]> git.itanic.dy.fi Git - linux-stable/blob - fs/ext4/ext4.h
ext4: use buckets for cr 1 block scan instead of rbtree
[linux-stable] / fs / ext4 / ext4.h
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  ext4.h
4  *
5  * Copyright (C) 1992, 1993, 1994, 1995
6  * Remy Card (card@masi.ibp.fr)
7  * Laboratoire MASI - Institut Blaise Pascal
8  * Universite Pierre et Marie Curie (Paris VI)
9  *
10  *  from
11  *
12  *  linux/include/linux/minix_fs.h
13  *
14  *  Copyright (C) 1991, 1992  Linus Torvalds
15  */
16
17 #ifndef _EXT4_H
18 #define _EXT4_H
19
20 #include <linux/refcount.h>
21 #include <linux/types.h>
22 #include <linux/blkdev.h>
23 #include <linux/magic.h>
24 #include <linux/jbd2.h>
25 #include <linux/quota.h>
26 #include <linux/rwsem.h>
27 #include <linux/rbtree.h>
28 #include <linux/seqlock.h>
29 #include <linux/mutex.h>
30 #include <linux/timer.h>
31 #include <linux/wait.h>
32 #include <linux/sched/signal.h>
33 #include <linux/blockgroup_lock.h>
34 #include <linux/percpu_counter.h>
35 #include <linux/ratelimit.h>
36 #include <crypto/hash.h>
37 #include <linux/falloc.h>
38 #include <linux/percpu-rwsem.h>
39 #include <linux/fiemap.h>
40 #ifdef __KERNEL__
41 #include <linux/compat.h>
42 #endif
43
44 #include <linux/fscrypt.h>
45 #include <linux/fsverity.h>
46
47 #include <linux/compiler.h>
48
49 /*
50  * The fourth extended filesystem constants/structures
51  */
52
53 /*
54  * with AGGRESSIVE_CHECK allocator runs consistency checks over
55  * structures. these checks slow things down a lot
56  */
57 #define AGGRESSIVE_CHECK__
58
59 /*
60  * with DOUBLE_CHECK defined mballoc creates persistent in-core
61  * bitmaps, maintains and uses them to check for double allocations
62  */
63 #define DOUBLE_CHECK__
64
65 /*
66  * Define EXT4FS_DEBUG to produce debug messages
67  */
68 #undef EXT4FS_DEBUG
69
70 /*
71  * Debug code
72  */
73 #ifdef EXT4FS_DEBUG
74 #define ext4_debug(f, a...)                                             \
75         do {                                                            \
76                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
77                         __FILE__, __LINE__, __func__);                  \
78                 printk(KERN_DEBUG f, ## a);                             \
79         } while (0)
80 #else
81 #define ext4_debug(fmt, ...)    no_printk(fmt, ##__VA_ARGS__)
82 #endif
83
84  /*
85   * Turn on EXT_DEBUG to enable ext4_ext_show_path/leaf/move in extents.c
86   */
87 #define EXT_DEBUG__
88
89 /*
90  * Dynamic printk for controlled extents debugging.
91  */
92 #ifdef CONFIG_EXT4_DEBUG
93 #define ext_debug(ino, fmt, ...)                                        \
94         pr_debug("[%s/%d] EXT4-fs (%s): ino %lu: (%s, %d): %s:" fmt,    \
95                  current->comm, task_pid_nr(current),                   \
96                  ino->i_sb->s_id, ino->i_ino, __FILE__, __LINE__,       \
97                  __func__, ##__VA_ARGS__)
98 #else
99 #define ext_debug(ino, fmt, ...)        no_printk(fmt, ##__VA_ARGS__)
100 #endif
101
102 #define ASSERT(assert)                                          \
103 do {                                                                    \
104         if (unlikely(!(assert))) {                                      \
105                 printk(KERN_EMERG                                       \
106                        "Assertion failure in %s() at %s:%d: '%s'\n",    \
107                        __func__, __FILE__, __LINE__, #assert);          \
108                 BUG();                                                  \
109         }                                                               \
110 } while (0)
111
112 /* data type for block offset of block group */
113 typedef int ext4_grpblk_t;
114
115 /* data type for filesystem-wide blocks number */
116 typedef unsigned long long ext4_fsblk_t;
117
118 /* data type for file logical block number */
119 typedef __u32 ext4_lblk_t;
120
121 /* data type for block group number */
122 typedef unsigned int ext4_group_t;
123
124 enum SHIFT_DIRECTION {
125         SHIFT_LEFT = 0,
126         SHIFT_RIGHT,
127 };
128
129 /*
130  * Flags used in mballoc's allocation_context flags field.
131  *
132  * Also used to show what's going on for debugging purposes when the
133  * flag field is exported via the traceport interface
134  */
135
136 /* prefer goal again. length */
137 #define EXT4_MB_HINT_MERGE              0x0001
138 /* blocks already reserved */
139 #define EXT4_MB_HINT_RESERVED           0x0002
140 /* metadata is being allocated */
141 #define EXT4_MB_HINT_METADATA           0x0004
142 /* first blocks in the file */
143 #define EXT4_MB_HINT_FIRST              0x0008
144 /* search for the best chunk */
145 #define EXT4_MB_HINT_BEST               0x0010
146 /* data is being allocated */
147 #define EXT4_MB_HINT_DATA               0x0020
148 /* don't preallocate (for tails) */
149 #define EXT4_MB_HINT_NOPREALLOC         0x0040
150 /* allocate for locality group */
151 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
152 /* allocate goal blocks or none */
153 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
154 /* goal is meaningful */
155 #define EXT4_MB_HINT_TRY_GOAL           0x0200
156 /* blocks already pre-reserved by delayed allocation */
157 #define EXT4_MB_DELALLOC_RESERVED       0x0400
158 /* We are doing stream allocation */
159 #define EXT4_MB_STREAM_ALLOC            0x0800
160 /* Use reserved root blocks if needed */
161 #define EXT4_MB_USE_ROOT_BLOCKS         0x1000
162 /* Use blocks from reserved pool */
163 #define EXT4_MB_USE_RESERVED            0x2000
164 /* Do strict check for free blocks while retrying block allocation */
165 #define EXT4_MB_STRICT_CHECK            0x4000
166 /* Large fragment size list lookup succeeded at least once for cr = 0 */
167 #define EXT4_MB_CR0_OPTIMIZED           0x8000
168 /* Avg fragment size rb tree lookup succeeded at least once for cr = 1 */
169 #define EXT4_MB_CR1_OPTIMIZED           0x00010000
170 struct ext4_allocation_request {
171         /* target inode for block we're allocating */
172         struct inode *inode;
173         /* how many blocks we want to allocate */
174         unsigned int len;
175         /* logical block in target inode */
176         ext4_lblk_t logical;
177         /* the closest logical allocated block to the left */
178         ext4_lblk_t lleft;
179         /* the closest logical allocated block to the right */
180         ext4_lblk_t lright;
181         /* phys. target (a hint) */
182         ext4_fsblk_t goal;
183         /* phys. block for the closest logical allocated block to the left */
184         ext4_fsblk_t pleft;
185         /* phys. block for the closest logical allocated block to the right */
186         ext4_fsblk_t pright;
187         /* flags. see above EXT4_MB_HINT_* */
188         unsigned int flags;
189 };
190
191 /*
192  * Logical to physical block mapping, used by ext4_map_blocks()
193  *
194  * This structure is used to pass requests into ext4_map_blocks() as
195  * well as to store the information returned by ext4_map_blocks().  It
196  * takes less room on the stack than a struct buffer_head.
197  */
198 #define EXT4_MAP_NEW            BIT(BH_New)
199 #define EXT4_MAP_MAPPED         BIT(BH_Mapped)
200 #define EXT4_MAP_UNWRITTEN      BIT(BH_Unwritten)
201 #define EXT4_MAP_BOUNDARY       BIT(BH_Boundary)
202 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
203                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY)
204
205 struct ext4_map_blocks {
206         ext4_fsblk_t m_pblk;
207         ext4_lblk_t m_lblk;
208         unsigned int m_len;
209         unsigned int m_flags;
210 };
211
212 /*
213  * Block validity checking, system zone rbtree.
214  */
215 struct ext4_system_blocks {
216         struct rb_root root;
217         struct rcu_head rcu;
218 };
219
220 /*
221  * Flags for ext4_io_end->flags
222  */
223 #define EXT4_IO_END_UNWRITTEN   0x0001
224
225 struct ext4_io_end_vec {
226         struct list_head list;          /* list of io_end_vec */
227         loff_t offset;                  /* offset in the file */
228         ssize_t size;                   /* size of the extent */
229 };
230
231 /*
232  * For converting unwritten extents on a work queue. 'handle' is used for
233  * buffered writeback.
234  */
235 typedef struct ext4_io_end {
236         struct list_head        list;           /* per-file finished IO list */
237         handle_t                *handle;        /* handle reserved for extent
238                                                  * conversion */
239         struct inode            *inode;         /* file being written to */
240         struct bio              *bio;           /* Linked list of completed
241                                                  * bios covering the extent */
242         unsigned int            flag;           /* unwritten or not */
243         refcount_t              count;          /* reference counter */
244         struct list_head        list_vec;       /* list of ext4_io_end_vec */
245 } ext4_io_end_t;
246
247 struct ext4_io_submit {
248         struct writeback_control *io_wbc;
249         struct bio              *io_bio;
250         ext4_io_end_t           *io_end;
251         sector_t                io_next_block;
252 };
253
254 /*
255  * Special inodes numbers
256  */
257 #define EXT4_BAD_INO             1      /* Bad blocks inode */
258 #define EXT4_ROOT_INO            2      /* Root inode */
259 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
260 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
261 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
262 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
263 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
264 #define EXT4_JOURNAL_INO         8      /* Journal inode */
265
266 /* First non-reserved inode for old ext4 filesystems */
267 #define EXT4_GOOD_OLD_FIRST_INO 11
268
269 /*
270  * Maximal count of links to a file
271  */
272 #define EXT4_LINK_MAX           65000
273
274 /*
275  * Macro-instructions used to manage several block sizes
276  */
277 #define EXT4_MIN_BLOCK_SIZE             1024
278 #define EXT4_MAX_BLOCK_SIZE             65536
279 #define EXT4_MIN_BLOCK_LOG_SIZE         10
280 #define EXT4_MAX_BLOCK_LOG_SIZE         16
281 #define EXT4_MAX_CLUSTER_LOG_SIZE       30
282 #ifdef __KERNEL__
283 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
284 #else
285 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
286 #endif
287 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
288 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
289                                          EXT4_SB(s)->s_cluster_bits)
290 #ifdef __KERNEL__
291 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
292 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
293 #else
294 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
295 #endif
296 #ifdef __KERNEL__
297 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
298 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
299 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
300 #else
301 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
302                                  EXT4_GOOD_OLD_INODE_SIZE : \
303                                  (s)->s_inode_size)
304 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
305                                  EXT4_GOOD_OLD_FIRST_INO : \
306                                  (s)->s_first_ino)
307 #endif
308 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
309 #define EXT4_MAX_BLOCKS(size, offset, blkbits) \
310         ((EXT4_BLOCK_ALIGN(size + offset, blkbits) >> blkbits) - (offset >> \
311                                                                   blkbits))
312
313 /* Translate a block number to a cluster number */
314 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
315 /* Translate a cluster number to a block number */
316 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
317 /* Translate # of blks to # of clusters */
318 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
319                                  (sbi)->s_cluster_bits)
320 /* Mask out the low bits to get the starting block of the cluster */
321 #define EXT4_PBLK_CMASK(s, pblk) ((pblk) &                              \
322                                   ~((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
323 #define EXT4_LBLK_CMASK(s, lblk) ((lblk) &                              \
324                                   ~((ext4_lblk_t) (s)->s_cluster_ratio - 1))
325 /* Fill in the low bits to get the last block of the cluster */
326 #define EXT4_LBLK_CFILL(sbi, lblk) ((lblk) |                            \
327                                     ((ext4_lblk_t) (sbi)->s_cluster_ratio - 1))
328 /* Get the cluster offset */
329 #define EXT4_PBLK_COFF(s, pblk) ((pblk) &                               \
330                                  ((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
331 #define EXT4_LBLK_COFF(s, lblk) ((lblk) &                               \
332                                  ((ext4_lblk_t) (s)->s_cluster_ratio - 1))
333
334 /*
335  * Structure of a blocks group descriptor
336  */
337 struct ext4_group_desc
338 {
339         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
340         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
341         __le32  bg_inode_table_lo;      /* Inodes table block */
342         __le16  bg_free_blocks_count_lo;/* Free blocks count */
343         __le16  bg_free_inodes_count_lo;/* Free inodes count */
344         __le16  bg_used_dirs_count_lo;  /* Directories count */
345         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
346         __le32  bg_exclude_bitmap_lo;   /* Exclude bitmap for snapshots */
347         __le16  bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
348         __le16  bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
349         __le16  bg_itable_unused_lo;    /* Unused inodes count */
350         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
351         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
352         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
353         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
354         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
355         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
356         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
357         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
358         __le32  bg_exclude_bitmap_hi;   /* Exclude bitmap block MSB */
359         __le16  bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
360         __le16  bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
361         __u32   bg_reserved;
362 };
363
364 #define EXT4_BG_INODE_BITMAP_CSUM_HI_END        \
365         (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
366          sizeof(__le16))
367 #define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END        \
368         (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
369          sizeof(__le16))
370
371 /*
372  * Structure of a flex block group info
373  */
374
375 struct flex_groups {
376         atomic64_t      free_clusters;
377         atomic_t        free_inodes;
378         atomic_t        used_dirs;
379 };
380
381 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
382 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
383 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
384
385 /*
386  * Macro-instructions used to manage group descriptors
387  */
388 #define EXT4_MIN_DESC_SIZE              32
389 #define EXT4_MIN_DESC_SIZE_64BIT        64
390 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
391 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
392 #ifdef __KERNEL__
393 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
394 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
395 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
396 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
397 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
398 #else
399 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
400 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
401 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
402 #endif
403
404 /*
405  * Constants relative to the data blocks
406  */
407 #define EXT4_NDIR_BLOCKS                12
408 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
409 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
410 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
411 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
412
413 /*
414  * Inode flags
415  */
416 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
417 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
418 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
419 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
420 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
421 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
422 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
423 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
424 /* Reserved for compression usage... */
425 #define EXT4_DIRTY_FL                   0x00000100
426 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
427 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
428         /* nb: was previously EXT2_ECOMPR_FL */
429 #define EXT4_ENCRYPT_FL                 0x00000800 /* encrypted file */
430 /* End compression flags --- maybe not all used */
431 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
432 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
433 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
434 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
435 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
436 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
437 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
438 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
439 #define EXT4_VERITY_FL                  0x00100000 /* Verity protected inode */
440 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
441 /* 0x00400000 was formerly EXT4_EOFBLOCKS_FL */
442
443 #define EXT4_DAX_FL                     0x02000000 /* Inode is DAX */
444
445 #define EXT4_INLINE_DATA_FL             0x10000000 /* Inode has inline data. */
446 #define EXT4_PROJINHERIT_FL             0x20000000 /* Create with parents projid */
447 #define EXT4_CASEFOLD_FL                0x40000000 /* Casefolded directory */
448 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
449
450 /* User modifiable flags */
451 #define EXT4_FL_USER_MODIFIABLE         (EXT4_SECRM_FL | \
452                                          EXT4_UNRM_FL | \
453                                          EXT4_COMPR_FL | \
454                                          EXT4_SYNC_FL | \
455                                          EXT4_IMMUTABLE_FL | \
456                                          EXT4_APPEND_FL | \
457                                          EXT4_NODUMP_FL | \
458                                          EXT4_NOATIME_FL | \
459                                          EXT4_JOURNAL_DATA_FL | \
460                                          EXT4_NOTAIL_FL | \
461                                          EXT4_DIRSYNC_FL | \
462                                          EXT4_TOPDIR_FL | \
463                                          EXT4_EXTENTS_FL | \
464                                          0x00400000 /* EXT4_EOFBLOCKS_FL */ | \
465                                          EXT4_DAX_FL | \
466                                          EXT4_PROJINHERIT_FL | \
467                                          EXT4_CASEFOLD_FL)
468
469 /* User visible flags */
470 #define EXT4_FL_USER_VISIBLE            (EXT4_FL_USER_MODIFIABLE | \
471                                          EXT4_DIRTY_FL | \
472                                          EXT4_COMPRBLK_FL | \
473                                          EXT4_NOCOMPR_FL | \
474                                          EXT4_ENCRYPT_FL | \
475                                          EXT4_INDEX_FL | \
476                                          EXT4_VERITY_FL | \
477                                          EXT4_INLINE_DATA_FL)
478
479 /* Flags that should be inherited by new inodes from their parent. */
480 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
481                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
482                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
483                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL |\
484                            EXT4_PROJINHERIT_FL | EXT4_CASEFOLD_FL |\
485                            EXT4_DAX_FL)
486
487 /* Flags that are appropriate for regular files (all but dir-specific ones). */
488 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL | EXT4_CASEFOLD_FL |\
489                            EXT4_PROJINHERIT_FL))
490
491 /* Flags that are appropriate for non-directories/regular files. */
492 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
493
494 /* The only flags that should be swapped */
495 #define EXT4_FL_SHOULD_SWAP (EXT4_HUGE_FILE_FL | EXT4_EXTENTS_FL)
496
497 /* Flags which are mutually exclusive to DAX */
498 #define EXT4_DAX_MUT_EXCL (EXT4_VERITY_FL | EXT4_ENCRYPT_FL |\
499                            EXT4_JOURNAL_DATA_FL | EXT4_INLINE_DATA_FL)
500
501 /* Mask out flags that are inappropriate for the given type of inode. */
502 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
503 {
504         if (S_ISDIR(mode))
505                 return flags;
506         else if (S_ISREG(mode))
507                 return flags & EXT4_REG_FLMASK;
508         else
509                 return flags & EXT4_OTHER_FLMASK;
510 }
511
512 /*
513  * Inode flags used for atomic set/get
514  */
515 enum {
516         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
517         EXT4_INODE_UNRM         = 1,    /* Undelete */
518         EXT4_INODE_COMPR        = 2,    /* Compress file */
519         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
520         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
521         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
522         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
523         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
524 /* Reserved for compression usage... */
525         EXT4_INODE_DIRTY        = 8,
526         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
527         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
528         EXT4_INODE_ENCRYPT      = 11,   /* Encrypted file */
529 /* End compression flags --- maybe not all used */
530         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
531         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
532         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
533         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
534         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
535         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
536         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
537         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
538         EXT4_INODE_VERITY       = 20,   /* Verity protected inode */
539         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
540 /* 22 was formerly EXT4_INODE_EOFBLOCKS */
541         EXT4_INODE_DAX          = 25,   /* Inode is DAX */
542         EXT4_INODE_INLINE_DATA  = 28,   /* Data in inode. */
543         EXT4_INODE_PROJINHERIT  = 29,   /* Create with parents projid */
544         EXT4_INODE_CASEFOLD     = 30,   /* Casefolded directory */
545         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
546 };
547
548 /*
549  * Since it's pretty easy to mix up bit numbers and hex values, we use a
550  * build-time check to make sure that EXT4_XXX_FL is consistent with respect to
551  * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost
552  * any extra space in the compiled kernel image, otherwise, the build will fail.
