Merge branch 'for-linus' of git://git.kernel.dk/linux-block
[linux-2.6.git] / fs / ext4 / 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 */
69 typedef int ext4_grpblk_t;
70
71 /* data type for filesystem-wide blocks number */
72 typedef unsigned long long ext4_fsblk_t;
73
74 /* data type for file logical block number */
75 typedef __u32 ext4_lblk_t;
76
77 /* data type for block group number */
78 typedef 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
114 struct 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 /* Sometimes (in the bigalloc case, from ext4_da_get_block_prep) the caller of
148  * ext4_map_blocks wants to know whether or not the underlying cluster has
149  * already been accounted for. EXT4_MAP_FROM_CLUSTER conveys to the caller that
150  * the requested mapping was from previously mapped (or delayed allocated)
151  * cluster. We use BH_AllocFromCluster only for this flag. BH_AllocFromCluster
152  * should never appear on buffer_head's state flags.
153  */
154 #define EXT4_MAP_FROM_CLUSTER   (1 << BH_AllocFromCluster)
155 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
156                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
157                                  EXT4_MAP_UNINIT | EXT4_MAP_FROM_CLUSTER)
158
159 struct ext4_map_blocks {
160         ext4_fsblk_t m_pblk;
161         ext4_lblk_t m_lblk;
162         unsigned int m_len;
163         unsigned int m_flags;
164 };
165
166 /*
167  * For delayed allocation tracking
168  */
169 struct mpage_da_data {
170         struct inode *inode;
171         sector_t b_blocknr;             /* start block number of extent */
172         size_t b_size;                  /* size of extent */
173         unsigned long b_state;          /* state of the extent */
174         unsigned long first_page, next_page;    /* extent of pages */
175         struct writeback_control *wbc;
176         int io_done;
177         int pages_written;
178         int retval;
179 };
180
181 /*
182  * Flags for ext4_io_end->flags
183  */
184 #define EXT4_IO_END_UNWRITTEN   0x0001
185 #define EXT4_IO_END_ERROR       0x0002
186 #define EXT4_IO_END_QUEUED      0x0004
187
188 struct ext4_io_page {
189         struct page     *p_page;
190         atomic_t        p_count;
191 };
192
193 #define MAX_IO_PAGES 128
194
195 typedef struct ext4_io_end {
196         struct list_head        list;           /* per-file finished IO list */
197         struct inode            *inode;         /* file being written to */
198         unsigned int            flag;           /* unwritten or not */
199         struct page             *page;          /* page struct for buffer write */
200         loff_t                  offset;         /* offset in the file */
201         ssize_t                 size;           /* size of the extent */
202         struct work_struct      work;           /* data work queue */
203         struct kiocb            *iocb;          /* iocb struct for AIO */
204         int                     result;         /* error value for AIO */
205         int                     num_io_pages;
206         struct ext4_io_page     *pages[MAX_IO_PAGES];
207 } ext4_io_end_t;
208
209 struct ext4_io_submit {
210         int                     io_op;
211         struct bio              *io_bio;
212         ext4_io_end_t           *io_end;
213         struct ext4_io_page     *io_page;
214         sector_t                io_next_block;
215 };
216
217 /*
218  * Special inodes numbers
219  */
220 #define EXT4_BAD_INO             1      /* Bad blocks inode */
221 #define EXT4_ROOT_INO            2      /* Root inode */
222 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
223 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
224 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
225 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
226 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
227 #define EXT4_JOURNAL_INO         8      /* Journal inode */
228
229 /* First non-reserved inode for old ext4 filesystems */
230 #define EXT4_GOOD_OLD_FIRST_INO 11
231
232 /*
233  * Maximal count of links to a file
234  */
235 #define EXT4_LINK_MAX           65000
236
237 /*
238  * Macro-instructions used to manage several block sizes
239  */
240 #define EXT4_MIN_BLOCK_SIZE             1024
241 #define EXT4_MAX_BLOCK_SIZE             65536
242 #define EXT4_MIN_BLOCK_LOG_SIZE         10
243 #define EXT4_MAX_BLOCK_LOG_SIZE         16
244 #ifdef __KERNEL__
245 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
246 #else
247 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
248 #endif
249 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
250 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
251                                          EXT4_SB(s)->s_cluster_bits)
252 #ifdef __KERNEL__
253 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
254 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
255 #else
256 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
257 #endif
258 #ifdef __KERNEL__
259 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
260 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
261 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
262 #else
263 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
264                                  EXT4_GOOD_OLD_INODE_SIZE : \
265                                  (s)->s_inode_size)
266 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
267                                  EXT4_GOOD_OLD_FIRST_INO : \
268                                  (s)->s_first_ino)
269 #endif
270 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
271
272 /* Translate a block number to a cluster number */
273 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
274 /* Translate a cluster number to a block number */
275 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
276 /* Translate # of blks to # of clusters */
277 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
278                                  (sbi)->s_cluster_bits)
279
280 /*
281  * Structure of a blocks group descriptor
282  */
283 struct ext4_group_desc
284 {
285         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
286         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
287         __le32  bg_inode_table_lo;      /* Inodes table block */
288         __le16  bg_free_blocks_count_lo;/* Free blocks count */
289         __le16  bg_free_inodes_count_lo;/* Free inodes count */
290         __le16  bg_used_dirs_count_lo;  /* Directories count */
291         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
292         __u32   bg_reserved[2];         /* Likely block/inode bitmap checksum */
293         __le16  bg_itable_unused_lo;    /* Unused inodes count */
294         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
295         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
296         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
297         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
298         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
299         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
300         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
301         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
302         __u32   bg_reserved2[3];
303 };
304
305 /*
306  * Structure of a flex block group info
307  */
308
309 struct flex_groups {
310         atomic_t free_inodes;
311         atomic_t free_clusters;
312         atomic_t used_dirs;
313 };
314
315 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
316 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
317 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
318
319 /*
320  * Macro-instructions used to manage group descriptors
321  */
322 #define EXT4_MIN_DESC_SIZE              32
323 #define EXT4_MIN_DESC_SIZE_64BIT        64
324 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
325 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
326 #ifdef __KERNEL__
327 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
328 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
329 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
330 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
331 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
332 #else
333 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
334 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
335 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
336 #endif
337
338 /*
339  * Constants relative to the data blocks
340  */
341 #define EXT4_NDIR_BLOCKS                12
342 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
343 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
344 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
345 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
346
347 /*
348  * Inode flags
349  */
350 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
351 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
352 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
353 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
354 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
355 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
356 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
357 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
358 /* Reserved for compression usage... */
359 #define EXT4_DIRTY_FL                   0x00000100
360 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
361 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
362 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
363 /* End compression flags --- maybe not all used */
364 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
365 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
366 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
367 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
368 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
369 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
370 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
371 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
372 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
373 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
374 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
375
376 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
377 #define EXT4_FL_USER_MODIFIABLE         0x004B80FF /* User modifiable flags */
378
379 /* Flags that should be inherited by new inodes from their parent. */
380 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
381                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
382                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
383                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
384
385 /* Flags that are appropriate for regular files (all but dir-specific ones). */
386 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
387
388 /* Flags that are appropriate for non-directories/regular files. */
389 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
390
391 /* Mask out flags that are inappropriate for the given type of inode. */
392 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
393 {
394         if (S_ISDIR(mode))
395                 return flags;
396         else if (S_ISREG(mode))
397                 return flags & EXT4_REG_FLMASK;
398         else
399                 return flags & EXT4_OTHER_FLMASK;
400 }
401
402 /*
403  * Inode flags used for atomic set/get
404  */
405 enum {
406         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
407         EXT4_INODE_UNRM         = 1,    /* Undelete */
408         EXT4_INODE_COMPR        = 2,    /* Compress file */
409         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
410         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
411         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
412         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
413         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
414 /* Reserved for compression usage... */
415         EXT4_INODE_DIRTY        = 8,
416         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
417         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
418         EXT4_INODE_ECOMPR       = 11,   /* Compression error */
419 /* End compression flags --- maybe not all used */
420         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
421         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
422         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
423         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
424         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
425         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
426         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
427         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
428         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
429         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
430         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
431 };
432
433 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
434 #define CHECK_FLAG_VALUE(FLAG) if (!TEST_FLAG_VALUE(FLAG)) { \
435         printk(KERN_EMERG "EXT4 flag fail: " #FLAG ": %d %d\n", \
436                 EXT4_##FLAG##_FL, EXT4_INODE_##FLAG); BUG_ON(1); }
437
438 /*
439  * Since it's pretty easy to mix up bit numbers and hex values, and we
440  * can't do a compile-time test for ENUM values, we use a run-time
441  * test to make sure that EXT4_XXX_FL is consistent with respect to
442  * EXT4_INODE_XXX.  If all is well the printk and BUG_ON will all drop
443  * out so it won't cost any extra space in the compiled kernel image.
