Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
[sfrench/cifs-2.6.git] / fs / ext4 / ext4.h
1 /*
2  *  ext4.h
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/include/linux/minix_fs.h
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  */
15
16 #ifndef _EXT4_H
17 #define _EXT4_H
18
19 #include <linux/types.h>
20 #include <linux/blkdev.h>
21 #include <linux/magic.h>
22 #include <linux/jbd2.h>
23 #include <linux/quota.h>
24 #include <linux/rwsem.h>
25 #include <linux/rbtree.h>
26 #include <linux/seqlock.h>
27 #include <linux/mutex.h>
28 #include <linux/timer.h>
29 #include <linux/wait.h>
30 #include <linux/blockgroup_lock.h>
31 #include <linux/percpu_counter.h>
32 #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(__func__, (inode), (fmt), ## a)
61
62 #define EXT4_ERROR_FILE(file, fmt, a...)        \
63         ext4_error_file(__func__, (file), (fmt), ## a)
64
65 /* data type for block offset of block group */
66 typedef int ext4_grpblk_t;
67
68 /* data type for filesystem-wide blocks number */
69 typedef unsigned long long ext4_fsblk_t;
70
71 /* data type for file logical block number */
72 typedef __u32 ext4_lblk_t;
73
74 /* data type for block group number */
75 typedef unsigned int ext4_group_t;
76
77 /*
78  * Flags used in mballoc's allocation_context flags field.
79  *
80  * Also used to show what's going on for debugging purposes when the
81  * flag field is exported via the traceport interface
82  */
83
84 /* prefer goal again. length */
85 #define EXT4_MB_HINT_MERGE              0x0001
86 /* blocks already reserved */
87 #define EXT4_MB_HINT_RESERVED           0x0002
88 /* metadata is being allocated */
89 #define EXT4_MB_HINT_METADATA           0x0004
90 /* first blocks in the file */
91 #define EXT4_MB_HINT_FIRST              0x0008
92 /* search for the best chunk */
93 #define EXT4_MB_HINT_BEST               0x0010
94 /* data is being allocated */
95 #define EXT4_MB_HINT_DATA               0x0020
96 /* don't preallocate (for tails) */
97 #define EXT4_MB_HINT_NOPREALLOC         0x0040
98 /* allocate for locality group */
99 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
100 /* allocate goal blocks or none */
101 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
102 /* goal is meaningful */
103 #define EXT4_MB_HINT_TRY_GOAL           0x0200
104 /* blocks already pre-reserved by delayed allocation */
105 #define EXT4_MB_DELALLOC_RESERVED       0x0400
106 /* We are doing stream allocation */
107 #define EXT4_MB_STREAM_ALLOC            0x0800
108
109
110 struct ext4_allocation_request {
111         /* target inode for block we're allocating */
112         struct inode *inode;
113         /* how many blocks we want to allocate */
114         unsigned int len;
115         /* logical block in target inode */
116         ext4_lblk_t logical;
117         /* the closest logical allocated block to the left */
118         ext4_lblk_t lleft;
119         /* the closest logical allocated block to the right */
120         ext4_lblk_t lright;
121         /* phys. target (a hint) */
122         ext4_fsblk_t goal;
123         /* phys. block for the closest logical allocated block to the left */
124         ext4_fsblk_t pleft;
125         /* phys. block for the closest logical allocated block to the right */
126         ext4_fsblk_t pright;
127         /* flags. see above EXT4_MB_HINT_* */
128         unsigned int flags;
129 };
130
131 /*
132  * Logical to physical block mapping, used by ext4_map_blocks()
133  *
134  * This structure is used to pass requests into ext4_map_blocks() as
135  * well as to store the information returned by ext4_map_blocks().  It
136  * takes less room on the stack than a struct buffer_head.
137  */
138 #define EXT4_MAP_NEW            (1 << BH_New)
139 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
140 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
141 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
142 #define EXT4_MAP_UNINIT         (1 << BH_Uninit)
143 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
144                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
145                                  EXT4_MAP_UNINIT)
146
147 struct ext4_map_blocks {
148         ext4_fsblk_t m_pblk;
149         ext4_lblk_t m_lblk;
150         unsigned int m_len;
151         unsigned int m_flags;
152 };
153
154 /*
155  * For delayed allocation tracking
156  */
157 struct mpage_da_data {
158         struct inode *inode;
159         sector_t b_blocknr;             /* start block number of extent */
160         size_t b_size;                  /* size of extent */
161         unsigned long b_state;          /* state of the extent */
162         unsigned long first_page, next_page;    /* extent of pages */
163         struct writeback_control *wbc;
164         int io_done;
165         int pages_written;
166         int retval;
167 };
168 #define EXT4_IO_UNWRITTEN       0x1
169 typedef struct ext4_io_end {
170         struct list_head        list;           /* per-file finished AIO list */
171         struct inode            *inode;         /* file being written to */
172         unsigned int            flag;           /* unwritten or not */
173         struct page             *page;          /* page struct for buffer write */
174         loff_t                  offset;         /* offset in the file */
175         ssize_t                 size;           /* size of the extent */
176         struct work_struct      work;           /* data work queue */
177 } ext4_io_end_t;
178
179 /*
180  * Special inodes numbers
181  */
182 #define EXT4_BAD_INO             1      /* Bad blocks inode */
183 #define EXT4_ROOT_INO            2      /* Root inode */
184 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
185 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
186 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
187 #define EXT4_JOURNAL_INO         8      /* Journal inode */
188
189 /* First non-reserved inode for old ext4 filesystems */
190 #define EXT4_GOOD_OLD_FIRST_INO 11
191
192 /*
193  * Maximal count of links to a file
194  */
195 #define EXT4_LINK_MAX           65000
196
197 /*
198  * Macro-instructions used to manage several block sizes
199  */
200 #define EXT4_MIN_BLOCK_SIZE             1024
201 #define EXT4_MAX_BLOCK_SIZE             65536
202 #define EXT4_MIN_BLOCK_LOG_SIZE         10
203 #ifdef __KERNEL__
204 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
205 #else
206 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
207 #endif
208 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
209 #ifdef __KERNEL__
210 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
211 #else
212 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
213 #endif
214 #ifdef __KERNEL__
215 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
216 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
217 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
218 #else
219 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
220                                  EXT4_GOOD_OLD_INODE_SIZE : \
221                                  (s)->s_inode_size)
222 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
223                                  EXT4_GOOD_OLD_FIRST_INO : \
224                                  (s)->s_first_ino)
225 #endif
226 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
227
228 /*
229  * Structure of a blocks group descriptor
230  */
231 struct ext4_group_desc
232 {
233         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
234         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
235         __le32  bg_inode_table_lo;      /* Inodes table block */
236         __le16  bg_free_blocks_count_lo;/* Free blocks count */
237         __le16  bg_free_inodes_count_lo;/* Free inodes count */
238         __le16  bg_used_dirs_count_lo;  /* Directories count */
239         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
240         __u32   bg_reserved[2];         /* Likely block/inode bitmap checksum */
241         __le16  bg_itable_unused_lo;    /* Unused inodes count */
242         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
243         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
244         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
245         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
246         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
247         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
248         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
249         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
250         __u32   bg_reserved2[3];
251 };
252
253 /*
254  * Structure of a flex block group info
255  */
256
257 struct flex_groups {
258         atomic_t free_inodes;
259         atomic_t free_blocks;
260         atomic_t used_dirs;
261 };
262
263 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
264 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
265 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
266
267 /*
268  * Macro-instructions used to manage group descriptors
269  */
270 #define EXT4_MIN_DESC_SIZE              32
271 #define EXT4_MIN_DESC_SIZE_64BIT        64
272 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
273 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
274 #ifdef __KERNEL__
275 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
276 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
277 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
278 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
279 #else
280 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
281 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
282 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
283 #endif
284
285 /*
286  * Constants relative to the data blocks
287  */
288 #define EXT4_NDIR_BLOCKS                12
289 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
290 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
291 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
292 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
293
294 /*
295  * Inode flags
296  */
297 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
298 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
299 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
300 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
301 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
302 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
303 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
304 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
305 /* Reserved for compression usage... */
306 #define EXT4_DIRTY_FL                   0x00000100
