ext4: Add new abstraction ext4_map_blocks() underneath ext4_get_blocks()
[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
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  * Logical to physical block mapping, used by ext4_map_blocks()
130  *
131  * This structure is used to pass requests into ext4_map_blocks() as
132  * well as to store the information returned by ext4_map_blocks().  It
133  * takes less room on the stack than a struct buffer_head.
134  */
135 #define EXT4_MAP_NEW            (1 << BH_New)
136 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
137 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
138 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
139 #define EXT4_MAP_UNINIT         (1 << BH_Uninit)
140 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
141                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
142                                  EXT4_MAP_UNINIT)
143
144 struct ext4_map_blocks {
145         ext4_fsblk_t m_pblk;
146         ext4_lblk_t m_lblk;
147         unsigned int m_len;
148         unsigned int m_flags;
149 };
150
151 /*
152  * For delayed allocation tracking
153  */
154 struct mpage_da_data {
155         struct inode *inode;
156         sector_t b_blocknr;             /* start block number of extent */
157         size_t b_size;                  /* size of extent */
158         unsigned long b_state;          /* state of the extent */
159         unsigned long first_page, next_page;    /* extent of pages */
160         struct writeback_control *wbc;
161         int io_done;
162         int pages_written;
163         int retval;
164 };
165 #define EXT4_IO_UNWRITTEN       0x1
166 typedef struct ext4_io_end {
167         struct list_head        list;           /* per-file finished AIO list */
168         struct inode            *inode;         /* file being written to */
169         unsigned int            flag;           /* unwritten or not */
170         struct page             *page;          /* page struct for buffer write */
171         loff_t                  offset;         /* offset in the file */
172         ssize_t                 size;           /* size of the extent */
173         struct work_struct      work;           /* data work queue */
174 } ext4_io_end_t;
175
176 /*
177  * Special inodes numbers
178  */
179 #define EXT4_BAD_INO             1      /* Bad blocks inode */
180 #define EXT4_ROOT_INO            2      /* Root inode */
181 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
182 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
183 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
184 #define EXT4_JOURNAL_INO         8      /* Journal inode */
185
186 /* First non-reserved inode for old ext4 filesystems */
187 #define EXT4_GOOD_OLD_FIRST_INO 11
188
189 /*
190  * Maximal count of links to a file
191  */
192 #define EXT4_LINK_MAX           65000
193
194 /*
195  * Macro-instructions used to manage several block sizes
196  */
197 #define EXT4_MIN_BLOCK_SIZE             1024
198 #define EXT4_MAX_BLOCK_SIZE             65536
199 #define EXT4_MIN_BLOCK_LOG_SIZE         10
200 #ifdef __KERNEL__
201 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
202 #else
203 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
204 #endif
205 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
206 #ifdef __KERNEL__
207 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
208 #else
209 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
210 #endif
211 #ifdef __KERNEL__
212 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
213 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
214 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
215 #else
216 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
217                                  EXT4_GOOD_OLD_INODE_SIZE : \
218                                  (s)->s_inode_size)
219 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
220                                  EXT4_GOOD_OLD_FIRST_INO : \
221                                  (s)->s_first_ino)
222 #endif
223 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
224
225 /*
226  * Structure of a blocks group descriptor
227  */
228 struct ext4_group_desc
229 {
230         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
231         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
232         __le32  bg_inode_table_lo;      /* Inodes table block */
233         __le16  bg_free_blocks_count_lo;/* Free blocks count */
234         __le16  bg_free_inodes_count_lo;/* Free inodes count */
235         __le16  bg_used_dirs_count_lo;  /* Directories count */
236         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
237         __u32   bg_reserved[2];         /* Likely block/inode bitmap checksum */
238         __le16  bg_itable_unused_lo;    /* Unused inodes count */
239         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
240         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
241         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
242         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
243         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
244         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
245         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
246         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
247         __u32   bg_reserved2[3];
248 };
249
250 /*
251  * Structure of a flex block group info
252  */
253
254 struct flex_groups {
255         atomic_t free_inodes;
256         atomic_t free_blocks;
257         atomic_t used_dirs;
258 };
259
260 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
261 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
262 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
263
264 /*
265  * Macro-instructions used to manage group descriptors
266  */
267 #define EXT4_MIN_DESC_SIZE              32
268 #define EXT4_MIN_DESC_SIZE_64BIT        64
269 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
270 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
271 #ifdef __KERNEL__
272 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
273 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
274 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
275 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
276 #else
277 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
278 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
279 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
280 #endif
281
282 /*
283  * Constants relative to the data blocks
284  */
285 #define EXT4_NDIR_BLOCKS                12
286 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
287 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
288 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
289 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
290
291 /*
292  * Inode flags
293  */
294 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
295 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
296 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
297 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
298 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
299 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
300 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
301 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
302 /* Reserved for compression usage... */
303 #define EXT4_DIRTY_FL                   0x00000100
304 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
305 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
306 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
307 /* End compression flags --- maybe not all used */
308 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
309 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
310 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
311 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
312 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
313 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
314 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
315 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
316 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
317 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
318 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
319
320 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
321 #define EXT4_FL_USER_MODIFIABLE         0x004B80FF /* User modifiable flags */
322
323 /* Flags that should be inherited by new inodes from their parent. */
324 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
325                            EXT4_SYNC_FL | EXT4_IMMUTABLE_FL | EXT4_APPEND_FL |\
326                            EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
327                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
328                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
329
330 /* Flags that are appropriate for regular files (all but dir-specific ones). */
331 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
332
333 /* Flags that are appropriate for non-directories/regular files. */
334 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
335
336 /* Mask out flags that are inappropriate for the given type of inode. */
337 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
338 {
339         if (S_ISDIR(mode))
340                 return flags;
341         else if (S_ISREG(mode))
342                 return flags & EXT4_REG_FLMASK;
343         else
344                 return flags & EXT4_OTHER_FLMASK;
345 }
346
347 /* Used to pass group descriptor data when online resize is done */
348 struct ext4_new_group_input {
349         __u32 group;            /* Group number for this data */
350         __u64 block_bitmap;     /* Absolute block number of block bitmap */
351         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
352         __u64 inode_table;      /* Absolute block number of inode table start */
353         __u32 blocks_count;     /* Total number of blocks in this group */
354         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
355         __u16 unused;
356 };
357
358 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
359 struct ext4_new_group_data {
360         __u32 group;
361         __u64 block_bitmap;
362         __u64 inode_bitmap;
363         __u64 inode_table;
364         __u32 blocks_count;
365         __u16 reserved_blocks;
366         __u16 unused;
367         __u32 free_blocks_count;
368 };
369
370 /*
371  * Flags used by ext4_get_blocks()
372  */
373         /* Allocate any needed blocks and/or convert an unitialized
374            extent to be an initialized ext4 */
375 #define EXT4_GET_BLOCKS_CREATE                  0x0001
376         /* Request the creation of an unitialized extent */
377 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
378 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
379                                                  EXT4_GET_BLOCKS_CREATE)
380         /* Caller is from the delayed allocation writeout path,
381            so set the magic i_delalloc_reserve_flag after taking the 
382            inode allocation semaphore for */
383 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
384         /* caller is from the direct IO path, request to creation of an
385         unitialized extents if not allocated, split the uninitialized
386         extent if blocks has been preallocated already*/
387 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
388 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
389 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
