Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
[sfrench/cifs-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
33 /*
34  * The fourth extended filesystem constants/structures
35  */
36
37 /*
38  * Define EXT4FS_DEBUG to produce debug messages
39  */
40 #undef EXT4FS_DEBUG
41
42 /*
43  * Debug code
44  */
45 #ifdef EXT4FS_DEBUG
46 #define ext4_debug(f, a...)                                             \
47         do {                                                            \
48                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
49                         __FILE__, __LINE__, __func__);                  \
50                 printk(KERN_DEBUG f, ## a);                             \
51         } while (0)
52 #else
53 #define ext4_debug(f, a...)     do {} while (0)
54 #endif
55
56 #define EXT4_ERROR_INODE(inode, fmt, a...) \
57         ext4_error_inode(__func__, (inode), (fmt), ## a);
58
59 #define EXT4_ERROR_FILE(file, fmt, a...)        \
60         ext4_error_file(__func__, (file), (fmt), ## a);
61
62 /* data type for block offset of block group */
63 typedef int ext4_grpblk_t;
64
65 /* data type for filesystem-wide blocks number */
66 typedef unsigned long long ext4_fsblk_t;
67
68 /* data type for file logical block number */
69 typedef __u32 ext4_lblk_t;
70
71 /* data type for block group number */
72 typedef unsigned int ext4_group_t;
73
74 /*
75  * Flags used in mballoc's allocation_context flags field.  
76  *
77  * Also used to show what's going on for debugging purposes when the
78  * flag field is exported via the traceport interface
79  */
80
81 /* prefer goal again. length */
82 #define EXT4_MB_HINT_MERGE              0x0001
83 /* blocks already reserved */
84 #define EXT4_MB_HINT_RESERVED           0x0002
85 /* metadata is being allocated */
86 #define EXT4_MB_HINT_METADATA           0x0004
87 /* first blocks in the file */
88 #define EXT4_MB_HINT_FIRST              0x0008
89 /* search for the best chunk */
90 #define EXT4_MB_HINT_BEST               0x0010
91 /* data is being allocated */
92 #define EXT4_MB_HINT_DATA               0x0020
93 /* don't preallocate (for tails) */
94 #define EXT4_MB_HINT_NOPREALLOC         0x0040
95 /* allocate for locality group */
96 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
97 /* allocate goal blocks or none */
98 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
99 /* goal is meaningful */
100 #define EXT4_MB_HINT_TRY_GOAL           0x0200
101 /* blocks already pre-reserved by delayed allocation */
102 #define EXT4_MB_DELALLOC_RESERVED       0x0400
103 /* We are doing stream allocation */
104 #define EXT4_MB_STREAM_ALLOC            0x0800
105
106
107 struct ext4_allocation_request {
108         /* target inode for block we're allocating */
109         struct inode *inode;
110         /* how many blocks we want to allocate */
111         unsigned int len;
112         /* logical block in target inode */
113         ext4_lblk_t logical;
114         /* the closest logical allocated block to the left */
115         ext4_lblk_t lleft;
116         /* the closest logical allocated block to the right */
117         ext4_lblk_t lright;
118         /* phys. target (a hint) */
119         ext4_fsblk_t goal;
120         /* phys. block for the closest logical allocated block to the left */
121         ext4_fsblk_t pleft;
122         /* phys. block for the closest logical allocated block to the right */
123         ext4_fsblk_t pright;
124         /* flags. see above EXT4_MB_HINT_* */
125         unsigned int flags;
126 };
127
128 /*
129  * For delayed allocation tracking
130  */
131 struct mpage_da_data {
132         struct inode *inode;
133         sector_t b_blocknr;             /* start block number of extent */
134         size_t b_size;                  /* size of extent */
135         unsigned long b_state;          /* state of the extent */
136         unsigned long first_page, next_page;    /* extent of pages */
137         struct writeback_control *wbc;
138         int io_done;
139         int pages_written;
140         int retval;
141 };
142 #define EXT4_IO_UNWRITTEN       0x1
143 typedef struct ext4_io_end {
144         struct list_head        list;           /* per-file finished AIO list */
145         struct inode            *inode;         /* file being written to */
146         unsigned int            flag;           /* unwritten or not */
147         struct page             *page;          /* page struct for buffer write */
148         loff_t                  offset;         /* offset in the file */
149         ssize_t                 size;           /* size of the extent */
150         struct work_struct      work;           /* data work queue */
151 } ext4_io_end_t;
152
153 /*
154  * Special inodes numbers
155  */
156 #define EXT4_BAD_INO             1      /* Bad blocks inode */
157 #define EXT4_ROOT_INO            2      /* Root inode */
158 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
159 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
160 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
161 #define EXT4_JOURNAL_INO         8      /* Journal inode */
162
163 /* First non-reserved inode for old ext4 filesystems */
164 #define EXT4_GOOD_OLD_FIRST_INO 11
165
166 /*
167  * Maximal count of links to a file
168  */
169 #define EXT4_LINK_MAX           65000
170
171 /*
172  * Macro-instructions used to manage several block sizes
173  */
174 #define EXT4_MIN_BLOCK_SIZE             1024
175 #define EXT4_MAX_BLOCK_SIZE             65536
176 #define EXT4_MIN_BLOCK_LOG_SIZE         10
177 #ifdef __KERNEL__
178 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
179 #else
180 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
181 #endif
182 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
183 #ifdef __KERNEL__
184 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
185 #else
186 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
187 #endif
188 #ifdef __KERNEL__
189 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
190 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
191 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
192 #else
193 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
194                                  EXT4_GOOD_OLD_INODE_SIZE : \
195                                  (s)->s_inode_size)
196 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
197                                  EXT4_GOOD_OLD_FIRST_INO : \
198                                  (s)->s_first_ino)
199 #endif
200 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
201
202 /*
203  * Structure of a blocks group descriptor
204  */
205 struct ext4_group_desc
206 {
207         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
208         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
209         __le32  bg_inode_table_lo;      /* Inodes table block */
210         __le16  bg_free_blocks_count_lo;/* Free blocks count */
211         __le16  bg_free_inodes_count_lo;/* Free inodes count */
212         __le16  bg_used_dirs_count_lo;  /* Directories count */
213         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
214         __u32   bg_reserved[2];         /* Likely block/inode bitmap checksum */
215         __le16  bg_itable_unused_lo;    /* Unused inodes count */
216         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
217         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
218         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
219         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
220         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
221         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
222         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
223         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
224         __u32   bg_reserved2[3];
225 };
226
227 /*
228  * Structure of a flex block group info
229  */
230
231 struct flex_groups {
232         atomic_t free_inodes;
233         atomic_t free_blocks;
234         atomic_t used_dirs;
235 };
236
237 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
238 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
239 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
240
241 /*
242  * Macro-instructions used to manage group descriptors
243  */
244 #define EXT4_MIN_DESC_SIZE              32
245 #define EXT4_MIN_DESC_SIZE_64BIT        64
246 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
247 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
248 #ifdef __KERNEL__
249 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
250 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
251 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
252 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
253 #else
254 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
255 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
256 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
257 #endif
258
259 /*
260  * Constants relative to the data blocks
261  */
262 #define EXT4_NDIR_BLOCKS                12
263 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
264 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
265 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
266 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
267
268 /*
269  * Inode flags
270  */
271 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
272 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
273 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
274 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
275 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
276 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
277 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
278 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
279 /* Reserved for compression usage... */
280 #define EXT4_DIRTY_FL                   0x00000100
281 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
282 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
283 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
284 /* End compression flags --- maybe not all used */
285 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
286 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
287 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
288 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