553  * It's important that these values are the same, since we are using
554  * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent
555  * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk
556  * values found in ext2, ext3 and ext4 filesystems, and of course the values
557  * defined in e2fsprogs.
558  *
559  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
560  */
561 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
562 #define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG))
563
564 static inline void ext4_check_flag_values(void)
565 {
566         CHECK_FLAG_VALUE(SECRM);
567         CHECK_FLAG_VALUE(UNRM);
568         CHECK_FLAG_VALUE(COMPR);
569         CHECK_FLAG_VALUE(SYNC);
570         CHECK_FLAG_VALUE(IMMUTABLE);
571         CHECK_FLAG_VALUE(APPEND);
572         CHECK_FLAG_VALUE(NODUMP);
573         CHECK_FLAG_VALUE(NOATIME);
574         CHECK_FLAG_VALUE(DIRTY);
575         CHECK_FLAG_VALUE(COMPRBLK);
576         CHECK_FLAG_VALUE(NOCOMPR);
577         CHECK_FLAG_VALUE(ENCRYPT);
578         CHECK_FLAG_VALUE(INDEX);
579         CHECK_FLAG_VALUE(IMAGIC);
580         CHECK_FLAG_VALUE(JOURNAL_DATA);
581         CHECK_FLAG_VALUE(NOTAIL);
582         CHECK_FLAG_VALUE(DIRSYNC);
583         CHECK_FLAG_VALUE(TOPDIR);
584         CHECK_FLAG_VALUE(HUGE_FILE);
585         CHECK_FLAG_VALUE(EXTENTS);
586         CHECK_FLAG_VALUE(VERITY);
587         CHECK_FLAG_VALUE(EA_INODE);
588         CHECK_FLAG_VALUE(INLINE_DATA);
589         CHECK_FLAG_VALUE(PROJINHERIT);
590         CHECK_FLAG_VALUE(CASEFOLD);
591         CHECK_FLAG_VALUE(RESERVED);
592 }
593
594 /* Used to pass group descriptor data when online resize is done */
595 struct ext4_new_group_input {
596         __u32 group;            /* Group number for this data */
597         __u64 block_bitmap;     /* Absolute block number of block bitmap */
598         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
599         __u64 inode_table;      /* Absolute block number of inode table start */
600         __u32 blocks_count;     /* Total number of blocks in this group */
601         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
602         __u16 unused;
603 };
604
605 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
606 struct compat_ext4_new_group_input {
607         u32 group;
608         compat_u64 block_bitmap;
609         compat_u64 inode_bitmap;
610         compat_u64 inode_table;
611         u32 blocks_count;
612         u16 reserved_blocks;
613         u16 unused;
614 };
615 #endif
616
617 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
618 struct ext4_new_group_data {
619         __u32 group;
620         __u64 block_bitmap;
621         __u64 inode_bitmap;
622         __u64 inode_table;
623         __u32 blocks_count;
624         __u16 reserved_blocks;
625         __u16 mdata_blocks;
626         __u32 free_clusters_count;
627 };
628
629 /* Indexes used to index group tables in ext4_new_group_data */
630 enum {
631         BLOCK_BITMAP = 0,       /* block bitmap */
632         INODE_BITMAP,           /* inode bitmap */
633         INODE_TABLE,            /* inode tables */
634         GROUP_TABLE_COUNT,
635 };
636
637 /*
638  * Flags used by ext4_map_blocks()
639  */
640         /* Allocate any needed blocks and/or convert an unwritten
641            extent to be an initialized ext4 */
642 #define EXT4_GET_BLOCKS_CREATE                  0x0001
643         /* Request the creation of an unwritten extent */
644 #define EXT4_GET_BLOCKS_UNWRIT_EXT              0x0002
645 #define EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT       (EXT4_GET_BLOCKS_UNWRIT_EXT|\
646                                                  EXT4_GET_BLOCKS_CREATE)
647         /* Caller is from the delayed allocation writeout path
648          * finally doing the actual allocation of delayed blocks */
649 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
650         /* caller is from the direct IO path, request to creation of an
651         unwritten extents if not allocated, split the unwritten
652         extent if blocks has been preallocated already*/
653 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
654 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
655 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
656                                          EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
657         /* Convert extent to initialized after IO complete */
658 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
659                                          EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
660         /* Eventual metadata allocation (due to growing extent tree)
661          * should not fail, so try to use reserved blocks for that.*/
662 #define EXT4_GET_BLOCKS_METADATA_NOFAIL         0x0020
663         /* Don't normalize allocation size (used for fallocate) */
664 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
665         /* Convert written extents to unwritten */
666 #define EXT4_GET_BLOCKS_CONVERT_UNWRITTEN       0x0100
667         /* Write zeros to newly created written extents */
668 #define EXT4_GET_BLOCKS_ZERO                    0x0200
669 #define EXT4_GET_BLOCKS_CREATE_ZERO             (EXT4_GET_BLOCKS_CREATE |\
670                                         EXT4_GET_BLOCKS_ZERO)
671         /* Caller will submit data before dropping transaction handle. This
672          * allows jbd2 to avoid submitting data before commit. */
673 #define EXT4_GET_BLOCKS_IO_SUBMIT               0x0400
674         /* Caller is in the atomic contex, find extent if it has been cached */
675 #define EXT4_GET_BLOCKS_CACHED_NOWAIT           0x0800
676
677 /*
678  * The bit position of these flags must not overlap with any of the
679  * EXT4_GET_BLOCKS_*.  They are used by ext4_find_extent(),
680  * read_extent_tree_block(), ext4_split_extent_at(),
681  * ext4_ext_insert_extent(), and ext4_ext_create_new_leaf().
682  * EXT4_EX_NOCACHE is used to indicate that the we shouldn't be
683  * caching the extents when reading from the extent tree while a
684  * truncate or punch hole operation is in progress.
685  */
686 #define EXT4_EX_NOCACHE                         0x40000000
687 #define EXT4_EX_FORCE_CACHE                     0x20000000
688 #define EXT4_EX_NOFAIL                          0x10000000
689
690 /*
691  * Flags used by ext4_free_blocks
692  */
693 #define EXT4_FREE_BLOCKS_METADATA               0x0001
694 #define EXT4_FREE_BLOCKS_FORGET                 0x0002
695 #define EXT4_FREE_BLOCKS_VALIDATED              0x0004
696 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE         0x0008
697 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
698 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
699 #define EXT4_FREE_BLOCKS_RERESERVE_CLUSTER      0x0040
700
701 /*
702  * ioctl commands
703  */
704 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
705 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
706 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
707 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
708 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
709 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
710 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
711 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
712 #define EXT4_IOC_MIGRATE                _IO('f', 9)
713  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
714  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
715 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
716 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
717 #define EXT4_IOC_RESIZE_FS              _IOW('f', 16, __u64)
718 #define EXT4_IOC_SWAP_BOOT              _IO('f', 17)
719 #define EXT4_IOC_PRECACHE_EXTENTS       _IO('f', 18)
720 /* ioctl codes 19--39 are reserved for fscrypt */
721 #define EXT4_IOC_CLEAR_ES_CACHE         _IO('f', 40)
722 #define EXT4_IOC_GETSTATE               _IOW('f', 41, __u32)
723 #define EXT4_IOC_GET_ES_CACHE           _IOWR('f', 42, struct fiemap)
724 #define EXT4_IOC_CHECKPOINT             _IOW('f', 43, __u32)
725
726 #define EXT4_IOC_SHUTDOWN _IOR ('X', 125, __u32)
727
728 /*
729  * Flags for going down operation
730  */
731 #define EXT4_GOING_FLAGS_DEFAULT                0x0     /* going down */
732 #define EXT4_GOING_FLAGS_LOGFLUSH               0x1     /* flush log but not data */
733 #define EXT4_GOING_FLAGS_NOLOGFLUSH             0x2     /* don't flush log nor data */
734
735 /*
736  * Flags returned by EXT4_IOC_GETSTATE
737  *
738  * We only expose to userspace a subset of the state flags in
739  * i_state_flags
740  */
741 #define EXT4_STATE_FLAG_EXT_PRECACHED   0x00000001
742 #define EXT4_STATE_FLAG_NEW             0x00000002
743 #define EXT4_STATE_FLAG_NEWENTRY        0x00000004
744 #define EXT4_STATE_FLAG_DA_ALLOC_CLOSE  0x00000008
745
746 /* flags for ioctl EXT4_IOC_CHECKPOINT */
747 #define EXT4_IOC_CHECKPOINT_FLAG_DISCARD        0x1
748 #define EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT        0x2
749 #define EXT4_IOC_CHECKPOINT_FLAG_DRY_RUN        0x4
750 #define EXT4_IOC_CHECKPOINT_FLAG_VALID          (EXT4_IOC_CHECKPOINT_FLAG_DISCARD | \
751                                                 EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT | \
752                                                 EXT4_IOC_CHECKPOINT_FLAG_DRY_RUN)
753
754 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
755 /*
756  * ioctl commands in 32 bit emulation
757  */
758 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
759 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
760 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
761 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
762 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
763 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
764 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
765 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
766 #endif
767
768 /*
769  * Returned by EXT4_IOC_GET_ES_CACHE as an additional possible flag.
770  * It indicates that the entry in extent status cache is for a hole.
771  */
772 #define EXT4_FIEMAP_EXTENT_HOLE         0x08000000
773
774 /* Max physical block we can address w/o extents */
775 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
776
777 /* Max logical block we can support */
778 #define EXT4_MAX_LOGICAL_BLOCK          0xFFFFFFFE
779
780 /*
781  * Structure of an inode on the disk
782  */
783 struct ext4_inode {
784         __le16  i_mode;         /* File mode */
785         __le16  i_uid;          /* Low 16 bits of Owner Uid */
786         __le32  i_size_lo;      /* Size in bytes */
787         __le32  i_atime;        /* Access time */
788         __le32  i_ctime;        /* Inode Change time */
789         __le32  i_mtime;        /* Modification time */
790         __le32  i_dtime;        /* Deletion Time */
791         __le16  i_gid;          /* Low 16 bits of Group Id */
792         __le16  i_links_count;  /* Links count */
793         __le32  i_blocks_lo;    /* Blocks count */
794         __le32  i_flags;        /* File flags */
795         union {
796                 struct {
797                         __le32  l_i_version;
798                 } linux1;
799                 struct {
800                         __u32  h_i_translator;
801                 } hurd1;
802                 struct {
803                         __u32  m_i_reserved1;
804                 } masix1;
805         } osd1;                         /* OS dependent 1 */
806         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
807         __le32  i_generation;   /* File version (for NFS) */
808         __le32  i_file_acl_lo;  /* File ACL */
809         __le32  i_size_high;
810         __le32  i_obso_faddr;   /* Obsoleted fragment address */
811         union {
812                 struct {
813                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
814                         __le16  l_i_file_acl_high;
815                         __le16  l_i_uid_high;   /* these 2 fields */
816                         __le16  l_i_gid_high;   /* were reserved2[0] */
817                         __le16  l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
818                         __le16  l_i_reserved;
819                 } linux2;
820                 struct {
821                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
822                         __u16   h_i_mode_high;
823                         __u16   h_i_uid_high;
824                         __u16   h_i_gid_high;
825                         __u32   h_i_author;
826                 } hurd2;
827                 struct {
828                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
829                         __le16  m_i_file_acl_high;
830                         __u32   m_i_reserved2[2];
831                 } masix2;
832         } osd2;                         /* OS dependent 2 */
833         __le16  i_extra_isize;
834         __le16  i_checksum_hi;  /* crc32c(uuid+inum+inode) BE */
835         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
836         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
837         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
838         __le32  i_crtime;       /* File Creation time */
839         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
840         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
841         __le32  i_projid;       /* Project ID */
842 };
843
844 struct move_extent {
845         __u32 reserved;         /* should be zero */
846         __u32 donor_fd;         /* donor file descriptor */
847         __u64 orig_start;       /* logical start offset in block for orig */
848         __u64 donor_start;      /* logical start offset in block for donor */
849         __u64 len;              /* block length to be moved */
850         __u64 moved_len;        /* moved block length */
851 };
852
853 #define EXT4_EPOCH_BITS 2
854 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
855 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
856
857 /*
858  * Extended fields will fit into an inode if the filesystem was formatted
859  * with large inodes (-I 256 or larger) and there are not currently any EAs
860  * consuming all of the available space. For new inodes we always reserve
861  * enough space for the kernel's known extended fields, but for inodes
862  * created with an old kernel this might not have been the case. None of
863  * the extended inode fields is critical for correct filesystem operation.
864  * This macro checks if a certain field fits in the inode. Note that
865  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
866  */
867 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
868         ((offsetof(typeof(*ext4_inode), field) +        \
869           sizeof((ext4_inode)->field))                  \
870         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
871             (einode)->i_extra_isize))                   \
872
873 /*
874  * We use an encoding that preserves the times for extra epoch "00":
875  *
876  * extra  msb of                         adjust for signed
877  * epoch  32-bit                         32-bit tv_sec to
878  * bits   time    decoded 64-bit tv_sec  64-bit tv_sec      valid time range
879  * 0 0    1    -0x80000000..-0x00000001  0x000000000 1901-12-13..1969-12-31
880  * 0 0    0    0x000000000..0x07fffffff  0x000000000 1970-01-01..2038-01-19
881  * 0 1    1    0x080000000..0x0ffffffff  0x100000000 2038-01-19..2106-02-07
882  * 0 1    0    0x100000000..0x17fffffff  0x100000000 2106-02-07..2174-02-25
883  * 1 0    1    0x180000000..0x1ffffffff  0x200000000 2174-02-25..2242-03-16
884  * 1 0    0    0x200000000..0x27fffffff  0x200000000 2242-03-16..2310-04-04
885  * 1 1    1    0x280000000..0x2ffffffff  0x300000000 2310-04-04..2378-04-22
886  * 1 1    0    0x300000000..0x37fffffff  0x300000000 2378-04-22..2446-05-10
887  *
888  * Note that previous versions of the kernel on 64-bit systems would
889  * incorrectly use extra epoch bits 1,1 for dates between 1901 and
890  * 1970.  e2fsck will correct this, assuming that it is run on the
891  * affected filesystem before 2242.
892  */
893
894 static inline __le32 ext4_encode_extra_time(struct timespec64 *time)
895 {
896         u32 extra =((time->tv_sec - (s32)time->tv_sec) >> 32) & EXT4_EPOCH_MASK;
897         return cpu_to_le32(extra | (time->tv_nsec << EXT4_EPOCH_BITS));
898 }
899
900 static inline void ext4_decode_extra_time(struct timespec64 *time,
901                                           __le32 extra)
902 {
903         if (unlikely(extra & cpu_to_le32(EXT4_EPOCH_MASK)))
904                 time->tv_sec += (u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK) << 32;
905         time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
906 }
907
908 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                           \
909 do {                                                                            \
910         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     {\
911                 (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);        \
912                 (raw_inode)->xtime ## _extra =                                  \
913                                 ext4_encode_extra_time(&(inode)->xtime);        \
914                 }                                                               \
915         else    \
916                 (raw_inode)->xtime = cpu_to_le32(clamp_t(int32_t, (inode)->xtime.tv_sec, S32_MIN, S32_MAX));    \
917 } while (0)
918
919 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
920 do {                                                                           \
921         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
922                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
923         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
924                 (raw_inode)->xtime ## _extra =                                 \
925                                 ext4_encode_extra_time(&(einode)->xtime);      \
926 } while (0)
927
928 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                           \
929 do {                                                                            \
930         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);        \
931         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) {    \
932                 ext4_decode_extra_time(&(inode)->xtime,                         \
933                                        raw_inode->xtime ## _extra);             \
934                 }                                                               \
935         else                                                                    \
936                 (inode)->xtime.tv_nsec = 0;                                     \
937 } while (0)
938
939
940 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
941 do {                                                                           \
942         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
943                 (einode)->xtime.tv_sec =                                       \
944                         (signed)le32_to_cpu((raw_inode)->xtime);               \
945         else                                                                   \
946                 (einode)->xtime.tv_sec = 0;                                    \
947         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
948                 ext4_decode_extra_time(&(einode)->xtime,                       \
949                                        raw_inode->xtime ## _extra);            \
950         else                                                                   \
951                 (einode)->xtime.tv_nsec = 0;                                   \
952 } while (0)
953
954 #define i_disk_version osd1.linux1.l_i_version
955
956 #if defined(__KERNEL__) || defined(__linux__)
957 #define i_reserved1     osd1.linux1.l_i_reserved1
958 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
959 #define i_blocks_high   osd2.linux2.l_i_blocks_high
960 #define i_uid_low       i_uid
961 #define i_gid_low       i_gid
962 #define i_uid_high      osd2.linux2.l_i_uid_high
963 #define i_gid_high      osd2.linux2.l_i_gid_high
964 #define i_checksum_lo   osd2.linux2.l_i_checksum_lo
965
966 #elif defined(__GNU__)
967
968 #define i_translator    osd1.hurd1.h_i_translator
969 #define i_uid_high      osd2.hurd2.h_i_uid_high
970 #define i_gid_high      osd2.hurd2.h_i_gid_high
971 #define i_author        osd2.hurd2.h_i_author
972
973 #elif defined(__masix__)
974
975 #define i_reserved1     osd1.masix1.m_i_reserved1
976 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
977 #define i_reserved2     osd2.masix2.m_i_reserved2
978
979 #endif /* defined(__KERNEL__) || defined(__linux__) */
980
981 #include "extents_status.h"
982 #include "fast_commit.h"
983
984 /*
985  * Lock subclasses for i_data_sem in the ext4_inode_info structure.
986  *
987  * These are needed to avoid lockdep false positives when we need to
988  * allocate blocks to the quota inode during ext4_map_blocks(), while
989  * holding i_data_sem for a normal (non-quota) inode.  Since we don't
990  * do quota tracking for the quota inode, this avoids deadlock (as
991  * well as infinite recursion, since it isn't turtles all the way
992  * down...)
993  *
994  *  I_DATA_SEM_NORMAL - Used for most inodes
995  *  I_DATA_SEM_OTHER  - Used by move_inode.c for the second normal inode
996  *                        where the second inode has larger inode number
997  *                        than the first
998  *  I_DATA_SEM_QUOTA  - Used for quota inodes only
999  */
1000 enum {
1001         I_DATA_SEM_NORMAL = 0,
1002         I_DATA_SEM_OTHER,
1003         I_DATA_SEM_QUOTA,
1004 };
1005
1006
1007 /*
1008  * fourth extended file system inode data in memory
1009  */
1010 struct ext4_inode_info {
1011         __le32  i_data[15];     /* unconverted */
1012         __u32   i_dtime;
1013         ext4_fsblk_t    i_file_acl;
1014
1015         /*
1016          * i_block_group is the number of the block group which contains
1017          * this file's inode.  Constant across the lifetime of the inode,
1018          * it is used for making block allocation decisions - we try to
1019          * place a file's data blocks near its inode block, and new inodes
1020          * near to their parent directory's inode.
1021          */
1022         ext4_group_t    i_block_group;
1023         ext4_lblk_t     i_dir_start_lookup;
1024 #if (BITS_PER_LONG < 64)
1025         unsigned long   i_state_flags;          /* Dynamic state flags */
1026 #endif
1027         unsigned long   i_flags;
1028
1029         /*
1030          * Extended attributes can be read independently of the main file
1031          * data. Taking i_rwsem even when reading would cause contention
1032          * between readers of EAs and writers of regular file data, so
1033          * instead we synchronize on xattr_sem when reading or changing
1034          * EAs.