390                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
391         /* Convert extent to initialized after IO complete */
392 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
393                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
394
395 /*
396  * Flags used by ext4_free_blocks
397  */
398 #define EXT4_FREE_BLOCKS_METADATA       0x0001
399 #define EXT4_FREE_BLOCKS_FORGET         0x0002
400 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
401
402 /*
403  * ioctl commands
404  */
405 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
406 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
407 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
408 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
409 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
410 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
411 #ifdef CONFIG_JBD2_DEBUG
412 #define EXT4_IOC_WAIT_FOR_READONLY      _IOR('f', 99, long)
413 #endif
414 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
415 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
416 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
417 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
418 #define EXT4_IOC_MIGRATE                _IO('f', 9)
419  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
420  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
421 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
422 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
423
424 /*
425  * ioctl commands in 32 bit emulation
426  */
427 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
428 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
429 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
430 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
431 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
432 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
433 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
434 #ifdef CONFIG_JBD2_DEBUG
435 #define EXT4_IOC32_WAIT_FOR_READONLY    _IOR('f', 99, int)
436 #endif
437 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
438 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
439
440
441 /*
442  *  Mount options
443  */
444 struct ext4_mount_options {
445         unsigned long s_mount_opt;
446         uid_t s_resuid;
447         gid_t s_resgid;
448         unsigned long s_commit_interval;
449         u32 s_min_batch_time, s_max_batch_time;
450 #ifdef CONFIG_QUOTA
451         int s_jquota_fmt;
452         char *s_qf_names[MAXQUOTAS];
453 #endif
454 };
455
456 /* Max physical block we can addres w/o extents */
457 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
458
459 /*
460  * Structure of an inode on the disk
461  */
462 struct ext4_inode {
463         __le16  i_mode;         /* File mode */
464         __le16  i_uid;          /* Low 16 bits of Owner Uid */
465         __le32  i_size_lo;      /* Size in bytes */
466         __le32  i_atime;        /* Access time */
467         __le32  i_ctime;        /* Inode Change time */
468         __le32  i_mtime;        /* Modification time */
469         __le32  i_dtime;        /* Deletion Time */
470         __le16  i_gid;          /* Low 16 bits of Group Id */
471         __le16  i_links_count;  /* Links count */
472         __le32  i_blocks_lo;    /* Blocks count */
473         __le32  i_flags;        /* File flags */
474         union {
475                 struct {
476                         __le32  l_i_version;
477                 } linux1;
478                 struct {
479                         __u32  h_i_translator;
480                 } hurd1;
481                 struct {
482                         __u32  m_i_reserved1;
483                 } masix1;
484         } osd1;                         /* OS dependent 1 */
485         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
486         __le32  i_generation;   /* File version (for NFS) */
487         __le32  i_file_acl_lo;  /* File ACL */
488         __le32  i_size_high;
489         __le32  i_obso_faddr;   /* Obsoleted fragment address */
490         union {
491                 struct {
492                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
493                         __le16  l_i_file_acl_high;
494                         __le16  l_i_uid_high;   /* these 2 fields */
495                         __le16  l_i_gid_high;   /* were reserved2[0] */
496                         __u32   l_i_reserved2;
497                 } linux2;
498                 struct {
499                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
500                         __u16   h_i_mode_high;
501                         __u16   h_i_uid_high;
502                         __u16   h_i_gid_high;
503                         __u32   h_i_author;
504                 } hurd2;
505                 struct {
506                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
507                         __le16  m_i_file_acl_high;
508                         __u32   m_i_reserved2[2];
509                 } masix2;
510         } osd2;                         /* OS dependent 2 */
511         __le16  i_extra_isize;
512         __le16  i_pad1;
513         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
514         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
515         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
516         __le32  i_crtime;       /* File Creation time */
517         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
518         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
519 };
520
521 struct move_extent {
522         __u32 reserved;         /* should be zero */
523         __u32 donor_fd;         /* donor file descriptor */
524         __u64 orig_start;       /* logical start offset in block for orig */
525         __u64 donor_start;      /* logical start offset in block for donor */
526         __u64 len;              /* block length to be moved */
527         __u64 moved_len;        /* moved block length */
528 };
529
530 #define EXT4_EPOCH_BITS 2
531 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
532 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
533
534 /*
535  * Extended fields will fit into an inode if the filesystem was formatted
536  * with large inodes (-I 256 or larger) and there are not currently any EAs
537  * consuming all of the available space. For new inodes we always reserve
538  * enough space for the kernel's known extended fields, but for inodes
539  * created with an old kernel this might not have been the case. None of
540  * the extended inode fields is critical for correct filesystem operation.
541  * This macro checks if a certain field fits in the inode. Note that
542  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
543  */
544 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
545         ((offsetof(typeof(*ext4_inode), field) +        \
546           sizeof((ext4_inode)->field))                  \
547         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
548             (einode)->i_extra_isize))                   \
549
550 static inline __le32 ext4_encode_extra_time(struct timespec *time)
551 {
552        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
553                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
554                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
555 }
556
557 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
558 {
559        if (sizeof(time->tv_sec) > 4)
560                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
561                                << 32;
562        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
563 }
564
565 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
566 do {                                                                           \
567         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
568         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
569                 (raw_inode)->xtime ## _extra =                                 \
570                                 ext4_encode_extra_time(&(inode)->xtime);       \
571 } while (0)
572
573 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
574 do {                                                                           \
575         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
576                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
577         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
578                 (raw_inode)->xtime ## _extra =                                 \
579                                 ext4_encode_extra_time(&(einode)->xtime);      \
580 } while (0)
581
582 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
583 do {                                                                           \
584         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
585         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
586                 ext4_decode_extra_time(&(inode)->xtime,                        \
587                                        raw_inode->xtime ## _extra);            \
588 } while (0)
589
590 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
591 do {                                                                           \
592         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
593                 (einode)->xtime.tv_sec =                                       \
594                         (signed)le32_to_cpu((raw_inode)->xtime);               \
595         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
596                 ext4_decode_extra_time(&(einode)->xtime,                       \
597                                        raw_inode->xtime ## _extra);            \
598 } while (0)
599
600 #define i_disk_version osd1.linux1.l_i_version
601
602 #if defined(__KERNEL__) || defined(__linux__)
603 #define i_reserved1     osd1.linux1.l_i_reserved1
604 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
605 #define i_blocks_high   osd2.linux2.l_i_blocks_high
606 #define i_uid_low       i_uid
607 #define i_gid_low       i_gid
608 #define i_uid_high      osd2.linux2.l_i_uid_high
609 #define i_gid_high      osd2.linux2.l_i_gid_high
610 #define i_reserved2     osd2.linux2.l_i_reserved2
611
612 #elif defined(__GNU__)
613
614 #define i_translator    osd1.hurd1.h_i_translator
615 #define i_uid_high      osd2.hurd2.h_i_uid_high
616 #define i_gid_high      osd2.hurd2.h_i_gid_high
617 #define i_author        osd2.hurd2.h_i_author
618
619 #elif defined(__masix__)
620
621 #define i_reserved1     osd1.masix1.m_i_reserved1
622 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
623 #define i_reserved2     osd2.masix2.m_i_reserved2
624
625 #endif /* defined(__KERNEL__) || defined(__linux__) */
626
627 /*
628  * storage for cached extent
629  */
630 struct ext4_ext_cache {
631         ext4_fsblk_t    ec_start;
632         ext4_lblk_t     ec_block;
633         __u32           ec_len; /* must be 32bit to return holes */
634         __u32           ec_type;
635 };
636
637 /*
638  * fourth extended file system inode data in memory
639  */
640 struct ext4_inode_info {
641         __le32  i_data[15];     /* unconverted */
642         __u32   i_flags;
643         ext4_fsblk_t    i_file_acl;
644         __u32   i_dtime;
645
646         /*
647          * i_block_group is the number of the block group which contains
648          * this file's inode.  Constant across the lifetime of the inode,
649          * it is ued for making block allocation decisions - we try to
650          * place a file's data blocks near its inode block, and new inodes
651          * near to their parent directory's inode.
