GNU Linux-libre 4.9.318-gnu1
[releases.git] / fs / ext4 / ext4.h
1 /*
2  *  ext4.h
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/include/linux/minix_fs.h
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  */
15
16 #ifndef _EXT4_H
17 #define _EXT4_H
18
19 #include <linux/types.h>
20 #include <linux/blkdev.h>
21 #include <linux/magic.h>
22 #include <linux/jbd2.h>
23 #include <linux/quota.h>
24 #include <linux/rwsem.h>
25 #include <linux/rbtree.h>
26 #include <linux/seqlock.h>
27 #include <linux/mutex.h>
28 #include <linux/timer.h>
29 #include <linux/version.h>
30 #include <linux/wait.h>
31 #include <linux/blockgroup_lock.h>
32 #include <linux/percpu_counter.h>
33 #include <linux/ratelimit.h>
34 #include <crypto/hash.h>
35 #include <linux/fscrypto.h>
36 #include <linux/falloc.h>
37 #include <linux/percpu-rwsem.h>
38 #ifdef __KERNEL__
39 #include <linux/compat.h>
40 #endif
41
42 /*
43  * The fourth extended filesystem constants/structures
44  */
45
46 /*
47  * with AGGRESSIVE_CHECK allocator runs consistency checks over
48  * structures. these checks slow things down a lot
49  */
50 #define AGGRESSIVE_CHECK__
51
52 /*
53  * with DOUBLE_CHECK defined mballoc creates persistent in-core
54  * bitmaps, maintains and uses them to check for double allocations
55  */
56 #define DOUBLE_CHECK__
57
58 /*
59  * Define EXT4FS_DEBUG to produce debug messages
60  */
61 #undef EXT4FS_DEBUG
62
63 /*
64  * Debug code
65  */
66 #ifdef EXT4FS_DEBUG
67 #define ext4_debug(f, a...)                                             \
68         do {                                                            \
69                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
70                         __FILE__, __LINE__, __func__);                  \
71                 printk(KERN_DEBUG f, ## a);                             \
72         } while (0)
73 #else
74 #define ext4_debug(fmt, ...)    no_printk(fmt, ##__VA_ARGS__)
75 #endif
76
77 /*
78  * Turn on EXT_DEBUG to get lots of info about extents operations.
79  */
80 #define EXT_DEBUG__
81 #ifdef EXT_DEBUG
82 #define ext_debug(fmt, ...)     printk(fmt, ##__VA_ARGS__)
83 #else
84 #define ext_debug(fmt, ...)     no_printk(fmt, ##__VA_ARGS__)
85 #endif
86
87 /* data type for block offset of block group */
88 typedef int ext4_grpblk_t;
89
90 /* data type for filesystem-wide blocks number */
91 typedef unsigned long long ext4_fsblk_t;
92
93 /* data type for file logical block number */
94 typedef __u32 ext4_lblk_t;
95
96 /* data type for block group number */
97 typedef unsigned int ext4_group_t;
98
99 enum SHIFT_DIRECTION {
100         SHIFT_LEFT = 0,
101         SHIFT_RIGHT,
102 };
103
104 /*
105  * Flags used in mballoc's allocation_context flags field.
106  *
107  * Also used to show what's going on for debugging purposes when the
108  * flag field is exported via the traceport interface
109  */
110
111 /* prefer goal again. length */
112 #define EXT4_MB_HINT_MERGE              0x0001
113 /* blocks already reserved */
114 #define EXT4_MB_HINT_RESERVED           0x0002
115 /* metadata is being allocated */
116 #define EXT4_MB_HINT_METADATA           0x0004
117 /* first blocks in the file */
118 #define EXT4_MB_HINT_FIRST              0x0008
119 /* search for the best chunk */
120 #define EXT4_MB_HINT_BEST               0x0010
121 /* data is being allocated */
122 #define EXT4_MB_HINT_DATA               0x0020
123 /* don't preallocate (for tails) */
124 #define EXT4_MB_HINT_NOPREALLOC         0x0040
125 /* allocate for locality group */
126 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
127 /* allocate goal blocks or none */
128 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
129 /* goal is meaningful */
130 #define EXT4_MB_HINT_TRY_GOAL           0x0200
131 /* blocks already pre-reserved by delayed allocation */
132 #define EXT4_MB_DELALLOC_RESERVED       0x0400
133 /* We are doing stream allocation */
134 #define EXT4_MB_STREAM_ALLOC            0x0800
135 /* Use reserved root blocks if needed */
136 #define EXT4_MB_USE_ROOT_BLOCKS         0x1000
137 /* Use blocks from reserved pool */
138 #define EXT4_MB_USE_RESERVED            0x2000
139
140 struct ext4_allocation_request {
141         /* target inode for block we're allocating */
142         struct inode *inode;
143         /* how many blocks we want to allocate */
144         unsigned int len;
145         /* logical block in target inode */
146         ext4_lblk_t logical;
147         /* the closest logical allocated block to the left */
148         ext4_lblk_t lleft;
149         /* the closest logical allocated block to the right */
150         ext4_lblk_t lright;
151         /* phys. target (a hint) */
152         ext4_fsblk_t goal;
153         /* phys. block for the closest logical allocated block to the left */
154         ext4_fsblk_t pleft;
155         /* phys. block for the closest logical allocated block to the right */
156         ext4_fsblk_t pright;
157         /* flags. see above EXT4_MB_HINT_* */
158         unsigned int flags;
159 };
160
161 /*
162  * Logical to physical block mapping, used by ext4_map_blocks()
163  *
164  * This structure is used to pass requests into ext4_map_blocks() as
165  * well as to store the information returned by ext4_map_blocks().  It
166  * takes less room on the stack than a struct buffer_head.
167  */
168 #define EXT4_MAP_NEW            (1 << BH_New)
169 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
170 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
171 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
172 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
173                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY)
174
175 struct ext4_map_blocks {
176         ext4_fsblk_t m_pblk;
177         ext4_lblk_t m_lblk;
178         unsigned int m_len;
179         unsigned int m_flags;
180 };
181
182 /*
183  * Flags for ext4_io_end->flags
184  */
185 #define EXT4_IO_END_UNWRITTEN   0x0001
186
187 /*
188  * For converting unwritten extents on a work queue. 'handle' is used for
189  * buffered writeback.
190  */
191 typedef struct ext4_io_end {
192         struct list_head        list;           /* per-file finished IO list */
193         handle_t                *handle;        /* handle reserved for extent
194                                                  * conversion */
195         struct inode            *inode;         /* file being written to */
196         struct bio              *bio;           /* Linked list of completed
197                                                  * bios covering the extent */
198         unsigned int            flag;           /* unwritten or not */
199         atomic_t                count;          /* reference counter */
200         loff_t                  offset;         /* offset in the file */
201         ssize_t                 size;           /* size of the extent */
202 } ext4_io_end_t;
203
204 struct ext4_io_submit {
205         struct writeback_control *io_wbc;
206         struct bio              *io_bio;
207         ext4_io_end_t           *io_end;
208         sector_t                io_next_block;
209 };
210
211 /*
212  * Special inodes numbers
213  */
214 #define EXT4_BAD_INO             1      /* Bad blocks inode */
215 #define EXT4_ROOT_INO            2      /* Root inode */
216 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
217 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
218 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
219 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
220 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
221 #define EXT4_JOURNAL_INO         8      /* Journal inode */
222
223 /* First non-reserved inode for old ext4 filesystems */
224 #define EXT4_GOOD_OLD_FIRST_INO 11
225
226 /*
227  * Maximal count of links to a file
228  */
229 #define EXT4_LINK_MAX           65000
230
231 /*
232  * Macro-instructions used to manage several block sizes
233  */
234 #define EXT4_MIN_BLOCK_SIZE             1024
235 #define EXT4_MAX_BLOCK_SIZE             65536
236 #define EXT4_MIN_BLOCK_LOG_SIZE         10
237 #define EXT4_MAX_BLOCK_LOG_SIZE         16
238 #define EXT4_MAX_CLUSTER_LOG_SIZE       30
239 #ifdef __KERNEL__
240 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
241 #else
242 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
243 #endif
244 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
245 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
246                                          EXT4_SB(s)->s_cluster_bits)
247 #ifdef __KERNEL__
248 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
249 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
250 #else
251 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
252 #endif
253 #ifdef __KERNEL__
254 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
255 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
256 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
257 #else
258 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
259                                  EXT4_GOOD_OLD_INODE_SIZE : \
260                                  (s)->s_inode_size)
261 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
262                                  EXT4_GOOD_OLD_FIRST_INO : \
263                                  (s)->s_first_ino)
264 #endif
265 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
266 #define EXT4_MAX_BLOCKS(size, offset, blkbits) \
267         ((EXT4_BLOCK_ALIGN(size + offset, blkbits) >> blkbits) - (offset >> \
268                                                                   blkbits))
269
270 /* Translate a block number to a cluster number */
271 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
272 /* Translate a cluster number to a block number */
273 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
274 /* Translate # of blks to # of clusters */
275 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
276                                  (sbi)->s_cluster_bits)
277 /* Mask out the low bits to get the starting block of the cluster */
278 #define EXT4_PBLK_CMASK(s, pblk) ((pblk) &                              \
279                                   ~((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
280 #define EXT4_LBLK_CMASK(s, lblk) ((lblk) &                              \
281                                   ~((ext4_lblk_t) (s)->s_cluster_ratio - 1))
282 /* Get the cluster offset */
283 #define EXT4_PBLK_COFF(s, pblk) ((pblk) &                               \
284                                  ((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
285 #define EXT4_LBLK_COFF(s, lblk) ((lblk) &                               \
286                                  ((ext4_lblk_t) (s)->s_cluster_ratio - 1))
287
288 /*
289  * Structure of a blocks group descriptor
290  */
291 struct ext4_group_desc
292 {
293         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
294         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
295         __le32  bg_inode_table_lo;      /* Inodes table block */
296         __le16  bg_free_blocks_count_lo;/* Free blocks count */
297         __le16  bg_free_inodes_count_lo;/* Free inodes count */
298         __le16  bg_used_dirs_count_lo;  /* Directories count */
299         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
300         __le32  bg_exclude_bitmap_lo;   /* Exclude bitmap for snapshots */
301         __le16  bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
302         __le16  bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
303         __le16  bg_itable_unused_lo;    /* Unused inodes count */
304         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
305         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
306         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
307         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
308         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
309         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
310         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
311         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
312         __le32  bg_exclude_bitmap_hi;   /* Exclude bitmap block MSB */
313         __le16  bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
314         __le16  bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
315         __u32   bg_reserved;
316 };
317
318 #define EXT4_BG_INODE_BITMAP_CSUM_HI_END        \
319         (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
320          sizeof(__le16))
321 #define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END        \
322         (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
323          sizeof(__le16))
324
325 /*
326  * Structure of a flex block group info
327  */
328
329 struct flex_groups {
330         atomic64_t      free_clusters;
331         atomic_t        free_inodes;
332         atomic_t        used_dirs;
333 };
334
335 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
336 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
337 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
338
339 /*
340  * Macro-instructions used to manage group descriptors
341  */
342 #define EXT4_MIN_DESC_SIZE              32
343 #define EXT4_MIN_DESC_SIZE_64BIT        64
344 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
345 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
346 #ifdef __KERNEL__
347 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
348 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
349 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
350 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
351 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
352 #else
353 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
354 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
355 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
356 #endif
357
358 /*
359  * Constants relative to the data blocks
360  */
361 #define EXT4_NDIR_BLOCKS                12
362 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
363 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
364 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
365 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
366
367 /*
368  * Inode flags
369  */
370 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
371 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
372 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
373 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
374 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
375 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
376 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
377 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
378 /* Reserved for compression usage... */
379 #define EXT4_DIRTY_FL                   0x00000100
380 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
381 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
382         /* nb: was previously EXT2_ECOMPR_FL */
383 #define EXT4_ENCRYPT_FL                 0x00000800 /* encrypted file */
384 /* End compression flags --- maybe not all used */
385 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
386 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
387 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
388 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
389 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
390 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
391 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
392 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
393 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
394 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
395 #define EXT4_INLINE_DATA_FL             0x10000000 /* Inode has inline data. */
396 #define EXT4_PROJINHERIT_FL             0x20000000 /* Create with parents projid */
397 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
398
399 #define EXT4_FL_USER_VISIBLE            0x304BDFFF /* User visible flags */
400 #define EXT4_FL_USER_MODIFIABLE         0x204380FF /* User modifiable flags */
401
402 #define EXT4_FL_XFLAG_VISIBLE           (EXT4_SYNC_FL | \
403                                          EXT4_IMMUTABLE_FL | \
404                                          EXT4_APPEND_FL | \
405                                          EXT4_NODUMP_FL | \
406                                          EXT4_NOATIME_FL | \
407                                          EXT4_PROJINHERIT_FL)
408
409 /* Flags that should be inherited by new inodes from their parent. */
410 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
411                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
412                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
413                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL |\
414                            EXT4_PROJINHERIT_FL)
415
416 /* Flags that are appropriate for regular files (all but dir-specific ones). */
417 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
418
419 /* Flags that are appropriate for non-directories/regular files. */
420 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
421
422 /* Mask out flags that are inappropriate for the given type of inode. */
423 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
424 {
425         if (S_ISDIR(mode))
426                 return flags;
427         else if (S_ISREG(mode))
428                 return flags & EXT4_REG_FLMASK;
429         else
430                 return flags & EXT4_OTHER_FLMASK;
431 }
432
433 /*
434  * Inode flags used for atomic set/get
435  */
436 enum {
437         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
438         EXT4_INODE_UNRM         = 1,    /* Undelete */
439         EXT4_INODE_COMPR        = 2,    /* Compress file */
440         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
441         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
442         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
443         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
444         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
445 /* Reserved for compression usage... */
446         EXT4_INODE_DIRTY        = 8,
447         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
448         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
449         EXT4_INODE_ENCRYPT      = 11,   /* Encrypted file */
450 /* End compression flags --- maybe not all used */
451         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
452         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
453         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
454         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
455         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
456         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
457         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
458         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
459         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
460         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
461         EXT4_INODE_INLINE_DATA  = 28,   /* Data in inode. */
462         EXT4_INODE_PROJINHERIT  = 29,   /* Create with parents projid */
463         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
464 };
465
466 /*
467  * Since it's pretty easy to mix up bit numbers and hex values, we use a
468  * build-time check to make sure that EXT4_XXX_FL is consistent with respect to
469  * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost
470  * any extra space in the compiled kernel image, otherwise, the build will fail.
471  * It's important that these values are the same, since we are using
472  * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent
473  * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk
474  * values found in ext2, ext3 and ext4 filesystems, and of course the values
475  * defined in e2fsprogs.
