ext4: Support for checksumming from journal triggers
[linux-2.6-block.git] / 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/types.h>
21 #include <linux/blkdev.h>
22 #include <linux/magic.h>
23 #include <linux/jbd2.h>
24 #include <linux/quota.h>
25 #include <linux/rwsem.h>
26 #include <linux/rbtree.h>
27 #include <linux/seqlock.h>
28 #include <linux/mutex.h>
29 #include <linux/timer.h>
30 #include <linux/wait.h>
31 #include <linux/sched/signal.h>
32 #include <linux/blockgroup_lock.h>
33 #include <linux/percpu_counter.h>
34 #include <linux/ratelimit.h>
35 #include <crypto/hash.h>
36 #include <linux/falloc.h>
37 #include <linux/percpu-rwsem.h>
38 #include <linux/fiemap.h>
39 #ifdef __KERNEL__
40 #include <linux/compat.h>
41 #endif
42
43 #include <linux/fscrypt.h>
44 #include <linux/fsverity.h>
45
46 #include <linux/compiler.h>
47
48 /*
49  * The fourth extended filesystem constants/structures
50  */
51
52 /*
53  * with AGGRESSIVE_CHECK allocator runs consistency checks over
54  * structures. these checks slow things down a lot
55  */
56 #define AGGRESSIVE_CHECK__
57
58 /*
59  * with DOUBLE_CHECK defined mballoc creates persistent in-core
60  * bitmaps, maintains and uses them to check for double allocations
61  */
62 #define DOUBLE_CHECK__
63
64 /*
65  * Define EXT4FS_DEBUG to produce debug messages
66  */
67 #undef EXT4FS_DEBUG
68
69 /*
70  * Debug code
71  */
72 #ifdef EXT4FS_DEBUG
73 #define ext4_debug(f, a...)                                             \
74         do {                                                            \
75                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
76                         __FILE__, __LINE__, __func__);                  \
77                 printk(KERN_DEBUG f, ## a);                             \
78         } while (0)
79 #else
80 #define ext4_debug(fmt, ...)    no_printk(fmt, ##__VA_ARGS__)
81 #endif
82
83  /*
84   * Turn on EXT_DEBUG to enable ext4_ext_show_path/leaf/move in extents.c
85   */
86 #define EXT_DEBUG__
87
88 /*
89  * Dynamic printk for controlled extents debugging.
90  */
91 #ifdef CONFIG_EXT4_DEBUG
92 #define ext_debug(ino, fmt, ...)                                        \
93         pr_debug("[%s/%d] EXT4-fs (%s): ino %lu: (%s, %d): %s:" fmt,    \
94                  current->comm, task_pid_nr(current),                   \
95                  ino->i_sb->s_id, ino->i_ino, __FILE__, __LINE__,       \
96                  __func__, ##__VA_ARGS__)
97 #else
98 #define ext_debug(ino, fmt, ...)        no_printk(fmt, ##__VA_ARGS__)
99 #endif
100
101 #define ASSERT(assert)                                          \
102 do {                                                                    \
103         if (unlikely(!(assert))) {                                      \
104                 printk(KERN_EMERG                                       \
105                        "Assertion failure in %s() at %s:%d: '%s'\n",    \
106                        __func__, __FILE__, __LINE__, #assert);          \
107                 BUG();                                                  \
108         }                                                               \
109 } while (0)
110
111 /* data type for block offset of block group */
112 typedef int ext4_grpblk_t;
113
114 /* data type for filesystem-wide blocks number */
115 typedef unsigned long long ext4_fsblk_t;
116
117 /* data type for file logical block number */
118 typedef __u32 ext4_lblk_t;
119
120 /* data type for block group number */
121 typedef unsigned int ext4_group_t;
122
123 enum SHIFT_DIRECTION {
124         SHIFT_LEFT = 0,
125         SHIFT_RIGHT,
126 };
127
128 /*
129  * Flags used in mballoc's allocation_context flags field.
130  *
131  * Also used to show what's going on for debugging purposes when the
132  * flag field is exported via the traceport interface
133  */
134
135 /* prefer goal again. length */
136 #define EXT4_MB_HINT_MERGE              0x0001
137 /* blocks already reserved */
138 #define EXT4_MB_HINT_RESERVED           0x0002
139 /* metadata is being allocated */
140 #define EXT4_MB_HINT_METADATA           0x0004
141 /* first blocks in the file */
142 #define EXT4_MB_HINT_FIRST              0x0008
143 /* search for the best chunk */
144 #define EXT4_MB_HINT_BEST               0x0010
145 /* data is being allocated */
146 #define EXT4_MB_HINT_DATA               0x0020
147 /* don't preallocate (for tails) */
148 #define EXT4_MB_HINT_NOPREALLOC         0x0040
149 /* allocate for locality group */
150 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
151 /* allocate goal blocks or none */
152 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
153 /* goal is meaningful */
154 #define EXT4_MB_HINT_TRY_GOAL           0x0200
155 /* blocks already pre-reserved by delayed allocation */
156 #define EXT4_MB_DELALLOC_RESERVED       0x0400
157 /* We are doing stream allocation */
158 #define EXT4_MB_STREAM_ALLOC            0x0800
159 /* Use reserved root blocks if needed */
160 #define EXT4_MB_USE_ROOT_BLOCKS         0x1000
161 /* Use blocks from reserved pool */
162 #define EXT4_MB_USE_RESERVED            0x2000
163 /* Do strict check for free blocks while retrying block allocation */
164 #define EXT4_MB_STRICT_CHECK            0x4000
165 /* Large fragment size list lookup succeeded at least once for cr = 0 */
166 #define EXT4_MB_CR0_OPTIMIZED           0x8000
167 /* Avg fragment size rb tree lookup succeeded at least once for cr = 1 */
168 #define EXT4_MB_CR1_OPTIMIZED           0x00010000
169 /* Perform linear traversal for one group */
170 #define EXT4_MB_SEARCH_NEXT_LINEAR      0x00020000
171 struct ext4_allocation_request {
172         /* target inode for block we're allocating */
173         struct inode *inode;
174         /* how many blocks we want to allocate */
175         unsigned int len;
176         /* logical block in target inode */
177         ext4_lblk_t logical;
178         /* the closest logical allocated block to the left */
179         ext4_lblk_t lleft;
180         /* the closest logical allocated block to the right */
181         ext4_lblk_t lright;
182         /* phys. target (a hint) */
183         ext4_fsblk_t goal;
184         /* phys. block for the closest logical allocated block to the left */
185         ext4_fsblk_t pleft;
186         /* phys. block for the closest logical allocated block to the right */
187         ext4_fsblk_t pright;
188         /* flags. see above EXT4_MB_HINT_* */
189         unsigned int flags;
190 };
191
192 /*
193  * Logical to physical block mapping, used by ext4_map_blocks()
194  *
195  * This structure is used to pass requests into ext4_map_blocks() as
196  * well as to store the information returned by ext4_map_blocks().  It
197  * takes less room on the stack than a struct buffer_head.
198  */
199 #define EXT4_MAP_NEW            BIT(BH_New)
200 #define EXT4_MAP_MAPPED         BIT(BH_Mapped)
201 #define EXT4_MAP_UNWRITTEN      BIT(BH_Unwritten)
202 #define EXT4_MAP_BOUNDARY       BIT(BH_Boundary)
203 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
204                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY)
205
206 struct ext4_map_blocks {
207         ext4_fsblk_t m_pblk;
208         ext4_lblk_t m_lblk;
209         unsigned int m_len;
210         unsigned int m_flags;
211 };
212
213 /*
214  * Block validity checking, system zone rbtree.
215  */
216 struct ext4_system_blocks {
217         struct rb_root root;
218         struct rcu_head rcu;
219 };
220
221 /*
222  * Flags for ext4_io_end->flags
223  */
224 #define EXT4_IO_END_UNWRITTEN   0x0001
225
226 struct ext4_io_end_vec {
227         struct list_head list;          /* list of io_end_vec */
228         loff_t offset;                  /* offset in the file */
229         ssize_t size;                   /* size of the extent */
230 };
231
232 /*
233  * For converting unwritten extents on a work queue. 'handle' is used for
234  * buffered writeback.
235  */
236 typedef struct ext4_io_end {
237         struct list_head        list;           /* per-file finished IO list */
238         handle_t                *handle;        /* handle reserved for extent
239                                                  * conversion */
240         struct inode            *inode;         /* file being written to */
241         struct bio              *bio;           /* Linked list of completed
242                                                  * bios covering the extent */
243         unsigned int            flag;           /* unwritten or not */
244         atomic_t                count;          /* reference counter */
245         struct list_head        list_vec;       /* list of ext4_io_end_vec */
246 } ext4_io_end_t;
247
248 struct ext4_io_submit {
249         struct writeback_control *io_wbc;
250         struct bio              *io_bio;
251         ext4_io_end_t           *io_end;
252         sector_t                io_next_block;
253 };
254
255 /*
256  * Special inodes numbers
257  */
258 #define EXT4_BAD_INO             1      /* Bad blocks inode */
259 #define EXT4_ROOT_INO            2      /* Root inode */
260 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
261 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
262 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
263 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
264 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
265 #define EXT4_JOURNAL_INO         8      /* Journal inode */
266
267 /* First non-reserved inode for old ext4 filesystems */
268 #define EXT4_GOOD_OLD_FIRST_INO 11
269
270 /*
271  * Maximal count of links to a file
272  */
273 #define EXT4_LINK_MAX           65000
274
275 /*
276  * Macro-instructions used to manage several block sizes
277  */
278 #define EXT4_MIN_BLOCK_SIZE             1024
279 #define EXT4_MAX_BLOCK_SIZE             65536
280 #define EXT4_MIN_BLOCK_LOG_SIZE         10
281 #define EXT4_MAX_BLOCK_LOG_SIZE         16
282 #define EXT4_MAX_CLUSTER_LOG_SIZE       30
283 #ifdef __KERNEL__
284 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
285 #else
286 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
287 #endif
288 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
289 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
290                                          EXT4_SB(s)->s_cluster_bits)
291 #ifdef __KERNEL__
292 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
293 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
294 #else
295 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
296 #endif
297 #ifdef __KERNEL__
298 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
299 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
300 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
301 #else
302 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
303                                  EXT4_GOOD_OLD_INODE_SIZE : \
304                                  (s)->s_inode_size)
305 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
306                                  EXT4_GOOD_OLD_FIRST_INO : \
307                                  (s)->s_first_ino)
308 #endif
309 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
310 #define EXT4_MAX_BLOCKS(size, offset, blkbits) \
311         ((EXT4_BLOCK_ALIGN(size + offset, blkbits) >> blkbits) - (offset >> \
312                                                                   blkbits))
313
314 /* Translate a block number to a cluster number */
315 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
316 /* Translate a cluster number to a block number */
317 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
318 /* Translate # of blks to # of clusters */
319 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
320                                  (sbi)->s_cluster_bits)
321 /* Mask out the low bits to get the starting block of the cluster */
322 #define EXT4_PBLK_CMASK(s, pblk) ((pblk) &                              \
323                                   ~((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
324 #define EXT4_LBLK_CMASK(s, lblk) ((lblk) &                              \
325                                   ~((ext4_lblk_t) (s)->s_cluster_ratio - 1))
326 /* Fill in the low bits to get the last block of the cluster */
327 #define EXT4_LBLK_CFILL(sbi, lblk) ((lblk) |                            \
328                                     ((ext4_lblk_t) (sbi)->s_cluster_ratio - 1))
329 /* Get the cluster offset */
330 #define EXT4_PBLK_COFF(s, pblk) ((pblk) &                               \
331                                  ((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
332 #define EXT4_LBLK_COFF(s, lblk) ((lblk) &                               \
333                                  ((ext4_lblk_t) (s)->s_cluster_ratio - 1))
334
335 /*
336  * Structure of a blocks group descriptor
337  */
338 struct ext4_group_desc
339 {
340         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
341         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
342         __le32  bg_inode_table_lo;      /* Inodes table block */
343         __le16  bg_free_blocks_count_lo;/* Free blocks count */
344         __le16  bg_free_inodes_count_lo;/* Free inodes count */
345         __le16  bg_used_dirs_count_lo;  /* Directories count */
346         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
347         __le32  bg_exclude_bitmap_lo;   /* Exclude bitmap for snapshots */
348         __le16  bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
349         __le16  bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
350         __le16  bg_itable_unused_lo;    /* Unused inodes count */
351         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
352         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
353         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
354         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
355         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
356         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
357         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
358         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
359         __le32  bg_exclude_bitmap_hi;   /* Exclude bitmap block MSB */
360         __le16  bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
361         __le16  bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
362         __u32   bg_reserved;
363 };
364
365 #define EXT4_BG_INODE_BITMAP_CSUM_HI_END        \
366         (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
367          sizeof(__le16))
368 #define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END        \
369         (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
370          sizeof(__le16))
371
372 /*
373  * Structure of a flex block group info
374  */
375
376 struct flex_groups {
377         atomic64_t      free_clusters;
378         atomic_t        free_inodes;
379         atomic_t        used_dirs;
380 };
381
382 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
383 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
384 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
385
386 /*
387  * Macro-instructions used to manage group descriptors
388  */
389 #define EXT4_MIN_DESC_SIZE              32
390 #define EXT4_MIN_DESC_SIZE_64BIT        64
391 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
392 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
393 #ifdef __KERNEL__
394 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
395 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
396 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
397 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
398 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
399 #else
400 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
401 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
402 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
403 #endif
404
405 /*
406  * Constants relative to the data blocks
407  */
408 #define EXT4_NDIR_BLOCKS                12
409 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
410 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
411 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
412 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
413
414 /*
415  * Inode flags
416  */
417 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
418 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
419 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
420 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
421 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
422 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
423 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
424 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
425 /* Reserved for compression usage... */
426 #define EXT4_DIRTY_FL                   0x00000100
427 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
428 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
429         /* nb: was previously EXT2_ECOMPR_FL */
430 #define EXT4_ENCRYPT_FL                 0x00000800 /* encrypted file */
431 /* End compression flags --- maybe not all used */
432 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
433 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
434 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
435 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
436 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
437 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
438 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
439 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
440 #define EXT4_VERITY_FL                  0x00100000 /* Verity protected inode */
441 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
442 /* 0x00400000 was formerly EXT4_EOFBLOCKS_FL */
443
444 #define EXT4_DAX_FL                     0x02000000 /* Inode is DAX */
445
446 #define EXT4_INLINE_DATA_FL             0x10000000 /* Inode has inline data. */
447 #define EXT4_PROJINHERIT_FL             0x20000000 /* Create with parents projid */
448 #define EXT4_CASEFOLD_FL                0x40000000 /* Casefolded directory */
449 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
450
451 /* User modifiable flags */
452 #define EXT4_FL_USER_MODIFIABLE         (EXT4_SECRM_FL | \
453                                          EXT4_UNRM_FL | \
454                                          EXT4_COMPR_FL | \
455                                          EXT4_SYNC_FL | \
456                                          EXT4_IMMUTABLE_FL | \
457                                          EXT4_APPEND_FL | \
458                                          EXT4_NODUMP_FL | \
459                                          EXT4_NOATIME_FL | \
460                                          EXT4_JOURNAL_DATA_FL | \
461                                          EXT4_NOTAIL_FL | \
462                                          EXT4_DIRSYNC_FL | \
463                                          EXT4_TOPDIR_FL | \
464                                          EXT4_EXTENTS_FL | \
465                                          0x00400000 /* EXT4_EOFBLOCKS_FL */ | \
466                                          EXT4_DAX_FL | \
467                                          EXT4_PROJINHERIT_FL | \
468                                          EXT4_CASEFOLD_FL)
469
470 /* User visible flags */
471 #define EXT4_FL_USER_VISIBLE            (EXT4_FL_USER_MODIFIABLE | \
472                                          EXT4_DIRTY_FL | \
473                                          EXT4_COMPRBLK_FL | \
474                                          EXT4_NOCOMPR_FL | \
475                                          EXT4_ENCRYPT_FL | \
476                                          EXT4_INDEX_FL | \
477                                          EXT4_VERITY_FL | \
478                                          EXT4_INLINE_DATA_FL)
479
480 /* Flags that should be inherited by new inodes from their parent. */
481 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
482                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
483                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
484                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL |\
485                            EXT4_PROJINHERIT_FL | EXT4_CASEFOLD_FL |\
486                            EXT4_DAX_FL)
487
488 /* Flags that are appropriate for regular files (all but dir-specific ones). */
489 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL | EXT4_CASEFOLD_FL |\
490                            EXT4_PROJINHERIT_FL))
491
492 /* Flags that are appropriate for non-directories/regular files. */
493 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
494
495 /* The only flags that should be swapped */
496 #define EXT4_FL_SHOULD_SWAP (EXT4_HUGE_FILE_FL | EXT4_EXTENTS_FL)
497
498 /* Flags which are mutually exclusive to DAX */
499 #define EXT4_DAX_MUT_EXCL (EXT4_VERITY_FL | EXT4_ENCRYPT_FL |\
500                            EXT4_JOURNAL_DATA_FL | EXT4_INLINE_DATA_FL)
501
502 /* Mask out flags that are inappropriate for the given type of inode. */
503 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
504 {
505         if (S_ISDIR(mode))
506                 return flags;
507         else if (S_ISREG(mode))
508                 return flags & EXT4_REG_FLMASK;
509         else
510                 return flags & EXT4_OTHER_FLMASK;
511 }
512
513 /*
514  * Inode flags used for atomic set/get
515  */
516 enum {
517         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
518         EXT4_INODE_UNRM         = 1,    /* Undelete */
519         EXT4_INODE_COMPR        = 2,    /* Compress file */
520         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
521         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
522         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
523         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
524         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
525 /* Reserved for compression usage... */
526         EXT4_INODE_DIRTY        = 8,
527         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
528         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
529         EXT4_INODE_ENCRYPT      = 11,   /* Encrypted file */
530 /* End compression flags --- maybe not all used */
531         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
532         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
533         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
534         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
535         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
536         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
537         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
538         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
539         EXT4_INODE_VERITY       = 20,   /* Verity protected inode */
540         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
541 /* 22 was formerly EXT4_INODE_EOFBLOCKS */
542         EXT4_INODE_DAX          = 25,   /* Inode is DAX */
543         EXT4_INODE_INLINE_DATA  = 28,   /* Data in inode. */
544         EXT4_INODE_PROJINHERIT  = 29,   /* Create with parents projid */
545         EXT4_INODE_CASEFOLD     = 30,   /* Casefolded directory */
546         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
547 };
548
549 /*
550  * Since it's pretty easy to mix up bit numbers and hex values, we use a
551  * build-time check to make sure that EXT4_XXX_FL is consistent with respect to
552  * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost
553  * any extra space in the compiled kernel image, otherwise, the build will fail.
