ext4: improve llseek error handling for overly large seek offsets
[linux-2.6.git] / fs / ext4 / ext4.h
1 /*
2  *  ext4.h
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/include/linux/minix_fs.h
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  */
15
16 #ifndef _EXT4_H
17 #define _EXT4_H
18
19 #include <linux/types.h>
20 #include <linux/blkdev.h>
21 #include <linux/magic.h>
22 #include <linux/jbd2.h>
23 #include <linux/quota.h>
24 #include <linux/rwsem.h>
25 #include <linux/rbtree.h>
26 #include <linux/seqlock.h>
27 #include <linux/mutex.h>
28 #include <linux/timer.h>
29 #include <linux/wait.h>
30 #include <linux/blockgroup_lock.h>
31 #include <linux/percpu_counter.h>
32 #ifdef __KERNEL__
33 #include <linux/compat.h>
34 #endif
35
36 /*
37  * The fourth extended filesystem constants/structures
38  */
39
40 /*
41  * Define EXT4FS_DEBUG to produce debug messages
42  */
43 #undef EXT4FS_DEBUG
44
45 /*
46  * Debug code
47  */
48 #ifdef EXT4FS_DEBUG
49 #define ext4_debug(f, a...)                                             \
50         do {                                                            \
51                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
52                         __FILE__, __LINE__, __func__);                  \
53                 printk(KERN_DEBUG f, ## a);                             \
54         } while (0)
55 #else
56 #define ext4_debug(f, a...)     do {} while (0)
57 #endif
58
59 #define EXT4_ERROR_INODE(inode, fmt, a...) \
60         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
61
62 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...)                 \
63         ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
64
65 #define EXT4_ERROR_FILE(file, fmt, a...)        \
66         ext4_error_file(__func__, __LINE__, (file), (fmt), ## a)
67
68 /* data type for block offset of block group */
69 typedef int ext4_grpblk_t;
70
71 /* data type for filesystem-wide blocks number */
72 typedef unsigned long long ext4_fsblk_t;
73
74 /* data type for file logical block number */
75 typedef __u32 ext4_lblk_t;
76
77 /* data type for block group number */
78 typedef unsigned int ext4_group_t;
79
80 /*
81  * Flags used in mballoc's allocation_context flags field.
82  *
83  * Also used to show what's going on for debugging purposes when the
84  * flag field is exported via the traceport interface
85  */
86
87 /* prefer goal again. length */
88 #define EXT4_MB_HINT_MERGE              0x0001
89 /* blocks already reserved */
90 #define EXT4_MB_HINT_RESERVED           0x0002
91 /* metadata is being allocated */
92 #define EXT4_MB_HINT_METADATA           0x0004
93 /* first blocks in the file */
94 #define EXT4_MB_HINT_FIRST              0x0008
95 /* search for the best chunk */
96 #define EXT4_MB_HINT_BEST               0x0010
97 /* data is being allocated */
98 #define EXT4_MB_HINT_DATA               0x0020
99 /* don't preallocate (for tails) */
100 #define EXT4_MB_HINT_NOPREALLOC         0x0040
101 /* allocate for locality group */
102 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
103 /* allocate goal blocks or none */
104 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
105 /* goal is meaningful */
106 #define EXT4_MB_HINT_TRY_GOAL           0x0200
107 /* blocks already pre-reserved by delayed allocation */
108 #define EXT4_MB_DELALLOC_RESERVED       0x0400
109 /* We are doing stream allocation */
110 #define EXT4_MB_STREAM_ALLOC            0x0800
111
112
113 struct ext4_allocation_request {
114         /* target inode for block we're allocating */
115         struct inode *inode;
116         /* how many blocks we want to allocate */
117         unsigned int len;
118         /* logical block in target inode */
119         ext4_lblk_t logical;
120         /* the closest logical allocated block to the left */
121         ext4_lblk_t lleft;
122         /* the closest logical allocated block to the right */
123         ext4_lblk_t lright;
124         /* phys. target (a hint) */
125         ext4_fsblk_t goal;
126         /* phys. block for the closest logical allocated block to the left */
127         ext4_fsblk_t pleft;
128         /* phys. block for the closest logical allocated block to the right */
129         ext4_fsblk_t pright;
130         /* flags. see above EXT4_MB_HINT_* */
131         unsigned int flags;
132 };
133
134 /*
135  * Logical to physical block mapping, used by ext4_map_blocks()
136  *
137  * This structure is used to pass requests into ext4_map_blocks() as
138  * well as to store the information returned by ext4_map_blocks().  It
139  * takes less room on the stack than a struct buffer_head.
140  */
141 #define EXT4_MAP_NEW            (1 << BH_New)
142 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
143 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
144 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
145 #define EXT4_MAP_UNINIT         (1 << BH_Uninit)
146 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
147                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
148                                  EXT4_MAP_UNINIT)
149
150 struct ext4_map_blocks {
151         ext4_fsblk_t m_pblk;
152         ext4_lblk_t m_lblk;
153         unsigned int m_len;
154         unsigned int m_flags;
155 };
156
157 /*
158  * For delayed allocation tracking
159  */
160 struct mpage_da_data {
161         struct inode *inode;
162         sector_t b_blocknr;             /* start block number of extent */
163         size_t b_size;                  /* size of extent */
164         unsigned long b_state;          /* state of the extent */
165         unsigned long first_page, next_page;    /* extent of pages */
166         struct writeback_control *wbc;
167         int io_done;
168         int pages_written;
169         int retval;
170 };
171 #define EXT4_IO_UNWRITTEN       0x1
172 typedef struct ext4_io_end {
173         struct list_head        list;           /* per-file finished IO list */
174         struct inode            *inode;         /* file being written to */
175         unsigned int            flag;           /* unwritten or not */
176         struct page             *page;          /* page struct for buffer write */
177         loff_t                  offset;         /* offset in the file */
178         ssize_t                 size;           /* size of the extent */
179         struct work_struct      work;           /* data work queue */
180         struct kiocb            *iocb;          /* iocb struct for AIO */
181         int                     result;         /* error value for AIO */
182 } ext4_io_end_t;
183
184 /*
185  * Special inodes numbers
186  */
187 #define EXT4_BAD_INO             1      /* Bad blocks inode */
188 #define EXT4_ROOT_INO            2      /* Root inode */
189 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
190 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
191 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
192 #define EXT4_JOURNAL_INO         8      /* Journal inode */
193
194 /* First non-reserved inode for old ext4 filesystems */
195 #define EXT4_GOOD_OLD_FIRST_INO 11
196
197 /*
198  * Maximal count of links to a file
199  */
200 #define EXT4_LINK_MAX           65000
201
202 /*
203  * Macro-instructions used to manage several block sizes
204  */
205 #define EXT4_MIN_BLOCK_SIZE             1024
206 #define EXT4_MAX_BLOCK_SIZE             65536
207 #define EXT4_MIN_BLOCK_LOG_SIZE         10
208 #define EXT4_MAX_BLOCK_LOG_SIZE         16
209 #ifdef __KERNEL__
210 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
211 #else
212 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
213 #endif
214 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
215 #ifdef __KERNEL__
216 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
217 #else
218 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
219 #endif
220 #ifdef __KERNEL__
221 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
222 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
223 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
224 #else
225 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
226                                  EXT4_GOOD_OLD_INODE_SIZE : \
227                                  (s)->s_inode_size)
228 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
229                                  EXT4_GOOD_OLD_FIRST_INO : \
230                                  (s)->s_first_ino)
231 #endif
232 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
233
234 /*
235  * Structure of a blocks group descriptor
236  */
237 struct ext4_group_desc
238 {
239         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
240         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
241         __le32  bg_inode_table_lo;      /* Inodes table block */
242         __le16  bg_free_blocks_count_lo;/* Free blocks count */
243         __le16  bg_free_inodes_count_lo;/* Free inodes count */
244         __le16  bg_used_dirs_count_lo;  /* Directories count */
245         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
246         __u32   bg_reserved[2];         /* Likely block/inode bitmap checksum */
247         __le16  bg_itable_unused_lo;    /* Unused inodes count */
248         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
249         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
250         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
251         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
252         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
253         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
254         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
255         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
256         __u32   bg_reserved2[3];
257 };
258
259 /*
260  * Structure of a flex block group info
261  */
262
263 struct flex_groups {
264         atomic_t free_inodes;
265         atomic_t free_blocks;
266         atomic_t used_dirs;
267 };
268
269 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
270 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
271 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
272
273 /*
274  * Macro-instructions used to manage group descriptors
275  */
276 #define EXT4_MIN_DESC_SIZE              32
277 #define EXT4_MIN_DESC_SIZE_64BIT        64
278 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
279 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
280 #ifdef __KERNEL__
281 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
282 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
283 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
284 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
285 #else
286 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
287 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
288 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
289 #endif
290
291 /*
292  * Constants relative to the data blocks
293  */
294 #define EXT4_NDIR_BLOCKS                12
295 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
296 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
297 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
298 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
299
300 /*
301  * Inode flags
302  */
303 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
304 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
305 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
306 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