289 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
290 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
291 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
292 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
293 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
294 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
295 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
296
297 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
298 #define EXT4_FL_USER_MODIFIABLE         0x004B80FF /* User modifiable flags */
299
300 /* Flags that should be inherited by new inodes from their parent. */
301 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
302                            EXT4_SYNC_FL | EXT4_IMMUTABLE_FL | EXT4_APPEND_FL |\
303                            EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
304                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
305                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
306
307 /* Flags that are appropriate for regular files (all but dir-specific ones). */
308 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
309
310 /* Flags that are appropriate for non-directories/regular files. */
311 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
312
313 /* Mask out flags that are inappropriate for the given type of inode. */
314 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
315 {
316         if (S_ISDIR(mode))
317                 return flags;
318         else if (S_ISREG(mode))
319                 return flags & EXT4_REG_FLMASK;
320         else
321                 return flags & EXT4_OTHER_FLMASK;
322 }
323
324 /* Used to pass group descriptor data when online resize is done */
325 struct ext4_new_group_input {
326         __u32 group;            /* Group number for this data */
327         __u64 block_bitmap;     /* Absolute block number of block bitmap */
328         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
329         __u64 inode_table;      /* Absolute block number of inode table start */
330         __u32 blocks_count;     /* Total number of blocks in this group */
331         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
332         __u16 unused;
333 };
334
335 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
336 struct ext4_new_group_data {
337         __u32 group;
338         __u64 block_bitmap;
339         __u64 inode_bitmap;
340         __u64 inode_table;
341         __u32 blocks_count;
342         __u16 reserved_blocks;
343         __u16 unused;
344         __u32 free_blocks_count;
345 };
346
347 /*
348  * Flags used by ext4_get_blocks()
349  */
350         /* Allocate any needed blocks and/or convert an unitialized
351            extent to be an initialized ext4 */
352 #define EXT4_GET_BLOCKS_CREATE                  0x0001
353         /* Request the creation of an unitialized extent */
354 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
355 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
356                                                  EXT4_GET_BLOCKS_CREATE)
357         /* Caller is from the delayed allocation writeout path,
358            so set the magic i_delalloc_reserve_flag after taking the 
359            inode allocation semaphore for */
360 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
361         /* caller is from the direct IO path, request to creation of an
362         unitialized extents if not allocated, split the uninitialized
363         extent if blocks has been preallocated already*/
364 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
365 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
366 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
367                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
368         /* Convert extent to initialized after IO complete */
369 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
370                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
371
372 /*
373  * Flags used by ext4_free_blocks
374  */
375 #define EXT4_FREE_BLOCKS_METADATA       0x0001
376 #define EXT4_FREE_BLOCKS_FORGET         0x0002
377 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
378
379 /*
380  * ioctl commands
381  */
382 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
383 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
384 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
385 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
386 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
387 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
388 #ifdef CONFIG_JBD2_DEBUG
389 #define EXT4_IOC_WAIT_FOR_READONLY      _IOR('f', 99, long)
390 #endif
391 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
392 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
393 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
394 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
395 #define EXT4_IOC_MIGRATE                _IO('f', 9)
396  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
397  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
398 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
399 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
400
401 /*
402  * ioctl commands in 32 bit emulation
403  */
404 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
405 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
406 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
407 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
408 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
409 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
410 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
411 #ifdef CONFIG_JBD2_DEBUG
412 #define EXT4_IOC32_WAIT_FOR_READONLY    _IOR('f', 99, int)
413 #endif
414 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
415 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
416
417
418 /*
419  *  Mount options
420  */
421 struct ext4_mount_options {
422         unsigned long s_mount_opt;
423         uid_t s_resuid;
424         gid_t s_resgid;
425         unsigned long s_commit_interval;
426         u32 s_min_batch_time, s_max_batch_time;
427 #ifdef CONFIG_QUOTA
428         int s_jquota_fmt;
429         char *s_qf_names[MAXQUOTAS];
430 #endif
431 };
432
433 /* Max physical block we can addres w/o extents */
434 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
435
436 /*
437  * Structure of an inode on the disk
438  */
439 struct ext4_inode {
440         __le16  i_mode;         /* File mode */
441         __le16  i_uid;          /* Low 16 bits of Owner Uid */
442         __le32  i_size_lo;      /* Size in bytes */
443         __le32  i_atime;        /* Access time */
444         __le32  i_ctime;        /* Inode Change time */
445         __le32  i_mtime;        /* Modification time */
446         __le32  i_dtime;        /* Deletion Time */
447         __le16  i_gid;          /* Low 16 bits of Group Id */
448         __le16  i_links_count;  /* Links count */
449         __le32  i_blocks_lo;    /* Blocks count */
450         __le32  i_flags;        /* File flags */
451         union {
452                 struct {
453                         __le32  l_i_version;
454                 } linux1;
455                 struct {
456                         __u32  h_i_translator;
457                 } hurd1;
458                 struct {
459                         __u32  m_i_reserved1;
460                 } masix1;
461         } osd1;                         /* OS dependent 1 */
462         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
463         __le32  i_generation;   /* File version (for NFS) */
464         __le32  i_file_acl_lo;  /* File ACL */
465         __le32  i_size_high;
466         __le32  i_obso_faddr;   /* Obsoleted fragment address */
467         union {
468                 struct {
469                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
470                         __le16  l_i_file_acl_high;
471                         __le16  l_i_uid_high;   /* these 2 fields */
472                         __le16  l_i_gid_high;   /* were reserved2[0] */
473                         __u32   l_i_reserved2;
474                 } linux2;
475                 struct {
476                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
477                         __u16   h_i_mode_high;
478                         __u16   h_i_uid_high;
479                         __u16   h_i_gid_high;
480                         __u32   h_i_author;
481                 } hurd2;
482                 struct {
483                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
484                         __le16  m_i_file_acl_high;
485                         __u32   m_i_reserved2[2];
486                 } masix2;
487         } osd2;                         /* OS dependent 2 */
488         __le16  i_extra_isize;
489         __le16  i_pad1;
490         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
491         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
492         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
493         __le32  i_crtime;       /* File Creation time */
494         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
495         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
496 };
497
498 struct move_extent {
499         __u32 reserved;         /* should be zero */
500         __u32 donor_fd;         /* donor file descriptor */
501         __u64 orig_start;       /* logical start offset in block for orig */
502         __u64 donor_start;      /* logical start offset in block for donor */
503         __u64 len;              /* block length to be moved */
504         __u64 moved_len;        /* moved block length */
505 };
506
507 #define EXT4_EPOCH_BITS 2
508 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
509 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
510
511 /*
512  * Extended fields will fit into an inode if the filesystem was formatted
513  * with large inodes (-I 256 or larger) and there are not currently any EAs
514  * consuming all of the available space. For new inodes we always reserve
515  * enough space for the kernel's known extended fields, but for inodes
516  * created with an old kernel this might not have been the case. None of
517  * the extended inode fields is critical for correct filesystem operation.
