Loading...
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * ext4.h
4 *
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
9 *
10 * from
11 *
12 * linux/include/linux/minix_fs.h
13 *
14 * Copyright (C) 1991, 1992 Linus Torvalds
15 */
16
17#ifndef _EXT4_H
18#define _EXT4_H
19
20#include <linux/types.h>
21#include <linux/blkdev.h>
22#include <linux/magic.h>
23#include <linux/jbd2.h>
24#include <linux/quota.h>
25#include <linux/rwsem.h>
26#include <linux/rbtree.h>
27#include <linux/seqlock.h>
28#include <linux/mutex.h>
29#include <linux/timer.h>
30#include <linux/version.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#ifdef __KERNEL__
40#include <linux/compat.h>
41#endif
42
43#include <linux/fscrypt.h>
44#include <linux/fsverity.h>
45
46#include <linux/compiler.h>
47
48/*
49 * The fourth extended filesystem constants/structures
50 */
51
52/*
53 * with AGGRESSIVE_CHECK allocator runs consistency checks over
54 * structures. these checks slow things down a lot
55 */
56#define AGGRESSIVE_CHECK__
57
58/*
59 * with DOUBLE_CHECK defined mballoc creates persistent in-core
60 * bitmaps, maintains and uses them to check for double allocations
61 */
62#define DOUBLE_CHECK__
63
64/*
65 * Define EXT4FS_DEBUG to produce debug messages
66 */
67#undef EXT4FS_DEBUG
68
69/*
70 * Debug code
71 */
72#ifdef EXT4FS_DEBUG
73#define ext4_debug(f, a...) \
74 do { \
75 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:", \
76 __FILE__, __LINE__, __func__); \
77 printk(KERN_DEBUG f, ## a); \
78 } while (0)
79#else
80#define ext4_debug(fmt, ...) no_printk(fmt, ##__VA_ARGS__)
81#endif
82
83/*
84 * Turn on EXT_DEBUG to get lots of info about extents operations.
85 */
86#define EXT_DEBUG__
87#ifdef EXT_DEBUG
88#define ext_debug(fmt, ...) printk(fmt, ##__VA_ARGS__)
89#else
90#define ext_debug(fmt, ...) no_printk(fmt, ##__VA_ARGS__)
91#endif
92
93/* data type for block offset of block group */
94typedef int ext4_grpblk_t;
95
96/* data type for filesystem-wide blocks number */
97typedef unsigned long long ext4_fsblk_t;
98
99/* data type for file logical block number */
100typedef __u32 ext4_lblk_t;
101
102/* data type for block group number */
103typedef unsigned int ext4_group_t;
104
105enum SHIFT_DIRECTION {
106 SHIFT_LEFT = 0,
107 SHIFT_RIGHT,
108};
109
110/*
111 * Flags used in mballoc's allocation_context flags field.
112 *
113 * Also used to show what's going on for debugging purposes when the
114 * flag field is exported via the traceport interface
115 */
116
117/* prefer goal again. length */
118#define EXT4_MB_HINT_MERGE 0x0001
119/* blocks already reserved */
120#define EXT4_MB_HINT_RESERVED 0x0002
121/* metadata is being allocated */
122#define EXT4_MB_HINT_METADATA 0x0004
123/* first blocks in the file */
124#define EXT4_MB_HINT_FIRST 0x0008
125/* search for the best chunk */
126#define EXT4_MB_HINT_BEST 0x0010
127/* data is being allocated */
128#define EXT4_MB_HINT_DATA 0x0020
129/* don't preallocate (for tails) */
130#define EXT4_MB_HINT_NOPREALLOC 0x0040
131/* allocate for locality group */
132#define EXT4_MB_HINT_GROUP_ALLOC 0x0080
133/* allocate goal blocks or none */
134#define EXT4_MB_HINT_GOAL_ONLY 0x0100
135/* goal is meaningful */
136#define EXT4_MB_HINT_TRY_GOAL 0x0200
137/* blocks already pre-reserved by delayed allocation */
138#define EXT4_MB_DELALLOC_RESERVED 0x0400
139/* We are doing stream allocation */
140#define EXT4_MB_STREAM_ALLOC 0x0800
141/* Use reserved root blocks if needed */
142#define EXT4_MB_USE_ROOT_BLOCKS 0x1000
143/* Use blocks from reserved pool */
144#define EXT4_MB_USE_RESERVED 0x2000
145
146struct ext4_allocation_request {
147 /* target inode for block we're allocating */
148 struct inode *inode;
149 /* how many blocks we want to allocate */
150 unsigned int len;
151 /* logical block in target inode */
152 ext4_lblk_t logical;
153 /* the closest logical allocated block to the left */
154 ext4_lblk_t lleft;
155 /* the closest logical allocated block to the right */
156 ext4_lblk_t lright;
157 /* phys. target (a hint) */
158 ext4_fsblk_t goal;
159 /* phys. block for the closest logical allocated block to the left */
160 ext4_fsblk_t pleft;
161 /* phys. block for the closest logical allocated block to the right */
162 ext4_fsblk_t pright;
163 /* flags. see above EXT4_MB_HINT_* */
164 unsigned int flags;
165};
166
167/*
168 * Logical to physical block mapping, used by ext4_map_blocks()
169 *
170 * This structure is used to pass requests into ext4_map_blocks() as
171 * well as to store the information returned by ext4_map_blocks(). It
172 * takes less room on the stack than a struct buffer_head.
173 */
174#define EXT4_MAP_NEW (1 << BH_New)
175#define EXT4_MAP_MAPPED (1 << BH_Mapped)
176#define EXT4_MAP_UNWRITTEN (1 << BH_Unwritten)
177#define EXT4_MAP_BOUNDARY (1 << BH_Boundary)
178#define EXT4_MAP_FLAGS (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
179 EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY)
180
181struct ext4_map_blocks {
182 ext4_fsblk_t m_pblk;
183 ext4_lblk_t m_lblk;
184 unsigned int m_len;
185 unsigned int m_flags;
186};
187
188/*
189 * Block validity checking, system zone rbtree.
190 */
191struct ext4_system_blocks {
192 struct rb_root root;
193 struct rcu_head rcu;
194};
195
196/*
197 * Flags for ext4_io_end->flags
198 */
199#define EXT4_IO_END_UNWRITTEN 0x0001
200
201/*
202 * For converting unwritten extents on a work queue. 'handle' is used for
203 * buffered writeback.
204 */
205typedef struct ext4_io_end {
206 struct list_head list; /* per-file finished IO list */
207 handle_t *handle; /* handle reserved for extent
208 * conversion */
209 struct inode *inode; /* file being written to */
210 struct bio *bio; /* Linked list of completed
211 * bios covering the extent */
212 unsigned int flag; /* unwritten or not */
213 atomic_t count; /* reference counter */
214 loff_t offset; /* offset in the file */
215 ssize_t size; /* size of the extent */
216} ext4_io_end_t;
217
218struct ext4_io_submit {
219 struct writeback_control *io_wbc;
220 struct bio *io_bio;
221 ext4_io_end_t *io_end;
222 sector_t io_next_block;
223};
224
225/*
226 * Special inodes numbers
227 */
228#define EXT4_BAD_INO 1 /* Bad blocks inode */
229#define EXT4_ROOT_INO 2 /* Root inode */
230#define EXT4_USR_QUOTA_INO 3 /* User quota inode */
231#define EXT4_GRP_QUOTA_INO 4 /* Group quota inode */
232#define EXT4_BOOT_LOADER_INO 5 /* Boot loader inode */
233#define EXT4_UNDEL_DIR_INO 6 /* Undelete directory inode */
234#define EXT4_RESIZE_INO 7 /* Reserved group descriptors inode */
235#define EXT4_JOURNAL_INO 8 /* Journal inode */
236
237/* First non-reserved inode for old ext4 filesystems */
238#define EXT4_GOOD_OLD_FIRST_INO 11
239
240/*
241 * Maximal count of links to a file
242 */
243#define EXT4_LINK_MAX 65000
244
245/*
246 * Macro-instructions used to manage several block sizes
247 */
248#define EXT4_MIN_BLOCK_SIZE 1024
249#define EXT4_MAX_BLOCK_SIZE 65536
250#define EXT4_MIN_BLOCK_LOG_SIZE 10
251#define EXT4_MAX_BLOCK_LOG_SIZE 16
252#define EXT4_MAX_CLUSTER_LOG_SIZE 30
253#ifdef __KERNEL__
254# define EXT4_BLOCK_SIZE(s) ((s)->s_blocksize)
255#else
256# define EXT4_BLOCK_SIZE(s) (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
257#endif
258#define EXT4_ADDR_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
259#define EXT4_CLUSTER_SIZE(s) (EXT4_BLOCK_SIZE(s) << \
260 EXT4_SB(s)->s_cluster_bits)
261#ifdef __KERNEL__
262# define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits)
263# define EXT4_CLUSTER_BITS(s) (EXT4_SB(s)->s_cluster_bits)
264#else
265# define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_log_block_size + 10)
266#endif
267#ifdef __KERNEL__
268#define EXT4_ADDR_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_addr_per_block_bits)
269#define EXT4_INODE_SIZE(s) (EXT4_SB(s)->s_inode_size)
270#define EXT4_FIRST_INO(s) (EXT4_SB(s)->s_first_ino)
271#else
272#define EXT4_INODE_SIZE(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
273 EXT4_GOOD_OLD_INODE_SIZE : \
274 (s)->s_inode_size)
275#define EXT4_FIRST_INO(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
276 EXT4_GOOD_OLD_FIRST_INO : \
277 (s)->s_first_ino)
278#endif
279#define EXT4_BLOCK_ALIGN(size, blkbits) ALIGN((size), (1 << (blkbits)))
280#define EXT4_MAX_BLOCKS(size, offset, blkbits) \
281 ((EXT4_BLOCK_ALIGN(size + offset, blkbits) >> blkbits) - (offset >> \
282 blkbits))
283
284/* Translate a block number to a cluster number */
285#define EXT4_B2C(sbi, blk) ((blk) >> (sbi)->s_cluster_bits)
286/* Translate a cluster number to a block number */
287#define EXT4_C2B(sbi, cluster) ((cluster) << (sbi)->s_cluster_bits)
288/* Translate # of blks to # of clusters */
289#define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
290 (sbi)->s_cluster_bits)
291/* Mask out the low bits to get the starting block of the cluster */
292#define EXT4_PBLK_CMASK(s, pblk) ((pblk) & \
293 ~((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
294#define EXT4_LBLK_CMASK(s, lblk) ((lblk) & \
295 ~((ext4_lblk_t) (s)->s_cluster_ratio - 1))
296/* Fill in the low bits to get the last block of the cluster */
297#define EXT4_LBLK_CFILL(sbi, lblk) ((lblk) | \
298 ((ext4_lblk_t) (sbi)->s_cluster_ratio - 1))
299/* Get the cluster offset */
300#define EXT4_PBLK_COFF(s, pblk) ((pblk) & \
301 ((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
302#define EXT4_LBLK_COFF(s, lblk) ((lblk) & \
303 ((ext4_lblk_t) (s)->s_cluster_ratio - 1))
304
305/*
306 * Structure of a blocks group descriptor
307 */
308struct ext4_group_desc
309{
310 __le32 bg_block_bitmap_lo; /* Blocks bitmap block */
311 __le32 bg_inode_bitmap_lo; /* Inodes bitmap block */
312 __le32 bg_inode_table_lo; /* Inodes table block */
313 __le16 bg_free_blocks_count_lo;/* Free blocks count */
314 __le16 bg_free_inodes_count_lo;/* Free inodes count */
315 __le16 bg_used_dirs_count_lo; /* Directories count */
316 __le16 bg_flags; /* EXT4_BG_flags (INODE_UNINIT, etc) */
317 __le32 bg_exclude_bitmap_lo; /* Exclude bitmap for snapshots */
318 __le16 bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
319 __le16 bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
320 __le16 bg_itable_unused_lo; /* Unused inodes count */
321 __le16 bg_checksum; /* crc16(sb_uuid+group+desc) */
322 __le32 bg_block_bitmap_hi; /* Blocks bitmap block MSB */
323 __le32 bg_inode_bitmap_hi; /* Inodes bitmap block MSB */
324 __le32 bg_inode_table_hi; /* Inodes table block MSB */
325 __le16 bg_free_blocks_count_hi;/* Free blocks count MSB */
326 __le16 bg_free_inodes_count_hi;/* Free inodes count MSB */
327 __le16 bg_used_dirs_count_hi; /* Directories count MSB */
328 __le16 bg_itable_unused_hi; /* Unused inodes count MSB */
329 __le32 bg_exclude_bitmap_hi; /* Exclude bitmap block MSB */
330 __le16 bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
331 __le16 bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
332 __u32 bg_reserved;
333};
334
335#define EXT4_BG_INODE_BITMAP_CSUM_HI_END \
336 (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
337 sizeof(__le16))
338#define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END \
339 (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
340 sizeof(__le16))
341
342/*
343 * Structure of a flex block group info
344 */
345
346struct flex_groups {
347 atomic64_t free_clusters;
348 atomic_t free_inodes;
349 atomic_t used_dirs;
350};
351
352#define EXT4_BG_INODE_UNINIT 0x0001 /* Inode table/bitmap not in use */
353#define EXT4_BG_BLOCK_UNINIT 0x0002 /* Block bitmap not in use */
354#define EXT4_BG_INODE_ZEROED 0x0004 /* On-disk itable initialized to zero */
355
356/*
357 * Macro-instructions used to manage group descriptors
358 */
359#define EXT4_MIN_DESC_SIZE 32
360#define EXT4_MIN_DESC_SIZE_64BIT 64
361#define EXT4_MAX_DESC_SIZE EXT4_MIN_BLOCK_SIZE
362#define EXT4_DESC_SIZE(s) (EXT4_SB(s)->s_desc_size)
363#ifdef __KERNEL__
364# define EXT4_BLOCKS_PER_GROUP(s) (EXT4_SB(s)->s_blocks_per_group)
365# define EXT4_CLUSTERS_PER_GROUP(s) (EXT4_SB(s)->s_clusters_per_group)
366# define EXT4_DESC_PER_BLOCK(s) (EXT4_SB(s)->s_desc_per_block)
367# define EXT4_INODES_PER_GROUP(s) (EXT4_SB(s)->s_inodes_per_group)
368# define EXT4_DESC_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_desc_per_block_bits)
369#else
370# define EXT4_BLOCKS_PER_GROUP(s) ((s)->s_blocks_per_group)
371# define EXT4_DESC_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
372# define EXT4_INODES_PER_GROUP(s) ((s)->s_inodes_per_group)
373#endif
374
375/*
376 * Constants relative to the data blocks
377 */
378#define EXT4_NDIR_BLOCKS 12
379#define EXT4_IND_BLOCK EXT4_NDIR_BLOCKS
380#define EXT4_DIND_BLOCK (EXT4_IND_BLOCK + 1)
381#define EXT4_TIND_BLOCK (EXT4_DIND_BLOCK + 1)
382#define EXT4_N_BLOCKS (EXT4_TIND_BLOCK + 1)
383
384/*
385 * Inode flags
386 */
387#define EXT4_SECRM_FL 0x00000001 /* Secure deletion */
388#define EXT4_UNRM_FL 0x00000002 /* Undelete */
389#define EXT4_COMPR_FL 0x00000004 /* Compress file */
390#define EXT4_SYNC_FL 0x00000008 /* Synchronous updates */
391#define EXT4_IMMUTABLE_FL 0x00000010 /* Immutable file */
392#define EXT4_APPEND_FL 0x00000020 /* writes to file may only append */
393#define EXT4_NODUMP_FL 0x00000040 /* do not dump file */
394#define EXT4_NOATIME_FL 0x00000080 /* do not update atime */
395/* Reserved for compression usage... */
396#define EXT4_DIRTY_FL 0x00000100
397#define EXT4_COMPRBLK_FL 0x00000200 /* One or more compressed clusters */
398#define EXT4_NOCOMPR_FL 0x00000400 /* Don't compress */
399 /* nb: was previously EXT2_ECOMPR_FL */
400#define EXT4_ENCRYPT_FL 0x00000800 /* encrypted file */
401/* End compression flags --- maybe not all used */
402#define EXT4_INDEX_FL 0x00001000 /* hash-indexed directory */
403#define EXT4_IMAGIC_FL 0x00002000 /* AFS directory */
404#define EXT4_JOURNAL_DATA_FL 0x00004000 /* file data should be journaled */
405#define EXT4_NOTAIL_FL 0x00008000 /* file tail should not be merged */
406#define EXT4_DIRSYNC_FL 0x00010000 /* dirsync behaviour (directories only) */
407#define EXT4_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/
408#define EXT4_HUGE_FILE_FL 0x00040000 /* Set to each huge file */
409#define EXT4_EXTENTS_FL 0x00080000 /* Inode uses extents */
410#define EXT4_VERITY_FL 0x00100000 /* Verity protected inode */
411#define EXT4_EA_INODE_FL 0x00200000 /* Inode used for large EA */
412#define EXT4_EOFBLOCKS_FL 0x00400000 /* Blocks allocated beyond EOF */
413#define EXT4_INLINE_DATA_FL 0x10000000 /* Inode has inline data. */
414#define EXT4_PROJINHERIT_FL 0x20000000 /* Create with parents projid */
415#define EXT4_CASEFOLD_FL 0x40000000 /* Casefolded file */
416#define EXT4_RESERVED_FL 0x80000000 /* reserved for ext4 lib */
417
418#define EXT4_FL_USER_VISIBLE 0x705BDFFF /* User visible flags */
419#define EXT4_FL_USER_MODIFIABLE 0x604BC0FF /* User modifiable flags */
420
421/* Flags we can manipulate with through EXT4_IOC_FSSETXATTR */
422#define EXT4_FL_XFLAG_VISIBLE (EXT4_SYNC_FL | \
423 EXT4_IMMUTABLE_FL | \
424 EXT4_APPEND_FL | \
425 EXT4_NODUMP_FL | \
426 EXT4_NOATIME_FL | \
427 EXT4_PROJINHERIT_FL)
428
429/* Flags that should be inherited by new inodes from their parent. */
430#define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
431 EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
432 EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
433 EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL |\
434 EXT4_PROJINHERIT_FL | EXT4_CASEFOLD_FL)
435
436/* Flags that are appropriate for regular files (all but dir-specific ones). */
437#define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL | EXT4_CASEFOLD_FL |\
438 EXT4_PROJINHERIT_FL))
439
440/* Flags that are appropriate for non-directories/regular files. */
441#define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
442
443/* The only flags that should be swapped */
444#define EXT4_FL_SHOULD_SWAP (EXT4_HUGE_FILE_FL | EXT4_EXTENTS_FL)
445
446/* Mask out flags that are inappropriate for the given type of inode. */
447static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
448{
449 if (S_ISDIR(mode))
450 return flags;
451 else if (S_ISREG(mode))
452 return flags & EXT4_REG_FLMASK;
453 else
454 return flags & EXT4_OTHER_FLMASK;
455}
456
457/*
458 * Inode flags used for atomic set/get
459 */
460enum {
461 EXT4_INODE_SECRM = 0, /* Secure deletion */
462 EXT4_INODE_UNRM = 1, /* Undelete */
463 EXT4_INODE_COMPR = 2, /* Compress file */
464 EXT4_INODE_SYNC = 3, /* Synchronous updates */
465 EXT4_INODE_IMMUTABLE = 4, /* Immutable file */
466 EXT4_INODE_APPEND = 5, /* writes to file may only append */
467 EXT4_INODE_NODUMP = 6, /* do not dump file */
468 EXT4_INODE_NOATIME = 7, /* do not update atime */
469/* Reserved for compression usage... */
470 EXT4_INODE_DIRTY = 8,
471 EXT4_INODE_COMPRBLK = 9, /* One or more compressed clusters */
472 EXT4_INODE_NOCOMPR = 10, /* Don't compress */
473 EXT4_INODE_ENCRYPT = 11, /* Encrypted file */
474/* End compression flags --- maybe not all used */
475 EXT4_INODE_INDEX = 12, /* hash-indexed directory */
476 EXT4_INODE_IMAGIC = 13, /* AFS directory */
477 EXT4_INODE_JOURNAL_DATA = 14, /* file data should be journaled */
478 EXT4_INODE_NOTAIL = 15, /* file tail should not be merged */
479 EXT4_INODE_DIRSYNC = 16, /* dirsync behaviour (directories only) */
480 EXT4_INODE_TOPDIR = 17, /* Top of directory hierarchies*/
481 EXT4_INODE_HUGE_FILE = 18, /* Set to each huge file */
482 EXT4_INODE_EXTENTS = 19, /* Inode uses extents */
483 EXT4_INODE_VERITY = 20, /* Verity protected inode */
484 EXT4_INODE_EA_INODE = 21, /* Inode used for large EA */
485 EXT4_INODE_EOFBLOCKS = 22, /* Blocks allocated beyond EOF */
486 EXT4_INODE_INLINE_DATA = 28, /* Data in inode. */
487 EXT4_INODE_PROJINHERIT = 29, /* Create with parents projid */
488 EXT4_INODE_RESERVED = 31, /* reserved for ext4 lib */
489};
490
491/*
492 * Since it's pretty easy to mix up bit numbers and hex values, we use a
493 * build-time check to make sure that EXT4_XXX_FL is consistent with respect to
494 * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost
495 * any extra space in the compiled kernel image, otherwise, the build will fail.
496 * It's important that these values are the same, since we are using
497 * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent
498 * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk
499 * values found in ext2, ext3 and ext4 filesystems, and of course the values
500 * defined in e2fsprogs.
501 *
502 * It's not paranoia if the Murphy's Law really *is* out to get you. :-)
503 */
504#define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
505#define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG))
506
507static inline void ext4_check_flag_values(void)
508{
509 CHECK_FLAG_VALUE(SECRM);
510 CHECK_FLAG_VALUE(UNRM);
511 CHECK_FLAG_VALUE(COMPR);
512 CHECK_FLAG_VALUE(SYNC);
513 CHECK_FLAG_VALUE(IMMUTABLE);
514 CHECK_FLAG_VALUE(APPEND);
515 CHECK_FLAG_VALUE(NODUMP);
516 CHECK_FLAG_VALUE(NOATIME);
517 CHECK_FLAG_VALUE(DIRTY);
518 CHECK_FLAG_VALUE(COMPRBLK);
519 CHECK_FLAG_VALUE(NOCOMPR);
520 CHECK_FLAG_VALUE(ENCRYPT);
521 CHECK_FLAG_VALUE(INDEX);
522 CHECK_FLAG_VALUE(IMAGIC);
523 CHECK_FLAG_VALUE(JOURNAL_DATA);
524 CHECK_FLAG_VALUE(NOTAIL);
525 CHECK_FLAG_VALUE(DIRSYNC);
526 CHECK_FLAG_VALUE(TOPDIR);
527 CHECK_FLAG_VALUE(HUGE_FILE);
528 CHECK_FLAG_VALUE(EXTENTS);
529 CHECK_FLAG_VALUE(VERITY);
530 CHECK_FLAG_VALUE(EA_INODE);
531 CHECK_FLAG_VALUE(EOFBLOCKS);
532 CHECK_FLAG_VALUE(INLINE_DATA);
533 CHECK_FLAG_VALUE(PROJINHERIT);
534 CHECK_FLAG_VALUE(RESERVED);
535}
536
537/* Used to pass group descriptor data when online resize is done */
538struct ext4_new_group_input {
539 __u32 group; /* Group number for this data */
540 __u64 block_bitmap; /* Absolute block number of block bitmap */
541 __u64 inode_bitmap; /* Absolute block number of inode bitmap */
542 __u64 inode_table; /* Absolute block number of inode table start */
543 __u32 blocks_count; /* Total number of blocks in this group */
544 __u16 reserved_blocks; /* Number of reserved blocks in this group */
545 __u16 unused;
546};
547
548#if defined(__KERNEL__) && defined(CONFIG_COMPAT)
549struct compat_ext4_new_group_input {
550 u32 group;
551 compat_u64 block_bitmap;
552 compat_u64 inode_bitmap;
553 compat_u64 inode_table;
554 u32 blocks_count;
555 u16 reserved_blocks;
556 u16 unused;
557};
558#endif
559
560/* The struct ext4_new_group_input in kernel space, with free_blocks_count */
561struct ext4_new_group_data {
562 __u32 group;
563 __u64 block_bitmap;
564 __u64 inode_bitmap;
565 __u64 inode_table;
566 __u32 blocks_count;
567 __u16 reserved_blocks;
568 __u16 mdata_blocks;
569 __u32 free_clusters_count;
570};
571
572/* Indexes used to index group tables in ext4_new_group_data */
573enum {
574 BLOCK_BITMAP = 0, /* block bitmap */
575 INODE_BITMAP, /* inode bitmap */
576 INODE_TABLE, /* inode tables */
577 GROUP_TABLE_COUNT,
578};
579
580/*
581 * Flags used by ext4_map_blocks()
582 */
583 /* Allocate any needed blocks and/or convert an unwritten
584 extent to be an initialized ext4 */
585#define EXT4_GET_BLOCKS_CREATE 0x0001
586 /* Request the creation of an unwritten extent */
587#define EXT4_GET_BLOCKS_UNWRIT_EXT 0x0002
588#define EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT (EXT4_GET_BLOCKS_UNWRIT_EXT|\
589 EXT4_GET_BLOCKS_CREATE)
590 /* Caller is from the delayed allocation writeout path
591 * finally doing the actual allocation of delayed blocks */
592#define EXT4_GET_BLOCKS_DELALLOC_RESERVE 0x0004
593 /* caller is from the direct IO path, request to creation of an
594 unwritten extents if not allocated, split the unwritten
595 extent if blocks has been preallocated already*/
596#define EXT4_GET_BLOCKS_PRE_IO 0x0008
597#define EXT4_GET_BLOCKS_CONVERT 0x0010
598#define EXT4_GET_BLOCKS_IO_CREATE_EXT (EXT4_GET_BLOCKS_PRE_IO|\
599 EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
600 /* Convert extent to initialized after IO complete */
601#define EXT4_GET_BLOCKS_IO_CONVERT_EXT (EXT4_GET_BLOCKS_CONVERT|\
602 EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
603 /* Eventual metadata allocation (due to growing extent tree)
604 * should not fail, so try to use reserved blocks for that.*/
605#define EXT4_GET_BLOCKS_METADATA_NOFAIL 0x0020
606 /* Don't normalize allocation size (used for fallocate) */
607#define EXT4_GET_BLOCKS_NO_NORMALIZE 0x0040
608 /* Request will not result in inode size update (user for fallocate) */
609#define EXT4_GET_BLOCKS_KEEP_SIZE 0x0080
610 /* Convert written extents to unwritten */
611#define EXT4_GET_BLOCKS_CONVERT_UNWRITTEN 0x0100
612 /* Write zeros to newly created written extents */
613#define EXT4_GET_BLOCKS_ZERO 0x0200
614#define EXT4_GET_BLOCKS_CREATE_ZERO (EXT4_GET_BLOCKS_CREATE |\
615 EXT4_GET_BLOCKS_ZERO)
616 /* Caller will submit data before dropping transaction handle. This
617 * allows jbd2 to avoid submitting data before commit. */
618#define EXT4_GET_BLOCKS_IO_SUBMIT 0x0400
619
620/*
621 * The bit position of these flags must not overlap with any of the
622 * EXT4_GET_BLOCKS_*. They are used by ext4_find_extent(),
623 * read_extent_tree_block(), ext4_split_extent_at(),
624 * ext4_ext_insert_extent(), and ext4_ext_create_new_leaf().
625 * EXT4_EX_NOCACHE is used to indicate that the we shouldn't be
626 * caching the extents when reading from the extent tree while a
627 * truncate or punch hole operation is in progress.
