ext4: reserve inodes and feature code for 'quota' feature
[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
112
113 struct ext4_allocation_request {
114         /* target inode for block we're allocating */
115         struct inode *inode;
116         /* how many blocks we want to allocate */
117         unsigned int len;
118         /* logical block in target inode */
119         ext4_lblk_t logical;
120         /* the closest logical allocated block to the left */
121         ext4_lblk_t lleft;
122         /* the closest logical allocated block to the right */
123         ext4_lblk_t lright;
124         /* phys. target (a hint) */
125         ext4_fsblk_t goal;
126         /* phys. block for the closest logical allocated block to the left */
127         ext4_fsblk_t pleft;
128         /* phys. block for the closest logical allocated block to the right */
129         ext4_fsblk_t pright;
130         /* flags. see above EXT4_MB_HINT_* */
131         unsigned int flags;
132 };
133
134 /*
135  * Logical to physical block mapping, used by ext4_map_blocks()
136  *
137  * This structure is used to pass requests into ext4_map_blocks() as
138  * well as to store the information returned by ext4_map_blocks().  It
139  * takes less room on the stack than a struct buffer_head.
140  */
141 #define EXT4_MAP_NEW            (1 << BH_New)
142 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
143 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
144 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
145 #define EXT4_MAP_UNINIT         (1 << BH_Uninit)
146 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
147                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
148                                  EXT4_MAP_UNINIT)
149
150 struct ext4_map_blocks {
151         ext4_fsblk_t m_pblk;
152         ext4_lblk_t m_lblk;
153         unsigned int m_len;
154         unsigned int m_flags;
155 };
156
157 /*
158  * For delayed allocation tracking
159  */
160 struct mpage_da_data {
161         struct inode *inode;
162         sector_t b_blocknr;             /* start block number of extent */
163         size_t b_size;                  /* size of extent */
164         unsigned long b_state;          /* state of the extent */
165         unsigned long first_page, next_page;    /* extent of pages */
166         struct writeback_control *wbc;
167         int io_done;
168         int pages_written;
169         int retval;
170 };
171
172 /*
173  * Flags for ext4_io_end->flags
174  */
175 #define EXT4_IO_END_UNWRITTEN   0x0001
176 #define EXT4_IO_END_ERROR       0x0002
177
178 struct ext4_io_page {
179         struct page     *p_page;
180         atomic_t        p_count;
181 };
182
183 #define MAX_IO_PAGES 128
184
185 typedef struct ext4_io_end {
186         struct list_head        list;           /* per-file finished IO list */
187         struct inode            *inode;         /* file being written to */
188         unsigned int            flag;           /* unwritten or not */
189         struct page             *page;          /* page struct for buffer write */
190         loff_t                  offset;         /* offset in the file */
191         ssize_t                 size;           /* size of the extent */
192         struct work_struct      work;           /* data work queue */
193         struct kiocb            *iocb;          /* iocb struct for AIO */
194         int                     result;         /* error value for AIO */
195         int                     num_io_pages;
196         struct ext4_io_page     *pages[MAX_IO_PAGES];
197 } ext4_io_end_t;
198
199 struct ext4_io_submit {
200         int                     io_op;
201         struct bio              *io_bio;
202         ext4_io_end_t           *io_end;
203         struct ext4_io_page     *io_page;
204         sector_t                io_next_block;
205 };
206
207 /*
208  * Special inodes numbers
209  */
210 #define EXT4_BAD_INO             1      /* Bad blocks inode */
211 #define EXT4_ROOT_INO            2      /* Root inode */
212 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
213 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
214 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
215 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
216 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
217 #define EXT4_JOURNAL_INO         8      /* Journal inode */
218
219 /* First non-reserved inode for old ext4 filesystems */
220 #define EXT4_GOOD_OLD_FIRST_INO 11
221
222 /*
223  * Maximal count of links to a file
224  */
225 #define EXT4_LINK_MAX           65000
226
227 /*
228  * Macro-instructions used to manage several block sizes
229  */
230 #define EXT4_MIN_BLOCK_SIZE             1024
231 #define EXT4_MAX_BLOCK_SIZE             65536
232 #define EXT4_MIN_BLOCK_LOG_SIZE         10
233 #define EXT4_MAX_BLOCK_LOG_SIZE         16
234 #ifdef __KERNEL__
235 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
236 #else
237 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
238 #endif
239 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
240 #ifdef __KERNEL__
241 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
242 #else
243 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
244 #endif
245 #ifdef __KERNEL__
246 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
247 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
248 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
249 #else
250 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
251                                  EXT4_GOOD_OLD_INODE_SIZE : \
252                                  (s)->s_inode_size)
253 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
254                                  EXT4_GOOD_OLD_FIRST_INO : \
255                                  (s)->s_first_ino)
256 #endif
257 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
258
259 /*
260  * Structure of a blocks group descriptor
261  */
262 struct ext4_group_desc
263 {
264         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
265         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
266         __le32  bg_inode_table_lo;      /* Inodes table block */
267         __le16  bg_free_blocks_count_lo;/* Free blocks count */
268         __le16  bg_free_inodes_count_lo;/* Free inodes count */
269         __le16  bg_used_dirs_count_lo;  /* Directories count */
270         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
271         __u32   bg_reserved[2];         /* Likely block/inode bitmap checksum */
272         __le16  bg_itable_unused_lo;    /* Unused inodes count */
273         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
274         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
275         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
276         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
277         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
278         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
279         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
280         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
281         __u32   bg_reserved2[3];
282 };
283
284 /*
285  * Structure of a flex block group info
286  */
287
288 struct flex_groups {
289         atomic_t free_inodes;
290         atomic_t free_blocks;
291         atomic_t used_dirs;
292 };
293
294 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
295 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
296 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
297
298 /*
299  * Macro-instructions used to manage group descriptors
300  */
301 #define EXT4_MIN_DESC_SIZE              32
302 #define EXT4_MIN_DESC_SIZE_64BIT        64
303 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
304 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
305 #ifdef __KERNEL__
306 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
307 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
308 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
309 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
310 #else
311 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
312 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
313 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
314 #endif
315
316 /*
317  * Constants relative to the data blocks
318  */
319 #define EXT4_NDIR_BLOCKS                12
320 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
321 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
322 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
323 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
324
325 /*
326  * Inode flags
327  */
328 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
329 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
330 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
331 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
332 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
333 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
334 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
335 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
336 /* Reserved for compression usage... */
337 #define EXT4_DIRTY_FL                   0x00000100
338 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
339 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
340 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
341 /* End compression flags --- maybe not all used */
342 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
343 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
344 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
345 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
346 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
347 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
348 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
349 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
350 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
351 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
352 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
353
354 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
355 #define EXT4_FL_USER_MODIFIABLE         0x004B80FF /* User modifiable flags */
356
357 /* Flags that should be inherited by new inodes from their parent. */
358 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
359                            EXT4_SYNC_FL | EXT4_IMMUTABLE_FL | EXT4_APPEND_FL |\
360                            EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
361                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
362                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
363
364 /* Flags that are appropriate for regular files (all but dir-specific ones). */
365 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
366
367 /* Flags that are appropriate for non-directories/regular files. */
368 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
369
370 /* Mask out flags that are inappropriate for the given type of inode. */
371 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
372 {
373         if (S_ISDIR(mode))
374                 return flags;
375         else if (S_ISREG(mode))
376                 return flags & EXT4_REG_FLMASK;
377         else
378                 return flags & EXT4_OTHER_FLMASK;
379 }
380
381 /*
382  * Inode flags used for atomic set/get
383  */
384 enum {
385         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
386         EXT4_INODE_UNRM         = 1,    /* Undelete */
387         EXT4_INODE_COMPR        = 2,    /* Compress file */
388         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
389         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
390         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
391         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
392         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
393 /* Reserved for compression usage... */
394         EXT4_INODE_DIRTY        = 8,
395         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
396         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
397         EXT4_INODE_ECOMPR       = 11,   /* Compression error */
398 /* End compression flags --- maybe not all used */
399         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
400         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
401         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
402         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
403         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
404         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
405         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
406         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
407         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
408         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
409         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
410 };
411
412 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
413 #define CHECK_FLAG_VALUE(FLAG) if (!TEST_FLAG_VALUE(FLAG)) { \
414         printk(KERN_EMERG "EXT4 flag fail: " #FLAG ": %d %d\n", \
415                 EXT4_##FLAG##_FL, EXT4_INODE_##FLAG); BUG_ON(1); }
416
417 /*
418  * Since it's pretty easy to mix up bit numbers and hex values, and we
419  * can't do a compile-time test for ENUM values, we use a run-time
420  * test to make sure that EXT4_XXX_FL is consistent with respect to
421  * EXT4_INODE_XXX.  If all is well the printk and BUG_ON will all drop
422  * out so it won't cost any extra space in the compiled kernel image.