518  * This macro checks if a certain field fits in the inode. Note that
519  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
520  */
521 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
522         ((offsetof(typeof(*ext4_inode), field) +        \
523           sizeof((ext4_inode)->field))                  \
524         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
525             (einode)->i_extra_isize))                   \
526
527 static inline __le32 ext4_encode_extra_time(struct timespec *time)
528 {
529        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
530                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
531                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
532 }
533
534 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
535 {
536        if (sizeof(time->tv_sec) > 4)
537                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
538                                << 32;
539        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
540 }
541
542 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
543 do {                                                                           \
544         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
545         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
546                 (raw_inode)->xtime ## _extra =                                 \
547                                 ext4_encode_extra_time(&(inode)->xtime);       \
548 } while (0)
549
550 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
551 do {                                                                           \
552         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
553                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
554         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
555                 (raw_inode)->xtime ## _extra =                                 \
556                                 ext4_encode_extra_time(&(einode)->xtime);      \
557 } while (0)
558
559 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
560 do {                                                                           \
561         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
562         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
563                 ext4_decode_extra_time(&(inode)->xtime,                        \
564                                        raw_inode->xtime ## _extra);            \
565 } while (0)
566
567 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
568 do {                                                                           \
569         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
570                 (einode)->xtime.tv_sec =                                       \
571                         (signed)le32_to_cpu((raw_inode)->xtime);               \
572         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
573                 ext4_decode_extra_time(&(einode)->xtime,                       \
574                                        raw_inode->xtime ## _extra);            \
575 } while (0)
576
577 #define i_disk_version osd1.linux1.l_i_version
578
579 #if defined(__KERNEL__) || defined(__linux__)
580 #define i_reserved1     osd1.linux1.l_i_reserved1
581 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
582 #define i_blocks_high   osd2.linux2.l_i_blocks_high
583 #define i_uid_low       i_uid
584 #define i_gid_low       i_gid
585 #define i_uid_high      osd2.linux2.l_i_uid_high
586 #define i_gid_high      osd2.linux2.l_i_gid_high
587 #define i_reserved2     osd2.linux2.l_i_reserved2
588
589 #elif defined(__GNU__)
590
591 #define i_translator    osd1.hurd1.h_i_translator
592 #define i_uid_high      osd2.hurd2.h_i_uid_high
593 #define i_gid_high      osd2.hurd2.h_i_gid_high
594 #define i_author        osd2.hurd2.h_i_author
595
596 #elif defined(__masix__)
597
598 #define i_reserved1     osd1.masix1.m_i_reserved1
599 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
600 #define i_reserved2     osd2.masix2.m_i_reserved2
601
602 #endif /* defined(__KERNEL__) || defined(__linux__) */
603
604 /*
605  * storage for cached extent
606  */
607 struct ext4_ext_cache {
608         ext4_fsblk_t    ec_start;
609         ext4_lblk_t     ec_block;
610         __u32           ec_len; /* must be 32bit to return holes */
611         __u32           ec_type;
612 };
613
614 /*
615  * fourth extended file system inode data in memory
616  */
617 struct ext4_inode_info {
618         __le32  i_data[15];     /* unconverted */
619         __u32   i_flags;
620         ext4_fsblk_t    i_file_acl;
621         __u32   i_dtime;
622
623         /*
624          * i_block_group is the number of the block group which contains
625          * this file's inode.  Constant across the lifetime of the inode,
626          * it is ued for making block allocation decisions - we try to
627          * place a file's data blocks near its inode block, and new inodes
628          * near to their parent directory's inode.
629          */
630         ext4_group_t    i_block_group;
631         unsigned long   i_state_flags;          /* Dynamic state flags */
632
633         ext4_lblk_t             i_dir_start_lookup;
634 #ifdef CONFIG_EXT4_FS_XATTR
635         /*
636          * Extended attributes can be read independently of the main file
637          * data. Taking i_mutex even when reading would cause contention
638          * between readers of EAs and writers of regular file data, so
639          * instead we synchronize on xattr_sem when reading or changing
640          * EAs.
641          */
642         struct rw_semaphore xattr_sem;
643 #endif
644
645         struct list_head i_orphan;      /* unlinked but open inodes */
646
647         /*
648          * i_disksize keeps track of what the inode size is ON DISK, not
649          * in memory.  During truncate, i_size is set to the new size by
650          * the VFS prior to calling ext4_truncate(), but the filesystem won't
651          * set i_disksize to 0 until the truncate is actually under way.
652          *
653          * The intent is that i_disksize always represents the blocks which
654          * are used by this file.  This allows recovery to restart truncate
655          * on orphans if we crash during truncate.  We actually write i_disksize
656          * into the on-disk inode when writing inodes out, instead of i_size.
657          *
658          * The only time when i_disksize and i_size may be different is when
659          * a truncate is in progress.  The only things which change i_disksize
660          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
661          */
662         loff_t  i_disksize;
663
664         /*
665          * i_data_sem is for serialising ext4_truncate() against
666          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
667          * data tree are chopped off during truncate. We can't do that in
668          * ext4 because whenever we perform intermediate commits during
669          * truncate, the inode and all the metadata blocks *must* be in a
670          * consistent state which allows truncation of the orphans to restart
671          * during recovery.  Hence we must fix the get_block-vs-truncate race
672          * by other means, so we have i_data_sem.
673          */
674         struct rw_semaphore i_data_sem;
675         struct inode vfs_inode;
676         struct jbd2_inode jinode;
677
678         struct ext4_ext_cache i_cached_extent;
679         /*
680          * File creation time. Its function is same as that of
681          * struct timespec i_{a,c,m}time in the generic inode.
682          */
683         struct timespec i_crtime;
684
685         /* mballoc */
686         struct list_head i_prealloc_list;
687         spinlock_t i_prealloc_lock;
688
689         /* ialloc */
690         ext4_group_t    i_last_alloc_group;
691
692         /* allocation reservation info for delalloc */
693         unsigned int i_reserved_data_blocks;
694         unsigned int i_reserved_meta_blocks;
695         unsigned int i_allocated_meta_blocks;
696         unsigned short i_delalloc_reserved_flag;
697         sector_t i_da_metadata_calc_last_lblock;
698         int i_da_metadata_calc_len;
699
700         /* on-disk additional length */
701         __u16 i_extra_isize;
702
703         spinlock_t i_block_reservation_lock;
704 #ifdef CONFIG_QUOTA
705         /* quota space reservation, managed internally by quota code */
706         qsize_t i_reserved_quota;
707 #endif
708
709         /* completed IOs that might need unwritten extents handling */
710         struct list_head i_completed_io_list;
711         spinlock_t i_completed_io_lock;
712         /* current io_end structure for async DIO write*/
713         ext4_io_end_t *cur_aio_dio;
714
715         /*
716          * Transactions that contain inode's metadata needed to complete
717          * fsync and fdatasync, respectively.