628 */
629#define EXT4_EX_NOCACHE 0x40000000
630#define EXT4_EX_FORCE_CACHE 0x20000000
631
632/*
633 * Flags used by ext4_free_blocks
634 */
635#define EXT4_FREE_BLOCKS_METADATA 0x0001
636#define EXT4_FREE_BLOCKS_FORGET 0x0002
637#define EXT4_FREE_BLOCKS_VALIDATED 0x0004
638#define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
639#define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER 0x0010
640#define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER 0x0020
641#define EXT4_FREE_BLOCKS_RERESERVE_CLUSTER 0x0040
642
643/*
644 * ioctl commands
645 */
646#define EXT4_IOC_GETFLAGS FS_IOC_GETFLAGS
647#define EXT4_IOC_SETFLAGS FS_IOC_SETFLAGS
648#define EXT4_IOC_GETVERSION _IOR('f', 3, long)
649#define EXT4_IOC_SETVERSION _IOW('f', 4, long)
650#define EXT4_IOC_GETVERSION_OLD FS_IOC_GETVERSION
651#define EXT4_IOC_SETVERSION_OLD FS_IOC_SETVERSION
652#define EXT4_IOC_GETRSVSZ _IOR('f', 5, long)
653#define EXT4_IOC_SETRSVSZ _IOW('f', 6, long)
654#define EXT4_IOC_GROUP_EXTEND _IOW('f', 7, unsigned long)
655#define EXT4_IOC_GROUP_ADD _IOW('f', 8, struct ext4_new_group_input)
656#define EXT4_IOC_MIGRATE _IO('f', 9)
657 /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
658 /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
659#define EXT4_IOC_ALLOC_DA_BLKS _IO('f', 12)
660#define EXT4_IOC_MOVE_EXT _IOWR('f', 15, struct move_extent)
661#define EXT4_IOC_RESIZE_FS _IOW('f', 16, __u64)
662#define EXT4_IOC_SWAP_BOOT _IO('f', 17)
663#define EXT4_IOC_PRECACHE_EXTENTS _IO('f', 18)
664#define EXT4_IOC_SET_ENCRYPTION_POLICY FS_IOC_SET_ENCRYPTION_POLICY
665#define EXT4_IOC_GET_ENCRYPTION_PWSALT FS_IOC_GET_ENCRYPTION_PWSALT
666#define EXT4_IOC_GET_ENCRYPTION_POLICY FS_IOC_GET_ENCRYPTION_POLICY
667/* ioctl codes 19--39 are reserved for fscrypt */
668#define EXT4_IOC_CLEAR_ES_CACHE _IO('f', 40)
669#define EXT4_IOC_GETSTATE _IOW('f', 41, __u32)
670#define EXT4_IOC_GET_ES_CACHE _IOWR('f', 42, struct fiemap)
671
672#define EXT4_IOC_FSGETXATTR FS_IOC_FSGETXATTR
673#define EXT4_IOC_FSSETXATTR FS_IOC_FSSETXATTR
674
675#define EXT4_IOC_SHUTDOWN _IOR ('X', 125, __u32)
676
677/*
678 * Flags for going down operation
679 */
680#define EXT4_GOING_FLAGS_DEFAULT 0x0 /* going down */
681#define EXT4_GOING_FLAGS_LOGFLUSH 0x1 /* flush log but not data */
682#define EXT4_GOING_FLAGS_NOLOGFLUSH 0x2 /* don't flush log nor data */
683
684/*
685 * Flags returned by EXT4_IOC_GETSTATE
686 *
687 * We only expose to userspace a subset of the state flags in
688 * i_state_flags
689 */
690#define EXT4_STATE_FLAG_EXT_PRECACHED 0x00000001
691#define EXT4_STATE_FLAG_NEW 0x00000002
692#define EXT4_STATE_FLAG_NEWENTRY 0x00000004
693#define EXT4_STATE_FLAG_DA_ALLOC_CLOSE 0x00000008
694
695#if defined(__KERNEL__) && defined(CONFIG_COMPAT)
696/*
697 * ioctl commands in 32 bit emulation
698 */
699#define EXT4_IOC32_GETFLAGS FS_IOC32_GETFLAGS
700#define EXT4_IOC32_SETFLAGS FS_IOC32_SETFLAGS
701#define EXT4_IOC32_GETVERSION _IOR('f', 3, int)
702#define EXT4_IOC32_SETVERSION _IOW('f', 4, int)
703#define EXT4_IOC32_GETRSVSZ _IOR('f', 5, int)
704#define EXT4_IOC32_SETRSVSZ _IOW('f', 6, int)
705#define EXT4_IOC32_GROUP_EXTEND _IOW('f', 7, unsigned int)
706#define EXT4_IOC32_GROUP_ADD _IOW('f', 8, struct compat_ext4_new_group_input)
707#define EXT4_IOC32_GETVERSION_OLD FS_IOC32_GETVERSION
708#define EXT4_IOC32_SETVERSION_OLD FS_IOC32_SETVERSION
709#endif
710
711/*
712 * Returned by EXT4_IOC_GET_ES_CACHE as an additional possible flag.
713 * It indicates that the entry in extent status cache is for a hole.
714 */
715#define EXT4_FIEMAP_EXTENT_HOLE 0x08000000
716
717/* Max physical block we can address w/o extents */
718#define EXT4_MAX_BLOCK_FILE_PHYS 0xFFFFFFFF
719
720/* Max logical block we can support */
721#define EXT4_MAX_LOGICAL_BLOCK 0xFFFFFFFF
722
723/*
724 * Structure of an inode on the disk
725 */
726struct ext4_inode {
727 __le16 i_mode; /* File mode */
728 __le16 i_uid; /* Low 16 bits of Owner Uid */
729 __le32 i_size_lo; /* Size in bytes */
730 __le32 i_atime; /* Access time */
731 __le32 i_ctime; /* Inode Change time */
732 __le32 i_mtime; /* Modification time */
733 __le32 i_dtime; /* Deletion Time */
734 __le16 i_gid; /* Low 16 bits of Group Id */
735 __le16 i_links_count; /* Links count */
736 __le32 i_blocks_lo; /* Blocks count */
737 __le32 i_flags; /* File flags */
738 union {
739 struct {
740 __le32 l_i_version;
741 } linux1;
742 struct {
743 __u32 h_i_translator;
744 } hurd1;
745 struct {
746 __u32 m_i_reserved1;
747 } masix1;
748 } osd1; /* OS dependent 1 */
749 __le32 i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
750 __le32 i_generation; /* File version (for NFS) */
751 __le32 i_file_acl_lo; /* File ACL */
752 __le32 i_size_high;
753 __le32 i_obso_faddr; /* Obsoleted fragment address */
754 union {
755 struct {
756 __le16 l_i_blocks_high; /* were l_i_reserved1 */
757 __le16 l_i_file_acl_high;
758 __le16 l_i_uid_high; /* these 2 fields */
759 __le16 l_i_gid_high; /* were reserved2[0] */
760 __le16 l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
761 __le16 l_i_reserved;
762 } linux2;
763 struct {
764 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */
765 __u16 h_i_mode_high;
766 __u16 h_i_uid_high;
767 __u16 h_i_gid_high;
768 __u32 h_i_author;
769 } hurd2;
770 struct {
771 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */
772 __le16 m_i_file_acl_high;
773 __u32 m_i_reserved2[2];
774 } masix2;
775 } osd2; /* OS dependent 2 */
776 __le16 i_extra_isize;
777 __le16 i_checksum_hi; /* crc32c(uuid+inum+inode) BE */
778 __le32 i_ctime_extra; /* extra Change time (nsec << 2 | epoch) */
779 __le32 i_mtime_extra; /* extra Modification time(nsec << 2 | epoch) */
780 __le32 i_atime_extra; /* extra Access time (nsec << 2 | epoch) */
781 __le32 i_crtime; /* File Creation time */
782 __le32 i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
783 __le32 i_version_hi; /* high 32 bits for 64-bit version */
784 __le32 i_projid; /* Project ID */
785};
786
787struct move_extent {
788 __u32 reserved; /* should be zero */
789 __u32 donor_fd; /* donor file descriptor */
790 __u64 orig_start; /* logical start offset in block for orig */
791 __u64 donor_start; /* logical start offset in block for donor */
792 __u64 len; /* block length to be moved */
793 __u64 moved_len; /* moved block length */
794};
795
796#define EXT4_EPOCH_BITS 2
797#define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
798#define EXT4_NSEC_MASK (~0UL << EXT4_EPOCH_BITS)
799
800/*
801 * Extended fields will fit into an inode if the filesystem was formatted
802 * with large inodes (-I 256 or larger) and there are not currently any EAs
803 * consuming all of the available space. For new inodes we always reserve
804 * enough space for the kernel's known extended fields, but for inodes
805 * created with an old kernel this might not have been the case. None of
806 * the extended inode fields is critical for correct filesystem operation.
807 * This macro checks if a certain field fits in the inode. Note that
808 * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
809 */
810#define EXT4_FITS_IN_INODE(ext4_inode, einode, field) \
811 ((offsetof(typeof(*ext4_inode), field) + \
812 sizeof((ext4_inode)->field)) \
813 <= (EXT4_GOOD_OLD_INODE_SIZE + \
814 (einode)->i_extra_isize)) \
815
816/*
817 * We use an encoding that preserves the times for extra epoch "00":
818 *
819 * extra msb of adjust for signed
820 * epoch 32-bit 32-bit tv_sec to
821 * bits time decoded 64-bit tv_sec 64-bit tv_sec valid time range
822 * 0 0 1 -0x80000000..-0x00000001 0x000000000 1901-12-13..1969-12-31
823 * 0 0 0 0x000000000..0x07fffffff 0x000000000 1970-01-01..2038-01-19
824 * 0 1 1 0x080000000..0x0ffffffff 0x100000000 2038-01-19..2106-02-07
825 * 0 1 0 0x100000000..0x17fffffff 0x100000000 2106-02-07..2174-02-25
826 * 1 0 1 0x180000000..0x1ffffffff 0x200000000 2174-02-25..2242-03-16
827 * 1 0 0 0x200000000..0x27fffffff 0x200000000 2242-03-16..2310-04-04
828 * 1 1 1 0x280000000..0x2ffffffff 0x300000000 2310-04-04..2378-04-22
829 * 1 1 0 0x300000000..0x37fffffff 0x300000000 2378-04-22..2446-05-10
830 *
831 * Note that previous versions of the kernel on 64-bit systems would
832 * incorrectly use extra epoch bits 1,1 for dates between 1901 and
833 * 1970. e2fsck will correct this, assuming that it is run on the
834 * affected filesystem before 2242.
835 */
836
837static inline __le32 ext4_encode_extra_time(struct timespec64 *time)
838{
839 u32 extra =((time->tv_sec - (s32)time->tv_sec) >> 32) & EXT4_EPOCH_MASK;
840 return cpu_to_le32(extra | (time->tv_nsec << EXT4_EPOCH_BITS));
841}
842
843static inline void ext4_decode_extra_time(struct timespec64 *time,
844 __le32 extra)
845{
846 if (unlikely(extra & cpu_to_le32(EXT4_EPOCH_MASK)))
847 time->tv_sec += (u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK) << 32;
848 time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
849}
850
851#define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode) \
852do { \
853 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) {\
854 (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec); \
855 (raw_inode)->xtime ## _extra = \
856 ext4_encode_extra_time(&(inode)->xtime); \
857 } \
858 else \
859 (raw_inode)->xtime = cpu_to_le32(clamp_t(int32_t, (inode)->xtime.tv_sec, S32_MIN, S32_MAX)); \
860} while (0)
861
862#define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode) \
863do { \
864 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \
865 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec); \
866 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \
867 (raw_inode)->xtime ## _extra = \
868 ext4_encode_extra_time(&(einode)->xtime); \
869} while (0)
870
871#define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode) \
872do { \
873 (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime); \
874 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) { \
875 ext4_decode_extra_time(&(inode)->xtime, \
876 raw_inode->xtime ## _extra); \
877 } \
878 else \
879 (inode)->xtime.tv_nsec = 0; \
880} while (0)
881
882
883#define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode) \
884do { \
885 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \
886 (einode)->xtime.tv_sec = \
887 (signed)le32_to_cpu((raw_inode)->xtime); \
888 else \
889 (einode)->xtime.tv_sec = 0; \
890 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \
891 ext4_decode_extra_time(&(einode)->xtime, \
892 raw_inode->xtime ## _extra); \
893 else \
894 (einode)->xtime.tv_nsec = 0; \
895} while (0)
896
897#define i_disk_version osd1.linux1.l_i_version
898
899#if defined(__KERNEL__) || defined(__linux__)
900#define i_reserved1 osd1.linux1.l_i_reserved1
901#define i_file_acl_high osd2.linux2.l_i_file_acl_high
902#define i_blocks_high osd2.linux2.l_i_blocks_high
903#define i_uid_low i_uid
904#define i_gid_low i_gid
905#define i_uid_high osd2.linux2.l_i_uid_high
906#define i_gid_high osd2.linux2.l_i_gid_high
907#define i_checksum_lo osd2.linux2.l_i_checksum_lo
908
909#elif defined(__GNU__)
910
911#define i_translator osd1.hurd1.h_i_translator
912#define i_uid_high osd2.hurd2.h_i_uid_high
913#define i_gid_high osd2.hurd2.h_i_gid_high
914#define i_author osd2.hurd2.h_i_author
915
916#elif defined(__masix__)
917
918#define i_reserved1 osd1.masix1.m_i_reserved1
919#define i_file_acl_high osd2.masix2.m_i_file_acl_high
920#define i_reserved2 osd2.masix2.m_i_reserved2
921
922#endif /* defined(__KERNEL__) || defined(__linux__) */
923
924#include "extents_status.h"
925
926/*
927 * Lock subclasses for i_data_sem in the ext4_inode_info structure.
928 *
929 * These are needed to avoid lockdep false positives when we need to
930 * allocate blocks to the quota inode during ext4_map_blocks(), while
931 * holding i_data_sem for a normal (non-quota) inode. Since we don't
932 * do quota tracking for the quota inode, this avoids deadlock (as
933 * well as infinite recursion, since it isn't turtles all the way
934 * down...)
935 *
936 * I_DATA_SEM_NORMAL - Used for most inodes
937 * I_DATA_SEM_OTHER - Used by move_inode.c for the second normal inode
938 * where the second inode has larger inode number
939 * than the first
940 * I_DATA_SEM_QUOTA - Used for quota inodes only
941 */
942enum {
943 I_DATA_SEM_NORMAL = 0,
944 I_DATA_SEM_OTHER,
945 I_DATA_SEM_QUOTA,
946};
947
948
949/*
950 * fourth extended file system inode data in memory
951 */
952struct ext4_inode_info {
953 __le32 i_data[15]; /* unconverted */
954 __u32 i_dtime;
955 ext4_fsblk_t i_file_acl;
956
957 /*
958 * i_block_group is the number of the block group which contains
959 * this file's inode. Constant across the lifetime of the inode,
960 * it is used for making block allocation decisions - we try to
961 * place a file's data blocks near its inode block, and new inodes
962 * near to their parent directory's inode.
963 */
964 ext4_group_t i_block_group;
965 ext4_lblk_t i_dir_start_lookup;
966#if (BITS_PER_LONG < 64)
967 unsigned long i_state_flags; /* Dynamic state flags */
968#endif
969 unsigned long i_flags;
970
971 /*
972 * Extended attributes can be read independently of the main file
973 * data. Taking i_mutex even when reading would cause contention
974 * between readers of EAs and writers of regular file data, so
975 * instead we synchronize on xattr_sem when reading or changing
976 * EAs.
977 */
978 struct rw_semaphore xattr_sem;
979
980 struct list_head i_orphan; /* unlinked but open inodes */
981
982 /*
983 * i_disksize keeps track of what the inode size is ON DISK, not
984 * in memory. During truncate, i_size is set to the new size by
985 * the VFS prior to calling ext4_truncate(), but the filesystem won't
986 * set i_disksize to 0 until the truncate is actually under way.
987 *
988 * The intent is that i_disksize always represents the blocks which
989 * are used by this file. This allows recovery to restart truncate
990 * on orphans if we crash during truncate. We actually write i_disksize
991 * into the on-disk inode when writing inodes out, instead of i_size.
992 *
993 * The only time when i_disksize and i_size may be different is when
994 * a truncate is in progress. The only things which change i_disksize
995 * are ext4_get_block (growth) and ext4_truncate (shrinkth).
996 */
997 loff_t i_disksize;
998
999 /*
1000 * i_data_sem is for serialising ext4_truncate() against
1001 * ext4_getblock(). In the 2.4 ext2 design, great chunks of inode's
1002 * data tree are chopped off during truncate. We can't do that in
1003 * ext4 because whenever we perform intermediate commits during
1004 * truncate, the inode and all the metadata blocks *must* be in a
1005 * consistent state which allows truncation of the orphans to restart
1006 * during recovery. Hence we must fix the get_block-vs-truncate race
1007 * by other means, so we have i_data_sem.
1008 */
1009 struct rw_semaphore i_data_sem;
1010 /*
1011 * i_mmap_sem is for serializing page faults with truncate / punch hole
1012 * operations. We have to make sure that new page cannot be faulted in
1013 * a section of the inode that is being punched. We cannot easily use
1014 * i_data_sem for this since we need protection for the whole punch
1015 * operation and i_data_sem ranks below transaction start so we have
1016 * to occasionally drop it.
1017 */
1018 struct rw_semaphore i_mmap_sem;
1019 struct inode vfs_inode;
1020 struct jbd2_inode *jinode;
1021
1022 spinlock_t i_raw_lock; /* protects updates to the raw inode */
1023
1024 /*
1025 * File creation time. Its function is same as that of
1026 * struct timespec64 i_{a,c,m}time in the generic inode.
1027 */
1028 struct timespec64 i_crtime;
1029
1030 /* mballoc */
1031 struct list_head i_prealloc_list;
1032 spinlock_t i_prealloc_lock;
1033
1034 /* extents status tree */
1035 struct ext4_es_tree i_es_tree;
1036 rwlock_t i_es_lock;
1037 struct list_head i_es_list;
1038 unsigned int i_es_all_nr; /* protected by i_es_lock */
1039 unsigned int i_es_shk_nr; /* protected by i_es_lock */
1040 ext4_lblk_t i_es_shrink_lblk; /* Offset where we start searching for
1041 extents to shrink. Protected by
1042 i_es_lock */
1043
1044 /* ialloc */
1045 ext4_group_t i_last_alloc_group;
1046
1047 /* allocation reservation info for delalloc */
1048 /* In case of bigalloc, this refer to clusters rather than blocks */
1049 unsigned int i_reserved_data_blocks;
1050 ext4_lblk_t i_da_metadata_calc_last_lblock;
1051 int i_da_metadata_calc_len;
1052
1053 /* pending cluster reservations for bigalloc file systems */
1054 struct ext4_pending_tree i_pending_tree;
1055
1056 /* on-disk additional length */
1057 __u16 i_extra_isize;
1058
1059 /* Indicate the inline data space. */
1060 u16 i_inline_off;
1061 u16 i_inline_size;
1062
1063#ifdef CONFIG_QUOTA
1064 /* quota space reservation, managed internally by quota code */
1065 qsize_t i_reserved_quota;
1066#endif
1067
1068 /* Lock protecting lists below */
1069 spinlock_t i_completed_io_lock;
1070 /*
1071 * Completed IOs that need unwritten extents handling and have
1072 * transaction reserved
1073 */
1074 struct list_head i_rsv_conversion_list;
1075 struct work_struct i_rsv_conversion_work;
1076 atomic_t i_unwritten; /* Nr. of inflight conversions pending */
1077
1078 spinlock_t i_block_reservation_lock;
1079
1080 /*
1081 * Transactions that contain inode's metadata needed to complete
1082 * fsync and fdatasync, respectively.
1083 */
1084 tid_t i_sync_tid;
1085 tid_t i_datasync_tid;
1086
1087#ifdef CONFIG_QUOTA
1088 struct dquot *i_dquot[MAXQUOTAS];
1089#endif
1090
1091 /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
1092 __u32 i_csum_seed;
1093
1094 kprojid_t i_projid;
1095};
1096
1097/*
1098 * File system states
1099 */
1100#define EXT4_VALID_FS 0x0001 /* Unmounted cleanly */
1101#define EXT4_ERROR_FS 0x0002 /* Errors detected */
1102#define EXT4_ORPHAN_FS 0x0004 /* Orphans being recovered */
1103
1104/*
1105 * Misc. filesystem flags
1106 */
1107#define EXT2_FLAGS_SIGNED_HASH 0x0001 /* Signed dirhash in use */
1108#define EXT2_FLAGS_UNSIGNED_HASH 0x0002 /* Unsigned dirhash in use */
1109#define EXT2_FLAGS_TEST_FILESYS 0x0004 /* to test development code */
1110
1111/*
1112 * Mount flags set via mount options or defaults
1113 */
1114#define EXT4_MOUNT_NO_MBCACHE 0x00001 /* Do not use mbcache */
1115#define EXT4_MOUNT_GRPID 0x00004 /* Create files with directory's group */
1116#define EXT4_MOUNT_DEBUG 0x00008 /* Some debugging messages */
1117#define EXT4_MOUNT_ERRORS_CONT 0x00010 /* Continue on errors */
1118#define EXT4_MOUNT_ERRORS_RO 0x00020 /* Remount fs ro on errors */
1119#define EXT4_MOUNT_ERRORS_PANIC 0x00040 /* Panic on errors */
1120#define EXT4_MOUNT_ERRORS_MASK 0x00070
1121#define EXT4_MOUNT_MINIX_DF 0x00080 /* Mimics the Minix statfs */
1122#define EXT4_MOUNT_NOLOAD 0x00100 /* Don't use existing journal*/
1123#ifdef CONFIG_FS_DAX
1124#define EXT4_MOUNT_DAX 0x00200 /* Direct Access */
1125#else
1126#define EXT4_MOUNT_DAX 0
1127#endif
1128#define EXT4_MOUNT_DATA_FLAGS 0x00C00 /* Mode for data writes: */
1129#define EXT4_MOUNT_JOURNAL_DATA 0x00400 /* Write data to journal */
1130#define EXT4_MOUNT_ORDERED_DATA 0x00800 /* Flush data before commit */
1131#define EXT4_MOUNT_WRITEBACK_DATA 0x00C00 /* No data ordering */
1132#define EXT4_MOUNT_UPDATE_JOURNAL 0x01000 /* Update the journal format */
1133#define EXT4_MOUNT_NO_UID32 0x02000 /* Disable 32-bit UIDs */
1134#define EXT4_MOUNT_XATTR_USER 0x04000 /* Extended user attributes */
1135#define EXT4_MOUNT_POSIX_ACL 0x08000 /* POSIX Access Control Lists */
1136#define EXT4_MOUNT_NO_AUTO_DA_ALLOC 0x10000 /* No auto delalloc mapping */
1137#define EXT4_MOUNT_BARRIER 0x20000 /* Use block barriers */
1138#define EXT4_MOUNT_QUOTA 0x40000 /* Some quota option set */
1139#define EXT4_MOUNT_USRQUOTA 0x80000 /* "old" user quota,
1140 * enable enforcement for hidden
1141 * quota files */
1142#define EXT4_MOUNT_GRPQUOTA 0x100000 /* "old" group quota, enable
1143 * enforcement for hidden quota
1144 * files */
1145#define EXT4_MOUNT_PRJQUOTA 0x200000 /* Enable project quota
1146 * enforcement */
1147#define EXT4_MOUNT_DIOREAD_NOLOCK 0x400000 /* Enable support for dio read nolocking */
1148#define EXT4_MOUNT_JOURNAL_CHECKSUM 0x800000 /* Journal checksums */
1149#define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
1150#define EXT4_MOUNT_WARN_ON_ERROR 0x2000000 /* Trigger WARN_ON on error */
1151#define EXT4_MOUNT_DELALLOC 0x8000000 /* Delalloc support */
1152#define EXT4_MOUNT_DATA_ERR_ABORT 0x10000000 /* Abort on file data write */
1153#define EXT4_MOUNT_BLOCK_VALIDITY 0x20000000 /* Block validity checking */
1154#define EXT4_MOUNT_DISCARD 0x40000000 /* Issue DISCARD requests */
1155#define EXT4_MOUNT_INIT_INODE_TABLE 0x80000000 /* Initialize uninitialized itables */
1156
1157/*
1158 * Mount flags set either automatically (could not be set by mount option)
1159 * based on per file system feature or property or in special cases such as
1160 * distinguishing between explicit mount option definition and default.
1161 */
1162#define EXT4_MOUNT2_EXPLICIT_DELALLOC 0x00000001 /* User explicitly
1163 specified delalloc */
1164#define EXT4_MOUNT2_STD_GROUP_SIZE 0x00000002 /* We have standard group
1165 size of blocksize * 8
1166 blocks */
1167#define EXT4_MOUNT2_HURD_COMPAT 0x00000004 /* Support HURD-castrated
1168 file systems */
1169
1170#define EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM 0x00000008 /* User explicitly
1171 specified journal checksum */
1172
1173#define clear_opt(sb, opt) EXT4_SB(sb)->s_mount_opt &= \
1174 ~EXT4_MOUNT_##opt
1175#define set_opt(sb, opt) EXT4_SB(sb)->s_mount_opt |= \
1176 EXT4_MOUNT_##opt
1177#define test_opt(sb, opt) (EXT4_SB(sb)->s_mount_opt & \
1178 EXT4_MOUNT_##opt)
1179
1180#define clear_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 &= \
1181 ~EXT4_MOUNT2_##opt
1182#define set_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 |= \
1183 EXT4_MOUNT2_##opt
1184#define test_opt2(sb, opt) (EXT4_SB(sb)->s_mount_opt2 & \
1185 EXT4_MOUNT2_##opt)
1186
1187#define ext4_test_and_set_bit __test_and_set_bit_le
1188#define ext4_set_bit __set_bit_le
1189#define ext4_set_bit_atomic ext2_set_bit_atomic
1190#define ext4_test_and_clear_bit __test_and_clear_bit_le
1191#define ext4_clear_bit __clear_bit_le
1192#define ext4_clear_bit_atomic ext2_clear_bit_atomic
1193#define ext4_test_bit test_bit_le
1194#define ext4_find_next_zero_bit find_next_zero_bit_le
1195#define ext4_find_next_bit find_next_bit_le
1196
1197extern void ext4_set_bits(void *bm, int cur, int len);
1198
1199/*
1200 * Maximal mount counts between two filesystem checks
1201 */
1202#define EXT4_DFL_MAX_MNT_COUNT 20 /* Allow 20 mounts */
1203#define EXT4_DFL_CHECKINTERVAL 0 /* Don't use interval check */
1204
1205/*
1206 * Behaviour when detecting errors
1207 */
1208#define EXT4_ERRORS_CONTINUE 1 /* Continue execution */
1209#define EXT4_ERRORS_RO 2 /* Remount fs read-only */
1210#define EXT4_ERRORS_PANIC 3 /* Panic */
1211#define EXT4_ERRORS_DEFAULT EXT4_ERRORS_CONTINUE
1212
1213/* Metadata checksum algorithm codes */
1214#define EXT4_CRC32C_CHKSUM 1
1215
1216/*
1217 * Structure of the super block
1218 */
1219struct ext4_super_block {
1220/*00*/ __le32 s_inodes_count; /* Inodes count */
1221 __le32 s_blocks_count_lo; /* Blocks count */
1222 __le32 s_r_blocks_count_lo; /* Reserved blocks count */
1223 __le32 s_free_blocks_count_lo; /* Free blocks count */
1224/*10*/ __le32 s_free_inodes_count; /* Free inodes count */
1225 __le32 s_first_data_block; /* First Data Block */
1226 __le32 s_log_block_size; /* Block size */
1227 __le32 s_log_cluster_size; /* Allocation cluster size */
1228/*20*/ __le32 s_blocks_per_group; /* # Blocks per group */
1229 __le32 s_clusters_per_group; /* # Clusters per group */
1230 __le32 s_inodes_per_group; /* # Inodes per group */
1231 __le32 s_mtime; /* Mount time */
1232/*30*/ __le32 s_wtime; /* Write time */
1233 __le16 s_mnt_count; /* Mount count */
1234 __le16 s_max_mnt_count; /* Maximal mount count */
1235 __le16 s_magic; /* Magic signature */
1236 __le16 s_state; /* File system state */
1237 __le16 s_errors; /* Behaviour when detecting errors */
1238 __le16 s_minor_rev_level; /* minor revision level */
1239/*40*/ __le32 s_lastcheck; /* time of last check */
1240 __le32 s_checkinterval; /* max. time between checks */
1241 __le32 s_creator_os; /* OS */
1242 __le32 s_rev_level; /* Revision level */
1243/*50*/ __le16 s_def_resuid; /* Default uid for reserved blocks */
1244 __le16 s_def_resgid; /* Default gid for reserved blocks */
1245 /*
1246 * These fields are for EXT4_DYNAMIC_REV superblocks only.
1247 *
1248 * Note: the difference between the compatible feature set and
1249 * the incompatible feature set is that if there is a bit set
1250 * in the incompatible feature set that the kernel doesn't
1251 * know about, it should refuse to mount the filesystem.
1252 *
1253 * e2fsck's requirements are more strict; if it doesn't know
1254 * about a feature in either the compatible or incompatible
1255 * feature set, it must abort and not try to meddle with
1256 * things it doesn't understand...