1035          */
1036         struct rw_semaphore xattr_sem;
1037
1038         /*
1039          * Inodes with EXT4_STATE_ORPHAN_FILE use i_orphan_idx. Otherwise
1040          * i_orphan is used.
1041          */
1042         union {
1043                 struct list_head i_orphan;      /* unlinked but open inodes */
1044                 unsigned int i_orphan_idx;      /* Index in orphan file */
1045         };
1046
1047         /* Fast commit related info */
1048
1049         /* For tracking dentry create updates */
1050         struct list_head i_fc_dilist;
1051         struct list_head i_fc_list;     /*
1052                                          * inodes that need fast commit
1053                                          * protected by sbi->s_fc_lock.
1054                                          */
1055
1056         /* Start of lblk range that needs to be committed in this fast commit */
1057         ext4_lblk_t i_fc_lblk_start;
1058
1059         /* End of lblk range that needs to be committed in this fast commit */
1060         ext4_lblk_t i_fc_lblk_len;
1061
1062         /* Number of ongoing updates on this inode */
1063         atomic_t  i_fc_updates;
1064
1065         /* Fast commit wait queue for this inode */
1066         wait_queue_head_t i_fc_wait;
1067
1068         /* Protect concurrent accesses on i_fc_lblk_start, i_fc_lblk_len */
1069         struct mutex i_fc_lock;
1070
1071         /*
1072          * i_disksize keeps track of what the inode size is ON DISK, not
1073          * in memory.  During truncate, i_size is set to the new size by
1074          * the VFS prior to calling ext4_truncate(), but the filesystem won't
1075          * set i_disksize to 0 until the truncate is actually under way.
1076          *
1077          * The intent is that i_disksize always represents the blocks which
1078          * are used by this file.  This allows recovery to restart truncate
1079          * on orphans if we crash during truncate.  We actually write i_disksize
1080          * into the on-disk inode when writing inodes out, instead of i_size.
1081          *
1082          * The only time when i_disksize and i_size may be different is when
1083          * a truncate is in progress.  The only things which change i_disksize
1084          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
1085          */
1086         loff_t  i_disksize;
1087
1088         /*
1089          * i_data_sem is for serialising ext4_truncate() against
1090          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
1091          * data tree are chopped off during truncate. We can't do that in
1092          * ext4 because whenever we perform intermediate commits during
1093          * truncate, the inode and all the metadata blocks *must* be in a
1094          * consistent state which allows truncation of the orphans to restart
1095          * during recovery.  Hence we must fix the get_block-vs-truncate race
1096          * by other means, so we have i_data_sem.
1097          */
1098         struct rw_semaphore i_data_sem;
1099         struct inode vfs_inode;
1100         struct jbd2_inode *jinode;
1101
1102         spinlock_t i_raw_lock;  /* protects updates to the raw inode */
1103
1104         /*
1105          * File creation time. Its function is same as that of
1106          * struct timespec64 i_{a,c,m}time in the generic inode.
1107          */
1108         struct timespec64 i_crtime;
1109
1110         /* mballoc */
1111         atomic_t i_prealloc_active;
1112         struct list_head i_prealloc_list;
1113         spinlock_t i_prealloc_lock;
1114
1115         /* extents status tree */
1116         struct ext4_es_tree i_es_tree;
1117         rwlock_t i_es_lock;
1118         struct list_head i_es_list;
1119         unsigned int i_es_all_nr;       /* protected by i_es_lock */
1120         unsigned int i_es_shk_nr;       /* protected by i_es_lock */
1121         ext4_lblk_t i_es_shrink_lblk;   /* Offset where we start searching for
1122                                            extents to shrink. Protected by
1123                                            i_es_lock  */
1124
1125         /* ialloc */
1126         ext4_group_t    i_last_alloc_group;
1127
1128         /* allocation reservation info for delalloc */
1129         /* In case of bigalloc, this refer to clusters rather than blocks */
1130         unsigned int i_reserved_data_blocks;
1131
1132         /* pending cluster reservations for bigalloc file systems */
1133         struct ext4_pending_tree i_pending_tree;
1134
1135         /* on-disk additional length */
1136         __u16 i_extra_isize;
1137
1138         /* Indicate the inline data space. */
1139         u16 i_inline_off;
1140         u16 i_inline_size;
1141
1142 #ifdef CONFIG_QUOTA
1143         /* quota space reservation, managed internally by quota code */
1144         qsize_t i_reserved_quota;
1145 #endif
1146
1147         /* Lock protecting lists below */
1148         spinlock_t i_completed_io_lock;
1149         /*
1150          * Completed IOs that need unwritten extents handling and have
1151          * transaction reserved
1152          */
1153         struct list_head i_rsv_conversion_list;
1154         struct work_struct i_rsv_conversion_work;
1155         atomic_t i_unwritten; /* Nr. of inflight conversions pending */
1156
1157         spinlock_t i_block_reservation_lock;
1158
1159         /*
1160          * Transactions that contain inode's metadata needed to complete
1161          * fsync and fdatasync, respectively.
1162          */
1163         tid_t i_sync_tid;
1164         tid_t i_datasync_tid;
1165
1166 #ifdef CONFIG_QUOTA
1167         struct dquot *i_dquot[MAXQUOTAS];
1168 #endif
1169
1170         /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
1171         __u32 i_csum_seed;
1172
1173         kprojid_t i_projid;
1174 };
1175
1176 /*
1177  * File system states
1178  */
1179 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
1180 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
1181 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
1182 #define EXT4_FC_REPLAY                  0x0020  /* Fast commit replay ongoing */
1183
1184 /*
1185  * Misc. filesystem flags
1186  */
1187 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
1188 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
1189 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
1190
1191 /*
1192  * Mount flags set via mount options or defaults
1193  */
1194 #define EXT4_MOUNT_NO_MBCACHE           0x00001 /* Do not use mbcache */
1195 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
1196 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
1197 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
1198 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
1199 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
1200 #define EXT4_MOUNT_ERRORS_MASK          0x00070
1201 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
1202 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
1203 #ifdef CONFIG_FS_DAX
1204 #define EXT4_MOUNT_DAX_ALWAYS           0x00200 /* Direct Access */
1205 #else
1206 #define EXT4_MOUNT_DAX_ALWAYS           0
1207 #endif
1208 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
1209 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
1210 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
1211 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
1212 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
1213 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
1214 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
1215 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
1216 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
1217 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
1218 #define EXT4_MOUNT_QUOTA                0x40000 /* Some quota option set */
1219 #define EXT4_MOUNT_USRQUOTA             0x80000 /* "old" user quota,
1220                                                  * enable enforcement for hidden
1221                                                  * quota files */
1222 #define EXT4_MOUNT_GRPQUOTA             0x100000 /* "old" group quota, enable
1223                                                   * enforcement for hidden quota
1224                                                   * files */
1225 #define EXT4_MOUNT_PRJQUOTA             0x200000 /* Enable project quota
1226                                                   * enforcement */
1227 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
1228 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
1229 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
1230 #define EXT4_MOUNT_WARN_ON_ERROR        0x2000000 /* Trigger WARN_ON on error */
1231 #define EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS 0x4000000
1232 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
1233 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
1234 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
1235 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
1236 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
1237
1238 /*
1239  * Mount flags set either automatically (could not be set by mount option)
1240  * based on per file system feature or property or in special cases such as
1241  * distinguishing between explicit mount option definition and default.
1242  */
1243 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
1244                                                       specified delalloc */
1245 #define EXT4_MOUNT2_STD_GROUP_SIZE      0x00000002 /* We have standard group
1246                                                       size of blocksize * 8
1247                                                       blocks */
1248 #define EXT4_MOUNT2_HURD_COMPAT         0x00000004 /* Support HURD-castrated
1249                                                       file systems */
1250 #define EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM   0x00000008 /* User explicitly
1251                                                 specified journal checksum */
1252
1253 #define EXT4_MOUNT2_JOURNAL_FAST_COMMIT 0x00000010 /* Journal fast commit */
1254 #define EXT4_MOUNT2_DAX_NEVER           0x00000020 /* Do not allow Direct Access */
1255 #define EXT4_MOUNT2_DAX_INODE           0x00000040 /* For printing options only */
1256 #define EXT4_MOUNT2_MB_OPTIMIZE_SCAN    0x00000080 /* Optimize group
1257                                                     * scanning in mballoc
1258                                                     */
1259
1260 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
1261                                                 ~EXT4_MOUNT_##opt
1262 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
1263                                                 EXT4_MOUNT_##opt
1264 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
1265                                          EXT4_MOUNT_##opt)
1266
1267 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
1268                                                 ~EXT4_MOUNT2_##opt
1269 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
1270                                                 EXT4_MOUNT2_##opt
1271 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
1272                                          EXT4_MOUNT2_##opt)
1273
1274 #define ext4_test_and_set_bit           __test_and_set_bit_le
1275 #define ext4_set_bit                    __set_bit_le
1276 #define ext4_set_bit_atomic             ext2_set_bit_atomic
1277 #define ext4_test_and_clear_bit         __test_and_clear_bit_le
1278 #define ext4_clear_bit                  __clear_bit_le
1279 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
1280 #define ext4_test_bit                   test_bit_le
1281 #define ext4_find_next_zero_bit         find_next_zero_bit_le
1282 #define ext4_find_next_bit              find_next_bit_le
1283
1284 extern void mb_set_bits(void *bm, int cur, int len);
1285
1286 /*
1287  * Maximal mount counts between two filesystem checks
1288  */
1289 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
1290 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
1291
1292 /*
1293  * Behaviour when detecting errors
1294  */
1295 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
1296 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
1297 #define EXT4_ERRORS_PANIC               3       /* Panic */
1298 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
1299
1300 /* Metadata checksum algorithm codes */
1301 #define EXT4_CRC32C_CHKSUM              1
1302
1303 #define EXT4_LABEL_MAX                  16
1304
1305 /*
1306  * Structure of the super block
1307  */
1308 struct ext4_super_block {
1309 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1310         __le32  s_blocks_count_lo;      /* Blocks count */
1311         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1312         __le32  s_free_blocks_count_lo; /* Free blocks count */
1313 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1314         __le32  s_first_data_block;     /* First Data Block */
1315         __le32  s_log_block_size;       /* Block size */
1316         __le32  s_log_cluster_size;     /* Allocation cluster size */
1317 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1318         __le32  s_clusters_per_group;   /* # Clusters per group */
1319         __le32  s_inodes_per_group;     /* # Inodes per group */
1320         __le32  s_mtime;                /* Mount time */
1321 /*30*/  __le32  s_wtime;                /* Write time */
1322         __le16  s_mnt_count;            /* Mount count */
1323         __le16  s_max_mnt_count;        /* Maximal mount count */
1324         __le16  s_magic;                /* Magic signature */
1325         __le16  s_state;                /* File system state */
1326         __le16  s_errors;               /* Behaviour when detecting errors */
1327         __le16  s_minor_rev_level;      /* minor revision level */
1328 /*40*/  __le32  s_lastcheck;            /* time of last check */
1329         __le32  s_checkinterval;        /* max. time between checks */
1330         __le32  s_creator_os;           /* OS */
1331         __le32  s_rev_level;            /* Revision level */
1332 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1333         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1334         /*
1335          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1336          *
1337          * Note: the difference between the compatible feature set and
1338          * the incompatible feature set is that if there is a bit set
1339          * in the incompatible feature set that the kernel doesn't
1340          * know about, it should refuse to mount the filesystem.
1341          *
1342          * e2fsck's requirements are more strict; if it doesn't know
1343          * about a feature in either the compatible or incompatible
1344          * feature set, it must abort and not try to meddle with
1345          * things it doesn't understand...
1346          */
1347         __le32  s_first_ino;            /* First non-reserved inode */
1348         __le16  s_inode_size;           /* size of inode structure */
1349         __le16  s_block_group_nr;       /* block group # of this superblock */
1350         __le32  s_feature_compat;       /* compatible feature set */
1351 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1352         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1353 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1354 /*78*/  char    s_volume_name[EXT4_LABEL_MAX];  /* volume name */
1355 /*88*/  char    s_last_mounted[64] __nonstring; /* directory where last mounted */
1356 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1357         /*
1358          * Performance hints.  Directory preallocation should only
1359          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1360          */
1361         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1362         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1363         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1364         /*
1365          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1366          */
1367 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1368 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1369         __le32  s_journal_dev;          /* device number of journal file */
1370         __le32  s_last_orphan;          /* start of list of inodes to delete */
1371         __le32  s_hash_seed[4];         /* HTREE hash seed */
1372         __u8    s_def_hash_version;     /* Default hash version to use */
1373         __u8    s_jnl_backup_type;
1374         __le16  s_desc_size;            /* size of group descriptor */
1375 /*100*/ __le32  s_default_mount_opts;
1376         __le32  s_first_meta_bg;        /* First metablock block group */
1377         __le32  s_mkfs_time;            /* When the filesystem was created */
1378         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1379         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1380 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1381         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1382         __le32  s_free_blocks_count_hi; /* Free blocks count */
1383         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1384         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1385         __le32  s_flags;                /* Miscellaneous flags */
1386         __le16  s_raid_stride;          /* RAID stride */
1387         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1388         __le64  s_mmp_block;            /* Block for multi-mount protection */
1389         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1390         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1391         __u8    s_checksum_type;        /* metadata checksum algorithm used */
1392         __u8    s_encryption_level;     /* versioning level for encryption */
1393         __u8    s_reserved_pad;         /* Padding to next 32bits */
1394         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1395         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1396         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1397         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1398                                               snapshot's future use */
1399         __le32  s_snapshot_list;        /* inode number of the head of the
1400                                            on-disk snapshot list */
1401 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1402         __le32  s_error_count;          /* number of fs errors */
1403         __le32  s_first_error_time;     /* first time an error happened */
1404         __le32  s_first_error_ino;      /* inode involved in first error */
1405         __le64  s_first_error_block;    /* block involved of first error */
1406         __u8    s_first_error_func[32] __nonstring;     /* function where the error happened */
1407         __le32  s_first_error_line;     /* line number where error happened */
1408         __le32  s_last_error_time;      /* most recent time of an error */
1409         __le32  s_last_error_ino;       /* inode involved in last error */
1410         __le32  s_last_error_line;      /* line number where error happened */
1411         __le64  s_last_error_block;     /* block involved of last error */
1412         __u8    s_last_error_func[32] __nonstring;      /* function where the error happened */
1413 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1414         __u8    s_mount_opts[64];
1415         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1416         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1417         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1418         __le32  s_backup_bgs[2];        /* groups with sparse_super2 SBs */
1419         __u8    s_encrypt_algos[4];     /* Encryption algorithms in use  */
1420         __u8    s_encrypt_pw_salt[16];  /* Salt used for string2key algorithm */
1421         __le32  s_lpf_ino;              /* Location of the lost+found inode */
1422         __le32  s_prj_quota_inum;       /* inode for tracking project quota */
1423         __le32  s_checksum_seed;        /* crc32c(uuid) if csum_seed set */
1424         __u8    s_wtime_hi;
1425         __u8    s_mtime_hi;
1426         __u8    s_mkfs_time_hi;
1427         __u8    s_lastcheck_hi;
1428         __u8    s_first_error_time_hi;
1429         __u8    s_last_error_time_hi;
1430         __u8    s_first_error_errcode;
1431         __u8    s_last_error_errcode;
1432         __le16  s_encoding;             /* Filename charset encoding */
1433         __le16  s_encoding_flags;       /* Filename charset encoding flags */
1434         __le32  s_orphan_file_inum;     /* Inode for tracking orphan inodes */
1435         __le32  s_reserved[94];         /* Padding to the end of the block */
1436         __le32  s_checksum;             /* crc32c(superblock) */
1437 };
1438
1439 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1440
1441 #ifdef __KERNEL__
1442
1443 /* Number of quota types we support */
1444 #define EXT4_MAXQUOTAS 3
1445
1446 #define EXT4_ENC_UTF8_12_1      1
1447
1448 /* Types of ext4 journal triggers */
1449 enum ext4_journal_trigger_type {
1450         EXT4_JTR_ORPHAN_FILE,
1451         EXT4_JTR_NONE   /* This must be the last entry for indexing to work! */
1452 };
1453
1454 #define EXT4_JOURNAL_TRIGGER_COUNT EXT4_JTR_NONE
1455
1456 struct ext4_journal_trigger {
1457         struct jbd2_buffer_trigger_type tr_triggers;
1458         struct super_block *sb;
1459 };
1460
1461 static inline struct ext4_journal_trigger *EXT4_TRIGGER(
1462                                 struct jbd2_buffer_trigger_type *trigger)
1463 {
1464         return container_of(trigger, struct ext4_journal_trigger, tr_triggers);
1465 }
1466
1467 #define EXT4_ORPHAN_BLOCK_MAGIC 0x0b10ca04
1468
1469 /* Structure at the tail of orphan block */
1470 struct ext4_orphan_block_tail {
1471         __le32 ob_magic;
1472         __le32 ob_checksum;
1473 };
1474
1475 static inline int ext4_inodes_per_orphan_block(struct super_block *sb)
1476 {
1477         return (sb->s_blocksize - sizeof(struct ext4_orphan_block_tail)) /
1478                         sizeof(u32);
1479 }
1480
1481 struct ext4_orphan_block {
1482         atomic_t ob_free_entries;       /* Number of free orphan entries in block */
1483         struct buffer_head *ob_bh;      /* Buffer for orphan block */
1484 };
1485
1486 /*
1487  * Info about orphan file.
