1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * ext4.h
4 *
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
9 *
10 * from
11 *
12 * linux/include/linux/minix_fs.h
13 *
14 * Copyright (C) 1991, 1992 Linus Torvalds
15 */
16
17 #ifndef _EXT4_H
18 #define _EXT4_H
19
20 #include <linux/refcount.h>
21 #include <linux/types.h>
22 #include <linux/blkdev.h>
23 #include <linux/magic.h>
24 #include <linux/jbd2.h>
25 #include <linux/quota.h>
26 #include <linux/rwsem.h>
27 #include <linux/rbtree.h>
28 #include <linux/seqlock.h>
29 #include <linux/mutex.h>
30 #include <linux/timer.h>
31 #include <linux/wait.h>
32 #include <linux/sched/signal.h>
33 #include <linux/blockgroup_lock.h>
34 #include <linux/percpu_counter.h>
35 #include <linux/ratelimit.h>
36 #include <crypto/hash.h>
37 #include <linux/falloc.h>
38 #include <linux/percpu-rwsem.h>
39 #include <linux/fiemap.h>
40 #ifdef __KERNEL__
41 #include <linux/compat.h>
42 #endif
43 #include <uapi/linux/ext4.h>
44
45 #include <linux/fscrypt.h>
46 #include <linux/fsverity.h>
47
48 #include <linux/compiler.h>
49
50 /*
51 * The fourth extended filesystem constants/structures
52 */
53
54 /*
55 * with AGGRESSIVE_CHECK allocator runs consistency checks over
56 * structures. these checks slow things down a lot
57 */
58 #define AGGRESSIVE_CHECK__
59
60 /*
61 * with DOUBLE_CHECK defined mballoc creates persistent in-core
62 * bitmaps, maintains and uses them to check for double allocations
63 */
64 #define DOUBLE_CHECK__
65
66 /*
67 * Define EXT4FS_DEBUG to produce debug messages
68 */
69 #undef EXT4FS_DEBUG
70
71 /*
72 * Debug code
73 */
74 #ifdef EXT4FS_DEBUG
75 #define ext4_debug(f, a...) \
76 do { \
77 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:", \
78 __FILE__, __LINE__, __func__); \
79 printk(KERN_DEBUG f, ## a); \
80 } while (0)
81 #else
82 #define ext4_debug(fmt, ...) no_printk(fmt, ##__VA_ARGS__)
83 #endif
84
85 /*
86 * Turn on EXT_DEBUG to enable ext4_ext_show_path/leaf/move in extents.c
87 */
88 #define EXT_DEBUG__
89
90 /*
91 * Dynamic printk for controlled extents debugging.
92 */
93 #ifdef CONFIG_EXT4_DEBUG
94 #define ext_debug(ino, fmt, ...) \
95 pr_debug("[%s/%d] EXT4-fs (%s): ino %lu: (%s, %d): %s:" fmt, \
96 current->comm, task_pid_nr(current), \
97 ino->i_sb->s_id, ino->i_ino, __FILE__, __LINE__, \
98 __func__, ##__VA_ARGS__)
99 #else
100 #define ext_debug(ino, fmt, ...) no_printk(fmt, ##__VA_ARGS__)
101 #endif
102
103 #define ASSERT(assert) \
104 do { \
105 if (unlikely(!(assert))) { \
106 printk(KERN_EMERG \
107 "Assertion failure in %s() at %s:%d: '%s'\n", \
108 __func__, __FILE__, __LINE__, #assert); \
109 BUG(); \
110 } \
111 } while (0)
112
113 /* data type for block offset of block group */
114 typedef int ext4_grpblk_t;
115
116 /* data type for filesystem-wide blocks number */
117 typedef unsigned long long ext4_fsblk_t;
118
119 /* data type for file logical block number */
120 typedef __u32 ext4_lblk_t;
121
122 /* data type for block group number */
123 typedef unsigned int ext4_group_t;
124
125 enum SHIFT_DIRECTION {
126 SHIFT_LEFT = 0,
127 SHIFT_RIGHT,
128 };
129
130 /*
131 * For each criteria, mballoc has slightly different way of finding
132 * the required blocks nad usually, higher the criteria the slower the
133 * allocation. We start at lower criterias and keep falling back to
134 * higher ones if we are not able to find any blocks. Lower (earlier)
135 * criteria are faster.
136 */
137 enum criteria {
138 /*
139 * Used when number of blocks needed is a power of 2. This
140 * doesn't trigger any disk IO except prefetch and is the
141 * fastest criteria.
142 */
143 CR_POWER2_ALIGNED,
144
145 /*
146 * Tries to lookup in-memory data structures to find the most
147 * suitable group that satisfies goal request. No disk IO
148 * except block prefetch.
149 */
150 CR_GOAL_LEN_FAST,
151
152 /*
153 * Same as CR_GOAL_LEN_FAST but is allowed to reduce the goal
154 * length to the best available length for faster allocation.
155 */
156 CR_BEST_AVAIL_LEN,
157
158 /*
159 * Reads each block group sequentially, performing disk IO if
160 * necessary, to find find_suitable block group. Tries to
161 * allocate goal length but might trim the request if nothing
162 * is found after enough tries.
163 */
164 CR_GOAL_LEN_SLOW,
165
166 /*
167 * Finds the first free set of blocks and allocates
168 * those. This is only used in rare cases when
169 * CR_GOAL_LEN_SLOW also fails to allocate anything.
170 */
171 CR_ANY_FREE,
172
173 /*
174 * Number of criterias defined.
175 */
176 EXT4_MB_NUM_CRS
177 };
178
179 /*
180 * Flags used in mballoc's allocation_context flags field.
181 *
182 * Also used to show what's going on for debugging purposes when the
183 * flag field is exported via the traceport interface
184 */
185
186 /* prefer goal again. length */
187 #define EXT4_MB_HINT_MERGE 0x0001
188 /* blocks already reserved */
189 #define EXT4_MB_HINT_RESERVED 0x0002
190 /* metadata is being allocated */
191 #define EXT4_MB_HINT_METADATA 0x0004
192 /* first blocks in the file */
193 #define EXT4_MB_HINT_FIRST 0x0008
194 /* search for the best chunk */
195 #define EXT4_MB_HINT_BEST 0x0010
196 /* data is being allocated */
197 #define EXT4_MB_HINT_DATA 0x0020
198 /* don't preallocate (for tails) */
199 #define EXT4_MB_HINT_NOPREALLOC 0x0040
200 /* allocate for locality group */
201 #define EXT4_MB_HINT_GROUP_ALLOC 0x0080
202 /* allocate goal blocks or none */
203 #define EXT4_MB_HINT_GOAL_ONLY 0x0100
204 /* goal is meaningful */
205 #define EXT4_MB_HINT_TRY_GOAL 0x0200
206 /* blocks already pre-reserved by delayed allocation */
207 #define EXT4_MB_DELALLOC_RESERVED 0x0400
208 /* We are doing stream allocation */
209 #define EXT4_MB_STREAM_ALLOC 0x0800
210 /* Use reserved root blocks if needed */
211 #define EXT4_MB_USE_ROOT_BLOCKS 0x1000
212 /* Use blocks from reserved pool */
213 #define EXT4_MB_USE_RESERVED 0x2000
214 /* Do strict check for free blocks while retrying block allocation */
215 #define EXT4_MB_STRICT_CHECK 0x4000
216 /* Large fragment size list lookup succeeded at least once for
217 * CR_POWER2_ALIGNED */
218 #define EXT4_MB_CR_POWER2_ALIGNED_OPTIMIZED 0x8000
219 /* Avg fragment size rb tree lookup succeeded at least once for
220 * CR_GOAL_LEN_FAST */
221 #define EXT4_MB_CR_GOAL_LEN_FAST_OPTIMIZED 0x00010000
222 /* Avg fragment size rb tree lookup succeeded at least once for
223 * CR_BEST_AVAIL_LEN */
224 #define EXT4_MB_CR_BEST_AVAIL_LEN_OPTIMIZED 0x00020000
225
226 struct ext4_allocation_request {
227 /* target inode for block we're allocating */
228 struct inode *inode;
229 /* how many blocks we want to allocate */
230 unsigned int len;
231 /* logical block in target inode */
232 ext4_lblk_t logical;
233 /* the closest logical allocated block to the left */
234 ext4_lblk_t lleft;
235 /* the closest logical allocated block to the right */
236 ext4_lblk_t lright;
237 /* phys. target (a hint) */
238 ext4_fsblk_t goal;
239 /* phys. block for the closest logical allocated block to the left */
240 ext4_fsblk_t pleft;
241 /* phys. block for the closest logical allocated block to the right */
242 ext4_fsblk_t pright;
243 /* flags. see above EXT4_MB_HINT_* */
244 unsigned int flags;
245 };
246
247 /*
248 * Logical to physical block mapping, used by ext4_map_blocks()
249 *
250 * This structure is used to pass requests into ext4_map_blocks() as
251 * well as to store the information returned by ext4_map_blocks(). It
252 * takes less room on the stack than a struct buffer_head.
253 */
254 #define EXT4_MAP_NEW BIT(BH_New)
255 #define EXT4_MAP_MAPPED BIT(BH_Mapped)
256 #define EXT4_MAP_UNWRITTEN BIT(BH_Unwritten)
257 #define EXT4_MAP_BOUNDARY BIT(BH_Boundary)
258 #define EXT4_MAP_DELAYED BIT(BH_Delay)
259 #define EXT4_MAP_FLAGS (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
260 EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
261 EXT4_MAP_DELAYED)
262
263 struct ext4_map_blocks {
264 ext4_fsblk_t m_pblk;
265 ext4_lblk_t m_lblk;
266 unsigned int m_len;
267 unsigned int m_flags;
268 };
269
270 /*
271 * Block validity checking, system zone rbtree.
272 */
273 struct ext4_system_blocks {
274 struct rb_root root;
275 struct rcu_head rcu;
276 };
277
278 /*
279 * Flags for ext4_io_end->flags
280 */
281 #define EXT4_IO_END_UNWRITTEN 0x0001
282
283 struct ext4_io_end_vec {
284 struct list_head list; /* list of io_end_vec */
285 loff_t offset; /* offset in the file */
286 ssize_t size; /* size of the extent */
287 };
288
289 /*
290 * For converting unwritten extents on a work queue. 'handle' is used for
291 * buffered writeback.
292 */
293 typedef struct ext4_io_end {
294 struct list_head list; /* per-file finished IO list */
295 handle_t *handle; /* handle reserved for extent
296 * conversion */
297 struct inode *inode; /* file being written to */
298 struct bio *bio; /* Linked list of completed
299 * bios covering the extent */
300 unsigned int flag; /* unwritten or not */
301 refcount_t count; /* reference counter */
302 struct list_head list_vec; /* list of ext4_io_end_vec */
303 } ext4_io_end_t;
304
305 struct ext4_io_submit {
306 struct writeback_control *io_wbc;
307 struct bio *io_bio;
308 ext4_io_end_t *io_end;
309 sector_t io_next_block;
310 };
311
312 /*
313 * Special inodes numbers
314 */
315 #define EXT4_BAD_INO 1 /* Bad blocks inode */
316 #define EXT4_ROOT_INO 2 /* Root inode */
317 #define EXT4_USR_QUOTA_INO 3 /* User quota inode */
318 #define EXT4_GRP_QUOTA_INO 4 /* Group quota inode */
319 #define EXT4_BOOT_LOADER_INO 5 /* Boot loader inode */
320 #define EXT4_UNDEL_DIR_INO 6 /* Undelete directory inode */
321 #define EXT4_RESIZE_INO 7 /* Reserved group descriptors inode */
322 #define EXT4_JOURNAL_INO 8 /* Journal inode */
323
324 /* First non-reserved inode for old ext4 filesystems */
325 #define EXT4_GOOD_OLD_FIRST_INO 11
326
327 /*
328 * Maximal count of links to a file
329 */
330 #define EXT4_LINK_MAX 65000
331
332 /*
333 * Macro-instructions used to manage several block sizes
334 */
335 #define EXT4_MIN_BLOCK_SIZE 1024
336 #define EXT4_MAX_BLOCK_SIZE 65536
337 #define EXT4_MIN_BLOCK_LOG_SIZE 10
338 #define EXT4_MAX_BLOCK_LOG_SIZE 16
339 #define EXT4_MAX_CLUSTER_LOG_SIZE 30
340 #ifdef __KERNEL__
341 # define EXT4_BLOCK_SIZE(s) ((s)->s_blocksize)
342 #else
343 # define EXT4_BLOCK_SIZE(s) (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
344 #endif
345 #define EXT4_ADDR_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
346 #define EXT4_CLUSTER_SIZE(s) (EXT4_BLOCK_SIZE(s) << \
347 EXT4_SB(s)->s_cluster_bits)
348 #ifdef __KERNEL__
349 # define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits)
350 # define EXT4_CLUSTER_BITS(s) (EXT4_SB(s)->s_cluster_bits)
351 #else
352 # define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_log_block_size + 10)
353 #endif
354 #ifdef __KERNEL__
355 #define EXT4_ADDR_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_addr_per_block_bits)
356 #define EXT4_INODE_SIZE(s) (EXT4_SB(s)->s_inode_size)
357 #define EXT4_FIRST_INO(s) (EXT4_SB(s)->s_first_ino)
358 #else
359 #define EXT4_INODE_SIZE(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
360 EXT4_GOOD_OLD_INODE_SIZE : \
361 (s)->s_inode_size)
362 #define EXT4_FIRST_INO(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
363 EXT4_GOOD_OLD_FIRST_INO : \
364 (s)->s_first_ino)
365 #endif
366 #define EXT4_BLOCK_ALIGN(size, blkbits) ALIGN((size), (1 << (blkbits)))
367 #define EXT4_MAX_BLOCKS(size, offset, blkbits) \
368 ((EXT4_BLOCK_ALIGN(size + offset, blkbits) >> blkbits) - (offset >> \
369 blkbits))
370
371 /* Translate a block number to a cluster number */
372 #define EXT4_B2C(sbi, blk) ((blk) >> (sbi)->s_cluster_bits)
373 /* Translate a cluster number to a block number */
374 #define EXT4_C2B(sbi, cluster) ((cluster) << (sbi)->s_cluster_bits)
375 /* Translate # of blks to # of clusters */
376 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
377 (sbi)->s_cluster_bits)
378 /* Mask out the low bits to get the starting block of the cluster */
379 #define EXT4_PBLK_CMASK(s, pblk) ((pblk) & \
380 ~((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
381 #define EXT4_LBLK_CMASK(s, lblk) ((lblk) & \
382 ~((ext4_lblk_t) (s)->s_cluster_ratio - 1))
383 /* Fill in the low bits to get the last block of the cluster */
384 #define EXT4_LBLK_CFILL(sbi, lblk) ((lblk) | \
385 ((ext4_lblk_t) (sbi)->s_cluster_ratio - 1))
386 /* Get the cluster offset */
387 #define EXT4_PBLK_COFF(s, pblk) ((pblk) & \
388 ((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
389 #define EXT4_LBLK_COFF(s, lblk) ((lblk) & \
390 ((ext4_lblk_t) (s)->s_cluster_ratio - 1))
391
392 /*
393 * Structure of a blocks group descriptor
394 */
395 struct ext4_group_desc
396 {
397 __le32 bg_block_bitmap_lo; /* Blocks bitmap block */
398 __le32 bg_inode_bitmap_lo; /* Inodes bitmap block */
399 __le32 bg_inode_table_lo; /* Inodes table block */
400 __le16 bg_free_blocks_count_lo;/* Free blocks count */
401 __le16 bg_free_inodes_count_lo;/* Free inodes count */
402 __le16 bg_used_dirs_count_lo; /* Directories count */
403 __le16 bg_flags; /* EXT4_BG_flags (INODE_UNINIT, etc) */
404 __le32 bg_exclude_bitmap_lo; /* Exclude bitmap for snapshots */
405 __le16 bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
406 __le16 bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
407 __le16 bg_itable_unused_lo; /* Unused inodes count */
408 __le16 bg_checksum; /* crc16(sb_uuid+group+desc) */
409 __le32 bg_block_bitmap_hi; /* Blocks bitmap block MSB */
410 __le32 bg_inode_bitmap_hi; /* Inodes bitmap block MSB */
411 __le32 bg_inode_table_hi; /* Inodes table block MSB */
412 __le16 bg_free_blocks_count_hi;/* Free blocks count MSB */
413 __le16 bg_free_inodes_count_hi;/* Free inodes count MSB */
414 __le16 bg_used_dirs_count_hi; /* Directories count MSB */
415 __le16 bg_itable_unused_hi; /* Unused inodes count MSB */
416 __le32 bg_exclude_bitmap_hi; /* Exclude bitmap block MSB */
417 __le16 bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
418 __le16 bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
419 __u32 bg_reserved;
420 };
421
422 #define EXT4_BG_INODE_BITMAP_CSUM_HI_END \
423 (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
424 sizeof(__le16))
425 #define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END \
426 (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
427 sizeof(__le16))
428
429 /*
430 * Structure of a flex block group info
431 */
432
433 struct flex_groups {
434 atomic64_t free_clusters;
435 atomic_t free_inodes;
436 atomic_t used_dirs;
437 };
438
439 #define EXT4_BG_INODE_UNINIT 0x0001 /* Inode table/bitmap not in use */
440 #define EXT4_BG_BLOCK_UNINIT 0x0002 /* Block bitmap not in use */
441 #define EXT4_BG_INODE_ZEROED 0x0004 /* On-disk itable initialized to zero */
442
443 /*
444 * Macro-instructions used to manage group descriptors
445 */
446 #define EXT4_MIN_DESC_SIZE 32
447 #define EXT4_MIN_DESC_SIZE_64BIT 64
448 #define EXT4_MAX_DESC_SIZE EXT4_MIN_BLOCK_SIZE
449 #define EXT4_DESC_SIZE(s) (EXT4_SB(s)->s_desc_size)
450 #ifdef __KERNEL__
451 # define EXT4_BLOCKS_PER_GROUP(s) (EXT4_SB(s)->s_blocks_per_group)
452 # define EXT4_CLUSTERS_PER_GROUP(s) (EXT4_SB(s)->s_clusters_per_group)
453 # define EXT4_DESC_PER_BLOCK(s) (EXT4_SB(s)->s_desc_per_block)
454 # define EXT4_INODES_PER_GROUP(s) (EXT4_SB(s)->s_inodes_per_group)
455 # define EXT4_DESC_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_desc_per_block_bits)
456 #else
457 # define EXT4_BLOCKS_PER_GROUP(s) ((s)->s_blocks_per_group)
458 # define EXT4_DESC_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
459 # define EXT4_INODES_PER_GROUP(s) ((s)->s_inodes_per_group)
460 #endif
461
462 /*
463 * Constants relative to the data blocks
464 */
465 #define EXT4_NDIR_BLOCKS 12
466 #define EXT4_IND_BLOCK EXT4_NDIR_BLOCKS
467 #define EXT4_DIND_BLOCK (EXT4_IND_BLOCK + 1)
468 #define EXT4_TIND_BLOCK (EXT4_DIND_BLOCK + 1)
469 #define EXT4_N_BLOCKS (EXT4_TIND_BLOCK + 1)
470
471 /*
472 * Inode flags
473 */
474 #define EXT4_SECRM_FL 0x00000001 /* Secure deletion */
475 #define EXT4_UNRM_FL 0x00000002 /* Undelete */
476 #define EXT4_COMPR_FL 0x00000004 /* Compress file */
477 #define EXT4_SYNC_FL 0x00000008 /* Synchronous updates */
478 #define EXT4_IMMUTABLE_FL 0x00000010 /* Immutable file */
479 #define EXT4_APPEND_FL 0x00000020 /* writes to file may only append */
480 #define EXT4_NODUMP_FL 0x00000040 /* do not dump file */
481 #define EXT4_NOATIME_FL 0x00000080 /* do not update atime */
482 /* Reserved for compression usage... */
483 #define EXT4_DIRTY_FL 0x00000100
484 #define EXT4_COMPRBLK_FL 0x00000200 /* One or more compressed clusters */
485 #define EXT4_NOCOMPR_FL 0x00000400 /* Don't compress */
486 /* nb: was previously EXT2_ECOMPR_FL */
487 #define EXT4_ENCRYPT_FL 0x00000800 /* encrypted file */
488 /* End compression flags --- maybe not all used */
489 #define EXT4_INDEX_FL 0x00001000 /* hash-indexed directory */
490 #define EXT4_IMAGIC_FL 0x00002000 /* AFS directory */
491 #define EXT4_JOURNAL_DATA_FL 0x00004000 /* file data should be journaled */
492 #define EXT4_NOTAIL_FL 0x00008000 /* file tail should not be merged */
493 #define EXT4_DIRSYNC_FL 0x00010000 /* dirsync behaviour (directories only) */
494 #define EXT4_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/
495 #define EXT4_HUGE_FILE_FL 0x00040000 /* Set to each huge file */
496 #define EXT4_EXTENTS_FL 0x00080000 /* Inode uses extents */
497 #define EXT4_VERITY_FL 0x00100000 /* Verity protected inode */
498 #define EXT4_EA_INODE_FL 0x00200000 /* Inode used for large EA */
499 /* 0x00400000 was formerly EXT4_EOFBLOCKS_FL */
500
501 #define EXT4_DAX_FL 0x02000000 /* Inode is DAX */
502
503 #define EXT4_INLINE_DATA_FL 0x10000000 /* Inode has inline data. */
504 #define EXT4_PROJINHERIT_FL 0x20000000 /* Create with parents projid */
505 #define EXT4_CASEFOLD_FL 0x40000000 /* Casefolded directory */
506 #define EXT4_RESERVED_FL 0x80000000 /* reserved for ext4 lib */
507
508 /* User modifiable flags */
509 #define EXT4_FL_USER_MODIFIABLE (EXT4_SECRM_FL | \
510 EXT4_UNRM_FL | \
511 EXT4_COMPR_FL | \
512 EXT4_SYNC_FL | \
513 EXT4_IMMUTABLE_FL | \
514 EXT4_APPEND_FL | \
515 EXT4_NODUMP_FL | \
516 EXT4_NOATIME_FL | \
517 EXT4_JOURNAL_DATA_FL | \
518 EXT4_NOTAIL_FL | \
519 EXT4_DIRSYNC_FL | \
520 EXT4_TOPDIR_FL | \
521 EXT4_EXTENTS_FL | \
522 0x00400000 /* EXT4_EOFBLOCKS_FL */ | \
523 EXT4_DAX_FL | \
524 EXT4_PROJINHERIT_FL | \
525 EXT4_CASEFOLD_FL)
526
527 /* User visible flags */
528 #define EXT4_FL_USER_VISIBLE (EXT4_FL_USER_MODIFIABLE | \
529 EXT4_DIRTY_FL | \
530 EXT4_COMPRBLK_FL | \
531 EXT4_NOCOMPR_FL | \
532 EXT4_ENCRYPT_FL | \
533 EXT4_INDEX_FL | \
534 EXT4_VERITY_FL | \
535 EXT4_INLINE_DATA_FL)
536
537 /* Flags that should be inherited by new inodes from their parent. */
538 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
539 EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
540 EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
541 EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL |\
542 EXT4_PROJINHERIT_FL | EXT4_CASEFOLD_FL |\
543 EXT4_DAX_FL)
544
545 /* Flags that are appropriate for regular files (all but dir-specific ones). */
546 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL | EXT4_CASEFOLD_FL |\
547 EXT4_PROJINHERIT_FL))
548
549 /* Flags that are appropriate for non-directories/regular files. */
550 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
551
552 /* The only flags that should be swapped */
553 #define EXT4_FL_SHOULD_SWAP (EXT4_HUGE_FILE_FL | EXT4_EXTENTS_FL)
554
555 /* Flags which are mutually exclusive to DAX */
556 #define EXT4_DAX_MUT_EXCL (EXT4_VERITY_FL | EXT4_ENCRYPT_FL |\
557 EXT4_JOURNAL_DATA_FL | EXT4_INLINE_DATA_FL)
558
559 /* Mask out flags that are inappropriate for the given type of inode. */
560 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
561 {
562 if (S_ISDIR(mode))
563 return flags;
564 else if (S_ISREG(mode))
565 return flags & EXT4_REG_FLMASK;
566 else
567 return flags & EXT4_OTHER_FLMASK;
568 }
569
570 /*
571 * Inode flags used for atomic set/get
572 */
573 enum {
574 EXT4_INODE_SECRM = 0, /* Secure deletion */
575 EXT4_INODE_UNRM = 1, /* Undelete */
576 EXT4_INODE_COMPR = 2, /* Compress file */
577 EXT4_INODE_SYNC = 3, /* Synchronous updates */
578 EXT4_INODE_IMMUTABLE = 4, /* Immutable file */
579 EXT4_INODE_APPEND = 5, /* writes to file may only append */
580 EXT4_INODE_NODUMP = 6, /* do not dump file */
581 EXT4_INODE_NOATIME = 7, /* do not update atime */
582 /* Reserved for compression usage... */
583 EXT4_INODE_DIRTY = 8,
584 EXT4_INODE_COMPRBLK = 9, /* One or more compressed clusters */
585 EXT4_INODE_NOCOMPR = 10, /* Don't compress */
586 EXT4_INODE_ENCRYPT = 11, /* Encrypted file */
587 /* End compression flags --- maybe not all used */
588 EXT4_INODE_INDEX = 12, /* hash-indexed directory */
589 EXT4_INODE_IMAGIC = 13, /* AFS directory */
590 EXT4_INODE_JOURNAL_DATA = 14, /* file data should be journaled */
591 EXT4_INODE_NOTAIL = 15, /* file tail should not be merged */
592 EXT4_INODE_DIRSYNC = 16, /* dirsync behaviour (directories only) */
593 EXT4_INODE_TOPDIR = 17, /* Top of directory hierarchies*/
594 EXT4_INODE_HUGE_FILE = 18, /* Set to each huge file */
595 EXT4_INODE_EXTENTS = 19, /* Inode uses extents */
596 EXT4_INODE_VERITY = 20, /* Verity protected inode */
597 EXT4_INODE_EA_INODE = 21, /* Inode used for large EA */
598 /* 22 was formerly EXT4_INODE_EOFBLOCKS */
599 EXT4_INODE_DAX = 25, /* Inode is DAX */
600 EXT4_INODE_INLINE_DATA = 28, /* Data in inode. */
601 EXT4_INODE_PROJINHERIT = 29, /* Create with parents projid */
602 EXT4_INODE_CASEFOLD = 30, /* Casefolded directory */
603 EXT4_INODE_RESERVED = 31, /* reserved for ext4 lib */
604 };
605
606 /*
607 * Since it's pretty easy to mix up bit numbers and hex values, we use a
608 * build-time check to make sure that EXT4_XXX_FL is consistent with respect to
609 * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost
610 * any extra space in the compiled kernel image, otherwise, the build will fail.