1257 */
1258 __le32 s_first_ino; /* First non-reserved inode */
1259 __le16 s_inode_size; /* size of inode structure */
1260 __le16 s_block_group_nr; /* block group # of this superblock */
1261 __le32 s_feature_compat; /* compatible feature set */
1262/*60*/ __le32 s_feature_incompat; /* incompatible feature set */
1263 __le32 s_feature_ro_compat; /* readonly-compatible feature set */
1264/*68*/ __u8 s_uuid[16]; /* 128-bit uuid for volume */
1265/*78*/ char s_volume_name[16]; /* volume name */
1266/*88*/ char s_last_mounted[64] __nonstring; /* directory where last mounted */
1267/*C8*/ __le32 s_algorithm_usage_bitmap; /* For compression */
1268 /*
1269 * Performance hints. Directory preallocation should only
1270 * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1271 */
1272 __u8 s_prealloc_blocks; /* Nr of blocks to try to preallocate*/
1273 __u8 s_prealloc_dir_blocks; /* Nr to preallocate for dirs */
1274 __le16 s_reserved_gdt_blocks; /* Per group desc for online growth */
1275 /*
1276 * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1277 */
1278/*D0*/ __u8 s_journal_uuid[16]; /* uuid of journal superblock */
1279/*E0*/ __le32 s_journal_inum; /* inode number of journal file */
1280 __le32 s_journal_dev; /* device number of journal file */
1281 __le32 s_last_orphan; /* start of list of inodes to delete */
1282 __le32 s_hash_seed[4]; /* HTREE hash seed */
1283 __u8 s_def_hash_version; /* Default hash version to use */
1284 __u8 s_jnl_backup_type;
1285 __le16 s_desc_size; /* size of group descriptor */
1286/*100*/ __le32 s_default_mount_opts;
1287 __le32 s_first_meta_bg; /* First metablock block group */
1288 __le32 s_mkfs_time; /* When the filesystem was created */
1289 __le32 s_jnl_blocks[17]; /* Backup of the journal inode */
1290 /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1291/*150*/ __le32 s_blocks_count_hi; /* Blocks count */
1292 __le32 s_r_blocks_count_hi; /* Reserved blocks count */
1293 __le32 s_free_blocks_count_hi; /* Free blocks count */
1294 __le16 s_min_extra_isize; /* All inodes have at least # bytes */
1295 __le16 s_want_extra_isize; /* New inodes should reserve # bytes */
1296 __le32 s_flags; /* Miscellaneous flags */
1297 __le16 s_raid_stride; /* RAID stride */
1298 __le16 s_mmp_update_interval; /* # seconds to wait in MMP checking */
1299 __le64 s_mmp_block; /* Block for multi-mount protection */
1300 __le32 s_raid_stripe_width; /* blocks on all data disks (N*stride)*/
1301 __u8 s_log_groups_per_flex; /* FLEX_BG group size */
1302 __u8 s_checksum_type; /* metadata checksum algorithm used */
1303 __u8 s_encryption_level; /* versioning level for encryption */
1304 __u8 s_reserved_pad; /* Padding to next 32bits */
1305 __le64 s_kbytes_written; /* nr of lifetime kilobytes written */
1306 __le32 s_snapshot_inum; /* Inode number of active snapshot */
1307 __le32 s_snapshot_id; /* sequential ID of active snapshot */
1308 __le64 s_snapshot_r_blocks_count; /* reserved blocks for active
1309 snapshot's future use */
1310 __le32 s_snapshot_list; /* inode number of the head of the
1311 on-disk snapshot list */
1312#define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1313 __le32 s_error_count; /* number of fs errors */
1314 __le32 s_first_error_time; /* first time an error happened */
1315 __le32 s_first_error_ino; /* inode involved in first error */
1316 __le64 s_first_error_block; /* block involved of first error */
1317 __u8 s_first_error_func[32] __nonstring; /* function where the error happened */
1318 __le32 s_first_error_line; /* line number where error happened */
1319 __le32 s_last_error_time; /* most recent time of an error */
1320 __le32 s_last_error_ino; /* inode involved in last error */
1321 __le32 s_last_error_line; /* line number where error happened */
1322 __le64 s_last_error_block; /* block involved of last error */
1323 __u8 s_last_error_func[32] __nonstring; /* function where the error happened */
1324#define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1325 __u8 s_mount_opts[64];
1326 __le32 s_usr_quota_inum; /* inode for tracking user quota */
1327 __le32 s_grp_quota_inum; /* inode for tracking group quota */
1328 __le32 s_overhead_clusters; /* overhead blocks/clusters in fs */
1329 __le32 s_backup_bgs[2]; /* groups with sparse_super2 SBs */
1330 __u8 s_encrypt_algos[4]; /* Encryption algorithms in use */
1331 __u8 s_encrypt_pw_salt[16]; /* Salt used for string2key algorithm */
1332 __le32 s_lpf_ino; /* Location of the lost+found inode */
1333 __le32 s_prj_quota_inum; /* inode for tracking project quota */
1334 __le32 s_checksum_seed; /* crc32c(uuid) if csum_seed set */
1335 __u8 s_wtime_hi;
1336 __u8 s_mtime_hi;
1337 __u8 s_mkfs_time_hi;
1338 __u8 s_lastcheck_hi;
1339 __u8 s_first_error_time_hi;
1340 __u8 s_last_error_time_hi;
1341 __u8 s_pad[2];
1342 __le16 s_encoding; /* Filename charset encoding */
1343 __le16 s_encoding_flags; /* Filename charset encoding flags */
1344 __le32 s_reserved[95]; /* Padding to the end of the block */
1345 __le32 s_checksum; /* crc32c(superblock) */
1346};
1347
1348#define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1349
1350#ifdef __KERNEL__
1351
1352/*
1353 * run-time mount flags
1354 */
1355#define EXT4_MF_MNTDIR_SAMPLED 0x0001
1356#define EXT4_MF_FS_ABORTED 0x0002 /* Fatal error detected */
1357#define EXT4_MF_TEST_DUMMY_ENCRYPTION 0x0004
1358
1359#ifdef CONFIG_FS_ENCRYPTION
1360#define DUMMY_ENCRYPTION_ENABLED(sbi) (unlikely((sbi)->s_mount_flags & \
1361 EXT4_MF_TEST_DUMMY_ENCRYPTION))
1362#else
1363#define DUMMY_ENCRYPTION_ENABLED(sbi) (0)
1364#endif
1365
1366/* Number of quota types we support */
1367#define EXT4_MAXQUOTAS 3
1368
1369#define EXT4_ENC_UTF8_12_1 1
1370
1371/*
1372 * Flags for ext4_sb_info.s_encoding_flags.
1373 */
1374#define EXT4_ENC_STRICT_MODE_FL (1 << 0)
1375
1376#define ext4_has_strict_mode(sbi) \
1377 (sbi->s_encoding_flags & EXT4_ENC_STRICT_MODE_FL)
1378
1379/*
1380 * fourth extended-fs super-block data in memory
1381 */
1382struct ext4_sb_info {
1383 unsigned long s_desc_size; /* Size of a group descriptor in bytes */
1384 unsigned long s_inodes_per_block;/* Number of inodes per block */
1385 unsigned long s_blocks_per_group;/* Number of blocks in a group */
1386 unsigned long s_clusters_per_group; /* Number of clusters in a group */
1387 unsigned long s_inodes_per_group;/* Number of inodes in a group */
1388 unsigned long s_itb_per_group; /* Number of inode table blocks per group */
1389 unsigned long s_gdb_count; /* Number of group descriptor blocks */
1390 unsigned long s_desc_per_block; /* Number of group descriptors per block */
1391 ext4_group_t s_groups_count; /* Number of groups in the fs */
1392 ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1393 unsigned long s_overhead; /* # of fs overhead clusters */
1394 unsigned int s_cluster_ratio; /* Number of blocks per cluster */
1395 unsigned int s_cluster_bits; /* log2 of s_cluster_ratio */
1396 loff_t s_bitmap_maxbytes; /* max bytes for bitmap files */
1397 struct buffer_head * s_sbh; /* Buffer containing the super block */
1398 struct ext4_super_block *s_es; /* Pointer to the super block in the buffer */
1399 struct buffer_head **s_group_desc;
1400 unsigned int s_mount_opt;
1401 unsigned int s_mount_opt2;
1402 unsigned int s_mount_flags;
1403 unsigned int s_def_mount_opt;
1404 ext4_fsblk_t s_sb_block;
1405 atomic64_t s_resv_clusters;
1406 kuid_t s_resuid;
1407 kgid_t s_resgid;
1408 unsigned short s_mount_state;
1409 unsigned short s_pad;
1410 int s_addr_per_block_bits;
1411 int s_desc_per_block_bits;
1412 int s_inode_size;
1413 int s_first_ino;
1414 unsigned int s_inode_readahead_blks;
1415 unsigned int s_inode_goal;
1416 u32 s_hash_seed[4];
1417 int s_def_hash_version;
1418 int s_hash_unsigned; /* 3 if hash should be signed, 0 if not */
1419 struct percpu_counter s_freeclusters_counter;
1420 struct percpu_counter s_freeinodes_counter;
1421 struct percpu_counter s_dirs_counter;
1422 struct percpu_counter s_dirtyclusters_counter;
1423 struct blockgroup_lock *s_blockgroup_lock;
1424 struct proc_dir_entry *s_proc;
1425 struct kobject s_kobj;
1426 struct completion s_kobj_unregister;
1427 struct super_block *s_sb;
1428#ifdef CONFIG_UNICODE
1429 struct unicode_map *s_encoding;
1430 __u16 s_encoding_flags;
1431#endif
1432
1433 /* Journaling */
1434 struct journal_s *s_journal;
1435 struct list_head s_orphan;
1436 struct mutex s_orphan_lock;
1437 unsigned long s_ext4_flags; /* Ext4 superblock flags */
1438 unsigned long s_commit_interval;
1439 u32 s_max_batch_time;
1440 u32 s_min_batch_time;
1441 struct block_device *journal_bdev;
1442#ifdef CONFIG_QUOTA
1443 /* Names of quota files with journalled quota */
1444 char __rcu *s_qf_names[EXT4_MAXQUOTAS];
1445 int s_jquota_fmt; /* Format of quota to use */
1446#endif
1447 unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1448 struct ext4_system_blocks __rcu *system_blks;
1449
1450#ifdef EXTENTS_STATS
1451 /* ext4 extents stats */
1452 unsigned long s_ext_min;
1453 unsigned long s_ext_max;
1454 unsigned long s_depth_max;
1455 spinlock_t s_ext_stats_lock;
1456 unsigned long s_ext_blocks;
1457 unsigned long s_ext_extents;
1458#endif
1459
1460 /* for buddy allocator */
1461 struct ext4_group_info ***s_group_info;
1462 struct inode *s_buddy_cache;
1463 spinlock_t s_md_lock;
1464 unsigned short *s_mb_offsets;
1465 unsigned int *s_mb_maxs;
1466 unsigned int s_group_info_size;
1467 unsigned int s_mb_free_pending;
1468 struct list_head s_freed_data_list; /* List of blocks to be freed
1469 after commit completed */
1470
1471 /* tunables */
1472 unsigned long s_stripe;
1473 unsigned int s_mb_stream_request;
1474 unsigned int s_mb_max_to_scan;
1475 unsigned int s_mb_min_to_scan;
1476 unsigned int s_mb_stats;
1477 unsigned int s_mb_order2_reqs;
1478 unsigned int s_mb_group_prealloc;
1479 unsigned int s_max_dir_size_kb;
1480 /* where last allocation was done - for stream allocation */
1481 unsigned long s_mb_last_group;
1482 unsigned long s_mb_last_start;
1483
1484 /* stats for buddy allocator */
1485 atomic_t s_bal_reqs; /* number of reqs with len > 1 */
1486 atomic_t s_bal_success; /* we found long enough chunks */
1487 atomic_t s_bal_allocated; /* in blocks */
1488 atomic_t s_bal_ex_scanned; /* total extents scanned */
1489 atomic_t s_bal_goals; /* goal hits */
1490 atomic_t s_bal_breaks; /* too long searches */
1491 atomic_t s_bal_2orders; /* 2^order hits */
1492 spinlock_t s_bal_lock;
1493 unsigned long s_mb_buddies_generated;
1494 unsigned long long s_mb_generation_time;
1495 atomic_t s_mb_lost_chunks;
1496 atomic_t s_mb_preallocated;
1497 atomic_t s_mb_discarded;
1498 atomic_t s_lock_busy;
1499
1500 /* locality groups */
1501 struct ext4_locality_group __percpu *s_locality_groups;
1502
1503 /* for write statistics */
1504 unsigned long s_sectors_written_start;
1505 u64 s_kbytes_written;
1506
1507 /* the size of zero-out chunk */
1508 unsigned int s_extent_max_zeroout_kb;
1509
1510 unsigned int s_log_groups_per_flex;
1511 struct flex_groups *s_flex_groups;
1512 ext4_group_t s_flex_groups_allocated;
1513
1514 /* workqueue for reserved extent conversions (buffered io) */
1515 struct workqueue_struct *rsv_conversion_wq;
1516
1517 /* timer for periodic error stats printing */
1518 struct timer_list s_err_report;
1519
1520 /* Lazy inode table initialization info */
1521 struct ext4_li_request *s_li_request;
1522 /* Wait multiplier for lazy initialization thread */
1523 unsigned int s_li_wait_mult;
1524
1525 /* Kernel thread for multiple mount protection */
1526 struct task_struct *s_mmp_tsk;
1527
1528 /* record the last minlen when FITRIM is called. */
1529 atomic_t s_last_trim_minblks;
1530
1531 /* Reference to checksum algorithm driver via cryptoapi */
1532 struct crypto_shash *s_chksum_driver;
1533
1534 /* Precomputed FS UUID checksum for seeding other checksums */
1535 __u32 s_csum_seed;
1536
1537 /* Reclaim extents from extent status tree */
1538 struct shrinker s_es_shrinker;
1539 struct list_head s_es_list; /* List of inodes with reclaimable extents */
1540 long s_es_nr_inode;
1541 struct ext4_es_stats s_es_stats;
1542 struct mb_cache *s_ea_block_cache;
1543 struct mb_cache *s_ea_inode_cache;
1544 spinlock_t s_es_lock ____cacheline_aligned_in_smp;
1545
1546 /* Ratelimit ext4 messages. */
1547 struct ratelimit_state s_err_ratelimit_state;
1548 struct ratelimit_state s_warning_ratelimit_state;
1549 struct ratelimit_state s_msg_ratelimit_state;
1550
1551 /* Barrier between changing inodes' journal flags and writepages ops. */
1552 struct percpu_rw_semaphore s_journal_flag_rwsem;
1553 struct dax_device *s_daxdev;
1554};
1555
1556static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1557{
1558 return sb->s_fs_info;
1559}
1560static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1561{
1562 return container_of(inode, struct ext4_inode_info, vfs_inode);
1563}
1564
1565static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1566{
1567 return ino == EXT4_ROOT_INO ||
1568 (ino >= EXT4_FIRST_INO(sb) &&
1569 ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1570}
1571
1572/*
1573 * Inode dynamic state flags
1574 */
1575enum {
1576 EXT4_STATE_JDATA, /* journaled data exists */
1577 EXT4_STATE_NEW, /* inode is newly created */
1578 EXT4_STATE_XATTR, /* has in-inode xattrs */
1579 EXT4_STATE_NO_EXPAND, /* No space for expansion */
1580 EXT4_STATE_DA_ALLOC_CLOSE, /* Alloc DA blks on close */
1581 EXT4_STATE_EXT_MIGRATE, /* Inode is migrating */
1582 EXT4_STATE_DIO_UNWRITTEN, /* need convert on dio done*/
1583 EXT4_STATE_NEWENTRY, /* File just added to dir */
1584 EXT4_STATE_MAY_INLINE_DATA, /* may have in-inode data */
1585 EXT4_STATE_EXT_PRECACHED, /* extents have been precached */
1586 EXT4_STATE_LUSTRE_EA_INODE, /* Lustre-style ea_inode */
1587 EXT4_STATE_VERITY_IN_PROGRESS, /* building fs-verity Merkle tree */
1588};
1589
1590#define EXT4_INODE_BIT_FNS(name, field, offset) \
1591static inline int ext4_test_inode_##name(struct inode *inode, int bit) \
1592{ \
1593 return test_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1594} \
1595static inline void ext4_set_inode_##name(struct inode *inode, int bit) \
1596{ \
1597 set_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1598} \
1599static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1600{ \
1601 clear_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1602}
1603
1604/* Add these declarations here only so that these functions can be
1605 * found by name. Otherwise, they are very hard to locate. */
1606static inline int ext4_test_inode_flag(struct inode *inode, int bit);
1607static inline void ext4_set_inode_flag(struct inode *inode, int bit);
1608static inline void ext4_clear_inode_flag(struct inode *inode, int bit);
1609EXT4_INODE_BIT_FNS(flag, flags, 0)
1610
1611/* Add these declarations here only so that these functions can be
1612 * found by name. Otherwise, they are very hard to locate. */
1613static inline int ext4_test_inode_state(struct inode *inode, int bit);
1614static inline void ext4_set_inode_state(struct inode *inode, int bit);
1615static inline void ext4_clear_inode_state(struct inode *inode, int bit);
1616#if (BITS_PER_LONG < 64)
1617EXT4_INODE_BIT_FNS(state, state_flags, 0)
1618
1619static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1620{
1621 (ei)->i_state_flags = 0;
1622}
1623#else
1624EXT4_INODE_BIT_FNS(state, flags, 32)
1625
1626static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1627{
1628 /* We depend on the fact that callers will set i_flags */
1629}
1630#endif
1631#else
1632/* Assume that user mode programs are passing in an ext4fs superblock, not
1633 * a kernel struct super_block. This will allow us to call the feature-test
1634 * macros from user land. */
1635#define EXT4_SB(sb) (sb)
1636#endif
1637
1638static inline bool ext4_verity_in_progress(struct inode *inode)
1639{
1640 return IS_ENABLED(CONFIG_FS_VERITY) &&
1641 ext4_test_inode_state(inode, EXT4_STATE_VERITY_IN_PROGRESS);
1642}
1643
1644#define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1645
1646/*
1647 * Codes for operating systems
1648 */
1649#define EXT4_OS_LINUX 0
1650#define EXT4_OS_HURD 1
1651#define EXT4_OS_MASIX 2
1652#define EXT4_OS_FREEBSD 3
1653#define EXT4_OS_LITES 4
1654
1655/*
1656 * Revision levels
1657 */
1658#define EXT4_GOOD_OLD_REV 0 /* The good old (original) format */
1659#define EXT4_DYNAMIC_REV 1 /* V2 format w/ dynamic inode sizes */
1660
1661#define EXT4_CURRENT_REV EXT4_GOOD_OLD_REV
1662#define EXT4_MAX_SUPP_REV EXT4_DYNAMIC_REV
1663
1664#define EXT4_GOOD_OLD_INODE_SIZE 128
1665
1666#define EXT4_EXTRA_TIMESTAMP_MAX (((s64)1 << 34) - 1 + S32_MIN)
1667#define EXT4_NON_EXTRA_TIMESTAMP_MAX S32_MAX
1668#define EXT4_TIMESTAMP_MIN S32_MIN
1669
1670/*
1671 * Feature set definitions
1672 */
1673
1674#define EXT4_FEATURE_COMPAT_DIR_PREALLOC 0x0001
1675#define EXT4_FEATURE_COMPAT_IMAGIC_INODES 0x0002
1676#define EXT4_FEATURE_COMPAT_HAS_JOURNAL 0x0004
1677#define EXT4_FEATURE_COMPAT_EXT_ATTR 0x0008
1678#define EXT4_FEATURE_COMPAT_RESIZE_INODE 0x0010
1679#define EXT4_FEATURE_COMPAT_DIR_INDEX 0x0020
1680#define EXT4_FEATURE_COMPAT_SPARSE_SUPER2 0x0200
1681
1682#define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER 0x0001
1683#define EXT4_FEATURE_RO_COMPAT_LARGE_FILE 0x0002
1684#define EXT4_FEATURE_RO_COMPAT_BTREE_DIR 0x0004
1685#define EXT4_FEATURE_RO_COMPAT_HUGE_FILE 0x0008
1686#define EXT4_FEATURE_RO_COMPAT_GDT_CSUM 0x0010
1687#define EXT4_FEATURE_RO_COMPAT_DIR_NLINK 0x0020
1688#define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE 0x0040
1689#define EXT4_FEATURE_RO_COMPAT_QUOTA 0x0100
1690#define EXT4_FEATURE_RO_COMPAT_BIGALLOC 0x0200
1691/*
1692 * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM). When
1693 * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
1694 * all other data structures' checksums. However, the METADATA_CSUM and
1695 * GDT_CSUM bits are mutually exclusive.
1696 */
1697#define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM 0x0400
1698#define EXT4_FEATURE_RO_COMPAT_READONLY 0x1000
1699#define EXT4_FEATURE_RO_COMPAT_PROJECT 0x2000
1700#define EXT4_FEATURE_RO_COMPAT_VERITY 0x8000
1701
1702#define EXT4_FEATURE_INCOMPAT_COMPRESSION 0x0001
1703#define EXT4_FEATURE_INCOMPAT_FILETYPE 0x0002
1704#define EXT4_FEATURE_INCOMPAT_RECOVER 0x0004 /* Needs recovery */
1705#define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 /* Journal device */
1706#define EXT4_FEATURE_INCOMPAT_META_BG 0x0010
1707#define EXT4_FEATURE_INCOMPAT_EXTENTS 0x0040 /* extents support */
1708#define EXT4_FEATURE_INCOMPAT_64BIT 0x0080
1709#define EXT4_FEATURE_INCOMPAT_MMP 0x0100
1710#define EXT4_FEATURE_INCOMPAT_FLEX_BG 0x0200
1711#define EXT4_FEATURE_INCOMPAT_EA_INODE 0x0400 /* EA in inode */
1712#define EXT4_FEATURE_INCOMPAT_DIRDATA 0x1000 /* data in dirent */
1713#define EXT4_FEATURE_INCOMPAT_CSUM_SEED 0x2000
1714#define EXT4_FEATURE_INCOMPAT_LARGEDIR 0x4000 /* >2GB or 3-lvl htree */
1715#define EXT4_FEATURE_INCOMPAT_INLINE_DATA 0x8000 /* data in inode */
1716#define EXT4_FEATURE_INCOMPAT_ENCRYPT 0x10000
1717#define EXT4_FEATURE_INCOMPAT_CASEFOLD 0x20000
1718
1719extern void ext4_update_dynamic_rev(struct super_block *sb);
1720
1721#define EXT4_FEATURE_COMPAT_FUNCS(name, flagname) \
1722static inline bool ext4_has_feature_##name(struct super_block *sb) \
1723{ \
1724 return ((EXT4_SB(sb)->s_es->s_feature_compat & \
1725 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname)) != 0); \
1726} \
1727static inline void ext4_set_feature_##name(struct super_block *sb) \
1728{ \
1729 ext4_update_dynamic_rev(sb); \
1730 EXT4_SB(sb)->s_es->s_feature_compat |= \
1731 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
1732} \
1733static inline void ext4_clear_feature_##name(struct super_block *sb) \
1734{ \
1735 EXT4_SB(sb)->s_es->s_feature_compat &= \
1736 ~cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
1737}
1738
1739#define EXT4_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
1740static inline bool ext4_has_feature_##name(struct super_block *sb) \
1741{ \
1742 return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
1743 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname)) != 0); \
1744} \
1745static inline void ext4_set_feature_##name(struct super_block *sb) \
1746{ \
1747 ext4_update_dynamic_rev(sb); \
1748 EXT4_SB(sb)->s_es->s_feature_ro_compat |= \
1749 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
1750} \
1751static inline void ext4_clear_feature_##name(struct super_block *sb) \
1752{ \
1753 EXT4_SB(sb)->s_es->s_feature_ro_compat &= \
1754 ~cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
1755}
1756
1757#define EXT4_FEATURE_INCOMPAT_FUNCS(name, flagname) \
1758static inline bool ext4_has_feature_##name(struct super_block *sb) \
1759{ \
1760 return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
1761 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname)) != 0); \
1762} \
1763static inline void ext4_set_feature_##name(struct super_block *sb) \
1764{ \
1765 ext4_update_dynamic_rev(sb); \
1766 EXT4_SB(sb)->s_es->s_feature_incompat |= \
1767 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
1768} \
1769static inline void ext4_clear_feature_##name(struct super_block *sb) \
1770{ \
1771 EXT4_SB(sb)->s_es->s_feature_incompat &= \
1772 ~cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
1773}
1774
1775EXT4_FEATURE_COMPAT_FUNCS(dir_prealloc, DIR_PREALLOC)
1776EXT4_FEATURE_COMPAT_FUNCS(imagic_inodes, IMAGIC_INODES)
1777EXT4_FEATURE_COMPAT_FUNCS(journal, HAS_JOURNAL)
1778EXT4_FEATURE_COMPAT_FUNCS(xattr, EXT_ATTR)
1779EXT4_FEATURE_COMPAT_FUNCS(resize_inode, RESIZE_INODE)
1780EXT4_FEATURE_COMPAT_FUNCS(dir_index, DIR_INDEX)
1781EXT4_FEATURE_COMPAT_FUNCS(sparse_super2, SPARSE_SUPER2)
1782
1783EXT4_FEATURE_RO_COMPAT_FUNCS(sparse_super, SPARSE_SUPER)
1784EXT4_FEATURE_RO_COMPAT_FUNCS(large_file, LARGE_FILE)
1785EXT4_FEATURE_RO_COMPAT_FUNCS(btree_dir, BTREE_DIR)
1786EXT4_FEATURE_RO_COMPAT_FUNCS(huge_file, HUGE_FILE)
1787EXT4_FEATURE_RO_COMPAT_FUNCS(gdt_csum, GDT_CSUM)
1788EXT4_FEATURE_RO_COMPAT_FUNCS(dir_nlink, DIR_NLINK)
1789EXT4_FEATURE_RO_COMPAT_FUNCS(extra_isize, EXTRA_ISIZE)
1790EXT4_FEATURE_RO_COMPAT_FUNCS(quota, QUOTA)
1791EXT4_FEATURE_RO_COMPAT_FUNCS(bigalloc, BIGALLOC)
1792EXT4_FEATURE_RO_COMPAT_FUNCS(metadata_csum, METADATA_CSUM)
1793EXT4_FEATURE_RO_COMPAT_FUNCS(readonly, READONLY)
1794EXT4_FEATURE_RO_COMPAT_FUNCS(project, PROJECT)
1795EXT4_FEATURE_RO_COMPAT_FUNCS(verity, VERITY)
1796
1797EXT4_FEATURE_INCOMPAT_FUNCS(compression, COMPRESSION)
1798EXT4_FEATURE_INCOMPAT_FUNCS(filetype, FILETYPE)
1799EXT4_FEATURE_INCOMPAT_FUNCS(journal_needs_recovery, RECOVER)
1800EXT4_FEATURE_INCOMPAT_FUNCS(journal_dev, JOURNAL_DEV)
1801EXT4_FEATURE_INCOMPAT_FUNCS(meta_bg, META_BG)
1802EXT4_FEATURE_INCOMPAT_FUNCS(extents, EXTENTS)
1803EXT4_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT)
1804EXT4_FEATURE_INCOMPAT_FUNCS(mmp, MMP)
1805EXT4_FEATURE_INCOMPAT_FUNCS(flex_bg, FLEX_BG)
1806EXT4_FEATURE_INCOMPAT_FUNCS(ea_inode, EA_INODE)
1807EXT4_FEATURE_INCOMPAT_FUNCS(dirdata, DIRDATA)
1808EXT4_FEATURE_INCOMPAT_FUNCS(csum_seed, CSUM_SEED)
1809EXT4_FEATURE_INCOMPAT_FUNCS(largedir, LARGEDIR)
1810EXT4_FEATURE_INCOMPAT_FUNCS(inline_data, INLINE_DATA)
1811EXT4_FEATURE_INCOMPAT_FUNCS(encrypt, ENCRYPT)
1812EXT4_FEATURE_INCOMPAT_FUNCS(casefold, CASEFOLD)
1813
1814#define EXT2_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR
1815#define EXT2_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1816 EXT4_FEATURE_INCOMPAT_META_BG)
1817#define EXT2_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1818 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1819 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1820
1821#define EXT3_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR
1822#define EXT3_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1823 EXT4_FEATURE_INCOMPAT_RECOVER| \
1824 EXT4_FEATURE_INCOMPAT_META_BG)
1825#define EXT3_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1826 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1827 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1828
1829#define EXT4_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR
1830#define EXT4_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1831 EXT4_FEATURE_INCOMPAT_RECOVER| \
1832 EXT4_FEATURE_INCOMPAT_META_BG| \
1833 EXT4_FEATURE_INCOMPAT_EXTENTS| \
1834 EXT4_FEATURE_INCOMPAT_64BIT| \
1835 EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1836 EXT4_FEATURE_INCOMPAT_EA_INODE| \
1837 EXT4_FEATURE_INCOMPAT_MMP | \
1838 EXT4_FEATURE_INCOMPAT_INLINE_DATA | \
1839 EXT4_FEATURE_INCOMPAT_ENCRYPT | \
1840 EXT4_FEATURE_INCOMPAT_CASEFOLD | \
1841 EXT4_FEATURE_INCOMPAT_CSUM_SEED | \
1842 EXT4_FEATURE_INCOMPAT_LARGEDIR)
1843#define EXT4_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1844 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1845 EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1846 EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1847 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1848 EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1849 EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1850 EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
1851 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
1852 EXT4_FEATURE_RO_COMPAT_QUOTA |\
1853 EXT4_FEATURE_RO_COMPAT_PROJECT |\
1854 EXT4_FEATURE_RO_COMPAT_VERITY)
1855
1856#define EXTN_FEATURE_FUNCS(ver) \
1857static inline bool ext4_has_unknown_ext##ver##_compat_features(struct super_block *sb) \
1858{ \
1859 return ((EXT4_SB(sb)->s_es->s_feature_compat & \
1860 cpu_to_le32(~EXT##ver##_FEATURE_COMPAT_SUPP)) != 0); \
1861} \
1862static inline bool ext4_has_unknown_ext##ver##_ro_compat_features(struct super_block *sb) \
1863{ \
1864 return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
1865 cpu_to_le32(~EXT##ver##_FEATURE_RO_COMPAT_SUPP)) != 0); \
1866} \
1867static inline bool ext4_has_unknown_ext##ver##_incompat_features(struct super_block *sb) \
1868{ \
1869 return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
1870 cpu_to_le32(~EXT##ver##_FEATURE_INCOMPAT_SUPP)) != 0); \
1871}
1872
1873EXTN_FEATURE_FUNCS(2)
1874EXTN_FEATURE_FUNCS(3)
1875EXTN_FEATURE_FUNCS(4)
1876
1877static inline bool ext4_has_compat_features(struct super_block *sb)
1878{
1879 return (EXT4_SB(sb)->s_es->s_feature_compat != 0);
1880}
1881static inline bool ext4_has_ro_compat_features(struct super_block *sb)
1882{
1883 return (EXT4_SB(sb)->s_es->s_feature_ro_compat != 0);
1884}
1885static inline bool ext4_has_incompat_features(struct super_block *sb)
1886{
1887 return (EXT4_SB(sb)->s_es->s_feature_incompat != 0);
1888}
1889
1890/*
1891 * Superblock flags
1892 */
1893#define EXT4_FLAGS_RESIZING 0
1894#define EXT4_FLAGS_SHUTDOWN 1
1895
1896static inline int ext4_forced_shutdown(struct ext4_sb_info *sbi)
1897{
1898 return test_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
1899}
1900
1901
1902/*
1903 * Default values for user and/or group using reserved blocks
1904 */
1905#define EXT4_DEF_RESUID 0
1906#define EXT4_DEF_RESGID 0
1907
1908/*
1909 * Default project ID
1910 */
1911#define EXT4_DEF_PROJID 0
1912
1913#define EXT4_DEF_INODE_READAHEAD_BLKS 32
1914
1915/*
1916 * Default mount options
1917 */
1918#define EXT4_DEFM_DEBUG 0x0001
1919#define EXT4_DEFM_BSDGROUPS 0x0002
1920#define EXT4_DEFM_XATTR_USER 0x0004
1921#define EXT4_DEFM_ACL 0x0008
1922#define EXT4_DEFM_UID16 0x0010
1923#define EXT4_DEFM_JMODE 0x0060
1924#define EXT4_DEFM_JMODE_DATA 0x0020
1925#define EXT4_DEFM_JMODE_ORDERED 0x0040
1926#define EXT4_DEFM_JMODE_WBACK 0x0060
1927#define EXT4_DEFM_NOBARRIER 0x0100
1928#define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1929#define EXT4_DEFM_DISCARD 0x0400
1930#define EXT4_DEFM_NODELALLOC 0x0800
1931
1932/*
1933 * Default journal batch times
1934 */
1935#define EXT4_DEF_MIN_BATCH_TIME 0
1936#define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1937
1938/*
1939 * Minimum number of groups in a flexgroup before we separate out
1940 * directories into the first block group of a flexgroup
1941 */
1942#define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1943
1944/*
1945 * Structure of a directory entry
1946 */
1947#define EXT4_NAME_LEN 255
1948
1949struct ext4_dir_entry {
1950 __le32 inode; /* Inode number */
1951 __le16 rec_len; /* Directory entry length */
1952 __le16 name_len; /* Name length */
1953 char name[EXT4_NAME_LEN]; /* File name */
1954};
1955
1956/*
1957 * The new version of the directory entry. Since EXT4 structures are
1958 * stored in intel byte order, and the name_len field could never be
1959 * bigger than 255 chars, it's safe to reclaim the extra byte for the
1960 * file_type field.