652          */
653         ext4_group_t    i_block_group;
654         unsigned long   i_state_flags;          /* Dynamic state flags */
655
656         ext4_lblk_t             i_dir_start_lookup;
657 #ifdef CONFIG_EXT4_FS_XATTR
658         /*
659          * Extended attributes can be read independently of the main file
660          * data. Taking i_mutex even when reading would cause contention
661          * between readers of EAs and writers of regular file data, so
662          * instead we synchronize on xattr_sem when reading or changing
663          * EAs.
664          */
665         struct rw_semaphore xattr_sem;
666 #endif
667
668         struct list_head i_orphan;      /* unlinked but open inodes */
669
670         /*
671          * i_disksize keeps track of what the inode size is ON DISK, not
672          * in memory.  During truncate, i_size is set to the new size by
673          * the VFS prior to calling ext4_truncate(), but the filesystem won't
674          * set i_disksize to 0 until the truncate is actually under way.
675          *
676          * The intent is that i_disksize always represents the blocks which
677          * are used by this file.  This allows recovery to restart truncate
678          * on orphans if we crash during truncate.  We actually write i_disksize
679          * into the on-disk inode when writing inodes out, instead of i_size.
680          *
681          * The only time when i_disksize and i_size may be different is when
682          * a truncate is in progress.  The only things which change i_disksize
683          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
684          */
685         loff_t  i_disksize;
686
687         /*
688          * i_data_sem is for serialising ext4_truncate() against
689          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
690          * data tree are chopped off during truncate. We can't do that in
691          * ext4 because whenever we perform intermediate commits during
692          * truncate, the inode and all the metadata blocks *must* be in a
693          * consistent state which allows truncation of the orphans to restart
694          * during recovery.  Hence we must fix the get_block-vs-truncate race
695          * by other means, so we have i_data_sem.
696          */
697         struct rw_semaphore i_data_sem;
698         struct inode vfs_inode;
699         struct jbd2_inode jinode;
700
701         struct ext4_ext_cache i_cached_extent;
702         /*
703          * File creation time. Its function is same as that of
704          * struct timespec i_{a,c,m}time in the generic inode.
705          */
706         struct timespec i_crtime;
707
708         /* mballoc */
709         struct list_head i_prealloc_list;
710         spinlock_t i_prealloc_lock;
711
712         /* ialloc */
713         ext4_group_t    i_last_alloc_group;
714
715         /* allocation reservation info for delalloc */
716         unsigned int i_reserved_data_blocks;
717         unsigned int i_reserved_meta_blocks;
718         unsigned int i_allocated_meta_blocks;
719         unsigned short i_delalloc_reserved_flag;
720         sector_t i_da_metadata_calc_last_lblock;
721         int i_da_metadata_calc_len;
722
723         /* on-disk additional length */
724         __u16 i_extra_isize;
725
726         spinlock_t i_block_reservation_lock;
727 #ifdef CONFIG_QUOTA
728         /* quota space reservation, managed internally by quota code */
729         qsize_t i_reserved_quota;
730 #endif
731
732         /* completed IOs that might need unwritten extents handling */
733         struct list_head i_completed_io_list;
734         spinlock_t i_completed_io_lock;
735         /* current io_end structure for async DIO write*/
736         ext4_io_end_t *cur_aio_dio;
737
738         /*
739          * Transactions that contain inode's metadata needed to complete
740          * fsync and fdatasync, respectively.
741          */
742         tid_t i_sync_tid;
743         tid_t i_datasync_tid;
744 };
745
746 /*
747  * File system states
748  */
749 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
750 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
751 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
752
753 /*
754  * Misc. filesystem flags
755  */
756 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
757 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
758 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
759
760 /*
761  * Mount flags
762  */
763 #define EXT4_MOUNT_OLDALLOC             0x00002  /* Don't use the new Orlov allocator */
764 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
765 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
766 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
767 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
768 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
769 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
770 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
771 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
772 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
773 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
774 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
775 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
776 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
777 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
778 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
779 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
780 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
781 #define EXT4_MOUNT_NOBH                 0x40000 /* No bufferheads */
782 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
783 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
784 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
785 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
786 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
787 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
788 #define EXT4_MOUNT_I_VERSION            0x2000000 /* i_version support */
789 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
790 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
791 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
792 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
793
794 #define clear_opt(o, opt)               o &= ~EXT4_MOUNT_##opt
795 #define set_opt(o, opt)                 o |= EXT4_MOUNT_##opt
796 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
797                                          EXT4_MOUNT_##opt)
798
799 #define ext4_set_bit                    ext2_set_bit
800 #define ext4_set_bit_atomic             ext2_set_bit_atomic
801 #define ext4_clear_bit                  ext2_clear_bit
802 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
803 #define ext4_test_bit                   ext2_test_bit
804 #define ext4_find_first_zero_bit        ext2_find_first_zero_bit
805 #define ext4_find_next_zero_bit         ext2_find_next_zero_bit
806 #define ext4_find_next_bit              ext2_find_next_bit
807
808 /*
809  * Maximal mount counts between two filesystem checks
810  */
811 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
812 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
813
814 /*
815  * Behaviour when detecting errors
816  */
817 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
818 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
819 #define EXT4_ERRORS_PANIC               3       /* Panic */
820 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
821
822 /*
823  * Structure of the super block
824  */
825 struct ext4_super_block {
826 /*00*/  __le32  s_inodes_count;         /* Inodes count */
827         __le32  s_blocks_count_lo;      /* Blocks count */
828         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
829         __le32  s_free_blocks_count_lo; /* Free blocks count */
830 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
831         __le32  s_first_data_block;     /* First Data Block */
832         __le32  s_log_block_size;       /* Block size */
833         __le32  s_obso_log_frag_size;   /* Obsoleted fragment size */
834 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
835         __le32  s_obso_frags_per_group; /* Obsoleted fragments per group */
836         __le32  s_inodes_per_group;     /* # Inodes per group */
837         __le32  s_mtime;                /* Mount time */
838 /*30*/  __le32  s_wtime;                /* Write time */
839         __le16  s_mnt_count;            /* Mount count */
840         __le16  s_max_mnt_count;        /* Maximal mount count */
841         __le16  s_magic;                /* Magic signature */
842         __le16  s_state;                /* File system state */
843         __le16  s_errors;               /* Behaviour when detecting errors */
844         __le16  s_minor_rev_level;      /* minor revision level */
845 /*40*/  __le32  s_lastcheck;            /* time of last check */
846         __le32  s_checkinterval;        /* max. time between checks */
847         __le32  s_creator_os;           /* OS */
848         __le32  s_rev_level;            /* Revision level */
849 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
850         __le16  s_def_resgid;           /* Default gid for reserved blocks */
851         /*
852          * These fields are for EXT4_DYNAMIC_REV superblocks only.
