver via cryptoapi */
1292         struct crypto_shash *s_chksum_driver;
1293
1294         /* Precomputed FS UUID checksum for seeding other checksums */
1295         __u32 s_csum_seed;
1296
1297         /* Reclaim extents from extent status tree */
1298         struct shrinker s_es_shrinker;
1299         struct list_head s_es_lru;
1300         unsigned long s_es_last_sorted;
1301         struct percpu_counter s_extent_cache_cnt;
1302         spinlock_t s_es_lru_lock ____cacheline_aligned_in_smp;
1303 };
1304
1305 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1306 {
1307         return sb->s_fs_info;
1308 }
1309 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1310 {
1311         return container_of(inode, struct ext4_inode_info, vfs_inode);
1312 }
1313
1314 static inline struct timespec ext4_current_time(struct inode *inode)
1315 {
1316         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1317                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1318 }
1319
1320 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1321 {
1322         return ino == EXT4_ROOT_INO ||
1323                 ino == EXT4_USR_QUOTA_INO ||
1324                 ino == EXT4_GRP_QUOTA_INO ||
1325                 ino == EXT4_BOOT_LOADER_INO ||
1326                 ino == EXT4_JOURNAL_INO ||
1327                 ino == EXT4_RESIZE_INO ||
1328                 (ino >= EXT4_FIRST_INO(sb) &&
1329                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1330 }
1331
1332 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
1333                                               struct ext4_io_end *io_end)
1334 {
1335         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
1336                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
1337                 atomic_inc(&EXT4_I(inode)->i_unwritten);
1338         }
1339 }
1340
1341 static inline ext4_io_end_t *ext4_inode_aio(struct inode *inode)
1342 {
1343         return inode->i_private;
1344 }
1345
1346 static inline void ext4_inode_aio_set(struct inode *inode, ext4_io_end_t *io)
1347 {
1348         inode->i_private = io;
1349 }
1350
1351 /*
1352  * Inode dynamic state flags
1353  */
1354 enum {
1355         EXT4_STATE_JDATA,               /* journaled data exists */
1356         EXT4_STATE_NEW,                 /* inode is newly created */
1357         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1358         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1359         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1360         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1361         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1362         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1363         EXT4_STATE_DELALLOC_RESERVED,   /* blks already reserved for delalloc */
1364         EXT4_STATE_DIOREAD_LOCK,        /* Disable support for dio read
1365                                            nolocking */
1366         EXT4_STATE_MAY_INLINE_DATA,     /* may have in-inode data */
1367         EXT4_STATE_ORDERED_MODE,        /* data=ordered mode */
1368 };
1369
1370 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1371 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1372 {                                                                       \
1373         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1374 }                                                                       \
1375 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1376 {                                                                       \
1377         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1378 }                                                                       \
1379 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1380 {                                                                       \
1381         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1382 }
1383
1384 EXT4_INODE_BIT_FNS(flag, flags, 0)
1385 #if (BITS_PER_LONG < 64)
1386 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1387
1388 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1389 {
1390         (ei)->i_state_flags = 0;
1391 }
1392 #else
1393 EXT4_INODE_BIT_FNS(state, flags, 32)
1394
1395 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1396 {
1397         /* We depend on the fact that callers will set i_flags */
1398 }
1399 #endif
1400 #else
1401 /* Assume that user mode programs are passing in an ext4fs superblock, not
1402  * a kernel struct super_block.  This will allow us to call the feature-test
1403  * macros from user land. */
1404 #define EXT4_SB(sb)     (sb)
1405 #endif
1406
1407 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1408
1409 /*
1410  * Codes for operating systems
1411  */
1412 #define EXT4_OS_LINUX           0
1413 #define EXT4_OS_HURD            1
1414 #define EXT4_OS_MASIX           2
1415 #define EXT4_OS_FREEBSD         3
1416 #define EXT4_OS_LITES           4
1417
1418 /*
1419  * Revision levels
1420  */
1421 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1422 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1423
1424 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1425 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1426
1427 #define EXT4_GOOD_OLD_INODE_SIZE 128
1428
1429 /*
1430  * Feature set definitions
1431  */
1432
1433 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1434         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1435 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1436         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1437 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1438         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1439 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1440         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1441 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1442         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1443 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1444         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1445 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1446         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1447 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1448         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1449 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1450         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1451
1452 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1453 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1454 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1455 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1456 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1457 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1458
1459 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1460 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1461 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1462 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1463 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1464 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1465 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1466 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1467 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1468 /*
1469  * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM).  When
1470  * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
1471  * all other data structures' checksums.  However, the METADATA_CSUM and
1472  * GDT_CSUM bits are mutually exclusive.
1473  */
1474 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM    0x0400
1475
1476 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1477 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1478 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1479 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1480 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1481 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1482 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1483 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1484 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1485 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1486 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1487 #define EXT4_FEATURE_INCOMPAT_BG_USE_META_CSUM  0x2000 /* use crc32c for bg */
1488 #define EXT4_FEATURE_INCOMPAT_LARGEDIR          0x4000 /* >2GB or 3-lvl htree */
1489 #define EXT4_FEATURE_INCOMPAT_INLINE_DATA       0x8000 /* data in inode */
1490
1491 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1492 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1493                                          EXT4_FEATURE_INCOMPAT_META_BG)
1494 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1495                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1496                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1497
1498 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1499 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1500                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1501                                          EXT4_FEATURE_INCOMPAT_META_BG)
1502 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1503                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1504                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1505
1506 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1507 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1508                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1509                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1510                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1511                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1512                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1513                                          EXT4_FEATURE_INCOMPAT_MMP |    \
1514                                          EXT4_FEATURE_INCOMPAT_INLINE_DATA)
1515 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1516                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1517                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1518                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1519                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1520                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1521                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1522                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
1523                                          EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
1524                                          EXT4_FEATURE_RO_COMPAT_QUOTA)
1525
1526 /*
1527  * Default values for user and/or group using reserved blocks
1528  */
1529 #define EXT4_DEF_RESUID         0
1530 #define EXT4_DEF_RESGID         0
1531
1532 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1533
1534 /*
1535  * Default mount options
1536  */
1537 #define EXT4_DEFM_DEBUG         0x0001
1538 #define EXT4_DEFM_BSDGROUPS     0x0002
1539 #define EXT4_DEFM_XATTR_USER    0x0004
1540 #define EXT4_DEFM_ACL           0x0008
1541 #define EXT4_DEFM_UID16         0x0010
1542 #define EXT4_DEFM_JMODE         0x0060
1543 #define EXT4_DEFM_JMODE_DATA    0x0020
1544 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1545 #define EXT4_DEFM_JMODE_WBACK   0x0060
1546 #define EXT4_DEFM_NOBARRIER     0x0100
1547 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1548 #define EXT4_DEFM_DISCARD       0x0400
1549 #define EXT4_DEFM_NODELALLOC    0x0800
1550
1551 /*
1552  * Default journal batch times
1553  */
1554 #define EXT4_DEF_MIN_BATCH_TIME 0
1555 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1556
1557 /*
1558  * Minimum number of groups in a flexgroup before we separate out
1559  * directories into the first block group of a flexgroup
1560  */
1561 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1562
1563 /*
1564  * Structure of a directory entry
1565  */
1566 #define EXT4_NAME_LEN 255
1567
1568 struct ext4_dir_entry {
1569         __le32  inode;                  /* Inode number */
1570         __le16  rec_len;                /* Directory entry length */
1571         __le16  name_len;               /* Name length */
1572         char    name[EXT4_NAME_LEN];    /* File name */
1573 };
1574
1575 /*
1576  * The new version of the directory entry.  Since EXT4 structures are
1577  * stored in intel byte order, and the name_len field could never be
1578  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1579  * file_type field.