554  * It's important that these values are the same, since we are using
555  * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent
556  * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk
557  * values found in ext2, ext3 and ext4 filesystems, and of course the values
558  * defined in e2fsprogs.
559  *
560  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
561  */
562 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
563 #define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG))
564
565 static inline void ext4_check_flag_values(void)
566 {
567         CHECK_FLAG_VALUE(SECRM);
568         CHECK_FLAG_VALUE(UNRM);
569         CHECK_FLAG_VALUE(COMPR);
570         CHECK_FLAG_VALUE(SYNC);
571         CHECK_FLAG_VALUE(IMMUTABLE);
572         CHECK_FLAG_VALUE(APPEND);
573         CHECK_FLAG_VALUE(NODUMP);
574         CHECK_FLAG_VALUE(NOATIME);
575         CHECK_FLAG_VALUE(DIRTY);
576         CHECK_FLAG_VALUE(COMPRBLK);
577         CHECK_FLAG_VALUE(NOCOMPR);
578         CHECK_FLAG_VALUE(ENCRYPT);
579         CHECK_FLAG_VALUE(INDEX);
580         CHECK_FLAG_VALUE(IMAGIC);
581         CHECK_FLAG_VALUE(JOURNAL_DATA);
582         CHECK_FLAG_VALUE(NOTAIL);
583         CHECK_FLAG_VALUE(DIRSYNC);
584         CHECK_FLAG_VALUE(TOPDIR);
585         CHECK_FLAG_VALUE(HUGE_FILE);
586         CHECK_FLAG_VALUE(EXTENTS);
587         CHECK_FLAG_VALUE(VERITY);
588         CHECK_FLAG_VALUE(EA_INODE);
589         CHECK_FLAG_VALUE(INLINE_DATA);
590         CHECK_FLAG_VALUE(PROJINHERIT);
591         CHECK_FLAG_VALUE(CASEFOLD);
592         CHECK_FLAG_VALUE(RESERVED);
593 }
594
595 /* Used to pass group descriptor data when online resize is done */
596 struct ext4_new_group_input {
597         __u32 group;            /* Group number for this data */
598         __u64 block_bitmap;     /* Absolute block number of block bitmap */
599         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
600         __u64 inode_table;      /* Absolute block number of inode table start */
601         __u32 blocks_count;     /* Total number of blocks in this group */
602         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
603         __u16 unused;
604 };
605
606 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
607 struct compat_ext4_new_group_input {
608         u32 group;
609         compat_u64 block_bitmap;
610         compat_u64 inode_bitmap;
611         compat_u64 inode_table;
612         u32 blocks_count;
613         u16 reserved_blocks;
614         u16 unused;
615 };
616 #endif
617
618 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
619 struct ext4_new_group_data {
620         __u32 group;
621         __u64 block_bitmap;
622         __u64 inode_bitmap;
623         __u64 inode_table;
624         __u32 blocks_count;
625         __u16 reserved_blocks;
626         __u16 mdata_blocks;
627         __u32 free_clusters_count;
628 };
629
630 /* Indexes used to index group tables in ext4_new_group_data */
631 enum {
632         BLOCK_BITMAP = 0,       /* block bitmap */
633         INODE_BITMAP,           /* inode bitmap */
634         INODE_TABLE,            /* inode tables */
635         GROUP_TABLE_COUNT,
636 };
637
638 /*
639  * Flags used by ext4_map_blocks()
640  */
641         /* Allocate any needed blocks and/or convert an unwritten
642            extent to be an initialized ext4 */
643 #define EXT4_GET_BLOCKS_CREATE                  0x0001
644         /* Request the creation of an unwritten extent */
645 #define EXT4_GET_BLOCKS_UNWRIT_EXT              0x0002
646 #define EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT       (EXT4_GET_BLOCKS_UNWRIT_EXT|\
647                                                  EXT4_GET_BLOCKS_CREATE)
648         /* Caller is from the delayed allocation writeout path
649          * finally doing the actual allocation of delayed blocks */
650 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
651         /* caller is from the direct IO path, request to creation of an
652         unwritten extents if not allocated, split the unwritten
653         extent if blocks has been preallocated already*/
654 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
655 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
656 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
657                                          EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
658         /* Convert extent to initialized after IO complete */
659 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
660                                          EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
661         /* Eventual metadata allocation (due to growing extent tree)
662          * should not fail, so try to use reserved blocks for that.*/
663 #define EXT4_GET_BLOCKS_METADATA_NOFAIL         0x0020
664         /* Don't normalize allocation size (used for fallocate) */
665 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
666         /* Convert written extents to unwritten */
667 #define EXT4_GET_BLOCKS_CONVERT_UNWRITTEN       0x0100
668         /* Write zeros to newly created written extents */
669 #define EXT4_GET_BLOCKS_ZERO                    0x0200
670 #define EXT4_GET_BLOCKS_CREATE_ZERO             (EXT4_GET_BLOCKS_CREATE |\
671                                         EXT4_GET_BLOCKS_ZERO)
672         /* Caller will submit data before dropping transaction handle. This
673          * allows jbd2 to avoid submitting data before commit. */
674 #define EXT4_GET_BLOCKS_IO_SUBMIT               0x0400
675
676 /*
677  * The bit position of these flags must not overlap with any of the
678  * EXT4_GET_BLOCKS_*.  They are used by ext4_find_extent(),
679  * read_extent_tree_block(), ext4_split_extent_at(),
680  * ext4_ext_insert_extent(), and ext4_ext_create_new_leaf().
681  * EXT4_EX_NOCACHE is used to indicate that the we shouldn't be
682  * caching the extents when reading from the extent tree while a
683  * truncate or punch hole operation is in progress.
684  */
685 #define EXT4_EX_NOCACHE                         0x40000000
686 #define EXT4_EX_FORCE_CACHE                     0x20000000
687 #define EXT4_EX_NOFAIL                          0x10000000
688
689 /*
690  * Flags used by ext4_free_blocks
691  */
692 #define EXT4_FREE_BLOCKS_METADATA               0x0001
693 #define EXT4_FREE_BLOCKS_FORGET                 0x0002
694 #define EXT4_FREE_BLOCKS_VALIDATED              0x0004
695 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE         0x0008
696 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
697 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
698 #define EXT4_FREE_BLOCKS_RERESERVE_CLUSTER      0x0040
699
700 /*
701  * ioctl commands
702  */
703 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
704 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
705 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
706 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
707 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
708 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
709 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
710 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
711 #define EXT4_IOC_MIGRATE                _IO('f', 9)
712  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
713  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
714 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
715 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
716 #define EXT4_IOC_RESIZE_FS              _IOW('f', 16, __u64)
717 #define EXT4_IOC_SWAP_BOOT              _IO('f', 17)
718 #define EXT4_IOC_PRECACHE_EXTENTS       _IO('f', 18)
719 /* ioctl codes 19--39 are reserved for fscrypt */
720 #define EXT4_IOC_CLEAR_ES_CACHE         _IO('f', 40)
721 #define EXT4_IOC_GETSTATE               _IOW('f', 41, __u32)
722 #define EXT4_IOC_GET_ES_CACHE           _IOWR('f', 42, struct fiemap)
723 #define EXT4_IOC_CHECKPOINT             _IOW('f', 43, __u32)
724
725 #define EXT4_IOC_SHUTDOWN _IOR ('X', 125, __u32)
726
727 /*
728  * Flags for going down operation
729  */
730 #define EXT4_GOING_FLAGS_DEFAULT                0x0     /* going down */
731 #define EXT4_GOING_FLAGS_LOGFLUSH               0x1     /* flush log but not data */
732 #define EXT4_GOING_FLAGS_NOLOGFLUSH             0x2     /* don't flush log nor data */
733
734 /*
735  * Flags returned by EXT4_IOC_GETSTATE
736  *
737  * We only expose to userspace a subset of the state flags in
738  * i_state_flags
739  */
740 #define EXT4_STATE_FLAG_EXT_PRECACHED   0x00000001
741 #define EXT4_STATE_FLAG_NEW             0x00000002
742 #define EXT4_STATE_FLAG_NEWENTRY        0x00000004
743 #define EXT4_STATE_FLAG_DA_ALLOC_CLOSE  0x00000008
744
745 /* flags for ioctl EXT4_IOC_CHECKPOINT */
746 #define EXT4_IOC_CHECKPOINT_FLAG_DISCARD        0x1
747 #define EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT        0x2
748 #define EXT4_IOC_CHECKPOINT_FLAG_DRY_RUN        0x4
749 #define EXT4_IOC_CHECKPOINT_FLAG_VALID          (EXT4_IOC_CHECKPOINT_FLAG_DISCARD | \
750                                                 EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT | \
751                                                 EXT4_IOC_CHECKPOINT_FLAG_DRY_RUN)
752
753 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
754 /*
755  * ioctl commands in 32 bit emulation
756  */
757 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
758 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
759 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
760 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
761 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
762 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
763 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
764 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
765 #endif
766
767 /*
768  * Returned by EXT4_IOC_GET_ES_CACHE as an additional possible flag.
769  * It indicates that the entry in extent status cache is for a hole.
770  */
771 #define EXT4_FIEMAP_EXTENT_HOLE         0x08000000
772
773 /* Max physical block we can address w/o extents */
774 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
775
776 /* Max logical block we can support */
777 #define EXT4_MAX_LOGICAL_BLOCK          0xFFFFFFFE
778
779 /*
780  * Structure of an inode on the disk
781  */
782 struct ext4_inode {
783         __le16  i_mode;         /* File mode */
784         __le16  i_uid;          /* Low 16 bits of Owner Uid */
785         __le32  i_size_lo;      /* Size in bytes */
786         __le32  i_atime;        /* Access time */
787         __le32  i_ctime;        /* Inode Change time */
788         __le32  i_mtime;        /* Modification time */
789         __le32  i_dtime;        /* Deletion Time */
790         __le16  i_gid;          /* Low 16 bits of Group Id */
791         __le16  i_links_count;  /* Links count */
792         __le32  i_blocks_lo;    /* Blocks count */
793         __le32  i_flags;        /* File flags */
794         union {
795                 struct {
796                         __le32  l_i_version;
797                 } linux1;
798                 struct {
799                         __u32  h_i_translator;
800                 } hurd1;
801                 struct {
802                         __u32  m_i_reserved1;
803                 } masix1;
804         } osd1;                         /* OS dependent 1 */
805         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
806         __le32  i_generation;   /* File version (for NFS) */
807         __le32  i_file_acl_lo;  /* File ACL */
808         __le32  i_size_high;
809         __le32  i_obso_faddr;   /* Obsoleted fragment address */
810         union {
811                 struct {
812                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
813                         __le16  l_i_file_acl_high;
814                         __le16  l_i_uid_high;   /* these 2 fields */
815                         __le16  l_i_gid_high;   /* were reserved2[0] */
816                         __le16  l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
817                         __le16  l_i_reserved;
818                 } linux2;
819                 struct {
820                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
821                         __u16   h_i_mode_high;
822                         __u16   h_i_uid_high;
823                         __u16   h_i_gid_high;
824                         __u32   h_i_author;
825                 } hurd2;
826                 struct {
827                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
828                         __le16  m_i_file_acl_high;
829                         __u32   m_i_reserved2[2];
830                 } masix2;
831         } osd2;                         /* OS dependent 2 */
832         __le16  i_extra_isize;
833         __le16  i_checksum_hi;  /* crc32c(uuid+inum+inode) BE */
834         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
835         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
836         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
837         __le32  i_crtime;       /* File Creation time */
838         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
839         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
840         __le32  i_projid;       /* Project ID */
841 };
842
843 struct move_extent {
844         __u32 reserved;         /* should be zero */
845         __u32 donor_fd;         /* donor file descriptor */
846         __u64 orig_start;       /* logical start offset in block for orig */
847         __u64 donor_start;      /* logical start offset in block for donor */
848         __u64 len;              /* block length to be moved */
849         __u64 moved_len;        /* moved block length */
850 };
851
852 #define EXT4_EPOCH_BITS 2
853 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
854 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
855
856 /*
857  * Extended fields will fit into an inode if the filesystem was formatted
858  * with large inodes (-I 256 or larger) and there are not currently any EAs
859  * consuming all of the available space. For new inodes we always reserve
860  * enough space for the kernel's known extended fields, but for inodes
861  * created with an old kernel this might not have been the case. None of
862  * the extended inode fields is critical for correct filesystem operation.
863  * This macro checks if a certain field fits in the inode. Note that
864  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
865  */
866 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
867         ((offsetof(typeof(*ext4_inode), field) +        \
868           sizeof((ext4_inode)->field))                  \
869         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
870             (einode)->i_extra_isize))                   \
871
872 /*
873  * We use an encoding that preserves the times for extra epoch "00":
874  *
875  * extra  msb of                         adjust for signed
876  * epoch  32-bit                         32-bit tv_sec to
877  * bits   time    decoded 64-bit tv_sec  64-bit tv_sec      valid time range
878  * 0 0    1    -0x80000000..-0x00000001  0x000000000 1901-12-13..1969-12-31
879  * 0 0    0    0x000000000..0x07fffffff  0x000000000 1970-01-01..2038-01-19
880  * 0 1    1    0x080000000..0x0ffffffff  0x100000000 2038-01-19..2106-02-07
881  * 0 1    0    0x100000000..0x17fffffff  0x100000000 2106-02-07..2174-02-25
882  * 1 0    1    0x180000000..0x1ffffffff  0x200000000 2174-02-25..2242-03-16
883  * 1 0    0    0x200000000..0x27fffffff  0x200000000 2242-03-16..2310-04-04
884  * 1 1    1    0x280000000..0x2ffffffff  0x300000000 2310-04-04..2378-04-22
885  * 1 1    0    0x300000000..0x37fffffff  0x300000000 2378-04-22..2446-05-10
886  *
887  * Note that previous versions of the kernel on 64-bit systems would
888  * incorrectly use extra epoch bits 1,1 for dates between 1901 and
889  * 1970.  e2fsck will correct this, assuming that it is run on the
890  * affected filesystem before 2242.
891  */
892
893 static inline __le32 ext4_encode_extra_time(struct timespec64 *time)
894 {
895         u32 extra =((time->tv_sec - (s32)time->tv_sec) >> 32) & EXT4_EPOCH_MASK;
896         return cpu_to_le32(extra | (time->tv_nsec << EXT4_EPOCH_BITS));
897 }
898
899 static inline void ext4_decode_extra_time(struct timespec64 *time,
900                                           __le32 extra)
901 {
902         if (unlikely(extra & cpu_to_le32(EXT4_EPOCH_MASK)))
903                 time->tv_sec += (u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK) << 32;
904         time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
905 }
906
907 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                           \
908 do {                                                                            \
909         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     {\
910                 (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);        \
911                 (raw_inode)->xtime ## _extra =                                  \
912                                 ext4_encode_extra_time(&(inode)->xtime);        \
913                 }                                                               \
914         else    \
915                 (raw_inode)->xtime = cpu_to_le32(clamp_t(int32_t, (inode)->xtime.tv_sec, S32_MIN, S32_MAX));    \
916 } while (0)
917
918 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
919 do {                                                                           \
920         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
921                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
922         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
923                 (raw_inode)->xtime ## _extra =                                 \
924                                 ext4_encode_extra_time(&(einode)->xtime);      \
925 } while (0)
926
927 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                           \
928 do {                                                                            \
929         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);        \
930         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) {    \
931                 ext4_decode_extra_time(&(inode)->xtime,                         \
932                                        raw_inode->xtime ## _extra);             \
933                 }                                                               \
934         else                                                                    \
935                 (inode)->xtime.tv_nsec = 0;                                     \
936 } while (0)
937
938
939 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
940 do {                                                                           \
941         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
942                 (einode)->xtime.tv_sec =                                       \
943                         (signed)le32_to_cpu((raw_inode)->xtime);               \
944         else                                                                   \
945                 (einode)->xtime.tv_sec = 0;                                    \
946         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
947                 ext4_decode_extra_time(&(einode)->xtime,                       \
948                                        raw_inode->xtime ## _extra);            \
949         else                                                                   \
950                 (einode)->xtime.tv_nsec = 0;                                   \
951 } while (0)
952
953 #define i_disk_version osd1.linux1.l_i_version
954
955 #if defined(__KERNEL__) || defined(__linux__)
956 #define i_reserved1     osd1.linux1.l_i_reserved1
957 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
958 #define i_blocks_high   osd2.linux2.l_i_blocks_high
959 #define i_uid_low       i_uid
960 #define i_gid_low       i_gid
961 #define i_uid_high      osd2.linux2.l_i_uid_high
962 #define i_gid_high      osd2.linux2.l_i_gid_high
963 #define i_checksum_lo   osd2.linux2.l_i_checksum_lo
964
965 #elif defined(__GNU__)
966
967 #define i_translator    osd1.hurd1.h_i_translator
968 #define i_uid_high      osd2.hurd2.h_i_uid_high
969 #define i_gid_high      osd2.hurd2.h_i_gid_high
970 #define i_author        osd2.hurd2.h_i_author
971
972 #elif defined(__masix__)
973
974 #define i_reserved1     osd1.masix1.m_i_reserved1
975 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
976 #define i_reserved2     osd2.masix2.m_i_reserved2
977
978 #endif /* defined(__KERNEL__) || defined(__linux__) */
979
980 #include "extents_status.h"
981 #include "fast_commit.h"
982
983 /*
984  * Lock subclasses for i_data_sem in the ext4_inode_info structure.
985  *
986  * These are needed to avoid lockdep false positives when we need to
987  * allocate blocks to the quota inode during ext4_map_blocks(), while
988  * holding i_data_sem for a normal (non-quota) inode.  Since we don't
989  * do quota tracking for the quota inode, this avoids deadlock (as
990  * well as infinite recursion, since it isn't turtles all the way
991  * down...)
992  *
993  *  I_DATA_SEM_NORMAL - Used for most inodes
994  *  I_DATA_SEM_OTHER  - Used by move_inode.c for the second normal inode
995  *                        where the second inode has larger inode number
996  *                        than the first
997  *  I_DATA_SEM_QUOTA  - Used for quota inodes only
998  */
999 enum {
1000         I_DATA_SEM_NORMAL = 0,
1001         I_DATA_SEM_OTHER,
1002         I_DATA_SEM_QUOTA,
1003 };
1004
1005
1006 /*
1007  * fourth extended file system inode data in memory
1008  */
1009 struct ext4_inode_info {
1010         __le32  i_data[15];     /* unconverted */
1011         __u32   i_dtime;
1012         ext4_fsblk_t    i_file_acl;
1013
1014         /*
1015          * i_block_group is the number of the block group which contains
1016          * this file's inode.  Constant across the lifetime of the inode,
1017          * it is used for making block allocation decisions - we try to
1018          * place a file's data blocks near its inode block, and new inodes
1019          * near to their parent directory's inode.