476  *
477  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
478  */
479 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
480 #define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG))
481
482 static inline void ext4_check_flag_values(void)
483 {
484         CHECK_FLAG_VALUE(SECRM);
485         CHECK_FLAG_VALUE(UNRM);
486         CHECK_FLAG_VALUE(COMPR);
487         CHECK_FLAG_VALUE(SYNC);
488         CHECK_FLAG_VALUE(IMMUTABLE);
489         CHECK_FLAG_VALUE(APPEND);
490         CHECK_FLAG_VALUE(NODUMP);
491         CHECK_FLAG_VALUE(NOATIME);
492         CHECK_FLAG_VALUE(DIRTY);
493         CHECK_FLAG_VALUE(COMPRBLK);
494         CHECK_FLAG_VALUE(NOCOMPR);
495         CHECK_FLAG_VALUE(ENCRYPT);
496         CHECK_FLAG_VALUE(INDEX);
497         CHECK_FLAG_VALUE(IMAGIC);
498         CHECK_FLAG_VALUE(JOURNAL_DATA);
499         CHECK_FLAG_VALUE(NOTAIL);
500         CHECK_FLAG_VALUE(DIRSYNC);
501         CHECK_FLAG_VALUE(TOPDIR);
502         CHECK_FLAG_VALUE(HUGE_FILE);
503         CHECK_FLAG_VALUE(EXTENTS);
504         CHECK_FLAG_VALUE(EA_INODE);
505         CHECK_FLAG_VALUE(EOFBLOCKS);
506         CHECK_FLAG_VALUE(INLINE_DATA);
507         CHECK_FLAG_VALUE(PROJINHERIT);
508         CHECK_FLAG_VALUE(RESERVED);
509 }
510
511 /* Used to pass group descriptor data when online resize is done */
512 struct ext4_new_group_input {
513         __u32 group;            /* Group number for this data */
514         __u64 block_bitmap;     /* Absolute block number of block bitmap */
515         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
516         __u64 inode_table;      /* Absolute block number of inode table start */
517         __u32 blocks_count;     /* Total number of blocks in this group */
518         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
519         __u16 unused;
520 };
521
522 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
523 struct compat_ext4_new_group_input {
524         u32 group;
525         compat_u64 block_bitmap;
526         compat_u64 inode_bitmap;
527         compat_u64 inode_table;
528         u32 blocks_count;
529         u16 reserved_blocks;
530         u16 unused;
531 };
532 #endif
533
534 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
535 struct ext4_new_group_data {
536         __u32 group;
537         __u64 block_bitmap;
538         __u64 inode_bitmap;
539         __u64 inode_table;
540         __u32 blocks_count;
541         __u16 reserved_blocks;
542         __u16 unused;
543         __u32 free_blocks_count;
544 };
545
546 /* Indexes used to index group tables in ext4_new_group_data */
547 enum {
548         BLOCK_BITMAP = 0,       /* block bitmap */
549         INODE_BITMAP,           /* inode bitmap */
550         INODE_TABLE,            /* inode tables */
551         GROUP_TABLE_COUNT,
552 };
553
554 /*
555  * Flags used by ext4_map_blocks()
556  */
557         /* Allocate any needed blocks and/or convert an unwritten
558            extent to be an initialized ext4 */
559 #define EXT4_GET_BLOCKS_CREATE                  0x0001
560         /* Request the creation of an unwritten extent */
561 #define EXT4_GET_BLOCKS_UNWRIT_EXT              0x0002
562 #define EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT       (EXT4_GET_BLOCKS_UNWRIT_EXT|\
563                                                  EXT4_GET_BLOCKS_CREATE)
564         /* Caller is from the delayed allocation writeout path
565          * finally doing the actual allocation of delayed blocks */
566 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
567         /* caller is from the direct IO path, request to creation of an
568         unwritten extents if not allocated, split the unwritten
569         extent if blocks has been preallocated already*/
570 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
571 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
572 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
573                                          EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
574         /* Convert extent to initialized after IO complete */
575 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
576                                          EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
577         /* Eventual metadata allocation (due to growing extent tree)
578          * should not fail, so try to use reserved blocks for that.*/
579 #define EXT4_GET_BLOCKS_METADATA_NOFAIL         0x0020
580         /* Don't normalize allocation size (used for fallocate) */
581 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
582         /* Request will not result in inode size update (user for fallocate) */
583 #define EXT4_GET_BLOCKS_KEEP_SIZE               0x0080
584         /* Convert written extents to unwritten */
585 #define EXT4_GET_BLOCKS_CONVERT_UNWRITTEN       0x0100
586         /* Write zeros to newly created written extents */
587 #define EXT4_GET_BLOCKS_ZERO                    0x0200
588 #define EXT4_GET_BLOCKS_CREATE_ZERO             (EXT4_GET_BLOCKS_CREATE |\
589                                         EXT4_GET_BLOCKS_ZERO)
590         /* Caller will submit data before dropping transaction handle. This
591          * allows jbd2 to avoid submitting data before commit. */
592 #define EXT4_GET_BLOCKS_IO_SUBMIT               0x0400
593
594 /*
595  * The bit position of these flags must not overlap with any of the
596  * EXT4_GET_BLOCKS_*.  They are used by ext4_find_extent(),
597  * read_extent_tree_block(), ext4_split_extent_at(),
598  * ext4_ext_insert_extent(), and ext4_ext_create_new_leaf().
599  * EXT4_EX_NOCACHE is used to indicate that the we shouldn't be
600  * caching the extents when reading from the extent tree while a
601  * truncate or punch hole operation is in progress.
602  */
603 #define EXT4_EX_NOCACHE                         0x40000000
604 #define EXT4_EX_FORCE_CACHE                     0x20000000
605
606 /*
607  * Flags used by ext4_free_blocks
608  */
609 #define EXT4_FREE_BLOCKS_METADATA       0x0001
610 #define EXT4_FREE_BLOCKS_FORGET         0x0002
611 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
612 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
613 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
614 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
615
616 /*
617  * ioctl commands
618  */
619 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
620 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
621 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
622 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
623 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
624 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
625 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
626 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
627 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
628 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
629 #define EXT4_IOC_MIGRATE                _IO('f', 9)
630  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
631  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
632 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
633 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
634 #define EXT4_IOC_RESIZE_FS              _IOW('f', 16, __u64)
635 #define EXT4_IOC_SWAP_BOOT              _IO('f', 17)
636 #define EXT4_IOC_PRECACHE_EXTENTS       _IO('f', 18)
637 #define EXT4_IOC_SET_ENCRYPTION_POLICY  FS_IOC_SET_ENCRYPTION_POLICY
638 #define EXT4_IOC_GET_ENCRYPTION_PWSALT  FS_IOC_GET_ENCRYPTION_PWSALT
639 #define EXT4_IOC_GET_ENCRYPTION_POLICY  FS_IOC_GET_ENCRYPTION_POLICY
640
641 #ifndef FS_IOC_FSGETXATTR
642 /* Until the uapi changes get merged for project quota... */
643
644 #define FS_IOC_FSGETXATTR               _IOR('X', 31, struct fsxattr)
645 #define FS_IOC_FSSETXATTR               _IOW('X', 32, struct fsxattr)
646
647 /*
648  * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
649  */
650 struct fsxattr {
651         __u32           fsx_xflags;     /* xflags field value (get/set) */
652         __u32           fsx_extsize;    /* extsize field value (get/set)*/
653         __u32           fsx_nextents;   /* nextents field value (get)   */
654         __u32           fsx_projid;     /* project identifier (get/set) */
655         unsigned char   fsx_pad[12];
656 };
657
658 /*
659  * Flags for the fsx_xflags field
660  */
661 #define FS_XFLAG_REALTIME       0x00000001      /* data in realtime volume */
662 #define FS_XFLAG_PREALLOC       0x00000002      /* preallocated file extents */
663 #define FS_XFLAG_IMMUTABLE      0x00000008      /* file cannot be modified */
664 #define FS_XFLAG_APPEND         0x00000010      /* all writes append */
665 #define FS_XFLAG_SYNC           0x00000020      /* all writes synchronous */
666 #define FS_XFLAG_NOATIME        0x00000040      /* do not update access time */
667 #define FS_XFLAG_NODUMP         0x00000080      /* do not include in backups */
668 #define FS_XFLAG_RTINHERIT      0x00000100      /* create with rt bit set */
669 #define FS_XFLAG_PROJINHERIT    0x00000200      /* create with parents projid */
670 #define FS_XFLAG_NOSYMLINKS     0x00000400      /* disallow symlink creation */
671 #define FS_XFLAG_EXTSIZE        0x00000800      /* extent size allocator hint */
672 #define FS_XFLAG_EXTSZINHERIT   0x00001000      /* inherit inode extent size */
673 #define FS_XFLAG_NODEFRAG       0x00002000      /* do not defragment */
674 #define FS_XFLAG_FILESTREAM     0x00004000      /* use filestream allocator */
675 #define FS_XFLAG_HASATTR        0x80000000      /* no DIFLAG for this */
676 #endif /* !defined(FS_IOC_FSGETXATTR) */
677
678 #define EXT4_IOC_FSGETXATTR             FS_IOC_FSGETXATTR
679 #define EXT4_IOC_FSSETXATTR             FS_IOC_FSSETXATTR
680
681 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
682 /*
683  * ioctl commands in 32 bit emulation
684  */
685 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
686 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
687 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
688 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
689 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
690 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
691 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
692 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
693 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
694 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
695 #endif
696
697 /* Max physical block we can address w/o extents */
698 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
699
700 /*
701  * Structure of an inode on the disk
702  */
703 struct ext4_inode {
704         __le16  i_mode;         /* File mode */
705         __le16  i_uid;          /* Low 16 bits of Owner Uid */
706         __le32  i_size_lo;      /* Size in bytes */
707         __le32  i_atime;        /* Access time */
708         __le32  i_ctime;        /* Inode Change time */
709         __le32  i_mtime;        /* Modification time */
710         __le32  i_dtime;        /* Deletion Time */
711         __le16  i_gid;          /* Low 16 bits of Group Id */
712         __le16  i_links_count;  /* Links count */
713         __le32  i_blocks_lo;    /* Blocks count */
714         __le32  i_flags;        /* File flags */
715         union {
716                 struct {
717                         __le32  l_i_version;
718                 } linux1;
719                 struct {
720                         __u32  h_i_translator;
721                 } hurd1;
722                 struct {
723                         __u32  m_i_reserved1;
724                 } masix1;
725         } osd1;                         /* OS dependent 1 */
726         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
727         __le32  i_generation;   /* File version (for NFS) */
728         __le32  i_file_acl_lo;  /* File ACL */
729         __le32  i_size_high;
730         __le32  i_obso_faddr;   /* Obsoleted fragment address */
731         union {
732                 struct {
733                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
734                         __le16  l_i_file_acl_high;
735                         __le16  l_i_uid_high;   /* these 2 fields */
736                         __le16  l_i_gid_high;   /* were reserved2[0] */
737                         __le16  l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
738                         __le16  l_i_reserved;
739                 } linux2;
740                 struct {
741                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
742                         __u16   h_i_mode_high;
743                         __u16   h_i_uid_high;
744                         __u16   h_i_gid_high;
745                         __u32   h_i_author;
746                 } hurd2;
747                 struct {
748                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
749                         __le16  m_i_file_acl_high;
750                         __u32   m_i_reserved2[2];
751                 } masix2;
752         } osd2;                         /* OS dependent 2 */
753         __le16  i_extra_isize;
754         __le16  i_checksum_hi;  /* crc32c(uuid+inum+inode) BE */
755         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
756         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
757         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
758         __le32  i_crtime;       /* File Creation time */
759         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
760         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
761         __le32  i_projid;       /* Project ID */
762 };
763
764 struct move_extent {
765         __u32 reserved;         /* should be zero */
766         __u32 donor_fd;         /* donor file descriptor */
767         __u64 orig_start;       /* logical start offset in block for orig */
768         __u64 donor_start;      /* logical start offset in block for donor */
769         __u64 len;              /* block length to be moved */
770         __u64 moved_len;        /* moved block length */
771 };
772
773 #define EXT4_EPOCH_BITS 2
774 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
775 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
776
777 /*
778  * Extended fields will fit into an inode if the filesystem was formatted
779  * with large inodes (-I 256 or larger) and there are not currently any EAs
780  * consuming all of the available space. For new inodes we always reserve
781  * enough space for the kernel's known extended fields, but for inodes
782  * created with an old kernel this might not have been the case. None of
783  * the extended inode fields is critical for correct filesystem operation.
784  * This macro checks if a certain field fits in the inode. Note that
785  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
786  */
787 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
788         ((offsetof(typeof(*ext4_inode), field) +        \
789           sizeof((ext4_inode)->field))                  \
790         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
791             (einode)->i_extra_isize))                   \
792
793 /*
794  * We use an encoding that preserves the times for extra epoch "00":
795  *
796  * extra  msb of                         adjust for signed
797  * epoch  32-bit                         32-bit tv_sec to
798  * bits   time    decoded 64-bit tv_sec  64-bit tv_sec      valid time range
799  * 0 0    1    -0x80000000..-0x00000001  0x000000000 1901-12-13..1969-12-31
800  * 0 0    0    0x000000000..0x07fffffff  0x000000000 1970-01-01..2038-01-19
801  * 0 1    1    0x080000000..0x0ffffffff  0x100000000 2038-01-19..2106-02-07
802  * 0 1    0    0x100000000..0x17fffffff  0x100000000 2106-02-07..2174-02-25
803  * 1 0    1    0x180000000..0x1ffffffff  0x200000000 2174-02-25..2242-03-16
804  * 1 0    0    0x200000000..0x27fffffff  0x200000000 2242-03-16..2310-04-04
805  * 1 1    1    0x280000000..0x2ffffffff  0x300000000 2310-04-04..2378-04-22
806  * 1 1    0    0x300000000..0x37fffffff  0x300000000 2378-04-22..2446-05-10
807  *
808  * Note that previous versions of the kernel on 64-bit systems would
809  * incorrectly use extra epoch bits 1,1 for dates between 1901 and
810  * 1970.  e2fsck will correct this, assuming that it is run on the
811  * affected filesystem before 2242.
812  */
813
814 static inline __le32 ext4_encode_extra_time(struct timespec *time)
815 {
816         u32 extra = sizeof(time->tv_sec) > 4 ?
817                 ((time->tv_sec - (s32)time->tv_sec) >> 32) & EXT4_EPOCH_MASK : 0;
818         return cpu_to_le32(extra | (time->tv_nsec << EXT4_EPOCH_BITS));
819 }
820
821 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
822 {
823         if (unlikely(sizeof(time->tv_sec) > 4 &&
824                         (extra & cpu_to_le32(EXT4_EPOCH_MASK)))) {
825 #if LINUX_VERSION_CODE < KERNEL_VERSION(4,20,0)
826                 /* Handle legacy encoding of pre-1970 dates with epoch
827                  * bits 1,1.  We assume that by kernel version 4.20,
828                  * everyone will have run fsck over the affected
829                  * filesystems to correct the problem.  (This
830                  * backwards compatibility may be removed before this
831                  * time, at the discretion of the ext4 developers.)