307 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
308 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
309 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
310 /* End compression flags --- maybe not all used */
311 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
312 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
313 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
314 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
315 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
316 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
317 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
318 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
319 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
320 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
321 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
322
323 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
324 #define EXT4_FL_USER_MODIFIABLE         0x004B80FF /* User modifiable flags */
325
326 /* Flags that should be inherited by new inodes from their parent. */
327 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
328                            EXT4_SYNC_FL | EXT4_IMMUTABLE_FL | EXT4_APPEND_FL |\
329                            EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
330                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
331                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
332
333 /* Flags that are appropriate for regular files (all but dir-specific ones). */
334 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
335
336 /* Flags that are appropriate for non-directories/regular files. */
337 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
338
339 /* Mask out flags that are inappropriate for the given type of inode. */
340 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
341 {
342         if (S_ISDIR(mode))
343                 return flags;
344         else if (S_ISREG(mode))
345                 return flags & EXT4_REG_FLMASK;
346         else
347                 return flags & EXT4_OTHER_FLMASK;
348 }
349
350 /*
351  * Inode flags used for atomic set/get
352  */
353 enum {
354         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
355         EXT4_INODE_UNRM         = 1,    /* Undelete */
356         EXT4_INODE_COMPR        = 2,    /* Compress file */
357         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
358         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
359         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
360         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
361         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
362 /* Reserved for compression usage... */
363         EXT4_INODE_DIRTY        = 8,
364         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
365         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
366         EXT4_INODE_ECOMPR       = 11,   /* Compression error */
367 /* End compression flags --- maybe not all used */
368         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
369         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
370         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
371         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
372         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
373         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
374         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
375         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
376         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
377         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
378         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
379 };
380
381 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
382 #define CHECK_FLAG_VALUE(FLAG) if (!TEST_FLAG_VALUE(FLAG)) { \
383         printk(KERN_EMERG "EXT4 flag fail: " #FLAG ": %d %d\n", \
384                 EXT4_##FLAG##_FL, EXT4_INODE_##FLAG); BUG_ON(1); }
385
386 /*
387  * Since it's pretty easy to mix up bit numbers and hex values, and we
388  * can't do a compile-time test for ENUM values, we use a run-time
389  * test to make sure that EXT4_XXX_FL is consistent with respect to
390  * EXT4_INODE_XXX.  If all is well the printk and BUG_ON will all drop
391  * out so it won't cost any extra space in the compiled kernel image.
392  * But it's important that these values are the same, since we are
393  * using EXT4_INODE_XXX to test for the flag values, but EXT4_XX_FL
394  * must be consistent with the values of FS_XXX_FL defined in
395  * include/linux/fs.h and the on-disk values found in ext2, ext3, and
396  * ext4 filesystems, and of course the values defined in e2fsprogs.
397  *
398  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
399  */
400 static inline void ext4_check_flag_values(void)
401 {
402         CHECK_FLAG_VALUE(SECRM);
403         CHECK_FLAG_VALUE(UNRM);
404         CHECK_FLAG_VALUE(COMPR);
405         CHECK_FLAG_VALUE(SYNC);
406         CHECK_FLAG_VALUE(IMMUTABLE);
407         CHECK_FLAG_VALUE(APPEND);
408         CHECK_FLAG_VALUE(NODUMP);
409         CHECK_FLAG_VALUE(NOATIME);
410         CHECK_FLAG_VALUE(DIRTY);
411         CHECK_FLAG_VALUE(COMPRBLK);
412         CHECK_FLAG_VALUE(NOCOMPR);
413         CHECK_FLAG_VALUE(ECOMPR);
414         CHECK_FLAG_VALUE(INDEX);
415         CHECK_FLAG_VALUE(IMAGIC);
416         CHECK_FLAG_VALUE(JOURNAL_DATA);
417         CHECK_FLAG_VALUE(NOTAIL);
418         CHECK_FLAG_VALUE(DIRSYNC);
419         CHECK_FLAG_VALUE(TOPDIR);
420         CHECK_FLAG_VALUE(HUGE_FILE);
421         CHECK_FLAG_VALUE(EXTENTS);
422         CHECK_FLAG_VALUE(EA_INODE);
423         CHECK_FLAG_VALUE(EOFBLOCKS);
424         CHECK_FLAG_VALUE(RESERVED);
425 }
426
427 /* Used to pass group descriptor data when online resize is done */
428 struct ext4_new_group_input {
429         __u32 group;            /* Group number for this data */
430         __u64 block_bitmap;     /* Absolute block number of block bitmap */
431         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
432         __u64 inode_table;      /* Absolute block number of inode table start */
433         __u32 blocks_count;     /* Total number of blocks in this group */
434         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
435         __u16 unused;
436 };
437
438 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
439 struct compat_ext4_new_group_input {
440         u32 group;
441         compat_u64 block_bitmap;
442         compat_u64 inode_bitmap;
443         compat_u64 inode_table;
444         u32 blocks_count;
445         u16 reserved_blocks;
446         u16 unused;
447 };
448 #endif
449
450 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
451 struct ext4_new_group_data {
452         __u32 group;
453         __u64 block_bitmap;
454         __u64 inode_bitmap;
455         __u64 inode_table;
456         __u32 blocks_count;
457         __u16 reserved_blocks;
458         __u16 unused;
459         __u32 free_blocks_count;
460 };
461
462 /*
463  * Flags used by ext4_get_blocks()
464  */
465         /* Allocate any needed blocks and/or convert an unitialized
466            extent to be an initialized ext4 */
467 #define EXT4_GET_BLOCKS_CREATE                  0x0001
468         /* Request the creation of an unitialized extent */
469 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
470 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
471                                                  EXT4_GET_BLOCKS_CREATE)
472         /* Caller is from the delayed allocation writeout path,
473            so set the magic i_delalloc_reserve_flag after taking the
474            inode allocation semaphore for */
475 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
476         /* caller is from the direct IO path, request to creation of an
477         unitialized extents if not allocated, split the uninitialized
478         extent if blocks has been preallocated already*/
479 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
480 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
481 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
482                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
483         /* Convert extent to initialized after IO complete */
484 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
485                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
486
487 /*
488  * Flags used by ext4_free_blocks
489  */
490 #define EXT4_FREE_BLOCKS_METADATA       0x0001
491 #define EXT4_FREE_BLOCKS_FORGET         0x0002
492 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
493
494 /*
495  * ioctl commands
496  */
497 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
498 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
499 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
500 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
501 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
502 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
503 #ifdef CONFIG_JBD2_DEBUG
504 #define EXT4_IOC_WAIT_FOR_READONLY      _IOR('f', 99, long)
505 #endif
506 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
507 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
508 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
509 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
510 #define EXT4_IOC_MIGRATE                _IO('f', 9)
511  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
512  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
513 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
514 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
515
516 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
517 /*
518  * ioctl commands in 32 bit emulation
519  */
520 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
521 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
522 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
523 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
524 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
525 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
526 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
527 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
528 #ifdef CONFIG_JBD2_DEBUG
529 #define EXT4_IOC32_WAIT_FOR_READONLY    _IOR('f', 99, int)
530 #endif
531 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
532 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
533 #endif
534
535
536 /*
537  *  Mount options
538  */
539 struct ext4_mount_options {
540         unsigned long s_mount_opt;
541         uid_t s_resuid;
542         gid_t s_resgid;
543         unsigned long s_commit_interval;