1020          */
1021         ext4_group_t    i_block_group;
1022         ext4_lblk_t     i_dir_start_lookup;
1023 #if (BITS_PER_LONG < 64)
1024         unsigned long   i_state_flags;          /* Dynamic state flags */
1025 #endif
1026         unsigned long   i_flags;
1027
1028         /*
1029          * Extended attributes can be read independently of the main file
1030          * data. Taking i_mutex even when reading would cause contention
1031          * between readers of EAs and writers of regular file data, so
1032          * instead we synchronize on xattr_sem when reading or changing
1033          * EAs.
1034          */
1035         struct rw_semaphore xattr_sem;
1036
1037         struct list_head i_orphan;      /* unlinked but open inodes */
1038
1039         /* Fast commit related info */
1040
1041         struct list_head i_fc_list;     /*
1042                                          * inodes that need fast commit
1043                                          * protected by sbi->s_fc_lock.
1044                                          */
1045
1046         /* Start of lblk range that needs to be committed in this fast commit */
1047         ext4_lblk_t i_fc_lblk_start;
1048
1049         /* End of lblk range that needs to be committed in this fast commit */
1050         ext4_lblk_t i_fc_lblk_len;
1051
1052         /* Number of ongoing updates on this inode */
1053         atomic_t  i_fc_updates;
1054
1055         /* Fast commit wait queue for this inode */
1056         wait_queue_head_t i_fc_wait;
1057
1058         /* Protect concurrent accesses on i_fc_lblk_start, i_fc_lblk_len */
1059         struct mutex i_fc_lock;
1060
1061         /*
1062          * i_disksize keeps track of what the inode size is ON DISK, not
1063          * in memory.  During truncate, i_size is set to the new size by
1064          * the VFS prior to calling ext4_truncate(), but the filesystem won't
1065          * set i_disksize to 0 until the truncate is actually under way.
1066          *
1067          * The intent is that i_disksize always represents the blocks which
1068          * are used by this file.  This allows recovery to restart truncate
1069          * on orphans if we crash during truncate.  We actually write i_disksize
1070          * into the on-disk inode when writing inodes out, instead of i_size.
1071          *
1072          * The only time when i_disksize and i_size may be different is when
1073          * a truncate is in progress.  The only things which change i_disksize
1074          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
1075          */
1076         loff_t  i_disksize;
1077
1078         /*
1079          * i_data_sem is for serialising ext4_truncate() against
1080          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
1081          * data tree are chopped off during truncate. We can't do that in
1082          * ext4 because whenever we perform intermediate commits during
1083          * truncate, the inode and all the metadata blocks *must* be in a
1084          * consistent state which allows truncation of the orphans to restart
1085          * during recovery.  Hence we must fix the get_block-vs-truncate race
1086          * by other means, so we have i_data_sem.
1087          */
1088         struct rw_semaphore i_data_sem;
1089         /*
1090          * i_mmap_sem is for serializing page faults with truncate / punch hole
1091          * operations. We have to make sure that new page cannot be faulted in
1092          * a section of the inode that is being punched. We cannot easily use
1093          * i_data_sem for this since we need protection for the whole punch
1094          * operation and i_data_sem ranks below transaction start so we have
1095          * to occasionally drop it.
1096          */
1097         struct rw_semaphore i_mmap_sem;
1098         struct inode vfs_inode;
1099         struct jbd2_inode *jinode;
1100
1101         spinlock_t i_raw_lock;  /* protects updates to the raw inode */
1102
1103         /*
1104          * File creation time. Its function is same as that of
1105          * struct timespec64 i_{a,c,m}time in the generic inode.
1106          */
1107         struct timespec64 i_crtime;
1108
1109         /* mballoc */
1110         atomic_t i_prealloc_active;
1111         struct list_head i_prealloc_list;
1112         spinlock_t i_prealloc_lock;
1113
1114         /* extents status tree */
1115         struct ext4_es_tree i_es_tree;
1116         rwlock_t i_es_lock;
1117         struct list_head i_es_list;
1118         unsigned int i_es_all_nr;       /* protected by i_es_lock */
1119         unsigned int i_es_shk_nr;       /* protected by i_es_lock */
1120         ext4_lblk_t i_es_shrink_lblk;   /* Offset where we start searching for
1121                                            extents to shrink. Protected by
1122                                            i_es_lock  */
1123
1124         /* ialloc */
1125         ext4_group_t    i_last_alloc_group;
1126
1127         /* allocation reservation info for delalloc */
1128         /* In case of bigalloc, this refer to clusters rather than blocks */
1129         unsigned int i_reserved_data_blocks;
1130
1131         /* pending cluster reservations for bigalloc file systems */
1132         struct ext4_pending_tree i_pending_tree;
1133
1134         /* on-disk additional length */
1135         __u16 i_extra_isize;
1136
1137         /* Indicate the inline data space. */
1138         u16 i_inline_off;
1139         u16 i_inline_size;
1140
1141 #ifdef CONFIG_QUOTA
1142         /* quota space reservation, managed internally by quota code */
1143         qsize_t i_reserved_quota;
1144 #endif
1145
1146         /* Lock protecting lists below */
1147         spinlock_t i_completed_io_lock;
1148         /*
1149          * Completed IOs that need unwritten extents handling and have
1150          * transaction reserved
1151          */
1152         struct list_head i_rsv_conversion_list;
1153         struct work_struct i_rsv_conversion_work;
1154         atomic_t i_unwritten; /* Nr. of inflight conversions pending */
1155
1156         spinlock_t i_block_reservation_lock;
1157
1158         /*
1159          * Transactions that contain inode's metadata needed to complete
1160          * fsync and fdatasync, respectively.
1161          */
1162         tid_t i_sync_tid;
1163         tid_t i_datasync_tid;
1164
1165 #ifdef CONFIG_QUOTA
1166         struct dquot *i_dquot[MAXQUOTAS];
1167 #endif
1168
1169         /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
1170         __u32 i_csum_seed;
1171
1172         kprojid_t i_projid;
1173 };
1174
1175 /*
1176  * File system states
1177  */
1178 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
1179 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
1180 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
1181 #define EXT4_FC_REPLAY                  0x0020  /* Fast commit replay ongoing */
1182
1183 /*
1184  * Misc. filesystem flags
1185  */
1186 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
1187 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
1188 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
1189
1190 /*
1191  * Mount flags set via mount options or defaults
1192  */
1193 #define EXT4_MOUNT_NO_MBCACHE           0x00001 /* Do not use mbcache */
1194 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
1195 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
1196 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
1197 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
1198 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
1199 #define EXT4_MOUNT_ERRORS_MASK          0x00070
1200 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
1201 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
1202 #ifdef CONFIG_FS_DAX
1203 #define EXT4_MOUNT_DAX_ALWAYS           0x00200 /* Direct Access */
1204 #else
1205 #define EXT4_MOUNT_DAX_ALWAYS           0
1206 #endif
1207 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
1208 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
1209 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
1210 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
1211 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
1212 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
1213 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
1214 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
1215 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
1216 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
1217 #define EXT4_MOUNT_QUOTA                0x40000 /* Some quota option set */
1218 #define EXT4_MOUNT_USRQUOTA             0x80000 /* "old" user quota,
1219                                                  * enable enforcement for hidden
1220                                                  * quota files */
1221 #define EXT4_MOUNT_GRPQUOTA             0x100000 /* "old" group quota, enable
1222                                                   * enforcement for hidden quota
1223                                                   * files */
1224 #define EXT4_MOUNT_PRJQUOTA             0x200000 /* Enable project quota
1225                                                   * enforcement */
1226 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
1227 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
1228 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
1229 #define EXT4_MOUNT_WARN_ON_ERROR        0x2000000 /* Trigger WARN_ON on error */
1230 #define EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS 0x4000000
1231 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
1232 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
1233 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
1234 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
1235 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
1236
1237 /*
1238  * Mount flags set either automatically (could not be set by mount option)
1239  * based on per file system feature or property or in special cases such as
1240  * distinguishing between explicit mount option definition and default.
1241  */
1242 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
1243                                                       specified delalloc */
1244 #define EXT4_MOUNT2_STD_GROUP_SIZE      0x00000002 /* We have standard group
1245                                                       size of blocksize * 8
1246                                                       blocks */
1247 #define EXT4_MOUNT2_HURD_COMPAT         0x00000004 /* Support HURD-castrated
1248                                                       file systems */
1249 #define EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM   0x00000008 /* User explicitly
1250                                                 specified journal checksum */
1251
1252 #define EXT4_MOUNT2_JOURNAL_FAST_COMMIT 0x00000010 /* Journal fast commit */
1253 #define EXT4_MOUNT2_DAX_NEVER           0x00000020 /* Do not allow Direct Access */
1254 #define EXT4_MOUNT2_DAX_INODE           0x00000040 /* For printing options only */
1255 #define EXT4_MOUNT2_MB_OPTIMIZE_SCAN    0x00000080 /* Optimize group
1256                                                     * scanning in mballoc
1257                                                     */
1258
1259 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
1260                                                 ~EXT4_MOUNT_##opt
1261 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
1262                                                 EXT4_MOUNT_##opt
1263 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
1264                                          EXT4_MOUNT_##opt)
1265
1266 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
1267                                                 ~EXT4_MOUNT2_##opt
1268 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
1269                                                 EXT4_MOUNT2_##opt
1270 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
1271                                          EXT4_MOUNT2_##opt)
1272
1273 #define ext4_test_and_set_bit           __test_and_set_bit_le
1274 #define ext4_set_bit                    __set_bit_le
1275 #define ext4_set_bit_atomic             ext2_set_bit_atomic
1276 #define ext4_test_and_clear_bit         __test_and_clear_bit_le
1277 #define ext4_clear_bit                  __clear_bit_le
1278 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
1279 #define ext4_test_bit                   test_bit_le
1280 #define ext4_find_next_zero_bit         find_next_zero_bit_le
1281 #define ext4_find_next_bit              find_next_bit_le
1282
1283 extern void ext4_set_bits(void *bm, int cur, int len);
1284
1285 /*
1286  * Maximal mount counts between two filesystem checks
1287  */
1288 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
1289 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
1290
1291 /*
1292  * Behaviour when detecting errors
1293  */
1294 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
1295 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
1296 #define EXT4_ERRORS_PANIC               3       /* Panic */
1297 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
1298
1299 /* Metadata checksum algorithm codes */
1300 #define EXT4_CRC32C_CHKSUM              1
1301
1302 /*
1303  * Structure of the super block
1304  */
1305 struct ext4_super_block {
1306 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1307         __le32  s_blocks_count_lo;      /* Blocks count */
1308         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1309         __le32  s_free_blocks_count_lo; /* Free blocks count */
1310 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1311         __le32  s_first_data_block;     /* First Data Block */
1312         __le32  s_log_block_size;       /* Block size */
1313         __le32  s_log_cluster_size;     /* Allocation cluster size */
1314 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1315         __le32  s_clusters_per_group;   /* # Clusters per group */
1316         __le32  s_inodes_per_group;     /* # Inodes per group */
1317         __le32  s_mtime;                /* Mount time */
1318 /*30*/  __le32  s_wtime;                /* Write time */
1319         __le16  s_mnt_count;            /* Mount count */
1320         __le16  s_max_mnt_count;        /* Maximal mount count */
1321         __le16  s_magic;                /* Magic signature */
1322         __le16  s_state;                /* File system state */
1323         __le16  s_errors;               /* Behaviour when detecting errors */
1324         __le16  s_minor_rev_level;      /* minor revision level */
1325 /*40*/  __le32  s_lastcheck;            /* time of last check */
1326         __le32  s_checkinterval;        /* max. time between checks */
1327         __le32  s_creator_os;           /* OS */
1328         __le32  s_rev_level;            /* Revision level */
1329 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1330         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1331         /*
1332          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1333          *
1334          * Note: the difference between the compatible feature set and
1335          * the incompatible feature set is that if there is a bit set
1336          * in the incompatible feature set that the kernel doesn't
1337          * know about, it should refuse to mount the filesystem.
1338          *
1339          * e2fsck's requirements are more strict; if it doesn't know
1340          * about a feature in either the compatible or incompatible
1341          * feature set, it must abort and not try to meddle with
1342          * things it doesn't understand...
1343          */
1344         __le32  s_first_ino;            /* First non-reserved inode */
1345         __le16  s_inode_size;           /* size of inode structure */
1346         __le16  s_block_group_nr;       /* block group # of this superblock */
1347         __le32  s_feature_compat;       /* compatible feature set */
1348 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1349         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1350 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1351 /*78*/  char    s_volume_name[16];      /* volume name */
1352 /*88*/  char    s_last_mounted[64] __nonstring; /* directory where last mounted */
1353 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1354         /*
1355          * Performance hints.  Directory preallocation should only
1356          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1357          */
1358         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1359         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1360         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1361         /*
1362          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1363          */
1364 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1365 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1366         __le32  s_journal_dev;          /* device number of journal file */
1367         __le32  s_last_orphan;          /* start of list of inodes to delete */
1368         __le32  s_hash_seed[4];         /* HTREE hash seed */
1369         __u8    s_def_hash_version;     /* Default hash version to use */
1370         __u8    s_jnl_backup_type;
1371         __le16  s_desc_size;            /* size of group descriptor */
1372 /*100*/ __le32  s_default_mount_opts;
1373         __le32  s_first_meta_bg;        /* First metablock block group */
1374         __le32  s_mkfs_time;            /* When the filesystem was created */
1375         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1376         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1377 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1378         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1379         __le32  s_free_blocks_count_hi; /* Free blocks count */
1380         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1381         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1382         __le32  s_flags;                /* Miscellaneous flags */
1383         __le16  s_raid_stride;          /* RAID stride */
1384         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1385         __le64  s_mmp_block;            /* Block for multi-mount protection */
1386         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1387         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1388         __u8    s_checksum_type;        /* metadata checksum algorithm used */
1389         __u8    s_encryption_level;     /* versioning level for encryption */
1390         __u8    s_reserved_pad;         /* Padding to next 32bits */
1391         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1392         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1393         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1394         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1395                                               snapshot's future use */
1396         __le32  s_snapshot_list;        /* inode number of the head of the
1397                                            on-disk snapshot list */
1398 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1399         __le32  s_error_count;          /* number of fs errors */
1400         __le32  s_first_error_time;     /* first time an error happened */
1401         __le32  s_first_error_ino;      /* inode involved in first error */
1402         __le64  s_first_error_block;    /* block involved of first error */
1403         __u8    s_first_error_func[32] __nonstring;     /* function where the error happened */
1404         __le32  s_first_error_line;     /* line number where error happened */
1405         __le32  s_last_error_time;      /* most recent time of an error */
1406         __le32  s_last_error_ino;       /* inode involved in last error */
1407         __le32  s_last_error_line;      /* line number where error happened */
1408         __le64  s_last_error_block;     /* block involved of last error */
1409         __u8    s_last_error_func[32] __nonstring;      /* function where the error happened */
1410 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1411         __u8    s_mount_opts[64];
1412         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1413         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1414         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1415         __le32  s_backup_bgs[2];        /* groups with sparse_super2 SBs */
1416         __u8    s_encrypt_algos[4];     /* Encryption algorithms in use  */
1417         __u8    s_encrypt_pw_salt[16];  /* Salt used for string2key algorithm */
1418         __le32  s_lpf_ino;              /* Location of the lost+found inode */
1419         __le32  s_prj_quota_inum;       /* inode for tracking project quota */
1420         __le32  s_checksum_seed;        /* crc32c(uuid) if csum_seed set */
1421         __u8    s_wtime_hi;
1422         __u8    s_mtime_hi;
1423         __u8    s_mkfs_time_hi;
1424         __u8    s_lastcheck_hi;
1425         __u8    s_first_error_time_hi;
1426         __u8    s_last_error_time_hi;
1427         __u8    s_first_error_errcode;
1428         __u8    s_last_error_errcode;
1429         __le16  s_encoding;             /* Filename charset encoding */
1430         __le16  s_encoding_flags;       /* Filename charset encoding flags */
1431         __le32  s_reserved[95];         /* Padding to the end of the block */
1432         __le32  s_checksum;             /* crc32c(superblock) */
1433 };
1434
1435 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1436
1437 #ifdef __KERNEL__
1438
1439 #ifdef CONFIG_FS_ENCRYPTION
1440 #define DUMMY_ENCRYPTION_ENABLED(sbi) ((sbi)->s_dummy_enc_policy.policy != NULL)
1441 #else
1442 #define DUMMY_ENCRYPTION_ENABLED(sbi) (0)
1443 #endif
1444
1445 /* Number of quota types we support */
1446 #define EXT4_MAXQUOTAS 3
1447
1448 #define EXT4_ENC_UTF8_12_1      1
1449
1450 /* Types of ext4 journal triggers */
1451 enum ext4_journal_trigger_type {
1452         EXT4_JTR_NONE   /* This must be the last entry for indexing to work! */