832                  */
833                 u64 extra_bits = le32_to_cpu(extra) & EXT4_EPOCH_MASK;
834                 if (extra_bits == 3 && ((time->tv_sec) & 0x80000000) != 0)
835                         extra_bits = 0;
836                 time->tv_sec += extra_bits << 32;
837 #else
838                 time->tv_sec += (u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK) << 32;
839 #endif
840         }
841         time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
842 }
843
844 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
845 do {                                                                           \
846         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
847         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
848                 (raw_inode)->xtime ## _extra =                                 \
849                                 ext4_encode_extra_time(&(inode)->xtime);       \
850 } while (0)
851
852 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
853 do {                                                                           \
854         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
855                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
856         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
857                 (raw_inode)->xtime ## _extra =                                 \
858                                 ext4_encode_extra_time(&(einode)->xtime);      \
859 } while (0)
860
861 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
862 do {                                                                           \
863         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
864         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
865                 ext4_decode_extra_time(&(inode)->xtime,                        \
866                                        raw_inode->xtime ## _extra);            \
867         else                                                                   \
868                 (inode)->xtime.tv_nsec = 0;                                    \
869 } while (0)
870
871 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
872 do {                                                                           \
873         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
874                 (einode)->xtime.tv_sec =                                       \
875                         (signed)le32_to_cpu((raw_inode)->xtime);               \
876         else                                                                   \
877                 (einode)->xtime.tv_sec = 0;                                    \
878         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
879                 ext4_decode_extra_time(&(einode)->xtime,                       \
880                                        raw_inode->xtime ## _extra);            \
881         else                                                                   \
882                 (einode)->xtime.tv_nsec = 0;                                   \
883 } while (0)
884
885 #define i_disk_version osd1.linux1.l_i_version
886
887 #if defined(__KERNEL__) || defined(__linux__)
888 #define i_reserved1     osd1.linux1.l_i_reserved1
889 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
890 #define i_blocks_high   osd2.linux2.l_i_blocks_high
891 #define i_uid_low       i_uid
892 #define i_gid_low       i_gid
893 #define i_uid_high      osd2.linux2.l_i_uid_high
894 #define i_gid_high      osd2.linux2.l_i_gid_high
895 #define i_checksum_lo   osd2.linux2.l_i_checksum_lo
896
897 #elif defined(__GNU__)
898
899 #define i_translator    osd1.hurd1.h_i_translator
900 #define i_uid_high      osd2.hurd2.h_i_uid_high
901 #define i_gid_high      osd2.hurd2.h_i_gid_high
902 #define i_author        osd2.hurd2.h_i_author
903
904 #elif defined(__masix__)
905
906 #define i_reserved1     osd1.masix1.m_i_reserved1
907 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
908 #define i_reserved2     osd2.masix2.m_i_reserved2
909
910 #endif /* defined(__KERNEL__) || defined(__linux__) */
911
912 #include "extents_status.h"
913
914 /*
915  * Lock subclasses for i_data_sem in the ext4_inode_info structure.
916  *
917  * These are needed to avoid lockdep false positives when we need to
918  * allocate blocks to the quota inode during ext4_map_blocks(), while
919  * holding i_data_sem for a normal (non-quota) inode.  Since we don't
920  * do quota tracking for the quota inode, this avoids deadlock (as
921  * well as infinite recursion, since it isn't turtles all the way
922  * down...)
923  *
924  *  I_DATA_SEM_NORMAL - Used for most inodes
925  *  I_DATA_SEM_OTHER  - Used by move_inode.c for the second normal inode
926  *                        where the second inode has larger inode number
927  *                        than the first
928  *  I_DATA_SEM_QUOTA  - Used for quota inodes only
929  */
930 enum {
931         I_DATA_SEM_NORMAL = 0,
932         I_DATA_SEM_OTHER,
933         I_DATA_SEM_QUOTA,
934 };
935
936
937 /*
938  * fourth extended file system inode data in memory
939  */
940 struct ext4_inode_info {
941         __le32  i_data[15];     /* unconverted */
942         __u32   i_dtime;
943         ext4_fsblk_t    i_file_acl;
944
945         /*
946          * i_block_group is the number of the block group which contains
947          * this file's inode.  Constant across the lifetime of the inode,
948          * it is ued for making block allocation decisions - we try to
949          * place a file's data blocks near its inode block, and new inodes
950          * near to their parent directory's inode.
951          */
952         ext4_group_t    i_block_group;
953         ext4_lblk_t     i_dir_start_lookup;
954 #if (BITS_PER_LONG < 64)
955         unsigned long   i_state_flags;          /* Dynamic state flags */
956 #endif
957         unsigned long   i_flags;
958
959         /*
960          * Extended attributes can be read independently of the main file
961          * data. Taking i_mutex even when reading would cause contention
962          * between readers of EAs and writers of regular file data, so
963          * instead we synchronize on xattr_sem when reading or changing
964          * EAs.
965          */
966         struct rw_semaphore xattr_sem;
967
968         struct list_head i_orphan;      /* unlinked but open inodes */
969
970         /*
971          * i_disksize keeps track of what the inode size is ON DISK, not
972          * in memory.  During truncate, i_size is set to the new size by
973          * the VFS prior to calling ext4_truncate(), but the filesystem won't
974          * set i_disksize to 0 until the truncate is actually under way.
975          *
976          * The intent is that i_disksize always represents the blocks which
977          * are used by this file.  This allows recovery to restart truncate
978          * on orphans if we crash during truncate.  We actually write i_disksize
979          * into the on-disk inode when writing inodes out, instead of i_size.
980          *
981          * The only time when i_disksize and i_size may be different is when
982          * a truncate is in progress.  The only things which change i_disksize
983          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
984          */
985         loff_t  i_disksize;
986
987         /*
988          * i_data_sem is for serialising ext4_truncate() against
989          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
990          * data tree are chopped off during truncate. We can't do that in
991          * ext4 because whenever we perform intermediate commits during
992          * truncate, the inode and all the metadata blocks *must* be in a
993          * consistent state which allows truncation of the orphans to restart
994          * during recovery.  Hence we must fix the get_block-vs-truncate race
995          * by other means, so we have i_data_sem.
996          */
997         struct rw_semaphore i_data_sem;
998         /*
999          * i_mmap_sem is for serializing page faults with truncate / punch hole
1000          * operations. We have to make sure that new page cannot be faulted in
1001          * a section of the inode that is being punched. We cannot easily use
1002          * i_data_sem for this since we need protection for the whole punch
1003          * operation and i_data_sem ranks below transaction start so we have
1004          * to occasionally drop it.
1005          */
1006         struct rw_semaphore i_mmap_sem;
1007         struct inode vfs_inode;
1008         struct jbd2_inode *jinode;
1009
1010         spinlock_t i_raw_lock;  /* protects updates to the raw inode */
1011
1012         /*
1013          * File creation time. Its function is same as that of
1014          * struct timespec i_{a,c,m}time in the generic inode.
1015          */
1016         struct timespec i_crtime;
1017
1018         /* mballoc */
1019         struct list_head i_prealloc_list;
1020         spinlock_t i_prealloc_lock;
1021
1022         /* extents status tree */
1023         struct ext4_es_tree i_es_tree;
1024         rwlock_t i_es_lock;
1025         struct list_head i_es_list;
1026         unsigned int i_es_all_nr;       /* protected by i_es_lock */
1027         unsigned int i_es_shk_nr;       /* protected by i_es_lock */
1028         ext4_lblk_t i_es_shrink_lblk;   /* Offset where we start searching for
1029                                            extents to shrink. Protected by
1030                                            i_es_lock  */
1031
1032         /* ialloc */
1033         ext4_group_t    i_last_alloc_group;
1034
1035         /* allocation reservation info for delalloc */
1036         /* In case of bigalloc, these refer to clusters rather than blocks */
1037         unsigned int i_reserved_data_blocks;
1038         unsigned int i_reserved_meta_blocks;
1039         unsigned int i_allocated_meta_blocks;
1040         ext4_lblk_t i_da_metadata_calc_last_lblock;
1041         int i_da_metadata_calc_len;
1042
1043         /* on-disk additional length */
1044         __u16 i_extra_isize;
1045
1046         /* Indicate the inline data space. */
1047         u16 i_inline_off;
1048         u16 i_inline_size;
1049
1050 #ifdef CONFIG_QUOTA
1051         /* quota space reservation, managed internally by quota code */
1052         qsize_t i_reserved_quota;
1053 #endif
1054
1055         /* Lock protecting lists below */
1056         spinlock_t i_completed_io_lock;
1057         /*
1058          * Completed IOs that need unwritten extents handling and have
1059          * transaction reserved
1060          */
1061         struct list_head i_rsv_conversion_list;
1062         struct work_struct i_rsv_conversion_work;
1063         atomic_t i_unwritten; /* Nr. of inflight conversions pending */
1064
1065         spinlock_t i_block_reservation_lock;
1066
1067         /*
1068          * Transactions that contain inode's metadata needed to complete
1069          * fsync and fdatasync, respectively.
1070          */
1071         tid_t i_sync_tid;
1072         tid_t i_datasync_tid;
1073
1074 #ifdef CONFIG_QUOTA
1075         struct dquot *i_dquot[MAXQUOTAS];
1076 #endif
1077
1078         /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
1079         __u32 i_csum_seed;
1080
1081         kprojid_t i_projid;
1082 };
1083
1084 /*
1085  * File system states
1086  */
1087 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
1088 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
1089 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
1090
1091 /*
1092  * Misc. filesystem flags
1093  */
1094 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
1095 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
1096 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
1097
1098 /*
1099  * Mount flags set via mount options or defaults
1100  */
1101 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
1102 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
1103 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
1104 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
1105 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
1106 #define EXT4_MOUNT_ERRORS_MASK          0x00070
1107 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
1108 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
1109 #ifdef CONFIG_FS_DAX
1110 #define EXT4_MOUNT_DAX                  0x00200 /* Direct Access */
1111 #else
1112 #define EXT4_MOUNT_DAX                  0
1113 #endif
1114 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
1115 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
1116 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
1117 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
1118 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
1119 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
1120 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
1121 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
1122 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
1123 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
1124 #define EXT4_MOUNT_QUOTA                0x40000 /* Some quota option set */
1125 #define EXT4_MOUNT_USRQUOTA             0x80000 /* "old" user quota,
1126                                                  * enable enforcement for hidden
1127                                                  * quota files */
1128 #define EXT4_MOUNT_GRPQUOTA             0x100000 /* "old" group quota, enable
1129                                                   * enforcement for hidden quota
1130                                                   * files */
1131 #define EXT4_MOUNT_PRJQUOTA             0x200000 /* Enable project quota
1132                                                   * enforcement */
1133 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
1134 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
1135 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
1136 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
1137 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
1138 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
1139 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
1140 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
1141
1142 /*
1143  * Mount flags set either automatically (could not be set by mount option)
1144  * based on per file system feature or property or in special cases such as
1145  * distinguishing between explicit mount option definition and default.
1146  */
1147 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
1148                                                       specified delalloc */
1149 #define EXT4_MOUNT2_STD_GROUP_SIZE      0x00000002 /* We have standard group
1150                                                       size of blocksize * 8
1151                                                       blocks */
1152 #define EXT4_MOUNT2_HURD_COMPAT         0x00000004 /* Support HURD-castrated
1153                                                       file systems */
1154
1155 #define EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM   0x00000008 /* User explicitly
1156                                                 specified journal checksum */
1157
1158 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
1159                                                 ~EXT4_MOUNT_##opt
1160 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
1161                                                 EXT4_MOUNT_##opt
1162 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
1163                                          EXT4_MOUNT_##opt)
1164
1165 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
1166                                                 ~EXT4_MOUNT2_##opt
1167 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
1168                                                 EXT4_MOUNT2_##opt
1169 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
1170                                          EXT4_MOUNT2_##opt)
1171
1172 #define ext4_test_and_set_bit           __test_and_set_bit_le
1173 #define ext4_set_bit                    __set_bit_le
1174 #define ext4_set_bit_atomic             ext2_set_bit_atomic
1175 #define ext4_test_and_clear_bit         __test_and_clear_bit_le
1176 #define ext4_clear_bit                  __clear_bit_le
1177 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
1178 #define ext4_test_bit                   test_bit_le
1179 #define ext4_find_next_zero_bit         find_next_zero_bit_le
1180 #define ext4_find_next_bit              find_next_bit_le
1181
1182 extern void ext4_set_bits(void *bm, int cur, int len);
1183
1184 /*
1185  * Maximal mount counts between two filesystem checks
1186  */
1187 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
1188 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
1189
1190 /*
1191  * Behaviour when detecting errors
1192  */
1193 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
1194 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
1195 #define EXT4_ERRORS_PANIC               3       /* Panic */
1196 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
1197
1198 /* Metadata checksum algorithm codes */
1199 #define EXT4_CRC32C_CHKSUM              1
1200
1201 /*
1202  * Structure of the super block
1203  */
1204 struct ext4_super_block {
1205 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1206         __le32  s_blocks_count_lo;      /* Blocks count */
1207         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1208         __le32  s_free_blocks_count_lo; /* Free blocks count */
1209 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1210         __le32  s_first_data_block;     /* First Data Block */
1211         __le32  s_log_block_size;       /* Block size */
1212         __le32  s_log_cluster_size;     /* Allocation cluster size */
1213 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1214         __le32  s_clusters_per_group;   /* # Clusters per group */
1215         __le32  s_inodes_per_group;     /* # Inodes per group */
1216         __le32  s_mtime;                /* Mount time */
1217 /*30*/  __le32  s_wtime;                /* Write time */
1218         __le16  s_mnt_count;            /* Mount count */
1219         __le16  s_max_mnt_count;        /* Maximal mount count */
1220         __le16  s_magic;                /* Magic signature */
1221         __le16  s_state;                /* File system state */
1222         __le16  s_errors;               /* Behaviour when detecting errors */
1223         __le16  s_minor_rev_level;      /* minor revision level */
1224 /*40*/  __le32  s_lastcheck;            /* time of last check */
1225         __le32  s_checkinterval;        /* max. time between checks */
1226         __le32  s_creator_os;           /* OS */
1227         __le32  s_rev_level;            /* Revision level */
1228 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1229         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1230         /*
1231          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1232          *
1233          * Note: the difference between the compatible feature set and
1234          * the incompatible feature set is that if there is a bit set
1235          * in the incompatible feature set that the kernel doesn't
1236          * know about, it should refuse to mount the filesystem.
1237          *
1238          * e2fsck's requirements are more strict; if it doesn't know
1239          * about a feature in either the compatible or incompatible
1240          * feature set, it must abort and not try to meddle with
1241          * things it doesn't understand...