444  * But it's important that these values are the same, since we are
445  * using EXT4_INODE_XXX to test for the flag values, but EXT4_XX_FL
446  * must be consistent with the values of FS_XXX_FL defined in
447  * include/linux/fs.h and the on-disk values found in ext2, ext3, and
448  * ext4 filesystems, and of course the values defined in e2fsprogs.
449  *
450  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
451  */
452 static inline void ext4_check_flag_values(void)
453 {
454         CHECK_FLAG_VALUE(SECRM);
455         CHECK_FLAG_VALUE(UNRM);
456         CHECK_FLAG_VALUE(COMPR);
457         CHECK_FLAG_VALUE(SYNC);
458         CHECK_FLAG_VALUE(IMMUTABLE);
459         CHECK_FLAG_VALUE(APPEND);
460         CHECK_FLAG_VALUE(NODUMP);
461         CHECK_FLAG_VALUE(NOATIME);
462         CHECK_FLAG_VALUE(DIRTY);
463         CHECK_FLAG_VALUE(COMPRBLK);
464         CHECK_FLAG_VALUE(NOCOMPR);
465         CHECK_FLAG_VALUE(ECOMPR);
466         CHECK_FLAG_VALUE(INDEX);
467         CHECK_FLAG_VALUE(IMAGIC);
468         CHECK_FLAG_VALUE(JOURNAL_DATA);
469         CHECK_FLAG_VALUE(NOTAIL);
470         CHECK_FLAG_VALUE(DIRSYNC);
471         CHECK_FLAG_VALUE(TOPDIR);
472         CHECK_FLAG_VALUE(HUGE_FILE);
473         CHECK_FLAG_VALUE(EXTENTS);
474         CHECK_FLAG_VALUE(EA_INODE);
475         CHECK_FLAG_VALUE(EOFBLOCKS);
476         CHECK_FLAG_VALUE(RESERVED);
477 }
478
479 /* Used to pass group descriptor data when online resize is done */
480 struct ext4_new_group_input {
481         __u32 group;            /* Group number for this data */
482         __u64 block_bitmap;     /* Absolute block number of block bitmap */
483         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
484         __u64 inode_table;      /* Absolute block number of inode table start */
485         __u32 blocks_count;     /* Total number of blocks in this group */
486         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
487         __u16 unused;
488 };
489
490 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
491 struct compat_ext4_new_group_input {
492         u32 group;
493         compat_u64 block_bitmap;
494         compat_u64 inode_bitmap;
495         compat_u64 inode_table;
496         u32 blocks_count;
497         u16 reserved_blocks;
498         u16 unused;
499 };
500 #endif
501
502 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
503 struct ext4_new_group_data {
504         __u32 group;
505         __u64 block_bitmap;
506         __u64 inode_bitmap;
507         __u64 inode_table;
508         __u32 blocks_count;
509         __u16 reserved_blocks;
510         __u16 unused;
511         __u32 free_blocks_count;
512 };
513
514 /* Indexes used to index group tables in ext4_new_group_data */
515 enum {
516         BLOCK_BITMAP = 0,       /* block bitmap */
517         INODE_BITMAP,           /* inode bitmap */
518         INODE_TABLE,            /* inode tables */
519         GROUP_TABLE_COUNT,
520 };
521
522 /*
523  * Flags used by ext4_map_blocks()
524  */
525         /* Allocate any needed blocks and/or convert an unitialized
526            extent to be an initialized ext4 */
527 #define EXT4_GET_BLOCKS_CREATE                  0x0001
528         /* Request the creation of an unitialized extent */
529 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
530 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
531                                                  EXT4_GET_BLOCKS_CREATE)
532         /* Caller is from the delayed allocation writeout path,
533            so set the magic i_delalloc_reserve_flag after taking the
534            inode allocation semaphore for */
535 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
536         /* caller is from the direct IO path, request to creation of an
537         unitialized extents if not allocated, split the uninitialized
538         extent if blocks has been preallocated already*/
539 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
540 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
541 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
542                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
543         /* Convert extent to initialized after IO complete */
544 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
545                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
546         /* Punch out blocks of an extent */
547 #define EXT4_GET_BLOCKS_PUNCH_OUT_EXT           0x0020
548         /* Don't normalize allocation size (used for fallocate) */
549 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
550         /* Request will not result in inode size update (user for fallocate) */
551 #define EXT4_GET_BLOCKS_KEEP_SIZE               0x0080
552
553 /*
554  * Flags used by ext4_free_blocks
555  */
556 #define EXT4_FREE_BLOCKS_METADATA       0x0001
557 #define EXT4_FREE_BLOCKS_FORGET         0x0002
558 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
559 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
560 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
561 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
562
563 /*
564  * Flags used by ext4_discard_partial_page_buffers
565  */
566 #define EXT4_DISCARD_PARTIAL_PG_ZERO_UNMAPPED   0x0001
567
568 /*
569  * ioctl commands
570  */
571 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
572 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
573 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
574 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
575 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
576 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
577 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
578 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
579 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
580 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
581 #define EXT4_IOC_MIGRATE                _IO('f', 9)
582  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
583  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
584 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
585 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
586 #define EXT4_IOC_RESIZE_FS              _IOW('f', 16, __u64)
587
588 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
589 /*
590  * ioctl commands in 32 bit emulation
591  */
592 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
593 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
594 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
595 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
596 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
597 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
598 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
599 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
600 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
601 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
602 #endif
603
604 /* Max physical block we can address w/o extents */
605 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
606
607 /*
608  * Structure of an inode on the disk
609  */
610 struct ext4_inode {
611         __le16  i_mode;         /* File mode */
612         __le16  i_uid;          /* Low 16 bits of Owner Uid */
613         __le32  i_size_lo;      /* Size in bytes */
614         __le32  i_atime;        /* Access time */
615         __le32  i_ctime;        /* Inode Change time */
616         __le32  i_mtime;        /* Modification time */
617         __le32  i_dtime;        /* Deletion Time */
618         __le16  i_gid;          /* Low 16 bits of Group Id */
619         __le16  i_links_count;  /* Links count */
620         __le32  i_blocks_lo;    /* Blocks count */
621         __le32  i_flags;        /* File flags */
622         union {
623                 struct {
624                         __le32  l_i_version;
625                 } linux1;
626                 struct {
627                         __u32  h_i_translator;
628                 } hurd1;
629                 struct {
630                         __u32  m_i_reserved1;
631                 } masix1;
632         } osd1;                         /* OS dependent 1 */
633         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
634         __le32  i_generation;   /* File version (for NFS) */
635         __le32  i_file_acl_lo;  /* File ACL */
636         __le32  i_size_high;
637         __le32  i_obso_faddr;   /* Obsoleted fragment address */
638         union {
639                 struct {
640                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
641                         __le16  l_i_file_acl_high;
642                         __le16  l_i_uid_high;   /* these 2 fields */
643                         __le16  l_i_gid_high;   /* were reserved2[0] */
644                         __u32   l_i_reserved2;
645                 } linux2;
646                 struct {
647                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
648                         __u16   h_i_mode_high;
649                         __u16   h_i_uid_high;
650                         __u16   h_i_gid_high;
651                         __u32   h_i_author;
652                 } hurd2;
653                 struct {
654                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
655                         __le16  m_i_file_acl_high;
656                         __u32   m_i_reserved2[2];
657                 } masix2;
658         } osd2;                         /* OS dependent 2 */
659         __le16  i_extra_isize;
660         __le16  i_pad1;
661         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
662         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
663         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
664         __le32  i_crtime;       /* File Creation time */
665         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
666         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
667 };
668
669 struct move_extent {
670         __u32 reserved;         /* should be zero */
671         __u32 donor_fd;         /* donor file descriptor */
672         __u64 orig_start;       /* logical start offset in block for orig */
673         __u64 donor_start;      /* logical start offset in block for donor */
674         __u64 len;              /* block length to be moved */
675         __u64 moved_len;        /* moved block length */
676 };
677
678 #define EXT4_EPOCH_BITS 2
679 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
680 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
681
682 /*
683  * Extended fields will fit into an inode if the filesystem was formatted
684  * with large inodes (-I 256 or larger) and there are not currently any EAs
685  * consuming all of the available space. For new inodes we always reserve
686  * enough space for the kernel's known extended fields, but for inodes
687  * created with an old kernel this might not have been the case. None of
688  * the extended inode fields is critical for correct filesystem operation.