307 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
308 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
309 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
310 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
311 /* Reserved for compression usage... */
312 #define EXT4_DIRTY_FL                   0x00000100
313 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
314 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
315 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
316 /* End compression flags --- maybe not all used */
317 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
318 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
319 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
320 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
321 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
322 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
323 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
324 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
325 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
326 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
327 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
328
329 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
330 #define EXT4_FL_USER_MODIFIABLE         0x004B80FF /* User modifiable flags */
331
332 /* Flags that should be inherited by new inodes from their parent. */
333 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
334                            EXT4_SYNC_FL | EXT4_IMMUTABLE_FL | EXT4_APPEND_FL |\
335                            EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
336                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
337                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
338
339 /* Flags that are appropriate for regular files (all but dir-specific ones). */
340 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
341
342 /* Flags that are appropriate for non-directories/regular files. */
343 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
344
345 /* Mask out flags that are inappropriate for the given type of inode. */
346 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
347 {
348         if (S_ISDIR(mode))
349                 return flags;
350         else if (S_ISREG(mode))
351                 return flags & EXT4_REG_FLMASK;
352         else
353                 return flags & EXT4_OTHER_FLMASK;
354 }
355
356 /*
357  * Inode flags used for atomic set/get
358  */
359 enum {
360         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
361         EXT4_INODE_UNRM         = 1,    /* Undelete */
362         EXT4_INODE_COMPR        = 2,    /* Compress file */
363         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
364         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
365         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
366         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
367         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
368 /* Reserved for compression usage... */
369         EXT4_INODE_DIRTY        = 8,
370         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
371         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
372         EXT4_INODE_ECOMPR       = 11,   /* Compression error */
373 /* End compression flags --- maybe not all used */
374         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
375         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
376         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
377         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
378         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
379         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
380         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
381         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
382         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
383         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
384         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
385 };
386
387 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
388 #define CHECK_FLAG_VALUE(FLAG) if (!TEST_FLAG_VALUE(FLAG)) { \
389         printk(KERN_EMERG "EXT4 flag fail: " #FLAG ": %d %d\n", \
390                 EXT4_##FLAG##_FL, EXT4_INODE_##FLAG); BUG_ON(1); }
391
392 /*
393  * Since it's pretty easy to mix up bit numbers and hex values, and we
394  * can't do a compile-time test for ENUM values, we use a run-time
395  * test to make sure that EXT4_XXX_FL is consistent with respect to
396  * EXT4_INODE_XXX.  If all is well the printk and BUG_ON will all drop
397  * out so it won't cost any extra space in the compiled kernel image.
398  * But it's important that these values are the same, since we are
399  * using EXT4_INODE_XXX to test for the flag values, but EXT4_XX_FL
400  * must be consistent with the values of FS_XXX_FL defined in
401  * include/linux/fs.h and the on-disk values found in ext2, ext3, and
402  * ext4 filesystems, and of course the values defined in e2fsprogs.
403  *
404  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
405  */
406 static inline void ext4_check_flag_values(void)
407 {
408         CHECK_FLAG_VALUE(SECRM);
409         CHECK_FLAG_VALUE(UNRM);
410         CHECK_FLAG_VALUE(COMPR);
411         CHECK_FLAG_VALUE(SYNC);
412         CHECK_FLAG_VALUE(IMMUTABLE);
413         CHECK_FLAG_VALUE(APPEND);
414         CHECK_FLAG_VALUE(NODUMP);
415         CHECK_FLAG_VALUE(NOATIME);
416         CHECK_FLAG_VALUE(DIRTY);
417         CHECK_FLAG_VALUE(COMPRBLK);
418         CHECK_FLAG_VALUE(NOCOMPR);
419         CHECK_FLAG_VALUE(ECOMPR);
420         CHECK_FLAG_VALUE(INDEX);
421         CHECK_FLAG_VALUE(IMAGIC);
422         CHECK_FLAG_VALUE(JOURNAL_DATA);
423         CHECK_FLAG_VALUE(NOTAIL);
424         CHECK_FLAG_VALUE(DIRSYNC);
425         CHECK_FLAG_VALUE(TOPDIR);
426         CHECK_FLAG_VALUE(HUGE_FILE);
427         CHECK_FLAG_VALUE(EXTENTS);
428         CHECK_FLAG_VALUE(EA_INODE);
429         CHECK_FLAG_VALUE(EOFBLOCKS);
430         CHECK_FLAG_VALUE(RESERVED);
431 }
432
433 /* Used to pass group descriptor data when online resize is done */
434 struct ext4_new_group_input {
435         __u32 group;            /* Group number for this data */
436         __u64 block_bitmap;     /* Absolute block number of block bitmap */
437         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
438         __u64 inode_table;      /* Absolute block number of inode table start */
439         __u32 blocks_count;     /* Total number of blocks in this group */
440         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
441         __u16 unused;
442 };
443
444 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
445 struct compat_ext4_new_group_input {
446         u32 group;
447         compat_u64 block_bitmap;
448         compat_u64 inode_bitmap;
449         compat_u64 inode_table;
450         u32 blocks_count;
451         u16 reserved_blocks;
452         u16 unused;
453 };
454 #endif
455
456 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
457 struct ext4_new_group_data {
458         __u32 group;
459         __u64 block_bitmap;
460         __u64 inode_bitmap;
461         __u64 inode_table;
462         __u32 blocks_count;
463         __u16 reserved_blocks;
464         __u16 unused;
465         __u32 free_blocks_count;
466 };
467
468 /*
469  * Flags used by ext4_map_blocks()
470  */
471         /* Allocate any needed blocks and/or convert an unitialized
472            extent to be an initialized ext4 */
473 #define EXT4_GET_BLOCKS_CREATE                  0x0001
474         /* Request the creation of an unitialized extent */
475 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
476 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
477                                                  EXT4_GET_BLOCKS_CREATE)
478         /* Caller is from the delayed allocation writeout path,
479            so set the magic i_delalloc_reserve_flag after taking the
480            inode allocation semaphore for */
481 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
482         /* caller is from the direct IO path, request to creation of an
483         unitialized extents if not allocated, split the uninitialized
484         extent if blocks has been preallocated already*/
485 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
486 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
487 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
488                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
489         /* Convert extent to initialized after IO complete */
490 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
491                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
492
493 /*
494  * Flags used by ext4_free_blocks
495  */
496 #define EXT4_FREE_BLOCKS_METADATA       0x0001
497 #define EXT4_FREE_BLOCKS_FORGET         0x0002
498 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
499
500 /*
501  * ioctl commands
502  */
503 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
504 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
505 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
506 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
507 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
508 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
509 #ifdef CONFIG_JBD2_DEBUG
510 #define EXT4_IOC_WAIT_FOR_READONLY      _IOR('f', 99, long)
511 #endif
512 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
513 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
514 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
515 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
516 #define EXT4_IOC_MIGRATE                _IO('f', 9)
517  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
518  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
519 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
520 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
521
522 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
523 /*
524  * ioctl commands in 32 bit emulation
525  */
526 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
527 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
528 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
529 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
530 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
531 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
532 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
533 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
534 #ifdef CONFIG_JBD2_DEBUG
535 #define EXT4_IOC32_WAIT_FOR_READONLY    _IOR('f', 99, int)
536 #endif
537 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
538 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
539 #endif
540
541
542 /*
543  *  Mount options
544  */
545 struct ext4_mount_options {
546         unsigned long s_mount_opt;
547         uid_t s_resuid;
548         gid_t s_resgid;
549         unsigned long s_commit_interval;
550         u32 s_min_batch_time, s_max_batch_time;
551 #ifdef CONFIG_QUOTA
552         int s_jquota_fmt;
553         char *s_qf_names[MAXQUOTAS];
554 #endif
555 };
556
557 /* Max physical block we can addres w/o extents */
558 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
559
560 /*
561  * Structure of an inode on the disk
562  */
563 struct ext4_inode {
564         __le16  i_mode;         /* File mode */