1488  */
1489 struct ext4_orphan_info {
1490         int of_blocks;                  /* Number of orphan blocks in a file */
1491         __u32 of_csum_seed;             /* Checksum seed for orphan file */
1492         struct ext4_orphan_block *of_binfo;     /* Array with info about orphan
1493                                                  * file blocks */
1494 };
1495
1496 /*
1497  * fourth extended-fs super-block data in memory
1498  */
1499 struct ext4_sb_info {
1500         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1501         unsigned long s_inodes_per_block;/* Number of inodes per block */
1502         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1503         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1504         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1505         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1506         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1507         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1508         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1509         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1510         unsigned long s_overhead;  /* # of fs overhead clusters */
1511         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1512         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1513         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1514         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1515         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1516         struct buffer_head * __rcu *s_group_desc;
1517         unsigned int s_mount_opt;
1518         unsigned int s_mount_opt2;
1519         unsigned long s_mount_flags;
1520         unsigned int s_def_mount_opt;
1521         ext4_fsblk_t s_sb_block;
1522         atomic64_t s_resv_clusters;
1523         kuid_t s_resuid;
1524         kgid_t s_resgid;
1525         unsigned short s_mount_state;
1526         unsigned short s_pad;
1527         int s_addr_per_block_bits;
1528         int s_desc_per_block_bits;
1529         int s_inode_size;
1530         int s_first_ino;
1531         unsigned int s_inode_readahead_blks;
1532         unsigned int s_inode_goal;
1533         u32 s_hash_seed[4];
1534         int s_def_hash_version;
1535         int s_hash_unsigned;    /* 3 if hash should be unsigned, 0 if not */
1536         struct percpu_counter s_freeclusters_counter;
1537         struct percpu_counter s_freeinodes_counter;
1538         struct percpu_counter s_dirs_counter;
1539         struct percpu_counter s_dirtyclusters_counter;
1540         struct percpu_counter s_sra_exceeded_retry_limit;
1541         struct blockgroup_lock *s_blockgroup_lock;
1542         struct proc_dir_entry *s_proc;
1543         struct kobject s_kobj;
1544         struct completion s_kobj_unregister;
1545         struct super_block *s_sb;
1546         struct buffer_head *s_mmp_bh;
1547
1548         /* Journaling */
1549         struct journal_s *s_journal;
1550         unsigned long s_ext4_flags;             /* Ext4 superblock flags */
1551         struct mutex s_orphan_lock;     /* Protects on disk list changes */
1552         struct list_head s_orphan;      /* List of orphaned inodes in on disk
1553                                            list */
1554         struct ext4_orphan_info s_orphan_info;
1555         unsigned long s_commit_interval;
1556         u32 s_max_batch_time;
1557         u32 s_min_batch_time;
1558         struct block_device *s_journal_bdev;
1559 #ifdef CONFIG_QUOTA
1560         /* Names of quota files with journalled quota */
1561         char __rcu *s_qf_names[EXT4_MAXQUOTAS];
1562         int s_jquota_fmt;                       /* Format of quota to use */
1563 #endif
1564         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1565         struct ext4_system_blocks __rcu *s_system_blks;
1566
1567 #ifdef EXTENTS_STATS
1568         /* ext4 extents stats */
1569         unsigned long s_ext_min;
1570         unsigned long s_ext_max;
1571         unsigned long s_depth_max;
1572         spinlock_t s_ext_stats_lock;
1573         unsigned long s_ext_blocks;
1574         unsigned long s_ext_extents;
1575 #endif
1576
1577         /* for buddy allocator */
1578         struct ext4_group_info ** __rcu *s_group_info;
1579         struct inode *s_buddy_cache;
1580         spinlock_t s_md_lock;
1581         unsigned short *s_mb_offsets;
1582         unsigned int *s_mb_maxs;
1583         unsigned int s_group_info_size;
1584         unsigned int s_mb_free_pending;
1585         struct list_head s_freed_data_list;     /* List of blocks to be freed
1586                                                    after commit completed */
1587         struct list_head s_discard_list;
1588         struct work_struct s_discard_work;
1589         atomic_t s_retry_alloc_pending;
1590         struct list_head *s_mb_avg_fragment_size;
1591         rwlock_t *s_mb_avg_fragment_size_locks;
1592         struct list_head *s_mb_largest_free_orders;
1593         rwlock_t *s_mb_largest_free_orders_locks;
1594
1595         /* tunables */
1596         unsigned long s_stripe;
1597         unsigned int s_mb_max_linear_groups;
1598         unsigned int s_mb_stream_request;
1599         unsigned int s_mb_max_to_scan;
1600         unsigned int s_mb_min_to_scan;
1601         unsigned int s_mb_stats;
1602         unsigned int s_mb_order2_reqs;
1603         unsigned int s_mb_group_prealloc;
1604         unsigned int s_mb_max_inode_prealloc;
1605         unsigned int s_max_dir_size_kb;
1606         /* where last allocation was done - for stream allocation */
1607         unsigned long s_mb_last_group;
1608         unsigned long s_mb_last_start;
1609         unsigned int s_mb_prefetch;
1610         unsigned int s_mb_prefetch_limit;
1611
1612         /* stats for buddy allocator */
1613         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1614         atomic_t s_bal_success; /* we found long enough chunks */
1615         atomic_t s_bal_allocated;       /* in blocks */
1616         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1617         atomic_t s_bal_groups_scanned;  /* number of groups scanned */
1618         atomic_t s_bal_goals;   /* goal hits */
1619         atomic_t s_bal_breaks;  /* too long searches */
1620         atomic_t s_bal_2orders; /* 2^order hits */
1621         atomic_t s_bal_cr0_bad_suggestions;
1622         atomic_t s_bal_cr1_bad_suggestions;
1623         atomic64_t s_bal_cX_groups_considered[4];
1624         atomic64_t s_bal_cX_hits[4];
1625         atomic64_t s_bal_cX_failed[4];          /* cX loop didn't find blocks */
1626         atomic_t s_mb_buddies_generated;        /* number of buddies generated */
1627         atomic64_t s_mb_generation_time;
1628         atomic_t s_mb_lost_chunks;
1629         atomic_t s_mb_preallocated;
1630         atomic_t s_mb_discarded;
1631         atomic_t s_lock_busy;
1632
1633         /* locality groups */
1634         struct ext4_locality_group __percpu *s_locality_groups;
1635
1636         /* for write statistics */
1637         unsigned long s_sectors_written_start;
1638         u64 s_kbytes_written;
1639
1640         /* the size of zero-out chunk */
1641         unsigned int s_extent_max_zeroout_kb;
1642
1643         unsigned int s_log_groups_per_flex;
1644         struct flex_groups * __rcu *s_flex_groups;
1645         ext4_group_t s_flex_groups_allocated;
1646
1647         /* workqueue for reserved extent conversions (buffered io) */
1648         struct workqueue_struct *rsv_conversion_wq;
1649
1650         /* timer for periodic error stats printing */
1651         struct timer_list s_err_report;
1652
1653         /* Lazy inode table initialization info */
1654         struct ext4_li_request *s_li_request;
1655         /* Wait multiplier for lazy initialization thread */
1656         unsigned int s_li_wait_mult;
1657
1658         /* Kernel thread for multiple mount protection */
1659         struct task_struct *s_mmp_tsk;
1660
1661         /* record the last minlen when FITRIM is called. */
1662         unsigned long s_last_trim_minblks;
1663
1664         /* Reference to checksum algorithm driver via cryptoapi */
1665         struct crypto_shash *s_chksum_driver;
1666
1667         /* Precomputed FS UUID checksum for seeding other checksums */
1668         __u32 s_csum_seed;
1669
1670         /* Reclaim extents from extent status tree */
1671         struct shrinker s_es_shrinker;
1672         struct list_head s_es_list;     /* List of inodes with reclaimable extents */
1673         long s_es_nr_inode;
1674         struct ext4_es_stats s_es_stats;
1675         struct mb_cache *s_ea_block_cache;
1676         struct mb_cache *s_ea_inode_cache;
1677         spinlock_t s_es_lock ____cacheline_aligned_in_smp;
1678
1679         /* Journal triggers for checksum computation */
1680         struct ext4_journal_trigger s_journal_triggers[EXT4_JOURNAL_TRIGGER_COUNT];
1681
1682         /* Ratelimit ext4 messages. */
1683         struct ratelimit_state s_err_ratelimit_state;
1684         struct ratelimit_state s_warning_ratelimit_state;
1685         struct ratelimit_state s_msg_ratelimit_state;
1686         atomic_t s_warning_count;
1687         atomic_t s_msg_count;
1688
1689         /* Encryption policy for '-o test_dummy_encryption' */
1690         struct fscrypt_dummy_policy s_dummy_enc_policy;
1691
1692         /*
1693          * Barrier between writepages ops and changing any inode's JOURNAL_DATA
1694          * or EXTENTS flag.
1695          */
1696         struct percpu_rw_semaphore s_writepages_rwsem;
1697         struct dax_device *s_daxdev;
1698         u64 s_dax_part_off;
1699 #ifdef CONFIG_EXT4_DEBUG
1700         unsigned long s_simulate_fail;
1701 #endif
1702         /* Record the errseq of the backing block device */
1703         errseq_t s_bdev_wb_err;
1704         spinlock_t s_bdev_wb_lock;
1705
1706         /* Information about errors that happened during this mount */
1707         spinlock_t s_error_lock;
1708         int s_add_error_count;
1709         int s_first_error_code;
1710         __u32 s_first_error_line;
1711         __u32 s_first_error_ino;
1712         __u64 s_first_error_block;
1713         const char *s_first_error_func;
1714         time64_t s_first_error_time;
1715         int s_last_error_code;
1716         __u32 s_last_error_line;
1717         __u32 s_last_error_ino;
1718         __u64 s_last_error_block;
1719         const char *s_last_error_func;
1720         time64_t s_last_error_time;
1721         /*
1722          * If we are in a context where we cannot update error information in
1723          * the on-disk superblock, we queue this work to do it.
1724          */
1725         struct work_struct s_error_work;
1726
1727         /* Ext4 fast commit sub transaction ID */
1728         atomic_t s_fc_subtid;
1729
1730         /*
1731          * After commit starts, the main queue gets locked, and the further
1732          * updates get added in the staging queue.
1733          */
1734 #define FC_Q_MAIN       0
1735 #define FC_Q_STAGING    1
1736         struct list_head s_fc_q[2];     /* Inodes staged for fast commit
1737                                          * that have data changes in them.
1738                                          */
1739         struct list_head s_fc_dentry_q[2];      /* directory entry updates */
1740         unsigned int s_fc_bytes;
1741         /*
1742          * Main fast commit lock. This lock protects accesses to the
1743          * following fields:
1744          * ei->i_fc_list, s_fc_dentry_q, s_fc_q, s_fc_bytes, s_fc_bh.
1745          */
1746         spinlock_t s_fc_lock;
1747         struct buffer_head *s_fc_bh;
1748         struct ext4_fc_stats s_fc_stats;
1749         tid_t s_fc_ineligible_tid;
1750 #ifdef CONFIG_EXT4_DEBUG
1751         int s_fc_debug_max_replay;
1752 #endif
1753         struct ext4_fc_replay_state s_fc_replay_state;
1754 };
1755
1756 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1757 {
1758         return sb->s_fs_info;
1759 }
1760 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1761 {
1762         return container_of(inode, struct ext4_inode_info, vfs_inode);
1763 }
1764
1765 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1766 {
1767         return ino == EXT4_ROOT_INO ||
1768                 (ino >= EXT4_FIRST_INO(sb) &&
1769                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1770 }
1771
1772 /*
1773  * Returns: sbi->field[index]
1774  * Used to access an array element from the following sbi fields which require
1775  * rcu protection to avoid dereferencing an invalid pointer due to reassignment
1776  * - s_group_desc
1777  * - s_group_info
1778  * - s_flex_group
1779  */
1780 #define sbi_array_rcu_deref(sbi, field, index)                             \
1781 ({                                                                         \
1782         typeof(*((sbi)->field)) _v;                                        \
1783         rcu_read_lock();                                                   \
1784         _v = ((typeof(_v)*)rcu_dereference((sbi)->field))[index];          \
1785         rcu_read_unlock();                                                 \
1786         _v;                                                                \
1787 })
1788
1789 /*
1790  * run-time mount flags
1791  */
1792 enum {
1793         EXT4_MF_MNTDIR_SAMPLED,
1794         EXT4_MF_FS_ABORTED,     /* Fatal error detected */
1795         EXT4_MF_FC_INELIGIBLE   /* Fast commit ineligible */
1796 };
1797
1798 static inline void ext4_set_mount_flag(struct super_block *sb, int bit)
1799 {
1800         set_bit(bit, &EXT4_SB(sb)->s_mount_flags);
1801 }
1802
1803 static inline void ext4_clear_mount_flag(struct super_block *sb, int bit)
1804 {
1805         clear_bit(bit, &EXT4_SB(sb)->s_mount_flags);
1806 }
1807
1808 static inline int ext4_test_mount_flag(struct super_block *sb, int bit)
1809 {
1810         return test_bit(bit, &EXT4_SB(sb)->s_mount_flags);
1811 }
1812
1813
1814 /*
1815  * Simulate_fail codes
1816  */
1817 #define EXT4_SIM_BBITMAP_EIO    1
1818 #define EXT4_SIM_BBITMAP_CRC    2
1819 #define EXT4_SIM_IBITMAP_EIO    3
1820 #define EXT4_SIM_IBITMAP_CRC    4
1821 #define EXT4_SIM_INODE_EIO      5
1822 #define EXT4_SIM_INODE_CRC      6
1823 #define EXT4_SIM_DIRBLOCK_EIO   7
1824 #define EXT4_SIM_DIRBLOCK_CRC   8
1825
1826 static inline bool ext4_simulate_fail(struct super_block *sb,
1827                                      unsigned long code)
1828 {
1829 #ifdef CONFIG_EXT4_DEBUG
1830         struct ext4_sb_info *sbi = EXT4_SB(sb);
1831
1832         if (unlikely(sbi->s_simulate_fail == code)) {
1833                 sbi->s_simulate_fail = 0;
1834                 return true;
1835         }
1836 #endif
1837         return false;
1838 }
1839
1840 static inline void ext4_simulate_fail_bh(struct super_block *sb,
1841                                          struct buffer_head *bh,
1842                                          unsigned long code)
1843 {
1844         if (!IS_ERR(bh) && ext4_simulate_fail(sb, code))
1845                 clear_buffer_uptodate(bh);
1846 }
1847
1848 /*
1849  * Error number codes for s_{first,last}_error_errno
1850  *
1851  * Linux errno numbers are architecture specific, so we need to translate
1852  * them into something which is architecture independent.   We don't define
1853  * codes for all errno's; just the ones which are most likely to be the cause
1854  * of an ext4_error() call.
1855  */
1856 #define EXT4_ERR_UNKNOWN         1
1857 #define EXT4_ERR_EIO             2
1858 #define EXT4_ERR_ENOMEM          3
1859 #define EXT4_ERR_EFSBADCRC       4
1860 #define EXT4_ERR_EFSCORRUPTED    5
1861 #define EXT4_ERR_ENOSPC          6
1862 #define EXT4_ERR_ENOKEY          7
1863 #define EXT4_ERR_EROFS           8
1864 #define EXT4_ERR_EFBIG           9
1865 #define EXT4_ERR_EEXIST         10
1866 #define EXT4_ERR_ERANGE         11
1867 #define EXT4_ERR_EOVERFLOW      12
1868 #define EXT4_ERR_EBUSY          13
1869 #define EXT4_ERR_ENOTDIR        14
1870 #define EXT4_ERR_ENOTEMPTY      15
1871 #define EXT4_ERR_ESHUTDOWN      16
1872 #define EXT4_ERR_EFAULT         17
1873
1874 /*
1875  * Inode dynamic state flags
1876  */
1877 enum {
1878         EXT4_STATE_JDATA,               /* journaled data exists */
1879         EXT4_STATE_NEW,                 /* inode is newly created */
1880         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1881         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1882         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1883         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1884         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1885         EXT4_STATE_MAY_INLINE_DATA,     /* may have in-inode data */
1886         EXT4_STATE_EXT_PRECACHED,       /* extents have been precached */
1887         EXT4_STATE_LUSTRE_EA_INODE,     /* Lustre-style ea_inode */
1888         EXT4_STATE_VERITY_IN_PROGRESS,  /* building fs-verity Merkle tree */
1889         EXT4_STATE_FC_COMMITTING,       /* Fast commit ongoing */
1890         EXT4_STATE_ORPHAN_FILE,         /* Inode orphaned in orphan file */
1891 };
1892
1893 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1894 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1895 {                                                                       \
1896         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1897 }                                                                       \
1898 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1899 {                                                                       \
1900         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1901 }                                                                       \
1902 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1903 {                                                                       \
1904         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1905 }
1906
1907 /* Add these declarations here only so that these functions can be
1908  * found by name.  Otherwise, they are very hard to locate. */
1909 static inline int ext4_test_inode_flag(struct inode *inode, int bit);
1910 static inline void ext4_set_inode_flag(struct inode *inode, int bit);
1911 static inline void ext4_clear_inode_flag(struct inode *inode, int bit);
1912 EXT4_INODE_BIT_FNS(flag, flags, 0)
1913
1914 /* Add these declarations here only so that these functions can be
1915  * found by name.  Otherwise, they are very hard to locate. */
1916 static inline int ext4_test_inode_state(struct inode *inode, int bit);
1917 static inline void ext4_set_inode_state(struct inode *inode, int bit);
1918 static inline void ext4_clear_inode_state(struct inode *inode, int bit);
1919 #if (BITS_PER_LONG < 64)
1920 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1921
1922 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1923 {
1924         (ei)->i_state_flags = 0;
1925 }
1926 #else
1927 EXT4_INODE_BIT_FNS(state, flags, 32)
1928
1929 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1930 {
1931         /* We depend on the fact that callers will set i_flags */
1932 }
1933 #endif
1934 #else
1935 /* Assume that user mode programs are passing in an ext4fs superblock, not
1936  * a kernel struct super_block.  This will allow us to call the feature-test
1937  * macros from user land. */
1938 #define EXT4_SB(sb)     (sb)
1939 #endif
1940
1941 static inline bool ext4_verity_in_progress(struct inode *inode)
1942 {
1943         return IS_ENABLED(CONFIG_FS_VERITY) &&
1944                ext4_test_inode_state(inode, EXT4_STATE_VERITY_IN_PROGRESS);
1945 }
1946
1947 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1948
1949 /*
1950  * Codes for operating systems
1951  */
1952 #define EXT4_OS_LINUX           0
1953 #define EXT4_OS_HURD            1
1954 #define EXT4_OS_MASIX           2
1955 #define EXT4_OS_FREEBSD         3
1956 #define EXT4_OS_LITES           4
1957
1958 /*
1959  * Revision levels
1960  */
1961 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1962 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1963
1964 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1965
1966 #define EXT4_GOOD_OLD_INODE_SIZE 128
1967
1968 #define EXT4_EXTRA_TIMESTAMP_MAX        (((s64)1 << 34) - 1  + S32_MIN)
1969 #define EXT4_NON_EXTRA_TIMESTAMP_MAX    S32_MAX
1970 #define EXT4_TIMESTAMP_MIN              S32_MIN
1971
1972 /*
1973  * Feature set definitions
1974  */
1975
1976 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1977 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1978 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1979 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1980 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1981 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1982 #define EXT4_FEATURE_COMPAT_SPARSE_SUPER2       0x0200
1983 /*
1984  * The reason why "FAST_COMMIT" is a compat feature is that, FS becomes
1985  * incompatible only if fast commit blocks are present in the FS. Since we
1986  * clear the journal (and thus the fast commit blocks), we don't mark FS as
1987  * incompatible. We also have a JBD2 incompat feature, which gets set when
1988  * there are fast commit blocks present in the journal.
1989  */
1990 #define EXT4_FEATURE_COMPAT_FAST_COMMIT         0x0400
1991 #define EXT4_FEATURE_COMPAT_STABLE_INODES       0x0800
1992 #define EXT4_FEATURE_COMPAT_ORPHAN_FILE         0x1000  /* Orphan file exists */
1993
1994 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1995 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1996 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1997 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1998 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1999 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
2000 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
2001 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
2002 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
2003 /*
2004  * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM).  When
2005  * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
2006  * all other data structures' checksums.  However, the METADATA_CSUM and
2007  * GDT_CSUM bits are mutually exclusive.