1961 */
1962struct ext4_dir_entry_2 {
1963 __le32 inode; /* Inode number */
1964 __le16 rec_len; /* Directory entry length */
1965 __u8 name_len; /* Name length */
1966 __u8 file_type;
1967 char name[EXT4_NAME_LEN]; /* File name */
1968};
1969
1970/*
1971 * This is a bogus directory entry at the end of each leaf block that
1972 * records checksums.
1973 */
1974struct ext4_dir_entry_tail {
1975 __le32 det_reserved_zero1; /* Pretend to be unused */
1976 __le16 det_rec_len; /* 12 */
1977 __u8 det_reserved_zero2; /* Zero name length */
1978 __u8 det_reserved_ft; /* 0xDE, fake file type */
1979 __le32 det_checksum; /* crc32c(uuid+inum+dirblock) */
1980};
1981
1982#define EXT4_DIRENT_TAIL(block, blocksize) \
1983 ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
1984 ((blocksize) - \
1985 sizeof(struct ext4_dir_entry_tail))))
1986
1987/*
1988 * Ext4 directory file types. Only the low 3 bits are used. The
1989 * other bits are reserved for now.
1990 */
1991#define EXT4_FT_UNKNOWN 0
1992#define EXT4_FT_REG_FILE 1
1993#define EXT4_FT_DIR 2
1994#define EXT4_FT_CHRDEV 3
1995#define EXT4_FT_BLKDEV 4
1996#define EXT4_FT_FIFO 5
1997#define EXT4_FT_SOCK 6
1998#define EXT4_FT_SYMLINK 7
1999
2000#define EXT4_FT_MAX 8
2001
2002#define EXT4_FT_DIR_CSUM 0xDE
2003
2004/*
2005 * EXT4_DIR_PAD defines the directory entries boundaries
2006 *
2007 * NOTE: It must be a multiple of 4
2008 */
2009#define EXT4_DIR_PAD 4
2010#define EXT4_DIR_ROUND (EXT4_DIR_PAD - 1)
2011#define EXT4_DIR_REC_LEN(name_len) (((name_len) + 8 + EXT4_DIR_ROUND) & \
2012 ~EXT4_DIR_ROUND)
2013#define EXT4_MAX_REC_LEN ((1<<16)-1)
2014
2015/*
2016 * If we ever get support for fs block sizes > page_size, we'll need
2017 * to remove the #if statements in the next two functions...
2018 */
2019static inline unsigned int
2020ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
2021{
2022 unsigned len = le16_to_cpu(dlen);
2023
2024#if (PAGE_SIZE >= 65536)
2025 if (len == EXT4_MAX_REC_LEN || len == 0)
2026 return blocksize;
2027 return (len & 65532) | ((len & 3) << 16);
2028#else
2029 return len;
2030#endif
2031}
2032
2033static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
2034{
2035 if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
2036 BUG();
2037#if (PAGE_SIZE >= 65536)
2038 if (len < 65536)
2039 return cpu_to_le16(len);
2040 if (len == blocksize) {
2041 if (blocksize == 65536)
2042 return cpu_to_le16(EXT4_MAX_REC_LEN);
2043 else
2044 return cpu_to_le16(0);
2045 }
2046 return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
2047#else
2048 return cpu_to_le16(len);
2049#endif
2050}
2051
2052/*
2053 * Hash Tree Directory indexing
2054 * (c) Daniel Phillips, 2001
2055 */
2056
2057#define is_dx(dir) (ext4_has_feature_dir_index((dir)->i_sb) && \
2058 ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
2059#define EXT4_DIR_LINK_MAX(dir) unlikely((dir)->i_nlink >= EXT4_LINK_MAX && \
2060 !(ext4_has_feature_dir_nlink((dir)->i_sb) && is_dx(dir)))
2061#define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
2062
2063/* Legal values for the dx_root hash_version field: */
2064
2065#define DX_HASH_LEGACY 0
2066#define DX_HASH_HALF_MD4 1
2067#define DX_HASH_TEA 2
2068#define DX_HASH_LEGACY_UNSIGNED 3
2069#define DX_HASH_HALF_MD4_UNSIGNED 4
2070#define DX_HASH_TEA_UNSIGNED 5
2071
2072static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
2073 const void *address, unsigned int length)
2074{
2075 struct {
2076 struct shash_desc shash;
2077 char ctx[4];
2078 } desc;
2079
2080 BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
2081
2082 desc.shash.tfm = sbi->s_chksum_driver;
2083 *(u32 *)desc.ctx = crc;
2084
2085 BUG_ON(crypto_shash_update(&desc.shash, address, length));
2086
2087 return *(u32 *)desc.ctx;
2088}
2089
2090#ifdef __KERNEL__
2091
2092/* hash info structure used by the directory hash */
2093struct dx_hash_info
2094{
2095 u32 hash;
2096 u32 minor_hash;
2097 int hash_version;
2098 u32 *seed;
2099};
2100
2101
2102/* 32 and 64 bit signed EOF for dx directories */
2103#define EXT4_HTREE_EOF_32BIT ((1UL << (32 - 1)) - 1)
2104#define EXT4_HTREE_EOF_64BIT ((1ULL << (64 - 1)) - 1)
2105
2106
2107/*
2108 * Control parameters used by ext4_htree_next_block
2109 */
2110#define HASH_NB_ALWAYS 1
2111
2112struct ext4_filename {
2113 const struct qstr *usr_fname;
2114 struct fscrypt_str disk_name;
2115 struct dx_hash_info hinfo;
2116#ifdef CONFIG_FS_ENCRYPTION
2117 struct fscrypt_str crypto_buf;
2118#endif
2119#ifdef CONFIG_UNICODE
2120 struct fscrypt_str cf_name;
2121#endif
2122};
2123
2124#define fname_name(p) ((p)->disk_name.name)
2125#define fname_len(p) ((p)->disk_name.len)
2126
2127/*
2128 * Describe an inode's exact location on disk and in memory
2129 */
2130struct ext4_iloc
2131{
2132 struct buffer_head *bh;
2133 unsigned long offset;
2134 ext4_group_t block_group;
2135};
2136
2137static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
2138{
2139 return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
2140}
2141
2142static inline bool ext4_is_quota_file(struct inode *inode)
2143{
2144 return IS_NOQUOTA(inode) &&
2145 !(EXT4_I(inode)->i_flags & EXT4_EA_INODE_FL);
2146}
2147
2148/*
2149 * This structure is stuffed into the struct file's private_data field
2150 * for directories. It is where we put information so that we can do
2151 * readdir operations in hash tree order.
2152 */
2153struct dir_private_info {
2154 struct rb_root root;
2155 struct rb_node *curr_node;
2156 struct fname *extra_fname;
2157 loff_t last_pos;
2158 __u32 curr_hash;
2159 __u32 curr_minor_hash;
2160 __u32 next_hash;
2161};
2162
2163/* calculate the first block number of the group */
2164static inline ext4_fsblk_t
2165ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
2166{
2167 return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
2168 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
2169}
2170
2171/*
2172 * Special error return code only used by dx_probe() and its callers.
2173 */
2174#define ERR_BAD_DX_DIR (-(MAX_ERRNO - 1))
2175
2176/* htree levels for ext4 */
2177#define EXT4_HTREE_LEVEL_COMPAT 2
2178#define EXT4_HTREE_LEVEL 3
2179
2180static inline int ext4_dir_htree_level(struct super_block *sb)
2181{
2182 return ext4_has_feature_largedir(sb) ?
2183 EXT4_HTREE_LEVEL : EXT4_HTREE_LEVEL_COMPAT;
2184}
2185
2186/*
2187 * Timeout and state flag for lazy initialization inode thread.
2188 */
2189#define EXT4_DEF_LI_WAIT_MULT 10
2190#define EXT4_DEF_LI_MAX_START_DELAY 5
2191#define EXT4_LAZYINIT_QUIT 0x0001
2192#define EXT4_LAZYINIT_RUNNING 0x0002
2193
2194/*
2195 * Lazy inode table initialization info
2196 */
2197struct ext4_lazy_init {
2198 unsigned long li_state;
2199 struct list_head li_request_list;
2200 struct mutex li_list_mtx;
2201};
2202
2203struct ext4_li_request {
2204 struct super_block *lr_super;
2205 struct ext4_sb_info *lr_sbi;
2206 ext4_group_t lr_next_group;
2207 struct list_head lr_request;
2208 unsigned long lr_next_sched;
2209 unsigned long lr_timeout;
2210};
2211
2212struct ext4_features {
2213 struct kobject f_kobj;
2214 struct completion f_kobj_unregister;
2215};
2216
2217/*
2218 * This structure will be used for multiple mount protection. It will be
2219 * written into the block number saved in the s_mmp_block field in the
2220 * superblock. Programs that check MMP should assume that if
2221 * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
2222 * to use the filesystem, regardless of how old the timestamp is.
2223 */
2224#define EXT4_MMP_MAGIC 0x004D4D50U /* ASCII for MMP */
2225#define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
2226#define EXT4_MMP_SEQ_FSCK 0xE24D4D50U /* mmp_seq value when being fscked */
2227#define EXT4_MMP_SEQ_MAX 0xE24D4D4FU /* maximum valid mmp_seq value */
2228
2229struct mmp_struct {
2230 __le32 mmp_magic; /* Magic number for MMP */
2231 __le32 mmp_seq; /* Sequence no. updated periodically */
2232
2233 /*
2234 * mmp_time, mmp_nodename & mmp_bdevname are only used for information
2235 * purposes and do not affect the correctness of the algorithm
2236 */
2237 __le64 mmp_time; /* Time last updated */
2238 char mmp_nodename[64]; /* Node which last updated MMP block */
2239 char mmp_bdevname[32]; /* Bdev which last updated MMP block */
2240
2241 /*
2242 * mmp_check_interval is used to verify if the MMP block has been
2243 * updated on the block device. The value is updated based on the
2244 * maximum time to write the MMP block during an update cycle.
2245 */
2246 __le16 mmp_check_interval;
2247
2248 __le16 mmp_pad1;
2249 __le32 mmp_pad2[226];
2250 __le32 mmp_checksum; /* crc32c(uuid+mmp_block) */
2251};
2252
2253/* arguments passed to the mmp thread */
2254struct mmpd_data {
2255 struct buffer_head *bh; /* bh from initial read_mmp_block() */
2256 struct super_block *sb; /* super block of the fs */
2257};
2258
2259/*
2260 * Check interval multiplier
2261 * The MMP block is written every update interval and initially checked every
2262 * update interval x the multiplier (the value is then adapted based on the
2263 * write latency). The reason is that writes can be delayed under load and we
2264 * don't want readers to incorrectly assume that the filesystem is no longer
2265 * in use.
2266 */
2267#define EXT4_MMP_CHECK_MULT 2UL
2268
2269/*
2270 * Minimum interval for MMP checking in seconds.
2271 */
2272#define EXT4_MMP_MIN_CHECK_INTERVAL 5UL
2273
2274/*
2275 * Maximum interval for MMP checking in seconds.
2276 */
2277#define EXT4_MMP_MAX_CHECK_INTERVAL 300UL
2278
2279/*
2280 * Function prototypes
2281 */
2282
2283/*
2284 * Ok, these declarations are also in <linux/kernel.h> but none of the
2285 * ext4 source programs needs to include it so they are duplicated here.
2286 */
2287# define NORET_TYPE /**/
2288# define ATTRIB_NORET __attribute__((noreturn))
2289# define NORET_AND noreturn,
2290
2291/* bitmap.c */
2292extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
2293void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
2294 struct ext4_group_desc *gdp,
2295 struct buffer_head *bh, int sz);
2296int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2297 struct ext4_group_desc *gdp,
2298 struct buffer_head *bh, int sz);
2299void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
2300 struct ext4_group_desc *gdp,
2301 struct buffer_head *bh);
2302int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2303 struct ext4_group_desc *gdp,
2304 struct buffer_head *bh);
2305
2306/* balloc.c */
2307extern void ext4_get_group_no_and_offset(struct super_block *sb,
2308 ext4_fsblk_t blocknr,
2309 ext4_group_t *blockgrpp,
2310 ext4_grpblk_t *offsetp);
2311extern ext4_group_t ext4_get_group_number(struct super_block *sb,
2312 ext4_fsblk_t block);
2313
2314extern unsigned int ext4_block_group(struct super_block *sb,
2315 ext4_fsblk_t blocknr);
2316extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
2317 ext4_fsblk_t blocknr);
2318extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
2319extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
2320 ext4_group_t group);
2321extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
2322 ext4_fsblk_t goal,
2323 unsigned int flags,
2324 unsigned long *count,
2325 int *errp);
2326extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
2327 s64 nclusters, unsigned int flags);
2328extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
2329extern void ext4_check_blocks_bitmap(struct super_block *);
2330extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
2331 ext4_group_t block_group,
2332 struct buffer_head ** bh);
2333extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
2334
2335extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
2336 ext4_group_t block_group);
2337extern int ext4_wait_block_bitmap(struct super_block *sb,
2338 ext4_group_t block_group,
2339 struct buffer_head *bh);
2340extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
2341 ext4_group_t block_group);
2342extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
2343 ext4_group_t block_group,
2344 struct ext4_group_desc *gdp);
2345ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
2346
2347#ifdef CONFIG_UNICODE
2348extern void ext4_fname_setup_ci_filename(struct inode *dir,
2349 const struct qstr *iname,
2350 struct fscrypt_str *fname);
2351#endif
2352
2353#ifdef CONFIG_FS_ENCRYPTION
2354static inline void ext4_fname_from_fscrypt_name(struct ext4_filename *dst,
2355 const struct fscrypt_name *src)
2356{
2357 memset(dst, 0, sizeof(*dst));
2358
2359 dst->usr_fname = src->usr_fname;
2360 dst->disk_name = src->disk_name;
2361 dst->hinfo.hash = src->hash;
2362 dst->hinfo.minor_hash = src->minor_hash;
2363 dst->crypto_buf = src->crypto_buf;
2364}
2365
2366static inline int ext4_fname_setup_filename(struct inode *dir,
2367 const struct qstr *iname,
2368 int lookup,
2369 struct ext4_filename *fname)
2370{
2371 struct fscrypt_name name;
2372 int err;
2373
2374 err = fscrypt_setup_filename(dir, iname, lookup, &name);
2375 if (err)
2376 return err;
2377
2378 ext4_fname_from_fscrypt_name(fname, &name);
2379
2380#ifdef CONFIG_UNICODE
2381 ext4_fname_setup_ci_filename(dir, iname, &fname->cf_name);
2382#endif
2383 return 0;
2384}
2385
2386static inline int ext4_fname_prepare_lookup(struct inode *dir,
2387 struct dentry *dentry,
2388 struct ext4_filename *fname)
2389{
2390 struct fscrypt_name name;
2391 int err;
2392
2393 err = fscrypt_prepare_lookup(dir, dentry, &name);
2394 if (err)
2395 return err;
2396
2397 ext4_fname_from_fscrypt_name(fname, &name);
2398
2399#ifdef CONFIG_UNICODE
2400 ext4_fname_setup_ci_filename(dir, &dentry->d_name, &fname->cf_name);
2401#endif
2402 return 0;
2403}
2404
2405static inline void ext4_fname_free_filename(struct ext4_filename *fname)
2406{
2407 struct fscrypt_name name;
2408
2409 name.crypto_buf = fname->crypto_buf;
2410 fscrypt_free_filename(&name);
2411
2412 fname->crypto_buf.name = NULL;
2413 fname->usr_fname = NULL;
2414 fname->disk_name.name = NULL;
2415
2416#ifdef CONFIG_UNICODE
2417 kfree(fname->cf_name.name);
2418 fname->cf_name.name = NULL;
2419#endif
2420}
2421#else /* !CONFIG_FS_ENCRYPTION */
2422static inline int ext4_fname_setup_filename(struct inode *dir,
2423 const struct qstr *iname,
2424 int lookup,
2425 struct ext4_filename *fname)
2426{
2427 fname->usr_fname = iname;
2428 fname->disk_name.name = (unsigned char *) iname->name;
2429 fname->disk_name.len = iname->len;
2430
2431#ifdef CONFIG_UNICODE
2432 ext4_fname_setup_ci_filename(dir, iname, &fname->cf_name);
2433#endif
2434
2435 return 0;
2436}
2437
2438static inline int ext4_fname_prepare_lookup(struct inode *dir,
2439 struct dentry *dentry,
2440 struct ext4_filename *fname)
2441{
2442 return ext4_fname_setup_filename(dir, &dentry->d_name, 1, fname);
2443}
2444
2445static inline void ext4_fname_free_filename(struct ext4_filename *fname)
2446{
2447#ifdef CONFIG_UNICODE
2448 kfree(fname->cf_name.name);
2449 fname->cf_name.name = NULL;
2450#endif
2451}
2452#endif /* !CONFIG_FS_ENCRYPTION */
2453
2454/* dir.c */
2455extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
2456 struct file *,
2457 struct ext4_dir_entry_2 *,
2458 struct buffer_head *, char *, int,
2459 unsigned int);
2460#define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset) \
2461 unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
2462 (de), (bh), (buf), (size), (offset)))
2463extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
2464 __u32 minor_hash,
2465 struct ext4_dir_entry_2 *dirent,
2466 struct fscrypt_str *ent_name);
2467extern void ext4_htree_free_dir_info(struct dir_private_info *p);
2468extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
2469 struct buffer_head *bh,
2470 void *buf, int buf_size,
2471 struct ext4_filename *fname,
2472 struct ext4_dir_entry_2 **dest_de);
2473void ext4_insert_dentry(struct inode *inode,
2474 struct ext4_dir_entry_2 *de,
2475 int buf_size,
2476 struct ext4_filename *fname);
2477static inline void ext4_update_dx_flag(struct inode *inode)
2478{
2479 if (!ext4_has_feature_dir_index(inode->i_sb))
2480 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
2481}
2482static const unsigned char ext4_filetype_table[] = {
2483 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
2484};
2485
2486static inline unsigned char get_dtype(struct super_block *sb, int filetype)
2487{
2488 if (!ext4_has_feature_filetype(sb) || filetype >= EXT4_FT_MAX)
2489 return DT_UNKNOWN;
2490
2491 return ext4_filetype_table[filetype];
2492}
2493extern int ext4_check_all_de(struct inode *dir, struct buffer_head *bh,
2494 void *buf, int buf_size);
2495
2496/* fsync.c */
2497extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2498
2499/* hash.c */
2500extern int ext4fs_dirhash(const struct inode *dir, const char *name, int len,
2501 struct dx_hash_info *hinfo);
2502
2503/* ialloc.c */
2504extern struct inode *__ext4_new_inode(handle_t *, struct inode *, umode_t,
2505 const struct qstr *qstr, __u32 goal,
2506 uid_t *owner, __u32 i_flags,
2507 int handle_type, unsigned int line_no,
2508 int nblocks);
2509
2510#define ext4_new_inode(handle, dir, mode, qstr, goal, owner, i_flags) \
2511 __ext4_new_inode((handle), (dir), (mode), (qstr), (goal), (owner), \
2512 i_flags, 0, 0, 0)
2513#define ext4_new_inode_start_handle(dir, mode, qstr, goal, owner, \
2514 type, nblocks) \
2515 __ext4_new_inode(NULL, (dir), (mode), (qstr), (goal), (owner), \
2516 0, (type), __LINE__, (nblocks))
2517
2518
2519extern void ext4_free_inode(handle_t *, struct inode *);
2520extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2521extern unsigned long ext4_count_free_inodes(struct super_block *);
2522extern unsigned long ext4_count_dirs(struct super_block *);
2523extern void ext4_check_inodes_bitmap(struct super_block *);
2524extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2525extern int ext4_init_inode_table(struct super_block *sb,
2526 ext4_group_t group, int barrier);
2527extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2528
2529/* mballoc.c */
2530extern const struct seq_operations ext4_mb_seq_groups_ops;
2531extern long ext4_mb_stats;
2532extern long ext4_mb_max_to_scan;
2533extern int ext4_mb_init(struct super_block *);
2534extern int ext4_mb_release(struct super_block *);
2535extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2536 struct ext4_allocation_request *, int *);
2537extern int ext4_mb_reserve_blocks(struct super_block *, int);
2538extern void ext4_discard_preallocations(struct inode *);
2539extern int __init ext4_init_mballoc(void);
2540extern void ext4_exit_mballoc(void);
2541extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2542 struct buffer_head *bh, ext4_fsblk_t block,
2543 unsigned long count, int flags);
2544extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2545 ext4_group_t ngroups);
2546extern int ext4_mb_add_groupinfo(struct super_block *sb,
2547 ext4_group_t i, struct ext4_group_desc *desc);
2548extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2549 ext4_fsblk_t block, unsigned long count);
2550extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2551extern void ext4_process_freed_data(struct super_block *sb, tid_t commit_tid);
2552
2553/* inode.c */
2554int ext4_inode_is_fast_symlink(struct inode *inode);
2555struct buffer_head *ext4_getblk(handle_t *, struct inode *, ext4_lblk_t, int);
2556struct buffer_head *ext4_bread(handle_t *, struct inode *, ext4_lblk_t, int);
2557int ext4_bread_batch(struct inode *inode, ext4_lblk_t block, int bh_count,
2558 bool wait, struct buffer_head **bhs);
2559int ext4_get_block_unwritten(struct inode *inode, sector_t iblock,
2560 struct buffer_head *bh_result, int create);
2561int ext4_get_block(struct inode *inode, sector_t iblock,
2562 struct buffer_head *bh_result, int create);
2563int ext4_dio_get_block(struct inode *inode, sector_t iblock,
2564 struct buffer_head *bh_result, int create);
2565int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2566 struct buffer_head *bh, int create);
2567int ext4_walk_page_buffers(handle_t *handle,
2568 struct buffer_head *head,
2569 unsigned from,
2570 unsigned to,
2571 int *partial,
2572 int (*fn)(handle_t *handle,
2573 struct buffer_head *bh));
2574int do_journal_get_write_access(handle_t *handle,
2575 struct buffer_head *bh);
2576#define FALL_BACK_TO_NONDELALLOC 1
2577#define CONVERT_INLINE_DATA 2
2578
2579typedef enum {
2580 EXT4_IGET_NORMAL = 0,
2581 EXT4_IGET_SPECIAL = 0x0001, /* OK to iget a system inode */
2582 EXT4_IGET_HANDLE = 0x0002 /* Inode # is from a handle */
2583} ext4_iget_flags;
2584
2585extern struct inode *__ext4_iget(struct super_block *sb, unsigned long ino,
2586 ext4_iget_flags flags, const char *function,
2587 unsigned int line);
2588
2589#define ext4_iget(sb, ino, flags) \
2590 __ext4_iget((sb), (ino), (flags), __func__, __LINE__)
2591
2592extern int ext4_write_inode(struct inode *, struct writeback_control *);
2593extern int ext4_setattr(struct dentry *, struct iattr *);
2594extern int ext4_getattr(const struct path *, struct kstat *, u32, unsigned int);
2595extern void ext4_evict_inode(struct inode *);
2596extern void ext4_clear_inode(struct inode *);
2597extern int ext4_file_getattr(const struct path *, struct kstat *, u32, unsigned int);
2598extern int ext4_sync_inode(handle_t *, struct inode *);
2599extern void ext4_dirty_inode(struct inode *, int);
2600extern int ext4_change_inode_journal_flag(struct inode *, int);
2601extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
2602extern int ext4_inode_attach_jinode(struct inode *inode);
2603extern int ext4_can_truncate(struct inode *inode);
2604extern int ext4_truncate(struct inode *);
2605extern int ext4_break_layouts(struct inode *);
2606extern int ext4_punch_hole(struct inode *inode, loff_t offset, loff_t length);
2607extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
2608extern void ext4_set_inode_flags(struct inode *);
2609extern int ext4_alloc_da_blocks(struct inode *inode);
2610extern void ext4_set_aops(struct inode *inode);
2611extern int ext4_writepage_trans_blocks(struct inode *);
2612extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
2613extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
2614 loff_t lstart, loff_t lend);
2615extern vm_fault_t ext4_page_mkwrite(struct vm_fault *vmf);
2616extern vm_fault_t ext4_filemap_fault(struct vm_fault *vmf);
2617extern qsize_t *ext4_get_reserved_space(struct inode *inode);
2618extern int ext4_get_projid(struct inode *inode, kprojid_t *projid);
2619extern void ext4_da_release_space(struct inode *inode, int to_free);
2620extern void ext4_da_update_reserve_space(struct inode *inode,
2621 int used, int quota_claim);
2622extern int ext4_issue_zeroout(struct inode *inode, ext4_lblk_t lblk,
2623 ext4_fsblk_t pblk, ext4_lblk_t len);
2624
2625/* indirect.c */
2626extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
2627 struct ext4_map_blocks *map, int flags);
2628extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
2629extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
2630extern void ext4_ind_truncate(handle_t *, struct inode *inode);
2631extern int ext4_ind_remove_space(handle_t *handle, struct inode *inode,
2632 ext4_lblk_t start, ext4_lblk_t end);
2633
2634/* ioctl.c */
2635extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
2636extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
2637
2638/* migrate.c */
2639extern int ext4_ext_migrate(struct inode *);
2640extern int ext4_ind_migrate(struct inode *inode);
2641
2642/* namei.c */
2643extern int ext4_dirblock_csum_verify(struct inode *inode,
2644 struct buffer_head *bh);
2645extern int ext4_orphan_add(handle_t *, struct inode *);
2646extern int ext4_orphan_del(handle_t *, struct inode *);
2647extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
2648 __u32 start_minor_hash, __u32 *next_hash);
2649extern int ext4_search_dir(struct buffer_head *bh,
2650 char *search_buf,
2651 int buf_size,
2652 struct inode *dir,
2653 struct ext4_filename *fname,
2654 unsigned int offset,
2655 struct ext4_dir_entry_2 **res_dir);
2656extern int ext4_generic_delete_entry(handle_t *handle,
2657 struct inode *dir,
2658 struct ext4_dir_entry_2 *de_del,
2659 struct buffer_head *bh,
2660 void *entry_buf,
2661 int buf_size,
2662 int csum_size);
2663extern bool ext4_empty_dir(struct inode *inode);
2664
2665/* resize.c */
2666extern int ext4_group_add(struct super_block *sb,
2667 struct ext4_new_group_data *input);
2668extern int ext4_group_extend(struct super_block *sb,
2669 struct ext4_super_block *es,
2670 ext4_fsblk_t n_blocks_count);
2671extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
2672
2673/* super.c */
2674extern struct buffer_head *ext4_sb_bread(struct super_block *sb,
2675 sector_t block, int op_flags);
2676extern int ext4_seq_options_show(struct seq_file *seq, void *offset);
2677extern int ext4_calculate_overhead(struct super_block *sb);
2678extern void ext4_superblock_csum_set(struct super_block *sb);
2679extern void *ext4_kvmalloc(size_t size, gfp_t flags);
2680extern void *ext4_kvzalloc(size_t size, gfp_t flags);
2681extern int ext4_alloc_flex_bg_array(struct super_block *sb,
2682 ext4_group_t ngroup);
2683extern const char *ext4_decode_error(struct super_block *sb, int errno,
2684 char nbuf[16]);
2685extern void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
2686 ext4_group_t block_group,
2687 unsigned int flags);
2688
2689extern __printf(4, 5)
2690void __ext4_error(struct super_block *, const char *, unsigned int,
2691 const char *, ...);
2692extern __printf(5, 6)
2693void __ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
2694 const char *, ...);
2695extern __printf(5, 6)
2696void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
2697 const char *, ...);
2698extern void __ext4_std_error(struct super_block *, const char *,
2699 unsigned int, int);
2700extern __printf(4, 5)
2701void __ext4_abort(struct super_block *, const char *, unsigned int,
2702 const char *, ...);
2703extern __printf(4, 5)
2704void __ext4_warning(struct super_block *, const char *, unsigned int,
2705 const char *, ...);
2706extern __printf(4, 5)
2707void __ext4_warning_inode(const struct inode *inode, const char *function,
2708 unsigned int line, const char *fmt, ...);
2709extern __printf(3, 4)
2710void __ext4_msg(struct super_block *, const char *, const char *, ...);
2711extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
2712 const char *, unsigned int, const char *);
2713extern __printf(7, 8)
2714void __ext4_grp_locked_error(const char *, unsigned int,
2715 struct super_block *, ext4_group_t,
2716 unsigned long, ext4_fsblk_t,
2717 const char *, ...);
2718
2719#define EXT4_ERROR_INODE(inode, fmt, a...) \
2720 ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
2721
2722#define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...) \
2723 ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
2724
2725#define EXT4_ERROR_FILE(file, block, fmt, a...) \
2726 ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
2727
2728#ifdef CONFIG_PRINTK
2729
2730#define ext4_error_inode(inode, func, line, block, fmt, ...) \
2731 __ext4_error_inode(inode, func, line, block, fmt, ##__VA_ARGS__)
2732#define ext4_error_file(file, func, line, block, fmt, ...) \
2733 __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__)
2734#define ext4_error(sb, fmt, ...) \
2735 __ext4_error(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2736#define ext4_abort(sb, fmt, ...) \
2737 __ext4_abort(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2738#define ext4_warning(sb, fmt, ...) \
2739 __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2740#define ext4_warning_inode(inode, fmt, ...) \
2741 __ext4_warning_inode(inode, __func__, __LINE__, fmt, ##__VA_ARGS__)
2742#define ext4_msg(sb, level, fmt, ...) \
2743 __ext4_msg(sb, level, fmt, ##__VA_ARGS__)
2744#define dump_mmp_msg(sb, mmp, msg) \
2745 __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg)
2746#define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...) \
2747 __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \
2748 fmt, ##__VA_ARGS__)
2749
2750#else
2751
2752#define ext4_error_inode(inode, func, line, block, fmt, ...) \
2753do { \
2754 no_printk(fmt, ##__VA_ARGS__); \
2755 __ext4_error_inode(inode, "", 0, block, " "); \
2756} while (0)
2757#define ext4_error_file(file, func, line, block, fmt, ...) \
2758do { \
2759 no_printk(fmt, ##__VA_ARGS__); \
2760 __ext4_error_file(file, "", 0, block, " "); \
2761} while (0)
2762#define ext4_error(sb, fmt, ...) \
2763do { \
2764 no_printk(fmt, ##__VA_ARGS__); \
2765 __ext4_error(sb, "", 0, " "); \
2766} while (0)
2767#define ext4_abort(sb, fmt, ...) \
2768do { \
2769 no_printk(fmt, ##__VA_ARGS__); \
2770 __ext4_abort(sb, "", 0, " "); \
2771} while (0)
2772#define ext4_warning(sb, fmt, ...) \
2773do { \
2774 no_printk(fmt, ##__VA_ARGS__); \
2775 __ext4_warning(sb, "", 0, " "); \
2776} while (0)
2777#define ext4_warning_inode(inode, fmt, ...) \
2778do { \
2779 no_printk(fmt, ##__VA_ARGS__); \
2780 __ext4_warning_inode(inode, "", 0, " "); \
2781} while (0)
2782#define ext4_msg(sb, level, fmt, ...) \
2783do { \
2784 no_printk(fmt, ##__VA_ARGS__); \
2785 __ext4_msg(sb, "", " "); \
2786} while (0)
2787#define dump_mmp_msg(sb, mmp, msg) \
2788 __dump_mmp_msg(sb, mmp, "", 0, "")
2789#define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...) \
2790do { \
2791 no_printk(fmt, ##__VA_ARGS__); \
2792 __ext4_grp_locked_error("", 0, sb, grp, ino, block, " "); \
2793} while (0)
2794
2795#endif
2796
2797extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
2798 __u32 compat);
2799extern int ext4_update_rocompat_feature(handle_t *handle,
2800 struct super_block *sb, __u32 rocompat);
2801extern int ext4_update_incompat_feature(handle_t *handle,
2802 struct super_block *sb, __u32 incompat);
2803extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
2804 struct ext4_group_desc *bg);
2805extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
2806 struct ext4_group_desc *bg);
2807extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
2808 struct ext4_group_desc *bg);
2809extern __u32 ext4_free_group_clusters(struct super_block *sb,
2810 struct ext4_group_desc *bg);
2811extern __u32 ext4_free_inodes_count(struct super_block *sb,
2812 struct ext4_group_desc *bg);
2813extern __u32 ext4_used_dirs_count(struct super_block *sb,
2814 struct ext4_group_desc *bg);
2815extern __u32 ext4_itable_unused_count(struct super_block *sb,
2816 struct ext4_group_desc *bg);
2817extern void ext4_block_bitmap_set(struct super_block *sb,
2818 struct ext4_group_desc *bg, ext4_fsblk_t blk);
2819extern void ext4_inode_bitmap_set(struct super_block *sb,
2820 struct ext4_group_desc *bg, ext4_fsblk_t blk);
2821extern void ext4_inode_table_set(struct super_block *sb,
2822 struct ext4_group_desc *bg, ext4_fsblk_t blk);
2823extern void ext4_free_group_clusters_set(struct super_block *sb,
2824 struct ext4_group_desc *bg,
2825 __u32 count);
2826extern void ext4_free_inodes_set(struct super_block *sb,
2827 struct ext4_group_desc *bg, __u32 count);
2828extern void ext4_used_dirs_set(struct super_block *sb,
2829 struct ext4_group_desc *bg, __u32 count);
2830extern void ext4_itable_unused_set(struct super_block *sb,
2831 struct ext4_group_desc *bg, __u32 count);
2832extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
2833 struct ext4_group_desc *gdp);
2834extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
2835 struct ext4_group_desc *gdp);
2836extern int ext4_register_li_request(struct super_block *sb,
2837 ext4_group_t first_not_zeroed);
2838
2839static inline int ext4_has_metadata_csum(struct super_block *sb)
2840{
2841 WARN_ON_ONCE(ext4_has_feature_metadata_csum(sb) &&
2842 !EXT4_SB(sb)->s_chksum_driver);
2843
2844 return ext4_has_feature_metadata_csum(sb) &&
2845 (EXT4_SB(sb)->s_chksum_driver != NULL);
2846}
2847
2848static inline int ext4_has_group_desc_csum(struct super_block *sb)
2849{
2850 return ext4_has_feature_gdt_csum(sb) || ext4_has_metadata_csum(sb);
2851}
2852
2853static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2854{
2855 return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2856 le32_to_cpu(es->s_blocks_count_lo);
2857}
2858
2859static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2860{
2861 return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2862 le32_to_cpu(es->s_r_blocks_count_lo);
2863}
2864
2865static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2866{
2867 return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2868 le32_to_cpu(es->s_free_blocks_count_lo);
2869}
2870
2871static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2872 ext4_fsblk_t blk)
2873{
2874 es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2875 es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2876}
2877
2878static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2879 ext4_fsblk_t blk)
2880{
2881 es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2882 es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2883}
2884
2885static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2886 ext4_fsblk_t blk)
2887{
2888 es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2889 es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2890}
2891
2892static inline loff_t ext4_isize(struct super_block *sb,
2893 struct ext4_inode *raw_inode)
2894{
2895 if (ext4_has_feature_largedir(sb) ||
2896 S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2897 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2898 le32_to_cpu(raw_inode->i_size_lo);
2899
2900 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2901}
2902
2903static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2904{
2905 raw_inode->i_size_lo = cpu_to_le32(i_size);
2906 raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2907}
2908
2909static inline
2910struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2911 ext4_group_t group)
2912{
2913 struct ext4_group_info ***grp_info;
2914 long indexv, indexh;
2915 BUG_ON(group >= EXT4_SB(sb)->s_groups_count);
2916 grp_info = EXT4_SB(sb)->s_group_info;
2917 indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2918 indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2919 return grp_info[indexv][indexh];
2920}
2921
2922/*
2923 * Reading s_groups_count requires using smp_rmb() afterwards. See
2924 * the locking protocol documented in the comments of ext4_group_add()
2925 * in resize.c
2926 */
2927static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2928{
2929 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2930
2931 smp_rmb();
2932 return ngroups;
2933}
2934
2935static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2936 ext4_group_t block_group)
2937{
2938 return block_group >> sbi->s_log_groups_per_flex;
2939}
2940
2941static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2942{
2943 return 1 << sbi->s_log_groups_per_flex;
2944}
2945
2946#define ext4_std_error(sb, errno) \
2947do { \
2948 if ((errno)) \
2949 __ext4_std_error((sb), __func__, __LINE__, (errno)); \
2950} while (0)
2951
2952#ifdef CONFIG_SMP
2953/* Each CPU can accumulate percpu_counter_batch clusters in their local
2954 * counters. So we need to make sure we have free clusters more
2955 * than percpu_counter_batch * nr_cpu_ids. Also add a window of 4 times.