611 * It's important that these values are the same, since we are using
612 * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent
613 * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk
614 * values found in ext2, ext3 and ext4 filesystems, and of course the values
615 * defined in e2fsprogs.
616 *
617 * It's not paranoia if the Murphy's Law really *is* out to get you. :-)
618 */
619 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1U << EXT4_INODE_##FLAG))
620 #define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG))
621
622 static inline void ext4_check_flag_values(void)
623 {
624 CHECK_FLAG_VALUE(SECRM);
625 CHECK_FLAG_VALUE(UNRM);
626 CHECK_FLAG_VALUE(COMPR);
627 CHECK_FLAG_VALUE(SYNC);
628 CHECK_FLAG_VALUE(IMMUTABLE);
629 CHECK_FLAG_VALUE(APPEND);
630 CHECK_FLAG_VALUE(NODUMP);
631 CHECK_FLAG_VALUE(NOATIME);
632 CHECK_FLAG_VALUE(DIRTY);
633 CHECK_FLAG_VALUE(COMPRBLK);
634 CHECK_FLAG_VALUE(NOCOMPR);
635 CHECK_FLAG_VALUE(ENCRYPT);
636 CHECK_FLAG_VALUE(INDEX);
637 CHECK_FLAG_VALUE(IMAGIC);
638 CHECK_FLAG_VALUE(JOURNAL_DATA);
639 CHECK_FLAG_VALUE(NOTAIL);
640 CHECK_FLAG_VALUE(DIRSYNC);
641 CHECK_FLAG_VALUE(TOPDIR);
642 CHECK_FLAG_VALUE(HUGE_FILE);
643 CHECK_FLAG_VALUE(EXTENTS);
644 CHECK_FLAG_VALUE(VERITY);
645 CHECK_FLAG_VALUE(EA_INODE);
646 CHECK_FLAG_VALUE(INLINE_DATA);
647 CHECK_FLAG_VALUE(PROJINHERIT);
648 CHECK_FLAG_VALUE(CASEFOLD);
649 CHECK_FLAG_VALUE(RESERVED);
650 }
651
652 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
653 struct compat_ext4_new_group_input {
654 u32 group;
655 compat_u64 block_bitmap;
656 compat_u64 inode_bitmap;
657 compat_u64 inode_table;
658 u32 blocks_count;
659 u16 reserved_blocks;
660 u16 unused;
661 };
662 #endif
663
664 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
665 struct ext4_new_group_data {
666 __u32 group;
667 __u64 block_bitmap;
668 __u64 inode_bitmap;
669 __u64 inode_table;
670 __u32 blocks_count;
671 __u16 reserved_blocks;
672 __u16 mdata_blocks;
673 __u32 free_clusters_count;
674 };
675
676 /* Indexes used to index group tables in ext4_new_group_data */
677 enum {
678 BLOCK_BITMAP = 0, /* block bitmap */
679 INODE_BITMAP, /* inode bitmap */
680 INODE_TABLE, /* inode tables */
681 GROUP_TABLE_COUNT,
682 };
683
684 /*
685 * Flags used by ext4_map_blocks()
686 */
687 /* Allocate any needed blocks and/or convert an unwritten
688 extent to be an initialized ext4 */
689 #define EXT4_GET_BLOCKS_CREATE 0x0001
690 /* Request the creation of an unwritten extent */
691 #define EXT4_GET_BLOCKS_UNWRIT_EXT 0x0002
692 #define EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT (EXT4_GET_BLOCKS_UNWRIT_EXT|\
693 EXT4_GET_BLOCKS_CREATE)
694 /* Caller is from the delayed allocation writeout path
695 * finally doing the actual allocation of delayed blocks */
696 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE 0x0004
697 /* caller is from the direct IO path, request to creation of an
698 unwritten extents if not allocated, split the unwritten
699 extent if blocks has been preallocated already*/
700 #define EXT4_GET_BLOCKS_PRE_IO 0x0008
701 #define EXT4_GET_BLOCKS_CONVERT 0x0010
702 #define EXT4_GET_BLOCKS_IO_CREATE_EXT (EXT4_GET_BLOCKS_PRE_IO|\
703 EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
704 /* Convert extent to initialized after IO complete */
705 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT (EXT4_GET_BLOCKS_CONVERT|\
706 EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
707 /* Eventual metadata allocation (due to growing extent tree)
708 * should not fail, so try to use reserved blocks for that.*/
709 #define EXT4_GET_BLOCKS_METADATA_NOFAIL 0x0020
710 /* Don't normalize allocation size (used for fallocate) */
711 #define EXT4_GET_BLOCKS_NO_NORMALIZE 0x0040
712 /* Convert written extents to unwritten */
713 #define EXT4_GET_BLOCKS_CONVERT_UNWRITTEN 0x0100
714 /* Write zeros to newly created written extents */
715 #define EXT4_GET_BLOCKS_ZERO 0x0200
716 #define EXT4_GET_BLOCKS_CREATE_ZERO (EXT4_GET_BLOCKS_CREATE |\
717 EXT4_GET_BLOCKS_ZERO)
718 /* Caller will submit data before dropping transaction handle. This
719 * allows jbd2 to avoid submitting data before commit. */
720 #define EXT4_GET_BLOCKS_IO_SUBMIT 0x0400
721 /* Caller is in the atomic contex, find extent if it has been cached */
722 #define EXT4_GET_BLOCKS_CACHED_NOWAIT 0x0800
723
724 /*
725 * The bit position of these flags must not overlap with any of the
726 * EXT4_GET_BLOCKS_*. They are used by ext4_find_extent(),
727 * read_extent_tree_block(), ext4_split_extent_at(),
728 * ext4_ext_insert_extent(), and ext4_ext_create_new_leaf().
729 * EXT4_EX_NOCACHE is used to indicate that the we shouldn't be
730 * caching the extents when reading from the extent tree while a
731 * truncate or punch hole operation is in progress.
732 */
733 #define EXT4_EX_NOCACHE 0x40000000
734 #define EXT4_EX_FORCE_CACHE 0x20000000
735 #define EXT4_EX_NOFAIL 0x10000000
736
737 /*
738 * Flags used by ext4_free_blocks
739 */
740 #define EXT4_FREE_BLOCKS_METADATA 0x0001
741 #define EXT4_FREE_BLOCKS_FORGET 0x0002
742 #define EXT4_FREE_BLOCKS_VALIDATED 0x0004
743 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
744 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER 0x0010
745 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER 0x0020
746 #define EXT4_FREE_BLOCKS_RERESERVE_CLUSTER 0x0040
747
748 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
749 /*
750 * ioctl commands in 32 bit emulation
751 */
752 #define EXT4_IOC32_GETVERSION _IOR('f', 3, int)
753 #define EXT4_IOC32_SETVERSION _IOW('f', 4, int)
754 #define EXT4_IOC32_GETRSVSZ _IOR('f', 5, int)
755 #define EXT4_IOC32_SETRSVSZ _IOW('f', 6, int)
756 #define EXT4_IOC32_GROUP_EXTEND _IOW('f', 7, unsigned int)
757 #define EXT4_IOC32_GROUP_ADD _IOW('f', 8, struct compat_ext4_new_group_input)
758 #define EXT4_IOC32_GETVERSION_OLD FS_IOC32_GETVERSION
759 #define EXT4_IOC32_SETVERSION_OLD FS_IOC32_SETVERSION
760 #endif
761
762 /* Max physical block we can address w/o extents */
763 #define EXT4_MAX_BLOCK_FILE_PHYS 0xFFFFFFFF
764
765 /* Max logical block we can support */
766 #define EXT4_MAX_LOGICAL_BLOCK 0xFFFFFFFE
767
768 /*
769 * Structure of an inode on the disk
770 */
771 struct ext4_inode {
772 __le16 i_mode; /* File mode */
773 __le16 i_uid; /* Low 16 bits of Owner Uid */
774 __le32 i_size_lo; /* Size in bytes */
775 __le32 i_atime; /* Access time */
776 __le32 i_ctime; /* Inode Change time */
777 __le32 i_mtime; /* Modification time */
778 __le32 i_dtime; /* Deletion Time */
779 __le16 i_gid; /* Low 16 bits of Group Id */
780 __le16 i_links_count; /* Links count */
781 __le32 i_blocks_lo; /* Blocks count */
782 __le32 i_flags; /* File flags */
783 union {
784 struct {
785 __le32 l_i_version;
786 } linux1;
787 struct {
788 __u32 h_i_translator;
789 } hurd1;
790 struct {
791 __u32 m_i_reserved1;
792 } masix1;
793 } osd1; /* OS dependent 1 */
794 __le32 i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
795 __le32 i_generation; /* File version (for NFS) */
796 __le32 i_file_acl_lo; /* File ACL */
797 __le32 i_size_high;
798 __le32 i_obso_faddr; /* Obsoleted fragment address */
799 union {
800 struct {
801 __le16 l_i_blocks_high; /* were l_i_reserved1 */
802 __le16 l_i_file_acl_high;
803 __le16 l_i_uid_high; /* these 2 fields */
804 __le16 l_i_gid_high; /* were reserved2[0] */
805 __le16 l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
806 __le16 l_i_reserved;
807 } linux2;
808 struct {
809 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */
810 __u16 h_i_mode_high;
811 __u16 h_i_uid_high;
812 __u16 h_i_gid_high;
813 __u32 h_i_author;
814 } hurd2;
815 struct {
816 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */
817 __le16 m_i_file_acl_high;
818 __u32 m_i_reserved2[2];
819 } masix2;
820 } osd2; /* OS dependent 2 */
821 __le16 i_extra_isize;
822 __le16 i_checksum_hi; /* crc32c(uuid+inum+inode) BE */
823 __le32 i_ctime_extra; /* extra Change time (nsec << 2 | epoch) */
824 __le32 i_mtime_extra; /* extra Modification time(nsec << 2 | epoch) */
825 __le32 i_atime_extra; /* extra Access time (nsec << 2 | epoch) */
826 __le32 i_crtime; /* File Creation time */
827 __le32 i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
828 __le32 i_version_hi; /* high 32 bits for 64-bit version */
829 __le32 i_projid; /* Project ID */
830 };
831
832 #define EXT4_EPOCH_BITS 2
833 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
834 #define EXT4_NSEC_MASK (~0UL << EXT4_EPOCH_BITS)
835
836 /*
837 * Extended fields will fit into an inode if the filesystem was formatted
838 * with large inodes (-I 256 or larger) and there are not currently any EAs
839 * consuming all of the available space. For new inodes we always reserve
840 * enough space for the kernel's known extended fields, but for inodes
841 * created with an old kernel this might not have been the case. None of
842 * the extended inode fields is critical for correct filesystem operation.
843 * This macro checks if a certain field fits in the inode. Note that
844 * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
845 */
846 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field) \
847 ((offsetof(typeof(*ext4_inode), field) + \
848 sizeof((ext4_inode)->field)) \
849 <= (EXT4_GOOD_OLD_INODE_SIZE + \
850 (einode)->i_extra_isize)) \
851
852 /*
853 * We use an encoding that preserves the times for extra epoch "00":
854 *
855 * extra msb of adjust for signed
856 * epoch 32-bit 32-bit tv_sec to
857 * bits time decoded 64-bit tv_sec 64-bit tv_sec valid time range
858 * 0 0 1 -0x80000000..-0x00000001 0x000000000 1901-12-13..1969-12-31
859 * 0 0 0 0x000000000..0x07fffffff 0x000000000 1970-01-01..2038-01-19
860 * 0 1 1 0x080000000..0x0ffffffff 0x100000000 2038-01-19..2106-02-07
861 * 0 1 0 0x100000000..0x17fffffff 0x100000000 2106-02-07..2174-02-25
862 * 1 0 1 0x180000000..0x1ffffffff 0x200000000 2174-02-25..2242-03-16
863 * 1 0 0 0x200000000..0x27fffffff 0x200000000 2242-03-16..2310-04-04
864 * 1 1 1 0x280000000..0x2ffffffff 0x300000000 2310-04-04..2378-04-22
865 * 1 1 0 0x300000000..0x37fffffff 0x300000000 2378-04-22..2446-05-10
866 *
867 * Note that previous versions of the kernel on 64-bit systems would
868 * incorrectly use extra epoch bits 1,1 for dates between 1901 and
869 * 1970. e2fsck will correct this, assuming that it is run on the
870 * affected filesystem before 2242.
871 */
872
873 static inline __le32 ext4_encode_extra_time(struct timespec64 ts)
874 {
875 u32 extra = ((ts.tv_sec - (s32)ts.tv_sec) >> 32) & EXT4_EPOCH_MASK;
876 return cpu_to_le32(extra | (ts.tv_nsec << EXT4_EPOCH_BITS));
877 }
878
879 static inline struct timespec64 ext4_decode_extra_time(__le32 base,
880 __le32 extra)
881 {
882 struct timespec64 ts = { .tv_sec = (signed)le32_to_cpu(base) };
883
884 if (unlikely(extra & cpu_to_le32(EXT4_EPOCH_MASK)))
885 ts.tv_sec += (u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK) << 32;
886 ts.tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
887 return ts;
888 }
889
890 #define EXT4_INODE_SET_XTIME_VAL(xtime, inode, raw_inode, ts) \
891 do { \
892 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) { \
893 (raw_inode)->xtime = cpu_to_le32((ts).tv_sec); \
894 (raw_inode)->xtime ## _extra = ext4_encode_extra_time(ts); \
895 } else \
896 (raw_inode)->xtime = cpu_to_le32(clamp_t(int32_t, (ts).tv_sec, S32_MIN, S32_MAX)); \
897 } while (0)
898
899 #define EXT4_INODE_SET_ATIME(inode, raw_inode) \
900 EXT4_INODE_SET_XTIME_VAL(i_atime, inode, raw_inode, inode_get_atime(inode))
901
902 #define EXT4_INODE_SET_MTIME(inode, raw_inode) \
903 EXT4_INODE_SET_XTIME_VAL(i_mtime, inode, raw_inode, inode_get_mtime(inode))
904
905 #define EXT4_INODE_SET_CTIME(inode, raw_inode) \
906 EXT4_INODE_SET_XTIME_VAL(i_ctime, inode, raw_inode, inode_get_ctime(inode))
907
908 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode) \
909 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \
910 EXT4_INODE_SET_XTIME_VAL(xtime, &((einode)->vfs_inode), \
911 raw_inode, (einode)->xtime)
912
913 #define EXT4_INODE_GET_XTIME_VAL(xtime, inode, raw_inode) \
914 (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra) ? \
915 ext4_decode_extra_time((raw_inode)->xtime, \
916 (raw_inode)->xtime ## _extra) : \
917 (struct timespec64) { \
918 .tv_sec = (signed)le32_to_cpu((raw_inode)->xtime) \
919 })
920
921 #define EXT4_INODE_GET_ATIME(inode, raw_inode) \
922 do { \
923 inode_set_atime_to_ts(inode, \
924 EXT4_INODE_GET_XTIME_VAL(i_atime, inode, raw_inode)); \
925 } while (0)
926
927 #define EXT4_INODE_GET_MTIME(inode, raw_inode) \
928 do { \
929 inode_set_mtime_to_ts(inode, \
930 EXT4_INODE_GET_XTIME_VAL(i_mtime, inode, raw_inode)); \
931 } while (0)
932
933 #define EXT4_INODE_GET_CTIME(inode, raw_inode) \
934 do { \
935 inode_set_ctime_to_ts(inode, \
936 EXT4_INODE_GET_XTIME_VAL(i_ctime, inode, raw_inode)); \
937 } while (0)
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 = \
943 EXT4_INODE_GET_XTIME_VAL(xtime, &(einode->vfs_inode), \
944 raw_inode); \
945 else \
946 (einode)->xtime = (struct timespec64){0, 0}; \
947 } while (0)
948
949 #define i_disk_version osd1.linux1.l_i_version
950
951 #if defined(__KERNEL__) || defined(__linux__)
952 #define i_reserved1 osd1.linux1.l_i_reserved1
953 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
954 #define i_blocks_high osd2.linux2.l_i_blocks_high
955 #define i_uid_low i_uid
956 #define i_gid_low i_gid
957 #define i_uid_high osd2.linux2.l_i_uid_high
958 #define i_gid_high osd2.linux2.l_i_gid_high
959 #define i_checksum_lo osd2.linux2.l_i_checksum_lo
960
961 #elif defined(__GNU__)
962
963 #define i_translator osd1.hurd1.h_i_translator
964 #define i_uid_high osd2.hurd2.h_i_uid_high
965 #define i_gid_high osd2.hurd2.h_i_gid_high
966 #define i_author osd2.hurd2.h_i_author
967
968 #elif defined(__masix__)
969
970 #define i_reserved1 osd1.masix1.m_i_reserved1
971 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
972 #define i_reserved2 osd2.masix2.m_i_reserved2
973
974 #endif /* defined(__KERNEL__) || defined(__linux__) */
975
976 #include "extents_status.h"
977 #include "fast_commit.h"
978
979 /*
980 * Lock subclasses for i_data_sem in the ext4_inode_info structure.
