Loading...
1/*
2 * ext4.h
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/include/linux/minix_fs.h
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 */
15
16#ifndef _EXT4_H
17#define _EXT4_H
18
19#include <linux/types.h>
20#include <linux/blkdev.h>
21#include <linux/magic.h>
22#include <linux/jbd2.h>
23#include <linux/quota.h>
24#include <linux/rwsem.h>
25#include <linux/rbtree.h>
26#include <linux/seqlock.h>
27#include <linux/mutex.h>
28#include <linux/timer.h>
29#include <linux/wait.h>
30#include <linux/blockgroup_lock.h>
31#include <linux/percpu_counter.h>
32#include <crypto/hash.h>
33#ifdef __KERNEL__
34#include <linux/compat.h>
35#endif
36
37/*
38 * The fourth extended filesystem constants/structures
39 */
40
41/*
42 * Define EXT4FS_DEBUG to produce debug messages
43 */
44#undef EXT4FS_DEBUG
45
46/*
47 * Debug code
48 */
49#ifdef EXT4FS_DEBUG
50#define ext4_debug(f, a...) \
51 do { \
52 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:", \
53 __FILE__, __LINE__, __func__); \
54 printk(KERN_DEBUG f, ## a); \
55 } while (0)
56#else
57#define ext4_debug(fmt, ...) no_printk(fmt, ##__VA_ARGS__)
58#endif
59
60#define EXT4_ERROR_INODE(inode, fmt, a...) \
61 ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
62
63#define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...) \
64 ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
65
66#define EXT4_ERROR_FILE(file, block, fmt, a...) \
67 ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
68
69/* data type for block offset of block group */
70typedef int ext4_grpblk_t;
71
72/* data type for filesystem-wide blocks number */
73typedef unsigned long long ext4_fsblk_t;
74
75/* data type for file logical block number */
76typedef __u32 ext4_lblk_t;
77
78/* data type for block group number */
79typedef unsigned int ext4_group_t;
80
81/*
82 * Flags used in mballoc's allocation_context flags field.
83 *
84 * Also used to show what's going on for debugging purposes when the
85 * flag field is exported via the traceport interface
86 */
87
88/* prefer goal again. length */
89#define EXT4_MB_HINT_MERGE 0x0001
90/* blocks already reserved */
91#define EXT4_MB_HINT_RESERVED 0x0002
92/* metadata is being allocated */
93#define EXT4_MB_HINT_METADATA 0x0004
94/* first blocks in the file */
95#define EXT4_MB_HINT_FIRST 0x0008
96/* search for the best chunk */
97#define EXT4_MB_HINT_BEST 0x0010
98/* data is being allocated */
99#define EXT4_MB_HINT_DATA 0x0020
100/* don't preallocate (for tails) */
101#define EXT4_MB_HINT_NOPREALLOC 0x0040
102/* allocate for locality group */
103#define EXT4_MB_HINT_GROUP_ALLOC 0x0080
104/* allocate goal blocks or none */
105#define EXT4_MB_HINT_GOAL_ONLY 0x0100
106/* goal is meaningful */
107#define EXT4_MB_HINT_TRY_GOAL 0x0200
108/* blocks already pre-reserved by delayed allocation */
109#define EXT4_MB_DELALLOC_RESERVED 0x0400
110/* We are doing stream allocation */
111#define EXT4_MB_STREAM_ALLOC 0x0800
112/* Use reserved root blocks if needed */
113#define EXT4_MB_USE_ROOT_BLOCKS 0x1000
114
115struct ext4_allocation_request {
116 /* target inode for block we're allocating */
117 struct inode *inode;
118 /* how many blocks we want to allocate */
119 unsigned int len;
120 /* logical block in target inode */
121 ext4_lblk_t logical;
122 /* the closest logical allocated block to the left */
123 ext4_lblk_t lleft;
124 /* the closest logical allocated block to the right */
125 ext4_lblk_t lright;
126 /* phys. target (a hint) */
127 ext4_fsblk_t goal;
128 /* phys. block for the closest logical allocated block to the left */
129 ext4_fsblk_t pleft;
130 /* phys. block for the closest logical allocated block to the right */
131 ext4_fsblk_t pright;
132 /* flags. see above EXT4_MB_HINT_* */
133 unsigned int flags;
134};
135
136/*
137 * Logical to physical block mapping, used by ext4_map_blocks()
138 *
139 * This structure is used to pass requests into ext4_map_blocks() as
140 * well as to store the information returned by ext4_map_blocks(). It
141 * takes less room on the stack than a struct buffer_head.
142 */
143#define EXT4_MAP_NEW (1 << BH_New)
144#define EXT4_MAP_MAPPED (1 << BH_Mapped)
145#define EXT4_MAP_UNWRITTEN (1 << BH_Unwritten)
146#define EXT4_MAP_BOUNDARY (1 << BH_Boundary)
147#define EXT4_MAP_UNINIT (1 << BH_Uninit)
148/* Sometimes (in the bigalloc case, from ext4_da_get_block_prep) the caller of
149 * ext4_map_blocks wants to know whether or not the underlying cluster has
150 * already been accounted for. EXT4_MAP_FROM_CLUSTER conveys to the caller that
151 * the requested mapping was from previously mapped (or delayed allocated)
152 * cluster. We use BH_AllocFromCluster only for this flag. BH_AllocFromCluster
153 * should never appear on buffer_head's state flags.
154 */
155#define EXT4_MAP_FROM_CLUSTER (1 << BH_AllocFromCluster)
156#define EXT4_MAP_FLAGS (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
157 EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
158 EXT4_MAP_UNINIT | EXT4_MAP_FROM_CLUSTER)
159
160struct ext4_map_blocks {
161 ext4_fsblk_t m_pblk;
162 ext4_lblk_t m_lblk;
163 unsigned int m_len;
164 unsigned int m_flags;
165};
166
167/*
168 * For delayed allocation tracking
169 */
170struct mpage_da_data {
171 struct inode *inode;
172 sector_t b_blocknr; /* start block number of extent */
173 size_t b_size; /* size of extent */
174 unsigned long b_state; /* state of the extent */
175 unsigned long first_page, next_page; /* extent of pages */
176 struct writeback_control *wbc;
177 int io_done;
178 int pages_written;
179 int retval;
180};
181
182/*
183 * Flags for ext4_io_end->flags
184 */
185#define EXT4_IO_END_UNWRITTEN 0x0001
186#define EXT4_IO_END_ERROR 0x0002
187#define EXT4_IO_END_QUEUED 0x0004
188#define EXT4_IO_END_DIRECT 0x0008
189#define EXT4_IO_END_IN_FSYNC 0x0010
190
191struct ext4_io_page {
192 struct page *p_page;
193 atomic_t p_count;
194};
195
196#define MAX_IO_PAGES 128
197
198/*
199 * For converting uninitialized extents on a work queue.
200 *
201 * 'page' is only used from the writepage() path; 'pages' is only used for
202 * buffered writes; they are used to keep page references until conversion
203 * takes place. For AIO/DIO, neither field is filled in.
204 */
205typedef struct ext4_io_end {
206 struct list_head list; /* per-file finished IO list */
207 struct inode *inode; /* file being written to */
208 unsigned int flag; /* unwritten or not */
209 struct page *page; /* for writepage() path */
210 loff_t offset; /* offset in the file */
211 ssize_t size; /* size of the extent */
212 struct work_struct work; /* data work queue */
213 struct kiocb *iocb; /* iocb struct for AIO */
214 int result; /* error value for AIO */
215 int num_io_pages; /* for writepages() */
216 struct ext4_io_page *pages[MAX_IO_PAGES]; /* for writepages() */
217} ext4_io_end_t;
218
219struct ext4_io_submit {
220 int io_op;
221 struct bio *io_bio;
222 ext4_io_end_t *io_end;
223 struct ext4_io_page *io_page;
224 sector_t io_next_block;
225};
226
227/*
228 * Special inodes numbers
229 */
230#define EXT4_BAD_INO 1 /* Bad blocks inode */
231#define EXT4_ROOT_INO 2 /* Root inode */
232#define EXT4_USR_QUOTA_INO 3 /* User quota inode */
233#define EXT4_GRP_QUOTA_INO 4 /* Group quota inode */
234#define EXT4_BOOT_LOADER_INO 5 /* Boot loader inode */
235#define EXT4_UNDEL_DIR_INO 6 /* Undelete directory inode */
236#define EXT4_RESIZE_INO 7 /* Reserved group descriptors inode */
237#define EXT4_JOURNAL_INO 8 /* Journal inode */
238
239/* First non-reserved inode for old ext4 filesystems */
240#define EXT4_GOOD_OLD_FIRST_INO 11
241
242/*
243 * Maximal count of links to a file
244 */
245#define EXT4_LINK_MAX 65000
246
247/*
248 * Macro-instructions used to manage several block sizes
249 */
250#define EXT4_MIN_BLOCK_SIZE 1024
251#define EXT4_MAX_BLOCK_SIZE 65536
252#define EXT4_MIN_BLOCK_LOG_SIZE 10
253#define EXT4_MAX_BLOCK_LOG_SIZE 16
254#ifdef __KERNEL__
255# define EXT4_BLOCK_SIZE(s) ((s)->s_blocksize)
256#else
257# define EXT4_BLOCK_SIZE(s) (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
258#endif
259#define EXT4_ADDR_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
260#define EXT4_CLUSTER_SIZE(s) (EXT4_BLOCK_SIZE(s) << \
261 EXT4_SB(s)->s_cluster_bits)
262#ifdef __KERNEL__
263# define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits)
264# define EXT4_CLUSTER_BITS(s) (EXT4_SB(s)->s_cluster_bits)
265#else
266# define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_log_block_size + 10)
267#endif
268#ifdef __KERNEL__
269#define EXT4_ADDR_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_addr_per_block_bits)
270#define EXT4_INODE_SIZE(s) (EXT4_SB(s)->s_inode_size)
271#define EXT4_FIRST_INO(s) (EXT4_SB(s)->s_first_ino)
272#else
273#define EXT4_INODE_SIZE(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
274 EXT4_GOOD_OLD_INODE_SIZE : \
275 (s)->s_inode_size)
276#define EXT4_FIRST_INO(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
277 EXT4_GOOD_OLD_FIRST_INO : \
278 (s)->s_first_ino)
279#endif
280#define EXT4_BLOCK_ALIGN(size, blkbits) ALIGN((size), (1 << (blkbits)))
281
282/* Translate a block number to a cluster number */
283#define EXT4_B2C(sbi, blk) ((blk) >> (sbi)->s_cluster_bits)
284/* Translate a cluster number to a block number */
285#define EXT4_C2B(sbi, cluster) ((cluster) << (sbi)->s_cluster_bits)
286/* Translate # of blks to # of clusters */
287#define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
288 (sbi)->s_cluster_bits)
289
290/*
291 * Structure of a blocks group descriptor
292 */
293struct ext4_group_desc
294{
295 __le32 bg_block_bitmap_lo; /* Blocks bitmap block */
296 __le32 bg_inode_bitmap_lo; /* Inodes bitmap block */
297 __le32 bg_inode_table_lo; /* Inodes table block */
298 __le16 bg_free_blocks_count_lo;/* Free blocks count */
299 __le16 bg_free_inodes_count_lo;/* Free inodes count */
300 __le16 bg_used_dirs_count_lo; /* Directories count */
301 __le16 bg_flags; /* EXT4_BG_flags (INODE_UNINIT, etc) */
302 __le32 bg_exclude_bitmap_lo; /* Exclude bitmap for snapshots */
303 __le16 bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
304 __le16 bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
305 __le16 bg_itable_unused_lo; /* Unused inodes count */
306 __le16 bg_checksum; /* crc16(sb_uuid+group+desc) */
307 __le32 bg_block_bitmap_hi; /* Blocks bitmap block MSB */
308 __le32 bg_inode_bitmap_hi; /* Inodes bitmap block MSB */
309 __le32 bg_inode_table_hi; /* Inodes table block MSB */
310 __le16 bg_free_blocks_count_hi;/* Free blocks count MSB */
311 __le16 bg_free_inodes_count_hi;/* Free inodes count MSB */
312 __le16 bg_used_dirs_count_hi; /* Directories count MSB */
313 __le16 bg_itable_unused_hi; /* Unused inodes count MSB */
314 __le32 bg_exclude_bitmap_hi; /* Exclude bitmap block MSB */
315 __le16 bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
316 __le16 bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
317 __u32 bg_reserved;
318};
319
320#define EXT4_BG_INODE_BITMAP_CSUM_HI_END \
321 (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
322 sizeof(__le16))
323#define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END \
324 (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
325 sizeof(__le16))
326
327/*
328 * Structure of a flex block group info
329 */
330
331struct flex_groups {
332 atomic_t free_inodes;
333 atomic_t free_clusters;
334 atomic_t used_dirs;
335};
336
337#define EXT4_BG_INODE_UNINIT 0x0001 /* Inode table/bitmap not in use */
338#define EXT4_BG_BLOCK_UNINIT 0x0002 /* Block bitmap not in use */
339#define EXT4_BG_INODE_ZEROED 0x0004 /* On-disk itable initialized to zero */
340
341/*
342 * Macro-instructions used to manage group descriptors
343 */
344#define EXT4_MIN_DESC_SIZE 32
345#define EXT4_MIN_DESC_SIZE_64BIT 64
346#define EXT4_MAX_DESC_SIZE EXT4_MIN_BLOCK_SIZE
347#define EXT4_DESC_SIZE(s) (EXT4_SB(s)->s_desc_size)
348#ifdef __KERNEL__
349# define EXT4_BLOCKS_PER_GROUP(s) (EXT4_SB(s)->s_blocks_per_group)
350# define EXT4_CLUSTERS_PER_GROUP(s) (EXT4_SB(s)->s_clusters_per_group)
351# define EXT4_DESC_PER_BLOCK(s) (EXT4_SB(s)->s_desc_per_block)
352# define EXT4_INODES_PER_GROUP(s) (EXT4_SB(s)->s_inodes_per_group)
353# define EXT4_DESC_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_desc_per_block_bits)
354#else
355# define EXT4_BLOCKS_PER_GROUP(s) ((s)->s_blocks_per_group)
356# define EXT4_DESC_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
357# define EXT4_INODES_PER_GROUP(s) ((s)->s_inodes_per_group)
358#endif
359
360/*
361 * Constants relative to the data blocks
362 */
363#define EXT4_NDIR_BLOCKS 12
364#define EXT4_IND_BLOCK EXT4_NDIR_BLOCKS
365#define EXT4_DIND_BLOCK (EXT4_IND_BLOCK + 1)
366#define EXT4_TIND_BLOCK (EXT4_DIND_BLOCK + 1)
367#define EXT4_N_BLOCKS (EXT4_TIND_BLOCK + 1)
368
369/*
370 * Inode flags
371 */
372#define EXT4_SECRM_FL 0x00000001 /* Secure deletion */
373#define EXT4_UNRM_FL 0x00000002 /* Undelete */
374#define EXT4_COMPR_FL 0x00000004 /* Compress file */
375#define EXT4_SYNC_FL 0x00000008 /* Synchronous updates */
376#define EXT4_IMMUTABLE_FL 0x00000010 /* Immutable file */
377#define EXT4_APPEND_FL 0x00000020 /* writes to file may only append */
378#define EXT4_NODUMP_FL 0x00000040 /* do not dump file */
379#define EXT4_NOATIME_FL 0x00000080 /* do not update atime */
380/* Reserved for compression usage... */
381#define EXT4_DIRTY_FL 0x00000100
382#define EXT4_COMPRBLK_FL 0x00000200 /* One or more compressed clusters */
383#define EXT4_NOCOMPR_FL 0x00000400 /* Don't compress */
384#define EXT4_ECOMPR_FL 0x00000800 /* Compression error */
385/* End compression flags --- maybe not all used */
386#define EXT4_INDEX_FL 0x00001000 /* hash-indexed directory */
387#define EXT4_IMAGIC_FL 0x00002000 /* AFS directory */
388#define EXT4_JOURNAL_DATA_FL 0x00004000 /* file data should be journaled */
389#define EXT4_NOTAIL_FL 0x00008000 /* file tail should not be merged */
390#define EXT4_DIRSYNC_FL 0x00010000 /* dirsync behaviour (directories only) */
391#define EXT4_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/
392#define EXT4_HUGE_FILE_FL 0x00040000 /* Set to each huge file */
393#define EXT4_EXTENTS_FL 0x00080000 /* Inode uses extents */
394#define EXT4_EA_INODE_FL 0x00200000 /* Inode used for large EA */
395#define EXT4_EOFBLOCKS_FL 0x00400000 /* Blocks allocated beyond EOF */
396#define EXT4_RESERVED_FL 0x80000000 /* reserved for ext4 lib */
397
398#define EXT4_FL_USER_VISIBLE 0x004BDFFF /* User visible flags */
399#define EXT4_FL_USER_MODIFIABLE 0x004B80FF /* User modifiable flags */
400
401/* Flags that should be inherited by new inodes from their parent. */
402#define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
403 EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
404 EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
405 EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
406
407/* Flags that are appropriate for regular files (all but dir-specific ones). */
408#define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
409
410/* Flags that are appropriate for non-directories/regular files. */
411#define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
412
413/* Mask out flags that are inappropriate for the given type of inode. */
414static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
415{
416 if (S_ISDIR(mode))
417 return flags;
418 else if (S_ISREG(mode))
419 return flags & EXT4_REG_FLMASK;
420 else
421 return flags & EXT4_OTHER_FLMASK;
422}
423
424/*
425 * Inode flags used for atomic set/get
426 */
427enum {
428 EXT4_INODE_SECRM = 0, /* Secure deletion */
429 EXT4_INODE_UNRM = 1, /* Undelete */
430 EXT4_INODE_COMPR = 2, /* Compress file */
431 EXT4_INODE_SYNC = 3, /* Synchronous updates */
432 EXT4_INODE_IMMUTABLE = 4, /* Immutable file */
433 EXT4_INODE_APPEND = 5, /* writes to file may only append */
434 EXT4_INODE_NODUMP = 6, /* do not dump file */
435 EXT4_INODE_NOATIME = 7, /* do not update atime */
436/* Reserved for compression usage... */
437 EXT4_INODE_DIRTY = 8,
438 EXT4_INODE_COMPRBLK = 9, /* One or more compressed clusters */
439 EXT4_INODE_NOCOMPR = 10, /* Don't compress */
440 EXT4_INODE_ECOMPR = 11, /* Compression error */
441/* End compression flags --- maybe not all used */
442 EXT4_INODE_INDEX = 12, /* hash-indexed directory */
443 EXT4_INODE_IMAGIC = 13, /* AFS directory */
444 EXT4_INODE_JOURNAL_DATA = 14, /* file data should be journaled */
445 EXT4_INODE_NOTAIL = 15, /* file tail should not be merged */
446 EXT4_INODE_DIRSYNC = 16, /* dirsync behaviour (directories only) */
447 EXT4_INODE_TOPDIR = 17, /* Top of directory hierarchies*/
448 EXT4_INODE_HUGE_FILE = 18, /* Set to each huge file */
449 EXT4_INODE_EXTENTS = 19, /* Inode uses extents */
450 EXT4_INODE_EA_INODE = 21, /* Inode used for large EA */
451 EXT4_INODE_EOFBLOCKS = 22, /* Blocks allocated beyond EOF */
452 EXT4_INODE_RESERVED = 31, /* reserved for ext4 lib */
453};
454
455#define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
456#define CHECK_FLAG_VALUE(FLAG) if (!TEST_FLAG_VALUE(FLAG)) { \
457 printk(KERN_EMERG "EXT4 flag fail: " #FLAG ": %d %d\n", \
458 EXT4_##FLAG##_FL, EXT4_INODE_##FLAG); BUG_ON(1); }
459
460/*
461 * Since it's pretty easy to mix up bit numbers and hex values, and we
462 * can't do a compile-time test for ENUM values, we use a run-time
463 * test to make sure that EXT4_XXX_FL is consistent with respect to
464 * EXT4_INODE_XXX. If all is well the printk and BUG_ON will all drop
465 * out so it won't cost any extra space in the compiled kernel image.
466 * But it's important that these values are the same, since we are
467 * using EXT4_INODE_XXX to test for the flag values, but EXT4_XX_FL
468 * must be consistent with the values of FS_XXX_FL defined in
469 * include/linux/fs.h and the on-disk values found in ext2, ext3, and
470 * ext4 filesystems, and of course the values defined in e2fsprogs.
471 *
472 * It's not paranoia if the Murphy's Law really *is* out to get you. :-)
473 */
474static inline void ext4_check_flag_values(void)
475{
476 CHECK_FLAG_VALUE(SECRM);
477 CHECK_FLAG_VALUE(UNRM);
478 CHECK_FLAG_VALUE(COMPR);
479 CHECK_FLAG_VALUE(SYNC);
480 CHECK_FLAG_VALUE(IMMUTABLE);
481 CHECK_FLAG_VALUE(APPEND);
482 CHECK_FLAG_VALUE(NODUMP);
483 CHECK_FLAG_VALUE(NOATIME);
484 CHECK_FLAG_VALUE(DIRTY);
485 CHECK_FLAG_VALUE(COMPRBLK);
486 CHECK_FLAG_VALUE(NOCOMPR);
487 CHECK_FLAG_VALUE(ECOMPR);
488 CHECK_FLAG_VALUE(INDEX);
489 CHECK_FLAG_VALUE(IMAGIC);
490 CHECK_FLAG_VALUE(JOURNAL_DATA);
491 CHECK_FLAG_VALUE(NOTAIL);
492 CHECK_FLAG_VALUE(DIRSYNC);
493 CHECK_FLAG_VALUE(TOPDIR);
494 CHECK_FLAG_VALUE(HUGE_FILE);
495 CHECK_FLAG_VALUE(EXTENTS);
496 CHECK_FLAG_VALUE(EA_INODE);
497 CHECK_FLAG_VALUE(EOFBLOCKS);
498 CHECK_FLAG_VALUE(RESERVED);
499}
500
501/* Used to pass group descriptor data when online resize is done */
502struct ext4_new_group_input {
503 __u32 group; /* Group number for this data */
504 __u64 block_bitmap; /* Absolute block number of block bitmap */
505 __u64 inode_bitmap; /* Absolute block number of inode bitmap */
506 __u64 inode_table; /* Absolute block number of inode table start */
507 __u32 blocks_count; /* Total number of blocks in this group */
508 __u16 reserved_blocks; /* Number of reserved blocks in this group */
509 __u16 unused;
510};
511
512#if defined(__KERNEL__) && defined(CONFIG_COMPAT)
513struct compat_ext4_new_group_input {
514 u32 group;
515 compat_u64 block_bitmap;
516 compat_u64 inode_bitmap;
517 compat_u64 inode_table;
518 u32 blocks_count;
519 u16 reserved_blocks;
520 u16 unused;
521};
522#endif
523
524/* The struct ext4_new_group_input in kernel space, with free_blocks_count */
525struct ext4_new_group_data {
526 __u32 group;
527 __u64 block_bitmap;
528 __u64 inode_bitmap;
529 __u64 inode_table;
530 __u32 blocks_count;
531 __u16 reserved_blocks;
532 __u16 unused;
533 __u32 free_blocks_count;
534};
535
536/* Indexes used to index group tables in ext4_new_group_data */
537enum {
538 BLOCK_BITMAP = 0, /* block bitmap */
539 INODE_BITMAP, /* inode bitmap */
540 INODE_TABLE, /* inode tables */
541 GROUP_TABLE_COUNT,
542};
543
544/*
545 * Flags used by ext4_map_blocks()
546 */
547 /* Allocate any needed blocks and/or convert an unitialized
548 extent to be an initialized ext4 */
549#define EXT4_GET_BLOCKS_CREATE 0x0001
550 /* Request the creation of an unitialized extent */
551#define EXT4_GET_BLOCKS_UNINIT_EXT 0x0002
552#define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT (EXT4_GET_BLOCKS_UNINIT_EXT|\
553 EXT4_GET_BLOCKS_CREATE)
554 /* Caller is from the delayed allocation writeout path,
555 so set the magic i_delalloc_reserve_flag after taking the
556 inode allocation semaphore for */
557#define EXT4_GET_BLOCKS_DELALLOC_RESERVE 0x0004
558 /* caller is from the direct IO path, request to creation of an
559 unitialized extents if not allocated, split the uninitialized
560 extent if blocks has been preallocated already*/
561#define EXT4_GET_BLOCKS_PRE_IO 0x0008
562#define EXT4_GET_BLOCKS_CONVERT 0x0010
563#define EXT4_GET_BLOCKS_IO_CREATE_EXT (EXT4_GET_BLOCKS_PRE_IO|\
564 EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
565 /* Convert extent to initialized after IO complete */
566#define EXT4_GET_BLOCKS_IO_CONVERT_EXT (EXT4_GET_BLOCKS_CONVERT|\
567 EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
568 /* Punch out blocks of an extent */
569#define EXT4_GET_BLOCKS_PUNCH_OUT_EXT 0x0020
570 /* Don't normalize allocation size (used for fallocate) */
571#define EXT4_GET_BLOCKS_NO_NORMALIZE 0x0040
572 /* Request will not result in inode size update (user for fallocate) */
573#define EXT4_GET_BLOCKS_KEEP_SIZE 0x0080
574
575/*
576 * Flags used by ext4_free_blocks
577 */
578#define EXT4_FREE_BLOCKS_METADATA 0x0001
579#define EXT4_FREE_BLOCKS_FORGET 0x0002
580#define EXT4_FREE_BLOCKS_VALIDATED 0x0004
581#define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
582#define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER 0x0010
583#define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER 0x0020
584
585/*
586 * Flags used by ext4_discard_partial_page_buffers
587 */
588#define EXT4_DISCARD_PARTIAL_PG_ZERO_UNMAPPED 0x0001
589
590/*
591 * ioctl commands
592 */
593#define EXT4_IOC_GETFLAGS FS_IOC_GETFLAGS
594#define EXT4_IOC_SETFLAGS FS_IOC_SETFLAGS
595#define EXT4_IOC_GETVERSION _IOR('f', 3, long)
596#define EXT4_IOC_SETVERSION _IOW('f', 4, long)
597#define EXT4_IOC_GETVERSION_OLD FS_IOC_GETVERSION
598#define EXT4_IOC_SETVERSION_OLD FS_IOC_SETVERSION
599#define EXT4_IOC_GETRSVSZ _IOR('f', 5, long)
600#define EXT4_IOC_SETRSVSZ _IOW('f', 6, long)
601#define EXT4_IOC_GROUP_EXTEND _IOW('f', 7, unsigned long)
602#define EXT4_IOC_GROUP_ADD _IOW('f', 8, struct ext4_new_group_input)
603#define EXT4_IOC_MIGRATE _IO('f', 9)
604 /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
605 /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
606#define EXT4_IOC_ALLOC_DA_BLKS _IO('f', 12)
607#define EXT4_IOC_MOVE_EXT _IOWR('f', 15, struct move_extent)
608#define EXT4_IOC_RESIZE_FS _IOW('f', 16, __u64)
609
610#if defined(__KERNEL__) && defined(CONFIG_COMPAT)
611/*
612 * ioctl commands in 32 bit emulation
613 */
614#define EXT4_IOC32_GETFLAGS FS_IOC32_GETFLAGS
615#define EXT4_IOC32_SETFLAGS FS_IOC32_SETFLAGS
616#define EXT4_IOC32_GETVERSION _IOR('f', 3, int)
617#define EXT4_IOC32_SETVERSION _IOW('f', 4, int)
618#define EXT4_IOC32_GETRSVSZ _IOR('f', 5, int)
619#define EXT4_IOC32_SETRSVSZ _IOW('f', 6, int)
620#define EXT4_IOC32_GROUP_EXTEND _IOW('f', 7, unsigned int)
621#define EXT4_IOC32_GROUP_ADD _IOW('f', 8, struct compat_ext4_new_group_input)
622#define EXT4_IOC32_GETVERSION_OLD FS_IOC32_GETVERSION
623#define EXT4_IOC32_SETVERSION_OLD FS_IOC32_SETVERSION
624#endif
625
626/* Max physical block we can address w/o extents */
627#define EXT4_MAX_BLOCK_FILE_PHYS 0xFFFFFFFF
628
629/*
630 * Structure of an inode on the disk
631 */
632struct ext4_inode {
633 __le16 i_mode; /* File mode */
634 __le16 i_uid; /* Low 16 bits of Owner Uid */
635 __le32 i_size_lo; /* Size in bytes */
636 __le32 i_atime; /* Access time */
637 __le32 i_ctime; /* Inode Change time */
638 __le32 i_mtime; /* Modification time */
639 __le32 i_dtime; /* Deletion Time */
640 __le16 i_gid; /* Low 16 bits of Group Id */
641 __le16 i_links_count; /* Links count */
642 __le32 i_blocks_lo; /* Blocks count */
643 __le32 i_flags; /* File flags */
644 union {
645 struct {
646 __le32 l_i_version;
647 } linux1;
648 struct {
649 __u32 h_i_translator;
650 } hurd1;
651 struct {
652 __u32 m_i_reserved1;
653 } masix1;
654 } osd1; /* OS dependent 1 */
655 __le32 i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
656 __le32 i_generation; /* File version (for NFS) */
657 __le32 i_file_acl_lo; /* File ACL */
658 __le32 i_size_high;
659 __le32 i_obso_faddr; /* Obsoleted fragment address */
660 union {
661 struct {
662 __le16 l_i_blocks_high; /* were l_i_reserved1 */
663 __le16 l_i_file_acl_high;
664 __le16 l_i_uid_high; /* these 2 fields */
665 __le16 l_i_gid_high; /* were reserved2[0] */
666 __le16 l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
667 __le16 l_i_reserved;
668 } linux2;
669 struct {
670 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */
671 __u16 h_i_mode_high;
672 __u16 h_i_uid_high;
673 __u16 h_i_gid_high;
674 __u32 h_i_author;
675 } hurd2;
676 struct {
677 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */
678 __le16 m_i_file_acl_high;
679 __u32 m_i_reserved2[2];
680 } masix2;
681 } osd2; /* OS dependent 2 */
682 __le16 i_extra_isize;
683 __le16 i_checksum_hi; /* crc32c(uuid+inum+inode) BE */
684 __le32 i_ctime_extra; /* extra Change time (nsec << 2 | epoch) */
685 __le32 i_mtime_extra; /* extra Modification time(nsec << 2 | epoch) */
686 __le32 i_atime_extra; /* extra Access time (nsec << 2 | epoch) */
687 __le32 i_crtime; /* File Creation time */
688 __le32 i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
689 __le32 i_version_hi; /* high 32 bits for 64-bit version */
690};
691
692struct move_extent {
693 __u32 reserved; /* should be zero */
694 __u32 donor_fd; /* donor file descriptor */
695 __u64 orig_start; /* logical start offset in block for orig */
696 __u64 donor_start; /* logical start offset in block for donor */
697 __u64 len; /* block length to be moved */
698 __u64 moved_len; /* moved block length */
699};
700
701#define EXT4_EPOCH_BITS 2
702#define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
703#define EXT4_NSEC_MASK (~0UL << EXT4_EPOCH_BITS)
704
705/*
706 * Extended fields will fit into an inode if the filesystem was formatted
707 * with large inodes (-I 256 or larger) and there are not currently any EAs
708 * consuming all of the available space. For new inodes we always reserve
709 * enough space for the kernel's known extended fields, but for inodes
710 * created with an old kernel this might not have been the case. None of
711 * the extended inode fields is critical for correct filesystem operation.