1453 };
1454
1455 #define EXT4_JOURNAL_TRIGGER_COUNT EXT4_JTR_NONE
1456
1457 struct ext4_journal_trigger {
1458         struct jbd2_buffer_trigger_type tr_triggers;
1459         struct super_block *sb;
1460 };
1461
1462 static inline struct ext4_journal_trigger *EXT4_TRIGGER(
1463                                 struct jbd2_buffer_trigger_type *trigger)
1464 {
1465         return container_of(trigger, struct ext4_journal_trigger, tr_triggers);
1466 }
1467
1468 /*
1469  * fourth extended-fs super-block data in memory
1470  */
1471 struct ext4_sb_info {
1472         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1473         unsigned long s_inodes_per_block;/* Number of inodes per block */
1474         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1475         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1476         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1477         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1478         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1479         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1480         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1481         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1482         unsigned long s_overhead;  /* # of fs overhead clusters */
1483         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1484         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1485         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1486         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1487         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1488         struct buffer_head * __rcu *s_group_desc;
1489         unsigned int s_mount_opt;
1490         unsigned int s_mount_opt2;
1491         unsigned long s_mount_flags;
1492         unsigned int s_def_mount_opt;
1493         ext4_fsblk_t s_sb_block;
1494         atomic64_t s_resv_clusters;
1495         kuid_t s_resuid;
1496         kgid_t s_resgid;
1497         unsigned short s_mount_state;
1498         unsigned short s_pad;
1499         int s_addr_per_block_bits;
1500         int s_desc_per_block_bits;
1501         int s_inode_size;
1502         int s_first_ino;
1503         unsigned int s_inode_readahead_blks;
1504         unsigned int s_inode_goal;
1505         u32 s_hash_seed[4];
1506         int s_def_hash_version;
1507         int s_hash_unsigned;    /* 3 if hash should be unsigned, 0 if not */
1508         struct percpu_counter s_freeclusters_counter;
1509         struct percpu_counter s_freeinodes_counter;
1510         struct percpu_counter s_dirs_counter;
1511         struct percpu_counter s_dirtyclusters_counter;
1512         struct percpu_counter s_sra_exceeded_retry_limit;
1513         struct blockgroup_lock *s_blockgroup_lock;
1514         struct proc_dir_entry *s_proc;
1515         struct kobject s_kobj;
1516         struct completion s_kobj_unregister;
1517         struct super_block *s_sb;
1518         struct buffer_head *s_mmp_bh;
1519
1520         /* Journaling */
1521         struct journal_s *s_journal;
1522         struct list_head s_orphan;
1523         struct mutex s_orphan_lock;
1524         unsigned long s_ext4_flags;             /* Ext4 superblock flags */
1525         unsigned long s_commit_interval;
1526         u32 s_max_batch_time;
1527         u32 s_min_batch_time;
1528         struct block_device *s_journal_bdev;
1529 #ifdef CONFIG_QUOTA
1530         /* Names of quota files with journalled quota */
1531         char __rcu *s_qf_names[EXT4_MAXQUOTAS];
1532         int s_jquota_fmt;                       /* Format of quota to use */
1533 #endif
1534         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1535         struct ext4_system_blocks __rcu *s_system_blks;
1536
1537 #ifdef EXTENTS_STATS
1538         /* ext4 extents stats */
1539         unsigned long s_ext_min;
1540         unsigned long s_ext_max;
1541         unsigned long s_depth_max;
1542         spinlock_t s_ext_stats_lock;
1543         unsigned long s_ext_blocks;
1544         unsigned long s_ext_extents;
1545 #endif
1546
1547         /* for buddy allocator */
1548         struct ext4_group_info ** __rcu *s_group_info;
1549         struct inode *s_buddy_cache;
1550         spinlock_t s_md_lock;
1551         unsigned short *s_mb_offsets;
1552         unsigned int *s_mb_maxs;
1553         unsigned int s_group_info_size;
1554         unsigned int s_mb_free_pending;
1555         struct list_head s_freed_data_list;     /* List of blocks to be freed
1556                                                    after commit completed */
1557         struct list_head s_discard_list;
1558         struct work_struct s_discard_work;
1559         atomic_t s_retry_alloc_pending;
1560         struct rb_root s_mb_avg_fragment_size_root;
1561         rwlock_t s_mb_rb_lock;
1562         struct list_head *s_mb_largest_free_orders;
1563         rwlock_t *s_mb_largest_free_orders_locks;
1564
1565         /* tunables */
1566         unsigned long s_stripe;
1567         unsigned int s_mb_max_linear_groups;
1568         unsigned int s_mb_stream_request;
1569         unsigned int s_mb_max_to_scan;
1570         unsigned int s_mb_min_to_scan;
1571         unsigned int s_mb_stats;
1572         unsigned int s_mb_order2_reqs;
1573         unsigned int s_mb_group_prealloc;
1574         unsigned int s_mb_max_inode_prealloc;
1575         unsigned int s_max_dir_size_kb;
1576         /* where last allocation was done - for stream allocation */
1577         unsigned long s_mb_last_group;
1578         unsigned long s_mb_last_start;
1579         unsigned int s_mb_prefetch;
1580         unsigned int s_mb_prefetch_limit;
1581
1582         /* stats for buddy allocator */
1583         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1584         atomic_t s_bal_success; /* we found long enough chunks */
1585         atomic_t s_bal_allocated;       /* in blocks */
1586         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1587         atomic_t s_bal_groups_scanned;  /* number of groups scanned */
1588         atomic_t s_bal_goals;   /* goal hits */
1589         atomic_t s_bal_breaks;  /* too long searches */
1590         atomic_t s_bal_2orders; /* 2^order hits */
1591         atomic_t s_bal_cr0_bad_suggestions;
1592         atomic_t s_bal_cr1_bad_suggestions;
1593         atomic64_t s_bal_cX_groups_considered[4];
1594         atomic64_t s_bal_cX_hits[4];
1595         atomic64_t s_bal_cX_failed[4];          /* cX loop didn't find blocks */
1596         atomic_t s_mb_buddies_generated;        /* number of buddies generated */
1597         atomic64_t s_mb_generation_time;
1598         atomic_t s_mb_lost_chunks;
1599         atomic_t s_mb_preallocated;
1600         atomic_t s_mb_discarded;
1601         atomic_t s_lock_busy;
1602
1603         /* locality groups */
1604         struct ext4_locality_group __percpu *s_locality_groups;
1605
1606         /* for write statistics */
1607         unsigned long s_sectors_written_start;
1608         u64 s_kbytes_written;
1609
1610         /* the size of zero-out chunk */
1611         unsigned int s_extent_max_zeroout_kb;
1612
1613         unsigned int s_log_groups_per_flex;
1614         struct flex_groups * __rcu *s_flex_groups;
1615         ext4_group_t s_flex_groups_allocated;
1616
1617         /* workqueue for reserved extent conversions (buffered io) */
1618         struct workqueue_struct *rsv_conversion_wq;
1619
1620         /* timer for periodic error stats printing */
1621         struct timer_list s_err_report;
1622
1623         /* Lazy inode table initialization info */
1624         struct ext4_li_request *s_li_request;
1625         /* Wait multiplier for lazy initialization thread */
1626         unsigned int s_li_wait_mult;
1627
1628         /* Kernel thread for multiple mount protection */
1629         struct task_struct *s_mmp_tsk;
1630
1631         /* record the last minlen when FITRIM is called. */
1632         atomic_t s_last_trim_minblks;
1633
1634         /* Reference to checksum algorithm driver via cryptoapi */
1635         struct crypto_shash *s_chksum_driver;
1636
1637         /* Precomputed FS UUID checksum for seeding other checksums */
1638         __u32 s_csum_seed;
1639
1640         /* Reclaim extents from extent status tree */
1641         struct shrinker s_es_shrinker;
1642         struct list_head s_es_list;     /* List of inodes with reclaimable extents */
1643         long s_es_nr_inode;
1644         struct ext4_es_stats s_es_stats;
1645         struct mb_cache *s_ea_block_cache;
1646         struct mb_cache *s_ea_inode_cache;
1647         spinlock_t s_es_lock ____cacheline_aligned_in_smp;
1648
1649         /* Journal triggers for checksum computation */
1650         struct ext4_journal_trigger s_journal_triggers[EXT4_JOURNAL_TRIGGER_COUNT];
1651
1652         /* Ratelimit ext4 messages. */
1653         struct ratelimit_state s_err_ratelimit_state;
1654         struct ratelimit_state s_warning_ratelimit_state;
1655         struct ratelimit_state s_msg_ratelimit_state;
1656         atomic_t s_warning_count;
1657         atomic_t s_msg_count;
1658
1659         /* Encryption policy for '-o test_dummy_encryption' */
1660         struct fscrypt_dummy_policy s_dummy_enc_policy;
1661
1662         /*
1663          * Barrier between writepages ops and changing any inode's JOURNAL_DATA
1664          * or EXTENTS flag.
1665          */
1666         struct percpu_rw_semaphore s_writepages_rwsem;
1667         struct dax_device *s_daxdev;
1668 #ifdef CONFIG_EXT4_DEBUG
1669         unsigned long s_simulate_fail;
1670 #endif
1671         /* Record the errseq of the backing block device */
1672         errseq_t s_bdev_wb_err;
1673         spinlock_t s_bdev_wb_lock;
1674
1675         /* Information about errors that happened during this mount */
1676         spinlock_t s_error_lock;
1677         int s_add_error_count;
1678         int s_first_error_code;
1679         __u32 s_first_error_line;
1680         __u32 s_first_error_ino;
1681         __u64 s_first_error_block;
1682         const char *s_first_error_func;
1683         time64_t s_first_error_time;
1684         int s_last_error_code;
1685         __u32 s_last_error_line;
1686         __u32 s_last_error_ino;
1687         __u64 s_last_error_block;
1688         const char *s_last_error_func;
1689         time64_t s_last_error_time;
1690         /*
1691          * If we are in a context where we cannot update error information in
1692          * the on-disk superblock, we queue this work to do it.
1693          */
1694         struct work_struct s_error_work;
1695
1696         /* Ext4 fast commit stuff */
1697         atomic_t s_fc_subtid;
1698         atomic_t s_fc_ineligible_updates;
1699         /*
1700          * After commit starts, the main queue gets locked, and the further
1701          * updates get added in the staging queue.
1702          */
1703 #define FC_Q_MAIN       0
1704 #define FC_Q_STAGING    1
1705         struct list_head s_fc_q[2];     /* Inodes staged for fast commit
1706                                          * that have data changes in them.
1707                                          */
1708         struct list_head s_fc_dentry_q[2];      /* directory entry updates */
1709         unsigned int s_fc_bytes;
1710         /*
1711          * Main fast commit lock. This lock protects accesses to the
1712          * following fields:
1713          * ei->i_fc_list, s_fc_dentry_q, s_fc_q, s_fc_bytes, s_fc_bh.
1714          */
1715         spinlock_t s_fc_lock;
1716         struct buffer_head *s_fc_bh;
1717         struct ext4_fc_stats s_fc_stats;
1718         u64 s_fc_avg_commit_time;
1719 #ifdef CONFIG_EXT4_DEBUG
1720         int s_fc_debug_max_replay;
1721 #endif
1722         struct ext4_fc_replay_state s_fc_replay_state;
1723 };
1724
1725 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1726 {
1727         return sb->s_fs_info;
1728 }
1729 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1730 {
1731         return container_of(inode, struct ext4_inode_info, vfs_inode);
1732 }
1733
1734 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1735 {
1736         return ino == EXT4_ROOT_INO ||
1737                 (ino >= EXT4_FIRST_INO(sb) &&
1738                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1739 }
1740
1741 /*
1742  * Returns: sbi->field[index]
1743  * Used to access an array element from the following sbi fields which require
1744  * rcu protection to avoid dereferencing an invalid pointer due to reassignment
1745  * - s_group_desc
1746  * - s_group_info
1747  * - s_flex_group
1748  */
1749 #define sbi_array_rcu_deref(sbi, field, index)                             \
1750 ({                                                                         \
1751         typeof(*((sbi)->field)) _v;                                        \
1752         rcu_read_lock();                                                   \
1753         _v = ((typeof(_v)*)rcu_dereference((sbi)->field))[index];          \
1754         rcu_read_unlock();                                                 \
1755         _v;                                                                \
1756 })
1757
1758 /*
1759  * run-time mount flags
1760  */
1761 enum {
1762         EXT4_MF_MNTDIR_SAMPLED,
1763         EXT4_MF_FS_ABORTED,     /* Fatal error detected */
1764         EXT4_MF_FC_INELIGIBLE,  /* Fast commit ineligible */
1765         EXT4_MF_FC_COMMITTING   /* File system underoing a fast
1766                                  * commit.
1767                                  */
1768 };
1769
1770 static inline void ext4_set_mount_flag(struct super_block *sb, int bit)
1771 {
1772         set_bit(bit, &EXT4_SB(sb)->s_mount_flags);
1773 }
1774
1775 static inline void ext4_clear_mount_flag(struct super_block *sb, int bit)
1776 {
1777         clear_bit(bit, &EXT4_SB(sb)->s_mount_flags);
1778 }
1779
1780 static inline int ext4_test_mount_flag(struct super_block *sb, int bit)
1781 {
1782         return test_bit(bit, &EXT4_SB(sb)->s_mount_flags);
1783 }
1784
1785
1786 /*
1787  * Simulate_fail codes
1788  */
1789 #define EXT4_SIM_BBITMAP_EIO    1
1790 #define EXT4_SIM_BBITMAP_CRC    2
1791 #define EXT4_SIM_IBITMAP_EIO    3
1792 #define EXT4_SIM_IBITMAP_CRC    4
1793 #define EXT4_SIM_INODE_EIO      5
1794 #define EXT4_SIM_INODE_CRC      6
1795 #define EXT4_SIM_DIRBLOCK_EIO   7
1796 #define EXT4_SIM_DIRBLOCK_CRC   8
1797
1798 static inline bool ext4_simulate_fail(struct super_block *sb,
1799                                      unsigned long code)
1800 {
1801 #ifdef CONFIG_EXT4_DEBUG
1802         struct ext4_sb_info *sbi = EXT4_SB(sb);
1803
1804         if (unlikely(sbi->s_simulate_fail == code)) {
1805                 sbi->s_simulate_fail = 0;
1806                 return true;
1807         }
1808 #endif
1809         return false;
1810 }
1811
1812 static inline void ext4_simulate_fail_bh(struct super_block *sb,
1813                                          struct buffer_head *bh,
1814                                          unsigned long code)
1815 {
1816         if (!IS_ERR(bh) && ext4_simulate_fail(sb, code))
1817                 clear_buffer_uptodate(bh);
1818 }
1819
1820 /*
1821  * Error number codes for s_{first,last}_error_errno
1822  *
1823  * Linux errno numbers are architecture specific, so we need to translate
1824  * them into something which is architecture independent.   We don't define
1825  * codes for all errno's; just the ones which are most likely to be the cause
1826  * of an ext4_error() call.
1827  */
1828 #define EXT4_ERR_UNKNOWN         1
1829 #define EXT4_ERR_EIO             2
1830 #define EXT4_ERR_ENOMEM          3
1831 #define EXT4_ERR_EFSBADCRC       4
1832 #define EXT4_ERR_EFSCORRUPTED    5
1833 #define EXT4_ERR_ENOSPC          6
1834 #define EXT4_ERR_ENOKEY          7
1835 #define EXT4_ERR_EROFS           8
1836 #define EXT4_ERR_EFBIG           9
1837 #define EXT4_ERR_EEXIST         10
1838 #define EXT4_ERR_ERANGE         11
1839 #define EXT4_ERR_EOVERFLOW      12
1840 #define EXT4_ERR_EBUSY          13
1841 #define EXT4_ERR_ENOTDIR        14
1842 #define EXT4_ERR_ENOTEMPTY      15
1843 #define EXT4_ERR_ESHUTDOWN      16
1844 #define EXT4_ERR_EFAULT         17
1845
1846 /*
1847  * Inode dynamic state flags
1848  */
1849 enum {
1850         EXT4_STATE_JDATA,               /* journaled data exists */
1851         EXT4_STATE_NEW,                 /* inode is newly created */
1852         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1853         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1854         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1855         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1856         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1857         EXT4_STATE_MAY_INLINE_DATA,     /* may have in-inode data */
1858         EXT4_STATE_EXT_PRECACHED,       /* extents have been precached */
1859         EXT4_STATE_LUSTRE_EA_INODE,     /* Lustre-style ea_inode */
1860         EXT4_STATE_VERITY_IN_PROGRESS,  /* building fs-verity Merkle tree */
1861         EXT4_STATE_FC_COMMITTING,       /* Fast commit ongoing */
1862 };
1863
1864 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1865 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1866 {                                                                       \
1867         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1868 }                                                                       \
1869 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1870 {                                                                       \
1871         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1872 }                                                                       \
1873 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1874 {                                                                       \
1875         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1876 }
1877
1878 /* Add these declarations here only so that these functions can be
1879  * found by name.  Otherwise, they are very hard to locate. */
1880 static inline int ext4_test_inode_flag(struct inode *inode, int bit);
1881 static inline void ext4_set_inode_flag(struct inode *inode, int bit);
1882 static inline void ext4_clear_inode_flag(struct inode *inode, int bit);
1883 EXT4_INODE_BIT_FNS(flag, flags, 0)
1884
1885 /* Add these declarations here only so that these functions can be
1886  * found by name.  Otherwise, they are very hard to locate. */
1887 static inline int ext4_test_inode_state(struct inode *inode, int bit);
1888 static inline void ext4_set_inode_state(struct inode *inode, int bit);
1889 static inline void ext4_clear_inode_state(struct inode *inode, int bit);
1890 #if (BITS_PER_LONG < 64)
1891 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1892
1893 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1894 {
1895         (ei)->i_state_flags = 0;
1896 }
1897 #else
1898 EXT4_INODE_BIT_FNS(state, flags, 32)
1899
1900 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1901 {
1902         /* We depend on the fact that callers will set i_flags */
1903 }
1904 #endif
1905 #else
1906 /* Assume that user mode programs are passing in an ext4fs superblock, not
1907  * a kernel struct super_block.  This will allow us to call the feature-test
1908  * macros from user land. */
1909 #define EXT4_SB(sb)     (sb)
1910 #endif
1911
1912 static inline bool ext4_verity_in_progress(struct inode *inode)
1913 {
1914         return IS_ENABLED(CONFIG_FS_VERITY) &&
1915                ext4_test_inode_state(inode, EXT4_STATE_VERITY_IN_PROGRESS);
1916 }
1917
1918 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1919
1920 /*
1921  * Codes for operating systems
1922  */
1923 #define EXT4_OS_LINUX           0
1924 #define EXT4_OS_HURD            1
1925 #define EXT4_OS_MASIX           2
1926 #define EXT4_OS_FREEBSD         3
1927 #define EXT4_OS_LITES           4
1928
1929 /*
1930  * Revision levels
1931  */
1932 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1933 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1934
1935 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1936
1937 #define EXT4_GOOD_OLD_INODE_SIZE 128
1938
1939 #define EXT4_EXTRA_TIMESTAMP_MAX        (((s64)1 << 34) - 1  + S32_MIN)
1940 #define EXT4_NON_EXTRA_TIMESTAMP_MAX    S32_MAX
1941 #define EXT4_TIMESTAMP_MIN              S32_MIN
1942
1943 /*
1944  * Feature set definitions
1945  */
1946
1947 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1948 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1949 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1950 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1951 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1952 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1953 #define EXT4_FEATURE_COMPAT_SPARSE_SUPER2       0x0200
1954 /*
1955  * The reason why "FAST_COMMIT" is a compat feature is that, FS becomes
1956  * incompatible only if fast commit blocks are present in the FS. Since we
1957  * clear the journal (and thus the fast commit blocks), we don't mark FS as
1958  * incompatible. We also have a JBD2 incompat feature, which gets set when
1959  * there are fast commit blocks present in the journal.
1960  */
1961 #define EXT4_FEATURE_COMPAT_FAST_COMMIT         0x0400
1962 #define EXT4_FEATURE_COMPAT_STABLE_INODES       0x0800
1963
1964 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1965 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1966 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1967 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1968 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1969 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1970 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1971 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1972 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1973 /*
1974  * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM).  When
1975  * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
1976  * all other data structures' checksums.  However, the METADATA_CSUM and
1977  * GDT_CSUM bits are mutually exclusive.