1580  */
1581 struct ext4_dir_entry_2 {
1582         __le32  inode;                  /* Inode number */
1583         __le16  rec_len;                /* Directory entry length */
1584         __u8    name_len;               /* Name length */
1585         __u8    file_type;
1586         char    name[EXT4_NAME_LEN];    /* File name */
1587 };
1588
1589 /*
1590  * This is a bogus directory entry at the end of each leaf block that
1591  * records checksums.
1592  */
1593 struct ext4_dir_entry_tail {
1594         __le32  det_reserved_zero1;     /* Pretend to be unused */
1595         __le16  det_rec_len;            /* 12 */
1596         __u8    det_reserved_zero2;     /* Zero name length */
1597         __u8    det_reserved_ft;        /* 0xDE, fake file type */
1598         __le32  det_checksum;           /* crc32c(uuid+inum+dirblock) */
1599 };
1600
1601 #define EXT4_DIRENT_TAIL(block, blocksize) \
1602         ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
1603                                         ((blocksize) - \
1604                                          sizeof(struct ext4_dir_entry_tail))))
1605
1606 /*
1607  * Ext4 directory file types.  Only the low 3 bits are used.  The
1608  * other bits are reserved for now.
1609  */
1610 #define EXT4_FT_UNKNOWN         0
1611 #define EXT4_FT_REG_FILE        1
1612 #define EXT4_FT_DIR             2
1613 #define EXT4_FT_CHRDEV          3
1614 #define EXT4_FT_BLKDEV          4
1615 #define EXT4_FT_FIFO            5
1616 #define EXT4_FT_SOCK            6
1617 #define EXT4_FT_SYMLINK         7
1618
1619 #define EXT4_FT_MAX             8
1620
1621 #define EXT4_FT_DIR_CSUM        0xDE
1622
1623 /*
1624  * EXT4_DIR_PAD defines the directory entries boundaries
1625  *
1626  * NOTE: It must be a multiple of 4
1627  */
1628 #define EXT4_DIR_PAD                    4
1629 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1630 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1631                                          ~EXT4_DIR_ROUND)
1632 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1633
1634 /*
1635  * If we ever get support for fs block sizes > page_size, we'll need
1636  * to remove the #if statements in the next two functions...
1637  */
1638 static inline unsigned int
1639 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1640 {
1641         unsigned len = le16_to_cpu(dlen);
1642
1643 #if (PAGE_CACHE_SIZE >= 65536)
1644         if (len == EXT4_MAX_REC_LEN || len == 0)
1645                 return blocksize;
1646         return (len & 65532) | ((len & 3) << 16);
1647 #else
1648         return len;
1649 #endif
1650 }
1651
1652 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1653 {
1654         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1655                 BUG();
1656 #if (PAGE_CACHE_SIZE >= 65536)
1657         if (len < 65536)
1658                 return cpu_to_le16(len);
1659         if (len == blocksize) {
1660                 if (blocksize == 65536)
1661                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1662                 else
1663                         return cpu_to_le16(0);
1664         }
1665         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1666 #else
1667         return cpu_to_le16(len);
1668 #endif
1669 }
1670
1671 /*
1672  * Hash Tree Directory indexing
1673  * (c) Daniel Phillips, 2001
1674  */
1675
1676 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1677                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1678                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1679 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1680 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1681
1682 /* Legal values for the dx_root hash_version field: */
1683
1684 #define DX_HASH_LEGACY          0
1685 #define DX_HASH_HALF_MD4        1
1686 #define DX_HASH_TEA             2
1687 #define DX_HASH_LEGACY_UNSIGNED 3
1688 #define DX_HASH_HALF_MD4_UNSIGNED       4
1689 #define DX_HASH_TEA_UNSIGNED            5
1690
1691 static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
1692                               const void *address, unsigned int length)
1693 {
1694         struct {
1695                 struct shash_desc shash;
1696                 char ctx[4];
1697         } desc;
1698         int err;
1699
1700         BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
1701
1702         desc.shash.tfm = sbi->s_chksum_driver;
1703         desc.shash.flags = 0;
1704         *(u32 *)desc.ctx = crc;
1705
1706         err = crypto_shash_update(&desc.shash, address, length);
1707         BUG_ON(err);
1708
1709         return *(u32 *)desc.ctx;
1710 }
1711
1712 #ifdef __KERNEL__
1713
1714 /* hash info structure used by the directory hash */
1715 struct dx_hash_info
1716 {
1717         u32             hash;
1718         u32             minor_hash;
1719         int             hash_version;
1720         u32             *seed;
1721 };
1722
1723
1724 /* 32 and 64 bit signed EOF for dx directories */
1725 #define EXT4_HTREE_EOF_32BIT   ((1UL  << (32 - 1)) - 1)
1726 #define EXT4_HTREE_EOF_64BIT   ((1ULL << (64 - 1)) - 1)
1727
1728
1729 /*
1730  * Control parameters used by ext4_htree_next_block
1731  */
1732 #define HASH_NB_ALWAYS          1
1733
1734
1735 /*
1736  * Describe an inode's exact location on disk and in memory
1737  */
1738 struct ext4_iloc
1739 {
1740         struct buffer_head *bh;
1741         unsigned long offset;
1742         ext4_group_t block_group;
1743 };
1744
1745 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1746 {
1747         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1748 }
1749
1750 /*
1751  * This structure is stuffed into the struct file's private_data field
1752  * for directories.  It is where we put information so that we can do
1753  * readdir operations in hash tree order.