712 * This macro checks if a certain field fits in the inode. Note that
713 * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
714 */
715#define EXT4_FITS_IN_INODE(ext4_inode, einode, field) \
716 ((offsetof(typeof(*ext4_inode), field) + \
717 sizeof((ext4_inode)->field)) \
718 <= (EXT4_GOOD_OLD_INODE_SIZE + \
719 (einode)->i_extra_isize)) \
720
721static inline __le32 ext4_encode_extra_time(struct timespec *time)
722{
723 return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
724 (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
725 ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
726}
727
728static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
729{
730 if (sizeof(time->tv_sec) > 4)
731 time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
732 << 32;
733 time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
734}
735
736#define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode) \
737do { \
738 (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec); \
739 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \
740 (raw_inode)->xtime ## _extra = \
741 ext4_encode_extra_time(&(inode)->xtime); \
742} while (0)
743
744#define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode) \
745do { \
746 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \
747 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec); \
748 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \
749 (raw_inode)->xtime ## _extra = \
750 ext4_encode_extra_time(&(einode)->xtime); \
751} while (0)
752
753#define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode) \
754do { \
755 (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime); \
756 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \
757 ext4_decode_extra_time(&(inode)->xtime, \
758 raw_inode->xtime ## _extra); \
759 else \
760 (inode)->xtime.tv_nsec = 0; \
761} while (0)
762
763#define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode) \
764do { \
765 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \
766 (einode)->xtime.tv_sec = \
767 (signed)le32_to_cpu((raw_inode)->xtime); \
768 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \
769 ext4_decode_extra_time(&(einode)->xtime, \
770 raw_inode->xtime ## _extra); \
771 else \
772 (einode)->xtime.tv_nsec = 0; \
773} while (0)
774
775#define i_disk_version osd1.linux1.l_i_version
776
777#if defined(__KERNEL__) || defined(__linux__)
778#define i_reserved1 osd1.linux1.l_i_reserved1
779#define i_file_acl_high osd2.linux2.l_i_file_acl_high
780#define i_blocks_high osd2.linux2.l_i_blocks_high
781#define i_uid_low i_uid
782#define i_gid_low i_gid
783#define i_uid_high osd2.linux2.l_i_uid_high
784#define i_gid_high osd2.linux2.l_i_gid_high
785#define i_checksum_lo osd2.linux2.l_i_checksum_lo
786
787#elif defined(__GNU__)
788
789#define i_translator osd1.hurd1.h_i_translator
790#define i_uid_high osd2.hurd2.h_i_uid_high
791#define i_gid_high osd2.hurd2.h_i_gid_high
792#define i_author osd2.hurd2.h_i_author
793
794#elif defined(__masix__)
795
796#define i_reserved1 osd1.masix1.m_i_reserved1
797#define i_file_acl_high osd2.masix2.m_i_file_acl_high
798#define i_reserved2 osd2.masix2.m_i_reserved2
799
800#endif /* defined(__KERNEL__) || defined(__linux__) */
801
802/*
803 * storage for cached extent
804 * If ec_len == 0, then the cache is invalid.
805 * If ec_start == 0, then the cache represents a gap (null mapping)
806 */
807struct ext4_ext_cache {
808 ext4_fsblk_t ec_start;
809 ext4_lblk_t ec_block;
810 __u32 ec_len; /* must be 32bit to return holes */
811};
812
813/*
814 * fourth extended file system inode data in memory
815 */
816struct ext4_inode_info {
817 __le32 i_data[15]; /* unconverted */
818 __u32 i_dtime;
819 ext4_fsblk_t i_file_acl;
820
821 /*
822 * i_block_group is the number of the block group which contains
823 * this file's inode. Constant across the lifetime of the inode,
824 * it is ued for making block allocation decisions - we try to
825 * place a file's data blocks near its inode block, and new inodes
826 * near to their parent directory's inode.
827 */
828 ext4_group_t i_block_group;
829 ext4_lblk_t i_dir_start_lookup;
830#if (BITS_PER_LONG < 64)
831 unsigned long i_state_flags; /* Dynamic state flags */
832#endif
833 unsigned long i_flags;
834
835#ifdef CONFIG_EXT4_FS_XATTR
836 /*
837 * Extended attributes can be read independently of the main file
838 * data. Taking i_mutex even when reading would cause contention
839 * between readers of EAs and writers of regular file data, so
840 * instead we synchronize on xattr_sem when reading or changing
841 * EAs.
842 */
843 struct rw_semaphore xattr_sem;
844#endif
845
846 struct list_head i_orphan; /* unlinked but open inodes */
847
848 /*
849 * i_disksize keeps track of what the inode size is ON DISK, not
850 * in memory. During truncate, i_size is set to the new size by
851 * the VFS prior to calling ext4_truncate(), but the filesystem won't
852 * set i_disksize to 0 until the truncate is actually under way.
853 *
854 * The intent is that i_disksize always represents the blocks which
855 * are used by this file. This allows recovery to restart truncate
856 * on orphans if we crash during truncate. We actually write i_disksize
857 * into the on-disk inode when writing inodes out, instead of i_size.
858 *
859 * The only time when i_disksize and i_size may be different is when
860 * a truncate is in progress. The only things which change i_disksize
861 * are ext4_get_block (growth) and ext4_truncate (shrinkth).
862 */
863 loff_t i_disksize;
864
865 /*
866 * i_data_sem is for serialising ext4_truncate() against
867 * ext4_getblock(). In the 2.4 ext2 design, great chunks of inode's
868 * data tree are chopped off during truncate. We can't do that in
869 * ext4 because whenever we perform intermediate commits during
870 * truncate, the inode and all the metadata blocks *must* be in a
871 * consistent state which allows truncation of the orphans to restart
872 * during recovery. Hence we must fix the get_block-vs-truncate race
873 * by other means, so we have i_data_sem.
874 */
875 struct rw_semaphore i_data_sem;
876 struct inode vfs_inode;
877 struct jbd2_inode *jinode;
878
879 struct ext4_ext_cache i_cached_extent;
880 /*
881 * File creation time. Its function is same as that of
882 * struct timespec i_{a,c,m}time in the generic inode.
883 */
884 struct timespec i_crtime;
885
886 /* mballoc */
887 struct list_head i_prealloc_list;
888 spinlock_t i_prealloc_lock;
889
890 /* ialloc */
891 ext4_group_t i_last_alloc_group;
892
893 /* allocation reservation info for delalloc */
894 /* In case of bigalloc, these refer to clusters rather than blocks */
895 unsigned int i_reserved_data_blocks;
896 unsigned int i_reserved_meta_blocks;
897 unsigned int i_allocated_meta_blocks;
898 ext4_lblk_t i_da_metadata_calc_last_lblock;
899 int i_da_metadata_calc_len;
900
901 /* on-disk additional length */
902 __u16 i_extra_isize;
903
904#ifdef CONFIG_QUOTA
905 /* quota space reservation, managed internally by quota code */
906 qsize_t i_reserved_quota;
907#endif
908
909 /* completed IOs that might need unwritten extents handling */
910 struct list_head i_completed_io_list;
911 spinlock_t i_completed_io_lock;
912 atomic_t i_ioend_count; /* Number of outstanding io_end structs */
913 /* current io_end structure for async DIO write*/
914 ext4_io_end_t *cur_aio_dio;
915 atomic_t i_aiodio_unwritten; /* Nr. of inflight conversions pending */
916
917 spinlock_t i_block_reservation_lock;
918
919 /*
920 * Transactions that contain inode's metadata needed to complete
921 * fsync and fdatasync, respectively.
922 */
923 tid_t i_sync_tid;
924 tid_t i_datasync_tid;
925
926 /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
927 __u32 i_csum_seed;
928};
929
930/*
931 * File system states
932 */
933#define EXT4_VALID_FS 0x0001 /* Unmounted cleanly */
934#define EXT4_ERROR_FS 0x0002 /* Errors detected */
935#define EXT4_ORPHAN_FS 0x0004 /* Orphans being recovered */
936
937/*
938 * Misc. filesystem flags
939 */
940#define EXT2_FLAGS_SIGNED_HASH 0x0001 /* Signed dirhash in use */
941#define EXT2_FLAGS_UNSIGNED_HASH 0x0002 /* Unsigned dirhash in use */
942#define EXT2_FLAGS_TEST_FILESYS 0x0004 /* to test development code */
943
944/*
945 * Mount flags
946 */
947#define EXT4_MOUNT_GRPID 0x00004 /* Create files with directory's group */
948#define EXT4_MOUNT_DEBUG 0x00008 /* Some debugging messages */
949#define EXT4_MOUNT_ERRORS_CONT 0x00010 /* Continue on errors */
950#define EXT4_MOUNT_ERRORS_RO 0x00020 /* Remount fs ro on errors */
951#define EXT4_MOUNT_ERRORS_PANIC 0x00040 /* Panic on errors */
952#define EXT4_MOUNT_ERRORS_MASK 0x00070
953#define EXT4_MOUNT_MINIX_DF 0x00080 /* Mimics the Minix statfs */
954#define EXT4_MOUNT_NOLOAD 0x00100 /* Don't use existing journal*/
955#define EXT4_MOUNT_DATA_FLAGS 0x00C00 /* Mode for data writes: */
956#define EXT4_MOUNT_JOURNAL_DATA 0x00400 /* Write data to journal */
957#define EXT4_MOUNT_ORDERED_DATA 0x00800 /* Flush data before commit */
958#define EXT4_MOUNT_WRITEBACK_DATA 0x00C00 /* No data ordering */
959#define EXT4_MOUNT_UPDATE_JOURNAL 0x01000 /* Update the journal format */
960#define EXT4_MOUNT_NO_UID32 0x02000 /* Disable 32-bit UIDs */
961#define EXT4_MOUNT_XATTR_USER 0x04000 /* Extended user attributes */
962#define EXT4_MOUNT_POSIX_ACL 0x08000 /* POSIX Access Control Lists */
963#define EXT4_MOUNT_NO_AUTO_DA_ALLOC 0x10000 /* No auto delalloc mapping */
964#define EXT4_MOUNT_BARRIER 0x20000 /* Use block barriers */
965#define EXT4_MOUNT_QUOTA 0x80000 /* Some quota option set */
966#define EXT4_MOUNT_USRQUOTA 0x100000 /* "old" user quota */
967#define EXT4_MOUNT_GRPQUOTA 0x200000 /* "old" group quota */
968#define EXT4_MOUNT_DIOREAD_NOLOCK 0x400000 /* Enable support for dio read nolocking */
969#define EXT4_MOUNT_JOURNAL_CHECKSUM 0x800000 /* Journal checksums */
970#define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
971#define EXT4_MOUNT_MBLK_IO_SUBMIT 0x4000000 /* multi-block io submits */
972#define EXT4_MOUNT_DELALLOC 0x8000000 /* Delalloc support */
973#define EXT4_MOUNT_DATA_ERR_ABORT 0x10000000 /* Abort on file data write */
974#define EXT4_MOUNT_BLOCK_VALIDITY 0x20000000 /* Block validity checking */
975#define EXT4_MOUNT_DISCARD 0x40000000 /* Issue DISCARD requests */
976#define EXT4_MOUNT_INIT_INODE_TABLE 0x80000000 /* Initialize uninitialized itables */
977
978#define EXT4_MOUNT2_EXPLICIT_DELALLOC 0x00000001 /* User explicitly
979 specified delalloc */
980
981#define clear_opt(sb, opt) EXT4_SB(sb)->s_mount_opt &= \
982 ~EXT4_MOUNT_##opt
983#define set_opt(sb, opt) EXT4_SB(sb)->s_mount_opt |= \
984 EXT4_MOUNT_##opt
985#define test_opt(sb, opt) (EXT4_SB(sb)->s_mount_opt & \
986 EXT4_MOUNT_##opt)
987
988#define clear_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 &= \
989 ~EXT4_MOUNT2_##opt
990#define set_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 |= \
991 EXT4_MOUNT2_##opt
992#define test_opt2(sb, opt) (EXT4_SB(sb)->s_mount_opt2 & \
993 EXT4_MOUNT2_##opt)
994
995#define ext4_test_and_set_bit __test_and_set_bit_le
996#define ext4_set_bit __set_bit_le
997#define ext4_set_bit_atomic ext2_set_bit_atomic
998#define ext4_test_and_clear_bit __test_and_clear_bit_le
999#define ext4_clear_bit __clear_bit_le
1000#define ext4_clear_bit_atomic ext2_clear_bit_atomic
1001#define ext4_test_bit test_bit_le
1002#define ext4_find_next_zero_bit find_next_zero_bit_le
1003#define ext4_find_next_bit find_next_bit_le
1004
1005extern void ext4_set_bits(void *bm, int cur, int len);
1006
1007/*
1008 * Maximal mount counts between two filesystem checks
1009 */
1010#define EXT4_DFL_MAX_MNT_COUNT 20 /* Allow 20 mounts */
1011#define EXT4_DFL_CHECKINTERVAL 0 /* Don't use interval check */
1012
1013/*
1014 * Behaviour when detecting errors
1015 */
1016#define EXT4_ERRORS_CONTINUE 1 /* Continue execution */
1017#define EXT4_ERRORS_RO 2 /* Remount fs read-only */
1018#define EXT4_ERRORS_PANIC 3 /* Panic */
1019#define EXT4_ERRORS_DEFAULT EXT4_ERRORS_CONTINUE
1020
1021/* Metadata checksum algorithm codes */
1022#define EXT4_CRC32C_CHKSUM 1
1023
1024/*
1025 * Structure of the super block
1026 */
1027struct ext4_super_block {
1028/*00*/ __le32 s_inodes_count; /* Inodes count */
1029 __le32 s_blocks_count_lo; /* Blocks count */
1030 __le32 s_r_blocks_count_lo; /* Reserved blocks count */
1031 __le32 s_free_blocks_count_lo; /* Free blocks count */
1032/*10*/ __le32 s_free_inodes_count; /* Free inodes count */
1033 __le32 s_first_data_block; /* First Data Block */
1034 __le32 s_log_block_size; /* Block size */
1035 __le32 s_log_cluster_size; /* Allocation cluster size */
1036/*20*/ __le32 s_blocks_per_group; /* # Blocks per group */
1037 __le32 s_clusters_per_group; /* # Clusters per group */
1038 __le32 s_inodes_per_group; /* # Inodes per group */
1039 __le32 s_mtime; /* Mount time */
1040/*30*/ __le32 s_wtime; /* Write time */
1041 __le16 s_mnt_count; /* Mount count */
1042 __le16 s_max_mnt_count; /* Maximal mount count */
1043 __le16 s_magic; /* Magic signature */
1044 __le16 s_state; /* File system state */
1045 __le16 s_errors; /* Behaviour when detecting errors */
1046 __le16 s_minor_rev_level; /* minor revision level */
1047/*40*/ __le32 s_lastcheck; /* time of last check */
1048 __le32 s_checkinterval; /* max. time between checks */
1049 __le32 s_creator_os; /* OS */
1050 __le32 s_rev_level; /* Revision level */
1051/*50*/ __le16 s_def_resuid; /* Default uid for reserved blocks */
1052 __le16 s_def_resgid; /* Default gid for reserved blocks */
1053 /*
1054 * These fields are for EXT4_DYNAMIC_REV superblocks only.
1055 *
1056 * Note: the difference between the compatible feature set and
1057 * the incompatible feature set is that if there is a bit set
1058 * in the incompatible feature set that the kernel doesn't
1059 * know about, it should refuse to mount the filesystem.
1060 *
1061 * e2fsck's requirements are more strict; if it doesn't know
1062 * about a feature in either the compatible or incompatible
1063 * feature set, it must abort and not try to meddle with
1064 * things it doesn't understand...
1065 */
1066 __le32 s_first_ino; /* First non-reserved inode */
1067 __le16 s_inode_size; /* size of inode structure */
1068 __le16 s_block_group_nr; /* block group # of this superblock */
1069 __le32 s_feature_compat; /* compatible feature set */
1070/*60*/ __le32 s_feature_incompat; /* incompatible feature set */
1071 __le32 s_feature_ro_compat; /* readonly-compatible feature set */
1072/*68*/ __u8 s_uuid[16]; /* 128-bit uuid for volume */
1073/*78*/ char s_volume_name[16]; /* volume name */
1074/*88*/ char s_last_mounted[64]; /* directory where last mounted */
1075/*C8*/ __le32 s_algorithm_usage_bitmap; /* For compression */
1076 /*
1077 * Performance hints. Directory preallocation should only
1078 * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1079 */
1080 __u8 s_prealloc_blocks; /* Nr of blocks to try to preallocate*/
1081 __u8 s_prealloc_dir_blocks; /* Nr to preallocate for dirs */
1082 __le16 s_reserved_gdt_blocks; /* Per group desc for online growth */
1083 /*
1084 * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1085 */
1086/*D0*/ __u8 s_journal_uuid[16]; /* uuid of journal superblock */
1087/*E0*/ __le32 s_journal_inum; /* inode number of journal file */
1088 __le32 s_journal_dev; /* device number of journal file */
1089 __le32 s_last_orphan; /* start of list of inodes to delete */
1090 __le32 s_hash_seed[4]; /* HTREE hash seed */
1091 __u8 s_def_hash_version; /* Default hash version to use */
1092 __u8 s_jnl_backup_type;
1093 __le16 s_desc_size; /* size of group descriptor */
1094/*100*/ __le32 s_default_mount_opts;
1095 __le32 s_first_meta_bg; /* First metablock block group */
1096 __le32 s_mkfs_time; /* When the filesystem was created */
1097 __le32 s_jnl_blocks[17]; /* Backup of the journal inode */
1098 /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1099/*150*/ __le32 s_blocks_count_hi; /* Blocks count */
1100 __le32 s_r_blocks_count_hi; /* Reserved blocks count */
1101 __le32 s_free_blocks_count_hi; /* Free blocks count */
1102 __le16 s_min_extra_isize; /* All inodes have at least # bytes */
1103 __le16 s_want_extra_isize; /* New inodes should reserve # bytes */
1104 __le32 s_flags; /* Miscellaneous flags */
1105 __le16 s_raid_stride; /* RAID stride */
1106 __le16 s_mmp_update_interval; /* # seconds to wait in MMP checking */
1107 __le64 s_mmp_block; /* Block for multi-mount protection */
1108 __le32 s_raid_stripe_width; /* blocks on all data disks (N*stride)*/
1109 __u8 s_log_groups_per_flex; /* FLEX_BG group size */
1110 __u8 s_checksum_type; /* metadata checksum algorithm used */
1111 __le16 s_reserved_pad;
1112 __le64 s_kbytes_written; /* nr of lifetime kilobytes written */
1113 __le32 s_snapshot_inum; /* Inode number of active snapshot */
1114 __le32 s_snapshot_id; /* sequential ID of active snapshot */
1115 __le64 s_snapshot_r_blocks_count; /* reserved blocks for active
1116 snapshot's future use */
1117 __le32 s_snapshot_list; /* inode number of the head of the
1118 on-disk snapshot list */
1119#define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1120 __le32 s_error_count; /* number of fs errors */
1121 __le32 s_first_error_time; /* first time an error happened */
1122 __le32 s_first_error_ino; /* inode involved in first error */
1123 __le64 s_first_error_block; /* block involved of first error */
1124 __u8 s_first_error_func[32]; /* function where the error happened */
1125 __le32 s_first_error_line; /* line number where error happened */
1126 __le32 s_last_error_time; /* most recent time of an error */
1127 __le32 s_last_error_ino; /* inode involved in last error */
1128 __le32 s_last_error_line; /* line number where error happened */
1129 __le64 s_last_error_block; /* block involved of last error */
1130 __u8 s_last_error_func[32]; /* function where the error happened */
1131#define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1132 __u8 s_mount_opts[64];
1133 __le32 s_usr_quota_inum; /* inode for tracking user quota */
1134 __le32 s_grp_quota_inum; /* inode for tracking group quota */
1135 __le32 s_overhead_clusters; /* overhead blocks/clusters in fs */
1136 __le32 s_reserved[108]; /* Padding to the end of the block */
1137 __le32 s_checksum; /* crc32c(superblock) */
1138};
1139
1140#define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1141
1142#ifdef __KERNEL__
1143
1144/*
1145 * run-time mount flags
1146 */
1147#define EXT4_MF_MNTDIR_SAMPLED 0x0001
1148#define EXT4_MF_FS_ABORTED 0x0002 /* Fatal error detected */
1149
1150/*
1151 * fourth extended-fs super-block data in memory
1152 */
1153struct ext4_sb_info {
1154 unsigned long s_desc_size; /* Size of a group descriptor in bytes */
1155 unsigned long s_inodes_per_block;/* Number of inodes per block */
1156 unsigned long s_blocks_per_group;/* Number of blocks in a group */
1157 unsigned long s_clusters_per_group; /* Number of clusters in a group */
1158 unsigned long s_inodes_per_group;/* Number of inodes in a group */
1159 unsigned long s_itb_per_group; /* Number of inode table blocks per group */
1160 unsigned long s_gdb_count; /* Number of group descriptor blocks */
1161 unsigned long s_desc_per_block; /* Number of group descriptors per block */
1162 ext4_group_t s_groups_count; /* Number of groups in the fs */
1163 ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1164 unsigned long s_overhead; /* # of fs overhead clusters */
1165 unsigned int s_cluster_ratio; /* Number of blocks per cluster */
1166 unsigned int s_cluster_bits; /* log2 of s_cluster_ratio */
1167 loff_t s_bitmap_maxbytes; /* max bytes for bitmap files */
1168 struct buffer_head * s_sbh; /* Buffer containing the super block */
1169 struct ext4_super_block *s_es; /* Pointer to the super block in the buffer */
1170 struct buffer_head **s_group_desc;
1171 unsigned int s_mount_opt;
1172 unsigned int s_mount_opt2;
1173 unsigned int s_mount_flags;
1174 unsigned int s_def_mount_opt;
1175 ext4_fsblk_t s_sb_block;
1176 kuid_t s_resuid;
1177 kgid_t s_resgid;
1178 unsigned short s_mount_state;
1179 unsigned short s_pad;
1180 int s_addr_per_block_bits;
1181 int s_desc_per_block_bits;
1182 int s_inode_size;
1183 int s_first_ino;
1184 unsigned int s_inode_readahead_blks;
1185 unsigned int s_inode_goal;
1186 spinlock_t s_next_gen_lock;
1187 u32 s_next_generation;
1188 u32 s_hash_seed[4];
1189 int s_def_hash_version;
1190 int s_hash_unsigned; /* 3 if hash should be signed, 0 if not */
1191 struct percpu_counter s_freeclusters_counter;
1192 struct percpu_counter s_freeinodes_counter;
1193 struct percpu_counter s_dirs_counter;
1194 struct percpu_counter s_dirtyclusters_counter;
1195 struct blockgroup_lock *s_blockgroup_lock;
1196 struct proc_dir_entry *s_proc;
1197 struct kobject s_kobj;
1198 struct completion s_kobj_unregister;
1199 struct super_block *s_sb;
1200
1201 /* Journaling */
1202 struct journal_s *s_journal;
1203 struct list_head s_orphan;
1204 struct mutex s_orphan_lock;
1205 unsigned long s_resize_flags; /* Flags indicating if there
1206 is a resizer */
1207 unsigned long s_commit_interval;
1208 u32 s_max_batch_time;
1209 u32 s_min_batch_time;
1210 struct block_device *journal_bdev;
1211#ifdef CONFIG_QUOTA
1212 char *s_qf_names[MAXQUOTAS]; /* Names of quota files with journalled quota */
1213 int s_jquota_fmt; /* Format of quota to use */
1214#endif
1215 unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1216 struct rb_root system_blks;
1217
1218#ifdef EXTENTS_STATS
1219 /* ext4 extents stats */
1220 unsigned long s_ext_min;
1221 unsigned long s_ext_max;
1222 unsigned long s_depth_max;
1223 spinlock_t s_ext_stats_lock;
1224 unsigned long s_ext_blocks;
1225 unsigned long s_ext_extents;
1226#endif
1227
1228 /* for buddy allocator */
1229 struct ext4_group_info ***s_group_info;
1230 struct inode *s_buddy_cache;
1231 spinlock_t s_md_lock;
1232 unsigned short *s_mb_offsets;
1233 unsigned int *s_mb_maxs;
1234
1235 /* tunables */
1236 unsigned long s_stripe;
1237 unsigned int s_mb_stream_request;
1238 unsigned int s_mb_max_to_scan;
1239 unsigned int s_mb_min_to_scan;
1240 unsigned int s_mb_stats;
1241 unsigned int s_mb_order2_reqs;
1242 unsigned int s_mb_group_prealloc;
1243 unsigned int s_max_writeback_mb_bump;
1244 /* where last allocation was done - for stream allocation */
1245 unsigned long s_mb_last_group;
1246 unsigned long s_mb_last_start;
1247
1248 /* stats for buddy allocator */
1249 atomic_t s_bal_reqs; /* number of reqs with len > 1 */
1250 atomic_t s_bal_success; /* we found long enough chunks */
1251 atomic_t s_bal_allocated; /* in blocks */
1252 atomic_t s_bal_ex_scanned; /* total extents scanned */
1253 atomic_t s_bal_goals; /* goal hits */
1254 atomic_t s_bal_breaks; /* too long searches */
1255 atomic_t s_bal_2orders; /* 2^order hits */
1256 spinlock_t s_bal_lock;
1257 unsigned long s_mb_buddies_generated;
1258 unsigned long long s_mb_generation_time;
1259 atomic_t s_mb_lost_chunks;
1260 atomic_t s_mb_preallocated;
1261 atomic_t s_mb_discarded;
1262 atomic_t s_lock_busy;
1263
1264 /* locality groups */
1265 struct ext4_locality_group __percpu *s_locality_groups;
1266
1267 /* for write statistics */
1268 unsigned long s_sectors_written_start;
1269 u64 s_kbytes_written;
1270
1271 unsigned int s_log_groups_per_flex;
1272 struct flex_groups *s_flex_groups;
1273
1274 /* workqueue for dio unwritten */
1275 struct workqueue_struct *dio_unwritten_wq;
1276
1277 /* timer for periodic error stats printing */
1278 struct timer_list s_err_report;
1279
1280 /* Lazy inode table initialization info */
1281 struct ext4_li_request *s_li_request;
1282 /* Wait multiplier for lazy initialization thread */
1283 unsigned int s_li_wait_mult;
1284
1285 /* Kernel thread for multiple mount protection */
1286 struct task_struct *s_mmp_tsk;
1287
1288 /* record the last minlen when FITRIM is called. */
1289 atomic_t s_last_trim_minblks;
1290
1291 /* Reference to checksum algorithm driver via cryptoapi */
1292 struct crypto_shash *s_chksum_driver;
1293
1294 /* Precomputed FS UUID checksum for seeding other checksums */
1295 __u32 s_csum_seed;
1296};
1297
1298static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1299{
1300 return sb->s_fs_info;
1301}
1302static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1303{
1304 return container_of(inode, struct ext4_inode_info, vfs_inode);
1305}
1306
1307static inline struct timespec ext4_current_time(struct inode *inode)
1308{
1309 return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1310 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1311}
1312
1313static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1314{
1315 return ino == EXT4_ROOT_INO ||
1316 ino == EXT4_JOURNAL_INO ||
1317 ino == EXT4_RESIZE_INO ||
1318 (ino >= EXT4_FIRST_INO(sb) &&
1319 ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1320}
1321
1322static inline void ext4_set_io_unwritten_flag(struct inode *inode,
1323 struct ext4_io_end *io_end)
1324{
1325 if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
1326 io_end->flag |= EXT4_IO_END_UNWRITTEN;
1327 atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten);
1328 }
1329}
1330
1331/*
1332 * Inode dynamic state flags
1333 */
1334enum {
1335 EXT4_STATE_JDATA, /* journaled data exists */
1336 EXT4_STATE_NEW, /* inode is newly created */
1337 EXT4_STATE_XATTR, /* has in-inode xattrs */
1338 EXT4_STATE_NO_EXPAND, /* No space for expansion */
1339 EXT4_STATE_DA_ALLOC_CLOSE, /* Alloc DA blks on close */
1340 EXT4_STATE_EXT_MIGRATE, /* Inode is migrating */
1341 EXT4_STATE_DIO_UNWRITTEN, /* need convert on dio done*/
1342 EXT4_STATE_NEWENTRY, /* File just added to dir */
1343 EXT4_STATE_DELALLOC_RESERVED, /* blks already reserved for delalloc */
1344};
1345
1346#define EXT4_INODE_BIT_FNS(name, field, offset) \
1347static inline int ext4_test_inode_##name(struct inode *inode, int bit) \
1348{ \
1349 return test_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1350} \
1351static inline void ext4_set_inode_##name(struct inode *inode, int bit) \
1352{ \
1353 set_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1354} \
1355static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1356{ \
1357 clear_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1358}
1359
1360EXT4_INODE_BIT_FNS(flag, flags, 0)
1361#if (BITS_PER_LONG < 64)
1362EXT4_INODE_BIT_FNS(state, state_flags, 0)
1363
1364static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1365{
1366 (ei)->i_state_flags = 0;
1367}
1368#else
1369EXT4_INODE_BIT_FNS(state, flags, 32)
1370
1371static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1372{
1373 /* We depend on the fact that callers will set i_flags */
1374}
1375#endif
1376#else
1377/* Assume that user mode programs are passing in an ext4fs superblock, not
1378 * a kernel struct super_block. This will allow us to call the feature-test
1379 * macros from user land. */
1380#define EXT4_SB(sb) (sb)
1381#endif
1382
1383#define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1384
1385/*
1386 * Codes for operating systems
1387 */
1388#define EXT4_OS_LINUX 0
1389#define EXT4_OS_HURD 1
1390#define EXT4_OS_MASIX 2
1391#define EXT4_OS_FREEBSD 3
1392#define EXT4_OS_LITES 4
1393
1394/*
1395 * Revision levels
1396 */
1397#define EXT4_GOOD_OLD_REV 0 /* The good old (original) format */
1398#define EXT4_DYNAMIC_REV 1 /* V2 format w/ dynamic inode sizes */
1399
1400#define EXT4_CURRENT_REV EXT4_GOOD_OLD_REV
1401#define EXT4_MAX_SUPP_REV EXT4_DYNAMIC_REV
1402
1403#define EXT4_GOOD_OLD_INODE_SIZE 128
1404
1405/*
1406 * Feature set definitions
1407 */
1408
1409#define EXT4_HAS_COMPAT_FEATURE(sb,mask) \
1410 ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1411#define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask) \
1412 ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1413#define EXT4_HAS_INCOMPAT_FEATURE(sb,mask) \
1414 ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1415#define EXT4_SET_COMPAT_FEATURE(sb,mask) \
1416 EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1417#define EXT4_SET_RO_COMPAT_FEATURE(sb,mask) \
1418 EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1419#define EXT4_SET_INCOMPAT_FEATURE(sb,mask) \
1420 EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1421#define EXT4_CLEAR_COMPAT_FEATURE(sb,mask) \
1422 EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1423#define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask) \
1424 EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1425#define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask) \
1426 EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1427
1428#define EXT4_FEATURE_COMPAT_DIR_PREALLOC 0x0001
1429#define EXT4_FEATURE_COMPAT_IMAGIC_INODES 0x0002
1430#define EXT4_FEATURE_COMPAT_HAS_JOURNAL 0x0004
1431#define EXT4_FEATURE_COMPAT_EXT_ATTR 0x0008
1432#define EXT4_FEATURE_COMPAT_RESIZE_INODE 0x0010
1433#define EXT4_FEATURE_COMPAT_DIR_INDEX 0x0020
1434
1435#define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER 0x0001
1436#define EXT4_FEATURE_RO_COMPAT_LARGE_FILE 0x0002
1437#define EXT4_FEATURE_RO_COMPAT_BTREE_DIR 0x0004
1438#define EXT4_FEATURE_RO_COMPAT_HUGE_FILE 0x0008
1439#define EXT4_FEATURE_RO_COMPAT_GDT_CSUM 0x0010
1440#define EXT4_FEATURE_RO_COMPAT_DIR_NLINK 0x0020
1441#define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE 0x0040
1442#define EXT4_FEATURE_RO_COMPAT_QUOTA 0x0100
1443#define EXT4_FEATURE_RO_COMPAT_BIGALLOC 0x0200
1444/*
1445 * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM). When
1446 * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
1447 * all other data structures' checksums. However, the METADATA_CSUM and
1448 * GDT_CSUM bits are mutually exclusive.