689  * This macro checks if a certain field fits in the inode. Note that
690  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
691  */
692 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
693         ((offsetof(typeof(*ext4_inode), field) +        \
694           sizeof((ext4_inode)->field))                  \
695         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
696             (einode)->i_extra_isize))                   \
697
698 static inline __le32 ext4_encode_extra_time(struct timespec *time)
699 {
700        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
701                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
702                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
703 }
704
705 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
706 {
707        if (sizeof(time->tv_sec) > 4)
708                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
709                                << 32;
710        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
711 }
712
713 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
714 do {                                                                           \
715         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
716         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
717                 (raw_inode)->xtime ## _extra =                                 \
718                                 ext4_encode_extra_time(&(inode)->xtime);       \
719 } while (0)
720
721 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
722 do {                                                                           \
723         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
724                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
725         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
726                 (raw_inode)->xtime ## _extra =                                 \
727                                 ext4_encode_extra_time(&(einode)->xtime);      \
728 } while (0)
729
730 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
731 do {                                                                           \
732         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
733         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
734                 ext4_decode_extra_time(&(inode)->xtime,                        \
735                                        raw_inode->xtime ## _extra);            \
736         else                                                                   \
737                 (inode)->xtime.tv_nsec = 0;                                    \
738 } while (0)
739
740 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
741 do {                                                                           \
742         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
743                 (einode)->xtime.tv_sec =                                       \
744                         (signed)le32_to_cpu((raw_inode)->xtime);               \
745         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
746                 ext4_decode_extra_time(&(einode)->xtime,                       \
747                                        raw_inode->xtime ## _extra);            \
748         else                                                                   \
749                 (einode)->xtime.tv_nsec = 0;                                   \
750 } while (0)
751
752 #define i_disk_version osd1.linux1.l_i_version
753
754 #if defined(__KERNEL__) || defined(__linux__)
755 #define i_reserved1     osd1.linux1.l_i_reserved1
756 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
757 #define i_blocks_high   osd2.linux2.l_i_blocks_high
758 #define i_uid_low       i_uid
759 #define i_gid_low       i_gid
760 #define i_uid_high      osd2.linux2.l_i_uid_high
761 #define i_gid_high      osd2.linux2.l_i_gid_high
762 #define i_reserved2     osd2.linux2.l_i_reserved2
763
764 #elif defined(__GNU__)
765
766 #define i_translator    osd1.hurd1.h_i_translator
767 #define i_uid_high      osd2.hurd2.h_i_uid_high
768 #define i_gid_high      osd2.hurd2.h_i_gid_high
769 #define i_author        osd2.hurd2.h_i_author
770
771 #elif defined(__masix__)
772
773 #define i_reserved1     osd1.masix1.m_i_reserved1
774 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
775 #define i_reserved2     osd2.masix2.m_i_reserved2
776
777 #endif /* defined(__KERNEL__) || defined(__linux__) */
778
779 /*
780  * storage for cached extent
781  * If ec_len == 0, then the cache is invalid.
782  * If ec_start == 0, then the cache represents a gap (null mapping)
783  */
784 struct ext4_ext_cache {
785         ext4_fsblk_t    ec_start;
786         ext4_lblk_t     ec_block;
787         __u32           ec_len; /* must be 32bit to return holes */
788 };
789
790 /*
791  * fourth extended file system inode data in memory
792  */
793 struct ext4_inode_info {
794         __le32  i_data[15];     /* unconverted */
795         __u32   i_dtime;
796         ext4_fsblk_t    i_file_acl;
797
798         /*
799          * i_block_group is the number of the block group which contains
800          * this file's inode.  Constant across the lifetime of the inode,
801          * it is ued for making block allocation decisions - we try to
802          * place a file's data blocks near its inode block, and new inodes
803          * near to their parent directory's inode.
804          */
805         ext4_group_t    i_block_group;
806         ext4_lblk_t     i_dir_start_lookup;
807 #if (BITS_PER_LONG < 64)
808         unsigned long   i_state_flags;          /* Dynamic state flags */
809 #endif
810         unsigned long   i_flags;
811
812 #ifdef CONFIG_EXT4_FS_XATTR
813         /*
814          * Extended attributes can be read independently of the main file
815          * data. Taking i_mutex even when reading would cause contention
816          * between readers of EAs and writers of regular file data, so
817          * instead we synchronize on xattr_sem when reading or changing
818          * EAs.
819          */
820         struct rw_semaphore xattr_sem;
821 #endif
822
823         struct list_head i_orphan;      /* unlinked but open inodes */
824
825         /*
826          * i_disksize keeps track of what the inode size is ON DISK, not
827          * in memory.  During truncate, i_size is set to the new size by
828          * the VFS prior to calling ext4_truncate(), but the filesystem won't
829          * set i_disksize to 0 until the truncate is actually under way.
830          *
831          * The intent is that i_disksize always represents the blocks which
832          * are used by this file.  This allows recovery to restart truncate
833          * on orphans if we crash during truncate.  We actually write i_disksize
834          * into the on-disk inode when writing inodes out, instead of i_size.
835          *
836          * The only time when i_disksize and i_size may be different is when
837          * a truncate is in progress.  The only things which change i_disksize
838          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
839          */
840         loff_t  i_disksize;
841
842         /*
843          * i_data_sem is for serialising ext4_truncate() against
844          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
845          * data tree are chopped off during truncate. We can't do that in
846          * ext4 because whenever we perform intermediate commits during
847          * truncate, the inode and all the metadata blocks *must* be in a
848          * consistent state which allows truncation of the orphans to restart
849          * during recovery.  Hence we must fix the get_block-vs-truncate race
850          * by other means, so we have i_data_sem.
851          */
852         struct rw_semaphore i_data_sem;
853         struct inode vfs_inode;
854         struct jbd2_inode *jinode;
855
856         struct ext4_ext_cache i_cached_extent;
857         /*
858          * File creation time. Its function is same as that of
859          * struct timespec i_{a,c,m}time in the generic inode.
860          */
861         struct timespec i_crtime;
862
863         /* mballoc */
864         struct list_head i_prealloc_list;
865         spinlock_t i_prealloc_lock;
866
867         /* ialloc */
868         ext4_group_t    i_last_alloc_group;
869
870         /* allocation reservation info for delalloc */
871         /* In case of bigalloc, these refer to clusters rather than blocks */
872         unsigned int i_reserved_data_blocks;
873         unsigned int i_reserved_meta_blocks;
874         unsigned int i_allocated_meta_blocks;
875         ext4_lblk_t i_da_metadata_calc_last_lblock;
876         int i_da_metadata_calc_len;
877
878         /* on-disk additional length */
879         __u16 i_extra_isize;
880
881 #ifdef CONFIG_QUOTA
882         /* quota space reservation, managed internally by quota code */
883         qsize_t i_reserved_quota;
884 #endif
885
886         /* completed IOs that might need unwritten extents handling */
887         struct list_head i_completed_io_list;
888         spinlock_t i_completed_io_lock;
889         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
890         /* current io_end structure for async DIO write*/
891         ext4_io_end_t *cur_aio_dio;
892         atomic_t i_aiodio_unwritten; /* Nr. of inflight conversions pending */
893
894         spinlock_t i_block_reservation_lock;
895
896         /*
897          * Transactions that contain inode's metadata needed to complete
898          * fsync and fdatasync, respectively.
899          */
900         tid_t i_sync_tid;
901         tid_t i_datasync_tid;
902 };
903
904 /*
905  * File system states
906  */
907 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
908 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
909 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
910
911 /*
912  * Misc. filesystem flags
913  */
914 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
915 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
916 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
917
918 /*
919  * Mount flags
920  */
921 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
922 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
923 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
924 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
925 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
926 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
927 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
928 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
929 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
930 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
931 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
932 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
933 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
934 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
935 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
936 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
937 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
938 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
939 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
940 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
941 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
942 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
943 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
944 #define EXT4_MOUNT_I_VERSION            0x2000000 /* i_version support */
945 #define EXT4_MOUNT_MBLK_IO_SUBMIT       0x4000000 /* multi-block io submits */
946 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
947 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
948 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
949 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
950 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
951
952 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
953                                                       specified delalloc */
954
955 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
956                                                 ~EXT4_MOUNT_##opt
957 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
958                                                 EXT4_MOUNT_##opt
959 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
960                                          EXT4_MOUNT_##opt)
961
962 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
963                                                 ~EXT4_MOUNT2_##opt
964 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
965                                                 EXT4_MOUNT2_##opt
966 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
967                                          EXT4_MOUNT2_##opt)
968
969 #define ext4_test_and_set_bit           __test_and_set_bit_le
970 #define ext4_set_bit                    __set_bit_le
971 #define ext4_set_bit_atomic             ext2_set_bit_atomic
972 #define ext4_test_and_clear_bit         __test_and_clear_bit_le
973 #define ext4_clear_bit                  __clear_bit_le
974 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
975 #define ext4_test_bit                   test_bit_le
976 #define ext4_find_next_zero_bit         find_next_zero_bit_le
977 #define ext4_find_next_bit              find_next_bit_le
978
979 extern void ext4_set_bits(void *bm, int cur, int len);
980
981 /*
982  * Maximal mount counts between two filesystem checks
983  */
984 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
985 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
986
987 /*
988  * Behaviour when detecting errors
989  */
990 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
991 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
992 #define EXT4_ERRORS_PANIC               3       /* Panic */
993 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
994
995 /*
996  * Structure of the super block
997  */
998 struct ext4_super_block {
999 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1000         __le32  s_blocks_count_lo;      /* Blocks count */
1001         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1002         __le32  s_free_blocks_count_lo; /* Free blocks count */
1003 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1004         __le32  s_first_data_block;     /* First Data Block */
1005         __le32  s_log_block_size;       /* Block size */
1006         __le32  s_log_cluster_size;     /* Allocation cluster size */
1007 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1008         __le32  s_clusters_per_group;   /* # Clusters per group */
1009         __le32  s_inodes_per_group;     /* # Inodes per group */
1010         __le32  s_mtime;                /* Mount time */
1011 /*30*/  __le32  s_wtime;                /* Write time */
1012         __le16  s_mnt_count;            /* Mount count */
1013         __le16  s_max_mnt_count;        /* Maximal mount count */
1014         __le16  s_magic;                /* Magic signature */
1015         __le16  s_state;                /* File system state */
1016         __le16  s_errors;               /* Behaviour when detecting errors */
1017         __le16  s_minor_rev_level;      /* minor revision level */
1018 /*40*/  __le32  s_lastcheck;            /* time of last check */
1019         __le32  s_checkinterval;        /* max. time between checks */
1020         __le32  s_creator_os;           /* OS */
1021         __le32  s_rev_level;            /* Revision level */
1022 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1023         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1024         /*
1025          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1026          *
1027          * Note: the difference between the compatible feature set and
1028          * the incompatible feature set is that if there is a bit set
1029          * in the incompatible feature set that the kernel doesn't
1030          * know about, it should refuse to mount the filesystem.