1754  */
1755 struct dir_private_info {
1756         struct rb_root  root;
1757         struct rb_node  *curr_node;
1758         struct fname    *extra_fname;
1759         loff_t          last_pos;
1760         __u32           curr_hash;
1761         __u32           curr_minor_hash;
1762         __u32           next_hash;
1763 };
1764
1765 /* calculate the first block number of the group */
1766 static inline ext4_fsblk_t
1767 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1768 {
1769         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1770                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1771 }
1772
1773 /*
1774  * Special error return code only used by dx_probe() and its callers.
1775  */
1776 #define ERR_BAD_DX_DIR  -75000
1777
1778 /*
1779  * Timeout and state flag for lazy initialization inode thread.
1780  */
1781 #define EXT4_DEF_LI_WAIT_MULT                   10
1782 #define EXT4_DEF_LI_MAX_START_DELAY             5
1783 #define EXT4_LAZYINIT_QUIT                      0x0001
1784 #define EXT4_LAZYINIT_RUNNING                   0x0002
1785
1786 /*
1787  * Lazy inode table initialization info
1788  */
1789 struct ext4_lazy_init {
1790         unsigned long           li_state;
1791         struct list_head        li_request_list;
1792         struct mutex            li_list_mtx;
1793 };
1794
1795 struct ext4_li_request {
1796         struct super_block      *lr_super;
1797         struct ext4_sb_info     *lr_sbi;
1798         ext4_group_t            lr_next_group;
1799         struct list_head        lr_request;
1800         unsigned long           lr_next_sched;
1801         unsigned long           lr_timeout;
1802 };
1803
1804 struct ext4_features {
1805         struct kobject f_kobj;
1806         struct completion f_kobj_unregister;
1807 };
1808
1809 /*
1810  * This structure will be used for multiple mount protection. It will be
1811  * written into the block number saved in the s_mmp_block field in the
1812  * superblock. Programs that check MMP should assume that if
1813  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1814  * to use the filesystem, regardless of how old the timestamp is.
1815  */
1816 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
1817 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1818 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
1819 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
1820
1821 struct mmp_struct {
1822         __le32  mmp_magic;              /* Magic number for MMP */
1823         __le32  mmp_seq;                /* Sequence no. updated periodically */
1824
1825         /*
1826          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1827          * purposes and do not affect the correctness of the algorithm
1828          */
1829         __le64  mmp_time;               /* Time last updated */
1830         char    mmp_nodename[64];       /* Node which last updated MMP block */
1831         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
1832
1833         /*
1834          * mmp_check_interval is used to verify if the MMP block has been
1835          * updated on the block device. The value is updated based on the
1836          * maximum time to write the MMP block during an update cycle.
1837          */
1838         __le16  mmp_check_interval;
1839
1840         __le16  mmp_pad1;
1841         __le32  mmp_pad2[226];
1842         __le32  mmp_checksum;           /* crc32c(uuid+mmp_block) */
1843 };
1844
1845 /* arguments passed to the mmp thread */
1846 struct mmpd_data {
1847         struct buffer_head *bh; /* bh from initial read_mmp_block() */
1848         struct super_block *sb;  /* super block of the fs */
1849 };
1850
1851 /*
1852  * Check interval multiplier
1853  * The MMP block is written every update interval and initially checked every
1854  * update interval x the multiplier (the value is then adapted based on the
1855  * write latency). The reason is that writes can be delayed under load and we
1856  * don't want readers to incorrectly assume that the filesystem is no longer
1857  * in use.
1858  */
1859 #define EXT4_MMP_CHECK_MULT             2UL
1860
1861 /*
1862  * Minimum interval for MMP checking in seconds.
1863  */
1864 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
1865
1866 /*
1867  * Maximum interval for MMP checking in seconds.
1868  */
1869 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
1870
1871 /*
1872  * Function prototypes
1873  */
1874
1875 /*
1876  * Ok, these declarations are also in <linux/kernel.h> but none of the
1877  * ext4 source programs needs to include it so they are duplicated here.
1878  */
1879 # define NORET_TYPE     /**/
1880 # define ATTRIB_NORET   __attribute__((noreturn))
1881 # define NORET_AND      noreturn,
1882
1883 /* bitmap.c */
1884 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
1885 void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1886                                 struct ext4_group_desc *gdp,
1887                                 struct buffer_head *bh, int sz);
1888 int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1889                                   struct ext4_group_desc *gdp,
1890                                   struct buffer_head *bh, int sz);
1891 void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1892                                 struct ext4_group_desc *gdp,
1893                                 struct buffer_head *bh);
1894 int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1895                                   struct ext4_group_desc *gdp,
1896                                   struct buffer_head *bh);
1897
1898 /* balloc.c */
1899 extern void ext4_get_group_no_and_offset(struct super_block *sb,
1900                                          ext4_fsblk_t blocknr,
1901                                          ext4_group_t *blockgrpp,
1902                                          ext4_grpblk_t *offsetp);
1903 extern ext4_group_t ext4_get_group_number(struct super_block *sb,
1904                                           ext4_fsblk_t block);
1905
1906 extern void ext4_validate_block_bitmap(struct super_block *sb,
1907                                        struct ext4_group_desc *desc,
1908                                        unsigned int block_group,
1909                                        struct buffer_head *bh);
1910 extern unsigned int ext4_block_group(struct super_block *sb,
1911                         ext4_fsblk_t blocknr);
1912 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1913                         ext4_fsblk_t blocknr);
1914 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1915 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1916                         ext4_group_t group);
1917 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1918                                          ext4_fsblk_t goal,
1919                                          unsigned int flags,
1920                                          unsigned long *count,
1921                                          int *errp);
1922 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
1923                                     s64 nclusters, unsigned int flags);
1924 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
1925 extern void ext4_check_blocks_bitmap(struct super_block *);
1926 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1927                                                     ext4_group_t block_group,
1928                                                     struct buffer_head ** bh);
1929 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1930
1931 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
1932                                                 ext4_group_t block_group);
1933 extern int ext4_wait_block_bitmap(struct super_block *sb,
1934                                   ext4_group_t block_group,
1935                                   struct buffer_head *bh);
1936 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1937                                                   ext4_group_t block_group);
1938 extern void ext4_init_block_bitmap(struct super_block *sb,
1939                                    struct buffer_head *bh,
1940                                    ext4_group_t group,
1941                                    struct ext4_group_desc *desc);
1942 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
1943                                               ext4_group_t block_group,
1944                                               struct ext4_group_desc *gdp);
1945 extern unsigned ext4_num_overhead_clusters(struct super_block *sb,
1946                                            ext4_group_t block_group,