2956 */
2957#define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2958#else
2959#define EXT4_FREECLUSTERS_WATERMARK 0
2960#endif
2961
2962/* Update i_disksize. Requires i_mutex to avoid races with truncate */
2963static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2964{
2965 WARN_ON_ONCE(S_ISREG(inode->i_mode) &&
2966 !inode_is_locked(inode));
2967 down_write(&EXT4_I(inode)->i_data_sem);
2968 if (newsize > EXT4_I(inode)->i_disksize)
2969 EXT4_I(inode)->i_disksize = newsize;
2970 up_write(&EXT4_I(inode)->i_data_sem);
2971}
2972
2973/* Update i_size, i_disksize. Requires i_mutex to avoid races with truncate */
2974static inline int ext4_update_inode_size(struct inode *inode, loff_t newsize)
2975{
2976 int changed = 0;
2977
2978 if (newsize > inode->i_size) {
2979 i_size_write(inode, newsize);
2980 changed = 1;
2981 }
2982 if (newsize > EXT4_I(inode)->i_disksize) {
2983 ext4_update_i_disksize(inode, newsize);
2984 changed |= 2;
2985 }
2986 return changed;
2987}
2988
2989int ext4_update_disksize_before_punch(struct inode *inode, loff_t offset,
2990 loff_t len);
2991
2992struct ext4_group_info {
2993 unsigned long bb_state;
2994 struct rb_root bb_free_root;
2995 ext4_grpblk_t bb_first_free; /* first free block */
2996 ext4_grpblk_t bb_free; /* total free blocks */
2997 ext4_grpblk_t bb_fragments; /* nr of freespace fragments */
2998 ext4_grpblk_t bb_largest_free_order;/* order of largest frag in BG */
2999 struct list_head bb_prealloc_list;
3000#ifdef DOUBLE_CHECK
3001 void *bb_bitmap;
3002#endif
3003 struct rw_semaphore alloc_sem;
3004 ext4_grpblk_t bb_counters[]; /* Nr of free power-of-two-block
3005 * regions, index is order.
3006 * bb_counters[3] = 5 means
3007 * 5 free 8-block regions. */
3008};
3009
3010#define EXT4_GROUP_INFO_NEED_INIT_BIT 0
3011#define EXT4_GROUP_INFO_WAS_TRIMMED_BIT 1
3012#define EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT 2
3013#define EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT 3
3014#define EXT4_GROUP_INFO_BBITMAP_CORRUPT \
3015 (1 << EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT)
3016#define EXT4_GROUP_INFO_IBITMAP_CORRUPT \
3017 (1 << EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT)
3018
3019#define EXT4_MB_GRP_NEED_INIT(grp) \
3020 (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
3021#define EXT4_MB_GRP_BBITMAP_CORRUPT(grp) \
3022 (test_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
3023#define EXT4_MB_GRP_IBITMAP_CORRUPT(grp) \
3024 (test_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
3025
3026#define EXT4_MB_GRP_WAS_TRIMMED(grp) \
3027 (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3028#define EXT4_MB_GRP_SET_TRIMMED(grp) \
3029 (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3030#define EXT4_MB_GRP_CLEAR_TRIMMED(grp) \
3031 (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3032
3033#define EXT4_MAX_CONTENTION 8
3034#define EXT4_CONTENTION_THRESHOLD 2
3035
3036static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
3037 ext4_group_t group)
3038{
3039 return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
3040}
3041
3042/*
3043 * Returns true if the filesystem is busy enough that attempts to
3044 * access the block group locks has run into contention.
3045 */
3046static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
3047{
3048 return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
3049}
3050
3051static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
3052{
3053 spinlock_t *lock = ext4_group_lock_ptr(sb, group);
3054 if (spin_trylock(lock))
3055 /*
3056 * We're able to grab the lock right away, so drop the
3057 * lock contention counter.
3058 */
3059 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
3060 else {
3061 /*
3062 * The lock is busy, so bump the contention counter,
3063 * and then wait on the spin lock.
3064 */
3065 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
3066 EXT4_MAX_CONTENTION);
3067 spin_lock(lock);
3068 }
3069}
3070
3071static inline void ext4_unlock_group(struct super_block *sb,
3072 ext4_group_t group)
3073{
3074 spin_unlock(ext4_group_lock_ptr(sb, group));
3075}
3076
3077/*
3078 * Block validity checking
3079 */
3080#define ext4_check_indirect_blockref(inode, bh) \
3081 ext4_check_blockref(__func__, __LINE__, inode, \
3082 (__le32 *)(bh)->b_data, \
3083 EXT4_ADDR_PER_BLOCK((inode)->i_sb))
3084
3085#define ext4_ind_check_inode(inode) \
3086 ext4_check_blockref(__func__, __LINE__, inode, \
3087 EXT4_I(inode)->i_data, \
3088 EXT4_NDIR_BLOCKS)
3089
3090/*
3091 * Inodes and files operations
3092 */
3093
3094/* dir.c */
3095extern const struct file_operations ext4_dir_operations;
3096
3097#ifdef CONFIG_UNICODE
3098extern const struct dentry_operations ext4_dentry_ops;
3099#endif
3100
3101/* file.c */
3102extern const struct inode_operations ext4_file_inode_operations;
3103extern const struct file_operations ext4_file_operations;
3104extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
3105
3106/* inline.c */
3107extern int ext4_get_max_inline_size(struct inode *inode);
3108extern int ext4_find_inline_data_nolock(struct inode *inode);
3109extern int ext4_init_inline_data(handle_t *handle, struct inode *inode,
3110 unsigned int len);
3111extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
3112
3113extern int ext4_readpage_inline(struct inode *inode, struct page *page);
3114extern int ext4_try_to_write_inline_data(struct address_space *mapping,
3115 struct inode *inode,
3116 loff_t pos, unsigned len,
3117 unsigned flags,
3118 struct page **pagep);
3119extern int ext4_write_inline_data_end(struct inode *inode,
3120 loff_t pos, unsigned len,
3121 unsigned copied,
3122 struct page *page);
3123extern struct buffer_head *
3124ext4_journalled_write_inline_data(struct inode *inode,
3125 unsigned len,
3126 struct page *page);
3127extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
3128 struct inode *inode,
3129 loff_t pos, unsigned len,
3130 unsigned flags,
3131 struct page **pagep,
3132 void **fsdata);
3133extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
3134 unsigned len, unsigned copied,
3135 struct page *page);
3136extern int ext4_try_add_inline_entry(handle_t *handle,
3137 struct ext4_filename *fname,
3138 struct inode *dir, struct inode *inode);
3139extern int ext4_try_create_inline_dir(handle_t *handle,
3140 struct inode *parent,
3141 struct inode *inode);
3142extern int ext4_read_inline_dir(struct file *filp,
3143 struct dir_context *ctx,
3144 int *has_inline_data);
3145extern int ext4_inlinedir_to_tree(struct file *dir_file,
3146 struct inode *dir, ext4_lblk_t block,
3147 struct dx_hash_info *hinfo,
3148 __u32 start_hash, __u32 start_minor_hash,
3149 int *has_inline_data);
3150extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
3151 struct ext4_filename *fname,
3152 struct ext4_dir_entry_2 **res_dir,
3153 int *has_inline_data);
3154extern int ext4_delete_inline_entry(handle_t *handle,
3155 struct inode *dir,
3156 struct ext4_dir_entry_2 *de_del,
3157 struct buffer_head *bh,
3158 int *has_inline_data);
3159extern bool empty_inline_dir(struct inode *dir, int *has_inline_data);
3160extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
3161 struct ext4_dir_entry_2 **parent_de,
3162 int *retval);
3163extern int ext4_inline_data_fiemap(struct inode *inode,
3164 struct fiemap_extent_info *fieinfo,
3165 int *has_inline, __u64 start, __u64 len);
3166
3167struct iomap;
3168extern int ext4_inline_data_iomap(struct inode *inode, struct iomap *iomap);
3169
3170extern int ext4_inline_data_truncate(struct inode *inode, int *has_inline);
3171
3172extern int ext4_convert_inline_data(struct inode *inode);
3173
3174static inline int ext4_has_inline_data(struct inode *inode)
3175{
3176 return ext4_test_inode_flag(inode, EXT4_INODE_INLINE_DATA) &&
3177 EXT4_I(inode)->i_inline_off;
3178}
3179
3180/* namei.c */
3181extern const struct inode_operations ext4_dir_inode_operations;
3182extern const struct inode_operations ext4_special_inode_operations;
3183extern struct dentry *ext4_get_parent(struct dentry *child);
3184extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
3185 struct ext4_dir_entry_2 *de,
3186 int blocksize, int csum_size,
3187 unsigned int parent_ino, int dotdot_real_len);
3188extern void ext4_initialize_dirent_tail(struct buffer_head *bh,
3189 unsigned int blocksize);
3190extern int ext4_handle_dirty_dirblock(handle_t *handle, struct inode *inode,
3191 struct buffer_head *bh);
3192extern int ext4_ci_compare(const struct inode *parent,
3193 const struct qstr *fname,
3194 const struct qstr *entry, bool quick);
3195
3196#define S_SHIFT 12
3197static const unsigned char ext4_type_by_mode[(S_IFMT >> S_SHIFT) + 1] = {
3198 [S_IFREG >> S_SHIFT] = EXT4_FT_REG_FILE,
3199 [S_IFDIR >> S_SHIFT] = EXT4_FT_DIR,
3200 [S_IFCHR >> S_SHIFT] = EXT4_FT_CHRDEV,
3201 [S_IFBLK >> S_SHIFT] = EXT4_FT_BLKDEV,
3202 [S_IFIFO >> S_SHIFT] = EXT4_FT_FIFO,
3203 [S_IFSOCK >> S_SHIFT] = EXT4_FT_SOCK,
3204 [S_IFLNK >> S_SHIFT] = EXT4_FT_SYMLINK,
3205};
3206
3207static inline void ext4_set_de_type(struct super_block *sb,
3208 struct ext4_dir_entry_2 *de,
3209 umode_t mode) {
3210 if (ext4_has_feature_filetype(sb))
3211 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
3212}
3213
3214/* readpages.c */
3215extern int ext4_mpage_readpages(struct address_space *mapping,
3216 struct list_head *pages, struct page *page,
3217 unsigned nr_pages, bool is_readahead);
3218extern int __init ext4_init_post_read_processing(void);
3219extern void ext4_exit_post_read_processing(void);
3220
3221/* symlink.c */
3222extern const struct inode_operations ext4_encrypted_symlink_inode_operations;
3223extern const struct inode_operations ext4_symlink_inode_operations;
3224extern const struct inode_operations ext4_fast_symlink_inode_operations;
3225
3226/* sysfs.c */
3227extern int ext4_register_sysfs(struct super_block *sb);
3228extern void ext4_unregister_sysfs(struct super_block *sb);
3229extern int __init ext4_init_sysfs(void);
3230extern void ext4_exit_sysfs(void);
3231
3232/* block_validity */
3233extern void ext4_release_system_zone(struct super_block *sb);
3234extern int ext4_setup_system_zone(struct super_block *sb);
3235extern int __init ext4_init_system_zone(void);
3236extern void ext4_exit_system_zone(void);
3237extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
3238 ext4_fsblk_t start_blk,
3239 unsigned int count);
3240extern int ext4_check_blockref(const char *, unsigned int,
3241 struct inode *, __le32 *, unsigned int);
3242
3243/* extents.c */
3244struct ext4_ext_path;
3245struct ext4_extent;
3246
3247/*
3248 * Maximum number of logical blocks in a file; ext4_extent's ee_block is
3249 * __le32.
3250 */
3251#define EXT_MAX_BLOCKS 0xffffffff
3252
3253extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
3254extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
3255extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
3256extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
3257 struct ext4_map_blocks *map, int flags);
3258extern int ext4_ext_truncate(handle_t *, struct inode *);
3259extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
3260 ext4_lblk_t end);
3261extern void ext4_ext_init(struct super_block *);
3262extern void ext4_ext_release(struct super_block *);
3263extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
3264 loff_t len);
3265extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
3266 loff_t offset, ssize_t len);
3267extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
3268 struct ext4_map_blocks *map, int flags);
3269extern int ext4_ext_calc_metadata_amount(struct inode *inode,
3270 ext4_lblk_t lblocks);
3271extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
3272 int num,
3273 struct ext4_ext_path *path);
3274extern int ext4_can_extents_be_merged(struct inode *inode,
3275 struct ext4_extent *ex1,
3276 struct ext4_extent *ex2);
3277extern int ext4_ext_insert_extent(handle_t *, struct inode *,
3278 struct ext4_ext_path **,
3279 struct ext4_extent *, int);
3280extern struct ext4_ext_path *ext4_find_extent(struct inode *, ext4_lblk_t,
3281 struct ext4_ext_path **,
3282 int flags);
3283extern void ext4_ext_drop_refs(struct ext4_ext_path *);
3284extern int ext4_ext_check_inode(struct inode *inode);
3285extern ext4_lblk_t ext4_ext_next_allocated_block(struct ext4_ext_path *path);
3286extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
3287 __u64 start, __u64 len);
3288extern int ext4_get_es_cache(struct inode *inode,
3289 struct fiemap_extent_info *fieinfo,
3290 __u64 start, __u64 len);
3291extern int ext4_ext_precache(struct inode *inode);
3292extern int ext4_collapse_range(struct inode *inode, loff_t offset, loff_t len);
3293extern int ext4_insert_range(struct inode *inode, loff_t offset, loff_t len);
3294extern int ext4_swap_extents(handle_t *handle, struct inode *inode1,
3295 struct inode *inode2, ext4_lblk_t lblk1,
3296 ext4_lblk_t lblk2, ext4_lblk_t count,
3297 int mark_unwritten,int *err);
3298extern int ext4_clu_mapped(struct inode *inode, ext4_lblk_t lclu);
3299
3300/* move_extent.c */
3301extern void ext4_double_down_write_data_sem(struct inode *first,
3302 struct inode *second);
3303extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
3304 struct inode *donor_inode);
3305extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
3306 __u64 start_orig, __u64 start_donor,
3307 __u64 len, __u64 *moved_len);
3308
3309/* page-io.c */
3310extern int __init ext4_init_pageio(void);
3311extern void ext4_exit_pageio(void);
3312extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
3313extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
3314extern int ext4_put_io_end(ext4_io_end_t *io_end);
3315extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
3316extern void ext4_io_submit_init(struct ext4_io_submit *io,
3317 struct writeback_control *wbc);
3318extern void ext4_end_io_rsv_work(struct work_struct *work);
3319extern void ext4_io_submit(struct ext4_io_submit *io);
3320extern int ext4_bio_write_page(struct ext4_io_submit *io,
3321 struct page *page,
3322 int len,
3323 struct writeback_control *wbc,
3324 bool keep_towrite);
3325
3326/* mmp.c */
3327extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
3328
3329/* verity.c */
3330extern const struct fsverity_operations ext4_verityops;
3331
3332/*
3333 * Add new method to test whether block and inode bitmaps are properly
3334 * initialized. With uninit_bg reading the block from disk is not enough
3335 * to mark the bitmap uptodate. We need to also zero-out the bitmap
3336 */
3337#define BH_BITMAP_UPTODATE BH_JBDPrivateStart
3338
3339static inline int bitmap_uptodate(struct buffer_head *bh)
3340{
3341 return (buffer_uptodate(bh) &&
3342 test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
3343}
3344static inline void set_bitmap_uptodate(struct buffer_head *bh)
3345{
3346 set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
3347}
3348
3349#define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
3350
3351/* For ioend & aio unwritten conversion wait queues */
3352#define EXT4_WQ_HASH_SZ 37
3353#define ext4_ioend_wq(v) (&ext4__ioend_wq[((unsigned long)(v)) %\
3354 EXT4_WQ_HASH_SZ])
3355extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
3356
3357extern int ext4_resize_begin(struct super_block *sb);
3358extern void ext4_resize_end(struct super_block *sb);
3359
3360static inline void ext4_set_io_unwritten_flag(struct inode *inode,
3361 struct ext4_io_end *io_end)
3362{
3363 if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
3364 io_end->flag |= EXT4_IO_END_UNWRITTEN;
3365 atomic_inc(&EXT4_I(inode)->i_unwritten);
3366 }
3367}
3368
3369static inline void ext4_clear_io_unwritten_flag(ext4_io_end_t *io_end)
3370{
3371 struct inode *inode = io_end->inode;
3372
3373 if (io_end->flag & EXT4_IO_END_UNWRITTEN) {
3374 io_end->flag &= ~EXT4_IO_END_UNWRITTEN;
3375 /* Wake up anyone waiting on unwritten extent conversion */
3376 if (atomic_dec_and_test(&EXT4_I(inode)->i_unwritten))
3377 wake_up_all(ext4_ioend_wq(inode));
3378 }
3379}
3380
3381extern const struct iomap_ops ext4_iomap_ops;
3382
3383static inline int ext4_buffer_uptodate(struct buffer_head *bh)
3384{
3385 /*
3386 * If the buffer has the write error flag, we have failed
3387 * to write out data in the block. In this case, we don't
3388 * have to read the block because we may read the old data
3389 * successfully.
3390 */
3391 if (!buffer_uptodate(bh) && buffer_write_io_error(bh))
3392 set_buffer_uptodate(bh);
3393 return buffer_uptodate(bh);
3394}
3395
3396#endif /* __KERNEL__ */
3397
3398#define EFSBADCRC EBADMSG /* Bad CRC detected */
3399#define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
3400
3401#endif /* _EXT4_H */
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * ext4.h
4 *
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
9 *
10 * from
11 *
12 * linux/include/linux/minix_fs.h
13 *
14 * Copyright (C) 1991, 1992 Linus Torvalds
15 */
16
17#ifndef _EXT4_H
18#define _EXT4_H
19
20#include <linux/refcount.h>
21#include <linux/types.h>
22#include <linux/blkdev.h>
23#include <linux/magic.h>
24#include <linux/jbd2.h>
25#include <linux/quota.h>
26#include <linux/rwsem.h>
27#include <linux/rbtree.h>
28#include <linux/seqlock.h>
29#include <linux/mutex.h>
30#include <linux/timer.h>
31#include <linux/wait.h>
32#include <linux/sched/signal.h>
33#include <linux/blockgroup_lock.h>
34#include <linux/percpu_counter.h>
35#include <linux/ratelimit.h>
36#include <crypto/hash.h>
37#include <linux/falloc.h>
38#include <linux/percpu-rwsem.h>
39#include <linux/fiemap.h>
40#ifdef __KERNEL__
41#include <linux/compat.h>
42#endif
43#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) \
104do { \
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 */
114typedef int ext4_grpblk_t;
115
116/* data type for filesystem-wide blocks number */
117typedef unsigned long long ext4_fsblk_t;
118
119/* data type for file logical block number */
120typedef __u32 ext4_lblk_t;
121
122/* data type for block group number */
123typedef unsigned int ext4_group_t;
124
125enum 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 */
137enum 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
226struct 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
263struct 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 */
273struct 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
283struct 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 */
293typedef 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
305struct 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 */
395struct 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
433struct 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. */
560static 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 */
573enum {
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
622static 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)
653struct 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 */
665struct 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 */
677enum {
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 */
771struct 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
873static 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
879static 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) \
891do { \
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) \
922do { \
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) \
928do { \
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) \
934do { \
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) \
940do { \
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 */
996enum {
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 */
1007struct 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 atomic_t i_unwritten; /* Nr. of inflight conversions pending */
1062
1063 /* Fast commit wait queue for this inode */
1064 wait_queue_head_t i_fc_wait;
1065
1066 /* Protect concurrent accesses on i_fc_lblk_start, i_fc_lblk_len */
1067 struct mutex i_fc_lock;
1068
1069 /*
1070 * i_disksize keeps track of what the inode size is ON DISK, not
1071 * in memory. During truncate, i_size is set to the new size by
1072 * the VFS prior to calling ext4_truncate(), but the filesystem won't
1073 * set i_disksize to 0 until the truncate is actually under way.
1074 *
1075 * The intent is that i_disksize always represents the blocks which
1076 * are used by this file. This allows recovery to restart truncate
1077 * on orphans if we crash during truncate. We actually write i_disksize
1078 * into the on-disk inode when writing inodes out, instead of i_size.
1079 *
1080 * The only time when i_disksize and i_size may be different is when
1081 * a truncate is in progress. The only things which change i_disksize
1082 * are ext4_get_block (growth) and ext4_truncate (shrinkth).
1083 */
1084 loff_t i_disksize;
1085
1086 /*
1087 * i_data_sem is for serialising ext4_truncate() against
1088 * ext4_getblock(). In the 2.4 ext2 design, great chunks of inode's
1089 * data tree are chopped off during truncate. We can't do that in
1090 * ext4 because whenever we perform intermediate commits during
1091 * truncate, the inode and all the metadata blocks *must* be in a
1092 * consistent state which allows truncation of the orphans to restart
1093 * during recovery. Hence we must fix the get_block-vs-truncate race
1094 * by other means, so we have i_data_sem.
1095 */
1096 struct rw_semaphore i_data_sem;
1097 struct inode vfs_inode;
1098 struct jbd2_inode *jinode;
1099
1100 spinlock_t i_raw_lock; /* protects updates to the raw inode */
1101
1102 /*
1103 * File creation time. Its function is same as that of
1104 * struct timespec64 i_{a,c,m}time in the generic inode.