565         __le16  i_uid;          /* Low 16 bits of Owner Uid */
566         __le32  i_size_lo;      /* Size in bytes */
567         __le32  i_atime;        /* Access time */
568         __le32  i_ctime;        /* Inode Change time */
569         __le32  i_mtime;        /* Modification time */
570         __le32  i_dtime;        /* Deletion Time */
571         __le16  i_gid;          /* Low 16 bits of Group Id */
572         __le16  i_links_count;  /* Links count */
573         __le32  i_blocks_lo;    /* Blocks count */
574         __le32  i_flags;        /* File flags */
575         union {
576                 struct {
577                         __le32  l_i_version;
578                 } linux1;
579                 struct {
580                         __u32  h_i_translator;
581                 } hurd1;
582                 struct {
583                         __u32  m_i_reserved1;
584                 } masix1;
585         } osd1;                         /* OS dependent 1 */
586         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
587         __le32  i_generation;   /* File version (for NFS) */
588         __le32  i_file_acl_lo;  /* File ACL */
589         __le32  i_size_high;
590         __le32  i_obso_faddr;   /* Obsoleted fragment address */
591         union {
592                 struct {
593                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
594                         __le16  l_i_file_acl_high;
595                         __le16  l_i_uid_high;   /* these 2 fields */
596                         __le16  l_i_gid_high;   /* were reserved2[0] */
597                         __u32   l_i_reserved2;
598                 } linux2;
599                 struct {
600                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
601                         __u16   h_i_mode_high;
602                         __u16   h_i_uid_high;
603                         __u16   h_i_gid_high;
604                         __u32   h_i_author;
605                 } hurd2;
606                 struct {
607                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
608                         __le16  m_i_file_acl_high;
609                         __u32   m_i_reserved2[2];
610                 } masix2;
611         } osd2;                         /* OS dependent 2 */
612         __le16  i_extra_isize;
613         __le16  i_pad1;
614         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
615         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
616         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
617         __le32  i_crtime;       /* File Creation time */
618         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
619         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
620 };
621
622 struct move_extent {
623         __u32 reserved;         /* should be zero */
624         __u32 donor_fd;         /* donor file descriptor */
625         __u64 orig_start;       /* logical start offset in block for orig */
626         __u64 donor_start;      /* logical start offset in block for donor */
627         __u64 len;              /* block length to be moved */
628         __u64 moved_len;        /* moved block length */
629 };
630
631 #define EXT4_EPOCH_BITS 2
632 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
633 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
634
635 /*
636  * Extended fields will fit into an inode if the filesystem was formatted
637  * with large inodes (-I 256 or larger) and there are not currently any EAs
638  * consuming all of the available space. For new inodes we always reserve
639  * enough space for the kernel's known extended fields, but for inodes
640  * created with an old kernel this might not have been the case. None of
641  * the extended inode fields is critical for correct filesystem operation.
642  * This macro checks if a certain field fits in the inode. Note that
643  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
644  */
645 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
646         ((offsetof(typeof(*ext4_inode), field) +        \
647           sizeof((ext4_inode)->field))                  \
648         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
649             (einode)->i_extra_isize))                   \
650
651 static inline __le32 ext4_encode_extra_time(struct timespec *time)
652 {
653        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
654                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
655                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
656 }
657
658 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
659 {
660        if (sizeof(time->tv_sec) > 4)
661                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
662                                << 32;
663        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
664 }
665
666 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
667 do {                                                                           \
668         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
669         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
670                 (raw_inode)->xtime ## _extra =                                 \
671                                 ext4_encode_extra_time(&(inode)->xtime);       \
672 } while (0)
673
674 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
675 do {                                                                           \
676         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
677                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
678         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
679                 (raw_inode)->xtime ## _extra =                                 \
680                                 ext4_encode_extra_time(&(einode)->xtime);      \
681 } while (0)
682
683 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
684 do {                                                                           \
685         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
686         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
687                 ext4_decode_extra_time(&(inode)->xtime,                        \
688                                        raw_inode->xtime ## _extra);            \
689 } while (0)
690
691 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
692 do {                                                                           \
693         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
694                 (einode)->xtime.tv_sec =                                       \
695                         (signed)le32_to_cpu((raw_inode)->xtime);               \
696         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
697                 ext4_decode_extra_time(&(einode)->xtime,                       \
698                                        raw_inode->xtime ## _extra);            \
699 } while (0)
700
701 #define i_disk_version osd1.linux1.l_i_version
702
703 #if defined(__KERNEL__) || defined(__linux__)
704 #define i_reserved1     osd1.linux1.l_i_reserved1
705 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
706 #define i_blocks_high   osd2.linux2.l_i_blocks_high
707 #define i_uid_low       i_uid
708 #define i_gid_low       i_gid
709 #define i_uid_high      osd2.linux2.l_i_uid_high
710 #define i_gid_high      osd2.linux2.l_i_gid_high
711 #define i_reserved2     osd2.linux2.l_i_reserved2
712
713 #elif defined(__GNU__)
714
715 #define i_translator    osd1.hurd1.h_i_translator
716 #define i_uid_high      osd2.hurd2.h_i_uid_high
717 #define i_gid_high      osd2.hurd2.h_i_gid_high
718 #define i_author        osd2.hurd2.h_i_author
719
720 #elif defined(__masix__)
721
722 #define i_reserved1     osd1.masix1.m_i_reserved1
723 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
724 #define i_reserved2     osd2.masix2.m_i_reserved2
725
726 #endif /* defined(__KERNEL__) || defined(__linux__) */
727
728 /*
729  * storage for cached extent
730  */
731 struct ext4_ext_cache {
732         ext4_fsblk_t    ec_start;
733         ext4_lblk_t     ec_block;
734         __u32           ec_len; /* must be 32bit to return holes */
735         __u32           ec_type;
736 };
737
738 /*
739  * fourth extended file system inode data in memory
740  */
741 struct ext4_inode_info {
742         __le32  i_data[15];     /* unconverted */
743         __u32   i_dtime;
744         ext4_fsblk_t    i_file_acl;
745
746         /*
747          * i_block_group is the number of the block group which contains
748          * this file's inode.  Constant across the lifetime of the inode,
749          * it is ued for making block allocation decisions - we try to
750          * place a file's data blocks near its inode block, and new inodes
751          * near to their parent directory's inode.
752          */
753         ext4_group_t    i_block_group;
754         unsigned long   i_state_flags;          /* Dynamic state flags */
755         unsigned long   i_flags;
756
757         ext4_lblk_t             i_dir_start_lookup;
758 #ifdef CONFIG_EXT4_FS_XATTR
759         /*
760          * Extended attributes can be read independently of the main file
761          * data. Taking i_mutex even when reading would cause contention
762          * between readers of EAs and writers of regular file data, so
763          * instead we synchronize on xattr_sem when reading or changing
764          * EAs.
765          */
766         struct rw_semaphore xattr_sem;
767 #endif
768
769         struct list_head i_orphan;      /* unlinked but open inodes */
770
771         /*
772          * i_disksize keeps track of what the inode size is ON DISK, not
773          * in memory.  During truncate, i_size is set to the new size by
774          * the VFS prior to calling ext4_truncate(), but the filesystem won't
775          * set i_disksize to 0 until the truncate is actually under way.
776          *
777          * The intent is that i_disksize always represents the blocks which
778          * are used by this file.  This allows recovery to restart truncate
779          * on orphans if we crash during truncate.  We actually write i_disksize
780          * into the on-disk inode when writing inodes out, instead of i_size.
781          *
782          * The only time when i_disksize and i_size may be different is when
783          * a truncate is in progress.  The only things which change i_disksize
784          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
785          */
786         loff_t  i_disksize;
787
788         /*
789          * i_data_sem is for serialising ext4_truncate() against
790          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
791          * data tree are chopped off during truncate. We can't do that in
792          * ext4 because whenever we perform intermediate commits during
793          * truncate, the inode and all the metadata blocks *must* be in a
794          * consistent state which allows truncation of the orphans to restart
795          * during recovery.  Hence we must fix the get_block-vs-truncate race
796          * by other means, so we have i_data_sem.
797          */
798         struct rw_semaphore i_data_sem;
799         struct inode vfs_inode;
800         struct jbd2_inode jinode;
801
802         struct ext4_ext_cache i_cached_extent;
803         /*
804          * File creation time. Its function is same as that of
805          * struct timespec i_{a,c,m}time in the generic inode.