1947                                            struct ext4_group_desc *gdp);
1948 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
1949
1950 /* dir.c */
1951 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1952                                   struct file *,
1953                                   struct ext4_dir_entry_2 *,
1954                                   struct buffer_head *, char *, int,
1955                                   unsigned int);
1956 #define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset)      \
1957         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
1958                                         (de), (bh), (buf), (size), (offset)))
1959 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1960                                     __u32 minor_hash,
1961                                     struct ext4_dir_entry_2 *dirent);
1962 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1963 extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
1964                              struct buffer_head *bh,
1965                              void *buf, int buf_size,
1966                              const char *name, int namelen,
1967                              struct ext4_dir_entry_2 **dest_de);
1968 void ext4_insert_dentry(struct inode *inode,
1969                         struct ext4_dir_entry_2 *de,
1970                         int buf_size,
1971                         const char *name, int namelen);
1972 static inline void ext4_update_dx_flag(struct inode *inode)
1973 {
1974         if (!EXT4_HAS_COMPAT_FEATURE(inode->i_sb,
1975                                      EXT4_FEATURE_COMPAT_DIR_INDEX))
1976                 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
1977 }
1978 static unsigned char ext4_filetype_table[] = {
1979         DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1980 };
1981
1982 static inline  unsigned char get_dtype(struct super_block *sb, int filetype)
1983 {
1984         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE) ||
1985             (filetype >= EXT4_FT_MAX))
1986                 return DT_UNKNOWN;
1987
1988         return ext4_filetype_table[filetype];
1989 }
1990
1991 /* fsync.c */
1992 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
1993
1994 /* hash.c */
1995 extern int ext4fs_dirhash(const char *name, int len, struct
1996                           dx_hash_info *hinfo);
1997
1998 /* ialloc.c */
1999 extern struct inode *__ext4_new_inode(handle_t *, struct inode *, umode_t,
2000                                       const struct qstr *qstr, __u32 goal,
2001                                       uid_t *owner, int handle_type,
2002                                       unsigned int line_no, int nblocks);
2003
2004 #define ext4_new_inode(handle, dir, mode, qstr, goal, owner) \
2005         __ext4_new_inode((handle), (dir), (mode), (qstr), (goal), (owner), \
2006                          0, 0, 0)
2007 #define ext4_new_inode_start_handle(dir, mode, qstr, goal, owner, \
2008                                     type, nblocks)                  \
2009         __ext4_new_inode(NULL, (dir), (mode), (qstr), (goal), (owner), \
2010                          (type), __LINE__, (nblocks))
2011
2012
2013 extern void ext4_free_inode(handle_t *, struct inode *);
2014 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2015 extern unsigned long ext4_count_free_inodes(struct super_block *);
2016 extern unsigned long ext4_count_dirs(struct super_block *);
2017 extern void ext4_check_inodes_bitmap(struct super_block *);
2018 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2019 extern int ext4_init_inode_table(struct super_block *sb,
2020                                  ext4_group_t group, int barrier);
2021 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2022
2023 /* mballoc.c */
2024 extern long ext4_mb_stats;
2025 extern long ext4_mb_max_to_scan;
2026 extern int ext4_mb_init(struct super_block *);
2027 extern int ext4_mb_release(struct super_block *);
2028 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2029                                 struct ext4_allocation_request *, int *);
2030 extern int ext4_mb_reserve_blocks(struct super_block *, int);
2031 extern void ext4_discard_preallocations(struct inode *);
2032 extern int __init ext4_init_mballoc(void);
2033 extern void ext4_exit_mballoc(void);
2034 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2035                              struct buffer_head *bh, ext4_fsblk_t block,
2036                              unsigned long count, int flags);
2037 extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2038                                    ext4_group_t ngroups);
2039 extern int ext4_mb_add_groupinfo(struct super_block *sb,
2040                 ext4_group_t i, struct ext4_group_desc *desc);
2041 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2042                                 ext4_fsblk_t block, unsigned long count);
2043 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2044
2045 /* inode.c */
2046 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
2047                                                 ext4_lblk_t, int, int *);
2048 struct buffer_head *ext4_bread(handle_t *, struct inode *,
2049                                                 ext4_lblk_t, int, int *);
2050 int ext4_get_block_write(struct inode *inode, sector_t iblock,
2051                          struct buffer_head *bh_result, int create);
2052 int ext4_get_block(struct inode *inode, sector_t iblock,
2053                                 struct buffer_head *bh_result, int create);
2054 int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2055                            struct buffer_head *bh, int create);
2056 int ext4_walk_page_buffers(handle_t *handle,
2057                            struct buffer_head *head,
2058                            unsigned from,
2059                            unsigned to,
2060                            int *partial,
2061                            int (*fn)(handle_t *handle,
2062                                      struct buffer_head *bh));
2063 int do_journal_get_write_access(handle_t *handle,
2064                                 struct buffer_head *bh);
2065 #define FALL_BACK_TO_NONDELALLOC 1
2066 #define CONVERT_INLINE_DATA      2
2067
2068 extern struct inode *ext4_iget(struct super_block *, unsigned long);
2069 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
2070 extern int  ext4_setattr(struct dentry *, struct iattr *);
2071 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
2072                                 struct kstat *stat);
2073 extern void ext4_evict_inode(struct inode *);
2074 extern void ext4_clear_inode(struct inode *);
2075 extern int  ext4_sync_inode(handle_t *, struct inode *);
2076 extern void ext4_dirty_inode(struct inode *, int);
2077 extern int ext4_change_inode_journal_flag(struct inode *, int);
2078 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
2079 extern int ext4_inode_attach_jinode(struct inode *inode);
2080 extern int ext4_can_truncate(struct inode *inode);
2081 extern void ext4_truncate(struct inode *);
2082 extern int ext4_punch_hole(struct inode *inode, loff_t offset, loff_t length);
2083 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
2084 extern void ext4_set_inode_flags(struct inode *);
2085 extern void ext4_get_inode_flags(struct ext4_inode_info *);
2086 extern int ext4_alloc_da_blocks(struct inode *inode);
2087 extern void ext4_set_aops(struct inode *inode);
2088 extern int ext4_writepage_trans_blocks(struct inode *);
2089 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
2090 extern int ext4_block_truncate_page(handle_t *handle,
2091                 struct address_space *mapping, loff_t from);
2092 extern int ext4_block_zero_page_range(handle_t *handle,
2093                 struct address_space *mapping, loff_t from, loff_t length);
2094 extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
2095                              loff_t lstart, loff_t lend);
2096 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2097 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
2098 extern void ext4_da_update_reserve_space(struct inode *inode,
2099                                         int used, int quota_claim);
2100
2101 /* indirect.c */
2102 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
2103                                 struct ext4_map_blocks *map, int flags);
2104 extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
2105                                 const struct iovec *iov, loff_t offset,
2106                                 unsigned long nr_segs);