981 *
982 * These are needed to avoid lockdep false positives when we need to
983 * allocate blocks to the quota inode during ext4_map_blocks(), while
984 * holding i_data_sem for a normal (non-quota) inode. Since we don't
985 * do quota tracking for the quota inode, this avoids deadlock (as
986 * well as infinite recursion, since it isn't turtles all the way
987 * down...)
988 *
989 * I_DATA_SEM_NORMAL - Used for most inodes
990 * I_DATA_SEM_OTHER - Used by move_inode.c for the second normal inode
991 * where the second inode has larger inode number
992 * than the first
993 * I_DATA_SEM_QUOTA - Used for quota inodes only
994 * I_DATA_SEM_EA - Used for ea_inodes only
995 */
996 enum {
997 I_DATA_SEM_NORMAL = 0,
998 I_DATA_SEM_OTHER,
999 I_DATA_SEM_QUOTA,
1000 I_DATA_SEM_EA
1001 };
1002
1003
1004 /*
1005 * fourth extended file system inode data in memory
1006 */
1007 struct ext4_inode_info {
1008 __le32 i_data[15]; /* unconverted */
1009 __u32 i_dtime;
1010 ext4_fsblk_t i_file_acl;
1011
1012 /*
1013 * i_block_group is the number of the block group which contains
1014 * this file's inode. Constant across the lifetime of the inode,
1015 * it is used for making block allocation decisions - we try to
1016 * place a file's data blocks near its inode block, and new inodes
1017 * near to their parent directory's inode.
1018 */
1019 ext4_group_t i_block_group;
1020 ext4_lblk_t i_dir_start_lookup;
1021 #if (BITS_PER_LONG < 64)
1022 unsigned long i_state_flags; /* Dynamic state flags */
1023 #endif
1024 unsigned long i_flags;
1025
1026 /*
1027 * Extended attributes can be read independently of the main file
1028 * data. Taking i_rwsem even when reading would cause contention
1029 * between readers of EAs and writers of regular file data, so
1030 * instead we synchronize on xattr_sem when reading or changing
1031 * EAs.
1032 */
1033 struct rw_semaphore xattr_sem;
1034
1035 /*
1036 * Inodes with EXT4_STATE_ORPHAN_FILE use i_orphan_idx. Otherwise
1037 * i_orphan is used.
1038 */
1039 union {
1040 struct list_head i_orphan; /* unlinked but open inodes */
1041 unsigned int i_orphan_idx; /* Index in orphan file */
1042 };
1043
1044 /* Fast commit related info */
1045
1046 /* For tracking dentry create updates */
1047 struct list_head i_fc_dilist;
1048 struct list_head i_fc_list; /*
1049 * inodes that need fast commit
1050 * protected by sbi->s_fc_lock.
1051 */
1052
1053 /* Start of lblk range that needs to be committed in this fast commit */
1054 ext4_lblk_t i_fc_lblk_start;
1055
1056 /* End of lblk range that needs to be committed in this fast commit */
1057 ext4_lblk_t i_fc_lblk_len;
1058
1059 /* Number of ongoing updates on this inode */
1060 atomic_t i_fc_updates;
1061
1062 /* Fast commit wait queue for this inode */
1063 wait_queue_head_t i_fc_wait;
1064
1065 /* Protect concurrent accesses on i_fc_lblk_start, i_fc_lblk_len */
1066 struct mutex i_fc_lock;
1067
1068 /*
1069 * i_disksize keeps track of what the inode size is ON DISK, not
1070 * in memory. During truncate, i_size is set to the new size by
1071 * the VFS prior to calling ext4_truncate(), but the filesystem won't
1072 * set i_disksize to 0 until the truncate is actually under way.
1073 *
1074 * The intent is that i_disksize always represents the blocks which
1075 * are used by this file. This allows recovery to restart truncate
1076 * on orphans if we crash during truncate. We actually write i_disksize
1077 * into the on-disk inode when writing inodes out, instead of i_size.
1078 *
1079 * The only time when i_disksize and i_size may be different is when
1080 * a truncate is in progress. The only things which change i_disksize
1081 * are ext4_get_block (growth) and ext4_truncate (shrinkth).
1082 */
1083 loff_t i_disksize;
1084
1085 /*
1086 * i_data_sem is for serialising ext4_truncate() against
1087 * ext4_getblock(). In the 2.4 ext2 design, great chunks of inode's
1088 * data tree are chopped off during truncate. We can't do that in
1089 * ext4 because whenever we perform intermediate commits during
1090 * truncate, the inode and all the metadata blocks *must* be in a
1091 * consistent state which allows truncation of the orphans to restart
1092 * during recovery. Hence we must fix the get_block-vs-truncate race
1093 * by other means, so we have i_data_sem.
1094 */
1095 struct rw_semaphore i_data_sem;
1096 struct inode vfs_inode;
1097 struct jbd2_inode *jinode;
1098
1099 spinlock_t i_raw_lock; /* protects updates to the raw inode */
1100
1101 /*
1102 * File creation time. Its function is same as that of
1103 * struct timespec64 i_{a,c,m}time in the generic inode.
1104 */
1105 struct timespec64 i_crtime;
1106
1107 /* mballoc */
1108 atomic_t i_prealloc_active;
1109 struct rb_root i_prealloc_node;
1110 rwlock_t i_prealloc_lock;
1111
1112 /* extents status tree */
1113 struct ext4_es_tree i_es_tree;
1114 rwlock_t i_es_lock;
1115 struct list_head i_es_list;
1116 unsigned int i_es_all_nr; /* protected by i_es_lock */
1117 unsigned int i_es_shk_nr; /* protected by i_es_lock */
1118 ext4_lblk_t i_es_shrink_lblk; /* Offset where we start searching for
1119 extents to shrink. Protected by
1120 i_es_lock */
1121
1122 /* ialloc */
1123 ext4_group_t i_last_alloc_group;
1124
1125 /* allocation reservation info for delalloc */
1126 /* In case of bigalloc, this refer to clusters rather than blocks */
1127 unsigned int i_reserved_data_blocks;
1128
1129 /* pending cluster reservations for bigalloc file systems */
1130 struct ext4_pending_tree i_pending_tree;
1131
1132 /* on-disk additional length */
1133 __u16 i_extra_isize;
1134
1135 /* Indicate the inline data space. */
1136 u16 i_inline_off;
1137 u16 i_inline_size;
1138
1139 #ifdef CONFIG_QUOTA
1140 /* quota space reservation, managed internally by quota code */
1141 qsize_t i_reserved_quota;
1142 #endif
1143
1144 /* Lock protecting lists below */
1145 spinlock_t i_completed_io_lock;
1146 /*
1147 * Completed IOs that need unwritten extents handling and have
1148 * transaction reserved
1149 */
1150 struct list_head i_rsv_conversion_list;
1151 struct work_struct i_rsv_conversion_work;
1152 atomic_t i_unwritten; /* Nr. of inflight conversions pending */
1153
1154 spinlock_t i_block_reservation_lock;
1155
1156 /*
1157 * Transactions that contain inode's metadata needed to complete
1158 * fsync and fdatasync, respectively.
1159 */
1160 tid_t i_sync_tid;
1161 tid_t i_datasync_tid;
1162
1163 #ifdef CONFIG_QUOTA
1164 struct dquot __rcu *i_dquot[MAXQUOTAS];
1165 #endif
1166
1167 /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
1168 __u32 i_csum_seed;
1169
1170 kprojid_t i_projid;
1171 };
1172
1173 /*
1174 * File system states
1175 */
1176 #define EXT4_VALID_FS 0x0001 /* Unmounted cleanly */
1177 #define EXT4_ERROR_FS 0x0002 /* Errors detected */
1178 #define EXT4_ORPHAN_FS 0x0004 /* Orphans being recovered */
1179 #define EXT4_FC_REPLAY 0x0020 /* Fast commit replay ongoing */
1180
1181 /*
1182 * Misc. filesystem flags
1183 */
1184 #define EXT2_FLAGS_SIGNED_HASH 0x0001 /* Signed dirhash in use */
1185 #define EXT2_FLAGS_UNSIGNED_HASH 0x0002 /* Unsigned dirhash in use */
1186 #define EXT2_FLAGS_TEST_FILESYS 0x0004 /* to test development code */
1187
1188 /*
1189 * Mount flags set via mount options or defaults
1190 */
1191 #define EXT4_MOUNT_NO_MBCACHE 0x00001 /* Do not use mbcache */
1192 #define EXT4_MOUNT_GRPID 0x00004 /* Create files with directory's group */
1193 #define EXT4_MOUNT_DEBUG 0x00008 /* Some debugging messages */
1194 #define EXT4_MOUNT_ERRORS_CONT 0x00010 /* Continue on errors */
1195 #define EXT4_MOUNT_ERRORS_RO 0x00020 /* Remount fs ro on errors */
1196 #define EXT4_MOUNT_ERRORS_PANIC 0x00040 /* Panic on errors */
1197 #define EXT4_MOUNT_ERRORS_MASK 0x00070
1198 #define EXT4_MOUNT_MINIX_DF 0x00080 /* Mimics the Minix statfs */
1199 #define EXT4_MOUNT_NOLOAD 0x00100 /* Don't use existing journal*/
1200 #ifdef CONFIG_FS_DAX
1201 #define EXT4_MOUNT_DAX_ALWAYS 0x00200 /* Direct Access */
1202 #else
1203 #define EXT4_MOUNT_DAX_ALWAYS 0
1204 #endif
1205 #define EXT4_MOUNT_DATA_FLAGS 0x00C00 /* Mode for data writes: */
1206 #define EXT4_MOUNT_JOURNAL_DATA 0x00400 /* Write data to journal */
1207 #define EXT4_MOUNT_ORDERED_DATA 0x00800 /* Flush data before commit */
1208 #define EXT4_MOUNT_WRITEBACK_DATA 0x00C00 /* No data ordering */
1209 #define EXT4_MOUNT_UPDATE_JOURNAL 0x01000 /* Update the journal format */
1210 #define EXT4_MOUNT_NO_UID32 0x02000 /* Disable 32-bit UIDs */
1211 #define EXT4_MOUNT_XATTR_USER 0x04000 /* Extended user attributes */
1212 #define EXT4_MOUNT_POSIX_ACL 0x08000 /* POSIX Access Control Lists */
1213 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC 0x10000 /* No auto delalloc mapping */
1214 #define EXT4_MOUNT_BARRIER 0x20000 /* Use block barriers */
1215 #define EXT4_MOUNT_QUOTA 0x40000 /* Some quota option set */
1216 #define EXT4_MOUNT_USRQUOTA 0x80000 /* "old" user quota,
1217 * enable enforcement for hidden
1218 * quota files */
1219 #define EXT4_MOUNT_GRPQUOTA 0x100000 /* "old" group quota, enable
1220 * enforcement for hidden quota
1221 * files */
1222 #define EXT4_MOUNT_PRJQUOTA 0x200000 /* Enable project quota
1223 * enforcement */
1224 #define EXT4_MOUNT_DIOREAD_NOLOCK 0x400000 /* Enable support for dio read nolocking */
1225 #define EXT4_MOUNT_JOURNAL_CHECKSUM 0x800000 /* Journal checksums */
1226 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
1227 #define EXT4_MOUNT_WARN_ON_ERROR 0x2000000 /* Trigger WARN_ON on error */
1228 #define EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS 0x4000000
1229 #define EXT4_MOUNT_DELALLOC 0x8000000 /* Delalloc support */
1230 #define EXT4_MOUNT_DATA_ERR_ABORT 0x10000000 /* Abort on file data write */
1231 #define EXT4_MOUNT_BLOCK_VALIDITY 0x20000000 /* Block validity checking */
1232 #define EXT4_MOUNT_DISCARD 0x40000000 /* Issue DISCARD requests */
1233 #define EXT4_MOUNT_INIT_INODE_TABLE 0x80000000 /* Initialize uninitialized itables */
1234
1235 /*
1236 * Mount flags set either automatically (could not be set by mount option)
1237 * based on per file system feature or property or in special cases such as
1238 * distinguishing between explicit mount option definition and default.
1239 */
1240 #define EXT4_MOUNT2_EXPLICIT_DELALLOC 0x00000001 /* User explicitly
1241 specified delalloc */
1242 #define EXT4_MOUNT2_STD_GROUP_SIZE 0x00000002 /* We have standard group
1243 size of blocksize * 8
1244 blocks */
1245 #define EXT4_MOUNT2_HURD_COMPAT 0x00000004 /* Support HURD-castrated
1246 file systems */
1247 #define EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM 0x00000008 /* User explicitly
1248 specified journal checksum */
1249
1250 #define EXT4_MOUNT2_JOURNAL_FAST_COMMIT 0x00000010 /* Journal fast commit */
1251 #define EXT4_MOUNT2_DAX_NEVER 0x00000020 /* Do not allow Direct Access */
1252 #define EXT4_MOUNT2_DAX_INODE 0x00000040 /* For printing options only */
1253 #define EXT4_MOUNT2_MB_OPTIMIZE_SCAN 0x00000080 /* Optimize group
1254 * scanning in mballoc
1255 */
1256 #define EXT4_MOUNT2_ABORT 0x00000100 /* Abort filesystem */
1257
1258 #define clear_opt(sb, opt) EXT4_SB(sb)->s_mount_opt &= \
1259 ~EXT4_MOUNT_##opt
1260 #define set_opt(sb, opt) EXT4_SB(sb)->s_mount_opt |= \
1261 EXT4_MOUNT_##opt
1262 #define test_opt(sb, opt) (EXT4_SB(sb)->s_mount_opt & \
1263 EXT4_MOUNT_##opt)
1264
1265 #define clear_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 &= \
1266 ~EXT4_MOUNT2_##opt
1267 #define set_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 |= \
1268 EXT4_MOUNT2_##opt
1269 #define test_opt2(sb, opt) (EXT4_SB(sb)->s_mount_opt2 & \
1270 EXT4_MOUNT2_##opt)
1271
1272 #define ext4_test_and_set_bit __test_and_set_bit_le
1273 #define ext4_set_bit __set_bit_le
1274 #define ext4_test_and_clear_bit __test_and_clear_bit_le
1275 #define ext4_clear_bit __clear_bit_le
1276 #define ext4_test_bit test_bit_le
1277 #define ext4_find_next_zero_bit find_next_zero_bit_le
1278 #define ext4_find_next_bit find_next_bit_le
1279
1280 extern void mb_set_bits(void *bm, int cur, int len);
1281
1282 /*
1283 * Maximal mount counts between two filesystem checks
1284 */
1285 #define EXT4_DFL_MAX_MNT_COUNT 20 /* Allow 20 mounts */
1286 #define EXT4_DFL_CHECKINTERVAL 0 /* Don't use interval check */
1287
1288 /*
1289 * Behaviour when detecting errors
1290 */
1291 #define EXT4_ERRORS_CONTINUE 1 /* Continue execution */
1292 #define EXT4_ERRORS_RO 2 /* Remount fs read-only */
1293 #define EXT4_ERRORS_PANIC 3 /* Panic */
1294 #define EXT4_ERRORS_DEFAULT EXT4_ERRORS_CONTINUE
1295
1296 /* Metadata checksum algorithm codes */
1297 #define EXT4_CRC32C_CHKSUM 1
1298
1299 #define EXT4_LABEL_MAX 16
1300
1301 /*
1302 * Structure of the super block
1303 */
1304 struct ext4_super_block {
1305 /*00*/ __le32 s_inodes_count; /* Inodes count */
1306 __le32 s_blocks_count_lo; /* Blocks count */
1307 __le32 s_r_blocks_count_lo; /* Reserved blocks count */
1308 __le32 s_free_blocks_count_lo; /* Free blocks count */
1309 /*10*/ __le32 s_free_inodes_count; /* Free inodes count */
1310 __le32 s_first_data_block; /* First Data Block */
1311 __le32 s_log_block_size; /* Block size */
1312 __le32 s_log_cluster_size; /* Allocation cluster size */
1313 /*20*/ __le32 s_blocks_per_group; /* # Blocks per group */
1314 __le32 s_clusters_per_group; /* # Clusters per group */
1315 __le32 s_inodes_per_group; /* # Inodes per group */
1316 __le32 s_mtime; /* Mount time */
1317 /*30*/ __le32 s_wtime; /* Write time */
1318 __le16 s_mnt_count; /* Mount count */
1319 __le16 s_max_mnt_count; /* Maximal mount count */
1320 __le16 s_magic; /* Magic signature */
1321 __le16 s_state; /* File system state */
1322 __le16 s_errors; /* Behaviour when detecting errors */
1323 __le16 s_minor_rev_level; /* minor revision level */
1324 /*40*/ __le32 s_lastcheck; /* time of last check */
1325 __le32 s_checkinterval; /* max. time between checks */
1326 __le32 s_creator_os; /* OS */
1327 __le32 s_rev_level; /* Revision level */
1328 /*50*/ __le16 s_def_resuid; /* Default uid for reserved blocks */
1329 __le16 s_def_resgid; /* Default gid for reserved blocks */
1330 /*
1331 * These fields are for EXT4_DYNAMIC_REV superblocks only.
1332 *
1333 * Note: the difference between the compatible feature set and
1334 * the incompatible feature set is that if there is a bit set
1335 * in the incompatible feature set that the kernel doesn't
1336 * know about, it should refuse to mount the filesystem.
1337 *
1338 * e2fsck's requirements are more strict; if it doesn't know
1339 * about a feature in either the compatible or incompatible
1340 * feature set, it must abort and not try to meddle with
1341 * things it doesn't understand...