1449 */
1450#define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM 0x0400
1451
1452#define EXT4_FEATURE_INCOMPAT_COMPRESSION 0x0001
1453#define EXT4_FEATURE_INCOMPAT_FILETYPE 0x0002
1454#define EXT4_FEATURE_INCOMPAT_RECOVER 0x0004 /* Needs recovery */
1455#define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 /* Journal device */
1456#define EXT4_FEATURE_INCOMPAT_META_BG 0x0010
1457#define EXT4_FEATURE_INCOMPAT_EXTENTS 0x0040 /* extents support */
1458#define EXT4_FEATURE_INCOMPAT_64BIT 0x0080
1459#define EXT4_FEATURE_INCOMPAT_MMP 0x0100
1460#define EXT4_FEATURE_INCOMPAT_FLEX_BG 0x0200
1461#define EXT4_FEATURE_INCOMPAT_EA_INODE 0x0400 /* EA in inode */
1462#define EXT4_FEATURE_INCOMPAT_DIRDATA 0x1000 /* data in dirent */
1463#define EXT4_FEATURE_INCOMPAT_BG_USE_META_CSUM 0x2000 /* use crc32c for bg */
1464#define EXT4_FEATURE_INCOMPAT_LARGEDIR 0x4000 /* >2GB or 3-lvl htree */
1465#define EXT4_FEATURE_INCOMPAT_INLINEDATA 0x8000 /* data in inode */
1466
1467#define EXT2_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR
1468#define EXT2_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1469 EXT4_FEATURE_INCOMPAT_META_BG)
1470#define EXT2_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1471 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1472 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1473
1474#define EXT3_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR
1475#define EXT3_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1476 EXT4_FEATURE_INCOMPAT_RECOVER| \
1477 EXT4_FEATURE_INCOMPAT_META_BG)
1478#define EXT3_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1479 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1480 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1481
1482#define EXT4_FEATURE_COMPAT_SUPP EXT2_FEATURE_COMPAT_EXT_ATTR
1483#define EXT4_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1484 EXT4_FEATURE_INCOMPAT_RECOVER| \
1485 EXT4_FEATURE_INCOMPAT_META_BG| \
1486 EXT4_FEATURE_INCOMPAT_EXTENTS| \
1487 EXT4_FEATURE_INCOMPAT_64BIT| \
1488 EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1489 EXT4_FEATURE_INCOMPAT_MMP)
1490#define EXT4_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1491 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1492 EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1493 EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1494 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1495 EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1496 EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1497 EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
1498 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)
1499
1500/*
1501 * Default values for user and/or group using reserved blocks
1502 */
1503#define EXT4_DEF_RESUID 0
1504#define EXT4_DEF_RESGID 0
1505
1506#define EXT4_DEF_INODE_READAHEAD_BLKS 32
1507
1508/*
1509 * Default mount options
1510 */
1511#define EXT4_DEFM_DEBUG 0x0001
1512#define EXT4_DEFM_BSDGROUPS 0x0002
1513#define EXT4_DEFM_XATTR_USER 0x0004
1514#define EXT4_DEFM_ACL 0x0008
1515#define EXT4_DEFM_UID16 0x0010
1516#define EXT4_DEFM_JMODE 0x0060
1517#define EXT4_DEFM_JMODE_DATA 0x0020
1518#define EXT4_DEFM_JMODE_ORDERED 0x0040
1519#define EXT4_DEFM_JMODE_WBACK 0x0060
1520#define EXT4_DEFM_NOBARRIER 0x0100
1521#define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1522#define EXT4_DEFM_DISCARD 0x0400
1523#define EXT4_DEFM_NODELALLOC 0x0800
1524
1525/*
1526 * Default journal batch times
1527 */
1528#define EXT4_DEF_MIN_BATCH_TIME 0
1529#define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1530
1531/*
1532 * Minimum number of groups in a flexgroup before we separate out
1533 * directories into the first block group of a flexgroup
1534 */
1535#define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1536
1537/*
1538 * Structure of a directory entry
1539 */
1540#define EXT4_NAME_LEN 255
1541
1542struct ext4_dir_entry {
1543 __le32 inode; /* Inode number */
1544 __le16 rec_len; /* Directory entry length */
1545 __le16 name_len; /* Name length */
1546 char name[EXT4_NAME_LEN]; /* File name */
1547};
1548
1549/*
1550 * The new version of the directory entry. Since EXT4 structures are
1551 * stored in intel byte order, and the name_len field could never be
1552 * bigger than 255 chars, it's safe to reclaim the extra byte for the
1553 * file_type field.
1554 */
1555struct ext4_dir_entry_2 {
1556 __le32 inode; /* Inode number */
1557 __le16 rec_len; /* Directory entry length */
1558 __u8 name_len; /* Name length */
1559 __u8 file_type;
1560 char name[EXT4_NAME_LEN]; /* File name */
1561};
1562
1563/*
1564 * This is a bogus directory entry at the end of each leaf block that
1565 * records checksums.
1566 */
1567struct ext4_dir_entry_tail {
1568 __le32 det_reserved_zero1; /* Pretend to be unused */
1569 __le16 det_rec_len; /* 12 */
1570 __u8 det_reserved_zero2; /* Zero name length */
1571 __u8 det_reserved_ft; /* 0xDE, fake file type */
1572 __le32 det_checksum; /* crc32c(uuid+inum+dirblock) */
1573};
1574
1575/*
1576 * Ext4 directory file types. Only the low 3 bits are used. The
1577 * other bits are reserved for now.
1578 */
1579#define EXT4_FT_UNKNOWN 0
1580#define EXT4_FT_REG_FILE 1
1581#define EXT4_FT_DIR 2
1582#define EXT4_FT_CHRDEV 3
1583#define EXT4_FT_BLKDEV 4
1584#define EXT4_FT_FIFO 5
1585#define EXT4_FT_SOCK 6
1586#define EXT4_FT_SYMLINK 7
1587
1588#define EXT4_FT_MAX 8
1589
1590#define EXT4_FT_DIR_CSUM 0xDE
1591
1592/*
1593 * EXT4_DIR_PAD defines the directory entries boundaries
1594 *
1595 * NOTE: It must be a multiple of 4
1596 */
1597#define EXT4_DIR_PAD 4
1598#define EXT4_DIR_ROUND (EXT4_DIR_PAD - 1)
1599#define EXT4_DIR_REC_LEN(name_len) (((name_len) + 8 + EXT4_DIR_ROUND) & \
1600 ~EXT4_DIR_ROUND)
1601#define EXT4_MAX_REC_LEN ((1<<16)-1)
1602
1603/*
1604 * If we ever get support for fs block sizes > page_size, we'll need
1605 * to remove the #if statements in the next two functions...
1606 */
1607static inline unsigned int
1608ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1609{
1610 unsigned len = le16_to_cpu(dlen);
1611
1612#if (PAGE_CACHE_SIZE >= 65536)
1613 if (len == EXT4_MAX_REC_LEN || len == 0)
1614 return blocksize;
1615 return (len & 65532) | ((len & 3) << 16);
1616#else
1617 return len;
1618#endif
1619}
1620
1621static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1622{
1623 if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1624 BUG();
1625#if (PAGE_CACHE_SIZE >= 65536)
1626 if (len < 65536)
1627 return cpu_to_le16(len);
1628 if (len == blocksize) {
1629 if (blocksize == 65536)
1630 return cpu_to_le16(EXT4_MAX_REC_LEN);
1631 else
1632 return cpu_to_le16(0);
1633 }
1634 return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1635#else
1636 return cpu_to_le16(len);
1637#endif
1638}
1639
1640/*
1641 * Hash Tree Directory indexing
1642 * (c) Daniel Phillips, 2001
1643 */
1644
1645#define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1646 EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1647 ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1648#define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1649#define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1650
1651/* Legal values for the dx_root hash_version field: */
1652
1653#define DX_HASH_LEGACY 0
1654#define DX_HASH_HALF_MD4 1
1655#define DX_HASH_TEA 2
1656#define DX_HASH_LEGACY_UNSIGNED 3
1657#define DX_HASH_HALF_MD4_UNSIGNED 4
1658#define DX_HASH_TEA_UNSIGNED 5
1659
1660static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
1661 const void *address, unsigned int length)
1662{
1663 struct {
1664 struct shash_desc shash;
1665 char ctx[crypto_shash_descsize(sbi->s_chksum_driver)];
1666 } desc;
1667 int err;
1668
1669 desc.shash.tfm = sbi->s_chksum_driver;
1670 desc.shash.flags = 0;
1671 *(u32 *)desc.ctx = crc;
1672
1673 err = crypto_shash_update(&desc.shash, address, length);
1674 BUG_ON(err);
1675
1676 return *(u32 *)desc.ctx;
1677}
1678
1679#ifdef __KERNEL__
1680
1681/* hash info structure used by the directory hash */
1682struct dx_hash_info
1683{
1684 u32 hash;
1685 u32 minor_hash;
1686 int hash_version;
1687 u32 *seed;
1688};
1689
1690
1691/* 32 and 64 bit signed EOF for dx directories */
1692#define EXT4_HTREE_EOF_32BIT ((1UL << (32 - 1)) - 1)
1693#define EXT4_HTREE_EOF_64BIT ((1ULL << (64 - 1)) - 1)
1694
1695
1696/*
1697 * Control parameters used by ext4_htree_next_block
1698 */
1699#define HASH_NB_ALWAYS 1
1700
1701
1702/*
1703 * Describe an inode's exact location on disk and in memory
1704 */
1705struct ext4_iloc
1706{
1707 struct buffer_head *bh;
1708 unsigned long offset;
1709 ext4_group_t block_group;
1710};
1711
1712static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1713{
1714 return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1715}
1716
1717/*
1718 * This structure is stuffed into the struct file's private_data field
1719 * for directories. It is where we put information so that we can do
1720 * readdir operations in hash tree order.
1721 */
1722struct dir_private_info {
1723 struct rb_root root;
1724 struct rb_node *curr_node;
1725 struct fname *extra_fname;
1726 loff_t last_pos;
1727 __u32 curr_hash;
1728 __u32 curr_minor_hash;
1729 __u32 next_hash;
1730};
1731
1732/* calculate the first block number of the group */
1733static inline ext4_fsblk_t
1734ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1735{
1736 return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1737 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1738}
1739
1740/*
1741 * Special error return code only used by dx_probe() and its callers.
1742 */
1743#define ERR_BAD_DX_DIR -75000
1744
1745void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1746 ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1747
1748/*
1749 * Timeout and state flag for lazy initialization inode thread.
1750 */
1751#define EXT4_DEF_LI_WAIT_MULT 10
1752#define EXT4_DEF_LI_MAX_START_DELAY 5
1753#define EXT4_LAZYINIT_QUIT 0x0001
1754#define EXT4_LAZYINIT_RUNNING 0x0002
1755
1756/*
1757 * Lazy inode table initialization info
1758 */
1759struct ext4_lazy_init {
1760 unsigned long li_state;
1761 struct list_head li_request_list;
1762 struct mutex li_list_mtx;
1763};
1764
1765struct ext4_li_request {
1766 struct super_block *lr_super;
1767 struct ext4_sb_info *lr_sbi;
1768 ext4_group_t lr_next_group;
1769 struct list_head lr_request;
1770 unsigned long lr_next_sched;
1771 unsigned long lr_timeout;
1772};
1773
1774struct ext4_features {
1775 struct kobject f_kobj;
1776 struct completion f_kobj_unregister;
1777};
1778
1779/*
1780 * This structure will be used for multiple mount protection. It will be
1781 * written into the block number saved in the s_mmp_block field in the
1782 * superblock. Programs that check MMP should assume that if
1783 * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1784 * to use the filesystem, regardless of how old the timestamp is.
1785 */
1786#define EXT4_MMP_MAGIC 0x004D4D50U /* ASCII for MMP */
1787#define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1788#define EXT4_MMP_SEQ_FSCK 0xE24D4D50U /* mmp_seq value when being fscked */
1789#define EXT4_MMP_SEQ_MAX 0xE24D4D4FU /* maximum valid mmp_seq value */
1790
1791struct mmp_struct {
1792 __le32 mmp_magic; /* Magic number for MMP */
1793 __le32 mmp_seq; /* Sequence no. updated periodically */
1794
1795 /*
1796 * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1797 * purposes and do not affect the correctness of the algorithm
1798 */
1799 __le64 mmp_time; /* Time last updated */
1800 char mmp_nodename[64]; /* Node which last updated MMP block */
1801 char mmp_bdevname[32]; /* Bdev which last updated MMP block */
1802
1803 /*
1804 * mmp_check_interval is used to verify if the MMP block has been
1805 * updated on the block device. The value is updated based on the
1806 * maximum time to write the MMP block during an update cycle.
1807 */
1808 __le16 mmp_check_interval;
1809
1810 __le16 mmp_pad1;
1811 __le32 mmp_pad2[226];
1812 __le32 mmp_checksum; /* crc32c(uuid+mmp_block) */
1813};
1814
1815/* arguments passed to the mmp thread */
1816struct mmpd_data {
1817 struct buffer_head *bh; /* bh from initial read_mmp_block() */
1818 struct super_block *sb; /* super block of the fs */
1819};
1820
1821/*
1822 * Check interval multiplier
1823 * The MMP block is written every update interval and initially checked every
1824 * update interval x the multiplier (the value is then adapted based on the
1825 * write latency). The reason is that writes can be delayed under load and we
1826 * don't want readers to incorrectly assume that the filesystem is no longer
1827 * in use.
1828 */
1829#define EXT4_MMP_CHECK_MULT 2UL
1830
1831/*
1832 * Minimum interval for MMP checking in seconds.
1833 */
1834#define EXT4_MMP_MIN_CHECK_INTERVAL 5UL
1835
1836/*
1837 * Maximum interval for MMP checking in seconds.
1838 */
1839#define EXT4_MMP_MAX_CHECK_INTERVAL 300UL
1840
1841/*
1842 * Function prototypes
1843 */
1844
1845/*
1846 * Ok, these declarations are also in <linux/kernel.h> but none of the
1847 * ext4 source programs needs to include it so they are duplicated here.
1848 */
1849# define NORET_TYPE /**/
1850# define ATTRIB_NORET __attribute__((noreturn))
1851# define NORET_AND noreturn,
1852
1853/* bitmap.c */
1854extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
1855void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1856 struct ext4_group_desc *gdp,
1857 struct buffer_head *bh, int sz);
1858int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1859 struct ext4_group_desc *gdp,
1860 struct buffer_head *bh, int sz);
1861void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1862 struct ext4_group_desc *gdp,
1863 struct buffer_head *bh, int sz);
1864int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1865 struct ext4_group_desc *gdp,
1866 struct buffer_head *bh, int sz);
1867
1868/* balloc.c */
1869extern void ext4_validate_block_bitmap(struct super_block *sb,
1870 struct ext4_group_desc *desc,
1871 unsigned int block_group,
1872 struct buffer_head *bh);
1873extern unsigned int ext4_block_group(struct super_block *sb,
1874 ext4_fsblk_t blocknr);
1875extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1876 ext4_fsblk_t blocknr);
1877extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1878extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1879 ext4_group_t group);
1880extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1881 ext4_fsblk_t goal,
1882 unsigned int flags,
1883 unsigned long *count,
1884 int *errp);
1885extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
1886 s64 nclusters, unsigned int flags);
1887extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
1888extern void ext4_check_blocks_bitmap(struct super_block *);
1889extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1890 ext4_group_t block_group,
1891 struct buffer_head ** bh);
1892extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1893
1894extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
1895 ext4_group_t block_group);
1896extern int ext4_wait_block_bitmap(struct super_block *sb,
1897 ext4_group_t block_group,
1898 struct buffer_head *bh);
1899extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1900 ext4_group_t block_group);
1901extern void ext4_init_block_bitmap(struct super_block *sb,
1902 struct buffer_head *bh,
1903 ext4_group_t group,
1904 struct ext4_group_desc *desc);
1905extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
1906 ext4_group_t block_group,
1907 struct ext4_group_desc *gdp);
1908extern unsigned ext4_num_overhead_clusters(struct super_block *sb,
1909 ext4_group_t block_group,
1910 struct ext4_group_desc *gdp);
1911ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
1912
1913/* dir.c */
1914extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1915 struct file *,
1916 struct ext4_dir_entry_2 *,
1917 struct buffer_head *, unsigned int);
1918#define ext4_check_dir_entry(dir, filp, de, bh, offset) \
1919 unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
1920 (de), (bh), (offset)))
1921extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1922 __u32 minor_hash,
1923 struct ext4_dir_entry_2 *dirent);
1924extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1925
1926/* fsync.c */
1927extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
1928extern int ext4_flush_completed_IO(struct inode *);
1929
1930/* hash.c */
1931extern int ext4fs_dirhash(const char *name, int len, struct
1932 dx_hash_info *hinfo);
1933
1934/* ialloc.c */
1935extern struct inode *ext4_new_inode(handle_t *, struct inode *, umode_t,
1936 const struct qstr *qstr, __u32 goal,
1937 uid_t *owner);
1938extern void ext4_free_inode(handle_t *, struct inode *);
1939extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1940extern unsigned long ext4_count_free_inodes(struct super_block *);
1941extern unsigned long ext4_count_dirs(struct super_block *);
1942extern void ext4_check_inodes_bitmap(struct super_block *);
1943extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1944extern int ext4_init_inode_table(struct super_block *sb,
1945 ext4_group_t group, int barrier);
1946extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
1947
1948/* mballoc.c */
1949extern long ext4_mb_stats;
1950extern long ext4_mb_max_to_scan;
1951extern int ext4_mb_init(struct super_block *);
1952extern int ext4_mb_release(struct super_block *);
1953extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1954 struct ext4_allocation_request *, int *);
1955extern int ext4_mb_reserve_blocks(struct super_block *, int);
1956extern void ext4_discard_preallocations(struct inode *);
1957extern int __init ext4_init_mballoc(void);
1958extern void ext4_exit_mballoc(void);
1959extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1960 struct buffer_head *bh, ext4_fsblk_t block,
1961 unsigned long count, int flags);
1962extern int ext4_mb_add_groupinfo(struct super_block *sb,
1963 ext4_group_t i, struct ext4_group_desc *desc);
1964extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
1965 ext4_fsblk_t block, unsigned long count);
1966extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
1967
1968/* inode.c */
1969struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1970 ext4_lblk_t, int, int *);
1971struct buffer_head *ext4_bread(handle_t *, struct inode *,
1972 ext4_lblk_t, int, int *);
1973int ext4_get_block(struct inode *inode, sector_t iblock,
1974 struct buffer_head *bh_result, int create);
1975
1976extern struct inode *ext4_iget(struct super_block *, unsigned long);
1977extern int ext4_write_inode(struct inode *, struct writeback_control *);
1978extern int ext4_setattr(struct dentry *, struct iattr *);
1979extern int ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1980 struct kstat *stat);
1981extern void ext4_evict_inode(struct inode *);
1982extern void ext4_clear_inode(struct inode *);
1983extern int ext4_sync_inode(handle_t *, struct inode *);
1984extern void ext4_dirty_inode(struct inode *, int);
1985extern int ext4_change_inode_journal_flag(struct inode *, int);
1986extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1987extern int ext4_can_truncate(struct inode *inode);
1988extern void ext4_truncate(struct inode *);
1989extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length);
1990extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
1991extern void ext4_set_inode_flags(struct inode *);
1992extern void ext4_get_inode_flags(struct ext4_inode_info *);
1993extern int ext4_alloc_da_blocks(struct inode *inode);
1994extern void ext4_set_aops(struct inode *inode);
1995extern int ext4_writepage_trans_blocks(struct inode *);
1996extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1997extern int ext4_discard_partial_page_buffers(handle_t *handle,
1998 struct address_space *mapping, loff_t from,
1999 loff_t length, int flags);
2000extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2001extern qsize_t *ext4_get_reserved_space(struct inode *inode);
2002extern void ext4_da_update_reserve_space(struct inode *inode,
2003 int used, int quota_claim);
2004
2005/* indirect.c */
2006extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
2007 struct ext4_map_blocks *map, int flags);
2008extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
2009 const struct iovec *iov, loff_t offset,
2010 unsigned long nr_segs);
2011extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
2012extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks, int chunk);
2013extern void ext4_ind_truncate(struct inode *inode);
2014
2015/* ioctl.c */
2016extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
2017extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
2018
2019/* migrate.c */
2020extern int ext4_ext_migrate(struct inode *);
2021
2022/* namei.c */
2023extern int ext4_dirent_csum_verify(struct inode *inode,
2024 struct ext4_dir_entry *dirent);
2025extern int ext4_orphan_add(handle_t *, struct inode *);
2026extern int ext4_orphan_del(handle_t *, struct inode *);
2027extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
2028 __u32 start_minor_hash, __u32 *next_hash);
2029
2030/* resize.c */
2031extern int ext4_group_add(struct super_block *sb,
2032 struct ext4_new_group_data *input);
2033extern int ext4_group_extend(struct super_block *sb,
2034 struct ext4_super_block *es,
2035 ext4_fsblk_t n_blocks_count);
2036extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
2037
2038/* super.c */
2039extern int ext4_calculate_overhead(struct super_block *sb);
2040extern int ext4_superblock_csum_verify(struct super_block *sb,
2041 struct ext4_super_block *es);
2042extern void ext4_superblock_csum_set(struct super_block *sb,
2043 struct ext4_super_block *es);
2044extern void *ext4_kvmalloc(size_t size, gfp_t flags);
2045extern void *ext4_kvzalloc(size_t size, gfp_t flags);
2046extern void ext4_kvfree(void *ptr);
2047extern __printf(4, 5)
2048void __ext4_error(struct super_block *, const char *, unsigned int,
2049 const char *, ...);
2050#define ext4_error(sb, message...) __ext4_error(sb, __func__, \
2051 __LINE__, ## message)
2052extern __printf(5, 6)
2053void ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
2054 const char *, ...);
2055extern __printf(5, 6)
2056void ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
2057 const char *, ...);
2058extern void __ext4_std_error(struct super_block *, const char *,
2059 unsigned int, int);
2060extern __printf(4, 5)
2061void __ext4_abort(struct super_block *, const char *, unsigned int,
2062 const char *, ...);
2063#define ext4_abort(sb, message...) __ext4_abort(sb, __func__, \
2064 __LINE__, ## message)
2065extern __printf(4, 5)
2066void __ext4_warning(struct super_block *, const char *, unsigned int,
2067 const char *, ...);
2068#define ext4_warning(sb, message...) __ext4_warning(sb, __func__, \
2069 __LINE__, ## message)
2070extern __printf(3, 4)
2071void ext4_msg(struct super_block *, const char *, const char *, ...);
2072extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
2073 const char *, unsigned int, const char *);
2074#define dump_mmp_msg(sb, mmp, msg) __dump_mmp_msg(sb, mmp, __func__, \
2075 __LINE__, msg)
2076extern __printf(7, 8)
2077void __ext4_grp_locked_error(const char *, unsigned int,
2078 struct super_block *, ext4_group_t,
2079 unsigned long, ext4_fsblk_t,
2080 const char *, ...);
2081#define ext4_grp_locked_error(sb, grp, message...) \
2082 __ext4_grp_locked_error(__func__, __LINE__, (sb), (grp), ## message)
2083extern void ext4_update_dynamic_rev(struct super_block *sb);
2084extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
2085 __u32 compat);
2086extern int ext4_update_rocompat_feature(handle_t *handle,
2087 struct super_block *sb, __u32 rocompat);
2088extern int ext4_update_incompat_feature(handle_t *handle,
2089 struct super_block *sb, __u32 incompat);
2090extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
2091 struct ext4_group_desc *bg);
2092extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
2093 struct ext4_group_desc *bg);
2094extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
2095 struct ext4_group_desc *bg);
2096extern __u32 ext4_free_group_clusters(struct super_block *sb,
2097 struct ext4_group_desc *bg);
2098extern __u32 ext4_free_inodes_count(struct super_block *sb,
2099 struct ext4_group_desc *bg);
2100extern __u32 ext4_used_dirs_count(struct super_block *sb,
2101 struct ext4_group_desc *bg);
2102extern __u32 ext4_itable_unused_count(struct super_block *sb,
2103 struct ext4_group_desc *bg);
2104extern void ext4_block_bitmap_set(struct super_block *sb,
2105 struct ext4_group_desc *bg, ext4_fsblk_t blk);
2106extern void ext4_inode_bitmap_set(struct super_block *sb,
2107 struct ext4_group_desc *bg, ext4_fsblk_t blk);
2108extern void ext4_inode_table_set(struct super_block *sb,
2109 struct ext4_group_desc *bg, ext4_fsblk_t blk);
2110extern void ext4_free_group_clusters_set(struct super_block *sb,
2111 struct ext4_group_desc *bg,
2112 __u32 count);
2113extern void ext4_free_inodes_set(struct super_block *sb,
2114 struct ext4_group_desc *bg, __u32 count);
2115extern void ext4_used_dirs_set(struct super_block *sb,
2116 struct ext4_group_desc *bg, __u32 count);
2117extern void ext4_itable_unused_set(struct super_block *sb,
2118 struct ext4_group_desc *bg, __u32 count);
2119extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
2120 struct ext4_group_desc *gdp);
2121extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
2122 struct ext4_group_desc *gdp);
2123
2124static inline int ext4_has_group_desc_csum(struct super_block *sb)
2125{
2126 return EXT4_HAS_RO_COMPAT_FEATURE(sb,
2127 EXT4_FEATURE_RO_COMPAT_GDT_CSUM |
2128 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM);
2129}
2130
2131static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2132{
2133 return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2134 le32_to_cpu(es->s_blocks_count_lo);
2135}
2136
2137static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2138{
2139 return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2140 le32_to_cpu(es->s_r_blocks_count_lo);
2141}
2142
2143static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2144{
2145 return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2146 le32_to_cpu(es->s_free_blocks_count_lo);
2147}
2148
2149static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2150 ext4_fsblk_t blk)
2151{
2152 es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2153 es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2154}
2155
2156static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2157 ext4_fsblk_t blk)
2158{
2159 es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2160 es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2161}
2162
2163static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2164 ext4_fsblk_t blk)
2165{
2166 es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2167 es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2168}
2169
2170static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2171{
2172 if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2173 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2174 le32_to_cpu(raw_inode->i_size_lo);
2175 else
2176 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2177}
2178
2179static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2180{
2181 raw_inode->i_size_lo = cpu_to_le32(i_size);
2182 raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2183}
2184
2185static inline
2186struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2187 ext4_group_t group)
2188{
2189 struct ext4_group_info ***grp_info;
2190 long indexv, indexh;
2191 grp_info = EXT4_SB(sb)->s_group_info;
2192 indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2193 indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2194 return grp_info[indexv][indexh];
2195}
2196
2197/*
2198 * Reading s_groups_count requires using smp_rmb() afterwards. See
2199 * the locking protocol documented in the comments of ext4_group_add()
2200 * in resize.c
2201 */
2202static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2203{
2204 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2205
2206 smp_rmb();
2207 return ngroups;
2208}
2209
2210static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2211 ext4_group_t block_group)
2212{
2213 return block_group >> sbi->s_log_groups_per_flex;
2214}
2215
2216static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2217{
2218 return 1 << sbi->s_log_groups_per_flex;
2219}
2220
2221#define ext4_std_error(sb, errno) \
2222do { \
2223 if ((errno)) \
2224 __ext4_std_error((sb), __func__, __LINE__, (errno)); \
2225} while (0)
2226
2227#ifdef CONFIG_SMP
2228/* Each CPU can accumulate percpu_counter_batch clusters in their local
2229 * counters. So we need to make sure we have free clusters more
2230 * than percpu_counter_batch * nr_cpu_ids. Also add a window of 4 times.
2231 */
2232#define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2233#else
2234#define EXT4_FREECLUSTERS_WATERMARK 0
2235#endif
2236
2237static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2238{
2239 /*
2240 * XXX: replace with spinlock if seen contended -bzzz
2241 */
2242 down_write(&EXT4_I(inode)->i_data_sem);
2243 if (newsize > EXT4_I(inode)->i_disksize)
2244 EXT4_I(inode)->i_disksize = newsize;
2245 up_write(&EXT4_I(inode)->i_data_sem);
2246 return ;
2247}
2248
2249struct ext4_group_info {
2250 unsigned long bb_state;
2251 struct rb_root bb_free_root;
2252 ext4_grpblk_t bb_first_free; /* first free block */
2253 ext4_grpblk_t bb_free; /* total free blocks */
2254 ext4_grpblk_t bb_fragments; /* nr of freespace fragments */
2255 ext4_grpblk_t bb_largest_free_order;/* order of largest frag in BG */
2256 struct list_head bb_prealloc_list;
2257#ifdef DOUBLE_CHECK
2258 void *bb_bitmap;
2259#endif
2260 struct rw_semaphore alloc_sem;
2261 ext4_grpblk_t bb_counters[]; /* Nr of free power-of-two-block
2262 * regions, index is order.
2263 * bb_counters[3] = 5 means
2264 * 5 free 8-block regions. */
2265};
2266
2267#define EXT4_GROUP_INFO_NEED_INIT_BIT 0
2268#define EXT4_GROUP_INFO_WAS_TRIMMED_BIT 1
2269
2270#define EXT4_MB_GRP_NEED_INIT(grp) \
2271 (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2272
2273#define EXT4_MB_GRP_WAS_TRIMMED(grp) \
2274 (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2275#define EXT4_MB_GRP_SET_TRIMMED(grp) \
2276 (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2277#define EXT4_MB_GRP_CLEAR_TRIMMED(grp) \
2278 (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2279
2280#define EXT4_MAX_CONTENTION 8
2281#define EXT4_CONTENTION_THRESHOLD 2
2282
2283static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2284 ext4_group_t group)
2285{
2286 return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2287}
2288
2289/*
2290 * Returns true if the filesystem is busy enough that attempts to
2291 * access the block group locks has run into contention.
2292 */
2293static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2294{
2295 return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2296}
2297
2298static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2299{
2300 spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2301 if (spin_trylock(lock))
2302 /*
2303 * We're able to grab the lock right away, so drop the
2304 * lock contention counter.
2305 */
2306 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2307 else {
2308 /*
2309 * The lock is busy, so bump the contention counter,
2310 * and then wait on the spin lock.
