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/version.h>
30#include <linux/wait.h>
31#include <linux/blockgroup_lock.h>
32#include <linux/percpu_counter.h>
33#include <linux/ratelimit.h>
34#include <crypto/hash.h>
35#include <linux/falloc.h>
36#ifdef __KERNEL__
37#include <linux/compat.h>
38#endif
39
40/*
41 * The fourth extended filesystem constants/structures
42 */
43
44/*
45 * with AGGRESSIVE_CHECK allocator runs consistency checks over
46 * structures. these checks slow things down a lot
47 */
48#define AGGRESSIVE_CHECK__
49
50/*
51 * with DOUBLE_CHECK defined mballoc creates persistent in-core
52 * bitmaps, maintains and uses them to check for double allocations
53 */
54#define DOUBLE_CHECK__
55
56/*
57 * Define EXT4FS_DEBUG to produce debug messages
58 */
59#undef EXT4FS_DEBUG
60
61/*
62 * Debug code
63 */
64#ifdef EXT4FS_DEBUG
65#define ext4_debug(f, a...) \
66 do { \
67 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:", \
68 __FILE__, __LINE__, __func__); \
69 printk(KERN_DEBUG f, ## a); \
70 } while (0)
71#else
72#define ext4_debug(fmt, ...) no_printk(fmt, ##__VA_ARGS__)
73#endif
74
75/*
76 * Turn on EXT_DEBUG to get lots of info about extents operations.
77 */
78#define EXT_DEBUG__
79#ifdef EXT_DEBUG
80#define ext_debug(fmt, ...) printk(fmt, ##__VA_ARGS__)
81#else
82#define ext_debug(fmt, ...) no_printk(fmt, ##__VA_ARGS__)
83#endif
84
85/* data type for block offset of block group */
86typedef int ext4_grpblk_t;
87
88/* data type for filesystem-wide blocks number */
89typedef unsigned long long ext4_fsblk_t;
90
91/* data type for file logical block number */
92typedef __u32 ext4_lblk_t;
93
94/* data type for block group number */
95typedef unsigned int ext4_group_t;
96
97enum SHIFT_DIRECTION {
98 SHIFT_LEFT = 0,
99 SHIFT_RIGHT,
100};
101
102/*
103 * Flags used in mballoc's allocation_context flags field.
104 *
105 * Also used to show what's going on for debugging purposes when the
106 * flag field is exported via the traceport interface
107 */
108
109/* prefer goal again. length */
110#define EXT4_MB_HINT_MERGE 0x0001
111/* blocks already reserved */
112#define EXT4_MB_HINT_RESERVED 0x0002
113/* metadata is being allocated */
114#define EXT4_MB_HINT_METADATA 0x0004
115/* first blocks in the file */
116#define EXT4_MB_HINT_FIRST 0x0008
117/* search for the best chunk */
118#define EXT4_MB_HINT_BEST 0x0010
119/* data is being allocated */
120#define EXT4_MB_HINT_DATA 0x0020
121/* don't preallocate (for tails) */
122#define EXT4_MB_HINT_NOPREALLOC 0x0040
123/* allocate for locality group */
124#define EXT4_MB_HINT_GROUP_ALLOC 0x0080
125/* allocate goal blocks or none */
126#define EXT4_MB_HINT_GOAL_ONLY 0x0100
127/* goal is meaningful */
128#define EXT4_MB_HINT_TRY_GOAL 0x0200
129/* blocks already pre-reserved by delayed allocation */
130#define EXT4_MB_DELALLOC_RESERVED 0x0400
131/* We are doing stream allocation */
132#define EXT4_MB_STREAM_ALLOC 0x0800
133/* Use reserved root blocks if needed */
134#define EXT4_MB_USE_ROOT_BLOCKS 0x1000
135/* Use blocks from reserved pool */
136#define EXT4_MB_USE_RESERVED 0x2000
137
138struct ext4_allocation_request {
139 /* target inode for block we're allocating */
140 struct inode *inode;
141 /* how many blocks we want to allocate */
142 unsigned int len;
143 /* logical block in target inode */
144 ext4_lblk_t logical;
145 /* the closest logical allocated block to the left */
146 ext4_lblk_t lleft;
147 /* the closest logical allocated block to the right */
148 ext4_lblk_t lright;
149 /* phys. target (a hint) */
150 ext4_fsblk_t goal;
151 /* phys. block for the closest logical allocated block to the left */
152 ext4_fsblk_t pleft;
153 /* phys. block for the closest logical allocated block to the right */
154 ext4_fsblk_t pright;
155 /* flags. see above EXT4_MB_HINT_* */
156 unsigned int flags;
157};
158
159/*
160 * Logical to physical block mapping, used by ext4_map_blocks()
161 *
162 * This structure is used to pass requests into ext4_map_blocks() as
163 * well as to store the information returned by ext4_map_blocks(). It
164 * takes less room on the stack than a struct buffer_head.
165 */
166#define EXT4_MAP_NEW (1 << BH_New)
167#define EXT4_MAP_MAPPED (1 << BH_Mapped)
168#define EXT4_MAP_UNWRITTEN (1 << BH_Unwritten)
169#define EXT4_MAP_BOUNDARY (1 << BH_Boundary)
170#define EXT4_MAP_FLAGS (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
171 EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY)
172
173struct ext4_map_blocks {
174 ext4_fsblk_t m_pblk;
175 ext4_lblk_t m_lblk;
176 unsigned int m_len;
177 unsigned int m_flags;
178};
179
180/*
181 * Flags for ext4_io_end->flags
182 */
183#define EXT4_IO_END_UNWRITTEN 0x0001
184
185/*
186 * For converting unwritten extents on a work queue. 'handle' is used for
187 * buffered writeback.
188 */
189typedef struct ext4_io_end {
190 struct list_head list; /* per-file finished IO list */
191 handle_t *handle; /* handle reserved for extent
192 * conversion */
193 struct inode *inode; /* file being written to */
194 struct bio *bio; /* Linked list of completed
195 * bios covering the extent */
196 unsigned int flag; /* unwritten or not */
197 atomic_t count; /* reference counter */
198 loff_t offset; /* offset in the file */
199 ssize_t size; /* size of the extent */
200} ext4_io_end_t;
201
202struct ext4_io_submit {
203 struct writeback_control *io_wbc;
204 struct bio *io_bio;
205 ext4_io_end_t *io_end;
206 sector_t io_next_block;
207};
208
209/*
210 * Special inodes numbers
211 */
212#define EXT4_BAD_INO 1 /* Bad blocks inode */
213#define EXT4_ROOT_INO 2 /* Root inode */
214#define EXT4_USR_QUOTA_INO 3 /* User quota inode */
215#define EXT4_GRP_QUOTA_INO 4 /* Group quota inode */
216#define EXT4_BOOT_LOADER_INO 5 /* Boot loader inode */
217#define EXT4_UNDEL_DIR_INO 6 /* Undelete directory inode */
218#define EXT4_RESIZE_INO 7 /* Reserved group descriptors inode */
219#define EXT4_JOURNAL_INO 8 /* Journal inode */
220
221/* First non-reserved inode for old ext4 filesystems */
222#define EXT4_GOOD_OLD_FIRST_INO 11
223
224/*
225 * Maximal count of links to a file
226 */
227#define EXT4_LINK_MAX 65000
228
229/*
230 * Macro-instructions used to manage several block sizes
231 */
232#define EXT4_MIN_BLOCK_SIZE 1024
233#define EXT4_MAX_BLOCK_SIZE 65536
234#define EXT4_MIN_BLOCK_LOG_SIZE 10
235#define EXT4_MAX_BLOCK_LOG_SIZE 16
236#ifdef __KERNEL__
237# define EXT4_BLOCK_SIZE(s) ((s)->s_blocksize)
238#else
239# define EXT4_BLOCK_SIZE(s) (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
240#endif
241#define EXT4_ADDR_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
242#define EXT4_CLUSTER_SIZE(s) (EXT4_BLOCK_SIZE(s) << \
243 EXT4_SB(s)->s_cluster_bits)
244#ifdef __KERNEL__
245# define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits)
246# define EXT4_CLUSTER_BITS(s) (EXT4_SB(s)->s_cluster_bits)
247#else
248# define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_log_block_size + 10)
249#endif
250#ifdef __KERNEL__
251#define EXT4_ADDR_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_addr_per_block_bits)
252#define EXT4_INODE_SIZE(s) (EXT4_SB(s)->s_inode_size)
253#define EXT4_FIRST_INO(s) (EXT4_SB(s)->s_first_ino)
254#else
255#define EXT4_INODE_SIZE(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
256 EXT4_GOOD_OLD_INODE_SIZE : \
257 (s)->s_inode_size)
258#define EXT4_FIRST_INO(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
259 EXT4_GOOD_OLD_FIRST_INO : \
260 (s)->s_first_ino)
261#endif
262#define EXT4_BLOCK_ALIGN(size, blkbits) ALIGN((size), (1 << (blkbits)))
263
264/* Translate a block number to a cluster number */
265#define EXT4_B2C(sbi, blk) ((blk) >> (sbi)->s_cluster_bits)
266/* Translate a cluster number to a block number */
267#define EXT4_C2B(sbi, cluster) ((cluster) << (sbi)->s_cluster_bits)
268/* Translate # of blks to # of clusters */
269#define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
270 (sbi)->s_cluster_bits)
271/* Mask out the low bits to get the starting block of the cluster */
272#define EXT4_PBLK_CMASK(s, pblk) ((pblk) & \
273 ~((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
274#define EXT4_LBLK_CMASK(s, lblk) ((lblk) & \
275 ~((ext4_lblk_t) (s)->s_cluster_ratio - 1))
276/* Get the cluster offset */
277#define EXT4_PBLK_COFF(s, pblk) ((pblk) & \
278 ((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
279#define EXT4_LBLK_COFF(s, lblk) ((lblk) & \
280 ((ext4_lblk_t) (s)->s_cluster_ratio - 1))
281
282/*
283 * Structure of a blocks group descriptor
284 */
285struct ext4_group_desc
286{
287 __le32 bg_block_bitmap_lo; /* Blocks bitmap block */
288 __le32 bg_inode_bitmap_lo; /* Inodes bitmap block */
289 __le32 bg_inode_table_lo; /* Inodes table block */
290 __le16 bg_free_blocks_count_lo;/* Free blocks count */
291 __le16 bg_free_inodes_count_lo;/* Free inodes count */
292 __le16 bg_used_dirs_count_lo; /* Directories count */
293 __le16 bg_flags; /* EXT4_BG_flags (INODE_UNINIT, etc) */
294 __le32 bg_exclude_bitmap_lo; /* Exclude bitmap for snapshots */
295 __le16 bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
296 __le16 bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
297 __le16 bg_itable_unused_lo; /* Unused inodes count */
298 __le16 bg_checksum; /* crc16(sb_uuid+group+desc) */
299 __le32 bg_block_bitmap_hi; /* Blocks bitmap block MSB */
300 __le32 bg_inode_bitmap_hi; /* Inodes bitmap block MSB */
301 __le32 bg_inode_table_hi; /* Inodes table block MSB */
302 __le16 bg_free_blocks_count_hi;/* Free blocks count MSB */
303 __le16 bg_free_inodes_count_hi;/* Free inodes count MSB */
304 __le16 bg_used_dirs_count_hi; /* Directories count MSB */
305 __le16 bg_itable_unused_hi; /* Unused inodes count MSB */
306 __le32 bg_exclude_bitmap_hi; /* Exclude bitmap block MSB */
307 __le16 bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
308 __le16 bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
309 __u32 bg_reserved;
310};
311
312#define EXT4_BG_INODE_BITMAP_CSUM_HI_END \
313 (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
314 sizeof(__le16))
315#define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END \
316 (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
317 sizeof(__le16))
318
319/*
320 * Structure of a flex block group info
321 */
322
323struct flex_groups {
324 atomic64_t free_clusters;
325 atomic_t free_inodes;
326 atomic_t used_dirs;
327};
328
329#define EXT4_BG_INODE_UNINIT 0x0001 /* Inode table/bitmap not in use */
330#define EXT4_BG_BLOCK_UNINIT 0x0002 /* Block bitmap not in use */
331#define EXT4_BG_INODE_ZEROED 0x0004 /* On-disk itable initialized to zero */
332
333/*
334 * Macro-instructions used to manage group descriptors
335 */
336#define EXT4_MIN_DESC_SIZE 32
337#define EXT4_MIN_DESC_SIZE_64BIT 64
338#define EXT4_MAX_DESC_SIZE EXT4_MIN_BLOCK_SIZE
339#define EXT4_DESC_SIZE(s) (EXT4_SB(s)->s_desc_size)
340#ifdef __KERNEL__
341# define EXT4_BLOCKS_PER_GROUP(s) (EXT4_SB(s)->s_blocks_per_group)
342# define EXT4_CLUSTERS_PER_GROUP(s) (EXT4_SB(s)->s_clusters_per_group)
343# define EXT4_DESC_PER_BLOCK(s) (EXT4_SB(s)->s_desc_per_block)
344# define EXT4_INODES_PER_GROUP(s) (EXT4_SB(s)->s_inodes_per_group)
345# define EXT4_DESC_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_desc_per_block_bits)
346#else
347# define EXT4_BLOCKS_PER_GROUP(s) ((s)->s_blocks_per_group)
348# define EXT4_DESC_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
349# define EXT4_INODES_PER_GROUP(s) ((s)->s_inodes_per_group)
350#endif
351
352/*
353 * Constants relative to the data blocks
354 */
355#define EXT4_NDIR_BLOCKS 12
356#define EXT4_IND_BLOCK EXT4_NDIR_BLOCKS
357#define EXT4_DIND_BLOCK (EXT4_IND_BLOCK + 1)
358#define EXT4_TIND_BLOCK (EXT4_DIND_BLOCK + 1)
359#define EXT4_N_BLOCKS (EXT4_TIND_BLOCK + 1)
360
361/*
362 * Inode flags
363 */
364#define EXT4_SECRM_FL 0x00000001 /* Secure deletion */
365#define EXT4_UNRM_FL 0x00000002 /* Undelete */
366#define EXT4_COMPR_FL 0x00000004 /* Compress file */
367#define EXT4_SYNC_FL 0x00000008 /* Synchronous updates */
368#define EXT4_IMMUTABLE_FL 0x00000010 /* Immutable file */
369#define EXT4_APPEND_FL 0x00000020 /* writes to file may only append */
370#define EXT4_NODUMP_FL 0x00000040 /* do not dump file */
371#define EXT4_NOATIME_FL 0x00000080 /* do not update atime */
372/* Reserved for compression usage... */
373#define EXT4_DIRTY_FL 0x00000100
374#define EXT4_COMPRBLK_FL 0x00000200 /* One or more compressed clusters */
375#define EXT4_NOCOMPR_FL 0x00000400 /* Don't compress */
376 /* nb: was previously EXT2_ECOMPR_FL */
377#define EXT4_ENCRYPT_FL 0x00000800 /* encrypted file */
378/* End compression flags --- maybe not all used */
379#define EXT4_INDEX_FL 0x00001000 /* hash-indexed directory */
380#define EXT4_IMAGIC_FL 0x00002000 /* AFS directory */
381#define EXT4_JOURNAL_DATA_FL 0x00004000 /* file data should be journaled */
382#define EXT4_NOTAIL_FL 0x00008000 /* file tail should not be merged */
383#define EXT4_DIRSYNC_FL 0x00010000 /* dirsync behaviour (directories only) */
384#define EXT4_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/
385#define EXT4_HUGE_FILE_FL 0x00040000 /* Set to each huge file */
386#define EXT4_EXTENTS_FL 0x00080000 /* Inode uses extents */
387#define EXT4_EA_INODE_FL 0x00200000 /* Inode used for large EA */
388#define EXT4_EOFBLOCKS_FL 0x00400000 /* Blocks allocated beyond EOF */
389#define EXT4_INLINE_DATA_FL 0x10000000 /* Inode has inline data. */
390#define EXT4_PROJINHERIT_FL 0x20000000 /* Create with parents projid */
391#define EXT4_RESERVED_FL 0x80000000 /* reserved for ext4 lib */
392
393#define EXT4_FL_USER_VISIBLE 0x304BDFFF /* User visible flags */
394#define EXT4_FL_USER_MODIFIABLE 0x204380FF /* User modifiable flags */
395
396#define EXT4_FL_XFLAG_VISIBLE (EXT4_SYNC_FL | \
397 EXT4_IMMUTABLE_FL | \
398 EXT4_APPEND_FL | \
399 EXT4_NODUMP_FL | \
400 EXT4_NOATIME_FL | \
401 EXT4_PROJINHERIT_FL)
402
403/* Flags that should be inherited by new inodes from their parent. */
404#define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
405 EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
406 EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
407 EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL |\
408 EXT4_PROJINHERIT_FL)
409
410/* Flags that are appropriate for regular files (all but dir-specific ones). */
411#define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
412
413/* Flags that are appropriate for non-directories/regular files. */
414#define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
415
416/* Mask out flags that are inappropriate for the given type of inode. */
417static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
418{
419 if (S_ISDIR(mode))
420 return flags;
421 else if (S_ISREG(mode))
422 return flags & EXT4_REG_FLMASK;
423 else
424 return flags & EXT4_OTHER_FLMASK;
425}
426
427/*
428 * Inode flags used for atomic set/get
429 */
430enum {
431 EXT4_INODE_SECRM = 0, /* Secure deletion */
432 EXT4_INODE_UNRM = 1, /* Undelete */
433 EXT4_INODE_COMPR = 2, /* Compress file */
434 EXT4_INODE_SYNC = 3, /* Synchronous updates */
435 EXT4_INODE_IMMUTABLE = 4, /* Immutable file */
436 EXT4_INODE_APPEND = 5, /* writes to file may only append */
437 EXT4_INODE_NODUMP = 6, /* do not dump file */
438 EXT4_INODE_NOATIME = 7, /* do not update atime */
439/* Reserved for compression usage... */
440 EXT4_INODE_DIRTY = 8,
441 EXT4_INODE_COMPRBLK = 9, /* One or more compressed clusters */
442 EXT4_INODE_NOCOMPR = 10, /* Don't compress */
443 EXT4_INODE_ENCRYPT = 11, /* Encrypted file */
444/* End compression flags --- maybe not all used */
445 EXT4_INODE_INDEX = 12, /* hash-indexed directory */
446 EXT4_INODE_IMAGIC = 13, /* AFS directory */
447 EXT4_INODE_JOURNAL_DATA = 14, /* file data should be journaled */
448 EXT4_INODE_NOTAIL = 15, /* file tail should not be merged */
449 EXT4_INODE_DIRSYNC = 16, /* dirsync behaviour (directories only) */
450 EXT4_INODE_TOPDIR = 17, /* Top of directory hierarchies*/
451 EXT4_INODE_HUGE_FILE = 18, /* Set to each huge file */
452 EXT4_INODE_EXTENTS = 19, /* Inode uses extents */
453 EXT4_INODE_EA_INODE = 21, /* Inode used for large EA */
454 EXT4_INODE_EOFBLOCKS = 22, /* Blocks allocated beyond EOF */
455 EXT4_INODE_INLINE_DATA = 28, /* Data in inode. */
456 EXT4_INODE_PROJINHERIT = 29, /* Create with parents projid */
457 EXT4_INODE_RESERVED = 31, /* reserved for ext4 lib */
458};
459
460/*
461 * Since it's pretty easy to mix up bit numbers and hex values, we use a
462 * build-time check to make sure that EXT4_XXX_FL is consistent with respect to
463 * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost
464 * any extra space in the compiled kernel image, otherwise, the build will fail.
465 * It's important that these values are the same, since we are using
466 * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent
467 * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk
468 * values found in ext2, ext3 and ext4 filesystems, and of course the values
469 * defined in e2fsprogs.
470 *
471 * It's not paranoia if the Murphy's Law really *is* out to get you. :-)
472 */
473#define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
474#define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG))
475
476static inline void ext4_check_flag_values(void)
477{
478 CHECK_FLAG_VALUE(SECRM);
479 CHECK_FLAG_VALUE(UNRM);
480 CHECK_FLAG_VALUE(COMPR);
481 CHECK_FLAG_VALUE(SYNC);
482 CHECK_FLAG_VALUE(IMMUTABLE);
483 CHECK_FLAG_VALUE(APPEND);
484 CHECK_FLAG_VALUE(NODUMP);
485 CHECK_FLAG_VALUE(NOATIME);
486 CHECK_FLAG_VALUE(DIRTY);
487 CHECK_FLAG_VALUE(COMPRBLK);
488 CHECK_FLAG_VALUE(NOCOMPR);
489 CHECK_FLAG_VALUE(ENCRYPT);
490 CHECK_FLAG_VALUE(INDEX);
491 CHECK_FLAG_VALUE(IMAGIC);
492 CHECK_FLAG_VALUE(JOURNAL_DATA);
493 CHECK_FLAG_VALUE(NOTAIL);
494 CHECK_FLAG_VALUE(DIRSYNC);
495 CHECK_FLAG_VALUE(TOPDIR);
496 CHECK_FLAG_VALUE(HUGE_FILE);
497 CHECK_FLAG_VALUE(EXTENTS);
498 CHECK_FLAG_VALUE(EA_INODE);
499 CHECK_FLAG_VALUE(EOFBLOCKS);
500 CHECK_FLAG_VALUE(INLINE_DATA);
501 CHECK_FLAG_VALUE(PROJINHERIT);
502 CHECK_FLAG_VALUE(RESERVED);
503}
504
505/* Used to pass group descriptor data when online resize is done */
506struct ext4_new_group_input {
507 __u32 group; /* Group number for this data */
508 __u64 block_bitmap; /* Absolute block number of block bitmap */
509 __u64 inode_bitmap; /* Absolute block number of inode bitmap */
510 __u64 inode_table; /* Absolute block number of inode table start */
511 __u32 blocks_count; /* Total number of blocks in this group */
512 __u16 reserved_blocks; /* Number of reserved blocks in this group */
513 __u16 unused;
514};
515
516#if defined(__KERNEL__) && defined(CONFIG_COMPAT)
517struct compat_ext4_new_group_input {
518 u32 group;
519 compat_u64 block_bitmap;
520 compat_u64 inode_bitmap;
521 compat_u64 inode_table;
522 u32 blocks_count;
523 u16 reserved_blocks;
524 u16 unused;
525};
526#endif
527
528/* The struct ext4_new_group_input in kernel space, with free_blocks_count */
529struct ext4_new_group_data {
530 __u32 group;
531 __u64 block_bitmap;
532 __u64 inode_bitmap;
533 __u64 inode_table;
534 __u32 blocks_count;
535 __u16 reserved_blocks;
536 __u16 unused;
537 __u32 free_blocks_count;
538};
539
540/* Indexes used to index group tables in ext4_new_group_data */
541enum {
542 BLOCK_BITMAP = 0, /* block bitmap */
543 INODE_BITMAP, /* inode bitmap */
544 INODE_TABLE, /* inode tables */
545 GROUP_TABLE_COUNT,
546};
547
548/*
549 * Flags used by ext4_map_blocks()
550 */
551 /* Allocate any needed blocks and/or convert an unwritten
552 extent to be an initialized ext4 */
553#define EXT4_GET_BLOCKS_CREATE 0x0001
554 /* Request the creation of an unwritten extent */
555#define EXT4_GET_BLOCKS_UNWRIT_EXT 0x0002
556#define EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT (EXT4_GET_BLOCKS_UNWRIT_EXT|\
557 EXT4_GET_BLOCKS_CREATE)
558 /* Caller is from the delayed allocation writeout path
559 * finally doing the actual allocation of delayed blocks */
560#define EXT4_GET_BLOCKS_DELALLOC_RESERVE 0x0004
561 /* caller is from the direct IO path, request to creation of an
562 unwritten extents if not allocated, split the unwritten
563 extent if blocks has been preallocated already*/
564#define EXT4_GET_BLOCKS_PRE_IO 0x0008
565#define EXT4_GET_BLOCKS_CONVERT 0x0010
566#define EXT4_GET_BLOCKS_IO_CREATE_EXT (EXT4_GET_BLOCKS_PRE_IO|\
567 EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
568 /* Convert extent to initialized after IO complete */
569#define EXT4_GET_BLOCKS_IO_CONVERT_EXT (EXT4_GET_BLOCKS_CONVERT|\
570 EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT)
571 /* Eventual metadata allocation (due to growing extent tree)
572 * should not fail, so try to use reserved blocks for that.*/
573#define EXT4_GET_BLOCKS_METADATA_NOFAIL 0x0020
574 /* Don't normalize allocation size (used for fallocate) */
575#define EXT4_GET_BLOCKS_NO_NORMALIZE 0x0040
576 /* Request will not result in inode size update (user for fallocate) */
577#define EXT4_GET_BLOCKS_KEEP_SIZE 0x0080
578 /* Convert written extents to unwritten */
579#define EXT4_GET_BLOCKS_CONVERT_UNWRITTEN 0x0100
580 /* Write zeros to newly created written extents */
581#define EXT4_GET_BLOCKS_ZERO 0x0200
582#define EXT4_GET_BLOCKS_CREATE_ZERO (EXT4_GET_BLOCKS_CREATE |\
583 EXT4_GET_BLOCKS_ZERO)
584
585/*
586 * The bit position of these flags must not overlap with any of the
587 * EXT4_GET_BLOCKS_*. They are used by ext4_find_extent(),
588 * read_extent_tree_block(), ext4_split_extent_at(),
589 * ext4_ext_insert_extent(), and ext4_ext_create_new_leaf().
590 * EXT4_EX_NOCACHE is used to indicate that the we shouldn't be
591 * caching the extents when reading from the extent tree while a
592 * truncate or punch hole operation is in progress.
