nfsd4: ignore WANT bits in open downgrade
[linux-2.6.git] / fs / ext4 / ext4.h
1 /*
2  *  ext4.h
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/include/linux/minix_fs.h
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  */
15
16 #ifndef _EXT4_H
17 #define _EXT4_H
18
19 #include <linux/types.h>
20 #include <linux/blkdev.h>
21 #include <linux/magic.h>
22 #include <linux/jbd2.h>
23 #include <linux/quota.h>
24 #include <linux/rwsem.h>
25 #include <linux/rbtree.h>
26 #include <linux/seqlock.h>
27 #include <linux/mutex.h>
28 #include <linux/timer.h>
29 #include <linux/wait.h>
30 #include <linux/blockgroup_lock.h>
31 #include <linux/percpu_counter.h>
32 #ifdef __KERNEL__
33 #include <linux/compat.h>
34 #endif
35
36 /*
37  * The fourth extended filesystem constants/structures
38  */
39
40 /*
41  * Define EXT4FS_DEBUG to produce debug messages
42  */
43 #undef EXT4FS_DEBUG
44
45 /*
46  * Debug code
47  */
48 #ifdef EXT4FS_DEBUG
49 #define ext4_debug(f, a...)                                             \
50         do {                                                            \
51                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
52                         __FILE__, __LINE__, __func__);                  \
53                 printk(KERN_DEBUG f, ## a);                             \
54         } while (0)
55 #else
56 #define ext4_debug(f, a...)     do {} while (0)
57 #endif
58
59 #define EXT4_ERROR_INODE(inode, fmt, a...) \
60         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
61
62 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...)                 \
63         ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
64
65 #define EXT4_ERROR_FILE(file, block, fmt, a...)                         \
66         ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
67
68 /* data type for block offset of block group */
69 typedef int ext4_grpblk_t;
70
71 /* data type for filesystem-wide blocks number */
72 typedef unsigned long long ext4_fsblk_t;
73
74 /* data type for file logical block number */
75 typedef __u32 ext4_lblk_t;
76
77 /* data type for block group number */
78 typedef unsigned int ext4_group_t;
79
80 /*
81  * Flags used in mballoc's allocation_context flags field.
82  *
83  * Also used to show what's going on for debugging purposes when the
84  * flag field is exported via the traceport interface
85  */
86
87 /* prefer goal again. length */
88 #define EXT4_MB_HINT_MERGE              0x0001
89 /* blocks already reserved */
90 #define EXT4_MB_HINT_RESERVED           0x0002
91 /* metadata is being allocated */
92 #define EXT4_MB_HINT_METADATA           0x0004
93 /* first blocks in the file */
94 #define EXT4_MB_HINT_FIRST              0x0008
95 /* search for the best chunk */
96 #define EXT4_MB_HINT_BEST               0x0010
97 /* data is being allocated */
98 #define EXT4_MB_HINT_DATA               0x0020
99 /* don't preallocate (for tails) */
100 #define EXT4_MB_HINT_NOPREALLOC         0x0040
101 /* allocate for locality group */
102 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
103 /* allocate goal blocks or none */
104 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
105 /* goal is meaningful */
106 #define EXT4_MB_HINT_TRY_GOAL           0x0200
107 /* blocks already pre-reserved by delayed allocation */
108 #define EXT4_MB_DELALLOC_RESERVED       0x0400
109 /* We are doing stream allocation */
110 #define EXT4_MB_STREAM_ALLOC            0x0800
111 /* Use reserved root blocks if needed */
112 #define EXT4_MB_USE_ROOT_BLOCKS         0x1000
113
114 struct ext4_allocation_request {
115         /* target inode for block we're allocating */
116         struct inode *inode;
117         /* how many blocks we want to allocate */
118         unsigned int len;
119         /* logical block in target inode */
120         ext4_lblk_t logical;
121         /* the closest logical allocated block to the left */
122         ext4_lblk_t lleft;
123         /* the closest logical allocated block to the right */
124         ext4_lblk_t lright;
125         /* phys. target (a hint) */
126         ext4_fsblk_t goal;
127         /* phys. block for the closest logical allocated block to the left */
128         ext4_fsblk_t pleft;
129         /* phys. block for the closest logical allocated block to the right */
130         ext4_fsblk_t pright;
131         /* flags. see above EXT4_MB_HINT_* */
132         unsigned int flags;
133 };
134
135 /*
136  * Logical to physical block mapping, used by ext4_map_blocks()
137  *
138  * This structure is used to pass requests into ext4_map_blocks() as
139  * well as to store the information returned by ext4_map_blocks().  It
140  * takes less room on the stack than a struct buffer_head.
141  */
142 #define EXT4_MAP_NEW            (1 << BH_New)
143 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
144 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
145 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
146 #define EXT4_MAP_UNINIT         (1 << BH_Uninit)
147 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
148                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
149                                  EXT4_MAP_UNINIT)
150
151 struct ext4_map_blocks {
152         ext4_fsblk_t m_pblk;
153         ext4_lblk_t m_lblk;
154         unsigned int m_len;
155         unsigned int m_flags;
156 };
157
158 /*
159  * For delayed allocation tracking
160  */
161 struct mpage_da_data {
162         struct inode *inode;
163         sector_t b_blocknr;             /* start block number of extent */
164         size_t b_size;                  /* size of extent */
165         unsigned long b_state;          /* state of the extent */
166         unsigned long first_page, next_page;    /* extent of pages */
167         struct writeback_control *wbc;
168         int io_done;
169         int pages_written;
170         int retval;
171 };
172
173 /*
174  * Flags for ext4_io_end->flags
175  */
176 #define EXT4_IO_END_UNWRITTEN   0x0001
177 #define EXT4_IO_END_ERROR       0x0002
178 #define EXT4_IO_END_QUEUED      0x0004
179
180 struct ext4_io_page {
181         struct page     *p_page;
182         atomic_t        p_count;
183 };
184
185 #define MAX_IO_PAGES 128
186
187 typedef struct ext4_io_end {
188         struct list_head        list;           /* per-file finished IO list */
189         struct inode            *inode;         /* file being written to */
190         unsigned int            flag;           /* unwritten or not */
191         struct page             *page;          /* page struct for buffer write */
192         loff_t                  offset;         /* offset in the file */
193         ssize_t                 size;           /* size of the extent */
194         struct work_struct      work;           /* data work queue */
195         struct kiocb            *iocb;          /* iocb struct for AIO */
196         int                     result;         /* error value for AIO */
197         int                     num_io_pages;
198         struct ext4_io_page     *pages[MAX_IO_PAGES];
199 } ext4_io_end_t;
200
201 struct ext4_io_submit {
202         int                     io_op;
203         struct bio              *io_bio;
204         ext4_io_end_t           *io_end;
205         struct ext4_io_page     *io_page;
206         sector_t                io_next_block;
207 };
208
209 /*
210  * Special inodes numbers
211  */
212 #define EXT4_BAD_INO             1      /* Bad blocks inode */
213 #define EXT4_ROOT_INO            2      /* Root inode */
214 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
215 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
216 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
217 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
218 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
219 #define EXT4_JOURNAL_INO         8      /* Journal inode */
220
221 /* First non-reserved inode for old ext4 filesystems */
222 #define EXT4_GOOD_OLD_FIRST_INO 11
223
224 /*
225  * Maximal count of links to a file
226  */
227 #define EXT4_LINK_MAX           65000
228
229 /*
230  * Macro-instructions used to manage several block sizes
231  */
232 #define EXT4_MIN_BLOCK_SIZE             1024
233 #define EXT4_MAX_BLOCK_SIZE             65536
234 #define EXT4_MIN_BLOCK_LOG_SIZE         10
235 #define EXT4_MAX_BLOCK_LOG_SIZE         16
236 #ifdef __KERNEL__
237 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
238 #else
239 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
240 #endif
241 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
242 #ifdef __KERNEL__
243 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
244 #else
245 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
246 #endif
247 #ifdef __KERNEL__
248 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
249 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
250 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
251 #else
252 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
253                                  EXT4_GOOD_OLD_INODE_SIZE : \
254                                  (s)->s_inode_size)
255 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
256                                  EXT4_GOOD_OLD_FIRST_INO : \
257                                  (s)->s_first_ino)
258 #endif
259 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
260
261 /*
262  * Structure of a blocks group descriptor
263  */
264 struct ext4_group_desc
265 {
266         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
267         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
268         __le32  bg_inode_table_lo;      /* Inodes table block */
269         __le16  bg_free_blocks_count_lo;/* Free blocks count */
270         __le16  bg_free_inodes_count_lo;/* Free inodes count */
271         __le16  bg_used_dirs_count_lo;  /* Directories count */
272         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
273         __u32   bg_reserved[2];         /* Likely block/inode bitmap checksum */
274         __le16  bg_itable_unused_lo;    /* Unused inodes count */
275         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
276         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
277         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
278         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
279         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
280         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
281         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
282         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
283         __u32   bg_reserved2[3];
284 };
285
286 /*
287  * Structure of a flex block group info
288  */
289
290 struct flex_groups {
291         atomic_t free_inodes;
292         atomic_t free_blocks;
293         atomic_t used_dirs;
294 };
295
296 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
297 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
298 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
299
300 /*
301  * Macro-instructions used to manage group descriptors
302  */
303 #define EXT4_MIN_DESC_SIZE              32
304 #define EXT4_MIN_DESC_SIZE_64BIT        64
305 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
306 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
307 #ifdef __KERNEL__
308 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
309 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
310 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
311 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
312 #else
313 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
314 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
315 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
316 #endif
317
318 /*
319  * Constants relative to the data blocks
320  */
321 #define EXT4_NDIR_BLOCKS                12
322 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
323 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
324 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
325 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
326
327 /*
328  * Inode flags
329  */
330 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
331 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
332 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
333 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
334 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
335 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
336 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
337 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
338 /* Reserved for compression usage... */
339 #define EXT4_DIRTY_FL                   0x00000100
340 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
341 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
342 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
343 /* End compression flags --- maybe not all used */
344 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
345 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
346 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
347 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
348 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
349 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
350 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
351 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
352 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
353 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
354 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
355
356 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
357 #define EXT4_FL_USER_MODIFIABLE         0x004B80FF /* User modifiable flags */
358
359 /* Flags that should be inherited by new inodes from their parent. */