806          */
807         struct timespec i_crtime;
808
809         /* mballoc */
810         struct list_head i_prealloc_list;
811         spinlock_t i_prealloc_lock;
812
813         /* ialloc */
814         ext4_group_t    i_last_alloc_group;
815
816         /* allocation reservation info for delalloc */
817         unsigned int i_reserved_data_blocks;
818         unsigned int i_reserved_meta_blocks;
819         unsigned int i_allocated_meta_blocks;
820         unsigned short i_delalloc_reserved_flag;
821         sector_t i_da_metadata_calc_last_lblock;
822         int i_da_metadata_calc_len;
823
824         /* on-disk additional length */
825         __u16 i_extra_isize;
826
827         spinlock_t i_block_reservation_lock;
828 #ifdef CONFIG_QUOTA
829         /* quota space reservation, managed internally by quota code */
830         qsize_t i_reserved_quota;
831 #endif
832
833         /* completed IOs that might need unwritten extents handling */
834         struct list_head i_completed_io_list;
835         spinlock_t i_completed_io_lock;
836         /* current io_end structure for async DIO write*/
837         ext4_io_end_t *cur_aio_dio;
838
839         /*
840          * Transactions that contain inode's metadata needed to complete
841          * fsync and fdatasync, respectively.
842          */
843         tid_t i_sync_tid;
844         tid_t i_datasync_tid;
845 };
846
847 /*
848  * File system states
849  */
850 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
851 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
852 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
853
854 /*
855  * Misc. filesystem flags
856  */
857 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
858 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
859 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
860
861 /*
862  * Mount flags
863  */
864 #define EXT4_MOUNT_OLDALLOC             0x00002  /* Don't use the new Orlov allocator */
865 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
866 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
867 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
868 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
869 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
870 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
871 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
872 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
873 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
874 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
875 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
876 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
877 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
878 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
879 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
880 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
881 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
882 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
883 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
884 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
885 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
886 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
887 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
888 #define EXT4_MOUNT_I_VERSION            0x2000000 /* i_version support */
889 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
890 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
891 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
892 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
893 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
894
895 #define clear_opt(o, opt)               o &= ~EXT4_MOUNT_##opt
896 #define set_opt(o, opt)                 o |= EXT4_MOUNT_##opt
897 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
898                                          EXT4_MOUNT_##opt)
899
900 #define ext4_set_bit                    ext2_set_bit
901 #define ext4_set_bit_atomic             ext2_set_bit_atomic
902 #define ext4_clear_bit                  ext2_clear_bit
903 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
904 #define ext4_test_bit                   ext2_test_bit
905 #define ext4_find_first_zero_bit        ext2_find_first_zero_bit
906 #define ext4_find_next_zero_bit         ext2_find_next_zero_bit
907 #define ext4_find_next_bit              ext2_find_next_bit
908
909 /*
910  * Maximal mount counts between two filesystem checks
911  */
912 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
913 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
914
915 /*
916  * Behaviour when detecting errors
917  */
918 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
919 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
920 #define EXT4_ERRORS_PANIC               3       /* Panic */
921 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
922
923 /*
924  * Structure of the super block
925  */
926 struct ext4_super_block {
927 /*00*/  __le32  s_inodes_count;         /* Inodes count */
928         __le32  s_blocks_count_lo;      /* Blocks count */
929         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
930         __le32  s_free_blocks_count_lo; /* Free blocks count */
931 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
932         __le32  s_first_data_block;     /* First Data Block */
933         __le32  s_log_block_size;       /* Block size */
934         __le32  s_obso_log_frag_size;   /* Obsoleted fragment size */
935 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
936         __le32  s_obso_frags_per_group; /* Obsoleted fragments per group */
937         __le32  s_inodes_per_group;     /* # Inodes per group */
938         __le32  s_mtime;                /* Mount time */
939 /*30*/  __le32  s_wtime;                /* Write time */
940         __le16  s_mnt_count;            /* Mount count */
941         __le16  s_max_mnt_count;        /* Maximal mount count */
942         __le16  s_magic;                /* Magic signature */
943         __le16  s_state;                /* File system state */
944         __le16  s_errors;               /* Behaviour when detecting errors */
945         __le16  s_minor_rev_level;      /* minor revision level */
946 /*40*/  __le32  s_lastcheck;            /* time of last check */
947         __le32  s_checkinterval;        /* max. time between checks */
948         __le32  s_creator_os;           /* OS */
949         __le32  s_rev_level;            /* Revision level */
950 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
951         __le16  s_def_resgid;           /* Default gid for reserved blocks */
952         /*
953          * These fields are for EXT4_DYNAMIC_REV superblocks only.
954          *
955          * Note: the difference between the compatible feature set and
956          * the incompatible feature set is that if there is a bit set
957          * in the incompatible feature set that the kernel doesn't
958          * know about, it should refuse to mount the filesystem.
959          *
960          * e2fsck's requirements are more strict; if it doesn't know
961          * about a feature in either the compatible or incompatible
962          * feature set, it must abort and not try to meddle with
963          * things it doesn't understand...
964          */
965         __le32  s_first_ino;            /* First non-reserved inode */
966         __le16  s_inode_size;           /* size of inode structure */
967         __le16  s_block_group_nr;       /* block group # of this superblock */
968         __le32  s_feature_compat;       /* compatible feature set */
969 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
970         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
971 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
972 /*78*/  char    s_volume_name[16];      /* volume name */
973 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
974 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
975         /*
976          * Performance hints.  Directory preallocation should only
977          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
978          */
979         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
980         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
981         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
982         /*
983          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
984          */
985 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
986 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
987         __le32  s_journal_dev;          /* device number of journal file */
988         __le32  s_last_orphan;          /* start of list of inodes to delete */
989         __le32  s_hash_seed[4];         /* HTREE hash seed */
990         __u8    s_def_hash_version;     /* Default hash version to use */
991         __u8    s_jnl_backup_type;
992         __le16  s_desc_size;            /* size of group descriptor */
993 /*100*/ __le32  s_default_mount_opts;
994         __le32  s_first_meta_bg;        /* First metablock block group */
995         __le32  s_mkfs_time;            /* When the filesystem was created */
996         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
997         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
998 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
999         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1000         __le32  s_free_blocks_count_hi; /* Free blocks count */
1001         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1002         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1003         __le32  s_flags;                /* Miscellaneous flags */
1004         __le16  s_raid_stride;          /* RAID stride */
1005         __le16  s_mmp_interval;         /* # seconds to wait in MMP checking */
1006         __le64  s_mmp_block;            /* Block for multi-mount protection */
1007         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1008         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1009         __u8    s_reserved_char_pad;
1010         __le16  s_reserved_pad;
1011         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1012         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1013         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1014         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1015                                               snapshot's future use */
1016         __le32  s_snapshot_list;        /* inode number of the head of the
1017                                            on-disk snapshot list */
1018 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1019         __le32  s_error_count;          /* number of fs errors */
1020         __le32  s_first_error_time;     /* first time an error happened */
1021         __le32  s_first_error_ino;      /* inode involved in first error */
1022         __le64  s_first_error_block;    /* block involved of first error */
1023         __u8    s_first_error_func[32]; /* function where the error happened */
1024         __le32  s_first_error_line;     /* line number where error happened */
1025         __le32  s_last_error_time;      /* most recent time of an error */
1026         __le32  s_last_error_ino;       /* inode involved in last error */
1027         __le32  s_last_error_line;      /* line number where error happened */
1028         __le64  s_last_error_block;     /* block involved of last error */
1029         __u8    s_last_error_func[32];  /* function where the error happened */
1030 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1031         __u8    s_mount_opts[64];
1032         __le32  s_reserved[112];        /* Padding to the end of the block */
1033 };
1034
1035 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1036
1037 #ifdef __KERNEL__
1038
1039 /*
1040  * run-time mount flags
1041  */
1042 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
1043 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
1044
1045 /*
1046  * fourth extended-fs super-block data in memory
1047  */
1048 struct ext4_sb_info {
1049         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1050         unsigned long s_inodes_per_block;/* Number of inodes per block */
1051         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1052         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1053         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1054         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1055         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1056         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1057         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1058         unsigned long s_overhead_last;  /* Last calculated overhead */
1059         unsigned long s_blocks_last;    /* Last seen block count */
1060         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1061         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1062         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1063         struct buffer_head **s_group_desc;
1064         unsigned int s_mount_opt;
1065         unsigned int s_mount_flags;
1066         ext4_fsblk_t s_sb_block;
1067         uid_t s_resuid;
1068         gid_t s_resgid;
1069         unsigned short s_mount_state;
1070         unsigned short s_pad;