718          */
719         tid_t i_sync_tid;
720         tid_t i_datasync_tid;
721 };
722
723 /*
724  * File system states
725  */
726 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
727 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
728 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
729
730 /*
731  * Misc. filesystem flags
732  */
733 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
734 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
735 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
736
737 /*
738  * Mount flags
739  */
740 #define EXT4_MOUNT_OLDALLOC             0x00002  /* Don't use the new Orlov allocator */
741 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
742 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
743 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
744 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
745 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
746 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
747 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
748 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
749 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
750 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
751 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
752 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
753 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
754 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
755 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
756 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
757 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
758 #define EXT4_MOUNT_NOBH                 0x40000 /* No bufferheads */
759 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
760 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
761 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
762 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
763 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
764 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
765 #define EXT4_MOUNT_I_VERSION            0x2000000 /* i_version support */
766 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
767 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
768 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
769 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
770
771 #define clear_opt(o, opt)               o &= ~EXT4_MOUNT_##opt
772 #define set_opt(o, opt)                 o |= EXT4_MOUNT_##opt
773 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
774                                          EXT4_MOUNT_##opt)
775
776 #define ext4_set_bit                    ext2_set_bit
777 #define ext4_set_bit_atomic             ext2_set_bit_atomic
778 #define ext4_clear_bit                  ext2_clear_bit
779 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
780 #define ext4_test_bit                   ext2_test_bit
781 #define ext4_find_first_zero_bit        ext2_find_first_zero_bit
782 #define ext4_find_next_zero_bit         ext2_find_next_zero_bit
783 #define ext4_find_next_bit              ext2_find_next_bit
784
785 /*
786  * Maximal mount counts between two filesystem checks
787  */
788 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
789 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
790
791 /*
792  * Behaviour when detecting errors
793  */
794 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
795 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
796 #define EXT4_ERRORS_PANIC               3       /* Panic */
797 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
798
799 /*
800  * Structure of the super block
801  */
802 struct ext4_super_block {
803 /*00*/  __le32  s_inodes_count;         /* Inodes count */
804         __le32  s_blocks_count_lo;      /* Blocks count */
805         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
806         __le32  s_free_blocks_count_lo; /* Free blocks count */
807 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
808         __le32  s_first_data_block;     /* First Data Block */
809         __le32  s_log_block_size;       /* Block size */
810         __le32  s_obso_log_frag_size;   /* Obsoleted fragment size */
811 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
812         __le32  s_obso_frags_per_group; /* Obsoleted fragments per group */
813         __le32  s_inodes_per_group;     /* # Inodes per group */
814         __le32  s_mtime;                /* Mount time */
815 /*30*/  __le32  s_wtime;                /* Write time */
816         __le16  s_mnt_count;            /* Mount count */
817         __le16  s_max_mnt_count;        /* Maximal mount count */
818         __le16  s_magic;                /* Magic signature */
819         __le16  s_state;                /* File system state */
820         __le16  s_errors;               /* Behaviour when detecting errors */
821         __le16  s_minor_rev_level;      /* minor revision level */
822 /*40*/  __le32  s_lastcheck;            /* time of last check */
823         __le32  s_checkinterval;        /* max. time between checks */
824         __le32  s_creator_os;           /* OS */
825         __le32  s_rev_level;            /* Revision level */
826 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
827         __le16  s_def_resgid;           /* Default gid for reserved blocks */
828         /*
829          * These fields are for EXT4_DYNAMIC_REV superblocks only.
830          *
831          * Note: the difference between the compatible feature set and
832          * the incompatible feature set is that if there is a bit set
833          * in the incompatible feature set that the kernel doesn't
834          * know about, it should refuse to mount the filesystem.
835          *
836          * e2fsck's requirements are more strict; if it doesn't know
837          * about a feature in either the compatible or incompatible
838          * feature set, it must abort and not try to meddle with
839          * things it doesn't understand...
840          */
841         __le32  s_first_ino;            /* First non-reserved inode */
842         __le16  s_inode_size;           /* size of inode structure */
843         __le16  s_block_group_nr;       /* block group # of this superblock */
844         __le32  s_feature_compat;       /* compatible feature set */
845 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
846         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
847 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
848 /*78*/  char    s_volume_name[16];      /* volume name */
849 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
850 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
851         /*
852          * Performance hints.  Directory preallocation should only
853          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
854          */
855         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
856         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
857         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
858         /*
859          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
860          */
861 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
862 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
863         __le32  s_journal_dev;          /* device number of journal file */
864         __le32  s_last_orphan;          /* start of list of inodes to delete */
865         __le32  s_hash_seed[4];         /* HTREE hash seed */
866         __u8    s_def_hash_version;     /* Default hash version to use */
867         __u8    s_reserved_char_pad;
868         __le16  s_desc_size;            /* size of group descriptor */
869 /*100*/ __le32  s_default_mount_opts;
870         __le32  s_first_meta_bg;        /* First metablock block group */
871         __le32  s_mkfs_time;            /* When the filesystem was created */
872         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
873         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
874 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
875         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
876         __le32  s_free_blocks_count_hi; /* Free blocks count */
877         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
878         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
879         __le32  s_flags;                /* Miscellaneous flags */
880         __le16  s_raid_stride;          /* RAID stride */
881         __le16  s_mmp_interval;         /* # seconds to wait in MMP checking */
882         __le64  s_mmp_block;            /* Block for multi-mount protection */
883         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
884         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
885         __u8    s_reserved_char_pad2;
886         __le16  s_reserved_pad;
887         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
888         __u32   s_reserved[160];        /* Padding to the end of the block */
889 };
890
891 #ifdef __KERNEL__
892
893 /*
894  * run-time mount flags
895  */
896 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
897 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
898
899 /*
900  * fourth extended-fs super-block data in memory
901  */
902 struct ext4_sb_info {
903         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
904         unsigned long s_inodes_per_block;/* Number of inodes per block */
905         unsigned long s_blocks_per_group;/* Number of blocks in a group */
906         unsigned long s_inodes_per_group;/* Number of inodes in a group */
907         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
908         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
909         unsigned long s_desc_per_block; /* Number of group descriptors per block */
910         ext4_group_t s_groups_count;    /* Number of groups in the fs */
911         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
912         unsigned long s_overhead_last;  /* Last calculated overhead */
913         unsigned long s_blocks_last;    /* Last seen block count */
914         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
915         struct buffer_head * s_sbh;     /* Buffer containing the super block */
916         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
917         struct buffer_head **s_group_desc;
918         unsigned int s_mount_opt;
919         unsigned int s_mount_flags;
920         ext4_fsblk_t s_sb_block;
921         uid_t s_resuid;
922         gid_t s_resgid;
923         unsigned short s_mount_state;
924         unsigned short s_pad;
925         int s_addr_per_block_bits;
926         int s_desc_per_block_bits;
927         int s_inode_size;
928         int s_first_ino;
929         unsigned int s_inode_readahead_blks;
930         unsigned int s_inode_goal;
931         spinlock_t s_next_gen_lock;
932         u32 s_next_generation;
933         u32 s_hash_seed[4];
934         int s_def_hash_version;
935         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
936         struct percpu_counter s_freeblocks_counter;
937         struct percpu_counter s_freeinodes_counter;
938         struct percpu_counter s_dirs_counter;
939         struct percpu_counter s_dirtyblocks_counter;
940         struct blockgroup_lock *s_blockgroup_lock;
941         struct proc_dir_entry *s_proc;
942         struct kobject s_kobj;