360 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
361                            EXT4_SYNC_FL | EXT4_IMMUTABLE_FL | EXT4_APPEND_FL |\
362                            EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
363                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
364                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
365
366 /* Flags that are appropriate for regular files (all but dir-specific ones). */
367 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
368
369 /* Flags that are appropriate for non-directories/regular files. */
370 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
371
372 /* Mask out flags that are inappropriate for the given type of inode. */
373 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
374 {
375         if (S_ISDIR(mode))
376                 return flags;
377         else if (S_ISREG(mode))
378                 return flags & EXT4_REG_FLMASK;
379         else
380                 return flags & EXT4_OTHER_FLMASK;
381 }
382
383 /*
384  * Inode flags used for atomic set/get
385  */
386 enum {
387         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
388         EXT4_INODE_UNRM         = 1,    /* Undelete */
389         EXT4_INODE_COMPR        = 2,    /* Compress file */
390         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
391         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
392         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
393         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
394         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
395 /* Reserved for compression usage... */
396         EXT4_INODE_DIRTY        = 8,
397         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
398         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
399         EXT4_INODE_ECOMPR       = 11,   /* Compression error */
400 /* End compression flags --- maybe not all used */
401         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
402         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
403         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
404         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
405         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
406         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
407         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
408         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
409         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
410         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
411         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
412 };
413
414 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
415 #define CHECK_FLAG_VALUE(FLAG) if (!TEST_FLAG_VALUE(FLAG)) { \
416         printk(KERN_EMERG "EXT4 flag fail: " #FLAG ": %d %d\n", \
417                 EXT4_##FLAG##_FL, EXT4_INODE_##FLAG); BUG_ON(1); }
418
419 /*
420  * Since it's pretty easy to mix up bit numbers and hex values, and we
421  * can't do a compile-time test for ENUM values, we use a run-time
422  * test to make sure that EXT4_XXX_FL is consistent with respect to
423  * EXT4_INODE_XXX.  If all is well the printk and BUG_ON will all drop
424  * out so it won't cost any extra space in the compiled kernel image.
425  * But it's important that these values are the same, since we are
426  * using EXT4_INODE_XXX to test for the flag values, but EXT4_XX_FL
427  * must be consistent with the values of FS_XXX_FL defined in
428  * include/linux/fs.h and the on-disk values found in ext2, ext3, and
429  * ext4 filesystems, and of course the values defined in e2fsprogs.
430  *
431  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
432  */
433 static inline void ext4_check_flag_values(void)
434 {
435         CHECK_FLAG_VALUE(SECRM);
436         CHECK_FLAG_VALUE(UNRM);
437         CHECK_FLAG_VALUE(COMPR);
438         CHECK_FLAG_VALUE(SYNC);
439         CHECK_FLAG_VALUE(IMMUTABLE);
440         CHECK_FLAG_VALUE(APPEND);
441         CHECK_FLAG_VALUE(NODUMP);
442         CHECK_FLAG_VALUE(NOATIME);
443         CHECK_FLAG_VALUE(DIRTY);
444         CHECK_FLAG_VALUE(COMPRBLK);
445         CHECK_FLAG_VALUE(NOCOMPR);
446         CHECK_FLAG_VALUE(ECOMPR);
447         CHECK_FLAG_VALUE(INDEX);
448         CHECK_FLAG_VALUE(IMAGIC);
449         CHECK_FLAG_VALUE(JOURNAL_DATA);
450         CHECK_FLAG_VALUE(NOTAIL);
451         CHECK_FLAG_VALUE(DIRSYNC);
452         CHECK_FLAG_VALUE(TOPDIR);
453         CHECK_FLAG_VALUE(HUGE_FILE);
454         CHECK_FLAG_VALUE(EXTENTS);
455         CHECK_FLAG_VALUE(EA_INODE);
456         CHECK_FLAG_VALUE(EOFBLOCKS);
457         CHECK_FLAG_VALUE(RESERVED);
458 }
459
460 /* Used to pass group descriptor data when online resize is done */
461 struct ext4_new_group_input {
462         __u32 group;            /* Group number for this data */
463         __u64 block_bitmap;     /* Absolute block number of block bitmap */
464         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
465         __u64 inode_table;      /* Absolute block number of inode table start */
466         __u32 blocks_count;     /* Total number of blocks in this group */
467         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
468         __u16 unused;
469 };
470
471 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
472 struct compat_ext4_new_group_input {
473         u32 group;
474         compat_u64 block_bitmap;
475         compat_u64 inode_bitmap;
476         compat_u64 inode_table;
477         u32 blocks_count;
478         u16 reserved_blocks;
479         u16 unused;
480 };
481 #endif
482
483 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
484 struct ext4_new_group_data {
485         __u32 group;
486         __u64 block_bitmap;
487         __u64 inode_bitmap;
488         __u64 inode_table;
489         __u32 blocks_count;
490         __u16 reserved_blocks;
491         __u16 unused;
492         __u32 free_blocks_count;
493 };
494
495 /*
496  * Flags used by ext4_map_blocks()
497  */
498         /* Allocate any needed blocks and/or convert an unitialized
499            extent to be an initialized ext4 */
500 #define EXT4_GET_BLOCKS_CREATE                  0x0001
501         /* Request the creation of an unitialized extent */
502 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
503 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
504                                                  EXT4_GET_BLOCKS_CREATE)
505         /* Caller is from the delayed allocation writeout path,
506            so set the magic i_delalloc_reserve_flag after taking the
507            inode allocation semaphore for */
508 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
509         /* caller is from the direct IO path, request to creation of an
510         unitialized extents if not allocated, split the uninitialized
511         extent if blocks has been preallocated already*/
512 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
513 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
514 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
515                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
516         /* Convert extent to initialized after IO complete */
517 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
518                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
519         /* Punch out blocks of an extent */
520 #define EXT4_GET_BLOCKS_PUNCH_OUT_EXT           0x0020
521         /* Don't normalize allocation size (used for fallocate) */
522 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
523
524 /*
525  * Flags used by ext4_free_blocks
526  */
527 #define EXT4_FREE_BLOCKS_METADATA       0x0001
528 #define EXT4_FREE_BLOCKS_FORGET         0x0002
529 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
530 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
531
532 /*
533  * ioctl commands
534  */
535 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
536 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
537 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
538 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
539 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
540 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
541 #ifdef CONFIG_JBD2_DEBUG
542 #define EXT4_IOC_WAIT_FOR_READONLY      _IOR('f', 99, long)
543 #endif
544 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
545 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
546 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
547 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
548 #define EXT4_IOC_MIGRATE                _IO('f', 9)
549  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
550  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
551 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
552 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
553
554 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
555 /*
556  * ioctl commands in 32 bit emulation
557  */
558 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
559 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
560 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
561 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
562 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
563 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
564 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
565 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
566 #ifdef CONFIG_JBD2_DEBUG
567 #define EXT4_IOC32_WAIT_FOR_READONLY    _IOR('f', 99, int)
568 #endif
569 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
570 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
571 #endif
572
573 /* Max physical block we can address w/o extents */
574 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
575
576 /*
577  * Structure of an inode on the disk
578  */
579 struct ext4_inode {
580         __le16  i_mode;         /* File mode */
581         __le16  i_uid;          /* Low 16 bits of Owner Uid */
582         __le32  i_size_lo;      /* Size in bytes */
583         __le32  i_atime;        /* Access time */
584         __le32  i_ctime;        /* Inode Change time */
585         __le32  i_mtime;        /* Modification time */
586         __le32  i_dtime;        /* Deletion Time */
587         __le16  i_gid;          /* Low 16 bits of Group Id */
588         __le16  i_links_count;  /* Links count */
589         __le32  i_blocks_lo;    /* Blocks count */
590         __le32  i_flags;        /* File flags */
591         union {
592                 struct {
593                         __le32  l_i_version;
594                 } linux1;
595                 struct {
596                         __u32  h_i_translator;
597                 } hurd1;
598                 struct {
599                         __u32  m_i_reserved1;
600                 } masix1;
601         } osd1;                         /* OS dependent 1 */
602         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
603         __le32  i_generation;   /* File version (for NFS) */
604         __le32  i_file_acl_lo;  /* File ACL */
605         __le32  i_size_high;
606         __le32  i_obso_faddr;   /* Obsoleted fragment address */
607         union {
608                 struct {
609                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
610                         __le16  l_i_file_acl_high;
611                         __le16  l_i_uid_high;   /* these 2 fields */
612                         __le16  l_i_gid_high;   /* were reserved2[0] */
613                         __u32   l_i_reserved2;
614                 } linux2;
615                 struct {
616                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
617                         __u16   h_i_mode_high;
618                         __u16   h_i_uid_high;
619                         __u16   h_i_gid_high;
620                         __u32   h_i_author;
621                 } hurd2;
622                 struct {
623                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
624                         __le16  m_i_file_acl_high;
625                         __u32   m_i_reserved2[2];
626                 } masix2;
627         } osd2;                         /* OS dependent 2 */
628         __le16  i_extra_isize;
629         __le16  i_pad1;
630         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
631         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
632         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
633         __le32  i_crtime;       /* File Creation time */
634         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
635         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
636 };
637
638 struct move_extent {
639         __u32 reserved;         /* should be zero */
640         __u32 donor_fd;         /* donor file descriptor */
641         __u64 orig_start;       /* logical start offset in block for orig */
642         __u64 donor_start;      /* logical start offset in block for donor */
643         __u64 len;              /* block length to be moved */
644         __u64 moved_len;        /* moved block length */
645 };
646
647 #define EXT4_EPOCH_BITS 2
648 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
649 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
650
651 /*
652  * Extended fields will fit into an inode if the filesystem was formatted
653  * with large inodes (-I 256 or larger) and there are not currently any EAs
654  * consuming all of the available space. For new inodes we always reserve
655  * enough space for the kernel's known extended fields, but for inodes
656  * created with an old kernel this might not have been the case. None of
657  * the extended inode fields is critical for correct filesystem operation.