853          *
854          * Note: the difference between the compatible feature set and
855          * the incompatible feature set is that if there is a bit set
856          * in the incompatible feature set that the kernel doesn't
857          * know about, it should refuse to mount the filesystem.
858          *
859          * e2fsck's requirements are more strict; if it doesn't know
860          * about a feature in either the compatible or incompatible
861          * feature set, it must abort and not try to meddle with
862          * things it doesn't understand...
863          */
864         __le32  s_first_ino;            /* First non-reserved inode */
865         __le16  s_inode_size;           /* size of inode structure */
866         __le16  s_block_group_nr;       /* block group # of this superblock */
867         __le32  s_feature_compat;       /* compatible feature set */
868 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
869         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
870 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
871 /*78*/  char    s_volume_name[16];      /* volume name */
872 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
873 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
874         /*
875          * Performance hints.  Directory preallocation should only
876          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
877          */
878         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
879         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
880         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
881         /*
882          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
883          */
884 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
885 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
886         __le32  s_journal_dev;          /* device number of journal file */
887         __le32  s_last_orphan;          /* start of list of inodes to delete */
888         __le32  s_hash_seed[4];         /* HTREE hash seed */
889         __u8    s_def_hash_version;     /* Default hash version to use */
890         __u8    s_reserved_char_pad;
891         __le16  s_desc_size;            /* size of group descriptor */
892 /*100*/ __le32  s_default_mount_opts;
893         __le32  s_first_meta_bg;        /* First metablock block group */
894         __le32  s_mkfs_time;            /* When the filesystem was created */
895         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
896         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
897 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
898         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
899         __le32  s_free_blocks_count_hi; /* Free blocks count */
900         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
901         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
902         __le32  s_flags;                /* Miscellaneous flags */
903         __le16  s_raid_stride;          /* RAID stride */
904         __le16  s_mmp_interval;         /* # seconds to wait in MMP checking */
905         __le64  s_mmp_block;            /* Block for multi-mount protection */
906         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
907         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
908         __u8    s_reserved_char_pad2;
909         __le16  s_reserved_pad;
910         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
911         __u32   s_reserved[160];        /* Padding to the end of the block */
912 };
913
914 #ifdef __KERNEL__
915
916 /*
917  * run-time mount flags
918  */
919 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
920 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
921
922 /*
923  * fourth extended-fs super-block data in memory
924  */
925 struct ext4_sb_info {
926         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
927         unsigned long s_inodes_per_block;/* Number of inodes per block */
928         unsigned long s_blocks_per_group;/* Number of blocks in a group */
929         unsigned long s_inodes_per_group;/* Number of inodes in a group */
930         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
931         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
932         unsigned long s_desc_per_block; /* Number of group descriptors per block */
933         ext4_group_t s_groups_count;    /* Number of groups in the fs */
934         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
935         unsigned long s_overhead_last;  /* Last calculated overhead */
936         unsigned long s_blocks_last;    /* Last seen block count */
937         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
938         struct buffer_head * s_sbh;     /* Buffer containing the super block */
939         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
940         struct buffer_head **s_group_desc;
941         unsigned int s_mount_opt;
942         unsigned int s_mount_flags;
943         ext4_fsblk_t s_sb_block;
944         uid_t s_resuid;
945         gid_t s_resgid;
946         unsigned short s_mount_state;
947         unsigned short s_pad;
948         int s_addr_per_block_bits;
949         int s_desc_per_block_bits;
950         int s_inode_size;
951         int s_first_ino;
952         unsigned int s_inode_readahead_blks;
953         unsigned int s_inode_goal;
954         spinlock_t s_next_gen_lock;
955         u32 s_next_generation;
956         u32 s_hash_seed[4];
957         int s_def_hash_version;
958         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
959         struct percpu_counter s_freeblocks_counter;
960         struct percpu_counter s_freeinodes_counter;
961         struct percpu_counter s_dirs_counter;
962         struct percpu_counter s_dirtyblocks_counter;
963         struct blockgroup_lock *s_blockgroup_lock;
964         struct proc_dir_entry *s_proc;
965         struct kobject s_kobj;
966         struct completion s_kobj_unregister;
967
968         /* Journaling */
969         struct inode *s_journal_inode;
970         struct journal_s *s_journal;
971         struct list_head s_orphan;
972         struct mutex s_orphan_lock;
973         struct mutex s_resize_lock;
974         unsigned long s_commit_interval;
975         u32 s_max_batch_time;
976         u32 s_min_batch_time;
977         struct block_device *journal_bdev;
978 #ifdef CONFIG_JBD2_DEBUG
979         struct timer_list turn_ro_timer;        /* For turning read-only (crash simulation) */
980         wait_queue_head_t ro_wait_queue;        /* For people waiting for the fs to go read-only */
981 #endif
982 #ifdef CONFIG_QUOTA
983         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
984         int s_jquota_fmt;                       /* Format of quota to use */
985 #endif
986         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
987         struct rb_root system_blks;
988
989 #ifdef EXTENTS_STATS
990         /* ext4 extents stats */
991         unsigned long s_ext_min;
992         unsigned long s_ext_max;
993         unsigned long s_depth_max;
994         spinlock_t s_ext_stats_lock;
995         unsigned long s_ext_blocks;
996         unsigned long s_ext_extents;
997 #endif
998
999         /* for buddy allocator */
1000         struct ext4_group_info ***s_group_info;
1001         struct inode *s_buddy_cache;
1002         long s_blocks_reserved;
1003         spinlock_t s_reserve_lock;
1004         spinlock_t s_md_lock;
1005         tid_t s_last_transaction;
1006         unsigned short *s_mb_offsets;
1007         unsigned int *s_mb_maxs;
1008
1009         /* tunables */
1010         unsigned long s_stripe;
1011         unsigned int s_mb_stream_request;
1012         unsigned int s_mb_max_to_scan;
1013         unsigned int s_mb_min_to_scan;
1014         unsigned int s_mb_stats;
1015         unsigned int s_mb_order2_reqs;
1016         unsigned int s_mb_group_prealloc;
1017         unsigned int s_max_writeback_mb_bump;
1018         /* where last allocation was done - for stream allocation */
1019         unsigned long s_mb_last_group;
1020         unsigned long s_mb_last_start;
1021
1022         /* stats for buddy allocator */
1023         spinlock_t s_mb_pa_lock;
1024         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1025         atomic_t s_bal_success; /* we found long enough chunks */
1026         atomic_t s_bal_allocated;       /* in blocks */
1027         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1028         atomic_t s_bal_goals;   /* goal hits */
1029         atomic_t s_bal_breaks;  /* too long searches */
1030         atomic_t s_bal_2orders; /* 2^order hits */
1031         spinlock_t s_bal_lock;
1032         unsigned long s_mb_buddies_generated;
1033         unsigned long long s_mb_generation_time;
1034         atomic_t s_mb_lost_chunks;
1035         atomic_t s_mb_preallocated;
1036         atomic_t s_mb_discarded;