2008  */
2009 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM    0x0400
2010 #define EXT4_FEATURE_RO_COMPAT_READONLY         0x1000
2011 #define EXT4_FEATURE_RO_COMPAT_PROJECT          0x2000
2012 #define EXT4_FEATURE_RO_COMPAT_VERITY           0x8000
2013 #define EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT   0x10000 /* Orphan file may be
2014                                                            non-empty */
2015
2016 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
2017 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
2018 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
2019 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
2020 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
2021 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
2022 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
2023 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
2024 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
2025 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
2026 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
2027 #define EXT4_FEATURE_INCOMPAT_CSUM_SEED         0x2000
2028 #define EXT4_FEATURE_INCOMPAT_LARGEDIR          0x4000 /* >2GB or 3-lvl htree */
2029 #define EXT4_FEATURE_INCOMPAT_INLINE_DATA       0x8000 /* data in inode */
2030 #define EXT4_FEATURE_INCOMPAT_ENCRYPT           0x10000
2031 #define EXT4_FEATURE_INCOMPAT_CASEFOLD          0x20000
2032
2033 extern void ext4_update_dynamic_rev(struct super_block *sb);
2034
2035 #define EXT4_FEATURE_COMPAT_FUNCS(name, flagname) \
2036 static inline bool ext4_has_feature_##name(struct super_block *sb) \
2037 { \
2038         return ((EXT4_SB(sb)->s_es->s_feature_compat & \
2039                 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname)) != 0); \
2040 } \
2041 static inline void ext4_set_feature_##name(struct super_block *sb) \
2042 { \
2043         ext4_update_dynamic_rev(sb); \
2044         EXT4_SB(sb)->s_es->s_feature_compat |= \
2045                 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
2046 } \
2047 static inline void ext4_clear_feature_##name(struct super_block *sb) \
2048 { \
2049         EXT4_SB(sb)->s_es->s_feature_compat &= \
2050                 ~cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
2051 }
2052
2053 #define EXT4_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
2054 static inline bool ext4_has_feature_##name(struct super_block *sb) \
2055 { \
2056         return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
2057                 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname)) != 0); \
2058 } \
2059 static inline void ext4_set_feature_##name(struct super_block *sb) \
2060 { \
2061         ext4_update_dynamic_rev(sb); \
2062         EXT4_SB(sb)->s_es->s_feature_ro_compat |= \
2063                 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
2064 } \
2065 static inline void ext4_clear_feature_##name(struct super_block *sb) \
2066 { \
2067         EXT4_SB(sb)->s_es->s_feature_ro_compat &= \
2068                 ~cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
2069 }
2070
2071 #define EXT4_FEATURE_INCOMPAT_FUNCS(name, flagname) \
2072 static inline bool ext4_has_feature_##name(struct super_block *sb) \
2073 { \
2074         return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
2075                 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname)) != 0); \
2076 } \
2077 static inline void ext4_set_feature_##name(struct super_block *sb) \
2078 { \
2079         ext4_update_dynamic_rev(sb); \
2080         EXT4_SB(sb)->s_es->s_feature_incompat |= \
2081                 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
2082 } \
2083 static inline void ext4_clear_feature_##name(struct super_block *sb) \
2084 { \
2085         EXT4_SB(sb)->s_es->s_feature_incompat &= \
2086                 ~cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
2087 }
2088
2089 EXT4_FEATURE_COMPAT_FUNCS(dir_prealloc,         DIR_PREALLOC)
2090 EXT4_FEATURE_COMPAT_FUNCS(imagic_inodes,        IMAGIC_INODES)
2091 EXT4_FEATURE_COMPAT_FUNCS(journal,              HAS_JOURNAL)
2092 EXT4_FEATURE_COMPAT_FUNCS(xattr,                EXT_ATTR)
2093 EXT4_FEATURE_COMPAT_FUNCS(resize_inode,         RESIZE_INODE)
2094 EXT4_FEATURE_COMPAT_FUNCS(dir_index,            DIR_INDEX)
2095 EXT4_FEATURE_COMPAT_FUNCS(sparse_super2,        SPARSE_SUPER2)
2096 EXT4_FEATURE_COMPAT_FUNCS(fast_commit,          FAST_COMMIT)
2097 EXT4_FEATURE_COMPAT_FUNCS(stable_inodes,        STABLE_INODES)
2098 EXT4_FEATURE_COMPAT_FUNCS(orphan_file,          ORPHAN_FILE)
2099
2100 EXT4_FEATURE_RO_COMPAT_FUNCS(sparse_super,      SPARSE_SUPER)
2101 EXT4_FEATURE_RO_COMPAT_FUNCS(large_file,        LARGE_FILE)
2102 EXT4_FEATURE_RO_COMPAT_FUNCS(btree_dir,         BTREE_DIR)
2103 EXT4_FEATURE_RO_COMPAT_FUNCS(huge_file,         HUGE_FILE)
2104 EXT4_FEATURE_RO_COMPAT_FUNCS(gdt_csum,          GDT_CSUM)
2105 EXT4_FEATURE_RO_COMPAT_FUNCS(dir_nlink,         DIR_NLINK)
2106 EXT4_FEATURE_RO_COMPAT_FUNCS(extra_isize,       EXTRA_ISIZE)
2107 EXT4_FEATURE_RO_COMPAT_FUNCS(quota,             QUOTA)
2108 EXT4_FEATURE_RO_COMPAT_FUNCS(bigalloc,          BIGALLOC)
2109 EXT4_FEATURE_RO_COMPAT_FUNCS(metadata_csum,     METADATA_CSUM)
2110 EXT4_FEATURE_RO_COMPAT_FUNCS(readonly,          READONLY)
2111 EXT4_FEATURE_RO_COMPAT_FUNCS(project,           PROJECT)
2112 EXT4_FEATURE_RO_COMPAT_FUNCS(verity,            VERITY)
2113 EXT4_FEATURE_RO_COMPAT_FUNCS(orphan_present,    ORPHAN_PRESENT)
2114
2115 EXT4_FEATURE_INCOMPAT_FUNCS(compression,        COMPRESSION)
2116 EXT4_FEATURE_INCOMPAT_FUNCS(filetype,           FILETYPE)
2117 EXT4_FEATURE_INCOMPAT_FUNCS(journal_needs_recovery,     RECOVER)
2118 EXT4_FEATURE_INCOMPAT_FUNCS(journal_dev,        JOURNAL_DEV)
2119 EXT4_FEATURE_INCOMPAT_FUNCS(meta_bg,            META_BG)
2120 EXT4_FEATURE_INCOMPAT_FUNCS(extents,            EXTENTS)
2121 EXT4_FEATURE_INCOMPAT_FUNCS(64bit,              64BIT)
2122 EXT4_FEATURE_INCOMPAT_FUNCS(mmp,                MMP)
2123 EXT4_FEATURE_INCOMPAT_FUNCS(flex_bg,            FLEX_BG)
2124 EXT4_FEATURE_INCOMPAT_FUNCS(ea_inode,           EA_INODE)
2125 EXT4_FEATURE_INCOMPAT_FUNCS(dirdata,            DIRDATA)
2126 EXT4_FEATURE_INCOMPAT_FUNCS(csum_seed,          CSUM_SEED)
2127 EXT4_FEATURE_INCOMPAT_FUNCS(largedir,           LARGEDIR)
2128 EXT4_FEATURE_INCOMPAT_FUNCS(inline_data,        INLINE_DATA)
2129 EXT4_FEATURE_INCOMPAT_FUNCS(encrypt,            ENCRYPT)
2130 EXT4_FEATURE_INCOMPAT_FUNCS(casefold,           CASEFOLD)
2131
2132 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
2133 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
2134                                          EXT4_FEATURE_INCOMPAT_META_BG)
2135 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
2136                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
2137                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
2138
2139 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
2140 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
2141                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
2142                                          EXT4_FEATURE_INCOMPAT_META_BG)
2143 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
2144                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
2145                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
2146
2147 #define EXT4_FEATURE_COMPAT_SUPP        (EXT4_FEATURE_COMPAT_EXT_ATTR| \
2148                                          EXT4_FEATURE_COMPAT_ORPHAN_FILE)
2149 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
2150                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
2151                                          EXT4_FEATURE_INCOMPAT_META_BG| \
2152                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
2153                                          EXT4_FEATURE_INCOMPAT_64BIT| \
2154                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
2155                                          EXT4_FEATURE_INCOMPAT_EA_INODE| \
2156                                          EXT4_FEATURE_INCOMPAT_MMP | \
2157                                          EXT4_FEATURE_INCOMPAT_INLINE_DATA | \
2158                                          EXT4_FEATURE_INCOMPAT_ENCRYPT | \
2159                                          EXT4_FEATURE_INCOMPAT_CASEFOLD | \
2160                                          EXT4_FEATURE_INCOMPAT_CSUM_SEED | \
2161                                          EXT4_FEATURE_INCOMPAT_LARGEDIR)
2162 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
2163                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
2164                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
2165                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
2166                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
2167                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
2168                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
2169                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
2170                                          EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
2171                                          EXT4_FEATURE_RO_COMPAT_QUOTA |\
2172                                          EXT4_FEATURE_RO_COMPAT_PROJECT |\
2173                                          EXT4_FEATURE_RO_COMPAT_VERITY |\
2174                                          EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT)
2175
2176 #define EXTN_FEATURE_FUNCS(ver) \
2177 static inline bool ext4_has_unknown_ext##ver##_compat_features(struct super_block *sb) \
2178 { \
2179         return ((EXT4_SB(sb)->s_es->s_feature_compat & \
2180                 cpu_to_le32(~EXT##ver##_FEATURE_COMPAT_SUPP)) != 0); \
2181 } \
2182 static inline bool ext4_has_unknown_ext##ver##_ro_compat_features(struct super_block *sb) \
2183 { \
2184         return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
2185                 cpu_to_le32(~EXT##ver##_FEATURE_RO_COMPAT_SUPP)) != 0); \
2186 } \
2187 static inline bool ext4_has_unknown_ext##ver##_incompat_features(struct super_block *sb) \
2188 { \
2189         return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
2190                 cpu_to_le32(~EXT##ver##_FEATURE_INCOMPAT_SUPP)) != 0); \
2191 }
2192
2193 EXTN_FEATURE_FUNCS(2)
2194 EXTN_FEATURE_FUNCS(3)
2195 EXTN_FEATURE_FUNCS(4)
2196
2197 static inline bool ext4_has_compat_features(struct super_block *sb)
2198 {
2199         return (EXT4_SB(sb)->s_es->s_feature_compat != 0);
2200 }
2201 static inline bool ext4_has_ro_compat_features(struct super_block *sb)
2202 {
2203         return (EXT4_SB(sb)->s_es->s_feature_ro_compat != 0);
2204 }
2205 static inline bool ext4_has_incompat_features(struct super_block *sb)
2206 {
2207         return (EXT4_SB(sb)->s_es->s_feature_incompat != 0);
2208 }
2209
2210 extern int ext4_feature_set_ok(struct super_block *sb, int readonly);
2211
2212 /*
2213  * Superblock flags
2214  */
2215 #define EXT4_FLAGS_RESIZING     0
2216 #define EXT4_FLAGS_SHUTDOWN     1
2217 #define EXT4_FLAGS_BDEV_IS_DAX  2
2218
2219 static inline int ext4_forced_shutdown(struct ext4_sb_info *sbi)
2220 {
2221         return test_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
2222 }
2223
2224 /*
2225  * Default values for user and/or group using reserved blocks
2226  */
2227 #define EXT4_DEF_RESUID         0
2228 #define EXT4_DEF_RESGID         0
2229
2230 /*
2231  * Default project ID
2232  */
2233 #define EXT4_DEF_PROJID         0
2234
2235 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
2236
2237 /*
2238  * Default mount options
2239  */
2240 #define EXT4_DEFM_DEBUG         0x0001
2241 #define EXT4_DEFM_BSDGROUPS     0x0002
2242 #define EXT4_DEFM_XATTR_USER    0x0004
2243 #define EXT4_DEFM_ACL           0x0008
2244 #define EXT4_DEFM_UID16         0x0010
2245 #define EXT4_DEFM_JMODE         0x0060
2246 #define EXT4_DEFM_JMODE_DATA    0x0020
2247 #define EXT4_DEFM_JMODE_ORDERED 0x0040
2248 #define EXT4_DEFM_JMODE_WBACK   0x0060
2249 #define EXT4_DEFM_NOBARRIER     0x0100
2250 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
2251 #define EXT4_DEFM_DISCARD       0x0400
2252 #define EXT4_DEFM_NODELALLOC    0x0800
2253
2254 /*
2255  * Default journal batch times
2256  */
2257 #define EXT4_DEF_MIN_BATCH_TIME 0
2258 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
2259
2260 /*
2261  * Minimum number of groups in a flexgroup before we separate out
2262  * directories into the first block group of a flexgroup
2263  */
2264 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
2265
2266 /*
2267  * Structure of a directory entry
2268  */
2269 #define EXT4_NAME_LEN 255
2270 /*
2271  * Base length of the ext4 directory entry excluding the name length
2272  */
2273 #define EXT4_BASE_DIR_LEN (sizeof(struct ext4_dir_entry_2) - EXT4_NAME_LEN)
2274
2275 struct ext4_dir_entry {
2276         __le32  inode;                  /* Inode number */
2277         __le16  rec_len;                /* Directory entry length */
2278         __le16  name_len;               /* Name length */
2279         char    name[EXT4_NAME_LEN];    /* File name */
2280 };
2281
2282
2283 /*
2284  * Encrypted Casefolded entries require saving the hash on disk. This structure
2285  * followed ext4_dir_entry_2's name[name_len] at the next 4 byte aligned
2286  * boundary.
2287  */
2288 struct ext4_dir_entry_hash {
2289         __le32 hash;
2290         __le32 minor_hash;
2291 };
2292
2293 /*
2294  * The new version of the directory entry.  Since EXT4 structures are
2295  * stored in intel byte order, and the name_len field could never be
2296  * bigger than 255 chars, it's safe to reclaim the extra byte for the
2297  * file_type field.
2298  */
2299 struct ext4_dir_entry_2 {
2300         __le32  inode;                  /* Inode number */
2301         __le16  rec_len;                /* Directory entry length */
2302         __u8    name_len;               /* Name length */
2303         __u8    file_type;              /* See file type macros EXT4_FT_* below */
2304         char    name[EXT4_NAME_LEN];    /* File name */
2305 };
2306
2307 /*
2308  * Access the hashes at the end of ext4_dir_entry_2
2309  */
2310 #define EXT4_DIRENT_HASHES(entry) \
2311         ((struct ext4_dir_entry_hash *) \
2312                 (((void *)(entry)) + \
2313                 ((8 + (entry)->name_len + EXT4_DIR_ROUND) & ~EXT4_DIR_ROUND)))
2314 #define EXT4_DIRENT_HASH(entry) le32_to_cpu(EXT4_DIRENT_HASHES(de)->hash)
2315 #define EXT4_DIRENT_MINOR_HASH(entry) \
2316                 le32_to_cpu(EXT4_DIRENT_HASHES(de)->minor_hash)
2317
2318 static inline bool ext4_hash_in_dirent(const struct inode *inode)
2319 {
2320         return IS_CASEFOLDED(inode) && IS_ENCRYPTED(inode);
2321 }
2322
2323 /*
2324  * This is a bogus directory entry at the end of each leaf block that
2325  * records checksums.
2326  */
2327 struct ext4_dir_entry_tail {
2328         __le32  det_reserved_zero1;     /* Pretend to be unused */
2329         __le16  det_rec_len;            /* 12 */
2330         __u8    det_reserved_zero2;     /* Zero name length */
2331         __u8    det_reserved_ft;        /* 0xDE, fake file type */
2332         __le32  det_checksum;           /* crc32c(uuid+inum+dirblock) */
2333 };
2334
2335 #define EXT4_DIRENT_TAIL(block, blocksize) \
2336         ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
2337                                         ((blocksize) - \
2338                                          sizeof(struct ext4_dir_entry_tail))))
2339
2340 /*
2341  * Ext4 directory file types.  Only the low 3 bits are used.  The
2342  * other bits are reserved for now.
2343  */
2344 #define EXT4_FT_UNKNOWN         0
2345 #define EXT4_FT_REG_FILE        1
2346 #define EXT4_FT_DIR             2
2347 #define EXT4_FT_CHRDEV          3
2348 #define EXT4_FT_BLKDEV          4
2349 #define EXT4_FT_FIFO            5
2350 #define EXT4_FT_SOCK            6
2351 #define EXT4_FT_SYMLINK         7
2352
2353 #define EXT4_FT_MAX             8
2354
2355 #define EXT4_FT_DIR_CSUM        0xDE
2356
2357 /*
2358  * EXT4_DIR_PAD defines the directory entries boundaries
2359  *
2360  * NOTE: It must be a multiple of 4
2361  */
2362 #define EXT4_DIR_PAD                    4
2363 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
2364 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
2365
2366 /*
2367  * The rec_len is dependent on the type of directory. Directories that are
2368  * casefolded and encrypted need to store the hash as well, so we add room for
2369  * ext4_extended_dir_entry_2. For all entries related to '.' or '..' you should
2370  * pass NULL for dir, as those entries do not use the extra fields.
2371  */
2372 static inline unsigned int ext4_dir_rec_len(__u8 name_len,
2373                                                 const struct inode *dir)
2374 {
2375         int rec_len = (name_len + 8 + EXT4_DIR_ROUND);
2376
2377         if (dir && ext4_hash_in_dirent(dir))
2378                 rec_len += sizeof(struct ext4_dir_entry_hash);
2379         return (rec_len & ~EXT4_DIR_ROUND);
2380 }
2381
2382 /*
2383  * If we ever get support for fs block sizes > page_size, we'll need
2384  * to remove the #if statements in the next two functions...