658  * This macro checks if a certain field fits in the inode. Note that
659  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
660  */
661 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
662         ((offsetof(typeof(*ext4_inode), field) +        \
663           sizeof((ext4_inode)->field))                  \
664         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
665             (einode)->i_extra_isize))                   \
666
667 static inline __le32 ext4_encode_extra_time(struct timespec *time)
668 {
669        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
670                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
671                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
672 }
673
674 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
675 {
676        if (sizeof(time->tv_sec) > 4)
677                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
678                                << 32;
679        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
680 }
681
682 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
683 do {                                                                           \
684         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
685         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
686                 (raw_inode)->xtime ## _extra =                                 \
687                                 ext4_encode_extra_time(&(inode)->xtime);       \
688 } while (0)
689
690 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
691 do {                                                                           \
692         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
693                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
694         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
695                 (raw_inode)->xtime ## _extra =                                 \
696                                 ext4_encode_extra_time(&(einode)->xtime);      \
697 } while (0)
698
699 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
700 do {                                                                           \
701         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
702         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
703                 ext4_decode_extra_time(&(inode)->xtime,                        \
704                                        raw_inode->xtime ## _extra);            \
705         else                                                                   \
706                 (inode)->xtime.tv_nsec = 0;                                    \
707 } while (0)
708
709 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
710 do {                                                                           \
711         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
712                 (einode)->xtime.tv_sec =                                       \
713                         (signed)le32_to_cpu((raw_inode)->xtime);               \
714         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
715                 ext4_decode_extra_time(&(einode)->xtime,                       \
716                                        raw_inode->xtime ## _extra);            \
717         else                                                                   \
718                 (einode)->xtime.tv_nsec = 0;                                   \
719 } while (0)
720
721 #define i_disk_version osd1.linux1.l_i_version
722
723 #if defined(__KERNEL__) || defined(__linux__)
724 #define i_reserved1     osd1.linux1.l_i_reserved1
725 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
726 #define i_blocks_high   osd2.linux2.l_i_blocks_high
727 #define i_uid_low       i_uid
728 #define i_gid_low       i_gid
729 #define i_uid_high      osd2.linux2.l_i_uid_high
730 #define i_gid_high      osd2.linux2.l_i_gid_high
731 #define i_reserved2     osd2.linux2.l_i_reserved2
732
733 #elif defined(__GNU__)
734
735 #define i_translator    osd1.hurd1.h_i_translator
736 #define i_uid_high      osd2.hurd2.h_i_uid_high
737 #define i_gid_high      osd2.hurd2.h_i_gid_high
738 #define i_author        osd2.hurd2.h_i_author
739
740 #elif defined(__masix__)
741
742 #define i_reserved1     osd1.masix1.m_i_reserved1
743 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
744 #define i_reserved2     osd2.masix2.m_i_reserved2
745
746 #endif /* defined(__KERNEL__) || defined(__linux__) */
747
748 /*
749  * storage for cached extent
750  * If ec_len == 0, then the cache is invalid.
751  * If ec_start == 0, then the cache represents a gap (null mapping)
752  */
753 struct ext4_ext_cache {
754         ext4_fsblk_t    ec_start;
755         ext4_lblk_t     ec_block;
756         __u32           ec_len; /* must be 32bit to return holes */
757 };
758
759 /*
760  * fourth extended file system inode data in memory
761  */
762 struct ext4_inode_info {
763         __le32  i_data[15];     /* unconverted */
764         __u32   i_dtime;
765         ext4_fsblk_t    i_file_acl;
766
767         /*
768          * i_block_group is the number of the block group which contains
769          * this file's inode.  Constant across the lifetime of the inode,
770          * it is ued for making block allocation decisions - we try to
771          * place a file's data blocks near its inode block, and new inodes
772          * near to their parent directory's inode.
773          */
774         ext4_group_t    i_block_group;
775         ext4_lblk_t     i_dir_start_lookup;
776 #if (BITS_PER_LONG < 64)
777         unsigned long   i_state_flags;          /* Dynamic state flags */
778 #endif
779         unsigned long   i_flags;
780
781 #ifdef CONFIG_EXT4_FS_XATTR
782         /*
783          * Extended attributes can be read independently of the main file
784          * data. Taking i_mutex even when reading would cause contention
785          * between readers of EAs and writers of regular file data, so
786          * instead we synchronize on xattr_sem when reading or changing
787          * EAs.
788          */
789         struct rw_semaphore xattr_sem;
790 #endif
791
792         struct list_head i_orphan;      /* unlinked but open inodes */
793
794         /*
795          * i_disksize keeps track of what the inode size is ON DISK, not
796          * in memory.  During truncate, i_size is set to the new size by
797          * the VFS prior to calling ext4_truncate(), but the filesystem won't
798          * set i_disksize to 0 until the truncate is actually under way.
799          *
800          * The intent is that i_disksize always represents the blocks which
801          * are used by this file.  This allows recovery to restart truncate
802          * on orphans if we crash during truncate.  We actually write i_disksize
803          * into the on-disk inode when writing inodes out, instead of i_size.
804          *
805          * The only time when i_disksize and i_size may be different is when
806          * a truncate is in progress.  The only things which change i_disksize
807          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
808          */
809         loff_t  i_disksize;
810
811         /*
812          * i_data_sem is for serialising ext4_truncate() against
813          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
814          * data tree are chopped off during truncate. We can't do that in
815          * ext4 because whenever we perform intermediate commits during
816          * truncate, the inode and all the metadata blocks *must* be in a
817          * consistent state which allows truncation of the orphans to restart
818          * during recovery.  Hence we must fix the get_block-vs-truncate race
819          * by other means, so we have i_data_sem.
820          */
821         struct rw_semaphore i_data_sem;
822         struct inode vfs_inode;
823         struct jbd2_inode *jinode;
824
825         struct ext4_ext_cache i_cached_extent;
826         /*
827          * File creation time. Its function is same as that of
828          * struct timespec i_{a,c,m}time in the generic inode.
829          */
830         struct timespec i_crtime;
831
832         /* mballoc */
833         struct list_head i_prealloc_list;
834         spinlock_t i_prealloc_lock;
835
836         /* ialloc */
837         ext4_group_t    i_last_alloc_group;
838
839         /* allocation reservation info for delalloc */
840         unsigned int i_reserved_data_blocks;
841         unsigned int i_reserved_meta_blocks;
842         unsigned int i_allocated_meta_blocks;
843         ext4_lblk_t i_da_metadata_calc_last_lblock;
844         int i_da_metadata_calc_len;
845
846         /* on-disk additional length */
847         __u16 i_extra_isize;
848
849 #ifdef CONFIG_QUOTA
850         /* quota space reservation, managed internally by quota code */
851         qsize_t i_reserved_quota;
852 #endif
853
854         /* completed IOs that might need unwritten extents handling */
855         struct list_head i_completed_io_list;
856         spinlock_t i_completed_io_lock;
857         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
858         /* current io_end structure for async DIO write*/
859         ext4_io_end_t *cur_aio_dio;
860         atomic_t i_aiodio_unwritten; /* Nr. of inflight conversions pending */
861
862         spinlock_t i_block_reservation_lock;
863
864         /*
865          * Transactions that contain inode's metadata needed to complete
866          * fsync and fdatasync, respectively.