423  * But it's important that these values are the same, since we are
424  * using EXT4_INODE_XXX to test for the flag values, but EXT4_XX_FL
425  * must be consistent with the values of FS_XXX_FL defined in
426  * include/linux/fs.h and the on-disk values found in ext2, ext3, and
427  * ext4 filesystems, and of course the values defined in e2fsprogs.
428  *
429  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
430  */
431 static inline void ext4_check_flag_values(void)
432 {
433         CHECK_FLAG_VALUE(SECRM);
434         CHECK_FLAG_VALUE(UNRM);
435         CHECK_FLAG_VALUE(COMPR);
436         CHECK_FLAG_VALUE(SYNC);
437         CHECK_FLAG_VALUE(IMMUTABLE);
438         CHECK_FLAG_VALUE(APPEND);
439         CHECK_FLAG_VALUE(NODUMP);
440         CHECK_FLAG_VALUE(NOATIME);
441         CHECK_FLAG_VALUE(DIRTY);
442         CHECK_FLAG_VALUE(COMPRBLK);
443         CHECK_FLAG_VALUE(NOCOMPR);
444         CHECK_FLAG_VALUE(ECOMPR);
445         CHECK_FLAG_VALUE(INDEX);
446         CHECK_FLAG_VALUE(IMAGIC);
447         CHECK_FLAG_VALUE(JOURNAL_DATA);
448         CHECK_FLAG_VALUE(NOTAIL);
449         CHECK_FLAG_VALUE(DIRSYNC);
450         CHECK_FLAG_VALUE(TOPDIR);
451         CHECK_FLAG_VALUE(HUGE_FILE);
452         CHECK_FLAG_VALUE(EXTENTS);
453         CHECK_FLAG_VALUE(EA_INODE);
454         CHECK_FLAG_VALUE(EOFBLOCKS);
455         CHECK_FLAG_VALUE(RESERVED);
456 }
457
458 /* Used to pass group descriptor data when online resize is done */
459 struct ext4_new_group_input {
460         __u32 group;            /* Group number for this data */
461         __u64 block_bitmap;     /* Absolute block number of block bitmap */
462         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
463         __u64 inode_table;      /* Absolute block number of inode table start */
464         __u32 blocks_count;     /* Total number of blocks in this group */
465         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
466         __u16 unused;
467 };
468
469 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
470 struct compat_ext4_new_group_input {
471         u32 group;
472         compat_u64 block_bitmap;
473         compat_u64 inode_bitmap;
474         compat_u64 inode_table;
475         u32 blocks_count;
476         u16 reserved_blocks;
477         u16 unused;
478 };
479 #endif
480
481 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
482 struct ext4_new_group_data {
483         __u32 group;
484         __u64 block_bitmap;
485         __u64 inode_bitmap;
486         __u64 inode_table;
487         __u32 blocks_count;
488         __u16 reserved_blocks;
489         __u16 unused;
490         __u32 free_blocks_count;
491 };
492
493 /*
494  * Flags used by ext4_map_blocks()
495  */
496         /* Allocate any needed blocks and/or convert an unitialized
497            extent to be an initialized ext4 */
498 #define EXT4_GET_BLOCKS_CREATE                  0x0001
499         /* Request the creation of an unitialized extent */
500 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
501 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
502                                                  EXT4_GET_BLOCKS_CREATE)
503         /* Caller is from the delayed allocation writeout path,
504            so set the magic i_delalloc_reserve_flag after taking the
505            inode allocation semaphore for */
506 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
507         /* caller is from the direct IO path, request to creation of an
508         unitialized extents if not allocated, split the uninitialized
509         extent if blocks has been preallocated already*/
510 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
511 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
512 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
513                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
514         /* Convert extent to initialized after IO complete */
515 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
516                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
517
518 /*
519  * Flags used by ext4_free_blocks
520  */
521 #define EXT4_FREE_BLOCKS_METADATA       0x0001
522 #define EXT4_FREE_BLOCKS_FORGET         0x0002
523 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
524
525 /*
526  * ioctl commands
527  */
528 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
529 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
530 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
531 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
532 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
533 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
534 #ifdef CONFIG_JBD2_DEBUG
535 #define EXT4_IOC_WAIT_FOR_READONLY      _IOR('f', 99, long)
536 #endif
537 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
538 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
539 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
540 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
541 #define EXT4_IOC_MIGRATE                _IO('f', 9)
542  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
543  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
544 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
545 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
546
547 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
548 /*
549  * ioctl commands in 32 bit emulation
550  */
551 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
552 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
553 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
554 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
555 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
556 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
557 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
558 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
559 #ifdef CONFIG_JBD2_DEBUG
560 #define EXT4_IOC32_WAIT_FOR_READONLY    _IOR('f', 99, int)
561 #endif
562 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
563 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
564 #endif
565
566 /* Max physical block we can address w/o extents */
567 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
568
569 /*
570  * Structure of an inode on the disk
571  */
572 struct ext4_inode {
573         __le16  i_mode;         /* File mode */
574         __le16  i_uid;          /* Low 16 bits of Owner Uid */
575         __le32  i_size_lo;      /* Size in bytes */
576         __le32  i_atime;        /* Access time */
577         __le32  i_ctime;        /* Inode Change time */
578         __le32  i_mtime;        /* Modification time */
579         __le32  i_dtime;        /* Deletion Time */
580         __le16  i_gid;          /* Low 16 bits of Group Id */
581         __le16  i_links_count;  /* Links count */
582         __le32  i_blocks_lo;    /* Blocks count */
583         __le32  i_flags;        /* File flags */
584         union {
585                 struct {
586                         __le32  l_i_version;
587                 } linux1;
588                 struct {
589                         __u32  h_i_translator;
590                 } hurd1;
591                 struct {
592                         __u32  m_i_reserved1;
593                 } masix1;
594         } osd1;                         /* OS dependent 1 */
595         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
596         __le32  i_generation;   /* File version (for NFS) */
597         __le32  i_file_acl_lo;  /* File ACL */
598         __le32  i_size_high;
599         __le32  i_obso_faddr;   /* Obsoleted fragment address */
600         union {
601                 struct {
602                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
603                         __le16  l_i_file_acl_high;
604                         __le16  l_i_uid_high;   /* these 2 fields */
605                         __le16  l_i_gid_high;   /* were reserved2[0] */
606                         __u32   l_i_reserved2;
607                 } linux2;
608                 struct {
609                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
610                         __u16   h_i_mode_high;
611                         __u16   h_i_uid_high;
612                         __u16   h_i_gid_high;
613                         __u32   h_i_author;
614                 } hurd2;
615                 struct {
616                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
617                         __le16  m_i_file_acl_high;
618                         __u32   m_i_reserved2[2];
619                 } masix2;
620         } osd2;                         /* OS dependent 2 */
621         __le16  i_extra_isize;
622         __le16  i_pad1;
623         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
624         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
625         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
626         __le32  i_crtime;       /* File Creation time */
627         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
628         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
629 };
630
631 struct move_extent {
632         __u32 reserved;         /* should be zero */
633         __u32 donor_fd;         /* donor file descriptor */
634         __u64 orig_start;       /* logical start offset in block for orig */
635         __u64 donor_start;      /* logical start offset in block for donor */
636         __u64 len;              /* block length to be moved */
637         __u64 moved_len;        /* moved block length */
638 };
639
640 #define EXT4_EPOCH_BITS 2
641 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
642 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
643
644 /*
645  * Extended fields will fit into an inode if the filesystem was formatted
646  * with large inodes (-I 256 or larger) and there are not currently any EAs
647  * consuming all of the available space. For new inodes we always reserve
648  * enough space for the kernel's known extended fields, but for inodes
649  * created with an old kernel this might not have been the case. None of
650  * the extended inode fields is critical for correct filesystem operation.