544         u32 s_min_batch_time, s_max_batch_time;
545 #ifdef CONFIG_QUOTA
546         int s_jquota_fmt;
547         char *s_qf_names[MAXQUOTAS];
548 #endif
549 };
550
551 /* Max physical block we can addres w/o extents */
552 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
553
554 /*
555  * Structure of an inode on the disk
556  */
557 struct ext4_inode {
558         __le16  i_mode;         /* File mode */
559         __le16  i_uid;          /* Low 16 bits of Owner Uid */
560         __le32  i_size_lo;      /* Size in bytes */
561         __le32  i_atime;        /* Access time */
562         __le32  i_ctime;        /* Inode Change time */
563         __le32  i_mtime;        /* Modification time */
564         __le32  i_dtime;        /* Deletion Time */
565         __le16  i_gid;          /* Low 16 bits of Group Id */
566         __le16  i_links_count;  /* Links count */
567         __le32  i_blocks_lo;    /* Blocks count */
568         __le32  i_flags;        /* File flags */
569         union {
570                 struct {
571                         __le32  l_i_version;
572                 } linux1;
573                 struct {
574                         __u32  h_i_translator;
575                 } hurd1;
576                 struct {
577                         __u32  m_i_reserved1;
578                 } masix1;
579         } osd1;                         /* OS dependent 1 */
580         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
581         __le32  i_generation;   /* File version (for NFS) */
582         __le32  i_file_acl_lo;  /* File ACL */
583         __le32  i_size_high;
584         __le32  i_obso_faddr;   /* Obsoleted fragment address */
585         union {
586                 struct {
587                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
588                         __le16  l_i_file_acl_high;
589                         __le16  l_i_uid_high;   /* these 2 fields */
590                         __le16  l_i_gid_high;   /* were reserved2[0] */
591                         __u32   l_i_reserved2;
592                 } linux2;
593                 struct {
594                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
595                         __u16   h_i_mode_high;
596                         __u16   h_i_uid_high;
597                         __u16   h_i_gid_high;
598                         __u32   h_i_author;
599                 } hurd2;
600                 struct {
601                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
602                         __le16  m_i_file_acl_high;
603                         __u32   m_i_reserved2[2];
604                 } masix2;
605         } osd2;                         /* OS dependent 2 */
606         __le16  i_extra_isize;
607         __le16  i_pad1;
608         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
609         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
610         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
611         __le32  i_crtime;       /* File Creation time */
612         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
613         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
614 };
615
616 struct move_extent {
617         __u32 reserved;         /* should be zero */
618         __u32 donor_fd;         /* donor file descriptor */
619         __u64 orig_start;       /* logical start offset in block for orig */
620         __u64 donor_start;      /* logical start offset in block for donor */
621         __u64 len;              /* block length to be moved */
622         __u64 moved_len;        /* moved block length */
623 };
624
625 #define EXT4_EPOCH_BITS 2
626 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
627 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
628
629 /*
630  * Extended fields will fit into an inode if the filesystem was formatted
631  * with large inodes (-I 256 or larger) and there are not currently any EAs
632  * consuming all of the available space. For new inodes we always reserve
633  * enough space for the kernel's known extended fields, but for inodes
634  * created with an old kernel this might not have been the case. None of
635  * the extended inode fields is critical for correct filesystem operation.
636  * This macro checks if a certain field fits in the inode. Note that
637  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
638  */
639 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
640         ((offsetof(typeof(*ext4_inode), field) +        \
641           sizeof((ext4_inode)->field))                  \
642         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
643             (einode)->i_extra_isize))                   \
644
645 static inline __le32 ext4_encode_extra_time(struct timespec *time)
646 {
647        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
648                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
649                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
650 }
651
652 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
653 {
654        if (sizeof(time->tv_sec) > 4)
655                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
656                                << 32;
657        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
658 }
659
660 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
661 do {                                                                           \
662         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
663         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
664                 (raw_inode)->xtime ## _extra =                                 \
665                                 ext4_encode_extra_time(&(inode)->xtime);       \
666 } while (0)
667
668 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
669 do {                                                                           \
670         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
671                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
672         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
673                 (raw_inode)->xtime ## _extra =                                 \
674                                 ext4_encode_extra_time(&(einode)->xtime);      \
675 } while (0)
676
677 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
678 do {                                                                           \
679         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
680         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
681                 ext4_decode_extra_time(&(inode)->xtime,                        \
682                                        raw_inode->xtime ## _extra);            \
683 } while (0)
684
685 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
686 do {                                                                           \
687         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
688                 (einode)->xtime.tv_sec =                                       \
689                         (signed)le32_to_cpu((raw_inode)->xtime);               \
690         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
691                 ext4_decode_extra_time(&(einode)->xtime,                       \
692                                        raw_inode->xtime ## _extra);            \
693 } while (0)
694
695 #define i_disk_version osd1.linux1.l_i_version
696
697 #if defined(__KERNEL__) || defined(__linux__)
698 #define i_reserved1     osd1.linux1.l_i_reserved1
699 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
700 #define i_blocks_high   osd2.linux2.l_i_blocks_high
701 #define i_uid_low       i_uid
702 #define i_gid_low       i_gid
703 #define i_uid_high      osd2.linux2.l_i_uid_high
704 #define i_gid_high      osd2.linux2.l_i_gid_high
705 #define i_reserved2     osd2.linux2.l_i_reserved2
706
707 #elif defined(__GNU__)
708
709 #define i_translator    osd1.hurd1.h_i_translator
710 #define i_uid_high      osd2.hurd2.h_i_uid_high
711 #define i_gid_high      osd2.hurd2.h_i_gid_high
712 #define i_author        osd2.hurd2.h_i_author
713
714 #elif defined(__masix__)
715
716 #define i_reserved1     osd1.masix1.m_i_reserved1
717 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
718 #define i_reserved2     osd2.masix2.m_i_reserved2
719
720 #endif /* defined(__KERNEL__) || defined(__linux__) */
721
722 /*
723  * storage for cached extent
724  */
725 struct ext4_ext_cache {
726         ext4_fsblk_t    ec_start;
727         ext4_lblk_t     ec_block;
728         __u32           ec_len; /* must be 32bit to return holes */
729         __u32           ec_type;
730 };
731
732 /*
733  * fourth extended file system inode data in memory
734  */
735 struct ext4_inode_info {
736         __le32  i_data[15];     /* unconverted */
737         __u32   i_dtime;
738         ext4_fsblk_t    i_file_acl;
739
740         /*
741          * i_block_group is the number of the block group which contains
742          * this file's inode.  Constant across the lifetime of the inode,
743          * it is ued for making block allocation decisions - we try to
744          * place a file's data blocks near its inode block, and new inodes
745          * near to their parent directory's inode.
746          */
747         ext4_group_t    i_block_group;
748         unsigned long   i_state_flags;          /* Dynamic state flags */
749         unsigned long   i_flags;
750
751         ext4_lblk_t             i_dir_start_lookup;
752 #ifdef CONFIG_EXT4_FS_XATTR
753         /*
754          * Extended attributes can be read independently of the main file
755          * data. Taking i_mutex even when reading would cause contention
756          * between readers of EAs and writers of regular file data, so
757          * instead we synchronize on xattr_sem when reading or changing
758          * EAs.
759          */
760         struct rw_semaphore xattr_sem;
761 #endif
762
763         struct list_head i_orphan;      /* unlinked but open inodes */
764
765         /*
766          * i_disksize keeps track of what the inode size is ON DISK, not
767          * in memory.  During truncate, i_size is set to the new size by
768          * the VFS prior to calling ext4_truncate(), but the filesystem won't
769          * set i_disksize to 0 until the truncate is actually under way.
770          *
771          * The intent is that i_disksize always represents the blocks which
772          * are used by this file.  This allows recovery to restart truncate
773          * on orphans if we crash during truncate.  We actually write i_disksize
774          * into the on-disk inode when writing inodes out, instead of i_size.