867          */
868         tid_t i_sync_tid;
869         tid_t i_datasync_tid;
870 };
871
872 /*
873  * File system states
874  */
875 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
876 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
877 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
878
879 /*
880  * Misc. filesystem flags
881  */
882 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
883 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
884 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
885
886 /*
887  * Mount flags
888  */
889 #define EXT4_MOUNT_OLDALLOC             0x00002  /* Don't use the new Orlov allocator */
890 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
891 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
892 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
893 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
894 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
895 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
896 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
897 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
898 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
899 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
900 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
901 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
902 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
903 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
904 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
905 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
906 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
907 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
908 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
909 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
910 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
911 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
912 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
913 #define EXT4_MOUNT_I_VERSION            0x2000000 /* i_version support */
914 #define EXT4_MOUNT_MBLK_IO_SUBMIT       0x4000000 /* multi-block io submits */
915 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
916 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
917 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
918 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
919 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
920
921 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
922                                                 ~EXT4_MOUNT_##opt
923 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
924                                                 EXT4_MOUNT_##opt
925 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
926                                          EXT4_MOUNT_##opt)
927
928 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
929                                                 ~EXT4_MOUNT2_##opt
930 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
931                                                 EXT4_MOUNT2_##opt
932 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
933                                          EXT4_MOUNT2_##opt)
934
935 #define ext4_set_bit                    __test_and_set_bit_le
936 #define ext4_set_bit_atomic             ext2_set_bit_atomic
937 #define ext4_clear_bit                  __test_and_clear_bit_le
938 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
939 #define ext4_test_bit                   test_bit_le
940 #define ext4_find_first_zero_bit        find_first_zero_bit_le
941 #define ext4_find_next_zero_bit         find_next_zero_bit_le
942 #define ext4_find_next_bit              find_next_bit_le
943
944 extern void ext4_set_bits(void *bm, int cur, int len);
945
946 /*
947  * Maximal mount counts between two filesystem checks
948  */
949 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
950 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
951
952 /*
953  * Behaviour when detecting errors
954  */
955 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
956 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
957 #define EXT4_ERRORS_PANIC               3       /* Panic */
958 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
959
960 /*
961  * Structure of the super block
962  */
963 struct ext4_super_block {
964 /*00*/  __le32  s_inodes_count;         /* Inodes count */
965         __le32  s_blocks_count_lo;      /* Blocks count */
966         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
967         __le32  s_free_blocks_count_lo; /* Free blocks count */
968 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
969         __le32  s_first_data_block;     /* First Data Block */
970         __le32  s_log_block_size;       /* Block size */
971         __le32  s_obso_log_frag_size;   /* Obsoleted fragment size */
972 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
973         __le32  s_obso_frags_per_group; /* Obsoleted fragments per group */
974         __le32  s_inodes_per_group;     /* # Inodes per group */
975         __le32  s_mtime;                /* Mount time */
976 /*30*/  __le32  s_wtime;                /* Write time */
977         __le16  s_mnt_count;            /* Mount count */
978         __le16  s_max_mnt_count;        /* Maximal mount count */
979         __le16  s_magic;                /* Magic signature */
980         __le16  s_state;                /* File system state */
981         __le16  s_errors;               /* Behaviour when detecting errors */
982         __le16  s_minor_rev_level;      /* minor revision level */
983 /*40*/  __le32  s_lastcheck;            /* time of last check */
984         __le32  s_checkinterval;        /* max. time between checks */
985         __le32  s_creator_os;           /* OS */
986         __le32  s_rev_level;            /* Revision level */
987 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
988         __le16  s_def_resgid;           /* Default gid for reserved blocks */
989         /*
990          * These fields are for EXT4_DYNAMIC_REV superblocks only.
991          *
992          * Note: the difference between the compatible feature set and
993          * the incompatible feature set is that if there is a bit set
994          * in the incompatible feature set that the kernel doesn't
995          * know about, it should refuse to mount the filesystem.
996          *
997          * e2fsck's requirements are more strict; if it doesn't know
998          * about a feature in either the compatible or incompatible
999          * feature set, it must abort and not try to meddle with
1000          * things it doesn't understand...
1001          */
1002         __le32  s_first_ino;            /* First non-reserved inode */
1003         __le16  s_inode_size;           /* size of inode structure */
1004         __le16  s_block_group_nr;       /* block group # of this superblock */
1005         __le32  s_feature_compat;       /* compatible feature set */
1006 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1007         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1008 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1009 /*78*/  char    s_volume_name[16];      /* volume name */
1010 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1011 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1012         /*
1013          * Performance hints.  Directory preallocation should only
1014          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1015          */
1016         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1017         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1018         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1019         /*
1020          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1021          */
1022 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1023 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1024         __le32  s_journal_dev;          /* device number of journal file */
1025         __le32  s_last_orphan;          /* start of list of inodes to delete */
1026         __le32  s_hash_seed[4];         /* HTREE hash seed */
1027         __u8    s_def_hash_version;     /* Default hash version to use */
1028         __u8    s_jnl_backup_type;
1029         __le16  s_desc_size;            /* size of group descriptor */
1030 /*100*/ __le32  s_default_mount_opts;
1031         __le32  s_first_meta_bg;        /* First metablock block group */
1032         __le32  s_mkfs_time;            /* When the filesystem was created */
1033         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1034         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1035 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1036         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1037         __le32  s_free_blocks_count_hi; /* Free blocks count */
1038         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1039         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1040         __le32  s_flags;                /* Miscellaneous flags */
1041         __le16  s_raid_stride;          /* RAID stride */
1042         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1043         __le64  s_mmp_block;            /* Block for multi-mount protection */
1044         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1045         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1046         __u8    s_reserved_char_pad;
1047         __le16  s_reserved_pad;
1048         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1049         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1050         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1051         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1052                                               snapshot's future use */
1053         __le32  s_snapshot_list;        /* inode number of the head of the
1054                                            on-disk snapshot list */
1055 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1056         __le32  s_error_count;          /* number of fs errors */
1057         __le32  s_first_error_time;     /* first time an error happened */
1058         __le32  s_first_error_ino;      /* inode involved in first error */
1059         __le64  s_first_error_block;    /* block involved of first error */
1060         __u8    s_first_error_func[32]; /* function where the error happened */
1061         __le32  s_first_error_line;     /* line number where error happened */
1062         __le32  s_last_error_time;      /* most recent time of an error */
1063         __le32  s_last_error_ino;       /* inode involved in last error */
1064         __le32  s_last_error_line;      /* line number where error happened */
1065         __le64  s_last_error_block;     /* block involved of last error */
1066         __u8    s_last_error_func[32];  /* function where the error happened */
1067 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1068         __u8    s_mount_opts[64];
1069         __le32  s_reserved[112];        /* Padding to the end of the block */
1070 };
1071
1072 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1073
1074 #ifdef __KERNEL__
1075
1076 /*
1077  * run-time mount flags
1078  */
1079 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
1080 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
1081
1082 /*
1083  * fourth extended-fs super-block data in memory
1084  */
1085 struct ext4_sb_info {
1086         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1087         unsigned long s_inodes_per_block;/* Number of inodes per block */
1088         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1089         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1090         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1091         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1092         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1093         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1094         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1095         unsigned long s_overhead_last;  /* Last calculated overhead */
1096         unsigned long s_blocks_last;    /* Last seen block count */
1097         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1098         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1099         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1100         struct buffer_head **s_group_desc;
1101         unsigned int s_mount_opt;
1102         unsigned int s_mount_opt2;
1103         unsigned int s_mount_flags;
1104         ext4_fsblk_t s_sb_block;
1105         uid_t s_resuid;
1106         gid_t s_resgid;
1107         unsigned short s_mount_state;
1108         unsigned short s_pad;
1109         int s_addr_per_block_bits;
1110         int s_desc_per_block_bits;
1111         int s_inode_size;
1112         int s_first_ino;
1113         unsigned int s_inode_readahead_blks;
1114         unsigned int s_inode_goal;
1115         spinlock_t s_next_gen_lock;
1116         u32 s_next_generation;
1117         u32 s_hash_seed[4];
1118         int s_def_hash_version;
1119         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1120         struct percpu_counter s_freeblocks_counter;
1121         struct percpu_counter s_freeinodes_counter;
1122         struct percpu_counter s_dirs_counter;
1123         struct percpu_counter s_dirtyblocks_counter;
1124         struct blockgroup_lock *s_blockgroup_lock;
1125         struct proc_dir_entry *s_proc;
1126         struct kobject s_kobj;
1127         struct completion s_kobj_unregister;
1128
1129         /* Journaling */
1130         struct journal_s *s_journal;
1131         struct list_head s_orphan;
1132         struct mutex s_orphan_lock;
1133         unsigned long s_resize_flags;           /* Flags indicating if there
1134                                                    is a resizer */
1135         unsigned long s_commit_interval;
1136         u32 s_max_batch_time;
1137         u32 s_min_batch_time;
1138         struct block_device *journal_bdev;
1139 #ifdef CONFIG_JBD2_DEBUG