1037         atomic_t s_lock_busy;
1038
1039         /* locality groups */
1040         struct ext4_locality_group __percpu *s_locality_groups;
1041
1042         /* for write statistics */
1043         unsigned long s_sectors_written_start;
1044         u64 s_kbytes_written;
1045
1046         unsigned int s_log_groups_per_flex;
1047         struct flex_groups *s_flex_groups;
1048
1049         /* workqueue for dio unwritten */
1050         struct workqueue_struct *dio_unwritten_wq;
1051 };
1052
1053 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1054 {
1055         return sb->s_fs_info;
1056 }
1057 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1058 {
1059         return container_of(inode, struct ext4_inode_info, vfs_inode);
1060 }
1061
1062 static inline struct timespec ext4_current_time(struct inode *inode)
1063 {
1064         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1065                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1066 }
1067
1068 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1069 {
1070         return ino == EXT4_ROOT_INO ||
1071                 ino == EXT4_JOURNAL_INO ||
1072                 ino == EXT4_RESIZE_INO ||
1073                 (ino >= EXT4_FIRST_INO(sb) &&
1074                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1075 }
1076
1077 /*
1078  * Inode dynamic state flags
1079  */
1080 enum {
1081         EXT4_STATE_JDATA,               /* journaled data exists */
1082         EXT4_STATE_NEW,                 /* inode is newly created */
1083         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1084         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1085         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1086         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1087         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1088 };
1089
1090 static inline int ext4_test_inode_state(struct inode *inode, int bit)
1091 {
1092         return test_bit(bit, &EXT4_I(inode)->i_state_flags);
1093 }
1094
1095 static inline void ext4_set_inode_state(struct inode *inode, int bit)
1096 {
1097         set_bit(bit, &EXT4_I(inode)->i_state_flags);
1098 }
1099
1100 static inline void ext4_clear_inode_state(struct inode *inode, int bit)
1101 {
1102         clear_bit(bit, &EXT4_I(inode)->i_state_flags);
1103 }
1104 #else
1105 /* Assume that user mode programs are passing in an ext4fs superblock, not
1106  * a kernel struct super_block.  This will allow us to call the feature-test
1107  * macros from user land. */
1108 #define EXT4_SB(sb)     (sb)
1109 #endif
1110
1111 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1112
1113 /*
1114  * Codes for operating systems
1115  */
1116 #define EXT4_OS_LINUX           0
1117 #define EXT4_OS_HURD            1
1118 #define EXT4_OS_MASIX           2
1119 #define EXT4_OS_FREEBSD         3
1120 #define EXT4_OS_LITES           4
1121
1122 /*
1123  * Revision levels
1124  */
1125 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1126 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1127
1128 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1129 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1130
1131 #define EXT4_GOOD_OLD_INODE_SIZE 128
1132
1133 /*
1134  * Feature set definitions
1135  */
1136
1137 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1138         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1139 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1140         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1141 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1142         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1143 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1144         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1145 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1146         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1147 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1148         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1149 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1150         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1151 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1152         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1153 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1154         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1155
1156 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1157 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1158 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1159 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1160 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1161 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1162
1163 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1164 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1165 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1166 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1167 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1168 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1169 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1170
1171 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1172 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1173 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1174 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1175 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1176 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1177 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1178 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1179 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1180 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1181 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1182
1183 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1184 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1185                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1186                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1187                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1188                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1189                                          EXT4_FEATURE_INCOMPAT_FLEX_BG)
1190 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1191                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1192                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1193                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1194                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1195                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1196                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE)
1197
1198 /*
1199  * Default values for user and/or group using reserved blocks
1200  */
1201 #define EXT4_DEF_RESUID         0
1202 #define EXT4_DEF_RESGID         0
1203
1204 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1205
1206 /*
1207  * Default mount options
1208  */
1209 #define EXT4_DEFM_DEBUG         0x0001
1210 #define EXT4_DEFM_BSDGROUPS     0x0002
1211 #define EXT4_DEFM_XATTR_USER    0x0004
1212 #define EXT4_DEFM_ACL           0x0008
1213 #define EXT4_DEFM_UID16         0x0010
1214 #define EXT4_DEFM_JMODE         0x0060
1215 #define EXT4_DEFM_JMODE_DATA    0x0020
1216 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1217 #define EXT4_DEFM_JMODE_WBACK   0x0060
1218
1219 /*
1220  * Default journal batch times
1221  */
1222 #define EXT4_DEF_MIN_BATCH_TIME 0
1223 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1224
1225 /*
1226  * Minimum number of groups in a flexgroup before we separate out
1227  * directories into the first block group of a flexgroup
1228  */
1229 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1230
1231 /*
1232  * Structure of a directory entry
1233  */
1234 #define EXT4_NAME_LEN 255
1235
1236 struct ext4_dir_entry {
1237         __le32  inode;                  /* Inode number */
1238         __le16  rec_len;                /* Directory entry length */
1239         __le16  name_len;               /* Name length */
1240         char    name[EXT4_NAME_LEN];    /* File name */
1241 };
1242
1243 /*
1244  * The new version of the directory entry.  Since EXT4 structures are
1245  * stored in intel byte order, and the name_len field could never be
1246  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1247  * file_type field.
1248  */
1249 struct ext4_dir_entry_2 {
1250         __le32  inode;                  /* Inode number */
1251         __le16  rec_len;                /* Directory entry length */
1252         __u8    name_len;               /* Name length */
1253         __u8    file_type;
1254         char    name[EXT4_NAME_LEN];    /* File name */
1255 };
1256
1257 /*
1258  * Ext4 directory file types.  Only the low 3 bits are used.  The
1259  * other bits are reserved for now.