775          *
776          * The only time when i_disksize and i_size may be different is when
777          * a truncate is in progress.  The only things which change i_disksize
778          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
779          */
780         loff_t  i_disksize;
781
782         /*
783          * i_data_sem is for serialising ext4_truncate() against
784          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
785          * data tree are chopped off during truncate. We can't do that in
786          * ext4 because whenever we perform intermediate commits during
787          * truncate, the inode and all the metadata blocks *must* be in a
788          * consistent state which allows truncation of the orphans to restart
789          * during recovery.  Hence we must fix the get_block-vs-truncate race
790          * by other means, so we have i_data_sem.
791          */
792         struct rw_semaphore i_data_sem;
793         struct inode vfs_inode;
794         struct jbd2_inode jinode;
795
796         struct ext4_ext_cache i_cached_extent;
797         /*
798          * File creation time. Its function is same as that of
799          * struct timespec i_{a,c,m}time in the generic inode.
800          */
801         struct timespec i_crtime;
802
803         /* mballoc */
804         struct list_head i_prealloc_list;
805         spinlock_t i_prealloc_lock;
806
807         /* ialloc */
808         ext4_group_t    i_last_alloc_group;
809
810         /* allocation reservation info for delalloc */
811         unsigned int i_reserved_data_blocks;
812         unsigned int i_reserved_meta_blocks;
813         unsigned int i_allocated_meta_blocks;
814         unsigned short i_delalloc_reserved_flag;
815         sector_t i_da_metadata_calc_last_lblock;
816         int i_da_metadata_calc_len;
817
818         /* on-disk additional length */
819         __u16 i_extra_isize;
820
821         spinlock_t i_block_reservation_lock;
822 #ifdef CONFIG_QUOTA
823         /* quota space reservation, managed internally by quota code */
824         qsize_t i_reserved_quota;
825 #endif
826
827         /* completed IOs that might need unwritten extents handling */
828         struct list_head i_completed_io_list;
829         spinlock_t i_completed_io_lock;
830         /* current io_end structure for async DIO write*/
831         ext4_io_end_t *cur_aio_dio;
832
833         /*
834          * Transactions that contain inode's metadata needed to complete
835          * fsync and fdatasync, respectively.
836          */
837         tid_t i_sync_tid;
838         tid_t i_datasync_tid;
839 };
840
841 /*
842  * File system states
843  */
844 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
845 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
846 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
847
848 /*
849  * Misc. filesystem flags
850  */
851 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
852 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
853 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
854
855 /*
856  * Mount flags
857  */
858 #define EXT4_MOUNT_OLDALLOC             0x00002  /* Don't use the new Orlov allocator */
859 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
860 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
861 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
862 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
863 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
864 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
865 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
866 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
867 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
868 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
869 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
870 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
871 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
872 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
873 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
874 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
875 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
876 #define EXT4_MOUNT_NOBH                 0x40000 /* No bufferheads */
877 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
878 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
879 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
880 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
881 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
882 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
883 #define EXT4_MOUNT_I_VERSION            0x2000000 /* i_version support */
884 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
885 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
886 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
887 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
888
889 #define clear_opt(o, opt)               o &= ~EXT4_MOUNT_##opt
890 #define set_opt(o, opt)                 o |= EXT4_MOUNT_##opt
891 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
892                                          EXT4_MOUNT_##opt)
893
894 #define ext4_set_bit                    ext2_set_bit
895 #define ext4_set_bit_atomic             ext2_set_bit_atomic
896 #define ext4_clear_bit                  ext2_clear_bit
897 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
898 #define ext4_test_bit                   ext2_test_bit
899 #define ext4_find_first_zero_bit        ext2_find_first_zero_bit
900 #define ext4_find_next_zero_bit         ext2_find_next_zero_bit
901 #define ext4_find_next_bit              ext2_find_next_bit
902
903 /*
904  * Maximal mount counts between two filesystem checks
905  */
906 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
907 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
908
909 /*
910  * Behaviour when detecting errors
911  */
912 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
913 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
914 #define EXT4_ERRORS_PANIC               3       /* Panic */
915 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
916
917 /*
918  * Structure of the super block
919  */
920 struct ext4_super_block {
921 /*00*/  __le32  s_inodes_count;         /* Inodes count */
922         __le32  s_blocks_count_lo;      /* Blocks count */
923         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
924         __le32  s_free_blocks_count_lo; /* Free blocks count */
925 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
926         __le32  s_first_data_block;     /* First Data Block */
927         __le32  s_log_block_size;       /* Block size */
928         __le32  s_obso_log_frag_size;   /* Obsoleted fragment size */
929 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
930         __le32  s_obso_frags_per_group; /* Obsoleted fragments per group */
931         __le32  s_inodes_per_group;     /* # Inodes per group */
932         __le32  s_mtime;                /* Mount time */
933 /*30*/  __le32  s_wtime;                /* Write time */
934         __le16  s_mnt_count;            /* Mount count */
935         __le16  s_max_mnt_count;        /* Maximal mount count */
936         __le16  s_magic;                /* Magic signature */
937         __le16  s_state;                /* File system state */
938         __le16  s_errors;               /* Behaviour when detecting errors */
939         __le16  s_minor_rev_level;      /* minor revision level */
940 /*40*/  __le32  s_lastcheck;            /* time of last check */
941         __le32  s_checkinterval;        /* max. time between checks */
942         __le32  s_creator_os;           /* OS */
943         __le32  s_rev_level;            /* Revision level */
944 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
945         __le16  s_def_resgid;           /* Default gid for reserved blocks */
946         /*
947          * These fields are for EXT4_DYNAMIC_REV superblocks only.
948          *
949          * Note: the difference between the compatible feature set and
950          * the incompatible feature set is that if there is a bit set
951          * in the incompatible feature set that the kernel doesn't
952          * know about, it should refuse to mount the filesystem.
953          *
954          * e2fsck's requirements are more strict; if it doesn't know
955          * about a feature in either the compatible or incompatible
956          * feature set, it must abort and not try to meddle with
957          * things it doesn't understand...