1140         struct timer_list turn_ro_timer;        /* For turning read-only (crash simulation) */
1141         wait_queue_head_t ro_wait_queue;        /* For people waiting for the fs to go read-only */
1142 #endif
1143 #ifdef CONFIG_QUOTA
1144         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
1145         int s_jquota_fmt;                       /* Format of quota to use */
1146 #endif
1147         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1148         struct rb_root system_blks;
1149
1150 #ifdef EXTENTS_STATS
1151         /* ext4 extents stats */
1152         unsigned long s_ext_min;
1153         unsigned long s_ext_max;
1154         unsigned long s_depth_max;
1155         spinlock_t s_ext_stats_lock;
1156         unsigned long s_ext_blocks;
1157         unsigned long s_ext_extents;
1158 #endif
1159         /* ext4 extent cache stats */
1160         unsigned long extent_cache_hits;
1161         unsigned long extent_cache_misses;
1162
1163         /* for buddy allocator */
1164         struct ext4_group_info ***s_group_info;
1165         struct inode *s_buddy_cache;
1166         spinlock_t s_md_lock;
1167         unsigned short *s_mb_offsets;
1168         unsigned int *s_mb_maxs;
1169
1170         /* tunables */
1171         unsigned long s_stripe;
1172         unsigned int s_mb_stream_request;
1173         unsigned int s_mb_max_to_scan;
1174         unsigned int s_mb_min_to_scan;
1175         unsigned int s_mb_stats;
1176         unsigned int s_mb_order2_reqs;
1177         unsigned int s_mb_group_prealloc;
1178         unsigned int s_max_writeback_mb_bump;
1179         /* where last allocation was done - for stream allocation */
1180         unsigned long s_mb_last_group;
1181         unsigned long s_mb_last_start;
1182
1183         /* stats for buddy allocator */
1184         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1185         atomic_t s_bal_success; /* we found long enough chunks */
1186         atomic_t s_bal_allocated;       /* in blocks */
1187         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1188         atomic_t s_bal_goals;   /* goal hits */
1189         atomic_t s_bal_breaks;  /* too long searches */
1190         atomic_t s_bal_2orders; /* 2^order hits */
1191         spinlock_t s_bal_lock;
1192         unsigned long s_mb_buddies_generated;
1193         unsigned long long s_mb_generation_time;
1194         atomic_t s_mb_lost_chunks;
1195         atomic_t s_mb_preallocated;
1196         atomic_t s_mb_discarded;
1197         atomic_t s_lock_busy;
1198
1199         /* locality groups */
1200         struct ext4_locality_group __percpu *s_locality_groups;
1201
1202         /* for write statistics */
1203         unsigned long s_sectors_written_start;
1204         u64 s_kbytes_written;
1205
1206         unsigned int s_log_groups_per_flex;
1207         struct flex_groups *s_flex_groups;
1208
1209         /* workqueue for dio unwritten */
1210         struct workqueue_struct *dio_unwritten_wq;
1211
1212         /* timer for periodic error stats printing */
1213         struct timer_list s_err_report;
1214
1215         /* Lazy inode table initialization info */
1216         struct ext4_li_request *s_li_request;
1217         /* Wait multiplier for lazy initialization thread */
1218         unsigned int s_li_wait_mult;
1219
1220         /* Kernel thread for multiple mount protection */
1221         struct task_struct *s_mmp_tsk;
1222
1223         /* record the last minlen when FITRIM is called. */
1224         atomic_t s_last_trim_minblks;
1225 };
1226
1227 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1228 {
1229         return sb->s_fs_info;
1230 }
1231 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1232 {
1233         return container_of(inode, struct ext4_inode_info, vfs_inode);
1234 }
1235
1236 static inline struct timespec ext4_current_time(struct inode *inode)
1237 {
1238         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1239                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1240 }
1241
1242 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1243 {
1244         return ino == EXT4_ROOT_INO ||
1245                 ino == EXT4_JOURNAL_INO ||
1246                 ino == EXT4_RESIZE_INO ||
1247                 (ino >= EXT4_FIRST_INO(sb) &&
1248                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1249 }
1250
1251 /*
1252  * Inode dynamic state flags
1253  */
1254 enum {
1255         EXT4_STATE_JDATA,               /* journaled data exists */
1256         EXT4_STATE_NEW,                 /* inode is newly created */
1257         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1258         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1259         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1260         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1261         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1262         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1263         EXT4_STATE_DELALLOC_RESERVED,   /* blks already reserved for delalloc */
1264 };
1265
1266 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1267 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1268 {                                                                       \
1269         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1270 }                                                                       \
1271 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1272 {                                                                       \
1273         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1274 }                                                                       \
1275 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1276 {                                                                       \
1277         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1278 }
1279
1280 EXT4_INODE_BIT_FNS(flag, flags, 0)
1281 #if (BITS_PER_LONG < 64)
1282 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1283
1284 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1285 {
1286         (ei)->i_state_flags = 0;
1287 }
1288 #else
1289 EXT4_INODE_BIT_FNS(state, flags, 32)
1290
1291 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1292 {
1293         /* We depend on the fact that callers will set i_flags */
1294 }
1295 #endif
1296 #else
1297 /* Assume that user mode programs are passing in an ext4fs superblock, not
1298  * a kernel struct super_block.  This will allow us to call the feature-test
1299  * macros from user land. */
1300 #define EXT4_SB(sb)     (sb)
1301 #endif
1302
1303 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1304
1305 /*
1306  * Codes for operating systems
1307  */
1308 #define EXT4_OS_LINUX           0
1309 #define EXT4_OS_HURD            1
1310 #define EXT4_OS_MASIX           2
1311 #define EXT4_OS_FREEBSD         3
1312 #define EXT4_OS_LITES           4
1313
1314 /*
1315  * Revision levels
1316  */
1317 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1318 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1319
1320 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1321 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1322
1323 #define EXT4_GOOD_OLD_INODE_SIZE 128
1324
1325 /*
1326  * Feature set definitions
1327  */
1328
1329 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1330         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1331 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1332         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1333 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1334         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1335 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1336         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1337 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1338         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1339 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1340         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1341 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1342         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1343 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1344         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1345 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1346         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1347
1348 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1349 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1350 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1351 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1352 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1353 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1354
1355 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1356 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1357 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1358 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1359 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1360 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1361 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1362 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1363
1364 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1365 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1366 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1367 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1368 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1369 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1370 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1371 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1372 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1373 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1374 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1375
1376 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1377 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1378                                          EXT4_FEATURE_INCOMPAT_META_BG)
1379 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1380                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1381                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1382
1383 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1384 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1385                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1386                                          EXT4_FEATURE_INCOMPAT_META_BG)
1387 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1388                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1389                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1390
1391 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1392 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1393                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1394                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1395                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1396                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1397                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1398                                          EXT4_FEATURE_INCOMPAT_MMP)
1399 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1400                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1401                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1402                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1403                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1404                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1405                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE)
1406
1407 /*
1408  * Default values for user and/or group using reserved blocks
1409  */
1410 #define EXT4_DEF_RESUID         0
1411 #define EXT4_DEF_RESGID         0
1412
1413 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1414
1415 /*
1416  * Default mount options
1417  */
1418 #define EXT4_DEFM_DEBUG         0x0001
1419 #define EXT4_DEFM_BSDGROUPS     0x0002
1420 #define EXT4_DEFM_XATTR_USER    0x0004
1421 #define EXT4_DEFM_ACL           0x0008
1422 #define EXT4_DEFM_UID16         0x0010
1423 #define EXT4_DEFM_JMODE         0x0060
1424 #define EXT4_DEFM_JMODE_DATA    0x0020
1425 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1426 #define EXT4_DEFM_JMODE_WBACK   0x0060
1427 #define EXT4_DEFM_NOBARRIER     0x0100
1428 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1429 #define EXT4_DEFM_DISCARD       0x0400
1430 #define EXT4_DEFM_NODELALLOC    0x0800
1431
1432 /*
1433  * Default journal batch times
1434  */
1435 #define EXT4_DEF_MIN_BATCH_TIME 0
1436 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1437
1438 /*
1439  * Minimum number of groups in a flexgroup before we separate out
1440  * directories into the first block group of a flexgroup
1441  */
1442 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1443
1444 /*
1445  * Structure of a directory entry
1446  */
1447 #define EXT4_NAME_LEN 255
1448
1449 struct ext4_dir_entry {
1450         __le32  inode;                  /* Inode number */
1451         __le16  rec_len;                /* Directory entry length */
1452         __le16  name_len;               /* Name length */
1453         char    name[EXT4_NAME_LEN];    /* File name */
1454 };
1455
1456 /*
1457  * The new version of the directory entry.  Since EXT4 structures are
1458  * stored in intel byte order, and the name_len field could never be
1459  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1460  * file_type field.
1461  */
1462 struct ext4_dir_entry_2 {
1463         __le32  inode;                  /* Inode number */
1464         __le16  rec_len;                /* Directory entry length */
1465         __u8    name_len;               /* Name length */
1466         __u8    file_type;
1467         char    name[EXT4_NAME_LEN];    /* File name */
1468 };
1469
1470 /*
1471  * Ext4 directory file types.  Only the low 3 bits are used.  The
1472  * other bits are reserved for now.