593 */
594#define EXT4_EX_NOCACHE 0x40000000
595#define EXT4_EX_FORCE_CACHE 0x20000000
596
597/*
598 * Flags used by ext4_free_blocks
599 */
600#define EXT4_FREE_BLOCKS_METADATA 0x0001
601#define EXT4_FREE_BLOCKS_FORGET 0x0002
602#define EXT4_FREE_BLOCKS_VALIDATED 0x0004
603#define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
604#define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER 0x0010
605#define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER 0x0020
606
607/* Encryption algorithms */
608#define EXT4_ENCRYPTION_MODE_INVALID 0
609#define EXT4_ENCRYPTION_MODE_AES_256_XTS 1
610#define EXT4_ENCRYPTION_MODE_AES_256_GCM 2
611#define EXT4_ENCRYPTION_MODE_AES_256_CBC 3
612#define EXT4_ENCRYPTION_MODE_AES_256_CTS 4
613
614#include "ext4_crypto.h"
615
616/*
617 * ioctl commands
618 */
619#define EXT4_IOC_GETFLAGS FS_IOC_GETFLAGS
620#define EXT4_IOC_SETFLAGS FS_IOC_SETFLAGS
621#define EXT4_IOC_GETVERSION _IOR('f', 3, long)
622#define EXT4_IOC_SETVERSION _IOW('f', 4, long)
623#define EXT4_IOC_GETVERSION_OLD FS_IOC_GETVERSION
624#define EXT4_IOC_SETVERSION_OLD FS_IOC_SETVERSION
625#define EXT4_IOC_GETRSVSZ _IOR('f', 5, long)
626#define EXT4_IOC_SETRSVSZ _IOW('f', 6, long)
627#define EXT4_IOC_GROUP_EXTEND _IOW('f', 7, unsigned long)
628#define EXT4_IOC_GROUP_ADD _IOW('f', 8, struct ext4_new_group_input)
629#define EXT4_IOC_MIGRATE _IO('f', 9)
630 /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
631 /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
632#define EXT4_IOC_ALLOC_DA_BLKS _IO('f', 12)
633#define EXT4_IOC_MOVE_EXT _IOWR('f', 15, struct move_extent)
634#define EXT4_IOC_RESIZE_FS _IOW('f', 16, __u64)
635#define EXT4_IOC_SWAP_BOOT _IO('f', 17)
636#define EXT4_IOC_PRECACHE_EXTENTS _IO('f', 18)
637#define EXT4_IOC_SET_ENCRYPTION_POLICY _IOR('f', 19, struct ext4_encryption_policy)
638#define EXT4_IOC_GET_ENCRYPTION_PWSALT _IOW('f', 20, __u8[16])
639#define EXT4_IOC_GET_ENCRYPTION_POLICY _IOW('f', 21, struct ext4_encryption_policy)
640
641#ifndef FS_IOC_FSGETXATTR
642/* Until the uapi changes get merged for project quota... */
643
644#define FS_IOC_FSGETXATTR _IOR('X', 31, struct fsxattr)
645#define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
646
647/*
648 * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
649 */
650struct fsxattr {
651 __u32 fsx_xflags; /* xflags field value (get/set) */
652 __u32 fsx_extsize; /* extsize field value (get/set)*/
653 __u32 fsx_nextents; /* nextents field value (get) */
654 __u32 fsx_projid; /* project identifier (get/set) */
655 unsigned char fsx_pad[12];
656};
657
658/*
659 * Flags for the fsx_xflags field
660 */
661#define FS_XFLAG_REALTIME 0x00000001 /* data in realtime volume */
662#define FS_XFLAG_PREALLOC 0x00000002 /* preallocated file extents */
663#define FS_XFLAG_IMMUTABLE 0x00000008 /* file cannot be modified */
664#define FS_XFLAG_APPEND 0x00000010 /* all writes append */
665#define FS_XFLAG_SYNC 0x00000020 /* all writes synchronous */
666#define FS_XFLAG_NOATIME 0x00000040 /* do not update access time */
667#define FS_XFLAG_NODUMP 0x00000080 /* do not include in backups */
668#define FS_XFLAG_RTINHERIT 0x00000100 /* create with rt bit set */
669#define FS_XFLAG_PROJINHERIT 0x00000200 /* create with parents projid */
670#define FS_XFLAG_NOSYMLINKS 0x00000400 /* disallow symlink creation */
671#define FS_XFLAG_EXTSIZE 0x00000800 /* extent size allocator hint */
672#define FS_XFLAG_EXTSZINHERIT 0x00001000 /* inherit inode extent size */
673#define FS_XFLAG_NODEFRAG 0x00002000 /* do not defragment */
674#define FS_XFLAG_FILESTREAM 0x00004000 /* use filestream allocator */
675#define FS_XFLAG_HASATTR 0x80000000 /* no DIFLAG for this */
676#endif /* !defined(FS_IOC_FSGETXATTR) */
677
678#define EXT4_IOC_FSGETXATTR FS_IOC_FSGETXATTR
679#define EXT4_IOC_FSSETXATTR FS_IOC_FSSETXATTR
680
681#if defined(__KERNEL__) && defined(CONFIG_COMPAT)
682/*
683 * ioctl commands in 32 bit emulation
684 */
685#define EXT4_IOC32_GETFLAGS FS_IOC32_GETFLAGS
686#define EXT4_IOC32_SETFLAGS FS_IOC32_SETFLAGS
687#define EXT4_IOC32_GETVERSION _IOR('f', 3, int)
688#define EXT4_IOC32_SETVERSION _IOW('f', 4, int)
689#define EXT4_IOC32_GETRSVSZ _IOR('f', 5, int)
690#define EXT4_IOC32_SETRSVSZ _IOW('f', 6, int)
691#define EXT4_IOC32_GROUP_EXTEND _IOW('f', 7, unsigned int)
692#define EXT4_IOC32_GROUP_ADD _IOW('f', 8, struct compat_ext4_new_group_input)
693#define EXT4_IOC32_GETVERSION_OLD FS_IOC32_GETVERSION
694#define EXT4_IOC32_SETVERSION_OLD FS_IOC32_SETVERSION
695#endif
696
697/* Max physical block we can address w/o extents */
698#define EXT4_MAX_BLOCK_FILE_PHYS 0xFFFFFFFF
699
700/*
701 * Structure of an inode on the disk
702 */
703struct ext4_inode {
704 __le16 i_mode; /* File mode */
705 __le16 i_uid; /* Low 16 bits of Owner Uid */
706 __le32 i_size_lo; /* Size in bytes */
707 __le32 i_atime; /* Access time */
708 __le32 i_ctime; /* Inode Change time */
709 __le32 i_mtime; /* Modification time */
710 __le32 i_dtime; /* Deletion Time */
711 __le16 i_gid; /* Low 16 bits of Group Id */
712 __le16 i_links_count; /* Links count */
713 __le32 i_blocks_lo; /* Blocks count */
714 __le32 i_flags; /* File flags */
715 union {
716 struct {
717 __le32 l_i_version;
718 } linux1;
719 struct {
720 __u32 h_i_translator;
721 } hurd1;
722 struct {
723 __u32 m_i_reserved1;
724 } masix1;
725 } osd1; /* OS dependent 1 */
726 __le32 i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
727 __le32 i_generation; /* File version (for NFS) */
728 __le32 i_file_acl_lo; /* File ACL */
729 __le32 i_size_high;
730 __le32 i_obso_faddr; /* Obsoleted fragment address */
731 union {
732 struct {
733 __le16 l_i_blocks_high; /* were l_i_reserved1 */
734 __le16 l_i_file_acl_high;
735 __le16 l_i_uid_high; /* these 2 fields */
736 __le16 l_i_gid_high; /* were reserved2[0] */
737 __le16 l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
738 __le16 l_i_reserved;
739 } linux2;
740 struct {
741 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */
742 __u16 h_i_mode_high;
743 __u16 h_i_uid_high;
744 __u16 h_i_gid_high;
745 __u32 h_i_author;
746 } hurd2;
747 struct {
748 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */
749 __le16 m_i_file_acl_high;
750 __u32 m_i_reserved2[2];
751 } masix2;
752 } osd2; /* OS dependent 2 */
753 __le16 i_extra_isize;
754 __le16 i_checksum_hi; /* crc32c(uuid+inum+inode) BE */
755 __le32 i_ctime_extra; /* extra Change time (nsec << 2 | epoch) */
756 __le32 i_mtime_extra; /* extra Modification time(nsec << 2 | epoch) */
757 __le32 i_atime_extra; /* extra Access time (nsec << 2 | epoch) */
758 __le32 i_crtime; /* File Creation time */
759 __le32 i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
760 __le32 i_version_hi; /* high 32 bits for 64-bit version */
761 __le32 i_projid; /* Project ID */
762};
763
764struct move_extent {
765 __u32 reserved; /* should be zero */
766 __u32 donor_fd; /* donor file descriptor */
767 __u64 orig_start; /* logical start offset in block for orig */
768 __u64 donor_start; /* logical start offset in block for donor */
769 __u64 len; /* block length to be moved */
770 __u64 moved_len; /* moved block length */
771};
772
773#define EXT4_EPOCH_BITS 2
774#define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
775#define EXT4_NSEC_MASK (~0UL << EXT4_EPOCH_BITS)
776
777/*
778 * Extended fields will fit into an inode if the filesystem was formatted
779 * with large inodes (-I 256 or larger) and there are not currently any EAs
780 * consuming all of the available space. For new inodes we always reserve
781 * enough space for the kernel's known extended fields, but for inodes
782 * created with an old kernel this might not have been the case. None of
783 * the extended inode fields is critical for correct filesystem operation.
784 * This macro checks if a certain field fits in the inode. Note that
785 * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
786 */
787#define EXT4_FITS_IN_INODE(ext4_inode, einode, field) \
788 ((offsetof(typeof(*ext4_inode), field) + \
789 sizeof((ext4_inode)->field)) \
790 <= (EXT4_GOOD_OLD_INODE_SIZE + \
791 (einode)->i_extra_isize)) \
792
793/*
794 * We use an encoding that preserves the times for extra epoch "00":
795 *
796 * extra msb of adjust for signed
797 * epoch 32-bit 32-bit tv_sec to
798 * bits time decoded 64-bit tv_sec 64-bit tv_sec valid time range
799 * 0 0 1 -0x80000000..-0x00000001 0x000000000 1901-12-13..1969-12-31
800 * 0 0 0 0x000000000..0x07fffffff 0x000000000 1970-01-01..2038-01-19
801 * 0 1 1 0x080000000..0x0ffffffff 0x100000000 2038-01-19..2106-02-07
802 * 0 1 0 0x100000000..0x17fffffff 0x100000000 2106-02-07..2174-02-25
803 * 1 0 1 0x180000000..0x1ffffffff 0x200000000 2174-02-25..2242-03-16
804 * 1 0 0 0x200000000..0x27fffffff 0x200000000 2242-03-16..2310-04-04
805 * 1 1 1 0x280000000..0x2ffffffff 0x300000000 2310-04-04..2378-04-22
806 * 1 1 0 0x300000000..0x37fffffff 0x300000000 2378-04-22..2446-05-10
807 *
808 * Note that previous versions of the kernel on 64-bit systems would
809 * incorrectly use extra epoch bits 1,1 for dates between 1901 and
810 * 1970. e2fsck will correct this, assuming that it is run on the
811 * affected filesystem before 2242.
812 */
813
814static inline __le32 ext4_encode_extra_time(struct timespec *time)
815{
816 u32 extra = sizeof(time->tv_sec) > 4 ?
817 ((time->tv_sec - (s32)time->tv_sec) >> 32) & EXT4_EPOCH_MASK : 0;
818 return cpu_to_le32(extra | (time->tv_nsec << EXT4_EPOCH_BITS));
819}
820
821static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
822{
823 if (unlikely(sizeof(time->tv_sec) > 4 &&
824 (extra & cpu_to_le32(EXT4_EPOCH_MASK)))) {
825#if LINUX_VERSION_CODE < KERNEL_VERSION(4,20,0)
826 /* Handle legacy encoding of pre-1970 dates with epoch
827 * bits 1,1. We assume that by kernel version 4.20,
828 * everyone will have run fsck over the affected
829 * filesystems to correct the problem. (This
830 * backwards compatibility may be removed before this
831 * time, at the discretion of the ext4 developers.)
832 */
833 u64 extra_bits = le32_to_cpu(extra) & EXT4_EPOCH_MASK;
834 if (extra_bits == 3 && ((time->tv_sec) & 0x80000000) != 0)
835 extra_bits = 0;
836 time->tv_sec += extra_bits << 32;
837#else
838 time->tv_sec += (u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK) << 32;
839#endif
840 }
841 time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
842}
843
844#define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode) \
845do { \
846 (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec); \
847 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \
848 (raw_inode)->xtime ## _extra = \
849 ext4_encode_extra_time(&(inode)->xtime); \
850} while (0)
851
852#define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode) \
853do { \
854 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \
855 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec); \
856 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \
857 (raw_inode)->xtime ## _extra = \
858 ext4_encode_extra_time(&(einode)->xtime); \
859} while (0)
860
861#define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode) \
862do { \
863 (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime); \
864 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \
865 ext4_decode_extra_time(&(inode)->xtime, \
866 raw_inode->xtime ## _extra); \
867 else \
868 (inode)->xtime.tv_nsec = 0; \
869} while (0)
870
871#define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode) \
872do { \
873 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \
874 (einode)->xtime.tv_sec = \
875 (signed)le32_to_cpu((raw_inode)->xtime); \
876 else \
877 (einode)->xtime.tv_sec = 0; \
878 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \
879 ext4_decode_extra_time(&(einode)->xtime, \
880 raw_inode->xtime ## _extra); \
881 else \
882 (einode)->xtime.tv_nsec = 0; \
883} while (0)
884
885#define i_disk_version osd1.linux1.l_i_version
886
887#if defined(__KERNEL__) || defined(__linux__)
888#define i_reserved1 osd1.linux1.l_i_reserved1
889#define i_file_acl_high osd2.linux2.l_i_file_acl_high
890#define i_blocks_high osd2.linux2.l_i_blocks_high
891#define i_uid_low i_uid
892#define i_gid_low i_gid
893#define i_uid_high osd2.linux2.l_i_uid_high
894#define i_gid_high osd2.linux2.l_i_gid_high
895#define i_checksum_lo osd2.linux2.l_i_checksum_lo
896
897#elif defined(__GNU__)
898
899#define i_translator osd1.hurd1.h_i_translator
900#define i_uid_high osd2.hurd2.h_i_uid_high
901#define i_gid_high osd2.hurd2.h_i_gid_high
902#define i_author osd2.hurd2.h_i_author
903
904#elif defined(__masix__)
905
906#define i_reserved1 osd1.masix1.m_i_reserved1
907#define i_file_acl_high osd2.masix2.m_i_file_acl_high
908#define i_reserved2 osd2.masix2.m_i_reserved2
909
910#endif /* defined(__KERNEL__) || defined(__linux__) */
911
912#include "extents_status.h"
913
914/*
915 * Lock subclasses for i_data_sem in the ext4_inode_info structure.
916 *
917 * These are needed to avoid lockdep false positives when we need to
918 * allocate blocks to the quota inode during ext4_map_blocks(), while
919 * holding i_data_sem for a normal (non-quota) inode. Since we don't
920 * do quota tracking for the quota inode, this avoids deadlock (as
921 * well as infinite recursion, since it isn't turtles all the way
922 * down...)
923 *
924 * I_DATA_SEM_NORMAL - Used for most inodes
925 * I_DATA_SEM_OTHER - Used by move_inode.c for the second normal inode
926 * where the second inode has larger inode number
927 * than the first
928 * I_DATA_SEM_QUOTA - Used for quota inodes only
929 */
930enum {
931 I_DATA_SEM_NORMAL = 0,
932 I_DATA_SEM_OTHER,
933 I_DATA_SEM_QUOTA,
934};
935
936
937/*
938 * fourth extended file system inode data in memory
939 */
940struct ext4_inode_info {
941 __le32 i_data[15]; /* unconverted */
942 __u32 i_dtime;
943 ext4_fsblk_t i_file_acl;
944
945 /*
946 * i_block_group is the number of the block group which contains
947 * this file's inode. Constant across the lifetime of the inode,
948 * it is ued for making block allocation decisions - we try to
949 * place a file's data blocks near its inode block, and new inodes
950 * near to their parent directory's inode.
951 */
952 ext4_group_t i_block_group;
953 ext4_lblk_t i_dir_start_lookup;
954#if (BITS_PER_LONG < 64)
955 unsigned long i_state_flags; /* Dynamic state flags */
956#endif
957 unsigned long i_flags;
958
959 /*
960 * Extended attributes can be read independently of the main file
961 * data. Taking i_mutex even when reading would cause contention
962 * between readers of EAs and writers of regular file data, so
963 * instead we synchronize on xattr_sem when reading or changing
964 * EAs.
965 */
966 struct rw_semaphore xattr_sem;
967
968 struct list_head i_orphan; /* unlinked but open inodes */
969
970 /*
971 * i_disksize keeps track of what the inode size is ON DISK, not
972 * in memory. During truncate, i_size is set to the new size by
973 * the VFS prior to calling ext4_truncate(), but the filesystem won't
974 * set i_disksize to 0 until the truncate is actually under way.
975 *
976 * The intent is that i_disksize always represents the blocks which
977 * are used by this file. This allows recovery to restart truncate
978 * on orphans if we crash during truncate. We actually write i_disksize
979 * into the on-disk inode when writing inodes out, instead of i_size.
980 *
981 * The only time when i_disksize and i_size may be different is when
982 * a truncate is in progress. The only things which change i_disksize
983 * are ext4_get_block (growth) and ext4_truncate (shrinkth).
984 */
985 loff_t i_disksize;
986
987 /*
988 * i_data_sem is for serialising ext4_truncate() against
989 * ext4_getblock(). In the 2.4 ext2 design, great chunks of inode's
990 * data tree are chopped off during truncate. We can't do that in
991 * ext4 because whenever we perform intermediate commits during
992 * truncate, the inode and all the metadata blocks *must* be in a
993 * consistent state which allows truncation of the orphans to restart
994 * during recovery. Hence we must fix the get_block-vs-truncate race
995 * by other means, so we have i_data_sem.
996 */
997 struct rw_semaphore i_data_sem;
998 /*
999 * i_mmap_sem is for serializing page faults with truncate / punch hole
1000 * operations. We have to make sure that new page cannot be faulted in
1001 * a section of the inode that is being punched. We cannot easily use
1002 * i_data_sem for this since we need protection for the whole punch
1003 * operation and i_data_sem ranks below transaction start so we have
1004 * to occasionally drop it.
1005 */
1006 struct rw_semaphore i_mmap_sem;
1007 struct inode vfs_inode;
1008 struct jbd2_inode *jinode;
1009
1010 spinlock_t i_raw_lock; /* protects updates to the raw inode */
1011
1012 /*
1013 * File creation time. Its function is same as that of
1014 * struct timespec i_{a,c,m}time in the generic inode.
1015 */
1016 struct timespec i_crtime;
1017
1018 /* mballoc */
1019 struct list_head i_prealloc_list;
1020 spinlock_t i_prealloc_lock;
1021
1022 /* extents status tree */
1023 struct ext4_es_tree i_es_tree;
1024 rwlock_t i_es_lock;
1025 struct list_head i_es_list;
1026 unsigned int i_es_all_nr; /* protected by i_es_lock */
1027 unsigned int i_es_shk_nr; /* protected by i_es_lock */
1028 ext4_lblk_t i_es_shrink_lblk; /* Offset where we start searching for
1029 extents to shrink. Protected by
1030 i_es_lock */
1031
1032 /* ialloc */
1033 ext4_group_t i_last_alloc_group;
1034
1035 /* allocation reservation info for delalloc */
1036 /* In case of bigalloc, these refer to clusters rather than blocks */
1037 unsigned int i_reserved_data_blocks;
1038 unsigned int i_reserved_meta_blocks;
1039 unsigned int i_allocated_meta_blocks;
1040 ext4_lblk_t i_da_metadata_calc_last_lblock;
1041 int i_da_metadata_calc_len;
1042
1043 /* on-disk additional length */
1044 __u16 i_extra_isize;
1045
1046 /* Indicate the inline data space. */
1047 u16 i_inline_off;
1048 u16 i_inline_size;
1049
1050#ifdef CONFIG_QUOTA
1051 /* quota space reservation, managed internally by quota code */
1052 qsize_t i_reserved_quota;
1053#endif
1054
1055 /* Lock protecting lists below */
1056 spinlock_t i_completed_io_lock;
1057 /*
1058 * Completed IOs that need unwritten extents handling and have
1059 * transaction reserved
1060 */
1061 struct list_head i_rsv_conversion_list;
1062 struct work_struct i_rsv_conversion_work;
1063 atomic_t i_unwritten; /* Nr. of inflight conversions pending */
1064
1065 spinlock_t i_block_reservation_lock;
1066
1067 /*
1068 * Transactions that contain inode's metadata needed to complete
1069 * fsync and fdatasync, respectively.
1070 */
1071 tid_t i_sync_tid;
1072 tid_t i_datasync_tid;
1073
1074#ifdef CONFIG_QUOTA
1075 struct dquot *i_dquot[MAXQUOTAS];
1076#endif
1077
1078 /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
1079 __u32 i_csum_seed;
1080
1081#ifdef CONFIG_EXT4_FS_ENCRYPTION
1082 /* Encryption params */
1083 struct ext4_crypt_info *i_crypt_info;
1084#endif
1085 kprojid_t i_projid;
1086};
1087
1088/*
1089 * File system states
1090 */
1091#define EXT4_VALID_FS 0x0001 /* Unmounted cleanly */
1092#define EXT4_ERROR_FS 0x0002 /* Errors detected */
1093#define EXT4_ORPHAN_FS 0x0004 /* Orphans being recovered */
1094
1095/*
1096 * Misc. filesystem flags
1097 */
1098#define EXT2_FLAGS_SIGNED_HASH 0x0001 /* Signed dirhash in use */
1099#define EXT2_FLAGS_UNSIGNED_HASH 0x0002 /* Unsigned dirhash in use */
1100#define EXT2_FLAGS_TEST_FILESYS 0x0004 /* to test development code */
1101
1102/*
1103 * Mount flags set via mount options or defaults
1104 */
1105#define EXT4_MOUNT_GRPID 0x00004 /* Create files with directory's group */
1106#define EXT4_MOUNT_DEBUG 0x00008 /* Some debugging messages */
1107#define EXT4_MOUNT_ERRORS_CONT 0x00010 /* Continue on errors */
1108#define EXT4_MOUNT_ERRORS_RO 0x00020 /* Remount fs ro on errors */
1109#define EXT4_MOUNT_ERRORS_PANIC 0x00040 /* Panic on errors */
1110#define EXT4_MOUNT_ERRORS_MASK 0x00070
1111#define EXT4_MOUNT_MINIX_DF 0x00080 /* Mimics the Minix statfs */
1112#define EXT4_MOUNT_NOLOAD 0x00100 /* Don't use existing journal*/
1113#ifdef CONFIG_FS_DAX
1114#define EXT4_MOUNT_DAX 0x00200 /* Direct Access */
1115#else
1116#define EXT4_MOUNT_DAX 0
1117#endif
1118#define EXT4_MOUNT_DATA_FLAGS 0x00C00 /* Mode for data writes: */
1119#define EXT4_MOUNT_JOURNAL_DATA 0x00400 /* Write data to journal */
1120#define EXT4_MOUNT_ORDERED_DATA 0x00800 /* Flush data before commit */
1121#define EXT4_MOUNT_WRITEBACK_DATA 0x00C00 /* No data ordering */
1122#define EXT4_MOUNT_UPDATE_JOURNAL 0x01000 /* Update the journal format */
1123#define EXT4_MOUNT_NO_UID32 0x02000 /* Disable 32-bit UIDs */
1124#define EXT4_MOUNT_XATTR_USER 0x04000 /* Extended user attributes */
1125#define EXT4_MOUNT_POSIX_ACL 0x08000 /* POSIX Access Control Lists */
1126#define EXT4_MOUNT_NO_AUTO_DA_ALLOC 0x10000 /* No auto delalloc mapping */
1127#define EXT4_MOUNT_BARRIER 0x20000 /* Use block barriers */
1128#define EXT4_MOUNT_QUOTA 0x80000 /* Some quota option set */
1129#define EXT4_MOUNT_USRQUOTA 0x100000 /* "old" user quota */
1130#define EXT4_MOUNT_GRPQUOTA 0x200000 /* "old" group quota */
1131#define EXT4_MOUNT_DIOREAD_NOLOCK 0x400000 /* Enable support for dio read nolocking */
1132#define EXT4_MOUNT_JOURNAL_CHECKSUM 0x800000 /* Journal checksums */
1133#define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
1134#define EXT4_MOUNT_DELALLOC 0x8000000 /* Delalloc support */
1135#define EXT4_MOUNT_DATA_ERR_ABORT 0x10000000 /* Abort on file data write */
1136#define EXT4_MOUNT_BLOCK_VALIDITY 0x20000000 /* Block validity checking */
1137#define EXT4_MOUNT_DISCARD 0x40000000 /* Issue DISCARD requests */
1138#define EXT4_MOUNT_INIT_INODE_TABLE 0x80000000 /* Initialize uninitialized itables */
1139
1140/*
1141 * Mount flags set either automatically (could not be set by mount option)
1142 * based on per file system feature or property or in special cases such as
1143 * distinguishing between explicit mount option definition and default.
1144 */
1145#define EXT4_MOUNT2_EXPLICIT_DELALLOC 0x00000001 /* User explicitly
1146 specified delalloc */
1147#define EXT4_MOUNT2_STD_GROUP_SIZE 0x00000002 /* We have standard group
1148 size of blocksize * 8
1149 blocks */
1150#define EXT4_MOUNT2_HURD_COMPAT 0x00000004 /* Support HURD-castrated
1151 file systems */
1152
1153#define EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM 0x00000008 /* User explicitly
1154 specified journal checksum */
1155
1156#define clear_opt(sb, opt) EXT4_SB(sb)->s_mount_opt &= \
1157 ~EXT4_MOUNT_##opt
1158#define set_opt(sb, opt) EXT4_SB(sb)->s_mount_opt |= \
1159 EXT4_MOUNT_##opt
1160#define test_opt(sb, opt) (EXT4_SB(sb)->s_mount_opt & \
1161 EXT4_MOUNT_##opt)
1162
1163#define clear_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 &= \
1164 ~EXT4_MOUNT2_##opt
1165#define set_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 |= \
1166 EXT4_MOUNT2_##opt
1167#define test_opt2(sb, opt) (EXT4_SB(sb)->s_mount_opt2 & \
1168 EXT4_MOUNT2_##opt)
1169
1170#define ext4_test_and_set_bit __test_and_set_bit_le
1171#define ext4_set_bit __set_bit_le
1172#define ext4_set_bit_atomic ext2_set_bit_atomic
1173#define ext4_test_and_clear_bit __test_and_clear_bit_le
1174#define ext4_clear_bit __clear_bit_le
1175#define ext4_clear_bit_atomic ext2_clear_bit_atomic
1176#define ext4_test_bit test_bit_le
1177#define ext4_find_next_zero_bit find_next_zero_bit_le
1178#define ext4_find_next_bit find_next_bit_le
1179
1180extern void ext4_set_bits(void *bm, int cur, int len);
1181
1182/*
1183 * Maximal mount counts between two filesystem checks
1184 */
1185#define EXT4_DFL_MAX_MNT_COUNT 20 /* Allow 20 mounts */
1186#define EXT4_DFL_CHECKINTERVAL 0 /* Don't use interval check */
1187
1188/*
1189 * Behaviour when detecting errors
1190 */
1191#define EXT4_ERRORS_CONTINUE 1 /* Continue execution */
1192#define EXT4_ERRORS_RO 2 /* Remount fs read-only */
1193#define EXT4_ERRORS_PANIC 3 /* Panic */
1194#define EXT4_ERRORS_DEFAULT EXT4_ERRORS_CONTINUE
1195
1196/* Metadata checksum algorithm codes */
1197#define EXT4_CRC32C_CHKSUM 1
1198
1199/*
1200 * Structure of the super block
1201 */
1202struct ext4_super_block {
1203/*00*/ __le32 s_inodes_count; /* Inodes count */
1204 __le32 s_blocks_count_lo; /* Blocks count */
1205 __le32 s_r_blocks_count_lo; /* Reserved blocks count */
1206 __le32 s_free_blocks_count_lo; /* Free blocks count */
1207/*10*/ __le32 s_free_inodes_count; /* Free inodes count */
1208 __le32 s_first_data_block; /* First Data Block */
1209 __le32 s_log_block_size; /* Block size */
1210 __le32 s_log_cluster_size; /* Allocation cluster size */
1211/*20*/ __le32 s_blocks_per_group; /* # Blocks per group */
1212 __le32 s_clusters_per_group; /* # Clusters per group */
1213 __le32 s_inodes_per_group; /* # Inodes per group */
1214 __le32 s_mtime; /* Mount time */
1215/*30*/ __le32 s_wtime; /* Write time */
1216 __le16 s_mnt_count; /* Mount count */
1217 __le16 s_max_mnt_count; /* Maximal mount count */
1218 __le16 s_magic; /* Magic signature */
1219 __le16 s_state; /* File system state */
1220 __le16 s_errors; /* Behaviour when detecting errors */
1221 __le16 s_minor_rev_level; /* minor revision level */
1222/*40*/ __le32 s_lastcheck; /* time of last check */
1223 __le32 s_checkinterval; /* max. time between checks */
1224 __le32 s_creator_os; /* OS */
1225 __le32 s_rev_level; /* Revision level */
1226/*50*/ __le16 s_def_resuid; /* Default uid for reserved blocks */
1227 __le16 s_def_resgid; /* Default gid for reserved blocks */
1228 /*
1229 * These fields are for EXT4_DYNAMIC_REV superblocks only.
1230 *
1231 * Note: the difference between the compatible feature set and
1232 * the incompatible feature set is that if there is a bit set
1233 * in the incompatible feature set that the kernel doesn't
1234 * know about, it should refuse to mount the filesystem.
1235 *
1236 * e2fsck's requirements are more strict; if it doesn't know
1237 * about a feature in either the compatible or incompatible
1238 * feature set, it must abort and not try to meddle with
1239 * things it doesn't understand...