943         struct completion s_kobj_unregister;
944
945         /* Journaling */
946         struct inode *s_journal_inode;
947         struct journal_s *s_journal;
948         struct list_head s_orphan;
949         struct mutex s_orphan_lock;
950         struct mutex s_resize_lock;
951         unsigned long s_commit_interval;
952         u32 s_max_batch_time;
953         u32 s_min_batch_time;
954         struct block_device *journal_bdev;
955 #ifdef CONFIG_JBD2_DEBUG
956         struct timer_list turn_ro_timer;        /* For turning read-only (crash simulation) */
957         wait_queue_head_t ro_wait_queue;        /* For people waiting for the fs to go read-only */
958 #endif
959 #ifdef CONFIG_QUOTA
960         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
961         int s_jquota_fmt;                       /* Format of quota to use */
962 #endif
963         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
964         struct rb_root system_blks;
965
966 #ifdef EXTENTS_STATS
967         /* ext4 extents stats */
968         unsigned long s_ext_min;
969         unsigned long s_ext_max;
970         unsigned long s_depth_max;
971         spinlock_t s_ext_stats_lock;
972         unsigned long s_ext_blocks;
973         unsigned long s_ext_extents;
974 #endif
975
976         /* for buddy allocator */
977         struct ext4_group_info ***s_group_info;
978         struct inode *s_buddy_cache;
979         long s_blocks_reserved;
980         spinlock_t s_reserve_lock;
981         spinlock_t s_md_lock;
982         tid_t s_last_transaction;
983         unsigned short *s_mb_offsets;
984         unsigned int *s_mb_maxs;
985
986         /* tunables */
987         unsigned long s_stripe;
988         unsigned int s_mb_stream_request;
989         unsigned int s_mb_max_to_scan;
990         unsigned int s_mb_min_to_scan;
991         unsigned int s_mb_stats;
992         unsigned int s_mb_order2_reqs;
993         unsigned int s_mb_group_prealloc;
994         unsigned int s_max_writeback_mb_bump;
995         /* where last allocation was done - for stream allocation */
996         unsigned long s_mb_last_group;
997         unsigned long s_mb_last_start;
998
999         /* stats for buddy allocator */
1000         spinlock_t s_mb_pa_lock;
1001         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1002         atomic_t s_bal_success; /* we found long enough chunks */
1003         atomic_t s_bal_allocated;       /* in blocks */
1004         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1005         atomic_t s_bal_goals;   /* goal hits */
1006         atomic_t s_bal_breaks;  /* too long searches */
1007         atomic_t s_bal_2orders; /* 2^order hits */
1008         spinlock_t s_bal_lock;
1009         unsigned long s_mb_buddies_generated;
1010         unsigned long long s_mb_generation_time;
1011         atomic_t s_mb_lost_chunks;
1012         atomic_t s_mb_preallocated;
1013         atomic_t s_mb_discarded;
1014         atomic_t s_lock_busy;
1015
1016         /* locality groups */
1017         struct ext4_locality_group __percpu *s_locality_groups;
1018
1019         /* for write statistics */
1020         unsigned long s_sectors_written_start;
1021         u64 s_kbytes_written;
1022
1023         unsigned int s_log_groups_per_flex;
1024         struct flex_groups *s_flex_groups;
1025
1026         /* workqueue for dio unwritten */
1027         struct workqueue_struct *dio_unwritten_wq;
1028 };
1029
1030 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1031 {
1032         return sb->s_fs_info;
1033 }
1034 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1035 {
1036         return container_of(inode, struct ext4_inode_info, vfs_inode);
1037 }
1038
1039 static inline struct timespec ext4_current_time(struct inode *inode)
1040 {
1041         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1042                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1043 }
1044
1045 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1046 {
1047         return ino == EXT4_ROOT_INO ||
1048                 ino == EXT4_JOURNAL_INO ||
1049                 ino == EXT4_RESIZE_INO ||
1050                 (ino >= EXT4_FIRST_INO(sb) &&
1051                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1052 }
1053
1054 /*
1055  * Inode dynamic state flags
1056  */
1057 enum {
1058         EXT4_STATE_JDATA,               /* journaled data exists */
1059         EXT4_STATE_NEW,                 /* inode is newly created */
1060         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1061         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1062         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1063         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1064         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1065 };
1066
1067 static inline int ext4_test_inode_state(struct inode *inode, int bit)
1068 {
1069         return test_bit(bit, &EXT4_I(inode)->i_state_flags);
1070 }
1071
1072 static inline void ext4_set_inode_state(struct inode *inode, int bit)
1073 {
1074         set_bit(bit, &EXT4_I(inode)->i_state_flags);
1075 }
1076
1077 static inline void ext4_clear_inode_state(struct inode *inode, int bit)
1078 {
1079         clear_bit(bit, &EXT4_I(inode)->i_state_flags);
1080 }
1081 #else
1082 /* Assume that user mode programs are passing in an ext4fs superblock, not
1083  * a kernel struct super_block.  This will allow us to call the feature-test
1084  * macros from user land. */
1085 #define EXT4_SB(sb)     (sb)
1086 #endif
1087
1088 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1089
1090 /*
1091  * Codes for operating systems
1092  */
1093 #define EXT4_OS_LINUX           0
1094 #define EXT4_OS_HURD            1
1095 #define EXT4_OS_MASIX           2
1096 #define EXT4_OS_FREEBSD         3
1097 #define EXT4_OS_LITES           4
1098
1099 /*
1100  * Revision levels
1101  */
1102 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1103 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1104
1105 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1106 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1107
1108 #define EXT4_GOOD_OLD_INODE_SIZE 128
1109
1110 /*
1111  * Feature set definitions
1112  */
1113
1114 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1115         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1116 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1117         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1118 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1119         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1120 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1121         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1122 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1123         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1124 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1125         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1126 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1127         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1128 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1129         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1130 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1131         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1132
1133 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1134 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1135 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1136 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1137 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1138 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1139
1140 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1141 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1142 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1143 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1144 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1145 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1146 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1147
1148 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1149 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1150 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1151 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1152 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1153 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1154 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1155 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1156 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1157 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1158 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1159
1160 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1161 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1162                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1163                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1164                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1165                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1166                                          EXT4_FEATURE_INCOMPAT_FLEX_BG)
1167 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1168                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1169                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1170                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1171                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1172                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1173                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE)
1174
1175 /*
1176  * Default values for user and/or group using reserved blocks
1177  */
1178 #define EXT4_DEF_RESUID         0
1179 #define EXT4_DEF_RESGID         0
1180
1181 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1182
1183 /*
1184  * Default mount options
1185  */
1186 #define EXT4_DEFM_DEBUG         0x0001
1187 #define EXT4_DEFM_BSDGROUPS     0x0002
1188 #define EXT4_DEFM_XATTR_USER    0x0004
1189 #define EXT4_DEFM_ACL           0x0008
1190 #define EXT4_DEFM_UID16         0x0010
1191 #define EXT4_DEFM_JMODE         0x0060
1192 #define EXT4_DEFM_JMODE_DATA    0x0020
1193 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1194 #define EXT4_DEFM_JMODE_WBACK   0x0060
1195
1196 /*
1197  * Default journal batch times
1198  */
1199 #define EXT4_DEF_MIN_BATCH_TIME 0
1200 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1201
1202 /*
1203  * Minimum number of groups in a flexgroup before we separate out
1204  * directories into the first block group of a flexgroup
1205  */
1206 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1207
1208 /*
1209  * Structure of a directory entry
1210  */
1211 #define EXT4_NAME_LEN 255
1212
1213 struct ext4_dir_entry {
1214         __le32  inode;                  /* Inode number */
1215         __le16  rec_len;                /* Directory entry length */
1216         __le16  name_len;               /* Name length */
1217         char    name[EXT4_NAME_LEN];    /* File name */
1218 };
1219
1220 /*
1221  * The new version of the directory entry.  Since EXT4 structures are
1222  * stored in intel byte order, and the name_len field could never be
1223  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1224  * file_type field.