651  * This macro checks if a certain field fits in the inode. Note that
652  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
653  */
654 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
655         ((offsetof(typeof(*ext4_inode), field) +        \
656           sizeof((ext4_inode)->field))                  \
657         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
658             (einode)->i_extra_isize))                   \
659
660 static inline __le32 ext4_encode_extra_time(struct timespec *time)
661 {
662        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
663                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
664                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
665 }
666
667 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
668 {
669        if (sizeof(time->tv_sec) > 4)
670                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
671                                << 32;
672        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
673 }
674
675 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
676 do {                                                                           \
677         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
678         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
679                 (raw_inode)->xtime ## _extra =                                 \
680                                 ext4_encode_extra_time(&(inode)->xtime);       \
681 } while (0)
682
683 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
684 do {                                                                           \
685         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
686                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
687         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
688                 (raw_inode)->xtime ## _extra =                                 \
689                                 ext4_encode_extra_time(&(einode)->xtime);      \
690 } while (0)
691
692 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
693 do {                                                                           \
694         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
695         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
696                 ext4_decode_extra_time(&(inode)->xtime,                        \
697                                        raw_inode->xtime ## _extra);            \
698         else                                                                   \
699                 (inode)->xtime.tv_nsec = 0;                                    \
700 } while (0)
701
702 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
703 do {                                                                           \
704         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
705                 (einode)->xtime.tv_sec =                                       \
706                         (signed)le32_to_cpu((raw_inode)->xtime);               \
707         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
708                 ext4_decode_extra_time(&(einode)->xtime,                       \
709                                        raw_inode->xtime ## _extra);            \
710         else                                                                   \
711                 (einode)->xtime.tv_nsec = 0;                                   \
712 } while (0)
713
714 #define i_disk_version osd1.linux1.l_i_version
715
716 #if defined(__KERNEL__) || defined(__linux__)
717 #define i_reserved1     osd1.linux1.l_i_reserved1
718 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
719 #define i_blocks_high   osd2.linux2.l_i_blocks_high
720 #define i_uid_low       i_uid
721 #define i_gid_low       i_gid
722 #define i_uid_high      osd2.linux2.l_i_uid_high
723 #define i_gid_high      osd2.linux2.l_i_gid_high
724 #define i_reserved2     osd2.linux2.l_i_reserved2
725
726 #elif defined(__GNU__)
727
728 #define i_translator    osd1.hurd1.h_i_translator
729 #define i_uid_high      osd2.hurd2.h_i_uid_high
730 #define i_gid_high      osd2.hurd2.h_i_gid_high
731 #define i_author        osd2.hurd2.h_i_author
732
733 #elif defined(__masix__)
734
735 #define i_reserved1     osd1.masix1.m_i_reserved1
736 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
737 #define i_reserved2     osd2.masix2.m_i_reserved2
738
739 #endif /* defined(__KERNEL__) || defined(__linux__) */
740
741 /*
742  * storage for cached extent
743  * If ec_len == 0, then the cache is invalid.
744  * If ec_start == 0, then the cache represents a gap (null mapping)
745  */
746 struct ext4_ext_cache {
747         ext4_fsblk_t    ec_start;
748         ext4_lblk_t     ec_block;
749         __u32           ec_len; /* must be 32bit to return holes */
750 };
751
752 /*
753  * fourth extended file system inode data in memory
754  */
755 struct ext4_inode_info {
756         __le32  i_data[15];     /* unconverted */
757         __u32   i_dtime;
758         ext4_fsblk_t    i_file_acl;
759
760         /*
761          * i_block_group is the number of the block group which contains
762          * this file's inode.  Constant across the lifetime of the inode,
763          * it is ued for making block allocation decisions - we try to
764          * place a file's data blocks near its inode block, and new inodes
765          * near to their parent directory's inode.
766          */
767         ext4_group_t    i_block_group;
768         ext4_lblk_t     i_dir_start_lookup;
769 #if (BITS_PER_LONG < 64)
770         unsigned long   i_state_flags;          /* Dynamic state flags */
771 #endif
772         unsigned long   i_flags;
773
774 #ifdef CONFIG_EXT4_FS_XATTR
775         /*
776          * Extended attributes can be read independently of the main file
777          * data. Taking i_mutex even when reading would cause contention
778          * between readers of EAs and writers of regular file data, so
779          * instead we synchronize on xattr_sem when reading or changing
780          * EAs.
781          */
782         struct rw_semaphore xattr_sem;
783 #endif
784
785         struct list_head i_orphan;      /* unlinked but open inodes */
786
787         /*
788          * i_disksize keeps track of what the inode size is ON DISK, not
789          * in memory.  During truncate, i_size is set to the new size by
790          * the VFS prior to calling ext4_truncate(), but the filesystem won't
791          * set i_disksize to 0 until the truncate is actually under way.
792          *
793          * The intent is that i_disksize always represents the blocks which
794          * are used by this file.  This allows recovery to restart truncate
795          * on orphans if we crash during truncate.  We actually write i_disksize
796          * into the on-disk inode when writing inodes out, instead of i_size.
797          *
798          * The only time when i_disksize and i_size may be different is when
799          * a truncate is in progress.  The only things which change i_disksize
800          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
801          */
802         loff_t  i_disksize;
803
804         /*
805          * i_data_sem is for serialising ext4_truncate() against
806          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
807          * data tree are chopped off during truncate. We can't do that in
808          * ext4 because whenever we perform intermediate commits during
809          * truncate, the inode and all the metadata blocks *must* be in a
810          * consistent state which allows truncation of the orphans to restart
811          * during recovery.  Hence we must fix the get_block-vs-truncate race
812          * by other means, so we have i_data_sem.
813          */
814         struct rw_semaphore i_data_sem;
815         struct inode vfs_inode;
816         struct jbd2_inode *jinode;
817
818         struct ext4_ext_cache i_cached_extent;
819         /*
820          * File creation time. Its function is same as that of
821          * struct timespec i_{a,c,m}time in the generic inode.
822          */
823         struct timespec i_crtime;
824
825         /* mballoc */
826         struct list_head i_prealloc_list;
827         spinlock_t i_prealloc_lock;
828
829         /* ialloc */
830         ext4_group_t    i_last_alloc_group;
831
832         /* allocation reservation info for delalloc */
833         unsigned int i_reserved_data_blocks;
834         unsigned int i_reserved_meta_blocks;
835         unsigned int i_allocated_meta_blocks;
836         ext4_lblk_t i_da_metadata_calc_last_lblock;
837         int i_da_metadata_calc_len;
838
839         /* on-disk additional length */
840         __u16 i_extra_isize;
841
842 #ifdef CONFIG_QUOTA
843         /* quota space reservation, managed internally by quota code */
844         qsize_t i_reserved_quota;
845 #endif
846
847         /* completed IOs that might need unwritten extents handling */
848         struct list_head i_completed_io_list;
849         spinlock_t i_completed_io_lock;
850         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
851         /* current io_end structure for async DIO write*/
852         ext4_io_end_t *cur_aio_dio;
853         atomic_t i_aiodio_unwritten; /* Nr. of inflight conversions pending */
854
855         spinlock_t i_block_reservation_lock;
856
857         /*
858          * Transactions that contain inode's metadata needed to complete
859          * fsync and fdatasync, respectively.