1071         int s_addr_per_block_bits;
1072         int s_desc_per_block_bits;
1073         int s_inode_size;
1074         int s_first_ino;
1075         unsigned int s_inode_readahead_blks;
1076         unsigned int s_inode_goal;
1077         spinlock_t s_next_gen_lock;
1078         u32 s_next_generation;
1079         u32 s_hash_seed[4];
1080         int s_def_hash_version;
1081         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1082         struct percpu_counter s_freeblocks_counter;
1083         struct percpu_counter s_freeinodes_counter;
1084         struct percpu_counter s_dirs_counter;
1085         struct percpu_counter s_dirtyblocks_counter;
1086         struct blockgroup_lock *s_blockgroup_lock;
1087         struct proc_dir_entry *s_proc;
1088         struct kobject s_kobj;
1089         struct completion s_kobj_unregister;
1090
1091         /* Journaling */
1092         struct inode *s_journal_inode;
1093         struct journal_s *s_journal;
1094         struct list_head s_orphan;
1095         struct mutex s_orphan_lock;
1096         struct mutex s_resize_lock;
1097         unsigned long s_commit_interval;
1098         u32 s_max_batch_time;
1099         u32 s_min_batch_time;
1100         struct block_device *journal_bdev;
1101 #ifdef CONFIG_JBD2_DEBUG
1102         struct timer_list turn_ro_timer;        /* For turning read-only (crash simulation) */
1103         wait_queue_head_t ro_wait_queue;        /* For people waiting for the fs to go read-only */
1104 #endif
1105 #ifdef CONFIG_QUOTA
1106         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
1107         int s_jquota_fmt;                       /* Format of quota to use */
1108 #endif
1109         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1110         struct rb_root system_blks;
1111
1112 #ifdef EXTENTS_STATS
1113         /* ext4 extents stats */
1114         unsigned long s_ext_min;
1115         unsigned long s_ext_max;
1116         unsigned long s_depth_max;
1117         spinlock_t s_ext_stats_lock;
1118         unsigned long s_ext_blocks;
1119         unsigned long s_ext_extents;
1120 #endif
1121
1122         /* for buddy allocator */
1123         struct ext4_group_info ***s_group_info;
1124         struct inode *s_buddy_cache;
1125         long s_blocks_reserved;
1126         spinlock_t s_reserve_lock;
1127         spinlock_t s_md_lock;
1128         tid_t s_last_transaction;
1129         unsigned short *s_mb_offsets;
1130         unsigned int *s_mb_maxs;
1131
1132         /* tunables */
1133         unsigned long s_stripe;
1134         unsigned int s_mb_stream_request;
1135         unsigned int s_mb_max_to_scan;
1136         unsigned int s_mb_min_to_scan;
1137         unsigned int s_mb_stats;
1138         unsigned int s_mb_order2_reqs;
1139         unsigned int s_mb_group_prealloc;
1140         unsigned int s_max_writeback_mb_bump;
1141         /* where last allocation was done - for stream allocation */
1142         unsigned long s_mb_last_group;
1143         unsigned long s_mb_last_start;
1144
1145         /* stats for buddy allocator */
1146         spinlock_t s_mb_pa_lock;
1147         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1148         atomic_t s_bal_success; /* we found long enough chunks */
1149         atomic_t s_bal_allocated;       /* in blocks */
1150         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1151         atomic_t s_bal_goals;   /* goal hits */
1152         atomic_t s_bal_breaks;  /* too long searches */
1153         atomic_t s_bal_2orders; /* 2^order hits */
1154         spinlock_t s_bal_lock;
1155         unsigned long s_mb_buddies_generated;
1156         unsigned long long s_mb_generation_time;
1157         atomic_t s_mb_lost_chunks;
1158         atomic_t s_mb_preallocated;
1159         atomic_t s_mb_discarded;
1160         atomic_t s_lock_busy;
1161
1162         /* locality groups */
1163         struct ext4_locality_group __percpu *s_locality_groups;
1164
1165         /* for write statistics */
1166         unsigned long s_sectors_written_start;
1167         u64 s_kbytes_written;
1168
1169         unsigned int s_log_groups_per_flex;
1170         struct flex_groups *s_flex_groups;
1171
1172         /* workqueue for dio unwritten */
1173         struct workqueue_struct *dio_unwritten_wq;
1174
1175         /* timer for periodic error stats printing */
1176         struct timer_list s_err_report;
1177
1178         /* Lazy inode table initialization info */
1179         struct ext4_li_request *s_li_request;
1180         /* Wait multiplier for lazy initialization thread */
1181         unsigned int s_li_wait_mult;
1182 };
1183
1184 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1185 {
1186         return sb->s_fs_info;
1187 }
1188 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1189 {
1190         return container_of(inode, struct ext4_inode_info, vfs_inode);
1191 }
1192
1193 static inline struct timespec ext4_current_time(struct inode *inode)
1194 {
1195         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1196                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1197 }
1198
1199 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1200 {
1201         return ino == EXT4_ROOT_INO ||
1202                 ino == EXT4_JOURNAL_INO ||
1203                 ino == EXT4_RESIZE_INO ||
1204                 (ino >= EXT4_FIRST_INO(sb) &&
1205                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1206 }
1207
1208 /*
1209  * Inode dynamic state flags
1210  */
1211 enum {
1212         EXT4_STATE_JDATA,               /* journaled data exists */
1213         EXT4_STATE_NEW,                 /* inode is newly created */
1214         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1215         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1216         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1217         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1218         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1219         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1220 };
1221
1222 #define EXT4_INODE_BIT_FNS(name, field)                                 \
1223 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1224 {                                                                       \
1225         return test_bit(bit, &EXT4_I(inode)->i_##field);                \
1226 }                                                                       \
1227 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1228 {                                                                       \
1229         set_bit(bit, &EXT4_I(inode)->i_##field);                        \
1230 }                                                                       \
1231 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1232 {                                                                       \
1233         clear_bit(bit, &EXT4_I(inode)->i_##field);                      \
1234 }
1235
1236 EXT4_INODE_BIT_FNS(flag, flags)
1237 EXT4_INODE_BIT_FNS(state, state_flags)
1238 #else
1239 /* Assume that user mode programs are passing in an ext4fs superblock, not
1240  * a kernel struct super_block.  This will allow us to call the feature-test
1241  * macros from user land. */
1242 #define EXT4_SB(sb)     (sb)
1243 #endif
1244
1245 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1246
1247 /*
1248  * Codes for operating systems
1249  */
1250 #define EXT4_OS_LINUX           0
1251 #define EXT4_OS_HURD            1
1252 #define EXT4_OS_MASIX           2
1253 #define EXT4_OS_FREEBSD         3
1254 #define EXT4_OS_LITES           4
1255
1256 /*
1257  * Revision levels
1258  */
1259 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1260 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1261
1262 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1263 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1264
1265 #define EXT4_GOOD_OLD_INODE_SIZE 128
1266
1267 /*
1268  * Feature set definitions
1269  */
1270
1271 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1272         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1273 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1274         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1275 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1276         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1277 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1278         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1279 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1280         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1281 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1282         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1283 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1284         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1285 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1286         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1287 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1288         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1289
1290 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1291 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1292 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1293 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1294 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1295 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1296
1297 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1298 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1299 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1300 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1301 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1302 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1303 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1304
1305 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1306 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1307 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1308 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1309 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1310 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1311 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1312 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1313 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1314 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1315 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1316
1317 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1318 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1319                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1320                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1321                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1322                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1323                                          EXT4_FEATURE_INCOMPAT_FLEX_BG)
1324 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1325                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1326                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1327                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1328                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1329                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1330                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE)
1331
1332 /*
1333  * Default values for user and/or group using reserved blocks
1334  */
1335 #define EXT4_DEF_RESUID         0
1336 #define EXT4_DEF_RESGID         0
1337
1338 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1339
1340 /*
1341  * Default mount options
1342  */
1343 #define EXT4_DEFM_DEBUG         0x0001
1344 #define EXT4_DEFM_BSDGROUPS     0x0002
1345 #define EXT4_DEFM_XATTR_USER    0x0004
1346 #define EXT4_DEFM_ACL           0x0008
1347 #define EXT4_DEFM_UID16         0x0010
1348 #define EXT4_DEFM_JMODE         0x0060
1349 #define EXT4_DEFM_JMODE_DATA    0x0020
1350 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1351 #define EXT4_DEFM_JMODE_WBACK   0x0060
1352 #define EXT4_DEFM_NOBARRIER     0x0100
1353 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1354 #define EXT4_DEFM_DISCARD       0x0400
1355 #define EXT4_DEFM_NODELALLOC    0x0800
1356
1357 /*
1358  * Default journal batch times
1359  */
1360 #define EXT4_DEF_MIN_BATCH_TIME 0
1361 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1362
1363 /*
1364  * Minimum number of groups in a flexgroup before we separate out
1365  * directories into the first block group of a flexgroup
1366  */
1367 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1368
1369 /*
1370  * Structure of a directory entry
1371  */
1372 #define EXT4_NAME_LEN 255
1373
1374 struct ext4_dir_entry {
1375         __le32  inode;                  /* Inode number */
1376         __le16  rec_len;                /* Directory entry length */
1377         __le16  name_len;               /* Name length */
1378         char    name[EXT4_NAME_LEN];    /* File name */
1379 };
1380
1381 /*
1382  * The new version of the directory entry.  Since EXT4 structures are
1383  * stored in intel byte order, and the name_len field could never be
1384  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1385  * file_type field.