2311 */
2312 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2313 EXT4_MAX_CONTENTION);
2314 spin_lock(lock);
2315 }
2316}
2317
2318static inline void ext4_unlock_group(struct super_block *sb,
2319 ext4_group_t group)
2320{
2321 spin_unlock(ext4_group_lock_ptr(sb, group));
2322}
2323
2324static inline void ext4_mark_super_dirty(struct super_block *sb)
2325{
2326 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
2327
2328 ext4_superblock_csum_set(sb, es);
2329 if (EXT4_SB(sb)->s_journal == NULL)
2330 sb->s_dirt =1;
2331}
2332
2333/*
2334 * Block validity checking
2335 */
2336#define ext4_check_indirect_blockref(inode, bh) \
2337 ext4_check_blockref(__func__, __LINE__, inode, \
2338 (__le32 *)(bh)->b_data, \
2339 EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2340
2341#define ext4_ind_check_inode(inode) \
2342 ext4_check_blockref(__func__, __LINE__, inode, \
2343 EXT4_I(inode)->i_data, \
2344 EXT4_NDIR_BLOCKS)
2345
2346/*
2347 * Inodes and files operations
2348 */
2349
2350/* dir.c */
2351extern const struct file_operations ext4_dir_operations;
2352
2353/* file.c */
2354extern const struct inode_operations ext4_file_inode_operations;
2355extern const struct file_operations ext4_file_operations;
2356extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2357
2358/* namei.c */
2359extern const struct inode_operations ext4_dir_inode_operations;
2360extern const struct inode_operations ext4_special_inode_operations;
2361extern struct dentry *ext4_get_parent(struct dentry *child);
2362
2363/* symlink.c */
2364extern const struct inode_operations ext4_symlink_inode_operations;
2365extern const struct inode_operations ext4_fast_symlink_inode_operations;
2366
2367/* block_validity */
2368extern void ext4_release_system_zone(struct super_block *sb);
2369extern int ext4_setup_system_zone(struct super_block *sb);
2370extern int __init ext4_init_system_zone(void);
2371extern void ext4_exit_system_zone(void);
2372extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2373 ext4_fsblk_t start_blk,
2374 unsigned int count);
2375extern int ext4_check_blockref(const char *, unsigned int,
2376 struct inode *, __le32 *, unsigned int);
2377
2378/* extents.c */
2379extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2380extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2381extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
2382 int chunk);
2383extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2384 struct ext4_map_blocks *map, int flags);
2385extern void ext4_ext_truncate(struct inode *);
2386extern int ext4_ext_punch_hole(struct file *file, loff_t offset,
2387 loff_t length);
2388extern void ext4_ext_init(struct super_block *);
2389extern void ext4_ext_release(struct super_block *);
2390extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2391 loff_t len);
2392extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
2393 ssize_t len);
2394extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2395 struct ext4_map_blocks *map, int flags);
2396extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2397 __u64 start, __u64 len);
2398/* move_extent.c */
2399extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2400 __u64 start_orig, __u64 start_donor,
2401 __u64 len, __u64 *moved_len);
2402
2403/* page-io.c */
2404extern int __init ext4_init_pageio(void);
2405extern void ext4_exit_pageio(void);
2406extern void ext4_ioend_wait(struct inode *);
2407extern void ext4_free_io_end(ext4_io_end_t *io);
2408extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2409extern int ext4_end_io_nolock(ext4_io_end_t *io);
2410extern void ext4_io_submit(struct ext4_io_submit *io);
2411extern int ext4_bio_write_page(struct ext4_io_submit *io,
2412 struct page *page,
2413 int len,
2414 struct writeback_control *wbc);
2415
2416/* mmp.c */
2417extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2418extern void ext4_mmp_csum_set(struct super_block *sb, struct mmp_struct *mmp);
2419extern int ext4_mmp_csum_verify(struct super_block *sb,
2420 struct mmp_struct *mmp);
2421
2422/* BH_Uninit flag: blocks are allocated but uninitialized on disk */
2423enum ext4_state_bits {
2424 BH_Uninit /* blocks are allocated but uninitialized on disk */
2425 = BH_JBDPrivateStart,
2426 BH_AllocFromCluster, /* allocated blocks were part of already
2427 * allocated cluster. Note that this flag will
2428 * never, ever appear in a buffer_head's state
2429 * flag. See EXT4_MAP_FROM_CLUSTER to see where
2430 * this is used. */
2431 BH_Da_Mapped, /* Delayed allocated block that now has a mapping. This
2432 * flag is set when ext4_map_blocks is called on a
2433 * delayed allocated block to get its real mapping. */
2434};
2435
2436BUFFER_FNS(Uninit, uninit)
2437TAS_BUFFER_FNS(Uninit, uninit)
2438BUFFER_FNS(Da_Mapped, da_mapped)
2439
2440/*
2441 * Add new method to test wether block and inode bitmaps are properly
2442 * initialized. With uninit_bg reading the block from disk is not enough
2443 * to mark the bitmap uptodate. We need to also zero-out the bitmap
2444 */
2445#define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2446
2447static inline int bitmap_uptodate(struct buffer_head *bh)
2448{
2449 return (buffer_uptodate(bh) &&
2450 test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2451}
2452static inline void set_bitmap_uptodate(struct buffer_head *bh)
2453{
2454 set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2455}
2456
2457#define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2458
2459/* For ioend & aio unwritten conversion wait queues */
2460#define EXT4_WQ_HASH_SZ 37
2461#define ext4_ioend_wq(v) (&ext4__ioend_wq[((unsigned long)(v)) %\
2462 EXT4_WQ_HASH_SZ])
2463#define ext4_aio_mutex(v) (&ext4__aio_mutex[((unsigned long)(v)) %\
2464 EXT4_WQ_HASH_SZ])
2465extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2466extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2467
2468#define EXT4_RESIZING 0
2469extern int ext4_resize_begin(struct super_block *sb);
2470extern void ext4_resize_end(struct super_block *sb);
2471
2472#endif /* __KERNEL__ */
2473
2474#include "ext4_extents.h"
2475
2476#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 cr = 0 */
217#define EXT4_MB_CR_POWER2_ALIGNED_OPTIMIZED 0x8000
218/* Avg fragment size rb tree lookup succeeded at least once for cr = 1 */
219#define EXT4_MB_CR_GOAL_LEN_FAST_OPTIMIZED 0x00010000
220/* Avg fragment size rb tree lookup succeeded at least once for cr = 1.5 */
221#define EXT4_MB_CR_BEST_AVAIL_LEN_OPTIMIZED 0x00020000
222
223struct ext4_allocation_request {
224 /* target inode for block we're allocating */
225 struct inode *inode;
226 /* how many blocks we want to allocate */
227 unsigned int len;
228 /* logical block in target inode */
229 ext4_lblk_t logical;
230 /* the closest logical allocated block to the left */
231 ext4_lblk_t lleft;
232 /* the closest logical allocated block to the right */
233 ext4_lblk_t lright;
234 /* phys. target (a hint) */
235 ext4_fsblk_t goal;
236 /* phys. block for the closest logical allocated block to the left */
237 ext4_fsblk_t pleft;
238 /* phys. block for the closest logical allocated block to the right */
239 ext4_fsblk_t pright;
240 /* flags. see above EXT4_MB_HINT_* */
241 unsigned int flags;
242};
243
244/*
245 * Logical to physical block mapping, used by ext4_map_blocks()
246 *
247 * This structure is used to pass requests into ext4_map_blocks() as
248 * well as to store the information returned by ext4_map_blocks(). It
249 * takes less room on the stack than a struct buffer_head.
250 */
251#define EXT4_MAP_NEW BIT(BH_New)
252#define EXT4_MAP_MAPPED BIT(BH_Mapped)
253#define EXT4_MAP_UNWRITTEN BIT(BH_Unwritten)
254#define EXT4_MAP_BOUNDARY BIT(BH_Boundary)
255#define EXT4_MAP_DELAYED BIT(BH_Delay)
256#define EXT4_MAP_FLAGS (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
257 EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
258 EXT4_MAP_DELAYED)
259
260struct ext4_map_blocks {
261 ext4_fsblk_t m_pblk;
262 ext4_lblk_t m_lblk;
263 unsigned int m_len;
264 unsigned int m_flags;
265};
266
267/*
268 * Block validity checking, system zone rbtree.
269 */
270struct ext4_system_blocks {
271 struct rb_root root;
272 struct rcu_head rcu;
273};
274
275/*
276 * Flags for ext4_io_end->flags
277 */
278#define EXT4_IO_END_UNWRITTEN 0x0001
279
280struct ext4_io_end_vec {
281 struct list_head list; /* list of io_end_vec */
282 loff_t offset; /* offset in the file */
283 ssize_t size; /* size of the extent */
284};
285
286/*
287 * For converting unwritten extents on a work queue. 'handle' is used for
288 * buffered writeback.
289 */
290typedef struct ext4_io_end {
291 struct list_head list; /* per-file finished IO list */
292 handle_t *handle; /* handle reserved for extent
293 * conversion */
294 struct inode *inode; /* file being written to */
295 struct bio *bio; /* Linked list of completed
296 * bios covering the extent */
297 unsigned int flag; /* unwritten or not */
298 refcount_t count; /* reference counter */
299 struct list_head list_vec; /* list of ext4_io_end_vec */
300} ext4_io_end_t;
301
302struct ext4_io_submit {
303 struct writeback_control *io_wbc;
304 struct bio *io_bio;
305 ext4_io_end_t *io_end;
306 sector_t io_next_block;
307};
308
309/*
310 * Special inodes numbers
311 */
312#define EXT4_BAD_INO 1 /* Bad blocks inode */
313#define EXT4_ROOT_INO 2 /* Root inode */
314#define EXT4_USR_QUOTA_INO 3 /* User quota inode */
315#define EXT4_GRP_QUOTA_INO 4 /* Group quota inode */
316#define EXT4_BOOT_LOADER_INO 5 /* Boot loader inode */
317#define EXT4_UNDEL_DIR_INO 6 /* Undelete directory inode */
318#define EXT4_RESIZE_INO 7 /* Reserved group descriptors inode */
319#define EXT4_JOURNAL_INO 8 /* Journal inode */
320
321/* First non-reserved inode for old ext4 filesystems */
322#define EXT4_GOOD_OLD_FIRST_INO 11
323
324/*
325 * Maximal count of links to a file
326 */
327#define EXT4_LINK_MAX 65000
328
329/*
330 * Macro-instructions used to manage several block sizes
331 */
332#define EXT4_MIN_BLOCK_SIZE 1024
333#define EXT4_MAX_BLOCK_SIZE 65536
334#define EXT4_MIN_BLOCK_LOG_SIZE 10
335#define EXT4_MAX_BLOCK_LOG_SIZE 16
336#define EXT4_MAX_CLUSTER_LOG_SIZE 30
337#ifdef __KERNEL__
338# define EXT4_BLOCK_SIZE(s) ((s)->s_blocksize)
339#else
340# define EXT4_BLOCK_SIZE(s) (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
341#endif
342#define EXT4_ADDR_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
343#define EXT4_CLUSTER_SIZE(s) (EXT4_BLOCK_SIZE(s) << \
344 EXT4_SB(s)->s_cluster_bits)
345#ifdef __KERNEL__
346# define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits)
347# define EXT4_CLUSTER_BITS(s) (EXT4_SB(s)->s_cluster_bits)
348#else
349# define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_log_block_size + 10)
350#endif
351#ifdef __KERNEL__
352#define EXT4_ADDR_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_addr_per_block_bits)
353#define EXT4_INODE_SIZE(s) (EXT4_SB(s)->s_inode_size)
354#define EXT4_FIRST_INO(s) (EXT4_SB(s)->s_first_ino)
355#else
356#define EXT4_INODE_SIZE(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
357 EXT4_GOOD_OLD_INODE_SIZE : \
358 (s)->s_inode_size)
359#define EXT4_FIRST_INO(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
360 EXT4_GOOD_OLD_FIRST_INO : \
361 (s)->s_first_ino)
362#endif
363#define EXT4_BLOCK_ALIGN(size, blkbits) ALIGN((size), (1 << (blkbits)))
364#define EXT4_MAX_BLOCKS(size, offset, blkbits) \
365 ((EXT4_BLOCK_ALIGN(size + offset, blkbits) >> blkbits) - (offset >> \
366 blkbits))
367
368/* Translate a block number to a cluster number */
369#define EXT4_B2C(sbi, blk) ((blk) >> (sbi)->s_cluster_bits)
370/* Translate a cluster number to a block number */
371#define EXT4_C2B(sbi, cluster) ((cluster) << (sbi)->s_cluster_bits)
372/* Translate # of blks to # of clusters */
373#define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
374 (sbi)->s_cluster_bits)
375/* Mask out the low bits to get the starting block of the cluster */
376#define EXT4_PBLK_CMASK(s, pblk) ((pblk) & \
377 ~((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
378#define EXT4_LBLK_CMASK(s, lblk) ((lblk) & \
379 ~((ext4_lblk_t) (s)->s_cluster_ratio - 1))
380/* Fill in the low bits to get the last block of the cluster */
381#define EXT4_LBLK_CFILL(sbi, lblk) ((lblk) | \
382 ((ext4_lblk_t) (sbi)->s_cluster_ratio - 1))
383/* Get the cluster offset */
384#define EXT4_PBLK_COFF(s, pblk) ((pblk) & \
385 ((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
386#define EXT4_LBLK_COFF(s, lblk) ((lblk) & \
387 ((ext4_lblk_t) (s)->s_cluster_ratio - 1))
388
389/*
390 * Structure of a blocks group descriptor
391 */
392struct ext4_group_desc
393{
394 __le32 bg_block_bitmap_lo; /* Blocks bitmap block */
395 __le32 bg_inode_bitmap_lo; /* Inodes bitmap block */
396 __le32 bg_inode_table_lo; /* Inodes table block */
397 __le16 bg_free_blocks_count_lo;/* Free blocks count */
398 __le16 bg_free_inodes_count_lo;/* Free inodes count */
399 __le16 bg_used_dirs_count_lo; /* Directories count */
400 __le16 bg_flags; /* EXT4_BG_flags (INODE_UNINIT, etc) */
401 __le32 bg_exclude_bitmap_lo; /* Exclude bitmap for snapshots */
402 __le16 bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
403 __le16 bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
404 __le16 bg_itable_unused_lo; /* Unused inodes count */
405 __le16 bg_checksum; /* crc16(sb_uuid+group+desc) */
406 __le32 bg_block_bitmap_hi; /* Blocks bitmap block MSB */
407 __le32 bg_inode_bitmap_hi; /* Inodes bitmap block MSB */
408 __le32 bg_inode_table_hi; /* Inodes table block MSB */
409 __le16 bg_free_blocks_count_hi;/* Free blocks count MSB */
410 __le16 bg_free_inodes_count_hi;/* Free inodes count MSB */
411 __le16 bg_used_dirs_count_hi; /* Directories count MSB */
412 __le16 bg_itable_unused_hi; /* Unused inodes count MSB */
413 __le32 bg_exclude_bitmap_hi; /* Exclude bitmap block MSB */
414 __le16 bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
415 __le16 bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
416 __u32 bg_reserved;
417};
418
419#define EXT4_BG_INODE_BITMAP_CSUM_HI_END \
420 (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
421 sizeof(__le16))
422#define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END \
423 (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
424 sizeof(__le16))
425
426/*
427 * Structure of a flex block group info
428 */
429
430struct flex_groups {
431 atomic64_t free_clusters;
432 atomic_t free_inodes;
433 atomic_t used_dirs;
434};
435
436#define EXT4_BG_INODE_UNINIT 0x0001 /* Inode table/bitmap not in use */
437#define EXT4_BG_BLOCK_UNINIT 0x0002 /* Block bitmap not in use */
438#define EXT4_BG_INODE_ZEROED 0x0004 /* On-disk itable initialized to zero */
439
440/*
441 * Macro-instructions used to manage group descriptors
442 */
443#define EXT4_MIN_DESC_SIZE 32
444#define EXT4_MIN_DESC_SIZE_64BIT 64
445#define EXT4_MAX_DESC_SIZE EXT4_MIN_BLOCK_SIZE
446#define EXT4_DESC_SIZE(s) (EXT4_SB(s)->s_desc_size)
447#ifdef __KERNEL__
448# define EXT4_BLOCKS_PER_GROUP(s) (EXT4_SB(s)->s_blocks_per_group)
449# define EXT4_CLUSTERS_PER_GROUP(s) (EXT4_SB(s)->s_clusters_per_group)
450# define EXT4_DESC_PER_BLOCK(s) (EXT4_SB(s)->s_desc_per_block)
451# define EXT4_INODES_PER_GROUP(s) (EXT4_SB(s)->s_inodes_per_group)
452# define EXT4_DESC_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_desc_per_block_bits)
453#else
454# define EXT4_BLOCKS_PER_GROUP(s) ((s)->s_blocks_per_group)
455# define EXT4_DESC_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
456# define EXT4_INODES_PER_GROUP(s) ((s)->s_inodes_per_group)
457#endif
458
459/*
460 * Constants relative to the data blocks
461 */
462#define EXT4_NDIR_BLOCKS 12
463#define EXT4_IND_BLOCK EXT4_NDIR_BLOCKS
464#define EXT4_DIND_BLOCK (EXT4_IND_BLOCK + 1)
465#define EXT4_TIND_BLOCK (EXT4_DIND_BLOCK + 1)
466#define EXT4_N_BLOCKS (EXT4_TIND_BLOCK + 1)
467
468/*
469 * Inode flags
470 */
471#define EXT4_SECRM_FL 0x00000001 /* Secure deletion */
472#define EXT4_UNRM_FL 0x00000002 /* Undelete */
473#define EXT4_COMPR_FL 0x00000004 /* Compress file */
474#define EXT4_SYNC_FL 0x00000008 /* Synchronous updates */
475#define EXT4_IMMUTABLE_FL 0x00000010 /* Immutable file */
476#define EXT4_APPEND_FL 0x00000020 /* writes to file may only append */
477#define EXT4_NODUMP_FL 0x00000040 /* do not dump file */
478#define EXT4_NOATIME_FL 0x00000080 /* do not update atime */
479/* Reserved for compression usage... */
480#define EXT4_DIRTY_FL 0x00000100
481#define EXT4_COMPRBLK_FL 0x00000200 /* One or more compressed clusters */
482#define EXT4_NOCOMPR_FL 0x00000400 /* Don't compress */
483 /* nb: was previously EXT2_ECOMPR_FL */
484#define EXT4_ENCRYPT_FL 0x00000800 /* encrypted file */
485/* End compression flags --- maybe not all used */
486#define EXT4_INDEX_FL 0x00001000 /* hash-indexed directory */
487#define EXT4_IMAGIC_FL 0x00002000 /* AFS directory */
488#define EXT4_JOURNAL_DATA_FL 0x00004000 /* file data should be journaled */
489#define EXT4_NOTAIL_FL 0x00008000 /* file tail should not be merged */
490#define EXT4_DIRSYNC_FL 0x00010000 /* dirsync behaviour (directories only) */
491#define EXT4_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/
492#define EXT4_HUGE_FILE_FL 0x00040000 /* Set to each huge file */
493#define EXT4_EXTENTS_FL 0x00080000 /* Inode uses extents */
494#define EXT4_VERITY_FL 0x00100000 /* Verity protected inode */
495#define EXT4_EA_INODE_FL 0x00200000 /* Inode used for large EA */
496/* 0x00400000 was formerly EXT4_EOFBLOCKS_FL */
497
498#define EXT4_DAX_FL 0x02000000 /* Inode is DAX */
499
500#define EXT4_INLINE_DATA_FL 0x10000000 /* Inode has inline data. */
501#define EXT4_PROJINHERIT_FL 0x20000000 /* Create with parents projid */
502#define EXT4_CASEFOLD_FL 0x40000000 /* Casefolded directory */
503#define EXT4_RESERVED_FL 0x80000000 /* reserved for ext4 lib */
504
505/* User modifiable flags */
506#define EXT4_FL_USER_MODIFIABLE (EXT4_SECRM_FL | \
507 EXT4_UNRM_FL | \
508 EXT4_COMPR_FL | \
509 EXT4_SYNC_FL | \
510 EXT4_IMMUTABLE_FL | \
511 EXT4_APPEND_FL | \
512 EXT4_NODUMP_FL | \
513 EXT4_NOATIME_FL | \
514 EXT4_JOURNAL_DATA_FL | \
515 EXT4_NOTAIL_FL | \
516 EXT4_DIRSYNC_FL | \
517 EXT4_TOPDIR_FL | \
518 EXT4_EXTENTS_FL | \
519 0x00400000 /* EXT4_EOFBLOCKS_FL */ | \
520 EXT4_DAX_FL | \
521 EXT4_PROJINHERIT_FL | \
522 EXT4_CASEFOLD_FL)
523
524/* User visible flags */
525#define EXT4_FL_USER_VISIBLE (EXT4_FL_USER_MODIFIABLE | \
526 EXT4_DIRTY_FL | \
527 EXT4_COMPRBLK_FL | \
528 EXT4_NOCOMPR_FL | \
529 EXT4_ENCRYPT_FL | \
530 EXT4_INDEX_FL | \
531 EXT4_VERITY_FL | \
532 EXT4_INLINE_DATA_FL)
533
534/* Flags that should be inherited by new inodes from their parent. */
535#define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
536 EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
537 EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
538 EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL |\
539 EXT4_PROJINHERIT_FL | EXT4_CASEFOLD_FL |\
540 EXT4_DAX_FL)
541
542/* Flags that are appropriate for regular files (all but dir-specific ones). */
543#define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL | EXT4_CASEFOLD_FL |\
544 EXT4_PROJINHERIT_FL))
545
546/* Flags that are appropriate for non-directories/regular files. */
547#define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
548
549/* The only flags that should be swapped */
550#define EXT4_FL_SHOULD_SWAP (EXT4_HUGE_FILE_FL | EXT4_EXTENTS_FL)
551
552/* Flags which are mutually exclusive to DAX */
553#define EXT4_DAX_MUT_EXCL (EXT4_VERITY_FL | EXT4_ENCRYPT_FL |\
554 EXT4_JOURNAL_DATA_FL | EXT4_INLINE_DATA_FL)
555
556/* Mask out flags that are inappropriate for the given type of inode. */
557static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
558{
559 if (S_ISDIR(mode))
560 return flags;
561 else if (S_ISREG(mode))
562 return flags & EXT4_REG_FLMASK;
563 else
564 return flags & EXT4_OTHER_FLMASK;
565}
566
567/*
568 * Inode flags used for atomic set/get
569 */
570enum {
571 EXT4_INODE_SECRM = 0, /* Secure deletion */
572 EXT4_INODE_UNRM = 1, /* Undelete */
573 EXT4_INODE_COMPR = 2, /* Compress file */
574 EXT4_INODE_SYNC = 3, /* Synchronous updates */
575 EXT4_INODE_IMMUTABLE = 4, /* Immutable file */
576 EXT4_INODE_APPEND = 5, /* writes to file may only append */
577 EXT4_INODE_NODUMP = 6, /* do not dump file */
578 EXT4_INODE_NOATIME = 7, /* do not update atime */
579/* Reserved for compression usage... */
580 EXT4_INODE_DIRTY = 8,
581 EXT4_INODE_COMPRBLK = 9, /* One or more compressed clusters */
582 EXT4_INODE_NOCOMPR = 10, /* Don't compress */
583 EXT4_INODE_ENCRYPT = 11, /* Encrypted file */
584/* End compression flags --- maybe not all used */
585 EXT4_INODE_INDEX = 12, /* hash-indexed directory */
586 EXT4_INODE_IMAGIC = 13, /* AFS directory */
587 EXT4_INODE_JOURNAL_DATA = 14, /* file data should be journaled */
588 EXT4_INODE_NOTAIL = 15, /* file tail should not be merged */
589 EXT4_INODE_DIRSYNC = 16, /* dirsync behaviour (directories only) */
590 EXT4_INODE_TOPDIR = 17, /* Top of directory hierarchies*/
591 EXT4_INODE_HUGE_FILE = 18, /* Set to each huge file */
592 EXT4_INODE_EXTENTS = 19, /* Inode uses extents */
593 EXT4_INODE_VERITY = 20, /* Verity protected inode */
594 EXT4_INODE_EA_INODE = 21, /* Inode used for large EA */
595/* 22 was formerly EXT4_INODE_EOFBLOCKS */
596 EXT4_INODE_DAX = 25, /* Inode is DAX */
597 EXT4_INODE_INLINE_DATA = 28, /* Data in inode. */
598 EXT4_INODE_PROJINHERIT = 29, /* Create with parents projid */
599 EXT4_INODE_CASEFOLD = 30, /* Casefolded directory */
600 EXT4_INODE_RESERVED = 31, /* reserved for ext4 lib */
601};
602
603/*
604 * Since it's pretty easy to mix up bit numbers and hex values, we use a
605 * build-time check to make sure that EXT4_XXX_FL is consistent with respect to
606 * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost
607 * any extra space in the compiled kernel image, otherwise, the build will fail.
608 * It's important that these values are the same, since we are using
609 * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent
610 * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk
611 * values found in ext2, ext3 and ext4 filesystems, and of course the values
612 * defined in e2fsprogs.
613 *
614 * It's not paranoia if the Murphy's Law really *is* out to get you. :-)
615 */
616#define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1U << EXT4_INODE_##FLAG))
617#define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG))
618
619static inline void ext4_check_flag_values(void)
620{
621 CHECK_FLAG_VALUE(SECRM);
622 CHECK_FLAG_VALUE(UNRM);
623 CHECK_FLAG_VALUE(COMPR);
624 CHECK_FLAG_VALUE(SYNC);
625 CHECK_FLAG_VALUE(IMMUTABLE);
626 CHECK_FLAG_VALUE(APPEND);
627 CHECK_FLAG_VALUE(NODUMP);
628 CHECK_FLAG_VALUE(NOATIME);
629 CHECK_FLAG_VALUE(DIRTY);
630 CHECK_FLAG_VALUE(COMPRBLK);
631 CHECK_FLAG_VALUE(NOCOMPR);
632 CHECK_FLAG_VALUE(ENCRYPT);
633 CHECK_FLAG_VALUE(INDEX);
634 CHECK_FLAG_VALUE(IMAGIC);
635 CHECK_FLAG_VALUE(JOURNAL_DATA);
636 CHECK_FLAG_VALUE(NOTAIL);
637 CHECK_FLAG_VALUE(DIRSYNC);
638 CHECK_FLAG_VALUE(TOPDIR);
639 CHECK_FLAG_VALUE(HUGE_FILE);
640 CHECK_FLAG_VALUE(EXTENTS);
641 CHECK_FLAG_VALUE(VERITY);
642 CHECK_FLAG_VALUE(EA_INODE);
643 CHECK_FLAG_VALUE(INLINE_DATA);
644 CHECK_FLAG_VALUE(PROJINHERIT);
645 CHECK_FLAG_VALUE(CASEFOLD);
646 CHECK_FLAG_VALUE(RESERVED);
647}
648
649#if defined(__KERNEL__) && defined(CONFIG_COMPAT)
650struct compat_ext4_new_group_input {
651 u32 group;
652 compat_u64 block_bitmap;
653 compat_u64 inode_bitmap;
654 compat_u64 inode_table;
655 u32 blocks_count;
656 u16 reserved_blocks;
657 u16 unused;
658};
659#endif
660
661/* The struct ext4_new_group_input in kernel space, with free_blocks_count */
662struct ext4_new_group_data {
663 __u32 group;
664 __u64 block_bitmap;
665 __u64 inode_bitmap;
666 __u64 inode_table;
667 __u32 blocks_count;
668 __u16 reserved_blocks;
669 __u16 mdata_blocks;
670 __u32 free_clusters_count;
671};
672
673/* Indexes used to index group tables in ext4_new_group_data */
674enum {
675 BLOCK_BITMAP = 0, /* block bitmap */
676 INODE_BITMAP, /* inode bitmap */
677 INODE_TABLE, /* inode tables */
678 GROUP_TABLE_COUNT,
679};
680
681/*
682 * Flags used by ext4_map_blocks()
683 */
684 /* Allocate any needed blocks and/or convert an unwritten
685 extent to be an initialized ext4 */
686#define EXT4_GET_BLOCKS_CREATE 0x0001
687 /* Request the creation of an unwritten extent */
688#define EXT4_GET_BLOCKS_UNWRIT_EXT 0x0002
689#define EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT (EXT4_GET_BLOCKS_UNWRIT_EXT|\
690 EXT4_GET_BLOCKS_CREATE)
691 /* Caller is from the delayed allocation writeout path
692 * finally doing the actual allocation of delayed blocks */
693#define EXT4_GET_BLOCKS_DELALLOC_RESERVE 0x0004
694 /* caller is from the direct IO path, request to creation of an
695 unwritten extents if not allocated, split the unwritten
696 extent if blocks has been preallocated already*/
697#define EXT4_GET_BLOCKS_PRE_IO 0x0008
698#define EXT4_GET_BLOCKS_CONVERT 0x0010
699#define EXT4_GET_BLOCKS_IO_CREATE_EXT (EXT4_GET_BLOCKS_PRE_IO|\
700 EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
701 /* Convert extent to initialized after IO complete */
702#define EXT4_GET_BLOCKS_IO_CONVERT_EXT (EXT4_GET_BLOCKS_CONVERT|\
703 EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
704 /* Eventual metadata allocation (due to growing extent tree)
705 * should not fail, so try to use reserved blocks for that.*/
706#define EXT4_GET_BLOCKS_METADATA_NOFAIL 0x0020
707 /* Don't normalize allocation size (used for fallocate) */
708#define EXT4_GET_BLOCKS_NO_NORMALIZE 0x0040
709 /* Convert written extents to unwritten */
710#define EXT4_GET_BLOCKS_CONVERT_UNWRITTEN 0x0100
711 /* Write zeros to newly created written extents */
712#define EXT4_GET_BLOCKS_ZERO 0x0200
713#define EXT4_GET_BLOCKS_CREATE_ZERO (EXT4_GET_BLOCKS_CREATE |\
714 EXT4_GET_BLOCKS_ZERO)
715 /* Caller will submit data before dropping transaction handle. This
716 * allows jbd2 to avoid submitting data before commit. */
717#define EXT4_GET_BLOCKS_IO_SUBMIT 0x0400
718 /* Caller is in the atomic contex, find extent if it has been cached */
719#define EXT4_GET_BLOCKS_CACHED_NOWAIT 0x0800
720
721/*
722 * The bit position of these flags must not overlap with any of the
723 * EXT4_GET_BLOCKS_*. They are used by ext4_find_extent(),
724 * read_extent_tree_block(), ext4_split_extent_at(),
725 * ext4_ext_insert_extent(), and ext4_ext_create_new_leaf().
726 * EXT4_EX_NOCACHE is used to indicate that the we shouldn't be
727 * caching the extents when reading from the extent tree while a
728 * truncate or punch hole operation is in progress.
729 */
730#define EXT4_EX_NOCACHE 0x40000000
731#define EXT4_EX_FORCE_CACHE 0x20000000
732#define EXT4_EX_NOFAIL 0x10000000
733
734/*
735 * Flags used by ext4_free_blocks
736 */
737#define EXT4_FREE_BLOCKS_METADATA 0x0001
738#define EXT4_FREE_BLOCKS_FORGET 0x0002
739#define EXT4_FREE_BLOCKS_VALIDATED 0x0004
740#define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
741#define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER 0x0010
742#define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER 0x0020
743#define EXT4_FREE_BLOCKS_RERESERVE_CLUSTER 0x0040
744
745#if defined(__KERNEL__) && defined(CONFIG_COMPAT)
746/*
747 * ioctl commands in 32 bit emulation
748 */
749#define EXT4_IOC32_GETVERSION _IOR('f', 3, int)
750#define EXT4_IOC32_SETVERSION _IOW('f', 4, int)
751#define EXT4_IOC32_GETRSVSZ _IOR('f', 5, int)
752#define EXT4_IOC32_SETRSVSZ _IOW('f', 6, int)
753#define EXT4_IOC32_GROUP_EXTEND _IOW('f', 7, unsigned int)
754#define EXT4_IOC32_GROUP_ADD _IOW('f', 8, struct compat_ext4_new_group_input)
755#define EXT4_IOC32_GETVERSION_OLD FS_IOC32_GETVERSION
756#define EXT4_IOC32_SETVERSION_OLD FS_IOC32_SETVERSION
757#endif
758
759/* Max physical block we can address w/o extents */
760#define EXT4_MAX_BLOCK_FILE_PHYS 0xFFFFFFFF
761
762/* Max logical block we can support */
763#define EXT4_MAX_LOGICAL_BLOCK 0xFFFFFFFE
764
765/*
766 * Structure of an inode on the disk
767 */
768struct ext4_inode {
769 __le16 i_mode; /* File mode */
770 __le16 i_uid; /* Low 16 bits of Owner Uid */
771 __le32 i_size_lo; /* Size in bytes */
772 __le32 i_atime; /* Access time */
773 __le32 i_ctime; /* Inode Change time */
774 __le32 i_mtime; /* Modification time */
775 __le32 i_dtime; /* Deletion Time */
776 __le16 i_gid; /* Low 16 bits of Group Id */
777 __le16 i_links_count; /* Links count */
778 __le32 i_blocks_lo; /* Blocks count */
779 __le32 i_flags; /* File flags */
780 union {
781 struct {
782 __le32 l_i_version;
783 } linux1;
784 struct {
785 __u32 h_i_translator;
786 } hurd1;
787 struct {
788 __u32 m_i_reserved1;
789 } masix1;
790 } osd1; /* OS dependent 1 */
791 __le32 i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
792 __le32 i_generation; /* File version (for NFS) */
793 __le32 i_file_acl_lo; /* File ACL */
794 __le32 i_size_high;
795 __le32 i_obso_faddr; /* Obsoleted fragment address */
796 union {
797 struct {
798 __le16 l_i_blocks_high; /* were l_i_reserved1 */
799 __le16 l_i_file_acl_high;
800 __le16 l_i_uid_high; /* these 2 fields */
801 __le16 l_i_gid_high; /* were reserved2[0] */
802 __le16 l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
803 __le16 l_i_reserved;
804 } linux2;
805 struct {
806 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */
807 __u16 h_i_mode_high;
808 __u16 h_i_uid_high;
809 __u16 h_i_gid_high;
810 __u32 h_i_author;
811 } hurd2;
812 struct {
813 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */
814 __le16 m_i_file_acl_high;
815 __u32 m_i_reserved2[2];
816 } masix2;
817 } osd2; /* OS dependent 2 */
818 __le16 i_extra_isize;
819 __le16 i_checksum_hi; /* crc32c(uuid+inum+inode) BE */
820 __le32 i_ctime_extra; /* extra Change time (nsec << 2 | epoch) */
821 __le32 i_mtime_extra; /* extra Modification time(nsec << 2 | epoch) */
822 __le32 i_atime_extra; /* extra Access time (nsec << 2 | epoch) */
823 __le32 i_crtime; /* File Creation time */
824 __le32 i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
825 __le32 i_version_hi; /* high 32 bits for 64-bit version */
826 __le32 i_projid; /* Project ID */
827};
828
829#define EXT4_EPOCH_BITS 2
830#define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
831#define EXT4_NSEC_MASK (~0UL << EXT4_EPOCH_BITS)
832
833/*
834 * Extended fields will fit into an inode if the filesystem was formatted
835 * with large inodes (-I 256 or larger) and there are not currently any EAs
836 * consuming all of the available space. For new inodes we always reserve
837 * enough space for the kernel's known extended fields, but for inodes
838 * created with an old kernel this might not have been the case. None of
839 * the extended inode fields is critical for correct filesystem operation.