2385  */
2386 static inline unsigned int
2387 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
2388 {
2389         unsigned len = le16_to_cpu(dlen);
2390
2391 #if (PAGE_SIZE >= 65536)
2392         if (len == EXT4_MAX_REC_LEN || len == 0)
2393                 return blocksize;
2394         return (len & 65532) | ((len & 3) << 16);
2395 #else
2396         return len;
2397 #endif
2398 }
2399
2400 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
2401 {
2402         BUG_ON((len > blocksize) || (blocksize > (1 << 18)) || (len & 3));
2403 #if (PAGE_SIZE >= 65536)
2404         if (len < 65536)
2405                 return cpu_to_le16(len);
2406         if (len == blocksize) {
2407                 if (blocksize == 65536)
2408                         return cpu_to_le16(EXT4_MAX_REC_LEN);
2409                 else
2410                         return cpu_to_le16(0);
2411         }
2412         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
2413 #else
2414         return cpu_to_le16(len);
2415 #endif
2416 }
2417
2418 /*
2419  * Hash Tree Directory indexing
2420  * (c) Daniel Phillips, 2001
2421  */
2422
2423 #define is_dx(dir) (ext4_has_feature_dir_index((dir)->i_sb) && \
2424                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
2425 #define EXT4_DIR_LINK_MAX(dir) unlikely((dir)->i_nlink >= EXT4_LINK_MAX && \
2426                     !(ext4_has_feature_dir_nlink((dir)->i_sb) && is_dx(dir)))
2427 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
2428
2429 /* Legal values for the dx_root hash_version field: */
2430
2431 #define DX_HASH_LEGACY                  0
2432 #define DX_HASH_HALF_MD4                1
2433 #define DX_HASH_TEA                     2
2434 #define DX_HASH_LEGACY_UNSIGNED         3
2435 #define DX_HASH_HALF_MD4_UNSIGNED       4
2436 #define DX_HASH_TEA_UNSIGNED            5
2437 #define DX_HASH_SIPHASH                 6
2438
2439 static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
2440                               const void *address, unsigned int length)
2441 {
2442         struct {
2443                 struct shash_desc shash;
2444                 char ctx[4];
2445         } desc;
2446
2447         BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
2448
2449         desc.shash.tfm = sbi->s_chksum_driver;
2450         *(u32 *)desc.ctx = crc;
2451
2452         BUG_ON(crypto_shash_update(&desc.shash, address, length));
2453
2454         return *(u32 *)desc.ctx;
2455 }
2456
2457 #ifdef __KERNEL__
2458
2459 /* hash info structure used by the directory hash */
2460 struct dx_hash_info
2461 {
2462         u32             hash;
2463         u32             minor_hash;
2464         int             hash_version;
2465         u32             *seed;
2466 };
2467
2468
2469 /* 32 and 64 bit signed EOF for dx directories */
2470 #define EXT4_HTREE_EOF_32BIT   ((1UL  << (32 - 1)) - 1)
2471 #define EXT4_HTREE_EOF_64BIT   ((1ULL << (64 - 1)) - 1)
2472
2473
2474 /*
2475  * Control parameters used by ext4_htree_next_block
2476  */
2477 #define HASH_NB_ALWAYS          1
2478
2479 struct ext4_filename {
2480         const struct qstr *usr_fname;
2481         struct fscrypt_str disk_name;
2482         struct dx_hash_info hinfo;
2483 #ifdef CONFIG_FS_ENCRYPTION
2484         struct fscrypt_str crypto_buf;
2485 #endif
2486 #if IS_ENABLED(CONFIG_UNICODE)
2487         struct fscrypt_str cf_name;
2488 #endif
2489 };
2490
2491 #define fname_name(p) ((p)->disk_name.name)
2492 #define fname_usr_name(p) ((p)->usr_fname->name)
2493 #define fname_len(p)  ((p)->disk_name.len)
2494
2495 /*
2496  * Describe an inode's exact location on disk and in memory
2497  */
2498 struct ext4_iloc
2499 {
2500         struct buffer_head *bh;
2501         unsigned long offset;
2502         ext4_group_t block_group;
2503 };
2504
2505 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
2506 {
2507         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
2508 }
2509
2510 static inline bool ext4_is_quota_file(struct inode *inode)
2511 {
2512         return IS_NOQUOTA(inode) &&
2513                !(EXT4_I(inode)->i_flags & EXT4_EA_INODE_FL);
2514 }
2515
2516 /*
2517  * This structure is stuffed into the struct file's private_data field
2518  * for directories.  It is where we put information so that we can do
2519  * readdir operations in hash tree order.
2520  */
2521 struct dir_private_info {
2522         struct rb_root  root;
2523         struct rb_node  *curr_node;
2524         struct fname    *extra_fname;
2525         loff_t          last_pos;
2526         __u32           curr_hash;
2527         __u32           curr_minor_hash;
2528         __u32           next_hash;
2529 };
2530
2531 /* calculate the first block number of the group */
2532 static inline ext4_fsblk_t
2533 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
2534 {
2535         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
2536                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
2537 }
2538
2539 /*
2540  * Special error return code only used by dx_probe() and its callers.
2541  */
2542 #define ERR_BAD_DX_DIR  (-(MAX_ERRNO - 1))
2543
2544 /* htree levels for ext4 */
2545 #define EXT4_HTREE_LEVEL_COMPAT 2
2546 #define EXT4_HTREE_LEVEL        3
2547
2548 static inline int ext4_dir_htree_level(struct super_block *sb)
2549 {
2550         return ext4_has_feature_largedir(sb) ?
2551                 EXT4_HTREE_LEVEL : EXT4_HTREE_LEVEL_COMPAT;
2552 }
2553
2554 /*
2555  * Timeout and state flag for lazy initialization inode thread.
2556  */
2557 #define EXT4_DEF_LI_WAIT_MULT                   10
2558 #define EXT4_DEF_LI_MAX_START_DELAY             5
2559 #define EXT4_LAZYINIT_QUIT                      0x0001
2560 #define EXT4_LAZYINIT_RUNNING                   0x0002
2561
2562 /*
2563  * Lazy inode table initialization info
2564  */
2565 struct ext4_lazy_init {
2566         unsigned long           li_state;
2567         struct list_head        li_request_list;
2568         struct mutex            li_list_mtx;
2569 };
2570
2571 enum ext4_li_mode {
2572         EXT4_LI_MODE_PREFETCH_BBITMAP,
2573         EXT4_LI_MODE_ITABLE,
2574 };
2575
2576 struct ext4_li_request {
2577         struct super_block      *lr_super;
2578         enum ext4_li_mode       lr_mode;
2579         ext4_group_t            lr_first_not_zeroed;
2580         ext4_group_t            lr_next_group;
2581         struct list_head        lr_request;
2582         unsigned long           lr_next_sched;
2583         unsigned long           lr_timeout;
2584 };
2585
2586 struct ext4_features {
2587         struct kobject f_kobj;
2588         struct completion f_kobj_unregister;
2589 };
2590
2591 /*
2592  * This structure will be used for multiple mount protection. It will be
2593  * written into the block number saved in the s_mmp_block field in the
2594  * superblock. Programs that check MMP should assume that if
2595  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
2596  * to use the filesystem, regardless of how old the timestamp is.
2597  */
2598 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
2599 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
2600 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
2601 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
2602
2603 struct mmp_struct {
2604         __le32  mmp_magic;              /* Magic number for MMP */
2605         __le32  mmp_seq;                /* Sequence no. updated periodically */
2606
2607         /*
2608          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
2609          * purposes and do not affect the correctness of the algorithm
2610          */
2611         __le64  mmp_time;               /* Time last updated */
2612         char    mmp_nodename[64];       /* Node which last updated MMP block */
2613         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
2614
2615         /*
2616          * mmp_check_interval is used to verify if the MMP block has been
2617          * updated on the block device. The value is updated based on the
2618          * maximum time to write the MMP block during an update cycle.
2619          */
2620         __le16  mmp_check_interval;
2621
2622         __le16  mmp_pad1;
2623         __le32  mmp_pad2[226];
2624         __le32  mmp_checksum;           /* crc32c(uuid+mmp_block) */
2625 };
2626
2627 /* arguments passed to the mmp thread */
2628 struct mmpd_data {
2629         struct buffer_head *bh; /* bh from initial read_mmp_block() */
2630         struct super_block *sb;  /* super block of the fs */
2631 };
2632
2633 /*
2634  * Check interval multiplier
2635  * The MMP block is written every update interval and initially checked every
2636  * update interval x the multiplier (the value is then adapted based on the
2637  * write latency). The reason is that writes can be delayed under load and we
2638  * don't want readers to incorrectly assume that the filesystem is no longer
2639  * in use.
2640  */
2641 #define EXT4_MMP_CHECK_MULT             2UL
2642
2643 /*
2644  * Minimum interval for MMP checking in seconds.
2645  */
2646 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
2647
2648 /*
2649  * Maximum interval for MMP checking in seconds.
2650  */
2651 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
2652
2653 /*
2654  * Function prototypes
2655  */
2656
2657 /*
2658  * Ok, these declarations are also in <linux/kernel.h> but none of the
2659  * ext4 source programs needs to include it so they are duplicated here.
2660  */
2661 # define NORET_TYPE     /**/
2662 # define ATTRIB_NORET   __attribute__((noreturn))
2663 # define NORET_AND      noreturn,
2664
2665 /* bitmap.c */
2666 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
2667 void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
2668                                 struct ext4_group_desc *gdp,
2669                                 struct buffer_head *bh, int sz);
2670 int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2671                                   struct ext4_group_desc *gdp,
2672                                   struct buffer_head *bh, int sz);
2673 void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
2674                                 struct ext4_group_desc *gdp,
2675                                 struct buffer_head *bh);
2676 int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2677                                   struct ext4_group_desc *gdp,
2678                                   struct buffer_head *bh);
2679
2680 /* balloc.c */
2681 extern void ext4_get_group_no_and_offset(struct super_block *sb,
2682                                          ext4_fsblk_t blocknr,
2683                                          ext4_group_t *blockgrpp,
2684                                          ext4_grpblk_t *offsetp);
2685 extern ext4_group_t ext4_get_group_number(struct super_block *sb,
2686                                           ext4_fsblk_t block);
2687
2688 extern unsigned int ext4_block_group(struct super_block *sb,
2689                         ext4_fsblk_t blocknr);
2690 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
2691                         ext4_fsblk_t blocknr);
2692 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
2693 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
2694                         ext4_group_t group);
2695 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
2696                                          ext4_fsblk_t goal,
2697                                          unsigned int flags,
2698                                          unsigned long *count,
2699                                          int *errp);
2700 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
2701                                     s64 nclusters, unsigned int flags);
2702 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
2703 extern void ext4_check_blocks_bitmap(struct super_block *);
2704 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
2705                                                     ext4_group_t block_group,
2706                                                     struct buffer_head ** bh);
2707 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
2708
2709 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
2710                                                 ext4_group_t block_group,
2711                                                 bool ignore_locked);
2712 extern int ext4_wait_block_bitmap(struct super_block *sb,
2713                                   ext4_group_t block_group,
2714                                   struct buffer_head *bh);
2715 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
2716                                                   ext4_group_t block_group);
2717 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
2718                                               ext4_group_t block_group,
2719                                               struct ext4_group_desc *gdp);
2720 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
2721
2722 #if IS_ENABLED(CONFIG_UNICODE)
2723 extern int ext4_fname_setup_ci_filename(struct inode *dir,
2724                                          const struct qstr *iname,
2725                                          struct ext4_filename *fname);
2726 #endif
2727
2728 /* ext4 encryption related stuff goes here crypto.c */
2729 #ifdef CONFIG_FS_ENCRYPTION
2730 extern const struct fscrypt_operations ext4_cryptops;
2731
2732 int ext4_fname_setup_filename(struct inode *dir, const struct qstr *iname,
2733                               int lookup, struct ext4_filename *fname);
2734
2735 int ext4_fname_prepare_lookup(struct inode *dir, struct dentry *dentry,
2736                               struct ext4_filename *fname);
2737
2738 void ext4_fname_free_filename(struct ext4_filename *fname);
2739
2740 int ext4_ioctl_get_encryption_pwsalt(struct file *filp, void __user *arg);
2741
2742 #else /* !CONFIG_FS_ENCRYPTION */
2743 static inline int ext4_fname_setup_filename(struct inode *dir,
2744                                             const struct qstr *iname,
2745                                             int lookup,
2746                                             struct ext4_filename *fname)
2747 {
2748         int err = 0;
2749         fname->usr_fname = iname;
2750         fname->disk_name.name = (unsigned char *) iname->name;
2751         fname->disk_name.len = iname->len;
2752
2753 #if IS_ENABLED(CONFIG_UNICODE)
2754         err = ext4_fname_setup_ci_filename(dir, iname, fname);
2755 #endif
2756
2757         return err;
2758 }
2759
2760 static inline int ext4_fname_prepare_lookup(struct inode *dir,
2761                                             struct dentry *dentry,
2762                                             struct ext4_filename *fname)
2763 {
2764         return ext4_fname_setup_filename(dir, &dentry->d_name, 1, fname);
2765 }
2766
2767 static inline void ext4_fname_free_filename(struct ext4_filename *fname)
2768 {
2769 #if IS_ENABLED(CONFIG_UNICODE)
2770         kfree(fname->cf_name.name);
2771         fname->cf_name.name = NULL;
2772 #endif
2773 }
2774
2775 static inline int ext4_ioctl_get_encryption_pwsalt(struct file *filp,
2776                                                    void __user *arg)
2777 {
2778         return -EOPNOTSUPP;
2779 }
2780 #endif /* !CONFIG_FS_ENCRYPTION */
2781
2782 /* dir.c */
2783 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
2784                                   struct file *,
2785                                   struct ext4_dir_entry_2 *,
2786                                   struct buffer_head *, char *, int,
2787                                   unsigned int);
2788 #define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset) \
2789         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
2790                                 (de), (bh), (buf), (size), (offset)))
2791 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
2792                                 __u32 minor_hash,
2793                                 struct ext4_dir_entry_2 *dirent,
2794                                 struct fscrypt_str *ent_name);
2795 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
2796 extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
2797                              struct buffer_head *bh,
2798                              void *buf, int buf_size,
2799                              struct ext4_filename *fname,
2800                              struct ext4_dir_entry_2 **dest_de);
2801 void ext4_insert_dentry(struct inode *dir, struct inode *inode,
2802                         struct ext4_dir_entry_2 *de,
2803                         int buf_size,
2804                         struct ext4_filename *fname);
2805 static inline void ext4_update_dx_flag(struct inode *inode)
2806 {
2807         if (!ext4_has_feature_dir_index(inode->i_sb) &&
2808             ext4_test_inode_flag(inode, EXT4_INODE_INDEX)) {
2809                 /* ext4_iget() should have caught this... */
2810                 WARN_ON_ONCE(ext4_has_feature_metadata_csum(inode->i_sb));
2811                 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
2812         }
2813 }
2814 static const unsigned char ext4_filetype_table[] = {
2815         DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
2816 };
2817
2818 static inline  unsigned char get_dtype(struct super_block *sb, int filetype)
2819 {
2820         if (!ext4_has_feature_filetype(sb) || filetype >= EXT4_FT_MAX)
2821                 return DT_UNKNOWN;
2822
2823         return ext4_filetype_table[filetype];
2824 }
2825 extern int ext4_check_all_de(struct inode *dir, struct buffer_head *bh,
2826                              void *buf, int buf_size);
2827
2828 /* fsync.c */
2829 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2830
2831 /* hash.c */
2832 extern int ext4fs_dirhash(const struct inode *dir, const char *name, int len,
2833                           struct dx_hash_info *hinfo);
2834
2835 /* ialloc.c */
2836 extern int ext4_mark_inode_used(struct super_block *sb, int ino);
2837 extern struct inode *__ext4_new_inode(struct user_namespace *, handle_t *,
2838                                       struct inode *, umode_t,
2839                                       const struct qstr *qstr, __u32 goal,
2840                                       uid_t *owner, __u32 i_flags,
2841                                       int handle_type, unsigned int line_no,
2842                                       int nblocks);
2843
2844 #define ext4_new_inode(handle, dir, mode, qstr, goal, owner, i_flags)          \
2845         __ext4_new_inode(&init_user_ns, (handle), (dir), (mode), (qstr),       \
2846                          (goal), (owner), i_flags, 0, 0, 0)
2847 #define ext4_new_inode_start_handle(mnt_userns, dir, mode, qstr, goal, owner, \
2848                                     type, nblocks)                  \
2849         __ext4_new_inode((mnt_userns), NULL, (dir), (mode), (qstr), (goal), (owner), \
2850                          0, (type), __LINE__, (nblocks))
2851
2852
2853 extern void ext4_free_inode(handle_t *, struct inode *);
2854 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2855 extern unsigned long ext4_count_free_inodes(struct super_block *);
2856 extern unsigned long ext4_count_dirs(struct super_block *);
2857 extern void ext4_check_inodes_bitmap(struct super_block *);
2858 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2859 extern int ext4_init_inode_table(struct super_block *sb,
2860                                  ext4_group_t group, int barrier);
2861 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2862
2863 /* fast_commit.c */
2864 int ext4_fc_info_show(struct seq_file *seq, void *v);
2865 void ext4_fc_init(struct super_block *sb, journal_t *journal);
2866 void ext4_fc_init_inode(struct inode *inode);
2867 void ext4_fc_track_range(handle_t *handle, struct inode *inode, ext4_lblk_t start,
2868                          ext4_lblk_t end);
2869 void __ext4_fc_track_unlink(handle_t *handle, struct inode *inode,
2870         struct dentry *dentry);
2871 void __ext4_fc_track_link(handle_t *handle, struct inode *inode,
2872         struct dentry *dentry);
2873 void ext4_fc_track_unlink(handle_t *handle, struct dentry *dentry);
2874 void ext4_fc_track_link(handle_t *handle, struct dentry *dentry);
2875 void __ext4_fc_track_create(handle_t *handle, struct inode *inode,
2876                             struct dentry *dentry);
2877 void ext4_fc_track_create(handle_t *handle, struct dentry *dentry);
2878 void ext4_fc_track_inode(handle_t *handle, struct inode *inode);
2879 void ext4_fc_mark_ineligible(struct super_block *sb, int reason, handle_t *handle);
2880 void ext4_fc_start_update(struct inode *inode);
2881 void ext4_fc_stop_update(struct inode *inode);
2882 void ext4_fc_del(struct inode *inode);
2883 bool ext4_fc_replay_check_excluded(struct super_block *sb, ext4_fsblk_t block);
2884 void ext4_fc_replay_cleanup(struct super_block *sb);
2885 int ext4_fc_commit(journal_t *journal, tid_t commit_tid);
2886 int __init ext4_fc_init_dentry_cache(void);
2887 void ext4_fc_destroy_dentry_cache(void);
2888 int ext4_fc_record_regions(struct super_block *sb, int ino,
2889                            ext4_lblk_t lblk, ext4_fsblk_t pblk,
2890                            int len, int replay);
2891
2892 /* mballoc.c */
2893 extern const struct seq_operations ext4_mb_seq_groups_ops;
2894 extern const struct seq_operations ext4_mb_seq_structs_summary_ops;
2895 extern long ext4_mb_stats;
2896 extern long ext4_mb_max_to_scan;
2897 extern int ext4_seq_mb_stats_show(struct seq_file *seq, void *offset);
2898 extern int ext4_mb_init(struct super_block *);
2899 extern int ext4_mb_release(struct super_block *);
2900 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2901                                 struct ext4_allocation_request *, int *);
2902 extern int ext4_mb_reserve_blocks(struct super_block *, int);
2903 extern void ext4_discard_preallocations(struct inode *, unsigned int);
2904 extern int __init ext4_init_mballoc(void);
2905 extern void ext4_exit_mballoc(void);
2906 extern ext4_group_t ext4_mb_prefetch(struct super_block *sb,
2907                                      ext4_group_t group,
2908                                      unsigned int nr, int *cnt);
2909 extern void ext4_mb_prefetch_fini(struct super_block *sb, ext4_group_t group,
2910                                   unsigned int nr);
2911
2912 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2913                              struct buffer_head *bh, ext4_fsblk_t block,
2914                              unsigned long count, int flags);
2915 extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2916                                    ext4_group_t ngroups);
2917 extern int ext4_mb_add_groupinfo(struct super_block *sb,
2918                 ext4_group_t i, struct ext4_group_desc *desc);
2919 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2920                                 ext4_fsblk_t block, unsigned long count);
2921 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2922 extern void ext4_process_freed_data(struct super_block *sb, tid_t commit_tid);
2923 extern void ext4_mb_mark_bb(struct super_block *sb, ext4_fsblk_t block,
2924                        int len, int state);
2925
2926 /* inode.c */
2927 void ext4_inode_csum_set(struct inode *inode, struct ext4_inode *raw,
2928                          struct ext4_inode_info *ei);
2929 int ext4_inode_is_fast_symlink(struct inode *inode);
2930 struct buffer_head *ext4_getblk(handle_t *, struct inode *, ext4_lblk_t, int);
2931 struct buffer_head *ext4_bread(handle_t *, struct inode *, ext4_lblk_t, int);
2932 int ext4_bread_batch(struct inode *inode, ext4_lblk_t block, int bh_count,
2933                      bool wait, struct buffer_head **bhs);
2934 int ext4_get_block_unwritten(struct inode *inode, sector_t iblock,
2935                              struct buffer_head *bh_result, int create);
2936 int ext4_get_block(struct inode *inode, sector_t iblock,
2937                    struct buffer_head *bh_result, int create);
2938 int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2939                            struct buffer_head *bh, int create);
2940 int ext4_walk_page_buffers(handle_t *handle,
2941                            struct inode *inode,