1386  */
1387 struct ext4_dir_entry_2 {
1388         __le32  inode;                  /* Inode number */
1389         __le16  rec_len;                /* Directory entry length */
1390         __u8    name_len;               /* Name length */
1391         __u8    file_type;
1392         char    name[EXT4_NAME_LEN];    /* File name */
1393 };
1394
1395 /*
1396  * Ext4 directory file types.  Only the low 3 bits are used.  The
1397  * other bits are reserved for now.
1398  */
1399 #define EXT4_FT_UNKNOWN         0
1400 #define EXT4_FT_REG_FILE        1
1401 #define EXT4_FT_DIR             2
1402 #define EXT4_FT_CHRDEV          3
1403 #define EXT4_FT_BLKDEV          4
1404 #define EXT4_FT_FIFO            5
1405 #define EXT4_FT_SOCK            6
1406 #define EXT4_FT_SYMLINK         7
1407
1408 #define EXT4_FT_MAX             8
1409
1410 /*
1411  * EXT4_DIR_PAD defines the directory entries boundaries
1412  *
1413  * NOTE: It must be a multiple of 4
1414  */
1415 #define EXT4_DIR_PAD                    4
1416 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1417 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1418                                          ~EXT4_DIR_ROUND)
1419 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1420
1421 /*
1422  * If we ever get support for fs block sizes > page_size, we'll need
1423  * to remove the #if statements in the next two functions...
1424  */
1425 static inline unsigned int
1426 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1427 {
1428         unsigned len = le16_to_cpu(dlen);
1429
1430 #if (PAGE_CACHE_SIZE >= 65536)
1431         if (len == EXT4_MAX_REC_LEN || len == 0)
1432                 return blocksize;
1433         return (len & 65532) | ((len & 3) << 16);
1434 #else
1435         return len;
1436 #endif
1437 }
1438
1439 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1440 {
1441         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1442                 BUG();
1443 #if (PAGE_CACHE_SIZE >= 65536)
1444         if (len < 65536)
1445                 return cpu_to_le16(len);
1446         if (len == blocksize) {
1447                 if (blocksize == 65536)
1448                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1449                 else
1450                         return cpu_to_le16(0);
1451         }
1452         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1453 #else
1454         return cpu_to_le16(len);
1455 #endif
1456 }
1457
1458 /*
1459  * Hash Tree Directory indexing
1460  * (c) Daniel Phillips, 2001
1461  */
1462
1463 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1464                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1465                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1466 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1467 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1468
1469 /* Legal values for the dx_root hash_version field: */
1470
1471 #define DX_HASH_LEGACY          0
1472 #define DX_HASH_HALF_MD4        1
1473 #define DX_HASH_TEA             2
1474 #define DX_HASH_LEGACY_UNSIGNED 3
1475 #define DX_HASH_HALF_MD4_UNSIGNED       4
1476 #define DX_HASH_TEA_UNSIGNED            5
1477
1478 #ifdef __KERNEL__
1479
1480 /* hash info structure used by the directory hash */
1481 struct dx_hash_info
1482 {
1483         u32             hash;
1484         u32             minor_hash;
1485         int             hash_version;
1486         u32             *seed;
1487 };
1488
1489 #define EXT4_HTREE_EOF  0x7fffffff
1490
1491 /*
1492  * Control parameters used by ext4_htree_next_block
1493  */
1494 #define HASH_NB_ALWAYS          1
1495
1496
1497 /*
1498  * Describe an inode's exact location on disk and in memory
1499  */
1500 struct ext4_iloc
1501 {
1502         struct buffer_head *bh;
1503         unsigned long offset;
1504         ext4_group_t block_group;
1505 };
1506
1507 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1508 {
1509         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1510 }
1511
1512 /*
1513  * This structure is stuffed into the struct file's private_data field
1514  * for directories.  It is where we put information so that we can do
1515  * readdir operations in hash tree order.
1516  */
1517 struct dir_private_info {
1518         struct rb_root  root;
1519         struct rb_node  *curr_node;
1520         struct fname    *extra_fname;
1521         loff_t          last_pos;
1522         __u32           curr_hash;
1523         __u32           curr_minor_hash;
1524         __u32           next_hash;
1525 };
1526
1527 /* calculate the first block number of the group */
1528 static inline ext4_fsblk_t
1529 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1530 {
1531         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1532                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1533 }
1534
1535 /*
1536  * Special error return code only used by dx_probe() and its callers.
1537  */
1538 #define ERR_BAD_DX_DIR  -75000
1539
1540 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1541                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1542
1543 extern struct proc_dir_entry *ext4_proc_root;
1544
1545 /*
1546  * Timeout and state flag for lazy initialization inode thread.
1547  */
1548 #define EXT4_DEF_LI_WAIT_MULT                   10
1549 #define EXT4_DEF_LI_MAX_START_DELAY             5
1550 #define EXT4_LAZYINIT_QUIT                      0x0001
1551 #define EXT4_LAZYINIT_RUNNING                   0x0002
1552
1553 /*
1554  * Lazy inode table initialization info
1555  */
1556 struct ext4_lazy_init {
1557         unsigned long           li_state;
1558
1559         wait_queue_head_t       li_wait_daemon;
1560         wait_queue_head_t       li_wait_task;
1561         struct timer_list       li_timer;
1562         struct task_struct      *li_task;
1563
1564         struct list_head        li_request_list;
1565         struct mutex            li_list_mtx;
1566 };
1567
1568 struct ext4_li_request {
1569         struct super_block      *lr_super;
1570         struct ext4_sb_info     *lr_sbi;
1571         ext4_group_t            lr_next_group;
1572         struct list_head        lr_request;
1573         unsigned long           lr_next_sched;
1574         unsigned long           lr_timeout;
1575 };
1576
1577 struct ext4_features {
1578         struct kobject f_kobj;
1579         struct completion f_kobj_unregister;
1580 };
1581
1582 /*
1583  * Function prototypes
1584  */
1585
1586 /*
1587  * Ok, these declarations are also in <linux/kernel.h> but none of the
1588  * ext4 source programs needs to include it so they are duplicated here.