1031          *
1032          * e2fsck's requirements are more strict; if it doesn't know
1033          * about a feature in either the compatible or incompatible
1034          * feature set, it must abort and not try to meddle with
1035          * things it doesn't understand...
1036          */
1037         __le32  s_first_ino;            /* First non-reserved inode */
1038         __le16  s_inode_size;           /* size of inode structure */
1039         __le16  s_block_group_nr;       /* block group # of this superblock */
1040         __le32  s_feature_compat;       /* compatible feature set */
1041 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1042         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1043 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1044 /*78*/  char    s_volume_name[16];      /* volume name */
1045 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1046 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1047         /*
1048          * Performance hints.  Directory preallocation should only
1049          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1050          */
1051         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1052         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1053         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1054         /*
1055          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1056          */
1057 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1058 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1059         __le32  s_journal_dev;          /* device number of journal file */
1060         __le32  s_last_orphan;          /* start of list of inodes to delete */
1061         __le32  s_hash_seed[4];         /* HTREE hash seed */
1062         __u8    s_def_hash_version;     /* Default hash version to use */
1063         __u8    s_jnl_backup_type;
1064         __le16  s_desc_size;            /* size of group descriptor */
1065 /*100*/ __le32  s_default_mount_opts;
1066         __le32  s_first_meta_bg;        /* First metablock block group */
1067         __le32  s_mkfs_time;            /* When the filesystem was created */
1068         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1069         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1070 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1071         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1072         __le32  s_free_blocks_count_hi; /* Free blocks count */
1073         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1074         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1075         __le32  s_flags;                /* Miscellaneous flags */
1076         __le16  s_raid_stride;          /* RAID stride */
1077         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1078         __le64  s_mmp_block;            /* Block for multi-mount protection */
1079         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1080         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1081         __u8    s_reserved_char_pad;
1082         __le16  s_reserved_pad;
1083         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1084         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1085         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1086         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1087                                               snapshot's future use */
1088         __le32  s_snapshot_list;        /* inode number of the head of the
1089                                            on-disk snapshot list */
1090 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1091         __le32  s_error_count;          /* number of fs errors */
1092         __le32  s_first_error_time;     /* first time an error happened */
1093         __le32  s_first_error_ino;      /* inode involved in first error */
1094         __le64  s_first_error_block;    /* block involved of first error */
1095         __u8    s_first_error_func[32]; /* function where the error happened */
1096         __le32  s_first_error_line;     /* line number where error happened */
1097         __le32  s_last_error_time;      /* most recent time of an error */
1098         __le32  s_last_error_ino;       /* inode involved in last error */
1099         __le32  s_last_error_line;      /* line number where error happened */
1100         __le64  s_last_error_block;     /* block involved of last error */
1101         __u8    s_last_error_func[32];  /* function where the error happened */
1102 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1103         __u8    s_mount_opts[64];
1104         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1105         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1106         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1107         __le32  s_reserved[109];        /* Padding to the end of the block */
1108 };
1109
1110 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1111
1112 #ifdef __KERNEL__
1113
1114 /*
1115  * run-time mount flags
1116  */
1117 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
1118 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
1119
1120 /*
1121  * fourth extended-fs super-block data in memory
1122  */
1123 struct ext4_sb_info {
1124         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1125         unsigned long s_inodes_per_block;/* Number of inodes per block */
1126         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1127         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1128         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1129         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1130         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1131         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1132         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1133         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1134         unsigned long s_overhead_last;  /* Last calculated overhead */
1135         unsigned long s_blocks_last;    /* Last seen block count */
1136         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1137         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1138         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1139         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1140         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1141         struct buffer_head **s_group_desc;
1142         unsigned int s_mount_opt;
1143         unsigned int s_mount_opt2;
1144         unsigned int s_mount_flags;
1145         ext4_fsblk_t s_sb_block;
1146         uid_t s_resuid;
1147         gid_t s_resgid;
1148         unsigned short s_mount_state;
1149         unsigned short s_pad;
1150         int s_addr_per_block_bits;
1151         int s_desc_per_block_bits;
1152         int s_inode_size;
1153         int s_first_ino;
1154         unsigned int s_inode_readahead_blks;
1155         unsigned int s_inode_goal;
1156         spinlock_t s_next_gen_lock;
1157         u32 s_next_generation;
1158         u32 s_hash_seed[4];
1159         int s_def_hash_version;
1160         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1161         struct percpu_counter s_freeclusters_counter;
1162         struct percpu_counter s_freeinodes_counter;
1163         struct percpu_counter s_dirs_counter;
1164         struct percpu_counter s_dirtyclusters_counter;
1165         struct blockgroup_lock *s_blockgroup_lock;
1166         struct proc_dir_entry *s_proc;
1167         struct kobject s_kobj;
1168         struct completion s_kobj_unregister;
1169
1170         /* Journaling */
1171         struct journal_s *s_journal;
1172         struct list_head s_orphan;
1173         struct mutex s_orphan_lock;
1174         unsigned long s_resize_flags;           /* Flags indicating if there
1175                                                    is a resizer */
1176         unsigned long s_commit_interval;
1177         u32 s_max_batch_time;
1178         u32 s_min_batch_time;
1179         struct block_device *journal_bdev;
1180 #ifdef CONFIG_QUOTA
1181         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
1182         int s_jquota_fmt;                       /* Format of quota to use */
1183 #endif
1184         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1185         struct rb_root system_blks;
1186
1187 #ifdef EXTENTS_STATS
1188         /* ext4 extents stats */
1189         unsigned long s_ext_min;
1190         unsigned long s_ext_max;
1191         unsigned long s_depth_max;
1192         spinlock_t s_ext_stats_lock;
1193         unsigned long s_ext_blocks;
1194         unsigned long s_ext_extents;
1195 #endif
1196         /* ext4 extent cache stats */
1197         unsigned long extent_cache_hits;
1198         unsigned long extent_cache_misses;
1199
1200         /* for buddy allocator */
1201         struct ext4_group_info ***s_group_info;
1202         struct inode *s_buddy_cache;
1203         spinlock_t s_md_lock;
1204         unsigned short *s_mb_offsets;
1205         unsigned int *s_mb_maxs;
1206
1207         /* tunables */
1208         unsigned long s_stripe;
1209         unsigned int s_mb_stream_request;
1210         unsigned int s_mb_max_to_scan;
1211         unsigned int s_mb_min_to_scan;
1212         unsigned int s_mb_stats;
1213         unsigned int s_mb_order2_reqs;
1214         unsigned int s_mb_group_prealloc;
1215         unsigned int s_max_writeback_mb_bump;
1216         /* where last allocation was done - for stream allocation */
1217         unsigned long s_mb_last_group;
1218         unsigned long s_mb_last_start;
1219
1220         /* stats for buddy allocator */
1221         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1222         atomic_t s_bal_success; /* we found long enough chunks */
1223         atomic_t s_bal_allocated;       /* in blocks */
1224         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1225         atomic_t s_bal_goals;   /* goal hits */
1226         atomic_t s_bal_breaks;  /* too long searches */
1227         atomic_t s_bal_2orders; /* 2^order hits */
1228         spinlock_t s_bal_lock;
1229         unsigned long s_mb_buddies_generated;
1230         unsigned long long s_mb_generation_time;
1231         atomic_t s_mb_lost_chunks;
1232         atomic_t s_mb_preallocated;
1233         atomic_t s_mb_discarded;
1234         atomic_t s_lock_busy;
1235
1236         /* locality groups */
1237         struct ext4_locality_group __percpu *s_locality_groups;
1238
1239         /* for write statistics */
1240         unsigned long s_sectors_written_start;
1241         u64 s_kbytes_written;
1242
1243         unsigned int s_log_groups_per_flex;
1244         struct flex_groups *s_flex_groups;
1245
1246         /* workqueue for dio unwritten */
1247         struct workqueue_struct *dio_unwritten_wq;
1248
1249         /* timer for periodic error stats printing */
1250         struct timer_list s_err_report;
1251
1252         /* Lazy inode table initialization info */
1253         struct ext4_li_request *s_li_request;
1254         /* Wait multiplier for lazy initialization thread */
1255         unsigned int s_li_wait_mult;
1256
1257         /* Kernel thread for multiple mount protection */
1258         struct task_struct *s_mmp_tsk;
1259
1260         /* record the last minlen when FITRIM is called. */
1261         atomic_t s_last_trim_minblks;
1262 };
1263
1264 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1265 {
1266         return sb->s_fs_info;
1267 }
1268 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1269 {
1270         return container_of(inode, struct ext4_inode_info, vfs_inode);
1271 }
1272
1273 static inline struct timespec ext4_current_time(struct inode *inode)
1274 {
1275         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1276                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1277 }