958          */
959         __le32  s_first_ino;            /* First non-reserved inode */
960         __le16  s_inode_size;           /* size of inode structure */
961         __le16  s_block_group_nr;       /* block group # of this superblock */
962         __le32  s_feature_compat;       /* compatible feature set */
963 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
964         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
965 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
966 /*78*/  char    s_volume_name[16];      /* volume name */
967 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
968 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
969         /*
970          * Performance hints.  Directory preallocation should only
971          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
972          */
973         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
974         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
975         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
976         /*
977          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
978          */
979 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
980 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
981         __le32  s_journal_dev;          /* device number of journal file */
982         __le32  s_last_orphan;          /* start of list of inodes to delete */
983         __le32  s_hash_seed[4];         /* HTREE hash seed */
984         __u8    s_def_hash_version;     /* Default hash version to use */
985         __u8    s_reserved_char_pad;
986         __le16  s_desc_size;            /* size of group descriptor */
987 /*100*/ __le32  s_default_mount_opts;
988         __le32  s_first_meta_bg;        /* First metablock block group */
989         __le32  s_mkfs_time;            /* When the filesystem was created */
990         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
991         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
992 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
993         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
994         __le32  s_free_blocks_count_hi; /* Free blocks count */
995         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
996         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
997         __le32  s_flags;                /* Miscellaneous flags */
998         __le16  s_raid_stride;          /* RAID stride */
999         __le16  s_mmp_interval;         /* # seconds to wait in MMP checking */
1000         __le64  s_mmp_block;            /* Block for multi-mount protection */
1001         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1002         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1003         __u8    s_reserved_char_pad2;
1004         __le16  s_reserved_pad;
1005         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1006         __u32   s_reserved[160];        /* Padding to the end of the block */
1007 };
1008
1009 #ifdef __KERNEL__
1010
1011 /*
1012  * run-time mount flags
1013  */
1014 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
1015 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
1016
1017 /*
1018  * fourth extended-fs super-block data in memory
1019  */
1020 struct ext4_sb_info {
1021         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1022         unsigned long s_inodes_per_block;/* Number of inodes per block */
1023         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1024         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1025         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1026         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1027         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1028         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1029         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1030         unsigned long s_overhead_last;  /* Last calculated overhead */
1031         unsigned long s_blocks_last;    /* Last seen block count */
1032         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1033         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1034         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1035         struct buffer_head **s_group_desc;
1036         unsigned int s_mount_opt;
1037         unsigned int s_mount_flags;
1038         ext4_fsblk_t s_sb_block;
1039         uid_t s_resuid;
1040         gid_t s_resgid;
1041         unsigned short s_mount_state;
1042         unsigned short s_pad;
1043         int s_addr_per_block_bits;
1044         int s_desc_per_block_bits;
1045         int s_inode_size;
1046         int s_first_ino;
1047         unsigned int s_inode_readahead_blks;
1048         unsigned int s_inode_goal;
1049         spinlock_t s_next_gen_lock;
1050         u32 s_next_generation;
1051         u32 s_hash_seed[4];
1052         int s_def_hash_version;
1053         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1054         struct percpu_counter s_freeblocks_counter;
1055         struct percpu_counter s_freeinodes_counter;
1056         struct percpu_counter s_dirs_counter;
1057         struct percpu_counter s_dirtyblocks_counter;
1058         struct blockgroup_lock *s_blockgroup_lock;
1059         struct proc_dir_entry *s_proc;
1060         struct kobject s_kobj;
1061         struct completion s_kobj_unregister;
1062
1063         /* Journaling */
1064         struct inode *s_journal_inode;
1065         struct journal_s *s_journal;
1066         struct list_head s_orphan;
1067         struct mutex s_orphan_lock;
1068         struct mutex s_resize_lock;
1069         unsigned long s_commit_interval;
1070         u32 s_max_batch_time;
1071         u32 s_min_batch_time;
1072         struct block_device *journal_bdev;
1073 #ifdef CONFIG_JBD2_DEBUG
1074         struct timer_list turn_ro_timer;        /* For turning read-only (crash simulation) */
1075         wait_queue_head_t ro_wait_queue;        /* For people waiting for the fs to go read-only */
1076 #endif
1077 #ifdef CONFIG_QUOTA
1078         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
1079         int s_jquota_fmt;                       /* Format of quota to use */
1080 #endif
1081         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1082         struct rb_root system_blks;
1083
1084 #ifdef EXTENTS_STATS
1085         /* ext4 extents stats */
1086         unsigned long s_ext_min;
1087         unsigned long s_ext_max;
1088         unsigned long s_depth_max;
1089         spinlock_t s_ext_stats_lock;
1090         unsigned long s_ext_blocks;
1091         unsigned long s_ext_extents;
1092 #endif
1093
1094         /* for buddy allocator */
1095         struct ext4_group_info ***s_group_info;
1096         struct inode *s_buddy_cache;
1097         long s_blocks_reserved;
1098         spinlock_t s_reserve_lock;
1099         spinlock_t s_md_lock;
1100         tid_t s_last_transaction;
1101         unsigned short *s_mb_offsets;
1102         unsigned int *s_mb_maxs;
1103
1104         /* tunables */
1105         unsigned long s_stripe;
1106         unsigned int s_mb_stream_request;
1107         unsigned int s_mb_max_to_scan;
1108         unsigned int s_mb_min_to_scan;
1109         unsigned int s_mb_stats;
1110         unsigned int s_mb_order2_reqs;
1111         unsigned int s_mb_group_prealloc;
1112         unsigned int s_max_writeback_mb_bump;
1113         /* where last allocation was done - for stream allocation */
1114         unsigned long s_mb_last_group;
1115         unsigned long s_mb_last_start;
1116
1117         /* stats for buddy allocator */
1118         spinlock_t s_mb_pa_lock;
1119         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1120         atomic_t s_bal_success; /* we found long enough chunks */
1121         atomic_t s_bal_allocated;       /* in blocks */
1122         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1123         atomic_t s_bal_goals;   /* goal hits */
1124         atomic_t s_bal_breaks;  /* too long searches */
1125         atomic_t s_bal_2orders; /* 2^order hits */
1126         spinlock_t s_bal_lock;
1127         unsigned long s_mb_buddies_generated;
1128         unsigned long long s_mb_generation_time;
1129         atomic_t s_mb_lost_chunks;
1130         atomic_t s_mb_preallocated;
1131         atomic_t s_mb_discarded;
1132         atomic_t s_lock_busy;
1133
1134         /* locality groups */
1135         struct ext4_locality_group __percpu *s_locality_groups;
1136
1137         /* for write statistics */
1138         unsigned long s_sectors_written_start;
1139         u64 s_kbytes_written;
1140
1141         unsigned int s_log_groups_per_flex;
1142         struct flex_groups *s_flex_groups;
1143
1144         /* workqueue for dio unwritten */
1145         struct workqueue_struct *dio_unwritten_wq;
1146 };
1147
1148 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1149 {
1150         return sb->s_fs_info;
1151 }
1152 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1153 {
1154         return container_of(inode, struct ext4_inode_info, vfs_inode);
1155 }
1156
1157 static inline struct timespec ext4_current_time(struct inode *inode)
1158 {
1159         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1160                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1161 }
1162
1163 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1164 {
1165         return ino == EXT4_ROOT_INO ||
1166                 ino == EXT4_JOURNAL_INO ||
1167                 ino == EXT4_RESIZE_INO ||
1168                 (ino >= EXT4_FIRST_INO(sb) &&
1169                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1170 }
1171
1172 /*
1173  * Inode dynamic state flags
1174  */
1175 enum {
1176         EXT4_STATE_JDATA,               /* journaled data exists */
1177         EXT4_STATE_NEW,                 /* inode is newly created */
1178         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1179         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1180         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1181         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1182         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1183         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1184 };
1185
1186 #define EXT4_INODE_BIT_FNS(name, field)                                 \
1187 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1188 {                                                                       \
1189         return test_bit(bit, &EXT4_I(inode)->i_##field);                \
1190 }                                                                       \
1191 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1192 {                                                                       \
1193         set_bit(bit, &EXT4_I(inode)->i_##field);                        \
1194 }                                                                       \
1195 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1196 {                                                                       \
1197         clear_bit(bit, &EXT4_I(inode)->i_##field);                      \
1198 }
1199
1200 EXT4_INODE_BIT_FNS(flag, flags)
1201 EXT4_INODE_BIT_FNS(state, state_flags)
1202 #else
1203 /* Assume that user mode programs are passing in an ext4fs superblock, not
1204  * a kernel struct super_block.  This will allow us to call the feature-test
1205  * macros from user land. */
1206 #define EXT4_SB(sb)     (sb)
1207 #endif
1208
1209 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1210
1211 /*
1212  * Codes for operating systems
1213  */
1214 #define EXT4_OS_LINUX           0
1215 #define EXT4_OS_HURD            1
1216 #define EXT4_OS_MASIX           2
1217 #define EXT4_OS_FREEBSD         3
1218 #define EXT4_OS_LITES           4
1219
1220 /*
1221  * Revision levels
1222  */
1223 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1224 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1225
1226 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1227 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1228
1229 #define EXT4_GOOD_OLD_INODE_SIZE 128
1230
1231 /*
1232  * Feature set definitions
1233  */
1234
1235 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1236         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1237 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1238         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1239 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1240         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1241 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1242         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1243 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1244         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1245 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1246         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1247 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1248         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1249 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1250         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1251 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1252         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1253
1254 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1255 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1256 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1257 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1258 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1259 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1260
1261 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1262 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1263 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1264 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1265 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1266 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1267 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1268
1269 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1270 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1271 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1272 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1273 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1274 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1275 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1276 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1277 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1278 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1279 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1280
1281 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1282 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1283                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1284                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1285                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1286                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1287                                          EXT4_FEATURE_INCOMPAT_FLEX_BG)
1288 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1289                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1290                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1291                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1292                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1293                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1294                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE)
1295
1296 /*
1297  * Default values for user and/or group using reserved blocks
1298  */
1299 #define EXT4_DEF_RESUID         0
1300 #define EXT4_DEF_RESGID         0
1301
1302 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1303
1304 /*
1305  * Default mount options
1306  */
1307 #define EXT4_DEFM_DEBUG         0x0001
1308 #define EXT4_DEFM_BSDGROUPS     0x0002
1309 #define EXT4_DEFM_XATTR_USER    0x0004
1310 #define EXT4_DEFM_ACL           0x0008
1311 #define EXT4_DEFM_UID16         0x0010
1312 #define EXT4_DEFM_JMODE         0x0060
1313 #define EXT4_DEFM_JMODE_DATA    0x0020
1314 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1315 #define EXT4_DEFM_JMODE_WBACK   0x0060
1316
1317 /*
1318  * Default journal batch times
1319  */
1320 #define EXT4_DEF_MIN_BATCH_TIME 0
1321 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1322
1323 /*
1324  * Minimum number of groups in a flexgroup before we separate out
1325  * directories into the first block group of a flexgroup
1326  */
1327 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1328
1329 /*
1330  * Structure of a directory entry
1331  */
1332 #define EXT4_NAME_LEN 255
1333
1334 struct ext4_dir_entry {
1335         __le32  inode;                  /* Inode number */
1336         __le16  rec_len;                /* Directory entry length */
1337         __le16  name_len;               /* Name length */
1338         char    name[EXT4_NAME_LEN];    /* File name */
1339 };
1340
1341 /*
1342  * The new version of the directory entry.  Since EXT4 structures are
1343  * stored in intel byte order, and the name_len field could never be
1344  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1345  * file_type field.