1589  */
1590 # define NORET_TYPE     /**/
1591 # define ATTRIB_NORET   __attribute__((noreturn))
1592 # define NORET_AND      noreturn,
1593
1594 /* bitmap.c */
1595 extern unsigned int ext4_count_free(struct buffer_head *, unsigned);
1596
1597 /* balloc.c */
1598 extern unsigned int ext4_block_group(struct super_block *sb,
1599                         ext4_fsblk_t blocknr);
1600 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1601                         ext4_fsblk_t blocknr);
1602 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1603 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1604                         ext4_group_t group);
1605 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1606                         ext4_fsblk_t goal, unsigned long *count, int *errp);
1607 extern int ext4_claim_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1608 extern int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1609 extern void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
1610                                 ext4_fsblk_t block, unsigned long count);
1611 extern ext4_fsblk_t ext4_count_free_blocks(struct super_block *);
1612 extern void ext4_check_blocks_bitmap(struct super_block *);
1613 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1614                                                     ext4_group_t block_group,
1615                                                     struct buffer_head ** bh);
1616 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1617 struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1618                                       ext4_group_t block_group);
1619 extern unsigned ext4_init_block_bitmap(struct super_block *sb,
1620                                        struct buffer_head *bh,
1621                                        ext4_group_t group,
1622                                        struct ext4_group_desc *desc);
1623 #define ext4_free_blocks_after_init(sb, group, desc)                    \
1624                 ext4_init_block_bitmap(sb, NULL, group, desc)
1625
1626 /* dir.c */
1627 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1628                                   struct ext4_dir_entry_2 *,
1629                                   struct buffer_head *, unsigned int);
1630 #define ext4_check_dir_entry(dir, de, bh, offset) \
1631         __ext4_check_dir_entry(__func__, __LINE__, (dir), (de), (bh), (offset))
1632 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1633                                     __u32 minor_hash,
1634                                     struct ext4_dir_entry_2 *dirent);
1635 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1636
1637 /* fsync.c */
1638 extern int ext4_sync_file(struct file *, int);
1639
1640 /* hash.c */
1641 extern int ext4fs_dirhash(const char *name, int len, struct
1642                           dx_hash_info *hinfo);
1643
1644 /* ialloc.c */
1645 extern struct inode *ext4_new_inode(handle_t *, struct inode *, int,
1646                                     const struct qstr *qstr, __u32 goal);
1647 extern void ext4_free_inode(handle_t *, struct inode *);
1648 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1649 extern unsigned long ext4_count_free_inodes(struct super_block *);
1650 extern unsigned long ext4_count_dirs(struct super_block *);
1651 extern void ext4_check_inodes_bitmap(struct super_block *);
1652 extern unsigned ext4_init_inode_bitmap(struct super_block *sb,
1653                                        struct buffer_head *bh,
1654                                        ext4_group_t group,
1655                                        struct ext4_group_desc *desc);
1656 extern void mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1657 extern int ext4_init_inode_table(struct super_block *sb,
1658                                  ext4_group_t group, int barrier);
1659
1660 /* mballoc.c */
1661 extern long ext4_mb_stats;
1662 extern long ext4_mb_max_to_scan;
1663 extern int ext4_mb_init(struct super_block *, int);
1664 extern int ext4_mb_release(struct super_block *);
1665 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1666                                 struct ext4_allocation_request *, int *);
1667 extern int ext4_mb_reserve_blocks(struct super_block *, int);
1668 extern void ext4_discard_preallocations(struct inode *);
1669 extern int __init init_ext4_mballoc(void);
1670 extern void exit_ext4_mballoc(void);
1671 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1672                              struct buffer_head *bh, ext4_fsblk_t block,
1673                              unsigned long count, int flags);
1674 extern int ext4_mb_add_groupinfo(struct super_block *sb,
1675                 ext4_group_t i, struct ext4_group_desc *desc);
1676 extern int ext4_mb_get_buddy_cache_lock(struct super_block *, ext4_group_t);
1677 extern void ext4_mb_put_buddy_cache_lock(struct super_block *,
1678                                                 ext4_group_t, int);
1679 /* inode.c */
1680 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1681                                                 ext4_lblk_t, int, int *);
1682 struct buffer_head *ext4_bread(handle_t *, struct inode *,
1683                                                 ext4_lblk_t, int, int *);
1684 int ext4_get_block(struct inode *inode, sector_t iblock,
1685                                 struct buffer_head *bh_result, int create);
1686
1687 extern struct inode *ext4_iget(struct super_block *, unsigned long);
1688 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
1689 extern int  ext4_setattr(struct dentry *, struct iattr *);
1690 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1691                                 struct kstat *stat);
1692 extern void ext4_evict_inode(struct inode *);
1693 extern void ext4_clear_inode(struct inode *);
1694 extern int  ext4_sync_inode(handle_t *, struct inode *);
1695 extern void ext4_dirty_inode(struct inode *);
1696 extern int ext4_change_inode_journal_flag(struct inode *, int);
1697 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1698 extern int ext4_can_truncate(struct inode *inode);
1699 extern void ext4_truncate(struct inode *);
1700 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
1701 extern void ext4_set_inode_flags(struct inode *);
1702 extern void ext4_get_inode_flags(struct ext4_inode_info *);
1703 extern int ext4_alloc_da_blocks(struct inode *inode);
1704 extern void ext4_set_aops(struct inode *inode);
1705 extern int ext4_writepage_trans_blocks(struct inode *);
1706 extern int ext4_meta_trans_blocks(struct inode *, int nrblocks, int idxblocks);
1707 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1708 extern int ext4_block_truncate_page(handle_t *handle,
1709                 struct address_space *mapping, loff_t from);
1710 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1711 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
1712 extern int flush_completed_IO(struct inode *inode);
1713 extern void ext4_da_update_reserve_space(struct inode *inode,
1714                                         int used, int quota_claim);
1715 /* ioctl.c */
1716 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1717 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1718
1719 /* migrate.c */
1720 extern int ext4_ext_migrate(struct inode *);
1721
1722 /* namei.c */
1723 extern int ext4_orphan_add(handle_t *, struct inode *);
1724 extern int ext4_orphan_del(handle_t *, struct inode *);
1725 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1726                                 __u32 start_minor_hash, __u32 *next_hash);
1727
1728 /* resize.c */
1729 extern int ext4_group_add(struct super_block *sb,
1730                                 struct ext4_new_group_data *input);
1731 extern int ext4_group_extend(struct super_block *sb,
1732                                 struct ext4_super_block *es,
1733                                 ext4_fsblk_t n_blocks_count);
1734
1735 /* super.c */
1736 extern void __ext4_error(struct super_block *, const char *, unsigned int,
1737                          const char *, ...)
1738         __attribute__ ((format (printf, 4, 5)));
1739 #define ext4_error(sb, message...)      __ext4_error(sb, __func__,      \
1740                                                      __LINE__, ## message)
1741 extern void ext4_error_inode(struct inode *, const char *, unsigned int,
1742                              ext4_fsblk_t, const char *, ...)
1743         __attribute__ ((format (printf, 5, 6)));
1744 extern void ext4_error_file(struct file *, const char *, unsigned int,
1745                             const char *, ...)
1746         __attribute__ ((format (printf, 4, 5)));
1747 extern void __ext4_std_error(struct super_block *, const char *,
1748                              unsigned int, int);
1749 extern void __ext4_abort(struct super_block *, const char *, unsigned int,
1750                        const char *, ...)
1751         __attribute__ ((format (printf, 4, 5)));
1752 #define ext4_abort(sb, message...)      __ext4_abort(sb, __func__, \
1753                                                        __LINE__, ## message)
1754 extern void __ext4_warning(struct super_block *, const char *, unsigned int,
1755                           const char *, ...)
1756         __attribute__ ((format (printf, 4, 5)));
1757 #define ext4_warning(sb, message...)    __ext4_warning(sb, __func__, \
1758                                                        __LINE__, ## message)
1759 extern void ext4_msg(struct super_block *, const char *, const char *, ...)
1760         __attribute__ ((format (printf, 3, 4)));
1761 extern void __ext4_grp_locked_error(const char *, unsigned int, \
1762                                     struct super_block *, ext4_group_t, \
1763                                     unsigned long, ext4_fsblk_t, \
1764                                     const char *, ...)