2107 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
2108 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
2109 extern void ext4_ind_truncate(handle_t *, struct inode *inode);
2110 extern int ext4_free_hole_blocks(handle_t *handle, struct inode *inode,
2111                                  ext4_lblk_t first, ext4_lblk_t stop);
2112
2113 /* ioctl.c */
2114 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
2115 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
2116
2117 /* migrate.c */
2118 extern int ext4_ext_migrate(struct inode *);
2119 extern int ext4_ind_migrate(struct inode *inode);
2120
2121 /* namei.c */
2122 extern int ext4_dirent_csum_verify(struct inode *inode,
2123                                    struct ext4_dir_entry *dirent);
2124 extern int ext4_orphan_add(handle_t *, struct inode *);
2125 extern int ext4_orphan_del(handle_t *, struct inode *);
2126 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
2127                                 __u32 start_minor_hash, __u32 *next_hash);
2128 extern int search_dir(struct buffer_head *bh,
2129                       char *search_buf,
2130                       int buf_size,
2131                       struct inode *dir,
2132                       const struct qstr *d_name,
2133                       unsigned int offset,
2134                       struct ext4_dir_entry_2 **res_dir);
2135 extern int ext4_generic_delete_entry(handle_t *handle,
2136                                      struct inode *dir,
2137                                      struct ext4_dir_entry_2 *de_del,
2138                                      struct buffer_head *bh,
2139                                      void *entry_buf,
2140                                      int buf_size,
2141                                      int csum_size);
2142
2143 /* resize.c */
2144 extern int ext4_group_add(struct super_block *sb,
2145                                 struct ext4_new_group_data *input);
2146 extern int ext4_group_extend(struct super_block *sb,
2147                                 struct ext4_super_block *es,
2148                                 ext4_fsblk_t n_blocks_count);
2149 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
2150
2151 /* super.c */
2152 extern int ext4_calculate_overhead(struct super_block *sb);
2153 extern int ext4_superblock_csum_verify(struct super_block *sb,
2154                                        struct ext4_super_block *es);
2155 extern void ext4_superblock_csum_set(struct super_block *sb);
2156 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
2157 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
2158 extern void ext4_kvfree(void *ptr);
2159 extern int ext4_alloc_flex_bg_array(struct super_block *sb,
2160                                     ext4_group_t ngroup);
2161 extern const char *ext4_decode_error(struct super_block *sb, int errno,
2162                                      char nbuf[16]);
2163
2164 extern __printf(4, 5)
2165 void __ext4_error(struct super_block *, const char *, unsigned int,
2166                   const char *, ...);
2167 extern __printf(5, 6)
2168 void __ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
2169                       const char *, ...);
2170 extern __printf(5, 6)
2171 void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
2172                      const char *, ...);
2173 extern void __ext4_std_error(struct super_block *, const char *,
2174                              unsigned int, int);
2175 extern __printf(4, 5)
2176 void __ext4_abort(struct super_block *, const char *, unsigned int,
2177                   const char *, ...);
2178 extern __printf(4, 5)
2179 void __ext4_warning(struct super_block *, const char *, unsigned int,
2180                     const char *, ...);
2181 extern __printf(3, 4)
2182 void __ext4_msg(struct super_block *, const char *, const char *, ...);
2183 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
2184                            const char *, unsigned int, const char *);
2185 extern __printf(7, 8)
2186 void __ext4_grp_locked_error(const char *, unsigned int,
2187                              struct super_block *, ext4_group_t,
2188                              unsigned long, ext4_fsblk_t,
2189                              const char *, ...);
2190
2191 #ifdef CONFIG_PRINTK
2192
2193 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2194         __ext4_error_inode(inode, func, line, block, fmt, ##__VA_ARGS__)
2195 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2196         __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__)
2197 #define ext4_error(sb, fmt, ...)                                        \
2198         __ext4_error(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2199 #define ext4_abort(sb, fmt, ...)                                        \
2200         __ext4_abort(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2201 #define ext4_warning(sb, fmt, ...)                                      \
2202         __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2203 #define ext4_msg(sb, level, fmt, ...)                           \
2204         __ext4_msg(sb, level, fmt, ##__VA_ARGS__)
2205 #define dump_mmp_msg(sb, mmp, msg)                                      \
2206         __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg)
2207 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2208         __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \
2209                                 fmt, ##__VA_ARGS__)
2210
2211 #else
2212
2213 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2214 do {                                                                    \
2215         no_printk(fmt, ##__VA_ARGS__);                                  \
2216         __ext4_error_inode(inode, "", 0, block, " ");                   \
2217 } while (0)
2218 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2219 do {                                                                    \
2220         no_printk(fmt, ##__VA_ARGS__);                                  \
2221         __ext4_error_file(file, "", 0, block, " ");                     \
2222 } while (0)
2223 #define ext4_error(sb, fmt, ...)                                        \
2224 do {                                                                    \
2225         no_printk(fmt, ##__VA_ARGS__);                                  \
2226         __ext4_error(sb, "", 0, " ");                                   \
2227 } while (0)
2228 #define ext4_abort(sb, fmt, ...)                                        \
2229 do {                                                                    \
2230         no_printk(fmt, ##__VA_ARGS__);                                  \
2231         __ext4_abort(sb, "", 0, " ");                                   \
2232 } while (0)
2233 #define ext4_warning(sb, fmt, ...)                                      \
2234 do {                                                                    \
2235         no_printk(fmt, ##__VA_ARGS__);                                  \
2236         __ext4_warning(sb, "", 0, " ");                                 \
2237 } while (0)
2238 #define ext4_msg(sb, level, fmt, ...)                                   \
2239 do {                                                                    \
2240         no_printk(fmt, ##__VA_ARGS__);                                  \
2241         __ext4_msg(sb, "", " ");                                        \
2242 } while (0)
2243 #define dump_mmp_msg(sb, mmp, msg)                                      \
2244         __dump_mmp_msg(sb, mmp, "", 0, "")
2245 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2246 do {                                                                    \
2247         no_printk(fmt, ##__VA_ARGS__);                          \
2248         __ext4_grp_locked_error("", 0, sb, grp, ino, block, " ");       \
2249 } while (0)
2250
2251 #endif
2252
2253 extern void ext4_update_dynamic_rev(struct super_block *sb);
2254 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
2255                                         __u32 compat);
2256 extern int ext4_update_rocompat_feature(handle_t *handle,
2257                                         struct super_block *sb, __u32 rocompat);
2258 extern int ext4_update_incompat_feature(handle_t *handle,
2259                                         struct super_block *sb, __u32 incompat);
2260 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
2261                                       struct ext4_group_desc *bg);
2262 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
2263                                       struct ext4_group_desc *bg);
