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