1473  */
1474 #define EXT4_FT_UNKNOWN         0
1475 #define EXT4_FT_REG_FILE        1
1476 #define EXT4_FT_DIR             2
1477 #define EXT4_FT_CHRDEV          3
1478 #define EXT4_FT_BLKDEV          4
1479 #define EXT4_FT_FIFO            5
1480 #define EXT4_FT_SOCK            6
1481 #define EXT4_FT_SYMLINK         7
1482
1483 #define EXT4_FT_MAX             8
1484
1485 /*
1486  * EXT4_DIR_PAD defines the directory entries boundaries
1487  *
1488  * NOTE: It must be a multiple of 4
1489  */
1490 #define EXT4_DIR_PAD                    4
1491 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1492 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1493                                          ~EXT4_DIR_ROUND)
1494 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1495
1496 /*
1497  * If we ever get support for fs block sizes > page_size, we'll need
1498  * to remove the #if statements in the next two functions...
1499  */
1500 static inline unsigned int
1501 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1502 {
1503         unsigned len = le16_to_cpu(dlen);
1504
1505 #if (PAGE_CACHE_SIZE >= 65536)
1506         if (len == EXT4_MAX_REC_LEN || len == 0)
1507                 return blocksize;
1508         return (len & 65532) | ((len & 3) << 16);
1509 #else
1510         return len;
1511 #endif
1512 }
1513
1514 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1515 {
1516         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1517                 BUG();
1518 #if (PAGE_CACHE_SIZE >= 65536)
1519         if (len < 65536)
1520                 return cpu_to_le16(len);
1521         if (len == blocksize) {
1522                 if (blocksize == 65536)
1523                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1524                 else
1525                         return cpu_to_le16(0);
1526         }
1527         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1528 #else
1529         return cpu_to_le16(len);
1530 #endif
1531 }
1532
1533 /*
1534  * Hash Tree Directory indexing
1535  * (c) Daniel Phillips, 2001
1536  */
1537
1538 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1539                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1540                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1541 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1542 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1543
1544 /* Legal values for the dx_root hash_version field: */
1545
1546 #define DX_HASH_LEGACY          0
1547 #define DX_HASH_HALF_MD4        1
1548 #define DX_HASH_TEA             2
1549 #define DX_HASH_LEGACY_UNSIGNED 3
1550 #define DX_HASH_HALF_MD4_UNSIGNED       4
1551 #define DX_HASH_TEA_UNSIGNED            5
1552
1553 #ifdef __KERNEL__
1554
1555 /* hash info structure used by the directory hash */
1556 struct dx_hash_info
1557 {
1558         u32             hash;
1559         u32             minor_hash;
1560         int             hash_version;
1561         u32             *seed;
1562 };
1563
1564 #define EXT4_HTREE_EOF  0x7fffffff
1565
1566 /*
1567  * Control parameters used by ext4_htree_next_block
1568  */
1569 #define HASH_NB_ALWAYS          1
1570
1571
1572 /*
1573  * Describe an inode's exact location on disk and in memory
1574  */
1575 struct ext4_iloc
1576 {
1577         struct buffer_head *bh;
1578         unsigned long offset;
1579         ext4_group_t block_group;
1580 };
1581
1582 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1583 {
1584         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1585 }
1586
1587 /*
1588  * This structure is stuffed into the struct file's private_data field
1589  * for directories.  It is where we put information so that we can do
1590  * readdir operations in hash tree order.
1591  */
1592 struct dir_private_info {
1593         struct rb_root  root;
1594         struct rb_node  *curr_node;
1595         struct fname    *extra_fname;
1596         loff_t          last_pos;
1597         __u32           curr_hash;
1598         __u32           curr_minor_hash;
1599         __u32           next_hash;
1600 };
1601
1602 /* calculate the first block number of the group */
1603 static inline ext4_fsblk_t
1604 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1605 {
1606         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1607                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1608 }
1609
1610 /*
1611  * Special error return code only used by dx_probe() and its callers.
1612  */
1613 #define ERR_BAD_DX_DIR  -75000
1614
1615 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1616                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1617
1618 /*
1619  * Timeout and state flag for lazy initialization inode thread.
1620  */
1621 #define EXT4_DEF_LI_WAIT_MULT                   10
1622 #define EXT4_DEF_LI_MAX_START_DELAY             5
1623 #define EXT4_LAZYINIT_QUIT                      0x0001
1624 #define EXT4_LAZYINIT_RUNNING                   0x0002
1625
1626 /*
1627  * Lazy inode table initialization info
1628  */
1629 struct ext4_lazy_init {
1630         unsigned long           li_state;
1631         struct list_head        li_request_list;
1632         struct mutex            li_list_mtx;
1633 };
1634
1635 struct ext4_li_request {
1636         struct super_block      *lr_super;
1637         struct ext4_sb_info     *lr_sbi;
1638         ext4_group_t            lr_next_group;
1639         struct list_head        lr_request;
1640         unsigned long           lr_next_sched;
1641         unsigned long           lr_timeout;
1642 };
1643
1644 struct ext4_features {
1645         struct kobject f_kobj;
1646         struct completion f_kobj_unregister;
1647 };
1648
1649 /*
1650  * This structure will be used for multiple mount protection. It will be
1651  * written into the block number saved in the s_mmp_block field in the
1652  * superblock. Programs that check MMP should assume that if
1653  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1654  * to use the filesystem, regardless of how old the timestamp is.
1655  */
1656 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
1657 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1658 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
1659 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
1660
1661 struct mmp_struct {
1662         __le32  mmp_magic;              /* Magic number for MMP */
1663         __le32  mmp_seq;                /* Sequence no. updated periodically */
1664
1665         /*
1666          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1667          * purposes and do not affect the correctness of the algorithm
1668          */
1669         __le64  mmp_time;               /* Time last updated */
1670         char    mmp_nodename[64];       /* Node which last updated MMP block */
1671         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
1672
1673         /*
1674          * mmp_check_interval is used to verify if the MMP block has been
1675          * updated on the block device. The value is updated based on the
1676          * maximum time to write the MMP block during an update cycle.
1677          */
1678         __le16  mmp_check_interval;
1679
1680         __le16  mmp_pad1;
1681         __le32  mmp_pad2[227];
1682 };
1683
1684 /* arguments passed to the mmp thread */
1685 struct mmpd_data {
1686         struct buffer_head *bh; /* bh from initial read_mmp_block() */
1687         struct super_block *sb;  /* super block of the fs */
1688 };
1689
1690 /*
1691  * Check interval multiplier
1692  * The MMP block is written every update interval and initially checked every
1693  * update interval x the multiplier (the value is then adapted based on the
1694  * write latency). The reason is that writes can be delayed under load and we
1695  * don't want readers to incorrectly assume that the filesystem is no longer
1696  * in use.
1697  */
1698 #define EXT4_MMP_CHECK_MULT             2UL
1699
1700 /*
1701  * Minimum interval for MMP checking in seconds.
1702  */
1703 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
1704
1705 /*
1706  * Maximum interval for MMP checking in seconds.
1707  */
1708 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
1709
1710 /*
1711  * Function prototypes
1712  */
1713
1714 /*
1715  * Ok, these declarations are also in <linux/kernel.h> but none of the
1716  * ext4 source programs needs to include it so they are duplicated here.