1278
1279 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1280 {
1281         return ino == EXT4_ROOT_INO ||
1282                 ino == EXT4_JOURNAL_INO ||
1283                 ino == EXT4_RESIZE_INO ||
1284                 (ino >= EXT4_FIRST_INO(sb) &&
1285                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1286 }
1287
1288 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
1289                                               struct ext4_io_end *io_end)
1290 {
1291         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
1292                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
1293                 atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten);
1294         }
1295 }
1296
1297 /*
1298  * Inode dynamic state flags
1299  */
1300 enum {
1301         EXT4_STATE_JDATA,               /* journaled data exists */
1302         EXT4_STATE_NEW,                 /* inode is newly created */
1303         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1304         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1305         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1306         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1307         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1308         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1309         EXT4_STATE_DELALLOC_RESERVED,   /* blks already reserved for delalloc */
1310 };
1311
1312 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1313 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1314 {                                                                       \
1315         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1316 }                                                                       \
1317 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1318 {                                                                       \
1319         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1320 }                                                                       \
1321 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1322 {                                                                       \
1323         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1324 }
1325
1326 EXT4_INODE_BIT_FNS(flag, flags, 0)
1327 #if (BITS_PER_LONG < 64)
1328 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1329
1330 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1331 {
1332         (ei)->i_state_flags = 0;
1333 }
1334 #else
1335 EXT4_INODE_BIT_FNS(state, flags, 32)
1336
1337 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1338 {
1339         /* We depend on the fact that callers will set i_flags */
1340 }
1341 #endif
1342 #else
1343 /* Assume that user mode programs are passing in an ext4fs superblock, not
1344  * a kernel struct super_block.  This will allow us to call the feature-test
1345  * macros from user land. */
1346 #define EXT4_SB(sb)     (sb)
1347 #endif
1348
1349 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1350
1351 /*
1352  * Codes for operating systems
1353  */
1354 #define EXT4_OS_LINUX           0
1355 #define EXT4_OS_HURD            1
1356 #define EXT4_OS_MASIX           2
1357 #define EXT4_OS_FREEBSD         3
1358 #define EXT4_OS_LITES           4
1359
1360 /*
1361  * Revision levels
1362  */
1363 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1364 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1365
1366 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1367 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1368
1369 #define EXT4_GOOD_OLD_INODE_SIZE 128
1370
1371 /*
1372  * Feature set definitions
1373  */
1374
1375 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1376         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1377 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1378         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1379 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1380         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1381 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1382         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1383 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1384         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1385 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1386         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1387 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1388         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1389 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1390         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1391 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1392         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1393
1394 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1395 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1396 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1397 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1398 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1399 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1400
1401 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1402 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1403 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1404 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1405 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1406 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1407 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1408 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1409 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1410 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM    0x0400
1411
1412 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1413 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1414 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1415 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1416 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1417 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1418 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1419 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1420 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1421 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1422 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1423 #define EXT4_FEATURE_INCOMPAT_INLINEDATA        0x2000 /* data in inode */
1424 #define EXT4_FEATURE_INCOMPAT_LARGEDIR          0x4000 /* >2GB or 3-lvl htree */
1425
1426 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1427 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1428                                          EXT4_FEATURE_INCOMPAT_META_BG)
1429 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1430                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1431                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1432
1433 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1434 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1435                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1436                                          EXT4_FEATURE_INCOMPAT_META_BG)
1437 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1438                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1439                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1440
1441 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1442 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1443                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1444                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1445                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1446                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1447                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1448                                          EXT4_FEATURE_INCOMPAT_MMP)
1449 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1450                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1451                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1452                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1453                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1454                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1455                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1456                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC)
1457
1458 /*
1459  * Default values for user and/or group using reserved blocks
1460  */
1461 #define EXT4_DEF_RESUID         0
1462 #define EXT4_DEF_RESGID         0
1463
1464 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1465
1466 /*
1467  * Default mount options
1468  */
1469 #define EXT4_DEFM_DEBUG         0x0001
1470 #define EXT4_DEFM_BSDGROUPS     0x0002
1471 #define EXT4_DEFM_XATTR_USER    0x0004
1472 #define EXT4_DEFM_ACL           0x0008
1473 #define EXT4_DEFM_UID16         0x0010
1474 #define EXT4_DEFM_JMODE         0x0060
1475 #define EXT4_DEFM_JMODE_DATA    0x0020
1476 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1477 #define EXT4_DEFM_JMODE_WBACK   0x0060
1478 #define EXT4_DEFM_NOBARRIER     0x0100
1479 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1480 #define EXT4_DEFM_DISCARD       0x0400
1481 #define EXT4_DEFM_NODELALLOC    0x0800
1482
1483 /*
1484  * Default journal batch times
1485  */
1486 #define EXT4_DEF_MIN_BATCH_TIME 0
1487 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1488
1489 /*
1490  * Minimum number of groups in a flexgroup before we separate out
1491  * directories into the first block group of a flexgroup
1492  */
1493 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1494
1495 /*
1496  * Structure of a directory entry
1497  */
1498 #define EXT4_NAME_LEN 255
1499
1500 struct ext4_dir_entry {
1501         __le32  inode;                  /* Inode number */
1502         __le16  rec_len;                /* Directory entry length */
1503         __le16  name_len;               /* Name length */
1504         char    name[EXT4_NAME_LEN];    /* File name */
1505 };
1506
1507 /*
1508  * The new version of the directory entry.  Since EXT4 structures are
1509  * stored in intel byte order, and the name_len field could never be
1510  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1511  * file_type field.
1512  */
1513 struct ext4_dir_entry_2 {
1514         __le32  inode;                  /* Inode number */
1515         __le16  rec_len;                /* Directory entry length */
1516         __u8    name_len;               /* Name length */
1517         __u8    file_type;
1518         char    name[EXT4_NAME_LEN];    /* File name */
1519 };
1520
1521 /*
1522  * Ext4 directory file types.  Only the low 3 bits are used.  The
1523  * other bits are reserved for now.
1524  */
1525 #define EXT4_FT_UNKNOWN         0
1526 #define EXT4_FT_REG_FILE        1
1527 #define EXT4_FT_DIR             2
1528 #define EXT4_FT_CHRDEV          3
1529 #define EXT4_FT_BLKDEV          4
1530 #define EXT4_FT_FIFO            5
1531 #define EXT4_FT_SOCK            6
1532 #define EXT4_FT_SYMLINK         7
1533
1534 #define EXT4_FT_MAX             8
1535
1536 /*
1537  * EXT4_DIR_PAD defines the directory entries boundaries
1538  *
1539  * NOTE: It must be a multiple of 4
1540  */
1541 #define EXT4_DIR_PAD                    4
1542 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1543 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1544                                          ~EXT4_DIR_ROUND)
1545 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1546
1547 /*
1548  * If we ever get support for fs block sizes > page_size, we'll need
1549  * to remove the #if statements in the next two functions...