1240 */
1241 __le32 s_first_ino; /* First non-reserved inode */
1242 __le16 s_inode_size; /* size of inode structure */
1243 __le16 s_block_group_nr; /* block group # of this superblock */
1244 __le32 s_feature_compat; /* compatible feature set */
1245/*60*/ __le32 s_feature_incompat; /* incompatible feature set */
1246 __le32 s_feature_ro_compat; /* readonly-compatible feature set */
1247/*68*/ __u8 s_uuid[16]; /* 128-bit uuid for volume */
1248/*78*/ char s_volume_name[16]; /* volume name */
1249/*88*/ char s_last_mounted[64]; /* directory where last mounted */
1250/*C8*/ __le32 s_algorithm_usage_bitmap; /* For compression */
1251 /*
1252 * Performance hints. Directory preallocation should only
1253 * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1254 */
1255 __u8 s_prealloc_blocks; /* Nr of blocks to try to preallocate*/
1256 __u8 s_prealloc_dir_blocks; /* Nr to preallocate for dirs */
1257 __le16 s_reserved_gdt_blocks; /* Per group desc for online growth */
1258 /*
1259 * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1260 */
1261/*D0*/ __u8 s_journal_uuid[16]; /* uuid of journal superblock */
1262/*E0*/ __le32 s_journal_inum; /* inode number of journal file */
1263 __le32 s_journal_dev; /* device number of journal file */
1264 __le32 s_last_orphan; /* start of list of inodes to delete */
1265 __le32 s_hash_seed[4]; /* HTREE hash seed */
1266 __u8 s_def_hash_version; /* Default hash version to use */
1267 __u8 s_jnl_backup_type;
1268 __le16 s_desc_size; /* size of group descriptor */
1269/*100*/ __le32 s_default_mount_opts;
1270 __le32 s_first_meta_bg; /* First metablock block group */
1271 __le32 s_mkfs_time; /* When the filesystem was created */
1272 __le32 s_jnl_blocks[17]; /* Backup of the journal inode */
1273 /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1274/*150*/ __le32 s_blocks_count_hi; /* Blocks count */
1275 __le32 s_r_blocks_count_hi; /* Reserved blocks count */
1276 __le32 s_free_blocks_count_hi; /* Free blocks count */
1277 __le16 s_min_extra_isize; /* All inodes have at least # bytes */
1278 __le16 s_want_extra_isize; /* New inodes should reserve # bytes */
1279 __le32 s_flags; /* Miscellaneous flags */
1280 __le16 s_raid_stride; /* RAID stride */
1281 __le16 s_mmp_update_interval; /* # seconds to wait in MMP checking */
1282 __le64 s_mmp_block; /* Block for multi-mount protection */
1283 __le32 s_raid_stripe_width; /* blocks on all data disks (N*stride)*/
1284 __u8 s_log_groups_per_flex; /* FLEX_BG group size */
1285 __u8 s_checksum_type; /* metadata checksum algorithm used */
1286 __u8 s_encryption_level; /* versioning level for encryption */
1287 __u8 s_reserved_pad; /* Padding to next 32bits */
1288 __le64 s_kbytes_written; /* nr of lifetime kilobytes written */
1289 __le32 s_snapshot_inum; /* Inode number of active snapshot */
1290 __le32 s_snapshot_id; /* sequential ID of active snapshot */
1291 __le64 s_snapshot_r_blocks_count; /* reserved blocks for active
1292 snapshot's future use */
1293 __le32 s_snapshot_list; /* inode number of the head of the
1294 on-disk snapshot list */
1295#define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1296 __le32 s_error_count; /* number of fs errors */
1297 __le32 s_first_error_time; /* first time an error happened */
1298 __le32 s_first_error_ino; /* inode involved in first error */
1299 __le64 s_first_error_block; /* block involved of first error */
1300 __u8 s_first_error_func[32]; /* function where the error happened */
1301 __le32 s_first_error_line; /* line number where error happened */
1302 __le32 s_last_error_time; /* most recent time of an error */
1303 __le32 s_last_error_ino; /* inode involved in last error */
1304 __le32 s_last_error_line; /* line number where error happened */
1305 __le64 s_last_error_block; /* block involved of last error */
1306 __u8 s_last_error_func[32]; /* function where the error happened */
1307#define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1308 __u8 s_mount_opts[64];
1309 __le32 s_usr_quota_inum; /* inode for tracking user quota */
1310 __le32 s_grp_quota_inum; /* inode for tracking group quota */
1311 __le32 s_overhead_clusters; /* overhead blocks/clusters in fs */
1312 __le32 s_backup_bgs[2]; /* groups with sparse_super2 SBs */
1313 __u8 s_encrypt_algos[4]; /* Encryption algorithms in use */
1314 __u8 s_encrypt_pw_salt[16]; /* Salt used for string2key algorithm */
1315 __le32 s_lpf_ino; /* Location of the lost+found inode */
1316 __le32 s_prj_quota_inum; /* inode for tracking project quota */
1317 __le32 s_checksum_seed; /* crc32c(uuid) if csum_seed set */
1318 __le32 s_reserved[98]; /* Padding to the end of the block */
1319 __le32 s_checksum; /* crc32c(superblock) */
1320};
1321
1322#define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1323
1324#ifdef __KERNEL__
1325
1326/*
1327 * run-time mount flags
1328 */
1329#define EXT4_MF_MNTDIR_SAMPLED 0x0001
1330#define EXT4_MF_FS_ABORTED 0x0002 /* Fatal error detected */
1331#define EXT4_MF_TEST_DUMMY_ENCRYPTION 0x0004
1332
1333#ifdef CONFIG_EXT4_FS_ENCRYPTION
1334#define DUMMY_ENCRYPTION_ENABLED(sbi) (unlikely((sbi)->s_mount_flags & \
1335 EXT4_MF_TEST_DUMMY_ENCRYPTION))
1336#else
1337#define DUMMY_ENCRYPTION_ENABLED(sbi) (0)
1338#endif
1339
1340/* Number of quota types we support */
1341#define EXT4_MAXQUOTAS 3
1342
1343/*
1344 * fourth extended-fs super-block data in memory
1345 */
1346struct ext4_sb_info {
1347 unsigned long s_desc_size; /* Size of a group descriptor in bytes */
1348 unsigned long s_inodes_per_block;/* Number of inodes per block */
1349 unsigned long s_blocks_per_group;/* Number of blocks in a group */
1350 unsigned long s_clusters_per_group; /* Number of clusters in a group */
1351 unsigned long s_inodes_per_group;/* Number of inodes in a group */
1352 unsigned long s_itb_per_group; /* Number of inode table blocks per group */
1353 unsigned long s_gdb_count; /* Number of group descriptor blocks */
1354 unsigned long s_desc_per_block; /* Number of group descriptors per block */
1355 ext4_group_t s_groups_count; /* Number of groups in the fs */
1356 ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1357 unsigned long s_overhead; /* # of fs overhead clusters */
1358 unsigned int s_cluster_ratio; /* Number of blocks per cluster */
1359 unsigned int s_cluster_bits; /* log2 of s_cluster_ratio */
1360 loff_t s_bitmap_maxbytes; /* max bytes for bitmap files */
1361 struct buffer_head * s_sbh; /* Buffer containing the super block */
1362 struct ext4_super_block *s_es; /* Pointer to the super block in the buffer */
1363 struct buffer_head **s_group_desc;
1364 unsigned int s_mount_opt;
1365 unsigned int s_mount_opt2;
1366 unsigned int s_mount_flags;
1367 unsigned int s_def_mount_opt;
1368 ext4_fsblk_t s_sb_block;
1369 atomic64_t s_resv_clusters;
1370 kuid_t s_resuid;
1371 kgid_t s_resgid;
1372 unsigned short s_mount_state;
1373 unsigned short s_pad;
1374 int s_addr_per_block_bits;
1375 int s_desc_per_block_bits;
1376 int s_inode_size;
1377 int s_first_ino;
1378 unsigned int s_inode_readahead_blks;
1379 unsigned int s_inode_goal;
1380 spinlock_t s_next_gen_lock;
1381 u32 s_next_generation;
1382 u32 s_hash_seed[4];
1383 int s_def_hash_version;
1384 int s_hash_unsigned; /* 3 if hash should be signed, 0 if not */
1385 struct percpu_counter s_freeclusters_counter;
1386 struct percpu_counter s_freeinodes_counter;
1387 struct percpu_counter s_dirs_counter;
1388 struct percpu_counter s_dirtyclusters_counter;
1389 struct blockgroup_lock *s_blockgroup_lock;
1390 struct proc_dir_entry *s_proc;
1391 struct kobject s_kobj;
1392 struct completion s_kobj_unregister;
1393 struct super_block *s_sb;
1394
1395 /* Journaling */
1396 struct journal_s *s_journal;
1397 struct list_head s_orphan;
1398 struct mutex s_orphan_lock;
1399 unsigned long s_resize_flags; /* Flags indicating if there
1400 is a resizer */
1401 unsigned long s_commit_interval;
1402 u32 s_max_batch_time;
1403 u32 s_min_batch_time;
1404 struct block_device *journal_bdev;
1405#ifdef CONFIG_QUOTA
1406 char *s_qf_names[EXT4_MAXQUOTAS]; /* Names of quota files with journalled quota */
1407 int s_jquota_fmt; /* Format of quota to use */
1408#endif
1409 unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1410 struct rb_root system_blks;
1411
1412#ifdef EXTENTS_STATS
1413 /* ext4 extents stats */
1414 unsigned long s_ext_min;
1415 unsigned long s_ext_max;
1416 unsigned long s_depth_max;
1417 spinlock_t s_ext_stats_lock;
1418 unsigned long s_ext_blocks;
1419 unsigned long s_ext_extents;
1420#endif
1421
1422 /* for buddy allocator */
1423 struct ext4_group_info ***s_group_info;
1424 struct inode *s_buddy_cache;
1425 spinlock_t s_md_lock;
1426 unsigned short *s_mb_offsets;
1427 unsigned int *s_mb_maxs;
1428 unsigned int s_group_info_size;
1429
1430 /* tunables */
1431 unsigned long s_stripe;
1432 unsigned int s_mb_stream_request;
1433 unsigned int s_mb_max_to_scan;
1434 unsigned int s_mb_min_to_scan;
1435 unsigned int s_mb_stats;
1436 unsigned int s_mb_order2_reqs;
1437 unsigned int s_mb_group_prealloc;
1438 unsigned int s_max_dir_size_kb;
1439 /* where last allocation was done - for stream allocation */
1440 unsigned long s_mb_last_group;
1441 unsigned long s_mb_last_start;
1442
1443 /* stats for buddy allocator */
1444 atomic_t s_bal_reqs; /* number of reqs with len > 1 */
1445 atomic_t s_bal_success; /* we found long enough chunks */
1446 atomic_t s_bal_allocated; /* in blocks */
1447 atomic_t s_bal_ex_scanned; /* total extents scanned */
1448 atomic_t s_bal_goals; /* goal hits */
1449 atomic_t s_bal_breaks; /* too long searches */
1450 atomic_t s_bal_2orders; /* 2^order hits */
1451 spinlock_t s_bal_lock;
1452 unsigned long s_mb_buddies_generated;
1453 unsigned long long s_mb_generation_time;
1454 atomic_t s_mb_lost_chunks;
1455 atomic_t s_mb_preallocated;
1456 atomic_t s_mb_discarded;
1457 atomic_t s_lock_busy;
1458
1459 /* locality groups */
1460 struct ext4_locality_group __percpu *s_locality_groups;
1461
1462 /* for write statistics */
1463 unsigned long s_sectors_written_start;
1464 u64 s_kbytes_written;
1465
1466 /* the size of zero-out chunk */
1467 unsigned int s_extent_max_zeroout_kb;
1468
1469 unsigned int s_log_groups_per_flex;
1470 struct flex_groups *s_flex_groups;
1471 ext4_group_t s_flex_groups_allocated;
1472
1473 /* workqueue for reserved extent conversions (buffered io) */
1474 struct workqueue_struct *rsv_conversion_wq;
1475
1476 /* timer for periodic error stats printing */
1477 struct timer_list s_err_report;
1478
1479 /* Lazy inode table initialization info */
1480 struct ext4_li_request *s_li_request;
1481 /* Wait multiplier for lazy initialization thread */
1482 unsigned int s_li_wait_mult;
1483
1484 /* Kernel thread for multiple mount protection */
1485 struct task_struct *s_mmp_tsk;
1486
1487 /* record the last minlen when FITRIM is called. */
1488 atomic_t s_last_trim_minblks;
1489
1490 /* Reference to checksum algorithm driver via cryptoapi */
1491 struct crypto_shash *s_chksum_driver;
1492
1493 /* Precomputed FS UUID checksum for seeding other checksums */
1494 __u32 s_csum_seed;
1495
1496 /* Reclaim extents from extent status tree */
1497 struct shrinker s_es_shrinker;
1498 struct list_head s_es_list; /* List of inodes with reclaimable extents */
1499 long s_es_nr_inode;
1500 struct ext4_es_stats s_es_stats;
1501 struct mb_cache *s_mb_cache;
1502 spinlock_t s_es_lock ____cacheline_aligned_in_smp;
1503
1504 /* Ratelimit ext4 messages. */
1505 struct ratelimit_state s_err_ratelimit_state;
1506 struct ratelimit_state s_warning_ratelimit_state;
1507 struct ratelimit_state s_msg_ratelimit_state;
1508};
1509
1510static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1511{
1512 return sb->s_fs_info;
1513}
1514static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1515{
1516 return container_of(inode, struct ext4_inode_info, vfs_inode);
1517}
1518
1519static inline struct timespec ext4_current_time(struct inode *inode)
1520{
1521 return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1522 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1523}
1524
1525static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1526{
1527 return ino == EXT4_ROOT_INO ||
1528 ino == EXT4_USR_QUOTA_INO ||
1529 ino == EXT4_GRP_QUOTA_INO ||
1530 ino == EXT4_BOOT_LOADER_INO ||
1531 ino == EXT4_JOURNAL_INO ||
1532 ino == EXT4_RESIZE_INO ||
1533 (ino >= EXT4_FIRST_INO(sb) &&
1534 ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1535}
1536
1537/*
1538 * Inode dynamic state flags
1539 */
1540enum {
1541 EXT4_STATE_JDATA, /* journaled data exists */
1542 EXT4_STATE_NEW, /* inode is newly created */
1543 EXT4_STATE_XATTR, /* has in-inode xattrs */
1544 EXT4_STATE_NO_EXPAND, /* No space for expansion */
1545 EXT4_STATE_DA_ALLOC_CLOSE, /* Alloc DA blks on close */
1546 EXT4_STATE_EXT_MIGRATE, /* Inode is migrating */
1547 EXT4_STATE_DIO_UNWRITTEN, /* need convert on dio done*/
1548 EXT4_STATE_NEWENTRY, /* File just added to dir */
1549 EXT4_STATE_DIOREAD_LOCK, /* Disable support for dio read
1550 nolocking */
1551 EXT4_STATE_MAY_INLINE_DATA, /* may have in-inode data */
1552 EXT4_STATE_ORDERED_MODE, /* data=ordered mode */
1553 EXT4_STATE_EXT_PRECACHED, /* extents have been precached */
1554};
1555
1556#define EXT4_INODE_BIT_FNS(name, field, offset) \
1557static inline int ext4_test_inode_##name(struct inode *inode, int bit) \
1558{ \
1559 return test_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1560} \
1561static inline void ext4_set_inode_##name(struct inode *inode, int bit) \
1562{ \
1563 set_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1564} \
1565static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1566{ \
1567 clear_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1568}
1569
1570/* Add these declarations here only so that these functions can be
1571 * found by name. Otherwise, they are very hard to locate. */
1572static inline int ext4_test_inode_flag(struct inode *inode, int bit);
1573static inline void ext4_set_inode_flag(struct inode *inode, int bit);
1574static inline void ext4_clear_inode_flag(struct inode *inode, int bit);
1575EXT4_INODE_BIT_FNS(flag, flags, 0)
1576
1577/* Add these declarations here only so that these functions can be
1578 * found by name. Otherwise, they are very hard to locate. */
1579static inline int ext4_test_inode_state(struct inode *inode, int bit);
1580static inline void ext4_set_inode_state(struct inode *inode, int bit);
1581static inline void ext4_clear_inode_state(struct inode *inode, int bit);
1582#if (BITS_PER_LONG < 64)
1583EXT4_INODE_BIT_FNS(state, state_flags, 0)
1584
1585static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1586{
1587 (ei)->i_state_flags = 0;
1588}
1589#else
1590EXT4_INODE_BIT_FNS(state, flags, 32)
1591
1592static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1593{
1594 /* We depend on the fact that callers will set i_flags */
1595}
1596#endif
1597#else
1598/* Assume that user mode programs are passing in an ext4fs superblock, not
1599 * a kernel struct super_block. This will allow us to call the feature-test
1600 * macros from user land. */
1601#define EXT4_SB(sb) (sb)
1602#endif
1603
1604/*
1605 * Returns true if the inode is inode is encrypted
1606 */
1607static inline int ext4_encrypted_inode(struct inode *inode)
1608{
1609#ifdef CONFIG_EXT4_FS_ENCRYPTION
1610 return ext4_test_inode_flag(inode, EXT4_INODE_ENCRYPT);
1611#else
1612 return 0;
1613#endif
1614}
1615
1616#define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1617
1618/*
1619 * Codes for operating systems
1620 */
1621#define EXT4_OS_LINUX 0
1622#define EXT4_OS_HURD 1
1623#define EXT4_OS_MASIX 2
1624#define EXT4_OS_FREEBSD 3
1625#define EXT4_OS_LITES 4
1626
1627/*
1628 * Revision levels
1629 */
1630#define EXT4_GOOD_OLD_REV 0 /* The good old (original) format */
1631#define EXT4_DYNAMIC_REV 1 /* V2 format w/ dynamic inode sizes */
1632
1633#define EXT4_CURRENT_REV EXT4_GOOD_OLD_REV
1634#define EXT4_MAX_SUPP_REV EXT4_DYNAMIC_REV
1635
1636#define EXT4_GOOD_OLD_INODE_SIZE 128
1637
1638/*
1639 * Feature set definitions
1640 */
1641
1642/* Use the ext4_{has,set,clear}_feature_* helpers; these will be removed */
1643#define EXT4_HAS_COMPAT_FEATURE(sb,mask) \
1644 ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1645#define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask) \
1646 ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1647#define EXT4_HAS_INCOMPAT_FEATURE(sb,mask) \
1648 ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1649#define EXT4_SET_COMPAT_FEATURE(sb,mask) \
1650 EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1651#define EXT4_SET_RO_COMPAT_FEATURE(sb,mask) \
1652 EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1653#define EXT4_SET_INCOMPAT_FEATURE(sb,mask) \
1654 EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1655#define EXT4_CLEAR_COMPAT_FEATURE(sb,mask) \
1656 EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1657#define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask) \
1658 EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1659#define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask) \
1660 EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1661
1662#define EXT4_FEATURE_COMPAT_DIR_PREALLOC 0x0001
1663#define EXT4_FEATURE_COMPAT_IMAGIC_INODES 0x0002
1664#define EXT4_FEATURE_COMPAT_HAS_JOURNAL 0x0004
1665#define EXT4_FEATURE_COMPAT_EXT_ATTR 0x0008
1666#define EXT4_FEATURE_COMPAT_RESIZE_INODE 0x0010
1667#define EXT4_FEATURE_COMPAT_DIR_INDEX 0x0020
1668#define EXT4_FEATURE_COMPAT_SPARSE_SUPER2 0x0200
1669
1670#define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER 0x0001
1671#define EXT4_FEATURE_RO_COMPAT_LARGE_FILE 0x0002
1672#define EXT4_FEATURE_RO_COMPAT_BTREE_DIR 0x0004
1673#define EXT4_FEATURE_RO_COMPAT_HUGE_FILE 0x0008
1674#define EXT4_FEATURE_RO_COMPAT_GDT_CSUM 0x0010
1675#define EXT4_FEATURE_RO_COMPAT_DIR_NLINK 0x0020
1676#define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE 0x0040
1677#define EXT4_FEATURE_RO_COMPAT_QUOTA 0x0100
1678#define EXT4_FEATURE_RO_COMPAT_BIGALLOC 0x0200
1679/*
1680 * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM). When
1681 * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
1682 * all other data structures' checksums. However, the METADATA_CSUM and
1683 * GDT_CSUM bits are mutually exclusive.
1684 */
1685#define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM 0x0400
1686#define EXT4_FEATURE_RO_COMPAT_READONLY 0x1000
1687#define EXT4_FEATURE_RO_COMPAT_PROJECT 0x2000
1688
1689#define EXT4_FEATURE_INCOMPAT_COMPRESSION 0x0001
1690#define EXT4_FEATURE_INCOMPAT_FILETYPE 0x0002
1691#define EXT4_FEATURE_INCOMPAT_RECOVER 0x0004 /* Needs recovery */
1692#define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 /* Journal device */
1693#define EXT4_FEATURE_INCOMPAT_META_BG 0x0010
1694#define EXT4_FEATURE_INCOMPAT_EXTENTS 0x0040 /* extents support */
1695#define EXT4_FEATURE_INCOMPAT_64BIT 0x0080
1696#define EXT4_FEATURE_INCOMPAT_MMP 0x0100
1697#define EXT4_FEATURE_INCOMPAT_FLEX_BG 0x0200
1698#define EXT4_FEATURE_INCOMPAT_EA_INODE 0x0400 /* EA in inode */
1699#define EXT4_FEATURE_INCOMPAT_DIRDATA 0x1000 /* data in dirent */
1700#define EXT4_FEATURE_INCOMPAT_CSUM_SEED 0x2000
1701#define EXT4_FEATURE_INCOMPAT_LARGEDIR 0x4000 /* >2GB or 3-lvl htree */
1702#define EXT4_FEATURE_INCOMPAT_INLINE_DATA 0x8000 /* data in inode */
1703#define EXT4_FEATURE_INCOMPAT_ENCRYPT 0x10000
1704
1705#define EXT4_FEATURE_COMPAT_FUNCS(name, flagname) \
1706static inline bool ext4_has_feature_##name(struct super_block *sb) \
1707{ \
1708 return ((EXT4_SB(sb)->s_es->s_feature_compat & \
1709 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname)) != 0); \
1710} \
1711static inline void ext4_set_feature_##name(struct super_block *sb) \
1712{ \
1713 EXT4_SB(sb)->s_es->s_feature_compat |= \
1714 cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
1715} \
1716static inline void ext4_clear_feature_##name(struct super_block *sb) \
1717{ \
1718 EXT4_SB(sb)->s_es->s_feature_compat &= \
1719 ~cpu_to_le32(EXT4_FEATURE_COMPAT_##flagname); \
1720}
1721
1722#define EXT4_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
1723static inline bool ext4_has_feature_##name(struct super_block *sb) \
1724{ \
1725 return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
1726 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname)) != 0); \
1727} \
1728static inline void ext4_set_feature_##name(struct super_block *sb) \
1729{ \
1730 EXT4_SB(sb)->s_es->s_feature_ro_compat |= \
1731 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
1732} \
1733static inline void ext4_clear_feature_##name(struct super_block *sb) \
1734{ \
1735 EXT4_SB(sb)->s_es->s_feature_ro_compat &= \
1736 ~cpu_to_le32(EXT4_FEATURE_RO_COMPAT_##flagname); \
1737}
1738
1739#define EXT4_FEATURE_INCOMPAT_FUNCS(name, flagname) \
1740static inline bool ext4_has_feature_##name(struct super_block *sb) \
1741{ \
1742 return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
1743 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname)) != 0); \
1744} \
1745static inline void ext4_set_feature_##name(struct super_block *sb) \
1746{ \
1747 EXT4_SB(sb)->s_es->s_feature_incompat |= \
1748 cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
1749} \
1750static inline void ext4_clear_feature_##name(struct super_block *sb) \
1751{ \
1752 EXT4_SB(sb)->s_es->s_feature_incompat &= \
1753 ~cpu_to_le32(EXT4_FEATURE_INCOMPAT_##flagname); \
1754}
1755
1756EXT4_FEATURE_COMPAT_FUNCS(dir_prealloc, DIR_PREALLOC)
1757EXT4_FEATURE_COMPAT_FUNCS(imagic_inodes, IMAGIC_INODES)
1758EXT4_FEATURE_COMPAT_FUNCS(journal, HAS_JOURNAL)
1759EXT4_FEATURE_COMPAT_FUNCS(xattr, EXT_ATTR)
1760EXT4_FEATURE_COMPAT_FUNCS(resize_inode, RESIZE_INODE)
1761EXT4_FEATURE_COMPAT_FUNCS(dir_index, DIR_INDEX)
1762EXT4_FEATURE_COMPAT_FUNCS(sparse_super2, SPARSE_SUPER2)
1763
1764EXT4_FEATURE_RO_COMPAT_FUNCS(sparse_super, SPARSE_SUPER)
1765EXT4_FEATURE_RO_COMPAT_FUNCS(large_file, LARGE_FILE)
1766EXT4_FEATURE_RO_COMPAT_FUNCS(btree_dir, BTREE_DIR)
1767EXT4_FEATURE_RO_COMPAT_FUNCS(huge_file, HUGE_FILE)
1768EXT4_FEATURE_RO_COMPAT_FUNCS(gdt_csum, GDT_CSUM)
1769EXT4_FEATURE_RO_COMPAT_FUNCS(dir_nlink, DIR_NLINK)
1770EXT4_FEATURE_RO_COMPAT_FUNCS(extra_isize, EXTRA_ISIZE)
1771EXT4_FEATURE_RO_COMPAT_FUNCS(quota, QUOTA)
1772EXT4_FEATURE_RO_COMPAT_FUNCS(bigalloc, BIGALLOC)
1773EXT4_FEATURE_RO_COMPAT_FUNCS(metadata_csum, METADATA_CSUM)
1774EXT4_FEATURE_RO_COMPAT_FUNCS(readonly, READONLY)
1775EXT4_FEATURE_RO_COMPAT_FUNCS(project, PROJECT)
1776
1777EXT4_FEATURE_INCOMPAT_FUNCS(compression, COMPRESSION)
1778EXT4_FEATURE_INCOMPAT_FUNCS(filetype, FILETYPE)
1779EXT4_FEATURE_INCOMPAT_FUNCS(journal_needs_recovery, RECOVER)
1780EXT4_FEATURE_INCOMPAT_FUNCS(journal_dev, JOURNAL_DEV)
1781EXT4_FEATURE_INCOMPAT_FUNCS(meta_bg, META_BG)
1782EXT4_FEATURE_INCOMPAT_FUNCS(extents, EXTENTS)
1783EXT4_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT)
1784EXT4_FEATURE_INCOMPAT_FUNCS(mmp, MMP)
1785EXT4_FEATURE_INCOMPAT_FUNCS(flex_bg, FLEX_BG)
1786EXT4_FEATURE_INCOMPAT_FUNCS(ea_inode, EA_INODE)
1787EXT4_FEATURE_INCOMPAT_FUNCS(dirdata, DIRDATA)
1788EXT4_FEATURE_INCOMPAT_FUNCS(csum_seed, CSUM_SEED)
1789EXT4_FEATURE_INCOMPAT_FUNCS(largedir, LARGEDIR)
1790EXT4_FEATURE_INCOMPAT_FUNCS(inline_data, INLINE_DATA)
1791EXT4_FEATURE_INCOMPAT_FUNCS(encrypt, ENCRYPT)
1792
1793#define EXT2_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR
1794#define EXT2_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1795 EXT4_FEATURE_INCOMPAT_META_BG)
1796#define EXT2_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1797 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1798 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1799
1800#define EXT3_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR
1801#define EXT3_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1802 EXT4_FEATURE_INCOMPAT_RECOVER| \
1803 EXT4_FEATURE_INCOMPAT_META_BG)
1804#define EXT3_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1805 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1806 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1807
1808#define EXT4_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR
1809#define EXT4_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1810 EXT4_FEATURE_INCOMPAT_RECOVER| \
1811 EXT4_FEATURE_INCOMPAT_META_BG| \
1812 EXT4_FEATURE_INCOMPAT_EXTENTS| \
1813 EXT4_FEATURE_INCOMPAT_64BIT| \
1814 EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1815 EXT4_FEATURE_INCOMPAT_MMP | \
1816 EXT4_FEATURE_INCOMPAT_INLINE_DATA | \
1817 EXT4_FEATURE_INCOMPAT_ENCRYPT | \
1818 EXT4_FEATURE_INCOMPAT_CSUM_SEED)
1819#define EXT4_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1820 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1821 EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1822 EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1823 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1824 EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1825 EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1826 EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
1827 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
1828 EXT4_FEATURE_RO_COMPAT_QUOTA |\
1829 EXT4_FEATURE_RO_COMPAT_PROJECT)
1830
1831#define EXTN_FEATURE_FUNCS(ver) \
1832static inline bool ext4_has_unknown_ext##ver##_compat_features(struct super_block *sb) \
1833{ \
1834 return ((EXT4_SB(sb)->s_es->s_feature_compat & \
1835 cpu_to_le32(~EXT##ver##_FEATURE_COMPAT_SUPP)) != 0); \
1836} \
1837static inline bool ext4_has_unknown_ext##ver##_ro_compat_features(struct super_block *sb) \
1838{ \
1839 return ((EXT4_SB(sb)->s_es->s_feature_ro_compat & \
1840 cpu_to_le32(~EXT##ver##_FEATURE_RO_COMPAT_SUPP)) != 0); \
1841} \
1842static inline bool ext4_has_unknown_ext##ver##_incompat_features(struct super_block *sb) \
1843{ \
1844 return ((EXT4_SB(sb)->s_es->s_feature_incompat & \
1845 cpu_to_le32(~EXT##ver##_FEATURE_INCOMPAT_SUPP)) != 0); \
1846}
1847
1848EXTN_FEATURE_FUNCS(2)
1849EXTN_FEATURE_FUNCS(3)
1850EXTN_FEATURE_FUNCS(4)
1851
1852static inline bool ext4_has_compat_features(struct super_block *sb)
1853{
1854 return (EXT4_SB(sb)->s_es->s_feature_compat != 0);
1855}
1856static inline bool ext4_has_ro_compat_features(struct super_block *sb)
1857{
1858 return (EXT4_SB(sb)->s_es->s_feature_ro_compat != 0);
1859}
1860static inline bool ext4_has_incompat_features(struct super_block *sb)
1861{
1862 return (EXT4_SB(sb)->s_es->s_feature_incompat != 0);
1863}
1864
1865/*
1866 * Default values for user and/or group using reserved blocks
1867 */
1868#define EXT4_DEF_RESUID 0
1869#define EXT4_DEF_RESGID 0
1870
1871/*
1872 * Default project ID
1873 */
1874#define EXT4_DEF_PROJID 0
1875
1876#define EXT4_DEF_INODE_READAHEAD_BLKS 32
1877
1878/*
1879 * Default mount options
1880 */
1881#define EXT4_DEFM_DEBUG 0x0001
1882#define EXT4_DEFM_BSDGROUPS 0x0002
1883#define EXT4_DEFM_XATTR_USER 0x0004
1884#define EXT4_DEFM_ACL 0x0008
1885#define EXT4_DEFM_UID16 0x0010
1886#define EXT4_DEFM_JMODE 0x0060
1887#define EXT4_DEFM_JMODE_DATA 0x0020
1888#define EXT4_DEFM_JMODE_ORDERED 0x0040
1889#define EXT4_DEFM_JMODE_WBACK 0x0060
1890#define EXT4_DEFM_NOBARRIER 0x0100
1891#define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1892#define EXT4_DEFM_DISCARD 0x0400
1893#define EXT4_DEFM_NODELALLOC 0x0800
1894
1895/*
1896 * Default journal batch times
1897 */
1898#define EXT4_DEF_MIN_BATCH_TIME 0
1899#define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1900
1901/*
1902 * Minimum number of groups in a flexgroup before we separate out
1903 * directories into the first block group of a flexgroup
1904 */
1905#define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1906
1907/*
1908 * Structure of a directory entry
1909 */
1910#define EXT4_NAME_LEN 255
1911
1912struct ext4_dir_entry {
1913 __le32 inode; /* Inode number */
1914 __le16 rec_len; /* Directory entry length */
1915 __le16 name_len; /* Name length */
1916 char name[EXT4_NAME_LEN]; /* File name */
1917};
1918
1919/*
1920 * The new version of the directory entry. Since EXT4 structures are
1921 * stored in intel byte order, and the name_len field could never be
1922 * bigger than 255 chars, it's safe to reclaim the extra byte for the
1923 * file_type field.