860          */
861         tid_t i_sync_tid;
862         tid_t i_datasync_tid;
863 };
864
865 /*
866  * File system states
867  */
868 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
869 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
870 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
871
872 /*
873  * Misc. filesystem flags
874  */
875 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
876 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
877 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
878
879 /*
880  * Mount flags
881  */
882 #define EXT4_MOUNT_OLDALLOC             0x00002  /* Don't use the new Orlov allocator */
883 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
884 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
885 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
886 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
887 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
888 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
889 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
890 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
891 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
892 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
893 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
894 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
895 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
896 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
897 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
898 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
899 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
900 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
901 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
902 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
903 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
904 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
905 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
906 #define EXT4_MOUNT_I_VERSION            0x2000000 /* i_version support */
907 #define EXT4_MOUNT_MBLK_IO_SUBMIT       0x4000000 /* multi-block io submits */
908 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
909 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
910 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
911 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
912 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
913
914 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
915                                                 ~EXT4_MOUNT_##opt
916 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
917                                                 EXT4_MOUNT_##opt
918 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
919                                          EXT4_MOUNT_##opt)
920
921 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
922                                                 ~EXT4_MOUNT2_##opt
923 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
924                                                 EXT4_MOUNT2_##opt
925 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
926                                          EXT4_MOUNT2_##opt)
927
928 #define ext4_set_bit                    __test_and_set_bit_le
929 #define ext4_set_bit_atomic             ext2_set_bit_atomic
930 #define ext4_clear_bit                  __test_and_clear_bit_le
931 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
932 #define ext4_test_bit                   test_bit_le
933 #define ext4_find_first_zero_bit        find_first_zero_bit_le
934 #define ext4_find_next_zero_bit         find_next_zero_bit_le
935 #define ext4_find_next_bit              find_next_bit_le
936
937 /*
938  * Maximal mount counts between two filesystem checks
939  */
940 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
941 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
942
943 /*
944  * Behaviour when detecting errors
945  */
946 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
947 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
948 #define EXT4_ERRORS_PANIC               3       /* Panic */
949 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
950
951 /*
952  * Structure of the super block
953  */
954 struct ext4_super_block {
955 /*00*/  __le32  s_inodes_count;         /* Inodes count */
956         __le32  s_blocks_count_lo;      /* Blocks count */
957         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
958         __le32  s_free_blocks_count_lo; /* Free blocks count */
959 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
960         __le32  s_first_data_block;     /* First Data Block */
961         __le32  s_log_block_size;       /* Block size */
962         __le32  s_obso_log_frag_size;   /* Obsoleted fragment size */
963 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
964         __le32  s_obso_frags_per_group; /* Obsoleted fragments per group */
965         __le32  s_inodes_per_group;     /* # Inodes per group */
966         __le32  s_mtime;                /* Mount time */
967 /*30*/  __le32  s_wtime;                /* Write time */
968         __le16  s_mnt_count;            /* Mount count */
969         __le16  s_max_mnt_count;        /* Maximal mount count */
970         __le16  s_magic;                /* Magic signature */
971         __le16  s_state;                /* File system state */
972         __le16  s_errors;               /* Behaviour when detecting errors */
973         __le16  s_minor_rev_level;      /* minor revision level */
974 /*40*/  __le32  s_lastcheck;            /* time of last check */
975         __le32  s_checkinterval;        /* max. time between checks */
976         __le32  s_creator_os;           /* OS */
977         __le32  s_rev_level;            /* Revision level */
978 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
979         __le16  s_def_resgid;           /* Default gid for reserved blocks */
980         /*
981          * These fields are for EXT4_DYNAMIC_REV superblocks only.
982          *
983          * Note: the difference between the compatible feature set and
984          * the incompatible feature set is that if there is a bit set
985          * in the incompatible feature set that the kernel doesn't
986          * know about, it should refuse to mount the filesystem.
987          *
988          * e2fsck's requirements are more strict; if it doesn't know
989          * about a feature in either the compatible or incompatible
990          * feature set, it must abort and not try to meddle with
991          * things it doesn't understand...
992          */
993         __le32  s_first_ino;            /* First non-reserved inode */
994         __le16  s_inode_size;           /* size of inode structure */
995         __le16  s_block_group_nr;       /* block group # of this superblock */
996         __le32  s_feature_compat;       /* compatible feature set */
997 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
998         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
999 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1000 /*78*/  char    s_volume_name[16];      /* volume name */
1001 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1002 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1003         /*
1004          * Performance hints.  Directory preallocation should only
1005          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1006          */
1007         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1008         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1009         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1010         /*
1011          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1012          */
1013 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1014 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1015         __le32  s_journal_dev;          /* device number of journal file */
1016         __le32  s_last_orphan;          /* start of list of inodes to delete */
1017         __le32  s_hash_seed[4];         /* HTREE hash seed */
1018         __u8    s_def_hash_version;     /* Default hash version to use */
1019         __u8    s_jnl_backup_type;
1020         __le16  s_desc_size;            /* size of group descriptor */
1021 /*100*/ __le32  s_default_mount_opts;
1022         __le32  s_first_meta_bg;        /* First metablock block group */
1023         __le32  s_mkfs_time;            /* When the filesystem was created */
1024         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1025         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1026 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1027         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1028         __le32  s_free_blocks_count_hi; /* Free blocks count */
1029         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1030         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1031         __le32  s_flags;                /* Miscellaneous flags */
1032         __le16  s_raid_stride;          /* RAID stride */
1033         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1034         __le64  s_mmp_block;            /* Block for multi-mount protection */
1035         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1036         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1037         __u8    s_reserved_char_pad;
1038         __le16  s_reserved_pad;
1039         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1040         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1041         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1042         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1043                                               snapshot's future use */
1044         __le32  s_snapshot_list;        /* inode number of the head of the
1045                                            on-disk snapshot list */
1046 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1047         __le32  s_error_count;          /* number of fs errors */
1048         __le32  s_first_error_time;     /* first time an error happened */
1049         __le32  s_first_error_ino;      /* inode involved in first error */
1050         __le64  s_first_error_block;    /* block involved of first error */
1051         __u8    s_first_error_func[32]; /* function where the error happened */
1052         __le32  s_first_error_line;     /* line number where error happened */
1053         __le32  s_last_error_time;      /* most recent time of an error */
1054         __le32  s_last_error_ino;       /* inode involved in last error */
1055         __le32  s_last_error_line;      /* line number where error happened */
1056         __le64  s_last_error_block;     /* block involved of last error */
1057         __u8    s_last_error_func[32];  /* function where the error happened */
1058 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1059         __u8    s_mount_opts[64];
1060         __le32  s_reserved[112];        /* Padding to the end of the block */
1061 };
1062
1063 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1064
1065 #ifdef __KERNEL__
1066
1067 /*
1068  * run-time mount flags
1069  */
1070 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
1071 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
1072
1073 /*
1074  * fourth extended-fs super-block data in memory
1075  */
1076 struct ext4_sb_info {
1077         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1078         unsigned long s_inodes_per_block;/* Number of inodes per block */
1079         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1080         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1081         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1082         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1083         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1084         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1085         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1086         unsigned long s_overhead_last;  /* Last calculated overhead */
1087         unsigned long s_blocks_last;    /* Last seen block count */
1088         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1089         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1090         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1091         struct buffer_head **s_group_desc;
1092         unsigned int s_mount_opt;
1093         unsigned int s_mount_opt2;
1094         unsigned int s_mount_flags;
1095         ext4_fsblk_t s_sb_block;
1096         uid_t s_resuid;
1097         gid_t s_resgid;
1098         unsigned short s_mount_state;
1099         unsigned short s_pad;
1100         int s_addr_per_block_bits;
1101         int s_desc_per_block_bits;
1102         int s_inode_size;
1103         int s_first_ino;
1104         unsigned int s_inode_readahead_blks;
1105         unsigned int s_inode_goal;
1106         spinlock_t s_next_gen_lock;
1107         u32 s_next_generation;
1108         u32 s_hash_seed[4];
1109         int s_def_hash_version;
1110         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1111         struct percpu_counter s_freeblocks_counter;
1112         struct percpu_counter s_freeinodes_counter;
1113         struct percpu_counter s_dirs_counter;
1114         struct percpu_counter s_dirtyblocks_counter;
1115         struct blockgroup_lock *s_blockgroup_lock;
1116         struct proc_dir_entry *s_proc;
1117         struct kobject s_kobj;
1118         struct completion s_kobj_unregister;
1119
1120         /* Journaling */
1121         struct journal_s *s_journal;
1122         struct list_head s_orphan;
1123         struct mutex s_orphan_lock;
1124         struct mutex s_resize_lock;
1125         unsigned long s_commit_interval;
1126         u32 s_max_batch_time;
1127         u32 s_min_batch_time;
1128         struct block_device *journal_bdev;
1129 #ifdef CONFIG_JBD2_DEBUG
1130         struct timer_list turn_ro_timer;        /* For turning read-only (crash simulation) */
1131         wait_queue_head_t ro_wait_queue;        /* For people waiting for the fs to go read-only */
1132 #endif
1133 #ifdef CONFIG_QUOTA
1134         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
1135         int s_jquota_fmt;                       /* Format of quota to use */
1136 #endif
1137         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1138         struct rb_root system_blks;
1139