2264 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
2265                                      struct ext4_group_desc *bg);
2266 extern __u32 ext4_free_group_clusters(struct super_block *sb,
2267                                       struct ext4_group_desc *bg);
2268 extern __u32 ext4_free_inodes_count(struct super_block *sb,
2269                                  struct ext4_group_desc *bg);
2270 extern __u32 ext4_used_dirs_count(struct super_block *sb,
2271                                 struct ext4_group_desc *bg);
2272 extern __u32 ext4_itable_unused_count(struct super_block *sb,
2273                                    struct ext4_group_desc *bg);
2274 extern void ext4_block_bitmap_set(struct super_block *sb,
2275                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2276 extern void ext4_inode_bitmap_set(struct super_block *sb,
2277                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2278 extern void ext4_inode_table_set(struct super_block *sb,
2279                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
2280 extern void ext4_free_group_clusters_set(struct super_block *sb,
2281                                          struct ext4_group_desc *bg,
2282                                          __u32 count);
2283 extern void ext4_free_inodes_set(struct super_block *sb,
2284                                 struct ext4_group_desc *bg, __u32 count);
2285 extern void ext4_used_dirs_set(struct super_block *sb,
2286                                 struct ext4_group_desc *bg, __u32 count);
2287 extern void ext4_itable_unused_set(struct super_block *sb,
2288                                    struct ext4_group_desc *bg, __u32 count);
2289 extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
2290                                        struct ext4_group_desc *gdp);
2291 extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
2292                                      struct ext4_group_desc *gdp);
2293 extern int ext4_register_li_request(struct super_block *sb,
2294                                     ext4_group_t first_not_zeroed);
2295
2296 static inline int ext4_has_group_desc_csum(struct super_block *sb)
2297 {
2298         return EXT4_HAS_RO_COMPAT_FEATURE(sb,
2299                                           EXT4_FEATURE_RO_COMPAT_GDT_CSUM |
2300                                           EXT4_FEATURE_RO_COMPAT_METADATA_CSUM);
2301 }
2302
2303 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2304 {
2305         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2306                 le32_to_cpu(es->s_blocks_count_lo);
2307 }
2308
2309 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2310 {
2311         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2312                 le32_to_cpu(es->s_r_blocks_count_lo);
2313 }
2314
2315 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2316 {
2317         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2318                 le32_to_cpu(es->s_free_blocks_count_lo);
2319 }
2320
2321 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2322                                          ext4_fsblk_t blk)
2323 {
2324         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2325         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2326 }
2327
2328 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2329                                               ext4_fsblk_t blk)
2330 {
2331         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2332         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2333 }
2334
2335 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2336                                            ext4_fsblk_t blk)
2337 {
2338         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2339         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2340 }
2341
2342 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2343 {
2344         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2345                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2346                         le32_to_cpu(raw_inode->i_size_lo);
2347         else
2348                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2349 }
2350
2351 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2352 {
2353         raw_inode->i_size_lo = cpu_to_le32(i_size);
2354         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2355 }
2356
2357 static inline
2358 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2359                                             ext4_group_t group)
2360 {
2361          struct ext4_group_info ***grp_info;
2362          long indexv, indexh;
2363          BUG_ON(group >= EXT4_SB(sb)->s_groups_count);
2364          grp_info = EXT4_SB(sb)->s_group_info;
2365          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2366          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2367          return grp_info[indexv][indexh];
2368 }
2369
2370 /*
2371  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2372  * the locking protocol documented in the comments of ext4_group_add()
2373  * in resize.c
2374  */
2375 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2376 {
2377         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2378
2379         smp_rmb();
2380         return ngroups;
2381 }
2382
2383 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2384                                              ext4_group_t block_group)
2385 {
2386         return block_group >> sbi->s_log_groups_per_flex;
2387 }
2388
2389 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2390 {
2391         return 1 << sbi->s_log_groups_per_flex;
2392 }
2393
2394 #define ext4_std_error(sb, errno)                               \
2395 do {                                                            \
2396         if ((errno))                                            \
2397                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2398 } while (0)
2399
2400 #ifdef CONFIG_SMP
2401 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2402  * counters. So we need to make sure we have free clusters more
2403  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2404  */
2405 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2406 #else
2407 #define EXT4_FREECLUSTERS_WATERMARK 0
2408 #endif
2409
2410 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2411 {
2412         /*
2413          * XXX: replace with spinlock if seen contended -bzzz
2414          */
2415         down_write(&EXT4_I(inode)->i_data_sem);
2416         if (newsize > EXT4_I(inode)->i_disksize)
2417                 EXT4_I(inode)->i_disksize = newsize;
2418         up_write(&EXT4_I(inode)->i_data_sem);
2419         return ;
2420 }
2421
2422 struct ext4_group_info {
2423         unsigned long   bb_state;
2424         struct rb_root  bb_free_root;
2425         ext4_grpblk_t   bb_first_free;  /* first free block */
2426         ext4_grpblk_t   bb_free;        /* total free blocks */
2427         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2428         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2429         struct          list_head bb_prealloc_list;
2430 #ifdef DOUBLE_CHECK
2431         void            *bb_bitmap;
2432 #endif
2433         struct rw_semaphore alloc_sem;
2434         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2435                                          * regions, index is order.
2436                                          * bb_counters[3] = 5 means
2437                                          * 5 free 8-block regions. */
2438 };
2439
2440 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2441 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2442
2443 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2444         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2445
2446 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2447         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2448 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2449         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2450 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2451         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2452
2453 #define EXT4_MAX_CONTENTION             8
2454 #define EXT4_CONTENTION_THRESHOLD       2
2455
2456 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2457                                               ext4_group_t group)
2458 {
2459         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2460 }
2461
2462 /*
2463  * Returns true if the filesystem is busy enough that attempts to
2464  * access the block group locks has run into contention.