1924 */
1925struct ext4_dir_entry_2 {
1926 __le32 inode; /* Inode number */
1927 __le16 rec_len; /* Directory entry length */
1928 __u8 name_len; /* Name length */
1929 __u8 file_type;
1930 char name[EXT4_NAME_LEN]; /* File name */
1931};
1932
1933/*
1934 * This is a bogus directory entry at the end of each leaf block that
1935 * records checksums.
1936 */
1937struct ext4_dir_entry_tail {
1938 __le32 det_reserved_zero1; /* Pretend to be unused */
1939 __le16 det_rec_len; /* 12 */
1940 __u8 det_reserved_zero2; /* Zero name length */
1941 __u8 det_reserved_ft; /* 0xDE, fake file type */
1942 __le32 det_checksum; /* crc32c(uuid+inum+dirblock) */
1943};
1944
1945#define EXT4_DIRENT_TAIL(block, blocksize) \
1946 ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
1947 ((blocksize) - \
1948 sizeof(struct ext4_dir_entry_tail))))
1949
1950/*
1951 * Ext4 directory file types. Only the low 3 bits are used. The
1952 * other bits are reserved for now.
1953 */
1954#define EXT4_FT_UNKNOWN 0
1955#define EXT4_FT_REG_FILE 1
1956#define EXT4_FT_DIR 2
1957#define EXT4_FT_CHRDEV 3
1958#define EXT4_FT_BLKDEV 4
1959#define EXT4_FT_FIFO 5
1960#define EXT4_FT_SOCK 6
1961#define EXT4_FT_SYMLINK 7
1962
1963#define EXT4_FT_MAX 8
1964
1965#define EXT4_FT_DIR_CSUM 0xDE
1966
1967/*
1968 * EXT4_DIR_PAD defines the directory entries boundaries
1969 *
1970 * NOTE: It must be a multiple of 4
1971 */
1972#define EXT4_DIR_PAD 4
1973#define EXT4_DIR_ROUND (EXT4_DIR_PAD - 1)
1974#define EXT4_DIR_REC_LEN(name_len) (((name_len) + 8 + EXT4_DIR_ROUND) & \
1975 ~EXT4_DIR_ROUND)
1976#define EXT4_MAX_REC_LEN ((1<<16)-1)
1977
1978/*
1979 * If we ever get support for fs block sizes > page_size, we'll need
1980 * to remove the #if statements in the next two functions...
1981 */
1982static inline unsigned int
1983ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1984{
1985 unsigned len = le16_to_cpu(dlen);
1986
1987#if (PAGE_SIZE >= 65536)
1988 if (len == EXT4_MAX_REC_LEN || len == 0)
1989 return blocksize;
1990 return (len & 65532) | ((len & 3) << 16);
1991#else
1992 return len;
1993#endif
1994}
1995
1996static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1997{
1998 if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1999 BUG();
2000#if (PAGE_SIZE >= 65536)
2001 if (len < 65536)
2002 return cpu_to_le16(len);
2003 if (len == blocksize) {
2004 if (blocksize == 65536)
2005 return cpu_to_le16(EXT4_MAX_REC_LEN);
2006 else
2007 return cpu_to_le16(0);
2008 }
2009 return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
2010#else
2011 return cpu_to_le16(len);
2012#endif
2013}
2014
2015/*
2016 * Hash Tree Directory indexing
2017 * (c) Daniel Phillips, 2001
2018 */
2019
2020#define is_dx(dir) (ext4_has_feature_dir_index((dir)->i_sb) && \
2021 ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
2022#define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
2023#define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
2024
2025/* Legal values for the dx_root hash_version field: */
2026
2027#define DX_HASH_LEGACY 0
2028#define DX_HASH_HALF_MD4 1
2029#define DX_HASH_TEA 2
2030#define DX_HASH_LEGACY_UNSIGNED 3
2031#define DX_HASH_HALF_MD4_UNSIGNED 4
2032#define DX_HASH_TEA_UNSIGNED 5
2033
2034static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
2035 const void *address, unsigned int length)
2036{
2037 struct {
2038 struct shash_desc shash;
2039 char ctx[4];
2040 } desc;
2041 int err;
2042
2043 BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
2044
2045 desc.shash.tfm = sbi->s_chksum_driver;
2046 desc.shash.flags = 0;
2047 *(u32 *)desc.ctx = crc;
2048
2049 err = crypto_shash_update(&desc.shash, address, length);
2050 BUG_ON(err);
2051
2052 return *(u32 *)desc.ctx;
2053}
2054
2055#ifdef __KERNEL__
2056
2057/* hash info structure used by the directory hash */
2058struct dx_hash_info
2059{
2060 u32 hash;
2061 u32 minor_hash;
2062 int hash_version;
2063 u32 *seed;
2064};
2065
2066
2067/* 32 and 64 bit signed EOF for dx directories */
2068#define EXT4_HTREE_EOF_32BIT ((1UL << (32 - 1)) - 1)
2069#define EXT4_HTREE_EOF_64BIT ((1ULL << (64 - 1)) - 1)
2070
2071
2072/*
2073 * Control parameters used by ext4_htree_next_block
2074 */
2075#define HASH_NB_ALWAYS 1
2076
2077struct ext4_filename {
2078 const struct qstr *usr_fname;
2079 struct ext4_str disk_name;
2080 struct dx_hash_info hinfo;
2081#ifdef CONFIG_EXT4_FS_ENCRYPTION
2082 struct ext4_str crypto_buf;
2083#endif
2084};
2085
2086#define fname_name(p) ((p)->disk_name.name)
2087#define fname_len(p) ((p)->disk_name.len)
2088
2089/*
2090 * Describe an inode's exact location on disk and in memory
2091 */
2092struct ext4_iloc
2093{
2094 struct buffer_head *bh;
2095 unsigned long offset;
2096 ext4_group_t block_group;
2097};
2098
2099static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
2100{
2101 return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
2102}
2103
2104/*
2105 * This structure is stuffed into the struct file's private_data field
2106 * for directories. It is where we put information so that we can do
2107 * readdir operations in hash tree order.
2108 */
2109struct dir_private_info {
2110 struct rb_root root;
2111 struct rb_node *curr_node;
2112 struct fname *extra_fname;
2113 loff_t last_pos;
2114 __u32 curr_hash;
2115 __u32 curr_minor_hash;
2116 __u32 next_hash;
2117};
2118
2119/* calculate the first block number of the group */
2120static inline ext4_fsblk_t
2121ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
2122{
2123 return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
2124 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
2125}
2126
2127/*
2128 * Special error return code only used by dx_probe() and its callers.
2129 */
2130#define ERR_BAD_DX_DIR (-(MAX_ERRNO - 1))
2131
2132/*
2133 * Timeout and state flag for lazy initialization inode thread.
2134 */
2135#define EXT4_DEF_LI_WAIT_MULT 10
2136#define EXT4_DEF_LI_MAX_START_DELAY 5
2137#define EXT4_LAZYINIT_QUIT 0x0001
2138#define EXT4_LAZYINIT_RUNNING 0x0002
2139
2140/*
2141 * Lazy inode table initialization info
2142 */
2143struct ext4_lazy_init {
2144 unsigned long li_state;
2145 struct list_head li_request_list;
2146 struct mutex li_list_mtx;
2147};
2148
2149struct ext4_li_request {
2150 struct super_block *lr_super;
2151 struct ext4_sb_info *lr_sbi;
2152 ext4_group_t lr_next_group;
2153 struct list_head lr_request;
2154 unsigned long lr_next_sched;
2155 unsigned long lr_timeout;
2156};
2157
2158struct ext4_features {
2159 struct kobject f_kobj;
2160 struct completion f_kobj_unregister;
2161};
2162
2163/*
2164 * This structure will be used for multiple mount protection. It will be
2165 * written into the block number saved in the s_mmp_block field in the
2166 * superblock. Programs that check MMP should assume that if
2167 * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
2168 * to use the filesystem, regardless of how old the timestamp is.
2169 */
2170#define EXT4_MMP_MAGIC 0x004D4D50U /* ASCII for MMP */
2171#define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
2172#define EXT4_MMP_SEQ_FSCK 0xE24D4D50U /* mmp_seq value when being fscked */
2173#define EXT4_MMP_SEQ_MAX 0xE24D4D4FU /* maximum valid mmp_seq value */
2174
2175struct mmp_struct {
2176 __le32 mmp_magic; /* Magic number for MMP */
2177 __le32 mmp_seq; /* Sequence no. updated periodically */
2178
2179 /*
2180 * mmp_time, mmp_nodename & mmp_bdevname are only used for information
2181 * purposes and do not affect the correctness of the algorithm
2182 */
2183 __le64 mmp_time; /* Time last updated */
2184 char mmp_nodename[64]; /* Node which last updated MMP block */
2185 char mmp_bdevname[32]; /* Bdev which last updated MMP block */
2186
2187 /*
2188 * mmp_check_interval is used to verify if the MMP block has been
2189 * updated on the block device. The value is updated based on the
2190 * maximum time to write the MMP block during an update cycle.
2191 */
2192 __le16 mmp_check_interval;
2193
2194 __le16 mmp_pad1;
2195 __le32 mmp_pad2[226];
2196 __le32 mmp_checksum; /* crc32c(uuid+mmp_block) */
2197};
2198
2199/* arguments passed to the mmp thread */
2200struct mmpd_data {
2201 struct buffer_head *bh; /* bh from initial read_mmp_block() */
2202 struct super_block *sb; /* super block of the fs */
2203};
2204
2205/*
2206 * Check interval multiplier
2207 * The MMP block is written every update interval and initially checked every
2208 * update interval x the multiplier (the value is then adapted based on the
2209 * write latency). The reason is that writes can be delayed under load and we
2210 * don't want readers to incorrectly assume that the filesystem is no longer
2211 * in use.
2212 */
2213#define EXT4_MMP_CHECK_MULT 2UL
2214
2215/*
2216 * Minimum interval for MMP checking in seconds.
2217 */
2218#define EXT4_MMP_MIN_CHECK_INTERVAL 5UL
2219
2220/*
2221 * Maximum interval for MMP checking in seconds.
2222 */
2223#define EXT4_MMP_MAX_CHECK_INTERVAL 300UL
2224
2225/*
2226 * Function prototypes
2227 */
2228
2229/*
2230 * Ok, these declarations are also in <linux/kernel.h> but none of the
2231 * ext4 source programs needs to include it so they are duplicated here.
2232 */
2233# define NORET_TYPE /**/
2234# define ATTRIB_NORET __attribute__((noreturn))
2235# define NORET_AND noreturn,
2236
2237/* bitmap.c */
2238extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
2239void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
2240 struct ext4_group_desc *gdp,
2241 struct buffer_head *bh, int sz);
2242int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2243 struct ext4_group_desc *gdp,
2244 struct buffer_head *bh, int sz);
2245void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
2246 struct ext4_group_desc *gdp,
2247 struct buffer_head *bh);
2248int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
2249 struct ext4_group_desc *gdp,
2250 struct buffer_head *bh);
2251
2252/* balloc.c */
2253extern void ext4_get_group_no_and_offset(struct super_block *sb,
2254 ext4_fsblk_t blocknr,
2255 ext4_group_t *blockgrpp,
2256 ext4_grpblk_t *offsetp);
2257extern ext4_group_t ext4_get_group_number(struct super_block *sb,
2258 ext4_fsblk_t block);
2259
2260extern unsigned int ext4_block_group(struct super_block *sb,
2261 ext4_fsblk_t blocknr);
2262extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
2263 ext4_fsblk_t blocknr);
2264extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
2265extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
2266 ext4_group_t group);
2267extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
2268 ext4_fsblk_t goal,
2269 unsigned int flags,
2270 unsigned long *count,
2271 int *errp);
2272extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
2273 s64 nclusters, unsigned int flags);
2274extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
2275extern void ext4_check_blocks_bitmap(struct super_block *);
2276extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
2277 ext4_group_t block_group,
2278 struct buffer_head ** bh);
2279extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
2280
2281extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
2282 ext4_group_t block_group);
2283extern int ext4_wait_block_bitmap(struct super_block *sb,
2284 ext4_group_t block_group,
2285 struct buffer_head *bh);
2286extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
2287 ext4_group_t block_group);
2288extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
2289 ext4_group_t block_group,
2290 struct ext4_group_desc *gdp);
2291ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
2292
2293/* crypto_policy.c */
2294int ext4_is_child_context_consistent_with_parent(struct inode *parent,
2295 struct inode *child);
2296int ext4_inherit_context(struct inode *parent, struct inode *child);
2297void ext4_to_hex(char *dst, char *src, size_t src_size);
2298int ext4_process_policy(const struct ext4_encryption_policy *policy,
2299 struct inode *inode);
2300int ext4_get_policy(struct inode *inode,
2301 struct ext4_encryption_policy *policy);
2302
2303/* crypto.c */
2304extern struct kmem_cache *ext4_crypt_info_cachep;
2305bool ext4_valid_contents_enc_mode(uint32_t mode);
2306uint32_t ext4_validate_encryption_key_size(uint32_t mode, uint32_t size);
2307extern struct workqueue_struct *ext4_read_workqueue;
2308struct ext4_crypto_ctx *ext4_get_crypto_ctx(struct inode *inode,
2309 gfp_t gfp_flags);
2310void ext4_release_crypto_ctx(struct ext4_crypto_ctx *ctx);
2311void ext4_restore_control_page(struct page *data_page);
2312struct page *ext4_encrypt(struct inode *inode,
2313 struct page *plaintext_page,
2314 gfp_t gfp_flags);
2315int ext4_decrypt(struct page *page);
2316int ext4_encrypted_zeroout(struct inode *inode, ext4_lblk_t lblk,
2317 ext4_fsblk_t pblk, ext4_lblk_t len);
2318extern const struct dentry_operations ext4_encrypted_d_ops;
2319
2320#ifdef CONFIG_EXT4_FS_ENCRYPTION
2321int ext4_init_crypto(void);
2322void ext4_exit_crypto(void);
2323static inline int ext4_sb_has_crypto(struct super_block *sb)
2324{
2325 return ext4_has_feature_encrypt(sb);
2326}
2327#else
2328static inline int ext4_init_crypto(void) { return 0; }
2329static inline void ext4_exit_crypto(void) { }
2330static inline int ext4_sb_has_crypto(struct super_block *sb)
2331{
2332 return 0;
2333}
2334#endif
2335
2336/* crypto_fname.c */
2337bool ext4_valid_filenames_enc_mode(uint32_t mode);
2338u32 ext4_fname_crypto_round_up(u32 size, u32 blksize);
2339unsigned ext4_fname_encrypted_size(struct inode *inode, u32 ilen);
2340int ext4_fname_crypto_alloc_buffer(struct inode *inode,
2341 u32 ilen, struct ext4_str *crypto_str);
2342int _ext4_fname_disk_to_usr(struct inode *inode,
2343 struct dx_hash_info *hinfo,
2344 const struct ext4_str *iname,
2345 struct ext4_str *oname);
2346int ext4_fname_disk_to_usr(struct inode *inode,
2347 struct dx_hash_info *hinfo,
2348 const struct ext4_dir_entry_2 *de,
2349 struct ext4_str *oname);
2350int ext4_fname_usr_to_disk(struct inode *inode,
2351 const struct qstr *iname,
2352 struct ext4_str *oname);
2353#ifdef CONFIG_EXT4_FS_ENCRYPTION
2354void ext4_fname_crypto_free_buffer(struct ext4_str *crypto_str);
2355int ext4_fname_setup_filename(struct inode *dir, const struct qstr *iname,
2356 int lookup, struct ext4_filename *fname);
2357void ext4_fname_free_filename(struct ext4_filename *fname);
2358#else
2359static inline
2360int ext4_setup_fname_crypto(struct inode *inode)
2361{
2362 return 0;
2363}
2364static inline void ext4_fname_crypto_free_buffer(struct ext4_str *p) { }
2365static inline int ext4_fname_setup_filename(struct inode *dir,
2366 const struct qstr *iname,
2367 int lookup, struct ext4_filename *fname)
2368{
2369 fname->usr_fname = iname;
2370 fname->disk_name.name = (unsigned char *) iname->name;
2371 fname->disk_name.len = iname->len;
2372 return 0;
2373}
2374static inline void ext4_fname_free_filename(struct ext4_filename *fname) { }
2375#endif
2376
2377
2378/* crypto_key.c */
2379void ext4_free_crypt_info(struct ext4_crypt_info *ci);
2380void ext4_free_encryption_info(struct inode *inode, struct ext4_crypt_info *ci);
2381int _ext4_get_encryption_info(struct inode *inode);
2382
2383#ifdef CONFIG_EXT4_FS_ENCRYPTION
2384int ext4_has_encryption_key(struct inode *inode);
2385
2386static inline int ext4_get_encryption_info(struct inode *inode)
2387{
2388 struct ext4_crypt_info *ci = EXT4_I(inode)->i_crypt_info;
2389
2390 if (!ci ||
2391 (ci->ci_keyring_key &&
2392 (ci->ci_keyring_key->flags & ((1 << KEY_FLAG_INVALIDATED) |
2393 (1 << KEY_FLAG_REVOKED) |
2394 (1 << KEY_FLAG_DEAD)))))
2395 return _ext4_get_encryption_info(inode);
2396 return 0;
2397}
2398
2399static inline struct ext4_crypt_info *ext4_encryption_info(struct inode *inode)
2400{
2401 return EXT4_I(inode)->i_crypt_info;
2402}
2403
2404#else
2405static inline int ext4_has_encryption_key(struct inode *inode)
2406{
2407 return 0;
2408}
2409static inline int ext4_get_encryption_info(struct inode *inode)
2410{
2411 return 0;
2412}
2413static inline struct ext4_crypt_info *ext4_encryption_info(struct inode *inode)
2414{
2415 return NULL;
2416}
2417#endif
2418
2419
2420/* dir.c */
2421extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
2422 struct file *,
2423 struct ext4_dir_entry_2 *,
2424 struct buffer_head *, char *, int,
2425 unsigned int);
2426#define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset) \
2427 unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
2428 (de), (bh), (buf), (size), (offset)))
2429extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
2430 __u32 minor_hash,
2431 struct ext4_dir_entry_2 *dirent,
2432 struct ext4_str *ent_name);
2433extern void ext4_htree_free_dir_info(struct dir_private_info *p);
2434extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
2435 struct buffer_head *bh,
2436 void *buf, int buf_size,
2437 struct ext4_filename *fname,
2438 struct ext4_dir_entry_2 **dest_de);
2439int ext4_insert_dentry(struct inode *dir,
2440 struct inode *inode,
2441 struct ext4_dir_entry_2 *de,
2442 int buf_size,
2443 struct ext4_filename *fname);
2444static inline void ext4_update_dx_flag(struct inode *inode)
2445{
2446 if (!ext4_has_feature_dir_index(inode->i_sb))
2447 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
2448}
2449static unsigned char ext4_filetype_table[] = {
2450 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
2451};
2452
2453static inline unsigned char get_dtype(struct super_block *sb, int filetype)
2454{
2455 if (!ext4_has_feature_filetype(sb) || filetype >= EXT4_FT_MAX)
2456 return DT_UNKNOWN;
2457
2458 return ext4_filetype_table[filetype];
2459}
2460extern int ext4_check_all_de(struct inode *dir, struct buffer_head *bh,
2461 void *buf, int buf_size);
2462
2463/* fsync.c */
2464extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2465
2466/* hash.c */
2467extern int ext4fs_dirhash(const char *name, int len, struct
2468 dx_hash_info *hinfo);
2469
2470/* ialloc.c */
2471extern struct inode *__ext4_new_inode(handle_t *, struct inode *, umode_t,
2472 const struct qstr *qstr, __u32 goal,
2473 uid_t *owner, int handle_type,
2474 unsigned int line_no, int nblocks);
2475
2476#define ext4_new_inode(handle, dir, mode, qstr, goal, owner) \
2477 __ext4_new_inode((handle), (dir), (mode), (qstr), (goal), (owner), \
2478 0, 0, 0)
2479#define ext4_new_inode_start_handle(dir, mode, qstr, goal, owner, \
2480 type, nblocks) \
2481 __ext4_new_inode(NULL, (dir), (mode), (qstr), (goal), (owner), \
2482 (type), __LINE__, (nblocks))
2483
2484
2485extern void ext4_free_inode(handle_t *, struct inode *);
2486extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2487extern unsigned long ext4_count_free_inodes(struct super_block *);
2488extern unsigned long ext4_count_dirs(struct super_block *);
2489extern void ext4_check_inodes_bitmap(struct super_block *);
2490extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2491extern int ext4_init_inode_table(struct super_block *sb,
2492 ext4_group_t group, int barrier);
2493extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2494
2495/* mballoc.c */
2496extern const struct file_operations ext4_seq_mb_groups_fops;
2497extern long ext4_mb_stats;
2498extern long ext4_mb_max_to_scan;
2499extern int ext4_mb_init(struct super_block *);
2500extern int ext4_mb_release(struct super_block *);
2501extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2502 struct ext4_allocation_request *, int *);
2503extern int ext4_mb_reserve_blocks(struct super_block *, int);
2504extern void ext4_discard_preallocations(struct inode *);
2505extern int __init ext4_init_mballoc(void);
2506extern void ext4_exit_mballoc(void);
2507extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2508 struct buffer_head *bh, ext4_fsblk_t block,
2509 unsigned long count, int flags);
2510extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2511 ext4_group_t ngroups);
2512extern int ext4_mb_add_groupinfo(struct super_block *sb,
2513 ext4_group_t i, struct ext4_group_desc *desc);
2514extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2515 ext4_fsblk_t block, unsigned long count);
2516extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2517
2518/* inode.c */
2519int ext4_inode_is_fast_symlink(struct inode *inode);
2520struct buffer_head *ext4_getblk(handle_t *, struct inode *, ext4_lblk_t, int);
2521struct buffer_head *ext4_bread(handle_t *, struct inode *, ext4_lblk_t, int);
2522int ext4_get_block_unwritten(struct inode *inode, sector_t iblock,
2523 struct buffer_head *bh_result, int create);
2524int ext4_dax_mmap_get_block(struct inode *inode, sector_t iblock,
2525 struct buffer_head *bh_result, int create);
2526int ext4_get_block(struct inode *inode, sector_t iblock,
2527 struct buffer_head *bh_result, int create);
2528int ext4_dio_get_block(struct inode *inode, sector_t iblock,
2529 struct buffer_head *bh_result, int create);
2530int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2531 struct buffer_head *bh, int create);
2532int ext4_walk_page_buffers(handle_t *handle,
2533 struct buffer_head *head,
2534 unsigned from,
2535 unsigned to,
2536 int *partial,
2537 int (*fn)(handle_t *handle,
2538 struct buffer_head *bh));
2539int do_journal_get_write_access(handle_t *handle,
2540 struct buffer_head *bh);
2541#define FALL_BACK_TO_NONDELALLOC 1
2542#define CONVERT_INLINE_DATA 2
2543
2544extern struct inode *ext4_iget(struct super_block *, unsigned long);
2545extern struct inode *ext4_iget_normal(struct super_block *, unsigned long);
2546extern int ext4_write_inode(struct inode *, struct writeback_control *);
2547extern int ext4_setattr(struct dentry *, struct iattr *);
2548extern int ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
2549 struct kstat *stat);
2550extern void ext4_evict_inode(struct inode *);
2551extern void ext4_clear_inode(struct inode *);
2552extern int ext4_sync_inode(handle_t *, struct inode *);
2553extern void ext4_dirty_inode(struct inode *, int);
2554extern int ext4_change_inode_journal_flag(struct inode *, int);
2555extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
2556extern int ext4_inode_attach_jinode(struct inode *inode);
2557extern int ext4_can_truncate(struct inode *inode);
2558extern void ext4_truncate(struct inode *);
2559extern int ext4_punch_hole(struct inode *inode, loff_t offset, loff_t length);
2560extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
2561extern void ext4_set_inode_flags(struct inode *);
2562extern void ext4_get_inode_flags(struct ext4_inode_info *);
2563extern int ext4_alloc_da_blocks(struct inode *inode);
2564extern void ext4_set_aops(struct inode *inode);
2565extern int ext4_writepage_trans_blocks(struct inode *);
2566extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
2567extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
2568 loff_t lstart, loff_t lend);
2569extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2570extern int ext4_filemap_fault(struct vm_area_struct *vma, struct vm_fault *vmf);
2571extern qsize_t *ext4_get_reserved_space(struct inode *inode);
2572extern int ext4_get_projid(struct inode *inode, kprojid_t *projid);
2573extern void ext4_da_update_reserve_space(struct inode *inode,
2574 int used, int quota_claim);
2575extern int ext4_issue_zeroout(struct inode *inode, ext4_lblk_t lblk,
2576 ext4_fsblk_t pblk, ext4_lblk_t len);
2577extern int ext4_get_next_extent(struct inode *inode, ext4_lblk_t lblk,
2578 unsigned int map_len,
2579 struct extent_status *result);
2580
2581/* indirect.c */
2582extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
2583 struct ext4_map_blocks *map, int flags);
2584extern ssize_t ext4_ind_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
2585 loff_t offset);
2586extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
2587extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
2588extern void ext4_ind_truncate(handle_t *, struct inode *inode);
2589extern int ext4_ind_remove_space(handle_t *handle, struct inode *inode,
2590 ext4_lblk_t start, ext4_lblk_t end);
2591
2592/* ioctl.c */
2593extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
2594extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
2595
2596/* migrate.c */
2597extern int ext4_ext_migrate(struct inode *);
2598extern int ext4_ind_migrate(struct inode *inode);
2599
2600/* namei.c */
2601extern int ext4_dirent_csum_verify(struct inode *inode,
2602 struct ext4_dir_entry *dirent);
2603extern int ext4_orphan_add(handle_t *, struct inode *);
2604extern int ext4_orphan_del(handle_t *, struct inode *);
2605extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
2606 __u32 start_minor_hash, __u32 *next_hash);
2607extern int ext4_search_dir(struct buffer_head *bh,
2608 char *search_buf,
2609 int buf_size,
2610 struct inode *dir,
2611 struct ext4_filename *fname,
2612 const struct qstr *d_name,
2613 unsigned int offset,
2614 struct ext4_dir_entry_2 **res_dir);
2615extern int ext4_generic_delete_entry(handle_t *handle,
2616 struct inode *dir,
2617 struct ext4_dir_entry_2 *de_del,
2618 struct buffer_head *bh,
2619 void *entry_buf,
2620 int buf_size,
2621 int csum_size);
2622extern int ext4_empty_dir(struct inode *inode);
2623
2624/* resize.c */
2625extern int ext4_group_add(struct super_block *sb,
2626 struct ext4_new_group_data *input);
2627extern int ext4_group_extend(struct super_block *sb,
2628 struct ext4_super_block *es,
2629 ext4_fsblk_t n_blocks_count);
2630extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
2631
2632/* super.c */
2633extern int ext4_seq_options_show(struct seq_file *seq, void *offset);
2634extern int ext4_calculate_overhead(struct super_block *sb);
2635extern void ext4_superblock_csum_set(struct super_block *sb);
2636extern void *ext4_kvmalloc(size_t size, gfp_t flags);
2637extern void *ext4_kvzalloc(size_t size, gfp_t flags);
2638extern int ext4_alloc_flex_bg_array(struct super_block *sb,
2639 ext4_group_t ngroup);
2640extern const char *ext4_decode_error(struct super_block *sb, int errno,
2641 char nbuf[16]);
2642
2643extern __printf(4, 5)
2644void __ext4_error(struct super_block *, const char *, unsigned int,
2645 const char *, ...);
2646extern __printf(5, 6)
2647void __ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
2648 const char *, ...);
2649extern __printf(5, 6)
2650void __ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
2651 const char *, ...);
2652extern void __ext4_std_error(struct super_block *, const char *,
2653 unsigned int, int);
2654extern __printf(4, 5)
2655void __ext4_abort(struct super_block *, const char *, unsigned int,
2656 const char *, ...);
2657extern __printf(4, 5)
2658void __ext4_warning(struct super_block *, const char *, unsigned int,
2659 const char *, ...);
2660extern __printf(4, 5)
2661void __ext4_warning_inode(const struct inode *inode, const char *function,
2662 unsigned int line, const char *fmt, ...);
2663extern __printf(3, 4)
2664void __ext4_msg(struct super_block *, const char *, const char *, ...);
2665extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
2666 const char *, unsigned int, const char *);
2667extern __printf(7, 8)
2668void __ext4_grp_locked_error(const char *, unsigned int,
2669 struct super_block *, ext4_group_t,
2670 unsigned long, ext4_fsblk_t,
2671 const char *, ...);
2672
2673#define EXT4_ERROR_INODE(inode, fmt, a...) \
2674 ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
2675
2676#define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...) \
2677 ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
2678
2679#define EXT4_ERROR_FILE(file, block, fmt, a...) \
2680 ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
2681
2682#ifdef CONFIG_PRINTK
2683
2684#define ext4_error_inode(inode, func, line, block, fmt, ...) \
2685 __ext4_error_inode(inode, func, line, block, fmt, ##__VA_ARGS__)
2686#define ext4_error_file(file, func, line, block, fmt, ...) \
2687 __ext4_error_file(file, func, line, block, fmt, ##__VA_ARGS__)
2688#define ext4_error(sb, fmt, ...) \
2689 __ext4_error(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2690#define ext4_abort(sb, fmt, ...) \
2691 __ext4_abort(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2692#define ext4_warning(sb, fmt, ...) \
2693 __ext4_warning(sb, __func__, __LINE__, fmt, ##__VA_ARGS__)
2694#define ext4_warning_inode(inode, fmt, ...) \
2695 __ext4_warning_inode(inode, __func__, __LINE__, fmt, ##__VA_ARGS__)
2696#define ext4_msg(sb, level, fmt, ...) \
2697 __ext4_msg(sb, level, fmt, ##__VA_ARGS__)
2698#define dump_mmp_msg(sb, mmp, msg) \
2699 __dump_mmp_msg(sb, mmp, __func__, __LINE__, msg)
2700#define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...) \
2701 __ext4_grp_locked_error(__func__, __LINE__, sb, grp, ino, block, \
2702 fmt, ##__VA_ARGS__)
2703
2704#else
2705
2706#define ext4_error_inode(inode, func, line, block, fmt, ...) \
2707do { \
2708 no_printk(fmt, ##__VA_ARGS__); \
2709 __ext4_error_inode(inode, "", 0, block, " "); \
2710} while (0)
2711#define ext4_error_file(file, func, line, block, fmt, ...) \
2712do { \
2713 no_printk(fmt, ##__VA_ARGS__); \
2714 __ext4_error_file(file, "", 0, block, " "); \
2715} while (0)
2716#define ext4_error(sb, fmt, ...) \
2717do { \
2718 no_printk(fmt, ##__VA_ARGS__); \
2719 __ext4_error(sb, "", 0, " "); \
2720} while (0)
2721#define ext4_abort(sb, fmt, ...) \
2722do { \
2723 no_printk(fmt, ##__VA_ARGS__); \
2724 __ext4_abort(sb, "", 0, " "); \
2725} while (0)
2726#define ext4_warning(sb, fmt, ...) \
2727do { \
2728 no_printk(fmt, ##__VA_ARGS__); \
2729 __ext4_warning(sb, "", 0, " "); \
2730} while (0)
2731#define ext4_warning_inode(inode, fmt, ...) \
2732do { \
2733 no_printk(fmt, ##__VA_ARGS__); \
2734 __ext4_warning_inode(inode, "", 0, " "); \
2735} while (0)
2736#define ext4_msg(sb, level, fmt, ...) \
2737do { \
2738 no_printk(fmt, ##__VA_ARGS__); \
2739 __ext4_msg(sb, "", " "); \
2740} while (0)
2741#define dump_mmp_msg(sb, mmp, msg) \
2742 __dump_mmp_msg(sb, mmp, "", 0, "")
2743#define ext4_grp_locked_error(sb, grp, ino, block, fmt, ...) \
2744do { \
2745 no_printk(fmt, ##__VA_ARGS__); \
2746 __ext4_grp_locked_error("", 0, sb, grp, ino, block, " "); \
2747} while (0)
2748
2749#endif
2750
2751extern void ext4_update_dynamic_rev(struct super_block *sb);
2752extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
2753 __u32 compat);
2754extern int ext4_update_rocompat_feature(handle_t *handle,
2755 struct super_block *sb, __u32 rocompat);
2756extern int ext4_update_incompat_feature(handle_t *handle,
2757 struct super_block *sb, __u32 incompat);
2758extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
2759 struct ext4_group_desc *bg);
2760extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
2761 struct ext4_group_desc *bg);
2762extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
2763 struct ext4_group_desc *bg);
2764extern __u32 ext4_free_group_clusters(struct super_block *sb,
2765 struct ext4_group_desc *bg);
2766extern __u32 ext4_free_inodes_count(struct super_block *sb,
2767 struct ext4_group_desc *bg);
2768extern __u32 ext4_used_dirs_count(struct super_block *sb,
2769 struct ext4_group_desc *bg);
2770extern __u32 ext4_itable_unused_count(struct super_block *sb,
2771 struct ext4_group_desc *bg);
2772extern void ext4_block_bitmap_set(struct super_block *sb,
2773 struct ext4_group_desc *bg, ext4_fsblk_t blk);
2774extern void ext4_inode_bitmap_set(struct super_block *sb,
2775 struct ext4_group_desc *bg, ext4_fsblk_t blk);
2776extern void ext4_inode_table_set(struct super_block *sb,
2777 struct ext4_group_desc *bg, ext4_fsblk_t blk);
2778extern void ext4_free_group_clusters_set(struct super_block *sb,
2779 struct ext4_group_desc *bg,
2780 __u32 count);
2781extern void ext4_free_inodes_set(struct super_block *sb,
2782 struct ext4_group_desc *bg, __u32 count);
2783extern void ext4_used_dirs_set(struct super_block *sb,
2784 struct ext4_group_desc *bg, __u32 count);
2785extern void ext4_itable_unused_set(struct super_block *sb,
2786 struct ext4_group_desc *bg, __u32 count);
2787extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
2788 struct ext4_group_desc *gdp);
2789extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
2790 struct ext4_group_desc *gdp);
2791extern int ext4_register_li_request(struct super_block *sb,
2792 ext4_group_t first_not_zeroed);
2793
2794static inline int ext4_has_group_desc_csum(struct super_block *sb)
2795{
2796 return ext4_has_feature_gdt_csum(sb) ||
2797 EXT4_SB(sb)->s_chksum_driver != NULL;
2798}
2799
2800static inline int ext4_has_metadata_csum(struct super_block *sb)
2801{
2802 WARN_ON_ONCE(ext4_has_feature_metadata_csum(sb) &&
2803 !EXT4_SB(sb)->s_chksum_driver);
2804
2805 return (EXT4_SB(sb)->s_chksum_driver != NULL);
2806}
2807static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2808{
2809 return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2810 le32_to_cpu(es->s_blocks_count_lo);
2811}
2812
2813static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2814{
2815 return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2816 le32_to_cpu(es->s_r_blocks_count_lo);
2817}
2818
2819static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2820{
2821 return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2822 le32_to_cpu(es->s_free_blocks_count_lo);
2823}
2824
2825static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2826 ext4_fsblk_t blk)
2827{
2828 es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2829 es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2830}
2831
2832static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2833 ext4_fsblk_t blk)
2834{
2835 es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2836 es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2837}
2838
2839static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2840 ext4_fsblk_t blk)
2841{
2842 es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2843 es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2844}
2845
2846static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2847{
2848 if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2849 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2850 le32_to_cpu(raw_inode->i_size_lo);
2851 else
2852 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2853}
2854
2855static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2856{
2857 raw_inode->i_size_lo = cpu_to_le32(i_size);
2858 raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2859}
2860
2861static inline
2862struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2863 ext4_group_t group)
2864{
2865 struct ext4_group_info ***grp_info;
2866 long indexv, indexh;
2867 BUG_ON(group >= EXT4_SB(sb)->s_groups_count);
2868 grp_info = EXT4_SB(sb)->s_group_info;
2869 indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2870 indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2871 return grp_info[indexv][indexh];
2872}
2873
2874/*
2875 * Reading s_groups_count requires using smp_rmb() afterwards. See
2876 * the locking protocol documented in the comments of ext4_group_add()
2877 * in resize.c
2878 */
2879static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2880{
2881 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2882
2883 smp_rmb();
2884 return ngroups;
2885}
2886
2887static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2888 ext4_group_t block_group)
2889{
2890 return block_group >> sbi->s_log_groups_per_flex;
2891}
2892
2893static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2894{
2895 return 1 << sbi->s_log_groups_per_flex;
2896}
2897
2898#define ext4_std_error(sb, errno) \
2899do { \
2900 if ((errno)) \
2901 __ext4_std_error((sb), __func__, __LINE__, (errno)); \
2902} while (0)
2903
2904#ifdef CONFIG_SMP
2905/* Each CPU can accumulate percpu_counter_batch clusters in their local
2906 * counters. So we need to make sure we have free clusters more
2907 * than percpu_counter_batch * nr_cpu_ids. Also add a window of 4 times.