1978  */
1979 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM    0x0400
1980 #define EXT4_FEATURE_RO_COMPAT_READONLY         0x1000
1981 #define EXT4_FEATURE_RO_COMPAT_PROJECT          0x2000
1982 #define EXT4_FEATURE_RO_COMPAT_VERITY           0x8000
1983
1984 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1985 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1986 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1987 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1988 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1989 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1990 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1991 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1992 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1993 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1994 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1995 #define EXT4_FEATURE_INCOMPAT_CSUM_SEED         0x2000
1996 #define EXT4_FEATURE_INCOMPAT_LARGEDIR          0x4000 /* >2GB or 3-lvl htree */
1997 #define EXT4_FEATURE_INCOMPAT_INLINE_DATA       0x8000 /* data in inode */
1998 #define EXT4_FEATURE_INCOMPAT_ENCRYPT           0x10000
1999 #define EXT4_FEATURE_INCOMPAT_CASEFOLD          0x20000
2000
2001 extern void ext4_update_dynamic_rev(struct super_block *sb);
2002
2003 #define EXT4_FEATURE_COMPAT_FUNCS(name, flagname) \
2004 static inline bool ext4_has_feature_##name(struct super_block *sb) \
2005 { \
2006         return ((EXT4_SB(sb)->s_es->s_feature_compat & \
2007                 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname)) != 0); \
2008 } \
2009 static inline void ext4_set_feature_##name(struct super_block *sb) \
2010 { \
2011         ext4_update_dynamic_rev(sb); \
2012         EXT4_SB(sb)->s_es->s_feature_compat |= \
2013                 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
2014 } \
2015 static inline void ext4_clear_feature_##name(struct super_block *sb) \
2016 { \
2017         EXT4_SB(sb)->s_es->s_feature_compat &= \
2018                 ~cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
2019 }
2020
2021 #define EXT4_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
2022 static inline bool ext4_has_feature_##name(struct super_block *sb) \
2023 { \
2024         return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
2025                 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname)) != 0); \
2026 } \
2027 static inline void ext4_set_feature_##name(struct super_block *sb) \
2028 { \
2029         ext4_update_dynamic_rev(sb); \
2030         EXT4_SB(sb)->s_es->s_feature_ro_compat |= \
2031                 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
2032 } \
2033 static inline void ext4_clear_feature_##name(struct super_block *sb) \
2034 { \
2035         EXT4_SB(sb)->s_es->s_feature_ro_compat &= \
2036                 ~cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
2037 }
2038
2039 #define EXT4_FEATURE_INCOMPAT_FUNCS(name, flagname) \
2040 static inline bool ext4_has_feature_##name(struct super_block *sb) \
2041 { \
2042         return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
2043                 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname)) != 0); \
2044 } \
2045 static inline void ext4_set_feature_##name(struct super_block *sb) \
2046 { \
2047         ext4_update_dynamic_rev(sb); \
2048         EXT4_SB(sb)->s_es->s_feature_incompat |= \
2049                 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
2050 } \
2051 static inline void ext4_clear_feature_##name(struct super_block *sb) \
2052 { \
2053         EXT4_SB(sb)->s_es->s_feature_incompat &= \
2054                 ~cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
2055 }
2056
2057 EXT4_FEATURE_COMPAT_FUNCS(dir_prealloc,         DIR_PREALLOC)
2058 EXT4_FEATURE_COMPAT_FUNCS(imagic_inodes,        IMAGIC_INODES)
2059 EXT4_FEATURE_COMPAT_FUNCS(journal,              HAS_JOURNAL)
2060 EXT4_FEATURE_COMPAT_FUNCS(xattr,                EXT_ATTR)
2061 EXT4_FEATURE_COMPAT_FUNCS(resize_inode,         RESIZE_INODE)
2062 EXT4_FEATURE_COMPAT_FUNCS(dir_index,            DIR_INDEX)
2063 EXT4_FEATURE_COMPAT_FUNCS(sparse_super2,        SPARSE_SUPER2)
2064 EXT4_FEATURE_COMPAT_FUNCS(fast_commit,          FAST_COMMIT)
2065 EXT4_FEATURE_COMPAT_FUNCS(stable_inodes,        STABLE_INODES)
2066
2067 EXT4_FEATURE_RO_COMPAT_FUNCS(sparse_super,      SPARSE_SUPER)
2068 EXT4_FEATURE_RO_COMPAT_FUNCS(large_file,        LARGE_FILE)
2069 EXT4_FEATURE_RO_COMPAT_FUNCS(btree_dir,         BTREE_DIR)
2070 EXT4_FEATURE_RO_COMPAT_FUNCS(huge_file,         HUGE_FILE)
2071 EXT4_FEATURE_RO_COMPAT_FUNCS(gdt_csum,          GDT_CSUM)
2072 EXT4_FEATURE_RO_COMPAT_FUNCS(dir_nlink,         DIR_NLINK)
2073 EXT4_FEATURE_RO_COMPAT_FUNCS(extra_isize,       EXTRA_ISIZE)
2074 EXT4_FEATURE_RO_COMPAT_FUNCS(quota,             QUOTA)
2075 EXT4_FEATURE_RO_COMPAT_FUNCS(bigalloc,          BIGALLOC)
2076 EXT4_FEATURE_RO_COMPAT_FUNCS(metadata_csum,     METADATA_CSUM)
2077 EXT4_FEATURE_RO_COMPAT_FUNCS(readonly,          READONLY)
2078 EXT4_FEATURE_RO_COMPAT_FUNCS(project,           PROJECT)
2079 EXT4_FEATURE_RO_COMPAT_FUNCS(verity,            VERITY)
2080
2081 EXT4_FEATURE_INCOMPAT_FUNCS(compression,        COMPRESSION)
2082 EXT4_FEATURE_INCOMPAT_FUNCS(filetype,           FILETYPE)
2083 EXT4_FEATURE_INCOMPAT_FUNCS(journal_needs_recovery,     RECOVER)
2084 EXT4_FEATURE_INCOMPAT_FUNCS(journal_dev,        JOURNAL_DEV)
2085 EXT4_FEATURE_INCOMPAT_FUNCS(meta_bg,            META_BG)
2086 EXT4_FEATURE_INCOMPAT_FUNCS(extents,            EXTENTS)
2087 EXT4_FEATURE_INCOMPAT_FUNCS(64bit,              64BIT)
2088 EXT4_FEATURE_INCOMPAT_FUNCS(mmp,                MMP)
2089 EXT4_FEATURE_INCOMPAT_FUNCS(flex_bg,            FLEX_BG)
2090 EXT4_FEATURE_INCOMPAT_FUNCS(ea_inode,           EA_INODE)
2091 EXT4_FEATURE_INCOMPAT_FUNCS(dirdata,            DIRDATA)
2092 EXT4_FEATURE_INCOMPAT_FUNCS(csum_seed,          CSUM_SEED)
2093 EXT4_FEATURE_INCOMPAT_FUNCS(largedir,           LARGEDIR)
2094 EXT4_FEATURE_INCOMPAT_FUNCS(inline_data,        INLINE_DATA)
2095 EXT4_FEATURE_INCOMPAT_FUNCS(encrypt,            ENCRYPT)
2096 EXT4_FEATURE_INCOMPAT_FUNCS(casefold,           CASEFOLD)
2097
2098 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
2099 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
2100                                          EXT4_FEATURE_INCOMPAT_META_BG)
2101 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
2102                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
2103                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
2104
2105 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
2106 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
2107                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
2108                                          EXT4_FEATURE_INCOMPAT_META_BG)
2109 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
2110                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
2111                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
2112
2113 #define EXT4_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
2114 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
2115                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
2116                                          EXT4_FEATURE_INCOMPAT_META_BG| \
2117                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
2118                                          EXT4_FEATURE_INCOMPAT_64BIT| \
2119                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
2120                                          EXT4_FEATURE_INCOMPAT_EA_INODE| \
2121                                          EXT4_FEATURE_INCOMPAT_MMP | \
2122                                          EXT4_FEATURE_INCOMPAT_INLINE_DATA | \
2123                                          EXT4_FEATURE_INCOMPAT_ENCRYPT | \
2124                                          EXT4_FEATURE_INCOMPAT_CASEFOLD | \
2125                                          EXT4_FEATURE_INCOMPAT_CSUM_SEED | \
2126                                          EXT4_FEATURE_INCOMPAT_LARGEDIR)
2127 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
2128                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
2129                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
2130                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
2131                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
2132                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
2133                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
2134                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
2135                                          EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
2136                                          EXT4_FEATURE_RO_COMPAT_QUOTA |\
2137                                          EXT4_FEATURE_RO_COMPAT_PROJECT |\
2138                                          EXT4_FEATURE_RO_COMPAT_VERITY)
2139
2140 #define EXTN_FEATURE_FUNCS(ver) \
2141 static inline bool ext4_has_unknown_ext##ver##_compat_features(struct super_block *sb) \
2142 { \
2143         return ((EXT4_SB(sb)->s_es->s_feature_compat & \
2144                 cpu_to_le32(~EXT##ver##_FEATURE_COMPAT_SUPP)) != 0); \
2145 } \
2146 static inline bool ext4_has_unknown_ext##ver##_ro_compat_features(struct super_block *sb) \
2147 { \
2148         return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
2149                 cpu_to_le32(~EXT##ver##_FEATURE_RO_COMPAT_SUPP)) != 0); \
2150 } \
2151 static inline bool ext4_has_unknown_ext##ver##_incompat_features(struct super_block *sb) \
2152 { \
2153         return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
2154                 cpu_to_le32(~EXT##ver##_FEATURE_INCOMPAT_SUPP)) != 0); \
2155 }
2156
2157 EXTN_FEATURE_FUNCS(2)
2158 EXTN_FEATURE_FUNCS(3)
2159 EXTN_FEATURE_FUNCS(4)
2160
2161 static inline bool ext4_has_compat_features(struct super_block *sb)
2162 {
2163         return (EXT4_SB(sb)->s_es->s_feature_compat != 0);
2164 }
2165 static inline bool ext4_has_ro_compat_features(struct super_block *sb)
2166 {
2167         return (EXT4_SB(sb)->s_es->s_feature_ro_compat != 0);
2168 }
2169 static inline bool ext4_has_incompat_features(struct super_block *sb)
2170 {
2171         return (EXT4_SB(sb)->s_es->s_feature_incompat != 0);
2172 }
2173
2174 /*
2175  * Superblock flags
2176  */
2177 #define EXT4_FLAGS_RESIZING     0
2178 #define EXT4_FLAGS_SHUTDOWN     1
2179 #define EXT4_FLAGS_BDEV_IS_DAX  2
2180
2181 static inline int ext4_forced_shutdown(struct ext4_sb_info *sbi)
2182 {
2183         return test_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
2184 }
2185
2186
2187 /*
2188  * Default values for user and/or group using reserved blocks
2189  */
2190 #define EXT4_DEF_RESUID         0
2191 #define EXT4_DEF_RESGID         0
2192
2193 /*
2194  * Default project ID
2195  */
2196 #define EXT4_DEF_PROJID         0
2197
2198 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
2199
2200 /*
2201  * Default mount options
2202  */
2203 #define EXT4_DEFM_DEBUG         0x0001
2204 #define EXT4_DEFM_BSDGROUPS     0x0002
2205 #define EXT4_DEFM_XATTR_USER    0x0004
2206 #define EXT4_DEFM_ACL           0x0008
2207 #define EXT4_DEFM_UID16         0x0010
2208 #define EXT4_DEFM_JMODE         0x0060
2209 #define EXT4_DEFM_JMODE_DATA    0x0020
2210 #define EXT4_DEFM_JMODE_ORDERED 0x0040
2211 #define EXT4_DEFM_JMODE_WBACK   0x0060
2212 #define EXT4_DEFM_NOBARRIER     0x0100
2213 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
2214 #define EXT4_DEFM_DISCARD       0x0400
2215 #define EXT4_DEFM_NODELALLOC    0x0800
2216
2217 /*
2218  * Default journal batch times
2219  */
2220 #define EXT4_DEF_MIN_BATCH_TIME 0
2221 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
2222
2223 /*
2224  * Minimum number of groups in a flexgroup before we separate out
2225  * directories into the first block group of a flexgroup
2226  */
2227 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
2228
2229 /*
2230  * Structure of a directory entry
2231  */
2232 #define EXT4_NAME_LEN 255
2233
2234 struct ext4_dir_entry {
2235         __le32  inode;                  /* Inode number */
2236         __le16  rec_len;                /* Directory entry length */
2237         __le16  name_len;               /* Name length */
2238         char    name[EXT4_NAME_LEN];    /* File name */
2239 };
2240
2241
2242 /*
2243  * Encrypted Casefolded entries require saving the hash on disk. This structure
2244  * followed ext4_dir_entry_2's name[name_len] at the next 4 byte aligned
2245  * boundary.
2246  */
2247 struct ext4_dir_entry_hash {
2248         __le32 hash;
2249         __le32 minor_hash;
2250 };
2251
2252 /*
2253  * The new version of the directory entry.  Since EXT4 structures are
2254  * stored in intel byte order, and the name_len field could never be
2255  * bigger than 255 chars, it's safe to reclaim the extra byte for the
2256  * file_type field.
2257  */
2258 struct ext4_dir_entry_2 {
2259         __le32  inode;                  /* Inode number */
2260         __le16  rec_len;                /* Directory entry length */
2261         __u8    name_len;               /* Name length */
2262         __u8    file_type;              /* See file type macros EXT4_FT_* below */
2263         char    name[EXT4_NAME_LEN];    /* File name */
2264 };
2265
2266 /*
2267  * Access the hashes at the end of ext4_dir_entry_2
2268  */
2269 #define EXT4_DIRENT_HASHES(entry) \
2270         ((struct ext4_dir_entry_hash *) \
2271                 (((void *)(entry)) + \
2272                 ((8 + (entry)->name_len + EXT4_DIR_ROUND) & ~EXT4_DIR_ROUND)))
2273 #define EXT4_DIRENT_HASH(entry) le32_to_cpu(EXT4_DIRENT_HASHES(de)->hash)
2274 #define EXT4_DIRENT_MINOR_HASH(entry) \
2275                 le32_to_cpu(EXT4_DIRENT_HASHES(de)->minor_hash)
2276
2277 static inline bool ext4_hash_in_dirent(const struct inode *inode)
2278 {
2279         return IS_CASEFOLDED(inode) && IS_ENCRYPTED(inode);
2280 }
2281
2282 /*
2283  * This is a bogus directory entry at the end of each leaf block that
2284  * records checksums.
2285  */
2286 struct ext4_dir_entry_tail {
2287         __le32  det_reserved_zero1;     /* Pretend to be unused */
2288         __le16  det_rec_len;            /* 12 */
2289         __u8    det_reserved_zero2;     /* Zero name length */
2290         __u8    det_reserved_ft;        /* 0xDE, fake file type */
2291         __le32  det_checksum;           /* crc32c(uuid+inum+dirblock) */
2292 };
2293
2294 #define EXT4_DIRENT_TAIL(block, blocksize) \
2295         ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
2296                                         ((blocksize) - \
2297                                          sizeof(struct ext4_dir_entry_tail))))
2298
2299 /*
2300  * Ext4 directory file types.  Only the low 3 bits are used.  The
2301  * other bits are reserved for now.
2302  */
2303 #define EXT4_FT_UNKNOWN         0
2304 #define EXT4_FT_REG_FILE        1
2305 #define EXT4_FT_DIR             2
2306 #define EXT4_FT_CHRDEV          3
2307 #define EXT4_FT_BLKDEV          4
2308 #define EXT4_FT_FIFO            5
2309 #define EXT4_FT_SOCK            6
2310 #define EXT4_FT_SYMLINK         7
2311
2312 #define EXT4_FT_MAX             8
2313
2314 #define EXT4_FT_DIR_CSUM        0xDE
2315
2316 /*
2317  * EXT4_DIR_PAD defines the directory entries boundaries
2318  *
2319  * NOTE: It must be a multiple of 4
2320  */
2321 #define EXT4_DIR_PAD                    4
2322 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
2323 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
2324
2325 /*
2326  * The rec_len is dependent on the type of directory. Directories that are
2327  * casefolded and encrypted need to store the hash as well, so we add room for
2328  * ext4_extended_dir_entry_2. For all entries related to '.' or '..' you should
2329  * pass NULL for dir, as those entries do not use the extra fields.
2330  */
2331 static inline unsigned int ext4_dir_rec_len(__u8 name_len,
2332                                                 const struct inode *dir)
2333 {
2334         int rec_len = (name_len + 8 + EXT4_DIR_ROUND);
2335
2336         if (dir && ext4_hash_in_dirent(dir))
2337                 rec_len += sizeof(struct ext4_dir_entry_hash);
2338         return (rec_len & ~EXT4_DIR_ROUND);
2339 }
2340
2341 /*
2342  * If we ever get support for fs block sizes > page_size, we'll need
2343  * to remove the #if statements in the next two functions...
2344  */
2345 static inline unsigned int
2346 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
2347 {
2348         unsigned len = le16_to_cpu(dlen);
2349
2350 #if (PAGE_SIZE >= 65536)
2351         if (len == EXT4_MAX_REC_LEN || len == 0)
2352                 return blocksize;
2353         return (len & 65532) | ((len & 3) << 16);
2354 #else
2355         return len;
2356 #endif
2357 }
2358
2359 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
2360 {
2361         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
2362                 BUG();
2363 #if (PAGE_SIZE >= 65536)
2364         if (len < 65536)
2365                 return cpu_to_le16(len);
2366         if (len == blocksize) {
2367                 if (blocksize == 65536)
2368                         return cpu_to_le16(EXT4_MAX_REC_LEN);
2369                 else
2370                         return cpu_to_le16(0);
2371         }
2372         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
2373 #else
2374         return cpu_to_le16(len);
2375 #endif
2376 }
2377
2378 /*
2379  * Hash Tree Directory indexing
2380  * (c) Daniel Phillips, 2001
2381  */
2382
2383 #define is_dx(dir) (ext4_has_feature_dir_index((dir)->i_sb) && \
2384                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
2385 #define EXT4_DIR_LINK_MAX(dir) unlikely((dir)->i_nlink >= EXT4_LINK_MAX && \
2386                     !(ext4_has_feature_dir_nlink((dir)->i_sb) && is_dx(dir)))
2387 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
2388
2389 /* Legal values for the dx_root hash_version field: */
2390
2391 #define DX_HASH_LEGACY                  0
2392 #define DX_HASH_HALF_MD4                1
2393 #define DX_HASH_TEA                     2
2394 #define DX_HASH_LEGACY_UNSIGNED         3
2395 #define DX_HASH_HALF_MD4_UNSIGNED       4
2396 #define DX_HASH_TEA_UNSIGNED            5
2397 #define DX_HASH_SIPHASH                 6
2398
2399 static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
2400                               const void *address, unsigned int length)
2401 {
2402         struct {
2403                 struct shash_desc shash;
2404                 char ctx[4];
2405         } desc;
2406
2407         BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
2408
2409         desc.shash.tfm = sbi->s_chksum_driver;
2410         *(u32 *)desc.ctx = crc;
2411
2412         BUG_ON(crypto_shash_update(&desc.shash, address, length));
2413
2414         return *(u32 *)desc.ctx;
2415 }
2416
2417 #ifdef __KERNEL__
2418
2419 /* hash info structure used by the directory hash */
2420 struct dx_hash_info
2421 {
2422         u32             hash;
2423         u32             minor_hash;
2424         int             hash_version;
2425         u32             *seed;
2426 };
2427
2428
2429 /* 32 and 64 bit signed EOF for dx directories */
2430 #define EXT4_HTREE_EOF_32BIT   ((1UL  << (32 - 1)) - 1)
2431 #define EXT4_HTREE_EOF_64BIT   ((1ULL << (64 - 1)) - 1)
2432
2433
2434 /*
2435  * Control parameters used by ext4_htree_next_block
2436  */
2437 #define HASH_NB_ALWAYS          1
2438
2439 struct ext4_filename {
2440         const struct qstr *usr_fname;
2441         struct fscrypt_str disk_name;
2442         struct dx_hash_info hinfo;
2443 #ifdef CONFIG_FS_ENCRYPTION
2444         struct fscrypt_str crypto_buf;
2445 #endif
2446 #ifdef CONFIG_UNICODE
2447         struct fscrypt_str cf_name;
2448 #endif
2449 };
2450
2451 #define fname_name(p) ((p)->disk_name.name)
2452 #define fname_usr_name(p) ((p)->usr_fname->name)
2453 #define fname_len(p)  ((p)->disk_name.len)
2454
2455 /*
2456  * Describe an inode's exact location on disk and in memory
2457  */
2458 struct ext4_iloc
2459 {
2460         struct buffer_head *bh;
2461         unsigned long offset;
2462         ext4_group_t block_group;
2463 };
2464
2465 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
2466 {
2467         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
2468 }
2469
2470 static inline bool ext4_is_quota_file(struct inode *inode)
2471 {
2472         return IS_NOQUOTA(inode) &&
2473                !(EXT4_I(inode)->i_flags & EXT4_EA_INODE_FL);
2474 }
2475
2476 /*
2477  * This structure is stuffed into the struct file's private_data field
2478  * for directories.  It is where we put information so that we can do
2479  * readdir operations in hash tree order.