1225  */
1226 struct ext4_dir_entry_2 {
1227         __le32  inode;                  /* Inode number */
1228         __le16  rec_len;                /* Directory entry length */
1229         __u8    name_len;               /* Name length */
1230         __u8    file_type;
1231         char    name[EXT4_NAME_LEN];    /* File name */
1232 };
1233
1234 /*
1235  * Ext4 directory file types.  Only the low 3 bits are used.  The
1236  * other bits are reserved for now.
1237  */
1238 #define EXT4_FT_UNKNOWN         0
1239 #define EXT4_FT_REG_FILE        1
1240 #define EXT4_FT_DIR             2
1241 #define EXT4_FT_CHRDEV          3
1242 #define EXT4_FT_BLKDEV          4
1243 #define EXT4_FT_FIFO            5
1244 #define EXT4_FT_SOCK            6
1245 #define EXT4_FT_SYMLINK         7
1246
1247 #define EXT4_FT_MAX             8
1248
1249 /*
1250  * EXT4_DIR_PAD defines the directory entries boundaries
1251  *
1252  * NOTE: It must be a multiple of 4
1253  */
1254 #define EXT4_DIR_PAD                    4
1255 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1256 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1257                                          ~EXT4_DIR_ROUND)
1258 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1259
1260 /*
1261  * Hash Tree Directory indexing
1262  * (c) Daniel Phillips, 2001
1263  */
1264
1265 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1266                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1267                       (EXT4_I(dir)->i_flags & EXT4_INDEX_FL))
1268 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1269 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1270
1271 /* Legal values for the dx_root hash_version field: */
1272
1273 #define DX_HASH_LEGACY          0
1274 #define DX_HASH_HALF_MD4        1
1275 #define DX_HASH_TEA             2
1276 #define DX_HASH_LEGACY_UNSIGNED 3
1277 #define DX_HASH_HALF_MD4_UNSIGNED       4
1278 #define DX_HASH_TEA_UNSIGNED            5
1279
1280 #ifdef __KERNEL__
1281
1282 /* hash info structure used by the directory hash */
1283 struct dx_hash_info
1284 {
1285         u32             hash;
1286         u32             minor_hash;
1287         int             hash_version;
1288         u32             *seed;
1289 };
1290
1291 #define EXT4_HTREE_EOF  0x7fffffff
1292
1293 /*
1294  * Control parameters used by ext4_htree_next_block
1295  */
1296 #define HASH_NB_ALWAYS          1
1297
1298
1299 /*
1300  * Describe an inode's exact location on disk and in memory
1301  */
1302 struct ext4_iloc
1303 {
1304         struct buffer_head *bh;
1305         unsigned long offset;
1306         ext4_group_t block_group;
1307 };
1308
1309 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1310 {
1311         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1312 }
1313
1314 /*
1315  * This structure is stuffed into the struct file's private_data field
1316  * for directories.  It is where we put information so that we can do
1317  * readdir operations in hash tree order.
1318  */
1319 struct dir_private_info {
1320         struct rb_root  root;
1321         struct rb_node  *curr_node;
1322         struct fname    *extra_fname;
1323         loff_t          last_pos;
1324         __u32           curr_hash;
1325         __u32           curr_minor_hash;
1326         __u32           next_hash;
1327 };
1328
1329 /* calculate the first block number of the group */
1330 static inline ext4_fsblk_t
1331 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1332 {
1333         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1334                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1335 }
1336
1337 /*
1338  * Special error return code only used by dx_probe() and its callers.
1339  */
1340 #define ERR_BAD_DX_DIR  -75000
1341
1342 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1343                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1344
1345 extern struct proc_dir_entry *ext4_proc_root;
1346
1347 /*
1348  * Function prototypes
1349  */
1350
1351 /*
1352  * Ok, these declarations are also in <linux/kernel.h> but none of the
1353  * ext4 source programs needs to include it so they are duplicated here.
1354  */
1355 # define NORET_TYPE     /**/
1356 # define ATTRIB_NORET   __attribute__((noreturn))
1357 # define NORET_AND      noreturn,
1358
1359 /* bitmap.c */
1360 extern unsigned int ext4_count_free(struct buffer_head *, unsigned);
1361
1362 /* balloc.c */
1363 extern unsigned int ext4_block_group(struct super_block *sb,
1364                         ext4_fsblk_t blocknr);
1365 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1366                         ext4_fsblk_t blocknr);
1367 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1368 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1369                         ext4_group_t group);
1370 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1371                         ext4_fsblk_t goal, unsigned long *count, int *errp);
1372 extern int ext4_claim_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1373 extern int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1374 extern void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
1375                                 ext4_fsblk_t block, unsigned long count);
1376 extern ext4_fsblk_t ext4_count_free_blocks(struct super_block *);
1377 extern void ext4_check_blocks_bitmap(struct super_block *);
1378 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1379                                                     ext4_group_t block_group,
1380                                                     struct buffer_head ** bh);
1381 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1382 struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1383                                       ext4_group_t block_group);
1384 extern unsigned ext4_init_block_bitmap(struct super_block *sb,
1385                                        struct buffer_head *bh,
1386                                        ext4_group_t group,
1387                                        struct ext4_group_desc *desc);
1388 #define ext4_free_blocks_after_init(sb, group, desc)                    \
1389                 ext4_init_block_bitmap(sb, NULL, group, desc)
1390
1391 /* dir.c */
1392 extern int ext4_check_dir_entry(const char *, struct inode *,
1393                                 struct ext4_dir_entry_2 *,
1394                                 struct buffer_head *, unsigned int);
1395 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1396                                     __u32 minor_hash,
1397                                     struct ext4_dir_entry_2 *dirent);
1398 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1399
1400 /* fsync.c */
1401 extern int ext4_sync_file(struct file *, struct dentry *, int);
1402
1403 /* hash.c */
1404 extern int ext4fs_dirhash(const char *name, int len, struct
1405                           dx_hash_info *hinfo);
1406
1407 /* ialloc.c */
1408 extern struct inode *ext4_new_inode(handle_t *, struct inode *, int,
1409                                     const struct qstr *qstr, __u32 goal);
1410 extern void ext4_free_inode(handle_t *, struct inode *);
1411 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1412 extern unsigned long ext4_count_free_inodes(struct super_block *);
1413 extern unsigned long ext4_count_dirs(struct super_block *);
1414 extern void ext4_check_inodes_bitmap(struct super_block *);
1415 extern unsigned ext4_init_inode_bitmap(struct super_block *sb,
1416                                        struct buffer_head *bh,
1417                                        ext4_group_t group,
1418                                        struct ext4_group_desc *desc);
1419 extern void mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1420
1421 /* mballoc.c */
1422 extern long ext4_mb_stats;
1423 extern long ext4_mb_max_to_scan;
1424 extern int ext4_mb_init(struct super_block *, int);
1425 extern int ext4_mb_release(struct super_block *);
1426 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1427                                 struct ext4_allocation_request *, int *);
1428 extern int ext4_mb_reserve_blocks(struct super_block *, int);
1429 extern void ext4_discard_preallocations(struct inode *);
1430 extern int __init init_ext4_mballoc(void);
1431 extern void exit_ext4_mballoc(void);
1432 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1433                              struct buffer_head *bh, ext4_fsblk_t block,
1434                              unsigned long count, int flags);
1435 extern int ext4_mb_add_groupinfo(struct super_block *sb,
1436                 ext4_group_t i, struct ext4_group_desc *desc);
1437 extern int ext4_mb_get_buddy_cache_lock(struct super_block *, ext4_group_t);
1438 extern void ext4_mb_put_buddy_cache_lock(struct super_block *,