2942                            struct buffer_head *head,
2943                            unsigned from,
2944                            unsigned to,
2945                            int *partial,
2946                            int (*fn)(handle_t *handle, struct inode *inode,
2947                                      struct buffer_head *bh));
2948 int do_journal_get_write_access(handle_t *handle, struct inode *inode,
2949                                 struct buffer_head *bh);
2950 #define FALL_BACK_TO_NONDELALLOC 1
2951 #define CONVERT_INLINE_DATA      2
2952
2953 typedef enum {
2954         EXT4_IGET_NORMAL =      0,
2955         EXT4_IGET_SPECIAL =     0x0001, /* OK to iget a system inode */
2956         EXT4_IGET_HANDLE =      0x0002  /* Inode # is from a handle */
2957 } ext4_iget_flags;
2958
2959 extern struct inode *__ext4_iget(struct super_block *sb, unsigned long ino,
2960                                  ext4_iget_flags flags, const char *function,
2961                                  unsigned int line);
2962
2963 #define ext4_iget(sb, ino, flags) \
2964         __ext4_iget((sb), (ino), (flags), __func__, __LINE__)
2965
2966 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
2967 extern int  ext4_setattr(struct user_namespace *, struct dentry *,
2968                          struct iattr *);
2969 extern int  ext4_getattr(struct user_namespace *, const struct path *,
2970                          struct kstat *, u32, unsigned int);
2971 extern void ext4_evict_inode(struct inode *);
2972 extern void ext4_clear_inode(struct inode *);
2973 extern int  ext4_file_getattr(struct user_namespace *, const struct path *,
2974                               struct kstat *, u32, unsigned int);
2975 extern int  ext4_sync_inode(handle_t *, struct inode *);
2976 extern void ext4_dirty_inode(struct inode *, int);
2977 extern int ext4_change_inode_journal_flag(struct inode *, int);
2978 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
2979 extern int ext4_get_fc_inode_loc(struct super_block *sb, unsigned long ino,
2980                           struct ext4_iloc *iloc);
2981 extern int ext4_inode_attach_jinode(struct inode *inode);
2982 extern int ext4_can_truncate(struct inode *inode);
2983 extern int ext4_truncate(struct inode *);
2984 extern int ext4_break_layouts(struct inode *);
2985 extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length);
2986 extern void ext4_set_inode_flags(struct inode *, bool init);
2987 extern int ext4_alloc_da_blocks(struct inode *inode);
2988 extern void ext4_set_aops(struct inode *inode);
2989 extern int ext4_writepage_trans_blocks(struct inode *);
2990 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
2991 extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
2992                              loff_t lstart, loff_t lend);
2993 extern vm_fault_t ext4_page_mkwrite(struct vm_fault *vmf);
2994 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
2995 extern int ext4_get_projid(struct inode *inode, kprojid_t *projid);
2996 extern void ext4_da_release_space(struct inode *inode, int to_free);
2997 extern void ext4_da_update_reserve_space(struct inode *inode,
2998                                         int used, int quota_claim);
2999 extern int ext4_issue_zeroout(struct inode *inode, ext4_lblk_t lblk,
3000                               ext4_fsblk_t pblk, ext4_lblk_t len);
3001
3002 /* indirect.c */
3003 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
3004                                 struct ext4_map_blocks *map, int flags);
3005 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
3006 extern void ext4_ind_truncate(handle_t *, struct inode *inode);
3007 extern int ext4_ind_remove_space(handle_t *handle, struct inode *inode,
3008                                  ext4_lblk_t start, ext4_lblk_t end);
3009
3010 /* ioctl.c */
3011 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
3012 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
3013 int ext4_fileattr_set(struct user_namespace *mnt_userns,
3014                       struct dentry *dentry, struct fileattr *fa);
3015 int ext4_fileattr_get(struct dentry *dentry, struct fileattr *fa);
3016 extern void ext4_reset_inode_seed(struct inode *inode);
3017 int ext4_update_overhead(struct super_block *sb);
3018
3019 /* migrate.c */
3020 extern int ext4_ext_migrate(struct inode *);
3021 extern int ext4_ind_migrate(struct inode *inode);
3022
3023 /* namei.c */
3024 extern int ext4_init_new_dir(handle_t *handle, struct inode *dir,
3025                              struct inode *inode);
3026 extern int ext4_dirblock_csum_verify(struct inode *inode,
3027                                      struct buffer_head *bh);
3028 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
3029                                 __u32 start_minor_hash, __u32 *next_hash);
3030 extern int ext4_search_dir(struct buffer_head *bh,
3031                            char *search_buf,
3032                            int buf_size,
3033                            struct inode *dir,
3034                            struct ext4_filename *fname,
3035                            unsigned int offset,
3036                            struct ext4_dir_entry_2 **res_dir);
3037 extern int ext4_generic_delete_entry(struct inode *dir,
3038                                      struct ext4_dir_entry_2 *de_del,
3039                                      struct buffer_head *bh,
3040                                      void *entry_buf,
3041                                      int buf_size,
3042                                      int csum_size);
3043 extern bool ext4_empty_dir(struct inode *inode);
3044
3045 /* resize.c */
3046 extern void ext4_kvfree_array_rcu(void *to_free);
3047 extern int ext4_group_add(struct super_block *sb,
3048                                 struct ext4_new_group_data *input);
3049 extern int ext4_group_extend(struct super_block *sb,
3050                                 struct ext4_super_block *es,
3051                                 ext4_fsblk_t n_blocks_count);
3052 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
3053 extern unsigned int ext4_list_backups(struct super_block *sb,
3054                                       unsigned int *three, unsigned int *five,
3055                                       unsigned int *seven);
3056
3057 /* super.c */
3058 extern struct buffer_head *ext4_sb_bread(struct super_block *sb,
3059                                          sector_t block, int op_flags);
3060 extern struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
3061                                                    sector_t block);
3062 extern void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
3063                                 bh_end_io_t *end_io);
3064 extern int ext4_read_bh(struct buffer_head *bh, int op_flags,
3065                         bh_end_io_t *end_io);
3066 extern int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait);
3067 extern void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block);
3068 extern int ext4_seq_options_show(struct seq_file *seq, void *offset);
3069 extern int ext4_calculate_overhead(struct super_block *sb);
3070 extern __le32 ext4_superblock_csum(struct super_block *sb,
3071                                    struct ext4_super_block *es);
3072 extern void ext4_superblock_csum_set(struct super_block *sb);
3073 extern int ext4_alloc_flex_bg_array(struct super_block *sb,
3074                                     ext4_group_t ngroup);
3075 extern const char *ext4_decode_error(struct super_block *sb, int errno,
3076                                      char nbuf[16]);
3077 extern void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
3078                                              ext4_group_t block_group,
3079                                              unsigned int flags);
3080
3081 extern __printf(7, 8)
3082 void __ext4_error(struct super_block *, const char *, unsigned int, bool,
3083                   int, __u64, const char *, ...);
3084 extern __printf(6, 7)
3085 void __ext4_error_inode(struct inode *, const char *, unsigned int,
3086                         ext4_fsblk_t, int, const char *, ...);
3087 extern __printf(5, 6)
3088 void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
3089                      const char *, ...);
3090 extern void __ext4_std_error(struct super_block *, const char *,
3091                              unsigned int, int);
3092 extern __printf(4, 5)
3093 void __ext4_warning(struct super_block *, const char *, unsigned int,
3094                     const char *, ...);
3095 extern __printf(4, 5)
3096 void __ext4_warning_inode(const struct inode *inode, const char *function,
3097                           unsigned int line, const char *fmt, ...);
3098 extern __printf(3, 4)
3099 void __ext4_msg(struct super_block *, const char *, const char *, ...);
3100 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
3101                            const char *, unsigned int, const char *);
3102 extern __printf(7, 8)
3103 void __ext4_grp_locked_error(const char *, unsigned int,
3104                              struct super_block *, ext4_group_t,
3105                              unsigned long, ext4_fsblk_t,
3106                              const char *, ...);
3107
3108 #define EXT4_ERROR_INODE(inode, fmt, a...) \
3109         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
3110
3111 #define EXT4_ERROR_INODE_ERR(inode, err, fmt, a...)                     \
3112         __ext4_error_inode((inode), __func__, __LINE__, 0, (err), (fmt), ## a)
3113
3114 #define ext4_error_inode_block(inode, block, err, fmt, a...)            \
3115         __ext4_error_inode((inode), __func__, __LINE__, (block), (err), \
3116                            (fmt), ## a)
3117
3118 #define EXT4_ERROR_FILE(file, block, fmt, a...)                         \
3119         ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
3120
3121 #define ext4_abort(sb, err, fmt, a...)                                  \
3122         __ext4_error((sb), __func__, __LINE__, true, (err), 0, (fmt), ## a)
3123
3124 #ifdef CONFIG_PRINTK
3125
3126 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
3127         __ext4_error_inode(inode, func, line, block, 0, fmt, ##__VA_ARGS__)
3128 #define ext4_error_inode_err(inode, func, line, block, err, fmt, ...)   \
3129         __ext4_error_inode((inode), (func), (line), (block),            \
3130                            (err), (fmt), ##__VA_ARGS__)
3131 #define ext4_error_file(file, func, line, block, fmt, ...)              \
3132         __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__)
3133 #define ext4_error(sb, fmt, ...)                                        \
3134         __ext4_error((sb), __func__, __LINE__, false, 0, 0, (fmt),      \
3135                 ##__VA_ARGS__)
3136 #define ext4_error_err(sb, err, fmt, ...)                               \
3137         __ext4_error((sb), __func__, __LINE__, false, (err), 0, (fmt),  \
3138                 ##__VA_ARGS__)
3139 #define ext4_warning(sb, fmt, ...)                                      \
3140         __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
3141 #define ext4_warning_inode(inode, fmt, ...)                             \
3142         __ext4_warning_inode(inode, __func__, __LINE__, fmt, ##__VA_ARGS__)
3143 #define ext4_msg(sb, level, fmt, ...)                           \
3144         __ext4_msg(sb, level, fmt, ##__VA_ARGS__)
3145 #define dump_mmp_msg(sb, mmp, msg)                                      \
3146         __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg)
3147 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
3148         __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \
3149                                 fmt, ##__VA_ARGS__)
3150
3151 #else
3152
3153 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
3154 do {                                                                    \
3155         no_printk(fmt, ##__VA_ARGS__);                                  \
3156         __ext4_error_inode(inode, "", 0, block, 0, " ");                \
3157 } while (0)
3158 #define ext4_error_inode_err(inode, func, line, block, err, fmt, ...)   \
3159 do {                                                                    \
3160         no_printk(fmt, ##__VA_ARGS__);                                  \
3161         __ext4_error_inode(inode, "", 0, block, err, " ");              \
3162 } while (0)
3163 #define ext4_error_file(file, func, line, block, fmt, ...)              \
3164 do {                                                                    \
3165         no_printk(fmt, ##__VA_ARGS__);                                  \
3166         __ext4_error_file(file, "", 0, block, " ");                     \
3167 } while (0)
3168 #define ext4_error(sb, fmt, ...)                                        \
3169 do {                                                                    \
3170         no_printk(fmt, ##__VA_ARGS__);                                  \
3171         __ext4_error(sb, "", 0, false, 0, 0, " ");                      \
3172 } while (0)
3173 #define ext4_error_err(sb, err, fmt, ...)                               \
3174 do {                                                                    \
3175         no_printk(fmt, ##__VA_ARGS__);                                  \
3176         __ext4_error(sb, "", 0, false, err, 0, " ");                    \
3177 } while (0)
3178 #define ext4_warning(sb, fmt, ...)                                      \
3179 do {                                                                    \
3180         no_printk(fmt, ##__VA_ARGS__);                                  \
3181         __ext4_warning(sb, "", 0, " ");                                 \
3182 } while (0)
3183 #define ext4_warning_inode(inode, fmt, ...)                             \
3184 do {                                                                    \
3185         no_printk(fmt, ##__VA_ARGS__);                                  \
3186         __ext4_warning_inode(inode, "", 0, " ");                        \
3187 } while (0)
3188 #define ext4_msg(sb, level, fmt, ...)                                   \
3189 do {                                                                    \
3190         no_printk(fmt, ##__VA_ARGS__);                                  \
3191         __ext4_msg(sb, "", " ");                                        \
3192 } while (0)
3193 #define dump_mmp_msg(sb, mmp, msg)                                      \
3194         __dump_mmp_msg(sb, mmp, "", 0, "")
3195 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
3196 do {                                                                    \
3197         no_printk(fmt, ##__VA_ARGS__);                          \
3198         __ext4_grp_locked_error("", 0, sb, grp, ino, block, " ");       \
3199 } while (0)
3200
3201 #endif
3202
3203 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
3204                                       struct ext4_group_desc *bg);
3205 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
3206                                       struct ext4_group_desc *bg);
3207 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
3208                                      struct ext4_group_desc *bg);
3209 extern __u32 ext4_free_group_clusters(struct super_block *sb,
3210                                       struct ext4_group_desc *bg);
3211 extern __u32 ext4_free_inodes_count(struct super_block *sb,
3212                                  struct ext4_group_desc *bg);
3213 extern __u32 ext4_used_dirs_count(struct super_block *sb,
3214                                 struct ext4_group_desc *bg);
3215 extern __u32 ext4_itable_unused_count(struct super_block *sb,
3216                                    struct ext4_group_desc *bg);
3217 extern void ext4_block_bitmap_set(struct super_block *sb,
3218                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
3219 extern void ext4_inode_bitmap_set(struct super_block *sb,
3220                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
3221 extern void ext4_inode_table_set(struct super_block *sb,
3222                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
3223 extern void ext4_free_group_clusters_set(struct super_block *sb,
3224                                          struct ext4_group_desc *bg,
3225                                          __u32 count);
3226 extern void ext4_free_inodes_set(struct super_block *sb,
3227                                 struct ext4_group_desc *bg, __u32 count);
3228 extern void ext4_used_dirs_set(struct super_block *sb,
3229                                 struct ext4_group_desc *bg, __u32 count);
3230 extern void ext4_itable_unused_set(struct super_block *sb,
3231                                    struct ext4_group_desc *bg, __u32 count);
3232 extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
3233                                        struct ext4_group_desc *gdp);
3234 extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
3235                                      struct ext4_group_desc *gdp);
3236 extern int ext4_register_li_request(struct super_block *sb,
3237                                     ext4_group_t first_not_zeroed);
3238
3239 static inline int ext4_has_metadata_csum(struct super_block *sb)
3240 {
3241         WARN_ON_ONCE(ext4_has_feature_metadata_csum(sb) &&
3242                      !EXT4_SB(sb)->s_chksum_driver);
3243
3244         return ext4_has_feature_metadata_csum(sb) &&
3245                (EXT4_SB(sb)->s_chksum_driver != NULL);
3246 }
3247
3248 static inline int ext4_has_group_desc_csum(struct super_block *sb)
3249 {
3250         return ext4_has_feature_gdt_csum(sb) || ext4_has_metadata_csum(sb);
3251 }
3252
3253 #define ext4_read_incompat_64bit_val(es, name) \
3254         (((es)->s_feature_incompat & cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT) \
3255                 ? (ext4_fsblk_t)le32_to_cpu(es->name##_hi) << 32 : 0) | \
3256                 le32_to_cpu(es->name##_lo))
3257
3258 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
3259 {
3260         return ext4_read_incompat_64bit_val(es, s_blocks_count);
3261 }
3262
3263 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
3264 {
3265         return ext4_read_incompat_64bit_val(es, s_r_blocks_count);
3266 }
3267
3268 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
3269 {
3270         return ext4_read_incompat_64bit_val(es, s_free_blocks_count);
3271 }
3272
3273 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
3274                                          ext4_fsblk_t blk)
3275 {
3276         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
3277         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
3278 }
3279
3280 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
3281                                               ext4_fsblk_t blk)
3282 {
3283         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
3284         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
3285 }
3286
3287 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
3288                                            ext4_fsblk_t blk)
3289 {
3290         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
3291         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
3292 }
3293
3294 static inline loff_t ext4_isize(struct super_block *sb,
3295                                 struct ext4_inode *raw_inode)
3296 {
3297         if (ext4_has_feature_largedir(sb) ||
3298             S_ISREG(le16_to_cpu(raw_inode->i_mode)))
3299                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
3300                         le32_to_cpu(raw_inode->i_size_lo);
3301
3302         return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
3303 }
3304
3305 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
3306 {
3307         raw_inode->i_size_lo = cpu_to_le32(i_size);
3308         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
3309 }
3310
3311 static inline
3312 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
3313                                             ext4_group_t group)
3314 {
3315          struct ext4_group_info **grp_info;
3316          long indexv, indexh;
3317          BUG_ON(group >= EXT4_SB(sb)->s_groups_count);
3318          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
3319          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
3320          grp_info = sbi_array_rcu_deref(EXT4_SB(sb), s_group_info, indexv);
3321          return grp_info[indexh];
3322 }
3323
3324 /*
3325  * Reading s_groups_count requires using smp_rmb() afterwards.  See
3326  * the locking protocol documented in the comments of ext4_group_add()
3327  * in resize.c
3328  */
3329 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
3330 {
3331         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
3332
3333         smp_rmb();
3334         return ngroups;
3335 }
3336
3337 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
3338                                              ext4_group_t block_group)
3339 {
3340         return block_group >> sbi->s_log_groups_per_flex;
3341 }
3342
3343 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
3344 {
3345         return 1 << sbi->s_log_groups_per_flex;
3346 }
3347
3348 #define ext4_std_error(sb, errno)                               \
3349 do {                                                            \
3350         if ((errno))                                            \
3351                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
3352 } while (0)
3353
3354 #ifdef CONFIG_SMP
3355 /* Each CPU can accumulate percpu_counter_batch clusters in their local
3356  * counters. So we need to make sure we have free clusters more
3357  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
3358  */
3359 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
3360 #else
3361 #define EXT4_FREECLUSTERS_WATERMARK 0
3362 #endif
3363
3364 /* Update i_disksize. Requires i_rwsem to avoid races with truncate */
3365 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
3366 {
3367         WARN_ON_ONCE(S_ISREG(inode->i_mode) &&
3368                      !inode_is_locked(inode));
3369         down_write(&EXT4_I(inode)->i_data_sem);
3370         if (newsize > EXT4_I(inode)->i_disksize)
3371                 WRITE_ONCE(EXT4_I(inode)->i_disksize, newsize);
3372         up_write(&EXT4_I(inode)->i_data_sem);
3373 }
3374
3375 /* Update i_size, i_disksize. Requires i_rwsem to avoid races with truncate */
3376 static inline int ext4_update_inode_size(struct inode *inode, loff_t newsize)
3377 {
3378         int changed = 0;
3379
3380         if (newsize > inode->i_size) {
3381                 i_size_write(inode, newsize);
3382                 changed = 1;
3383         }
3384         if (newsize > EXT4_I(inode)->i_disksize) {
3385                 ext4_update_i_disksize(inode, newsize);
3386                 changed |= 2;
3387         }
3388         return changed;
3389 }
3390
3391 int ext4_update_disksize_before_punch(struct inode *inode, loff_t offset,
3392                                       loff_t len);
3393
3394 struct ext4_group_info {
3395         unsigned long   bb_state;
3396 #ifdef AGGRESSIVE_CHECK
3397         unsigned long   bb_check_counter;
3398 #endif
3399         struct rb_root  bb_free_root;
3400         ext4_grpblk_t   bb_first_free;  /* first free block */
3401         ext4_grpblk_t   bb_free;        /* total free blocks */
3402         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
3403         int             bb_avg_fragment_size_order;     /* order of average
3404                                                            fragment in BG */
3405         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
3406         ext4_group_t    bb_group;       /* Group number */
3407         struct          list_head bb_prealloc_list;
3408 #ifdef DOUBLE_CHECK
3409         void            *bb_bitmap;
3410 #endif
3411         struct rw_semaphore alloc_sem;
3412         struct list_head bb_avg_fragment_size_node;
3413         struct list_head bb_largest_free_order_node;
3414         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
3415                                          * regions, index is order.