1550  */
1551 static inline unsigned int
1552 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1553 {
1554         unsigned len = le16_to_cpu(dlen);
1555
1556 #if (PAGE_CACHE_SIZE >= 65536)
1557         if (len == EXT4_MAX_REC_LEN || len == 0)
1558                 return blocksize;
1559         return (len & 65532) | ((len & 3) << 16);
1560 #else
1561         return len;
1562 #endif
1563 }
1564
1565 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1566 {
1567         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1568                 BUG();
1569 #if (PAGE_CACHE_SIZE >= 65536)
1570         if (len < 65536)
1571                 return cpu_to_le16(len);
1572         if (len == blocksize) {
1573                 if (blocksize == 65536)
1574                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1575                 else
1576                         return cpu_to_le16(0);
1577         }
1578         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1579 #else
1580         return cpu_to_le16(len);
1581 #endif
1582 }
1583
1584 /*
1585  * Hash Tree Directory indexing
1586  * (c) Daniel Phillips, 2001
1587  */
1588
1589 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1590                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1591                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1592 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1593 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1594
1595 /* Legal values for the dx_root hash_version field: */
1596
1597 #define DX_HASH_LEGACY          0
1598 #define DX_HASH_HALF_MD4        1
1599 #define DX_HASH_TEA             2
1600 #define DX_HASH_LEGACY_UNSIGNED 3
1601 #define DX_HASH_HALF_MD4_UNSIGNED       4
1602 #define DX_HASH_TEA_UNSIGNED            5
1603
1604 #ifdef __KERNEL__
1605
1606 /* hash info structure used by the directory hash */
1607 struct dx_hash_info
1608 {
1609         u32             hash;
1610         u32             minor_hash;
1611         int             hash_version;
1612         u32             *seed;
1613 };
1614
1615 #define EXT4_HTREE_EOF  0x7fffffff
1616
1617 /*
1618  * Control parameters used by ext4_htree_next_block
1619  */
1620 #define HASH_NB_ALWAYS          1
1621
1622
1623 /*
1624  * Describe an inode's exact location on disk and in memory
1625  */
1626 struct ext4_iloc
1627 {
1628         struct buffer_head *bh;
1629         unsigned long offset;
1630         ext4_group_t block_group;
1631 };
1632
1633 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1634 {
1635         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1636 }
1637
1638 /*
1639  * This structure is stuffed into the struct file's private_data field
1640  * for directories.  It is where we put information so that we can do
1641  * readdir operations in hash tree order.
1642  */
1643 struct dir_private_info {
1644         struct rb_root  root;
1645         struct rb_node  *curr_node;
1646         struct fname    *extra_fname;
1647         loff_t          last_pos;
1648         __u32           curr_hash;
1649         __u32           curr_minor_hash;
1650         __u32           next_hash;
1651 };
1652
1653 /* calculate the first block number of the group */
1654 static inline ext4_fsblk_t
1655 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1656 {
1657         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1658                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1659 }
1660
1661 /*
1662  * Special error return code only used by dx_probe() and its callers.
1663  */
1664 #define ERR_BAD_DX_DIR  -75000
1665
1666 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1667                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1668
1669 /*
1670  * Timeout and state flag for lazy initialization inode thread.
1671  */
1672 #define EXT4_DEF_LI_WAIT_MULT                   10
1673 #define EXT4_DEF_LI_MAX_START_DELAY             5
1674 #define EXT4_LAZYINIT_QUIT                      0x0001
1675 #define EXT4_LAZYINIT_RUNNING                   0x0002
1676
1677 /*
1678  * Lazy inode table initialization info
1679  */
1680 struct ext4_lazy_init {
1681         unsigned long           li_state;
1682         struct list_head        li_request_list;
1683         struct mutex            li_list_mtx;
1684 };
1685
1686 struct ext4_li_request {
1687         struct super_block      *lr_super;
1688         struct ext4_sb_info     *lr_sbi;
1689         ext4_group_t            lr_next_group;
1690         struct list_head        lr_request;
1691         unsigned long           lr_next_sched;
1692         unsigned long           lr_timeout;
1693 };
1694
1695 struct ext4_features {
1696         struct kobject f_kobj;
1697         struct completion f_kobj_unregister;
1698 };
1699
1700 /*
1701  * This structure will be used for multiple mount protection. It will be
1702  * written into the block number saved in the s_mmp_block field in the
1703  * superblock. Programs that check MMP should assume that if
1704  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1705  * to use the filesystem, regardless of how old the timestamp is.
1706  */
1707 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
1708 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1709 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
1710 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
1711
1712 struct mmp_struct {
1713         __le32  mmp_magic;              /* Magic number for MMP */
1714         __le32  mmp_seq;                /* Sequence no. updated periodically */
1715
1716         /*
1717          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1718          * purposes and do not affect the correctness of the algorithm
1719          */
1720         __le64  mmp_time;               /* Time last updated */
1721         char    mmp_nodename[64];       /* Node which last updated MMP block */
1722         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
1723
1724         /*
1725          * mmp_check_interval is used to verify if the MMP block has been
1726          * updated on the block device. The value is updated based on the
1727          * maximum time to write the MMP block during an update cycle.
1728          */
1729         __le16  mmp_check_interval;
1730
1731         __le16  mmp_pad1;
1732         __le32  mmp_pad2[227];
1733 };
1734
1735 /* arguments passed to the mmp thread */
1736 struct mmpd_data {
1737         struct buffer_head *bh; /* bh from initial read_mmp_block() */
1738         struct super_block *sb;  /* super block of the fs */
1739 };
1740
1741 /*
1742  * Check interval multiplier
1743  * The MMP block is written every update interval and initially checked every
1744  * update interval x the multiplier (the value is then adapted based on the
1745  * write latency). The reason is that writes can be delayed under load and we
1746  * don't want readers to incorrectly assume that the filesystem is no longer
1747  * in use.
1748  */
1749 #define EXT4_MMP_CHECK_MULT             2UL
1750
1751 /*
1752  * Minimum interval for MMP checking in seconds.
1753  */
1754 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
1755
1756 /*
1757  * Maximum interval for MMP checking in seconds.
1758  */
1759 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
1760
1761 /*
1762  * Function prototypes
1763  */
1764
1765 /*
1766  * Ok, these declarations are also in <linux/kernel.h> but none of the
1767  * ext4 source programs needs to include it so they are duplicated here.
1768  */
1769 # define NORET_TYPE     /**/
1770 # define ATTRIB_NORET   __attribute__((noreturn))
1771 # define NORET_AND      noreturn,
1772
1773 /* bitmap.c */
1774 extern unsigned int ext4_count_free(struct buffer_head *, unsigned);
1775
1776 /* balloc.c */
1777 extern unsigned int ext4_block_group(struct super_block *sb,
1778                         ext4_fsblk_t blocknr);
1779 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1780                         ext4_fsblk_t blocknr);
1781 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1782 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1783                         ext4_group_t group);
1784 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1785                                          ext4_fsblk_t goal,
1786                                          unsigned int flags,
1787                                          unsigned long *count,
1788                                          int *errp);
1789 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
1790                                     s64 nclusters, unsigned int flags);
1791 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
1792 extern void ext4_check_blocks_bitmap(struct super_block *);
1793 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1794                                                     ext4_group_t block_group,
1795                                                     struct buffer_head ** bh);
1796 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1797 struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1798                                       ext4_group_t block_group);
1799 extern void ext4_init_block_bitmap(struct super_block *sb,
1800                                    struct buffer_head *bh,
1801                                    ext4_group_t group,
1802                                    struct ext4_group_desc *desc);
1803 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
1804                                               ext4_group_t block_group,
1805                                               struct ext4_group_desc *gdp);
1806 extern unsigned ext4_num_overhead_clusters(struct super_block *sb,
1807                                            ext4_group_t block_group,
1808                                            struct ext4_group_desc *gdp);
1809 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
1810
1811 /* dir.c */
1812 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1813                                   struct file *,
1814                                   struct ext4_dir_entry_2 *,
1815                                   struct buffer_head *, unsigned int);
1816 #define ext4_check_dir_entry(dir, filp, de, bh, offset)                 \
1817         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
1818                                         (de), (bh), (offset)))
1819 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1820                                     __u32 minor_hash,
1821                                     struct ext4_dir_entry_2 *dirent);
1822 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1823
1824 /* fsync.c */
1825 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
1826 extern int ext4_flush_completed_IO(struct inode *);
1827
1828 /* hash.c */
1829 extern int ext4fs_dirhash(const char *name, int len, struct
1830                           dx_hash_info *hinfo);
1831
1832 /* ialloc.c */
1833 extern struct inode *ext4_new_inode(handle_t *, struct inode *, umode_t,
1834                                     const struct qstr *qstr, __u32 goal,
1835                                     uid_t *owner);
1836 extern void ext4_free_inode(handle_t *, struct inode *);
1837 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1838 extern unsigned long ext4_count_free_inodes(struct super_block *);
1839 extern unsigned long ext4_count_dirs(struct super_block *);
1840 extern void ext4_check_inodes_bitmap(struct super_block *);
1841 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1842 extern int ext4_init_inode_table(struct super_block *sb,
1843                                  ext4_group_t group, int barrier);
1844
1845 /* mballoc.c */
1846 extern long ext4_mb_stats;
1847 extern long ext4_mb_max_to_scan;
1848 extern int ext4_mb_init(struct super_block *, int);
1849 extern int ext4_mb_release(struct super_block *);
1850 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1851                                 struct ext4_allocation_request *, int *);
1852 extern int ext4_mb_reserve_blocks(struct super_block *, int);
1853 extern void ext4_discard_preallocations(struct inode *);
1854 extern int __init ext4_init_mballoc(void);
1855 extern void ext4_exit_mballoc(void);
1856 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1857                              struct buffer_head *bh, ext4_fsblk_t block,
1858                              unsigned long count, int flags);
1859 extern int ext4_mb_add_groupinfo(struct super_block *sb,
1860                 ext4_group_t i, struct ext4_group_desc *desc);
1861 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
1862                                 ext4_fsblk_t block, unsigned long count);
1863 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
1864
1865 /* inode.c */
1866 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1867                                                 ext4_lblk_t, int, int *);
1868 struct buffer_head *ext4_bread(handle_t *, struct inode *,
1869                                                 ext4_lblk_t, int, int *);
1870 int ext4_get_block(struct inode *inode, sector_t iblock,
