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