2908 */
2909#define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2910#else
2911#define EXT4_FREECLUSTERS_WATERMARK 0
2912#endif
2913
2914/* Update i_disksize. Requires i_mutex to avoid races with truncate */
2915static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2916{
2917 WARN_ON_ONCE(S_ISREG(inode->i_mode) &&
2918 !inode_is_locked(inode));
2919 down_write(&EXT4_I(inode)->i_data_sem);
2920 if (newsize > EXT4_I(inode)->i_disksize)
2921 EXT4_I(inode)->i_disksize = newsize;
2922 up_write(&EXT4_I(inode)->i_data_sem);
2923}
2924
2925/* Update i_size, i_disksize. Requires i_mutex to avoid races with truncate */
2926static inline int ext4_update_inode_size(struct inode *inode, loff_t newsize)
2927{
2928 int changed = 0;
2929
2930 if (newsize > inode->i_size) {
2931 i_size_write(inode, newsize);
2932 changed = 1;
2933 }
2934 if (newsize > EXT4_I(inode)->i_disksize) {
2935 ext4_update_i_disksize(inode, newsize);
2936 changed |= 2;
2937 }
2938 return changed;
2939}
2940
2941int ext4_update_disksize_before_punch(struct inode *inode, loff_t offset,
2942 loff_t len);
2943
2944struct ext4_group_info {
2945 unsigned long bb_state;
2946 struct rb_root bb_free_root;
2947 ext4_grpblk_t bb_first_free; /* first free block */
2948 ext4_grpblk_t bb_free; /* total free blocks */
2949 ext4_grpblk_t bb_fragments; /* nr of freespace fragments */
2950 ext4_grpblk_t bb_largest_free_order;/* order of largest frag in BG */
2951 struct list_head bb_prealloc_list;
2952#ifdef DOUBLE_CHECK
2953 void *bb_bitmap;
2954#endif
2955 struct rw_semaphore alloc_sem;
2956 ext4_grpblk_t bb_counters[]; /* Nr of free power-of-two-block
2957 * regions, index is order.
2958 * bb_counters[3] = 5 means
2959 * 5 free 8-block regions. */
2960};
2961
2962#define EXT4_GROUP_INFO_NEED_INIT_BIT 0
2963#define EXT4_GROUP_INFO_WAS_TRIMMED_BIT 1
2964#define EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT 2
2965#define EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT 3
2966
2967#define EXT4_MB_GRP_NEED_INIT(grp) \
2968 (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2969#define EXT4_MB_GRP_BBITMAP_CORRUPT(grp) \
2970 (test_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
2971#define EXT4_MB_GRP_IBITMAP_CORRUPT(grp) \
2972 (test_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, &((grp)->bb_state)))
2973
2974#define EXT4_MB_GRP_WAS_TRIMMED(grp) \
2975 (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2976#define EXT4_MB_GRP_SET_TRIMMED(grp) \
2977 (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2978#define EXT4_MB_GRP_CLEAR_TRIMMED(grp) \
2979 (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2980
2981#define EXT4_MAX_CONTENTION 8
2982#define EXT4_CONTENTION_THRESHOLD 2
2983
2984static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2985 ext4_group_t group)
2986{
2987 return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2988}
2989
2990/*
2991 * Returns true if the filesystem is busy enough that attempts to
2992 * access the block group locks has run into contention.
2993 */
2994static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2995{
2996 return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2997}
2998
2999static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
3000{
3001 spinlock_t *lock = ext4_group_lock_ptr(sb, group);
3002 if (spin_trylock(lock))
3003 /*
3004 * We're able to grab the lock right away, so drop the
3005 * lock contention counter.
3006 */
3007 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
3008 else {
3009 /*
3010 * The lock is busy, so bump the contention counter,
3011 * and then wait on the spin lock.
3012 */
3013 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
3014 EXT4_MAX_CONTENTION);
3015 spin_lock(lock);
3016 }
3017}
3018
3019static inline void ext4_unlock_group(struct super_block *sb,
3020 ext4_group_t group)
3021{
3022 spin_unlock(ext4_group_lock_ptr(sb, group));
3023}
3024
3025/*
3026 * Block validity checking
3027 */
3028#define ext4_check_indirect_blockref(inode, bh) \
3029 ext4_check_blockref(__func__, __LINE__, inode, \
3030 (__le32 *)(bh)->b_data, \
3031 EXT4_ADDR_PER_BLOCK((inode)->i_sb))
3032
3033#define ext4_ind_check_inode(inode) \
3034 ext4_check_blockref(__func__, __LINE__, inode, \
3035 EXT4_I(inode)->i_data, \
3036 EXT4_NDIR_BLOCKS)
3037
3038/*
3039 * Inodes and files operations
3040 */
3041
3042/* dir.c */
3043extern const struct file_operations ext4_dir_operations;
3044
3045/* file.c */
3046extern const struct inode_operations ext4_file_inode_operations;
3047extern const struct file_operations ext4_file_operations;
3048extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
3049
3050/* inline.c */
3051extern int ext4_get_max_inline_size(struct inode *inode);
3052extern int ext4_find_inline_data_nolock(struct inode *inode);
3053extern int ext4_init_inline_data(handle_t *handle, struct inode *inode,
3054 unsigned int len);
3055extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
3056
3057extern int ext4_readpage_inline(struct inode *inode, struct page *page);
3058extern int ext4_try_to_write_inline_data(struct address_space *mapping,
3059 struct inode *inode,
3060 loff_t pos, unsigned len,
3061 unsigned flags,
3062 struct page **pagep);
3063extern int ext4_write_inline_data_end(struct inode *inode,
3064 loff_t pos, unsigned len,
3065 unsigned copied,
3066 struct page *page);
3067extern struct buffer_head *
3068ext4_journalled_write_inline_data(struct inode *inode,
3069 unsigned len,
3070 struct page *page);
3071extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
3072 struct inode *inode,
3073 loff_t pos, unsigned len,
3074 unsigned flags,
3075 struct page **pagep,
3076 void **fsdata);
3077extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
3078 unsigned len, unsigned copied,
3079 struct page *page);
3080extern int ext4_try_add_inline_entry(handle_t *handle,
3081 struct ext4_filename *fname,
3082 struct inode *dir, struct inode *inode);
3083extern int ext4_try_create_inline_dir(handle_t *handle,
3084 struct inode *parent,
3085 struct inode *inode);
3086extern int ext4_read_inline_dir(struct file *filp,
3087 struct dir_context *ctx,
3088 int *has_inline_data);
3089extern int htree_inlinedir_to_tree(struct file *dir_file,
3090 struct inode *dir, ext4_lblk_t block,
3091 struct dx_hash_info *hinfo,
3092 __u32 start_hash, __u32 start_minor_hash,
3093 int *has_inline_data);
3094extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
3095 struct ext4_filename *fname,
3096 const struct qstr *d_name,
3097 struct ext4_dir_entry_2 **res_dir,
3098 int *has_inline_data);
3099extern int ext4_delete_inline_entry(handle_t *handle,
3100 struct inode *dir,
3101 struct ext4_dir_entry_2 *de_del,
3102 struct buffer_head *bh,
3103 int *has_inline_data);
3104extern int empty_inline_dir(struct inode *dir, int *has_inline_data);
3105extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
3106 struct ext4_dir_entry_2 **parent_de,
3107 int *retval);
3108extern int ext4_inline_data_fiemap(struct inode *inode,
3109 struct fiemap_extent_info *fieinfo,
3110 int *has_inline, __u64 start, __u64 len);
3111extern int ext4_try_to_evict_inline_data(handle_t *handle,
3112 struct inode *inode,
3113 int needed);
3114extern void ext4_inline_data_truncate(struct inode *inode, int *has_inline);
3115
3116extern int ext4_convert_inline_data(struct inode *inode);
3117
3118static inline int ext4_has_inline_data(struct inode *inode)
3119{
3120 return ext4_test_inode_flag(inode, EXT4_INODE_INLINE_DATA) &&
3121 EXT4_I(inode)->i_inline_off;
3122}
3123
3124/* namei.c */
3125extern const struct inode_operations ext4_dir_inode_operations;
3126extern const struct inode_operations ext4_special_inode_operations;
3127extern struct dentry *ext4_get_parent(struct dentry *child);
3128extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
3129 struct ext4_dir_entry_2 *de,
3130 int blocksize, int csum_size,
3131 unsigned int parent_ino, int dotdot_real_len);
3132extern void initialize_dirent_tail(struct ext4_dir_entry_tail *t,
3133 unsigned int blocksize);
3134extern int ext4_handle_dirty_dirent_node(handle_t *handle,
3135 struct inode *inode,
3136 struct buffer_head *bh);
3137#define S_SHIFT 12
3138static unsigned char ext4_type_by_mode[S_IFMT >> S_SHIFT] = {
3139 [S_IFREG >> S_SHIFT] = EXT4_FT_REG_FILE,
3140 [S_IFDIR >> S_SHIFT] = EXT4_FT_DIR,
3141 [S_IFCHR >> S_SHIFT] = EXT4_FT_CHRDEV,
3142 [S_IFBLK >> S_SHIFT] = EXT4_FT_BLKDEV,
3143 [S_IFIFO >> S_SHIFT] = EXT4_FT_FIFO,
3144 [S_IFSOCK >> S_SHIFT] = EXT4_FT_SOCK,
3145 [S_IFLNK >> S_SHIFT] = EXT4_FT_SYMLINK,
3146};
3147
3148static inline void ext4_set_de_type(struct super_block *sb,
3149 struct ext4_dir_entry_2 *de,
3150 umode_t mode) {
3151 if (ext4_has_feature_filetype(sb))
3152 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
3153}
3154
3155/* readpages.c */
3156extern int ext4_mpage_readpages(struct address_space *mapping,
3157 struct list_head *pages, struct page *page,
3158 unsigned nr_pages);
3159
3160/* symlink.c */
3161extern const struct inode_operations ext4_encrypted_symlink_inode_operations;
3162extern const struct inode_operations ext4_symlink_inode_operations;
3163extern const struct inode_operations ext4_fast_symlink_inode_operations;
3164
3165/* sysfs.c */
3166extern int ext4_register_sysfs(struct super_block *sb);
3167extern void ext4_unregister_sysfs(struct super_block *sb);
3168extern int __init ext4_init_sysfs(void);
3169extern void ext4_exit_sysfs(void);
3170
3171/* block_validity */
3172extern void ext4_release_system_zone(struct super_block *sb);
3173extern int ext4_setup_system_zone(struct super_block *sb);
3174extern int __init ext4_init_system_zone(void);
3175extern void ext4_exit_system_zone(void);
3176extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
3177 ext4_fsblk_t start_blk,
3178 unsigned int count);
3179extern int ext4_check_blockref(const char *, unsigned int,
3180 struct inode *, __le32 *, unsigned int);
3181
3182/* extents.c */
3183struct ext4_ext_path;
3184struct ext4_extent;
3185
3186/*
3187 * Maximum number of logical blocks in a file; ext4_extent's ee_block is
3188 * __le32.
3189 */
3190#define EXT_MAX_BLOCKS 0xffffffff
3191
3192extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
3193extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
3194extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
3195extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
3196 struct ext4_map_blocks *map, int flags);
3197extern void ext4_ext_truncate(handle_t *, struct inode *);
3198extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
3199 ext4_lblk_t end);
3200extern void ext4_ext_init(struct super_block *);
3201extern void ext4_ext_release(struct super_block *);
3202extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
3203 loff_t len);
3204extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
3205 loff_t offset, ssize_t len);
3206extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
3207 struct ext4_map_blocks *map, int flags);
3208extern int ext4_ext_calc_metadata_amount(struct inode *inode,
3209 ext4_lblk_t lblocks);
3210extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
3211 int num,
3212 struct ext4_ext_path *path);
3213extern int ext4_can_extents_be_merged(struct inode *inode,
3214 struct ext4_extent *ex1,
3215 struct ext4_extent *ex2);
3216extern int ext4_ext_insert_extent(handle_t *, struct inode *,
3217 struct ext4_ext_path **,
3218 struct ext4_extent *, int);
3219extern struct ext4_ext_path *ext4_find_extent(struct inode *, ext4_lblk_t,
3220 struct ext4_ext_path **,
3221 int flags);
3222extern void ext4_ext_drop_refs(struct ext4_ext_path *);
3223extern int ext4_ext_check_inode(struct inode *inode);
3224extern int ext4_find_delalloc_range(struct inode *inode,
3225 ext4_lblk_t lblk_start,
3226 ext4_lblk_t lblk_end);
3227extern int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk);
3228extern ext4_lblk_t ext4_ext_next_allocated_block(struct ext4_ext_path *path);
3229extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
3230 __u64 start, __u64 len);
3231extern int ext4_ext_precache(struct inode *inode);
3232extern int ext4_collapse_range(struct inode *inode, loff_t offset, loff_t len);
3233extern int ext4_insert_range(struct inode *inode, loff_t offset, loff_t len);
3234extern int ext4_swap_extents(handle_t *handle, struct inode *inode1,
3235 struct inode *inode2, ext4_lblk_t lblk1,
3236 ext4_lblk_t lblk2, ext4_lblk_t count,
3237 int mark_unwritten,int *err);
3238
3239/* move_extent.c */
3240extern void ext4_double_down_write_data_sem(struct inode *first,
3241 struct inode *second);
3242extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
3243 struct inode *donor_inode);
3244extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
3245 __u64 start_orig, __u64 start_donor,
3246 __u64 len, __u64 *moved_len);
3247
3248/* page-io.c */
3249extern int __init ext4_init_pageio(void);
3250extern void ext4_exit_pageio(void);
3251extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
3252extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
3253extern int ext4_put_io_end(ext4_io_end_t *io_end);
3254extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
3255extern void ext4_io_submit_init(struct ext4_io_submit *io,
3256 struct writeback_control *wbc);
3257extern void ext4_end_io_rsv_work(struct work_struct *work);
3258extern void ext4_io_submit(struct ext4_io_submit *io);
3259extern int ext4_bio_write_page(struct ext4_io_submit *io,
3260 struct page *page,
3261 int len,
3262 struct writeback_control *wbc,
3263 bool keep_towrite);
3264
3265/* mmp.c */
3266extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
3267
3268/*
3269 * Add new method to test whether block and inode bitmaps are properly
3270 * initialized. With uninit_bg reading the block from disk is not enough
3271 * to mark the bitmap uptodate. We need to also zero-out the bitmap
3272 */
3273#define BH_BITMAP_UPTODATE BH_JBDPrivateStart
3274
3275static inline int bitmap_uptodate(struct buffer_head *bh)
3276{
3277 return (buffer_uptodate(bh) &&
3278 test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
3279}
3280static inline void set_bitmap_uptodate(struct buffer_head *bh)
3281{
3282 set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
3283}
3284
3285/*
3286 * Disable DIO read nolock optimization, so new dioreaders will be forced
3287 * to grab i_mutex
3288 */
3289static inline void ext4_inode_block_unlocked_dio(struct inode *inode)
3290{
3291 ext4_set_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
3292 smp_mb();
3293}
3294static inline void ext4_inode_resume_unlocked_dio(struct inode *inode)
3295{
3296 smp_mb();
3297 ext4_clear_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
3298}
3299
3300#define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
3301
3302/* For ioend & aio unwritten conversion wait queues */
3303#define EXT4_WQ_HASH_SZ 37
3304#define ext4_ioend_wq(v) (&ext4__ioend_wq[((unsigned long)(v)) %\
3305 EXT4_WQ_HASH_SZ])
3306extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
3307
3308#define EXT4_RESIZING 0
3309extern int ext4_resize_begin(struct super_block *sb);
3310extern void ext4_resize_end(struct super_block *sb);
3311
3312static inline void ext4_set_io_unwritten_flag(struct inode *inode,
3313 struct ext4_io_end *io_end)
3314{
3315 if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
3316 io_end->flag |= EXT4_IO_END_UNWRITTEN;
3317 atomic_inc(&EXT4_I(inode)->i_unwritten);
3318 }
3319}
3320
3321static inline void ext4_clear_io_unwritten_flag(ext4_io_end_t *io_end)
3322{
3323 struct inode *inode = io_end->inode;
3324
3325 if (io_end->flag & EXT4_IO_END_UNWRITTEN) {
3326 io_end->flag &= ~EXT4_IO_END_UNWRITTEN;
3327 /* Wake up anyone waiting on unwritten extent conversion */
3328 if (atomic_dec_and_test(&EXT4_I(inode)->i_unwritten))
3329 wake_up_all(ext4_ioend_wq(inode));
3330 }
3331}
3332
3333#endif /* __KERNEL__ */
3334
3335#define EFSBADCRC EBADMSG /* Bad CRC detected */
3336#define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
3337
3338#endif /* _EXT4_H */
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#ifdef __KERNEL__
33#include <linux/compat.h>
34#endif
35
36/*
37 * The fourth extended filesystem constants/structures
38 */
39
40/*
41 * Define EXT4FS_DEBUG to produce debug messages
42 */
43#undef EXT4FS_DEBUG
44
45/*
46 * Debug code
47 */
48#ifdef EXT4FS_DEBUG
49#define ext4_debug(f, a...) \
50 do { \
51 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:", \
52 __FILE__, __LINE__, __func__); \
53 printk(KERN_DEBUG f, ## a); \
54 } while (0)
55#else
56#define ext4_debug(f, a...) do {} while (0)
57#endif
58
59#define EXT4_ERROR_INODE(inode, fmt, a...) \
60 ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
61
62#define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...) \
63 ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
64
65#define EXT4_ERROR_FILE(file, block, fmt, a...) \
66 ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
67
68/* data type for block offset of block group */
69typedef int ext4_grpblk_t;
70
71/* data type for filesystem-wide blocks number */
72typedef unsigned long long ext4_fsblk_t;
73
74/* data type for file logical block number */
75typedef __u32 ext4_lblk_t;
76
77/* data type for block group number */
78typedef unsigned int ext4_group_t;
79
80/*
81 * Flags used in mballoc's allocation_context flags field.
82 *
83 * Also used to show what's going on for debugging purposes when the
84 * flag field is exported via the traceport interface
85 */
86
87/* prefer goal again. length */
88#define EXT4_MB_HINT_MERGE 0x0001
89/* blocks already reserved */
90#define EXT4_MB_HINT_RESERVED 0x0002
91/* metadata is being allocated */
92#define EXT4_MB_HINT_METADATA 0x0004
93/* first blocks in the file */
94#define EXT4_MB_HINT_FIRST 0x0008
95/* search for the best chunk */
96#define EXT4_MB_HINT_BEST 0x0010
97/* data is being allocated */
98#define EXT4_MB_HINT_DATA 0x0020
99/* don't preallocate (for tails) */
100#define EXT4_MB_HINT_NOPREALLOC 0x0040
101/* allocate for locality group */
102#define EXT4_MB_HINT_GROUP_ALLOC 0x0080
103/* allocate goal blocks or none */
104#define EXT4_MB_HINT_GOAL_ONLY 0x0100
105/* goal is meaningful */
106#define EXT4_MB_HINT_TRY_GOAL 0x0200
107/* blocks already pre-reserved by delayed allocation */
108#define EXT4_MB_DELALLOC_RESERVED 0x0400
109/* We are doing stream allocation */
110#define EXT4_MB_STREAM_ALLOC 0x0800
111/* Use reserved root blocks if needed */
112#define EXT4_MB_USE_ROOT_BLOCKS 0x1000
113
114struct ext4_allocation_request {
115 /* target inode for block we're allocating */
116 struct inode *inode;
117 /* how many blocks we want to allocate */
118 unsigned int len;
119 /* logical block in target inode */
120 ext4_lblk_t logical;
121 /* the closest logical allocated block to the left */
122 ext4_lblk_t lleft;
123 /* the closest logical allocated block to the right */
124 ext4_lblk_t lright;
125 /* phys. target (a hint) */
126 ext4_fsblk_t goal;
127 /* phys. block for the closest logical allocated block to the left */
128 ext4_fsblk_t pleft;
129 /* phys. block for the closest logical allocated block to the right */
130 ext4_fsblk_t pright;
131 /* flags. see above EXT4_MB_HINT_* */
132 unsigned int flags;
133};
134
135/*
136 * Logical to physical block mapping, used by ext4_map_blocks()
137 *
138 * This structure is used to pass requests into ext4_map_blocks() as
139 * well as to store the information returned by ext4_map_blocks(). It
140 * takes less room on the stack than a struct buffer_head.