1342 */
1343 __le32 s_first_ino; /* First non-reserved inode */
1344 __le16 s_inode_size; /* size of inode structure */
1345 __le16 s_block_group_nr; /* block group # of this superblock */
1346 __le32 s_feature_compat; /* compatible feature set */
1347 /*60*/ __le32 s_feature_incompat; /* incompatible feature set */
1348 __le32 s_feature_ro_compat; /* readonly-compatible feature set */
1349 /*68*/ __u8 s_uuid[16]; /* 128-bit uuid for volume */
1350 /*78*/ char s_volume_name[EXT4_LABEL_MAX] __nonstring; /* volume name */
1351 /*88*/ char s_last_mounted[64] __nonstring; /* directory where last mounted */
1352 /*C8*/ __le32 s_algorithm_usage_bitmap; /* For compression */
1353 /*
1354 * Performance hints. Directory preallocation should only
1355 * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1356 */
1357 __u8 s_prealloc_blocks; /* Nr of blocks to try to preallocate*/
1358 __u8 s_prealloc_dir_blocks; /* Nr to preallocate for dirs */
1359 __le16 s_reserved_gdt_blocks; /* Per group desc for online growth */
1360 /*
1361 * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1362 */
1363 /*D0*/ __u8 s_journal_uuid[16]; /* uuid of journal superblock */
1364 /*E0*/ __le32 s_journal_inum; /* inode number of journal file */
1365 __le32 s_journal_dev; /* device number of journal file */
1366 __le32 s_last_orphan; /* start of list of inodes to delete */
1367 __le32 s_hash_seed[4]; /* HTREE hash seed */
1368 __u8 s_def_hash_version; /* Default hash version to use */
1369 __u8 s_jnl_backup_type;
1370 __le16 s_desc_size; /* size of group descriptor */
1371 /*100*/ __le32 s_default_mount_opts;
1372 __le32 s_first_meta_bg; /* First metablock block group */
1373 __le32 s_mkfs_time; /* When the filesystem was created */
1374 __le32 s_jnl_blocks[17]; /* Backup of the journal inode */
1375 /* 64bit support valid if EXT4_FEATURE_INCOMPAT_64BIT */
1376 /*150*/ __le32 s_blocks_count_hi; /* Blocks count */
1377 __le32 s_r_blocks_count_hi; /* Reserved blocks count */
1378 __le32 s_free_blocks_count_hi; /* Free blocks count */
1379 __le16 s_min_extra_isize; /* All inodes have at least # bytes */
1380 __le16 s_want_extra_isize; /* New inodes should reserve # bytes */
1381 __le32 s_flags; /* Miscellaneous flags */
1382 __le16 s_raid_stride; /* RAID stride */
1383 __le16 s_mmp_update_interval; /* # seconds to wait in MMP checking */
1384 __le64 s_mmp_block; /* Block for multi-mount protection */
1385 __le32 s_raid_stripe_width; /* blocks on all data disks (N*stride)*/
1386 __u8 s_log_groups_per_flex; /* FLEX_BG group size */
1387 __u8 s_checksum_type; /* metadata checksum algorithm used */
1388 __u8 s_encryption_level; /* versioning level for encryption */
1389 __u8 s_reserved_pad; /* Padding to next 32bits */
1390 __le64 s_kbytes_written; /* nr of lifetime kilobytes written */
1391 __le32 s_snapshot_inum; /* Inode number of active snapshot */
1392 __le32 s_snapshot_id; /* sequential ID of active snapshot */
1393 __le64 s_snapshot_r_blocks_count; /* reserved blocks for active
1394 snapshot's future use */
1395 __le32 s_snapshot_list; /* inode number of the head of the
1396 on-disk snapshot list */
1397 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1398 __le32 s_error_count; /* number of fs errors */
1399 __le32 s_first_error_time; /* first time an error happened */
1400 __le32 s_first_error_ino; /* inode involved in first error */
1401 __le64 s_first_error_block; /* block involved of first error */
1402 __u8 s_first_error_func[32] __nonstring; /* function where the error happened */
1403 __le32 s_first_error_line; /* line number where error happened */
1404 __le32 s_last_error_time; /* most recent time of an error */
1405 __le32 s_last_error_ino; /* inode involved in last error */
1406 __le32 s_last_error_line; /* line number where error happened */
1407 __le64 s_last_error_block; /* block involved of last error */
1408 __u8 s_last_error_func[32] __nonstring; /* function where the error happened */
1409 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1410 __u8 s_mount_opts[64];
1411 __le32 s_usr_quota_inum; /* inode for tracking user quota */
1412 __le32 s_grp_quota_inum; /* inode for tracking group quota */
1413 __le32 s_overhead_clusters; /* overhead blocks/clusters in fs */
1414 __le32 s_backup_bgs[2]; /* groups with sparse_super2 SBs */
1415 __u8 s_encrypt_algos[4]; /* Encryption algorithms in use */
1416 __u8 s_encrypt_pw_salt[16]; /* Salt used for string2key algorithm */
1417 __le32 s_lpf_ino; /* Location of the lost+found inode */
1418 __le32 s_prj_quota_inum; /* inode for tracking project quota */
1419 __le32 s_checksum_seed; /* crc32c(uuid) if csum_seed set */
1420 __u8 s_wtime_hi;
1421 __u8 s_mtime_hi;
1422 __u8 s_mkfs_time_hi;
1423 __u8 s_lastcheck_hi;
1424 __u8 s_first_error_time_hi;
1425 __u8 s_last_error_time_hi;
1426 __u8 s_first_error_errcode;
1427 __u8 s_last_error_errcode;
1428 __le16 s_encoding; /* Filename charset encoding */
1429 __le16 s_encoding_flags; /* Filename charset encoding flags */
1430 __le32 s_orphan_file_inum; /* Inode for tracking orphan inodes */
1431 __le32 s_reserved[94]; /* 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 /* Number of quota types we support */
1440 #define EXT4_MAXQUOTAS 3
1441
1442 #define EXT4_ENC_UTF8_12_1 1
1443
1444 /* Types of ext4 journal triggers */
1445 enum ext4_journal_trigger_type {
1446 EXT4_JTR_ORPHAN_FILE,
1447 EXT4_JTR_NONE /* This must be the last entry for indexing to work! */
1448 };
1449
1450 #define EXT4_JOURNAL_TRIGGER_COUNT EXT4_JTR_NONE
1451
1452 struct ext4_journal_trigger {
1453 struct jbd2_buffer_trigger_type tr_triggers;
1454 struct super_block *sb;
1455 };
1456
1457 static inline struct ext4_journal_trigger *EXT4_TRIGGER(
1458 struct jbd2_buffer_trigger_type *trigger)
1459 {
1460 return container_of(trigger, struct ext4_journal_trigger, tr_triggers);
1461 }
1462
1463 #define EXT4_ORPHAN_BLOCK_MAGIC 0x0b10ca04
1464
1465 /* Structure at the tail of orphan block */
1466 struct ext4_orphan_block_tail {
1467 __le32 ob_magic;
1468 __le32 ob_checksum;
1469 };
1470
1471 static inline int ext4_inodes_per_orphan_block(struct super_block *sb)
1472 {
1473 return (sb->s_blocksize - sizeof(struct ext4_orphan_block_tail)) /
1474 sizeof(u32);
1475 }
1476
1477 struct ext4_orphan_block {
1478 atomic_t ob_free_entries; /* Number of free orphan entries in block */
1479 struct buffer_head *ob_bh; /* Buffer for orphan block */
1480 };
1481
1482 /*
1483 * Info about orphan file.
1484 */
1485 struct ext4_orphan_info {
1486 int of_blocks; /* Number of orphan blocks in a file */
1487 __u32 of_csum_seed; /* Checksum seed for orphan file */
1488 struct ext4_orphan_block *of_binfo; /* Array with info about orphan
1489 * file blocks */
1490 };
1491
1492 /*
1493 * fourth extended-fs super-block data in memory
1494 */
1495 struct ext4_sb_info {
1496 unsigned long s_desc_size; /* Size of a group descriptor in bytes */
1497 unsigned long s_inodes_per_block;/* Number of inodes per block */
1498 unsigned long s_blocks_per_group;/* Number of blocks in a group */
1499 unsigned long s_clusters_per_group; /* Number of clusters in a group */
1500 unsigned long s_inodes_per_group;/* Number of inodes in a group */
1501 unsigned long s_itb_per_group; /* Number of inode table blocks per group */
1502 unsigned long s_gdb_count; /* Number of group descriptor blocks */
1503 unsigned long s_desc_per_block; /* Number of group descriptors per block */
1504 ext4_group_t s_groups_count; /* Number of groups in the fs */
1505 ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1506 unsigned long s_overhead; /* # of fs overhead clusters */
1507 unsigned int s_cluster_ratio; /* Number of blocks per cluster */
1508 unsigned int s_cluster_bits; /* log2 of s_cluster_ratio */
1509 loff_t s_bitmap_maxbytes; /* max bytes for bitmap files */
1510 struct buffer_head * s_sbh; /* Buffer containing the super block */
1511 struct ext4_super_block *s_es; /* Pointer to the super block in the buffer */
1512 /* Array of bh's for the block group descriptors */
1513 struct buffer_head * __rcu *s_group_desc;
1514 unsigned int s_mount_opt;
1515 unsigned int s_mount_opt2;
1516 unsigned long s_mount_flags;
1517 unsigned int s_def_mount_opt;
1518 unsigned int s_def_mount_opt2;
1519 ext4_fsblk_t s_sb_block;
1520 atomic64_t s_resv_clusters;
1521 kuid_t s_resuid;
1522 kgid_t s_resgid;
1523 unsigned short s_mount_state;
1524 unsigned short s_pad;
1525 int s_addr_per_block_bits;
1526 int s_desc_per_block_bits;
1527 int s_inode_size;
1528 int s_first_ino;
1529 unsigned int s_inode_readahead_blks;
1530 unsigned int s_inode_goal;
1531 u32 s_hash_seed[4];
1532 int s_def_hash_version;
1533 int s_hash_unsigned; /* 3 if hash should be unsigned, 0 if not */
1534 struct percpu_counter s_freeclusters_counter;
1535 struct percpu_counter s_freeinodes_counter;
1536 struct percpu_counter s_dirs_counter;
1537 struct percpu_counter s_dirtyclusters_counter;
1538 struct percpu_counter s_sra_exceeded_retry_limit;
1539 struct blockgroup_lock *s_blockgroup_lock;
1540 struct proc_dir_entry *s_proc;
1541 struct kobject s_kobj;
1542 struct completion s_kobj_unregister;
1543 struct super_block *s_sb;
1544 struct buffer_head *s_mmp_bh;
1545
1546 /* Journaling */
1547 struct journal_s *s_journal;
1548 unsigned long s_ext4_flags; /* Ext4 superblock flags */
1549 struct mutex s_orphan_lock; /* Protects on disk list changes */
1550 struct list_head s_orphan; /* List of orphaned inodes in on disk
1551 list */
1552 struct ext4_orphan_info s_orphan_info;
1553 unsigned long s_commit_interval;
1554 u32 s_max_batch_time;
1555 u32 s_min_batch_time;
1556 struct file *s_journal_bdev_file;
1557 #ifdef CONFIG_QUOTA
1558 /* Names of quota files with journalled quota */
1559 char __rcu *s_qf_names[EXT4_MAXQUOTAS];
1560 int s_jquota_fmt; /* Format of quota to use */
1561 #endif
1562 unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1563 struct ext4_system_blocks __rcu *s_system_blks;
1564
1565 #ifdef EXTENTS_STATS
1566 /* ext4 extents stats */
1567 unsigned long s_ext_min;
1568 unsigned long s_ext_max;
1569 unsigned long s_depth_max;
1570 spinlock_t s_ext_stats_lock;
1571 unsigned long s_ext_blocks;
1572 unsigned long s_ext_extents;
1573 #endif
1574
1575 /* for buddy allocator */
1576 struct ext4_group_info ** __rcu *s_group_info;
1577 struct inode *s_buddy_cache;
1578 spinlock_t s_md_lock;
1579 unsigned short *s_mb_offsets;
1580 unsigned int *s_mb_maxs;
1581 unsigned int s_group_info_size;
1582 unsigned int s_mb_free_pending;
1583 struct list_head s_freed_data_list[2]; /* List of blocks to be freed
1584 after commit completed */
1585 struct list_head s_discard_list;
1586 struct work_struct s_discard_work;
1587 atomic_t s_retry_alloc_pending;
1588 struct list_head *s_mb_avg_fragment_size;
1589 rwlock_t *s_mb_avg_fragment_size_locks;
1590 struct list_head *s_mb_largest_free_orders;
1591 rwlock_t *s_mb_largest_free_orders_locks;
1592
1593 /* tunables */
1594 unsigned long s_stripe;
1595 unsigned int s_mb_max_linear_groups;
1596 unsigned int s_mb_stream_request;
1597 unsigned int s_mb_max_to_scan;
1598 unsigned int s_mb_min_to_scan;
1599 unsigned int s_mb_stats;
1600 unsigned int s_mb_order2_reqs;
1601 unsigned int s_mb_group_prealloc;
1602 unsigned int s_max_dir_size_kb;
1603 /* where last allocation was done - for stream allocation */
1604 unsigned long s_mb_last_group;
1605 unsigned long s_mb_last_start;
1606 unsigned int s_mb_prefetch;
1607 unsigned int s_mb_prefetch_limit;
1608 unsigned int s_mb_best_avail_max_trim_order;
1609
1610 /* stats for buddy allocator */
1611 atomic_t s_bal_reqs; /* number of reqs with len > 1 */
1612 atomic_t s_bal_success; /* we found long enough chunks */
1613 atomic_t s_bal_allocated; /* in blocks */
1614 atomic_t s_bal_ex_scanned; /* total extents scanned */
1615 atomic_t s_bal_cX_ex_scanned[EXT4_MB_NUM_CRS]; /* total extents scanned */
1616 atomic_t s_bal_groups_scanned; /* number of groups scanned */
1617 atomic_t s_bal_goals; /* goal hits */
1618 atomic_t s_bal_len_goals; /* len goal hits */
1619 atomic_t s_bal_breaks; /* too long searches */
1620 atomic_t s_bal_2orders; /* 2^order hits */
1621 atomic_t s_bal_p2_aligned_bad_suggestions;
1622 atomic_t s_bal_goal_fast_bad_suggestions;
1623 atomic_t s_bal_best_avail_bad_suggestions;
1624 atomic64_t s_bal_cX_groups_considered[EXT4_MB_NUM_CRS];
1625 atomic64_t s_bal_cX_hits[EXT4_MB_NUM_CRS];
1626 atomic64_t s_bal_cX_failed[EXT4_MB_NUM_CRS]; /* cX loop didn't find blocks */
1627 atomic_t s_mb_buddies_generated; /* number of buddies generated */
1628 atomic64_t s_mb_generation_time;
1629 atomic_t s_mb_lost_chunks;
1630 atomic_t s_mb_preallocated;
1631 atomic_t s_mb_discarded;
1632 atomic_t s_lock_busy;
1633
1634 /* locality groups */
1635 struct ext4_locality_group __percpu *s_locality_groups;
1636
1637 /* for write statistics */
1638 unsigned long s_sectors_written_start;
1639 u64 s_kbytes_written;
1640
1641 /* the size of zero-out chunk */
1642 unsigned int s_extent_max_zeroout_kb;
1643
1644 unsigned int s_log_groups_per_flex;
1645 struct flex_groups * __rcu *s_flex_groups;
1646 ext4_group_t s_flex_groups_allocated;
1647
1648 /* workqueue for reserved extent conversions (buffered io) */
1649 struct workqueue_struct *rsv_conversion_wq;
1650
1651 /* timer for periodic error stats printing */
1652 struct timer_list s_err_report;
1653
1654 /* Lazy inode table initialization info */
1655 struct ext4_li_request *s_li_request;
1656 /* Wait multiplier for lazy initialization thread */
1657 unsigned int s_li_wait_mult;
1658
1659 /* Kernel thread for multiple mount protection */
1660 struct task_struct *s_mmp_tsk;
1661
1662 /* record the last minlen when FITRIM is called. */
1663 unsigned long s_last_trim_minblks;
1664
1665 /* Reference to checksum algorithm driver via cryptoapi */
1666 struct crypto_shash *s_chksum_driver;
1667
1668 /* Precomputed FS UUID checksum for seeding other checksums */
1669 __u32 s_csum_seed;
1670
1671 /* Reclaim extents from extent status tree */
1672 struct shrinker *s_es_shrinker;
1673 struct list_head s_es_list; /* List of inodes with reclaimable extents */
1674 long s_es_nr_inode;
1675 struct ext4_es_stats s_es_stats;
1676 struct mb_cache *s_ea_block_cache;
1677 struct mb_cache *s_ea_inode_cache;
1678 spinlock_t s_es_lock ____cacheline_aligned_in_smp;
1679
1680 /* Journal triggers for checksum computation */
1681 struct ext4_journal_trigger s_journal_triggers[EXT4_JOURNAL_TRIGGER_COUNT];
1682
1683 /* Ratelimit ext4 messages. */
1684 struct ratelimit_state s_err_ratelimit_state;
1685 struct ratelimit_state s_warning_ratelimit_state;
1686 struct ratelimit_state s_msg_ratelimit_state;
1687 atomic_t s_warning_count;
1688 atomic_t s_msg_count;
1689
1690 /* Encryption policy for '-o test_dummy_encryption' */
1691 struct fscrypt_dummy_policy s_dummy_enc_policy;
1692
1693 /*
1694 * Barrier between writepages ops and changing any inode's JOURNAL_DATA
1695 * or EXTENTS flag or between writepages ops and changing DELALLOC or
1696 * DIOREAD_NOLOCK mount options on remount.
1697 */
1698 struct percpu_rw_semaphore s_writepages_rwsem;
1699 struct dax_device *s_daxdev;
1700 u64 s_dax_part_off;
1701 #ifdef CONFIG_EXT4_DEBUG
1702 unsigned long s_simulate_fail;
1703 #endif
1704 /* Record the errseq of the backing block device */
1705 errseq_t s_bdev_wb_err;
1706 spinlock_t s_bdev_wb_lock;
1707
1708 /* Information about errors that happened during this mount */
1709 spinlock_t s_error_lock;
1710 int s_add_error_count;
1711 int s_first_error_code;
1712 __u32 s_first_error_line;
1713 __u32 s_first_error_ino;
1714 __u64 s_first_error_block;
1715 const char *s_first_error_func;
1716 time64_t s_first_error_time;
1717 int s_last_error_code;
1718 __u32 s_last_error_line;
1719 __u32 s_last_error_ino;
1720 __u64 s_last_error_block;
1721 const char *s_last_error_func;
1722 time64_t s_last_error_time;
1723 /*
1724 * If we are in a context where we cannot update the on-disk
1725 * superblock, we queue the work here. This is used to update
1726 * the error information in the superblock, and for periodic
1727 * updates of the superblock called from the commit callback
1728 * function.
1729 */
1730 struct work_struct s_sb_upd_work;
1731
1732 /* Ext4 fast commit sub transaction ID */
1733 atomic_t s_fc_subtid;
1734
1735 /*
1736 * After commit starts, the main queue gets locked, and the further
1737 * updates get added in the staging queue.
1738 */
1739 #define FC_Q_MAIN 0
1740 #define FC_Q_STAGING 1
1741 struct list_head s_fc_q[2]; /* Inodes staged for fast commit
1742 * that have data changes in them.
1743 */
1744 struct list_head s_fc_dentry_q[2]; /* directory entry updates */
1745 unsigned int s_fc_bytes;
1746 /*
1747 * Main fast commit lock. This lock protects accesses to the
1748 * following fields:
1749 * ei->i_fc_list, s_fc_dentry_q, s_fc_q, s_fc_bytes, s_fc_bh.
1750 */
1751 spinlock_t s_fc_lock;
1752 struct buffer_head *s_fc_bh;
1753 struct ext4_fc_stats s_fc_stats;
1754 tid_t s_fc_ineligible_tid;
1755 #ifdef CONFIG_EXT4_DEBUG
1756 int s_fc_debug_max_replay;
1757 #endif
1758 struct ext4_fc_replay_state s_fc_replay_state;
1759 };
1760
1761 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1762 {
1763 return sb->s_fs_info;
1764 }
1765 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1766 {
1767 return container_of(inode, struct ext4_inode_info, vfs_inode);
1768 }
1769
1770 static inline int ext4_writepages_down_read(struct super_block *sb)
1771 {
1772 percpu_down_read(&EXT4_SB(sb)->s_writepages_rwsem);
1773 return memalloc_nofs_save();
1774 }
1775
1776 static inline void ext4_writepages_up_read(struct super_block *sb, int ctx)
1777 {
1778 memalloc_nofs_restore(ctx);
1779 percpu_up_read(&EXT4_SB(sb)->s_writepages_rwsem);
1780 }
1781
1782 static inline int ext4_writepages_down_write(struct super_block *sb)
1783 {
1784 percpu_down_write(&EXT4_SB(sb)->s_writepages_rwsem);
1785 return memalloc_nofs_save();
1786 }
1787
1788 static inline void ext4_writepages_up_write(struct super_block *sb, int ctx)
1789 {
1790 memalloc_nofs_restore(ctx);
1791 percpu_up_write(&EXT4_SB(sb)->s_writepages_rwsem);
1792 }
1793
1794 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1795 {
1796 return ino == EXT4_ROOT_INO ||
1797 (ino >= EXT4_FIRST_INO(sb) &&
1798 ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1799 }
1800
1801 /*
1802 * Returns: sbi->field[index]
1803 * Used to access an array element from the following sbi fields which require
1804 * rcu protection to avoid dereferencing an invalid pointer due to reassignment
1805 * - s_group_desc
1806 * - s_group_info
1807 * - s_flex_group
1808 */
1809 #define sbi_array_rcu_deref(sbi, field, index) \
1810 ({ \
1811 typeof(*((sbi)->field)) _v; \
1812 rcu_read_lock(); \
1813 _v = ((typeof(_v)*)rcu_dereference((sbi)->field))[index]; \
1814 rcu_read_unlock(); \
1815 _v; \
1816 })
1817
1818 /*
1819 * run-time mount flags
1820 */
1821 enum {
1822 EXT4_MF_MNTDIR_SAMPLED,
1823 EXT4_MF_FC_INELIGIBLE /* Fast commit ineligible */
1824 };
1825
1826 static inline void ext4_set_mount_flag(struct super_block *sb, int bit)
1827 {
1828 set_bit(bit, &EXT4_SB(sb)->s_mount_flags);
1829 }
1830
1831 static inline void ext4_clear_mount_flag(struct super_block *sb, int bit)
1832 {
1833 clear_bit(bit, &EXT4_SB(sb)->s_mount_flags);
1834 }
1835
1836 static inline int ext4_test_mount_flag(struct super_block *sb, int bit)
1837 {
1838 return test_bit(bit, &EXT4_SB(sb)->s_mount_flags);
1839 }
1840
1841
1842 /*
1843 * Simulate_fail codes
1844 */
1845 #define EXT4_SIM_BBITMAP_EIO 1
1846 #define EXT4_SIM_BBITMAP_CRC 2
1847 #define EXT4_SIM_IBITMAP_EIO 3
1848 #define EXT4_SIM_IBITMAP_CRC 4
1849 #define EXT4_SIM_INODE_EIO 5
1850 #define EXT4_SIM_INODE_CRC 6
1851 #define EXT4_SIM_DIRBLOCK_EIO 7
1852 #define EXT4_SIM_DIRBLOCK_CRC 8
1853
1854 static inline bool ext4_simulate_fail(struct super_block *sb,
1855 unsigned long code)
1856 {
1857 #ifdef CONFIG_EXT4_DEBUG
1858 struct ext4_sb_info *sbi = EXT4_SB(sb);
1859
1860 if (unlikely(sbi->s_simulate_fail == code)) {
1861 sbi->s_simulate_fail = 0;
1862 return true;
1863 }
1864 #endif
1865 return false;
1866 }
1867
1868 static inline void ext4_simulate_fail_bh(struct super_block *sb,
1869 struct buffer_head *bh,
1870 unsigned long code)
1871 {
1872 if (!IS_ERR(bh) && ext4_simulate_fail(sb, code))
1873 clear_buffer_uptodate(bh);
1874 }
1875
1876 /*
1877 * Error number codes for s_{first,last}_error_errno
1878 *
1879 * Linux errno numbers are architecture specific, so we need to translate
1880 * them into something which is architecture independent. We don't define
1881 * codes for all errno's; just the ones which are most likely to be the cause
1882 * of an ext4_error() call.