1260  */
1261 #define EXT4_FT_UNKNOWN         0
1262 #define EXT4_FT_REG_FILE        1
1263 #define EXT4_FT_DIR             2
1264 #define EXT4_FT_CHRDEV          3
1265 #define EXT4_FT_BLKDEV          4
1266 #define EXT4_FT_FIFO            5
1267 #define EXT4_FT_SOCK            6
1268 #define EXT4_FT_SYMLINK         7
1269
1270 #define EXT4_FT_MAX             8
1271
1272 /*
1273  * EXT4_DIR_PAD defines the directory entries boundaries
1274  *
1275  * NOTE: It must be a multiple of 4
1276  */
1277 #define EXT4_DIR_PAD                    4
1278 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1279 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1280                                          ~EXT4_DIR_ROUND)
1281 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1282
1283 /*
1284  * Hash Tree Directory indexing
1285  * (c) Daniel Phillips, 2001
1286  */
1287
1288 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1289                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1290                       (EXT4_I(dir)->i_flags & EXT4_INDEX_FL))
1291 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1292 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1293
1294 /* Legal values for the dx_root hash_version field: */
1295
1296 #define DX_HASH_LEGACY          0
1297 #define DX_HASH_HALF_MD4        1
1298 #define DX_HASH_TEA             2
1299 #define DX_HASH_LEGACY_UNSIGNED 3
1300 #define DX_HASH_HALF_MD4_UNSIGNED       4
1301 #define DX_HASH_TEA_UNSIGNED            5
1302
1303 #ifdef __KERNEL__
1304
1305 /* hash info structure used by the directory hash */
1306 struct dx_hash_info
1307 {
1308         u32             hash;
1309         u32             minor_hash;
1310         int             hash_version;
1311         u32             *seed;
1312 };
1313
1314 #define EXT4_HTREE_EOF  0x7fffffff
1315
1316 /*
1317  * Control parameters used by ext4_htree_next_block
1318  */
1319 #define HASH_NB_ALWAYS          1
1320
1321
1322 /*
1323  * Describe an inode's exact location on disk and in memory
1324  */
1325 struct ext4_iloc
1326 {
1327         struct buffer_head *bh;
1328         unsigned long offset;
1329         ext4_group_t block_group;
1330 };
1331
1332 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1333 {
1334         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1335 }
1336
1337 /*
1338  * This structure is stuffed into the struct file's private_data field
1339  * for directories.  It is where we put information so that we can do
1340  * readdir operations in hash tree order.
1341  */
1342 struct dir_private_info {
1343         struct rb_root  root;
1344         struct rb_node  *curr_node;
1345         struct fname    *extra_fname;
1346         loff_t          last_pos;
1347         __u32           curr_hash;
1348         __u32           curr_minor_hash;
1349         __u32           next_hash;
1350 };
1351
1352 /* calculate the first block number of the group */
1353 static inline ext4_fsblk_t
1354 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1355 {
1356         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1357                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1358 }
1359
1360 /*
1361  * Special error return code only used by dx_probe() and its callers.
1362  */
1363 #define ERR_BAD_DX_DIR  -75000
1364
1365 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1366                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1367
1368 extern struct proc_dir_entry *ext4_proc_root;
1369
1370 /*
1371  * Function prototypes
1372  */
1373
1374 /*
1375  * Ok, these declarations are also in <linux/kernel.h> but none of the
1376  * ext4 source programs needs to include it so they are duplicated here.
1377  */
1378 # define NORET_TYPE     /**/
1379 # define ATTRIB_NORET   __attribute__((noreturn))
1380 # define NORET_AND      noreturn,
1381
1382 /* bitmap.c */
1383 extern unsigned int ext4_count_free(struct buffer_head *, unsigned);
1384
1385 /* balloc.c */
1386 extern unsigned int ext4_block_group(struct super_block *sb,
1387                         ext4_fsblk_t blocknr);
1388 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1389                         ext4_fsblk_t blocknr);
1390 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1391 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1392                         ext4_group_t group);
1393 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1394                         ext4_fsblk_t goal, unsigned long *count, int *errp);
1395 extern int ext4_claim_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1396 extern int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1397 extern void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
1398                                 ext4_fsblk_t block, unsigned long count);
1399 extern ext4_fsblk_t ext4_count_free_blocks(struct super_block *);
1400 extern void ext4_check_blocks_bitmap(struct super_block *);
1401 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1402                                                     ext4_group_t block_group,
1403                                                     struct buffer_head ** bh);
1404 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1405 struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1406                                       ext4_group_t block_group);
1407 extern unsigned ext4_init_block_bitmap(struct super_block *sb,
1408                                        struct buffer_head *bh,
1409                                        ext4_group_t group,
1410                                        struct ext4_group_desc *desc);
1411 #define ext4_free_blocks_after_init(sb, group, desc)                    \
1412                 ext4_init_block_bitmap(sb, NULL, group, desc)
1413
1414 /* dir.c */
1415 extern int ext4_check_dir_entry(const char *, struct inode *,
1416                                 struct ext4_dir_entry_2 *,
1417                                 struct buffer_head *, unsigned int);
1418 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1419                                     __u32 minor_hash,
1420                                     struct ext4_dir_entry_2 *dirent);
1421 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1422
1423 /* fsync.c */
1424 extern int ext4_sync_file(struct file *, struct dentry *, int);
1425
1426 /* hash.c */
1427 extern int ext4fs_dirhash(const char *name, int len, struct
1428                           dx_hash_info *hinfo);
1429
1430 /* ialloc.c */
1431 extern struct inode *ext4_new_inode(handle_t *, struct inode *, int,
1432                                     const struct qstr *qstr, __u32 goal);
1433 extern void ext4_free_inode(handle_t *, struct inode *);
1434 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1435 extern unsigned long ext4_count_free_inodes(struct super_block *);
1436 extern unsigned long ext4_count_dirs(struct super_block *);
1437 extern void ext4_check_inodes_bitmap(struct super_block *);
1438 extern unsigned ext4_init_inode_bitmap(struct super_block *sb,
1439                                        struct buffer_head *bh,
1440                                        ext4_group_t group,
1441                                        struct ext4_group_desc *desc);
1442 extern void mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1443
1444 /* mballoc.c */
1445 extern long ext4_mb_stats;
1446 extern long ext4_mb_max_to_scan;
1447 extern int ext4_mb_init(struct super_block *, int);
1448 extern int ext4_mb_release(struct super_block *);
1449 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1450                                 struct ext4_allocation_request *, int *);
1451 extern int ext4_mb_reserve_blocks(struct super_block *, int);
1452 extern void ext4_discard_preallocations(struct inode *);
1453 extern int __init init_ext4_mballoc(void);
1454 extern void exit_ext4_mballoc(void);
1455 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1456                              struct buffer_head *bh, ext4_fsblk_t block,
1457                              unsigned long count, int flags);
1458 extern int ext4_mb_add_groupinfo(struct super_block *sb,
1459                 ext4_group_t i, struct ext4_group_desc *desc);
1460 extern int ext4_mb_get_buddy_cache_lock(struct super_block *, ext4_group_t);
1461 extern void ext4_mb_put_buddy_cache_lock(struct super_block *,
1462                                                 ext4_group_t, int);
1463 /* inode.c */
1464 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1465                                                 ext4_lblk_t, int, int *);
1466 struct buffer_head *ext4_bread(handle_t *, struct inode *,
1467                                                 ext4_lblk_t, int, int *);