141 */
142#define EXT4_MAP_NEW (1 << BH_New)
143#define EXT4_MAP_MAPPED (1 << BH_Mapped)
144#define EXT4_MAP_UNWRITTEN (1 << BH_Unwritten)
145#define EXT4_MAP_BOUNDARY (1 << BH_Boundary)
146#define EXT4_MAP_UNINIT (1 << BH_Uninit)
147#define EXT4_MAP_FLAGS (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
148 EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
149 EXT4_MAP_UNINIT)
150
151struct ext4_map_blocks {
152 ext4_fsblk_t m_pblk;
153 ext4_lblk_t m_lblk;
154 unsigned int m_len;
155 unsigned int m_flags;
156};
157
158/*
159 * For delayed allocation tracking
160 */
161struct mpage_da_data {
162 struct inode *inode;
163 sector_t b_blocknr; /* start block number of extent */
164 size_t b_size; /* size of extent */
165 unsigned long b_state; /* state of the extent */
166 unsigned long first_page, next_page; /* extent of pages */
167 struct writeback_control *wbc;
168 int io_done;
169 int pages_written;
170 int retval;
171};
172
173/*
174 * Flags for ext4_io_end->flags
175 */
176#define EXT4_IO_END_UNWRITTEN 0x0001
177#define EXT4_IO_END_ERROR 0x0002
178#define EXT4_IO_END_QUEUED 0x0004
179
180struct ext4_io_page {
181 struct page *p_page;
182 atomic_t p_count;
183};
184
185#define MAX_IO_PAGES 128
186
187typedef struct ext4_io_end {
188 struct list_head list; /* per-file finished IO list */
189 struct inode *inode; /* file being written to */
190 unsigned int flag; /* unwritten or not */
191 struct page *page; /* page struct for buffer write */
192 loff_t offset; /* offset in the file */
193 ssize_t size; /* size of the extent */
194 struct work_struct work; /* data work queue */
195 struct kiocb *iocb; /* iocb struct for AIO */
196 int result; /* error value for AIO */
197 int num_io_pages;
198 struct ext4_io_page *pages[MAX_IO_PAGES];
199} ext4_io_end_t;
200
201struct ext4_io_submit {
202 int io_op;
203 struct bio *io_bio;
204 ext4_io_end_t *io_end;
205 struct ext4_io_page *io_page;
206 sector_t io_next_block;
207};
208
209/*
210 * Special inodes numbers
211 */
212#define EXT4_BAD_INO 1 /* Bad blocks inode */
213#define EXT4_ROOT_INO 2 /* Root inode */
214#define EXT4_USR_QUOTA_INO 3 /* User quota inode */
215#define EXT4_GRP_QUOTA_INO 4 /* Group quota inode */
216#define EXT4_BOOT_LOADER_INO 5 /* Boot loader inode */
217#define EXT4_UNDEL_DIR_INO 6 /* Undelete directory inode */
218#define EXT4_RESIZE_INO 7 /* Reserved group descriptors inode */
219#define EXT4_JOURNAL_INO 8 /* Journal inode */
220
221/* First non-reserved inode for old ext4 filesystems */
222#define EXT4_GOOD_OLD_FIRST_INO 11
223
224/*
225 * Maximal count of links to a file
226 */
227#define EXT4_LINK_MAX 65000
228
229/*
230 * Macro-instructions used to manage several block sizes
231 */
232#define EXT4_MIN_BLOCK_SIZE 1024
233#define EXT4_MAX_BLOCK_SIZE 65536
234#define EXT4_MIN_BLOCK_LOG_SIZE 10
235#define EXT4_MAX_BLOCK_LOG_SIZE 16
236#ifdef __KERNEL__
237# define EXT4_BLOCK_SIZE(s) ((s)->s_blocksize)
238#else
239# define EXT4_BLOCK_SIZE(s) (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
240#endif
241#define EXT4_ADDR_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
242#ifdef __KERNEL__
243# define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits)
244#else
245# define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_log_block_size + 10)
246#endif
247#ifdef __KERNEL__
248#define EXT4_ADDR_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_addr_per_block_bits)
249#define EXT4_INODE_SIZE(s) (EXT4_SB(s)->s_inode_size)
250#define EXT4_FIRST_INO(s) (EXT4_SB(s)->s_first_ino)
251#else
252#define EXT4_INODE_SIZE(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
253 EXT4_GOOD_OLD_INODE_SIZE : \
254 (s)->s_inode_size)
255#define EXT4_FIRST_INO(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
256 EXT4_GOOD_OLD_FIRST_INO : \
257 (s)->s_first_ino)
258#endif
259#define EXT4_BLOCK_ALIGN(size, blkbits) ALIGN((size), (1 << (blkbits)))
260
261/*
262 * Structure of a blocks group descriptor
263 */
264struct ext4_group_desc
265{
266 __le32 bg_block_bitmap_lo; /* Blocks bitmap block */
267 __le32 bg_inode_bitmap_lo; /* Inodes bitmap block */
268 __le32 bg_inode_table_lo; /* Inodes table block */
269 __le16 bg_free_blocks_count_lo;/* Free blocks count */
270 __le16 bg_free_inodes_count_lo;/* Free inodes count */
271 __le16 bg_used_dirs_count_lo; /* Directories count */
272 __le16 bg_flags; /* EXT4_BG_flags (INODE_UNINIT, etc) */
273 __u32 bg_reserved[2]; /* Likely block/inode bitmap checksum */
274 __le16 bg_itable_unused_lo; /* Unused inodes count */
275 __le16 bg_checksum; /* crc16(sb_uuid+group+desc) */
276 __le32 bg_block_bitmap_hi; /* Blocks bitmap block MSB */
277 __le32 bg_inode_bitmap_hi; /* Inodes bitmap block MSB */
278 __le32 bg_inode_table_hi; /* Inodes table block MSB */
279 __le16 bg_free_blocks_count_hi;/* Free blocks count MSB */
280 __le16 bg_free_inodes_count_hi;/* Free inodes count MSB */
281 __le16 bg_used_dirs_count_hi; /* Directories count MSB */
282 __le16 bg_itable_unused_hi; /* Unused inodes count MSB */
283 __u32 bg_reserved2[3];
284};
285
286/*
287 * Structure of a flex block group info
288 */
289
290struct flex_groups {
291 atomic_t free_inodes;
292 atomic_t free_blocks;
293 atomic_t used_dirs;
294};
295
296#define EXT4_BG_INODE_UNINIT 0x0001 /* Inode table/bitmap not in use */
297#define EXT4_BG_BLOCK_UNINIT 0x0002 /* Block bitmap not in use */
298#define EXT4_BG_INODE_ZEROED 0x0004 /* On-disk itable initialized to zero */
299
300/*
301 * Macro-instructions used to manage group descriptors
302 */
303#define EXT4_MIN_DESC_SIZE 32
304#define EXT4_MIN_DESC_SIZE_64BIT 64
305#define EXT4_MAX_DESC_SIZE EXT4_MIN_BLOCK_SIZE
306#define EXT4_DESC_SIZE(s) (EXT4_SB(s)->s_desc_size)
307#ifdef __KERNEL__
308# define EXT4_BLOCKS_PER_GROUP(s) (EXT4_SB(s)->s_blocks_per_group)
309# define EXT4_DESC_PER_BLOCK(s) (EXT4_SB(s)->s_desc_per_block)
310# define EXT4_INODES_PER_GROUP(s) (EXT4_SB(s)->s_inodes_per_group)
311# define EXT4_DESC_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_desc_per_block_bits)
312#else
313# define EXT4_BLOCKS_PER_GROUP(s) ((s)->s_blocks_per_group)
314# define EXT4_DESC_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
315# define EXT4_INODES_PER_GROUP(s) ((s)->s_inodes_per_group)
316#endif
317
318/*
319 * Constants relative to the data blocks
320 */
321#define EXT4_NDIR_BLOCKS 12
322#define EXT4_IND_BLOCK EXT4_NDIR_BLOCKS
323#define EXT4_DIND_BLOCK (EXT4_IND_BLOCK + 1)
324#define EXT4_TIND_BLOCK (EXT4_DIND_BLOCK + 1)
325#define EXT4_N_BLOCKS (EXT4_TIND_BLOCK + 1)
326
327/*
328 * Inode flags
329 */
330#define EXT4_SECRM_FL 0x00000001 /* Secure deletion */
331#define EXT4_UNRM_FL 0x00000002 /* Undelete */
332#define EXT4_COMPR_FL 0x00000004 /* Compress file */
333#define EXT4_SYNC_FL 0x00000008 /* Synchronous updates */
334#define EXT4_IMMUTABLE_FL 0x00000010 /* Immutable file */
335#define EXT4_APPEND_FL 0x00000020 /* writes to file may only append */
336#define EXT4_NODUMP_FL 0x00000040 /* do not dump file */
337#define EXT4_NOATIME_FL 0x00000080 /* do not update atime */
338/* Reserved for compression usage... */
339#define EXT4_DIRTY_FL 0x00000100
340#define EXT4_COMPRBLK_FL 0x00000200 /* One or more compressed clusters */
341#define EXT4_NOCOMPR_FL 0x00000400 /* Don't compress */
342#define EXT4_ECOMPR_FL 0x00000800 /* Compression error */
343/* End compression flags --- maybe not all used */
344#define EXT4_INDEX_FL 0x00001000 /* hash-indexed directory */
345#define EXT4_IMAGIC_FL 0x00002000 /* AFS directory */
346#define EXT4_JOURNAL_DATA_FL 0x00004000 /* file data should be journaled */
347#define EXT4_NOTAIL_FL 0x00008000 /* file tail should not be merged */
348#define EXT4_DIRSYNC_FL 0x00010000 /* dirsync behaviour (directories only) */
349#define EXT4_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/
350#define EXT4_HUGE_FILE_FL 0x00040000 /* Set to each huge file */
351#define EXT4_EXTENTS_FL 0x00080000 /* Inode uses extents */
352#define EXT4_EA_INODE_FL 0x00200000 /* Inode used for large EA */
353#define EXT4_EOFBLOCKS_FL 0x00400000 /* Blocks allocated beyond EOF */
354#define EXT4_RESERVED_FL 0x80000000 /* reserved for ext4 lib */
355
356#define EXT4_FL_USER_VISIBLE 0x004BDFFF /* User visible flags */
357#define EXT4_FL_USER_MODIFIABLE 0x004B80FF /* User modifiable flags */
358
359/* Flags that should be inherited by new inodes from their parent. */
360#define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
361 EXT4_SYNC_FL | EXT4_IMMUTABLE_FL | EXT4_APPEND_FL |\
362 EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
363 EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
364 EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
365
366/* Flags that are appropriate for regular files (all but dir-specific ones). */
367#define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
368
369/* Flags that are appropriate for non-directories/regular files. */
370#define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
371
372/* Mask out flags that are inappropriate for the given type of inode. */
373static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
374{
375 if (S_ISDIR(mode))
376 return flags;
377 else if (S_ISREG(mode))
378 return flags & EXT4_REG_FLMASK;
379 else
380 return flags & EXT4_OTHER_FLMASK;
381}
382
383/*
384 * Inode flags used for atomic set/get
385 */
386enum {
387 EXT4_INODE_SECRM = 0, /* Secure deletion */
388 EXT4_INODE_UNRM = 1, /* Undelete */
389 EXT4_INODE_COMPR = 2, /* Compress file */
390 EXT4_INODE_SYNC = 3, /* Synchronous updates */
391 EXT4_INODE_IMMUTABLE = 4, /* Immutable file */
392 EXT4_INODE_APPEND = 5, /* writes to file may only append */
393 EXT4_INODE_NODUMP = 6, /* do not dump file */
394 EXT4_INODE_NOATIME = 7, /* do not update atime */
395/* Reserved for compression usage... */
396 EXT4_INODE_DIRTY = 8,
397 EXT4_INODE_COMPRBLK = 9, /* One or more compressed clusters */
398 EXT4_INODE_NOCOMPR = 10, /* Don't compress */
399 EXT4_INODE_ECOMPR = 11, /* Compression error */
400/* End compression flags --- maybe not all used */
401 EXT4_INODE_INDEX = 12, /* hash-indexed directory */
402 EXT4_INODE_IMAGIC = 13, /* AFS directory */
403 EXT4_INODE_JOURNAL_DATA = 14, /* file data should be journaled */
404 EXT4_INODE_NOTAIL = 15, /* file tail should not be merged */
405 EXT4_INODE_DIRSYNC = 16, /* dirsync behaviour (directories only) */
406 EXT4_INODE_TOPDIR = 17, /* Top of directory hierarchies*/
407 EXT4_INODE_HUGE_FILE = 18, /* Set to each huge file */
408 EXT4_INODE_EXTENTS = 19, /* Inode uses extents */
409 EXT4_INODE_EA_INODE = 21, /* Inode used for large EA */
410 EXT4_INODE_EOFBLOCKS = 22, /* Blocks allocated beyond EOF */
411 EXT4_INODE_RESERVED = 31, /* reserved for ext4 lib */
412};
413
414#define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
415#define CHECK_FLAG_VALUE(FLAG) if (!TEST_FLAG_VALUE(FLAG)) { \
416 printk(KERN_EMERG "EXT4 flag fail: " #FLAG ": %d %d\n", \
417 EXT4_##FLAG##_FL, EXT4_INODE_##FLAG); BUG_ON(1); }
418
419/*
420 * Since it's pretty easy to mix up bit numbers and hex values, and we
421 * can't do a compile-time test for ENUM values, we use a run-time
422 * test to make sure that EXT4_XXX_FL is consistent with respect to
423 * EXT4_INODE_XXX. If all is well the printk and BUG_ON will all drop
424 * out so it won't cost any extra space in the compiled kernel image.
425 * But it's important that these values are the same, since we are
426 * using EXT4_INODE_XXX to test for the flag values, but EXT4_XX_FL
427 * must be consistent with the values of FS_XXX_FL defined in
428 * include/linux/fs.h and the on-disk values found in ext2, ext3, and
429 * ext4 filesystems, and of course the values defined in e2fsprogs.
430 *
431 * It's not paranoia if the Murphy's Law really *is* out to get you. :-)
432 */
433static inline void ext4_check_flag_values(void)
434{
435 CHECK_FLAG_VALUE(SECRM);
436 CHECK_FLAG_VALUE(UNRM);
437 CHECK_FLAG_VALUE(COMPR);
438 CHECK_FLAG_VALUE(SYNC);
439 CHECK_FLAG_VALUE(IMMUTABLE);
440 CHECK_FLAG_VALUE(APPEND);
441 CHECK_FLAG_VALUE(NODUMP);
442 CHECK_FLAG_VALUE(NOATIME);
443 CHECK_FLAG_VALUE(DIRTY);
444 CHECK_FLAG_VALUE(COMPRBLK);
445 CHECK_FLAG_VALUE(NOCOMPR);
446 CHECK_FLAG_VALUE(ECOMPR);
447 CHECK_FLAG_VALUE(INDEX);
448 CHECK_FLAG_VALUE(IMAGIC);
449 CHECK_FLAG_VALUE(JOURNAL_DATA);
450 CHECK_FLAG_VALUE(NOTAIL);
451 CHECK_FLAG_VALUE(DIRSYNC);
452 CHECK_FLAG_VALUE(TOPDIR);
453 CHECK_FLAG_VALUE(HUGE_FILE);
454 CHECK_FLAG_VALUE(EXTENTS);
455 CHECK_FLAG_VALUE(EA_INODE);
456 CHECK_FLAG_VALUE(EOFBLOCKS);
457 CHECK_FLAG_VALUE(RESERVED);
458}
459
460/* Used to pass group descriptor data when online resize is done */
461struct ext4_new_group_input {
462 __u32 group; /* Group number for this data */
463 __u64 block_bitmap; /* Absolute block number of block bitmap */
464 __u64 inode_bitmap; /* Absolute block number of inode bitmap */
465 __u64 inode_table; /* Absolute block number of inode table start */
466 __u32 blocks_count; /* Total number of blocks in this group */
467 __u16 reserved_blocks; /* Number of reserved blocks in this group */
468 __u16 unused;
469};
470
471#if defined(__KERNEL__) && defined(CONFIG_COMPAT)
472struct compat_ext4_new_group_input {
473 u32 group;
474 compat_u64 block_bitmap;
475 compat_u64 inode_bitmap;
476 compat_u64 inode_table;
477 u32 blocks_count;
478 u16 reserved_blocks;
479 u16 unused;
480};
481#endif
482
483/* The struct ext4_new_group_input in kernel space, with free_blocks_count */
484struct ext4_new_group_data {
485 __u32 group;
486 __u64 block_bitmap;
487 __u64 inode_bitmap;
488 __u64 inode_table;
489 __u32 blocks_count;
490 __u16 reserved_blocks;
491 __u16 unused;
492 __u32 free_blocks_count;
493};
494
495/*
496 * Flags used by ext4_map_blocks()
497 */
498 /* Allocate any needed blocks and/or convert an unitialized
499 extent to be an initialized ext4 */
500#define EXT4_GET_BLOCKS_CREATE 0x0001
501 /* Request the creation of an unitialized extent */
502#define EXT4_GET_BLOCKS_UNINIT_EXT 0x0002
503#define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT (EXT4_GET_BLOCKS_UNINIT_EXT|\
504 EXT4_GET_BLOCKS_CREATE)
505 /* Caller is from the delayed allocation writeout path,
506 so set the magic i_delalloc_reserve_flag after taking the
507 inode allocation semaphore for */
508#define EXT4_GET_BLOCKS_DELALLOC_RESERVE 0x0004
509 /* caller is from the direct IO path, request to creation of an
510 unitialized extents if not allocated, split the uninitialized
511 extent if blocks has been preallocated already*/
512#define EXT4_GET_BLOCKS_PRE_IO 0x0008
513#define EXT4_GET_BLOCKS_CONVERT 0x0010
514#define EXT4_GET_BLOCKS_IO_CREATE_EXT (EXT4_GET_BLOCKS_PRE_IO|\
515 EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
516 /* Convert extent to initialized after IO complete */
517#define EXT4_GET_BLOCKS_IO_CONVERT_EXT (EXT4_GET_BLOCKS_CONVERT|\
518 EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
519 /* Punch out blocks of an extent */
520#define EXT4_GET_BLOCKS_PUNCH_OUT_EXT 0x0020
521 /* Don't normalize allocation size (used for fallocate) */
522#define EXT4_GET_BLOCKS_NO_NORMALIZE 0x0040
523
524/*
525 * Flags used by ext4_free_blocks
526 */
527#define EXT4_FREE_BLOCKS_METADATA 0x0001
528#define EXT4_FREE_BLOCKS_FORGET 0x0002
529#define EXT4_FREE_BLOCKS_VALIDATED 0x0004
530#define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
531
532/*
533 * ioctl commands
534 */
535#define EXT4_IOC_GETFLAGS FS_IOC_GETFLAGS
536#define EXT4_IOC_SETFLAGS FS_IOC_SETFLAGS
537#define EXT4_IOC_GETVERSION _IOR('f', 3, long)
538#define EXT4_IOC_SETVERSION _IOW('f', 4, long)
539#define EXT4_IOC_GETVERSION_OLD FS_IOC_GETVERSION
540#define EXT4_IOC_SETVERSION_OLD FS_IOC_SETVERSION
541#ifdef CONFIG_JBD2_DEBUG
542#define EXT4_IOC_WAIT_FOR_READONLY _IOR('f', 99, long)
543#endif
544#define EXT4_IOC_GETRSVSZ _IOR('f', 5, long)
545#define EXT4_IOC_SETRSVSZ _IOW('f', 6, long)
546#define EXT4_IOC_GROUP_EXTEND _IOW('f', 7, unsigned long)
547#define EXT4_IOC_GROUP_ADD _IOW('f', 8, struct ext4_new_group_input)
548#define EXT4_IOC_MIGRATE _IO('f', 9)
549 /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
550 /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
551#define EXT4_IOC_ALLOC_DA_BLKS _IO('f', 12)
552#define EXT4_IOC_MOVE_EXT _IOWR('f', 15, struct move_extent)
553
554#if defined(__KERNEL__) && defined(CONFIG_COMPAT)
555/*
556 * ioctl commands in 32 bit emulation
557 */
558#define EXT4_IOC32_GETFLAGS FS_IOC32_GETFLAGS
559#define EXT4_IOC32_SETFLAGS FS_IOC32_SETFLAGS
560#define EXT4_IOC32_GETVERSION _IOR('f', 3, int)
561#define EXT4_IOC32_SETVERSION _IOW('f', 4, int)
562#define EXT4_IOC32_GETRSVSZ _IOR('f', 5, int)
563#define EXT4_IOC32_SETRSVSZ _IOW('f', 6, int)
564#define EXT4_IOC32_GROUP_EXTEND _IOW('f', 7, unsigned int)
565#define EXT4_IOC32_GROUP_ADD _IOW('f', 8, struct compat_ext4_new_group_input)
566#ifdef CONFIG_JBD2_DEBUG
567#define EXT4_IOC32_WAIT_FOR_READONLY _IOR('f', 99, int)
568#endif
569#define EXT4_IOC32_GETVERSION_OLD FS_IOC32_GETVERSION
570#define EXT4_IOC32_SETVERSION_OLD FS_IOC32_SETVERSION
571#endif
572
573/* Max physical block we can address w/o extents */
574#define EXT4_MAX_BLOCK_FILE_PHYS 0xFFFFFFFF
575
576/*
577 * Structure of an inode on the disk
578 */
579struct ext4_inode {
580 __le16 i_mode; /* File mode */
581 __le16 i_uid; /* Low 16 bits of Owner Uid */
582 __le32 i_size_lo; /* Size in bytes */
583 __le32 i_atime; /* Access time */
584 __le32 i_ctime; /* Inode Change time */
585 __le32 i_mtime; /* Modification time */
586 __le32 i_dtime; /* Deletion Time */
587 __le16 i_gid; /* Low 16 bits of Group Id */
588 __le16 i_links_count; /* Links count */
589 __le32 i_blocks_lo; /* Blocks count */
590 __le32 i_flags; /* File flags */
591 union {
592 struct {
593 __le32 l_i_version;
594 } linux1;
595 struct {
596 __u32 h_i_translator;
597 } hurd1;
598 struct {
599 __u32 m_i_reserved1;
600 } masix1;
601 } osd1; /* OS dependent 1 */
602 __le32 i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
603 __le32 i_generation; /* File version (for NFS) */
604 __le32 i_file_acl_lo; /* File ACL */
605 __le32 i_size_high;
606 __le32 i_obso_faddr; /* Obsoleted fragment address */
607 union {
608 struct {
609 __le16 l_i_blocks_high; /* were l_i_reserved1 */
610 __le16 l_i_file_acl_high;
611 __le16 l_i_uid_high; /* these 2 fields */
612 __le16 l_i_gid_high; /* were reserved2[0] */
613 __u32 l_i_reserved2;
614 } linux2;
615 struct {
616 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */
617 __u16 h_i_mode_high;
618 __u16 h_i_uid_high;
619 __u16 h_i_gid_high;
620 __u32 h_i_author;
621 } hurd2;
622 struct {
623 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */
624 __le16 m_i_file_acl_high;
625 __u32 m_i_reserved2[2];
626 } masix2;
627 } osd2; /* OS dependent 2 */
628 __le16 i_extra_isize;
629 __le16 i_pad1;
630 __le32 i_ctime_extra; /* extra Change time (nsec << 2 | epoch) */
631 __le32 i_mtime_extra; /* extra Modification time(nsec << 2 | epoch) */
632 __le32 i_atime_extra; /* extra Access time (nsec << 2 | epoch) */
633 __le32 i_crtime; /* File Creation time */
634 __le32 i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
635 __le32 i_version_hi; /* high 32 bits for 64-bit version */
636};
637
638struct move_extent {
639 __u32 reserved; /* should be zero */
640 __u32 donor_fd; /* donor file descriptor */
641 __u64 orig_start; /* logical start offset in block for orig */
642 __u64 donor_start; /* logical start offset in block for donor */
643 __u64 len; /* block length to be moved */
644 __u64 moved_len; /* moved block length */
645};
646
647#define EXT4_EPOCH_BITS 2
648#define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
649#define EXT4_NSEC_MASK (~0UL << EXT4_EPOCH_BITS)
650
651/*
652 * Extended fields will fit into an inode if the filesystem was formatted
653 * with large inodes (-I 256 or larger) and there are not currently any EAs
654 * consuming all of the available space. For new inodes we always reserve
655 * enough space for the kernel's known extended fields, but for inodes
656 * created with an old kernel this might not have been the case. None of
657 * the extended inode fields is critical for correct filesystem operation.
658 * This macro checks if a certain field fits in the inode. Note that
659 * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
660 */
661#define EXT4_FITS_IN_INODE(ext4_inode, einode, field) \
662 ((offsetof(typeof(*ext4_inode), field) + \
663 sizeof((ext4_inode)->field)) \
664 <= (EXT4_GOOD_OLD_INODE_SIZE + \
665 (einode)->i_extra_isize)) \
666
667static inline __le32 ext4_encode_extra_time(struct timespec *time)
668{
669 return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
670 (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
671 ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
672}
673
674static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
675{
676 if (sizeof(time->tv_sec) > 4)
677 time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
678 << 32;
679 time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
680}
681
682#define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode) \
683do { \
684 (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec); \
685 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \
686 (raw_inode)->xtime ## _extra = \
687 ext4_encode_extra_time(&(inode)->xtime); \
688} while (0)
689
690#define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode) \
691do { \
692 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \
693 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec); \
694 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \
695 (raw_inode)->xtime ## _extra = \
696 ext4_encode_extra_time(&(einode)->xtime); \
697} while (0)
698
699#define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode) \
700do { \
701 (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime); \
702 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \
703 ext4_decode_extra_time(&(inode)->xtime, \
704 raw_inode->xtime ## _extra); \
705 else \
706 (inode)->xtime.tv_nsec = 0; \
707} while (0)
708
709#define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode) \
710do { \
711 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \
712 (einode)->xtime.tv_sec = \
713 (signed)le32_to_cpu((raw_inode)->xtime); \
714 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \
715 ext4_decode_extra_time(&(einode)->xtime, \
716 raw_inode->xtime ## _extra); \
717 else \
718 (einode)->xtime.tv_nsec = 0; \
719} while (0)
720
721#define i_disk_version osd1.linux1.l_i_version
722
723#if defined(__KERNEL__) || defined(__linux__)
724#define i_reserved1 osd1.linux1.l_i_reserved1
725#define i_file_acl_high osd2.linux2.l_i_file_acl_high
726#define i_blocks_high osd2.linux2.l_i_blocks_high
727#define i_uid_low i_uid
728#define i_gid_low i_gid
729#define i_uid_high osd2.linux2.l_i_uid_high
730#define i_gid_high osd2.linux2.l_i_gid_high
731#define i_reserved2 osd2.linux2.l_i_reserved2
732
733#elif defined(__GNU__)
734
735#define i_translator osd1.hurd1.h_i_translator
736#define i_uid_high osd2.hurd2.h_i_uid_high
737#define i_gid_high osd2.hurd2.h_i_gid_high
738#define i_author osd2.hurd2.h_i_author
739
740#elif defined(__masix__)
741
742#define i_reserved1 osd1.masix1.m_i_reserved1
743#define i_file_acl_high osd2.masix2.m_i_file_acl_high
744#define i_reserved2 osd2.masix2.m_i_reserved2
745
746#endif /* defined(__KERNEL__) || defined(__linux__) */
747
748/*
749 * storage for cached extent
750 * If ec_len == 0, then the cache is invalid.
751 * If ec_start == 0, then the cache represents a gap (null mapping)
752 */
753struct ext4_ext_cache {
754 ext4_fsblk_t ec_start;
755 ext4_lblk_t ec_block;
756 __u32 ec_len; /* must be 32bit to return holes */
757};
758
759/*
760 * fourth extended file system inode data in memory
761 */
762struct ext4_inode_info {
763 __le32 i_data[15]; /* unconverted */
764 __u32 i_dtime;
765 ext4_fsblk_t i_file_acl;
766
767 /*
768 * i_block_group is the number of the block group which contains
769 * this file's inode. Constant across the lifetime of the inode,
770 * it is ued for making block allocation decisions - we try to
771 * place a file's data blocks near its inode block, and new inodes
772 * near to their parent directory's inode.
773 */
774 ext4_group_t i_block_group;
775 ext4_lblk_t i_dir_start_lookup;
776#if (BITS_PER_LONG < 64)
777 unsigned long i_state_flags; /* Dynamic state flags */
778#endif
779 unsigned long i_flags;
780
781#ifdef CONFIG_EXT4_FS_XATTR
782 /*
783 * Extended attributes can be read independently of the main file
784 * data. Taking i_mutex even when reading would cause contention
785 * between readers of EAs and writers of regular file data, so
786 * instead we synchronize on xattr_sem when reading or changing
787 * EAs.
788 */
789 struct rw_semaphore xattr_sem;
790#endif
791
792 struct list_head i_orphan; /* unlinked but open inodes */
793
794 /*
795 * i_disksize keeps track of what the inode size is ON DISK, not
796 * in memory. During truncate, i_size is set to the new size by
797 * the VFS prior to calling ext4_truncate(), but the filesystem won't
798 * set i_disksize to 0 until the truncate is actually under way.
799 *
800 * The intent is that i_disksize always represents the blocks which
801 * are used by this file. This allows recovery to restart truncate
802 * on orphans if we crash during truncate. We actually write i_disksize
803 * into the on-disk inode when writing inodes out, instead of i_size.
804 *
805 * The only time when i_disksize and i_size may be different is when
806 * a truncate is in progress. The only things which change i_disksize
807 * are ext4_get_block (growth) and ext4_truncate (shrinkth).
808 */
809 loff_t i_disksize;
810
811 /*
812 * i_data_sem is for serialising ext4_truncate() against
813 * ext4_getblock(). In the 2.4 ext2 design, great chunks of inode's
814 * data tree are chopped off during truncate. We can't do that in
815 * ext4 because whenever we perform intermediate commits during
816 * truncate, the inode and all the metadata blocks *must* be in a
817 * consistent state which allows truncation of the orphans to restart
818 * during recovery. Hence we must fix the get_block-vs-truncate race
819 * by other means, so we have i_data_sem.