1105 */
1106 struct timespec64 i_crtime;
1107
1108 /* mballoc */
1109 atomic_t i_prealloc_active;
1110
1111 /* allocation reservation info for delalloc */
1112 /* In case of bigalloc, this refer to clusters rather than blocks */
1113 unsigned int i_reserved_data_blocks;
1114 struct rb_root i_prealloc_node;
1115 rwlock_t i_prealloc_lock;
1116
1117 /* extents status tree */
1118 struct ext4_es_tree i_es_tree;
1119 rwlock_t i_es_lock;
1120 struct list_head i_es_list;
1121 unsigned int i_es_all_nr; /* protected by i_es_lock */
1122 unsigned int i_es_shk_nr; /* protected by i_es_lock */
1123 ext4_lblk_t i_es_shrink_lblk; /* Offset where we start searching for
1124 extents to shrink. Protected by
1125 i_es_lock */
1126
1127 /* ialloc */
1128 ext4_group_t i_last_alloc_group;
1129
1130 /* pending cluster reservations for bigalloc file systems */
1131 struct ext4_pending_tree i_pending_tree;
1132
1133 /* on-disk additional length */
1134 __u16 i_extra_isize;
1135
1136 /* Indicate the inline data space. */
1137 u16 i_inline_off;
1138 u16 i_inline_size;
1139
1140#ifdef CONFIG_QUOTA
1141 /* quota space reservation, managed internally by quota code */
1142 qsize_t i_reserved_quota;
1143#endif
1144
1145 /* Lock protecting lists below */
1146 spinlock_t i_completed_io_lock;
1147 /*
1148 * Completed IOs that need unwritten extents handling and have
1149 * transaction reserved
1150 */
1151 struct list_head i_rsv_conversion_list;
1152 struct work_struct i_rsv_conversion_work;
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
1280extern 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 */
1304struct 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 */
1445enum 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
1452struct ext4_journal_trigger {
1453 struct jbd2_buffer_trigger_type tr_triggers;
1454 struct super_block *sb;
1455};
1456
1457static 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 */
1466struct ext4_orphan_block_tail {
1467 __le32 ob_magic;
1468 __le32 ob_checksum;
1469};
1470
1471static 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
1477struct 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 */
1485struct 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 */
1495struct 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 /* Atomic write unit values in bytes */
1733 unsigned int s_awu_min;
1734 unsigned int s_awu_max;
1735
1736 /* Ext4 fast commit sub transaction ID */
1737 atomic_t s_fc_subtid;
1738
1739 /*
1740 * After commit starts, the main queue gets locked, and the further
1741 * updates get added in the staging queue.
1742 */
1743#define FC_Q_MAIN 0
1744#define FC_Q_STAGING 1
1745 struct list_head s_fc_q[2]; /* Inodes staged for fast commit
1746 * that have data changes in them.
1747 */
1748 struct list_head s_fc_dentry_q[2]; /* directory entry updates */
1749 unsigned int s_fc_bytes;
1750 /*
1751 * Main fast commit lock. This lock protects accesses to the
1752 * following fields:
1753 * ei->i_fc_list, s_fc_dentry_q, s_fc_q, s_fc_bytes, s_fc_bh.
1754 */
1755 spinlock_t s_fc_lock;
1756 struct buffer_head *s_fc_bh;
1757 struct ext4_fc_stats s_fc_stats;
1758 tid_t s_fc_ineligible_tid;
1759#ifdef CONFIG_EXT4_DEBUG
1760 int s_fc_debug_max_replay;
1761#endif
1762 struct ext4_fc_replay_state s_fc_replay_state;
1763};
1764
1765static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1766{
1767 return sb->s_fs_info;
1768}
1769static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1770{
1771 return container_of(inode, struct ext4_inode_info, vfs_inode);
1772}
1773
1774static inline int ext4_writepages_down_read(struct super_block *sb)
1775{
1776 percpu_down_read(&EXT4_SB(sb)->s_writepages_rwsem);
1777 return memalloc_nofs_save();
1778}
1779
1780static inline void ext4_writepages_up_read(struct super_block *sb, int ctx)
1781{
1782 memalloc_nofs_restore(ctx);
1783 percpu_up_read(&EXT4_SB(sb)->s_writepages_rwsem);
1784}
1785
1786static inline int ext4_writepages_down_write(struct super_block *sb)
1787{
1788 percpu_down_write(&EXT4_SB(sb)->s_writepages_rwsem);
1789 return memalloc_nofs_save();
1790}
1791
1792static inline void ext4_writepages_up_write(struct super_block *sb, int ctx)
1793{
1794 memalloc_nofs_restore(ctx);
1795 percpu_up_write(&EXT4_SB(sb)->s_writepages_rwsem);
1796}
1797
1798static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1799{
1800 return ino == EXT4_ROOT_INO ||
1801 (ino >= EXT4_FIRST_INO(sb) &&
1802 ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1803}
1804
1805/*
1806 * Returns: sbi->field[index]
1807 * Used to access an array element from the following sbi fields which require
1808 * rcu protection to avoid dereferencing an invalid pointer due to reassignment
1809 * - s_group_desc
1810 * - s_group_info
1811 * - s_flex_group
1812 */
1813#define sbi_array_rcu_deref(sbi, field, index) \
1814({ \
1815 typeof(*((sbi)->field)) _v; \
1816 rcu_read_lock(); \
1817 _v = ((typeof(_v)*)rcu_dereference((sbi)->field))[index]; \
1818 rcu_read_unlock(); \
1819 _v; \
1820})
1821
1822/*
1823 * run-time mount flags
1824 */
1825enum {
1826 EXT4_MF_MNTDIR_SAMPLED,
1827 EXT4_MF_FC_INELIGIBLE /* Fast commit ineligible */
1828};
1829
1830static inline void ext4_set_mount_flag(struct super_block *sb, int bit)
1831{
1832 set_bit(bit, &EXT4_SB(sb)->s_mount_flags);
1833}
1834
1835static inline void ext4_clear_mount_flag(struct super_block *sb, int bit)
1836{
1837 clear_bit(bit, &EXT4_SB(sb)->s_mount_flags);
1838}
1839
1840static inline int ext4_test_mount_flag(struct super_block *sb, int bit)
1841{
1842 return test_bit(bit, &EXT4_SB(sb)->s_mount_flags);
1843}
1844
1845
1846/*
1847 * Simulate_fail codes
1848 */
1849#define EXT4_SIM_BBITMAP_EIO 1
1850#define EXT4_SIM_BBITMAP_CRC 2
1851#define EXT4_SIM_IBITMAP_EIO 3
1852#define EXT4_SIM_IBITMAP_CRC 4
1853#define EXT4_SIM_INODE_EIO 5
1854#define EXT4_SIM_INODE_CRC 6
1855#define EXT4_SIM_DIRBLOCK_EIO 7
1856#define EXT4_SIM_DIRBLOCK_CRC 8
1857
1858static inline bool ext4_simulate_fail(struct super_block *sb,
1859 unsigned long code)
1860{
1861#ifdef CONFIG_EXT4_DEBUG
1862 struct ext4_sb_info *sbi = EXT4_SB(sb);
1863
1864 if (unlikely(sbi->s_simulate_fail == code)) {
1865 sbi->s_simulate_fail = 0;
1866 return true;
1867 }
1868#endif
1869 return false;
1870}
1871
1872/*
1873 * Error number codes for s_{first,last}_error_errno
1874 *
1875 * Linux errno numbers are architecture specific, so we need to translate
1876 * them into something which is architecture independent. We don't define
1877 * codes for all errno's; just the ones which are most likely to be the cause
1878 * of an ext4_error() call.
1879 */
1880#define EXT4_ERR_UNKNOWN 1
1881#define EXT4_ERR_EIO 2
1882#define EXT4_ERR_ENOMEM 3
1883#define EXT4_ERR_EFSBADCRC 4
1884#define EXT4_ERR_EFSCORRUPTED 5
1885#define EXT4_ERR_ENOSPC 6
1886#define EXT4_ERR_ENOKEY 7
1887#define EXT4_ERR_EROFS 8
1888#define EXT4_ERR_EFBIG 9
1889#define EXT4_ERR_EEXIST 10
1890#define EXT4_ERR_ERANGE 11
1891#define EXT4_ERR_EOVERFLOW 12
1892#define EXT4_ERR_EBUSY 13
1893#define EXT4_ERR_ENOTDIR 14
1894#define EXT4_ERR_ENOTEMPTY 15
1895#define EXT4_ERR_ESHUTDOWN 16
1896#define EXT4_ERR_EFAULT 17
1897
1898/*
1899 * Inode dynamic state flags
1900 */
1901enum {
1902 EXT4_STATE_NEW, /* inode is newly created */
1903 EXT4_STATE_XATTR, /* has in-inode xattrs */
1904 EXT4_STATE_NO_EXPAND, /* No space for expansion */
1905 EXT4_STATE_DA_ALLOC_CLOSE, /* Alloc DA blks on close */
1906 EXT4_STATE_EXT_MIGRATE, /* Inode is migrating */
1907 EXT4_STATE_NEWENTRY, /* File just added to dir */
1908 EXT4_STATE_MAY_INLINE_DATA, /* may have in-inode data */
1909 EXT4_STATE_EXT_PRECACHED, /* extents have been precached */
1910 EXT4_STATE_LUSTRE_EA_INODE, /* Lustre-style ea_inode */
1911 EXT4_STATE_VERITY_IN_PROGRESS, /* building fs-verity Merkle tree */
1912 EXT4_STATE_FC_COMMITTING, /* Fast commit ongoing */
1913 EXT4_STATE_ORPHAN_FILE, /* Inode orphaned in orphan file */
1914};
1915
1916#define EXT4_INODE_BIT_FNS(name, field, offset) \
1917static inline int ext4_test_inode_##name(struct inode *inode, int bit) \
1918{ \
1919 return test_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1920} \
1921static inline void ext4_set_inode_##name(struct inode *inode, int bit) \
1922{ \
1923 set_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1924} \
1925static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1926{ \
1927 clear_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1928}
1929
1930/* Add these declarations here only so that these functions can be
1931 * found by name. Otherwise, they are very hard to locate. */
1932static inline int ext4_test_inode_flag(struct inode *inode, int bit);
1933static inline void ext4_set_inode_flag(struct inode *inode, int bit);
1934static inline void ext4_clear_inode_flag(struct inode *inode, int bit);
1935EXT4_INODE_BIT_FNS(flag, flags, 0)
1936
1937/* Add these declarations here only so that these functions can be
1938 * found by name. Otherwise, they are very hard to locate. */
1939static inline int ext4_test_inode_state(struct inode *inode, int bit);
1940static inline void ext4_set_inode_state(struct inode *inode, int bit);
1941static inline void ext4_clear_inode_state(struct inode *inode, int bit);
1942#if (BITS_PER_LONG < 64)
1943EXT4_INODE_BIT_FNS(state, state_flags, 0)
1944
1945static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1946{
1947 (ei)->i_state_flags = 0;
1948}
1949#else
1950EXT4_INODE_BIT_FNS(state, flags, 32)
1951
1952static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1953{
1954 /* We depend on the fact that callers will set i_flags */
1955}
1956#endif
1957#else
1958/* Assume that user mode programs are passing in an ext4fs superblock, not
1959 * a kernel struct super_block. This will allow us to call the feature-test
1960 * macros from user land. */
1961#define EXT4_SB(sb) (sb)
1962#endif
1963
1964static inline bool ext4_verity_in_progress(struct inode *inode)
1965{
1966 return IS_ENABLED(CONFIG_FS_VERITY) &&
1967 ext4_test_inode_state(inode, EXT4_STATE_VERITY_IN_PROGRESS);
1968}
1969
1970#define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1971
1972/*
1973 * Codes for operating systems
1974 */
1975#define EXT4_OS_LINUX 0
1976#define EXT4_OS_HURD 1
1977#define EXT4_OS_MASIX 2
1978#define EXT4_OS_FREEBSD 3
1979#define EXT4_OS_LITES 4
1980
1981/*
1982 * Revision levels
1983 */
1984#define EXT4_GOOD_OLD_REV 0 /* The good old (original) format */
1985#define EXT4_DYNAMIC_REV 1 /* V2 format w/ dynamic inode sizes */
1986
1987#define EXT4_MAX_SUPP_REV EXT4_DYNAMIC_REV
1988
1989#define EXT4_GOOD_OLD_INODE_SIZE 128
1990
1991#define EXT4_EXTRA_TIMESTAMP_MAX (((s64)1 << 34) - 1 + S32_MIN)
1992#define EXT4_NON_EXTRA_TIMESTAMP_MAX S32_MAX
1993#define EXT4_TIMESTAMP_MIN S32_MIN
1994
1995/*
1996 * Feature set definitions
1997 */
1998
1999#define EXT4_FEATURE_COMPAT_DIR_PREALLOC 0x0001
2000#define EXT4_FEATURE_COMPAT_IMAGIC_INODES 0x0002
2001#define EXT4_FEATURE_COMPAT_HAS_JOURNAL 0x0004
2002#define EXT4_FEATURE_COMPAT_EXT_ATTR 0x0008
2003#define EXT4_FEATURE_COMPAT_RESIZE_INODE 0x0010
2004#define EXT4_FEATURE_COMPAT_DIR_INDEX 0x0020
2005#define EXT4_FEATURE_COMPAT_SPARSE_SUPER2 0x0200
2006/*
2007 * The reason why "FAST_COMMIT" is a compat feature is that, FS becomes
2008 * incompatible only if fast commit blocks are present in the FS. Since we
2009 * clear the journal (and thus the fast commit blocks), we don't mark FS as
2010 * incompatible. We also have a JBD2 incompat feature, which gets set when
2011 * there are fast commit blocks present in the journal.
2012 */
2013#define EXT4_FEATURE_COMPAT_FAST_COMMIT 0x0400
2014#define EXT4_FEATURE_COMPAT_STABLE_INODES 0x0800
2015#define EXT4_FEATURE_COMPAT_ORPHAN_FILE 0x1000 /* Orphan file exists */
2016
2017#define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER 0x0001
2018#define EXT4_FEATURE_RO_COMPAT_LARGE_FILE 0x0002
2019#define EXT4_FEATURE_RO_COMPAT_BTREE_DIR 0x0004
2020#define EXT4_FEATURE_RO_COMPAT_HUGE_FILE 0x0008
2021#define EXT4_FEATURE_RO_COMPAT_GDT_CSUM 0x0010
2022#define EXT4_FEATURE_RO_COMPAT_DIR_NLINK 0x0020
2023#define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE 0x0040
2024#define EXT4_FEATURE_RO_COMPAT_QUOTA 0x0100
2025#define EXT4_FEATURE_RO_COMPAT_BIGALLOC 0x0200
2026/*
2027 * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM). When
2028 * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
2029 * all other data structures' checksums. However, the METADATA_CSUM and
2030 * GDT_CSUM bits are mutually exclusive.
2031 */
2032#define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM 0x0400
2033#define EXT4_FEATURE_RO_COMPAT_READONLY 0x1000
2034#define EXT4_FEATURE_RO_COMPAT_PROJECT 0x2000
2035#define EXT4_FEATURE_RO_COMPAT_VERITY 0x8000
2036#define EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT 0x10000 /* Orphan file may be
2037 non-empty */
2038
2039#define EXT4_FEATURE_INCOMPAT_COMPRESSION 0x0001
2040#define EXT4_FEATURE_INCOMPAT_FILETYPE 0x0002
2041#define EXT4_FEATURE_INCOMPAT_RECOVER 0x0004 /* Needs recovery */
2042#define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 /* Journal device */
2043#define EXT4_FEATURE_INCOMPAT_META_BG 0x0010
2044#define EXT4_FEATURE_INCOMPAT_EXTENTS 0x0040 /* extents support */
2045#define EXT4_FEATURE_INCOMPAT_64BIT 0x0080
2046#define EXT4_FEATURE_INCOMPAT_MMP 0x0100
2047#define EXT4_FEATURE_INCOMPAT_FLEX_BG 0x0200
2048#define EXT4_FEATURE_INCOMPAT_EA_INODE 0x0400 /* EA in inode */
2049#define EXT4_FEATURE_INCOMPAT_DIRDATA 0x1000 /* data in dirent */
2050#define EXT4_FEATURE_INCOMPAT_CSUM_SEED 0x2000
2051#define EXT4_FEATURE_INCOMPAT_LARGEDIR 0x4000 /* >2GB or 3-lvl htree */
2052#define EXT4_FEATURE_INCOMPAT_INLINE_DATA 0x8000 /* data in inode */
2053#define EXT4_FEATURE_INCOMPAT_ENCRYPT 0x10000
2054#define EXT4_FEATURE_INCOMPAT_CASEFOLD 0x20000
2055
2056extern void ext4_update_dynamic_rev(struct super_block *sb);
2057
2058#define EXT4_FEATURE_COMPAT_FUNCS(name, flagname) \
2059static inline bool ext4_has_feature_##name(struct super_block *sb) \
2060{ \
2061 return ((EXT4_SB(sb)->s_es->s_feature_compat & \
2062 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname)) != 0); \
2063} \
2064static inline void ext4_set_feature_##name(struct super_block *sb) \
2065{ \
2066 ext4_update_dynamic_rev(sb); \
2067 EXT4_SB(sb)->s_es->s_feature_compat |= \
2068 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
2069} \
2070static inline void ext4_clear_feature_##name(struct super_block *sb) \
2071{ \
2072 EXT4_SB(sb)->s_es->s_feature_compat &= \
2073 ~cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
2074}
2075
2076#define EXT4_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
2077static inline bool ext4_has_feature_##name(struct super_block *sb) \
2078{ \
2079 return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
2080 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname)) != 0); \
2081} \
2082static inline void ext4_set_feature_##name(struct super_block *sb) \
2083{ \
2084 ext4_update_dynamic_rev(sb); \
2085 EXT4_SB(sb)->s_es->s_feature_ro_compat |= \
2086 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
2087} \
2088static inline void ext4_clear_feature_##name(struct super_block *sb) \
2089{ \
2090 EXT4_SB(sb)->s_es->s_feature_ro_compat &= \
2091 ~cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
2092}
2093
2094#define EXT4_FEATURE_INCOMPAT_FUNCS(name, flagname) \
2095static inline bool ext4_has_feature_##name(struct super_block *sb) \
2096{ \
2097 return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
2098 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname)) != 0); \
2099} \
2100static inline void ext4_set_feature_##name(struct super_block *sb) \
2101{ \
2102 ext4_update_dynamic_rev(sb); \
2103 EXT4_SB(sb)->s_es->s_feature_incompat |= \
2104 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
2105} \
2106static inline void ext4_clear_feature_##name(struct super_block *sb) \
2107{ \
2108 EXT4_SB(sb)->s_es->s_feature_incompat &= \
2109 ~cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
2110}
2111
2112EXT4_FEATURE_COMPAT_FUNCS(dir_prealloc, DIR_PREALLOC)
2113EXT4_FEATURE_COMPAT_FUNCS(imagic_inodes, IMAGIC_INODES)
2114EXT4_FEATURE_COMPAT_FUNCS(journal, HAS_JOURNAL)
2115EXT4_FEATURE_COMPAT_FUNCS(xattr, EXT_ATTR)
2116EXT4_FEATURE_COMPAT_FUNCS(resize_inode, RESIZE_INODE)
2117EXT4_FEATURE_COMPAT_FUNCS(dir_index, DIR_INDEX)
2118EXT4_FEATURE_COMPAT_FUNCS(sparse_super2, SPARSE_SUPER2)
2119EXT4_FEATURE_COMPAT_FUNCS(fast_commit, FAST_COMMIT)
2120EXT4_FEATURE_COMPAT_FUNCS(stable_inodes, STABLE_INODES)
2121EXT4_FEATURE_COMPAT_FUNCS(orphan_file, ORPHAN_FILE)
2122
2123EXT4_FEATURE_RO_COMPAT_FUNCS(sparse_super, SPARSE_SUPER)
2124EXT4_FEATURE_RO_COMPAT_FUNCS(large_file, LARGE_FILE)
2125EXT4_FEATURE_RO_COMPAT_FUNCS(btree_dir, BTREE_DIR)
2126EXT4_FEATURE_RO_COMPAT_FUNCS(huge_file, HUGE_FILE)
2127EXT4_FEATURE_RO_COMPAT_FUNCS(gdt_csum, GDT_CSUM)
2128EXT4_FEATURE_RO_COMPAT_FUNCS(dir_nlink, DIR_NLINK)
2129EXT4_FEATURE_RO_COMPAT_FUNCS(extra_isize, EXTRA_ISIZE)
2130EXT4_FEATURE_RO_COMPAT_FUNCS(quota, QUOTA)
2131EXT4_FEATURE_RO_COMPAT_FUNCS(bigalloc, BIGALLOC)
2132EXT4_FEATURE_RO_COMPAT_FUNCS(metadata_csum, METADATA_CSUM)
2133EXT4_FEATURE_RO_COMPAT_FUNCS(readonly, READONLY)
2134EXT4_FEATURE_RO_COMPAT_FUNCS(project, PROJECT)
2135EXT4_FEATURE_RO_COMPAT_FUNCS(verity, VERITY)
2136EXT4_FEATURE_RO_COMPAT_FUNCS(orphan_present, ORPHAN_PRESENT)
2137
2138EXT4_FEATURE_INCOMPAT_FUNCS(compression, COMPRESSION)
2139EXT4_FEATURE_INCOMPAT_FUNCS(filetype, FILETYPE)
2140EXT4_FEATURE_INCOMPAT_FUNCS(journal_needs_recovery, RECOVER)
2141EXT4_FEATURE_INCOMPAT_FUNCS(journal_dev, JOURNAL_DEV)
2142EXT4_FEATURE_INCOMPAT_FUNCS(meta_bg, META_BG)
2143EXT4_FEATURE_INCOMPAT_FUNCS(extents, EXTENTS)
2144EXT4_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT)
2145EXT4_FEATURE_INCOMPAT_FUNCS(mmp, MMP)
2146EXT4_FEATURE_INCOMPAT_FUNCS(flex_bg, FLEX_BG)
2147EXT4_FEATURE_INCOMPAT_FUNCS(ea_inode, EA_INODE)
2148EXT4_FEATURE_INCOMPAT_FUNCS(dirdata, DIRDATA)
2149EXT4_FEATURE_INCOMPAT_FUNCS(csum_seed, CSUM_SEED)
2150EXT4_FEATURE_INCOMPAT_FUNCS(largedir, LARGEDIR)
2151EXT4_FEATURE_INCOMPAT_FUNCS(inline_data, INLINE_DATA)
2152EXT4_FEATURE_INCOMPAT_FUNCS(encrypt, ENCRYPT)
2153EXT4_FEATURE_INCOMPAT_FUNCS(casefold, CASEFOLD)
2154
2155#define EXT2_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR
2156#define EXT2_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
2157 EXT4_FEATURE_INCOMPAT_META_BG)
2158#define EXT2_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
2159 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
2160 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
2161
2162#define EXT3_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR
2163#define EXT3_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
2164 EXT4_FEATURE_INCOMPAT_RECOVER| \
2165 EXT4_FEATURE_INCOMPAT_META_BG)
2166#define EXT3_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
2167 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
2168 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
2169
2170#define EXT4_FEATURE_COMPAT_SUPP (EXT4_FEATURE_COMPAT_EXT_ATTR| \
2171 EXT4_FEATURE_COMPAT_ORPHAN_FILE)
2172#define EXT4_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
2173 EXT4_FEATURE_INCOMPAT_RECOVER| \
2174 EXT4_FEATURE_INCOMPAT_META_BG| \
2175 EXT4_FEATURE_INCOMPAT_EXTENTS| \
2176 EXT4_FEATURE_INCOMPAT_64BIT| \
2177 EXT4_FEATURE_INCOMPAT_FLEX_BG| \
2178 EXT4_FEATURE_INCOMPAT_EA_INODE| \
2179 EXT4_FEATURE_INCOMPAT_MMP | \
2180 EXT4_FEATURE_INCOMPAT_INLINE_DATA | \
2181 EXT4_FEATURE_INCOMPAT_ENCRYPT | \
2182 EXT4_FEATURE_INCOMPAT_CASEFOLD | \
2183 EXT4_FEATURE_INCOMPAT_CSUM_SEED | \
2184 EXT4_FEATURE_INCOMPAT_LARGEDIR)
2185#define EXT4_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
2186 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
2187 EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
2188 EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
2189 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
2190 EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
2191 EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
2192 EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
2193 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
2194 EXT4_FEATURE_RO_COMPAT_QUOTA |\
2195 EXT4_FEATURE_RO_COMPAT_PROJECT |\
2196 EXT4_FEATURE_RO_COMPAT_VERITY |\
2197 EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT)
2198
2199#define EXTN_FEATURE_FUNCS(ver) \
2200static inline bool ext4_has_unknown_ext##ver##_compat_features(struct super_block *sb) \
2201{ \
2202 return ((EXT4_SB(sb)->s_es->s_feature_compat & \
2203 cpu_to_le32(~EXT##ver##_FEATURE_COMPAT_SUPP)) != 0); \
2204} \
2205static inline bool ext4_has_unknown_ext##ver##_ro_compat_features(struct super_block *sb) \
2206{ \
2207 return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
2208 cpu_to_le32(~EXT##ver##_FEATURE_RO_COMPAT_SUPP)) != 0); \
2209} \
2210static inline bool ext4_has_unknown_ext##ver##_incompat_features(struct super_block *sb) \
2211{ \
2212 return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
2213 cpu_to_le32(~EXT##ver##_FEATURE_INCOMPAT_SUPP)) != 0); \
2214}
2215
2216EXTN_FEATURE_FUNCS(2)
2217EXTN_FEATURE_FUNCS(3)
2218EXTN_FEATURE_FUNCS(4)
2219
2220static inline bool ext4_has_compat_features(struct super_block *sb)
2221{
2222 return (EXT4_SB(sb)->s_es->s_feature_compat != 0);
2223}
2224static inline bool ext4_has_ro_compat_features(struct super_block *sb)
2225{
2226 return (EXT4_SB(sb)->s_es->s_feature_ro_compat != 0);
2227}
2228static inline bool ext4_has_incompat_features(struct super_block *sb)
2229{
2230 return (EXT4_SB(sb)->s_es->s_feature_incompat != 0);
2231}
2232
2233extern int ext4_feature_set_ok(struct super_block *sb, int readonly);
2234
2235/*
2236 * Superblock flags
2237 */
2238#define EXT4_FLAGS_RESIZING 0
2239#define EXT4_FLAGS_SHUTDOWN 1
2240#define EXT4_FLAGS_BDEV_IS_DAX 2
2241
2242static inline int ext4_forced_shutdown(struct super_block *sb)
2243{
2244 return test_bit(EXT4_FLAGS_SHUTDOWN, &EXT4_SB(sb)->s_ext4_flags);
2245}
2246
2247/*
2248 * Default values for user and/or group using reserved blocks
2249 */
2250#define EXT4_DEF_RESUID 0
2251#define EXT4_DEF_RESGID 0
2252
2253/*
2254 * Default project ID
2255 */
2256#define EXT4_DEF_PROJID 0
2257
2258#define EXT4_DEF_INODE_READAHEAD_BLKS 32
2259
2260/*
2261 * Default mount options
2262 */
2263#define EXT4_DEFM_DEBUG 0x0001
2264#define EXT4_DEFM_BSDGROUPS 0x0002
2265#define EXT4_DEFM_XATTR_USER 0x0004
2266#define EXT4_DEFM_ACL 0x0008
2267#define EXT4_DEFM_UID16 0x0010
2268#define EXT4_DEFM_JMODE 0x0060
2269#define EXT4_DEFM_JMODE_DATA 0x0020
2270#define EXT4_DEFM_JMODE_ORDERED 0x0040
2271#define EXT4_DEFM_JMODE_WBACK 0x0060
2272#define EXT4_DEFM_NOBARRIER 0x0100
2273#define EXT4_DEFM_BLOCK_VALIDITY 0x0200
2274#define EXT4_DEFM_DISCARD 0x0400
2275#define EXT4_DEFM_NODELALLOC 0x0800
2276
2277/*
2278 * Default journal batch times
2279 */
2280#define EXT4_DEF_MIN_BATCH_TIME 0
2281#define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
2282
2283/*
2284 * Minimum number of groups in a flexgroup before we separate out
2285 * directories into the first block group of a flexgroup
2286 */
2287#define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
2288
2289/*
2290 * Structure of a directory entry
2291 */
2292#define EXT4_NAME_LEN 255
2293/*
2294 * Base length of the ext4 directory entry excluding the name length
2295 */
2296#define EXT4_BASE_DIR_LEN (sizeof(struct ext4_dir_entry_2) - EXT4_NAME_LEN)
2297
2298struct ext4_dir_entry {
2299 __le32 inode; /* Inode number */
2300 __le16 rec_len; /* Directory entry length */
2301 __le16 name_len; /* Name length */
2302 char name[EXT4_NAME_LEN]; /* File name */
2303};
2304
2305
2306/*
2307 * Encrypted Casefolded entries require saving the hash on disk. This structure
2308 * followed ext4_dir_entry_2's name[name_len] at the next 4 byte aligned
2309 * boundary.
2310 */
2311struct ext4_dir_entry_hash {
2312 __le32 hash;
2313 __le32 minor_hash;
2314};
2315
2316/*
2317 * The new version of the directory entry. Since EXT4 structures are
2318 * stored in intel byte order, and the name_len field could never be
2319 * bigger than 255 chars, it's safe to reclaim the extra byte for the
2320 * file_type field.
2321 */
2322struct ext4_dir_entry_2 {
2323 __le32 inode; /* Inode number */
2324 __le16 rec_len; /* Directory entry length */
2325 __u8 name_len; /* Name length */
2326 __u8 file_type; /* See file type macros EXT4_FT_* below */
2327 char name[EXT4_NAME_LEN]; /* File name */
2328};
2329
2330/*
2331 * Access the hashes at the end of ext4_dir_entry_2
2332 */
2333#define EXT4_DIRENT_HASHES(entry) \
2334 ((struct ext4_dir_entry_hash *) \
2335 (((void *)(entry)) + \
2336 ((8 + (entry)->name_len + EXT4_DIR_ROUND) & ~EXT4_DIR_ROUND)))
2337#define EXT4_DIRENT_HASH(entry) le32_to_cpu(EXT4_DIRENT_HASHES(entry)->hash)
2338#define EXT4_DIRENT_MINOR_HASH(entry) \
2339 le32_to_cpu(EXT4_DIRENT_HASHES(entry)->minor_hash)
2340
2341static inline bool ext4_hash_in_dirent(const struct inode *inode)
2342{
2343 return IS_CASEFOLDED(inode) && IS_ENCRYPTED(inode);
2344}
2345
2346/*
2347 * This is a bogus directory entry at the end of each leaf block that
2348 * records checksums.