1439                                                 ext4_group_t, int);
1440 /* inode.c */
1441 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1442                                                 ext4_lblk_t, int, int *);
1443 struct buffer_head *ext4_bread(handle_t *, struct inode *,
1444                                                 ext4_lblk_t, int, int *);
1445 int ext4_get_block(struct inode *inode, sector_t iblock,
1446                                 struct buffer_head *bh_result, int create);
1447
1448 extern struct inode *ext4_iget(struct super_block *, unsigned long);
1449 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
1450 extern int  ext4_setattr(struct dentry *, struct iattr *);
1451 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1452                                 struct kstat *stat);
1453 extern void ext4_delete_inode(struct inode *);
1454 extern int  ext4_sync_inode(handle_t *, struct inode *);
1455 extern void ext4_dirty_inode(struct inode *);
1456 extern int ext4_change_inode_journal_flag(struct inode *, int);
1457 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1458 extern int ext4_can_truncate(struct inode *inode);
1459 extern void ext4_truncate(struct inode *);
1460 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
1461 extern void ext4_set_inode_flags(struct inode *);
1462 extern void ext4_get_inode_flags(struct ext4_inode_info *);
1463 extern int ext4_alloc_da_blocks(struct inode *inode);
1464 extern void ext4_set_aops(struct inode *inode);
1465 extern int ext4_writepage_trans_blocks(struct inode *);
1466 extern int ext4_meta_trans_blocks(struct inode *, int nrblocks, int idxblocks);
1467 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1468 extern int ext4_block_truncate_page(handle_t *handle,
1469                 struct address_space *mapping, loff_t from);
1470 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1471 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
1472 extern int flush_completed_IO(struct inode *inode);
1473 extern void ext4_da_update_reserve_space(struct inode *inode,
1474                                         int used, int quota_claim);
1475 /* ioctl.c */
1476 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1477 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1478
1479 /* migrate.c */
1480 extern int ext4_ext_migrate(struct inode *);
1481
1482 /* namei.c */
1483 extern unsigned int ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize);
1484 extern __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize);
1485 extern int ext4_orphan_add(handle_t *, struct inode *);
1486 extern int ext4_orphan_del(handle_t *, struct inode *);
1487 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1488                                 __u32 start_minor_hash, __u32 *next_hash);
1489
1490 /* resize.c */
1491 extern int ext4_group_add(struct super_block *sb,
1492                                 struct ext4_new_group_data *input);
1493 extern int ext4_group_extend(struct super_block *sb,
1494                                 struct ext4_super_block *es,
1495                                 ext4_fsblk_t n_blocks_count);
1496
1497 /* super.c */
1498 extern void __ext4_error(struct super_block *, const char *, const char *, ...)
1499         __attribute__ ((format (printf, 3, 4)));
1500 #define ext4_error(sb, message...)      __ext4_error(sb, __func__, ## message)
1501 extern void ext4_error_inode(const char *, struct inode *, const char *, ...)
1502         __attribute__ ((format (printf, 3, 4)));
1503 extern void ext4_error_file(const char *, struct file *, const char *, ...)
1504         __attribute__ ((format (printf, 3, 4)));
1505 extern void __ext4_std_error(struct super_block *, const char *, int);
1506 extern void ext4_abort(struct super_block *, const char *, const char *, ...)
1507         __attribute__ ((format (printf, 3, 4)));
1508 extern void __ext4_warning(struct super_block *, const char *,
1509                           const char *, ...)
1510         __attribute__ ((format (printf, 3, 4)));
1511 #define ext4_warning(sb, message...)    __ext4_warning(sb, __func__, ## message)
1512 extern void ext4_msg(struct super_block *, const char *, const char *, ...)
1513         __attribute__ ((format (printf, 3, 4)));
1514 extern void ext4_grp_locked_error(struct super_block *, ext4_group_t,
1515                                 const char *, const char *, ...)
1516         __attribute__ ((format (printf, 4, 5)));
1517 extern void ext4_update_dynamic_rev(struct super_block *sb);
1518 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1519                                         __u32 compat);
1520 extern int ext4_update_rocompat_feature(handle_t *handle,
1521                                         struct super_block *sb, __u32 rocompat);
1522 extern int ext4_update_incompat_feature(handle_t *handle,
1523                                         struct super_block *sb, __u32 incompat);
1524 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1525                                       struct ext4_group_desc *bg);
1526 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1527                                       struct ext4_group_desc *bg);
1528 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1529                                      struct ext4_group_desc *bg);
1530 extern __u32 ext4_free_blks_count(struct super_block *sb,
1531                                 struct ext4_group_desc *bg);
1532 extern __u32 ext4_free_inodes_count(struct super_block *sb,
1533                                  struct ext4_group_desc *bg);
1534 extern __u32 ext4_used_dirs_count(struct super_block *sb,
1535                                 struct ext4_group_desc *bg);
1536 extern __u32 ext4_itable_unused_count(struct super_block *sb,
1537                                    struct ext4_group_desc *bg);
1538 extern void ext4_block_bitmap_set(struct super_block *sb,
1539                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1540 extern void ext4_inode_bitmap_set(struct super_block *sb,
1541                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1542 extern void ext4_inode_table_set(struct super_block *sb,
1543                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
1544 extern void ext4_free_blks_set(struct super_block *sb,
1545                                struct ext4_group_desc *bg, __u32 count);
1546 extern void ext4_free_inodes_set(struct super_block *sb,
1547                                 struct ext4_group_desc *bg, __u32 count);
1548 extern void ext4_used_dirs_set(struct super_block *sb,
1549                                 struct ext4_group_desc *bg, __u32 count);
1550 extern void ext4_itable_unused_set(struct super_block *sb,
1551                                    struct ext4_group_desc *bg, __u32 count);
1552 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
1553                                    struct ext4_group_desc *gdp);
1554 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
1555                                        struct ext4_group_desc *gdp);
1556
1557 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
1558 {
1559         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
1560                 le32_to_cpu(es->s_blocks_count_lo);
1561 }
1562
1563 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
1564 {
1565         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
1566                 le32_to_cpu(es->s_r_blocks_count_lo);
1567 }
1568
1569 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
1570 {
1571         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
1572                 le32_to_cpu(es->s_free_blocks_count_lo);
1573 }
1574
1575 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
1576                                          ext4_fsblk_t blk)
1577 {
1578         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
1579         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
1580 }
1581
1582 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
1583                                               ext4_fsblk_t blk)
1584 {
1585         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
1586         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
1587 }
1588
1589 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
1590                                            ext4_fsblk_t blk)
1591 {
1592         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
1593         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
1594 }
1595
1596 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
1597 {
1598         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
1599                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
1600                         le32_to_cpu(raw_inode->i_size_lo);
1601         else