820 */
821 struct rw_semaphore i_data_sem;
822 struct inode vfs_inode;
823 struct jbd2_inode *jinode;
824
825 struct ext4_ext_cache i_cached_extent;
826 /*
827 * File creation time. Its function is same as that of
828 * struct timespec i_{a,c,m}time in the generic inode.
829 */
830 struct timespec i_crtime;
831
832 /* mballoc */
833 struct list_head i_prealloc_list;
834 spinlock_t i_prealloc_lock;
835
836 /* ialloc */
837 ext4_group_t i_last_alloc_group;
838
839 /* allocation reservation info for delalloc */
840 unsigned int i_reserved_data_blocks;
841 unsigned int i_reserved_meta_blocks;
842 unsigned int i_allocated_meta_blocks;
843 ext4_lblk_t i_da_metadata_calc_last_lblock;
844 int i_da_metadata_calc_len;
845
846 /* on-disk additional length */
847 __u16 i_extra_isize;
848
849#ifdef CONFIG_QUOTA
850 /* quota space reservation, managed internally by quota code */
851 qsize_t i_reserved_quota;
852#endif
853
854 /* completed IOs that might need unwritten extents handling */
855 struct list_head i_completed_io_list;
856 spinlock_t i_completed_io_lock;
857 atomic_t i_ioend_count; /* Number of outstanding io_end structs */
858 /* current io_end structure for async DIO write*/
859 ext4_io_end_t *cur_aio_dio;
860 atomic_t i_aiodio_unwritten; /* Nr. of inflight conversions pending */
861
862 spinlock_t i_block_reservation_lock;
863
864 /*
865 * Transactions that contain inode's metadata needed to complete
866 * fsync and fdatasync, respectively.
867 */
868 tid_t i_sync_tid;
869 tid_t i_datasync_tid;
870};
871
872/*
873 * File system states
874 */
875#define EXT4_VALID_FS 0x0001 /* Unmounted cleanly */
876#define EXT4_ERROR_FS 0x0002 /* Errors detected */
877#define EXT4_ORPHAN_FS 0x0004 /* Orphans being recovered */
878
879/*
880 * Misc. filesystem flags
881 */
882#define EXT2_FLAGS_SIGNED_HASH 0x0001 /* Signed dirhash in use */
883#define EXT2_FLAGS_UNSIGNED_HASH 0x0002 /* Unsigned dirhash in use */
884#define EXT2_FLAGS_TEST_FILESYS 0x0004 /* to test development code */
885
886/*
887 * Mount flags
888 */
889#define EXT4_MOUNT_OLDALLOC 0x00002 /* Don't use the new Orlov allocator */
890#define EXT4_MOUNT_GRPID 0x00004 /* Create files with directory's group */
891#define EXT4_MOUNT_DEBUG 0x00008 /* Some debugging messages */
892#define EXT4_MOUNT_ERRORS_CONT 0x00010 /* Continue on errors */
893#define EXT4_MOUNT_ERRORS_RO 0x00020 /* Remount fs ro on errors */
894#define EXT4_MOUNT_ERRORS_PANIC 0x00040 /* Panic on errors */
895#define EXT4_MOUNT_MINIX_DF 0x00080 /* Mimics the Minix statfs */
896#define EXT4_MOUNT_NOLOAD 0x00100 /* Don't use existing journal*/
897#define EXT4_MOUNT_DATA_FLAGS 0x00C00 /* Mode for data writes: */
898#define EXT4_MOUNT_JOURNAL_DATA 0x00400 /* Write data to journal */
899#define EXT4_MOUNT_ORDERED_DATA 0x00800 /* Flush data before commit */
900#define EXT4_MOUNT_WRITEBACK_DATA 0x00C00 /* No data ordering */
901#define EXT4_MOUNT_UPDATE_JOURNAL 0x01000 /* Update the journal format */
902#define EXT4_MOUNT_NO_UID32 0x02000 /* Disable 32-bit UIDs */
903#define EXT4_MOUNT_XATTR_USER 0x04000 /* Extended user attributes */
904#define EXT4_MOUNT_POSIX_ACL 0x08000 /* POSIX Access Control Lists */
905#define EXT4_MOUNT_NO_AUTO_DA_ALLOC 0x10000 /* No auto delalloc mapping */
906#define EXT4_MOUNT_BARRIER 0x20000 /* Use block barriers */
907#define EXT4_MOUNT_QUOTA 0x80000 /* Some quota option set */
908#define EXT4_MOUNT_USRQUOTA 0x100000 /* "old" user quota */
909#define EXT4_MOUNT_GRPQUOTA 0x200000 /* "old" group quota */
910#define EXT4_MOUNT_DIOREAD_NOLOCK 0x400000 /* Enable support for dio read nolocking */
911#define EXT4_MOUNT_JOURNAL_CHECKSUM 0x800000 /* Journal checksums */
912#define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
913#define EXT4_MOUNT_I_VERSION 0x2000000 /* i_version support */
914#define EXT4_MOUNT_MBLK_IO_SUBMIT 0x4000000 /* multi-block io submits */
915#define EXT4_MOUNT_DELALLOC 0x8000000 /* Delalloc support */
916#define EXT4_MOUNT_DATA_ERR_ABORT 0x10000000 /* Abort on file data write */
917#define EXT4_MOUNT_BLOCK_VALIDITY 0x20000000 /* Block validity checking */
918#define EXT4_MOUNT_DISCARD 0x40000000 /* Issue DISCARD requests */
919#define EXT4_MOUNT_INIT_INODE_TABLE 0x80000000 /* Initialize uninitialized itables */
920
921#define clear_opt(sb, opt) EXT4_SB(sb)->s_mount_opt &= \
922 ~EXT4_MOUNT_##opt
923#define set_opt(sb, opt) EXT4_SB(sb)->s_mount_opt |= \
924 EXT4_MOUNT_##opt
925#define test_opt(sb, opt) (EXT4_SB(sb)->s_mount_opt & \
926 EXT4_MOUNT_##opt)
927
928#define clear_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 &= \
929 ~EXT4_MOUNT2_##opt
930#define set_opt2(sb, opt) EXT4_SB(sb)->s_mount_opt2 |= \
931 EXT4_MOUNT2_##opt
932#define test_opt2(sb, opt) (EXT4_SB(sb)->s_mount_opt2 & \
933 EXT4_MOUNT2_##opt)
934
935#define ext4_set_bit __test_and_set_bit_le
936#define ext4_set_bit_atomic ext2_set_bit_atomic
937#define ext4_clear_bit __test_and_clear_bit_le
938#define ext4_clear_bit_atomic ext2_clear_bit_atomic
939#define ext4_test_bit test_bit_le
940#define ext4_find_first_zero_bit find_first_zero_bit_le
941#define ext4_find_next_zero_bit find_next_zero_bit_le
942#define ext4_find_next_bit find_next_bit_le
943
944extern void ext4_set_bits(void *bm, int cur, int len);
945
946/*
947 * Maximal mount counts between two filesystem checks
948 */
949#define EXT4_DFL_MAX_MNT_COUNT 20 /* Allow 20 mounts */
950#define EXT4_DFL_CHECKINTERVAL 0 /* Don't use interval check */
951
952/*
953 * Behaviour when detecting errors
954 */
955#define EXT4_ERRORS_CONTINUE 1 /* Continue execution */
956#define EXT4_ERRORS_RO 2 /* Remount fs read-only */
957#define EXT4_ERRORS_PANIC 3 /* Panic */
958#define EXT4_ERRORS_DEFAULT EXT4_ERRORS_CONTINUE
959
960/*
961 * Structure of the super block
962 */
963struct ext4_super_block {
964/*00*/ __le32 s_inodes_count; /* Inodes count */
965 __le32 s_blocks_count_lo; /* Blocks count */
966 __le32 s_r_blocks_count_lo; /* Reserved blocks count */
967 __le32 s_free_blocks_count_lo; /* Free blocks count */
968/*10*/ __le32 s_free_inodes_count; /* Free inodes count */
969 __le32 s_first_data_block; /* First Data Block */
970 __le32 s_log_block_size; /* Block size */
971 __le32 s_obso_log_frag_size; /* Obsoleted fragment size */
972/*20*/ __le32 s_blocks_per_group; /* # Blocks per group */
973 __le32 s_obso_frags_per_group; /* Obsoleted fragments per group */
974 __le32 s_inodes_per_group; /* # Inodes per group */
975 __le32 s_mtime; /* Mount time */
976/*30*/ __le32 s_wtime; /* Write time */
977 __le16 s_mnt_count; /* Mount count */
978 __le16 s_max_mnt_count; /* Maximal mount count */
979 __le16 s_magic; /* Magic signature */
980 __le16 s_state; /* File system state */
981 __le16 s_errors; /* Behaviour when detecting errors */
982 __le16 s_minor_rev_level; /* minor revision level */
983/*40*/ __le32 s_lastcheck; /* time of last check */
984 __le32 s_checkinterval; /* max. time between checks */
985 __le32 s_creator_os; /* OS */
986 __le32 s_rev_level; /* Revision level */
987/*50*/ __le16 s_def_resuid; /* Default uid for reserved blocks */
988 __le16 s_def_resgid; /* Default gid for reserved blocks */
989 /*
990 * These fields are for EXT4_DYNAMIC_REV superblocks only.
991 *
992 * Note: the difference between the compatible feature set and
993 * the incompatible feature set is that if there is a bit set
994 * in the incompatible feature set that the kernel doesn't
995 * know about, it should refuse to mount the filesystem.
996 *
997 * e2fsck's requirements are more strict; if it doesn't know
998 * about a feature in either the compatible or incompatible
999 * feature set, it must abort and not try to meddle with
1000 * things it doesn't understand...
1001 */
1002 __le32 s_first_ino; /* First non-reserved inode */
1003 __le16 s_inode_size; /* size of inode structure */
1004 __le16 s_block_group_nr; /* block group # of this superblock */
1005 __le32 s_feature_compat; /* compatible feature set */
1006/*60*/ __le32 s_feature_incompat; /* incompatible feature set */
1007 __le32 s_feature_ro_compat; /* readonly-compatible feature set */
1008/*68*/ __u8 s_uuid[16]; /* 128-bit uuid for volume */
1009/*78*/ char s_volume_name[16]; /* volume name */
1010/*88*/ char s_last_mounted[64]; /* directory where last mounted */
1011/*C8*/ __le32 s_algorithm_usage_bitmap; /* For compression */
1012 /*
1013 * Performance hints. Directory preallocation should only
1014 * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1015 */
1016 __u8 s_prealloc_blocks; /* Nr of blocks to try to preallocate*/
1017 __u8 s_prealloc_dir_blocks; /* Nr to preallocate for dirs */
1018 __le16 s_reserved_gdt_blocks; /* Per group desc for online growth */
1019 /*
1020 * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1021 */
1022/*D0*/ __u8 s_journal_uuid[16]; /* uuid of journal superblock */
1023/*E0*/ __le32 s_journal_inum; /* inode number of journal file */
1024 __le32 s_journal_dev; /* device number of journal file */
1025 __le32 s_last_orphan; /* start of list of inodes to delete */
1026 __le32 s_hash_seed[4]; /* HTREE hash seed */
1027 __u8 s_def_hash_version; /* Default hash version to use */
1028 __u8 s_jnl_backup_type;
1029 __le16 s_desc_size; /* size of group descriptor */
1030/*100*/ __le32 s_default_mount_opts;
1031 __le32 s_first_meta_bg; /* First metablock block group */
1032 __le32 s_mkfs_time; /* When the filesystem was created */
1033 __le32 s_jnl_blocks[17]; /* Backup of the journal inode */
1034 /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1035/*150*/ __le32 s_blocks_count_hi; /* Blocks count */
1036 __le32 s_r_blocks_count_hi; /* Reserved blocks count */
1037 __le32 s_free_blocks_count_hi; /* Free blocks count */
1038 __le16 s_min_extra_isize; /* All inodes have at least # bytes */
1039 __le16 s_want_extra_isize; /* New inodes should reserve # bytes */
1040 __le32 s_flags; /* Miscellaneous flags */
1041 __le16 s_raid_stride; /* RAID stride */
1042 __le16 s_mmp_update_interval; /* # seconds to wait in MMP checking */
1043 __le64 s_mmp_block; /* Block for multi-mount protection */
1044 __le32 s_raid_stripe_width; /* blocks on all data disks (N*stride)*/
1045 __u8 s_log_groups_per_flex; /* FLEX_BG group size */
1046 __u8 s_reserved_char_pad;
1047 __le16 s_reserved_pad;
1048 __le64 s_kbytes_written; /* nr of lifetime kilobytes written */
1049 __le32 s_snapshot_inum; /* Inode number of active snapshot */
1050 __le32 s_snapshot_id; /* sequential ID of active snapshot */
1051 __le64 s_snapshot_r_blocks_count; /* reserved blocks for active
1052 snapshot's future use */
1053 __le32 s_snapshot_list; /* inode number of the head of the
1054 on-disk snapshot list */
1055#define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1056 __le32 s_error_count; /* number of fs errors */
1057 __le32 s_first_error_time; /* first time an error happened */
1058 __le32 s_first_error_ino; /* inode involved in first error */
1059 __le64 s_first_error_block; /* block involved of first error */
1060 __u8 s_first_error_func[32]; /* function where the error happened */
1061 __le32 s_first_error_line; /* line number where error happened */
1062 __le32 s_last_error_time; /* most recent time of an error */
1063 __le32 s_last_error_ino; /* inode involved in last error */
1064 __le32 s_last_error_line; /* line number where error happened */
1065 __le64 s_last_error_block; /* block involved of last error */
1066 __u8 s_last_error_func[32]; /* function where the error happened */
1067#define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1068 __u8 s_mount_opts[64];
1069 __le32 s_reserved[112]; /* Padding to the end of the block */
1070};
1071
1072#define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1073
1074#ifdef __KERNEL__
1075
1076/*
1077 * run-time mount flags
1078 */
1079#define EXT4_MF_MNTDIR_SAMPLED 0x0001
1080#define EXT4_MF_FS_ABORTED 0x0002 /* Fatal error detected */
1081
1082/*
1083 * fourth extended-fs super-block data in memory
1084 */
1085struct ext4_sb_info {
1086 unsigned long s_desc_size; /* Size of a group descriptor in bytes */
1087 unsigned long s_inodes_per_block;/* Number of inodes per block */
1088 unsigned long s_blocks_per_group;/* Number of blocks in a group */
1089 unsigned long s_inodes_per_group;/* Number of inodes in a group */
1090 unsigned long s_itb_per_group; /* Number of inode table blocks per group */
1091 unsigned long s_gdb_count; /* Number of group descriptor blocks */
1092 unsigned long s_desc_per_block; /* Number of group descriptors per block */
1093 ext4_group_t s_groups_count; /* Number of groups in the fs */
1094 ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1095 unsigned long s_overhead_last; /* Last calculated overhead */
1096 unsigned long s_blocks_last; /* Last seen block count */
1097 loff_t s_bitmap_maxbytes; /* max bytes for bitmap files */
1098 struct buffer_head * s_sbh; /* Buffer containing the super block */
1099 struct ext4_super_block *s_es; /* Pointer to the super block in the buffer */
1100 struct buffer_head **s_group_desc;
1101 unsigned int s_mount_opt;
1102 unsigned int s_mount_opt2;
1103 unsigned int s_mount_flags;
1104 ext4_fsblk_t s_sb_block;
1105 uid_t s_resuid;
1106 gid_t s_resgid;
1107 unsigned short s_mount_state;
1108 unsigned short s_pad;
1109 int s_addr_per_block_bits;
1110 int s_desc_per_block_bits;
1111 int s_inode_size;
1112 int s_first_ino;
1113 unsigned int s_inode_readahead_blks;
1114 unsigned int s_inode_goal;
1115 spinlock_t s_next_gen_lock;
1116 u32 s_next_generation;
1117 u32 s_hash_seed[4];
1118 int s_def_hash_version;
1119 int s_hash_unsigned; /* 3 if hash should be signed, 0 if not */
1120 struct percpu_counter s_freeblocks_counter;
1121 struct percpu_counter s_freeinodes_counter;
1122 struct percpu_counter s_dirs_counter;
1123 struct percpu_counter s_dirtyblocks_counter;
1124 struct blockgroup_lock *s_blockgroup_lock;
1125 struct proc_dir_entry *s_proc;
1126 struct kobject s_kobj;
1127 struct completion s_kobj_unregister;
1128
1129 /* Journaling */
1130 struct journal_s *s_journal;
1131 struct list_head s_orphan;
1132 struct mutex s_orphan_lock;
1133 unsigned long s_resize_flags; /* Flags indicating if there
1134 is a resizer */
1135 unsigned long s_commit_interval;
1136 u32 s_max_batch_time;
1137 u32 s_min_batch_time;
1138 struct block_device *journal_bdev;
1139#ifdef CONFIG_JBD2_DEBUG
1140 struct timer_list turn_ro_timer; /* For turning read-only (crash simulation) */
1141 wait_queue_head_t ro_wait_queue; /* For people waiting for the fs to go read-only */
1142#endif
1143#ifdef CONFIG_QUOTA
1144 char *s_qf_names[MAXQUOTAS]; /* Names of quota files with journalled quota */
1145 int s_jquota_fmt; /* Format of quota to use */
1146#endif
1147 unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1148 struct rb_root system_blks;
1149
1150#ifdef EXTENTS_STATS
1151 /* ext4 extents stats */
1152 unsigned long s_ext_min;
1153 unsigned long s_ext_max;
1154 unsigned long s_depth_max;
1155 spinlock_t s_ext_stats_lock;
1156 unsigned long s_ext_blocks;
1157 unsigned long s_ext_extents;
1158#endif
1159 /* ext4 extent cache stats */
1160 unsigned long extent_cache_hits;
1161 unsigned long extent_cache_misses;
1162
1163 /* for buddy allocator */
1164 struct ext4_group_info ***s_group_info;
1165 struct inode *s_buddy_cache;
1166 spinlock_t s_md_lock;
1167 unsigned short *s_mb_offsets;
1168 unsigned int *s_mb_maxs;
1169
1170 /* tunables */
1171 unsigned long s_stripe;
1172 unsigned int s_mb_stream_request;
1173 unsigned int s_mb_max_to_scan;
1174 unsigned int s_mb_min_to_scan;
1175 unsigned int s_mb_stats;
1176 unsigned int s_mb_order2_reqs;
1177 unsigned int s_mb_group_prealloc;
1178 unsigned int s_max_writeback_mb_bump;
1179 /* where last allocation was done - for stream allocation */
1180 unsigned long s_mb_last_group;
1181 unsigned long s_mb_last_start;
1182
1183 /* stats for buddy allocator */
1184 atomic_t s_bal_reqs; /* number of reqs with len > 1 */
1185 atomic_t s_bal_success; /* we found long enough chunks */
1186 atomic_t s_bal_allocated; /* in blocks */
1187 atomic_t s_bal_ex_scanned; /* total extents scanned */
1188 atomic_t s_bal_goals; /* goal hits */
1189 atomic_t s_bal_breaks; /* too long searches */
1190 atomic_t s_bal_2orders; /* 2^order hits */
1191 spinlock_t s_bal_lock;
1192 unsigned long s_mb_buddies_generated;
1193 unsigned long long s_mb_generation_time;
1194 atomic_t s_mb_lost_chunks;
1195 atomic_t s_mb_preallocated;
1196 atomic_t s_mb_discarded;
1197 atomic_t s_lock_busy;
1198
1199 /* locality groups */
1200 struct ext4_locality_group __percpu *s_locality_groups;
1201
1202 /* for write statistics */
1203 unsigned long s_sectors_written_start;
1204 u64 s_kbytes_written;
1205
1206 unsigned int s_log_groups_per_flex;
1207 struct flex_groups *s_flex_groups;
1208
1209 /* workqueue for dio unwritten */
1210 struct workqueue_struct *dio_unwritten_wq;
1211
1212 /* timer for periodic error stats printing */
1213 struct timer_list s_err_report;
1214
1215 /* Lazy inode table initialization info */
1216 struct ext4_li_request *s_li_request;
1217 /* Wait multiplier for lazy initialization thread */
1218 unsigned int s_li_wait_mult;
1219
1220 /* Kernel thread for multiple mount protection */
1221 struct task_struct *s_mmp_tsk;
1222
1223 /* record the last minlen when FITRIM is called. */
1224 atomic_t s_last_trim_minblks;
1225};
1226
1227static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1228{
1229 return sb->s_fs_info;
1230}
1231static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1232{
1233 return container_of(inode, struct ext4_inode_info, vfs_inode);
1234}
1235
1236static inline struct timespec ext4_current_time(struct inode *inode)
1237{
1238 return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1239 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1240}
1241
1242static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1243{
1244 return ino == EXT4_ROOT_INO ||
1245 ino == EXT4_JOURNAL_INO ||
1246 ino == EXT4_RESIZE_INO ||
1247 (ino >= EXT4_FIRST_INO(sb) &&
1248 ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1249}
1250
1251/*
1252 * Inode dynamic state flags
1253 */
1254enum {
1255 EXT4_STATE_JDATA, /* journaled data exists */
1256 EXT4_STATE_NEW, /* inode is newly created */
1257 EXT4_STATE_XATTR, /* has in-inode xattrs */
1258 EXT4_STATE_NO_EXPAND, /* No space for expansion */
1259 EXT4_STATE_DA_ALLOC_CLOSE, /* Alloc DA blks on close */
1260 EXT4_STATE_EXT_MIGRATE, /* Inode is migrating */
1261 EXT4_STATE_DIO_UNWRITTEN, /* need convert on dio done*/
1262 EXT4_STATE_NEWENTRY, /* File just added to dir */
1263 EXT4_STATE_DELALLOC_RESERVED, /* blks already reserved for delalloc */
1264};
1265
1266#define EXT4_INODE_BIT_FNS(name, field, offset) \
1267static inline int ext4_test_inode_##name(struct inode *inode, int bit) \
1268{ \
1269 return test_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1270} \
1271static inline void ext4_set_inode_##name(struct inode *inode, int bit) \
1272{ \
1273 set_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1274} \
1275static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1276{ \
1277 clear_bit(bit + (offset), &EXT4_I(inode)->i_##field); \
1278}
1279
1280EXT4_INODE_BIT_FNS(flag, flags, 0)
1281#if (BITS_PER_LONG < 64)
1282EXT4_INODE_BIT_FNS(state, state_flags, 0)
1283
1284static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1285{
1286 (ei)->i_state_flags = 0;
1287}
1288#else
1289EXT4_INODE_BIT_FNS(state, flags, 32)
1290
1291static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1292{
1293 /* We depend on the fact that callers will set i_flags */
1294}
1295#endif
1296#else
1297/* Assume that user mode programs are passing in an ext4fs superblock, not
1298 * a kernel struct super_block. This will allow us to call the feature-test
1299 * macros from user land. */
1300#define EXT4_SB(sb) (sb)
1301#endif
1302
1303#define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1304
1305/*
1306 * Codes for operating systems
1307 */
1308#define EXT4_OS_LINUX 0
1309#define EXT4_OS_HURD 1
1310#define EXT4_OS_MASIX 2
1311#define EXT4_OS_FREEBSD 3
1312#define EXT4_OS_LITES 4
1313
1314/*
1315 * Revision levels
1316 */
1317#define EXT4_GOOD_OLD_REV 0 /* The good old (original) format */
1318#define EXT4_DYNAMIC_REV 1 /* V2 format w/ dynamic inode sizes */
1319
1320#define EXT4_CURRENT_REV EXT4_GOOD_OLD_REV
1321#define EXT4_MAX_SUPP_REV EXT4_DYNAMIC_REV
1322
1323#define EXT4_GOOD_OLD_INODE_SIZE 128
1324
1325/*
1326 * Feature set definitions
1327 */
1328
1329#define EXT4_HAS_COMPAT_FEATURE(sb,mask) \
1330 ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1331#define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask) \
1332 ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1333#define EXT4_HAS_INCOMPAT_FEATURE(sb,mask) \
1334 ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1335#define EXT4_SET_COMPAT_FEATURE(sb,mask) \
1336 EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1337#define EXT4_SET_RO_COMPAT_FEATURE(sb,mask) \
1338 EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1339#define EXT4_SET_INCOMPAT_FEATURE(sb,mask) \
1340 EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1341#define EXT4_CLEAR_COMPAT_FEATURE(sb,mask) \
1342 EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1343#define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask) \
1344 EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1345#define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask) \
1346 EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1347
1348#define EXT4_FEATURE_COMPAT_DIR_PREALLOC 0x0001
1349#define EXT4_FEATURE_COMPAT_IMAGIC_INODES 0x0002
1350#define EXT4_FEATURE_COMPAT_HAS_JOURNAL 0x0004
1351#define EXT4_FEATURE_COMPAT_EXT_ATTR 0x0008
1352#define EXT4_FEATURE_COMPAT_RESIZE_INODE 0x0010
1353#define EXT4_FEATURE_COMPAT_DIR_INDEX 0x0020
1354
1355#define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER 0x0001
1356#define EXT4_FEATURE_RO_COMPAT_LARGE_FILE 0x0002
1357#define EXT4_FEATURE_RO_COMPAT_BTREE_DIR 0x0004
1358#define EXT4_FEATURE_RO_COMPAT_HUGE_FILE 0x0008
1359#define EXT4_FEATURE_RO_COMPAT_GDT_CSUM 0x0010
1360#define EXT4_FEATURE_RO_COMPAT_DIR_NLINK 0x0020
1361#define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE 0x0040
1362#define EXT4_FEATURE_RO_COMPAT_QUOTA 0x0100
1363
1364#define EXT4_FEATURE_INCOMPAT_COMPRESSION 0x0001
1365#define EXT4_FEATURE_INCOMPAT_FILETYPE 0x0002
1366#define EXT4_FEATURE_INCOMPAT_RECOVER 0x0004 /* Needs recovery */
1367#define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 /* Journal device */
1368#define EXT4_FEATURE_INCOMPAT_META_BG 0x0010
1369#define EXT4_FEATURE_INCOMPAT_EXTENTS 0x0040 /* extents support */
1370#define EXT4_FEATURE_INCOMPAT_64BIT 0x0080
1371#define EXT4_FEATURE_INCOMPAT_MMP 0x0100
1372#define EXT4_FEATURE_INCOMPAT_FLEX_BG 0x0200
1373#define EXT4_FEATURE_INCOMPAT_EA_INODE 0x0400 /* EA in inode */
1374#define EXT4_FEATURE_INCOMPAT_DIRDATA 0x1000 /* data in dirent */
1375
1376#define EXT2_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR
1377#define EXT2_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1378 EXT4_FEATURE_INCOMPAT_META_BG)
1379#define EXT2_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1380 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1381 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1382
1383#define EXT3_FEATURE_COMPAT_SUPP EXT4_FEATURE_COMPAT_EXT_ATTR
1384#define EXT3_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1385 EXT4_FEATURE_INCOMPAT_RECOVER| \
1386 EXT4_FEATURE_INCOMPAT_META_BG)
1387#define EXT3_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1388 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1389 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1390
1391#define EXT4_FEATURE_COMPAT_SUPP EXT2_FEATURE_COMPAT_EXT_ATTR
1392#define EXT4_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1393 EXT4_FEATURE_INCOMPAT_RECOVER| \
1394 EXT4_FEATURE_INCOMPAT_META_BG| \
1395 EXT4_FEATURE_INCOMPAT_EXTENTS| \
1396 EXT4_FEATURE_INCOMPAT_64BIT| \
1397 EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1398 EXT4_FEATURE_INCOMPAT_MMP)
1399#define EXT4_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1400 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1401 EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1402 EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1403 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1404 EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1405 EXT4_FEATURE_RO_COMPAT_HUGE_FILE)
1406
1407/*
1408 * Default values for user and/or group using reserved blocks
1409 */
1410#define EXT4_DEF_RESUID 0
1411#define EXT4_DEF_RESGID 0
1412
1413#define EXT4_DEF_INODE_READAHEAD_BLKS 32
1414
1415/*
1416 * Default mount options
1417 */
1418#define EXT4_DEFM_DEBUG 0x0001
1419#define EXT4_DEFM_BSDGROUPS 0x0002
1420#define EXT4_DEFM_XATTR_USER 0x0004
1421#define EXT4_DEFM_ACL 0x0008
1422#define EXT4_DEFM_UID16 0x0010
1423#define EXT4_DEFM_JMODE 0x0060
1424#define EXT4_DEFM_JMODE_DATA 0x0020
1425#define EXT4_DEFM_JMODE_ORDERED 0x0040
1426#define EXT4_DEFM_JMODE_WBACK 0x0060
1427#define EXT4_DEFM_NOBARRIER 0x0100
1428#define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1429#define EXT4_DEFM_DISCARD 0x0400
1430#define EXT4_DEFM_NODELALLOC 0x0800
1431
1432/*
1433 * Default journal batch times
1434 */
1435#define EXT4_DEF_MIN_BATCH_TIME 0
1436#define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1437
1438/*
1439 * Minimum number of groups in a flexgroup before we separate out
1440 * directories into the first block group of a flexgroup
1441 */
1442#define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1443
1444/*
1445 * Structure of a directory entry
1446 */
1447#define EXT4_NAME_LEN 255
1448
1449struct ext4_dir_entry {
1450 __le32 inode; /* Inode number */
1451 __le16 rec_len; /* Directory entry length */
1452 __le16 name_len; /* Name length */
1453 char name[EXT4_NAME_LEN]; /* File name */
1454};
1455
1456/*
1457 * The new version of the directory entry. Since EXT4 structures are
1458 * stored in intel byte order, and the name_len field could never be
1459 * bigger than 255 chars, it's safe to reclaim the extra byte for the
1460 * file_type field.