1140 #ifdef EXTENTS_STATS
1141         /* ext4 extents stats */
1142         unsigned long s_ext_min;
1143         unsigned long s_ext_max;
1144         unsigned long s_depth_max;
1145         spinlock_t s_ext_stats_lock;
1146         unsigned long s_ext_blocks;
1147         unsigned long s_ext_extents;
1148 #endif
1149         /* ext4 extent cache stats */
1150         unsigned long extent_cache_hits;
1151         unsigned long extent_cache_misses;
1152
1153         /* for buddy allocator */
1154         struct ext4_group_info ***s_group_info;
1155         struct inode *s_buddy_cache;
1156         spinlock_t s_md_lock;
1157         unsigned short *s_mb_offsets;
1158         unsigned int *s_mb_maxs;
1159
1160         /* tunables */
1161         unsigned long s_stripe;
1162         unsigned int s_mb_stream_request;
1163         unsigned int s_mb_max_to_scan;
1164         unsigned int s_mb_min_to_scan;
1165         unsigned int s_mb_stats;
1166         unsigned int s_mb_order2_reqs;
1167         unsigned int s_mb_group_prealloc;
1168         unsigned int s_max_writeback_mb_bump;
1169         /* where last allocation was done - for stream allocation */
1170         unsigned long s_mb_last_group;
1171         unsigned long s_mb_last_start;
1172
1173         /* stats for buddy allocator */
1174         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1175         atomic_t s_bal_success; /* we found long enough chunks */
1176         atomic_t s_bal_allocated;       /* in blocks */
1177         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1178         atomic_t s_bal_goals;   /* goal hits */
1179         atomic_t s_bal_breaks;  /* too long searches */
1180         atomic_t s_bal_2orders; /* 2^order hits */
1181         spinlock_t s_bal_lock;
1182         unsigned long s_mb_buddies_generated;
1183         unsigned long long s_mb_generation_time;
1184         atomic_t s_mb_lost_chunks;
1185         atomic_t s_mb_preallocated;
1186         atomic_t s_mb_discarded;
1187         atomic_t s_lock_busy;
1188
1189         /* locality groups */
1190         struct ext4_locality_group __percpu *s_locality_groups;
1191
1192         /* for write statistics */
1193         unsigned long s_sectors_written_start;
1194         u64 s_kbytes_written;
1195
1196         unsigned int s_log_groups_per_flex;
1197         struct flex_groups *s_flex_groups;
1198
1199         /* workqueue for dio unwritten */
1200         struct workqueue_struct *dio_unwritten_wq;
1201
1202         /* timer for periodic error stats printing */
1203         struct timer_list s_err_report;
1204
1205         /* Lazy inode table initialization info */
1206         struct ext4_li_request *s_li_request;
1207         /* Wait multiplier for lazy initialization thread */
1208         unsigned int s_li_wait_mult;
1209
1210         /* Kernel thread for multiple mount protection */
1211         struct task_struct *s_mmp_tsk;
1212 };
1213
1214 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1215 {
1216         return sb->s_fs_info;
1217 }
1218 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1219 {
1220         return container_of(inode, struct ext4_inode_info, vfs_inode);
1221 }
1222
1223 static inline struct timespec ext4_current_time(struct inode *inode)
1224 {
1225         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1226                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1227 }
1228
1229 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1230 {
1231         return ino == EXT4_ROOT_INO ||
1232                 ino == EXT4_JOURNAL_INO ||
1233                 ino == EXT4_RESIZE_INO ||
1234                 (ino >= EXT4_FIRST_INO(sb) &&
1235                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1236 }
1237
1238 /*
1239  * Inode dynamic state flags
1240  */
1241 enum {
1242         EXT4_STATE_JDATA,               /* journaled data exists */
1243         EXT4_STATE_NEW,                 /* inode is newly created */
1244         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1245         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1246         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1247         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1248         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1249         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1250         EXT4_STATE_DELALLOC_RESERVED,   /* blks already reserved for delalloc */
1251 };
1252
1253 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1254 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1255 {                                                                       \
1256         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1257 }                                                                       \
1258 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1259 {                                                                       \
1260         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1261 }                                                                       \
1262 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1263 {                                                                       \
1264         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1265 }
1266
1267 EXT4_INODE_BIT_FNS(flag, flags, 0)
1268 #if (BITS_PER_LONG < 64)
1269 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1270
1271 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1272 {
1273         (ei)->i_state_flags = 0;
1274 }
1275 #else
1276 EXT4_INODE_BIT_FNS(state, flags, 32)
1277
1278 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1279 {
1280         /* We depend on the fact that callers will set i_flags */
1281 }
1282 #endif
1283 #else
1284 /* Assume that user mode programs are passing in an ext4fs superblock, not
1285  * a kernel struct super_block.  This will allow us to call the feature-test
1286  * macros from user land. */
1287 #define EXT4_SB(sb)     (sb)
1288 #endif
1289
1290 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1291
1292 /*
1293  * Codes for operating systems
1294  */
1295 #define EXT4_OS_LINUX           0
1296 #define EXT4_OS_HURD            1
1297 #define EXT4_OS_MASIX           2
1298 #define EXT4_OS_FREEBSD         3
1299 #define EXT4_OS_LITES           4
1300
1301 /*
1302  * Revision levels
1303  */
1304 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1305 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1306
1307 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1308 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1309
1310 #define EXT4_GOOD_OLD_INODE_SIZE 128
1311
1312 /*
1313  * Feature set definitions
1314  */
1315
1316 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1317         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1318 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1319         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1320 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1321         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1322 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1323         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1324 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1325         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1326 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1327         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1328 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1329         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1330 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1331         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1332 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1333         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1334
1335 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1336 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1337 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1338 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1339 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1340 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1341
1342 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1343 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1344 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1345 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1346 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1347 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1348 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1349 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1350
1351 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1352 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1353 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1354 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1355 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1356 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1357 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1358 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1359 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1360 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1361 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1362
1363 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1364 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1365                                          EXT4_FEATURE_INCOMPAT_META_BG)
1366 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1367                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1368                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1369
1370 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1371 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1372                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1373                                          EXT4_FEATURE_INCOMPAT_META_BG)
1374 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1375                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1376                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1377
1378 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1379 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1380                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1381                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1382                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1383                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1384                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1385                                          EXT4_FEATURE_INCOMPAT_MMP)
1386 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1387                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1388                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1389                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1390                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1391                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1392                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE)
1393
1394 /*
1395  * Default values for user and/or group using reserved blocks
1396  */
1397 #define EXT4_DEF_RESUID         0
1398 #define EXT4_DEF_RESGID         0
1399
1400 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1401
1402 /*
1403  * Default mount options
1404  */
1405 #define EXT4_DEFM_DEBUG         0x0001
1406 #define EXT4_DEFM_BSDGROUPS     0x0002
1407 #define EXT4_DEFM_XATTR_USER    0x0004
1408 #define EXT4_DEFM_ACL           0x0008
1409 #define EXT4_DEFM_UID16         0x0010
1410 #define EXT4_DEFM_JMODE         0x0060
1411 #define EXT4_DEFM_JMODE_DATA    0x0020
1412 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1413 #define EXT4_DEFM_JMODE_WBACK   0x0060
1414 #define EXT4_DEFM_NOBARRIER     0x0100
1415 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1416 #define EXT4_DEFM_DISCARD       0x0400
1417 #define EXT4_DEFM_NODELALLOC    0x0800
1418
1419 /*
1420  * Default journal batch times
1421  */
1422 #define EXT4_DEF_MIN_BATCH_TIME 0
1423 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1424
1425 /*
1426  * Minimum number of groups in a flexgroup before we separate out
1427  * directories into the first block group of a flexgroup
1428  */
1429 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1430
1431 /*
1432  * Structure of a directory entry
1433  */
1434 #define EXT4_NAME_LEN 255
1435
1436 struct ext4_dir_entry {
1437         __le32  inode;                  /* Inode number */
1438         __le16  rec_len;                /* Directory entry length */
1439         __le16  name_len;               /* Name length */
1440         char    name[EXT4_NAME_LEN];    /* File name */
1441 };
1442
1443 /*
1444  * The new version of the directory entry.  Since EXT4 structures are
1445  * stored in intel byte order, and the name_len field could never be
1446  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1447  * file_type field.
1448  */
1449 struct ext4_dir_entry_2 {
1450         __le32  inode;                  /* Inode number */
1451         __le16  rec_len;                /* Directory entry length */
1452         __u8    name_len;               /* Name length */
1453         __u8    file_type;
1454         char    name[EXT4_NAME_LEN];    /* File name */
1455 };
1456
1457 /*
1458  * Ext4 directory file types.  Only the low 3 bits are used.  The
1459  * other bits are reserved for now.
1460  */
1461 #define EXT4_FT_UNKNOWN         0
1462 #define EXT4_FT_REG_FILE        1
1463 #define EXT4_FT_DIR             2
1464 #define EXT4_FT_CHRDEV          3
1465 #define EXT4_FT_BLKDEV          4
1466 #define EXT4_FT_FIFO            5
1467 #define EXT4_FT_SOCK            6
1468 #define EXT4_FT_SYMLINK         7
1469
1470 #define EXT4_FT_MAX             8
1471
1472 /*
1473  * EXT4_DIR_PAD defines the directory entries boundaries
1474  *
1475  * NOTE: It must be a multiple of 4
1476  */
1477 #define EXT4_DIR_PAD                    4
1478 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1479 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1480                                          ~EXT4_DIR_ROUND)
1481 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1482
1483 /*
1484  * If we ever get support for fs block sizes > page_size, we'll need
1485  * to remove the #if statements in the next two functions...