1871                                 struct buffer_head *bh_result, int create);
1872
1873 extern struct inode *ext4_iget(struct super_block *, unsigned long);
1874 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
1875 extern int  ext4_setattr(struct dentry *, struct iattr *);
1876 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1877                                 struct kstat *stat);
1878 extern void ext4_evict_inode(struct inode *);
1879 extern void ext4_clear_inode(struct inode *);
1880 extern int  ext4_sync_inode(handle_t *, struct inode *);
1881 extern void ext4_dirty_inode(struct inode *, int);
1882 extern int ext4_change_inode_journal_flag(struct inode *, int);
1883 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1884 extern int ext4_can_truncate(struct inode *inode);
1885 extern void ext4_truncate(struct inode *);
1886 extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length);
1887 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
1888 extern void ext4_set_inode_flags(struct inode *);
1889 extern void ext4_get_inode_flags(struct ext4_inode_info *);
1890 extern int ext4_alloc_da_blocks(struct inode *inode);
1891 extern void ext4_set_aops(struct inode *inode);
1892 extern int ext4_writepage_trans_blocks(struct inode *);
1893 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1894 extern int ext4_discard_partial_page_buffers(handle_t *handle,
1895                 struct address_space *mapping, loff_t from,
1896                 loff_t length, int flags);
1897 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1898 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
1899 extern void ext4_da_update_reserve_space(struct inode *inode,
1900                                         int used, int quota_claim);
1901
1902 /* indirect.c */
1903 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
1904                                 struct ext4_map_blocks *map, int flags);
1905 extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
1906                                 const struct iovec *iov, loff_t offset,
1907                                 unsigned long nr_segs);
1908 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
1909 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks, int chunk);
1910 extern void ext4_ind_truncate(struct inode *inode);
1911
1912 /* ioctl.c */
1913 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1914 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1915
1916 /* migrate.c */
1917 extern int ext4_ext_migrate(struct inode *);
1918
1919 /* namei.c */
1920 extern int ext4_orphan_add(handle_t *, struct inode *);
1921 extern int ext4_orphan_del(handle_t *, struct inode *);
1922 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1923                                 __u32 start_minor_hash, __u32 *next_hash);
1924
1925 /* resize.c */
1926 extern int ext4_group_add(struct super_block *sb,
1927                                 struct ext4_new_group_data *input);
1928 extern int ext4_group_extend(struct super_block *sb,
1929                                 struct ext4_super_block *es,
1930                                 ext4_fsblk_t n_blocks_count);
1931 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
1932
1933 /* super.c */
1934 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
1935 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
1936 extern void ext4_kvfree(void *ptr);
1937 extern __printf(4, 5)
1938 void __ext4_error(struct super_block *, const char *, unsigned int,
1939                   const char *, ...);
1940 #define ext4_error(sb, message...)      __ext4_error(sb, __func__,      \
1941                                                      __LINE__, ## message)
1942 extern __printf(5, 6)
1943 void ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
1944                       const char *, ...);
1945 extern __printf(5, 6)
1946 void ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
1947                      const char *, ...);
1948 extern void __ext4_std_error(struct super_block *, const char *,
1949                              unsigned int, int);
1950 extern __printf(4, 5)
1951 void __ext4_abort(struct super_block *, const char *, unsigned int,
1952                   const char *, ...);
1953 #define ext4_abort(sb, message...)      __ext4_abort(sb, __func__, \
1954                                                        __LINE__, ## message)
1955 extern __printf(4, 5)
1956 void __ext4_warning(struct super_block *, const char *, unsigned int,
1957                     const char *, ...);
1958 #define ext4_warning(sb, message...)    __ext4_warning(sb, __func__, \
1959                                                        __LINE__, ## message)
1960 extern __printf(3, 4)
1961 void ext4_msg(struct super_block *, const char *, const char *, ...);
1962 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
1963                            const char *, unsigned int, const char *);
1964 #define dump_mmp_msg(sb, mmp, msg)      __dump_mmp_msg(sb, mmp, __func__, \
1965                                                        __LINE__, msg)
1966 extern __printf(7, 8)
1967 void __ext4_grp_locked_error(const char *, unsigned int,
1968                              struct super_block *, ext4_group_t,
1969                              unsigned long, ext4_fsblk_t,
1970                              const char *, ...);
1971 #define ext4_grp_locked_error(sb, grp, message...) \
1972         __ext4_grp_locked_error(__func__, __LINE__, (sb), (grp), ## message)
1973 extern void ext4_update_dynamic_rev(struct super_block *sb);
1974 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1975                                         __u32 compat);
1976 extern int ext4_update_rocompat_feature(handle_t *handle,
1977                                         struct super_block *sb, __u32 rocompat);
1978 extern int ext4_update_incompat_feature(handle_t *handle,
1979                                         struct super_block *sb, __u32 incompat);
1980 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1981                                       struct ext4_group_desc *bg);
1982 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1983                                       struct ext4_group_desc *bg);
1984 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1985                                      struct ext4_group_desc *bg);
1986 extern __u32 ext4_free_group_clusters(struct super_block *sb,
1987                                       struct ext4_group_desc *bg);
1988 extern __u32 ext4_free_inodes_count(struct super_block *sb,
1989                                  struct ext4_group_desc *bg);
1990 extern __u32 ext4_used_dirs_count(struct super_block *sb,
1991                                 struct ext4_group_desc *bg);
1992 extern __u32 ext4_itable_unused_count(struct super_block *sb,
1993                                    struct ext4_group_desc *bg);
1994 extern void ext4_block_bitmap_set(struct super_block *sb,
1995                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1996 extern void ext4_inode_bitmap_set(struct super_block *sb,
1997                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1998 extern void ext4_inode_table_set(struct super_block *sb,
1999                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
2000 extern void ext4_free_group_clusters_set(struct super_block *sb,
2001                                          struct ext4_group_desc *bg,
2002                                          __u32 count);
2003 extern void ext4_free_inodes_set(struct super_block *sb,
2004                                 struct ext4_group_desc *bg, __u32 count);
2005 extern void ext4_used_dirs_set(struct super_block *sb,
2006                                 struct ext4_group_desc *bg, __u32 count);
2007 extern void ext4_itable_unused_set(struct super_block *sb,
2008                                    struct ext4_group_desc *bg, __u32 count);
2009 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
2010                                    struct ext4_group_desc *gdp);
2011 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
2012                                        struct ext4_group_desc *gdp);
2013
2014 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2015 {
2016         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2017                 le32_to_cpu(es->s_blocks_count_lo);
2018 }
2019
2020 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2021 {
2022         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2023                 le32_to_cpu(es->s_r_blocks_count_lo);
2024 }
2025
2026 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2027 {
2028         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2029                 le32_to_cpu(es->s_free_blocks_count_lo);
2030 }
2031
2032 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2033                                          ext4_fsblk_t blk)
2034 {
2035         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2036         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2037 }
2038
2039 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2040                                               ext4_fsblk_t blk)
2041 {
2042         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2043         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2044 }
2045
2046 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2047                                            ext4_fsblk_t blk)
2048 {
2049         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2050         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2051 }
2052
2053 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2054 {
2055         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2056                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2057                         le32_to_cpu(raw_inode->i_size_lo);
2058         else
2059                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2060 }
2061
2062 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2063 {
2064         raw_inode->i_size_lo = cpu_to_le32(i_size);
2065         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2066 }
2067
2068 static inline
2069 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2070                                             ext4_group_t group)
2071 {
2072          struct ext4_group_info ***grp_info;
2073          long indexv, indexh;
2074          grp_info = EXT4_SB(sb)->s_group_info;
2075          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2076          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2077          return grp_info[indexv][indexh];
2078 }
2079
2080 /*
2081  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2082  * the locking protocol documented in the comments of ext4_group_add()
2083  * in resize.c
2084  */
2085 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2086 {
2087         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2088
2089         smp_rmb();
2090         return ngroups;
2091 }
2092
2093 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2094                                              ext4_group_t block_group)
2095 {
2096         return block_group >> sbi->s_log_groups_per_flex;
2097 }
2098
2099 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2100 {
2101         return 1 << sbi->s_log_groups_per_flex;
2102 }
2103
2104 #define ext4_std_error(sb, errno)                               \
2105 do {                                                            \
2106         if ((errno))                                            \
2107                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2108 } while (0)
2109
2110 #ifdef CONFIG_SMP
2111 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2112  * counters. So we need to make sure we have free clusters more
2113  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2114  */
2115 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2116 #else
2117 #define EXT4_FREECLUSTERS_WATERMARK 0
2118 #endif
2119
2120 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2121 {
2122         /*
2123          * XXX: replace with spinlock if seen contended -bzzz
2124          */
2125         down_write(&EXT4_I(inode)->i_data_sem);
2126         if (newsize > EXT4_I(inode)->i_disksize)
2127                 EXT4_I(inode)->i_disksize = newsize;
2128         up_write(&EXT4_I(inode)->i_data_sem);
2129         return ;
2130 }
2131
2132 struct ext4_group_info {
2133         unsigned long   bb_state;
2134         struct rb_root  bb_free_root;
2135         ext4_grpblk_t   bb_first_free;  /* first free block */
2136         ext4_grpblk_t   bb_free;        /* total free blocks */
2137         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2138         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2139         struct          list_head bb_prealloc_list;
2140 #ifdef DOUBLE_CHECK
2141         void            *bb_bitmap;
2142 #endif
2143         struct rw_semaphore alloc_sem;
2144         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2145                                          * regions, index is order.