840 * This macro checks if a certain field fits in the inode. Note that
841 * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
842 */
843#define EXT4_FITS_IN_INODE(ext4_inode, einode, field) \
844 ((offsetof(typeof(*ext4_inode), field) + \
845 sizeof((ext4_inode)->field)) \
846 <= (EXT4_GOOD_OLD_INODE_SIZE + \
847 (einode)->i_extra_isize)) \
848
849/*
850 * We use an encoding that preserves the times for extra epoch "00":
851 *
852 * extra msb of adjust for signed
853 * epoch 32-bit 32-bit tv_sec to
854 * bits time decoded 64-bit tv_sec 64-bit tv_sec valid time range
855 * 0 0 1 -0x80000000..-0x00000001 0x000000000 1901-12-13..1969-12-31
856 * 0 0 0 0x000000000..0x07fffffff 0x000000000 1970-01-01..2038-01-19
857 * 0 1 1 0x080000000..0x0ffffffff 0x100000000 2038-01-19..2106-02-07
858 * 0 1 0 0x100000000..0x17fffffff 0x100000000 2106-02-07..2174-02-25
859 * 1 0 1 0x180000000..0x1ffffffff 0x200000000 2174-02-25..2242-03-16
860 * 1 0 0 0x200000000..0x27fffffff 0x200000000 2242-03-16..2310-04-04
861 * 1 1 1 0x280000000..0x2ffffffff 0x300000000 2310-04-04..2378-04-22
862 * 1 1 0 0x300000000..0x37fffffff 0x300000000 2378-04-22..2446-05-10
863 *
864 * Note that previous versions of the kernel on 64-bit systems would
865 * incorrectly use extra epoch bits 1,1 for dates between 1901 and
866 * 1970. e2fsck will correct this, assuming that it is run on the
867 * affected filesystem before 2242.
868 */
869
870static inline __le32 ext4_encode_extra_time(struct timespec64 ts)
871{
872 u32 extra = ((ts.tv_sec - (s32)ts.tv_sec) >> 32) & EXT4_EPOCH_MASK;
873 return cpu_to_le32(extra | (ts.tv_nsec << EXT4_EPOCH_BITS));
874}
875
876static inline struct timespec64 ext4_decode_extra_time(__le32 base,
877 __le32 extra)
878{
879 struct timespec64 ts = { .tv_sec = (signed)le32_to_cpu(base) };
880
881 if (unlikely(extra & cpu_to_le32(EXT4_EPOCH_MASK)))
882 ts.tv_sec += (u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK) << 32;
883 ts.tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
884 return ts;
885}
886
887#define EXT4_INODE_SET_XTIME_VAL(xtime, inode, raw_inode, ts) \
888do { \
889 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) { \
890 (raw_inode)->xtime = cpu_to_le32((ts).tv_sec); \
891 (raw_inode)->xtime ## _extra = ext4_encode_extra_time(ts); \
892 } else \
893 (raw_inode)->xtime = cpu_to_le32(clamp_t(int32_t, (ts).tv_sec, S32_MIN, S32_MAX)); \
894} while (0)
895
896#define EXT4_INODE_SET_ATIME(inode, raw_inode) \
897 EXT4_INODE_SET_XTIME_VAL(i_atime, inode, raw_inode, inode_get_atime(inode))
898
899#define EXT4_INODE_SET_MTIME(inode, raw_inode) \
900 EXT4_INODE_SET_XTIME_VAL(i_mtime, inode, raw_inode, inode_get_mtime(inode))
901
902#define EXT4_INODE_SET_CTIME(inode, raw_inode) \
903 EXT4_INODE_SET_XTIME_VAL(i_ctime, inode, raw_inode, inode_get_ctime(inode))
904
905#define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode) \
906 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \
907 EXT4_INODE_SET_XTIME_VAL(xtime, &((einode)->vfs_inode), \
908 raw_inode, (einode)->xtime)
909
910#define EXT4_INODE_GET_XTIME_VAL(xtime, inode, raw_inode) \
911 (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra) ? \
912 ext4_decode_extra_time((raw_inode)->xtime, \
913 (raw_inode)->xtime ## _extra) : \
914 (struct timespec64) { \
915 .tv_sec = (signed)le32_to_cpu((raw_inode)->xtime) \
916 })
917
918#define EXT4_INODE_GET_ATIME(inode, raw_inode) \
919do { \
920 inode_set_atime_to_ts(inode, \
921 EXT4_INODE_GET_XTIME_VAL(i_atime, inode, raw_inode)); \
922} while (0)
923
924#define EXT4_INODE_GET_MTIME(inode, raw_inode) \
925do { \
926 inode_set_mtime_to_ts(inode, \
927 EXT4_INODE_GET_XTIME_VAL(i_mtime, inode, raw_inode)); \
928} while (0)
929
930#define EXT4_INODE_GET_CTIME(inode, raw_inode) \
931do { \
932 inode_set_ctime_to_ts(inode, \
933 EXT4_INODE_GET_XTIME_VAL(i_ctime, inode, raw_inode)); \
934} while (0)
935
936#define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode) \
937do { \
938 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \
939 (einode)->xtime = \
940 EXT4_INODE_GET_XTIME_VAL(xtime, &(einode->vfs_inode), \
941 raw_inode); \
942 else \
943 (einode)->xtime = (struct timespec64){0, 0}; \
944} while (0)
945
946#define i_disk_version osd1.linux1.l_i_version
947
948#if defined(__KERNEL__) || defined(__linux__)
949#define i_reserved1 osd1.linux1.l_i_reserved1
950#define i_file_acl_high osd2.linux2.l_i_file_acl_high
951#define i_blocks_high osd2.linux2.l_i_blocks_high
952#define i_uid_low i_uid
953#define i_gid_low i_gid
954#define i_uid_high osd2.linux2.l_i_uid_high
955#define i_gid_high osd2.linux2.l_i_gid_high
956#define i_checksum_lo osd2.linux2.l_i_checksum_lo
957
958#elif defined(__GNU__)
959
960#define i_translator osd1.hurd1.h_i_translator
961#define i_uid_high osd2.hurd2.h_i_uid_high
962#define i_gid_high osd2.hurd2.h_i_gid_high
963#define i_author osd2.hurd2.h_i_author
964
965#elif defined(__masix__)
966
967#define i_reserved1 osd1.masix1.m_i_reserved1
968#define i_file_acl_high osd2.masix2.m_i_file_acl_high
969#define i_reserved2 osd2.masix2.m_i_reserved2
970
971#endif /* defined(__KERNEL__) || defined(__linux__) */
972
973#include "extents_status.h"
974#include "fast_commit.h"
975
976/*
977 * Lock subclasses for i_data_sem in the ext4_inode_info structure.
978 *
979 * These are needed to avoid lockdep false positives when we need to
980 * allocate blocks to the quota inode during ext4_map_blocks(), while
981 * holding i_data_sem for a normal (non-quota) inode. Since we don't
982 * do quota tracking for the quota inode, this avoids deadlock (as
983 * well as infinite recursion, since it isn't turtles all the way
984 * down...)
985 *
986 * I_DATA_SEM_NORMAL - Used for most inodes
987 * I_DATA_SEM_OTHER - Used by move_inode.c for the second normal inode
988 * where the second inode has larger inode number
989 * than the first
990 * I_DATA_SEM_QUOTA - Used for quota inodes only
991 * I_DATA_SEM_EA - Used for ea_inodes only
992 */
993enum {
994 I_DATA_SEM_NORMAL = 0,
995 I_DATA_SEM_OTHER,
996 I_DATA_SEM_QUOTA,
997 I_DATA_SEM_EA
998};
999
1000
1001/*
1002 * fourth extended file system inode data in memory
1003 */
1004struct ext4_inode_info {
1005 __le32 i_data[15]; /* unconverted */
1006 __u32 i_dtime;
1007 ext4_fsblk_t i_file_acl;
1008
1009 /*
1010 * i_block_group is the number of the block group which contains
1011 * this file's inode. Constant across the lifetime of the inode,
1012 * it is used for making block allocation decisions - we try to
1013 * place a file's data blocks near its inode block, and new inodes
1014 * near to their parent directory's inode.
1015 */
1016 ext4_group_t i_block_group;
1017 ext4_lblk_t i_dir_start_lookup;
1018#if (BITS_PER_LONG < 64)
1019 unsigned long i_state_flags; /* Dynamic state flags */
1020#endif
1021 unsigned long i_flags;
1022
1023 /*
1024 * Extended attributes can be read independently of the main file
1025 * data. Taking i_rwsem even when reading would cause contention
1026 * between readers of EAs and writers of regular file data, so
1027 * instead we synchronize on xattr_sem when reading or changing
1028 * EAs.
1029 */
1030 struct rw_semaphore xattr_sem;
1031
1032 /*
1033 * Inodes with EXT4_STATE_ORPHAN_FILE use i_orphan_idx. Otherwise
1034 * i_orphan is used.
1035 */
1036 union {
1037 struct list_head i_orphan; /* unlinked but open inodes */
1038 unsigned int i_orphan_idx; /* Index in orphan file */
1039 };
1040
1041 /* Fast commit related info */
1042
1043 /* For tracking dentry create updates */
1044 struct list_head i_fc_dilist;
1045 struct list_head i_fc_list; /*
1046 * inodes that need fast commit
1047 * protected by sbi->s_fc_lock.
1048 */
1049
1050 /* Start of lblk range that needs to be committed in this fast commit */
1051 ext4_lblk_t i_fc_lblk_start;
1052
1053 /* End of lblk range that needs to be committed in this fast commit */
1054 ext4_lblk_t i_fc_lblk_len;
1055
1056 /* Number of ongoing updates on this inode */
1057 atomic_t i_fc_updates;
1058
1059 /* Fast commit wait queue for this inode */
1060 wait_queue_head_t i_fc_wait;
1061
1062 /* Protect concurrent accesses on i_fc_lblk_start, i_fc_lblk_len */
1063 struct mutex i_fc_lock;
1064
1065 /*
1066 * i_disksize keeps track of what the inode size is ON DISK, not
1067 * in memory. During truncate, i_size is set to the new size by
1068 * the VFS prior to calling ext4_truncate(), but the filesystem won't
1069 * set i_disksize to 0 until the truncate is actually under way.
1070 *
1071 * The intent is that i_disksize always represents the blocks which
1072 * are used by this file. This allows recovery to restart truncate
1073 * on orphans if we crash during truncate. We actually write i_disksize
1074 * into the on-disk inode when writing inodes out, instead of i_size.
1075 *
1076 * The only time when i_disksize and i_size may be different is when
1077 * a truncate is in progress. The only things which change i_disksize
1078 * are ext4_get_block (growth) and ext4_truncate (shrinkth).
1079 */
1080 loff_t i_disksize;
1081
1082 /*
1083 * i_data_sem is for serialising ext4_truncate() against
1084 * ext4_getblock(). In the 2.4 ext2 design, great chunks of inode's
1085 * data tree are chopped off during truncate. We can't do that in
1086 * ext4 because whenever we perform intermediate commits during
1087 * truncate, the inode and all the metadata blocks *must* be in a
1088 * consistent state which allows truncation of the orphans to restart
1089 * during recovery. Hence we must fix the get_block-vs-truncate race
1090 * by other means, so we have i_data_sem.
1091 */
1092 struct rw_semaphore i_data_sem;
1093 struct inode vfs_inode;
1094 struct jbd2_inode *jinode;
1095
1096 spinlock_t i_raw_lock; /* protects updates to the raw inode */
1097
1098 /*
1099 * File creation time. Its function is same as that of
1100 * struct timespec64 i_{a,c,m}time in the generic inode.
1101 */
1102 struct timespec64 i_crtime;
1103
1104 /* mballoc */
1105 atomic_t i_prealloc_active;
1106 struct rb_root i_prealloc_node;
1107 rwlock_t i_prealloc_lock;
1108
1109 /* extents status tree */
1110 struct ext4_es_tree i_es_tree;
1111 rwlock_t i_es_lock;
1112 struct list_head i_es_list;
1113 unsigned int i_es_all_nr; /* protected by i_es_lock */
1114 unsigned int i_es_shk_nr; /* protected by i_es_lock */
1115 ext4_lblk_t i_es_shrink_lblk; /* Offset where we start searching for
1116 extents to shrink. Protected by
1117 i_es_lock */
1118
1119 /* ialloc */
1120 ext4_group_t i_last_alloc_group;
1121
1122 /* allocation reservation info for delalloc */
1123 /* In case of bigalloc, this refer to clusters rather than blocks */
1124 unsigned int i_reserved_data_blocks;
1125
1126 /* pending cluster reservations for bigalloc file systems */
1127 struct ext4_pending_tree i_pending_tree;
1128
1129 /* on-disk additional length */
1130 __u16 i_extra_isize;
1131
1132 /* Indicate the inline data space. */
1133 u16 i_inline_off;
1134 u16 i_inline_size;
1135
1136#ifdef CONFIG_QUOTA
1137 /* quota space reservation, managed internally by quota code */
1138 qsize_t i_reserved_quota;
1139#endif
1140
1141 /* Lock protecting lists below */
1142 spinlock_t i_completed_io_lock;
1143 /*
1144 * Completed IOs that need unwritten extents handling and have
1145 * transaction reserved
1146 */
1147 struct list_head i_rsv_conversion_list;
1148 struct work_struct i_rsv_conversion_work;
1149 atomic_t i_unwritten; /* Nr. of inflight conversions pending */
1150
1151 spinlock_t i_block_reservation_lock;
1152
1153 /*
1154 * Transactions that contain inode's metadata needed to complete
1155 * fsync and fdatasync, respectively.
1156 */
1157 tid_t i_sync_tid;
1158 tid_t i_datasync_tid;
1159
1160#ifdef CONFIG_QUOTA
1161 struct dquot *i_dquot[MAXQUOTAS];
1162#endif
1163
1164 /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
1165 __u32 i_csum_seed;
1166
1167 kprojid_t i_projid;
1168};
1169
1170/*
1171 * File system states
1172 */
1173#define EXT4_VALID_FS 0x0001 /* Unmounted cleanly */
1174#define EXT4_ERROR_FS 0x0002 /* Errors detected */
1175#define EXT4_ORPHAN_FS 0x0004 /* Orphans being recovered */
1176#define EXT4_FC_REPLAY 0x0020 /* Fast commit replay ongoing */
1177
1178/*
1179 * Misc. filesystem flags
1180 */
1181#define EXT2_FLAGS_SIGNED_HASH 0x0001 /* Signed dirhash in use */
1182#define EXT2_FLAGS_UNSIGNED_HASH 0x0002 /* Unsigned dirhash in use */
1183#define EXT2_FLAGS_TEST_FILESYS 0x0004 /* to test development code */
1184
1185/*
1186 * Mount flags set via mount options or defaults
1187 */
1188#define EXT4_MOUNT_NO_MBCACHE 0x00001 /* Do not use mbcache */
1189#define EXT4_MOUNT_GRPID 0x00004 /* Create files with directory's group */
1190#define EXT4_MOUNT_DEBUG 0x00008 /* Some debugging messages */
1191#define EXT4_MOUNT_ERRORS_CONT 0x00010 /* Continue on errors */
1192#define EXT4_MOUNT_ERRORS_RO 0x00020 /* Remount fs ro on errors */
1193#define EXT4_MOUNT_ERRORS_PANIC 0x00040 /* Panic on errors */
1194#define EXT4_MOUNT_ERRORS_MASK 0x00070
1195#define EXT4_MOUNT_MINIX_DF 0x00080 /* Mimics the Minix statfs */
1196#define EXT4_MOUNT_NOLOAD 0x00100 /* Don't use existing journal*/
1197#ifdef CONFIG_FS_DAX
1198#define EXT4_MOUNT_DAX_ALWAYS 0x00200 /* Direct Access */
1199#else
1200#define EXT4_MOUNT_DAX_ALWAYS 0
1201#endif
1202#define EXT4_MOUNT_DATA_FLAGS 0x00C00 /* Mode for data writes: */
1203#define EXT4_MOUNT_JOURNAL_DATA 0x00400 /* Write data to journal */
1204#define EXT4_MOUNT_ORDERED_DATA 0x00800 /* Flush data before commit */
1205#define EXT4_MOUNT_WRITEBACK_DATA 0x00C00 /* No data ordering */
1206#define EXT4_MOUNT_UPDATE_JOURNAL 0x01000 /* Update the journal format */
1207#define EXT4_MOUNT_NO_UID32 0x02000 /* Disable 32-bit UIDs */
1208#define EXT4_MOUNT_XATTR_USER 0x04000 /* Extended user attributes */
1209#define EXT4_MOUNT_POSIX_ACL 0x08000 /* POSIX Access Control Lists */
1210#define EXT4_MOUNT_NO_AUTO_DA_ALLOC 0x10000 /* No auto delalloc mapping */
1211#define EXT4_MOUNT_BARRIER 0x20000 /* Use block barriers */
1212#define EXT4_MOUNT_QUOTA 0x40000 /* Some quota option set */
1213#define EXT4_MOUNT_USRQUOTA 0x80000 /* "old" user quota,
1214 * enable enforcement for hidden
1215 * quota files */
1216#define EXT4_MOUNT_GRPQUOTA 0x100000 /* "old" group quota, enable
1217 * enforcement for hidden quota
1218 * files */
1219#define EXT4_MOUNT_PRJQUOTA 0x200000 /* Enable project quota
1220 * enforcement */
1221#define EXT4_MOUNT_DIOREAD_NOLOCK 0x400000 /* Enable support for dio read nolocking */
1222#define EXT4_MOUNT_JOURNAL_CHECKSUM 0x800000 /* Journal checksums */
1223#define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
1224#define EXT4_MOUNT_WARN_ON_ERROR 0x2000000 /* Trigger WARN_ON on error */
1225#define EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS 0x4000000
1226#define EXT4_MOUNT_DELALLOC 0x8000000 /* Delalloc support */
1227#define EXT4_MOUNT_DATA_ERR_ABORT 0x10000000 /* Abort on file data write */
1228#define EXT4_MOUNT_BLOCK_VALIDITY 0x20000000 /* Block validity checking */
1229#define EXT4_MOUNT_DISCARD 0x40000000 /* Issue DISCARD requests */
1230#define EXT4_MOUNT_INIT_INODE_TABLE 0x80000000 /* Initialize uninitialized itables */
1231
1232/*
1233 * Mount flags set either automatically (could not be set by mount option)
1234 * based on per file system feature or property or in special cases such as
1235 * distinguishing between explicit mount option definition and default.
1236 */
1237#define EXT4_MOUNT2_EXPLICIT_DELALLOC 0x00000001 /* User explicitly
1238 specified delalloc */
1239#define EXT4_MOUNT2_STD_GROUP_SIZE 0x00000002 /* We have standard group
1240 size of blocksize * 8
1241 blocks */
1242#define EXT4_MOUNT2_HURD_COMPAT 0x00000004 /* Support HURD-castrated
1243 file systems */
1244#define EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM 0x00000008 /* User explicitly
1245 specified journal checksum */
1246
1247#define EXT4_MOUNT2_JOURNAL_FAST_COMMIT 0x00000010 /* Journal fast commit */
1248#define EXT4_MOUNT2_DAX_NEVER 0x00000020 /* Do not allow Direct Access */
1249#define EXT4_MOUNT2_DAX_INODE 0x00000040 /* For printing options only */
1250#define EXT4_MOUNT2_MB_OPTIMIZE_SCAN 0x00000080 /* Optimize group
1251 * scanning in mballoc
1252 */
1253#define EXT4_MOUNT2_ABORT 0x00000100 /* Abort filesystem */
1254
1255#define clear_opt(sb, opt) EXT4_SB(sb)->s_mount_opt &= \
1256 ~EXT4_MOUNT_##opt
1257#define set_opt(sb, opt) EXT4_SB(sb)->s_mount_opt |= \
1258 EXT4_MOUNT_##opt
1259#define test_opt(sb, opt) (EXT4_SB(sb)->s_mount_opt & \
1260 EXT4_MOUNT_##opt)
1261
1262#define clear_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 &= \
1263 ~EXT4_MOUNT2_##opt
1264#define set_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 |= \
1265 EXT4_MOUNT2_##opt
1266#define test_opt2(sb, opt) (EXT4_SB(sb)->s_mount_opt2 & \
1267 EXT4_MOUNT2_##opt)
1268
1269#define ext4_test_and_set_bit __test_and_set_bit_le
1270#define ext4_set_bit __set_bit_le
1271#define ext4_test_and_clear_bit __test_and_clear_bit_le
1272#define ext4_clear_bit __clear_bit_le
1273#define ext4_test_bit test_bit_le
1274#define ext4_find_next_zero_bit find_next_zero_bit_le
1275#define ext4_find_next_bit find_next_bit_le
1276
1277extern void mb_set_bits(void *bm, int cur, int len);
1278
1279/*
1280 * Maximal mount counts between two filesystem checks
1281 */
1282#define EXT4_DFL_MAX_MNT_COUNT 20 /* Allow 20 mounts */
1283#define EXT4_DFL_CHECKINTERVAL 0 /* Don't use interval check */
1284
1285/*
1286 * Behaviour when detecting errors
1287 */
1288#define EXT4_ERRORS_CONTINUE 1 /* Continue execution */
1289#define EXT4_ERRORS_RO 2 /* Remount fs read-only */
1290#define EXT4_ERRORS_PANIC 3 /* Panic */
1291#define EXT4_ERRORS_DEFAULT EXT4_ERRORS_CONTINUE
1292
1293/* Metadata checksum algorithm codes */
1294#define EXT4_CRC32C_CHKSUM 1
1295
1296#define EXT4_LABEL_MAX 16
1297
1298/*
1299 * Structure of the super block
1300 */
1301struct ext4_super_block {
1302/*00*/ __le32 s_inodes_count; /* Inodes count */
1303 __le32 s_blocks_count_lo; /* Blocks count */
1304 __le32 s_r_blocks_count_lo; /* Reserved blocks count */
1305 __le32 s_free_blocks_count_lo; /* Free blocks count */
1306/*10*/ __le32 s_free_inodes_count; /* Free inodes count */
1307 __le32 s_first_data_block; /* First Data Block */
1308 __le32 s_log_block_size; /* Block size */
1309 __le32 s_log_cluster_size; /* Allocation cluster size */
1310/*20*/ __le32 s_blocks_per_group; /* # Blocks per group */
1311 __le32 s_clusters_per_group; /* # Clusters per group */
1312 __le32 s_inodes_per_group; /* # Inodes per group */
1313 __le32 s_mtime; /* Mount time */
1314/*30*/ __le32 s_wtime; /* Write time */
1315 __le16 s_mnt_count; /* Mount count */
1316 __le16 s_max_mnt_count; /* Maximal mount count */
1317 __le16 s_magic; /* Magic signature */
1318 __le16 s_state; /* File system state */
1319 __le16 s_errors; /* Behaviour when detecting errors */
1320 __le16 s_minor_rev_level; /* minor revision level */
1321/*40*/ __le32 s_lastcheck; /* time of last check */
1322 __le32 s_checkinterval; /* max. time between checks */
1323 __le32 s_creator_os; /* OS */
1324 __le32 s_rev_level; /* Revision level */
1325/*50*/ __le16 s_def_resuid; /* Default uid for reserved blocks */
1326 __le16 s_def_resgid; /* Default gid for reserved blocks */
1327 /*
1328 * These fields are for EXT4_DYNAMIC_REV superblocks only.
1329 *
1330 * Note: the difference between the compatible feature set and
1331 * the incompatible feature set is that if there is a bit set
1332 * in the incompatible feature set that the kernel doesn't
1333 * know about, it should refuse to mount the filesystem.
1334 *
1335 * e2fsck's requirements are more strict; if it doesn't know
1336 * about a feature in either the compatible or incompatible
1337 * feature set, it must abort and not try to meddle with
1338 * things it doesn't understand...
1339 */
1340 __le32 s_first_ino; /* First non-reserved inode */
1341 __le16 s_inode_size; /* size of inode structure */
1342 __le16 s_block_group_nr; /* block group # of this superblock */
1343 __le32 s_feature_compat; /* compatible feature set */
1344/*60*/ __le32 s_feature_incompat; /* incompatible feature set */
1345 __le32 s_feature_ro_compat; /* readonly-compatible feature set */
1346/*68*/ __u8 s_uuid[16]; /* 128-bit uuid for volume */
1347/*78*/ char s_volume_name[EXT4_LABEL_MAX]; /* volume name */
1348/*88*/ char s_last_mounted[64] __nonstring; /* directory where last mounted */
1349/*C8*/ __le32 s_algorithm_usage_bitmap; /* For compression */
1350 /*
1351 * Performance hints. Directory preallocation should only
1352 * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1353 */
1354 __u8 s_prealloc_blocks; /* Nr of blocks to try to preallocate*/
1355 __u8 s_prealloc_dir_blocks; /* Nr to preallocate for dirs */
1356 __le16 s_reserved_gdt_blocks; /* Per group desc for online growth */
1357 /*
1358 * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1359 */
1360/*D0*/ __u8 s_journal_uuid[16]; /* uuid of journal superblock */
1361/*E0*/ __le32 s_journal_inum; /* inode number of journal file */
1362 __le32 s_journal_dev; /* device number of journal file */
1363 __le32 s_last_orphan; /* start of list of inodes to delete */
1364 __le32 s_hash_seed[4]; /* HTREE hash seed */
1365 __u8 s_def_hash_version; /* Default hash version to use */
1366 __u8 s_jnl_backup_type;
1367 __le16 s_desc_size; /* size of group descriptor */
1368/*100*/ __le32 s_default_mount_opts;
1369 __le32 s_first_meta_bg; /* First metablock block group */
1370 __le32 s_mkfs_time; /* When the filesystem was created */
1371 __le32 s_jnl_blocks[17]; /* Backup of the journal inode */
1372 /* 64bit support valid if EXT4_FEATURE_INCOMPAT_64BIT */
1373/*150*/ __le32 s_blocks_count_hi; /* Blocks count */
1374 __le32 s_r_blocks_count_hi; /* Reserved blocks count */
1375 __le32 s_free_blocks_count_hi; /* Free blocks count */
1376 __le16 s_min_extra_isize; /* All inodes have at least # bytes */
1377 __le16 s_want_extra_isize; /* New inodes should reserve # bytes */
1378 __le32 s_flags; /* Miscellaneous flags */
1379 __le16 s_raid_stride; /* RAID stride */
1380 __le16 s_mmp_update_interval; /* # seconds to wait in MMP checking */
1381 __le64 s_mmp_block; /* Block for multi-mount protection */
1382 __le32 s_raid_stripe_width; /* blocks on all data disks (N*stride)*/
1383 __u8 s_log_groups_per_flex; /* FLEX_BG group size */
1384 __u8 s_checksum_type; /* metadata checksum algorithm used */
1385 __u8 s_encryption_level; /* versioning level for encryption */
1386 __u8 s_reserved_pad; /* Padding to next 32bits */
1387 __le64 s_kbytes_written; /* nr of lifetime kilobytes written */
1388 __le32 s_snapshot_inum; /* Inode number of active snapshot */
1389 __le32 s_snapshot_id; /* sequential ID of active snapshot */
1390 __le64 s_snapshot_r_blocks_count; /* reserved blocks for active
1391 snapshot's future use */
1392 __le32 s_snapshot_list; /* inode number of the head of the
1393 on-disk snapshot list */
1394#define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1395 __le32 s_error_count; /* number of fs errors */
1396 __le32 s_first_error_time; /* first time an error happened */
1397 __le32 s_first_error_ino; /* inode involved in first error */
1398 __le64 s_first_error_block; /* block involved of first error */
1399 __u8 s_first_error_func[32] __nonstring; /* function where the error happened */
1400 __le32 s_first_error_line; /* line number where error happened */
1401 __le32 s_last_error_time; /* most recent time of an error */
1402 __le32 s_last_error_ino; /* inode involved in last error */
1403 __le32 s_last_error_line; /* line number where error happened */
1404 __le64 s_last_error_block; /* block involved of last error */
1405 __u8 s_last_error_func[32] __nonstring; /* function where the error happened */
1406#define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1407 __u8 s_mount_opts[64];
1408 __le32 s_usr_quota_inum; /* inode for tracking user quota */
1409 __le32 s_grp_quota_inum; /* inode for tracking group quota */
1410 __le32 s_overhead_clusters; /* overhead blocks/clusters in fs */
1411 __le32 s_backup_bgs[2]; /* groups with sparse_super2 SBs */
1412 __u8 s_encrypt_algos[4]; /* Encryption algorithms in use */
1413 __u8 s_encrypt_pw_salt[16]; /* Salt used for string2key algorithm */
1414 __le32 s_lpf_ino; /* Location of the lost+found inode */
1415 __le32 s_prj_quota_inum; /* inode for tracking project quota */
1416 __le32 s_checksum_seed; /* crc32c(uuid) if csum_seed set */
1417 __u8 s_wtime_hi;
1418 __u8 s_mtime_hi;
1419 __u8 s_mkfs_time_hi;
1420 __u8 s_lastcheck_hi;
1421 __u8 s_first_error_time_hi;
1422 __u8 s_last_error_time_hi;
1423 __u8 s_first_error_errcode;
1424 __u8 s_last_error_errcode;
1425 __le16 s_encoding; /* Filename charset encoding */
1426 __le16 s_encoding_flags; /* Filename charset encoding flags */
1427 __le32 s_orphan_file_inum; /* Inode for tracking orphan inodes */
1428 __le32 s_reserved[94]; /* Padding to the end of the block */
1429 __le32 s_checksum; /* crc32c(superblock) */
1430};
1431
1432#define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1433
1434#ifdef __KERNEL__
1435
1436/* Number of quota types we support */
1437#define EXT4_MAXQUOTAS 3
1438
1439#define EXT4_ENC_UTF8_12_1 1
1440
1441/* Types of ext4 journal triggers */
1442enum ext4_journal_trigger_type {
1443 EXT4_JTR_ORPHAN_FILE,
1444 EXT4_JTR_NONE /* This must be the last entry for indexing to work! */
1445};
1446
1447#define EXT4_JOURNAL_TRIGGER_COUNT EXT4_JTR_NONE
1448
1449struct ext4_journal_trigger {
1450 struct jbd2_buffer_trigger_type tr_triggers;
1451 struct super_block *sb;
1452};
1453
1454static inline struct ext4_journal_trigger *EXT4_TRIGGER(
1455 struct jbd2_buffer_trigger_type *trigger)
1456{
1457 return container_of(trigger, struct ext4_journal_trigger, tr_triggers);
1458}
1459
1460#define EXT4_ORPHAN_BLOCK_MAGIC 0x0b10ca04
1461
1462/* Structure at the tail of orphan block */
1463struct ext4_orphan_block_tail {
1464 __le32 ob_magic;
1465 __le32 ob_checksum;
1466};
1467
1468static inline int ext4_inodes_per_orphan_block(struct super_block *sb)
1469{
1470 return (sb->s_blocksize - sizeof(struct ext4_orphan_block_tail)) /
1471 sizeof(u32);
1472}
1473
1474struct ext4_orphan_block {
1475 atomic_t ob_free_entries; /* Number of free orphan entries in block */
1476 struct buffer_head *ob_bh; /* Buffer for orphan block */
1477};
1478
1479/*
1480 * Info about orphan file.