2480  */
2481 struct dir_private_info {
2482         struct rb_root  root;
2483         struct rb_node  *curr_node;
2484         struct fname    *extra_fname;
2485         loff_t          last_pos;
2486         __u32           curr_hash;
2487         __u32           curr_minor_hash;
2488         __u32           next_hash;
2489 };
2490
2491 /* calculate the first block number of the group */
2492 static inline ext4_fsblk_t
2493 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
2494 {
2495         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
2496                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
2497 }
2498
2499 /*
2500  * Special error return code only used by dx_probe() and its callers.
2501  */
2502 #define ERR_BAD_DX_DIR  (-(MAX_ERRNO - 1))
2503
2504 /* htree levels for ext4 */
2505 #define EXT4_HTREE_LEVEL_COMPAT 2
2506 #define EXT4_HTREE_LEVEL        3
2507
2508 static inline int ext4_dir_htree_level(struct super_block *sb)
2509 {
2510         return ext4_has_feature_largedir(sb) ?
2511                 EXT4_HTREE_LEVEL : EXT4_HTREE_LEVEL_COMPAT;
2512 }
2513
2514 /*
2515  * Timeout and state flag for lazy initialization inode thread.
2516  */
2517 #define EXT4_DEF_LI_WAIT_MULT                   10
2518 #define EXT4_DEF_LI_MAX_START_DELAY             5
2519 #define EXT4_LAZYINIT_QUIT                      0x0001
2520 #define EXT4_LAZYINIT_RUNNING                   0x0002
2521
2522 /*
2523  * Lazy inode table initialization info
2524  */
2525 struct ext4_lazy_init {
2526         unsigned long           li_state;
2527         struct list_head        li_request_list;
2528         struct mutex            li_list_mtx;
2529 };
2530
2531 enum ext4_li_mode {
2532         EXT4_LI_MODE_PREFETCH_BBITMAP,
2533         EXT4_LI_MODE_ITABLE,
2534 };
2535
2536 struct ext4_li_request {
2537         struct super_block      *lr_super;
2538         enum ext4_li_mode       lr_mode;
2539         ext4_group_t            lr_first_not_zeroed;
2540         ext4_group_t            lr_next_group;
2541         struct list_head        lr_request;
2542         unsigned long           lr_next_sched;
2543         unsigned long           lr_timeout;
2544 };
2545
2546 struct ext4_features {
2547         struct kobject f_kobj;
2548         struct completion f_kobj_unregister;
2549 };
2550
2551 /*
2552  * This structure will be used for multiple mount protection. It will be
2553  * written into the block number saved in the s_mmp_block field in the
2554  * superblock. Programs that check MMP should assume that if
2555  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
2556  * to use the filesystem, regardless of how old the timestamp is.
2557  */
2558 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
2559 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
2560 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
2561 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
2562
2563 struct mmp_struct {
2564         __le32  mmp_magic;              /* Magic number for MMP */
2565         __le32  mmp_seq;                /* Sequence no. updated periodically */
2566
2567         /*
2568          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
2569          * purposes and do not affect the correctness of the algorithm
2570          */
2571         __le64  mmp_time;               /* Time last updated */
2572         char    mmp_nodename[64];       /* Node which last updated MMP block */
2573         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
2574
2575         /*
2576          * mmp_check_interval is used to verify if the MMP block has been
2577          * updated on the block device. The value is updated based on the
2578          * maximum time to write the MMP block during an update cycle.
2579          */
2580         __le16  mmp_check_interval;
2581
2582         __le16  mmp_pad1;
2583         __le32  mmp_pad2[226];
2584         __le32  mmp_checksum;           /* crc32c(uuid+mmp_block) */
2585 };
2586
2587 /* arguments passed to the mmp thread */
2588 struct mmpd_data {
2589         struct buffer_head *bh; /* bh from initial read_mmp_block() */
2590         struct super_block *sb;  /* super block of the fs */
2591 };
2592
2593 /*
2594  * Check interval multiplier
2595  * The MMP block is written every update interval and initially checked every
2596  * update interval x the multiplier (the value is then adapted based on the
2597  * write latency). The reason is that writes can be delayed under load and we
2598  * don't want readers to incorrectly assume that the filesystem is no longer
2599  * in use.
2600  */
2601 #define EXT4_MMP_CHECK_MULT             2UL
2602
2603 /*
2604  * Minimum interval for MMP checking in seconds.
2605  */
2606 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
2607
2608 /*
2609  * Maximum interval for MMP checking in seconds.
2610  */
2611 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
2612
2613 /*
2614  * Function prototypes
2615  */
2616
2617 /*
2618  * Ok, these declarations are also in <linux/kernel.h> but none of the
2619  * ext4 source programs needs to include it so they are duplicated here.
2620  */
2621 # define NORET_TYPE     /**/
2622 # define ATTRIB_NORET   __attribute__((noreturn))
2623 # define NORET_AND      noreturn,
2624
2625 /* bitmap.c */
2626 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
2627 void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
2628                                 struct ext4_group_desc *gdp,
2629                                 struct buffer_head *bh, int sz);
2630 int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2631                                   struct ext4_group_desc *gdp,
2632                                   struct buffer_head *bh, int sz);
2633 void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
2634                                 struct ext4_group_desc *gdp,
2635                                 struct buffer_head *bh);
2636 int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2637                                   struct ext4_group_desc *gdp,
2638                                   struct buffer_head *bh);
2639
2640 /* balloc.c */
2641 extern void ext4_get_group_no_and_offset(struct super_block *sb,
2642                                          ext4_fsblk_t blocknr,
2643                                          ext4_group_t *blockgrpp,
2644                                          ext4_grpblk_t *offsetp);
2645 extern ext4_group_t ext4_get_group_number(struct super_block *sb,
2646                                           ext4_fsblk_t block);
2647
2648 extern unsigned int ext4_block_group(struct super_block *sb,
2649                         ext4_fsblk_t blocknr);
2650 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
2651                         ext4_fsblk_t blocknr);
2652 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
2653 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
2654                         ext4_group_t group);
2655 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
2656                                          ext4_fsblk_t goal,
2657                                          unsigned int flags,
2658                                          unsigned long *count,
2659                                          int *errp);
2660 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
2661                                     s64 nclusters, unsigned int flags);
2662 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
2663 extern void ext4_check_blocks_bitmap(struct super_block *);
2664 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
2665                                                     ext4_group_t block_group,
2666                                                     struct buffer_head ** bh);
2667 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
2668
2669 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
2670                                                 ext4_group_t block_group,
2671                                                 bool ignore_locked);
2672 extern int ext4_wait_block_bitmap(struct super_block *sb,
2673                                   ext4_group_t block_group,
2674                                   struct buffer_head *bh);
2675 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
2676                                                   ext4_group_t block_group);
2677 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
2678                                               ext4_group_t block_group,
2679                                               struct ext4_group_desc *gdp);
2680 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
2681
2682 #ifdef CONFIG_UNICODE
2683 extern int ext4_fname_setup_ci_filename(struct inode *dir,
2684                                          const struct qstr *iname,
2685                                          struct ext4_filename *fname);
2686 #endif
2687
2688 #ifdef CONFIG_FS_ENCRYPTION
2689 static inline void ext4_fname_from_fscrypt_name(struct ext4_filename *dst,
2690                                                 const struct fscrypt_name *src)
2691 {
2692         memset(dst, 0, sizeof(*dst));
2693
2694         dst->usr_fname = src->usr_fname;
2695         dst->disk_name = src->disk_name;
2696         dst->hinfo.hash = src->hash;
2697         dst->hinfo.minor_hash = src->minor_hash;
2698         dst->crypto_buf = src->crypto_buf;
2699 }
2700
2701 static inline int ext4_fname_setup_filename(struct inode *dir,
2702                                             const struct qstr *iname,
2703                                             int lookup,
2704                                             struct ext4_filename *fname)
2705 {
2706         struct fscrypt_name name;
2707         int err;
2708
2709         err = fscrypt_setup_filename(dir, iname, lookup, &name);
2710         if (err)
2711                 return err;
2712
2713         ext4_fname_from_fscrypt_name(fname, &name);
2714
2715 #ifdef CONFIG_UNICODE
2716         err = ext4_fname_setup_ci_filename(dir, iname, fname);
2717 #endif
2718         return err;
2719 }
2720
2721 static inline int ext4_fname_prepare_lookup(struct inode *dir,
2722                                             struct dentry *dentry,
2723                                             struct ext4_filename *fname)
2724 {
2725         struct fscrypt_name name;
2726         int err;
2727
2728         err = fscrypt_prepare_lookup(dir, dentry, &name);
2729         if (err)
2730                 return err;
2731
2732         ext4_fname_from_fscrypt_name(fname, &name);
2733
2734 #ifdef CONFIG_UNICODE
2735         err = ext4_fname_setup_ci_filename(dir, &dentry->d_name, fname);
2736 #endif
2737         return err;
2738 }
2739
2740 static inline void ext4_fname_free_filename(struct ext4_filename *fname)
2741 {
2742         struct fscrypt_name name;
2743
2744         name.crypto_buf = fname->crypto_buf;
2745         fscrypt_free_filename(&name);
2746
2747         fname->crypto_buf.name = NULL;
2748         fname->usr_fname = NULL;
2749         fname->disk_name.name = NULL;
2750
2751 #ifdef CONFIG_UNICODE
2752         kfree(fname->cf_name.name);
2753         fname->cf_name.name = NULL;
2754 #endif
2755 }
2756 #else /* !CONFIG_FS_ENCRYPTION */
2757 static inline int ext4_fname_setup_filename(struct inode *dir,
2758                                             const struct qstr *iname,
2759                                             int lookup,
2760                                             struct ext4_filename *fname)
2761 {
2762         int err = 0;
2763         fname->usr_fname = iname;
2764         fname->disk_name.name = (unsigned char *) iname->name;
2765         fname->disk_name.len = iname->len;
2766
2767 #ifdef CONFIG_UNICODE
2768         err = ext4_fname_setup_ci_filename(dir, iname, fname);
2769 #endif
2770
2771         return err;
2772 }
2773
2774 static inline int ext4_fname_prepare_lookup(struct inode *dir,
2775                                             struct dentry *dentry,
2776                                             struct ext4_filename *fname)
2777 {
2778         return ext4_fname_setup_filename(dir, &dentry->d_name, 1, fname);
2779 }
2780
2781 static inline void ext4_fname_free_filename(struct ext4_filename *fname)
2782 {
2783 #ifdef CONFIG_UNICODE
2784         kfree(fname->cf_name.name);
2785         fname->cf_name.name = NULL;
2786 #endif
2787 }
2788 #endif /* !CONFIG_FS_ENCRYPTION */
2789
2790 /* dir.c */
2791 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
2792                                   struct file *,
2793                                   struct ext4_dir_entry_2 *,
2794                                   struct buffer_head *, char *, int,
2795                                   unsigned int);
2796 #define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset) \
2797         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
2798                                 (de), (bh), (buf), (size), (offset)))
2799 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
2800                                 __u32 minor_hash,
2801                                 struct ext4_dir_entry_2 *dirent,
2802                                 struct fscrypt_str *ent_name);
2803 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
2804 extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
2805                              struct buffer_head *bh,
2806                              void *buf, int buf_size,
2807                              struct ext4_filename *fname,
2808                              struct ext4_dir_entry_2 **dest_de);
2809 void ext4_insert_dentry(struct inode *dir, struct inode *inode,
2810                         struct ext4_dir_entry_2 *de,
2811                         int buf_size,
2812                         struct ext4_filename *fname);
2813 static inline void ext4_update_dx_flag(struct inode *inode)
2814 {
2815         if (!ext4_has_feature_dir_index(inode->i_sb) &&
2816             ext4_test_inode_flag(inode, EXT4_INODE_INDEX)) {
2817                 /* ext4_iget() should have caught this... */
2818                 WARN_ON_ONCE(ext4_has_feature_metadata_csum(inode->i_sb));
2819                 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
2820         }
2821 }
2822 static const unsigned char ext4_filetype_table[] = {
2823         DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
2824 };
2825
2826 static inline  unsigned char get_dtype(struct super_block *sb, int filetype)
2827 {
2828         if (!ext4_has_feature_filetype(sb) || filetype >= EXT4_FT_MAX)
2829                 return DT_UNKNOWN;
2830
2831         return ext4_filetype_table[filetype];
2832 }
2833 extern int ext4_check_all_de(struct inode *dir, struct buffer_head *bh,
2834                              void *buf, int buf_size);
2835
2836 /* fsync.c */
2837 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2838
2839 /* hash.c */
2840 extern int ext4fs_dirhash(const struct inode *dir, const char *name, int len,
2841                           struct dx_hash_info *hinfo);
2842
2843 /* ialloc.c */
2844 extern int ext4_mark_inode_used(struct super_block *sb, int ino);
2845 extern struct inode *__ext4_new_inode(struct user_namespace *, handle_t *,
2846                                       struct inode *, umode_t,
2847                                       const struct qstr *qstr, __u32 goal,
2848                                       uid_t *owner, __u32 i_flags,
2849                                       int handle_type, unsigned int line_no,
2850                                       int nblocks);
2851
2852 #define ext4_new_inode(handle, dir, mode, qstr, goal, owner, i_flags)          \
2853         __ext4_new_inode(&init_user_ns, (handle), (dir), (mode), (qstr),       \
2854                          (goal), (owner), i_flags, 0, 0, 0)
2855 #define ext4_new_inode_start_handle(mnt_userns, dir, mode, qstr, goal, owner, \
2856                                     type, nblocks)                  \
2857         __ext4_new_inode((mnt_userns), NULL, (dir), (mode), (qstr), (goal), (owner), \
2858                          0, (type), __LINE__, (nblocks))
2859
2860
2861 extern void ext4_free_inode(handle_t *, struct inode *);
2862 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2863 extern unsigned long ext4_count_free_inodes(struct super_block *);
2864 extern unsigned long ext4_count_dirs(struct super_block *);
2865 extern void ext4_check_inodes_bitmap(struct super_block *);
2866 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2867 extern int ext4_init_inode_table(struct super_block *sb,
2868                                  ext4_group_t group, int barrier);
2869 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2870
2871 /* fast_commit.c */
2872 int ext4_fc_info_show(struct seq_file *seq, void *v);
2873 void ext4_fc_init(struct super_block *sb, journal_t *journal);
2874 void ext4_fc_init_inode(struct inode *inode);
2875 void ext4_fc_track_range(handle_t *handle, struct inode *inode, ext4_lblk_t start,
2876                          ext4_lblk_t end);
2877 void __ext4_fc_track_unlink(handle_t *handle, struct inode *inode,
2878         struct dentry *dentry);
2879 void __ext4_fc_track_link(handle_t *handle, struct inode *inode,
2880         struct dentry *dentry);
2881 void ext4_fc_track_unlink(handle_t *handle, struct dentry *dentry);
2882 void ext4_fc_track_link(handle_t *handle, struct dentry *dentry);
2883 void __ext4_fc_track_create(handle_t *handle, struct inode *inode,
2884                             struct dentry *dentry);
2885 void ext4_fc_track_create(handle_t *handle, struct dentry *dentry);
2886 void ext4_fc_track_inode(handle_t *handle, struct inode *inode);
2887 void ext4_fc_mark_ineligible(struct super_block *sb, int reason);
2888 void ext4_fc_start_ineligible(struct super_block *sb, int reason);
2889 void ext4_fc_stop_ineligible(struct super_block *sb);
2890 void ext4_fc_start_update(struct inode *inode);
2891 void ext4_fc_stop_update(struct inode *inode);
2892 void ext4_fc_del(struct inode *inode);
2893 bool ext4_fc_replay_check_excluded(struct super_block *sb, ext4_fsblk_t block);
2894 void ext4_fc_replay_cleanup(struct super_block *sb);
2895 int ext4_fc_commit(journal_t *journal, tid_t commit_tid);
2896 int __init ext4_fc_init_dentry_cache(void);
2897
2898 /* mballoc.c */
2899 extern const struct seq_operations ext4_mb_seq_groups_ops;
2900 extern const struct seq_operations ext4_mb_seq_structs_summary_ops;
2901 extern long ext4_mb_stats;
2902 extern long ext4_mb_max_to_scan;
2903 extern int ext4_seq_mb_stats_show(struct seq_file *seq, void *offset);
2904 extern int ext4_mb_init(struct super_block *);
2905 extern int ext4_mb_release(struct super_block *);
2906 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2907                                 struct ext4_allocation_request *, int *);
2908 extern int ext4_mb_reserve_blocks(struct super_block *, int);
2909 extern void ext4_discard_preallocations(struct inode *, unsigned int);
2910 extern int __init ext4_init_mballoc(void);
2911 extern void ext4_exit_mballoc(void);
2912 extern ext4_group_t ext4_mb_prefetch(struct super_block *sb,
2913                                      ext4_group_t group,
2914                                      unsigned int nr, int *cnt);
2915 extern void ext4_mb_prefetch_fini(struct super_block *sb, ext4_group_t group,
2916                                   unsigned int nr);
2917
2918 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2919                              struct buffer_head *bh, ext4_fsblk_t block,
2920                              unsigned long count, int flags);
2921 extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2922                                    ext4_group_t ngroups);
2923 extern int ext4_mb_add_groupinfo(struct super_block *sb,
2924                 ext4_group_t i, struct ext4_group_desc *desc);
2925 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2926                                 ext4_fsblk_t block, unsigned long count);
2927 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2928 extern void ext4_process_freed_data(struct super_block *sb, tid_t commit_tid);
2929 extern void ext4_mb_mark_bb(struct super_block *sb, ext4_fsblk_t block,
2930                        int len, int state);
2931
2932 /* inode.c */
2933 void ext4_inode_csum_set(struct inode *inode, struct ext4_inode *raw,
2934                          struct ext4_inode_info *ei);
2935 int ext4_inode_is_fast_symlink(struct inode *inode);
2936 struct buffer_head *ext4_getblk(handle_t *, struct inode *, ext4_lblk_t, int);
2937 struct buffer_head *ext4_bread(handle_t *, struct inode *, ext4_lblk_t, int);
2938 int ext4_bread_batch(struct inode *inode, ext4_lblk_t block, int bh_count,
2939                      bool wait, struct buffer_head **bhs);
2940 int ext4_get_block_unwritten(struct inode *inode, sector_t iblock,
2941                              struct buffer_head *bh_result, int create);
2942 int ext4_get_block(struct inode *inode, sector_t iblock,
2943                    struct buffer_head *bh_result, int create);
2944 int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2945                            struct buffer_head *bh, int create);
2946 int ext4_walk_page_buffers(handle_t *handle,
2947                            struct inode *inode,
2948                            struct buffer_head *head,
2949                            unsigned from,
2950                            unsigned to,
2951                            int *partial,
2952                            int (*fn)(handle_t *handle, struct inode *inode,
2953                                      struct buffer_head *bh));
2954 int do_journal_get_write_access(handle_t *handle, struct inode *inode,
2955                                 struct buffer_head *bh);
2956 #define FALL_BACK_TO_NONDELALLOC 1
2957 #define CONVERT_INLINE_DATA      2
2958
2959 typedef enum {
2960         EXT4_IGET_NORMAL =      0,
2961         EXT4_IGET_SPECIAL =     0x0001, /* OK to iget a system inode */
2962         EXT4_IGET_HANDLE =      0x0002  /* Inode # is from a handle */
2963 } ext4_iget_flags;
2964
2965 extern struct inode *__ext4_iget(struct super_block *sb, unsigned long ino,
2966                                  ext4_iget_flags flags, const char *function,
2967                                  unsigned int line);
2968
2969 #define ext4_iget(sb, ino, flags) \
2970         __ext4_iget((sb), (ino), (flags), __func__, __LINE__)
2971
2972 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
2973 extern int  ext4_setattr(struct user_namespace *, struct dentry *,
2974                          struct iattr *);