1486  */
1487 static inline unsigned int
1488 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1489 {
1490         unsigned len = le16_to_cpu(dlen);
1491
1492 #if (PAGE_CACHE_SIZE >= 65536)
1493         if (len == EXT4_MAX_REC_LEN || len == 0)
1494                 return blocksize;
1495         return (len & 65532) | ((len & 3) << 16);
1496 #else
1497         return len;
1498 #endif
1499 }
1500
1501 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1502 {
1503         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1504                 BUG();
1505 #if (PAGE_CACHE_SIZE >= 65536)
1506         if (len < 65536)
1507                 return cpu_to_le16(len);
1508         if (len == blocksize) {
1509                 if (blocksize == 65536)
1510                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1511                 else
1512                         return cpu_to_le16(0);
1513         }
1514         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1515 #else
1516         return cpu_to_le16(len);
1517 #endif
1518 }
1519
1520 /*
1521  * Hash Tree Directory indexing
1522  * (c) Daniel Phillips, 2001
1523  */
1524
1525 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1526                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1527                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1528 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1529 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1530
1531 /* Legal values for the dx_root hash_version field: */
1532
1533 #define DX_HASH_LEGACY          0
1534 #define DX_HASH_HALF_MD4        1
1535 #define DX_HASH_TEA             2
1536 #define DX_HASH_LEGACY_UNSIGNED 3
1537 #define DX_HASH_HALF_MD4_UNSIGNED       4
1538 #define DX_HASH_TEA_UNSIGNED            5
1539
1540 #ifdef __KERNEL__
1541
1542 /* hash info structure used by the directory hash */
1543 struct dx_hash_info
1544 {
1545         u32             hash;
1546         u32             minor_hash;
1547         int             hash_version;
1548         u32             *seed;
1549 };
1550
1551 #define EXT4_HTREE_EOF  0x7fffffff
1552
1553 /*
1554  * Control parameters used by ext4_htree_next_block
1555  */
1556 #define HASH_NB_ALWAYS          1
1557
1558
1559 /*
1560  * Describe an inode's exact location on disk and in memory
1561  */
1562 struct ext4_iloc
1563 {
1564         struct buffer_head *bh;
1565         unsigned long offset;
1566         ext4_group_t block_group;
1567 };
1568
1569 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1570 {
1571         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1572 }
1573
1574 /*
1575  * This structure is stuffed into the struct file's private_data field
1576  * for directories.  It is where we put information so that we can do
1577  * readdir operations in hash tree order.
1578  */
1579 struct dir_private_info {
1580         struct rb_root  root;
1581         struct rb_node  *curr_node;
1582         struct fname    *extra_fname;
1583         loff_t          last_pos;
1584         __u32           curr_hash;
1585         __u32           curr_minor_hash;
1586         __u32           next_hash;
1587 };
1588
1589 /* calculate the first block number of the group */
1590 static inline ext4_fsblk_t
1591 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1592 {
1593         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1594                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1595 }
1596
1597 /*
1598  * Special error return code only used by dx_probe() and its callers.
1599  */
1600 #define ERR_BAD_DX_DIR  -75000
1601
1602 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1603                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1604
1605 /*
1606  * Timeout and state flag for lazy initialization inode thread.
1607  */
1608 #define EXT4_DEF_LI_WAIT_MULT                   10
1609 #define EXT4_DEF_LI_MAX_START_DELAY             5
1610 #define EXT4_LAZYINIT_QUIT                      0x0001
1611 #define EXT4_LAZYINIT_RUNNING                   0x0002
1612
1613 /*
1614  * Lazy inode table initialization info
1615  */
1616 struct ext4_lazy_init {
1617         unsigned long           li_state;
1618         struct list_head        li_request_list;
1619         struct mutex            li_list_mtx;
1620 };
1621
1622 struct ext4_li_request {
1623         struct super_block      *lr_super;
1624         struct ext4_sb_info     *lr_sbi;
1625         ext4_group_t            lr_next_group;
1626         struct list_head        lr_request;
1627         unsigned long           lr_next_sched;
1628         unsigned long           lr_timeout;
1629 };
1630
1631 struct ext4_features {
1632         struct kobject f_kobj;
1633         struct completion f_kobj_unregister;
1634 };
1635
1636 /*
1637  * This structure will be used for multiple mount protection. It will be
1638  * written into the block number saved in the s_mmp_block field in the
1639  * superblock. Programs that check MMP should assume that if
1640  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1641  * to use the filesystem, regardless of how old the timestamp is.
1642  */
1643 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
1644 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1645 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
1646 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
1647
1648 struct mmp_struct {
1649         __le32  mmp_magic;              /* Magic number for MMP */
1650         __le32  mmp_seq;                /* Sequence no. updated periodically */
1651
1652         /*
1653          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1654          * purposes and do not affect the correctness of the algorithm
1655          */
1656         __le64  mmp_time;               /* Time last updated */
1657         char    mmp_nodename[64];       /* Node which last updated MMP block */
1658         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
1659
1660         /*
1661          * mmp_check_interval is used to verify if the MMP block has been
1662          * updated on the block device. The value is updated based on the
1663          * maximum time to write the MMP block during an update cycle.
1664          */
1665         __le16  mmp_check_interval;
1666
1667         __le16  mmp_pad1;
1668         __le32  mmp_pad2[227];
1669 };
1670
1671 /* arguments passed to the mmp thread */
1672 struct mmpd_data {
1673         struct buffer_head *bh; /* bh from initial read_mmp_block() */
1674         struct super_block *sb;  /* super block of the fs */
1675 };
1676
1677 /*
1678  * Check interval multiplier
1679  * The MMP block is written every update interval and initially checked every
1680  * update interval x the multiplier (the value is then adapted based on the
1681  * write latency). The reason is that writes can be delayed under load and we
1682  * don't want readers to incorrectly assume that the filesystem is no longer
1683  * in use.
1684  */
1685 #define EXT4_MMP_CHECK_MULT             2UL
1686
1687 /*
1688  * Minimum interval for MMP checking in seconds.
1689  */
1690 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
1691
1692 /*
1693  * Maximum interval for MMP checking in seconds.
1694  */
1695 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
1696
1697 /*
1698  * Function prototypes
1699  */
1700
1701 /*
1702  * Ok, these declarations are also in <linux/kernel.h> but none of the
1703  * ext4 source programs needs to include it so they are duplicated here.