2146                                          * bb_counters[3] = 5 means
2147                                          * 5 free 8-block regions. */
2148 };
2149
2150 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2151 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2152
2153 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2154         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2155
2156 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2157         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2158 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2159         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2160 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2161         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2162
2163 #define EXT4_MAX_CONTENTION             8
2164 #define EXT4_CONTENTION_THRESHOLD       2
2165
2166 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2167                                               ext4_group_t group)
2168 {
2169         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2170 }
2171
2172 /*
2173  * Returns true if the filesystem is busy enough that attempts to
2174  * access the block group locks has run into contention.
2175  */
2176 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2177 {
2178         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2179 }
2180
2181 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2182 {
2183         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2184         if (spin_trylock(lock))
2185                 /*
2186                  * We're able to grab the lock right away, so drop the
2187                  * lock contention counter.
2188                  */
2189                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2190         else {
2191                 /*
2192                  * The lock is busy, so bump the contention counter,
2193                  * and then wait on the spin lock.
2194                  */
2195                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2196                                   EXT4_MAX_CONTENTION);
2197                 spin_lock(lock);
2198         }
2199 }
2200
2201 static inline void ext4_unlock_group(struct super_block *sb,
2202                                         ext4_group_t group)
2203 {
2204         spin_unlock(ext4_group_lock_ptr(sb, group));
2205 }
2206
2207 static inline void ext4_mark_super_dirty(struct super_block *sb)
2208 {
2209         if (EXT4_SB(sb)->s_journal == NULL)
2210                 sb->s_dirt =1;
2211 }
2212
2213 /*
2214  * Block validity checking
2215  */
2216 #define ext4_check_indirect_blockref(inode, bh)                         \
2217         ext4_check_blockref(__func__, __LINE__, inode,                  \
2218                             (__le32 *)(bh)->b_data,                     \
2219                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2220
2221 #define ext4_ind_check_inode(inode)                                     \
2222         ext4_check_blockref(__func__, __LINE__, inode,                  \
2223                             EXT4_I(inode)->i_data,                      \
2224                             EXT4_NDIR_BLOCKS)
2225
2226 /*
2227  * Inodes and files operations
2228  */
2229
2230 /* dir.c */
2231 extern const struct file_operations ext4_dir_operations;
2232
2233 /* file.c */
2234 extern const struct inode_operations ext4_file_inode_operations;
2235 extern const struct file_operations ext4_file_operations;
2236 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2237
2238 /* namei.c */
2239 extern const struct inode_operations ext4_dir_inode_operations;
2240 extern const struct inode_operations ext4_special_inode_operations;
2241 extern struct dentry *ext4_get_parent(struct dentry *child);
2242
2243 /* symlink.c */
2244 extern const struct inode_operations ext4_symlink_inode_operations;
2245 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2246
2247 /* block_validity */
2248 extern void ext4_release_system_zone(struct super_block *sb);
2249 extern int ext4_setup_system_zone(struct super_block *sb);
2250 extern int __init ext4_init_system_zone(void);
2251 extern void ext4_exit_system_zone(void);
2252 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2253                                  ext4_fsblk_t start_blk,
2254                                  unsigned int count);
2255 extern int ext4_check_blockref(const char *, unsigned int,
2256                                struct inode *, __le32 *, unsigned int);
2257
2258 /* extents.c */
2259 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2260 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2261 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
2262                                        int chunk);
2263 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2264                                struct ext4_map_blocks *map, int flags);
2265 extern void ext4_ext_truncate(struct inode *);
2266 extern int ext4_ext_punch_hole(struct file *file, loff_t offset,
2267                                 loff_t length);
2268 extern void ext4_ext_init(struct super_block *);
2269 extern void ext4_ext_release(struct super_block *);
2270 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2271                           loff_t len);
2272 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
2273                           ssize_t len);
2274 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2275                            struct ext4_map_blocks *map, int flags);
2276 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2277                         __u64 start, __u64 len);
2278 /* move_extent.c */
2279 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2280                              __u64 start_orig, __u64 start_donor,
2281                              __u64 len, __u64 *moved_len);
2282
2283 /* page-io.c */
2284 extern int __init ext4_init_pageio(void);
2285 extern void ext4_exit_pageio(void);
2286 extern void ext4_ioend_wait(struct inode *);
2287 extern void ext4_free_io_end(ext4_io_end_t *io);
2288 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2289 extern int ext4_end_io_nolock(ext4_io_end_t *io);
2290 extern void ext4_io_submit(struct ext4_io_submit *io);
2291 extern int ext4_bio_write_page(struct ext4_io_submit *io,
2292                                struct page *page,
2293                                int len,
2294                                struct writeback_control *wbc);
2295
2296 /* mmp.c */
2297 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2298
2299 /* BH_Uninit flag: blocks are allocated but uninitialized on disk */
2300 enum ext4_state_bits {
2301         BH_Uninit       /* blocks are allocated but uninitialized on disk */
2302           = BH_JBDPrivateStart,
2303         BH_AllocFromCluster,    /* allocated blocks were part of already
2304                                  * allocated cluster. Note that this flag will
2305                                  * never, ever appear in a buffer_head's state
2306                                  * flag. See EXT4_MAP_FROM_CLUSTER to see where
2307                                  * this is used. */
2308         BH_Da_Mapped,   /* Delayed allocated block that now has a mapping. This
2309                          * flag is set when ext4_map_blocks is called on a
2310                          * delayed allocated block to get its real mapping. */
2311 };
2312
2313 BUFFER_FNS(Uninit, uninit)
2314 TAS_BUFFER_FNS(Uninit, uninit)
2315 BUFFER_FNS(Da_Mapped, da_mapped)
2316
2317 /*
2318  * Add new method to test wether block and inode bitmaps are properly
2319  * initialized. With uninit_bg reading the block from disk is not enough
2320  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2321  */
2322 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2323
2324 static inline int bitmap_uptodate(struct buffer_head *bh)
2325 {
2326         return (buffer_uptodate(bh) &&
2327                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2328 }
2329 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2330 {
2331         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2332 }
2333
2334 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2335
2336 /* For ioend & aio unwritten conversion wait queues */
2337 #define EXT4_WQ_HASH_SZ         37
2338 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
2339                                             EXT4_WQ_HASH_SZ])
2340 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
2341                                              EXT4_WQ_HASH_SZ])
2342 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2343 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2344
2345 #define EXT4_RESIZING   0
2346 extern int ext4_resize_begin(struct super_block *sb);
2347 extern void ext4_resize_end(struct super_block *sb);
2348
2349 #endif  /* __KERNEL__ */
2350
2351 #include "ext4_extents.h"
2352
2353 #endif  /* _EXT4_H */