1602                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
1603 }
1604
1605 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
1606 {
1607         raw_inode->i_size_lo = cpu_to_le32(i_size);
1608         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
1609 }
1610
1611 static inline
1612 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
1613                                             ext4_group_t group)
1614 {
1615          struct ext4_group_info ***grp_info;
1616          long indexv, indexh;
1617          grp_info = EXT4_SB(sb)->s_group_info;
1618          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
1619          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
1620          return grp_info[indexv][indexh];
1621 }
1622
1623 /*
1624  * Reading s_groups_count requires using smp_rmb() afterwards.  See
1625  * the locking protocol documented in the comments of ext4_group_add()
1626  * in resize.c
1627  */
1628 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
1629 {
1630         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
1631
1632         smp_rmb();
1633         return ngroups;
1634 }
1635
1636 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
1637                                              ext4_group_t block_group)
1638 {
1639         return block_group >> sbi->s_log_groups_per_flex;
1640 }
1641
1642 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
1643 {
1644         return 1 << sbi->s_log_groups_per_flex;
1645 }
1646
1647 #define ext4_std_error(sb, errno)                               \
1648 do {                                                            \
1649         if ((errno))                                            \
1650                 __ext4_std_error((sb), __func__, (errno));      \
1651 } while (0)
1652
1653 #ifdef CONFIG_SMP
1654 /* Each CPU can accumulate percpu_counter_batch blocks in their local
1655  * counters. So we need to make sure we have free blocks more
1656  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
1657  */
1658 #define EXT4_FREEBLOCKS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
1659 #else
1660 #define EXT4_FREEBLOCKS_WATERMARK 0
1661 #endif
1662
1663 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
1664 {
1665         /*
1666          * XXX: replace with spinlock if seen contended -bzzz
1667          */
1668         down_write(&EXT4_I(inode)->i_data_sem);
1669         if (newsize > EXT4_I(inode)->i_disksize)
1670                 EXT4_I(inode)->i_disksize = newsize;
1671         up_write(&EXT4_I(inode)->i_data_sem);
1672         return ;
1673 }
1674
1675 struct ext4_group_info {
1676         unsigned long   bb_state;
1677         struct rb_root  bb_free_root;
1678         ext4_grpblk_t   bb_first_free;  /* first free block */
1679         ext4_grpblk_t   bb_free;        /* total free blocks */
1680         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
1681         struct          list_head bb_prealloc_list;
1682 #ifdef DOUBLE_CHECK
1683         void            *bb_bitmap;
1684 #endif
1685         struct rw_semaphore alloc_sem;
1686         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
1687                                          * regions, index is order.
1688                                          * bb_counters[3] = 5 means
1689                                          * 5 free 8-block regions. */
1690 };
1691
1692 #define EXT4_GROUP_INFO_NEED_INIT_BIT   0
1693
1694 #define EXT4_MB_GRP_NEED_INIT(grp)      \
1695         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
1696
1697 #define EXT4_MAX_CONTENTION             8
1698 #define EXT4_CONTENTION_THRESHOLD       2
1699
1700 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
1701                                               ext4_group_t group)
1702 {
1703         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
1704 }
1705
1706 /*
1707  * Returns true if the filesystem is busy enough that attempts to
1708  * access the block group locks has run into contention.
1709  */
1710 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
1711 {
1712         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
1713 }
1714
1715 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
1716 {
1717         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
1718         if (spin_trylock(lock))
1719                 /*
1720                  * We're able to grab the lock right away, so drop the
1721                  * lock contention counter.
1722                  */
1723                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
1724         else {
1725                 /*
1726                  * The lock is busy, so bump the contention counter,
1727                  * and then wait on the spin lock.
1728                  */
1729                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
1730                                   EXT4_MAX_CONTENTION);
1731                 spin_lock(lock);
1732         }
1733 }
1734
1735 static inline void ext4_unlock_group(struct super_block *sb,
1736                                         ext4_group_t group)
1737 {
1738         spin_unlock(ext4_group_lock_ptr(sb, group));
1739 }
1740
1741 /*
1742  * Inodes and files operations
1743  */
1744
1745 /* dir.c */
1746 extern const struct file_operations ext4_dir_operations;
1747
1748 /* file.c */
1749 extern const struct inode_operations ext4_file_inode_operations;
1750 extern const struct file_operations ext4_file_operations;
1751
1752 /* namei.c */
1753 extern const struct inode_operations ext4_dir_inode_operations;
1754 extern const struct inode_operations ext4_special_inode_operations;
1755 extern struct dentry *ext4_get_parent(struct dentry *child);
1756
1757 /* symlink.c */
1758 extern const struct inode_operations ext4_symlink_inode_operations;
1759 extern const struct inode_operations ext4_fast_symlink_inode_operations;
1760
1761 /* block_validity */
1762 extern void ext4_release_system_zone(struct super_block *sb);
1763 extern int ext4_setup_system_zone(struct super_block *sb);
1764 extern int __init init_ext4_system_zone(void);
1765 extern void exit_ext4_system_zone(void);
1766 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
1767                                  ext4_fsblk_t start_blk,
1768                                  unsigned int count);
1769
1770 /* extents.c */
1771 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
1772 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
1773 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
1774                                        int chunk);
1775 extern int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
1776                                ext4_lblk_t iblock, unsigned int max_blocks,
1777                                struct buffer_head *bh_result, int flags);
1778 extern void ext4_ext_truncate(struct inode *);
1779 extern void ext4_ext_init(struct super_block *);
1780 extern void ext4_ext_release(struct super_block *);
1781 extern long ext4_fallocate(struct inode *inode, int mode, loff_t offset,
1782                           loff_t len);
1783 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
1784                           ssize_t len);
1785 extern int ext4_get_blocks(handle_t *handle, struct inode *inode,
1786                            sector_t block, unsigned int max_blocks,
1787                            struct buffer_head *bh, int flags);
1788 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1789                         __u64 start, __u64 len);
1790 /* move_extent.c */
1791 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
1792                              __u64 start_orig, __u64 start_donor,
1793                              __u64 len, __u64 *moved_len);
1794
1795
1796 /* BH_Uninit flag: blocks are allocated but uninitialized on disk */
1797 enum ext4_state_bits {
1798         BH_Uninit       /* blocks are allocated but uninitialized on disk */
1799           = BH_JBDPrivateStart,
1800 };
1801
1802 BUFFER_FNS(Uninit, uninit)
1803 TAS_BUFFER_FNS(Uninit, uninit)
1804
1805 /*
1806  * Add new method to test wether block and inode bitmaps are properly
1807  * initialized. With uninit_bg reading the block from disk is not enough
1808  * to mark the bitmap uptodate. We need to also zero-out the bitmap
1809  */
1810 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
1811
1812 static inline int bitmap_uptodate(struct buffer_head *bh)
1813 {
1814         return (buffer_uptodate(bh) &&
1815                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
1816 }
1817 static inline void set_bitmap_uptodate(struct buffer_head *bh)
1818 {
1819         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
1820 }
1821
1822 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
1823
1824 #endif  /* __KERNEL__ */
1825
1826 #endif  /* _EXT4_H */