1704  */
1705 # define NORET_TYPE     /**/
1706 # define ATTRIB_NORET   __attribute__((noreturn))
1707 # define NORET_AND      noreturn,
1708
1709 /* bitmap.c */
1710 extern unsigned int ext4_count_free(struct buffer_head *, unsigned);
1711
1712 /* balloc.c */
1713 extern unsigned int ext4_block_group(struct super_block *sb,
1714                         ext4_fsblk_t blocknr);
1715 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1716                         ext4_fsblk_t blocknr);
1717 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1718 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1719                         ext4_group_t group);
1720 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1721                         ext4_fsblk_t goal, unsigned long *count, int *errp);
1722 extern int ext4_claim_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1723 extern ext4_fsblk_t ext4_count_free_blocks(struct super_block *);
1724 extern void ext4_check_blocks_bitmap(struct super_block *);
1725 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1726                                                     ext4_group_t block_group,
1727                                                     struct buffer_head ** bh);
1728 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1729 struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1730                                       ext4_group_t block_group);
1731 extern unsigned ext4_init_block_bitmap(struct super_block *sb,
1732                                        struct buffer_head *bh,
1733                                        ext4_group_t group,
1734                                        struct ext4_group_desc *desc);
1735 #define ext4_free_blocks_after_init(sb, group, desc)                    \
1736                 ext4_init_block_bitmap(sb, NULL, group, desc)
1737
1738 /* dir.c */
1739 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1740                                   struct file *,
1741                                   struct ext4_dir_entry_2 *,
1742                                   struct buffer_head *, unsigned int);
1743 #define ext4_check_dir_entry(dir, filp, de, bh, offset)                 \
1744         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
1745                                         (de), (bh), (offset)))
1746 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1747                                     __u32 minor_hash,
1748                                     struct ext4_dir_entry_2 *dirent);
1749 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1750
1751 /* fsync.c */
1752 extern int ext4_sync_file(struct file *, int);
1753 extern int ext4_flush_completed_IO(struct inode *);
1754
1755 /* hash.c */
1756 extern int ext4fs_dirhash(const char *name, int len, struct
1757                           dx_hash_info *hinfo);
1758
1759 /* ialloc.c */
1760 extern struct inode *ext4_new_inode(handle_t *, struct inode *, int,
1761                                     const struct qstr *qstr, __u32 goal);
1762 extern void ext4_free_inode(handle_t *, struct inode *);
1763 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1764 extern unsigned long ext4_count_free_inodes(struct super_block *);
1765 extern unsigned long ext4_count_dirs(struct super_block *);
1766 extern void ext4_check_inodes_bitmap(struct super_block *);
1767 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1768 extern int ext4_init_inode_table(struct super_block *sb,
1769                                  ext4_group_t group, int barrier);
1770
1771 /* mballoc.c */
1772 extern long ext4_mb_stats;
1773 extern long ext4_mb_max_to_scan;
1774 extern int ext4_mb_init(struct super_block *, int);
1775 extern int ext4_mb_release(struct super_block *);
1776 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1777                                 struct ext4_allocation_request *, int *);
1778 extern int ext4_mb_reserve_blocks(struct super_block *, int);
1779 extern void ext4_discard_preallocations(struct inode *);
1780 extern int __init ext4_init_mballoc(void);
1781 extern void ext4_exit_mballoc(void);
1782 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1783                              struct buffer_head *bh, ext4_fsblk_t block,
1784                              unsigned long count, int flags);
1785 extern int ext4_mb_add_groupinfo(struct super_block *sb,
1786                 ext4_group_t i, struct ext4_group_desc *desc);
1787 extern void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
1788                                 ext4_fsblk_t block, unsigned long count);
1789 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
1790
1791 /* inode.c */
1792 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1793                                                 ext4_lblk_t, int, int *);
1794 struct buffer_head *ext4_bread(handle_t *, struct inode *,
1795                                                 ext4_lblk_t, int, int *);
1796 int ext4_get_block(struct inode *inode, sector_t iblock,
1797                                 struct buffer_head *bh_result, int create);
1798
1799 extern struct inode *ext4_iget(struct super_block *, unsigned long);
1800 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
1801 extern int  ext4_setattr(struct dentry *, struct iattr *);
1802 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1803                                 struct kstat *stat);
1804 extern void ext4_evict_inode(struct inode *);
1805 extern void ext4_clear_inode(struct inode *);
1806 extern int  ext4_sync_inode(handle_t *, struct inode *);
1807 extern void ext4_dirty_inode(struct inode *);
1808 extern int ext4_change_inode_journal_flag(struct inode *, int);
1809 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1810 extern int ext4_can_truncate(struct inode *inode);
1811 extern void ext4_truncate(struct inode *);
1812 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
1813 extern void ext4_set_inode_flags(struct inode *);
1814 extern void ext4_get_inode_flags(struct ext4_inode_info *);
1815 extern int ext4_alloc_da_blocks(struct inode *inode);
1816 extern void ext4_set_aops(struct inode *inode);
1817 extern int ext4_writepage_trans_blocks(struct inode *);
1818 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1819 extern int ext4_block_truncate_page(handle_t *handle,
1820                 struct address_space *mapping, loff_t from);
1821 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1822 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
1823 extern void ext4_da_update_reserve_space(struct inode *inode,
1824                                         int used, int quota_claim);
1825 /* ioctl.c */
1826 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1827 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1828
1829 /* migrate.c */
1830 extern int ext4_ext_migrate(struct inode *);
1831
1832 /* namei.c */
1833 extern int ext4_orphan_add(handle_t *, struct inode *);
1834 extern int ext4_orphan_del(handle_t *, struct inode *);
1835 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1836                                 __u32 start_minor_hash, __u32 *next_hash);
1837
1838 /* resize.c */
1839 extern int ext4_group_add(struct super_block *sb,
1840                                 struct ext4_new_group_data *input);
1841 extern int ext4_group_extend(struct super_block *sb,
1842                                 struct ext4_super_block *es,
1843                                 ext4_fsblk_t n_blocks_count);
1844
1845 /* super.c */
1846 extern void __ext4_error(struct super_block *, const char *, unsigned int,
1847                          const char *, ...)
1848         __attribute__ ((format (printf, 4, 5)));
1849 #define ext4_error(sb, message...)      __ext4_error(sb, __func__,      \
1850                                                      __LINE__, ## message)
1851 extern void ext4_error_inode(struct inode *, const char *, unsigned int,
1852                              ext4_fsblk_t, const char *, ...)
1853         __attribute__ ((format (printf, 5, 6)));
1854 extern void ext4_error_file(struct file *, const char *, unsigned int,
1855                             ext4_fsblk_t, const char *, ...)
1856         __attribute__ ((format (printf, 5, 6)));
1857 extern void __ext4_std_error(struct super_block *, const char *,
1858                              unsigned int, int);
1859 extern void __ext4_abort(struct super_block *, const char *, unsigned int,
1860                        const char *, ...)
1861         __attribute__ ((format (printf, 4, 5)));
1862 #define ext4_abort(sb, message...)      __ext4_abort(sb, __func__, \
1863                                                        __LINE__, ## message)
1864 extern void __ext4_warning(struct super_block *, const char *, unsigned int,
1865                           const char *, ...)
1866         __attribute__ ((format (printf, 4, 5)));
1867 #define ext4_warning(sb, message...)    __ext4_warning(sb, __func__, \
1868                                                        __LINE__, ## message)
1869 extern void ext4_msg(struct super_block *, const char *, const char *, ...)
1870         __attribute__ ((format (printf, 3, 4)));
1871 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
1872                            const char *, unsigned int, const char *);
1873 #define dump_mmp_msg(sb, mmp, msg)      __dump_mmp_msg(sb, mmp, __func__, \
1874                                                        __LINE__, msg)
1875 extern void __ext4_grp_locked_error(const char *, unsigned int, \
1876                                     struct super_block *, ext4_group_t, \
1877                                     unsigned long, ext4_fsblk_t, \
1878                                     const char *, ...)