1883 */
1884 #define EXT4_ERR_UNKNOWN 1
1885 #define EXT4_ERR_EIO 2
1886 #define EXT4_ERR_ENOMEM 3
1887 #define EXT4_ERR_EFSBADCRC 4
1888 #define EXT4_ERR_EFSCORRUPTED 5
1889 #define EXT4_ERR_ENOSPC 6
1890 #define EXT4_ERR_ENOKEY 7
1891 #define EXT4_ERR_EROFS 8
1892 #define EXT4_ERR_EFBIG 9
1893 #define EXT4_ERR_EEXIST 10
1894 #define EXT4_ERR_ERANGE 11
1895 #define EXT4_ERR_EOVERFLOW 12
1896 #define EXT4_ERR_EBUSY 13
1897 #define EXT4_ERR_ENOTDIR 14
1898 #define EXT4_ERR_ENOTEMPTY 15
1899 #define EXT4_ERR_ESHUTDOWN 16
1900 #define EXT4_ERR_EFAULT 17
1901
1902 /*
1903 * Inode dynamic state flags
1904 */
1905 enum {
1906 EXT4_STATE_NEW, /* inode is newly created */
1907 EXT4_STATE_XATTR, /* has in-inode xattrs */
1908 EXT4_STATE_NO_EXPAND, /* No space for expansion */
1909 EXT4_STATE_DA_ALLOC_CLOSE, /* Alloc DA blks on close */
1910 EXT4_STATE_EXT_MIGRATE, /* Inode is migrating */
1911 EXT4_STATE_NEWENTRY, /* File just added to dir */
1912 EXT4_STATE_MAY_INLINE_DATA, /* may have in-inode data */
1913 EXT4_STATE_EXT_PRECACHED, /* extents have been precached */
1914 EXT4_STATE_LUSTRE_EA_INODE, /* Lustre-style ea_inode */
1915 EXT4_STATE_VERITY_IN_PROGRESS, /* building fs-verity Merkle tree */
1916 EXT4_STATE_FC_COMMITTING, /* Fast commit ongoing */
1917 EXT4_STATE_ORPHAN_FILE, /* Inode orphaned in orphan file */
1918 };
1919
1920 #define EXT4_INODE_BIT_FNS(name, field, offset) \
1921 static inline int ext4_test_inode_##name(struct inode *inode, int bit) \
1922 { \
1923 return test_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1924 } \
1925 static inline void ext4_set_inode_##name(struct inode *inode, int bit) \
1926 { \
1927 set_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1928 } \
1929 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1930 { \
1931 clear_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1932 }
1933
1934 /* Add these declarations here only so that these functions can be
1935 * found by name. Otherwise, they are very hard to locate. */
1936 static inline int ext4_test_inode_flag(struct inode *inode, int bit);
1937 static inline void ext4_set_inode_flag(struct inode *inode, int bit);
1938 static inline void ext4_clear_inode_flag(struct inode *inode, int bit);
1939 EXT4_INODE_BIT_FNS(flag, flags, 0)
1940
1941 /* Add these declarations here only so that these functions can be
1942 * found by name. Otherwise, they are very hard to locate. */
1943 static inline int ext4_test_inode_state(struct inode *inode, int bit);
1944 static inline void ext4_set_inode_state(struct inode *inode, int bit);
1945 static inline void ext4_clear_inode_state(struct inode *inode, int bit);
1946 #if (BITS_PER_LONG < 64)
1947 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1948
1949 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1950 {
1951 (ei)->i_state_flags = 0;
1952 }
1953 #else
1954 EXT4_INODE_BIT_FNS(state, flags, 32)
1955
1956 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1957 {
1958 /* We depend on the fact that callers will set i_flags */
1959 }
1960 #endif
1961 #else
1962 /* Assume that user mode programs are passing in an ext4fs superblock, not
1963 * a kernel struct super_block. This will allow us to call the feature-test
1964 * macros from user land. */
1965 #define EXT4_SB(sb) (sb)
1966 #endif
1967
1968 static inline bool ext4_verity_in_progress(struct inode *inode)
1969 {
1970 return IS_ENABLED(CONFIG_FS_VERITY) &&
1971 ext4_test_inode_state(inode, EXT4_STATE_VERITY_IN_PROGRESS);
1972 }
1973
1974 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1975
1976 /*
1977 * Codes for operating systems
1978 */
1979 #define EXT4_OS_LINUX 0
1980 #define EXT4_OS_HURD 1
1981 #define EXT4_OS_MASIX 2
1982 #define EXT4_OS_FREEBSD 3
1983 #define EXT4_OS_LITES 4
1984
1985 /*
1986 * Revision levels
1987 */
1988 #define EXT4_GOOD_OLD_REV 0 /* The good old (original) format */
1989 #define EXT4_DYNAMIC_REV 1 /* V2 format w/ dynamic inode sizes */
1990
1991 #define EXT4_MAX_SUPP_REV EXT4_DYNAMIC_REV
1992
1993 #define EXT4_GOOD_OLD_INODE_SIZE 128
1994
1995 #define EXT4_EXTRA_TIMESTAMP_MAX (((s64)1 << 34) - 1 + S32_MIN)
1996 #define EXT4_NON_EXTRA_TIMESTAMP_MAX S32_MAX
1997 #define EXT4_TIMESTAMP_MIN S32_MIN
1998
1999 /*
2000 * Feature set definitions
2001 */
2002
2003 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC 0x0001
2004 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES 0x0002
2005 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL 0x0004
2006 #define EXT4_FEATURE_COMPAT_EXT_ATTR 0x0008
2007 #define EXT4_FEATURE_COMPAT_RESIZE_INODE 0x0010
2008 #define EXT4_FEATURE_COMPAT_DIR_INDEX 0x0020
2009 #define EXT4_FEATURE_COMPAT_SPARSE_SUPER2 0x0200
2010 /*
2011 * The reason why "FAST_COMMIT" is a compat feature is that, FS becomes
2012 * incompatible only if fast commit blocks are present in the FS. Since we
2013 * clear the journal (and thus the fast commit blocks), we don't mark FS as
2014 * incompatible. We also have a JBD2 incompat feature, which gets set when
2015 * there are fast commit blocks present in the journal.
2016 */
2017 #define EXT4_FEATURE_COMPAT_FAST_COMMIT 0x0400
2018 #define EXT4_FEATURE_COMPAT_STABLE_INODES 0x0800
2019 #define EXT4_FEATURE_COMPAT_ORPHAN_FILE 0x1000 /* Orphan file exists */
2020
2021 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER 0x0001
2022 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE 0x0002
2023 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR 0x0004
2024 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE 0x0008
2025 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM 0x0010
2026 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK 0x0020
2027 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE 0x0040
2028 #define EXT4_FEATURE_RO_COMPAT_QUOTA 0x0100
2029 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC 0x0200
2030 /*
2031 * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM). When
2032 * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
2033 * all other data structures' checksums. However, the METADATA_CSUM and
2034 * GDT_CSUM bits are mutually exclusive.
2035 */
2036 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM 0x0400
2037 #define EXT4_FEATURE_RO_COMPAT_READONLY 0x1000
2038 #define EXT4_FEATURE_RO_COMPAT_PROJECT 0x2000
2039 #define EXT4_FEATURE_RO_COMPAT_VERITY 0x8000
2040 #define EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT 0x10000 /* Orphan file may be
2041 non-empty */
2042
2043 #define EXT4_FEATURE_INCOMPAT_COMPRESSION 0x0001
2044 #define EXT4_FEATURE_INCOMPAT_FILETYPE 0x0002
2045 #define EXT4_FEATURE_INCOMPAT_RECOVER 0x0004 /* Needs recovery */
2046 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 /* Journal device */
2047 #define EXT4_FEATURE_INCOMPAT_META_BG 0x0010
2048 #define EXT4_FEATURE_INCOMPAT_EXTENTS 0x0040 /* extents support */
2049 #define EXT4_FEATURE_INCOMPAT_64BIT 0x0080
2050 #define EXT4_FEATURE_INCOMPAT_MMP 0x0100
2051 #define EXT4_FEATURE_INCOMPAT_FLEX_BG 0x0200
2052 #define EXT4_FEATURE_INCOMPAT_EA_INODE 0x0400 /* EA in inode */
2053 #define EXT4_FEATURE_INCOMPAT_DIRDATA 0x1000 /* data in dirent */
2054 #define EXT4_FEATURE_INCOMPAT_CSUM_SEED 0x2000
2055 #define EXT4_FEATURE_INCOMPAT_LARGEDIR 0x4000 /* >2GB or 3-lvl htree */
2056 #define EXT4_FEATURE_INCOMPAT_INLINE_DATA 0x8000 /* data in inode */
2057 #define EXT4_FEATURE_INCOMPAT_ENCRYPT 0x10000
2058 #define EXT4_FEATURE_INCOMPAT_CASEFOLD 0x20000
2059
2060 extern void ext4_update_dynamic_rev(struct super_block *sb);
2061
2062 #define EXT4_FEATURE_COMPAT_FUNCS(name, flagname) \
2063 static inline bool ext4_has_feature_##name(struct super_block *sb) \
2064 { \
2065 return ((EXT4_SB(sb)->s_es->s_feature_compat & \
2066 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname)) != 0); \
2067 } \
2068 static inline void ext4_set_feature_##name(struct super_block *sb) \
2069 { \
2070 ext4_update_dynamic_rev(sb); \
2071 EXT4_SB(sb)->s_es->s_feature_compat |= \
2072 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
2073 } \
2074 static inline void ext4_clear_feature_##name(struct super_block *sb) \
2075 { \
2076 EXT4_SB(sb)->s_es->s_feature_compat &= \
2077 ~cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
2078 }
2079
2080 #define EXT4_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
2081 static inline bool ext4_has_feature_##name(struct super_block *sb) \
2082 { \
2083 return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
2084 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname)) != 0); \
2085 } \
2086 static inline void ext4_set_feature_##name(struct super_block *sb) \
2087 { \
2088 ext4_update_dynamic_rev(sb); \
2089 EXT4_SB(sb)->s_es->s_feature_ro_compat |= \
2090 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
2091 } \
2092 static inline void ext4_clear_feature_##name(struct super_block *sb) \
2093 { \
2094 EXT4_SB(sb)->s_es->s_feature_ro_compat &= \
2095 ~cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
2096 }
2097
2098 #define EXT4_FEATURE_INCOMPAT_FUNCS(name, flagname) \
2099 static inline bool ext4_has_feature_##name(struct super_block *sb) \
2100 { \
2101 return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
2102 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname)) != 0); \
2103 } \
2104 static inline void ext4_set_feature_##name(struct super_block *sb) \
2105 { \
2106 ext4_update_dynamic_rev(sb); \
2107 EXT4_SB(sb)->s_es->s_feature_incompat |= \
2108 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
2109 } \
2110 static inline void ext4_clear_feature_##name(struct super_block *sb) \
2111 { \
2112 EXT4_SB(sb)->s_es->s_feature_incompat &= \
2113 ~cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
2114 }
2115
2116 EXT4_FEATURE_COMPAT_FUNCS(dir_prealloc, DIR_PREALLOC)
2117 EXT4_FEATURE_COMPAT_FUNCS(imagic_inodes, IMAGIC_INODES)
2118 EXT4_FEATURE_COMPAT_FUNCS(journal, HAS_JOURNAL)
2119 EXT4_FEATURE_COMPAT_FUNCS(xattr, EXT_ATTR)
2120 EXT4_FEATURE_COMPAT_FUNCS(resize_inode, RESIZE_INODE)
2121 EXT4_FEATURE_COMPAT_FUNCS(dir_index, DIR_INDEX)
2122 EXT4_FEATURE_COMPAT_FUNCS(sparse_super2, SPARSE_SUPER2)
2123 EXT4_FEATURE_COMPAT_FUNCS(fast_commit, FAST_COMMIT)
2124 EXT4_FEATURE_COMPAT_FUNCS(stable_inodes, STABLE_INODES)
2125 EXT4_FEATURE_COMPAT_FUNCS(orphan_file, ORPHAN_FILE)
2126
2127 EXT4_FEATURE_RO_COMPAT_FUNCS(sparse_super, SPARSE_SUPER)
2128 EXT4_FEATURE_RO_COMPAT_FUNCS(large_file, LARGE_FILE)
2129 EXT4_FEATURE_RO_COMPAT_FUNCS(btree_dir, BTREE_DIR)
2130 EXT4_FEATURE_RO_COMPAT_FUNCS(huge_file, HUGE_FILE)
2131 EXT4_FEATURE_RO_COMPAT_FUNCS(gdt_csum, GDT_CSUM)
2132 EXT4_FEATURE_RO_COMPAT_FUNCS(dir_nlink, DIR_NLINK)
2133 EXT4_FEATURE_RO_COMPAT_FUNCS(extra_isize, EXTRA_ISIZE)
2134 EXT4_FEATURE_RO_COMPAT_FUNCS(quota, QUOTA)
2135 EXT4_FEATURE_RO_COMPAT_FUNCS(bigalloc, BIGALLOC)
2136 EXT4_FEATURE_RO_COMPAT_FUNCS(metadata_csum, METADATA_CSUM)
2137 EXT4_FEATURE_RO_COMPAT_FUNCS(readonly, READONLY)
2138 EXT4_FEATURE_RO_COMPAT_FUNCS(project, PROJECT)
2139 EXT4_FEATURE_RO_COMPAT_FUNCS(verity, VERITY)
2140 EXT4_FEATURE_RO_COMPAT_FUNCS(orphan_present, ORPHAN_PRESENT)
2141
2142 EXT4_FEATURE_INCOMPAT_FUNCS(compression, COMPRESSION)
2143 EXT4_FEATURE_INCOMPAT_FUNCS(filetype, FILETYPE)
2144 EXT4_FEATURE_INCOMPAT_FUNCS(journal_needs_recovery, RECOVER)
2145 EXT4_FEATURE_INCOMPAT_FUNCS(journal_dev, JOURNAL_DEV)
2146 EXT4_FEATURE_INCOMPAT_FUNCS(meta_bg, META_BG)
2147 EXT4_FEATURE_INCOMPAT_FUNCS(extents, EXTENTS)
2148 EXT4_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT)
2149 EXT4_FEATURE_INCOMPAT_FUNCS(mmp, MMP)
2150 EXT4_FEATURE_INCOMPAT_FUNCS(flex_bg, FLEX_BG)
2151 EXT4_FEATURE_INCOMPAT_FUNCS(ea_inode, EA_INODE)
2152 EXT4_FEATURE_INCOMPAT_FUNCS(dirdata, DIRDATA)
2153 EXT4_FEATURE_INCOMPAT_FUNCS(csum_seed, CSUM_SEED)
2154 EXT4_FEATURE_INCOMPAT_FUNCS(largedir, LARGEDIR)
2155 EXT4_FEATURE_INCOMPAT_FUNCS(inline_data, INLINE_DATA)
2156 EXT4_FEATURE_INCOMPAT_FUNCS(encrypt, ENCRYPT)
2157 EXT4_FEATURE_INCOMPAT_FUNCS(casefold, CASEFOLD)
2158
2159 #define EXT2_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR
2160 #define EXT2_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
2161 EXT4_FEATURE_INCOMPAT_META_BG)
2162 #define EXT2_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
2163 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
2164 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
2165
2166 #define EXT3_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR
2167 #define EXT3_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
2168 EXT4_FEATURE_INCOMPAT_RECOVER| \
2169 EXT4_FEATURE_INCOMPAT_META_BG)
2170 #define EXT3_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
2171 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
2172 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
2173
2174 #define EXT4_FEATURE_COMPAT_SUPP (EXT4_FEATURE_COMPAT_EXT_ATTR| \
2175 EXT4_FEATURE_COMPAT_ORPHAN_FILE)
2176 #define EXT4_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
2177 EXT4_FEATURE_INCOMPAT_RECOVER| \
2178 EXT4_FEATURE_INCOMPAT_META_BG| \
2179 EXT4_FEATURE_INCOMPAT_EXTENTS| \
2180 EXT4_FEATURE_INCOMPAT_64BIT| \
2181 EXT4_FEATURE_INCOMPAT_FLEX_BG| \
2182 EXT4_FEATURE_INCOMPAT_EA_INODE| \
2183 EXT4_FEATURE_INCOMPAT_MMP | \
2184 EXT4_FEATURE_INCOMPAT_INLINE_DATA | \
2185 EXT4_FEATURE_INCOMPAT_ENCRYPT | \
2186 EXT4_FEATURE_INCOMPAT_CASEFOLD | \
2187 EXT4_FEATURE_INCOMPAT_CSUM_SEED | \
2188 EXT4_FEATURE_INCOMPAT_LARGEDIR)
2189 #define EXT4_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
2190 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
2191 EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
2192 EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
2193 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
2194 EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
2195 EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
2196 EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
2197 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
2198 EXT4_FEATURE_RO_COMPAT_QUOTA |\
2199 EXT4_FEATURE_RO_COMPAT_PROJECT |\
2200 EXT4_FEATURE_RO_COMPAT_VERITY |\
2201 EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT)
2202
2203 #define EXTN_FEATURE_FUNCS(ver) \
2204 static inline bool ext4_has_unknown_ext##ver##_compat_features(struct super_block *sb) \
2205 { \
2206 return ((EXT4_SB(sb)->s_es->s_feature_compat & \
2207 cpu_to_le32(~EXT##ver##_FEATURE_COMPAT_SUPP)) != 0); \
2208 } \
2209 static inline bool ext4_has_unknown_ext##ver##_ro_compat_features(struct super_block *sb) \
2210 { \
2211 return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
2212 cpu_to_le32(~EXT##ver##_FEATURE_RO_COMPAT_SUPP)) != 0); \
2213 } \
2214 static inline bool ext4_has_unknown_ext##ver##_incompat_features(struct super_block *sb) \
2215 { \
2216 return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
2217 cpu_to_le32(~EXT##ver##_FEATURE_INCOMPAT_SUPP)) != 0); \
2218 }
2219
2220 EXTN_FEATURE_FUNCS(2)
2221 EXTN_FEATURE_FUNCS(3)
2222 EXTN_FEATURE_FUNCS(4)
2223
2224 static inline bool ext4_has_compat_features(struct super_block *sb)
2225 {
2226 return (EXT4_SB(sb)->s_es->s_feature_compat != 0);
2227 }
2228 static inline bool ext4_has_ro_compat_features(struct super_block *sb)
2229 {
2230 return (EXT4_SB(sb)->s_es->s_feature_ro_compat != 0);
2231 }
2232 static inline bool ext4_has_incompat_features(struct super_block *sb)
2233 {
2234 return (EXT4_SB(sb)->s_es->s_feature_incompat != 0);
2235 }
2236
2237 extern int ext4_feature_set_ok(struct super_block *sb, int readonly);
2238
2239 /*
2240 * Superblock flags
2241 */
2242 #define EXT4_FLAGS_RESIZING 0
2243 #define EXT4_FLAGS_SHUTDOWN 1
2244 #define EXT4_FLAGS_BDEV_IS_DAX 2
2245
2246 static inline int ext4_forced_shutdown(struct super_block *sb)
2247 {
2248 return test_bit(EXT4_FLAGS_SHUTDOWN, &EXT4_SB(sb)->s_ext4_flags);
2249 }
2250
2251 /*
2252 * Default values for user and/or group using reserved blocks
2253 */
2254 #define EXT4_DEF_RESUID 0
2255 #define EXT4_DEF_RESGID 0
2256
2257 /*
2258 * Default project ID
2259 */
2260 #define EXT4_DEF_PROJID 0
2261
2262 #define EXT4_DEF_INODE_READAHEAD_BLKS 32
2263
2264 /*
2265 * Default mount options
2266 */
2267 #define EXT4_DEFM_DEBUG 0x0001
2268 #define EXT4_DEFM_BSDGROUPS 0x0002
2269 #define EXT4_DEFM_XATTR_USER 0x0004
2270 #define EXT4_DEFM_ACL 0x0008
2271 #define EXT4_DEFM_UID16 0x0010
2272 #define EXT4_DEFM_JMODE 0x0060
2273 #define EXT4_DEFM_JMODE_DATA 0x0020
2274 #define EXT4_DEFM_JMODE_ORDERED 0x0040
2275 #define EXT4_DEFM_JMODE_WBACK 0x0060
2276 #define EXT4_DEFM_NOBARRIER 0x0100
2277 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
2278 #define EXT4_DEFM_DISCARD 0x0400
2279 #define EXT4_DEFM_NODELALLOC 0x0800
2280
2281 /*
2282 * Default journal batch times
2283 */
2284 #define EXT4_DEF_MIN_BATCH_TIME 0
2285 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
2286
2287 /*
2288 * Minimum number of groups in a flexgroup before we separate out
2289 * directories into the first block group of a flexgroup
2290 */
2291 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
2292
2293 /*
2294 * Structure of a directory entry
2295 */
2296 #define EXT4_NAME_LEN 255
2297 /*
2298 * Base length of the ext4 directory entry excluding the name length
2299 */
2300 #define EXT4_BASE_DIR_LEN (sizeof(struct ext4_dir_entry_2) - EXT4_NAME_LEN)
2301
2302 struct ext4_dir_entry {
2303 __le32 inode; /* Inode number */
2304 __le16 rec_len; /* Directory entry length */
2305 __le16 name_len; /* Name length */
2306 char name[EXT4_NAME_LEN]; /* File name */
2307 };
2308
2309
2310 /*
2311 * Encrypted Casefolded entries require saving the hash on disk. This structure
2312 * followed ext4_dir_entry_2's name[name_len] at the next 4 byte aligned
2313 * boundary.