2465  */
2466 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2467 {
2468         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2469 }
2470
2471 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2472 {
2473         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2474         if (spin_trylock(lock))
2475                 /*
2476                  * We're able to grab the lock right away, so drop the
2477                  * lock contention counter.
2478                  */
2479                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2480         else {
2481                 /*
2482                  * The lock is busy, so bump the contention counter,
2483                  * and then wait on the spin lock.
2484                  */
2485                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2486                                   EXT4_MAX_CONTENTION);
2487                 spin_lock(lock);
2488         }
2489 }
2490
2491 static inline void ext4_unlock_group(struct super_block *sb,
2492                                         ext4_group_t group)
2493 {
2494         spin_unlock(ext4_group_lock_ptr(sb, group));
2495 }
2496
2497 /*
2498  * Block validity checking
2499  */
2500 #define ext4_check_indirect_blockref(inode, bh)                         \
2501         ext4_check_blockref(__func__, __LINE__, inode,                  \
2502                             (__le32 *)(bh)->b_data,                     \
2503                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2504
2505 #define ext4_ind_check_inode(inode)                                     \
2506         ext4_check_blockref(__func__, __LINE__, inode,                  \
2507                             EXT4_I(inode)->i_data,                      \
2508                             EXT4_NDIR_BLOCKS)
2509
2510 /*
2511  * Inodes and files operations
2512  */
2513
2514 /* dir.c */
2515 extern const struct file_operations ext4_dir_operations;
2516
2517 /* file.c */
2518 extern const struct inode_operations ext4_file_inode_operations;
2519 extern const struct file_operations ext4_file_operations;
2520 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2521 extern void ext4_unwritten_wait(struct inode *inode);
2522
2523 /* inline.c */
2524 extern int ext4_has_inline_data(struct inode *inode);
2525 extern int ext4_get_inline_size(struct inode *inode);
2526 extern int ext4_get_max_inline_size(struct inode *inode);
2527 extern int ext4_find_inline_data_nolock(struct inode *inode);
2528 extern void ext4_write_inline_data(struct inode *inode,
2529                                    struct ext4_iloc *iloc,
2530                                    void *buffer, loff_t pos,
2531                                    unsigned int len);
2532 extern int ext4_prepare_inline_data(handle_t *handle, struct inode *inode,
2533                                     unsigned int len);
2534 extern int ext4_init_inline_data(handle_t *handle, struct inode *inode,
2535                                  unsigned int len);
2536 extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
2537
2538 extern int ext4_readpage_inline(struct inode *inode, struct page *page);
2539 extern int ext4_try_to_write_inline_data(struct address_space *mapping,
2540                                          struct inode *inode,
2541                                          loff_t pos, unsigned len,
2542                                          unsigned flags,
2543                                          struct page **pagep);
2544 extern int ext4_write_inline_data_end(struct inode *inode,
2545                                       loff_t pos, unsigned len,
2546                                       unsigned copied,
2547                                       struct page *page);
2548 extern struct buffer_head *
2549 ext4_journalled_write_inline_data(struct inode *inode,
2550                                   unsigned len,
2551                                   struct page *page);
2552 extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
2553                                            struct inode *inode,
2554                                            loff_t pos, unsigned len,
2555                                            unsigned flags,
2556                                            struct page **pagep,
2557                                            void **fsdata);
2558 extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
2559                                          unsigned len, unsigned copied,
2560                                          struct page *page);
2561 extern int ext4_try_add_inline_entry(handle_t *handle, struct dentry *dentry,
2562                                      struct inode *inode);
2563 extern int ext4_try_create_inline_dir(handle_t *handle,
2564                                       struct inode *parent,
2565                                       struct inode *inode);
2566 extern int ext4_read_inline_dir(struct file *filp,
2567                                 struct dir_context *ctx,
2568                                 int *has_inline_data);
2569 extern int htree_inlinedir_to_tree(struct file *dir_file,
2570                                    struct inode *dir, ext4_lblk_t block,
2571                                    struct dx_hash_info *hinfo,
2572                                    __u32 start_hash, __u32 start_minor_hash,
2573                                    int *has_inline_data);
2574 extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
2575                                         const struct qstr *d_name,
2576                                         struct ext4_dir_entry_2 **res_dir,
2577                                         int *has_inline_data);
2578 extern int ext4_delete_inline_entry(handle_t *handle,
2579                                     struct inode *dir,
2580                                     struct ext4_dir_entry_2 *de_del,
2581                                     struct buffer_head *bh,
2582                                     int *has_inline_data);
2583 extern int empty_inline_dir(struct inode *dir, int *has_inline_data);
2584 extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
2585                                         struct ext4_dir_entry_2 **parent_de,
2586                                         int *retval);
2587 extern int ext4_inline_data_fiemap(struct inode *inode,
2588                                    struct fiemap_extent_info *fieinfo,
2589                                    int *has_inline);
2590 extern int ext4_try_to_evict_inline_data(handle_t *handle,
2591                                          struct inode *inode,
2592                                          int needed);
2593 extern void ext4_inline_data_truncate(struct inode *inode, int *has_inline);
2594
2595 extern int ext4_convert_inline_data(struct inode *inode);
2596
2597 /* namei.c */
2598 extern const struct inode_operations ext4_dir_inode_operations;
2599 extern const struct inode_operations ext4_special_inode_operations;
2600 extern struct dentry *ext4_get_parent(struct dentry *child);
2601 extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
2602                                  struct ext4_dir_entry_2 *de,
2603                                  int blocksize, int csum_size,
2604                                  unsigned int parent_ino, int dotdot_real_len);
2605 extern void initialize_dirent_tail(struct ext4_dir_entry_tail *t,
2606                                    unsigned int blocksize);
2607 extern int ext4_handle_dirty_dirent_node(handle_t *handle,
2608                                          struct inode *inode,
2609                                          struct buffer_head *bh);
2610 #define S_SHIFT 12
2611 static unsigned char ext4_type_by_mode[S_IFMT >> S_SHIFT] = {
2612         [S_IFREG >> S_SHIFT]    = EXT4_FT_REG_FILE,
2613         [S_IFDIR >> S_SHIFT]    = EXT4_FT_DIR,
2614         [S_IFCHR >> S_SHIFT]    = EXT4_FT_CHRDEV,