2349 */
2350struct ext4_dir_entry_tail {
2351 __le32 det_reserved_zero1; /* Pretend to be unused */
2352 __le16 det_rec_len; /* 12 */
2353 __u8 det_reserved_zero2; /* Zero name length */
2354 __u8 det_reserved_ft; /* 0xDE, fake file type */
2355 __le32 det_checksum; /* crc32c(uuid+inum+dirblock) */
2356};
2357
2358#define EXT4_DIRENT_TAIL(block, blocksize) \
2359 ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
2360 ((blocksize) - \
2361 sizeof(struct ext4_dir_entry_tail))))
2362
2363/*
2364 * Ext4 directory file types. Only the low 3 bits are used. The
2365 * other bits are reserved for now.
2366 */
2367#define EXT4_FT_UNKNOWN 0
2368#define EXT4_FT_REG_FILE 1
2369#define EXT4_FT_DIR 2
2370#define EXT4_FT_CHRDEV 3
2371#define EXT4_FT_BLKDEV 4
2372#define EXT4_FT_FIFO 5
2373#define EXT4_FT_SOCK 6
2374#define EXT4_FT_SYMLINK 7
2375
2376#define EXT4_FT_MAX 8
2377
2378#define EXT4_FT_DIR_CSUM 0xDE
2379
2380/*
2381 * EXT4_DIR_PAD defines the directory entries boundaries
2382 *
2383 * NOTE: It must be a multiple of 4
2384 */
2385#define EXT4_DIR_PAD 4
2386#define EXT4_DIR_ROUND (EXT4_DIR_PAD - 1)
2387#define EXT4_MAX_REC_LEN ((1<<16)-1)
2388
2389/*
2390 * The rec_len is dependent on the type of directory. Directories that are
2391 * casefolded and encrypted need to store the hash as well, so we add room for
2392 * ext4_extended_dir_entry_2. For all entries related to '.' or '..' you should
2393 * pass NULL for dir, as those entries do not use the extra fields.
2394 */
2395static inline unsigned int ext4_dir_rec_len(__u8 name_len,
2396 const struct inode *dir)
2397{
2398 int rec_len = (name_len + 8 + EXT4_DIR_ROUND);
2399
2400 if (dir && ext4_hash_in_dirent(dir))
2401 rec_len += sizeof(struct ext4_dir_entry_hash);
2402 return (rec_len & ~EXT4_DIR_ROUND);
2403}
2404
2405/*
2406 * If we ever get support for fs block sizes > page_size, we'll need
2407 * to remove the #if statements in the next two functions...
2408 */
2409static inline unsigned int
2410ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
2411{
2412 unsigned len = le16_to_cpu(dlen);
2413
2414#if (PAGE_SIZE >= 65536)
2415 if (len == EXT4_MAX_REC_LEN || len == 0)
2416 return blocksize;
2417 return (len & 65532) | ((len & 3) << 16);
2418#else
2419 return len;
2420#endif
2421}
2422
2423static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
2424{
2425 BUG_ON((len > blocksize) || (blocksize > (1 << 18)) || (len & 3));
2426#if (PAGE_SIZE >= 65536)
2427 if (len < 65536)
2428 return cpu_to_le16(len);
2429 if (len == blocksize) {
2430 if (blocksize == 65536)
2431 return cpu_to_le16(EXT4_MAX_REC_LEN);
2432 else
2433 return cpu_to_le16(0);
2434 }
2435 return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
2436#else
2437 return cpu_to_le16(len);
2438#endif
2439}
2440
2441/*
2442 * Hash Tree Directory indexing
2443 * (c) Daniel Phillips, 2001
2444 */
2445
2446#define is_dx(dir) (ext4_has_feature_dir_index((dir)->i_sb) && \
2447 ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
2448#define EXT4_DIR_LINK_MAX(dir) unlikely((dir)->i_nlink >= EXT4_LINK_MAX && \
2449 !(ext4_has_feature_dir_nlink((dir)->i_sb) && is_dx(dir)))
2450#define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
2451
2452/* Legal values for the dx_root hash_version field: */
2453
2454#define DX_HASH_LEGACY 0
2455#define DX_HASH_HALF_MD4 1
2456#define DX_HASH_TEA 2
2457#define DX_HASH_LEGACY_UNSIGNED 3
2458#define DX_HASH_HALF_MD4_UNSIGNED 4
2459#define DX_HASH_TEA_UNSIGNED 5
2460#define DX_HASH_SIPHASH 6
2461#define DX_HASH_LAST DX_HASH_SIPHASH
2462
2463static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
2464 const void *address, unsigned int length)
2465{
2466 struct {
2467 struct shash_desc shash;
2468 char ctx[4];
2469 } desc;
2470
2471 BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
2472
2473 desc.shash.tfm = sbi->s_chksum_driver;
2474 *(u32 *)desc.ctx = crc;
2475
2476 BUG_ON(crypto_shash_update(&desc.shash, address, length));
2477
2478 return *(u32 *)desc.ctx;
2479}
2480
2481#ifdef __KERNEL__
2482
2483/* hash info structure used by the directory hash */
2484struct dx_hash_info
2485{
2486 u32 hash;
2487 u32 minor_hash;
2488 int hash_version;
2489 u32 *seed;
2490};
2491
2492
2493/* 32 and 64 bit signed EOF for dx directories */
2494#define EXT4_HTREE_EOF_32BIT ((1UL << (32 - 1)) - 1)
2495#define EXT4_HTREE_EOF_64BIT ((1ULL << (64 - 1)) - 1)
2496
2497
2498/*
2499 * Control parameters used by ext4_htree_next_block
2500 */
2501#define HASH_NB_ALWAYS 1
2502
2503struct ext4_filename {
2504 const struct qstr *usr_fname;
2505 struct fscrypt_str disk_name;
2506 struct dx_hash_info hinfo;
2507#ifdef CONFIG_FS_ENCRYPTION
2508 struct fscrypt_str crypto_buf;
2509#endif
2510#if IS_ENABLED(CONFIG_UNICODE)
2511 struct qstr cf_name;
2512#endif
2513};
2514
2515#define fname_name(p) ((p)->disk_name.name)
2516#define fname_usr_name(p) ((p)->usr_fname->name)
2517#define fname_len(p) ((p)->disk_name.len)
2518
2519/*
2520 * Describe an inode's exact location on disk and in memory
2521 */
2522struct ext4_iloc
2523{
2524 struct buffer_head *bh;
2525 unsigned long offset;
2526 ext4_group_t block_group;
2527};
2528
2529static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
2530{
2531 return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
2532}
2533
2534static inline bool ext4_is_quota_file(struct inode *inode)
2535{
2536 return IS_NOQUOTA(inode) &&
2537 !(EXT4_I(inode)->i_flags & EXT4_EA_INODE_FL);
2538}
2539
2540/*
2541 * This structure is stuffed into the struct file's private_data field
2542 * for directories. It is where we put information so that we can do
2543 * readdir operations in hash tree order.
2544 */
2545struct dir_private_info {
2546 struct rb_root root;
2547 struct rb_node *curr_node;
2548 struct fname *extra_fname;
2549 loff_t last_pos;
2550 __u32 curr_hash;
2551 __u32 curr_minor_hash;
2552 __u32 next_hash;
2553 u64 cookie;
2554 bool initialized;
2555};
2556
2557/* calculate the first block number of the group */
2558static inline ext4_fsblk_t
2559ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
2560{
2561 return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
2562 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
2563}
2564
2565/*
2566 * Special error return code only used by dx_probe() and its callers.
2567 */
2568#define ERR_BAD_DX_DIR (-(MAX_ERRNO - 1))
2569
2570/* htree levels for ext4 */
2571#define EXT4_HTREE_LEVEL_COMPAT 2
2572#define EXT4_HTREE_LEVEL 3
2573
2574static inline int ext4_dir_htree_level(struct super_block *sb)
2575{
2576 return ext4_has_feature_largedir(sb) ?
2577 EXT4_HTREE_LEVEL : EXT4_HTREE_LEVEL_COMPAT;
2578}
2579
2580/*
2581 * Timeout and state flag for lazy initialization inode thread.
2582 */
2583#define EXT4_DEF_LI_WAIT_MULT 10
2584#define EXT4_DEF_LI_MAX_START_DELAY 5
2585#define EXT4_LAZYINIT_QUIT 0x0001
2586#define EXT4_LAZYINIT_RUNNING 0x0002
2587
2588/*
2589 * Lazy inode table initialization info
2590 */
2591struct ext4_lazy_init {
2592 unsigned long li_state;
2593 struct list_head li_request_list;
2594 struct mutex li_list_mtx;
2595};
2596
2597enum ext4_li_mode {
2598 EXT4_LI_MODE_PREFETCH_BBITMAP,
2599 EXT4_LI_MODE_ITABLE,
2600};
2601
2602struct ext4_li_request {
2603 struct super_block *lr_super;
2604 enum ext4_li_mode lr_mode;
2605 ext4_group_t lr_first_not_zeroed;
2606 ext4_group_t lr_next_group;
2607 struct list_head lr_request;
2608 unsigned long lr_next_sched;
2609 unsigned long lr_timeout;
2610};
2611
2612struct ext4_features {
2613 struct kobject f_kobj;
2614 struct completion f_kobj_unregister;
2615};
2616
2617/*
2618 * This structure will be used for multiple mount protection. It will be
2619 * written into the block number saved in the s_mmp_block field in the
2620 * superblock. Programs that check MMP should assume that if
2621 * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
2622 * to use the filesystem, regardless of how old the timestamp is.
2623 */
2624#define EXT4_MMP_MAGIC 0x004D4D50U /* ASCII for MMP */
2625#define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
2626#define EXT4_MMP_SEQ_FSCK 0xE24D4D50U /* mmp_seq value when being fscked */
2627#define EXT4_MMP_SEQ_MAX 0xE24D4D4FU /* maximum valid mmp_seq value */
2628
2629struct mmp_struct {
2630 __le32 mmp_magic; /* Magic number for MMP */
2631 __le32 mmp_seq; /* Sequence no. updated periodically */
2632
2633 /*
2634 * mmp_time, mmp_nodename & mmp_bdevname are only used for information
2635 * purposes and do not affect the correctness of the algorithm
2636 */
2637 __le64 mmp_time; /* Time last updated */
2638 char mmp_nodename[64]; /* Node which last updated MMP block */
2639 char mmp_bdevname[32]; /* Bdev which last updated MMP block */
2640
2641 /*
2642 * mmp_check_interval is used to verify if the MMP block has been
2643 * updated on the block device. The value is updated based on the
2644 * maximum time to write the MMP block during an update cycle.
2645 */
2646 __le16 mmp_check_interval;
2647
2648 __le16 mmp_pad1;
2649 __le32 mmp_pad2[226];
2650 __le32 mmp_checksum; /* crc32c(uuid+mmp_block) */
2651};
2652
2653/* arguments passed to the mmp thread */
2654struct mmpd_data {
2655 struct buffer_head *bh; /* bh from initial read_mmp_block() */
2656 struct super_block *sb; /* super block of the fs */
2657};
2658
2659/*
2660 * Check interval multiplier
2661 * The MMP block is written every update interval and initially checked every
2662 * update interval x the multiplier (the value is then adapted based on the
2663 * write latency). The reason is that writes can be delayed under load and we
2664 * don't want readers to incorrectly assume that the filesystem is no longer
2665 * in use.
2666 */
2667#define EXT4_MMP_CHECK_MULT 2UL
2668
2669/*
2670 * Minimum interval for MMP checking in seconds.
2671 */
2672#define EXT4_MMP_MIN_CHECK_INTERVAL 5UL
2673
2674/*
2675 * Maximum interval for MMP checking in seconds.
2676 */
2677#define EXT4_MMP_MAX_CHECK_INTERVAL 300UL
2678
2679/*
2680 * Function prototypes
2681 */
2682
2683/*
2684 * Ok, these declarations are also in <linux/kernel.h> but none of the
2685 * ext4 source programs needs to include it so they are duplicated here.
2686 */
2687# define NORET_TYPE /**/
2688# define ATTRIB_NORET __attribute__((noreturn))
2689# define NORET_AND noreturn,
2690
2691/* bitmap.c */
2692extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
2693void ext4_inode_bitmap_csum_set(struct super_block *sb,
2694 struct ext4_group_desc *gdp,
2695 struct buffer_head *bh);
2696int ext4_inode_bitmap_csum_verify(struct super_block *sb,
2697 struct ext4_group_desc *gdp,
2698 struct buffer_head *bh);
2699void ext4_block_bitmap_csum_set(struct super_block *sb,
2700 struct ext4_group_desc *gdp,
2701 struct buffer_head *bh);
2702int ext4_block_bitmap_csum_verify(struct super_block *sb,
2703 struct ext4_group_desc *gdp,
2704 struct buffer_head *bh);
2705
2706/* balloc.c */
2707extern void ext4_get_group_no_and_offset(struct super_block *sb,
2708 ext4_fsblk_t blocknr,
2709 ext4_group_t *blockgrpp,
2710 ext4_grpblk_t *offsetp);
2711extern ext4_group_t ext4_get_group_number(struct super_block *sb,
2712 ext4_fsblk_t block);
2713
2714extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
2715extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
2716 ext4_group_t group);
2717extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
2718 ext4_fsblk_t goal,
2719 unsigned int flags,
2720 unsigned long *count,
2721 int *errp);
2722extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
2723 s64 nclusters, unsigned int flags);
2724extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
2725extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
2726 ext4_group_t block_group,
2727 struct buffer_head ** bh);
2728extern struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2729 ext4_group_t group);
2730extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
2731
2732extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
2733 ext4_group_t block_group,
2734 bool ignore_locked);
2735extern int ext4_wait_block_bitmap(struct super_block *sb,
2736 ext4_group_t block_group,
2737 struct buffer_head *bh);
2738extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
2739 ext4_group_t block_group);
2740extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
2741 ext4_group_t block_group,
2742 struct ext4_group_desc *gdp);
2743ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
2744
2745#if IS_ENABLED(CONFIG_UNICODE)
2746extern int ext4_fname_setup_ci_filename(struct inode *dir,
2747 const struct qstr *iname,
2748 struct ext4_filename *fname);
2749
2750static inline void ext4_fname_free_ci_filename(struct ext4_filename *fname)
2751{
2752 kfree(fname->cf_name.name);
2753 fname->cf_name.name = NULL;
2754}
2755#else
2756static inline int ext4_fname_setup_ci_filename(struct inode *dir,
2757 const struct qstr *iname,
2758 struct ext4_filename *fname)
2759{
2760 return 0;
2761}
2762
2763static inline void ext4_fname_free_ci_filename(struct ext4_filename *fname)
2764{
2765}
2766#endif
2767
2768/* ext4 encryption related stuff goes here crypto.c */
2769#ifdef CONFIG_FS_ENCRYPTION
2770extern const struct fscrypt_operations ext4_cryptops;
2771
2772int ext4_fname_setup_filename(struct inode *dir, const struct qstr *iname,
2773 int lookup, struct ext4_filename *fname);
2774
2775int ext4_fname_prepare_lookup(struct inode *dir, struct dentry *dentry,
2776 struct ext4_filename *fname);
2777
2778void ext4_fname_free_filename(struct ext4_filename *fname);
2779
2780int ext4_ioctl_get_encryption_pwsalt(struct file *filp, void __user *arg);
2781
2782#else /* !CONFIG_FS_ENCRYPTION */
2783static inline int ext4_fname_setup_filename(struct inode *dir,
2784 const struct qstr *iname,
2785 int lookup,
2786 struct ext4_filename *fname)
2787{
2788 fname->usr_fname = iname;
2789 fname->disk_name.name = (unsigned char *) iname->name;
2790 fname->disk_name.len = iname->len;
2791
2792 return ext4_fname_setup_ci_filename(dir, iname, fname);
2793}
2794
2795static inline int ext4_fname_prepare_lookup(struct inode *dir,
2796 struct dentry *dentry,
2797 struct ext4_filename *fname)
2798{
2799 return ext4_fname_setup_filename(dir, &dentry->d_name, 1, fname);
2800}
2801
2802static inline void ext4_fname_free_filename(struct ext4_filename *fname)
2803{
2804 ext4_fname_free_ci_filename(fname);
2805}
2806
2807static inline int ext4_ioctl_get_encryption_pwsalt(struct file *filp,
2808 void __user *arg)
2809{
2810 return -EOPNOTSUPP;
2811}
2812#endif /* !CONFIG_FS_ENCRYPTION */
2813
2814/* dir.c */
2815extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
2816 struct file *,
2817 struct ext4_dir_entry_2 *,
2818 struct buffer_head *, char *, int,
2819 unsigned int);
2820#define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset) \
2821 unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
2822 (de), (bh), (buf), (size), (offset)))
2823extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
2824 __u32 minor_hash,
2825 struct ext4_dir_entry_2 *dirent,
2826 struct fscrypt_str *ent_name);
2827extern void ext4_htree_free_dir_info(struct dir_private_info *p);
2828extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
2829 struct buffer_head *bh,
2830 void *buf, int buf_size,
2831 struct ext4_filename *fname,
2832 struct ext4_dir_entry_2 **dest_de);
2833void ext4_insert_dentry(struct inode *dir, struct inode *inode,
2834 struct ext4_dir_entry_2 *de,
2835 int buf_size,
2836 struct ext4_filename *fname);
2837static inline void ext4_update_dx_flag(struct inode *inode)
2838{
2839 if (!ext4_has_feature_dir_index(inode->i_sb) &&
2840 ext4_test_inode_flag(inode, EXT4_INODE_INDEX)) {
2841 /* ext4_iget() should have caught this... */
2842 WARN_ON_ONCE(ext4_has_feature_metadata_csum(inode->i_sb));
2843 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
2844 }
2845}
2846static const unsigned char ext4_filetype_table[] = {
2847 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
2848};
2849
2850static inline unsigned char get_dtype(struct super_block *sb, int filetype)
2851{
2852 if (!ext4_has_feature_filetype(sb) || filetype >= EXT4_FT_MAX)
2853 return DT_UNKNOWN;
2854
2855 return ext4_filetype_table[filetype];
2856}
2857extern int ext4_check_all_de(struct inode *dir, struct buffer_head *bh,
2858 void *buf, int buf_size);
2859
2860/* fsync.c */
2861extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2862
2863/* hash.c */
2864extern int ext4fs_dirhash(const struct inode *dir, const char *name, int len,
2865 struct dx_hash_info *hinfo);
2866
2867/* ialloc.c */
2868extern int ext4_mark_inode_used(struct super_block *sb, int ino);
2869extern struct inode *__ext4_new_inode(struct mnt_idmap *, handle_t *,
2870 struct inode *, umode_t,
2871 const struct qstr *qstr, __u32 goal,
2872 uid_t *owner, __u32 i_flags,
2873 int handle_type, unsigned int line_no,
2874 int nblocks);
2875
2876#define ext4_new_inode(handle, dir, mode, qstr, goal, owner, i_flags) \
2877 __ext4_new_inode(&nop_mnt_idmap, (handle), (dir), (mode), (qstr), \
2878 (goal), (owner), i_flags, 0, 0, 0)
2879#define ext4_new_inode_start_handle(idmap, dir, mode, qstr, goal, owner, \
2880 type, nblocks) \
2881 __ext4_new_inode((idmap), NULL, (dir), (mode), (qstr), (goal), (owner), \
2882 0, (type), __LINE__, (nblocks))
2883
2884
2885extern void ext4_free_inode(handle_t *, struct inode *);
2886extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2887extern unsigned long ext4_count_free_inodes(struct super_block *);
2888extern unsigned long ext4_count_dirs(struct super_block *);
2889extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2890extern int ext4_init_inode_table(struct super_block *sb,
2891 ext4_group_t group, int barrier);
2892extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2893
2894/* fast_commit.c */
2895int ext4_fc_info_show(struct seq_file *seq, void *v);
2896void ext4_fc_init(struct super_block *sb, journal_t *journal);
2897void ext4_fc_init_inode(struct inode *inode);
2898void ext4_fc_track_range(handle_t *handle, struct inode *inode, ext4_lblk_t start,
2899 ext4_lblk_t end);
2900void __ext4_fc_track_unlink(handle_t *handle, struct inode *inode,
2901 struct dentry *dentry);
2902void __ext4_fc_track_link(handle_t *handle, struct inode *inode,
2903 struct dentry *dentry);
2904void ext4_fc_track_unlink(handle_t *handle, struct dentry *dentry);
2905void ext4_fc_track_link(handle_t *handle, struct dentry *dentry);
2906void __ext4_fc_track_create(handle_t *handle, struct inode *inode,
2907 struct dentry *dentry);
2908void ext4_fc_track_create(handle_t *handle, struct dentry *dentry);
2909void ext4_fc_track_inode(handle_t *handle, struct inode *inode);
2910void ext4_fc_mark_ineligible(struct super_block *sb, int reason, handle_t *handle);
2911void ext4_fc_start_update(struct inode *inode);
2912void ext4_fc_stop_update(struct inode *inode);
2913void ext4_fc_del(struct inode *inode);
2914bool ext4_fc_replay_check_excluded(struct super_block *sb, ext4_fsblk_t block);
2915void ext4_fc_replay_cleanup(struct super_block *sb);
2916int ext4_fc_commit(journal_t *journal, tid_t commit_tid);
2917int __init ext4_fc_init_dentry_cache(void);
2918void ext4_fc_destroy_dentry_cache(void);
2919int ext4_fc_record_regions(struct super_block *sb, int ino,
2920 ext4_lblk_t lblk, ext4_fsblk_t pblk,
2921 int len, int replay);
2922
2923/* mballoc.c */
2924extern const struct seq_operations ext4_mb_seq_groups_ops;
2925extern const struct seq_operations ext4_mb_seq_structs_summary_ops;
2926extern int ext4_seq_mb_stats_show(struct seq_file *seq, void *offset);
2927extern int ext4_mb_init(struct super_block *);
2928extern void ext4_mb_release(struct super_block *);
2929extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2930 struct ext4_allocation_request *, int *);
2931extern void ext4_discard_preallocations(struct inode *);
2932extern int __init ext4_init_mballoc(void);
2933extern void ext4_exit_mballoc(void);
2934extern ext4_group_t ext4_mb_prefetch(struct super_block *sb,
2935 ext4_group_t group,
2936 unsigned int nr, int *cnt);
2937extern void ext4_mb_prefetch_fini(struct super_block *sb, ext4_group_t group,
2938 unsigned int nr);
2939
2940extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2941 struct buffer_head *bh, ext4_fsblk_t block,
2942 unsigned long count, int flags);
2943extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2944 ext4_group_t ngroups);
2945extern int ext4_mb_add_groupinfo(struct super_block *sb,
2946 ext4_group_t i, struct ext4_group_desc *desc);
2947extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2948 ext4_fsblk_t block, unsigned long count);
2949extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2950extern void ext4_process_freed_data(struct super_block *sb, tid_t commit_tid);
2951extern void ext4_mb_mark_bb(struct super_block *sb, ext4_fsblk_t block,
2952 int len, bool state);
2953static inline bool ext4_mb_cr_expensive(enum criteria cr)
2954{
2955 return cr >= CR_GOAL_LEN_SLOW;
2956}
2957
2958/* inode.c */
2959void ext4_inode_csum_set(struct inode *inode, struct ext4_inode *raw,
2960 struct ext4_inode_info *ei);
2961int ext4_inode_is_fast_symlink(struct inode *inode);
2962struct buffer_head *ext4_getblk(handle_t *, struct inode *, ext4_lblk_t, int);
2963struct buffer_head *ext4_bread(handle_t *, struct inode *, ext4_lblk_t, int);
2964int ext4_bread_batch(struct inode *inode, ext4_lblk_t block, int bh_count,
2965 bool wait, struct buffer_head **bhs);
2966int ext4_get_block_unwritten(struct inode *inode, sector_t iblock,
2967 struct buffer_head *bh_result, int create);
2968int ext4_get_block(struct inode *inode, sector_t iblock,
2969 struct buffer_head *bh_result, int create);
2970int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2971 struct buffer_head *bh, int create);
2972int ext4_walk_page_buffers(handle_t *handle,
2973 struct inode *inode,
2974 struct buffer_head *head,
2975 unsigned from,
2976 unsigned to,
2977 int *partial,
2978 int (*fn)(handle_t *handle, struct inode *inode,
2979 struct buffer_head *bh));
2980int do_journal_get_write_access(handle_t *handle, struct inode *inode,
2981 struct buffer_head *bh);
2982#define FALL_BACK_TO_NONDELALLOC 1
2983#define CONVERT_INLINE_DATA 2
2984
2985typedef enum {
2986 EXT4_IGET_NORMAL = 0,
2987 EXT4_IGET_SPECIAL = 0x0001, /* OK to iget a system inode */
2988 EXT4_IGET_HANDLE = 0x0002, /* Inode # is from a handle */
2989 EXT4_IGET_BAD = 0x0004, /* Allow to iget a bad inode */
2990 EXT4_IGET_EA_INODE = 0x0008 /* Inode should contain an EA value */
2991} ext4_iget_flags;
2992
2993extern struct inode *__ext4_iget(struct super_block *sb, unsigned long ino,
2994 ext4_iget_flags flags, const char *function,
2995 unsigned int line);
2996
2997#define ext4_iget(sb, ino, flags) \
2998 __ext4_iget((sb), (ino), (flags), __func__, __LINE__)
2999
3000extern int ext4_write_inode(struct inode *, struct writeback_control *);
3001extern int ext4_setattr(struct mnt_idmap *, struct dentry *,
3002 struct iattr *);
3003extern u32 ext4_dio_alignment(struct inode *inode);
3004extern int ext4_getattr(struct mnt_idmap *, const struct path *,
3005 struct kstat *, u32, unsigned int);
3006extern void ext4_evict_inode(struct inode *);
3007extern void ext4_clear_inode(struct inode *);
3008extern int ext4_file_getattr(struct mnt_idmap *, const struct path *,
3009 struct kstat *, u32, unsigned int);
3010extern void ext4_dirty_inode(struct inode *, int);
3011extern int ext4_change_inode_journal_flag(struct inode *, int);
3012extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
3013extern int ext4_get_fc_inode_loc(struct super_block *sb, unsigned long ino,
3014 struct ext4_iloc *iloc);
3015extern int ext4_inode_attach_jinode(struct inode *inode);
3016extern int ext4_can_truncate(struct inode *inode);
3017extern int ext4_truncate(struct inode *);
3018extern int ext4_break_layouts(struct inode *);
3019extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length);
3020extern void ext4_set_inode_flags(struct inode *, bool init);
3021extern int ext4_alloc_da_blocks(struct inode *inode);
3022extern void ext4_set_aops(struct inode *inode);
3023extern int ext4_writepage_trans_blocks(struct inode *);
3024extern int ext4_normal_submit_inode_data_buffers(struct jbd2_inode *jinode);
3025extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
3026extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
3027 loff_t lstart, loff_t lend);
3028extern vm_fault_t ext4_page_mkwrite(struct vm_fault *vmf);
3029extern qsize_t *ext4_get_reserved_space(struct inode *inode);
3030extern int ext4_get_projid(struct inode *inode, kprojid_t *projid);
3031extern void ext4_da_release_space(struct inode *inode, int to_free);
3032extern void ext4_da_update_reserve_space(struct inode *inode,
3033 int used, int quota_claim);
3034extern int ext4_issue_zeroout(struct inode *inode, ext4_lblk_t lblk,
3035 ext4_fsblk_t pblk, ext4_lblk_t len);
3036
3037/* indirect.c */
3038extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
3039 struct ext4_map_blocks *map, int flags);
3040extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
3041extern void ext4_ind_truncate(handle_t *, struct inode *inode);
3042extern int ext4_ind_remove_space(handle_t *handle, struct inode *inode,
3043 ext4_lblk_t start, ext4_lblk_t end);
3044
3045/* ioctl.c */
3046extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
3047extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
3048int ext4_fileattr_set(struct mnt_idmap *idmap,
3049 struct dentry *dentry, struct fileattr *fa);
3050int ext4_fileattr_get(struct dentry *dentry, struct fileattr *fa);
3051extern void ext4_reset_inode_seed(struct inode *inode);
3052int ext4_update_overhead(struct super_block *sb, bool force);
3053int ext4_force_shutdown(struct super_block *sb, u32 flags);
3054
3055/* migrate.c */
3056extern int ext4_ext_migrate(struct inode *);
3057extern int ext4_ind_migrate(struct inode *inode);
3058
3059/* namei.c */
3060extern int ext4_init_new_dir(handle_t *handle, struct inode *dir,
3061 struct inode *inode);
3062extern int ext4_dirblock_csum_verify(struct inode *inode,
3063 struct buffer_head *bh);
3064extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
3065 __u32 start_minor_hash, __u32 *next_hash);
3066extern int ext4_search_dir(struct buffer_head *bh,
3067 char *search_buf,
3068 int buf_size,
3069 struct inode *dir,
3070 struct ext4_filename *fname,
3071 unsigned int offset,
3072 struct ext4_dir_entry_2 **res_dir);
3073extern int ext4_generic_delete_entry(struct inode *dir,
3074 struct ext4_dir_entry_2 *de_del,
3075 struct buffer_head *bh,
3076 void *entry_buf,
3077 int buf_size,
3078 int csum_size);
3079extern bool ext4_empty_dir(struct inode *inode);
3080
3081/* resize.c */
3082extern void ext4_kvfree_array_rcu(void *to_free);
3083extern int ext4_group_add(struct super_block *sb,
3084 struct ext4_new_group_data *input);
3085extern int ext4_group_extend(struct super_block *sb,
3086 struct ext4_super_block *es,
3087 ext4_fsblk_t n_blocks_count);
3088extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
3089extern unsigned int ext4_list_backups(struct super_block *sb,
3090 unsigned int *three, unsigned int *five,
3091 unsigned int *seven);
3092
3093/* super.c */
3094extern struct buffer_head *ext4_sb_bread(struct super_block *sb,
3095 sector_t block, blk_opf_t op_flags);
3096extern struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
3097 sector_t block);
3098extern void ext4_read_bh_nowait(struct buffer_head *bh, blk_opf_t op_flags,
3099 bh_end_io_t *end_io, bool simu_fail);
3100extern int ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags,
3101 bh_end_io_t *end_io, bool simu_fail);
3102extern int ext4_read_bh_lock(struct buffer_head *bh, blk_opf_t op_flags, bool wait);