1242          */
1243         __le32  s_first_ino;            /* First non-reserved inode */
1244         __le16  s_inode_size;           /* size of inode structure */
1245         __le16  s_block_group_nr;       /* block group # of this superblock */
1246         __le32  s_feature_compat;       /* compatible feature set */
1247 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1248         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1249 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1250 /*78*/  char    s_volume_name[16];      /* volume name */
1251 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1252 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1253         /*
1254          * Performance hints.  Directory preallocation should only
1255          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1256          */
1257         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1258         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1259         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1260         /*
1261          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1262          */
1263 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1264 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1265         __le32  s_journal_dev;          /* device number of journal file */
1266         __le32  s_last_orphan;          /* start of list of inodes to delete */
1267         __le32  s_hash_seed[4];         /* HTREE hash seed */
1268         __u8    s_def_hash_version;     /* Default hash version to use */
1269         __u8    s_jnl_backup_type;
1270         __le16  s_desc_size;            /* size of group descriptor */
1271 /*100*/ __le32  s_default_mount_opts;
1272         __le32  s_first_meta_bg;        /* First metablock block group */
1273         __le32  s_mkfs_time;            /* When the filesystem was created */
1274         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1275         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1276 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1277         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1278         __le32  s_free_blocks_count_hi; /* Free blocks count */
1279         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1280         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1281         __le32  s_flags;                /* Miscellaneous flags */
1282         __le16  s_raid_stride;          /* RAID stride */
1283         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1284         __le64  s_mmp_block;            /* Block for multi-mount protection */
1285         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1286         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1287         __u8    s_checksum_type;        /* metadata checksum algorithm used */
1288         __u8    s_encryption_level;     /* versioning level for encryption */
1289         __u8    s_reserved_pad;         /* Padding to next 32bits */
1290         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1291         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1292         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1293         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1294                                               snapshot's future use */
1295         __le32  s_snapshot_list;        /* inode number of the head of the
1296                                            on-disk snapshot list */
1297 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1298         __le32  s_error_count;          /* number of fs errors */
1299         __le32  s_first_error_time;     /* first time an error happened */
1300         __le32  s_first_error_ino;      /* inode involved in first error */
1301         __le64  s_first_error_block;    /* block involved of first error */
1302         __u8    s_first_error_func[32]; /* function where the error happened */
1303         __le32  s_first_error_line;     /* line number where error happened */
1304         __le32  s_last_error_time;      /* most recent time of an error */
1305         __le32  s_last_error_ino;       /* inode involved in last error */
1306         __le32  s_last_error_line;      /* line number where error happened */
1307         __le64  s_last_error_block;     /* block involved of last error */
1308         __u8    s_last_error_func[32];  /* function where the error happened */
1309 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1310         __u8    s_mount_opts[64];
1311         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1312         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1313         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1314         __le32  s_backup_bgs[2];        /* groups with sparse_super2 SBs */
1315         __u8    s_encrypt_algos[4];     /* Encryption algorithms in use  */
1316         __u8    s_encrypt_pw_salt[16];  /* Salt used for string2key algorithm */
1317         __le32  s_lpf_ino;              /* Location of the lost+found inode */
1318         __le32  s_prj_quota_inum;       /* inode for tracking project quota */
1319         __le32  s_checksum_seed;        /* crc32c(uuid) if csum_seed set */
1320         __le32  s_reserved[98];         /* Padding to the end of the block */
1321         __le32  s_checksum;             /* crc32c(superblock) */
1322 };
1323
1324 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1325
1326 #ifdef __KERNEL__
1327
1328 /*
1329  * run-time mount flags
1330  */
1331 #define EXT4_MF_MNTDIR_SAMPLED          0x0001
1332 #define EXT4_MF_FS_ABORTED              0x0002  /* Fatal error detected */
1333 #define EXT4_MF_TEST_DUMMY_ENCRYPTION   0x0004
1334
1335 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1336 #define DUMMY_ENCRYPTION_ENABLED(sbi) (unlikely((sbi)->s_mount_flags & \
1337                                                 EXT4_MF_TEST_DUMMY_ENCRYPTION))
1338 #else
1339 #define DUMMY_ENCRYPTION_ENABLED(sbi) (0)
1340 #endif
1341
1342 /* Number of quota types we support */
1343 #define EXT4_MAXQUOTAS 3
1344
1345 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1346 #define EXT4_KEY_DESC_PREFIX "ext4:"
1347 #define EXT4_KEY_DESC_PREFIX_SIZE 5
1348 #endif
1349
1350 /*
1351  * fourth extended-fs super-block data in memory
1352  */
1353 struct ext4_sb_info {
1354         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1355         unsigned long s_inodes_per_block;/* Number of inodes per block */
1356         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1357         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1358         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1359         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1360         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1361         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1362         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1363         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1364         unsigned long s_overhead;  /* # of fs overhead clusters */
1365         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1366         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1367         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1368         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1369         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1370         struct buffer_head * __rcu *s_group_desc;
1371         unsigned int s_mount_opt;
1372         unsigned int s_mount_opt2;
1373         unsigned int s_mount_flags;
1374         unsigned int s_def_mount_opt;
1375         ext4_fsblk_t s_sb_block;
1376         atomic64_t s_resv_clusters;
1377         kuid_t s_resuid;
1378         kgid_t s_resgid;
1379         unsigned short s_mount_state;
1380         unsigned short s_pad;
1381         int s_addr_per_block_bits;
1382         int s_desc_per_block_bits;
1383         int s_inode_size;
1384         int s_first_ino;
1385         unsigned int s_inode_readahead_blks;
1386         unsigned int s_inode_goal;
1387         spinlock_t s_next_gen_lock;
1388         u32 s_next_generation;
1389         u32 s_hash_seed[4];
1390         int s_def_hash_version;
1391         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1392         struct percpu_counter s_freeclusters_counter;
1393         struct percpu_counter s_freeinodes_counter;
1394         struct percpu_counter s_dirs_counter;
1395         struct percpu_counter s_dirtyclusters_counter;
1396         struct blockgroup_lock *s_blockgroup_lock;
1397         struct proc_dir_entry *s_proc;
1398         struct kobject s_kobj;
1399         struct completion s_kobj_unregister;
1400         struct super_block *s_sb;
1401
1402         /* Journaling */
1403         struct journal_s *s_journal;
1404         struct list_head s_orphan;
1405         struct mutex s_orphan_lock;
1406         unsigned long s_resize_flags;           /* Flags indicating if there
1407                                                    is a resizer */
1408         unsigned long s_commit_interval;
1409         u32 s_max_batch_time;
1410         u32 s_min_batch_time;
1411         struct block_device *journal_bdev;
1412 #ifdef CONFIG_QUOTA
1413         char *s_qf_names[EXT4_MAXQUOTAS];       /* Names of quota files with journalled quota */
1414         int s_jquota_fmt;                       /* Format of quota to use */
1415 #endif
1416         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1417         struct rb_root system_blks;
1418
1419 #ifdef EXTENTS_STATS
1420         /* ext4 extents stats */
1421         unsigned long s_ext_min;
1422         unsigned long s_ext_max;
1423         unsigned long s_depth_max;
1424         spinlock_t s_ext_stats_lock;
1425         unsigned long s_ext_blocks;
1426         unsigned long s_ext_extents;
1427 #endif
1428
1429         /* for buddy allocator */
1430         struct ext4_group_info ** __rcu *s_group_info;
1431         struct inode *s_buddy_cache;
1432         spinlock_t s_md_lock;
1433         unsigned short *s_mb_offsets;
1434         unsigned int *s_mb_maxs;
1435         unsigned int s_group_info_size;
1436         unsigned int s_mb_free_pending;
1437
1438         /* tunables */
1439         unsigned long s_stripe;
1440         unsigned int s_mb_stream_request;
1441         unsigned int s_mb_max_to_scan;
1442         unsigned int s_mb_min_to_scan;
1443         unsigned int s_mb_stats;
1444         unsigned int s_mb_order2_reqs;
1445         unsigned int s_mb_group_prealloc;
1446         unsigned int s_max_dir_size_kb;
1447         /* where last allocation was done - for stream allocation */
1448         unsigned long s_mb_last_group;
1449         unsigned long s_mb_last_start;
1450
1451         /* stats for buddy allocator */
1452         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1453         atomic_t s_bal_success; /* we found long enough chunks */
1454         atomic_t s_bal_allocated;       /* in blocks */
1455         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1456         atomic_t s_bal_goals;   /* goal hits */
1457         atomic_t s_bal_breaks;  /* too long searches */
1458         atomic_t s_bal_2orders; /* 2^order hits */
1459         spinlock_t s_bal_lock;
1460         unsigned long s_mb_buddies_generated;
1461         unsigned long long s_mb_generation_time;
1462         atomic_t s_mb_lost_chunks;
1463         atomic_t s_mb_preallocated;
1464         atomic_t s_mb_discarded;
1465         atomic_t s_lock_busy;
1466
1467         /* locality groups */
1468         struct ext4_locality_group __percpu *s_locality_groups;
1469
1470         /* for write statistics */
1471         unsigned long s_sectors_written_start;
1472         u64 s_kbytes_written;
1473
1474         /* the size of zero-out chunk */
1475         unsigned int s_extent_max_zeroout_kb;
1476
1477         unsigned int s_log_groups_per_flex;
1478         struct flex_groups * __rcu *s_flex_groups;
1479         ext4_group_t s_flex_groups_allocated;
1480
1481         /* workqueue for reserved extent conversions (buffered io) */
1482         struct workqueue_struct *rsv_conversion_wq;
1483
1484         /* timer for periodic error stats printing */
1485         struct timer_list s_err_report;
1486
1487         /* Lazy inode table initialization info */
1488         struct ext4_li_request *s_li_request;
1489         /* Wait multiplier for lazy initialization thread */
1490         unsigned int s_li_wait_mult;
1491
1492         /* Kernel thread for multiple mount protection */
1493         struct task_struct *s_mmp_tsk;
1494
1495         /* record the last minlen when FITRIM is called. */
1496         atomic_t s_last_trim_minblks;
1497
1498         /* Reference to checksum algorithm driver via cryptoapi */
1499         struct crypto_shash *s_chksum_driver;
1500
1501         /* Precomputed FS UUID checksum for seeding other checksums */
1502         __u32 s_csum_seed;
1503
1504         /* Reclaim extents from extent status tree */
1505         struct shrinker s_es_shrinker;
1506         struct list_head s_es_list;     /* List of inodes with reclaimable extents */
1507         long s_es_nr_inode;
1508         struct ext4_es_stats s_es_stats;
1509         struct mb_cache *s_mb_cache;
1510         spinlock_t s_es_lock ____cacheline_aligned_in_smp;
1511
1512         /* Ratelimit ext4 messages. */
1513         struct ratelimit_state s_err_ratelimit_state;
1514         struct ratelimit_state s_warning_ratelimit_state;
1515         struct ratelimit_state s_msg_ratelimit_state;
1516
1517         /*
1518          * Barrier between writepages ops and changing any inode's JOURNAL_DATA
1519          * or EXTENTS flag.
1520          */
1521         struct percpu_rw_semaphore s_writepages_rwsem;
1522
1523         /* Encryption support */
1524 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1525         u8 key_prefix[EXT4_KEY_DESC_PREFIX_SIZE];
1526         u8 key_prefix_size;
1527 #endif
1528 };
1529
1530 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1531 {
1532         return sb->s_fs_info;
1533 }
1534 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1535 {
1536         return container_of(inode, struct ext4_inode_info, vfs_inode);
1537 }
1538
1539 static inline struct timespec ext4_current_time(struct inode *inode)
1540 {
1541         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1542                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1543 }
1544
1545 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1546 {
1547         return ino == EXT4_ROOT_INO ||
1548                 (ino >= EXT4_FIRST_INO(sb) &&
1549                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1550 }
1551
1552 /*
1553  * Returns: sbi->field[index]
1554  * Used to access an array element from the following sbi fields which require
1555  * rcu protection to avoid dereferencing an invalid pointer due to reassignment
1556  * - s_group_desc
1557  * - s_group_info
1558  * - s_flex_group
1559  */
1560 #define sbi_array_rcu_deref(sbi, field, index)                             \
1561 ({                                                                         \
1562         typeof(*((sbi)->field)) _v;                                        \
1563         rcu_read_lock();                                                   \
1564         _v = ((typeof(_v)*)rcu_dereference((sbi)->field))[index];          \
1565         rcu_read_unlock();                                                 \
1566         _v;                                                                \
1567 })
1568
1569 /*
1570  * Inode dynamic state flags
1571  */
1572 enum {
1573         EXT4_STATE_JDATA,               /* journaled data exists */
1574         EXT4_STATE_NEW,                 /* inode is newly created */
1575         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1576         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1577         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1578         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1579         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1580         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1581         EXT4_STATE_DIOREAD_LOCK,        /* Disable support for dio read
1582                                            nolocking */
1583         EXT4_STATE_MAY_INLINE_DATA,     /* may have in-inode data */
1584         EXT4_STATE_EXT_PRECACHED,       /* extents have been precached */
1585 };
1586
1587 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1588 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1589 {                                                                       \
1590         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1591 }                                                                       \
1592 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1593 {                                                                       \
1594         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1595 }                                                                       \
1596 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1597 {                                                                       \
1598         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1599 }
1600
1601 /* Add these declarations here only so that these functions can be
1602  * found by name.  Otherwise, they are very hard to locate. */
1603 static inline int ext4_test_inode_flag(struct inode *inode, int bit);
1604 static inline void ext4_set_inode_flag(struct inode *inode, int bit);
1605 static inline void ext4_clear_inode_flag(struct inode *inode, int bit);
1606 EXT4_INODE_BIT_FNS(flag, flags, 0)
1607
1608 /* Add these declarations here only so that these functions can be
1609  * found by name.  Otherwise, they are very hard to locate. */
1610 static inline int ext4_test_inode_state(struct inode *inode, int bit);
1611 static inline void ext4_set_inode_state(struct inode *inode, int bit);
1612 static inline void ext4_clear_inode_state(struct inode *inode, int bit);
1613 #if (BITS_PER_LONG < 64)
1614 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1615
1616 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1617 {
1618         (ei)->i_state_flags = 0;
1619 }
1620 #else
1621 EXT4_INODE_BIT_FNS(state, flags, 32)
1622
1623 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1624 {
1625         /* We depend on the fact that callers will set i_flags */
1626 }
1627 #endif
1628 #else
1629 /* Assume that user mode programs are passing in an ext4fs superblock, not
1630  * a kernel struct super_block.  This will allow us to call the feature-test
1631  * macros from user land. */
1632 #define EXT4_SB(sb)     (sb)
1633 #endif
1634
1635 /*
1636  * Returns true if the inode is inode is encrypted
1637  */
1638 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1639
1640 /*
1641  * Codes for operating systems
1642  */
1643 #define EXT4_OS_LINUX           0
1644 #define EXT4_OS_HURD            1
1645 #define EXT4_OS_MASIX           2
1646 #define EXT4_OS_FREEBSD         3
1647 #define EXT4_OS_LITES           4
1648
1649 /*
1650  * Revision levels
1651  */
1652 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1653 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1654
1655 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1656 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1657
1658 #define EXT4_GOOD_OLD_INODE_SIZE 128
1659
1660 /*
1661  * Feature set definitions
1662  */
1663
1664 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1665 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1666 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1667 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1668 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1669 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1670 #define EXT4_FEATURE_COMPAT_SPARSE_SUPER2       0x0200
1671
1672 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1673 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1674 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1675 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1676 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1677 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1678 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1679 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1680 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1681 /*
1682  * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM).  When
1683  * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
1684  * all other data structures' checksums.  However, the METADATA_CSUM and
1685  * GDT_CSUM bits are mutually exclusive.