1765         __attribute__ ((format (printf, 7, 8)));
1766 #define ext4_grp_locked_error(sb, grp, message...) \
1767         __ext4_grp_locked_error(__func__, __LINE__, (sb), (grp), ## message)
1768 extern void ext4_update_dynamic_rev(struct super_block *sb);
1769 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1770                                         __u32 compat);
1771 extern int ext4_update_rocompat_feature(handle_t *handle,
1772                                         struct super_block *sb, __u32 rocompat);
1773 extern int ext4_update_incompat_feature(handle_t *handle,
1774                                         struct super_block *sb, __u32 incompat);
1775 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1776                                       struct ext4_group_desc *bg);
1777 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1778                                       struct ext4_group_desc *bg);
1779 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1780                                      struct ext4_group_desc *bg);
1781 extern __u32 ext4_free_blks_count(struct super_block *sb,
1782                                 struct ext4_group_desc *bg);
1783 extern __u32 ext4_free_inodes_count(struct super_block *sb,
1784                                  struct ext4_group_desc *bg);
1785 extern __u32 ext4_used_dirs_count(struct super_block *sb,
1786                                 struct ext4_group_desc *bg);
1787 extern __u32 ext4_itable_unused_count(struct super_block *sb,
1788                                    struct ext4_group_desc *bg);
1789 extern void ext4_block_bitmap_set(struct super_block *sb,
1790                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1791 extern void ext4_inode_bitmap_set(struct super_block *sb,
1792                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1793 extern void ext4_inode_table_set(struct super_block *sb,
1794                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
1795 extern void ext4_free_blks_set(struct super_block *sb,
1796                                struct ext4_group_desc *bg, __u32 count);
1797 extern void ext4_free_inodes_set(struct super_block *sb,
1798                                 struct ext4_group_desc *bg, __u32 count);
1799 extern void ext4_used_dirs_set(struct super_block *sb,
1800                                 struct ext4_group_desc *bg, __u32 count);
1801 extern void ext4_itable_unused_set(struct super_block *sb,
1802                                    struct ext4_group_desc *bg, __u32 count);
1803 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
1804                                    struct ext4_group_desc *gdp);
1805 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
1806                                        struct ext4_group_desc *gdp);
1807
1808 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
1809 {
1810         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
1811                 le32_to_cpu(es->s_blocks_count_lo);
1812 }
1813
1814 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
1815 {
1816         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
1817                 le32_to_cpu(es->s_r_blocks_count_lo);
1818 }
1819
1820 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
1821 {
1822         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
1823                 le32_to_cpu(es->s_free_blocks_count_lo);
1824 }
1825
1826 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
1827                                          ext4_fsblk_t blk)
1828 {
1829         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
1830         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
1831 }
1832
1833 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
1834                                               ext4_fsblk_t blk)
1835 {
1836         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
1837         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
1838 }
1839
1840 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
1841                                            ext4_fsblk_t blk)
1842 {
1843         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
1844         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
1845 }
1846
1847 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
1848 {
1849         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
1850                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
1851                         le32_to_cpu(raw_inode->i_size_lo);
1852         else
1853                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
1854 }
1855
1856 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
1857 {
1858         raw_inode->i_size_lo = cpu_to_le32(i_size);
1859         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
1860 }
1861
1862 static inline
1863 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
1864                                             ext4_group_t group)
1865 {
1866          struct ext4_group_info ***grp_info;
1867          long indexv, indexh;
1868          grp_info = EXT4_SB(sb)->s_group_info;
1869          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
1870          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
1871          return grp_info[indexv][indexh];
1872 }
1873
1874 /*
1875  * Reading s_groups_count requires using smp_rmb() afterwards.  See
1876  * the locking protocol documented in the comments of ext4_group_add()
1877  * in resize.c
1878  */
1879 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
1880 {
1881         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
1882
1883         smp_rmb();
1884         return ngroups;
1885 }
1886
1887 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
1888                                              ext4_group_t block_group)
1889 {
1890         return block_group >> sbi->s_log_groups_per_flex;
1891 }
1892
1893 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
1894 {
1895         return 1 << sbi->s_log_groups_per_flex;
1896 }
1897
1898 #define ext4_std_error(sb, errno)                               \
1899 do {                                                            \
1900         if ((errno))                                            \
1901                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
1902 } while (0)
1903
1904 #ifdef CONFIG_SMP
1905 /* Each CPU can accumulate percpu_counter_batch blocks in their local
1906  * counters. So we need to make sure we have free blocks more
1907  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
1908  */
1909 #define EXT4_FREEBLOCKS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
1910 #else
1911 #define EXT4_FREEBLOCKS_WATERMARK 0
1912 #endif
1913
1914 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
1915 {
1916         /*
1917          * XXX: replace with spinlock if seen contended -bzzz
1918          */
1919         down_write(&EXT4_I(inode)->i_data_sem);
1920         if (newsize > EXT4_I(inode)->i_disksize)
1921                 EXT4_I(inode)->i_disksize = newsize;
1922         up_write(&EXT4_I(inode)->i_data_sem);
1923         return ;
1924 }
1925
1926 struct ext4_group_info {
1927         unsigned long   bb_state;
1928         struct rb_root  bb_free_root;
1929         ext4_grpblk_t   bb_first_free;  /* first free block */
1930         ext4_grpblk_t   bb_free;        /* total free blocks */
1931         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
1932         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
1933         struct          list_head bb_prealloc_list;
1934 #ifdef DOUBLE_CHECK
1935         void            *bb_bitmap;
1936 #endif
1937         struct rw_semaphore alloc_sem;
1938         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
1939                                          * regions, index is order.
1940                                          * bb_counters[3] = 5 means
1941                                          * 5 free 8-block regions. */
1942 };
1943
1944 #define EXT4_GROUP_INFO_NEED_INIT_BIT   0
1945
1946 #define EXT4_MB_GRP_NEED_INIT(grp)      \
1947         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
1948
1949 #define EXT4_MAX_CONTENTION             8
1950 #define EXT4_CONTENTION_THRESHOLD       2
1951
1952 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
1953                                               ext4_group_t group)
1954 {
1955         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
1956 }
1957
1958 /*
1959  * Returns true if the filesystem is busy enough that attempts to
1960  * access the block group locks has run into contention.
1961  */
1962 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
1963 {
1964         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
1965 }
1966
1967 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
1968 {
1969         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
1970         if (spin_trylock(lock))
1971                 /*
1972                  * We're able to grab the lock right away, so drop the
1973                  * lock contention counter.
1974                  */
1975                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
1976         else {
1977                 /*
1978                  * The lock is busy, so bump the contention counter,
1979                  * and then wait on the spin lock.
1980                  */
1981                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
1982                                   EXT4_MAX_CONTENTION);
1983                 spin_lock(lock);
1984         }
1985 }
1986
1987 static inline void ext4_unlock_group(struct super_block *sb,
1988                                         ext4_group_t group)
1989 {
1990         spin_unlock(ext4_group_lock_ptr(sb, group));
1991 }
1992
1993 static inline void ext4_mark_super_dirty(struct super_block *sb)
1994 {
1995         if (EXT4_SB(sb)->s_journal == NULL)
1996                 sb->s_dirt =1;
1997 }
1998
1999 /*
2000  * Inodes and files operations
2001  */
2002
2003 /* dir.c */
2004 extern const struct file_operations ext4_dir_operations;
2005
2006 /* file.c */
2007 extern const struct inode_operations ext4_file_inode_operations;
2008 extern const struct file_operations ext4_file_operations;
2009 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2010
2011 /* namei.c */
2012 extern const struct inode_operations ext4_dir_inode_operations;
2013 extern const struct inode_operations ext4_special_inode_operations;
2014 extern struct dentry *ext4_get_parent(struct dentry *child);
2015
2016 /* symlink.c */
2017 extern const struct inode_operations ext4_symlink_inode_operations;
2018 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2019
2020 /* block_validity */
2021 extern void ext4_release_system_zone(struct super_block *sb);
2022 extern int ext4_setup_system_zone(struct super_block *sb);
2023 extern int __init init_ext4_system_zone(void);
2024 extern void exit_ext4_system_zone(void);
2025 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2026                                  ext4_fsblk_t start_blk,
2027                                  unsigned int count);
2028
2029 /* extents.c */
2030 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2031 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2032 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
2033                                        int chunk);
2034 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2035                                struct ext4_map_blocks *map, int flags);
2036 extern void ext4_ext_truncate(struct inode *);
2037 extern void ext4_ext_init(struct super_block *);
2038 extern void ext4_ext_release(struct super_block *);
2039 extern long ext4_fallocate(struct inode *inode, int mode, loff_t offset,
2040                           loff_t len);
2041 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
2042                           ssize_t len);
2043 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2044                            struct ext4_map_blocks *map, int flags);
2045 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2046                         __u64 start, __u64 len);
2047 /* move_extent.c */
2048 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2049                              __u64 start_orig, __u64 start_donor,
2050                              __u64 len, __u64 *moved_len);
2051
2052
2053 /* BH_Uninit flag: blocks are allocated but uninitialized on disk */
2054 enum ext4_state_bits {
2055         BH_Uninit       /* blocks are allocated but uninitialized on disk */
2056           = BH_JBDPrivateStart,
2057 };
2058
2059 BUFFER_FNS(Uninit, uninit)
2060 TAS_BUFFER_FNS(Uninit, uninit)
2061
2062 /*
2063  * Add new method to test wether block and inode bitmaps are properly
2064  * initialized. With uninit_bg reading the block from disk is not enough
2065  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2066  */
2067 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2068
2069 static inline int bitmap_uptodate(struct buffer_head *bh)
2070 {
2071         return (buffer_uptodate(bh) &&
2072                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2073 }
2074 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2075 {
2076         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2077 }
2078
2079 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2080
2081 #endif  /* __KERNEL__ */
2082
2083 #endif  /* _EXT4_H */