3103extern void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block);
3104extern int ext4_seq_options_show(struct seq_file *seq, void *offset);
3105extern int ext4_calculate_overhead(struct super_block *sb);
3106extern __le32 ext4_superblock_csum(struct super_block *sb,
3107 struct ext4_super_block *es);
3108extern void ext4_superblock_csum_set(struct super_block *sb);
3109extern int ext4_alloc_flex_bg_array(struct super_block *sb,
3110 ext4_group_t ngroup);
3111extern const char *ext4_decode_error(struct super_block *sb, int errno,
3112 char nbuf[16]);
3113extern void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
3114 ext4_group_t block_group,
3115 unsigned int flags);
3116extern unsigned int ext4_num_base_meta_blocks(struct super_block *sb,
3117 ext4_group_t block_group);
3118
3119extern __printf(7, 8)
3120void __ext4_error(struct super_block *, const char *, unsigned int, bool,
3121 int, __u64, const char *, ...);
3122extern __printf(6, 7)
3123void __ext4_error_inode(struct inode *, const char *, unsigned int,
3124 ext4_fsblk_t, int, const char *, ...);
3125extern __printf(5, 6)
3126void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
3127 const char *, ...);
3128extern void __ext4_std_error(struct super_block *, const char *,
3129 unsigned int, int);
3130extern __printf(4, 5)
3131void __ext4_warning(struct super_block *, const char *, unsigned int,
3132 const char *, ...);
3133extern __printf(4, 5)
3134void __ext4_warning_inode(const struct inode *inode, const char *function,
3135 unsigned int line, const char *fmt, ...);
3136extern __printf(3, 4)
3137void __ext4_msg(struct super_block *, const char *, const char *, ...);
3138extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
3139 const char *, unsigned int, const char *);
3140extern __printf(7, 8)
3141void __ext4_grp_locked_error(const char *, unsigned int,
3142 struct super_block *, ext4_group_t,
3143 unsigned long, ext4_fsblk_t,
3144 const char *, ...);
3145
3146#define EXT4_ERROR_INODE(inode, fmt, a...) \
3147 ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
3148
3149#define EXT4_ERROR_INODE_ERR(inode, err, fmt, a...) \
3150 __ext4_error_inode((inode), __func__, __LINE__, 0, (err), (fmt), ## a)
3151
3152#define ext4_error_inode_block(inode, block, err, fmt, a...) \
3153 __ext4_error_inode((inode), __func__, __LINE__, (block), (err), \
3154 (fmt), ## a)
3155
3156#define EXT4_ERROR_FILE(file, block, fmt, a...) \
3157 ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
3158
3159#define ext4_abort(sb, err, fmt, a...) \
3160 __ext4_error((sb), __func__, __LINE__, true, (err), 0, (fmt), ## a)
3161
3162#ifdef CONFIG_PRINTK
3163
3164#define ext4_error_inode(inode, func, line, block, fmt, ...) \
3165 __ext4_error_inode(inode, func, line, block, 0, fmt, ##__VA_ARGS__)
3166#define ext4_error_inode_err(inode, func, line, block, err, fmt, ...) \
3167 __ext4_error_inode((inode), (func), (line), (block), \
3168 (err), (fmt), ##__VA_ARGS__)
3169#define ext4_error_file(file, func, line, block, fmt, ...) \
3170 __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__)
3171#define ext4_error(sb, fmt, ...) \
3172 __ext4_error((sb), __func__, __LINE__, false, 0, 0, (fmt), \
3173 ##__VA_ARGS__)
3174#define ext4_error_err(sb, err, fmt, ...) \
3175 __ext4_error((sb), __func__, __LINE__, false, (err), 0, (fmt), \
3176 ##__VA_ARGS__)
3177#define ext4_warning(sb, fmt, ...) \
3178 __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
3179#define ext4_warning_inode(inode, fmt, ...) \
3180 __ext4_warning_inode(inode, __func__, __LINE__, fmt, ##__VA_ARGS__)
3181#define ext4_msg(sb, level, fmt, ...) \
3182 __ext4_msg(sb, level, fmt, ##__VA_ARGS__)
3183#define dump_mmp_msg(sb, mmp, msg) \
3184 __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg)
3185#define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...) \
3186 __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \
3187 fmt, ##__VA_ARGS__)
3188
3189#else
3190
3191#define ext4_error_inode(inode, func, line, block, fmt, ...) \
3192do { \
3193 no_printk(fmt, ##__VA_ARGS__); \
3194 __ext4_error_inode(inode, "", 0, block, 0, " "); \
3195} while (0)
3196#define ext4_error_inode_err(inode, func, line, block, err, fmt, ...) \
3197do { \
3198 no_printk(fmt, ##__VA_ARGS__); \
3199 __ext4_error_inode(inode, "", 0, block, err, " "); \
3200} while (0)
3201#define ext4_error_file(file, func, line, block, fmt, ...) \
3202do { \
3203 no_printk(fmt, ##__VA_ARGS__); \
3204 __ext4_error_file(file, "", 0, block, " "); \
3205} while (0)
3206#define ext4_error(sb, fmt, ...) \
3207do { \
3208 no_printk(fmt, ##__VA_ARGS__); \
3209 __ext4_error(sb, "", 0, false, 0, 0, " "); \
3210} while (0)
3211#define ext4_error_err(sb, err, fmt, ...) \
3212do { \
3213 no_printk(fmt, ##__VA_ARGS__); \
3214 __ext4_error(sb, "", 0, false, err, 0, " "); \
3215} while (0)
3216#define ext4_warning(sb, fmt, ...) \
3217do { \
3218 no_printk(fmt, ##__VA_ARGS__); \
3219 __ext4_warning(sb, "", 0, " "); \
3220} while (0)
3221#define ext4_warning_inode(inode, fmt, ...) \
3222do { \
3223 no_printk(fmt, ##__VA_ARGS__); \
3224 __ext4_warning_inode(inode, "", 0, " "); \
3225} while (0)
3226#define ext4_msg(sb, level, fmt, ...) \
3227do { \
3228 no_printk(fmt, ##__VA_ARGS__); \
3229 __ext4_msg(sb, "", " "); \
3230} while (0)
3231#define dump_mmp_msg(sb, mmp, msg) \
3232 __dump_mmp_msg(sb, mmp, "", 0, "")
3233#define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...) \
3234do { \
3235 no_printk(fmt, ##__VA_ARGS__); \
3236 __ext4_grp_locked_error("", 0, sb, grp, ino, block, " "); \
3237} while (0)
3238
3239#endif
3240
3241extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
3242 struct ext4_group_desc *bg);
3243extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
3244 struct ext4_group_desc *bg);
3245extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
3246 struct ext4_group_desc *bg);
3247extern __u32 ext4_free_group_clusters(struct super_block *sb,
3248 struct ext4_group_desc *bg);
3249extern __u32 ext4_free_inodes_count(struct super_block *sb,
3250 struct ext4_group_desc *bg);
3251extern __u32 ext4_used_dirs_count(struct super_block *sb,
3252 struct ext4_group_desc *bg);
3253extern __u32 ext4_itable_unused_count(struct super_block *sb,
3254 struct ext4_group_desc *bg);
3255extern void ext4_block_bitmap_set(struct super_block *sb,
3256 struct ext4_group_desc *bg, ext4_fsblk_t blk);
3257extern void ext4_inode_bitmap_set(struct super_block *sb,
3258 struct ext4_group_desc *bg, ext4_fsblk_t blk);
3259extern void ext4_inode_table_set(struct super_block *sb,
3260 struct ext4_group_desc *bg, ext4_fsblk_t blk);
3261extern void ext4_free_group_clusters_set(struct super_block *sb,
3262 struct ext4_group_desc *bg,
3263 __u32 count);
3264extern void ext4_free_inodes_set(struct super_block *sb,
3265 struct ext4_group_desc *bg, __u32 count);
3266extern void ext4_used_dirs_set(struct super_block *sb,
3267 struct ext4_group_desc *bg, __u32 count);
3268extern void ext4_itable_unused_set(struct super_block *sb,
3269 struct ext4_group_desc *bg, __u32 count);
3270extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
3271 struct ext4_group_desc *gdp);
3272extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
3273 struct ext4_group_desc *gdp);
3274extern int ext4_register_li_request(struct super_block *sb,
3275 ext4_group_t first_not_zeroed);
3276
3277static inline int ext4_has_metadata_csum(struct super_block *sb)
3278{
3279 WARN_ON_ONCE(ext4_has_feature_metadata_csum(sb) &&
3280 !EXT4_SB(sb)->s_chksum_driver);
3281
3282 return ext4_has_feature_metadata_csum(sb) &&
3283 (EXT4_SB(sb)->s_chksum_driver != NULL);
3284}
3285
3286static inline int ext4_has_group_desc_csum(struct super_block *sb)
3287{
3288 return ext4_has_feature_gdt_csum(sb) || ext4_has_metadata_csum(sb);
3289}
3290
3291#define ext4_read_incompat_64bit_val(es, name) \
3292 (((es)->s_feature_incompat & cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT) \
3293 ? (ext4_fsblk_t)le32_to_cpu(es->name##_hi) << 32 : 0) | \
3294 le32_to_cpu(es->name##_lo))
3295
3296static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
3297{
3298 return ext4_read_incompat_64bit_val(es, s_blocks_count);
3299}
3300
3301static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
3302{
3303 return ext4_read_incompat_64bit_val(es, s_r_blocks_count);
3304}
3305
3306static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
3307{
3308 return ext4_read_incompat_64bit_val(es, s_free_blocks_count);
3309}
3310
3311static inline void ext4_blocks_count_set(struct ext4_super_block *es,
3312 ext4_fsblk_t blk)
3313{
3314 es->s_blocks_count_lo = cpu_to_le32((u32)blk);
3315 es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
3316}
3317
3318static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
3319 ext4_fsblk_t blk)
3320{
3321 es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
3322 es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
3323}
3324
3325static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
3326 ext4_fsblk_t blk)
3327{
3328 es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
3329 es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
3330}
3331
3332static inline loff_t ext4_isize(struct super_block *sb,
3333 struct ext4_inode *raw_inode)
3334{
3335 if (ext4_has_feature_largedir(sb) ||
3336 S_ISREG(le16_to_cpu(raw_inode->i_mode)))
3337 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
3338 le32_to_cpu(raw_inode->i_size_lo);
3339
3340 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
3341}
3342
3343static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
3344{
3345 raw_inode->i_size_lo = cpu_to_le32(i_size);
3346 raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
3347}
3348
3349/*
3350 * Reading s_groups_count requires using smp_rmb() afterwards. See
3351 * the locking protocol documented in the comments of ext4_group_add()
3352 * in resize.c
3353 */
3354static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
3355{
3356 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3357
3358 smp_rmb();
3359 return ngroups;
3360}
3361
3362static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
3363 ext4_group_t block_group)
3364{
3365 return block_group >> sbi->s_log_groups_per_flex;
3366}
3367
3368static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
3369{
3370 return 1 << sbi->s_log_groups_per_flex;
3371}
3372
3373#define ext4_std_error(sb, errno) \
3374do { \
3375 if ((errno)) \
3376 __ext4_std_error((sb), __func__, __LINE__, (errno)); \
3377} while (0)
3378
3379#ifdef CONFIG_SMP
3380/* Each CPU can accumulate percpu_counter_batch clusters in their local
3381 * counters. So we need to make sure we have free clusters more
3382 * than percpu_counter_batch * nr_cpu_ids. Also add a window of 4 times.
3383 */
3384#define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
3385#else
3386#define EXT4_FREECLUSTERS_WATERMARK 0
3387#endif
3388
3389/* Update i_disksize. Requires i_rwsem to avoid races with truncate */
3390static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
3391{
3392 WARN_ON_ONCE(S_ISREG(inode->i_mode) &&
3393 !inode_is_locked(inode));
3394 down_write(&EXT4_I(inode)->i_data_sem);
3395 if (newsize > EXT4_I(inode)->i_disksize)
3396 WRITE_ONCE(EXT4_I(inode)->i_disksize, newsize);
3397 up_write(&EXT4_I(inode)->i_data_sem);
3398}
3399
3400/* Update i_size, i_disksize. Requires i_rwsem to avoid races with truncate */
3401static inline int ext4_update_inode_size(struct inode *inode, loff_t newsize)
3402{
3403 int changed = 0;
3404
3405 if (newsize > inode->i_size) {
3406 i_size_write(inode, newsize);
3407 changed = 1;
3408 }
3409 if (newsize > EXT4_I(inode)->i_disksize) {
3410 ext4_update_i_disksize(inode, newsize);
3411 changed |= 2;
3412 }
3413 return changed;
3414}
3415
3416int ext4_update_disksize_before_punch(struct inode *inode, loff_t offset,
3417 loff_t len);
3418
3419struct ext4_group_info {
3420 unsigned long bb_state;
3421#ifdef AGGRESSIVE_CHECK
3422 unsigned long bb_check_counter;
3423#endif
3424 struct rb_root bb_free_root;
3425 ext4_grpblk_t bb_first_free; /* first free block */
3426 ext4_grpblk_t bb_free; /* total free blocks */
3427 ext4_grpblk_t bb_fragments; /* nr of freespace fragments */
3428 int bb_avg_fragment_size_order; /* order of average
3429 fragment in BG */
3430 ext4_grpblk_t bb_largest_free_order;/* order of largest frag in BG */
3431 ext4_group_t bb_group; /* Group number */
3432 struct list_head bb_prealloc_list;
3433#ifdef DOUBLE_CHECK
3434 void *bb_bitmap;
3435#endif
3436 struct rw_semaphore alloc_sem;
3437 struct list_head bb_avg_fragment_size_node;
3438 struct list_head bb_largest_free_order_node;
3439 ext4_grpblk_t bb_counters[]; /* Nr of free power-of-two-block
3440 * regions, index is order.
3441 * bb_counters[3] = 5 means
3442 * 5 free 8-block regions. */
3443};
3444
3445#define EXT4_GROUP_INFO_NEED_INIT_BIT 0
3446#define EXT4_GROUP_INFO_WAS_TRIMMED_BIT 1
3447#define EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT 2
3448#define EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT 3
3449#define EXT4_GROUP_INFO_BBITMAP_CORRUPT \
3450 (1 << EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT)
3451#define EXT4_GROUP_INFO_IBITMAP_CORRUPT \
3452 (1 << EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT)
3453#define EXT4_GROUP_INFO_BBITMAP_READ_BIT 4
3454
3455#define EXT4_MB_GRP_NEED_INIT(grp) \
3456 (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
3457#define EXT4_MB_GRP_BBITMAP_CORRUPT(grp) \
3458 (test_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
3459#define EXT4_MB_GRP_IBITMAP_CORRUPT(grp) \
3460 (test_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
3461
3462#define EXT4_MB_GRP_WAS_TRIMMED(grp) \
3463 (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3464#define EXT4_MB_GRP_SET_TRIMMED(grp) \
3465 (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3466#define EXT4_MB_GRP_CLEAR_TRIMMED(grp) \
3467 (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3468#define EXT4_MB_GRP_TEST_AND_SET_READ(grp) \
3469 (test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_READ_BIT, &((grp)->bb_state)))
3470
3471#define EXT4_MAX_CONTENTION 8
3472#define EXT4_CONTENTION_THRESHOLD 2
3473
3474static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
3475 ext4_group_t group)
3476{
3477 return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
3478}
3479
3480/*
3481 * Returns true if the filesystem is busy enough that attempts to
3482 * access the block group locks has run into contention.
3483 */
3484static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
3485{
3486 return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
3487}
3488
3489static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
3490{
3491 spinlock_t *lock = ext4_group_lock_ptr(sb, group);
3492 if (spin_trylock(lock))
3493 /*
3494 * We're able to grab the lock right away, so drop the
3495 * lock contention counter.
3496 */
3497 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
3498 else {
3499 /*
3500 * The lock is busy, so bump the contention counter,
3501 * and then wait on the spin lock.
3502 */
3503 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
3504 EXT4_MAX_CONTENTION);
3505 spin_lock(lock);
3506 }
3507}
3508
3509static inline void ext4_unlock_group(struct super_block *sb,
3510 ext4_group_t group)
3511{
3512 spin_unlock(ext4_group_lock_ptr(sb, group));
3513}
3514
3515#ifdef CONFIG_QUOTA
3516static inline bool ext4_quota_capable(struct super_block *sb)
3517{
3518 return (test_opt(sb, QUOTA) || ext4_has_feature_quota(sb));
3519}
3520
3521static inline bool ext4_is_quota_journalled(struct super_block *sb)
3522{
3523 struct ext4_sb_info *sbi = EXT4_SB(sb);
3524
3525 return (ext4_has_feature_quota(sb) ||
3526 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]);
3527}
3528int ext4_enable_quotas(struct super_block *sb);
3529#endif
3530
3531/*
3532 * Block validity checking
3533 */
3534#define ext4_check_indirect_blockref(inode, bh) \
3535 ext4_check_blockref(__func__, __LINE__, inode, \
3536 (__le32 *)(bh)->b_data, \
3537 EXT4_ADDR_PER_BLOCK((inode)->i_sb))
3538
3539#define ext4_ind_check_inode(inode) \
3540 ext4_check_blockref(__func__, __LINE__, inode, \
3541 EXT4_I(inode)->i_data, \
3542 EXT4_NDIR_BLOCKS)
3543
3544/*
3545 * Inodes and files operations
3546 */
3547
3548/* dir.c */
3549extern const struct file_operations ext4_dir_operations;
3550
3551/* file.c */
3552extern const struct inode_operations ext4_file_inode_operations;
3553extern const struct file_operations ext4_file_operations;
3554extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
3555
3556/* inline.c */
3557extern int ext4_get_max_inline_size(struct inode *inode);
3558extern int ext4_find_inline_data_nolock(struct inode *inode);
3559extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
3560
3561int ext4_readpage_inline(struct inode *inode, struct folio *folio);
3562extern int ext4_try_to_write_inline_data(struct address_space *mapping,
3563 struct inode *inode,
3564 loff_t pos, unsigned len,
3565 struct folio **foliop);
3566int ext4_write_inline_data_end(struct inode *inode, loff_t pos, unsigned len,
3567 unsigned copied, struct folio *folio);
3568extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
3569 struct inode *inode,
3570 loff_t pos, unsigned len,
3571 struct folio **foliop,
3572 void **fsdata);
3573extern int ext4_try_add_inline_entry(handle_t *handle,
3574 struct ext4_filename *fname,
3575 struct inode *dir, struct inode *inode);
3576extern int ext4_try_create_inline_dir(handle_t *handle,
3577 struct inode *parent,
3578 struct inode *inode);
3579extern int ext4_read_inline_dir(struct file *filp,
3580 struct dir_context *ctx,
3581 int *has_inline_data);
3582extern int ext4_inlinedir_to_tree(struct file *dir_file,
3583 struct inode *dir, ext4_lblk_t block,
3584 struct dx_hash_info *hinfo,
3585 __u32 start_hash, __u32 start_minor_hash,
3586 int *has_inline_data);
3587extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
3588 struct ext4_filename *fname,
3589 struct ext4_dir_entry_2 **res_dir,
3590 int *has_inline_data);
3591extern int ext4_delete_inline_entry(handle_t *handle,
3592 struct inode *dir,
3593 struct ext4_dir_entry_2 *de_del,
3594 struct buffer_head *bh,
3595 int *has_inline_data);
3596extern bool empty_inline_dir(struct inode *dir, int *has_inline_data);
3597extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
3598 struct ext4_dir_entry_2 **parent_de,
3599 int *retval);
3600extern void *ext4_read_inline_link(struct inode *inode);
3601
3602struct iomap;
3603extern int ext4_inline_data_iomap(struct inode *inode, struct iomap *iomap);
3604
3605extern int ext4_inline_data_truncate(struct inode *inode, int *has_inline);
3606
3607extern int ext4_convert_inline_data(struct inode *inode);
3608
3609static inline int ext4_has_inline_data(struct inode *inode)
3610{
3611 return ext4_test_inode_flag(inode, EXT4_INODE_INLINE_DATA) &&
3612 EXT4_I(inode)->i_inline_off;
3613}
3614
3615/* namei.c */
3616extern const struct inode_operations ext4_dir_inode_operations;
3617extern const struct inode_operations ext4_special_inode_operations;
3618extern struct dentry *ext4_get_parent(struct dentry *child);
3619extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
3620 struct ext4_dir_entry_2 *de,
3621 int blocksize, int csum_size,
3622 unsigned int parent_ino, int dotdot_real_len);
3623extern void ext4_initialize_dirent_tail(struct buffer_head *bh,
3624 unsigned int blocksize);
3625extern int ext4_handle_dirty_dirblock(handle_t *handle, struct inode *inode,
3626 struct buffer_head *bh);
3627extern int __ext4_unlink(struct inode *dir, const struct qstr *d_name,
3628 struct inode *inode, struct dentry *dentry);
3629extern int __ext4_link(struct inode *dir, struct inode *inode,
3630 struct dentry *dentry);
3631
3632#define S_SHIFT 12
3633static const unsigned char ext4_type_by_mode[(S_IFMT >> S_SHIFT) + 1] = {
3634 [S_IFREG >> S_SHIFT] = EXT4_FT_REG_FILE,
3635 [S_IFDIR >> S_SHIFT] = EXT4_FT_DIR,
3636 [S_IFCHR >> S_SHIFT] = EXT4_FT_CHRDEV,
3637 [S_IFBLK >> S_SHIFT] = EXT4_FT_BLKDEV,
3638 [S_IFIFO >> S_SHIFT] = EXT4_FT_FIFO,
3639 [S_IFSOCK >> S_SHIFT] = EXT4_FT_SOCK,
3640 [S_IFLNK >> S_SHIFT] = EXT4_FT_SYMLINK,
3641};
3642
3643static inline void ext4_set_de_type(struct super_block *sb,
3644 struct ext4_dir_entry_2 *de,
3645 umode_t mode) {
3646 if (ext4_has_feature_filetype(sb))
3647 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
3648}
3649
3650/* readpages.c */
3651extern int ext4_mpage_readpages(struct inode *inode,
3652 struct readahead_control *rac, struct folio *folio);
3653extern int __init ext4_init_post_read_processing(void);
3654extern void ext4_exit_post_read_processing(void);
3655
3656/* symlink.c */
3657extern const struct inode_operations ext4_encrypted_symlink_inode_operations;
3658extern const struct inode_operations ext4_symlink_inode_operations;
3659extern const struct inode_operations ext4_fast_symlink_inode_operations;
3660
3661/* sysfs.c */
3662extern void ext4_notify_error_sysfs(struct ext4_sb_info *sbi);
3663extern int ext4_register_sysfs(struct super_block *sb);
3664extern void ext4_unregister_sysfs(struct super_block *sb);
3665extern int __init ext4_init_sysfs(void);
3666extern void ext4_exit_sysfs(void);
3667
3668/* block_validity */
3669extern void ext4_release_system_zone(struct super_block *sb);
3670extern int ext4_setup_system_zone(struct super_block *sb);
3671extern int __init ext4_init_system_zone(void);
3672extern void ext4_exit_system_zone(void);
3673extern int ext4_inode_block_valid(struct inode *inode,
3674 ext4_fsblk_t start_blk,
3675 unsigned int count);
3676extern int ext4_check_blockref(const char *, unsigned int,
3677 struct inode *, __le32 *, unsigned int);
3678extern int ext4_sb_block_valid(struct super_block *sb, struct inode *inode,
3679 ext4_fsblk_t start_blk, unsigned int count);
3680
3681
3682/* extents.c */
3683struct ext4_ext_path;
3684struct ext4_extent;
3685
3686/*
3687 * Maximum number of logical blocks in a file; ext4_extent's ee_block is
3688 * __le32.
3689 */
3690#define EXT_MAX_BLOCKS 0xffffffff
3691
3692extern void ext4_ext_tree_init(handle_t *handle, struct inode *inode);
3693extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
3694extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
3695 struct ext4_map_blocks *map, int flags);
3696extern int ext4_ext_truncate(handle_t *, struct inode *);
3697extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
3698 ext4_lblk_t end);
3699extern void ext4_ext_init(struct super_block *);
3700extern void ext4_ext_release(struct super_block *);
3701extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
3702 loff_t len);
3703extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
3704 loff_t offset, ssize_t len);
3705extern int ext4_convert_unwritten_io_end_vec(handle_t *handle,
3706 ext4_io_end_t *io_end);
3707extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
3708 struct ext4_map_blocks *map, int flags);
3709extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
3710 int num,
3711 struct ext4_ext_path *path);
3712extern struct ext4_ext_path *ext4_ext_insert_extent(
3713 handle_t *handle, struct inode *inode,
3714 struct ext4_ext_path *path,
3715 struct ext4_extent *newext, int gb_flags);
3716extern struct ext4_ext_path *ext4_find_extent(struct inode *, ext4_lblk_t,
3717 struct ext4_ext_path *,
3718 int flags);
3719extern void ext4_free_ext_path(struct ext4_ext_path *);
3720extern int ext4_ext_check_inode(struct inode *inode);
3721extern ext4_lblk_t ext4_ext_next_allocated_block(struct ext4_ext_path *path);
3722extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
3723 __u64 start, __u64 len);
3724extern int ext4_get_es_cache(struct inode *inode,
3725 struct fiemap_extent_info *fieinfo,
3726 __u64 start, __u64 len);
3727extern int ext4_ext_precache(struct inode *inode);
3728extern 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);
3732extern int ext4_clu_mapped(struct inode *inode, ext4_lblk_t lclu);
3733extern int ext4_datasem_ensure_credits(handle_t *handle, struct inode *inode,
3734 int check_cred, int restart_cred,
3735 int revoke_cred);
3736extern void ext4_ext_replay_shrink_inode(struct inode *inode, ext4_lblk_t end);
3737extern int ext4_ext_replay_set_iblocks(struct inode *inode);
3738extern int ext4_ext_replay_update_ex(struct inode *inode, ext4_lblk_t start,
3739 int len, int unwritten, ext4_fsblk_t pblk);
3740extern int ext4_ext_clear_bb(struct inode *inode);
3741
3742
3743/* move_extent.c */
3744extern void ext4_double_down_write_data_sem(struct inode *first,
3745 struct inode *second);
3746extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
3747 struct inode *donor_inode);
3748extern 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 */
3753extern int __init ext4_init_pageio(void);
3754extern void ext4_exit_pageio(void);
3755extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
3756extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
3757extern int ext4_put_io_end(ext4_io_end_t *io_end);
3758extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
3759extern void ext4_io_submit_init(struct ext4_io_submit *io,
3760 struct writeback_control *wbc);
3761extern void ext4_end_io_rsv_work(struct work_struct *work);
3762extern void ext4_io_submit(struct ext4_io_submit *io);
3763int ext4_bio_write_folio(struct ext4_io_submit *io, struct folio *page,
3764 size_t len);
3765extern struct ext4_io_end_vec *ext4_alloc_io_end_vec(ext4_io_end_t *io_end);
3766extern struct ext4_io_end_vec *ext4_last_io_end_vec(ext4_io_end_t *io_end);
3767
3768/* mmp.c */
3769extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
3770
3771/* mmp.c */
3772extern void ext4_stop_mmpd(struct ext4_sb_info *sbi);
3773
3774/* verity.c */
3775extern const struct fsverity_operations ext4_verityops;
3776
3777/* orphan.c */
3778extern int ext4_orphan_add(handle_t *, struct inode *);
3779extern int ext4_orphan_del(handle_t *, struct inode *);
3780extern void ext4_orphan_cleanup(struct super_block *sb,
3781 struct ext4_super_block *es);
3782extern void ext4_release_orphan_info(struct super_block *sb);
3783extern int ext4_init_orphan_info(struct super_block *sb);
3784extern int ext4_orphan_file_empty(struct super_block *sb);
3785extern 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
3797static inline int bitmap_uptodate(struct buffer_head *bh)
3798{
3799 return (buffer_uptodate(bh) &&
3800 test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
3801}
3802static 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])
3811extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
3812
3813extern int ext4_resize_begin(struct super_block *sb);
3814extern int ext4_resize_end(struct super_block *sb, bool update_backups);
3815
3816static 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
3825static 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
3837extern const struct iomap_ops ext4_iomap_ops;
3838extern const struct iomap_ops ext4_iomap_overwrite_ops;
3839extern const struct iomap_ops ext4_iomap_report_ops;
3840
3841static 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
3854static inline bool ext4_inode_can_atomic_write(struct inode *inode)
3855{
3856
3857 return S_ISREG(inode->i_mode) && EXT4_SB(inode->i_sb)->s_awu_min > 0;
3858}
3859
3860extern int ext4_block_write_begin(handle_t *handle, struct folio *folio,
3861 loff_t pos, unsigned len,
3862 get_block_t *get_block);
3863#endif /* __KERNEL__ */
3864
3865#define EFSBADCRC EBADMSG /* Bad CRC detected */
3866#define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
3867
3868#endif /* _EXT4_H */