1481 */
1482struct ext4_orphan_info {
1483 int of_blocks; /* Number of orphan blocks in a file */
1484 __u32 of_csum_seed; /* Checksum seed for orphan file */
1485 struct ext4_orphan_block *of_binfo; /* Array with info about orphan
1486 * file blocks */
1487};
1488
1489/*
1490 * fourth extended-fs super-block data in memory
1491 */
1492struct ext4_sb_info {
1493 unsigned long s_desc_size; /* Size of a group descriptor in bytes */
1494 unsigned long s_inodes_per_block;/* Number of inodes per block */
1495 unsigned long s_blocks_per_group;/* Number of blocks in a group */
1496 unsigned long s_clusters_per_group; /* Number of clusters in a group */
1497 unsigned long s_inodes_per_group;/* Number of inodes in a group */
1498 unsigned long s_itb_per_group; /* Number of inode table blocks per group */
1499 unsigned long s_gdb_count; /* Number of group descriptor blocks */
1500 unsigned long s_desc_per_block; /* Number of group descriptors per block */
1501 ext4_group_t s_groups_count; /* Number of groups in the fs */
1502 ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1503 unsigned long s_overhead; /* # of fs overhead clusters */
1504 unsigned int s_cluster_ratio; /* Number of blocks per cluster */
1505 unsigned int s_cluster_bits; /* log2 of s_cluster_ratio */
1506 loff_t s_bitmap_maxbytes; /* max bytes for bitmap files */
1507 struct buffer_head * s_sbh; /* Buffer containing the super block */
1508 struct ext4_super_block *s_es; /* Pointer to the super block in the buffer */
1509 /* Array of bh's for the block group descriptors */
1510 struct buffer_head * __rcu *s_group_desc;
1511 unsigned int s_mount_opt;
1512 unsigned int s_mount_opt2;
1513 unsigned long s_mount_flags;
1514 unsigned int s_def_mount_opt;
1515 unsigned int s_def_mount_opt2;
1516 ext4_fsblk_t s_sb_block;
1517 atomic64_t s_resv_clusters;
1518 kuid_t s_resuid;
1519 kgid_t s_resgid;
1520 unsigned short s_mount_state;
1521 unsigned short s_pad;
1522 int s_addr_per_block_bits;
1523 int s_desc_per_block_bits;
1524 int s_inode_size;
1525 int s_first_ino;
1526 unsigned int s_inode_readahead_blks;
1527 unsigned int s_inode_goal;
1528 u32 s_hash_seed[4];
1529 int s_def_hash_version;
1530 int s_hash_unsigned; /* 3 if hash should be unsigned, 0 if not */
1531 struct percpu_counter s_freeclusters_counter;
1532 struct percpu_counter s_freeinodes_counter;
1533 struct percpu_counter s_dirs_counter;
1534 struct percpu_counter s_dirtyclusters_counter;
1535 struct percpu_counter s_sra_exceeded_retry_limit;
1536 struct blockgroup_lock *s_blockgroup_lock;
1537 struct proc_dir_entry *s_proc;
1538 struct kobject s_kobj;
1539 struct completion s_kobj_unregister;
1540 struct super_block *s_sb;
1541 struct buffer_head *s_mmp_bh;
1542
1543 /* Journaling */
1544 struct journal_s *s_journal;
1545 unsigned long s_ext4_flags; /* Ext4 superblock flags */
1546 struct mutex s_orphan_lock; /* Protects on disk list changes */
1547 struct list_head s_orphan; /* List of orphaned inodes in on disk
1548 list */
1549 struct ext4_orphan_info s_orphan_info;
1550 unsigned long s_commit_interval;
1551 u32 s_max_batch_time;
1552 u32 s_min_batch_time;
1553 struct bdev_handle *s_journal_bdev_handle;
1554#ifdef CONFIG_QUOTA
1555 /* Names of quota files with journalled quota */
1556 char __rcu *s_qf_names[EXT4_MAXQUOTAS];
1557 int s_jquota_fmt; /* Format of quota to use */
1558#endif
1559 unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1560 struct ext4_system_blocks __rcu *s_system_blks;
1561
1562#ifdef EXTENTS_STATS
1563 /* ext4 extents stats */
1564 unsigned long s_ext_min;
1565 unsigned long s_ext_max;
1566 unsigned long s_depth_max;
1567 spinlock_t s_ext_stats_lock;
1568 unsigned long s_ext_blocks;
1569 unsigned long s_ext_extents;
1570#endif
1571
1572 /* for buddy allocator */
1573 struct ext4_group_info ** __rcu *s_group_info;
1574 struct inode *s_buddy_cache;
1575 spinlock_t s_md_lock;
1576 unsigned short *s_mb_offsets;
1577 unsigned int *s_mb_maxs;
1578 unsigned int s_group_info_size;
1579 unsigned int s_mb_free_pending;
1580 struct list_head s_freed_data_list[2]; /* List of blocks to be freed
1581 after commit completed */
1582 struct list_head s_discard_list;
1583 struct work_struct s_discard_work;
1584 atomic_t s_retry_alloc_pending;
1585 struct list_head *s_mb_avg_fragment_size;
1586 rwlock_t *s_mb_avg_fragment_size_locks;
1587 struct list_head *s_mb_largest_free_orders;
1588 rwlock_t *s_mb_largest_free_orders_locks;
1589
1590 /* tunables */
1591 unsigned long s_stripe;
1592 unsigned int s_mb_max_linear_groups;
1593 unsigned int s_mb_stream_request;
1594 unsigned int s_mb_max_to_scan;
1595 unsigned int s_mb_min_to_scan;
1596 unsigned int s_mb_stats;
1597 unsigned int s_mb_order2_reqs;
1598 unsigned int s_mb_group_prealloc;
1599 unsigned int s_max_dir_size_kb;
1600 /* where last allocation was done - for stream allocation */
1601 unsigned long s_mb_last_group;
1602 unsigned long s_mb_last_start;
1603 unsigned int s_mb_prefetch;
1604 unsigned int s_mb_prefetch_limit;
1605 unsigned int s_mb_best_avail_max_trim_order;
1606
1607 /* stats for buddy allocator */
1608 atomic_t s_bal_reqs; /* number of reqs with len > 1 */
1609 atomic_t s_bal_success; /* we found long enough chunks */
1610 atomic_t s_bal_allocated; /* in blocks */
1611 atomic_t s_bal_ex_scanned; /* total extents scanned */
1612 atomic_t s_bal_cX_ex_scanned[EXT4_MB_NUM_CRS]; /* total extents scanned */
1613 atomic_t s_bal_groups_scanned; /* number of groups scanned */
1614 atomic_t s_bal_goals; /* goal hits */
1615 atomic_t s_bal_len_goals; /* len goal hits */
1616 atomic_t s_bal_breaks; /* too long searches */
1617 atomic_t s_bal_2orders; /* 2^order hits */
1618 atomic_t s_bal_p2_aligned_bad_suggestions;
1619 atomic_t s_bal_goal_fast_bad_suggestions;
1620 atomic_t s_bal_best_avail_bad_suggestions;
1621 atomic64_t s_bal_cX_groups_considered[EXT4_MB_NUM_CRS];
1622 atomic64_t s_bal_cX_hits[EXT4_MB_NUM_CRS];
1623 atomic64_t s_bal_cX_failed[EXT4_MB_NUM_CRS]; /* cX loop didn't find blocks */
1624 atomic_t s_mb_buddies_generated; /* number of buddies generated */
1625 atomic64_t s_mb_generation_time;
1626 atomic_t s_mb_lost_chunks;
1627 atomic_t s_mb_preallocated;
1628 atomic_t s_mb_discarded;
1629 atomic_t s_lock_busy;
1630
1631 /* locality groups */
1632 struct ext4_locality_group __percpu *s_locality_groups;
1633
1634 /* for write statistics */
1635 unsigned long s_sectors_written_start;
1636 u64 s_kbytes_written;
1637
1638 /* the size of zero-out chunk */
1639 unsigned int s_extent_max_zeroout_kb;
1640
1641 unsigned int s_log_groups_per_flex;
1642 struct flex_groups * __rcu *s_flex_groups;
1643 ext4_group_t s_flex_groups_allocated;
1644
1645 /* workqueue for reserved extent conversions (buffered io) */
1646 struct workqueue_struct *rsv_conversion_wq;
1647
1648 /* timer for periodic error stats printing */
1649 struct timer_list s_err_report;
1650
1651 /* Lazy inode table initialization info */
1652 struct ext4_li_request *s_li_request;
1653 /* Wait multiplier for lazy initialization thread */
1654 unsigned int s_li_wait_mult;
1655
1656 /* Kernel thread for multiple mount protection */
1657 struct task_struct *s_mmp_tsk;
1658
1659 /* record the last minlen when FITRIM is called. */
1660 unsigned long s_last_trim_minblks;
1661
1662 /* Reference to checksum algorithm driver via cryptoapi */
1663 struct crypto_shash *s_chksum_driver;
1664
1665 /* Precomputed FS UUID checksum for seeding other checksums */
1666 __u32 s_csum_seed;
1667
1668 /* Reclaim extents from extent status tree */
1669 struct shrinker *s_es_shrinker;
1670 struct list_head s_es_list; /* List of inodes with reclaimable extents */
1671 long s_es_nr_inode;
1672 struct ext4_es_stats s_es_stats;
1673 struct mb_cache *s_ea_block_cache;
1674 struct mb_cache *s_ea_inode_cache;
1675 spinlock_t s_es_lock ____cacheline_aligned_in_smp;
1676
1677 /* Journal triggers for checksum computation */
1678 struct ext4_journal_trigger s_journal_triggers[EXT4_JOURNAL_TRIGGER_COUNT];
1679
1680 /* Ratelimit ext4 messages. */
1681 struct ratelimit_state s_err_ratelimit_state;
1682 struct ratelimit_state s_warning_ratelimit_state;
1683 struct ratelimit_state s_msg_ratelimit_state;
1684 atomic_t s_warning_count;
1685 atomic_t s_msg_count;
1686
1687 /* Encryption policy for '-o test_dummy_encryption' */
1688 struct fscrypt_dummy_policy s_dummy_enc_policy;
1689
1690 /*
1691 * Barrier between writepages ops and changing any inode's JOURNAL_DATA
1692 * or EXTENTS flag or between writepages ops and changing DELALLOC or
1693 * DIOREAD_NOLOCK mount options on remount.
1694 */
1695 struct percpu_rw_semaphore s_writepages_rwsem;
1696 struct dax_device *s_daxdev;
1697 u64 s_dax_part_off;
1698#ifdef CONFIG_EXT4_DEBUG
1699 unsigned long s_simulate_fail;
1700#endif
1701 /* Record the errseq of the backing block device */
1702 errseq_t s_bdev_wb_err;
1703 spinlock_t s_bdev_wb_lock;
1704
1705 /* Information about errors that happened during this mount */
1706 spinlock_t s_error_lock;
1707 int s_add_error_count;
1708 int s_first_error_code;
1709 __u32 s_first_error_line;
1710 __u32 s_first_error_ino;
1711 __u64 s_first_error_block;
1712 const char *s_first_error_func;
1713 time64_t s_first_error_time;
1714 int s_last_error_code;
1715 __u32 s_last_error_line;
1716 __u32 s_last_error_ino;
1717 __u64 s_last_error_block;
1718 const char *s_last_error_func;
1719 time64_t s_last_error_time;
1720 /*
1721 * If we are in a context where we cannot update the on-disk
1722 * superblock, we queue the work here. This is used to update
1723 * the error information in the superblock, and for periodic
1724 * updates of the superblock called from the commit callback
1725 * function.
1726 */
1727 struct work_struct s_sb_upd_work;
1728
1729 /* Ext4 fast commit sub transaction ID */
1730 atomic_t s_fc_subtid;
1731
1732 /*
1733 * After commit starts, the main queue gets locked, and the further
1734 * updates get added in the staging queue.
1735 */
1736#define FC_Q_MAIN 0
1737#define FC_Q_STAGING 1
1738 struct list_head s_fc_q[2]; /* Inodes staged for fast commit
1739 * that have data changes in them.
1740 */
1741 struct list_head s_fc_dentry_q[2]; /* directory entry updates */
1742 unsigned int s_fc_bytes;
1743 /*
1744 * Main fast commit lock. This lock protects accesses to the
1745 * following fields:
1746 * ei->i_fc_list, s_fc_dentry_q, s_fc_q, s_fc_bytes, s_fc_bh.
1747 */
1748 spinlock_t s_fc_lock;
1749 struct buffer_head *s_fc_bh;
1750 struct ext4_fc_stats s_fc_stats;
1751 tid_t s_fc_ineligible_tid;
1752#ifdef CONFIG_EXT4_DEBUG
1753 int s_fc_debug_max_replay;
1754#endif
1755 struct ext4_fc_replay_state s_fc_replay_state;
1756};
1757
1758static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1759{
1760 return sb->s_fs_info;
1761}
1762static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1763{
1764 return container_of(inode, struct ext4_inode_info, vfs_inode);
1765}
1766
1767static inline int ext4_writepages_down_read(struct super_block *sb)
1768{
1769 percpu_down_read(&EXT4_SB(sb)->s_writepages_rwsem);
1770 return memalloc_nofs_save();
1771}
1772
1773static inline void ext4_writepages_up_read(struct super_block *sb, int ctx)
1774{
1775 memalloc_nofs_restore(ctx);
1776 percpu_up_read(&EXT4_SB(sb)->s_writepages_rwsem);
1777}
1778
1779static inline int ext4_writepages_down_write(struct super_block *sb)
1780{
1781 percpu_down_write(&EXT4_SB(sb)->s_writepages_rwsem);
1782 return memalloc_nofs_save();
1783}
1784
1785static inline void ext4_writepages_up_write(struct super_block *sb, int ctx)
1786{
1787 memalloc_nofs_restore(ctx);
1788 percpu_up_write(&EXT4_SB(sb)->s_writepages_rwsem);
1789}
1790
1791static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1792{
1793 return ino == EXT4_ROOT_INO ||
1794 (ino >= EXT4_FIRST_INO(sb) &&
1795 ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1796}
1797
1798/*
1799 * Returns: sbi->field[index]
1800 * Used to access an array element from the following sbi fields which require
1801 * rcu protection to avoid dereferencing an invalid pointer due to reassignment
1802 * - s_group_desc
1803 * - s_group_info
1804 * - s_flex_group
1805 */
1806#define sbi_array_rcu_deref(sbi, field, index) \
1807({ \
1808 typeof(*((sbi)->field)) _v; \
1809 rcu_read_lock(); \
1810 _v = ((typeof(_v)*)rcu_dereference((sbi)->field))[index]; \
1811 rcu_read_unlock(); \
1812 _v; \
1813})
1814
1815/*
1816 * run-time mount flags
1817 */
1818enum {
1819 EXT4_MF_MNTDIR_SAMPLED,
1820 EXT4_MF_FC_INELIGIBLE /* Fast commit ineligible */
1821};
1822
1823static inline void ext4_set_mount_flag(struct super_block *sb, int bit)
1824{
1825 set_bit(bit, &EXT4_SB(sb)->s_mount_flags);
1826}
1827
1828static inline void ext4_clear_mount_flag(struct super_block *sb, int bit)
1829{
1830 clear_bit(bit, &EXT4_SB(sb)->s_mount_flags);
1831}
1832
1833static inline int ext4_test_mount_flag(struct super_block *sb, int bit)
1834{
1835 return test_bit(bit, &EXT4_SB(sb)->s_mount_flags);
1836}
1837
1838
1839/*
1840 * Simulate_fail codes
1841 */
1842#define EXT4_SIM_BBITMAP_EIO 1
1843#define EXT4_SIM_BBITMAP_CRC 2
1844#define EXT4_SIM_IBITMAP_EIO 3
1845#define EXT4_SIM_IBITMAP_CRC 4
1846#define EXT4_SIM_INODE_EIO 5
1847#define EXT4_SIM_INODE_CRC 6
1848#define EXT4_SIM_DIRBLOCK_EIO 7
1849#define EXT4_SIM_DIRBLOCK_CRC 8
1850
1851static inline bool ext4_simulate_fail(struct super_block *sb,
1852 unsigned long code)
1853{
1854#ifdef CONFIG_EXT4_DEBUG
1855 struct ext4_sb_info *sbi = EXT4_SB(sb);
1856
1857 if (unlikely(sbi->s_simulate_fail == code)) {
1858 sbi->s_simulate_fail = 0;
1859 return true;
1860 }
1861#endif
1862 return false;
1863}
1864
1865static inline void ext4_simulate_fail_bh(struct super_block *sb,
1866 struct buffer_head *bh,
1867 unsigned long code)
1868{
1869 if (!IS_ERR(bh) && ext4_simulate_fail(sb, code))
1870 clear_buffer_uptodate(bh);
1871}
1872
1873/*
1874 * Error number codes for s_{first,last}_error_errno
1875 *
1876 * Linux errno numbers are architecture specific, so we need to translate
1877 * them into something which is architecture independent. We don't define
1878 * codes for all errno's; just the ones which are most likely to be the cause
1879 * of an ext4_error() call.
1880 */
1881#define EXT4_ERR_UNKNOWN 1
1882#define EXT4_ERR_EIO 2
1883#define EXT4_ERR_ENOMEM 3
1884#define EXT4_ERR_EFSBADCRC 4
1885#define EXT4_ERR_EFSCORRUPTED 5
1886#define EXT4_ERR_ENOSPC 6
1887#define EXT4_ERR_ENOKEY 7
1888#define EXT4_ERR_EROFS 8
1889#define EXT4_ERR_EFBIG 9
1890#define EXT4_ERR_EEXIST 10
1891#define EXT4_ERR_ERANGE 11
1892#define EXT4_ERR_EOVERFLOW 12
1893#define EXT4_ERR_EBUSY 13
1894#define EXT4_ERR_ENOTDIR 14
1895#define EXT4_ERR_ENOTEMPTY 15
1896#define EXT4_ERR_ESHUTDOWN 16
1897#define EXT4_ERR_EFAULT 17
1898
1899/*
1900 * Inode dynamic state flags
1901 */
1902enum {
1903 EXT4_STATE_NEW, /* inode is newly created */
1904 EXT4_STATE_XATTR, /* has in-inode xattrs */
1905 EXT4_STATE_NO_EXPAND, /* No space for expansion */
1906 EXT4_STATE_DA_ALLOC_CLOSE, /* Alloc DA blks on close */
1907 EXT4_STATE_EXT_MIGRATE, /* Inode is migrating */
1908 EXT4_STATE_NEWENTRY, /* File just added to dir */
1909 EXT4_STATE_MAY_INLINE_DATA, /* may have in-inode data */
1910 EXT4_STATE_EXT_PRECACHED, /* extents have been precached */
1911 EXT4_STATE_LUSTRE_EA_INODE, /* Lustre-style ea_inode */
1912 EXT4_STATE_VERITY_IN_PROGRESS, /* building fs-verity Merkle tree */
1913 EXT4_STATE_FC_COMMITTING, /* Fast commit ongoing */
1914 EXT4_STATE_ORPHAN_FILE, /* Inode orphaned in orphan file */
1915};
1916
1917#define EXT4_INODE_BIT_FNS(name, field, offset) \
1918static inline int ext4_test_inode_##name(struct inode *inode, int bit) \
1919{ \
1920 return test_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1921} \
1922static inline void ext4_set_inode_##name(struct inode *inode, int bit) \
1923{ \
1924 set_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1925} \
1926static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1927{ \
1928 clear_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1929}
1930
1931/* Add these declarations here only so that these functions can be
1932 * found by name. Otherwise, they are very hard to locate. */
1933static inline int ext4_test_inode_flag(struct inode *inode, int bit);
1934static inline void ext4_set_inode_flag(struct inode *inode, int bit);
1935static inline void ext4_clear_inode_flag(struct inode *inode, int bit);
1936EXT4_INODE_BIT_FNS(flag, flags, 0)
1937
1938/* Add these declarations here only so that these functions can be
1939 * found by name. Otherwise, they are very hard to locate. */
1940static inline int ext4_test_inode_state(struct inode *inode, int bit);
1941static inline void ext4_set_inode_state(struct inode *inode, int bit);
1942static inline void ext4_clear_inode_state(struct inode *inode, int bit);
1943#if (BITS_PER_LONG < 64)
1944EXT4_INODE_BIT_FNS(state, state_flags, 0)
1945
1946static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1947{
1948 (ei)->i_state_flags = 0;
1949}
1950#else
1951EXT4_INODE_BIT_FNS(state, flags, 32)
1952
1953static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1954{
1955 /* We depend on the fact that callers will set i_flags */
1956}
1957#endif
1958#else
1959/* Assume that user mode programs are passing in an ext4fs superblock, not
1960 * a kernel struct super_block. This will allow us to call the feature-test
1961 * macros from user land. */
1962#define EXT4_SB(sb) (sb)
1963#endif
1964
1965static inline bool ext4_verity_in_progress(struct inode *inode)
1966{
1967 return IS_ENABLED(CONFIG_FS_VERITY) &&
1968 ext4_test_inode_state(inode, EXT4_STATE_VERITY_IN_PROGRESS);
1969}
1970
1971#define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1972
1973/*
1974 * Codes for operating systems
1975 */
1976#define EXT4_OS_LINUX 0
1977#define EXT4_OS_HURD 1
1978#define EXT4_OS_MASIX 2
1979#define EXT4_OS_FREEBSD 3
1980#define EXT4_OS_LITES 4
1981
1982/*
1983 * Revision levels
1984 */
1985#define EXT4_GOOD_OLD_REV 0 /* The good old (original) format */
1986#define EXT4_DYNAMIC_REV 1 /* V2 format w/ dynamic inode sizes */
1987
1988#define EXT4_MAX_SUPP_REV EXT4_DYNAMIC_REV
1989
1990#define EXT4_GOOD_OLD_INODE_SIZE 128
1991
1992#define EXT4_EXTRA_TIMESTAMP_MAX (((s64)1 << 34) - 1 + S32_MIN)
1993#define EXT4_NON_EXTRA_TIMESTAMP_MAX S32_MAX
1994#define EXT4_TIMESTAMP_MIN S32_MIN
1995
1996/*
1997 * Feature set definitions
1998 */
1999
2000#define EXT4_FEATURE_COMPAT_DIR_PREALLOC 0x0001
2001#define EXT4_FEATURE_COMPAT_IMAGIC_INODES 0x0002
2002#define EXT4_FEATURE_COMPAT_HAS_JOURNAL 0x0004
2003#define EXT4_FEATURE_COMPAT_EXT_ATTR 0x0008
2004#define EXT4_FEATURE_COMPAT_RESIZE_INODE 0x0010
2005#define EXT4_FEATURE_COMPAT_DIR_INDEX 0x0020
2006#define EXT4_FEATURE_COMPAT_SPARSE_SUPER2 0x0200
2007/*
2008 * The reason why "FAST_COMMIT" is a compat feature is that, FS becomes
2009 * incompatible only if fast commit blocks are present in the FS. Since we
2010 * clear the journal (and thus the fast commit blocks), we don't mark FS as
2011 * incompatible. We also have a JBD2 incompat feature, which gets set when
2012 * there are fast commit blocks present in the journal.
2013 */
2014#define EXT4_FEATURE_COMPAT_FAST_COMMIT 0x0400
2015#define EXT4_FEATURE_COMPAT_STABLE_INODES 0x0800
2016#define EXT4_FEATURE_COMPAT_ORPHAN_FILE 0x1000 /* Orphan file exists */
2017
2018#define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER 0x0001
2019#define EXT4_FEATURE_RO_COMPAT_LARGE_FILE 0x0002
2020#define EXT4_FEATURE_RO_COMPAT_BTREE_DIR 0x0004
2021#define EXT4_FEATURE_RO_COMPAT_HUGE_FILE 0x0008
2022#define EXT4_FEATURE_RO_COMPAT_GDT_CSUM 0x0010
2023#define EXT4_FEATURE_RO_COMPAT_DIR_NLINK 0x0020
2024#define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE 0x0040
2025#define EXT4_FEATURE_RO_COMPAT_QUOTA 0x0100
2026#define EXT4_FEATURE_RO_COMPAT_BIGALLOC 0x0200
2027/*
2028 * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM). When
2029 * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
2030 * all other data structures' checksums. However, the METADATA_CSUM and
2031 * GDT_CSUM bits are mutually exclusive.
2032 */
2033#define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM 0x0400
2034#define EXT4_FEATURE_RO_COMPAT_READONLY 0x1000
2035#define EXT4_FEATURE_RO_COMPAT_PROJECT 0x2000
2036#define EXT4_FEATURE_RO_COMPAT_VERITY 0x8000
2037#define EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT 0x10000 /* Orphan file may be
2038 non-empty */
2039
2040#define EXT4_FEATURE_INCOMPAT_COMPRESSION 0x0001
2041#define EXT4_FEATURE_INCOMPAT_FILETYPE 0x0002
2042#define EXT4_FEATURE_INCOMPAT_RECOVER 0x0004 /* Needs recovery */
2043#define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 /* Journal device */
2044#define EXT4_FEATURE_INCOMPAT_META_BG 0x0010
2045#define EXT4_FEATURE_INCOMPAT_EXTENTS 0x0040 /* extents support */
2046#define EXT4_FEATURE_INCOMPAT_64BIT 0x0080
2047#define EXT4_FEATURE_INCOMPAT_MMP 0x0100
2048#define EXT4_FEATURE_INCOMPAT_FLEX_BG 0x0200
2049#define EXT4_FEATURE_INCOMPAT_EA_INODE 0x0400 /* EA in inode */
2050#define EXT4_FEATURE_INCOMPAT_DIRDATA 0x1000 /* data in dirent */
2051#define EXT4_FEATURE_INCOMPAT_CSUM_SEED 0x2000
2052#define EXT4_FEATURE_INCOMPAT_LARGEDIR 0x4000 /* >2GB or 3-lvl htree */
2053#define EXT4_FEATURE_INCOMPAT_INLINE_DATA 0x8000 /* data in inode */
2054#define EXT4_FEATURE_INCOMPAT_ENCRYPT 0x10000
2055#define EXT4_FEATURE_INCOMPAT_CASEFOLD 0x20000
2056
2057extern void ext4_update_dynamic_rev(struct super_block *sb);
2058
2059#define EXT4_FEATURE_COMPAT_FUNCS(name, flagname) \
2060static inline bool ext4_has_feature_##name(struct super_block *sb) \
2061{ \
2062 return ((EXT4_SB(sb)->s_es->s_feature_compat & \
2063 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname)) != 0); \
2064} \
2065static inline void ext4_set_feature_##name(struct super_block *sb) \
2066{ \
2067 ext4_update_dynamic_rev(sb); \
2068 EXT4_SB(sb)->s_es->s_feature_compat |= \
2069 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
2070} \
2071static inline void ext4_clear_feature_##name(struct super_block *sb) \
2072{ \
2073 EXT4_SB(sb)->s_es->s_feature_compat &= \
2074 ~cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
2075}
2076
2077#define EXT4_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
2078static inline bool ext4_has_feature_##name(struct super_block *sb) \
2079{ \
2080 return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
2081 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname)) != 0); \
2082} \
2083static inline void ext4_set_feature_##name(struct super_block *sb) \
2084{ \
2085 ext4_update_dynamic_rev(sb); \
2086 EXT4_SB(sb)->s_es->s_feature_ro_compat |= \
2087 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
2088} \
2089static inline void ext4_clear_feature_##name(struct super_block *sb) \
2090{ \
2091 EXT4_SB(sb)->s_es->s_feature_ro_compat &= \
2092 ~cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
2093}
2094
2095#define EXT4_FEATURE_INCOMPAT_FUNCS(name, flagname) \
2096static inline bool ext4_has_feature_##name(struct super_block *sb) \
2097{ \
2098 return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
2099 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname)) != 0); \
2100} \
2101static inline void ext4_set_feature_##name(struct super_block *sb) \
2102{ \
2103 ext4_update_dynamic_rev(sb); \
2104 EXT4_SB(sb)->s_es->s_feature_incompat |= \
2105 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
2106} \
2107static inline void ext4_clear_feature_##name(struct super_block *sb) \
2108{ \
2109 EXT4_SB(sb)->s_es->s_feature_incompat &= \
2110 ~cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
2111}
2112
2113EXT4_FEATURE_COMPAT_FUNCS(dir_prealloc, DIR_PREALLOC)
2114EXT4_FEATURE_COMPAT_FUNCS(imagic_inodes, IMAGIC_INODES)
2115EXT4_FEATURE_COMPAT_FUNCS(journal, HAS_JOURNAL)
2116EXT4_FEATURE_COMPAT_FUNCS(xattr, EXT_ATTR)
2117EXT4_FEATURE_COMPAT_FUNCS(resize_inode, RESIZE_INODE)
2118EXT4_FEATURE_COMPAT_FUNCS(dir_index, DIR_INDEX)
2119EXT4_FEATURE_COMPAT_FUNCS(sparse_super2, SPARSE_SUPER2)
2120EXT4_FEATURE_COMPAT_FUNCS(fast_commit, FAST_COMMIT)
2121EXT4_FEATURE_COMPAT_FUNCS(stable_inodes, STABLE_INODES)
2122EXT4_FEATURE_COMPAT_FUNCS(orphan_file, ORPHAN_FILE)
2123
2124EXT4_FEATURE_RO_COMPAT_FUNCS(sparse_super, SPARSE_SUPER)
2125EXT4_FEATURE_RO_COMPAT_FUNCS(large_file, LARGE_FILE)
2126EXT4_FEATURE_RO_COMPAT_FUNCS(btree_dir, BTREE_DIR)
2127EXT4_FEATURE_RO_COMPAT_FUNCS(huge_file, HUGE_FILE)
2128EXT4_FEATURE_RO_COMPAT_FUNCS(gdt_csum, GDT_CSUM)
2129EXT4_FEATURE_RO_COMPAT_FUNCS(dir_nlink, DIR_NLINK)
2130EXT4_FEATURE_RO_COMPAT_FUNCS(extra_isize, EXTRA_ISIZE)
2131EXT4_FEATURE_RO_COMPAT_FUNCS(quota, QUOTA)
2132EXT4_FEATURE_RO_COMPAT_FUNCS(bigalloc, BIGALLOC)
2133EXT4_FEATURE_RO_COMPAT_FUNCS(metadata_csum, METADATA_CSUM)
2134EXT4_FEATURE_RO_COMPAT_FUNCS(readonly, READONLY)
2135EXT4_FEATURE_RO_COMPAT_FUNCS(project, PROJECT)
2136EXT4_FEATURE_RO_COMPAT_FUNCS(verity, VERITY)
2137EXT4_FEATURE_RO_COMPAT_FUNCS(orphan_present, ORPHAN_PRESENT)
2138
2139EXT4_FEATURE_INCOMPAT_FUNCS(compression, COMPRESSION)
2140EXT4_FEATURE_INCOMPAT_FUNCS(filetype, FILETYPE)
2141EXT4_FEATURE_INCOMPAT_FUNCS(journal_needs_recovery, RECOVER)
2142EXT4_FEATURE_INCOMPAT_FUNCS(journal_dev, JOURNAL_DEV)
2143EXT4_FEATURE_INCOMPAT_FUNCS(meta_bg, META_BG)
2144EXT4_FEATURE_INCOMPAT_FUNCS(extents, EXTENTS)
2145EXT4_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT)
2146EXT4_FEATURE_INCOMPAT_FUNCS(mmp, MMP)
2147EXT4_FEATURE_INCOMPAT_FUNCS(flex_bg, FLEX_BG)
2148EXT4_FEATURE_INCOMPAT_FUNCS(ea_inode, EA_INODE)
2149EXT4_FEATURE_INCOMPAT_FUNCS(dirdata, DIRDATA)
2150EXT4_FEATURE_INCOMPAT_FUNCS(csum_seed, CSUM_SEED)
2151EXT4_FEATURE_INCOMPAT_FUNCS(largedir, LARGEDIR)
2152EXT4_FEATURE_INCOMPAT_FUNCS(inline_data, INLINE_DATA)
2153EXT4_FEATURE_INCOMPAT_FUNCS(encrypt, ENCRYPT)
2154EXT4_FEATURE_INCOMPAT_FUNCS(casefold, CASEFOLD)
2155
2156#define EXT2_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR
2157#define EXT2_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
2158 EXT4_FEATURE_INCOMPAT_META_BG)
2159#define EXT2_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
2160 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
2161 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
2162
2163#define EXT3_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR
2164#define EXT3_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
2165 EXT4_FEATURE_INCOMPAT_RECOVER| \
2166 EXT4_FEATURE_INCOMPAT_META_BG)
2167#define EXT3_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
2168 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
2169 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
2170
2171#define EXT4_FEATURE_COMPAT_SUPP (EXT4_FEATURE_COMPAT_EXT_ATTR| \
2172 EXT4_FEATURE_COMPAT_ORPHAN_FILE)
2173#define EXT4_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
2174 EXT4_FEATURE_INCOMPAT_RECOVER| \
2175 EXT4_FEATURE_INCOMPAT_META_BG| \
2176 EXT4_FEATURE_INCOMPAT_EXTENTS| \
2177 EXT4_FEATURE_INCOMPAT_64BIT| \
2178 EXT4_FEATURE_INCOMPAT_FLEX_BG| \
2179 EXT4_FEATURE_INCOMPAT_EA_INODE| \
2180 EXT4_FEATURE_INCOMPAT_MMP | \
2181 EXT4_FEATURE_INCOMPAT_INLINE_DATA | \
2182 EXT4_FEATURE_INCOMPAT_ENCRYPT | \
2183 EXT4_FEATURE_INCOMPAT_CASEFOLD | \
2184 EXT4_FEATURE_INCOMPAT_CSUM_SEED | \
2185 EXT4_FEATURE_INCOMPAT_LARGEDIR)
2186#define EXT4_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
2187 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
2188 EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
2189 EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
2190 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
2191 EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
2192 EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
2193 EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
2194 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
2195 EXT4_FEATURE_RO_COMPAT_QUOTA |\
2196 EXT4_FEATURE_RO_COMPAT_PROJECT |\
2197 EXT4_FEATURE_RO_COMPAT_VERITY |\
2198 EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT)
2199
2200#define EXTN_FEATURE_FUNCS(ver) \
2201static inline bool ext4_has_unknown_ext##ver##_compat_features(struct super_block *sb) \
2202{ \
2203 return ((EXT4_SB(sb)->s_es->s_feature_compat & \
2204 cpu_to_le32(~EXT##ver##_FEATURE_COMPAT_SUPP)) != 0); \
2205} \
2206static inline bool ext4_has_unknown_ext##ver##_ro_compat_features(struct super_block *sb) \
2207{ \
2208 return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
2209 cpu_to_le32(~EXT##ver##_FEATURE_RO_COMPAT_SUPP)) != 0); \
2210} \
2211static inline bool ext4_has_unknown_ext##ver##_incompat_features(struct super_block *sb) \
2212{ \
2213 return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
2214 cpu_to_le32(~EXT##ver##_FEATURE_INCOMPAT_SUPP)) != 0); \
2215}
2216
2217EXTN_FEATURE_FUNCS(2)
2218EXTN_FEATURE_FUNCS(3)
2219EXTN_FEATURE_FUNCS(4)
2220
2221static inline bool ext4_has_compat_features(struct super_block *sb)
2222{
2223 return (EXT4_SB(sb)->s_es->s_feature_compat != 0);
2224}
2225static inline bool ext4_has_ro_compat_features(struct super_block *sb)
2226{
2227 return (EXT4_SB(sb)->s_es->s_feature_ro_compat != 0);
2228}
2229static inline bool ext4_has_incompat_features(struct super_block *sb)
2230{
2231 return (EXT4_SB(sb)->s_es->s_feature_incompat != 0);
2232}
2233
2234extern int ext4_feature_set_ok(struct super_block *sb, int readonly);
2235
2236/*
2237 * Superblock flags
2238 */
2239#define EXT4_FLAGS_RESIZING 0
2240#define EXT4_FLAGS_SHUTDOWN 1
2241#define EXT4_FLAGS_BDEV_IS_DAX 2
2242
2243static inline int ext4_forced_shutdown(struct super_block *sb)
2244{
2245 return test_bit(EXT4_FLAGS_SHUTDOWN, &EXT4_SB(sb)->s_ext4_flags);
2246}
2247
2248/*
2249 * Default values for user and/or group using reserved blocks
2250 */
2251#define EXT4_DEF_RESUID 0
2252#define EXT4_DEF_RESGID 0
2253
2254/*
2255 * Default project ID
2256 */
2257#define EXT4_DEF_PROJID 0
2258
2259#define EXT4_DEF_INODE_READAHEAD_BLKS 32
2260
2261/*
2262 * Default mount options
2263 */
2264#define EXT4_DEFM_DEBUG 0x0001
2265#define EXT4_DEFM_BSDGROUPS 0x0002
2266#define EXT4_DEFM_XATTR_USER 0x0004
2267#define EXT4_DEFM_ACL 0x0008
2268#define EXT4_DEFM_UID16 0x0010
2269#define EXT4_DEFM_JMODE 0x0060
2270#define EXT4_DEFM_JMODE_DATA 0x0020
2271#define EXT4_DEFM_JMODE_ORDERED 0x0040
2272#define EXT4_DEFM_JMODE_WBACK 0x0060
2273#define EXT4_DEFM_NOBARRIER 0x0100
2274#define EXT4_DEFM_BLOCK_VALIDITY 0x0200
2275#define EXT4_DEFM_DISCARD 0x0400
2276#define EXT4_DEFM_NODELALLOC 0x0800
2277
2278/*
2279 * Default journal batch times
2280 */
2281#define EXT4_DEF_MIN_BATCH_TIME 0
2282#define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
2283
2284/*
2285 * Minimum number of groups in a flexgroup before we separate out
2286 * directories into the first block group of a flexgroup
2287 */
2288#define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
2289
2290/*
2291 * Structure of a directory entry
2292 */
2293#define EXT4_NAME_LEN 255
2294/*
2295 * Base length of the ext4 directory entry excluding the name length
2296 */
2297#define EXT4_BASE_DIR_LEN (sizeof(struct ext4_dir_entry_2) - EXT4_NAME_LEN)
2298
2299struct ext4_dir_entry {
2300 __le32 inode; /* Inode number */
2301 __le16 rec_len; /* Directory entry length */
2302 __le16 name_len; /* Name length */
2303 char name[EXT4_NAME_LEN]; /* File name */
2304};
2305
2306
2307/*
2308 * Encrypted Casefolded entries require saving the hash on disk. This structure
2309 * followed ext4_dir_entry_2's name[name_len] at the next 4 byte aligned
2310 * boundary.