1346  */
1347 struct ext4_dir_entry_2 {
1348         __le32  inode;                  /* Inode number */
1349         __le16  rec_len;                /* Directory entry length */
1350         __u8    name_len;               /* Name length */
1351         __u8    file_type;
1352         char    name[EXT4_NAME_LEN];    /* File name */
1353 };
1354
1355 /*
1356  * Ext4 directory file types.  Only the low 3 bits are used.  The
1357  * other bits are reserved for now.
1358  */
1359 #define EXT4_FT_UNKNOWN         0
1360 #define EXT4_FT_REG_FILE        1
1361 #define EXT4_FT_DIR             2
1362 #define EXT4_FT_CHRDEV          3
1363 #define EXT4_FT_BLKDEV          4
1364 #define EXT4_FT_FIFO            5
1365 #define EXT4_FT_SOCK            6
1366 #define EXT4_FT_SYMLINK         7
1367
1368 #define EXT4_FT_MAX             8
1369
1370 /*
1371  * EXT4_DIR_PAD defines the directory entries boundaries
1372  *
1373  * NOTE: It must be a multiple of 4
1374  */
1375 #define EXT4_DIR_PAD                    4
1376 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1377 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1378                                          ~EXT4_DIR_ROUND)
1379 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1380
1381 /*
1382  * Hash Tree Directory indexing
1383  * (c) Daniel Phillips, 2001
1384  */
1385
1386 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1387                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1388                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1389 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1390 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1391
1392 /* Legal values for the dx_root hash_version field: */
1393
1394 #define DX_HASH_LEGACY          0
1395 #define DX_HASH_HALF_MD4        1
1396 #define DX_HASH_TEA             2
1397 #define DX_HASH_LEGACY_UNSIGNED 3
1398 #define DX_HASH_HALF_MD4_UNSIGNED       4
1399 #define DX_HASH_TEA_UNSIGNED            5
1400
1401 #ifdef __KERNEL__
1402
1403 /* hash info structure used by the directory hash */
1404 struct dx_hash_info
1405 {
1406         u32             hash;
1407         u32             minor_hash;
1408         int             hash_version;
1409         u32             *seed;
1410 };
1411
1412 #define EXT4_HTREE_EOF  0x7fffffff
1413
1414 /*
1415  * Control parameters used by ext4_htree_next_block
1416  */
1417 #define HASH_NB_ALWAYS          1
1418
1419
1420 /*
1421  * Describe an inode's exact location on disk and in memory
1422  */
1423 struct ext4_iloc
1424 {
1425         struct buffer_head *bh;
1426         unsigned long offset;
1427         ext4_group_t block_group;
1428 };
1429
1430 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1431 {
1432         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1433 }
1434
1435 /*
1436  * This structure is stuffed into the struct file's private_data field
1437  * for directories.  It is where we put information so that we can do
1438  * readdir operations in hash tree order.
1439  */
1440 struct dir_private_info {
1441         struct rb_root  root;
1442         struct rb_node  *curr_node;
1443         struct fname    *extra_fname;
1444         loff_t          last_pos;
1445         __u32           curr_hash;
1446         __u32           curr_minor_hash;
1447         __u32           next_hash;
1448 };
1449
1450 /* calculate the first block number of the group */
1451 static inline ext4_fsblk_t
1452 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1453 {
1454         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1455                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1456 }
1457
1458 /*
1459  * Special error return code only used by dx_probe() and its callers.
1460  */
1461 #define ERR_BAD_DX_DIR  -75000
1462
1463 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1464                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1465
1466 extern struct proc_dir_entry *ext4_proc_root;
1467
1468 /*
1469  * Function prototypes
1470  */
1471
1472 /*
1473  * Ok, these declarations are also in <linux/kernel.h> but none of the
1474  * ext4 source programs needs to include it so they are duplicated here.