1686  */
1687 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM    0x0400
1688 #define EXT4_FEATURE_RO_COMPAT_READONLY         0x1000
1689 #define EXT4_FEATURE_RO_COMPAT_PROJECT          0x2000
1690
1691 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1692 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1693 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1694 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1695 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1696 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1697 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1698 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1699 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1700 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1701 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1702 #define EXT4_FEATURE_INCOMPAT_CSUM_SEED         0x2000
1703 #define EXT4_FEATURE_INCOMPAT_LARGEDIR          0x4000 /* >2GB or 3-lvl htree */
1704 #define EXT4_FEATURE_INCOMPAT_INLINE_DATA       0x8000 /* data in inode */
1705 #define EXT4_FEATURE_INCOMPAT_ENCRYPT           0x10000
1706
1707 #define EXT4_FEATURE_COMPAT_FUNCS(name, flagname) \
1708 static inline bool ext4_has_feature_##name(struct super_block *sb) \
1709 { \
1710         return ((EXT4_SB(sb)->s_es->s_feature_compat & \
1711                 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname)) != 0); \
1712 } \
1713 static inline void ext4_set_feature_##name(struct super_block *sb) \
1714 { \
1715         EXT4_SB(sb)->s_es->s_feature_compat |= \
1716                 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
1717 } \
1718 static inline void ext4_clear_feature_##name(struct super_block *sb) \
1719 { \
1720         EXT4_SB(sb)->s_es->s_feature_compat &= \
1721                 ~cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
1722 }
1723
1724 #define EXT4_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
1725 static inline bool ext4_has_feature_##name(struct super_block *sb) \
1726 { \
1727         return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
1728                 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname)) != 0); \
1729 } \
1730 static inline void ext4_set_feature_##name(struct super_block *sb) \
1731 { \
1732         EXT4_SB(sb)->s_es->s_feature_ro_compat |= \
1733                 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
1734 } \
1735 static inline void ext4_clear_feature_##name(struct super_block *sb) \
1736 { \
1737         EXT4_SB(sb)->s_es->s_feature_ro_compat &= \
1738                 ~cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
1739 }
1740
1741 #define EXT4_FEATURE_INCOMPAT_FUNCS(name, flagname) \
1742 static inline bool ext4_has_feature_##name(struct super_block *sb) \
1743 { \
1744         return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
1745                 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname)) != 0); \
1746 } \
1747 static inline void ext4_set_feature_##name(struct super_block *sb) \
1748 { \
1749         EXT4_SB(sb)->s_es->s_feature_incompat |= \
1750                 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
1751 } \
1752 static inline void ext4_clear_feature_##name(struct super_block *sb) \
1753 { \
1754         EXT4_SB(sb)->s_es->s_feature_incompat &= \
1755                 ~cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
1756 }
1757
1758 EXT4_FEATURE_COMPAT_FUNCS(dir_prealloc,         DIR_PREALLOC)
1759 EXT4_FEATURE_COMPAT_FUNCS(imagic_inodes,        IMAGIC_INODES)
1760 EXT4_FEATURE_COMPAT_FUNCS(journal,              HAS_JOURNAL)
1761 EXT4_FEATURE_COMPAT_FUNCS(xattr,                EXT_ATTR)
1762 EXT4_FEATURE_COMPAT_FUNCS(resize_inode,         RESIZE_INODE)
1763 EXT4_FEATURE_COMPAT_FUNCS(dir_index,            DIR_INDEX)
1764 EXT4_FEATURE_COMPAT_FUNCS(sparse_super2,        SPARSE_SUPER2)
1765
1766 EXT4_FEATURE_RO_COMPAT_FUNCS(sparse_super,      SPARSE_SUPER)
1767 EXT4_FEATURE_RO_COMPAT_FUNCS(large_file,        LARGE_FILE)
1768 EXT4_FEATURE_RO_COMPAT_FUNCS(btree_dir,         BTREE_DIR)
1769 EXT4_FEATURE_RO_COMPAT_FUNCS(huge_file,         HUGE_FILE)
1770 EXT4_FEATURE_RO_COMPAT_FUNCS(gdt_csum,          GDT_CSUM)
1771 EXT4_FEATURE_RO_COMPAT_FUNCS(dir_nlink,         DIR_NLINK)
1772 EXT4_FEATURE_RO_COMPAT_FUNCS(extra_isize,       EXTRA_ISIZE)
1773 EXT4_FEATURE_RO_COMPAT_FUNCS(quota,             QUOTA)
1774 EXT4_FEATURE_RO_COMPAT_FUNCS(bigalloc,          BIGALLOC)
1775 EXT4_FEATURE_RO_COMPAT_FUNCS(metadata_csum,     METADATA_CSUM)
1776 EXT4_FEATURE_RO_COMPAT_FUNCS(readonly,          READONLY)
1777 EXT4_FEATURE_RO_COMPAT_FUNCS(project,           PROJECT)
1778
1779 EXT4_FEATURE_INCOMPAT_FUNCS(compression,        COMPRESSION)
1780 EXT4_FEATURE_INCOMPAT_FUNCS(filetype,           FILETYPE)
1781 EXT4_FEATURE_INCOMPAT_FUNCS(journal_needs_recovery,     RECOVER)
1782 EXT4_FEATURE_INCOMPAT_FUNCS(journal_dev,        JOURNAL_DEV)
1783 EXT4_FEATURE_INCOMPAT_FUNCS(meta_bg,            META_BG)
1784 EXT4_FEATURE_INCOMPAT_FUNCS(extents,            EXTENTS)
1785 EXT4_FEATURE_INCOMPAT_FUNCS(64bit,              64BIT)
1786 EXT4_FEATURE_INCOMPAT_FUNCS(mmp,                MMP)
1787 EXT4_FEATURE_INCOMPAT_FUNCS(flex_bg,            FLEX_BG)
1788 EXT4_FEATURE_INCOMPAT_FUNCS(ea_inode,           EA_INODE)
1789 EXT4_FEATURE_INCOMPAT_FUNCS(dirdata,            DIRDATA)
1790 EXT4_FEATURE_INCOMPAT_FUNCS(csum_seed,          CSUM_SEED)
1791 EXT4_FEATURE_INCOMPAT_FUNCS(largedir,           LARGEDIR)
1792 EXT4_FEATURE_INCOMPAT_FUNCS(inline_data,        INLINE_DATA)
1793 EXT4_FEATURE_INCOMPAT_FUNCS(encrypt,            ENCRYPT)
1794
1795 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1796 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1797                                          EXT4_FEATURE_INCOMPAT_META_BG)
1798 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1799                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1800                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1801
1802 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1803 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1804                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1805                                          EXT4_FEATURE_INCOMPAT_META_BG)
1806 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1807                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1808                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1809
1810 #define EXT4_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1811 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1812                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1813                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1814                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1815                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1816                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1817                                          EXT4_FEATURE_INCOMPAT_MMP | \
1818                                          EXT4_FEATURE_INCOMPAT_INLINE_DATA | \
1819                                          EXT4_FEATURE_INCOMPAT_ENCRYPT | \
1820                                          EXT4_FEATURE_INCOMPAT_CSUM_SEED)
1821 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1822                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1823                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1824                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1825                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1826                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1827                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1828                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
1829                                          EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
1830                                          EXT4_FEATURE_RO_COMPAT_QUOTA |\
1831                                          EXT4_FEATURE_RO_COMPAT_PROJECT)
1832
1833 #define EXTN_FEATURE_FUNCS(ver) \
1834 static inline bool ext4_has_unknown_ext##ver##_compat_features(struct super_block *sb) \
1835 { \
1836         return ((EXT4_SB(sb)->s_es->s_feature_compat & \
1837                 cpu_to_le32(~EXT##ver##_FEATURE_COMPAT_SUPP)) != 0); \
1838 } \
1839 static inline bool ext4_has_unknown_ext##ver##_ro_compat_features(struct super_block *sb) \
1840 { \
1841         return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
1842                 cpu_to_le32(~EXT##ver##_FEATURE_RO_COMPAT_SUPP)) != 0); \
1843 } \
1844 static inline bool ext4_has_unknown_ext##ver##_incompat_features(struct super_block *sb) \
1845 { \
1846         return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
1847                 cpu_to_le32(~EXT##ver##_FEATURE_INCOMPAT_SUPP)) != 0); \
1848 }
1849
1850 EXTN_FEATURE_FUNCS(2)
1851 EXTN_FEATURE_FUNCS(3)
1852 EXTN_FEATURE_FUNCS(4)
1853
1854 static inline bool ext4_has_compat_features(struct super_block *sb)
1855 {
1856         return (EXT4_SB(sb)->s_es->s_feature_compat != 0);
1857 }
1858 static inline bool ext4_has_ro_compat_features(struct super_block *sb)
1859 {
1860         return (EXT4_SB(sb)->s_es->s_feature_ro_compat != 0);
1861 }
1862 static inline bool ext4_has_incompat_features(struct super_block *sb)
1863 {
1864         return (EXT4_SB(sb)->s_es->s_feature_incompat != 0);
1865 }
1866
1867 /*
1868  * Default values for user and/or group using reserved blocks
1869  */
1870 #define EXT4_DEF_RESUID         0
1871 #define EXT4_DEF_RESGID         0
1872
1873 /*
1874  * Default project ID
1875  */
1876 #define EXT4_DEF_PROJID         0
1877
1878 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1879
1880 /*
1881  * Default mount options
1882  */
1883 #define EXT4_DEFM_DEBUG         0x0001
1884 #define EXT4_DEFM_BSDGROUPS     0x0002
1885 #define EXT4_DEFM_XATTR_USER    0x0004
1886 #define EXT4_DEFM_ACL           0x0008
1887 #define EXT4_DEFM_UID16         0x0010
1888 #define EXT4_DEFM_JMODE         0x0060
1889 #define EXT4_DEFM_JMODE_DATA    0x0020
1890 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1891 #define EXT4_DEFM_JMODE_WBACK   0x0060
1892 #define EXT4_DEFM_NOBARRIER     0x0100
1893 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1894 #define EXT4_DEFM_DISCARD       0x0400
1895 #define EXT4_DEFM_NODELALLOC    0x0800
1896
1897 /*
1898  * Default journal batch times
1899  */
1900 #define EXT4_DEF_MIN_BATCH_TIME 0
1901 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1902
1903 /*
1904  * Minimum number of groups in a flexgroup before we separate out
1905  * directories into the first block group of a flexgroup
1906  */
1907 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1908
1909 /*
1910  * Structure of a directory entry
1911  */
1912 #define EXT4_NAME_LEN 255
1913
1914 struct ext4_dir_entry {
1915         __le32  inode;                  /* Inode number */
1916         __le16  rec_len;                /* Directory entry length */
1917         __le16  name_len;               /* Name length */
1918         char    name[EXT4_NAME_LEN];    /* File name */
1919 };
1920
1921 /*
1922  * The new version of the directory entry.  Since EXT4 structures are
1923  * stored in intel byte order, and the name_len field could never be
1924  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1925  * file_type field.
1926  */
1927 struct ext4_dir_entry_2 {
1928         __le32  inode;                  /* Inode number */
1929         __le16  rec_len;                /* Directory entry length */
1930         __u8    name_len;               /* Name length */
1931         __u8    file_type;
1932         char    name[EXT4_NAME_LEN];    /* File name */
1933 };
1934
1935 /*
1936  * This is a bogus directory entry at the end of each leaf block that
1937  * records checksums.
1938  */
1939 struct ext4_dir_entry_tail {
1940         __le32  det_reserved_zero1;     /* Pretend to be unused */
1941         __le16  det_rec_len;            /* 12 */
1942         __u8    det_reserved_zero2;     /* Zero name length */
1943         __u8    det_reserved_ft;        /* 0xDE, fake file type */
1944         __le32  det_checksum;           /* crc32c(uuid+inum+dirblock) */
1945 };
1946
1947 #define EXT4_DIRENT_TAIL(block, blocksize) \
1948         ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
1949                                         ((blocksize) - \
1950                                          sizeof(struct ext4_dir_entry_tail))))
1951
1952 /*
1953  * Ext4 directory file types.  Only the low 3 bits are used.  The
1954  * other bits are reserved for now.
1955  */
1956 #define EXT4_FT_UNKNOWN         0
1957 #define EXT4_FT_REG_FILE        1
1958 #define EXT4_FT_DIR             2
1959 #define EXT4_FT_CHRDEV          3
1960 #define EXT4_FT_BLKDEV          4
1961 #define EXT4_FT_FIFO            5
1962 #define EXT4_FT_SOCK            6
1963 #define EXT4_FT_SYMLINK         7
1964
1965 #define EXT4_FT_MAX             8
1966
1967 #define EXT4_FT_DIR_CSUM        0xDE
1968
1969 /*
1970  * EXT4_DIR_PAD defines the directory entries boundaries
1971  *
1972  * NOTE: It must be a multiple of 4
1973  */
1974 #define EXT4_DIR_PAD                    4
1975 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1976 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1977                                          ~EXT4_DIR_ROUND)
1978 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1979
1980 /*
1981  * If we ever get support for fs block sizes > page_size, we'll need
1982  * to remove the #if statements in the next two functions...
1983  */
1984 static inline unsigned int
1985 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1986 {
1987         unsigned len = le16_to_cpu(dlen);
1988
1989 #if (PAGE_SIZE >= 65536)
1990         if (len == EXT4_MAX_REC_LEN || len == 0)
1991                 return blocksize;
1992         return (len & 65532) | ((len & 3) << 16);
1993 #else
1994         return len;
1995 #endif
1996 }
1997
1998 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1999 {
2000         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
2001                 BUG();
2002 #if (PAGE_SIZE >= 65536)
2003         if (len < 65536)
2004                 return cpu_to_le16(len);
2005         if (len == blocksize) {
2006                 if (blocksize == 65536)
2007                         return cpu_to_le16(EXT4_MAX_REC_LEN);
2008                 else
2009                         return cpu_to_le16(0);
2010         }
2011         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
2012 #else
2013         return cpu_to_le16(len);
2014 #endif
2015 }
2016
2017 /*
2018  * Hash Tree Directory indexing
2019  * (c) Daniel Phillips, 2001
2020  */
2021
2022 #define is_dx(dir) (ext4_has_feature_dir_index((dir)->i_sb) && \
2023                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
2024 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
2025 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
2026
2027 /* Legal values for the dx_root hash_version field: */
2028
2029 #define DX_HASH_LEGACY          0
2030 #define DX_HASH_HALF_MD4        1
2031 #define DX_HASH_TEA             2
2032 #define DX_HASH_LEGACY_UNSIGNED 3
2033 #define DX_HASH_HALF_MD4_UNSIGNED       4
2034 #define DX_HASH_TEA_UNSIGNED            5
2035
2036 static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
2037                               const void *address, unsigned int length)
2038 {
2039         struct {
2040                 struct shash_desc shash;
2041                 char ctx[4];
2042         } desc;
2043         int err;
2044
2045         BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
2046
2047         desc.shash.tfm = sbi->s_chksum_driver;
2048         desc.shash.flags = 0;
2049         *(u32 *)desc.ctx = crc;
2050
2051         err = crypto_shash_update(&desc.shash, address, length);
2052         BUG_ON(err);
2053
2054         return *(u32 *)desc.ctx;
2055 }
2056
2057 #ifdef __KERNEL__
2058
2059 /* hash info structure used by the directory hash */
2060 struct dx_hash_info
2061 {
2062         u32             hash;
2063         u32             minor_hash;
2064         int             hash_version;
2065         u32             *seed;
2066 };
2067
2068
2069 /* 32 and 64 bit signed EOF for dx directories */
2070 #define EXT4_HTREE_EOF_32BIT   ((1UL  << (32 - 1)) - 1)
2071 #define EXT4_HTREE_EOF_64BIT   ((1ULL << (64 - 1)) - 1)
2072
2073
2074 /*
2075  * Control parameters used by ext4_htree_next_block
2076  */
2077 #define HASH_NB_ALWAYS          1
2078
2079 struct ext4_filename {
2080         const struct qstr *usr_fname;
2081         struct fscrypt_str disk_name;
2082         struct dx_hash_info hinfo;
2083 #ifdef CONFIG_EXT4_FS_ENCRYPTION
2084         struct fscrypt_str crypto_buf;
2085 #endif
2086 };
2087
2088 #define fname_name(p) ((p)->disk_name.name)
2089 #define fname_len(p)  ((p)->disk_name.len)
2090
2091 /*
2092  * Describe an inode's exact location on disk and in memory
2093  */
2094 struct ext4_iloc
2095 {
2096         struct buffer_head *bh;
2097         unsigned long offset;
2098         ext4_group_t block_group;
2099 };
2100
2101 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
2102 {
2103         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
2104 }
2105
2106 /*
2107  * This structure is stuffed into the struct file's private_data field
2108  * for directories.  It is where we put information so that we can do
2109  * readdir operations in hash tree order.
2110  */
2111 struct dir_private_info {
2112         struct rb_root  root;
2113         struct rb_node  *curr_node;
2114         struct fname    *extra_fname;
2115         loff_t          last_pos;
2116         __u32           curr_hash;
2117         __u32           curr_minor_hash;
2118         __u32           next_hash;
2119 };
2120
2121 /* calculate the first block number of the group */
2122 static inline ext4_fsblk_t
2123 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
2124 {
2125         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
2126                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
2127 }
2128
2129 /*
2130  * Special error return code only used by dx_probe() and its callers.