2311 */
2312struct ext4_dir_entry_hash {
2313 __le32 hash;
2314 __le32 minor_hash;
2315};
2316
2317/*
2318 * The new version of the directory entry. Since EXT4 structures are
2319 * stored in intel byte order, and the name_len field could never be
2320 * bigger than 255 chars, it's safe to reclaim the extra byte for the
2321 * file_type field.
2322 */
2323struct ext4_dir_entry_2 {
2324 __le32 inode; /* Inode number */
2325 __le16 rec_len; /* Directory entry length */
2326 __u8 name_len; /* Name length */
2327 __u8 file_type; /* See file type macros EXT4_FT_* below */
2328 char name[EXT4_NAME_LEN]; /* File name */
2329};
2330
2331/*
2332 * Access the hashes at the end of ext4_dir_entry_2
2333 */
2334#define EXT4_DIRENT_HASHES(entry) \
2335 ((struct ext4_dir_entry_hash *) \
2336 (((void *)(entry)) + \
2337 ((8 + (entry)->name_len + EXT4_DIR_ROUND) & ~EXT4_DIR_ROUND)))
2338#define EXT4_DIRENT_HASH(entry) le32_to_cpu(EXT4_DIRENT_HASHES(de)->hash)
2339#define EXT4_DIRENT_MINOR_HASH(entry) \
2340 le32_to_cpu(EXT4_DIRENT_HASHES(de)->minor_hash)
2341
2342static inline bool ext4_hash_in_dirent(const struct inode *inode)
2343{
2344 return IS_CASEFOLDED(inode) && IS_ENCRYPTED(inode);
2345}
2346
2347/*
2348 * This is a bogus directory entry at the end of each leaf block that
2349 * records checksums.
2350 */
2351struct ext4_dir_entry_tail {
2352 __le32 det_reserved_zero1; /* Pretend to be unused */
2353 __le16 det_rec_len; /* 12 */
2354 __u8 det_reserved_zero2; /* Zero name length */
2355 __u8 det_reserved_ft; /* 0xDE, fake file type */
2356 __le32 det_checksum; /* crc32c(uuid+inum+dirblock) */
2357};
2358
2359#define EXT4_DIRENT_TAIL(block, blocksize) \
2360 ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
2361 ((blocksize) - \
2362 sizeof(struct ext4_dir_entry_tail))))
2363
2364/*
2365 * Ext4 directory file types. Only the low 3 bits are used. The
2366 * other bits are reserved for now.
2367 */
2368#define EXT4_FT_UNKNOWN 0
2369#define EXT4_FT_REG_FILE 1
2370#define EXT4_FT_DIR 2
2371#define EXT4_FT_CHRDEV 3
2372#define EXT4_FT_BLKDEV 4
2373#define EXT4_FT_FIFO 5
2374#define EXT4_FT_SOCK 6
2375#define EXT4_FT_SYMLINK 7
2376
2377#define EXT4_FT_MAX 8
2378
2379#define EXT4_FT_DIR_CSUM 0xDE
2380
2381/*
2382 * EXT4_DIR_PAD defines the directory entries boundaries
2383 *
2384 * NOTE: It must be a multiple of 4
2385 */
2386#define EXT4_DIR_PAD 4
2387#define EXT4_DIR_ROUND (EXT4_DIR_PAD - 1)
2388#define EXT4_MAX_REC_LEN ((1<<16)-1)
2389
2390/*
2391 * The rec_len is dependent on the type of directory. Directories that are
2392 * casefolded and encrypted need to store the hash as well, so we add room for
2393 * ext4_extended_dir_entry_2. For all entries related to '.' or '..' you should
2394 * pass NULL for dir, as those entries do not use the extra fields.
2395 */
2396static inline unsigned int ext4_dir_rec_len(__u8 name_len,
2397 const struct inode *dir)
2398{
2399 int rec_len = (name_len + 8 + EXT4_DIR_ROUND);
2400
2401 if (dir && ext4_hash_in_dirent(dir))
2402 rec_len += sizeof(struct ext4_dir_entry_hash);
2403 return (rec_len & ~EXT4_DIR_ROUND);
2404}
2405
2406/*
2407 * If we ever get support for fs block sizes > page_size, we'll need
2408 * to remove the #if statements in the next two functions...
2409 */
2410static inline unsigned int
2411ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
2412{
2413 unsigned len = le16_to_cpu(dlen);
2414
2415#if (PAGE_SIZE >= 65536)
2416 if (len == EXT4_MAX_REC_LEN || len == 0)
2417 return blocksize;
2418 return (len & 65532) | ((len & 3) << 16);
2419#else
2420 return len;
2421#endif
2422}
2423
2424static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
2425{
2426 BUG_ON((len > blocksize) || (blocksize > (1 << 18)) || (len & 3));
2427#if (PAGE_SIZE >= 65536)
2428 if (len < 65536)
2429 return cpu_to_le16(len);
2430 if (len == blocksize) {
2431 if (blocksize == 65536)
2432 return cpu_to_le16(EXT4_MAX_REC_LEN);
2433 else
2434 return cpu_to_le16(0);
2435 }
2436 return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
2437#else
2438 return cpu_to_le16(len);
2439#endif
2440}
2441
2442/*
2443 * Hash Tree Directory indexing
2444 * (c) Daniel Phillips, 2001
2445 */
2446
2447#define is_dx(dir) (ext4_has_feature_dir_index((dir)->i_sb) && \
2448 ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
2449#define EXT4_DIR_LINK_MAX(dir) unlikely((dir)->i_nlink >= EXT4_LINK_MAX && \
2450 !(ext4_has_feature_dir_nlink((dir)->i_sb) && is_dx(dir)))
2451#define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
2452
2453/* Legal values for the dx_root hash_version field: */
2454
2455#define DX_HASH_LEGACY 0
2456#define DX_HASH_HALF_MD4 1
2457#define DX_HASH_TEA 2
2458#define DX_HASH_LEGACY_UNSIGNED 3
2459#define DX_HASH_HALF_MD4_UNSIGNED 4
2460#define DX_HASH_TEA_UNSIGNED 5
2461#define DX_HASH_SIPHASH 6
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 fscrypt_str 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};
2554
2555/* calculate the first block number of the group */
2556static inline ext4_fsblk_t
2557ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
2558{
2559 return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
2560 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
2561}
2562
2563/*
2564 * Special error return code only used by dx_probe() and its callers.
2565 */
2566#define ERR_BAD_DX_DIR (-(MAX_ERRNO - 1))
2567
2568/* htree levels for ext4 */
2569#define EXT4_HTREE_LEVEL_COMPAT 2
2570#define EXT4_HTREE_LEVEL 3
2571
2572static inline int ext4_dir_htree_level(struct super_block *sb)
2573{
2574 return ext4_has_feature_largedir(sb) ?
2575 EXT4_HTREE_LEVEL : EXT4_HTREE_LEVEL_COMPAT;
2576}
2577
2578/*
2579 * Timeout and state flag for lazy initialization inode thread.
2580 */
2581#define EXT4_DEF_LI_WAIT_MULT 10
2582#define EXT4_DEF_LI_MAX_START_DELAY 5
2583#define EXT4_LAZYINIT_QUIT 0x0001
2584#define EXT4_LAZYINIT_RUNNING 0x0002
2585
2586/*
2587 * Lazy inode table initialization info
2588 */
2589struct ext4_lazy_init {
2590 unsigned long li_state;
2591 struct list_head li_request_list;
2592 struct mutex li_list_mtx;
2593};
2594
2595enum ext4_li_mode {
2596 EXT4_LI_MODE_PREFETCH_BBITMAP,
2597 EXT4_LI_MODE_ITABLE,
2598};
2599
2600struct ext4_li_request {
2601 struct super_block *lr_super;
2602 enum ext4_li_mode lr_mode;
2603 ext4_group_t lr_first_not_zeroed;
2604 ext4_group_t lr_next_group;
2605 struct list_head lr_request;
2606 unsigned long lr_next_sched;
2607 unsigned long lr_timeout;
2608};
2609
2610struct ext4_features {
2611 struct kobject f_kobj;
2612 struct completion f_kobj_unregister;
2613};
2614
2615/*
2616 * This structure will be used for multiple mount protection. It will be
2617 * written into the block number saved in the s_mmp_block field in the
2618 * superblock. Programs that check MMP should assume that if
2619 * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
2620 * to use the filesystem, regardless of how old the timestamp is.
2621 */
2622#define EXT4_MMP_MAGIC 0x004D4D50U /* ASCII for MMP */
2623#define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
2624#define EXT4_MMP_SEQ_FSCK 0xE24D4D50U /* mmp_seq value when being fscked */
2625#define EXT4_MMP_SEQ_MAX 0xE24D4D4FU /* maximum valid mmp_seq value */
2626
2627struct mmp_struct {
2628 __le32 mmp_magic; /* Magic number for MMP */
2629 __le32 mmp_seq; /* Sequence no. updated periodically */
2630
2631 /*
2632 * mmp_time, mmp_nodename & mmp_bdevname are only used for information
2633 * purposes and do not affect the correctness of the algorithm
2634 */
2635 __le64 mmp_time; /* Time last updated */
2636 char mmp_nodename[64]; /* Node which last updated MMP block */
2637 char mmp_bdevname[32]; /* Bdev which last updated MMP block */
2638
2639 /*
2640 * mmp_check_interval is used to verify if the MMP block has been
2641 * updated on the block device. The value is updated based on the
2642 * maximum time to write the MMP block during an update cycle.
2643 */
2644 __le16 mmp_check_interval;
2645
2646 __le16 mmp_pad1;
2647 __le32 mmp_pad2[226];
2648 __le32 mmp_checksum; /* crc32c(uuid+mmp_block) */
2649};
2650
2651/* arguments passed to the mmp thread */
2652struct mmpd_data {
2653 struct buffer_head *bh; /* bh from initial read_mmp_block() */
2654 struct super_block *sb; /* super block of the fs */
2655};
2656
2657/*
2658 * Check interval multiplier
2659 * The MMP block is written every update interval and initially checked every
2660 * update interval x the multiplier (the value is then adapted based on the
2661 * write latency). The reason is that writes can be delayed under load and we
2662 * don't want readers to incorrectly assume that the filesystem is no longer
2663 * in use.
2664 */
2665#define EXT4_MMP_CHECK_MULT 2UL
2666
2667/*
2668 * Minimum interval for MMP checking in seconds.
2669 */
2670#define EXT4_MMP_MIN_CHECK_INTERVAL 5UL
2671
2672/*
2673 * Maximum interval for MMP checking in seconds.
2674 */
2675#define EXT4_MMP_MAX_CHECK_INTERVAL 300UL
2676
2677/*
2678 * Function prototypes
2679 */
2680
2681/*
2682 * Ok, these declarations are also in <linux/kernel.h> but none of the
2683 * ext4 source programs needs to include it so they are duplicated here.
2684 */
2685# define NORET_TYPE /**/
2686# define ATTRIB_NORET __attribute__((noreturn))
2687# define NORET_AND noreturn,
2688
2689/* bitmap.c */
2690extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
2691void ext4_inode_bitmap_csum_set(struct super_block *sb,
2692 struct ext4_group_desc *gdp,
2693 struct buffer_head *bh, int sz);
2694int ext4_inode_bitmap_csum_verify(struct super_block *sb,
2695 struct ext4_group_desc *gdp,
2696 struct buffer_head *bh, int sz);
2697void ext4_block_bitmap_csum_set(struct super_block *sb,
2698 struct ext4_group_desc *gdp,
2699 struct buffer_head *bh);
2700int ext4_block_bitmap_csum_verify(struct super_block *sb,
2701 struct ext4_group_desc *gdp,
2702 struct buffer_head *bh);
2703
2704/* balloc.c */
2705extern void ext4_get_group_no_and_offset(struct super_block *sb,
2706 ext4_fsblk_t blocknr,
2707 ext4_group_t *blockgrpp,
2708 ext4_grpblk_t *offsetp);
2709extern ext4_group_t ext4_get_group_number(struct super_block *sb,
2710 ext4_fsblk_t block);
2711
2712extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
2713extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
2714 ext4_group_t group);
2715extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
2716 ext4_fsblk_t goal,
2717 unsigned int flags,
2718 unsigned long *count,
2719 int *errp);
2720extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
2721 s64 nclusters, unsigned int flags);
2722extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
2723extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
2724 ext4_group_t block_group,
2725 struct buffer_head ** bh);
2726extern struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2727 ext4_group_t group);
2728extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
2729
2730extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
2731 ext4_group_t block_group,
2732 bool ignore_locked);
2733extern int ext4_wait_block_bitmap(struct super_block *sb,
2734 ext4_group_t block_group,
2735 struct buffer_head *bh);
2736extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
2737 ext4_group_t block_group);
2738extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
2739 ext4_group_t block_group,
2740 struct ext4_group_desc *gdp);
2741ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
2742
2743#if IS_ENABLED(CONFIG_UNICODE)
2744extern int ext4_fname_setup_ci_filename(struct inode *dir,
2745 const struct qstr *iname,
2746 struct ext4_filename *fname);
2747#endif
2748
2749/* ext4 encryption related stuff goes here crypto.c */
2750#ifdef CONFIG_FS_ENCRYPTION
2751extern const struct fscrypt_operations ext4_cryptops;
2752
2753int ext4_fname_setup_filename(struct inode *dir, const struct qstr *iname,
2754 int lookup, struct ext4_filename *fname);
2755
2756int ext4_fname_prepare_lookup(struct inode *dir, struct dentry *dentry,
2757 struct ext4_filename *fname);
2758
2759void ext4_fname_free_filename(struct ext4_filename *fname);
2760
2761int ext4_ioctl_get_encryption_pwsalt(struct file *filp, void __user *arg);
2762
2763#else /* !CONFIG_FS_ENCRYPTION */
2764static inline int ext4_fname_setup_filename(struct inode *dir,
2765 const struct qstr *iname,
2766 int lookup,
2767 struct ext4_filename *fname)
2768{
2769 int err = 0;
2770 fname->usr_fname = iname;
2771 fname->disk_name.name = (unsigned char *) iname->name;
2772 fname->disk_name.len = iname->len;
2773
2774#if IS_ENABLED(CONFIG_UNICODE)
2775 err = ext4_fname_setup_ci_filename(dir, iname, fname);
2776#endif
2777
2778 return err;
2779}
2780
2781static inline int ext4_fname_prepare_lookup(struct inode *dir,
2782 struct dentry *dentry,
2783 struct ext4_filename *fname)
2784{
2785 return ext4_fname_setup_filename(dir, &dentry->d_name, 1, fname);
2786}
2787
2788static inline void ext4_fname_free_filename(struct ext4_filename *fname)
2789{
2790#if IS_ENABLED(CONFIG_UNICODE)
2791 kfree(fname->cf_name.name);
2792 fname->cf_name.name = NULL;
2793#endif
2794}
2795
2796static inline int ext4_ioctl_get_encryption_pwsalt(struct file *filp,
2797 void __user *arg)
2798{
2799 return -EOPNOTSUPP;
2800}
2801#endif /* !CONFIG_FS_ENCRYPTION */
2802
2803/* dir.c */
2804extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
2805 struct file *,
2806 struct ext4_dir_entry_2 *,
2807 struct buffer_head *, char *, int,
2808 unsigned int);
2809#define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset) \
2810 unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
2811 (de), (bh), (buf), (size), (offset)))
2812extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
2813 __u32 minor_hash,
2814 struct ext4_dir_entry_2 *dirent,
2815 struct fscrypt_str *ent_name);
2816extern void ext4_htree_free_dir_info(struct dir_private_info *p);
2817extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
2818 struct buffer_head *bh,
2819 void *buf, int buf_size,
2820 struct ext4_filename *fname,
2821 struct ext4_dir_entry_2 **dest_de);
2822void ext4_insert_dentry(struct inode *dir, struct inode *inode,
2823 struct ext4_dir_entry_2 *de,
2824 int buf_size,
2825 struct ext4_filename *fname);
2826static inline void ext4_update_dx_flag(struct inode *inode)
2827{
2828 if (!ext4_has_feature_dir_index(inode->i_sb) &&
2829 ext4_test_inode_flag(inode, EXT4_INODE_INDEX)) {
2830 /* ext4_iget() should have caught this... */
2831 WARN_ON_ONCE(ext4_has_feature_metadata_csum(inode->i_sb));
2832 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
2833 }
2834}
2835static const unsigned char ext4_filetype_table[] = {
2836 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
2837};
2838
2839static inline unsigned char get_dtype(struct super_block *sb, int filetype)
2840{
2841 if (!ext4_has_feature_filetype(sb) || filetype >= EXT4_FT_MAX)
2842 return DT_UNKNOWN;
2843
2844 return ext4_filetype_table[filetype];
2845}
2846extern int ext4_check_all_de(struct inode *dir, struct buffer_head *bh,
2847 void *buf, int buf_size);
2848
2849/* fsync.c */
2850extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2851
2852/* hash.c */
2853extern int ext4fs_dirhash(const struct inode *dir, const char *name, int len,
2854 struct dx_hash_info *hinfo);
2855
2856/* ialloc.c */
2857extern int ext4_mark_inode_used(struct super_block *sb, int ino);
2858extern struct inode *__ext4_new_inode(struct mnt_idmap *, handle_t *,
2859 struct inode *, umode_t,
2860 const struct qstr *qstr, __u32 goal,
2861 uid_t *owner, __u32 i_flags,
2862 int handle_type, unsigned int line_no,
2863 int nblocks);
2864
2865#define ext4_new_inode(handle, dir, mode, qstr, goal, owner, i_flags) \
2866 __ext4_new_inode(&nop_mnt_idmap, (handle), (dir), (mode), (qstr), \
2867 (goal), (owner), i_flags, 0, 0, 0)
2868#define ext4_new_inode_start_handle(idmap, dir, mode, qstr, goal, owner, \
2869 type, nblocks) \
2870 __ext4_new_inode((idmap), NULL, (dir), (mode), (qstr), (goal), (owner), \
2871 0, (type), __LINE__, (nblocks))
2872
2873
2874extern void ext4_free_inode(handle_t *, struct inode *);
2875extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2876extern unsigned long ext4_count_free_inodes(struct super_block *);
2877extern unsigned long ext4_count_dirs(struct super_block *);
2878extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2879extern int ext4_init_inode_table(struct super_block *sb,
2880 ext4_group_t group, int barrier);
2881extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2882
2883/* fast_commit.c */
2884int ext4_fc_info_show(struct seq_file *seq, void *v);
2885void ext4_fc_init(struct super_block *sb, journal_t *journal);
2886void ext4_fc_init_inode(struct inode *inode);
2887void ext4_fc_track_range(handle_t *handle, struct inode *inode, ext4_lblk_t start,
2888 ext4_lblk_t end);
2889void __ext4_fc_track_unlink(handle_t *handle, struct inode *inode,
2890 struct dentry *dentry);
2891void __ext4_fc_track_link(handle_t *handle, struct inode *inode,
2892 struct dentry *dentry);
2893void ext4_fc_track_unlink(handle_t *handle, struct dentry *dentry);
2894void ext4_fc_track_link(handle_t *handle, struct dentry *dentry);
2895void __ext4_fc_track_create(handle_t *handle, struct inode *inode,
2896 struct dentry *dentry);
2897void ext4_fc_track_create(handle_t *handle, struct dentry *dentry);
2898void ext4_fc_track_inode(handle_t *handle, struct inode *inode);
2899void ext4_fc_mark_ineligible(struct super_block *sb, int reason, handle_t *handle);
2900void ext4_fc_start_update(struct inode *inode);
2901void ext4_fc_stop_update(struct inode *inode);
2902void ext4_fc_del(struct inode *inode);
2903bool ext4_fc_replay_check_excluded(struct super_block *sb, ext4_fsblk_t block);
2904void ext4_fc_replay_cleanup(struct super_block *sb);
2905int ext4_fc_commit(journal_t *journal, tid_t commit_tid);
2906int __init ext4_fc_init_dentry_cache(void);
2907void ext4_fc_destroy_dentry_cache(void);
2908int ext4_fc_record_regions(struct super_block *sb, int ino,
2909 ext4_lblk_t lblk, ext4_fsblk_t pblk,
2910 int len, int replay);
2911
2912/* mballoc.c */
2913extern const struct seq_operations ext4_mb_seq_groups_ops;
2914extern const struct seq_operations ext4_mb_seq_structs_summary_ops;
2915extern int ext4_seq_mb_stats_show(struct seq_file *seq, void *offset);
2916extern int ext4_mb_init(struct super_block *);
2917extern void ext4_mb_release(struct super_block *);
2918extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2919 struct ext4_allocation_request *, int *);
2920extern void ext4_discard_preallocations(struct inode *);
2921extern int __init ext4_init_mballoc(void);
2922extern void ext4_exit_mballoc(void);
2923extern ext4_group_t ext4_mb_prefetch(struct super_block *sb,
2924 ext4_group_t group,
2925 unsigned int nr, int *cnt);
2926extern void ext4_mb_prefetch_fini(struct super_block *sb, ext4_group_t group,
2927 unsigned int nr);
2928
2929extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2930 struct buffer_head *bh, ext4_fsblk_t block,
2931 unsigned long count, int flags);
2932extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2933 ext4_group_t ngroups);
2934extern int ext4_mb_add_groupinfo(struct super_block *sb,
2935 ext4_group_t i, struct ext4_group_desc *desc);
2936extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2937 ext4_fsblk_t block, unsigned long count);
2938extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2939extern void ext4_process_freed_data(struct super_block *sb, tid_t commit_tid);
2940extern void ext4_mb_mark_bb(struct super_block *sb, ext4_fsblk_t block,
2941 int len, bool state);
2942static inline bool ext4_mb_cr_expensive(enum criteria cr)
2943{
2944 return cr >= CR_GOAL_LEN_SLOW;
2945}
2946
2947/* inode.c */
2948void ext4_inode_csum_set(struct inode *inode, struct ext4_inode *raw,
2949 struct ext4_inode_info *ei);
2950int ext4_inode_is_fast_symlink(struct inode *inode);
2951struct buffer_head *ext4_getblk(handle_t *, struct inode *, ext4_lblk_t, int);
2952struct buffer_head *ext4_bread(handle_t *, struct inode *, ext4_lblk_t, int);
2953int ext4_bread_batch(struct inode *inode, ext4_lblk_t block, int bh_count,
2954 bool wait, struct buffer_head **bhs);
2955int ext4_get_block_unwritten(struct inode *inode, sector_t iblock,
2956 struct buffer_head *bh_result, int create);
2957int ext4_get_block(struct inode *inode, sector_t iblock,
2958 struct buffer_head *bh_result, int create);
2959int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2960 struct buffer_head *bh, int create);
2961int ext4_walk_page_buffers(handle_t *handle,
2962 struct inode *inode,
2963 struct buffer_head *head,
2964 unsigned from,
2965 unsigned to,
2966 int *partial,
2967 int (*fn)(handle_t *handle, struct inode *inode,
2968 struct buffer_head *bh));
2969int do_journal_get_write_access(handle_t *handle, struct inode *inode,
2970 struct buffer_head *bh);
2971#define FALL_BACK_TO_NONDELALLOC 1
2972#define CONVERT_INLINE_DATA 2
2973
2974typedef enum {
2975 EXT4_IGET_NORMAL = 0,
2976 EXT4_IGET_SPECIAL = 0x0001, /* OK to iget a system inode */
2977 EXT4_IGET_HANDLE = 0x0002, /* Inode # is from a handle */
2978 EXT4_IGET_BAD = 0x0004, /* Allow to iget a bad inode */
2979 EXT4_IGET_EA_INODE = 0x0008 /* Inode should contain an EA value */
2980} ext4_iget_flags;
2981
2982extern struct inode *__ext4_iget(struct super_block *sb, unsigned long ino,
2983 ext4_iget_flags flags, const char *function,
2984 unsigned int line);
2985
2986#define ext4_iget(sb, ino, flags) \
2987 __ext4_iget((sb), (ino), (flags), __func__, __LINE__)
2988
2989extern int ext4_write_inode(struct inode *, struct writeback_control *);
2990extern int ext4_setattr(struct mnt_idmap *, struct dentry *,
2991 struct iattr *);
2992extern u32 ext4_dio_alignment(struct inode *inode);
2993extern int ext4_getattr(struct mnt_idmap *, const struct path *,
2994 struct kstat *, u32, unsigned int);
2995extern void ext4_evict_inode(struct inode *);
2996extern void ext4_clear_inode(struct inode *);
2997extern int ext4_file_getattr(struct mnt_idmap *, const struct path *,
2998 struct kstat *, u32, unsigned int);
2999extern void ext4_dirty_inode(struct inode *, int);
3000extern int ext4_change_inode_journal_flag(struct inode *, int);
3001extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
3002extern int ext4_get_fc_inode_loc(struct super_block *sb, unsigned long ino,
3003 struct ext4_iloc *iloc);
3004extern int ext4_inode_attach_jinode(struct inode *inode);
3005extern int ext4_can_truncate(struct inode *inode);
3006extern int ext4_truncate(struct inode *);
3007extern int ext4_break_layouts(struct inode *);
3008extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length);
3009extern void ext4_set_inode_flags(struct inode *, bool init);
3010extern int ext4_alloc_da_blocks(struct inode *inode);
3011extern void ext4_set_aops(struct inode *inode);
3012extern int ext4_writepage_trans_blocks(struct inode *);
3013extern int ext4_normal_submit_inode_data_buffers(struct jbd2_inode *jinode);
3014extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
3015extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
3016 loff_t lstart, loff_t lend);
3017extern vm_fault_t ext4_page_mkwrite(struct vm_fault *vmf);
3018extern qsize_t *ext4_get_reserved_space(struct inode *inode);
3019extern int ext4_get_projid(struct inode *inode, kprojid_t *projid);
3020extern void ext4_da_release_space(struct inode *inode, int to_free);
3021extern void ext4_da_update_reserve_space(struct inode *inode,
3022 int used, int quota_claim);
3023extern int ext4_issue_zeroout(struct inode *inode, ext4_lblk_t lblk,
3024 ext4_fsblk_t pblk, ext4_lblk_t len);
3025
3026/* indirect.c */
3027extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
3028 struct ext4_map_blocks *map, int flags);
3029extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
3030extern void ext4_ind_truncate(handle_t *, struct inode *inode);
3031extern int ext4_ind_remove_space(handle_t *handle, struct inode *inode,
3032 ext4_lblk_t start, ext4_lblk_t end);
3033
3034/* ioctl.c */
3035extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
3036extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
3037int ext4_fileattr_set(struct mnt_idmap *idmap,
3038 struct dentry *dentry, struct fileattr *fa);
3039int ext4_fileattr_get(struct dentry *dentry, struct fileattr *fa);
3040extern void ext4_reset_inode_seed(struct inode *inode);
3041int ext4_update_overhead(struct super_block *sb, bool force);
3042int ext4_force_shutdown(struct super_block *sb, u32 flags);
3043
3044/* migrate.c */
3045extern int ext4_ext_migrate(struct inode *);
3046extern int ext4_ind_migrate(struct inode *inode);
3047
3048/* namei.c */
3049extern int ext4_init_new_dir(handle_t *handle, struct inode *dir,
3050 struct inode *inode);
3051extern int ext4_dirblock_csum_verify(struct inode *inode,
3052 struct buffer_head *bh);
3053extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
3054 __u32 start_minor_hash, __u32 *next_hash);
3055extern int ext4_search_dir(struct buffer_head *bh,
3056 char *search_buf,
3057 int buf_size,
3058 struct inode *dir,
3059 struct ext4_filename *fname,
3060 unsigned int offset,
3061 struct ext4_dir_entry_2 **res_dir);
3062extern int ext4_generic_delete_entry(struct inode *dir,
3063 struct ext4_dir_entry_2 *de_del,
3064 struct buffer_head *bh,
3065 void *entry_buf,
3066 int buf_size,
3067 int csum_size);
3068extern bool ext4_empty_dir(struct inode *inode);
3069
3070/* resize.c */
3071extern void ext4_kvfree_array_rcu(void *to_free);
3072extern int ext4_group_add(struct super_block *sb,
3073 struct ext4_new_group_data *input);
3074extern int ext4_group_extend(struct super_block *sb,
3075 struct ext4_super_block *es,
3076 ext4_fsblk_t n_blocks_count);
3077extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
3078extern unsigned int ext4_list_backups(struct super_block *sb,
3079 unsigned int *three, unsigned int *five,
3080 unsigned int *seven);
3081
3082/* super.c */
3083extern struct buffer_head *ext4_sb_bread(struct super_block *sb,
3084 sector_t block, blk_opf_t op_flags);
3085extern struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
3086 sector_t block);
3087extern void ext4_read_bh_nowait(struct buffer_head *bh, blk_opf_t op_flags,
3088 bh_end_io_t *end_io);
3089extern int ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags,
3090 bh_end_io_t *end_io);
3091extern int ext4_read_bh_lock(struct buffer_head *bh, blk_opf_t op_flags, bool wait);
3092extern void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block);
3093extern int ext4_seq_options_show(struct seq_file *seq, void *offset);
3094extern int ext4_calculate_overhead(struct super_block *sb);
3095extern __le32 ext4_superblock_csum(struct super_block *sb,
3096 struct ext4_super_block *es);
3097extern void ext4_superblock_csum_set(struct super_block *sb);
3098extern int ext4_alloc_flex_bg_array(struct super_block *sb,
3099 ext4_group_t ngroup);
3100extern const char *ext4_decode_error(struct super_block *sb, int errno,
3101 char nbuf[16]);
3102extern void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
3103 ext4_group_t block_group,
3104 unsigned int flags);
3105extern unsigned int ext4_num_base_meta_blocks(struct super_block *sb,
3106 ext4_group_t block_group);
3107
3108extern __printf(7, 8)
3109void __ext4_error(struct super_block *, const char *, unsigned int, bool,
3110 int, __u64, const char *, ...);
3111extern __printf(6, 7)
3112void __ext4_error_inode(struct inode *, const char *, unsigned int,
3113 ext4_fsblk_t, int, const char *, ...);
3114extern __printf(5, 6)
3115void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
3116 const char *, ...);
3117extern void __ext4_std_error(struct super_block *, const char *,
3118 unsigned int, int);