1468 int ext4_get_block(struct inode *inode, sector_t iblock,
1469                                 struct buffer_head *bh_result, int create);
1470
1471 extern struct inode *ext4_iget(struct super_block *, unsigned long);
1472 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
1473 extern int  ext4_setattr(struct dentry *, struct iattr *);
1474 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1475                                 struct kstat *stat);
1476 extern void ext4_delete_inode(struct inode *);
1477 extern int  ext4_sync_inode(handle_t *, struct inode *);
1478 extern void ext4_dirty_inode(struct inode *);
1479 extern int ext4_change_inode_journal_flag(struct inode *, int);
1480 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1481 extern int ext4_can_truncate(struct inode *inode);
1482 extern void ext4_truncate(struct inode *);
1483 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
1484 extern void ext4_set_inode_flags(struct inode *);
1485 extern void ext4_get_inode_flags(struct ext4_inode_info *);
1486 extern int ext4_alloc_da_blocks(struct inode *inode);
1487 extern void ext4_set_aops(struct inode *inode);
1488 extern int ext4_writepage_trans_blocks(struct inode *);
1489 extern int ext4_meta_trans_blocks(struct inode *, int nrblocks, int idxblocks);
1490 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1491 extern int ext4_block_truncate_page(handle_t *handle,
1492                 struct address_space *mapping, loff_t from);
1493 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1494 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
1495 extern int flush_completed_IO(struct inode *inode);
1496 extern void ext4_da_update_reserve_space(struct inode *inode,
1497                                         int used, int quota_claim);
1498 /* ioctl.c */
1499 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1500 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1501
1502 /* migrate.c */
1503 extern int ext4_ext_migrate(struct inode *);
1504
1505 /* namei.c */
1506 extern unsigned int ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize);
1507 extern __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize);
1508 extern int ext4_orphan_add(handle_t *, struct inode *);
1509 extern int ext4_orphan_del(handle_t *, struct inode *);
1510 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1511                                 __u32 start_minor_hash, __u32 *next_hash);
1512
1513 /* resize.c */
1514 extern int ext4_group_add(struct super_block *sb,
1515                                 struct ext4_new_group_data *input);
1516 extern int ext4_group_extend(struct super_block *sb,
1517                                 struct ext4_super_block *es,
1518                                 ext4_fsblk_t n_blocks_count);
1519
1520 /* super.c */
1521 extern void __ext4_error(struct super_block *, const char *, const char *, ...)
1522         __attribute__ ((format (printf, 3, 4)));
1523 #define ext4_error(sb, message...)      __ext4_error(sb, __func__, ## message)
1524 extern void ext4_error_inode(const char *, struct inode *, const char *, ...)
1525         __attribute__ ((format (printf, 3, 4)));
1526 extern void ext4_error_file(const char *, struct file *, const char *, ...)
1527         __attribute__ ((format (printf, 3, 4)));
1528 extern void __ext4_std_error(struct super_block *, const char *, int);
1529 extern void ext4_abort(struct super_block *, const char *, const char *, ...)
1530         __attribute__ ((format (printf, 3, 4)));
1531 extern void __ext4_warning(struct super_block *, const char *,
1532                           const char *, ...)
1533         __attribute__ ((format (printf, 3, 4)));
1534 #define ext4_warning(sb, message...)    __ext4_warning(sb, __func__, ## message)
1535 extern void ext4_msg(struct super_block *, const char *, const char *, ...)
1536         __attribute__ ((format (printf, 3, 4)));
1537 extern void ext4_grp_locked_error(struct super_block *, ext4_group_t,
1538                                 const char *, const char *, ...)
1539         __attribute__ ((format (printf, 4, 5)));
1540 extern void ext4_update_dynamic_rev(struct super_block *sb);
1541 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1542                                         __u32 compat);
1543 extern int ext4_update_rocompat_feature(handle_t *handle,
1544                                         struct super_block *sb, __u32 rocompat);
1545 extern int ext4_update_incompat_feature(handle_t *handle,
1546                                         struct super_block *sb, __u32 incompat);
1547 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1548                                       struct ext4_group_desc *bg);
1549 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1550                                       struct ext4_group_desc *bg);
1551 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1552                                      struct ext4_group_desc *bg);
1553 extern __u32 ext4_free_blks_count(struct super_block *sb,
1554                                 struct ext4_group_desc *bg);
1555 extern __u32 ext4_free_inodes_count(struct super_block *sb,
1556                                  struct ext4_group_desc *bg);
1557 extern __u32 ext4_used_dirs_count(struct super_block *sb,
1558                                 struct ext4_group_desc *bg);
1559 extern __u32 ext4_itable_unused_count(struct super_block *sb,
1560                                    struct ext4_group_desc *bg);
1561 extern void ext4_block_bitmap_set(struct super_block *sb,
1562                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1563 extern void ext4_inode_bitmap_set(struct super_block *sb,
1564                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1565 extern void ext4_inode_table_set(struct super_block *sb,
1566                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
1567 extern void ext4_free_blks_set(struct super_block *sb,
1568                                struct ext4_group_desc *bg, __u32 count);
1569 extern void ext4_free_inodes_set(struct super_block *sb,
1570                                 struct ext4_group_desc *bg, __u32 count);
1571 extern void ext4_used_dirs_set(struct super_block *sb,
1572                                 struct ext4_group_desc *bg, __u32 count);
1573 extern void ext4_itable_unused_set(struct super_block *sb,
1574                                    struct ext4_group_desc *bg, __u32 count);
1575 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
1576                                    struct ext4_group_desc *gdp);
1577 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
1578                                        struct ext4_group_desc *gdp);
1579
1580 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
1581 {
1582         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
1583                 le32_to_cpu(es->s_blocks_count_lo);
1584 }
1585
1586 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
1587 {
1588         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
1589                 le32_to_cpu(es->s_r_blocks_count_lo);
1590 }
1591
1592 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
1593 {
1594         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
1595                 le32_to_cpu(es->s_free_blocks_count_lo);
1596 }
1597
1598 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
1599                                          ext4_fsblk_t blk)
1600 {
1601         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
1602         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
1603 }
1604
1605 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
1606                                               ext4_fsblk_t blk)
1607 {
1608         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
1609         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
1610 }
1611
1612 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
1613                                            ext4_fsblk_t blk)
1614 {
1615         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
1616         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
1617 }
1618
1619 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
1620 {
1621         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
1622                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
1623                         le32_to_cpu(raw_inode->i_size_lo);
1624         else
1625                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
1626 }
1627
1628 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
1629 {
1630         raw_inode->i_size_lo = cpu_to_le32(i_size);