1879         __attribute__ ((format (printf, 7, 8)));
1880 #define ext4_grp_locked_error(sb, grp, message...) \
1881         __ext4_grp_locked_error(__func__, __LINE__, (sb), (grp), ## message)
1882 extern void ext4_update_dynamic_rev(struct super_block *sb);
1883 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1884                                         __u32 compat);
1885 extern int ext4_update_rocompat_feature(handle_t *handle,
1886                                         struct super_block *sb, __u32 rocompat);
1887 extern int ext4_update_incompat_feature(handle_t *handle,
1888                                         struct super_block *sb, __u32 incompat);
1889 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1890                                       struct ext4_group_desc *bg);
1891 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1892                                       struct ext4_group_desc *bg);
1893 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1894                                      struct ext4_group_desc *bg);
1895 extern __u32 ext4_free_blks_count(struct super_block *sb,
1896                                 struct ext4_group_desc *bg);
1897 extern __u32 ext4_free_inodes_count(struct super_block *sb,
1898                                  struct ext4_group_desc *bg);
1899 extern __u32 ext4_used_dirs_count(struct super_block *sb,
1900                                 struct ext4_group_desc *bg);
1901 extern __u32 ext4_itable_unused_count(struct super_block *sb,
1902                                    struct ext4_group_desc *bg);
1903 extern void ext4_block_bitmap_set(struct super_block *sb,
1904                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1905 extern void ext4_inode_bitmap_set(struct super_block *sb,
1906                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
1907 extern void ext4_inode_table_set(struct super_block *sb,
1908                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
1909 extern void ext4_free_blks_set(struct super_block *sb,
1910                                struct ext4_group_desc *bg, __u32 count);
1911 extern void ext4_free_inodes_set(struct super_block *sb,
1912                                 struct ext4_group_desc *bg, __u32 count);
1913 extern void ext4_used_dirs_set(struct super_block *sb,
1914                                 struct ext4_group_desc *bg, __u32 count);
1915 extern void ext4_itable_unused_set(struct super_block *sb,
1916                                    struct ext4_group_desc *bg, __u32 count);
1917 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
1918                                    struct ext4_group_desc *gdp);
1919 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
1920                                        struct ext4_group_desc *gdp);
1921
1922 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
1923 {
1924         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
1925                 le32_to_cpu(es->s_blocks_count_lo);
1926 }
1927
1928 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
1929 {
1930         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
1931                 le32_to_cpu(es->s_r_blocks_count_lo);
1932 }
1933
1934 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
1935 {
1936         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
1937                 le32_to_cpu(es->s_free_blocks_count_lo);
1938 }
1939
1940 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
1941                                          ext4_fsblk_t blk)
1942 {
1943         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
1944         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
1945 }
1946
1947 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
1948                                               ext4_fsblk_t blk)
1949 {
1950         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
1951         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
1952 }
1953
1954 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
1955                                            ext4_fsblk_t blk)
1956 {
1957         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
1958         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
1959 }
1960
1961 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
1962 {
1963         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
1964                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
1965                         le32_to_cpu(raw_inode->i_size_lo);
1966         else
1967                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
1968 }
1969
1970 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
1971 {
1972         raw_inode->i_size_lo = cpu_to_le32(i_size);
1973         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
1974 }
1975
1976 static inline
1977 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
1978                                             ext4_group_t group)
1979 {
1980          struct ext4_group_info ***grp_info;
1981          long indexv, indexh;
1982          grp_info = EXT4_SB(sb)->s_group_info;
1983          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
1984          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
1985          return grp_info[indexv][indexh];
1986 }
1987
1988 /*
1989  * Reading s_groups_count requires using smp_rmb() afterwards.  See
1990  * the locking protocol documented in the comments of ext4_group_add()
1991  * in resize.c
1992  */
1993 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
1994 {
1995         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
1996
1997         smp_rmb();
1998         return ngroups;
1999 }
2000
2001 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2002                                              ext4_group_t block_group)
2003 {
2004         return block_group >> sbi->s_log_groups_per_flex;
2005 }
2006
2007 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2008 {
2009         return 1 << sbi->s_log_groups_per_flex;
2010 }
2011
2012 #define ext4_std_error(sb, errno)                               \
2013 do {                                                            \
2014         if ((errno))                                            \
2015                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2016 } while (0)
2017
2018 #ifdef CONFIG_SMP
2019 /* Each CPU can accumulate percpu_counter_batch blocks in their local
2020  * counters. So we need to make sure we have free blocks more
2021  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2022  */
2023 #define EXT4_FREEBLOCKS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2024 #else
2025 #define EXT4_FREEBLOCKS_WATERMARK 0
2026 #endif
2027
2028 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2029 {
2030         /*
2031          * XXX: replace with spinlock if seen contended -bzzz
2032          */
2033         down_write(&EXT4_I(inode)->i_data_sem);
2034         if (newsize > EXT4_I(inode)->i_disksize)
2035                 EXT4_I(inode)->i_disksize = newsize;
2036         up_write(&EXT4_I(inode)->i_data_sem);
2037         return ;
2038 }
2039
2040 struct ext4_group_info {
2041         unsigned long   bb_state;
2042         struct rb_root  bb_free_root;
2043         ext4_grpblk_t   bb_first_free;  /* first free block */
2044         ext4_grpblk_t   bb_free;        /* total free blocks */
2045         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2046         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2047         struct          list_head bb_prealloc_list;
2048 #ifdef DOUBLE_CHECK
2049         void            *bb_bitmap;
2050 #endif
2051         struct rw_semaphore alloc_sem;
2052         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2053                                          * regions, index is order.
2054                                          * bb_counters[3] = 5 means
2055                                          * 5 free 8-block regions. */
2056 };
2057
2058 #define EXT4_GROUP_INFO_NEED_INIT_BIT   0
2059
2060 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2061         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2062
2063 #define EXT4_MAX_CONTENTION             8
2064 #define EXT4_CONTENTION_THRESHOLD       2
2065
2066 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2067                                               ext4_group_t group)
2068 {
2069         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2070 }
2071
2072 /*
2073  * Returns true if the filesystem is busy enough that attempts to
2074  * access the block group locks has run into contention.
2075  */
2076 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2077 {
2078         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2079 }
2080
2081 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2082 {
2083         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2084         if (spin_trylock(lock))
2085                 /*
2086                  * We're able to grab the lock right away, so drop the
2087                  * lock contention counter.
2088                  */
2089                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2090         else {
2091                 /*
2092                  * The lock is busy, so bump the contention counter,
2093                  * and then wait on the spin lock.
2094                  */
2095                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2096                                   EXT4_MAX_CONTENTION);
2097                 spin_lock(lock);
2098         }
2099 }
2100
2101 static inline void ext4_unlock_group(struct super_block *sb,
2102                                         ext4_group_t group)
2103 {
2104         spin_unlock(ext4_group_lock_ptr(sb, group));
2105 }
2106
2107 static inline void ext4_mark_super_dirty(struct super_block *sb)
2108 {
2109         if (EXT4_SB(sb)->s_journal == NULL)
2110                 sb->s_dirt =1;
2111 }
2112
2113 /*
2114  * Inodes and files operations
2115  */
2116
2117 /* dir.c */
2118 extern const struct file_operations ext4_dir_operations;
2119
2120 /* file.c */
2121 extern const struct inode_operations ext4_file_inode_operations;
2122 extern const struct file_operations ext4_file_operations;
2123 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2124
2125 /* namei.c */
2126 extern const struct inode_operations ext4_dir_inode_operations;
2127 extern const struct inode_operations ext4_special_inode_operations;
2128 extern struct dentry *ext4_get_parent(struct dentry *child);
2129
2130 /* symlink.c */
2131 extern const struct inode_operations ext4_symlink_inode_operations;
2132 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2133
2134 /* block_validity */
2135 extern void ext4_release_system_zone(struct super_block *sb);
2136 extern int ext4_setup_system_zone(struct super_block *sb);
2137 extern int __init ext4_init_system_zone(void);
2138 extern void ext4_exit_system_zone(void);
2139 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2140                                  ext4_fsblk_t start_blk,
2141                                  unsigned int count);
2142
2143 /* extents.c */
2144 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2145 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2146 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
2147                                        int chunk);
2148 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2149                                struct ext4_map_blocks *map, int flags);
2150 extern void ext4_ext_truncate(struct inode *);
2151 extern void ext4_ext_init(struct super_block *);
2152 extern void ext4_ext_release(struct super_block *);
2153 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2154                           loff_t len);
2155 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
2156                           ssize_t len);
2157 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2158                            struct ext4_map_blocks *map, int flags);
2159 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2160                         __u64 start, __u64 len);
2161 /* move_extent.c */
2162 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2163                              __u64 start_orig, __u64 start_donor,
2164                              __u64 len, __u64 *moved_len);
2165
2166 /* page-io.c */
2167 extern int __init ext4_init_pageio(void);
2168 extern void ext4_exit_pageio(void);
2169 extern void ext4_ioend_wait(struct inode *);
2170 extern void ext4_free_io_end(ext4_io_end_t *io);
2171 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2172 extern int ext4_end_io_nolock(ext4_io_end_t *io);
2173 extern void ext4_io_submit(struct ext4_io_submit *io);
2174 extern int ext4_bio_write_page(struct ext4_io_submit *io,
2175                                struct page *page,
2176                                int len,
2177                                struct writeback_control *wbc);
2178
2179 /* mmp.c */
2180 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2181
2182 /* BH_Uninit flag: blocks are allocated but uninitialized on disk */
2183 enum ext4_state_bits {
2184         BH_Uninit       /* blocks are allocated but uninitialized on disk */
2185           = BH_JBDPrivateStart,
2186 };
2187
2188 BUFFER_FNS(Uninit, uninit)
2189 TAS_BUFFER_FNS(Uninit, uninit)
2190
2191 /*
2192  * Add new method to test wether block and inode bitmaps are properly
2193  * initialized. With uninit_bg reading the block from disk is not enough
2194  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2195  */
2196 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2197
2198 static inline int bitmap_uptodate(struct buffer_head *bh)
2199 {
2200         return (buffer_uptodate(bh) &&
2201                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2202 }
2203 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2204 {
2205         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2206 }
2207
2208 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2209
2210 /* For ioend & aio unwritten conversion wait queues */
2211 #define EXT4_WQ_HASH_SZ         37
2212 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
2213                                             EXT4_WQ_HASH_SZ])
2214 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
2215                                              EXT4_WQ_HASH_SZ])
2216 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2217 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2218
2219 #endif  /* __KERNEL__ */
2220
2221 #endif  /* _EXT4_H */