2615         [S_IFBLK >> S_SHIFT]    = EXT4_FT_BLKDEV,
2616         [S_IFIFO >> S_SHIFT]    = EXT4_FT_FIFO,
2617         [S_IFSOCK >> S_SHIFT]   = EXT4_FT_SOCK,
2618         [S_IFLNK >> S_SHIFT]    = EXT4_FT_SYMLINK,
2619 };
2620
2621 static inline void ext4_set_de_type(struct super_block *sb,
2622                                 struct ext4_dir_entry_2 *de,
2623                                 umode_t mode) {
2624         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE))
2625                 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
2626 }
2627
2628
2629 /* symlink.c */
2630 extern const struct inode_operations ext4_symlink_inode_operations;
2631 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2632
2633 /* block_validity */
2634 extern void ext4_release_system_zone(struct super_block *sb);
2635 extern int ext4_setup_system_zone(struct super_block *sb);
2636 extern int __init ext4_init_system_zone(void);
2637 extern void ext4_exit_system_zone(void);
2638 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2639                                  ext4_fsblk_t start_blk,
2640                                  unsigned int count);
2641 extern int ext4_check_blockref(const char *, unsigned int,
2642                                struct inode *, __le32 *, unsigned int);
2643
2644 /* extents.c */
2645 struct ext4_ext_path;
2646 struct ext4_extent;
2647
2648 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2649 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2650 extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
2651 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2652                                struct ext4_map_blocks *map, int flags);
2653 extern void ext4_ext_truncate(handle_t *, struct inode *);
2654 extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
2655                                  ext4_lblk_t end);
2656 extern void ext4_ext_init(struct super_block *);
2657 extern void ext4_ext_release(struct super_block *);
2658 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2659                           loff_t len);
2660 extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
2661                                           loff_t offset, ssize_t len);
2662 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2663                            struct ext4_map_blocks *map, int flags);
2664 extern int ext4_ext_calc_metadata_amount(struct inode *inode,
2665                                          ext4_lblk_t lblocks);
2666 extern int ext4_extent_tree_init(handle_t *, struct inode *);
2667 extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
2668                                                    int num,
2669                                                    struct ext4_ext_path *path);
2670 extern int ext4_can_extents_be_merged(struct inode *inode,
2671                                       struct ext4_extent *ex1,
2672                                       struct ext4_extent *ex2);
2673 extern int ext4_ext_insert_extent(handle_t *, struct inode *,
2674                                   struct ext4_ext_path *,
2675                                   struct ext4_extent *, int);
2676 extern struct ext4_ext_path *ext4_ext_find_extent(struct inode *, ext4_lblk_t,
2677                                                   struct ext4_ext_path *);
2678 extern void ext4_ext_drop_refs(struct ext4_ext_path *);
2679 extern int ext4_ext_check_inode(struct inode *inode);
2680 extern int ext4_find_delalloc_range(struct inode *inode,
2681                                     ext4_lblk_t lblk_start,
2682                                     ext4_lblk_t lblk_end);
2683 extern int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk);
2684 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2685                         __u64 start, __u64 len);
2686
2687
2688 /* move_extent.c */
2689 extern void ext4_double_down_write_data_sem(struct inode *first,
2690                                             struct inode *second);
2691 extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
2692                                           struct inode *donor_inode);
2693 void ext4_inode_double_lock(struct inode *inode1, struct inode *inode2);
2694 void ext4_inode_double_unlock(struct inode *inode1, struct inode *inode2);
2695 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2696                              __u64 start_orig, __u64 start_donor,
2697                              __u64 len, __u64 *moved_len);
2698
2699 /* page-io.c */
2700 extern int __init ext4_init_pageio(void);
2701 extern void ext4_exit_pageio(void);
2702 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2703 extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
2704 extern int ext4_put_io_end(ext4_io_end_t *io_end);
2705 extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
2706 extern void ext4_io_submit_init(struct ext4_io_submit *io,
2707                                 struct writeback_control *wbc);
2708 extern void ext4_end_io_rsv_work(struct work_struct *work);
2709 extern void ext4_io_submit(struct ext4_io_submit *io);
2710 extern int ext4_bio_write_page(struct ext4_io_submit *io,
2711                                struct page *page,
2712                                int len,
2713                                struct writeback_control *wbc);
2714
2715 /* mmp.c */
2716 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2717 extern void ext4_mmp_csum_set(struct super_block *sb, struct mmp_struct *mmp);
2718 extern int ext4_mmp_csum_verify(struct super_block *sb,
2719                                 struct mmp_struct *mmp);
2720
2721 /*
2722  * Note that these flags will never ever appear in a buffer_head's state flag.
2723  * See EXT4_MAP_... to see where this is used.
2724  */
2725 enum ext4_state_bits {
2726         BH_Uninit       /* blocks are allocated but uninitialized on disk */
2727          = BH_JBDPrivateStart,
2728         BH_AllocFromCluster,    /* allocated blocks were part of already
2729                                  * allocated cluster. */
2730 };
2731
2732 /*
2733  * Add new method to test whether block and inode bitmaps are properly
2734  * initialized. With uninit_bg reading the block from disk is not enough
2735  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2736  */
2737 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2738
2739 static inline int bitmap_uptodate(struct buffer_head *bh)
2740 {
2741         return (buffer_uptodate(bh) &&
2742                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2743 }
2744 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2745 {
2746         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2747 }
2748
2749 /*
2750  * Disable DIO read nolock optimization, so new dioreaders will be forced
2751  * to grab i_mutex
2752  */
2753 static inline void ext4_inode_block_unlocked_dio(struct inode *inode)
2754 {
2755         ext4_set_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2756         smp_mb();
2757 }
2758 static inline void ext4_inode_resume_unlocked_dio(struct inode *inode)
2759 {
2760         smp_mb();
2761         ext4_clear_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2762 }
2763
2764 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2765
2766 /* For ioend & aio unwritten conversion wait queues */
2767 #define EXT4_WQ_HASH_SZ         37
2768 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
2769                                             EXT4_WQ_HASH_SZ])
2770 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
2771                                              EXT4_WQ_HASH_SZ])
2772 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2773 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2774
2775 #define EXT4_RESIZING   0
2776 extern int ext4_resize_begin(struct super_block *sb);
2777 extern void ext4_resize_end(struct super_block *sb);
2778
2779 #endif  /* __KERNEL__ */
2780
2781 #endif  /* _EXT4_H */
This page took 0.189772 seconds and 4 git commands to generate.