1717  */
1718 # define NORET_TYPE     /**/
1719 # define ATTRIB_NORET   __attribute__((noreturn))
1720 # define NORET_AND      noreturn,
1721
1722 /* bitmap.c */
1723 extern unsigned int ext4_count_free(struct buffer_head *, unsigned);
1724
1725 /* balloc.c */
1726 extern unsigned int ext4_block_group(struct super_block *sb,
1727                         ext4_fsblk_t blocknr);
1728 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1729                         ext4_fsblk_t blocknr);
1730 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1731 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1732                         ext4_group_t group);
1733 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1734                                          ext4_fsblk_t goal,
1735                                          unsigned int flags,
1736                                          unsigned long *count,
1737                                          int *errp);
1738 extern int ext4_claim_free_blocks(struct ext4_sb_info *sbi,
1739                                   s64 nblocks, unsigned int flags);
1740 extern ext4_fsblk_t ext4_count_free_blocks(struct super_block *);
1741 extern void ext4_check_blocks_bitmap(struct super_block *);
1742 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1743                                                     ext4_group_t block_group,
1744                                                     struct buffer_head ** bh);
1745 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1746 struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1747                                       ext4_group_t block_group);
1748 extern unsigned ext4_init_block_bitmap(struct super_block *sb,
1749                                        struct buffer_head *bh,
1750                                        ext4_group_t group,
1751                                        struct ext4_group_desc *desc);
1752 #define ext4_free_blocks_after_init(sb, group, desc)                    \
1753                 ext4_init_block_bitmap(sb, NULL, group, desc)
1754 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
1755
1756 /* dir.c */
1757 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1758                                   struct file *,
1759                                   struct ext4_dir_entry_2 *,
1760                                   struct buffer_head *, unsigned int);
1761 #define ext4_check_dir_entry(dir, filp, de, bh, offset)                 \
1762         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
1763                                         (de), (bh), (offset)))
1764 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1765                                     __u32 minor_hash,
1766                                     struct ext4_dir_entry_2 *dirent);
1767 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1768
1769 /* fsync.c */
1770 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
1771 extern int ext4_flush_completed_IO(struct inode *);
1772
1773 /* hash.c */
1774 extern int ext4fs_dirhash(const char *name, int len, struct
1775                           dx_hash_info *hinfo);
1776
1777 /* ialloc.c */
1778 extern struct inode *ext4_new_inode(handle_t *, struct inode *, int,
1779                                     const struct qstr *qstr, __u32 goal);
1780 extern void ext4_free_inode(handle_t *, struct inode *);
1781 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1782 extern unsigned long ext4_count_free_inodes(struct super_block *);
1783 extern unsigned long ext4_count_dirs(struct super_block *);
1784 extern void ext4_check_inodes_bitmap(struct super_block *);
1785 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1786 extern int ext4_init_inode_table(struct super_block *sb,
1787                                  ext4_group_t group, int barrier);
1788
1789 /* mballoc.c */
1790 extern long ext4_mb_stats;
1791 extern long ext4_mb_max_to_scan;
1792 extern int ext4_mb_init(struct super_block *, int);
1793 extern int ext4_mb_release(struct super_block *);
1794 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1795                                 struct ext4_allocation_request *, int *);
1796 extern int ext4_mb_reserve_blocks(struct super_block *, int);
1797 extern void ext4_discard_preallocations(struct inode *);
1798 extern int __init ext4_init_mballoc(void);
1799 extern void ext4_exit_mballoc(void);
1800 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1801                              struct buffer_head *bh, ext4_fsblk_t block,
1802                              unsigned long count, int flags);
1803 extern int ext4_mb_add_groupinfo(struct super_block *sb,
1804                 ext4_group_t i, struct ext4_group_desc *desc);
1805 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
1806                                 ext4_fsblk_t block, unsigned long count);
1807 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
1808
1809 /* inode.c */
1810 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1811                                                 ext4_lblk_t, int, int *);
1812 struct buffer_head *ext4_bread(handle_t *, struct inode *,
1813                                                 ext4_lblk_t, int, int *);
1814 int ext4_get_block(struct inode *inode, sector_t iblock,
1815                                 struct buffer_head *bh_result, int create);
1816
1817 extern struct inode *ext4_iget(struct super_block *, unsigned long);
1818 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
1819 extern int  ext4_setattr(struct dentry *, struct iattr *);
1820 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1821                                 struct kstat *stat);
1822 extern void ext4_evict_inode(struct inode *);
1823 extern void ext4_clear_inode(struct inode *);
1824 extern int  ext4_sync_inode(handle_t *, struct inode *);
1825 extern void ext4_dirty_inode(struct inode *, int);
1826 extern int ext4_change_inode_journal_flag(struct inode *, int);
1827 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1828 extern int ext4_can_truncate(struct inode *inode);
1829 extern void ext4_truncate(struct inode *);
1830 extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length);
1831 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
1832 extern void ext4_set_inode_flags(struct inode *);
1833 extern void ext4_get_inode_flags(struct ext4_inode_info *);
1834 extern int ext4_alloc_da_blocks(struct inode *inode);
1835 extern void ext4_set_aops(struct inode *inode);
1836 extern int ext4_writepage_trans_blocks(struct inode *);
1837 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1838 extern int ext4_block_truncate_page(handle_t *handle,
1839                 struct address_space *mapping, loff_t from);
1840 extern int ext4_block_zero_page_range(handle_t *handle,
1841                 struct address_space *mapping, loff_t from, loff_t length);
1842 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1843 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
1844 extern void ext4_da_update_reserve_space(struct inode *inode,
1845                                         int used, int quota_claim);
1846
1847 /* indirect.c */
1848 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
1849                                 struct ext4_map_blocks *map, int flags);
1850 extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
1851                                 const struct iovec *iov, loff_t offset,
1852                                 unsigned long nr_segs);
1853 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
1854 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks, int chunk);
1855 extern void ext4_ind_truncate(struct inode *inode);
1856
1857 /* ioctl.c */
1858 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1859 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1860
1861 /* migrate.c */
1862 extern int ext4_ext_migrate(struct inode *);
1863
1864 /* namei.c */
1865 extern int ext4_orphan_add(handle_t *, struct inode *);
1866 extern int ext4_orphan_del(handle_t *, struct inode *);
1867 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1868                                 __u32 start_minor_hash, __u32 *next_hash);
1869
1870 /* resize.c */
1871 extern int ext4_group_add(struct super_block *sb,
1872                                 struct ext4_new_group_data *input);
1873 extern int ext4_group_extend(struct super_block *sb,
1874                                 struct ext4_super_block *es,
1875                                 ext4_fsblk_t n_blocks_count);
1876
1877 /* super.c */
1878 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
1879 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
1880 extern void ext4_kvfree(void *ptr);
1881 extern void __ext4_error(struct super_block *, const char *, unsigned int,
1882                          const char *, ...)
1883         __attribute__ ((format (printf, 4, 5)));
1884 #define ext4_error(sb, message...)      __ext4_error(sb, __func__,      \
1885                                                      __LINE__, ## message)
1886 extern void ext4_error_inode(struct inode *, const char *, unsigned int,
1887                              ext4_fsblk_t, const char *, ...)
1888         __attribute__ ((format (printf, 5, 6)));
1889 extern void ext4_error_file(struct file *, const char *, unsigned int,
1890                             ext4_fsblk_t, const char *, ...)
1891         __attribute__ ((format (printf, 5, 6)));
1892 extern void __ext4_std_error(struct super_block *, const char *,
1893                              unsigned int, int);
1894 extern void __ext4_abort(struct super_block *, const char *, unsigned int,
1895                        const char *, ...)
1896         __attribute__ ((format (printf, 4, 5)));
1897 #define ext4_abort(sb, message...)      __ext4_abort(sb, __func__, \
1898                                                        __LINE__, ## message)
1899 extern void __ext4_warning(struct super_block *, const char *, unsigned int,
1900                           const char *, ...)
1901         __attribute__ ((format (printf, 4, 5)));
1902 #define ext4_warning(sb, message...)    __ext4_warning(sb, __func__, \
1903                                                        __LINE__, ## message)
1904 extern void ext4_msg(struct super_block *, const char *, const char *, ...)
1905         __attribute__ ((format (printf, 3, 4)));
1906 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
1907                            const char *, unsigned int, const char *);
1908 #define dump_mmp_msg(sb, mmp, msg)      __dump_mmp_msg(sb, mmp, __func__, \
1909                                                        __LINE__, msg)
1910 extern void __ext4_grp_locked_error(const char *, unsigned int, \
1911                                     struct super_block *, ext4_group_t, \
1912                                     unsigned long, ext4_fsblk_t, \
1913                                     const char *, ...)