2975 extern int  ext4_getattr(struct user_namespace *, const struct path *,
2976                          struct kstat *, u32, unsigned int);
2977 extern void ext4_evict_inode(struct inode *);
2978 extern void ext4_clear_inode(struct inode *);
2979 extern int  ext4_file_getattr(struct user_namespace *, const struct path *,
2980                               struct kstat *, u32, unsigned int);
2981 extern int  ext4_sync_inode(handle_t *, struct inode *);
2982 extern void ext4_dirty_inode(struct inode *, int);
2983 extern int ext4_change_inode_journal_flag(struct inode *, int);
2984 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
2985 extern int ext4_get_fc_inode_loc(struct super_block *sb, unsigned long ino,
2986                           struct ext4_iloc *iloc);
2987 extern int ext4_inode_attach_jinode(struct inode *inode);
2988 extern int ext4_can_truncate(struct inode *inode);
2989 extern int ext4_truncate(struct inode *);
2990 extern int ext4_break_layouts(struct inode *);
2991 extern int ext4_punch_hole(struct inode *inode, loff_t offset, loff_t length);
2992 extern void ext4_set_inode_flags(struct inode *, bool init);
2993 extern int ext4_alloc_da_blocks(struct inode *inode);
2994 extern void ext4_set_aops(struct inode *inode);
2995 extern int ext4_writepage_trans_blocks(struct inode *);
2996 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
2997 extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
2998                              loff_t lstart, loff_t lend);
2999 extern vm_fault_t ext4_page_mkwrite(struct vm_fault *vmf);
3000 extern vm_fault_t ext4_filemap_fault(struct vm_fault *vmf);
3001 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
3002 extern int ext4_get_projid(struct inode *inode, kprojid_t *projid);
3003 extern void ext4_da_release_space(struct inode *inode, int to_free);
3004 extern void ext4_da_update_reserve_space(struct inode *inode,
3005                                         int used, int quota_claim);
3006 extern int ext4_issue_zeroout(struct inode *inode, ext4_lblk_t lblk,
3007                               ext4_fsblk_t pblk, ext4_lblk_t len);
3008
3009 /* indirect.c */
3010 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
3011                                 struct ext4_map_blocks *map, int flags);
3012 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
3013 extern void ext4_ind_truncate(handle_t *, struct inode *inode);
3014 extern int ext4_ind_remove_space(handle_t *handle, struct inode *inode,
3015                                  ext4_lblk_t start, ext4_lblk_t end);
3016
3017 /* ioctl.c */
3018 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
3019 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
3020 int ext4_fileattr_set(struct user_namespace *mnt_userns,
3021                       struct dentry *dentry, struct fileattr *fa);
3022 int ext4_fileattr_get(struct dentry *dentry, struct fileattr *fa);
3023 extern void ext4_reset_inode_seed(struct inode *inode);
3024
3025 /* migrate.c */
3026 extern int ext4_ext_migrate(struct inode *);
3027 extern int ext4_ind_migrate(struct inode *inode);
3028
3029 /* namei.c */
3030 extern int ext4_init_new_dir(handle_t *handle, struct inode *dir,
3031                              struct inode *inode);
3032 extern int ext4_dirblock_csum_verify(struct inode *inode,
3033                                      struct buffer_head *bh);
3034 extern int ext4_orphan_add(handle_t *, struct inode *);
3035 extern int ext4_orphan_del(handle_t *, struct inode *);
3036 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
3037                                 __u32 start_minor_hash, __u32 *next_hash);
3038 extern int ext4_search_dir(struct buffer_head *bh,
3039                            char *search_buf,
3040                            int buf_size,
3041                            struct inode *dir,
3042                            struct ext4_filename *fname,
3043                            unsigned int offset,
3044                            struct ext4_dir_entry_2 **res_dir);
3045 extern int ext4_generic_delete_entry(struct inode *dir,
3046                                      struct ext4_dir_entry_2 *de_del,
3047                                      struct buffer_head *bh,
3048                                      void *entry_buf,
3049                                      int buf_size,
3050                                      int csum_size);
3051 extern bool ext4_empty_dir(struct inode *inode);
3052
3053 /* resize.c */
3054 extern void ext4_kvfree_array_rcu(void *to_free);
3055 extern int ext4_group_add(struct super_block *sb,
3056                                 struct ext4_new_group_data *input);
3057 extern int ext4_group_extend(struct super_block *sb,
3058                                 struct ext4_super_block *es,
3059                                 ext4_fsblk_t n_blocks_count);
3060 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
3061
3062 /* super.c */
3063 extern struct buffer_head *ext4_sb_bread(struct super_block *sb,
3064                                          sector_t block, int op_flags);
3065 extern struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
3066                                                    sector_t block);
3067 extern void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
3068                                 bh_end_io_t *end_io);
3069 extern int ext4_read_bh(struct buffer_head *bh, int op_flags,
3070                         bh_end_io_t *end_io);
3071 extern int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait);
3072 extern void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block);
3073 extern int ext4_seq_options_show(struct seq_file *seq, void *offset);
3074 extern int ext4_calculate_overhead(struct super_block *sb);
3075 extern void ext4_superblock_csum_set(struct super_block *sb);
3076 extern int ext4_alloc_flex_bg_array(struct super_block *sb,
3077                                     ext4_group_t ngroup);
3078 extern const char *ext4_decode_error(struct super_block *sb, int errno,
3079                                      char nbuf[16]);
3080 extern void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
3081                                              ext4_group_t block_group,
3082                                              unsigned int flags);
3083
3084 extern __printf(7, 8)
3085 void __ext4_error(struct super_block *, const char *, unsigned int, bool,
3086                   int, __u64, const char *, ...);
3087 extern __printf(6, 7)
3088 void __ext4_error_inode(struct inode *, const char *, unsigned int,
3089                         ext4_fsblk_t, int, const char *, ...);
3090 extern __printf(5, 6)
3091 void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
3092                      const char *, ...);
3093 extern void __ext4_std_error(struct super_block *, const char *,
3094                              unsigned int, int);
3095 extern __printf(4, 5)
3096 void __ext4_warning(struct super_block *, const char *, unsigned int,
3097                     const char *, ...);
3098 extern __printf(4, 5)
3099 void __ext4_warning_inode(const struct inode *inode, const char *function,
3100                           unsigned int line, const char *fmt, ...);
3101 extern __printf(3, 4)
3102 void __ext4_msg(struct super_block *, const char *, const char *, ...);
3103 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
3104                            const char *, unsigned int, const char *);
3105 extern __printf(7, 8)
3106 void __ext4_grp_locked_error(const char *, unsigned int,
3107                              struct super_block *, ext4_group_t,
3108                              unsigned long, ext4_fsblk_t,
3109                              const char *, ...);
3110
3111 #define EXT4_ERROR_INODE(inode, fmt, a...) \
3112         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
3113
3114 #define EXT4_ERROR_INODE_ERR(inode, err, fmt, a...)                     \
3115         __ext4_error_inode((inode), __func__, __LINE__, 0, (err), (fmt), ## a)
3116
3117 #define ext4_error_inode_block(inode, block, err, fmt, a...)            \
3118         __ext4_error_inode((inode), __func__, __LINE__, (block), (err), \
3119                            (fmt), ## a)
3120
3121 #define EXT4_ERROR_FILE(file, block, fmt, a...)                         \
3122         ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
3123
3124 #define ext4_abort(sb, err, fmt, a...)                                  \
3125         __ext4_error((sb), __func__, __LINE__, true, (err), 0, (fmt), ## a)
3126
3127 #ifdef CONFIG_PRINTK
3128
3129 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
3130         __ext4_error_inode(inode, func, line, block, 0, fmt, ##__VA_ARGS__)
3131 #define ext4_error_inode_err(inode, func, line, block, err, fmt, ...)   \
3132         __ext4_error_inode((inode), (func), (line), (block),            \
3133                            (err), (fmt), ##__VA_ARGS__)
3134 #define ext4_error_file(file, func, line, block, fmt, ...)              \
3135         __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__)
3136 #define ext4_error(sb, fmt, ...)                                        \
3137         __ext4_error((sb), __func__, __LINE__, false, 0, 0, (fmt),      \
3138                 ##__VA_ARGS__)
3139 #define ext4_error_err(sb, err, fmt, ...)                               \
3140         __ext4_error((sb), __func__, __LINE__, false, (err), 0, (fmt),  \
3141                 ##__VA_ARGS__)
3142 #define ext4_warning(sb, fmt, ...)                                      \
3143         __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
3144 #define ext4_warning_inode(inode, fmt, ...)                             \
3145         __ext4_warning_inode(inode, __func__, __LINE__, fmt, ##__VA_ARGS__)
3146 #define ext4_msg(sb, level, fmt, ...)                           \
3147         __ext4_msg(sb, level, fmt, ##__VA_ARGS__)
3148 #define dump_mmp_msg(sb, mmp, msg)                                      \
3149         __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg)
3150 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
3151         __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \
3152                                 fmt, ##__VA_ARGS__)
3153
3154 #else
3155
3156 #define ext4_error_inode(inode, func, line, block, fmt, ...)            \
3157 do {                                                                    \
3158         no_printk(fmt, ##__VA_ARGS__);                                  \
3159         __ext4_error_inode(inode, "", 0, block, 0, " ");                \
3160 } while (0)
3161 #define ext4_error_inode_err(inode, func, line, block, err, fmt, ...)   \
3162 do {                                                                    \
3163         no_printk(fmt, ##__VA_ARGS__);                                  \
3164         __ext4_error_inode(inode, "", 0, block, err, " ");              \
3165 } while (0)
3166 #define ext4_error_file(file, func, line, block, fmt, ...)              \
3167 do {                                                                    \
3168         no_printk(fmt, ##__VA_ARGS__);                                  \
3169         __ext4_error_file(file, "", 0, block, " ");                     \
3170 } while (0)
3171 #define ext4_error(sb, fmt, ...)                                        \
3172 do {                                                                    \
3173         no_printk(fmt, ##__VA_ARGS__);                                  \
3174         __ext4_error(sb, "", 0, false, 0, 0, " ");                      \
3175 } while (0)
3176 #define ext4_error_err(sb, err, fmt, ...)                               \
3177 do {                                                                    \
3178         no_printk(fmt, ##__VA_ARGS__);                                  \
3179         __ext4_error(sb, "", 0, false, err, 0, " ");                    \
3180 } while (0)
3181 #define ext4_warning(sb, fmt, ...)                                      \
3182 do {                                                                    \
3183         no_printk(fmt, ##__VA_ARGS__);                                  \
3184         __ext4_warning(sb, "", 0, " ");                                 \
3185 } while (0)
3186 #define ext4_warning_inode(inode, fmt, ...)                             \
3187 do {                                                                    \
3188         no_printk(fmt, ##__VA_ARGS__);                                  \
3189         __ext4_warning_inode(inode, "", 0, " ");                        \
3190 } while (0)
3191 #define ext4_msg(sb, level, fmt, ...)                                   \
3192 do {                                                                    \
3193         no_printk(fmt, ##__VA_ARGS__);                                  \
3194         __ext4_msg(sb, "", " ");                                        \
3195 } while (0)
3196 #define dump_mmp_msg(sb, mmp, msg)                                      \
3197         __dump_mmp_msg(sb, mmp, "", 0, "")
3198 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...)            \
3199 do {                                                                    \
3200         no_printk(fmt, ##__VA_ARGS__);                          \
3201         __ext4_grp_locked_error("", 0, sb, grp, ino, block, " ");       \
3202 } while (0)
3203
3204 #endif
3205
3206 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
3207                                       struct ext4_group_desc *bg);
3208 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
3209                                       struct ext4_group_desc *bg);
3210 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
3211                                      struct ext4_group_desc *bg);
3212 extern __u32 ext4_free_group_clusters(struct super_block *sb,
3213                                       struct ext4_group_desc *bg);
3214 extern __u32 ext4_free_inodes_count(struct super_block *sb,
3215                                  struct ext4_group_desc *bg);
3216 extern __u32 ext4_used_dirs_count(struct super_block *sb,
3217                                 struct ext4_group_desc *bg);
3218 extern __u32 ext4_itable_unused_count(struct super_block *sb,
3219                                    struct ext4_group_desc *bg);
3220 extern void ext4_block_bitmap_set(struct super_block *sb,
3221                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
3222 extern void ext4_inode_bitmap_set(struct super_block *sb,
3223                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
3224 extern void ext4_inode_table_set(struct super_block *sb,
3225                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
3226 extern void ext4_free_group_clusters_set(struct super_block *sb,
3227                                          struct ext4_group_desc *bg,
3228                                          __u32 count);
3229 extern void ext4_free_inodes_set(struct super_block *sb,
3230                                 struct ext4_group_desc *bg, __u32 count);
3231 extern void ext4_used_dirs_set(struct super_block *sb,
3232                                 struct ext4_group_desc *bg, __u32 count);
3233 extern void ext4_itable_unused_set(struct super_block *sb,
3234                                    struct ext4_group_desc *bg, __u32 count);
3235 extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
3236                                        struct ext4_group_desc *gdp);
3237 extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
3238                                      struct ext4_group_desc *gdp);
3239 extern int ext4_register_li_request(struct super_block *sb,
3240                                     ext4_group_t first_not_zeroed);
3241
3242 static inline int ext4_has_metadata_csum(struct super_block *sb)
3243 {
3244         WARN_ON_ONCE(ext4_has_feature_metadata_csum(sb) &&
3245                      !EXT4_SB(sb)->s_chksum_driver);
3246
3247         return ext4_has_feature_metadata_csum(sb) &&
3248                (EXT4_SB(sb)->s_chksum_driver != NULL);
3249 }
3250
3251 static inline int ext4_has_group_desc_csum(struct super_block *sb)
3252 {
3253         return ext4_has_feature_gdt_csum(sb) || ext4_has_metadata_csum(sb);
3254 }
3255
3256 #define ext4_read_incompat_64bit_val(es, name) \
3257         (((es)->s_feature_incompat & cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT) \
3258                 ? (ext4_fsblk_t)le32_to_cpu(es->name##_hi) << 32 : 0) | \
3259                 le32_to_cpu(es->name##_lo))
3260
3261 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
3262 {
3263         return ext4_read_incompat_64bit_val(es, s_blocks_count);
3264 }
3265
3266 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
3267 {
3268         return ext4_read_incompat_64bit_val(es, s_r_blocks_count);
3269 }
3270
3271 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
3272 {
3273         return ext4_read_incompat_64bit_val(es, s_free_blocks_count);
3274 }
3275
3276 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
3277                                          ext4_fsblk_t blk)
3278 {
3279         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
3280         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
3281 }
3282
3283 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
3284                                               ext4_fsblk_t blk)
3285 {
3286         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
3287         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
3288 }
3289
3290 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
3291                                            ext4_fsblk_t blk)
3292 {
3293         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
3294         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
3295 }
3296
3297 static inline loff_t ext4_isize(struct super_block *sb,
3298                                 struct ext4_inode *raw_inode)
3299 {
3300         if (ext4_has_feature_largedir(sb) ||
3301             S_ISREG(le16_to_cpu(raw_inode->i_mode)))
3302                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
3303                         le32_to_cpu(raw_inode->i_size_lo);
3304
3305         return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
3306 }
3307
3308 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
3309 {
3310         raw_inode->i_size_lo = cpu_to_le32(i_size);
3311         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
3312 }
3313
3314 static inline
3315 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
3316                                             ext4_group_t group)
3317 {
3318          struct ext4_group_info **grp_info;
3319          long indexv, indexh;
3320          BUG_ON(group >= EXT4_SB(sb)->s_groups_count);
3321          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
3322          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
3323          grp_info = sbi_array_rcu_deref(EXT4_SB(sb), s_group_info, indexv);
3324          return grp_info[indexh];
3325 }
3326
3327 /*
3328  * Reading s_groups_count requires using smp_rmb() afterwards.  See
3329  * the locking protocol documented in the comments of ext4_group_add()
3330  * in resize.c
3331  */
3332 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
3333 {
3334         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
3335
3336         smp_rmb();
3337         return ngroups;
3338 }
3339
3340 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
3341                                              ext4_group_t block_group)
3342 {
3343         return block_group >> sbi->s_log_groups_per_flex;
3344 }
3345
3346 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
3347 {
3348         return 1 << sbi->s_log_groups_per_flex;
3349 }
3350
3351 #define ext4_std_error(sb, errno)                               \
3352 do {                                                            \
3353         if ((errno))                                            \
3354                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
3355 } while (0)
3356
3357 #ifdef CONFIG_SMP
3358 /* Each CPU can accumulate percpu_counter_batch clusters in their local
3359  * counters. So we need to make sure we have free clusters more
3360  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
3361  */
3362 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
3363 #else
3364 #define EXT4_FREECLUSTERS_WATERMARK 0
3365 #endif
3366
3367 /* Update i_disksize. Requires i_mutex to avoid races with truncate */
3368 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
3369 {
3370         WARN_ON_ONCE(S_ISREG(inode->i_mode) &&
3371                      !inode_is_locked(inode));
3372         down_write(&EXT4_I(inode)->i_data_sem);
3373         if (newsize > EXT4_I(inode)->i_disksize)
3374                 WRITE_ONCE(EXT4_I(inode)->i_disksize, newsize);
3375         up_write(&EXT4_I(inode)->i_data_sem);
3376 }
3377
3378 /* Update i_size, i_disksize. Requires i_mutex to avoid races with truncate */
3379 static inline int ext4_update_inode_size(struct inode *inode, loff_t newsize)
3380 {
3381         int changed = 0;
3382
3383         if (newsize > inode->i_size) {
3384                 i_size_write(inode, newsize);
3385                 changed = 1;
3386         }
3387         if (newsize > EXT4_I(inode)->i_disksize) {
3388                 ext4_update_i_disksize(inode, newsize);
3389                 changed |= 2;
3390         }
3391         return changed;
3392 }
3393
3394 int ext4_update_disksize_before_punch(struct inode *inode, loff_t offset,
3395                                       loff_t len);
3396
3397 struct ext4_group_info {
3398         unsigned long   bb_state;
3399 #ifdef AGGRESSIVE_CHECK
3400         unsigned long   bb_check_counter;
3401 #endif
3402         struct rb_root  bb_free_root;
3403         ext4_grpblk_t   bb_first_free;  /* first free block */
3404         ext4_grpblk_t   bb_free;        /* total free blocks */
3405         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
3406         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
3407         ext4_group_t    bb_group;       /* Group number */
3408         struct          list_head bb_prealloc_list;
3409 #ifdef DOUBLE_CHECK
3410         void            *bb_bitmap;
3411 #endif
3412         struct rw_semaphore alloc_sem;
3413         struct rb_node  bb_avg_fragment_size_rb;
3414         struct list_head bb_largest_free_order_node;
3415         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
3416                                          * regions, index is order.