1475  */
1476 # define NORET_TYPE     /**/
1477 # define ATTRIB_NORET   __attribute__((noreturn))
1478 # define NORET_AND      noreturn,
1479
1480 /* bitmap.c */
1481 extern unsigned int ext4_count_free(struct buffer_head *, unsigned);
1482
1483 /* balloc.c */
1484 extern unsigned int ext4_block_group(struct super_block *sb,
1485                         ext4_fsblk_t blocknr);
1486 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1487                         ext4_fsblk_t blocknr);
1488 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1489 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1490                         ext4_group_t group);
1491 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1492                         ext4_fsblk_t goal, unsigned long *count, int *errp);
1493 extern int ext4_claim_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1494 extern int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1495 extern void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
1496                                 ext4_fsblk_t block, unsigned long count);
1497 extern ext4_fsblk_t ext4_count_free_blocks(struct super_block *);
1498 extern void ext4_check_blocks_bitmap(struct super_block *);
1499 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1500                                                     ext4_group_t block_group,
1501                                                     struct buffer_head ** bh);
1502 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1503 struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1504                                       ext4_group_t block_group);
1505 extern unsigned ext4_init_block_bitmap(struct super_block *sb,
1506                                        struct buffer_head *bh,
1507                                        ext4_group_t group,
1508                                        struct ext4_group_desc *desc);
1509 #define ext4_free_blocks_after_init(sb, group, desc)                    \
1510                 ext4_init_block_bitmap(sb, NULL, group, desc)
1511
1512 /* dir.c */
1513 extern int ext4_check_dir_entry(const char *, struct inode *,
1514                                 struct ext4_dir_entry_2 *,
1515                                 struct buffer_head *, unsigned int);
1516 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1517                                     __u32 minor_hash,
1518                                     struct ext4_dir_entry_2 *dirent);
1519 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1520
1521 /* fsync.c */
1522 extern int ext4_sync_file(struct file *, int);
1523
1524 /* hash.c */
1525 extern int ext4fs_dirhash(const char *name, int len, struct
1526                           dx_hash_info *hinfo);
1527
1528 /* ialloc.c */
1529 extern struct inode *ext4_new_inode(handle_t *, struct inode *, int,
1530                                     const struct qstr *qstr, __u32 goal);
1531 extern void ext4_free_inode(handle_t *, struct inode *);
1532 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1533 extern unsigned long ext4_count_free_inodes(struct super_block *);
1534 extern unsigned long ext4_count_dirs(struct super_block *);
1535 extern void ext4_check_inodes_bitmap(struct super_block *);
1536 extern unsigned ext4_init_inode_bitmap(struct super_block *sb,
1537                                        struct buffer_head *bh,
1538                                        ext4_group_t group,
1539                                        struct ext4_group_desc *desc);
1540 extern void mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1541
1542 /* mballoc.c */
1543 extern long ext4_mb_stats;
1544 extern long ext4_mb_max_to_scan;
1545 extern int ext4_mb_init(struct super_block *, int);
1546 extern int ext4_mb_release(struct super_block *);
1547 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1548                                 struct ext4_allocation_request *, int *);
1549 extern int ext4_mb_reserve_blocks(struct super_block *, int);
1550 extern void ext4_discard_preallocations(struct inode *);
1551 extern int __init init_ext4_mballoc(void);
1552 extern void exit_ext4_mballoc(void);
1553 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1554                              struct buffer_head *bh, ext4_fsblk_t block,
1555                              unsigned long count, int flags);
1556 extern int ext4_mb_add_groupinfo(struct super_block *sb,
1557                 ext4_group_t i, struct ext4_group_desc *desc);
1558 extern int ext4_mb_get_buddy_cache_lock(struct super_block *, ext4_group_t);
1559 extern void ext4_mb_put_buddy_cache_lock(struct super_block *,
1560                                                 ext4_group_t, int);
1561 /* inode.c */
1562 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1563                                                 ext4_lblk_t, int, int *);
1564 struct buffer_head *ext4_bread(handle_t *, struct inode *,
1565                                                 ext4_lblk_t, int, int *);
1566 int ext4_get_block(struct inode *inode, sector_t iblock,
1567                                 struct buffer_head *bh_result, int create);
1568
1569 extern struct inode *ext4_iget(struct super_block *, unsigned long);
1570 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
1571 extern int  ext4_setattr(struct dentry *, struct iattr *);
1572 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1573                                 struct kstat *stat);
1574 extern void ext4_delete_inode(struct inode *);
1575 extern int  ext4_sync_inode(handle_t *, struct inode *);
1576 extern void ext4_dirty_inode(struct inode *);
1577 extern int ext4_change_inode_journal_flag(struct inode *, int);
1578 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1579 extern int ext4_can_truncate(struct inode *inode);
1580 extern void ext4_truncate(struct inode *);
1581 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
1582 extern void ext4_set_inode_flags(struct inode *);
1583 extern void ext4_get_inode_flags(struct ext4_inode_info *);
1584 extern int ext4_alloc_da_blocks(struct inode *inode);
1585 extern void ext4_set_aops(struct inode *inode);
1586 extern int ext4_writepage_trans_blocks(struct inode *);
1587 extern int ext4_meta_trans_blocks(struct inode *, int nrblocks, int idxblocks);
1588 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1589 extern int ext4_block_truncate_page(handle_t *handle,
1590                 struct address_space *mapping, loff_t from);
1591 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1592 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
1593 extern int flush_completed_IO(struct inode *inode);
1594 extern void ext4_da_update_reserve_space(struct inode *inode,
1595                                         int used, int quota_claim);
1596 /* ioctl.c */
1597 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1598 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1599
1600 /* migrate.c */
1601 extern int ext4_ext_migrate(struct inode *);
1602
1603 /* namei.c */
1604 extern unsigned int ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize);
1605 extern __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize);
1606 extern int ext4_orphan_add(handle_t *, struct inode *);
1607 extern int ext4_orphan_del(handle_t *, struct inode *);
1608 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1609                                 __u32 start_minor_hash, __u32 *next_hash);
1610
1611 /* resize.c */
1612 extern int ext4_group_add(struct super_block *sb,
1613                                 struct ext4_new_group_data *input);
1614 extern int ext4_group_extend(struct super_block *sb,
1615                                 struct ext4_super_block *es,
1616                                 ext4_fsblk_t n_blocks_count);
1617
1618 /* super.c */
1619 extern void __ext4_error(struct super_block *, const char *, const char *, ...)
1620         __attribute__ ((format (printf, 3, 4)));
1621 #define ext4_error(sb, message...)      __ext4_error(sb, __func__, ## message)
1622 extern void ext4_error_inode(const char *, struct inode *, const char *, ...)
1623         __attribute__ ((format (printf, 3, 4)));
1624 extern void ext4_error_file(const char *, struct file *, const char *, ...)
1625         __attribute__ ((format (printf, 3, 4)));
1626 extern void __ext4_std_error(struct super_block *, const char *, int);
1627 extern void ext4_abort(struct super_block *, const char *, const char *, ...)
1628         __attribute__ ((format (printf, 3, 4)));
1629 extern void __ext4_warning(struct super_block *, const char *,
1630                           const char *, ...)
1631         __attribute__ ((format (printf, 3, 4)));
1632 #define ext4_warning(sb, message...)    __ext4_warning(sb, __func__, ## message)
1633 extern void ext4_msg(struct super_block *, const char *, const char *, ...)
1634         __attribute__ ((format (printf, 3, 4)));
1635 extern void ext4_grp_locked_error(struct super_block *, ext4_group_t,
1636                                 const char *, const char *, ...)