1461 */
1462struct ext4_dir_entry_2 {
1463 __le32 inode; /* Inode number */
1464 __le16 rec_len; /* Directory entry length */
1465 __u8 name_len; /* Name length */
1466 __u8 file_type;
1467 char name[EXT4_NAME_LEN]; /* File name */
1468};
1469
1470/*
1471 * Ext4 directory file types. Only the low 3 bits are used. The
1472 * other bits are reserved for now.
1473 */
1474#define EXT4_FT_UNKNOWN 0
1475#define EXT4_FT_REG_FILE 1
1476#define EXT4_FT_DIR 2
1477#define EXT4_FT_CHRDEV 3
1478#define EXT4_FT_BLKDEV 4
1479#define EXT4_FT_FIFO 5
1480#define EXT4_FT_SOCK 6
1481#define EXT4_FT_SYMLINK 7
1482
1483#define EXT4_FT_MAX 8
1484
1485/*
1486 * EXT4_DIR_PAD defines the directory entries boundaries
1487 *
1488 * NOTE: It must be a multiple of 4
1489 */
1490#define EXT4_DIR_PAD 4
1491#define EXT4_DIR_ROUND (EXT4_DIR_PAD - 1)
1492#define EXT4_DIR_REC_LEN(name_len) (((name_len) + 8 + EXT4_DIR_ROUND) & \
1493 ~EXT4_DIR_ROUND)
1494#define EXT4_MAX_REC_LEN ((1<<16)-1)
1495
1496/*
1497 * If we ever get support for fs block sizes > page_size, we'll need
1498 * to remove the #if statements in the next two functions...
1499 */
1500static inline unsigned int
1501ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1502{
1503 unsigned len = le16_to_cpu(dlen);
1504
1505#if (PAGE_CACHE_SIZE >= 65536)
1506 if (len == EXT4_MAX_REC_LEN || len == 0)
1507 return blocksize;
1508 return (len & 65532) | ((len & 3) << 16);
1509#else
1510 return len;
1511#endif
1512}
1513
1514static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1515{
1516 if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1517 BUG();
1518#if (PAGE_CACHE_SIZE >= 65536)
1519 if (len < 65536)
1520 return cpu_to_le16(len);
1521 if (len == blocksize) {
1522 if (blocksize == 65536)
1523 return cpu_to_le16(EXT4_MAX_REC_LEN);
1524 else
1525 return cpu_to_le16(0);
1526 }
1527 return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1528#else
1529 return cpu_to_le16(len);
1530#endif
1531}
1532
1533/*
1534 * Hash Tree Directory indexing
1535 * (c) Daniel Phillips, 2001
1536 */
1537
1538#define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1539 EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1540 ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1541#define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1542#define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1543
1544/* Legal values for the dx_root hash_version field: */
1545
1546#define DX_HASH_LEGACY 0
1547#define DX_HASH_HALF_MD4 1
1548#define DX_HASH_TEA 2
1549#define DX_HASH_LEGACY_UNSIGNED 3
1550#define DX_HASH_HALF_MD4_UNSIGNED 4
1551#define DX_HASH_TEA_UNSIGNED 5
1552
1553#ifdef __KERNEL__
1554
1555/* hash info structure used by the directory hash */
1556struct dx_hash_info
1557{
1558 u32 hash;
1559 u32 minor_hash;
1560 int hash_version;
1561 u32 *seed;
1562};
1563
1564#define EXT4_HTREE_EOF 0x7fffffff
1565
1566/*
1567 * Control parameters used by ext4_htree_next_block
1568 */
1569#define HASH_NB_ALWAYS 1
1570
1571
1572/*
1573 * Describe an inode's exact location on disk and in memory
1574 */
1575struct ext4_iloc
1576{
1577 struct buffer_head *bh;
1578 unsigned long offset;
1579 ext4_group_t block_group;
1580};
1581
1582static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1583{
1584 return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1585}
1586
1587/*
1588 * This structure is stuffed into the struct file's private_data field
1589 * for directories. It is where we put information so that we can do
1590 * readdir operations in hash tree order.
1591 */
1592struct dir_private_info {
1593 struct rb_root root;
1594 struct rb_node *curr_node;
1595 struct fname *extra_fname;
1596 loff_t last_pos;
1597 __u32 curr_hash;
1598 __u32 curr_minor_hash;
1599 __u32 next_hash;
1600};
1601
1602/* calculate the first block number of the group */
1603static inline ext4_fsblk_t
1604ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1605{
1606 return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1607 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1608}
1609
1610/*
1611 * Special error return code only used by dx_probe() and its callers.
1612 */
1613#define ERR_BAD_DX_DIR -75000
1614
1615void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1616 ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1617
1618/*
1619 * Timeout and state flag for lazy initialization inode thread.
1620 */
1621#define EXT4_DEF_LI_WAIT_MULT 10
1622#define EXT4_DEF_LI_MAX_START_DELAY 5
1623#define EXT4_LAZYINIT_QUIT 0x0001
1624#define EXT4_LAZYINIT_RUNNING 0x0002
1625
1626/*
1627 * Lazy inode table initialization info
1628 */
1629struct ext4_lazy_init {
1630 unsigned long li_state;
1631 struct list_head li_request_list;
1632 struct mutex li_list_mtx;
1633};
1634
1635struct ext4_li_request {
1636 struct super_block *lr_super;
1637 struct ext4_sb_info *lr_sbi;
1638 ext4_group_t lr_next_group;
1639 struct list_head lr_request;
1640 unsigned long lr_next_sched;
1641 unsigned long lr_timeout;
1642};
1643
1644struct ext4_features {
1645 struct kobject f_kobj;
1646 struct completion f_kobj_unregister;
1647};
1648
1649/*
1650 * This structure will be used for multiple mount protection. It will be
1651 * written into the block number saved in the s_mmp_block field in the
1652 * superblock. Programs that check MMP should assume that if
1653 * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1654 * to use the filesystem, regardless of how old the timestamp is.
1655 */
1656#define EXT4_MMP_MAGIC 0x004D4D50U /* ASCII for MMP */
1657#define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1658#define EXT4_MMP_SEQ_FSCK 0xE24D4D50U /* mmp_seq value when being fscked */
1659#define EXT4_MMP_SEQ_MAX 0xE24D4D4FU /* maximum valid mmp_seq value */
1660
1661struct mmp_struct {
1662 __le32 mmp_magic; /* Magic number for MMP */
1663 __le32 mmp_seq; /* Sequence no. updated periodically */
1664
1665 /*
1666 * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1667 * purposes and do not affect the correctness of the algorithm
1668 */
1669 __le64 mmp_time; /* Time last updated */
1670 char mmp_nodename[64]; /* Node which last updated MMP block */
1671 char mmp_bdevname[32]; /* Bdev which last updated MMP block */
1672
1673 /*
1674 * mmp_check_interval is used to verify if the MMP block has been
1675 * updated on the block device. The value is updated based on the
1676 * maximum time to write the MMP block during an update cycle.
1677 */
1678 __le16 mmp_check_interval;
1679
1680 __le16 mmp_pad1;
1681 __le32 mmp_pad2[227];
1682};
1683
1684/* arguments passed to the mmp thread */
1685struct mmpd_data {
1686 struct buffer_head *bh; /* bh from initial read_mmp_block() */
1687 struct super_block *sb; /* super block of the fs */
1688};
1689
1690/*
1691 * Check interval multiplier
1692 * The MMP block is written every update interval and initially checked every
1693 * update interval x the multiplier (the value is then adapted based on the
1694 * write latency). The reason is that writes can be delayed under load and we
1695 * don't want readers to incorrectly assume that the filesystem is no longer
1696 * in use.
1697 */
1698#define EXT4_MMP_CHECK_MULT 2UL
1699
1700/*
1701 * Minimum interval for MMP checking in seconds.
1702 */
1703#define EXT4_MMP_MIN_CHECK_INTERVAL 5UL
1704
1705/*
1706 * Maximum interval for MMP checking in seconds.
1707 */
1708#define EXT4_MMP_MAX_CHECK_INTERVAL 300UL
1709
1710/*
1711 * Function prototypes
1712 */
1713
1714/*
1715 * Ok, these declarations are also in <linux/kernel.h> but none of the
1716 * ext4 source programs needs to include it so they are duplicated here.
1717 */
1718# define NORET_TYPE /**/
1719# define ATTRIB_NORET __attribute__((noreturn))
1720# define NORET_AND noreturn,
1721
1722/* bitmap.c */
1723extern unsigned int ext4_count_free(struct buffer_head *, unsigned);
1724
1725/* balloc.c */
1726extern unsigned int ext4_block_group(struct super_block *sb,
1727 ext4_fsblk_t blocknr);
1728extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1729 ext4_fsblk_t blocknr);
1730extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1731extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1732 ext4_group_t group);
1733extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1734 ext4_fsblk_t goal,
1735 unsigned int flags,
1736 unsigned long *count,
1737 int *errp);
1738extern int ext4_claim_free_blocks(struct ext4_sb_info *sbi,
1739 s64 nblocks, unsigned int flags);
1740extern ext4_fsblk_t ext4_count_free_blocks(struct super_block *);
1741extern void ext4_check_blocks_bitmap(struct super_block *);
1742extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1743 ext4_group_t block_group,
1744 struct buffer_head ** bh);
1745extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1746struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1747 ext4_group_t block_group);
1748extern unsigned ext4_init_block_bitmap(struct super_block *sb,
1749 struct buffer_head *bh,
1750 ext4_group_t group,
1751 struct ext4_group_desc *desc);
1752#define ext4_free_blocks_after_init(sb, group, desc) \
1753 ext4_init_block_bitmap(sb, NULL, group, desc)
1754ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
1755
1756/* dir.c */
1757extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1758 struct file *,
1759 struct ext4_dir_entry_2 *,
1760 struct buffer_head *, unsigned int);
1761#define ext4_check_dir_entry(dir, filp, de, bh, offset) \
1762 unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
1763 (de), (bh), (offset)))
1764extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1765 __u32 minor_hash,
1766 struct ext4_dir_entry_2 *dirent);
1767extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1768
1769/* fsync.c */
1770extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
1771extern int ext4_flush_completed_IO(struct inode *);
1772
1773/* hash.c */
1774extern int ext4fs_dirhash(const char *name, int len, struct
1775 dx_hash_info *hinfo);
1776
1777/* ialloc.c */
1778extern struct inode *ext4_new_inode(handle_t *, struct inode *, int,
1779 const struct qstr *qstr, __u32 goal);
1780extern void ext4_free_inode(handle_t *, struct inode *);
1781extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1782extern unsigned long ext4_count_free_inodes(struct super_block *);
1783extern unsigned long ext4_count_dirs(struct super_block *);
1784extern void ext4_check_inodes_bitmap(struct super_block *);
1785extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1786extern int ext4_init_inode_table(struct super_block *sb,
1787 ext4_group_t group, int barrier);
1788
1789/* mballoc.c */
1790extern long ext4_mb_stats;
1791extern long ext4_mb_max_to_scan;
1792extern int ext4_mb_init(struct super_block *, int);
1793extern int ext4_mb_release(struct super_block *);
1794extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1795 struct ext4_allocation_request *, int *);
1796extern int ext4_mb_reserve_blocks(struct super_block *, int);
1797extern void ext4_discard_preallocations(struct inode *);
1798extern int __init ext4_init_mballoc(void);
1799extern void ext4_exit_mballoc(void);
1800extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1801 struct buffer_head *bh, ext4_fsblk_t block,
1802 unsigned long count, int flags);
1803extern int ext4_mb_add_groupinfo(struct super_block *sb,
1804 ext4_group_t i, struct ext4_group_desc *desc);
1805extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
1806 ext4_fsblk_t block, unsigned long count);
1807extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
1808
1809/* inode.c */
1810struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1811 ext4_lblk_t, int, int *);
1812struct buffer_head *ext4_bread(handle_t *, struct inode *,
1813 ext4_lblk_t, int, int *);
1814int ext4_get_block(struct inode *inode, sector_t iblock,
1815 struct buffer_head *bh_result, int create);
1816
1817extern struct inode *ext4_iget(struct super_block *, unsigned long);
1818extern int ext4_write_inode(struct inode *, struct writeback_control *);
1819extern int ext4_setattr(struct dentry *, struct iattr *);
1820extern int ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1821 struct kstat *stat);
1822extern void ext4_evict_inode(struct inode *);
1823extern void ext4_clear_inode(struct inode *);
1824extern int ext4_sync_inode(handle_t *, struct inode *);
1825extern void ext4_dirty_inode(struct inode *, int);
1826extern int ext4_change_inode_journal_flag(struct inode *, int);
1827extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1828extern int ext4_can_truncate(struct inode *inode);
1829extern void ext4_truncate(struct inode *);
1830extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length);
1831extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
1832extern void ext4_set_inode_flags(struct inode *);
1833extern void ext4_get_inode_flags(struct ext4_inode_info *);
1834extern int ext4_alloc_da_blocks(struct inode *inode);
1835extern void ext4_set_aops(struct inode *inode);
1836extern int ext4_writepage_trans_blocks(struct inode *);
1837extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1838extern int ext4_block_truncate_page(handle_t *handle,
1839 struct address_space *mapping, loff_t from);
1840extern int ext4_block_zero_page_range(handle_t *handle,
1841 struct address_space *mapping, loff_t from, loff_t length);
1842extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1843extern qsize_t *ext4_get_reserved_space(struct inode *inode);
1844extern void ext4_da_update_reserve_space(struct inode *inode,
1845 int used, int quota_claim);
1846
1847/* indirect.c */
1848extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
1849 struct ext4_map_blocks *map, int flags);
1850extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
1851 const struct iovec *iov, loff_t offset,
1852 unsigned long nr_segs);
1853extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
1854extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks, int chunk);
1855extern void ext4_ind_truncate(struct inode *inode);
1856
1857/* ioctl.c */
1858extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1859extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1860
1861/* migrate.c */
1862extern int ext4_ext_migrate(struct inode *);
1863
1864/* namei.c */
1865extern int ext4_orphan_add(handle_t *, struct inode *);
1866extern int ext4_orphan_del(handle_t *, struct inode *);
1867extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1868 __u32 start_minor_hash, __u32 *next_hash);
1869
1870/* resize.c */
1871extern int ext4_group_add(struct super_block *sb,
1872 struct ext4_new_group_data *input);
1873extern int ext4_group_extend(struct super_block *sb,
1874 struct ext4_super_block *es,
1875 ext4_fsblk_t n_blocks_count);
1876
1877/* super.c */
1878extern void *ext4_kvmalloc(size_t size, gfp_t flags);
1879extern void *ext4_kvzalloc(size_t size, gfp_t flags);
1880extern void ext4_kvfree(void *ptr);
1881extern void __ext4_error(struct super_block *, const char *, unsigned int,
1882 const char *, ...)
1883 __attribute__ ((format (printf, 4, 5)));
1884#define ext4_error(sb, message...) __ext4_error(sb, __func__, \
1885 __LINE__, ## message)
1886extern void ext4_error_inode(struct inode *, const char *, unsigned int,
1887 ext4_fsblk_t, const char *, ...)
1888 __attribute__ ((format (printf, 5, 6)));
1889extern void ext4_error_file(struct file *, const char *, unsigned int,
1890 ext4_fsblk_t, const char *, ...)
1891 __attribute__ ((format (printf, 5, 6)));
1892extern void __ext4_std_error(struct super_block *, const char *,
1893 unsigned int, int);
1894extern void __ext4_abort(struct super_block *, const char *, unsigned int,
1895 const char *, ...)
1896 __attribute__ ((format (printf, 4, 5)));
1897#define ext4_abort(sb, message...) __ext4_abort(sb, __func__, \
1898 __LINE__, ## message)
1899extern void __ext4_warning(struct super_block *, const char *, unsigned int,
1900 const char *, ...)
1901 __attribute__ ((format (printf, 4, 5)));
1902#define ext4_warning(sb, message...) __ext4_warning(sb, __func__, \
1903 __LINE__, ## message)
1904extern void ext4_msg(struct super_block *, const char *, const char *, ...)
1905 __attribute__ ((format (printf, 3, 4)));
1906extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
1907 const char *, unsigned int, const char *);
1908#define dump_mmp_msg(sb, mmp, msg) __dump_mmp_msg(sb, mmp, __func__, \
1909 __LINE__, msg)
1910extern void __ext4_grp_locked_error(const char *, unsigned int, \
1911 struct super_block *, ext4_group_t, \
1912 unsigned long, ext4_fsblk_t, \
1913 const char *, ...)
1914 __attribute__ ((format (printf, 7, 8)));
1915#define ext4_grp_locked_error(sb, grp, message...) \
1916 __ext4_grp_locked_error(__func__, __LINE__, (sb), (grp), ## message)
1917extern void ext4_update_dynamic_rev(struct super_block *sb);
1918extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1919 __u32 compat);
1920extern int ext4_update_rocompat_feature(handle_t *handle,
1921 struct super_block *sb, __u32 rocompat);
1922extern int ext4_update_incompat_feature(handle_t *handle,
1923 struct super_block *sb, __u32 incompat);
1924extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1925 struct ext4_group_desc *bg);
1926extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1927 struct ext4_group_desc *bg);
1928extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1929 struct ext4_group_desc *bg);
1930extern __u32 ext4_free_blks_count(struct super_block *sb,
1931 struct ext4_group_desc *bg);
1932extern __u32 ext4_free_inodes_count(struct super_block *sb,
1933 struct ext4_group_desc *bg);
1934extern __u32 ext4_used_dirs_count(struct super_block *sb,
1935 struct ext4_group_desc *bg);
1936extern __u32 ext4_itable_unused_count(struct super_block *sb,
1937 struct ext4_group_desc *bg);
1938extern void ext4_block_bitmap_set(struct super_block *sb,
1939 struct ext4_group_desc *bg, ext4_fsblk_t blk);
1940extern void ext4_inode_bitmap_set(struct super_block *sb,
1941 struct ext4_group_desc *bg, ext4_fsblk_t blk);
1942extern void ext4_inode_table_set(struct super_block *sb,
1943 struct ext4_group_desc *bg, ext4_fsblk_t blk);
1944extern void ext4_free_blks_set(struct super_block *sb,
1945 struct ext4_group_desc *bg, __u32 count);
1946extern void ext4_free_inodes_set(struct super_block *sb,
1947 struct ext4_group_desc *bg, __u32 count);
1948extern void ext4_used_dirs_set(struct super_block *sb,
1949 struct ext4_group_desc *bg, __u32 count);
1950extern void ext4_itable_unused_set(struct super_block *sb,
1951 struct ext4_group_desc *bg, __u32 count);
1952extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
1953 struct ext4_group_desc *gdp);
1954extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
1955 struct ext4_group_desc *gdp);
1956
1957static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
1958{
1959 return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
1960 le32_to_cpu(es->s_blocks_count_lo);
1961}
1962
1963static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
1964{
1965 return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
1966 le32_to_cpu(es->s_r_blocks_count_lo);
1967}
1968
1969static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
1970{
1971 return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
1972 le32_to_cpu(es->s_free_blocks_count_lo);
1973}
1974
1975static inline void ext4_blocks_count_set(struct ext4_super_block *es,
1976 ext4_fsblk_t blk)
1977{
1978 es->s_blocks_count_lo = cpu_to_le32((u32)blk);
1979 es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
1980}
1981
1982static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
1983 ext4_fsblk_t blk)
1984{
1985 es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
1986 es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
1987}
1988
1989static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
1990 ext4_fsblk_t blk)
1991{
1992 es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
1993 es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
1994}
1995
1996static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
1997{
1998 if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
1999 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2000 le32_to_cpu(raw_inode->i_size_lo);
2001 else
2002 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2003}
2004
2005static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2006{
2007 raw_inode->i_size_lo = cpu_to_le32(i_size);
2008 raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2009}
2010
2011static inline
2012struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2013 ext4_group_t group)
2014{
2015 struct ext4_group_info ***grp_info;
2016 long indexv, indexh;
2017 grp_info = EXT4_SB(sb)->s_group_info;
2018 indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2019 indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2020 return grp_info[indexv][indexh];
2021}
2022
2023/*
2024 * Reading s_groups_count requires using smp_rmb() afterwards. See
2025 * the locking protocol documented in the comments of ext4_group_add()
2026 * in resize.c
2027 */
2028static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2029{
2030 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2031
2032 smp_rmb();
2033 return ngroups;
2034}
2035
2036static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2037 ext4_group_t block_group)
2038{
2039 return block_group >> sbi->s_log_groups_per_flex;
2040}
2041
2042static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2043{
2044 return 1 << sbi->s_log_groups_per_flex;
2045}
2046
2047#define ext4_std_error(sb, errno) \
2048do { \
2049 if ((errno)) \
2050 __ext4_std_error((sb), __func__, __LINE__, (errno)); \
2051} while (0)
2052
2053#ifdef CONFIG_SMP
2054/* Each CPU can accumulate percpu_counter_batch blocks in their local
2055 * counters. So we need to make sure we have free blocks more
2056 * than percpu_counter_batch * nr_cpu_ids. Also add a window of 4 times.
2057 */
2058#define EXT4_FREEBLOCKS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2059#else
2060#define EXT4_FREEBLOCKS_WATERMARK 0
2061#endif
2062
2063static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2064{
2065 /*
2066 * XXX: replace with spinlock if seen contended -bzzz
2067 */
2068 down_write(&EXT4_I(inode)->i_data_sem);
2069 if (newsize > EXT4_I(inode)->i_disksize)
2070 EXT4_I(inode)->i_disksize = newsize;
2071 up_write(&EXT4_I(inode)->i_data_sem);
2072 return ;
2073}
2074
2075struct ext4_group_info {
2076 unsigned long bb_state;
2077 struct rb_root bb_free_root;
2078 ext4_grpblk_t bb_first_free; /* first free block */
2079 ext4_grpblk_t bb_free; /* total free blocks */
2080 ext4_grpblk_t bb_fragments; /* nr of freespace fragments */
2081 ext4_grpblk_t bb_largest_free_order;/* order of largest frag in BG */
2082 struct list_head bb_prealloc_list;
2083#ifdef DOUBLE_CHECK
2084 void *bb_bitmap;
2085#endif
2086 struct rw_semaphore alloc_sem;
2087 ext4_grpblk_t bb_counters[]; /* Nr of free power-of-two-block
2088 * regions, index is order.
2089 * bb_counters[3] = 5 means
2090 * 5 free 8-block regions. */
2091};
2092
2093#define EXT4_GROUP_INFO_NEED_INIT_BIT 0
2094#define EXT4_GROUP_INFO_WAS_TRIMMED_BIT 1
2095
2096#define EXT4_MB_GRP_NEED_INIT(grp) \
2097 (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2098
2099#define EXT4_MB_GRP_WAS_TRIMMED(grp) \
2100 (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2101#define EXT4_MB_GRP_SET_TRIMMED(grp) \
2102 (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2103#define EXT4_MB_GRP_CLEAR_TRIMMED(grp) \
2104 (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2105
2106#define EXT4_MAX_CONTENTION 8
2107#define EXT4_CONTENTION_THRESHOLD 2
2108
2109static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2110 ext4_group_t group)
2111{
2112 return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2113}
2114
2115/*
2116 * Returns true if the filesystem is busy enough that attempts to
2117 * access the block group locks has run into contention.
2118 */
2119static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2120{
2121 return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2122}
2123
2124static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2125{
2126 spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2127 if (spin_trylock(lock))
2128 /*
2129 * We're able to grab the lock right away, so drop the
2130 * lock contention counter.
2131 */
2132 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2133 else {
2134 /*
2135 * The lock is busy, so bump the contention counter,
2136 * and then wait on the spin lock.
2137 */
2138 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2139 EXT4_MAX_CONTENTION);
2140 spin_lock(lock);
2141 }
2142}
2143
2144static inline void ext4_unlock_group(struct super_block *sb,
2145 ext4_group_t group)
2146{
2147 spin_unlock(ext4_group_lock_ptr(sb, group));
2148}
2149
2150static inline void ext4_mark_super_dirty(struct super_block *sb)
2151{
2152 if (EXT4_SB(sb)->s_journal == NULL)
2153 sb->s_dirt =1;
2154}
2155
2156/*
2157 * Block validity checking
2158 */
2159#define ext4_check_indirect_blockref(inode, bh) \
2160 ext4_check_blockref(__func__, __LINE__, inode, \
2161 (__le32 *)(bh)->b_data, \
2162 EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2163
2164#define ext4_ind_check_inode(inode) \
2165 ext4_check_blockref(__func__, __LINE__, inode, \
2166 EXT4_I(inode)->i_data, \
2167 EXT4_NDIR_BLOCKS)
2168
2169/*
2170 * Inodes and files operations
2171 */
2172
2173/* dir.c */
2174extern const struct file_operations ext4_dir_operations;
2175
2176/* file.c */
2177extern const struct inode_operations ext4_file_inode_operations;
2178extern const struct file_operations ext4_file_operations;
2179extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2180
2181/* namei.c */
2182extern const struct inode_operations ext4_dir_inode_operations;
2183extern const struct inode_operations ext4_special_inode_operations;
2184extern struct dentry *ext4_get_parent(struct dentry *child);
2185
2186/* symlink.c */
2187extern const struct inode_operations ext4_symlink_inode_operations;
2188extern const struct inode_operations ext4_fast_symlink_inode_operations;
2189
2190/* block_validity */
2191extern void ext4_release_system_zone(struct super_block *sb);
2192extern int ext4_setup_system_zone(struct super_block *sb);
2193extern int __init ext4_init_system_zone(void);
2194extern void ext4_exit_system_zone(void);
2195extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2196 ext4_fsblk_t start_blk,
2197 unsigned int count);
2198extern int ext4_check_blockref(const char *, unsigned int,
2199 struct inode *, __le32 *, unsigned int);
2200
2201/* extents.c */
2202extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2203extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2204extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
2205 int chunk);
2206extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2207 struct ext4_map_blocks *map, int flags);
2208extern void ext4_ext_truncate(struct inode *);
2209extern int ext4_ext_punch_hole(struct file *file, loff_t offset,
2210 loff_t length);
2211extern void ext4_ext_init(struct super_block *);
2212extern void ext4_ext_release(struct super_block *);
2213extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2214 loff_t len);
2215extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
2216 ssize_t len);
2217extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2218 struct ext4_map_blocks *map, int flags);
2219extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2220 __u64 start, __u64 len);
2221/* move_extent.c */
2222extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2223 __u64 start_orig, __u64 start_donor,
2224 __u64 len, __u64 *moved_len);
2225
2226/* page-io.c */
2227extern int __init ext4_init_pageio(void);
2228extern void ext4_exit_pageio(void);
2229extern void ext4_ioend_wait(struct inode *);
2230extern void ext4_free_io_end(ext4_io_end_t *io);
2231extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2232extern int ext4_end_io_nolock(ext4_io_end_t *io);
2233extern void ext4_io_submit(struct ext4_io_submit *io);
2234extern int ext4_bio_write_page(struct ext4_io_submit *io,
2235 struct page *page,
2236 int len,
2237 struct writeback_control *wbc);
2238
2239/* mmp.c */
2240extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2241
2242/* BH_Uninit flag: blocks are allocated but uninitialized on disk */
2243enum ext4_state_bits {
2244 BH_Uninit /* blocks are allocated but uninitialized on disk */
2245 = BH_JBDPrivateStart,
2246};
2247
2248BUFFER_FNS(Uninit, uninit)
2249TAS_BUFFER_FNS(Uninit, uninit)
2250
2251/*
2252 * Add new method to test wether block and inode bitmaps are properly
2253 * initialized. With uninit_bg reading the block from disk is not enough
2254 * to mark the bitmap uptodate. We need to also zero-out the bitmap
2255 */
2256#define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2257
2258static inline int bitmap_uptodate(struct buffer_head *bh)
2259{
2260 return (buffer_uptodate(bh) &&
2261 test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2262}
2263static inline void set_bitmap_uptodate(struct buffer_head *bh)
2264{
2265 set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2266}
2267
2268#define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2269
2270/* For ioend & aio unwritten conversion wait queues */
2271#define EXT4_WQ_HASH_SZ 37
2272#define ext4_ioend_wq(v) (&ext4__ioend_wq[((unsigned long)(v)) %\
2273 EXT4_WQ_HASH_SZ])
2274#define ext4_aio_mutex(v) (&ext4__aio_mutex[((unsigned long)(v)) %\
2275 EXT4_WQ_HASH_SZ])
2276extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2277extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2278
2279#define EXT4_RESIZING 0
2280extern int ext4_resize_begin(struct super_block *sb);
2281extern void ext4_resize_end(struct super_block *sb);
2282
2283#endif /* __KERNEL__ */
2284
2285#endif /* _EXT4_H */