3416                                          * bb_counters[3] = 5 means
3417                                          * 5 free 8-block regions. */
3418 };
3419
3420 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
3421 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
3422 #define EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT     2
3423 #define EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT     3
3424 #define EXT4_GROUP_INFO_BBITMAP_CORRUPT         \
3425         (1 << EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT)
3426 #define EXT4_GROUP_INFO_IBITMAP_CORRUPT         \
3427         (1 << EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT)
3428 #define EXT4_GROUP_INFO_BBITMAP_READ_BIT        4
3429
3430 #define EXT4_MB_GRP_NEED_INIT(grp)      \
3431         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
3432 #define EXT4_MB_GRP_BBITMAP_CORRUPT(grp)        \
3433         (test_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
3434 #define EXT4_MB_GRP_IBITMAP_CORRUPT(grp)        \
3435         (test_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
3436
3437 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
3438         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3439 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
3440         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3441 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
3442         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3443 #define EXT4_MB_GRP_TEST_AND_SET_READ(grp)      \
3444         (test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_READ_BIT, &((grp)->bb_state)))
3445
3446 #define EXT4_MAX_CONTENTION             8
3447 #define EXT4_CONTENTION_THRESHOLD       2
3448
3449 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
3450                                               ext4_group_t group)
3451 {
3452         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
3453 }
3454
3455 /*
3456  * Returns true if the filesystem is busy enough that attempts to
3457  * access the block group locks has run into contention.
3458  */
3459 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
3460 {
3461         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
3462 }
3463
3464 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
3465 {
3466         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
3467         if (spin_trylock(lock))
3468                 /*
3469                  * We're able to grab the lock right away, so drop the
3470                  * lock contention counter.
3471                  */
3472                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
3473         else {
3474                 /*
3475                  * The lock is busy, so bump the contention counter,
3476                  * and then wait on the spin lock.
3477                  */
3478                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
3479                                   EXT4_MAX_CONTENTION);
3480                 spin_lock(lock);
3481         }
3482 }
3483
3484 static inline void ext4_unlock_group(struct super_block *sb,
3485                                         ext4_group_t group)
3486 {
3487         spin_unlock(ext4_group_lock_ptr(sb, group));
3488 }
3489
3490 #ifdef CONFIG_QUOTA
3491 static inline bool ext4_quota_capable(struct super_block *sb)
3492 {
3493         return (test_opt(sb, QUOTA) || ext4_has_feature_quota(sb));
3494 }
3495
3496 static inline bool ext4_is_quota_journalled(struct super_block *sb)
3497 {
3498         struct ext4_sb_info *sbi = EXT4_SB(sb);
3499
3500         return (ext4_has_feature_quota(sb) ||
3501                 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]);
3502 }
3503 int ext4_enable_quotas(struct super_block *sb);
3504 #endif
3505
3506 /*
3507  * Block validity checking
3508  */
3509 #define ext4_check_indirect_blockref(inode, bh)                         \
3510         ext4_check_blockref(__func__, __LINE__, inode,                  \
3511                             (__le32 *)(bh)->b_data,                     \
3512                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
3513
3514 #define ext4_ind_check_inode(inode)                                     \
3515         ext4_check_blockref(__func__, __LINE__, inode,                  \
3516                             EXT4_I(inode)->i_data,                      \
3517                             EXT4_NDIR_BLOCKS)
3518
3519 /*
3520  * Inodes and files operations
3521  */
3522
3523 /* dir.c */
3524 extern const struct file_operations ext4_dir_operations;
3525
3526 /* file.c */
3527 extern const struct inode_operations ext4_file_inode_operations;
3528 extern const struct file_operations ext4_file_operations;
3529 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
3530
3531 /* inline.c */
3532 extern int ext4_get_max_inline_size(struct inode *inode);
3533 extern int ext4_find_inline_data_nolock(struct inode *inode);
3534 extern int ext4_init_inline_data(handle_t *handle, struct inode *inode,
3535                                  unsigned int len);
3536 extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
3537
3538 extern int ext4_readpage_inline(struct inode *inode, struct page *page);
3539 extern int ext4_try_to_write_inline_data(struct address_space *mapping,
3540                                          struct inode *inode,
3541                                          loff_t pos, unsigned len,
3542                                          struct page **pagep);
3543 extern int ext4_write_inline_data_end(struct inode *inode,
3544                                       loff_t pos, unsigned len,
3545                                       unsigned copied,
3546                                       struct page *page);
3547 extern struct buffer_head *
3548 ext4_journalled_write_inline_data(struct inode *inode,
3549                                   unsigned len,
3550                                   struct page *page);
3551 extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
3552                                            struct inode *inode,
3553                                            loff_t pos, unsigned len,
3554                                            struct page **pagep,
3555                                            void **fsdata);
3556 extern int ext4_try_add_inline_entry(handle_t *handle,
3557                                      struct ext4_filename *fname,
3558                                      struct inode *dir, struct inode *inode);
3559 extern int ext4_try_create_inline_dir(handle_t *handle,
3560                                       struct inode *parent,
3561                                       struct inode *inode);
3562 extern int ext4_read_inline_dir(struct file *filp,
3563                                 struct dir_context *ctx,
3564                                 int *has_inline_data);
3565 extern int ext4_inlinedir_to_tree(struct file *dir_file,
3566                                   struct inode *dir, ext4_lblk_t block,
3567                                   struct dx_hash_info *hinfo,
3568                                   __u32 start_hash, __u32 start_minor_hash,
3569                                   int *has_inline_data);
3570 extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
3571                                         struct ext4_filename *fname,
3572                                         struct ext4_dir_entry_2 **res_dir,
3573                                         int *has_inline_data);
3574 extern int ext4_delete_inline_entry(handle_t *handle,
3575                                     struct inode *dir,
3576                                     struct ext4_dir_entry_2 *de_del,
3577                                     struct buffer_head *bh,
3578                                     int *has_inline_data);
3579 extern bool empty_inline_dir(struct inode *dir, int *has_inline_data);
3580 extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
3581                                         struct ext4_dir_entry_2 **parent_de,
3582                                         int *retval);
3583 extern int ext4_inline_data_fiemap(struct inode *inode,
3584                                    struct fiemap_extent_info *fieinfo,
3585                                    int *has_inline, __u64 start, __u64 len);
3586 extern void *ext4_read_inline_link(struct inode *inode);
3587
3588 struct iomap;
3589 extern int ext4_inline_data_iomap(struct inode *inode, struct iomap *iomap);
3590
3591 extern int ext4_inline_data_truncate(struct inode *inode, int *has_inline);
3592
3593 extern int ext4_convert_inline_data(struct inode *inode);
3594
3595 static inline int ext4_has_inline_data(struct inode *inode)
3596 {
3597         return ext4_test_inode_flag(inode, EXT4_INODE_INLINE_DATA) &&
3598                EXT4_I(inode)->i_inline_off;
3599 }
3600
3601 /* namei.c */
3602 extern const struct inode_operations ext4_dir_inode_operations;
3603 extern const struct inode_operations ext4_special_inode_operations;
3604 extern struct dentry *ext4_get_parent(struct dentry *child);
3605 extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
3606                                  struct ext4_dir_entry_2 *de,
3607                                  int blocksize, int csum_size,
3608                                  unsigned int parent_ino, int dotdot_real_len);
3609 extern void ext4_initialize_dirent_tail(struct buffer_head *bh,
3610                                         unsigned int blocksize);
3611 extern int ext4_handle_dirty_dirblock(handle_t *handle, struct inode *inode,
3612                                       struct buffer_head *bh);
3613 extern int __ext4_unlink(handle_t *handle, struct inode *dir, const struct qstr *d_name,
3614                          struct inode *inode);
3615 extern int __ext4_link(struct inode *dir, struct inode *inode,
3616                        struct dentry *dentry);
3617
3618 #define S_SHIFT 12
3619 static const unsigned char ext4_type_by_mode[(S_IFMT >> S_SHIFT) + 1] = {
3620         [S_IFREG >> S_SHIFT]    = EXT4_FT_REG_FILE,
3621         [S_IFDIR >> S_SHIFT]    = EXT4_FT_DIR,
3622         [S_IFCHR >> S_SHIFT]    = EXT4_FT_CHRDEV,
3623         [S_IFBLK >> S_SHIFT]    = EXT4_FT_BLKDEV,
3624         [S_IFIFO >> S_SHIFT]    = EXT4_FT_FIFO,
3625         [S_IFSOCK >> S_SHIFT]   = EXT4_FT_SOCK,
3626         [S_IFLNK >> S_SHIFT]    = EXT4_FT_SYMLINK,
3627 };
3628
3629 static inline void ext4_set_de_type(struct super_block *sb,
3630                                 struct ext4_dir_entry_2 *de,
3631                                 umode_t mode) {
3632         if (ext4_has_feature_filetype(sb))
3633                 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
3634 }
3635
3636 /* readpages.c */
3637 extern int ext4_mpage_readpages(struct inode *inode,
3638                 struct readahead_control *rac, struct page *page);
3639 extern int __init ext4_init_post_read_processing(void);
3640 extern void ext4_exit_post_read_processing(void);
3641
3642 /* symlink.c */
3643 extern const struct inode_operations ext4_encrypted_symlink_inode_operations;
3644 extern const struct inode_operations ext4_symlink_inode_operations;
3645 extern const struct inode_operations ext4_fast_symlink_inode_operations;
3646
3647 /* sysfs.c */
3648 extern void ext4_notify_error_sysfs(struct ext4_sb_info *sbi);
3649 extern int ext4_register_sysfs(struct super_block *sb);
3650 extern void ext4_unregister_sysfs(struct super_block *sb);
3651 extern int __init ext4_init_sysfs(void);
3652 extern void ext4_exit_sysfs(void);
3653
3654 /* block_validity */
3655 extern void ext4_release_system_zone(struct super_block *sb);
3656 extern int ext4_setup_system_zone(struct super_block *sb);
3657 extern int __init ext4_init_system_zone(void);
3658 extern void ext4_exit_system_zone(void);
3659 extern int ext4_inode_block_valid(struct inode *inode,
3660                                   ext4_fsblk_t start_blk,
3661                                   unsigned int count);
3662 extern int ext4_check_blockref(const char *, unsigned int,
3663                                struct inode *, __le32 *, unsigned int);
3664 extern int ext4_sb_block_valid(struct super_block *sb, struct inode *inode,
3665                                 ext4_fsblk_t start_blk, unsigned int count);
3666
3667
3668 /* extents.c */
3669 struct ext4_ext_path;
3670 struct ext4_extent;
3671
3672 /*
3673  * Maximum number of logical blocks in a file; ext4_extent's ee_block is
3674  * __le32.
3675  */
3676 #define EXT_MAX_BLOCKS  0xffffffff
3677
3678 extern void ext4_ext_tree_init(handle_t *handle, struct inode *inode);
3679 extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
3680 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
3681                                struct ext4_map_blocks *map, int flags);
3682 extern int ext4_ext_truncate(handle_t *, struct inode *);
3683 extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
3684                                  ext4_lblk_t end);
3685 extern void ext4_ext_init(struct super_block *);
3686 extern void ext4_ext_release(struct super_block *);
3687 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
3688                           loff_t len);
3689 extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
3690                                           loff_t offset, ssize_t len);
3691 extern int ext4_convert_unwritten_io_end_vec(handle_t *handle,
3692                                              ext4_io_end_t *io_end);
3693 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
3694                            struct ext4_map_blocks *map, int flags);
3695 extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
3696                                                    int num,
3697                                                    struct ext4_ext_path *path);
3698 extern int ext4_ext_insert_extent(handle_t *, struct inode *,
3699                                   struct ext4_ext_path **,
3700                                   struct ext4_extent *, int);
3701 extern struct ext4_ext_path *ext4_find_extent(struct inode *, ext4_lblk_t,
3702                                               struct ext4_ext_path **,
3703                                               int flags);
3704 extern void ext4_ext_drop_refs(struct ext4_ext_path *);
3705 extern int ext4_ext_check_inode(struct inode *inode);
3706 extern ext4_lblk_t ext4_ext_next_allocated_block(struct ext4_ext_path *path);
3707 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
3708                         __u64 start, __u64 len);
3709 extern int ext4_get_es_cache(struct inode *inode,
3710                              struct fiemap_extent_info *fieinfo,
3711                              __u64 start, __u64 len);
3712 extern int ext4_ext_precache(struct inode *inode);
3713 extern int ext4_swap_extents(handle_t *handle, struct inode *inode1,
3714                                 struct inode *inode2, ext4_lblk_t lblk1,
3715                              ext4_lblk_t lblk2,  ext4_lblk_t count,
3716                              int mark_unwritten,int *err);
3717 extern int ext4_clu_mapped(struct inode *inode, ext4_lblk_t lclu);
3718 extern int ext4_datasem_ensure_credits(handle_t *handle, struct inode *inode,
3719                                        int check_cred, int restart_cred,
3720                                        int revoke_cred);
3721 extern void ext4_ext_replay_shrink_inode(struct inode *inode, ext4_lblk_t end);
3722 extern int ext4_ext_replay_set_iblocks(struct inode *inode);
3723 extern int ext4_ext_replay_update_ex(struct inode *inode, ext4_lblk_t start,
3724                 int len, int unwritten, ext4_fsblk_t pblk);
3725 extern int ext4_ext_clear_bb(struct inode *inode);
3726
3727
3728 /* move_extent.c */
3729 extern void ext4_double_down_write_data_sem(struct inode *first,
3730                                             struct inode *second);
3731 extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
3732                                           struct inode *donor_inode);
3733 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
3734                              __u64 start_orig, __u64 start_donor,
3735                              __u64 len, __u64 *moved_len);
3736
3737 /* page-io.c */
3738 extern int __init ext4_init_pageio(void);
3739 extern void ext4_exit_pageio(void);
3740 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
3741 extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
3742 extern int ext4_put_io_end(ext4_io_end_t *io_end);
3743 extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
3744 extern void ext4_io_submit_init(struct ext4_io_submit *io,
3745                                 struct writeback_control *wbc);
3746 extern void ext4_end_io_rsv_work(struct work_struct *work);
3747 extern void ext4_io_submit(struct ext4_io_submit *io);
3748 extern int ext4_bio_write_page(struct ext4_io_submit *io,
3749                                struct page *page,
3750                                int len,
3751                                bool keep_towrite);
3752 extern struct ext4_io_end_vec *ext4_alloc_io_end_vec(ext4_io_end_t *io_end);
3753 extern struct ext4_io_end_vec *ext4_last_io_end_vec(ext4_io_end_t *io_end);
3754
3755 /* mmp.c */
3756 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
3757
3758 /* mmp.c */
3759 extern void ext4_stop_mmpd(struct ext4_sb_info *sbi);
3760
3761 /* verity.c */
3762 extern const struct fsverity_operations ext4_verityops;
3763
3764 /* orphan.c */
3765 extern int ext4_orphan_add(handle_t *, struct inode *);
3766 extern int ext4_orphan_del(handle_t *, struct inode *);
3767 extern void ext4_orphan_cleanup(struct super_block *sb,
3768                                 struct ext4_super_block *es);
3769 extern void ext4_release_orphan_info(struct super_block *sb);
3770 extern int ext4_init_orphan_info(struct super_block *sb);
3771 extern int ext4_orphan_file_empty(struct super_block *sb);
3772 extern void ext4_orphan_file_block_trigger(
3773                                 struct jbd2_buffer_trigger_type *triggers,
3774                                 struct buffer_head *bh,
3775                                 void *data, size_t size);
3776
3777 /*
3778  * Add new method to test whether block and inode bitmaps are properly
3779  * initialized. With uninit_bg reading the block from disk is not enough
3780  * to mark the bitmap uptodate. We need to also zero-out the bitmap
3781  */
3782 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
3783
3784 static inline int bitmap_uptodate(struct buffer_head *bh)
3785 {
3786         return (buffer_uptodate(bh) &&
3787                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
3788 }
3789 static inline void set_bitmap_uptodate(struct buffer_head *bh)
3790 {
3791         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
3792 }
3793
3794 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
3795
3796 /* For ioend & aio unwritten conversion wait queues */
3797 #define EXT4_WQ_HASH_SZ         37
3798 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
3799                                             EXT4_WQ_HASH_SZ])
3800 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
3801
3802 extern int ext4_resize_begin(struct super_block *sb);
3803 extern void ext4_resize_end(struct super_block *sb);
3804
3805 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
3806                                               struct ext4_io_end *io_end)
3807 {
3808         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
3809                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
3810                 atomic_inc(&EXT4_I(inode)->i_unwritten);
3811         }
3812 }
3813
3814 static inline void ext4_clear_io_unwritten_flag(ext4_io_end_t *io_end)
3815 {
3816         struct inode *inode = io_end->inode;
3817
3818         if (io_end->flag & EXT4_IO_END_UNWRITTEN) {
3819                 io_end->flag &= ~EXT4_IO_END_UNWRITTEN;
3820                 /* Wake up anyone waiting on unwritten extent conversion */
3821                 if (atomic_dec_and_test(&EXT4_I(inode)->i_unwritten))
3822                         wake_up_all(ext4_ioend_wq(inode));
3823         }
3824 }
3825
3826 extern const struct iomap_ops ext4_iomap_ops;
3827 extern const struct iomap_ops ext4_iomap_overwrite_ops;
3828 extern const struct iomap_ops ext4_iomap_report_ops;
3829
3830 static inline int ext4_buffer_uptodate(struct buffer_head *bh)
3831 {
3832         /*
3833          * If the buffer has the write error flag, we have failed
3834          * to write out data in the block.  In this  case, we don't
3835          * have to read the block because we may read the old data
3836          * successfully.
3837          */
3838         if (buffer_write_io_error(bh))
3839                 set_buffer_uptodate(bh);
3840         return buffer_uptodate(bh);
3841 }
3842
3843 #endif  /* __KERNEL__ */
3844
3845 #define EFSBADCRC       EBADMSG         /* Bad CRC detected */
3846 #define EFSCORRUPTED    EUCLEAN         /* Filesystem is corrupted */
3847
3848 #endif  /* _EXT4_H */