1914         __attribute__ ((format (printf, 7, 8)));
1915 #define ext4_grp_locked_error(sb, grp, message...) \
1916         __ext4_grp_locked_error(__func__, __LINE__, (sb), (grp), ## message)
1917 extern void ext4_update_dynamic_rev(struct super_block *sb);
1918 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1919                                         __u32 compat);
1920 extern int ext4_update_rocompat_feature(handle_t *handle,
1921                                         struct super_block *sb, __u32 rocompat);
1922 extern int ext4_update_incompat_feature(handle_t *handle,
1923                                         struct super_block *sb, __u32 incompat);
1924 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1925                                       struct ext4_group_desc *bg);
1926 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1927                                       struct ext4_group_desc *bg);
1928 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1929                                      struct ext4_group_desc *bg);
1930 extern __u32 ext4_free_blks_count(struct super_block *sb,
1931                                 struct ext4_group_desc *bg);
1932 extern __u32 ext4_free_inodes_count(struct super_block *sb,
1933                                  struct ext4_group_desc *bg);
1934 extern __u32 ext4_used_dirs_count(struct super_block *sb,
1935                                 struct ext4_group_desc *bg);
1936 extern __u32 ext4_itable_unused_count(struct super_block *sb,
1937                                    struct ext4_group_desc *bg);
1938 extern void ext4_block_bitmap_set(struct super_block *sb,
1939                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1940 extern void ext4_inode_bitmap_set(struct super_block *sb,
1941                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1942 extern void ext4_inode_table_set(struct super_block *sb,
1943                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
1944 extern void ext4_free_blks_set(struct super_block *sb,
1945                                struct ext4_group_desc *bg, __u32 count);
1946 extern void ext4_free_inodes_set(struct super_block *sb,
1947                                 struct ext4_group_desc *bg, __u32 count);
1948 extern void ext4_used_dirs_set(struct super_block *sb,
1949                                 struct ext4_group_desc *bg, __u32 count);
1950 extern void ext4_itable_unused_set(struct super_block *sb,
1951                                    struct ext4_group_desc *bg, __u32 count);
1952 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
1953                                    struct ext4_group_desc *gdp);
1954 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
1955                                        struct ext4_group_desc *gdp);
1956
1957 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
1958 {
1959         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
1960                 le32_to_cpu(es->s_blocks_count_lo);
1961 }
1962
1963 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
1964 {
1965         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
1966                 le32_to_cpu(es->s_r_blocks_count_lo);
1967 }
1968
1969 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
1970 {
1971         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
1972                 le32_to_cpu(es->s_free_blocks_count_lo);
1973 }
1974
1975 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
1976                                          ext4_fsblk_t blk)
1977 {
1978         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
1979         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
1980 }
1981
1982 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
1983                                               ext4_fsblk_t blk)
1984 {
1985         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
1986         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
1987 }
1988
1989 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
1990                                            ext4_fsblk_t blk)
1991 {
1992         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
1993         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
1994 }
1995
1996 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
1997 {
1998         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
1999                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2000                         le32_to_cpu(raw_inode->i_size_lo);
2001         else
2002                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2003 }
2004
2005 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2006 {
2007         raw_inode->i_size_lo = cpu_to_le32(i_size);
2008         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2009 }
2010
2011 static inline
2012 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2013                                             ext4_group_t group)
2014 {
2015          struct ext4_group_info ***grp_info;
2016          long indexv, indexh;
2017          grp_info = EXT4_SB(sb)->s_group_info;
2018          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2019          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2020          return grp_info[indexv][indexh];
2021 }
2022
2023 /*
2024  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2025  * the locking protocol documented in the comments of ext4_group_add()
2026  * in resize.c
2027  */
2028 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2029 {
2030         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2031
2032         smp_rmb();
2033         return ngroups;
2034 }
2035
2036 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2037                                              ext4_group_t block_group)
2038 {
2039         return block_group >> sbi->s_log_groups_per_flex;
2040 }
2041
2042 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2043 {
2044         return 1 << sbi->s_log_groups_per_flex;
2045 }
2046
2047 #define ext4_std_error(sb, errno)                               \
2048 do {                                                            \
2049         if ((errno))                                            \
2050                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2051 } while (0)
2052
2053 #ifdef CONFIG_SMP
2054 /* Each CPU can accumulate percpu_counter_batch blocks in their local
2055  * counters. So we need to make sure we have free blocks more
2056  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2057  */
2058 #define EXT4_FREEBLOCKS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2059 #else
2060 #define EXT4_FREEBLOCKS_WATERMARK 0
2061 #endif
2062
2063 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2064 {
2065         /*
2066          * XXX: replace with spinlock if seen contended -bzzz
2067          */
2068         down_write(&EXT4_I(inode)->i_data_sem);
2069         if (newsize > EXT4_I(inode)->i_disksize)
2070                 EXT4_I(inode)->i_disksize = newsize;
2071         up_write(&EXT4_I(inode)->i_data_sem);
2072         return ;
2073 }
2074
2075 struct ext4_group_info {
2076         unsigned long   bb_state;
2077         struct rb_root  bb_free_root;
2078         ext4_grpblk_t   bb_first_free;  /* first free block */
2079         ext4_grpblk_t   bb_free;        /* total free blocks */
2080         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2081         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2082         struct          list_head bb_prealloc_list;
2083 #ifdef DOUBLE_CHECK
2084         void            *bb_bitmap;
2085 #endif
2086         struct rw_semaphore alloc_sem;
2087         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2088                                          * regions, index is order.
2089                                          * bb_counters[3] = 5 means
2090                                          * 5 free 8-block regions. */
2091 };
2092
2093 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2094 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2095
2096 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2097         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2098
2099 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2100         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2101 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2102         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2103 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2104         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2105
2106 #define EXT4_MAX_CONTENTION             8
2107 #define EXT4_CONTENTION_THRESHOLD       2
2108
2109 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2110                                               ext4_group_t group)
2111 {
2112         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2113 }
2114
2115 /*
2116  * Returns true if the filesystem is busy enough that attempts to
2117  * access the block group locks has run into contention.
2118  */
2119 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2120 {
2121         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2122 }
2123
2124 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2125 {
2126         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2127         if (spin_trylock(lock))
2128                 /*
2129                  * We're able to grab the lock right away, so drop the
2130                  * lock contention counter.
2131                  */
2132                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2133         else {
2134                 /*
2135                  * The lock is busy, so bump the contention counter,
2136                  * and then wait on the spin lock.
2137                  */
2138                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2139                                   EXT4_MAX_CONTENTION);
2140                 spin_lock(lock);
2141         }
2142 }
2143
2144 static inline void ext4_unlock_group(struct super_block *sb,
2145                                         ext4_group_t group)
2146 {
2147         spin_unlock(ext4_group_lock_ptr(sb, group));
2148 }
2149
2150 static inline void ext4_mark_super_dirty(struct super_block *sb)
2151 {
2152         if (EXT4_SB(sb)->s_journal == NULL)
2153                 sb->s_dirt =1;
2154 }
2155
2156 /*
2157  * Block validity checking
2158  */
2159 #define ext4_check_indirect_blockref(inode, bh)                         \
2160         ext4_check_blockref(__func__, __LINE__, inode,                  \
2161                             (__le32 *)(bh)->b_data,                     \
2162                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2163
2164 #define ext4_ind_check_inode(inode)                                     \
2165         ext4_check_blockref(__func__, __LINE__, inode,                  \
2166                             EXT4_I(inode)->i_data,                      \
2167                             EXT4_NDIR_BLOCKS)
2168
2169 /*
2170  * Inodes and files operations
2171  */
2172
2173 /* dir.c */
2174 extern const struct file_operations ext4_dir_operations;
2175
2176 /* file.c */
2177 extern const struct inode_operations ext4_file_inode_operations;
2178 extern const struct file_operations ext4_file_operations;
2179 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2180
2181 /* namei.c */
2182 extern const struct inode_operations ext4_dir_inode_operations;
2183 extern const struct inode_operations ext4_special_inode_operations;
2184 extern struct dentry *ext4_get_parent(struct dentry *child);
2185
2186 /* symlink.c */
2187 extern const struct inode_operations ext4_symlink_inode_operations;
2188 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2189
2190 /* block_validity */
2191 extern void ext4_release_system_zone(struct super_block *sb);
2192 extern int ext4_setup_system_zone(struct super_block *sb);
2193 extern int __init ext4_init_system_zone(void);
2194 extern void ext4_exit_system_zone(void);
2195 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2196                                  ext4_fsblk_t start_blk,
2197                                  unsigned int count);
2198 extern int ext4_check_blockref(const char *, unsigned int,
2199                                struct inode *, __le32 *, unsigned int);
2200
2201 /* extents.c */
2202 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2203 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2204 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
2205                                        int chunk);
2206 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2207                                struct ext4_map_blocks *map, int flags);
2208 extern void ext4_ext_truncate(struct inode *);
2209 extern int ext4_ext_punch_hole(struct file *file, loff_t offset,
2210                                 loff_t length);
2211 extern void ext4_ext_init(struct super_block *);
2212 extern void ext4_ext_release(struct super_block *);
2213 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2214                           loff_t len);
2215 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
2216                           ssize_t len);
2217 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2218                            struct ext4_map_blocks *map, int flags);
2219 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2220                         __u64 start, __u64 len);
2221 /* move_extent.c */
2222 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2223                              __u64 start_orig, __u64 start_donor,
2224                              __u64 len, __u64 *moved_len);
2225
2226 /* page-io.c */
2227 extern int __init ext4_init_pageio(void);
2228 extern void ext4_exit_pageio(void);
2229 extern void ext4_ioend_wait(struct inode *);
2230 extern void ext4_free_io_end(ext4_io_end_t *io);
2231 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2232 extern int ext4_end_io_nolock(ext4_io_end_t *io);
2233 extern void ext4_io_submit(struct ext4_io_submit *io);
2234 extern int ext4_bio_write_page(struct ext4_io_submit *io,
2235                                struct page *page,
2236                                int len,
2237                                struct writeback_control *wbc);
2238
2239 /* mmp.c */
2240 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2241
2242 /* BH_Uninit flag: blocks are allocated but uninitialized on disk */
2243 enum ext4_state_bits {
2244         BH_Uninit       /* blocks are allocated but uninitialized on disk */
2245           = BH_JBDPrivateStart,
2246 };
2247
2248 BUFFER_FNS(Uninit, uninit)
2249 TAS_BUFFER_FNS(Uninit, uninit)
2250
2251 /*
2252  * Add new method to test wether block and inode bitmaps are properly
2253  * initialized. With uninit_bg reading the block from disk is not enough
2254  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2255  */
2256 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2257
2258 static inline int bitmap_uptodate(struct buffer_head *bh)
2259 {
2260         return (buffer_uptodate(bh) &&
2261                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2262 }
2263 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2264 {
2265         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2266 }
2267
2268 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2269
2270 /* For ioend & aio unwritten conversion wait queues */
2271 #define EXT4_WQ_HASH_SZ         37
2272 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
2273                                             EXT4_WQ_HASH_SZ])
2274 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
2275                                              EXT4_WQ_HASH_SZ])
2276 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2277 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2278
2279 #define EXT4_RESIZING   0
2280 extern int ext4_resize_begin(struct super_block *sb);
2281 extern void ext4_resize_end(struct super_block *sb);
2282
2283 #endif  /* __KERNEL__ */
2284
2285 #endif  /* _EXT4_H */