2131  */
2132 #define ERR_BAD_DX_DIR  (-(MAX_ERRNO - 1))
2133
2134 /*
2135  * Timeout and state flag for lazy initialization inode thread.
2136  */
2137 #define EXT4_DEF_LI_WAIT_MULT                   10
2138 #define EXT4_DEF_LI_MAX_START_DELAY             5
2139 #define EXT4_LAZYINIT_QUIT                      0x0001
2140 #define EXT4_LAZYINIT_RUNNING                   0x0002
2141
2142 /*
2143  * Lazy inode table initialization info
2144  */
2145 struct ext4_lazy_init {
2146         unsigned long           li_state;
2147         struct list_head        li_request_list;
2148         struct mutex            li_list_mtx;
2149 };
2150
2151 struct ext4_li_request {
2152         struct super_block      *lr_super;
2153         struct ext4_sb_info     *lr_sbi;
2154         ext4_group_t            lr_next_group;
2155         struct list_head        lr_request;
2156         unsigned long           lr_next_sched;
2157         unsigned long           lr_timeout;
2158 };
2159
2160 struct ext4_features {
2161         struct kobject f_kobj;
2162         struct completion f_kobj_unregister;
2163 };
2164
2165 /*
2166  * This structure will be used for multiple mount protection. It will be
2167  * written into the block number saved in the s_mmp_block field in the
2168  * superblock. Programs that check MMP should assume that if
2169  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
2170  * to use the filesystem, regardless of how old the timestamp is.
2171  */
2172 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
2173 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
2174 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
2175 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
2176
2177 struct mmp_struct {
2178         __le32  mmp_magic;              /* Magic number for MMP */
2179         __le32  mmp_seq;                /* Sequence no. updated periodically */
2180
2181         /*
2182          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
2183          * purposes and do not affect the correctness of the algorithm
2184          */
2185         __le64  mmp_time;               /* Time last updated */
2186         char    mmp_nodename[64];       /* Node which last updated MMP block */
2187         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
2188
2189         /*
2190          * mmp_check_interval is used to verify if the MMP block has been
2191          * updated on the block device. The value is updated based on the
2192          * maximum time to write the MMP block during an update cycle.
2193          */
2194         __le16  mmp_check_interval;
2195
2196         __le16  mmp_pad1;
2197         __le32  mmp_pad2[226];
2198         __le32  mmp_checksum;           /* crc32c(uuid+mmp_block) */
2199 };
2200
2201 /* arguments passed to the mmp thread */
2202 struct mmpd_data {
2203         struct buffer_head *bh; /* bh from initial read_mmp_block() */
2204         struct super_block *sb;  /* super block of the fs */
2205 };
2206
2207 /*
2208  * Check interval multiplier
2209  * The MMP block is written every update interval and initially checked every
2210  * update interval x the multiplier (the value is then adapted based on the
2211  * write latency). The reason is that writes can be delayed under load and we
2212  * don't want readers to incorrectly assume that the filesystem is no longer
2213  * in use.
2214  */
2215 #define EXT4_MMP_CHECK_MULT             2UL
2216
2217 /*
2218  * Minimum interval for MMP checking in seconds.
2219  */
2220 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
2221
2222 /*
2223  * Maximum interval for MMP checking in seconds.
2224  */
2225 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
2226
2227 /*
2228  * Function prototypes
2229  */
2230
2231 /*
2232  * Ok, these declarations are also in <linux/kernel.h> but none of the
2233  * ext4 source programs needs to include it so they are duplicated here.
2234  */
2235 # define NORET_TYPE     /**/
2236 # define ATTRIB_NORET   __attribute__((noreturn))
2237 # define NORET_AND      noreturn,
2238
2239 /* bitmap.c */
2240 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
2241 void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
2242                                 struct ext4_group_desc *gdp,
2243                                 struct buffer_head *bh, int sz);
2244 int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2245                                   struct ext4_group_desc *gdp,
2246                                   struct buffer_head *bh, int sz);
2247 void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
2248                                 struct ext4_group_desc *gdp,
2249                                 struct buffer_head *bh);
2250 int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2251                                   struct ext4_group_desc *gdp,
2252                                   struct buffer_head *bh);
2253
2254 /* balloc.c */
2255 extern void ext4_get_group_no_and_offset(struct super_block *sb,
2256                                          ext4_fsblk_t blocknr,
2257                                          ext4_group_t *blockgrpp,
2258                                          ext4_grpblk_t *offsetp);
2259 extern ext4_group_t ext4_get_group_number(struct super_block *sb,
2260                                           ext4_fsblk_t block);
2261
2262 extern unsigned int ext4_block_group(struct super_block *sb,
2263                         ext4_fsblk_t blocknr);
2264 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
2265                         ext4_fsblk_t blocknr);
2266 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
2267 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
2268                         ext4_group_t group);
2269 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
2270                                          ext4_fsblk_t goal,
2271                                          unsigned int flags,
2272                                          unsigned long *count,
2273                                          int *errp);
2274 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
2275                                     s64 nclusters, unsigned int flags);
2276 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
2277 extern void ext4_check_blocks_bitmap(struct super_block *);
2278 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
2279                                                     ext4_group_t block_group,
2280                                                     struct buffer_head ** bh);
2281 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
2282
2283 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
2284                                                 ext4_group_t block_group);
2285 extern int ext4_wait_block_bitmap(struct super_block *sb,
2286                                   ext4_group_t block_group,
2287                                   struct buffer_head *bh);
2288 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
2289                                                   ext4_group_t block_group);
2290 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
2291                                               ext4_group_t block_group,
2292                                               struct ext4_group_desc *gdp);
2293 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
2294
2295 static inline int ext4_sb_has_crypto(struct super_block *sb)
2296 {
2297         return ext4_has_feature_encrypt(sb);
2298 }
2299
2300 static inline bool ext4_encrypted_inode(struct inode *inode)
2301 {
2302         return ext4_test_inode_flag(inode, EXT4_INODE_ENCRYPT);
2303 }
2304
2305 #ifdef CONFIG_EXT4_FS_ENCRYPTION
2306 static inline int ext4_fname_setup_filename(struct inode *dir,
2307                         const struct qstr *iname,
2308                         int lookup, struct ext4_filename *fname)
2309 {
2310         struct fscrypt_name name;
2311         int err;
2312
2313         memset(fname, 0, sizeof(struct ext4_filename));
2314
2315         err = fscrypt_setup_filename(dir, iname, lookup, &name);
2316
2317         fname->usr_fname = name.usr_fname;
2318         fname->disk_name = name.disk_name;
2319         fname->hinfo.hash = name.hash;
2320         fname->hinfo.minor_hash = name.minor_hash;
2321         fname->crypto_buf = name.crypto_buf;
2322         return err;
2323 }
2324
2325 static inline void ext4_fname_free_filename(struct ext4_filename *fname)
2326 {
2327         struct fscrypt_name name;
2328
2329         name.crypto_buf = fname->crypto_buf;
2330         fscrypt_free_filename(&name);
2331
2332         fname->crypto_buf.name = NULL;
2333         fname->usr_fname = NULL;
2334         fname->disk_name.name = NULL;
2335 }
2336 #else
2337 static inline int ext4_fname_setup_filename(struct inode *dir,
2338                 const struct qstr *iname,
2339                 int lookup, struct ext4_filename *fname)
2340 {
2341         fname->usr_fname = iname;
2342         fname->disk_name.name = (unsigned char *) iname->name;
2343         fname->disk_name.len = iname->len;
2344         return 0;
2345 }
2346 static inline void ext4_fname_free_filename(struct ext4_filename *fname) { }
2347
2348 #define fscrypt_set_d_op(i)
2349 #define fscrypt_get_ctx                 fscrypt_notsupp_get_ctx
2350 #define fscrypt_release_ctx             fscrypt_notsupp_release_ctx
2351 #define fscrypt_encrypt_page            fscrypt_notsupp_encrypt_page
2352 #define fscrypt_decrypt_page            fscrypt_notsupp_decrypt_page
2353 #define fscrypt_decrypt_bio_pages       fscrypt_notsupp_decrypt_bio_pages
2354 #define fscrypt_pullback_bio_page       fscrypt_notsupp_pullback_bio_page
2355 #define fscrypt_restore_control_page    fscrypt_notsupp_restore_control_page
2356 #define fscrypt_zeroout_range           fscrypt_notsupp_zeroout_range
2357 #define fscrypt_ioctl_set_policy        fscrypt_notsupp_ioctl_set_policy
2358 #define fscrypt_ioctl_get_policy        fscrypt_notsupp_ioctl_get_policy
2359 #define fscrypt_has_permitted_context   fscrypt_notsupp_has_permitted_context
2360 #define fscrypt_inherit_context         fscrypt_notsupp_inherit_context
2361 #define fscrypt_get_encryption_info     fscrypt_notsupp_get_encryption_info
2362 #define fscrypt_put_encryption_info     fscrypt_notsupp_put_encryption_info
2363 #define fscrypt_setup_filename          fscrypt_notsupp_setup_filename
2364 #define fscrypt_free_filename           fscrypt_notsupp_free_filename
2365 #define fscrypt_fname_encrypted_size    fscrypt_notsupp_fname_encrypted_size
2366 #define fscrypt_fname_alloc_buffer      fscrypt_notsupp_fname_alloc_buffer
2367 #define fscrypt_fname_free_buffer       fscrypt_notsupp_fname_free_buffer
2368 #define fscrypt_fname_disk_to_usr       fscrypt_notsupp_fname_disk_to_usr
2369 #define fscrypt_fname_usr_to_disk       fscrypt_notsupp_fname_usr_to_disk
2370 #endif
2371
2372 /* dir.c */
2373 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
2374                                   struct file *,
2375                                   struct ext4_dir_entry_2 *,
2376                                   struct buffer_head *, char *, int,
2377                                   unsigned int);
2378 #define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset)      \
2379         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
2380                                         (de), (bh), (buf), (size), (offset)))
2381 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
2382                                 __u32 minor_hash,
2383                                 struct ext4_dir_entry_2 *dirent,
2384                                 struct fscrypt_str *ent_name);
2385 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
2386 extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
2387                              struct buffer_head *bh,
2388                              void *buf, int buf_size,
2389                              struct ext4_filename *fname,
2390                              struct ext4_dir_entry_2 **dest_de);
2391 int ext4_insert_dentry(struct inode *dir,
2392                        struct inode *inode,
2393                        struct ext4_dir_entry_2 *de,
2394                        int buf_size,
2395                        struct ext4_filename *fname);
2396 static inline void ext4_update_dx_flag(struct inode *inode)
2397 {
2398         if (!ext4_has_feature_dir_index(inode->i_sb) &&
2399             ext4_test_inode_flag(inode, EXT4_INODE_INDEX)) {
2400                 /* ext4_iget() should have caught this... */
2401                 WARN_ON_ONCE(ext4_has_feature_metadata_csum(inode->i_sb));
2402                 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
2403         }
2404 }
2405 static unsigned char ext4_filetype_table[] = {
2406         DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
2407 };
2408
2409 static inline  unsigned char get_dtype(struct super_block *sb, int filetype)
2410 {
2411         if (!ext4_has_feature_filetype(sb) || filetype >= EXT4_FT_MAX)
2412                 return DT_UNKNOWN;
2413
2414         return ext4_filetype_table[filetype];
2415 }
2416 extern int ext4_check_all_de(struct inode *dir, struct buffer_head *bh,
2417                              void *buf, int buf_size);
2418
2419 /* fsync.c */
2420 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2421
2422 /* hash.c */
2423 extern int ext4fs_dirhash(const char *name, int len, struct
2424                           dx_hash_info *hinfo);
2425
2426 /* ialloc.c */
2427 extern struct inode *__ext4_new_inode(handle_t *, struct inode *, umode_t,
2428                                       const struct qstr *qstr, __u32 goal,
2429                                       uid_t *owner, int handle_type,
2430                                       unsigned int line_no, int nblocks);
2431
2432 #define ext4_new_inode(handle, dir, mode, qstr, goal, owner) \
2433         __ext4_new_inode((handle), (dir), (mode), (qstr), (goal), (owner), \
2434                          0, 0, 0)
2435 #define ext4_new_inode_start_handle(dir, mode, qstr, goal, owner, \
2436                                     type, nblocks)                  \
2437         __ext4_new_inode(NULL, (dir), (mode), (qstr), (goal), (owner), \
2438                          (type), __LINE__, (nblocks))
2439
2440
2441 extern void ext4_free_inode(handle_t *, struct inode *);
2442 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2443 extern unsigned long ext4_count_free_inodes(struct super_block *);
2444 extern unsigned long ext4_count_dirs(struct super_block *);
2445 extern void ext4_check_inodes_bitmap(struct super_block *);
2446 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2447 extern int ext4_init_inode_table(struct super_block *sb,
2448                                  ext4_group_t group, int barrier);
2449 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2450
2451 /* mballoc.c */
2452 extern const struct file_operations ext4_seq_mb_groups_fops;
2453 extern long ext4_mb_stats;
2454 extern long ext4_mb_max_to_scan;
2455 extern int ext4_mb_init(struct super_block *);
2456 extern int ext4_mb_release(struct super_block *);
2457 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2458                                 struct ext4_allocation_request *, int *);
2459 extern int ext4_mb_reserve_blocks(struct super_block *, int);
2460 extern void ext4_discard_preallocations(struct inode *);
2461 extern int __init ext4_init_mballoc(void);
2462 extern void ext4_exit_mballoc(void);
2463 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2464                              struct buffer_head *bh, ext4_fsblk_t block,
2465                              unsigned long count, int flags);
2466 extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2467                                    ext4_group_t ngroups);
2468 extern int ext4_mb_add_groupinfo(struct super_block *sb,
2469                 ext4_group_t i, struct ext4_group_desc *desc);
2470 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2471                                 ext4_fsblk_t block, unsigned long count);
2472 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2473
2474 /* inode.c */
2475 int ext4_inode_is_fast_symlink(struct inode *inode);
2476 struct buffer_head *ext4_getblk(handle_t *, struct inode *, ext4_lblk_t, int);
2477 struct buffer_head *ext4_bread(handle_t *, struct inode *, ext4_lblk_t, int);
2478 int ext4_get_block_unwritten(struct inode *inode, sector_t iblock,
2479                              struct buffer_head *bh_result, int create);
2480 int ext4_dax_get_block(struct inode *inode, sector_t iblock,
2481                        struct buffer_head *bh_result, int create);
2482 int ext4_get_block(struct inode *inode, sector_t iblock,
2483                    struct buffer_head *bh_result, int create);
2484 int ext4_dio_get_block(struct inode *inode, sector_t iblock,
2485                        struct buffer_head *bh_result, int create);
2486 int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2487                            struct buffer_head *bh, int create);
2488 int ext4_walk_page_buffers(handle_t *handle,
2489                            struct buffer_head *head,
2490                            unsigned from,
2491                            unsigned to,
2492                            int *partial,
2493                            int (*fn)(handle_t *handle,
2494                                      struct buffer_head *bh));
2495 int do_journal_get_write_access(handle_t *handle,
2496                                 struct buffer_head *bh);
2497 #define FALL_BACK_TO_NONDELALLOC 1
2498 #define CONVERT_INLINE_DATA      2
2499
2500 typedef enum {
2501         EXT4_IGET_NORMAL =      0,
2502         EXT4_IGET_SPECIAL =     0x0001, /* OK to iget a system inode */
2503         EXT4_IGET_HANDLE =      0x0002  /* Inode # is from a handle */
2504 } ext4_iget_flags;
2505
2506 extern struct inode *__ext4_iget(struct super_block *sb, unsigned long ino,
2507                                  ext4_iget_flags flags, const char *function,
2508                                  unsigned int line);
2509
2510 #define ext4_iget(sb, ino, flags) \
2511         __ext4_iget((sb), (ino), (flags), __func__, __LINE__)
2512
2513 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
2514 extern int  ext4_setattr(struct dentry *, struct iattr *);
2515 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
2516                                 struct kstat *stat);
2517 extern void ext4_evict_inode(struct inode *);
2518 extern void ext4_clear_inode(struct inode *);
2519 extern int  ext4_sync_inode(handle_t *, struct inode *);
2520 extern void ext4_dirty_inode(struct inode *, int);
2521 extern int ext4_change_inode_journal_flag(struct inode *, int);
2522 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
2523 extern int ext4_inode_attach_jinode(struct inode *inode);
2524 extern int ext4_can_truncate(struct inode *inode);
2525 extern void ext4_truncate(struct inode *);
2526 extern int ext4_punch_hole(struct inode *inode, loff_t offset, loff_t length);
2527 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
2528 extern void ext4_set_inode_flags(struct inode *);
2529 extern void ext4_get_inode_flags(struct ext4_inode_info *);
2530 extern int ext4_alloc_da_blocks(struct inode *inode);
2531 extern void ext4_set_aops(struct inode *inode);