1631         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
1632 }
1633
1634 static inline
1635 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
1636                                             ext4_group_t group)
1637 {
1638          struct ext4_group_info ***grp_info;
1639          long indexv, indexh;
1640          grp_info = EXT4_SB(sb)->s_group_info;
1641          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
1642          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
1643          return grp_info[indexv][indexh];
1644 }
1645
1646 /*
1647  * Reading s_groups_count requires using smp_rmb() afterwards.  See
1648  * the locking protocol documented in the comments of ext4_group_add()
1649  * in resize.c
1650  */
1651 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
1652 {
1653         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
1654
1655         smp_rmb();
1656         return ngroups;
1657 }
1658
1659 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
1660                                              ext4_group_t block_group)
1661 {
1662         return block_group >> sbi->s_log_groups_per_flex;
1663 }
1664
1665 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
1666 {
1667         return 1 << sbi->s_log_groups_per_flex;
1668 }
1669
1670 #define ext4_std_error(sb, errno)                               \
1671 do {                                                            \
1672         if ((errno))                                            \
1673                 __ext4_std_error((sb), __func__, (errno));      \
1674 } while (0)
1675
1676 #ifdef CONFIG_SMP
1677 /* Each CPU can accumulate percpu_counter_batch blocks in their local
1678  * counters. So we need to make sure we have free blocks more
1679  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
1680  */
1681 #define EXT4_FREEBLOCKS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
1682 #else
1683 #define EXT4_FREEBLOCKS_WATERMARK 0
1684 #endif
1685
1686 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
1687 {
1688         /*
1689          * XXX: replace with spinlock if seen contended -bzzz
1690          */
1691         down_write(&EXT4_I(inode)->i_data_sem);
1692         if (newsize > EXT4_I(inode)->i_disksize)
1693                 EXT4_I(inode)->i_disksize = newsize;
1694         up_write(&EXT4_I(inode)->i_data_sem);
1695         return ;
1696 }
1697
1698 struct ext4_group_info {
1699         unsigned long   bb_state;
1700         struct rb_root  bb_free_root;
1701         ext4_grpblk_t   bb_first_free;  /* first free block */
1702         ext4_grpblk_t   bb_free;        /* total free blocks */
1703         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
1704         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
1705         struct          list_head bb_prealloc_list;
1706 #ifdef DOUBLE_CHECK
1707         void            *bb_bitmap;
1708 #endif
1709         struct rw_semaphore alloc_sem;
1710         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
1711                                          * regions, index is order.
1712                                          * bb_counters[3] = 5 means
1713                                          * 5 free 8-block regions. */
1714 };
1715
1716 #define EXT4_GROUP_INFO_NEED_INIT_BIT   0
1717
1718 #define EXT4_MB_GRP_NEED_INIT(grp)      \
1719         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
1720
1721 #define EXT4_MAX_CONTENTION             8
1722 #define EXT4_CONTENTION_THRESHOLD       2
1723
1724 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
1725                                               ext4_group_t group)
1726 {
1727         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
1728 }
1729
1730 /*
1731  * Returns true if the filesystem is busy enough that attempts to
1732  * access the block group locks has run into contention.
1733  */
1734 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
1735 {
1736         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
1737 }
1738
1739 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
1740 {
1741         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
1742         if (spin_trylock(lock))
1743                 /*
1744                  * We're able to grab the lock right away, so drop the
1745                  * lock contention counter.
1746                  */
1747                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
1748         else {
1749                 /*
1750                  * The lock is busy, so bump the contention counter,
1751                  * and then wait on the spin lock.
1752                  */
1753                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
1754                                   EXT4_MAX_CONTENTION);
1755                 spin_lock(lock);
1756         }
1757 }
1758
1759 static inline void ext4_unlock_group(struct super_block *sb,
1760                                         ext4_group_t group)
1761 {
1762         spin_unlock(ext4_group_lock_ptr(sb, group));
1763 }
1764
1765 /*
1766  * Inodes and files operations
1767  */
1768
1769 /* dir.c */
1770 extern const struct file_operations ext4_dir_operations;
1771
1772 /* file.c */
1773 extern const struct inode_operations ext4_file_inode_operations;
1774 extern const struct file_operations ext4_file_operations;
1775
1776 /* namei.c */
1777 extern const struct inode_operations ext4_dir_inode_operations;
1778 extern const struct inode_operations ext4_special_inode_operations;
1779 extern struct dentry *ext4_get_parent(struct dentry *child);
1780
1781 /* symlink.c */
1782 extern const struct inode_operations ext4_symlink_inode_operations;
1783 extern const struct inode_operations ext4_fast_symlink_inode_operations;
1784
1785 /* block_validity */
1786 extern void ext4_release_system_zone(struct super_block *sb);
1787 extern int ext4_setup_system_zone(struct super_block *sb);
1788 extern int __init init_ext4_system_zone(void);
1789 extern void exit_ext4_system_zone(void);
1790 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
1791                                  ext4_fsblk_t start_blk,
1792                                  unsigned int count);
1793
1794 /* extents.c */
1795 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
1796 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
1797 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
1798                                        int chunk);
1799 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
1800                                struct ext4_map_blocks *map, int flags);
1801 extern void ext4_ext_truncate(struct inode *);
1802 extern void ext4_ext_init(struct super_block *);
1803 extern void ext4_ext_release(struct super_block *);
1804 extern long ext4_fallocate(struct inode *inode, int mode, loff_t offset,
1805                           loff_t len);
1806 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
1807                           ssize_t len);
1808 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
1809                            struct ext4_map_blocks *map, int flags);
1810 extern int ext4_get_blocks(handle_t *handle, struct inode *inode,
1811                            sector_t block, unsigned int max_blocks,
1812                            struct buffer_head *bh, int flags);
1813 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1814                         __u64 start, __u64 len);
1815 /* move_extent.c */
1816 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
1817                              __u64 start_orig, __u64 start_donor,
1818                              __u64 len, __u64 *moved_len);
1819
1820
1821 /* BH_Uninit flag: blocks are allocated but uninitialized on disk */
1822 enum ext4_state_bits {
1823         BH_Uninit       /* blocks are allocated but uninitialized on disk */
1824           = BH_JBDPrivateStart,
1825 };
1826
1827 BUFFER_FNS(Uninit, uninit)
1828 TAS_BUFFER_FNS(Uninit, uninit)
1829
1830 /*
1831  * Add new method to test wether block and inode bitmaps are properly
1832  * initialized. With uninit_bg reading the block from disk is not enough
1833  * to mark the bitmap uptodate. We need to also zero-out the bitmap
1834  */
1835 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
1836
1837 static inline int bitmap_uptodate(struct buffer_head *bh)
1838 {
1839         return (buffer_uptodate(bh) &&
1840                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
1841 }
1842 static inline void set_bitmap_uptodate(struct buffer_head *bh)
1843 {
1844         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
1845 }
1846
1847 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
1848
1849 #endif  /* __KERNEL__ */
1850
1851 #endif  /* _EXT4_H */