1637         __attribute__ ((format (printf, 4, 5)));
1638 extern void ext4_update_dynamic_rev(struct super_block *sb);
1639 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1640                                         __u32 compat);
1641 extern int ext4_update_rocompat_feature(handle_t *handle,
1642                                         struct super_block *sb, __u32 rocompat);
1643 extern int ext4_update_incompat_feature(handle_t *handle,
1644                                         struct super_block *sb, __u32 incompat);
1645 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1646                                       struct ext4_group_desc *bg);
1647 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1648                                       struct ext4_group_desc *bg);
1649 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1650                                      struct ext4_group_desc *bg);
1651 extern __u32 ext4_free_blks_count(struct super_block *sb,
1652                                 struct ext4_group_desc *bg);
1653 extern __u32 ext4_free_inodes_count(struct super_block *sb,
1654                                  struct ext4_group_desc *bg);
1655 extern __u32 ext4_used_dirs_count(struct super_block *sb,
1656                                 struct ext4_group_desc *bg);
1657 extern __u32 ext4_itable_unused_count(struct super_block *sb,
1658                                    struct ext4_group_desc *bg);
1659 extern void ext4_block_bitmap_set(struct super_block *sb,
1660                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1661 extern void ext4_inode_bitmap_set(struct super_block *sb,
1662                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1663 extern void ext4_inode_table_set(struct super_block *sb,
1664                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
1665 extern void ext4_free_blks_set(struct super_block *sb,
1666                                struct ext4_group_desc *bg, __u32 count);
1667 extern void ext4_free_inodes_set(struct super_block *sb,
1668                                 struct ext4_group_desc *bg, __u32 count);
1669 extern void ext4_used_dirs_set(struct super_block *sb,
1670                                 struct ext4_group_desc *bg, __u32 count);
1671 extern void ext4_itable_unused_set(struct super_block *sb,
1672                                    struct ext4_group_desc *bg, __u32 count);
1673 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
1674                                    struct ext4_group_desc *gdp);
1675 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
1676                                        struct ext4_group_desc *gdp);
1677
1678 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
1679 {
1680         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
1681                 le32_to_cpu(es->s_blocks_count_lo);
1682 }
1683
1684 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
1685 {
1686         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
1687                 le32_to_cpu(es->s_r_blocks_count_lo);
1688 }
1689
1690 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
1691 {
1692         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
1693                 le32_to_cpu(es->s_free_blocks_count_lo);
1694 }
1695
1696 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
1697                                          ext4_fsblk_t blk)
1698 {
1699         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
1700         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
1701 }
1702
1703 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
1704                                               ext4_fsblk_t blk)
1705 {
1706         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
1707         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
1708 }
1709
1710 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
1711                                            ext4_fsblk_t blk)
1712 {
1713         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
1714         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
1715 }
1716
1717 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
1718 {
1719         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
1720                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
1721                         le32_to_cpu(raw_inode->i_size_lo);
1722         else
1723                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
1724 }
1725
1726 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
1727 {
1728         raw_inode->i_size_lo = cpu_to_le32(i_size);
1729         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
1730 }
1731
1732 static inline
1733 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
1734                                             ext4_group_t group)
1735 {
1736          struct ext4_group_info ***grp_info;
1737          long indexv, indexh;
1738          grp_info = EXT4_SB(sb)->s_group_info;
1739          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
1740          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
1741          return grp_info[indexv][indexh];
1742 }
1743
1744 /*
1745  * Reading s_groups_count requires using smp_rmb() afterwards.  See
1746  * the locking protocol documented in the comments of ext4_group_add()
1747  * in resize.c
1748  */
1749 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
1750 {
1751         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
1752
1753         smp_rmb();
1754         return ngroups;
1755 }
1756
1757 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
1758                                              ext4_group_t block_group)
1759 {
1760         return block_group >> sbi->s_log_groups_per_flex;
1761 }
1762
1763 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
1764 {
1765         return 1 << sbi->s_log_groups_per_flex;
1766 }
1767
1768 #define ext4_std_error(sb, errno)                               \
1769 do {                                                            \
1770         if ((errno))                                            \
1771                 __ext4_std_error((sb), __func__, (errno));      \
1772 } while (0)
1773
1774 #ifdef CONFIG_SMP
1775 /* Each CPU can accumulate percpu_counter_batch blocks in their local
1776  * counters. So we need to make sure we have free blocks more
1777  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
1778  */
1779 #define EXT4_FREEBLOCKS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
1780 #else
1781 #define EXT4_FREEBLOCKS_WATERMARK 0
1782 #endif
1783
1784 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
1785 {
1786         /*
1787          * XXX: replace with spinlock if seen contended -bzzz
1788          */
1789         down_write(&EXT4_I(inode)->i_data_sem);
1790         if (newsize > EXT4_I(inode)->i_disksize)
1791                 EXT4_I(inode)->i_disksize = newsize;
1792         up_write(&EXT4_I(inode)->i_data_sem);
1793         return ;
1794 }
1795
1796 struct ext4_group_info {
1797         unsigned long   bb_state;
1798         struct rb_root  bb_free_root;
1799         ext4_grpblk_t   bb_first_free;  /* first free block */
1800         ext4_grpblk_t   bb_free;        /* total free blocks */
1801         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
1802         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
1803         struct          list_head bb_prealloc_list;
1804 #ifdef DOUBLE_CHECK
1805         void            *bb_bitmap;
1806 #endif
1807         struct rw_semaphore alloc_sem;
1808         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
1809                                          * regions, index is order.
1810                                          * bb_counters[3] = 5 means
1811                                          * 5 free 8-block regions. */
1812 };
1813
1814 #define EXT4_GROUP_INFO_NEED_INIT_BIT   0
1815
1816 #define EXT4_MB_GRP_NEED_INIT(grp)      \
1817         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
1818
1819 #define EXT4_MAX_CONTENTION             8
1820 #define EXT4_CONTENTION_THRESHOLD       2
1821
1822 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
1823                                               ext4_group_t group)
1824 {
1825         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
1826 }
1827
1828 /*
1829  * Returns true if the filesystem is busy enough that attempts to
1830  * access the block group locks has run into contention.
1831  */
1832 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
1833 {
1834         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
1835 }
1836
1837 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
1838 {
1839         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
1840         if (spin_trylock(lock))
1841                 /*
1842                  * We're able to grab the lock right away, so drop the
1843                  * lock contention counter.
1844                  */
1845                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
1846         else {
1847                 /*
1848                  * The lock is busy, so bump the contention counter,
1849                  * and then wait on the spin lock.
1850                  */
1851                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
1852                                   EXT4_MAX_CONTENTION);
1853                 spin_lock(lock);
1854         }
1855 }
1856
1857 static inline void ext4_unlock_group(struct super_block *sb,
1858                                         ext4_group_t group)
1859 {
1860         spin_unlock(ext4_group_lock_ptr(sb, group));
1861 }
1862
1863 /*
1864  * Inodes and files operations
1865  */
1866
1867 /* dir.c */
1868 extern const struct file_operations ext4_dir_operations;
1869
1870 /* file.c */
1871 extern const struct inode_operations ext4_file_inode_operations;
1872 extern const struct file_operations ext4_file_operations;
1873
1874 /* namei.c */
1875 extern const struct inode_operations ext4_dir_inode_operations;
1876 extern const struct inode_operations ext4_special_inode_operations;
1877 extern struct dentry *ext4_get_parent(struct dentry *child);
1878
1879 /* symlink.c */
1880 extern const struct inode_operations ext4_symlink_inode_operations;
1881 extern const struct inode_operations ext4_fast_symlink_inode_operations;
1882
1883 /* block_validity */
1884 extern void ext4_release_system_zone(struct super_block *sb);
1885 extern int ext4_setup_system_zone(struct super_block *sb);
1886 extern int __init init_ext4_system_zone(void);
1887 extern void exit_ext4_system_zone(void);
1888 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
1889                                  ext4_fsblk_t start_blk,
1890                                  unsigned int count);
1891
1892 /* extents.c */
1893 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
1894 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
1895 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
1896                                        int chunk);
1897 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
1898                                struct ext4_map_blocks *map, int flags);
1899 extern void ext4_ext_truncate(struct inode *);
1900 extern void ext4_ext_init(struct super_block *);
1901 extern void ext4_ext_release(struct super_block *);
1902 extern long ext4_fallocate(struct inode *inode, int mode, loff_t offset,
1903                           loff_t len);
1904 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
1905                           ssize_t len);
1906 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
1907                            struct ext4_map_blocks *map, int flags);
1908 extern int ext4_get_blocks(handle_t *handle, struct inode *inode,
1909                            sector_t block, unsigned int max_blocks,
1910                            struct buffer_head *bh, int flags);
1911 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1912                         __u64 start, __u64 len);
1913 /* move_extent.c */
1914 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
1915                              __u64 start_orig, __u64 start_donor,
1916                              __u64 len, __u64 *moved_len);
1917
1918
1919 /* BH_Uninit flag: blocks are allocated but uninitialized on disk */
1920 enum ext4_state_bits {
1921         BH_Uninit       /* blocks are allocated but uninitialized on disk */
1922           = BH_JBDPrivateStart,
1923 };
1924
1925 BUFFER_FNS(Uninit, uninit)
1926 TAS_BUFFER_FNS(Uninit, uninit)
1927
1928 /*
1929  * Add new method to test wether block and inode bitmaps are properly
1930  * initialized. With uninit_bg reading the block from disk is not enough
1931  * to mark the bitmap uptodate. We need to also zero-out the bitmap
1932  */
1933 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
1934
1935 static inline int bitmap_uptodate(struct buffer_head *bh)
1936 {
1937         return (buffer_uptodate(bh) &&
1938                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
1939 }
1940 static inline void set_bitmap_uptodate(struct buffer_head *bh)
1941 {
1942         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
1943 }
1944
1945 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
1946
1947 #endif  /* __KERNEL__ */
1948
1949 #endif  /* _EXT4_H */