2532 extern int ext4_writepage_trans_blocks(struct inode *);
2533 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
2534 extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
2535                              loff_t lstart, loff_t lend);
2536 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2537 extern int ext4_filemap_fault(struct vm_area_struct *vma, struct vm_fault *vmf);
2538 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
2539 extern int ext4_get_projid(struct inode *inode, kprojid_t *projid);
2540 extern void ext4_da_update_reserve_space(struct inode *inode,
2541                                         int used, int quota_claim);
2542 extern int ext4_issue_zeroout(struct inode *inode, ext4_lblk_t lblk,
2543                               ext4_fsblk_t pblk, ext4_lblk_t len);
2544 extern int ext4_get_next_extent(struct inode *inode, ext4_lblk_t lblk,
2545                                 unsigned int map_len,
2546                                 struct extent_status *result);
2547
2548 /* indirect.c */
2549 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
2550                                 struct ext4_map_blocks *map, int flags);
2551 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
2552 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
2553 extern void ext4_ind_truncate(handle_t *, struct inode *inode);
2554 extern int ext4_ind_remove_space(handle_t *handle, struct inode *inode,
2555                                  ext4_lblk_t start, ext4_lblk_t end);
2556
2557 /* ioctl.c */
2558 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
2559 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
2560
2561 /* migrate.c */
2562 extern int ext4_ext_migrate(struct inode *);
2563 extern int ext4_ind_migrate(struct inode *inode);
2564
2565 /* namei.c */
2566 extern int ext4_dirent_csum_verify(struct inode *inode,
2567                                    struct ext4_dir_entry *dirent);
2568 extern int ext4_orphan_add(handle_t *, struct inode *);
2569 extern int ext4_orphan_del(handle_t *, struct inode *);
2570 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
2571                                 __u32 start_minor_hash, __u32 *next_hash);
2572 extern int ext4_search_dir(struct buffer_head *bh,
2573                            char *search_buf,
2574                            int buf_size,
2575                            struct inode *dir,
2576                            struct ext4_filename *fname,
2577                            const struct qstr *d_name,
2578                            unsigned int offset,
2579                            struct ext4_dir_entry_2 **res_dir);
2580 extern int ext4_generic_delete_entry(handle_t *handle,
2581                                      struct inode *dir,
2582                                      struct ext4_dir_entry_2 *de_del,
2583                                      struct buffer_head *bh,
2584                                      void *entry_buf,
2585                                      int buf_size,
2586                                      int csum_size);
2587 extern bool ext4_empty_dir(struct inode *inode);
2588
2589 /* resize.c */
2590 extern void ext4_kvfree_array_rcu(void *to_free);
2591 extern int ext4_group_add(struct super_block *sb,
2592                                 struct ext4_new_group_data *input);
2593 extern int ext4_group_extend(struct super_block *sb,
2594                                 struct ext4_super_block *es,
2595                                 ext4_fsblk_t n_blocks_count);
2596 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
2597
2598 /* super.c */
2599 extern int ext4_seq_options_show(struct seq_file *seq, void *offset);
2600 extern int ext4_calculate_overhead(struct super_block *sb);
2601 extern void ext4_superblock_csum_set(struct super_block *sb);
2602 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
2603 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
2604 extern int ext4_alloc_flex_bg_array(struct super_block *sb,
2605                                     ext4_group_t ngroup);
2606 extern const char *ext4_decode_error(struct super_block *sb, int errno,
2607                                      char nbuf[16]);
2608
2609 extern __printf(4, 5)
2610 void __ext4_error(struct super_block *, const char *, unsigned int,
2611                   const char *, ...);
2612 extern __printf(5, 6)
2613 void __ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
2614                       const char *, ...);
2615 extern __printf(5, 6)
2616 void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
2617                      const char *, ...);
2618 extern void __ext4_std_error(struct super_block *, const char *,
2619                              unsigned int, int);
2620 extern __printf(4, 5)
2621 void __ext4_abort(struct super_block *, const char *, unsigned int,
2622                   const char *, ...);
2623 extern __printf(4, 5)
2624 void __ext4_warning(struct super_block *, const char *, unsigned int,
2625                     const char *, ...);
2626 extern __printf(4, 5)
2627 void __ext4_warning_inode(const struct inode *inode, const char *function,
2628                           unsigned int line, const char *fmt, ...);
2629 extern __printf(3, 4)
2630 void __ext4_msg(struct super_block *, const char *, const char *, ...);
2631 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
2632                            const char *, unsigned int, const char *);
2633 extern __printf(7, 8)
2634 void __ext4_grp_locked_error(const char *, unsigned int,
2635                              struct super_block *, ext4_group_t,
2636                              unsigned long, ext4_fsblk_t,
2637                              const char *, ...);
2638
2639 #define EXT4_ERROR_INODE(inode, fmt, a...) \
2640         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
2641
2642 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...)                 \
2643         ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
2644
2645 #define EXT4_ERROR_FILE(file, block, fmt, a...)                         \
2646         ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
2647
2648 #ifdef CONFIG_PRINTK
2649
2650 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2651         __ext4_error_inode(inode, func, line, block, fmt, ##__VA_ARGS__)
2652 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2653         __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__)
2654 #define ext4_error(sb, fmt, ...)                                        \
2655         __ext4_error(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2656 #define ext4_abort(sb, fmt, ...)                                        \
2657         __ext4_abort(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2658 #define ext4_warning(sb, fmt, ...)                                      \
2659         __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2660 #define ext4_warning_inode(inode, fmt, ...)                             \
2661         __ext4_warning_inode(inode, __func__, __LINE__, fmt, ##__VA_ARGS__)
2662 #define ext4_msg(sb, level, fmt, ...)                           \
2663         __ext4_msg(sb, level, fmt, ##__VA_ARGS__)
2664 #define dump_mmp_msg(sb, mmp, msg)                                      \
2665         __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg)
2666 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2667         __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \
2668                                 fmt, ##__VA_ARGS__)
2669
2670 #else
2671
2672 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
2673 do {                                                                    \
2674         no_printk(fmt, ##__VA_ARGS__);                                  \
2675         __ext4_error_inode(inode, "", 0, block, " ");                   \
2676 } while (0)
2677 #define ext4_error_file(file, func, line, block, fmt, ...)              \
2678 do {                                                                    \
2679         no_printk(fmt, ##__VA_ARGS__);                                  \
2680         __ext4_error_file(file, "", 0, block, " ");                     \
2681 } while (0)
2682 #define ext4_error(sb, fmt, ...)                                        \
2683 do {                                                                    \
2684         no_printk(fmt, ##__VA_ARGS__);                                  \
2685         __ext4_error(sb, "", 0, " ");                                   \
2686 } while (0)
2687 #define ext4_abort(sb, fmt, ...)                                        \
2688 do {                                                                    \
2689         no_printk(fmt, ##__VA_ARGS__);                                  \
2690         __ext4_abort(sb, "", 0, " ");                                   \
2691 } while (0)
2692 #define ext4_warning(sb, fmt, ...)                                      \
2693 do {                                                                    \
2694         no_printk(fmt, ##__VA_ARGS__);                                  \
2695         __ext4_warning(sb, "", 0, " ");                                 \
2696 } while (0)
2697 #define ext4_warning_inode(inode, fmt, ...)                             \
2698 do {                                                                    \
2699         no_printk(fmt, ##__VA_ARGS__);                                  \
2700         __ext4_warning_inode(inode, "", 0, " ");                        \
2701 } while (0)
2702 #define ext4_msg(sb, level, fmt, ...)                                   \
2703 do {                                                                    \
2704         no_printk(fmt, ##__VA_ARGS__);                                  \
2705         __ext4_msg(sb, "", " ");                                        \
2706 } while (0)
2707 #define dump_mmp_msg(sb, mmp, msg)                                      \
2708         __dump_mmp_msg(sb, mmp, "", 0, "")
2709 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
2710 do {                                                                    \
2711         no_printk(fmt, ##__VA_ARGS__);                          \
2712         __ext4_grp_locked_error("", 0, sb, grp, ino, block, " ");       \
2713 } while (0)
2714
2715 #endif
2716
2717 extern void ext4_update_dynamic_rev(struct super_block *sb);
2718 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
2719                                         __u32 compat);
2720 extern int ext4_update_rocompat_feature(handle_t *handle,
2721                                         struct super_block *sb, __u32 rocompat);
2722 extern int ext4_update_incompat_feature(handle_t *handle,
2723                                         struct super_block *sb, __u32 incompat);
2724 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
2725                                       struct ext4_group_desc *bg);
2726 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
2727                                       struct ext4_group_desc *bg);
2728 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
2729                                      struct ext4_group_desc *bg);
2730 extern __u32 ext4_free_group_clusters(struct super_block *sb,
2731                                       struct ext4_group_desc *bg);
2732 extern __u32 ext4_free_inodes_count(struct super_block *sb,
2733                                  struct ext4_group_desc *bg);
2734 extern __u32 ext4_used_dirs_count(struct super_block *sb,
2735                                 struct ext4_group_desc *bg);
2736 extern __u32 ext4_itable_unused_count(struct super_block *sb,
2737                                    struct ext4_group_desc *bg);
2738 extern void ext4_block_bitmap_set(struct super_block *sb,
2739                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2740 extern void ext4_inode_bitmap_set(struct super_block *sb,
2741                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2742 extern void ext4_inode_table_set(struct super_block *sb,
2743                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
2744 extern void ext4_free_group_clusters_set(struct super_block *sb,
2745                                          struct ext4_group_desc *bg,
2746                                          __u32 count);
2747 extern void ext4_free_inodes_set(struct super_block *sb,
2748                                 struct ext4_group_desc *bg, __u32 count);
2749 extern void ext4_used_dirs_set(struct super_block *sb,
2750                                 struct ext4_group_desc *bg, __u32 count);
2751 extern void ext4_itable_unused_set(struct super_block *sb,
2752                                    struct ext4_group_desc *bg, __u32 count);
2753 extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
2754                                        struct ext4_group_desc *gdp);
2755 extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
2756                                      struct ext4_group_desc *gdp);
2757 extern int ext4_register_li_request(struct super_block *sb,
2758                                     ext4_group_t first_not_zeroed);
2759
2760 static inline int ext4_has_group_desc_csum(struct super_block *sb)
2761 {
2762         return ext4_has_feature_gdt_csum(sb) ||
2763                EXT4_SB(sb)->s_chksum_driver != NULL;
2764 }
2765
2766 static inline int ext4_has_metadata_csum(struct super_block *sb)
2767 {
2768         WARN_ON_ONCE(ext4_has_feature_metadata_csum(sb) &&
2769                      !EXT4_SB(sb)->s_chksum_driver);
2770
2771         return (EXT4_SB(sb)->s_chksum_driver != NULL);
2772 }
2773 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2774 {
2775         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2776                 le32_to_cpu(es->s_blocks_count_lo);
2777 }
2778
2779 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2780 {
2781         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2782                 le32_to_cpu(es->s_r_blocks_count_lo);
2783 }
2784
2785 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2786 {
2787         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2788                 le32_to_cpu(es->s_free_blocks_count_lo);
2789 }
2790
2791 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2792                                          ext4_fsblk_t blk)
2793 {
2794         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2795         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2796 }
2797
2798 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2799                                               ext4_fsblk_t blk)
2800 {
2801         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2802         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2803 }
2804
2805 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2806                                            ext4_fsblk_t blk)
2807 {
2808         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2809         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2810 }
2811
2812 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2813 {
2814         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2815                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2816                         le32_to_cpu(raw_inode->i_size_lo);
2817         else
2818                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2819 }
2820
2821 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2822 {
2823         raw_inode->i_size_lo = cpu_to_le32(i_size);
2824         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2825 }
2826
2827 static inline
2828 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2829                                             ext4_group_t group)
2830 {
2831          struct ext4_group_info **grp_info;
2832          long indexv, indexh;
2833          BUG_ON(group >= EXT4_SB(sb)->s_groups_count);
2834          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2835          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2836          grp_info = sbi_array_rcu_deref(EXT4_SB(sb), s_group_info, indexv);
2837          return grp_info[indexh];
2838 }
2839
2840 /*
2841  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2842  * the locking protocol documented in the comments of ext4_group_add()
2843  * in resize.c
2844  */
2845 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2846 {
2847         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2848
2849         smp_rmb();
2850         return ngroups;
2851 }
2852
2853 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2854                                              ext4_group_t block_group)
2855 {
2856         return block_group >> sbi->s_log_groups_per_flex;
2857 }
2858
2859 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2860 {
2861         return 1 << sbi->s_log_groups_per_flex;
2862 }
2863
2864 #define ext4_std_error(sb, errno)                               \
2865 do {                                                            \
2866         if ((errno))                                            \
2867                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2868 } while (0)
2869
2870 #ifdef CONFIG_SMP
2871 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2872  * counters. So we need to make sure we have free clusters more
2873  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2874  */
2875 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2876 #else
2877 #define EXT4_FREECLUSTERS_WATERMARK 0
2878 #endif
2879
2880 /* Update i_disksize. Requires i_mutex to avoid races with truncate */
2881 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2882 {
2883         WARN_ON_ONCE(S_ISREG(inode->i_mode) &&
2884                      !inode_is_locked(inode));
2885         down_write(&EXT4_I(inode)->i_data_sem);
2886         if (newsize > EXT4_I(inode)->i_disksize)
2887                 WRITE_ONCE(EXT4_I(inode)->i_disksize, newsize);
2888         up_write(&EXT4_I(inode)->i_data_sem);
2889 }
2890
2891 /* Update i_size, i_disksize. Requires i_mutex to avoid races with truncate */
2892 static inline int ext4_update_inode_size(struct inode *inode, loff_t newsize)
2893 {
2894         int changed = 0;
2895
2896         if (newsize > inode->i_size) {
2897                 i_size_write(inode, newsize);
2898                 changed = 1;
2899         }
2900         if (newsize > EXT4_I(inode)->i_disksize) {
2901                 ext4_update_i_disksize(inode, newsize);
2902                 changed |= 2;
2903         }
2904         return changed;
2905 }
2906
2907 int ext4_update_disksize_before_punch(struct inode *inode, loff_t offset,
2908                                       loff_t len);
2909
2910 struct ext4_group_info {
2911         unsigned long   bb_state;
2912         struct rb_root  bb_free_root;
2913         ext4_grpblk_t   bb_first_free;  /* first free block */
2914         ext4_grpblk_t   bb_free;        /* total free blocks */
2915         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2916         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2917         struct          list_head bb_prealloc_list;
2918 #ifdef DOUBLE_CHECK
2919         void            *bb_bitmap;
2920 #endif
2921         struct rw_semaphore alloc_sem;
2922         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2923                                          * regions, index is order.