2314 */
2315 struct ext4_dir_entry_hash {
2316 __le32 hash;
2317 __le32 minor_hash;
2318 };
2319
2320 /*
2321 * The new version of the directory entry. Since EXT4 structures are
2322 * stored in intel byte order, and the name_len field could never be
2323 * bigger than 255 chars, it's safe to reclaim the extra byte for the
2324 * file_type field.
2325 */
2326 struct ext4_dir_entry_2 {
2327 __le32 inode; /* Inode number */
2328 __le16 rec_len; /* Directory entry length */
2329 __u8 name_len; /* Name length */
2330 __u8 file_type; /* See file type macros EXT4_FT_* below */
2331 char name[EXT4_NAME_LEN]; /* File name */
2332 };
2333
2334 /*
2335 * Access the hashes at the end of ext4_dir_entry_2
2336 */
2337 #define EXT4_DIRENT_HASHES(entry) \
2338 ((struct ext4_dir_entry_hash *) \
2339 (((void *)(entry)) + \
2340 ((8 + (entry)->name_len + EXT4_DIR_ROUND) & ~EXT4_DIR_ROUND)))
2341 #define EXT4_DIRENT_HASH(entry) le32_to_cpu(EXT4_DIRENT_HASHES(de)->hash)
2342 #define EXT4_DIRENT_MINOR_HASH(entry) \
2343 le32_to_cpu(EXT4_DIRENT_HASHES(de)->minor_hash)
2344
2345 static inline bool ext4_hash_in_dirent(const struct inode *inode)
2346 {
2347 return IS_CASEFOLDED(inode) && IS_ENCRYPTED(inode);
2348 }
2349
2350 /*
2351 * This is a bogus directory entry at the end of each leaf block that
2352 * records checksums.
2353 */
2354 struct ext4_dir_entry_tail {
2355 __le32 det_reserved_zero1; /* Pretend to be unused */
2356 __le16 det_rec_len; /* 12 */
2357 __u8 det_reserved_zero2; /* Zero name length */
2358 __u8 det_reserved_ft; /* 0xDE, fake file type */
2359 __le32 det_checksum; /* crc32c(uuid+inum+dirblock) */
2360 };
2361
2362 #define EXT4_DIRENT_TAIL(block, blocksize) \
2363 ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
2364 ((blocksize) - \
2365 sizeof(struct ext4_dir_entry_tail))))
2366
2367 /*
2368 * Ext4 directory file types. Only the low 3 bits are used. The
2369 * other bits are reserved for now.
2370 */
2371 #define EXT4_FT_UNKNOWN 0
2372 #define EXT4_FT_REG_FILE 1
2373 #define EXT4_FT_DIR 2
2374 #define EXT4_FT_CHRDEV 3
2375 #define EXT4_FT_BLKDEV 4
2376 #define EXT4_FT_FIFO 5
2377 #define EXT4_FT_SOCK 6
2378 #define EXT4_FT_SYMLINK 7
2379
2380 #define EXT4_FT_MAX 8
2381
2382 #define EXT4_FT_DIR_CSUM 0xDE
2383
2384 /*
2385 * EXT4_DIR_PAD defines the directory entries boundaries
2386 *
2387 * NOTE: It must be a multiple of 4
2388 */
2389 #define EXT4_DIR_PAD 4
2390 #define EXT4_DIR_ROUND (EXT4_DIR_PAD - 1)
2391 #define EXT4_MAX_REC_LEN ((1<<16)-1)
2392
2393 /*
2394 * The rec_len is dependent on the type of directory. Directories that are
2395 * casefolded and encrypted need to store the hash as well, so we add room for
2396 * ext4_extended_dir_entry_2. For all entries related to '.' or '..' you should
2397 * pass NULL for dir, as those entries do not use the extra fields.
2398 */
2399 static inline unsigned int ext4_dir_rec_len(__u8 name_len,
2400 const struct inode *dir)
2401 {
2402 int rec_len = (name_len + 8 + EXT4_DIR_ROUND);
2403
2404 if (dir && ext4_hash_in_dirent(dir))
2405 rec_len += sizeof(struct ext4_dir_entry_hash);
2406 return (rec_len & ~EXT4_DIR_ROUND);
2407 }
2408
2409 /*
2410 * If we ever get support for fs block sizes > page_size, we'll need
2411 * to remove the #if statements in the next two functions...
2412 */
2413 static inline unsigned int
2414 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
2415 {
2416 unsigned len = le16_to_cpu(dlen);
2417
2418 #if (PAGE_SIZE >= 65536)
2419 if (len == EXT4_MAX_REC_LEN || len == 0)
2420 return blocksize;
2421 return (len & 65532) | ((len & 3) << 16);
2422 #else
2423 return len;
2424 #endif
2425 }
2426
2427 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
2428 {
2429 BUG_ON((len > blocksize) || (blocksize > (1 << 18)) || (len & 3));
2430 #if (PAGE_SIZE >= 65536)
2431 if (len < 65536)
2432 return cpu_to_le16(len);
2433 if (len == blocksize) {
2434 if (blocksize == 65536)
2435 return cpu_to_le16(EXT4_MAX_REC_LEN);
2436 else
2437 return cpu_to_le16(0);
2438 }
2439 return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
2440 #else
2441 return cpu_to_le16(len);
2442 #endif
2443 }
2444
2445 /*
2446 * Hash Tree Directory indexing
2447 * (c) Daniel Phillips, 2001
2448 */
2449
2450 #define is_dx(dir) (ext4_has_feature_dir_index((dir)->i_sb) && \
2451 ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
2452 #define EXT4_DIR_LINK_MAX(dir) unlikely((dir)->i_nlink >= EXT4_LINK_MAX && \
2453 !(ext4_has_feature_dir_nlink((dir)->i_sb) && is_dx(dir)))
2454 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
2455
2456 /* Legal values for the dx_root hash_version field: */
2457
2458 #define DX_HASH_LEGACY 0
2459 #define DX_HASH_HALF_MD4 1
2460 #define DX_HASH_TEA 2
2461 #define DX_HASH_LEGACY_UNSIGNED 3
2462 #define DX_HASH_HALF_MD4_UNSIGNED 4
2463 #define DX_HASH_TEA_UNSIGNED 5
2464 #define DX_HASH_SIPHASH 6
2465
2466 static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
2467 const void *address, unsigned int length)
2468 {
2469 struct {
2470 struct shash_desc shash;
2471 char ctx[4];
2472 } desc;
2473
2474 BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
2475
2476 desc.shash.tfm = sbi->s_chksum_driver;
2477 *(u32 *)desc.ctx = crc;
2478
2479 BUG_ON(crypto_shash_update(&desc.shash, address, length));
2480
2481 return *(u32 *)desc.ctx;
2482 }
2483
2484 #ifdef __KERNEL__
2485
2486 /* hash info structure used by the directory hash */
2487 struct dx_hash_info
2488 {
2489 u32 hash;
2490 u32 minor_hash;
2491 int hash_version;
2492 u32 *seed;
2493 };
2494
2495
2496 /* 32 and 64 bit signed EOF for dx directories */
2497 #define EXT4_HTREE_EOF_32BIT ((1UL << (32 - 1)) - 1)
2498 #define EXT4_HTREE_EOF_64BIT ((1ULL << (64 - 1)) - 1)
2499
2500
2501 /*
2502 * Control parameters used by ext4_htree_next_block
2503 */
2504 #define HASH_NB_ALWAYS 1
2505
2506 struct ext4_filename {
2507 const struct qstr *usr_fname;
2508 struct fscrypt_str disk_name;
2509 struct dx_hash_info hinfo;
2510 #ifdef CONFIG_FS_ENCRYPTION
2511 struct fscrypt_str crypto_buf;
2512 #endif
2513 #if IS_ENABLED(CONFIG_UNICODE)
2514 struct qstr cf_name;
2515 #endif
2516 };
2517
2518 #define fname_name(p) ((p)->disk_name.name)
2519 #define fname_usr_name(p) ((p)->usr_fname->name)
2520 #define fname_len(p) ((p)->disk_name.len)
2521
2522 /*
2523 * Describe an inode's exact location on disk and in memory
2524 */
2525 struct ext4_iloc
2526 {
2527 struct buffer_head *bh;
2528 unsigned long offset;
2529 ext4_group_t block_group;
2530 };
2531
2532 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
2533 {
2534 return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
2535 }
2536
2537 static inline bool ext4_is_quota_file(struct inode *inode)
2538 {
2539 return IS_NOQUOTA(inode) &&
2540 !(EXT4_I(inode)->i_flags & EXT4_EA_INODE_FL);
2541 }
2542
2543 /*
2544 * This structure is stuffed into the struct file's private_data field
2545 * for directories. It is where we put information so that we can do
2546 * readdir operations in hash tree order.
2547 */
2548 struct dir_private_info {
2549 struct rb_root root;
2550 struct rb_node *curr_node;
2551 struct fname *extra_fname;
2552 loff_t last_pos;
2553 __u32 curr_hash;
2554 __u32 curr_minor_hash;
2555 __u32 next_hash;
2556 };
2557
2558 /* calculate the first block number of the group */
2559 static inline ext4_fsblk_t
2560 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
2561 {
2562 return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
2563 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
2564 }
2565
2566 /*
2567 * Special error return code only used by dx_probe() and its callers.
2568 */
2569 #define ERR_BAD_DX_DIR (-(MAX_ERRNO - 1))
2570
2571 /* htree levels for ext4 */
2572 #define EXT4_HTREE_LEVEL_COMPAT 2
2573 #define EXT4_HTREE_LEVEL 3
2574
2575 static inline int ext4_dir_htree_level(struct super_block *sb)
2576 {
2577 return ext4_has_feature_largedir(sb) ?
2578 EXT4_HTREE_LEVEL : EXT4_HTREE_LEVEL_COMPAT;
2579 }
2580
2581 /*
2582 * Timeout and state flag for lazy initialization inode thread.
2583 */
2584 #define EXT4_DEF_LI_WAIT_MULT 10
2585 #define EXT4_DEF_LI_MAX_START_DELAY 5
2586 #define EXT4_LAZYINIT_QUIT 0x0001
2587 #define EXT4_LAZYINIT_RUNNING 0x0002
2588
2589 /*
2590 * Lazy inode table initialization info
2591 */
2592 struct ext4_lazy_init {
2593 unsigned long li_state;
2594 struct list_head li_request_list;
2595 struct mutex li_list_mtx;
2596 };
2597
2598 enum ext4_li_mode {
2599 EXT4_LI_MODE_PREFETCH_BBITMAP,
2600 EXT4_LI_MODE_ITABLE,
2601 };
2602
2603 struct ext4_li_request {
2604 struct super_block *lr_super;
2605 enum ext4_li_mode lr_mode;
2606 ext4_group_t lr_first_not_zeroed;
2607 ext4_group_t lr_next_group;
2608 struct list_head lr_request;
2609 unsigned long lr_next_sched;
2610 unsigned long lr_timeout;
2611 };
2612
2613 struct ext4_features {
2614 struct kobject f_kobj;
2615 struct completion f_kobj_unregister;
2616 };
2617
2618 /*
2619 * This structure will be used for multiple mount protection. It will be
2620 * written into the block number saved in the s_mmp_block field in the
2621 * superblock. Programs that check MMP should assume that if
2622 * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
2623 * to use the filesystem, regardless of how old the timestamp is.
2624 */
2625 #define EXT4_MMP_MAGIC 0x004D4D50U /* ASCII for MMP */
2626 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
2627 #define EXT4_MMP_SEQ_FSCK 0xE24D4D50U /* mmp_seq value when being fscked */
2628 #define EXT4_MMP_SEQ_MAX 0xE24D4D4FU /* maximum valid mmp_seq value */
2629
2630 struct mmp_struct {
2631 __le32 mmp_magic; /* Magic number for MMP */
2632 __le32 mmp_seq; /* Sequence no. updated periodically */
2633
2634 /*
2635 * mmp_time, mmp_nodename & mmp_bdevname are only used for information
2636 * purposes and do not affect the correctness of the algorithm
2637 */
2638 __le64 mmp_time; /* Time last updated */
2639 char mmp_nodename[64]; /* Node which last updated MMP block */
2640 char mmp_bdevname[32]; /* Bdev which last updated MMP block */
2641
2642 /*
2643 * mmp_check_interval is used to verify if the MMP block has been
2644 * updated on the block device. The value is updated based on the
2645 * maximum time to write the MMP block during an update cycle.
2646 */
2647 __le16 mmp_check_interval;
2648
2649 __le16 mmp_pad1;
2650 __le32 mmp_pad2[226];
2651 __le32 mmp_checksum; /* crc32c(uuid+mmp_block) */
2652 };
2653
2654 /* arguments passed to the mmp thread */
2655 struct mmpd_data {
2656 struct buffer_head *bh; /* bh from initial read_mmp_block() */
2657 struct super_block *sb; /* super block of the fs */
2658 };
2659
2660 /*
2661 * Check interval multiplier
2662 * The MMP block is written every update interval and initially checked every
2663 * update interval x the multiplier (the value is then adapted based on the
2664 * write latency). The reason is that writes can be delayed under load and we
2665 * don't want readers to incorrectly assume that the filesystem is no longer
2666 * in use.
2667 */
2668 #define EXT4_MMP_CHECK_MULT 2UL
2669
2670 /*
2671 * Minimum interval for MMP checking in seconds.
2672 */
2673 #define EXT4_MMP_MIN_CHECK_INTERVAL 5UL
2674
2675 /*
2676 * Maximum interval for MMP checking in seconds.
2677 */
2678 #define EXT4_MMP_MAX_CHECK_INTERVAL 300UL
2679
2680 /*
2681 * Function prototypes
2682 */
2683
2684 /*
2685 * Ok, these declarations are also in <linux/kernel.h> but none of the
2686 * ext4 source programs needs to include it so they are duplicated here.
2687 */
2688 # define NORET_TYPE /**/
2689 # define ATTRIB_NORET __attribute__((noreturn))
2690 # define NORET_AND noreturn,
2691
2692 /* bitmap.c */
2693 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
2694 void ext4_inode_bitmap_csum_set(struct super_block *sb,
2695 struct ext4_group_desc *gdp,
2696 struct buffer_head *bh, int sz);
2697 int ext4_inode_bitmap_csum_verify(struct super_block *sb,
2698 struct ext4_group_desc *gdp,
2699 struct buffer_head *bh, int sz);
2700 void ext4_block_bitmap_csum_set(struct super_block *sb,
2701 struct ext4_group_desc *gdp,
2702 struct buffer_head *bh);
2703 int ext4_block_bitmap_csum_verify(struct super_block *sb,
2704 struct ext4_group_desc *gdp,
2705 struct buffer_head *bh);
2706
2707 /* balloc.c */
2708 extern void ext4_get_group_no_and_offset(struct super_block *sb,
2709 ext4_fsblk_t blocknr,
2710 ext4_group_t *blockgrpp,
2711 ext4_grpblk_t *offsetp);
2712 extern ext4_group_t ext4_get_group_number(struct super_block *sb,
2713 ext4_fsblk_t block);
2714
2715 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
2716 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
2717 ext4_group_t group);
2718 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
2719 ext4_fsblk_t goal,
2720 unsigned int flags,
2721 unsigned long *count,
2722 int *errp);
2723 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
2724 s64 nclusters, unsigned int flags);
2725 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
2726 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
2727 ext4_group_t block_group,
2728 struct buffer_head ** bh);
2729 extern struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2730 ext4_group_t group);
2731 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
2732
2733 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
2734 ext4_group_t block_group,
2735 bool ignore_locked);
2736 extern int ext4_wait_block_bitmap(struct super_block *sb,
2737 ext4_group_t block_group,
2738 struct buffer_head *bh);
2739 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
2740 ext4_group_t block_group);
2741 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
2742 ext4_group_t block_group,
2743 struct ext4_group_desc *gdp);
2744 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
2745
2746 #if IS_ENABLED(CONFIG_UNICODE)
2747 extern int ext4_fname_setup_ci_filename(struct inode *dir,
2748 const struct qstr *iname,
2749 struct ext4_filename *fname);
2750
2751 static inline void ext4_fname_free_ci_filename(struct ext4_filename *fname)
2752 {
2753 kfree(fname->cf_name.name);
2754 fname->cf_name.name = NULL;
2755 }
2756 #else
2757 static inline int ext4_fname_setup_ci_filename(struct inode *dir,
2758 const struct qstr *iname,
2759 struct ext4_filename *fname)
2760 {
2761 return 0;
2762 }
2763
2764 static inline void ext4_fname_free_ci_filename(struct ext4_filename *fname)
2765 {
2766 }
2767 #endif
2768
2769 /* ext4 encryption related stuff goes here crypto.c */
2770 #ifdef CONFIG_FS_ENCRYPTION
2771 extern const struct fscrypt_operations ext4_cryptops;
2772
2773 int ext4_fname_setup_filename(struct inode *dir, const struct qstr *iname,
2774 int lookup, struct ext4_filename *fname);
2775
2776 int ext4_fname_prepare_lookup(struct inode *dir, struct dentry *dentry,
2777 struct ext4_filename *fname);
2778
2779 void ext4_fname_free_filename(struct ext4_filename *fname);
2780
2781 int ext4_ioctl_get_encryption_pwsalt(struct file *filp, void __user *arg);
2782
2783 #else /* !CONFIG_FS_ENCRYPTION */
2784 static inline int ext4_fname_setup_filename(struct inode *dir,
2785 const struct qstr *iname,
2786 int lookup,
2787 struct ext4_filename *fname)
2788 {
2789 fname->usr_fname = iname;
2790 fname->disk_name.name = (unsigned char *) iname->name;
2791 fname->disk_name.len = iname->len;
2792
2793 return ext4_fname_setup_ci_filename(dir, iname, fname);
2794 }
2795
2796 static inline int ext4_fname_prepare_lookup(struct inode *dir,
2797 struct dentry *dentry,
2798 struct ext4_filename *fname)
2799 {
2800 return ext4_fname_setup_filename(dir, &dentry->d_name, 1, fname);
2801 }
2802
2803 static inline void ext4_fname_free_filename(struct ext4_filename *fname)
2804 {
2805 ext4_fname_free_ci_filename(fname);
2806 }
2807
2808 static inline int ext4_ioctl_get_encryption_pwsalt(struct file *filp,
2809 void __user *arg)
2810 {
2811 return -EOPNOTSUPP;
2812 }
2813 #endif /* !CONFIG_FS_ENCRYPTION */
2814
2815 /* dir.c */
2816 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
2817 struct file *,
2818 struct ext4_dir_entry_2 *,
2819 struct buffer_head *, char *, int,
2820 unsigned int);
2821 #define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset) \
2822 unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
2823 (de), (bh), (buf), (size), (offset)))
2824 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
2825 __u32 minor_hash,
2826 struct ext4_dir_entry_2 *dirent,
2827 struct fscrypt_str *ent_name);
2828 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
2829 extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
2830 struct buffer_head *bh,
2831 void *buf, int buf_size,
2832 struct ext4_filename *fname,
2833 struct ext4_dir_entry_2 **dest_de);
2834 void ext4_insert_dentry(struct inode *dir, struct inode *inode,
2835 struct ext4_dir_entry_2 *de,
2836 int buf_size,
2837 struct ext4_filename *fname);
2838 static inline void ext4_update_dx_flag(struct inode *inode)
2839 {
2840 if (!ext4_has_feature_dir_index(inode->i_sb) &&
2841 ext4_test_inode_flag(inode, EXT4_INODE_INDEX)) {
2842 /* ext4_iget() should have caught this... */
2843 WARN_ON_ONCE(ext4_has_feature_metadata_csum(inode->i_sb));
2844 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
2845 }
2846 }
2847 static const unsigned char ext4_filetype_table[] = {
2848 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
2849 };
2850
2851 static inline unsigned char get_dtype(struct super_block *sb, int filetype)
2852 {
2853 if (!ext4_has_feature_filetype(sb) || filetype >= EXT4_FT_MAX)
2854 return DT_UNKNOWN;
2855
2856 return ext4_filetype_table[filetype];
2857 }
2858 extern int ext4_check_all_de(struct inode *dir, struct buffer_head *bh,
2859 void *buf, int buf_size);
2860
2861 /* fsync.c */
2862 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2863
2864 /* hash.c */
2865 extern int ext4fs_dirhash(const struct inode *dir, const char *name, int len,
2866 struct dx_hash_info *hinfo);
2867
2868 /* ialloc.c */
2869 extern int ext4_mark_inode_used(struct super_block *sb, int ino);
2870 extern struct inode *__ext4_new_inode(struct mnt_idmap *, handle_t *,
2871 struct inode *, umode_t,
2872 const struct qstr *qstr, __u32 goal,
2873 uid_t *owner, __u32 i_flags,
2874 int handle_type, unsigned int line_no,
2875 int nblocks);
2876
2877 #define ext4_new_inode(handle, dir, mode, qstr, goal, owner, i_flags) \
2878 __ext4_new_inode(&nop_mnt_idmap, (handle), (dir), (mode), (qstr), \
2879 (goal), (owner), i_flags, 0, 0, 0)
2880 #define ext4_new_inode_start_handle(idmap, dir, mode, qstr, goal, owner, \
2881 type, nblocks) \
2882 __ext4_new_inode((idmap), NULL, (dir), (mode), (qstr), (goal), (owner), \
2883 0, (type), __LINE__, (nblocks))
2884
2885
2886 extern void ext4_free_inode(handle_t *, struct inode *);
2887 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2888 extern unsigned long ext4_count_free_inodes(struct super_block *);
2889 extern unsigned long ext4_count_dirs(struct super_block *);
2890 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2891 extern int ext4_init_inode_table(struct super_block *sb,
2892 ext4_group_t group, int barrier);
2893 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2894
2895 /* fast_commit.c */
2896 int ext4_fc_info_show(struct seq_file *seq, void *v);
2897 void ext4_fc_init(struct super_block *sb, journal_t *journal);
2898 void ext4_fc_init_inode(struct inode *inode);
2899 void ext4_fc_track_range(handle_t *handle, struct inode *inode, ext4_lblk_t start,
2900 ext4_lblk_t end);
2901 void __ext4_fc_track_unlink(handle_t *handle, struct inode *inode,
2902 struct dentry *dentry);
2903 void __ext4_fc_track_link(handle_t *handle, struct inode *inode,
2904 struct dentry *dentry);
2905 void ext4_fc_track_unlink(handle_t *handle, struct dentry *dentry);
2906 void ext4_fc_track_link(handle_t *handle, struct dentry *dentry);
2907 void __ext4_fc_track_create(handle_t *handle, struct inode *inode,
2908 struct dentry *dentry);
2909 void ext4_fc_track_create(handle_t *handle, struct dentry *dentry);