3417                                          * bb_counters[3] = 5 means
3418                                          * 5 free 8-block regions. */
3419 };
3420
3421 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
3422 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
3423 #define EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT     2
3424 #define EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT     3
3425 #define EXT4_GROUP_INFO_BBITMAP_CORRUPT         \
3426         (1 << EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT)
3427 #define EXT4_GROUP_INFO_IBITMAP_CORRUPT         \
3428         (1 << EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT)
3429 #define EXT4_GROUP_INFO_BBITMAP_READ_BIT        4
3430
3431 #define EXT4_MB_GRP_NEED_INIT(grp)      \
3432         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
3433 #define EXT4_MB_GRP_BBITMAP_CORRUPT(grp)        \
3434         (test_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
3435 #define EXT4_MB_GRP_IBITMAP_CORRUPT(grp)        \
3436         (test_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
3437
3438 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
3439         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3440 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
3441         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3442 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
3443         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3444 #define EXT4_MB_GRP_TEST_AND_SET_READ(grp)      \
3445         (test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_READ_BIT, &((grp)->bb_state)))
3446
3447 #define EXT4_MAX_CONTENTION             8
3448 #define EXT4_CONTENTION_THRESHOLD       2
3449
3450 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
3451                                               ext4_group_t group)
3452 {
3453         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
3454 }
3455
3456 /*
3457  * Returns true if the filesystem is busy enough that attempts to
3458  * access the block group locks has run into contention.
3459  */
3460 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
3461 {
3462         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
3463 }
3464
3465 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
3466 {
3467         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
3468         if (spin_trylock(lock))
3469                 /*
3470                  * We're able to grab the lock right away, so drop the
3471                  * lock contention counter.
3472                  */
3473                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
3474         else {
3475                 /*
3476                  * The lock is busy, so bump the contention counter,
3477                  * and then wait on the spin lock.
3478                  */
3479                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
3480                                   EXT4_MAX_CONTENTION);
3481                 spin_lock(lock);
3482         }
3483 }
3484
3485 static inline void ext4_unlock_group(struct super_block *sb,
3486                                         ext4_group_t group)
3487 {
3488         spin_unlock(ext4_group_lock_ptr(sb, group));
3489 }
3490
3491 #ifdef CONFIG_QUOTA
3492 static inline bool ext4_quota_capable(struct super_block *sb)
3493 {
3494         return (test_opt(sb, QUOTA) || ext4_has_feature_quota(sb));
3495 }
3496
3497 static inline bool ext4_is_quota_journalled(struct super_block *sb)
3498 {
3499         struct ext4_sb_info *sbi = EXT4_SB(sb);
3500
3501         return (ext4_has_feature_quota(sb) ||
3502                 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]);
3503 }
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                                          unsigned flags,
3543                                          struct page **pagep);
3544 extern int ext4_write_inline_data_end(struct inode *inode,
3545                                       loff_t pos, unsigned len,
3546                                       unsigned copied,
3547                                       struct page *page);
3548 extern struct buffer_head *
3549 ext4_journalled_write_inline_data(struct inode *inode,
3550                                   unsigned len,
3551                                   struct page *page);
3552 extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
3553                                            struct inode *inode,
3554                                            loff_t pos, unsigned len,
3555                                            unsigned flags,
3556                                            struct page **pagep,
3557                                            void **fsdata);
3558 extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
3559                                          unsigned len, unsigned copied,
3560                                          struct page *page);
3561 extern int ext4_try_add_inline_entry(handle_t *handle,
3562                                      struct ext4_filename *fname,
3563                                      struct inode *dir, struct inode *inode);
3564 extern int ext4_try_create_inline_dir(handle_t *handle,
3565                                       struct inode *parent,
3566                                       struct inode *inode);
3567 extern int ext4_read_inline_dir(struct file *filp,
3568                                 struct dir_context *ctx,
3569                                 int *has_inline_data);
3570 extern int ext4_inlinedir_to_tree(struct file *dir_file,
3571                                   struct inode *dir, ext4_lblk_t block,
3572                                   struct dx_hash_info *hinfo,
3573                                   __u32 start_hash, __u32 start_minor_hash,
3574                                   int *has_inline_data);
3575 extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
3576                                         struct ext4_filename *fname,
3577                                         struct ext4_dir_entry_2 **res_dir,
3578                                         int *has_inline_data);
3579 extern int ext4_delete_inline_entry(handle_t *handle,
3580                                     struct inode *dir,
3581                                     struct ext4_dir_entry_2 *de_del,
3582                                     struct buffer_head *bh,
3583                                     int *has_inline_data);
3584 extern bool empty_inline_dir(struct inode *dir, int *has_inline_data);
3585 extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
3586                                         struct ext4_dir_entry_2 **parent_de,
3587                                         int *retval);
3588 extern int ext4_inline_data_fiemap(struct inode *inode,
3589                                    struct fiemap_extent_info *fieinfo,
3590                                    int *has_inline, __u64 start, __u64 len);
3591
3592 struct iomap;
3593 extern int ext4_inline_data_iomap(struct inode *inode, struct iomap *iomap);
3594
3595 extern int ext4_inline_data_truncate(struct inode *inode, int *has_inline);
3596
3597 extern int ext4_convert_inline_data(struct inode *inode);
3598
3599 static inline int ext4_has_inline_data(struct inode *inode)
3600 {
3601         return ext4_test_inode_flag(inode, EXT4_INODE_INLINE_DATA) &&
3602                EXT4_I(inode)->i_inline_off;
3603 }
3604
3605 /* namei.c */
3606 extern const struct inode_operations ext4_dir_inode_operations;
3607 extern const struct inode_operations ext4_special_inode_operations;
3608 extern struct dentry *ext4_get_parent(struct dentry *child);
3609 extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
3610                                  struct ext4_dir_entry_2 *de,
3611                                  int blocksize, int csum_size,
3612                                  unsigned int parent_ino, int dotdot_real_len);
3613 extern void ext4_initialize_dirent_tail(struct buffer_head *bh,
3614                                         unsigned int blocksize);
3615 extern int ext4_handle_dirty_dirblock(handle_t *handle, struct inode *inode,
3616                                       struct buffer_head *bh);
3617 extern int __ext4_unlink(handle_t *handle, struct inode *dir, const struct qstr *d_name,
3618                          struct inode *inode);
3619 extern int __ext4_link(struct inode *dir, struct inode *inode,
3620                        struct dentry *dentry);
3621
3622 #define S_SHIFT 12
3623 static const unsigned char ext4_type_by_mode[(S_IFMT >> S_SHIFT) + 1] = {
3624         [S_IFREG >> S_SHIFT]    = EXT4_FT_REG_FILE,
3625         [S_IFDIR >> S_SHIFT]    = EXT4_FT_DIR,
3626         [S_IFCHR >> S_SHIFT]    = EXT4_FT_CHRDEV,
3627         [S_IFBLK >> S_SHIFT]    = EXT4_FT_BLKDEV,
3628         [S_IFIFO >> S_SHIFT]    = EXT4_FT_FIFO,
3629         [S_IFSOCK >> S_SHIFT]   = EXT4_FT_SOCK,
3630         [S_IFLNK >> S_SHIFT]    = EXT4_FT_SYMLINK,
3631 };
3632
3633 static inline void ext4_set_de_type(struct super_block *sb,
3634                                 struct ext4_dir_entry_2 *de,
3635                                 umode_t mode) {
3636         if (ext4_has_feature_filetype(sb))
3637                 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
3638 }
3639
3640 /* readpages.c */
3641 extern int ext4_mpage_readpages(struct inode *inode,
3642                 struct readahead_control *rac, struct page *page);
3643 extern int __init ext4_init_post_read_processing(void);
3644 extern void ext4_exit_post_read_processing(void);
3645
3646 /* symlink.c */
3647 extern const struct inode_operations ext4_encrypted_symlink_inode_operations;
3648 extern const struct inode_operations ext4_symlink_inode_operations;
3649 extern const struct inode_operations ext4_fast_symlink_inode_operations;
3650
3651 /* sysfs.c */
3652 extern void ext4_notify_error_sysfs(struct ext4_sb_info *sbi);
3653 extern int ext4_register_sysfs(struct super_block *sb);
3654 extern void ext4_unregister_sysfs(struct super_block *sb);
3655 extern int __init ext4_init_sysfs(void);
3656 extern void ext4_exit_sysfs(void);
3657
3658 /* block_validity */
3659 extern void ext4_release_system_zone(struct super_block *sb);
3660 extern int ext4_setup_system_zone(struct super_block *sb);
3661 extern int __init ext4_init_system_zone(void);
3662 extern void ext4_exit_system_zone(void);
3663 extern int ext4_inode_block_valid(struct inode *inode,
3664                                   ext4_fsblk_t start_blk,
3665                                   unsigned int count);
3666 extern int ext4_check_blockref(const char *, unsigned int,
3667                                struct inode *, __le32 *, unsigned int);
3668
3669 /* extents.c */
3670 struct ext4_ext_path;
3671 struct ext4_extent;
3672
3673 /*
3674  * Maximum number of logical blocks in a file; ext4_extent's ee_block is
3675  * __le32.
3676  */
3677 #define EXT_MAX_BLOCKS  0xffffffff
3678
3679 extern void ext4_ext_tree_init(handle_t *handle, struct inode *inode);
3680 extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
3681 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
3682                                struct ext4_map_blocks *map, int flags);
3683 extern int ext4_ext_truncate(handle_t *, struct inode *);
3684 extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
3685                                  ext4_lblk_t end);
3686 extern void ext4_ext_init(struct super_block *);
3687 extern void ext4_ext_release(struct super_block *);
3688 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
3689                           loff_t len);
3690 extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
3691                                           loff_t offset, ssize_t len);
3692 extern int ext4_convert_unwritten_io_end_vec(handle_t *handle,
3693                                              ext4_io_end_t *io_end);
3694 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
3695                            struct ext4_map_blocks *map, int flags);
3696 extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
3697                                                    int num,
3698                                                    struct ext4_ext_path *path);
3699 extern int ext4_ext_insert_extent(handle_t *, struct inode *,
3700                                   struct ext4_ext_path **,
3701                                   struct ext4_extent *, int);
3702 extern struct ext4_ext_path *ext4_find_extent(struct inode *, ext4_lblk_t,
3703                                               struct ext4_ext_path **,
3704                                               int flags);
3705 extern void ext4_ext_drop_refs(struct ext4_ext_path *);
3706 extern int ext4_ext_check_inode(struct inode *inode);
3707 extern ext4_lblk_t ext4_ext_next_allocated_block(struct ext4_ext_path *path);
3708 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
3709                         __u64 start, __u64 len);
3710 extern int ext4_get_es_cache(struct inode *inode,
3711                              struct fiemap_extent_info *fieinfo,
3712                              __u64 start, __u64 len);
3713 extern int ext4_ext_precache(struct inode *inode);
3714 extern int ext4_swap_extents(handle_t *handle, struct inode *inode1,
3715                                 struct inode *inode2, ext4_lblk_t lblk1,
3716                              ext4_lblk_t lblk2,  ext4_lblk_t count,
3717                              int mark_unwritten,int *err);
3718 extern int ext4_clu_mapped(struct inode *inode, ext4_lblk_t lclu);
3719 extern int ext4_datasem_ensure_credits(handle_t *handle, struct inode *inode,
3720                                        int check_cred, int restart_cred,
3721                                        int revoke_cred);
3722 extern void ext4_ext_replay_shrink_inode(struct inode *inode, ext4_lblk_t end);
3723 extern int ext4_ext_replay_set_iblocks(struct inode *inode);
3724 extern int ext4_ext_replay_update_ex(struct inode *inode, ext4_lblk_t start,
3725                 int len, int unwritten, ext4_fsblk_t pblk);
3726 extern int ext4_ext_clear_bb(struct inode *inode);
3727
3728
3729 /* move_extent.c */
3730 extern void ext4_double_down_write_data_sem(struct inode *first,
3731                                             struct inode *second);
3732 extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
3733                                           struct inode *donor_inode);
3734 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
3735                              __u64 start_orig, __u64 start_donor,
3736                              __u64 len, __u64 *moved_len);
3737
3738 /* page-io.c */
3739 extern int __init ext4_init_pageio(void);
3740 extern void ext4_exit_pageio(void);
3741 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
3742 extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
3743 extern int ext4_put_io_end(ext4_io_end_t *io_end);
3744 extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
3745 extern void ext4_io_submit_init(struct ext4_io_submit *io,
3746                                 struct writeback_control *wbc);
3747 extern void ext4_end_io_rsv_work(struct work_struct *work);
3748 extern void ext4_io_submit(struct ext4_io_submit *io);
3749 extern int ext4_bio_write_page(struct ext4_io_submit *io,
3750                                struct page *page,
3751                                int len,
3752                                bool keep_towrite);
3753 extern struct ext4_io_end_vec *ext4_alloc_io_end_vec(ext4_io_end_t *io_end);
3754 extern struct ext4_io_end_vec *ext4_last_io_end_vec(ext4_io_end_t *io_end);
3755
3756 /* mmp.c */
3757 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
3758
3759 /* mmp.c */
3760 extern void ext4_stop_mmpd(struct ext4_sb_info *sbi);
3761
3762 /* verity.c */
3763 extern const struct fsverity_operations ext4_verityops;
3764
3765 /*
3766  * Add new method to test whether block and inode bitmaps are properly
3767  * initialized. With uninit_bg reading the block from disk is not enough
3768  * to mark the bitmap uptodate. We need to also zero-out the bitmap
3769  */
3770 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
3771
3772 static inline int bitmap_uptodate(struct buffer_head *bh)
3773 {
3774         return (buffer_uptodate(bh) &&
3775                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
3776 }
3777 static inline void set_bitmap_uptodate(struct buffer_head *bh)
3778 {
3779         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
3780 }
3781
3782 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
3783
3784 /* For ioend & aio unwritten conversion wait queues */
3785 #define EXT4_WQ_HASH_SZ         37
3786 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
3787                                             EXT4_WQ_HASH_SZ])
3788 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
3789
3790 extern int ext4_resize_begin(struct super_block *sb);
3791 extern void ext4_resize_end(struct super_block *sb);
3792
3793 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
3794                                               struct ext4_io_end *io_end)
3795 {
3796         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
3797                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
3798                 atomic_inc(&EXT4_I(inode)->i_unwritten);
3799         }
3800 }
3801
3802 static inline void ext4_clear_io_unwritten_flag(ext4_io_end_t *io_end)
3803 {
3804         struct inode *inode = io_end->inode;
3805
3806         if (io_end->flag & EXT4_IO_END_UNWRITTEN) {
3807                 io_end->flag &= ~EXT4_IO_END_UNWRITTEN;
3808                 /* Wake up anyone waiting on unwritten extent conversion */
3809                 if (atomic_dec_and_test(&EXT4_I(inode)->i_unwritten))
3810                         wake_up_all(ext4_ioend_wq(inode));
3811         }
3812 }
3813
3814 extern const struct iomap_ops ext4_iomap_ops;
3815 extern const struct iomap_ops ext4_iomap_overwrite_ops;
3816 extern const struct iomap_ops ext4_iomap_report_ops;
3817
3818 static inline int ext4_buffer_uptodate(struct buffer_head *bh)
3819 {
3820         /*
3821          * If the buffer has the write error flag, we have failed
3822          * to write out data in the block.  In this  case, we don't
3823          * have to read the block because we may read the old data
3824          * successfully.
3825          */
3826         if (buffer_write_io_error(bh))
3827                 set_buffer_uptodate(bh);
3828         return buffer_uptodate(bh);
3829 }
3830
3831 #endif  /* __KERNEL__ */
3832
3833 #define EFSBADCRC       EBADMSG         /* Bad CRC detected */
3834 #define EFSCORRUPTED    EUCLEAN         /* Filesystem is corrupted */
3835
3836 #endif  /* _EXT4_H */