3119extern __printf(4, 5)
3120void __ext4_warning(struct super_block *, const char *, unsigned int,
3121 const char *, ...);
3122extern __printf(4, 5)
3123void __ext4_warning_inode(const struct inode *inode, const char *function,
3124 unsigned int line, const char *fmt, ...);
3125extern __printf(3, 4)
3126void __ext4_msg(struct super_block *, const char *, const char *, ...);
3127extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
3128 const char *, unsigned int, const char *);
3129extern __printf(7, 8)
3130void __ext4_grp_locked_error(const char *, unsigned int,
3131 struct super_block *, ext4_group_t,
3132 unsigned long, ext4_fsblk_t,
3133 const char *, ...);
3134
3135#define EXT4_ERROR_INODE(inode, fmt, a...) \
3136 ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
3137
3138#define EXT4_ERROR_INODE_ERR(inode, err, fmt, a...) \
3139 __ext4_error_inode((inode), __func__, __LINE__, 0, (err), (fmt), ## a)
3140
3141#define ext4_error_inode_block(inode, block, err, fmt, a...) \
3142 __ext4_error_inode((inode), __func__, __LINE__, (block), (err), \
3143 (fmt), ## a)
3144
3145#define EXT4_ERROR_FILE(file, block, fmt, a...) \
3146 ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
3147
3148#define ext4_abort(sb, err, fmt, a...) \
3149 __ext4_error((sb), __func__, __LINE__, true, (err), 0, (fmt), ## a)
3150
3151#ifdef CONFIG_PRINTK
3152
3153#define ext4_error_inode(inode, func, line, block, fmt, ...) \
3154 __ext4_error_inode(inode, func, line, block, 0, fmt, ##__VA_ARGS__)
3155#define ext4_error_inode_err(inode, func, line, block, err, fmt, ...) \
3156 __ext4_error_inode((inode), (func), (line), (block), \
3157 (err), (fmt), ##__VA_ARGS__)
3158#define ext4_error_file(file, func, line, block, fmt, ...) \
3159 __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__)
3160#define ext4_error(sb, fmt, ...) \
3161 __ext4_error((sb), __func__, __LINE__, false, 0, 0, (fmt), \
3162 ##__VA_ARGS__)
3163#define ext4_error_err(sb, err, fmt, ...) \
3164 __ext4_error((sb), __func__, __LINE__, false, (err), 0, (fmt), \
3165 ##__VA_ARGS__)
3166#define ext4_warning(sb, fmt, ...) \
3167 __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
3168#define ext4_warning_inode(inode, fmt, ...) \
3169 __ext4_warning_inode(inode, __func__, __LINE__, fmt, ##__VA_ARGS__)
3170#define ext4_msg(sb, level, fmt, ...) \
3171 __ext4_msg(sb, level, fmt, ##__VA_ARGS__)
3172#define dump_mmp_msg(sb, mmp, msg) \
3173 __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg)
3174#define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...) \
3175 __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \
3176 fmt, ##__VA_ARGS__)
3177
3178#else
3179
3180#define ext4_error_inode(inode, func, line, block, fmt, ...) \
3181do { \
3182 no_printk(fmt, ##__VA_ARGS__); \
3183 __ext4_error_inode(inode, "", 0, block, 0, " "); \
3184} while (0)
3185#define ext4_error_inode_err(inode, func, line, block, err, fmt, ...) \
3186do { \
3187 no_printk(fmt, ##__VA_ARGS__); \
3188 __ext4_error_inode(inode, "", 0, block, err, " "); \
3189} while (0)
3190#define ext4_error_file(file, func, line, block, fmt, ...) \
3191do { \
3192 no_printk(fmt, ##__VA_ARGS__); \
3193 __ext4_error_file(file, "", 0, block, " "); \
3194} while (0)
3195#define ext4_error(sb, fmt, ...) \
3196do { \
3197 no_printk(fmt, ##__VA_ARGS__); \
3198 __ext4_error(sb, "", 0, false, 0, 0, " "); \
3199} while (0)
3200#define ext4_error_err(sb, err, fmt, ...) \
3201do { \
3202 no_printk(fmt, ##__VA_ARGS__); \
3203 __ext4_error(sb, "", 0, false, err, 0, " "); \
3204} while (0)
3205#define ext4_warning(sb, fmt, ...) \
3206do { \
3207 no_printk(fmt, ##__VA_ARGS__); \
3208 __ext4_warning(sb, "", 0, " "); \
3209} while (0)
3210#define ext4_warning_inode(inode, fmt, ...) \
3211do { \
3212 no_printk(fmt, ##__VA_ARGS__); \
3213 __ext4_warning_inode(inode, "", 0, " "); \
3214} while (0)
3215#define ext4_msg(sb, level, fmt, ...) \
3216do { \
3217 no_printk(fmt, ##__VA_ARGS__); \
3218 __ext4_msg(sb, "", " "); \
3219} while (0)
3220#define dump_mmp_msg(sb, mmp, msg) \
3221 __dump_mmp_msg(sb, mmp, "", 0, "")
3222#define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...) \
3223do { \
3224 no_printk(fmt, ##__VA_ARGS__); \
3225 __ext4_grp_locked_error("", 0, sb, grp, ino, block, " "); \
3226} while (0)
3227
3228#endif
3229
3230extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
3231 struct ext4_group_desc *bg);
3232extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
3233 struct ext4_group_desc *bg);
3234extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
3235 struct ext4_group_desc *bg);
3236extern __u32 ext4_free_group_clusters(struct super_block *sb,
3237 struct ext4_group_desc *bg);
3238extern __u32 ext4_free_inodes_count(struct super_block *sb,
3239 struct ext4_group_desc *bg);
3240extern __u32 ext4_used_dirs_count(struct super_block *sb,
3241 struct ext4_group_desc *bg);
3242extern __u32 ext4_itable_unused_count(struct super_block *sb,
3243 struct ext4_group_desc *bg);
3244extern void ext4_block_bitmap_set(struct super_block *sb,
3245 struct ext4_group_desc *bg, ext4_fsblk_t blk);
3246extern void ext4_inode_bitmap_set(struct super_block *sb,
3247 struct ext4_group_desc *bg, ext4_fsblk_t blk);
3248extern void ext4_inode_table_set(struct super_block *sb,
3249 struct ext4_group_desc *bg, ext4_fsblk_t blk);
3250extern void ext4_free_group_clusters_set(struct super_block *sb,
3251 struct ext4_group_desc *bg,
3252 __u32 count);
3253extern void ext4_free_inodes_set(struct super_block *sb,
3254 struct ext4_group_desc *bg, __u32 count);
3255extern void ext4_used_dirs_set(struct super_block *sb,
3256 struct ext4_group_desc *bg, __u32 count);
3257extern void ext4_itable_unused_set(struct super_block *sb,
3258 struct ext4_group_desc *bg, __u32 count);
3259extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
3260 struct ext4_group_desc *gdp);
3261extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
3262 struct ext4_group_desc *gdp);
3263extern int ext4_register_li_request(struct super_block *sb,
3264 ext4_group_t first_not_zeroed);
3265
3266static inline int ext4_has_metadata_csum(struct super_block *sb)
3267{
3268 WARN_ON_ONCE(ext4_has_feature_metadata_csum(sb) &&
3269 !EXT4_SB(sb)->s_chksum_driver);
3270
3271 return ext4_has_feature_metadata_csum(sb) &&
3272 (EXT4_SB(sb)->s_chksum_driver != NULL);
3273}
3274
3275static inline int ext4_has_group_desc_csum(struct super_block *sb)
3276{
3277 return ext4_has_feature_gdt_csum(sb) || ext4_has_metadata_csum(sb);
3278}
3279
3280#define ext4_read_incompat_64bit_val(es, name) \
3281 (((es)->s_feature_incompat & cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT) \
3282 ? (ext4_fsblk_t)le32_to_cpu(es->name##_hi) << 32 : 0) | \
3283 le32_to_cpu(es->name##_lo))
3284
3285static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
3286{
3287 return ext4_read_incompat_64bit_val(es, s_blocks_count);
3288}
3289
3290static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
3291{
3292 return ext4_read_incompat_64bit_val(es, s_r_blocks_count);
3293}
3294
3295static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
3296{
3297 return ext4_read_incompat_64bit_val(es, s_free_blocks_count);
3298}
3299
3300static inline void ext4_blocks_count_set(struct ext4_super_block *es,
3301 ext4_fsblk_t blk)
3302{
3303 es->s_blocks_count_lo = cpu_to_le32((u32)blk);
3304 es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
3305}
3306
3307static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
3308 ext4_fsblk_t blk)
3309{
3310 es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
3311 es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
3312}
3313
3314static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
3315 ext4_fsblk_t blk)
3316{
3317 es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
3318 es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
3319}
3320
3321static inline loff_t ext4_isize(struct super_block *sb,
3322 struct ext4_inode *raw_inode)
3323{
3324 if (ext4_has_feature_largedir(sb) ||
3325 S_ISREG(le16_to_cpu(raw_inode->i_mode)))
3326 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
3327 le32_to_cpu(raw_inode->i_size_lo);
3328
3329 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
3330}
3331
3332static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
3333{
3334 raw_inode->i_size_lo = cpu_to_le32(i_size);
3335 raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
3336}
3337
3338/*
3339 * Reading s_groups_count requires using smp_rmb() afterwards. See
3340 * the locking protocol documented in the comments of ext4_group_add()
3341 * in resize.c
3342 */
3343static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
3344{
3345 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3346
3347 smp_rmb();
3348 return ngroups;
3349}
3350
3351static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
3352 ext4_group_t block_group)
3353{
3354 return block_group >> sbi->s_log_groups_per_flex;
3355}
3356
3357static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
3358{
3359 return 1 << sbi->s_log_groups_per_flex;
3360}
3361
3362#define ext4_std_error(sb, errno) \
3363do { \
3364 if ((errno)) \
3365 __ext4_std_error((sb), __func__, __LINE__, (errno)); \
3366} while (0)
3367
3368#ifdef CONFIG_SMP
3369/* Each CPU can accumulate percpu_counter_batch clusters in their local
3370 * counters. So we need to make sure we have free clusters more
3371 * than percpu_counter_batch * nr_cpu_ids. Also add a window of 4 times.
3372 */
3373#define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
3374#else
3375#define EXT4_FREECLUSTERS_WATERMARK 0
3376#endif
3377
3378/* Update i_disksize. Requires i_rwsem to avoid races with truncate */
3379static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
3380{
3381 WARN_ON_ONCE(S_ISREG(inode->i_mode) &&
3382 !inode_is_locked(inode));
3383 down_write(&EXT4_I(inode)->i_data_sem);
3384 if (newsize > EXT4_I(inode)->i_disksize)
3385 WRITE_ONCE(EXT4_I(inode)->i_disksize, newsize);
3386 up_write(&EXT4_I(inode)->i_data_sem);
3387}
3388
3389/* Update i_size, i_disksize. Requires i_rwsem to avoid races with truncate */
3390static inline int ext4_update_inode_size(struct inode *inode, loff_t newsize)
3391{
3392 int changed = 0;
3393
3394 if (newsize > inode->i_size) {
3395 i_size_write(inode, newsize);
3396 changed = 1;
3397 }
3398 if (newsize > EXT4_I(inode)->i_disksize) {
3399 ext4_update_i_disksize(inode, newsize);
3400 changed |= 2;
3401 }
3402 return changed;
3403}
3404
3405int ext4_update_disksize_before_punch(struct inode *inode, loff_t offset,
3406 loff_t len);
3407
3408struct ext4_group_info {
3409 unsigned long bb_state;
3410#ifdef AGGRESSIVE_CHECK
3411 unsigned long bb_check_counter;
3412#endif
3413 struct rb_root bb_free_root;
3414 ext4_grpblk_t bb_first_free; /* first free block */
3415 ext4_grpblk_t bb_free; /* total free blocks */
3416 ext4_grpblk_t bb_fragments; /* nr of freespace fragments */
3417 int bb_avg_fragment_size_order; /* order of average
3418 fragment in BG */
3419 ext4_grpblk_t bb_largest_free_order;/* order of largest frag in BG */
3420 ext4_group_t bb_group; /* Group number */
3421 struct list_head bb_prealloc_list;
3422#ifdef DOUBLE_CHECK
3423 void *bb_bitmap;
3424#endif
3425 struct rw_semaphore alloc_sem;
3426 struct list_head bb_avg_fragment_size_node;
3427 struct list_head bb_largest_free_order_node;
3428 ext4_grpblk_t bb_counters[]; /* Nr of free power-of-two-block
3429 * regions, index is order.
3430 * bb_counters[3] = 5 means
3431 * 5 free 8-block regions. */
3432};
3433
3434#define EXT4_GROUP_INFO_NEED_INIT_BIT 0
3435#define EXT4_GROUP_INFO_WAS_TRIMMED_BIT 1
3436#define EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT 2
3437#define EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT 3
3438#define EXT4_GROUP_INFO_BBITMAP_CORRUPT \
3439 (1 << EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT)
3440#define EXT4_GROUP_INFO_IBITMAP_CORRUPT \
3441 (1 << EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT)
3442#define EXT4_GROUP_INFO_BBITMAP_READ_BIT 4
3443
3444#define EXT4_MB_GRP_NEED_INIT(grp) \
3445 (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
3446#define EXT4_MB_GRP_BBITMAP_CORRUPT(grp) \
3447 (test_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
3448#define EXT4_MB_GRP_IBITMAP_CORRUPT(grp) \
3449 (test_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
3450
3451#define EXT4_MB_GRP_WAS_TRIMMED(grp) \
3452 (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3453#define EXT4_MB_GRP_SET_TRIMMED(grp) \
3454 (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3455#define EXT4_MB_GRP_CLEAR_TRIMMED(grp) \
3456 (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
3457#define EXT4_MB_GRP_TEST_AND_SET_READ(grp) \
3458 (test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_READ_BIT, &((grp)->bb_state)))
3459
3460#define EXT4_MAX_CONTENTION 8
3461#define EXT4_CONTENTION_THRESHOLD 2
3462
3463static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
3464 ext4_group_t group)
3465{
3466 return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
3467}
3468
3469/*
3470 * Returns true if the filesystem is busy enough that attempts to
3471 * access the block group locks has run into contention.
3472 */
3473static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
3474{
3475 return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
3476}
3477
3478static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
3479{
3480 spinlock_t *lock = ext4_group_lock_ptr(sb, group);
3481 if (spin_trylock(lock))
3482 /*
3483 * We're able to grab the lock right away, so drop the
3484 * lock contention counter.
3485 */
3486 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
3487 else {
3488 /*
3489 * The lock is busy, so bump the contention counter,
3490 * and then wait on the spin lock.
3491 */
3492 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
3493 EXT4_MAX_CONTENTION);
3494 spin_lock(lock);
3495 }
3496}
3497
3498static inline void ext4_unlock_group(struct super_block *sb,
3499 ext4_group_t group)
3500{
3501 spin_unlock(ext4_group_lock_ptr(sb, group));
3502}
3503
3504#ifdef CONFIG_QUOTA
3505static inline bool ext4_quota_capable(struct super_block *sb)
3506{
3507 return (test_opt(sb, QUOTA) || ext4_has_feature_quota(sb));
3508}
3509
3510static inline bool ext4_is_quota_journalled(struct super_block *sb)
3511{
3512 struct ext4_sb_info *sbi = EXT4_SB(sb);
3513
3514 return (ext4_has_feature_quota(sb) ||
3515 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]);
3516}
3517int ext4_enable_quotas(struct super_block *sb);
3518#endif
3519
3520/*
3521 * Block validity checking
3522 */
3523#define ext4_check_indirect_blockref(inode, bh) \
3524 ext4_check_blockref(__func__, __LINE__, inode, \
3525 (__le32 *)(bh)->b_data, \
3526 EXT4_ADDR_PER_BLOCK((inode)->i_sb))
3527
3528#define ext4_ind_check_inode(inode) \
3529 ext4_check_blockref(__func__, __LINE__, inode, \
3530 EXT4_I(inode)->i_data, \
3531 EXT4_NDIR_BLOCKS)
3532
3533/*
3534 * Inodes and files operations
3535 */
3536
3537/* dir.c */
3538extern const struct file_operations ext4_dir_operations;
3539
3540/* file.c */
3541extern const struct inode_operations ext4_file_inode_operations;
3542extern const struct file_operations ext4_file_operations;
3543extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
3544
3545/* inline.c */
3546extern int ext4_get_max_inline_size(struct inode *inode);
3547extern int ext4_find_inline_data_nolock(struct inode *inode);
3548extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
3549
3550int ext4_readpage_inline(struct inode *inode, struct folio *folio);
3551extern int ext4_try_to_write_inline_data(struct address_space *mapping,
3552 struct inode *inode,
3553 loff_t pos, unsigned len,
3554 struct page **pagep);
3555int ext4_write_inline_data_end(struct inode *inode, loff_t pos, unsigned len,
3556 unsigned copied, struct folio *folio);
3557extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
3558 struct inode *inode,
3559 loff_t pos, unsigned len,
3560 struct page **pagep,
3561 void **fsdata);
3562extern int ext4_try_add_inline_entry(handle_t *handle,
3563 struct ext4_filename *fname,
3564 struct inode *dir, struct inode *inode);
3565extern int ext4_try_create_inline_dir(handle_t *handle,
3566 struct inode *parent,
3567 struct inode *inode);
3568extern int ext4_read_inline_dir(struct file *filp,
3569 struct dir_context *ctx,
3570 int *has_inline_data);
3571extern int ext4_inlinedir_to_tree(struct file *dir_file,
3572 struct inode *dir, ext4_lblk_t block,
3573 struct dx_hash_info *hinfo,
3574 __u32 start_hash, __u32 start_minor_hash,
3575 int *has_inline_data);
3576extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
3577 struct ext4_filename *fname,
3578 struct ext4_dir_entry_2 **res_dir,
3579 int *has_inline_data);
3580extern int ext4_delete_inline_entry(handle_t *handle,
3581 struct inode *dir,
3582 struct ext4_dir_entry_2 *de_del,
3583 struct buffer_head *bh,
3584 int *has_inline_data);
3585extern bool empty_inline_dir(struct inode *dir, int *has_inline_data);
3586extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
3587 struct ext4_dir_entry_2 **parent_de,
3588 int *retval);
3589extern void *ext4_read_inline_link(struct inode *inode);
3590
3591struct iomap;
3592extern int ext4_inline_data_iomap(struct inode *inode, struct iomap *iomap);
3593
3594extern int ext4_inline_data_truncate(struct inode *inode, int *has_inline);
3595
3596extern int ext4_convert_inline_data(struct inode *inode);
3597
3598static inline int ext4_has_inline_data(struct inode *inode)
3599{
3600 return ext4_test_inode_flag(inode, EXT4_INODE_INLINE_DATA) &&
3601 EXT4_I(inode)->i_inline_off;
3602}
3603
3604/* namei.c */
3605extern const struct inode_operations ext4_dir_inode_operations;
3606extern const struct inode_operations ext4_special_inode_operations;
3607extern struct dentry *ext4_get_parent(struct dentry *child);
3608extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
3609 struct ext4_dir_entry_2 *de,
3610 int blocksize, int csum_size,
3611 unsigned int parent_ino, int dotdot_real_len);
3612extern void ext4_initialize_dirent_tail(struct buffer_head *bh,
3613 unsigned int blocksize);
3614extern int ext4_handle_dirty_dirblock(handle_t *handle, struct inode *inode,
3615 struct buffer_head *bh);
3616extern int __ext4_unlink(struct inode *dir, const struct qstr *d_name,
3617 struct inode *inode, struct dentry *dentry);
3618extern int __ext4_link(struct inode *dir, struct inode *inode,
3619 struct dentry *dentry);
3620
3621#define S_SHIFT 12
3622static const unsigned char ext4_type_by_mode[(S_IFMT >> S_SHIFT) + 1] = {
3623 [S_IFREG >> S_SHIFT] = EXT4_FT_REG_FILE,
3624 [S_IFDIR >> S_SHIFT] = EXT4_FT_DIR,
3625 [S_IFCHR >> S_SHIFT] = EXT4_FT_CHRDEV,
3626 [S_IFBLK >> S_SHIFT] = EXT4_FT_BLKDEV,
3627 [S_IFIFO >> S_SHIFT] = EXT4_FT_FIFO,
3628 [S_IFSOCK >> S_SHIFT] = EXT4_FT_SOCK,
3629 [S_IFLNK >> S_SHIFT] = EXT4_FT_SYMLINK,
3630};
3631
3632static inline void ext4_set_de_type(struct super_block *sb,
3633 struct ext4_dir_entry_2 *de,
3634 umode_t mode) {
3635 if (ext4_has_feature_filetype(sb))
3636 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
3637}
3638
3639/* readpages.c */
3640extern int ext4_mpage_readpages(struct inode *inode,
3641 struct readahead_control *rac, struct folio *folio);
3642extern int __init ext4_init_post_read_processing(void);
3643extern void ext4_exit_post_read_processing(void);
3644
3645/* symlink.c */
3646extern const struct inode_operations ext4_encrypted_symlink_inode_operations;
3647extern const struct inode_operations ext4_symlink_inode_operations;
3648extern const struct inode_operations ext4_fast_symlink_inode_operations;
3649
3650/* sysfs.c */
3651extern void ext4_notify_error_sysfs(struct ext4_sb_info *sbi);
3652extern int ext4_register_sysfs(struct super_block *sb);
3653extern void ext4_unregister_sysfs(struct super_block *sb);
3654extern int __init ext4_init_sysfs(void);
3655extern void ext4_exit_sysfs(void);
3656
3657/* block_validity */
3658extern void ext4_release_system_zone(struct super_block *sb);
3659extern int ext4_setup_system_zone(struct super_block *sb);
3660extern int __init ext4_init_system_zone(void);
3661extern void ext4_exit_system_zone(void);
3662extern int ext4_inode_block_valid(struct inode *inode,
3663 ext4_fsblk_t start_blk,
3664 unsigned int count);
3665extern int ext4_check_blockref(const char *, unsigned int,
3666 struct inode *, __le32 *, unsigned int);
3667extern int ext4_sb_block_valid(struct super_block *sb, struct inode *inode,
3668 ext4_fsblk_t start_blk, unsigned int count);
3669
3670
3671/* extents.c */
3672struct ext4_ext_path;
3673struct ext4_extent;
3674
3675/*
3676 * Maximum number of logical blocks in a file; ext4_extent's ee_block is
3677 * __le32.
3678 */
3679#define EXT_MAX_BLOCKS 0xffffffff
3680
3681extern void ext4_ext_tree_init(handle_t *handle, struct inode *inode);
3682extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
3683extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
3684 struct ext4_map_blocks *map, int flags);
3685extern int ext4_ext_truncate(handle_t *, struct inode *);
3686extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
3687 ext4_lblk_t end);
3688extern void ext4_ext_init(struct super_block *);
3689extern void ext4_ext_release(struct super_block *);
3690extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
3691 loff_t len);
3692extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
3693 loff_t offset, ssize_t len);
3694extern int ext4_convert_unwritten_io_end_vec(handle_t *handle,
3695 ext4_io_end_t *io_end);
3696extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
3697 struct ext4_map_blocks *map, int flags);
3698extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
3699 int num,
3700 struct ext4_ext_path *path);
3701extern int ext4_ext_insert_extent(handle_t *, struct inode *,
3702 struct ext4_ext_path **,
3703 struct ext4_extent *, int);
3704extern struct ext4_ext_path *ext4_find_extent(struct inode *, ext4_lblk_t,
3705 struct ext4_ext_path **,
3706 int flags);
3707extern void ext4_free_ext_path(struct ext4_ext_path *);
3708extern int ext4_ext_check_inode(struct inode *inode);
3709extern ext4_lblk_t ext4_ext_next_allocated_block(struct ext4_ext_path *path);
3710extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
3711 __u64 start, __u64 len);
3712extern int ext4_get_es_cache(struct inode *inode,
3713 struct fiemap_extent_info *fieinfo,
3714 __u64 start, __u64 len);
3715extern int ext4_ext_precache(struct inode *inode);
3716extern int ext4_swap_extents(handle_t *handle, struct inode *inode1,
3717 struct inode *inode2, ext4_lblk_t lblk1,
3718 ext4_lblk_t lblk2, ext4_lblk_t count,
3719 int mark_unwritten,int *err);
3720extern int ext4_clu_mapped(struct inode *inode, ext4_lblk_t lclu);
3721extern int ext4_datasem_ensure_credits(handle_t *handle, struct inode *inode,
3722 int check_cred, int restart_cred,
3723 int revoke_cred);
3724extern void ext4_ext_replay_shrink_inode(struct inode *inode, ext4_lblk_t end);
3725extern int ext4_ext_replay_set_iblocks(struct inode *inode);
3726extern int ext4_ext_replay_update_ex(struct inode *inode, ext4_lblk_t start,
3727 int len, int unwritten, ext4_fsblk_t pblk);
3728extern int ext4_ext_clear_bb(struct inode *inode);
3729
3730
3731/* move_extent.c */
3732extern void ext4_double_down_write_data_sem(struct inode *first,
3733 struct inode *second);
3734extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
3735 struct inode *donor_inode);
3736extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
3737 __u64 start_orig, __u64 start_donor,
3738 __u64 len, __u64 *moved_len);
3739
3740/* page-io.c */
3741extern int __init ext4_init_pageio(void);
3742extern void ext4_exit_pageio(void);
3743extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
3744extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
3745extern int ext4_put_io_end(ext4_io_end_t *io_end);
3746extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
3747extern void ext4_io_submit_init(struct ext4_io_submit *io,
3748 struct writeback_control *wbc);
3749extern void ext4_end_io_rsv_work(struct work_struct *work);
3750extern void ext4_io_submit(struct ext4_io_submit *io);
3751int ext4_bio_write_folio(struct ext4_io_submit *io, struct folio *page,
3752 size_t len);
3753extern struct ext4_io_end_vec *ext4_alloc_io_end_vec(ext4_io_end_t *io_end);
3754extern struct ext4_io_end_vec *ext4_last_io_end_vec(ext4_io_end_t *io_end);
3755
3756/* mmp.c */
3757extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
3758
3759/* mmp.c */
3760extern void ext4_stop_mmpd(struct ext4_sb_info *sbi);
3761
3762/* verity.c */
3763extern const struct fsverity_operations ext4_verityops;
3764
3765/* orphan.c */
3766extern int ext4_orphan_add(handle_t *, struct inode *);
3767extern int ext4_orphan_del(handle_t *, struct inode *);
3768extern void ext4_orphan_cleanup(struct super_block *sb,
3769 struct ext4_super_block *es);
3770extern void ext4_release_orphan_info(struct super_block *sb);
3771extern int ext4_init_orphan_info(struct super_block *sb);
3772extern int ext4_orphan_file_empty(struct super_block *sb);
3773extern void ext4_orphan_file_block_trigger(
3774 struct jbd2_buffer_trigger_type *triggers,
3775 struct buffer_head *bh,
3776 void *data, size_t size);
3777
3778/*
3779 * Add new method to test whether block and inode bitmaps are properly
3780 * initialized. With uninit_bg reading the block from disk is not enough
3781 * to mark the bitmap uptodate. We need to also zero-out the bitmap
3782 */
3783#define BH_BITMAP_UPTODATE BH_JBDPrivateStart
3784
3785static inline int bitmap_uptodate(struct buffer_head *bh)
3786{
3787 return (buffer_uptodate(bh) &&
3788 test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
3789}
3790static inline void set_bitmap_uptodate(struct buffer_head *bh)
3791{
3792 set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
3793}
3794
3795/* For ioend & aio unwritten conversion wait queues */
3796#define EXT4_WQ_HASH_SZ 37
3797#define ext4_ioend_wq(v) (&ext4__ioend_wq[((unsigned long)(v)) %\
3798 EXT4_WQ_HASH_SZ])
3799extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
3800
3801extern int ext4_resize_begin(struct super_block *sb);
3802extern int ext4_resize_end(struct super_block *sb, bool update_backups);
3803
3804static inline void ext4_set_io_unwritten_flag(struct inode *inode,
3805 struct ext4_io_end *io_end)
3806{
3807 if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
3808 io_end->flag |= EXT4_IO_END_UNWRITTEN;
3809 atomic_inc(&EXT4_I(inode)->i_unwritten);
3810 }
3811}
3812
3813static inline void ext4_clear_io_unwritten_flag(ext4_io_end_t *io_end)
3814{
3815 struct inode *inode = io_end->inode;
3816
3817 if (io_end->flag & EXT4_IO_END_UNWRITTEN) {
3818 io_end->flag &= ~EXT4_IO_END_UNWRITTEN;
3819 /* Wake up anyone waiting on unwritten extent conversion */
3820 if (atomic_dec_and_test(&EXT4_I(inode)->i_unwritten))
3821 wake_up_all(ext4_ioend_wq(inode));
3822 }
3823}
3824
3825extern const struct iomap_ops ext4_iomap_ops;
3826extern const struct iomap_ops ext4_iomap_overwrite_ops;
3827extern const struct iomap_ops ext4_iomap_report_ops;
3828
3829static inline int ext4_buffer_uptodate(struct buffer_head *bh)
3830{
3831 /*
3832 * If the buffer has the write error flag, we have failed
3833 * to write out data in the block. In this case, we don't
3834 * have to read the block because we may read the old data
3835 * successfully.
3836 */
3837 if (buffer_write_io_error(bh))
3838 set_buffer_uptodate(bh);
3839 return buffer_uptodate(bh);
3840}
3841
3842#endif /* __KERNEL__ */
3843
3844#define EFSBADCRC EBADMSG /* Bad CRC detected */
3845#define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
3846
3847#endif /* _EXT4_H */