2910 void ext4_fc_track_inode(handle_t *handle, struct inode *inode);
2911 void ext4_fc_mark_ineligible(struct super_block *sb, int reason, handle_t *handle);
2912 void ext4_fc_start_update(struct inode *inode);
2913 void ext4_fc_stop_update(struct inode *inode);
2914 void ext4_fc_del(struct inode *inode);
2915 bool ext4_fc_replay_check_excluded(struct super_block *sb, ext4_fsblk_t block);
2916 void ext4_fc_replay_cleanup(struct super_block *sb);
2917 int ext4_fc_commit(journal_t *journal, tid_t commit_tid);
2918 int __init ext4_fc_init_dentry_cache(void);
2919 void ext4_fc_destroy_dentry_cache(void);
2920 int ext4_fc_record_regions(struct super_block *sb, int ino,
2921 ext4_lblk_t lblk, ext4_fsblk_t pblk,
2922 int len, int replay);
2923
2924 /* mballoc.c */
2925 extern const struct seq_operations ext4_mb_seq_groups_ops;
2926 extern const struct seq_operations ext4_mb_seq_structs_summary_ops;
2927 extern int ext4_seq_mb_stats_show(struct seq_file *seq, void *offset);
2928 extern int ext4_mb_init(struct super_block *);
2929 extern void ext4_mb_release(struct super_block *);
2930 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2931 struct ext4_allocation_request *, int *);
2932 extern void ext4_discard_preallocations(struct inode *);
2933 extern int __init ext4_init_mballoc(void);
2934 extern void ext4_exit_mballoc(void);
2935 extern ext4_group_t ext4_mb_prefetch(struct super_block *sb,
2936 ext4_group_t group,
2937 unsigned int nr, int *cnt);
2938 extern void ext4_mb_prefetch_fini(struct super_block *sb, ext4_group_t group,
2939 unsigned int nr);
2940
2941 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2942 struct buffer_head *bh, ext4_fsblk_t block,
2943 unsigned long count, int flags);
2944 extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2945 ext4_group_t ngroups);
2946 extern int ext4_mb_add_groupinfo(struct super_block *sb,
2947 ext4_group_t i, struct ext4_group_desc *desc);
2948 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2949 ext4_fsblk_t block, unsigned long count);
2950 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2951 extern void ext4_process_freed_data(struct super_block *sb, tid_t commit_tid);
2952 extern void ext4_mb_mark_bb(struct super_block *sb, ext4_fsblk_t block,
2953 int len, bool state);
2954 static inline bool ext4_mb_cr_expensive(enum criteria cr)
2955 {
2956 return cr >= CR_GOAL_LEN_SLOW;
2957 }
2958
2959 /* inode.c */
2960 void ext4_inode_csum_set(struct inode *inode, struct ext4_inode *raw,
2961 struct ext4_inode_info *ei);
2962 int ext4_inode_is_fast_symlink(struct inode *inode);
2963 struct buffer_head *ext4_getblk(handle_t *, struct inode *, ext4_lblk_t, int);
2964 struct buffer_head *ext4_bread(handle_t *, struct inode *, ext4_lblk_t, int);
2965 int ext4_bread_batch(struct inode *inode, ext4_lblk_t block, int bh_count,
2966 bool wait, struct buffer_head **bhs);
2967 int ext4_get_block_unwritten(struct inode *inode, sector_t iblock,
2968 struct buffer_head *bh_result, int create);
2969 int ext4_get_block(struct inode *inode, sector_t iblock,
2970 struct buffer_head *bh_result, int create);
2971 int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2972 struct buffer_head *bh, int create);
2973 int ext4_walk_page_buffers(handle_t *handle,
2974 struct inode *inode,
2975 struct buffer_head *head,
2976 unsigned from,
2977 unsigned to,
2978 int *partial,
2979 int (*fn)(handle_t *handle, struct inode *inode,
2980 struct buffer_head *bh));
2981 int do_journal_get_write_access(handle_t *handle, struct inode *inode,
2982 struct buffer_head *bh);
2983 #define FALL_BACK_TO_NONDELALLOC 1
2984 #define CONVERT_INLINE_DATA 2
2985
2986 typedef enum {
2987 EXT4_IGET_NORMAL = 0,
2988 EXT4_IGET_SPECIAL = 0x0001, /* OK to iget a system inode */
2989 EXT4_IGET_HANDLE = 0x0002, /* Inode # is from a handle */
2990 EXT4_IGET_BAD = 0x0004, /* Allow to iget a bad inode */
2991 EXT4_IGET_EA_INODE = 0x0008 /* Inode should contain an EA value */
2992 } ext4_iget_flags;
2993
2994 extern struct inode *__ext4_iget(struct super_block *sb, unsigned long ino,
2995 ext4_iget_flags flags, const char *function,
2996 unsigned int line);
2997
2998 #define ext4_iget(sb, ino, flags) \
2999 __ext4_iget((sb), (ino), (flags), __func__, __LINE__)
3000
3001 extern int ext4_write_inode(struct inode *, struct writeback_control *);
3002 extern int ext4_setattr(struct mnt_idmap *, struct dentry *,
3003 struct iattr *);
3004 extern u32 ext4_dio_alignment(struct inode *inode);
3005 extern int ext4_getattr(struct mnt_idmap *, const struct path *,
3006 struct kstat *, u32, unsigned int);
3007 extern void ext4_evict_inode(struct inode *);
3008 extern void ext4_clear_inode(struct inode *);
3009 extern int ext4_file_getattr(struct mnt_idmap *, const struct path *,
3010 struct kstat *, u32, unsigned int);
3011 extern void ext4_dirty_inode(struct inode *, int);
3012 extern int ext4_change_inode_journal_flag(struct inode *, int);
3013 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
3014 extern int ext4_get_fc_inode_loc(struct super_block *sb, unsigned long ino,
3015 struct ext4_iloc *iloc);
3016 extern int ext4_inode_attach_jinode(struct inode *inode);
3017 extern int ext4_can_truncate(struct inode *inode);
3018 extern int ext4_truncate(struct inode *);
3019 extern int ext4_break_layouts(struct inode *);
3020 extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length);
3021 extern void ext4_set_inode_flags(struct inode *, bool init);
3022 extern int ext4_alloc_da_blocks(struct inode *inode);
3023 extern void ext4_set_aops(struct inode *inode);
3024 extern int ext4_writepage_trans_blocks(struct inode *);
3025 extern int ext4_normal_submit_inode_data_buffers(struct jbd2_inode *jinode);
3026 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
3027 extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
3028 loff_t lstart, loff_t lend);
3029 extern vm_fault_t ext4_page_mkwrite(struct vm_fault *vmf);
3030 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
3031 extern int ext4_get_projid(struct inode *inode, kprojid_t *projid);
3032 extern void ext4_da_release_space(struct inode *inode, int to_free);
3033 extern void ext4_da_update_reserve_space(struct inode *inode,
3034 int used, int quota_claim);
3035 extern int ext4_issue_zeroout(struct inode *inode, ext4_lblk_t lblk,
3036 ext4_fsblk_t pblk, ext4_lblk_t len);
3037
3038 /* indirect.c */
3039 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
3040 struct ext4_map_blocks *map, int flags);
3041 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
3042 extern void ext4_ind_truncate(handle_t *, struct inode *inode);
3043 extern int ext4_ind_remove_space(handle_t *handle, struct inode *inode,
3044 ext4_lblk_t start, ext4_lblk_t end);
3045
3046 /* ioctl.c */
3047 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
3048 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
3049 int ext4_fileattr_set(struct mnt_idmap *idmap,
3050 struct dentry *dentry, struct fileattr *fa);
3051 int ext4_fileattr_get(struct dentry *dentry, struct fileattr *fa);
3052 extern void ext4_reset_inode_seed(struct inode *inode);
3053 int ext4_update_overhead(struct super_block *sb, bool force);
3054 int ext4_force_shutdown(struct super_block *sb, u32 flags);
3055
3056 /* migrate.c */
3057 extern int ext4_ext_migrate(struct inode *);
3058 extern int ext4_ind_migrate(struct inode *inode);
3059
3060 /* namei.c */
3061 extern int ext4_init_new_dir(handle_t *handle, struct inode *dir,
3062 struct inode *inode);
3063 extern int ext4_dirblock_csum_verify(struct inode *inode,
3064 struct buffer_head *bh);
3065 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
3066 __u32 start_minor_hash, __u32 *next_hash);
3067 extern int ext4_search_dir(struct buffer_head *bh,
3068 char *search_buf,
3069 int buf_size,
3070 struct inode *dir,
3071 struct ext4_filename *fname,
3072 unsigned int offset,
3073 struct ext4_dir_entry_2 **res_dir);
3074 extern int ext4_generic_delete_entry(struct inode *dir,
3075 struct ext4_dir_entry_2 *de_del,
3076 struct buffer_head *bh,
3077 void *entry_buf,
3078 int buf_size,
3079 int csum_size);
3080 extern bool ext4_empty_dir(struct inode *inode);
3081
3082 /* resize.c */
3083 extern void ext4_kvfree_array_rcu(void *to_free);
3084 extern int ext4_group_add(struct super_block *sb,
3085 struct ext4_new_group_data *input);
3086 extern int ext4_group_extend(struct super_block *sb,
3087 struct ext4_super_block *es,
3088 ext4_fsblk_t n_blocks_count);
3089 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
3090 extern unsigned int ext4_list_backups(struct super_block *sb,
3091 unsigned int *three, unsigned int *five,
3092 unsigned int *seven);
3093
3094 /* super.c */
3095 extern struct buffer_head *ext4_sb_bread(struct super_block *sb,
3096 sector_t block, blk_opf_t op_flags);
3097 extern struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
3098 sector_t block);
3099 extern void ext4_read_bh_nowait(struct buffer_head *bh, blk_opf_t op_flags,
3100 bh_end_io_t *end_io);
3101 extern int ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags,
3102 bh_end_io_t *end_io);
3103 extern int ext4_read_bh_lock(struct buffer_head *bh, blk_opf_t op_flags, bool wait);
3104 extern void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block);
3105 extern int ext4_seq_options_show(struct seq_file *seq, void *offset);
3106 extern int ext4_calculate_overhead(struct super_block *sb);
3107 extern __le32 ext4_superblock_csum(struct super_block *sb,
3108 struct ext4_super_block *es);
3109 extern void ext4_superblock_csum_set(struct super_block *sb);
3110 extern int ext4_alloc_flex_bg_array(struct super_block *sb,
3111 ext4_group_t ngroup);
3112 extern const char *ext4_decode_error(struct super_block *sb, int errno,
3113 char nbuf[16]);
3114 extern void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
3115 ext4_group_t block_group,
3116 unsigned int flags);
3117 extern unsigned int ext4_num_base_meta_blocks(struct super_block *sb,
3118 ext4_group_t block_group);
3119
3120 extern __printf(7, 8)
3121 void __ext4_error(struct super_block *, const char *, unsigned int, bool,
3122 int, __u64, const char *, ...);
3123 extern __printf(6, 7)
3124 void __ext4_error_inode(struct inode *, const char *, unsigned int,
3125 ext4_fsblk_t, int, const char *, ...);
3126 extern __printf(5, 6)
3127 void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
3128 const char *, ...);
3129 extern void __ext4_std_error(struct super_block *, const char *,
3130 unsigned int, int);
3131 extern __printf(4, 5)
3132 void __ext4_warning(struct super_block *, const char *, unsigned int,
3133 const char *, ...);
3134 extern __printf(4, 5)
3135 void __ext4_warning_inode(const struct inode *inode, const char *function,
3136 unsigned int line, const char *fmt, ...);
3137 extern __printf(3, 4)
3138 void __ext4_msg(struct super_block *, const char *, const char *, ...);
3139 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
3140 const char *, unsigned int, const char *);
3141 extern __printf(7, 8)
3142 void __ext4_grp_locked_error(const char *, unsigned int,
3143 struct super_block *, ext4_group_t,
3144 unsigned long, ext4_fsblk_t,
3145 const char *, ...);
3146
3147 #define EXT4_ERROR_INODE(inode, fmt, a...) \
3148 ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
3149
3150 #define EXT4_ERROR_INODE_ERR(inode, err, fmt, a...) \
3151 __ext4_error_inode((inode), __func__, __LINE__, 0, (err), (fmt), ## a)
3152
3153 #define ext4_error_inode_block(inode, block, err, fmt, a...) \
3154 __ext4_error_inode((inode), __func__, __LINE__, (block), (err), \
3155 (fmt), ## a)
3156
3157 #define EXT4_ERROR_FILE(file, block, fmt, a...) \
3158 ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
3159
3160 #define ext4_abort(sb, err, fmt, a...) \
3161 __ext4_error((sb), __func__, __LINE__, true, (err), 0, (fmt), ## a)
3162
3163 #ifdef CONFIG_PRINTK
3164
3165 #define ext4_error_inode(inode, func, line, block, fmt, ...) \
3166 __ext4_error_inode(inode, func, line, block, 0, fmt, ##__VA_ARGS__)
3167 #define ext4_error_inode_err(inode, func, line, block, err, fmt, ...) \
3168 __ext4_error_inode((inode), (func), (line), (block), \
3169 (err), (fmt), ##__VA_ARGS__)
3170 #define ext4_error_file(file, func, line, block, fmt, ...) \
3171 __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__)
3172 #define ext4_error(sb, fmt, ...) \
3173 __ext4_error((sb), __func__, __LINE__, false, 0, 0, (fmt), \
3174 ##__VA_ARGS__)
3175 #define ext4_error_err(sb, err, fmt, ...) \
3176 __ext4_error((sb), __func__, __LINE__, false, (err), 0, (fmt), \
3177 ##__VA_ARGS__)
3178 #define ext4_warning(sb, fmt, ...) \
3179 __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
3180 #define ext4_warning_inode(inode, fmt, ...) \
3181 __ext4_warning_inode(inode, __func__, __LINE__, fmt, ##__VA_ARGS__)
3182 #define ext4_msg(sb, level, fmt, ...) \
3183 __ext4_msg(sb, level, fmt, ##__VA_ARGS__)
3184 #define dump_mmp_msg(sb, mmp, msg) \
3185 __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg)
3186 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...) \
3187 __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \
3188 fmt, ##__VA_ARGS__)
3189
3190 #else
3191
3192 #define ext4_error_inode(inode, func, line, block, fmt, ...) \
3193 do { \
3194 no_printk(fmt, ##__VA_ARGS__); \
3195 __ext4_error_inode(inode, "", 0, block, 0, " "); \
3196 } while (0)
3197 #define ext4_error_inode_err(inode, func, line, block, err, fmt, ...) \
3198 do { \
3199 no_printk(fmt, ##__VA_ARGS__); \
3200 __ext4_error_inode(inode, "", 0, block, err, " "); \
3201 } while (0)
3202 #define ext4_error_file(file, func, line, block, fmt, ...) \
3203 do { \
3204 no_printk(fmt, ##__VA_ARGS__); \
3205 __ext4_error_file(file, "", 0, block, " "); \
3206 } while (0)
3207 #define ext4_error(sb, fmt, ...) \
3208 do { \
3209 no_printk(fmt, ##__VA_ARGS__); \
3210 __ext4_error(sb, "", 0, false, 0, 0, " "); \
3211 } while (0)
3212 #define ext4_error_err(sb, err, fmt, ...) \
3213 do { \
3214 no_printk(fmt, ##__VA_ARGS__); \
3215 __ext4_error(sb, "", 0, false, err, 0, " "); \
3216 } while (0)
3217 #define ext4_warning(sb, fmt, ...) \
3218 do { \
3219 no_printk(fmt, ##__VA_ARGS__); \
3220 __ext4_warning(sb, "", 0, " "); \
3221 } while (0)
3222 #define ext4_warning_inode(inode, fmt, ...) \
3223 do { \
3224 no_printk(fmt, ##__VA_ARGS__); \
3225 __ext4_warning_inode(inode, "", 0, " "); \
3226 } while (0)
3227 #define ext4_msg(sb, level, fmt, ...) \
3228 do { \
3229 no_printk(fmt, ##__VA_ARGS__); \
3230 __ext4_msg(sb, "", " "); \
3231 } while (0)
3232 #define dump_mmp_msg(sb, mmp, msg) \
3233 __dump_mmp_msg(sb, mmp, "", 0, "")
3234 #define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...) \
3235 do { \
3236 no_printk(fmt, ##__VA_ARGS__); \
3237 __ext4_grp_locked_error("", 0, sb, grp, ino, block, " "); \
3238 } while (0)
3239
3240 #endif
3241
3242 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
3243 struct ext4_group_desc *bg);
3244 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
3245 struct ext4_group_desc *bg);
3246 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
3247 struct ext4_group_desc *bg);
3248 extern __u32 ext4_free_group_clusters(struct super_block *sb,
3249 struct ext4_group_desc *bg);
3250 extern __u32 ext4_free_inodes_count(struct super_block *sb,
3251 struct ext4_group_desc *bg);
3252 extern __u32 ext4_used_dirs_count(struct super_block *sb,
3253 struct ext4_group_desc *bg);
3254 extern __u32 ext4_itable_unused_count(struct super_block *sb,
3255 struct ext4_group_desc *bg);
3256 extern void ext4_block_bitmap_set(struct super_block *sb,
3257 struct ext4_group_desc *bg, ext4_fsblk_t blk);
3258 extern void ext4_inode_bitmap_set(struct super_block *sb,
3259 struct ext4_group_desc *bg, ext4_fsblk_t blk);
3260 extern void ext4_inode_table_set(struct super_block *sb,
3261 struct ext4_group_desc *bg, ext4_fsblk_t blk);
3262 extern void ext4_free_group_clusters_set(struct super_block *sb,
3263 struct ext4_group_desc *bg,
3264 __u32 count);
3265 extern void ext4_free_inodes_set(struct super_block *sb,
3266 struct ext4_group_desc *bg, __u32 count);
3267 extern void ext4_used_dirs_set(struct super_block *sb,
3268 struct ext4_group_desc *bg, __u32 count);
3269 extern void ext4_itable_unused_set(struct super_block *sb,
3270 struct ext4_group_desc *bg, __u32 count);
3271 extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
3272 struct ext4_group_desc *gdp);
3273 extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
3274 struct ext4_group_desc *gdp);
3275 extern int ext4_register_li_request(struct super_block *sb,
3276 ext4_group_t first_not_zeroed);
3277
3278 static inline int ext4_has_metadata_csum(struct super_block *sb)
3279 {
3280 WARN_ON_ONCE(ext4_has_feature_metadata_csum(sb) &&
3281 !EXT4_SB(sb)->s_chksum_driver);
3282
3283 return ext4_has_feature_metadata_csum(sb) &&
3284 (EXT4_SB(sb)->s_chksum_driver != NULL);
3285 }
3286
3287 static inline int ext4_has_group_desc_csum(struct super_block *sb)
3288 {
3289 return ext4_has_feature_gdt_csum(sb) || ext4_has_metadata_csum(sb);
3290 }
3291
3292 #define ext4_read_incompat_64bit_val(es, name) \
3293 (((es)->s_feature_incompat & cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT) \
3294 ? (ext4_fsblk_t)le32_to_cpu(es->name##_hi) << 32 : 0) | \
3295 le32_to_cpu(es->name##_lo))
3296
3297 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
3298 {
3299 return ext4_read_incompat_64bit_val(es, s_blocks_count);
3300 }
3301
3302 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
3303 {
3304 return ext4_read_incompat_64bit_val(es, s_r_blocks_count);
3305 }
3306
3307 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
3308 {
3309 return ext4_read_incompat_64bit_val(es, s_free_blocks_count);
3310 }
3311
3312 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
3313 ext4_fsblk_t blk)
3314 {
3315 es->s_blocks_count_lo = cpu_to_le32((u32)blk);
3316 es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
3317 }
3318
3319 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
3320 ext4_fsblk_t blk)
3321 {
3322 es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
3323 es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
3324 }
3325
3326 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
3327 ext4_fsblk_t blk)
3328 {
3329 es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
3330 es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
3331 }
3332
3333 static inline loff_t ext4_isize(struct super_block *sb,
3334 struct ext4_inode *raw_inode)
3335 {
3336 if (ext4_has_feature_largedir(sb) ||
3337 S_ISREG(le16_to_cpu(raw_inode->i_mode)))
3338 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
3339 le32_to_cpu(raw_inode->i_size_lo);
3340
3341 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
3342 }
3343
3344 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
3345 {
3346 raw_inode->i_size_lo = cpu_to_le32(i_size);
3347 raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
3348 }
3349
3350 /*
3351 * Reading s_groups_count requires using smp_rmb() afterwards. See
3352 * the locking protocol documented in the comments of ext4_group_add()
3353 * in resize.c
3354 */
3355 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
3356 {
3357 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3358
3359 smp_rmb();
3360 return ngroups;
3361 }
3362
3363 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
3364 ext4_group_t block_group)
3365 {
3366 return block_group >> sbi->s_log_groups_per_flex;
3367 }
3368
3369 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
3370 {
3371 return 1 << sbi->s_log_groups_per_flex;
3372 }
3373
3374 #define ext4_std_error(sb, errno) \
3375 do { \
3376 if ((errno)) \
3377 __ext4_std_error((sb), __func__, __LINE__, (errno)); \
3378 } while (0)
3379
3380 #ifdef CONFIG_SMP
3381 /* Each CPU can accumulate percpu_counter_batch clusters in their local
3382 * counters. So we need to make sure we have free clusters more
3383 * than percpu_counter_batch * nr_cpu_ids. Also add a window of 4 times.