2924                                          * bb_counters[3] = 5 means
2925                                          * 5 free 8-block regions. */
2926 };
2927
2928 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2929 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2930 #define EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT     2
2931 #define EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT     3
2932
2933 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2934         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2935 #define EXT4_MB_GRP_BBITMAP_CORRUPT(grp)        \
2936         (test_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
2937 #define EXT4_MB_GRP_IBITMAP_CORRUPT(grp)        \
2938         (test_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
2939
2940 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2941         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2942 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2943         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2944 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2945         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2946
2947 #define EXT4_MAX_CONTENTION             8
2948 #define EXT4_CONTENTION_THRESHOLD       2
2949
2950 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2951                                               ext4_group_t group)
2952 {
2953         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2954 }
2955
2956 /*
2957  * Returns true if the filesystem is busy enough that attempts to
2958  * access the block group locks has run into contention.
2959  */
2960 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2961 {
2962         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2963 }
2964
2965 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2966 {
2967         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2968         if (spin_trylock(lock))
2969                 /*
2970                  * We're able to grab the lock right away, so drop the
2971                  * lock contention counter.
2972                  */
2973                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2974         else {
2975                 /*
2976                  * The lock is busy, so bump the contention counter,
2977                  * and then wait on the spin lock.
2978                  */
2979                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2980                                   EXT4_MAX_CONTENTION);
2981                 spin_lock(lock);
2982         }
2983 }
2984
2985 static inline void ext4_unlock_group(struct super_block *sb,
2986                                         ext4_group_t group)
2987 {
2988         spin_unlock(ext4_group_lock_ptr(sb, group));
2989 }
2990
2991 /*
2992  * Block validity checking
2993  */
2994 #define ext4_check_indirect_blockref(inode, bh)                         \
2995         ext4_check_blockref(__func__, __LINE__, inode,                  \
2996                             (__le32 *)(bh)->b_data,                     \
2997                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2998
2999 #define ext4_ind_check_inode(inode)                                     \
3000         ext4_check_blockref(__func__, __LINE__, inode,                  \
3001                             EXT4_I(inode)->i_data,                      \
3002                             EXT4_NDIR_BLOCKS)
3003
3004 /*
3005  * Inodes and files operations
3006  */
3007
3008 /* dir.c */
3009 extern const struct file_operations ext4_dir_operations;
3010
3011 /* file.c */
3012 extern const struct inode_operations ext4_file_inode_operations;
3013 extern const struct file_operations ext4_file_operations;
3014 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
3015
3016 /* inline.c */
3017 extern int ext4_get_max_inline_size(struct inode *inode);
3018 extern int ext4_find_inline_data_nolock(struct inode *inode);
3019 extern int ext4_init_inline_data(handle_t *handle, struct inode *inode,
3020                                  unsigned int len);
3021 extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
3022
3023 extern int ext4_readpage_inline(struct inode *inode, struct page *page);
3024 extern int ext4_try_to_write_inline_data(struct address_space *mapping,
3025                                          struct inode *inode,
3026                                          loff_t pos, unsigned len,
3027                                          unsigned flags,
3028                                          struct page **pagep);
3029 extern int ext4_write_inline_data_end(struct inode *inode,
3030                                       loff_t pos, unsigned len,
3031                                       unsigned copied,
3032                                       struct page *page);
3033 extern struct buffer_head *
3034 ext4_journalled_write_inline_data(struct inode *inode,
3035                                   unsigned len,
3036                                   struct page *page);
3037 extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
3038                                            struct inode *inode,
3039                                            loff_t pos, unsigned len,
3040                                            unsigned flags,
3041                                            struct page **pagep,
3042                                            void **fsdata);
3043 extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
3044                                          unsigned len, unsigned copied,
3045                                          struct page *page);
3046 extern int ext4_try_add_inline_entry(handle_t *handle,
3047                                      struct ext4_filename *fname,
3048                                      struct inode *dir, struct inode *inode);
3049 extern int ext4_try_create_inline_dir(handle_t *handle,
3050                                       struct inode *parent,
3051                                       struct inode *inode);
3052 extern int ext4_read_inline_dir(struct file *filp,
3053                                 struct dir_context *ctx,
3054                                 int *has_inline_data);
3055 extern int htree_inlinedir_to_tree(struct file *dir_file,
3056                                    struct inode *dir, ext4_lblk_t block,
3057                                    struct dx_hash_info *hinfo,
3058                                    __u32 start_hash, __u32 start_minor_hash,
3059                                    int *has_inline_data);
3060 extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
3061                                         struct ext4_filename *fname,
3062                                         const struct qstr *d_name,
3063                                         struct ext4_dir_entry_2 **res_dir,
3064                                         int *has_inline_data);
3065 extern int ext4_delete_inline_entry(handle_t *handle,
3066                                     struct inode *dir,
3067                                     struct ext4_dir_entry_2 *de_del,
3068                                     struct buffer_head *bh,
3069                                     int *has_inline_data);
3070 extern bool empty_inline_dir(struct inode *dir, int *has_inline_data);
3071 extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
3072                                         struct ext4_dir_entry_2 **parent_de,
3073                                         int *retval);
3074 extern int ext4_inline_data_fiemap(struct inode *inode,
3075                                    struct fiemap_extent_info *fieinfo,
3076                                    int *has_inline, __u64 start, __u64 len);
3077 extern void ext4_inline_data_truncate(struct inode *inode, int *has_inline);
3078
3079 extern int ext4_convert_inline_data(struct inode *inode);
3080
3081 static inline int ext4_has_inline_data(struct inode *inode)
3082 {
3083         return ext4_test_inode_flag(inode, EXT4_INODE_INLINE_DATA) &&
3084                EXT4_I(inode)->i_inline_off;
3085 }
3086
3087 /* namei.c */
3088 extern const struct inode_operations ext4_dir_inode_operations;
3089 extern const struct inode_operations ext4_special_inode_operations;
3090 extern struct dentry *ext4_get_parent(struct dentry *child);
3091 extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
3092                                  struct ext4_dir_entry_2 *de,
3093                                  int blocksize, int csum_size,
3094                                  unsigned int parent_ino, int dotdot_real_len);
3095 extern void initialize_dirent_tail(struct ext4_dir_entry_tail *t,
3096                                    unsigned int blocksize);
3097 extern int ext4_handle_dirty_dirent_node(handle_t *handle,
3098                                          struct inode *inode,
3099                                          struct buffer_head *bh);
3100 #define S_SHIFT 12
3101 static unsigned char ext4_type_by_mode[S_IFMT >> S_SHIFT] = {
3102         [S_IFREG >> S_SHIFT]    = EXT4_FT_REG_FILE,
3103         [S_IFDIR >> S_SHIFT]    = EXT4_FT_DIR,
3104         [S_IFCHR >> S_SHIFT]    = EXT4_FT_CHRDEV,
3105         [S_IFBLK >> S_SHIFT]    = EXT4_FT_BLKDEV,
3106         [S_IFIFO >> S_SHIFT]    = EXT4_FT_FIFO,
3107         [S_IFSOCK >> S_SHIFT]   = EXT4_FT_SOCK,
3108         [S_IFLNK >> S_SHIFT]    = EXT4_FT_SYMLINK,
3109 };
3110
3111 static inline void ext4_set_de_type(struct super_block *sb,
3112                                 struct ext4_dir_entry_2 *de,
3113                                 umode_t mode) {
3114         if (ext4_has_feature_filetype(sb))
3115                 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
3116 }
3117
3118 /* readpages.c */
3119 extern int ext4_mpage_readpages(struct address_space *mapping,
3120                                 struct list_head *pages, struct page *page,
3121                                 unsigned nr_pages);
3122
3123 /* symlink.c */
3124 extern const struct inode_operations ext4_encrypted_symlink_inode_operations;
3125 extern const struct inode_operations ext4_symlink_inode_operations;
3126 extern const struct inode_operations ext4_fast_symlink_inode_operations;
3127
3128 /* sysfs.c */
3129 extern int ext4_register_sysfs(struct super_block *sb);
3130 extern void ext4_unregister_sysfs(struct super_block *sb);
3131 extern int __init ext4_init_sysfs(void);
3132 extern void ext4_exit_sysfs(void);
3133
3134 /* block_validity */
3135 extern void ext4_release_system_zone(struct super_block *sb);
3136 extern int ext4_setup_system_zone(struct super_block *sb);
3137 extern int __init ext4_init_system_zone(void);
3138 extern void ext4_exit_system_zone(void);
3139 extern int ext4_inode_block_valid(struct inode *inode,
3140                                   ext4_fsblk_t start_blk,
3141                                   unsigned int count);
3142 extern int ext4_check_blockref(const char *, unsigned int,
3143                                struct inode *, __le32 *, unsigned int);
3144
3145 /* extents.c */
3146 struct ext4_ext_path;
3147 struct ext4_extent;
3148
3149 /*
3150  * Maximum number of logical blocks in a file; ext4_extent's ee_block is
3151  * __le32.
3152  */
3153 #define EXT_MAX_BLOCKS  0xffffffff
3154
3155 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
3156 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
3157 extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
3158 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
3159                                struct ext4_map_blocks *map, int flags);
3160 extern void ext4_ext_truncate(handle_t *, struct inode *);
3161 extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
3162                                  ext4_lblk_t end);
3163 extern void ext4_ext_init(struct super_block *);
3164 extern void ext4_ext_release(struct super_block *);
3165 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
3166                           loff_t len);
3167 extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
3168                                           loff_t offset, ssize_t len);
3169 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
3170                            struct ext4_map_blocks *map, int flags);
3171 extern int ext4_ext_calc_metadata_amount(struct inode *inode,
3172                                          ext4_lblk_t lblocks);
3173 extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
3174                                                    int num,
3175                                                    struct ext4_ext_path *path);
3176 extern int ext4_can_extents_be_merged(struct inode *inode,
3177                                       struct ext4_extent *ex1,
3178                                       struct ext4_extent *ex2);
3179 extern int ext4_ext_insert_extent(handle_t *, struct inode *,
3180                                   struct ext4_ext_path **,
3181                                   struct ext4_extent *, int);
3182 extern struct ext4_ext_path *ext4_find_extent(struct inode *, ext4_lblk_t,
3183                                               struct ext4_ext_path **,
3184                                               int flags);
3185 extern void ext4_ext_drop_refs(struct ext4_ext_path *);
3186 extern int ext4_ext_check_inode(struct inode *inode);
3187 extern int ext4_find_delalloc_range(struct inode *inode,
3188                                     ext4_lblk_t lblk_start,
3189                                     ext4_lblk_t lblk_end);
3190 extern int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk);
3191 extern ext4_lblk_t ext4_ext_next_allocated_block(struct ext4_ext_path *path);
3192 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
3193                         __u64 start, __u64 len);
3194 extern int ext4_ext_precache(struct inode *inode);
3195 extern int ext4_collapse_range(struct inode *inode, loff_t offset, loff_t len);
3196 extern int ext4_insert_range(struct inode *inode, loff_t offset, loff_t len);
3197 extern int ext4_swap_extents(handle_t *handle, struct inode *inode1,
3198                                 struct inode *inode2, ext4_lblk_t lblk1,
3199                              ext4_lblk_t lblk2,  ext4_lblk_t count,
3200                              int mark_unwritten,int *err);
3201
3202 /* move_extent.c */
3203 extern void ext4_double_down_write_data_sem(struct inode *first,
3204                                             struct inode *second);
3205 extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
3206                                           struct inode *donor_inode);
3207 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
3208                              __u64 start_orig, __u64 start_donor,
3209                              __u64 len, __u64 *moved_len);
3210
3211 /* page-io.c */
3212 extern int __init ext4_init_pageio(void);
3213 extern void ext4_exit_pageio(void);
3214 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
3215 extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
3216 extern int ext4_put_io_end(ext4_io_end_t *io_end);
3217 extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
3218 extern void ext4_io_submit_init(struct ext4_io_submit *io,
3219                                 struct writeback_control *wbc);
3220 extern void ext4_end_io_rsv_work(struct work_struct *work);
3221 extern void ext4_io_submit(struct ext4_io_submit *io);
3222 extern int ext4_bio_write_page(struct ext4_io_submit *io,
3223                                struct page *page,
3224                                int len,
3225                                struct writeback_control *wbc,
3226                                bool keep_towrite);
3227
3228 /* mmp.c */
3229 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
3230
3231 /*
3232  * Add new method to test whether block and inode bitmaps are properly
3233  * initialized. With uninit_bg reading the block from disk is not enough
3234  * to mark the bitmap uptodate. We need to also zero-out the bitmap
3235  */
3236 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
3237
3238 static inline int bitmap_uptodate(struct buffer_head *bh)
3239 {
3240         return (buffer_uptodate(bh) &&
3241                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
3242 }
3243 static inline void set_bitmap_uptodate(struct buffer_head *bh)
3244 {
3245         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
3246 }
3247
3248 /*
3249  * Disable DIO read nolock optimization, so new dioreaders will be forced
3250  * to grab i_mutex
3251  */
3252 static inline void ext4_inode_block_unlocked_dio(struct inode *inode)
3253 {
3254         ext4_set_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
3255         smp_mb();
3256 }
3257 static inline void ext4_inode_resume_unlocked_dio(struct inode *inode)
3258 {
3259         smp_mb();
3260         ext4_clear_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
3261 }
3262
3263 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
3264
3265 /* For ioend & aio unwritten conversion wait queues */
3266 #define EXT4_WQ_HASH_SZ         37
3267 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
3268                                             EXT4_WQ_HASH_SZ])
3269 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
3270
3271 #define EXT4_RESIZING   0
3272 extern int ext4_resize_begin(struct super_block *sb);
3273 extern void ext4_resize_end(struct super_block *sb);
3274
3275 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
3276                                               struct ext4_io_end *io_end)
3277 {
3278         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
3279                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
3280                 atomic_inc(&EXT4_I(inode)->i_unwritten);
3281         }
3282 }
3283
3284 static inline void ext4_clear_io_unwritten_flag(ext4_io_end_t *io_end)
3285 {
3286         struct inode *inode = io_end->inode;
3287
3288         if (io_end->flag & EXT4_IO_END_UNWRITTEN) {
3289                 io_end->flag &= ~EXT4_IO_END_UNWRITTEN;
3290                 /* Wake up anyone waiting on unwritten extent conversion */
3291                 if (atomic_dec_and_test(&EXT4_I(inode)->i_unwritten))
3292                         wake_up_all(ext4_ioend_wq(inode));
3293         }
3294 }
3295
3296 static inline bool ext4_aligned_io(struct inode *inode, loff_t off, loff_t len)
3297 {
3298         int blksize = 1 << inode->i_blkbits;
3299
3300         return IS_ALIGNED(off, blksize) && IS_ALIGNED(len, blksize);
3301 }
3302
3303 #endif  /* __KERNEL__ */
3304
3305 #define EFSBADCRC       EBADMSG         /* Bad CRC detected */
3306 #define EFSCORRUPTED    EUCLEAN         /* Filesystem is corrupted */
3307
3308 #endif  /* _EXT4_H */