3384 */
3385 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
3386 #else
3387 #define EXT4_FREECLUSTERS_WATERMARK 0
3388 #endif
3389
3390 /* Update i_disksize. Requires i_rwsem to avoid races with truncate */
3391 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
3392 {
3393 WARN_ON_ONCE(S_ISREG(inode->i_mode) &&
3394 !inode_is_locked(inode));
3395 down_write(&EXT4_I(inode)->i_data_sem);
3396 if (newsize > EXT4_I(inode)->i_disksize)
3397 WRITE_ONCE(EXT4_I(inode)->i_disksize, newsize);
3398 up_write(&EXT4_I(inode)->i_data_sem);
3399 }
3400
3401 /* Update i_size, i_disksize. Requires i_rwsem to avoid races with truncate */
3402 static inline int ext4_update_inode_size(struct inode *inode, loff_t newsize)
3403 {
3404 int changed = 0;
3405
3406 if (newsize > inode->i_size) {
3407 i_size_write(inode, newsize);
3408 changed = 1;
3409 }
3410 if (newsize > EXT4_I(inode)->i_disksize) {
3411 ext4_update_i_disksize(inode, newsize);
3412 changed |= 2;
3413 }
3414 return changed;
3415 }
3416
3417 int ext4_update_disksize_before_punch(struct inode *inode, loff_t offset,
3418 loff_t len);
3419
3420 struct ext4_group_info {
3421 unsigned long bb_state;
3422 #ifdef AGGRESSIVE_CHECK
3423 unsigned long bb_check_counter;
3424 #endif
3425 struct rb_root bb_free_root;
3426 ext4_grpblk_t bb_first_free; /* first free block */
3427 ext4_grpblk_t bb_free; /* total free blocks */
3428 ext4_grpblk_t bb_fragments; /* nr of freespace fragments */
3429 int bb_avg_fragment_size_order; /* order of average
3430 fragment in BG */
3431 ext4_grpblk_t bb_largest_free_order;/* order of largest frag in BG */
3432 ext4_group_t bb_group; /* Group number */
3433 struct list_head bb_prealloc_list;
3434 #ifdef DOUBLE_CHECK
3435 void *bb_bitmap;
3436 #endif
3437 struct rw_semaphore alloc_sem;
3438 struct list_head bb_avg_fragment_size_node;
3439 struct list_head bb_largest_free_order_node;
3440 ext4_grpblk_t bb_counters[]; /* Nr of free power-of-two-block
3441 * regions, index is order.
3442 * bb_counters[3] = 5 means
3443 * 5 free 8-block regions. */
3444 };
3445
3446 #define EXT4_GROUP_INFO_NEED_INIT_BIT 0
3447 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT 1
3448 #define EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT 2
3449 #define EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT 3
3450 #define EXT4_GROUP_INFO_BBITMAP_CORRUPT \
3451 (1 << EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT)
3452 #define EXT4_GROUP_INFO_IBITMAP_CORRUPT \
3453 (1 << EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT)
3454 #define EXT4_GROUP_INFO_BBITMAP_READ_BIT 4
3455
3456 #define EXT4_MB_GRP_NEED_INIT(grp) \
3457 (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
3458 #define EXT4_MB_GRP_BBITMAP_CORRUPT(grp) \
3459 (test_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
3460 #define EXT4_MB_GRP_IBITMAP_CORRUPT(grp) \
3461 (test_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
3462
3463 #define EXT4_MB_GRP_WAS_TRIMMED(grp) \
3464 (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3465 #define EXT4_MB_GRP_SET_TRIMMED(grp) \
3466 (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3467 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp) \
3468 (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3469 #define EXT4_MB_GRP_TEST_AND_SET_READ(grp) \
3470 (test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_READ_BIT, &((grp)->bb_state)))
3471
3472 #define EXT4_MAX_CONTENTION 8
3473 #define EXT4_CONTENTION_THRESHOLD 2
3474
3475 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
3476 ext4_group_t group)
3477 {
3478 return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
3479 }
3480
3481 /*
3482 * Returns true if the filesystem is busy enough that attempts to
3483 * access the block group locks has run into contention.
3484 */
3485 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
3486 {
3487 return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
3488 }
3489
3490 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
3491 {
3492 spinlock_t *lock = ext4_group_lock_ptr(sb, group);
3493 if (spin_trylock(lock))
3494 /*
3495 * We're able to grab the lock right away, so drop the
3496 * lock contention counter.
3497 */
3498 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
3499 else {
3500 /*
3501 * The lock is busy, so bump the contention counter,
3502 * and then wait on the spin lock.
3503 */
3504 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
3505 EXT4_MAX_CONTENTION);
3506 spin_lock(lock);
3507 }
3508 }
3509
3510 static inline void ext4_unlock_group(struct super_block *sb,
3511 ext4_group_t group)
3512 {
3513 spin_unlock(ext4_group_lock_ptr(sb, group));
3514 }
3515
3516 #ifdef CONFIG_QUOTA
3517 static inline bool ext4_quota_capable(struct super_block *sb)
3518 {
3519 return (test_opt(sb, QUOTA) || ext4_has_feature_quota(sb));
3520 }
3521
3522 static inline bool ext4_is_quota_journalled(struct super_block *sb)
3523 {
3524 struct ext4_sb_info *sbi = EXT4_SB(sb);
3525
3526 return (ext4_has_feature_quota(sb) ||
3527 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]);
3528 }
3529 int ext4_enable_quotas(struct super_block *sb);
3530 #endif
3531
3532 /*
3533 * Block validity checking
3534 */
3535 #define ext4_check_indirect_blockref(inode, bh) \
3536 ext4_check_blockref(__func__, __LINE__, inode, \
3537 (__le32 *)(bh)->b_data, \
3538 EXT4_ADDR_PER_BLOCK((inode)->i_sb))
3539
3540 #define ext4_ind_check_inode(inode) \
3541 ext4_check_blockref(__func__, __LINE__, inode, \
3542 EXT4_I(inode)->i_data, \
3543 EXT4_NDIR_BLOCKS)
3544
3545 /*
3546 * Inodes and files operations
3547 */
3548
3549 /* dir.c */
3550 extern const struct file_operations ext4_dir_operations;
3551
3552 /* file.c */
3553 extern const struct inode_operations ext4_file_inode_operations;
3554 extern const struct file_operations ext4_file_operations;
3555 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
3556
3557 /* inline.c */
3558 extern int ext4_get_max_inline_size(struct inode *inode);
3559 extern int ext4_find_inline_data_nolock(struct inode *inode);
3560 extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
3561
3562 int ext4_readpage_inline(struct inode *inode, struct folio *folio);
3563 extern int ext4_try_to_write_inline_data(struct address_space *mapping,
3564 struct inode *inode,
3565 loff_t pos, unsigned len,
3566 struct page **pagep);
3567 int ext4_write_inline_data_end(struct inode *inode, loff_t pos, unsigned len,
3568 unsigned copied, struct folio *folio);
3569 extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
3570 struct inode *inode,
3571 loff_t pos, unsigned len,
3572 struct page **pagep,
3573 void **fsdata);
3574 extern int ext4_try_add_inline_entry(handle_t *handle,
3575 struct ext4_filename *fname,
3576 struct inode *dir, struct inode *inode);
3577 extern int ext4_try_create_inline_dir(handle_t *handle,
3578 struct inode *parent,
3579 struct inode *inode);
3580 extern int ext4_read_inline_dir(struct file *filp,
3581 struct dir_context *ctx,
3582 int *has_inline_data);
3583 extern int ext4_inlinedir_to_tree(struct file *dir_file,
3584 struct inode *dir, ext4_lblk_t block,
3585 struct dx_hash_info *hinfo,
3586 __u32 start_hash, __u32 start_minor_hash,
3587 int *has_inline_data);
3588 extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
3589 struct ext4_filename *fname,
3590 struct ext4_dir_entry_2 **res_dir,
3591 int *has_inline_data);
3592 extern int ext4_delete_inline_entry(handle_t *handle,
3593 struct inode *dir,
3594 struct ext4_dir_entry_2 *de_del,
3595 struct buffer_head *bh,
3596 int *has_inline_data);
3597 extern bool empty_inline_dir(struct inode *dir, int *has_inline_data);
3598 extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
3599 struct ext4_dir_entry_2 **parent_de,
3600 int *retval);
3601 extern void *ext4_read_inline_link(struct inode *inode);
3602
3603 struct iomap;
3604 extern int ext4_inline_data_iomap(struct inode *inode, struct iomap *iomap);
3605
3606 extern int ext4_inline_data_truncate(struct inode *inode, int *has_inline);
3607
3608 extern int ext4_convert_inline_data(struct inode *inode);
3609
3610 static inline int ext4_has_inline_data(struct inode *inode)
3611 {
3612 return ext4_test_inode_flag(inode, EXT4_INODE_INLINE_DATA) &&
3613 EXT4_I(inode)->i_inline_off;
3614 }
3615
3616 /* namei.c */
3617 extern const struct inode_operations ext4_dir_inode_operations;
3618 extern const struct inode_operations ext4_special_inode_operations;
3619 extern struct dentry *ext4_get_parent(struct dentry *child);
3620 extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
3621 struct ext4_dir_entry_2 *de,
3622 int blocksize, int csum_size,
3623 unsigned int parent_ino, int dotdot_real_len);
3624 extern void ext4_initialize_dirent_tail(struct buffer_head *bh,
3625 unsigned int blocksize);
3626 extern int ext4_handle_dirty_dirblock(handle_t *handle, struct inode *inode,
3627 struct buffer_head *bh);
3628 extern int __ext4_unlink(struct inode *dir, const struct qstr *d_name,
3629 struct inode *inode, struct dentry *dentry);
3630 extern int __ext4_link(struct inode *dir, struct inode *inode,
3631 struct dentry *dentry);
3632
3633 #define S_SHIFT 12
3634 static const unsigned char ext4_type_by_mode[(S_IFMT >> S_SHIFT) + 1] = {
3635 [S_IFREG >> S_SHIFT] = EXT4_FT_REG_FILE,
3636 [S_IFDIR >> S_SHIFT] = EXT4_FT_DIR,
3637 [S_IFCHR >> S_SHIFT] = EXT4_FT_CHRDEV,
3638 [S_IFBLK >> S_SHIFT] = EXT4_FT_BLKDEV,
3639 [S_IFIFO >> S_SHIFT] = EXT4_FT_FIFO,
3640 [S_IFSOCK >> S_SHIFT] = EXT4_FT_SOCK,
3641 [S_IFLNK >> S_SHIFT] = EXT4_FT_SYMLINK,
3642 };
3643
3644 static inline void ext4_set_de_type(struct super_block *sb,
3645 struct ext4_dir_entry_2 *de,
3646 umode_t mode) {
3647 if (ext4_has_feature_filetype(sb))
3648 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
3649 }
3650
3651 /* readpages.c */
3652 extern int ext4_mpage_readpages(struct inode *inode,
3653 struct readahead_control *rac, struct folio *folio);
3654 extern int __init ext4_init_post_read_processing(void);
3655 extern void ext4_exit_post_read_processing(void);
3656
3657 /* symlink.c */
3658 extern const struct inode_operations ext4_encrypted_symlink_inode_operations;
3659 extern const struct inode_operations ext4_symlink_inode_operations;
3660 extern const struct inode_operations ext4_fast_symlink_inode_operations;
3661
3662 /* sysfs.c */
3663 extern void ext4_notify_error_sysfs(struct ext4_sb_info *sbi);
3664 extern int ext4_register_sysfs(struct super_block *sb);
3665 extern void ext4_unregister_sysfs(struct super_block *sb);
3666 extern int __init ext4_init_sysfs(void);
3667 extern void ext4_exit_sysfs(void);
3668
3669 /* block_validity */
3670 extern void ext4_release_system_zone(struct super_block *sb);
3671 extern int ext4_setup_system_zone(struct super_block *sb);
3672 extern int __init ext4_init_system_zone(void);
3673 extern void ext4_exit_system_zone(void);
3674 extern int ext4_inode_block_valid(struct inode *inode,
3675 ext4_fsblk_t start_blk,
3676 unsigned int count);
3677 extern int ext4_check_blockref(const char *, unsigned int,
3678 struct inode *, __le32 *, unsigned int);
3679 extern int ext4_sb_block_valid(struct super_block *sb, struct inode *inode,
3680 ext4_fsblk_t start_blk, unsigned int count);
3681
3682
3683 /* extents.c */
3684 struct ext4_ext_path;
3685 struct ext4_extent;
3686
3687 /*
3688 * Maximum number of logical blocks in a file; ext4_extent's ee_block is
3689 * __le32.
3690 */
3691 #define EXT_MAX_BLOCKS 0xffffffff
3692
3693 extern void ext4_ext_tree_init(handle_t *handle, struct inode *inode);
3694 extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
3695 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
3696 struct ext4_map_blocks *map, int flags);
3697 extern int ext4_ext_truncate(handle_t *, struct inode *);
3698 extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
3699 ext4_lblk_t end);
3700 extern void ext4_ext_init(struct super_block *);
3701 extern void ext4_ext_release(struct super_block *);
3702 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
3703 loff_t len);
3704 extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
3705 loff_t offset, ssize_t len);
3706 extern int ext4_convert_unwritten_io_end_vec(handle_t *handle,
3707 ext4_io_end_t *io_end);
3708 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
3709 struct ext4_map_blocks *map, int flags);
3710 extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
3711 int num,
3712 struct ext4_ext_path *path);
3713 extern int ext4_ext_insert_extent(handle_t *, struct inode *,
3714 struct ext4_ext_path **,
3715 struct ext4_extent *, int);
3716 extern struct ext4_ext_path *ext4_find_extent(struct inode *, ext4_lblk_t,
3717 struct ext4_ext_path **,
3718 int flags);
3719 extern void ext4_free_ext_path(struct ext4_ext_path *);
3720 extern int ext4_ext_check_inode(struct inode *inode);
3721 extern ext4_lblk_t ext4_ext_next_allocated_block(struct ext4_ext_path *path);
3722 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
3723 __u64 start, __u64 len);
3724 extern int ext4_get_es_cache(struct inode *inode,
3725 struct fiemap_extent_info *fieinfo,
3726 __u64 start, __u64 len);
3727 extern int ext4_ext_precache(struct inode *inode);
3728 extern int ext4_swap_extents(handle_t *handle, struct inode *inode1,
3729 struct inode *inode2, ext4_lblk_t lblk1,
3730 ext4_lblk_t lblk2, ext4_lblk_t count,
3731 int mark_unwritten,int *err);
3732 extern int ext4_clu_mapped(struct inode *inode, ext4_lblk_t lclu);
3733 extern int ext4_datasem_ensure_credits(handle_t *handle, struct inode *inode,
3734 int check_cred, int restart_cred,
3735 int revoke_cred);
3736 extern void ext4_ext_replay_shrink_inode(struct inode *inode, ext4_lblk_t end);
3737 extern int ext4_ext_replay_set_iblocks(struct inode *inode);
3738 extern int ext4_ext_replay_update_ex(struct inode *inode, ext4_lblk_t start,
3739 int len, int unwritten, ext4_fsblk_t pblk);
3740 extern int ext4_ext_clear_bb(struct inode *inode);
3741
3742
3743 /* move_extent.c */
3744 extern void ext4_double_down_write_data_sem(struct inode *first,
3745 struct inode *second);
3746 extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
3747 struct inode *donor_inode);
3748 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
3749 __u64 start_orig, __u64 start_donor,
3750 __u64 len, __u64 *moved_len);
3751
3752 /* page-io.c */
3753 extern int __init ext4_init_pageio(void);
3754 extern void ext4_exit_pageio(void);
3755 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
3756 extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
3757 extern int ext4_put_io_end(ext4_io_end_t *io_end);
3758 extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
3759 extern void ext4_io_submit_init(struct ext4_io_submit *io,
3760 struct writeback_control *wbc);
3761 extern void ext4_end_io_rsv_work(struct work_struct *work);
3762 extern void ext4_io_submit(struct ext4_io_submit *io);
3763 int ext4_bio_write_folio(struct ext4_io_submit *io, struct folio *page,
3764 size_t len);
3765 extern struct ext4_io_end_vec *ext4_alloc_io_end_vec(ext4_io_end_t *io_end);
3766 extern struct ext4_io_end_vec *ext4_last_io_end_vec(ext4_io_end_t *io_end);
3767
3768 /* mmp.c */
3769 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
3770
3771 /* mmp.c */
3772 extern void ext4_stop_mmpd(struct ext4_sb_info *sbi);
3773
3774 /* verity.c */
3775 extern const struct fsverity_operations ext4_verityops;
3776
3777 /* orphan.c */
3778 extern int ext4_orphan_add(handle_t *, struct inode *);
3779 extern int ext4_orphan_del(handle_t *, struct inode *);
3780 extern void ext4_orphan_cleanup(struct super_block *sb,
3781 struct ext4_super_block *es);
3782 extern void ext4_release_orphan_info(struct super_block *sb);
3783 extern int ext4_init_orphan_info(struct super_block *sb);
3784 extern int ext4_orphan_file_empty(struct super_block *sb);
3785 extern void ext4_orphan_file_block_trigger(
3786 struct jbd2_buffer_trigger_type *triggers,
3787 struct buffer_head *bh,
3788 void *data, size_t size);
3789
3790 /*
3791 * Add new method to test whether block and inode bitmaps are properly
3792 * initialized. With uninit_bg reading the block from disk is not enough
3793 * to mark the bitmap uptodate. We need to also zero-out the bitmap
3794 */
3795 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
3796
3797 static inline int bitmap_uptodate(struct buffer_head *bh)
3798 {
3799 return (buffer_uptodate(bh) &&
3800 test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
3801 }
3802 static inline void set_bitmap_uptodate(struct buffer_head *bh)
3803 {
3804 set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
3805 }
3806
3807 /* For ioend & aio unwritten conversion wait queues */
3808 #define EXT4_WQ_HASH_SZ 37
3809 #define ext4_ioend_wq(v) (&ext4__ioend_wq[((unsigned long)(v)) %\
3810 EXT4_WQ_HASH_SZ])
3811 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
3812
3813 extern int ext4_resize_begin(struct super_block *sb);
3814 extern int ext4_resize_end(struct super_block *sb, bool update_backups);
3815
3816 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
3817 struct ext4_io_end *io_end)
3818 {
3819 if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
3820 io_end->flag |= EXT4_IO_END_UNWRITTEN;
3821 atomic_inc(&EXT4_I(inode)->i_unwritten);
3822 }
3823 }
3824
3825 static inline void ext4_clear_io_unwritten_flag(ext4_io_end_t *io_end)
3826 {
3827 struct inode *inode = io_end->inode;
3828
3829 if (io_end->flag & EXT4_IO_END_UNWRITTEN) {
3830 io_end->flag &= ~EXT4_IO_END_UNWRITTEN;
3831 /* Wake up anyone waiting on unwritten extent conversion */
3832 if (atomic_dec_and_test(&EXT4_I(inode)->i_unwritten))
3833 wake_up_all(ext4_ioend_wq(inode));
3834 }
3835 }
3836
3837 extern const struct iomap_ops ext4_iomap_ops;
3838 extern const struct iomap_ops ext4_iomap_overwrite_ops;
3839 extern const struct iomap_ops ext4_iomap_report_ops;
3840
3841 static inline int ext4_buffer_uptodate(struct buffer_head *bh)
3842 {
3843 /*
3844 * If the buffer has the write error flag, we have failed
3845 * to write out data in the block. In this case, we don't
3846 * have to read the block because we may read the old data
3847 * successfully.
3848 */
3849 if (buffer_write_io_error(bh))
3850 set_buffer_uptodate(bh);
3851 return buffer_uptodate(bh);
3852 }
3853
3854 #endif /* __KERNEL__ */
3855
3856 #define EFSBADCRC EBADMSG /* Bad CRC detected */
3857 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
3858
3859 #endif /* _EXT4_H */
3860
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.